Fundamentals of Structural Geology (Pollard D.D., Fletcher R.C.)

Fundamentals of Structural Geology
Fundamentals of Structural Geology provides a new working in related disciplines, including geophysics,
framework for the investigation of geological struc- rock mechanics, eld mapping, hydrogeology, petro-
tures by integrating eld mapping and mechanical leum and geotechnical engineering, and natural
analysis. It emphasizes the observational data, hazard mitigation. The book is supported by a
modern mapping technology, principles of contin- website (www.cambridge.org/ 0521839270) hosting
uum mechanics, and the mathematical and compu- images from the book, additional colour images,
tational skills, necessary to map, describe, model, student exercises and MATLAB scripts. Solutions to
and explain deformation in the Earths lithosphere the exercises are available to instructors.
quantitatively.
Assuming a basic knowledge of physical geology, dav i d p o l l a r d is the Morris Professor of Earth
introductory calculus, and physics, this advanced Sciences in the Department of Geological and
textbook builds on more traditional courses that Environmental Sciences at Stanford University
emphasize descriptive terminology, geometric tech- where he co-directs the program in Structural
niques, and kinematics. In a signicant departure Geology and Geomechanics. He and his students are
from conventional textbooks on the subject, differ- using quantitative eld data and principles of struc-
ential geometry is introduced and applied to quan- tural geology, combined with laboratory and com-
tify descriptions of geological structures. Differential puter modeling, to address questions about processes
geometry integrates the spatial information conven- of faulting, fracturing, and rock deformation. The
tionally found on maps with orientation data from research aims to understand how faults and fractures
stereograms to provide reproducible descriptions of evolve in the Earths crust; how they affect the ow
geological structures. By starting from the funda- of magma, groundwater, and hydrocarbons; and
mental conservation laws of mass and momentum, what role fractures play in earthquake generation
the constitutive laws of material behavior, and the and volcanic eruption
kinematic relationships for strain and rate of defor-
mation, the authors demonstrate the relevance of r a y m o n d f l e t c h e r is a Research Professor in
solid and uid mechanics to structural geology. the Department of Geosciences at the Pennsylvania
The constitutive relations used in the book are State University. He and his collaborators study the
sufciently elementary to enable students to gain continuous deformation of rock as in the emplace-
physical insight from analytical solutions, but are ment of mantled gneiss domes, rock folding, and
adequately realistic to provide compelling correla- basin and range necking. He also works on processes
tions to observational data. linking chemical aspects of mineral growth or disso-
This book offers a modern quantitative approach lution in rocks and deformation. Currently he is
to structural geology for advanced undergraduate studying folding near the base of ice sheets, and the
and graduate students and researchers in structural evolution of structures and rheological behavior of
geology and tectonics. It will also interest those composite rock masses.
Fundamentals of
Structural Geology
David D. Pollard
Stanford University
and
Raymond C. Fletcher
The Pennsylvania State University
University Printing House, Cambridge cb2 8bs, United Kingdom
Published in the United States of America by Cambridge University Press, New York
Cambridge University Press is part of the University of Cambridge.

It furthers the Universitys mission by disseminating knowledge in the pursuit of
education, learning and research at the highest international levels of excellence.
www.cambridge.org
Information on this title: www.cambridge.org/9780521839273
D. D. Pollard and R. C. Fletcher 2005
This publication is in copyright. Subject to statutory exception
and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.
First published 2005
4th printing 2010
A catalogue record for this publication is available from the British Library
isbn 978-0-521-83927-3 Hardback
Cambridge University Press has no responsibility for the persistence or accuracy of
URLs for external or third-party internet websites referred to in this publication,
and does not guarantee that any content on such websites is, or will remain, accurate
orappropriate.
Contents
Preface page ix
Acknowledgments xi
Chapter 1 Motivations and opportunities 1

1.1 Earthquake hazards in southern California 2
1.2 Radar lineaments on Venus 9
1.3 Faulting in a North Sea hydrocarbon reservoir 11
1.4 Anticracks in southern France 16
1.5 Mountain building on the Colorado Plateau 20
1.6 Concluding remarks 24
Chapter 2 Structural mapping techniques and tools 25

2.1 Geographic coordinates and map projections 27
2.2 Local coordinates and position vectors 34
2.3 Orientations of structural elements 52
2.4 Structural mapping using GPS technology 69
Chapter 3 Characterizing structures using differential

geometry 75
3.1 The concept and description of lineations 77
3.2 The concept and description of curved surfaces 91
3.3 Applications of differential geometry to structural geology 114
Chapter 4 Physical quantities, fields, dimensions, and

scaling 120
4.1 Physical quantities and the continuum 121
4.2 Physical dimensions and dimensional analysis 127
4.3 Dimensionless groups and the scaling of structural processes 132
4.4 Scaled laboratory models 143
Chapter 5 Deformation and flow 152

5.1 Rock deformation: some observations and a simple
description 154
5.2 Evolving geometry of a structure: kinematic models, velocity
models, and deformation 158
5.3 Relation between deformation and velocity elds 168
5.4 Velocity elds: the instantaneous state of motion 177
5.5 General results 183
vi CONTENTS
Chapter 6 Force, traction, and stress 194

6.1 Concepts of force and traction 196
6.2 Concept and analysis of stress 207
6.3 State of stress in the Earth 227
Chapter 7 Conservation of mass and momentum 243

7.1 Particle dynamics 245
7.2 Rigid-body dynamics and statics 248
7.3 Conservation of mass and momentum in a deformable
continuum 260
7.4 Field equations for the elastic solid and viscous uid 276
Chapter 8 Elastic deformation 287

8.1 Estimating rock properties from geological eld tests 288
8.2 The idealized elastic material 292
8.3 Quasi-static displacement boundary value problems 299
8.4 Quasi-static traction boundary value problems 308
8.5 Elastic properties from laboratory and engineering
eld tests 319
8.6 Elastic heterogeneity and anisotropy 323
Chapter 9 Brittle behavior 333

9.1 Brittle deformation in the laboratory and in the eld 334
9.2 Strength of laboratory samples 337
9.3 Brittle failure in a eld of homogeneous stress 357
9.4 Brittle failure in a eld of heterogeneous stress 364
9.5 Fracture propagation and fault growth 371
Chapter 10 Viscous flow 384

10.1 Rock deformation by viscous ow 385
10.2 Constitutive relations for isotropic viscous uids 386
10.3 Plane and antiplane ow 388
10.4 Viscous ow in layers: mullions and folds 396
10.5 Flow of anisotropic viscous uids 416
Chapter 11 Rheological behavior 421

11.1 Departures from linear viscous ow 422
11.2 Boudinage and the non-linear power-law uid 423
11.3 Coupling of viscous ow and macroscopic diffusional
transport 440
CONTENTS vii
11.4 Continuum properties of composite materials 446

11.5 Anisotropic uids and internal instability 450
Chapter 12 Model development and methodology 456

12.1 Idealization of eld observations 457
12.2 Selection of general boundary conditions 462
12.3 A methodology for the practice of structural geology 474
References 478
Index 497
Preface
Fundamentals of Structural Geology is a textbook that and a physics course covering mechanics and
emphasizes modern techniques of eld data heat. We consider these courses to be the essential
acquisition and analysis, the principles of contin- mathematical and scientic pre-requisites for a
uum mechanics, and the mathematical and com- course using this textbook. Elementary concepts
putational skills necessary to describe, model, of vector analysis, matrix theory, linear algebra,
and explain quantitatively the deformation of ordinary and partial differential equations, and
rock in Earths lithosphere. computer programming with Matlab are used
With precise location data now available from throughout, but are introduced in such a way that
the Global Positioning System (GPS) and powerful a formal course in these subjects, while helpful,
computer systems now transportable in a back- should not be considered a pre-requisite. The
pack, the quantity of reproducible eld data has authors view this textbook as appropriate for a
increased dramatically. These new data sets rst course in structural geology, but recognize
demand better methods for describing the geom- that many students will come to a course using
etry of structures, and we address this demand by this book after a traditional course that empha-
introducing the basic concepts of differential sizes the descriptive terminology, geometric tech-
geometry, which provide unambiguous descrip- niques, and kinematic concepts of the discipline.
tions of curved lineations and surfaces in three Although designed as a text for students, this
dimensions. Data sets from a variety of eld areas book also should be useful as a reference for
are provided via the textbook website to promote researchers in structural geology, and as an aid for
the practice of opening eld notebooks to the updating instructors and professionals who have
entire community of researchers, and as input for been exposed only to traditional courses and text-
student exercises (see below). books on the subject. Furthermore, this book
Textbooks in structural geology provide el- should be attractive to scientists in related dis-
ements of continuum mechanics (e.g. separate ciplines (geophysics, rock mechanics, tectonics,
chapters on stress and strain), but rarely are these geotechnical engineering, and petroleum engi-
concepts tied together with constitutive laws or neering) who are looking for a modern summary of
formulated into equations of motion or equilib- the fundamentals of structural geology. We encour-
rium to solve boundary or initial value problems. age students and professionals from these disci-
These textbooks largely beg the questions: what plines to learn about the modern methods and
methodology should one adopt to solve the prob- tools of structural geology so that they can effec-
lems of structural geology; and what are the fun- tively interact with geologists on multi-disciplinary
damental constructs that must be acknowledged projects.
and honored? These constructs are the conserva- One of the opportunities and challenges of
tion laws of mass, momentum, and energy, com- publishing a textbook in the twenty-rst century
bined with the constitutive laws for material is the fact that the printed volume is no longer the
behavior and the kinematic relationships for only vehicle for communication between authors
strain and rate of deformation. We use these con- and readers. Accordingly, we have prepared a
structs to build a rational methodology for the homepage for Fundamentals of Structural Geology
investigation of tectonic processes and their struc- that is available on the World Wide Web (www.
tural products. cambridge.org/0521839270) and provides the fol-
This textbook is designed for senior under- lowing supplementary materials for readers,
graduate students and graduate students who instructors, and students:
have taken an introductory physical geology
course, mathematics courses that include differ- Full color images for all outcrop photographs
ential and integral calculus in several variables, used in the text
x PREFACE
Full color images for key graphical results used ization environment and speed, and professional
in the text programmers have written applications such as
Supplementary outcrop photographs, maps, Matlab that provide most of the computational
and cross sections tools needed by structural geologists.
A repository for supplementary images con- For the authors of this textbook, it is not
tributed by readers sufcient to focus on understanding the struc-
Exercises for students that reinforce the con- tural history of the Earth as an arcane academic
cepts introduced in the text exercise. We believe that structural geologists
Data sets from eld mapping campaigns for use can make important contributions in natural
in the exercises resource recovery (including water, oil, gas, and
Solutions to the exercises for instructors with minerals), in the assessment of natural hazards
password protection (including earthquakes, landslides, and volcanic
Sample Matlab m-les for the exercises eruptions), and in the management of the en-
Sample Matlab m-les for recreation of graph- vironment (for example the long-term storage of
ical gures found in the text radioactive materials and the contamination of
A repository for exercises and Matlab m-les fractured aquifers by hazardous chemicals). It is
contributed by readers the authors hope that students and instructors
Errata alike will be as captivated as we have been by the
remarkable opportunities and challenges of struc-
With a laptop connected to the Web and an LCD tural geology. Great satisfaction in the practice of
projector instructors can use the color outcrop this science is achieved when one successfully
images in the classroom to illustrate geological brings together the beauty of the natural world
concepts, and run the m-les with their own and the physical world of continuum mechanics
choice of parameters for a dynamic demonstra- to achieve a better understanding of rock defor-
tion of the mechanical concepts. We envision mation and the development of structures. By
readers of the textbook having this website open doing so one contributes to the knowledge of
on their desktop to enhance their learning experi- Earths remarkable history and to the solution of
ence. Today desktop PCs provide the necessary important practical problems facing society
CPU power, 3D graphics cards provide the visual- today.
Acknowledgments
David Pollard would like to acknowledge four Taixu Bai, Laurent Maerten, Scott S. Young, Frantz
teachers who shaped his understanding of struc- Maerten, Stephan Bergbauer, Peter Eichhubl,
tural geology as an undergraduate and graduate PhillipG.Resor,KurtR.Sternlof,PatriciaE.Fiore,Ian
student. Donald B. McIntyre of Pomona College W. Mynatt, W. Ashley Grifth, Nicolas Bellahsen,
provided the spark that ignited his curiosity about Gaurav Chopra, and J. Ole Kaven.
the subject and put it in an historical context. David Pollard would like to thank John Suppe
Arvid M. Johnson of Stanford University intro- of Princeton University and Patience A. Cowie of
duced him to the tools of mechanics and to a ratio- Edinburgh University for hosting sabbaticals that
nal way to approach physical processes in the eld provided important time for development of the
and laboratory. John G. Ramsay of Imperial College materials presented here. He gratefully acknowl-
taught him how to measure deformation in edges the help of the staff of the Department of
outcrop and investigate the geometry and kin- Geological and Environmental Sciences and the
ematics of rock subject to ductile deformation. Branner Earth Sciences Library at Stanford Uni-
Neville J. Price of Imperial College introduced him versity. Also, he extends special thanks to the
to rock mechanics and the analysis of rock subject Seeley G. Mudd Science Library at Pomona College
to brittle deformation. These teachers provided a and to the National Cello Institute for providing an
diversity of viewpoints of structural geology that idyllic venue for preparation of the manuscript.
was fascinating as well as challenging, and the Raymond Fletcher would like to acknowledge
origins of many of the themes played out in this several people who contributed to his education
textbook can be traced directly to their classrooms. as a structural geologist. William F. Brace (MIT)
Arvid Johnsons role in the formative stages of awarded him a C in the undergraduate structural
work on the textbook was particularly important. geology course, giving useful incentive for further
David Pollard was privileged to study with stu- study of a subject that Bills treatment showed to
dents who were colleagues at Pomona College, consist of an intriguing combination of eld obser-
Stanford University, and Imperial College, and later vation and mechanical analysis. Bill Brace also gave
to work with students in a teaching and advisory excellent advice on what not to do as a Ph.D.
capacity at the University of Rochester, the US research project prior to the arrival at Brown
Geological Survey (Menlo Park), and Stanford University of his Ph.D. advisor William M. Chapple.
University. Many of these students have partici- Bill Chapple provided guidance in formulating a
pated in research that helped to shape the concepts tractable complete mechanical model for the
and methods described in this book. They include: emplacement of a gneiss dome and M. A. Jaswon
Atilla Aydin, Zeev Reches, Gary R. Holzhausen, John pointed him toward a method of analysis. Inter-
W. Cosgrove, Otto H. Muller, David R. Dockstader, action with Bill Chapple over many years contin-
Paul T. Delaney, Paul Segall, Jon H. Fink, J. Russell ued to enrich his experience. The foundation for
Dyer, Russell K. Davies, Laurie L. Erickson, Marie D. his understanding of continuum mechanics was
Jackson, Peter C. Wallmann, Stephen J. Martel, provided by the lucid presentation of this subject
Allan M. Rubin, Larry G. Mastin, Jon E. Olson, Sarah in a two-semester course at Brown University by
D. Saltzer, Scott S. Zeller, Andrew L. Thomas, Carl E. E. T. Onat. Arvid M. Johnson introduced him to the
Renshaw, Roland Brgmann, Pauline M. Mollema, disciplined mapping of small-scale structures in
Marco Antonellini, Haiqing Wu, Peter P. the eld interspersed with more freewheeling dis-
Christiansen,Stephen K. Matthi, JoshuaJ.Roering, cussions of mechanical modeling. Memorable dis-
J. Ramn Arrowsmith, George Hilley, Emanuel J. M. cussions over coffee and pastry with Bernard Hallet
Willemse, Michele L. Cooke, Elissa Koenig, Juliet G. continue to provide him with imaginative ideas,
Crider, W. Lansing Taylor, Simon A. Kattenhorn, such as treating the Basin-and-Range Province as a
xii ACKNOWLEDGMENTS
string of blood sausages. He has beneted from George Gazonas, Bill Kilsdonk, Frank Irwin,
and enjoyed collaborations with former graduate Duncan Mardon, and Tom Patton.
students Judi Chester, Russell Davies, Jon Fink,
Chapter 1
Motivations and opportunities
Mt. Hillers, southern Henry Mountains, UT. The mountain is The sciences do not try to explain, they hardly even try
cored by igneous rock and surrounded by upturned beds of to interpret, they mainly make models. By a model is
sandstone and shale. G. K. Gilbert coined the term meant a mathematical construct which, with the addi-
laccolite for these structures in the late 1870s and tion of certain verbal interpretations, describes
proposed models for this process of mountain building based observed phenomena. The justication of such a math-
on mechanical principles. Inset: Frontispiece from G. K. ematical construct is solely and precisely that it is
Gilberts Report on the Geology of the Henry Mountains expected to work (quote from John von Neumann;
(Gilbert, 1877). To the rear of this illustration the Gleick, 1987, p. 273).
sedimentary strata form the structural dome of
Mt. Ellsworth, and to the front the eroded remnant of the
dome represents the current topography of this mountain.
Photograph by D. D. Pollard.
2 MOTIVATIONS AND OPPORTUNITIES
I
n this chapter we motivate the study of struc- President of the Geological Society of Americas
tural geology by introducing selected topics Structural Geology and Tectonic Division, con-
that illustrate the extraordinary breadth of cluded:
interesting problems and important practical We are at the end of the era when an unquestioning
applications of this discipline. For example, we public belief in the benets of basic scientic research
use the Imperial Valley earthquake of 1979 along almost automatically led to increased budgets at the
the San Andreas Fault zone to describe tech- NSF (National Science Foundation) Program level.
niques for geological hazard analysis. In a second Already, NSF management and the Congress want to
example the lineaments visible in radar images hear arguments about how research, and especially
of Venus provide the data for investigating tec- new programs, will address important social issues:
tonic processes on a planet other than our own. environmental changes and hazards, exploitation,
This is followed by an investigation of normal waste, and recycling of natural resources, and the like
(Cowan, 1992).
faulting in a hydrocarbon reservoir under the
North Sea, off the coast of Norway, to introduce Thus, whether a career in the Earth sciences takes
an application to petroleum exploration and one to industry or to academia or to a government
production. Then we describe the pattern of laboratory, the structural geologist should know
small faults, veins, and solution surfaces from how to address problems of social importance. To
an exposure in southern France, an example that this end, we integrate aspects of active tectonics,
demonstrates the practice of structural geology engineering geology, and petroleum geology into
at the human scale. The concept of anticracks this book to show how structural geology can
that emerged from this academic investigation contribute to solving problems in these areas.
is now being used to help explain the origin of Most inhabitants of southern California are
huge earthquakes a hundred kilometers below familiar with earthquakes and the geological
Earths surface. Finally, we describe a mecha- hazard associated with living in an active tectonic
nism for mountain building that was discovered province, although the recurrence time of major
in the Henry Mountains of southern Utah in the events is great enough to instill a sense of com-
late nineteenth century by one of the pioneers of placency in many citizens. On the other hand,
structural geology, G. K. Gilbert. Earth scientists and government ofcials are
The frontispiece for this chapter is a photo- acutely aware that destructive earthquakes could
graph of Mt. Hillers in the southern Henry occur at any moment. Teams of scientists and
Mountains. Like all the photographs that appear engineers supported by federal and local govern-
as grayscale images in this book, a color image ments are monitoring the continuing activity of
of this photograph is available at the textbook the faults in this area and have tools in place to
website along with images of related exposures capture data from the next signicant event (Yeats
and scenes. These are presented as monitor et al., 1997).
resolution images for quick viewing with a web What are the data that these scientists and
browser or for LCD projection in the classroom for engineers are hoping to capture? Perhaps the
teaching purposes. most fundamental aspect of faulting is the fact
that the rock and soil on either side of the fault
slip past one another. There is relative motion of
1.1 Earthquake hazards in these two masses more or less parallel to the fault
surface. For example, Fig. 1.1 is a photograph
southern California taken across the trace of the Imperial Fault in the
Imperial Valley of southern California shortly
Academic researchers have learned that society after a magnitude 6.5 earthquake struck on
may not be content to continue funding the October 15, 1979. The vertical surface just behind
arcane studies of ancient rocks that have been the the observers feet is one surface of the fault
mainstay of the National Science Foundations exposed at the time of the earthquake. Relative to
Tectonics Program in the past. Darrel Cowan, then the ground on which the observer is standing, slip
1.1 EARTHQUAKE HAZARDS IN SOUTHERN CALIFORNIA 3
on the fault offset the two small drainage chan-

nels upward and to the right. By identifying soil
particles (say in the bottom of the drainage
channel) that were adjacent before the slip, one
can make precise measurements of the offset.
Using a tape measure, the geologist records the
horizontal, strike slip component of relative
motion as about 5 cm, and the upward, dip slip
component as about 20 cm.
To characterize the behavior of a fault, one
would like to know the magnitudes and direc-
tions of this relative motion in terms of the
displacements, velocities, and accelerations of Fig 1.1 Ground rupture along the northern trace of the
originally adjacent particles over the entire fault. Imperial Fault in southern California after the October 15,
1979, magnitude 6.5 earthquake. View is to the southwest.
The relative motion of particles is directly mea-
The strike and dip components of slip are identified based on
surable only at (or very near) the surface of the
the offsets of the small stream channels. The relative motion
Earth for active faults, and yet the fault might is right-lateral strike slip (5cm) and dip slip (20cm) down
extend to depths of 10 km or more. Furthermore, to the northeast. See website for color image. Photograph by
one would like to know the distributions of these D. D. Pollard.
quantities over the entire time the two surfaces of
the fault were in relative motion. In other words
one would like to know the spatial and temporal the earthquake and a time after the earthquake,
distributions of displacement, velocity, and accel- and cannot measure the velocities or accelera-
eration for particles of rock or soil in the vicinity tions that occurred during the slip event.
of the fault. Given such information we could Although the data gathered by geologists
begin to understand the mechanisms that control provide the most direct measurement of slip at
fault slip and, perhaps, be in a position to be the Earths surface, they only record the slip at
predictive about such events. certain points along the fault and these data
may not be similar to the distribution of slip at
1.1.1 Contributions from geology, depth. For example, the offset of a fence line at the
geodesy, and geophysics surface may be strongly inuenced by a thick
Figure 1.2 is a schematic illustration of some of layer of relatively soft soil or unconsolidated
the tools used to monitor the slip across faults in sediments overlying the more rigid rock below.
active tectonic regions (Thatcher and Bonilla, Models are required to interpolate the surface slip
1989). The illustration in Fig. 1.2a represents a ver- between these data points and to extrapolate
tical cross section along the fault with contours of these surface measurements to the sub-surface.
slip magnitude. The tools used to estimate the slip Using elasticity theory, one could specify remote
distribution fall within three different disciplines stresses and stresses along the fault as boundary
in the Earth sciences: namely geology, geodesy, conditions and solve for the slip distribution over
and geophysics. The geologist measures the offset the fault surface. One could search for boundary
of geological structures and formations across a conditions that produced a slip distribution best
fault at the surface as well as the offset of what- matching the slip measured at the surface. Of
ever cultural markers might be present (Fig. 1.2b). course the model parameters themselves may be
By walking along the surface trace of the fault, the poorly constrained, and there may be many possi-
structural geologist can gather data on many dif- ble slip distributions at depth that are consistent
ferent types of geological and cultural features with data from the surface. None-the-less, such
and plot a graph of fault slip at the surface versus modeling exercises are the only way for the geolo-
distance along the fault. Usually the geologist gist to extrapolate data from the surface to the
records only the total slip between a time before sub-surface.
(a) stable monuments. In some cases the instruments

are permanently mounted at the survey locations
Earthquake and record data that can be used to calculate
slip distribution velocities and accelerations. In these respects the
geodetic data can provide a better constraint on
the deformation associated with faulting.
(b) On the other hand the benchmarks usually are
Geologic data not located at the fault itself, so they do not
Surface slip
directly record fault slip, even at the surface.

Rather, a model (usually based on elasticity
Distance theory) is employed that requires as input the
location and geometry of the fault and the
(c) mechanical behavior of the rock mass underlying
Geodetic data the geodetic network. These models usually treat
Geodetic slip
the fault as a set of segments, each with a constant

slip, so the output is slip at the surface for differ-
Distance ent segments of the fault (Fig. 1.2c). The geodeti-
cally inferred slip is consistent with the changes
(d) in line lengths or angles between the benchmarks
Seismologic data of the array, but clearly depends upon the chosen
Moment release
segment geometry and the other model par-
distribution
ameters. More elaborate models are capable of
calculating slip distributions at depth from the
geodetic data. Because the geodetic data come
from widely scattered locations away from the
(e) InSAR data fault, the geometry and mechanical behavior of
Slip distribution the sub-surface materials over a large volume of
from inversion rock must be provided as model input.
The third category of data is taken from seis-
mograms recorded both in the vicinity of the
fault and at distant stations at the time of the
earthquake (Fig. 1.2d). Although the locations of
Fig 1.2 Schematic diagram of four different methods for the seismographs may be even more remote from
estimating the slip on a fault (Thatcher and Bonilla, 1989). the fault than the geodetic benchmarks, these
The actual slip is contoured on the fault surface in (a). instruments continuously record the shaking of
Illustrations (b)(d) show how geologists, geodesists, and the ground due to the passage of seismic waves
seismologists gather data (left column), and graphical generated at the fault. Therefore, they can
representations of these data are shown to the right. provide a wealth of data for inferring the behav-
(e) Interferometric synthetic aperture radar (InSAR) data
ior of the fault. In this example pulses on the seis-
provide the field of displacement at the surface near a fault
mogram are correlated to areas on the fault at
which can be inverted to estimate the slip distribution.
depth that slipped at slightly different times or
at different distances from the recording instru-
The geodesist measures the changes in ment. What is actually calculated is the seismic
lengths, angles, and/or elevations between sur- moment on the fault over these areas, but this can,
veyed benchmarks usually located at scattered in principle, be related to the average slip. By
points some distance from the fault (Fig. 1.2c). combining data from many seismographs a
Such measurements are often more precise than picture of the moment release distribution on
geological measurements because high-precision the fault can be constructed. In practice the
instruments are used to gather the data and the instruments may not be ideally located, and
bench marks are xed to carefully designed and there may not be as many as one would desire.
Models of the sub-surface fault geometry are

33o 15'
needed as well as the mechanical properties
(seismic wave velocities) of the rock from the SALTON SEA
EA
fault to the location of the seismographs.
ST
The use of interferometric synthetic aperture
radar (InSAR) for the detection of ground dis- 111
placements associated with earthquakes was 86
HIG
highlighted in articles appearing in the early
HLI
1990s (Massonnet et al., 1993; Prescott, 1993;
NE
Zebker et al., 1994). The radar signal is transmitted 33o 15'
78
from a satellite to the ground surface where it is
reected back to the satellite and recorded as a 111
Brawley
set of pixels making up an image of the surface. Fault
Knowledge of the travel time and speed of the zone
CANAL
Im
0 km 10
pe
signal provide the information necessary to cal-
ria
l
culate the range, or distance, from the satellite
to each reective site on the surface. If the same 8
region is imaged at two different times, for 32o 45'
Fa
111
example before and after the earthquake, the dif-
u
lt
ference between the two images can be used to cal-
ATES
culate the component of the surface displacement UNITED ST
MEXICO Epicenter
directed toward the satellite. The resulting image 115o 45' 10/15/79
115o 30'
(Fig. 1.2e), called an interferogram, is similar to a
contour map of the displacement component on
Fig 1.3 Map of the region affected by the October 15,
which the white and black bands (called fringes) 1979, earthquake in southern California (Wosser et al.,
are the contours. The fault segments are shown as 1982). The epicenter is shown as a star in the lower right-
ne white lines superimposed on this image. By hand corner.
invoking a model (usually based on elasticity
theory) for the location and geometry of the fault
segments and the mechanical behavior of the the concepts and contributions from geophysics
rock mass, one may use this displacement dis- and geodesy to the study of faulting. In addition
tribution on Earths surface to calculate the important insights are attained from studying the
corresponding slip distribution on the fault. The effects of faulting on the geomorphology of the
abundance of data provides considerable con- landscape (Arrowsmith et al., 1996; Arrowsmith
straint on the unknown slip distribution below et al., 1998). The most comprehensive view of
Earths surface and very exciting avenues for new faults and the faulting process will come from an
research on faulting. integration of all these data and that integration
It should be obvious from this discussion that will be most effective in the context of building
the different disciplines contribute information well-constrained models.
that is based on different observations in different
locations and over different length and time 1.1.2 Conceptual and mechanical models
scales. Yet scientists from all three disciplines are for the 1979 earthquake rupture
studying the same physical phenomenon, fault- On October 15, 1979, the magnitude 6.5 earth-
ing, and they are using the same tools to build quake rupture began just south of the USMexico
their models, namely elasticity theory. In this border and spread approximately 35 km to the
textbook we focus on the geological data and the north into southern California (Fig. 1.3), breaking
models that are used to relate measurements ground along the trace of the Imperial Fault
of slip to fault behavior. On the other hand each (Johnson et al., 1982; Wosser et al., 1982). Many
discipline is providing important pieces of the agricultural features such as fence lines and
puzzle, so structural geologists should be aware of canals provided markers to measure the slip
Map view B
A A'
B'
Cross sectional view Shear

A A' B B' resistence
0
u
~3 km/ s1 FR FR
A
10 ?
Depth
QP QP
20
(km)
Fig 1.4 Three views of a crustal-scale strike slip fault. Map on faults? Some answers to this question have
view illustrates the fault as a zone of deformation. Cross
come from research by scientists and engineers
section AA in the fault plane includes a contour map of the
over the past few decades, but much remains to be
slip (u) which goes to zero at the fault tipline and is greatest
near the hypocenter (star). Cross section BB perpendicular
understood.
to the fault plane suggests that slip mechanisms are frictional In the previous section we described how geol-
resistance (FR) in the upper part of the crust and localized ogists, geodesists, and geophysicists use models to
quasi-plastic flow (QP) in the lower part. The graph at the extrapolate information on displacements or
right indicates a linearly increasing resistance to shearing with accelerations from the locations where data are
depth to the brittleductile transition, and then a non-linear measured on the Earths surface to the fault in the
decreasing resistance to shearing with depth. Reprinted from sub-surface. These models help us to understand
Sibson (1989) with permission from Elsevier.
the behavior of faults where they cannot be
observed directly and they provide insights
across the fault trace. The farmers, homeowners, concerning earthquake faulting as a structural
businesses, and municipalities in the Imperial process. The faulting process is conceptualized at
Valley, mostly around the town of El Centro, sus- the crustal scale in Fig. 1.4 for a vertical fault with
tained over twenty million dollars in damage. strike slip motion (Sibson, 1989). Each view of this
Fortunately, there was no loss of life and few cata- conceptual fault model reveals different aspects
strophic failures of man-made structures in this of faulting at the crustal scale. The map view
event. On the other hand, earthquakes of similar shows a zone of fractures and deformation, rather
magnitude often are accompanied by many than two surfaces in contact. This suggests that
deaths in regions with less stringent building faults can be more complex than a single fracture
codes, or no building codes at all. These events and that shearing of material in a fault zone may
testify to the destructive power of earthquakes characterize the deformation rather than slip
and to the need to understand such hazards. between two surfaces. The vertical cross section
Because earthquakes are generated by sudden slip viewed parallel to the fault indicates that fric-
on faults, we need to understand the mechanisms tional resistance (labeled FR in Fig. 1.4) to slip on
and behaviors of faults in order to develop a fault operates to depths of perhaps 10 km and
informed hazard mitigation policy. Just what are plastic ow (labeled QP) is associated with dis-
the causes and consequences of dynamic rupture tributed shearing in a zone at deeper levels. Thus,
the mechanisms of faulting may change with (a) Map view of rupture
depth as temperature and pressure increase, such 0 10 20 30 km
that brittle fracture and friction dominate at Distance along rupture
shallow depths and ductile ow dominates at Epicenter
Imperial Fault
greater depths. In this conceptual model the resis-
tance to shearing increases with depth to this Br
aw
le
transition and then decreases with depth. In a ver- y
Fa
ul
tical section viewed perpendicular to the fault N t
(AA), dynamic shearing begins at depth, near
the brittleductile transition and spreads out over Vertical cross sections of rupture
(b) Rupture time (s)
the fault surface at a velocity of about 3 km s1, 0
12
eventually reaching the Earths surface.
Hypocenter
The Imperial Valley earthquake is noteworthy 5 10
because it occurred within a dense array of geo- 6

8
4
10 2
detic and geophysical instruments and there were
abundant cultural features for the geologists to (c) Slip duration (s)
measure at the surface (Savage et al., 1979). The 0
0.4 0.4
mechanical model reviewed here was constructed 5 0.8
1.2
using data from the seismographs and strong 1.6
motion instruments that monitored this event 10 1.2
(Archuleta, 1984). The results are not unique and (d) Strike slip offset (m)
the choice of model parameters could be debated, 0 0.2
but that is not the issue here. This model provides 0.2
5
an excellent example of the insight one can gain 0.6
1.0 1.0
about phenomena that are otherwise totally inac- 10
1.4 1.0
0.6 0.6
cessible to direct observation.
Figure 1.5a is a map of the rupture traces for (e) Dip slip offset (m)
0
both the Imperial and Brawley Faults as compiled 0.08 0.16
by geologists from observations at the Earths 5
0
surface. The photograph shown in Fig. 1.1 was
10
taken near the northern end of the Imperial Fault.
The map also shows rupture traces along the 0 10 20 30 km
Brawley Fault that trend oblique to the Imperial
Fault. Apparently the Brawley Fault slipped at Fig 1.5 Map and cross sections of the Imperial Fault and
about the same time as the Imperial Fault, but the the Brawley Fault for the October 15, 1979, earthquake in
southern California (Archuleta, 1984): (a) map of the rupture
relative motion on the Brawley Fault was primar-
trace; (b)(e) vertical cross sections parallel to the fault trace
ily dip slip. Note that the southern half of the
with contours of the model rupture time, slip duration,
Imperial Fault rupture trace is drawn as continu- strike slip offset, and dip slip offset.
ous, whereas it is drawn as composed of discrete
segments in the northern half. Also shown on
Fig. 1.5a is the rupture epicenter, the point at the paragraph brings up interesting questions about
surface of the Earth immediately above the point faulting. Why did the rupture not break to the
where rupture initiated, as inferred by geophysi- surface immediately over the epicenter? Why
cists from seismic records. This location is depen- would a second fault rupture at the same time as
dent upon a model for the seismic wave velocities the Imperial Fault, and why is the trace of the
of the crustal rocks. Note that the epicenter was second fault obliquely oriented? What does the
approximately 5 km south of the southernmost discontinuous nature of the rupture trace tell us
surface break. about faulting? Some of these questions can be
Each of the observations made in the previous addressed with models for the rupture process.
The mechanical model for the October 15, these dots and the hypocenter the model fault has
1979, earthquake event considers only the slipped and stopped, whereas to the north and
rupture along the Imperial Fault (Archuleta, above the 8-s contour the fault has not yet slipped.
1984). The lower four panels of Fig. 1.5 are graphs At any particular location on the model fault the
of different physical quantities calculated using slipping occurred over a period of time ranging
the model and plotted on a vertical planar section from a fraction of a second to almost 2 s as the
that approximates the more complex geometry of rupture front passed, and then slipping stopped.
the actual fault as suggested by the mapped trace Figures 1.5d and e show, respectively, two com-
in the rst panel. The model fault is about 12 km ponents of slip between the model fault surfaces
in depth (ordinate) and 35 km in length (abscissa). after the rupture has completed its propagation
The physical quantities (rupture time, slip dura- from the hypocenter to the northern termination.
tion, strike slip, and dip slip) are represented by Strike slip varies from 1.4 m near the bottom
contours of equal magnitude on these graphs. center of the fault to a few decimeters or less at
Together these panels provide a remarkable visu- the surface. The strike slip is zero along the south-
alization of the model slip event from the Earths ern portion of the fault at the surface and this is
surface to the bottom of the rupture. consistent with the observations shown on the
Figure 1.5b illustrates the position of the map in the rst panel. Note that the surface mea-
leading edge of the model rupture to the north of surements of slip, amounting to about 20 cm,
the point of rupture initiation, the hypocenter, at under-represent the slip at depth by a factor of
times measured in seconds after initiation. What eight or more. The model fault slipped much
happened to the south of the hypocenter is more at depth than at the surface. Dip slip is con-
ignored on these panels. At a given time, say 4 s, centrated near the surface at the northern end of
that portion of the fault between the hypocenter the model fault with magnitudes approaching
and the 4-s contour has slipped, while elsewhere a few decimeters. This is consistent with the
on the fault no slip has occurred. Clearly, slip on geological observations (see Fig. 1.1) that indicate
the model fault does not initiate everywhere the rocks on the northeastern side of this part of
simultaneously. Rather, the model rupture initi- the Imperial Fault went down relative to those
ated at a point, at the hypocentral depth of about on the southwestern side. The amount of dip slip
8 km. Then, the rupture front advanced rapidly to at the surface (up to about 20 cm) also is consis-
the north and less rapidly upward toward the tent with the eld observations. The relative
surface. The rupture took a total time of about 12 s motion on the Brawley Fault was also nearly pure
to spread the 35 km to the north end of the model dip slip with the northwestern side down. In fact,
fault. Thus, the average rupture velocity was about the region between the Imperial and Brawley
3 km s1 toward the north, approximately the Faults is a topographic depression occupied by a
speed of seismic shear waves. (usually) dry lake-bed. This suggests that the rela-
Figure 1.5c shows the total time that originally tive motion experienced during the 1979 earth-
adjacent particles on the two surfaces of the quake is typical of the recent geological history of
model fault were in relative motion. For example, this fault system.
along the contour labeled 1.6 s the two surfaces This mechanical model gives us a picture
slipped for a total time of less than 2 s. You might of active faulting that is reasonably consistent
nd this surprising given the fact that the total with the available surcial and seismic data from
duration of faulting was about 12 s. Clearly all the 1979 event. It informs our intuition about
parts of the model fault were not slipping at the the physical process of faulting and provides a
same time. This is illustrated in the previous panel glimpse into possible behavior along the Imperial
by the pattern of dots next to the 8-s contour. Fault at depth. Building models such as this one
These dots cover the relatively small portion of and using these models to understand the process
the fault that has already slipped and is still in the of faulting is an exciting area of research in which
process of slipping at the moment that the structural geologists can participate (Segall and
rupture front lies along the 8-s contour. Between Pollard, 1980; Aydin and Schultz, 1990; Cowie and
1.2 RADAR LINEAMENTS ON VENUS 9
212o 213o 214o 215o 216o E

Scholz, 1992; Dawers et al., 1993; Brgmann et al.,
1994; Muller et al., 2003).
-13o
1.2 Radar lineaments on Venus

-14o
In his book The Assayer Galileo Galilei apparently
wrote the following:
-15o
The Universe, which stands continually open to our
gaze, cannot be understood unless one rst learns to
comprehend the language and read the letters in
which it is composed. It is written in the language of -16o
mathematics . . . (Gregory, 1990). 100km
S
Today the power of modern telescopes and satel-
lite exploration of the planets provides countless Fig 1.6 Left-looking F-MIDR 15s214 radar image of the
surface of Venus from the Magellan mission (Koenig and
opportunities to investigate structures in rock
Pollard, 1998). Look angle is approximately 40. Note radial
beyond Earth. pattern of lineaments centered at 15S, 215E. One degree
The Magellan mission to Venus produced radar of latitude or longitude is about 100 km.
images of much of the planets surface and many
of the structures observed on these images appar-
ently are related to volcanism (Head et al., 1992). selves are not lled with magma; rather they are
One of the most interesting classes of structures is the ephemeral manifestation of a large rising
composed of radar lineaments (paired bright and body of magma beneath the surface. The second
dark lines or single bright lines on the image) that type of lineament pattern is interpreted as having
appear to radiate from a central focal point like formed as magma-lled fractures, dikes, that prop-
spokes on a bicycle wheel (Fig. 1.6). Grosls and agated upward and radially outward from a
Head (1994) have identied more than 160 such central magma conduit at shallow depths under
radial systems on Venus and have interpreted the the volcanic edice. The dikes act as the passage-
lineaments as fractures cutting the surface of ways for magma ow to the surface from the
the Venusian crust. At rst glance the patterns are central conduit. Here we only consider the second
reminiscent of radial fracture patterns you might type, which apparently makes up more than 70%
have seen in a pane of tempered glass where it has of the radial patterns identied on Venus.
been struck by a rock; however, these lineament
patterns are enormous. The pattern at (15S, 1.2.1 Conceptual and mechanical models
215E), shown in more detail in Fig. 1.7, is about for graben formation
200 km in diameter and the average diameter for The volcanic edice at (15S, 215E) stands about
all such patterns identied on Venus is 325 km, 1 km above the surrounding plains. Elongate and
with some as great as 2000 km! These patterns are lobate gray regions (labeled rlf on Fig. 1.7) on the
intriguing in their symmetry and awe inspiring in surface of this edice are interpreted as lava ows
their size; they clearly warrant our attention as (Koenig and Pollard, 1998), which spread down
structural geologists. the anks of this large volcano before solidifying.
Grosls and Head (1994) developed conceptual Also visible on the radar image are paired radial
models to distinguish and interpret two types of bright and dark lines that extend up to 50 km
radial lineament patterns. For the rst type they down the slope (labeled g on Fig. 1.7). These are
suggest the fractures are formed by doming and interpreted as graben, linear depressions about 1
stretching of the Venusian crust over a body of to 2 km in width bounded by normal faults along
molten rock, magma, that ows upward from a which the central block of rock has moved down-
source reservoir at depth. The fractures them- ward. When obliquely incident radar signals are
Graben formed
above dike
0.0
2 Normal fault 0.2
Depth (km)
0.0 0.4
4 0.2
g 0.4
0.4
0.2
6 Contours of
Magma-filled 0.4 horizontal
dike 0.2 0.0 stress
0.0
8
5 0.0
0 5
tlf
0.0
f Distance (km)
rlf
Fig 1.8 Schematic block diagram of a graben bounded by
two normal faults and underlain by a dike (Rubin and Pollard,
50 km 1988; Koenig and Pollard, 1998). The horizontal stress
component induced by opening of the dike is contoured on
the front view showing a stress shadow (negative,
Fig 1.7 Left-looking F-MIDR 15s214 radar image of the
compressive stress) to either side of the dike and a stress
southeast quadrant of the volcanic edifice centered at 15S,
concentration (positive, tensile stress) near the dike tips.
215E. Surface structures are identified (Koenig and Pollard,
1998) as graben (g), fractures (f), terminal lava flows (tlf ),
and radial lava flows (rlf ). The graben and fractures
Horizontal stretching is caused by tensile stresses
radiate from the volcanic center in the upper left corner of
that tend to pull the rock apart. Therefore, one
the image.
needs to determine if dike opening at depth could
induce tensile stresses near the Earths surface,
reected off the fault surface on one side of the where normal faults bound the graben. Such as-
graben, returning a bright lineament, the fault sociations are found in volcanic regions on Earth,
surface on the other side of the graben lies in a including those in Iceland and Afar.
radar shadow and forms a dark lineament. Thus The mechanical model of this phenomenon is
the surface topography accounts for the closely based on principles that are formulated into a set
spaced pairs of bright and dark lines on the image. of mathematical equations known as the theory of
Figure 1.8 illustrates the conceptual model for elasticity (Timoshenko and Goodier, 1970). This
the development of graben over igneous dikes. theory and the relevant equations are described in
The pairs of normal faults bounding graben on detail later in this textbook. For the moment you
Venus are interpreted as having formed because of only need to know that these equations can, for
the local horizontal stretching of the rock imme- example, be solved to determine the stress
diately over or ahead of vertical dikes. Because the distribution in the rock mass surrounding a dike.
dikes apparently propagated upward and outward This formulation is called a boundary value problem
from a central magma chamber located under the because one prescribes the stresses on the bound-
summit of the edice at (15 S, 215E), the graben aries of a body and the governing equations of elas-
form a radial pattern on the anks of this volcano. ticity theory are used to calculate the stresses in
A couple of questions come to mind when the interior. In this case one boundary represents
thinking about the origin of graben as described in Earths surface, which is free of stress, and the
the previous paragraph. Does the opening of a dike other boundaries represent the dike walls that are
actually lead to stretching at the surface? If it does, subjected to stresses equal to the outward-directed
why should two normal faults form to either side of pressure of the magma. The magma pressure
the dike instead of one immediately over the dike? pushes the dike walls apart and distorts the sur-
These questions have been addressed by studying rounding rock mass, thereby inducing a change in
the physical relationships between normal faults the stress distribution that is not easy to imagine
and dikes (Rubin and Pollard, 1988; Rubin, 1990). without the aid of elasticity theory.
1.3 FAULTING IN A NORTH SEA HYDROCARBON RESERVOIR 11
On the front face of the block diagram in Fig. 1.8 After many years of observing geology departments,
values of the horizontal stress are shown using a our experience indicates that they have adjusted to the
contour plot. This illustrates quantitatively how the scientic revolution of recent decades, but have largely
opening of a model dike changes the stress eld in ignored the technical, economic, and social changes
the surrounding rock. To either side of the dike neg- that inuence the practice of geology and thereby
have ignored the professional lives of their students.
ative (compressive) stresses are induced that push
inward on any small element of rock. These Structural geologists can make important contri-
induced compressive stresses would tend to prevent butions to natural resource recovery, including
normal faults from forming because they increase water, oil, gas, and minerals. They can play key
the frictional resistance to sliding. However, imme- roles in the management of the environment, for
diately over the dike, positive (tensile) stresses are example in the long-term storage of radioactive
induced. These tensile stresses stretch the rock and materials and the contamination of fractured
could contribute to the formation of normal faults. aquifers by hazardous chemicals.
The distribution of horizontal stress at the surface One of the most interesting new areas of
has two maxima located symmetrically about the research in structural geology related to natural
plane of the model dike. These two maxima corre- resources is the investigation of folds and faults
late with the two normal faults that develop to using data from seismic reection surveys
either side of the dike plane. In this way the model (Kattenhorn and Pollard, 2001). These data are
has successfully addressed the questions we asked gathered by the petroleum industry to image
about the relationship between dikes and graben. hydrocarbon reservoirs (Sheriff and Geldart,
Does the success of the model provide all one 1995). To carry out a reection survey in a sedi-
needs to know to interpret the radar lineaments mentary basin, seismic waves are generated by
on Venus? Certainly it is supportive of the concept impulsively striking the surface of the Earth.
that dikes can induce slip on normal faults, but it Depending upon the depth of investigation
does not prove that dikes actually exist beneath required, these waves are generated using explo-
the graben. Some of the paired radial bright and sive charges, mechanically driven vibrators, or
dark lines on the radar image at (15S, 215E) can simply a hand-held hammer. Some of the waves
be traced down the anks of the volcano where travel down into the Earth, reect off sedimentary
they merge into a single bright radar lineament layers or other structures, and travel back to the
(labeled f on Fig. 1.7). These single lineaments surface where they are recorded by a string of
are interpreted as open ssures that formed as portable seismographs laid out along the survey
dikes neared the surface (Koenig and Pollard, line. The depth to different reecting horizons
1998). This interpretation is consistent with the can be computed by identifying the two-way travel
presence of dark lobate regions, interpreted as lava time (down and back) for each reection, and by
ows, emerging from near the distal end of some knowing the velocity for acoustic waves in the
ssures and extending outward in a radial direc- rock. A series of impulses is generated along the
tion (labeled tlf on Fig. 1.7). These ows provide survey line and reections are recorded at each
compelling evidence that dikes underlie the radial seismograph. The abundant data help to reduce
fractures and graben, and are the conduits for the the uncertainty in identifying and locating the
escape of magma from beneath the volcano. reecting horizons using a variety of data pro-
cessing techniques.
The output of this processing is a seismic
1.3 Faulting in a North Sea reection cross section (seismic section) of the Earth
immediately under the survey line. For example,
hydrocarbon reservoir consider the northern part of the North Sea
(Fig. 1.9a) on the Norwegian Continental Shelf
R. N. Farvolden and J. A. Cherry (1991) wondered if (Maerten et al., 2000; Maerten et al., 2002). Figure
geology departments are preparing their students 1.9b is a seismic section taken from approximately
for the twenty-rst century: 150 km west of the Norwegian coast (Faerseth
(a) the reectors in the pile of sedimentary layers.

N Note that these reectors are not continuous
Oseberg
across the section from west to east, suggesting
l
that they have been offset by normal faults or
truncated by unconformities.
l
l
l
l
l
l
l
l
l l l l l l l l l l
NORWAY Specic sedimentary horizons are interpreted

l l l l l
l
l
l
in Fig. 1.9b including the top of the Brent Group,

l
l
l
l
l l
l
l l
l
l l
l
l l l l l
l
l l l l l l
the top of the Heather Formation, and the base of

l l
l l
l
l
l l l
l
l l l l
l
l
l
l l l l l
Bergen the Cretaceous. Interpretations of the normal

l
l l
l l l l
l
l l
l l
rm
Shetland faults (thick steeply inclined white lines) were

r a b en
l
l l
l l l
o
l
l
Platform
Platf
l
Oslo drawn and labeled by Faerseth and his colleagues

l
l
l l
l
l l
l l
l l
gG
l
l
l l l l
l
from the Norwegian oil company Norsk Hydro

l
a
l
l l
l
l l
l
Hord
l
Vikin
l l
l
l l
ASA. The faults strike approximately northsouth

l l
l l
l l l
so seismic sections along eastwest trends such as

NORTH SEA in Fig. 1.9b should display these faults as viewed
100 km along their strike. The two east-dipping faults
offset the top of the Brent Group, but do not offset
(b) W E the overlying base of the Cretaceous. Because
1500 these faults offset formations of upper Jurassic
Two way travel tme (ms)
1550
age, but do not extend into the overlying
1600
1650 Cretaceous strata, they are interpreted as having
1700 formed during Jurassic extension of the basin. The
1750
west-dipping fault interpreted in Fig. 1.9b cuts the
1800
1850 top of the Brent and extends slightly above the
1900 base of the Cretaceous. It also offsets one of
1950
the east-dipping faults. Because the west-dipping
2000
faults in this region systematically offset the east-
dipping faults, Faerseth and his colleagues con-
Fig 1.9 (a) Regional map showing the location of Oseberg clude that the west-dipping faults are younger.
Syd Field on the Norwegian Continental Shelf. A combination of factors, some related to
(b) Interpreted seismic data from the northern part of the improved equipment for acquiring such data and
North Sea approximately 150km west of the Norwegian
others related to improved software and hardware
coast. Two-way travel time plotted versus distance along
for processing such data, have changed the
section. Interpreted horizons include the base of the
Cretaceous, the top of the Heather Formation, and the top
quality and resolution of seismic imaging tech-
of the Brent Group. Reprinted from Maerten et al. (2002) nology dramatically in the past few decades.
and Faerseth et al. (1997) with permission from Elsevier. These developments are largely driven by the need
for oil and gas companies to improve their explo-
ration strategies through better images of sedi-
et al., 1997). For offshore surveys the seismographs mentary basins, and to improve their production
are towed in long strings behind a ship and the strategies through better characterizations of par-
sources for the acoustic waves are implosions ticular reservoirs within these basins. The older
created by submerged devices called air guns. The technique is essentially two dimensional, provid-
seismic section plots position from west to east ing an image of the strata on a vertical cross
along the survey line (see kilometer scale) versus section along the survey line. By capturing numer-
two-way travel time measured in milliseconds ous closely spaced parallel lines in two orthogonal
(ms). For a typical wave velocity of 2 km s1, the directions, enough information is gathered to
bottom of the section at 2000 ms 2 s represents a interpolate the reecting horizons between these
depth of about 2 km. The processed data appear as two-dimensional surveys and produce a three-
a series of bright and dark stripes representing dimensional image of the horizons within a
volume of rock. The seismic section in Fig. 1.9 has 2 km

been combined with hundreds of additional sec-
tions surveyed both parallel and perpendicular to
this one. Using such three-dimensional surveys N
the opportunity exists to characterize and map
the three-dimensional geometry of the faults.
Both the acquisition and the processing of
seismic reection data are in the domain of geo- Omega North
physics so we will not dwell on this aspect of the
subject. On the other hand the images captured
by these surveys illuminate both the sedimento-
logical and the structural heterogeneities in
the rock mass. These images have provided us Well 8
with the challenging problem of understanding
folding and faulting in three dimensions, no
longer limited by poor exposure and arbitrary ero-
sional slices through these structures. Instead of
searching for the few exposures of a poorly
revealed fault, hoping to measure the offset of one Well 10
or two geological markers, we are able to measure
the offset of continuous reecting horizons at
many points across a fault. Instead of locating the Omega South
two points in space that represent the fault termi-
nations at either end of an exposed fault trace, we
are able to locate the fault tipline at many points in
three dimensions. Of course there is a limit to the
resolution of these images, now about 10-m offset
for good reectors, but the techniques are improv-
ing yearly. Horizontal
separation
1.3.1 Conceptual and mechanical models Dip
Fault
for fault linkage
As an example of using structural geology to
analyze three-dimensional seismic data we turn to Base of Dip
Brent separation
the Oseberg Syd Field operated by the Norwegian
company Norsk Hydro (Faerseth et al., 1997). The
faulting in two parts of this eld, called Omega
North and Omega South (Fig. 1.10), has been inves- Fig 1.10 Map of normal faults in part of the Oseberg Syd
Field in the northern North Sea based on the structural
tigated using a three-dimensional seismic survey
separation of the base of the Brent Formation. Wells No. 8
(Maerten, 2000; Maerten et al., 2000, 2002). The
and No. 10 penetrate the producing reservoir and data from
map depicts the base of the Brent Formation, one these wells suggest the Omega North and Omega South
of the hydrocarbon reservoirs. The black stripes of regions are isolated from one another by sealing faults. The
variable width represent the normal faults that dashed square indicates the area of faulting investigated by
cut this formation and separate pieces of the Brent modeling. Inset: Schematic vertical cross section that
Formation in map view. The major faults in this illustrates how the width of the fault stripe on the map (gray
eld strike approximately northsouth and dip to on cross section) depends upon the dip of the normal fault
the west. A second set of faults is less well devel- and the dip separation of a particular horizon. Reprinted
from Maerten et al. (2000) by permission of the AAPG
oped, striking approximately northwestsouth-
whose permission is required for future use.
east and dipping to the southwest.
The inset on Fig. 1.10 illustrates in cross producing the hydrocarbons from this reservoir
section how the horizontal component of separa- (Maerten et al., 1999, 2000). For example, if part of
tion (sometimes called the heave) and the dip of the target reservoir is separated from a well by a
the normal fault are represented by the width of sealing fault those hydrocarbons beyond the fault
the black stripes on maps. Greater stripe widths cannot be produced and another well must be
(gray on the inset) imply greater horizontal separ- drilled. If the second well is drilled and it turns
ations of two points that originally were adjacent out that the compartments do have a uid con-
to one another at the base of the Brent Formation. nection because the faults are not continuous, or
Two faults with the same dip are displayed as do not extend as far as originally interpreted, the
stripes with lesser and greater widths (horizontal unnecessary expense can be signicant. Drilling
separations) corresponding to lesser and greater one additional well in these offshore settings can
dip separation. Two faults with the same dip sep- be a multi-million dollar proposition. Clearly the
aration are displayed as stripes with lesser and design of the production strategy is dependent
greater widths corresponding to greater and upon accurate interpretation of the fault geome-
lesser dips. In regions where the dips of all the try and sealing properties. Here we demonstrate
normal faults are about the same, the widths of how models of fault slip distributions developed
the stripes on the structure map are used to infer by structural geologists can be helpful in making
the relative magnitude of dip separations. such design decisions.
Similarly, if the dip does not vary along the length The two compartments of the Brent reservoir
of a fault, the changing width of the stripe is used in the Oseberg Syd Field are labeled in Fig. 1.10 as
to infer changes in the distribution of dip separa- Omega North and South. Both are bounded to the
tion. west and east by major faults, presumed to be
From information gathered by drilling and sealing. Data from well No. 8, which intersects the
recovering core samples, measuring a variety of Brent Formation in Omega North, and data from
indicators in the wells, and running ow and pres- well No. 10, which intersects the formation in
sure tests in the sub-surface, it is believed that the Omega South, suggest that there is a uid pres-
rocks in the fault zones shown in Fig. 1.10 are rel- sure difference between the northern and south-
atively impermeable to uid ow compared to the ern compartments. However, the interpreted fault
unfaulted portions of the sandstone reservoir. pattern does not dene a complete compartment.
Therefore these are referred to as sealing faults. A The two northwestsoutheast striking faults just
number of mechanisms for changing the perme- to the southwest of Well No. 8 do not form a com-
ability of rock in a fault zone have been identied plete barrier between Omega South and Omega
(Jones et al., 1998). For example, the crushing of North. Perhaps there are other explanations for
sandstone grains during shearing in the fault the pressure difference, but the hypothesis inves-
zone can ll the pores with fragments, thereby tigated here is that these two faults are actually
decreasing the porosity and the permeability linked and thereby separate the reservoir into two
(Antonellini and Aydin, 1994, 1995; Antonellini et compartments.
al., 1994). Also, it is possible for very ne-grained This hypothesis was tested using a mechanical
clays to be dragged or injected into the fault zone model that relates fault geometry, rock properties,
to form a seal (Aydin and Eyal, 2002). If the faults and tectonic loading to fault slip (Crider and
are all of the sealing type, then they can divide the Pollard, 1998; Maerten et al., 2000; Crider, 2001).
reservoir into a number of isolated compartments The modeling method subdivides faults into many
with little or no uid communications between small triangular elements, as seen for a non-planar
compartments. The map shown in Fig. 1.10 sug- fault with irregular tipline in Fig. 1.11, and these
gests that the Brent Formation could be divided elements approximate the three-dimensional
into many compartments separated by the sealing geometry. The model is based on elasticity theory
faults. and the numerical method used to solve the gov-
The presence of fault-sealed compartments erning equations is called the boundary element
has important implications for locating wells and method (Crouch and Stareld, 1983). Each element
closed contours distributed along the fault at

z
y Tipline
about mid-height. On a single fault that is isolated
from its neighbors one would expect a single
maximum in slip located more or less at the center
x
of the fault.
The geometry of faults A, B, and C were repre-
sented in the model as originally interpreted
from the seismic data. The slip distribution on
fault A (Fig. 1.12b) was computed by imposing
Fault Tipline boundary conditions on the model that are con-
surface
sistent with the overall deformation recorded by
the fault heaves across the entire Oseberg Syd
Element U SS
Field. The interpreted slip distribution on fault A
U DS
and the computed slip distribution on model
fault A are roughly similar, but only two of the
three maxima are seen in the model distribution.
Fig 1.11 Model of a normal fault taken from seismic data The fact that the intersection of faults A and C
and analyzed using the boundary element code Poly3D. In produced two maxima suggests that fault B
this illustration the fault has an irregular tipline and an extends to the southeast until it truncates against
irregular surface. The fault surface is divided into many small fault A. The computed slip distribution for this
triangular boundary elements. Inset: Dip slip, UDS, and strike
new model geometry (Fig. 1.12c) has three
slip, USS, are constant on an element but vary from element
to element to model the slip distribution on the fault.
maxima, one on each side of the two lines of inter-
Reprinted from Maerten et al. (2002) with permission from section of faults B and C with fault A. The model
Elsevier. slip distribution and the interpreted slip distribu-
tion are not identical, but the major features are
remarkably similar. This correspondence suggests
may be assigned a strike slip, USS, and a dip slip, that it would be well worth the effort to look
UDS, component (Fig. 1.11, inset) based on the avail- again at the seismic data to determine if the
able data, or the magnitude and direction of the linkage of fault B with fault A is permitted by the
shear stress drop during slip may be assigned. data. Fault B could extend and link, but have slip
Given these local boundary conditions and the that is below the resolution of the data. Or, the
state of stress or strain in the region before slip on seismic interpreter could have overlooked the
the faults, the numerical method solves the elastic linkage. It would also be worth checking to see if
boundary value problem for the displacement, the slip distribution on fault B is suggestive of
strain, and stress elds in the region surrounding linkage with fault A.
the fault. The mechanical models described here provide
The basis for the investigation was the three- encouraging results for the further evaluation of
dimensional seismic data set used for the original the geometry of sealing faults. In addition, as
interpretation of the faults and the sedimentary small-scale opening fractures, and their counter-
horizons in the Oseberg Syd Field. From this inter- part compaction bands, propagate through reser-
pretation the geometries of faults A, B, and C were voirs they can have a signicant effect on bulk
dened and visualized (Fig. 1.12). These faults are permeability (Taylor et al., 1999; Aydin, 2000;
not planar structures, nor are their tiplines simple Taylor and Pollard, 2000). Working out the rela-
elliptical shapes. The slip on model fault A, as tionships among the faults that can be imaged
interpreted using the offset of reecting horizons using seismic techniques and these sub-seismic
on the seismic data, was examined and contoured fractures is a challenging task. One can easily
to produce the slip distribution in Fig. 1.12a. This imagine how structural geologists and geophysi-
slip distribution is somewhat unusual in that it cists, working together with high-quality three-
has three distinct maxima, each dened by a set of dimensional seismic data, could improve the
Interpreted slip from 3D seismic data

(a) N
1 km NW SE
B C
0 100 m
1 km
Slip
(b) Computed slip from model 1
C
B
(c) Computed slip from model 2
C
B
Fig 1.12 Fault maps and slip distributions on fault A

between Omega North and Omega South in the Oseberg 1.4 Anticracks in southern France
Syd Field in the northern North Sea. (a) Slip as determined
by interpretation of seismic reflection data. (b) Slip from a Mapping, describing, and analyzing geologic
boundary element model that includes faults A, B, and C as
structures can help one appreciate new and beau-
interpreted. (c) Slip for a boundary element model that
tiful facets of the natural world. One of the great
extends fault B to link with fault A. See website for color
image. Reprinted from Maerten et al. (2002) with permission scientists of the early twentieth century, Henri
from Elsevier. Poincar, said it this way:
The scientist does not study nature because it is useful to
do so. He studies it because he takes pleasure in it; and
interpretation of fault and fracture geometries
he takes pleasure in it because it is beautiful. If nature
and provide input for a more effective production
were not beautiful, it would not be worth knowing and
strategy. Using models to clarify the geometry of life would not be worth living . . . It is because simplicity
faults in hydrocarbon reservoirs is a timely and vastness are both beautiful that we seek by prefer-
example of an application of modern structural ence simple facts and vast facts; that we take delight,
geology to a problem of economic importance to now in following the giant courses of the stars, now in
society. scrutinizing with a microscope that prodigious small-
1.4 ANTICRACKS IN SOUTHERN FRANCE 17
oriented parallel to the bottom of the photograph.

Parts of these faults are highlighted in the photo-
graph because they contain a thin layer of white
Vein calcite between the two surfaces of the faults.
Solution
surface Near the left termination of the middle fault two
thin white structures extend toward the lower
Fault left-hand corner of the photograph. They have
tapered shapes, being thickest at the fault and
Solution thinning with distance from the fault to zero at
Vein Fault surface
their distal terminations. These structures are
5 cm interpreted as veins that broke open the limestone
and propagated away from the fault. Each vein
Fig 1.13 Outcrop photograph from Les Matelles in lled with groundwater from the surrounding
southern France showing three small faults with traces rock mass because opening reduced the local uid
parallel to the top of the photograph. Each fault has an pressure, and the mineral calcite was precipitated
antisymmetric distribution of veins (filled with white calcite)
from this solution. Note that a few veins also
and solution surfaces (dark wavy bands). See website for
trend toward the upper right-hand corner of the
color image. Photograph by J.-P. Petit. Reprinted from Petit
and Mattauer (1995) with permission from Elsevier.
photograph from near the right-hand termina-
tion of this small fault.
Very different looking structures extend from
ness which is also a vastness, and now in seeking in geo- near the left-hand termination of the fault toward
logical ages the traces of the past that attracts us the upper left-hand corner of the photograph.
because of its remoteness (Chandrasekhar, 1979).
Instead of smoothly tapered cracks with white
Once bitten by the bug of structural geology llings, these structures are less regular and are
it is impossible to walk up to an outcrop display- marked by dark blotches. The dark material is
ing a particularly ornate structure or complex made up of insoluble minerals, probably clays
structural relationship (for example, as pictured that are found dispersed throughout the lime-
in Fig. 1.13) and not exclaim: WOW, look at this! stone in minor quantities. Here they are concen-
Such was the reaction of the authors of this text- trated within what is called a solution seam,
book to the outcrops of Jurassic limestone composed of two surfaces of limestone on either
exposed at Les Matelles in the Languedoc region side of the insoluble material. As the name
of southern France (Petit and Mattauer, 1995). implies this structure is interpreted as forming
Actually, our rst glimpse of these structures where the limestone has dissolved and the solu-
came by way of an article in the journal ble components (calcium carbonate) have been
Tectonophysics (Rispoli, 1981). We were working transported away in the groundwater, either by
at the US Geological Survey in Menlo Park, diffusion within stagnant groundwater or by
California, when the article appeared and both of groundwater ow, leaving the insoluble minerals
us were immediately captivated by the systematic (Rutter, 1983; Mardon, 1988).
geometrical relationships among the structures
illustrated on the published maps. Despite the 1.4.2 Conceptual and mechanical models
fact that we were working on quite different prob- for veins and solution surfaces
lems for the Earthquake Studies Branch of the near a fault
Survey at the time, we resolved to seek an expla- The combination of veins, solution surfaces, and
nation for the systematic relationships among fault is depicted in the conceptual model shown
these small structures. schematically in Fig. 1.14. We understand that vein
surfaces move away from one another as the vein
1.4.1 Cracks and anticracks opens. On the other hand, we can think of the solu-
The structures pictured in Fig. 1.13 include three tion surfaces as closing or anticracks. As limestone
small faults, each about 20 cm in trace length and dissolved at the surfaces and was transported away
60
y 2
Quadrant 2 Quadrant 1 40
compression tension
Rhomb cavity 1
20
Left step { x
0 0
y
Vein Solution
Fault segment surface -20
-1
Quadrant 3 Quadrant 4 -40
tension compression
-2 -60
-2 -1 0 1 2
Fig 1.14 Schematic illustration of a left-lateral fault with x
veins and solution surfaces emanating from the fault tips in an
Fig 1.15 Contour map of the normal stress acting parallel
antisymmetric pattern (Fletcher and Pollard, 1981). Pairs of
to a fault modeled as a surface of displacement discontinuity
arrows indicate stretching or shortening of the rock
with uniform left-lateral slip (Crouch and Starfield, 1983).
associated with slip on the fault. This deformation is
Note the pattern of tensile stress (positive) in the first and
accommodated by the formation of the veins and solution
third quadrants, and compressive stress (negative) in the
surfaces, respectively.
second and fourth quadrants. Also note that the stress
magnitudes increase toward the model fault tips. See website
in the groundwater, the material to either side of for color image.
the solution surface moved inward. Note how the
veins and solution surfaces are distributed into
the four quadrants dened by a coordinate system Laboratory investigations have shown that the
centered on the fault with the x-axis parallel to the mechanical response of rock to stress is, in part,
trace of the fault. The conceptual model consists of dependent upon the magnitude of the normal
a system of cracks and anticracks distributed stress. For example, modest positive values of the
about the small fault in what is called an anti- normal stress (on the order of 1 to 10 MPa) corre-
symmetric conguration. The presence of veins late with extension of the rock mass and the devel-
and solution surfaces with this particular anti- opment of opening cracks. As the normal stress
symmetric arrangement is diagnostic of left-lateral becomes more compressive, pressure solution
slip on the fault. If the veins are in the second and may become an active deformation mechanism in
fourth quadrants, and the solution surfaces are in rocks with soluble components. Although the
the rst and third quadrants, the arrangement is exact magnitudes of the tensile and compressive
diagnostic of right-lateral slip. stresses necessary to induce these structures is
Our conceptual model for the relative motion not well known, it is clear that tension is neces-
of the surfaces of the cracks and anticracks (Fig. sary to induce opening cracks and compression is
1.14) suggests that the limestone in the rst and necessary to induce solution surfaces.
third quadrants extended parallel to the fault and The mechanical model we employ consists of a
the limestone in the second and fourth quadrants single two-dimensional fault (Fig. 1.15) with left-
contracted parallel to the fault. How do these lateral relative motion (Pollard and Segall, 1987).
deformations relate to the state of stress? Why do The normal stress acting on planes perpendicular
these structures initiate near the fault termina- to the model fault is calculated on a grid of points,
tions and propagate outward into these quad- and these values are contoured to produce the
rants? Why do the structures stop propagating at gure. The contours appear symmetric about the
a short distance from the fault? These and other model fault, but note that the stress is, in fact, dis-
questions can be addressed with a mechanical continuous across the fault surfaces, having the
model for the state of stress near a fault. same magnitude but opposite sign for adjacent
1.4 ANTICRACKS IN SOUTHERN FRANCE 19
points on the two surfaces. The stress, as indicated near the terminations of the faults. This is the
by the values associated with each contour, is posi- region of greatest stress concentration, and there-
tive (tensile) in the rst and third quadrants and fore is the locality where secondary structures are
negative (compressive) in the second and fourth most likely to form. Finally, the stress distribution
quadrants. Furthermore there is a stress concentra- offers an explanation for the limited extent of the
tion at the terminations of the model fault: the secondary structures. The stress decreases away
values associated with the contours increase from the fault tips toward much lower values at a
toward the fault tip and the spacing between con- distance that scales with the length of the fault.
tours decreases. In fact the stress becomes so great This decrease in stress is consistent with the ter-
very near the model fault tips that the contours mination of the veins and solution surfaces at
merge into a pattern that is no longer distinguish- modest distances from the fault tips. Although
able at the scale of this gure. Thus we have omitted this two-dimensional model provides important
the contouring in a small region around each tip. insights, additional understanding of the faulting
Figure 1.15 is an example of a plot prepared process may be achieved using three-dimensional
using MATLAB, the computational and graphics models (Willemse et al., 1996; Willemse, 1997;
engine that we employ throughout this textbook. Martel and Boger, 1998).
The m-file used to compute the values of the stress The authors became intrigued by the struc-
component at the grid points and to prepare the tures at Les Matelles over twenty years ago and, in
contour plot is available at the textbook website. the course of investigating models for their for-
There, a color version of the contour plot is view- mation, conceived of the concept of anticracks. At
able along with contour plots of other stress com- that time we had no practical applications for this
ponents. This procedure is followed throughout concept in mind. Nor, to our knowledge, did the
the textbook where grayscale gures are used to exposure at Les Matelles gure signicantly in the
reduce printing costs and color versions are avail- solution of any problem relevant to society. For us
able at the website. this was an academic exercise, motivated by a
To relate the veins and solution surfaces at the strong (and inexplicable) urge to understand these
Les Matelles outcrop to the left-lateral faults we structures, and nothing more. However, explana-
have to recognize that the veins and solution sur- tions for certain features of very deep and large-
faces are secondary structures and the faults are the magnitude earthquakes now utilize the concept of
primary structures. In other words, the veins and anticracks (Green and Burnley, 1989). Also, the evo-
solution surfaces formed in response to the stress lution of fault zones in limestone through a
changes in the rock mass as slip developed on the complex sequence of vein development, solution
faults. We can correlate the symmetry of these sec- surface development, and slip on solution surfaces
ondary structures with the symmetry of the stress has been documented and interpreted using the
eld about the model fault (Fig. 1.15). The veins are anticrack concept (Willemse et al., 1996). In addi-
cracks that are pulled open by tensile stresses and tion, the development of compaction bands in
therefore are associated with the eld of tensile porous sandstone has been explained using the
stress in quadrants 1 and 3. Conversely, the solu- anticrack concept (Mollema and Antonellini,
tion surfaces form in response to elevated com- 1996). These tabular zones of localized compaction
pressive stresses and therefore are associated with appear to propagate as anticracks in response to
the eld of compressive stress in quadrants 2 and 4. elevated compression, and they also have a
The correlation between the model and the signicant effect on the permeability of reservoirs
exposure observations is supported by the fact and aquifers in porous sandstone (Sternlof et al.,
that the veins and solution surfaces do not cross 2004). Perhaps the time and the taxpayers dollars
the fault surfaces where the model indicates a spent working on the seemingly arcane concept of
discontinuity (change in sign) of the stress. anticracks at the Earthquake Studies Branch of the
Furthermore, the model provides an explanation US Geological Survey can be justied in light of
for the initiation of these secondary structures these applications.
1.5 Mountain building on the

Colorado Plateau
We are geologists because we love beautiful mineral
specimens or ne fossils or magnicent mountains
(Woodford, 1956).
MOUNT HILLERS
Many structural geologists would argue that the
premier research topic in this discipline is moun-
tain building. Mountain building produces many
of the most dramatic landscapes on our planet T. Diorite Porphyry MOUNT HOLMES
and provides countless opportunities to escape
K. Mancos Group
from urban settings into the wilderness for dif-
ferent forms of recreation. Mountains profoundly
J. Morrison Fmn
affect local climates and provide challenges to
builders of roads and dams. Few can say that their J. San Rafael
lives are not affected in some way by mountains.
J.-Tr. Glen
The obvious question is: why are they there? No Canyon Grp MOUNT
single answer is credible and geologists have Tr. Chinle, ELLSWORTH
identied several tectonic processes that lead to Moenkopi Fms 37o 45'
mountain building. Here we describe one such P. Cutler Fm Utah
process because it is rather simple to understand
and because it informs us about an interesting HENRY MOUNTAINS
chapter in the development of structural geology 0 1 2 3 4 5 km
in the latter half of the nineteenth century.
110o 45' 110o
Our story begins when John Wesley Powell
stopped briey on one of his harrowing boat trips Fig 1.16 Simplified geological map of the southern Henry
down the Colorado River (186971) and climbed Mountains. Mt. Holmes, Mt. Ellsworth, and Mt. Hillers are
up out of the deep canyon to look around. Across structural domes. T, Tertiary; K, Cretaceous; J, Jurassic; Tr,
the plateau country of southern Utah Powell saw Triassic; P, Permian. Reprinted from Jackson and Pollard
several mountain peaks, capped with massive (1988) with permission from The Geological Society of
America.
gray rock that clearly was not sedimentary. One of
the nearby peaks, later named Mt. Hillers, is
shown on the frontispiece for this chapter. Powell
gave the range a name, the Henry Mountains, in layer-cake geology are three mountains, Mt.
honor of the distinguished physicist Joseph Holmes, Mt. Ellsworth, and Mt. Hillers, each asso-
Henry. In hindsight it is tting that this range, ciated with igneous rocks (black on the map). Note
which would reveal to others one of the funda- how the strata circle Mt. Hillers, with older sedi-
mental physical processes of mountain building, mentary units exposed toward the center of the
should be named for a physicist. mountain. This is a clear indication that the
Figure 1.16 is a geological map of the region mountain is a structural dome: the strata have been
Powell could see from his perch on the side of the elevated over the center of the mountain relative
nearby canyon. The Mesozoic strata of this region to the anks.
range in age from the Permian Cutler Formation Powell wondered about the origin of these
through the Upper Cretaceous Mancos Group. spectacular mountains, but his primary mission
Throughout much of the nearby plateau they are was exploration of the river so he did little more
inclined just a few degrees to the west, so pro- than gaze at them from a distance. A few years
gressively younger rocks crop out from east to later, as Chief of the Geological Survey in
west across this map. Disturbing this simple Washington, DC, Powell sent a young geologist,
1.5 MOUNTAIN BUILDING ON THE COLORADO PLATEAU 21
Grove Karl Gilbert, to investigate the Wasatch (a)

4
Plateau, just west of the Henry Mountains, and to
carry out a topographic and geologic survey of the 1 4
3 2 1
region. Gilberts career and his many contribu-
tions to geology are described in the book Grove
Karl Gilbert: A Great Engine of Research (Pyne, 1980).
Gilbert worked from horseback covering an area N 35o W Sea level
from Salina, Utah, to the western edge of the
Henry Mountains from late June to mid August of (b) 4
1875 (Gilbert, 1877). The chronicle of this trip is
clearly laid out in his eld notebooks and these 1 43 2 1
have been reproduced and annotated by Charles
Hunt (Hunt, 1988a).
The frontispiece from Gilberts Report on the
Geology of the Henry Mountains, published in 1877, is
reproduced as an inset on the frontispiece for this N 35o W Sea level
chapter. This drawing illustrates the form of the
Fig 1.17 Cross sections of Mt. Hillers (Gilbert, 1877, Figs.
displaced (and now eroded) sedimentary strata
25, 26). (a) Only the exposed formations are depicted with
forming the structural dome that is M. Ellsworth.
appropriate inclinations. (b) Idealized representation of the
The base of this diagram is taken at modern sea sub-surface structure showing the bottom of the laccolith
level and the strata are broadly divided into four and the bending of strata onto the flanks of the dome.
layers: the upper Paleozoic at the bottom, fol-
lowed by the Triassic and Jurassic, then the
Cretaceous, and nally the Tertiary at the top. The 1.5.1 Conceptual and mechanical models
near half of the diagram displays the current for laccolith formation
topography of Mt. Ellsworth and the remote half Although exposures in the three southern Henry
shows the form of the strata in the structural Mountains do not reveal the bottoms of the larger
dome. In the preface to this report Gilbert com- masses of igneous rock or the conduits that fed
ments as follows: magma into them (Fig. 1.17a), some of the smaller
bodies on the anks of these mountains have well-
Two months would be far too short a period in which
to survey a thousand square miles in Pennsylvania or exposed oors. Based in part on analogy to the
Illinois, but among the Colorado Plateaus it proved smaller intrusions, Gilbert concluded that the
sufcient. A few comprehensive views from mountain uplifted strata of the major domes formed as
tops gave the general distribution of the formations, magma owed upward from an unknown source,
and the remainder of the time was spent in the exami- perhaps in dikes, and then spread laterally
nation of the localities which best displayed the pecu- between two strata as thin sills. Some of these sills
liar features of the structure. So thorough was the continued to thicken by pushing up the overlying
display and so satisfactory the examination, that in strata into structural domes (Fig. 1.17b). The
preparing my report I have felt less than ever before resulting structure was named a laccolite by
the desire to revisit the eld and prove my conclusions
Gilbert and is now called a laccolith. The progres-
by more extended observation (Gilbert, 1877, p. vii).
sion from feeder dike to sill to laccolith, and the
Those peculiar features of the structure revealed concomitant formation of a structural dome, was
the two major processes that shaped the Henry the new conceptual model for mountain building
Mountains, upward displacement of the strata and offered by Gilbert in his 1877 report (Gilbert,
the progress of erosion that carved the mountains 1877).
as we know them today. As structural geologists it Gilbert wondered if there was a systematic
is the rst of these processes that captures our relationship between the depth of emplacement
attention, while the geomorphologist would nat- of a laccolith and its diameter. One might imagine
urally focus on the second. that magma, insinuating between horizontal
strata, would nd it easier to spread laterally at 2a

shallower depths, because there is less overbur-
den. If this conjecture were correct, one would
expect to nd laccoliths with lesser diameters Cylindrical
exposed lower in the stratigraphic sequence. On fault
the other hand one might imagine that magma S S
Rock piston
would have to spread farther at greater depth to d
gain the leverage necessary to push the overbur-
den upward. If this conjecture were correct, one
would expect to nd laccoliths with greater diam-
eters exposed lower in the stratigraphic sequence. Magma Pm Pw
Both conjectures cannot be correct.
To evaluate these conjectures and thereby
clarify the mechanics of laccolith formation Fig 1.18 Gilberts piston-cylinder mechanical model for
Gilbert idealized the laccolith as a cylindrical laccolith formation: 2a, piston diameter; d, depth to laccolith
bottom; S, shear strength of cylindrical fault; Pm Pw, driving
chamber of magma pushing upward on a rigid
pressure.
piston representing the overlying strata (Fig. 1.18).
The mechanical model considers the piston of
overburden alone and ignores the dynamics of base of the piston (Fig. 1.18) is evaluated as the
the magma and the surrounding rock. The magma pressure, Pm, times the surface area of the
boundary of the piston is a cylindrical and verti- piston, a2. Recall that pressure is a force per unit
cal fault. The relevant geometric parameters are area, so this force is simply the pressure times the
the diameter of the piston, 2a, and the depth of surface area. The magnitude of the downward
overburden, d, above the base of the laccolith. directed force along the fault is given by the shear
This depth is equal to the height of the rock strength, S, times the area of the cylindrical fault,
piston and the surrounding fault. Note that the 2ad. The shear strength is dened as the shear
slip on this fault would be equal to the thickness stress acting on the fault just before slip and shear
of the laccolith. The cylindrical fault is an ideal- stress is the force per unit area. Finally, the mag-
ization of what is observed to be a exure of the nitude of the downward directed force due to
strata. We describe the procedure of idealization gravity is given by the pressure, Pw, due to the
in the development of mechanical models in the weight of the piston times the surface area of the
last chapter of this book. piston, a2. In these expressions upward directed
To derive a mechanical relationship between forces are taken as positive.
the diameter and depth Gilbert considered the Newtons Second Law, for the case of zero
forces acting on the piston and sought a relation- acceleration, requires the net force acting on the
ship among these forces based on Newtons Laws piston to be zero. Summing the forces identied
of Motion (Newton, 1687), in particular Newtons in the previous paragraph and setting this sum to
Second Law, F ma, where F is the net force, m is zero we have the equilibrium equation:
the mass, and a is the linear acceleration. He
(Pm Pw)a2 2Sad 0 (1.1)
simplied this equation by considering the accel-
eration to be zero. Of course the piston must accel- The rst term in this equation is the driving force
erate as slip develops on the fault, but Gilbert for upward motion of the piston. Clearly the
chose to consider the moment just before slip magma pressure must exceed the lithostatic pres-
begins, when the upward force due to the magma sure for uplift. The second term is the resisting
pressure is just sufcient to balance the weight of force due to the shear strength of rock along the
the overburden and the shear force resisting slip fault. Note that the driving force increases as the
on the fault. In other words the system is in a state square of the piston radius, whereas the resisting
of mechanical equilibrium. force increases only in proportion to the radius.
The magnitude of the upward force on the For small radii there may be insufcient driving
1.5 MOUNTAIN BUILDING ON THE COLORADO PLATEAU 23
Table 1.1. Gilberts field data on laccolite diameter.
Zone Formations Laccolite name Diameter (miles)
Upper Blue Gate Shale Sentinel 0.7

Tununk Shale Geikie 0.8
A 0.9
Marvin 1.0
Jukes 1.4
Peale 1.8
Flaming Gorge Shale Steward 1.0
B 1.1
Newberry 1.8
C 1.9
Lower Dana 2.0
Greater Holmes 2.1
Lesser Holmes 2.1
Ellsworth 2.3
Pulpit 2.3
Maze 2.8
Crescent 3.6
Hillers 3.9
force to overcome the resistance to faulting. This correct according to this model for laccolith for-
does not preclude the lateral growth of a sill, but mation.
does preclude the development of the laccolith. The conceptual and mechanical models for-
From this relationship Gilbert inferred that short mulated by Gilbert for laccolith formation pro-
sills would be incapable of producing a structural vided him with a linear relationship between the
dome of the kind observed in the Henry diameter of laccoliths and their depth of burial.
Mountains. On the other hand if the magma were He devoted much of his time in the eld to gath-
able to spread far enough laterally as a sill, the ering data (Table 1.1) on the horizontal dimen-
driving force would equal the resisting force and, sions of laccoliths and their stratigraphic
in the next increment of growth, the fault would positions in order to test this relationship
develop and the overburden would begin to dis- (Gilbert, 1877, p. 86). Although the diameters
place upward to form a laccolith. given in this table are uncertain because of incom-
Solving the equilibrium equation for the plete exposure of the larger laccoliths in the
radius of the piston, Gilbert found: Henry Mountains, Gilbert came to the conclusion
that these data were consistent with the relation-
2dS
a (1.2) ship he had derived. None of the diameters of lac-
P m Pw
coliths from the upper zone exceed the diameters
He noted the linear relationship between piston of those from the lower zone. For its simplicity
radius, a, and depth of overburden, d, and inferred and the insight gained from it, this model is
that, for a given (positive) driving pressure, Pm remarkably successful.
Pw, and fault shear strength, S, magma must Many questions about laccolith formation
spread to a greater radius at a greater depth in cannot be addressed with Gilberts model. What
order for the overburden to be pushed upward are the feeder conduits? How does the sill propa-
along the fault. The second conjecture made gate laterally from this feeder to a diameter where
above (laccoliths with greater diameters should be the laccolith can form? At what rate did the
found lower in the stratigraphic sequence) is magma ow in these conduits? How might one
account for the resistance to bending of the strata tonic processes and their products over relevant
over the laccolith? How did the temperature vary length and time scales by combining observational
as the hot magma invaded the cold sedimentary data with both conceptual and mechanical
strata? We develop other models in later chapters models. For each of the examples given we
to address some of these questions. described the phenomenon and then offered a
Because the bottom contacts of the larger simple conceptual model that illustrates the key
igneous intrusions are not exposed, other con- elements of the physical process. Then we
ceptual models for the development of the struc- described a mechanical model used to understand
tural domes could be viable. Based on geological the physical process. A second objective for this
mapping of the Henry Mountain range and sur- chapter was to illustrate reductionism, a scientic
rounding plateau country, Hunt (1953) proposed methodology in which complex processes are
that each of the major mountain peaks is under- broken up into their (relative few) fundamental
lain by a cylindrical stock of igneous rock that elements in order to understand each element in
extends well below the bottom contacts that were isolation. Then, the elements are put back together
inferred by Gilbert. This hypothesis has been chal- to understand how the whole system operates. This
lenged, based on more recent geological mapping methodology is espoused throughout the book.
in the southern Henry Mountains and cross sec- Some of the topics in the introduction satisfy
tions that were constructed from new structural our objective of demonstrating the possible roles
data (Jackson and Pollard, 1988). The two hypothe- a structural geologist can take in society. For
ses for the shape of the magma chamber, stock or example, career opportunities may be available
laccolith, were debated in the literature and a undertaking geological hazard assessment with
variety of geological and geophysical evidence the US Geological Survey (USGS), being a member
brought to bear on the subject (Hunt, 1988b). of a production analysis team for an international
oil company, or participating in planetary explo-
ration with the National Aeronautics and Space
1.6 Concluding remarks Administration (NASA). We hope that these exam-
ples motivate students to learn more about the
One of our objectives for this chapter was to intro- fundamentals of structural geology by reading
duce a strategy for comprehending complex tec- this textbook.
Chapter 2
Structural mapping techniques and tools
Structural geologists use a GPS receiver to determine the formed. Thus order in the sky became chaos. Random.
UTM coordinates of the point under their feet on this But even then . . . Even then, what Coyote did was evil,
outcrop of Aztec sandstone in the Valley of Fire, NV. but was there not a pattern, too, in the evil deed?
Photograph by D. D. Pollard. That had not been the time in Leaphorns life when
he had patience for the old metaphysics. He remem-
bered telling Haskie Jim about modern astronomy and
You think these raindrops are random? his uncle had the cosmic mechanics of gravity and velocity.
asked. And Leaphorn had been surprised. Hed said of Leaphorn had said something like Even so, you
course they were random. Didnt his uncle think they couldnt expect to nd anything except randomness
were random? in the way the rain fell. And Haskie Jim had watched
The stars, Haskie Jim said. We have a legend about the rain awhile, silently. And then he had said, and Joe
how First Man and First Woman, over by Huerfano Leaphorn still remembered not just the words but the
Mesa, had the stars in their blanket and were placing old mans face when he said them: I think from where
them carefully in the sky. And then Coyote grabbed we stand the rain seems random. If we could stand
the blanket and whirled it around and ung them into somewhere else, we would see the order in it.
the darkness and that is how the Milky Way was (Hillerman, 1990, pp. 21314.)
26 STRUCTURAL MAPPING TECHNIQUES AND TOOLS
U
nlike the laboratory-bound physicist, vectors are dened with reference to a particular
structural geologists usually make mea- coordinate system. UTM coordinates may not be
surements and gather data in the eld, the best choice during a mapping campaign, and
where, because of poor exposures and the limited usually they are not the most convenient for mod-
view that erosion provides, structures may seem eling and data analysis. To change from one coor-
chaotically arranged. Yet, when each measure- dinate system to another one uses transformation
ment is tied to a particular geographic location equations, which make use of vector concepts and
and represented on a map, and when sets of mea- operations.
surements are organized by orientation and Because much of the structural information
plotted on an appropriate graph, systematic rela- recorded at exposures can be reduced to the ori-
tionships and patterns emerge that can be inter- entations in space of planar or linear elements, we
preted. In making such interpretations we have introduce the techniques to measure, record, and
achieved what Haskie Jim was alluding to, we have analyze these orientations. Special projections are
found where to stand to view the structures to used to depict a set of orientations on a at piece
bring order to what otherwise appears to be a of paper a common one being the stereographic
random phenomenon. projection. Here, the basic procedures for plotting
In pursuit of this objective we begin this orientation data on a stereogram are described in
chapter by nding our bearings in terms of the such a way that they can be implemented on a
coordinate systems commonly employed to locate computer. Combinations of structural maps and
outcrops in the eld and to construct maps of stereograms are used to visualize, respectively, the
geological structures. This includes an introduc- spatial distributions and the orientations of geo-
tion to the geographic coordinate systems used logical structures. Examples are provided to illus-
for the Global Positioning System (GPS), which is trate how this can be done.
becoming the standard tool for structural Finally we describe a modern mapping cam-
mapping (see frontispiece to Chapter 2). Also, we paign that utilizes GPS technology to create a
introduce the Universal Transverse Mercator precise structure contour map in a region where
(UTM) projection, used to project geographic four sets of faults intersect to disrupt and fold the
information from the curved surface of the Earth surrounding strata. With new technology the
onto a at piece of paper. Structural geologists mapping was accomplished in a fraction of the
should be capable of measuring the location of time required using traditional methods, with
key exposures where eld data are taken, and much greater control on the shapes of the
then constructing the map projections that deformed strata and on the fault offsets and slip
record the spatial distributions of these data. directions because of the abundance of quantita-
Structural maps should convey the geometry of tive data. The GPS technology is rapidly evolving
structures to other geologists in a form that is so we anticipate a new generation of structural
quantitatively precise and readily visualized and maps that will provide the impetus for more
analyzed. specic modeling of tectonic processes and the
A position vector uniquely determines the data better to constrain those models.
location of every exposure and every point at
which structural data are collected in the eld.
Vectors have many uses in structural geology, 2.1 Geographic coordinates and
from locating an exposure, to describing the
shape of folded strata, to representing quantities
map projections
such as displacement and velocity in the physical
laws that underlie the modeling of tectonic 2.1.1 Geographic coordinates: the
processes. The position vector is introduced here Global Positioning System (GPS)
along with some of the vector concepts and nota- The rst order of business for any mapping project
tions useful for structural mapping. Position is to locate oneself on Earths surface, either to
2.1 GEOGRAPHIC COORDINATES AND MAP PROJECTIONS 27
identify the positions of outcrops or sampling (a) Latitude N

localities, or simply to avoid getting lost! For
mapping or sampling on the Earths surface it is
common to use a geographic coordinate system
based on lines of constant latitude and longitude
(Fig. 2.1a). Lines of constant latitude trend from
east to west and are numbered in degrees, 0 to 90, Equator
W E
both north and south from the equator to the
poles. Lines of constant longitude trend from
north to south and pass through the poles. They
are numbered in degrees, 0 to 180, both east and
west from Greenwich, UK, so there are 360 in
total. To be more precise in locating position on Longitude
S
the Earths surface, each degree is subdivided into
sixty minutes (1 60) and each minute is sub- (b)
divided into sixty seconds (1 60). 400 38o 50 N
The scheme of subdividing circles, such as the
lines of latitude or longitude, into 360 sectors Q
dates back at least to the ancient Babylonians 300
(Beckmann, 1971), who used a base 60 numbering
200
system. Imagine having to recall sixty different
names for your basic counting numbers! The 38o 30 N
Babylonians knew that the perimeter of a hexagon 100 m
is exactly equal to six times the radius, R, of the cir-
cumscribed circle (Fig. 2.1c). Apparently thinking
123o 40 W 123o 20 W
that each equilateral triangle with side length R
denes a unit sector of the circle, they divided
(c) Circle, circumference = 2pR
each of the six sectors of the circumscribed circle
into 60 sub-sectors, based upon their numbering
system, and this results in 360 sub-sectors for the R
entire circle. This system dees the simple power
of ten relationships of a metric system, but we are R
stuck with it. 60o
We indicate latitude and longitude on the
edges of some of the maps used in this text when
they cover sufcient area to warrant this coordi-
nate system (Fig. 2.1b). Note how the lines of con-
stant longitude in this example are designated to
Hexagon,
be west (W) of Greenwich and the lines of constant
perimeter = 6R
latitude are designated to be north (N) of the
equator. Distance above or below the chosen Fig 2.1 Geographic coordinates consisting of latitude,
datum usually is indicated on such a map using longitude, and elevation. (a) Lines of constant latitude and
topographic contours, lines of equal elevation, to longitude. (b) Topographic map with a datum of mean sea
produce a topographic map. Here the contour level and contour interval of 100m. (c) Hexagon with
interval is 100 m and the datum is mean sea level. circumscribed circle used by Babylonians to divide the circle
The two geographic coordinates, elevation, and into 360 sectors.
datum completely dene the location of the
outcrop at the point Q:
Latitude: 38 45 N (a) Rotation axis GRS-80

Longitude: 123 42 W ellipsoid
Elevation (mean sea level datum): 425 m
Note that the precision of the location given here
is only to the nearest degree and the elevation is g Geoid
extrapolated from contours with 100 m intervals Down
Semi-minor axis
(not very precise). Center
The convention in geodesy is to use a mathe-
matical model for the Earths shape that is an
oblate ellipsoid: a three-dimensional surface b
formed by rotating an ellipse about its minor axis.
If the semi-major and semi-minor axes of the
ellipse are called a and b, respectively, the com-
pression or attening of the ellipsoid at the poles
a
is dened as f (a b)/a. The rotation of the Earth
does cause some extension of the equatorial Semi-major axis
radius and some attening at the poles so a b,
and the semi-major axis is the radius of the equa- (b) WGS-84
torial circle of the ellipsoid. A standard model ellipsoid datum
used by geodesists is called the Geodetic Reference
System 1980 ellipsoid, or GRS-80 for short 3.6 m
(Hofmann-Wellenhof et al., 1997, p. 293). The Earth's
center of GRS-80 is located at the true center of the surface
Earth and the semi-minor axis is parallel to the
Earths axis of rotation (Fig. 2.2a). The length of +28.9 m
the semi-minor axis is b 6 356 752.3 m, and the
length of the semi-major axis, extending from the
center of the Earth to the equator, is a
6 378 137.0 m. Note that this ellipsoid is attened
only by 21 384.7 m (about 21 km) so f 0.003 352 8. Sea level
In other words the ellipsoid is only about 0.3% dif- Geoid datum
ferent than a perfect sphere. An early ellipsoidal
Different elevations of the
model for Earth was calculated in 1830 by Everest,
geology building at Stanford University
working in India, where he estimated a
6 377 276 m and b 6 356 075 m, so the attening Fig 2.2 Cross section of Earth. (a) Mathematical model
is f 0.003 324 4 (Richardus and Adler, 1972, p. 23). ellipsoid and physically defined geoid. Local gravitational
Different ellipsoids are used in different regions acceleration vector, g, defines down. (b) Elevations of
because they minimize discrepancies with the Geology Building at Stanford University relative to different
local mean sea level. datums.
To put the degreeminutesecond measures of
distance in perspective we use the semi-major axis
These odd values and the awkwardness of con-
of GRS-80 as the radius, R, of the equatorial circle,
verting degreesminutesseconds to standard dis-
and calculate the circumference as C 2R
tance units make this geographic coordinate
40 075 016.7 m. The following relationships are
system less than ideal, but it is conventional to use
found for distances along this equator:
it and conventions of this long standing (perhaps
1 degree 111 319.5 m three to four thousand years) are difcult to
1 minute 1 855.3 m abandon.
1 second 30.9 m
The third coordinate in the geographic system System datum for 1984 (WGS-84) and this is almost
is approximately parallel to a radial line from identical to the ellipsoid called GRS-80. However,
Earths center and is referred to as elevation. this ellipsoid may be separated from the mean sea
Often, elevation is referenced to a sea level datum, level geoid by a considerable distance. For
that is mean sea level as determined by tidal example, the WGS-84 ellipsoid is about 32.5 m
gauges. This is not an unreasonable choice, but it above the mean sea level geoid at Stanford
is subject to perturbations caused by water tem- University in central California (Fig. 2.2b). Thus,
perature variations and currents, and it is limited the reported elevation relative to the WGS-84
in applications to the coastline. So-called leveling datum for a location just outside the Geology
surveys are performed to determine elevations Building at Stanford University using a high-pre-
away from the coast as heights above (or below) cision GPS receiver is 3.6 m, but the building
another datum called the geoid. The geoid is a clearly is not below sea level! The elevation rela-
physically dened surface (Fig. 2.2a) that is every- tive to the mean sea level geoid is about 28.9 m.
where perpendicular to the local direction of the This difference is not due to errors in measure-
acceleration of gravity, g, so it is everywhere per- ment, but rather to use of the different datums.
pendicular to the direction we call down from A metric coordinate system for location on the
observations of falling objects. Leveling instru- Earths surface is becoming popular because of
ments (for example a bubble level) are capable of the accessibility of inexpensive receivers for the
very precise determinations of the local direction GPS and the fact that a metric system is easier to
of gravity and therefore can determine the shape manipulate. The GPS provides locations by mea-
of the geoid. For a uid Earth, the geoid would be suring the times of travel of radio signals from a
a perfect ellipsoid, but the Earths geoid has a set of satellites to a receiving antenna held at the
very irregular shape. These irregularities mean desired location (Hofmann-Wellenhof et al., 1997).
that the local direction of gravitational accelera- The travel time measurement requires very
tion (down) does not point directly toward the precise (atomic) clocks on the satellites and
center of the Earth. Indeed, this acceleration sophisticated electronic techniques for synchro-
varies from place to place, because of topography nizing the clock in the receiver to the satellite
and variable rock density. In some applications clocks. Using these clocks, the time of sending
the particular geoid chosen as the datum is the and receiving a radio signal from a particular
one that best approximates mean sea level at the satellite can be differenced to compute the travel
nearest coastline. time, t, for the signal from that satellite. Knowing
Thus, an important question to ask about el- the velocity of the radio signal, v, corrected for
evations is: what is the datum? Is it a mathemati- atmospheric delays, and the travel time, t, the dis-
cally dened ellipsoid or a physically dened tance from the antenna to that satellite is com-
geoid (see Fig. 2.2a)? This information should be puted as d vt. Knowing the positions of all the
provided on any map that depicts elevation. On satellites (provided by the government operators
many older maps in the continental United of the system) and the respective distances from
States, for example, the datum used is the the antenna to each satellite, the location of the
North American Datum for 1927 (NAD-27). This antenna can be computed. A minimum of four
coincides with the ellipsoid that closely approxi- satellite distances is required, three to determine
mates the geoid in North America, but the center location and a fourth to synchronize the clocks.
of this ellipsoid is about 100 m from the actual Any redundant data are used to rene the preci-
center of the Earth. Another familiar datum is the sion of the location.
North American Datum for 1983 (NAD-83) that Locations can be determined to better than
appears, for example, on many topographic and 1 m using portable receivers, and these are an
geologic maps published by the US Geological excellent choice for mapping most geological
Survey. Today the datum used for the Global structures (Chapter 2 Frontispiece). The two geol-
Positioning System (GPS) is the World Geodetic ogists shown on the frontispiece for this chapter
are standing on an exposure of Aztec sandstone in tion and then are unwrapped to a at sheet for
the Valley of Fire, Utah. The GPS antenna, seen display. The projection is done primarily for con-
just above the left-hand shoulder of one of the venience, to avoid having to use three-dimen-
geologists, is on a short mast extending from the sional representations such as globes. Ideally one
geologists backpack. In the geologists hands is would want the map projection to represent geo-
the computer used to enter data and control the graphic information without distortion. For
receiver. The most precise (geodetic grade) GPS example, any two curves of equal length on the
provide locations to about 0.01 m using equip- sphere (or ellipsoid) should project to two curves
ment transported in vehicles and set up over of equal length on the at map. This carto-
stable benchmarks. The more precise measure- graphic criterion is called equidistance because
ments are done using a technique of differential such a projection correctly represents distances.
corrections that reduces errors by comparing the The cartographic criterion called conformality
signal received by the roving antenna to the signal refers to the correct representation of shapes on
received at a nearby base station where the loca- the map and requires that angles on the sphere
tion is xed and well known. (or ellipsoid) project to the same angles on the
At the time of publication of this book the GPS map. The third cartographic criterion, equiva-
is the preferred technology for locating ones posi- lency, requires the correct representation of areas
tion on Earths surface while mapping geological from the sphere (or ellipsoid) to the map.
structures. Because this is a rapidly developing Unfortunately, it is generally not possible to
technology we have only described the rudimen- achieve undistorted representations that meet
tary features of the system and refer interested all three criteria simultaneously; so different
readers to books and manufacturers manuals for projections (apparently there are about two
details concerning usage, precision, and instru- hundred in use) are chosen for different purposes
mentation (Committee on the Future of the with the objective of minimizing the distortion
Global Positioning System, 1995; Hofmann- or honoring one or other of the three carto-
Wellenhof et al., 1997). graphic criteria (Richardus and Adler, 1972).
To illustrate the procedure of projection we
2.1.2 Map projections: the Universal consider a spherical datum of radius R, take the
Transverse Mercator (UTM) so-called perspective point (view point) as the
projection center of the sphere, C, and project points from
Different forms of map projections are used to the sphere onto a plane that is tangent to the
depict geographic information located on Earths sphere at the point O (Fig. 2.3a). The point O should
curved surface on a at piece of paper (Alpha et al., be centrally located within the region where
1988). This procedure is a necessary and crucial structural mapping is planned. The north pole of
part of making maps that accurately record and the sphere is labeled N. Points on the sphere are
convey the eld data of structural geology. In this projected to the plane along straight lines that
section we describe what is meant by map projec- emanate from the perspective point. This is called
tion, discuss some of the complications of this pro- the gnomonic projection and it is one member of
cedure, and provide some illustrative examples. In the class of azimuthal projections that provide
particular we describe the UTM projection, which images similar to what one would observe on a
is becoming the standard for geological and topo- photograph taken from space along a line of sight
graphical maps. coincident with the normal to the datum ellip-
A curved surface is taken as the datum, either soid or sphere at the point O (Richardus and Adler,
a spherical or an ellipsoidal model of Earth, and 1972, Chapter 4). Among the attributes of the gno-
the projection surface is taken as a plane, monic projection is the fact that directions from
cone, or cylinder that is tangent to the datum the point O are not distorted. Also, straight lines
(Richardus and Adler, 1972). The cone and cylin- from this point represent arcs of great circles (the
der are wrapped around the datum for projec- intersections of the sphere with planes that pass
Proje dene the Cartesian coordinates (X, Y) of projected

(a) ction
plane points in terms of the geographic coordinates of
Y points on the sphere.
P' N Consider an arbitrary point, P(, ), on the
sphere (Fig. 2.3a) with latitude, , and longitude,
O P . Since the origin may lie on any meridian, the
Perspective relative longitude of P is determined by the dif-
0 point ference, O, between the longitude of the
meridian passing through the origin and that
C
passing through the point P. Use is made of the
Equator angle measured from the radial line CO to the
0 radial line CP. The point P is the projection of
the point P onto the projection plane. The angle
is measured in the projection plane from OX to the
Datum sphere line OP (Fig. 2.3b). The Cartesian coordinates of
of radius R
the projected point, P(X, Y ), are derived from the
(b) Projection plane trigonometry of Fig. 2.3 as functions of the radius,
R, and the two angles and (Richardus and
Y, Northing Adler, 1972, p. 59):
A P'
R sin cos
X OB OPcos R tan cos
cos
(2.1)
R sin sin
Y OA OPsin R tan sin
cos
X, Easting
In terms of the geographic coordinates, latitude
O B and longitude, the Cartesian coordinates of the
point P on the projection plane are (Richardus
Fig 2.3 The gnomonic projection (Richardus and Adler, and Adler, 1972, p. 59):
1972). (a) Spherical datum with projection plane tangent to
the sphere at point O. Symbols are identified in the text. R( cos sin )
(b) Cartesian coordinates (X, Y) of the projection plane. X
sin O sin cos O cos cos
(2.2)
R( cos O sin sin O cos cos )
through the center of the sphere) and therefore Y
sin O sin cos O cos cos
the shortest paths from this point to any other on
the sphere. Equations (2.2) are the mapping equations for
The point O(O, O) has a latitude O and longi- the gnomonic projection from a spherical datum
tude O and lies on an arbitrary meridian of as illustrated in Fig. 2.3. Given the radius of the
the sphere. This point is the origin of a two- datum and the geographic coordinates (latitude,
dimensional Cartesian coordinate system on the , and longitude, ) of any point on the datum,
projection plane (Fig. 2.3b). The X-axis is co- and the latitude, O, and relative longitude, , of
incident with the projection of a great circle the origin of the projection plane, these equations
through O and perpendicular to the central provide the Cartesian coordinates of that point on
meridian with positive X pointing to the east. The the projection plane (the map). For O /2 the
Y-axis is coincident with the projection of this point O is at the north pole and this is referred to
central meridian and positive Y points to the as a polar gnomonic projection. For O 0 the
north. The X- and Y-axes are referred to as easting point O is on the equator and this is referred to as
and northing, respectively. The objective is to a transverse gnomonic projection. If the point O is
(a) (c) Zone 10 84o N
12
W
0
126 o
oW
Northing
Equator
Equator
Easting
10 000 000 m
Central Y
(b) meridian
80o S
500 000 m
Equator X
Fig 2.4 The Transverse Mercator (TM) projection

(Hofmann-Wellenhof et al., 1997). (a) Spherical or ellipsoidal
Line of datum with right circular cylinder used for projection. (b)
latitude Cylinder unwrapped to form the projection plane with lines
of latitude and longitude. (c) Universal Transverse Mercator
Line of
longitude (UTM) grid (dashed lines) superimposed on projection.
on the Greenwich meridian . All forms of projection maps sections of the spherical or ellip-
this projection are limited in that an entire hemi- soidal datum onto a cylinder with its axis parallel
sphere cannot be projected: the boundary of the to the plane of the equator (Fig. 2.4a). The cylinder
hemisphere would plot at an innite distance is tangent to the ellipsoid along a particular line of
from the origin. longitude known as the central meridian. For this
One of the more common projections in use projection the equator and the central meridian
today is the Transverse Mercator (TM) projection in are straight lines whereas all other lines of longi-
which a spherical or ellipsoidal datum is projected tude and latitude are curved (Fig. 2.4b). Lines of
onto a right circular cylinder. The mathematical longitude to either side of the central line are
procedure used to accomplish the projection is concave toward this line. Lines of latitude, other
called conformal mapping (Hofmann-Wellenhof et than the equator, are concave toward their respec-
al., 1997, p. 287). As the name implies this trans- tive poles. Because this is a conformal projection
formation and others in this class preserve the the lines of longitude are everywhere orthogonal
angular relations of curves on the datum. The TM to the lines of latitude.
Taking a simple example for the sake of illus- these lines are projected as straight lines and they
tration consider the TM projection for a spherical are orthogonal to one another (Fig. 2.4c). Recall
datum of radius R. The Cartesian coordinates are that the other lines of longitude and latitude are
chosen so the X-axis is coincident with the equator projected as curved lines. The UTM grid is a
and Y-axis is coincident with the central meridian. Cartesian (rectangular) metric grid overlaid on
They are related to the latitude, , and longitude, this projection. The western edge of the UTM grid
, as (Richardus and Adler, 1972, p. 101): is a line drawn parallel to the central meridian but
500 000 m (about 4.5) to the west. The eastern
1
X R ln
2
1 cos cos
1 cos cos (2.3)
edge of the UTM grid is 500 000 m to the east of the
central meridian. Values along this axis are
Y R tan 1( cot sin )
referred to as false eastings because of the shift in
origin, and these values are measured parallel to
This projection is most appropriate where the the equator, starting at the western edge of the
mapped objects are organized along a particular UTM grid. Thus, the central line of longitude is at
meridian, or where the datum can be mapped sep- a false easting (X-coordinate value) of 500 000 m.
arately in narrow strips aligned with different Northing is measured in the northern hemi-
lines of longitude. sphere from the equator toward the north and
Most GPS receivers have onboard computers parallel to the central line of longitude. In the
that are capable of reporting locations using the southern hemisphere the zero northing is shifted
so-called Universal Transverse Mercator (UTM) projec- to the south, so the equator is at a false northing
tion, which is a modication of the TM projection. of 10 000 000 m. Note that for the depiction of
The UTM system was designed to meet a number zone 10 shown in Fig. 2.4c the easting axis is
of criteria including conformality to minimize stretched relative to the northing axis. At true
directional errors, a minimum number of zones, scale the zone in each hemisphere would be a very
limited errors in scale, a common referencing thin strip that is ten times taller than it is wide.
grid, and limited convergence of lines of longi- To specify the UTM coordinates of a particular
tude toward the poles (Richardus and Adler, 1972, point on Earths surface, the hemisphere and the
p. 138). The design of this system has proved to be zone number and the datum must be identied,
quite good so UTM coordinates are found on most along with the appropriate easting, northing, and
modern maps. To set up the UTM system the Earth elevation. Thus, for example, the UTM coordinates
is divided into 60 zones (the ancient Babylonians of a location just outside the Geology Building at
would like this choice!), each spanning 6 of lon- Stanford University are reported as follows:
gitude and each extending from 80 S to 84 N, and
Northern hemisphere
these zones are individually projected (e.g. Fig.
Zone M10
2.4c). The zones are numbered consecutively
Datum WGS-84
toward the east from M1, which spans from 180
Easting: 573 218.49 m
W to 174 W longitude, with a central line of lon-
Northing: 4 142 572.31 m
gitude at 177 W. Thus, for example, zones M10
Elevation: 3.6 m
through M19 span the United States from just off
the west coast (126 W line of longitude) to just off The horizontal precision of the measured coordi-
the east coast (66 W line of longitude). The central nates is about 3 cm for easting and northing and
meridian of zone M10 is 123 W and the central the vertical precision is about 10 cm. Recall that
meridian of zone M19 is 69 W. The projection the mean sea level elevation is about 28.9 m and
cylinder is tangent to the central meridian if the the negative elevation listed here is referenced to
datum is taken as a sphere. the WGS-84 ellipsoid (Fig. 2.2b).
To establish the UTM coordinates within a par- At the time of writing of this book the UTM
ticular zone, the central line of longitude and the projection is the preferred coordinate system for
equator are used as reference lines because both of the preparation of geological and structural
maps. Formulae to compute the coordinates on transformation from one coordinate system to
the UTM grid given geographic coordinates on an another because these are basic tools for struc-
ellipsoidal datum, and to calculate the geographic tural mapping.
coordinates on the ellipsoidal datum given the It is useful to understand the concept of
UTM coordinates have been derived (Richardus vectors in general, and to be familiar with specic
and Adler, 1972). These take several pages to write techniques for manipulating vectors, because
down and are not repeated here. Many GPS structural geologists employ them for modeling
systems have these formulae built in and make as well as mapping. This should come as no sur-
the computations at the push of a button. prise because the evolution of geologic structures
is primarily a physical process and the physical
laws that describe such a process are written as
2.2 Local coordinates and position vector equations. This section focuses on position
vectors, but also serves as a summary of some
vectors general vector concepts that we build upon in
later sections where, for example, we use vectors
Points and sets of points can be dened only relative to to characterize the shapes of folded geological sur-
(i.e., as functions of ) a coordinate system, never faces and to visualize velocity elds within a
absolutely. The coordinate system is the unavoidable deforming rock mass. Other general vector con-
residue of the eradication of the ego in that geomet- cepts are introduced as needed in later sections
rico-physical world which reason sifts from the given
and chapters.
using objectivity as its standard a nal scanty
token in this objective sphere that existence is only
given and can only be given as the intentional content 2.2.1 Locating data using local
of the processes of consciousness of a pure, sense- coordinates and position vectors
giving ego (Weyl, 1987). Geographical coordinates, based on the UTM
grid, may not be the best choice for mapping in
Students rst learn to deal with points and sets the eld, but their use is becoming universal for
of points described relative to a chosen coordi- the nal presentation of structural maps. For
nate system in elementary courses in mathemat- large-scale maps awkwardness arises because the
ics. This procedure was apparently conceived by central line of longitude for each UTM zone is
Ren Descartes in the early seventeenth century 500 000 m to the east of the origin (Fig. 2.4c), so a
and has become one of the most powerful tools typical easting would have six digits with meter
ever developed for scientists and engineers precision and eight digits if centimeter precision
(Davis and Hersh, 1986; Aczel, 2000). In honor of were required. Similar numbers of digits are
Descartes contribution the most familiar coordi- required for the northing unless, for example, the
nate system we use is referred to as the Cartesian location is near the equator in the northern
coordinate system. With this coordinate system and hemisphere. To reduce this cumbersome number
the concept of a position vector one can locate of digits it is practical to select a local origin
outcrops or the point where samples are col- within the mapped region. Local origins are also
lected in the eld relative to a local origin. In commonly employed in modeling and data analy-
other words the origin is located in the region sis. Furthermore, a local origin may be necessi-
being mapped or on the actual outcrop for very tated by the use of surveying equipment that
large-scale mapping, instead of at an arbitrary references locations to the instrument itself
point determined by a global projection such as rather than the UTM system. Topographical, geo-
the UTM grid. To put the map or data in a global logical, and structural maps can be prepared
context one transforms the position vectors from using a coordinate system consisting of the
the local coordinate system to the UTM system. easting, northing, and elevation for which the
Here we introduce the position vector, several datum is the local elevation of the instrument
local coordinate systems, and equations for the and the easting and northing are measured in the
2.2 LOCAL COORDINATES AND POSITION VECTORS 35
horizontal plane from a local origin at the instru- (a)

ment site. These maps also can be prepared using
a local Cartesian coordinate system with x- and y-
Dike
axes in the local (horizontal) datum and the z-axis
Dike
x z
vertical.
As an example consider the region (Fig. 2.5a)
near Ship Rock, New Mexico, where volcanic y
necks and associated igneous dikes crop out in
the Mancos Shale of the San Juan Basin (Delaney Volcanic Volcanic
neck neck
and Pollard, 1981). In this photograph, taken
from the largest volcanic edice called Ship Rock,
three smaller volcanic necks are seen in the fore- (b)
y
ground and three dikes crop out along low ridges
Northeastern Ship Rock Dike
that trend out into the basin to the northeast. A
rough scale for the photograph is provided by a
one-lane dirt track, about 2 to 3 m wide, that scars 1) 1)
p(2 q(2
the desert landscape just to the south (to the N
right) of the middle dike. As read from the topo-
56o
graphic map for this quadrangle, the UTM coor- 10 m
0 5
dinates of a point near the western (proximal) x
termination of this dike, called the northeastern 0
dike, are:
Fig 2.5 Structural mapping at Ship Rock, NM, using local
Northern hemisphere
coordinates and position vectors (Delaney and Pollard,
Zone M12
1981). (a) Photograph of volcanic necks and igneous dikes
Datum NAD-29 with local Cartesian coordinate system. (b) Aerial
Easting: 694 000 m photograph of dike outcrop with local coordinate system and
Northing: 4 063 000 m position vectors, p(21) and q(21), marking contact.
Elevation: 1675 m Photograph by D. D. Pollard.
The precision of these coordinates is about 30 m

for easting and northing and the vertical precision plays a crucial role in the mechanics of dike for-
is about 5 m. To map the dikes and volcanic necks mation this kind of representation is inadequate
with greater precision a surveying instrument, for structural investigations. The northeastern
such as a total station, could be set up at this loca- dike at Ship Rock is composed of thirty-ve dis-
tion and a new origin and datum established at the crete segments ranging from less than 1 m to
instrument site. On Fig. 2.5a we have positioned a about 7 m in thickness with an average of about
Cartesian coordinate system there with the (x, y)- 2.3 m (Delaney and Pollard, 1981). Therefore, to
plane as the datum. Total stations are capable of achieve a precision of about ten percent of the
determining the local coordinates of points rela- average thickness, say 20 cm, the two adjacent
tive to the instrument location over distances up to contacts must be located to within about a
a few kilometers with precisions ranging from a decimeter. Instruments were not available in 1980
few centimeters to about a decimeter. to survey the contacts directly. Instead, vertical
On most geological maps dikes are repre- photographs were taken from a low-ying aircraft
sented as solid lines with a uniform thickness and the central portions enlarged eight times to
determined by the diameter of the pen or the yield a map scale of about 1 : 228. Mapping was
minimum line width of the printer used to done directly on the enlarged photos with a pen
produce the map, rather than the scaled thick- diameter of about 0.24 mm, equivalent to about
ness of the exposed dike. Because dike thickness 5 cm on the ground. Although the exposure of the
(a) (b) z
z
ez az
p
pz
ay
p
y
O
ey y
ex ax O
x c
px
(c) z x
z
py
pz z
(d)
P(x, y, z)
y z O
pz P p
p y
c
p
O x
y
O a
y
x x x b
Fig 2.6 The position vector, p. (a) Base vectors, e, and from scalar quantities, which have a magnitude,
direction angles, . (b) Components (px, py, pz) of the and possibly both positive and negative values,
position vector. (c) Rotation from the (x, y, z) coordinate but scalars lack a specic direction.
system to (x y z). (d) Translation of the coordinate system. In general, a vector may be written as a linear
combination of three base vectors and three quan-
dike introduced some uncertainty in locating the tities called the scalar components of the vector. For
contact, this procedure generally provided the position vectors we choose a set of three mutually
required precision. A portion of the rst photo- orthogonal base vectors (ex, ey, ez) each of unit
graph in the series is shown in Fig. 2.5b along with magnitude and each directed, respectively, from
the local coordinate system. the origin along the positive axes (Ox, Oy, Oz) of a
Every point in three-dimensional space is right-handed Cartesian coordinate system (Fig.
uniquely determined by a position vector, p, that 2.6a):
species the position of the point relative to a
ex (1, 0, 0), ey (0, 1, 0), ez (0, 0, 1) (2.4)
xed origin, O (Fig. 2.6a). The position vector is
visualized as an arrow extending from the origin These equations give the values of the three scalar
to the point. In other words the tail of this vector components for each unit base vector where it is
is at the origin and the head is at the designated understood that the respective components are
point. As with all vectors, the position vector has for the three coordinate directions (Ox, Oy, Oz).
both a magnitude and a direction. The length of Mutually orthogonal unit base vectors are said
the arrow is drawn proportional to the magni- to form an orthonormal basis (Malvern, 1969). In
tude, and the shaft and head of the arrow pre- some notations different symbols are used to
scribe the direction. In this text, vector quantities distinguish each base vector (e.g. i, j, k), but here
are written in boldface type to distinguish them a common symbol is used with subscripts to
associate each base vector with a particular coor- orthogonal projections of v onto lines drawn
dinate direction. Most vector quantities used in parallel to the base vectors (ex, ey, ez), respectively.
this book, such as the position vector or velocity If the two points are coincident v 0 (0, 0, 0) is
vector, are free to take on any orientation and the zero vector.
therefore do not carry subscripts. The coordinate The meter (m) is the physical unit used to
axes, and therefore the base vectors, comprise a measure distance, and this is carried by the posi-
right-handed system. By right handed we mean tion vector components, not the base vectors.
that the axes (Ox, Oy, Oz) extend, respectively, in Thus one would write, for example, py 22.5 m
the same mutual orientations as the thumb, and |ey| 1. Each base vector has a magnitude of
index nger, and middle nger of the right hand one but does not carry a physical unit. The base
when these ngers are directed orthogonal to one vectors provide the directional information nec-
another. Base vectors need not be orthogonal to essary to compose a position vector. The three-
one another nor of unit magnitude, and Cartesian dimensional space dened in terms of position
coordinates can be left-handed, but these alterna- vector components using the basis (ex, ey, ez) is
tives are not employed in this book. referred to as Euclidean space (Lipschutz, 1969).
The scalar components of a vector are quanti- When the base vectors change direction
ties that can be visualized as the orthogonal pro- because of a rotation of the coordinate system
jections of the vector arrow onto lines in arbitrary about the origin (Fig. 2.6c), the components of the
but designated directions. In general, scalar composition vector p for an arbitrary point P change,
ponents are written with the same symbol as the but the magnitude and direction of the position
vector, but the type is not bold and each compo- vector do not change. For example, the projection
nent has a subscript that identies the line pro- pz of p onto the rotated coordinate axis Oz is dif-
jected upon. For the position vector, p, the ferent from the projection pz onto the original
designated directions are parallel to the base coordinate axis Oz. Thus we say that (px, py, pz) are
vectors (Fig. 2.6b), in other words lines parallel to the components of p with respect to a particular
the respective coordinate axes. The three compo- basis (ex, ey, ez), and this basis must be dened in
nents of the position vector are written (px, py, pz). order to interpret the components. On the other
Thus, the equation for the position vector with hand if the base vectors and coordinate system are
the basis (ex, ey, ez) is written as the following translated to a new origin, O, but not rotated (Fig.
linear combination of scalar components and 2.6d), a new position vector, p, extends to the
base vectors: given point, P, and this vector has a different mag-
nitude and/or direction than p. In general, two
p pxex pyey pzez (2.5)
vectors are equal only if they have the same mag-
Although it is convenient to speak of a position nitude and direction, so the position vectors for
vector simply as p, and this notation is useful for the point P referred to different origins are not
writing compact vector equations, most calcula- equal, despite the fact that they dene the loca-
tions are done using the scalar components. tion of the same point. In this sense position
Apparently Descartes introduced the notion of vectors are so-called xed vectors (Malvern, 1969):
using vector components, dened with respect to they emanate from a particular point, the origin
a coordinate system in order to develop analytical of the coordinate system, and the location of this
geometry (Fung, 1969, p. 22). origin must be specied in order to dene a set of
An arbitrary vector, v, may have its tail at any position vectors. Position vectors are useful quan-
point (x1, y1, z1) and its head at any other point (x2, tities for dening the locations of points, but they
y2, z2) so this vector is written: lack the attributes of those vectors that we think
of as physical entities, such as force or velocity,
v (x2 x1)ex ( y2 y1)ey (z2 z1)ez
(2.6) which are independent of an arbitrarily dened
vxex vyey vzez
coordinate system.
The quantities (vx, vy, vz) are the scalar components The magnitude of any vector is calculated as
of the vector, which can be interpreted as the the square root of the sum of the squares of its
scalar components and this quantity is written Table 2.1. Position vectors for ship rock dike.
with the same symbol as the vector but not in
boldface type. For the position vector this rela- Counter, i px (m) py (m) qx (m) qy (m)
tionship may be deduced from Fig. 2.6b and the
theorem of Pythagoras. The squared length of the 1 3.9 15.7 3.9 15.7
position vector is p2 (pz)2 c2, but c2 (px)2 (py)2. 2 4.6 15.4 4.6 16.0
Thus, the magnitude, p, of the position vector p is: 3 5.8 15.4 5.8 16.0
4 7.0 15.3 7.0 16.1
p |p| p2x p2y p2z 0 (2.7) 5 8.1 15.2 8.1 16.0
The magnitude of a vector, written symbolically ... ... ... ... ...
using ||, is equivalent to the scaled length of the 2720 2900.2 6.1 2900.2 4.4
vector arrow, and is always greater than or equal 2721 2901.4 6.2 2901.4 4.7
to zero. The components of the position vector are 2722 2902.5 6.4 2902.5 5.0
proportional to the vector magnitude and the 2723 2903.7 6.5 2903.7 5.2
cosines of the angles that the vector makes with
2724 2904.8 5.9 2904.8 5.9
the positive coordinate axes (Fig. 2.6a):
px p cos x, py p cos y, pz p cos z (2.8)
think of the photograph-based map as the projec-
The set of three angles (x, y, z) are called the tion of the dike outcrop onto the (x, y)-datum plane
direction angles of the position vector, and the of the local coordinate system. Then, for example,
cosines of these angles are referred to as the direc- the position vector for the counter i 21, shown in
tion cosines. The direction angles are measured in Fig. 2.5b, is written in the form of (2.5) as:
the planes dened by the vector and the respec-
p pxex pyey (26.7 m)ex (18.2 m)ey (2.9)
tive coordinate axis, and each is taken as the
smaller of the two angles between the vector and Because all of the vectors used to quantify the dike
the positive coordinate axis. contact lie in the (x, y)-plane, the direction angle
At Ship Rock (Fig. 2.5a), the UTM coordinates of z /2 (Fig. 2.6a) and, for example, using (2.8) and
the local origin near the western (proximal) ter- i 21 we have:
mination of the middle dike are given above. The
local Cartesian coordinate system is oriented with
the z-axis vertical (upward) and the x-axis directed
py p cos y p cos
2
x p sin x
in an azimuth of 056, which is approximately and

parallel to the outcrop trace of the dike. Two sets px p cos x (2.10)
of position vectors, p(i) and q(i), trace out the so
contact between the Mancos Shale and the
x tan 1( py px) 34.3
igneous rock of the dike with p(i) along the north-
western side and q(i) along the southeastern side For two-dimensional cases only one independent
(Fig. 2.5b). The spacing is about 1 m between direction angle is required to orient the position
members of a given set and there are 2724 vectors.
members of each set that dene the shapes of the Digitization of the dike map from Ship Rock is
thirty-ve dike segments. The rst few and last few carried out such that the x-components of p and q
members are given in Table 2.1 and the entire are identical for each value of the counter. Also,
data set is provided at the textbook website. the dike segments are approximately parallel to
The z-components of the position vectors are the x-axis of the local coordinate system.
not listed in Table 2.1. In fact the dike outcrop Therefore, for a given value of the counter, the
varies in elevation by about 55 m over the nearly dike thickness, t(i), is approximately equal to the
3-km distance from the southwestern to the difference between the y-components of the two
northeastern termination, and these elevation position vectors, which is equal to the magnitude
differences were not measured. In effect one can of the vector difference:
t(i) py(i) qy(i) |p(i) q(i)| (2.11) its nal thickness simply by opening a large crack.
The physical insights and new data gained through
This can be understood by considering the general precise quantitative mapping are the reward for
expression for the sum of two vectors: the effort required to learn and implement the
techniques.
if r v w, then rx vx wx,
ry vy wy, rz vz wz (2.12)
2.2.2 Transformation of position and
The difference of the two vectors may be thought basis vectors, and coordinate
of as r v (w). In other words the sum (or dif- systems
ference) of two vectors is a vector with scalar com- For the presentation and publication of maps and
ponents that are the sum (or difference) of the structural data sets it is appropriate to reference
respective components. This result can be gener- locations to the UTM grid (Fig. 2.4). For example,
alized to the sum or difference of any number of the local Cartesian coordinate system (x, y, z) used
vectors. For a given value of the counter the x-com- to dene the position vectors for the contact of
ponents of the position vectors from Table 2.1 are the northeastern Ship Rock dike (Fig. 2.7) can be
equal, and the z-components are zero for all these transformed to the UTM grid through a two-step
vectors, so the magnitude (2.7) of the vector dif- procedure starting with a clockwise rotation
ference reduces to the difference of the y-compo- through 34 about a vertical axis to align the x-axis
nents. The set of thickness values so determined is and y-axis with the easting and northing direc-
used in a mechanical model for the opening of the tions on the grid. This is followed by a translation
dike segments that leads to estimates for the stiff- of the local origin to the origin for the northern
ness of the host rock (Delaney and Pollard, 1981). hemisphere of zone 12, which is 694 000 m to the
Several segments of the northeastern Ship Rock west, 4 063 000 m to the south, and 1675 m down.
dike are shown on Fig. 2.7, a structure map using In general, changing (transforming) from one
local Cartesian coordinates (Fig. 2.5b). The precise rectangular coordinate system to another may
mapping method described above brings out a host involve three independent operations: a rotation
of structural features (Delaney and Pollard, 1981) about a xed origin (Fig. 2.6c) without changing
that otherwise would be unrecorded on typical the orthogonality or sense of the axes; a transla-
quadrangle-scale maps where thin constant-width tion of the origin (Fig. 2.6d) without changing the
lines only display the location and trace of dikes. orientation or sense of the axes; and a reection
Instead we note that the dike is composed of sub- that only changes the sense of the axes (right
parallel but offset segments forming an echelon handed to left handed or vice versa) while main-
array. Segments numbered 13 to 15 are offset by a taining their orientation and the origin. Because
few meters and have blunt, rounded terminations, we utilize only right-handed coordinate systems,
whereas segments 15 to 21 are offset by as much as the operation of reection is not considered
15 m and have more tapered terminations. The further. Here we illustrate rotation and transla-
contacts of overlapping segments have a distinct tion using the Ship Rock example and then gen-
asymmetry such that adjacent contacts curve eralize these operations for the transformation of
toward their respective terminations and distal vectors and rectangular coordinate systems in
contacts are relatively straight. At x 1000 m, on three dimensions.
segment 12, both contacts bulge outward to form The azimuth of the local x-axis at Ship Rock is
a putative volcanic neck; and at x 1310 m, on 56 measured clockwise from north in the hori-
segment 19, one contact bulges outward such that zontal plane (Fig. 2.5). Therefore, the transforma-
the total thickness almost doubles. Both of these tion to a new Cartesian coordinate system (x, y)
structures are associated with breccias that that shares the same origin but is aligned with the
suggest the local increases in thickness are related UTM grid axes is a clockwise rotation of 34 about
to fracturing and brecciation of the host rock and the vertical z-axis. This is a two-dimensional trans-
transport of the breccia by owing magma. At formation in which every point, P(x, y), referred to
these locations the dike apparently did not attain the old coordinate system on the map takes on
Meters
50
13
Km
9 Tmn 11 12
Km Thb
Tmb Thb Tmb 10 Tmb
83
Thb
7 Tmn Thb
8
0
800 900
16
14 Thb
15 75 Tmb
Thb
Thb Thb Tmn
Km
Tmn
Km
1000 1100
Km 18 19 87 80
Tmb
17 Thb
Thb 23
87 90
80 Km
82
1200 1300
Tmn Minette
Contact of dike with exposed host rock (km)
Thb Heterobreccia
Contact located to within 0.05 m
Contact approximately located; possible error greater that 0.1 m Tmb
Tmb Monobreccia
28 Strike and dip of beds
Mancos Shale
25 Strike and dip of joints
Km Meters
50
82 Strike of vertical joints
Km
9 Segment number 21 74 Thb
Thb Tmn
20 Thb 77
86
Km
Tmn
88
0
1400 1500
Fig 2.7 Structural map of several segments of the

northeastern Ship Rock dike (Delaney and Pollard, 1981).
new coordinates, P(x, y), which are found using (a) in a

y xs
the following equations: y
in a
x x cos y sin ys
(2.13) P(x, y)
y x sin y cos a
sa x
o
The angle is the counterclockwise angle from a xc
the Ox to Ox and the trigonometry used to derive
these equations is illustrated in Fig. 2.8a. For the a
os
local coordinate system at Ship Rock the angle yc a
34. x
O
The x- and y-components of position vectors p
and q (Fig. 2.8b) change under the rotational
transformation just described, but the position (b) y px
y
vectors themselves are unchanged. Because the
components of an arbitrary position vector, p,
that locates a point, P(x, y), are equivalent to the
coordinates of that point, the two-dimensional x
rotational transformation of position vectors p py
about the z-axis follows directly from (2.13):
px px cos py sin a x
(2.14) p y
py px sin py cos O p x
Here px and py are the components of the same
position vector, p, locating the same point, P(x, y),
but this vector is referred to the new (x, y)
y
(c) y
coordinate system.
For a two-dimensional rotational transforma-
tion the basis for the new (x, y) coordinate system
is not the same as that for the old (x, y)-system (Fig. x
ey ey
2.8c). However, the base vector ex only has an x-
component, and ey only has a y-component, so the ex
a
rotational transformation to the new base vectors x
ex
(ex, ey) may be taken directly from (2.13):
O
ex ex cos ey sin
(2.15)
ey ex sin ey cos Fig 2.8 Rotation of Cartesian coordinates in two
dimensions. (a) New coordinates of the point P.
These are vector equations that relate two different (b) Components of the position vector under the old (x, y)
pairs of orthogonal unit vectors. One pair aligns and new (x y) coordinate systems. (c) Base vectors, e, for
with the positive axes of the old coordinate system the old and new coordinate system.
and the other with the positive axes of the new
system. The fact that both ex and ey are unit vectors
follows from the Pythagorean relation sin2 the new (x, y)-coordinate system back to the old
cos2 1. In contrast to the vector equations (2.15), (x, y)-system merely reverses the operation just
the rotational transformation illustrated in Fig. described. The inverse form of the rotation equa-
2.8b is for the single position vector, p, and utilizes tions is found by solving (2.13) for x and y:
scalar equations (2.14) for the components of that
vector referred to the new coordinate system. x xcos ysin
(2.16)
Transforming the coordinates of a point from y xsin ycos
The inverse transformation for the components of (a) z

a position vector, or for the base vectors, follows
from (2.16) by analogy with (2.14) and (2.15).
Coordinates of points on the Ship Rock map in z
the rotated local system (x, y, z), and compo- y
nents of position vectors referred to that system,
are transformed to the UTM grid by a translation
of the origin using the following equations: (z, x )
Easting x 694 000 m y
Northing y 4 063 000 m (2.17)
x
O
Elevation z 1675 m (x, x )
(y, x )
Notice that the transformation is accomplished x
by adding to the local coordinates the coordinates
of the local origin referred to the UTM system.
(b) z
The transformation of local coordinates at the
Ship Rock site to the UTM grid is an example of
a translation from one rectangular (Cartesian)
system to another rectangular system with paral-
lel axes (Fig. 2.6d). For the purpose of generalizing w
this discussion we refer to (x, y, z) as the old system v
|w|
and (x, y, z) as the new system, with origins at O |v| cos u
and O, respectively. For every point, P(x, y, z), O y
referred to the old coordinate system, the rectan-
gular coordinates in the new system are found
using the following expressions:
x x a x
y y b (2.18)
z z c Fig 2.9 Rotation of Cartesian coordinates in three
dimensions. (a) Direction angles relating old (x, y, z) and new
Here the constants (a, b, and c) are the coordinates
(x y z) axes. (b) Geometric interpretation of scalar product
of the origin, O, of the new system as written with of two arbitrary vectors, v and w.
respect to the old system. That is, the origin of the
old system is translated a distance a along x and a
distance b along y and a distance c along z from the two angles in the plane dened by Ox and Ox.
the origin of the old system (Fig. 2.6d). The inverse This direction angle is referred to as (x, x).
of this transformation is found algebraically by Similarly, the direction angle (y, x) relates the y-
rearranging the three equations to solve for x, y, axis to the x-axis, and the direction angle (z, x)
and z, respectively. relates the z-axis to the x-axis. Three direction
The two-dimensional rotational transforma- angles are dened similarly for the y-axis and for
tion (2.13) can be generalized to three dimensions the z-axis, bringing the total to nine angles.
if the old and new coordinate systems share a The old and new coordinate systems (Fig. 2.9a)
common origin. The coordinate axes (Ox, Oy, Oz) of have bases (ex, ey, ez) and (ex, ey, ez), respectively.
the old system are related to the coordinate axes To calculate the new basis vectors from the old
(Ox, Oy, Oz) of the new system using nine direction we need to dene the scalar product of two vectors.
angles. For example, in Fig. 2.9a the positive x-axis For two arbitrary vectors, v and w, the scalar
is related to the positive x-axis by the smaller of product is:
v w |v||w| cos , for 0 The direction cosines and basis vectors may be
vxwx vywy vzwz (2.19) organized into a table that facilitates rotational
transformations in three dimensions:
Because the scalar product is written with a dot
ex ey ez
between the two vectors, it is referred to as the dot
product. Here is the smaller of the two angles ex mxx mxy mxz
(2.23)
between the lines directed parallel to v and w ey myx myy myz
measured in the plane dened by the two vectors ez mzx mzy mzz
(Fig. 2.9b), so the product |v|cos is the orthogo- The new and old basis vectors are listed in the rst
nal projection of v onto the line parallel to w. In row and rst column, respectively, and the direc-
other words |v|cos is the component of v in the tion cosines ll out the table such that their two
direction of w. Therefore, the scalar product may subscripts match the subscript of the basis vector
be interpreted geometrically as the product of the heading the row and the column, respectively.
component of v along w and the magnitude of w. Each new basis vector is composed of a linear
Alternatively, the scalar product may be inter- combination of the old basis vectors and the
preted as the product of the component of w direction cosines in the corresponding column of
along v and the magnitude of v. (2.23):
Two important relationships among the
vectors dening a basis are expressed con- ex mxxex myxey mzxez
veniently using the scalar product (2.19): ey mxyex myyey mzyez (2.24)
ex ex ey ey ez ez 1 ez mxzex myzey mzzez
ex ey ey ez ez ex 0 (2.20)
Each old basis vector is composed of a linear com-
These relations are interpreted by recalling that bination of the new basis vectors and the direc-
orthonormal basis vectors are of unit magnitude tion cosines in the corresponding row of (2.23):
and they are mutually orthogonal to one another.
ex mxxex mxyey mxzez
For the rotational transformation we want, for
example, to project ex, ey, and ez onto the line par- ey myxex myyey myzez (2.25)
allel to ex (Fig. 2.9a). These projections are accom- ez mzxex mzyey mzzez
plished using the scalar product (2.19) and
recalling that the base vectors are unit vectors: This is the inverse rotational transformation. The
basis vectors are unit vectors so (2.7) requires that
ex ex |ex||ex|cos (x, x) cos (x, x) mxx the sum of the squares of the direction cosines in
ey ex |ey||ex| cos (y, x) cos (y, x) myx (2.21) any row or column of (2.23) equal one. For
ez ex |ez||ex|cos (z, x) cos (z, x) mzx example:
Analogous procedures are used to project ex, ey, |ey| (mxy)2 (myy)2 (mzy)2 1,
and ez onto lines parallel to ey and ez and these so (mxy)2 (myy)2 (mzy)2 1 (2.26)
operations dene nine direction cosines that relate
The direction cosines as arranged in (2.23) com-
the old basis to the new basis:
prise a square matrix and the rotational transfor-
mxx cos (x, x), mxy cos (x, y), mxz cos (x, z) mation will be described in terms of matrix
operations later.
myx cos (y, x), myy cos (y, y), myz cos (y, z)
It was pointed out with respect to the two-
mzx cos (z, x), mzy cos (z, y), mzz cos (z, z)
dimensional form of the rotational transforma-
(2.22)
tion that basis vectors and components of
Here the double subscripts on the direction position vectors and coordinates of points all
cosines refer to the reference (old) axis and the transform using similar equations. This principle
transformed (new) axis, respectively. extends to three dimensions so we can construct
a table of directional cosines for rotation of posi-

tion vector components using (2.23) as a template: (a) Km
px py pz Km

px mxx mxy mxz
(2.27) Thb
py myx myy myz Tmn
pz mzx mzy mzz Tmn Thb
Tmn
Tmn Thb
For example, one uses mzx for the orthogonal pro- Thb
N
jection of the z-component of the position vector Km
onto the x-axis. Each new component of the posi- Tmn 0 10 20 Meters
m
tion vector is composed of a linear combination of
the old components and the direction cosines in the
corresponding column of (2.27). Each old compo-
(b) Plug Plug
nent of the position vector is composed of a linear Tmn Tmn
combination of the new components and the direc-
tion cosines in the corresponding row of (2.27). TmnTmn
Because the coordinates of an arbitrary point,
P(x, y, z), are equivalent to the components of the Km
Thb
position vector, p, for that point, the coordinates Tmn Thb Km
Km Tmn
can replace the components in (2.27) to form a Tmn
Thb Thb Km Dike Tmn
table of direction cosines for the rotational trans- Tmn N
Thb Thb 0 Dike
2m
formation of coordinates: Tmn N
x y z
0 2 Meters
x mxx mxy mxz
(2.28) (c)
y myx myy myz
z mzx mzy mzz
From this table we understand, for example, that

the direction cosine myz is used for the orthogonal
projection of the y-coordinate of P onto the z-axis.
Each new coordinate of P is composed of a linear
combination of the old coordinates and the direc-
tion cosines in the corresponding column of
(2.28). The inverse transformation equations are
formed using the rows of this table. In summary
(d) y r cos u
(2.23), (2.27), and (2.28) provide the equations for
both forward (old to new) and inverse (new to old) r = constant
rotational transformations for orthonormal base P(r, u)
vectors, for components of position vectors, and r =R r s r sin u
for coordinates of points referred to rectangular u
Cartesian coordinate systems. O x

Transformation equations that commonly are u = constant
employed in modeling structures include those
for the cylindrical coordinate system. As an example
consider the volcanic neck shown in the lower- Fig 2.10 Volcanic neck near Ship Rock, NM (Delaney and
right corner of the photograph from Ship Rock Pollard, 1981). (a) Structural map of the neck and associated
(Fig. 2.5a). This edice, once a conduit for magma, dikes. (b) Detailed map of the neck and a dike. (c) Aerial
photograph of the neck. (d) Polar coordinate system used in
is about 30 m in diameter and roughly circular in
models of volcanic necks.
map view (Fig. 2.10a, c). The neck is composed pri-
marily of a volcanic igneous rock called minette, z

(a)
but breccias of minette and Mancos Shale crop
out around the perimeter. From the side of the
neck a thin dike extends about 440 m to the
northeast along a gently curving outcrop (Fig.
2.5a), and a dike with similar trend extends from P(r, u, z)
the opposite side of the neck a few meters to the
southwest. The crosscutting relationships (Fig. p
ez
2.10b) between the minette of the dike and the
ey y
minette and breccias of the neck suggest that the O
dike formed rst and solidied while local brec- ex
r
ciation and erosion of the Mancos Shale and the u
dike rock enabled the neck to grow in diameter to
its present size (Delaney and Pollard, 1981). This x s
interpretation is surprising because the analogy
to hydraulic fracturing of wells (Hubbert and (b) z
Willis, 1957) would suggest that the neck formed
rst and that pressurized magma in the neck frac-
tured the contact, initiating dike propagation
into the Mancos Shale. The interpretation that
dikes form rst and necks grow from them is sup- p
ported by observations on active volcanoes where w
short-lived ssure eruptions precede longer-lived ez
P(r, w, u)
eruptions from cylindrical vents located along ey y
the ssure. O
ex
The forms of the volcanic necks at Ship Rock
motivate the choice of a cylindrical coordinate u
system for idealizing the geometry and facilitat-
ing modeling (Fig. 2.11a). For reference consider x
the Cartesian axes (Ox, Oy, Oz) with Oz parallel to
Fig 2.11 Position vector, p, and base vector, e, for two
the cylindrical axis. The basic elements of the
coordinate systems. (a) Cylindrical coordinate system.
cylindrical system are an origin, O, a radial dis- (b) Spherical coordinate system.
tance, r, measured from the origin, an angle, ,
measured from the line Ox, and an axial distance,
z, measured from the origin. Both r and are mea- the axial coordinate z is identical for the two
sured in the plane perpendicular to the cylindri- transformations. Following standard conventions
cal axis and the angle is positive if clockwise when the positive square root is used so r is always a pos-
looking in the positive direction of Oz. Given an itive number. The line Ox in the Cartesian system
arbitrary point P(r, , z) in cylindrical coordinates, and that in the polar system must be identical,
the Cartesian coordinates are found using the fol- sharing the same origin and direction.
lowing transformation equations (Selby, 1975, The position vector, p, used to locate the point
p. 385): P(r, , z) in the cylindrical coordinate system (Fig.
2.11a) is written as a linear combination of the
x r cos , y r sin , z z
(2.29)
scalar components and the three mutually

y orthogonal unit base vectors (ex, ey, ez) of the
r x2 y2, tan 1 , zz
x Cartesian system:
p pxex pyey pzez
The second line of (2.29) contains the correspond-
ing inverse transformation equations. Note that (r cos )ex (r sin )ey (z)ez (2.30)
The Cartesian coordinates dened in the rst line ing to the governing equations for heat conduc-
of (2.29) are the scalar components of the position tion with the boundary condition that the tem-
vector. These scalar components carry the units of perature far from the neck remains zero:
length (meters) as do the radius, r, and axial
length, z. The angle, , is measured in radians
which may be thought of as the ratio of arc length
IC: for t 0
T Tm at r R
T 0 at r R
(2.31)
to radial length, s/r. BC: at r T 0 for t 0
For modeling purposes it is helpful to reduce
the number of coordinates from three to two by A uniform temperature could be added every-
idealizing the structure as one that is perfectly where to account for the ambient temperature
cylindrical. For the volcanic neck at Ship Rock before the development of the volcanic neck.
(Fig. 2.10c), we would say that it has a similar The natural symmetry of some geological
geometry for any horizontal cross section within structures motivates use of a coordinate system
several tens of meters of the current outcrop, so that bears some resemblance to the cylindrical
we can ignore spatial variations in geometry, system, but admits very eccentric shapes. The map
material properties, and boundary conditions of dike segments (Fig. 2.7) suggests that an ellipti-
with the axial coordinate z. Then we focus on the cal cross section may be a good approximation for
two-dimensional polar coordinates (r, ) and dene some of these. Segment 16, for example, is 136 m
the problem in terms of these two coordinates long, has a maximum thickness of 3.4 m, and is
(Fig. 2.10d). Radial lines of constant and circles approximately symmetric about planes that pass
of constant r form an orthogonal network cen- through its middle parallel to its length and thick-
tered on the origin. The contact between the ness (Fig. 2.12a). The contact between the minette
minette of the neck and the Mancos Shale is and the Mancos Shale is relatively smooth and
dened as a particular circle, r R15 m, with a slowly tapers toward the distal terminations. We
radius that best approximates that of the neck. have scant information about the extent or shape
The alternative would be to dene the contact in of this segment with depth, or with height above
terms of the Cartesian coordinates as the circle (x2 the current outcrop. Unlike the necks which
y2)1/2 R, but this is mathematically more cum- stand well above the surrounding shale, the dike
bersome. segments reveal little of their three-dimensional
Similarly, the boundary (BC) and initial (IC) geometry. Over the few meters of local relief the
conditions for models of the volcanic neck are contact maintains a near vertical dip and this is
simplied using polar rather than Cartesian coor- consistent over the 55 m of elevation change from
dinates. For example consider a model for the con- one end of the dike to the other. Where exposed
ductive heat ow from the hot magma into the the terminations of individual segments are
cold host rock. The cylindrical geometry of the vol- steeply plunging. Based on this limited informa-
canic neck dictates a temperature eld that varies tion we adopt two-dimensional elliptical coordi-
spatially only in the radial coordinate direction, nates, which may be visualized as a set of confocal
thereby reducing the problem to one spatial ellipses and hyperbolae (Fig. 2.12b).
dimension. Because the governing equation for For reference consider the Cartesian axes (Ox,
heat conduction is written with temperature as Oy, Oz) with Oz perpendicular to the plane of inter-
the dependent variable, the initial conditions typ- est. We postulate that the dike segment has a
ically constrain the temperature eld at some similar geometry for any horizontal cross section
specied time and the boundary conditions con- within several tens of meters of the current
strain the temperature eld at specied locations outcrop, so we can ignore spatial variations in
on the surface(s) of the body. Here we consider geometry, material properties, and boundary con-
initial conditions at the time t 0, such that the ditions with the coordinate z. Then the basic ele-
temperature is uniform, Tm, throughout the neck ments of the two-dimensional elliptical system
and zero throughout the host rock. Then let the are an origin, O, the line Ox, a coordinate , a coor-
temperature eld, T(r, t) evolve with time accord- dinate , and a focal length, f, measured from the
(a) The position vector, p, used to locate the point

P(, ) in the two-dimensional elliptical coordi-
nate system (Fig. 2.12b) is written as a linear com-
bination of the scalar components (2.32) and the
two mutually orthogonal unit base vectors (ex, ey)
of the Cartesian system:
p pxex pyey
(2.33)
( f cosh cos )ex ( f sinh sin )ey
The scalar components carry the units of length

(b) j = constant (meters) because f has units of length. Thus the
y focal length determines the length scale of this
P(j, h) coordinate system.
p For a given hyperbola ( constant), as
j = jO b
ranges from numbers much less than one to
numbers greater than one the ellipses range from
O x very eccentric to nearly circular. In the limit, as
0 the ellipse collapses down onto the x-axis,
f extending from one focus to the other. Later we
use such highly eccentric ellipses as models for
a fractures and faults in rock. For a given ellipse (
h = constant constant), as the angle ranges from 0 to 360,
points vary in position around the ellipse in a
Fig 2.12 (a) Aerial photograph of one segment of the
counterclockwise direction starting at the inter-
northeastern dike at Ship Rock (Delaney and Pollard, 1981).
(b) Elliptical coordinate system used in models of dike
section with the positive x-axis where 0. The
segments. angle 90 at the intersection of the ellipse and
the positive y-axis; 180 at the intersection
with the negative x-axis, and so forth.
origin along the line Ox (Fig. 2.12b). The ellipses
The semi-major axis, a, and semi-minor axis, b,
and hyperbolae are called confocal because they
of a particular ellipse ( constant) are:
share common foci located at x f.
Given an arbitrary point P(, ) in elliptical a f cosh , b f sinh (2.34)
coordinates, the Cartesian coordinates are found
For segment 16 of the Ship Rock dike (Fig. 2.12a)
using the following transformation equations
we have a 68 m and b 1.7 m, so the particular
(Timoshenko and Goodier, 1970):
ellipse, O, dening the contact is:
x f cosh cos , y f sinh sin (2.32)
The functions cosh and sinh are hyperbolic func-

tions. Just as in the polar coordinate system where
O tanh1
b
a
tanh1(0.025) 0.025 (2.35)
any circle centered at the origin would be repre- The specication of boundary conditions for prob-
sented by a constant value of r, each of the family lems related to dike segments is particularly
of confocal ellipses is represented by a constant simple using elliptical coordinates. For example,
value of . However, unlike the coordinate r, which if the magma pressure, pm, exerts traction, t, only
is a length measured in meters, the coordinate perpendicular to the contact, we write:
is just a number. Also, as each radial line in the
polar system is associated with a particular value
of the angle , each of the family of confocal
BC: on O
t pm
t 0
(2.36)
hyperbolae is associated with a particular value of Here t is the traction component acting on the
the angle . elliptical surface, O, in the -coordinate direction
(a) Ch 45 20 mm, Cv 43 20 mm,

(2.37)
Cd 45 23 mm
Here Ch and Cv are the horizontal and vertical

semi-axes of the cavities measured from the
center to the periphery in the plane of the inter-
secting joint, and Cd is the horizontal semi-axis
measured perpendicular to this plane. Although
the mean values are nearly identical for the three
semi-axes, the standard deviations indicate that
(b) the semi-axes can be quite different for any par-
ticular cavity. None-the-less the cavities may be
idealized as spherical in shape as a rst approxi-
mation in the procedure of modeling their
mechanical behavior.
Again for reference we consider the Cartesian
axes (Ox, Oy, Oz). The basic elements of the spheri-
cal system are an origin, O, a radial distance, ,
measured from the origin, an angle measured
from the line Oz, and an angle, , measured from
1m the line Ox (Fig. 2.11b). The angle is measured in
the plane dened by the Oz-axis and the position
Fig 2.13 Cavities that served as nucleation sites for joints
vector for the point in question and is the smaller
in the Soreq Dolomite, Israel. (a) Photograph of cliff face
with joints and cavities. (b) Map of outcrop. Reprinted from
of the two angles between these directions in that
Weinberger (2001) with permission from Elsevier. plane. The angle is measured in the (x, y)-plane
and is positive if clockwise when looking in the
positive direction of Oz. Curves of constant are
and t is the traction component acting on that analogous to lines of latitude and curves of con-
surface in the -coordinate direction. stant are analogous to lines of longitude.
The last coordinate system we consider here is Together these curves form an orthogonal
motivated by geological structures that are network on any sphere of constant radius .
approximately spherical in shape. A good example Given an arbitrary point P(, , ) in spherical
comes from the Soreq Formation, a stack of coordinates, the Cartesian coordinates are found
dolomite layers separated by thin marls that crops using the following transformation equations
out in central Israel and contains two systematic (Selby, 1975, p. 385):
sets of joints (Weinberger, 2001). The west-striking
joint set is older than the northnorthwest-
x cos sin , y sin sin, z cos
striking set and is spatially associated with cavi-
ties that apparently are dissolved anhydrite
nodules (Fig. 2.13). The surfaces of joints of this set
x2 y2 z2, cos 1 z
x2 y2 z2,
are decorated with both plumose structures and

rib markings (Hodgson, 1961; DeGraff and Aydin,
1987; Aydin and DeGraff, 1988; Bahat, 1991) indi-
tan 1
y
x
(2.38)
cating that each joint nucleated at a particular The second and third lines contain the inverse
cavity and propagated vertically and laterally transformation equations. The position vector, p,
away from it. The three semi-axes of 14 cavities in used to locate an arbitrary point P(, , ) in spher-
a single layer were measured and provide the fol- ical coordinates (Fig. 2.11b) is written using the
lowing statistics: rst line of (2.38) as:
p pxex pyey pzez The direction cosines of (2.28) are referred to as

mij where both indices i and j range from 1 to 3, so
( cos sin )ex ( sin sin )ey ( cos )ez
the double index implies nine quantities that can
(2.39) be arranged in a table as:
In other words p is a linear combination of the
x1 x2 x3
Cartesian coordinates from (2.38) and the base
x1 m11 m12 m13
vectors (ex, ey, ez). (2.41)
x2 m21 m22 m23
Points on a particular sphere, such as the
x3 m31 m32 m33
perimeter of one of the cavities in the Soreq
Formation (Fig. 2.13) are identied by specifying Note that the rst subscript on the direction
O 45 mm, where O is the cavity radius. If the cosines refers to the reference (old) axis and the
cavities contained no uid at the time of joint for- second subscript refers to the transformed (new)
mation, the traction, t, acting on the perimeter axis. Thus, one uses m31 for the orthogonal pro-
was identically zero and the boundary conditions jection of the x3-coordinate of a point onto the x1-
are written: axis. Similarly, one uses m23 for the orthogonal
BC: on O, t t t 0 (2.40) projection of the x2-coordinate of a point onto the
x3-axis.
The three traction components act on the surface Additional efciency is gained in writing equa-
of the spherical cavity in the directions of increas- tions with indicial notation by introducing the
ing coordinates , , and , respectively. so-called summation convention. This convention
establishes that a repeated subscript in any given
2.2.3 Indicial and matrix notations and term of an equation implies summation with
equations respect to that index over its range. Using (2.41)
The systematics of (2.28) motivate a notation, the equations for the rotational transformation
referred to as indicial notation, that greatly reduces from old coordinate, xi, to new coordinates, xi , are
the effort of writing the rotational transforma- written:
tion equations and facilitates recalling them from
3
memory. This notation also applies to many of the
equations from continuum mechanics that are
x1 m11x1 m21x2 m31x3 m
i1
i1xi mi1xi
3
used for a variety of different purposes in struc-
tural geology, so it is important to understand
x2 m12x1 m22x2 m32x3 m
i1
i2xi mi2xi (2.42)
how to read, formulate, and decompose equations 3
written using indicial notation. Some of the basic x3 m13x1 m23x2 m33x3 m
i1
i3xi mi3xi
elements of indicial notation are introduced

here as they specically relate to coordinate trans- In the last step of each equation the summation
formations; others are introduced as needed symbol is replaced by the summation convention.
throughout the text. Recognizing the pattern of subscripts in these
Rather than naming the Cartesian coordinates equations, one may compose the forward and
(x, y, z) they are named (x1, x2, x3) in which a single inverse rotational transformation equations as:
letter, x, is used with subscripts 1, 2, and 3 to dis-
xj mij xi, xi mijxj (2.43)
tinguish the three axes. Although this appears to
have made the notation more cumbersome, A repeated subscript in any term is called a dummy
efciency is achieved in that one can refer to the index: any letter may be used for this subscript
three coordinates simply as xi where it is under- because it only functions to inform the reader
stood that i ranges from 1 to 3. In this context the that summation is required over the specied
subscript i is called an index. Similarly, one can range. In contrast any non-repeated subscript in a
refer to the three basis vectors as ei and the three given term is called a free index because the reader
components of the position vector as pi. is free to choose any particular value within the
specied range. Because there is one free index as a table or array of numbers motivates consider-
which ranges from 1 to 3 in each of these equa- ation of the concept and mathematical properties
tions, each may be expanded into the three of a matrix. This is further motivated by the fact
equations. The forward and inverse rotational that the computational engine, MATLAB, used for
transformations for the components of the old the exercises and many of the graphical illustra-
and new position vectors ( pi, pj) and the old and tions in this text, treats all data sets as arrays of
new basis vectors (ei, ej) are constructed similarly. numbers and offers many useful functions that
The denition of the direction cosines in operate on matrices. Furthermore, many of the
terms of the base vectors is written using indicial constructs of continuum mechanics can be
notation and the scalar product (2.19) as: described and manipulated as matrices. As we
have just done with indicial notation, some of the
mij ei ej (2.44)
basic concepts of matrices are introduced here as
Here there are no repeated indices but both i and they specically relate to coordinate transforma-
j range from 1 to 3, so this equation expands into tions; others are introduced as needed through-
nine equations for the direction cosines. The out the text.
scalar product of two arbitrary vectors, v and w A matrix is a rectangular array of numbers
(Fig. 2.9b), is expressed using indicial notation as: with each element of the array designated by its
position in the array according to the row and
v w v i wi (2.45)
column number: an m by n matrix has m rows and
Here the repeated index requires summation of n columns. For example, consider the position
the product of the components over the range 1 to vectors that were used to describe the contact of
3. the dike at Ship Rock (Fig. 2.5). In general, a posi-
The facts that the base vectors are of unit mag- tion vector, p, is written using indicial notation
nitude and orthogonal to one another are and in expanded form as follows:
expressed by the two equations:
p piei p1e1 p2e2 p3e3 (2.50)
e1 e1 e2 e2 e3 e3 1
(2.46) The three components (p1, p2, p3) of the position
e1 e2 e2 e3 e3 e1 0
vector can be thought of as either a 1 by 3 row
These conditions motivate the denition of a new matrix or a 3 by 1 column matrix:
quantity, ij called the Kronecker delta that nds
considerable usage with indicial notation:

P1

1, if i j P [P1 P2 P3 ], or P P2 (2.51)
ij , for (i, j 1, 2, 3) (2.47)
0, if i j P3
The conditions of unit magnitude and orthogonal Notice that the single subscript for each vector
orientations (2.46) are succinctly written using component, pi, is replaced by a double subscript
the Kronecker delta: and that these subscripts refer, respectively, to the
row number and column number. There are no
ei ej ij (2.48)
restrictions on the number of elements in a row
Furthermore, the conditions that the squares of or column matrix, or on the relationships among
direction cosines in each row and each column of those elements, although our example happens to
(2.41) sum to one are written: use three elements that are components of a posi-
tion vector. Sometimes row and column matrices
mkimkj ij, mikmjk ij (2.49)
are referred to as vectors, but we restrict that
The rst equation applies to the direction cosines term to quantities in which the elements have the
in each column (sum over the rst index) and the properties of vectors.
second equation applies to each row (sum over the As a second example consider the set of
second index). direction cosines used to relate the old and new
The arrangement of direction cosines in (2.41) basis vectors for a rotational transformation of
coordinates (2.41). Using indicial notation and in the two matrices and the sequence in which they
expanded form these direction cosines are can be multiplied. As an example, consider the 2
written: by 3 matrix, Q, and the 3 by 1 matrix, R:
m11 m12 m13

R1
mij m21
m31
m22
m32
m23
m33
(2.52)
Q Q1
Q1
Q2
Q2
Q3
Q3
, R R2
R3
(2.56)
The matrix representation of the direction

These may be multiplied in the sequence QR and
cosines follows quite naturally because the row
the multiplication is carried out as though each
and column numbers of the matrix, M, are the
row of Q and the column of R are vectors and one
same as the indices of mij:
wishes to form their scalar products:

M11 M12 M13
M M21
M31
M22
M32
M23
M33
(2.53) QR Q 11R11 Q 12R 21 Q 13R 31
Q 21R11 Q 22R 21 Q 23R 31
S (2.57)
The result is the matrix S of order 2 by 1: a matrix

This is a 3 by 3 matrix and any matrix where the with the same number of rows as the rst matrix,
number of rows and columns are identical, m n, Q, and the same number of columns as the second
is termed a square matrix. The number of rows and matrix, R. The two matrices, Q and R, cannot be
columns is referred to as the order of a matrix multiplied in the sequence RQ.
and, in general, there are no restrictions on the Now consider the general case of a matrix Q of
order of a matrix. order m by l and a matrix R of order l by n. The mul-
Just as two vectors are added (or subtracted) tiplication of these two matrices can be symbol-
component by component (2.12) two matrices are ized as follows:
added (or subtracted) element by element. For
example, the addition of two 3 by 2 matrices, Q QR [m by l][l by n] [m by n] S (2.58)
and R, is carried out as: Note that the order of the resulting matrix S is m
by n. In general, two matrices can be multiplied in

Q 11 R 11 Q 12 R 12
the sequence QR if the number of columns of Q is
Q R Q 21 R21 Q 22 R 22 (2.54)
equal to the number of rows of R. Indicial notation
Q 31 R31 Q 32 R 32
provides a succinct way to describe the elements
This results in a row matrix of exactly the same of the matrix S (Malvern, 1969, p. 41):
order, 3 by 2. The addition (or subtraction) of

matrices requires that they be of the same order. i 1, . . . , m
The 3 by 2 matrix R is scaled by a constant Sij Q ik Rkj, for j 1, . . . , n (2.59)
numerical factor as follows: k 1, . . . , l
Here Sij is the element in the ith row and jth

kR11 kR12
kR kR21 kR22 column of the matrix S. Because of the repeated
(2.55)
kR31 kR32
index k, each element is the sum of l terms. If the
number of rows of Q is the same as the number of
In general, the multiplication of a matrix by a con- columns of R, m n, the multiplication can
stant is accomplished by multiplying each proceed in the reverse sequence, RQ, but the two
element of the matrix by that constant. The divi- products are not equal, QR RQ.
sion of a matrix by a constant is equivalent to mul- Multiplication with square matrices is com-
tiplying by the reciprocal of that constant. monly encountered in applications to physical
The multiplication of two matrices follows problems and they have special properties that can
rules that are similar, in part, to those for the be illustrated using the rotational transformation
scalar product of two vectors (2.19), but there are equations. Recall the table (2.41) that relates the
important distinctions that restrict the order of old, xi, and new, xj, Cartesian coordinates using the
direction cosines, mij. These three quantities may describe the structures underfoot. Most geological
be written as two column matrices and one square structures may be idealized as three-dimensional
matrix: curved surfaces or curved lines. Examples are sur-
faces that truncate older formations such as a

X1 X1 M11 M12 M13 fault offsetting sedimentary bedding, or a dike
X X2 , X X2 , M M21 M22 M23 (2.60) that cuts across an igneous contact. Or there may
X3 X3 M31 M32 M33 be surfaces within a mass of rock dened by the
alignment of platy minerals, as in an igneous or
The inverse rotational transformation, xi mijxj, metamorphic foliation. Curvilinear structures
may be constructed directly by matrix multiplica- also may be composed of aligned mineral grains,
tion, X MX, because the number of columns of as in a metamorphic lineation. The intersection of
M is the same as the number of rows of X, and the two curved surfaces, for example the intersection
summation required by the indicial notation is of two faults, would dene a curved linear struc-
consistent with that of matrix multiplication: ture. Regardless of the specic nature of these
curved surfaces and lines we need techniques for

X1 M11 M12 M13 X1
measuring their orientations in the eld and
X2 M21 M22 M23 X2 for recording these orientations on a map at the
X3 M31 M32 M33 X3 position determined by the UTM coordinates.
(2.61)

M11X11 M12X21 M13X31 Techniques are introduced here along with a pro-
M21X11 M22X21 M23X31 jection that is useful for visualizing the relative
M31X11 M32X21 M33X31 orientations of such structures.
However, the forward rotational transformation, 2.3.1 Orientations of linear and planar
xj mijxi, is not represented as X MX, despite structural elements
the fact that the number of columns of M is the Most curved structural surfaces may be approxi-
same as the number of rows of X, because the sum- mated locally by a planar element that is tangential
mation implied by the indicial notation is not to the surface at the point of measurement.
consistent with matrix multiplication. Instead Similarly, most curvilinear structures may be
one must rst take the transpose of M, symbolized approximated locally by a linear element that is tan-
as M T, and then compute the product as X M TX: gential to the curve at the point of measurement.
What are actually recorded by the structural geol-

X 1 M11 M21 M31 X1 ogist at an exposure are the orientations of these
X 2 M12 M22 M32 X2 structural elements. The exposure photographs in
X 3 M13 M23 M33 X3 Fig. 2.14 show a number of geological structures

M11X11 M21X21 M31X31 that can be approximated in this way with planar
M12 X11 M22 X21 M32 X31 (2.62)
and linear elements. These exposures are from the
M13 X11 M23 X21 M33 X31 northern part of the San Rafael Swell in the
Colorado Plateau province of central Utah (Kelly,
Note that the transpose, M T, is found by inter- 1955). In Fig. 2.14a a member of the Chimney Rock
changing the rows and columns of M. fault system juxtaposes beds of limestone, silt-
stone, and mudstone of the Middle Jurassic
Carmel Formation (to the left) against the massive
2.3 Orientations of structural Jurassic Navajo Sandstone (to the right). Because
the Carmel Formation immediately overlies the
elements Navajo Sandstone in the normal stratigraphic
sequence, we deduce that the Carmel Formation
Given the complete UTM geographic coordinates exposed in this photograph has moved downward
for the position of a particular outcrop, the next on the fault relative to the Navajo Sandstone.
step in most structural studies is to measure and The fault pictured in Fig. 2.14a is a steeply
2.3 ORIENTATIONS OF STRUCTURAL ELEMENTS 53
(a) is not a good approximation. Many such elements

positioned along the exposed trace of the fault
could provide a good representation at the scale
of the trace length. For smaller areas at this expo-
sure there is a roughness associated with undula-
tions and small steps on the exposed surfaces
within the fault zone (Fig. 2.14b). To capture this
roughness a map at the outcrop scale with many
Fault
rmation
Carmel Fo surface Navajo individual planar elements properly positioned
Sandstone and oriented, could provide a good representa-
tion. For still smaller areas individual grains of
sand provide a roughness that precludes approxi-
mation as a single planar element. Similarly, beds
of the Carmel Formation appear to be roughly
planar at a scale of several square meters and are
gently inclined toward the left-hand side of the
photograph (Fig. 2.14a). They too have undulations
at both larger and smaller scales that would
(b) require a redenition of the size of the planar
element. For our purposes the orientations of the
fault and the beds at this outcrop can be repre-
sented by the orientations of planar elements that
locally approximate these structures at the scale
of about a square meter.
On many exposures of the Navajo Sandstone,
where a member of the Chimney Rock fault
system cuts it, curvilinear structures are visible
called slickenlines or slickensides (Fig. 2.14b). These
line
Fault surface structures apparently resulted from the frictional

Slicken
sliding of the two rock masses over one another

along the fault. If this is a correct interpretation
the slickenlines trend in the direction of relative
motion. Although the slickenlines are gently
curved in detail, we can approximate them with
linear elements with lengths from a few centime-
ters to several decimeters. The orientations of the
Fig 2.14 Outcrop in Navajo Sandstone and Carmel slickenlines at this exposure are represented by
Formation near Chimney Rock, UT (Maerten et al., 2001). the orientations of these linear elements. Because
(a) Fault surface that can be approximated locally using a there is a range of orientations over the exposure,
planar structural element. (b) Slickenlines that can be the question being addressed might require a
approximated using a linear element. Photograph by D. D.
detailed map of the surface with many distinct
Pollard.
orientations at different locations on the surface.
For our purposes we will take a single (average)
inclined and somewhat undulating tabular zone orientation to represent the population.
of deformed rock. Within that zone are surfaces The orientations of the planar and linear ele-
that can be approximated as planar over areas of ments that approximate geologic structures in
a few square meters. If we were to consider a much outcrop are dened relative to a local geographic
larger area, we would nd that the fault bends coordinate system composed of east, north, and
and even is segmented, so a single planar element up (Fig. 2.15). Structural geologists use a small
hand-held instrument that combines a compass (a) Up

r th as
and an inclinometer with bubble levels to No
measure the necessary angles between the planar
tion
or linear elements and this coordinate system.
e direc
Instructions for the use of such instruments are Strik East
Dip dir
o f ection
found in books on geological eld methods Line
(Compton, 1962; Davis and Reynolds, 1996) and strike
most students will be introduced to their use at Horizontal fd
geological eld camps. For the local coordinate plane Planar element
system up is determined as the opposite direc- that locally
tion to the local gravitational acceleration using approximates
a structure
the bubble levels, and this determines the hori-
zontal plane that contains north and east. North (b) r th ap
No Plunge
is measured with the compass (corrected for the Up
direction
magnetic declination) and east is 90 clockwise East
from north when looking down on the horizontal Plunge,
plane. fp
Consider rst the planar element that approx-
imates the fault zone at the outcrop shown in Fig.
2.14a. Imagine a small lake lapping up against the
Vertical Linear element that locally
fault. Because the water surface is a geoid (every-
plane approximates a structure
where perpendicular to the down direction), the
intersection of the water and the planar element
denes a horizontal line that is referred to as the (c) Up
line of strike. Of course this line, and any line for r th
No
that matter, points in two directions, so we must
e
establish a convention to specify which of the two Strik n
tio
directions to measure. The convention is one of direc
East
many so-called right-hand rules that are used in Rake, ur
structural geology. Position yourself on the struc-
ture in the eld (or imagine positioning yourself Planar
element
on the planar element that approximates that
structure) and look along the line of strike such Linear element
that the structure (planar element) slopes down to contained in
your right. The dip direction is the direction the planar element
outstretched ngers of your right hand would
Fig 2.15 Diagrams used to define orientations of
point if you raised your right arm to the horizon- structural elements. Different azimuth angles, , and angles
tal plane (Fig. 2.15a). In other words the strike of inclination, , are distinguished by appropriate subscripts.
direction is along the line of strike such that the (a) Strike and dip of a planar element. (b) Plunge direction
dip direction is to the right. The strike and dip and plunge of a linear element. (c) Rake of a linear element in
directions are orthogonal to one another in the a planar element.
horizontal plane.
Either the azimuth of the strike direction or
the azimuth of the dip direction is used to relate 270 (west) to just less than 360, and it is a con-
these directions to the local geographic coordi- vention to use three digits when specifying any
nate system. The azimuth is the clockwise angle, , azimuth. The azimuth of the strike direction, s,
looking down on the horizontal plane, measured is the strike of the planar element that approxi-
from north to that direction. Thus, azimuths vary mates the structure (Fig. 2.15a). In casual conver-
from 000 (north) to 090 (east) to 180 (south) to sation this would be referred to as the strike of the
structure. For the example illustrated in Fig. 2.14a element using strike and dip. The relations
we would refer to the strike of the fault, and the between dip direction and strike direction are:
fact that the fault is not simply a plane in space
d s 90, 0 d 90 (2.63)
would be left unspoken. To understand what
someone means by the strike of a fault, both Two special cases are noteworthy. For a horizontal
the nature of the surface within the fault zone planar element the dip is zero, so the strike and
and the size of the planar element should be dip direction are undened and we would write
identied. (UDF, 00) for the strike (or dip) direction and dip:
Knowing the strike of a planar element
s UDF d, d 0 (2.64)
enables one to identify the orientation of the line
of intersection of that element with a horizontal For a vertical planar element the dip is 90, but
plane. There are an innite number of elements there are two possible strike (and dip) directions,
that share the same strike, but vary in inclination, and either one is suitable. For example, the strike
so we have to dene a second angle that is a and dip of a vertical plane with line of strike
measure of this inclination in order to determine recorded in the eld as (136, 90) could equally well
uniquely the orientation of a particular planar be recorded as (316, 90).
element relative to the local geographic coordi- Next consider the orientations of linear ele-
nate system. This angle, d, is measured in a verti- ments that approximate structures such as the
cal plane that contains the dip direction, from the slickenlines in Fig. 2.14b. These also are dened
dip direction to the planar element in question with respect to the local geographic coordinate
(Fig. 2.15a). This angle is referred to as the dip of system composed of east, north, and up (Fig.
the planar element that approximates the struc- 2.15b). The rst step is to imagine a vertical plane
ture, and by convention two digits are used to that contains the linear element. Position your-
specify the dip. Thus, a dip of seven degrees would self over the structure at the exposure (or imagine
be written 07. Note that the angle measured from positioning yourself over the line segment) such
the horizontal plane to the planar element in ver- that the structure (line segment) is inclined
tical planes of other orientations would be downward in front of you. The plunge direction is
smaller than the dip. Such an angle is referred to the direction of your view in the horizontal plane.
as an apparent dip. Apparent dips are commonly The azimuth of the plunge direction, p, is some-
observed and measured in the eld where expo- times referred to as the trend of the linear
sures cut obliquely across structures. element and it is specied using three digits. The
Only two angles are necessary to reference a angle p measured in the vertical plane from the
planar element to the local geographic coordinate plunge direction down to the linear element is
system. Some use the strike and dip (s, d), dened as the plunge and it is specied using two
whereas others use the dip direction and dip (d, digits.
d). Because of the construction of some eld Two angles, the plunge direction and the
instruments it may be more convenient to plunge (p, p) are necessary to reference a linear
measure and record the strike. On the other element to the local geographic coordinate
hand, the dip direction and dip require the system and these would be recorded, for example,
identication of only one direction and that direc- as (356, 58) indicating a linear trend just a few
tion corresponds more directly to the inclination degrees west of north and plunging 58 in that
of the planar element that is measured. If the dip direction. Again, two special cases are noteworthy.
direction and dip were used, the number pair For a vertical linear element the plunge is 90, so
(085, 37) would indicate a planar element with a the plunge direction is undened:
dip direction just 5 to the north of east and an
p UDF, p 90 (2.65)
inclination of 37 in that same direction. The
degree symbol is left off for recording conve- This line would be recorded as (UDF, 90). For a hor-
nience. The number pair (355, 37), plus the right- izontal element the plunge is 0, so there are two
hand rule, would identify the same planar possible plunge directions and either one can be
used. Thus, the same horizontal linear element linear elements. For example, note the different
could be recorded as (022, 00) or (202, 00). symbols for joints and veins or dikes. In this way a
Some curvilinear structures lie in curved sur- lot of information is conveyed in a compact form
faces that are well-dened structures themselves. on the map, and structures can be related to one
For example, the slickenlines shown in Fig. 2.14b are another in terms of their locations.
in the surface of a fault. For these cases the strike
and dip of the planar element that represents the 2.3.2 Stereographic projection of
surface are recorded, along with an angle known as structural elements
the rake that can be measured on the exposure with It is useful to have a graphical means to visualize
a protractor (Fig. 2.15c). The rake is the angle, r, the attitudes of planar and linear structural ele-
measured in the plane of the element from the ments. The most effective tools for this purpose
strike direction down to the linear element. It is are a family of projections that create an image of
recorded using three digits and can vary from 000 the elements on a at piece of paper. The locations
(the linear element is parallel to the line of strike of of the structures are not recorded in this image,
the planar element) to just less than 180: but their orientations relative to the geographic
coordinate system are recorded. Here we intro-
000 r 180 (2.66)
duce one of these projections, the so-called stereo-
For r 090 the linear element is inclined graphic projection. Details concerning the use of
directly down the dip of the planar element. Of this projection, and other members of this family
course one could measure the plunge direction of projections, can be found in books devoted to
and plunge of this line, but often it is simpler to the subject (Phillips, 1954; Ragan, 1985; Marshak
measure the strike and dip of the planar element and Mitra, 1988). In some courses in structural
and the rake angle in this planar element. geology much of the students time is committed
Structural geologists refer to the attitude of a to the manipulation of these projections by hand
structure and by that they mean the orientation in and many of the geometric problems encountered
space, relative to the local geographic coordinate in eldwork are described in these reference
system, of the planar or linear element that works. Here we adopt the more analytical
approximates (is tangential to) the structure at the approach described by Goodman and Shi (1985,
point of measurement. Thus, the attitude of a p. 56) that avoids the tedium and inaccuracy of
fault at a particular location would be recorded as hand constructions. We also take advantage of the
the strike and dip, or the dip direction and dip. The visualization power of modern computer applica-
attitude of a slickenline on that fault would be tions for plotting quantitative eld data.
recorded as the plunge direction and the plunge. We begin by reviewing the basic concepts and
These measurements are represented on maps present the analytical expressions necessary to
using symbols and numbers placed at the appro- plot stereographic projections of planar and
priate location. Some of the symbols used on struc- linear structural elements. Consider a linear
tural maps are illustrated in Fig. 2.16, extracted element xed in space at the center, C, of a trans-
from a more extensive table of symbols in a parent sphere called the reference sphere (Fig. 2.17a).
manual of eld geology (Compton, 1962). For most Only the part of the linear element extending
of these symbols, longer line segments are drawn from C to the point P on the sphere is shown.
parallel to the strike direction so the azimuth can Points such as P are projected onto the equatorial
be determined with reference to the north direc- plane of this sphere and the intersection of the
tion on the map. Shorter line segments indicate sphere and this plane is called the reference circle.
the dip direction and arrows indicate the plunge Points on the reference circle represent the four
direction. Numbers set near the shorter line seg- compass directions (north, east, south, and west),
ments record the dip or plunge angle. The style of and the axis perpendicular to the equatorial plane
the line segments is used to distinguish different intersects the top of the sphere at the zenith, Z. In
structures that are approximated as planar or the view shown in this gure the sphere is rotated
40
Contact, showing dip 90 Strike of vertical bedding;
stratigraphic tops to north (right)
90
Contact, vertical (left) and
63 overturned (right) Horizontal bedding
72 90 Fault, showing dips Strike and dip of foliations
Fault, showing trend and plunge Strike of vertical foliations

75 D
U of linear features (D, down-
thrown side; U, up-thrown side)
Horizontal foliations
Fault, showing relative
horizontal movement Strike and dip of joints
(left) and veins or dikes (right)
T Thrust faults; T
or saw-teeth in upper plate Strike of vertical joints (left)
and veins or dikes (right)
Normal fault;
hachures on down-thrown side Horizontal joints
(left) and veins or dikes (right)
10
Anticline, showing trace of axial
plane and plunge of axis 30 Trend and plunge of lineation
31
Syncline, showing trace of axial
90 Vertical lineation
plane and plunge of axis
10 Strike and dip of bedding Trend of horizontal lineation
70 Strike and dip of

overturned bedding
Fig 2.16 Selection of symbols used on structural maps to

only the intersection of the linear element with
indicate type of structure and orientation (Compton, 1962).
the lower hemisphere is projected. The point
where the linear element would intersect the
somewhat about the eastwest axis, so the upper upper hemisphere projects to a point on the equa-
surface of the equatorial plane inside the refer- torial plane that is outside the reference circle.
ence circle is visible as a shaded elliptical surface. The scales for plotting the point representing
Thus, the zenith has moved toward the observer the linear element are provided by the meridional
and appears somewhat below what appears as the stereographic net, sometimes called the Wulff net
top of the sphere. The zenith is chosen as the per- after G. V. Wulff who published a version in 1902,
spective point for the stereographic projections or the equal angle net, or the stereonet (Phillips,
described here. One also may choose the bottom 1954). Later in this section we show how to con-
of the reference sphere as the perspective point. struct the stereonet itself, but we start with a
The intersection of the linear element with the simpler construction, which provides the tech-
lower hemisphere is a point, P, and this point is nique for plotting linear elements. An example of
projected to the point P on the equatorial plane a stereonet is given in Fig. 2.17b where we note
along the line ZP from the zenith. By convention that the net is composed of a family of arcs of
(a) N
Reference (b)
Projection Z sphere Stereonet
of point
Equatorial
N, y plane Great
circle
W C E, x
P W E
Reference Reference
S o circle
circle 33
Small =
Point, P circle Fp
Linear
element ap = 222o
S
N, y
(c) Z (d)
Equatorial Reference
Vertical circle
plane plane
y g
R
P Plunge C
direction W E, x
C Fp y ap
R P x
P
re ge
n
Trace of projection plane

io
S
di lun
ct
P
Fig 2.17 Stereographic projection of linear element.

(a) Reference sphere. (b) Meridional stereographic projection direction and plunge of (222, 33) is plotted on the
or equal angle net. (c) Vertical plane passing through center stereonet in Fig. 2.17b. Plotting the point repre-
of reference sphere and containing linear element. senting the element by hand is accomplished on a
(d) Equatorial plane. sheet of transparent material pinned through the
center of the net and marked with the north direc-
tion at the top of the net. The overlay is rotated
great circles (meridians) running from north to counterclockwise through an angle equal to the
south and an orthogonal family of arcs of small azimuth of the plunge direction and the plunge
circles. Both the great and small circular arcs are angle is scaled off from the reference circle at
restricted to the interior of the reference circle. north toward the center using the small circles for
The plunge direction, p, of the linear element is a scale. On this net the small circles are drawn
measured from north clockwise around the refer- every 10. The point so identied is marked on the
ence circle. The plunge angle, p, is measured overlay and the overlay is rotated back to the orig-
from the reference circle toward the center to the inal orientation as shown in Fig. 2.17b.
appropriate point along the radial line that is Locating the point representing a linear
coincident with the plunge direction. Thus, linear element on a stereographic projection may be
elements with shallow plunges project as points done analytically, so a computer can plot the pro-
lying close to, but just inside the reference circle, jection rather than your hand. Consider a verti-
whereas steeply plunging linear elements project cal plane passing through the center of the
as points near the center, C. sphere and containing the plunge direction (Fig.
An example of a linear element with plunge 2.17c). The radius of the sphere is R. The angle
PCP is the plunge, p, and we dene the angle at the center, C, of the transparent reference
CZP as . The distance from the center, C, to the sphere (Fig. 2.18a). For the sake of an example we
projected point is: take the dip direction and dip as (118, 26). The
intersection of the planar element with the
CP R tan (2.67)
sphere is a so-called great circle, because it is a
Because CP CZ R, the triangle CZPC is isosceles circle and because this circle has the largest pos-
and the angle ZPC also is . Therefore, the two sible radius of all those formed by planes of this
angles are related as: orientation intersecting the sphere. In fact it has
the same radius as the reference sphere, R. A
1
45 p (2.68) planar element not passing through the center of
2
the sphere also intersects the sphere to form a
Substituting for in the previous equation we circle, but this is called a small circle, because it has
write the distance of the projected point from the a radius that is less than the radius of the refer-
center as: ence sphere. Only the intersection of the planar
element with the lower hemisphere is drawn and
1
CP R tan 45 p
2 (2.69) a straight line marks the intersection of this
element with the equatorial plane. The stereo-
CP varies from R to 0 as the plunge angle varies graphic projection is constructed by connecting
from 0 to 90. lines of sight from the zenith, Z, to points such as
The next step is to determine the coordinates G on the half great circle. The line ZG intersects
of the projected point, P, relative to a Cartesian the equatorial plane at the point G, which is the
coordinate system with center at C and the x-axis projection of the point G. All possible lines ZG
and y-axis positive toward east and north respec- from the zenith to the half great circle intersect
tively (Fig. 2.17d). The point representing the the equatorial plane along a circular arc, and this
linear element is located along the radial line in arc is the stereographic projection of the half
the plunge direction at the distance CP from the great circle. Note that the radius of this circular
center. In the Cartesian system the coordinates of arc is greater than the radius of the reference
the point P are related to the angle , measured circle, R, unless the dip of the planar element is
counterclockwise from Ox to the line CP, as: zero. In this special case the half great circle rep-
resenting the planar element is coincident with
x CPcos ,y CPsin (2.70)
the reference circle.
Furthermore, the angle is related to the plunge The line of strike, s 28, connects the end
direction, p, as 90 p, so the coordinates of points of the projected half great circle at the ref-
the point P are: erence circle (Fig. 2.18b). The strike direction is
that direction viewed along the line of strike with
x CPsin p,y CPcos p (2.71)
the trace of the circular arc to the right. Any
Utilizing (2.69) to substitute for the distance CP, azimuth, such as the strike direction, s, or the
we have: dip direction, d, is measured from north clock-
wise around the reference circle. The dip angle,
1

x R tan 45 p sin p
2
d, is measured from the reference circle to the
circular arc along a radial line in the equatorial
(2.72) plane that is coincident with the dip direction.
1

y R tan 45 p cos p
2
Thus, planar elements with shallow dips project
as nearly complete half circular arcs lying close to
These are the equations used to plot the projec- the reference circle, whereas steeply dipping
tion of a linear element on a stereonet of radius R, planar elements project as nearly straight lines
given the azimuth of plunge, p, and angle of approaching the line of strike. Constructing the
plunge, p. circular arc by hand is accomplished on a sheet
Next consider a planar element xed in space of transparent material pined through the center
n
tio
dir rike
ec
St
(a) Zenith, Z (b) N as
Equatorial
plane Projection
Equatorial Projection of half
plane of half great great circle
N circle
e
trik
C
s
W E W C E
of
G
Planar
e
ad
Lin
S element
f
d D
G G
Half great circle dir ip
Sphere ec
tio
n
S
N, y
(c) Z (d) (x, y)
Vertical
plane n Q R'
Q R R Dip C b a
direction d
Q P k R E, x
W
C C h C
D
fd P d ip
Trace of P f ire
Trace of d ctio
projection n
planar
element Projection
Equatorial of half
plane S great circle
Fig 2.18 Stereographic projection of planar element. puter. Consider a vertical plane passing through
(a) Reference sphere. (b) Projection of half great circle onto the center of the sphere, C, and containing the
equatorial plane. (c) Vertical plane passing through center dip direction (Fig. 2.18c). The horizontal line is the
of reference sphere and containing the dip direction.
trace of the projection (equatorial) plane and the
(d) Equatorial plane with projection of whole circle.
line segment PQ is the trace of the planar
element. The point P is the projection of P onto
of the stereonet and marked with the north the equatorial plane from a perspective at the
direction at the top of the net. The overlay is then zenith, Z, and the point Q is the projection of the
rotated counterclockwise through the angle of point Q. The points P and Q lie on the projected
strike and the dip is scaled off from the reference circle and the line segment PQ is a diameter of
circle at east toward the center using the great that circle. An arc of this circle is the stereo-
circles of the net for a scale. The great circle so graphic projection of the half great circle, so we
identied is traced on the overlay and the overlay seek the center and radius of this projected circle
is rotated back to the original orientation to in order to plot it.
produce the graphical representation of the The angle PCP is the dip, d, of the planar
planar element shown in Fig. 2.18b. element and the angle CZP is dened as . By the
To avoid the tedious and inaccurate process of same argument leading to (2.68), the two angles
hand construction we seek analytical expressions are related as 45 d/2. Using this relation
for plotting the circular arc representing a planar and the triangle CZPC, the angle ZPC is related to
element on a stereographic projection by com- the dip as:
1
angle ZPC 90 45 d (2.73) N
2
Because PQ is a diameter of the sphere, QZP is a

Stereonet
right angle, so QZPQ is a right triangle. If the line as = 064o
segment CZ bisects the hypothenuse of this right
triangle, then CP CZ and we dene this length
o
as R, which is the radius of the projected circle. 16
Furthermore, because the triangle CZPC is =1
ur E
isosceles, the angles ZPC and CZP are equal. P
Using this fact and the previous equation we have:
1
45 d angle CZC (2.74)
2
ad = 154o
Solving for the angle CZC we nd that this angle ap = 197o fd = 58o
must be equal to the angle of dip, d. The S
trigonometry of the right triangle CZCC provides fp = 50o
the distance, CC, from the center of the sphere to
Fig 2.19 Meridional stereographic net with projection of
the center of the projected circle and the radius of
linear element within planar element.
that circle, R:
R
CC R tan d, R (2.75)
cos d
The second step in each of these equations follows
As the dip of the planar element varies from 0 to from the fact that the angle is related to the dip
90, the distance CC varies from 0 to , and the direction, d, as 270 d. Substituting (2.75)
radius of the projected circle varies from R to . and (2.77) in (2.76) we have:
Thus, the projected arc of the half great circle
varies from being coincident with half the refer- x R tan d sin d (R cos d) cos
ence circle to being coincident with the straight (2.78)
y R tan d cos d (R cos d) sin
line of strike.
For plotting purposes a Cartesian coordinate These are the equations used to plot the projected
system is established with origin at the center, C, circle representing the orientation of a planar
of the equatorial plane and the x-axis and y-axis element on a stereonet of radius R, given the
are taken as positive toward east and north, azimuth of dip, d, and the angle of dip, d. To plot
respectively (Fig. 2.18d). The coordinates of points the whole circle one uses the range 0 2. To
on the projected circle are given by: restrict the plot to the circular arc lying within
the reference circle (Fig. 2.18d), in other words the
x h R cos
y k R sin 0 2 (2.76)
projection of the half great circle, the further con-
dition on the coordinates is (x2 y2)1/2 R.
Slickenlines lying in the plane of a fault (Fig.
Here (h, k) are the coordinates of the center, C, of 2.14) are idealized as linear elements contained
the projected circle and R is the radius. The point within planar elements on a stereographic pro-
C is at a distance CC on a radial line oriented at jection (Fig. 2.19). The linear element projects to
180 from the dip direction. The coordinates (h, k) the point P and falls on the great circle repre-
are related to the angle , measured counter- senting the planar element. The rake angle, r, is
clockwise from Ox to this radial line: measured from the point on the reference circle,
representing the strike direction of the planar
h CCcos CCsin d
element, along the great circle to the point P. The
(2.77)
k CCsin CCcos d example shown in Fig. 2.19 is for a linear element
N
Stereonet
ap = 298o
fp = 64o
P
W 90 o E
Q
ad = 118o
fd = 26o
S
Fig 2.20 Meridional stereographic net with projection of
linear element that is perpendicular to a planar element.
Fig 2.21 Meridional stereographic net plotted both inside

with a rake of 116 lying in a planar element with and outside the reference circle.
strike and dip of (064, 58).
Any plane can be oriented in three-dimensional
space by identifying the orientation of the per- The dip interval between successive great circles is
pendicular line, referred to as the normal to a arbitrary and is taken as d 10 for the con-
planar element. Consider, for example, the planar struction of this net. For plotting purposes the
element with strike and dip (028, 26) shown in Fig. number of great circles in each set is n (90/d)
2.20. To understand the relationship of the normal 1. For these two sets of great circles the general
to the other attributes of planar elements con- plotting equations for planar elements (2.78)
sider a line element that lies in the planar reduce to:
element and plunges with the same angle as the
dip. This line element is represented by the point x R tan d (R cos d) cos
Q on the stereonet whereas the normal plots at (2.80)
y (R cos d) sin
point P. These points lie along a straight line, ori-
ented in the dip direction that passes through the To plot great circles covering the full stereonet
center of the stereonet. The smaller of the two (with the exception of points near the zenith) one
angles between the normal and the linear element uses the range 0 2. To restrict the plot to
measured in this vertical plane is 90. The point P the interior of the reference circle the further con-
sometimes is referred to as the pole of a planar dition on the coordinates is (x2 y2)1/2 R.
element. There are two sets of small circles on the
The nal topic in our discussion of stereo- meridional stereonet. In general any small circle
graphic projections is the construction of the on the stereonet can be thought of as the projec-
meridional stereographic net itself (Fig. 2.21). The tion of the intersection of a cone with the refer-
net is composed of two sets of great circles repre- ence sphere (Goodman and Shi, 1985, p. 71). The
senting the projections of planes with common apex of the cone is at the center of the sphere and
dip directions either to the east or to the west, and the cone itself can be generated by a set of lines
dip angles between 0 and 90: that make a common angle, , with a vector, Rv/v
that extends from the center to the perimeter of
d 90, 0 d 90 and
the reference sphere (Fig. 2.22). The vector is nor-
d 270, 0 d 90 (2.79) malized by dividing each component by the vector
Rv/v
Equatorial Z, z
plane v
P
W
S C
P N, y
E, x
Reference
sphere
A
Fig 2.22 Reference sphere with arbitrary vector v and

normalized vector Rv/v extending from center to perimeter.
magnitude, v. This produces a unit vector, which

then is scaled by the radius of the sphere, R. This
normalized vector is related to its components as: Fig 2.23 Lambert equal area projection or Schmidt net.
Rv
v

Rvx
v x
e
Rvy
v y
e
Rvz
v z
e (2.81)
these require more elaborate techniques for plot-
The radius of the small circle, R, and the coordi- ting structural data on stereonets. Most of these
nates of the center of the small circle (h, k) are graphical constructions can be derived from the
(Goodman and Shi, 1985, p. 75): elementary concepts and procedures introduced
here and therefore can be implemented on a com-
puter rather than a piece of paper. For example,
R sin (Rvx v)
R , h , common problems include determining true dip
(Rvz v) cos (Rvz v) cos
from apparent dip of a planar element, and deter-
(Rvy v) mining the orientation of the line of intersection
k (2.82)
(Rvz v) cos of two planar elements (Marshak and Mitra, 1988,
Chapter 5). An example of a more complicated
Substituting these equations into (2.76) we nd
problem is based on the fact that a lineation in a
the coordinates of points on a projected small
bedding surfaces within a cylindrical fold follows
circle centered on the vector Rv/v with apical
a path along a small circle as the surface is
angle 2.
unfolded about a horizontal fold axis (Marshak
For the two sets of small circles on the merid-
and Mitra, 1988, p. 119). For additional coverage of
ional stereonet, all of the vectors are directed
these and other graphical constructions using the
north or south, along the y-axis, so Rvy/v R and
stereonet, one should refer to specialized books
Rvx/v 0 Rvz/v. Therefore the coordinates of
on geometric techniques (Phillips, 1954; Ragan,
points on these small circles are:
1985; Marshak and Mitra, 1988).
x R tan cos
(2.83) 2.3.3 Equal area projection and graphical
y (R cos ) R tan sin
orientation statistics
The angular interval between successive small Projections other than the stereographic projec-
circles is arbitrary and is taken as 10 for the tion have been invoked to solve important geo-
construction of this net (Fig. 2.21). metrical problems in structural geology. For
There are many applications for stereographic example, the Lambert equal area projection is associ-
projections in structural geology, and some of ated with the so-called Schmidt net (Fig. 2.23), which
(a) (b) statistics of structural elements that dene a rock

fabric. Fabric in this context refers to the internal
arrangement of the physical constituents that
make up the rock mass. For example, planar
fabrics in metamorphic rocks (Turner and Weiss,
1963, p. 97) can be composed of layers of different
rock types (Fig. 2.24a), a set of sub-parallel frac-
(c) (d) tures (Fig. 2.24b), or a set of similarly oriented
platy mineral grains (Fig. 2.24c). In metamorphic
rocks such structures are referred to as a meta-
morphic foliation, but planar fabrics can be found
in sedimentary and igneous rocks. In these illus-
trations there is little doubt that the con-
stituents are arranged in a very orderly manner
such that the normals to the different layers or
(e) (f) fractures or platy mineral grains are oriented in
almost exactly the same direction. The con-
stituents of a rock mass also may be arranged to
form a metamorphic lineation (Turner and Weiss,
1963, p. 102). Examples include elongate clusters
of mineral grains (Fig. 2.24d) and individual pris-
matic grains (Fig. 2.24e) that point in almost the
same direction. Platy mineral grains (Fig. 2.24f)
Fig 2.24 Schematic illustrations of rock fabrics. Planar that contain a particular direction form a
fabrics consisting of (a) different rock types, (b) sub-parallel lineation.
fractures, and (c) platy mineral grains. Linear fabrics
It is not uncommon, however, for rock fabrics
consisting of (d) elongate clusters of mineral grains,
to be less obvious than the schematic illustrations
(e) prismatic grains, and (f) platy mineral grains with
common direction. Reprinted from Turner and Weiss (1963) of Fig. 2.24. In these instances it is necessary to
with permission from McGraw-Hill. analyze the orientations of the constituents and
determine whether or not these data could have
resulted from a random sampling of a population
looks very much like the stereonet, but has the that has no preferred orientation. In cases such as
desirable feature that areas bounded by pairs of slickenlines the lineation may have a direction
adjacent great and small circles, each separated by that must be considered (Davis, 1986). An analysis
the same number of degrees, have the same of orientation or direction data can be purely
surface area (Phillips, 1954). In other words this graphical, in which case the Lambert equal area
net satises the cartographic criterion of equiva- projection and the Schmidt net are well suited to
lency, i.e. the correct representation of areas. Note the task of preparing the so-called fabric diagram.
that such areas clearly are not of equal size on the Methods have been devised for hand contouring
stereonet (Fig. 2.21), where pairs of great and small the number of points per unit area on the
circles are separated by 10. For example, a 10 by Schmidt net (Marshak and Mitra, 1988, p. 148). For
10 area near the center of this net is smaller than example, a counting circle with an area that is 1%
one near east or west along the reference circle. of the area of the net is positioned at every inter-
Therefore, a set of points plotted near the center of section of a regular grid covering the net. The
the stereonet would appear to be more densely number of points in the counter is associated with
clustered than a set with the same angular rela- each grid intersection and these numbers are con-
tions plotted near the reference circle. toured. A nested set of contours encloses a cluster
The need for equivalency is most apparent in of points and serves to identify a direction of
structural studies that address the orientation preferred orientation. Graphical methods that
include considerations of statistics (Fisher, 1953) (a) Z

have advantages over these contouring methods, Vertical
so examples are presented after we describe how plane
to prepare the fabric diagram.
The rst step in the preparation of the fabric R n
diagram is the plotting of points that represent Plunge
the orientations of line elements (e.g. the normals direction
C fp
to foliations or the lineations) on the Schmidt net.
Consider a reference sphere of radius R with an 2n
R
arbitrarily oriented line segment passing through P
the center, C, and intersecting the lower hemi- Trace of P
sphere at the point P. We view the line segment CP projection plane A
in a vertical plane that contains the plunge direc- R
tion (Fig. 2.25a). Unlike the stereographic projec-
tion, the equal area projection plane is tangent to
the sphere at the antipode, A. Given the relation (b)
(2.68) between the angle and the plunge angle, N, y ge n
l un ctio
p, the angle ACP is: P ire
d
Angle ACP 180 90 p 90 p 2 Projection x P

plane
(2.84) ap y
Because AC CP R, the triangle ACPA is isosceles W A E, x

and angle CPA is equal to angle PAC, so:
R
1
Angle CPA (180 2) 90 (2.85)
2
Because the angle ZPC is , the angle ZPA is a right

angle and the triangle AZPA is a right triangle with Reference
hypothenuse 2R. Using this right triangle the dis- S circle
tance, AP, from the antipode to the point in ques-
tion is:
AP 2R sin (2.86)
(c) Sphere
Equatorial
It is a property of the Lambert equal area projec- plane
tion that the distance AP from the antipode to the
projected point, P, is equal to the distance AP.
Substituting for the angle using (2.68) we relate
C N, y
Q
this distance to the plunge angle:
xl
1
AP AP 2R sin 45 p
2 (2.87) w
Q(x, y, z)
z
Geometrically this step in the Lambert projection
can be accomplished by turning a circle with Fig 2.25 Lambert equal area projection of a linear
element. (a) Vertical plane containing the linear element.
center at A through the point in question onto the
(b) Equatorial projection plane. (c) Reference sphere.
projection plane (Ragan, 1985, p. 273).
The next step in the projection procedure con- An arbitrary point, Q (x, y, z), at the intersection of
siders the projection plane itself (Fig. 2.25b). a particular meridian and parallel corresponds to
Linear elements with zero plunge project onto the a linear element through the center of the sphere.
reference circle in this view, so (2.87) gives the This point projects onto the horizontal (x, y)-plane
radius of the reference circle as: at Q and the distance CQ (x2 y2)1/2. Further-
more, the cosine of the plunge angle, p, is CQ/R.
R 2R sin (45) R 2 (2.88) Using these relationships as shown in Fig. 2.25c,
the plunge direction and plunge are:
The arbitrarily oriented point P is associated with
a plunge direction, p, measured clockwise from 1
p tan 1( yx),
north. Taking a Cartesian coordinate system with 2
x- and y-axes coincident with east and north, p cos 1[(x2 y2)12R] (2.92)
respectively, the coordinates of the point P are:
Given the longitude and latitude of points on a par-
x APsin p, y APcos p (2.89) ticular meridian or parallel, the rst set of equa-
tions establishes the Cartesian coordinates and the
It is convenient to scale the reference circle of the second provides the plunge direction and plunge.
equal area projection so its radius, R, is equal to These two angles are used in (2.90) to project and
the radius of the reference sphere, R. This is plot the points. The longitude intervals between
accomplished by multiplying AP by 12 2. successive meridians is arbitrary and is taken as
Substituting for AP from (2.87) and scaling the 10 for the construction of Fig. 2.23. The range of
radius we have: longitudes is 0 180 to cover the lower hemi-
sphere. Similarly the latitude intervals between
1

x R 2 sin 45 p sin p
2
successive parallels is taken as 10, and the
range of latitudes is 90 90.
(2.90) Given the equations to construct the Schmidt
1

y R 2 sin 45 p cos p
2
net and to plot points representing linear ele-
ments on a fabric diagram, the most important
These are the coordinates of a point on the question is whether or not the distribution of
Schmidt net of radius R representing the orienta- points has a statistically signicant preferred ori-
tion of a linear element with plunge direction, p, entation (Kamb, 1959a). This question is addressed
and plunge angle, p. For foliations or other graphically by relating the area of the counter
planar fabrics the azimuth and plunge of the used in the construction of contours on the
normal, n and n, are substituted in these equa- diagram to the total number of points, N, in the
tions to prepare the fabric diagram. population. The counter is a circle with area Ac
The Schmidt net itself is plotted as the merid- that is some fraction, 0 f 1, of the total area An
ians and parallels of a sphere of radius R, oriented of the Schmidt net:
such that the poles, north and south, are on the
Ac r2
horizontal y-axis (Fig. 2.25c). Note that this illus- f (2.93)
An R2
tration of the sphere is rotated slightly about the
vertical z-axis so the equatorial plane is visible. The center of the counting circle is positioned at
Longitude angles, , are measured from the hori- every intersection of an r by r square grid laid over
zontal x-axis around the equatorial perimeter and the projection. The number of points within the
latitude angles, , are measured along a meridian counter is recorded for each intersection and
from the equator toward the poles. The equations these numbers are contoured. What distinguishes
relating longitude and latitude to the Cartesian this method from those mentioned earlier is the
coordinates are: choice of the radius, r, of the counting circle.
To understand how r is determined consider a
x R cos cos , y R sin , z R cos sin (2.91) set of N points that have statistically uniform ori-
3R
entations over the entire net. In other words this r (2.96)
N 9
population has no preferred orientation. As the
counter is randomly positioned on the net the Given the radius of the Schmidt net, R, and the
number of points within the counter, n, varies. number of points, N, in the population (2.96) pro-
The distribution of n is a binomial distribution vides the radius of the counter and the distance
because the counting circle divides the popula- between the intersection points of the square grid
tion into two mutually exclusive sets: those inside that overlays the net. Using the counter and grid
the counter and those outside. On average the so dened one can construct the orientation
number of points in the counter will be fN density diagram to display graphically the statis-
because the area of the counter is a fraction f of tical signicance of a data set containing the ori-
the area of the net. The mean, m, and standard entations of a linear fabric or the normals to a
deviation, s, of this binomial distribution are planar fabric.
(Krumbein and Graybill, 1965, p. 102):
2.3.4 Field and model angles and
m f N, s [ f N(1 f )]12 (2.94)
analytical orientation statistics
Recall that the standard deviation is a measure of We turn now to the relationships between geo-
the spread of the distribution about the mean. As graphic angles and coordinates, and the angles and
the counter gets very small, f 0, m 0, and s coordinates used in data analysis and model con-
( fN)1/2. As the counter approaches the size of the struction. Recall that the orientations of planar
net, f 1, m N, and s 0. and linear elements that approximate geological
It is recommended (Kamb, 1959a) that the structures are measured in the eld using two geo-
radius of the counter be chosen such that m 3s. graphic angles, the azimuth, , and the inclina-
In other words, for the population with no pre- tion, . Different terms are associated with the
ferred orientation, the number of points within azimuth of strike (s), dip (d), plunge (p), and
the counter, on average, would be three times the normal (n), and appropriate subscripts distin-
standard deviation. The fabric diagrams can be guish these. Similarly, different subscripts for the
contoured at values of 0, 2s, 4s, 6s, etc. Contours inclination angle, , distinguish the dip (d),
drawn from counts using this prescription are plunge (p), and plunge of the normal (n). For any
very smooth. If such large counts cluster in one planar element the azimuth and plunge of the
region of the fabric diagram and produce closed normal line (pole) may be used to specify the orien-
contours with values greater than 3s, one can tation, so in fact we only need to consider how line
interpret the population as having a preferred ori- segments are oriented in three-dimensional space
entation. Substituting the expressions for the to account for the orientations of all linear and
mean and standard deviation from (2.94) into the planar elements that approximate geological struc-
condition m 3s and solving for f, we nd: tures. For data analysis and model computations it
is convenient to describe the orientation of any line
9
f (2.95) segment using three direction angles (x, y, z) that
N9
relate the line to a Cartesian coordinate system,
Note that a 1% counter area, sometimes chosen rather than the geographic system. Here we intro-
arbitrarily for the contouring of fabric diagrams, duce the relationships that transform eld data in
corresponds to N 891. This is an unusually large the geographic coordinate system to a Cartesian
number of points for fabric studies. Implemen- system using these direction angles. The textbook
tations using a 1% area and fewer than 891 points by Groshong (1999) provides additional discussion
are likely to produce irregular contours that have of these and other techniques of three-dimensional
no statistical signicance because the counter geometry as used in structural geology.
area is too small. Consider the orientation of the line segment OP
Substituting the ratio of areas for f in the pre- relative to the orthogonal geographic coordinate
vious equation, and solving for the radius of the system composed of the axes east, north, and up
counter, r, we have: (Fig. 2.26a). Regardless of their specic geological
(a) Up upward directed line segment (Goodman, 1980,

North
p. 145), whereas structural geologists take the
B
D downward directed segment and refer to the
a angle as the plunge. Perhaps geologists typically
gaze downward to observe structures in outcrops
O East and mining engineers gaze upward to observe
f A blocks of rock that may fall on their heads!
The line OP (Fig. 2.26a) projects onto the hori-
zontal (east, north) plane as the line OD, and this
P line in turn projects onto the east and north axes
as the lines OA and OB, respectively. The following
Vertical plane three trigonometric equations relate these lines
to the geographic angles:
C
OA OB OD
(b) z sin , cos , cos (2.98)
y OD OD OP
B Here it is understood that OP and OD are inher-

az ently positive and OC is negative, but OA and OB are
ay positive or negative depending upon their loca-
A x tion on the positive or negative extensions of the
O ax respective coordinate axes. This accounts for the
range of the azimuth and plunge.
Now consider a Cartesian coordinate system
P composed of axes x, y, and z (Fig. 2.26b) that shares
the origin at O with the geographic system, and is
oriented such that the respective axes are coinci-
dent with east, north, and up. The line OP projects
C onto the x-, y-, and z-axes as the lines OA, OB, and
Fig 2.26 Field and model angles. (a) Geographic OC. The direction angle x is measured in the
coordinate system with plunge direction, , and plunge angle, plane POA; the angle y is measured in the plane
, of line element. (b) Cartesian coordinate system with POB; and the angle z is measured in the plane
direction angles (x, y, z) of line element. POC. Each direction angle is the smaller of the two
possible angles from the line OP to the positive
extensions of the respective coordinate axis. Thus,
meaning, all azimuths, , are dened as angles the ranges of the direction angles are:
measured clockwise from north in the horizontal
plane to the vertical plane (ODPC) that contains the
0 x , 0 y , z (2.99)
line. All plunges, , are dened as the angle mea- 2
sured downward in this vertical plane from the
horizontal plane to the line. Thus the ranges of The following three trigonometric equations
these two geographic angles are restricted as relate the projections of OP onto the coordinate
follows: axes and the line OP itself:
OA OB OC
0 2, 0 (2.97) cos x , cos y , cos z (2.100)
2 OP OP OP
It is interesting to note that mining engineers These are the direction cosines for the line OP with
dene the normal to planar elements as the respect to the Cartesian coordinate system.
2.4 STRUCTURAL MAPPING USING GPS TECHNOLOGY 69
The transformation equations are constructed Note that the magnitude of the resultant
from the preceding trigonometric identities as vector does not have a unit value. The components
follows: of the resultant vector are found as the sums of
the respective components of the set of unit
OA OD
cos x sin cos vectors:
OD OP
OB OD n n n
cos y cos cos (2.101)
OD OP Ux u (i), U u (i), U u (i)
i1
x y
i1
y z
i1
z (2.105)
cos z
OC
OP

cos sin
2 The direction cosines of the resultant vector are
given by the ratios of the components to the mag-
In this way the direction cosines for the line OP are nitude of this vector:
calculated from the azimuth and plunge of the
linear element. The azimuth and plunge are cos x Ux U, cos y Uy U,
recovered from the direction cosines using: (2.106)
cos z Uz U
tan1 cos x
cos y, sin1[cos z] (2.102) The azimuth and plunge of the resultant vector
(mean direction) are found using (2.102). If the
To compute the full range for the azimuth, the unit vectors representing the direction data are
signs of both the numerator and the denominator widely scattered the magnitude U is small com-
in the arctangent function must be used. Most pared to n, whereas for tightly clustered data the
computer languages offer a function such as magnitude of U approaches n. The spherical vari-
ATAN2(XNUM, YNUM) that explicitly uses the two ance is dened (Davis, 1986, p. 334):
arguments with their signs.
The relationships we have just derived are used s 2s (n U)n (2.107)
to develop an analytical method for determining
the mean direction for a set of n linear elements This is a measure of the clustering of the direction
or normals to planar elements (Davis, 1986). data about the mean.
Consider each member of the set to be a unit vector,
u(i), where i 1 to n. For example, the line segment
OP (Fig. 2.26) could represent one such unit vector:
2.4 Structural mapping using GPS
u(i) ux(i)ex uy(i)ey uz(i)ez technology
(2.103)
u(i) [u2x (i) u2y (i) u2z (i)]12 1
2.4.1 The Chimney Rock fault array
The components of any vector are equal to the The Chimney Rock fault array crops out on the
vector magnitude times the respective direction northern San Rafael Swell (Fig. 2.27) and is
cosine (2.8), so in the case of a unit vector the com- exposed over an area of about 25 km2 where the
ponents are the direction cosines. Using this rela- local stratigraphy (Fig. 2.28) is composed of the
tionship and (2.101) the components of the unit Jurassic Navajo Sandstone and overlying Carmel
vector may be related to the azimuth, (i), and Formation (Maerten, 2000; Maerten et al., 2001;
plunge, (i), of the line element or normal. Davatzes and Aydin, 2003). The lower Carmel is
The mean direction for a set of linear elements predominantly shale, sandy shale, and limestone
or normals taken as unit vectors is dened as the beds. The top of the Navajo and three resistant
direction of the resultant vector, U: limestone layers in the lower Carmel provided
excellent marker horizons for mapping in this
U Uxex Uyey Uzez
(2.104) region and for determining the location, orienta-
U [U x2 U y2 U z2]12 tion, and offset on the faults. The traces of the
meters
0 15
1
Chimney Rock Alternating gray
2
00
fault array limestone and shale
9
0
30
90 0
00
60
Blue-gray limestone
0
0 0 (mapped layer)

12
60
0

30
1500 0 Blue sandy shale
Base Carmel Formation

10
0
Green
1800
30 River Yellowish calcareous
39o 0 sandstone
Sandy shale
0
210
600
e ll
Sw
el
Yellowish sandstone
a fa
fold
nR 900
5
Sa
ket
Utah N Gray limestone

oc
e rp
t
Wa San Rafael Alternating yellowish
limestone and shale
1200
Swell
Reddish limestone
Reddish shale and
111o sandstone alternating
with limestone beds
Henry Mountains Basin 0 10 20 30 0 Unconformity,
km weathered surface
Top Navajo
Sandstone
Fig 2.27 Structure contour map of San Rafael Swell region Cross-laminated
of south central Utah (Maerten, 2000). Contour interval is aeolian sandstone
300feet. Location of Chimney Rock fault array indicated with iron oxide (reddish)
within rectangle.
Fig 2.28 Local stratigraphic section at top of Navaho

Sandstone and base of Carmel Formation in Chimney Rock
faults are shown on a map (Fig. 2.29) to trend area. Reprinted from Maerten et al. (2001) with permission
predominantly to the northwest or to the north- from Elsevier.
east. One might suppose that the fault system is
composed of two sets based upon these two strike
directions; however, this view does not take into cate that the slip was approximately down the dip
consideration the dip of the faults. Some of these of these faults. Therefore they are referred to as
faults dip to the north (solid lines), but others dip slip faults. Furthermore, because the hanging-
with a similar strike dip to the south (dotted wall moved downward relative to the footwall
lines). Therefore, based on the three-dimensional these are referred to as normal faults. If the hang-
orientations of these faults, there are a total of ingwall moved up relative to the footwall they
four different sets (Krantz, 1988). would be called reverse faults.
The traces of the Chimney Rock faults range in The structure contour map (Fig. 2.29) is based
length from a few hundred meters to as much as on the elevation of the top of the blue-gray lime-
5 km, and the offsets of the sedimentary marker stone near the bottom of the Carmel Formation
horizons range up to almost 40 m, measured in (Fig. 2.28). This resistant layer is about a meter
the fault plane and down the dip of the fault (Fig. thick and forms prominent ledges throughout
2.30a). As depicted in this sketch, the block of rock the mapped region. The elevation of the top of the
above the fault, referred to as the hangingwall, has blue-gray limestone gradually decreases from
apparently moved down relative to the block west to east across the map, so it is inclined to the
below the fault, which is referred to as the footwall. east. To estimate a typical dip angle, compare the
Slickenlines on the fault surfaces (Fig. 2.14b) indi- following points on Fig. 2.29:
540 000 541 000 542 000 543 000 544 000 545 000
4 345 000
N
4 345 000
The Chimney Rock fault system NM

Faults and structure contours
at the base of the Carmel Formation 14o
4 344 000
3o
1680
1660
4 344 000
1640
1700
fault
North
0
1700
0
1680
172
164
166
4 343 000
1720
4 343 000
Gla
1620
ss F
1700
ault
1680
0
0
166
164
170
0
66
rt F ault1
90
Sho
1620
4 342 000
16
4 342 000
1720
1660
1680
00
17
4341000
4 341 000
1640
Lit
tl eF
au
lt
10
lt Normal fault dipping

Fau
17
1700
to the north
1660
er r y
1680
Bl u e b 00 Normal fault dipping

17 to the south
4 340 000
4 340 000
0
168
40
17
20
0
166
17
F re
00 nch
1640
17 ma
La S nF
al au
0
lt
0
168
Fa
166
ult
4 339 000
4 339 000
UTM coordinate system, 0 1000 2000

zone 12 North, datum WGS84. METERS
Altitude measured from mean sea level (msl). Contour interval 10 m
540 000 541 000 542 000 543 000 544 000 545 000
Fig 2.29 Structural contour map of blue-gray limestone
unit near base of Carmel Formation illustrating offset by WGS-84. Noting the positions of these points
faults of the Chimney Rock array (Maerten, 2000). along the northing grid line 4 344 000 m and the
fact that they are separated by 1000 m horizon-
tally and 49 m vertically with no intervening
Easting: 542 000 m Easting: 543 000 m
faults, we calculate:
Northing: 4 344 000 m Northing: 4 344 000 m
Elevation: 1693 m Elevation: 1644 m
The complete positions would include informa-
d tan 1 1693 m 1644 m
1000 m
3 (2.108)
tion from the map that indicates it is in the north- The strike of the blue-gray limestone is every-
ern hemisphere, Zone M12, and the datum is where parallel to the structure contours so the
(a) slip W contours that intersect the trace of the fault

Dip
t
faul N (shown schematically in Fig. 2.30a). Now continue
following the 1680-m contour to the north on Fig.
2.29 until it encounters the Frenchman Fault and
1680-m contour
note that this fault dips to the south. The 1680-m
contour steps a total of about 150 m to the east
Slickenline Blue-gray across this fault. The relative upward motion of
Footwall limestone
Hangingwall
the footwall block on the north side of the
Frenchman Fault is responsible for this disconti-
nuity in the contour (Fig. 2.30a).
The sense of relative motion of the footwall
(b) Strike slip W
and hangingwall need not be oriented along the
fault N
dip of the fault. Another possibility is illustrated
in Fig. 2.30b where the slickenlines are oriented
parallel to the strike direction. The fault on the
1680-m contour left side of this figure is a strike slip fault and has
Blue-gray slipped such that the hangingwall block moved to
Slickenline
Footwall limestone the west relative to the footwall block. The fault
Hangingwall on the left side of this figure is a left-lateral fault
because, when looking across the fault, the block
on the opposite side appears to have moved to the
Fig 2.30 Schematic illustration of a pair of faults bounding
left. The fault on the right side of this figure is a
a graben. (a) Dip slip. (b) Strike slip.
right-lateral fault. Given sufficient slip in the strike
direction across the fault on the left side of Fig.
strike at the intersection of the 1670 contour and 2.30b, the 1680-m contour would step to the west
the 4 344 000 m northing grid line is due north as much as the same contour steps to the west on
(Fig. 2.29). The map symbol for strike and dip of the dip slip fault illustrated on the left in Fig.
bedding (Fig. 2.16) is placed at this intersection 2.30a. Thus, the sense and magnitude of offset of
and recorded as (000, 03). One can judge from the the structure contours are not diagnostic of the
spacings and orientations of the structure con- direction or magnitude of slip.
tours that this is a typical strike and dip, at least On the other hand the slickenline directions
over the eastern portion of the map. If the blue- and the offset of sedimentary horizons are diag-
gray limestone were planar, the contours would nostic of the slip direction and magnitude of dip
be straight lines and uniformly spaced. The undu- slip on the Chimney Rock faults (Fig. 2.31). The
lations in these lines suggest that this limestone magnitude of the dip slip increases from zero at
layer is locally folded, perhaps as a result of the the eastern termination of the Blueberry Fault to
faulting or the development of the San Rafael about 16 m and then jumps to about 30 m across
Swell (Fig. 2.27). the intersection of the La Sal Fault. Between the
The sense in which the structure contours on La Sal and Little Faults the dip slip increases to
Fig. 2.29 are discontinuous across the normal about 35 m and then decreases to just less than
faults depends upon the direction of inclination 30 m before jumping to about 16 m across the
of the faults. For example, follow the 1680-m Little Fault. The rake of slickenlines is about 90
contour from near the southeast corner of the (pure dip slip) near the eastern termination of
mapped region toward the north and note that the Blueberry Fault and decreases to about 70 at
this contour steps about 300 m to the west across the intersection with the La Sal Fault. Across the
the La Sal Fault. The La Sal Fault dips to the north La Sal Fault the rake jumps to about 105 and
and the relative downward motion of the hang- then decreases toward the Little Fault, across
ingwall block on the north side of this fault is which the rake jumps to 65 and then increases
responsible for the discontinuity in this and other back to 90.
60 60
La Sal Little
Blueberry Fault Fault Fault
50 70
Rake ()
40 80o
Dip slip (m)
NE SW
30 90
20 100
?
10 110
0 120
0 500 1000 1500 2000 2500 3000
Distance from eastern tip (m)
Fig 2.31 Distribution of dip slip magnitude and rake of
ments in the TDC1TM data collector running Asset
slickenlines on Blueberry Fault of the Chimney Rock array
(Maerten, 2000).
SurveyorTM software that prompts the user for
particular field measurements using a data dic-
tionary. The receiver and data collector are
These data demonstrate that the Chimney carried in the field in a small backpack (fron-
Rock faults have a systematic variation of slip tispiece, Chapter 2) and the data are later down-
magnitude and direction and that the intersec- loaded to a laptop computer running the
tions of faults are associated with discontinuities PathFinder OfficeTM software. This software is
in both magnitude and direction. While one used to create a custom data dictionary, visualize
could characterize the Blueberry Fault as a the map data, process the coordinate data using a
normal fault with about 30 m of dip slip, this technique called differential correction, and
would not reveal the interesting slip variations export the final data to other software or to a
that provide important insights regarding the printer. Coordinate data can be collected at preset
mechanical behavior of this fault and its neigh- time intervals as the geologist walks throughout
bors. The data necessary to quantify these slip the field area, or at specific sites where particular
variations are obtainable using GPS technology. structures crop out.
The Asset SurveyorTM software enables the
2.4.2 GPS technology and mapping structural geologist to create a data dictionary
techniques specifically tailored to the structures and terrain
The combination of fault traces and structure in the region being mapped. In terms of recording
contours on Fig. 2.29 provide considerable insight quantitative data, keyed to locations, this is a
about the geometry of the structures in the significant improvement on the traditional geolo-
Chimney Rock region. The construction of this gist notebook. For this region the basic features
map was facilitated by a sub-meter precision GPS were the four distinctive sedimentary layers and
receiver combined with a data collector and a the faults, so a data dictionary could contain the
laptop computer (Maerten et al., 2001). This is a entries shown in Table 2.2 (Maerten et al., 2001).
remarkable tool for modern structural investiga- Each feature can be recorded at a particular site as
tions. Specifically, the ProXL SystemTM manufac- point data or at many sites along a traverse as
tured by Trimble Navigation Limited was used to line data. For example, as the geologist walks
receive the GPS signals and process the coordi- along the ledge (Fig. 2.28) formed by one of the
nate data to determine locations to within less resistant limestone layers (C1, C2, C3) or the top of
than 1 m, given sufficient satellite signals. These the Navajo Formation (Nav), the receiver can
data were combined with other field measure- collect location coordinates every few steps,
Table 2.2. Data dictionary. to construct the map shown in Fig. 2.29 in fteen
days using the GPS system (Maerten et al., 2001). A
Features Attributes Options total of 14 863 locations (easting, northing, eleva-
tion) were collected to dene the tops of the four
Layer (point) Type C1, C2, C3, Nav sedimentary layers. The thickness of the local
Strike 0 to 360 stratigraphic section (Fig. 2.28) was used to extrap-
Dip 0 to 90 olate from a given measurement to the elevation
comments of the top of the blue-gray limestone at each loca-
Layer (line) Type C1, C2, C3, Nav tion. These combined data were used to construct
comment the structure contour map. In addition, a total of
Fault (point) Strike 0 to 360 1768 locations were collected to dene the traces
Dip 0 to 90 of the faults, and the orientations of these faults
Rake of 0 to 180 were measured at 537 locations. By precisely and
slickenline efciently digitizing the three-dimensional geom-
Surface Good, etry of the deformed layers and the faults in this
quality medium, region, the GPS mapping technology provides an
poor effective mapping tool for the structural geologist.
Fault offset 0 m to 100 m
Faulted layer C1, C2, C3, Nav
comments 2.5 Concluding remarks
Fault (line) Faulted layer C1, C2, C3, Nav
comments A number of new technologies are transforming
structural mapping. These include the Global
Positioning System (GPS), used to locate outcrops
creating a very dense data set along the traverse. and structures on Earths surface more rapidly and
For each feature there are a number of common precisely than one could have imagined when the
attributes such as the type of layer (C1 or Nav), authors of this textbook began their careers. High-
the orientation (strike and dip), and comments. resolution aerial photography and scanning tech-
The options determine such things as the names nologies enable one to create digital images with
of the layer types, or the magnitude of the offset pixel widths of a few centimeters on the ground.
on a fault. At particular sites, the orientation of a Airbourne Laser Swath Mapping (ALSM) produces
layer or a fault can be measured using a compass topographic maps with decimeter precision. These
and inclinometer, and these numbers are recorded technologies are remarkable and they are rapidly
as the strike and dip angles. The denitions of evolving. There is no doubt that this revolution in
these angles were given earlier in this chapter. our ability to image Earths surface will usher in a
A single geologist collected the data necessary new generation of structural maps and data.
Chapter 3
Characterizing structures using differential

geometry
Two aerial views of the southeastern margin of the San Rafael

Swell, UT. Mesozoic clasitic sedimentary rocks are upturned
in the Waterpocket monocline. Photographs by D. D. Pollard.
The strange combination of mathematics and physics is

a Greek invention, pioneered by Archimedes. Modern
science is a mythical monster: half-goat, half-bird. The
student of physics is led simultaneously to the labora-
tory, to face the phenomena of physical reality; and to
the math course, to forget about the phenomena and to
contemplate pure abstractions. That this hybrid exis-
tence is at all fertile is amazing: we use it, because we
have discovered its effectiveness through experience.
The structure of the application of mathematics to
physics by Archimedes, then, is this: by making
explicit, clear assumptions, one draws the logical impli-
cations of the assumptions, which then have to hold for
the world as long as the assumptions themselves do. willing to say very little, as long as what we say is well
Mathematics may have little to say, directly, argued. Good arguments are good starting points for
about the physical world, but it is the only way to say truly productive discussion, and so it is not surprising
anything at all with any certainty. The bet of modern that the mathematical route has been so productive in
science following on Archimedes is that we are modern science (Netz, 2000).
76 CHARACTERIZING STRUCTURES USING DIFFERENTIAL GEOMETRY
I
n the previous chapter we illustrated examples tations. It may be clear that a particular lineation
of geological surfaces, such as the top of the has a different plunge direction and plunge at dif-
Triassic Chinle Formation throughout the San ferent exposures, but how does the attitude vary
Rafael Swell in southern Utah, using structure from one location to the next? The lineation may
contours (Fig. 2.27). The more detailed shape of approximate a three-dimensional curve, so we
the top of the blue-gray limestone bed near the need to understand how to describe the shape of
bottom of the Carmel Formation in the Chimney such curves. The spatial variation of plunge direc-
Rock area is shown by the structure contours in tion and plunge can be inferred qualitatively from
Fig. 2.29. The frontispiece of this chapter shows the distribution of attitude symbols on a struc-
two aerial photographs of the exposed surfaces of tural map, but differential geometry provides the
Jurassic sandstone formations on the ank of the tools for the quantication and analysis of these
Waterpocket Fold that denes the southeastern spatial variations.
margin of the San Rafael Swell. Notice how the Relatively little use of differential geometry is
sandstone formations bend over the monoclinal found in the twentieth-century literature of
exure and also bend as the strike of the beds structural geology, despite the obvious need to
changes along the length of the fold. Monoclinal describe the complex shapes of curved lineations
exures are a continuing focus of structural inves- and surfaces, and the attractiveness of accom-
tigation in this region (Reches, 1978; Reches and plishing this in a quantitative manner (Mallet,
Johnson, 1978; Cooke et al., 2000; Johnson and 2002). This literature provides few examples that
Johnson, 2000). Clearly these surfaces are not we can use to illustrate the concepts introduced
planar, but what is their shape? A primary task for here. Furthermore, eld techniques for deducing
structural geologists is to describe and character- the shapes of curves or surfaces from scattered
ize such surfaces and this may be accomplished in location and orientation data are just now being
a mathematically rigorous and complete manner devised and tested. Therefore this chapter focuses
using concepts and tools from differential geometry, on the principles and methods of differential
the branch of mathematics that brings the power geometry that appear to have the greatest poten-
of vector calculus to geometry (Gauss, 1827). Here tial for application to structural geology. This
we review some of the elementary concepts of dif- introduction is meant to encourage the use of
ferential geometry that are helpful to quantify these principles and methods and thereby
the departure of geological surfaces from a plane provide, in the words of Reveil Netz (2000), good
(Mallet, 2002). starting points for truly productive discussion of
Structural data typically are gathered at scat- the geometry of geological structures.
tered exposures as point measurements and the Position vectors are used to describe points,
locations of these points should be identied curves, and surfaces in differential geometry.
using geographic or local coordinates and position Many other vector quantities, such as the tangent
vectors. The measured data include the local atti- and curvature vector, are derived from the posi-
tudes of planar and linear elements that approxi- tion vector and used extensively in this chapter.
mate, for example, a foliation (Figs. 2.24ac) or Therefore it is necessary to understand the basic
lineation (Figs. 2.24df) at the point of measure- concepts of vectors, and to be familiar with
ment (Cloos, 1946; Turner and Weiss, 1963). specic techniques for manipulating vectors as
Plotting the attitudes of a set of structural ele- introduced in the previous chapter before reading
ments on a stereographic projection enables one this chapter. For in-depth treatments of differen-
to compare the orientations of different members tial geometry that provide a rigorous mathemati-
of the set. While serving a useful purpose in their cal basis, the reader is referred to textbooks on the
own right, stereographic projections provide subject (Struik, 1961; Stoker, 1969). In particular,
an incomplete characterization of foliations and the book by Lipschutz (1969) is a source for much
lineations, because these projections lack any of the material in this chapter and provides many
information about the spatial variations of orien- useful exercises and worked examples.
3.1 THE CONCEPT AND DESCRIPTION OF LINEATIONS 77
3.1 The concept and description (a) F

of lineations
3.1.1 Discrete, superficial, and

penetrative lineations
Before proceeding to the analytical description of
lineations using differential geometry it is useful
to step back and review some qualitative aspects of
these structures. Some lineations are dened by
the intersection of two geological surfaces that
separate one volume of rock from another. For
example, a fault (FF, Fig. 3.1a) separates the rela- F
tively young and undeformed rocks of a sedimen- (b) I
tary basin from the older and more deformed
sedimentary or metamorphic rocks of the adja-
cent mountain range. An igneous contact (II, Fig.
3.1b) separates older deformed sedimentary or
metamorphic rocks from the younger rocks of an
igneous intrusion. An angular unconformity
(UU, Fig. 3.1c) separates older sedimentary or
I
metamorphic rocks from the overlying sedimen-
(c)
tary strata. In these two-dimensional illustrations,
taken from the textbook by Turner and Weiss
(1963), the fault is shown as a straight line and the U
igneous contact and unconformity are shown
U
as curved lines. In general such structures are
three-dimensional surfaces that may be roughly
planar or highly curved. The intersections of these
surfaces with other similar surfaces, or with the
Fig 3.1 Two distinct rock volumes separated across
surfaces of the sedimentary strata or metamor-
geological surfaces including: (a) a fault, FF; (b) an igneous
phic foliations, are curved lineations in three-
contact, II; and (c) an angular unconformity, UU.
dimensional space. Intersections of sedimentary or metamorphic layers with
To generalize these concepts consider two these surfaces define discrete lineations. Reprinted from
roughly planar and continuous geological sur- Turner and Weiss (1963) with permission from McGraw-Hill.
faces, say an igneous contact and an unconfor-
mity that intersect one another. To the extent
that the surfaces are planar, the intersection
denes a straight line (Fig. 3.2a). At points along fore is sometimes referred to as a non-penetrative
the intersection the attitudes, measured as the lineation.
plunge direction and plunge, would be approxi- The igneous contact and unconformity may be
mately equal and the resulting points would plot continuous curved surfaces and in this case their
in a very tight cluster on a stereogram. The type intersection denes a continuous curve in three-
of lineation we have just described is called a dimensional space (Fig. 3.2b). Measures of plunge
discrete lineation because it is made up of the set direction and plunge at discrete outcrops along
of points common to two discrete surfaces. such a curved lineation would result in distinctly
Unlike the lineations shown in Figs. 2.24df this different attitudes and the set of points would plot
type does not permeate the rock mass and there- along an arc across the stereogram. Because the
(a) intrusive contact. An example (Fig. 3.3a) is taken

from a small fault in the Lake Edison granodiorite
in the Sierra Nevada of central California. There a
set of opening fractures formed in the granodior-
ite, presumably due to contraction during cooling
(Bergbauer and Martel, 1999), and these fractures
were lled with hydrothermal minerals, predom-
inantly quartz, epidote, and chlorite (Segall and
Pollard, 1983a). During a later tectonic event
sufcient shear stress was resolved across these
weak surfaces to promote slip and the consequent
shearing of the hydrothermal minerals between
the two fault surfaces generated the slickenlines
(Segall and Pollard, 1983b; Martel et al., 1988). A
second example (Fig. 3.3b) is taken from the intru-
sive contact of a diorite porphyry sill in the Henry
Mountains, Utah. Here shearing of the highly
(b) viscous magma against the sandstone host rock
generated the slickenlines (Johnson and Pollard,
1973), which are composed of fragmented
feldspar grains.
Both examples of slickenlines in Fig. 3.3 are
taken from individual exposures but nearby expo-
sures also display these structures. If one could
observe the entire fault surface or igneous contact
surface, we suggest that the lineations would
cover much if not all of these surfaces. This type
of lineation is called a supercial lineation because
it is only found on a discrete surface. Furthermore
we suggest that the linear elements observed
locally would form coherent patterns over these
surfaces, reecting the continuous relative
motion of the two sides of the fault and the con-
Fig 3.2 Discrete lineations defined by intersections of
geological surfaces. (a) Straight lineation at intersection of tinuous relative motion of the magma against the
two planar surfaces. (b) Curved lineation at intersection of host rock. That is, one could dene a set of three-
curved surfaces. Reprinted from Turner and Weiss (1963) dimensional curves lying in these surfaces that
with permission from McGraw-Hill. are everywhere parallel to the local directions of
the slickenlines. These curves are the trajectories
of the relative motion of the surfaces. As with any
spatial relations of these outcrops are not depicted continuous curve the geometric attributes of
on the stereogram one could not reconstruct the these trajectories can be dened and analyzed
curved intersection from these data. However, if using differential geometry.
the geographic coordinates of these points were In contrast consider foliations composed of
recorded along with the attitudes one could recon- lithologic layering (Fig. 2.24a) or a set of sub-
struct the curved intersection and study its geo- parallel fractures (Fig. 2.24b), or the preferred ori-
metric attributes using differential geometry. entation of tabular mineral grains (Fig. 2.24c).
A second type of lineation exists only on dis- Where two roughly planar foliations with different
crete surfaces such as the surfaces of a fault or attitudes exist in the same rock mass, say litho-
(a) outcrops would be approximately equal and the

resulting points would plot in a very tight cluster
on a stereogram. A possible alternative is that the
lithologic layering is folded and the fractures are
roughly planar (Fig. 3.4b). In this case the intersec-
tions dene continuous curves in three-dimen-
sional space that permeate the rock mass, so
attitudes from different outcrops would plot at
scattered locations on the stereogram.
An outstanding challenge for structural geolo-
gists is the development of procedures to dene
individual curves in three-dimensional space
from scattered measurements of the attitudes
(b) and outcrop locations of a penetrative lineation.
The underlying presumption is that the linear
fabric elements from scattered outcrops are part
of a coherent and continuous pattern of curves in
three dimensions, much like the ow lines of a
three-dimensional steady-state ow problem in
uid mechanics. The ow lines are everywhere
parallel to the local velocity vector. We are not pre-
suming that penetrative lineations have any par-
ticular relationship to a velocity eld. Rather we
are advocating the study of such lineations in a
three-dimensional spatial context that could lead
to their quantitative characterization using dif-
ferential geometry. With such a characterization
in hand one could model the structures using con-
tinuum mechanics and test hypotheses concern-
ing relationships between the fabric and the
velocity eld.
3.1.2 Parametric representation of

curves
Given the intuitive concept that a curve is a set of
points, arranged side by side in some orderly
Fig 3.3 Superficial lineations on geological surfaces. and continuous distribution, it should not be sur-
(a) Slickenlines on fault surface in granitic rock of the Sierra prising that position vectors, which dene the
Nevada, California (Segall and Pollard, 1983b). (b) Lineations
locations of points, are used to dene curves. To
on igneous contact in Henry Mountains, UT (Johnson and
distinguish individual points clearly from the set
Pollard, 1973). Photographs by D. D. Pollard.
of points composing a curve we use the symbol p
for the position vector of a point and c for the
logic layering and a set of fractures, the mutual curve. The spatial continuity of the set of points
intersections dene a penetrative lineation that per- composing a curve is achieved by dening c as a
meates the rock mass (Fig. 3.4a). To the extent that continuous function. Because c is a vector quan-
both foliations are planar, the intersections dene tity these functions are called vector functions.
straight lineations. Attitudes gathered at different Curves in three-dimensional Euclidean space are
(a) trace out the curve. The vector equation (3.1) is

called the parametric representation of the curve,
and the real variable t is an arbitrary parameter
for this representation.
For example, consider the parametric repre-
sentation for a circular helix (Fig. 3.5a) dened as
(Lipschutz, 1969, p. 63):
c(t) a (cos t)ex a( sin t)ey btez

(circular helix) (3.2)
The components of this vector function and the

ranges of the constants, a and b, are:
cx(t) a(cos t), cy(t) a(sin t), cz(t) bt,

a 0, b (3.3)
(b)
The points on this curve lie on a right cylinder of
radius a with the cylindrical axis coincident with
the z-axis. As the parameter increases from t 0 to
t 2 the point on the curve advances in the z-
direction a distance 2|b|, and the x and y com-
ponents return to their original values. As t
continues to increase the points continue to
advance in the z-direction encircling the z-axis.
For a 0 and b 0, (3.2) reduces to the special case
of a circle of radius a in the (x, y)-plane (Fig. 3.5b):
c(t) a(cos t)ex a(sin t)ey (circle) (3.4)
In this representation of a circle the parameter t is

the counterclockwise angle measured in radians
from 0 at the positive x-axis.
Fig 3.4 Penetrative lineations. (a) Intersections of two
As we review the concepts of differential geom-
penetrative planar foliations define straight lineations.
etry the circular helix (3.2) and the circle (3.4) are
(b) Intersections of two penetrative curved foliations define
curved lineations. Reprinted from Turner and Weiss (1963)
used as examples because they are well known
with permission from McGraw-Hill. and because they are easily visualized. In the
classic review of lineations by Ernst Cloos (1946)
geological lineations are not reported with shapes
dened in terms of vector functions of a single that approximate a complete circle or the full
real variable, t, such that: cycle of a circular helix. However, lineations are
reported that lie on surfaces or within layers of
c(t) cx(t)ex cy(t)ey cz(t)ez (3.1)
sedimentary or metamorphic rock that approxi-
The three scalar functions (cx(t), cy(t), cz(t)) are the mate a cylindrical fold (Turner and Weiss, 1963,
components of the vector function with respect to pp. 1239). Folds are termed cylindrical if a
the base vectors (ex, ey, ez). These functions, along straight line moving parallel to itself can generate
with the base vectors, determine the position the surfaces of the layers. The straight-line gener-
vectors for all points on the curve: as t varies ator is called the fold axis. In Fig. 3.6a the idealized
smoothly from one value to another, the points shape of a folded surface containing a lineation is
(a) z (a) z
y
Arc of
circle
Fo
ld a
xis
Lineation
t = 2p
c y x
2p|b |
t=0
Cylindrically
folded surface
x |a |
(b)
(b) y z y
a
Arc of
circular helix
Fo
c ld a
xis
s Lineation
t
x x
t=0
Cylindrically
folded surface
Fig 3.5 (a) Circular helix, defined by position vector c, Fig 3.6 (a) Lineations on cylindrical fold with circular
with radius a and pitch b. (b) Special case of a circle where profile shape lie on arcs of circles. (b) Lineations oblique to
b 0. Arbitrary parameter is t; arc length is s. fold axis lie on arcs of a circular helix.
composed of two halves of a circular cylinder eations where they are continuous. Can the arc of
joined along the dashed line. This is just one of a circle or circular helix approximate a given lin-
the many possible shapes for a cylindrical fold. For eation, or is a more complex shape required? In
this special case, lineations that are perpendicu- part our inability to address this question is due
lar to the fold axis approximate the arc of a circle. to poor exposure and therefore inadequate data,
Lineations that are not perpendicular to the fold but in part it is due to the lack of quantitative
axis may approximate the arc of a circular helix tools to describe and analyze curved lines in three-
(Fig. 3.6b). Thus even these elementary examples dimensional space. Differential geometry pro-
of curved lines from differential geometry may vides these tools.
have application to geological lineations.
The lineations illustrated in Fig. 3.6 are shown 3.1.3 The unit tangent vector
as short tick marks on the folded surface. This is Recall that the local orientation of a curvilinear
in keeping with the fact that eld measurements structure at the exposure is measured as the ori-
usually are limited to orientations determined at entation of the line element that is tangential to
isolated exposures. Not much is known about the the lineation. It should come as no surprise, then,
continuity of lineations from exposure to expo- that we make use of the unit tangent vector, t, along
sure, or about the three-dimensional shapes of lin- a curve. Some care is needed to distinguish the
symbol for the arbitrary parameter of a curve, t, (a)

from that for the unit tangent vector, t. This vector s
is dened by considering the position vector to be c(s+s)-
a function of a special parameter, s, such that c(s) s+s
s=0
|dc/ds| 1. This parameter is the length of an arc
of the curve from some arbitrary initial point z
where s 0 (Fig. 3.7a), so it is referred to as the arc c(s)
length. Now consider the difference between the c(s+s)
position vectors for two points on the curve, say
s and s s. Notice that this is a secant to the curve
between these two points. Dividing this difference y
by the arc length, s, and taking the limit as this
length goes to zero, we are left with the denition x
of the derivative of the vector function c with
respect to the arc length s (Lipschutz, 1969, p. 61): (b) Tangent t(s) Secant
line line
c(s s) c(s) dc s
lim t(s) (3.5)
s0 s ds s+s
In the limit, as s goes to zero the secant becomes

parallel to the curve and of the same length as the
z c(s)
arc. Therefore, this derivative is the unit tangent
vector, t(s), at the point c(s).
The relationship between the unit tangent y
vector and the curve at any point c(s) can be
thought of intuitively in terms of a straight line x
passing through that point called the tangent
line. To envision the tangent line, rst consider Fig 3.7 Diagrams to define unit tangent vector t.
the secant line (Fig. 3.7b) that passes through the (a) Difference between two position vectors, c(s s) and
two points, c(s) and c(s s) on the curve. In the c(s), for curve defines vector parallel to the secant line. (b) In
limit, as s goes to zero the secant line becomes the limit, as the arc length, s, goes to zero the secant line
is parallel to the tangent line.
the tangent line, and becomes parallel to the unit
tangent vector at the point c(s). In this sense the
tangent line is the one straight line, of an innite
Comparing this equation with (3.2), we note that,
number of differently oriented straight lines
despite being called natural, this representation
through the point, that has the closest contact
is somewhat more cumbersome to write down
with the curve and best quanties the orientation
because the arbitrary parameter, t, is replaced
of the curve at that point. One can say that it is the
with (a2b2)1/2s. For a 0 and b 0 the equation
best tting straight line to the curve at that
for the circular helix reduces to the natural rep-
point.
resentation of the circle:
When the arc length, s, is used as the parame-
ter, the equation for the curve is called the natural c(s) a cos (sa)ex a sin (sa)ey (3.7)
representation of the curve. For the circular helix
the natural representation is (Lipschutz, 1969, In comparing this equation with (3.4) recall that
p. 52): the central angle of a sector of a circle measured
in radians is equal to the ratio of the arc length to
c(s) a cos[(a2 b2)12s]ex the radius, that is t s/a (Fig. 3.5b).
a sin[(a2 b2)12s]ey b(a2 b2)12sez The natural representation of a curve permits
(3.6) a direct calculation of the unit tangent vector by
taking the derivative with respect to the arc (a) z

length, t dc/ds. For an arbitrary parameter, t, the
unit tangent vector is dened using the chain rule
and the fact that ds/dt |dc/dt| (Lipschutz, 1969,
p. 61): k t
t
dc dc dt dc ds dc

ds dt ds dt dt dt | |
dc
dt
(3.8) c
y
This is a more general equation for the unit
tangent vector because it does not depend upon
the parameter being the arc length. Recalling x
that c is a vector function of a single real vari-
(b) y
able, t or s, how does one take the derivative of
such a function? The answer is that one takes the
t=/2
derivative of each component with respect to the t=-ex
variable and uses these as the components of a
new vector. For example, given c(t) as in (3.1), t=+ey
then: k= -e y /a
dc(t) dcx(t) dcy(t) dc (t) k=-ex/a x
e e z ez (3.9) t=0
dt dt x dt y dt
The function c(t) is differentiable at some particu-

lar value of the variable, say t t0, if each compo-
a
nent is differentiable at that point. Note that the
derivative of a vector function is a vector function:
higher-order derivatives may be calculated follow-
ing the same procedure, and the standard formu- Fig 3.8 (a) Unit tangent vector, t, and curvature vector, k,
lae for derivatives of the common functions apply on circular helix. (b) Unit tangent and curvature vectors on a
(Selby, 1975). circle.
For the circular helix (Fig. 3.8a) the derivative
and absolute value of the derivative of the vector By substitution, notice that the tangent vector for
function are found using (3.2): the circle varies with the parameter t as:
dc
a(sin t)ex a(cos t)ey bez,
dt t 0 2 32
(3.13)
| |
dc
dt
(a2 b2)1 2 (3.10)
t ey ex ey ex
This describes a set of unit vectors that are,

Equations (3.10) are substituted into (3.8) to nd
indeed, tangent to the circle (Fig. 3.8b).
the unit tangent vector for the circular helix:
The unit tangent vector for the circular helix
t(t) (a2 b2)12[ a(sin t)ex a(cos t)ey bez] maintains a constant angular relationship to the
(3.11) z-axis for any position along the circular helix
(Fig. 3.8a). This angle is found using the scalar
As b goes to zero the helix collapses into a circle product of the tangent vector (3.11) and the base
on the (x, y)-plane (Fig. 3.8b) and the unit tangent vector ez:
vector becomes:
t(t) (sin t)ex (cos t)ey (3.12) t ez b(a2 b2) 12 |t||ez|cos cos (3.14)
In the last step we used the fact that both vectors 3.1.4 The curvature vector and the
have unit magnitude, so the product of their scalar curvature
absolute values is one. Solving for the angle : Although the unit tangent vector is an important
geometric quantity for the characterization of
cos 1(t ez) cos 1[b(a2 b2)12] constant curves it is not one of the two fundamental prop-
(3.15) erties that uniquely determine the shape of a
curve. The rst of these fundamental geometric
As b goes to zero the helix collapses into a circle quantities is the curvature. The curvature vector, k,
on the (x, y)-plane and the angle goes to /2. This is dened for a natural representation of a curve,
special relationship is illustrated in Fig. 3.8b c(s), as the derivative of the unit tangent vector
where it is seen that t is perpendicular to the z- with respect to the natural parameter s. This deriv-
axis for all values of the arbitrary parameter t. ative is dened using the standard limiting pro-
For a lineation observed at exposure, such as cedure from calculus (Lipschutz, 1969, p. 62):
the slickenlines in Fig. 3.3, the orientation deter-
mined in the eld using geographic angles can be t(s s) t(s) dt
lim k(s) (3.17)
related to the unit tangent vector, t. Although s0 s ds
slickenlines may be curved, they are approxi- This denition, and the earlier denition of the
mated locally with linear elements whose orien- unit tangent vector (3.5), imply that the curvature
tation is measured using the plunge direction, p, vector is equivalent to the second derivative of
and plunge, p. We assert that such a linear the vector function c with respect to the natural
element is parallel to the unit tangent vector of a parameter: k(s) d2c/ds2. Therefore, denition of
three-dimensional curve that passes through the the curvature vector requires that the curve, c(s),
point of measurement, so it has the same direc- have a continuous second derivative over the
tion angles. Furthermore, because t is a unit interval of interest. In general, the curvature
vector, the scalar components are equivalent to vector is directed away from the curve on its
the direction cosines. We nd using (2.101): concave side (Fig. 3.8). In other words the curva-
ture vector points in the direction that the curve
tx cos x sin pcos p
is turning.
ty cos y cos pcos p (3.16)
The specic orientation of the curvature
tz cos z sin p
vector is determined by the fact that the tangent
In this way we relate the orientation data taken at vector is constant in magnitude; in fact it is a unit
scattered exposures on a continuous geological lin- vector. In general, if v is an arbitrary vector func-
eation to the components of the unit tangent tion such that |v| constant (not necessarily a
vectors at correlative points on a three-dimensional unit vector), then from the properties of the scalar
curve. product we have v v |v||v| cos 0 constant.
As an example, consider the lineation Differentiating this scalar product using the stan-
described by the plunge direction, p 222, and dard product rule (Selby, 1975):
plunge, p 33 (Fig. 2.17). Substituting these
dv dv dv
values into (3.16) we have tx 0.561, ty 0.623, v v 0,so v 0 (3.18)
dt dt dt
tz 0.545 and |t| 1. Recall that the geo-
graphic coordinates (east, north, up) correspond Given that neither v nor dv/dt is zero, a zero scalar
to the Cartesian coordinates (x, y, z), as shown in product implies that cos 0 and /2. Thus,
Fig. 2.15, to reconcile the three negative compo- the curvature vector is orthogonal to the unit
nents of the tangent vector with the orientation tangent vector.
of this lineation as plotted on the stereogram. Consider Fig. 3.9a to understand how the
Also note that these components combine to give magnitude of the curvature vector is related to
a unit magnitude, as expected for the tangent the change in orientation of the tangent vector
vector. with respect to position along a curve (Lipschutz,
(a) t(s) p. 472) to approximate the function for small

angles. Dividing both sides of this equation by the
differential arc length, s, and taking the limit as
this length goes to zero:
z
| lim
s0
t(s s) t(s)
s | d
lim (s)
s0 s ds
c(s) (3.20)
c(s + s)
t(s + s) The left-hand side of this equation is the magni-
tude of the curvature vector (3.17), |k(s)| (s), a
y quantity known as the scalar curvature. From this
x relationship one can understand that the scalar
curvature is equivalent to the spatial rate of
(b) change of the orientation of the unit tangent
t(s)
vector with arc length along the curve. The scalar
curvature, (s), is called an intrinsic property of a
u curve (Lipschutz, 1969) because it is one of two
t(s + s) t(s) quantities that uniquely denes the shape of a
t (s + s) curve. Where the orientation of the unit tangent
vector changes more rapidly with position along
the curve, the curvature is greater. A point on the
curve where the curvature is zero is called an
Fig 3.9 Diagrams to define scalar curvature, . (a) Two inection point.
tangent vectors, t(s s) and t(s), on curve separated by arc For an arbitrary parametric representation of
length s. (b) Change in angle, , between two tangent a curve, c(t), the curvature vector and the scalar
vectors with respect to arc length s defines scalar curvature curvature are calculated using (Lipschutz, 1969,
in limit as arc length goes to zero. p. 65):
1969, p. 73). Here the parameter is the arc length,

s, and the unit tangent vectors t(s) and t(s s)
k(t)
dt
dt | dc
dt|, |k(t)| k2x k2y k2z
(3.21)
are shown with different orientations at their
respective points on the curve, separated by some That is, one rst calculates the unit tangent vector
small (differential) arc length s. Bringing the using (3.8) and then calculates the derivative of
tails of these two unit vectors together forms an that vector function. To calculate the scalar cur-
isosceles triangle (Fig. 3.9b) with sides of unit vature one takes the absolute value of the curva-
length, a base of length |t(s s) t(s)|, and a dif- ture vector.
ferential angle between the sides, . The general To calculate the curvature vector and the
relationship among the sides of equal length, a, scalar curvature for the circular helix (3.2) we take
the base, b, and the angle between the sides, , the derivative of the unit tangent vector (3.11) and
for an isosceles triangle is b/a 2 sin(/2). use (3.21) to nd:
Substituting b |t(ss) t(s)| and a 1 we
have, for small angles: k(t) a(a2 b2)1[(cos t)ex (sin t)ey],
a(a2 b2)1 (3.22)
|t(s s) t(s) | 2 sin (2) (3.19)
Note that the curvature vector lies in the (x, y)-
Here we have used the rst term of the series plane (there is no component in the z-direction), it
expansion for the sine function (Selby, 1975, is orthogonal to the tangent vector, and it points
away from the curve on its concave side toward (a) s

the z-axis (Fig. 3.8a). Unlike the tangent vector, t,
the curvature vector, k, is not generally a unit c(s)
k(s)
vector and here it has a magnitude given by that
s+s
is a constant related to the radius, a, and the pitch,
b, of the helix. For the circle in the (x, y)-plane (Fig. s-s
C
3.8b), the curvature vector and scalar curvature
are found from (3.22) by letting b go to zero:
k(t) (1a)[(cos t)ex (sin t)ey], 1a (3.23)
By substitution, notice that the curvature vector

for the circle varies with the parameter t as: C
t 0 2 32
(3.24) (b) y t = -0.5
k ex a ey a ex a ey a 7 = 0.707
= 1.414
This curvature vector is orthogonal to the circle
and directed toward the center of the circle (Fig. 6
3.8b). It is not a unit vector unless a 1. The scalar
curvature is inversely proportional to the radius
of the circle. 5
A positive scalar quantity called the radius of t =0
curvature, , is motivated by (3.23) and is dened =2
for an arbitrary parametric representation of a = 0.5
4
curve, c(t), as the reciprocal of the scalar curvature
(Lipschutz, 1969, p. 63):
1 3
(t) (3.25)
(t)
From (3.23) we have a, so the radius of curva-

ture of a circle is the radius of that circle. For the -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 t
circular helix (Fig. 3.8) the geometric relationship -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 3 3.5 x
is not so obvious, but using (3.22) we see that the
Fig 3.10 Diagram to define radius of curvature (Lipschutz,
radius of curvature is a constant everywhere 1969). (a) Radius of curvature, , defined as radius of
along the helix with a value (a2 b2)/a. The best-fitting circle to curve at point s. (b) Scalar curvature, ,
radius of curvature for the circular helix is the and radius of curvature, (s), at two points on a parabola.
radius of a circle that is tangent to the helix and
lies in the plane dened by the unit tangent vector
and the curvature vector. In general, for an arbi- for the curve illustrated in Fig. 3.10a (Lipschutz,
trary curved line, the radius of curvature is a func- 1969, p. 1). Given the radius of curvature, (s)
tion of the parameter t, and therefore must be 1/|k(s)|, at any point along this curve, for what
calculated for every point along the curve. For any circle is this the radius? To address this question
straight line or segment of a straight line, and also consider a circle, C, that passes through the curve
for any point of inection along a curve the radius at the three points, c(s s), c(s), and c(s s). In
of curvature is innite because the scalar curva- the limit, as s goes to zero the circle C becomes
ture is zero. the circle C and the radius of C is equal to the
To acquire a more intuitive understanding of radius of curvature of the curve at the point c(s).
the radius of curvature consider once again the This circle lies on the concave side of the curve
natural parametric representation of a curve, c(s), and the curvature vector for the circle, C, and the
curve are identical at the point c(s). The circle C is The magnitude of the cross product is propor-
the one circle, of an innite number of different tional to the magnitudes of the two vectors and
circles that pass through this point, that has the the sine of the smaller angle, , between these
closest contact with the curve, and therefore pro- vectors measured in the plane that they dene:
vides a geometric visualization of the radius of
curvature. One can say that it is the best-tting |v w| |v||w|sin ,0 (3.29)
circle to the curve at that point. These general relationships for the cross (or
For the natural representation of a curve, c(s), vector) product are used extensively in differen-
recall that the curvature vector is the second tial geometry and are used here to calculate the
derivative of the position vector: k(s) d2c/ds2, so scalar curvature.
the scalar curvature is dened as (s) |d2c/ds2|. Again taking the circular helix (3.2) as an
Given this denition, one might be tempted to example we use (3.26) to calculate the scalar cur-
associate curvature with the second derivative of vature. The second derivative of the vector func-
a function, say y f(x), representing a plane curve. tion c(t) for the helix is:
In fact, this is commonly done in casual conversa-
tions where one says the rst derivative, dy/dx, is d2c
a(cos t)ex a(sin t)ey (3.30)
the slope of the curve, and the second derivative, dt2
d2y/dx2, is the curvature. However, such state-
The absolute value of the cross product of the two
ments involve special conditions and approxima-
derivatives is found using (3.27):
tions that are spelled out in the following
paragraphs. dc d2c
ab( sin t)ex ab( cos t)ey a2ez
If c c(t) is the arbitrary parametric represen- dt dt2
tation of a curve, as in (3.1), then a general
denition of the scalar curvature is (Lipschutz, |dc d2c

dt dt2 |
a(a2 b2)12 (3.31)
1969, p. 64):
Then the quotient in (3.26) is:
| || |
3
dc d2c dc
(t) (3.26) a(a2 b2)12
dt dt2 dt (t) a(a2 b2)1 (3.32)
[(a2 b2)12]3
When a curve is dened using a parameter other
then the arc length, s, the scalar curvature is not This is exactly the result found in (3.22): the scalar
simply the absolute value of the second derivative curvature of the circular helix is a constant
of the vector function c(t). Compare (3.26) to the depending upon both the radius and the pitch.
earlier denition of the scalar curvature (3.21): The general equation for the scalar curvature
both are useful denitions. (3.26) may be specialized for a plane curve, for
To calculate the scalar curvature using (3.26) example a curve that lies entirely within the (x, y)-
one uses the cross product of two vectors, say v and plane:
w, found from the components of these vectors as
c(t) cx(t)ex cy(t)ey (3.33)
(Selby, 1975, p. 556):
Here the vector function c(t) lacks any component
v w (vywz vzwy)ex (vzwx vxwz)ey
along the z-axis and the two non-zero components
(vxwy vywx)ez (3.27)
are functions of the arbitrary parameter, t.
The cross (or vector) product also may be evalu- Substituting into (3.26) we nd the scalar curva-
ated as the determinant of the matrix formed by ture for this parametric representation of the
the base vectors and components as follows: plane curve (Varberg and Purcell, 1992, p. 623):

ex vx wx
v w det ey vy wy (3.28) (t) | dcx d2cy dcy d2cx

dt dt2 dt dt2 |
dcx
dt
2

dcy
dt
2 32
ez vz wz (3.34)
Again we see that the curvature is not simply (x) 2(4x2 8x 5)32 (3.40)
related to the second derivative of the vector func-
For x t1, this is identical to the curvature of
tion c(t).
the parabola found in (3.36) using the parametric
The circular helix is not a planar curve, so we
representation as a function of t.
turn to the parametric representation of a para-
If the squared slope of the function, y f(x) in
bolic curve that lies entirely in the (x, y)-plane (Fig.
the denominator of (3.38) is small compared to
3.10b):
one, then the curvature may be approximated as:
c(t) (t 1)ex (t2 3)ey (3.35)
| |
2
2 y dy (3.41)
Taking the rst and second derivatives of the com- (x) ,for 1
x2 dx
ponents, the scalar curvature is found using
(3.34): For the parabola of Fig. 3.10b this approximation
|2 0| 2 gives a constant value, 2, for all x. This is exact
(t) 2 (3.36)
[(1) (2t)2]32 (1 4t2)32 at the base of the parabola where x 1 and the
slope is zero, but is in error by 182% where
For t 0.5 we nd 0.707 so 1.414 and a
0.707 at x 0.5 and the slope is 45o. For plane
circle of this radius is shown tangent to the
curves described by functions of the form y f(x),
parabola at this point in Fig. 3.10b. For t 0 the
with squared slopes that are not small compared
scalar curvature 2 and the radius of curvature
to one, (3.38) is the appropriate equation for scalar
is 0.5. A circle with this radius of curvature is
curvature. For the general parametric representa-
shown tangent to the point at the base of the
tion of a plane curve (3.34) is the appropriate equa-
parabola. Clearly there is a signicant change in
tion. For the general parametric representation of
curvature along a relatively short arc of the
a curve that is not conned to a plane (3.26) must
parabola.
be used.
We further specialize the representation of
the curve such that cx(x) x and cy(x) y. In other
words the arbitrary parameter in (3.33) is the x-
3.1.5 The unit principal normal vector
coordinate, and the derivatives of the two compo-
and binormal vector
We have already mentioned that the curvature
nents of c may be rewritten:
vector is not generally a unit vector, that it is
dcx dx d2c dcy dy d2cy d2y orthogonal to the unit tangent vector, and that it
1, 2x 0, , 2 2
dt dx dt dt dx dt dx is directed away from the curve on the concave
(3.37) side. As the curve c(s) passes through an inection
point (Fig. 3.11a), the curvature vector, k(s), goes to
Upon substitution into (3.34) the curvature takes zero magnitude and thereafter switches direction
the form often introduced in calculus textbooks to the other side of the curve. In the interest of
(Varberg and Purcell, 1992, p. 623): working with a geometric quantity that is less
erratic in both magnitude and direction, a unit
2 32
(x) | |
d2y
dx2
1
dy
dx
(3.38) vector is dened as parallel to the curvature
vector, but directed to remain continuous along
the curve wherever possible (Fig. 3.11b):
Again note that the scalar curvature is not simply
the absolute value of the second derivative of the k(s)
n(s) (3.42)
function y f(x). The parabolic plane curve illus- |k (s)|
trated in Fig. 3.10b may be written in this form using This vector is called the unit principal normal vector
x t1 and substituting for t in y t2 3 to nd: for the natural representation of the curve c(s)
(Lipschutz, 1969, p. 64). The choice of sign in the
y x2 2x 4 (3.39)
numerator is used to keep this normal vector from
The curvature for this function is found using switching direction arbitrarily from one side of
(3.38) as: the curve to the other at points of inection. For
(a) (a) z
z
b(t)
Inflection
y point n(t) t(t)
x
k(s)
c(t)
c(s) y
(b)
z x
(b)
y b(s)
x n(s) n(s)
Inflection
point t(s)
db/ds
c(s)
Fig 3.11 Diagrams to define unit principal normal z
(db/ds) n
vector, n. (a) Curve with curvature vectors, k(s). (b) Same c(s)
curve with unit principal normal vectors, n(s). y
x
example, in Fig. 3.11b the positive sign is used to
Fig 3.12 (a) Circular helix with moving trihedron defined
the left of the inection point so n(s) and k(s) are by unit tangent, principal normal, and binormal vectors (t, n,
similarly directed away from the concave side of b) all functions of the arbitrary parameter t. (b) Derivative of
the curve. However, to the right of the inection the binormal vector, db/ds, used to define the scalar
point the negative sign is used so n(s) and k(s) are torsion, .
oppositely directed. Although k(s) varies in mag-
nitude and remains directed away from the
concave side, n(s) is a unit vector directed consis- in the (x, y)-plane. However, as determined by
tently to one side of the curve. (3.11), the unit tangent vector for the circular
For an arbitrary parametric representation of helix is not in the (x, y)-plane unless b 0. Because
a curve, c(t), the unit principal normal vector is there are no inection points along the helix or
(Lipschutz, 1969, p. 66): the circle the choice of signs is arbitrary.
In order to identify the second property that
k(t) k(t) uniquely describes curves we again consider the
n(t) (3.43)
|k (t) | (t) natural parametric representation of a curve, c(s),
For the circular helix, using (3.22) in (3.43) with and dene a unit vector, b(s), called the unit binor-
the positive sign we have: mal vector, which is normal to the plane contain-
ing the unit tangent vector, t(s), and the unit
n(t) [( cos t)ex ( sin t)ey] (3.44)
principal normal vector, n(s):
Note that n(t) is independent of the pitch, b, of the
b(s) t(s) n(s) (3.45)
helix, so the principal normal vector given here is
the same as that for the circle. This unit vector is The three unit vectors [t(s), n(s), b(s)] form the so-
directed toward the z-axis and away from the called moving trihedron for the curve (Fig. 3.12b),
curve on its concave side for all values of the par- which can be thought of as traveling along the
ameter t (Fig. 3.12a). Also, note that this vector lies curve with change in arc length, s. The binormal
vector is directed following the conventional of the binormal vector, b, with arc length, s, but
right-hand rule for vector (cross) products: curl only that part of the change in orientation that
the ngers of your right hand from t toward n and projects onto the plane normal to the tangent
your thumb points in the direction of b. vector, t. In other words the torsion describes the
For an arbitrary parametric representation of component of rotation of the binormal vector
a curve, c(t), the unit binormal vector is the cross about the tangent line with change in position
product of the tangent and principal normal along the curve. The torsion is called an intrinsic
vectors, both written as functions of the parame- property of a curve because it serves, along with
ter t (Lipschutz, 1969, p. 68): the scalar curvature, to dene the shape of the
curve uniquely.
b(t) t(t) n(t) (3.46) For an arbitrary parametric representation of
Referring back to (3.29) and recalling that the the curve c(t) the torsion is (Lipschutz, 1969,
tangent vector, t(t), and the principal normal pp. 6970):
vector, n(t), are mutually orthogonal, the smaller
angle between them is /2, so sin 1.
Furthermore, both of these vectors have unit mag-
(t) | |
db
dt
dc
dt
n (3.49)
nitudes, so the binormal vector, b(t), also is a unit

The torsion also may be calculated in terms of the
vector. The moving trihedron (Fig. 3.12b) is com-
rst three derivatives of the vector function c(t):
posed of three orthogonal unit vectors.
|
For the circular helix (Fig. 3.12a) the unit
|
2
dc d2c d3c dc d2c
binormal vector is found using (3.11) and (3.44) in (t) (3.50)
dt dt2 dt3 dt dt2
(3.46):
This property of the curve depends upon the arbi-
b(t) (a2 b2)12[b( sin t)ex b( cos t)ey aez]
trary representation of the curve having deriva-
(3.47) tives of order 3 or greater that are continuous. The
For the special case where b goes to zero, resulting numerator of (3.50) is sometimes written without
in a circle, the unit binormal vector is b(t) ez. The the scalar and vector product symbols and
circle lies in the (x, y)-plane, which also contains without the inner parenthesis. In this form it is
the tangent vector and the principal normal referred to as a triple scalar product. The following
vector, and the unit binormal vector is parallel to determinant provides a handy way to evaluate a
the z-axis. triple scalar product:

ux vx wx
3.1.6 The scalar torsion [uvw] u (v w) det uy vy wy (3.51)
The unit binormal vector is used to dene the uz vz wz
second intrinsic geometric property of curves,
namely the torsion. For the natural parametric rep- Here u, v, and w are arbitrary vectors and the
resentation of a curve, c(s), the torsion is dened determinant is composed of their components.
as (Lipschutz, 1969, p. 69): For the circular helix the torsion is calculated
using (3.49). Taking the derivative of the unit

db binormal vector (3.47) and (3.10) we have:
(s) n (3.48)
ds
db
(a2 b2)12[b( cos t)ex b( sin t)ey],
Because b(s) is a constant (unit) vector function, dt
the derivative db/ds is orthogonal to b(s), and
therefore lies in the plane containing t(s) and n(s) | |
dc
dt
(a2 b2)12 (3.52)
(Fig. 3.12b). The scalar product of db/ds and n
determines the component of the vector db/ds on Taking the scalar product of the ratio of these two
an axis for which n is the base vector. Thus, the quantities with the unit principal normal vector
torsion is a measure of the change in orientation (3.44), the torsion of the circular helix is:
3.2 THE CONCEPT AND DESCRIPTION OF CURVED SURFACES 91
(a2 b2)1[b( cos t)ex b( sin t)ey] must address the continuity and spatial variation
[ ( cos t)ex ( sin t)ey] b(a2 b2)1 (3.53) of these structures. In this way eld data can be
put in the context of differential geometry and
The torsion is constant and proportional to the one can begin to analyze the shapes of lineations
pitch, b. For b 0, 0 and the circular helix is in nature. Some of the tools developed to analyze
called right handed, that is the axis of the helix is curves are used to describe and analyze curved
parallel to the thumb of your right hand when surfaces, the subject of the next section of this
your ngers bend to follow the curve (Fig. 3.12a). chapter.
For b 0, 0 and the helix is left handed.
Because b 0 for the circle, (3.53) demonstrates
that the torsion is zero for that special case. A 3.2 The concept and description
curve with zero torsion lies entirely within a
single plane, and that plane contains both the
of curved surfaces
unit tangent vector and unit principal normal
vector. Such a curve is called a plane curve and the 3.2.1 Discrete, folded, and penetrative
torsion for all plane curves is zero (Lipschutz, geological surfaces
1969, p. 70). Curved surfaces are found in a wide variety of geo-
In summary, we have introduced the paramet- logical contexts making up many different kinds
ric representation of curves as dened by the of structures. For example, discrete geological sur-
vector function, c(t), where t is an arbitrary par- faces include faults (Fig. 3.1a), igneous contacts
ameter. The shape of such curves is determined by (Fig. 3.1b), and unconformities (Fig. 3.1c) all of
the curvature, (t), and the torsion, (t), both of which locally resemble a planar surface, but
which are scalar functions of position along the viewed more broadly are curved surfaces that can
curve. The curvature measures the change in ori- be characterized using the principles of differen-
entation of the unit tangent vector, t(t), and the tial geometry introduced in this section. In each
torsion measures the rotation of the binormal case the curvature of these surfaces has implica-
vector, b(t), about the tangent line. These two tions for their origin and the physical processes
vectors, along with the principal unit normal involved in their evolution. For example, faults
vector, n(t), form a mutually orthogonal set of unit may be curved because they evolved from discrete
vectors called the moving trihedron of the curved segments that do not lie in a plane (Segall and
line. Pollard, 1983b; Martel et al., 1988). Furthermore,
We have shown how the geographic angles the curvature of a fault surface may constrain the
used to measure the orientation of lineations at direction and magnitude of slip during an earth-
the exposure are related to the components of the quake (Carena and Suppe, 2002). The curvature of
unit tangent vector (3.16), and how to calculate the contact of an igneous dike may be used to
the curvature from the spatial variation of the deduce the stiffness of the surrounding host rock
tangent vector (3.26). To apply the concepts of and the distribution of magma pressure (Delaney
curved lines from differential geometry to curvi- and Pollard, 1981). The curvature of an angular
linear structures observed at exposure, lineations unconformity provides information about the
must be sufciently continuous so that the sedimentary processes that shaped that surface.
tangent, normal, and binormal vector functions To make the appropriate deductions about the
and their rst derivatives with respect to the physical processes involved in the formation of
natural parameter can be dened. Data that faults, igneous contacts, unconformities, and
would be suitable for such a study are available in other discrete geological surfaces one must quan-
the geological record, but the common measure- titatively characterize the shapes of these sur-
ments of lineation attitudes at discrete points are faces.
not sufcient. The tools we have introduced here Sedimentary and metamorphic layering com-
are suitable for an analysis of the spatial varia- monly is folded and the shapes of the surfaces of
tions of lineations, so the gathering of eld data the folded layers have been the subject of many
H
(a)
(a)
(b)
(b)
(b)
H
Fig 3.13 Some geometric attributes of folded geological

surfaces. (a) Cylindrical fold with straight fold hinge, HH.
(b) Non-cylindrical fold with hinge that is a plane curve.
Reprinted from Turner and Weiss (1963) with permission (c)
from McGraw-Hill.
geometric classication schemes (Turner and

Weiss, 1963; Fleuty, 1964; Ramsay, 1967; Hudle-
ston, 1973; Twiss and Moores, 1992). For example,
structural geologists identify the line or curve
joining points of greatest curvature on such sur-
faces as the fold hinge (Fig. 3.13). Where the hinge
is approximately a straight line and that line is
capable of generating the folded surface when
moved perpendicular to itself (Fig. 3.13a), the Fig 3.14 More geometric attributes of folded geological
direction of the line is termed the fold axis and the surfaces. Successive hinges define axial surfaces that may be
surface is called a cylindrical fold. The hinge may be (a) planar, (b) cylindrical, or (c) non-cylindrical. Reprinted
a plane curve (Fig. 3.13b) or may curve in three from Turner and Weiss (1963) with permission from
dimensions. For a stack of layers in a fold the McGraw-Hill.
surface that contains successive hinges is referred
to as the axial surface. The shape of the axial isogon, fold tightness, and limb to hinge ratio.
surface may be approximately planar (Fig. 3.14a), While descriptions of folded surfaces and layers
cylindrical (Fig. 3.14b), or non-cylindrical (Fig. may be enhanced using these attributes, the
3.14c). Other geometric attributes of folded sur- resulting classication schemes are not mathe-
faces include the crest line, trough line, dip matically rigorous, they do not provide the data
necessary to reproduce the surface, and they do plane, not a line. This leads to the denition of a
not provide adequate measures for quantitatively curved surface as a continuous vector function of
comparing the shapes of the different surfaces or two scalar variables (u, v), called the parameters of
the spatial variations in shape of a single surface. the surface, such that s s(u, v). The two parame-
Descriptions of folded surfaces using differential ters may be thought of as the coordinates of
geometry do not have these shortcomings. points on a plane, called the parameter plane, and
Metamorphic rocks composed of the system- those points map onto the surface according to
atic arrangement of layers of different rock types the vector function s(u, v). As the two parameters
(Fig. 2.24a), a set of sub-parallel fractures (Fig. vary, the heads of the successive position vectors
2.24b), or a set of similarly oriented platy mineral sweep out the curved surface in three-dimen-
grains (Fig. 2.24c) are said to have a penetrative sional space.
planar fabric (Turner and Weiss, 1963, p. 97). The To understand the analytical denition of a
name indicates that the constituents locally curved surface we begin by describing the coordi-
dene a surface that resembles a plane, but the nate systems used for the two-dimensional par-
normal to this planar fabric may systematically ameter plane and the three-dimensional curved
change orientation from exposure to exposure. In surface (Fig. 3.15). The two Cartesian axes (Ou, Ov)
these cases it may be possible to represent this and the associated base vectors (eu, ev) dene the
spatial variation in orientation with a set of parameter plane on which the two coordinates
curved surfaces such that a given surface is every- are the parameters u and v. The three Cartesian
where tangent to the locally planar fabric. axes (Ox, Oy, Oz) and the associated base vectors (ex,
Procedures to dene individual continuous sur- ey, ez) comprise the system for the curved surface.
faces in three-dimensional space from scattered The position vectors, s, for the curved surface can
measurements of a penetrative planar fabric have be written as a function of the two parameters
not been devised, so that is a noteworthy chal- (Lipschutz, 1969, p. 128):
lenge for structural geologists. With such a char-
s(u, v) sx(u, v)ex sy(u, v)ey sz(u, v)ez (3.54)
acterization in hand one could model fabrics
using continuum mechanics and test hypotheses The three scalar functions [sx(u,v), sy(u,v), sz(u,v)] are
concerning their orientation using differential the components of the vector function, s(u, v),
geometry. with respect to the base vectors (ex, ey, ez). These
functions, along with the base vectors, determine
3.2.2 Parametric representations of the position vectors for all points on the curved
curved surfaces surface. The vector equation (3.54) is called the
We began our discussion of curved lines by appeal- parametric representation of the surface. Compare
ing to the intuitive notion of a set of points the facts that a single variable parameterizes the
arranged in an orderly and continuous fashion to curved line (3.1) and a pair of variables parame-
form a curve in three-dimensional space. Because terizes the curved surface (3.54).
individual points in three-dimensional space are Any point in the parameter plane (Fig. 3.15a)
identied by their position vectors, p, this led to may be dened by a two-dimensional position
the denition of a curved line as a continuous vector w ueu vev with respect to the base
vector function of a single scalar variable t, called vectors eu and ev. Thus, the position vectors for the
the arbitrary parameter of the curve, such that c curved surface, s, in three-dimensional space
c(t). As t increases in value the heads of succes- are determined by a vector function of the two-
sive position vectors trace out the curved line. The dimensional vector variable, w, that is s s(w).
analytical description of curved surfaces also may However, because the components of the vector w
be approached from the intuitive notion of a set are the two parameters (u, v), we can speak of the
of points arranged in some continuous fashion in surface as a function of these two scalar parame-
three-dimensional space. However, sufciently ters, s s(u, v), and that is what we will do in the
close to any particular point, the neighboring following discussion. An individual point (uo, vo) in
points are distributed such that they resemble a the parameter plane (Fig. 3.15a) maps onto the
surface (Fig. 3.15b) using the position vector p(uo, (a)

vo). Similarly, the coordinate lines u uo and v vo v
u = uo
in the parameter plane map onto the curves c(uo,
v) and c(u, vo) on the surface. This hierarchy of
points, curves and a surface is fundamental to the
ev Parameter
concepts of differential geometry.
plane p(uo,vo)
For example, consider the parametric repre-
v = vo
sentation of a particular curved surface (Fig. 3.16)
in three-dimensional space (Lipschutz, 1969, w
p. 151):
s (u v)ex (u v)ey 2(u2 v2)ez (3.55) O eu u
The components of this vector function are equal

to the coordinates x, y, and z in the three-dimen-
(b)
sional space containing the surface: s
s(u, v)
x u v, y u v, z 2(u2 v2) (3.56) v
Surface
c(u, vo) c(uo, v)
Adding and then subtracting the rst two equa-
tions to eliminate v and then u, and then substi-
tuting the resulting equations into the third z s
equation, we have: ez
) u
1 1 ,v o
u (x y),v (x y),so z x2 y2 ( uo
2 2 p
(3.57) y
O
The last of these equations is in the standard form ex ey
of an elliptic paraboloid (Selby, 1975, p. 400) and is
x
the special case where sections parallel to the (x,
y)-plane are circles. In the geological context the Fig 3.15 Parametric representation of a curved surface
patch of this surface near the origin is similar in (a) Two-dimensional parameter plane with parameters u and
shape to the surfaces of formations that are v. Lines uuo and vvo in the parameter plane map to v- and
deformed into a basin-shaped fold. u-parameter curves on the surface. (b) Three-dimensional
How do the coordinate lines u uo and v vo in surface defined by vector function of two parameters, s(u, v).
the parameter plane map onto this elliptic parab- Reprinted from Pollard et al. (2004) with permission from
oloid? In other words what are the curves c(uo, v) The Geological Society of London.
and c(u, vo) on the surface s(u, v)? For example,

setting v vo we have x y 2vo, so y x 2vo. This
equation denes a plane that is parallel to the z- value of uo varies, the set of v-parameter curves is
axis and intersects the (x, y)-plane along a line dened. In this way the two sets of coordinate
with a unit positive slope. This plane intersects lines, v constant and u constant, that cover the
the surface along a parabola to form the curve parameter plane map to the two sets of curves
c(u, vo) which is referred to as a u-parameter curve that cover the three-dimensional surface.
on the surface (Fig. 3.16a). As the value of vo varies, All three components of the vector function
the set of u-parameter curves is dened. Similarly, s(u, v) in (3.54) may be complicated functions of u
by setting u uo one denes a plane that is paral- and v, subject only to constraints that insure the
lel to the z-axis and intersects the (x, y)-plane along functions are continuous and can be differenti-
a line with a unit negative slope. This plane also ated, and that the surface has a well-dened
intersects the surface along a parabola, forming tangent plane at each point (Lipschutz, 1969,
the curve c(uo, v) which is referred to as a v-par- p. 150). However, in many cases of interest in struc-
ameter curve on the surface (Fig. 3.16b). As the tural geology a simpler form of the parametric
(a) s(u, v) u = 0.8

z v
(a) +1, +1
v = 0.7
p(0.8, 0.7)
v = constant
u = constant
c(u, vo ) u
Parameter
plane
y 1, 1
(b)
v-parameter
u-parameter curve
curve c(0.8, v)
x 2 z
c(u, 0.7)
y = x 2vo p(0.8, 0.7)

y
1
(b) z s( u, v )
s(u, v) +1
x 0
0 1
1 0 +1
c( u o ,v )
Fig 3.17 The same elliptic paraboloid shown in Fig. 3.16
but represented as a Monge patch with parameter plane
superimposed on the (x, y)-plane so u and v are equivalent to
x and y. (a) Parameter plane. (b) Wire-frame diagram of
elliptic paraboloid: the u- and v-parameter curves, c(u, vo) and
c(uo, v), are parabolas in (x, z)- and (y, z)-planes, respectively,
y
that define two of the sets of wires.
y = x + 2u o function of the two parameters, g(u, v). A paramet-

ric representation of a surface in which two of the
x
components of s are the parameters and the third
Fig 3.16 Elliptic paraboloid with circular sections parallel component is a function of the parameters is
to the (x, y)-plane. (a) The u-parameter curve, c(u, vo), is a referred to as a Monge patch, named after the
parabola. (b) The v-parameter curve, c(uo, v), also is a French mathematician Gaspard Monge who lived
parabola. from 1746 to 1818 and is regarded as one of the
founders of differential geometry.
representation of a surface, or of a patch of a Practical applications of the Monge patch to
surface, may be found such that: structural mapping become obvious if one thinks
of the parameter plane as superimposed on the
s(u, v) uex vey g(u, v)ez (3.58)
(x, y)-plane in three-dimensional space (Fig. 3.17a),
Note that the components, sx and sy, of s are so the parameters u and v are equivalent to the
equated to the parameters u and v, and only sz is a coordinates x and y. Then the vector function for
a surface that is the target for mapping may be corresponding point on the curved surface in
written: three-dimensional space.
s(x, y) xex yey g(x, y)ez (3.59)
3.2.3 The tangent plane, tangent vector,
Here we consider the (x, y)-plane as the horizontal and unit normal vector
plane of a local Cartesian coordinate system estab- We continue to use the more general parametric
lished in the region being mapped, and values of representation of a surface (3.54) to develop the
g(x, y) are the measured elevations of exposures of theoretical concepts necessary to characterize
the surface relative to the local origin. Similarly, surfaces, but recognize the Monge patch (3.58) as
the parameter plane can be superimposed on the a useful representation for mapping. Because the
UTM grid for the region in which case (3.58) is parametric representation of a surface, s(u, v),
transformed such that sx easting, sy northing, describes a vector function of two variable para-
and sz elevation. meters, there is a partial derivative associated
To gain further insights about the Monge with each parameter. To calculate the partial
patch consider the parameter plane (Fig. 3.17a) derivative s/u, for example, one takes the deriv-
with a rectangular grid of lines, u constant and ative with respect to u of each component of the
v constant, parallel to the coordinate axes and vector function while holding v constant, and
use the following parametric representation of a then uses these as the components of a new vector
curved surface which is in the form (3.58) of a function. Thus, the partial derivatives of s(u, v)
Monge patch (Lipschutz, 1969, p. 185): with respect to the two parameters are (Lipschutz,
1969, p. 126):
s(u, v) uex vey (u2 v2)ez (3.60)
s(u, v) sx sy s
Using the rst two components to eliminate the ex ey z ez
u u u u
parameters u and v from the expression for the (3.61)
s(u,v) sx sy sz
third component, and noting that the compo- ex ey ez
v v v v
nents of the vector are equivalent to the coordi-
nates, the equation z x2 y2 is found which is the Recall from (3.8) that for the natural representa-
same elliptic paraboloid illustrated in Fig. 3.16. tion of a curve, c(s), the derivative with respect to
However, in contrast to that representation, the u- the arc length, s, is the unit tangent vector.
and v-parameter curves for (3.60) are the intersec- Because the partial derivative s/u is taken with
tions of the surface with planes parallel to the v constant, this is equivalent to taking the deriv-
(x, z)-plane and the (y, z)-plane, respectively. ative along any one of the u-parameter curves, for
Note how three sets of curved lines are used to example, c(u, 0.7) as shown in Fig. 3.17b. Thus, the
represent the elliptic paraboloid in the wire- partial derivative, s/u, is a vector that is tangent
frame diagram (Fig. 3.17b). The intersections of to a u-parameter curve and points in the direction
the curved surface with planes parallel to the (x, z)- of increasing u. Similarly, s/v, is a vector that is
plane form a set of parabolas, as do the intersec- tangent to a v-parameter curve and points in the
tions of the curved surface with planes parallel to direction of increasing v. These tangent vectors
the (y, z)-plane. The intersections of the curved are not necessarily unit vectors because the u- and
surface with planes parallel to the (x, y)-plane form v-parameter curves are not necessarily the natural
a set of circles. A particular u-parameter curve, representations of these curves.
dened by the vector function c(u, 0.7), and a par- As an example consider the partial derivatives
ticular v-parameter curve, dened by c(0.8, v), are of the parametric representation for the elliptic
highlighted in Fig. 3.17b. The point (0.8, 0.7) on the paraboloid (3.60) illustrated in Fig. 3.17b:
parameter plane maps to the point p(0.8, 0.7) on s s
ex 2uez, ey 2vez (3.62)
the curved surface at the intersection of the two u v
highlighted curves. In this way every coordinate Note that the tangent vectors, s/u, for the u-
line and every point in the two-dimensional par- parameter curves lie in planes that are parallel to
ameter plane have a corresponding curve and a the (x, z)-plane. Consider the particular u-parameter
(a) s s(u, v) are not generally one. The tangent vectors,

= e + 1.6ez
u x s/v, for the v-parameter curves have a similar
form, but all lie in planes parallel to the (y, z)-plane.
The two partial derivatives (3.61) of the vector
function for a curved surface are used to dene
1.6 the parametric representation of planes, P, that
z are tangent to the surface. In general, the family
1.6 of tangent planes for the surface, s(u, v), are dened
c(u, 0.7) as (Lipschutz, 1969, p. 158):
1.2 p(0.8, 0.7)
s s
P s h k , h, k (3.63)
u v
0.8 1.0
This equation may be understood intuitively by
s
=e considering the vector function for the tangent
0.4 p(0, 0.7) u x
plane at the arbitrary point on the surface desig-
1.0
y, v = 0.7 nated by the point p(uo, vo) on the parameter plane:
0.0 x, u
1 0 +1
P(uo, vo) s(uo, vo) h
s
u |
uo,vo
k
s
v |
uo,vo
,
(3.64)
(b) h, k
z
2 s The rst term on the right-hand side is the posi-
k
v s tion vector for the point on the curved surface.
h
N u The second and third terms extend the position
vector parallel to the tangent vectors at this point
by arbitrary distances proportional to the vari-
1 ables h and k. As h and k range over the entire set
of real numbers, this equation denes all possible
7)
points on the tangent plane.

0.
x, u = 0.8
8,
P(u, v) For example, consider the particular point on

0.
y, v = 0.7 y
p(
the parameter plane p(0.8, 0.7), and its mapping

z = 1.13 +1
onto the curved surface s(u, v) illustrated in Fig.
x
1 +1
3.17b. The parametric representation of the
1 tangent plane, P, to this elliptic paraboloid (3.60)
at the designated point is:
Fig 3.18 Tangent vectors and tangent planes to a curved
surface. (a) Tangent vector to the elliptic paraboloid in the P(0.8, 0.7) 0.8ex 0.7ey 1.13ez h(ex 1.6ez)
direction of a u-parameter curve. (b) Tangent plane, P(u, v),
k(ey 1.4ez), h, k
to the elliptic paraboloid with unit normal vector, N.
(3.65)
A portion of this tangent plane is illustrated along

curve, c(u, 0.7), as shown in Fig. 3.18a. At the point with a portion of the wire-frame diagram for the
p(0, 0.7) on this curve the slope is zero and the mag- surface in Fig. 3.18b. Note that the rst three
nitude of the tangent vector is one, because s/u terms of the right-hand side of this vector equa-
ex. As u increases from 0, the inclinations of the tion locate the designated point, p(0.8, 0.7), on the
tangent vectors increase in proportion to u, just as surface. The fourth term is the tangent vector to
the slope of the parabolic curve increases. Thus, at the u-parameter curve, s/u, evaluated at this
the point p(0.8, 0.7) that tangent vector is s/u ex point and scaled by the arbitrary variable h. The
1.6ez. Clearly, the magnitudes of the tangent fth term is the tangent vector to the v-parameter
vectors as calculated by the partial derivatives of curve, s/v, evaluated at this point and scaled by
the arbitrary variable k. As h and k vary, the sum (a)

of these terms denes vectors that cover the v Parameter
tangent plane, P. plane
The tangent plane plays an important role in
p(uo, vo)
describing any curved surface. Furthermore, any v = vo
geological surface observed at exposure is approx- u = u(t ),
imated locally with planar elements whose orien- v = v(t)
tations are measured using angles such as the
strike, s, and dip, d. These planar elements are
O
tangent planes to the geological surface at the u = uo u s(u, v )
O
point of measurement and would be described
quantitatively by (3.63) if a parametric represen- N
(b) c(u, vo) s
tation of the geological surface were known.
Now we are in a position to bring the concepts v
of the curved line and curved surface together to T
understand the geometry of an arbitrary curved
line lying on a particular surface. This concept is c[u(t), v(t)] s
necessary because the curvature at a point on a u
surface may vary with direction, and these direc- z c(uo, v)
tions are dened in terms of curves passing y
through the point and lying on the curved surface.
Consider an arbitrary curve (Fig. 3.19a) in the par- x
ameter plane (u, v) that is dened by the functions u = uo + t cos U
u u(t), v v(t) and passes through the point p(uo, (c) v
v = vo + t sin U
vo). The two coordinate lines, u uo and v vo, are m
parallel to the axes and also pass through this
point. The parametric representation of the (uo, vo ) u
surface, s(u, v), along with the u-parameter curve, vo
c(u, vo), and the v-parameter curve, c(uo, v), are
shown in Fig. 3.19b. The arbitrary curve in the Parameter
parameter plane is a function of the two surface plane
O
parameters, u and v, and these are, in turn, func- O uo u
tions of the one curve parameter, t. Thus, the
parametric representation of the arbitrary curve is Fig 3.19 Diagrams to define the tangent vector, T, to a
given by the vector function c[u(t), v(t)]. curved surface, s(u, v), in the direction of an arbitrary curve,
The tangent vector, T, to the arbitrary curve is c[u(t), v(t)]. (a) Parameter plane. (b) Three-dimensional
given by the derivative of the vector function surface. (c) Parameter plane with line in arbitrary direction.
Reprinted from Pollard et al. (2004) with permission from
c[u(t), v(t)] with respect to the parameter, t. Unlike
The Geological Society of London.
the unit tangent vector, t, dened in (3.8), this
tangent vector is not normalized by its magni-
tude. Because c is a vector function of two variable In the last step we use the fact that the partial
parameters that are, in turn, functions of a single derivatives, s/u and s/v, are the tangent
variable parameter, the derivative is evaluated vectors to the u- and v-parameter curves, respec-
using the chain rule as (Lipschutz, 1969, p. 158): tively (Fig. 3.19b). As we have shown in Fig. 3.18b,
these two vectors lie in the tangent plane and,
dc[u(t), v(t)] dc(u, vo) du dc(uo, v) dv indeed, are used to dene the tangent plane (3.63)
T
dt du dt dv dt to the curved surface at a designated point.
s du s dv Because the tangent vector T is linearly dependent
(3.66)
u dt v dt upon these two partial derivatives, it also lies in
dc
the tangent plane. In this way the tangent vector ( cos )ex ( sin )ey
dt
to the surface s(u, v) at an arbitrary point in any
arbitrary direction is related to the partial deriva- (2uo cos 2vo sin 2t)ez (3.71)
tives of the parametric representation of the
At the arbitrary point p(uo, vo) the curve and its
curved surface at that point.
derivative are found by setting t 0 in these equa-
The following example provides geometric
tions. The tangent vectors anywhere on the ellip-
interpretations for the quantities du/dt and dv/dt
tic paraboloid (3.60) in the direction specied by
in (3.66) and further insight about the tangent
the direction cosines of the arbitrary line are cal-
vector T. Consider a straight line in the parameter
culated using (3.66) as:
plane through the arbitrary point p(uo, vo) and
having an arbitrary slope, m (Fig. 3.19c). The equa- T(u, v) (ex 2uez) cos (ey 2vez) sin (3.72)
tion for this line, given in the standard point-slope
form, is: This is equivalent to the tangent vector calculated
for the curved line in the preceding equation,
v vo m(u uo) (3.67) given our denitions of the parameters u and v as
To write this equation in parametric form con- functions of t.
sider the parameter, t, to be the coordinate mea- The orientation of the tangent plane (3.63) is
sured along the line from the arbitrary point uniquely determined by either of the two unit
p(uo, vo). The parameters u and v are related to the normal vectors to that plane. The choice between
parameter t by noting that the slope m tan , these two oppositely directed vectors is deter-
where is the angle from the positive u-axis to the mined by a right-hand rule: the unit normal
line, so: vector, N, makes a right-handed orthogonal
system with the two tangent vectors, s/u and
u uo t cos ,v vo t sin (3.68) s/v (Fig. 3.19b) and is dened as (Lipschutz, 1969,
By varying the angle, , lines with any orientation p. 158):
can be used to specify the direction of the tangent
s s
vector, T, at a point on the surface. The derivatives
u v
of u(t) and v(t) as used in the denition of the N (3.73)
tangent vector (3.66) are: |
s s

u v |
du
dt
dv
dt

cos , sin cos
2 (3.69) Recall that the vector (cross) product of two arbi-
These are the direction cosines of the angles trary vectors, v w, is normal to the plane con-
between the arbitrary line and the two coordinate taining v and w and that the thumb of your right
axes in the parameter plane. In (3.66) these two hand points in the direction of v w when your
quantities scale the tangent vectors for the u- and ngers curl from v toward w. Using (3.73) the unit
v-parameter curves to determine the tangent normal at any point on a surface can be calculated
vector, T, for the curve c[u(t), v(t)]. The ratio from its parametric representation s(u, v).
(dv/dt)/(du/dt) tan determines the direction of As an example consider the unit normal vector
the tangent line to this curve. for the elliptic paraboloid (3.60). Using (3.28) the
Returning to the example of the elliptic parab- vector product of the two tangent vectors is:
oloid (3.60), the straight line in the parameter

ex 1 0
plane (Fig. 3.19c) maps onto this surface as the s s
det ey 0 1
curve: u v
ez 2u 2v
c[u(t), v(t)] (uo t cos )ex (vo t sin )ey 2uex 2vey 1ez (3.74)
(u2o 2uot cos v2o
2vot sin t2)ez (3.70) Therefore, the unit normal vector is:
The tangent vector anywhere along this curve is: N (2uex 2vey 1ez)(4u2 4v2 1)1 2 (3.75)
At the origin in the parameter plane, u 0 v, we (a)

nd N ez. As expected, this is a unit vector
pointed away from the concave side of the surface
along the positive z-axis. For the point p(0.8, 0.7)
shown in Fig. 3.17, we calculate the unit normal d d
vector as:
N(0.8, 0.7) 0.681 01 ex 0.595 88 ey

0.425 63 ez (3.76)
(b)
Given the signs of the three components or the
visualization of the surface near this point in Fig.
3.18b, one can conclude that N(0.8, 0.7) points
away from the surface on its concave side and that
it obeys the right-hand rule with respect to the
two tangent vectors.
3.2.4 Dike and joint surfaces idealized as

helicoids (c)
Observations and mapping of opening fractures
in rock, including basaltic dikes (Delaney and
Pollard, 1981) and joints (Woodworth, 1896) (d)
suggest that the surfaces of some of these frac-
tures can be idealized as helicoids (Pollard et al.,
1982). Specically, the traces of some echelon
Fig 3.20 Surface structures of joints from pelitic rocks of
fractures are approximately straight when the Mystic River region, MA. (a) Joint plane with plumose
viewed in cross sectional exposures that are per- structure and fringe fracture surfaces. (b) Plumose structure
pendicular to the propagation direction (e.g. Fig. on fringe fracture surfaces. (c) Cross section of fringe
3.20c, d). However, exposures at different levels showing echelon fractures. (d) Cross section of fringe with
(serial cross sections) reveal different orienta- echelon fractures and cross fractures. Reproduced from
tions, such that the surfaces appear to twist about Woodworth (1896) with permission from The Museum of
an axis that is parallel to the propagation direc- Science, Boston.
tion (Fig. 3.20a, b). A straight line that is perpen-
dicular to the propagation axis would sweep out
The u-parameter curves on the surface are straight
these fracture surfaces if it were rotated about
lines that intersect and are perpendicular to the z-
the axis and translated along it. If the spatial rate
axis. Each v-parameter curve is a helix that inter-
of rotation is constant the twisted surface so pro-
sects the x-axis and curves around the z-axis. The
duced is a helicoid. To understand how to test the
tangent vectors to the u- and v-parameter curves
hypothesis that some dike and joint surfaces
are found using (3.61) such that:
approximate helicoids, we review the characteris-
tics of this class of surfaces using differential s
geometry. ( cos v)ex ( sin v)ey
u (3.78)
The parametric representation of helicoids is s
(u sin v)ex (u cos v)ey (c)ez
based on (3.54) where v constant (an angle) and v
u constant (a length) are coordinate lines in
Using these partial derivatives in (3.73) the unit
the parameter plane that map onto the u- and v-
normal vector at any point on the helicoid is:
parameter curves on the helicoidal surface
dened by the following vector function:
N(u, v) (1 c2 u2)[(c sin v)ex (c cos v)ey (u)ez]
s(u, v) (u cos v)ex (u sin v)ey (cv)ez (3.77) (3.79)
Line of Because the scalar product of unit vectors is equal

y breakdown to the cosine of the angle between them, the twist
angle is equal to the parameter v. The component
of s(0, v) in the z-direction is sz cv, and this is
b equal to the z-coordinate, so dz/dv c. The spatial
+b x
rate of twist of the surface is dened as dv/dz 1/c.
N(0,0) The patch of the helicoidal surface (Fig. 3.21) that
cv
we take as an analytical description of a fracture
surface, covers the range 0 v , so is the
Mid-line maximum twist angle at the distal edge.
For v 0 the helicoidal surface (3.77) is coinci-
dent with the x-axis (Fig. 3.21) such that the posi-
z
N(0, v) tion vector and unit normal are:
s(u, 0) (u)ex,N(u, 0) (1 c2 u2)
[(c)ey (u)ez ] (3.82)
v
Position on this part of the surface is given by the
Fig 3.21 Patch of a helicoidal surface taken as a model for value of u. For the analytical description of a heli-
the surface of a fringe joint. N is the unit normal vector to
coidal fracture surface we take the range b u
the surface; 1/c is the spatial rate of twist; and is the twist
b so the width of the fracture is 2b and the frac-
angle at the distal edge. Reprinted from Pollard et al. (2004)
with permission from The Geological Society of London. ture mid-line is coincident with the z-axis. We
take the x-axis as the intersection between a
planar fracture surface that lies in the (x, z)-plane
For a given value of the constant c, the unit where z 0 and a helicoidal fracture surface that
normal completely determines the orientation of twists about the positive z-axis. Thus, the x-axis is
the surface at any point specied by the parame- equivalent to the line of breakdown from a single
ters u and v. parent (main) fracture to multiple echelon frac-
The patch of a helicoidal surface used to model tures (feather fractures, twist hackle) in the fringe
a fracture surface is illustrated in Fig. 3.21. To of a joint or dike (Fig. 3.20a, line dd).
appreciate the geometric meaning of the parame- For u 0 and v 0 the unit normal for the heli-
ters and constant associated with the helicoid coidal surface is N(0, 0) ey, which is in the neg-
consider the following special cases. For u 0 the ative y-coordinate direction and is parallel to the
surface (3.77) is coincident with the z-axis such unit normal for the planar fracture surface (Fig.
that the position vector and unit normal are: 3.21). However, for all other points along the x-
axis, 0 |u| b and v 0, there is a component of
s(0, v) (cv)ez,N(0, v) ( sin v)ex ( cos v)ey (3.80) the unit normal in the z-coordinate direction that
Position on this part of the surface is determined by is proportional to u/(c2 u2)1/2. Thus, the planar
the value of v scaled by the constant c. The unit and helicoidal fracture surfaces may be continu-
normal is independent of c and varies, for example, ous with one another along the line of breakdown
from ey for v 0 to ex for v /2 as the local (x-axis), but there is a discontinuity in the orien-
surface twists from the (x, z)-plane to the (y, z)-plane. tation of the two surfaces except at the mid-line of
We dene the twist of the helicoidal surface the helicoidal surface (u 0, v 0). This may have
(Fig. 3.21) as the angle between the reference unit important implications for the growth of an
normal, N(0, 0), and the unit normal at the point echelon fracture surface.
in question along the mid-line, N(0, v). The scalar The geometry of twist hackle has been hypoth-
product of these two unit vectors is: esized to be similar to a helicoidal surface (Pollard
et al., 2004). We test this hypothesis using an area
N(0, 0) N(0, v) [(1)ey] [( sin v)ex (white rectangular box) on a joint surface found
( cos v)ey] cos v (3.81) on a hand-sized sample of chert (Fig. 3.22). The
(a)
Hackle
z
x
Main join
t
Fig 3.23 Oblique rendering of a portion of the joint fringe
in Fig. 3.22 from scanned data. Reprinted from Pollard et al.
(2004) with permission from The Geological Society of
London.
(b) Main join

t the main joint surface. The elevation of the hackle
Hackle
is greater than zero for lines between x 17 and

19.5 mm (except for the plucked portion), and less
than zero for lines between x 20.5 and 23 mm.
Furthermore, the ridge along x 17 mm increases
Fig 3.22 Photograph of the surface of a joint in chert with
in elevation and the trough along x 23 mm
plumose structure merging into twist hackle. Reprinted from decreases in elevation away from the main joint.
Pollard et al. (2004) with permission from The Geological These observations are consistent with the
Society of London. hypothesis that the hackle approximates a heli-
coidal surface.
Recall that the centerline of a helicoid is
data set consists of 5244 points measured with a dened as the line along which the parameter u
laser scanner that cover a small portion of the 0. The difference between the orientation of the
main joint surface, a portion of the breakdown unit normal at the proximal edge of the helicoid
zone, most of one coarse hackle in the fringe and that at any position along the centerline was
region, and the adjacent cross fractures. An dened as the twist angle, which is equal to the
oblique rendering of this area (Fig. 3.23) depicts parameter v. Using data for the twist hackle (Fig.
the local geometry of the main joint surface in the 3.23; Pollard et al., 2004) we calculate the change
background, the breakdown zone, the hackle, and in orientation of the unit normal vector, N, rela-
the two cross fractures. Note that this view fore- tive to a reference point on the main joint surface,
shortens the z-axis but leaves the x- and y-axes along the grid line x 20 mm, our interpretation
approximately equal. This rendering of the of the centerline (Fig. 3.24). The orientation of the
surface does not capture the sharp discontinuity unit normal relative to the reference value
of the exposed surface where the cross fractures changes systematically with the z-coordinate. On
meet the hackle, but it does provide an accurate the main joint surface (z 4 mm) the change in
image of the main joint, the hackle, and the cross angle is roughly constant and less than about 4.
fractures. Also, it should be noted that a small Over the breakdown zone (4 z 19 mm) the
piece of the hackle surface has been plucked away angle increases to about 25. For the distal portion
from the breakdown zone where the ridge along x of the hackle (19 z 37 mm) the angle oscillates
18 mm would have intersected the main joint about a more-or-less constant value. The spatial
surface. rate of twist (1/c) is the slope of the curve plotted
The centerline of the hackle (Fig. 3.23) lies on Fig. 3.24 in the breakdown zone. For the grid
somewhere between the lines x 19.5 and line x 20 mm, the slope is roughly constant over
20.5 mm, and the elevation of the surface there is the range 7 z 18 mm and is about 2 mm1
approximately zero, equal to the local elevation of (0.035 mm1). A constant rate of twist is consistent
main joint break down zone hackle fundamental forms. These quantities are intro-
Change in N (o)
30 duced in this and the following sections. The rst

25
20 fundamental form, I, at an arbitrary point on a
15
10 curved surface, s(u, v), is a measure of the differ-
5
0 ential arc length of curves lying on the surface
-10 -5 0 5 10 15 20 25 30 35 40
z [mm] and oriented in all possible directions at that
point. To dene this property of a curved surface
Fig 3.24 Graph of orientation of unit normal vector, N, consider the two points, p(u, v) and p(udu, vdv),
versus distance along the mid-line of a fringe fracture surface that lie along an arbitrary line, u u(t) and v v(t),
from the joint pictured in Figs. 3.22 and 3.23. Reprinted from in the parameter plane and are separated by an
Pollard et al. (2004) with permission from The Geological
arbitrarily small distance (Fig. 3.25a). These points
Society of London.
map onto the curved surface using the position
vectors p(u, v) and p(udu, vdv) along the curve
with the hypothesis that the hackle approximates c[u(t), v(t)]. A tangent vector to this arbitrary curve
a helicoidal surface. is dened using (3.66) such that (Lipschutz, 1969,
In summary, we have introduced the elemen- p. 171):
tary concepts to describe quantitatively a surface
s s
(3.54) using a parametric representation dened dc Tdt du dv (3.83)
u v
by the vector function of two parameters, s(u, v).
The partial derivatives of this function with Because this vector is a differential quantity that
respect to each parameter (3.61) are tangent is parallel to T, we refer to it as the differential
vectors to the respective parameter curves on the tangent vector. As shown in Fig. 3.25b, the differ-
surface and these serve to dene the tangent ential tangent vector is not exactly parallel to the
planes for the surface. Structural geologists quan- secant line that passes through the two points
tify the orientation of a geological surface at an p(u, v) and p(udu, vdv) on the curve. However,
exposure by measuring strike and dip or dip and recall from our discussion of Fig. 3.7 that, as the
dip direction, and these measurements use the distance between the two points goes to zero, the
concept of the tangent plane at a point on a tangent line and the secant line become parallel
surface. The tangent vector to an arbitrary curve and the tangent line becomes the best-tting line
on the surface (3.66) lies in the tangent plane. This to the curve at the point in question. In this limit
concept will be used in the next two sections the magnitude of the differential tangent vector,
where we introduce measures of arc length and dc, becomes equal to the arc length of the curve
curvature. The normal vector (3.73) to the curved c[u(t), v(t)] on the surface s(u, v).
surface is equivalent to the pole to a geological The rst fundamental form, I, is a differential
surface as plotted on a stereogram. In this way we quantity dened as the scalar product of the dif-
understand the relationships among the analyti- ferential tangent vector, dc, with itself (Lipschutz,
cal descriptions of a curved surface and the every- 1969, p. 171). Using (3.83) the rst fundamental
day techniques used by structural geologists. For form is expanded as follows:
the particular example of twist hackle in the
fringe region of a joint surface we have shown
how the normal vector can be used to test the
I dc dc
s
u
s
du dv
v
s
u
s
du dv
v

hypothesis that the hackle surface is helicoidal s s s s s s
(Pollard et al., 2004). du2 2 dudv dv2
u u u v v v
(3.84)
3.2.5 The first fundamental form, arc
length, and surface area The coefcients in this equation are scalar quan-
A continuous curved surface is completely tities with particular geometric interpretations.
described at an arbitrary point in terms of two dif- Because of their role in dening I, and in the cal-
ferential quantities called the rst and second culation of useful quantities such as arc length
(a) v Parameter The scalar quantities E, F, and G are called the

plane coefcients of the rst fundamental form. The
u = u(t), v = v(t) values of the coefcients depend upon the choice
of parameters used to represent the surface, but
b the rst fundamental form itself (3.85) is invari-
ant with respect to this choice (Lipschutz, 1969,
p. 172). In this sense I is a property of the surface
a and plays a fundamental role in dening arc
lengths on the surface.
O Recall that the scalar product of any vector, v,
O u
(b) with itself is equal to the squared magnitude of
the vector: v v |v||v| |v|2. Thus, the rst fun-
damental form may be interpreted as:
p(u + du, v + dv)
I |dc|2 0 (3.86)
dv
p(u,v) In other words it is the square of the differential
arc length (3.83) of the curve c[u(t), v(t)] on the
surface s(u, v), so it is a positive quantity. The rst
du fundamental form refers to the arc length of
s(u, v) curves in all directions at a particular point on the
surface, and the differential parameters du and dv
(c) in (3.85) are used to dene a particular direction.
Tangent p(u, v) + dc
plane In general, I 0 if and only if du 0 and dv 0.
dc Recall that the scalar product of two arbit-
rary vectors, v and w, may be written: v w
|v||w|cos , where is the smaller angle between
c[u(t), v(t)] the two vectors. The coefcients of the rst fun-
damental form as dened in the second of (3.85)
z are scalar products of the tangent vectors to the u-
p(u, v) p(u + du, v + dv)
and v-parameter curves. Therefore, they can be
y interpreted geometrically as:
x
Fig 3.25 Diagrams to define first fundamental form for a

E
u| |
s 2
, F | || |
s
u
s
v
cos ,
surface. (a) Parameter plane with arbitrary curve uu(t), v

v(t). (b) Fine-scale view of curve. (c) Three-dimensional view
G
v| |
s 2
(3.87)
of surface and mapped curve, c[u(t), v(t)], with differential

Here is the smaller angle between the two
tangent vector, dc, that approximates the arc of the curve.
tangent vectors at a particular point on the
The Geological Society of London. surface. From these equations we understand that
E and G are, respectively, the squares of the mag-
nitudes (lengths) of the tangent vectors to the u-
and area on the curved surface, they are denoted and v-parameter curves, so they satisfy the rela-
with the special symbols E, F, and G. Using this tionships E 0 and G 0. It is also the case that EG
notation the rst fundamental form is written F2 0. Furthermore, the u- and v-parameter
(Lipschutz, 1969, p. 171): curves are orthogonal ( 90) if and only if F 0
(Lipschutz, 1969, p. 173).
I E du2 2F dudv G dv2 As an example we compute the coefcients of
s s s s s s the rst fundamental form for the elliptic parab-
E ,F ,G (3.85)
u u u v v v oloid (3.60):
E (ex 2uez) (ex 2uez) 1 4u2 (a) Parameter plane

v
F (ex 2uez) (ey 2vez) 4uv (3.88)
G (ey 2vez) (ey 2vez) 1 4v2
dv
Where u 0 we have E 1, so the tangent vectors (u, v)
to the u-parameter curves along u 0 are unit
vectors, a result we illustrated in Fig. 3.18a. A
similar result is found for the v-parameter curves
by noting that G 1 for v 0. Because F is not zero, u
O
the u- and v-parameter curves are not orthogonal,
O
as is readily conrmed by glancing at Fig. 3.17b.
The rst fundamental form (3.85) at a point on du s(u, v)
a surface, s f(u, v), is the square of the differen-
tial arc length of a curve through that point. To (b)
dc(uo, v)
illustrate the geometric meaning of this quantity
p(u + du,
and to show how it is utilized in practical applica- dA v + dv)
tions consider the points t a and t b along the
arbitrary curve u u(t), v v(t) in the parameter dc(u, vo)
plane (Fig. 3.25a). This curve maps onto the
surface as c[u(t), v(t)] and the differential tangent z
vector, dc, anywhere along this curve is dened by p(u, v) c(u, vo)
(3.83). Integrating the magnitude (length) of the c(uo, v)
differential tangent vector, |dc| |dc/dt|dt, from x y
a to b we nd the arc length, s, of the curve on the
surface (Lipschutz, 1969, p. 173): Fig 3.26 Diagrams to define the differential surface area.
(a) Parameter plane with coordinate lines parallel to u and v.
b b
| |
12 (b) Surface with u- and v-parameter curves and parallelogram
dc dc dc
s dt dt approximating surface area between curves. Reprinted from
dt dt dt
a a Pollard et al. (2004) with permission from The Geological
b
E dt 2F dt dt G dt
du 2 du dv dv 2 12 Society of London.
dt (3.89)
a
onto the surface as a curvilinear grid (Fig. 3.26b).

Here the magnitude of dc/dt is taken as the square
The differential tangent vectors at the point p(u, v)
root of the scalar product of this vector with itself,
in the directions of the u- and v-parameter curves
and this scalar product is associated with the
are found from (3.83):
coefcients of the rst fundamental form using
(3.85). The derivatives du/dt and dv/dt establish the s s
dc (u, vo) du,dc (uo, v) dv (3.90)
direction in which the arc length is measured: u v
they are the direction cosines for the curve in the
These vectors form two sides of a small parallelo-
parameter plane that maps onto the surface as
gram and the differential area, dA, of this planar
c[u(t), v(t)] (Fig. 3.19).
gure is used to approximate the area of the
The coefcients of the rst fundamental form
curved surface between the adjacent parameter
also are useful for calculating the area of a surface,
curves. The differential area is:
given its parametric representation. Consider
adjacent members of the two families of coordi- dA |dc(u, vo)||dc(uo, v)| sin (3.91)
nate lines on the parameter plane that are sepa-
rated by small differential distances du and dv (Fig. This equation follows from the fact that the area
3.26a). These lines partition the parameter plane of a parallelogram with sides a and b with
into a rectangular grid that is, in turn, mapped included angle is A absin.
The formula for dA is similar to (3.29) for the magnitude of the tangent vector to the v-parame-
absolute value of the cross product of two vectors. ter curves. In this case the vector has a magnitude
Therefore, the differential area of the parallelo- (c2 u2)1/2. From the general parametric represen-
gram can be calculated as: tation for a circular helix (3.2) we consider a par-
ticular example that has a radius uo and pitch c
dA |dc (u, vo) dc (uo, v) | | s s

u v
dudv | (Lipschutz, 1969, p. 63):

s s s s 12 c(t) (uo cos t)ex (uo sin t)ey (ct)ez (3.97)
dudv (3.92)
u v u v
Here t is the single arbitrary parameter for the
An identity relating scalar and vector products curve, and the tangent vector is:
can be used to convert this equation to an equa- dc
tion that involves the coefcients of the rst fun- (uo sin t)ex (uo cos t)ey (c)ez (3.98)
dt
damental form (Lipschutz, 1969, p. 10):
Comparing these equations to (3.77) and the
(a b) (c d) (a c)(b d) (a d)(b c) (3.93) second of (3.78) it is clear that each v-parameter
Substituting the appropriate vector cross prod- curve on the helicoid is a helix with radius uo and
pitch c. Because neither tangent vector (3.78) is
ucts into this identity yields:
zero but the coefcient F 0, the u- and v-parame-
s s

u v

s s

u v ter curves are orthogonal everywhere on the heli-
coidal surface. The set of straight lines and the set
of helixes cover the helicoidal surface with an

s s

u u s s

v v

s s

u v s s

u v orthogonal network.
The arc length of any curve on a surface is cal-
EG F2 (3.94) culated using (3.89) and for the helicoid this
Combining these results the differential area of reduces to:
the curved surface is dA [EG F2]12 dudv. t2
dt (c u ) dt
du 2 dv 2 12
Because EG F2 0, the square root of this quan- s 2 2 dt (3.99)
tity is a real number. The area of a curved surface t1
is found by integrating the differential area over
the surface: For a coordinate line v constant which maps
onto the helicoid as a u-parameter curve with u
A (EG F ) 2 12
dudv
(3.95)
t, we have du/dt 1, dv/dt 0. Taking the limits as
t1 b and t2 b we nd s 2b, just what is
The surface area is a function of the coefcients of expected for a straight line. For a coordinate line
the rst fundamental form and, in turn, of the u constant uo that maps onto the helicoid as a
parameters u and v. v-parameter curve (a helix) with v t, we have
Consider the helicoidal surface (3.77), which du/dt 0, dv/dt 1. Taking the limits as t1 0 and
we have identied as a geometric model, for twist t2 we nd the length of this segment of the
hackle (Figs. 3.22 and 3.23). The coefcients of the helix, s (u2o c2)1/2. Along the z-axis uo 0 and s
rst fundamental form are found using (3.85) c, which is the length of the surface (Fig. 3.21).
such that: The area of the patch of a helicoid that we take
as a fracture surface (Fig. 3.21) is found using the
E 1,F 0,G c2 u2 (3.96) coefcients of the rst fundamental form in (3.95)
In general E is the squared magnitude of the with the limits b u b and 0 v . From
tangent vector to the u-parameter curves. This symmetry this is equivalent to twice the area
vector for the helicoid is a unit vector that lies in using the range 0 u b such that:
the (x, y)-plane and is not a function of u, so the u- b
parameter curve is a straight line perpendicular to A2 c u dudv

2 2
(3.100)
the z-axis (Fig. 3.21). Similarly, G is the squared 0 0
length c. Taking the n fractures as a model for

Relative surface area, A(n)/A(1)
1
0.9 the twist hackle in the fringe region of a joint
0.8 (Fig. 3.22) the non-intuitive result is that the
0.7 p /180 surface area decreases as the number of frac-
p / 60
0.6 p / 20 tures increases (Fig. 3.27). On this gure each
p/6
0.5 curve corresponds to a different twist angle, .
0.4 For a twist angle of 1 ( /180) the surface area
0.3 of ten fractures is 99.5% of that for the single
0.2 fracture, only marginally less. However, for a
0.1 twist angle of 30 ( /6) the surface area of ten
0 fractures is 36.1% of that for the single fracture,
0 5 10 15 20 25 30 35 40 45 50 dramatically less. Because the energy required to
Number of partial fractures form a fracture in brittle materials scales with
the fracture surface area (Lawn and Wilshaw,
Fig 3.27 Graph of relative surface area versus number of
1975) this result shows that the breakdown of
partial fractures for helicoidal model of fringe fractures.
Relative surface area decreases with number of fractures. joints into hackle with helicoidal shapes is con-
Reprinted from Pollard et al. (1982) with permission from sistent with a condition of lesser energy
The Geological Society of America. expended during propagation.
3.2.6 The second fundamental form,

The inner integral is evaluated using (Selby, 1975, surface shape, and normal
p. 424): curvature
The second fundamental form provides a measure
1

a2 x2 dx xa2 x2 a2 ln (x a2 x2)
2
of the shape at any point on a continuous curved
surface. To understand how this is accomplished
(3.101) we focus on a very small part of the parameter
plane so lengths along the coordinate axes are
The area of the helicoidal fracture surface is: measured using the differential quantities du and
dv (Fig. 3.28a). Consider the arbitrary curve u

A b b2 c2 c2 ln b b2 c2
c (3.102) u(t), v v(t) in the parameter plane which maps to
the curve c[u(t), v(t)] on the curved surface. At an
As the spatial rate of twist goes to zero, 1/c 0, the arbitrary point along this curve the differential
area goes to that of a rectangular plane, A A0. tangent vector, dc, is dened by (3.83) and this
For unit half-width (b 1) and unit length ( c 1), vector lies in the tangent plane to the surface (Fig.
we have A0 2, and the normalized area, A/A0, 3.28b). The unit vector, N, at this arbitrary point is
increases slowly and non-linearly with rate of a function of the two parameters u and v, such
twist. that the differential is:
Using the dimensionless ratio b/c in (3.102), the
N N
area of the helicoidal fracture surface takes the dN du dv (3.104)
u v
form (Pollard et al., 2004):
The vector dN is a measure of the change in ori-

2 2
A c2 b b 1 ln b b 1 entation of N with position along the curve on the
c c c c surface and, in this sense, it is a measure of the
shape of the surface. Also, because N is constant in
(3.103)
magnitude, the vector dN is orthogonal to N and
This relationship demonstrates that the surface therefore lies in the tangent plane (Lipschutz,
area of n helicoidal fractures, each of half-width 1969).
b and length c, is less than the surface area of Although dc and dN both lie in the tangent
a single helicoidal fracture of half-width nb and plane of the surface (Fig. 3.28b), these vectors
(a) dv
Parameter plane II dN dc
N
u
N
du dv
v
s
u
s
du dv
v
p(u, v)
u = u(t),

N s

u u
du2 N s N s

u v v u
dudv
v = v(t) N s

v v
dv2 (3.105)
du The coefcients in this equation are given partic-

O ular symbolic names: L, M, and N. Using these
O s(u, v) symbols the second fundamental form is written:
N
(b)
dN II Ldu2 2Mdudv Ndv2
c[u(t), v(t)] dc
L N s

u u
, M
1 N s N s

2 u v v u
,
z v)
N
N s

v v (3.106)
(u,
p The quantities L, M, and N are functions of the two
y
Tangent parameters u and v, and are called the coefcients
x plane of the second fundamental form of the surface.
Do not confuse the scalar quantity N, and the
(c) Osculating
vector quantity N. The coefcients depend upon
plane N n(t)
the choice of parameters used to represent the
u surface, but the second fundamental form itself is
invariant with respect to this choice (Lipschutz,
c[u(t), v(t)] t(t) 1969), and in this sense II is a property of the
surface. Note that II characterizes the changing
shape of the surface in all directions at a particu-
z v) lar point and that the differential parameters du
u,
p( and dv dene the direction.
y k(t)
The coefcients of the second fundamental
x form (3.106) may be rewritten in a different way
that is useful for computations. The unit normal
Fig 3.28 Diagrams to define second fundamental form for vector, N, to the surface s(u, v) is perpendicular to
a surface. (a) Parameter plane with arbitrary curve uu(t), v the vectors that are tangent to the u- and v-par-
v(t). (b) Surface with curve, c[u(t), v(t)], unit normal vector, ameter curves (Fig. 3.19b). Therefore, for example,
N, differential tangent vector, dc, and differential normal the scalar product of N and the tangent vector
vector, dN. (c) Surface with osculating plane containing the s/u is zero, and this product may be expanded as
unit principal normal vector, n(t), and curvature vector, k(t). follows:

The Geological Society of London. s s 2s N s
N 0 N N 2
u u u u u u
(3.107)
are not necessarily parallel to one another. The
shape of the surface in the particular direction Rearranging the right-hand side of this expression
specied by dc is characterized by the scalar we have:
product of the two vectors, dN and dc, and this
product is used to dene the second fundamen-
tal form, II:
N
2s
u2

N s
u u
L
(3.108)
Similar expansions yield the other coefcients of

the second fundamental:
2s 2s 2s
LN 2
,M N ,N N 2 (3.109)
u uv v
These coefcients are the scalar product of the
unit normal vector and the respective second
partial derivatives of the surface, s(u, v).
One can use the coefcients of the second fun-
damental form to characterize the shape of a
surface in the vicinity of a particular point as
follows (Lipschutz, 1969):

0, elliptic point
LN M2 0, parabolic point (3.110)
0, hyperbolic point
L M N 0,planar point
For the parabolic point not all of the coefcients

are zero but the combination LN M2 is zero. The
three non-planar characteristic shapes are illus-
trated in Fig. 3.29. For the elliptic point the local
surface lies entirely on one side of the tangent
plane to that point. Planes that are parallel to the
tangent plane and intersect the local surface cut
out elliptical curves. For the parabolic point the
local surface is cylindrical and may lie on one or
both sides of the tangent plane to that point.
Planes that are parallel to the tangent plane inter-
sect the local surface in one or two straight lines.
For the hyperbolic point the local surface lies on
both sides of the tangent plane to that point. The
local surface intersects the tangent plane along
two lines where the surface passes from one side
to the other of the tangent plane. For the special
case where all the coefcients of the second fun-
damental form are zero, the local surface is
planar.
As an example consider the coefcients of the
second fundamental form for the helicoidal
surface (Fig. 3.21). The unit normal vector is given
in (3.79) and the second partial derivatives of
s(u, v) are found from (3.78) to be:
2s 2s
2
0, ( sin v)ex ( cos v)ey,
u uv
2s Fig 3.29 Wire-frame images of three characteristic shapes
(u cos v)ex (u sin v)ey (3.111)
v2 of a surface near an arbitrary point: (a) elliptic, (b) parabolic,
and (c) hyperbolic. Reprinted from Pollard et al. (2004) with
Taking the scalar product of the unit normal
permission from The Geological Society of London.
vector and the respective derivatives as indicated
in (3.109) the coefcients of the second funda- curvature of this curve times the cosine of the
mental form for the helicoidal surface are: angle between n and N. If the osculating plane of
the curve contains the unit normal vector for the
L 0, M c c2 u2, N 0 (3.112)
surface, then n and N are parallel, and n . On
Using (3.110) the local shape of the helicoidal the other hand, if the osculating plane of the
surface is determined by the sign of LN M2 curve is parallel to the tangent plane for the
c2/(c2 u2) 0. With the exception of the case c surface, then n 0.
0, which describes a planar surface, the sign of The relationship given in (3.114) illustrates the
this quantity is negative, so every point on the fact that the curvature of an arbitrary curve at a
helicoidal surface is hyperbolic. point on a surface is greater than or equal to the
Recall from (3.21) that the shape of a curved normal curvature of the surface in the direction
line is characterized, in part, using the curvature of the curve at that point. Familiar examples are
vector, k, and the scalar curvature, |k|. For a the circles of latitude and longitude on a sphere
curve on a surface s(u, v) two analogous measures of radius R (Fig. 2.1a). The normal curvature in any
of shape are the normal curvature vector, kn, and direction at any point on the sphere is a constant,
the normal curvature, n. Both of these quantities n 1/R. Circles of longitude are the intersection
are dened by considering an arbitrary curve u of the sphere with planes that pass through the
u(t), v v(t) in the parameter plane (Fig. 3.28a) center and the poles. These curves have the same
which maps to the curve c[u(t), v(t)] on the surface radius as the sphere and therefore their curvature
(Fig. 3.28c). At a point on this curve the curvature is 1/R. However, circles of latitude, except the
vector k is a function of the parameter t; it lies in equatorial circle, have a lesser radius, () R,
the osculating plane of the curve; and it extends and therefore a greater curvature, 1 ().
away from the concave side of the curve. The oscu- These circles are the intersections of the sphere
lating plane of a curve is the plane that contains with planes that are parallel to the equatorial
both the unit tangent vector, t(t), and the unit prin- plane and do not pass through the center. The
cipal normal vector, n(t). The unit normal vector, unit principal normal vector, n(t), for these circles
N, is perpendicular to the surface; it is a function and the unit normal vector, N, for the sphere are
of the parameters u and v; and it may not lie in the not parallel. As the circles of latitude approach
osculating plane of the curve. The normal curva- the poles of the sphere, their radii of curvature
ture vector, kn, and the normal curvature, n, are approach zero, the osculating plane of the circle
dened in terms of k(t) and N(u, v) as: approaches the tangent plane of the sphere, and
the curvature of the circle becomes greater and
kn (k N)N,n k N (3.113)
greater. This exemplies the fact that curves on
Because N is a unit vector we understand from surfaces provide the direction in which the
(3.113) that n, is a scalar quantity equal to the normal curvature of the surface is measured, but
component of k along N. Also, kn is a vector of the normal curvature is not necessarily equal to
magnitude n with the same, or the opposite, the curvature of the curve.
direction as N. If one chooses two differently directed curves
Recall that the direction of the unit principal through the same point on a surface (Fig. 3.30a),
normal vector, n(t), is chosen for consistency along the respective values of the normal curvature, n,
the curve (Fig. 3.11). If we choose the direction of for the surface may be different. On the other
n(t) for the arbitrary curve c[u(t), v(t)] on the surface hand, the curvature, , at a point on a curve is a
s(u, v) (Fig. 3.28c) such that the angle, , between unique property of the curve. The normal curva-
n(t) and N(u, v) is in the range 0 /2, then ture, n, at a point on a curved surface varies in a
(Lipschutz, 1969): smooth and systematic manner with the direc-
tion of the tangent line through the point of inter-
n cos ,0 2 (3.114)
est, from a maximum value, 1, to a minimum
In other words, the normal curvature associated value, 2. These two values of normal curvature, 1
with a particular curve on a surface is equal to the and 2, are called the principal normal curvatures.
to 2
(a) N ne
n t li K 2
e h
ng wit 1.5
1
Ta rve
c u
1 2
Tangent line
Normal curvature
to
curve with K 3
n 0.5
a
0 4
v)
Ta rve
u, cu
ng w
z (
en ith
p 0.5 5
tl K1
in
y
e
1
to
x
1.5
(b) N
c(u, vo ) s 2
0 45 90 135 180 225 270 315 360
v
Alpha ()
T
c[u(t), v(t)] Uo Fig 3.31 Graph of normal curvature versus angle with
z examples for elliptical dome (1), parabolic antiform (2),
hyperbolic saddle (3), parabolic synform (4), and elliptical
s
y c(uo, v) basin (5). Reprinted from Pollard et al. (2004) with
u permission from The Geological Society of London.
x
Fig 3.30 Diagrams to define variation of normal curvature shapes illustrated in Fig. 3.29. We use the Monge
with direction at a point on a surface. (a) Angle measured
patch (3.58) where the (x, y)-plane is the horizontal
in the tangent plane from direction of maximum principal
plane to give the surface an orientation (up or
normal curvature, 1, to direction of normal curvature n.
(b) Angle o measured in the tangent plane from tangent to down) and relate it to geological structures. For
u-parameter curve to direction of principal normal the sake of this illustration we consider particular
curvature. cases where 1 2 1. The elliptic paraboloid
(3.60) in this context is analogous to the surfaces
of sedimentary beds in a basin-shaped structure
The variation of normal curvature with direc- (Fig. 3.17b). For points where both principal cur-
tion is of the same form for all surfaces with vatures are positive (concave upward) or negative
continuous second partial derivatives such that (concave downward) the shape is elliptical and the
(Lipschutz, 1969): structure is a basin or a dome. For points where
n 1 cos 2 2 sin 2 (3.115) one principal curvature is zero and the other is
positive or negative the shape is parabolic and the
This relationship is known as Eulers Theorem and structure is a cylindrical synform or antiform. For
the angle is measured in the tangent plane from points where the principal curvatures are of dif-
the direction of the tangent line corresponding ferent signs the shape is hyperbolic and the struc-
to the curvature 1 to that corresponding to n. ture is a saddle.
The directions of the tangent lines associated Because the normal curvature is a property of
with the extreme values of normal curvature are a surface at any point, we anticipate that it can be
called the principal directions of normal curvature written as a function of the fundamental forms.
and they are orthogonal. The derivation uses (3.21) to replace the curvature
The signs of the principal normal curvatures vector in (3.113) with the derivative of the tangent
are related to the shape and orientation of a vector written as a function of the arbitrary par-
surface in the vicinity of a point as dened in ameter t (Lipschutz, 1969):
(3.110). For example in Fig. 3.31 the distributions
of normal curvature n at a point as a function of
the angle are plotted for the fundamental
n k N dt
dt
N | |
dc
dt
(3.116)
Because the tangent vector, t, is perpendicular to dened in (3.77). The coefcients of the rst and
the normal vector, N, the derivative of their scalar second fundamental forms were derived from
product is zero: (3.85) and (3.106) above such that:
d dt dN 2c
(t N) N t 0, I (du)2 (c2 u2)(dv)2,II (dudv)
dt dt dt c2 u2
dt dN
so N t (3.117) (3.121)
dt dt
Using the last equation in (3.116) and substituting In the form of (3.119) the normal curvature for the
(3.8) for the tangent vector we have: helicoidal surface is:

n t
dN
dt /| |
dc
dt

dc dN

dt dt / dc dc

dt dt n
2c du dv

c2 u2 dt dt

(3.118) du 2 dv 2 (3.122)
(c2 u2)
dt dt
The derivatives of c and N with respect to t may be
rewritten as derivatives of s and N with respect to On a u-parameter curve u t, du/dt 1, and dv/dt
u and v using (3.83) and (3.104). Comparing the 0, so the normal curvature is n 0, just what one
results with the expressions for the coefcients of would expect for a straight line. Recall that such a
the fundamental forms, (3.85) and (3.106), leads to straight line is the generating line for the heli-
the denition of the normal curvature in terms of coidal surface when it is moved perpendicular to
these coefcients: itself and rotated about the z-axis which is the
mid-line of the helicoid (Fig. 3.21).
L(dudt)2 2M(dudt)(dvdt) N(dvdt)2 A v-parameter curve on the helicoidal surface
n
E(dudt)2 2F(dudt)(dvdt) G(dvdt)2 (Fig. 3.21) is a circular helix with radius uo and
(3.119) pitch c. On this curve v t, du/dt 0, and dv/dt 1,
From the discussion of (3.66) and Fig. 3.19 recall so the normal curvature is n 0. The curvature
that the ratio of the derivatives dv/dt and du/dt vector, k, for the helix is not zero (3.22) and is
determine the direction of the tangent line to the directed in the (x, y)-plane toward the z-axis. The
arbitrary curve c[u(t), v(t)] at the point in question, unit normal vector, N, for the helicoidal surface
so the normal curvature depends upon this direc- along a v-parameter curve is orthogonal to the cur-
tion. In addition the normal curvature depends vature vector, so the scalar product in (3.113)
upon the coefcients of the rst and second fun- denes the normal curvature as k N n 0. In
damental forms. other words the curvature vector for the helix
The normal curvature may be written in terms does not resolve any component onto the line
of the differentials, du and dv, using (3.119) normal to the helicoidal surface. This result is
(Lipschutz, 1969): non-intuitive because the v-parameter curve
clearly has a non-zero curvature, but the normal
L du2 2M dudv N dv2 II
n (3.120) curvature of the surface along the tangent line to
E du2 2F dudv G dv2 I
this curve is zero.
The ratio of the differentials, dv : du, determines
the direction of the tangent line to the arbitrary 3.2.7 Principal normal curvatures,
curve c[u(t), v(t)], and these differentials are Gaussian, and mean curvature
referred to as the direction numbers of the tangent Eulers Theorem (3.115) is used to calculate the
line. From (3.120) we understand that the normal normal curvature n in the direction of any line
curvature n at an arbitrary point on the surface tangent to a surface, s(u, v), given the principal
s(u, v) in the direction of this tangent line is equal normal curvatures, 1 and 2, at a point on the
to the ratio of the second to the rst fundamental surface. Here we describe how to calculate the
form. magnitudes of the two principal normal curva-
To illustrate the concept of normal curvature, tures and the principal directions. Recall from cal-
consider again the helicoidal surface (Fig. 3.21) as culus that the maximum and minimum values of
a function of two variables are found by setting for the coefcients of the fundamental forms,
the partial derivatives of the function to zero. (3.96) and (3.112), to nd tan o 1 c2 u2. Here
From (3.120) we note that the normal curvature, o is the angle measured in the tangent plane
n, is a function of the two differential quantities, to the surface from the tangent line for the u-
du and dv, which determine the direction of the parameter curve to the tangent line for the two
tangent line on the surface at a particular point. principal directions (Fig. 3.30b). Along the mid-
Thus, the derivative of n with respect to each of line, u 0, of the helicoidal surface (Fig. 3.21) the
these direction numbers is set equal to zero and tangent of the principal directions is equal to the
evaluated for the principal directions, identied spatial rate of twist, 1/c.
by duo and dvo: The magnitudes of the principal normal cur-
vatures, 1 and 2, are found by rearranging
n
du | (duo, dvo) 0,
n
dv | (duo, dvo) 0 (3.123) (3.124) to factor out the two differentials duo and
dvo:
We substitute n II/I, use the formula for the
derivative of a quotient, and write the partial (L oE)duo (M oF)dvo 0
(3.127)
derivatives of I and II in terms of their coefcients (M oF)duo (N oG)dvo 0
using (3.85) and (3.106). When the resulting
expressions are evaluated for the principal direc- These two linear equations have a simultaneous
tions, duo and dvo, the normal curvature takes on solution (duo, dvo) if the determinant of the
extreme values, o II/I, satisfying the following coefcients of the left side is zero. Expanding the
linear equations (Lipschutz, 1969): determinant produces a quadratic equation in o:
(Lduo Mdvo) (Eduo Fdvo)(o) 0 (EG F2)2o ( EN 2FM GL)o

(3.124)
(Mduo Ndvo) (Fduo Gdvo)(o) 0 (LN M2) 0 (3.128)
These equations have a simultaneous solution (1, For example, the magnitudes of the principal
o) if the determinant made up of the coefcients normal curvatures for the helicoidal surface (3.77)
on the left-hand side is zero. Expanding the deter- are found by substitution of (3.96) and (3.112) to
minant produces a quadratic equation in the nd 1, 2 c(c2 u2). The two principal curva-
direction numbers for the principal directions, tures are equal in magnitude and opposite in sign.
duo and dvo: They are independent of the parameter v, and so
are constant along any particular circular helix
(LF ME)(duo)2 (LG NE)duodvo
that is a v-parameter curve. Along the mid-line the
(MG NF)(dvo)2 0 (3.125)
principal curvatures are equal in magnitude to
The ratio of the direction numbers, dvo : duo tan the spatial rate of twist, 1/c. For a given rate of
o, and this determines the angles, o and o /2, twist the principal curvatures decrease in magni-
between the tangent to the u-parameter curve and tude with distance, u, from the mid-line.
the tangent to the principal directions (Fig. 3.30b). Equation (3.128) may have two real and
Dividing (3.125) by (duo)2 and substituting for unequal roots, 1 and 2, or two real and equal
the ratio of direction numbers, we nd a qua- non-zero roots, or two zero roots (Lipschutz, 1969).
dratic equation in tan o: The second case pertains to elliptical points (Fig.
3.29a) at which the normal curvature is non-zero
(MG NF) tan 2o (LG NE) tan o
but the same in all directions and the ratios of
(LF ME) 0 (3.126)
respective fundamental coefcients are constant:
The principal directions for the normal curvature
L M N
at a point on a surface are found from (3.126) n constant (3.129)
E F G
using the standard formula for the solution of a
quadratic equation. Using the helicoidal surface This is referred to as an umbilical point. The third
(3.77) as an example, the principal directions of case is the planar point where the normal curva-
the normal curvature are found by substituting ture is zero in all directions.
Kg < 0 Kg = 0 Kg > 0 circular helix. Along the mid-line of the helicoidal

surface, u 0, the Gaussian curvature is a con-
stant equal to the negative of the squared rate of
Km < 0 twist.
We have identied the three possible non-
Synform Basin planar shapes (hyperbolic, parabolic, and elliptic)
in the vicinity of a particular point on a surface
Km = 0 based on the sign of the numerator in (3.132) and
Saddle Plane these shapes are illustrated in Fig. 3.29. Because
the denominator in (3.132) always is positive, the
sign of the Gaussian curvature is determined by
Km > 0
the numerator. Thus the sign of g may be used to
Antiform Dome distinguish these three non-planar shapes. Taken
together, the signs of the Gaussian and the mean
Fig 3.32 Table of six different characteristic shapes of
curvature may be used to categorize surfaces with
geological surfaces near an arbitrary point categorized by the
signs of the Gaussian curvature, g, and mean normal respect to orientation in keeping with geological
curvature, m. Reprinted from Bergbauer and Pollard (2003) conventions (Roberts, 2001). Consider the Monge
with permission from Elsevier. patch (3.58) with the (x, y)-plane horizontal and
positive z upward. Six shapes are distinguished in
Fig. 3.32 by noting that the sign of the mean cur-
Dividing each term of (3.128) by the rst term vature is different for elliptic shapes that are
in parentheses we have: domes and basins, and for parabolic shapes that
are cylindrical antiforms and synforms (Berg-
2o 2mo g 0 (3.130)
bauer and Pollard, 2003). Therefore, for g 0 the
The rst constant, m, is the average of the two surface is elliptic; it is a dome if m 0 and a basin
principal normal curvatures and is referred to as if m 0. For g 0 the surface is parabolic; it is
the mean principal normal curvature: antiformal if m 0, planar if m 0, and synfor-
mal if m 0. For g 0 the surface is hyperbolic.
(EN 2FM GL) 1
m ( 1 2) (3.131) This categorization provides a simple way to
2(EG F 2) 2
describe geological surfaces using the concepts of
The second constant, g, is the product of the prin- differential geometry.
cipal curvatures and is called the Gaussian curva-
ture:
(LN M2) 3.3 Applications of differential
g 1 2 (3.132)
(EG F2) geometry to structural geology
For example, the mean curvature for the heli-
coidal surface is found by substituting the In the introduction to this chapter we stated that
coefcients of the fundamental forms, (3.96) and a primary task for structural geologists is to
(3.112), into (3.131) to nd m 0. The fact that the describe and characterize the lineations and sur-
mean normal curvature is zero is consistent with faces that make up the structures we use to
the two principal curvatures being equal and unravel the history of deformation in a region and
opposite in sign. Surfaces that satisfy the condi- to understand how the rocks in Earths crust
tion of zero mean curvature are called minimal sur- deform. We noted, with some surprise, that rela-
faces. The Gaussian curvature for the helicoidal tively little use of differential geometry is found
surface is found by substituting (3.96) and (3.112) in the twentieth-century literature of structural
into (3.132) to nd g c2(c2 u2)2. The Gaussian geology, despite the fact that this mathematical
curvature is not a function of the parameter v and subject provides the only rigorous, complete and
therefore is constant along a given v-parameter self-consistent method to describe and character-
3.3 APPLICATIONS TO STRUCTURAL GEOLOGY 115
ize geological structures. A keyword search of surface that are everywhere parallel to the local
GEOREF, a geological reference data base in July direction of relative motion, and then use the cur-
2003 using the exact phrase differential geome- vature and torsion to characterize the shapes of
try uncovered only fourteen references from these curves.
1963 to 2003. Of these ten are applications to To illustrate the fact that slickenlines do vary
other geoscience subjects such as gravity and systematically with position on a fault, and to
geodesy, seismology, hydrogeology, and paleon- appreciate some of the challenges inherent to the
tology. The four remaining consider the three- investigation of supercial lineations we turn to a
dimensional geometry of folds, the deformation data set from the Chimney Rock fault array (Fig.
of Earths surface due to neotectonic crustal 2.29). The four sets of faults in this region are dis-
motion (Zakarevicius, 2000; Grachev et al., 2001), played on the structure contour map constructed
and the normal curvature of geological surfaces on the base of the Carmel Formation. Note, for
(Bergbauer and Pollard, 2003). A search on the example, that individual contours on this map are
phrase Gaussian curvature again uncovered truncated by the Frenchman Fault. When traced
fourteen references including three on the esti- to the north across the fault the sense of step is
mation of strain and the prediction of fractures consistently to the east. However, the magnitude
within folds (Lisle, 1994, 2000; Ozkaya, 2002). of the step decreases toward both terminations of
This section includes examples that provide this fault. This change in step magnitude suggests
insights into how differential geometry can be that the magnitude of the slip decreases from the
applied to problems in structural geology. These mid-section of the Frenchman Fault toward the
examples are works in progress and we expect terminations and, indeed, the slip must go to zero
more details to emerge during on-going studies. at the terminations by denition.
None-the-less we hope that these examples will The distribution of dip slip is plotted versus
encourage others to apply differential geometry position along the trace of the Blueberry Fault in
to structural problems. Fig. 2.31. Note how the magnitude of the dip slip
(gray boxes) increases from zero at the eastern ter-
3.3.1 Characterizing the shapes of mination to more than 30 m near the middle of
lineations on discrete surfaces the fault. The slip distribution is not continuous,
Lineations are found on discrete geological sur- but jumps abruptly where members of other fault
faces such as faults and intrusive contacts (Fig. sets intersect the Blueberry Fault. The distribu-
3.3). These supercial lineations typically are tion of slickenline rake (black diamonds) is
aligned on an exposure such that multiple mea- plotted versus position along the trace of the
surements of the orientations of linear elements Blueberry Fault. The rakes are approximately 90
would have a standard deviation of few degrees. (down dip) near the northeastern and southwest-
However, faults and intrusive contacts may pass ern terminations of the fault, but decrease more
through rocks with different mechanical proper- or less systematically to about 70 as one
ties, they may be inuenced mechanically by adja- approaches the intersections with the La Sal and
cent faults or intrusions, and they may be curved Little Faults. Across these faults the rakes abruptly
surfaces. Thus one should expect supercial lin- increase. The rakes are systematically greater than
eations to vary in orientation over the surfaces on 90 (inclined toward the east) between the inter-
which they are found. On a fault, for example, sections with the La Sal and Little Faults, and less
slickenlines should form a systematic pattern that than 90 (inclined toward the west) on the distal
reects the relative motion of the two surfaces sides of these intersections. These changes reect
during frictional sliding. Of course the direction the mechanical interaction of the faults and are
of relative motion at a point may change as a fault consistent with elastic models of this interaction
develops, leading to overprinting of slickenlines (Maerten, 2000).
with different orientations. Where overprinting is Exposure of the faults at Chimney Rock are
not an issue, one should be able to dene a set of adequate to document the lateral variation in
three-dimensional curves lying on the fault rake of the slickenlines over a distance of almost
Small ~N
hut D 26
C C- B 20
B-
A2
Hinge line
A1
Fig 3.33 Oblique aerial photograph of the nose of the

tion in curvature and torsion of these curved lines
Emigrant Gap anticline, WY. Reprinted from Bergbauer and
completely characterizes the geometry of relative
Pollard (2004) with permission from The Geological Society
of America. slip on the model fault. The local orientations of
these curves can be compared with the exposure
data for consistency.
3 km, but inadequate to document the vertical
variation. However, model fault surfaces can be 3.3.2 Characterizing the shapes of folded
constructed and elastic boundary value problems surfaces
can be solved to nd the variation in slip direction The Emigrant Gap anticline is a doubly plunging
with position over the entire model fault surface. fold exposed for about 30 km along a north-
Then curves can be dened such that their westsoutheast trend near Casper, Wyoming
tangent line is everywhere parallel to the local rel- (Bergbauer and Pollard, 2004). Sandstone beds of
ative slip direction. The three-dimensional varia- the Frontier Formation crop out on the fold limbs
and are continuously exposed around the fold
N hinge (Fig. 3.33). The top of the lowest-most sand-
stone bed (labeled A1) of the Frontier Formation
b (345, 4) denes a somewhat asymmetric anticlinal
surface, plunging gently to the north, with
approximately planar limbs and a rounded hinge.
The arc length of the exposed surface is on the
order of 500 m, and the amplitude is about 75 m.
On a stereographic projection (Fig. 3.34), 543 poles
W E to bedding dene a great circle (labeled ), with a
p tighter cluster of poles representing the eastern
limb and a broader cluster representing the
n = 543 western limb. The pole to the great circle (labeled
) denes an approximate fold axis with trend
and plunge of 345, 04.
The top of the A1 sandstone was sampled at
2529 points using a TrimbleTM Pro XL GPS receiver
S with a vertical precision of 0.5 to 1.5 m and hori-
zontal precision of less than 0.7 m. Fig. 3.35a is a
Fig 3.34 Stereographic projection of 543 poles to Frontier map of the GPS locations on the A1 surface with a
Formation sandstone beds on the Emigrant Gap anticline. local coordinate system in which x easting, y
Reprinted from Bergbauer and Pollard (2004) with
northing, and z up. The map region is about
permission from The Geological Society of America.
500 m wide and 2300 m long. The limbs of the
3.3 APPLICATIONS TO STRUCTURAL GEOLOGY 117
(a) (b)
100
z (m)
8 50
0
0 100 200 300 400 500 600
100
2000
z (m)
7 50
0
0 100 200 300 400 500 600
100
z (m)
6 50
1500 0
0 100 200 300 400 500 600
100
z (m)
5 50
0
0 100 200 300 400 500 600
100
4
z (m)
1000 50
0
y (m)
0 100 200 300 400 500 600

100
z (m)
3 50
0
0 100 200 300 400 500 600
500 100
2
z (m)
50
0
0 100 200 300 400 500 600
100
z (m)
1 50
0
0 0 100 200 300 400 500 600
z 0 500 x (m)
x (m)
Fig 3.35 (a) Map view of Emigrant Gap anticline with 2529 not exposed. Eight cross sections of the folded
GPS survey points (Bergbauer, 2002). (b) Eight profiles of surface were used to constrain the shape of the
fold shape constrained by GPS data. surface (Fig. 3.35b). From these cross sections one
can infer that the anticline is asymmetric, with
folded surface could be sampled only in the south- the west ank exhibiting steeper dips than the
ern part of the map area, and the hinge only is east ank, and the hinge is rounded. The cross sec-
exposed and sampled near the northern terminations were constructed using GPS points located
tion of the outcrop. The GPS data were used to within 150 m of each cross section. Cross sections
create a digital model of the surface, despite the 1 through 6 are weakly constrained near the fold
limitation that the A1 sandstone is exposed only hinge, whereas cross sections 7 and 8 are weakly
over about 25% of the map area. Clearly, sufcient constrained on the fold limbs.
data control is lacking to make conclusive state- A model surface was constructed by interpola-
ments about the geometry where the surface is tion and ltering (Bergbauer and Pollard, 2004)
(a) kmin (b) kmax

x 103
3.5
2000 2000
3
2.5
1500 1500
2
Curvature (m1)
y (m)
y (m)
1.5
1000 1000
1
0.5
500 0 500
0.5
Hinge line
Crest line
0 1 0
0 500 0 500
x (m) x (m)
Fig 3.36 (a) Map view of Emigrant Gap anticline with
contours of minimum principal normal curvature, 2, and tick parallel to the trend of the fold hinge, and
marks parallel to direction of minimum curvature (Bergbauer, signicantly change directions only across the
2002). (b) Contours of maximum principal normal curvature, gentle surface undulations. The directions of
1, and tick marks parallel to direction of maximum maximum curvature (white ticks, Fig. 3.36b) are
curvature. approximately perpendicular to the fold hinge.
It is standard practice in structural geology to
that is approximately cylindrical in shape and idealize folds as cylindrical structures. The prin-
exhibits a smoothly rounded hinge region. Gentle cipal normal curvatures provide a quantitative
surface undulations, which trend obliquely to the measure of the departure from a cylindrical
hinge line, are superimposed on the more or less shape. For a surface to be perfectly cylindrical one
cylindrical shape. Apart from these gentle undu- principal curvature must be zero everywhere, and
lations, the limbs of the modeled surface are the other principal curvature must have the same
approximately planar. Fig. 3.36 shows the distrib- distribution on every cross section taken perpen-
utions of the maximum, 1, and minimum, 2, dicular to the fold axis. Not only is the minimum
principal normal curvatures. Values of the principal curvature non-zero across the modeled
minimum curvature range from 1 e3 to 1 e3 bedding surface (Fig. 3.36a), the maximum curva-
m1 and values of the maximum curvature range ture distributions differ from one cross section to
from 1. 7 e4 to 3.8 e3 m1. Areas with elevated another (Fig. 3.36b). Discrimination of cylindrical
magnitudes of minimum curvature trend and non-cylindrical areas across the surface is
obliquely across the fold due to the gentle surface possible using the categorization depicted in Fig.
undulations. The directions of minimum curva- 3.32. Based on the signs of the Gaussian and mean
ture, shown as white ticks in Fig. 3.36a, are sub- curvatures at every grid point, the surface can be
3.4 CONCLUDING REMARKS 119
decomposed into areas that are locally shaped like tive geometry and stereographic projection did in
one or the other of the six different characteristic the second half of the twentieth century. This is
shapes (Bergbauer and Pollard, 2003). Performing not the intention of the authors. We view differ-
this analysis on the modeled bedding surface ential geometry as the appropriate mathematical
shows that the surface is composed primarily of machinery to characterize structures, but this
domal and saddle-like areas, which reect the characterization is just one step in an investiga-
gentle undulations superimposed on the broader tion which ultimately must include consideration
fold shape. The modeled bedding surface does of the constitutive properties of rock and models
not contain any cylindrically shaped areas of of deformation based on the equations of motion
signicant extent. We suggest that the characteri- (Guiton et al., 2003).
zation of folded surfaces using differential geom- Justications for learning differential geome-
etry will provide new insights concerning the try are several. The structures encountered in
process of folding (Fisher and Wiklerson, 2000; Earths crust are three dimensional with spatial
Lisle, 2000; Bergbauer and Pollard, 2004). variations in size and shape that only can be
accounted for using a geometry that involves the
spatial derivatives of such things as orientation and
3.4 Concluding remarks curvature. Plotting orientation data on a stereo-
graphic projection eliminates the opportunity
The objective of this chapter is to introduce struc- to visualize and analyze these spatial changes.
tural geologists to the elementary concepts of dif- Furthermore, to proceed with modeling one needs
ferential geometry that serve to characterize to write boundary conditions that refer explicitly
curves (lineations) and surfaces in three-dimen- to geometry of surfaces. Finally, we now have
sional space. One could imagine that these con- precise eld data on the three-dimensional shape
cepts and the tools that follow from them might of surfaces from new technology such as GPS and
capture the attention of structural geologists, we need to know how to describe these surfaces and
much as the concepts and tools related to descrip- how to compare them to a model result.
Chapter 4
Physical quantities, fields, dimensions,

and scaling
Earths crust under the state of Texas being lifted by a crane. In physical science a rst essential step in the direction
Scaling laws demonstrate that the good state of Texas is of learning any subject is to nd principles of numeri-
utterly incapable of self-support. Reprinted from Hubbert cal reckoning and practical methods for measuring
(1945) by permission of the AAPG whose permission is some quality connected with it. I often say that when
required for further use. you can measure what you are speaking about, and
express it in numbers you know something about it;
but when you cannot measure it, when you cannot
express it in numbers your knowledge is of a meager
and unsatisfactory kind: it may be the beginning of
knowledge, but you have scarcely, in your thoughts,
advanced to the stage of science, whatever the matter
may be (Thomson, 1891).
4.1 PHYSICAL QUANTITIES AND THE CONTINUUM 121
I
n an insightful article about research in tools to understand the scaling of structural phe-
geology in the early twentieth century M. King nomena, and to set up scaled laboratory experi-
Hubbert refers to Sir William Thomson (Lord ments to model the development of structures.
Kelvin) as the Patron Saint of geologists, includ-
ing himself, who espouse a quantitative agenda,
and he cites this quotation from Thomson as their 4.1 Physical quantities and
guiding credo (Hubbert, 1974). Thomson is not
advocating numeration purely for the sake of col-
the continuum
lecting numbers; rather this is a call to measure
relevant physical quantities and express them as 4.1.1 Fundamental and derived quantities
numbers. Thomson was a physicist, not a geolo- Structural geology is concerned with deformation
gist, but Hubbert recognized the importance of of rock and this is largely a physical process,
quantication in the geological sciences and was although chemical processes can play important
a leader among geologists of his generation in this roles. Most of the physical quantities we use in
regard (Hubbert, 1972). In part, Hubbert was react- this textbook can be described in terms of four
ing to the popularity of descriptive taxonomy for fundamental quantities for mechanical systems,
geologists of the twentieth century, structural namely length, mass, time, and temperature.
geologists being no exception: an introductory Associated with each fundamental quantity are
textbook published in 1987 provides a glossary of actual objects (e.g. a cylinder of platinumiridium
terms with over 350 entries that beginning stu- alloy residing at Svres, France, and assigned a
dents might be expected to master (Dennis, 1987). mass of one kilogram), or devices with prescribed
In a playful reaction to the plethora of terms for procedures (e.g. a device to measure the duration
intrusive forms at mid-century Charles Hunt com- of 9 192 631 770 periods of radiation correspond-
ments on the feeder to the Trachyte Mesa laccol- ing to transitions of the cesium-133 atom and
ith: Because the form has certain resemblances to assigned a time of one second; Mechtly, 1973).
the woody structure of the cane cactus the name These are used as standards to dene the quanti-
cactolith might be used and dened as a quasi-hor- ties, and copies of the standards are used for
izontal chonolith composed of anastomosing duc- everyday measurement. For example, one would
toliths whose distal ends curl like a harpolith, thin compare an unknown mass to a copy of the stan-
like a sphenolith, or bulge discordantly like an dard kilogram using a balance. For structural
akmolith or ethmolith (Hunt, 1953). geologists most measurements are made using
Here we introduce some of the concepts and classical physical principles that predate relativ-
the tools necessary to practice structural geology ity, quantum mechanics, and the physics of
in a manner that Hubbert would have understood atomic and sub-atomic particles. As two modern
and Thomson would have appreciated. We begin physicists point out:
this chapter by dening the basic physical quanti- Observations are formulated in the language of classi-
ties used to describe and measure Earth struc- cal physics because that is the language used to record
tures, and agree on their units of measure. This measurements with macroscopic instruments. That
leads to a discussion of the continuum, the math- statement does not imply that the measuring instru-
ematical idealization that forms the basis for ments follow classical physics instead of quantum
most of our thinking about the spatial and tem- physics, a wrong opinion some writers ascribe incor-
poral variations of the relevant physical quanti- rectly to Bohr. Instead our statement implies that the
ties. These so-called eld quantities are dened at special nature, in particular the larger size, of measur-
every point in the continuum and are inferred to ing instruments allows the description of their behav-
ior in classical terms (Feshbach and Weisskopf, 1988).
be measurable in the rock mass. Next we consider
physical dimensions and explain how dimen- Such measurements determine a numerical value
sional analysis is used to check the consistency of for the physical quantity.
equations and to construct graphs of physical It is not the measured number itself that is
quantities. Dimensional analysis provides the useful, but rather that number in combination
122 PHYSICAL QUANTITIES, FIELDS, DIMENSIONS, AND SCALING
Table 4.1. Physical quantities, units, and symbols.
Quantity Unit Symbol
Fundamental physical quantities and SI units

Length meter m
Mass kilogram kg
Time second s
Temperature kelvin K
Some derived quantities and SI units
Area square meter m2
Volume cubic meter m3
Displacement meter m
Velocity meter per second m s1
Acceleration meter per second squared m s2
Mass density kilogram per cubic meter k g m3
Force newton N
Traction, stress pascal Pa
Pressure pascal Pa
Work, energy joule J
Common quantities and units
Time year (annum) a
Temperature degree Celsius C
Plane angle radian rad
with appropriate units of measure. A myriad of units volume, V [] m3 (4.1)

of measure for the fundamental quantities have
been invented and many are in use today, but we acceleration, a []m s2 (4.2)
will use those now recognized as part of the
mass density, []kg m3 (4.3)
Systme Internationale dUnits or SI system
(Table 4.1). Use of the four units (meter, kilogram, thermal expansion, []K 1 (4.4)
second, and kelvin) for the fundamental physical
quantities and the development of an interna- These relationships illustrate how the units of
tional regulatory system can be traced back his- derived quantities are made up of products and
torically about two hundred years. For example, powers of the units for the fundamental quanti-
the meter and the kilogram were created by ties.
members of the Paris Academy of Sciences and There are a few derived quantities that are par-
adopted by the National Assembly of France in ticularly important and have been given special
1795 (Mechtly, 1973). The annual review by R. A. names, usually to honor a person responsible for
Nelson in Physics Today (Nelson, 2003) provides a introducing or clarifying the usage of the quan-
useful summary of metric practice, and the article tity. Perhaps the most famous person in this
by D. Kind and T. Quinn summarizes the status of regard is the English natural philosopher Sir Isaac
metrology (the science of measurement) at the end Newton (16421727) after whom the unit of force
of the twentieth century (Kind and Quinn, 1999). is named. Newtons second law F ma establishes
The units for mechanical quantities commonly the relationship among force, F, mass, m, and
used in structural geology are derived using the acceleration, a, from which the units follow:
units of the fundamental quantities. For example,
reading the symbol [] has units of we have: force, F []kg m s2 N (4.5)
Note that proper names for the SI units, such as Unit weight
newton and kelvin, are not capitalized, but the 1m = 26670 N m3
1
corresponding units themselves, N and K, are cap-
italized. The unit of stress follows from the
2
concept of a force per unit area:
stress, [](kg m s 2) m2 N m2 Pa 3

Stress on base
(4.6)
= 26670 N m2
The SI unit for stress is the pascal, named after the 1000 m
French mathematician and physicist Blaise Pascal
(162362). Equations (4.5) and (4.6) illustrate how
the special units of derived quantities can be con- 998
verted to products and powers of units of the fun-
damental quantities. 999
To put the newton and the pascal into a
geological context consider the weight per unit Stress on base of column
1000
volume of granite, one of the most common
~27 MN m2 = 27 MPa
crustal rocks. Measurement of 155 different
samples of granite produced a range from 2.516
104 N m3 to 2.809 104 N m3. These unit weights
come from Table 4.1 of Memoir 97 of the Fig 4.1 Stack of 1m cubes of rock 1km high results in a
Geological Society of America (Daly et al., 1966). vertical stress of about 27MPa.
Clearly not all rocks called granite have the same
unit weight, but we take the reported mean value
of 2.667 104 N m3 for this calculation. Thus, one quantity with the least signicant gures. Thus,
cubic meter of this granite weighs 2.667 104 N at when the weight of the granite cube is divided by
sea level. The sea level weight of one of the authors the authors weight, the quotient is rounded off to
of this textbook in archaic units is 140 pounds three signicant gures. When adding or sub-
force, which is 6.23 102 N. Thus, the weight of tracting a set of numbers, they are arranged by
the granite cube is greater than that of the author place (hundreds, tens, ones, tenths, etc.) and the
by a factor of 4.28 101. result is rounded off to the least place that con-
Implicit in the preceding paragraph are two tains signicant gures in all the numbers of the
concepts, scientic notation and signicant gures set.
that are standard practice for a scientist working Now imagine the cube of granite positioned
with numerical data. For example, the mean unit below 999 other such cubes (Fig. 4.1) and calculate
weight of granite is given in scientic notation the force per unit area (stress) acting on the
with four signicant gures. In scientic notation bottom of this granite column:
a value is represented by a number, with only one stress, (2.667 104 N m3)(1.000 103 m)
digit to the left of the decimal place, multiplied by
2.667 107 Pa (4.7)
a power of ten. The power indicates the order of
magnitude of the quantity. The total number of The unit weights reported for sedimentary,
digits to the left and right of the decimal place is metamorphic, and igneous rocks in Memoir 97
the number of signicant gures. They are called range from 1.44 104 N m3 (sandsiltclay) to
signicant because they recur consistently during 3.392 104 N m3 (eclogite). We infer that the
repeated measurements. The weight of the author stress acting in the vertical direction at one thou-
has three signicant gures. sand meters depth in the Earth is likely to fall in
When multiplying or dividing two quantities the range from about 14 to 34 million pascals. This
in scientic notation the number of signicant inference neglects the possible mechanical con-
gures of the result is the same as that of the straints that the surrounding rock might place on
such a column, but we will evaluate those lateral Table 4.2. Selected SI prefixes and symbols.
constraints in a later chapter. A handy rule of
thumb is: the vertical stress in the Earths crust Prefix Symbol Multiple of
due to the weight of overlying rock increases with
depth at a rate of about 25 million pascals per giga G 109
thousand meters. mega M 106
A few units, not part of the ofcial SI system, kilo k 103
are in such common usage in the geological liter- deci d 101
ature that we refer to them throughout the text. centi c 102
The annum, a, is used for one year when measur- milli m 103
ing the age (time before present) of rocks and min-
micro 106
nano n 109
erals. The unit degree Celsius, C, is equivalent to
the unit kelvin, but the scales are offset such that
the number of degrees Celsius is less than the tables to facilitate unit conversion. These conver-
number of kelvin by the constant 273.15. Some sions take the form of the common examples
important derived quantities are made up of shown in Table 4.3.
ratios of fundamental quantities in which the
units cancel out. For example, one measure of 4.1.3 The material continuum
deformation called stretch, is dened as the nal Geometric and physical quantities used in struc-
length of a material line segment divided by its tural geology (e.g. strike and dip, mass density,
original length, so the stretch is devoid of units. stretch, displacement) usually are measured at
Because angles are dened as ratios of circular arc scattered locations or isolated exposures and the
lengths to radial lengths, they too are devoid of values so obtained commonly are used to charac-
units. However, it is customary to assign the unit terize a volume of rock that surrounds each
radian to angles. location. For example, in their monograph on
metamorphic tectonites Turner and Weiss (1963)
4.1.2 SI prefixes and conversion factors emphasize that one of the foundations of struc-
One of the beauties of the SI system is the ease tural analysis, as conceived by Bruno Sander in
with which quantities are manipulated in simple the second half of the twentieth century (Sander,
powers of ten by placing different prexes on the 1970), is the concept that a deformed rock mass is
units (Table 4.2). Other prexes exist that extend separable into volumes of statistically homoge-
the range of values both upward and downward, neous fabric that are investigated independently.
but these are less commonly called for in struc- One of the principal tools of such an investigation
tural geology. The prexes and symbols from is the stereonet on which orientation data are
Table 4.2 are attached to the front of the respec- plotted, devoid of any connection to location. This
tive unit or symbol. For example, using the viewpoint begs the question: how does the physi-
symbols k and M the rule of thumb stated in the cal quantity under investigation vary from one
previous section says that the vertical stress volume to an adjacent volume? The spatial varia-
increases with depth at a rate of 25 megapascals tion of physical quantities is unapproachable
per kilometer or 25 MPa km1. Because the ages using this method, in part because it avoids the
of rock formations typically fall in the range of use of calculus and the underlying principles of
millions of years the units are written mega- that mathematical discipline.
annum (Ma). The alternative, advocated here, is to embrace
Because there are many examples of archaic calculus and use it to investigate how physical
units in the literature of structural geology, one quantities such as the poles to planar elements,
needs to be procient converting to the SI system. temperature, velocity, and stress vary in space and
The CRC Handbook of Chemistry and Physics (Lide, time as structures evolve. In this context physical
2004) and The International System of Units (Mechtly, quantities are dened at a mathematical point by
1973) are useful references that contain extensive a limiting process in which an element of the
Table 4.3. Selected conversions from archaic to abandon this level of understanding until they had no
SI units. alternative. However, the development of physics in
the twentieth century has been a progressive move-
From To Multiply by ment away from visualizable models and toward
abstract mathematical models.
To convert length Physicists now talk of elds in a much more
inch meter 2.54102 abstract way than Faraday or Maxwell did. A eld is
foot meter 3.048 101 now thought of as a way of assigning numbers to a
mile meter 1.609 344103 region of space, much as a temperature map assigns a
temperature to every point on the earths surface.
To convert mass Although this description makes a eld seem very
kgf s2 m1 kilogram 9.806 65 abstract, it proves to be a very rich way of talking
pound mass kilogram 4.535 924101 about nature. In fact, physicists today talk about elds
(lbm) in exactly the same way as they talk about material
To convert time objects (Gregory, 1990).
hour second 3.60 103
annum second 3.153 6107 This viewpoint was of tremendous value to physi-
To convert temperature cists and engineers throughout the twentieth
Celsius kelvin T(K)T(C) century, but few structural geologists adopted
273.15 this perspective.
Fahrenheit kelvin T(K)(5/9)[T(F) Perhaps our most familiar experience with the
459.67] concept of a continuum comes with the realiza-
To convert force tion that given any two real numbers one can
kilogram newton 9.806 65 choose another that falls between the rst two.
force (kgf) Because of this property the set of real numbers is
pound newton 4.448 222 called continuous, or we would say it forms a con-
force (lbf) tinuum. This concept is applied every time we con-
dyne newton 1105 struct a graph of a continuous function and give a
scale to the ordinate and abscissa. We know that
To convert pressure, traction, or stress
atm pascal 1.01105 we can choose any scale for the axes and the func-
bar pascal 1.00 105 tion will plot without gaps. Because physicists
dyne/cm 2
pascal 1101 assert, based on intuition, that time and space can
2
lbf/in (psi) pascal 6.894 757103 be represented by real numbers, it is natural to
think of time and space as continuous. The fertile
To convert angle
imaginations of mathematicians have come up
degree radian 3.141 59/180
with functions that are discontinuous and some
of these have applications in structural geology.
For example, faults may be modeled as a surface
material is shrunk down about that point. We of discontinuity in a function for the displace-
understand that such a viewpoint cannot be taken ment eld. None-the-less, the displacement eld
literally if the element becomes too small (e.g. is adequately represented as continuous in the
smaller than a single pore in a sandstone), but the rock surrounding the fault.
denition provides the necessary mathematical Structural geologists seek to describe the
properties to interpret and explain geologic struc- motion of particles in a rock mass as it deforms
tures in which these point quantities vary contin- under the action of prescribed forces. It would be
uously in space and time. In other words these are useful to assign material properties or calculate
eld quantities dened in a material continuum. physical quantities at arbitrary points within
We feel we understand something when we can the rock mass. For this endeavor we construct a
picture how the wheels and levers must t together in material continuum for which the mass density,
order for it to work. Physicists were reluctant to momentum, and energy are well dened at every
the formal denition of the density at the point P

Mass density (kg m3)
5000
Rock mass in the material continuum is:
density
4000
3000
lim
n
Mn
Vn
(4.9)
2000 Continuum mass If we could carry out measurements on a rock

density
Lattice sample at smaller and smaller volumes we would
1000 defect Pore plot an irregular curve (Fig. 4.2, solid line). Below
0 a certain size particular mineral grains or pores
Molecular grain rock crustal
might alter the result; below that size, lattice
Length scale defects in a particular mineral could cause irreg-
ularities in the curve; and eventually individual
Fig 4.2 Mass density plotted versus length scale for rock
over the range from crustal to molecular. Major molecules, atoms, or sub-atomic particles would
discontinuities are caused by pore and lattice defects. become the important contributors to the mass
density. For these small volumes we have lost sight
of the density of the rock sample, yet our arbi-
point. This can, however, present conceptual dif- trarily chosen point P is, in principle, even
culties. For example, consider the mass density, , smaller.
of a sample of rock as determined by: The solution to this practical problem is to
accept the material continuum as a description of
M the rock sample while recognizing that there are
(4.8)
V constraints on the volume below which the
Here M is the mass and V is the volume of the denition (4.9) has no meaningful application.
sample. Mass is equivalent to weight divided by For a physicist, examining a crystal of quartz or a
the known acceleration of gravity and it is a drop of water, the continuum concept for mass
straightforward matter to weigh the sample. If density is reconciled with the concepts of particle
the sample is cylindrical the volume is readily cal- physics by insisting that the ratio, M/V, should
culated after measuring the height and diameter. only be calculated for length scales, L, much
With these numbers in hand, (4.8) is used to cal- greater than the intermolecular spacing in quartz
culate the mass density of the sample. Suppose and much greater than the mean free molecular
we decide to use this density as representative of path in the water. That is, the volume must be
the rock mass under consideration. Plotting much greater than a cube with sides of length
density as a function of volume for a homoge- about 1010 m. For structural geologists this
neous material continuum we have a continuous volume is a useful limit if we are considering
line of constant value (Fig. 4.2, dashed line). We deformation of individual mineral grains. How-
must admit that this density may not represent ever, the volume must include many of the differ-
the rock mass at scales of the lithosphere or at ent constituent grains to give a representative
scales of individual grains: there our sample size mass density for rock. Typically several cubic mil-
would have to be adjusted to determine a mean- limeters, perhaps up to several cubic centimeters,
ingful density. would be necessary to get meaningful densities
To characterize the mass density at a point P in for rock samples, depending on the grain size and
the continuum consider a sequence of volumes, heterogeneity of the sample. Thus, a few cubic
Vi, of mass, Mi, each containing the point and centimeters might represent the lower limit in
ordered from largest, i 1, to smallest i n, such volume for application of the continuum model
that the volume approaches zero as n approaches in terms of rock density.
innity. In this limit the largest dimension of the Given what we have learned in the twentieth
sample approaches zero, so the volume converges century from particle physicists about the basic
to the point, not to a surface or a curve containing building blocks of solids and uids we must
the point (Malvern, 1969). Using this procedure understand that the material continuum is an
4.2 PHYSICAL DIMENSIONS AND DIMENSIONAL ANALYSIS 127
abstraction or idealization of nature and we must of force times displacement. Thus, when analyz-
address its limitations. As John Wheeler reminds ing the relationships among various quantities it
us in his forward to The Continuum: A Critical is instructive to consider the dimensions of those
Examination of the Foundation of Analysis by quantities. In this section we introduce the
Hermann Weyl: dimensions commonly encountered in mechani-
cal processes.
For the advancing army of physics, battling for many a
Dimensional analysis is a useful tool for
decade with heat and sound, elds and particles, gravi-
working with and understanding theoretical con-
tation and spacetime geometry, the cavalry of mathe-
matics, galloping out ahead, provided what it thought structs in all of science and engineering. The
to be the rationale for the real number system. En- paper by M. K. Hubbert (1937) puts the use of
counter with the quantum has taught us, however, dimensional analysis in a geological context and
that we acquire our knowledge in bits; that the contin- relies on the methods put forward in the book by
uum is forever beyond our reach. Yet for daily work the P. W. Bridgman (1931). In this section we use
concept of the continuum has been and will continue dimensional analysis to understand whether a
to be as indispensable for physics as it is for mathemat- given equation, which reportedly describes some
ics. In either eld of endeavor, in any given enterprise, aspect of rock deformation, is consistent from a
we can adopt the continuum and give up absolute dimensional point of view. If not, the equation is
rigor, or adopt rigor and give up the continuum, but
invalid and should be discarded. Then we intro-
we cant pursue both approaches at the same time in
duce the technique for plotting dimensionless
the same application (Weyl, 1987).
graphs and illustrate why this is the preferred
We refer to mathematical points at which density method to present scientic results.
is dened in a continuum mechanical model of a
faulted rock mass. Those points and the 4.2.1 Dimensionally homogeneous
motions or material properties attributed to equations
them, must be thought of as representative of a The dimensions of the fundamental mechanical
nite piece of rock, perhaps several cubic cen- quantities (length, mass, time, and temperature)
timeters in volume. are given as: L, M, T, and respectively. The
dimensions of derived quantities are composed of
products and powers of these fundamental
4.2 Physical dimensions and dimensions. Reading {} has dimensions of we
have, for example:
dimensional analysis
area, A{}L2 (4.10)
One can express physical quantities in terms of volume, V{}L3 (4.11)
many different units of measure and a particular displacement, u{}L (4.12)
quantity can take on very different numerical
values under the different systems of units. For velocity, v{}L T 1 (4.13)
example, 1 m 2.85 103 printer points 3.94 acceleration, a{}L T 2 (4.14)
101 inches 3.28 100 feet 4.97 103 furlongs
mass density, {}M L 3 (4.15)
6.21 104 miles. However, the underlying
physics must be independent of the choice of force, F{}M L T 2 (4.16)
units: it cant depend on the length of the Kings stress, {}M L 1 T2 (4.17)
Foot! This leads us to the concept that there is
thermal expansion, {}1 (4.18)
something more fundamental than the units
attached to a physical quantity and this is the phys- stretch, S{}L L 1 L0 1 (4.19)
ical dimension of that quantity. Regardless of the
3.141 592 65, . . . , {}1 (4.20)
units chosen, for example, for mechanical work
(newton meter, pound force foot, or dyne cen- Both the stretch and the angle are dimensionless
timeter) this physical quantity has the dimensions physical quantities, but we use the symbol l
The subscripts, a and w, attached to density, , and

Atmosphere temperature, T, refer to that property of the aes-
thenosphere and of the ocean water, respectively. To
water: Tw , rw
Sea seawater evaluate (4.21) we assign the appropriate dimen-
sions to each quantity:
E
change in elevation, E{}L (4.22)
mass density, a and w {}M L 3 (4.23)
temperature, Ta and Tw {} (4.24)
Plate thermal expansion, {}1 (4.25)
motion Oceanic
lithosphere thermal diffusivity, {}L2 T 1 (4.26)
time, t{}T (4.27)
A basic principle of dimensional analysis is: an

equation is dimensionally homogeneous if every term
Magma flow
has the same dimensions. To be meaningful in a
physical context it is necessary for an equation to
be dimensionally homogeneous. This is not
Aesthenosphere: Ta, ra sufcient because one could construct a dimen-
sionally homogeneous equation that does not
obey the fundamental laws of physics. Further-
Fig 4.3 Schematic vertical cross section through oceanic
lithosphere and upper aesthenosphere at a mid-ocean ridge more, being dimensionally homogeneous does
and spreading center (Davis and Lister, 1974). not imply that the equation is the only, or even
the best, description of the event or process under
consideration. This test merely is a starting point
rather than 0, so quantities such as these can be in the evaluation of equations.
included in the algebraic manipulation of dimen- We apply the principle of dimensional homo-
sional equations (Obert and Duvall, 1967). geneity to (4.21) by noting that the left-hand side
Similarly, the symbol 1 is used for dimensionless is the elevation change which has dimensions of
constants such as . length, E{}L. The right-hand side is analyzed
Dimensional analysis provides a procedure to by substituting the dimensional symbols and can-
evaluate equations involving physical quantities: celing exponents where appropriate:
those purporting to describe some physical object,
{}
event, or process. This procedure enables one to a
t
2 (Tw Ta)
check whether an equation might be in error by a w
being dimensionally inconsistent. This is probably

M L3 L2T1T
the rst thing you should do when confronted (1)(1)( )
M L3 M L3 1
with an unfamiliar equation, especially a very
complicated one. For example, suppose that you (M L3)(M1 L3)(1)() L2 T1 T
are reading the geological literature and you come
M0 L0 0 L2 T 0 L2 L (4.28)
across the following equation for the change in
elevation, E, of the seaoor due to cooling and Note that the sum or difference of two terms with
thermal contraction of oceanic crust (Fig. 4.3) as it the same dimensions can be shortened to a single
is transported away from a mid-ocean ridge by term with those same dimensions, so ( ) is
plate motion (Davis and Lister, 1974): written as (). Also, exponents are added for terms
that are multiplied, and these may cancel to

t
E a
2 (Tw Ta) (4.21) produce a dimensionless term, as in 1 0 1.
a w After simplifying the right-hand side we nd
dimensions of length: this equation is dimension- (a)

ally homogeneous.
4.2.2 Dimensionless equations and graphs

Having corrected dimensional inconsistencies,
the next step in the analysis of an unfamiliar
Aplite dike
equation is to plot it in graphical form so you can
visualize the relationships between the different
variables. For other than the simplest equations Offset
with one dependent and one independent vari-
able, there are several possibilities for plotting
graphs. Here we introduce the concept of dimen-
sionless equations and graphs because they turn
out to be a particularly instructive form for clari- Fault
fying the relationships among the different vari-
ables. Fig. 4.4a shows an exposure of a fault in
granitic rock with offset aplite dikes indicating ~50 cm
apparent left-lateral separations of several tens of
centimeters. By walking along this fault one could
gather data on offset markers. A schematic map (b)
X
(Fig. 4.4b) illustrates such a fault trace with a few
offset markers at different positions along the
fault. The trace length of the fault is W 384 m.
The positions of each marker are determined by 100 m
the eld coordinate axis, X, oriented along the
trace of the fault. A plot of the measured offset, O,
versus position, X, is shown in Fig. 4.5a. Note that
the offset is zero at either end of the fault by
denition. These points may or may not be
exposed, so their locations are not necessarily O
well dened. W
The following equation, which we will learn
more about later in the book, describes the rela-
tive displacement, u, or slip along a two-dimen-
sional model fault in an elastic material (Pollard
and Segall, 1987): Y
Offset, O, is
exaggerated
u 2
1
G
a2 x2 (4.29)
Fig 4.4 (a) Photograph of fault in granitic rock exposure
offsetting aplite dikes from the Sierra Nevada, CA (Segall and
The variables and their dimensions are given by:
Pollard, 1983b). (b) Schematic map of fault with offset dikes
relative displacement, u{}L (4.30) and zenoliths. Photograph by D. D. Pollard.
shear stress drop, {}M L 1 T 2 (4.31)

Poissons ratio, {}1 (4.32) The relative displacement, u, is analogous to the
shear modulus, G{}M L 1 T 2 (4.33) offset, O, measured across the fault and the model
fault length, 2a, is analogous to the outcrop trace
half-length, a{}L (4.34) length of the fault, W. The shear stress drop, ,
spatial coordinate, x{}L (4.35) is the change in shear stress acting on the model
(a) O (m) Because the origin of the coordinate system for

0.4 (4.29) is at the model fault center, the eld coor-
dinate, X, must be transformed to have an origin
0.2 at the center of the natural fault (Fig. 4.5b). This is
X (m) done using x X (W/2). The solid curve on this
0.0 graph represents the relative displacements for
0 100 200 300 400 the model. This curve was calculated using a
192 m and noting that the offset at x 0 is about
(b) u (m) 0.4 m. Substituting these values, we nd:
0.4 m
0.2 m
0.4 m 2
1
G
(192 m),

x (m) 1

so 2 0.002 (4.36)
192 m 0.0 +192 m G
This value for the leading terms on the right-hand

(c) u/a
Fault #3 side of (4.29) was used to plot the solid curve on
C3
Fig. 4.5b. Because we have adjusted the numerical
Fault #2 value of this leading term to t the data at x 0,
C2
Fault #1
the solid curve goes exactly through the datum
C1 point there. The curve goes through the data at
x/a the ends of the fault, x 192 m, where the offset
1.0 +1.0 is zero by denition.
0.0
If we wanted to describe several different
(d) uG/ 2a(1 n) faults, we could measure offset markers for each
and plot the offset as a function of distance along
+1.0 each fault. It would be difcult to compare the dif-
ferent faults, because each would be on a different
x/a graph. However, we can generalize the eld data
1.0 +1.0 by dividing measurements of offset by the half-
0.0
length of the fault. Carrying out the analogous
Fig 4.5 Graphs of offset or displacement discontinuity operation for each side of the model equation we
versus position along the trace of a fault. (a) Offset versus nd the following dimensionless equation for the
position using field coordinate system. (b) Displacement relative displacements:
discontinuity versus position with origin at fault middle.
(c) Normalized displacement discontinuity versus normalized
position. (d) Generic plot of dimensionless displacement
u
a
2(1 )

G
1
x2
a2
(4.37)
discontinuity versus dimensionless distance.
The terms in this equation are numbers, dimen-
sionless quantities, and dimensionless ratios.
fault during slip. Poissons ratio, , and the shear Using this dimensionless form we can plot eld
modulus, G, are two properties of the elastic ma- data from different faults on the same graph (Fig.
terial surrounding the model fault. The combina- 4.5c). When normalized in this way the abscissa
tion G/(1 ) can be thought of as the stiffness of values of all such data sets range from x/a 1.0
this material. Inspection of (4.29) shows that the to x/a 1.0, but the ordinate values for a partic-
equation is dimensionally homogeneous. The ular x/a may be quite different. The values at the
actual distribution of slip on faults and models to center of the fault traces (x 0) dene a set of con-
explain these are part of a growing literature stants (C1, C2, C3, etc.) equal to the quantity
(Cowie and Scholz, 1992; Dawers et al., 1993; 2(1 )/G for each fault.
Brgmann et al., 1994; Cowie and Shipton, 1998). Although the dimensionless equation we have
just derived is useful, it is possible to represent the

(a)
generic model fault on a single curve by moving
the term 2(1 )/G to the left-hand side of (4.37):

uG x2
1 2 (4.38)
2 a(1 ) a
Plotting the left-hand side as a function of the

right we generate the generic curve for dimen-
sionless relative displacement versus dimension-
less position (Fig. 4.5d). In this form one would
refer to the relative displacement as having been
normalized by the maximum relative displace-
ment, 2(1 )/G, the value of the displacement
at the middle of the fault trace. The ends of the
fault are given by x/a 1 and x/a 1.
In Figure 4.5d the curve for relative displace-
ment is symmetric and goes to zero at the ends of Dike
the fault. At the middle of the fault, the offset of
geologic markers would be greatest and, on this
dimensionless graph, would have a magnitude of
(b) Model Host
1. All possible distributions of relative displace-
dike rock
ment for faults that approximate the behavior of
this theoretical model would scatter about a x
single curve on this plot. The scatter would reect
At t =0 T = Tm T =0
errors in measurement and mechanical differ-
ences between the model and natural fault. All of 2a
the data sets for a set of faults could be plotted on
this graph and thereby could be compared to each (c) 1.0 0
0.1 .05
0.02
0.
0
other and to the model.

1
0.2
0.8 kt/a2 = 0.00
Next we consider a model for heat conduction
Temperature, T/Tm
near an intrusion of magma to illustrate the 0.5

dimensionless equation and graph for the contin- 0.6
uous temperature variation in space and time 1
(Carslaw and Jaeger, 1959; Cathles, 1977). In Fig. 0.4 2
4.6a the eroded remnants of an igneous dike are
5
pictured and a glance at this photograph suggests 0.2
that the shape of the dike is roughly tabular. The
length along the outcrop and the height along the 0.0
canyon face are much greater than the dike thick- 0.0 1.0 2.0 3.0
ness. From observations on active volcanoes we Distance, x/a
know that the time scale for emplacement of Fig 4.6 (a) Photograph of basaltic dike exposure from the
some basaltic dikes can be small relative to the San Rafael Swell, UT (Delaney et al., 1986). (b) Thermal
time scale for signicant heat loss into the sur- conduction model with initial conditions of elevated
rounding host rock (Delaney and Pollard, 1982). temperature in the dike and zero in the surroundings.
Although both the tabular shape and the relative (c) Graph of normalized temperature versus distance with
time scales just mentioned must be reconsidered curves representing successive normalized times (Carslaw
to understand the details of dike emplacement, and Jaeger, 1959). Photograph by D. D. Pollard. Graph
reprinted from Carslaw and Jaeger (1959) by permission of
these postulates serve to constrain an instructive
Oxford University Press.
model for the temperature eld.
The model (Fig. 4.6b) is based on the solution contact, x/a 1, changes to Tm/2. As time increases,
for conduction of heat in one dimension, x, away temperatures in the model dike decrease and
from the tabular region a x a in an innite those in the immediate surroundings increase
body with homogeneous thermal properties and then decrease. For t/a2 5, corresponding to
(Lovering, 1935, 1936; Jaeger, 1957). The tabular a time of about 58 days for a dike 1 m thick, the
region is taken as a model for a dike of thickness temperature has dropped to about 20% of its
2a. The only thermal property is the diffusivity, , initial value and risen to a comparable tempera-
which typically has values near 1 106 m2 s1 for ture in the immediate surroundings.
rocks (Lee and Delaney, 1987). The initial condi-
tions (IC) on the temperature, T, are dened at the
arbitrary time, t 0, and the boundary condition 4.3 Dimensionless groups and the
(BC) on the temperature is dened at an innite
distance from the model dike:
scaling of structural processes
IC: for t 0,T Tm for a x a
Models of geologic structures provide insights
IC: for t 0, T 0 for a x a (4.39) about deformation in Earths crust, some of
which come from studying dimensionless groups
BC: at x , T 0 for all t
of variables. In this section we explore examples of
Here Tm is the initial temperature throughout the these dimensionless groups and show how they
model dike, the initial temperature is zero every- are used to understand the scaling of structural
where else, and very far from the model dike the processes. The direct method to identify dimension-
temperature remains zero for all times. A con- less groups considers the governing differential
stant ambient temperature, Ta, may be added to equations for the process and manipulates these
the solution for all positions and times. to isolate the dimensionless groups (Bird et al.,
The distribution of temperature in space, x, 1960, pp. 107, 185, 338). While the direct method
and time, t, normalized by the initial tempera- is preferred, it is not applicable if the governing
ture, Tm, is (Carslaw and Jaeger, 1959): equations are unknown or in doubt. In Chapter 12
we discuss the procedure for selecting the gov-
T(x, t) 1
Tm 2
erf
ax
2 t
erf
ax

2 t
,
erning equations and general boundary condi-
tions of a problem. We begin our discussion of the
x , t 0 (4.40) direct method with the bending of sedimentary
layers over a laccolith and then consider the ow
The function erf() is the error function whose of magma through a sill.
values are tabulated in reference books (Carslaw Next we introduce the Rayleigh method of
and Jaeger, 1959). Recalling from (4.26) that the dimensional analysis (Brodkey and Hershey, 1988)
dimensions of thermal diffusivity are L2 T1, the with an example that addresses the buoyant rise
terms in parentheses in (4.40) are dimensionless of salt through a sedimentary basin in an intru-
so the equation is dimensionally homogeneous. sive form called a diapir. This method does not
In Figure 4.6c the normalized temperature dis- rely on knowledge of the governing equations for
tribution is plotted as a function of distance from viscous ow, but does require a complete knowl-
the centerline of the dike using values of t/a2 as edge of all the variables relevant to the process.
a parameter that is a proxy for time. In this way, Underlying this method is a theorem introduced
for a given diffusivity and dike thickness, each by Buckingham (1915) and based on the necessity
curve represents the distribution of temperature for equations that describe physical processes to
for a particular snapshot in time. Note that the be dimensionally homogeneous. The Rayleigh
initial temperature eld is portrayed by the line method itself cannot assure one that the dimen-
labeled t/a2 0, and the eld for subsequent sionless groups so determined are correct, and an
times has successively greater values of this par- erroneous result will be found (with no warning)
ameter. In the rst instant the temperature at the if the number of variables is too few or too many.
4.3 DIMENSIONLESS GROUPS AND SCALING 133
The choice of a method depends upon ones (a)

condence in selecting the governing equations
versus ones condence in selecting all the vari-
ables (but no more) and properly grouping them.
4.3.1 Bending over a laccolith:

the direct method
The rst example is taken from the theory of
bending of thin elastic plates under lateral loads
(Timoshenko and Woinowsky-Krieger, 1959). This
theory has been applied to the study of laccol-
ithic intrusions (Johnson, 1970; Pollard and
z
Johnson, 1973; Jackson and Pollard, 1990) in the (b)
Henry Mountains of southern Utah. Fig. 4.7a is a
photograph of an outcrop on the ank of
Trachyte Mesa, where the edge of a small laccol- D B
ith, composed of diorite porphyry, is exposed uo
p x
and a few beds of Entrada Sandstone are bent H Magma
over the laccolith. Only half of the laccolith
model is illustrated here (Fig. 4.7b) because we
postulate that it is symmetric about its center.
L/2
Most of the overburden has been eroded from
this site, but stratigraphic studies suggest that
(c)
the depth, D, was a few kilometers at the time of
magma intrusion. Note how the prominent
Entrada sandstone layer is horizontal at the left
side of the photograph, then bends concave
upward against the diorite porphyry, and then
bends concave downward and attens out over
the top of the laccolith.
The plate theory model represents a layer of
sedimentary rock with height, H, and length, L,
which overlies the laccolith (Fig. 4.7b). The strata
are continuous beyond the periphery of the
laccolith so L refers to that portion of a layer
immediately above the intrusion. The x-axis is
directed along the middle surface of the layer.
Fig 4.7 (a) Photograph of distal edge of a laccolith
Conceptually, the model supposes that magma is
exposure from the Henry Mountains, UT (Pollard and
intruded upward through some unspecied con- Johnson, 1973). (b) Elastic plate model of bending strata over
duit and then spreads laterally under the layer the laccolith. (c) Laboratory model with viscous fluid injected
in question. The driving force for bending the under elastic layer to simulate laccolith formation.
layer is provided by the net upward pressure, p, Photograph by D. D. Pollard.
which is taken to be constant. This pressure is
given by the difference between the magma pres-
sure and the pressure due to the weight of the dimensions as pressure, that is B {} M L1 T2. It
overburden, gD, where is the average density is possible to consider a variable pressure due to
and g is the acceleration of gravity. Bending is magma ow, a variable resistance to bending, and
resisted by the elastic stiffness, B, a constant resistance to bending provided by shear stress
material property of the layer with the same transmitted between adjacent layers (Pollard and
Johnson, 1973; Koch et al., 1981), but these effects Next one substitutes the normalized variables
are ignored in order to introduce dimensional and differential operator into the differential
analysis using a simple model. These effects, in equation:
part, may account for the atter top of the
1 d4 u d4u*z 12p
Trachyte Mesa laccolith (Fig. 4.7a) as compared to (uou*z) 4o (4.44)
L d(x*)
4 4 L d(x*)4 BH3
the laboratory model laccolith with a single layer
(Fig. 4.7c). The nal step is to rearrange the equation to
The governing differential equation for the group the constants into a single dimensionless
bending plate model is presented without deriva- group:
tion (Johnson, 1970), because our purpose is to
demonstrate the direct method of dimensional
analysis. This equation is based upon simplifying
d4u*z
d(x*) 4
12
p L4
B uoH3 (4.45)
postulates about the kinematics of bending that The left-hand side of (4.45) is the dimensionless
are valid if the layer is thin compared to its length differential operator acting on the dimensionless
(Timoshenko and Woinowsky-Krieger, 1959). The dependent variable. The term in square brackets
differential equation for the vertical deection, on the right-hand side is the dimensionless group
uz, of the middle surface of the layer is: we have identied for this differential equation.
In some contexts dimensionless groups are
d4uz 12p
(4.41) referred to as the scale factors. Note that the
dx4 BH3
dimensionless group identied in (4.45) contains
The rst step in the analysis is to identify the vari- the elastic stiffness of the bent layer, B, the
ables, and to understand their roles in the physi- length, L, and height, H, of the layer, and the net
cal process. Here, the spatial coordinate, x, is the upward pressure, p, acting on the layer. The
only independent variable, and the vertical powers to which these quantities are raised in the
deection, uz, is the only dependent variable. A dimensionless group inform us about the relative
solution to (4.41) is uz f (x), a function that sensitivity of the deection, uz u*u z o
, to varia-
describes the distribution of deection with positions in these physical quantities. For example,
tion along the layer. the deection scales directly with the fourth
The next step is to make the variables dimen- power of the length, L. Thus, all else being equal,
sionless (to normalize them) by dividing each by a two layers that differ in length by a factor of two
characteristic value of a quantity with the same would differ in deection by a factor of sixteen.
dimensions. The natural choice for normalizing x Changing the height also has a dramatic effect on
is L, the length of the layer in the x-direction. For the bending whereas changing the rock stiffness
a characteristic vertical deection, we choose the or the net upward pressure by a comparable
value at the center of the plate, uo uz (x 0). These factor has relatively little effect because the stiff-
choices are arbitrary, but are motivated by the ness, B, and the net pressure, p, enter the dimen-
geometry and the symmetry of the problem. The sionless group to the rst power. If the height is
normalized variables are written with a super- doubled, the deection decreases by a factor of
script *: eight, but if the stiffness is doubled the deection
is decreased by a factor of two. Similarly, doubling
x u the net pressure increases the deection by a
x* , u*z z (4.42)
L uo factor of two.
The differential operator in (4.41) also must be In this manner one can assess the importance
normalized. In this case d4/dx4 {} L4, and this of different physical quantities for the outcome of
operator is normalized using the length of the a tectonic process. Interestingly, this assessment
layer: does not require one to solve the differential equa-
tion (4.41). By determining the sensitivity of the
d4 d4 dependent variable to the various parameters that
L4 4 (4.43)
d(x*)4 dx affect that variable, one can design a strategy for
eld and laboratory investigations that focuses on (a)

measurement of the most sensitive parameters.
For example, errors in the eld measurement of
length or height of the layer over the laccolith are
of greater consequence than comparable errors in
the laboratory measurement of rock stiffness. Of
course, such assessments depend upon the
correct selection of the governing equation.
Recent analyses based on somewhat different gov-
erning equations, geometry, and boundary condi-
tions for the deformation of strata over laccoliths
have provided additional insights (Kerr and
Pollard, 1998; Zenzri and Keer, 2001).
(b) z
4.3.2 Magma flow in a conduit: the direct
method
The direct method of dimensional analysis is illus- r, h
trated using a problem from the theory of uid x
dynamics for an isothermal viscous uid (Bird vo W
p1 p2
et al., 1960, p. 71). Viscous ow theory has been vx
applied to a myriad of problems in structural
geology including the folding of ductile strata,
the development of salt domes, and rebound of
the Earths crust after glacial unloading (Johnson
L
and Fletcher, 1994). The geological examples we
refer to here are the sills of Shonkin Sag, Montana
(Fig. 4.8a), thin horizontal conduits through (c)
which a viscous magma owed (Hurlbut and
Griggs, 1939; Pollard et al., 1975). The governing
equation for this ow is presented without
detailed derivation, because our purpose, again, is
to demonstrate the direct method of dimensional
analysis.
The model is a parallel-sided conduit of width,
W, lled with a viscous uid (Fig. 4.8b). The
Newtonian viscosity, , measures the resistance to
ow and is postulated to be constant in space and
time. The viscosity has the same dimensions as
pressure multiplied by time, that is {}
M L1 T1. The mass density of the magma, , also
is taken as a constant. The length of the conduit,
L, is very great compared to the width, as is the Fig 4.8 (a) Photograph of laccolith exposure (left) and a
dimension out of the (x, z)-plane of view. Here set of sills (right) from the Shonkin Sag, MT (Pollard et al.,
the origin of coordinates is at the center of the 1975). (b) Viscous flow model between parallel plates.
conduit and the z-axis is parallel to the width of (c) Reynolds experiments of dye injected into viscous fluid
the conduit. We postulate that the only non-zero flowing in a tube (Van Dyke, 1982): upper sketch shows
component of velocity, vx, is directed along the laminar flow regime and lower sketches show turbulent flow
regimes. Photograph of exposure by D. D. Pollard.
length of the conduit, and the pressure decrease,
Laboratory photographs by N. H. Johannesen and C. Lowe.
p1p2, drives this ow.
The governing equation for the ow of viscous malized by a physical quantity that shares the
magma in a sill is derived from two general prin- same dimensions. It is customary in uid dynam-
ciples, conservation of mass and conservation of ics to select a characteristic length and a charac-
momentum. The rst of these dictates that vx teristic velocity for this purpose. Here the only
cannot vary in the direction of ow, but it can characteristic length is the width of the conduit,
vary across the conduit in the z-direction, and it W. We select the velocity, vo, at the center of the
can vary in time. Thus, conditions on the velocity conduit to be characteristic. This is the maximum
components are: velocity, but the selection is arbitrary so we could
have selected the average velocity. The normalized
vx f (z, t) only,vy 0 vz (4.46)
variables are dened as:
The velocity component, vx, is one of the depen- x z v
x* , z* , t* o t ,
dent variables of this problem. W W W
The second principle, conservation of momen- v p po
v*x x , p* (4.48)
tum, introduces the forces acting on volume ele- vo v o2
ments of the magma. The decrease in pressure The characteristic velocity and distance are used
from one side of the element to the other intro- in the ratio vo/W to dene a dimensionless time.
duces a net force in the direction of this pressure Also, a reference pressure, po, is subtracted from
decrease. The pressure, p, is the second dependent the pressure and then the combination v 2o is used
variable in this problem and it can vary with posi- to normalize this reduced pressure. The reference
tion, x, along the direction of ow, and with time. pressure could be that at the entrance to the sill.
Viscous drag introduces another force on the The differential operators are normalized as
volume element and this is proportional to the follows:
viscosity, . The gravitational force acts in the ver-
tical direction, and therefore does not contribute W
, W ,
to ow in the horizontal conduit, and is ignored. t* vo t x* x
Under the restriction of constant density and vis- 2 2
W2 2 (4.49)
cosity, the pressure forces and viscous forces are (z*)2 z
capable of producing accelerations in the magma
The normalized variables (4.48) and differential
described by the equation:
operators (4.49) are substituted into the governing
vx p 2v equation (4.47) to nd:
2x (4.47)
t x z
vo 1
(v v* ) ( v 2o p* po)
This is a special case of the more general, W t* o x W x*
three-dimensional equations of motion called the 1 2
2 (v v* ) (4.50)
NavierStokes equations, developed by Navier in W (z*)2 o x
1822 (Bird et al., 1960, p. 81). The left-hand side of
Bringing the constants outside the derivatives
this equation is the mass per unit volume times
and eliminating the derivative of the constant ref-
the acceleration (time derivative of the velocity).
erence pressure this equation becomes:
The right-hand side is the sum of the pressure and
viscous forces per unit volume. In essence this vo2 v*x v 2 p* vo 2v*x
o (4.51)
equation is a specialized expression of Newtons W t* W x* W 2 (z*)2
Second Law of Motion written in the order ma F,
Note that the two combinations of physical con-
where m is the mass, a is the acceleration, and F is
stants in this equation are dimensional; they both
the net force acting on a uid element.
have dimensions of force per unit volume; and
There are two dependent variables, velocity
each is associated with the magnitude of a differ-
and pressure, and three independent variables,
ent force acting in the ow system:
the x- and z-coordinates and time in (4.47). In addi-
tion, there are two uid constants, mass density vo2
inertial force per unit volume (4.52)
and viscosity. Each of the variables must be nor- W
vo referred to as laminar ow. At a greater velocity

viscous force per unit volume (4.53)
W2 (middle illustration) uid particles follow cir-
Viscous forces are related to the product of the vis- cuitous paths, both along and across the axis of
cosity and velocity divided by the square of the the pipe. Thus, the stream of dye mixes with the
characteristic length. If the viscosity is doubled, adjacent uid and thereby spreads across the
but the width and velocity remained unchanged, entire pipe with distance from the point of injec-
we would expect the viscous forces to double. tion. Using special visualization techniques (lower
Similarly, if the width of the conduit is cut in illustration) this new ow regime can be seen as a
half, but the viscosity and velocity remained un- complex set of eddies that vary rapidly with time.
changed, we would expect the viscous forces to This is referred to as turbulent ow.
increase by a factor of four. Similarly, if the velocity The transition from laminar to turbulent ow
were doubled, the inertial forces would increase by in a pipe occurs at Reynolds Numbers ranging
a factor of four, but the viscous forces would only from 2000 to 13 000, depending upon the rough-
double. In this way we understand that these com- ness of the pipe and the geometry of the entrance.
binations of quantities are the scale factors for the When the product of the diameter, velocity, and
forces acting on the uid. density, divided by the viscosity is less than 2000,
Dividing all three terms of (4.51) by v 2o /W, a viscous forces dominate over inertial forces and
single dimensionless group is identied: the ow is laminar. In the literature of uid
mechanics this is referred to as low Reynolds
v*x
t*
p*
o2
x*
v W 2 2v*x
vo W (z*)2
(4.54) Number flow. The great viscosity of magma relative
to typical products of conduit width, velocity, and
This group is a ratio of the viscous force (4.53) to density, usually places them in the laminar ow
the inertial force (4.52). This dimensionless group regime. There may be little direct evidence for
usually is written as the reciprocal of the form ow laminae in igneous rock, although regular
given here and called the Reynolds Number. It is patterns of zenoliths or crystals may be sugges-
associated with the name of a famous uid tive. Observations of modern eruptions suggest
mechanician, Osborne Reynolds, who studied the that streams of lava approximate laminar ow.
transition from laminar to turbulent ow in con- None-the-less, the inference that magma ow is in
duits (Reynolds, 1883): the laminar regime usually is based on estimates
of Reynolds Number and analogies to laboratory
Wvo experiments using other liquids.
Reynolds Number Re ,

inertial force
(4.55) 4.3.3 Rise of a salt diapir: the Rayleigh
viscous force method
The magnitude of Reynolds Number can be One of the more interesting and challenging prob-
used to characterize the transition in style of ow lems in structural geology is the rise of salt (or
between two dramatically different ow regimes. magma) from depth toward the surface (Fig. 4.9a),
This was demonstrated in the classic experiments and the associated deformation of the host rock
by Reynolds in which he injected dye into the uid (Trusheim, 1960; Braunstein and OBrien, 1968).
owing through a pipe (Fig. 4.8c) and observed The less dense salt responds to the forces of buoy-
how the ow changed as a function of the veloc- ancy and ows upward while the surrounding
ity, while the pipe diameter and the uid density rock mass deforms in a uid or ductile manner
and viscosity remained constant (White, 1974). At and ows out of the way. This conceptual model of
relatively low velocity (upper illustration) the ow salt intrusion is quite different from that pro-
eld is very regular, so the stream of dye is per- posed by G. K. Gilbert (1877) for the emplacement
fectly straight, regardless of the position of injec- of magma in the Henry Mountains laccoliths (Fig.
tion from near the wall to the center of the pipe. 4.7a). The laccoliths apparently formed at a rela-
Clearly the path of any particle of uid is straight tively shallow level in the crust where the sur-
and parallel to the walls of the pipe. This is rounding rock mass deformed largely as an elastic
(a) are the density of the sphere, s, and the density
of the host uid, f . Since the sphere is rising
because of buoyancy, another parameter of this
problem must be the gravitational acceleration, g.
The velocity elds inside and outside the sphere
are complex, but here we focus only on the veloc-
ity of the sphere, v, relative to the static host at a
great distance from the sphere, and consider that
to be the dependent variable. For this conceptual
model we postulate that the ow is steady, so the
(b) velocity is constant and time does not enter the
rf , hf problem. Also we postulate that the ow is
isothermal, so heat transfer from the body to the
surroundings is ignored. We do not specify any
v distance scale that would place boundaries on the
rs , hs size of the surrounding uid mass. Conceptually,
the body rises forever in a host uid of innite
g
extent. Finally, the direction of rise is tacitly
assumed to be in the opposite direction of the
gravitational acceleration, so no coordinate axes
2R are explicitly required to dene this problem.
Despite all of the simplifying postulates made
in the previous paragraph, we have identied six
Fig 4.9 (a) Schematic diagram of salt diapers. (b) Stokes quantities that apparently affect the velocity of
model for a viscous sphere rising in a viscous fluid. Schematic
the sphere. In an experimental approach to this
diagram reprinted from Trusheim (1960) by permission of
problem, each quantity would be systematically
the AAPG whose permission is required for further use.
varied, as all others are held constant, in order to
discover their relationships. The number of exper-
and brittle solid that bent and fractured in iments would appear to be daunting; however,
response to the advancing magma. Masses of salt dimensional analysis helps to reduce the number
(or magma) whose rise is largely accommodated of variables for experimentation. The rst step is
by ductile ow of the surrounding rock are called to list all the physical quantities and identify their
diapirs. We illustrate the Rayleigh method of dimensions:
dimensional analysis in the context of diapirs by
radius of sphere, R{}L (4.56)
considering the slow rise of a buoyant viscous
sphere in another viscous uid of greater density relative velocity of sphere, v{}L T 1 (4.57)
(Fig. 4.9b). The solution by C. G. Stokes dates to the density of host fluid, f {}M L 3 (4.58)
middle of the nineteenth century and has found
density of sphere, s {}M L 3 (4.59)
innumerable applications in engineering and
science (White, 1974, p. 211). However, we analyze viscosity of host uid,f {}M L 1 T 1 (4.60)
this problem without the benet of the governing viscosity of sphere, s {}M L 1 T 1 (4.61)
equations or their solution by employing dimen-
acceleration of gravity, g{}L T 2 (4.62)
sional analysis.
A postulate, born out by the Stokes solution, is There are seven quantities in the three dimen-
that the rising body of viscous uid maintains a sions: length (L), mass (M), and time (T). We may
spherical form. Therefore the size and shape of reduce this number by making the additional
this body is completely specied by its radius, R. assumption that the densities and the accelera-
The viscosity of the sphere, s, and the viscosity of tion of gravity enter only through the difference
the host uid, f, are both considered constant, as in specic weights of the two uids:
difference in specic weights, groups and the exponents for these can be used to
determine the other three exponents.
g ( f s )g{}M L 2 T2 (4.63)
We choose the exponents c and e, and solve for
Thus, the revised list includes only ve indepen- the other exponents in terms of these:
dent quantities. Furthermore, the quantity mass
d c e
density times gravitational acceleration is a
measure of one of the forces acting in ow b 2c d e
regimes such as the rising diapir: 2c (c e) e c (4.69)
a b 2c d e
g gravitational force per unit volume (4.64)
( c) 2c (c e) e 2c
Recall that two other forces, inertial and viscous,
The exponents a, b, and d are removed from (4.65)
were dened in (4.52) and (4.53).
by substitution:
The Rayleigh method takes each of the ve

R2(g) c s e
independent quantities and raises it to an R2cvc ( g)c (ce) es
unknown integral or fractional exponent, here f v f f
given by the symbols a through e. An objective of constant (4.70)
the analysis is to determine these exponents and
The terms in square brackets are the two dimen-
use them to identify the dimensionless groups.
sionless groups for this process. The dimensional
The quantities, raised to these unknown powers,
analysis provides no additional information
are multiplied together and it is asserted that
about the values of the exponents. That informa-
their product is equal to a constant:
tion is discovered through laboratory experimen-
Ra v b ( g) c fd se constant (4.65) tation. However, we now have only two quantities
to work with instead of the original ve, so the
The appropriate dimensional expressions are sub-
design of the necessary experiments is greatly
stituted for the physical quantities in (4.65):
simplied.
La (L T1)b (M L2 T2)c (M L1 T1)d (M L1 T1)e The rst term in square brackets in (4.70) con-
La(Lb Tb )(Mc L2c T2c )(Md Ld Td )(Me Le Te) tains two measures of force per unit volume of
1 (4.66) sphere. The gravitational force per unit volume
(4.64) is proportional to the density difference
Because the product of the quantities raised to the between the host uid and the sphere, and to the
unknown powers is a constant, the product of the acceleration of gravity. The viscous force per unit
dimensional terms raised to these powers must be volume (4.53) is proportional to the viscosity of
dimensionless: the product must be equal to one. the host uid and the relative velocity, and
This implies that the product of each dimensional inversely proportional to the square of the sphere
term raised to the given powers is equal to one: radius. The rise of the sphere can thus be seen as
LaLbL2cLdLe Lab2cde 1 dependent upon a competition between the grav-
itational and viscous forces. Considering the
Mc Md Me Mcde 1 (4.67)
powers to which the variables are raised, we note
Tb T2c T d Te Tb2cde 1 that the relative velocity is most sensitive to
From (4.67) we conclude that the sum of the expo- changes in the radius of the sphere.
nents for each dimensional term is zero: The second dimensionless group identied in
(4.70) is the ratio of viscosities for the sphere and
a b 2c d e 0 host. Because the viscosity of the sphere might be
cde0 (4.68) either zero (an open hole) or innite (a rigid body),
b 2c d e 0 it is advisable to use the following dimensionless
group:
By this procedure we have reformulated the ve

e
unknown exponents into three equations. This f s
(4.71)
suggests that there are only two dimensionless f s
The quantity in square brackets ranges from 1 (a)

to 1 as the sphere changes from rigid (s )
to an open hole (s 0). This dimensionless
group is eliminated from consideration if the vis-
cosity of the sphere and the host uid are the
same.
The so-called Buckingham Theorem was pub-
lished by E. Buckingham in 1915, and often is cited
in discussions of dimensional analysis (Bucking-
ham, 1915; Bird et al., 1960; Brodkey and Hershey,
1988). This theorem is based upon the necessity
for dimensional homogeneity of equations that
describe physical phenomena. That is, every term
in such an equation, when written in terms of the (b)
dimensions of the four fundamental quantities,
must be made up of the same powers of each Hinge
Hi
b
quantity. The theorem states: ng
Lim
e
Lim
The number of dimensionless groups for a particular Th
ick
b
physical process is equal to the number of variables nes
,H
less the number of dimensions represented in those s, H
ss
ne
variables.
ick
Th
Since there are ve physical quantities in the

A rc
rising sphere problem as conceptualized here, and le n g
th , L
a
there are three dimensions, there are only two
independent dimensionless groups according to
the theorem. The Rayleigh method determines (c) Viscosity, h1
two groups and that is consistent with the
theorem.
Viscosity, h Thickness, H
4.3.4 Dimension analysis applied to the Dxx
folding process Amplitude, A
A fundamental question in structural geology
concerns the length scale of structures compris-
ing an array. A notable example is that of an array Wavelength, L
of folds (Fig. 4.10a), but the question pertains to
many other structures. Convenient measures of
length scales for folds include the distance along Fig 4.10 Fold styles, terminology, and modeling.
a particular surface from hinge to hinge and the (a) Photograph of folds in metamorphic rocks. (b) Sketch of a
thickness between adjacent surfaces (Fig. 4.10b). portion of a fold illustrating the hinges and limbs, and
In a deformed terrain, folds will generally occur at suggesting field measurements of arc length and thickness.
many scales; from single layers a few millimeters (c) Model geometry and parameters described in the text.
Photograph by R. C. Fletcher.
in thickness, H, with arc lengths, La, of a few cen-
timeters to composite rock layers several kilo-
meters in thickness having arc lengths of ten or
more kilometers. adjacent surface traces bounding a fold are likely
The simpler example we address here is a to vary, as are thicknesses at different positions
folded layer embedded in deformed metamorphic along the fold limbs. However, average values
rock (Fig. 4.10a). Hinge-to-hinge arc lengths for two from multiple measurements along a train of
folds provide quantitative estimates of the regu- within this region at an instant of time, with the
larity of folding. The arc length is commonly, but aim of analyzing the rate of change of quantities
incorrectly, called the fold wavelength (Fig. 4.10b). of interest.
Our analysis considers the folded quartz vein to be The contrasting possibilities of interest are: (i)
isolated. What we principally mean by this is a negligible positive (or negative) value of the rate
that the folding is independent of that in nearby of change of fold amplitude, dA/dt, in which case
layers. That is, we do not see other nearby layers the layer will undergo nearly uniform thickening;
that are folded in concert with them. and (ii) a large positive value of dA/dt, correspond-
Note that the original vein thickness was not ing to the marked folding or buckling of the layer.
uniform and that this has affected the regularity To think about that, we have to imagine, in a con-
of the folding, but the arc length to thickness crete fashion, what the properties of the rocks
ratios fall within a modest range. What might involved are under the conditions that the folds
this regularity mean? Clearly, it cannot easily be formed. Clearly, they have deformed in a more or
ascribed to something built into the rock prior less continuous fashion. Although it takes a
to folding, since the regularity is present only in rather large leap, we might assume, for simplicity,
this layer. We must then ascribe it to some mech- they behave like other stiff, but still deformable,
anism inherent in the process of folding itself. uid-like media with which we are familiar, and
Before discovering an explanation, and, as an aid treat them as viscous uids. We then suppose the
to this, we consider a perfectly regular, periodic layer and medium have viscosities and 1,
structure of the same sort (Fig. 4.10c). This con- respectively.
sists of a layer embedded in a uniform medium, We now have a reasonably clear idea of a
with the upper and lower surfaces of the layer in model that might address some aspect of the
the form of in-phase sinusoidal surfaces with folding process so we turn to dimensional analy-
wavelength, L, and amplitude, A. The innite, per- sis. The quantities involved in the model, with
fectly periodic fold train is drawn with a low limb their dimensions, are:
dip.
The truly periodic fold train is an idealization rate of change in fold amplitude,
of the geometric form of a fold train and is used dAdt{}L T 1 (4.72)
here to isolate a small volume of rock containing bulk rate of shortening, Dxx {}T 1 (4.73)
a single fold from the remainder of the layer. fold amplitude, A{}L (4.74)
Because of the periodicity, the two vertical planes
layer thickness, H{}L (4.75)
in Fig. 4.10c are mirror planes of symmetry. For
wavelength, L{}L (4.76)
such a plane, two conditions apply. First, in the
deformation, which will be idealized to uniform layer viscosity, {}M L 1 T 1 (4.77)
layer-parallel shortening, except for the ow asso- medium viscosity, 1 {}M L 1 T 1 (4.78)
ciated with the folding, a particle cannot pass
through a mirror plane. Second, the shear stress There are seven physical quantities involving
must vanish at a mirror plane. three dimensions, M, L, and T, so the Buckingham
We model the case of a segment of the layer Theorem indicates that there are four dimen-
containing a single troughcresttrough fold sionless groups. Since M occurs only in the vis-
with a horizontal span given by the wavelength, L, cosities, one dimensionless group must be the
and an arc length, La L. To remind ourselves of viscosity ratio
the conditions that must apply at them, we 1
replace the bounding mirror planes by rigid R (4.79)

platens with smooth, frictionless, vertical sur-
faces. We suppose that these approach each other It is useful to choose groups that are relatively
at a rate corresponding to a rate of deformation, simple and that have a concrete physical or geo-
Dxx. As a natural starting point, we shall be con- metrical interpretation. An appealing choice is
cerned only in the elds of velocity and stress the aspect ratio, layer thickness to wavelength,
which we write in the form: We term this ratio the dimensionless, or normal-
ized, rate of growth.
k 2

H
L
(4.80) We now write (4.83) as a function of the other
dimensionless groups:
Since 2/L is the wavenumber, we may call
dAdt
(4.80) a dimensionless or normalized wavenum- f (k, A, R) (4.84)
A |Dxx |
ber. The aspect ratio, H/L, is appealing because it
immediately brings to mind the data on the ratio This states that the rate of growth, normalized by
of thickness to arc length, H/La, which we obtained the rate of shortening, may be expressed, for any
to express fold regularity. Including the wave- pair of viscous materials for which the viscosity
number in (4.80) anticipates special results of ratio is given, as a function of the aspect ratio H/L
folding theory. or the dimensionless wavenumber, k, and the
As a third dimensionless group, we take the limb dip, A. If we are correct in supposing that
maximum slope of the folded surface, or the limb the folds, or the positions of fold hinges, are estab-
dip of the fold: lished when the deections of the layer are still
small, we may then limit ourselves to discovering
tan A 2

A
L
(4.81) something about the growth rate under the
restriction A 1. In that case this dimension-
Here is the maximum dip angle. With the aspect less group makes a negligible contribution to the
ratio, k, and the dip, , we may readily visualize relation and write (4.84):
the fold form. One might wonder why the wave- dAdt
length, L, rather than the arc length, La, has been q(k, R) (4.85)
A |Dxx |
chosen. The implied reason is that we have tacitly
assumed that in the process of folding the thick- Here q designates the unknown function.
ness/arc length ratios, which clearly involve a If we further suppose that the relative rate of
random element, are selected during the initial growth of the fold takes place rapidly, so that the
phase of folding, when the folds have low aspect ratio remains the same, k may be treated as
limb dip. In the geometric idealization of such a a constant. Hence, the function on the right-hand
fold the arc length approximates the wavelength, side will be a constant for a given pair of materi-
La L. als, R, and a given aspect ratio, k. As we will show
Only one dimensionless group remains to be later, k will vary according to k(t) k(0) exp(2Dxxt),
set. This must involve the rate of change of ampli- where k(0) is its initial value. But, if q is a constant,
tude, dA/dt, and the rate of deformation, Dxx, we may integrate the above equation to obtain the
which have so far been left out. Excluding the vis- description of fold amplitude growth:
cosities, only two quantities contain the time, so A(k, t) A(k, 0) exp(q |Dxx |t) (4.86)
the dimensionless group will have the form:
Here, we are led to think about initial fold ampli-
dAdt tude. Does this mean, somehow, that the regular
(4.82)
X |Dxx | length scale of folds in the fold train is pre-
Here X is a quantity with the dimension of length, gured in the form of the layer? The answer is yes,
and must be selected from the possibilities H, L, in the sense that we can hardly expect to get some-
and A. Since the anticipated result is that dA/dt thing, i.e. folding, if the layer is uniform in thick-
0, and shortening implies Dxx 0, we take its ness and has perfectly planar surfaces to begin
absolute value to avoid negative dimensionless with. However, the answer is no, if we suppose that
groups. We choose X A, so the dimensionless the initial layer form contains a kind of perfect
group is: although very greatly diminished template of the
nal fold train form. What is present initially is
dAdt imperfection in the form of a gentle irregular, or
(4.83)
A |Dxx | random, waviness of the layer surfaces.
4.4 SCALED LABORATORY MODELS 143
We know that such waviness can be broken results that we will obtain in a later chapter, this
up, mathematically, into its Fourier wavelength relation is
components. Here, we shall suppose merely for
the sake of simplicity in visualization and analy- Ld 2
(4.87)
sis, that such waviness can be thought of as an H (6R)1 3
innite sum or superposition of two-dimensional
A given value for (L/H)mean, implies a particular vis-
cylindrical sinusoidal waves with axes normal to
cosity ratio R 1/. Issues arise in judging the
a single direction of layer shortening, each with
validity of this result, including whether it is
a different wavelength, L. As with many other
appropriate to treat rocks under the conditions of
wave phenomena, we suppose, further, that each
deformation as viscous uids. However, it is satis-
component behaves independent of all the
fying to obtain information on the fundamental
others, at least in approximation. Then, the equa-
properties of rock by dimensional analysis. Such
tion for A(k, t) may be thought of as describing the
properties clearly cannot be inferred from eld
growth or amplication, of each component, as a
observations alone.
function of its k-value. If then, q(k, R) varies in a
suitable manner, and, in particular, if it has a
single maximum at some value of k kd corre-
sponding to the ratio Ld/H, the amplication of 4.4 Scaled laboratory models
the fold components will be selective. Here the
subscript d refers to the dominant wavelength. It seems inevitable that model experiments coupled
That is, as time or layer shortening goes on, the with theoretical analysis of the dynamics of tectonic
component at Ld/H will receive the greatest processes will contribute greatly to a sound, coherent
amplication, and the shape of the layer will be theory of structural geology and tectonics. By running
scale models of tectonic events, one may ultimately
dominated by a superposition of fold components
hope to separate the physically possible from the phys-
with L/H at or near this value. In this manner, the
ically impossible hypotheses, and the former may be
regularity seen in the nal conguration will be studied in detail to illustrate tectonic processes to an
established. extent not otherwise possible (Ramberg, 1967).
When the folded form of the layer, which will
inherit substantial irregularity from the random- Both the length scale and the time scale for many
ness in amplitude and phase of the initial wavi- tectonic processes make direct observation impos-
ness, reaches some stage in its development, the sible. In terms of length, we have no difculty
independent growth of individual wavelength observing the surface of the Earth at the neces-
components will cease, and the form established sary scale, but observations are extremely limited
at that point, as in the positions of fold hinges, at depth. Mines and wells are few and far between,
will be locked in. If folding or buckling accom- and modern imaging technologies (e.g. three-
plishes a shortening of the span of the layer with dimensional seismic reection), while vastly
minor changes in its thickness, the spacing of improved, typically provide data only from loca-
hinges along the layer, in terms of arc length, will tions of interest to the oil and gas industry. In
tend to be preserved. Thus, the data collected from terms of the time scale, some tectonic processes
our fold train may be referred back to the time of take millions of years to develop and their charac-
cessation of selective amplication. teristic rates prevent most attempts to monitor or
It is reasonable to suggest that the mean value investigate the phenomena directly. If, as Hans
of L/H provides an estimate of the value Ld/H, or the Ramberg suggests, we can make models in the lab-
so-called dominant wavelength/thickness ratio. If we oratory that reproduce these processes, we can
knew the form of the function q(k, R) we might directly observe the model structures as they
then nd Ld/H as a function of the viscosity ratio, develop and gain important insights. Rambergs
R, and thus use the L/H data to estimate the ratio opinion was written at a time when numerical
of the viscosities of the medium and the layer at models of tectonic processes were still under
the time the folding occurred. To anticipate development, and these now offer an alternative
to laboratory models. In this section the con- (a)

straints imposed by building a model with much
smaller length scale and much shorter time scale
are examined, and rules are described to ensure
similarity between the natural process and the
model process.
M. King Hubbert (1937) wrote one of the earli- Lp
est papers that advocates a laboratory modeling
approach to problems in structural geology and
points out the requirements for scaling a model. Hp
These activities continue today and, indeed, the
number, range, and sophistication of such experi-
ments have increased greatly in the last few
decades, with several multi-investigator laborato- (b) (c)
ries active. For example, researchers at the
Applied Geodynamics Laboratory of the Bureau of
Economic Geology, University of Texas, have
studied the rise of salt domes and salt withdrawal
(Vendeville et al., 1995; Ge and Jackson, 1998).
Those at the Fault Dynamics Project at Royal
Holloway, University of London, have studied a Fig 4.11 (a) Exposure photograph of banded metamorphic
variety of structural styles and faulting mecha- rock with small folds. (b) Same exposure reduced by a factor
nisms in extensional tectonic settings (McClay et of two with geometric similarity maintained. (c) Same
al., 1991; McClay and White, 1995). Researchers in exposure but not geometrically similar. Photograph by D. D.
France at Universit Rennes are using laboratory Pollard.
models to study the development of growth fault
systems (Manduit and Brun, 1998), while others at
Universit de Montpellier are investigating fault- and rarely on a piece of paper. These two types of
ing in accretionary wedges (Gutscher et al., 1998). experiments have something in common: both
Researchers in Canada at the Experimental types obey the fundamental laws of mechanics.
Tectonics Laboratory at Queens University are The theoretical models do this by design, whereas
investigating foldfault relationships and the the laboratory models do this because they are
inuence of stratigraphic heterogeneities on part of the natural world from which those laws
faulting (Liu and Dixon, 1990, 1991). Despite the were derived. The same dimensionless groups of
difculty of mimicking the behavior of rock over physical quantities that appear in the governing
geological time with sand and putty on a labora- equations for a particular process in the Earth
tory bench, laboratory model studies can provide should be used to scale the laboratory models of
important quantitative insight, and in most this process.
instances they record interesting relationships
between the applied loading conditions and the 4.4.1 Geometric and kinematic similarity
development of structures that appear similar to Figure 4.11a is a photograph of an exposure of a dis-
those in the Earth. In some cases model experi- tinctly banded metamorphic rock that displays
ments have been performed on actual rock beautiful folds. The 50-mm lens cap (lower center)
samples at elevated pressures to investigate fault- provides a length scale, so we know that the actual
ing (Patton et al., 1998) or folding (Couples and wavelength of the prominent gray band near the
Lewis, 1998). middle of the photo is about Lp 212 mm and the
Laboratory experiments are termed analog or height of this layer measured at its lowest point is
physical models whereas computational or numeri- about Hp 12 mm. We refer to the outcrop with
cal models are performed usually on a computer the banded fold as the prototype. It is simple to
make a visual model of this outcrop that honors experiment. The time is measured from some arbi-
the geometry of the prototype using image pro- trary moment, often at the initiation of the
cessing software (Fig. 4.11b). Here the wavelength process, and the model ratio for time is dened as:
and height are halved in value, so Lm 106 mm and
Tm
Hm 6 mm. The model ratio, Lr, for the lengths is: Tr (4.91)
Tp
Lm 1
Lr (4.88) Given this ratio, one can compare the prototype
Lp 2
and a model at corresponding times during the
Any other length that we measure on this outcrop development of the process. We say that the pro-
is related to the corresponding length in the totype and the model are kinematically similar if
model by this same ratio. For example, for the they are geometrically similar at every corre-
height of the gray band we have: sponding time over the duration of the process. By
corresponding time we mean a time for the pro-
Hm 1
Lr (4.89) totype process and a time for the model process
Hp 2
that are related by the model ratio for time.
The prototype and model are geometrically similar. Kinematic similarity can be understood in
The image of the outcrop shown in Fig. 4.11a is terms of two motion pictures, one of the model
scaled down to a smaller size to t on the page of and the other of the prototype. Lets say the
this book. Thus, the lens cap displayed on the model ratio for time is Tr 3.1 1011, so each
gure is not 50 mm in diameter, but it represents model second represents one thousand prototype
50 mm on the actual exposure. The photographic years. The camera recording the model process
image is, itself, a geometrically scaled model of shoots at a speed of one frame per second and the
the exposure and the process of reproduction of camera recording the prototype process shoots at
this image maintains geometric similarity. a speed of one frame per thousand years. If each
Another model of the exposure is shown in successive pair of frames of the two motion pic-
Fig. 4.11c. Is this geometrically similar? The eld tures is geometrically similar, the two processes
of view has changed somewhat, but the image also are kinematically similar. The corresponding
seems to have many similarities to the prototype frames may have different length scales, but
shown in Fig. 4.11a. On the other hand, a careful lengths throughout the prototype and model
inspection reveals that the fold shape is distorted obey the model ratio for lengths.
relative to the prototype. We measure the wave- Turning to a geological example, it is likely
length in this model as Lm 106 mm and the that the layers shown in Fig. 4.11a had lesser
height as Hm 12 mm. Comparing the model amplitudes and greater wavelengths at an earlier
ratios we nd: time in the folding process. In Fig. 4.12 three dif-
ferent stages in the hypothetical one million year
Lm 1 H 1
Lr, m Lr (4.90) development of the prototype fold are illustrated,
Lp 2 Hp 1
assuming that the deformation conserved
The scaling in the horizontal direction is the same volume and that the shortening in the direction
as that used to produce Fig. 4.11b, but lengths of the measured wavelength is simply the recip-
measured in the vertical direction are the same as rocal of the elongation perpendicular to this
those in the prototype. Thus, the height of the direction. We start to record the process at an
layer in the prototype and model are identical. arbitrary time (0 s) shown in Fig. 4.12a. Five
This model does not preserve geometric simi- hundred thousand years into the process the pro-
larity. Comparing the lens cap in the three gures totype fold would look like the image in Fig.
conrms the scaling relations. 4.12b, and at the end of the one million years the
Tectonic processes may be very slow to develop, fold would have attained the shape observed in
but they are not static. Thus, the relative time outcrop today (Fig. 4.12c).
scales for the model, Tm, and the prototype, Tp, must For a kinematically similar model of this
be considered carefully when designing a model folding process using the model ratio proposed
(a) Time, t0 typically involve the dimensions of length, mass,

and time. The model ratios for these fundamental
quantities are:
(b) Time, t1 Lm M T
Lr, m Mr, m Tr (4.92)
Lp Mp Tp
To maintain strict similarity, all correspond-

ing lengths, masses, and times in the model and
the prototype must adhere to the given model
(c) Time, t2 ratios.
Model length scales typically are chosen for
convenient experimentation on a laboratory
bench, without the necessity of a microscope, so
they vary from a decimeter to perhaps a meter.
Prototype length scales can range from a meter to
hundreds of kilometers. Model materials usually
do not vary in mass density by more than a factor
Fig 4.12 Snapshots of folding process with kinematic
of two from the densities of common rocks, so the
similarity. (a) Initial time. (b) After 500 000 years. (c) After
1 000 000 years. Photograph by D. D. Pollard. model ratio for density is of order one: r m/p
100. However, the model ratio for mass is the
model ratio for density times the cube of the
above, Tr 3.1 1011, one model second is one model ratio for length. Thus, the length ratio can
thousand prototype years. At the beginning (0 s) impose very great differences between the mass of
and at the ending (1000 s) of the process, the model and that of the prototype. The time scales
model would look exactly like the images in Fig. for experiments are determined by convenience
4.12a and c, respectively. At an intermediate and necessity, and a few minutes to a few days
stage, after 500 s, an image of the model would duration is typical. Some tectonic processes (e.g.
look exactly like Fig. 4.12b. In order to achieve fracture propagation) may operate at these
this kinematic similarity the motions of all parti- human time scales, but most are believed to
cles in the model, when appropriately scaled for develop over thousands to millions of years.
the model length and time ratios, must mimic Common ranges for the model ratios of the fun-
the motions of corresponding particles in the damental quantities are:
prototype.
106 Lr 1,1018 Mr 1,
10 10 Tr 1 (4.93)
4.4.2 Dynamic similarity
Most of us are familiar with the concept of geo- The model ratios for all three fundamental quan-
metric similarity and can easily recognize intu- tities vary from very small numbers to about
itively when certain lengths are distorted relative unity.
to others. Most of us are less familiar with kine- The derived physical quantities can be evalu-
matic similarity, but we can recognize changes in ated in terms of their model ratios as well. Some
time when a motion picture of everyday scenes is important examples are:
speeded up or slowed down.
On the other hand our intuition usually is not
Vm L3m
well developed when it comes to judging the sim- volume, L3r (4.94)
Vp L3p
ilarity of a prototype and model with respect to
forces. In this regard we must turn to calculations am Lm T2
m
acceleration, Lr T2 (4.95)
to test for dynamic similarity and these calculations ap Lp T2
p
r
m Mm L3
m follows that any ratio of two different forces in
mass density, Mr L3
r (4.96)
p Mp L3
p the prototype must equal the corresponding ratio
of those different forces in the model. For
Fm Mm Lm T2
force, m
Mr Lr T2
r (4.97) example, considering a process in which inertial,
Fp Mp Lp T2
p viscous and gravitational forces are present, we
nd:
m Mm L1
m Tm
2
stress, Mr L1
r Tr
2 (4.98)
p Mp Lp Tp
1 2 Fmi Fpi
, or Rem Rep (4.100)
Fmv Fpv
The derived quantities are reduced to their equiv-
alent fundamental quantities and these are used Fmi Fpi
, or Frm Frp (4.101)
to dene the model ratios. Fmg Fpg
Notice that the model ratio for forces contains
the ratios for length, time, and mass. Length and Recall that Re is the dimensionless group called
time are considered through their respective the Reynolds Number, dened in (4.55). Dynamic
model ratios to assure geometric and kinematic similarity requires that the Reynolds Numbers for
similarity. Mass usually is not considered inde- the model and prototype be identical. These
pendently, but rather through the analysis of the numbers can be evaluated using a characteristic
force ratio, which is used to evaluate the dynamic length, wo, a characteristic velocity, vo, the density,
similarity between a model and the prototype. To , and the viscosity, , for both the model and pro-
assure similarity one must identify all of the dif- totype. The Froude Number, Fr, measures the rela-
ferent forces acting in the tectonic process under tive importance of inertial and gravitational
investigation. For example, in our study of the forces. From (4.52) and (4.64) we have:
dimensionless groups for ow of magma in a
v2o
conduit (Fig. 4.8) and rise of a salt diapir (Fig. 4.9) Froude Number Fr ,
gwo
we identied inertial forces (4.52) caused by the
change in velocity with time, viscous forces (4.53) inertial force
(4.102)
caused by the drag of magma against the side of gravitational force
the conduit, and gravitational forces (4.64)
The Froude Numbers for the model and prototype
related to a density contrast. A model and proto-
can be evaluated in terms of characteristic veloci-
type that are geometrically and kinematically
ties and lengths, and the acceleration of gravity,
similar, are said to be dynamically similar if the
to assure dynamic similarity.
model ratios for all of the forces acting on any two
This example points out a second important
corresponding particles are equal (Ramberg,
role for dimensionless groups in structural
1967, p. 4):
geology. We have already shown how useful they
Fmi Fmv Fmg are for understanding and interpreting the equa-
Fme
Fr (4.99) tions that govern mathematical models of tec-
Fpi Fpv Fpg Fpe
tonic processes. Now we see that they are useful in
the design of scaled laboratory model experi-
Here the subscripts m and p refer to the model
ments of these processes. Besides inertial, viscous,
and prototype as before, and the subscripts i, v, g,
and gravitational forces, there are likely to be
and e refer to inertial, viscous, gravitational, and
forces associated with spatial gradients in pres-
elastic forces, respectively. There may be other
sure or stress:
forces that should be considered.
It would be a daunting task to evaluate the p
forces acting on all corresponding particles for a pressure (stress) force per unit volume
wo
prototype and a model, but fortunately a simpler (4.103)
procedure usually is adequate. Because the ratios
of particular forces in the prototype and model all Here p refers to a characteristic change in pres-
must be equal to a common model ratio, Fr, it sure (or stress) from one location to another. These
forces are used to dene additional dimensionless (a)

groups that may be useful in model design
(Ramberg, 1967, p. 40):
pwo
Stokes Number St ,
vow 2o
pressure force
(4.104)
viscous force
g
Ramberg Number Ra ,
vo w2o
gravitational force
(4.105)
viscous force
g
Smoluchowski Number Sm ,
pwo
gravitational force
(4.106)
pressure force
(b)
4 5
Ramberg has referred to the ratio of gravitation to
viscous forces as a nameless ratio but we 3 6
suggest the Ramberg Number in recognition of the
major contribution he made to tectonic model- 2
1 7
ing. Ramberg suggested the name for the
Smoluchowski Number because of the two papers 8
published by M. Smoluchowski in which a theory
for buckling of elastic layers on a viscous founda-
tion is developed along with associated model
Fig 4.13 Centrifuge models and equipment (Ramberg,
experiments (Smoluchowski, 1909). The Stokes
1967). (a) Model of salt dome development. (b) Centrifuge
Number is named for one of the pioneers in uid
used to increase the body force: 1, model in centrifuge cup;
dynamics who solved some of the classic prob- 2, stroboscopic light reflector; 3, camera; 4, camera
lems for slow viscous ow. For strict dynamic sim- electronics; 5, stroboscope; 6, temperature and speed
ilarity each of these numbers for the prototype control unit; 7, motor; 8, refrigerator.
must be equal to the corresponding number for
the model.
The procedure for tectonic model design may 2.16 103 kg m3
be further simplied when some of the character- g 9.8 m s2
istic forces are much smaller than others. For
example, inertial forces typically are much The radius of the prototype salt dome ranges
smaller than either the viscous or gravitational from 1.5 to 3 km, and the velocity ranges from
forces, so the Reynolds and Froude Numbers are 1 cm a1 to 1 m a1. These values are inferred
very small. For example, in the rise of a salt dome from eld observations and interpretations of
the relevant parameters take on the following geological records. The density and viscosity are
ranges (Ramberg, 1967, pp. 489): values estimated from laboratory measurements.
Density is well known, but viscosity is poorly
1.5 103 m wo 3 103 m constrained because it can vary over many orders
of magnitude depending upon temperature, pres-
3 1010 m s1 vo 3 107 m s1
sure, and chemical environment. Based on the
1 106 N s m2 1 1016 N s m2 (4.107) ranges of values given above, the Reynolds
Number for the rising salt dome would range

from 1.9 106 to 9.7 1020, and the Froude
Lm
Lp
T 2
m ,so Tm Tp
Tp
Lm
Lp
(4.109)
Number from 6.1 1018 to 3.1 1024. These very

small ratios justify ignoring the effects of inertial Using a reasonable range of model ratios for
forces when scaling models of salt dome develop- length (106 Lr 1) and for the time scales for
ment. A laboratory model of salt dome develop- tectonic processes (102 s Tp 1015 s), we calculate
ment is illustrated in Fig. 4.13a (Ramberg, 1967, a range of model time scales from Tm 101 s to Tm
p. 123). 1015 s. At the lower end of this range the experi-
Many experimental observations of viscous menter would have few problems, but clearly the
ow in conduits have demonstrated that the upper end is unattainable. For example, a
Reynolds Number must be greater than approxi- common model length scale is ten centimeters
mately 1 103 for the ow regime to become (0.1 m), corresponding to a prototype length scale
turbulent (Bird et al., 1960, pp. 1838). For of one kilometer (1000 m). A common time scale
Reynolds Numbers less than this value, ow in for the duration of tectonic processes is one
conduits is laminar (Fig. 4.8b). Because the range million years (106 years). Given these values, the
of Reynolds Numbers estimated for the rising salt model time scale is ten thousand years (104 years),
dome (and many other tectonic processes involv- not a practical duration for experiments designed
ing ow) is many orders of magnitude less than to be observed by humans!
this transition value, the ow regime is laminar. The constraint on the model time scale
The precise scaling of inertial forces between the imposed by similar accelerations of gravity acting
model and prototype is not necessary as long as on the prototype and the model is not fatal to all
the model is well within the laminar ow regime. model experiments in tectonics because the iner-
The condition for strict dynamic similarity in tial forces associated with accelerations in the
terms of inertial forces between model and proto- prototype may be insignicant compared to other
type can be relaxed without compromising the forces. Recall, for example, that ow of many geo-
usefulness of the experimental results. logical materials is in the realm of low Reynolds
Number and low Froude Number ow. With very
4.4.3 Gravitational forces: a problem for small ratios of inertial to viscous forces (4.55) and
model similarity? inertial to gravitational forces (4.102), the inertial
It was noted in the preceding section that all cor- forces can be ignored.
responding lengths, masses, and times in the There is, however, a constraint imposed on the
model and the prototype must adhere to the given strength of model materials that must deform
model ratios to maintain strict similarity. In under their own weight to simulate the defor-
general these are thought of as independent mation of very large masses of rock. Consider-
ratios, to be chosen at the convenience of the ing the model ratio for gravitational forces we
experimenter, or as constrained by the available have:
materials. However, models of tectonic processes
often are carried out on a lab bench where the Fmg mgmL3m mL3m
(4.110)
acceleration of gravity is approximately 9.8 m s2, Fpg p gp L3p p L3p
and this value is not signicantly different for the
prototype. Thus, the model ratio for accelerations Here the ratio of gravitational acceleration in the
is approximately one. Examining the model ratio model and prototype is approximately one. Given
for accelerations we nd: this ratio of gravitational forces, the correspond-
ing ratio of stresses and strengths is:
ar Lr T2
r 1,so Lr T r
2 (4.108)
Thus, the model length and time ratios are not mg mLm Lm
(4.111)
independent. Solving for the model time: pg pLp Lp
Here the ratio of densities in the model and pro- to increase the body force in the model by placing
totype is approximately one. This analysis leads to it in a centrifuge (Fig. 4.13b) (Ramberg, 1967;
the conclusion that the strength ratio must scale Dixon and Summers, 1985). For example, if the
as the length ratio. Using a reasonable range of dominant forces operating in the prototype are
model ratios for length (106 Lr 1) we conclude pressure, gravitational, and viscous forces, the
that the model ratios for strength must be in this Stokes (4.104), Ramberg (4.105), and Smolu-
same range (106 r 1). The upper end of this chowski (4.106) Numbers must be examined and
range corresponds to laboratory experiments on shown to be equivalent to the corresponding
meter-scale prototypes such as small folds and numbers for the model. Thus, comparing the
these can be effectively modeled with relatively Smoluchowski Numbers for the model and pro-
strong materials (Ramberg, 1963). At the lower totype, one must show that:
end of this range the experimenter must try to
model processes with length scales of kilometers
or tens of kilometers in the laboratory. If gravi-

g wo
p m

g wo
p p
(4.112)
tational forces induce the deformation, very weak If the gravitational force as measured by g is
materials are required for the model to meet the approximately the same in the prototype and
constraint imposed by the model ratio for model and the model ratio for lengths is 106,
strengths. then the model ratio for pressure (stress,
M. King Hubbert (1945) brought this point to strength) also must be 106 and this is not easily
the attention of geologists in an article entitled attained. On the other hand if the body force in
Strength of the Earth that examined the appar- the model can be articially increased in a cen-
ent contradiction between the great strength of trifuge, such that the model ratio for accelera-
a hand sample of rock and the modest strength tions is 103, then the model ratio for strength
of a huge rock mass containing innumerable need only be 103. Materials with strengths
folds and faults that witness to its apparent required to meet this constraint are readily
weakness. To drive home this point Hubbert pro- obtained for laboratory experiments. Similar con-
posed the operation illustrated in the fron- clusions about the appropriate viscosity for
tispiece to this chapter in which the state of model materials are found by examining the
Texas is lifted by a huge crane. Of course this Ramberg Number.
would be impossible to implement, so he consid-
ered a laboratory model of such an operation. By
considering the scaling of this model he con- 4.5 Concluding remarks
cluded that it would be impossible under the
existing force of gravity. In this chapter we introduced the material contin-
Consequently, if we tried to lift such a block in the uum, a construct that has produced astounding
manner indicated . . . the eyebolts would pull out; if we results in both fundamental physics and applied
should support it on a pair of saw horses, its middle engineering, including solid deformation, uid
would collapse; were we to place it upon a horizontal ow, and heat transport. Most human-made objects,
table, its sides would fall off. In fact, to lift it at all from automobiles to spacecraft, from bridges to
would require the use of a scoop shovel . . . The dams, from golf clubs to bicycles, are designed using
inescapable conclusion, therefore, is that the good continuum mechanical principles. Furthermore,
state of Texas is utterly incapable of self-support many of these objects are built on or with machines
(Hubbert, 1945).
that were, themselves, designed using these princi-
The scaling of model experiments has proven ples. The success of this way of thought should be
to be problematic because materials that ow beyond dispute, but curiously the application of
and fracture under their own weight are rare, yet continuum mechanics to structural geology lags
large masses of rock subject to gravitational considerably behind applications in other scientic
forces do just that. A solution to this dilemma is and engineering disciplines. This textbook is
designed to encourage students of structural groups, and the scaling of tectonic processes. These
geology to use the concept of the material contin- should be readily available tools to take in the eld
uum. We also reviewed some bread and butter as mapping progresses and questions arise, and to
techniques for the practicing structural geologist use at the desk when puzzling over the result of a
including dimensional analysis, dimensionless computation.
Chapter 5
Deformation and flow
Nearly undeformed (upper left) and deformed (lower right)

oids with spherulitic cores viewed in thin section. Matrix is
calcite with some mud and average ratios of long to short
axes are 1.16 and 1.56 respectively. Samples from near
Harrisonburg, VA. Reprinted from Cloos (1971, Plates 9 and
11) with permission of The John Hopkins University Press.
A clear separation between geometrical and dynamic deformation which can be considered independently
considerations was maintained by Becker, the of force, mass, chemical composition, elasticity, heat,
American geologist, at a very early date, and he magnetism and electricity; and it is of greatest use to
referred to the English physicist Thomson (Lord Kelvin), science for such properties to be considered as a rst
who says very clearly: We can see, therefore, that there step (Sander, 1970, p. 12).
are many attributes of movement, displacement, and
DEFORMATION AND FLOW 153
M
ost structural geologists think about (a)
deformation and ow in an inverse
problem mode: from the nal state back
toward the initial state of the deformed body of
rock. For example, the lower right photograph in
the frontispiece for this chapter shows deformed
oids with elliptical shapes: ratios of long to short
axes are about 1.56 (Cloos, 1947, 1971). The upper
left photograph shows nearly undeformed oids
with approximately circular shapes: ratios of long
to short axes are about 1.16. One can think of the
deformed oids being transformed back toward
an initial state much like that of the nearly unde-
formed oids. In other words the ellipsoidal par-
ticles become nearly spherical. This viewpoint is
natural since the primary observational data of
structural geologists are eld observations of
(b)
deformed rock. y
The above admonition by the Austrian geolo- P(x,y)
gist Bruno Sander, written in 1948 (Sander, 1970), b
a x
suggests that we set aside much of the physics in O
our initial study of rock deformation and focus ex-
clusively on kinematics. Becker and Sander are
prominent participants in the early history of
a
structural geology. Their advice is followed to this
date by an influential school of structural geolo- Fig 5.1 (a) Broken and extended fossil belemnite with
gists, to which the present authors do not sub- quartz, q, and calcite, c, filling. (b) The deformed shape of an
scribe (Fletcher and Pollard, 1999; Pollard, 2000). initial unit circle with radius of one undeformed belemnite
is an ellipse with semi-axes a and b. The trace of cleavage
Instead, we take kinematics as an integral part
(lineation) is parallel to the long axis of this strain ellipse and
of a complete mechanical analysis, including
the orientation and stretch of the belemnite are represented
enough physics to formulate a well-posed problem. by the radial line OP. Reprinted from Badoux (1963) with
Therefore we devote this chapter to the subject of permission of Universit de Lausanne.
kinematics and succeeding chapters to the other
elements of a complete mechanics.
Other types of observations enter a synthesis such processes as faulting, folding, or mountain
leading to understanding of a process of deforma- building at plate margins. Field observations of
tion and its products. For example, controlled many sorts, including the deformed shapes of
laboratory experiments produce detailed infor- objects such as oids may be selected to constrain
mation on the behavior of rock materials under a forward model built upon a synthesis of much
load, ranging from brittle failure at modest tem- previous information (Nur et al., 1986; Pachell,
perature and pressure to slow creeping ow at et al., 2003).
high temperature and pressure. Moreover, theor- From the more traditional perspective the
etical study of deformation and failure of a wide result of a process of deformation, such as one of
range of materials and human-made structures the deformed objects shown in Fig. 5.1 and in the
over the last two centuries has provided a rened frontispiece to this chapter, is described, and the
picture of the processes involved and the laws gov- relationship between its present and initial forms
erning them. The considerable insight and infor- is worked out. Applied to one or more such
mation provided by these additional sources tend objects, this procedure may lead to an estimate
to favor the formulation of a forward problem for of the strain of the rock containing them. The
154 DEFORMATION AND FLOW
procedure chiey involves considerations of geom- direction in which the rock has undergone the
etry, although assumptions must be made as to the greatest extension (Wood and Oertel, 1980). A
initial forms of the objects. However, as the above belemnite oriented parallel to the cleavage would
quotation suggests, no reference is made to the show the maximum value of the ratio l/l0, provid-
underlying physical process that resulted in the ing a measure of the bulk rock deformation in that
deformation. We begin this chapter by working direction.
through a few examples. Interpretation of the Estimation of the maximum stretch from the
examples involves many simplifying assumptions information available at the exposure, requires
that set aside complicating factors, which we other assumptions. One is that the stretch of the
return to after developing the other parts of a com- belemnite was equal to that of any other linear
plete mechanics in subsequent chapters. material line in the rock with the same orienta-
tion, for example, one lying along the line l on
the gure, or one much farther from the belem-
5.1 Rock deformation: some nite. This assumption implies homogeneity of defor-
mation within some volume of rock containing
observations and a simple the belemnite. In detail, the belemnite itself has
description not undergone a homogeneous deformation, but
has broken into pieces that have separated by
5.1.1 Deformed belemnite means of a process involving precipitation of
The closely lined segments of the stretched belem- material into the gaps. The belemnite might have
nite in Fig. 5.1a were initially connected to form an been more resistant to extension than the sur-
intact fossil (Badoux, 1963). First quartz and later rounding rock, and thus its stretch would under-
calcite was precipitated into the gaps between the estimate that of the rock.
segments as stretching occurred. Measuring the We have introduced the notion of homo-
nal span occupied by the segments, l, and initial geneous deformation in a simple way. As we show
length, l0, yields a measure of the elongation of a later in this chapter the technical denition of a
material line, the belemnite or its centerline. An deformation is the transformation that relates
assumption is that the segments did not stretch: the positions of particles in a body in one state to
the belemnite fractured into pieces that behaved as that in another. Here we speak of these two states
rigid objects embedded in a deformable medium, as the initial state and the nal state. A particle is
the rock containing it. The nal length, measured an element of mass within the rock of small
from the gure, is l 153/60", measured with a dimensions relative to the length scale of interest.
scale marked off in 60ths of an inch, and the sum Here, the only length scale is the length of the
of the segment lengths is l0 103/60". The ratio l/l0 belemnite, so small is taken relative to that. The
1.48 termed the stretch of a material line is a position of a particle is given by its position vector
dimensionless measure of the elongation. Thus, between a coordinate origin and the particle. In
any units may be used for the measurements. The this more formal context homogeneity is dened
use of fossil belemnites to estimate strain goes back as a linear transformation between positions in the
at least to the late-nineteenth century (Cloos, 1946; initial and nal states.
Hossain, 1979). The intent here is to estimate, from the elonga-
The motivation for study of the belemnite is not tion of the belemnite and the angle between it and
for its sake alone, but for what it might tell us cleavage, the maximum elongation in the plane of
about the deformation of the rock containing it. cleavage. We also assume, for this example, that
The belemnite lies in the bedding plane and is the plane of cleavage is normal to the bed contain-
inclined at an angle to cleavage, represented by ing the belemnite. The data given provide no basis
the parallel line segments in the gure. Much for this assumption, but allow us to consider only
research has shown that the plane of cleavage is the plane of particles making up the bedding
normal to the direction in the rock that has under- surface and thereby reduce the problem to two
gone the greatest shortening, and contains the dimensions.
5.1 EXAMPLES OF ROCK DEFORMATION 155
Finite strains in the W. Helvetic nappes, 0 1 2 km

Valasi, Switzerland
UH Wildhorn Nappe UH
Diablerets Nappe
UH
NW
Ultra-Helvetic nappes SE
Morcles Nappe
Top of Cretaceous
Thrust contact
Top of Jurassic
Unit circle
Strain elipse (XZ)
0 1 2
km
Aiguilles Rouges Massif
Fig 5.2 Strain distribution represented by strain ellipses in

a down plunge section of folds through the western Helvetic
was deformed to an ellipse with the same area. The
nappes (Ramsay and Huber, 1983). condition for this is that ab 1. Using this, a 1.61
and b 1/a 0.62. The ellipse, with lines repre-
senting the cleavage and belemnite, is shown in
A linear transformation will take a circular Fig. 5.1b. If the cleavage were normal to the plane
locus on the bed surface and transform it into an of the drawing in Fig. 5.1a, the ellipse would be a
ellipse. Since the long axis of the ellipse will lie principal section of the strain ellipsoid, the three-
parallel to the cleavage and we know the angle dimensional surface that would result from the
between the cleavage and the centerline of the homogeneous deformation of a spherical surface.
belemnite, we may try to compute the maximum Interpretation of naturally deformed objects to
stretch from the stretch of the belemnite, l/l0 yield the strain for some representative volume of
1.48 and the angle of 25. We write the para- rock has a large literature in structural geology.
metric equations for an ellipse as: The volume of rock considered is kept sufci-
x a cos ently small so that the deformation may be approx-
(5.1) imated as homogeneous. Much systematic
y b sin
research has led to results of the type shown in Fig.
If the radius of the ellipse is 1.48 at 25, we 5.2, here for the 10-km scale structures of the
have the relation: western Helvetic nappes of the Alpine orogen
a2 cos 2 (25) b2 sin2 (25) (1.48)2 (5.2) (Ramsay and Huber, 1987). Principal sections of
strain ellipsoids that are approximately vertical are
Since this is only one equation in two unknowns, plotted in a vertical section through three stacked
we need further information or assumptions in and folded sheets of sedimentary rock, called
order to compute the maximum stretch. One nappes, in the western Alps in Fig. 5.2. These give a
approach would be to nd another, differently synoptic picture of the deformation in the strongly
oriented belemnite in the bedding plane, thus deformed nappes. The distributions of their mag-
providing another equation of the above form. nitude and orientation, simultaneously repre-
Another assumption is to suppose that an initial sented in this gure, may be used to think about
unit circle of radius one undeformed belemnite, the process of nappe emplacement. For example,
(a) (a) 7177 A C

58136
3.
3.46
18
1.9
2.
4
37 728 E
2.
2.0
2. 69 B
9
5 91
3.4
5.
16 .31
4
630 E+F
728 H
4.4
85 B" 5.39 58-2
629A LURAY INGHAM BERRYVILLE

4.8
4
(b)
717 5.67
B.C
l0
(b)
l
Fig 5.3 (a) Folded quartz vein in schist; the lens cap is 5 cm
in diameter. (b) Sketch of fold with initial arc lenth, lo, and
final length, l. Photograph by R. C. Fletcher.
sliding or shearing of these vast rock masses, one

upon the other, is indicated by the very large
strains (large aspect ratio strain ellipses) present at 1.39 1.55 2.21 3.46
the base of each nappe. Notice that the strain varies
spatially. In the right-side-up strata at the upper Fig 5.4 (a) Cross sections of the South Mountain Fold.
part of the nappe, the strain is small to moderate, Mean shapes of deformed oids are given as the strike-
normal section of the strain ellipsoid, its axial ratio, and its
relative to the much larger strains near its base. The
orientation. (b) Deformed olite sections, with aspect ratios.
belemnite of Fig. 5.1 comes from a quarry at
Reprinted from Cloos (1971, Figs. 28 and 2) with permission
Leytron, Switzerland, just above the axial surface of The Johns Hopkins University Press.
of the Morcles Nappe, at which the beds switch
from overturned below to right-side-up above.
maximum elongation occurred at right-angles to
5.1.2 Other examples: folds and folded the contraction in the plane of the gure, and the
olites area in the plane was unchanged in the deforma-
The quartz vein in Fig. 5.3 may be used to provide tion, the corresponding elongation would be the
information on the strain in the rock containing reciprocal of the above ratio, or 1.42, and this
it. The rock may have been deformed before the stretch refers to the material lines perpendicular
quartz vein formed, so the information pertains to the line labeled l.
only to the subsequent deformation. If it is The horizontally attened and vertically elon-
assumed the vein buckled into the folded form gated oids of Fig. 5.4b were initially approxi-
without changing its arc length, the nal length l mately spherical in form. The folded layers in
260/60" and the initial length l0 370/60" give a which the oids occur, called olites, were ini-
ratio l/l0 0.70. This ratio is less than one, and, tially at lying and uniformly thick. The vertical
also called a stretch, represents a contraction in cross section of the fold illustrated in Fig. 5.4a
the direction parallel to the line labeled l. If (Cloos, 1947, 1971) shows the current traces of two
5.1 EXAMPLES OF ROCK DEFORMATION 157
layer interfaces. To change them to this form, a the belemnite of Fig. 5.1 in the initial and nal
spatial distribution of relative motion between states. We refer these positions to coordinate axes
particles occurred. Each elliptical section in the xed in the bedding plane, with origin xed to a
gure represents the strain ellipse for some rock particular particle, e.g. the tip of the belemnite.
volume obtained by combining many oid shape Under the postulate of homogeneous deforma-
measurements, assuming homogeneous deformation, the nal coordinates of a particle, x and y,
tion over the sample area, and that these objects, may be related to its initial coordinates, denoted
as the belemnite in Fig. 5.1, underwent the same X and Y, by linear equations describing a linear
deformation as the rock. If oids were stiffer than transformation:
the matrix around them, they would have a strain x x(X, Y ) AX BY
less than that of the bulk rock. The numbers given (5.3)
y y(X, Y ) CX DY
are the long to short axial ratios, a quantitative
measure of the strain. Since the initial ratio would Here A, B, C, and D are constants. The mathemati-
have been unity for spherical oids, a ratio of 4 cal statement x x(X, Y), where x occurs on both
represents an increase in the long diameter by a sides, is common in continuum mechanics and
factor of 2 and a decrease in the short diameter by other branches of physics. The x on the left-hand
a factor of 12 . side denotes the numerical value of the function
Data on the distribution of the strain in the x on the right-hand side for any pair of arguments
fold provide a constraint in addition to layer X, Y. To use another letter means that our descrip-
shape, on the deformation that took place. For tion will involve many additional letters and that
example, we might suppose that the fold was we will have to keep track of what they refer to,
formed when the rock layer was bent into a form e.g. if we had written x f(X, Y), we would have to
of roughly a semi-circle with little change in remember that f went with x. This way, the expres-
thickness. The ratio of the diameter to the cir- sion is self-referential. The inverse transforma-
cumference of a semi-circle, 2/, would then cor- tion, between the nal and initial positions, is:
respond to the bulk shortening of the rock mass X X (x, y) ax by
containing it. The square of the reciprocal of it, (5.4)
Y Y(x, y) cx dy
2.46, might then correspond to an oid axial ratio.
This value falls within the range of ratios given. The constants a, b, c, and d may be expressed in
Since bending implies extension at the upper terms of A, B, C, and D by solving (5.3) for X and Y
surface of the layer and contraction at the lower and comparing coefcients. We obtain:
surface, and such a distribution is not indicated
a D(AD BC)
by the data, this model for the folding is not con-
b B(AD BC)
sistent with the deformation of the olites. (5.5)
c C(AD BC)
The assumption that the oids maintain con-
d A(AD BC)
stant volume allowed Cloos to determine that
material lines normal to the plane of section kept Likewise, we nd:
their initial length (Cloos, 1947, 1971). The area of
A d(ad bc)
an ellipse is ab, where a and b are the lengths of
B b(ad bc)
the semi-axes, and (4/3)abc is the volume of an (5.6)
C c(ad bc)
ellipsoid, where c is the length of the plane-
D a(ad bc)
normal semi-axis. Scale the dimensions so that
abc a30 1, where a0 1 is the radius of the initial The particle at the origin of coordinates stays
sphere. Then, if b 1/a, c 1, and a material line in there.
the normal direction has not changed its length. In a homogeneous transformation, a straight
material line is transformed into another one
5.1.3 Linear transformations with different position, length, and orientation. A
Consider the positions of particles, such as those unit circle in the undeformed state, X 2Y 2 1, is
that might lie in the bedding plane occupied by transformed into an ellipse. We used this result,
with further assumptions, to draw a strain ellipse (a)

associated with the deformed belemnite. It 36o 00'
CALIFORNIA
should be clear, from the four constants required
to obtain the full transformation, that the data
afforded by this example is not sufcient to deter- 0 10 mm a 1
mine these and, hence, the initial positions of all Velocity
particles that now lie on the strain ellipse. Later,
we return to such transformations to study their San
35o 30' Andreas
properties in a more formal way.
Fault
5.2 Evolving geometry of a PACIFIC 0 20 km

OCEAN Distance
structure: kinematic models,
121o 00' 120o 30' 120o 00'
velocity models, and
deformation Fig 5.5 (a) Velocity vectors from geodetic measurements
near the San Andreas Fault in central California (Harris and
Segall, 1987).
5.2.1 Geometric and kinematic models
We now consider forward models for the evolu-
tion of a structure in a deforming rock mass, such geometric objects such as curved surfaces, trian-
as the nappe structures in Fig. 5.2 or the folds in gles, or polyhedra. Without the signicance
Figs. 5.3 or 5.4. We look at what sequence of steps attached to these objects, there would be little
this might involve, while continuing to avoid dis- reason to painstakingly work out their forms and
cussion of a complete physical description of the disposition in space. Only the nal or current state
process. Instead, we focus on geometric and kine- of the body is described. We may imagine, or be
matic models, which include descriptions of the able to reconstruct, an initial state.
initial and current positions of particles, and of A hypothetical picture of the continuous
the motion of particles between the two states. change in geometry between initial and nal
A rst step in the study of a deformed or deform- states might be conceived. Such a description
ing structure is the geometric description of its would be termed a geometric model of the evolution
form and internal structure. Topography is impor- of the form of the structure, as described, for
tant in the description of an active mountain range example, by the changes in shape of the bed sur-
or volcanic edice because it both reects and faces in Fig. 5.4a. In a kinematic model, the motions
affects the dynamic process that created the struc- of all particles in the body between the initial and
ture and may continue to modify it. The descrip- nal states are described. If the observed nal state
tion of the internal structure would include the and the imagined initial states are viewed as snap-
congurations of surfaces such as layer or forma- shots, a kinematic model may be compared to a
tion boundaries and fault surfaces or fault zones, movie showing all intervening states. Carrying the
the position, attitude, and form of minor folds, and analogy a bit further, such a movie would not have
so forth. A description of the belemnite of Fig. 5.1 a sound track that described the dynamics or
would include details seen in the gure, but would causal elements of the process resulting in the
involve a more precise and complete description as motion and geometric evolution. To achieve a com-
in a detailed map of the belemnite fragments, the plete description of the process of formation of the
intervening veins, and the details of cleavage ori- structure, we must incorporate geometry and
entation in the surrounding rock. This is generally kinematics within a complete mechanical model.
termed a description of the structures geometry. It
is more than that, because special signicance is 5.2.2 Velocity fields
invariably attached to the features whose forms A description of the motion of particles, as in a
are documented, i.e. they are not merely abstract kinematic model, at one instant of time is given
5.2 KINEMATIC MODELS, VELOCITY MODELS, AND DEFORMATION 159
(b)
Trilateration station
displacement with 2s error elipse
Trilateration station held fixed
2 m displacement
HAWAII
19o 30'
0 10 km
19o 20'
PACIFIC
OCEAN
155o 30' 155o 20' 155o 10' 155o 00' 154o 50'
Fig 5.5 (cont.) (b) Horizontal displacement of monuments

at the surface of Kilauea Volcano, HI (Delaney et al., 1998). magma from a deep source, motion of magma
The displacements, up to several meters, are small relative to within the volcano, or other processes such as
the scale of the area. They represent motion over twenty faulting (Delaney, 1990; Delaney et al., 1993, 1998).
years, so the greatest average velocity magnitude is about Two lithospheric plates may move as nearly
20 cm a1. rigid bodies, the relative motion between them
being taken up by adjustment at the plate bound-
by the velocity eld. The velocity eld is expressed ary, generally over a zone of from tens to hundreds
in terms of the spatial distribution of the velocity of kilometers in width. In the case of the volcanic
vectors of particles currently occupying positions edice, the horizontal motions at the surface are
referred to a reference coordinate system. seen to vary in a more or less continuous manner
Studies in neotectonics often start with the obser- (Fig. 5.5b). To account for this, we may imagine a
vation of such a velocity eld. Precise repeated sur- continuous distribution of motion within the
veying of xed monuments at the Earths surface edice and its surroundings and this may lead to
allows one to determine the current velocity dis- a detailed mechanical model of the edice (Owen
tribution at Earths surface (Segall and Harris, et al., 1995). Motions within the edice are
1986). For example, the relative motion between inferred, while those at the surface are measured,
two lithospheric plates (Fig. 5.5a) is determined and serve to constrain the model. Rigid plate
from the relative changes in position of monu- motions constrain models for the internal
ments on either plate over some interval of time motions within the Earth and for the complex
(Harris and Segall, 1987; Murray et al., 2001; Murray local motions in the plate boundary zones. They
and Segall, 2002). The motions of monuments dis- constrain, but do not determine such motions. In
tributed over the surface of a volcanic edice (Fig. the geological literature, the word control is often
5.5b) may be measured to monitor the inux of used. The sense seems to be somewhere between
y, Y (x, y) observed, it is seen that the discontinuous motion

of particles along the surface is associated with a
continuous distribution of motion within the sur-
rounding rock.
In this chapter, we chiey consider motion and
y u uy deformation in two dimensions. This may be an
x
adequate rst approximation in structural geology
and tectonics, as indicated by the geometry of the
(X, Y )
resulting structures. A mountain belt may span
X thousands of kilometers along strike, but only a
Y ux
few hundred kilometers in width. The approxima-
0 x, X tion may work for the evolution of a single fold in
a portion of a mountain belt, such as that in Fig.
X 5.7, or for a set of folds in an outcrop. Use of verti-
x cal cross sections or of down-plunge sections (Fig.
5.3 and 5.4) to illustrate roughly two-dimensional
Fig 5.6 Initial and final position vectors, X and x, with structure is based on this approximation.
components (X, Y ) and (x, y). The displacement vector u has
Restriction to two dimensions results in
components (ux, uy) (x X, y Y ).
simpler presentation of concepts and ease in
obtaining results. Concepts are often more readily
constrain and determine, both of which, unlike visualized and grasped. Fewer quantities are
control, have distinct and separate meanings. involved two, rather than three, components of
The change in position of a monument over vectors, and four, rather than nine, components of
the time interval between two surveys is its dis- second-rank tensors. Relations and graphical con-
placement, the difference between the nal position structions can be carried out in two dimensions,
(during the nal survey) and the initial position so that, for example, true lengths and angles are
(during the rst survey), as measured relative to represented, and we need only plane geometry.
some convenient reference frame. The compo- Many results can be obtained in closed form or by
nents of displacement (Fig. 5.6) are: graphical construction.
On the other hand, to honor the geometry of
ux x X
(5.7) structures, which may be determined to great
uy y Y
accuracy and detail by modern surveying
Dividing the displacement by the time interval methods such as GPS, and to take the systematic
yields an estimate for velocity. It is perhaps not suit- study and interpretation of structures at all scales
able to do this if the motion was the result of a to a new level of renement, it will be necessary
sudden seismogenic faulting event. Displacement to abandon the approximation of two dimension-
and velocity are vectors. Since these quantities ality. In this book, we present and apply differen-
generally vary from monument to monument, tial geometry as one step in this direction.
and between them, they comprise vector elds. The
study of these elds and of other derivative elds 5.2.3 Deformation associated with the
is called: kinematics (1840): a branch of dynamics emplacement of an igneous pluton
that deals with aspects of motion apart from con- An array of large batholiths, 50 to 100 km in diam-
siderations of mass and force (Websters Ninth New eter, intrude greenstone (metamorphosed vol-
Collegiate Dictionary). canic and shallow intrusive rocks of intermediate
Discontinuous motion, as illustrated by plate to basic composition) and associated sediments
motion, is familiar from observations of faulting, (Fig. 5.8) in the Archean craton of Zimbabwe,
in which initially neighboring particles are dis- southern Africa. Ramsay carried out a structural
placed across the fault surface. When an entire investigation of one of the smaller and more sym-
fault surface, terminating within the rock mass is metrical structures, the Chindamora batholith
0 8 16 km
Fig 5.7 Map view of a segment of the South Mountain

Anticline showing that the distribution of strain is roughly country rock, or of previously solidied intrusive
independent of position along strike, suggesting the validity of units (cognate xenoliths). The distribution of xeno-
a two-dimensional approximation. Identical ellipses only lith shapes within the batholith is shown in Fig.
indicate the trend of the long axis of the strain ellipse; the 5.9b. More strongly attened xenoliths are found
axial ratio in the cross section is plotted by each. Axial ratios in the earlier emplaced rocks and nearer the
are comparable in strike-parallel belts. Reprinted from Cloos periphery. The exception, in both inward order of
(1971, Plate 1) with permission of The Johns Hopkins
intrusion and presence of strongly attened xeno-
University Press.
liths, is the last-intruded western granite.
The initial forms of the xenoliths would have
(Ramsay, 1989). The batholith is zoned (Fig. 5.9a) in been irregular, but roughly equant, so we take the
the order of intrusion, with the earliest intruded mean form as a sphere. The nal form for a given
unit at its periphery. Ramsay determined the local population of xenoliths averages out to a
mean shapes of populations of xenoliths (Greek: pancake-shaped ellipsoid with one short axis and
foreign stones), consisting of fragments of the two sub-equal long axes. Elliptical sections
N (a)
Chindamora batholith - rock types
Tonalite
Granodiorite
0 50 100 150 km Adamellite
Western adamellitic
granite
Greenstone belt
17o 30' S
0 10 km
Granitic batholiths
o
Greenstone belts 31 15' E
(b)
Finite strain
Fig 5.8 Array of batholiths in the Archean craton of
Zimbabwe. Reprinted from Ramsay (1989) with permission
from Elsevier.
1
showing the short and long axis are demonstrated 2
for each population. All xenolith axial lengths are
1
2
changed from their initial values.
Ramsay proposed an ingenious model to 2
explain the shape variation of the xenoliths in the 0 5 10 km
12
1
Chindamora batholith. The model addresses the
radial increase of deformation outward and suc-
1 strain elipse X/Y = X/Z
cessive emplacement of magma at the batholith 2
composite strains, 1 from greenstone
center. He supposed that the deformation of a xenoliths, 2 from tonalite xenoliths
xenolith begins when it is captured near the
solidication front in mostly crystallized material Fig 5.9 (a) Compositional zoning of the Chindamora
that is strong, but still hot and plastic. The xeno- batholith. (b) Distribution of representative elliptical sections
lith is then deformed with its host as further for local populations of xenoliths. Where two sections are
magma is injected into the center of the intru- shown, the one with greater eccentricity is for country rock
sion, inating it. He presents a quantitative model xenoliths embedded in xenoliths of the prior solidified
magmatic phase, that with smaller eccentricity is for the
of pluton ination, which he treats as an expand-
cognate xenoliths. Reprinted from Ramsay (1989) with
ing spherical or hemispherical body forcibly
permission from Elsevier.
deforming the surrounding greenstone.
Ination by means of magma supply from a
dike-like or columnar conduit does not seem a and the hemispherical ination model then appro-
plausible mechanism for the vast array of domical ximates the ow within the domes (Ramsay, 1989).
batholiths of the Zimbabwe greenstone belt. In an
alternative model, an entire layer underlying the 5.2.4 Inflation of a spherical shell: a
greenstone might have undergone melting or kinematic model
partial melting. Organized upwelling of domical Consider the ination of a spherical shell, or
masses of this layer might then have occurred (Fig. hollow sphere. Ination would be attributed to the
5.8). Ramsay makes a suggestion along these lines action of an internal pressure, but we are not
presently interested in the causal mechanism. By (a)

symmetry, the ination, although of a three-
du
dimensional object, is mathematically one dimen-
=2
r
sional, depending only on radius. The object is the 4
r1
volume contained between two concentric spheri- 2 .0
r=
cal surfaces. We may readily grasp the nature of the R

R
deformation from Fig. 5.10a. Focus on two bound-
=1
R
.2
ing spherical surfaces, with initial radii R and R1. 1=
1
Let R1 be the radius of the surface occupied by the
2 1 0 1 2
xenolith when it starts to deform and R be a nearby
radius in the solid shell. The shell is assumed to (b)
have the same volume after deformation, when its
inner and outer radii are r1 and r, respectively. Then:
3
4 4
(r 3 r 31) (R3 R31),
3 3
or r 3 r 31 R3 R31 (5.8)
2
Since we are concerned with the deformation
of the xenolith, a body of very small dimensions
relative to the pluton, we let R R1 R. Since R
R and r r, we can throw out terms in 1
which R or r are squared or cubed. For example:
r 3 r 31 (r1 r)3 r 31
r 31 3r 21r 3r1(r)2 (r)3 r31 0
2 1 0 1 2
3r21r (5.9)
With such approximation, substitution into (5.8) Fig 5.10 (a) Expanding sphere (or hemisphere) model for
pluton emplacement (Ramsay, 1989). (b) Elliptical principal
yields:
cross sections of initial spheres with ratio of final radius to
radius at incorporation.
r R2
r21r R21R,or 21 (5.10)
R r1
semi-axes, a and c:
Written in terms of innitesimal quantities, the
relation is exact: dr 2c c
(5.13)
dr R21 r1d 2a a
(5.11)
dR r21 Eliminating d between these:
To describe the deformation of the xenolith, c dr R1
(5.14)
consider the small elements shown in Fig. 5.10a. a dR r1
Since we have not described how to treat the Substitution from (5.11) gives the desired result:
deformation of a sphere to an ellipsoid we take a

1 3
c R1
somewhat rougher approach. Suppose the initial (5.15)
a r1
spherical xenolith is just enclosed by the element.
Then the ratio of the elements dimensions must This states that the current position of the xeno-
be unity, or: lith, r1, and its shape, given by the ratio c/a, deter-
mine the initial position, R1. Xenolith cross
dR
1 (5.12) sections for values r1/R1 1, 1.5, 2, 2.5, and 3 are
R1d shown in Fig. 5.10b.
In the nal state, we suppose the element just How may this description of a single xenolith
encloses an ellipsoid with long and short principal be used to support or refute the notion that the
7 (a)
15 8
6 1
9
Axial ratio, a/c
10 0.8
10 5
1 2 3 4 t 11 0.6
15 0.4
5
a 20 0.2
gd 25
wg
1 0
0 5 10 15 20 25 30
Final radius, r1 0.2
0.4
Fig 5.11 Plot of xenolith axial ratio, a/c, versus final radius,
r1, for the Chindamora pluton. Data symbols refer to t, 0.6
tonalite; gd, granodiorite; a, adamellite; wg, western granite. 0.8
Reprinted from Ramsay (1989) with permission from Elsevier.
1
mechanism of pluton emplacement is balloon- 1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1
ing? Ramsay plots the axial ratio, a/c, versus the
(b)
nal radius of the corresponding xenolith popu- 1
lation (Fig. 5.11). The points are scattered, but lie
0.8
in relatively localized regions. If the capture
radius, R1, is used as a parameter, each value of it 0.6
corresponds to a curve in r1, a/c-space. Only certain 0.4
of these curves will sweep through the regions for
0.2
each intrusive rock type. The range in values of R1
for these curves then corresponds to the position 0
of the partial solidication front at which the 0.2
xenolith is frozen into the expanding plastic shell. 0.4
The model of an inating pluton with spheri-
0.6
cal symmetry is a kinematic model because it pre-
scribes the motion of the particles within the 0.8
body from the initial to nal states. No informa- 1
tion is required beyond the spherical symmetry
during ination, and the concept of a variable 1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1
radius at which xenoliths are incorporated into
Fig 5.12 A buoyant viscous sphere rising in a viscous fluid
the deforming shell. Given the t between the (Lamb, 1945). (a) Velocity vectors. (b) Streamlines followed
strain data from the eld and the model relation by particles. Surfaces across which particles do not move,
(Fig. 5.11), the model is a viable one. Can you and within which the fluid is confined during the rise, are
propose further tests of it? toroidal, like the external surface of a donut; the traces of
several such surfaces are shown.
5.2.5 Internal deformation in a rising
spherical diaper country rock. The model has also been used to
A second model of pluton emplacement has been simulate the rise of hot, buoyant diapirs, or
popular for over fty years (Grout, 1945; White- mantle plumes, through the mantle (Anderson,
head and Luther, 1975). Earlier studies used labo- 1975; Whitehead and Luther, 1975; Ribe and
ratory models and the principles of model scaling Christensen, 1999).
(Hubbert, 1937) discussed in Chapter 4. The model This model is not simply a kinematic model,
consists of the rise of an approximately spherical as the explicit involvement of materials of well-
mass of viscous uid, representing the pluton, in characterized behavior and properties, i.e. viscos-
another viscous uid, representing hot, plastic ity and density, and , and the acceleration of
gravity, g, indicates. As we shall see, the velocity By symmetry, particles in the sphere move in
eld, although expressed by (5.19), is not easy to vertical planes through its center, so that it will
conceive of in an ad hoc manner. sufce to consider a description conned to one of
The remarkable solution of Stokes for a rigid these planes, the (x, y)-plane. Accordingly, the
sphere rising or sinking in a viscous uid was mathematical dependence of the solution is two
modied by others to treat the rise of a viscous dimensional. The velocity of a particle, referred to
sphere (Lamb, 1945). A sphere of viscous uid with the spatial coordinates, is independent of time.
density 1, intrusion viscosity 1, and radius R will The components of velocity, vx and vy , are:
rise in a viscous medium with density and vis-

2 2
cosity if the density difference, 1, is posi- r x xy
vx C 2 1 ,vy C (5.19)
tive. Experiment and theory show that the sphere R R r2
maintains its shape if the rate of rise is slow
where
enough (Fig. 5.12). We use the results to illustrate
the motion within the sphere, but specically to
determine the deformation within it and internal
r (x2 y2), C
1 ( 1)gR2
3 2 31
structures that might form between initial and
nal states, separated by an interval of rise. The velocity eld can be represented by the
The steady, or time-independent, rate of rise of stream function, a scalar quantity whose contours
the sphere is: are the streamlines (Fig. 5.12b):

2
r y
V
3
2 ( 1)gR2
( 1)
(2 31) (5.16) (x, y) C 1
R
r2 sin 2, sin
r
(5.20)
You are likely to be familiar with the special case Here is the angle in the section from the hori-
of a rigid sphere, 1 , sinking in a viscous uid, zontal. In the gure, contours of the stream
which may be used to estimate the rate of settling function at equal interval are plotted. The veloc-
of sediment particles in water: ity vector (Fig. 5.12a) is tangent to the stream-
lines and the speed, or magnitude of the velocity
V
9
2 ( 1)gR2
(5.17) vector, is inversely proportional to the contour
spacing. Thus, the maximum speed, relative to
The other limiting case applies, for example, to an origin at the center of the sphere, occurs at
the rise of a gas bubble, 1 0, in a viscous uid the surface of the sphere at its equator.
such as a basaltic magma: Because the ow within the sphere is steady in a
reference frame with origin at the sphere center,
V
3
1 ( 1)gR2
(5.18) particles remain on the streamlines as the
sphere rises. The steady internal motion is driven
The speed of rising or sinking only varies by a by the same steady rate of recovery of gravita-
factor of 3/2 in going from a rigid sphere (5.17) to tional potential energy that drives the rise of the
one that has low viscosity relative to its surround- sphere. The external ow is also steady if
ings, 1/ 1, (5.18). referred to the coordinate system rising with the
Position within the sphere is referred to sphere.
Cartesian coordinates that are xed at its center, To determine the nal, or current, position of
with coordinate x in the vertical direction of any particle in the sphere, for a given initial posi-
motion, and y and z in the horizontal plane. As the tion, we must follow it over the course of the inter-
ow continues, these positions will be occupied by val of rise using the relations (5.19). This is done
different particles. We may continue to think of the numerically, since, while the velocity at any posi-
spatial coordinates x, y, and z as giving the current tion is constant, particles move along paths
position of a particle which occupied an initial through the sphere along which the velocity
position with coordinates X, Y, and Z. We may not changes. If x(t; X, Y ) and y(t; X, Y ) are the coordi-
be able to specify the initial coordinates usefully. nates of the current position of a particle that
occupied position with coordinates X and Y at

time t 0: 1
x(0; X, Y ) X 0.5
(5.21)
y(0; X, Y ) Y
The position at a small increment of time t later 0

is:
0.5
x(t t; X, Y)
x(t; X, Y ) vx[x(t; X, Y ), y(t; X, Y)]t
(5.22) 1
y(t t; X, Y )
y(t; X, Y ) vy[x(t; X, Y ), y(t; X, Y)]t
1.5
For certain motions, possibly even in the present
case, this equation, or its equivalent, may be inte- 2
grated in closed form. Here, however, we assign t 1.5 1 0.5 0 0.5 1 1.5
a small but nite value, advance the particle to its
new position, and repeat this process many times Fig 5.13 A buoyant viscous sphere rising in a viscous fluid
(Lamb, 1945). Initial square grid of particles in vertical
to achieve the desired result.
diametral plane of sphere and deformed grid after a sphere
It is most convenient to refer the interval
rise of one radius.
between initial and nal states to the amount of
rise of the sphere, expressed in units of its radius,
R. The amount of rise is Vt, but from (5.16) and the three-dimensional surface in the deformed
(5.19): state is an ellipsoid. The sections look only
V 2( 1) approximately like ellipses because of their large
2 (5.23) size. The distribution of deformation, with more
C
attened forms outward and with long dimen-
The intrusion viscosity, 1, is much less than that sions concentrically oriented is similar to that of
of the country rock, . Time will be expressed in deformed xenoliths in the Chindamora pluton
units such that we may set C1 in (5.19). In unit (Fig. 5.9) or the model for it (Fig. 5.10b).
time, the sphere will rise by an amount equal to This result suggests that the rising sphere
its diameter, 2R. model might provide an alternative to the
Figure 5.13 shows the deformed state of a grid inating intrusion model. However, this would
whose intersection points correspond to the nal not seem to be supported by the roughly concen-
positions of a set of particles that initially lay at tric pattern of intrusive rock types in that pluton.
the intersection points of the square grid in the In this model, a concentric pattern can be pro-
gure after a rise of one sphere radius. The duced. Consider a set of equally spaced horizontal
deformed grid lines are constructed by connect- surfaces in the initial state (Fig. 5.15a). After a rise
ing grid points with straight segments. To a good equal to the sphere diameter, these surfaces are
approximation, the line segments in the nal deformed into those shown in Fig. 5.15b. The
state correspond to the material lines connecting entire surfaces are formed by spinning this section
grid points in the initial state, because the grid around the vertical axis. If the magma body had
elements are small enough. developed a layered compositional sequence prior
As another example, Fig. 5.14b shows the nal to its rise, then something resembling a concen-
positions and forms of an array of material circles tric distribution of magma types might arise. The
in the initial state (Fig. 5.14a) after rise of one process of multiple injection proposed by Ramsay
sphere diameter. The circles may be interpreted is supported by his observations. It is not clear,
as sections of initially spherical surfaces. however, that multiple injection occurs by intru-
Provided an initial circle is innitesimally small, sion into the center of an expanding spherical or
(a) (a)
1 1
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0.2 0.2
0.4 0.4
0.6 0.6
0.8 0.8
1 1
1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1 1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1
(b) (b)
1 1
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0.2 0.2
0.4 0.4
0.6 0.6
0.8 0.8
1 1
1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1 1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1
Fig 5.14 A buoyant viscous sphere rising in a viscous fluid Fig 5.15 A buoyant viscous sphere rising in a viscous fluid
(Lamb, 1945). (a) Initial array of circular material lines in (Lamb, 1945). (a) Initial traces of equally spaced horizontal
vertical diametral plane of sphere. (b) Deformed final state material planes. (b) Deformed traces after rise of sphere of
array after rise of one sphere diameter. one diameter.
hemispherical volume, and a scheme adapted to a Thus, the stretching of a thin layer at the margin
rising sub-spherical intrusion might be worked of the intrusion must be the same as that of a
out. In any case, the more elaborate model should thin layer in the adjacent wall rock. The thin-
follow Stokes example and use a complete ning shown by the deformed initial spheres near
mechanics. the surface must substantially under-estimate the
Thinning of the wall rock at the top of the thinning at the contact with the overlying uid
sphere can be determined from the Stokes model. medium, since a large fraction of this has been
A boundary condition was imposed in the model greatly thinned to make room for the material
that required the uids to stick at their interface. represented by the grid. To estimate the thinning
at the contact, we use a method similar to that deformed internal contacts, and the stretching of
used to nd the ratio of deformed ellipsoid axes in the surrounding country rock. If these features
the spherical shell model. conform to the predictions of the model, they
Consider a pill-shaped element of material serve to support it. If they do not, a markedly dif-
whose upper surface coincides with the surface of ferent model must then be formulated. Either
the sphere at its apex. The element is taken to be outcome would advance our knowledge of the
sufciently small that the surface of the intrusion process of pluton emplacement.
is approximated by the horizontal tangent plane.
In the initial state, before any rise of the sphere,
the radius of the pill is Y and its thickness is X. 5.3 Relation between deformation
The volume of the pill is V (Y )2X. After some
amount of rise, in the nal state, the radius and
and velocity fields
thickness of the pill are y and x, and its volume
is v (y)2x. Since the volumes are equal, y/Y 5.3.1 Chevron folds
(x/X)1/2 . The nal form of the pill may be The folds, seen in cross section in Fig. 5.16, are
found by tracking the position of the particle with called chevron folds because of their straight limbs
initial position (R, Y ) on the tangent plane, or and narrow, sharp hinges (Ryan and Smith, 1998).
with initial position (R X, 0) on the vertical axis A chevron is composed of two stripes that meet
of the sphere. For the former, the velocity away at a sharp angle with the apex generally up, as
from the center point is: seen in insignia of rank on military uniforms.
Prominent quartz veins, in the form of saddle
d Cy d(y) C
vy(R, y) (y) , or dt (5.24) reefs occur at the crests of the folds. Many more
dt R y R
veins are present, including those along the limbs
Integration from t0, and between the limits Y and parallel to bedding, and in fault zones whose
and y corresponding to the initial and nal posi- location appears to be controlled in part by the
tions of the particle at the end of the pill diame- prior presence of the fold structure. Important,
ter, yields: also, is the remarkable dike that is emplaced
along the hinge surface of the anticline. The
ln

y
Y

C
R
y
t*, so exp
Y
C
R
t*
(5.25) quartz veins contain high-grade gold ore, so that
the process that formed them and determined
their distribution is of much commercial interest.
Here t* is the time of rise yielding the desired hor-
Here we consider how the folds might have
izontal stretching of the pill-shaped element. A
formed.
reduction in pill thickness by a factor of 0.1 cor-
Evidence for slip between layers in the form of
responds to an increase in pill radius of
slickenlines is found in association with the bed-
10 3.2, or:
parallel quartz veins. Slipping layers generally

C consist of many individual beds and are 10 m in
t* ln 10 1.15 (5.26)
R thickness. The length of the fold limbs in the
example from Bendigo, Australia (Ryan and
The distance of rise of the sphere is:
Smith, 1998), is about 300 to 400 m or more, so
R that the ratio of the thickness of a slip-surface

x* Vt* V 1.15 (5.27) bounded layer to limb length is less than 1/100.
C
The representative bedding traces shown in the
Using (5.23), the amount of rise of the sphere is x cross sections do not show all interfaces on which
2.3R, or slightly more that its entire diameter. slip has occurred; they are accurately drawn from
The Stokes model for the rise of a sphere in a data collected in the mine. Not all chevron folds
ductily deforming (viscous) medium thus pro- show evidence for slip (Fig. 5.17), but it is com-
vides a rich source of potential interpretations of monly enough observed in this fold type to be
features of a pluton, such as included xenoliths, viewed as characteristic.
5.3 RELATION BETWEEN DEFORMATION AND VELOCITY FIELDS 169
CENTRAL VICTORIA NOVA SCOTIA

BENDIGO (Great Extended Hustlers) GOLDENVILLE
100 m
NOVA SCOTIA
GOLDENVILLE
Discovery 1861
Grade 15g/t
Total Production 6.7 million grams
Au to 300 m 6.5 million grams
Depth 325 m (unknown 0 cm 3
below this depth)
CENTRAL VICTORIA
BENDIGO (Great Extended Hustlers) Fig 5.17 Chevron folds in phyllitearenite layers, Poudre
Discovery 1851 Canyon, Colorado. Photograph by R. C. Fletcher.
Grade 15g l1
Total Production 32 million grams
Au to 300 m 0.5 million grams
Depth 1200 m+
The presence of these and other features

Workings
results in a tendency for the rock to split more
Fault easily along surfaces parallel to the cleavage
Dyke hence, the name. Cleavage is inferred to lie
Quartz vein normal to the direction of maximum shortening
Bedding
in the rock mass. In many cases, this inference is
Extent of mining in
most Meguma Deposits supported by direct evidence in the form of bodies
of known initial shape now attened in the plane
of cleavage; an example is (iii).
Fig 5.16 Chevron folds as seen in mine sections from Chevron folds are common in sequences of
Bendigo, Victoria, Australia, and Goldenville, Nova Scotia. strata made up of thick stiff or strong sand-
Reprinted from Ryan and Smith (1998) with permission from stone layers separated by thin soft or weak
Elsevier.
shale layers, or their metamorphosed equivalents.
In our discussion of postulated or actual rock
A second feature observed in the Bendigo fold properties, pairs of terms like stiff and soft, or
is cleavage (Ryan and Smith, 1998). Cleavage is strong and weak, are relative, and, often, as here,
present in both the stiff graywacke sandstone do not refer to an explicit type of material behav-
layers and the interbedded softer metamorphosed ior, such as elasticity or viscous ow. Later, we do
pelitic rocks. Briey, cleavage is a ne-scaled, introduce this specicity and quantitative preci-
structure often called a fabric that is pervasive, sion. Also, with regard to dimensions of rock
with features such as: (i) a strong planar align- bodies, we will often use the terms large and
ment of platy mineral grains, such as micas; (ii) small or thin and thick in the same relative
small-scale crinkles, which may themselves sense. That is, a dimension of the object under dis-
have the aspect of chevron folds, formed in rocks cussion, such as layer thickness, has two or more
possessing a prior strong orientation of the sort values, among which relations such as layer A is
just mentioned, which may rst form in the com- thick (thin, sub-equal) relative to layer B may be
paction of a clay-rich sediment (shale); or (iii) stated. They are also seen in folding at a much
planar alignment of grains whose attened shape smaller scale (Fig. 5.17). The chevron fold form is
is due to the deformation of the rock itself; or (iv) so common and striking, apart from its associa-
discrete, sub-parallel surfaces of dissolution, tion with gold deposits, that structural geologists
often indicated by build up of insoluble residue have been inspired to develop models for the
such as clay in limestone, referred to as solution folding process and fold evolution from an initial
seams. state.
Because a structural geologist can observe the

nal structure, and can infer the initial form of
the rock mass involved, kinematic models are
popular. A suite of structures may be available
from which stages in the development of the
structure may be inferred. A kinematic model is
formulated in three steps.
1. The geometry of the component elements of
the structure, here the individual layers
between slip surfaces, is idealized, providing a
geometric model.
2. From the geometric model, a procedure for
approximating the form of the structure
throughout its evolution is devised: an evolu-
tionary model.
3. Finally, consistent with the evolutionary
model, a set of particle motions is devised to
provide the motion and deformation of all indi-
vidual rock elements within the structure. This Fig 5.18 De Sitter (1964) model for chevron folding. The
is the kinematic model proper. small circle on the right limb is the location of an imaginary
pin that acts as the center of rotation for this layer.
5.3.2 Models for chevron folds

Most characteristic of chevron folds are long all earlier, lower limb dip stages. This reects a
straight limbs in relation to narrow, tight hinges. lack of any better understanding of how such a
An explanation for both features is required, form might arise. The requirement of original
though one might expect that the explanation for seed folds present in the layered material is
one might provide that for the other complemen- physically unrealistic, but in the evolutionary
tary feature. What we mean by explanation in this model, they are taken to be present.
case is detailed answers to questions such as: We examine two kinematic models for
How do these features come about? What condi- chevron folds that are based on different mecha-
tions and circumstances are involved? What must nisms of folding. In a model rst proposed by the
be the mechanical behavior of the rock layers and structural geologist L. U. De Sitter (1964), the fold
their interfaces? hinges are assumed to be broken, so the folding
While natural chevron folds have variable mechanism consists solely of the sliding and rota-
layer thickness and form of individual limbs and tion of the rigid layers in each fold limb. A set of
hinges, those of Fig. 5.16 and 5.17 show regularity folds of very low limb dip, with broken hinges and
in layer thickness. Those in Fig. 5.16 also show a straight limbs, is postulated as an initial state. A
remarkable persistence in form in the vertical model fold produced in this manner is shown in
direction. In the geometric models for chevron Fig. 5.18. Following the geometric idealization, all
folds that we consider, these two regularities are layers have equal thickness. The fold is also taken
idealized: (i) fold limbs continue vertically to be symmetrical across the axial surface. The
without change in horizontal span or limb dip; kinematics of a single fold limb of this type is also
and (ii) all layers making up a fold limb have the sometimes referred to as that of a bookshelf model,
same thickness. A third idealization is that the because it is similar to what ideally happens when
fold limbs are straight and the hinge regions are support is taken away from a set of slightly tilted
narrow. In the evolutionary models for chevron books on a at surface, and the books slide in
folding considered here, this idealization, present concert. In this version of the model, which
in the folds observed in nature, is carried back to differs somewhat from the original De Sitter
model, the gaps opened up at the broken hinges (a)

are exactly compensated by interpenetrations
between layers. If, instead of physically impossible
interpenetration, the layer ends are dissolved and
the material deposited in the gaps, both processes
allowed by the presence of pore water and the
slowness of the folding process, structures some-
what like the saddle reefs seen in the fold in Fig.
5.16 are produced. De Sitters aim in breaking
the layers was to produce gaps of this sort.
However, he backed the layers away, so that they
only touched at single points, there was no inter-
penetration, and the gaps were larger.
In a second model, no slip between layers in
the fold limbs occurs. An initial seed fold is
attened homogeneously to produce a fold with
larger limb dips. An example of such attening is
(b)
shown in Fig. 5.19b, in which the seed fold is that
in Fig. 5.19a. The shapes of some of the folds in Fig.
5.19b approximate those of the individual layers
in the chevron fold in Fig. 5.16. Note, however, that
these folds are not developed in a stack of layers.
In this model, continuity of the layers through the
hinges as seen in Fig. 5.16 is maintained. In Fig.
5.20, the model is used to tighten an initial
chevron fold with lower limb dip.
We now analyze the particle motions in the
two models. The rst model accounts for the pres-
ence of slip surfaces seen in many, but not all,
chevron folds. Since the layers are rigid, it does
not account for the presence of cleavage in them.
Conversely, the second model accounts for the
presence of cleavage, and leads to an orientation
of cleavage parallel to the axial plane of the fold,
but it does not account for inter-layer slip. We may
combine the two mechanisms to provide a kine-
matic model for chevron folds in which both
observed features are present.
5.3.3 De Sitter model: rotation of and

slip of rigid layers
The De Sitter model for chevron folding (De Sitter,
1964) is remarkably simple but at the expense of
whatever complication must really go on in the Fig 5.19 (a) Folded fibrous calcite veins in limey shale;
region of the fold hinges! This complication can sample width 2 cm. (b) The image in (a) has been flattened
be dealt with in several ways, one of which has in the horizontal direction by a factor of 1/2, and extended in
been described. The fold limbs rotate as the span a vertical direction by a factor of 2. This illustrates the
folding mechanism of the second kinematic model.
of the limb, S, normal to the axial plane of the
fold, is reduced (Fig. 5.21a). Only a single limb is
(a) y
H
d
L
(b) d
Step 1 Step 0
u
Fig 5.20 Chevron fold obtained by flattening a chevron

fold of smaller limb dip; the amount of flattening is the same
as that in Fig. 5.19b.
Step 2 d
shown here. Limb length, L, does not change; the

Fig 5.21 (a) Rotation of stack of layers of length, L, and
layer thickness is H. The layer centered at the thickness, H, through angle, . (b) Two-step procedure for
origin of coordinates is supposed to be pinned at forming a chevron fold limb by the De Sitter model (De
this position. This artice is necessary to provide Sitter, 1964).
a reference point from which the relative motions
of other particles in the fold limb can be deter-
mined. The pinned particle stays where it is S cos
during folding. This denition of a reference (5.29)
S0 cos 0
point or coordinate origin is generally necessary
in thinking about motion and deformation. Since the absolute scale of the structure has no
At limb dip : signicance this measure seems a good choice.
S/S0 is a measure of bulk strain, since it gives the
S L cos (5.28)
reduction in the horizontal dimension of the fold
A measure of the progress of folding might be the limb, and its inverse, the increase of an initial
limb dip itself, or the shortening of the span of the segment of the limb in its vertical dimension. But
limb, S. The relation (5.28) yields an initial span S0 S/S0 does not conform to a proper denition of
for the initial value 0. With folding, S decreases strain in terms of initial and nal lengths of a
from its initial value S0, and a dimensionless material line, because S and S0 are not the nal and
measure of the change is: initial lengths of the same material element.
Since the layers are rigid in the model, a material y

line element within an individual layer does not
change its length and no such element undergoes
(x, y)
strain. It will be useful to formulate a concept of
bulk strain for the fold limb as a whole.
A rock mass cut by numerous faults may
undergo a bulk strain for a set of displacements on
r =R
r sin u
the faults. The present model affords a simple (X, Y )
example of such a situation, the faults being the u R
layer-parallel or bed-parallel slip surfaces. Since R sin
rock bodies may deform by the relative motion of
x
many approximately rigid elements, down to indi- R cos
vidual grains or grain fragments, this concept of
bulk strain has many applications. In the present r cosu
case, the slip across each interface is:
Fig 5.22 Rigid-body rotation about the origin through the
u H( tan tan 0) (5.30) positive, anticlockwise angle .
Here 0 is the initial dip of the seed fold. The

negative sign is adopted because the fold limb of layer with the Y-coordinate of its center point. For
Fig. 5.21a may be produced from the unfolded the nth layer upward, this would be nH, and down-
conguration by rst sliding each layer above the ward, nH. To avoid such complexity, at least ini-
pinned layer at the origin to the left by this tially, let us agree to nd the functions (5.31) that
amount and then rotating the whole limb by . apply only to the particles located on the center-
The physics convention is that a positive rotation lines of the layers. For these particles, the func-
is counterclockwise (Fig. 5.21b). tions may be written as though the particle
We now seek a description of the two-dimen- displacements were continuous. We thus obtain a
sional deformation that corresponds to the De description of the bulk deformation of the fold
Sitter chevron fold model: limb.
We write down continuous functions that
x x(X, Y )
(5.31) capture exactly the initial and nal positions of
y y(X, Y )
the particles on the mid-planes. These are most
The description (5.31) will be developed using the easily found by consulting Fig. 5.21b and combin-
xed coordinate axes of Fig. 5.21 to which initial ing the results for the two steps. Coordinates of
and nal coordinates will be referred. Since the the intermediate positions are:
layers rotate about axes parallel to z and particles
move in cross-sectional planes of constant z we
have z Z. The particle at the origin remains
x (1) X

u
H
Y X ( tan )Y
(5.33)
y (1) Y
there:
A positive anticlockwise rotation about an axis
x(0, 0) 0 through the origin by an angle results in new
(5.32)
y(0, 0) 0 particle coordinates (Fig. 5.22):
If (5.31) is to be specied for all particles in the x X cos Y sin
body for the rigid layer model, called the De Sitter (5.34)
y X sin Y cos
model, they must be specied for each and every
layer, since the slip between layers means that Recall the expression of x and y in terms of the
these functions cannot be continuous functions angle between the positive x-axis and the radius
of initial position. This could be done in a compact r or the distance between the origin and the posi-
form by identifying the pair of functions for each tion (x, y):
x r cos , y r sin
(5.35)
X R cos , Y R sin
But, from Fig. 5.22:

, r R so
x R cos ( ) L0
(5.36)
cos (R cos ) sin (R sin )
X cos Y sin
The expression for y is similarly derived. To carry S0

out the second step, we substitute for the initial
coordinates X and Y in (5.34) the intermediate
coordinates x(1) and y (1) of (5.33) for the rotation
, yielding:
x (X Y tan ) cos Y sin X cos

L
y (X Y tan ) sin Y cos (5.37)
X sin Y
1
cos
These continuous functions give the nal posi-

tions of mid-plane particles but not the positions
of particles off the mid-planes (Fig. 5.23). A mater- S
ial line of length L0 is set out in the initial state. In
Fig 5.23 Continuous approximation to the deformation of
the nal state, this has been cut up into a series of
the De Sitter model (De Sitter, 1964).
segments that are inclined at the same angle to
the layer surfaces and whose aggregate length is
L0. The continuous medium approximation to the deforms any material line in the body to another
deformed line is the straight line through the material line with, in general, different length
mid-points of length L. The ratio L/L0 is a measure and orientation. The deformations (5.33) and
of the strain of this line. On the other hand, the (5.34) combined to achieve the nal deformation
measure of bulk shortening S/S0 cannot be are also homogeneous.
identied as the deformation of a material line in The simplicity of the De Sitter model is such
this manner. that the change in fold shape as folding contin-
The relations (5.37) may be written: ues, or the evolutionary model, and the motions
x Fxx X FxyY involved, or the kinematic model, are immedi-
ately apparent, although we still have to write
y Fyx X FyyY
these down in explicit form. However, a descrip-
(5.38) tion of the deformation involves only the initial
Fxx cos , Fxy 0
and a current or nal state, and not the path
1
Fyx sin , Fyy between them. Before completing the kinematic
cos
model, we study a second model for chevron
This homogeneous linear transformation describes folding.
the homogeneous deformation of the fold limb.
Homogeneous refers to the fact that Fxx, Fyx, . . . , are 5.3.4 Homogeneous flattening model
independent of X and Y, and linear to the fact that The presence of cleavage in the Bendigo chevron
the initial coordinates enter linearly. As is shown folds (Fig. 5.16) means that they could not have
by the example, a homogeneous deformation formed solely by the rotation of rigid layers in the
fold limbs with localized bending or breaking of

the layers in the narrow hinges. Accordingly, we
introduce another model for fold formation in
which layers do not behave as rigid bodies. We
begin with the limiting case in which inter-layer
slip is excluded.
Start with the conguration shown in Fig.
5.20, which corresponds with perfect chevron
folds of low limb dip. As above, consider a func-
tional relationship that describes the nal posi-
tion of any particle in the body in terms of its
initial position. We choose the simple relation-
ship:
x FxxX
Fxx = 1 Fxx = 0.5

1 (5.39)
y FyyY Y
Fxx Fig 5.24 Initial chevrons (Fxx 1) and flattened chevrons
(Fxx 0.5). Circular markers become strain ellipses.
Here Fxx and Fyy are constants. The initial and nal
fold forms are two-dimensional cylindrical forms
with generator or axis parallel to the z-axis.
First consider what happens to a rectangular We may compute a relationship between the
element of area in the (x, y)-plane whose initial initial and nal dips, 0 and , or the ratio of the
corners are the points (0, 0), (0, X), (X, Y ), and (0, Y ). principal axes of the strain ellipse from the trans-
Its area is XY. In the deformed state, the element formation (5.39). Without imposing the restric-
remains a rectangle, with corners at (0, 0), (0,x), (x, tion of constant area:
y), and (0, y) and area:
xy (Fxx X)(FyyY) (Fxx Fyy) XY XY (5.40)

tan
Fyy
Fxx
tan 0
b Fyy S
Thus, by choosing Fyy 1/Fxx in (5.39) we prescribe , Fxx (5.41)
a Fxx S0
that the cross-sectional area of the body, or any
part of it, remains the same. Referring to a broad Here b and a are the vertical and horizontal semi-
range of structure cross sections, structural geol- axes of the strain ellipse. In contrast to the ratio
ogists refer to this circumstance as one in which S/S0 obtained in the rotating rigid layer model,
area is conserved. Keep in mind that conservation of this quantity does describe a strain: here the
area is NOT a law of nature, such as conservation of change in length of a horizontal material line
mass. Does the De Sitter model conserve area? between the initial and nal states.
Application of the transformation (5.39) with The type of folding, or the fold mechanism, of
Fxx 1, to the conguration in Fig. 5.20, results in the homogeneous attening model is often
a tighter fold. An example is shown for Fxx 0.5. termed passive folding (Donath and Parker, 1974),
Fig. 5.24 shows a model of this type for Fxx 0.5, since it corresponds to a situation in which the
which starts with a chevron seed fold with a limb mechanical properties of layers or interfaces play
dip of 15. Several circles in the initial congura- no active role. It is as though the layers were com-
tion are deformed into ellipses. By our hypothesis posed of materials with the same isotropic
for the relationship between deformation and mechanical properties, which allowed for a stiff
cleavage, the model cleavage is oriented vertically, uid-like behavior, and surfaces of easy slip were
normal to the short axis of the ellipse. Application not present. One may also apply this model to dif-
of (5.39) requires a computation for individual ferent seed fold forms, such as that dened by a
lines, circles, and other loci. set of sinusoidal surfaces (Fig. 5.25). Since the

(a) 1 (5.42)
(b) y (1) X sin (1) Y
cos (1)
Here, we have assumed that the initial dip is van-

ishingly small. To obtain the nal form, the trans-
formation (5.39) is applied to (5.42):
Fxx =1, Fxy =0
Fyx =0, Fyy =1/Fxx x Fxx
(2)x(1)

1 (1) (5.43)
y y
(c) Fxx =0.5, Fxy =0 F (2)
xx
Fxy =0, Fyy =1/Fxx
Substituting from (5.42) we have:
Fxx = 0.5, Fxy =1 x (Fxx

(2) cos (1)) X

Fyx = 0, Fyy =1/Fxx 1 1 (5.44)
y X sin (1) Y
Fxx
(2) cos (1)
Fig 5.25 Folds with sinusoidal initial and final surface forms
produced by passive folding: (a) the seed fold, (b) flattening, Notice that the forms of the coefcients are:
(c) a combination of flattening and shearing. The coefficients
defining the deformations are given.
x Fxx X FxyY
(5.45)
y Fyx X FyyY
In these equations the coefcients are:

variety is endless we should begin to think about
what folds in rock actually look like and how they Fxx Fxx
(2)F (1) F (2)F (1)
xx xy yx
might have formed as constrained by the laws of Fxy Fxx

(2)F (1) F (2)F (1)
xy xy yy
physics and explicitly dened material properties. (5.46)
These considerations will be taken up in later Fyx Fyx
(2)F (1) F (2)F (1)
xx yy yx
chapters for now we are limited to geometry and Fyy Fyx(2)Fxy

(1) F (2)F (1)
yy yy
kinematics.
(1)
The quantities Fxx ,. . . , are given in (5.42). Here,
5.3.5 Inter-layer slip and homogeneous and throughout this section, we describe the De
flattening Sitter model in terms of its continuous medium
Since both the rigid layer slip model and the approximation. We can appreciate that the alge-
homogeneous deformation models of geometry braic complexity of such a description increases
and kinematics are quite simple, we may combine rapidly, although the operation itself is straight-
them to obtain a chevron fold model that shows forward. Imagine applying each of the two mech-
both characteristic features: inter-layer slip and anisms alternately over many steps!
cleavage. One simple procedure for combining the The nal dip is:
models is to rst apply one model, or mechanism, 2
to achieve a fold form from an initial congura-
tion, and then apply the other mechanism to the
tan 1
Fxx
(2)
tan (1)
b
a
tan (1) (5.47)
structure produced to achieve the nal fold. This

The nal limb dip may be measured, and if
mirrors the two-part procedure of rst slipping
strain markers allowed b/a to be determined, (5.47)
the layers in the fold limb, and then subjecting
provides an estimate of the limb dip at which
the whole to a rigid-body rotation, or vice versa
layer-parallel slip ceased according to the model!
(Fig. 5.21b). For example, beginning with the De
Because the circular sections are simply
Sitter model, the fold is obtained by rst applying
rotated during the layer-slip stage, and no slip
the transformation:
occurs during the second stage, cleavage in the
x(1) X cos (1) fold will be vertical. If we observe a chevron fold
5.4 VELOCITY FIELDS: THE INSTANTANEOUS STATE OF MOTION 177
with such vertical cleavage, and with evidence for nal state, but of the continuous progression of
inter-layer slip, we might then conclude that our states between them. The term progressive deforma-
model is appropriate. If the cleavage were not ver- tion is often used by structural geologists to denote
tical except within the hinge region where sym- such a progression, whether or not it can be
metry requires it another kinematic model must described in detail (Ramsay and Huber, 1983, 1987).
be concocted. We do not claim that geometric, That is, any rock mass may be viewed as having
evolutionary, and kinematic models are the nal undergone a progressive deformation, the end
goal of our study, but they do provide a means of result of which is the suite of structures and
organizing observations and sorting out hypothe- deformed objects that we see in the exposure. On
ses as to folding mechanism. the other hand, the term progressive deformation
does not appear in the literature or texts of contin-
uum mechanics, which also deal with the defor-
5.4 Velocity fields: the mation of materials, including rocks. This term
seems to be the special invention of structural
instantaneous state of motion geologists. In the context of our discussion, the
description of a progressive deformation would
We have described the production of the limb of a appear to consist of a specication of the four
chevron fold in terms of the homogeneous defor- coefcients Fxx(t), Fxy(t), Fyx(t), and Fyy(t) as functions
mation, as described by the transformation from of time.
initial to nal coordinates of the particles in the Consider the rate of change in the positions of
body through relations of the form (5.45). The particles in the body. This description refers
quantities Fxx, Fxy, Fyx, and Fyy are the components explicitly to the physical variable time t, and thus
of a second-order tensor, the deformation gradi- leads toward a consideration of the physical
ent tensor. In the models so far developed, the processes responsible for folding. Taking the
coefcients have been obtained for two simple derivatives of the expressions in (5.45) with
kinds of deformation. Then, we obtained the respect to time:
coefcients for a sequence of two deformations in
dx dFxx dFxy
succession, one of each kind. This was consistent X Y
dt dt dt
with two observed features of chevron folds: inter- (5.48)
dy dFyx dFyy
layer slip and cleavage. However, this composite X Y
dt dt dt
model seems articial. It suggests a process in
which slip surfaces are initiated between rigid Since X and Y are the initial coordinates of the par-
layers and a fold forms in this manner up to a ticle, we do not operate on these. But the rates of
certain dip, at which point a completely different change of the current coordinates of the particles
mechanism of folding sets in. The model describes are their velocity components, or:
neither the initiation of slip nor its cessation. The
opposite sequence might also be considered. This dx dy
vx, vy (5.49)
composite process is plausible if the two mecha- dt dt
nisms operated during episodes separated in
The velocity eld at any time will depend on the
time, in which the conditions were markedly dif-
current conditions, properties, and forces
ferent. For example, lower temperature and pres-
applied to the body. Such a description is gener-
sure and the presence of uids might have been
ally developed in terms of position and time in
associated with an episode of inter-layer slip and
the body, so that in the two-dimensional case
higher temperature and pressure with an episode
under consideration:
of homogeneous deformation.
Since a parameter such as changes continu- vx vx(x, y, t, material properties, applied forces)
ously, the description obtained might be inter- vy vy(x, y, t, material properties, applied forces)
preted not only as the description of an initial and (5.50)
Here, x and y are used to describe the current coor- Lxx(Fxx X FxyY) Lxy(Fyx X FyyY)
dinates of some particle in the body, but they are
also used to describe position in space. The use of Collecting terms in X and Y and carrying out the
these quantities to describe current position of same operation using vy :
particular particles and position within the body dFxx
creates no problem. For example, the velocity at a Lxx Fxx Lxy Fyx
dt
particular position is an attribute of the particle dFxy
currently occupying this position. The mathemat- Lxx Fxy Lxy Fyy
dt
ical description of the variation in velocity with dFyx (5.53)
position may be smoothly varying. While we may Lyx Fxx Lyy Fyx
dt
formally evaluate the function elsewhere than dFyy
within the body, e.g. within a hole in the body, it Lyx Fxy Lyy Fyy
dt
is clear that the values of quantities there have no
physical signicance. To illustrate how we think about velocity elds of
The velocity eld within the body at a particu- the simple type (5.51) and how we incorporate
lar instant of time will depend upon the physical them into (5.53) and solve for the state of defor-
state of the body, its properties, and the forces mation, we go back to our two chevron fold
applied to it. Thus, rather than the current state models. We show how we can combine the two
of deformation, with reference to some initial models into one that allows the mechanisms of
conguration, it is this that we must look to if we attening and inter-layer slip and layer rotation to
want to understand the process producing the go on simultaneously. The key concept is that the
structure of interest. We will examine this rela- separate velocity elds for the two mechanisms at
tionship in this textbook. An example given in any instant are additive.
this chapter is that of the Stokes solution. Initial and nal positions of particles for
However, we also want to know how the history of attening are related by:
the instantaneous state of motion, or the velocity x Fxx X,y FyyY (5.54)
eld, gives rise to the current or nal structure;
and within the limited context of kinematics that The set of differential equations (5.53) is incom-
is something we may accomplish in this chapter. plete without a set of initial conditions. Initially,
In the present case of a homogeneous defor- the coordinates of particle position x and y are just
mation, as in a single limb of an idealized chevron equal to X and Y, so:
fold, the deformation is given by the coefcients
Fxx(0) 1,Fxy(0) 0,
Fxx, Fxy, Fyx, and Fyy. Moreover, the velocity eld (5.55)
Fyx(0) 0,Fyy(0) 1
within the fold limb where we again consider
the smooth equivalent of the rigid layer model
To have (5.54) at any time during the folding
can be written:
process, we must require that the coefcients Fxy
vx Lxxx Lxy y and Fyx are always zero. Examination of (5.53) indi-
(5.51)
vy Lyxx Lyy y cates that the quantities Lxy and Lyx must always be
zero, and (5.53) reduces to:
Here the coefcients are uniform in the limb.
Here, x and y denote both the spatial coordinates dFxx dFyy
Lxx Fxx, L yy Fyy (5.56)
and the current positions of particles. Combining dt dt
(5.45) and (5.48) through (5.51), we obtain relations
The equivalent velocity eld is:
for the rates of change of the components Fxx,. . . :
vx Lxx x,vy Lyy y (5.57)
vx
dx
dt

dFxx
dt
X
dFxy
dt
Y
To integrate (5.56) with the initial conditions
Lxxx Lxy y (5.52) (5.55), we need to know how Lxx and Lyy vary with
time. In the simplest case they are constant.

Integration then yields:
Fxx(t) exp(Lxxt)
(5.58)
Fyy(t) exp(Lyyt)
The condition of constant area requires Fyy(t)

1/Fxx(t), or, from (5.58):
Lyy Lxx (5.59)
Indeed, constant area requires (5.59) whatever the

time variation of these quantities. The time frame
during which structures form is an interesting
topic brought up by considerations of velocity. It
is signicant to questions such as that of the con-
nection between folding and gold mineralization
at Bendigo (Fig. 5.16). The velocity eld (5.57) is
illustrated by plotting velocity vectors on a square
Fig 5.26 Velocity field (5.57) illustrated by plotting the
grid (Fig. 5.26).
velocity vectors for particles located on a square grid.
Consider the second model, in which rigid
layers rotate and slide relative to each other.
Recall that our model for this is a smooth repre- arises as to how to specify it. Further, the
sentation of the bulk deformation. The deforma- coefcients are functions of the current dip; in
tion is given by: the rst model this was not the case.
To get some feeling for (5.62) we evaluate the
x Fxx X X cos quantities:

(5.60)

1 d
1
y Fyx X FyyY X sin Y vx x tan
cos dt
(5.63)
If this applies throughout the folding process, we

1
d
must require Fxy(t) 0. Since examination of the vy x(1 tan 2 ) y tan
dt
form (5.53) indicates that this requires that Lxy
Velocity elds for 0 and 22.5 are shown in
vanish, (5.53) reduces to:
Fig. 5.27. The result for 45 is the same as that
dFxx shown in Fig. 5.26 for the attening model! That
Lxx Fxx
dt this must be so is seen by substituting tan 1 for
dFyx 45 in (5.63).
Lyx Fxx Lyy Fyx (5.61)
dt The velocity eld must change as the dip
dFyy changes, and you can perhaps visualize how
Lyy Fyy
dt these examples correspond to a combination of
In (5.60), Fyy 1/Fxx, and the restriction Lyy Lxx smoothed inter-layer slip and simultaneous rigid
also holds. We may use (5.60) to compute the rotation of the entire limb. To aid visualization,
coefcients for the velocity eld in (5.61): we decompose the velocity eld into these two
component parts. The rigid-body rotation will
d
Lxx Lyy tan always have the same form, but the part corre-
dt (5.62) sponding to the sliding will vary in magnitude
d with limb dip. For example, this decomposition
Lyx (1 tan2 )
dt is done for 22.5 in Fig. 5.28. The decomposi-
These expressions depend upon the rate of change tion of (5.63) is obtained by noting that the rigid-
of limb dip with time, and a question immediately body rotation (Fig. 5.28b) is expressed as:

1
d (a)
(vx)rotation y
dt
(5.64)

1
d
(vy)rotation x
dt
The shear is expressed by the velocity eld

obtained by subtracting (5.64) from (5.63). The
latter looks complicated, but yields a result like
that in Fig. 5.28a for any choice of .
5.4.1 Combined model for chevron

folding
We now formulate a kinematic model that com-
bines the De Sitter model, model 1, and the homo-
geneous attening model, model 2. The velocity
eld for the combined model is formed by summing
the two velocity elds in some proportion:
vx f1vx(1) f2vx(2) (b)
vy f1vy(1) f2vy(2) (5.65)
f1 f2 1
The fractions f1 and f2 might be taken to be func-

tions of time or limb dip. This model is ad hoc,
since we have advanced no physical principles
that would allow us to x f1 and f2. We present
these models to illustrate results that might
approximately simulate the development of
chevron folds and to provide some experience in
thinking about kinematics and deformation.
The fact that the velocity eld of the De Sitter
model (model 1) is expressed in terms of the rate
of change in dip, d/dt, creates difculty. We
might use the dip itself as a time-like variable, and
assign some arbitrary constant value to d/dt.
However, the dip is also changed by homogeneous Fig 5.27 Velocity field for the rigid layer chevron fold
attening (model 2). As one way of proceeding, we model: (a) for 0, (b) for 22.5.
express the progress of folding by means of the
rate of change in the span of the limb, normalized
by the span itself. For model 1:
The total relative rate of change of S, as might have

1 dS
(1)
1 (1) been anticipated, is:
v (S, 0)
S dt S x

1 (1)
L S tan
d
(5.66)

1 dS
S dt
L(1)
xx Lxx Lxx
(2) (5.68)
S xx dt
A simple combined model is one in which the
For model 2: individual contributions are in constant ratio:
(2)

1 dS
S dt
L(2)
xx (5.67) L(1)
xx
Lxx
1R
, L(2)
xx
RLxx
1R
L(2)
,R xx
L(1)
xx
(5.69)
(a)
Fig 5.29 Velocity field for combined model at 22.5 for

(b)
R 1. Compare with Figs. 5.28b and 5.27.
If both mechanisms contribute equally, R 1; the

velocity eld at 22.5 is shown in Fig. 5.29.
To follow the fold structure from an initial to
a nal state, we must rst determine how varies
with the relative change in limb span:
S
exp(Lxxt) (5.71)
S0
Here Lxx is taken to be constant, for simplicity.

From (5.66):
d(1) L(1)
xx (5.72)
dt tan
Figure 5.30 shows how to compute the rate of

Fig 5.28 Decomposition of velocity field for model 1 for change in dip associated with homogeneous
22.5: (a) layer-parallel slip or shear, (b) rigid-body attening:
rotation.
( tan d)
tan ( d)
(1 d tan )
One might then ask which value of R best simulates
a natural chevron fold in terms of amount of inter- [ tan vy (1, tan ) dt]
(5.73)
layer slip and homogeneous layer deformation on [1 vy (1, tan ) dt]
the fold limb. With the choice of Lxx as dening the
rate of the process, we may replace f1and f2 in (5.65) Substituting for model 2, we obtain:
using (5.69). The combined model is: d(2) 2 tan L(2)
xx
(5.74)
vx Lxxx dt (1 tan2 )
vy

1
1R
1
tan
tan x y Lxx (5.70)
Combining (5.73) and (5.74) with the expressions
(1)
for Lxx (2)
and Lxx :
y 60
1 50
x
0 R=0
d
40 1
tan d 2
Limb dip (o)

dd
5
30
vy (1, tan d)dt 10
vx (1, tan d)dt 20

Infinite
Fig 5.30 Trigonometric relations used to compute the rate
of change in dip, , associated with homogeneous flattening. 10
d L(1) 2 tan L(2) 0.5 0.6 0.7 0.8 0.9 1.0

xx xx
S/S0
dt tan (1 tan2 )
Fig 5.31 Limb dip as a function of S/S0 for composite

1

1R
1

2R tan
L
tan (1 tan2 ) xx (5.75) models with R as a parameter; initial dip of the seed fold
was 5.
Since, for constant Lxx, S/S0 is given by (5.71), we

may use (5.75) to obtain either (t) or (S/S0). In examples of chevron folds, which have limb dips
either case, we must integrate (5.75) numerically. of 50 to 70. Folding is initiated for seed folds with
It is somewhat surprising that a relatively simple (0) 5, meant to correspond to a dip that might
kinematic model produces such a complicated be achieved by initial buckling before the chevron
relation! Since: kinematics sets in. The limit R corresponds to
passive, purely kinematic amplication; the limit
d d R 0 corresponds to deck-of-cards folding in
tan (1 tan2 ) (5.76)
dt dt which the limb length does not change. If this
The relation (5.75) may be recast into the some- models natural chevron folds, attainment of ade-
what simpler form: quate dip would imply that a model with R 10 is
appropriate.
d
dt
tan
1
1R
1 tan2
tan
2R tan Lxx 5.4.2 Forward integration of the motion:
steady and non-steady velocity fields
(5.77) Because we know what the velocity eld is as a
function of limb dip and can keep track of that, it
A part of the description of chevron folding, is possible to follow the paths and current posi-
according to the present kinematic model, would tions of particles in the fold limb by numerical
be the variation of limb dip with relative fold span. integration. Given the current position of a par-
Integration of (5.75) for R 0, 1, 2, 5, 10, and ticle and the limb dip, we compute its velocity. We
yields the curves shown in Fig. 5.31. Here, the ratio then move the particle a small distance by the
of nal to initial fold span is given as S/S0 et, position increments x and y, so that the new
where t is expressed in units of 1/Lxx, and is used as positions are:
a measure of deformation or time. Note that the
progression of folding in the gure is from right to
x(t t) x(t) vx[x(t), y(t); (t)]t
left. The results shown extend to S/S0 0.5, a plau- (5.78)
sible typical amount of shortening for natural y(t t) y(t) vy[x(t), y(t); (t)]t
5.5 GENERAL RESULTS 183
lines; for three-dimensional motions, they might

be spherical surfaces, planes, or lines. These are
material loci, made up of particles. Figure 5.32
shows initial loci consisting of two perpendicular
lines of equal length and a circle at several stages
during folding following the homogeneous
attening model. Also shown is the continuous
path, or trajectory, taken by the intersection point
of the two line segments. The more nearly hori-
zontal line segment is taken to represent the
initial tilted bed surface segment, and so its incli-
nation equals the limb dip of the fold.
To determine the shape of a fold limb layer for
the combined mechanism model the velocity eld
(5.70) is used, together with (5.75). Results for the
De Sitter mechanism alone (R 0) and equal con-
tributions from each mechanism (R 1) are
shown in Fig. 5.33. The limbs end up with the
same span, S. The fact that the De Sitter model
produces a slightly larger strain, as measured by
the deformation of the initial circular locus is
puzzling. Recall that in this model, layers simply
undergo rigid-body rotation. Recall, however, that
the combined model is based on the continuous-
Fig 5.32 Initial and deformed material loci for deformation approximation to the De Sitter
homogeneous flattening model. model, which does not represent discrete layers.
The deformation shown is the bulk deformation.
It would be observed if a large circle had been
Here (t) is required to dene the velocity eld for inscribed on a stack of thin layers such as a card
the general model. This method uses quantities deck.
evaluated at the initial position of the particle. The velocity eld for the homogeneous atten-
Since the velocity will generally change continuing model is steady state: the velocity at any spatial
ously between the initial and nal positions, a position does not change with time. The velocity
gradual divergence between the computed trajec- of a particle moving through the velocity eld
tory of the particle and the true trajectory may does change. This steady velocity eld has been
develop. To minimize this, we use small time represented by vector arrows on a grid of posi-
increments. Alternatively, the velocity may be tions in Fig. 5.26. The continuous-deformation
computed at the new position, and the average variant of the De Sitter model, or the combined
velocity for the two positions used to re-compute mechanism model do not yield steady velocity
the new position. This method involves a slightly elds because the velocity distributions depend
lengthier computer program, but it provides upon , as illustrated by the velocity elds shown
more stable results. In the present examples this in Figs. 5.27 through 5.29.
rened method is not used.
Following individual particles may be infor-
mative, such as convective motion in a magma
5.5 General results
chamber. We will often follow particle loci that
allow us to visualize the velocity eld or the Our study of kinematic models for chevron folds
current state of deformation. Such initial loci for involved general concepts in deformation and
two-dimensional motions might be circles or kinematics. These are basic tools in the study of
(a) Time = 0.5, R = 0 handle the non-homogeneous case? The secret

here is to start by transforming our picture of
homogeneous deformation down to the innitesi-
mal region known as the neighborhood of a par-
ticle. You may guess this from your experience
with calculus.
5.5.1 Deformation in the neighborhood

of a particle: two-dimensional case
In a two-dimensional plane or axisymmetric
deformation the initial and nal positions of
particles remain in the same plane. A uniform
additional displacement of all particles, or trans-
lation, an operation that maintains the initial
orientation of the plane, would not affect this.
An example is the uniform additional upward
motion of the particles in the rising viscous
sphere. A description relative to an external xed
(b) Time = 0.5, R = 1 coordinate would include this translation. For
example, we might wish to describe the motion
of particles in the sphere relative to a coordinate
system xed with respect to the stationary uid
medium far from it.
From our discussion of examples of deforma-
tion, such as the olites of the South Mountain
anticline (Fig. 5.4) or the deformed xenoliths of
the Chindamora batholith (Fig. 5.9) you have
likely noticed that a description of deformation is
local. The range in variation between individual
olites within a hand sample may be large, but it
represents a statistical or random variation attrib-
uted to the facts that the initial shapes of olites
deviate from perfect spheres, and that they are
not homogeneous in their properties. One must
sample a volume that is representative of a homo-
Fig 5.33 Fold limbs produced by the combined mechanism geneous deformation. On the other hand the
model; initial layer dip of 10.
ratios of the axial lengths of the ellipsoidal shapes
of olites or xenoliths vary between sample loca-
the processes and products of rock deformation. tions and these variations may be representative
Here we introduce additional ideas and operations of a heterogeneous deformation.
and refer the interested reader to standard refer- Although they clearly and dramatically indi-
ences (Malvern, 1969). It seems appropriate at this cate that deformation has taken place, the axial
point to take up a few formal topics and questions. ratios of olites and their orientations do not
Deformation and ow certainly cant be restricted fully quantify the deformation. As another
to the homogeneous case over large volumes of example to introduce our present task, consider
rock: how would all the interesting structures that the deformed grid within the viscous sphere (Fig.
we observe in the eld have formed? We should be 5.13). This shows a change in the shape, size, and
suspicious from our strenuous, but not wholly sat- orientation of initially square material elements
isfying, modeling of chevron folds. How do we after an interval of ow within the sphere. These
(a) dx would no longer provide an adequate repre-

sentation of the line element.
dx dy Writing quantities in component form, the
innitesimal material line in the deformed state
is described by the vector components:
dX u dx
dY
x x x
dx dX dY
dX X Y
y y (5.79)
X dy dX dY
X Y
dx
The partial derivatives are referred to coordinates
(b) in the initial state and are called deformation gradi-
ds ents. For convenience, we write (5.79) in the form:
dX
u dx FxxdX FxydY
(5.80)
dy FyxdX FyydY
dS
We have encountered expressions of this form
Fig 5.34 (a) Infinitesimal vectors dX and dx with before in the case of homogeneous deformation
components (dX, dY) and (dx, dy) lie along a material line in chevron fold models, (5.45). The array of quan-
segment in the initial and final states, respectively. The
tities Fxx, Fxy, Fyx, and Fyy is a second-order tensor, since
vectors emanate from the initial and final positions of the
it satises the general denition:
same particle. Only the end points of the position vectors, X
and x, for that particle are indicated. The displacement A second-order tensor associates a vector with each
vector u has components (ux, uy). (b) The vectors are direction in space by means of a relation that is linear
represented in terms of polar coordinates (dS, ) and (ds, and homogeneous in the direction cosines (Prager,
), respectively. 1961).
The vector components are (dx, dy), and the direc-

attributes of the elements vary smoothly within tion cosines of the direction in space are the quan-
the body, and they are informative of the style tities dX/dS, dY/dS (Fig. 5.34). Second-order tensors,
and magnitude of deformation, but they do not relating pairs of vectors, have many physical appli-
provide a complete description. In the idealized cations (Malvern, 1969).
continuous medium of this model, we may Another description of the vector dx is found
dene the quantities describing deformation at a using components of the displacement vector, u,
mathematical point in terms of tensors based by noting that:
upon gradients in deformation and displace-
x X ux(X, Y)
ment. (5.81)
Consider a particle that occupies position (X, y Y uy(X, Y)
Y) in the initial state and position (x, y) (X ux, Y
The partial derivatives of these quantities are:
uy) in the nal state (Fig. 5.34a). An arbitrarily
oriented material line of particles that is innites- x u x u
imal in length is overlain by the vector dX with 1 x, x
X X Y Y
components (dX, dY) in the initial state. This ma- (5.82)
y uy y uy
terial line is translated, stretched, and rotated and , 1
X X Y Y
then is overlain by the vector dx with components
(dx, dy) in the deformed state. Innitesimal refers The partial derivatives of ux and uy in (5.82) are the
to the fact that, if the material line were not displacement gradients. The components of dx as
innitesimal in length, it would generally show functions of the displacement gradients are
some curvature in the nal state, and the vector found by substituting (5.82) into (5.79).
The innitesimal material line is represented which applies when the square of the extension is
by the vector dX, and therefore by its length dS much less than the extension and much less than
[(dX)2 (dY)2]1/2 and direction , measured coun- one. Because the innitesimal strain is dened in
terclockwise from the positive X-axis such that the terms of the displacement gradients we write the
components are (Fig. 5.34b): quadratic elongation (5.85) in terms of the dis-
placement gradients using (5.82). For small strains
dX dS cos , dY dS sin (5.83) we dont have to pay attention to the distinction
The vector dx may be similarly expressed in terms between position (x, y) and position (X, Y). Then,
of quantities ds and : the derivative operators with respect to initial
coordinates, /X and /Y, may be replaced by the
dx ds cos ,dy ds sin (5.84) operators /x and /y. If the displacement gradi-
2 2 1/2 ents all are much less than one, we may approxi-
Here ds [(dx) (dy) ] is the length of the ma-
mate the quadratic elongation by discarding
terial line in the deformed state.
quantities in multiples of the gradients such as
Deformation, as we have seen examples of it,
(ux/x)2 or (ux/y) (uy/x) to nd:
involves strain and rotation. A familiar aspect of
strain is a change in the length of a material line
which may be quantied using the square of the
stretch, (ds/dS)2, a dimensionless quantity referred
n
1 ux uy
2 x

y

1 ux uy
2 x

y
cos 2
to as the quadratic elongation. For an arbitrarily ori-
ented material line segment of innitesimal

1 u y ux
2 x

y
sin 2 (5.87)
length we dene the quadratic elongation using

(5.80) and (5.83): The innitesimal strain components in two
dimensions are:

2
ds (dx)2 (dy)2

ux u y 1 u y ux
dS (dS)2 xx , yy , xy yx
x y 2 x y
(Fxx cos Fxy sin )2
(5.88)
(Fyx cos Fyy sin )2
(5.85) Then, the extension at a point in the direction
12 (Fxx
2 F2 F2 F2 )
xy yx yy may be written in terms of these components by
substitution into (5.87):
12 (Fxx
2 F 2 F 2 F 2 ) cos 2
xy yx yy
(Fxx Fxy Fyx Fyy) sin 2 n 12 (xx yy) 12 (xx yy) cos 2 xy sin 2
(5.89)
The last two lines are found using the standard
double angle formulae (Selby, 1975). This is an The two-dimensional innitesimal strain tensor com-
exact description of change in length in two ponents form a symmetric array because xy yx:
dimensions at a point in terms of the deformation
gradients and the orientation of the material line
in the initial state.
xx
yx
xy
yy (5.90)
Another useful measure of change in line If all its components vanish, no innitesimal
length is the extension, dened as n (ds dS)/dS material line element at the point under consid-
(ds/dS) 1. The extension is related to the qua- eration will undergo a change in length.
dratic elongation as: Turning now to the concept of rotation, a
rigid-body rotation in the neighborhood of a par-

2
ds
(1 n)2 1 2n 2n 1 2n (5.86) ticle is illustrated in Fig. 5.35. An arbitrarily ori-
dS
ented innitesimal material line represented by
In the last step we have utilized the approxima- the vector dX in the initial state is rotated
tion taken to dene the so-called innitesimal strain through the angle , and there represented by
Y, y
sin 16 3
dS cos ( + v)
cos 1 12 2 1 (5.94)
dx
Under these conditions the derivative operators

dS sin ( + v)
with respect to initial coordinates, /X and /Y,

S
in (5.93) may be replaced by /x and /y, and the

=d
u dX
displacement gradients are related to the
ds
v dS innitesimal angle of pure rotation as:

dS sin

ux ux
dS cos
X, x x
uy
y
uy
0
0 (5.95)
x y
Fig 5.35 Pure rotation through the positive,
counterclockwise, angle . Infinitesimal vectors dX
and dx lie along a material line segment in the initial and final Combining (5.88) for the innitesimal strain
states, respectively. The vectors emanate from the same and (5.95) for the innitesimal pure rotation, the
particle fixed at the origin. The vectors may be represented displacement gradients may be expanded into
in terms of polar coordinates (dS, ) and (ds, ), respectively, symmetric and antisymmetric parts:
where dS ds.

ux 1 uy ux

x 2 x y

dx. The material line does not change length, so 1 uy ux uy
that ds dS. If all such material lines behave this
2 x y y
way, the local deformation is a pure rotation. From
Fig. 5.35:

1 uy ux
dx dS cos ( ) 0
(5.91) 2 x y

dS( cos cos sin sin ) 1 uy ux
0
Repeating these steps for dy and associating the 2 x y
trigonometric functions of the angle of rotation,
with the components of Fij we have:
Fxx cos , Fxy sin
xx
yx
xy
yy

0

0 (5.96)
(5.92)
Fyx sin , Fyy cos Here the innitesimal strain is symmetric and the
innitesimal rotation is antisymmetric.
Substitution of these expressions into (5.85)
When the strain and rotation are small, or
conrms the condition ds/dS 1.
innitesimal in the sense that squares of quanti-
The displacement gradients for pure rotation
ties can be discarded, special results apply. The
are obtained by substituting (5.91) and (5.92) into
deformation can be broken up into a rotation and
(5.80) and (5.82):
a strain, applied in either order, since the contri-
ux u butions are additive:
1 cos , x sin
X Y
u y
X
uy
sin , 1 cos
Y
(5.93) Fxx
Fyx
Fxy
Fyy

1 xx
yx
xy
1 yy (5.97)
If the rotation is small, 1, then the series As we shall see in later chapters, small strains
expansion of the trigonometric functions gives: and rotations are signicant in elasticity and its
application in models for the formation of joints, In the deformed state the vectors have the compo-
dikes, faults, or other structures. nents:
Notice that there are two applications of the
dx1 (Fxx cos Fxy sin)dS
innitesimal concept in the preceding discus-
sion. The rst is introduced in calculus and con- dy1 (Fyx cos Fyy sin )dS
cerns the treatment of a material line element (5.100)
dx2 (Fxx sin Fxy cos )dS
of arbitrary orientation that is represented by
the innitesimal vector dX with components dy2 (Fyx sin Fyy cos )dS
(dX, dY) in the initial state and dx with compo-
The cosine of the angle between these two vectors
nents (dx, dy) in the deformed state. These
is obtained by forming their scalar product:
vectors are contained within a neighborhood of
the particle or spatial point from which they dx1 dx2 (dx1)(dx2) (dy1)(dy2)
emanate that is sufciently small so one may
ignore the non-linear terms in the Taylor series (dx1)2 (dy1)2 (dx2)2 (dy2)2 cos
expansion about that particle or point. For (5.101)
example we have:
In the second line is the angle between the
x
X
x 1 2x

dx dX dY 2 X2 (dX)
Y
2 vectors. Substituting from (5.100) we have:
2
2x
XY
2x

(dX)(dY) 2 (dY )2
Y (5.98)
cos
1
2
Fxx
2 F 2 F 2 F 2 sin 2
xy yx yy
For (5.79) only the rst-derivative terms were kept.
The resulting measures of strain at a point, such
(Fxx Fxy Fyx Fyy) cos 2
as the quadratic elongation (5.85), are exact and
involve no approximations and therefore no

1 2
(F F 2 F 2 F 2 )
2 xx xy yx yy
errors of analysis. The second application is in the 1 2
simplifying approximation of innitesimal strain (Fxx 2 F 2 F 2 ) cos 2
Fxy yx yy
2
and rotation. If n 1 and 1, we discard

12
terms higher than rst-order in these quantities. (Fxx Fxy Fyx Fyy) sin 2
This introduces errors of analysis which may or
may not be tolerable (see Section 5.5.3).
An aspect of strain that is quite distinct from

1 2
(F F 2 F 2 F 2 )
2 xx xy yx yy
the change in length of material lines is shear. 1 2
Shear is a measure of the change in angle between (Fxx 2 F 2 F 2 ) cos 2
Fxy yx yy
2
two initially perpendicular material line ele-

12 1
ments (Fig. 5.36). Consider two initial innitesimal (Fxx Fxy Fyx Fyy) sin 2
(5.102)
vectors directed at right-angles to each other: dX1
with components (dX1, dY1) and dX2 with compo-
The change in angle between the two vectors is
nents (dX2, dY2). The rst innitesimal vector is ori-
the angle of shear, or /2 , where:
ented at the arbitrary angle . Since we are only
interested in the angle between the material line
elements in the deformed state, both vectors are
given the same length dS. In the initial state the
cos cos
2
sin (5.103)
vectors have the components: Considering only the angle change between the
two orthogonal material lines for the case 0,
dX1 dS cos ,dY1 dS sin the components of the two representative vectors
(5.99)
dX2 dS sin ,dY2 dS cos (5.99) reduce to:
Y, y This relation also gives the value of the compo-

dx2 nent xy of the innitesimal strain tensor for the
transformation of coordinates from axes (X, Y) to
dX2 dx1 axes (X, Y) rotated by . The geometric interpre-
b
tation to xy is one-half the change in angle
dS cos
between material line elements parallel to the

dS
p/2 coordinate axes X and Y in the initial state.

dX1
dS In summary we note that homogeneous deforma-
dS sin
tion consists of deformation within a nite region
dS cos X, x of a body for which the components of the defor-
mation gradient tensor (5.80) are uniform. The
dS sin nite region may extend to the entire body, but
Fig 5.36 Shear is the angle change between two material need not do so. It is always possible to express the
line segments initially perpendicular to one another, here deformation with respect to a coordinate system
represented by the pair of infinitesimal vectors dX1 and dX2. whose origin lies within this nite region. This is
In the final state these are dx1 and dx2. precisely the same logic as our use of a coordinate
system xed to the particle to describe the defor-
dX1 dS, dY1 0, mation in its neighborhood. Thus, for such a nite
(5.104) region, we may write as in the earlier sections the
dX2 0, dY2 dS relations given in (5.45). Everything said above
with reference to the neighborhood of a point
For this special case (5.102) reduces to: applies to the homogeneous deformation through-
out the nite region.
(Fxx Fxy Fyx Fyy)
cos sin (5.105)
(Fxx2 Fyx2 ) (Fxy2 Fyy2 ) 5.5.2 Deformation in the neighborhood
Since the results for the innitesimal strain of a particle: three-dimensional
and rotation case are much simpler, we write case
(5.102) in terms of the displacement gradient Here we summarize the three-dimensional case
using (5.82) and eliminate all products of these for deformation in the neighborhood of a particle
gradients. Then, using (5.88) for the strain compo- and write down general expressions for the defor-
nents we have: mation tensor and the strain tensor in terms of
deformation gradients and displacement gradi-
cos [(xx yy) sin 2 2xy cos 2] ents. Because there is little additional physical
{[(1 xx yy) (xx yy) cos 2 insight to be gained beyond the discussion of the
two-dimensional case given in the previous
2xy sin 2]12 [(1 xx yy)
section we rely more on indicial notation here. On
(xx yy) cos 2 2xy sin 2]12} 1 the other hand, rock deformation is inherently
(5.106) three-dimensional so one needs to be acquainted
with these forms of the basic kinematic equa-
The numerator is rst order in the small quanti- tions. While far from being an exhaustive descrip-
ties already, so we have: tion of strain and deformation (see for example
cos (xxyy)sin 2 2xy cos2 (5.107) Malvern, 1969) we provide a basis consistent with
the level of continuum mechanics used in this
Since the quantity is small, cos sin , so the textbook. Means (1976) provides an explanatory
angle of shear is related to the innitesimal strain review of strain and deformation in the context of
components as: structural geology.
Expanding the discussion of Fig. 5.34 to three
2 12 (xx yy) sin 2 xy cos 2 (5.108) dimensions, consider a particle located by position
xk xk
vector X with components Xi (X1, X2, X3) in the CIJ (5.111)
XI XJ
initial state that is located by position vector x with
components xi (x1, x2, x3) in the current state. An A counterpart to this tensor, which is used to
arbitrarily oriented material line extending from compute (dS)2 using partial derivatives of the
that particle in the initial state along the innites- initial coordinates with respect to the current
imal vector, dX, with components (dX1, dX2, dX3) is coordinates, is referred to as the Cauchy deforma-
translated, rotated, and stretched to lie along the tion tensor (Malvern, 1969).
innitesimal vector, dx, with components (dx1, In the Lagrangian formulation the change in
dx2, dx3) in the current state. Innitesimal refers to the squared length of the material line is related
the fact that, if the vector were not innitesimal in to the deformation gradients as:
length, the material line element would be curved
in the current state and the vector dx would not
provide an adequate representation. In other
(ds)2 (dS)2 dXI
xk xk
XI XJ
IJ dXJ (5.112)
words, the neighborhood of the particle is so

This result leads to the denition of the
dened that the deformation within it is reason-
Lagrangian strain tensor:
ably taken as homogeneous.
The components of the vector dx in the current
state may be calculated from the components of
the corresponding vector dX in the initial state as:
EIJ
1 xk xk
2 XI XJ
IJ (5.113)
This is an exact description of the strain at a point

xk
dxk dX FkMdXM (5.109) in a continuum which involves no approxima-
XM M
tions, and there are no restrictions upon the mag-
Here and in what follows capital indices are used nitude of the components. The Lagrangian strain
for coordinates in the initial state and lower case tensor (or its counterpart in the Eulerian formu-
indices for coordinates in the current state, and lation) often is referred to in the literature of
these indices have the range 1, 2, 3. The FkM are structural geology as the nite strain in contrast to
components of the three-dimensional deforma- the innitesimal strain, but it should be under-
tion gradient tensor for which the partial deriva- stood that these tensors are inclusive of all strain
tives are taken with respect to the initial magnitudes whether small or large.
coordinates. In other words this is the tensor that Comparing (5.111) and (5.113), the Lagrangian
associates the vector, dX, representing innitesi- strain tensor is related to the Green deformation
mal material lines of all possible orientations at tensor as 2EIJ CIJ IJ. These tensors are symmet-
the point X in the initial state with the vector, dx, ric and both have three orthogonal principal axes
representing those material lines in the current at the point X in the initial state, the correspond-
state. This is referred to as the Lagrangian formu- ing directions of which coincide (Malvern, 1969).
lation of the deformation gradients in contrast to The Green deformation tensor reduces to zero and
the Eulerian formulation in which the derivatives the Lagrangian strain tensor reduces to one as the
are taken with respect to the current coordinates magnitudes of the deformation gradients all go to
(Malvern, 1969, Section 4.5). zero.
The length of the material line in the initial
state is the magnitude of the innitesimal vector, 5.5.3 Errors associated with use of
dS |dX|, and the length in the current state is the infinitesimal strains
magnitude ds |dx|. The square of the current One of the rst questions that a structural geolo-
length is related to the deformation gradients as: gist should ask when taking up a problem related
x x to deformation in Earths crust is: should the
(ds)2 dXI k k dXJ (5.110) strains be approximated with the innitesimal
XI XJ
strain components? A positive answer opens the
This result leads to the denition of the Green door to the possibility that linear elasticity may be
deformation tensor: employed to model the deformation. A negative
answer means that the strain should be described displacement gradients referred to the initial
using the Lagrangian strain tensor (5.113) or its coordinates. Under these conditions the differ-
Eulerian counterpart. To address this question ences between partial derivatives taken with
one can, for example, compare the Lagrangian respect to the initial coordinates and those taken
and innitesimal strains for a given set of defor- with respect to the current coordinates are negli-
mation or displacement gradients and calculate gible, so the distinctions made here between the
the error. Whether or not that error is acceptable two sets of coordinates are ignored and the strain
depends upon the application. If one is looking for components are written:
order of magnitude results from a calculation,
large errors may be tolerable. If one is working
under strict engineering guidelines, only very
ij
1 ui uj

2 xj xi (5.118)
small errors may be tolerable. The two-dimensional forms of these equations

To estimate the error we write the Lagrangian referred to Cartesian coordinates are given in
strain tensor as given in (5.113) in terms of the dis- (5.88).
placement gradients by noting that: As an example of error analysis consider the
xi Xi ui (X1, X2, X3, t) (5.114) photographs of oids from the South Mountain
fold shown in the frontispiece for this chapter.
In other words the displacement components, ui, Recall that the average ratio of long to short axes
are functions of the coordinates in the initial state for the less deformed sample (on the left) is 1.16,
and of time. Here we have reverted to using only and that for the more deformed sample (on the
lower case indices in order to facilitate compari- right) is 1.56. For the sake of this example we
son with the innitesimal strain components as consider plane deformation (zero displacement
conventionally written. Taking the partial deriva- perpendicular to the photograph) and take the
tives of the xi as described in (5.114) and substitut- coordinate axes (X, Y) parallel to the average ori-
ing these into (5.113) we nd the Lagrangian entations of the long and short axes, respectively
components of strain: (no rotation of material lines that coincide with
the principal directions in the deformed state).
Eij

1 ui uj uk uk

2 X j Xi Xi Xj (5.115) For the more deformed sample we take the dis-
placement gradients in the plane as:
The innitesimal components of strain are
ux u
dened using the rst two terms on the right- 0.25, x 0,
X Y
hand side: (5.119)
u y u y
0, 0.20

1 ui uj X Y
ij (5.116)
2 Xj Xi
The components of the Lagrangian strain (5.115)
The error introduced by using (5.116) instead of are:
(5.115) may be dened as:
Exx 0.30, Exy 0, Eyx 0, Eyy 0.18
Eij ij (5.120)
eij 100 (5.117)
Eij
The components of the innitesimal strain (5.116)
This evaluates the error, eij, on a component by are:
component basis.
If squares and products of the displacement xx 0.25, xy 0, yx 0, yy 0.20
gradients are small enough compared with the (5.121)
gradients themselves to result in tolerable errors The errors in the two normal components of
then the last term on the right-hand side of (5.115) strain are found using (5.117):
may be dropped. What is left are the components
of the innitesimal strain written in terms of exx 17%, eyy 11% (5.122)
123o40' 123o30' axis perpendicular to the vertical fault surface,

and the z-axis vertical. Note that the displacement
vectors are approximately parallel to the fault
trace. Comparing the displacements of monu-
ments 5, 4, and 1, the displacement is seen to
decrease in magnitude with distance perpendicu-
39o00' lar from the fault. Furthermore, comparing mon-
uments 1 and 2, or 3 and 4, or 9 and 10, it is
apparent that the displacement does not vary
signicantly with distance parallel to the fault. We
know from investigations of many subsequent
earthquakes on the San Andreas Fault that rup-
tures typically extend to depths no greater that
about 10 to 15 km, so the slip is likely to have varied
from 4 m at the surface to zero at these depths.
Taking our observations of the displacement
y
vectors on Fig. 5.37 as representative of the dis-

z
placement eld throughout this region we char-

acterize the displacement components as:
x
38o50'
ux ux (y, z), uy 0, uz 0 (5.123)
Fig 5.37 Map of the horizontal displacement of twelve
monuments near Pt. Arena, CA, during the 1906 San The displacement of monument 5, adjacent to the
Francisco earthquake (Lawson, 1908; Pollard and Segall, fault zone, was about 2.5 m and the displacement
1987). of monument 1, at a distance of 13.5 km from the
fault zone, was about 1 m. Using these values, and
an estimated depth of faulting of 12.5 km, we esti-
In contrast the errors associated with the less mate the displacement gradients as:
deformed sample are 7% and 4%, respectively.
Whether or not these errors are tolerable depends ux ux 1.5 m
1.1 104
upon the magnitude of errors associated with y y 13.5 103 m
data collection and the application of the analysis. ux ux 2.5 m
2.0 104
As a second example consider the displace- z z 12.5 103 m
ment eld associated with faulting during the (5.124)
great San Francisco earthquake of 1906 along the
San Andreas Fault (Fig. 5.37). In the vicinity of Although the slip on the fault was several meters,
Point Arena near the northern-most trace of the the displacement gradients are so small that
fault the relative horizontal displacements of squares and products of these gradients may be
monuments were calculated from triangulation neglected. This conclusion admits use of innites-
surveys taken before and after the earthquake imal strains and suggests that linear elasticity
(Lawson, 1908). The displacement vectors on this theory would be an appropriate tool to investigate
map demonstrate that slip on the fault was right the faulting process (Pollard and Segall, 1987).
lateral, that the offset across the fault zone was
about 4 m, and that Earths surface displaced
during the earthquake at least 15 km away from
5.6 Concluding remarks
the fault by about 1 m. Can we use innitesimal
strains to characterize this deformation? In this chapter, we examined deformed objects in
We choose a Cartesian coordinate system and rocks belemnites, folded veins, concretions, and
reference frame on the trace of the fault with the xenoliths and indicated how a quantitative
x-axis horizontal and parallel to the trace, the y- measure of strain might be obtained from mea-
surements of them. Several examples demonstrate between observations of geometry and strain and
that strain typically has a continuous variation detailed models of structural evolution, we con-
within structures, providing one constraint on the sidered geometric and ad hoc kinematic models of
process of formation. More detailed treatment of chevron folds. We study a mechanical model for
practical methods of strain estimation from their origin in a later chapter. The kinematic
deformed geological objects are given in other model presented here provides a simple example
textbooks of structural geology (Ramsay and with which to illustrate the formal treatment of
Huber, 1983). Interpretation of deformed xenoliths strain and rotation as integrals that follow a par-
in the Chindamora batholith by means of an ticle moving through a temporally and possibly
appealing, but ad hoc, kinematic model illustrates spatially varying velocity eld. In the rising viscous
a methodology aimed at constraining the mecha- sphere example, strain ellipses were computed by
nism of intrusion by means of strain measure- following dense sets of particles by numerical
ments. In contrast, we showed how a simple means. The chapter concluded with a more formal
steady-state distribution of velocity in a rising treatment of plane deformation in two dimen-
diapir, provided by the complete mechanical sions, and descriptions of deformation and strain
model afforded by the Stokes solution, might be in three dimensions. In this discussion we point
used to follow the positions of particles, thus gen- out how to evaluate the errors expected when one
erating a detailed picture of the evolution of the chooses to employ the innitesimal strain in the
strain distribution in the body. As another bridge analysis of problems in structural geology.
Chapter 6
Force, traction, and stress
Photoelastic image of maximum shear stress contours in The concept of stress is the heart of our subject. It is
grains of model rock. Stress is concentrated at grain the unique way continuum mechanics has for specify-
contacts. Inset: photoelastic image of three circular disks ing the interaction between one part of a material
with point contact loads. Reprinted from Gallagher et al. body and another (Fung, 1969, p. 41).
(1974) with permission from Elsevier.
FORCE, TRACTION, AND STRESS 195
I
n this chapter we dene the relationships (a) y (b) y
F
among forces, tractions, and stresses. One of
f
the rst concepts encountered in a physics
a
class is that of the resultant force, F, acting on a
particle with mass, m, and the associated linear r
x z x
acceleration, a, of that particle in the direction m
that the force acts (Fig. 6.1a). For a rigid body (Fig. z T
Point
6.1b) one considers, for example, the resultant
torque, , about the axis z, due to the force, f,
acting at position, r, and the associated angular (c) t(n)
acceleration. For a deformable body the traction n
vector, t(n), is a measure of force per unit area
acting on the surface of a body (Fig. 6.1c), where
the surface has an orientation specied by the
outward unit normal vector, n. This surface can t(n) n
Area
be the exterior boundary of a rock mass or an
imagined surface within the rock mass. The trac-
tion vector is dened at a point on such a surface
Syy
in a limiting process as the area of a small
element of this surface shrinks toward zero about Syz Syx
(d) Sxy
the point. Similarly, if one imagines a small
Szy
cubical element of a given orientation at a point Szx S xx
within a body (Fig. 6.1d), one can dene the trac- Sxz
tions acting on all six sides as the volume shrinks y Szz
toward zero. The components of the traction on
each side dene the components of the stress Volume
acting on the cubical element with reference to
the chosen coordinate system, and this collection
of forces per unit area is referred to as a tensor x
quantity. Normal and shear components of the
stress tensor are directed perpendicular and par- z
allel to the sides of the cubical element, respec-
tively. In this hierarchy of concepts the force
vector acts on a point mass; the torque vector acts Fig 6.1 Force, torque, traction, and stress are illustrated
about an axis; the force per unit area, or traction as: (a) force vector, F, acting on a particle of mass, m, with
vector, acts on a surface element; and the set of corresponding acceleration vector, a; (b) torque vector, ,
forces per unit area, or stress tensor, acts on a due to force, f, acting at position vector, r; (c) traction
volume element. vector, t(n), acting on surface element with outward normal
The traction and stress are of interest to the n; (d) stress tensor, ij, acting on volume element with edges
structural geologist because structures develop as parallel to coordinate axes.
the rocks of the Earths crust strain and ow, and
the distribution of this deformation is related
to the stresses acting within the rock mass and tities with spatial and temporal variations. For
the tractions acting on its surfaces. In this and example, the frontispiece for this chapter is a pho-
later chapters we show how the concepts of trac- tographic visualization of the distribution of
tion and stress can be applied to understand the shear stress in the grains of a model sandstone
origin and evolution of geological structures. In (Gallager et al., 1974). An understanding of the pos-
most natural examples the traction and stress sible variations of these elds in the Earth is of
vary with position and time: they are eld quan- fundamental importance to structural geologists.
196 FORCE, TRACTION, AND STRESS
6.1 Concepts of force and traction (a)

b
In the context of rock masses that behave as
deformable solids or uids two classes of forces
are recognized: body forces act on volume elements
within the rock mass and surface forces act on dV dV
surface elements, either the actual surface of the
rock mass or imaginary surfaces within it. z
Common examples of body forces are those due to
gravity and magnetic attraction, both of which y
act at a distance rather than through the direct
contact to two objects. In contrast surface forces
x
are those due to the direct contact of one object
dy i =1
with another. For an imaginary surface within a (b)
rock mass one can think of the surface force due
to the rock mass on one side of that surface in i =2
contact with the rock mass on the other side.
dz
Because rock is a porous material one has to con-
sider the nature of such a contact and the i =3
denition of surface forces with some care. Body
and surface forces are dened at arbitrary points P
in the continuous medium that we take as a
model for the rock mass. In this section we con- z dx
sider body and surface forces and the traction
vector; in the next section we take up the stress y
tensor.
x
6.1.1 Body force df3
If the vector b is the body force per unit mass df2 df1
acting on an innitesimal volume element dV
(Fig. 6.2a) then the resultant of all body forces Fig 6.2 (a) Body force, b, per unit mass acting on
acting on the nite volume, V, is (Malvern, 1969): infinitesimal volume element, dV, within a finite volume, V.
(b) Sequence of finite volumes, Vi, with resultant body

b dV ex bx dV ey by dV ez bz dV forces, fi.
V V V V
(6.1)
volume with sides aligned with the coordinate
Here is the mass density and (bx, by, bz) are the system as in Fig. 6.2b), the volume integral may be
Cartesian components of b, and both the density evaluated as a triple integral with appropriate
and body force may vary with the spatial coordi- ranges for the limits in the three coordinates:
nates. The resultant body force is the vector sum of z2 y2 x2
the body forces acting on all innitesimal ele-
ments within the nite volume V. The density and b dx dy dz
z1 y1 x1
(6.2)
the body force per unit mass may vary in time, in
which case (6.1) is considered to represent a given As indicated in (6.1) each component may be
instant in time. In principle, if the spatial varia- written as a separate triple integral and these are
tions of b are known as functions of the three added to compute the resultant body force.
coordinates, and if the shape of the nite volume In general we want to characterize the body
is relatively simple (for example a rectangular force at any arbitrary point P in the deformable
6.1 CONCEPTS OF FORCE AND TRACTION 197
solid or owing uid and this requires an ideal- Taylor, 2003). For a Cartesian coordinate system
ization of the rock mass as a material continuum. with z-axis vertical and directed upward (Fig. 6.2b)
For this purpose we consider a sequence of nite the components of the body force per unit mass
volumes, Vi, each with resultant body force, fi, are:
and average density, i, and each containing the
point in question (Fig. 6.2b). The nite volumes are bx 0, by 0, bz g* (6.4)
ordered from largest, i 1, to smallest i n, such
that the volume approaches zero as n . In this For the rectangular body of volume V shown in
limit the largest dimension of the sample Fig. 6.2b with side lengths x x2 x1, y y2 y1,
approaches zero, so the volume converges to the and z z2 z1 the resultant body force magni-
point P, not to a surface or a curve containing the tude is:
point (Malvern, 1969). Furthermore, the resultant z2 y2 x2
body force varies as one considers successive

volumes in the sequence and approaches zero in
bz dx dy dz g*xyz g*V
z1 y1 x1
(6.5)
the limit. The formal denition of the body force
per unit mass at the point P in the material con- Here density is taken as uniform so the calcula-
tinuum is: tion is greatly simplied.
The body force per unit volume is the vector
b lim
n
n
fn
Vn
(6.3)
quantity g. Under conditions where both mass
density and the gravitational acceleration are
taken as known values, constant in time and
Although both the numerator and the denomina- uniform in space, the components of the body
tor in (6.3) approach zero, a fundamental postu- force per unit volume with coordinates as in Fig.
late is that the ratio approaches a denite value 6.2 are:
at each and every point in the continuum. The
difculty encountered when rationalizing deni- bx 0, by 0, bz g* (6.6)
tions such as this with our knowledge of the prop-
The gravitational body force plays an important
erties of real materials is resolved as follows:
role in many tectonic processes from the buoyant
When we pass, however, to the innitesimal limit dV rise of magma and salt through the crust to the
(volume) or dS (surface), we are dealing with a hypo- loading and unloading of buried rock masses
thetical concept, a continuum or continuous medium, during mountain building and erosion.
whose justication depends not on any study of actual
materials in the small, but rather on the efcacy and 6.1.2 Surface force: the traction vector
utility of the concept in enabling us to describe and
In order to discuss deformation in the Earth we
predict the behavior of actual materials in the large,
need a way of talking about and quantifying the
i.e., the macroscopic behavior (Malvern, 1969).
distribution of forces acting on an arbitrary
For rock one must be aware that application of the surface within a rock mass. For example, we
material continuum may be problematic at the might ask what were the forces distributed on the
grain scale and at the intermolecular scale where surfaces of a fault that would cause it to slip? Or,
sharp discontinuities break up an otherwise con- what are the forces distributed on the surfaces of
tinuous body. a dike that would cause it to open? Questions like
For most problems in structural geology these require us to understand and use the trac-
gravity is the only signicant body force, so the tion vector. The traction vector also is used to
body force per unit mass is the gravitational accel- dene the distribution of forces (or lack thereof)
eration, b g. Furthermore, the magnitude of acting on any external surfaces of a rock mass. For
this body force, g, usually does not vary signi- example, the Earths surface is characterized as
cantly over length scales from meters to kilo- being a traction-free surface. Of course large build-
meters, so it may be taken as uniform and equal to ings, dams, and other engineering structures
the standard value g* 9.806 65 m s2 (Mohr and impose non-zero tractions and the lling of large
reservoirs imposes sufcient traction to depress d df

the surface over a broad region. The weight of the x
dA
atmosphere provides a normal traction (1 atm
P P
0.1 MPa) and the wind imposes some shear trac-
tion, but all of these tractions acting on Earths
surface usually are ignored because the tractions
involved in the tectonic processes leading to the n
development of geological structures are signi- P
cantly greater in magnitude. (+)
Consider a surface, S, arbitrarily located Rock mass
S (-)
within the Earth and shown in cross section as the
dashed curve in Fig. 6.3. This is not necessarily a
physical boundary between rocks of different
lithologies, nor does it have a particular scale. The Fig 6.3 At point P on an arbitrary surface S within a rock
surface could be located within a single rock unit mass the outward unit normal is n. Inset shows force vectors
or within a single mineral grain. The surface acting on surface with area A containing the point P.
passes through the point P and the orientation of Resultant force is f and resultant torque is .
the surface at this point is given by the outward
unit normal vector n. We refer to the positive ( )
side of the surface with reference to how n is through the point (Malvern, 1969). The resultant
directed. Given the orientation of n, the surface S forces on each successive patch may vary in mag-
bounds that portion of the rock mass on the neg- nitude and direction, but they also approach zero
ative side, indicated by the gray swath on the cross in this limit. The ratio of resultant force to surface
section. area is dened as the traction vector, t(n):
The forces acting on the surface S could have a
complex distribution, varying from point to point
in both magnitude and direction. This is schemat-
t(n) lim
n
fn
An
(6.7)
ically illustrated in an enlarged view (Fig. 6.3, The French mathematician Augustine-Louis
inset) by a collection of arrows with their tails or Cauchy (17891857) apparently rst contemplated
heads on a small patch of the surface of area A this ratio and proposed that it approaches a
that contains the point P. These forces account for denite value in this limit (Fung, 1969). The trac-
the mechanical action of the rock mass on the tion vector sometimes is referred to as the stress
positive side of the patch and they may be repre- vector, but the stress is a different construct so
sented by a resultant force, f, acting at the cen- this choice of words is confusing and should be
troid, P, of the patch, and a resultant torque, , avoided.
acting about an axis through the centroid. The We have written the traction vector as t(n) in
resultant force and torque are not restricted as to (6.7) to emphasize that this quantity is a function
direction. In some contexts the resultant torque is of the orientation of the surface upon which it
referred to as the resultant moment. acts. This orientation is specied by the vector n of
In a thought experiment the rock mass is ide- unit magnitude, directed outward and normal to
alized as a material continuum and we consider a the surface (Fig. 6.4a). An innite number of sur-
sequence of patches of the designated surface, S, faces (S1, S2, S3, . . .) with different normal curva-
with nite areas, Ai, each associated with a dif- tures may be constructed through the point P, all
ferent resultant force, fi, and each containing the having the same normal, n, at that point. The trac-
point P. The patches are ordered from largest, tion vectors acting on all of these surfaces at P are
i 1, to smallest i n, such that their surface areas identical in magnitude and direction. On the
approach zero as n . In this limit the largest other hand two surfaces may be constructed
dimension of the patch approaches zero, so the through the same point P with different normals,
patch converges to the point P and not to a curve n(1) and n(2) (Fig. 6.4b). The traction vectors, t[n(1)]
(a) t(n) tion vector for that force. The magnitude of this
position vector is the distance from the centroid
to the point of application of the force on the
n surface, and in the limit this distance goes to zero,
so Cauchys proposal has an intuitive appeal. Note
that a resultant torque may exist on a surface of
P nite area. For example, as a layer of sedimentary
rock is bent during folding most cross-sectional
Surface surfaces of the layer are loaded by forces that con-
S1 tribute to a net torque. This torque is called a
bending moment and these moments play a
prominent role in theories of bending and folding
S2
S3 (Timoshenko, 1958; Timoshenko and Woinowsky-
Krieger, 1959). The theory of coupled stresses,
which will not be considered further in this text-
book, admits the possibility of a continuous dis-
(b) t[n(2)] tribution of torques per unit area, with a limit at
n(1)
a point on a surface that is different from zero
(Malvern, 1969).
t[n(1)] n(2) The traction at a point on a surface is a
measure of the force per unit area imparted by the
material from one side of the surface to the ma-
Surface terial on the other side. This concept can be stated
succinctly as (Fung, 1969, p. 51):
P
t[n(2)] t[n(1)] (6.9)
S1
Here it is understood that n(1) and n(2) are unit
normals at the same point on a surface viewed
S2 from opposite sides, so n(1) n(2). To interpret
(6.9) consider Fig. 6.5a in which a rock mass is
Fig 6.4 Relations among surfaces Si through point P and divided into two parts, 1 and 2, by the surface, S,
traction vector at point P. (a) Surfaces with different
illustrated here in cross section. The outward unit
curvatures but same outward unit normal, n, have the same
normal vector for part 1 is n(1), and that for part 2
traction. (b) Surfaces with different normals have different
tractions regardless of their curvature.
is n(2): these are oppositely directed vectors. At the
point P the rock of part 2 exerts a traction, t[n(1)],
on part 1. As drawn, part 2 is pulling on part 1 at
an angle that is somewhat oblique to n(1). Now,
and t[n(2)], acting on these surfaces are different imagine removing part 2 and replacing the
in magnitude and direction. mechanical action of part 2 on part 1 with the
In a thought experiment similar to that appropriate distribution of tractions. If this oper-
leading to (6.7) the ratio of resultant torque to ation were done accurately, according to Cauchy,
surface area is considered (Fig. 6.3). Cauchy appar- nothing about the mechanical state of part 1
ently contemplated this ratio and proposed that would change. Now consider the same surface
the limiting value approaches zero: with part 1 removed (Fig. 6.5b). What traction
would have to be applied at point P to replace the
lim
n
n
An
0 (6.8) mechanical action of part 1 on part 2? According
to Cauchys concept, we would have to apply a
Torque is the vector product of the force acting at traction, here called t[n(2)], of magnitude equal to
a given distance from the centroid and the posi- t[n(1)], but oppositely directed.
(a) To summarize, Cauchy apparently enunciated

t[n(1)] most of the following points concerning the
traction vector in 1822 (Truesdell, 1961):
Part 2
n(1) 1. The traction is a vector quantity that acts at a
point on an imaginary or real surface of
arbitrary orientation (Fig. 6.3), specied by the
outward unit normal vector, in the interior or
P
on the exterior of a body.
Surface, S 2. The traction measures the limiting ratio of
Part 1
resultant force to surface area on a patch of the
surface as this patch shrinks down about a
point.
3. Different surfaces with the same orientation at
a common point are acted upon by the same
traction (Fig. 6.4a), but differently oriented sur-
(b) faces through that same point are acted upon
by different tractions (Fig. 6.4b).
4. The traction vector can vary in orientation
Surface, S from acting normal to the surface to acting tan-
gential to the surface.
Part 2 5. The traction at a point on a surface is equal and
opposite to the traction that acts at that same
point for the same surface with opposite
P outward unit normal vector (Fig. 6.5).
Part 1
6. The physical dimensions of traction are force
n(2) per unit area, M L T2 L2 M L1 T2.
7. The SI units for the traction are N m2 Pa.
t[n(2)] The conceptualization of the traction vector was a

major accomplishment in the history of develop-
ment of continuum mechanics.
Fig 6.5 Surface, S, separates two parts of a body (Fung,
1969). (a) At point P part 2 exerts a traction t[n(1)] on part 6.1.3 Application of the traction vector
1. (b) At point P part 1 exerts a traction t[n(2)] on part 2. to rock
These two tractions are equal in magnitude and oppositely
The denition of the traction vector, t(n), at a
directed.
point on a surface with outward unit normal, n,
depends upon the limit (6.7) in which a small
patch shrinks toward the point and the ratio of
The relationship expressed in (6.9) is an exten- resultant force acting on the patch to the area of
sion of Newtons Third Law: to every action there is the patch converges smoothly to a denite value.
always opposed an equal reaction; or, the mutual Such a denition is entirely appropriate for the
actions of two bodies upon each other are always idealized material continuum, but the applica-
equal, and directed to contrary parts (Resnick and tion of this concept to rock requires careful con-
Halliday, 1977, p. 79). Here action refers to force, so sideration of scale and the constituent properties.
equal and opposite forces describe the mechani- As can be appreciated by glancing at Fig. 6.6a, rock
cal response of one body upon another. By consid- viewed at the grain scale can be highly hetero-
ering a small patch of a surface, and the resultant geneous with a multitude of sharp discontinuities
force acting on that patch, this law is extended to in material properties. This image of sandstone is
the traction vector. a few hundred micrometers across and shows
several sand grains, some grain fragments,

(a)
cement, and pores. The grain-scale heterogeneity
of rock leads to a very complex distribution of
forces, tractions, and stress (see frontispiece of
this chapter). On the other hand the image of
sandstone in Fig. 6.6b is a few decimeters across
and shows, with the exception of the prominent
zone of deformation bands, a remarkably homo-
geneous and apparently continuous material.
From these images important questions arise
about mechanical behavior of rock. Can one
ignore the grain-scale heterogeneity when consid-
ering the behavior of rock at the scale of an expo-
sure? For what scale of problems must one
explicitly include the geometry and differing
50 m material properties of individual mineral grains?
At the grain scale (Fig. 6.6a), the boundaries of
each grain would be explicitly dened and trac-
(b)
tions on these surfaces would serve as boundary
conditions. For example, the grains might be ide-
alized as continuous and homogeneous elastic
spheres with traction-free surfaces, except on
areas of contact where the non-zero normal and
shear tractions would account for the distribution
of forces transmitted across these areas of contact.
The traction on any surface at points in the space
between spheres would have no meaning in this
context. The traction on any surface at points
within each sphere would be determined by the
boundary conditions. In contrast the sandstone at
the exposure scale (Fig. 6.6b) could be idealized as
a homogeneous elastic solid, perhaps innite in
extent with no internal or external boundaries,
except around the zone of deformation bands.
The traction on any surface at any point would
ignore the grain-scale heterogeneity and would be
interpreted as described below.
To appreciate the differences between the
grain-scale and the exposure-scale interpretation
of the traction in sandstone consider an imagi-
nary surface that cuts across many grains and
50 mm pores. On this surface the traction may vary from
negligible (within a pore) to order 10 to 100 MPa
Fig 6.6 Scale dependence of homogeneity in a rock mass. (near a grain-to-grain contact). For exposure-scale
(a) Sandstone at 50 m scale showing grains and pores: problems we seek a patch size on the surface
material is heterogeneous (photograph courtesy of Xavier du where a meaningful average of the grain-scale
Barnard). (b) Sandstone at 50-mm scale: material is uctuations is achieved. The size of such a patch
approximately homogeneous except for zone of deformation
may be estimated using a model in which a bed
bands. Photograph by D. D. Pollard.
of springs (Fig. 6.7a) replaces the mechanical
(a) F The relationship between the force, F, and the

displacement, u, of a spring with constant, K, is
i =1 2 3... i=N
u F Ku, so, for the one-dimensional loading system
illustrated in Fig. 6.7a, we have:
F Ku,Fi Kiu (6.10)
Ki H
Here F is the total force magnitude applied to the
platen and K is the effective spring constant. By
effective spring constant we mean the constant for a
single spring that would have the same relation-
Wi ship between force and displacement as the entire
W
bed of springs. The effective spring constant is the
sum of all the individual constants, just as the total
width of the array is the sum of individual widths:
(b) 2.0
N N
1.8 K Ki,W Wi
i1 i1
(6.11)
1.6
1.4 It is a characteristic of springs arranged in paral-
1.2 lel that the spring constants are additive.
Because the platen that loads the bed of
tp / t
1.0
springs (Fig. 6.7a) applies the same displacement,
0.8
u, to each constituent, the ratio of total force to
0.6 effective spring constant, F/K, must be the same as
0.4 the ratio of any individual force to the individual
0.2 spring constant, Fi /Ki. For the purpose of under-
0 standing the denition of the traction vector, we
0 10 20 30 40 50 60 70 80 90 100 substitute the appropriate traction, multiplied by
Number of grains the surface area on which it acts, for the forces
acting on the constituents to nd:
Fig 6.7 Model for grain-scale heterogeneity of rock.
(a) Bed of springs with different stiffnesses accounts for
tDW ti DWi
u (6.12)
grains and pores. (b) Partial traction normalized by total K Ki
traction approaches a unit value as number of grains
Here t is the total traction on the platen of area DW,
increases. Reprinted from Amadei and Stephansson (1997)
and the ti are the individual tractions that replace
with kind permission of Springer Science and Business Media.
the mechanical action of the platen on each con-
stituent of area DWi. Because they have a negligible
behavior of the individual constituents of the spring constant, those constituents representing
sandstone at the grain scale (Amadei and pores carry a negligible traction regardless of
Stephansson, 1997, p. 50). A row of N grains and their width. However, for a given displacement,
pores is shown in two dimensions where the this equation demonstrates that the traction
individual widths, Wi, combine to a total width, acting on particular constituents increases in pro-
W. The height, H, before loading, and the depth portion to decreases in width. Also, the traction
out of this plane, D, are constants. The stiffness of increases in proportion to increases in the spring
each constituent is approximated with spring constant. In other words, thinner constituents of
constant, Ki, and the force acting on each spring is the same stiffness carry greater traction, and
Fi. The spring constant for pores would be very stiffer constituents of the same width carry
small compared with that for grains. The greater traction.
loading system is a rigid platen that imposes the Consider an area that is only a part of the total
same displacement, u, on all the springs. area, and the partial traction, tp, transmitted from
the platen to the constituents under this partial area would vary slowly and approximately contin-
area. The partial traction is dened as: uously throughout the rock. The area dened by
this procedure is referred to as the representative
n n
F u K elementary area, because it includes enough con-

i i
im im 1mN stituents to be representative of the rock as a
tp , (6.13)
n n mnN
D
im
Wi D
im
Wi whole. In this way the concept of the traction
vector, which presupposes a material continuum,
Note that the partial traction is found by is extended to rock that is both heterogeneous
summing the forces acting on the appropriate and discontinuous at the grain scale (Fig. 6.6a),
constituents, and then dividing by the partial but effectively homogeneous and continuous at
area that this net force acts upon. The traction the scale of a hand sample or exposure (Fig. 6.6b).
acting on the entire platen for the total of N con-
stituents is: 6.1.4 Variation of the traction with
N orientation of the surface
F uK
u Ki
i1
The traction vector at a point, P, varies with the
t N (6.14) orientation of the surface, S, upon which it acts
DW DW
D Wi
i1
(Fig. 6.4b). To understand how the traction varies
consider a small tetrahedral element with three
We assume here that N is great enough so that a orthogonal sides that are parallel to the respective
meaningful measure of the traction is assured. coordinate planes. The fourth side is a patch of the
To compare the partial and total tractions for surface S with area A that includes the point P
different areas sampled consider their ratio: and is arbitrarily inclined to the coordinate
planes (Fig. 6.8a). This is referred to as the Cauchy
n N
K W tetrahedron because it was introduced in publica-

i i
tp im i1 1mN tions by Cauchy in 1823 and 1827 (Malvern, 1969).
, (6.15)
t n N mnN The orientation of the patch is determined by

im
Wi
i1
Ki
the outward-directed unit normal vector n,
which makes direction angles (x, y, z) with the
As the partial area approaches the total area in
coordinate axes. Each angle is taken as the smaller
size, that is as m goes to 1 and n goes to N, the ratio
of the two in the plane containing n and the
tp /t goes to 1. In other words, the partial traction
respective coordinate axis. Because n is a unit
approaches the value of the total traction, as
vector, the components (nx, ny, nz) are the direction
expected. For partial areas less than the total area,
cosines:
the ratio may differ from one, and this difference
is a measure of the variability in traction intro- nx cos x,ny cos y,nz cos z (6.16)
duced by the heterogeneity in stiffness and width
Recalling the denition of the magnitude of a
of the individual constituents.
vector (2.7), these components are related as:
An example of many calculations of the ratio
tp /t for different sample areas from a model rock (nx)2 (ny)2 (nz)2 1 (6.17)
is shown in Fig. 6.7b (Amadei and Stephansson,
The areas of the three orthogonal sides of this
1997, p. 50). For small numbers of constituents in
element are found by projection of A onto the
the partial sample the scatter of the ratio about 1
coordinate planes such that:
is signicant, but the inclusion of 50 constituents
reduces the scatter to less than 10%. Since typical Ax Anx, Ay Any, Az Anz (6.18)
grains in medium sandstone are less than
0.50 mm in diameter, an area on the order of For this derivation n points into the positive
10 mm2 would be large enough to average out octant, but the resulting equations actually apply
most of the variability in partial traction. Using to all orientations on n. These relationships are
this area the ratio of resultant force to surface used below to derive equations that describe the
(a) dAx z is possible to choose an orientation of Cartesian

coordinates such that the tractions acting on
az the mutually orthogonal surfaces of a cubical
n element have zero tangential components (Fig. 6.8
inset). That is, the tractions are exactly perpendic-
ax ay ular to the surfaces on which they act. On the
dAy nz
ny three positive sides of this element (those with
y
outward normals directed along the positive coor-
P
x nx dinate axes) we label these tractions t1, t2, and t3,
and note that their components are (t1, 0, 0), (0, t2,
0), and (0, 0, t3). In keeping with convention the
t3 tractions are ordered such that t1 t2 t3. In the
dA
limit as this element shrinks to the point P we use
dAz (6.9) to determine that the tractions on the nega-
t2 tive sides are t1, t2, and t3 and their compo-
(b) z t1 nents are (t1, 0, 0), (0, t2, 0), and (0, 0, t3).
Now imagine that the tetrahedral element is
n cut free of the cubical element and the surround-
t1 ing material and the mechanical action of this
t(n)
material on the element is replaced with the
appropriate tractions (Fig. 6.8b). For the moment
t2 ty tz we ignore any body forces acting on the element.
P tx Although we understand that forces are distrib-
uted over the surfaces of the element, the trac-
y tions are drawn as single arrows with their heads
or tails at the mid-points of the sides and each rep-
resents the appropriate ratio of resultant force to
area as dened in (6.7). The tractions acting on the
t3
three mutually orthogonal sides are t1, t2, and
x t3 and they are directed along the negative coor-
Fig 6.8 Cauchy tetrahedron used to define traction dinate axes. The traction, t(n), acts on the patch
variation with orientation of plane. (a) Surface areas of sides with outward normal, n, and has components
parallel to coordinate planes are (Ax, Ay, Az); direction [tx(n), ty(n), tz(n)] in the coordinate directions. In
angles for outward unit normal, n, are (x, y, z); and general there are no restrictions on the orienta-
components of the normal vector are (nx, ny, nz). (b) Traction tion of this traction vector: it may be parallel,
vectors acting on sides parallel to coordinate planes are inclined, or perpendicular to n.
(t1, t2, t3) and components of the traction vector t(n)
To nd the relationship among the tractions
are (tx, ty, tz).
acting on the tetrahedral element of Fig. 6.8b we
use Newtons Second Law, F ma, and postulate
variation of the traction vector with the orienta- static equilibrium, a 0, so the net force in
tion of n. each coordinate direction is zero. Taking the x-
In what follows we consider the limiting case coordinate direction as an example and comput-
in which the length of the longest edge of the ing the force components as the respective
tetrahedron (Fig. 6.8a) goes toward zero, so the traction components times the areas of the sur-
sides converge toward the point P and the trac- faces on which these tractions act, the sum of the
tions acting on the four sides act at that point. The force components is written:
orientation of the coordinate planes that dene
the three orthogonal sides of the element is not f [t (n)]A t A
x x x x
(6.19)
arbitrary. In a later section we show that it always [tx(n)]A t1Ax 0
Rearranging (6.19), we use (6.18) to eliminate the Traction space ty (n)

t1
areas and write tx(n) t1nx. Here the absolute value
sign in (6.18) has been dropped, so this relation-
ship applies to the full range of orientations of n. t2 t(n) t2
By similar arguments we have:
O t1 tx (n)
tx(n) t1nx,ty(n) t2ny,tz(n) t3nz (6.20)
t3 t3
Note that the components of t(n) have opposite
signs to the respective components of the trac-
tz (n)
tions acting on the orthogonal sides of the
element. Because, for example, the traction t1 is
Fig 6.9 The traction ellipsoid is constructed with
directed in the negative x-direction, the compo-
coordinate axes (tx, ty, tz) in traction space. The semi-axes of
nent tx(n) must be positive for a balance of forces.
the ellipsoid are coincident with the traction vectors (t1, t2,
As the inclined patch rotates toward parallelism t3).
with the (y, z)-plane such that n is directed along
the positive x-axis, we have nx 1, ny 0, and
nz 0 so the y- and z-components of t(n) go to zero prolate ellipsoid t1 t2 t3, the oblate ellipsoid
and tx(n) t1, in keeping with (6.9). On the other t1 t2 t3, and the sphere t1 t2 t3.
hand, as n approaches the negative x-axis, we have The symmetry and continuity of all possible
nx 1, so tx(n) t1. Although we derived (6.20) traction variations at a point are claried by
by ignoring body forces and postulating static the traction ellipsoid, which apparently was
equilibrium, we show in Chapter 7 that this rela- described by Gabriel Lam (17951870) between
tionship is more generally applicable to problems 1820 and 1830 (Fung, 1965, p. 76). This construc-
of both solid deformation and uid ow that tion commonly is referred to as the stress ellip-
include body forces and admit accelerations in soid, but the quantity being plotted is the traction
the equations of motion. vector. To emphasize the fact that traction and
The relationships in (6.20) demonstrate that stress are distinct physical quantities we use the
the traction vector t(n) is equal to the special trac- name traction ellipsoid. The traction vectors t 1, t 2,
tions t1, t2, and t3 as n is successively directed in and t 3 are shown in Fig. 6.9 aligned with the
the positive x-, y-, and z-directions, and it is equal respective positive axes of the traction component
to t1, t2, and t3 as n is successively directed in coordinates. The traction vectors t 1, t 2, and
the negative x-, y-, and z-directions. But how does t 3 would be aligned with the respective negative
t(n) vary for intermediate orientations? Solving axes, and t(n) is drawn with all positive compo-
each of (6.20) for the respective component of the nents, consistent with the balance of forces on the
unit vector and substituting these into (6.17) we tetrahedral element of Fig. 6.8b.
have: The traction ellipsoid provides a useful visual-
ization of the traction variation at a point, but
[tx(n)]2 [ty(n)]2 [tz(n)]2
1 (6.21) does not by itself reveal the orientation of the
[t1]2 [t2]2 [t3]2
surface on which a particular traction vector acts.
This equation is in the standard form for an ellip- The special tractions t 1, t 2, and t 3 are exactly par-
soid drawn in traction space with coordinate allel to the unit normal vector for the surface on
axes tx(n), ty(n), and tz(n), (Fig. 6.9). The semi-major, which they act, but for the general case the trac-
semi-intermediate, and semi-minor axes of the tion vector t(n) is not parallel to n (Fig. 6.8b). A
ellipsoid have lengths t1, t2, and t3, respectively. If second graphical construction, called the traction-
the set of traction vectors for surfaces with all pos- director surface, provides a tool for the visualization
sible orientations through the point P are drawn of the surfaces on which the tractions act
with tails at the origin O in traction space, the trac- (Timoshenko and Goodier, 1970). We develop the
tion ellipsoid (6.21) is the locus of points at the relevant equations in three dimensions, but plot
heads of these vectors. Special cases include the the special case (nz 0) in two dimensions for
t1 0.8, t2 0.5, t3 arbitrary

(a) (6.22)
1 tx(n) 0.4,ty(n) 0.433,tz(n) 0
Traction space
0.9
Note that t3 may have any value, but (6.20) requires
0.8
tz(n) 0 for tractions in the plane nz 0. In other
0.7 words the symmetry of the traction variation pre-
Trace of
0.6 cludes any contribution from t3 on surfaces that
t2
traction
0.5 ellipsoid
are parallel to the z-axis.
The three-dimensional surface in physical space
ty(n)
0.4 called the traction-director surface is dened as

0.3 (Timoshenko and Goodier, 1970, p. 222):
)
t(n
0.2 x2 y2 z2
1 (6.23)
0.1 t1 t2 t3
t1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
For tractions scaled as t t/1MPa m2 this is an
ellipsoid drawn in physical space with coordinate
tx(n) axes x, y, and z. The semi-major, semi-intermedi-
ate, and semi-minor axes of the ellipsoid (6.23)
have lengths that are the positive square roots of
(b)
1 t1, t2, and t3, respectively. The trace of the traction-
Physical space n director surface for the two-dimensional example
0.9 using (6.22) is illustrated in Fig. 6.10b. The tangent
0.8 plane to the traction-director surface at an arbi-
trary point (x0, y0, z0) is:
0.7
0.6 (x0, y0, 0)
xx0 yy0 zz0
1 (6.24)
Tangent plane
t1 t2 t3
0.5
y
)
t(n
The so-called normal form of the equation for a

o
0.4
lt
h
lle
plane at a perpendicular distance h from the

ra
0.3
pa
Trace of origin and oriented by the unit normal vector n is

ne
0.2 traction-director (Selby, 1975):

Li
surface
0.1 xnx yny znz
1 (6.25)
0 h h h
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x This plane is parallel to that on which the traction
vector t(n) acts. The tangent plane (6.24) and the
Fig 6.10 (a) Plot of the trace of a traction ellipsoid in first plane dened by (6.25) are identical if:
quadrant of the (tx, ty)-plane in traction space (MPa) with a
particular traction vector, t(n). (b) Plot of the trace of the hx0 hy hz
t1 , t 0, t 0 (6.26)
traction-director surface in the (x, y)-plane of physical space. nx 2 ny 3 nz
The tangent plane at the point of intersection of a line
parallel to t(n) is parallel to the plane on which this traction Substituting these expressions for the magni-
acts. tudes of t1, t2, and t3 into (6.20) we have:
tx(n) hx0, ty(n) hy0, tz(n) hz0 (6.27)

ease of representation. One quadrant of the
[tx(n), ty(n)]-plane in traction space is shown in Fig. In other words the scaled components of the trac-
6.10a with a particular traction vector t(n) and the tion vector t(n) are proportional, respectively, to
cross-sectional trace of the traction ellipsoid. For the coordinates of the point (x0, y0, z0) on the trac-
this example we take the following values (MPa): tion-director surface, so the line drawn from the
6.2 CONCEPT AND ANALYSIS OF STRESS 207
origin through the point (x0, y0, z0) is parallel to enigmatic. This makes it challenging to develop an
this traction vector (Fig. 6.10b). intuitive understanding for stress by simply
The unit normal vector, n, determines the ori- observing or mapping geologic structures.
entation of the tangent plane to the traction- However, certain structures can provide com-
director surface at the point (x0, y0, z0) and this pelling data, because they have a simple geometri-
plane is parallel to the surface on which the trac- cal relationship to some aspect of the stress eld.
tion vector t(n) acts. Given the magnitudes of the Some of the best examples are vertical igneous
special tractions t1, t2, and t3 and the components dikes, formed as magma was injected into frac-
of t(n), the components of n from (6.20) are: tures that tend to be oriented perpendicular to the
tx(n) ty(n) tz(n) direction of least horizontal compressive stress. In
nx , ny , nz (6.28) such a case, a set of curves drawn parallel to the
t1 t2 t3
pattern of dikes provides a map of the orientation
The following generalizations can be made about of the stress trajectories. Here we use a map
the traction vector at a point by studying the trac- pattern of dikes to introduce the concept of stress
tion ellipsoid and the traction-director surface: and to relate this to the traction vector.
Many vertical dikes crop out in the Raton Basin
1. The traction vector t(n) varies continuously in
of southeastern Colorado. The photograph shown
magnitude and direction as the surface on
in Fig. 6.11 includes an outcrop of a large dike
which it acts changes orientation.
trending northward from near the base of West
2. The special tractions t 1, t 2, and t 3 act on three
Spanish Peak. Many of the dikes of this region are
mutually orthogonal planes and have zero tan-
more resistant to erosion than the sedimentary
gential components, so they are perpendicular
host rock and crop out as prominent vertical walls
to the respective plane on which they act.
capping long ridges that form a radial pattern
3. The magnitudes of the tractions t1, t2, and t3 are
about the peak (Johnson, 1961, 1968). The radial
represented by the lengths of the semi-axes of
dikes are classied based on their composition
the traction ellipsoid: they are equivalent to
and Fig. 6.12a shows only those dikes of syenite
the extreme values of the normal component
and syenodiorite composition. The inference is
of the traction at a point.
that rocks of similar composition represent one
4. The traction t(n) is not parallel to the unit
period of magmatic activity and, perhaps, one
normal vector n except on surfaces where this
regional stress eld.
traction is equal to either t 1, t 2, or t 3.
The sedimentary rocks of the Raton Basin (Fig.
Other facts about the traction vector are described 6.12a) form the broad La Veta syncline with a
after we introduce the stress tensor in the next steeply dipping western limb that abuts the older
section. The traction vector provides the link rocks of the Sangre de Cristo Mountains, and a
between the concept of a distributed surface force gently dipping eastern limb that merges laterally
and the stress. It also provides the means to with the sub-horizontal formations of the Great
describe boundary conditions in terms of distrib- Plains. The sedimentary rocks folded into the Le
utions of forces acting on the internal or external Veta syncline are cut by the igneous rock and
surfaces of the material continuum. some deformed and metamorphosed sedimen-
tary rock making up West and East Spanish Peaks.
These impressive mountains rise almost 2 km
6.2 Concept and analysis of stress above the topography of the surrounding plain
and the pattern of igneous dikes seems to radiate
The shape of a deformed fossil (Fig. 5.1), the offset from West Peak. It was this systematic map
of a marker horizon across a fault (Fig. 2.14), and a pattern that led Helmer Od (1957) to propose a
multitude of other geological structures (Ramsay correspondence between the dike pattern and the
and Huber, 1983) provide direct evidence relevant stress distribution at the time of dike formation.
to the kinematics of deformation; however, eld He suggested that the dike pattern should corre-
evidence relevant to the state of stress is more spond to the pattern of stress trajectories.
Fig 6.11 Photograph of East and West Spanish Peaks in

southeastern Colorado with a large radial dike emanating against side b and thereby contributes to the
from West Peak. stretching of the circumference of the chamber as
it expands due to the magma pressure. Cauchys
We infer that a magma chamber at depth relationship (6.9) states that the tractions acting
below West Peak fed the igneous dikes making up on opposite sides of this element are equal in mag-
the radial map pattern. The magma pressure nitude and oppositely directed in the limit as the
acting in this chamber was apparently sufcient element shrinks toward a point:
to fracture the host rock adjacent to the chamber
t(c) t(a)andt(d) t(b) (6.29)
(or at least to open pre-existing fractures) so the
magma could invade these fractures and form the Thus, the magma pressure induces a pair of trac-
dikes. In Fig. 6.12b we show a schematic drawing tions of magnitude t(a) pushing inward and a pair
of one-quarter of a vertical cylindrical magma of tractions of magnitude t(b) pulling outward
chamber in a horizontal cross section below the on element A. The former pair compresses the
current surface and an example of a potential dike element in the radial direction and is used to
path extending outward from the chamber wall. dene a component of compressive stress. The latter
Other idealized chamber shapes are spherical pair extends the element in a circumferential
(Anderson, 1936; Mogi, 1958) or sill-like (Johnson direction and is used to dene a component of
and Pollard, 1973; Pollard and Johnson, 1973; tensile stress. If the outward-directed pair of trac-
Fialko et al., 2001). The magma pressure pushes tions is great enough, the rock will pull apart
outward on the chamber wall and induces a local (fracture) along a line oriented perpendicular to
stress eld in the surrounding host rock. We use these tractions and the magma may invade this
these geological structures to introduce the fracture to initiate dike formation.
concept of stress and the physical quantity called Along the potential dike path (Fig. 6.12b) a
the stress tensor. cubical element B has inward directed tractions
acting parallel to the path and outward-directed
6.2.1 The stress tensor tractions acting perpendicular to the path. The
Consider a small cubical element, A, with one side directions and magnitudes of these tractions
oriented tangential to the wall of the magma change with the distance from the magma
chamber (shown in map view, Fig. 6.12b). A trac- chamber. One special feature of the elements
tion of magnitude t(a) pushes on side a of this along the potential dike path as drawn in Fig.
element with the same force per unit area as the 6.12b is that the tractions act only normal to the
magma pressure, p. The pressure in the chamber element sides. Thus, the compressive and tensile
also induces a traction of magnitude t(b) that pulls stresses are referred to as normal stress components.
(a) y
ne
syncli
Radial dikes 37 30' (a)
85
ta
b
Sangre de Cristo Mountains
7
La Ve
ty(c) t(a) ty(a)

t(c) c C a
East x
West Peak
Peak
d
tx(c) tx(a)
0 5 km
85 Raton Basin syy
Distance
10500' 104 45' (b)
(b) sxx
Magma chamber Host rock
y y
t(a) = p
x
t(d ) C
A x
syx
(c)
t(b) t(c)
sxy
Potential
B dike path
Fig 6.12 (a) Map of radial dike system focused on West

Fig 6.13 Relations among the traction vector components
Spanish Peak, CO (Johnson, 1961). Reprinted from Muller
and stress tensor components. (a) Traction vectors acting on
and Pollard (1977) with permission of Birkhanser-Verlag.
element. (b) Normal stress components. (c) Shear stress
(b) Schematic illustration of magma chamber below West Peak
components.
with trace of radial dike and elements with traction vectors.
If we were to consider an element C, not oriented tx(c) tx(a),ty(c) ty(a) (6.30)

with two sides parallel to a potential dike path, we
would nd that the tractions are oblique to the Because of this special relationship between tx(a)
sides and have both normal and tangential com- and tx(c), a single quantity, xx, is dened as the
ponents. normal component of stress and is represented
Viewing element C in two dimensions in Fig. schematically as two open-headed arrows drawn
6.13a we set up a Cartesian coordinate system perpendicular to sides a and c, with lengths and
with x perpendicular to side a and y perpendicu- directions consistent with the respective traction
lar to side b. The traction vector acting on side a components (Fig. 6.13b). The two subscripts signal
resolves into components, tx(a) and ty(a), and the that this stress component is acting on opposing
traction acting on side c resolves into components, planes with normal vectors that are parallel to the
tx(c) and ty(c). In the limit as the element shrinks to x-axis, and it is acting in an orientation parallel to
a point (6.9) requires these two tractions to be the x-axis. This is referred to as the onin convention
equal in magnitude and oppositely directed, so for subscripts. We draw pairs of stress arrows with
their respective components are: open heads to distinguish stress components from
traction vectors and note that stress component pression is positive and tension is negative. Shear
arrows always are drawn in pairs. The normal stress components directed as shown in Fig. 6.13c
stress component yy is dened using a similar are given negative signs. This choice is motivated
argument. by the fact that the normal stress typically is com-
The combination of the two tangential traction pressive at depth in the Earth, because it is related
components, ty(a) and ty(c), in Fig. 6.13a are used to the weight of the overlying rock. A disincentive
along with the condition (6.30) to dene xy as a for choosing this convention is the awkward
shear component of stress. This quantity is repre- result that components of the displacement
sented schematically as two arrows drawn parallel vector, when related to the stress components, are
to sides a and c, with lengths and directions con- found to be positive in the negative coordinate
sistent with the respective traction components directions. This is at variance with standard prac-
(Fig. 6.13c). The two subscripts are chosen because tice for dening vector components. Students of
this stress component is acting on two opposing structural geology should be able to work with
planes with normals parallel to the x-axis and it is both sign conventions, because both are used
acting in an orientation parallel to the y-axis. The throughout the relevant literature.
arrows have open heads, but only half of the head A simple procedure for remembering how to
is drawn to distinguish the shear stress from the draw all eight arrows representing the stress com-
normal stress. The shear stress component yx is ponents in two dimensions is to consider the
dened using a similar argument. outward normal vector to the plane under con-
There is an apparent contradiction in alge- sideration. If that normal points in the positive
braic signs between traction components and coordinate direction (as on sides a and b, Fig. 6.13),
stress components. For example, one of the arrows a positive stress component arrow should point in
for xx points in the positive x-direction and the the positive coordinate direction. If that normal
other arrow points in the negative x-direction (Fig. points in the negative coordinate direction (sides
6.13b). This stress component is made up of two c and d), then a positive arrow should point in the
traction components that have opposite signs (Fig. negative coordinate direction. This produces a set
6.13a). On the other hand, we need to give a single of arrows consistent with the sign convention for
algebraic sign to the stress component. The tradi- positive stress components (Means, 1976).
tional choice, motivated by considering the sign By extension of the reasoning behind the
conventions for the traction components point- denition of the Cartesian components of stress
ing into the rst quadrant on sides a and b (Fig. in two dimensions (Fig. 6.13), there are nine
6.13a), is to assign positive signs to stress compo- components of stress in three dimensions (Fig.
nents that correspond to these positive traction 6.14), one normal component and two shear
components. This makes a tensile normal stress, rep- components acting on each of the three pairs of
resented by outward directed arrows, a positive opposing sides of the cubical element. Arrows
quantity, whereas a compressive normal stress, rep- representing these nine components are shown
resented by inward directed arrows, is a negative acting on the visible sides of the element. These
quantity. Similarly, the two shear stresses are pos- components are drawn in their positive orienta-
itive as drawn in Fig. 6.13c. This is the convention tions and it is understood that arrows drawn in
used in most of the physics and engineering liter- the opposite directions on the opposing sides of
ature. Because many of the concepts and analysis the element make up the hidden member of
methods we use in structural geology are taken each pair.
from that literature, this is an attractive choice The nine components of stress are not inde-
and we will use it throughout this book, unless pendent. This is understood here by treating the
otherwise stated. cubical element in Fig. 6.14 as nite in size with
Most practitioners of soil and rock mechanics side lengths x, y, and z; considering the state
(Jaeger and Cook, 1979), as well as many structural of stress to be homogeneous throughout; and
geologists and geophysicists, use the opposite con- ignoring the effects of body forces. We imagine
vention for signs of the stress components: com- cutting the element free of the surroundings and
z components. For example, the shear stress xy on

the positive x-side of the element is associated
szz
with a moment about the z-axis (stress multiplied
by area and lever arm) that is (xyyz )(x/2). On the
szy
negative x-side, the shear stress xy is associated
szx syz dz with an exactly equivalent moment for a homo-
geneous stress state. On the positive and negative
syy y y-sides of the element the shear stress yx is asso-
sxz
ciated with moments about the z-axis and these
sxy syx are both equal to (yxxz )(y /2). These are the
only stress components that contribute to the
moment about the z-axis. Therefore the sum of
dx the moments about the z-axis is written:
sxx
x
dy m 2( z
1 1
xyyz)(2 x) 2(yxxz)(2 y) 0
Fig 6.14 Cartesian components of the stress tensor acting (6.33)

on a volume element.
Because the sides of the element are of nite
length, the net moment is zero only if the two
replacing the mechanical action of the sur-
shear stresses are equal. Similar arguments lead
roundings by the appropriate stress components.
to the conclusion that there are only six indepen-
Next we consider the proposition that the net
dent stress components in three dimensions
force and net moment must be zero for static
because the shear stresses are related as:
equilibrium (no linear or angular acceleration).
For example, the stress components on the posi- xy yx, yz zy, zx xz (6.34)
tive x-side of the element are associated with the
following forces (stress multiplied by area) in the This constraint means that the following matrix
x-, y-, and z-directions: representation of the state of stress is symmetric:

xxyz, xyyz, xzyz xx xy xz
yx yy yz (6.35)
on x 12 x (6.31)
zx zy zz
On the negative x-side the equivalent stress com- The normal stress components are placed along
ponents give rise to negative forces of the same the main diagonal in this matrix and the equiva-
magnitudes in the x-, y-, and z-directions: lent shear stress components are placed in sym-
xxyz, xyyz, xzyz metric locations about this diagonal. The rows of
this matrix contain, respectively, the stress com-
on x 12 x (6.32) ponents on the x-, y-, and z-sides of the cubical
element pictured in Fig. 6.14.
Similar arguments for the y- and z-sides demon- Other notations for the stress components are
strate that the forces in the three coordinate direc- found in the literature. For example, the symbol
tions exactly balance because, for a homogeneous may replace for all shear stresses (e.g. xy, yz, zx)
stress state, the corresponding force components to distinguish normal and shear stress compo-
on opposing sides of the element are equal and nents (Timoshenko and Goodier, 1970). Or, the
opposite. The condition of force equilibrium normal stress components may have only one sub-
places no constraints on the stress components. script (e.g. x, y, z) because the subscripts are
However, consideration of moment equi- identical (Jaeger and Cook, 1979). To accommo-
librium does constrain some of the stress date the use of indicial notation, the Cartesian
coordinate axes (x, y, z) may be replaced by (x1, x2, szz

x3) and the stress components are referred to in
aggregate as ij, where it is understood that the z szu szr
ranges of the indices are i 1, 2, 3 and j 1, 2, 3.
The constraint on the shear stress components
(6.34) is expressed using indicial notation as: r
srr
ij ji (6.36) sur suu
sru
This notation also facilitates the use of matrix u srz suz
algebra because the subscripts for the stress com- O x
ponents correspond to the row and column
numbers of the following matrix: Fig 6.15 Cylindrical components of the stress tensor
acting on a volume element.

11 12 13
21 22 23 (6.37)
31 32 33
also have ignored moments due to distributed
Recalling the onin convention for subscripts body or surface couples because most problems in
the rst row contains the stress components structural geology have found satisfactory corre-
on the sides of the cubical element perpendicular spondence to nature without invoking coupled
to the x1-axis and these components are directed, stresses (Malvern, 1969). The constraint on the
respectively, in the x1-, x2-, and x3-directions. shear stress components (6.34) that leads to a sym-
Unlike the displacement, velocity, or accelera- metric matrix of stress components (6.35) is lost in
tion the physical quantity we call the stress is not the presence of coupled stresses.
dened at a point in the continuum by a single Some problems in structural geology can be
vector with three components. In fact it takes six idealized using cylindrical symmetry about an
traction vectors, three of which are independent axis in three dimensions. The map view of the
because of (6.9), acting on the orthogonal sides of dike pattern at Spanish Peaks is a possible
the cubical element to determine the nine stress example (Fig. 6.12a) because the dikes appear to
components. The state of stress at a point in the con- radiate from a point near the center of West
tinuum is completely dened by the nine compo- Spanish Peak. A practical example would be the
nents of (6.35), six of which are independent cylindrical hole cut by a drilling rig to produce
because of (6.34). Stress is referred to as a second- water or hydrocarbons from porous formations at
order tensor to distinguish it from vectors, which are depth. For these and other problems it is useful to
rst-order tensors, and scalars, which are zero- dene the stress components in terms of a cylin-
order tensors. To complete the denition of stress drical coordinate system (Fig. 6.15) which is com-
at a point, one would include the appropriate unit posed of a cylindrical axis, Oz, a perpendicular
of measure, along with the three coordinates of the axis, Ox, a radial distance, r, and a counterclock-
point. If the stress state is a function of time then wise angle, , from the Ox-axis to the radial line.
the denition would include the appropriate time. The six independent stress components (rr,
We reduced the nine stress components to six r, rz, , z, zz) are dened on a small element
by postulating static equilibrium for the nite that has sides parallel to radial lines and concen-
cubical element, no body forces, and a homo- tric circles with centers at the origin. These are
geneous state of stress. In Chapter 7, after intro- called the cylindrical components of stress. Shear
ducing the conservation laws, we show that (6.34) components on adjacent faces must be of equal
is not so restricted, but applies to problems of uid magnitude to prevent angular accelerations:
dynamics in which elements experience both
r r, rz zr, z z (6.38)
linear and angular accelerations, to deformation
in the presence of body forces, and to hetero- Note that the onin convention for subscripts is
geneous stress states in solids and uids. Here we followed: r acts on the sides with normals
(a) z words we consider the most general case of

loading. As a consequence, each traction vector
may have three non-zero components, and each of
n these is equated to one of the nine stress compo-
t(n) nents (Fig. 6.16a). Only one of the two arrows rep-
sxx
resenting each stress component is shown
sxy tz(n)
syx because we have cut away half of the cubical
syy ty(n)
element used to dene the stress components (Fig.
P sxz tx(n) 6.14). On the inclined side of the tetrahedron the
szx y traction t(n) represents the mechanical action of
syz
that removed half and, in general, this vector has
szy
three non-zero components. We consider the lim-
dA iting case in which the length of the longest edge
x szz of the tetrahedron goes to zero so the sides con-
verge on the point P and investigate the relation-
ship among the stress and traction components
(b) tx(n) n
for surfaces through that point.
s
The force components acting on the sides of
y t(n) the tetrahedral element are the products of the
y n
stress or traction components and the respective
ty(n) areas of these sides. To nd the relationship
sxx x
x x
among these components we again invoke
Newtons Second Law, F ma, with the postulate of
sxy
static equilibrium, a 0, so the net force in each
syx coordinate direction is zero. Taking the x-coordi-
nate direction as an example, the sum of the force
syy components in x is written:
Fig 6.16 Relations among traction vector and stress fx [tx(n)]A xxAx yxAy zxAz 0
tensor components. (a) Cauchy tetrahedron with traction (6.39)
t(n) acting on surface with outward unit normal n and stress
Note that the component of force due to the trac-
components acting on coordinate planes. (b) Two-
dimensional relations among traction vector and stress tion component tx(n) is balanced by that due to the
tensor components. stress components acting in the x-direction on the
x-, y-, and z-sides of the element. Rearranging (6.39)
using (6.18), and following similar arguments for
parallel to the radial, or r-direction, and in the the balance of forces in the y- and z-directions, we
circumferential, or -direction. nd three equations that collectively are known
as Cauchys Formula:
6.2.2 Cauchys formula tx(n) xxnx yxny zxnz
We nd important relationships among the trac-
ty(n) xynx yyny zynz (6.40)
tion and stress components by studying the
tz(n) xznx yzny zznz
Cauchy tetrahedron (Fig. 6.16a). Recall that the
inclined side of this element is given an area A Cauchys Formula instructs us that the traction
and the areas of the orthogonal sides (Ax, Ay, Az) vector, t(n), on a surface of any orientation
are found by projection of A onto the coordinate (dened by the outward unit normal vector n)
planes (6.18). Unlike the case illustrated in Fig. through a given point, is completely determined
6.8b, here the traction vectors acting on the by the nine (six independent) components of the
orthogonal sides are not constrained to be paral- stress tensor at that point. We derived Cauchys
lel to the respective normal vectors. In other Formula ignoring body forces and postulating
static equilibrium, but these restrictions can be z

relaxed. After introducing the conservation of
linear momentum in Chapter 7 we use this physi-
cal law alone to derive Cauchys Formula, thereby
y
showing that it is applicable to problems of solid
deformation and uid ow including body forces
x
and accelerations.
Cauchys Formula is expressed using indicial
notation for the Cartesian coordinates xi as: z
ti(n) jinj (6.41) szz
Here it is understood that the repeated index on szy
the right-hand side implies summation over the t(n)
range of j. Given the range of i, (6.41) expands to szx
three equations, one for each component of the
sxz ty(n) tz(n)
traction vector. Using a matrix representation for syy n tx(n)
the components of the traction vector, the stress sxy
tensor, and the unit normal vector, Cauchys y
Formula may be expressed:
sxx Boundary

t1(n) 11 21 31 n1
t2(n) 12 22 32 n2 (6.42) x
t3(n) 13 23 33 n3
Fig 6.17 Volume element with one surface coincident with
Note that the matrix of stress components is the surface of a body. Components of the traction vector, t(n),
transpose of (6.37). The matrix representation acting on this surface are related to certain stress
(6.42) emphasizes the fact that the stress tensor components acting on the element.
can be thought of as a linear operator that gives
the traction vector as a function of the unit
normal vector. The components of stress are con- tx(n) yx xy, ty(n) yy,
sistent with the denition of a second-order tensor
tz(n) yz zy (6.43)
quantity in that they associate a vector (the trac-
tion) with any direction in space as determined by The other components of the stress tensor, namely
the respective direction cosines (components of xx, xz zx, and zz, are not determined by the
the unit normal vector). traction acting on this boundary. These compo-
Cauchys Formula relates the tractions, acting nents may be calculated by solving a boundary
as boundary conditions for models of geologic struc- value problem, but are not given by the boundary
tures, to the stress components on a cubical condition itself.
element adjacent to that boundary. For example, Cauchys Formula given by (6.40) reduces to
consider the body shown in Fig. 6.17 and a small two dimensions for conditions of plane deforma-
element that has one side coincident with the tion. For example, taking the (x, y)-plane as the
boundary of the body. For convenience we choose plane of interest (Fig. 6.16b) we have:
a coordinate system with the y-axis normal to that
nz 0, xz 0 zx, yz 0 zy (6.44)
side, and the other two axes parallel to the edges
of that side. The outward unit normal vector to The two out-of-plane shear stresses must be
the boundary of the body at that point, has com- zero by denition, and the out-of-plane normal
ponents nx 0, ny 1, and nz 0, so the traction stress which generally is not zero is eliminated
and stress components are related according to from (6.40) because the direction cosine, nz,
Cauchys Formula: is zero. The other two direction cosines are
related to one another such that nx cos x and z

ny cos[(/2)x] sin x, so the two-dimensional
form of Cauchys Formula is: |ts|
tx(n) xxcosx yxsinx n
(6.45) u
ty(n) xycosx yysinx
sxx
Here x is the counterclockwise angle measured
syx sxy t(n)
from the Ox-axis to the outward unit normal n for syy
the plane on which t(n) acts. sxz
P y
6.2.3 Normal and shear tractions on a syz szx |tn|
surface szy
For some problems in structural geology it is nec-
szz
essary to calculate the normal and tangential
(shear) traction components acting on an arbi- x
trarily oriented surface within the rock mass as a
function of a homogeneous state of stress. For Fig 6.18 Cauchy tetrahedron with traction vector, t(n),
example, slip on a fault may be thought of as a decomposed into a normal and a shear component.
frictional sliding process that is driven by the
shear component of the traction vector acting on
the fault surface (Wallace, 1951; Bott, 1959; ship v (v n)n n (v n), (Malvern, 1969). This
Morris et al., 1996). To the extent that the rock is used to resolve the traction vector, t, into two
mass on one side of a fault pushes against the vector components that are, respectively, normal
adjacent fault surface the corresponding (nega- and tangential to the surface:
tive) normal component of the traction deters
frictional sliding. Thus, for a set of faults with the t (t n)n n (t n) (6.46)
same frictional strength, those carrying the In order to simplify the presentation in this
greatest shear traction and the least (negative) section we write t(n) as t and understand that this
normal traction would be favored for slip. As a and all other vectors act on the surface with
second example, opening of a joint may be outward unit normal n.
thought of as driven by the pull of the adjacent In the rst term on the right-hand side of (6.46)
rock mass on the prospective fracture surface we use the fact that the component of any vector
(Pollard and Aydin, 1988). For a rock mass that is parallel to a unit vector is given by their scalar
isotropic with respect to tensile strength the product to calculate the normal component of t:
surface with the greatest (positive) normal trac-
tion would be favored for jointing. t n |t||n| cos |t| cos tn (6.47)
The three-dimensional relationship between
Because the range of the angle is 0 , this
the traction vector, t(n), and the unit normal
scalar product may be positive or negative and,
vector, n, on an arbitrarily oriented surface is illus-
correspondingly, the vector t would pull or push
trated in Fig. 6.18. The angle is the smaller angle
on the surface. In other words, the direction of the
between these two vectors in the plane which they
vector component of t that is normal to the
dene (dashed rectangle) with their tails at the
surface is given by [sgn(tn)]n.
point P. The Cauchy tetrahedron is pictured with
For the purpose of computing the normal com-
stress components acting on the coordinate planes
ponent of t recall that a scalar product can be
representing a general state of stress. Any vector, v,
written as the sum of the products of the respec-
may be resolved into two vector components that
tive components:
are, respectively, parallel and perpendicular to an
arbitrary unit normal vector, n, with the relation- t n txnx tyny tznz (6.48)
Cauchys Formula (6.40) is used to write (6.48) in (6.52) by dividing each component of n (t n) by
terms of the stress components: |ts|.
The two-dimensional resolution in the (x, y)-
tn xxn2x yyn2y zzn2z 2xynxny 2yznynz
plane of the x- and y-components of the traction t
2zxnznx (6.49) onto the n- and s-axes can be derived by inspection
Given the state of stress and the orientation of a of Fig. 6.16b:
surface at a point, (6.49) is used to calculate the
tn txcos x tysin x
normal component of the traction vector. (6.54)
ts txsin x tycos x
We use the second term on the right-hand side
of (6.46) to calculate the tangential (shear) compo- Here the (n, s)-coordinate axes are arranged with n
nent of t. The magnitude of the vector t n is the perpendicular and outward from the surface on
magnitude of this component: which t acts and s tangential to this surface. Both
n and s are in the (x, y)-plane and s is directed such
|t n | |t ||n | sin |t | sin |ts | (6.50)
that n is to the right when looking in positive s (a
Given the range 0 , the quantity |t| sin is right-hand rule). Given these two reference direc-
always positive. Having specied only one refer- tions, tn and ts may be either positive or negative.
ence direction, n, we cannot distinguish positive The two-dimensional forms of Cauchys Formula,
and negative signs for the tangential component (6.45), are used to write (6.54) in terms of the stress
of t. However, the vector t n is directed perpen- components:
dicular to the plane dened by t and n, and lies in
tn xx cos 2x yy sin 2x 2xy sin x cos x
the surface on which t acts because it is perpen-
dicular to n. Recall that a vector product may be ts (xx yy) sin x cos x (6.55)
written in terms of the components as: xy( cos 2x sin 2x)
t n (tynz tzny)ex (tznx txnz)ey Given a two-dimensional state of stress in the (y, z)-
(txny tynx)ez (6.51) or the (z, x)-plane, (6.55) may be used with appro-
priate exchange of subscripts.
Similarly, the second vector product of (6.46) may
be written using (6.51) as:
n (t n) [(1 n2x )tx nxnyty nxnztz]ex
6.2.4 Principal values and principal axes
of normal stress
[nxnytx (1 n2y )ty nynztz]ey In Section 6.1.4 we derived the equation for the
[nznxtx nznyty (1 n2z )tz ]ez traction ellipsoid (6.21) by postulating without
(6.52) derivation that three orthogonal surfaces at a
point could be so oriented that the traction vector
This vector is the resolution of the traction t onto acting on each is directed parallel to the respec-
the surface with outward unit normal vector n. tive normal vector. In other words the tangential
Cauchys Formula (6.40) may be used to write component of the traction on each orthogonal
(6.52) in terms of the stress components. The mag- surface is identically zero. The tractions acting on
nitude of the tangential (shear) component of t these three surfaces (t1, t2, and t3) correspond to
may be calculated as |n (t n)| using (6.52) or, the semi-axes of the traction ellipsoid (Fig. 6.9)
noting in Fig. 6.18 that |tn| and |ts| are the and are ordered such that t1 t2 t3. Here we
lengths of the sides of a right triangle and |t| is derive these relationships in terms of the state of
the length of the hypotenuse, one may use the fol- stress at a point by equating the normal compo-
lowing (positive) square root: nent of the traction vector to the normal stress.
|ts | |t |2 |tn |2 We show that the normal stress takes on extreme
(6.53)
values in three orthogonal directions and dene
The direction cosines of the vector component of each of these as a principal normal stress. The prin-
t tangential to the surface are calculated from cipal normal stresses play important roles in
theories of failure, fracture, and faulting of rock, This cubic equation for the unknown normal stress,
topics that we consider in later chapters. nn, has three real roots (Bell, 1920), and these are
We begin by considering an arbitrary state of the three principal stresses 1, 2, 3 which are
stress at a point dened by six independent com- ordered such that 1 2 3. Standard algebraic
ponents and ask if a plane through this point techniques exist to solve such a cubic equation
exists upon which the shear component of the (Selby, 1975) and to show for all possible values of
traction vector vanishes. Referring to Fig. 6.18, the the stress components that the three roots are real.
traction vector, t, would be parallel to the outward Given the values of the principal stresses each
unit normal vector, n, and the only stress compo- may be substituted, successively, for the normal
nent associated with this plane would be the stress, nn, in (6.57) and the three components of
normal stress, nn. The components of the traction the unit normal vector for each principal stress
vector acting on this plane in the arbitrarily axis may be determined (Fig. 6.19a). The three
chosen Cartesian coordinate system would be: components of the unit normal vectors that are
parallel to the principal axes are written:
tx(n) nnnx, ty(n) nnny,
(nx1, ny1, nz1) components of n(1)
tz(n) nnnz (6.56)
(nx2, ny2, nz2) components of n(2) (6.60)
These same traction components are related to (nx3, ny3, nz3) components of n(3)
the stress components through Cauchys Formula.
Substituting the right-hand sides of (6.56) for the Here, for example, the component nz1 is the direc-
traction components in (6.40), and rearranging tion cosine used to project the base vector for the
yields three linear equations for the unknown z-axis onto the axis parallel to the direction of 1.
normal stress, nn, and the three unknown com- To show that the principal axes are orthogonal
ponents of the unit normal vector (nx, ny, nz). We use nn 1 in (6.57) and multiply each equation
include (6.17) to make a set of four equations in by the respective components of n(2), (Jaeger and
four unknowns: Cook, 1979, p. 20):
(xx nn)nx yxny zxnz 0 (xx 1)nx1nx2 yxny1nx2 zxnz1nx2 0

xynx (yy nn)ny zynz 0 xynx1ny2 (yy 1)ny1ny2 zynz1ny2 0 (6.61)
xznx yzny (zz nn)nz 0 (6.57) xznx1nz2 yzny1nz2 (zz 1)nz1nz2 0
(nx)2 (ny)2 (nz)2 1
Next use nn 2 in (6.57) and multiply each equa-
Equations (6.57) have solutions for the compo- tion by the respective components of n(1):
nents of the unit normal vector only if the deter-
(xx 2)nx2nx1 yxny2nx1 zxnz2nx1 0
minant of the coefcients is equal to zero (Gere
and Weaver, 1965): xynx2ny1 (yy 2)ny2ny1 zynz2ny1 0 (6.62)
xznx2nz1 yzny2nz1 (zz 2)nz2nz1 0
| |
xx nn yx zx
Adding the three equations (6.61); then adding
xy yy nn zy 0 (6.58)
the three equations (6.62); and nally subtracting
xz yz xx nn
the second sum from the rst and invoking the
Expanding the determinant and invoking the symmetry condition (6.34) we nd:
symmetry of the stress tensor (6.34) we have: (1 2)(nx1nx2 ny1ny2 nz1nz2) 0 (6.63)
3nn (xx yy zz)nn

2 (
xx yy yy zz
If these two principal stresses are not equal, then
zzxx xy2 yz2 zx2 )nn the second term in parentheses must be zero.
Note that the second term is the scalar product of
(xxyyzz 2xyyzzx xxyz2
the two unit vectors, n(1) and n(2), and from the
yyzx2 zzxy2 ) 0 (6.59) denition of the scalar product (2.19) we have
(a) z (a)
n(3) n(2)
z
0.2
0.1
0
0.1 1
0.2
0.5 0.5
(z, 1) 0
x
y y 0 0.5
0.5 1
n(1) O
(x, 1)
(y, 1) (b)
x
z
z 0.5
(b)
0
0.5
2
s2 1.5
s3 1 1
0.5 0.5
0
0 0.5 x
y 0.5 1
1 1.5
2
s1 y
Fig 6.20 (a) Traction ellipsoid for state of stress with
principal axes parallel to coordinate axes. (b) Traction
ellipsoid for state of stress defined in (6.68) with principal
axes not parallel to coordinate axes.
x
Fig 6.19 (a) Geometric relations among the Cartesian

The stress state 1 1, 2 12, 3 14(MPa) is illus-
coordinates and the unit normal vectors directed parallel to trated in Fig. 6.20a as a traction ellipsoid. In this
the axes of principal stresses. (b) Volume element on which illustration we have chosen xy yz zx 0, so
principal stresses act. the principal axes of stress are aligned with the
coordinate axes, and xx 1, yy 2, zz 3.
If the three principal stresses were equal
n(1) n(2) cos 0. Here is the angle between (1 2 3) the state of stress would be isotropic
the two unit vectors which, in this case, must be and the traction ellipsoid would degenerate to a
/2. A similar derivation for the other two pairs of sphere. In this case any orthogonal set of axes may
principal stresses shows that they are mutually serve as the principal axes. If two of the principal
orthogonal. stresses were equal the traction vectors would
If the Cartesian coordinate system is chosen to trace out an ellipsoid with the one unequal axis
coincide with the axes of principal stress at a being the axis of revolution. In the plane perpen-
point then all the off-diagonal terms in the matrix dicular to the axis of revolution any two orthogo-
of stress components (6.35) are zero and the diag- nal axes may serve as the principal axes and only
onal terms are the principal stresses: the principal axis parallel to the axis of revolution
is unique. Because the three principal stresses are

1 0 0 unequal in this example the traction vectors trace
0 2 0 (6.64) out an ellipsoid with three unequal axes which
0 0 3 are coincident with the coordinate axes (Fig.
6.20a). The complete stress analysis, including three roots that are the so-called eigenvalues or, as
principal stresses, trajectories, and all compo- we refer to them in this context, the principal
nents, is shown in color on the textbook website. values 1, 2, 3. The eigenvalues of an asymmet-
Equations (6.57) may be solved using matrix ric matrix can be real or complex, but the special
algebra and this is a classic problem called an case of a symmetric matrix always yields real
eigenvalue problem, which has many applications eigenvalues (Gere and Weaver, 1965), so the sym-
in the sciences and engineering (Gere and Weaver, metric matrix of stress components always yields
1965). Here we refer to this as a principal value real principal stresses. Real eigenvalues may be
problem since we are looking for the principal positive, negative, or zero and likewise the princi-
values of stress, and (6.57) is written: pal stresses may be positive (tension), negative
(compression), or zero subject only to the con-

xx yx zx nx nx straint that 1 2 3.
xy yy zy ny ny (6.65) Each eigenvalue of a matrix is associated with
nn
xz yz zz nz nz a vector called the eigenvector. For the matrix of
A solution to (6.65) is the vector n with compo- stress components the eigenvectors are the unit
nents (nx, ny, nz), which determines the orientation normal vectors for the planes on which the prin-
of a principal plane with respect to the Cartesian cipal stresses act. Having determined the values of
coordinates (x, y, z). The product of the matrix of the three principal stresses from solving (6.59),
stress components and the unit normal compo- these are substituted separately into (6.67) to
nents is equal to the product of a scalar, nn, and obtain three simultaneous equations to be solved
the unit normal components, where this scalar is for the components of the normal vector n(1), n(2),
the unknown principal stress. Rearranging (6.65) or n(3), associated with each principal stress. In
we have the following homogeneous equation: general, the eigenvectors for a symmetric matrix
are orthogonal to one another (Gere and Weaver,

1965).

xx yx zx 1 0 0 nx 0
xy yy zy nn 0 1 0 ny 0 As an example, consider a point in a contin-
xz yz zz 0 0 1 nz 0 uum where the following symmetric matrix pro-
vides the stress components referred to a specied
(6.66)
Cartesian coordinate system (MPa):
This equation is called homogeneous because the

2 34 1

right-hand side is the null vector. The second xx xy xz 4
3 1
matrix on the left-hand side is called the identity yx yy yz 4 1 2 (6.68)
1 1 1
matrix because multiplication of a matrix of the zx zy zz 4 2 2
same order by it results in the same matrix.

Carrying out the multiplication and subtrac- This state of stress is associated with a traction
tion indicated in (6.66) we have the set of equa- ellipsoid oriented such that none of the ellip-
tions (6.57) in matrix form: soidal axes correspond to the coordinate axes (Fig.
6.20b). The principal values of stress are (MPa):

xx nn yx zx nx 0
1 2.5080, 2 0.8261, 3 0.1659 (6.69)
xy yy nn zy ny 0
xz yz zz nn nz 0 The corresponding components of the principal
vectors are:
(6.67)

One solution to (6.67) is the null vector n, referred nx1 0.8398 nx2 0.5256
to as a trivial solution, but this is of no interest for ny1 0.4930 , ny2 0.6333 ,
the physical problem we are considering. Non- nz1 nz2 0.5681
0.2273
trivial solutions exist only if the determinant of

the coefcient matrix is zero, as stated in (6.58), nx3 0.1361
and this leads to the cubic equation (6.59) in the ny3 0.5965 (6.70)
unknown normal stress, nn. This equation has nz3 0.7909
At the point under consideration we may repre- (a) host rock

sent the state of stress with three lines, oriented 1
1
parallel to the principal vectors (6.70) and scaled magma chamber
to the magnitudes of the principal stresses (6.69). x
For a two-dimensional stress eld, such as
1
that associated with conditions of plane strain
1
(Fig. 6.16), we take the plane of interest as the (x, y)-
plane so the matrix of stress components is: x

xx xy 0 principal
yx yy 0 (6.71) stress
1
0 0 zz trajectory
x
The two in-plane principal stresses are
(Timoshenko and Goodier, 1970):
1 1
2 (xx yy) [4 (xx yy) xy]
2 2 12
(b)
1 1
(6.72)
2 (xx yy) [4 (xx yy) xy]
2 2 12
The third principal stress is zz (xx yy).

Because Poissons ratio has a range 0 12 , it is
not possible to specify which will be the greatest,
1, intermediate, 2, and least principal stress, 3,
before making these calculations.
The direction of greatest in-plane principal
stresses is determined by:

1 2xy
1 tan 1 (6.73)
2 xx yy
The angle 1 is measured from Ox counterclock-

wise to the axis of maximum principal stress.
Fig 6.21 (a) Map of principal stress trajectories near the
Equation (6.73) enables one to calculate the ori-
magma chamber under West Spanish Peak, CO.
entation of the principal stresses at any point in (b) Visualization of three-dimensional stress field near normal
a two-dimensional eld of spatially varying faults in a North Sea hydrocarbon reservoir. Reprinted from
stress, such as that around the model magma Maerten et al. (2002) with permission of Elsevier.
chamber beneath West Spanish Peak at the
time the radial dikes formed (Fig. 6.21a). Because
these orientations vary smoothly from point to
point in the plane, it is possible to construct system of stress surfaces would exist at every
smoothly turning curves that are everywhere point being tangential to the three principal
parallel to one of the local principal stresses. stress axes, but this is not generally true (Treagus
Similarly one can construct a set of curves every- and Lisle, 1997). None-the-less it is always possible
where parallel to the other local principal stress, to determine the principal stress magnitudes and
and these two families of curves are orthogonal orientations at a point and these may be visual-
to one another. These curves are called principal ized using three-dimensional graphical tech-
stress trajectories. niques (Fig. 6.21b). In this gure the pairs of small
One might suppose that the concept of planes intersect along the direction of the inter-
stress trajectories could be extended to a three- mediate principal stress, 2, and the smaller angle
dimensional stress eld such that an orthogonal between these planes is bisected by the least
principal stress, 3 (maximum compressive the orientation of a coordinate system. Rather, the
stress). The triangulated surfaces are model fault constitutive law and strength of a particular ma-
surfaces for a hydrocarbon reservoir in the North terial should be a property of the material itself
Sea and the principal stresses were used to predict and the ambient conditions of temperature and
the location and orientation of small faults in this pressure.
reservoir (Maerten et al., 2002).
The orientation of the Cartesian coordinate 6.2.5 Maximum shear stresses
system in the preceding discussion of principal Given the principal normal stresses (1, 2, 3) and
stresses was arbitrary. Any orientation could be their orientations, one can calculate the variation
chosen and the resulting principal stresses and in the shear traction magnitude, |ts|, with the ori-
principal axes would be identical. In other words entation, n, of the plane on which it acts. We
the three roots of the cubic equation (6.59) are the equate |ts| to the magnitude of the shear stress
same, regardless of coordinate system, so the |ns| and seek orientations of the planes on which
coefcients of this equation must not vary for a the shear stress attains extreme values because
given state of stress. Rewriting this cubic equation these quantities play important roles in rock
by collecting the stress components into constant deformation, particularly faulting. As shown in
coefcients we have: Fig. 6.18 and written in (6.53) the magnitudes of
3nn I1 2nn I2nn I3 0 (6.74) the traction vector, t, and the stresses ns and nn
are related such that:
The invariant coefcients are:
2ns |t |2 2nn (6.77)
I1 xx yy zz
Here we have equated the normal traction com-
I2 xxyy yyzz zzxx 2xy 2yz 2zx
ponent, tn, to the normal stress nn acting on the
I3 xxyyzz 2xyyzzx xxyz2 yyzx2 zzxy2 plane of interest. To write the quantities on the
(6.75) right-hand side of (6.77) in terms of the principal
stresses a Cartesian coordinate system is chosen
These combinations of the Cartesian stress com- with positive x-, y-, and z-axes in the directions of
ponents are referred to as the stress invariants. the unit vectors n(1), n(2), and n(3) respectively
The denitions of the stress invariants in (Fig. 6.22a). Cauchys Formula, (6.40), then reduces
terms of the Cartesian components are reduced to to tx 1nx, ty 2ny, tz 3nz and the squared
denitions in terms of the principal stresses by magnitude of the traction vector is the sum of the
rotating the coordinate system until it aligns with squared components:
the principal axes. This is equivalent to setting the
shear stress components to zero and equating the |t |2 12n2x 22n 2y 32n 2z (6.78)
normal components to the principal stresses: The squared normal stress is taken from (6.49)
I1 1 2 3 with the shear stress components equal to zero
and the normal stress components equal to the
I2 12 23 32 (6.76)
principal stresses:
I3 123
2nn (1n2x 2n2y 3n2z )2 (6.79)
Note in particular that the sum of the three
normal stress components is invariant. This has a Substituting (6.78) and (6.79) into (6.77) we have:
rather simple interpretation as three times the
ns2 12n 2x 22n2y 32 n2z (1n2x 2n2y 3n2z )2
mean normal stress. The other invariants do not
(6.80)
have such simple interpretations but all are
employed in the development of constitutive laws When this equation is expanded there are terms
for isotropic materials and in theories of failure in each of the principal stresses that can be
because, it is argued, such laws and theories rearranged using (6.17) and the following example
should not depend upon an arbitrary choice for (Fung, 1965):
12(n2x n 4x ) 12n2x (1 n2x ) 12nx2(n 2y n 2z ) (6.81) (a) z n(3)

Equation (6.80) then takes the form (Jaeger and
Cook, 1979): s3
2 ( )2n2n2 ( )2n2n2
ns 1 2 x y 2 3 y z
(3 1)2n2z n2x (6.82) s2 n(2)

Note that the shear stress is a function of the prin-
cipal stress differences. y
The goal is to identify the extreme values of
the shear stress as a function of the components
of n. However, these three components are not x s1
independent, so we use (6.17) to eliminate nz from n(1)
(6.82) and write the shear stress as a function of
the two independent variables, nx and ny: (b) z n
2ns (3 1)2n2x (1 n2x ) (2 3)2n2y (1 n2y )
snn
[(1 2)2 (2 3)2 (3 1)2]n2x n2y
(6.83) ay = 45o
s
The maximum and minimum, or stationary,
ns
values of the shear stress are found by taking the y
derivatives of (6.83) with respect to nx and ny:
dns
2ns (3 1)22nx(1 2n2x ) [(1 2)2
dnx x
(2 3)2 (3 1)2]2nxn2y
Fig 6.22 (a) Volume element oriented such that principal
dns (6.84)
stresses act on sides. (b) Volume element rotated such that
2ns ( 2 3 )22n y(1 2ny ) [(1 2)
2 2
dny the maximum shear stress acts on sides.
(2 3)2 (3 1)2]n2x 2ny
The right-hand sides of (6.84) are zero if nx 0 ny, (6.83) such that |ns | 12 |(2 3)|. Substituting
but this refers to a principal plane on which the the components into (6.79) and equating the
shear stress is zero (minimum). normal traction component to the normal stress,
Taking nx 0 the right-hand side of the rst of nn, we nd the magnitude of the normal stress
(6.84) is zero and the second reduces to: acting on these planes is |nn | 12 |(2 3)|. The
steps of this paragraph are repeated taking nx 0
dns
2ns (2 3)22ny(1 2n2y ) 0 (6.85) to nd a second set of components for n and then
dny
the entire analysis is repeated after eliminating nx
For 2 3 0, the second term in parentheses or ny from (6.82) to nd a third set of components.
must be zero, and this requires n2y 12. Using (6.17) All three sets are given in the Table 6.1 along with
we nd n2z 12 and conclude that the components the magnitudes of the maximum shear stresses
of n for the planes carrying this shear stress and the magnitudes of the normal stresses on
are: nx 0, ny 12, and nz 12. These compo- these planes.
nents dene four planes that contain the x-axis The shear stress we have identied as 12 |1 3 |
(the direction of 1) and bisect the y- and z-axes in Table 6.1 always is the greatest in magnitude,
(the directions of 2 and 3), (Fig. 6.22b). The mag- because 1 2 3, but the order of the other
nitude of the shear stress acting on these planes two depends upon the particular values of the prin-
is found by substituting the components into cipal normal stresses. The maximum shear stresses
Table 6.1.Components of n and magnitudes of

(a) |sns|
maximum shear stress and normal stress.
E
nx ny nz |ns| |nn| H
D
0 12 12 1
|2 3 | 1
|2 3 | G
2 2 F I
12 0 12 1
|1 3 | 1
|1 3 | 2az
2 2 2ax
2ay snn
12 12 0 1
|1 2 | 1
|1 2 |
2 2 s3 B s2 C A s1
y s2
act on planes that contain one of the principal s1

normal stress directions and bisect the angle (b) y x
between the other two, so the normals to these z s3
planes make angles of 45o with the principal axes ay
of normal stress (Fig. 6.22). The magnitudes of the
I D
maximum shear stresses are one-half of the differ-
ence between the associated principal normal G
ax
stresses. Recall that the principal normal stresses n
act on planes that carry zero shear stress. In con- F
trast, the maximum shear stresses act on planes x
that have a normal stress equal in magnitude to the
average of the associated principal normal stresses. H
E
z
az
6.2.6 The Mohr diagram: visualizing
Fig 6.23 (a) Mohr space in which the magnitude of the
stress variation for a given stress shear stress is plotted versus the normal stress on surfaces
state with outward unit normal n. (b) Three-dimensional space
The variation of normal and shear stress with the with normal vector n related to angles and points in Mohr
orientation of the boundary on which they act in a space. Reprinted from Jaeger and Cook (1979) with the kind
region of homogeneous stress can be appreciated permission of Mrs. Jennifer D. Cook.
using a graphical construction called the Mohr
diagram. Apparently the German civil engineer
Otto Mohr rst described this graphical constructions act on the faces. These faces carry the
tion in 1882 (Mohr, 1882; Timoshenko and Goodier, maximum, intermediate, and minimum princi-
1970). Since then it has been used in countless text- pal normal stresses ( 1, 2, 3) and the principal
books and journal articles, becoming a standard directions n(1), n(2), n(3) are parallel to the x-, y-,
tool for stress analysis. The need for graphical con- and z-axes respectively (Fig. 6.23 inset). On an arbi-
structions for the analysis of stress variation has trarily oriented plane with outward unit normal
largely been made obsolete by computers. The tech- n the traction vector may be resolved into normal
nique is reviewed here for historical reasons and and shear components (Fig. 6.18) and we consider
to reinforce the intuitive understanding of these the associated normal and shear stress for this
variations. construction. In Mohr space (Fig. 6.23a) one plots
Detailed derivations of the equations behind the magnitude of the shear stress, |ns|, on the
the Mohr diagram are provided elsewhere ordinate and the normal stress, nn, on the
(Malvern, 1969; Jaeger and Cook, 1979), so we only abscissa. Three half-circles are plotted with
describe the graphical result. Consider a cubic centers along the abscissa, and their intersections
element oriented in such a way that no shear trac- with the abscissa correspond to the principal
stresses. The circle with a center at A traces out are entirely to the right of the origin, the normal
coordinate pairs (nn, |ns|) from (1, 0) to ( 2, 0) as stresses on all possible planes are positive so they
the angle 2x varies from 0 to for surfaces that are tensions. If the circles are to the left of the
contain the z-axis and the principal direction n(3) origin, the normal stresses are negative so they
such that z /2 (see Fig. 6.23b). Similar state- are compressions. If the circles straddle the
ments follow for the circles with centres at B and origin, then some planes carry tensile stresses and
C that trace out coordinate pairs from (2, 0) to others carry compressive stresses. The greatest
(3, 0) as 2y varies from 0 to and x /2, and shear stress on a plane that contains the y-axis and
from (3, 0) to (1, 0) as 2z varies from 0 to and n(2) is found on the circle with center at C where
y /2. The center of the circle at C corresponds 2z 90. In other words, in physical space this
to the normal stress (1 3)/2 and the radius of shear stress acts on a plane oriented at 45 to the
that circle is the magnitude of maximum shear principal directions n(1) and n(3).
stress (1 3)/2. The other stresses from Table 6.1
are related to the centers and radii of the circles 6.2.7 Variation of stress components
centered at A and B. with orientation of the coordinate
We now identify where, in Mohr space, coordi- system
nate pairs (nn, |ns|) would plot that act on planes Given the stress components referred to a
oblique to all three coordinate axes. The normal Cartesian coordinate system with a particular ori-
vector, n, to the arbitrarily oriented plane in phys- entation, it is useful to calculate the stress compo-
ical space (Fig. 6.23b) extends from the origin to nents referred to a Cartesian system with another
the perimeter of a unit sphere. The direction orientation. In other words the two coordinate
angles for this vector are x, y, and z. The cone systems are related by a rotation about a common
swept out by rotating about the Ox-axis with con- origin. This procedure is somewhat similar to the
stant angle x intersects the unit sphere along the transformation of coordinates by rotation that was
dashed circle DE. The coordinate pairs (nn, |ns|) described in Chapter 2, but here the equations are
acting on surfaces with these orientations are rep- different because the relative surface areas upon
resented in Mohr space along the dashed circle which the stress components act must be taken
DE with center at B. Similarly, the cone about into consideration (Jaeger and Cook, 1979,
the Oy-axis with constant angle y traces the pp. 245). It should come as no surprise that we
dashed circle FG on the unit sphere and coordi- employ Cauchys Formula in the derivation.
nate pairs (nn, |ns|) acting on these surfaces are The given stress components are referred to
represented in Mohr space along the dashed circle axes of the rst coordinate system (x, y, z) and we
FG with center at C. Finally, the cone for con- seek the corresponding stress components referred
stant angle z traces the dashed circle HI on the to axes of a second coordinate system (x, y, z ).
unit sphere and coordinate pairs (nn, |ns|) are Basis vectors from the common origin, O, and
represented in Mohr space along the dashed circle directed along the positive axes Ox, Oy, and Oz are
HI with center at A. The common intersection used in the transformation equations and these are
of the three dashed circles in Mohr space provides dened using the direction cosines of the angles
the normal stress and the magnitude of the shear between the respective coordinate axes (Fig. 6.24a):
stress on the oblique plane with normal vector n.
ex mxxex myxey mzxez
All possible coordinate pairs fall in the gray region
ey mxyex myyey mzyez (6.86)
between the three half-circles.
ez mxzex myzey mzzez
From the Mohr diagram (Fig. 6.23a) we observe
that the principal stresses act on orthogonal The double subscripts on the direction cosines mij
planes. For example, the point (1, 0) in Mohr refer to the reference axis and the transformed
space is associated with the double angle 2z , axis, respectively. For example, myx is the cosine of
so in physical space we have z /2. Because the the angle (y, x ).
principal stresses plot on the abscissa, the princi- The normal stress component, xx, acts on the
pal planes carry no shear stress. If the Mohr circles plane with normal ex and in a direction parallel to
(a) z xx t(ex) ex
xxm2xx yym2yx zzm2zx 2xymxxmyx
z 2yzmyxmzx 2zxmzxmxx (6.88)
y
ez The shear stress component acting on this plane
ey and in a direction parallel to the y-axis is found by
(z, x) resolving the traction vector t(ex) onto the Oy-axis.
y In other words it is the scalar product of the trac-
ex
x tion vector and the unit normal vector ey. The trac-
O tion vector components are given in (6.87) and the
(x, x) components of ey are given in (6.86) so the shear
(y, x)
x stress is:
xy t(ex) ey
(b) y
xxmxxmxy yymyxmyy zzmzxmzy
y
xy(mxymyx myymxx)
syy
x yz(myymzx mzymyx)
syx
zx(mzymxx mxymzx) (6.89)
ax
sxy Resolving the traction vector t(ex) onto the Oz-
sxx axis determines the other shear stress component
x
acting on this plane.
Following a similar procedure for planes with
normal vectors ey and ez the other four transfor-
mation equations are (Jaeger and Cook, 1979,
pp. 246):
yy xxm2xy yym2yy zzm2zy

Fig 6.24 (a) Transformation of the stress components
from the (x, y, z)-coordinate system to the (x, y, z)- 2xymxymyy 2yzmyymzy 2zxmzymxy
coordinate system. Basis vectors and direction angles are (6.90)
shown. (b) Two-dimensional transformation.
zz xxm2xz yym2yz zzm2zz
this basis vector. This stress component is found by 2xymxzmyz 2yzmyzmzz 2zxmzzmxz
resolving the traction vector t(ex) acting on this (6.91)
plane onto the Ox axis. In other words the normal
stress is the scalar product of the traction vector yz xxmxymxz yymyymyz zzmzymzz
and the basis vector ex. The traction vector compo-
xy(mxymyz mxzmyy)
nents are found from Cauchys Formula, (6.40), by
noting that the normal to the plane of interest is ex: yz(myymzz myzmzy)
zx(mzymxz mzzmxy) (6.92)
tx(ex) xxmxx yxmyx zxmzx
ty(ex) xymxx yymyx zymzx (6.87)
zx xxmxzmxx yymyzmyx zzmzzmzx
tz(ex) xzmxx yzmyx zzmzx
xy(mxzmyx mxxmyz)
Using these components of the traction vector yz(myzmzx myxmzz)
and the components of ex from (6.86), the normal
zx(mzzmxx mzxmxz) (6.93)
stress is:
Although these equations look complex, a few a given point in the body determined by the polar
minutes study reveals patterns that reect the coordinates r and (Fig. 6.15). One can imagine
fact that the three normal components transform sliding the origin of coordinates to the point in
in a similar fashion, as do the three shear compo- question so the geometry is much like that in Fig.
nents. Furthermore the sequence of subscripts is 6.24b where the coordinates (x, y) are oriented in
systematic and repetitive. the same direction as the coordinates (r, ). If we
The transformation equations for stress com- take the angle x and the angle to be equal, the
ponents, (6.88) through (6.93), are more general transformation from Cartesian stress components
than one might suppose. Equations of this same to polar stress components is identical to (6.95)
form are, for example, appropriate for the trans- after substituting for x and making the appro-
formation of the innitesimal strain compo- priate substitutions for the stress components:
nents and other physical quantities (Fung, 1969,
rr xx cos 2x yy sin 2x 2xy sin x cos x
pp. 325). These quantities are collectively known
as tensors of rank two. They are dened by two char- xx sin 2x yy cos 2x 2xy sin x cos x
acteristics: r (xx yy) sin x cos x
xy( cos 2x sin 2x) (6.96)
1. they are composed of nine components in the
(x, y, z)-coordinate system; and The inverse transformation, from polar to
2. they transform to the (x, y, z)-coordinate Cartesian stress components, is found by rst
system following equations of the form pro- exchanging the subscripts in (6.96) and then
vided above. changing the sign of x. In this case only the sign
of terms containing sin x change.
The transformation of stress components in two
dimensions follows from the three-dimensional
expressions given above. For example consider the 6.2.8 An example: stress analysis at the
(x, y)-coordinate system transformed to the (x, y)- grain scale
coordinate system by a rotation about the out-of- The individual sand grains of Fig. 6.6a are in
plane z-axis through an angle x, measured from Ox contact over small areas between the pores, and
to Ox (Fig. 6.24b). The following relationships are the force that one grain exerts on another is
helpful: entirely transmitted through that area of contact.
The relatively great forces and small contact areas
combine to produce signicant concentrations

mxx mxy mxz cos x sin x 0
myx myy myz sin x cos x 0 (6.94) and complex distributions of stress within indi-
mzx mzy mzz 0 0 1 vidual grains. The spatial variation of the
maximum shear stress within grains can be visu-
From these relationships the transformation alized using a technique called photoelasticity
equations for the normal and shear stress compo- (Frocht, 1948). The frontispiece for this chapter is
nents are: a photograph from a photoelastic model experi-
ment that illustrates the shear stress distribution
xx xx cos 2x yy sin 2x 2xy sin x cos x induced in simulated grains of a porous sand-
stone by the forces acting on the contacts (Price,
yy xx sin 2x yy cos 2x 2xy sin x cos x
1966; Gallagher et al., 1974). Each grain in this
xy (xx yy) sin x cos x model is cut from a sheet of transparent and opti-
xy( cos 2x sin 2x) (6.95) cally isotropic material, such as polycarbonate.
The pattern of black bands in the photograph of
Similar equations enable a two-dimensional these grains is equivalent to a contour map of
transformation in the other coordinate planes. the magnitude of the maximum shear stress at
The transformation equations from Cartesian each point.
to polar stress components, or vice versa, follow A somewhat simpler picture emerges if we con-
from those just derived. Consider the stress state at sider only one grain loaded by opposed forces of
6.3 STATE OF STRESS IN THE EARTH 227
(a) P maximum shear stress ranges from zero along the

sides of the disk, to greater values in the interior,
to the greatest value near the points of application
of the applied forces. Because the stress is concen-
trated near these two points, one would infer that
fractures would initiate at these points if the
stress were great enough. Indeed, microscopic
examinations of deformed sandstone often reveal
fractures emanating from grain contacts
(Gallagher et al., 1974) and examples can be seen in
Fig. 6.6a.
Photoelasticity is a useful technique for direct
visualization of part of the stress eld in labora-
tory models of geologic structures. An alternative
is to use mathematical models, based on elasticity
theory (see Chapter 8). For example, Fig. 6.25b is a
contour plot of the maximum shear stress (see
P Table 6.1) for a circular disk of thickness, t, and
(b) diameter, d, loaded by opposing point forces of
1
stress ss magnitude P. The Cartesian stress components are
3 (Frocht, 1948):
0.8
0.6
0.4
2.5 xx
t
2P (R y)x2 (R y)x2 1
r 41

r 42

d (6.97)

2
0.2 2P (R y)3 (R y)3 1
yy (6.98)
0
t r 41 r 42 d
1.5

0.2 2P (R y)2x (R y)2x
xy (6.99)
0.4 1 t r 41 r 42
0.6 Here the terms in the denominators are

0.5
0.8 r 21 x2 (R y)2 and r 22 x2 (R y)2.
There is a remarkable similarity between the
1 1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1
stress eld calculated from the mathematical
Fig 6.25 Elastic models for the stress distribution in a model (6.97) and the stress eld visualized in the
single circular grain subject to point forces. (a) Photoelastic analogous laboratory experiment (Fig. 6.25a).
image of maximum shear stress contours (Frocht, 1948). (b) Many such examples have demonstrated the
Maximum shear stress contours from the two-dimensional efcacy of continuum mechanics for predicting
solution to the elastic boundary value problem. stress variations in solid materials (Frocht, 1948).
Given the remarkable computational power of
modern computers and the availability of analyti-
magnitude P (Fig. 6.25a). Again a laboratory exper- cal and numerical methods to solve problems in
iment using photoelasticity provides an image of elasticity, there is little need to turn to photoelas-
the maximum shear stress distribution (Frocht, tic experiments today.
1948). This could be a model for a single grain of
sand compressed between two other grains by the
weight of the overlying rock. The geometry of 6.3 State of stress in the Earth
the sand grain is idealized as circular and the
mechanical actions of the neighboring grains on In the book The Dynamics of Faulting and Dyke Form-
this grain are approximated as point forces. The ation with Applications to Britain, E. M. Anderson
described a state of stress that provides a good ref- Stress states at greater depths must be extrapo-
erence state for investigations of problems in lated from these data, inferred from studying
structural geology: data recorded on seismographs during earth-
quakes, or calculated from models. Seismic data
It is possible to imagine a condition in which the indicate the orientations of the principal stresses
lateral pressure from all sides increases steadily with
in the vicinity of a signicant earthquake
depth, so as to be everywhere equal to the vertical.
(Engelder, 1993). Typically these events range
This will not often happen in nature, but it forms a
convenient standard of reference, and may be dened
from a few kilometers depth to a few tens of kilo-
as the standard state (Anderson, 1951, pp. 13, 148). meters, so this method extends our knowledge of
the stress state throughout much of the Earths
From this description we understand that the crust. The magnitudes of the stresses are not
normal stress components (pressure) are equal determined by this method, and there can be con-
and the shear stress components are zero. In other siderable uncertainty about the orientations
words this is an isotropic state of stress and the mag- (McKenzie, 1969). None-the-less, these so-called
nitude is determined by the weight of overlying fault-plane solutions have proved to be very valuable
rocks. In this section we dene a state of stress in compiling maps of the principal stress orienta-
that is consistent with Andersons concept and tions (Zoback, 1992).
then describe data from eld measurements that In most in-situ stress measurement data sets
show typical variations from this state with the vertical normal stress ranges from zero at the
depth. Two techniques for measuring the state of surface to about 50 MPa at 2 km depth (Brown and
stress at shallow depths are described and the Hoek, 1978), more or less following a linear distri-
data from such tests are summarized. Finally, we bution (Fig. 6.26a). This is consistent with the ver-
provide examples at both the outcrop and the tical normal stress being related simply to the
crustal scale that illustrate how tectonic states of weight of the overlying rock mass. On the other
stress act to supplement the standard state, and hand, the horizontal components of normal stress
cause different styles of deformation. vary in a less systematic fashion with depth
Can stress be measured in the Earth? The mea- (Fig. 6.26b), possibly reecting differing tectonic
surement is not direct in the sense that one mea- loading conditions. Here, the ratio of horizontal
sures a distance directly with a ruler. Instead, to vertical stress is shown to be widely scattered
calculations using measured values of other phys- near the surface and converging toward values of
ical quantities and/or a model are required. one or less at depths greater than 2 km.
Several techniques for so-called in-situ stress mea- The direction of principal stress and its varia-
surement have been developed and used at expo- tion in map view across continents, plate bound-
sures and in boreholes and mines (Engelder, 1993). aries, and other tectonic features may be compiled
Amadei and Stephansson (1997) describe these from the point measurements. Usually these data
techniques and also document much of the avail- are presented in terms of the direction of the most
able data. These data have been used by mining compressive normal stress acting in the horizontal
engineers in the design of underground openings, plane, near the Earths surface (Fig. 6.27). In this
and by civil engineers in the design of founda- particular gure the authors have focused on
tions for dams and other large construction pro- California and the orientation of this stress near
jects. As structural geologists, our interest in the the San Andreas Fault system (Zoback et al., 1987).
state of stress stems from the fact that the evolu- They nd that the direction of the greatest com-
tion of geologic structures depends upon the tem- pression is remarkably consistent across this
poral and spatial variations of stress. region, being more or less from northeast to south-
Because in-situ stress measurement techniques west. This direction varies from somewhat oblique
require direct access to the rock mass, stress mea- to nearly perpendicular to the trace of the San
surements have only been made at shallow depths Andreas Fault zone. Where the most compressive
in the crust, typically less than a few kilometers. stress is oblique to the fault, one can think of this
(a) 0 illustrated there, the state of stress envisioned by

Anderson is dened as follows:
500
sv = 0.027z xx yy zz g*z
Depth below surface (m)
1000 xy y x y z zy zx x z 0 (6.100)
Fx Fy 0,Fz g*
1500
Here g* is the weight per unit volume of rock.
Australia Positive z is upward from the surface, so the
2000
USA normal stress components are negative (compres-
Canada sive) below the surface, and they increase in mag-
Scandinavia
2500 South Africa nitude linearly with depth for a constant unit
Other regions weight.
3000 Values for the unit weight of rock vary from
0 10 20 30 40 50 60 70
Vertical stress, sv (MPa) about 2.0 104 to 3.5 104 N m3 with a pure
quartzite having a unit weight of 2.65 104 N m3
(b) (Daly et al., 1966). Summarizing many individual
0
measurements of the vertical stress component,
zz, a value of 0.0265 MPa m1 is reported for data
from around the world at depths ranging from
500
100 to 3000 m (McGarr and Gay, 1978), a value of
Depth below surface (m)
0.0285 MPa m1 over the depth range from 0 to

1000 2300 m is reported for the Canadian Shield
(Herget, 1993), and the data shown in Fig. 6.26
K = 1500 + 0.5
1500 z provide an average value of 0.027 MPa m1, again
down to about 3000 m (Brown and Hoek, 1978).
2000 These data are roughly consistent with the rule of
thumb: the vertical compressive stress gradient
100 + 0.3 with depth is about 25 MPa km1. Anderson
2500 K=
z suggested that (6.100) may not be common, but it
is a good place to start an investigation, and it
3000
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 has become known as Andersons standard state
K = sHa/sv (Hafner, 1951). Where measurements of the verti-
cal stress depart from the standard state, there
Fig 6.26 Variation of the stress components to depths of usually are obvious explanations in terms of the
3 km from in-situ measurements. (a) Vertical normal stress. local topography, geological heterogeneities, or
(b) Horizontal normal stress normalized by vertical stress. evidence for tectonic activity (Amadei and
Reprinted from Brown and Hoek (1978) with permission of
Stephansson, 1997).
Elsevier.
In a body of water (or other viscous uid) with
constant density and at rest, the stress compo-
stress as promoting shearing. Where this stress is nents are the same as Andersons standard state
perpendicular to the fault, the cause of shearing is (6.100). This is referred to as a hydrostatic stress or
a puzzle that remains to be resolved. hydrostatic pressure. Hydrostatic is not a very
good term to describe the stress state in a body of
6.3.1 Andersons standard state and rock since the prex hydro implies water. By
variations of stress with depth analogy, however, a state of stress in the Earth that
For the coordinate system shown in Fig. 6.28, and is isotropic and simply proportional to the average
the stress components and body force component rock density, the local acceleration of gravity, and
Idaho
42o
Oregon
Nevada
Kla California
Ca
ma
sc
th
ad
Mo
eM
u
ou
nta
nta
ins
ins
40o
Sie
Sa
rra
n
Ne
Gr
An
vad
ea
dr
t
ea
38o
Ha
s
yw
M
ar
SAN FRANCISCO
ou
d
nt
ai
ns
CALAVERAS Ne
va
Va
Ca d a
ll
lifo
ey
Gr
Sa orio
rn
eg
COALINGA ia
n
Co
as
t
36o
OCK
GARL
Ra
A
ng
SAN LUIS OBISPO

es
Mojave Desert
Ari
zon
B CAJON PASS
a
CARRIZO PLAIN Ran ges Eastern
C Traverse Transverse
SANTA BARBARA Ranges
o Borehole
34 elongation
Focal Sa
Pe
n SA
n
mechanism NEWPORT-INGLEWOOD Ja N
nin
c AN
in
su
Hydraulic to D
R
la
fracturing SAN BERNADINO EA

Ra
S
ng
Geological
es
0 100 200 km SAN DIEGO
32o
124o 122o 120o 118o 116o 114o
D
Fig 6.27 Directions of greatest compressive horizontal
stress from various measurement techniques near the San
0

xx yy zz g dz
(6.101)
Andreas Fault system in California. Reprinted with
permission from Zoback et al. (1987), copyright 1987. xy yx yz zy zx xz 0
Jaeger and Cook refer to this as Heims rule after
depth is referred to as a lithostatic state of stress. For the Swiss geologist Albert Heim (Jaeger and Cook,
a variable density, , and variable gravitational 1979, p. 371). For the investigation of many struc-
acceleration, g, with respect to depth, the stress tural problems, it is not necessary to consider the
components in a body of rock at depth, z D, continuous variations in these quantities. Rather,
may follow the relationship: an average or representative unit weight can be
adopted because the greatest variation of typical z

(a)
unit weights is only a factor of two. However, in
some engineering applications, the level of detail y Traction-free
suggested by (6.101) may be prudent (Amadei and
surface
x
Stephansson, 1997, pp. 415).
The following linear relationships summarize
data on the variations in principal stresses with
depths to 2300 m from the Canadian Shield
(Herget, 1993):
1 1.4 MPa (0.0225 MPa m1)z
2 6.4 MPa (0.0293 MPa m1)z (6.102)
3 12.1 MPa (0.0403 MPa m1)z
szz
Based upon data from Sweden, over the depth
range from 0 to 1000 m, the principal stress mag-
nitudes vary with depth as (Stephansson, 1993): sxx
1 0.8 MPa (0.020 MPa m1)z
syy
2 5.1 MPa (0.029 MPa m1)z (6.103)
3 10.8 MPa (0.037 MPa m1)z sxx = syy = szz = pg*z
Fz = rg*
These best-tting linear relationships dene gra- sxy = syz = szx = 0
dients that range from 0.020 to 0.040 MPa m1.
The stress magnitudes at the surface (z 0) range (b)
z Traction-free
from approximately 0 to greater than 12 MPa, and surface
all the principal stresses are compressive over the x
range of depths. The conclusion that the state of
stress is anisotropic (different principal stress
magnitudes in different directions) should not
come as a surprise to structural geologists, szz
ux = 0,
because the formation of most geologic structures Fz = rg* uy = 0,
requires an anisotropic stress state. tz = 0
sxx
Because the surface of the Earth is essentially
free of shear tractions (an exception being shear
induced by wind), one of the principal stress direc-
tions must be normal to the surface, and this prin- Perfect confinement
cipal stress must be zero in magnitude. For the
Canadian Shield data set the calculated value is
1.4 MPa and for the Sweden data set it is ux = 0, uy = 0, uz = 0
0.8 MPa. This is probably indicative of the error
Fig 6.28 Models for the state of stress variation with
introduced by tting a linear relationship to scat-
depth. (a) Andersons standard state (Anderson, 1951).
tered data. The two horizontal principal stresses
(b) The state of perfect confinement.
are not constrained to be zero at the surface and,
indeed, they can take on surprisingly great mag-
nitudes in compression. For example, 3 at the are vertical and horizontal at shallow depths is
surface for the Canadian data is 12 MPa, roughly supported by the results of many analyses of earth-
equivalent to the compressive stress under a quake focal mechanisms from around the globe
column of rock 500 m high. (Zoback et al., 1989; Engelder, 1993). Stress data
The concept that the principal stress directions from mines in Canada and South Africa indicate
that the divergence in orientation of the most (a) 165 observations

steeply inclined principal stress can be up to 30
from vertical (McGarr and Gay, 1978; Herget, 1993).
The orientation data from the Canadian Shield are
presented as contour plots of 165 observations on
lower hemisphere stereonets (Fig. 6.29). There is a
tight clustering of data for the greatest principal
stress, 1, about the vertical axis (plunge of 90), but
some measurements plunge as shallowly as 60.
The other two principal stress orientations are
widely scattered in azimuth, but the intermediate
5%
2% per 1% area
principal stress is approximately northwest
1%
southeast and the minimum principal stress
(greatest compression) is approximately north- (b)
eastsouthwest.
Despite the general tendency for the principal
stresses to be vertical and horizontal, and for their
magnitudes to increase linearly with depth,
exceptions may occur, particularly near the
surface of the Earth, and in the presence of
signicant topographic variations. Figure 6.30
illustrates the spatial variations in the magni-
tudes of the three stress components in the (x, y)- (c)
plane under a long symmetric (two-dimensional)
ridge (Savage et al., 1985). Note that the y-axis is
vertical and positive upward, and that both axes
are scaled by the height of the ridge, b, above the
origin. The stress components are scaled by g*b,
the expected magnitude of the principal stresses
at a depth equal to the height of the ridge, accord-
ing to Andersons standard state. In this model,
the properties of the rock are assumed to be
8%
homogeneous, isotropic, and elastic. The loading 6%
of this ridge is entirely due to gravity, but others per 1% area
3%
report examples that include the effects of a hori- 1%
zontal tectonic compression (Savage and Swolfs,
1986; Pan et al., 1995). Fig 6.29 Orientations of principal stresses from the
For Andersons standard state and no topogra- Canadian Shield. (a) Minimum principal stress (greatest
phy, contours of the horizontal component of compression). (b) Intermediate principal stress. (c) Maximum
principal stress (least compression). Reprinted from Herget
normal stress, xx/g*b, would be equally spaced
(1993) with permission of Elsevier.
horizontal lines, with the 0-contour at the trac-
tion-free surface. Those contours under the sym-
metric ridge (Fig. 6.30a) are signicantly perturbed 6.30b) and are simpler than contours for the hori-
from this simple pattern. At depths below the zontal normal stress. At depths greater than the
origin that are about equivalent to the ridge height of the ridge, the contours of vertical
height the contour pattern simplies to sub- normal stress show only minor perturbations due
horizontal lines of approximately equal spacing. to the topography. Note, however, that the magni-
Contours of the vertical normal stress, yy/g*b, tudes of the two normal stress components at
mimic the shape of the topographic surface (Fig. these depths are signicantly different, with the
(a) 1 (b)
Sxx /Rg*b 1
b Syy /Rg*b
b
0 0
1 1
y/b
y/b
2 2
3 3
4 4
0 1 2 3 4 0 1 2 3 4
x/b x/b
(c) 1
b Sxy /Rg*b contours of shear stress are quite complex in
pattern, but a 0-contour extends from the ridge
0 top vertically downward, indicating that the prin-
cipal stress directions are vertical and horizontal
1 along that symmetry line. The small values of
shear stress relative to the two normal stresses
y/b
imply that the principal stress directions are

2 approximately parallel to the coordinate axes.
0.06
The boundary conditions on the sides of the

0.00
model (Fig. 6.30) are a zero horizontal displace-

3 ment. This horizontal constraint results in compres-
sive horizontal stresses throughout the model,
with a magnitude that depends upon the nature
4
0 1 2 3 4 of the constraint and the material properties. To
x/b quantify the stress state under conditions of a hor-
izontal constraint, consider a model with a trac-
Fig 6.30 Elastic model for the variation in stress tion-free upper surface and no topography (Fig.
components due to gravity under a long symmetric ridge. 6.28b), and composed of an elastic material,
Components are normalized by g*b, where b is the ridge
loaded only by gravity. The lateral edges of this
height. (a) Horizontal normal stress. (b) Vertical normal
model are not constrained in the vertical direc-
stress. (c) Shear stress. Reprinted from Savage et al. (1985)
with permission of Elsevier. tion, so tz 0 there. However, they are con-
strained to have zero horizontal displacements, so
ux 0 uy. The cylindrical rollers between the
horizontal component being less than the vertical rigid platen and the model are meant to imply
component. This is caused by the remote bound- traction-free vertical motion, but no horizontal
ary conditions of no horizontal displacement. displacements. A similar picture would illustrate
Contours of the shear stress, xy/g*b, would not the conditions in the (y, z)-plane. Under gravita-
appear on a plot representing Andersons standard tional loading the model contracts vertically and
state because that component would be zero every- the lateral constraints induce horizontal com-
where. For the symmetric ridge (Fig. 6.30c) the pressive stresses. The stress state at any depth
(z 0), along with the gravitational force compo- (a) Vertical section
nents per unit volume, Fi, are given by:
Fluid injection

xx yy ; zz g*z
1 zz
xy yx yz zy zx xz 0 (6.104)
Pressure Hydraulic
Fx Fy 0, Fz g* D
transducer fracture
This state of stress is referred to as a state of perfect
connement.
The relationship between the vertical and hori-
zontal components of stress (6.104) depends upon
Packer
an elastic property called Poissons ratio, . For this
discussion it is sufcient to understand that SV
Poissons ratio is a dimensionless number with
values that range between 0 and 1/2. For 0.25, the SH
value used for the models illustrated in Fig. 6.30, the
horizontal stress is one-third of the vertical stress
according to (6.104). For 1/3, the horizontal stress
is one-half of the vertical stress, and as goes to 1/2, Sh
the vertical and horizontal stress components (b) Horizontal section SH
become equal. Thus, Andersons standard state is
consistent with an elastic, perfectly constrained
rock mass only if Poissons ratio is equal to 1/2.
Wellbore
u
6.3.2 Measurement of in-situ stress:
hydraulic fractures and wellbore
breakouts Hydraulic
Two common methods for stress measurement at
fracture
depth in the Earths crust involve data taken from
wellbores. The rst method is based upon per-
turbing the local state of stress near a wellbore by
increasing the internal uid pressure until the Fig 6.31 Schematic illustrations of hydraulic fracture
wall of the wellbore fractures (Fig. 6.31). Because generation from a wellbore. (a) Vertical cross section in the
the fracture is induced by uid pressure, this is plane of the fracture and containing the wellbore.
(b) Horizontal cross section through the fracture and
referred to as the hydraulic fracturing method. The
wellbore. State of stress in absence of fracture and wellbore
objective is to determine the magnitudes and ori-
is (SV, SH, Sh).
entations of the three principal stresses at the site
of the measurement, so this is referred to as the in-
situ stress. Here we introduce the elementary con- surface, it is presumed that one principal stress is
cepts and theory behind these tests. Amadei and vertical. The magnitude of this compressive stress
Stephansson describe hydraulic methods for is called SV, and it is aligned with the vertical axis
stress determination in more detail, and evaluate of the wellbore (Fig. 6.31a). Furthermore, it is pre-
more general conditions for these tests (Amadei sumed that the magnitude of this stress is deter-
and Stephansson, 1997). Other techniques for esti- mined by the average unit weight, g*, of the
mating stress in the Earths crust involve the overlying rock and the depth, D:
interpretation of earthquake data (Hanks, 1977;
SV g*D (6.105)
Scholz, 1990) and the interpretation of geological
structures (Zoback et al., 1989; Zoback, 1992). Given SV, the problem is reduced to nding the
Based on proximity of the Earths traction-free magnitudes of the greater and lesser principal
Pc = Fracture initiation pressure this record the subsequent gradually declining

pressure occurs as the injected uid ows into the
Pr = Fracture re-opening pressure
Shut in fracture and the fracture tip propagates away
Shut in
Pressure
Ps = Shut in from the wellbore. A sharp drop in pressure

pressure Ps = Shut in pressure
occurs when pumping ceases and further ow
Po = Formation
pore pressure into or out of the section is prevented by closing
Cycle 1 Cycle 2 the appropriate valves. This is the so-called shut-in
phase, and the gradually declining pressure after
shut-in is associated with leakage of uid from the
Flow rate
section or the fracture into the adjacent rock. The

decrease in slope of the pressure versus time
Time record, shortly after shut-in, is interpreted as the
closing of the fracture, because this would curtail
Fig 6.32 Plots of pressure versus time and flow rate
versus time for hydraulic fracturing. Pressures used to infer
signicant leakage from the fracture walls and
the state of stress are indicated. Reprinted from Enever et al. thus tend to stabilize the pressure. This so-called
(1992) with permission of Elsevier. shut-in pressure, Ps, may be difcult to detect, but it
plays a crucial role in the determination of the
stress state. On the next cycle of uid injection
stresses in the horizontal plane, SH and Sh, respec- into the packed off section of the wellbore, the re-
tively, and their orientation (azimuth), (Fig. 6.31b). opening pressure, Pr, is interpreted as that necessary
It is not known a priori whether SV is the least, to open the fracture at the wellbore. The decrease
intermediate, or greatest principal stress. It is gen- in slope of the pressure versus time record reects
erally presumed that all of these stresses are com- the enhanced ow into the fracture. The second
pressive, so SH is the greatest compressive stress and subsequent injections provide additional
and Sh is the least compressive stress in the hori- measures of the shut-in pressure.
zontal plane. The pressure record from the The various pressures recorded during the
hydraulic fracturing test is used to determine the hydraulic fracturing procedure are used to esti-
magnitudes of these stresses. Their orientation is mate the stress state in the horizontal plane
determined by assuming that the least compres- (Fig. 6.32b) by employing models that relate these
sive stress, Sh, is perpendicular to the fracture, and physical quantities (Hubbert and Willis, 1957;
the fracture orientation is detected in the wellbore Scheidegger, 1962; Fairhurst, 1964; Haimson and
using devices called impression packers or bore- Fairhurst, 1967). The model reviewed here is based
hole televiewers (Amadei and Stephansson, 1997). on elasticity theory. The rock surrounding the
To carry out the hydraulic fracturing proce- borehole is postulated to behave as a homoge-
dure a short section of the well is sealed off from neous and isotropic solid with respect to its elastic
the uid pressure above and below with so-called properties. More complete models would include
straddle packers, inatable rubber tubes that the effects of uids, present in the rock before the
press against the sides of the wellbore (Fig. 6.31a). procedure and leaking into the pores of the sur-
The uid pressure is monitored with an elec- rounding rock from the wellbore and the fracture.
tronic pressure transducer, and it is controlled by The model presented here is based upon the solu-
pumping uid into the section between the strad- tion for the two-dimensional stress state around a
dle packers at a constant rate. Schematic pressure circular hole loaded by a uniform remote stress.
versus time and uid ow versus time records are This solution apparently was derived rst by G.
shown in Fig. 6.32, with different characteristic Kirsh and published in 1898, and it has been
pressures identied during two cycles of uid veried in numerous laboratory studies by direct
injection (Enever et al., 1992). The breakdown pres- strain measurement and photoelastic investiga-
sure, Pc, is the greatest pressure recorded on the tions (Timoshenko and Goodier, 1970, p. 90).
rst cycle of injection, and this is interpreted as The geometry of the problem consists of a cir-
the pressure at which the fracturing initiates. In cular hole of radius R and a polar coordinate
Linear sur sr u problem consists of equations for the three polar

elastic suu srr
stress components (Fig. 6.33) everywhere in the
solid r horizontal plane that match the conditions
s specied above at the edge of the hole and at great
t(n) = P
distances from the hole. The functions that solve
u Sh
this problem are:
n
SH

O x 1 R 2
R 2
rr (SH Sh) 1 P
2 r r
Circular Remote stress

2 4
1 R R
R hole r >> R, u = 0o (SH Sh) 1 4 3 cos 2
2 r r
Fig 6.33 Geometry and stress components for the Kirsh 6.108)
solution to the elastic boundary value problem of a circular

hole in an infinite body (Jaeger and Cook, 1979). 1 R 2
R 4
r (SH Sh) 1 2 3 sin 2
2 r r
(6.109)
system with origin at the center of the hole (Fig.
6.33). For the sake of mathematical convenience

2 2
1 R R
the elastic body extends to an innite distance (SH Sh) 1 P
2 r r
from the hole, but the solution approximates

4
nite bodies that extend to distances that are 1 R
(SH Sh) 1 3 cos 2 (6.110)
great compared to R. The internal boundary con- 2 r
ditions for this problem are specied in terms of
the traction acting on the edge of the hole: According to Jaeger and Cook this is, perhaps, the
most important solution for the discipline of rock
BC: on r R,tn P,ts 0 (6.106) mechanics (Jaeger and Cook, 1979, pp. 249), and it
has been used for a number of important applica-
This traction presses against the hole boundary tions in structural geology as well. These rather
with a magnitude equal to P, and has no shear complicated looking equations can be reduced to
component, so this is equivalent to a static uid some simple relationships between the stress
pressure. The remote boundary conditions are components and the pressure in the hole by ana-
written in terms of the stress components decay- lyzing how the stresses are distributed about the
ing to some uniform values at innite distances: hole.
We learn from the Kirsh solution how the
stresses are distributed with radial distance from

rr 12 (SH Sh) 12 (SH Sh)cos 2 the hole. For example, in Fig. 6.34a we plot the
r
BC: as , r 21 (SH Sh)sin12
1
radial and circumferential stress components for
R 2 (SH Sh) 2 (SH Sh)cos 2 the case of uniaxial remote compression. These
stress components are normalized by the magni-
(6.107)
tude of the remote stress, SH, and are plotted as
For 0 the stress components are rr SH, r they are distributed along the radial line, 0o,
0, and Sh. In other words this is a biaxial from the edge of the hole, r/R 1, to a distance r/R
state of compressive stress with the radial compo- 5. The shear stress is zero along this line of sym-
nent being the greatest compression, SH, and the metry. For uniaxial compression the radial stress
circumferential component being the least com- is zero at the edge of the hole, it increases slightly
pression, Sh. SH and Sh are the magnitudes of the to a tensile maximum, and then steadily
remote principal stresses, because the shear stress decreases toward the remote compressive value
is zero for this orientation. rr/SH 1. The circumferential stress at the edge
A solution to this elastic boundary value of the hole is tensile and of the same magnitude
as the remote stress, /SH 1. The circumfer- (a)

ential stress decreases rapidly away from the hole 1
At r/R = : srr = SH = 1
to a compressive minimum, and then increases 0.8
Normalized stress component

suu = 0 = sru
slowly toward zero with greater distance from the 0.6
At r/R = 1: srr = 0 = sru
hole. 0.4
suu /SH
The greatest stress perturbations occur at or 0.2
sru /SH
near the edge of the hole. Based on theoretical 0
models (Hubbert and Willis, 1957) and laboratory 0.2
experiments (Haimson, 1968), researchers have 0.4
srr /SH
argued that opening fractures, not shear fractures
0.6
are the common result of increased uid pressure
0.8
in wellbores. These fractures typically trend along
1
the axis of vertical wellbores, so it is the circum- 1 1.5 2 2.5 3 3.5 4 4.5 5
ferential component of stress that would act to r /R , u = 0 o
(b)
initiate and open the fracture. Therefore, we focus
1
attention on the value of the circumferential
stress at the edge of the hole in the elastic model. Normalized stress component 0.5
srr /SH = 0 = sru /SH
For the case of internal pressure, the circumfer- 0
ential stress is the same everywhere around the
0.5
hole and is equal in magnitude to the applied
pressure, /P 1. However, for the case of a uni- 1 suu /SH
axial remote compression the circumferential 1.5
stress varies systematically from a tension at
0, where /SH 1, to a compression at 2 At r/R = : srr = SH = 1
90, where /SH 3 (Fig. 6.34b). That is, the suu = 0 = sru
2.5
At r/R = 1: srr = 0 = sru
remote compression induces a local tension along
3
the edge of the hole that is oriented perpendicu- 0 10 20 30 40 50 60 70 80 90
lar to the applied stress, and it induces a local u (o), r/R = 1
compression along the edge of the hole that is ori-
Fig 6.34 Plots of the polar stress components for the
ented parallel to the applied stress. Here we focus circular hole problem (Jaeger and Cook, 1979). (a)
on the local tensile stress. The greatest tensile Components versus distance from the hole edge for uniaxial
stress occurs at r R and 0. From symmetry, remote compression. (b) Components versus position on the
this same stress is induced at 180. Using these hole edge for uniaxial remote compression.
conditions in (6.108) through (6.110) we nd the
stress state at these points is:
essary to do this is referred to as the breakdown

r r P pressure, Pc:
r
At 1 and 0, : r 0
R For T, P Pc 3Sh SH T (6.113)
P SH 3Sh
(6.111) By measuring the breakdown pressure during the
hydraulic fracturing procedure, and by measur-
For a given state of remote biaxial compres- ing the tensile strength in the laboratory for a
sion, the pressures necessary to induce a circum- sample of the formation being fractured, two of
ferential tension at these points are: the four quantities in this equation can be deter-
mined (Scheidegger, 1962).
P 3Sh SH (6.112)
Because the appropriate tensile strength is
To initiate an opening fracture in otherwise that for the rock at the in-situ conditions of stress,
unfractured rock, the local tensile stress must temperature, etc., and because these conditions
equal the tensile strength, T, and the pressure nec- may be difcult to reproduce in the laboratory, an
alternative method usually is employed (Brede- ular to the direction of Sh, so the fracture orienta-
hoeft et al., 1976). During the second (and subse- tion determines the orientation of SH and Sh. The
quent) cycles of uid injection, the pressure other principal stress is presumed to be vertical.
versus time record is monitored and then ana- Alternatively, one can use the breakdown pres-
lyzed carefully to identify a change in slope of the sure, Pc, and a measure of the tensile strength, T,
rising pressure that signals the reopening of the in the determination of SH.
fracture. The elastic model suggests that the re- The second wellbore procedure provides a
opening pressure, Pr, is that necessary to increase direct determination of the orientation of the in-
the circumferential stress just slightly above zero: situ stresses, SH and Sh. As with the hydraulic frac-
turing method, the simplest interpretations
For 0,P Pr 3Sh SH (6.114)
depend upon a vertical wellbore and the pre-
This interpretation of the pressure versus time sumption that one of the principal stresses is ver-
record implies that the in-situ tensile strength is: tical. A variety of instruments, including the
borehole camera, dipmeter, acoustic televiewer,
T P c Pr (6.115)
and electrical resistance microscanner are capa-
The nal measure of pressure used to estimate ble of measuring the shape of the wellbore
the in-situ stresses is that just sufcient to hold the (Amadei and Stephansson, 1997, p. 308). Although
developed fracture open against the least com- the drilling bit is designed to cut a cylindrical
pressive horizontal stress, Sh. The relationship hole, the records from these instruments demon-
between this pressure and stress is found from a strate that sections of some wellbores are not
different solution to the elastic boundary value cylindrical, but instead have systematic increases
problem, because the geometry now is that of a in radii along two diametrically opposed zones
crack, not a circular hole. The opening, u, of a (Fig. 6.35). These zones are referred to as wellbore
crack in an elastic body is proportional to the dif- breakouts because it is inferred that the hole was
ference between the internal pressure, P, that enlarged by the breakage of rock, due to a local
forces the walls apart, and the remote compres- stress concentration, and the subsequent spalling
sive stress, here taken as Sh, that pushes them of the rock fragments into the wellbore.
together (Pollard and Segall, 1987): The geometry of the zones of broken rock
associated with a wellbore breakout suggest that
u P Sh (6.116)
these are not a result of a single fracture extend-
Thus, the pressure and the stress must be just ing perpendicular to the wellbore, as in the
about equal as the crack starts to open, or as it hydraulic fracturing procedure (Fig. 6.31). Instead,
closes. Just as the fracture surfaces come together, it has been proposed that a set of shear fractures
a second knee in the pressure record indicates the oriented oblique to the wellbore (Fig. 6.35), or a set
shut-in pressure, Ps, and this is interpreted as equal of opening fractures oriented parallel to the well-
in magnitude to the least compressive stress: bore, is responsible for the fragmentation of the
rock (Zoback, 1985; Zheng et al., 1989). In either
For u 0,P Ps Sh (6.117)
case the stress concentration induced by drilling
Given the depth of overburden, D, and its the hole is held responsible for the fracturing.
average unit weight, g*, and the two pressures, Ps Once the rock is fractured, ow of the drilling
and Pr, read from the pressure versus time record, uid carries the fragments away, leaving the open
one can calculate all three in-situ principal stresses breakout.
as: The Kirsh solution for the elastic boundary
value problem of a circular hole subject to inter-
SV g*D
nal pressure and remote biaxial compressive
S h Ps (6.118)
stresses provides the equations necessary for an
SH 3Ps Pc T 3Ps Pr
elementary analysis of the stress concentration
This interpretation presumes that a vertical frac- that may cause breakouts. To assure that
ture has propagated in a plane that is perpendic- hydraulic fractures have not initiated, and that
For the sake of an example we choose the loading

Sh conditions SH/P 2 and Sh/P 1, and note that the
Wellbore circumferential stress would be four times greater
SH
breakout than the radial stress. This stress difference could
be responsible for the fracturing that creates the
r>R breakout zone.
Remote stress One of the more extensive demonstrations of
r >> R the utility of wellbore breakouts for determining
r=R the orientation of the horizontal principal
stresses came from a study in the Western
Canadian Basin (Bell and Babcock, 1986). A map
from this study (Fig. 6.36) shows the locations of
154 wells throughout the basin, and at each loca-
tion the average direction of the wellbore break-
outs is indicated by a short line. The solid lines
indicate the average of the dominant population
of breakouts whereas the dashed lines represent
minor populations. Note the length scale on the
map and the systematic nature of the breakout
orientations. Given the interpretation described
above, that the breakouts align with the direc-
Shear fractures tion of Sh, one would conclude that the greatest
horizontal compressive stress, SH, is dominantly
oriented along a northeastsouthwest trend, per-
pendicular to the boundary between the Rocky
Mountains and the Canadian Basin. This system-
atic trend continues for over 1000 km along the
Opening fractures range front.
Fig 6.35 Schematic illustration of wellbore breakouts.

Inferred directions of greatest, SH, and least, Sh, horizontal
6.3.3 Dike pattern at the Spanish Peaks
compressive stress remote from the wellbore are shown.
Ernest M. Anderson hypothesized a mechanical
Insets show breakouts generated from shear fractures and relationship between igneous dikes and the state
from opening fractures. of stress in Earths crust: dikes are intruded per-
pendicular to the least compressive principal
stress (Anderson, 1972). Anderson recognized that
existing fractures have not opened due to the uid
igneous dikes are a form of opening fracture
pressure in the wellbore, consider cases where:
driven by the competition between the internal
P 3Sh SH (6.119) uid pressure and the remote compressive stress
acting across the dike plane. He suggested that the
According to our previous analysis of the Kirsh
most favorable orientation for the dike plane is
solution, the greatest stress concentration is at
perpendicular to 1, the least compressive stress.
the hole boundary where the boundary is ori-
In a region of the crust subject to a homogeneous
ented parallel to SH. The stress state at these two
stress eld one would expect dikes to be planar
points is:
sub-parallel structures. On the other hand if the
stress eld exhibits spatial or temporal variations

rr P
r 3 one would expect individual dikes to be curved,
At 1 and , : r 0
R 2 2 reecting propagation along a curved stress tra-
P 3SH Sh
jectory. The radial dike set in the Spanish Peaks
(6.120) region (Fig. 6.12a) presumably reects variations
0 100
RO km
CK
Y
M
O
UN
TA
IN
CA
S
NA
DI
AN
BA
SI
N
LEGEND
Well location number
Major population
Minor population
108o
116o 112o
Fig 6.36 Map of western Canada with directions of least (SH Sh)d2
0.5 (6.121)
horizontal compressive stress inferred from wellbore Pr2o
breakouts. Reprinted from Bell and Babcock (1986) with
For the model result shown in Fig. 6.37 the orien-
permission of Elsevier.
tation of the greatest regional compression is 082o
and the estimated difference between the princi-
in the orientation of the stress eld, so these vari- pal values is (SH Sh) 0.03P. Given an estimated
ations can be deduced from the pattern of dikes. upper bound for the magma pressure of
The traces of dikes are correlated (Fig. 6.37) to the P 100 MPa, based on the strength of rock sur-
trajectories of maximum compressive stress deter- rounding the magma chamber, the regional stress
mined from the solution to a boundary value difference is less than 5 MPa. For comparison an
problem of elasticity theory (Pollard and Muller, average value of 2.1 MPa is reported for 26 mea-
1976; Muller and Pollard, 1977). For comparison surements of horizontal principal stress differ-
purposes the trajectories and dike traces are pre- ence from seven wells in the Piceance Basin of
sented in alternating vertical strips across the Colorado using the hydraulic fracturing tech-
map. Here the correlation is surprisingly good nique (Bredehoeft et al., 1976).
and there is a unique relationship among the The analysis used to estimate the regional
values for the remote principal horizontal stress stress difference is based upon the solution for an
components, SH and Sh, the magma pressure, P, elastic body with a pressurized hole (the magma
the radius of the magma chamber, ro, and the dis- chamber), a rigid boundary (the Sangre de Cristo
tance from the magma chamber to the Sangre de range), and a regional stress eld. No account is
Cristo Mountain front, d: made of the perturbation of the stress eld by the
37o 30'
Stress
trajectories
Radial
dikes
0 5 km
Distance
37o 15'
105o 00' 105o 45'
Fig 6.37 Map of Spanish Peaks region of southeastern relationship one can map the chains of cinder
Colorado with traces of dikes (dotted lines) and trajectories
cones just like the exposures of dikes at the
of greatest compressive horizontal stress (short solid lines)
Spanish Peaks and use these data to estimate the
from an elastic solution. Reprinted from Muller and Pollard
(1977) with permission of Birkhanser-Verlag. horizontal principal stress orientations along vol-
canic arcs like the Aleutians (Fig. 6.38). Note that
the estimated motion vectors for the Pacic plate
emplacement of the dikes, yet each dike will relative to the North American plate are sub-par-
perturb the local stress eld and could thereby allel to the maximum horizontal compression
inuence the propagation path of the next dike inferred from the cinder cones. Apparently the
(Meriaux and Lister, 2002). Also, this analysis convergence direction of the two plates is
neglects the inuence of pre-existing fractures, reected in the compression direction within the
which may guide the propogating dike (Ziv and volcanoes near the Earths surface. This informa-
Rubin, 2000), and the effects of topography (Fialko tion is crucial to an understanding of the struc-
and Rubin, 1999). tural and volcanic history of such regions.
Andersons concept has been applied to the
injection of dikes under active volcanoes in the
Aleutian Arc (Nakamura, 1977; Nakamura et al.,
1977). A direct relationship between chains of
cinder cones on large volcanoes and the orienta- In the Newtonian context of rigid-body dynamics
tion of sub-surface dikes was inferred by suggest- it is understood that forces are associated with
ing that dikes provided the conduits for ow of accelerations. On the other hand, in the context of
magma to the surface of the volcano. With this a deformable solid the stress is associated with
1000 km
Fig 6.38 Map of Alaska and Aleutian Arc with directions of

greatest compressive horizontal stress (short solid lines) with ow in a ductile shear zone might ask: what
from chains of cinder cones on volcanic centers and from stress state produced a certain rate of deforma-
active faults (Nakamura, 1977). Reprinted from Nakamura et tion? In this chapter we have described the attrib-
al. (1977) with permission of Birkhanser-Verlag. utes of the traction vector and the stress tensor
and shown how to manipulate these quantities.
Occasionally structural geologists have claimed
strain, and in the context of a owing uid the that the kinematic quantities (displacement,
stress is associated with the rate of deformation. velocity, acceleration, strain, rate of deformation,
The boundary conditions on the solid and on the etc.) are more central to problem solving in struc-
uid may be described in terms of the tractions tural geology and this has led others to claim that
distributed over the bounding surfaces, both the dynamic quantities (force, traction, stress) are
external and internal to the body under investi- more central. Neither claim is defensible because
gation. The structural geologist concerned with both sets of quantities are required to solve prob-
deformation of the brittle crust might ask: what lems. Both appear in the equations of motion
stress drop across a fault produced a certain strain described in Chapter 7 and they are inextricably
measured at Earths surface using geodetic instru- linked to one another by constitutive laws
ments? Or, the structural geologist concerned described in Chapters 8, 10, and 11.
Chapter 7
Conservation of mass and momentum
Structural block diagram of a part of the Penninic Alps, language of mathematics. The goal is to construct a
Switzerland (Argand, 1911).
working model of the universe out of commonplace
notions: ideas of number and order and measures of
At the heart of all this calculation lies the deeply held time and distance. With such a working model, we
conviction that natural phenomena are, in essence, can leap ahead in time and predict what the other-
the consequence of just a small number of physical wise opaque future has in store for us (Peterson,
laws, and that these laws are best expressed in the 1993).
244 CONSERVATION OF MASS AND MOMENTUM
I
n the context of structural geology we can con- moments during rare events such as earthquakes.
struct a working model of mountain building In the absence of appreciable accelerations, con-
from those small number of physical laws and servation of momentum requires that the resul-
then leap backward in time and understand the tant forces and torques acting on a rock mass are
development of geological structures such as those negligible. This condition ensures that the linear
depicted in the structural block diagram (Chapter and angular momenta are nearly constant with
7, frontispiece) of a part of the Penninic Alps of respect to time. These requirements are funda-
Switzerland constructed by Emile Argand and pub- mental and should be examined at an early stage
lished in 1911 (Argand, 1911). This is one of the ear- of any program of modeling geologic structures.
liest published block diagrams in the literature of Our review of concepts introduced in the
structural geology (McIntyre and Weiss, 1956; typical college physics class leaves us short of
Howarth, 1999) and it illustrates what was known having all of the necessary tools of mechanics to
in the early part of the twentieth century about analyze tectonic processes. Given our human time
one of the most interesting and complex regions of scale and the poor resolution of our eyes for dis-
folding and faulting in that mountain chain. cerning small changes in the shapes of objects,
Among the small number of physical laws that rock does seem quite rigid. One might think that
can be employed to understand tectonic processes the dynamics of rigid bodies would be as far as we
and their structural products are those of mass, have to go to tackle problems of rock deformation
momentum, and energy conservation. Newtons in the Earth. Indeed, the early practitioners of
Second Law of Motion, for example, relating force, plate tectonics conceptualized plates of the litho-
mass, and linear acceleration is embodied in a sphere as thin rigid masses slowing moving over a
generalized statement of momentum conserva- mobile aesthenosphere. Similarly, the typical mid-
tion. Kinetics is the branch of mechanics that con- twentieth century structural geology textbook
siders the action of forces and torques on particles introduced students to faulting with diagrams
and rigid bodies, and their resulting accelera- showing rigid blocks moving relative to one
tions. This should be familiar ground for students another (Billings, 1972), and laboratory exercises
recently exposed to the mechanics section of a utilized painted wooden blocks to illustrate the
college physics course. However, we offer a short patterns of offset strata (Fig. 7.1). While these
review for those who would benet from a second gures and blocks may be instructive guides to
encounter with these topics. This review serves understanding the map patterns of faulted strata,
another important purpose. Textbooks in struc- the perceptive students might ask: what happens
tural geology typically fail to make a clear con- near the end of a fault? Clearly a rigid block model
nection between the material taught in the could not provide a satisfactory answer, because
mechanics section of an introductory college the model fault has no end.
physics course and the mechanical concepts Moving beyond the rigid block models, it is nec-
employed in analyzing geological structures. By essary to consider a continuous and deformable
making that connection explicit students are pre- body of rock and to broaden our perspective to con-
pared to use the mechanics effectively and with a strain explicitly how mass and momentum are
condence that stems from understanding the conserved throughout such a body. Here we postu-
fundamentals of the discipline. late that temperature changes and chemical
We begin with the concepts of linear and changes within the body are negligible, so the
angular momentum as treated in particle dynamics model is isothermal and isochemical. If heat ow and
and generalize this to rigid-body dynamics. While chemical reactions play an important role, than
the methods of classical dynamics have applica- conservation of energy and conservation of chem-
tion in many familiar human endeavors (from ical species must be included. With these limita-
tracking satellites to playing snooker) where accel- tions in mind, conservation of mass leads to the
erations are key to understanding, the rock masses equation of continuity and the conservation of
that comprise Earths crust do not experience momentum leads to the equations of motion for the
appreciable accelerations, apart from those brief material continuum. In turn, the equations of
7.1 PARTICLE DYNAMICS 245
(a) (a) z f(1)

f(2)
Particle
of mass m
Acceleration, a
y
Resultant f(n)
x force, F
(b) (b) z
f(1)
f(2)
Particle
of mass m
Velocity, v
y
Fig 7.1 Rigid structural block diagrams of normal fault
Linear f(n)
x momentum, p
cutting inclined sedimentary strata (a) Faulted blocks.
(b) Left-hand block eroded to remove fault scarp. Reprinted
Fig 7.2 (a) Particle of mass, m, with set of forces, f(i),
from Billings (1972) by permission of Peavson Education, Inc.,
resultant force, F, and acceleration, a. (b) Particle of mass, m,
Upper Saddle River, NJ.
with set of forces, f(i), linear momentum, p, and velocity, v.
motion are specialized for the linear elastic solid with respect to a Cartesian coordinate system (x, y,
and the linear viscous uid, two of the most useful z) and kinematic quantities, such as velocity and
material behaviors in studies of structural geology. acceleration, are dened with respect to an iner-
Finally, we consider the equations of motion for a tial frame of reference (Fig. 7.2). An inertial frame of
body that experiences negligible accelerations reference is xed in space or moving with a constant
and derive the equations of equilibrium. velocity relative to the distant stars. Newtons
Second Law describes the relationship between
the acceleration of such a particle, relative to an
7.1 Particle dynamics inertial frame of reference, and the forces acting
upon it:
7.1.1 Force and linear momentum
F ma (7.1)
Consider the motion of a single particle under the
action of applied forces (Resnick and Halliday, Here the vector a is the acceleration of the particle,
1977, Chapter 5). In this idealized context a parti- the vector F is the resultant force acting on the par-
cle is considered to be an isolated quantity of mass ticle, and the scalar m is the mass of the particle.
residing at a point: it has neither a size nor a shape. What Newton meant by resultant force, F, is
Furthermore the mass is postulated to be constant the vector summation of all the forces f(1), f(2),
with respect to time, so conservation of mass is . . . , f(n) acting on the particle (Fig. 7.2a). Here the
satised by denition. The particle is positioned numbers in parentheses identify the different
forces. The three components, Fx, Fy, and Fz of the

resultant force are equal to the sum of the respec-
tive components of all forces acting on the parti-
cle. For example, the x-component is calculated as:
n
Fx fx(1) fx(2) fx(n) f (i) ma
i1
x x
(7.2)
Here n is the total number of different forces and

ax is the x-component of acceleration. The compo-
nents Fy and Fz follow by changing subscripts. Note
that the component of acceleration is directly pro-
portional to the component of resultant force. Said
another way, the acceleration components are in
the same ratio as the resultant force components:
a x Fx a y Fy a z Fz
, , (7.3)
a y F y a z F z ax F x
Therefore, the acceleration takes place in exactly

the same direction as the action of the resultant
force. This phenomenon is referred to as a linear
acceleration because it takes place in a line, the line
of action of the resultant force.
The relationship among force, mass, and accel-
eration expressed as Newtons Second Law (7.1)
can be rearranged to place either acceleration or Fig 7.3 Portrait of Sir Isaac Newton by Sir James Thornhill
force alone on the left-hand side of the equation. in 1712. The original is at Woolsthorpe Manor, UK,
Thus, one can think of either the acceleration or birthplace and family home of Newton. Reproduced from a
the force as the dependent variable to be calcu- photographic image with the permission of the National
lated given the mass and the other quantity. One Trust Photographic Library (NTPL/John Hammond).
might conclude from this mathematical manipu-
lation that acceleration of a particle causes a calculations, physicists treat either a or F as depen-
force, or that application of a force on a particle dent upon the other, based on the necessities of
causes acceleration. Newtons position on this the problem at hand, and generally demure on
question of causality is suggested in a recent questions of causality. In conversations, physicists
translation of The Principia where the rst and usually follow Newton and speak of forces causing
second laws are written as: accelerations. This seems to be an effective way to
Every body preserves in its state of being at rest or of moving proceed, despite the obvious duality of thought.
uniformly straight forward, except insofar as it is compelled Apparently, advances in physics do not depend
to change its state by forces impressed. upon a resolution of this question, so it is largely
A change in motion is proportional to the motive force ignored. On the other hand the question of causal-
impressed and takes place along the straight line in which ity has attracted a good deal of attention from
that force is impressed (Newton, 1687, p. 416).
modern philosophers who discuss so-called causal
Apparently, for Newton (Fig. 7.3), forces cause accel- asymmetries or the direction of causation in philo-
erations and not the other way around. Modern sophical terms (Sosa and Tooley, 1993; Hausman,
physicists are more ambivalent about this ques- 1998; Pearl, 2000).
tion, probably because of the inherent symmetry The linear momentum, p, of a particle is a vector
in equations like F ma and because of philosoph- quantity dened as the product of the particle
ical concerns about the nature of causality. In mass and its velocity (Fig. 7.2b):
7.1 PARTICLE DYNAMICS 247
p mv (7.4) z A
fco
Just as the force and acceleration have the same s
direction, the linear momentum and velocity E
have the same direction because their respective n B
components are in the same ratio, for example

px/py vx/vy. f
O y
Given constant resultant force acting on a par-
r
ticle of known mass, one can calculate its acceler-

fsin
ation using (7.1). The instantaneous velocity of the D
particle is found using the fact that the time rate
x C
of change of velocity is the acceleration; so one Particle of
can integrate a constant acceleration over time to mass m
calculate the changing velocity. In differential
Fig 7.4 Schematic diagram to define the torque, , with
form dv adt, and integrating:
respect to the origin, O, of a Cartesian coordinate system as

dv a dt,so v at v(0) (7.5)
the vector product of the radial position vector, r, and the
net force, f, acting on a particle of mass, m.
This constant, v(0), is the velocity at the initial

time, t 0. Although we tend to think of time as (Fig. 7.4). The position and force vectors lie in the
having no beginning or end in daily life, in ideal- plane OABC (shaded in the gure) that has a unit
izing these problems in mechanics we specify an normal vector n. The position of the mass and the
arbitrary beginning of a process at the initial direction of the force are arbitrary with respect to
time. The velocity of a particle changes linearly the coordinate origin and axes, but the normal
with time, if the acceleration is constant, and emanates from the origin and it is perpendicular
from (7.4) we infer that the momentum changes to the plane containing f and r. The angle is the
linearly with time under these same conditions. smaller of the two angles measured in the plane
Constant acceleration of a particle is, in turn, OABC between the lines of action of the radial
associated with a constant resultant force. vector, r, and force vector, f.
The time rate of change of linear momentum The torque is proportional to the distance
is equivalent to the resultant force acting on the from the origin to the point of application of the
particle. This is demonstrated using (7.1) and (7.4) force, and also is proportional to the component
as follows (Resnick and Halliday, 1977, p. 168): of the force acting perpendicular to the position
d d d vector in the plane OABC (Fig. 7.4). The torque is the
p (mv) m v ma F (7.6) vector product of the radial position vector, r, and
dt dt dt
the force vector, f:
Here it is understood that the particle mass does
not change with time. rf (7.7)
By denition this vector is directed perpendicular

7.1.2 Torque and angular momentum to the plane containing the two crossed vectors, so
The position of the particle with respect to the the torque is a vector parallel to n (Fig. 7.4). The
origin of coordinates and inertial frame of refer- direction of the torque vector is determined by
ence plays no role in relating force and linear aligning the thumb of your right hand with n
momentum (7.6), but position is central in the such that your ngers curl from r toward f. The
relationship between torque and angular momen- direction of the torque is the direction in which
tum. Therefore, consider a force vector, f, acting your thumb points: a right-hand convention.
on a particle of mass, m, located by the radial posi- The magnitude of the torque is:
tion vector, r, drawn from the origin, O, of a coordi-
nate system (x, y, z) and inertial frame of reference r ( f sin ), for 0 (7.8)
Here r and f are the magnitudes of the radial and the particular location of the origin of the coordi-
force vectors, respectively, and f sin is the com- nate system, O, and inertial frame of reference
ponent of f along the line DE drawn perpendicular (Fig. 7.4). That is, we speak of the torque with
to the radial vector and in the plane containing r respect to the origin, or the angular momentum
and f (Fig. 7.4). Thus, force acting parallel to the with respect to the origin. If we were to move the
position vector produces no torque and force location of the origin these quantities would
acting at right-angles to the position vector will change in magnitude and direction. Further-
contribute all of its magnitude to the torque. more, the particle is not tied to the origin, so it
The angular momentum, , is dened in terms will move along a path dened by the line of
of the position vector of the particle, r, and the action of the applied force, f, rather than spin
linear momentum as: about an axis of rotation parallel to n. The line
dened by n is not an axis of rotation for the par-
r mv r p (7.9) ticle per se, but rather it is the line along which
Because the linear momentum acts in the same the torque and angular momentum are directed.
direction as the force, the angular momentum, ,
acts in the same direction as the torque, (Fig. 7.4).
That is, acts along the normal n with a direction 7.2 Rigid-body dynamics and
determined by the right-hand convention. The statics
magnitude of the angular momentum is:
In this section we generalize the relationships of

r (p sin ) r (mv sin ),for 0
particle dynamics so they apply to an aggregate of
(7.10)
particles making up a rigid body, one that does not
If the position vector and the linear momentum change shape or size with time, and does not ex-
are parallel to one another, the magnitude of the perience any gain or loss of mass. This rigid body,
angular momentum is zero. If the velocity of like the particle, satises conservation of mass by
the particle is exactly perpendicular to r, then the denition. The rst step toward understanding the
magnitude of the angular momentum is simply dynamics of a rigid body is to dene the center of
the product of the distance from the origin, the mass. Then we relate the forces acting on the body
mass, and the magnitude of the velocity, rmv. to its linear and angular momentum. Conserva-
Recall that the time rate of change of the linear tion of momentum is the underlying principle for
momentum is equivalent to the force acting on a rigid-body dynamics and provides the necessary
particle (7.6). There is an analogous relationship conditions for static equilibrium of such a body.
for the time rate of change of the angular momen-
tum and the torque, which is derived from (7.9) as 7.2.1 Center of mass
follows (Resnick and Halliday, 1977, p. 234): To dene the center of mass no restrictions need
d d dp dr be placed on the size of the body: it may represent
(r p) r p
dt dt dt dt a few cubic meters or many cubic kilometers.
r f v mv r f (7.11) Also, no restrictions are placed on the complexity
of the geological structures within the body. For
Here the standard form for the derivative of a example, consider the structural block diagram
product is used with the proviso that, for a cross (Chapter 7, frontispiece) of a part of the Penninic
product, the order of variables must be preserved. Alps (Argand, 1911). The distance across the front
Also, note that the cross product of two parallel of the block is about 75 km and the distance from
vectors, such as v and mv, is always zero. The front to back is about 50 km, so this is a crustal-
torque is equal to the time rate of change of the scale diagram. Note how the map pattern of
angular momentum. the rocks is reected in the cross section on the
It is important to recognize that the physical front of the block. This repetition of patterns
quantities we have dened in this section refer to occurs because the structures are plunging to the
7.2 RIGID-BODY DYNAMICS AND STATICS 249
z volumes of the elements that make up its con-

stituent parts as:
y Center 1 m
of mass V
V i1 i i
(7.12)
x
Here it is understood that V is the total mass of
the body and iVi is the mass of each volume
element. The total mass, V, contained in B does
r* not change with time: it is conserved. The small
volumes, Vi, do not change with time because
the body is rigid throughout. The individual den-
ri sities, i, are uniform within each element but
may vary spatially throughout the body. On the
other hand, these individual densities do not vary
in time: there is no mass transport from one
Vi, ri
volume element to another. In the limit as Vi
goes to zero this collection of volume elements is
equivalent to a system of particles, and the
Rigid body, B motion of each element is governed by the re-
lationships reviewed in the previous section.
Fig 7.5 Schematic diagram to define the center of mass, Depending upon the forces that are applied to
r*, of a rigid body made up of m volume elements, Vi, of this rigid body an individual volume element may
mass density, i. move in a complex manner that involves both
translations and rotations with respect to the ref-
erence frame. Any two elements, however, do not
southwest and interpreted to be continuous. In move relative to one another, because the body is
Fig. 7.5 we schematically represent Argands block rigid by denition. There is one position in this
diagram as a rigid body called B, which has an body, called the center of mass (Fig. 7.5), that moves
average density, , and total volume, V, neither of in the same manner as a single particle of con-
which change with time. You might wonder how centrated mass, V, when subject to the resultant
the folds and faults could have formed if the block of all forces applied to the body. The center of
were rigid and, of course, the answer is that they mass is quite special in that it behaves like the par-
could not. However, the center of mass can be cal- ticles we dealt with in the previous section, but
culated for an arbitrary instant in time during the there is only one such center of mass in any rigid
deformation using the method introduced here. body. The center of mass has a position vector, r*,
Furthermore, Argands example serves to empha- that is (Resnick and Halliday, 1977, Chapter 9,
size that this is not a vacuous exercise in mathe- p. 164):
matical physics, but rather a crucial step toward m
understanding the deformation of Earths crust. r* V r V
i1
i i i (7.13)
The rigid body of Fig. 7.5 is lled with m (not to
be confused with the mass, m) small volume ele- In other words, the product of the position vector
ments of rock, Vi, each having a uniform density, for the center of mass of the body and the total
i. Each volume element is located with a position mass is equal to the vector sum of the product of
vector, ri, radial to the origin, O, of the (x, y, z)- the position vectors for each volume element and
coordinate system and inertial frame of reference. their respective masses.
Fixing a coordinate system at some location in or The coordinates of the center of mass of the
on the Earth usually provides a suitable inertial body are the individual components of the pos-
frame of reference. The average density of the ition vector r*. For example, the x-coordinate of r*
body is related to the individual densities and is the component x*:
z center of mass if they are uniformly distributed

f2
(see next section). We generalize the relation-
y Center ship between resultant force and linear momen-
of mass tum for an innitesimal particle (7.6) to a form
appropriate for a rigid body acted upon by inter-
x nal and external forces by expanding the sum-
mation in (7.13):
Vr * 1V1r1 2V2r2 mVmrm (7.15)
r* The left-hand side is the product of the mass of the

entire body and the position vector for the center
of mass, and the right-hand side is the sum of the
F(s) products of the mass of each volume element and
f1 its position vector.
F(b) Next we take time derivatives of the terms on
the right-hand side of (7.15). Because the volume
v* and density of any element are constant in time
fn
we nd:
Rigid body, B
d d d
Fig 7.6 Schematic diagram to define linear momentum, P,
1V1 r 2V2 r2 mVm rm
dt 1 dt dt
of a rigid body acted upon by a set of surface forces, f(i), with a
resultant surface force, F(s), and a resultant body force, F(b). 1V1v1 2V2v2 mVmvm (7.16)
1 m Taking the time derivative again and using

x* x V
V i1 i i i
(7.14) Newtons Second Law (7.1) we have:
The densities of the volume elements, i, are not d d d

1V1 v1 2V2 v2 mVm vm
necessarily the same, so these equations account dt dt dt
for the spatial heterogeneity of density in 1V1a1 2V2a2 mVma m F
deformed rock masses such as that depicted in
(7.17)
Argands block diagram from the Penninic Alps
(Chapter 7, frontispiece). Given the structural Each term is equivalent to the resultant force
architecture as worked out by Argand, and the acting on a particular volume element, and their
densities of the individual rock units, as could be vector sum is the resultant force, F, acting on the
determined by a standard laboratory procedure, body.
one could calculate the components of the center In The Principia Newtons Third Law is stated:
of mass using (7.14) and similar expressions for
the y- and z-components. To any action there is always an opposite and equal reaction;
in other words, the actions of two bodies upon each other are
always equal and always opposite in direction (Newton,
7.2.2 Conservation of linear momentum 1687, p. 417).
In Fig. 7.6 we represent the rock mass as a rigid
body acted upon by arbitrary surface forces and It follows that the internal forces acting between
body forces and use this to seek relationships any two adjacent volume elements are equal in
among these forces and the linear momentum magnitude and oppositely directed. Thus, accord-
of the body. The resultant of the surface forces, ing to Newtons Third Law, the vector sum of these
F(s), is thought of as acting at the center of mass internal forces is zero. What remains on the right-
of the body. The resultant of the body forces, hand side of (7.17) is the sum of the external
F(b), also may be thought of as acting at the surface forces and the body forces:
d2 d2 d2
1V1 2r1 2V2 2r2 mVm 2rm torques acting on the body and is derived in
dt dt dt
Section 7.2.4.
F (s) F (b) (7.18)
7.2.3 Evaluation of surface and body
Note that the body force acts on all m elements of forces
the body, but the n surface forces act at particular We replace the mechanical action of the exterior
locations on the exterior surface of the body (Fig. rock mass on the body schematically in Fig. 7.6
7.6). with a set of surface forces, f1, f2, . . . , fn, acting at
Because neither the volume nor the average discrete points. The resultant of these n surface
density is a function of time for the rigid body, the forces is their vector sum which acts at the center
rst time derivative of the left-hand side of (7.15) is: of mass of the body:
d n
V r* Vv* P
dt
(7.19) F(s) fj
j1
(7.22)
This relationship denes the linear momentum, P,

This concept will be generalized in Section 7.2.5 to
of the entire rigid body with respect to the inertial
account for a continuous distribution of forces
reference frame. The velocity at any location other
per unit area, taken as the tractions acting on the
than the center of mass may be different, but the
surface of the body.
behavior of the body as a whole is characterized by
Evaluation of the body force focuses on the
v*. Analogous to (7.6), the time rate of change of
weight of the rock mass because this usually is the
the linear momentum, P, from (7.19) is equal to
only signicant contributor. Weight is the force
the sum of the surface and body force resultants
exerted on a rock body by the gravitational attrac-
dened in (7.18):
tion of the Earth. Consider the body shown in Fig.
d d 7.6 subdivided into m volume elements each of
P V v* F(s) F(b) (7.20)
dt dt mass iVi. The weight of each element is:
This relationship expresses one of those small wi iVigi (7.23)
number of physical laws mentioned at the begin-
ning of the chapter that summarize our under- Here gi is the local acceleration of gravity at the pos-
standing of natural phenomena. Here it is the law ition of the element. The direction of the acceler-
of conservation of linear momentum. For a rigid body, ation of gravity, g, is downward by denition and
subject to given surface and body force resultants, therefore denes the local vertical. The accelera-
linear momentum is neither created nor destroyed tion of gravity is found by measurement to vary
spontaneously, but changes with time in strict slightly from place to place over the Earths
accordance to the action of these resultant forces. surface and shallow interior due to local varia-
For conditions where linear momentum does tions in rock density and distance from the
not change with time, the conservation of linear Earths center. In other words it is a eld quantity
momentum (7.20) dictates that the resultant of all that varies with position, so we specied that gi is
external forces must be zero: the local value at the position of the element.
Because the Earths crust deforms as faults slip
d
if P 0,then F (s) F (b) 0 (7.21) and mountains grow, or as erosion brings down
dt
the height of mountains, the local acceleration of
This is one of two conditions required for static gravity can vary in time as well as space. These
equilibrium. The linear momentum of individual variations in g are studied by geophysicists and
volume elements of a rigid body may change with are used to infer density variations in the Earth
time, but the linear momentum of the body as a (Turcotte and Schubert, 1982, Chapter 5).
whole, with respect to an inertial frame of refer- Variations of g on and within the crust can
ence, must remain constant if the force resultants account for body forces that are different from
are zero. The second condition for static equilib- those one would calculate assuming a uniform g,
rium depends upon a balance of the external but these differences usually are insignicant for
problems in structural geology. For example, T(s)+T(b) z

density contrasts among typical rock types found
in the Earths crust could cause variations in g, f2
but common densities only range from about 2 y Center
103 to 3 103 kg m3 (Clark, 1966). Similarly, of mass
although g changes with distance from the center
x
of the Earth, most of our attention in structural
geology is focused on the outer shell of the Earth
with a thickness less than 100 km, relative to the
Earths radius of over 6000 km. The difference
between the equatorial radius (about 6378 km) r1 r*
and the polar radius (about 6357 km) is only
21 km, and the difference between the highest
mountain peak (almost 9 km above sea level) and f1
the deepest parts of the ocean oor (about 11 km _
rVg*
below sea level) is only about 20 km. Neither of
these differences is large compared to the radius
of Earth. Thus, the variations in g within the crust fn
Rigid body, B
are likely to be small, so we treat the magnitude
and direction of the gravitational eld as uniform Fig 7.7 Schematic diagram to define angular momentum,
in space and constant in time. , of a rigid body acted upon by a set of forces, f(i ), with a
It is convenient from a computational point of resultant torque, T(s), due to surface forces, and a resultant
view to replace the action of all the body forces, torque, T(b), due to the body forces.
wi, with the resultant force acting at a center of
gravity. The center of gravity and the center of n
mass are coincident if the gravity eld over the f Vg * 0

j1
j (7.26)
body is uniform in magnitude. The resultant body
force (weight) of the body is: This is a restatement of (7.21) and represents the
m rst condition for static equilibrium of the rigid
F(b) w Vg*
i1
i (7.24) body subject to discrete surface forces and a
uniform gravitational acceleration.
Here the acceleration of gravity is treated as a con-
stant, g*, with a magnitude g* 9.8 m s2. This 7.2.4 Conservation of angular
body force can be thought of as acting at the momentum
center of gravity, which is the center of mass. For When calculating the angular momentum for an
a non-uniform acceleration of gravity the center of isolated particle all forces act at the point where
gravity and center of mass may not be coincident. the particle resides (Fig. 7.4), but the points of appli-
As a consequence of the discrete surface forces cation of the forces may differ for a body with nite
(7.22) and the uniform gravity eld (7.24), the con- shape and size (Fig. 7.7). To calculate the angular
servation of linear momentum (7.20) is written: momentum the space occupied by this body is
n called B and it is lled with m distinct volume ele-
d d
dt
P V v*
dt j1
f j Vg* (7.25) ments, Vi, each having a uniform density, i. In
keeping with the previous section we consider
If the sum of the surface forces and the weight in gravity to be the only body force and the gravita-
each coordinate direction is zero, the time rate of tional acceleration to be uniform. Each element is
change of the linear momentum is zero. For located with a position vector, ri, radial to the
the rigid body this means that the velocity of the origin of the coordinate system and inertial frame
center of mass does not change with time and the of reference. The body has an average density, ,
surface and body forces are related as: and total volume, V, and these quantities are
related to the individual densities and volumes of momentum is neither created nor destroyed
its constituent elements using (7.15). spontaneously, but changes with time in strict
Taking the time derivative of the right-hand accordance with the action of these forces.
side of (7.15) with volume and density of each If the angular momentum does not change
element considered constant gives (7.16), the appreciably with time one postulates that it is
vector sum of the linear momenta for the m el- exactly zero for modeling purposes. The conserva-
ements. The cross products of the respective position of angular momentum then requires that the
tion vectors and linear momenta are the angular resultant of all external torques is zero:
momenta of the elements, and the vector sum of n
d
these is: if
dt
0, then [r f ] r* Vg* 0
j1
j j (7.31)
r1 1V1v1 r2 2V2v2 . . . rm mVmvm
This is the second condition used to establish the
(7.27)
static equilibrium of a rigid body. The other condi-
From (7.11) the time derivative of the angular tion is (7.21). Note that the angular momentum of
momentum is the torque, so the resultant torque, individual elements of the body may change with
T, acting on the rigid body is: time, but the angular momentum of the body as
d d a whole, with respect to an inertial frame of ref-
(r 1V1v1) (r2 2V2v2) erence, is constant in time if the resultant torque
dt 1 dt
d is zero.
(rm mVmvm) T (7.28)
dt
This summation includes both the torques caused
7.2.5 Static equilibrium in integral form
for the rigid continuum
by external forces acting on the body and those
When contemplating the equilibrium of a body of
caused by internal forces acting among the el-
rock it would rarely be practical to think about a
ements. Newtons Third Law is invoked to justify
large number of volume elements, each with a par-
ignoring torques caused by internal forces
ticular density or momentum. While this device
(Resnick and Halliday, 1977).
has clear pedagogical advantages for introducing
What remains after eliminating the contribu-
the conservation laws for a rigid body, a more prag-
tions of the internal forces are the torques caused
matic approach is to invoke the continuum and
by each of the n surface forces, fj, applied at pos-
let the physical quantities under discussion be
itions, rj, and the torque caused by the uniform
dened at every point of the body. Then summa-
gravitational body force, Vg*, acting at the center
tion over the number of elements is replaced by
of mass, r* (Fig. 7.7):
integration. One of founders of structural geology
n
in the nineteenth century, Grove Karl Gilbert (Fig.
[r f ] r * Vg* T(s) T(b)
j1
j j (7.29)
7.8a), developed his mechanical model for laccol-
This is the resultant external torque with respect ith formation (Fig. 1.18) using this approach to
to the origin, O, and inertial frame of reference. equilibrium. The concept of equilibrium, whether
From (7.11) the time rate of change of the angular used in the consideration of mountain building or
momentum, , of the body as a whole is equal to of erosion of the landscape was central to Gilberts
the resultant torque caused by all external forces method of investigation (Pyne, 1980).
acting on the body: The concept of a center of mass is extended to
n
the continuum by considering and V to be the
d
dt

j1

[rj fj ] r* Vg* (7.30) average density and total volume, neither of
which change with time because the body is rigid.
This equation expresses another of those small The mass of this body is conserved by denition.
number of physical laws mentioned at the begin- An integral over the volume of the body is derived
ning of the chapter, in this case the law of conserva- from (7.13) by letting the element volumes, Vi,
tion of angular momentum. For a rigid body, subject shrink to an innitesimal size, dV, so the position
to given surface forces and gravity, angular vector of the center of mass is:

1
r* r dV
V (7.32) (a)
V
Here the mass density is an integrable function of

the spatial coordinates, (x, y, z).
Perhaps nowhere in the literature of structural
geology (Pyne, 1980) is the concept of equilibrium
used more decisively to guide the development of
conceptual models than in the writing of G. K.
Gilbert (Fig. 7.8a).
In Fig. 7.8b the rock mass is represented as a
material continuum, bounded by the surface S,
and points are located by position vectors, r, asso-
ciated with a coordinate system and inertial
frame of reference. We consider the intensity of
distributed forces acting over the surface S as
described at a particular point by the traction
vector, t(n), acting on the surface element, S,
with outward unit normal vector n. Both the trac-
tion vector and the unit normal vector are func-
tions of position on S and the resultant force (b) z
acting on the area, S, is approximated as t(n)S.
Letting the area element shrink toward zero, we y
write S as the differential quantity dS, and take Surface, S
the integral of the force, t(n)dS, over the entire x
surface:
r

F(s) t(n) dS (7.33)
S
n
S
This integral is the resultant surface force, F(s), r

acting on the rigid body.
t(n)
The body force is the weight and this can be
V, r
described for a small volume, V, within the body
of rock (Fig. 7.8b) as the product of its average
density, , and the local acceleration of gravity, g*,
treated as uniform. The quantity g* is the weight rg*
per unit volume. Letting the small volume shrink
toward zero and writing V as the differential Fig 7.8 (a) Grove Karl Gilbert. Reprinted from Hunt
quantity dV, we integrate the weight over the (1988) with permission of The Geological Society of
entire volume of the body: America. (b) Schematic diagram to define the static
equilibrium of a continuous rigid body with a distribution of

F(b) g*dV (7.34)
traction vectors, t(n), acting on surface elements, S, and a
distribution of weights per unit volume, g*, acting on
V
volume elements, V.
This integral expression represents the resultant
body force due to gravity, F(b), and it may be
thought of as acting at the center of mass of the In both cases the torque is found by considering
body. the location of each surface and volume element
The surface and body forces may produce in terms of the position vector, r, and forming the
torques about the origin of the coordinate system. cross product of the position vector with the force.
For the tractions acting on the surface of the body (a)

the resultant torque, T(s), is:

T(s) [ r t (n)] dS (7.35)
S
For the body force acting over the volume of the

body the resultant torque, T(b), is:

T(b) [r g*] dV (7.36)
V
The sum of these two integrals represents the

resultant torque acting on the body, and this
torque is thought of as a vector located at the
origin of coordinates, O.
A restatement of conservation of linear and
angular momentum for the continuous rigid
body in equilibrium is that the resultant external
force and torque due to surface tractions and (b)
gravity must be zero:
t(n) dS g* dV 0 (7.37)

S V
[r t(n) ] dS (r g*) dV 0 (7.38) Fig 7.9 (a) M. King Hubbert: photograph reproduced with
permission of the School of Earth Sciences, Stanford
S V
University. (b) Vertical cross section through idealized fold
All of the physical quantities in these equations and thrust mountain belt. Reprinted from Hubbert (1951)
are dened at each and every point in the contin- with permission of The Geological Society of America.
uum. None of these quantities are functions of
time, but the traction and density may vary spa-
understanding for a variety of empirically well-known
tially and must be integrable functions of the
geologic structures (Hubbert, 1951).
spatial coordinates. Gravitational acceleration is
taken as uniform in space and constant in time. In this paper Hubbert provides the motivation for
laboratory experiments using the sandbox technique.
7.2.6 An example: Appalachian fold and Similar techniques using model materials such as
thrust mountain belt sand, clay, and plaster continue to provide insights
In 1951 M. King Hubbert (Fig. 7.9a) published a concerning the development of crustal-scale struc-
paper in the Geological Society of America tures to this day (Fossen and Gabrielsen, 1996;
Bulletin entitled Mechanical basis for certain Wang and Davis, 1996; Guglielmo et al., 2000;
familiar geologic structures. The abstract reads Ackermann et al., 2001; Cobbold et al., 2001; McClay
as follows: and Bonora, 2001). The sandbox used by Hubbert
contained a rigid platen that produced a set of
A simple experiment with loose sand shows that this
thrust faults when moved laterally (Fig. 7.10). The
material exhibits faulting under deformational stress
in a manner remarkably similar to rocks. Moreover, motion of the platen apparently increased the hor-
the sand experiment is amenable to theoretical analy- izontal compressive stress while the vertical stress
sis with good agreement between predicted and remained essentially that caused by gravity. These
observed behavior. The same theoretical treatment, stress changes led to the development of the faults
with slight modication, is also applicable to the apparently analogous to those in fold and thrust
behavior of rocks, and appears to afford a basis of mountain belts.
(a) pioneers of modern structural geology in the

twentieth century.
In Hubberts paper a brief section called
Application of the Newtonian laws of motion
points out how popular concepts of fold and
thrust faulted mountain belts, such as the
Appalachians, suffered from an incomplete con-
sideration of equilibrium (Hubbert, 1951). The fold
and thrust mountain belts of the world were char-
acterized as having a train of folded strata with
increasing amplitude and degree of asymmetry in
(b) one direction, perpendicular to the belt (Fig. 7.9b).
The most intense deformation is found on one
side of the belt, the thrust faults characteristically
dip toward the region of greater intensity of defor-
mation, and the folds are overturned away from
this region. The challenge for structural geologists
at the time was to offer an explanation for the
mechanical cause of the asymmetric style of
mountain building. Hubbert describes a notion,
apparently formulated and popularized by J. D.
Dana (1847a, b), that the cause of fold and thrust
Fig 7.10 Photographs of sandbox model apparatus used to
mountain belts was a one-sided, active thrust from
investigate the development of normal and thrust faults.
the side adjacent to the greatest deformation.
(a) Sandbox with undeformed layers. (b) Platen has moved to
the right generating a set of thrust faults. Reprinted from
According to this notion the thrusting force dissi-
Hubbert (1951) with permission of The Geological Society of pated across the mountain belt and became
America. insignicant on the side away from the intense
deformation.
The one-sided thrust hypothesis is depicted in
the context of a free-body diagram showing greater
Hubbert also demonstrated how an elemen- horizontal forces on one side than the other (Fig.
tary analysis of homogeneous stress states in the 7.11a). The volume of rock has a long dimension,
sand could be combined with a criterion for shear L, oriented across the mountain belt, a short
failure, called the Coulomb criterion, to explain dimension or width, W, and a height, H. The free
the relationship between the stress state and body is a thin slice of rock across the belt that is
faulting (Hubbert, 1951). This method for analyz- meant to be representative of any other slice
ing faulting is in common use today and has been taken at other locations along the belt: all paral-
extended to the study of earthquake aftershocks lel slices would be similar in terms of geometry
and the triggering of smaller earthquakes by and loading conditions. The body is free in the
major earthquakes (King et al., 1994; Stein et al., sense that it is cut away from the rest of Earths
1996; Harris and Simpson, 1998; Cocco and Rice, crust and the mechanical action of the exterior
2002). These studies depend upon knowledge of rock mass is replaced by a distribution of trac-
the state of stress around the faults in question, tions acting on the surface and a distribution of
and this usually comes from solutions to bound- gravitational forces acting on the body. According
ary value problems that obey conservation of to the notion of an active thrust, the greater
mass and momentum. Here we show how the forces cause the intense deformation on the left-
boundary conditions for such a model are con- hand side, whereas the lesser forces are consis-
strained by the conditions of mechanical equilib- tent with the lack of deformation on the
rium. M. King Hubbert (Fig. 7.9a) was one of the right-hand side (Fig. 7.9b).
(a) z does not satisfy conservation of linear momen-

y
W L tum as embodied in the equilibrium equations
(7.37) and (7.38). He proposed to correct the free
O x body as drawn in Fig. 7.11a with a distribution of
shear tractions on the bottom of the block acting
to oppose the active thrust (Fig. 7.11c). In what
H follows we will examine the tectonic forces acting
on a model for a fold and thrust mountain belt as
conceived by Hubbert and determine the relative
CL Rg*z Rg*z magnitudes of the tractions acting on the free
body. Small changes in linear and angular
(b) momentum with respect to time are acknowl-
z W edged for a mountain belt because we know that
O y the deformation began at some time and later
ended. Also as the thrust faults and folds accom-
modated shortening across the mountain belt,
velocities certainly changed locally. None-the-less,
at any given time throughout the history of moun-
H tain building the forces in any coordinate direc-
tion must (nearly) balance one another and the
net torques about the origin in any coordinate
direction must (nearly) balance one another.
We begin by evaluating the tractions acting in
Rg*z Rg*z the x-coordinate direction on all six sides of the
body (Fig. 7.11). On the left side the normal trac-
(c) z
y tion is composed of a tectonic part with a magni-
W L tude, CL, and a lithostatic part related to the
weight of the overlying rock, whereas the right
O x side carries only the lithostatic tractions:
BC: on x 0,0 y W,
Tz(L) Rg*H H Tz(R) H z 0; tx CL g*z
(7.39)
SB BC: on x L,0 y W,
H z 0; tx g*z
CB
The tectonic traction is uniformly distributed over
Fig 7.11 Free-body diagrams for fold and thrust mountain the left side whereas the lithostatic traction
belt. (a) One-sided, active thrust concept with unbalanced increases linearly with depth. Both parts of this
tectonic traction, CL, and lithostatic tractions, g *z. traction push against the body and are positive in
(b) Lithostatic tractions on cross sections in (y, z)-plane. sign (note z is negative below the surface). The
(c) Additional loading necessary for static equilibrium tractions are drawn as arrows pointed at the mid-
includes shear tractions SB on the base, and TZ on the sides, section of the body, but it should be recognized
and normal traction CB on the base.
that the traction is uniformly distributed over the
entire side of the free body. The traction on the
The explanation based on an active thrust right side pushes against the body in the negative
apparently was satisfactory for many structural x-direction, so its sign is negative.
geologists of the day; however, as Hubbert pointed The top surface of the free body is traction
out, the distribution of surface and body forces free and on the bottom surface the shear traction
proposed by Hubbert is taken as uniform with a The shear traction is equal to the tectonic traction
magnitude, SB: times the ratio of the height to the length of the
body.
BC: on 0 x L, 0 y W, z 0; tx 0 The only y-components of the tractions acting
BC: on 0 x L, 0 y W, (7.40) on the free body (Fig. 7.11b) are those related to the
z H; tx SB lithostatic loading and these exactly balance one
another. The z-components of the surface trac-
On the front and back sides of the free body we tions and the body forces are shown on the free
follow Hubberts suggestion that there are no body in Fig. 7.11c. The presence of a shear traction
shear tractions: on the bottom of the body raises a question about
BC: on 0 x L,y 0, the possibility of a shear traction acting in the z-
H z 0;tx 0 direction on the left and right sides. For the
(7.41) moment we do not explicitly dene these trac-
BC: on 0 x L,y W, tions, because their distributions and magnitudes
H z 0;tx 0 are best addressed in terms of the conservation of
angular momentum, but account for them as
These conditions would apply for a linear moun-
follows:
tain belt.
For equilibrium the integral of the x-compo- BC: on x 0,0 y W,
nents of the tractions over the surfaces on which H z 0;Tz Tz(L)
they act must sum to zero (7.37): (7.45)
BC: on x L,0 y W,
0 W 0 W
H z 0;Tz Tz(R)
(CL g*z) dy dz (g*z) dy dz
H 0
W L
H 0
Here Tz(L) and Tz(R) are unspecied functions rep-
(S ) dx dy 0
resenting the traction components in the z-direc-
B (7.42) tion on the left- and right-hand sides of the body.
0 0
The top surface is traction free, but the bottom
Because the integrands are constant the inner surface is subject to a uniform traction in the z-
denite integrals over the limits 0 to W with direction of magnitude CB:
respect to y and 0 to L with respect to x produce
BC: on 0 x L,0 y W,
the constants W and L, respectively. The remain-
z 0;Tz 0
ing integrals are evaluated over their respective
(7.46)
limits and the resulting expression is simplied as BC: on 0 x L,0 y W,
follows: z H;Tz CB
| | |
0 0 W
1 1 On the front and back sides of the body the sym-
W (CLz g*z2) W g*z2 L SB y 0
2 H 2 H 0
metry of the loading dictates that there are no
CLWH SBWL (7.43)
shear tractions in the z-direction:
The net force caused by the tectonic loading on
BC: on 0 x L,y 0,
the left side of the free body is exactly balanced by
H z 0; Tz 0
the net force caused by the resisting shear loading
on the bottom of the body. The quantitative rela- BC: on 0 x L,y W,
tionship between the magnitudes of the tectonic H z 0;Tz 0 (7.47)
traction on the left side, CL, and the resisting shear
traction on the bottom, SB, is: Referring to (7.37), we integrate the z-compo-
nents of the tractions over the surfaces on which
SB CL
H
L
(7.44) they act, integrate the body force over the volume,
and set the sum to zero:
0 W 0 W W L
z
Tz(L) dy dz Tz(R) dy dz C dx dy
B
(a) L/2
H 0 H 0 0 0
Center
0 W L
of mass
(g*) dx dy dz 0 (7.48)
H 0 0 rg*V
H/2 r* b*
We postulate that Tz(L) and Tz(R) produce net O r x
surface forces that are equal in magnitude and
CB
oppositely directed, so they exactly balance and
the rst two integrals cancel one another.
Carrying out the integration, we have: (b)
z
CL L
CB g*H (7.49)
The normal traction on the bottom of the free

body is the product of the density, the uniform r
acceleration of gravity, and the height of the H r
body. b
The distributions of shear tractions on the left x
and right sides of the free body are constrained O
using conservation of angular momentum (7.38).
SR(1 z/H)
Because of the two-dimensional nature of the
Fig 7.12 Free-body diagrams for fold and thrust mountain
body we only consider the torques produced by belt. (a) Torque due to weight and normal tractions on base.
forces acting in the (x, z)-plane (Fig. 7.12). The (b) Torque due to tractions on right and left side.
torques acting on the free body can be calculated
with respect to any origin so it is helpful to
direction), and that caused by the weight is posi-
choose an origin that simplies the calculation.
tive (it acts into the page in the positive y-direc-
Choosing the lower left corner for the origin elim-
tion). Combining the appropriate terms and
inates torques produced by the shear tractions on
integrating over the bottom surface and the
the left side and the bottom. Recall that the mag-
volume we have:
nitude of the cross product of two vectors is the
WL H W L
xC sin (2) dx dy 2 sin *

product of the magnitude of the rst vector, the L
magnitude of the second vector, and the sine of B
the smaller angle between the lines of action of 0 0 0 0 0
the two vectors. Thus, we evaluate the torque pro- (g* sin *) dx dy dz 0
duced by the normal traction on the base of the (7.51)
free body, and that due to the weight acting at the 12 CBWL2 12 g*L2WH 0
center of gravity as (Fig. 7.12a):
From (7.49) we know that CB g*H, so these
On 0 x L,0 y W,z 0, torques exactly balance.
Next consider the torques produced by the tec-
or r x; Tz CB,sin sin (2)
tonic traction acting on the left side of the body
(7.50) and the shear traction acting on the right side
At x 12L, y 12W,z 12H, (Fig. 7.12b). The shear traction on any vertical
surface must be zero where that surface intersects
L
or r r* ;g g*,sin sin * the traction-free surface of the Earth. Thus, we
2 sin *
postulate that the distribution of shear traction
The torque caused by the bottom traction is nega- on the right side varies linearly from zero at the
tive (it acts out of the page in the negative y- top to a value SR at the bottom:
On x 0,0 y W,0 z H, 7.3 Conservation of mass and

or r z; Tx CL,sin sin (2) momentum in a deformable
(7.52)
On x L,0 y W,0 z H, continuum
or r
L
sin z
;Tz SR 1 ,sin
H
L
r The previous sections of this chapter have devel-
oped the concept of momentum conservation
Combining the appropriate terms and integrat- within the context of particle dynamics and of
ing over the left and right sides of the body we rigid-body dynamics and statics. Despite the fact
have: that equilibrium analysis of a rigid body is an
important step to conrm the applicability of pre-
WH WH
sin
scribed tectonic boundary conditions, this analy-

L
zCL sin (2) dy dz
sis does not consider the deformation of the body
0 0 0 0
itself. For structural geologists the deformation of

z
SR 1 sin () dy dz 0
H
(7.53) a rock mass is the central focus, so we turn now to
conservation of mass and momentum in a
CLWH SRLW 0 deformable body of rock idealized as a material
continuum. These relationships underlie the
Using (7.44) we nd SR is related to the magnitude more specialized equations that we use later in
of the basal shear traction as: the textbook to model rock deformation during
folding, faulting, fabric development, and the for-
SR CL
H
L
SB (7.54) mation of many other geological structures.
We note here that both of the conservation
The shear traction on the right side in the lower laws are developed without regard for particular
right corner (x L, z 0) is equal to the shear trac- material properties other than mass density, so
tion on the bottom, because the respective shear they apply quite generally to the entire spectrum
stresses must be equal. The shear tractions on the of rock deformation, from silicate magmas that
left- and right-hand sides are: ow like a uid to rocks that deform like a mal-
leable solid to rocks that deform like an elastic
Tz(L) Tz(R) SB 1 z
H (7.55)
solid. We do restrict our attention to conditions
where temperature uctuations and the associ-
ated changes in density and ow of heat can be
This completes the equilibrium analysis of
ignored. In addition, we ignore chemical reac-
Hubberts Appalachian free body. This, or the
tions and the associated changes in density and
analogous procedure using relationships that
concentration of chemical species. Thus, the rela-
include the time rate of change of the linear and
tionships derived here apply, strictly speaking, to
angular momentum, is a vital step in any model-
a material that is isothermal and isochemical.
ing program to conrm that the prescribed trac-
Although few tectonic processes would obey these
tions and body forces are consistent with the
restricted conditions in detail, we show by
conservation of momentum. Without such
example that they do not preclude gaining con-
conrmation, any discussion of the development
siderable insight into those processes.
of structures lacks the necessary foundation in
mechanics to be credible, and it should be
7.3.1 Referential and spatial descriptions
ignored.
of motion
Recall from the discussion of kinematics in
Chapter 5 that two different sets of coordinates
may be used to describe the positions of particles in
a deforming rock mass: (X, Y, Z) are the coordinates
7.3 THE DEFORMABLE CONTINUUM 261
of a particle in the initial state, and (x, y, z) are the

(a)
coordinates of a particle in the current state (Fig. 5.6).
The coordinates (X, Y, Z) are components of the posi-
tion vector X and (x, y, z) are components of x. The
(x, y)
Y, y
two sets of coordinates usually are measured with
u
respect to the same origin and axes. Each coordi-
(X, Y)
nate in the current state is given by the sum of the x
respective coordinate in the initial state and the
corresponding component of the displacement X
vector, u: X, x
0

x X ux
10 cm
x X u,or y Y uy (7.56)
z Z uz
In this and subsequent chapters we consider

deformation primarily in the context of either an
elastic solid or a viscous uid. While the chosen
(b) Y, y
coordinates are independent of material proper-
ties, the theories of elastic solid mechanics and
viscous uid mechanics have been developed (X, Y)
from quite different points of view. In this section
we consider how these different points of view X
X, x
impact the analytical descriptions of motion and
0
develop concepts that are used in subsequent sec-
tions to apply the principles of conservation of
mass and momentum to a material continuum.
The deformation of an elastic solid is viewed 10 cm
with the initial state taken as the reference. In an
engineering context this state usually is associ-
ated with zero external loads. In other words
there are no surface or body forces acting on the
Fig 7.13 Illustration of the referential description of
elastic solid. Then, a given loading is applied, and
motion, xx(X, t), using a small fault in granitic rock of the
the particles displace to the current positions: it is
Sierra Nevada, CA (Segall and Pollard, 1983). The position
understood that they return to the initial posi- vector x locates a particle in the current state that was
tions when all the loads are removed. Because the located by the position vector X in the initial state.
conguration of particles changes with time as Photograph by D. D. Pollard.
the loading changes, but always returns to the
initial state upon unloading, this unloaded con-
dition is thought of as the natural state of the displacements go to zero and the particles return
elastic body. In the geological context the initial to the initial positions.
state may be quite arbitrarily chosen and usually For example, consider the exposure of granitic
is associated with a pre-existing loading condition rock (Fig. 7.13a) from the Sierra Nevada of
involving both surface and body forces. Displace- California where an aplite dike is offset about a
ments corresponding to this loading condition decimeter by a fault with apparent left-lateral
are undened. One then considers some change motion (Segall and Pollard, 1983a). The geological
of loading and the corresponding displacement inference is that this structure can be restored
of particles to their current positions. If the from the current state to an initial state in which
loading reverts to that of the initial state these it was continuous across the fault (Fig. 7.13b). The
mechanical inference is that the faulting can be

described by elastic deformation, specically the
displacement eld relating the initial and current
states. We do not suppose that the offset aplite
dike would return to its initial conguration if
the rock mass were excised from the Sierra
Nevada today and all external loads relaxed. This
supposition ignores the loading that existed
before faulting and the mechanical behavior of
the rock mass over the 80 million years since the
faulting event. Rather, we suppose that the defor-
mation at the time of faulting is adequately
described as elastic. This supposition is supported
by modern studies of deformation associated with
active faulting and by laboratory studies of the
constitutive properties of rock.
Taking the coordinates (X, Y, Z) and time, t, as
the independent variables, we develop the so-
called referential description of motion (Malvern,
1969). At the time just prior to t 0 the particles Fig 7.14 Portrait of the Italian mathematician Joseph-Louis
of rock are in their initial conguration around Lagrange who was born in Turin in 1736 and baptised
Giuseppe Lodovico Lagrangia. The coordinates (X, Y, Z) used
the un-slipped fault (Fig. 7.13b), which is in
in the referential description of motion are referred to as the
mechanical equilibrium with respect to the pre- Lagrangian coordinates. Reproduced with the permission of
existing loads. Under the prescribed change in the Department of Special Collections, Stanford University
surface and/or body forces initiated at time t 0, Library.
the particles displace to their current coordinates
such that:
A complete referential description of the

x x(X, Y, Z, t)
motion of particles from the initial state would be
x x(X, t),or y y(X, Y, Z, t) (7.57) a function x(X, t) that describes the continuous
z z(X, Y, Z, t)
change of position for every particle, each origi-
nally at a particular point X, for all times from t
Recall that the x (or x, y, and z) on the left-hand
0 to t the current time. In other words, the paths
sides of these equations is the value of the func-
followed by the particles are traced out by the
tion whereas the same symbols on the right-hand
position vectors x according to (7.57). We do not
sides signify the function itself. After some
know this functional relationship for the particles
unknown duration of time, slip ceased on the fault
near the fault (Fig. 7.13), so must be content with
and the conguration of particles attained that
a two-state description of the proposed elastic
seen in the exposure today (Fig. 7.13a). The position
deformation. However, elastic models could be
vector x locates the particle in the current state
investigated that would track the particles as they
that was located by the position vector X in the
accelerated from their initial positions, attained
initial state. This is called the referential descrip-
some peak velocity, and decelerated to their
tion of motion because it refers back to the initial
current positions. Given such a function that is
state. It also is called the Lagrangian description of
differentiable with respect to time one may calcu-
motion after the Italian mathematician Joseph-
late the velocity, v, of an arbitrary particle as the
Louis Lagrange (17361813), (Fig. 7.14). The coordi-
vector function G:
nates (X, Y, Z) are called the Lagrangian coordinates,
or sometimes the material coordinates because they
describe the initial positions of material particles.
v
x
|
t X
G(X, t) (7.58)
Here the subscript after the vertical bar indicates may suggest directions and magnitudes of rela-
x also is a function of the independent variable X, tive velocities. In this context current may be
which is held constant during partial differentia- taken within the time frame of active deforma-
tion with respect to time. The material coordi- tion and certainly does not mean the present day.
nates are held constant because we want to track As an example, consider another exposure
the velocity of a particular particle. For this from the Sierra Nevada of California where a
reason the partial derivative in (7.58) is referred to mac dike about 30 cm thick cuts the metamor-
as the material time derivative and the resulting phic host rock (Fig. 7.15a). The igneous rock is com-
velocity is called the particle velocity (Malvern, posed of a very ne-grained black groundmass
1969). To understand this viewpoint an analogy is interspersed with lath-shaped white phenocrysts
drawn to an observer, stationed at the origin of (probably feldspar) that are arranged in an evoca-
the material coordinates (X, Y) in Fig. 7.13, who tive pattern across the dike (Fig. 7.15b). Near the
measures the time rate of change of the particle at two contacts with the host rock the phenocrysts
every position along its path from X to x as the tend to have their long axes parallel to the
fault slips. contact, whereas near the mid-line of the dike the
Given a function x(X, t) that is twice differen- long axes are perpendicular to the contact.
tiable with respect to time, one may obtain the Furthermore, there appears to be a greater con-
particle acceleration, a: centration of phenocrysts near the mid-line. The
geological inference is that magma from an
a | |
2x

v
G(X, t)
t2 X t X t
(7.59) unknown source forced open this fracture and
owed through it for some period of time in the
Again, we do not know the function x(X, t) for the direction that the pencil is pointing. The phe-
particles near the fault (Fig. 7.13), but an elastic nocrysts were organized into a systematic pattern
model would provide the particle accelerations that eventually became frozen in place as
according to (7.59). The particle velocity (7.58) and enough heat was lost to solidify the magma. The
particle acceleration (7.59) are kinematic quanti- mechanical inference is that the injection of
ties employed in elastic solid mechanics. magma into the dike can be described by the ow
We turn now to the ow of a viscous uid, of a viscous uid containing a number of lath-
which is viewed with the current state taken as shaped solid objects.
the reference. Under given loading conditions a We adopt a spatial description of motion which
particular rate of deformation is associated with takes the coordinates (x, y, z), and time, t, as the
the particle at each coordinate position in the independent variables (Malvern, 1969). These
current state and the initial conguration of coordinates describe positions in space such that
those particles is not prescribed. Upon unloading the point (x, y) in Fig. 7.15b is associated with a par-
each particle simply remains at rest at its current ticle of a particular phenocryst at a given time,
position. A special case is that in which the veloc- but would be associated with a particle of the
ity eld does not change with time and the rate of ground mass or a different phenocryst at a later
deformation is constant at any position in the time as the magma ows through that xed point.
ow. This is referred to as a steady state of ow for This viewpoint brings attention to given points in
the viscous uid. In the geological context, a space rather then to given particles. With each
nearly steady state may exist for some period of successive instant in time the particle that was at
time, preceded and followed by periods of accel- a given position may move away and a new parti-
erating or decelerating ow. For the structural cle may move into the eld of view, but the posi-
geologists standing on an exposure of igneous tion in space remains xed. This is called the
rock the frozen pattern of aligned xenoliths, spatial description of motion. It also is called the
phenocrysts, or vesicles may suggest how this Eulerian description of motion after the Swiss
material once owed. Similarly, for an exposure of mathematician Leonhard Euler (170783), (Fig.
highly deformed metamorphic rock, although 7.16). The coordinates (x, y, z) are called the Eulerian
ow has long since ceased, the folds or boudinage coordinates and also the spatial coordinates.
Fig 7.15 Illustration of the spatial description of motion in

(a)
which the velocity is a function of the current location and
time, vg(x, t). The igneous dike from the Sierra Nevada is
about 30 cm thick and contains a pattern of phenocrysts that
suggest the flow direction and relative magnitude.
Photograph by D. D. Pollard.
(b)
x y
(x, y)
x v
In contrast to the referential description of model magma would be described at every posi-
motion, particle paths are not a primary feature tion as a function of time, providing the vector
of the spatial description, and the initial positions function:
of particles usually are not dened. For example,
v g(x, t) (7.60)
the initial positions of the phenocrysts in the
mac dike (Fig. 7.15) are unknown. A complete The quantity dened in (7.60) is called the local
spatial description of motion would be a function velocity because it refers to the velocity at a given
that describes the velocity, v, at every position, x, location. Note that both (7.58) and (7.60) are equa-
for all times from t 0 to t the current state. In tions for velocity, v, so the particle velocity at a
our eld example we might consider a steady particular location and time is the same as the
velocity eld of a viscous uid containing lath- local velocity in that location at that time.
shaped solid objects within a conduit of xed However, the functions G(X, t) and g(x, t) are dif-
width equal to that of the dike. The velocity of the ferent because the former describes the velocity of
velocity of a given particle that follows one of the

stream lines illustrated in Fig. 5.12b changes with
position, being greater where the stream lines are
more closely spaced. This particle and every other
particle, except those at the stagnation points at
the top and bottom of the sphere, accelerate and
decelerate as they circulate within the sphere.
Clearly one cannot take the local rate of change of
velocity (7.61) as the particle acceleration (7.59) in
this particular case.
It is important to clarify the general relation-
ship between the particle acceleration (7.59) and
the local rate of change of velocity (7.61). We do
this by showing how to calculate the material
time derivative of a quantity given a spatial
description of the kinematics of that quantity
(Malvern, 1969). We start with the function g(x, t)
that describes the local velocity (7.60). The under-
lying premise is that particle motion may be
dened by a function x(X, t) for the referential
description of motion (7.57). This relationship is
substituted for the spatial coordinates, x, in the
Fig 7.16 Portrait of the Swiss mathematician Leonhard function for the local velocity to transform it to a
Euler who was born in Basel, Switzerland in 1707. The function of the material coordinates X:
coordinates (x, y, z) used in the spatial description of motion
v g[x(X, t), t] G(X, t) (7.62)
are referred to as the Eulerian coordinates. Reproduced with
the permission of the Department of Special Collections, Because we have started with a spatial description
Stanford University Library. of motion the referential description given by the
function x x(X, t) may not be known, but knowl-
edge of this function is not necessary to dene the
a given particle along its path, while the latter material time derivative of (7.62). Recalling the
describes the velocity of all particles that pass Chain Rule of calculus, if z f(x, y) but x x(r, s)
through a given location. and y y(r, s) one takes the partial derivative of z
The time derivative of the local velocity (7.60) with respect to s holding r constant as (Varberg
at a given current location is: and Purcell, 1992):
v
|
g(x, t)
t x t
(7.61)
z
| | | | |

z x

z y
s r x y s r y x s r
(7.63)
The quantity dened in (7.61) is referred to as the Making the appropriate associations for the quan-
local rate of change of velocity (Malvern, 1969) and tities in the vector function (7.62) we have:
not as the acceleration because it does not neces-
sarily dene the particle acceleration (7.59) at the
position x and time t. To appreciate this apparent
v
|
v x
| | | |

v t
t X x t t X t x t X
(7.64)
contradiction recall the velocity distribution The left-hand side is the material time derivative
(5.19) within a rising viscous sphere (Fig. 5.12). of the velocity which is, by (7.59), the particle
This is an example of steady ow, which means that acceleration, a. The rst partial derivative on the
all of the kinematic quantities are constant in right-hand side may be calculated from the local
time at every point. Thus, at every current pos- velocity (7.60), which is given. The second partial
ition x the time derivative of velocity as dened in derivative is the particle velocity (7.58), but that is
(7.61) is identically zero. On the other hand the equivalent to the local velocity, v, dened in (7.60).
The third partial derivative is the local rate of of uid mechanics, as developed using the spatial
change of velocity (7.61), which may be calculated description of motion, the operator (7.67) is used to
from (7.60). The fourth partial derivative evaluates calculate the rate of change of material properties
to one. that play roles in the fundamental principles of
Using the above interpretations for the partial conservation of mass and momentum. Apparently
derivatives in (7.64) and rearranging them, we the concepts embodied in (7.67) can be traced back
have an expression for the particle acceleration to publications of Euler (Fig. 7.16) in 1770 and
written in terms of the local velocity using a Lagrange (Fig. 7.14) in 1783 (Malvern, 1969).
spatial description of motion: The material time derivative operator (7.67)
may be applied to scalar, vector (7.65), or tensor
a |
v
v
v
t x x t| (7.65) functions that describe a property of the material
in terms of the spatial coordinates, x, and time, t
The rst term on the right-hand side is the local (Malvern, 1969). Suppose the mass density is
rate of change of velocity at the current position known as (x, t). The material time derivative
x. What distinguishes this term from the particle of this scalar quantity is:
acceleration at that position is the second term,
which is a product of the local velocity and the
spatial derivative of velocity at the current time t.
D
|
Dt t X
v

x t| (7.68)
The second term may be interpreted as the rate of The rst term on the right-hand side of (7.68)
change of velocity due to the ow of material at describes the local rate of change of density at the
velocity v through a spatially varying velocity current position x. The second term describes the
eld. For steady ow, such as that within the rate of change of density at the current time t due
rising viscous sphere (Fig. 5.12), the rst term on to the ow of material at velocity v with a spatially
the right-hand side of (7.65) is zero, so particle varying density eld. In terms of the velocity com-
accelerations are entirely due to the spatial varia- ponents (7.68) is:
tions in velocity as described by the second term. D
For example, the x-component of acceleration vx vy vz (7.69)
Dt t x y z
from (7.65) is written:
Here is it understood that the velocity compo-
v v v v nents are known functions of the current position
ax x vx x vy x vz x (7.66)
t x y z and time (7.60). Furthermore, the mass density is
The components ay and az follow by change of sub- a known function of the current position and
scripts. Here is it understood that the velocity time, so the partial derivatives are taken with the
components are known functions of the current appropriate independent variables held constant.
position and time, v g(x, t), so the partial deriva- Other scalar properties of the material are oper-
tives are taken with the appropriate independent ated upon and interpreted similarly.
variables held constant as indicated in (7.65).
The operation characterized in (7.65) for calcu- 7.3.2 Conservation of mass: the
lating the particle acceleration from the local equation of continuity
velocity may be generalized to calculate the What constraints must be imposed to assure that
material time derivative of any quantity associated the relative motions of particles in a deforming
with the material, given a spatial description of rock mass obey the fundamental law of mass con-
its kinematics: servation? To address this question we adopt the
spatial description of motion and consider a xed
D
|

Dt t X t x |
v

x t | (7.67) volume element within a deforming material con-
tinuum (Fig. 7.17). The center of the element is at
The operator D/Dt is also referred to as the substan- the arbitrary point specied by the position vector
tial derivative, but material time derivative is more x with components (x, y, z) which are the current
descriptive of its role. For example in the context coordinates. The sides of the element are parallel
v(x, t) tity is the product of mass density and the veloc-

ity, v, and so it is a function of the spatial coor-
(rvx)|x dx/2dydz dinates and time. Because the element sides are
parallel to the coordinate directions, the total
(rvx)|x + dx/2dydz mass ux through a particular side is propor-
tional to the component of velocity acting per-
dz x pendicular to that side, and to the surface area of
that side. For example, through the left side of the
z element, as viewed in Fig. 7.17, the mass rate
dy
y would be vxyz, where and vx are evaluated at
dx x x/2. The arrows normal to the sides of the
x element are meant to represent the mass rates
Fig 7.17 Schematic diagram to define the conservation of
through the entire side, not just through a point
mass as material moves through a fixed volume element with at the middle. Similarly, the total mass rate
velocity, v(x, t). For example, the mass flux through the left- through the right side would be vxyz, where
hand side of the element is the product of the mass density, and vx are evaluated at x x/2. The difference
, and the velocity component, vx, evaluated at that side between the rate in through the left side and the
times the surface area of the side. rate out through the right-hand side is:
( vx) |xx2y z ( vx) |xx2y z (7.72)
to the coordinate axes and the lengths of the sides
are x, y, and z. As material moves through the Because the velocity components vy and vz are par-
volume element, the velocity at the center is a allel to these sides, they cannot contribute to this
function of the current location and time, v(x, t). part of the mass rate.
Mass is accounted for in terms of the mass density, Now consider a set of n elements, each con-
which also is a function of the current location taining x, with successively smaller volumes such
and time, (x, t). The conservation of mass for that the volume approaches zero in the limit as n
the volume element is prescribed as follows (Bird . In this limit the largest dimension of the
et al., 1960): element approaches zero, so the volume con-
verges to the central point at x and not to a surface

rate of rate of rate of or line. The partial derivative of the x-component
mass mass mass (7.70) of the mass ux, vx, with respect to x is dened in
increase in out this limit as:
lim ( vx) |xx2 ( vx) |xx2
In other words, mass may enter and leave the ( v ) (7.73)
x x n x
element and the total mass of the element may
change with time, but mass is neither created nor Equating (7.71) and (7.72), dividing through by the
destroyed within the element. We assume that volume, and using (7.73) we have:
mass is not converted to energy in the processes

that generate geological structures (Wilczek, ( vx) (7.74)
t x
2004).
The rate of accumulation of mass in the Both the rate of change of density and the spatial
element, the left-hand side of (7.70), is measured derivative of the component of mass ux are
by the temporal derivative of density evaluated at evaluated at the current location x in this one-
x and multiplied by the volume, xyz: dimensional description of mass conservation.
Conservation of mass during motion in one

x y z (7.71) coordinate direction (7.74) is, perhaps, more inter-
t
esting than one might surmise. Because both
The right-hand side of (7.70) is accounted for using density and velocity may be functions of x, the
the mass ux per unit volume. This vector quan- derivative of their product is:
(a) vx For the arbitrary vector, u, we have:
d vx
r = uniform

u ex ey ez
x y

z
(uxex uyey uzez)
u u y uz
x x (7.77)
x y z
dx
(b) r Note that this operation is similar to the scalar
product of two vectors: each partial derivative
dr operates on the respective component of the
vx = uniform
vector and the resulting sum is a scalar quantity.
Using the del operator, the rate of change of
x
x d x/2 x + d x/2 density may be written (Bird et al., 1960):

Fig 7.18 Graphs to illustrate the two terms on the right- ( v) (7.78)
hand side of (7.75) that account for the mass conservation in t
one-dimensional flow. (a) Rate of change of density as a This scalar equation is a spatial description of the
function of a spatial change in velocity with uniform density. conservation of mass because it describes
(b) Rate of change of density as a function of a spatial change
changes at a xed point in space and both the
in density with uniform velocity.
density and the velocity components are
expressed as functions of the spatial coordinates.

v The relationship in (7.78) is called the equation of
x vx (7.75)
t x x continuity.
In component form the continuity equation
The rst term in parentheses describes the rate of
(7.78) is written:
change of density if the density does not vary with
x, but the velocity does vary with x (Fig. 7.18a). For
example, where the velocity increases with x, the

t

x
( vx) ( vy) ( vz)
y z (7.79)
material in the vicinity of x is stretched, so the
density there decreases with time in proportion to Because both the density and the velocity com-
vx/x. The second term in parentheses describes ponents may be functions of the spatial coordi-
the rate of change of density if the velocity does nates, the partial derivatives in (7.79) expand as
not vary with x, but the density does vary with x follows:
(Fig. 7.18b). Where the density increases with x,
motion in the x-direction carries material with
lesser density into the vicinity of x, so the density

t

vx vy vz

x y z

vx vy vz
x y

z
there decreases with time in proportion to vx and (7.80)
/x. Density at a xed point in a material con-
tinuum can change by either one (or both) of Notice that the sum of the left-hand side and the terms
these two independent mechanisms while mass is in the second parentheses on the right-hand side are
conserved. the material time derivative of density as dened in
The one-dimensional relationship (7.74) is gen- (7.69). Therefore we can write the continuity equation
eralized for mass uxes through all six sides of the as (Malvern, 1969):
element using the differential operator, , (called
del) on the mass ux vector, v. When operating
on a vector quantity is dened as the vector (Bird
D
Dt

vx vy vz

x y z
v (7.81)
et al., 1960):
Here the operation v is called the divergence of

ex ey ez (7.76) the velocity vector eld which sometimes is
x y z
written div v. The product of the density and the
divergence of the velocity may be interpreted as v(x,t)

the rate of change of density of the particle at the vz(vx)|z+z/2xy vy(vx)|y+y/2xz
point x because the material in the vicinity of
this particle is stretching in one or more of the vx(vx)|x-x/2yz
coordinate directions, and this stretch is not
exactly compensated for by a contraction in the vx(vx)|x+x/2yz
other coordinate directions. Here it is understood
that the spatial derivatives of the velocity com- z x
ponents are evaluated at the current position
z y
and time.
Perhaps the most common postulate employed y x vz(vx)|z-z/2xy
in setting up models in structural geology involv- vy(vx)|y-y/2xz
ing viscous uid mechanics is that the rock is x
incompressible. This means that the density in the
innitesimal element surrounding any particle Fig 7.19 Schematic diagram to define the conservation
of linear momentum as material moves through a fixed
does not change with time, so the right-hand side
volume element with momentum, v(x, t). For example,
of (7.81) is identically zero:
the x-component of momentum flux through the left-hand
vx vy vz side of the element is the product of the velocity
v0 (7.82) component, vx, and the momentum, vx, evaluated at that
x y z
side times the surface area of the side. The x-component of
momentum flux includes terms from all other sides of the
In other words the divergence of the velocity
element.
vector eld is zero. Near any particle in the mater-
ial continuum and for all relevant times the veloc-
ity gradients are constrained such that a stretch in
one coordinate direction is compensated exactly analogous equation for a deformable material
by contractions in the other coordinate directions. continuum by adopting the spatial description of
Equation (7.82) assures conservation of mass motion and considering a volume element (Fig.
for the incompressible but deformable material 7.19) that is xed in space with respect to the coor-
continuum. dinate origin. In a word equation we have (Bird et
It follows from (7.81) and (7.82) that the mater- al., 1960):
ial time derivative of density is zero for the incom-

pressible material, D/Dt 0. For a body that is rate of rate of
homogeneous with respect to density and incom- momentum momentum
pressible: increase in

rate of resultant
0, and 0 (7.83) momentum of all (7.84)
x y z t
out forces
These are the most constrained conditions for the
material continuum subject to conservation of Each term in (7.84) is taken per unit volume. The
mass: the mass density is uniform in space and center of the element is at an arbitrary point
constant in time. specied by the position vector x with compo-
nents (x, y, z) which are the current coordinates.
The sides of the element are parallel to the coor-
7.3.3 Conservation of linear momentum: dinate axes, and the lengths of the sides are x, y,
the equations of motion and z. The momentum associated with the point
Conservation of linear momentum for the rigid at the center of the element is a vector function
body is described by (7.20). Here we derive the of the current location and time, v(x, t), with
components in all three coordinate directions. lim vx( vx) |xx2 vx( vx) |xx2
v ( v ) (7.87)
The rate of increase of the momentum is: x x x n x
Multiplying both sides of (7.87) by the volume and
( v) (7.85)
t comparing this to (7.86), we see how the negative
This vector accounts for the left-hand side of of the left-hand side of (7.87) accounts for one com-
(7.84). ponent of the rate of x-momentum change. The
As material moves through the xed element components of the rate of x-momentum change
(Fig. 7.19), momentum is carried in and out paral- through the front and back of the element, and
lel to the three coordinate directions in proportion through the bottom and top of the element, are
to the respective velocity components. This trans- similarly dened so the net rate of change of x
port of momentum is measured by the momentum momentum is:

ux per unit volume, which is the product of the v ( v )y z x vy( vx)x z y
velocity and the momentum, v( v). For example, x x x y

on the left-hand side of the element we have vx(vx), vz( vx)x y z (7.88)
where and vx are evaluated at x x/2. This quan- z
tity times the area, yz, is the momentum ux Here it is understood that the derivatives are eval-
through the left-hand side, vx(vx)yz. Similarly, uated at the point x. The net rate of change of y
the momentum ux through the right-hand side is and z momentum each have three components
vx(vx)yz, where and vx are evaluated at x x/2. that are found using a similar procedure.
We account for the difference between the rate of The one-dimensional relationship (7.88) for the
momentum in through the left side and the rate rate of change of momentum is generalized to
out through the right-hand side as: three dimensions using the differential operator
on the momentum ux, v( v). Because the
vx( vx) |xx2y z vx( vx) |xx2y z (7.86) momentum ux is a product of two vectors,
referred to as a dyadic product, this operation is
Because the velocity components vy and vz are par- somewhat different from that dened in (7.77).
allel to these sides, they cannot contribute to this The dyadic product is a special form of second-
part of the rate of momentum change. On the rank tensor with nine components. For example,
other hand, vy can carry x momentum, vx the dyadic product of the two arbitrary vectors, u
through the front and back sides of the element, and w is (Bird et al., 1960):
and vz can carry x momentum, vx through the

uxwx uxwy uxwz
bottom and top of the element. The pairs of
uw uywx uywy u y wz (7.89)
arrows normal to the sides of the element in Fig.
u z wx uzwy u z wz
7.19 are meant to represent these three uxes of x
momentum across the element. The order of the Using the del operator (7.76) on this dyadic
two velocity vector components is in keeping with product we have:
an onin subscript convention: e.g. vy(vz)xz is
the momentum ux on a side with normal par-
allel to the y-coordinate of the momentum in
uw
uxwx uywx uzwx
x

y

z
ex

the z-direction. uxwy u ywy uzwy
To convert the nite difference in (7.86) to a ey (7.90)
x y z
partial derivative a set of n elements (Fig. 7.19) is
considered with successively smaller volumes
that contain x and approach zero in the limit as n
uxwz uywz uzwz
x

y

z
ez
so the elements converge on the central point
at x. In this limit the partial derivative of the The rate of change of momentum per unit
momentum ux, vx(vx), with respect to x is: volume is found by dividing (7.88) and its coun-
(szx)|z +z/2dxdy (syx)| y +y/2dxdz (xx) |xx2 y z (xx) |xx2y z (7.92)
As the successively smaller elements converge on

the point x, the partial derivative of the stress
(sxx)|x +x/2dydz component, xx, with respect to x is:
(sxx)|x x/2dydz
(rgx*)|xdxdydz xx lim (xx) |xx2 (xx) |xx2
dz x (7.93)
x n x
z dy Multiplying both sides of (7.93) by the volume and
y dx comparing this to (7.92) we see how the left-hand
(szx)|z z/2dxdy
side of (7.93) accounts for part of the net force in
(syx)|y y/2dxdz the x-direction. The x-component of force due to
x
shear stresses on the front and back of the
Fig 7.20 Schematic diagram to define the resultant force element, and on the bottom and top of the
acting on fixed volume element. For example, the x-
element, are similarly dened with reference to
component of the resultant force on the left-hand side of the
Fig. 7.20 so the resultant of surface forces in the x-
element is the product of the normal stress component, xx,
and the surface area of the side. The x-component of direction is:
resultant force includes terms from all other sides of the
( ) y z x (yx) x z y (zx) x y z
element and involves both shear and normal stresses. The x xx y z
body force component per unit volume, g*x, times the
(7.94)
volume also contributes to the resultant force.
Here it is understood that the derivatives are eval-
uated at the point x. The resultants of surface
forces in the y- and z-directions are found by
terparts by the volume of the element and using
similar procedures.
the del operator:
The one-dimensional relationship for the

surface force (7.94) is generalized to three dimen-
[v (v)] v ( v ) v ( v ) v ( v ) e sions using the differential operator (7.76) on
x x x y y x z z x x
the stress tensor, , following the procedure intro-

v ( v ) v ( v ) v ( v ) e
x x y y y y z z y y duced in (7.91):

v ( v ) v ( v ) v ( v ) e
x x z y y z z z z z

xx yx zx
x

y

z x
e

(7.91) xy yy zy
x

y

z y
e
This vector accounts for the rst and second terms
on the right-hand side of (7.84).
The nal term to evaluate in (7.84) is the
xz yx zz
x

y

z z
e

(7.95)
resultant force. The resultant surface force in This vector is the resultant surface force per unit
the x-coordinate direction is related to the x- volume. The body force per unit volume acting on
components of the tractions acting on the sides of the element is taken as:
the xed volume element (Fig. 7.20). It is conven-
g* (7.96)
tional to use the equivalent stress components
instead of the traction components to account for This vector accounts for the other part of the last
the surface forces. Thus, for example, the net term of (7.84).
surface force associated with the normal stress Collecting terms from (7.85), (7.91), (7.95), and
component, xx, on the left- and right-hand sides (7.96) we have a statement of the conservation of
of the element is: linear momentum (Bird et al., 1960, p. 78):
v vy v
(v) [v (v)] g* (7.97) v x x vx v x z
t x y z
This is a spatial description of the conservation of

vxvx vxvy vxvz (7.100)
linear momentum because it describes changes at x y z
a xed point in space where the density, velocity, v v v
stress, and acceleration of gravity are expressed as v x x vy x v z x
x y z
functions of the spatial coordinates. This is
referred to as the equation of motion because it These terms are rearranged as follows:
governs the motion of a deforming material
subject to conservation of linear momentum. It is
important to note that (7.97) does not depend
v v v
vx x vy x vz x
x y z
upon particular properties of the material, such
as elasticity or viscosity, so it applies to any body
that can be suitably characterized as a material
vx vx
x

vx
x
vy
y
v
vy z vz
y z

z
continuum. (7.101)
Written out in component form, the rst of
the three equations of motion given by (7.97) is: The equation of continuity (7.80) shows that the
second term on the right-hand side of (7.99) is
equal to the second term of (7.101), so these two
( v ) vx( vx) vy( vx) vz( vx)
t x x y z terms are eliminated. What remains is the mater-
ial time derivative of the velocity multiplied by
xx yx zx
g*x (7.98) the mass density. Similar results are obtained for
x y z
the other two equations of motion such that (7.97)
The second and third equations follow by cyclic may be rewritten:
substitution of the subscripts. On the left-hand
Dv
side of (7.98) is the time derivative of the x-compo- g* (7.102)
Dt
nents of momentum per unit volume. The rst
three terms on the right-hand side are spatial Using indicial notation the equations of motion
derivatives of the momentum ux per unit in this form are:
volume. The next three terms are the spatial deriv-
D vi ji
atives of the stress components and the nal term g*i (7.103)
Dt xj
is the component of gravitational body force per
unit volume. The equations of motion (7.97) may Here it is understood that the indices i and j range
be written in a more general form by replacing the over the three spatial coordinates (x, y, z). Written
gravitational body force, g*, with a generic body in component form we have:
force, F(b), but applications to structural geology
usually require only the gravitational body force, Dvx xx yx zx
g*x (7.104)
and this usually is taken as constant in time and Dt x y z
uniform in space. Dvy xy yy zy
The equations of motion can be rewritten to g*y (7.105)
Dt x y z
refer to a particle traveling with the deforming
material at the position x and current time t. For Dvz xz yz zz
g*z (7.106)
example, consider (7.98) and expand the partial Dt x y z
derivative on the left-hand side: This form of the equations of motion is credited to
Augustine-Louis Cauchy (17891857) (Fig. 7.21) and
v
( v ) x vx (7.99) is referred to as Cauchys First Law of Motion
t x t t
(Malvern, 1969, p. 214). Recall from (7.65) that the
The rst three terms on the right-hand side of material time derivative of velocity is the particle
(7.98) are expanded as: acceleration, a, so the left-hand side of (7.102) is
y Surface, S
x
dS
n
Volume, V
x
t(n)
dV
rv(x, t)
rg*
Fig 7.22 Schematic diagram to define the conservation of
linear momentum for a fixed volume element based upon the
integration of the tractions, t(n), acting on surface elements,
S, over the surface, S, and the integration of the unit
Fig 7.21 Portrait of the French mathematician Augustine-
weights, g*, acting on volume elements V, over the
Louis Cauchy who was born in Paris, France, in 1789
volume, V.
(OConnor and Robertson, 2004). Cauchys First Law of
Motion is given in (7.104) through (7.106).
continuum one imagines a set of particles of

the mass per unit volume times the particle accel- given total mass occupying a volume, V, with
eration. The terms on the right-hand side are the bounding surface, S, at a given instant in time, t
resultant surface and body forces per unit volume. (Fig. 7.22). The momentum is given using the
Therefore one can interpret (7.102) as a statement spatial description of motion as v(x, t) and this is
of Newtons Second Law, ma F, set here in the integrated over the volume. A distribution of
context of a deformable continuum. Again it is surface forces per unit area is represented by the
important to emphasize that these equations of equivalent tractions, t(n), and this is integrated
motion are independent of constitutive proper- over the surface. A distribution of body forces per
ties so they apply to any material that can be ide- unit volume is taken as that due to the unit
alized as a continuum. weight, g*, and these are integrated over the
One of the basic postulates of continuum volume. Then, the momentum principle is
mechanics is the momentum principle (Malvern, expressed as:
1969, p. 213):

D
vdV t dS g*dV
the time rate of change of the total momentum of a Dt (7.107)
V S V
given set of particles equals the vector sum of all the
external forces acting on the particles of the set, pro-
On the left-hand side the material time derivative
vided Newtons Third Law of action and reaction
is used because the momentum is given using the
governs the internal forces.
spatial description of motion.
Recall that a similar statement, applied to a single Using the steps that we describe in the next
particle, was expressed in (7.6) and again in (7.20) section the stress tensor is substituted for the trac-
for the rigid body. In the context of a material tion vector on the right-hand side of (7.107) and
the surface integral is transformed to a volume tions, t(n), acting on the surface S. The cross
integral using the divergence theorem (Malvern, product of the position vector and the traction, x
1969, p. 200). Reynolds transport theorem is used t, is integrated over the surface as in (7.35) to
to move the material time derivative inside the nd the resultant torque due to surface forces. A
integral on the left-hand side (Malvern, 1969, distribution of body forces per unit volume is rep-
p. 210). Because the integrals refer to any arbitrary resented by the unit weight, g*, and the cross
volume the integrand on the left-hand side must product x g* is integrated over the volume as in
equal the sum of the integrands on the right at (7.36) to nd the resultant torque due to body
each point in the material continuum. The result- forces. Newtons Third Law is invoked to argue
ing equation is Cauchys First Law of Motion that the forces due to internal interactions are
(7.102). equal, opposite, and collinear so they produce no
resultant torque on the body.
7.3.4 Conservation of angular The underlying postulate is that the material
momentum: symmetry of the time derivative of the total angular momentum is
stress tensor equal to the vector sum of the resultant torques:
Conservation of angular momentum for the ma-

terial continuum is described using the momen- D
(x v) dV (x t) dS (x g*) dV
tum principle, given in the previous quotation, Dt
V S V
with the word angular inserted before momen- (7.108)
tum and the words all external torques substi-
tuted for all external forces. A similar statement, The material time derivative is used because the
applied to a single particle, was expressed in (7.11) angular momentum must be that associated with
and again in (7.30) for the rigid body. One might particles.
suppose, based on what we have derived in the pre- The next step in the derivation involves writing
vious section for conservation of linear momen- the vector cross products in terms of the vector
tum that the conservation of angular momentum components. Recall from (3.27) that the vector
in a deforming continuum would lead to a set of product u of two arbitrary vectors, v and w, is
equations relating density, velocity, stress, and dened in terms of their Cartesian components as:
gravitational acceleration in a form analogous to u v w (vywz vzwy)ex (vzwx vxwz)ey
Cauchys First Law of Motion (7.102). However, a
(vxwy vywx)ez (7.109)
derivation based on extending this concept to the
continuum (Malvern, 1969, p. 215) leads to a This expansion of (7.108) would involve a large
remarkably simple set of equations, relating only number of terms, but is condensed using indicial
the shear stress components to one another such notation and the permutation symbol epqr:
that the stress tensor is symmetric. The derivation

and its implications are described here. 0, when any two indices are equal;
In the context of a material continuum con- 1, when indices are (x, y, z),
sider a set of particles of given total mass occupy- epqr
( y, z, x), or (z, x, y);
ing a volume, V, with bounding surface, S, at a 1, when indices are (x, z, y),
given instant in time, t (Fig. 7.22). For the particle
(y, x, z), or (z, y, x).
at x, the momentum per unit volume, v, is a
(7.110)
function of the current coordinates and time, so
the angular momentum per unit volume at that Employing (7.110) the components of the vector
location, dened as x v, also is a function of x product u are written:
and t. This quantity is integrated over the volume
up epqrvqwr (7.111)
to obtain the total angular momentum. The body
is acted upon by a distribution of surface forces Here it is understood that the indices p, q, and r
per unit area represented by the equivalent trac- range over x, y, and z and that repeated indices on
the right-hand side imply summation over that referred to as the divergence theorem or Gausss
range. Each component of u is composed of nine theorem (Malvern, 1969, p. 200). Written in terms of
terms but eight of these are reduced to zero by the components of an arbitrary vector, u, this
(7.110) and the remaining terms are those found in theorem is:
(7.109).

The integral equation for the conservation of u
uini dS i dV (7.116)
angular momentum (7.108) is rewritten using the xi
S V
vector components and the permutation symbol On the left-hand side the component of the vector
as: u directed normal to the surface S is integrated
over that surface. On the right-hand side the diver-

D
e x v dV epqr xqtr dS epqr xqg*r dV gence of the vector u is integrated over the volume
Dt pqr q r
V S V V bounded by that surface. Applying the diver-
(7.112) gence theorem to (7.115) we have:

The left-hand side of (7.112) can be rearranged xqsr
epqr xqsrns dS epqr dV
using Reynolds transport theorem (Malvern, 1969, xs
S V
p. 210) which applies to any scalar, vector, or
tensor function. Given such a function, Q, of
the current coordinates and time, the material

epqr xq
V
sr
xs
sr
xq
xsdV
time derivative of the volume integral may be

sr
rearranged according to this theorem as follows: epqr (xq qssr) dV (7.117)
xs
V

D DQ
Q dV dV (7.113) This transformation enables us to consider the
Dt Dt
V V right-hand side of (7.112) as a single volume inte-
gral:
Therefore the left-hand side of (7.112) may be

written:

sr
epqr xq g*r qr dV (7.118)
xs

D D
e x v dV epqr (xqvr) dV V
Dt pqr q r Dt The last term in square brackets follows from the
V V
fact that qssr qr.
V
Dv
epqr xq r vr
Dt
Dxq
Dt
dV Using (7.114) and (7.118) the rearranged and
transformed integral equation for the conserva-
tion of angular momentum (7.112) becomes:

Dvr
epqr xq dV (7.114)
Dt

Dvr

sr

V
In the second step Dxq/Dt vq by denition and epqr xq dV epqr xq g*r qr dV
Dt xs
V V
epqrvrvq 0 using (7.110).
(7.119)
In the rst term on the right-hand side of
(7.112) Cauchys Formula, tr srns, is used to
Buried within this equation is Cauchys First Law
replace the traction vector components with the
of Motion (7.103), such that the left-hand side of
stress tensor components, sr, where ns are the
(7.119) exactly cancels the terms in parentheses on
components of the outward unit normal to
the right-hand side leaving:
the surface S:

epqrqr dV 0
(7.120)

epqr xqtr dS epqr xqsr ns dS (7.115) V
S S This relation must hold for an arbitrary volume

The surface integral of a vector quantity may be V and therefore the integrand must be zero.
transformed to a volume integral using what is Employing (7.110) we have:
7.4 Field equations for the elastic

p x, yz zy 0
epqrqr 0, so for p y, zx xz 0 (7.121)
p z, xy yx 0
solid and viscous fluid
In general:
The equations of motion for the material contin-
uum, (7.103) and (7.122), follow from conservation
qr rq,q r (7.122)
of linear and angular momentum and are inde-
This expression of conservation of angular pendent of the constitutive properties of the
momentum is credited to Augustine-Louis Cauchy material in motion. In the context of structural
(1789857), (Fig. 7.21), and is referred to as geology these equations account equally well for
Cauchys Second Law of Motion (Malvern, 1969, the dramatic motion associated with earthquake
p. 214). ruptures and the imperceptible motion of tec-
It is noteworthy that the preceding derivation tonic plates between major earthquake events.
does not presume a homogeneous state of stress However, to apply these equations to a particular
or static equilibrium, and it does not neglect body problem it is necessary to select an appropriate set
forces as we did in the Chapter 6 to rst derive of constitutive equations that relate the stress com-
(7.122) and justify the symmetry of the stress ponents to the strain or rate of deformation com-
tensor. The symmetry of the stress tensor applies ponents and thereby explicitly dene the
to the material continuum with surface and body mechanical behavior of the material. The consti-
forces that induce heterogeneous stress states and tutive equations are used to eliminate the stress
both linear and angular accelerations. This con- components from the equations of motion,
clusion is independent of the material properties thereby reducing the number of dependent vari-
so applies to elastic solids and viscous uids. ables. In this way more specialized equations of
Perhaps because it is so broadly applicable, and motion are derived for elastic solids, viscous
because it does not contain the kinematic quanti- uids, and other materials.
ties, (7.122) is rarely mentioned as an integral part In some circumstances the constitutive prop-
of the equations of motion. The symmetry of the erties can be measured directly in laboratory
stress tensor usually is tacitly presumed in appli- tests, so the appropriate behavior can be identi-
cations of continuum mechanics to structural ed and used. In other cases one postulates the
geology and that is the case throughout this text- constitutive properties, based upon inferences
book. In what follows, when referring to the equa- from eld observations. In Chapters 8 and 10 we
tions of motion, we mean (7.102) and equations describe the elastic solid and the viscous uid in
derived from it, but (7.122) is understood. some detail, including testing methods designed
There are two caveats regarding the symmetry for both the laboratory and the eld. Here we
of the stress tensor. There must be no distributed derive the governing equations for the linear
surface and body couples that would lead to isotropic elastic solid and the linear isotropic
couple stresses and a non-symmetric stress tensor viscous uid from the more general equations of
(Malvern, 1969, p. 217). Furthermore, for problems motion derived in the previous section. Those
of elasticity one adopts the referential description equations, you will recall, are already specialized
of motion rather than the spatial description used to isothermal and isochemical conditions. In
here, and one supposes that equilibrium is estab- doing so we show how these equations are put
lished for the body in the undeformed state. For into forms that have immediate practical appli-
cases where the strains are taken as innitesimal cations in structural geology. Our focus is limited
it may be plausible to argue that the initial coor- to these two most elementary constitutive laws
dinates and the current coordinates of particles because they provide an appropriate introduc-
are so little different that this supposition is tion to the subject and because they are adequate
reasonable. Examples of instability, such as buck- approximations for the deformation or ow
ling of thin beams and plates, demonstrate that accompanying the development of many geologi-
this supposition is not always valid. cal structures.
7.4 ELASTIC AND VISCOUS FIELD EQUATIONS 277
7.4.1 Field equations for the linear acknowledged in applications of the linear theory
isotropic elastic solid and could prove to be problematic.
The Lagrangian (material) coordinates X are Further simplications of the equations of
taken as independent variables. Because there are motion (7.103) are achieved by considering the
only two states to compare, and it is the displace- material derivatives of the kinematic quantities.
ment that measures the change in position of any For example, components of the particle velocity
particle from the reference to the current state, are linearized as:
the equations of motion (7.103) must be written in
terms of the displacement u rather than the Dux ux u u u u
vx vx x vy x vz x x
velocity v. Because the equations of motion (7.103) Dt t x y z t
are written in terms of a spatial description of (7.125)
motion, conservation of momentum is satised
in the current conguration of the deforming Here products of the velocity components and the
body. To evaluate the conditions under which displacement gradients must be small compared
conservation of momentum would be satised in to the time rate of change of displacement.
the reference state, the stress tensor is written in Similarly, components of the particle acceleration
a form other than that introduced by Cauchy. are linearized as:
Consideration of this so-called PiolaKirchhoff
stress tensor is beyond the scope of this text so we Dvx vx v v v v 2u
ax vx x vy x vz x x 2x
refer the interested reader to other sources for Dt t x y z t t
the details of the evaluation (Fung, 1965; (7.126)
Malvern, 1969). In summary, equations of motion
of the same form as (7.103) are written for the ref- Here products of velocity components and the
erence state using the PiolaKirchhoff stress velocity gradients must be small compared to the
tensor. There equations may be transformed to time rate of change of velocity.
those using the Cauchy stress tensor, but terms Given the simplications and linearizations
appear that may be approximated, for example, described above, the equations of motion in the
as: reference state are written:
ux uy uz 2ui ji

1 higher-order terms 1 g*i (7.127)
X Y Z t2 Xj
(7.123) It is understood that the density and acceleration
of gravity are evaluated as functions of the mater-
In other words, displacement gradients with
ial coordinates, and the partial derivatives of the
respect to the material coordinates, Xi, must be
displacement components with respect to time are
small compared to unity. This is the same approx-
taken with the material coordinates held constant.
imation that is made to reduce nite strains to the
What may seem like inconsequential changes of
innitesimal strains, and therefore it is in keeping
notation between (7.127) and (7.103) involve the
with the well-understood postulates of linear elas-
omission of terms as described above that could be
ticity theory. In addition, however, terms appear
signicant in a given application. Each of these
in the equations of motion that may be approxi-
should be evaluated to understand the degree of
mated, for example, as:
accuracy of the solutions that are employed.

ux u u The right-hand side of (7.127) contains spatial
1 x x xx (7.124) gradients of the Cauchy stress components
x xx y yx z zx
which are eliminated in favor of the displace-
Here products of displacement gradients with ments by rst relating the displacement gradi-
respect to the spatial coordinates, xi, and stress ents to the strain components and then relating
components must be small compared to the the strains to the stresses. These steps involve
stress components. This approximation is rarely additional linearizations that are consistent
with those mentioned above. The general equa-

tion for the nite strain tensor using the refer-
ential description of motion is (Malvern, 1969,
p. 160):
Eij

1 ui uj uk uk

2 Xj Xi Xi Xj (7.128)
The Eij are referred to as the Lagrangian nite

strain components, because the partial deriva-
tives are taken with respect to the Lagrangian
(material) coordinates, Xi, and this description of
motion is associated with Joseph-Louis Lagrange
(Fig. 7.14). Particular longitudinal and shear com-
ponents of the strain tensor are linearized to nd
the corresponding innitesimal strain compo-
nents, for example, as follows:

2 2 2
ux 1 ux u y uz ux
Exx
X 2 X X X X
Exy

1 ux uy
2 Y X

1 ux ux uy uy uz uz

2 X Y X Y X Y

1 ux uy Fig 7.23 Disputed portrait of the English natural scientist
(7.129) Robert Hooke who was born in 1635 on the Isle of Wight,
2 Y X
England (see Phillip Ball, Nature, v. 433, p. 197, Jan. 2005).
Note that the non-linear terms are products of the The linear form of Hookes Law for the isotropic elastic
material is given in (7.131).
displacement gradients and these must be small
compared to the displacement gradients them-
selves. With approximations such as these the
innitesimal strain tensor, ij, is related to the dis- Newton. Chapter 8 is devoted to a discussion of the
placement gradients as: measurement of elastic material properties and
various forms of Hookes Law. The general linear
ij
1 ui uj

2 Xj Xi (7.130) form is simplied here for an isotropic material:
one in which the elastic constants are not depen-
A nite strain tensor also may be written using dent upon direction. For such a material the
the spatial description of motion and when a stressstrain relationships are:
similar linearization is carried out an equation of
the same form as (7.130) emerges in which the ij 2Gij kkij (7.131)
partial derivatives are taken with respect to the Here G is the elastic shear modulus and is Lams
Eulerian (spatial) coordinates, xi. Typically, the constant. These are properties of the material and
distinction between these two descriptions of the they have the same units and dimensions as
kinematics is overlooked in applications of linear stress. Expanding (7.131), typical normal and
elasticity. shear components of stress are related to the
For the linear elastic material the stress com- innitesimal strains as:
ponents are related to the innitesimal strain
components using constitutive equations called xx 2Gxx (xx yy zz)
Hookes Law after Robert Hooke (Fig. 7.23) the (7.132)
xy 2Gxy
English polymath and contemporary of Isaac
The three equations of motion (7.127), six kine-

matic equations (7.130), and six constitutive equa-
tions (7.131) form a complete set of eld equations
for the linearized isotropic and isothermal elastic
material. The independent variables are the three
material coordinates and time (X, Y, Z, t). The
dependent variables (unknowns) are the three dis-
placement components (ux, uy, uz), the six innites-
imal strain components (xx, yy, zz, xy, yz, zx), and
the six stress components (xx, yy, zz, xy, yz, zx).
Usually, the mass density, , the components of
gravitational acceleration, g*i, and the two elastic
constants, G and , are taken as given by labora-
tory or eld data, and in most applications these
quantities are postulated to be uniform in space
and constant in time.
There are fteen eld equations and fteen
unknowns for the elastic boundary value problem
as posed above. This problem is put in a more prac-
tical form by replacing the stress components in
(7.127) with displacement components using the
kinematic equations (7.130) and Hookes Law
(7.131), for example as:
Fig 7.24 Bust of the French mathematician and engineer

u ux uy uz Claude Louis Marie Henri Navier who was born in Dijon,
xx 2G x France, in 1785 (OConnor and Robertson, 2004). His name,
X X Y Z
(7.133) along with that of Stokes, is associated with the velocity

ux uy equations of motion for the viscous fluid (7.170).
xy G
Y X
The other stress components follow by similar

steps. Then (7.127) is written:
2 u x
t 2
G

2ux 2ux 2ux

X2 Y2 Z2
(G )

2u i 2u i 2 u k 2 u x 2 u y 2uz
G (G ) g*i (7.134) g*x (7.135)
t2 XkXk XiXk X2 XY XZ

These are known as Naviers displacement equa- 2 u y 2u y 2u y 2u y
tions of motion (Fung, 1969, p. 261) after the 2
G (G )
t X2 Y2 Z2
French mathematician and scholar of engineer-
ing science Claude Louis Marie Henri Navier
(17851836), (Fig. 7.24). Apparently they were
2ux 2uy 2uz

XY Y2 YZ
g*y (7.136)
introduced by Navier in 1821 with only one elastic

constant and corrected in 1822 by Cauchy
(Malvern, 1969). It is understood that the partial

2 u z
t 2
G
2uz 2uz 2uz

X2 Y2 Z2
(G )
derivatives of the displacement components with

respect to time are taken with the material coor-
dinates held constant.
2ux

2uy 2uz

XZ YZ Z2
g*z (7.137)
In component form Naviers equations of The independent variables are the three material
motion (7.134) are: coordinates and time (X, Y, Z, t) and the dependent
variables (unknowns) are the three displacement conserved. Typically, however, the structural geol-
components (ux, uy, uz). A solution would be three ogist is not confronted with data in the eld that
equations for the displacement components as directly constrain the particle velocity or acceler-
functions of the material coordinates and time. ation. Rather it is the displacement from some
The three equations, (7.135)(7.137), are solved for inferred initial conguration to the current
an elastic body of prescribed geometry subject to conguration, as in the opening of a dike or the
boundary conditions dened at every point on the slip on a fault. Therefore, the structural geologist
exterior and interior boundaries in terms of the generally approaches problems related to the
three displacement components. Derivatives of development of structures from a point of view in
the displacement components with respect to which the left-hand sides of (7.134) are set to zero
time provide the particle velocity as in (7.125) and and Naviers displacement equations of motion
the particle acceleration as in (7.126). Derivatives become:
of the displacement components with respect to
the material coordinates (7.130) provide the inni- 2 u i 2 uk
G (G ) g*i 0 (7.138)
tesimal strain components, and the isotropic XkXk XiXk
forms of Hookes Law (7.131) provide the stress
The motion of particles is described by the dis-
components. In this way all of the relevant physi-
placement from the reference state to the current
cal quantities are accounted for as functions of
state, while the details of the path followed, veloc-
the material coordinates and time, and the fun-
ities, and accelerations are ignored.
damental laws of conservation of mass and
The restrictions imposed to derive (7.138) put
momentum are obeyed.
the problem in the realm of quasi-static equilib-
The most familiar applications of solutions to
rium. We use the prex quasi because this is
the dynamic equations of elasticity (7.134) are to
not a problem of a static rigid body, but rather
bodies set in motion by sudden loading or unload-
one in which the body deforms and the relative
ing, for example from an explosion or the impact
displacement of particles is accounted for by the
of two bodies in motion, that generate waves
strain eld. In component form the quasi-static
within the elastic solid (Achenbach, 1973), or from
versions of Naviers displacement equations are
the rapid propagation of a fracture (Freund, 1979).
identical to (7.135)(7.137) with the left-hand
Well-known geological examples include the
sides set to zero. A solution to these three equa-
motion immediately following an explosive vol-
tions would be three equations for the displace-
canic eruption, the impact of a meteor (Melosh,
ment components as functions of the material
1989), or the rupture of a fault (Li, 1987; Kostrov
coordinates. The strains are computed from the
and Das, 1988; Scholz, 1990). In these cases rock
displacements using the kinematic equations
particles close to the impulsive event are set in
(7.130) and the stresses follow from Hookes Law
motion rst, while the rest of the body is unaf-
(7.131).
fected. Seismic waves propagate outward from the
For some problems in structural geology it is
source with speeds on the order of a few kilo-
more appropriate to formulate the elastic bound-
meters per second and set the rest of the rock
ary value problem in terms of the stress compo-
mass in motion (Aki and Richards, 1980).
nents. Taking the equations of motion as (7.127)
and supposing that the products of mass density
7.4.2 Quasi-static equilibrium for the and linearized accelerations are insignicant
linear isotropic elastic solid compared to the gradients in stress and the body
The equations of motion (7.134) for the isotropic
forces per unit volume, we have:
and isothermal linear elastic material place no
restrictions on the magnitudes of the velocity or ji
acceleration of any particle in the continuum. g*i 0 (7.139)
Xj
They describe a material that may be accelerating
or decelerating, but these changes are always These equations of quasi-static equilibrium are
related to the appropriate forces, so momentum is expanded in component form as:
xx yx zx

X

Y

Z
g*x 0 (7.140)
ZX Y X
2yy yz zx xy

Y

Z
0
(7.147)
xy yy zy

X

Y

Z
g*y 0 (7.141)
2zz

XY Z X
yz zx xy

Y

Z
0
(7.148)
xz yz zz

g*z 0 (7.142) These conditions are derived by assuming the
X Y Z
existence of single-valued functions for the dis-
These three equations are not sufcient to deter- placement components with continuous third
mine the six independent stress components. It is partial derivatives (second partial derivatives of
necessary to include the so-called equations of the strain components). It can be shown (Malvern,
compatibility to solve problems of linear elasticity 1969, p. 187) that the six compatibility equations
formulated in terms of the stress components. represent only three independent conditions.
The necessity of adding equations of compati- Also, it can be proved that the compatibility equa-
bility may be appreciated by comparing the for- tions are necessary and sufcient conditions for
mulations of the elastic problem in terms of the existence of single-valued displacements in a
stresses or displacements. For the latter case, solu- simply connected body. By ruling out a uniform
tion of Naviers three equations of motion (7.138) translation or rigid rotation of the body these con-
yields three displacement components; deriva- ditions ensure that a unique displacement distri-
tives of these with respect to the material coordi- bution is derivable from the strain distribution.
nates (7.130) provide the innitesimal strains; and To obtain a consistent set of governing equa-
Hookes Law (7.131) provides the stresses without tions in terms of the stress components, the com-
ambiguity. On the other hand, given a stress state patibility equations are transformed from strains
that satises the equilibrium conditions (7.139), to stresses using Hookes Law. In the form given in
Hookes Law may be used to determine the strains (7.131) we have the stress components as a function
without ambiguity, but the kinematic equations of the strain components and the two elastic con-
(7.130) for determining the displacements present stants, G and . For substitution into the compati-
the difculty. There are six partial differential bility equations, (7.131) must be solved algebraically
equations to determine three displacement com- for the strain components. This alternate form of
ponents, so the solution is over-determined. In Hookes Law customarily is written using two dif-
general, relationships must exist among the ferent constants, Youngs modulus E and Poissons
strain components that eliminate this ambiguity, ratio , for the isotropic elastic material. For such a
and these are known as St. Venants compatibility material the strainstress relationships are:
equations (Malvern, 1969, p. 183).
1
For a derivation of the compatibility equations ij (7.149)
E ij E kk ij
we refer the interested reader to other sources
(Fung, 1965; Malvern, 1969). St. Venants compati- Youngs modulus has the same units and dimen-
bility equations for the innitesimal strain com- sions as stress, and Poissons ratio is dimensionless.
ponents are: Expanding typical normal and shear components
of strain we have:
2xx 2yy 2xy

2 2 0 (7.143) 1
Y2 X XY xx (yy zz)
E xx
2yy 2zz 2yz
2 2 0 (7.144) 1
Z2 Y YZ xy (7.150)
E xy
2zz 2xx 2
2
2 2 zx 0 (7.145) Refer to Chapter 8 for a discussion of the mea-
X Z ZX
surement of E and , and for equations relating

2xx

YZ X

yz zx xy
X

Y

Z
0
(7.146) the isotropic elastic constants, only two of which
are independent.
The derivation of the compatibility equations boundary conditions are dened as traction dis-
in terms of the stress components is given in tributions on the internal and external bound-
detail elsewhere (Malvern, 1969, p. 502). In short, aries of the body. Because of the inherent difculty
the stress components are substituted for the of solving the compatibility equations in addition
strain components using (7.149) and the equilib- to those of equilibrium the number of analytical
rium equations (7.139) are used to simplify the solutions for three-dimensional problems with
resulting equations and nd the so-called stress components as the dependent variable is
BeltramiMichell compatibility equations for the small. The more common approach is to take the
isothermal and isotropic linear elastic material. three displacement components as the dependent
In the most general form of these equations there variables and solve Naviers equations of motion
are partial derivatives of the body force per unit (7.138). On the other hand many approaches and
volume with respect to the material coordinates. solutions exist for two-dimensional problems,
To be consistent with our development thus far examples of which are presented in Chapter 8.
the body force is taken as mass density times grav-
itational acceleration, g*, which is postulated to
be uniform in space and constant in time. Under 7.4.3 Equations of motion for the linear
these conditions terms containing the body force isotropic viscous fluid
drop out of the BeltramiMichell compatibility Fluid ow is investigated by focusing attention on
equations which reduce to: the current state of the body and ignoring what-
ever might have been described as an initial or ref-
1 2
2ij 0 (7.151) erence state. Therefore the spatial description of
1 iXj
motion is adopted and we take the Eulerian
As mentioned above, these six equations repre- (spatial) coordinates and time (x, y, z, t) as the inde-
sent only three independent conditions. These pendent variables. This is one of several ways in
compatibility equations and the equilibrium which the analysis of uids differs from that of
equations (7.139) form a complete set for three- elastic solids, for which the referential descrip-
dimensional problems in elastic theory where the tion of motion is adopted. Instead of choosing the
six stress components are the dependent vari- displacement components of particles, the veloc-
ables. ity components at given positions are chosen as
Expanding typical members of (7.151) in com- dependent variables. Rather than the innitesi-
ponent form we have, for example: mal strains, the fundamental kinematic quanti-
ties used to describe the deformation are the
2
2
2 2

2 2 xx
X Y Z
1 2
( yy zz) 0
1 X2 xx
components of the rate of deformation tensor
(Malvern, 1969, p. 145):
2
2
2 2
2 2
X Y Z xy
1 2
( xy zz) 0
1 XY xx Dij

1 vi vj

2 xj xi (7.153)
(7.152)
Note both the similarity and the difference
Note that these relationships depend entirely between this quantity and the strain as dened in
upon second derivatives of stress components (7.130). The innitesimal strain, ij, is proportional
with respect to the material coordinates. to partial derivatives of the displacement compo-
Therefore functions for the stress components nents, ui, with respect to the material coordinates,
that are constant or linear in the material coordi- Xi, whereas the rate of deformation, Dij, is propor-
nates will automatically satisfy the compatibility tional to partial derivatives of the velocity compo-
conditions (Timoshenko and Goodier, 1970, nents, vi, with respect to the spatial coordinates,
Chapter 9). Such functions are solutions to the xi. Furthermore, the rate of deformation is not
three-dimensional elastic problem if they satisfy limited to small velocity gradients in the manner
the equilibrium conditions (7.139) and the pre- that the innitesimal strain is limited to small dis-
scribed boundary conditions. In these cases the placement gradients.
Perhaps the most fundamental mechanical

distinction between a uid and a solid is that the
greatest shear stress is zero everywhere within a
uid at rest and within a uid in a state of
uniform velocity. This property sometimes is
described by stating that a uid at rest (or in
uniform motion) is incapable of supporting a
shear stress. Recall from Chapter 6, Table 6.1, that
the maximum shear stresses are dened in terms
of the principal normal stresses as:
1
2 (1 3), 12 (2 3), 12 (1 2) (7.154)
By denition 1 2 3, so all of the principal

shear stresses are positive or zero. If the greatest
principal shear stress is zero, then all three prin-
cipal shear stresses must be zero, and the stress
state must be isotropic: 11 22 33 for any ori-
entation of the coordinate axes, so principal direc-
tions are not dened. In this mechanical context,
the static pressure, p0, is dened as the negative of Fig 7.25 Photograph of the Irish mathematician and fluid
the uniform normal stress: dynamicist George Gabriel Stokes who was born in Skreen,
Ireland, in 1819 (OConnor and Robertson, 2004). His name,
p0 ij ij (7.155) along with that of Navier, is associated with the velocity
equations of motion for the viscous fluid (7.160).
If the uid is in motion such that the velocity is

not uniform, the mean normal pressure, p, is dened Using (7.158) guarantees that the thermodynamic
as the negative of the mean value of the principal pressure, p, reduces to the static pressure, p0,
normal stresses: when the uid comes to rest or to a state of
uniform velocity, but it may not be equal to the
p 13 (1 2 3),12 (1 3) 0 (7.156) mean normal pressure (7.156) dened as a func-
tion of the normal stress components for a uid in
In this case the greatest shear stress must be motion. In keeping with the postulates employed
greater than zero and the state of stress is not earlier in this chapter we limit our attention to
isotropic. barotropic ows, those in which the thermo-
Pressure also is dened in a thermodynamic dynamic pressure is independent of temperature,
context. For a static uid in thermodynamic equi- so the equation of state is of the form:
librium the pressure, absolute temperature, and
f (p, ) 0 (7.159)
mass density are related by an equation of state
(Malvern, 1969, p. 295): These thermodynamic characterizations of pres-
sure must be reconciled with the mechanical
F ( p0, T, ) 0 (7.157) denitions of pressure from the preceding para-
graph and this can be done by considering the
For investigations of uid ow it is assumed that constitutive law for the uid.
the so-called thermodynamic pressure, p, is dened The constitutive law for the viscous uid con-
using this same relationship, even when the uid tinuum was developed by the Irish mathemati-
is in motion: cian and hydrodynamicist George Gabriel Stokes
(18191903), (Fig. 7.25). He proposed that the state
F(p, T, ) 0 (7.158) of stress is determined by a combination of the
thermodynamic pressure and a linear function of In other words the uid is incompressible. Another
the rate of deformation tensor. Any uid obeying possibility is that the bulk viscosity (7.161) is iden-
this general linear form or simplications of it is tically zero:
referred to as a Newtonian viscous uid because of
Newtons insightful investigations of viscous ow. 0,so 23 and p p (7.164)
The general linear form is simplied here for an
isotropic uid: one in which the viscous constants Apparently this was suggested by Stokes and so it
are not dependent upon direction. For such a uid is referred to as the Stokes condition. If either (7.164)
the stressrate of deformation relationships are: or (7.163) is satised the constitutive law is
reduced to one material constant, and coinciden-
ij pij 2 Dij Dkkij (7.160) tally from (7.161) the mean normal pressure is
Here and are the two material constants that equal to the thermodynamic pressure.
characterize the viscosity of the uid. Note both the Employing the Stokes condition the constitu-
similarity and the difference between this constitutive law for the linear and isotropic viscous uid
tive law and Hookes Law for the elastic solid (7.131). is:
The last two terms on the right-hand side are of the
same form, but the rate of deformation replaces the ij pij 2 Dij 23 Dkkij (7.165)
innitesimal strain and the constants have a differ-
In the absence of any gradients in velocity the
ent meaning. Here the additional term is that con-
normal stress components are equal to the nega-
taining the thermodynamic pressure, p. For the
tive of the thermodynamic pressure and the stress
uid at rest the normal stress components are equal
state is isotropic. In the analysis of geologic struc-
to the negative of the thermodynamic pressure.
tures it is commonly postulated that the rock
Two further simplications of the constitutive
mass is incompressible: in other words the mass
of the law for the Newtonian viscous uid (7.160)
density is constant. Using this constraint as
lead to a reduction of the number of material con-
described by (7.163) the constitutive law (7.165)
stants to one. Both of these follow from the rela-
becomes:
tionship between the mean normal pressure, p
and the thermodynamic pressure, p (Malvern, ij pij 2 Dij (7.166)
1969, p. 299):
Expanding (7.166) in component form, typical

2

p p Dkk p Dkk
3
(7.161)
stress components are:
Here is a material property referred to as the bulk

viscosity and Dkk is the rate of change of volume.
xx p 2
vx
x
v vy
,xy x
y x
yx
The mean normal pressure is the sum of the ther- (7.167)

modynamic pressure (present in the absence of
The stress components are linearly related to the
volume change due to ow) and the pressure
rate of deformation components and the New-
caused by the change in volume due to ow.
tonian viscosity is the proportionality constant.
Interpretation of (7.161) is facilitated by recalling
Substituting for the stresses in one of Cauchys
that conservation of mass leads to the continuity
First Laws of Motion (7.104) using the constitutive
equation (7.81) which is rewritten here as:
equations (7.166) we have, for example:
1 D
Dt
x
v vy vz

x y z
Dkk (7.162)
Dvx
Dt
p

2v 2v
2 2x 2x
x x y
2vy
xy
One possibility is that the material time deriva-
tive of the density is identically zero:
2vz 2vx

xz z2 g*x (7.168)
D Three of the partial derivatives in the parentheses

0, so Dkk 0 and p p (7.163)
Dt can be eliminated as follows:
2vx 2vy 2vz

x2 xy xz x x y z

vx vy vz
0 (7.169)
derivatives of the velocity components give the
rate of deformation components. The pressure
and rate of deformation are used with the consti-
Using (7.162) and (7.163) the term in parentheses is tutive laws (7.167) to calculate the stress compo-
zero because the material is incompressible. nents.
Similar steps reduce the other equations of The NavierStokes equations are the subject of
motion leaving: classic textbooks in hydrodynamics (Lamb, 1945)
Dvi p 2vi and uid mechanics (Landau and Lifshitz, 1960).
g *i (7.170) Especially important in the context of structural
Dt xi xkxk
geology is slow viscous ow, referred to as creep-
These are the NavierStokes equations for the ow ing ow or low Reynolds Number ow (Happel
of a linear, isotropic, and incompressible viscous and Brenner, 1965), in which products of mass
uid with constant mass density. density and material time derivatives of the veloc-
In summary, we have four equations in four ity components are considered negligible and
unknowns. Conservation of mass for the incom- the left-hand side of (7.170) is taken as zero.
pressible uid is taken from (7.162) in component Applications to particular problems include
form as: folding of viscous layers (Johnson and Fletcher,
vx vy vz 1994) and the geodynamics of Earths crust
0 (7.171) (Turcotte and Schubert, 1982; Ranalli, 1987). Low
x y z
Reynolds Number ow includes fully developed
Conservation of linear momentum is embodied in (steady-state) laminar ow in conduits, ow
the NavierStokes equations (7.170), which in com- around immersed objects, ow in narrow but vari-
ponent form are: able aperture conduits, and ow in porous mate-
rials (White, 1974, p. 202).

Dvx
Dt
p

x

2vx 2vx 2vx

x2 y2 z2
g*x
(7.172)

Dv y
Dt
p

y

2vy 2v y 2vy

x2 y2 z2
g*y (7.173)
In the title for a paper (Fletcher and Pollard,
1999) published in the twentieth anniversary

Dvz
Dt
p

z

2vz 2vz 2vz

x2 y2 z2
g*z
(7.174) special issue of the Journal of Structural Geology the
authors of this textbook asked the question:
Conservation of angular momentum is implicit in Can we understand tectonic processes and their
Cauchys Second Law of Motion and the symmetry structural products without appeal to a complete
of the stress tensor (7.122). In applications to mechanics? Our answer was, and is, no. We
structural geology the mass density, , the accel- argued that the majority of structural geologists
eration of gravity, g*, and the viscosity, , com- in the twentieth century worked with isolated
monly are taken as given by laboratory or fragments of continuum mechanics (strain
eld data. Thus, the four dependent variables analysis, Mohrs circles, homogeneous stress
(unknowns) are the velocity components (vx, vy, vz) states) which naturally led to the development of
and pressure, p, and these are sought as functions ad hoc models. In particular the possibility
of the independent variables which are the three that mechanical quantities such as displace-
spatial coordinates and time (x, y, z, t). Given the ment, velocity, and stress vary continuously in
velocity components the constitutive equations space and time was largely ignored. To address
(7.167) are used to calculate the stress compo- these variations in three-dimensional space and
nents. A solution to the NavierStokes equations time requires the mathematical concept of
provides three equations for the velocity compo- partial differentiation with which one can for-
nents and one equation for the pressure as func- mulate the governing equations of continuity
tions of the spatial coordinates and time. Spatial and motion, and set up boundary value and
initial value problems. It is these problems, and links between the concepts of calculus and
their solutions, that constitute the essential physics as taught in typical undergraduate
ingredients of model studies in structural courses and the concepts of continuum mechan-
geology. In this chapter we have provided the ics as required to practice structural geology.
Chapter 8
Elastic deformation
Experiment Theory
() ()
() ()
Comparison of displacement field components near an edge The conceptual success of the [innitesimal theory of
dislocation from phase images of experiments (left column) elasticity and the linear theory of viscosity] is perhaps
and from anisotropic elastic theory (right column). (a) and the broadest we know in science: in terms of them we
(b) displacement component ux parallel to bottom edge of face, explain, and in varying amount control, our
image. (c) and (d) displacement component uy parallel to left daily environment: winds and tides, earthquakes and
edge of image. Photograph reproduced from (Hytch et al., sounds, structures and mechanisms, sailing and ying,
2003) with kind permission of Martin J. Hytch. heat and light (Truesdell and Noll, 1965).
288 ELASTIC DEFORMATION
I
n this chapter we describe how the elastic prop- this theory has made to structural geology.
erties of rock are measured in the laboratory Next, the measurement techniques used to
and provide tables of numbers representing the determine elastic properties in the laboratory and
range of values for different rock types. However, at engineering eld sites are described. This leads
the need to understand and measure the resis- to a discussion of how elastic properties vary with
tance to deformation of rocks goes well beyond the size of the rock mass being tested. Con-
the simple accumulation of numbers in hand- siderations of scale effects are closely linked to
books of rock properties. To paraphrase Truesdell those of heterogeneity with respect to elastic
and Noll (1965), the aim of structural geology is to properties. To build intuition a solution to a
construct mathematical models that enable us, boundary value problem is examined for a circu-
from use of knowledge gathered in a few observa- lar inclusion with different elastic stiffness and
tions, to predict by logical processes the outcomes compressibility than the surrounding material.
in many other circumstances. To analyze a geo- Finally, we consider how the elastic properties of
logic structure one must choose the appropriate rock may vary with orientation, in other words
boundary or initial value problem to serve as a how anisotropic is rock with respect to elastic
mechanical model. To formulate such a problem properties? After reviewing Hookes Law for
one must postulate a particular mechanical anisotropic materials in general, and providing
behavior. That is, one must say exactly what the some representative values of the elastic moduli
relationship is between the stress acting within a from laboratory measurements, we describe a
material and some measure of the deformation, solution to a boundary value problem for an
usually strain or rate of deformation. These orthotropic elastic material. While it is well
relationships are called constitutive equations. For known that rock masses at the scale of Earths
example, researchers studying the displacement crust can be heterogeneous and anisotropic to
eld around an edge dislocation in silicon as some degree, it is noteworthy how well isotropic
revealed by electron microscopy (Chapter 8, fron- and homogeneous elastic models correspond to
tispiece) postulated an anisotropic linear elastic measured deformation. This is illustrated using
constitutive law and calculated model displace- data on surface displacements during the 1999
ments that are remarkably similar to those Hector Mine earthquake.
observed.
We begin this chapter by describing the defor-
mation of the Mancos Shale associated with the
emplacement of a basaltic dike near Ship Rock, 8.1 Estimating rock properties
New Mexico, about 30 million years ago. In this
example the shale was compressed as magma
from geological field tests
pressure forced the dike open, so the thickness of
the dike gives us a measure of the resistance to To determine how rock masses resist deformation
deformation of the surrounding rock. Next we one must conceive an experiment and build or
introduce the formal concept of linear elasticity identify a testing apparatus that controls the
as rst envisioned by Robert Hooke in 1676, along applied loads and facilitates measurement of the
with the constitutive equations that connect resulting deformation. In the context of elastic
stress to strain for idealized elastic materials. solids the experiment is designed to determine a
These constitutive equations are central to the quantitative relationship between the stress and
innitesimal theory of an elastic continuum. As the strain. Here we focus on such experiments,
Truesdell and Noll (1965) suggest in the opening conducted in the eld rather than in a laboratory,
quotation for this chapter, elastic theory has because this enables one to estimate the elastic
played an enormous role in the development of properties of rock at length scales of kilometers
science and engineering. Some examples are which are relevant to the geological structures
reviewed to illustrate the type of contributions under consideration.
8.1 ESTIMATING ROCK PROPERTIES FROM FIELD TESTS 289
8.1.1 A field test at Ship Rock, New (a)

Mexico
The apparatus for this eld test is an igneous
dike that crops out near the base of a volcanic
edice called Ship Rock in northwestern New
Mexico (Fig. 8.1a). In this oblique aerial photo-
graph Ship Rock is the most prominent feature,
reaching about 600 m above the surrounding
plain. This is the remnant of a volcanic conduit
that fed surface eruptions and lava ows about 30
million years ago (Delaney and Pollard, 1981). It is
one of the most spectacular geological features of
the picturesque four corners region where the
states of New Mexico, Arizona, Utah, and
Colorado adjoin. The Navajo people who live in
this region call this edice Tse Bitai or Winged
Rock and it plays a signicant role in their (b)
n
creation story. Kmu Tm
A geological map (Fig. 8.1b) reveals several
dikes that form a crude radial pattern around the Ship Rock
base of Ship Rock. These dikes, largely composed Tmn
of a Tertiary volcanic rock called minette (Tmn), Ttb
crop out in the nearly at-lying upper part of the
Cretaceous age Mancos Shale (Kmu) which is
made up of thinly bedded shales, siltstones, and
36o 40'
sandstones. An estimated 1 km of sedimentary
rock has been eroded from this region since dike
Tmn
New Mexico
emplacement, so the top of the Ship Rock edice
(composed of Tertiary tuff breccia, Ttb) was prob-
ably within a few hundred meters of the old land Kmu
surface at the time lava erupted from the central
conduit of this volcano. Because the exposed dikes
range up to 10 km in outcrop length it seems likely
that some of them also erupted to feed lava ows. 0 1 2 3 4 km
Today approximately 1 km of sedimentary rock
lies below the ground surface and overlie a crys- 108o 50'
talline basement of Precambrian age. The limb Fig 8.1 (a) Oblique aerial photograph of Ship Rock, NM
dips on a gentle syncline that cuts across the (photograph by D. L. Baars). Height of edifice is about 600 m
northeastern dike are less than 5 and these dips and length of the northeastern dike in the foreground is
are typical for the exposed sedimentary strata about 3 km. (b) Map of region near Ship Rock: Kmu,
over much of the area. Cretaceous Mancos Shale; Tmn, Tertiary minette; Ttb,
The basic idea behind our use of dikes as natural Tertiary tuff breccia (Delaney and Pollard, 1981).
testing machines is that magma pressure in dikes
pushed against and elastically compressed the sur-
rounding sedimentary rock mass. Our objective is Mancos Shale as four springs that resist the com-
to estimate the resistance of the sedimentary rock pression (Fig. 8.2a). The mechanical relationship
to this loading. For the sake of this example we ide- between applied force, F, and the relative displace-
alize the dike walls as two rigid plates and the ment, u, of the two ends of an elastic spring is:
(a) T, equivalent to the thickness of the dike, and the

Potential dike plane
plates move symmetrically apart, so the displace-
K/2 K/2 ment is u T/2.
We idealize the magma as a uid under
C pressure, P, which is injected between the two
L inner plates (Fig. 8.2b). In order for this uid to
Mancos Mancos
Shale Shale squeeze between the two plates and push them
apart, the pressure must exceed the compressive
K/2 K/2 tectonic stress holding the plates together. The
L displacement of the inner plates will be propor-
tional to the amount by which P exceeds C, and
(b) T this quantity, P C, is called the driving pressure.
y Thus the force, F, associated with the plate
separation is F (P C)WL. Note that the total
force acting on the spring after injection of the
P x uid is PWL, but spring compression due to the
L
Igneous tectonic force, CWL, took place before injection
dike and is not related directly to the opening of the
plates.
Substituting for the force and displacement in
L Displacement, u (8.1), we have:

Fig 8.2 Spring and block model for dike. (a) Before dike T
emplacement compression, C, acts across the prospective (P C)WL K (8.2)
2
dike plane; K/2 is the spring constant. (b) Pressure, P,
compresses springs and model dike opens with displacement, The equivalent relationship between stress and
u, in both directions.
strain is found by rearranging this equation:
F Ku (8.1) (P C)

K T2
W L
E
T2
L (8.3)
The spring constant, K, measures the stiffness, or Here the left-hand side is the applied stress and
resistance, to deformation. The springs are con- the right-hand side is a constant, E, times the
nected to two outer rigid plates, initially loaded by resultant strain associated with opening of the
a compressive stress of magnitude, C, represent- idealized dike. Recall that normal strain is a
ing the tectonic loading at the time of dike change in length divided by the original length of
emplacement. This compressive stress is transmit- a line element. Here the change in the original
ted through the springs (which have shortened spring length, L, is given by T/2, so the term in
accordingly), thereby holding the two inner plates parentheses on the right-hand side is the normal
tightly pressed together. strain. The constant, E, is called Youngs modulus
The plates have a length, L, the outcrop length of elasticity. It measures the resistance of a ma-
of the dike, and a width (out of the plane of view) terial to change in length (strain) under an
given by W, so their area is WL. Before dike applied normal stress.
emplacement, the distance, L, separates the inner The northeastern dike at Ship Rock has an
and outer plates. At this point in our thought outcrop length L 2900 m and an average thick-
experiment the two outer plates are xed in ness T 2.3 m. Using (8.3) the ratio of driving pres-
place and thereafter are not allowed to move, so sure to Youngs modulus is estimated as:
we can measure the relative displacement of the
spring ends as the plates are pushed apart. We (P C) T
0.0004 (8.4)
refer to the distance between the inner plates as E 2L
8.1 ESTIMATING ROCK PROPERTIES FROM FIELD TESTS 291
By measuring dike length and thickness, and Table 8.1. Geologic eld tests using Ship Rock
nding independent estimates for the magma dikes.
pressure and compressive stress, we can use (8.4)
to estimate Youngs modulus for the rock mass Dike name Host rock E (GPa)
deformed by the dike.
The geological evidence suggests that the Northeastern Mancos Shale 5
current outcrop of the northeastern dike was Southern Mancos Shale 4
about 1 km below the ground surface at the time
Small Mancos Shale 12
of dike intrusion, and we assume that the dike
erupted at that surface. We take m 2600 kg m3 multiple dike segments yields a Youngs modulus
and r 2400 kg m3 as the magma and host rock of about 2 GPa, not very different from the single
densities, g* 9.8 m s2 as the acceleration of fracture estimate (Table 8.1). If the dike did not
gravity, and D 103 m as the height of the magma propagate to the surface, the pressure would not
column and thickness of overburden. The magma necessarily be the hydrostatic value used above.
pressure and compressive stress in the Mancos Furthermore, the dike is unlikely to be two dimen-
Shale at the depth of the current outcrop are cal- sional and the effect of the three-dimensional
culated as P mgD 25.5 MPa and C r gD 23.5 form on opening should be addressed using solu-
MPa. Therefore, the driving pressure was (P C) tions to a three-dimensional boundary value
2 MPa and Youngs modulus is estimated from problem of elasticity.
(8.4) as E 5 GPa. The southern dike has a length
of about 9 km and an average thickness of about 8.1.2 A generalized geological field
10 m, and the small dike just to the south of the method for estimating rock
northeastern dike has a length of about 1 km and properties
an average thickness of about 0.5 m. These data Geological eld methods for estimating rock
provide additional estimates of Youngs modulus properties are useful for two reasons. Most
(see Table 8.1). We do not suggest that these obvious is the fact that many geological struc-
values of Youngs modulus necessarily are repre- tures are too large to submit to laboratory study;
sentative of other large rock masses, and they cer- their size puts them beyond the capability of
tainly are not known with the same precision human engineering in terms of their length scale.
expected for laboratory measurements. On the Furthermore, if time is an important variable,
other hand, we believe that the method described then laboratory tests are incapable of duplicating
here is important and should be applied to other geological time scales. Note that we are not refer-
igneous dikes. ring here to scaled model experiments that seek
The northeastern dike at Ship Rock is not con- to simulate geological processes (see Chapter 4),
tinuous along the outcrop, but is divided into 35 but rather to the measurement of the physical
echelon segments separated by Mancos Shale properties of rock. Second, the physical and chem-
(Delaney and Pollard, 1981). Also, the thickness of ical conditions under which many geological
the dike is partly attributable to erosion of the structures have formed is unknown, and even if
dike wall by the owing magma. Apparently these conditions could be deciphered, they may
the hot magma caused thermal fracturing of the not be reproducible in the laboratory. Thus we are
Mancos Shale at the contact and this fractured faced with a considerable challenge to determine
rock was locally removed by the magma. Clearly the properties of rock masses at depth in the
such a process is not included in the elastic model Earth at geological length and time scales.
of fracture dilation, so thickness measurements One approach to this problem is to use tests
should be corrected accordingly. Furthermore, under natural conditions and at natural length
one might want to make a small correction for and time scales. That is, let nature do the experi-
shrinkage of the igneous rock as it cools from ment and look for an appropriate way to interpret
magmatic temperatures. Using a numerical solu- what has been done. The method is summarized
tion to the elastic boundary value problem for in six steps as follows:
1. Identify a natural experiment run under the gists rock hammer hits an exposure. If the expo-
conditions and at the scale of interest. sure is fresh granite, the hammer springs back
2. Use mapping techniques in the eld to charac- quickly with a high-pitched ringing and vibrates
terize the structures, identify the lithologies, vigorously for a second or two. The rock also gives
and measure the relevant geometric parame- off a sharp audible report. Perhaps a few small
ters. chips of rock or metal shoot out from the point of
3. Infer the sequence of deformation and appro- impact if the blow is particularly aggressive, but
priate boundary conditions for loading and lighter blows permanently deform neither the
displacement (or other relevant physical quan- hammer nor the rock. Both rock and hammer
tities) from the eld data. return nearly to their shapes just before the
4. Set up and solve, or borrow from the literature, impact. If you held your hand on the exposure
the appropriate mechanical problem (usually a near the point of the hammer blow you would feel
boundary or initial value problem from con- vibrations in the granite, a result of waves propa-
tinuum mechanics). gating out from the point of impact. Other types
5. Derive from the solution to this problem an of rock respond in a similar fashion, but with
equation for the physical property of interest in some quantitative differences. For example, sand-
terms of the measured and/or inferred quanti- stone might respond with a duller sound and the
ties. hammer will not seem to jump back as quickly.
6. Use the derived equation and available data to When materials return essentially to their origi-
estimate the physical property. nal shape after the applied loading is removed, we
say they are elastic.
This method is not limited to rocks with elastic A linear relationship between force and exten-
properties, but is generally applicable to any ma- sion is credited to the English natural philosopher
terial behavior described by a well-dened consti- Robert Hooke (16351703) who published a state-
tutive law. Nor is it limited to dikes, but is ment on the subject in 1676 in the following
generally applicable to any geological structure remarkable form (Gordon, 1976):
that can be modeled using continuum mechanics.
ceiiinosssttuu
Hooke apparently was intent on laying claim to

8.2 The idealized elastic material this area of research, without providing the
specics, before the appearance of a lengthier
The general concept of an elastic material is one in treatise. The anagram has the following solution
which the current conguration depends only on published in 1679 by Hooke:
the initial (unstressed) conguration and the
Ut tensio sic vis
current state of stress, and not on the history of
deformation from the initial to the current state J. E. Gordon provides the following translation
(Truesdell and Noll, 1965). In most applications of (Gordon, 1976):
elasticity theory one does not study the most
As the extension, so the force
general elastic material, but rather one in which
the stress is linearly related to the innitesimal In other words, the extension is proportional to
strain. Because the innitesimal strain is an approx- the force.
imation these elastic models must be considered an Despite his obfuscation in presenting this dis-
approximation. Never-the-less linear elastic theory covery, Robert Hooke is honored by having this
has provided a wide variety of useful solutions for relationship, and a generalized version that
mechanical problems in structural geology. extends the simple one-dimensional concept to
three dimensions, referred to as Hookes Law of
8.2.1 The elastic solid linear elasticity. The book The Abyss of Time by
One common experience with the mechanical Claude C. Albritton (1980) contains an informative
behavior of rock comes at the moment the geolo- account of Hookes life and his scientic contribu-
8.2 THE IDEALIZED ELASTIC MATERIAL 293
tions to geology. The incomplete, one-dimensional (a) Y

W Initial state Area, A
form of the elasticity model introduced by Hooke
was further developed in the 1700s by Bernoulli
H Elastic bar X
and Euler in order to describe the deformation of
beams of materials used in construction. The Z
B
development of a complete model was impossible y
at that time, because it depended on understand- (b)
w Current state
ing the general concepts of stress and strain, h f
Extended elastic bar
which were not formulated until 1822 by Cauchy x
z
(Malvern, 1969). b
We construct in our minds a perfectly elastic
solid, one for which the extension is completely (c) Y,y
recoverable upon release of the stress. In order to
focus on mechanical relationships, the tempera-
xx= a
ture is postulated to be constant, so we consider
isothermal conditions, and the bar is postulated to X,x
have homogeneous material properties. To quantify
Z,z
the behavior, consider a rectangular bar of this
material with length B in the initial state (Fig. Fig 8.3 Idealized bar used to define elastic constants. (a) In
8.3a). The length of the undeformed bar is parallel the initial unloaded state the length of the bar is B. (b) In the
to the X-axis of the chosen coordinate system. The current loaded state the bar has extended to a length b.
undeformed width and height of the bar are W (c) Element from the bar showing lateral contraction
and H, respectively, and the cross-sectional area is accompanying longitudinal extension.
A WH.
At some time, t, the bar is loaded by a force of
magnitude, f, acting perpendicular to the ends stress (extensional strain) rst quadrant and the
and directed outward. The force arrows (Fig. 8.3b) compressive stress (contractional strain) third
are meant to be schematic; the actual force is uni- quadrant. The solid is non-linear elastic if the values
formly distributed over the ends, so the stress follow a curved path, but return along the same
within the bar is perfectly homogeneous. In curve to zero extension (the origin of the graph)
response to the applied force the bar stretches to upon complete unloading. If the values follow a
a deformed length b parallel to the x-axis of the straight-line segment and return to zero exten-
coordinates for the current state. In earlier chap- sion, the solid is linear elastic. If the loading and
ters you learned that the stress and the extension unloading paths are identical, the deformation is
can be dened at every point within such a bar by said to be reversible, and this is a necessary
taking limits on the local ratios of force to surface attribute of all elastic deformation whether it is
area and change in length to length. Here, linear or non-linear. In contrast, if the deforma-
because these quantities are postulated to be the tion is irreversible (does not follow the same curve
same everywhere in the bar and act along the axis upon unloading), it is inelastic. Notice that time
of the bar, we will refer to them as the axial stress, has not appeared in our discussion of the elastic
a, and axial extension, ea, at any point. The axial solid. The idealized elastic material will extend or
stress and axial extension are: shorten in proportion to the applied stress as this
stress is increased and decreased on any time
f bB
a , ea (8.5) scale.
A B
Now we place an additional restriction on the
Values of axial stress, a, are plotted versus axial behavior of the elastic solid: namely we limit our
extension, ea, to represent the loading and discussion to elastic solids that are isotropic with
unloading of the bar in graphical form (Fig. 8.4). respect to the material properties that relate
Note that this graph includes both the tensile stress and extension. For the elastic bar (Fig. 8.3)
da
E(tangent) f (ea) (8.7)
Axial stress
sa = f/A
linear dea
elastic
non-linear
loading elastic The local slope is the rst derivative of the axial
stress with respect to the axial extension and is a
sa function of the extension.
unloading
r e Because extension is dimensionless, Youngs
o
m tiff ea
compression tension
s modulus and the tangent modulus have the same

less dimensions as stress and carry the same units as
stiff
Axial s
les iff Axial extension, stress:
contraction st ea = (b B)/B
Youngs modulus, E {} M L1 T2, and
unloading E [] Nm2 Pa (8.8)
re
mo
f
stif
g
din
Approximate values of Youngs modulus for

loa
tic
common materials are (Eshbach, 1961):

un
las
~e
loading
steel (spring), E 200 GPa
copper, E 110 GPa
tic
linear aluminum, E 70 GPa

las
elastic
ine
redwood (dry), E 9 GPa

plexiglas, E 3 GPa
Fig 8.4 Plot of axial stress versus axial extension for linear
rock, E 1 to 100 GPa
elastic behavior (thick straight line), non-linear elastic
behavior (curved line), and non-linear inelastic behavior. It is interesting to note that E for rocks covers most
of the range of these familiar natural and syn-
thetic solids. Rocks in the higher part of this range
this would mean that stretching along its length
would ring when hit by a geologists hammer,
(in the x-coordinate direction) would produce
whereas those in the lower part would respond
exactly the same plot of stress versus extension
with a dull thud. Youngs modulus is a property of
as stretching the bar along its width (in the z-
a linear elastic solid which, in qualitative terms,
coordinate direction), or, for that matter, in any
characterizes how stiff it is in response to applied
other direction. For a linear and isotropic elastic
stress: stiffer elastic solids have steeper slopes on
solid the proportionality constant in these one-
the stress versus extension graph (Fig. 8.4) and
dimensional stretching experiments is called
therefore greater Youngs moduli. In contrast,
Youngs modulus of elasticity after the English scien-
softer elastic solids have less steep slopes.
tist, Thomas Young. This modulus is customarily
Most of the direct evidence for the elasticity of
indicated with the symbol E and is dened as the
rock comes from laboratory experiments con-
ratio of the change in axial stress, a, to the cor-
ducted on small samples of rock. Figure. 8.5 pro-
responding change in axial extension, ea (Fig.
vides four examples of axial stress plotted versus
8.4). For the linear elastic solid the ratio of these
axial contraction for uniaxial compression tests
changes is the same for any point along the
conducted in stiff testing machines (Jaeger and
loading path, so:
Cook, 1979, Fig. 4.2.3). Two of these rock types
(Rand Quartzite and Solenhofen Limestone)
a a
E constant (8.6) display a nearly linear and apparently reversible
ea ea
behavior to axial compressions of about 200 MPa
For non-linear elastic solids the slope changes con- and axial contractions of 3 to 4 103. We say
tinuously along the curve of stress versus exten- apparently reversible because the data are not
sion. In this case a so-called tangent elastic modulus shown for unloading these samples, but the
can be dened at every point as the slope of a authors imply that the behavior was dominantly
tangent line (the dashed line segment on Fig. 8.4): elastic. The Karroo Dolerite displays a nearly
Axial contraction (x103) there is no stress acting in this lateral direction,

8 6 4 2 0 6 4 2 0
the bar is thinner. The perpendicular stress, p,
0 0 and the perpendicular extension, ep, are respec-
Gosford Rand
Sandstone Quartzite 70 tively:
14
140 wW
28 p 0,ep (8.9)
210 W
Axial stress (MPa)

42 280 Usually when the axial stress is tensile, the per-
pendicular extension is a negative number
6 4 2 0 6 4 2 0
0 0
because the width of the bar reduces, that is W
Karroo Solenhofen w. For an isotropic bar the extension in any direc-
70 70
Dolerite Limestone tion, perpendicular to the specimen axis, would
140 140 be the same. To characterize this behavior a
210
210
second elastic property, called Poissons ratio is
280 dened:
280
350 ep (8.10)
e
a
Because the extension in the numerator and

Fig 8.5 Uniaxial compression tests in a stiff testing
machine for four different rock types. Reprinted from Jaeger denominator of (8.10) usually are of opposite sign,
and Cook (1979) with the kind permission of Mrs. Jennifer D. a negative sign is used in the denition to make
Cook. this quantity a positive number.
As a ratio of dimensionless quantities,
Poissons ratio is dimensionless itself, and carries
linear and apparently reversible behavior to axial no units.
compression of about 300 MPa and axial con-
Poissons ratio, {} L0 1 (8.11)
traction of about 4 103. All of these samples
have a distinct non-linear behavior near the peak Different elastic materials have different values of
axial stress and this was followed by brittle Poissons ratio and these fall in the range 0.0
failure. (Jaeger and Cook 1979, pp. 82) conclude 0.5, with those materials at the lower end of this
that these tests show how unimportant the range being compressible and those at the upper
regions OA and BC [see Fig. 9.5] are in many end being incompressible. For a given axial exten-
practical cases and therefore that the assumption sion, a bar with a greater value of Poissons ratio
of linear elasticity up to failure really is a good would thin more than a bar with a lesser value.
one. The Gosford sandstone, in contrast, displays Because Poissons ratio does not depend on the
a non-linear behavior with increasing stiffness as sign of the extension, a greater value of also
the compression is increased. This rock is dis- implies that a bar would thicken more for a given
tinctly softer than the others and has a lower axial shortening. Values of Poissons ratio for
peak stress before loss of load-carrying capacity. common materials include:
Using linear elastic behavior to model this rock
rubber, 0.5
in the pre-failure stressstrain regime could be
cork, 0.0
problematic.
rock, 0.1 to 0.3
When you stretch a rubber band it becomes
thinner. The more you stretch the band, the Rubber is a material that is nearly incompressible:
thinner it becomes. The elastic bar (Fig. 8.3) is it maintains a nearly constant volume by thicken-
drawn to reect this well-known property of ing just enough to compensate for a given shorten-
elastic materials. The original width, W, is ing. In contrast, materials like cork or foam rubber,
decreased to a nal width, w, in a direction per- with a Poissons ratio of nearly zero, are said to
pendicular to the applied force. Even though be compressible, because they can be shortened
without getting thicker. This property of cork

(a) y
makes it possible to insert a cylindrical piece into
s yy = s xx
the neck of a bottle by pressing on its end. In con- s
trast, a rubber stopper must be tapered. Rocks are n
somewhat compressible.
g = 45o
tn
8.2.2 Stressstrain relationships for the
homogeneous, isotropic, linear x
ts
elastic solid s xx
For a homogeneous and isotropic linear elastic
bar under isothermal conditions being stretched
in the x-coordinate direction by an applied stress
xx (Fig. 8.3) the extension in x is proportional to
the stress and the proportionality constant is the
reciprocal of Youngs modulus (8.6). The exten- (b) y
y x

sions in y or z are proportional to the stress and
the proportionality constant is the Poissons ratio s
divided by Youngs modulus. That is xx xx E y
x
= g = 45 o
x
s
y
and yy xx zz xx so yy zz (E)xx. s
t
Consider successively applying a normal stress in
each coordinate direction and accounting for all
x
of the possible normal strains. A consequence of
the assumption of small strains and rotations is
that the strain state is independent of the order in
which the stresses are applied to the body.
Therefore, these strain states can be superimposed
Fig 8.6 Element used to define relation between shear
and we have: stress and shear strain for linear elastic material

1 (Timoshenko and Goodier, 1970). (a) Tension and
xx (yy zz)
E xx
compression of equal magnitudes. (b) Shear stress on
element oriented at 45 to that in (a).
yy
1
(zz xx)
E yy (8.12)
on a diagonal boundary with outward unit
zz
1
(xx yy)
E zz normal, n, oriented at 45 are:
For the isotropic material these relationships do tn xx cos 2 45 xx sin 2 45 0

not depend upon the orientation of the coordi-
ts (yy xx) sin 45 cos 45 xx (8.13)
nate system and hold, therefore, for any choice of
orthogonal coordinate axes. These are the normal and shear traction compo-
The relationships between the innitesimal nents acting on the x-face of an inclined cubic
shear strains and the shear components of stress element, where the x-axis makes an angle of 45 to
must be dened for a complete description of the the x-axis (Fig. 8.6b). The same result is found for the
strainstress relationships (Timoshenko and other faces of this inclined element, so the normal
Goodier, 1970, p. 9). Consider a cubic element with stress components both are zero, and the shear
only normal stress components acting on the x- stress component is equal to the negative of xx.
and y-faces of equal magnitude and opposite sign, Using (8.12) the two normal strain components
such that yy xx and xy 0 (Fig. 8.6a). From for the stress state shown in Fig. 8.6a are found to
Cauchys Formula the traction components acting be equal in magnitude and opposite in sign, so
yy xx. In other words the extension in the The stress components appear as dependent
x-direction is equal to the contraction in the variables in (8.19) to be calculated from the strain
y-direction. The element inclined at 45 experi- components (independent variables) and elastic
ences only a shear strain. This follows from the moduli. In contrast, the strain components appear
fact that the strains transform just as the stress as dependent variables in (8.18) to be calculated
components (Fung, 1969): from the stress components (independent vari-
ables) and elastic moduli. Some have suggested that
xx xx cos 2 45o xx sin 2 45o 0 the possibility of treating either stress or strain as
the dependent variables implies that there is no
xy (yy xx) sin 45o cos 45o xx (8.14)
cause and effect relationship between stress and
The nal step is to relate the shear stress and shear strain (Marrett and Peacock, 1999). This notion is
strain on the inclined element. From (8.12) where contrary to the concept of Newtonian mechanics in
yy xx and zz 0, we have: which force is described as the causative agent and
acceleration is the resulting effect.
xx
1

yy
E xx
1
E xx (8.15) The natural extension of Newtons concept to
the elastic continuum is to view stress or traction
But we have just learned, from resolving stresses as the causative agent and strain or displacement
and strains on the inclined element, that xx as the resulting effect. Thus (8.18) and (8.19) simply
xy and xx xy, so: show how stress components can be calculated
from strain components, or vice versa, and
1 nothing more profound is implied concerning the
xy xy (8.16) physical framework of Newtonian cause and effect.
E
The constitutive equations for the linear and
Thus, the shear strain is proportional to the shear isotropic elastic solid are used extensively in the
stress acting on the inclined element and the con- analysis of geologic structures, often with little
stant of proportionality is (1 )/E. experimental justication. Ideally, the rocks
The relationship between shear stress and would be sampled in the eld area, brought back
innitesimal shear strain (8.16) can be generalized to the laboratory, and tested to reveal their
for all of the components in three dimensions to mechanical properties. Usually, the facilities and
give: funding for such testing is not available. Even if
1 1 1 rock samples were tested rigorously, would the
xy xy,yz yz,zx zx
E E E measured properties correspond to those millions
(8.17) of years ago when the geologic structures under
Equations (8.12) and (8.17) are a statement of investigation actually formed? Identifying those
Hookes Law for the isotropic elastic material. conditions is a challenging problem for structural
These six equations are written using indicial geologists. More often than not, justication for
notation as: using the elastic model comes a posteriori, after
1 the solution to the boundary value problem pro-
ij ij kkij (8.18)
E E vides a compelling correlation to eld observa-
Given the innitesimal strain components, (8.18) tions. Examples of such correlations are found
can be rearranged as: throughout this book and in the cited literature.
ij
E
1
ij
E

(1 )(1 2) kk ij
(8.19) 8.2.3 Relations among elastic moduli for
isotropic materials
It is clear from (8.18) that when all the stress com- To characterize further the isotropic and linear
ponents are identically zero, all the strain com- elastic material we derive the relation between
ponents are zero. This describes the initial, volumetric strain and pressure. The innitesimal
unloaded state of the elastic body. volumetric strain is given by the change in volume
of an element divided by the original volume, The sum of the normal stress components also is
V/V. To understand how this quantity is related to invariant for any rotation of the coordinate
the innitesimal strain components consider a system. For an isotropic state of compressive stress
sphere of radius L in the initial state that deforms the negative of the uniform normal stress is the
to an ellipsoid with half axial lengths L1, L2, and L3 static pressure, po xx yy zz. Some use
in the current state. Recall that the normal the phrase hydrostatic pressure or hydrostatic
innitesimal strain of a line element is equivalent compression for this quantity, but this should be
to the extension of that line element. Thus, the avoided, as the subject here is the deformation of
normal strain of the line that becomes the major elastic solids, not of water. Because the volumetric
axis of the ellipsoid is 1 (L1 L)/L, so L1 (1 1) L. strain (8.22) is the sum of the three normal strain
The other axial lengths are similarly related to the components, we add these as dened in (8.12) and
principal strains. The volume of the sphere is Vs use (8.23) with m p0 to nd:
4 3 4
3 L and that of the ellipsoid is Ve 3 (L1L2L3)
1 2 3(1 2) 1
which may be approximated as: kk kk m p0 (8.24)
E E K
Ve 43 (1 1)(1 2)(1 3)L3 Here K is called the bulk modulus, which relates the
43 (1 1 2 3)L3 (8.20) innitesimal volumetric strain to the pressure for
an isotropic state of stress.
The approximation is obtained by neglecting
Considering Youngs modulus and Poissons
products of the principal strains because they are
ratio as the two independent moduli of the
very small compared to one or compared to the
isotropic elastic material, and using (8.24), the
strains themselves.
bulk modulus is written:
It is one of the basic attributes of the innites-
imal strain tensor that the sum of the normal E
K (8.25)
strains is invariant for any rotation of the orthog- 3(1 2)
onal coordinate axes, so we can write the volume
Note that the bulk modulus, K, approaches an
of the ellipsoid:
innite value as Poissons ratio approaches 0.5.
Ve 43 (1 xx yy zz)L3 (8.21) This is consistent with the characterization of
such materials as being incompressible. Most
Substituting for the volumes of the ellipsoid and
liquids are nearly incompressible, whereas gases
sphere in the equation for the volumetric strain,
are highly compressible. Very porous rocks are
we have:
somewhat compressible, whereas rocks with low
V Ve Vs 43 (1 xx yy zz) 43 porosity tend to be less compressible. For the per-

V Vs 4
3
fectly compressible material, 0, so the bulk
xx yy zz kk (8.22) modulus is K E/3.
The elastic shear modulus, G, is used to relate
Note that the last term in (8.19) is proportional to
shear stress to shear strain, as in (8.17), from
the volumetric strain. Thus, each normal stress
which we have:
component is related to the corresponding
normal strain component and the volumetric E
G (8.26)
strain. Also, for the perfectly compressible mater- 2(1 )
ial, 0, each normal stress is simply propor-
The factor of one-half appears because we are
tional to the corresponding normal strain,
using the tensor convention to dene shear strain.
because the last term in (8.19) is zero.
This shear strain is one-half the magnitude of the
Next we relate the volumetric strain to the
so-called engineering shear strain and the shear
mean normal stress, m, dened as the average of
modulus was originally dened using the engi-
the three normal stress components:
neering convention (Fung, 1969). For perfectly
compressible material, 0, so G E/2, and for
m 13 (xx yy zz) 13 kk (8.23)
incompressible material, 1/2, so G E/3.
8.3 QUASI-STATIC DISPLACEMENT BOUNDARY VALUE PROBLEMS 299
Gabriel Lam (17951870) introduced another Sokolnikoff, 1956; Timoshenko and Goodier,
elastic constant, , in 1852 (Fung, 1969) and this is 1970; Barber, 1992).
related to Youngs modulus and Poissons ratio as: The mathematical structure of the theory of
elasticity makes it possible to solve boundary
Ev
(8.27) value problems of elasticity using displacement,
(1 )(1 2)
traction, or mixed boundary conditions. This fact
Poissons ratio is related to Lams constant as: led to a debate about whether boundary displace-
ments cause the elastic solid to deform (Marrett
(8.28) and Peacock, 1999; Tikoff and Wojtal, 1999;
2(G )
Peacock and Marrett, 2000; Pollard, 2000). In the
The constant characterizes the spectrum of authors opinion the opportunity to employ dis-
behaviors from perfectly compressible materials, placement boundary conditions is no more than
0, for which 0, to incompressible materials, a consequence of the underlying mathematical
1/2, for which . Taken together, and G relationships among traction and displacement
sometimes are called Lams constants. Equations components. In a Newtonian context one would
(8.19) may be written in terms of Lams constants seek an explanation for the specied displace-
using (8.26) and (8.27) such that: ments in terms of forces applied exterior to the
model boundary, and these forces would be the
ij 2Gij kkij (8.29)
ultimate causative agents for the resulting defor-
The relatively simple behavior of the isotropic and mation in the interior. On the other hand, where
linear elastic solid makes it ideally suited for tractions are the prescribed boundary conditions
analysis in continuum mechanical models. One one may refer to the associated forces as the
can solve problems that involve quite complex causative agents for the deformation.
geometries and boundary conditions by restrict-
ing the material behavior in this way, and the 8.3.1 Two-dimensional plane strain
solutions provide important insights about the solutions for cylindrical structures
origins of some geologic structures. All structures in the Earth are three dimensional,
but there are circumstances in which it is appro-
priate to ignore some of the components of stress,
8.3 Quasi-static displacement strain, and displacement. This has practical impli-
cations because the mathematical complexity of
boundary value problems the boundary value problem is greatly reduced.
One circumstance involves structures that are
One may formulate a problem in elasticity very long in one dimension relative to their size in
theory in terms of the displacement compo- the other two dimensions. If the geometry of such
nents, or in terms of the stress components, as a structure does not change signicantly along its
the dependent variables. In this section we take length, it may be described as a cylindrical structure.
the equations of motion written in terms of the A common geological example would be the
displacement components as the dependent surface of a sedimentary layer, folded into a shape
variables. A specic example is provided using that may be approximated by moving the fold axis
the two-dimensional solution for an edge dislo- through space without changing its orientation
cation which has surprisingly broad applica- (Fig. 8.7a). Other common examples include the
tions in structural geology, ranging from surfaces of blade-like dikes in volcanic rift zones
micrometer-scale defects in mineral grains to and of vertical joints conned between two hori-
plate-bounding faults at continental margins zontal sedimentary layers (Fig. 8.7b).
(Weertman and Weertman, 1964; Weertman, The special case of deformation that applies to
1996). Textbooks on elasticity theory provide two-dimensional cylindrical structures is called
many other useful solutions for displacement plane strain. Here the (x, y)-plane is taken as the
boundary value problems (Muskhelishvili, 1954; plane of interest (Fig. 8.7), and we postulate that
(a) material coordinates (X, Y, Z) with the spatial coor-

y dinates (x, y, z) in keeping with the consequences
of innitesimal strains.
Using (8.30) as a constraint on the displace-
ment vector, the kinematic relationships among
spatial gradients in the displacement compo-
nents and the innitesimal strain components
(7.129) simplify to the following:
x
xx
ux
x
uy
,yy ,xy
y
1 ux uy
2 y

x (8.33)
z
The normal strains in the x- and y-directions and
(b) the shear strain in the (x, y)-plane are only func-
tions of the x- and y-coordinates. The out-of-
y plane strain components are identically zero:
zz 0, yz 0, zx 0. The name plane strain
follows directly from these two facts about the
strain components.
The in-plane stress components are propor-
x tional to the strain components through Hookes
Law (8.29), which reduces to:
u
xx (2G )xx yy,
(8.34)
z u y = f 2 (x , y ) yy xx (2G )yy, xy 2Gxy
The three stress components in the (x, y)-plane are
uz = 0
u x = f1 ( x , y ) functions only of the x- and y-coordinates. The
normal stress in the z-direction, zz, is not gener-
Fig 8.7 Geological structures approximating plane strain ally zero despite the fact that the displacement
conditions. (a) Cylindrical fold. (b) Blade-shaped dike or joint. and normal strain components in that direction
are zero:

the out-of-plane displacement component, uz, is zz (xx yy) ( yy)
2(G ) xx
identically zero, whereas the two in-plane compo-
nents are functions of x and y only: (xx yy) (8.35)
ux f1(x, y), uy f2(x, y), uz 0 (8.30) The out-of-plane shear stress components are
zero: yz 0 and zx 0.
These conditions demand that neither the geom-
etry nor the loading conditions change along the 8.3.2 A complete solution in two
z-axis. Based upon these displacement conditions, dimensions: the edge dislocation
and ignoring body forces, the quasi-static form of
Naviers displacement equations of motion (7.138) The dislocation is an object worthy of study. Its exis-
reduce to: tence permits metals to be plastically deformed with
ease, a circumstance upon which our modern
2ux 2u 2 u y
(2G ) 2
G 2x (G ) 0 (8.31) technology is so dependent. . . . The dislocation also
x y xy
permits nonmetallic crystalline materials to be
2uy 2uy 2ux
G 2 (2G ) 2 (G ) 0 (8.32) plastically deformed. . . . Thus the dislocation plays
x y xy a commanding role in those grandest of all
These equations govern the spatial distribution of deformations on earth: the upheavals that have
the two displacements in the (x, y)-plane. Note that produced the mountain ranges and the continents
here and in what follows we have replaced the themselves (Weertman and Weertman, 1964, p. 1) .
The edge dislocation provides an insightful example (a)

of a structure that can be modeled using the two-
dimensional plane strain form of the equilibrium 100m
equations, (8.31)(8.32), (Hirth and Lothe, 1982).
Figure 8.8a shows dislocation loops (ne white
lines) in a crystal of silicon observed at the 100-m
scale by chemical etching (Friedel, 1964, Fig. 1.19).
Dislocations move through mineral grains during
plastic deformation of rock as a result of the
forces imposed during tectonic events (Poirier,
1985). Surprisingly, the quasi-static elastic solu-
tion for the displacement, strain, and stress in the
vicinity of an edge dislocation at the micrometer-
scale also can be used to model kilometer-scale
geological structures, including those as diverse
Burgers
as igneous dikes and plate-bounding strike slip (b) Extra half- vector
faults. In this broader context models utilizing plane b = bxex
dislocation solutions have provided a deeper
understanding of the physical processes that
shape mountain ranges and continents. In this 1 2 3 4 5
f s
section we describe the physical nature of the 4
edge dislocation at the crystal-lattice scale and 1
then examine the elastic solution and show how 3

z 2 x
pairs of edge dislocations approximate the defor-
mation near fractures and faults (Weertman and 2
3
Weertman, 1964).
Figure 8.8b is a schematic two-dimensional 1
4
illustration of a crystalline lattice that contains
an edge dislocation. Note the extra column of 5 4 3 2 1
atoms that appears to distort locally the other-
wise regular lattice near the bottom atom in the
y
column. The origin of the coordinate system (x, y, x
z) is chosen to be coincident with this atom.
Although only a few atoms are illustrated, we
imagine the lattice extending a very great dis- +sxy +exy
tance in all coordinate directions. Furthermore
we imagine that every lattice plane that is paral-
y
lel to the (x, y)-plane is identical. Thus, the column Fig 8.8 (a) Dislocation loops (fine white lines) in silicon
of atoms is actually an extra half-plane of atoms in (Friedel, 1964). (b) Two-dimensional illustration of crystalline
the (y, z)-plane, and the straight line parallel to the lattice distorted by an edge dislocation with Burgers vector,
z-axis that marks the base of the half-plane is the b (Weertman and Weertman, 1964). Signs of shear stress
dislocation line. In this sense the edge dislocation is and strain shown in inset.
a linear defect in the crystalline solid, but dislo-
cation lines also may form loops as illustrated in dislocation are measured by completing a circuit
Fig. 8.8a. A tangent vector, t, parallel to the dislo- along the rows and columns of atoms in a plane
cation line, orients the dislocation in the crystal that is perpendicular to the dislocation line, in
lattice. The direction of t is arbitrary, but here it is this case the (x, y)-plane (Fig. 8.8b). By convention
taken in the positive z-coordinate direction. the circuit is taken clockwise when the view is in
The magnitude and direction of the edge the direction of the tangent vector, and the circuit
is made with an equal number of steps (from atom (a) Tangent z

to atom) along parallel paths. For example, start- vector, t
ing at the arbitrarily chosen atom s, one could
take four steps down, ve steps to the left, four
steps up, and ve steps to the right, nishing at x
atom f. If the lattice were undistorted, s and
Glide Burgers
f would be the same atom. The fact that the
plane vector, b,
circuit does not close reveals the presence of the
bx > 0
edge dislocation. The magnitude of the vector, b
bxex, extending from atom f to atom s mea-
sures the distortion of the lattice and this vector Dislocation y
is directed in the positive x-coordinate direction, line
i.e. bx 0. This is called the Burgers vector of the dis-
location. For an edge dislocation the Burgers (b)
vector is always perpendicular to the dislocation Tangent z
line. For the so-called screw dislocation the Burgers Burgers vector, t
vector is parallel to the dislocation line. Dislo- vector, b,
cation loops (Fig. 8.8a) are composed of segments bx < 0
of edge, screw, and mixed dislocations. x
An edge dislocation line can move through a Glide
crystalline solid parallel to the plane containing plane
the tangent vector and the Burgers vector. This
plane is called the glide plane. A fundamental
feature of the deformation associated with the
Dislocation y
edge dislocation is that the displacement eld is
line
discontinuous across that portion of the glide
plane shown as shaded in Fig. 8.9. The edge dislo- Fig 8.9 Geometric relationships among the glide plane,
cation is symbolized in this gure with an tangent vector, Burgers vector, and dislocation line
inverted T such that the cross-bar lies in the (Weertman and Weertman, 1964). (a) Positive edge
glide plane parallel to the Burgers vector, and the dislocation. (b) Negative edge dislocation.
upright bar points in the direction of the extra
half-plane of atoms. Note that the vector compo- the dislocation line (Fig. 8.10a). This small volume
nent bx is positive when the extra half-plane is called the dislocation core, and the radius of the
extends in the negative y-coordinate direction. core is estimated using the theoretical strength, S,
The extra half-plane of atoms is located 90o coun- of solids. For example, as we demonstrate below,
terclockwise from the direction of the Burgers the magnitude of the shear stress, s, at a radial
vector when looking in the direction of the distance, r, from the edge dislocation is propor-
tangent vector. tional to the elastic shear modulus, G, and to the
Dislocations exist at the atomic scale (Fig. 8.8) magnitude of the Burgers vector, b, and inversely
where the concept of a continuum is violated, yet proportional to r, such that |s|~Gb2 r. The
the continuum concept is inherent to elastic range of theoretical strengths, S, as a function of
theory. Furthermore, the distortions of the crystal the elastic shear modulus is G/30 S G/3
lattice very near the dislocation are likely to be (Weertman and Weertman, 1964, p. 35). Conser-
greater than the limiting strains imposed on the vatively, taking the lower end of the range for
linear theory of elasticity. None-the-less, elastic strength and setting the radius of the core, rc, to
solutions have proved invaluable for the investi- be that at which the stress equals the strength, we
gation of these defects with the proviso that the estimate rc 5b. Note that the shear stress, s,
mechanical elds so calculated are not applicable from the elastic solution is singular at the dis-
inside a small cylindrical volume that surrounds location where r 0, suggesting that inelastic
(a) ux
b
2
tan 1
y
x

G
2G
xy
x2 y2
(8.36)
r = 5b
Dislocation
Elastic
region
b
uy
2
G
2(2G )
ln
x2 y2
C
core
Glide x

G
2G
y2
x2 y2
(8.37)
plane Here the arbitrary constant C in the expression for

uy has dimensions of length squared, so the argu-
u ment of the log term is dimensionless, and we
assign C a value of one. This constant provides a
Edge uniform translation parallel to the y-axis that can
dislocation r
be ignored in the displacement eld and has no
y
effect on the strain or stress elds which depend
on spatial derivatives of the displacements.
Of the four terms in the displacement equa-
(b) tions, (8.36)(8.37), only the inverse tangent term
is discontinuous for a circuit around the disloca-
Normalized displacement terms, u/b
0.5 tion line (Fig. 8.10b, ux term 1). Here the radius is
0.4
ux term 1 taken as r 10b, outside the dislocation core. Note
0.3
how the contribution from this term to the nor-
0.2 uy term 1
malized displacement, ux/b, is zero at 0 and
0.1
ux term 2 decreases linearly with to a minimum of 0.5 at
0
0.1
, the position of the glide plane. Then, the
0.2 uy term 2 normalized displacement jumps discontinuously
0.3 to 0.5 across the glide plane and thereafter
0.4 decreases linearly to zero at 2. The relative
0.5 displacement, or displacement discontinuity,
across the glide plane (x 0, y 0) is equal to the
50 100 150 200 250 300 350
magnitude of the Burgers vector, in this case a unit
(o)
value. The other three terms in the displacement
Fig 8.10 (a) Schematic illustration of edge dislocation and equations are necessary to satisfy the governing
dislocation core inside of which the deformation is inelastic. equilibrium equations, but do not contribute to
(b) Plot of terms in the displacement equations, (8.36) and the displacement discontinuity.
(8.37), versus angle defined in (a). The kinematic equations (8.33) are used to
determine the strain components:
deformation would occur there. Outside the core,

and over length scales greater than the radius of
the core, the continuum concept is valid and the
xx
by
2
(3G 2)x2 Gy2
(2G )(x2 y2)2 (8.38)
assumptions of the linear theory are not violated.

This has been veried by direct observation of the
displacement eld (see Chapter 8 frontispiece).
yy
by
2
(G 2)x2 Gy2
(2G )(x2 y2)2 (8.39)

The elastic solution for the edge dislocation is bx 2(G )(x2 y2)
premised on the conditions that the deformation xy (8.40)
2 (2G )(x2 y2)2
may be taken as quasi-static and plain strain. The
solution to the Naviers displacement equations of The denominators for each of the strain compo-
motion, (8.31) and (8.32), is (Weertman and nents is proportional to r 4 where r (x2 y2)1/2 is
Weertman, 1964, p. 36): the radial distance from the dislocation in the
(x, y)-plane. Noting that x r cos and y r sin , (a) sxx

the numerators are proportional to br3, so the 100
strain components are proportional to b/r. Thus, 1000
as the distance to the dislocation line becomes
very small, the strain becomes very large. In fact, 500
there is a mathematical singularity in all of the
strain components at the dislocation line (as r 0
y/b
0, ij ). This is mathematically correct, but is 0
non-physical, so we restrict attention to the region
500
outside the dislocation core (r 5b), where the
strain components are nite.
Hookes Law for plane strain conditions (8.34) 1000
is used to determine the stress components: 100

100 0 100
xx
by
2
2G(G )(3x2 y2)
(2G )(x2 y2)2 (8.41)
(b)
x /b
syy
100 1000
yy
by
2
2G(G )(x2 y2)
(2G )(x2 y2)2 (8.42) 800
600

bx 2G(G )(x2 y2) 400
xy (8.43) 200
2 (2G )(x2 y2)2
0 0
y/b
Like the strain components, the stress compo-

200
nents are proportional to b/r, so they also are sin-
400
gular at the dislocation line. As mentioned above
we restrict attention to the region outside the dis- 600
location core where the stress components are 800

less than the strength of the material. Note that 100 1000
the presence of the edge dislocation alters the 100 0 100
normal stress, zz, parallel to the dislocation line x /b
(c) sxy
according to (8.35). 100 1000
The distributions of the stress components
800
near a positive edge dislocation are illustrated in
600
Fig. 8.11 as contour maps for a region that is 200b
400
on a side, omitting the dislocation core where r
5b, and using G 3 104 MPa. The y-axis is pos- 200
itive downward, so the view is in the direction of 0 0
y/b
the tangent vector and the extra half-plane 200

extends upward from the origin along y 0. All of 400
the stress components decrease in magnitude 600
away from the dislocation line (as r , ij 0). 800
The normal stress, xx, is tensile in the region y
100 1000
0, and compressive in the region y 0 (Fig. 8.11a). 100 0 100
That is, a compressive stress is induced on both x/ b
sides of the extra half-plane of atoms, and a tensile
stress is induced off the end of the half-plane. The Fig 8.11 Contour maps of stress components near a
main lobes of contours of the normal stress, yy, positive edge dislocation omitting the dislocation core.
(a) Normal stress, xx. (b) Normal stress, yy. (c) Shear
are similarly distributed, with compression to
stress, xy.
both sides and tension off the end of the extra
half-plane of atoms (Fig. 8.11b). In a thought (a) (b)

experiment one can cleave the crystal and insert
the extra half-plane, thereby pushing the atoms
aside and inducing this compressive stress. The
wedging action of the inserted half-plane of atoms
induces a tensile stress beyond the dislocation (c) (d)
line. Of course this is not the way an edge disloca-
tion is created, but the effect on the stress eld is
similar. Because of this similarity one can use the
edge dislocation to model the wedging action of
the uid pressure in a dike or vein. Fig 8.12 Pairs of edge dislocations used to model
The main lobes of contours of the shear stress, geological structures. (a) Opening fractures including joints,
veins, dikes, and sills. (b) Closing fractures including solution
xy (Fig. 8.11c), are aligned with and symmetric
surfaces and compaction bands. (c) Map view of right-lateral
about the glide plane (x-axis). The shear stress
strike slip fault. (d) Map view of left-lateral strike slip fault.
lobes are positive along the slipped portion (x 0),
and are negative along the un-slipped portion of
the glide plane (x 0). These signs are consistent tangent vectors for both dislocations in the pair
with the shear strain along the glide plane. For are directed into the page. Two edge dislocations
example, note that distortion of the originally with parallel but offset glide planes and opposite
square lattice to the left of the dislocation line in signs constitute an opening or a closing defect
Fig. 8.8b is bottom to the right. With the positive in the lattice, depending upon their congura-
y-axis directed downward this distortion is a position. If the extra planar segment of atoms extends
tive shear strain and is associated with a positive between the two dislocations (Fig. 8.12a), the sur-
shear stress. To the right of the dislocation line rounding lattice must spread apart to accommo-
the distortion is bottom to the left, so the shear date this defect, so this is described as an
strain and stress are negative. In a thought exper- opening distortion. If the extra half-planes
iment to move the edge dislocation from its extend away from the two dislocations (Fig. 8.12b),
former location to its current location one can the surrounding lattice distorts into the gap left
shear the columns of atoms in a negative sense between the dislocations, so this is described as
(top to right) by application of a negative shear closing. The pair of edge dislocations that cor-
stress until the original bonds (dashed lines in Fig. respond to an opening displacement discontinu-
8.8b) break and new bonds (solid lines) are estab- ity serves as a model for joints, veins, dikes, and
lished on adjacent columns, thereby leaving the sills. The pair that corresponds to a closing dis-
half-plane above the dislocation unconnected and placement discontinuity serves as a model for
appearing to be extra. The displacement discon- solution surfaces and compaction bands.
tinuity across the glide plane of the edge disloca- If the two edge dislocations share the same
tion is top to the right, similar to right-lateral slip glide plane and have opposite signs the planar
across a fault or shear fracture. By analogy, slip on segment between them is a sliding defect in the
such a structure would be induced by a remotely lattice. Columns of atoms that were originally
applied negative shear stress and would induce a continuous across this plane are offset by relative
negative shear stress concentration off the end of motion parallel to the glide plane and perpendic-
the structure and a positive shear stress (drop) to ular to the dislocation lines. The sense of sliding
the sides. depends upon the conguration of the disloca-
In two-dimensional analyses, dislocations are tions such that the relative motion is always
viewed as innitely long straight lines oriented toward the adjacent extra half-plane of atoms. In
perpendicular to the eld of view. Pairs of disloca- map view, right-lateral (Fig. 8.12c) and left-lateral
tions can be used to create defects in the crystal strike slip faults (Fig. 8.12d) may be modeled with
lattice where the extra plane segment of atoms is appropriate pairs of edge dislocations. These
bounded in extent (Fig. 8.12). In this gure the gures may be viewed as arbitrary cross sections
within the crust so these pairings of edge disloca- 124o W 120o W 116o W
tions become models for thrust and normal faults
as well. 42o N
0 200 km
8.3.3 A compelling example of elastic 40o N
deformation: the Hector Mine
earthquake
To provide a specic example at the kilometer- 38o N
scale and to motivate the further application of
linear elastic models to structural problems we 36o N
describe the surface displacement eld for the
magnitude 7.1 Hector Mine earthquake that 34o N
occurred on October 16, 1999, in southern
California (Fig. 8.13). This earthquake is associated
with the earlier Landers earthquake (Sieh et al.,
1993), a magnitude 7.5 event that occurred on
June 28, 1992. Both earthquakes occurred in a A B
desert region to the northeast of the San Andreas
Fault. The desert terrain provided an excellent
opportunity to observe and measure the rupture
traces at the surface and to map the details of the northeast br
shear zones and the crustal-scale deformation
La
vic
(Massonnet et al., 1993). The Hector Mine earth-

La
quake was associated with surface rupture on

anch
ke
several faults, including Lavic Lake, Bullion, and

F
au
Mesquite Lake, which are arranged in a complex

lt
pattern trending from northerly to northwesterly B'

over a distance of almost 50 km (Treiman et al.,
2002). The individual fault segments are arranged
in parallel, intersecting, echelon, and curving pat-
terns. The offset of natural and cultural features
was dominantly right-lateral strike slip with a N
maximum of about 5 m. Here we focus on the C
broad pattern of displacements at the Earths D
surface as revealed by synthetic aperture radar (SAR)
Ea
E
st We
interferometry (Zebker et al., 1994; Price and

Bu st
Sandwell, 1998).
llio Bu
M
D'
n llio
es
The use of SAR for the detection of ground dis-

Fa n
qu
ul Fa
ite
placements associated with tectonic events such 0 10 km C'

t
La
as earthquakes was highlighted in an article

ke
ul
appearing in Nature in July of 1993 (Prescott,

Fa
E'
ul
1993). The radar signal is transmitted from a satel- Mapped surface rupture
t
A'
lite (in this case ERS-1 from an altitude of 785 km) Surface trace of model
to the ground surface where it is reected back to
the satellite and recorded as a set of pixels Fig 8.13 Location map for the Hector Mine earthquake,
making up an image of the surface, each pixel showing California with box indicating region of earthquake
representing an area of about 100 m2 on the rupture, expanded in lower part of figure. Reprinted from
Maerten et al. (2005) with permission of the Seismological
ground. Knowledge of the travel time and speed
Society of America.
of the signal provides the information necessary
throughout this region, both tens of kilometers

from the fault traces, and, in the very near eld,
within a few kilometers of the traces. The dis-
placement eld has been modeled using both ver-
tical (Price and Burgmann, 2002) and steeply
dipping ( Jonsson et al., 2002) rectangular fault seg-
ments. The latter investigation suggests that the
segments dip about 83 to the east, but usage of
dipping rectangular segments leads to gaps and
overlaps in the fault geometry that may be un-
realistic.
The model presented here (Maerten et al., 2005)
consists of a linear elastic half-space with a set of
six fault surfaces that honor the details of the
observed surface ruptures at a kilometer scale
Fig 8.14 Synthetic aperture radar interferogram for time (Fig. 8.15). The half-space is bounded by a planar
spanning Hector Mine earthquake. Reprinted from Jonsson et surface that extends out to innite distances hori-
al. (2002) with permission of the Seismological Society of zontally and is free of tractions. The model also
America. extends to an innite depth, but neither of these
features is problematic because the faults are very
to calculate the range, or distance, from the satel- small compared to the radius of curvature of the
lite to each reective site on the surface. If the Earth. Each fault is composed of triangular ele-
same region is imaged at two different times, and ments, with an average side length of about
if the reectivity of the surface has not changed 2.6 km that t together in a continuous surface.
signicantly, the difference between the two Using the data from the interferogram (Fig. 8.14)
images can be used to calculate the component of and the given geometry of the faults, the slip dis-
the surface displacement directed along the line tributions on the faults was determined using a
between the satellite and the ground surface for linear inversion. The slip is partitioned among the
each pixel. Because the satellite will not be in pre- six faults in a manner similar to that observed
cisely the same position for the two images the along the rupture traces. Most impressive,
effects of the topography must be removed. The however, is the correspondence between the syn-
resulting image is called an interferogram. Note thetic and the actual interferograms.
that the data contained on the interferogram are As you might imagine the success of this
not the displacement vector at each pixel, but one method of observation for monitoring and mod-
component of that vector. eling displacement changes on the order of a few
To investigate deformation associated with centimeters over regions on a crustal scale has
the Hector Mine earthquake, SAR images for stimulated a great deal of interest (Fialko et al.,
September 15, 1999, and October 20, 1999, were 2003). It is clear that this technology has the
selected and compared (Jonsson et al., 2002). The potential for a wide variety of applications in
interferogram (Fig. 8.14) shows a pattern of white structural geology. For our present purpose,
and black bands (called fringes) representing con- however, the most compelling conclusion from
tours of the displacement component with each this study is the applicability of models using
cycle representing an interval of 10 cm. The more linear elastic properties and innitesimal strains
prominent fault segments are shown as ne lines to deformation in the Earth. Although the
superimposed on this image. In this broad region, authors of the quote at the beginning of this
about 90 80 km in size, there is a distinct chapter wrote about the efcacy of the linear
pattern of fringes with discreet lobes that extend theory of elasticity long before the Hector Mine
outward to the northwest and southeast. This earthquake they likely would nd the correspon-
technology can resolve the displacement eld dence between Figs. 8.14 and 8.15 very satisfying.
8.4 Quasi-static traction boundary

value problems
In this section we take the equations of equilibrium,

written in terms of the stresses, as the dependent
variables. Constraints on the innitesimal strain
eld that arise from the kinematic relations require
that we incorporate conditions of compatibility,
written in terms of the stress components, into the
set of governing equations. Solving these equations
for the stress components, the strains are calculated
using Hookes Law and the displacements are cal-
culated by integration of the kinematic relations.
The Airy stress function is introduced in Cartesian
coordinates and used to solve a complete two-
dimensional problem for a rectangular region of
Earths crust subject to simple tectonic and gravita-
tional loading conditions. The generalized solution
for two-dimensional problems in polar coordinates
is introduced and used to model a cylindrical valley
in an elastic half-space.
8.4.1 The Airy stress function in two-

dimensional Cartesian coordinates
The six independent stress components (8.19) are
reduced to three for two-dimensional problems
(Timoshenko and Goodier, 1970, p. 15). One class of
such problems, referred to as plane stress, applies to
very thin plates with no tractions applied to either
surface. Plane stress is widely used in engineering
applications, but it has fewer applications in struc-
tural geology. Although the crust of the Earth can
be approximated geometrically as a thin plate or
spherical shell in the context of plate tectonics,
only the upper surface is traction free (neglecting
wind shear and atmospheric pressure). Sedi-
mentary strata also can be approximated geomet-
rically as thin plates, but they are bounded on
both sides by layers that impose non-zero trac-
tions. Therefore, the stress conditions in the in-
Fig 8.15 Elastic models used to investigate the Hector
terior of these thin plates are unlikely to be
Mine earthquake. (a) Six fault surfaces. (b) Fault surfaces
approximated by plane stress conditions.
composed of triangular dislocation elements. (c) Model
interferogram. Reprinted from Maerten et al. (2005) with A second class of two-dimensional problems,
permission of the Seismological Society of America. called plane strain, was dened in (8.30) and led to
a pair of governing equations, (8.31) and (8.32),
with the two in-plane displacement components
as the dependent variables. Here, in contrast, the
8.4 QUASI-STATIC TRACTION BOUNDARY VALUE PROBLEMS 309
(a) tx(n) xxnx yxny

t(n) (8.45)
ty(n) xynx yyny
y The governing equations for the two-dimen-

sional traction boundary value problem include
the equations of static equilibrium written in
x terms of the stress components (7.1397.141).
Taking the (x, y)-plane as the plane of interest and
Boundary using (8.44) these equations reduce to:
xx yx
0 (8.46)
x y
xy yy
g* 0 (8.47)
x y
tx(n) Mass density, , and gravitational acceleration, g*,

(b) are presumed to be known constants. Also, the y-
y t(n) coordinate is taken as vertical and positive
n upward so the body force per unit volume acts in
ty(n) the negative y-direction with a magnitude g*.
Y a These equilibrium equations assure that the
sxx
spatial variation of the stress components in the
sxy
x (x, y)-plane are such that the net force is zero in
both coordinate directions on every element in
syx the body. The implicit condition, xy yx, assures
that the net torque about the z-axis is zero. These
syy conditions apply to every innitesimal element,
as well as every macroscopic portion of the body,
Fig 8.16 Schematic illustration of traction boundary value and to the entire body.
problem. (a) Two-dimensional plane strain conditions with The innitesimal strain components are calcu-
tractions, t(n), distributed on internal and external
lated from the stress components using (8.18):
boundaries. (b) Element used to define relations among the
traction components and stress components. 1
xx [xx(1 2) yy (1 )] (8.48)
E
1
dependent variables are the three in-plane stress yy [yy (1 2) xx (1 )] (8.49)
components which only are functions of x and y: E
1
xy xy (8.50)
xx f1 (x, y), xy f2 (x, y), yy f3 (x, y) (8.44) E
The out-of-plane-normal stress is given by (8.35), The out-of-plane strain components are zero. The
and the out-of-plane shear stresses are zero. nal step in a complete solution is the determi-
Boundary conditions are specied in terms of the nation of the in-plane displacement compo-
distribution of tractions, t(n), acting on exterior nents. Using the kinematic equations for plane
and interior surfaces of the body (Fig. 8.16a). The strain (8.33), one must determine two displace-
outward unit normal, n, lies in the (x, y)-plane ment components from three strain compo-
everywhere on these surfaces. These tractions nents. This problem is over-determined and the
induce a distribution of stresses throughout the relationship among the strain components that
body, and sufciently close to a boundary the appropriately restricts their spatial variations is
stresses must be in equilibrium with the tractions the rst of St. Venants compatibility equations
according to Cauchys Formula (Fig. 8.16b): (7.143):
2xx 2yy 2xy

2 2 0 (8.51) This is the governing equation for plane strain
y2 x xy
problems of elasticity with constant gravitational
This is the two-dimensional compatibility equation, body force and stress components as the depen-
written in terms of the strain components. dent variables. Solutions to particular problems
Because we are taking the stress components require a stress function that satises (8.56) and
as dependent variables, (8.48)(8.50) are used to provides functions for the stress components
rewrite the compatibility equation in terms of the using (8.55) that satisfy the traction boundary
stress components: conditions. Textbooks in elasticity theory give
2 2 numerous examples of Airy stress functions that
2
[(1 )xx yy] 2 [(1 )yy xx] apply to a wide variety of problems in structural
y x
geology (Timoshenko and Goodier, 1970; Jaeger
2xy
2 0 (8.52) and Cook, 1979; Barber, 1992).
xy
Particular solutions for the biharmonic equa-
Differentiating (8.46) with respect to x, and (8.47) tion in the form of polynomials serve to illustrate
with respect to y with and g* constant, and simple states of stress. For example, choosing only
adding we nd: terms up to rst order and using C0, C1, and C2
for constants, the Airy stress function is
2yy 2xx 2xy
2
2
2 0 (8.53) (x, y) C0 C1x C2y. Substituting this stress
y x xy
function into (8.55) we nd:
Adding (8.52) and (8.53) we have:
xx g *y;xy 0;yy g *y (8.57)
2
2
y x
2

2 (xx yy) 0 (8.54) Note that the shear stress is zero and the con-
stants play no role in the functions for the stress
The compatibility equation for plane strain and
components. A particular region in the (x, y)-plane
constant body forces reduces to this harmonic
(shaded rectangle in Fig. 8.17a) is illustrated with
equation for the sum of the two in-plane normal
the traction boundary conditions (8.45) corre-
stress components.
sponding to this solution. The upper surface is
Solutions for the three in-plane-stress compo-
traction free and therefore could represent
nents must satisfy equilibrium (8.46)(8.47) and
Earths surface. If the material is incompressible,
compatibility (8.54). A method for solving these
0.5, this Airy stress function gives the isotropic
equations was proposed by G. B. Airy in 1862
state of stress equivalent to Andersons standard
(Timoshenko and Goodier, 1970, p. 32) and the
state: xx yy zz g*y. For values of Poissons
function used in this method, (x, y), commonly is
ratio less than 0.5, the out-of-plane-normal stress,
referred to as the Airy stress function. The stress
zz 2g*y, is somewhat less than the in-plane-
components are written in terms of this stress
normal stresses. For the perfectly compressible
function as:
material, 0, the out-of-plane-normal stress is
2 2 zero.
xx g *y,yy 2 g *y,
y 2 x The region of interest in the (x, y)-plane is arbi-
2 (8.55) trarily chosen, but the mathematical solution for
xy the state of stress applies everywhere in the plane.
xy
This can lead to states of stress that satisfy the gov-
Direct substitution demonstrates that these func- erning equation (8.56), but are not appropriate for
tions satisfy the equilibrium equations, (8.46) the physical problem under consideration. For
(8.47), and substitution into the compatibility example, if the region of interest in Fig. 8.17a were
equation (8.54) yields the biharmonic partial dif- to be extended upward to positive values of y, the
ferential equation: resulting normal stresses would be tensile (posi-
4 4 4 tive), where y is positive. This would not represent
2 2 2 4 0 (8.56)
x 4 x y y stresses induced by the weight of the body and
(a) y L
y
tx = 0, ty = 0 x
tx = 0, ty = 0 x
tx = rg*y tx = rg*y
ty = 0 H ty = 0
tx = rg*y H tx = CL + rg*y
ty = Cy ty = Cy
tx = 0, ty = rg*H
(b) tx = CH, ty = rg*H
y Fig 8.18 Schematic illustration of plane strain elastic

problem used to investigate crustal-scale faulting (Hafner,
C3 D
1951; Hubbert, 1951).
C5 C4 x C5
C5 are constants with dimensions the same as
stress. The in-plane stress components are:
H
C4 C4 xx C5 g*( y D), yy C3 g*( y D),
rg*(y D) xy C4 (8.59)

rg*(y D) C4
The two normal stresses include a linearly varying
C3 part that accounts for the gravitational body force
and constant terms, C3 and C5, that represent a
Fig 8.17 Schematic illustration of plane strain elastic uniform tectonic loading (Fig. 8.17b). The tectonic
problem with traction distributions on rectangular tractions acting on the boundary of the rectangu-
boundaries (Timoshenko and Goodier, 1970). (a) Free lar region are drawn for positive values of these
surface at top and lithostatic loading in the interior of a half-
constants, and the normal stresses associated
space. (b) Interior of half-space with lithostatic and tectonic
with these constant terms would be tensile.
loading.
Changing the sign of the constants produces com-
pressive tectonic stresses within the region. The
therefore would be a physically inappropriate sign of the tectonic shear stress depends on the
application of this solution. sign of C4.
If the region of interest is entirely below the
traction-free surface, (8.55) can be modied to 8.4.2 From stress to strain to
account for a translation of the origin to a depth, displacement fields in Cartesian
D, at the top of this region (Fig. 8.17b): coordinates
To illustrate the complete solution for a plane
2 2
xx 2
g*( y D);xy ; strain problem with traction boundary condi-
y xy
tions, including stress, strain, and displacement
2 elds, we consider the following stress function
yy g *( y D) (8.58)
x2 (x, y) 12 Cxy2. Here C is taken as a positive con-
We use this conguration to investigate higher- stant and the coordinate system and regions of
order terms in the stress function and thereby interest are illustrated in Fig. 8.18. This stress func-
introduce stresses that may be attributed to tec- tion was used by M. King Hubbert to investigate
tonic forces. For example, consider the stress func- the mechanical basis for crustal-scale faulting
tion (x, y) 12 C3x2 C4xy 12 C5 y2. Here, C3, C4, and (Hubbert, 1951), and by Willy Hafner to investigate
thrust faulting in fold and thrust mountain belts Note that the strains scale with the same con-
(Hafner, 1951). The in-plane stress components are stant, C, that scales the stresses. Also, the strains
found using (8.55): are inversely proportional to Youngs modulus
and directly proportional to a factor that includes
xx Cx g *y,yy g *y,xy Cy (8.60) Poissons ratio. Although there is no tectonic
normal stress in the y-direction, there is a tectonic
Both normal stress components include a contri-
normal strain in this direction, induced by the tec-
bution from the weight of the material that
tonic normal stress in the x-direction. The region
increases in compression with depth (negative y).
x 0 (Fig. 8.18) is one of horizontal tectonic exten-
The tectonic part of the horizontal normal stress
sion (xx 0), so the vertical strain is a contraction
varies linearly in the x-direction and is zero along
(yy 0). The region x 0 is one of horizontal tec-
the (x 0)-coordinate axis. The tectonic shear
tonic contraction (xx 0), so the vertical strain is
stress varies linearly in the y-direction from zero
an extension (yy 0). In both regions the strains
along the (y 0)-coordinate axis, taken here as
increase in magnitude linearly from x 0.
the traction-free representation of the Earths
The tectonic displacement eld is found by
surface. The only regions of interest in the (x, y)-
integrating the kinematic equations (8.33) after
plane are in the third and fourth quadrants of the
substituting (8.61):
coordinate system, where y 0 and the gravita-
tional normal stresses are compressive (negative).
For a rectangular region in the fourth quadrant
(x 0), the tectonic normal stress is tensile and

ux xxdx E 2
1 2 1 2
Cx f1(y) C1 (8.62)
increases linearly with distance from the left-

hand side of the region (Fig. 8.18). On the other
uy yydy

E
2
Cxy f2(x) C2 (8.63)
hand, for a rectangular region in the third quad-

rant (x 0), the tectonic normal stress is com- To evaluate the arbitrary functions and constants
pressive. we use the kinematic equation for the shear strain
Hookes Law (8.48)(8.50) provides the in-plane (8.33) rearranged as follows:
strain components, but one should consider if the
contribution from the weight of the material is
relevant. This part of the strain eld represents
ux uy
y

x
2xy 0 (8.64)
the deformation that would be experienced by the Substituting for the displacement components
elastic material if gravity were turned off in the from (8.62) and (8.63) and for the shear strain from
initial state and then turned on for the nal state. (8.61) we nd:
For the Earth, the force of gravity is not turned on
and off. Therefore, most problems in structural
geology are better posed by comparing an initial
df2(x) df1( y)
dx

dy

2
E
Cy 2
1
E
Cy 0
state with gravitation loading, to a nal state with (8.65)
gravitation loading plus the appropriate tectonic
forces. In this context one should ignore the con- The rst term is only a function of x and the last
tributions of gravitational loading to the strain three terms are only functions of y so:
and displacement elds.
Substituting (8.60) for the stress components
in Hookes Law, without the term g*y, the in-
df2(x)
dx
C3 1
dy
df ( y) 2
E
Cy 2
1
E
Cy
plane tectonic strain components are: (8.66)
1 2 2
xx Cx,yy Cx, Rearranging and integrating:
E E
xy
1
E
Cy (8.61)
f2(x) C3dx C3x C4 (8.67)
(a) y ux = C 6 not appear in the expressions for the tectonic

L
uy = C 7 strain components (8.61) and stress components
(8.60). This is consistent with the concept that
x rigid translations and rigid rotations do not con-
tribute to the strain or stress elds. Therefore,
these terms are of no relevance to studies that
H depend upon the strain or stress elds. In most
applications the rigid motions are ignored.
8.4.3 Generalized solution for plane

(b) y cylindrical (polar) coordinates
The natural coordinate system for some models of
x geological structures is made up of two in-plane
coordinates, r and , and one out-of-plane coordi-
nate, z (Fig. 8.20). If the structure is very long in
the z-coordinate direction, it may be approxi-
H
mated using plane strain conditions where uz 0
everywhere and the in-plane displacements, ur
and u are only functions of the two polar coordi-
nates, r and . Under these conditions the three
L independent stress components are only func-
Fig 8.19 Schematic illustration of linear and constant tions of the polar coordinates:
terms in displacement equations (8.69) and (8.70).
(a) Constant terms are rigid translation. (b) Linear terms rr g1(r, ),r g2(r, ), g3(r, ) (8.71)
are rigid rotation.
The out-of-plane normal stress is proportional to
Poissons ratio, and the out-of-plane shear stresses

are zero: zz (rr ), rz 0, z 0. Taking
f1( y)
2 2
E
Cy C3 dy
the cylindrical z-axis and the reference axis Ox as
horizontal, perpendicular to the direction of the
gravitational body force, F, near Earths surface,
E 2
2 2 1 2
Cy C3 y C5 (8.68) the components of the body force are:
The functions f1(y) and f2(x) are substituted into Fr g * sin ,F g* cos ,Fz 0 (8.72)
(8.62) and (8.63), and the constants are combined For these conditions the equilibrium equations
to give the displacement components: are (Timoshenko and Goodier, 1970):
ux E 2
1 2 1 2
Cx
E
2 2 1 2
2
Cy C3 y C6 rr 1 r rr
r

r

r
Fr 0 (8.73)
(8.69) r 1 2r

2 F 0 (8.74)
uy Cxy C3x C7 (8.70) r r r
E
The equilibrium equations are satised, as may be
The last two terms in each displacement equation shown by substitution, with polar stress compo-
have a special interpretation. The constants, C6 nents related to an Airy stress function, (r, ), as
and C7, represent a rigid translation of the region of follows (Timoshenko and Goodier, 1970):
interest (Fig. 8.19a). The two terms with the con-
stant C3 represent a rigid rotation of the region 1 1 2
rr g*r sin (8.75)
about the origin (Fig. 8.19b). These constants do r r r2 2

1 (a)
r (8.76)
r r su r sru srr
2 suu
g*r sin (8.77) Fu
r2
Fr
To solve a particular problem the stress compo-
nents must satisfy the specied traction boundary
conditions following Cauchys Formula.
Hookes Law is dened by noting that one may
rotate the Cartesian coordinates (x, y) about the z-
axis to coincide with the polar coordinates (r, ). r u
Then, the strainstress relationships in polar
coordinates are written by exchanging each x
z
Cartesian subscript in (8.48)(8.50) with its polar
counterpart:
(b) (c)
1
rr [rr(1 2) (1 )] (8.78)
E
1
[(1 2) rr (1 )] (8.79)
E
1
r (8.80)
E r
The kinematic equations in polar coordinates are (d) (e)

(Malvern, 1969):
ur
rr (8.81)
r

1
u
u
r r (8.82)
(f) (g)
r

1 ur
2r
u
r u
r (8.83)
The three strain components are used to deter-

mine two displacement components, so these
strains must satisfy a compatibility condition
which is written in terms of the Airy stress func-
tion, giving the biharmonic equation in polar Fig 8.20 (a) Cylindrical coordinates (r, , z) and associated
coordinates (Malvern, 1969): polar stress components. (b) Schematic illustrations of six
different elastic boundary value problems solved by selected
2 1 1 2

r2 r r r2 2 2 1 1 2

r2 r r r2 2
0 terms from Michells general solution in polar coordinates
(8.71) (Michell, 1899).
(8.84)
Solutions to this equation for stress states that In 1899, J. H. Michell derived a generalized
satisfy particular boundary conditions are Airy stress function in polar coordinates, from
reviewed in textbooks on elasticity theory which many particular solutions can be extracted
(Timoshenko and Goodier, 1970; Barber, 1992). (Michell, 1899; Timoshenko and Goodier, 1970):
(r,) a0 ln r b0r2 c0r2 ln r d0r2 a0 r

tr = 0, tu = 0

1 u x
ar sin (b1r 3 a1r1 b1r ln r) cos z
2 1

1
cr
2 1
cos (d1r 3 c1r1 d1r ln r) sin

D = r sin u
(a r
n2
n
n b r n2 a rn b rn2) cos n
n n n 2R

(c r
n2
n
nd
nr
n2 crn d rn2) sin n
n n suu
(8.85)
For r >> R,
For example, the rst three terms of (8.85) solve srr = suu
srr
the problem of symmetric loading of a hollow = rg*r sin u
cylinder (Fig. 8.20b), which is used for in-situ deter-
Fig 8.21 Schematic illustration of elastic half-space with
mination of elastic moduli in boreholes. The
half-cylindrical cut from traction-free surface used as model
fourth term is used to solve the problem of an for a valley.
elastic half-space with a distributed constant
normal traction (Fig. 8.20c). This nds many appli-
cations to loading of Earths surface. The last term and the z-axis is parallel to the valley. The shape of
on the rst line provides a pure shear stress as in the valley is idealized as a circular cylinder and it
a twisted rod (Fig. 8.20d). Terms on the second and is very long in the z-direction compared to the
third lines are used to solve the problem of a point radius, R. The geometry in the plane of interest
force in an innite body (Fig. 8.20e). Combinations is that of the half-space, r sin () 0, with a semi-
of point forces have been applied as earthquake circular cut that removes the elastic material
source mechanisms and the integration of the where r R.
point force solution over a boundary is a standard The surface of the elastic half-space with the
technique for developing solutions to new prob- cut is traction free. Thus, the conditions on the
lems. The rst term on the third line is used below three segments of the boundary are:

to solve the problem of a cylindrical valley loaded r R, 2
by gravity (Fig. 8.20f). The terms on the fourth line BC: on r R, , tr 0, t 0
with n 2 are the solution for a cylindrical hole in
r R, 2 (8.86)
a body with a uniform normal stress at innity
(Fig. 8.20g). This has been called the most impor- Prior to the valley being incised the polar stress
tant problem in rock mechanics (Jaeger and Cook, components are given by (8.75)(8.77) without the
1979, pp. 249) and apparently was rst obtained by terms containing the Airy stress function:
B. Kirsch (1898; Timoshenko and Goodier, 1970,
rr g *r sin ,r 0 (8.87)
p. 90). It has been applied to the stress distribution
around boreholes and tunnels, and to a variety of Because the valley provides only a local perturba-
in-situ stress measurement techniques. tion of this stress state the boundary conditions in
the remote eld are:
8.4.4 The stress state induced by gravity

rr g * r sin
near a valley
As an example of a particular solution taken from BC: as r and sin 0, g * r sin
Michells generalized solution (8.85), consider the r 0
state of stress near a valley excised into an other- (8.88)
wise featureless (planar) terrain (Fig. 8.21). The The normal stress components are equal com-
polar coordinate system (r, ) is chosen such that pressions that increase in magnitude linearly
the x- and z-axes are parallel to the planar surface with depth, D r sin.
To account for the stress perturbation of the (a)

srr (valley)
valley, the rst term in the third line of (8.85) is 0
taken with the constant c1 g*R2, so the Airy 500 20
stress function is: 1000
1500
(r, ) 12 g *R2r cos (8.89) 15
2000
The polar stress components are found using 2500
(8.75)(8.77): 3000
10

R 2 3500
rr (g* r sin ) ,r 0, 0 5
r 4000
(8.90)
4500
5000 0
Everywhere on the circular boundary, r R, the 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
perturbing radial stress is just sufcient to reduce

the gravitational radial stress to zero. Both the
gravitation and perturbing shear stress are zero.
Therefore the boundary conditions on the valley (b) srr
0
wall are satised. Because the sin terms are zero
500 20
on the horizontal segments of the boundary, it is
30
traction free. Because the perturbing stress scales 1000
40
with (R2/r), its contribution goes toward zero for 1500
50
radial distances that are large compared to the 2000
60
valley radius and the stress state approaches that 2500
70
due to gravity alone (8.88). 3000 80
The distribution of the perturbing radial stress 3500 90
is illustrated in Fig. 8.22a for the particular case 4000 100
where R 103 m and g* 0.025 MPa m1. The per- 110
4500
turbing stress is tensile and greatest at the valley 120
5000
bottom, decreasing toward zero at the upper 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
edges of the valley. Contours of equal radial stress

curve around the base of the valley and the dis-
tance between the contours increases with radial (c) suu
distance because of the (R2/r) distribution. These 0
characteristics of the stress distribution indicate 20
500
that there is a concentration of stress at the valley 30
1000
bottom. For this case the perturbing radial tensile 40
1500
stress there is 25 MPa. In other words the stress 50
concentration is equal in magnitude to the gravi- 2000
60
tationally induced compressive stress at 1 km 2500 70
depth, but of opposite sign. 3000 80
The total radial stress, found by adding (8.87) 3500 90
and (8.90), is illustrated in Fig. 8.22b. Note how the 100
4000
contours of equal radial stress are depressed 110
4500
under the valley, but return to nearly horizontal 120
5000
lines for depths greater than a few valley radii. 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
Because the gravitational stress at the valley
Fig 8.22 Contour maps of stress components near an
bottom is 25 MPa, the total radial stress is
idealized valley. (a) Perturbing normal stress, rr. (b) Total
exactly zero there (and everywhere else on the
normal stress, rr. (c) Normal stress, .
boundary). The circumferential normal stress is
not perturbed by the valley, so it remains com- 109o 45' 109o 30'
Span CITY OF MOAB
pressive and increases in magnitude linearly with ish V
alley
Salt
depth below the horizontal surface. Contours of Antic
line
(Coll
the total circumferential stress are straight hori- apse
38
Kings Bottom d)
o 30'
zontal lines throughout the half-space including Sa
lt
Sy
ncl
the vicinity of the valley (Fig. 8.22c). The total cir- ine
Cane Creek
cumferential stress at the valley bottom is: Salt Anticline
Colorado
(r R, 3 2) g*R (8.91)
103
River
o 30'
In other words, the normal stress acting in the
horizontal direction is compressive, and scales in Gray
s Pa
sture
Salt
magnitude with the average unit weight of the
30'
Sync
line
38 o
surrounding rock mass and with the vertical dis- MEANDER
tance from the valley rim to the valley oor. ANTICLINE
38
Because the polar component of total shear stress
o
15
is zero, the stress trajectories are radial lines (least 110o
'
compressive stress) and circumferential curves
Gr
(greatest compressive stress). ee
n
Ri
ve
One of the more interesting structures associ- r
ated with deeply incised canyons are so-called

valley anticlines (Potter and McGill, 1978). Potter THE LOOPS
and McGill identify two types of valley anticlines.

15'
2
The rst type is characterized as broad upright
38 o
109
A Outlines of
folds that follow major drainages and, although 3 Needles
o
Miles
fault
45'
the folding tends to die out upward, it may deect 0 5 205.5 4
zone
strata all the way to the valley rim, up to 600 m B
C 5
Miles
D 6 38o
above the valley oor. The second type is smaller,
sometimes with straighter limbs and tighter Home Springs
Anticline
hinges, and usually is conned to thin-bedded 7
strata on the valley oor. The spatial association of

Gypsum Canyon
landforms and structures often leads to an inter- 8
pretation that the landform (drainage channel in 38o 110o
this case) was superimposed on the structure
Fig 8.23 Map of Colorado River near Canyonlands
because deformation related to the structure
National Park, UT, showing trace of the Meander Anticline.
made the rock more susceptible to erosion.
Reprinted from Huntoon (1982) with permission of The
However, valley anticlines are interpreted in just Geological Society of America.
the opposite way: the uvial valley predated the
anticline and the location of the fold was deter-
mined by the course of the river. the drainage, and even here a secondary anticline
Superb outcrops of both types of valley anti- suggests that canyon incision in one of the loops
clines are found in Canyonlands National Park, did generate a fold. Viewed from the Colorado
southeastern Utah (Potter and McGill, 1978; River (Fig. 8.24a), north of the conuence with the
Huntoon, 1982). The major structure, called the Green River, the Meander Anticline appears as a
Meander Anticline, follows the Colorado River for gentle symmetrical fold with a hinge line nearly
approximately 35 km and the trace of the fold coincident with the center of the river. At Mile
hinge is almost perfectly coincident with the 205.5, the limb dips have increased to 50, and
meandering course of the river (Fig. 8.23). Only farther downstream the fold becomes asymmetric
where the river takes two very pronounced me- with limb dips of only a few degrees to the west
anders at The Loops does the anticline cut across and up to 30 to the east.
(a) 700 m thick near the northern edge of the Needles

fault zone (Fig. 8.23) to zero near the southern
edge. There is ample evidence for gravitational
sliding of the Needles fault zone, where normal
faults bound a set of grabens that have extended
toward the river (McGill and Stromquist, 1979;
Cartwright et al., 1995; Moore and Schultz, 1999).
Huntoon (1982, Fig. 5) suggests that models for
the development of the Meander Anticline can be
(b) discriminated by comparing the stress states
implied by the models to the stress states deduced
from geological structures at two key locations,
the Needles fault zone and the axis of the
Meander Anticline. Specically he deduces from
the normal faults in the Needles fault zone that
the greatest compressive stress is oriented verti-
cally there. Along the axis of the anticline he doc-
uments thrust faults that strike approximately
parallel to the fold axis and dip both eastward and
Fig 8.24 Photographs along the Colorado River. (a) The
westward at low angles (Fig. 8.24b). From these
Meander Anticline viewed north of the confluence with the
Green River. (b) Thrust faults striking approximately parallel
thrust faults he deduces that the greatest com-
to the fold axis near the Needles fault zone. Reprinted from pressive stress is oriented horizontally and per-
Huntoon (1982) with permission of The Geological Society pendicular to the fold axis. Huntoon concludes
of America. Photographs by Peter Huntoon. that only the sliding plate model is consistent
with these two different stress states and suggests
that the sliding plate produces compression in
While most geologists who have written the rocks under the canyon and leads to the for-
recently about the Meander Anticline agree that mation of the Meander Anticline. Potter and
the meandering course of the Colorado River McGill (1978) conclude that excess horizontal
developed before the anticline, a number of compressive stress was responsible for the small
hypotheses have been offered to explain the valley anticlines.
folding. For example, there is some debate about The linear elastic solution for the stress distri-
the role of evaporites (gypsum with minor lime- bution near a cylindrical valley is inappropriate
stone and shale) from the Paradox Member of the for addressing questions about the owage of the
Pennsylvanian age Hermosa Formation that evaporite or about the sliding of the overlying
underlies much of this region. Some have sug- plate on the evaporate because these processes
gested that the evaporites have owed laterally involve inelastic deformation. However, the elastic
and upward toward the canyon bottom because of solution does provide a useful tool for evaluating
stress changes caused by the rapid incision of the the stress distribution due to gravity acting on the
river (Potter and McGill, 1978). There is ample evi- valley alone. The stress perturbation due to forma-
dence for such owage in the four salt plugs tion of a cylindrical valley is a radial tension con-
located along the southern portion of the anti- centrated near the valley bottom (Fig. 8.22a). This
cline (Fig. 8.23, marked A, B, C, D) where evapor- radial tension reduces the lithostatic compression
ites have pierced through the overlying strata to that acted perpendicular to the future valley walls
the surface. Others have suggested that the plate and bottom to zero, thereby unloading the rock
of strata overlying the evaporites is sliding toward mass adjacent to the valley. This unloading occurs
the canyon under the inuence of gravity, while primarily below the valley and undoubtedly
the evaporites remain essentially in place played an important role in the development of
(Huntoon, 1982). This plate varies from about the Needles fault zone. The orientation of the
8.5 LABORATORY AND ENGINEERING FIELD TESTS 319
stress trajectories to the side of the valley, with (a) Uniaxial compression
greatest compression approximately vertical and applied by machine
least compression approximately horizontal is Hydraulic
consistent with the normal faulting. Lead piston head
wires Force
transducer
Hardened steel
8.5 Elastic properties from platen
laboratory and engineering

field tests
Rock specimen
The measurement of rock properties in the lab-
oratory traditionally falls within the discipline Hardened steel platen
called rock mechanics, although a broad variety of Hemispherical cap
scientists and engineers, including geophysicists Hemispherical seat
and structural geologists as well as mining and
civil engineers, contribute to this activity. Struc-
tural geologists use the data from rock property Testing machine steel frame
testing, so we review the procedures and results (b)

while leaving aside many of the technical details.
Some of the details of experiments to determine Electrical
the elastic properties are described in textbooks resistance
strain gage
on rock mechanics (Jaeger and Cook, 1979).
L
Although we have referred to this activity as an
experiment the procedure has become routine
and is considered by some to be a standardized
engineering test. We present selected values for Lead
Youngs modulus and Poissons ratio. wires
8.5.1 Measuring the elastic properties of (c)

laboratory samples Tie-bar
The most direct laboratory experiment to deter-
mine Youngs modulus of elasticity for rock is the Upper cross head
uniaxial compression test. In this test a machine Fluid pressure supply
applies a normal compressive force through a set
Piston
of steel platens onto the circular ends of a cylin- Hydraulic cylinder
drical rock specimen (Fig. 8.25a). The machine, Rod
not shown in the gure, applies the force through
Platen
a screw-driven or hydraulically driven piston.
Because the force is applied only parallel to the Holding device
with specimen
axis of the specimen, this test is called uniaxial.
Indeed, the intention is to induce only a normal
stress, a, of constant magnitude acting parallel to Lower cross head
the axis of the specimen, and then to measure the

axial extension, ea. Youngs modulus for a linear
elastic material would be the ratio of these two Fig 8.25 Equipment used for uniaxial compression testing
quantities (8.6). of rock. (a) Sample with platens, hemispherical cap, and force
transducer. (b) Electrical resistance strain gage. (c) Hydraulic
Although the test sounds simple enough, prob-
testing machine.
lems can lead to spurious data. For example, if the
specimen ends are not perfectly parallel, bending (a) Axial extension, ea
can be induced. Also, depending on differences in 0.003 0.002 0.001 0
0
the elastic properties of the platen and specimen,
and on the friction at their contacts, the specimen
25
Axial stress, sa (MPa)

may be constrained by the platens or pushed radi-
ally outward as the platens expand under load. Colorado
granite 50
Both of these phenomena are accompanied by a
non-uniform stress state on the scale of the entire
Georgia 75
specimen. To understand the possible severity and granite
consequences of this non-uniformity, the state of
100
stress can be calculated using a boundary value
problem for a cylindrical body with the appropri- 125
ate tractions or displacements on the ends (Peng,
1971). The hemispherical cap and seat (Fig. 8.25a), (b) Axial extension, ea
special platen material, and other standard 0.0018 0.0012 0.0006 0
testing procedures are designed to minimize the 0
specimen-scale non-uniformities in the stress 5
eld. At the grain scale the stress and strain elds
Limestone 10
Axial stress, sa (MPa)

are likely to be very non-uniform. The presump- with cracks
tion is that the stress and strain represent average 15
values over the volume of the sample. Because of Limestone
20
the difculties inherent to the testing procedure
we will refer to laboratory results as apparent 25
Youngs moduli.
30
An electronic strain gage can be glued onto the
specimen to determine the axial extension (Fig. 35
8.25b). These gages are composed of thin wires
Fig 8.26 Uniaxial compression test results (Obert and
that undergo a change in electrical resistance as Duvall, 1967). (a) Nearly linear behavior (Georgia granite)
they are stretched or shortened along with the and non-linear behavior (Colorado granite). (b) Loading and
specimen. This change in resistance is trans- unloading data for limestone with cracks and without cracks.
formed into an electronic signal calibrated to the
extension magnitude and read onto a digital
display, chart recorder, or computer storage signal calibrated to the force magnitude. The
device. Thus electrical resistance change is mea- cross-sectional area of the specimen is calculated
sured and axial extension is estimated from this from a measurement of the diameter and this is
measurement via a calibration of the strain gage. used to calculate the axial stress, which is
The gage must be large compared to the grain size assumed to represent the average stress through-
to measure an average extension. Alternatively, out the specimen. Neither the stress nor the strain
mechanical dial gages or electronic displacement are measured directly.
transducers can be attached to the specimen or Data from the uniaxial test include values of
platens to measure the changes in length that, the average normal stress and extension acting
with the original length, are used to calculate the along the cylindrical axis of the rock specimen. If
axial extension. extensions are very small, we assume the strain
A force transducer placed between the piston can be characterized as innitesimal. From these
and the specimen measures the normal force data a graph of stress versus extension can be pre-
acting on the specimen (Fig. 8.25a). In fact, most pared and the slope of the curve used to estimate
force transducers use a strain gage to measure the the apparent Youngs modulus, or the tangent
extension of a small metal part within the trans- Youngs modulus. The relationship between axial
ducer and then transform this into an electronic stress and extension for a specimen of Georgia
8.5 LABORATORY AND ENGINEERING FIELD TESTS 321
Table 8.2. Rock mechanics laboratory tests for Table 8.3. Rock mechanics laboratory tests for
apparent Youngs modulus (GPa). apparent Poissons ratio.
Rock type From To Mean Rock type From To Mean
Quartzite 70 105 90 Quartzite 0.11 0.25 0.16

Gneiss 16 103 68 Gneiss 0.10 0.40 0.22
Basalt 16 101 63 Basalt 0.13 0.38 0.22
Granite 10 74 45 Granite 0.10 0.39 0.23
Limestone 1 92 48 Limestone 0.08 0.39 0.25
Sandstone 10 46 22 Sandstone 0.10 0.40 0.24
Shale 10 44 28 Shale 0.10 0.19 0.14
Pittsburgh coal 1.5 3.7 3.2 Witbank coal 0.33 0.37 0.35
granite is given in Fig. 8.26a (Obert and Duvall, A few generalizations can be made from the
1967). The set of line segments drawn between apparent elastic moduli given in Table 8.2. A par-
successive data points on this graph form a nearly ticular rock type is not associated with a particu-
straight line. In contrast Colorado granite displays lar modulus. This should come as no surprise
a distinctly non-linear behavior. The straight-line because rock classication schemes are based on
segments are not parallel, but these data could be ranges of values of chemical composition, miner-
used to calculate a tangent modulus. alogy, and texture. You have seen (Fig. 8.26b) that
Both granite specimens (Fig. 8.26a) failed by even samples from the same rock mass can exhibit
fracturing shortly after the most compressive different moduli. The metamorphic and igneous
stress shown on the graph was imposed. The rocks typically have greater values than the clastic
results we have given do not include possible data sedimentary rocks, although strongly weathered
on unloading to lesser compressive stresses, which granite can be less stiff than well-indurated sand-
would test whether the deformation of these stone. Speaking qualitatively, we would describe
granite specimens was reversible before fracturing rocks with values around 100 GPa as being very
began. Unloading data are shown in Fig. 8.26b for stiff, whereas rocks with values around 1 GPa
two limestone specimens (Obert and Duvall, 1967). would be termed very soft. From Table 8.2 we draw
One shows a nearly linear relationship between the following rule of thumb: laboratory specimens of
stress and extension, a relatively large modulus, rock have Youngs moduli that range from about 1 to
and approximately reversible behavior. The other 100 GPa with a typical value of about 50 GPa. Rocks
is distinctly non-linear, has lesser moduli, and does are very stiff relative to our experience with softer
not follow the same path in loading and unload- elastic materials like rubber, but they are elastic
ing. This non-linear behavior is attributed to the none-the-less.
presence of abundant microcracks in the second of When cylindrical specimens of rock are short-
the limestone specimens. ened axially in a uniaxial compression test (Fig.
8.25), they respond by expanding perpendicular
8.5.2 Apparent Youngs modulus and to the applied load. The negative ratio of this per-
Poissons ratio pendicular extension to the axial extension is
Representative values of the apparent Youngs Poissons ratio (8.10). For most rock specimens
modulus for selected metamorphic, igneous, and Poissons ratio is approximately constant only
sedimentary rock types are given in Table 8.2 over restricted ranges of loading, time, and speci-
(Bieniawski, 1984). Youngs modulus and the men size. None-the-less, representative values for
tangent modulus are not distinguished in these the same suite of rock types selected for Table 8.2
data. This table is not meant to be complete: one are given in Table 8.3 (Bieniawski, 1984).
can turn to handbooks that have more extensive Unlike values of Youngs modulus that vary
tabulations (Clark, 1966). over two orders of magnitude, values of Poissons
dependence of rock properties. One engineering

Surrounding rock mass eld test is called a plate-bearing test (Fig. 8.27)
because the applied load is transmitted to the
rock surface by a circular bearing plate
Bearing plate
(Goodman, 1980). Hydraulic jacks load the plate
50 to while displacement gages measure the motion of
100 cm the rock face. To calculate the applied stress and
Hydraulic jack resulting strain, these quantities are substituted
into an expression derived from the solution to a
boundary value problem of elasticity. The charac-
teristic length for these tests is the diameter of the
bearing plate, which typically ranges from 50 to
100 cm. In contrast, the diameter of typical lab-
Fig 8.27 Schematic illustration of hydraulic jack used in oratory specimens ranges from 2.5 to 10 cm. Table
tests for the elastic stiffness of rock surrounding an 8.4 presents values for apparent Youngs moduli
underground opening. at laboratory and engineering eld scales, and
the ratios of eld modulus to laboratory modulus
ratio for all of the rock types listed vary only from as selected from a more extensive data set
0.08 to 0.40. We develop the following rule of (Bieniawski, 1984).
thumb from these data: laboratory specimens of rock The rock masses tested in the eld usually are
have Poissons ratios that range from about 0.1 to 0.4 softer than laboratory samples of the same rock.
with a typical value of about 0.25. Only one site, the mica project, of the fteen given
in the original data set, contradicts this general-
8.5.3 Elastic properties of rock from ization and data from that site show nearly the
engineering field tests same Youngs modulus at the two scales. We con-
In light of the heterogeneous nature of rock clude that increasing the characteristic length for
masses as observed at laboratory, outcrop, and test specimens from centimeters to several meters
crustal scales, it seems natural to suspect that may have little or no effect on stiffness, but
physical properties, such as the elastic moduli, usually it will decrease the stiffness by as much as
would be length-scale dependent (Pinto da Cunha, two orders of magnitude. Because the conditions
1990). Rock engineers have developed methods to of the rock are likely to be quite different from site
estimate the in-situ properties of rock masses to site, similar rock types can behave very differ-
around underground openings and beneath large ently. For example crystalline rocks from the LG-2
engineering structures like dams, and these site and the Tehachapi Tunnel site have practi-
tests provide valuable insight about the scale cally identical laboratory moduli, yet their eld
Table 8.4. Laboratory and engineering field tests for Youngs modulus (GPa).
Project site Rock type E (plate) E (lab) E (plate)/E (lab)
LG-2 project Granite 50.0 80.0 0.62

Churchill Falls Gneiss 41.5 55.0 0.75
Dworshak Dam Granite/gneiss 23.5 51.7 0.45
Tumut 2 Gneiss/granite 6.9 59.1 0.12
Tehachapi Tunnel Diorite gneiss 4.8 77.9 0.06
Mica project Quartzite gneiss 27.6 27.0 1.04
Elandsberg Graywacke 39.6 73.4 0.54
Waldeck II Graywacke 5.0 20.0 0.25
York Canyon Shaley sandstone 0.65 43.4 0.015
8.6 ELASTIC HETEROGENEITY AND ANISOTROPY 323
moduli differ by an order of magnitude. The s2r

granite from LG-2 was described as massive, Surroundings: y
whereas the diorite gneiss from Tehachapi was Gs, vs
described as fractured. Apparently, the presence
of fractures (and perhaps other heterogeneities) at suu(R+, p/2)
s1r
the eld scale produced a profound change in the
rock stiffness. The three sedimentary rock types s2i suu(R+, 0)
s1i r
listed in Table 8.4 are consistent with the general-
ization that greater size correlates with lesser u x
stiffness. They also illustrate the broad range of
possible behaviors: for some graywackes the stiff- Inclusion:
ur G i , vi
ness decreased by only a factor of two, whereas for uu
some shaley sandstones the decrease was by a
factor of more than sixty. 2R
8.6 Elastic heterogeneity and

anisotropy Fig 8.28 The plain strain elastic problem of a cylindrical
inclusion with shear modulus and Poissons ratio, Gi and i, in
surroundings with Gs and s. Loading is by remote principal
A rock mass is heterogeneous with respect to an stresses, 1r and 2r ( Jaeger and Cook, 1979).
elastic property if the value of that property varies
from place to place. A rock mass is anisotropic with
respect to an elastic property if the value of that of minerals than the surrounding rock, and they
property varies with orientation at any given may have a different grain size and texture. As
place. Thus heterogeneity refers to spatial varia- such, the elastic properties of the inclusion are
tion and anisotropy refers to directional variation. likely to be different than those of the surround-
These terms apply to any physical property so it is ings. A very instructive boundary value problem is
important to specify that property. A rock mass the elastic solution for a circular inclusion (Jaeger
with no spatial variation in a particular property and Cook, 1979, pp. 261).
is said to be homogeneous with respect to that prop- The inclusion has a radius, R, and the sur-
erty, and one with no direction variation is said to rounding elastic material extends to an innite
be isotropic with respect to that property. In this distance from the origin of coordinates at the
section we describe heterogeneities and aniso- center of the inclusion (Fig. 8.28). Both a Cartesian
tropies with respect to elastic properties and show (x, y)- and a polar (r, )-coordinate system are
how these affect the solutions to certain elastic employed. The inclusion has a shear modulus, Gi,
boundary value problems. and Poissons ratio, i, and it is embedded in sur-
roundings with shear modulus, Gs, and Poissons
8.6.1 Deformation of a heterogeneous ratio, s. The following combinations of these con-
elastic material stants are used for the plane strain conditions
Some heterogeneities serve to amplify locally a specied here:
remote tension or compression, whereas others
i 3 4i,s 3 4s,k Gi Gs (8.92)
actually convert a remote compression into a local
tension or vice versa. To keep the analysis simple The interface between the inclusion and the sur-
and analytical we consider an inclusion of roundings is bonded, so the inclusion remains
foreign material such as a fossil, clast, or xeno- xed to the surroundings regardless of the
lith in an otherwise homogeneous rock mass applied loading. This boundary condition is
(Wiltschko and Sutton, 1982). Typically these specied by equating like components of the dis-
foreign objects are composed of a different suite placement vector on either side of the boundary:
BC: ur(r R, ) ur(r R, ), Remarkably, the principal stresses within the
(8.93) inclusion are simply proportional to their respec-
u(r R, ) u(r R, )
tive values in the remote eld and the propor-
It follows that the traction acting on the inclusion tionality constant is the same for each principal
is equal and opposite to the traction acting on the stress. The magnitude of the stress within the
surroundings at every point of the contact. This inclusion changes with the ratio of the shear
implies that certain components of the stress are moduli, k, but this quantity is always positive, so
continuous across the interface: the sign of the internal stress is always the same
as the sign of the respective remote stress. Thus, a
rr(r R, ) rr(r R, ), remote tension induces tension within the inclu-
(8.94) sion, and a remote compression induces compres-
r (r R, ) r (r R, )
sion within the inclusion.
We consider the case of biaxial remote stresses, 1r Recalling that k Gi/Gs, the above equations
and 2r , acting at innite distance in the x- and y- can be used to determine how the stress state
directions, respectively: changes within the inclusion as a function of this
ratio, again for the case where both Poissons
BC: at x , y : ratios are equal to 14 :
(8.95)
xx r1, xy 0 yx, yy r2
i1 i2 0,for Gi Gs
(8.99)
Because the shear stress components are zero in
the remote eld the normal components are prin- i1 32 r1 and i2 32 r2,for Gi Gs
cipal stresses. In general the stress in a softer inclusion is lesser
A remarkable and non-intuitive result found in magnitude than that in the surroundings, and
by solving this boundary value problem is that the the stress in a stiffer inclusion is greater. In the
state of stress in the inclusion is homogeneous. limit as the stiffness of the inclusion goes to zero
The principal stresses within the inclusion, 1i and the behavior changes to that of a circular hole
2i , are ( Jaeger and Cook, 1979, pp. 262): in elastic surroundings. The greatest stress that
can be induced within a stiffer inclusion is an
[k(s 2) i] k(s 1) r
ixx i1 increase over the remotely applied stress by a
2(2k i 1) (ks 1) 1
factor of 32.
[k(s 2) i 2] k (s 1) r Outside the inclusion the state of stress is het-
2 (8.96)
2(2k i 1) (ks 1) erogeneous but it changes with distance from the
[k(s 2) i 2] k(s 1) r inclusion to approach the remote state of stress.
iyy i2 1 The stress components in the surroundings are
2(2k i 1) (ks 1)
(Jaeger and Cook, 1979, p. 250):
[k(s 2) i] k (s 1) r

(8.97) 1 BR2
2(2k i 1) (ks 1) 2 rr ( r1 r2) 1 2
2 r
The homogeneity of stress and strain inside the
inclusion is true for all possible values of the
elastic moduli. The stress and strain also are
1

2AR2 3CR4
( r1 r2) 1 2 4 cos 2
2 r r
homogeneous within an elliptical inclusion, and (8.100)
within a three-dimensional ellipsoidal inclusion
(Muskhelishvili, 1954; Eshelby, 1957).
1

BR2
( r1 r2) 1 2
2 r
For the sake of easily interpretable results, con-
sider the special case where vi 0.25 vs, so i 2
s. The stress everywhere inside the inclusion is:
1
2 r
3CR4
( r1 r2) 1 4 cos 2
(8.101)
i1
3k
r , i2
2k 1 1
3k
r
2k 1 2
(8.98)
1
2 r
AR2 3CR4
r ( r1 r2) 1 2 4 sin 2
r (8.102)
The stress components are functions of the polar applied stress, and the stress is zero on the inter-
coordinates, r and , and the constants A, B, and C face that is parallel to the applied stress. In this
in these equations are related to the elastic con- sense the stiff inclusion creates a stress shadow in
stants as: the surrounding softer material. The stiffer inclu-
sion carries more of the applied load than adja-
2(1 k) 1 k(s 1)
A ;B i ; cent regions by up to a factor of 3/2. The adjacent
k s 1 2k i 1
regions carry less of the applied load by up to a
k1
C (8.103) factor of 1/2. These results provide a useful bench-
k s 1
mark for assessing the effect of stiffer or softer
Consider the two points just outside the inclu- rock masses on the state of stress.
sion, where the r R , and the polar angles are Distributions of the Cartesian stress compo-
0 and /2. There we calculate the circumfer- nents are illustrated in Fig. 8.29 for a uniaxial
ential stress using (8.101). Again, consider the case stress in the remote eld r1 1 and a softer inclu-
where the two Poissons ratios are equal to 14 , so sion with shear modulus Gi 10 GPa and Poissons
the relevant constants are A 2(1 k)(1 2k), ratio i 14 embedded in stiffer surroundings with
B 12 A, and C 12 A. Using these coordinates shear modulus Gs 30 GPa and Poissons ratio s
1
and constants we nd: 4 . Note that the stress components are uniform
within the inclusion and that both the y-compo-
(r R, 0) 1k

r
1 2k 1
3
r
1 2k 2 nent of normal stress and the shear stress are zero.
The x-component of normal stress within the
(8.104) softer inclusion is diminished in value relative to
the remote stress. The contour patterns outside the
(r R, 2) 3

r
1 2k 1
1k
r
1 2k 2 inclusion are quite complex, but they clearly illus-
trate the facts that perturbations due to the inclu-
(8.105) sion are symmetric, they decrease with distance,
and they are negligible at radial distances r 5R.
Because the coefcients that multiply the remote
stresses are just interchanged for the two points,
8.6.2 Generalized Hookes Law for
consider only a uniaxial stress, 1, acting along
anisotropic rocks and minerals
the x-axis:
The most general linear relationship among the
components of strain and stress is one in which

r1, for Gi Gs
(r R, 0) 0, for Gi Gs (8.106) each of the nine components of strain, ij, is lin-
1 r early related to the nine components of stress, ij.
2 1, for Gi Gs
For example, given a Cartesian coordinate system
with orthogonal axes x, y, and z, the equation for

3 r1, for Gi Gs

r R,

2 r1, for Gi Gs (8.107)
0, for Gi Gs
the normal strain component, exx, may be written
as (Nye, 1985, Chapter VIII):
xx s1111xx s1112xy s1113xz
If the inclusion is much softer than the sur-
roundings, Gi Gs, the circumferential stress s1121yx s1122yy s1123yz
component is similar to that for an open circular
s1131zx s1132zy s1133zz (8.108)
hole: on the interface that is perpendicular to the
applied stress the local stress is equal in magni- The sijkl are constants of proportionality and the
tude but opposite in sign, and there is a stress con- four subscripts on each s correspond to the two sub-
centration factor of 3 on the interface that is scripts of the strain component e, followed by the
parallel to the applied stress. two subscripts of the stress component , with sub-
If the inclusion is much stiffer than the sur- scripts (1, 2, 3) corresponding to (x, y, z), respec-
roundings, Gi Gs, there is a diminution factor of tively. There are eight more equations of this form,
1/2 on the interface that is perpendicular to the each one linearly relating a component of strain to
(a)4 xx
material is one in which relatively small stresses
1.6
cause relatively large strains. In other words, the
3 1.5 respective constants for a compliant material are
2 1..4 greater than those for a stiff material.
1..3 The units and dimensions of compliance are
1
1..2 the inverse of those for stress:
0 1..1
compliance, s[ ] N1m2 Pa1
1.0 (8.109)
-1 compliance, s{ } M 1L T 2
0..9
-2 The symmetry of the innitesimal strain com-
0..8
-3
ponents and of the stress components enables a
0..7
reduction in the number of independent compli-
0..6
-4
-4 -2 0 2 4 ances to 36 according to:
(b) yy
4 sijkl sjikl and sijkl sijlk (8.110)
0.15
3 0.10 Therefore, the linear anisotropic elastic material
2
0.05 is one in which each of the six independent com-
0.00 ponents of strain is linearly related to the six inde-
1
-0.05 pendent components of stress (Lekhnitskii, 1963;
0 -0.10 Jaeger and Cook, 1979):
-0.15 xx s11xx s12yy s13zz s14yz
-1
-0.20
-2
s15zx s16xy (8.111)
-0.25
-3 -0.30 yy s21xx s22yy s23zz s24yz
-0.35
-4 s25zx s26xy (8.112)
-4 -2 0 2 4
(c) xy
4
zz s31xx s32yy s33zz s34yz
3 0..2 s35zx s36xy (8.113)
2
0..1
yz s41xx s42yy s43zz s44yz
1
0.0 s45zx s46xy (8.114)
0
-1 -0.1 zx s51xx s52yy s53zz s54yz

s55zx s56xy (8.115)
-2 -0.2
-3 xy s61xx s62yy s63zz s64yz

-0.3
-4 s65zx s66xy (8.116)
-4 -2 0 2 4
Fig 8.29 Contour maps of stress components near The sij form a matrix of constants of proportionality
cylindrical inclusion. (a) Normal stress, xx. (b) Normal which also are referred to as compliances. Here the
stress, yy. (c) Shear stress, xy. subscript notation for the compliances has been
simplied, based on the order of listing the strain
the nine components of stress, and each one and stress components. The rst subscript corre-
having nine proportionality constants. Altogether sponds to the rank of the strain component in the
there are 81 constants that are referred to as com- column on the left-hand side. The second subscript
pliances. These constants make up a fourth-rank corresponds to the rank of the stress component in
tensor quantity (Nye, 1985, p. 133). A compliant each row on the right-hand side. For isothermal and
reversible deformation, considerations of the common minerals is less than 21 because of their
elastic strain energy (Nye, 1985, p. 136) prove that symmetry. There are seven systems of crystal sym-
this matrix of compliances is symmetric: metry (triclinic, monoclinic, rhombic, tetragonal,
trigonal, hexagonal, or cubic) and specic symme-
sij sji (8.117)
try classes within each system (Lekhnitskii, 1963,
This symmetry further reduces the number of pp. 2632). Triclinic crystals have 21 different com-
independent compliances to 21, so the matrix of pliances, but as the symmetry increases the
compliances can be represented in the following number of compliances, as referred to crystallo-
manner: graphic reference axes, decreases. For example,
some trigonal crystals have only six independent

s11 s12 s13 s14 s15 s16 compliances (Nye, 1985, table 9):
s22 s23 s24 s25 s26
s11 s22, s12, s44 s55, s33, s13 s23,
s33 s34 s35 s36
(8.118) s14 s24 2s56, s66 2(s11 s12) (8.120)
s44 s45 s46
s55 s56 All hexagonal crystals have ve independent
s66 compliances:
It is understood that the terms below the diagonal s11 s22, s12, s44 s55, s33,
are equivalent to their counterparts above the s13 s23, s66 2(s11 s12) (8.121)
diagonal. Cubic crystals have three independent compli-
It is intuitive that solids will extend in the ances:
direction of an applied tensile stress. Further- s11 s22 s33, s12 s13 s23, s44 s55 s66
more, we have experience with solids (e.g. a (8.122)
rubber band) that contract in the directions per-
pendicular to an applied tension. Earlier in this The compliances not listed in these relationships
chapter we dened Poissons ratio as the elastic are identically zero.
property of an isotropic material that related such Measured values for the compliances of a few
lateral contractions to the extension in the direc- well-known minerals from the trigonal, hexag-
tion of the applied tensile stress. However, the onal, and cubic systems are given in Table 8.5
general anisotropic elastic material exhibits (Birch, 1966; Nye, 1985). The units are 1011N1/m2
behavior that is not so intuitive. For example, if a (1011 Pa1) and all refer to room temperature
uniaxial tensile stress is applied along the z-axis, unless otherwise noted.
the only stress component is zz, yet all six inde- Recall that the range of Youngs moduli for
pendent strain components would be non-zero: rock is from about 109 to 1011 Pa, so the inverse
Youngs moduli would range from 1011 to 109
xx s13zz,yy s23zz,zz s33zz, Pa1, roughly the same range as that of the values
(8.119)
yz s43zz,zx s53zz,xy s63zz of compliance in this table. Note that the miner-
als from each crystal system span a similar range
It is expected that a rectangular block of this
of compliances with the exception of ice, which is
material would extend parallel to the tension and
extraordinarily compliant. Also, minerals of the
contract perpendicular to this tension. However,
same composition, but different systems, such
there are shear strains induced by the tensile
as - and -quartz, have different compliances.
stress such that the block would have non-rectan-
Usually, greater temperatures correspond to
gular sides in the loaded state.
greater compliances, as in the case of halite.
Finally, minerals known for their hardness,
8.6.3 Compliances for anisotropic such as diamond, have very low compliances.
minerals and rocks Because rock is made up of many different
Minerals are natural examples of crystalline solids minerals, and these minerals are anisotropic with
that are anisotropic with respect to elastic proper- respect to elastic properties, it is natural to
ties; however, the number of compliances for many suppose that rock would be anisotropic. However,
Table 8.5. Compliances of a few minerals (1011Pa1).
Crystal System s11 s12 s44 s33 s13 s14
Calcite trigonal 1.13 0.37 4.03 1.75 0.43 0.91

-Quartz trigonal 1.28 0.15 2.00 0.96 0.11 0.45
Hematite trigonal 0.44 0.10 1.19 0.44 0.02 0.08
Ice (10C) hexagonal 10.24 4.22 33.00 8.37 1.90
-Quartz (600C) hexagonal 0.94 0.06 2.77 1.06 0.26
Apatite hexagonal 0.75 0.10 1.51 1.09 0.40
Halite (900 K) cubic 4.79 1.34 9.47
Halite (600 K) cubic 3.19 0.76 8.49
Halite (300 K) cubic 2.28 0.45 7.81
Galena cubic 1.20 0.30 4.00
Diamond cubic 0.14 0.04 0.23
in many examples the individual mineral grains that there are ve independent compliances for
are arranged with random orientations. Taking a rocks with this form of symmetry and that the
sample of such a rock that is large compared to compliances have the same relationship to one
the grain size, the elastic properties would be another as in the hexagonal crystal class. Rocks
approximately isotropic. Calculation procedures with these elastic properties are referred to being
have been derived to estimate the average transversally isotropic.
isotropic elastic constants for a rock from knowl- The strainstress relations for the transversally
edge of the elastic properties and abundances of isotropic rock can be written using variants of the
each mineral (Hearmon, 1961). more familiar elastic constants, Youngs modulus
Some rocks at the hand-sample scale are and Poissons ratio as follows:
anisotropic with respect to elastic properties
1
because the constituent grains are not randomly xx (xx yy) zz (8.127)
E E
oriented. For example, slates have a strong pre-
ferred orientation of the platy minerals. Such 1
yy (yy xx) zz (8.128)
rocks are likely to have an axis of elastic symme- E E
try perpendicular to the plane of the mineral
1
fabric. Within the plane of the fabric the rock may zz ( yy) (8.129)
E zz E xx
be approximately isotropic. Such a material is
described as having a plane of isotropy, or an axis 1 1 2(1 )
yz , , xy (8.130)
(perpendicular to this plane) of rotational sym- G yz zx G zx xy E
metry (Lekhnitskii, 1963, p. 24) and the linear
Here E and are Youngs modulus and Poissons
strainstress equations reduce to:
ratio for any direction of applied normal stress
within the plane of isotropy. The constants E and
xx s11xx s12yy s13zz (8.123)
are Youngs modulus and Poissons ratio for an
yy s12xx s11yy s13zz (8.124) applied normal stress along the axis of symmetry.
The constant G is the shear modulus for an
zz s13xx s13yy s33zz (8.125)
applied shear stress in any plane that contains the
yz s44yz, zx s44zx, xy 2(s11 s12)xy axis of symmetry.
For plane strain deformation (8.30) with uz 0
(8.126)
consider the possibility that there are two
Here the z-axis is the axis of rotational symmetry orthogonal axes of elastic symmetry in the (x, y)-
and the (x, y)-plane is the plane of isotropy. Notice plane. These conditions require four compliances
in contrast to the two elastic moduli for the Table 8.6. Youngs modulus of a few rocks in
isotropic material (8.48)(8.50) and dene the so- orthogonal directions (GPa).
called orthotropic solid:
Rock type Perpendicular Parallel Parallel
xx s11xx s12yy, yy s12xx s22yy, (A) (B)
xy s66xy (8.131)
Gneiss 18.6 23.1 12.4
In terms of Youngs moduli and Poissons ratios, Marble 49.3 62.7 71.7
these equations are written: Granite 30.4 27.4 44.2
1 1 Limestone 33.4 41.0 37.2
xx 21 , 12 xx yy, Sandstone 6.0 6.7 8.8
E1 xx E2 yy yy E1 E2
1
Sandstone 7.1 10.6 11.2
xy xy (8.132) Oil shale 12.4 21.4
G
Oil shale 21.1 33.2
What appear to be ve constants are only four
because the two Poissons ratios are related:
21 12 E circular hole (Fig. 8.30) in an orthotropic solid
, so 12 21 1 (8.133) loaded by a remote uniaxial stress ( Jaeger and
E2 E1 E2
Cook, 1979, pp. 2979). This is a two-dimensional,
E1 and 12 are Youngs modulus and Poissons ratio plane strain solution, so the hole represents a
for an applied normal stress along the x-axis, long cylindrical opening perpendicular to the
whereas E2 and 21 are Youngs modulus and (x, y)-plane. The in-plane stress components are
Poissons ratio for an applied normal stress along related to the Airy stress function by:
the y-axis. G is the shear modulus in the (x, y)-
plane. 2 2 2
xx 2
,yy 2 ,xy (8.136)
For isotropic materials, the number of inde- y x xy
pendent compliances reduces to two because:
Substituting these equations into the constitutive
s11 s22 s33, s12 s13 s23, equations for the orthotropic elastic material
s44 s55 s66 2(s11 s12) (8.134) (8.131):
How these compliances relate to the isotropic 2 2 2 2
elastic constants introduced earlier can be xx s11 2
s12 2 ,yy s12 2 s22 2 ,
y x y x
deduced from the strainstress relationships 2
written using the compliances and (8.134), and xy s66 (8.137)
xy
comparing these to (8.18), (Nye, 1985, p. 143):
Substituting the constitutive equations into the
1 1 compatibility equation written in terms of the
s11 , s12 , 2(s11 s12) (8.135)
E E G strain components, we have:
Data on anisotropic elastic properties for rock 4 4 4
samples are presented in Table 8.6 for measure- s22 (s66 2s12) 2 2 s11 4 0 (8.138)
x4 x y y
ments of the modulus of elasticity (Youngs
modulus) perpendicular to bedding or foliation (if Dividing through by s22 we dene the following
any) and in two orthogonal directions parallel to constants:
bedding or foliation (Obert and Duvall, 1967, p. 486). s11 s 2s12
12 C1, 66 1 2 C2
s22 s22
8.6.4 Deformation of an anisotropic
(8.139)
elastic body
To assess the importance of elastic anisotropy The compatibility equation can be rearranged as
during deformation of a rock mass we consider a follows:
For the purpose of illustrating the effects of

y
r elastic anisotropy on the state of stress we con-
s1 sider the problem of the circular hole and select
suu(R+, p/2) the following values for the elastic constants: E1
40 GPa, E2 20 GPa, G 10 GPa, 12 0.2, 21 0.1.
r suu(R+, 0) This would be considered a very anisotropic rock,
x judging from Table 8.6: the ratio E1/E2 2 is
u
greater than any of the examples given there. Only
four constants are independent and these ve con-
stants are self-consistent according to (8.133).
2R Using (8.144), we nd C1 0.5 and C2 1.8, and
solving the quadratic equation (8.143) we have 1
Orthotropic elastic: (1.457, 0.343). Using the rst of (8.141), the
E1, N12, E2, N21, G second constant is 2 (0.343, 1.457). Note that
there is only one pair of independent constants
Fig 8.30 The plain strain elastic problem of a cylindrical given by the solution to the quadratic equation.
hole in an orthotropic elastic material. Loading is by remote This follows from (8.140) where one can see that
uniaxial principal stress, 1r (Jaeger and Cook, 1979). the constants 1 and 2 are interchangeable.
Stress states for the elastic boundary value

2 2 2 2 problem are given in terms of yet more constants
2
1 2 2
2 2 0 (8.140) that are related to the constants 1 and 2 as
x y x y
follows:
Note that the sum and the product of the con-
stants 1 and 2 are related to the compliances 1 1 1
(8.139), so one must solve these two equations 1 0.094,2 2 0.261
1 1 2 1
simultaneously to nd these relations. Multi-
plying the second of (8.139) by 1 we have: (8.145)
12 C1, 21 12 C21 (8.141) Now consider the circular hole in an innite
anisotropic body with uniaxial stress, 1, at an
Subtracting the second equation from the rst innite distance acting in the Ox-direction (Fig.
equation we nd the following quadratic equa- 8.30). The x- and y-axes are symmetry axes with
tion: respect to the anisotropy. At the edge of the hole,
the circumferential normal stress is (Jaeger and
21 C21 C1 0 (8.142)
Cook, 1979, pp. 298):
The two solutions for this quadratic equation are:
(1 1)(1 2)(1 1 2 12 2 cos 2)

1 1 (1 21 21 cos 2)(1 22 22 cos 2)
2 C2 2 C 2 4C1,
1
1 2 where C 22 4C1 0
(8.143) (8.146)
The restriction on the values of C1 and C2 follows

The stress distribution as a function of position, ,
from the fact that the compliances are real
around half the hole is illustrated in Fig. 8.31
numbers, so we are looking for solutions that are
where triangles mark the curve for the isotropic
real values of 1, given real values of C1 and C2.
material, diamonds mark the curve for loading
These constants are related to the familiar elastic
parallel to the direction of greater Youngs
constants as follows:
modulus, E1, and squares mark the curve for
s11 E2 E loading parallel to the direction of lesser Youngs
C1 ,C2 2 221 (8.144)
s22 E1 G modulus, E2.
Orthotropic Plate: Uniaxial Tension 350

4 b
Load // E1
Isotropic 300 Confining
Breaking stress (MPa)

3
Load // E2 pressure
250 35.0 MPa
2
10.5 MPa
/1r
200 3.5 MPa

1
150

0
20 40 60 80 100 120 140 160
100
-1
50
-2
Position Theta (degree) 0
0o 15o 30o 45o 60o 75o 90o
Fig 8.31 Plot of circumferential stress component, , at
Inclination, b, of cleavage to specimen axis ()
edge of cylindrical hole in orthotropic elastic material with
uniaxial remote stress, 1r. Three different cases are:
isotropic (triangles), loading parallel to E1 (diamonds), and Fig 8.32 Plot of strength versus inclination of cleavage for
loading parallel to E2 (squares). triaxial tests of Martinsburg Shale for three different
confining plessures: , 3.5 Mpa; , 10.5 Mpa; , 35 Mpa.
Reprinted from Donath (1961) with permission of The
The normalized circumferential stress (Fig. Geological Society of America.
8.31) for the isotropic material varies from 1 at
0 to 3 at /2. In comparison, the normal- of rocks are much more important in failure
ized circumferential stress for the anisotropic under inhomogeneous stresses than effects of
material dened above, and loaded parallel to the anisotropy of elasticity (Jaeger and Cook, 1979,
direction of greater Youngs modulus, varies from pp. 299). For example, data on the anisotropy of
0.737 at 0 to 3.475 at /2. Thus, the breaking strength under triaxial compression for
local stress acting perpendicular to the remotely samples of Martinsburg Shale are shown in Fig.
applied stress is diminished somewhat less than 8.32 where values range over one order of magni-
that for the isotropic material, and the local stress tude depending upon the orientation of the cleav-
acting parallel to the applied stress is concen- age (Donath, 1961). The greatest strength is
trated somewhat more. If the remote loading is measured when the cleavage is perpendicular to
parallel to the lesser Youngs modulus of this the compression and the least when the cleavage
anisotropic material, the circumferential stress is at an angle of about 30o to the compression.
varies from 1.443 at 0 to 2.722 at /2.
Thus, the local stress is diminished somewhat
more than that for the isotropic material, and
concentrated somewhat less. For the degree of
anisotropy used here, the differences in stress con- Since the earliest investigations of geologic struc-
centration and diminution relative to the tures, geologists have used qualitative observa-
isotropic material are modest. Using the isotropic tions and everyday words to interpret how rocks
solution to model this anisotropic rock does not behave as structures evolve. For example, two com-
change the qualitative nature of the stress distri- monplace words, competent and incompetent, have
bution and the sign changes are preserved from been used to explain the style and relative magni-
one side of the hole to the other. tude of deformation. Being colloquial expres-
In assessing the effect of elastic anisotropy sions, these words have been used quite freely to
Jaeger and Cook conclude that the ratio E1/E2 2 describe the mechanical behaviors of rock: com-
as used in Fig. 8.31 is rather extreme, so it is prob- petent implying less easily deformed and incom-
ably true that the effects of anistropy of strength petent more readily deformed under a given set of
forces. While these words are, perhaps, appropri- stress and strain. To understand the tectonic
ate for ordinary conversations about structures, processes that lead to the development of geolog-
they lack the formal denition that might make ical structures one must choose an appropriate
them useful in a quantitative analysis. In this boundary or initial value problem to serve as a
chapter we have introduced the concept of elas- mechanical model. To formulate such a problem
ticity and the elastic properties of rock. These one must postulate a particular mechanical
properties are useful in quantitative investiga- behavior. Here we have chosen Hookes Law for the
tions of geologic structures because they provide linear elastic solid; in Chapter 10 we choose
precise measures of the relationship between Newtons relation for the linear viscous uid.
Chapter 9
Brittle behavior
Jointed limestone bed at Lilstock Beach on the southern of direction, truly wonderful . . . The observer becomes
coast of the Bristol Channel, England. Photograph by D. D. so absorbed with the scene that he unconsciously
Pollard. begins to feel as if the rocks under and around him
were in process of being illimitably cleft from north to
An admirer of Nature may be excused becoming enrap- southas if the earths crust were in course of splitting
tured when he takes a view from any of these noble up from one pole to the other; and he only rids himself
terraces [in County Clare, Ireland]. Looking north, or of the feeling to become bewildered with the question,
south, his eyes are riveted on vast surfaces of gray lime- as to what mysterious agent produced the singular
stone rocks, split up to an extent, and with a regularity phenomenon he is contemplating (King, 1875).
334 BRITTLE BEHAVIOR
I 9.1 Brittle deformation in the

n the preceding chapter we learned that the
mechanical behavior of rock under certain con-
ditions can be approximated with a linear
laboratory and in the field
elastic material, a mathematical construct formu-
lated using Hookes Law to relate stress and Laboratory experiments on rock samples loaded
innitesimal strain. The elastic material is useful beyond their elastic limit under a variety of con-
for describing both ancient and modern deforma- ditions help us to understand what happens with
tion in the Earth at a variety of length and time the onset of inelastic deformation in the Earth
scales. However, the limestone described by King (Griggs and Handin, 1960b; Paterson, 1978; Carter
as illimitably cleft from north to south provides et al., 1981; Wong, 1982a,b; Reches, 1983; Reches
an evocative example of fracturing which is inelas- and Dieterich, 1983; Duba et al., 1990). Griggs and
tic, non-recoverable deformation. Even if the frac- Handin (1960a) provide a useful summary in their
ture surfaces were pushed back together they illustration (Fig. 9.1) based upon decades of experi-
would not heal. Similarly, the limestone bed pic- ence testing rock in the laboratory under condi-
tured in the frontispiece for this chapter from tions of triaxial compression and extension. In
Lilstock Beach on the southern margin of the these experiments cylindrical specimens are
Bristol Channel, England, is broken by numerous subject to a uniform compressive radial stress
fractures, providing visual evidence of inelastic called the conning pressure and an axial compres-
behavior (Rawnsley et al., 1998; Engelder and sive stress. The experiments are referred to as
Peacock, 2001). In this chapter we contemplate the extension tests (Fig. 9.1a) if the axial stress is the
singular phenomenon that bewildered King and least compressive stress (designated 1 according
describe the modern concepts, laboratory data, to the sign conventions adopted here) or compres-
and fracture mechanics required to address his sion tests (Fig. 9.1b) if the axial stress is the greatest
question. compressive stress (designated 3).
A few fracture tips are visible on the bedding Rock specimens in extension tests (Fig. 9.1a) typ-
surface of the Lilstock exposure (Chapter 9, front- ically fail by the formation of an extension fracture,
ispiece). We infer that the stress state reached the oriented perpendicular to the least compressive
rock strength near the tips of the fractures as they stress, at axial strains less than about 1%. The rela-
propagated through the limestone, so the local tive motion of the surfaces of extension fractures is
stressstrain behavior was non-linear and irrecov- dominantly opening. In compression tests at low
erable there. Presumably the tips now visible rep- conning pressures (Fig. 9.1b), splitting fractures
resent those points where the local stress form parallel to the greatest compressive stress at
dropped below the rock strength and propagation strains from about 1 to 5% (Fig. 9.2a). Splitting frac-
ceased. We also infer, based on experiments and tures also open although wedge-shaped fractures
theory to be described in this chapter, that the with dominantly shearing motion may develop
stress state remained within the elastic range in near the ends of specimens. At modest conning
much of the rock between the fractures. This pressures (Fig. 9.1c), shear fractures form at an acute
limited extent of inelastic deformation is charac- angle to the greatest compressive stress at strains
teristic of materials that have deformed in a brittle from about 2 to 8% (Fig. 9.2b). At greater conning
state. We seek to understand the consequences of pressures (Fig. 9.1d), the deformation is distributed
brittle behavior for the state of stress and strain, across a shear zone at strains from about 5 to 10%
the nature of the physical mechanisms that (Fig. 9.2c). At high conning pressure (Fig. 9.1e),
operate during such deformation, and the shearing is pervasive throughout the specimen,
reasons why inelastic deformation may be local- which deforms in what is called the ductile state and
ized into thin tabular zones. maintains its integrity to strains greater than 10%.
Distributed ow is a characteristic feature of
ductile deformation, a topic that we take up in a
later chapter. Brittle deformation, in contrast, is
localized into discrete fractures or tabular zones.
9.1 BRITTLE DEFORMATION IN THE LABORATORY AND FIELD 335
Extension test Compression test, confining pressure increasing
(a) (b) s3 (c) (d) (e)

s1
s2 = s3 s1 = s2
Extension Splitting Shear Shear Distributed

fracture fracture fracture zone shearing
Typical axial strain
15% 28% 510% >10%
at fracture = <1%
Fig 9.1 Schematic representation of brittle failure styles in formation of faults or joints. On the other hand
triaxial tests. (a) Extension test. (b)(e) Compression test laboratory tests may produce useful values of
with confining pressure increasing to the right. Reprinted material properties such as elastic stiffness, fric-
from Griggs and Handin (1960a) with permission of The
tional strength, or fracture toughness over a range
Geological Society of America.
of conning pressures, temperatures, and strain
rates. Although laboratory triaxial experiments
Jaeger and Cook comment on the naming of may provide few insights about the processes of
inclined fractures in compression test specimens faulting or jointing, knowledge of the material
(Fig. 9.1c and 9.2b) as follows: properties obtained from them is vital for the
Griggs and Handin (1960) . . . call it [the inclined frac- development of models of faulting and jointing.
ture] a fault because of its correspondence with geo- Ultimately the laboratory specimens depicted
logical faulting, and they have been followed by many in the rst three columns of Fig. 9.1 break apart
writers; however, it seems preferable to conne the into two or more fragments because extension or
term fault to the geological context and to retain the shear fractures propagate across the rock cylin-
term shear fracture in the experimental context
ders, which thereby loose their ability to support
(Jaeger and Cook, 1979, pp. 92).
the applied compressive stresses. On the other
Here we choose to follow the recommendation of hand, joints and faults are contained within
Jaeger and Cook so the two phenomena, shear Earths crust, so any behavior of laboratory speci-
fractures in laboratory specimens and faults in mens that is dependent upon the cylindrical geom-
Earths crust, are clearly distinguished. Further- etry or the properties of the testing apparatus in
more, we extend this recommendation to other contact with the specimen is unlikely to be dupli-
styles of localized deformation. For example, cated in nature. For example, consider the joints
extension fracture and splitting fracture are terms and faults in the inter-layered limestone and shale
appropriate for laboratory fractures that open, beds of the Blue Lias Formation at Lilstock Beach.
whereas we conne the term joint to the geologi- These are exposed in cross section (Fig. 9.3) along
cal context. What may seem like a semantic detail the steep wave-cut cliff just inland from the expo-
is important because it emphasizes the fact that sure (Chapter 9, frontispiece) that reveals the joint
laboratory tests rarely duplicate the length and traces on a single limestone bedding surface.
time scales, the materials, or the conditions of In this cliff exposure vertical joints are seen to
(a) (b) (c)
Fig 9.2 Examples of brittle failure in triaxial test specimens Laboratory experiments other than the stan-
of Ohtawa basalt. (a) 0.1 MPa confining pressure. (b) 49 MPa dard triaxial test are motivated by eld observa-
confining pressure. (c) 98 MPa confining pressure. Reprinted
tions of joints and faults. For example, exposures
from Hoshino et al. (1972) with permission of The
that reveal the tips of joints (Chapter 9, fronti-
Geological Survey of Japan.
spiece) raise the question: what is the resistance
to propagation of a single opening fracture in
fracture the limestone beds but terminate at or rock? To understand how the concentrated tensile
near interfaces with adjacent shale beds. Clearly a stresses along the advancing fracture tip are
given limestone bed did not break along a single resisted by rock strength one may conduct an
fracture: multiple parallel joints formed with a experiment (Fig. 9.4a) where control of the oppo-
regular spacing and ultimately several differently sitely directed displacement of the two pins
oriented sets of joints developed. Two faults crop enables one to monitor fracture propagation and
out in this exposure that cut across and locally measure the so-called fracture toughness of the rock
disrupt the otherwise nearly horizontal bedding. specimen (Atkinson and Meredith, 1987). Speci-
Unlike a single open joint, each fault is composed mens made up of layers of the same or different
of multiple closely spaced surfaces and each materials enable the researcher to investigate
surface is in intimate contact with another opening fractures that stop or propagate across
surface. Members of this set of faults do not termi- layer interfaces (Renshaw and Pollard, 1995;
nate within the exposure so their extent is McConaughy and Engelder, 2001). Exposures of
unknown, but detailed knowledge of the stratigra- faults such as those at Lilstock Beach (Fig. 9.3)
phy has been used to estimate a few meters to a few raise the question: what is the resistance to
tens of meters of slip (Engelder and Peacock, 2001). sliding of one specimen in direct contact with
The juxtaposition of limestone and shale beds, the another of the same or dissimilar rock type?
magnitude of the slip, the rotation of the fractured Laboratory studies (Fig. 9.4b) of frictional strength
limestone blocks, and the owage of the shale are help to address this question. While fracture
some of the characteristics of these faults that are toughness or friction experiments focus attention
not duplicated in laboratory experiments. on mechanical aspects of opening fractures or
9.2 STRENGTH OF LABORATORY SAMPLES 337
Fig 9.3 Joints and faults in inter-layered limestone and (a)

shale of the Blue Lias Formation at Lilstock Beach, Bristol
Channel, England. Photograph by D. D. Pollard.
sliding surfaces that may be more relevant to the

processes of jointing or faulting in the Earth than
the triaxial experiment, the relevant material
heterogeneities (sedimentary bedding, lithologic
contacts, other joints or faults, etc.) that may play
crucial roles in these processes typically are not
included at laboratory scales. The challenge for W
the structural geologist is to understand these (b)
complex processes when direct laboratory simu-
lations are not possible.
F
9.2 Strength of laboratory samples
Rock samples tested in the laboratory under con-

ditions that promote brittle behavior cannot be
loaded to arbitrary values of stress, but are
limited to values less than the strength of the
sample. Because the strength of certain engineer-
Fig 9.4 Alternative testing configurations to the triaxial
ing materials, like steel, is greater than that of
test. (a) Crack propagation test. (b) Frictional sliding test.
many rocks, it is possible to construct a laboratory
sa Here f is the force recorded by the load cell, A is the

Tu measured cross-sectional area of the sample, Li is
the measured length of the undeformed sample,
ep ep O and L Li is the displacement of the platen recor-
ea
ded by the displacement transducer. Alternatively,
the strain may be measured by electrical resistance
A strain gages attached directly to the sample (Fig.
8.25). The axial stress should be adjusted for
D changes in the cross-sectional area, and non-uni-
formities in the state of stress caused by shear trac-
tions imparted to the sample by differences in
B lateral extension of the platen and the sample
should be considered in the design of the experi-
Cu ment (Peng, 1971). Furthermore, axial strain is not
C a component of the strain tensor as dened in
Chapter 5, but is a measure of deformation com-
Fig 9.5 Generic axial stress versus axial extension curves monly called the extension or the elongation. We
for uniaxial tension and compression tests. Uniaxial tensile
refer to this quantity as the axial strain since that
strength, Tu, and compressive strength, Cu.
is the convention used in rock testing laboratories.
The generic stressstrain curve (Fig. 9.5) illus-
testing machine that can load rock samples to trates several important concepts and facilitates
their limiting values of stress. Furthermore, the the denition of key terminology. For samples
load-bearing members of the machine can be loaded in compression, the segment OA repre-
designed with greater cross-sectional area than sents non-linear elastic adjustments (e.g. closing
the sample, so the net force induces a lesser stress of cracks) and is followed by a nearly linear elastic
in the machine than in the sample. An example segment AB with a slope that would be Youngs
of such a machine is shown in Fig. 8.25. The outer modulus if the behavior were strictly linear. The
steel frame supports the load transmitted segment BC is concave upward because inelastic
through the sample from a mobile platen to a deformation mechanisms within the sample act
xed platen. A hydraulic pump provides pressure to decrease the stiffness. These mechanisms are
to move a piston that is attached to the mobile irreversible, so unloading on the dashed path
platen. A computer controls the ow of the results in a permanent axial strain ep. Upon reload-
hydraulic uid (and thereby the motion of this ing of the deformed sample, the stressstrain plot
platen) based upon electronic readings from a returns approximately to the original curve. The
displacement transducer and load cell that rock is said to be in a ductile state because it is accu-
monitor the deformation and loading of the mulating permanent strain while subject to an
sample. increasing magnitude of compressive stress. The
Data from such tests on cylindrical samples of slope of the curve continues to decline as the
rock usually are recorded on graphs of axial stress, sample shortens until the minimum stress (great-
a, plotted versus axial strain, ea (Fig. 9.5), and the est compression) is reached at point C. For uniax-
so-called stressstrain curve is used to understand ial tests, the stress at this point is the negative of
and characterize the behavior of the sample the uniaxial compressive strength, Cu. The generic
(Jaeger and Cook, 1979). Note that tension and tensile test has a similar stressstrain curve (Fig.
extension are taken as positive for this stress 9.5), here plotted in the rst quadrant because
strain graph. The axial stress and axial strain are cal- both stress and strain are positive quantities. The
culated as follows: tensile stress is limited by the uniaxial tensile
strength, Tu.
f L Li The point C on Fig. 9.5 represents the begin-
a ,ea (9.1)
A Li ning of the process of failure. The segment CD has
a negative slope and the inelastic mechanisms the stressstrain curve led researchers to try to
operating there are responsible for a degradation capture the complete stressstrain curve and to under-
of the load-carrying capacity as the sample contin- stand what inuence the testing machine might
ues to shorten and the failure process develops. have on the specimen behavior during failure
The sample is said to be in a brittle state because it is (Hudson et al., 1972; Jaeger and Cook, 1979, p. 177).
accumulating permanent strain as the magnitude Consider the schematic illustration of a
of the compressive stress decreases. Unloading testing machine (Fig. 8.25) and note that it is com-
during this brittle deformation along the dashed posed of two basic components: an hydraulic actua-
path results in a larger permanent strain, ep, and tor (including piston and cylinder, rod, and platen)
reloading returns approximately to the original for applying a force to the specimen and a reaction
curve. Tests typically end at a strain limited by the frame (tie-bars and cross-heads) for supporting this
apparatus (e.g. point D) or complete failure of the force. When uid pressure is increased in the
sample. actuator the piston is driven downward applying
a compression to the specimen and shortening it.
9.2.1 Soft and stiff testing machines The downward directed force is transmitted
Given a well-designed testing machine, the testing through the specimen to the lower cross-head.
procedure to determine strength is straightfor- The actuator also applies an equivalent upward
ward: load the sample until it starts to lose its directed force to the upper cross-head so the tie-
capacity to carry the load, and record the extreme bars are placed in tension and elongate, while the
value of stress. However, to understand the pro- cross-head bends. Both the actuator and the reac-
cess of failure one must understand the mechani- tion frame deform and are capable of storing
cal interplay between the sample and the testing elastic energy. If the mechanical parts of the
machine. It turns out that the machines can play testing machine combine to be very much stiffer
a strong, and even dominant, role in the outcome than the rock specimen, the machine would
of mechanical property tests and experiments. deform very little and we could ignore its role in
Indeed, the particular action of these machines the test. However, specimens are explosively dis-
has strongly inuenced our understanding of how integrated in some testing machines because
rock specimens behave when they are fracturing, the energy stored in the machine is released into
so we need to understand this action and ask if the specimen even though no uid is pumped
it might duplicate the behavior of the Earth into the actuator to do additional work. In other
during natural deformation events. This action, as words, the system composed of the machine and
we learn in this section, depends upon whether the specimen self-destructs.
the machine is soft or stiff relative to the rock To understand this process the initial defor-
sample. mation of the specimen is idealized (Fig. 9.6a)
For many discussions of extension and shear with a spring of constant stiffness, Cs. The
fracturing in the literature of rock mechanics the machine frame is idealized with two springs of
stressstrain curve after failure initiates is not stiffness Cm/2 in parallel and tied together with an
shown because the testing machine was incapable upper and lower rigid bar. All of these springs are
of tracking these quantities accurately during the tied to a rigid basal support which is taken as the
rapid failure event. Shortly after reaching the peak reference frame for measuring displacements.
stress (Fig. 9.5, point C), the data stream was ter- The vertical y-axis is positive downward so forces
minated because the specimen disintegrated into and displacement acting downward are positive.
a pile of rock chips and dust with a loud bang. As Two applied forces idealize the actuator: Fs acting
early as 1943 it was clear that explosive disinte- downward on the specimen spring and Fm acting
gration was not necessarily a natural behavior, but upward on the rigid bar joining the two machine
rather elastic energy is stored in the cylinder and springs (Fig. 9.6b). These forces are identical in
in the machine . . . and the release of this energy magnitude and opposite in direction, so Fs Fm.
causes the breakdown of the cylinder (Whitney, Under the applied force, Fs, the specimen spring
1943). Curiosity about the post-peak-stress part of contracts and its upper end moves toward the
(a) (b) (a)

um
Fm = P Fs
Cm/2 Cm/2 Fs = P Slope
us = Cs Slope = f (us)
Cs
P
Fs
x
F
F 2
(c) (d) y
1
u
us
u u
u = (b) (c)
Fm Fm
u u
um um
Fig 9.6 Idealized specimen and testing machine used to 3 4
Slope
investigate stability of tests in soft and stiff testing machines. Fm =Cm
P Fm P
Slope
lower support with a displacement, us (Fig. 9.6b). = Cm
On a force versus displacement graph (Fig. 9.7a)
the linear relationship, Fs Csus, plots in the rst
Fig 9.7 Force versus displacement plots for idealized
quadrant because both quantities are positive.
specimen and machine of Figure 9.6. (a) Specimen. (b) Stiff
Meanwhile, under the action of the force, Fm, the
testing machine. (c) Soft testing machine.
machine springs extend and their upper ends
move away from the lower support with a dis-
placement, um (Fig. 9.6b). The linear relationship, The work done in this one-dimensional system
Fm Cmum, for the combined machine springs by a constant applied force on an object is dened
plots in the third quadrant (Fig. 9.7b, c) because as the product of the force and displacement mag-
both quantities are negative. We show nitudes, W Fu. Work is positive if the force acts
stressstrain behaviors for a stiff machine in the same direction as the displacement of the
(greater slope) and a soft machine (lesser slope). object, so a positive W implies that work is done
For a rock specimen tested into the inelastic on the object whereas a negative W implies work
portion of the stressstrain curve the force, Fs, is is done by the object. Because the force changes
not related linearly to the displacement, us, but continuously with displacement we dene work
takes on a relationship we characterize in a as the integral:
general way as Fs f(us). The slope of the u2
stressstrain curve for the specimen is quantied
by the rst derivative of the function f with
W12 f (u) du
u1
(9.2)
respect to us which we write as f (us). In the initial
stage of the test f (us) Cs, so the slope is positive The work is equivalent to the area under the curve
and constant (Fig. 9.7a), whereas after inelastic f(u) on a force versus displacement graph between
deformation begins f(us) is variable. At the peak the two limiting displacements, u1 and u2. We do
value of stress the slope is zero and thereafter not specify a particular function, f (u), and inte-
f (us) is negative. For the machine the linear rela- grate it, but rather we geometrically determine
tionship is retraced as the load decreases and the the incremental work as one of the areas, 1
displacement fully recovers in the post-peak through 4 (Fig. 9.7) and compare them (Hudson
elastic regime (Fig. 9.7b, c). et al., 1972).
Consider a test in which loading has pro- are those that are not prevented by the boundary
gressed along the elastic portion of the forcedis- conditions, such as the xed lower bar. Applying
placement curve until Fs F (Fig. 9.7a). For the this principle to the system illustrated in Fig. 9.7
next small increment of force, F, the displace- we add the two equations in (9.5) and set the
ment increases by u and the increment of work results equal to zero:
done by the force on the specimen spring is:
Wm Ws 12 Cm | f(us) | (u)2 0 (9.6)
W Fu 12 Fu (9.3)
The system is in static equilibrium only if the
W is called the virtual work and u is called the machine stiffness, Cm, and the magnitude of the
virtual displacement. We use (9.3) to calculate the specimen stiffness, |f (us)|, are identical.
virtual work associated with the specimen and To understand the implications of (9.6) we
testing machine in the post-peak stress regime. In examine the two cases where equilibrium is not
that regime, suppose the actuator applies forces attained. First, consider the machine to be stiffer
Fs P and Fm P so the system is in equilibrium. than the specimen, so Cm |f (us)| and area 2 is
Imagine replacing the actuator with a perfectly greater than area 3 (Fig. 9.7a, b). For this condi-
rigid bar that transmits this force and couples the tion, the work that the machine spring can do on
specimen and machine springs so they must dis- the rest of the system is not sufcient to meet the
place by the same amount (Fig. 9.6c). If a down- requirements of the work done on the specimen
ward virtual displacement, u, is imposed on the spring. To achieve the downward displacement,
upper horizontal bar (Fig. 9.6d), the forces within u, work must be done by some external force, for
the springs change as: example the hydraulic pump supplies more uid
to the actuator, thereby increasing the applied
Fm Cmu
(9.4) force. These testing conditions are described as
Fs f (us)u | f(us) |u, post-peak stress stable because the system will remain in static
Using (9.3) the virtual work for the machine and equilibrium unless some external action is taken.
specimen are: If this external action is carefully controlled, the
test can proceed under stable conditions and the
complete stressstrain curve will be captured.
Wm Pu 12 Cm(u)2
(9.5) Now suppose the machine is softer than the
Ws Pu 12 | f(us)|(u)2 specimen spring, so Cm |f (us)| and area 2 is less
than area 4 (Fig. 9.7a, c). The machine spring can
Note that u is small (less than unity), so (u)2 is do more work than the specimen spring requires;
very small. Thus, the leading terms in (9.5) are so if the distance between the upper bar and the
larger than the second terms, so Wm is a negative specimen spring is not changed, the machine will
quantity and Ws is positive. That is, work is done drive the top of the specimen downward at an
by the machine on the specimen. accelerating rate. For an actual testing machine,
The next step in the mechanical analysis is to this could lead to a catastrophic result, and the
determine if the system is stable by asking: what stressstrain information for the remainder of the
are the conditions under which the system will test would be lost. These testing conditions are
spontaneously evolve to a different state? Such a described as unstable because the system self-
spontaneous response could represent cata- destructs unless some external action is taken, for
strophic failure of the specimen. To address this example the hydraulic pump drains uid out of
question we apply the principle of virtual work: for the actuator, thus decreasing the applied force.
any possible virtual displacement of an elastic This must happen at a sufcient rate for the system
body in equilibrium, the total virtual work done to remain in equilibrium. Another solution is to
by the internal forces, the body forces, and the control the test using the displacement of the
surface forces must vanish (Venkatraman and specimen, us, instead of the force from the actua-
Patel, 1970, p. 120). Possible virtual displacements tor. Modern testing machines use servo-control
systems in which computers, connected to trans-

Uniaxial tension
ducers monitoring force and displacement, are applied by testing
capable of feeding back the corrective action to machine
sensitive hydraulic valves in order to stabilize
the test. Hemispherical
cap and seat
Because Cm is always positive, instability req-
uires that the derivative of the specimen force Cement
displacement curve, f (us), be negative according to
Grip fixture
(9.6). Therefore we conclude that instability is only
possible in the post-failure region of the typical
stressstrain curve. Furthermore, instability
requires that the magnitude of f (us) exceed that of Dogbone
Cm. Thus, unstable specimen behavior is dened as: Tensile sample
fracture
| f(us) | Cm and f(us) 0, unstable (9.7)
A soft testing machine is one in which the post-
failure behavior is unstable. It should be clear that
Cement
rock testing to failure and beyond is not a trivial
endeavor, but these issues are, for the most part,
addressed by modern servo-controlled testing Hemispherical
machines. cap and seat
Lead
wires Force transducer
9.2.2 Uniaxial tensile and compressive
strength Testing machine frame
Denitions of the uniaxial tensile strength, Tu, and
Fig 9.8 Specimen holder for uniaxial tensile strength test
uniaxial compressive strength, Cu, are written in
(Obert and Duvall, 1967). Dogbone sample is broken by a
terms of the extreme values of the axial stress, a,
tensile fracture.
and are conditional on the presumed homo-
geneous state of stress written here in terms of
the principal values:
Duvall, 1967). The ends of the rock specimen are
Tu max(a),1 0, 2 0 3 (9.8) held by cementing them onto end pieces,
attached to the testing machine through hemi-
Cu |min(a) |,1 0 2, 3 0 (9.9) spherical caps and seats. The objective is to record
the maximum value of the axial stress (9.8) and
The tests usually are conducted at room (atmos-
thereby measure the strength. These tests are
pheric) pressure, so the actual values of the two
designed to impart a uniform stress state through-
equal principal stresses are about 0.1 MPa. This
out the sample; however, by cementing the speci-
is negligible compared to typical rock strengths in
men to the end pieces it is forced to contract
uniaxial tests which are of the order 10 to 100 MPa.
radially and extend axially in concert with the
The uniaxial tensile and compressive strengths
end piece. This constraint can induce elevated and
have the same units and dimensions as stress:
non-uniform stresses within the sample that can
strength, Tu or Cu [ ] N m2 Pa (9.10) lead to premature failure. By machining the spec-
imen to a smaller radius near its mid-section (Fig.
Tu{}M L1 T2, Cu{}M L1 T 2 (9.11)
9.8) a dogbone-shaped sample is prepared that
Strength is a scalar quantity and does not carry a carries greater stress at the mid-section in pro-
sign. portion to the reduced cross-sectional area (Jaeger
A schematic illustration of a uniaxial tensile and Cook, 1979). With this shape, failure can be
strength apparatus is shown in Fig. 9.8 (Obert and induced away from the end constraints, in a
Table 9.1. Rock mechanics laboratory tests for Table 9.2. Rock mechanics laboratory tests for
uniaxial tensile strength (MPa). uniaxial compressive strength (MPa).
Rock type From To Mean Rock type From To Mean
Quartzite 17 28 25 Quartzite 200 304 252

Gneiss 3 21 14 Gneiss 73 340 159
Basalt 2 28 13 Basalt 42 355 150
Granite 3 39 12 Granite 30 324 166
Limestone 2 40 12 Limestone 48 210 102
Sandstone 3 7 5 Sandstone 40 179 96
Shale 2 5 3 Shale 36 172 95
Pittsburgh coal 1.9 3.2 2.5 Pittsburgh coal 14 30 22
region that has a more homogeneous state of along the axis of the specimen and record the
stress. magnitude of its minimum (most compressive)
A data set (Table 9.1) for uniaxial tensile value (9.9). Friction between the end platens and
strengths (Bieniawski, 1984) is given for the same the sample may constrain the sample to expand
suite of rocks used to tabulate elastic properties laterally in concert with the platens. The resulting
(refer to Tables 8.2 and 8.3). stress state near the ends can be non-uniform and
The uniaxial tensile strengths range over about not uniaxial. The effects of the end platens on the
one order of magnitude among all rock types tab- local stress state and the consequences for style
ulated, and even among different samples of the and localization of deformation in uniaxial com-
same lithology. The tensile strengths for crys- pression have been investigated (Peng, 1971; Peng
talline rocks can be as small as those for clastic sed- and Johnson, 1972). Proposed methods to com-
imentary rocks, but typically they are somewhat pensate for this effect include matching the
greater. We conclude that laboratory specimens of elastic constants of the end platens to the sample
rock have uniaxial tensile strengths that range and inserting materials between the ends and the
from about 2 to 40 MPa with a typical value of platens to reduce the friction. These and other
about 10 MPa. renements of testing procedures have resulted in
Comparing the apparent Youngs moduli from more accurate determinations of uniaxial com-
Table 8.2 to the tensile strengths of Table 9.1, pressive strengths (Bieniawski and Bernede, 1979).
notice that the Youngs moduli are several thou- Selected uniaxial compressive strengths are
sand times greater than the tensile strengths. given in Table 9.2 (Bieniawski, 1984).
Using 1 : 4000 as representative of the ratio of These uniaxial compressive strengths range
tensile strength to Youngs modulus, the value of over about one order of magnitude. Crystalline
the axial extension at failure would be rocks tend to have greater compressive strengths
ea(max) Tu E 2.5 104. A practical conse- than clastic sedimentary rocks, but some granites
quence of this great difference between Youngs are weaker than some shales. We conclude that
modulus and the uniaxial tensile strength is that laboratory specimens of rock have uniaxial com-
axial extensions in uniaxial test samples are very pressive strengths that range from about 30 to
small. In most cases the deformation is approxi- 350 MPa with a typical value of about 150 MPa.
mately elastic for these small extensions. We con- Comparing Tables 9.1 and 9.2 we draw another
clude that rocks have an elastic stiffness on the conclusion: typical laboratory samples of rock
order of 103 times the tensile strength under these subject to uniaxial loading are weaker in tension
conditions. than in compression by about one order of magni-
An apparatus for conducting uniaxial com- tude. This has profound implications for the devel-
pressive tests is shown in Fig. 8.25. Again, the opment of brittle deformation in Earths crust and
objective is to induce a uniform normal stress its interpretation by structural geologists.
The minimum value of the axial strain (con- (a) (b)

traction) before failure is determined by compar-
ing the Youngs moduli and uniaxial compressive Wedge
strengths taken from Tables 8.2 and 9.2, respec-
tively. Using 1 : 300 as a representative value for Flaw
the ratio of compressive strength to Youngs
modulus, we have ea(min) Cu E 3.3 103.
This and the preceding example should reinforce
the important concept that stiffness and strength
are different by three or four orders of magnitude,
and that the magnitudes of the axial strains at the (c) (d)
initiation of failure for these brittle materials are
on the order of 103 to 104.
Given the very small elastic strains that Softer
precede failure in uniaxial tests one might con- Sliding
clude that elastic deformation is unimportant for crack
structural geologists. This would be a mistake. For
one thing, it is common to have a sample Stiffer
length, L, measured in kilometers for structures
in the Earth, so typical displacements would be on
the order of decimeters to meters. For example, Fig 9.9 Schematic examples of microscopic mechanisms of
the dike at Ship Rock (Fig. 2.7) is about 3 km long deformation at the grain scale in rock during strength tests.
and the adjacent rock was displaced about 1 m as (a) Microcrack growth from flaws within mineral grains. (b)
the dike opened. Some of the fault segments that Wedging of one grain between neighbors with grain
ruptured during the Hector Mine earthquake (Fig. boundary sliding. (c) Lateral extension of soft grain promotes
8.13) are more than 10 km long and slip up to crack growth in adjacent stiff grain. (d) Slip of inclined flaw
induces wing cracks.
several meters was recorded along their surface
traces. Yet the surface displacement eld clearly
correlates with that of an elastic model. Thus, for nisms of deformation during strength tests of
structural geologists, the elastic strains that rock, both in uniaxial and multi-axial compres-
accompany brittle fracture play an important role sion (Tapponnier and Brace, 1976; Wong, 1982b;
in crustal deformation. Kranz, 1983). The mechanisms of deformation
The most prominent macroscopic mechanism include the nucleation, opening, and propagation
that acts to limit the stress under conditions of of isolated microcracks from aws within grains
uniaxial testing is fracture. The word macroscopic is (Fig. 9.9a). Loads are concentrated at contacts
used here to refer to phenomena, observable with between grains (Fig. 9.9b) and one grain may be
the unaided eye, such as extension fractures, split- driven into the other, opening a crack. The exten-
ting fractures, and shear fractures (Figs. 9.1 and sion of a compliant (soft) grain parallel to the
9.2). In contrast, the word microscopic refers to the contact with a stiffer grain (Fig. 9.9c) can drive the
grain-scale phenomena within the sample that growth of opening microcracks in the stiffer grain.
individually are visible only with a microscope or Inclined grain boundaries, pre-existing cracks, or
hand lens. The details of the grain-scale deforma- cleavages can slip during loading in compression,
tion during these experiments, the ways in which thereby creating tensile stress concentrations near
this deformation proceeds to weaken the whole the tips of the slipping surfaces from which
specimen over the course of the test, and the strain opening (wing) cracks nucleate (Fig. 9.9d).
localization into discrete fractures along which the In conclusion, uniaxial tests measure the bulk
specimen breaks can be quite complex (Peng and strength of the collection of constituent mineral
Johnson, 1972). A great deal of research has been grains, cracks, pores, and other heterogeneities
devoted to identifying the microscopic mecha- that make up particular samples. As such these
samples may or may not be representative of the s3

strength in the natural setting of that rock, or of s2
the state that rock was in during the tectonic event
of interest to the structural geologist. Both the uni- Tu
s1
formity of the stress and the homogeneity of the 0
, s 3=
rock sample need to be evaluated when interpret- s 1= s 2
ing the results of laboratory strength tests. Finally,
the strength criteria (9.8) and (9.9) do not explicitly
address the physical mechanisms responsible for
3
=s
the loss of load-carrying capacity; they simply
2
=s
assert that this happens at a certain stress level.
1
s
9.2.3 Polyaxial strength, stress invariants, Cu
and stress deviation
s1 = Pc = s2
Below the traction-free surface of the Earth the
state of stress may be approximated by Andersons
C t
standard state, introduced in Chapter 6, in which
the principal stresses are equal in magnitude
(1 2 3) and become more compressive lin-
early with depth. This isotropic state of stress is
likely to be supplemented in regions of tectonic
activity by stresses that result in a polyaxial state of
Fig 9.10 Plot in principal stress space with uniaxial tensile
stress: the principal stresses have different magni-
and compressive strengths, Tu and Cu, and triaxial
tudes, none of which is likely to be zero. In prin-
compressive strength, Ct.
cipal stress space (Fig. 9.10), the uniaxial tensile
strength plots as a point (Tu, 0, 0) on the positive 1-
axis and the uniaxial compressive strength plots the strength. We dene a criterion for failure as a
as a point (0, 0, Cu) on the negative 3-axis. The function of the principal stresses that represents
paths representing these tests are straight lines the failure surface and write this function in terms
from the origin. A third path lies in the plane (1 of the stress invariants as dened in Chapter 6. Recall
2) and extends into the octant where all prin- that the invariants have the special property that
cipal stresses are compressive along the straight they do not change magnitude with the orienta-
line 1 2 3. This represents the isotropic com- tion of the Cartesian coordinate system. Because
pressive loading envisioned by Anderson and physical properties of a material, such as strength,
usually is followed initially in laboratory triax- should not depend upon the arbitrary orientation
ial tests. Then, two of the principal stresses are of a coordinate system, it is natural to dene such
held constant and equal to what is called the properties in terms of the stress invariants (Jaeger
conning pressure, Pc, while the third principal and Cook, 1979). Criteria for failure under uniaxial
stress becomes more compressive until failure at conditions are equivalent to placing limiting
the point (Pc, Pc, Ct), the triaxial compres- values on the rst invariant of the stress tensor,
sive strength (see next section). Strength in the I1 1 2 3, when only one principal stress is
context of a polyaxial state of stress is represented non-zero:
by a surface, called the failure surface, that passes
through these three points. The complete failure I1 T0,2 0 3 (9.12)
surface separates possible states of stress between
I1 C0,1 0 2 (9.13)
the origin and the surface from impossible states
of stress on and beyond the surface. Note that this pre-supposes that the initiation of
The combination of principal stresses (1, 2, failure depends only upon the current homo-
3) at any point on the failure surface represents geneous state of stress. That is, failure does not
depend upon stress gradients or the history of In practice, strength has been dened using
loading. many different combinations of the stress compo-
For a polyaxial state of stress it is commonly nents as dictated by the many different kinds of
assumed that failure is primarily related to the dis- testing machines and sample congurations that
tortion of a material, whereas changes in volume have been invented. Some of these are based on a
are of secondary importance. Recall from our dis- mathematical premise, such as dependence upon
cussion of the bulk modulus of elasticity in Chapter invariants of stress or of stress deviation, whereas
8 that volume change is proportional to the mean others are based on physical arguments for the
normal stress, m 13 (1 2 3) 13 (I1). This is causes and mechanisms of failure. Usually these
the normal stress that acts on the octahedral strength criteria may be characterized as some
planes, the eight planes with normals that are functional relationship among the components
equally inclined to the directions of principal of principal stress (Jaeger and Cook, 1979):
stress. The vertices of the octahedron dened by
1 f ( 2, 3) (9.18)
these planes lie on the principal stress axes.
Therefore, the mean normal stress is sometimes Each of these functions denes a failure surface in
referred to as the octahedral normal stress. The mean principal stress space. These criteria do not take
normal stress is subtracted from each normal possible spatial gradients in the stress eld into
stress component to dene the components of the account, so they must be applied on a point-by-
stress deviation tensor as follows: point basis in a heterogeneous eld of stress.
sxx xx m, syy yy m, szz zz m,
9.2.4 Triaxial strength, confining
syz yz, szx zx, sxy xy (9.14) pressure, and pore pressure
Note that the shear stress components are identi- Largely because of the technical difculties in the
cal to the shear stress deviation components. design of a true polyaxial apparatus, the most
The invariants of stress deviation are (Jaeger common procedure used in rock mechanics is the
and Cook, 1979): triaxial test. The conditions imposed are illustrated
in principal stress space (Fig. 9.10) and the appar-
J1 sxx syy szz 0 atus is shown in a schematic cross section (Fig.
J2 (sxxsyy syyszz szzsxx) s2xy s2yz s2zx 9.11). More complete and precise engineering
(9.15)
J3 sxxsyyszz 2sxy syzszx sxxs2yz drawings are available (Griggs and Handin,
syys2zx szzs2xy 1960b). The apparatus itself is placed between the
The second invariant of stress deviation is related two platens of a testing machine (Fig. 8.25), which
to the shear stress acting on the octahedral planes provides the axial load. A key feature of the appar-
which is called the octahedral shear stress, o: atus is a port for supplying uid, under pressure
called the conning pressure, Pc, to the region
1
o [(1 2)2 (2 3)2 (3 1)2 ] 12 between the inner wall of the pressure vessel and
3
the cylindrical jacket surrounding the rock

2
J (9.16) sample. The jacket, often rubber or malleable
3 2
metal like copper, is impermeable to uids and
For materials that failure due to shearing one pos- more easily deformed than the sample itself. A
sible criterion using stress deviation invariants is separate port can supply a different uid, under
based on the octahedral shear stress attaining a pressure called the pore pressure, Pp, directly to the
critical value, Oo, taken as a constant (Jaeger and sample surface, and thence to the internal pores
Cook, 1979): of the rock. Because Pp Pc, the pore uid does not
inate the jacket and stays within the rock pores.
Oo max (o) (9.17)
A furnace can be attached to the pressure vessel to
Here Oo is the octahedral shear strength of the mate- heat the sample and the vessel to a designated
rial (Hobbs, 1962). temperature. Here we examine the role that
Force applied stress, a r, but they change in shape for non-
by testing machine isotropic stress states, and these shape changes
are believed to be related to failure. Recall,
Piston head
however, that an anisotropic elastic material will
Force transducer distort under isotropic loading (Chapter 8), so
Pore fluid this presumption is appropriate, at best, for
Pressure, Pp
samples that are isotropic with respect to elastic
Platen
properties.
Seal Positive differential stress corresponds to
extension tests and negative differential stress
Deformable corresponds to compression tests (Fig. 9.1). Extreme
jacket
values of the differential stress are recorded as the
Rock sample differential strength for extension, De, and compres-
sion, Dc, respectively:
Confining fluid
pressure, Pc
De max (),1 Pc, 2 Pc 3
Pressure vessel (9.21)
Dc |min () |,1 Pc 2, 3 Pc
(9.22)
Testing machine frame
Some experimentalists use the extreme value of
Fig 9.11 Schematic illustration of a triaxial testing the axial stress as the measure of strength in tri-
apparatus. axial tests (Jaeger and Cook, 1979):
Ct |min(a) |,1 Pc 2, 3 Pc (9.23)

conning pressure and pore pressure play in
determining rock strength under conditions of For a compression test this is referred to as the tri-
brittle deformation. axial compressive strength (Fig. 9.10).
The design objective for triaxial vessels (Fig. As an example of results from triaxial testing
9.11) is to achieve a uniform state of stress and consider data from a study of sedimentary rocks
pore pressure throughout the rock sample. The from the Tertiary basins of Japan (Hoshino et al.,
axial and radial stress components are presumed 1972). One hundred different rocks, ranging in
to be related to the applied force, f, and conning age from Pliocene to Oligocene, were deformed in
pressure, Pc, as: triaxial compression at room temperature, and
a f A 1 or 3 the differential strengths, Dc, were recorded at
(9.19) conning pressures ranging from 0.1 MPa (atmos-
r Pc 2 and 3, or 1 and 2
pheric pressure) to 245 MPa (equivalent to about
The apparatus is called triaxial, but it is not 10 km depth). The lithologies were primarily clay-
capable of imposing normal stresses of different stone, siltstone, shale, and sandstone, although a
magnitudes in three coordinate directions, only few volcanic rocks were included. All of the
in the axial and radial directions. To record the samples were dried (atmospheric pore pressure),
results of a triaxial test graphically, some experi- and all were cored so the cylindrical axis was per-
mentalists plot the differential stress, , versus the pendicular to the sedimentary bedding. The tri-
axial strain, ea (Clark, 1966), where: axial compressive strengths (9.23) for three
sandstones are used to plot the stress state at
a r (9.20)
failure (Fig. 9.12) in the principal stress plane
The rationale is that rock samples only change in dened by the conning pressure and the axial
volume during loading to an isotropic state of compression.
s1 = s2 (MPa) tions remained the same. Given an increase of

250 200 150 100 50 0 compressive stress due to the weight of overlying
0 rock of about 25 MPa km1, these tests could rep-
resent deformation of the Ohtawa basalt at the
s3
s 2= Earths surface, at about 2 km depth, and at about
s 1=
200 4 km depth. The macroscopic deformation mech-
anisms changed signicantly in these tests from
wedge fracture with axial splitting at the lowest
s3 (MPa)
conning pressure, to a localized shear fracture
oblique to the sample axis at the intermediate
conning pressure, to a broad network of oblique
shear fractures at the highest conning pressure.
600 Clearly the magnitude of the conning pressure
plays an important role in determining the mech-
anisms that govern the strength of this rock.
800 An important concept concerning the
Maze strength of soils that are saturated with water or
Maze other uid was introduced by Karl Terzaghi in
Furukawa 1923 and later found application in studies of
1000
rock, concrete, and other porous and permeable
Fig 9.12 Plot of triaxial strength data in principal stress
solids (Terzaghi, 1943). The concept depends upon
space for three sandstones from Tertiary basins of Japan
the uid lling all the pores of the material and
(Hoshino et al., 1972).
these pores must be homogeneously and perva-
sively distributed throughout at a scale that is
The uniaxial compressive strengths for the small compared to the scale of interest. Further-
three sandstones range from 46 to 113 to 122 MPa, more the pores must be interconnected in such a
so there is considerable variability in strength way that local changes in uid pressure during
even among samples representing the same lithol- deformation are rapidly equilibrated by ow
ogy. The two samples of Maze sandstone, taken through the network of pores. Finally, the concept
from the same formation, differ by almost a factor is purely mechanical, so the uid must not react
of three. Presumably these differences reect chemically with the solid. Under these condi-
subtle differences in the constituents of these tions, Terzaghi discovered that the deformation
rocks that are not reected in their lithologic and and failure of soil samples in the laboratory
formation names. The triaxial compressive depended upon the so-called effective stress state, as
strengths, Ct 3, increase as conning pres- opposed to the stress state as ordinarily dened
sure, Pc 1, increases. Note that the general (Nur and Byerlee, 1971).
form of the experimental data for a particular The effective stress state is related to the stress
suite of samples approximates a linear relation- state by adding the pore uid pressure to the
ship between the principal stress components. normal stress components:
Later in this chapter we introduce a failure cri-
terion, called the Coulomb criterion that is con- xx Pp, yy yy Pp, zz zz Pp,
xx
sistent with a linear relationship between the yz yz, zx zx, xy
xy (9.24)
principal stress components, 1 and 3, at failure.
In Figure 9.2 samples of Ohtawa basalt are Do not confuse the components of effective stress
shown after deformation in the same triaxial (9.24) with the components of stress deviation (9.14)
testing apparatus used in the previous example dened by subtracting the mean normal stress
(Hoshino et al., 1972). These particular samples from each of the normal stress components. One
were deformed at conning pressures of 0.1, 49, can visualize the role of pore pressure by consi-
and 98 MPa, respectively, while all other condi- dering the typical stress state to be compressive
Table 9.3. Triaxial test results on Berea Taking the conning and pore pressures to 500 MPa
Sandstone (MPa). resulted in a differential strength of 63 MPa, within
the range of values for the tests at lower pressures.
Pc (Pp 0) Dc Pc Pp Dc Clearly, the pore pressure serves to neutralize the
role of conning pressure in increasing the differ-
0 72 0 72 ential strength of Berea Sandstone.
50 159 50 82 Consider the triaxial strength test data on
100 242,248 100 78
Berea Sandstone plotted in principal stress space
150 150 75
(Fig. 9.13a). Conning pressures were varied from
200 418,432 200 60,64
0 to 200 MPa, representative of depths to about
500 500 63
8 km. The tests were conducted at room tempera-
ture and pore pressures from 0 to 175 MPa. For
(negative) at depth. There, the normal stress com- each condition of conning and pore pressure, a
ponents act inward on the boundaries of a repre- data point represents the principal stress state at
sentative element of porous rock, whereas the uid failure. For a given maximum principal stress
pressure pushes outward in all directions from (1 Pc), say 200 MPa, the minimum principal
within the pores. In this way the pore pressure stress (3 a) at failure is less compressive as
serves to counterbalance the compressive stress. the pore pressure increases. This variation in
Since the pore pressure acts equally in all directions strength is considerably greater than the varia-
it does not inuence the shear stress components. tion between samples tested at the same pore
Two experimental observations, carried out pressure, so it is deemed to be signicant. It is not
using triaxial testing procedures, led Terzaghi to possible to summarize these data with a single
the concept of effective stress. When the axial and line or curve.
radial stress are of equal magnitude, the sample is If Terzaghis concept has merit, the strength of
subjected to an externally imposed isotropic state Berea Sandstone at a given effective conning
of stress, a r Pc. As this isotropic state of pressure should be constant. The effective conning
stress changes, the volume of the sample changes. pressure is dened as the difference between the
On the other hand if the pore pressure is conning pressure and the pore pressure:
increased in magnitude at the same rate as this
Pc Pc Pp (9.25)
isotropic stress so Pc Pp, Terzaghi noted that the
volume of the sample did not change appreciably. In Figure 9.13b the data for Berea Sandstone are
Thus, the volume change is related to the effective plotted in effective principal stress space. On the
conning pressure, Pc Pp. abscissa the maximum principal effective stress is
The second observation of Terzaghi was that plotted, and this is equivalent to the negative of
the strength of laboratory samples increased the effective conning pressure, 1 Pc. The
signicantly with conning pressure, but did not minimum effective stress, 3, is plotted on the
increase appreciably if the pore pressure was ordinate. All data points at a given effective
increased in concert with the conning pressure. conning pressure have essentially the same
We use experimental data for Berea Sandstone in strength, thereby verifying Terzaghis concept.
Table 9.3 to illustrate this phenomenon (Handin et Furthermore, the collection of data points repre-
al., 1963). senting effective stress states at failure collapse
For this porous and permeable sandstone the approximately (presumably within the experi-
differential strength, Dc, increased from 72 to over mental error) onto a straight line. This fact sug-
400 MPa with increases in conning pressure from gests that a single failure surface can be dened in
0 to 200 MPa and zero pore pressure. With pore terms of the effective principal stresses.
pressure equal to conning pressure the differen- The pore pressure also effects the transition
tial strength varied non-systematically over the from brittle to ductile behavior (Handin et al.,
range 60 to 82 MPa. These changes are probably 1963). This is illustrated (Fig. 9.14) on axial stress
within experimental error of being constant. versus axial strain curves for Indiana Limestone at
44 480
(a) s1 = s2 = Pc (MPa)
250 200 150 100 50 0 0.0
0
Force applied to piston (N)

100 27.6
=s 3
s 1= s 2 34.5
200
s3 (MPa)
41.4
22 241
Pore pressure
0 MPa
50 MPa
500
100 MPa
55.2
600
150 MPa
175 MPa 62.1
700
(b) s1 = s2 (MPa) 68.9 MPa
0
250 200 150 100 50 0
0
0 2.54 5.08 7.62 104
100 Piston displacement (m)
= s3
s 1= s 2
200 Fig 9.14 Plot of axial force versus piston displacement for
triaxial test of Indiana Limestone at a constant confining
s3 (MPa)
pressure of 69 MPa and variable pore pressure from 0 to

69 MPa. Reprinted from Robinson (1959) with permission of
Pore pressure the Colorado School of Mines Library.
0 MPa
500
50 MPa
100 MPa
600 9.2.5 Frictional strength
150 MPa
175 MPa
700 Though simply expressed, the laws of friction encapsu-
late a host of microscopic and nanoscopic phenomena
Fig 9.13 Plots of triaxial strength data for Berea Sandstone whose elucidation has become one of the most fasci-
at a variety of pore pressures (Handin et al., 1963). nating pursuits in applied physics (Hhner and
(a) Principal stress space. (b) Effective principal stress space. Spencer, 1998).
Leonardo da Vinci rst addressed questions about
a xed conning pressure Pc 69 MPa and pore friction through a series of experiments on
pressures varying from Pp 0 to 69 MPa (Robinson, sliding objects down an inclined plane (Resnick
1959). For all the tests the initial behavior is nearly and Halliday, 1977). Since he lived from 1452 to
linear and presumably elastic. For those tests con- 1519, long before force was clearly dened by
ducted at pore pressures ranging from 0 to Newton, his concepts were entirely empirical.
34.5 MPa (effective conning pressures of 69 to These concepts are consistent with a simple pro-
34.5 MPa) the axial stress becomes more compres- portionality between the magnitudes of two
sive, or is approximately constant, as strain accu- forces: F acting parallel to a horizontal plane at
mulates beyond the elastic limit. This corresponds the moment motion begins, and W acting down-
to ductile behavior. In contrast, for greater pore ward across the plane and being the weight of the
pressures (lesser effective conning pressures), object. The force F is said to initiate the motion
the axial stress becomes less compressive and the while the weight W resists the motion. A some-
samples lose some load-carrying capacity as what more general relationship admits lesser
inelastic strain accumulates. This is characteristic values of the force F before the object moves, and
of brittle behavior. In general, samples tested at retains the proportionality:
low effective conning pressures exhibit brittle
behavior; whereas those tested at high effective F sW (static);
(9.26)
conning pressure are ductile (Heard, 1960). F sW (sliding initiates)
The quantity s is called the static friction and this (a)

proportionality is known as Amontons Law after y
Guillaume Amonton who published his experi-
mental results in 1699 (Bowden and Tabor, 1950).
In this relationship the friction is independent of
the apparent area of contact. There is no relative
motion if the inequality is satised, because the
force F is insufcient to overcome the frictional
resistance between the object and the surface. W f
y =W W
Sliding initiates at the moment the force equals co x =W
the product of the static friction and the weight. sf sin
U f
Today, the microscopic to macroscopic phenom-
ena associated with frictional sliding on faults is
one of the most fascinating pursuits of Earth x
scientists studying earthquakes (Dieterich, 1986,
W
1994; Linker and Dieterich, 1992; Lockner and
Beeler, 2003). ds
On an inclined surface (Fig. 9.15a) the force (b)
that initiates sliding is the component of weight d
acting tangentially to the slope, Wx. The force
resisting this motion is the component acting F
v
normal to the slope, Wy. For an angle of inclina-
tion specied by , the static friction is equal to
the ratio of these forces and to the tangent of the
angle at the moment sliding initiates:
s | |
Wx

sin
Wy cos
tan ,0 90 (9.27)
W
The absolute value is taken to assure a positive (c)
sign. In this context the angle is referred to as
the angle of friction.
A conceptual model for laboratory friction
experiments includes a surface in contact with an B
C
object, sometimes called the slider, and a spring
Force, F
A E
that represents the elastic behavior of the testing
machine (Fig. 9.15b). As the right end of the spring
moves with a steady velocity v imposed by the
testing machine, the force F in the spring D
increases linearly with the displacement d of the
point at the right-hand end of the spring, while
the slider remains xed, ds 0. When the equality
Displacement, d
in (9.26) is satised the slider begins to displace
and the plot of force versus displacement becomes Fig 9.15 Schematic illustrations of concepts related to
non-linear (Fig. 9.15c). The behavior of this system friction. (a) Block on inclined plane slides under the action of
in the simplest cases can be described either as gravity. (b) Block on horizontal plane slides under action of
stable sliding or as stick slip sliding (Byerlee, 1978; force transmitted through a spring. (c) Plot of force versus
Dieterich, 1981). In stable sliding, the slider accel- displacement shows stable sliding (OABC) and stick slip
erates gradually to a constant velocity equal to v (OABDE). Reprinted from Byerlee (1978) with permission of
Birkhanser-Verlag.
and maintains this speed with a constant spring
Table 9.4. Static friction. A

o 5.17
Rock type From To
Shear traction, |ts| (MPa)

o
Gabbro 0.48 0.67 B
x
Granite 0.18 0.66 C 3.45
Gneiss 0.61 0.71 x o
D o
Marble 0.62 0.75
x
Quartzite 0.48 0.67 x
Trachyte 0.56 0.68 x o

1.72
Sandstone 0.51 0.68 x
x
force (path OABC). On the other hand, stick slip
behavior involves the sudden acceleration of the
slider with concurrent drop in the spring force.
5.17 3.45 1.72 0
Eventually the force recovers and such events may Normal traction, tn (MPa)
be repeated many times during a single experi-
ment (path OABDE, etc.). This behavior has been Fig 9.16 Plot of shear traction versus normal traction for
suggested as the laboratory counterpart of the laboratory friction tests on sliding surfaces in marble (A),
mechanism responsible for earthquakes (Brace trachyte (B), trachyte with smoother surfaces (C), and
and Byerlee, 1966). sandstone (D). Reprinted from Jaeger and Cook (1979) with
The actual conguration of specimen and the kind permission of Mrs. Jennifer D. Cook.
testing machine for laboratory friction experi-
ments is quite variable (Jaeger and Cook, 1979):
F dW (constant sliding velocity) (9.28)
some are not much more sophisticated than the
conceptual model of a slider on a surface (Fig. Here d is the dynamic friction. The dynamic fric-
9.15b). A common design utilizes the standard tri- tion typically is less than the static friction.
axial test with a diagonal saw cut across cylindri- Although static friction, s, is fundamentally
cal specimens (Fig. 9.11). Another design involves dened in terms of the applied forces (9.26), one
pushing a block of rock between two adjacent may divide through by the apparent contact area
blocks (Fig. 9.4b), and still another (Tullis and and write a comparable relationship for the static
Tullis, 1986; Tullis, 1988) employs the counter- friction in terms of the shear traction, ts, and
rotation of two hollow cylinders. Reported values normal traction, tn, acting on the surfaces:
of the static friction do not vary widely. In Table
9.4 a few representative values are recorded
|ts | stn, tn 0 (sliding initiates) (9.29)
(Jaeger and Cook, 1979). The absolute value is used on the left-hand side
The rule of thumb we take from these data is because the relationship should not depend upon
that values of the static friction for common rock the arbitrary sign of the shear traction. Recall that
types typically range from about 0.5 to 0.8, with a positive normal traction pulls on the surface,
0.6 being a good general estimate. Apparently whereas a negative traction pushes against it. The
there is some ambiguity in the reported values of normal traction must be zero or negative to insure
the static friction because some researchers use that the two surfaces stay in contact.
the force required to initiate sliding (point A, Fig. Laboratory data using an apparatus similar in
9.15c), others use the greatest value (point B), and design to that shown in Fig. 9.4b approximate a
still others use the value after a stable sliding at a linear relationship between the shear and normal
constant velocity is achieved (point C). For the case tractions as sliding initiates for marble (A), tra-
of sliding at a constant velocity a second value of chyte (B), trachyte with smoother surfaces (C), and
friction usually is dened as: sandstone (D), (Fig. 9.16). The surfaces of the test
specimens are described as moderately rough Table 9.5. Frictional strength and coefcient of
and the behavior is believed to be representative friction.
of many rock types ( Jaeger and Cook, 1979). Note
that the intercept on the shear traction axis is not Rock type Sf (MPa) c
zero for any of the data sets, so a more general
linear relationship than (9.29) is required: Marble 1.10 0.75
Trachyte 0.41 0.68
|ts | Sf ctn,tn 0 (sliding initiates) (9.30) Gabbro 0.38 0.66
Granite 0.31 0.64
The intercept on the ordinate, Sf , is the frictional
Sandstone 0.28 0.51
strength in the absence of any normal traction and
the slope c is the coefcient of friction. This linear
relationship is identical in form to Coulombs cri- on the magnitude of the normal traction acting on
terion for the shear strength of intact solids which the sliding surface (Byerlee, 1978). The rst class
is described later in this chapter (9.38). The includes normal traction conditions to 5 MPa and
Coulomb criterion applies to the interior of a con- would therefore relate to very shallow conditions
tinuous solid rather than to discrete surfaces in in the Earth, typically less than a few hundred
frictional contact. The mathematical similarity of meters depth. This is the environment of the engi-
these equations should not obscure the fact that neering geologist and civil engineer (Barton, 1973).
the former describes a friction experiment where At moderate normal tractions, from 5 to
sliding is induced along two surfaces in contact, 100 MPa, many laboratory results for maximum
whereas the latter describes the initiation of a friction (point B on Fig. 9.15c) plot close to a line
shear fracture in an otherwise unbroken solid. with zero intercept and a slope equal to a static fric-
Laboratory experiments also have demonstrated tion of 0.85 (Fig. 9.17a). Rock types include sand-
that friction is dependent upon the velocity of stone, graywacke, limestone, quartzite, gneiss,
sliding and the time of contact (Dieterich, 1979a, granite, granodiorite, and gabbro. These results are
b, 1981; Kilgore et al., 1993). applicable to underground excavations, well-bore
Given data that dene the linear relationship problems, and sliding on faults down to about 4 km
(9.30), the static friction, s, is derived from the depth. Apparently the friction is not highly depen-
frictional strength and coefcient of friction as dent on lithology (with a few exceptions), nor is it
follows: particularly dependent on the roughness of the
|t s |S sliding surfaces. The data set for maximum friction
s f c, in the range from 100 MPa up to about
tn tn (9.31)
tn 0 (sliding initiates) 1500 MPa is applicable to sliding on faults at
depths from 4 to perhaps 60 km (Fig. 9.17b). A linear
For values of the normal traction less than or com- relationship with an intercept of 50 MPa and a
parable to the frictional strength, the rst term on slope equal to a coefcient of friction of about 0.6
the right-hand side makes a signicant contribu- ts much of the data (excluding materials such as
tion to the static friction. Under these conditions montmorillonite, vermiculite, and illite). For
it is not appropriate to equate the coefcient of normal traction magnitudes greater than a few
friction and the static friction. If the normal trac- hundred MPa, the coefcient of friction is essen-
tion is much greater than the frictional strength, tially equal to the static friction (9.31).
the rst term becomes insignicant and the The frictional behavior of most rocks in the
coefcient of friction approaches the value of the ranges of normal traction specied in Fig. 9.17a and
static friction. Some laboratory values for the fric- b can be characterized as follows (Byerlee, 1978):
tional strength and the coefcient of friction
from (9.30) are reported in Table 9.5. |ts | 0.85tn, 5 tn 100 MPa
Laboratory data on the friction of rock surfaces |ts | 50 MPa 0.6tn, (9.32)
may be categorized into three broad classes based 100 tn 2000 MPa
(a) 100 roughly perpendicular to the uniaxial tensile

stress within the specimen. The relative motion of
s |=
|t
0 the two fracture surfaces is approximately per-

.8 80 pendicular to the fracture plane, so we refer to
5
n this structure as an opening fracture. In materials as
t
heterogeneous at the grain scale as rock, it should

60
not be surprising that the growth of a macro-
scopic opening fracture can involve many differ-
ent microscopic deformation mechanisms. The
40
double cantilever beam testing procedure (Fig.
9.18a) provides the necessary control on the rate
20
of propagation to study these grain-scale mecha-
nisms (Hoagland et al., 1973). In these tests on spec-
imens of Salem Limestone and Berea Sandstone a
wedge was driven between two steel pins attached
100 80 60 40 20 0 on either side of a pre-cut notch in the rock
Normal traction, tn MPa sample. As the notch was opened by this wedging
(b) action, a fracture initiated in the region of stress
concentration at the notch tip and propagated
|t |
s = through the specimen. The rate of fracture propa-

5
0 gation was directly related to the rate of advance
0.6
tn 1000
of the wedge. The load applied to the wedge and
the opening displacement between the pins was
recorded, along with observations of the micro-
scopic deformation, acoustic emissions, and
500
macroscopic fracture length.
A representative example of the load versus
opening displacement records from these tests
(Fig. 9.18a) is correlated to drawings of the samples
2000 1500 1000 500 0
at different stages of deformation (Fig. 9.18b).
Normal traction, tn (MPa)
Examination of the samples before loading
Fig 9.17 Plot of shear traction versus normal traction for revealed minor microcrack damage near the
laboratory friction tests. (a) Tests at normal tractions from notch, presumably associated with sample prepa-
5 to 100 MPa. (b) Tests at normal tractions from 100 ration. The initial loading produced a relationship
to 1500 MPa. Reprinted from Byerlee (1978) with between load and displacement that is approxi-
permission of Birkhanser-Verlag. mately linear and reversible, consistent with
elastic behavior, and only minor microcracking
was detected. As loading increased the test record
These empirical equations, sometimes referred to was characterized by some acoustic emissions, a
as Byerlees Law, have proven to be useful in many distinct non-linearity in the loaddisplacement
practical applications. curve, and the development of abundant micro-
cracks at the notch tip forming a so-called damage
9.2.6 Fracture toughness zone. When the advance of the wedge was stopped
For the uniaxial tensile test conguration, a in this region the acoustic emissions continued,
typical macroscopic event is that a single fracture but at a decaying rate, and the load decreased
nucleates somewhere within or on the surface of somewhat with time. This demonstrated that
the specimen; this fracture propagates across the there was a time dependence to microcrack
specimen; and the specimen is split into two growth and to the development of the damage
parts (Fig. 9.8). The fracture surfaces are oriented zone.
Attainment of peak loading for the double can- (a)

tilever beam samples (Fig. 9.18a) roughly coincided
with the initiation of a macroscopic opening frac-
2
ture, propagating from the notch tip, and away
from the advancing wedge. After some fracture
growth a steady state was achieved, during con-
stant advancement of the wedge, in which the
fracture tip propagated through the sample
Load
accompanied by a zone of microcracking and
other grain-scale damage. One moment in this
steady-state process of fracture propagation is 1
illustrated schematically in Fig. 9.18c. The damage
zone (hachured region) develops in front of the
fracture tip and is left behind as a wake along
the side of the fracture surfaces. The spatial Rock
density of the microcracking is greater near the 0
macroscopic fracture surfaces (double hachured
region). Displacement
The phenomenon observed in these fracture (b)
propagation experiments is unlikely to occur in a
material with homogeneous strength. Rather, the 0
material would break at the point of greatest stress
concentration (in this case the notch tip) and the
fracture would propagate with little or no damage
zone. Because most rocks are highly hetero- 1
geneous at the grain scale, there are likely to be
many weak points where inelastic deformation
can proceed at local stress levels less than that at
the point of greatest stress concentration. The
microscopic deformation mechanisms active in
these damage zones include the growth of existing 2
microcracks within mineral grains (Fig. 9.19a), the
nucleation of microcracks at aws within grains
(Fig. 9.19b), the opening of grain boundaries (Fig.
9.19c), and the shearing of grain boundaries (Fig.
9.19c) (Friedman et al., 1972; Hoagland et al., 1973; (c)
Peck et al., 1985; Labuz et al., 1987). The roughness
of the fracture surfaces corresponds in part to the
grain size, because the macroscopic fracture prop-
agates both around and through individual
grains, seeking the path of least resistance.
Consider an opening fracture (Fig. 9.20) that Fig 9.18 Double cantilever beam testing procedure for
has not propagated through a laboratory speci- opening fracture propagation in Salem Limestone. (a) Plot of
men. The specimen is subjected to an applied load versus displacement (inset illustrates testing apparatus).
stress, a, that is less than the uniaxial tensile (b) Drawings of samples at different loads showing
strength, Tu. We would like to know the value of development of microcrack damage zone. (c) Schematic
applied stress required to initiate and continue illustration of steady-state fracture propagation with damage
zone. Reprinted from Hoagland et al. (1973) with permission
fracture propagation. Put another way, what is
of Springer-Verlag.
the resistance of a rock to fracture propagation?
(a) (b) (a) sa
Microcrack y
syy
Flaw x
x
(c) (d)
Grain 1 2b
2a
Grain 2
Fig 9.19 Schematic examples of microscopic deformation

mechanisms in damage zone during opening fracture
propagation. (a) Microcrack growth within mineral grains.
(b) Growth of cracks from flaws. (c) Opening of grain
(b)
boundaries. (d) Shearing of grain boundaries as grains pull
apart.
The answer comes from the engineering disci-

pline called fracture mechanics (Rice, 1968; Lawn
and Wilshaw, 1975; Kanninen and Popelar, 1985;
Anderson, 1995). By fracture propagation we mean
that two new surfaces develop along an extension
of the fracture plane that merge with and become
part of the older fracture surfaces (Fig. 9.20 inset).
Pulling the rock apart depends most directly on
the local tensile stress, yy, acting near the frac-
ture tip. The founders of fracture mechanics dis-
covered how to relate the remotely applied stress
to the local stress near the fracture tip (Grifth, Fig 9.20 Opening crack and associated stress. (a) Crack
1921, 1924; Irwin, 1958). They tackled this subject to applied tensile stress, a (inset shows crack tip
problem using elastic boundary value problems stress yy). (b) Photoelastic experimental image of maximum
for the stress eld around cracks subject to remote shear stress field near opening crack. Stress is concentrated
tension. The local stress near the crack tip is given in two lobes that converge on the crack tip.
approximately by:

12
a
yy a ,for x a,y 0 (9.33) for many loading congurations, the relationship
2x
between the local stress and the distance is the
Here, 2a is the fracture length and x is distance same: the local (crack tip) stress, yy, varies approx-
away from the tip in the plane of the fracture. imately as one over the square root of the distance
For many sample and fracture geometries, and from the tip. This is referred to as the near-tip stress
9.3 BRITTLE FAILURE IN A FIELD OF HOMOGENEOUS STRESS 357
distribution. As the fracture tip is approached, x Table 9.6. Fracture toughness (MPa m1/2).
becomes smaller and smaller relative to 2a, so yy
becomes larger and larger relative to a. This indi- Rock type From To
cates that the stress is highly concentrated near
the fracture tip and it is this stress concentration Granite 1.66 3.52
(Fig. 9.20b) that promotes fracture propagation. Basalt 0.99 3.75
Because the near-tip stress has the same depen- Quartzite 1.31 2.10
dence on distance for all opening fractures, we Marble 0.87 1.49
can redirect our attention to the other quantities Limestone 0.86 1.65
in (9.33) and group them into a new parameter Sandstone 0.34 2.66
called the stress intensity, KI:
Shale 0.17 2.61
K I a a (9.34)
values are for tests conducted at room temperature
Here the subscript I stands for mode I fracture and and atmospheric pressure.
indicates that the relative displacements of the The rule of thumb we take from this data set
fracture surfaces produce an opening motion. is: values of fracture toughness for common rock
Later we consider the other two modes of fracture types tested at room temperature and atmos-
in which the relative displacements of the frac- pheric pressure range from about 0.1 to 4.0 with a
ture surfaces produce a shearing motion. For representative value of 1.0 MPa m1/2. These con-
other fracture geometries and other arrange- cepts have found applications to hydraulic frac-
ments of the applied loads, equations comparable turing of wellbores (Rummel, 1987).
to (9.34) are tabulated in engineering handbooks
(Tada et al., 1973).
The stress intensity is a measure of the magni- 9.3 Brittle failure in a field of
tude of the local stresses anywhere in the fracture
tip region. Laboratory experiments have shown
homogeneous stress
that fracture propagation depends on this local
stress eld and that one can write a propagation Field observations and laboratory tests serve to
criterion in terms of the stress intensity reaching motivate the development of a theory for the
a critical value (Atkinson, 1987). The criterion is: failure of rock samples subject to stress states that
resolve both compression and shear across poten-
K IC K I (at propagation) (9.35)
tial fracture surfaces. It is anticipated that a
Here, K IC is called the critical stress intensity or theory for shear strength will help to explain the
fracture toughness. The units and dimensions of development of shear fractures in laboratory spec-
fracture toughness may be worked out from (9.34) imens and, perhaps, be useful in extrapolating
and are: laboratory data to faults in Earths crust. Given
such a theory, and the appropriate data from eld
fracture toughness [ ] MPa m12
(9.36) observations and laboratory experiments, the
KIC {} M L 12 T2
structural geologist should be in a position to esti-
Fracture toughness is a property that measures mate the magnitude of the stresses at the time of
the resistance of a particular material to the prop- faulting. This would provide a sound physical
agation of a fracture. As such it should be inde- basis for interpreting the geologic history of
pendent of fracture size or geometry, but it may faulted rock masses.
depend on such things as the temperature,
conning pressure, and chemical environment. 9.3.1 Coulombs concept of failure in
A variety of laboratory procedures for measur- shear
ing fracture toughness have been devised (Atkinson Conceptually one might postulate that shear frac-
and Meredith, 1987), and representative values for turing is caused by shear stresses and that the sign
selected rock types are given in Table 9.6. These of the shear stress is irrelevant to the strength. For
an arbitrary polyaxial loading characterized by (a) y s3

the principal stresses (1, 2, and 3) we derived
n
the magnitude of the maximum shear stress in
Chapter 6 as s 12 |1 3 |. This shear stress is
independent of the intermediate principal stress, y
s1
2. We dene the maximum shear strength, Sm, as x
s2
this shear stress at its limiting value:
Potential
Sm s(max),1 3 (9.37) shear
fracture
The planes carrying the maximum shear stress
form an orthogonal pair that intersect along the
intermediate principal stress axis and are bisected
by the 1- and 3-axes. That is, the normals to these
planes lie in the (1, 3)-plane and make angles of (b) sn
45 with the 1-axis. At failure this criterion
would predict the initiation of orthogonal shear
ss
fractures in these two orientations.
The concept of shear failure embodied in (9.37)
is consistent with the uniaxial tensile and com- Potential
pressive strength criteria, (9.8) and (9.9), but it shear
neglects the fundamental insight provided by fracture
Coulomb. Coulomb studied the frictional charac-
Fig 9.21 Schematic illustrations used to derive Coulomb
teristics of materials in the eighteenth century
criterion for shear failure. (a) Rock mass subject to
and hypothesized that shear fracturing was
homogeneous stress state with dashed lines representing
driven by the applied shear stress and resisted by potential shear fractures. (b) Element with side parallel to
a combination of the cohesive strength (adherence) potential shear fracture subject to normal and shear stress,
and the normal compressive stress acting across n and s.
the predicted fracture surface (Coulomb, 1773;
Jaeger and Cook, 1979). Accordingly, the compres-
sive stress plays a role in shear fracture similar to failure begins. As the shear fracture initiates, the
the role of the normal traction in sliding friction homogeneous stress eld would change to a het-
(9.30): greater compression leads to greater resis- erogeneous eld with complex spatial variations
tance to shear fracture, just as greater inward near the shear fracture. The orientation of poten-
directed normal traction leads to greater resis- tial shear fractures (dashed lines in Fig. 9.21) is
tance to sliding. specied by an outward unit normal vector n.
To investigate Coulombs criterion consider a Because of the symmetry of the stress tensor two
rock mass subject to a homogeneous state of stress orientations of potential shear fracture exist.
characterized by the principal stress components, These orientations are symmetric with the princi-
1, 2, and 3. The intermediate principal stress pal stress axes but, as we show below, they are
plays no role in the criterion, so we focus on the orthogonal only in the special case (9.37) where
plane containing the maximum and minimum the normal stress plays no role in failure.
principal stresses and develop the theory in two Coulombs criterion may be stated as a linear
dimensions (Fig. 9.21). In this gure the two prin- relationship between the shear and normal
cipal stresses are compressive, but combinations stresses acting on the surfaces of a volume
of tension and compression are permitted if they element that are parallel to the potential fracture
do not lead to tensile failure. We refer to potential plane (Fig. 9.21):
shear fracture when discussing the Coulomb cri-
terion, because nothing in the criterion explicitly |s | S0 i n,
addresses what happens once the process of 1 Tu (shear fracture initiates) (9.38)
The constant S0 is referred to as the inherent shear

strength because this is the resistance to shear frac-
ture when the normal stress is zero. Because shear
fractures develop independently of the sense of
shearing, the absolute value of the shear stress is
used. The effect of the normal stress is modied by
a constant, i, called the coefcient of internal fric-
tion. It should be understood that surfaces in fric-
tional contact do not exist in the context of the
Coulomb criterion, so the name is misleading.
The linear relationship that relates shear and
normal traction components on sliding surfaces
is given by (9.30). The uniaxial tensile strength, Tu,
is used in (9.38) to restrict the stress state to those
that lead to shear fracture rather than tensile frac-
ture. Because the actual stress state may be poly-
axial, this restriction may have to include all
principal stresses and depend upon a more gener-
alized tensile strength criterion (Bourne and
Willemse, 2001).
In (9.38) and what follows, the effective normal
stress, n n Pp, may be substituted for the
normal stress to account for pore uid pressure as
in (9.24):
|s | S0 i (n Pp),

(9.39)
1 Tu (shear fracture initiates)
Because the normal stress usually is compressive Fig 9.22 Map of aftershocks following the 1979
(negative) and the pore pressure is a positive Homestead Valley earthquake and the Coulomb stress on
number, the addition of pore pressure mitigates planes parallel to the main fault at a depth of 3 km. Reprinted
the effect of the compression. In other words, the from King et al. (1994) with permission of the Seismological
Society of America.
greater the pore pressure the lesser the shear
stress required to induce shear fractures.
The Coulomb criterion (9.38) is used to dene
stress is viewed as the dependent variable and
the Coulomb stress, C:
would be calculated using the shear and normal
C |s | i n, stress acting on the potential shear fracture
(9.40) planes.
1 Tu and C S0
Two other important applications of the
The Coulomb stress may be used to compare dif- Coulomb stress (9.40) are to relate aftershock dis-
ferent potential shear fracture orientations in a tributions to stress changes after major earth-
deforming rock mass and assess which is closer to quakes, and to assess the likelihood of one
failure. These could be differently oriented planes earthquake event triggering another. For example,
at the same location subject to the same principal the spatial distribution of aftershocks following
stress state, or two planes at different locations the 1979 Homestead Valley earthquake (Fig. 9.22)
subject to different stress states. The potential have been shown to correlate with the distribution
fracture surfaces could be in different rock types of Coulomb stress change (Stein and Lisowski,
with different coefcients of internal friction and 1983). Here the contours represent equal values of
different pore pressures. In all cases the Coulomb C on planes parallel to the main fault at a depth
of 3 km. The Coulomb stress was calculated by s3

y 2gC
solving the boundary value problem for an elastic
half-space with a fault that is 5.5 km long and 6 km
high with a maximum slip of 0.5 m tapering to
zero at the fault tip-lines. Two years of aftershock
locations are plotted on the same map and show
n n
a distinct correlation to regions of elevated gC
Coulomb stress change. The locations and fault s1 gC x
plane solutions for earthquakes during the 1984
Morgan Hill seismic activity also show a corre-
lation to Coulomb stress magnitudes and the
orientations of planes on which the maximum
Coulomb stress acts (Oppenheimer et al., 1988). The
locations of earthquakes that may have been trig- Potential
gered by the 1992 Landers earthquake also corre-
shear
fracture
late with regions of elevated Coulomb stress (King
et al., 1994). These and other examples demon-
strate the efcacy of this criterion for interpreting
shear failure at crustal scales in regions of active Fig 9.23 Potential shear fracture planes for the Coulomb
criterion.
faulting.
To carry out these analyses the Coulomb stress
is evaluated in terms of the maximum and
The positive sign is used for in the rst and third
minimum principal stresses, 1 and 3, rather
quadrants and the negative sign is used for in
than the shear and normal components acting on
the second and fourth quadrants. This expression
the potential shear fracture. This is done employ-
is used to determine the Coulomb stress when one
ing Cauchys Formula (6.55):
knows (or postulates) the orientation, , of the
potential shear fracture (Fig. 9.21), the coefcient
n tn 12 (1 3) 12 (1 3) cos 2 of internal friction, and the homogeneous stress
(9.41)
state in terms of the principal stresses or the effec-
s ts 12 (1 3) sin 2 tive principal stresses, 1 1 Pp, 3 3 Pp,
in the presence of pore uids.
Here is the angle between the Ox-axis (here taken
parallel to the direction in which 1 acts) and the
normal, n, to the potential shear fracture (Fig.
9.3.2 Predicting the orientation and
9.21). Substituting these expressions into (9.40) we
initiation of potential shear
have:
fractures
In some applications it is useful to predict the ori-
C | 12 (1 3) sin 2 entation of potential shear fractures. This is accom-
|
plished by determining the orientation of the two
(9.42)
12 (1 3) 12 (1 3) cos 2 planes on which the Coulomb stress is maximized
for a given state of stress. The normals to these
In order to proceed we exchange the absolute planes make angles, C, to the axis of maximum
value sign and the negative sign in the rst term principal stress (Fig. 9.23), and we refer to these as
on the right-hand side of this expression for a the critical Coulomb angles. Taking the derivative of
sign. Using these conditions (9.42) is written: C with respect to the angle using (9.43), we nd:
dC 1
C 12 (1 3)( sin 2 i cos 2) 12 (1 3) i ( 3)( 2 cos 2C 2 i sin 2C) 0
d 2 1
1 Tu and C S0 (9.43) (9.44)
Rearranging to solve for the critical Coulomb Table 9.7. Rock mechanics laboratory tests for
angles: the coefcient of internal friction.
1
C tan 1
2
1
i
(9.45) Rock Type i C ()
The principal stress planes (C 0, /2, , and

Frederick Diabase 1.7 15
Westerly Granite 1.4 18
3/2) are excluded because no shear stress is
Witwatersrand Quartzite 1.0 23
resolved on these planes. Restricting attention to
Bowral Trachyte 1.0 23
the rst two quadrants, and noting that the
Cheshire Quartzite 0.9 24
coefcient of internal friction is positive by
Carrara Marble 0.7 28
denition, the positive sign is associated with a
Gosford Sandstone 0.5 32
critical angle in the range 0 C /4 and the
negative sign puts the angle in the range 3/4
C . with the corresponding critical angle, C, in the
Based on (9.45) the potential shear fractures rst quadrant (Brace, 1964; Hoek, 1965; Jaeger and
are oblique to the principal stress axes and at Hoskins, 1966a, b).
equal acute angles of 45 or less to the direction of The following rule of thumb is consistent with
least principal (most compressive) stress, 3 (Fig. these laboratory results: the Coulomb criterion
9.23). In other words the 3-axis is the acute bisec- predicts shear fractures to form in rock specimens
tor of the potential shear fracture planes. These on planes that are oriented at acute angles
are referred to as conjugate shear fractures. It should between about 15 and 30 to the direction of
be understood that the criterion does not specify maximum compressive stress.
a particular location, only the orientation of these Now that a relationship for the orientations of
fractures. All of the parallel dashed lines drawn in the potential shear fractures is established (9.45),
this gure are equally likely to become shear frac- we can determine the Coulomb stress acting on
tures in a eld of homogeneous stress. The two ori- these particular planes. This is called the critical
entations of potential shear fractures shown in Coulomb stress, CC, to emphasize that it is the
Fig. 9.23 have the opposite senses of shearing: one Coulomb stress acting on planes oriented at the
is right lateral, the other is left lateral. critical Coulomb angles, C, for a given stress state
The orientations of the potential shear frac- and internal friction. Because this value is the
tures (9.45) are independent of the magnitudes of same on both potential shear fractures we can,
the principal stresses. In the context of the without loss of generality, consider only the rst
Coulomb theory these orientations are only quadrant and write (9.43):
dependent on the coefcient of internal friction.
As i 0, C 45 and 135 , so potential shear CC 12 (1 3)( sin 2C i cos 2c) 12 (1 3)i
fractures make angles of 45 with the direction
1 Tu and CC S0 (9.46)
of 3 in this limit. This is the orientation predicted
by the maximum shear stress criterion (9.37). As This expression may be simplied using the fol-
the coefcient of internal friction approaches lowing trigonometric relationships that are valid
zero, the contribution of the normal stress in the rst quadrant (Selby, 1975):
becomes insignicant, so the two criteria are con-
(1 i) 1
sistent in this respect. For i 1, the potential sin 2C
1 (1 i)2 1 2i
shear fractures make angles of 22.5 with the
direction of maximum compressive stress. In (9.47)
1 i
general, as i increases the potential shear frac- cos 2C
1 (1 i)2 1 2i
ture orientations converge on the 3-axis.
Values of the coefcient of internal friction for Substituting these relationships into (9.46) and
seven different rocks are given in Table 9.7 along rearranging, we nd:
CC 12 (1 3)(1 2i)12 12 (1 3) i (a) N

(9.48)
1 Tu and CC S0
Note that the critical Coulomb stress is a function

of the coefcient of internal friction, half the
principal stress difference (equivalent to the
maximum shearing stress), and half the principal
stress sum (equivalent to the mean normal stress
in the plane of interest). As before one may use
the effective principal stresses, 1 1 Pp,
3 3 Pp, to account for pore uid
pressure.
The inherent shear strength is dened as the
critical Coulomb stress reaching a limiting value:
S0 max (CC), (9.49)
This relationship is useful in analyses where the

state of stress is computed using the solution to a
boundary value problem. Given the principal (b) N
stresses at these points, the coefcient of internal
friction, and the inherent shear strength, one can
use (9.48) to compute the normalized critical
Coulomb stress, CC/S0, over the region of interest.
Failure is predicted when this quantity equals
one. By computing the orientations of the princi-
pal stresses and using (9.45), one can predict the
orientations of the potential shear fractures. Then
CC/S0 is contoured over the region of interest,
thereby identifying the areas most prone to shear
failure. This methodology was developed by
Hubbert (1951) and Hafner (1951), and many
others have followed their example for analyzing
the state of stress in Earths crust and the devel-
opment of faults using the Coulomb criterion 2 km
(Sanford, 1959; Couples, 1977; Segall and Pollard,
1980; Bourne and Willemse, 2001; Crider, 2001; Fig 9.24 Comparison of fault orientations interpreted
Guiton et al., 2003). from a seismic reflection survey in the Oseberg Field, North
It should be emphasized that the stress state Sea, and orientations predicted using the Coulomb criterion.
will change as soon as the rst fault initiates. To (a) Major faults (black) and secondary faults (gray) with tick
continue the analysis one must resort to numeri- marks representing secondary fault orientations. (b) Tick
marks representing predicted orientations. Reprinted from
cal solutions of the boundary value problem in
Maerten et al. (2002) with permission of Elsevier.
which the fault is explicitly included as a surface
of displacement discontinuity. An example was
described in Chapter 1 for normal faults from the stress oriented eastwest (direction of crustal
Oseberg Field of the North Sea (Fig. 1.10) extension), the intermediate compressive stress
(Maerten, 2000; Maerten et al., 2002). Based on the northsouth, and the greatest compressive stress
Coulomb criterion one would interpret the major vertical. However, the three-dimensional seismic
faults (broad lines, Fig. 9.24a) as forming in a data reveal other faults with lesser throw (light
regional stress eld with the least compressive gray lines, Fig. 9.24a) cutting the same part of the
sedimentary sequence and trending oblique to s3

the major faults. It was hypothesized that these
faults developed in the same regional stress eld,
but slip on the major faults locally perturbed this Tu s1
eld causing the smaller faults to form in oblique
orientations. To test this hypothesis the geometry Stress states not
of the major faults was used as input for an possible because
elastic model. When subjected to an eastwest of definition s1 >s3
Cu s3 = Cu/2
extension the model faults slipped and the ori-
entations of the principal stresses where com- Possible
puted. A comparison of the strikes of smaller stress
states
faults interpreted from the seismic survey and
3
s
=
the calculated directions of intermediate princi-
1
s
pal stress (Fig. 9.24b) suggests that the Coulomb Stress states
Failure not possible
criterion provides a good correlation at many surface because of
locations. failure
9.3.3 The Coulomb criterion as a failure

Fig 9.25 Principal stress space with two linear failure
surface in stress space surfaces associated with the Coulomb criterion and the
Failure in tension and in shear may be dened tensile strength criterion.
using plots of a failure surface in principal stress
space based on theoretical criteria and these may
Cu
be compared to laboratory data. Consider a two- S0 (9.51)
2[(1 2i)12 i]
dimensional view of principal stress space that
contains the maximum and minimum principal Recall that S0 is the strength of a potential shear
stresses (Fig. 9.25). By denition, 1 3, so the fracture that has no normal traction acting upon
more darkly shaded portion of this gure is off it, a loading condition that is difcult to achieve.
limits. Also, the only restriction on the interme- In principle, one can use the intercept and slope
diate stress is 1 2 3, so this plane can be of the failure surface obtained from experimen-
shifted along the 2-axis accordingly. To plot the tal data to determine Cu and i, and then calcu-
failure surface we set the critical Coulomb stress late S0 using (9.51). In practice this can be
equal to the inherent shear strength, CC S0, problematic because the specimen might fail by
using (9.48) and separate the principal stresses: axial splitting and the failure surface might not
be linear. Nevertheless, values for the inherent
3 2S0[(1 2i)12 i]
(1 2i)12 i

(1 2i)12 i 1
shear strength sometimes are quoted in the liter-
ature. Before applying such values one should be
Cu [(1 2i)12 i]21 (9.50) aware how they were determined, and how well
the laboratory tests conform to the basic postu-
In this linear relationship the rst term on the lates of the Coulomb criterion.
right-hand side is the uniaxial compressive There are restrictions on the extension of the
strength, Cu, the coefcient of the second term is failure surface (Fig. 9.24) into the fourth quad-
the slope of the failure surface which is positive rant. In particular, the intersection of this line
and greater than or equal to one. The region below with the 1-axis is not a measure of the uniaxial
this line (lightly shaded) is off limits because tensile strength, Tu, because the criterion is
shear failure prevents the stress from attaining restricted to potential shear fracture surfaces. To
these values. understand how this restriction may limit the
According to the Coulomb criterion (9.50) the failure surface we substitute the second of (9.47)
uniaxial compressive strength, Cu, is related to the into the rst of (9.41) and postulate that the
inherent shear strength, S0, as: normal stress across the potential shear fracture
must not be tensile. Then, separating the princi- (9.50) to account for pore uid pressure, the re-
pal stresses we nd: lationship remains linear and compares favorably
to data (Fig. 9.13b) from triaxial compression tests
1[(1 2i)12 i] 3[(1 2i)12 i] 0 (9.52) on Berea Sandstone (Handin et al., 1963).
Rearranging (9.50) to put it in a similar form:
1[(1 2i)12 i] 3[(1 2i)12 i] 2S0 9.4 Brittle failure in a field of
(9.53) heterogeneous stress
These two expressions must coincide at the point
in stress space where the failure surface and the The Coulomb criterion, reviewed in the previous
restriction on the stress state are both satised. section, asserts that solids subject to a homoge-
Eliminating 3 by adding (9.52) and (9.53), and neous stress state lose their load-carrying capacity
then substituting for 1 in (9.52) we nd: when a certain combination of shear and normal
stress acting on potential shear fractures reaches
3 S0[(1 2i)12 i] 12 Cu (9.54) a critical value. The Coulomb criterion is cali-
brated using laboratory measurements of the uni-
Thus, the restriction n 0 on the potential shear axial compressive strength and the coefcient of
fracture truncates the failure surface where 3 is internal friction. Although this criterion provides
equal to the negative of half the uniaxial compres- a reasonable t to some laboratory data and has
sive strength (Fig. 9.25). For greater values of 3, it been applied to numerous practical problems
has been suggested that the uniaxial tensile with some degree of success, it does not explicitly
strength would limit the stress state such that address development of the fractures themselves
(Paul, 1961): in the heterogeneous stress state induced by the
fractures. Here we introduce methods for calcu-
1 Tu,for 1 3 12 Cu (9.55) lating the perturbed stress eld around holes and
cracks and use these to investigate brittle failure
In this way the intersection of the failure surface in a eld of heterogeneous stress (Anderson, 1951;
with the (1, 3)-plane is composed of two straight Hubbert, 1951; Chinnery, 1961, 1963, 1966; Cooke
lines. The lightly shaded region (Fig. 9.25) below and Underwood, 2001; Bourne, 2003).
and to the right of these lines is off limits because
of shear or tensile failure. 9.4.1 The boundary value problem of
A number of laboratory studies have compared C. E. Inglis
data on the stress state at failure during triaxial One of the most cited articles in the literature of
tests to the Coulomb criterion. These studies fracture mechanics was presented by C. E. Inglis in
include those that focus on the effects of 1913 to the Royal Institute of Naval Architects in
conning pressure, temperature, deformation England (Inglis, 1913). The title of his article,
rate, and pore pressure on the strength of a wide Stresses in a plate due to the presence of cracks
variety of rocks (Handin and Hager, 1957, 1958; and sharp corners, might seem somewhat
Handin et al., 1963). As an example, we return to abstract for a group of naval architects, but one of
the data selected from a study of the mechanical the principal causes of ship disasters in those days
properties of sedimentary rocks from Tertiary was the growth of fractures in plates making up
basins of Japan (Hoshino et al., 1972). Recall that in the hulls of ships. To address this problem Inglis
this study the rocks were deformed in triaxial solved the elastic boundary value problem for an
compression tests at room temperature with no elliptical hole with major diameter, 2a, and minor
pore pressure. When plotted in principal stress diameter, 2b, in an elastic plate (Fig. 9.26).
space (Fig. 9.12) these data approximate a linear At the time Inglis took up this problem, engi-
relationship of the form (9.50). Using the effective neers knew that holes (e.g. those cut in deck plates
principal stresses, 1 1 Pp, 3 3 Pp, in to make hatches) would alter the local state of
9.4 BRITTLE FAILURE IN A FIELD OF HETEROGENEOUS STRESS 365
r
syy (app) ~ syy pro rata was not a good rule of thumb for these
calculations.
Inglis solved this problem analytically and to
Elastic plate help make the mathematical problem tractable
syy (x = a, y = 0)
y he considered a plate that is innite in extent.
After solving the boundary value problem for the
syy (ave) innite plate it is possible to show that this solu-
tion closely approximates a nite plate that is only
2b x several times bigger than the major diameter of
ts tn the hole. For a remote boundary condition on the
plate, he specied that a uniform normal stress,
2a ryy, acts perpendicular to the major diameter:
BC: as (x2 y2)12 ,xx 0,
xy 0,yy ryy (9.57)
For the internal boundary Inglis specied a trac-

W tion-free condition:
x2 y2
Fig 9.26 Schematic illustration of elastic boundary value BC: on 1,tn 0 ts (9.58)
a 2 b2
problem for an elliptical hole subject to symmetric uniaxial
applied stress (Inglis, 1913). This condition ignores atmospheric pressure.
Inglis solved the governing equations of linear
elasticity theory for the state of stress everywhere
stress, but they disregarded the shape of the hole in the plate. The rather complex equations he
and focused their attention on the cross-sectional derived can be reduced for special cases to gain
area of material outside the hole (Gordon, 1976). physical insight. For example, the normal stress at
For example, consider a uniformly applied stress, the ends of the major diameter (Fig. 9.26) is:
yy (app), acting on the top and bottom edges of a
square plate of side length, W, and thickness, D
(Fig. 9.26). The total force acting on the plate would
yy (x a, y 0) ryy 2a
b
1
(9.59)
be the applied stress times the cross-sectional area Note that the local stress is proportional to the
of the plate, WD. At the mid-section of the plate, y remote stress, ryy, and is related to the shape of
0, the cross-sectional area is reduced to (W 2a)D. the hole through the ratio of major to minor
The engineers did not know how to calculate the diameters. Comparing (9.56) to (9.59) it is clear
stress acting across the mid-section, so they simply that the average stress can be a very poor estimate
dealt with the average value, yy (ave). Equilibrium of the maximum stress. Inglis related this result
of forces requires that the same total force is trans- to the fracture of ship hulls by noting:
mitted across any section of the plate, so
yy(app) WD yy(ave) (W 2a)D. Solving for the When a/b 1,000, the tension at x a, y 0 is 2,001
average stress: times the mean tension. The ellipse in this latter case
would appear as a ne crack, and a very small pull
yy(ave) yy(app) W
W 2a
, y 0 (9.56)
applied to the plate across the crack would set up a
tension at the ends sufcient to start a tear in the
material (Inglis, 1913).
The Latin phrase used by the engineers to describe
this method, and perhaps to obscure their lack This solution to the elastic boundary value
of an exact calculation, was pro rata. In other problem provided, for the rst time, a quantita-
words the ratio of average stress to applied stress tive prediction of the great increase in stress at
was the same as the ratio of plate width to notches, corners, and crack tips and thereby led to
reduced plate width. Inglis was able to show that an understanding of why such geometric features
posed a great danger for engineering structures. cracks with which Mr. Inglis paper deals are math-
According to (9.59) a remotely applied stress that ematical and not real ones (Inglis, 1913).
is well within the normal safety factors for hull
Despite such vacuous criticism, the solution to
plates could be amplied locally by the presence
the boundary value problem derived by Inglis
of a crack to values well above that predicted by
became a cornerstone of one of the most success-
(9.56), and perhaps well above the strength of the
ful engineering endeavors of the twentieth
material.
century, the development of engineering fracture
For practical purposes (9.59) suggests that the
mechanics (Lawn and Wilshaw, 1975; Kanninen
concentration of tensile stress near the ends of
and Popelar, 1985; Anderson, 1995).
cracks in engineering structures can locally break
the material apart and create a propagating frac-
9.4.2 Fracture initiation at Griffith flaws
ture that might lead to catastrophic failure of the
A. A. Grifth addressed the concept of fracture in-
structure. A plausible propagation criterion
itiation in solids in two articles in the early 1920s.
would be that the stress at the fracture tip reaches
He is best known for his analysis of fractures in
the uniaxial tensile strength, Tu, of the material at
terms of a macroscopic energy balance, but in his
that point: yy (x a, y 0) Tu. Substituting
second paper he addressed the local stress con-
this condition into (9.59) and rearranging to solve
centration near the ends of crack-like holes and
for the remote stress at propagation, we have
related this to fracture initiation (Grifth, 1921,

b 1924). Perhaps the most important contribution
ryy Tu , for b a (9.60)
2a of Grifths research was his demonstration that
solids contain sub-microscopic aws that act to
By measuring the uniaxial tensile strength of a increase the tensile stress locally and thereby in-
material in the laboratory (Table 9.1) and knowing itiate tensile fracture growth. The presence of
the geometry of any crack that might exist in the unseen aws was a non-intuitive concept, consid-
structure, an engineer could use (9.60) to place ering that Grifth was working on laboratory
limits on the applied stress in order to prevent glass specimens that appeared nearly awless
further fracture propagation. Although a vast under the microscope.
improvement over the pro rata method, this Grifth began his research with the assertion
approach is somewhat impractical because of the that the analysis of C. E. Inglis (Inglis, 1913) for
difculty of measuring the minor axis of cracks. the state of stress around holes and notches in
This difculty may be addressed by characterizing elastic plates could be used to estimate the intrin-
the stress concentration in terms of the stress sic tensile strength of solids. By intrinsic strength
intensity factor (9.33), which is only a function of Grifth meant the greatest stress that the solid
crack length. could endure before rupture of the bonds
Although the paper of Inglis contains a care- between the atoms, ions, or molecules that hold
fully constructed analysis and a thorough investi- the solid together. Apparently, he envisioned the
gation of the results in practical terms, he met bonds starting to break in the vicinity of a stress
with considerable resistance to his new concepts. concentration created by a aw, so the breaking
A prominent engineer wrote in the proceedings of bonds formed a more-or-less planar displace-
the Royal Institute: ment discontinuity, perhaps along a lattice plane
I regret to say that, although I have studied them with if the solid were crystalline. Grifth used the
Professor Loves book on elasticity at my elbow, I nd stress (9.59) at the end of a very eccentric ellipti-
that the notations, which are doubtless very con- cal hole (b a) in an elastic plate (Fig. 9.26) to
venient for the discussion of vibrations and other com- model the aw and used the radius of curvature,
plicated subjects, are so elaborate that I have not been rc b2/a, to characterize the local shape of the
able to do more than apply some simple tests to the aw:
results, and do not now feel satised that the math-

a 12
ematical deductions are fair representations of practi-
yy (x a, y 0) 2 ryy , for rc a (9.61)
cal cases; in other words, the holes, corners, and rc
Grifth suggested that the appropriate radius of large compared to molecular dimensions. These
curvature in (9.61) would be the intermolecular Grifth aws serve to concentrate the stress and
spacing for the solid, so rc 5 1010 m, and he thereby weaken the material so that the macro-
determined the value of the product ryya12 by scopic measures of strength are much less than
experimentation in order to estimate the intrinsic the intrinsic strength. Fossils, clasts of different
tensile strength. lithologies, pore cavities, grain boundaries, and
Grifth introduced a small scratch of known microcracks are some of the many possible aws
length, 2a, onto the surface of cylindrical tubes or that can provide the stress concentration neces-
spherical bulbs of glass with a glass cutter. Then sary to initiate fractures in rock.
he pressurized the vessels with gas until they
burst, and recorded the ultimate pressure. The 9.4.3 Griffiths criteria for brittle failure
relationship between the gas pressure and the In the second of his two classic papers on fracture
remote stress component acting across the Grifth explored the problem of fracture initia-
scratch, ryy, was determined from the boundary tion and failure under a biaxial remote stress,
value problem for internal pressure in elastic using the solution of Inglis for an elliptical cavity
tubes and spheres. Grifth discovered that the in an elastic material (Inglis, 1913; Grifth, 1924).
product, ryya12 2.63 105 N m32 was nearly Here we review Grifths analysis, but, to make it
constant for the scratch lengths he tested, from 4 more relevant to geological applications, we
to 23 mm. Substituting into (9.61) Grifth evalu- include uid pressure acting on the cavity walls.
ated the stress at the tip of the incipient fracture The limiting case of an elliptical cavity with very
as 2.3 104 MPa and suggested this was an esti- great aspect ratio, a/b 1, is used to model a
mate for the intrinsic strength, Ti. However, this crack-like aw. Flaws of all possible orientations
estimate is very great relative to measured uni- relative to the remote principal stress axes are
axial tensile strengths which typically are of the examined, and the one with the greatest induced
order 101 to 102 MPa (Table 9.1). tensile stress is identied as the most danger-
To understand this puzzling result, Grifth ous. When this tensile stress equals the uniaxial
divided the estimated intrinsic strength by tensile strength the initiation of a tensile fracture
Youngs modulus for glass, E 6.2 104 MPa, to is predicted and this condition is assumed to be
estimate a strain of 0.37 at failure. He realized that coincident with failure of the material. This pro-
Hookes Law probably would not hold at these cedure is used to dene a failure surface in prin-
levels of strain and he knew that the concepts of cipal stress space and that surface is compared to
an elastic continuum were on shaky ground at the laboratory data.
molecular scale. Therefore he inferred that the The boundary conditions at an innite dis-
intrinsic tensile strength would be somewhat less tance from the elliptical hole (Fig. 9.27) consist of
than the value calculated above. Grifth sug- uniformly distributed normal stresses that are the
gested that a reasonable order-of-magnitude value principal stresses, 1 and 3, directed along the x-
was: and y-axes, respectively:
E BC: at x2 y2 ,xx 1,
yy(x a, y 0) Ti 6 103 MPa (9.62)
10
xy 0,yy 3 (9.63)
This value is greater than the uniaxial tensile
Recall from the discussion of elliptical coordi-
strengths for glass specimens, the strongest of
nates in Chapter 2 that constant denes a
which Grifth measured at Tu 1.7 102 MPa, and
family of confocal ellipses with 0 being the par-
many specimens had uniaxial tensile strengths as
ticular ellipse designated as the surface of the
much as two orders of magnitude less than the
hole. The boundary conditions there consist of a
intrinsic strength. These results inspired the con-
uniform normal traction of magnitude P and no
clusion mentioned above: brittle solids such as
shear traction:
glass must contain a myriad of aws, too small to
be detected by the optical microscope, but very BC: on 0, tn P, ts 0 (9.64)
s3
y
t P 2a
b
1 C
2a
b
1
(9.66)
The remote compression is concentrated by a

factor that is proportional to the axial ratio, a/b, as
2b 2a Inglis discovered (9.59). The internal pressure
180o s1 results in a tensile stress concentration also pro-
ts 90o portional to the axial ratio.
jo tn x
h The elliptical hole induces tangential stresses
P
that can be of the same sign as the remotely
270o b applied stress, or of the opposite sign. This result
0o
particularly intrigued Grifth because he was
shh interested in local tensile fracture under a com-
pressive remote stress. To develop a criterion for
failure Grifth approximated (9.65) for a crack-
like aw, b/a 1, and determined the maximum
tangential stress, t(max), as a function of pos-
Fig 9.27 Schematic illustration of elastic boundary value
ition, , using t 0. Then he determined the
problem for an inclined elliptical hole subject to biaxial extreme value of t(max), as a function of the ori-
applied stress (Jaeger and Cook, 1979). entation of the aw, , using t(max) 0. The
aw with this orientation, c, was designated the
most dangerous aw because t(max) would be
The angle is the elliptical coordinate that deter-
the greatest for a given loading condition. Grifth
mines the position on the hole boundary and
found two solutions for the most dangerous aw.
varies from 0 to 2 counterclockwise from the
Under some biaxial loading conditions the most
right-hand tip. The hole may be inclined to the
dangerous aw is inclined to the principal stress
remote principal stress directions and this incli-
axes, 0 c 90; under other conditions the
nation is specied by the angle measured from
most dangerous aw is symmetric to the principal
the major diameter to the Ox-axis.
stresses, c 90, with long dimension perpendic-
Grifth postulated that fracture initiation
ular to the greatest principal stress, 1. The fol-
would occur somewhere along the hole boundary
lowing relationships dene these two cases. If
and that the incipient opening fracture would be
31 3 4P 0, then:
oriented perpendicular to the boundary. Thus he
focused attention on the normal stress compo- ( 1 3) 2
nent, (0), acting tangential to the hole max(t) , inclined aw
40(1 3 2P)
boundary (Fig. 9.27). To simplify the notation we (9.67)
refer to this stress as t (Jaeger and Cook, 1979):
(1 3 2P)2ab If 31 3 4P 0, then:

t P
(a2 b 2) (a2 b 2) cos 2
2(1 P)
(1 3)[(a b)2 cos 2( ) (a2 b2) cos 2] max(t) , symmetric flaw (9.68)
0
(a2 b2) (a2 b2) cos 2
Because these equations contain the shape of the
(9.65)
most dangerous aw, 0, Grifth realized that the
Some familiar results are contained in (9.65). For theory could not be related to the strength of
example, the tangential stress at the tip of an ellip- materials under various loading conditions
tical hole, 0, oriented with long axis parallel unless the shapes of all crack-like aws were
to the x-axis, 0, and loaded by internal pres- known. Without knowing these shapes, his result
sure, P, and remote compressive stress, 3 C, is: had little practical value.
To solve the dilemma Grifth proposed to

eliminate 0 from these equations by relating it to
the uniaxial tensile strength, Tu. Considering the
symmetric case, c 90, he suggested that a lab-
oratory specimen would fail if 1 P Tu (here we
added the internal pressure). Substituting this
relationship into (9.68) we have t(max) 2Tu 0.
This is the predicted stress at the tip of the most
dangerous symmetric aw just as a fracture initi-
ates. Substituting this relationship into (9.67) and
(9.68), the unwanted term 0 is eliminated, leaving
the Grifth criteria for failure. If 31 3 4P 0
then:
(1 3)2 8Tu(1 3 2P) 0, inclined flaw
(9.69)
In this case a crack would propagate away from

the inclined aw oblique to its long axis, forming
a so-called wing crack (Segall and Pollard, 1983;
Cruikshank and Aydin, 1995; Cooke, 1997;
Willemse and Pollard, 1998). Experimental inves-
tigations using plexiglass and glass (Fig. 9.28)
show that the newly created cracks propagate
along curved paths until they are approximately
parallel to 3, and then stop (Brace and
Bombolakis, 1963; Hoek and Bieniawski, 1965). On
the other hand if 31 3 4P 0 then:
Fig 9.28 Photograph of laboratory experiment with three
1 P Tu 0, symmetric flaw (9.70) echelon cracks subject to compression. Wing cracks
propagated from near the ends of these into a direction
In this case a crack would propagate away from approximately parallel to the applied compression (Brace and
the tip of the symmetric aw parallel to its long Bombolakis, 1963).
axis.
Because 1 3, legitimate values of the stress
9.4.4 The Griffith criteria as a failure state plot below the isotropic stress line, 1 3,
surface in stress space thus eliminating the shaded region of Fig. 9.29a
The two conditions (9.69) and (9.70) can be plotted from consideration. Grifth supposed that frac-
on a graph of 1/Tu versus 3/Tu, where we neglect ture initiation would limit possible states of stress
the effects of uid pressure. To construct this to values above the solid curve. By solving (9.69)
failure surface (9.69) is rearranged into the stan- for the special case of a uniaxial compressive
dard quadratic form and solved for 1: stress, where 1 0 and 3 Cu, we nd Cu 8Tu:
the uniaxial compressive strength is predicted to
1 12 (8Tu 23) (9.71) be eight times the uniaxial tensile strength. This
[(8Tu 23)2 4(8Tu3 23)]12 prediction is within a factor of two or three for
most rocks (refer to Tables 9.1 and 9.2), but the
This equation is divided by Tu, and plotted in Fig. discrepancy suggests that the Grifth theory fails
9.29a as the curved part of the failure surface. The to address some important aspects of failure in
other part is the straight line, 1/Tu 1. compression.
(a) s 3 /To surface of an inclined aw. Grifth supposed that

6 5 4 3 2 1 2
this crack would continue to propagate and even-
s1 / To tually break the body into two parts, thereby
2 limiting the state of stress. In fact (Fig. 9.28),
s 1 = To additional compression is necessary for contin-
s3
s 1= 4 ued propagation, so this phenomenon will not
6 necessarily lead to failure of the entire body.
Experiments on sets of aws oriented oblique to
Co/ T o the direction of maximum compression show how
10 shearing induces growth of tensile fractures that
may link to adjacent aws and create a through-
n 12 going structure (Nemat-Nasser and Horii, 1982).
tio
tia
ini 14 Laboratory experiments have been carried out
e
tur in triaxial testing vessels and compared to various
ac
Fr 16 strength criteria (Hoek and Bieniawski, 1965). In
18 Fig. 9.29b the stress state at failure is normalized
by the uniaxial compressive strength for 32 dif-
20
ferent rocks, including sandstones, limestones,
(b) s3/Co shales, marbles, quartzites, gneisses, basalt, and
granite. The solid line on this graph is the original
2.0 1.5 1.0 0.5 Grifth criterion, (9.69) and (9.70), which under-
s1/Co
estimates the strength of most rocks at large com-
pressive stresses. In part, this discrepancy is
1
related to a boundary condition that Grifth over-
s3
s 1= looked. Under compressive stress, crack-like aws
are likely to close, so their walls will be in fric-
2
tional contact. A modied Grifth criterion
Griffith criterion depends on the coefcient of friction, c, and pre-
Modified criterion dicts a somewhat greater strength (McClintock
Laboratory data
3
and Walsh, 1962). Comparing the modied cri-
terion to the strength data one nds that most fall
within the bounds of the criterion for a range of
friction, 0.5 c 1.0 (Fig. 9.29b).
4
In summary, we have shown that the criterion
of strength based on fracture initiation at aws, as
5
proposed by Grifth, and suitably modied to
m = 0.5 m = 1.0 account for closing of aws, is a reasonably good
predictor of laboratory triaxial test data. It gives
Fig 9.29 Plots in principal stress space normalized by
uniaxial tensile strength, Tu. (a) Griffith criteria for failure an explicit value of strength under any biaxial
(Griffith, 1924). (b) Laboratory data compared to Griffith loading condition when calibrated by the uniaxial
criteria and modified criteria to include flaw closure and tensile strength. It also predicts the orientation of
friction (McClintock and Walsh, 1962). Reprinted from fracture, perpendicular to the greatest principal
Jaeger and Cook (1979) with the kind permission of Mrs. stress, for the cases where this fracture initiates at
Jennifer D. Cook. the end of the aw. Most importantly it provides
an explicit mechanism for failure, growth of tensile
If the initial state of stress changed such that fractures from stress concentrations near pre-
its representation as a point on Fig. 9.29a moved to existing aws. Grifth provided an important
the curved portion of the solid curve, Grifths cri- building block for our understanding of strength
terion predicts that a crack would initiate on the and brittle deformation in Earths crust.
9.5 FRACTURE PROPAGATION AND FAULT GROWTH 371
9.5 Fracture propagation and fault 9.5.1 Propagation of joints, veins, and
dikes
growth Fractures are idealized as two surfaces with
mirror image geometry that are in contact in
Criteria for failure in a homogeneous stress the initial unloaded state and are bounded in
state, such as the Coulomb criterion (9.50), extent by a common curve called the tipline.
inform us about the limiting stress conditions Sufciently close to the tipline the shape of this
that are obtained at failure, but do not explicitly curve is approximately straight and the surfaces
include the structure (e.g. a shear fracture) that are approximately planar. We adopt a Cartesian
is associated with the process of failure. Criteria coordinate system with the y-axis normal to the
for failure in a heterogeneous stress state, such plane of the surfaces and the z-axis parallel to the
as the Grifth criteria, (9.69) and (9.70), include tipline (Fig. 9.30). Upon loading of the elastic
the stress concentrating structure and describe body the fracture surfaces move relative to one
the initiation of cracking, but do not address the another and this motion may be classied
propagation of the crack or other possible events according to the coordinate directions we have
and mechanisms that may be involved in the chosen (Kanninen and Popelar, 1985). Relative
evolution of structures such as joints and faults. motion in the y-coordinate direction is referred
By combining solutions to elastic boundary value to as Mode I or the opening mode; relative motion
problems with principles of fracture mechanics in the x-coordinate direction is mode II or the
one can explore the processes of fracturing and sliding mode; and relative motion in the z-coordi-
faulting in rock from the initiation stage through nate direction is mode III or the tearing mode.
a stage of propagation or development to the Mode I is associated with geologic structures
eventual cessation of tectonic activity as the such as joints, veins, and dikes, whereas modes II
structure attains the size and conguration we and III are associated with shear fractures and
observe in exposure today. faults. Modes II and III both involve a shearing
The evolution of structures in brittle rock is a motion of the surfaces with the former being per-
large topic with interesting examples that are too pendicular to the tipline and the latter being par-
numerous to be described in detail here. Instead allel to the tipline. This classication may appear
we focus on a few examples that are meant to arbitrary, but because of the symmetry of the
provide a summary of the methodology of inves- fracture tip the elastic stress elds near the tip
tigation and some insight concerning the kinds of are uniquely distinguished by these modes, and
results one might expect. The rst example is the each mode is associated with a different style of
propagation of opening fractures such as joints, propagation.
veins, and dikes. Earlier in this chapter we There are many solutions in the literature for
described the loading conditions necessary for linear elastic problems that involve fractures (Sih,
propagation in terms of the stress intensity at the 1973; Tada et al., 1973). A common feature of these
fracture tip reaching the fracture toughness of is that the stress components in the vicinity of the
the rock (9.35). Here we show how the loading con- fracture tip are distributed in a way that depends
ditions determine the path that the fracture largely upon the fracture mode (Irwin, 1957;
follows as it propagates. The second example is the Williams, 1957). In contrast, the magnitudes of
growth of faults in granite and in sandstone. In the stress components depend upon the fracture
the granite it is the propagation of opening frac- geometry away from the tipline and the loading
tures to link adjacent sheared joints that enables conditions. This interesting and perhaps non-
the faults to grow in length. In the sandstone it is intuitive fact comes about because the near-tip
the clustering of deformation bands that enables stress distributions are dominated by the local
the fault to grow in thickness. Through these geometry of the fracture tips, which are taken as
examples of fault growth we make the point that identical for all modes, and by the relative motion
faulting is a process that can involve several dif- of the fracture surfaces, which are uniquely dis-
ferent physical mechanisms. tinguished by the modes (Fig. 9.30).
(a) (b) (c)

r y y
y
x x
x
u
z z
z
Mode I, opening Mode II, sliding Mode III, tearing
Fig 9.30 Three modes of fracture. (a) Opening mode I.

(b) Sliding mode II. (c) Tearing mode III. Reprinted from The term K II is the mode II stress intensity factor.
Kanninen and Popelar (1985) with permission of Oxford For pure mode III fracture tips, the stress compo-
University Press. nents are approximated as:
To examine the stress distributions in the

vicinity of the fracture tip it is convenient to use a
xz
yz

K III
sin (2)
(2r)12 cos (2) (9.74)
polar coordinate system (r, ) with origin at the tip The term KIII is the mode III stress intensity factor.
(Fig. 9.30). What we mean by the vicinity of the One may summarize the equations for the near-
fracture tip (Fig. 9.20) is a distance r 0.01a, tip stress components as follows:
where 2a is a characteristic length of the fracture ij [KI fij() KIIgij() KIIIhij()](2r)12 (9.75)
(Pollard and Segall, 1987). When analytical solu-
tions for the plane strain stress components are Here the indices i and j range over x, y, and z. This
expanded about the fracture tip and higher-order clearly demonstrates the separation of stress mag-
terms in powers of r are eliminated, what remains nitude (intensity) from stress distribution. Note
is proportional to r 1/2 and contains trigonomet- that several of the trigonometric functions in
ric functions of of order unity. For pure mode I (9.75) are zero.
fracture tips, the stress components in the (x, y)- Because the stress intensity and the inverse
plane are approximated as: square root of the radial distance are common to
all stress components (9.75), the components are

xx cos (2)[1 sin (2) sin (32)] most easily compared by plotting the trigonomet-
yy KI cos (2)[1 sin (2) sin (32)] ric functions over the range , from
xy (2r)12 sin (2) cos (2) cos (32) one fracture surface around to the other (Fig.
(9.72) 9.31). Positive values are associated with tensile
K I is the mode I stress intensity, which has units and normal stresses. Some instructive results con-
dimensions given in (9.36) and has a value that cerning geologic structures are found in these
depends upon the fracture geometry and loading plots. For mode I the function fyy is proportional to
conditions. Stress intensity factors are tabulated the normal stress component, yy, acting perpen-
in engineering handbooks (Tada et al., 1973). dicular to the plane of the opening fracture. One
For pure mode II fracture tips, the components might expect this stress component to have a
are approximated as: maximum value just ahead of the fracture tip in
the plane of the fracture, 0. Instead, yy has

xx sin (2)[2 cos (2) cos (32)]
two equal maxima to either side of the fracture
yy KII sin (2)[cos (2) cos (32)]
(2r)12 cos (2)[1 sin (2) sin (32)] plane (Fig. 9.31a). These maxima may, in part,
xy explain the cloud of microcracks that develop
(9.73) around the opening fracture tip in the laboratory
experiments on Salem Limestone and Berea

Sandstone illustrated schematically in Fig. 9.18.
(a) Mode I
The dual maxima for fyy (Fig. 9.31a) provide an
fyy
explanation for the secondary deformation
observed where igneous dikes have propagated to
within a few tens of meters of the ground surface fxx
in volcanic rift zones (Fig. 9.32). This map records
open vertical cracks and normal faults in the
Keanakakoi ash deposit of the Kau Desert in the fxy
southwest rift zone of Kilauea Volcano caused by
the eruptive event of December 31, 1974 (Pollard et
al., 1983). Lava erupted from 85 echelon ssures
forming a set that extend about 4.5 km along the
rift. The map shows the northeast end of one
ssure with the lava that issued from it and (b) Mode II
owed primarily to the south. Lava from the next
ssure to the northeast also owed to the south
and merged with the lava from this ssure but did
not cover the adjacent Keanakakoi ash. In the ash gxy
deposit a set of open vertical cracks formed at the
time of the eruption, which are approximately
parallel to the ssure and cluster into two groups,
one to either side of the projection of the ssure
gyy
to the northeast. We interpret these two clusters gxx
of cracks as forming because of the dual maxima
in the normal stress component, yy (Fig. 9.31a).
Here the upward propagating dike is the mode I
fracture and this normal stress component would
be parallel to the ground surface and perpendicu-
lar to the cracks.
The distribution of gxx (Fig. 9.31b) for the mode
II fracture provides an explanation for the sec- (c)
ondary structures mapped in association with Mode III
small left-lateral faults in limestone from the
Languedoc region of southern France (Fletcher
and Pollard, 1981; Rispoli, 1981). We described
hyz
these structures in Chapter 1 (Figs. 1.13, 1.14)
where it was pointed out that opening veins and
closing solution surfaces form on opposite sides of
the faults near their tips. For mode II fracture, the
function gxx is proportional to the normal stress hxz
component, xx, acting parallel to the plane of the
sliding fracture. This stress component has a
maximum (greatest tension) on one surface of the
fracture, , and a minimum (greatest com-
pression) on the other surface, . Consider- Fig 9.31 Plots of spatial variation with angle of near
ing the small left-lateral faults to be approximated fracture tip Cartesian stress components. (a) Mode I.
by mode II fractures, extreme values of the normal (b) Mode II. (c) Mode III. Reprinted from Lawn and Wilshaw
(1975) with permission of Cambridge University Press.
stress with opposite signs on the adjacent surfaces
Cracks
Fissure
Cracks
Fig 9.32 Map of open fissures and normal faults in vicinity of the fracture where it is perturbed, and
Keanakakoi ash deposit associated with the 1974 fissure this perturbation is greatest in the near-tip stress
eruption on the southwestern rift of Kilauea Volcano. eld as dened in (9.75). Before the fracture devel-
Numbers are horizontal (H) and vertical (V) separations in ops, the stress eld is uniform throughout the
millimetres across the cracks. Reprinted from Pollard et al. body and equal to the remote stress. The symme-
(1983) with permission of Elsevier.
try of the opening displacements for straight
joints and dikes suggests that these fractures form
is consistent with the formations of veins and solu- symmetrically with respect to the remote princi-
tion surfaces there. Because the graph in Fig. 9.31b pal stress axes. The greatest mode I stress intensity
is for right-lateral slip, one must change the sign is achieved if the fracture is perpendicular to the
of gxx to apply the result to the Languedoc faults. least compressive (most tensile) remote principal
The traces of many joints, veins, and dikes in stress, so that is the preferred orientation.
outcrop are remarkably straight (Fig. 2.12) or Furthermore, the near-tip stress eld tends to
gently curved (Chapter 9, frontispiece). In a rock guide the fracture tip into an orientation that is
mass that is isotropic with respect to fracture symmetric with the remote principal stress axes
toughness, the direction of propagation for these (Cotterell and Rice, 1980).
mode I fractures is determined by the stress eld, For the pure opening fracture (Fig. 9.33a) there
and the systematic traces suggest that the princi- is no resolved remote shear stress on the plane of
pal stress axes were uniformly oriented or the fracture. For the near-tip stress eld (9.72), we
smoothly varying on the length scale of the frac- have K I 0, but fyx 0 at 0 (Fig. 9.31a) so the
tures. Here we distinguish two regions, one far near-tip shear stress yx 0 on the next increment
from the fracture and the other very near the frac- of growth in the plane of the fracture. Under these
ture tip relative to the in-plane fracture length. conditions the mode I fracture is predicted to
We describe these as the remote stress eld and propagate along a straight path. However, if this
the near-tip stress eld, respectively. The remote fracture propagates straight into a region with
stress eld is thought of as uniform, except in the differently oriented remote principal stress axes,
(a) (b) (c)

W
No Resolved Resolved
resolved mode II mode III
shear shear shear
Fig 9.33 Schematic illustrations of fracture propagation

K I sin (0) K II[3 cos (0) 1] 0 (9.78)
path for dominantly opening mode fractures. (a) Pure mode I.
(b) Mixed modes I and II. (c) Mixed modes I and III.
Reprinted from Pollard and Aydin (1988) with permission of The angle 0 is the predicted orientation of the
The Geological Society of America. next increment of fracture propagation.
For pure mode I loading (K II 0), the fracture is
predicted to propagate in its established plane,
some shear stress, yx and/or yz, would be resolved 0 0, according to (9.78). Right-lateral shearing of
on the fracture plane. Here we consider the case the fracture is associated with a positive K II which
where a remote shear stress, yx, induces a non- corresponds to a negative 0 and a clockwise
zero shearing mode stress intensity, K II. For the turning of the fracture. If the introduction of
near-tip stress eld (9.73), we have K II 0 and mode II loading is associated with propagation
gyx 1 at 0 (Fig. 9.31b) so the near-tip shear through a smoothly varying stress eld, the frac-
stress yx 0 on the extension of the fracture ture may follow a curved path. On the other hand
plane (the patch bounded by dashed lines in Fig. if the fracture does not propagate while
9.33b). signicant mode II loading is added, eventual
We postulate that the next increment of frac- propagation may occur at a sharp angle to the
ture propagation is parallel to the radial plane former fracture plane. For example, under pure
carrying the greatest local circumferential stress, mode II loading (K I 0), the fracture path is pre-
(max), (Erdogan and Sih, 1963). The circumfer- dicted to take a sharp kink and propagate at
ential stress in the near-tip eld for mixed mode angles of 0 70.5. Here the negative sign cor-
III loading is found by transforming (9.72) and responds to a positive KII. In this way the geometry
(9.73): of fracture traces at exposure may be used to infer
the loading conditions during fracture propaga-
tion (Pollard and Aydin, 1988; Olson and Pollard,
(2 r)12 cos ( 12 ) K I cos 2 ( 12 ) 2 K II sin ()
3
1989; Olsen, 1993; Willemse and Pollard, 1998;

(9.76)
Kattenhorn et al.,2000).
The extreme values of this stress component are The postulates leading to the prediction of
found by differentiating (9.76) with respect to fracture paths in mixed mode III loading using
and setting the result to zero: (9.78) have been tested in controlled laboratory
experiments, some of which focus on kinked
cos ( 12 ) {K I sin () K II[3 cos () 1]} 0 (9.77) paths (Erdogan and Sih, 1963) and others on
curved paths (Thomas and Pollard, 1993). Here we
This equation has two solutions where , but describe biaxial tests of curved paths in thin
these correspond to zero values on the stress-free sheets of polymethyl methacrylate (PMMA), a
fracture surfaces which are of no interest. The rel- transparent and nominally isotropic plastic called
evant solution for fracture propagation is found plexiglass. Two narrow, parallel slots were milled
where the terms in braces sum to zero (Erdogan into the PMMA with echelon geometry, a constant
and Sih, 1963): parallel separation of 19 cm, and a variable
perpendicular spacing of 1.0, 3.0, and 6.0 cm (Fig. (a)

-
9.34). The biaxial remote loading was a tension
acting perpendicular to the slots, and a stress
acting parallel to the slots that was an equal
tension (all around tension, ATT), or zero stress
(uniaxial loading, UNI), or a compression of equal
(b) -
magnitude (crack parallel compression, CPC). A
small crack was notched into the proximal tips of
the slots to initiate fracture propagation.
Typically the fracture paths (dashed curves, Fig.
9.34) are nearly straight and perpendicular to the
remotely applied tensile stress (parallel to
the starter slot) until the fracture tip entered the
stress eld perturbed by the second slot. For
the smallest spacing the mechanical interaction (c) -
between the fracture and slot is greatest and the
paths have the greatest curvature. The mechani-
cal interaction is most pronounced for the AAT
loading and least for the CPC loading so the paths
for a given spacing have the greatest curvature for
the AAT loading. When mechanical interaction
dominates (lesser spacing and AAT loading) the
fracture path rst turns away from the slot and
then turns toward it. When the remote loading
dominates (greater spacing and CPC loading) the
paths are nearly straight and perpendicular to the
applied tensile stress.
The laboratory experiments were modeled
(Thomas and Pollard, 1993) using a numerical
computer code and the Boundary Element Fig 9.34 Comparison of fracture propagation paths in
Method (Crouch and Stareld, 1983) which is laboratory experiments using PMMA (dashed curves) and
based on linear elastic theory. The sample bound- paths predicted by solutions to elastic boundary value
ary was subject to displacement and/or traction problems (solid curves). Loading conditions indicated by
boundary conditions to match those applied by inset. (a) 1-cm initial spacing. (b) 3-cm initial spacing. (c) 6-cm
the testing machine. The slots were modeled as a initial spacing. Reprinted from Thomas and Pollard (1993)
set of boundary elements with traction-free con- with permission of Elsevier.
ditions. The biaxial remote loading on the sample

was prescribed according to one of the ratios (ATT, experimental curves, suggesting that the maxi-
UNI, or CPC) and the elastic boundary value mum circumferential stress provides a reasonable
problem was solved. The loading was increased criterion for the direction of continuous opening
until the stress intensity factors, K I and K II, fracture propagation in mixed mode III loading
satised the fracture criterion. Then a new bound- conditions.
ary element was added to the current end of the Examples of fracture paths that are similar to
fracture in the orientation 0 determined by those from the laboratory experiments in PMMA
(9.78). That element perturbs the local stress eld (Fig. 9.34) are found for a variety of length scales
so the boundary value problem was solved again and structures (Fig. 9.35). Traces of cracks in glass
and the next increment of fracture path deter- at a 25-m scale, hydrothermal veins in granitic
mined. The numerical fracture paths (solid rock at a 25-cm scale, basaltic dikes in shale at a
curves, Fig. 9.34) are remarkably similar to the 250-m scale, and oceanic ridges along the East
(a) eld (9.74), we have K III 0 and hyz 1 at 0 (Fig.

9.31c), so the near-tip shear stress yz 0 on the
extension of the fracture plane (the patch
bounded by dashed lines in Fig. 9.33c). We postu-
25 m
late that the next increment of fracture propaga-
tion is perpendicular to the direction of greatest
Crack local tensile stress for mixed mode IIII loading
(Pollard et al., 1982; Olson and Pollard, 1991). This
(b) direction denes a set of planes that contain the
y-axis and are rotated about this axis through an
angle 0. With continued propagation the frac-
25 cm ture breaks down into a set of echelon fractures
that extend from the tipline where the mode III
Vein
loading was introduced (Fig. 3.20, 3.22). Natural
examples include echelon dike segments, echelon
(c) vein segments, and hackle on joints (Woodworth,
250 m 1896; Nicholson and Ejiofor, 1987; Pollard and
Dike Aydin, 1988; Cooke and Pollard, 1996).
We have focused on the propagation of a single
opening mode fracture and the resulting geome-
try and surface textures. However, there are sys-
(d) tematic relationships among the members of an
opening mode fracture set (Renshaw and Pollard,
1994; Wu and Pollard, 1995; Renshaw and Park,
1997; Renshaw, 2000). For example, many studies
Ridge 2.5 km
have shown that joints in sedimentary rocks may
have a regular spacing that is linearly related to
Fig 9.35 Examples of paths of dominantly opening the thickness of the jointed unit (Narr and Suppe,
fractures showing mechanical interaction in different 1991; Gross, 1993; Gross et al., 1995). Linear elas-
materials at different scales (Pollard et al., 1982). (a) Glass at ticity has provided important insights about this
25m. (b) Granite at 25 cm. (c) Shale at 250 m. (d) Basalt at
phenomenon through the investigation of the
2.5 km.
stress distribution between two adjacent opening
mode fractures in the middle layer of a three-layer
elastic model subject to extension (Bai and
Pacic Rise at a 2.5-km scale all appear to have Pollard, 2000; Bai et al., 2000). These models reveal
followed curving, hook-shaped paths that are that the stress changes from tensile to compres-
characteristic of opening fracture propagation sion when the ratio of fracture spacing to layer
under conditions of signicant mechanical inter- thickness falls below a critical value of about one.
action in a material that is nominally isotropic This stress transition denes the condition of frac-
with respect to fracture toughness and elastic ture saturation: continued extension of the layers
properties (Pollard and Aydin, 1984). Some of is accommodated by fracture opening rather than
these even display such subtle effects as the the initiation of new fractures.
paths rst turning away from the neighboring We conclude that the propagation of opening
fracture. fractures in brittle rock is a process that can be con-
As implied by Fig. 9.33c the propagation of an ceptualized in terms of the increase in area of the
opening fracture is perturbed in a more complex two fracture surfaces as the fracture tipline
way if a remote shear stress, yz, is resolved on the advances. Particles on these surfaces that once
fracture plane and induces a non-zero tearing were bonded together are separated by a displace-
mode stress intensity, K III. For the near-tip stress ment discontinuity and each increment of new
(a) (b) (c)
Unfaulted Fault zone

joint
Joints
Left step
Dike Splay fractures Single
faults
Right step
Fig 9.36 Schematic illustration of development of faults in

granitic rocks of the Sierra Nevada (Segall and Pollard, 1983; 1999). These joints are organized into domains
Martel et al., 1988). (a) Joints propagate to form one set of with horizontal dimensions of several tens to a
opening fractures. (b) Some joints slip to form left-lateral few hundreds of meters within which the strikes
faults with wing cracks near tips linking to neighboring faults. vary by only a few degrees, but between which the
(c) Adjacent left-lateral faults form boundaries of fault zones. strikes can differ by as much as 20. Over glaciated
Reprinted from Brgmann et al. (1994) with permission of outcrops and cliff exposures tens of meters in
Elsevier.
extent the traces of individual joint segments are
nearly straight, indicating that they are approxi-
surface area is oriented approximately perpendicu- mately planar in three dimensions. The joint seg-
lar to the local maximum tensile stress. This simple ments are arranged in echelon patterns (Fig.
conceptual model may require modication to 9.37a) and trains of segments interpreted as a
explain features of some opening fractures such as single joint may be several tens of meters in
a process zone of microcracks (Fig. 9.18), or the length. Individual joints are up to a centimeter in
development of plastic deformation beyond the thickness and spacing between adjacent joints
near-tip region, or the propagation of fractures in ranges from a few decimeters to a few tens of
non-elastic materials, but tools exist in the litera- meters.
ture of fracture mechanics to investigate these phe- The lled joints cross-cut older aplite dikes,
nomena (Kanninen and Popelar, 1985; Renshaw the margins of which were separated in a direc-
and Harvey, 1994; Anderson, 1995). tion perpendicular to the plane of the joint. These
small normal separations are clearly distin-
9.5.2 Growth of faults in granite and guished from left-lateral offsets (Fig. 9.37b),
sandstone ranging from a few millimeters to 2 m, that doc-
In contrast to opening fractures, the development ument the next event in the development of the
of faults apparently is a more complicated process faults. The hydrothermal minerals within the
that can involve several different physical mecha- faults have acquired a mylonitic fabric (Fig. 9.37c),
nisms which are not readily conceptualized in whereas those mineral grains within nearby
terms of the propagation of a fracture tipline and unsheared joints are not deformed (Segall and
the increase in area of two surfaces that formerly Pollard, 1983; Segall and Simpson, 1986). The
were bonded together. As an example we consider presence of these two parallel structures, joints
the growth of faults in granitic rock of the Sierra and faults, containing the same mineral assem-
Nevada, California, in the Late Cretaceous (Segall blage in the same exposure suggests that the
et al., 1990). The oldest fracturing event, probably faults once were members of the joint set. This
related to cooling of the Lake Edison Granodiorite, interpretation is further supported by the fact
was the formation of a set of nearly vertical east- that joints exist in some exposures without
northeast-striking joints (Fig. 9.36a), which are faults, but exposures containing faults always
lled primarily with epidote and chlorite and include joints. We interpret the faults as sheared
quartz deposited from a hydrothermal uid joints rather than shear fractures. Sheared
(Segall and Pollard, 1983; Bergbauer and Martel, joints develop when a slip event nucleates and
(a) (b)
(c) (d)
(e) (f)
Fig 9.37 Photographs at exposures depicting stages in

Apparently, the slip events that propagated
growth of faults in granitic rock. (a) Joints sealed with
along the former joints were unable to nucleate a
epidote, chlorite, and quartz. (b) Sheared joints become left-
lateral faults and offset older aplite dike. (c) Hydrothermal
shear fracture in the unbroken granite when the
minerals in sheared joints have a mylonitic fabric. (d) slip reached the tip of a joint segment. Instead, an
Extensional step between two fault segements. (e) Fault zone opening fracture, called a splay fracture, nucle-
between adjacent left-lateral faults with one set of inclined ated on one side of the sheared joint near the tip
fractures linking boundaries. (f) Compound fault zone with and propagated in an oblique direction counter-
about 100 m of left-lateral offset. Photograph by D. D. Pollard. clockwise from the trace of the fault (Fig. 9.36b).
These splay fractures are explained using the
propagates along a pre-existing joint, separating stress distribution near the tip of a mode II frac-
particles with a shearing displacement disconti- ture (Fig. 9.31b). The normal stress, xx, acting par-
nuity that were separated originally with an allel to the fracture is greatest in magnitude on
opening displacement discontinuity. the fracture surface where and is tensile
(a) (b) (c) Slip

Offset surface
marker
DB
0 1 2
m
Fig 9.38 Schematic illustration of development of faults in
porous sandstones of the San Rafael Desert (Aydin, 1978; faults with oblique fractures (Martel et al., 1988;
Aydin and Johnson, 1978). (a) Single shearing deformation Martel and Pollard, 1989). These fault zones typi-
band (DB) offsets marker horizon by a few millimeters. (b) cally are 0.5 to 3 m in width, reecting the spacing
Deformation bands cluster into a zone with offset of several inherited from the original joint set. The most
centimeters. (c) A slip surface develops on one margin of a
prominent and earliest formed fractures within
zone with offset of several meters. Reprinted from Davatzes
the zone strike at an acute counterclockwise
and Aydin (2003) with permission of Elsevier.
angle to the two bounding surfaces of the zone
and do not cross-cut those boundaries (Fig. 9.37e).
on one surface and compressive on the other. The Left-lateral offset of older structures is localized
splay fractures extend no more than a few meters on the former faults that bound these zones and
from the fault surfaces and near their ends strike can be as great as 10 m. In contrast to the
from 15 to 60 counterclockwise from the trend of mylonitic textures found in the faults (Fig. 9.37c),
the faults. the same epidote, chlorite, and quartz assemblage
Where two echelon fault segments are in the boundary faults is characterized by cata-
arranged with a left step (looking toward the tip, clastic textures. The fault zones can be up to 1 km
the next segment is to the left), splay fractures in length and are composed of segments a few
emanating from both segments may link to the tens of meters in length joined end-to-end at steps
adjacent segment and some may open to form and bends. This segmentation reects the geome-
rhomb-shaped cavities up to 10 cm wide (Fig. try of the earlier formed joint set and its domains.
9.37d). These extensional steps range from a few The granitic rock outside the two bounding sur-
centimeters to approximately 1 m in width faces of the fault zone is fractured near the steps
(normal distance between the segments) and are and bends, but elsewhere the fracturing is strictly
up to 2 m in length (overlap of the segments). The conned to the zone. Many of the internal frac-
slip on fault segments may be transferred to adja- tures show evidence only for opening but some
cent segments in a succession along strike by this also are sheared. Right-lateral deformation was
linkage mechanism such that faults more than accommodated on kink bands (Davies and
100 m in trace length are formed. The length of Pollard, 1986). The nal stage identied in the
faults apparently increased not by shear fracture development of the faults is the side-to-side
propagation, but rather by the end-to-end linkage linkage of adjacent fault zones and faults with
of echelon segments through opening splay frac- oblique fractures to form a compound fault zone
tures in extensional steps. At some locations near (Fig. 9.37f) about 10 m in width and several kilo-
the contact with the younger Mono Creek Granite, meters long (Martel, 1990). These zones offset
right steps display well-developed ductile fabrics older structures by as much as 100 m in a left-
with a foliation oriented oblique to the fault seg- lateral sense.
ments and these also may serve to transfer slip The second example considers fault develop-
(Brgmann and Pollard, 1992, 1994). ment in porous sandstone (Aydin, 1978; Aydin and
The next stage in the development of faults in Johnson, 1978). These faults were identied,
the Lake Edison Granodiorite (Fig. 9.36c) involved mapped, and described from outcrops of Entrada
the side-to-side linkage of parallel and adjacent and Navajo Sandstones in the San Rafael Desert of
(a) (b)
(c) sequential development. Slip surfaces only exist

in association with zones of deformation bands
yet some zones lack a slip surface. Similarly, zones
are always composed of individual deformation
bands, but individual bands are commonly found
isolated from any neighboring bands. From these
systematic occurrences it was deduced that indi-
vidual bands formed rst; these occasionally clus-
tered into zones; and some zones became the
locus of slip surfaces.
The rst stage in the process is the growth of
shearing deformation bands (Fig. 9.38a) across
which there are relative displacements parallel to
the band of a few millimeters to a few centimeters
(Fig. 9.39a). There is no evidence of discrete sliding
surfaces within the bands, nor is there evidence of
pre-existing structures such as joints that could
Fig 9.39 Photographs of exposures depicting stages in have served to localize the deformation. Perhaps
growth of faults in sandstone (Aydin, 1977). (a) Single these structures propagate as mode II fractures
deformation band in Entrada Sandstone. (b) Zone of (Petit and Barquins, 1988). The deformation
deformation bands with sub-parallel bands in vertical section within the bands includes both the relative
and anastamosing bands in horizontal section. (c) Zone of
motion of sand grains and the fracturing of grains
deformation bands with slip surface on the left margin that
offsets Entrada Sandstone beds by about 7 m. Photographs
accompanied by a pronounced decrease (60 to
reprinted with permission of A. Adyin. 75%) in porosity. A typical deformation band is
tabular in shape, up to a few millimeters in thick-
ness and tens of meters in outcrop length. Single
southeastern Utah (Aydin, 1977). There are three bands are organized into sub-parallel sets.
structural elements found in the exposures: indi- The second stage in the development of faults
vidual shearing deformation bands, zones of in these sandstones is the clustering of two or
these bands, and slip surfaces (Fig. 9.38). The more individual deformation bands into a zone
spatial relationships among these elements (Fig. 9.38b). The deformation bands within a zone
provide the evidence necessary to interpret their have approximately the same orientation as the
zone itself but locally vary in orientation (Fig. together. It is conceivable that the slip surfaces in
9.39b). The traces of bands are nearly straight on sandstone propagate in this way, but they require
cross sections parallel to the direction of offset, the previous development of the zone of defor-
but form a wavy pattern on cross sections perpen- mation bands in order to nucleate.
dicular to the offset. Lenses of sandstone may be The development of faults in granitic rock is
preserved between wavy bands but the bands not expected to conform to the example presented
rarely appear to cross one another. The thickness here unless a pre-existing set of weak surfaces is
of a zone increases simply by the addition of defor- present. If such anisotropy is absent, or if more
mation bands. The relative shearing displacement than one set exists, we would anticipate a different
across a zone is the sum of that across the outcome. For porous sandstone the sequence from
members and may be as great as 25 to 30 cm for a deformation band to slip surface described above
zone containing 100 bands. is not the only possibility. For example, in the pres-
The third stage in the development of these ence of a set of pre-existing joints the mechanisms
faults is the localization of discrete surfaces (Fig. involved in the evolution of faults in sandstone is
9.38c) with slickenlines and striations indicating quite different (Flodin and Aydin, 2004; Meyers
slip. These slip surfaces are found on the margins and Aydin, 2004). Furthermore, for a different
of thick zones of deformation bands (Fig. 9.39c) lithology, such as interbedded limestone and
and they accommodate several decimeters to shale, the mechanisms are different than those
several meters of offset. There is a reduction in described for granite or sandstone (Peacock, 1991;
grain size and porosity within a few millimeters Peacock and Sanderson, 1991, 1994; Willemse et al.,
of the slip surfaces and the surfaces themselves 1997; Cooke, 1997). Some phenomena described
may be highly polished relative to the sandstone. here may be explained using quasi-static elastic
Investigations at other localities suggest that slip models; others may require solutions for dynamic
surfaces nucleate in small patches which link to elastic problems (Rice, 1980; Poliakov et al., 2002).
form an anastomosing network (Shipton and
Cowie, 2001).
This review of the physical mechanisms
involved in faulting of granitic rock in the Sierra
Nevada (Fig. 9.36) and of sandstone in the San The dominant behavior of rock in Earths upper
Rafael Desert (Fig. 9.38) serves to make the point crust is elastic and brittle at scales that range from
that faulting is a more complicated process than that of mineral grains to the crust itself. Brittle
the development of opening fractures such as the deformation is manifest in rock structures includ-
veins or dikes described in the previous section. ing microcracks, joints, veins, dikes, deformation
Multiple physical mechanisms are involved and bands, compaction bands, and faults. In the lab-
these may differ depending upon the rock type oratory, extension and shear fractures form as
and tectonic setting (Brgmann et al., 1994). The loading conditions reach the strength of samples.
fault zones in granite begin to form with rela- While these laboratory experiments do not neces-
tively little internal deformation except for a sarily reproduce the mechanisms responsible for
single set of oblique fractures. The fault zones in outcrop-scale structures, they do provide impor-
sandstone only form by the accumulation of tant data regarding strength, friction, and frac-
bands accounting for considerable internal defor- ture toughness. The strength of rock samples
mation. The thickness of zones in granite is deter- typically increases with increasing conning pres-
mined by the original spacing of joints whereas sure and decreases with increasing pore pressure.
the thickness of zones in sandstone is determined The fact that tensile strength is about one order of
by how many deformation bands cluster together. magnitude less than compressive strength helps
In neither case is the growth of these faults to rationalize the abundance of opening fractures
readily conceptualized in terms of the propaga- in the crust despite the nominal compressive
tion of a fracture tipline and the increase in area stress regime due to the overburden weight.
of two surfaces that formerly were bonded Grifths concept of stress concentration at aws
and the tools of linear elastic fracture mechanics play important roles in the development of faults.
are usefully employed to understand the initia- These and other contributions represent a con-
tion, propagation, and pattern development of siderable achievement in understanding the
opening fractures. The Coulomb criterion for phenomena of faulting (Sibson, 1986, 1987, 1989),
shear fracture initiation has found many insight- and a comparable body of literature has advanced
ful applications to faulting and earthquake phe- our understanding of the mechanics of faulting
nomena. On the other hand, exposure studies of (Rudnicki, 1977, 1979, 1980; Rice, 1980, 1992), but
faults reveal that stress and material heterogen- the research to integrate these different views of
eity not incorporated into the Coulomb criterion faulting stands before us.
Chapter 10
Viscous flow
Straight-limbed chevron-like folds in Cretaceous strata in the

large a part of Mr. Kings geological observations, that
footwall of the Lewis Thrust, Canadian Rockies, from the
the duty of enquiring into the physics of this enor-
Kananaskis Highway between Banff and Blairmore, looking
mously complicated subject devolved seriously upon
south. Photograph by D. Wiltschko.
me. Above all things some form of reliable working
hypothesis was to be discovered; and this is what the
This bulletin is the second contribution to the general present bulletin endeavors to accomplish. I believe the
investigation of the physical constants of rocks, the physical hypothesis has been found and that the data
experiments concerning which follow a general plan afford substantial corroboration of Maxwells theory of
devised by Mr. Clarence King. Questions bearing the viscosity of solids (Barus, 1891).
directly on the viscosity of rock masses make up so
10.1 ROCK DEFORMATION BY VISCOUS FLOW 385
10.1 Rock deformation by viscous (a)

flow
A
s the above quotation indicates, over one
hundred years ago, the rst director of the
United States Geological Survey, Clarence
King, hired a scientist to determine the viscosity of
rock. Searching through Kings account of the
survey of the fortieth parallel, we nd no mention
of Questions bearing directly on the viscosity of
rock masses . . . King did not write much, and so
it appears his thoughts on this subject and the (b)
observations that motivated them may have been
lost. King undoubtedly looked at rock masses that
contained fold structures like those shown on the
frontispiece for this chapter and in Fig. 10.1a,
which would have suggested to him that rocks
underwent continuous ow. Barus worked at ele-
vated temperature with steel and other sub-
stances, but not, to our knowledge, with rock. The
rst experimental studies in which rock was
deformed in a continuous manner were those of
Adams and Nicholson (Adams and Nicholson,
1901). Although interesting results were obtained,
research on the rheological behavior of rocks Fig 10.1 (a) Multi-layer folds (Moine Formation, Scotland).
apparently stopped until Griggs took it up in the (b) Multi-layer folds in alternating dark carbonatelight
1930s, working with technology developed by anhydrite, now gypsum, annual layer pairs (Castile Formation,
Bridgman, the 1946 Nobel Prize winner in physics Permian Basin, near Carlsbad, New Mexico). Photographs by
for his studies of the behavior of materials at high D. V. Wiltschko and K. Cruickshank.
pressure (Griggs, 1939). Examples of ordinary
viscous uid include cooking oils, lubricating ice, as in glaciers, and of salt, has been treated as
oils, molasses and honey, asphalt, molten glass, though these materials behave as viscous uids,
and magmas with modest volume fractions of and numerous researchers have modeled mantle
crystals and bubbles. Direct experience with some convection as ow in a viscous uid. The rst
of these materials reveals the large temperature detailed models of rock folding treat the buckling
dependence of their viscosity. of a viscous layer (Ramberg, 1960; Biot, 1961;
The use of the constitutive relations for an Chapple, 1968; Dietrich, 1969; Dieterich and
isotropic Newtonian viscous uid, dened below, Carter, 1969; Dieterich and Onat, 1969).
as a model for the rheological behavior of rock in Newton (1687) addressed the denition of a
mathematical models of rock deformation has a viscous uid in The Principia: The resistance
long history. Haskell (1937) used it for homo- which arises from the lack of lubricity in the parts
geneous crust and mantle in a model used to in- of a uid other things being equal is propor-
terpret measurements of glacial rebound. Cathles tional to the velocity by which the parts of the
(1975) treats the same phenomenon by represent- uid are being separated from each other. The
ing the crust and mantle as a series of layers with statement is a bit hard to interpret, but we
different viscosity, thus dealing approximately imagine Newton may have had two parallel plates
with the large effects of temperature and compo- containing a uid in mind, and the velocity is that
sition on rheological behavior. The slow ow of of one plate relative to the other in the direction
386 VISCOUS FLOW
of their plane. Lack of lubricity is viscosity. vx(x, h)

Resistance is shear stress and separation velocity
must be the velocity gradient perpendicular to Fx
the plane of the plates. We study Newtons uid in
this chapter, although much experimental work vx(x, y)
has demonstrated that glacier ice and other rocks h
y
at high homologous temperatures (temperature/
temperature of melting) and slow rates of defor- 0 x
mation behave as non-Newtonian uids. By slow we
mean rates of order Dxx 1014 s1, which would
for example account for a shortening to 50% in 2.2 L
Ma. None-the-less, many viscous boundary value Fig 10.2 An interpretation of Newtons thought
problems used by engineers have application to experiment.
questions motivated by geological eld observa-
tions. Several are discussed in the classic text
Elasticity, Fracture, and Flow . . . (Jaeger, 1964a), ness h, length L, and depth W is placed on the
including the ow of a viscous uid between planar horizontal surface of a rigid substrate
approaching or separating rigid plates (Robin and roughened to prevent slippage so vx(x, 0) 0.
Cruden, 1994), which we discuss later in this Although not drawn that way, L and W are sup-
chapter. posed to be much greater than h, so that condi-
The aim this chapter is to establish familiarity tions at the periphery of the sheet will not have a
with the formulation and analysis of models major effect on the interior ow. A rigid plate is
involving slow, creeping ow that have applica- placed on top of the uid and a horizontal force,
tion in structural geology. Models are developed Fx, is applied to it. The plate is rough so that the
analytically, although some require numerical uid adheres to it and attains the same velocity.
implementation to evaluate expressions or Because the plate spreads out the application of
contour quantities. Current research in this area the force in a uniform manner over the surface of
often uses numerical codes based on the nite- the uid, once a bit of motion takes place, the hor-
element, the nite-difference, or other methods izontal traction is:
for solving the eld equations subject to boundary tx Fx LW (10.1)
conditions. The models or boundary value prob-
lems examined here provide preparatory experi- Recall from Cauchys formula (Chapter 6) that
ence and insight necessary for further study using this traction component is related to the stress
these numerical methods, and a better funda- components as tx xxnx yxny zxnz. Since the
mental understanding of these phenomena. outward normal to the surface has components nx
0, ny 1, and nz 0, the shear stress at the
surface is:
10.2 Constitutive relations for yx Fx LW (10.2)
isotropic viscous fluids Gravity would cause the uid to ooze out of the
sides of the apparatus in Fig. 10.2, but to a sub-
10.2.1 Newtonian viscous fluid stantial extent, the rigid plates prevent that from
Consider an interpretation of Newtons thought happening. We shall suppose that the upper plate
experiment shown in cross section in Fig. 10.2. is not so thick and of such a large density that its
We could imagine performing an experiment like presence would contribute to outward ow at the
this, but it would create a mess because the uid edges. We hypothesize that the uid a few thick-
is not wholly contained. A slab of a uid, such as nesses, h, away from the edge will not sense
warm asphalt, of suitably large viscosity, thick- its presence, so there will be no dependence of
10.2 CONSTITUTIVE RELATIONS FOR ISOTROPIC VISCOUS FLUIDS 387
velocity or stress on x. Ignoring the weight of the Dxx [xx (yy zz)][2G(1 )]
uid itself and that of the rigid plate, only the
stress component xy is non-zero. Then, in the Dyy [yy (zz xx)] [2G(1 )]
absence of accelerations and gravity, Cauchys Dzz [zz (xx yy)] [2G(1 )]
Laws of Motion (Chapter 7) reduce to the single (10.8)
stress equilibrium equation: Dyz yz 2G
yx xy Dzx zx 2G

0 (10.3)
y x Dxy xy 2G
In other words, the shear stress in the (x, y)-plane
is constant within the layer of uid. If the uid is incompressible, the instantaneous
For a layer composed of an isotropic rate of change in volume is zero:
Newtonian viscous uid, experiments show that
1 dV
the velocity of the plate is proportional to the Dxx Dyy Dzz 0 (10.9)
V dt
shear stress xy or:
Substituting from (10.8), this requires 1 2.
vx(h) Cxy (10.4)
Using this in (10.8) and replacing G with the vis-
Here C is a constant with dimensions M1 L2 T. cosity, , as indicated by the experiment of Fig.
Only the rate of deformation component Dxy is 10.2, we have:
non-zero (Chapter 5), and because vy 0:
1 vx Dxx [xx (13)(xx yy zz)] 2
Dxy (10.5)
2 y Dyy [yy (13)(xx yy zz)] 2
For a given stress, experiments show that C is pro- Dzz [zz (13)(xx yy zz)] 2
portional to the thickness of the uid, h, so the
Dyz yz 2 (10.10)
velocity prole is linear and:
vx(h) 2Dxyh Cxy (10.6) Dzx zx 2
Dxy xy 2
Taking the constants C/h , we write:
Dxy xy 2 (10.7) The constitutive relations for a viscous uid
have the same form as those for an incompress-
Here is a material constant, the viscosity, with ible elastic solid, and the kinematic equations for
dimensions of M L1 T1. Other things being equal, the rate of deformation in terms of velocity gra-
the shear stress is proportional to the velocity gradients have the same form as the kinematic equa-
dient and the proportionality constant is twice tions for innitesimal strain in terms of
the Newtonian viscosity. displacement gradients. Thus, there is a com-
This experimental result is a special case of plete formal equivalence between the equations
the full set of constitutive relations for an governing the deformation of an elastic solid and
isotropic uid. A general linear relation between those governing the ow of a viscous uid. A
the innitesimal strain tensor and the stress solution to a problem for the deformation of an
tensor for an isotropic elastic solid was given in elastic body is associated with an equivalent
Chapter 8. The relations desired here may be solution to a problem for ow of a viscous body.
obtained from these by replacing the compo- This equivalence is described through the
nents of innitesimal strain with the compo- Correspondence Principle of Maurice Biot, and it
nents of the rate of deformation tensor, and extends to all viscoelastic substances, whose
writing G and for the shear modulus, G, and behavior combines elastic and viscous responses
Poissons ratio, : (Biot, 1965).
388 VISCOUS FLOW
10.3 Plane and antiplane flow The two constitutive relations are:
Dxz xz 2

10.3.1 Governing equations Dyz yz 2 (10.18)
The ows considered in this chapter depend only
on coordinates x and y. Two sorts of ow satisfy The single equilibrium equation is:
this restriction, and interesting cases have both xz yz
going on simultaneously. For plane steady ow: fz (10.19)
x y
vx vx(x, y), vy vy(x, y), vz 0 (10.11) Here fz is the body force component per unit
volume acting in the z-direction. Equations
Non-zero components of the rate of deformation
(10.17)(10.19) are a complete set of governing
tensor are:
equations for antiplane ow in an isotropic
Dxx
vx
x
vy
, Dyy , Dxy
y
1 vy vx

2 x y (10.12)
incompressible viscous uid. Antiplane ows are
not treated further in this chapter.
For Dzz 0, (10.10) gives:
10.3.2 Flow down an inclined plane
zz (xx yy) 2 (10.13) Consider an innite sheet of viscous uid of
uniform thickness h supported by the planar
The constitutive relations for plane ow in an
surface of a rigid substrate sloping at an angle
isotropic uid are:
(Fig. 10.3), to which it adheres. Gravity is turned on,
Dxx Dyy (xx yy) 4 so the uid ows downhill like a glacier and we
(10.14)
take x parallel to the slope and y normal to it. The
Dxy xy 2 boundary conditions for adherence to the sub-
The equilibrium equations for plane ow are: strate are:
xx yx vx(x, 0) 0, vy(x, 0) 0 (10.20)

fx
x y
(10.15) At the upper surface, the normal and shear trac-
xy yy tions vanish, yielding:
fy
x y
yy(x, h) 0, xy(x, h) 0 (10.21)
Here fx and fy are the body force components per
unit volume. Here, and in what follows, it is The equilibrium equations are:
understood that xy yx. With the y-axis vertical xx xy
and upward, and gravity as the only body force, g sin
x y
the components would evaluate as fx 0 and fy (10.22)
xy yy
g. Equations (10.12)(10.15) are the governing g cos
equations for plane ow of an isotropic viscous x y
uid. Here the body force components are those shown
A second ow independent of the coordinate z in Fig. 10.3.
is the antiplane ow where the velocity compo- Supposing that neither density nor viscosity
nents are constrained as: vary in the x-direction, and if the density is
vx 0, vy 0, vz vz(x, y) (10.16) uniform, we may set the rst terms in (10.22)
equal to zero, integrate, and apply the boundary
There are only two non-zero components of the conditions (10.21), obtaining:
rate of deformation tensor:
xy g sin (y h)
1 vz 1 vz (10.23)
Dxz ,Dyz (10.17) yy g cos (y h)
2 x 2 y
10.3 PLANE AND ANTIPLANE FLOW 389
a y the recumbent form of a fold nappe is developed

(Fig. 10.4b). Actual fold nappes are shown in Fig.
5.2 and in the frontispiece for Chapter 7.
h rg si The ow law of ice under conditions in many
n a
glaciers is non-Newtonian. Ice is incompressible
0
rg and, for simplicity, we may treat it as isotropic,
cos a although most glacier ice is strongly anisotropic.
rg x The complete constitutive relations for a non-
linear isotropic and incompressible uid are more
complicated than the linear relations (Chapter
11), but the one-dimensional relation between the
Fig 10.3 Section through the model for glacier-like
rate of shearing and the shear stress is:
gravitational flow.
Dxy B(xy
2 )[(n1) 2]
xy (10.26)
The stress distribution is statically determinate in All other deviatoric stress components are zero.
this case. That is, it only depends on the stress The form of (10.26) takes care of signs, since Dxy
boundary conditions and the equilibrium equa- and xy must have the same sign whatever the
tions. This means that we might consider any kind value of n 1. If, as here, both are positive, we may
of uid owing downhill provided its properties write, more simply:
only vary in the y-direction. Only Dxy is non-zero,
and for the uniform viscous uid: Dxy Bxy
n (10.27)
vx xy B is a material constant, expressed in SI units as

(g ) sin (y h) (10.24) (MPa)n s, and n is the stress exponent. The isotropic
y
Newtonian viscous uid (10.7) is a special case of
Integrating and applying the velocity boundary this so-called power-law uid for which n 1 and B
condition (10.20) yields: 1/2. The rheological behavior of ice under
glacier ow conditions and of many rock-forming
vx(y) (g ) sin ( 12 y2 hy ) (10.25) minerals such as quartz and olivine are approxi-
mated by such a law with n 3 (Kirby and
Glacier-like ow has been viewed as a possible Kronenberg, 1987; Evans and Kohlstedt, 1995).
mode of emplacement of large thrust sheets and Since (10.27) may be applied to the ow on an
fold nappes in which gravity is interpreted as the inclined plane, we substitute from the rst of
dominant factor (Hudleston, 1992). For example, (10.23) and obtain:
the northeasterly ow of supracrustal rocks of the

n
vx y
southern Canadian Rockies is envisioned as a 2B gh sin () n 1 (10.28)
result of gravitational spreading from a region of y h
buoyant upwelling of metamorphic and plutonic This may be integrated to give:
rocks in the core of the fold belt (Price, 1973). In

n1
another example, thrusting and folding of rock of 2Bh y
n1
vx c gh sin () n 1 (10.29)
the Pyrenees Mountains (Fig. 10.4a) is envisioned h
as a result of the upwelling of lower crust or
Here c is a constant of integration. Ignoring the pos-
mantle, here called the asthenolith (Choukroune
sibility of the uid sliding on its supporting plane,
and Seguret, 1973). If the lower crust is much less
we take the velocity to vanish at the base, y 0, so c
viscous because of its composition than the
equals the term in braces, and the nal result is:
underlying mantle, downslope motion of the
crust off such a high might be accomplished

n1
chiey by a glacier-like ow. If we apply the veloc- 2Bh y
vx [gh sin ()]n 1 1 (10.30)
ity eld (10.25) to an initially upright fold form, n1 h
390 VISCOUS FLOW
(a)
Sialic crust, r = 2.7 Asthenolith,

Intermediary layer, r = 3. r = 2.5
Arch re
imed ssu
ian pre
upward
about 40 km
(b)
y
y
x x
Fig 10.4 (a) Schematic diagram suggesting the mechanism
for thrusting and folding in the Pyrenees Mountains. If a mass of rock did ow in approximately
(b) Deformation of initial upright folds by glacier-like this fashion, and was cut by erosion, measure-
gravitational flow. Reprinted from Choukroune and Seguret ment of the shear strain distribution would allow
(1973) with permission of John Wiley & Sons. one to estimate n. Given the velocity eld, which
in this case is steady, we may determine the dis-
tribution of strain in the body for a given amount
The term in the rst braces is the velocity at the of surface displacement (Fig. 10.5b). Total dis-
surface of the layer, y h, since the second term placement is different for each prole in order to
equals 1 when y h. For given material properties, separate them. The deformed shapes of small
n and B, one may evaluate this velocity. Alterna- circles initially situated along the y/h-axis were
tively, we may measure glacier thicknesses, slopes, computed from a series of initial particle pos-
and surface velocities to estimate these proper- itions. Because of the gradient in shear rate, the
ties, in the spirit of a eld experiment. Notice that resulting loci only approximate ellipses. For
the surface velocity is proportional to hn 1. example, see the strongly deformed circle at the
An effect of the stress exponent, n, may be seen base of the sheet for n 10. Procedures in Chapter
by plotting the dimensionless ratio vx/vx(h) against 5 provide exact results: since the velocity gradi-
y/h for various n (Fig. 10.5a). As n increases, shear- ent tensor for a particle is constant as it moves
ing becomes strongly concentrated downward along a line at constant height y, the displace-
with increasing shear stress. For example, velocity ment gradient tensor may be computed by direct
proles through the thickness of a glacier may be integration.
measured by melting a vertical borehole and
placing a cable in it (Raymond, 1971; Meier et al., 10.3.3 Flow between approaching or
1974). After an interval of time, the cable may be separating rigid plates
re-occupied by sliding a melting device along it. The ow of uid outward from approaching rigid
Measurement of the inclination of the cable as a plates has been in the literature of structural
function of inextensible wire length from the geology for several decades (Jaeger, 1964a). It has
surface then allows the prole to be determined been applied to deformation in an orogenic belt
by integration from the surface position. between lithospheric plates whose relative motion
(a) 1 (a)
0.8
0.6
y/h
n=1
0.4
3 (b)
x
0.2 10
V +V
100
0
0 0.2 0.4 0.6 0.8 1
vx(y)/vx(h)
y
0 2L
(b) 1
0.8
2h
0.6
n=1
y/h
3
0.4
10
Fig 10.6 (a) Conceptual model for deformation in an
0.2
orogenic belt between lithospheric plates with normal and
tangential relative motion; reprinted from Robin and Cruden
0 (1994) with permission of Elsevier. (b) Axes and parameters
0 0.2 0.4 0.6 0.6 1.0 1.2 for flow between approaching parallel plates (Jaeger, 1964).
x/h
Fig 10.5 (a) Normalized velocity profiles for n1, 3, 10,

and 100. (b) Strain ellipses after surface displacements x/h1 shown in Fig. 10.6b. The plates are of width 2L, so
for n 1, 3, and 10. that the slit in which the viscous uid ows is
open to the outside at x L. To satisfy the end
conditions exactly greatly increases the difculty
is both normal to and tangent to their boundary of the problem, and we almost ignore them. The
(Fig. 10.6a): the case of transpression (Robin and idea used is that these conditions will only
Cruden, 1994). The problem may be done either in signicantly affect the ow within a few slit
plane ow or in radially symmetric ow in (r, z)- widths from the end. We thus restrict attention to
coordinates; the former is the case of interest here. the case L/h 1, so that these regions will be
This boundary value problem is historically inter- small relative to the total length.
esting because it demonstrated to structural geol- If the uid sticks to the surfaces of the plates,
ogists that a ow of some complexity in terms of the boundary conditions there are:
velocity and stress elds could be worked out by
elementary means. vx(x,h) 0 vx(x, h)
(10.31)
The conguration and coordinate axes are vy(x,h) V vy(x, h)
392 VISCOUS FLOW
The magnitude of the velocity of each plate is V. (a) (b)

We select the coordinate axes so that the ow will
have maximum symmetry:
vx(x,y) vx(x, y)
vy(x,y) vy(x, y) (10.32)
vx(x, y) vx(x, y)
That is, vx is an even function of y and an odd func-

tion of x, and vy is an odd function of y. The second
restriction implies that vx(0, y) 0. A vertical
mirror plane of symmetry is present at x 0. Such
a plane behaves like the frictionless surface of
a rigid medium, preventing horizontal ow.
Symmetry also requires that the shear traction on x
the (x 0)-plane be zero.
Fluid ows in opposite directions from the y
central mirror plane of symmetry. The uid ux
at some distance |x| from the mirror plane must Fig 10.7 (a) Streamlines and (b) velocity vectors for flow
be equal to the volume per unit depth swept out between approaching parallel plates.
by the approaching plates per unit time, or 2V|x|.
Both conditions can be met by setting vx x. The
simplest function satisfying this condition and
(10.31) is: Substituting the rst of these into each of the
equilibrium equations (10.15) without gravity
vx A(y h)( y h)x A( y2 h2)x (10.33)
yields:
From incompressibility the velocity gradients are xx yy
related as: 2 Ax, 2 Ay (10.38)
x y
vy v
x A( y2 h2) (10.34) Integrating each of (10.38) we nd:
y x
xx Ax2 f1( y)
Integrating (10.34) we nd:
(10.39)
yy Ay2 f2(x)
(10.35)
vy A ( 13 y 3 h2y ) By combining the second relation in (10.37) with
(10.39), we obtain:
This velocity distribution satises the conditions
(10.31) if: 4 A( y2 h2) Ax2 f1( y) C1 Ay2 f2(x)
(10.40)
A 3V2h3 (10.36)
The velocity eld over a distance of 4h from the The arbitrary functions f1(y) and f2(x) are:
mid-plane is illustrated by streamlines and vectors
f1( y) A(3y2 2h2) C
in Fig. 10.7. (10.41)
f2(x) A(x2 2h2) C
From the constitutive relations (10.14) and
kinematic relations (10.12) we have: Then the two normal stress components from
(10.39) are:
xy
vy vx

x y
2Axy
(10.37) xx A(x2 3y2 2h2) C
v (10.42)
xx yy 4 x 4 A( y2 h2) yy A(x2 y2 2h2) C
x
The constant C may be determined by specify- 10.3.4 Lubrication theory: steady-state

ing boundary conditions at the ends of the plates. solution for an accretionary
For example, suppose that the mean normal pres- wedge
sure, p 13 (1 2 3) 12 (xx yy), is zero there. An accretionary wedge (Fig. 10.8) is built up at the
Specically we take: edge of a plate under which another subducts
(Chapple, 1978; Davis et al., 1983; Batt et al., 2001;
BC: xx yy 0 at x L, y 0 (10.43) Brandon, 2004). Such accretionary wedges have
Since h/L 1, ignore the terms in h2 in compari- been treated as though the material involved
son with those in L2 and take: exhibited sand-like, or plastic, behavior (Dahlen et
al., 1984; Hilley and Strecker, 2004; Hilley et al.,
C A L2 (10.44) 2004). Here we develop a simple model that
The two normal stress components for these assumes the material behaves as a viscous uid.
boundary conditions are: This assumption has been used by Emerman and
Turcotte (1983), who rst worked out the model
xx A(x2 L2 3y2 2h2) presented here. As they did, we could also obtain
(10.45) a solution for the non-linear power-law uid dis-
yy A(x2 L2 y2 2h2) cussed in Section 10.3.2. In contrast to the bound-
The average normal stress acting on the plates ary value problem for the ow down an inclined
to produce an approach velocity of 2V is: plane, the analysis set up here has many more
L
steps, and involves more approximations than the
(x, h) dx
1 problem for ow between approaching parallel
yy(ave) yy (10.46)
L plates.
0
A schematic illustration of the model
Substituting from the second of (10.45) we nd: conguration (Fig. 10.8) has a feature termed a
backstop. We do not wish to motivate and justify
yy(ave) 23 A L2 VL2h3 (10.47) this peculiar feature exhaustively here, but some
discussion seems necessary. An accretionary
We note that the approach velocity, V, of the wedge is composed of sediments and other ma-
plates scales with the applied stress and h3, terials scraping off a subducting plate, and the
whereas it is inversely proportional to the viscos- analogy has been drawn between this process and
ity and L2. pushing a mass of material across a horizontal or
Robin and Cruden (1994) use this ow eld to tilting surface as between a snowplow mounted at
simulate the deformation in a slab of material the front of a truck and the road surface. The rel-
lying between two lithospheric plates (Fig. 10.7a) ative motion is inverted here so the subducting
that have a component of normal motion as well plate moves, while the plow stays xed. The wedge
as tangential relative motion, i.e. the deformation of material on the subducting plate moves in a
termed transpression. In this case, the region direction opposite to that of subduction. The
shown in Fig. 10.7b is imagined as vertical, with backstop is identied with the plow and taken to
material in it extruding out from the top to form be vertical. No plow is present towards the rear of
a kind of mountain belt. The trans- part of the a natural accretionary wedge, so this must stand
motion is shearing across the region, but in and in as an approximation for some other feature.
out of the plane. The shearing motion may be Progressing from the toe (Fig. 10.8), an accre-
superposed on the -pression motion derived tionary wedge has a topographic culmination, to
here, since the governing equations are linear. the rear of which it loses topography. The reason
Although the resulting velocity eld is easy to for this is that the subducting plate piles up ma-
visualize, the distribution of strain and rotation terial as long as it exerts a shearing motion at the
in the uid slab is complicated; the results must base of the wedge, but once the plate surface con-
be computed and puzzled out. tacts lower, hotter material of negligible strength,
394 VISCOUS FLOW
(a) Distance from deformation front (km)

0 20 40 60 80 100 120 140 160 180 200 220 240 260
Mt. Olympus
Deformation Shelf West Structural lid East

front edge coast coast
0 0
Depth (km)
10 Juan de Fuca plate (all crus rz

on
e
10
t and mantle subducted) ea
20 CASCADIA WEDGE etro sh 20
r crustal structure
30 Moho S here poorly resolved 30
30
6 km/my
(b) Austroalpine
Jura Mtns. Swiss foreland nappes
0 0
Depth (km)
Helvetic Pennine nappes

10 10
nappes
20 ALPINE 20
Moho
e
on
30 WEDGE
rz
30
ea
sh
tro
Moho S 40
re
50
~5 km
/my
(c) L/4
Erosion A
we focus on the prominent seaward side of the
Backstop
wedge. We speak of the backstop and occasionally

H of plowing as in these physical analogies.
Under- To solve for the steady-state shape of the
plating Strong
wedge, h(x), and the ow and stresses within it, we
U sin will employ the approximations of lubrication
(lx) Weak
theory (Batchelor, 1967; Schlichting, 1979). The
idea behind lubrication theory as applied here is
Fig 10.8 Comparison of the structure of (a) the Cascadia
that the wedge is slender: the maximum thick-
margin, Washington State, and (b) the European Alps,
ness is much smaller than its width, so h(0)/L
Switzerland (Brandon, 2004). (c) Schematic diagram of an
accretionary wedge model. 1, where h(0) is the maximum height at the back-
stop. We suppose that the velocity of plowing is
sufciently small for this to be so. For such a body,
this ceases to occur. The topographic culmination it is natural to suppose that quantities vary much
occurs approximately above this position. In the more rapidly in the vertical direction than in the
present case, at least, we take the backstop to rep- horizontal direction. Thus, horizontal derivatives
resent the position of the topographic culmina- of quantities of the same sort, e. g. stress compo-
tion and, by making it vertical and frictionless, we nents, may be neglected relative to vertical deriv-
approximate the accretionary wedge as having atives. In the analysis, we postulate that the unit
mirror symmetry. An accretionary wedge does weight, g, and viscosity, , are uniform while rec-
continue beyond the topographic culmination, ognizing these quantities would vary spatially in
but it tends to possess a strong asymmetry. Here, an actual accretionary wedge.
The rst step is to drop the x-derivative in the subducting plate at the backstop are not provided.
second equation in (10.15) and write: Indeed, no conditions at a backstop are specied
in the formulation of this problem! Rigorously, we
yy y g (10.48)
cannot even speak of its presence because it has
Integrating this equation with the condition that not entered the formulation and analysis.
the normal stress vanish at the top of the wedge, At this point, a complete approximate solution
approximated as yy(x, h) 0, yields: for stress and velocity in the wedge is expressed in
terms of the unknown prole h(x,t). To obtain this,
yy g( y h),h h(x, t) (10.49) we write the relation for conservation of mass
between vertical surfaces at x and x dx for wedge
In the kinematic relations (10.12), derivatives of vx material of uniform and constant density:
occur in Dxy and Dxx, and that in Dxy is larger, by
h J
hypothesis, so Dxy Dxx. Discarding the derivative dx Jx(x) Jx(x dx) x dx (10.55)
t x
vy x in Dxy, we have:
Here Jx is the volume ux across a vertical surface
Dxy 12 (vx y) (10.50) per unit strike length, or:
h
From (10.14) we conclude that xy |xx yy|.
Using these approximations in the rst equilib-

Jx(x) vx(x, y)dy (10.56)
0
rium equation (10.15), the result (10.49) yields:
Substituting (10.54) with the condition vx(0) V
xy yy h into (10.56) and integrating:
xx g (10.51)
y x x x
gh3 h
Integrating, and using the requirement that the Jx Vh (10.57)
3 x
shear traction vanish at the upper surface,
approximated by xy(h) 0, we obtain: Because the wedge slopes toward its toe, h/x 0,
h the rst term is positive, signifying a ux toward
xy g ( y h) (10.52) the toe. The second negative term represents the
x
motion of the subducting plate relative to a xed
We may now use (10.14), (10.50), and (10.52) to backstop.
write: Substituting (10.57) into (10.55), yields a partial
vx g h differential equation in h:
( y h) (10.53)
y x
Integration gives:

h gh3 h
t x 3 x
Vh
(10.58)
vx
g h y2
x 2
hy vx(0) (10.54) We look for a steady-state solution, in which the
form of the wedge is unchanged, and set ht 0.
If the viscous wedge does not slip at its base, Integrating the resulting ordinary differential
vx(0) V . You may wonder how we are able to equation yields:
move the wedge along with our plow with this no-
gh3 h
slip condition. The notion is that the uid is Vh C (10.59)
3 x
scraped off right at the backstop, but the mass
of uid is not sliding over the surface as a whole. This equation satises the condition that the
As in the analysis of the ow between wedge has a nite width, L, that is h(L)0, if C 0.
approaching rigid plates (Section 10.3.3), end con- Factoring out an h, integrating again, and re-
ditions on the wedge are ignored. Thus, the applying this condition gives (Emerman and
details as to how the material is scraped off the Turcotte, 1983):
396 VISCOUS FLOW
(a) Typical values for an accretionary wedge are h0

h(x)/h(0) 20 km, L 200 km, 2600 kg m3 and a subduc-
tion velocity U 5 cm a1 (Emerman and Turcotte,
1983; Batt et al., 2001). The relation (10.62) yields a
wedge viscosity 0.7 1020 Pa s1. This is a plaus-
Backstop
Accretionary ible estimate for rock viscosity in an accretionary

wedge wedge (Emerman and Turcotte, 1983). An estimate
of 1021 Pa s1 has been obtained for the weak upper
x/L mantle (Haskell, 1937). The lubrication theory
model for a viscous accretionary wedge may be
late
ct ing p used to assess the effect of factors such as rate of
S ubdu supply of material, rate of erosion, and viscosity
on wedge dynamics.
(b)
10.4 Viscous flow in layers:
h(x)/h(0)
mullions and folds
The great variety of structures produced by rock

Backstop
Accretionary deformation and their variation in form and asso-

wedge ciation may be usefully grouped into a few cat-
egories. One category contains structures
produced by the deformation of layers. Examples
in this chapter include the folds seen in the front-
ispiece and in Fig. 10.1; additional examples
x/L follow. Other categories are the crack-like struc-
Subducting plate tures described in Chapter 9 such as faults, joints,
dikes, veins, and anticracks, and inclusions of one
Fig 10.9 Schematic diagrams of accretionary wedges material embedded in another. Models for struc-
forward of a backstop. (a) Inclined boundary between wedge
tures in each category are commonly formulated
and subducting plate. (b) Horizontal boundary.
using special sets of mathematical tools. Here we
focus on the deformation of layers.
13
h
9VL
g
1
x
L
(10.60) 10.4.1 Biharmonic equation and its
solution
The proles of the wedges in Fig. 10.9 have the In Chapter 8 we introduced the three components
form specied by (10.60), where h in both cases is of stress in the (x, y)-plane written as partial deriv-
taken as the total vertical height above the planar atives of the Airy stress function, (x,y):
base. The solution applies to a nite wedge of any
2 2 2
cross-sectional area A, where: xx 2
, xy ,yy 2 (10.63)
y xy x
L

4 13

243VL These expressions satisfy the two-dimensional
A h(x)dx (10.61)
64g stress equilibrium equations (10.15) and thus
0
The maximum wedge thickness at x 0 from replace three unknown stress components in a
(10.60) is: problem of plane ow with a single unknown
function from which they may be derived.

13
9VL The plane ow kinematic relations (10.12) are
h0 (10.62)
g three equations in two unknowns, vx and vy.
10.4 VISCOUS FLOW IN LAYERS: MULLIONS AND FOLDS 397
Eliminating the two components of velocity (a) Grenoble Pelvoux Massif

Belledonne Massif SE
between them yields the rate of deformation com-
patibility relation:
2Dxx 2Dxy 2Dyy

2 2 0 (10.64)
y2 xy x
0 5 10 km
In order to obtain this equation by taking partial (b)

derivatives, and even before that to write the rate
of deformation components in terms of partial
derivatives of the velocity components, the veloc- 15 cm
ity components must be continuous and have con-

tinuous partial derivatives to third order. This
continuity is a prior necessary condition for deriv-
ing (10.64). Structural geologists often refer to the
10 cm 15 cm
compatibility condition as a condition required to
insure that the velocity eld (or displacement
eld) is smooth, so that pieces of the body
described are not slipping or separating relative to
each other. This interpretation has things back- 50 cm 5 cm
wards: continuity is required to write down (10.64).
From (10.14) and (10.63) the rate of deforma-
tion components in terms of the stress function
are:
Fig 10.10 Mullion structures: (a) large scale (Ramsay,
Dxx Dyy
1 2 2

4 y 2 x2,Dxy
1 2
2 xy
1967); (b) small scale (reprinted from Sokoutis (1987) with
permission of Elsevier).
(10.65)
Substituting into (10.64), we nd:
10.4.2 Mullions

2 2 1 2 2 Mullion structure was the rst geological struc-
ture to be investigated using the nite-element
y2 x2 4 y2 x2
method (Dieterich and Onat, 1969). Here, we will
2
2

1 2
xy 2 xy
0 (10.66) learn something about its formation from a
simple analytical study. This structure, illustrated
at scales of centimeters and kilometers in
In the present case, we suppose that the viscosity, Fig. 10.10, consists of a lobe-and-cusp morphology
, is not a function of position so we obtain: of interfaces between different rock types and is
4 4 4 interpreted to have formed during interface-
2 2 2 4 0 (10.67) parallel shortening of large magnitude. Lobate
y 4 y x x
forms are developed in the stiffer rock with the
This is the biharmonic equation, often written in the softer, or less viscous, rock-lling cusps. One
condensed form: might suppose that the surfaces of the more
viscous rock layers were puckered, so as to
4 2(2) 0 (10.68)
undergo little surface-parallel shortening. If so,
Having reduced the general set of equations for one could lay out the surface trace in a straight
plane ow, (10.12)(10.15), to a single equation in line and use this to estimate the original length of
one unknown function, (x, y), the next step is to the interface. Then, as done for the folded vein in
set up some boundary value problems of interest Fig. 5.3, one could estimate the amount of bulk
and solve them. shortening.
398 VISCOUS FLOW
The mullion structures in Fig. 10.10a have a L

variable cusp-to-cusp span that is much smaller y
D*xx
than the vertical dimensions, not indicated in the
gure, of the bodies at whose interface they lie. x
Both are probably layer-like or sheet-like in form. 0
Large mullions in this gure have spans of about
10 km, but many smaller parasitic mullions
have dimensions of a kilometer, so mullions
apparently form at multiple scales. These obser-
vations lead us to infer that mullion initiation in n y tn(1) (1)
ts
s
this case may not depend on a characteristic
dimension of the initial conguration, such as a
layer thickness. We thus consider a conguration
u x ts(2) tn
(2)
consisting of two viscous half-spaces. In contrast,

the mullions in Fig. 10.10b show a regular span on Fig 10.11 Portion of interface for the mullion model, with
the decimeter scale with cusp-to-cusp arc lengths axes and tractions locally tangent and normal to the
comparable to the thickness of the layer separat- interface.
ing them. The model developed here will not give
insight into this aspect of their deformation.
Suppose two materials undergo uniform shor- deformation continues. This completes the solu-
tening parallel to a planar interface without the tion for the case of a perfectly planar interface,
formation of mullion structure. Quantities in the which we term the basic state.
upper half-space are identied with the super- Now suppose that mullion structure is formed
script (1), and those in the lower half-space with by the amplication of an irregular waviness
(2). Assuming plane ow, (10.12)(10.15), the present on the interface before shortening. We
homogeneous shortening is described by: treat the interface as a cylindrical surface of sinu-
soidal form (Fig. 10.11) with axis normal to the (x,
D(1)
xx Dxx Dxx
(2)
y)-plane of ow, amplitude A, wavenumber , and
Dyy
(1) D (2) D D (10.69) wavelength L 2/:
yy yy xx
D(1)
xy 0 Dxy
(2) y (x, t) A(t) cos (t) x (10.72)
From (10.69) and the constitutive relations (10.14) Both the amplitude and wavenumber are func-
we have: tions of time, t. The interface in the case of mul-
lions shown in Fig. 10.10a is not periodic, nor does
xx yy) 41 ( xx yy) 42 Dxx
( (1) (2)
(10.72) describe the lobe-and-cusp asymmetry of
(10.70)
xy xy xy 0
(1) (2) the mullion interface. However, a sum of wave-
length components provides the possibility of
The stress components yy and xy are the same in extending the solution to cover these variations.
both half-spaces, as required by the continuity By considering a periodic interface form, we
of traction at the planar interface. The surface- isolate a region of interest that is a single wave-
parallel normal stress must be different unless length wide. The vertical planes at x L/2 may
the viscosities are the same. be treated as though they were the frictionless
Taking the coordinate origin at the interface, planar surfaces of two rigid platens, between
the velocity components, with the origin xed, are: which the two media with their sinusoidal inter-
v (1)
face shorten. Because the shear stress must vanish
x v x Dxxx
(2)
(10.71) on a mirror plane of symmetry, these planes have
v (1)
y v y Dyy y Dxx y
(2)
the property of a frictionless surface.
These satisfy the condition that adjacent particles The local slope with maximum at x L/4 of
in the two viscous uids remain neighbors as magnitude A is:
tan x ( A) sin x (10.73) ponents of stress (Fig. 10.11). We might also have
used the components with respect to the coordi-
Taking A 1, we use the approximations:
nate axes x and y. The present choice is more gen-
tan 33 ( A) sin x erally useful; for example, if the interface were
sin 36 ( A) sin x (10.74) frictionless, the shear component of the traction
would be required to vanish. It would then be nec-
cos 1 22 1
essary to write the velocity boundary conditions
We seek a solution accurate to terms proportional in terms of normal and tangential components
to A. In the derivation, if terms proportional to also, since slip on a frictionless interface would
(A)2 or higher are encountered, they are deleted. exclude a condition being set on the tangential
We cannot hope to simulate the development of a component of velocity.
large-amplitude mullion (Fig. 10.10a) with this Expressing the tractions in terms of the stress
approximation. Instead, we study mullion initia- components referred to the local coordinate axes,
tion and the inception of an asymmetric lobe-and- n and s:
cusp structure.
t (1)
n nn, t n nn
(1) (2) (2)
For the wavy interface, the basic-state solution (10.79)
no longer satises all boundary conditions at the s ns , t s ns
t (1) (1) (2) (2)
interface to the desired accuracy, as we now In terms of the stress components, the boundary
demonstrate by introduction of a correction term conditions are:
in the form of additional sets of velocity, stress,
nn (x, ) nn (x, )
(1) (2)
and rate of deformation components. The veloc- (10.80)
ity components in the two media are the sum of ns (x, ) ns (x, )
(1) (2)
the respective components from the basic state

These conditions must be written in terms of the
and from this corrective state, so they have the
components of stress referred to the (x, y)-coordi-
form:
nates used in the eld equations, in which the
v (1) ~(1) ~(1)
x vx vx , vy vy vy
(1) (1) (1)
solution will be expressed. Using the transforma-
(10.75) tion law we have:
v (2) ~(2) ~(2)
x vx vx , vy vy vy
(2) (2) (2)
The stress and rate of deformation components nn xx sin 2 2xy sin cos yy cos 2
are composed of similar sums. Suppose that the
interface is welded, so velocity components of ns (xx yy) sin cos xy( cos 2 sin 2)
adjacent particles across it are equal: (10.81)
v (1) Using (10.74), and retaining only terms up to those

x (x, ) v x (x, )
(2)
(10.76) proportional to A, we have:
y (x, ) v y (x, )
v (1) (2)
nn 2xy yy

Since the basic-state velocity components already (10.82)
satisfy these conditions, (10.76) reduce to: ns (xxyy) xy
~ ~(2)
x (x, ) v x (x, )
v (1) No correction to the basic-state solution is nec-
(10.77)
~
v (1) ) ~
v (2) essary if the interface is perfectly planar (A 0);
y (x, y (x, )
that is, the perturbing solution is identically zero.
The boundary conditions on the normal and Therefore we postulate that the components of
shear components of the traction on the interface stress and velocity in the perturbing solution are
are: themselves proportional to A. Since xy 0, we
t (1) have xy ~ xy and the term xy in the rst of
s (x, ) t s (x, )
(2)
(10.78) (10.82) can be discarded because it is a product of
t (1)
n (x, ) t n (x, )
(2)
two terms, both of which are proportional to A.
The tractions will be evaluated for each medium Applying these conditions, and the forms (10.74),
in terms of the interface geometry and the com- the stress boundary conditions (10.82) become:
400 VISCOUS FLOW
yy (x, ) yy (x, )
~
(1) ~ (2) The solution to (10.88) may be written:
~
(1) (1)
xy (x, ) ( xx yy)[(A) sin x] (10.83) F( y) 2{[a b(y1)]ey [c d(y 1)]ey}
~ xy (x, ) ( xx yy)[( A) sin x]
(2) (2)
(10.89)
The second relation in (10.83) is consistent with Here a, b, c, and d are arbitrary coefcients whose
the hypothesis that the perturbing ow is propor- values are xed by the boundary conditions. The
tional to A. Indeed, any information that the term ey is multiplied by a b to give the resulting
interface is not planar, and, hence, that a per- stress components a symmetric form. The factor
turbing ow must be present, comes only from 2 is added to make the expressions for the veloc-
this condition. ity components simpler in form. The stress com-
The second boundary condition in (10.83) ponents associated with the perturbing ow are
determines the x-dependence of the required solu- found from (10.63):
tion of the biharmonic equation (10.67). Two
~
xx 2{[a b(y 1)]ey
pieces of information are supplied. First, the con-
dition must be satised for all values of x on the [c d(y1)]e y}cosx
nearly planar interface y (x). This suggests the ~
yy 2{[a b(y1)]ey (10.90)
possibility that the function (x, y) and, from it, [c d(y 1)]ey }cos x
the components of stress and velocity, might be
~
xy 2[(a by)ey(c dy)e y] sin x
written in a separable form:
(x, y) F( y)G(x) (10.84) From the constitutive relations (10.14) and the
kinematic relations (10.12) for plane ow we have:
Since ~xy 2 xy, the second condition in ~x ~
v xx ~
yy
(10.83) could be satised if G(x) cos x, giving: [(a by)ey
x 4
(c dy)e y] cosx (10.91)
~
xy(x, ) (dFdy)y sin x (dFdy)y0 sin x
(10.85) Integrating (10.91) the x-component of velocity is:
To the present approximation, this condition ~

v x [(a by)ey (c dy)e y] sin x (10.92)
becomes:
The y-component of velocity is found from
the condition of incompressibility written as
(dF (1)dy)y0 sin x ( A)( (1)
xx yy) sin x ~y y v
~x x and (10.92):
v
(dF (2)dy)y0 sin x ( A)( (2)
xx yy) sin x
~
v y {[a b(y 1)]ey [c d(y 1)] e y}cosx
(10.86)
(10.93)
The common factor sin x may be cancelled in Using this solution for the components of stress
(10.86). and velocity, we are able to solve a great many
The stress function is taken as: interesting problems for layers. Our rst applica-
tion is in the analysis of the mullion. Later, we use
(x, y) (1 2) F ( y) cos x (10.87)
this solution to study the buckling of viscous
Here the factor 1/2 leads to simpler expressions layers (Fig. 10.1a and Fig. 10.14). Other examples
for the stress components. Substituting (10.87) are discussed elsewhere (Johnson and Fletcher,
into the biharmonic equation (10.67) we obtain 1994).
the ordinary differential equation: In the present case, the solutions for the per-
turbing ows in the upper and lower half-spaces
d4F d2F are special cases of the more general forms just
22 2 4F 0 (10.88)
dy 4 dy obtained. Since the ows in the two half-spaces
must be distinct, there are eight arbitrary con- such particles are not dissolved away or eroded,
stants to be xed using boundary conditions. nor is new material added to the interface. Using
These constants are: a1, b1, c1, d1, a2, b2, c2, d2, the solution for the coefcients, (10.95) and
where subscripts 1 and 2 refer to the upper and (10.96), and the expressions for the velocity com-
lower half-spaces, respectively. At the interface ponents, (10.92) and (10.93), we may evaluate the
there are only four boundary conditions, (10.77) velocity vectors for a large number of particles on
and (10.83), thus indicating that we have not yet the interface, y Acos x, and increment their
specied four boundary conditions elsewhere. positions by multiplying the velocity vectors by a
Considering the y-component of velocity (10.93) in small quantity t. The velocity components in the
the upper half-space, y 0, notice that the lower layer are:
coefcients a1 and b1 multiply terms proportional
to ey, which grow in magnitude without limit as x Dxxx a2(1 y)e sin x
v (2) y
(10.97)
y increases. Since a relatively gentle waviness at v (2)
y Dxx y a2ye cos x
y
the interface cannot be expected to produce an This includes the basic state and the perturbing
ever-increasing ow away from the interface, we ows. If the viscosity of the lower medium is
take a1 b1 0. Considering the same situation in greater than that of the upper, 2 1, then a2 0
the lower half-space, we conclude that c2 d2 0. for shortening, Dxx 0. It is useful to re-cast the
The remaining four coefcients are xed by expressions (10.97) with (10.95) and (10.96) into a
application of the four boundary conditions at dimensionless form, using |Dxx|L as the charac-
the interface. Begin with the condition on the teristic velocity:
continuity of the shear stress, (10.86). Using
(10.90), and substituting y in the functions con-
tained in the expressions ~ ~ (2)
xy and xy , we obtain:
(1)
v (2)
x
|Dxx|L
x 1 1R
Sgn (Dxx)
L 1R 2 A
L
21(c1 d1)e (A)41Dxx

(1 y)ey sin x

22(a2 b2)e ( A)42Dxx (10.94) (10.98)

(2)
Recall our hypothesis that the perturbing ow is vy y 1 1R 2 A
Sgn (Dxx)
proportional to A. This can be the case only if the |Dxx|L L 1R L

coefcients themselves are proportional to this
quantity. Since (A) cos x~A, we expand the yeycos x
exponential functions to rst order in this quan- Here the dimensionless quantities are 2A/L A
tity: e 1 and e 1. Retaining only and R 2/1. If the half-spaces are undergoing
terms linearly proportional to A we nd shortening, Sgn(Dxx) 1, but for interface-
1c1 2a2 2(2 1)Dxx(A). The condition on parallel extension, Sgn(Dxx) 1.
continuity of vx, (10.77), treated in the same way, Figure 10.12a shows the perturbing part of the
yields c1 a2. Combining these equations we have: velocity eld in the lower, more viscous half-space
c1 a2 2[(1 2)(1 2)]Dxx A (10.95) for R 10. Notice how the perturbing ow is con-
centrated near the surface and has zero vertical
The remaining two boundary conditions on the velocity there. The horizontal ow away from the
normal component of stress and the vertical com- crest of the sinusoidal perturbation in surface
ponent of velocity yield: shape broadens the lobes and tightens the cusps
in the interface form. This is kinematically
c1 d1 0 a2 b2 (10.96)
amplied in the basic state of uniform shortening.
From (10.95), if the viscosities of the two media are Figure 10.12b shows contours of the perturbing
equal, the perturbing ow vanishes. part of the horizontal normal stress. A horizontal
We consider the deformation of the interface zero contour lies at y L/, below which this
by following particles on it. We postulate that stress component changes sign, from tensile to
402 VISCOUS FLOW
(a)
(a)
(b)
(b)
Fig 10.13 (a) Profiles from mullion model computed by
0.9 0.9
incrementing positions of particles. (b) Profiles computed
0.5 using the interface evolution equations.
0.5 0.5
0.1 0.1 0.1

material time derivative, derived in Chapter 7. If f is
0.02 0.02 0.02 a scalar property of a material element the ma-
0.00
terial time derivative is:
0.02
0.02 0.02 Df f f f f
vx vy vz (10.99)
Dt t x y z
0.00 0.00
A special property of a particle on the interface
is that it stays on the interface. For the two-
dimensional case considered here there is no
dependence on the coordinate z. We then take as
Fig 10.12 (a) Velocity vectors for the mullion perturbing the property:
flow. (b) Contours of the perturbing horizontal normal
stress, normalized by its maximum value. f (x, y, t) y (x, t) 0 (10.100)
While other particles may be free to take arbitrary

compressive under the crest of the structure. The positions, or x- and y-coordinates, the position of a
contours shown below the zero contour are only particle on the interface is constrained to have a
0.02 of the maximum; these contours are also special relationship between its x- and y-coordi-
represented above the zero contour. nates:
Figure 10.13a shows a sinusoidal surface with D
an initial maximum slope of 0.02 radians and new [y (x, t)] 0 (10.101)
Dt
surfaces obtained by evaluating the velocity
vector at points on the surface and using them to Using (10.99) with x, y, and t as the independent
move particles for small dimensionless incre- variables we nd:
ments of |Dxx|t for R 10. The proles are shifted

vertically to separate them. The change in hori- vy(x, ) vx(x, ) (10.102)
t x
zontal span of the surfaces indicates the nite
homogeneous shortening. The surface begins to This is the interface evolution equation.
assume a lobe-and-cusp form with broader anti- Recall that we have obtained an approximate
cline and tighter anking synclines. The last two solution for the sinusoidal shape perturbation,
proles have slopes too large for a good approxi- Acos x. Such a solution applies to an arbitrary
mation by the rst-order solution, but are shown two-dimensional, or cylindrical, shape perturba-
here to clarify the mullion-like forms. tion because: (1) this may be treated as a sum of
Another means of following the evolution of Fourier components with form allowing for varia-
an interface or surface requires the use of the tion in phase; and (2) to the approximation of the
present solution, such components are linearly lation for dA/dt. In fact, if the missing term is
independent, i.e. non-interacting. However, for a included, the 2 in the second equation must be
periodic perturbation, we may analyze how the cancelled so (10.105) over-estimates the rate at
incipient mullion structure evolves at larger which the lobe-and-cusp structure develops. The
amplitude and interface slope while keeping the more accurate result is shown in Fig. 10.13b,
mathematics manageable. A step in this direction which is computed using this method.
is to substitute the solution into the interface To improve upon the analysis of stress, veloc-
evolution equation (10.102). ity, morphological development, and strain
Imagine that the initially sinusoidal interface within mullion structures, one may appeal to
develops into a form better approximated by: numerical codes (Dieterich and Onat, 1969), or
obtain more accurate results by extending the
(x, t) A cos x Acos 2x (10.103) analysis used here to higher orders of approxi-
Here the initial amplitudes are A(0) A0 and mation (Johnson and Fletcher, 1994). The rela-
A(0) 0. The asymmetric lobe-and-cusp modi- tively simple method of analysis and results
cation to the sinusoidal form is represented by the provided here offers insight that helps us to
second term. When velocity components are substi- understand how the lobe-and-cusp morphology
tuted in (10.102), we retain contributions propor- develops. In essence, the cusps occur due to a
tional to cos 2x. In the expansion of v (2)
y (x, ), we
small compressive stress concentration offset by a
retain terms in a2, that is those proportional to stress reduction in the material that extends
(A)2. Substituting (10.103) and (10.98) into (10.102): outward into the less viscous uid (Fig. 10.12b).
We may follow the strain distribution as well as
dA
dt
cos x A sin x x
d
dt

dA
dt
cos 2x the interface morphology using the present
results, or to see to what extent a tendency to pre-

d serve arc length along the interface exists. Both
A sin 2 x 2x
dt the kinematics and the stress distribution
suggest that a variety of minor structures might
Dxx(A cos x A cos 2x) a2(A cos x) cos x
be formed close to the interface in a region of
(Dxxx a2 sin x)(A sin x2 Asin 2x) marked strain variation.
(1.104)
10.4.3 Folding of a single viscous layer
We nd, according to this computation, that A is
The method of analysis developed in the preced-
proportional to (A)2. The product a2(A) also is
ing section may be applied to the folding of a
proportional to (A)2. Since d dt Dxx, the
more viscous layer isolated in a less viscous
terms proportional to x on both sides cancel.
medium, giving insight into how folds such as
Then, since this relation must be satised at all x,
those in Fig. 10.1a and 10.14 are initiated. The
the terms in cos x and cos 2x on each side must
latter shows a train of folds in a single layer of
be equal, requiring:
coarse K-feldspar/quartz leucosome sandwiched
dAdt Dxx A between two layers of quartzplagioclasebiotite
dAdt Dxx A 2[(1 R) (1 R)]Dxx A( A)
gneiss of somewhat different composition. The
layer is irregular, because either it formed that
(10.105)
way, in part, or because it was possibly stretched
Here we use the relation cos 2x cos2x sin2x in an irregular fashion before it was then folded
and have substituted for a2. These are a coupled in layer-parallel shortening. Of particular interest
set of two equations in the shape of the interface is the rough regularity in the arc lengths between
expressed in terms of a primary sinusoidal form fold hinges. Other features, such as the tendency
and a rst-harmonic form with one-half the wave- to nd quartz lling in the tightly appressed
length. Because we have not obtained a complete hinges and the fold forms are also of interest. The
solution accurate to terms proportional to (A)2, lobe-and-cusp forms seen in the mullion struc-
there will generally be terms missing in the re- tures are also approximately seen in the forms of
404 VISCOUS FLOW
(a)
(b)
Fig 10.14 (a) Single-layer fold in a leucosome layer

embedded in a fine-grained quartzplagioclasebiotite gneiss. the perturbation to the upper and lower surfaces
(b) Working drawing from which values of Larc and layer may be written:
thickness were measured. Photograph by R. C. Fletcher.
y h Bcos (x ) h B cos cos x
each layer surface; this morphology gives nearly Bsin sin x
uniform thickness, ignoring the irregularities. An
y h Bcos (x ) h B cos cos x
interpretation of the quartz llings is that the
B sin sin x (10.106)
layer pulled away from the gneiss, with the separ-
ation accompanied by precipitation of quartz, so
The parts of the perturbation at L 2/ that are
an open cavity never formed. Longer arc length
proportional to cos x and sin x do not interact in
folds on the left show less quartz inlling than
the case of folding in layer-parallel shortening,
the tight group of smaller arc length/thickness
which we study here. Hence, we examine only the
ratio folds on the right. The lower working
case of a pair of in-phase surface forms, those pro-
drawing indicates measures of the hinge-to-
portional to cos x. We then re-write (10.106) in
hinge arc lengths and local layer thickness that
the form:
were made in order to quantify the regularity in
this and other proles from this fold train giving
y h (Bb Bs) cos x
the data shown in Fig. 10.20.
y h (Bb Bs) cos x
A regular but not periodic train of folds is in-
itiated by a process of selective amplication Bb 12 (Bcos B cos ) (10.107)
acting on the slight initial irregularity present in
the surfaces of a layer. The multi-layer folds of Fig. Bs 12 (Bcos B cos )
10.1b are remarkably regular. The present analysis
will allow us to understand how such regularity Any pair of in-phase surface perturbations at a
comes about. given L may be decomposed into a buckle fold form
Irregularity on the two surfaces may be math- and a pinch-and-swell form with amplitudes Bb and
ematically decomposed into a set of waveforms at Bs (buckle and swell), respectively (Fig. 10.15).
different wavelength, L, which is a continuous These waveforms are selectively amplied when
variable. At each wavelength, the component in the slopes along the layer surfaces are small. Each
(a) z' A
In-phase
perturbation y
x
0 2h
z'' A
=
L
Fold form
(b)
+ (x, y) (x, y)
(L/2 + x, y) (L/2 x, y)
Pinch-and-swell
form
Fig 10.16 (a) Axes and parameters used in analysis of low-

Fig 10.15 Decomposition of in-phase perturbation into slope folding. (b) Symmetry of the velocity vector.
fold and pinch-and-swell parts.
Consider a layer of thickness H 2h of viscous

pair of component waveforms, on the two layer sur- uid of viscosity embedded in a viscous uid of
faces, at a particular wavelength, L, gives rise to an viscosity 1 (Fig. 10.16). As in the mullion analysis,
independent perturbing ow within the layer and we prescribe a basic state that exactly describes
surrounding medium, i.e. the components are the ow if the layer surfaces are perfectly plane
non-interacting. The perturbing ow associated and parallel. We then consider the approximate
with it causes the amplitude of each pair of com- solution for the perturbing ow when the two sur-
ponent waveforms to grow or decay. In layer-paral- faces of the layer are sinusoidal, with the same
lel shortening, growth always occurs, and is most amplitude and phase. The two surfaces are:
rapid at the dominant wavelength, Ld, better rep- h A cos x
resented as the ratio Ld/H, where H is the layer thick- (10.108)
h A cos x
ness. The rate of growth decreases for wavelengths
that are longer or shorter than Ld. We might guess The basic-state ow is taken to be uniform layer-
that fold arc lengths from one anticlinal or syncli- parallel shortening, the same as that for the
nal hinge to the next divided by the local layer mullion problem. The rates of deformation and
thickness would approximately equal Ld/H. stress in the lower and upper half-spaces of the
We now set up and solve the problem of cylin- embedding medium are equal. These are given in
drical folding in plane ow in layer-parallel short- (10.71) and (10.70). In the present case, we retain
ening. To go from the results to an interpretation superscript or subscript 1 to denote the upper
of the fold arc length/thickness statistics derived half-space and use no superscripts or subscripts
from fold trains or fold train segments such as for layer quantities. The symmetry of the
that shown in Fig. 10.14 requires additional steps problem will be exploited, so that it will not be
that will be discussed later. This interpretation necessary to deal with the lower half-space and
will yield an estimate of the ratio of the viscosity the boundary conditions at the lower layer/
of the host to that of the layer. Here, we postulate medium interface. In a demonstration of this the
that these materials may be adequately repre- superscript 2 will be used for quantities in the
sented as linear viscous uids. lower half-space.
406 VISCOUS FLOW
To exploit the symmetry seen in Fig. 10.16b, the mullion problem. However, we refer them to a
take the coordinate origin for the layer at its mid- coordinate system in which the x-coordinate is the
plane. The velocity vector is shown for four par- same as that for the layer, but the origin is moved
ticles in symmetric relation to each other. Vertical up an amount h to the mean interface position.
mirror planes of symmetry are present at at x 0, Then, the velocity components in the upper half-
L/2, and L/2, with centers of symmetry at x space are:
L/4 and y 0. The y-component of the velocity ~ x (c1 d1y)e
v (1) y sin x
vector for the perturbing ow in the layer then (10.113)
satises ~ v y(x, y) ~
v y(x, y), so ~v y is an even func- ~ y [c1 d1(y 1)]e
v (1) y cos x
tion of x. We accordingly use the general form The stress components in the upper half-space are:
(10.93). The center of symmetry requires
~
v y(L2 x, y) ~ v y(x, y). Using these conditions ~ xx 21[c1 d1(y 1)]e
(1) y cos x
we obtain: ~ yy 21[c1 d1(y 1)]e

(1) y cos x (10.114)
~ xy 21(c1 d1y)e
(1) y sin x
{[a b(y 1)]e y [c d(y 1)]ey}
cos[(2L)(L 2 x)] The coefcients in the solution are deter-
{[a b(y 1)]e y[c d(y 1)]ey} mined from the four boundary conditions at the
upper surface; these sufce to determine the
cos x
coefcients a, b, c1, and d1. The additional two con-
{[a b(y 1)]e y [c d(y 1)]ey} stants for the lower half-space, a2 and b2, are
cos x (10.109) related to c1 and d1 by symmetry conditions.
Approximations used in the mullion problem are
For this to hold at all x and y, the coefcients of used here, and the detailed computations are not
ey, ey, yey, and yey must be equal, requiring repeated. For example, in a term of the form ae,
c a and d b. Substituting into (10.93), the y- we take e eh(1 A cos x) eh. If no slip
component of perturbing velocity is: occurs at the interface:
~ v x(x, h) ~
~ v (1)
x (x, 0)
v y {a(ey ey) b[y(ey ey) (10.115)
~ ~
v y(x, h) v y (x, 0)
(1)
(ey ey)] }cos x (10.110)
Recalling that the origin for the upper half-space
Thus, ~ v y is an even function of y, and symmetry
is taken at the mean base of this medium to sim-
reduces the number of arbitrary coefcients from
plify algebraic expressions, these conditions yield:
four to two. The x-component of perturbing veloc-
ity is: a(eh eh) bh(eh eh) c1
a(eh eh) b[h(eh eh) (eh eh)]
~
v x [a(ey ey) by(ey ey)] sin x
(c1 d1) (10.116)
(10.111)
Continuity of the normal and shear tractions
The perturbing stress components in the layer are: requires:
~
yy(x, h) ~ yy (x, 0)
(1)
~
xx 2{a(ey ey) b[y(ey ey)
~
xy(x, h) 4Dxx( A) sin x
(ey ey)]} cos x
~ (1)(x, 0) 4 D ( A) sin x
xy 1 xx (10.117)
~
yy 2{a(ey ey) b[y(ey ey) (10.112)
These yield:
(ey ey)]} cos x
2{a(eh eh) b[h(eh eh) (eheh)])
~
xy 2[a(ey ey) by(ey ey)] sin x 21(c1 d1) 2[a(eh eh)
(10.118)
bh(eheh)] 4Dxx A
In the upper half-space, the perturbing velocity
and stress components are the same as those in 21c141DxxA
Coefcients c1 and c1 d1 may be eliminated

between these pairs of relations to give 1/100
(ab)(sinh h R cosh h) bh(cosh h
R = 1/50
R sinh h) 0 1
10
(10.119)
a(cosh h R sinh h) bh(sinh h 1/20
Sgn(Dxx)(2 + q)
R cosh h) (1 R)Dxx A
1/10
Here R 1 . Solving for the constants we
nd:
a (1M)[h(cosh h R sinh h)

(sinh h R cosh h)] 2(1 R)Dxx A
b (1M)(sinh h R cosh h)2(1R)Dxx A (10.120)

0
10
M k(1R2)(12k)[(1 R2)sinh k 2R cosh k] 1 2
100 10 10
Here k 2h H, where H is the full layer thick- log10(L/H)
ness.
Fig 10.17 Sgn(Dxx)(2q) versus log10(L/H) showing
The rate of growth of perturbation amplitude,
maxima, for R0.01, 0.02, 0.05, and 0.1.
from (10.102) and (10.110) is:
dA A plot of Sgn(Dxx)(2 q) versus log10(L/H) is

Dxx A 2{a cosh (h)
dt shown in Fig. 10.17 for R 1/100, 1/50, 1/20, and
b[h sinh h cosh h]} (10.121) 1/10. Maximum growth rate occurs for a compo-
Upon further substitution: nent at the dominant wavelength (Biot, 1961), Ld/H,
for which q qd. Since growth is exponential for
dA
dt
1
2k(1R)
(1R2)[(1 R2) sinh k 2R cosh k] each component and the components do not inter-
act as long as A 1, Ld/H provides an estimate of
Sgn(Dxx) |Dxx |A (10.122)
the fold arc length/thickness ratio that might be
seen in a natural or experimentally produced fold
train. When maximum slopes in the folding layer
The description of the evolution of layer shape for reach 10 to 15, the linear independence of wave-
the single sinusoidal perturbation is completed length components breaks down, current positions
by the relations dLdt Dxx L and dHdt Dxx H of fold hinges are locked-in, and folding contin-
from which: ues at approximately constant layer thickness.
Amplication for a 10% thickening of the layer,
dkdt 2Dxxk (10.123)
or for Dxxt 0.1, is exp[0.1(2 q)]. For
It may be more useful to think in terms of the amplication by a factor of 10, 0.1(q 2) 2.3, or q
growth rate in the maximum slope, A, because 21. This corresponds to a strong folding insta-
this is a dimensionless quantity and we perceive a bility, since a perturbation at the dominant wave-
fold chiey by its maximum limb dip or slope. length with initial dip of 0.01 radian (0.5)
Since d dt Dxx, we have: would reach a dip of 15 in about 15% layer thick-
ening. This requires a viscosity ratio R 1/50.
d( A) Folding instability is weaker for a viscosity ratio R
(2 q) Sgn(Dxx) |Dxx |( A) (10.124)
dt 1/20, and 25% layer thickening is required to
Here the quantity q is dened from (10.122) as: yield amplication by a factor of 10. For R 1/10 a
thickening of 46% is required.
2k(1R) If R 1, (10.125) may be expanded to leading
q (10.125)
(1R2)[(1 R2) sinh k 2R cosh k] terms in k to obtain:
408 VISCOUS FLOW
(a) 18 (b) 30
16
25
14
20
12
2 + qd
L d /H
15
10
10
8
5
6
4 0
2 1.5 1 0.5 0 2 1.5 1 0.5 0
log10(R) log10(R)
Fig 10.18 (a) Ld/H versus R; dotted curve is thin-plate

kd 2 (Ld H) (6R)13
approximation. (b) 2 qd versus R. (10.127)
qd (169)13R23
q2 / k2 2R
6
Rk
k (10.126)
Ld/H and qd are plotted in Fig. 10.18 from a numer-
ical evaluation of (10.125) and the approxima-
tions (10.127). The solutions for Ld/H and qd do not
Typically, only the rst two terms are given in the admit of values less than 2 for Ld/H, or of more
so-called thin-plate approximation (Biot, 1961), than moderately strong folding instability with
but the third term improves the approximation qd 10 for Ld/H less than about 10. Thus, the folds
(Fig. 10.17). The expression (10.126) sums two in Fig. 10.14 cannot be interpreted from these
effects. The part of the curve descending as L/H relations. It is necessary to follow amplication of
decreases is dominated by the term k2/6 in the fold components as layer thickening continues,
denominator, and is tied to the bending resis- and the value of L/H of a component decreases.
tance of the layer; the part of the curve descend- This is done by numerically integrating (10.124)
ing as L/H increases is dominated by the terms R using:
and is associated with resistance of the medium to 1
layer deection. At the dominant wavelength, the d ln ( A) Sgn(Dxx)(2 q)dt (10.128)
A
two effects are balanced to give a maximum
growth rate. with an initial value (0) A(0). The result for R
Using only the two terms of (10.126) we have: 1/20 at several values of layer thickening is shown
(a)
20
102
18
Amplification, l(t)A(t)/l(0)A(0)
16
14
Frequency
12
101 10
8
6
4
2
100
100 101 102 0
log10 (L/H) 2 3 4 5 6 7 8 9 10
Larc/H
(b)
Fig 10.19 Amplification of A as a function of L/H with 50
layer thickening H/H0 from 1.2 to 2.4 in increments of 0.2.
Trajectories of individual components are for initial values 45
L/H 2, 4, 6, . . . , 18, 20, 30, 40, 60, 80, and 100.
40
Cumulative number
35
in Fig. 10.19. The value Lmax/H of the component
30
with the maximum amplication is Ld/H at H/H0
1.1 and 6.5 at H/H0 2. Curves of amplication 25
versus L/H for individual fold components are the
20
lines rising to the left.
The above relations allow interpretation of 15
data collected from fold trains of the sort shown in 10
Fig. 10.14, if the layer and surrounding medium
may be approximated as uniform viscous uids. 5
The data set for this prole and others from the 0
same fold train is presented as a histogram of fold 100 Larc/H 101
arc length to thickness ratio and also a cumulative
frequency distribution in Fig. 10.20. An interpreta- Fig 10.20 Frequency distribution of Larc/H for fold train, an
tion of the data to estimate the viscosity ratio and example of which is shown in Fig. 10.14: (a) histogram, (b)
in this case the amount of uniform layer thicken- cumulative frequency.
ing H/H0 may be based on the variations of
maximum amplication and values Lmax/H at Lmax/H. Despite the notation, Larc is not a wave-
which it occurs, as given in Fig. 10.21. Details of the length but simply, in this case, twice the hinge-to-
argument are given elsewhere (Sherwin and hinge arc length in a fold train, a quantity that has
Chapple, 1968). Simulations of fold train formed a random distribution. For the fold train from
by combining wavelength components of specied which Fig. 10.14 comes, the average value is
amplitude derived from an initial white rough- Larc H 4.92. Thus, values of R and H/H0 will lie
ness amplitude spectrum (all components have along a contour close to that for the 5 contour in
the same initial slope) showed that the mean of the Fig. 10.21. From an assessment of the degree of reg-
fold arc length thickness ratio, Larc/H, is a good esti- ularity of folding in terms of the dispersion (stan-
mate of the value at maximum amplication, dard deviation/mean) of the distribution shown in
410 VISCOUS FLOW
3.5
currently active accretionary wedges like those of
200
40
60
400
0
2.4 1000
30
10
Taiwan and the Himalaya (Hilley and Strecker,
20 2004) and the Olympic Mountains of Washington
2.2 4 State (Batt et al., 2001; Brandon, 2004), or in
15 extinct wedges. Here, we formulate and analyze a
2.0 model for ow, deformation, and stress in the
4.5
H/H0
10 region of the backstop of an accretionary wedge,

1.8 using another solution for a viscous layer.
5 A model for an accretionary wedge based on a
1.6 5 layer of uniform thickness requires a substantial
5.5
approximation that dispenses with the tapered
1.4 6
form of the entire wedge. Thus, it cannot model
7 the toe region. Our goal here is to visualize the
1.2 2
8 processes occurring near the rear of an accre-
10 9
12 tionary wedge, where exhumation of rock passing
0 0.04 0.08 0.12 0.16 0.2 through the wedge takes place by a combination
log10(R) of horizontal thickening and concomitant uplift
Fig 10.21 Summary model results from which R and H/H0
and erosion, with or without near-surface exten-
may be estimated. sion. Figure 10.8c illustrates features that the
model will attempt to capture and we have earlier
used it to discuss the central concept of a backstop.
Fig. 10.20, the maximum amplication may be esti- Near the backstop, particle motion approaches
mated, and a value in the range from 15 to 40 is vertical. Vertical velocity at the surface acts to
indicated. The method is not given here. The cor- increase the height of the wedge, but is offset by
responding arc of the contour Larc/H 4.9 then erosion and tectonic denudation that produces
gives a range in R from 1/11.5 to 1/8.3 and a corre- thinning by ow or normal faulting. Material may
sponding range in H/H0 from 1.7 to 1.9. Thus, in be added to or removed from the wedge by under-
modeling rock masses containing coarse K- plating or basal erosion.
feldspar-rich pegmatite, granite, or gneiss in ner- The present model seeks to address the details
grained quartzplagioclasebiotite gneiss, the of wedge behavior in the region of exhumation
former may be given a viscosity approximately ten that cannot be modeled by lubrication theory,
times that of the latter. Evidently, a more extensive since what is involved is chiey sub-horizontal
and detailed study along these lines would provide compression or extension, not sub-horizontal
estimates that are more reliable and their uncer- shear. Treatment of the backstop as a vertical
tainty. mirror plane suggests that we treat the wedge as
a layer segment bounded by two such symmetry
10.4.4 Circulating cell model for a planes (Fig. 10.22), one of which must then be a
steady-state accretionary wedge frontstop. This is an unrealistic feature, but we
The lubrication theory model for a viscous accre- only apply the rear half of the model to ow near
tionary wedge presented in Section 10.3.4, from the backstop, the segment of width L/4.
which only the simplest steady-state example was A cylindrical topographic surface with ampli-
extracted, provides a basis for assessing the tude A slopes to the right, opposite to the direc-
inuence of factors such as rate of supply of mation of basal motion due to subduction drag. The
terial, rate of erosion, or viscosity on wedge model is composed of a homogeneous viscous
dynamics. This model is less useful in providing uid of viscosity and density . To complete its
details of internal deformation and trajectories of specication, we need to set the boundary condi-
rock volumes within the wedge. Such information tions that determine the internal ow. The condi-
is essential for interpreting eld observations of tions at the lateral bounding surfaces are the
deformation in the rocks exposed by erosion in symmetry conditions:
y boundary conditions are expressed in terms of

the total ow. The normal stress components xx
A
and yy in fact include the lithostatic basic state,
but to simplify notation, this is not written out.
H The appropriate forms of the velocity compo-
nents, (10.92) and (10.93), and the stress compo-
x nents, (10.90), are those used for the perturbing
ows in the previous models for mullions and
0
folds. The x-dependences of these are appropri-
L/2 ate, with vy proportional to cos x. Indeed, the
boundary conditions on subduction drag and
Fig 10.22 Circulating cell model.
underplating were chosen to conform to this
sinusoidal dependence. Again, by imposing con-
ditions that are somewhat articial, we obtain a
vx(0, y) vx(L 2, y) 0 description of wedge dynamics that is remark-
(10.129)
xy(0, y) xy(L2, y) 0 ably simple, and thus capable of providing
sharp, if restricted, physical insight into accre-
The conditions at the base of the model are those
tionary wedge dynamics.
that drive the ow, since any topography would
Development of the boundary conditions
decay under the action of gravity alone. We adopt
(10.130), (10.131), and (10.132) leads to the rela-
conditions that lead to a simple solution and ow
tions:
pattern. We earlier considered a condition in
which the base of the wedge adheres to the sub- a c U
ducting plate. However, estimated horizontal abcd W (10.133)
motions within a natural wedge are generally (a bk)ek (c dk)e k 0
much smaller than the subduction velocity. Here, [a b(k 1)]ek [c d(k 1)]e k gH A2k
we specify the basal velocity without relation to
subduction velocity as: Here, we have re-named the coefcients using a
prime. The dimensions of the coefcients are
vx(x, 0) U sin x (10.130)
those of velocity, as indicated by the rst pair of
Thus, coupling between subducting plate and relations in (10.133). Using the magnitude of sub-
wedge is taken to be greatest at the mid-point of duction drag, U, as the reference velocity, we
the cell, x L/4, from which it tapers off toward divide (10.133) through by it to obtain:
the backstop and the frontstop. We specify a dis-
ac1
tribution of underplating:
abcd
vy(x, 0) W cos x (10.131) (a bk)ek (c dk)e k 0 (10.134)
[a b(k 1)]ek [c d(k 1)]e k
This is maximum at the backstop and, forward of
the mid-point x L/4, corresponds instead to basal (gH2U)(A k)
erosion. Here a, b, c, and d are dimensionless.
At the upper surface, y H A cos x, and the Recall that accretionary wedges gain material
normal and shear traction vanish. These condi- by frontal accretion and by underplating. In this
tions are approximated by a boundary condition case, frontal accretion corresponds in a rough
on the plane y H where: sense to the ux of material across the vertical
xz(x, H) 0 plane at x L/4 (Fig. 10.22), where the ow is hori-
(10.132) zontal. Wedges loose material by erosion. In
zz(x, H) g A cos x
nature, wedges do not have a vertical backstop,
Note that in the present model the basic state is and material can ow within the wedge past the
one of lithostatic stress and no ow. Thus, the point of highest surface relief and erosion may
412 VISCOUS FLOW
take place on both sides of this. A very simple wedge divided by its mean or maximum thick-
modication of the present model may be made ness. The second dimensionless group is the ratio
to account for this, but we avoid the temptation to of the magnitude of underplating to that of sub-
introduce it in order to achieve the simplest duction drag. The third dimensionless group is
approximation to wedge dynamics. Moreover, the the ratio of the lithostatic stress to a quantity
effect of any such modication to a model is best with the dimensions of stress that is proportional
appreciated by rst studying the behavior before to the viscosity and K. The parameter K is associ-
it is introduced. ated with thinning of the wedge by erosion: it is
Prior to solving (10.134), we consider the not a rate of deformation, but has the same
special restriction to the case of a steady-state dimensions as one (time)1. The third dimension-
wedge, in which the amount of material lost by less group may be interpreted as the ratio of the
erosion and outward ow is just balanced by the relative outward ux of material by gravity-driven
inward ow, and the balance is achieved by time- glacier ow within the wedge to that by erosion.
independent regimens of inux, underplating, Note that gH/2 is a rate of deformation. Each of
and erosion. For the present model, this requires these dimensionless groups, then, has a clear and
that the vertical ux of material in the interval of concrete connection with major aspects of wedge
interest, 0 x L/4, and at the upper surface, is form and dynamics, and a connection with mea-
just balanced by erosion. We use a rate of erosion surable quantities such as wedge form, the ampli-
proportional to the local topographic relief tude of topographic relief, and rate of erosion or
which, in view of the sinusoidal variation of exhumation, or erosional ux. These parameters
topography, is described by: also establish the deformation within the wedge
and the strain observed in rocks exposed at the
(d Adt)erosion KA (10.135) surface.
Before solving for the coefcients, we obtain
In the steady state, this loss is just balanced by the one other result that emphasizes the simplicity of
ux associated with vy(x, H) over the same interval the model and, by inference, that of the dynamics
in x. This gives the condition: of a natural accretionary wedge when viewed at a
suitably large length scale and increment of time.
K Ass U{[a b(k 1)]ek [c d(k 1)]e k} 0 The velocity and stress distributions may be
(10.136) obtained if we know the stream function, ,
Here Ass is the amplitude of the steady-state topo- where vx y and vy x. For this case:
graphic relief.
Substitution of (10.136) into the fourth equa- UH(1H){[a b(y1)]ey
tion in (10.134) yields the set of equations: [c d(y 1)]ey} sin x (10.139)
ac1 We want to obtain a self-consistent set of velocity

abcd and stress components, or an equivalent expres-
(a bk)ek (c dk) ek 0 (10.137) sion of the stream function, that contain the
(1k)[a b(k1)]ek (1 k) dimensionless groups. Further, since accretionary
[c d(k 1)]ek 0 wedges are slender, in the sense that H/(L/2) 1,
or k 1, we seek polynomial expressions in z
This indicates that the ow and stress distribution
that contain only leading terms in k or z k. This
in the steady-state accretionary wedge is a func-
requires that we obtain a suitable approximate
tion of only three dimensionless parameters:
solution to the boundary conditions (10.137) and
a corresponding expansion of the stream function
k 2(LH), WU, gH(2K ) (10.138)
in (10.139).
The rst dimensionless group contains the aspect When all expressions are expanded uniformly
ratio of the wedge, (L/2)/H, which may be identied to terms proportional to (z)3, the results are not
roughly with the backstop-to-toe width of the self-consistent in the sense that the velocity
components do not exactly satisfy the condition

of incompressibility, and the stress components

UH
k 3y2 k
y 2
2k 1 1
y
3k sin x
do not satisfy the equations of stress equilibrium. (10.145)

For these conditions to be satised by the polyno-
mial approximations, several terms must then be While this approximation may be used to
dropped; the result is: compute the pattern of streamlines and other
quantities, its chief utility is in indicating the rel-
(U)[(abcd) y(a c) ( y22)
ative simplicity of the model dynamics and in
(a b c d) ( y 36)
obtaining closed-form solutions for quantities of
(a 2b c 2d)] sin x (10.140) interest. In the computations whose results are
given below, the exact relations, with coefcients
The displacement components are:
derived from (10.137) are used.
vx U[(a c) y(a b c d) Again, insofar as the model simulates the
(y22)(a 2b c 2d)] sin x dynamics of a natural accretionary wedge in the
vy U[(a b c d) y(a c) region in which exhumation takes place, only
(y22)(a b c d) three dimensionless parameters are involved,
(y36) (a 2b c 2d)] cos x (10.141) either (10.138) or k, , and . In addition, the
topography scales with the height 3U/gH, from
The stress components are: (10.144). It is of interest, then, to estimate these
xy 2U[(ac) y (a b cd) parameters from eld observations.
A further conclusion concerns the role of
(y22)(a 2bc 2d)] sin x
nite strain in the mechanics of this large-scale
yy 2U[(ab c d) y(ac)
tectonic structure. The nite deformation may be
(y22)(a b c d) computed, but it plays no role in the underlying
(y36)(a 2bc 2d)] cos x physics. Rheological properties are invariably
xx 2U[(a b cd) altered during deformation, but any model for
y(a 2bc 2d)] cos x (10.142)
this transformation would be described by a set of
evolution equations for their rates of change, and
It is somewhat easier to solve for the strain itself would enter as neither dependent nor
coefcients in the approximate solution by using independent variables. The computation of nite
the k 1 expansions in the boundary conditions strain, especially its distribution in outcrop, does
(10.134). The result is: provide a powerful means of constraining any
model for wedge dynamics.
a 12SA2k
Figure 10.23 compares streamlines computed
c 12 SA2 k (10.143) from the approximation and the exact solution
b SA2kSA22 12k
for the case 0.1, 0.2, and k /10. The exact
d SA2kSA2212k solution is obtained for a wedge of maximum
These satisfy the boundary conditions to within thickness H 20 km and a topographic-culmina-
residuals that are a small factor k2 times the tion-to-toe distance L/2 200 km. The approxima-
remaining terms. From the condition (10.136): tion is excellent; for other values of and it may
not be so good. The change in slope of the stream-
Ass (1k2)(3U gH)( k)(1 ) lines indicates a transition from motion toward
(W kU)(K k2gH3) (10.144) the backstop below about one-third of the wedge
(1k2)(3K gH) 3(2k2) thickness and ow away from it above that depth.
This implies a component of horizontal exten-
The dimensionless parameter is preferred over sion superposed on earlier shortening for rock
because it is zero, rather than innite, when the volumes traveling along the streamlines. Because
rate of erosion goes to zero. The approximate form the ow is steady, the streamlines are particle
of the stream function is: trajectories.
414 VISCOUS FLOW
0. 3
0. 2
2py/L
0. 1
0.0
0.0 0. 5 1.0 1. 5 2.0 2. 5 3.0
2px/L
Fig 10.23 Exact and approximate (higher in each pair)
streamlines for a circulating cell model.
for (k 3)(21). In the present case, k /10
and 1.2, and 0.15.
In Fig. 10.24, streamlines are plotted for four
Using the Olympic accretionary wedge as an pairs of and with k /10. In these gures, the
example (Fig. 10.8a), we may roughly estimate the small A/H 1 surface amplitude is not shown
model parameters (Batt et al., 2001; Brandon, and the upper surface is taken as nominally plane.
2004). A rate of erosion 1 km Ma1 is associated Figure 10.24a with 0.1 and 1 is a possible
with a topographic amplitude A 1.5 km, giving approximation for the Olympic accretionary
an erosion constant K 2/3 Ma1. Using a density wedge. If 0, strong horizontal extension
2700 kg m3, acceleration of gravity g 9.8 m occurs at the upper surface, with large magnitude
s2, and H 20 km, we obtain 1.2, much larger shearing of opposite sign corresponding to the
than the value used in Fig. 10.23. The ratio W/U inward and outward uxes of material in the
can be estimated from interpretation of observa- wedge ow. Vigorous underplating (Fig. 10.24c)
tions. Strain measured from rock at exposure eliminates much of the shearing associated with
gives no indication of extension, which is favored basal drag. Vigorous erosion (Fig. 10.24d) elimi-
by larger values of , i.e. of underplating. This pro- nates the near-surface extension.
vides an upper limit on . While this estimate may The nite deformation may be computed by
be established by examining model streamlines, it integration to determine the displacement gradi-
is simpler to examine the rate of deformation and ent tensor Fij along particle trajectories. In plane
strain of a particle underplated at the backstop. ow the evolution equations reduce to:
The approximate solution for the velocity compo- DFxx Dt Dxx Fxx (Dxy z)Fyx
nents provides convenient closed-form results:
DFxy Dt Dxx Fxy (Dxy z)Fyy
(10.147)
Dxx(0, y)
U
1
3
2k2
k
1
2y
y2
k (10.146)
DFyx Dt (Dyx z)Fxx DyyFyx
DFyy Dt (Dyx z)Fxy DyyFyy
Using L= 400 km, or twice the width of the back-
stop-to-toe distance, the magnitude of the rate of Here z 12(vy x vx y). Since the velocity
deformation, U, is 0.5 1014 s1 times the value components are known, integration may be
of U expressed in cm a1. The horizontal rate of carried out numerically. A simpler means of
deformation is a minimum (greatest negative obtaining an illustration of the strain ellipses is to
value) at the base of the wedge, and increases follow the deformation of an initial circle of par-
upwards. Near-surface extension occurs for any ticles by incrementing their positions for a succes-
greater than a value which gives zero at y k, or sion of small time steps, but this operation does
(a) components of the displacement gradient tensor

are Fyy 1/Fxx. We then have:
DFyy Fyy Fyy

vy Dyy Fyy (10.148)
Dt t y
(b) This is a special case of the general relations given

in (10.147). Since the ow is steady, Fyy t 0 and
since vy(0, y) and Fyy(0, y) are only functions of y,
(10.148) reduces to:
dFyy dvy 1 1
(c) vy Dyy Fyy F , or dFyy dvy
dy dy yy Fyy vy
(10.149)
Integrating between y1 and y2, we have:

L/20
(d) Fyy (y2) vy( y2)
(10.150)
Fyy( y1) vy( y1)
The total strain of a particle rising from the base

0 L/4 of the wedge as a function of its height is:
Fig 10.24 Streamlines for four sets of model parameters: Fyy(0, y) vy(0, y)W
(a) 0.1, 1; (b) 0, 0; (c) 0.1, 0; (d)
0, 1.

1

3
y 2k2
k
1
y2
y3
3k
(10.151)
not directly supply the tensor F of the center point.
A result for 0.1, 1, and k /4 is shown in The ratio of vertical to horizontal axes of the
Fig. 10.25. In preparing the gure, a dense set of strain ellipse, or the quadratic elongation, is (Fyy)2.
particles lying on a circle of very small radius was This is plotted versus the dimensionless height
followed to assure an adequate approximation to from the base of the wedge, y/H y/k in Fig. 10.26
homogeneity; the resulting strain ellipse was then for k /10, 1, and 0.1 0.15, 0.2, 0.4, and 1.
enlarged by a factor of about 10. Only the deforma- The maximum on the curve indicates the height
tion after the element has been underplated can be at which the horizontal rate of deformation is zero,
evaluated, as indicated by the circular locus at the and goes from shortening to extension. The qua-
base of the wedge. Although we do not expect that dratic elongation is then a maximum. If the curve
near-surface ductile ow as in this model will be crosses the line (Fyy)2 1, the initial vertical elonga-
present in natural accretionary wedges, pressure tion is cancelled out, and the strain thereafter is
solution may result in ductile deformation to vertical shortening.
within a few kilometers of the surface. If material In the simple circulating cell model for the
is then not strongly deformed as it moves upward, region of an accretionary wedge about one-third to
as indicated by the streamlines of Fig. 10.24a, the one-half its length from its culmination a complete
model strain distribution will still be comparable picture of wedge dynamics may be established.
to a natural outcrop pattern. This depends on only a few dimensionless groups:
An impression of strain magnitude near the k, , and , plus the scaling quantities U for veloc-
backstop and its dependence on the dimension- ity, U for rate of deformation, and 3U/gH for
less parameters may be obtained from a closed- topographic amplitude. Semi-quantitative ts may
form expression for the strain of a particle likely be obtained between model parameters and
underplated at the backstop and rising vertically observed and estimated quantities from a natural
along it. At the backstop, the only non-zero accretionary wedge.
416 VISCOUS FLOW
0.4
0.3
2py/L
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1
2px/L
Fig 10.25 Strain ellipses for particles rising along geneous functions of the components of stress.
trajectories at 0, 0.2, 0.4, 0.6, and 0.8 of the half-cell width
We then write:
for the case 0.1, 1.
Dxx a11xx a12yy a13zz a14yz
10.5 Flow of anisotropic viscous a15zx a16xy
fluids Dyy a12xx a22yy a23zz a24yz

a25zx a26xy
Rocks composed of layers, such as sedimentary Dzz a13xx a23yy a33zz a34yz
and metasedimentary rocks, those made up of
less-continuous layers or of lenticular bodies, a35zx a36xy
(10.152)
such as gneisses and some sedimentary rocks, and 2Dyz a14xx a24yy a34zz a44yz
those with a strong foliation, compositional
banding, and alignment of mineral grains, will a45zx a46xy
exhibit anisotropic rheological behavior. Such 2Dzx a15xx a25yy a35zz a45yz
rocks often contain structures, such as folds, at
a55zx a56xy
several scales. The dimensions of the smallest
structures scale with the thickness of individual 2Dxy a16xx a26yy a36zz a46yz
layers, lenses, or even mineral grains, as in mica
a56zx a66xy
schists. The dimensions of larger scale structures
may be tens to hundreds of times the thickness of This notation follows that in Lehknitskii (1963),
components, as are the chevron folds described in who has written a useful treatise on the deforma-
Chapter 5. We now formulate constitutive relation of anisotropic elastic bodies. The relations
tions for an anisotropic viscous uid. This will (10.152) are equivalent to those for the elastic
then be used in a mechanical model for chevron solid, but with the strain components, ij,
folds. replaced by the rate of deformation components,
Dij. The form of the coefcients is based on the
10.5.1 Constitutive relations for identications: 1 with xx, 2 with yy, 3 with zz, 4
anisotropic viscous fluids with yz, 5 with zx, and 6 with xy.
As in the case of an anisotropic elastic solid The symmetric aij matrix from (10.152) has (6
(Chapter 8), formal derivation of general relations 6)/2 3 21 distinct elements. Since the material
may start with the postulate that the components is incompressible, (10.9) applies, requiring the six
of the rate of deformation are linear and homo- conditions:
10.5 FLOW OF ANISOTROPIC VISCOUS FLUIDS 417
ow is a mirror plane of symmetry. Let plane ow

4 take place in the (x, y)-plane for the orthorhombic
material of (10.154). We may then eliminate the
a=0.10
Quadratic elongation, (Fyy)2
normal component of stress zz by the condition

that Dzz 0, or from (10.154):
3
0.15 a a
zz 13 xx 23 yy (10.155)
a33 a33
2 0.20
Using this in the equations from (10.154) for Dxx
and Dyy, and retaining the relation for Dxy, we have:
0.40
1 Dxx b11xx b12yy
1.00 Dyy b12xx b22yy (10.156)
0 2Dxy a66xy
0 0.2 0.4 0.6 0.8 1
z/H Here b11 a11 a213 / a33 and b12 a12 a13a23 / a33. Evo-
Fig 10.26 Quadratic elongation (a/b) for particles rising king incompressibility, we obtain b11 b12 0 and
along the backstop, for 1 and 0.1, 0.15, 0.2, 0.4, and b12 b22 0, so b22 b12 b11. Substituting for the
1.
two independent constants in (10.156) the consti-
tutive relations for plane ow are:
ai1 ai2 ai3 0, i 1, 2, 3, 4, 5, 6 (10.153)
Dxx (xxyy) 4n
The number of independent coefcients for a
Dyy (xx yy) 4n (10.157)
general anisotropic viscous uid is reduced from
21 to 15. Dxy xy 2s
Following the discussion of anisotropic elastic
Here n and s are the principal viscosities for
materials with certain symmetry restrictions, we
plane ow in the (x, y)-plane (10.154). For plane
write for an orthorhombic material with three
ow in other symmetry planes, the numerical
mutually perpendicular mirror planes of symme-
values of these quantities would change.
try, initially setting aside the conditions (10.153):
We may further suppose that the material of
interest is transversely isotropic, with equivalent
Dxx a11xx a12yy a13zz
directions x and z, so that only the y-axis is a
Dyy a12xx a22yy a23zz unique direction within the material. The
coefcients in (10.154) then have the additional
Dzz a13xx a23yy a33zz
(10.154) restrictions:
2Dyz a44yz
a23 a12, a33 a11, a44 a66, a55 2(a11a13)
2Dzx a55zx
(10.158)
2Dxy a66xy
The last condition arises from isotropy in the (x, z)-
This simpler form of the constitutive relations plane, i.e. that the constants do not change for
only applies when the axes x, y, and z are principal such a material when the coordinate axes are
axes of anisotropy, xed in material directions changed by an arbitrary rotation about the y-axis.
coinciding with the intersections of pairs of An incompressible, transversely isotropic aniso-
mirror planes in this case. tropic viscous uid is then described by only two
We will investigate a plane ow problem for an parameters, the principal viscosities n and s of
incompressible anisotropic viscous uid. One (10.157).
requirement for plane ow is that the plane of Such a material may be thought of as an
418 VISCOUS FLOW

1
approximation to a layered medium made up of f1 f2
s (10.165)
alternating layers of stiff and soft isotropic 1 2
viscous uid when the scale of the ow of interest
The ratio of the principal viscosities is:
is much greater than the layer thickness, as in the

case of chevron folds. Suppose the layers have n f f
thicknesses h1 and h2 and viscosities 1 and 2, and m ( f11 f22) 1 2 (10.166)
s 1 2
take the axes x and z parallel to layering so the y-
axis is normal to the layering. The bulk properties This is a measure of the degree of anisotropy.
of this composite material are obtained as follows. Inspection shows that m 1. It may be shown that
In layer-parallel extension, the rates of extension the maximum of m occurs at f1 f2 0.5.
in the two layer types must be equal. In the two in- The full constitutive relations for the trans-
plane directions: versely isotropic anisotropic uid, in principal
coordinates, are:
xx D xx Dxx, D zz D zz Dzz
D (1) (2) (1) (2) (10.159)
For simplicity, consider plane ow in the (x, y)- Dxx xx 13 (xx yy zz) 2n
plane, so that for example: Dyy yy 13 (xx yy zz) 2n
(10.167)
Dzz zz 13 (xx yy zz) 2n
D(1)
zz (1) 1
zz 3 ( (1) (1)
xx yy zz ) 21 0
Dyz yz 2s, Dzx zx 2n, Dxy xy 2s
zz ( xx yy) 2
(1) (10.160)
(1)
Then, developing the rst of (10.159): 10.5.2 Chevron folding of an anisotropic

( (1)
xx yy) 41 ( xx yy) 42 (xxyy) 4n
(2) viscous fluid
(10.161) In Chapter 5, we studied several kinematic models
for chevron folding in a stack of layers of equal
The components of normal stress acting normal thickness between which slip could take place.
to the interface are equal. The last expression in Here, we consider a simple but complete mechan-
(10.161) identies a bulk layer-parallel normal ical model that makes use of the continuum
stress xx and a bulk viscosity in layer-parallel approximation of a nely layered medium by an
shortening or extension n. The bulk normal anisotropic viscous uid. We will be able to
stress component parallel to the layering is: extract, by a separate procedure, the amount of
inter-layer slip as a function of limb dip as well as
xx h2 xx
h1 (1) (2)
xx (10.162) the deformation of the layers. In the continuum
h1 h2 approximation of this material it is not possible to
Combining (10.161) and (10.162): treat in detail the deformation in the region of a
fold hinge, whose dimension will be a few times
h11 h22 the thickness of an individual layer. The approxi-
n f11 f22 (10.163)
h1 h2 mation to chevron folding is supported by the
Here f1 and f2 are the thickness (volume) fractions large fraction of the rock volume occupied by the
of the alternating layer types. The bulk viscosity in straight fold limbs. Recall that the continuum
layer-parallel shear is obtained by taking the approximation only applies to ow at a scale that
average of the rate of shear: is much greater than the individual layer thick-
ness. However, for the homogeneously deforming
Dxy f1D (1)
xy f2D xy f1
(2)

2
xy
1
f2
2
xy
2

xy
2s
fold limbs, an exact interpretation of the macro-
scopic solution may be made to give the quantities
for component layers or surfaces.
(10.164)
As in Chapter 5, consider the evolution of sym-
Rearranging, the bulk viscosity in layer-parallel metric, periodic chevron folds formed by shorten-
shear is: ing normal to their axial planes (Fig. 10.27). The
10.5 FLOW OF ANISOTROPIC VISCOUS FLUIDS 419
y sxx (xxyy) 2, sxy xy (10.169)
Consider the deformation of a segment of the

trace of a folded surface on the left-dipping limb
of the fold (Fig. 10.27) whose projection onto the
x-axis is taken to be 1 in arbitrary units. The rate
of change in limb dip as well as in the length of
the fold limb may be obtained by considering the
velocity of the particle on its end:
vx(1, tan ) Dxx

(10.170)
vy(1, tan ) Dxx tan 2Dxy
From Fig. 10.28, the new limb dip, d, after an

innitesimal time dt is given by:
tan vy (1, tan )dt

tan ( d) (10.171)
1 vx(1, tan )dt
Combining (1.170) and (1.171) we have:
d d
tan u tan (1 tan 2) 2(Dxy Dxx tan )
dt dt
u
x (10.172)
1 Symmetry requires that the shear component

of the deviatoric stress sxy equals zero. Thus, the
Fig 10.27 Portion of a set of periodic chevron folds for a relations in (10.168) reduce to:
stack of layers with inter-layer slip; shown are 1/2 the spans
of the right-dipping and left-dipping limbs and the fixed (x, y)- 4nDxx [(1 m) (1 m) cos 4]sxx
coordinate axes, and the limb dip . 4nDxy (1 m) sin 4 sxx (10.173)
The rst of (10.173) is used to eliminate sxx from

constitutive equations for a transversely isotropic
the expression for Dxy, and substitution of the
anisotropic viscous uid are given in (10.157). We
result into (10.172) yields:
also showed how these could be identied as the
bulk constitutive relations for a layered material d d
consisting of alternating stiff and soft isotropic tan (1 tan 2)
dt dt
viscous layers. Now let x and y denote the princi-
pal axes of anisotropy as in (10.157) and take the
x-axis xed in the direction of bulk shortening
2
(1 m) sin 4
[(1 m) (1 m) cos 4]
tan Dxx
(10.174)
normal to the fold axial plane. The constitutive
relations for the homogeneous material in the This relation may be integrated numerically to
left-dipping fold limb are then, by transformation follow the evolution in limb dip with shortening.
of coordinates: As in Chapter 5, it is necessary to introduce a peri-
odic set of seed folds with initial limb dip 0.
4nDxx [(1 m) (1 m) cos 4 ]sxx
Examples of fold evolution in terms of limb dip,
(1 m) sin 4sxy , for several values of the viscosity ratio, m n/s,
4nDxy (1m) sin 4sxx are given in Fig. 10.29 for m 1, 1.5, 2, 3, 4, 10, and
. Note that fold growth corresponds to a decrease
[(1 m) (1 m) cos 4]sxy (10.168)
in fold span and so progress is from right to left on
Here the stress components are the deviatoric this gure. For the isotropic uid, m 1, fold
stress components dened as: growth is the passive kinematic amplication of
420 VISCOUS FLOW
y vy (1, tan u)dt

60
vx (1, tan u)dt

50
m = infinity
Limb dip, u (o)

40
tan u 30 2 3 4 10
du 1.5
20
u 1
x 10
1
0
Fig 10.28 Change in angle and length of layer segment 0.5 0.6 0.7 0.8 0.9 1
after a time increment dt. Fold span (S/S0)
Fig 10.29 Limb dip of chevron folds as a function of

a surface in homogeneous shortening. In the deck current to initial fold span S/S0 for m 1, 1.5, 2, 3, 4, 10, and
of cards limit, m , where the normal viscosity .
greatly exceeds the shear viscosity, the component
layers do not change their thickness. A modest
value of m 1.5 leads to signicant fold derived and applied to natural examples: glacier-
amplication, and a value m 10 behaves nearly like ow driven by gravity; the -pression part of
like the limiting case of rigid layers. Since the transpression; the gross dynamics of accretionary
notion that natural chevron folds begin with limb wedges; and deformation of layers that have
dips as low as 5 is not rmly established (Fletcher roughly planar surfaces as in mullion structure,
and Pollard, 1999), the relation between limb dip folding, and ow within the exhumation region
and fold span is not necessarily that characteristic of an accretionary wedge. Other models for ow
of the natural folds. None-the-less, the methodol- in a viscous uid may be obtained by taking
ogy described here illustrates the application of advantage of the correspondence principle to
anisotropic viscous ow and suggests a fruitful convert elastic solutions presented in this text.
path for research on the behavior of layered geo- An important conclusion is that nite strain,
logical materials. while it plays a prominent role in modern struc-
tural investigations, does not directly enter the
formulation or development of forward models of
10.6 Concluding remarks ductile rock deformation, which are centered on
the velocity and stress elds. Evidently, nite
In this chapter, we have used the constitutive re- strain plays a signicant role, as in the interpreta-
lations for a linear (Newtonian) viscous uid to tion of natural single-layer folds, in constraining
complete a mechanical description for the large, such models. However, the many attempts to forge
permanent ductile deformation of rock, both models implicitly or explicitly involving causal
isotropic and anisotropic. Models have been aspects in terms of nite strain or other attributes
restricted to plane ow, for which solutions are of the deformed state, as embodied in the words
readily obtained and the results can be under- the strain was partitioned . . . or in the common
stood from illustrations of a cross section. misconception surrounding strain compatibil-
Another set of solutions may be obtained for ity, must prove to be essentially vacuous. Experi-
antiplane ow, the equations for which are only ence with forward models of rock deformation
introduced here. Solutions for plane ow are may result in a re-examination of this issue.
Chapter 11
Rheological behavior
Pressure solution in sandstone from the Olympic No mathematical theory can completely describe the
accretionary wedge, Olympic Mountains, NW Washington complex world around us. Every theory is aimed at a
State, USA. The horizontal dimension is 10 mm. certain class of phenomena, formulates their essential
Photograph by J. M. Rahl. features, and disregards what is of minor importance.
The theory meets its limits of applicability where a dis-
regarded inuence becomes important (Flugge, 1967).
422 RHEOLOGICAL BEHAVIOR
11.1 Departures from linear In one instance (Groshong, 1975), layer thick-
ening associated with a well-developed set of folds
viscous flow in a limestone layer embedded in shale was only
10%. Interpretation of data for these folds
L
aboratory determinations of steady-state rock (Fletcher, 1974) is consistent with the type of non-
creep and eld observations indicate that linear relation observed in the laboratory for
ductile rocks may not be well-approximated creep of carbonate rock. It has been suggested that
by homogeneous, isotropic, incompressible, and the limited range in fold arc length/thickness
linear viscous uids of uniform viscosity. In this ratios 4 to 6 is a manifestation of highly non-
chapter, we consider other constitutive relations linear behavior (Smith, 1977, 1979). Further evi-
that broaden the range in rock behavior described dence for markedly non-linear behavior is offered
and provide a basis for understanding some of the by examples of necking in layer-parallel exten-
differences arising. Because the linear Newtonian sion, which cannot occur in a Newtonian layer.
viscous uid is the simplest material that under- Alternatively, interpretation of the relative vis-
goes large permanent deformation, the formula- cosities of component rocks in a deformed con-
tion and analysis of models using it and their glomerate has been offered as evidence of linear
application to interpret a set of eld observations viscous behavior (Treagus and Treagus, 2002).
is always a useful rst step (see Chapter 10). The Motivated by these questions, we study necking
results obtained establish a benchmark from and folding of layers of non-linear power-law uid
which to understand differences in behavior asso- in this chapter.
ciated with other constitutive relations. Whether Inhomogeneity in rock masses occurs at a wide
the new model results in a large or subtle contrast range in scale, from less than a grain diameter to
in behavior relative to an already well-understood the tens or hundreds of kilometers appropriate to
viscous model, we will achieve a better under- the rst-order dynamics of crustal deformation. A
standing of the reasons for the differences and a useful postulate is that at the scale of interest the
greater condence in data interpretation. rock may be treated as a continuum whose behav-
For example, initiation of a regular train of ior is approximated by constitutive relations con-
folds by selective amplication was studied for the taining only a few rheological parameters. The
Newtonian viscous layer in Chapter 10. Data from question then arises as to how these parameters
natural fold populations yield a limited range in might be estimated from the three major deter-
mean fold arc length/thickness ratios of about 4 to minants of the bulk behavior: the volume frac-
6. Interpretation of the data in terms of the tions of the signicant mechanically distinct
folding of a Newtonian layer implies: (i) a modest components, their individual constitutive re-
layer to host viscosity ratio, 1/R, of about 10 to 20; lations, and the phase geometry of the composite
(ii) a weak folding instability, so that a large layer- material. Answers to this question would provide
parallel shortening is required to establish a a better understanding of the degree of complex-
regular fold train, e.g. a doubling of layer thick- ity of rocks and a basis for accepting or maintain-
ness. On the other hand, laboratory studies of ing doubts about the approximation of a locally
steady-state rock creep suggest that much larger homogeneous continuum with spatially varying
contrasts in effective viscosity should be common properties. In this chapter, we use elementary
(Table 11.1), with equivalent 1/R values of about methods to analyze composites made up of two
100 to 1000 or more, and that constitutive rela- isotropic viscous components. Composite materi-
tions for rock creep should generally be non- als must generally exhibit anisotropy in their rhe-
linear. Do the laboratory results extrapolate to ological behavior, with isotropic behavior being
natural rock behavior at much smaller rates of only a special case. We therefore include consid-
deformation and lower temperatures? Answers to eration of anisotropy.
this question not only affect conclusions as to Anisotropic materials such as foliated or
small-scale folding, but also the modeling of large- layered rock may have nearly uniform properties
scale crustal deformation. with respect to their principal axes of anisotropy.
11.2 BOUDINAGE AND THE NON-LINEAR POWER-LAW FLUID 423
Table 11.1. Rheological constants for a few quartzites and carbonates.
Material n log10 A0 Q log10 B log10eff

(MPan s1) (kJ mol1) (MPan s1) (Pa s1)
Quartzite (d) 2.8 5.463 184 19.1 23.0

Quartzite (d) 2.9 5.30 170 17.9 21.7
Quartzite (w) 2.6 1.35 230 18.6 22.7
Quartzite (w) 1.8 2.54 151 13.9 18.8
Quartzite (w) 4.0 9.4 135 19.0 21.7
Marble 8.3 3.9 260 22.2 20.6
Limestone 3.4 3.4 298 18.9 20.7
Deformation of a rock mass containing small in Britain and America, the smaller type of sausage is
random perturbations in the orientation of the more common, and these are seen hanging in strings,
principal axes may give rise to their strong selec- end to end. Transverse sections of non-equidimensional
tive amplication leading to folding, internal boudins remind the unwary of the latter. This misinter-
boudinage, or other structures (Cobbold et al., pretation of Lohests original description has unfortu-
nately been made in at least two papers (Wilson and
1971). Here, we examine the internal instability in
Cosgrove, 1982).
plane ow of an anisotropic viscous uid.
A third deviation from the Newtonian uid is Necking as exhibited in pinch-and-swell structures
through compressibility that is mediated by the in stiff layers (Fig. 11.1a) is an example indicating
transport by intergranular diffusion or by Darcy rheological non-linearity in natural rock deforma-
ow of a mobile component one that is soluble tion. The ability to work molten glass to produce
in an intergranular pore uid or acts as a weak sheets of uniform thickness depends on the
uid in a much stiffer solid framework, as in a absence of a necking instability in a linear viscous
partly melted rock. In this chapter, we consider uid. In structural geology, boudinage includes
the former alternative, and model some of the the continuous necking that produces pinch-and-
effects that arise from it in folding and necking swell structures (Fig. 11.1a) and the processes that
when the medium containing a stiff layer exhibits produce discretely segmented boudins (Fig. 11.1b,
this kind of behavior in combination with c). While discretely segmented in cross section,
Newtonian viscosity. they may coalesce in the axial direction. Discrete
All deviations from the homogeneous New- boudins may often have undergone an initial
tonian viscous uid lead to effects that produce episode of continuous necking. Two mechanisms
observable features in the structures and internal of segmentation occur: mode I cracks that con-
fabrics of deformed rocks, and they are of sub- temporaneously ll with precipitated minerals
stantial interest in application to the interpreta- (Fig. 11.1b), and mode II faults or shear bands (Fig
tion of eld data. 11.1c). After separation, boudins tend to undergo
further deformation that results in a variety of
striking forms. The modeling of boudinage pre-
sented here is restricted to continuous necking.
11.2 Boudinage and the non-linear It is not clear whether examples of necking of
power-law fluid rock layers are chiey a consequence of non-
linearity derived from strain rate thinning, in which
Confusion with regards to the signicance of the term the rate of deformation increases more rapidly
boudin has undoubtedly arisen in some countries than linearly with the deviatoric stress, or that
because of the different ways in which sausages are dis- derived from strain softening, in which an element
played in shops. On the continent of Europe, large weakens as it deforms, since both contribute to
boudins are found lying side-by-side on grocers slabs; necking (Neurath and Smith, 1982). We rst give a
(a) A model for large-scale crustal necking is pre-

sented to illustrate the roles of gravity and
surcial processes acting on topography.
11.2.1 Plane-sections-remain-plane
analysis for a free plate
Let the material be the isotropic incompressible
power-law uid already treated in the glacier-like
ow of Chapter 10. Consider a free plate of cylin-
drical form whose thickness varies as in Fig. 11.2.
It is not necessary to assume such a restricted
form, but we do suppose in the present case that
layer thickness varies slowly with length and that
(b)
the thickness variation is small relative to the
mean thickness. For this periodic shape, these
requirements are:
A 1 and A h 1 (11.1)
Here 2/L is the wavenumber and A is the

amplitude of the sinusoidal perturbation. We
later use this conguration to study initiation of
necking in an embedded layer. Here, though, we
more generally consider the deformation of a
layer of thickness H(x, t) that satises the equiva-
(c) lent conditions:
H(x, t) H(t) H~ (x, t)

(H~ ) H~
1 and 1 (11.2)
x H
~
Here H is the mean thickness and H is the devia-
tion from it. The thickness variation of the plate
is also cylindrical with a constant prole in the
layer direction normal to x. The plate undergoes
plane deformation when subjected to an axial
force per unit depth of constant magnitude Fx 0.
Fig 11.1 (a) Continuous necking in a gneiss layer 6 cm in As here, the use of H for the full thickness of a
thickness. (b) Discrete boudins separated by deformed layer, but also h for the half-thickness, is done to
quartz precipitated into the boudin gaps; hammer length avoid having to write H/2 in boundary conditions
40 cm. (c) Boudins deformed and partially separated by
when the coordinate origin is taken at the center
right-dipping normal faults; horizontal span 10 m.
of a layer to exploit symmetry.
The mean axial normal stress on a vertical
surface through the plate is:
simple analysis of necking of a free plate, or
unconned layer. We then show how selective Fx
xx(x, t) (11.3)
amplication occurs in layer extension, giving H(x, t)
necking, and, in layer-parallel shortening, giving This stress is taken to be uniform on a vertical
folding, in a layer of homogeneous and isotropic surface across the plate. We only treat the case of
power-law uid embedded in a weaker medium. a free plate with zero traction on its surface.
y Linearizing the relation (11.7) for a perturba-

~
tion in thickness H about a mean value H yields:

A ~
dH dH
[Hn1 (n 1)Hn2H~ ] C (11.8)
dt dt
x Expansion and separation of mean and perturb-
2h
0 ing parts gives:
n1 dH
H C
dt

~
1 dH dH
~
(n 1)H 0
L H dt dt
(11.9)
~
dH
Fig 11.2 Sinusoidal pinch-and-swell perturbation. (n 1)Dxx H
~
dt
The last relation indeed shows that the perturba-

As a second related approximation, we adopt tion H~
does not grow for a viscous layer, n 1, and
the familiar plane-sections-remain-plane approxi- grows only slowly unless n is large. Values of n
mation used in engineering (Timoshenko and obtained in the laboratory for high-temperature
Young, 1968) or perhaps formerly used when steady-state creep of rock tend to be modest, typi-
numerical codes were not readily available. Each cally ranging from 3 to 5 (Table 11.1), so that only
vertical section of the plate thus is treated as a weak necking instability arises from the non-
undergoing homogeneous extension: linear relation between the rate of deformation
1 H and the deviatoric stress. For the sinusoidal per-
Dxx (11.4) turbation (Fig. 11.2) comparison with (11.9), with H~
H t
~
2h and H A cos x, yields:
The applicable single-component relation for a
power-law uid relates Dxx and the deviatoric dA
(n 1)Dxx A (11.10)
stress component, sxx, which for plane ow is: dt
We may use this approximation to follow the
1 1F
sxx xx x (11.5) change in shape of the layer segment (Fig. 11.2) by
2 2H
completing (11.10) with relations for the changes
The constitutive relation is: in wavelength L and mean thickness 2h. These are:
Dxx B(s2xx) (n1) 2 sxx (11.6) dL
DxxL
dt
Substituting (11.4) and (11.5) into (11.6) we have: (11.11)
dh
Dxx h
n dt
Dxx
1 dH
H dt
B 1 Fx
2 H
C
n
H (11.7) The relations (11.10) and (11.11) are only appropri-
dH ate under the conditions (11.1). The initial and
so, Hn1 C
dt nal shapes for a stretch of 1.29 for a material
This is the relation for change in vertical dimen- with n 5 and an initial conguration with h(0)
sion of an innitesimal material element of cross- 1 and A(0) 0.1 is given in Fig. 11.3. This example
sectional area dA Hdx and thus the derivative is looks good, but amplitude growth from (11.10)
a material time derivative. We could use the con- may exceed layer thinning, so that the maximum
stancy of area dA, expressing conservation of mass thickness of the nal form ends up larger than in-
for an incompressible material in plane ow, to itially! If n 10 is used in this example, such a
establish the relative positions of sections along result is obtained. One has to be careful with
the layer, but this problem is set aside. approximations.
y A convenient isotropic non-linear form for the

constitutive relations (Nye, 1953; Calladine and
Drucker, 1962) is:
x Dij BJ2(n1) 2sij, J2 12 skl skl,
sij ij 13 kkij (11.13)
Here the scalar J2 is the second isotropic invariant

Fig 11.3 Evolution of an initial pinch-and-swell of the deviatoric stress. Thus, all quantities in
perturbation in a free plate after a stretch of 1.29.
(11.13) are independent of the orientation of the
coordinate axes with respect to the material, and
The expectation for a layer embedded in a the relations are those for an isotropic material.
medium is that the degree or strength of instabil- We may write, in analogy with a Newtonian
ity will be less than that for the mechanically viscous uid:
unconstrained layer. In part to study this aspect, 1
we formulate a complete tensor characterization Dij s
2eff ij (11.14)
of a power-law uid and use it to analyze necking
2eff [BJ (n1)
2
2]1
and folding in the manner of the treatment given
in Chapter 10. Here eff is an effective viscosity that is generally
position dependent through the local value of the
11.2.2 Constitutive relations for an stress. A comparison of the form (11.13) with the
isotropic power-law fluid and a experimental relationship (11.12), reversing the
plastic solid usual sign convention used for experiments, is
A general tensor relationship between the compo- obtained as follows. Taking x1 for the direction
nents of the rate of deformation and those of the along the cylindrical sample axis, we have:
deviatoric stress is needed to treat arbitrary defor-
mations of the power-law uid. In experiments
s1 1 13 (1 2 3) 23 (1 3)
(Evans and Kohlstedt, 1995) on right-circular cylin-
ders of rock under axial stress, 1, generally com- (11.15)
s2 s3 13 (1 3)
pressive, and uniform radial or circumferential
stress, 3, acting on the surface of the cylinder,
J2 12 (s21 s22 s23) 13 (1 3)2
results are usually given in the form:
Then:
d1
dt
A0 exp

Q
R T 1
( 3) n (11.12)
D1 B 13 (1 3)2
(n1) 2
2
3 ( 1 3)
.
Here d1dt 1 is the axial strain rate. Because
A0 exp (QRT)(13)n
experiments are conducted in compression, the
experimentalists treat a rate of shortening and a
B [(3)(n1)2 (32)] A0 exp(QRT) (11.16)
compressive stress as positive. The constant pre-
ceding the power of the difference between axial The coefcient B is expressed in units (MPa)n s1
and circumferential normal stress components is in terms of constants given in the literature.
written as the product of a pre-exponential con- Constants for several polycrystalline quartzites
stant, A0, and a term expressing temperature and marbles are given in Table 11.1 (Evans and
dependence, where Q is an activation energy, R is Kohlstedt, 1995). Values of effective viscosity
the gas constant, and T is the absolute tempera- (11.14) at a temperature of 400 C and a maximum
ture. Primes are introduced since the symbols Q , shear stress of 10 MPa were computed from these
R, and T will be used for other quantities. In our quantities and are also given.
development, we retain the convention of taking In later analysis, the inverse of (11.13) is used.
compressive stress as negative. To derive this, rst we form:
I2 12 DijDij B2J(n1)
2
1
2 sijsij B J2
2 n
(11.17)
J2 (I2B2)1n
(ref)
sxy /sxy
Using (11.17) in (11.13) we have:
sij B1nI(n1)
2
2nD
ij (11.18)
n =infinite,
It is useful to make a comparison between the rigid-plastic solid
relations for the Newtonian viscous uid, the 1
power-law uid, and a third material, the
rigid, perfectly plastic solid. For simplicity, the 10
results for plane ow are given, for which 3
sxx syy 12 (xx yy), sxy xy. All materials are
incompressible so the rates of deformation are n =1
related as:
Dxx Dyy 0 (11.19) 0

0 1 (ref)
Dxy /Dxy
The constitutive relations for the viscous uid are:
Fig 11.4 Normalized one-dimensional relations between
Dxx (xx yy) 4
rate of deformation and deviatoric stress for power-law fluids
(11.20)
Dxy xy 4 with different stress exponents n.
The constitutive relations for the power-law uid

are: If the deviatoric stress component 12 (xx yy) is
zero, we may write the normalized relation for
Dxx BJ2(n1) 2 12 (xx yy), Dxy BJ (n1)2
2 xy the power-law material:

1n
(11.21) xy Dxy
where J2 14 (xx yy)2 2xy (ref ) (11.23)
(ref )
xy Dxy

The constitutive relations for the rigid, perfectly

plastic solid at yield are: Here the value Dxy(ref ) occurs at (ref ). The relation is
xy
shown in Fig. 11.4 for the viscous uid, n 1, and
Dxx 12 (xx yy), Dxy xy non-linear power-law uids with n 3, 10, and .
The last may be readily identied with the rigid-
where J2 14 (xx yy)2 2xy K 2 (11.22) (ref ) for it is the yield stress K.
plastic solid where xy
K is the maximum shear stress attained at yield- The rate of deformation is zero below the yield
ing of the material. Here (x, y) is not a ma- stress; at the yield stress, the rate of deformation
terial constant but is a function of position. The is indeterminate from the relation (11.23) alone;
rst pair of conditions in (11.22) is a statement and the stress cannot exceed the yield value.
of isotropy alone: the principal axes of the rate of
deformation tensor coincide with the principal 11.2.3 Linearization of the constitutive
axes of the stress tensor or the deviatoric stress relations and solution for low-
tensor. To complement the addition of a new slope necking and folding of a
unknown, an additional equation in the stress power-law layer
components, the yield condition is specied. For The fact that necking of an embedded layer or
stresses below this condition, the material is rigid inter-layer sequence produces pinch-and-swell
and, further, the stress state cannot lie outside the structures with regularity in neck-to-neck span
yield condition, which is a circle in (12 (xx yy), to mean layer thickness implies selective ampli-
xy)-space. All three materials satisfy the govern- cation of an initial random waviness in layer
ing equations of stress equilibrium and the kine- surfaces. This behavior depends upon the linear
matic relations. independence of wavelength components in the
random waviness. Selective amplication may plane and only the normal, principal components
then only occur when the constitutive rela- of sij are non-zero, yielding:
tions of a non-linear material can be linearized
2 D~ xx s~xx (n 1)2nJ2 sxx (sxx s~xx syy s~yy szz s~zz)
about a basic state of ow, since only then do the
wavelength components in the perturbing ow 2 D~ yy s~yy (n 1)2nJ2 syy (sxx s~xx syy s~yy szz s~zz)
not interact. Linear independence also requires
2 D~ xy s~xy (11.27)
that linearization of the boundary conditions,
when only terms proportional to A are re- Here D~ xx D~ yy 0 . Because of the proportionality
tained, is accurate, as in our analysis of viscous of components of the rate of deformation and
folding. deviatoric stress in the isotropic relations for the
Here, the basic state is taken as uniform layer- basic state in (11.26), the basic-state deviatoric
parallel extension or shortening, which simply stress components and J2 may be replaced by
changes the sign of Dxx. To emphasize this, we basic-state rate of deformation components and
write: I2 12 Dkl Dkl such that:
Dxx Sgn(Dxx) |Dxx | (11.24) 2 2 2 2
I2 12 (D xx D yy D zz) D xx(1 2 ) (11.28)
Since we have not studied folding in layer-parallel
shortening for a non-linear uid layer, we also The quantity Dzz Dxx is the ratio of the basic-
consider that here. We shall see that other types state rate of deformation component parallel to
of instability in the shortening or extension of a the axis of the shape perturbation to that acting
layer occur; for example shortening of a soft layer in the plane of the layer and normal to the axis.
between stiff media produces mullions (Smith, The plane ow case is then 0.
1975). Because restrictions are placed on the com-
To add interest to the analysis, consider a basic- ponents of the rate of deformation, it is more con-
state ow in which the principal rates of defor- venient to use the inverse relations (11.18). The
mation, Dxx, Dyy, Dzz, are all non-zero, with axes x linearized form is obtained in indicial notation:
and z in the plane of the layer and y normal to it.
The perturbing ow is restricted to a plane ow
associated with a cylindrical component in the
(n1) 2n
sij s~ij B1n I 2
1
(n 1) ~

D D (D D
2nI2 kl kl ij
~
ij)
(11.29)
shape perturbation whose axis lies parallel to the
principal axis of the basic state, . From (11.29) we extract the basic-state and pertur-
To linearize the constitutive relations, it is con- bation relations:
venient initially to carry out the computation
using indicial notation: sij B1n I 2(n1)2n Dij

(n 1) (11.30)

(n1)2
1 s~ij B1nI(n1) ~
2n D
ij
~
D D D
Dij Dij B (sklskl 2skl s~kl)
~
(sij s~ij) 2
2nI2 ij kl kl
2
In the present case, we introduce two restrictions:

BJ2(n1)2 1
(n 1)skl s~kl
2J2
(sij s~ij) (i) the only non-zero components are the principal
components Dxx, Dyy, and Dzz; and (ii) the perturb-
(11.25)
ing ow is plane, with non-zero components
~ ~ ~
From the second line of (11.25) we nd: D xx Dyy, Dxy .
The relations (11.30) give for the
perturbing ow:
Dij BJ2(n1)2 sij

D~ ij BJ2(n1)2 s~ij
(n 1)
2J2
sijskl s~kl
(11.26)
s~xx 2 1
(n 1) ~
D (D Dyy) D
2nI2 xx xx xx

(n 1)
We now reduce the second set of six equations, s~yy 2 1 ~
D (D Dxx) D (11.31)
2nI2 yy yy yy
where indices i and j take values 1, 2, and 3, for the
restrictions of interest: the perturbing ow is s~xy 2 D
~
xy
d4V 1 1
d2V

(n1) 2n
Here 2 B1nI 2 and is the effective viscos- 2 1 2 2 4V 0 (11.38)
ity (11.14) in the basic state, which is uniform within dy4 n x ny dy
the layer. These may be further condensed to: The solution may be written:
s~xx 2
1 ~
D
nx xx V (a cos y b sin y)ey
(c cos y d sin y)ey (11.39)

1 ~
s~yy 2 D (11.32) Here:
ny yy
n, 1 n
1 1
s~xy 2 D
~
xy
(11.40)

Here the inverse of nx and ny are dened: 3(n 1) 2
1
n n 1
1 (n 1) 4(1 2)
1 D (D Dyy)
nx 2nI2 xx xx
(11.33) To obtain the expression for n in (11.40) we used
1 (n 1) the relations (11.28) and the following:
1 D (D Dyy)
ny 2nI2 yy xx
2 1 1
(11.41)
The stress and deviatoric stress components are n nx ny
related by:
From (11.36) and (11.40), the expressions for
~ xx
2
1 ~

D p~
nx xx
the velocity components are:
~ yy 2
1 ~
D p~ (11.34)
v~x a (n^ 1)b cos y
ny yy
b (n^ 1) a sin y ey
2 Dxy ~

~ xy
c (n^ 1)d cos y
Here p~ 13 (~ xx ~ yy ~ zz)
is the perturbing part
of the pressure, or the negative of the mean
d (n^ 1)c sin y ey sin x
normal perturbing stress. For plane ow, the con- v
~ [(a cos y b sin y)e y
y
dition of incompressibility is:
(c cos y d sin y)e y ] cos x (11.42)
~ ~
v~x v~y
Dxx Dyy 0 (11.35)
x x From the third equation in (11.32) and (11.42) the
This is automatically satised by the separable stress component ~ xy may be obtained, and from
expressions: the stress equilibrium equations, the components
~ xx and ~ yy are obtained by differentiation and
1 dV integration:
v~x sin x

dy (11.36)
~ 22
xy a (n 1)b cos y
v~x V cos x
Substituting (11.36) and (11.32), using (11.34), into b (n 1)a sin y ey
the stress equilibrium equations (10.15):
c (n 1)d cos y

~xx ~xy
x

y
0 d (n 1)c sin y e y sin x
(11.37)
~xy ~yy ~ 2[(a cos y b sin y) ey

0 yy
x y
(c cos y d sin y) ey ] cos x (11.43)
Eliminating between the two equilibrium equa-
p~
tions, we obtain an ordinary differential equation The even more complicated expressions for
in V V(y): the layer-parallel normal component
~
xx and the
pressure p~ are not used in solving for the velocity 22 a (n 1)b cos h (e h eh)
eld for necking or folding, and are not given

here. b (n 1)a sin h(eh eh)
The solution for low-slope or innitesimal- 2 (Dxx Dyy) A(1 R)
amplitude necking refers in this case to the ow
set up by a symmetric pinch-and-swell component 2121 c1 (n1 1)d1 (11.49)
(Fig. 11.2). The analysis of the folding of a single
layer in Chapter 10 provides most of the details. 2[a cos h(eh eh)
Here, the mirror planes of symmetry immediately b sin h(eh eh)] 211c1 (11.50)
indicate that the vertical component of the per-
turbing velocity is odd in y when the coordinate Here, both the layer and the medium are power-
origin is taken at the center of the layer and that law uids. Since the coefcients for the medium
it is even in x, so: may be eliminated between pairs of equations in
(11.47) through (11.50), this system reduces to a
v~y(x, y) v~y(x,y) pair of equations that may be solved for the
(11.44)
v~y(x, y) v~y (x, y) coefcients a and b.
The interface evolution equation is developed
The rst condition in (11.44) implies c a, d b. as in the viscous folding example, yielding for the
The expressions for the velocity components may rate of change in amplitude, A, rather than slope,
be simplied accordingly, giving: A:
v~ x a (n 1)b cos y(e y ey )
b (n 1)a sin y (ey ey) sin x

dA
dt
(1 )
(11.45)
v~y [a cos y (ey ey)

n(2 )(1 R)

(1 Q ) 2(nsin1)k [(1 Q )(e
(b sin y)(ey ey)]cos x
2 2 k e k) 2Q (ek ek)]
The stress components are:
Sgn(Dxx)|Dxx|A
xy
~ 22 a (n 1)b cos y (ey ey)
q(k; n, n1, R, )Sgn (Dxx) |Dxx| A (11.51)

b (n 1)a sin y (ey ey) sin x Here:

~
yy 2[a cos y(ey ey) 1 1 1
Q R (11.52)
b sin y(ey ey)] cos x (11.46)
Using the same approximations as in the analysis The principal argument k 2 (HL) of the func-
for the buckling of a viscous layer, we obtain four tion q species the wavelength to layer thickness
relations from the boundary conditions at the ratio of the cylindrical sinusoidal perturbation.
upper sinusoidal surface: The rst three dimensionless parameters describe
the rheological behavior of the layer and medium:

a (n 1)b cos h (eh eh) n is the stress exponent of the layer, n1 that of the
medium, and R is the ratio of the effective viscos-
b (n 1)a sin h (eh eh) ity of the medium to that of the layer. Finally,
describes the basic-state ow to which the pinch-
1 c1 (n 1)d1 (11.47) and-swell component is responding. Since Dxx is
greater than zero for axis-normal extension, 0
a cos h(eh eh) b sin h(eh eh) c1 implies an additional axis-parallel extension, and
(11.48) 0 implies axis-parallel shortening. Note that
the effective stress exponent of the layer, n, is a identical to (11.51) except for a single change from
function of both n and (11.40); the stress expo- a positive to a negative sign in the denominator of
nent n1 is hidden in the quantity Q. The function the second term in braces, just after the quantity
q may be termed the relative rate of amplication (1 Q2). This is simply a consequence of a symme-
factor; it is dimensionless. try in folding opposite to that expressed by (11.44),
Since Dyy Dxx, the layer thickness, wave- or, in folding, v~y(x, y) v~y(x, y). Regular mullion
length, and wavenumber satisfy the relations: structures (Fig. 10.10b) forming in shortening of a
soft layer between stiff half-spaces are treated by
dH dL d assigning values of R 1 in either of these two
Dyy H, Dxx L, Dxx (11.53) relations. Mullions may be approximately sym-
dt dt dt
metric about the layer mid-plane, in which case
What seem like only moderate extensions of the (11.51) applies, or less commonly, asymmetric, in
model for initiation of folding of a single layer of which case the relation with the sign change
viscous uid embedded in a viscous medium, here would apply. Given a set of natural pinch-and-
for the complementary case of necking, lead to a swell structures, folds, or mullions, leading ques-
somewhat daunting range in behavior. If the layer tions would be whether the present model could
and medium are non-linear power-law uids, two produce them and what ranges of parameters are
stress exponents are introduced, n and n1, and if necessary.
the basic-state ow has a component of stretching As in the viscous folding problem, it is useful
or shortening parallel to the perturbation axis, a to expand the relative rate of amplication factor,
feature not uncommon in natural deformation, q, to low-order terms in k. The result is:
the parameter enters. In contrast, the behavior in
the viscous folding model only varies with the vis- n(2 )(1 R)
q (1 ) (11.54)

cosity ratio R. Thus, from a single parameter space, k2 2Q Qk 2
2 1
we must now consider a four-dimensional space! 6 k 3 n
In interpreting a set of data from trains of
The free-plate result is obtained by setting Q R
natural folds or pinch-and-swell structures, all
0:
four parameters may play a signicant role. Thus,
the relation (11.51), although relatively compli-
cated, allows for a full study of the variation in
behavior with the four parameters, the viscosity
1
(q)free plate (1 ) n 1
2
1 121 k
2
ratio R, the stress exponent of the layer n, that of

the medium n1, and the deformation rate ratio .

(n 1) 1
1 2
12 1
2
1
k n 1 k2 1
12
Both eld data and experimental data such as
those in Table 11.1 may be brought to bear. To treat
(n 1) 12 n 1 (11.55)
necking in layer extension, we know that we must The last expression corresponds to the plane-sec-
consider n 1. In most examples presented here, tions-remain-plane approximation (11.10) but now
the parameters are reduced to two by setting n1 contains information on the effect of additional
1, the Newtonian viscous limit, and 0. An shortening ( 0) or extension ( 0) along the
exception is the result for mullions in a layer for axis of the perturbation. The second line indicates
which computed results are obtained later (Fig. a wavenumber, k, dependent deviation from the
11.12), which only form if n1 1, and for which the plane sections result. The plane sections approxi-
dependence on n1 is key. Because the parameter mation for q, the last line in (11.55), is contoured
Q ~ nn1 plays a signicant role in the relation in (Fig. 11.5) as a function of the intrinsic stress
(11.51), restriction to n1 1 bypasses interesting exponent, n, and the rate of deformation ratio, .
behavior (Smith, 1977). This shows a substantial decrease in instability, as
Perhaps remarkably, the result for amplica- measured by the relative rate of amplication q,
tion of a fold component is given by a relation away from a plane ow basic state, 0. This
4 1000 10
n = 10, n1 = 1
0
0.001
3
5
100 R = 0.01
log10(n)
q
2
40 0.1
20
q =10 0
1
4
2
1
0
0
1 0 +1 5
j 100 101 102
L/H
Fig 11.5 Contours of q for plane-sections-remain-plane
approximation for the free plate in (n, )-space. Fig 11.6 q-spectra for n 10 and R 0, 0.001, 0.01, and
0.1; heavy lines show the exact results, light lines the
approximation for k 1.
decrease is less for extension parallel to the pinch-
and-swell axis ( 0) than for shortening ( 0)
parallel to the axis. For example, if n 101, q 100 the amplication, and thus the regularity in an
for plane ow, but only q 8 for an axial rate of array of structures. The q-spectra for R 0.05 and
extension one-half that of the axis-normal exten- n 10, 100, and 10 000, and (Fig. 11.7) show mul-
sion, that is for 0.5. For axis-normal contrac- tiple maxima that become prominent for n 100.
tion, 0.5 and the relative rate of amplication These arise because of the increasing dominance
factor is q 3. Thus, the model suggests that pinch- in the variation of q with k (11.51) of the sinusoidal
and-swell structures are favored in nearly plane term, sin [k1 (1n)] sin k, n 1 over the
deformation. exponential term, exp ( k 1n ) 1, n 1. The
The dependence on L/H, where H 2h is the transition between two modes of necking or
thickness of the layer, is shown by plotting q folding, one at modest stress exponent n, and one
versus L/H (Fig.11.6) for necking in plane ow for n at large stress exponent n 10100, is also tied to
10, n1 1, and several values of R. The expansion this transition in dependence. The second mode is
for k 1 yields a good approximation (dashed one of resonance folding or necking in which:
lines) down to L/H 10. While the stress exponent The competent layer does not act mechanically
n 10 is large compared with values estimated in as a coherent unit, but, instead, the irregularities
laboratory experiments, the necking instability is on one interface produce motions that deform
relatively weak, so that a large basic-state stretch the other and vice versa (Smith, 1979). This
is required to get signicant amplication. For behavior is associated with the dominance of
example, an amplication of 10 at the dominant sinusoidal versus exponential variation in veloc-
wavelength at which q is a maximum, for qd 9 ity and stress components in the y-direction,
would require a stretch S exp(ln 10qd) 1.29. which result in the dependences of q on k just
We denote the variation of q with wavelength noted. It is only shown for k 1, or for L/H 6,
or wavenumber the q-spectrum. From it the pos- when the sinusoidal dependence in k begins to
ition of maximum is at the dominant wave- become apparent.
length/thickness ratio, Ld/H, and the relative The coherent layer mode is exemplied by
sharpness of the peak indicates how selective is the pure folding mode of a single viscous layer, as
20
L/H = 0.8
10000
15
10 100
n =10 n = 1, L/H = 4
5
3
0
q
5
L/H = 4
10
15
20
100 L/H 101
n = 1, L/H = 2
Fig 11.7 Multiple peaks in qq(L/H) as a function of n
10, 100, 10 000 and the n approximation for R0.05, n1
3, 0. Velocity fields for the two maxima at L/H4 and
0.8 and the minimum at L/H4/3 are shown in Fig. 11.9.
illustrated by the perturbing velocity elds (Fig.

11.8). These are velocity elds in unconned or free
layers. Even though the sinusoidal perturbation in
this case is only imposed at the upper surface, not
directly indicated in the gure, a nearly pure
n = 10 000, L/H = 2
folding mode, with approximately uniform verti-
cal velocity at the fold hinges, develops at L/H 4.
At L/H 2, the layer behaves approximately as a
half-space, with the velocity decreasing exponen-
tially away from the surface at which the shape
perturbation is present. For the highly non-linear
layer, the sinusoidal dependence results, as in n = 10 000, L/H = 4
Smiths description, at L/H 4 or k /2, in a
maximum vertical velocity at the lower surface Fig 11.8 Coherent layer and resonance behaviors for
and zero vertical velocity at the upper surface at layers with n1 and 10 000, R 0.05 for folding at L/H4
which the shape perturbation is present. For L/H and 2 driven by a sinusoidal perturbation at only the upper
surface of the layer.
2, or k , the perturbing ow is distributed with
uniform intensity throughout the layer, but no
modication of the shape perturbation occurs. In Before further considering resonance effects
the coherent mode in necking, because the rate of in necking, we rst obtain results for folding and
extension is nearly uniform on a section, the ver- necking in the limit n . Denote As as the ampli-
tical velocity varies linearly through the section, tude of the pinch-and-swell mode and Ab for that
in contrast to the uniform vertical velocity in of the folding mode. Taking account of the sign
folding, and is zero at the mid-point. For such change in (11.51) for folding, and considering only
cases the plane-sections-remain-plane approxima- plane ow ( 0), (11.49) reduces in this limit for
tion for k 1 gives good results. necking to:
dAs n1 (1 R) Sgn(Dxx) sin k |Dxx|As

dt R 1 [R (k sin k)2 n1 ]
(11.56)
n
1 (1 R) Sgn(Dxx) sin k |Dxx |As
R
For folding: L/H = 4/1

dAb n (1 R) Sgn(Dxx) sin k |Dxx|Ab
1
dt R 1 [R(k sin k)2 n1 ]
(11.57)
n1
(1 R) Sgn(Dxx) sin k |Dxx |Ab
R
We may use these results to compute the rates
of growth for the amplitudes of the upper and
lower interfaces, A and A, where:
A Ab As,A Ab As (11.58)
Using the approximations for R 0:

dA n1 L/H = 4/3
(1 R) Sgn(Dxx) sin k |Dxx |A
dt R
(11.59)
dA n1
(1 R) Sgn(Dxx) sin k |Dxx |A
dt R
To re-conrm the sense of these relations, in
folding Sgn ( Dxx ) 1 and both A and A are pos-
itive: the perturbation grows if sin k 0. In
necking in extension, Sgn ( Dxx ) 1 and A is
negative, A is positive: again, the perturbation
grows if sin k 0.
The discrepancy from the approximations in
(11.59) is due to the difference between the dis-
carded terms in the denominators of (11.56) and
(11.57), which are small if R n1 1. The sinu-
soidal dependence in (11.59) is reected in a ver-
tical dependence in the velocity eld (Fig. 11.9)
that corresponds to the propagation of distur-
bances at either layer surface, resulting in either
cancellation or reinforcement as a function of
wavenumber k. Here, as opposed to the cases in
Fig. 11.8, a full pinch-and-swell perturbation is L/H = 4/5
imposed, and both interfaces have sinusoidal
shape perturbations. Resonance corresponds to Fig 11.9 Perturbing velocity distribution for necking in a
layer with n10 000 at L/H4/1, 4/3, and 4/5 showing
the case of maximum reinforcement. Further,
resonance behavior.
the simple dependency on k through the func-
tion sin k immediately allows us to read off the
results. Maximum rate of amplication will or L/H 4/2, 4/4, . . . The results shown in Fig. 11.9
occur for sin k 1, or for k /2, 5/2, . . . , or L/H are equivalent to the peak positive and negative
4/1, 4/5, . . . , with a maximum rate of decay in values in Fig. 11.7 at L/H 4/1, 4/3, and 4/5. The
perturbation at k 3/2, 7/2, . . . , or L/H 4/3, case of neither growth nor decay, L/H 2, is
4/7, . . . , and zero rate of change at k , 2, . . . , shown in Fig. 11.8.
4 (a)
400
3 200
4.1
100
40 2
log10(n)
20 qd = 10
2
Ld/H = 4.5
5
1 6
10 8
20 15
(b)
0
0
3 2 1
log10(R)
Fig 11.10 Contours of qd and Ld/H in (n, R)-space for

n11. Fig 11.11 (a) Boudin exposed in a shear zone (near
Middletown, CT). (b) Boudin profile, with 2 : 1 vertical
exaggeration: actual (light line) and symmetric (heavy line),
Comparison with natural structures requires with rectangles having same area and same maximum
thickness, done for entire structure and two individual
that we consider the effect of the mechanical
boudins. Photograph by R. C. Fletcher.
constraint of the embedding medium. This is
summarized in Fig. 11.10 for the case of plane
ow ( 0) and a Newtonian viscous medium (n1 structures, the corresponding distances between
1) by contouring the dominant wavelength to necks and the layer thickness attained before
thickness ratio, Ld/H, at which the rate of extreme attenuation of the necks may be esti-
amplication is maximum and the value of q for mated by returning the area between necks in the
it, qd, in (n, R)-space. This result allows an initial observed structures to a rectangle whose vertical
interpretation of a train of pinch-and-swell struc- dimension is the current maximum thickness. An
tures from which a mean neck-to-neck span to impression of this operation may be obtained by
mean thickness could be estimated. Thus, associ- examining Fig. 11.3.
ating this with the theoretical value of Ld/H, we This operation has been carried out for a
may trace the appropriate contour of this quan- natural boudin cross section (Fig. 11.11). The
tity on the gure to obtain possible values of n boudin, or pair of boudins, is in a pegmatite layer
and R. If we infer from the strong development of embedded in gneiss. The bulk deformation of the
necking that the maximum relative rate of gneiss consisted of extension plus right-lateral
amplication qd was large, e.g. qd 20, we may (positive) shear, as indicated by right-dipping
further limit the possible values of n and R to normal faults that produce a few sharp offsets in
those corresponding to the upper-left portion of the upper and lower surfaces (Fig. 11.11a). The
the gure. exposure is within a shallowly dipping right-
In folding, a strong tendency of the layer to lateral shear zone near Middletown, Connecticut.
maintain uniform thickness after the locking in The boudin prole (Fig. 11.11b) was measured
of a regular sequence of folds allows us to use the from an approximate median line, producing the
fold arc length to thickness ratios as those estab- actual slightly asymmetric form. A symmetric
lished at the end of selective amplication and form was created by locally adjusting the proles
the inception of nite amplitude folding. To esti- to lie at half the local layer thickness from the
mate better the value of Ld/H for pinch-and-swell median plane. The adjusted prole differs by only
a modest amount from the actual prole. The pair Mullion, n = 3

4
of boudins is separated from any other segments 10 40
of the same layer by a distance at least comparable 30
25
to its length. While part of the deformation is tied 20
8
to the normal faulting, including the central neck 15
3
between the boudins, the overall form is more
qd =10
log10(n1)
suggestive of a continuous ductile necking.
Detailed study to support or refute this suggestion
6
has not been carried out by us. None-the-less, this 2 5
hypothesis provides a useful example for the
present discussion.
In Fig. 11.11b, we have constructed the rectan- Ld /H = 6 10
2 8
gles approximating the initial boudin form for 1
the entire structure, supposing the central neck
to be a later feature, and for the two individual
boudins or pinch-and-swell structures. Their
0.2 0.6 1.0 1.4 1.8
aspect ratios are, for the entire structure, 4.2, and log10(R)
for the individuals, 2.3 on the left and 2.7 on the
right. The value for the entire structure is consis- Fig 11.12 qd and Ld/H for mullion structures for n 3 and
tent with model values of Ld/H for n 1, but n1 2.
those for the individual structures are not. It
might be suggested that the latter reect the ten-
dency for a mechanically isolated segment to The strength of the instability is small unless the
divide in two sub-equal segments if it becomes host has a very large stress exponent, so that
unstable with respect to necking, in this case mullion structures such as those seen in Fig. 10.2b
chiey by faulting. The present model applies may be inferred to represent large amounts of
only to a continuous layer or a segment with very layer-parallel shortening. This might explain the
large aspect ratio, and does not apply directly to signicant difference between the span-to-thick-
a process involving discrete faults (but see the ness ratio of these structures and the values of
next section). Ld/H 6.
The present model may also be applied to the
initiation of the regular mullion structures 11.2.4 A model for large-scale crustal
shown in Fig. 10.2b. The lobe-and-cusp morphol- necking
ogy indicates that the layer has the lower effective Necking may manifest itself at crustal and litho-
viscosity, or R 1. These structures are produced spheric scales, on the Earth, on other planets and
in layer-parallel shortening, so that Sgn ( Dxx ) 1. on the moon of Jupiter, Ganymede (Fink and
Using these and plane ow ( 0), (11.51) then pro- Fletcher, 1981; Collins et al., 1998; Patel et al., 1999;
vides a relation describing the growth or decay of Dombard and McKinnon, 2001). The strikingly
the pinch-and-swell perturbations whose selective regular succession of basins and ranges that form
amplication give rise to the mullions. Appreci- the dominant structure in the Basin and Range
able instability requires that the host be non- Province of the western United States (Fig. 11.13)
linear. We thus assign a typical stress exponent have been interpreted as the result of necking of
(Table 11.1) to the layer, n 3, and determine the the strong brittle layer of the crust (Fletcher and
variation of qd and Ld/H in (n1, R)-space (Fig. 11.12). Hallet, 1983). The crust is broken up into segments
Note that n1 2. This gure is read in the same approximately 30 km in width (Fig. 11.14). These
manner as Fig. 11.10, with values of n and n1 deter- are superposed on a subtler necking at a scale
mining Ld/H and qd, the d again denoting the 100 km in which the strong upper layer of the
dominant or most rapidly amplifying component. mantle lithosphere plays a signicant role (Zuber
10
9
8
7
Frequency
6
5
4
3
2
1
0
10 15 20 25 30 35 40 45 50 55 60
Spacing between ranges (km)
Fig 11.14 Histogram of spans between range centers

(Fletcher and Hallet, 1983).
0 100
km
space. More complicated and realistic models
have been treated (Fletcher and Hallet, 1983;
Zuber et al., 1986). In contrast to necking of an
embedded layer at the scale of an exposure,
gravity plays an important role in crustal
necking, and there is an essential asymmetry
between boundary conditions at the top and base
Fig 11.13 Basin-and-range structure in the state of of the plastic layer. A further interesting effect
Nevada; spans between range centers were measured on the involves the role of erosion, sediment transport,
transects (Fletcher and Hallet, 1983). and deposition.
While the perturbing ow in a plastic layer at
yield may be treated as that in a power-law layer
et al., 1986). The latter expresses itself as an alter-
in which the stress exponent n tends to innity, it
nation in dip of the dominant basin-bounding
is instructive to use the plastic yield condition and
normal faults.
ow law directly. For the plane basic state and per-
To model this process, we postulate that in the
turbing ows considered here, these are given by
brittle crustal layer faulting was initially more dis-
(11.22). The layer is at yield in uniform shortening
tributed, with faults of all length scales giving a
so that for the basic-state and perturbing stresses,
roughly continuous deformation at scales of a few
the yield condition (11.22) gives, for mean and per-
kilometers or less. Thus, in the initial phase of
turbing stresses:
extension a regular array of pinch-and-swell struc-
tures with spans of 30 km forms. The develop- xx yy 2K Sgn ( Dxx )
ment of major discrete normal fault zones (11.60)

~ ~ 0
xx yy
bounding the basins is then superposed on this
template. Here K is the shear stress and it is independent of
A rigid-plastic layer undergoing extension mean stress. The Airy stress function for the per-
affords an approximate model for the brittle turbing stress then satises:
behavior of the strong upper crustal layer. In the
present model, we will simulate the lower ductile 2 2
0 (11.61)
crust as a homogeneous Newtonian viscous half- y 2 x2
The appropriate form of the solution for may be Combining (11.62) and (11.67):
written:
b 2 |Dxx |A S
cos k
Sgn(Dxx) sin k
1
2
K
|Dxx |
(b cos y d sin y)cos x (11.62)
k
(11.68)
The associated perturbing stress components are:

d 2 |Dxx |A S

sin k
k
Sgn(Dxx) cos k

~xx
~yy
| |
K
Dxx
(b cos y d sin y) cos x
(11.63)
where as before k 2 (H/L). The four boundary
conditions at the layer/half-space interface,
A cos x, are:
~xy
| |
K
Dxx
(b sin y d cos y) sinx
v~x(x,0) v~ (1)
x (x, 0), v y (x, 0) v y (x, 0)
~ ~ (1)
~ (x, 0)
~ (1) (x, 0)
The perturbing stress components are gener- yy yy
(11.69)
ated from the yield condition and, by use of the ~ (x, 0) 2K Sgn(D )( A) sin x
xy xx
Airy stress function, from the equations of stress
xy (x, 0) 41Dxx(A) sin x
~ (1)
equilibrium. The choice of the constant factors in
(11.62) is merely to provide convenient expres- Substituting from (11.66) and (11.62) together with
sions for both the stress and velocity components. expressions for a viscous half-space (Chapter 10)
From the ow law in (11.22): into (11.69) gives:
Dxx K Sgn ( Dxx ) (c b) a1, a a1 b1

~ D
~ ~ K Sgn ( D ) K
D xx yy xx
(11.64) d 2K Sgn (Dxx) (A) 21 a1 41Dxx (A)
|Dxx |
~
Dxy
~xy K
b 21 (a1 b1) (11.70)
Substitution into the equation of compatibility |Dxx |
(10.64) yields:
From (11.70) the remaining two constants for the
~
2 ~
2 |Dxx |Sgn(Dxx) plastic layer, a and c, may be obtained.
22
y 2 x2 K2 The evolution equations for the two interface
(b cos y d sin y)cos x (11.65)
amplitudes A and A are obtained as in the previ-
ous examples. Here, the results may be written
This equation has both a homogeneous solution, out explicitly. They are moderately complex in
for zero right-hand side, and the particular solu- form and a detailed exegesis of the terms has not
tion for the given right-hand side. When the result been worked out, but some informative features
is substituted into the relations (11.64) and these may be pointed out. The evolution equations are:
are integrated, we obtain:
v
~
x [(a d by) sin y (c b dy) cos y] sin x
dA
dt
S
D xx A 2 S (1 R cos k sin k)
Rk

v
~
y [(a by) cos y (c dy) sin y] cos x
Sgn(Dxx) sin2k |Dxx |A
(11.66)
Note that the perturbing stress components

(11.62) contain only two arbitrary constants b and
d. These constants are xed by the conditions on
dA
dt
2
D xx A S
R k
cos k
Sgn (Dxx)sin k |Dxx |A
the vanishing of tractions at the top of the layer, (11.71)
on the surface H A cos x. These relations The behavior depends upon three dimension-
give, to the present approximation: less groups. The strength ratio R 21 |Dxx | K is
~ (x, H) g Acos x
equivalent to the earlier effective viscosity ratio in
yy
(11.67) models for folding and necking. Note, however,
~ (x, H) 2K Sgn ( D ) ( A)sin x
the interesting dependence on the absolute rate
xy xx
of shortening, since the shear strength K of the fault scarps (Andrews and Hanks, 1985) supposes
brittle upper crust is independent of rate. At slow that the ux of material is proportional to slope.
rates of extension, R will be smaller, and the Conservation of mass, excluding differences
degree of instability will be larger. The dimen- in density between rock and sediments, then
sionless group S gH2K is the ratio of a litho- requires that the rate of change in elevation be
static stress to the stress difference at yield in the proportional to the negative of the divergence of
plastic layer. A more realistic model for the the volume ux. Any adjustment towards an iso-
strength of the crust would introduce a mean static state, often included in models for the
stress dependence on the strength. Effectively, inlling of sedimentary basins and concomitant
only a cohesion-like quantity is used here, erosion at their periphery, is already included in
although its value may be assigned to account for the solution for the velocity eld. Applying
a mean value of the strength of the brittle crustal Cullings model on a component-by-component
layer that is dependent on the thickness H. basis, we obtain:
Further discussion of this issue may be found in
the papers cited earlier. The third dimensionless
group is k 2/(L/H).

dA
dt surface
P2A Mk2 |Dxx |A
Note that the relations (11.71) have the form: M P ( H2 |Dxx | )

(11.74)
dA Here M P ( H 2 |D xx | ) enters as an additional

q11A q12 A
dt dimensionless group and P is a diffusion constant.
(11.72)
dA The quantity (11.74) is then added to q11.
q21A q22 A
dt The value of the dimensionless group M might
be estimated and the use of Cullings model for
In the limiting case S 0, (11.71) becomes:
topographic decay at scales of approximately 10
dA to 100 km might be assessed by detailed study,
Sgn(Dxx) |Dxx |
dt but this is beyond the scope of the present treat-

1 ment. More simply, if the rate of reduction by
(1 2 sin 2 k) A 2 1 sin k A
R erosion is 1, 10, 100 m Ma1, or 1 km Ma1 for a

relief of 1 km at a wavelength of 30 km, a layer

dA
dt
2

Sgn(Dxx) |Dxx | sin k A A
R (11.73) thickness H 10 km, and a rate of deformation
|Dxx | 1015 s1, then M 0.01, 0.1, 1, or 10. The
Excluding the terms for kinematic amplication, layer thickness, H, is irrelevant to the surface
with factors 1 in brackets, and a term of order process but is conveniently used here to replace
unity in q11, the dominant behavior for R 1 by k. Since the proposed relief may be large, its
exhibits resonance, in the sense that the reduction by a factor of ten results in an increase
amplication at each surface is driven by the in M by the same factor. Accordingly, M 1100
amplitude at the opposite surface. This suggests might be a reasonable range for this dimension-
that the damping of topography by erosion, sedi- less number.
ment transport, and deposition may have a large Evolution of the interfaces from arbitrary
effect on the necking instability. This damping initial values A(0) and A(0) tends to produce a
may be incorporated into the evolution equations form that grows with the positive eigenvalue of
by simply adding a term to dAdt in (11.71). That the system (11.72) or:
is, the current state of motion depends on the
q 12 (q11 q22)
(q 11 q22)
2q q
current topography, not on its rate of change, and 1
4 12 21
so the contribution to the latter by surface (11.75)
A q q22 q
processes is simply additive. 12
A q21 q q11
A variety of models for topographic damping
might be used. One rst proposed by Culling Here the ratio of amplitudes associated with
(1960) and used by others to model the decay of this eigenvalue is given by the second line. The
dominant wavelength is that maximizing the S =1

eigenvalue q. 20
As with the necking of a single layer at n 1,
15
resonance leads to the excitation of multiple qd =10
peaks in rate of amplication at a sequence of 10
1
values of L/H 1. Here, we shall consider only the 20
8
log10(R)
maximum value of Ld/H. A question of principal 40 6
interest is how the existence of the necking insta-
bility and its length scale constrains the dimen- Ld/H = 4.6 5
100
sionless groups S, R, and M. Alternatively, we may
2
ask which a priori estimates of these quantities, as 200
of M above, would be consistent with necking
400
at the basin-and-range scale 30 km. Overlap
between these two sets of values would support
the model. 1000
An estimate for mean strength of the brittle 3
2 1 0 1 2
crust in extension, using a friction angle of 30 and log10(M)
no cohesion is K gH/3. Including cohesion, K
will be larger than this by a multiple greater than Fig 11.15 Contours of qd and Ld/H for the crustal necking
one and certainly less than two for H 10 km. model in (R, M)-space for S1.
Thus, we expect 3/2 S 3/4, and for 1 1019 to
1021 Pa s1 and Dxx 1015 s1, R 0.0002 to 0.02.
The latter values are remarkably small relative to behavior (Fletcher and Hallet, 1983), but compara-
effective viscosity ratios obtained from interpretable to the present one, provide a better t.
tion of natural fold data; if Dxx 1014 s1, R
0.0020.2. With these estimates in mind, we may
illustrate application of the model to the basin- 11.3 Coupling of viscous flow and
and-range structures by computing Ld/H, qd, and macroscopic diffusional
(A/A)d for S 1 over a wide range in R and M. transport
Contours of these quantities, determined numeri-
cally, are shown in Fig. 11.15. Under a wide range of conditions, pressure solu-
The regularity of basin-and-range structure tion is a signicant mechanism of deformation in
(Fig. 11.12) suggests a strong necking instability, so rocks that contain a large volume fraction of mod-
that as previously discussed for folding (Chapter erately soluble minerals such as calcite and quartz.
10) we require qd 20. This quantity depends Pressure solution taking place at the grain scale
chiey on R (Fig. 11.15) and the constraint requires (Chapter 11, frontispiece) can result in pervasive
R 0.05. We may use the mean value of the mea- ductile deformation. For example, in this sand-
sured spans as an estimate of Ld, since the amount stone, dissolution occurs on grain surfaces and on
of extension associated with the later-stage basin- surfaces of the dark anastamosing solution seams
and-range structure is modest. Since the smallest approximately normal to the direction of maxi-
values of Ld/H are 4.6 to 4.8, this implies H mum compression, horizontal in the gure.
6.5 km. Although this is smaller than the typical Dissolved material diffuses to grain surfaces sub-
depths to the brittleductile transition in conti- jected to the intermediate or least compression
nental crust of 10 to 15 km, it may be reasonable and precipitates there, often in the form of bers,
in view of the higher thermal gradient in this as here. Precipitation occurred onto the ends of
region. We thus conclude that the present simple the bers as they incipiently pulled away from the
model is a plausible one for the initiation of basin- sub-horizontal surfaces of the grains. The rate of
and-range structure. Other models with a more deformation is determined by the rate of diffusive
realistic characterization of crustal rheological transport from sites of dissolution to those of
11.3 VISCOUS FLOW AND MACROSCOPIC DIFFUSIONAL TRANSPORT 441
re-precipitation, by the kinetics of dissolution and

precipitation, by rates of mechanical adjustment,
or by some combination of these. There is a depen-
dence on grain size, with ne-grained rocks
deforming more rapidly than coarse-grained rocks
at the same stress. If both pressure solution and
crystal plasticity contribute signicantly to defor-
mation, a coarser-grained rock may deform chiey
by the latter mechanism while a ner-grained rock
deforms chiey by the former. A rock deforming by
pressure solution will behave approximately as a
linear viscous uid.
The sandstone (frontispiece) comes from a
Fig 11.16 Multi-layer boudinage in gneiss showing
location in the internal part of the Cascades accre-
segregation of material into boudin necks; horizontal span
tionary wedge, southeast of the topographic high
6 m. Photograph by R. C. Fletcher.
point at Mt. Olympus (Washington, USA). There,
the cleavage is nearly vertical and normal to the
direction of plate convergence. The thin section is
than diffusion. In this section, we treat the trans-
normal to the cleavage. The principal stretches for
port as diffusion; descriptions for either diffusion
the strain produced by pressure solution are 0.7
or Darcy ow are formally nearly equivalent.
normal to cleavage, 1.17 parallel to cleavage in the
Volume loss or gain involves the net transport of
plane of the gure, and 0.98 in the direction
the dissolved component over a macroscopic dis-
normal to these (Feehan and Brandon, 1999). The
tance grain size. This is the subject of interest to
product of the stretches is 0.80, which implies
us here, although not at the scale required to
that 20% of the initial volume has been removed
produce the loss of silica in the rock volumes
in the deformation. Since the observed deforma-
sampled by the sandstone described above. Rather,
tion most likely took place after the rock was com-
we will consider dissolution with negative dilata-
pacted to negligible porosity, volume loss must
tion and precipitation with positive dilatation with
come from loss of dissolved material, chiey
diffusional transport mediating between these at
quartz, from an initial volume of rock.
the scale of a structure in outcrop. Negative dilata-
tion is shown schematically in Fig. 11.17, where the
11.3.1 A model coupling deformation
smaller volume on the right contains the same
and diffusion in a viscous fluid
number of inert or insoluble marker particles.
The specic behavior to be considered here relates
To treat macroscopic transport by diffusion in
to the segregation of material in sites such as
a deforming rock, we make the following assump-
boudin necks (Fig. 11.1b, Fig. 11.16). Figure 11.16
tions:
shows the neck region in a boudin in a rock con-
sisting of inter-layered dark amphibolite and light 1. Pressure solution is pervasive, so that dilata-
felsic layers, and represents a case of multi-layer tion is continuous at the scale of interest. This
boudinage rather than that of a single stiff layer assumption may be valid even if material is dis-
in a soft medium, the case studied in this chapter. solved along discrete solution seams and pre-
The neck is shown by the sharp in-folding of layers cipitated into discrete veins, provided the scale
on the right-hand side of the gure and by the of transport is much greater than the dimen-
inlling of the roughly lenticular boudin gap sions and spacing of the seams or veins dis-
volume. When this occurred, the light-colored persed within the rock volume.
inlling may have been a melt. This local segrega- 2. Dilatation is isotropic. This assumption is less
tion may be an intermediary in the wholesale seg- satisfactory if dissolution occurs on seams and
regation of melt from a partly melted rock, the precipitation in veins, since these are generally
transport in this case being by Darcy ow rather strongly aligned.
Jx and Jy are the components of the macroscopic

mass ux vector per unit depth and V0 is the
specic volume per unit mass of the soluble
mineral.
Mass ux is related to the gradient of the bulk
chemical potential of the dissolving or precipitat-
ing solid phase in the intergranular lm, , or:

Jx M ,Jy M
x y
Fig 11.17 Schematic diagram of homogeneous negative (11.78)
dilatation with retention of inert marker particles. Area loss f c0
where M
volume loss 20%. RT
Here is the diffusivity of the mineral component

in aqueous uid, f is an effective lm porosity, is
3. Only plane deformation is treated, and dilata- the lm tortuosity, c0 is the mean concentration of
tion is assumed isotropic in the plane of ow. the solid component in the aqueous lm, R is the
4. The material dissolving or precipitating makes gas constant, and T is the absolute temperature.
up much of the rock and is uniformly available If the kinetics of dissolution and precipitation are
throughout the volume of interest. rapid, relative to transport, the mean chemical
5. Distortion of elements takes place as though potential in an element of the rock may be written
the material were an isotropic incompressible (Kamb, 1959a):
viscous uid: distortion and dilatation are addi-
tive parts of the rate of deformation, but other-
0 12 (xx yy)V0 (11.79)
wise not directly coupled.
With these assumptions, expressions for the Combining (11.77), (11.78), and (11.79) we have:
rate of deformation are:
Dxx
1
( yy) D
D 2
2
x y
2

2 (xx yy)
(11.80)
4 xx
1
Dyy (xx yy) D
4
(11.76)
where

f c0V02
4RT
1 The dimensions of are [ ] M1 L3 T.
Dxy xy
2 Combining (11.78) through (11.80) and (11.76)
Here twice D is the rate of dilatation, referring to yields the desired constitutive relations for the
the rate of change in the area of an element in the plane ow, including the diffusional transport of
plane of ow. the soluble mineral component. An inhomo-
In the case of pressure solution, the rate of geneous distribution of pressure or mean stress in
dilatation, twice D, is associated with dissolution, a viscous medium deforming in plane ow has
transport, and precipitation of the soluble, volu- been used to assess dissolution, diffusion, and pre-
metrically dominant, mineral in the rock. A cipitation within it (Stephansson, 1974), where
further assumption, point 4, is that there is no the divergence of the gradient in mean stress is
porosity within the material to act as a sink, nor used as a measure of dilatation. However, as we
can porosity be created by dissolution. Thus, dis- have learned, the Airy stress function, , in such
solution or precipitation must be balanced by a material satises the biharmonic equation, so
transport that at the scale of interest is by diffu- that:

sion along intergranular uid lms. Then: 2 2 2 2 2 2
( yy) 0
x2 y2 xx

x2 y2 x2 y2

J Jy
2D x V0 (11.77)
x y (11.81)
Thus, in the present formulation, such a stress dis- The rate of dilatation is:
tribution would give no dilatation, even though
vx vy
the gradient in mean stress is not zero. The re- 2D (1 2)(mey ney) cos x
x y
lation (11.81) is obtained by substituting the stress
(11.86)
components into the constitutive relations for an
incompressible viscous uid, and these in turn into
the equation of compatibility. Repeating this 11.3.2 Necking of a power-law layer
here, but with the present constitutive relations, embedded in a viscous medium
we obtain: with macroscopic transport
To illustrate some aspects of the behavior of per-

2
2 2 2 2 vasive pressure solution (Fletcher, 1982), we
2
2 1 2 2
2 0 (11.82)
x y x y return to the study of boudinage, which serves as
a focus of interest and attention in this chapter.
To treat examples of deformation and diffu-
Consider the necking of a power-law uid layer in
sional transport in layered rock congurations,
a medium of the type formulated here. Pressure
we again seek a solution that is separable in x and
solution does not take place within the layer. The
y. This is:
relations derived from the boundary conditions
(x, y)
2
2
[a b(ly 1)]ey mey
are nearly the same as in (11.47) through (11.50),
but the right-hand sides are replaced according to:
[c d(ly 1)]ey ney cos x vsx(x, h) (c1 g1) sin x
~

1 v y(x, h) (c1 d1 g1) cos x
~
where 1 (11.83)
22
~yy (x, h) 21(c1 d1 g1) cos x (11.87)
The four terms containing the constants a, b, c,
~xy(x, h) 21(c1 g1) sin x
and d in this expression are identical to those used
in Chapter 10, and satisfy the biharmonic equa- 4 Dxx(A) (1 1 )
tion. The remaining two terms containing m and
Here we have also used the restriction to a plane
n yield a non-zero dilatation rate, 2D. The quantity
basic-state ow. Since there are now ve co-
has the dimensions of a length squared, and
efcients to be xed, the additional one in this
we write L* (2).
case being g1, the equivalent of n in the relations
The stress components derived from (11.83)
(11.83) to (11.86), another boundary condition is
are:
required. This is the vanishing of the normal com-
xx 2 {[a b(y 1)]ey [c d(y 1)]ey ponent of the diffusional ux at the interface.
2(mey ney)} cos x While there is a basic state of uniform exten-
sion, the rate of dilatation in the case of a layer in
yy 2 {[a b(y 1)]ey [c d(y 1)]ey which diffusion does not occur is zero. Since there
(mey ney)} cos x (11.84) is a jump in layer-parallel normal stress, and
hence mean stress, across the mediumlayer
xy 2 {[(a by)]ey (c dy)ey interface, the condition of zero diffusional ux
(mey ney)} sin x may not correspond to what is seen in a natural
deformation. In extension, the mean stress in the
The velocity components derived using (11.76) and layer is less than that in the medium, and the
(11.80) are: component of a soluble mineral in the medium
vx [(a by)ey (c dy)ey mey would tend to diffuse into the stiff layer and pre-
ney] sin x
cipitate. If such a situation were posited, the addi-
(11.85) tional boundary condition would be on the
vy {[a b(y 1)]ey [c d(y 1)]ey
continuity of the normal ux at the interface, but
(mey ney)} cos x for simplicity, we exclude such diffusion.
2
enters, but parameters and , which also refer to
12 10
the medium, are not subscripted. At any layer
stress exponent, n, qd increases with L*/H and the
qd = 1000
1 increase is signicant for large n at which the
400
5 200 necking instability is otherwise moderate to
10 100 strong. This effect saturates as L*/H becomes
log10(L*/H)
20
40 comparable to Ld/H, or at approximately L*/H 10.
0 That is, as transport becomes efcient, the driving
gradient itself is lowered, so Ld/H and qd reach lim-
10 3.5 iting values and then do not change as L*/H
1 4 increases further.
At large n, or the plastic layer limit, the insta-
10 8 6 5 Ld /H = 4.5 bility becomes very large. Since much boudinage
occurs by separation of the stiff layer into discrete
2 segments by faults or shear zones, we might antic-
0 1 2 3 4
log10(n) ipate that this behavior is indeed characteristic of
layers forming this type of boudin. The present
model would thus be expected to provide a satis-
Fig 11.18 qd and Ld/H contoured in (n, L*/H)-space for
factory model for any initial phase of continuous
necking for viscosity ratio R 0.05.
necking that might have preceded through-going
faulting, as in the basin-and-range model.
The effect of diffusional mass transport in the
In the absence of basic-state transport, the con-
medium on necking is further illustrated by plot-
dition of zero normal ux at the interface, to rst
ting (Fig. 11.19) deformed grids and velocity elds
order in slope A, is:
for cases of no diffusion (L*/H 0) and vigorous dif-

1 ~ (1) ~ (1) fusion (L*/H 3.2). Only the grid deformation due
~(1)
J y (x, 0) ( xx ) 0 (11.88)
y 2 xx (x,0)
to the perturbing ow is shown; in (Fig. 11.19a),
the apparent thickening in the swell of the layer
To the present approximation, this yields the con-
or in the medium adjacent to the pinch would
dition:
be offset by the uniform basic-state extension.
1 Without diffusion (L*/H 0), the medium does not
d1 ( 2 1)g1 0 (11.89)
2 thin appreciably over the swell nor thicken over
the pinch or neck, but with diffusion (L*/H 3.2),
The ve equations obtained by combining
these effects are marked. Notice that the elements
(11.87) and (11.89) are solved numerically for the
above the swell also shrink, while those over the
constants a, b, c1, d1, and g1. The added complexity
neck undergo a positive dilatation. A third effect,
relative to the necking problem considered above
associated with a partial reduction of the mechan-
suggests that closed-form results may be laborious
ical constraint from the medium in the case of dif-
to obtain and may not easily provide physical
fusion, is that the perturbing deformation is
insight. We do not attempt this here.
markedly reduced away from the layer. The last
The effect on necking of dilatation mediated
effect is emphasized in the perturbing velocity
by diffusional transport in the host is shown by
elds (Fig. 11.19b); the medium velocity being
contouring qd and Ld/H in (n, L*/H)-space at xed vis-
much greater in the case of no diffusion except in
cosity ratio R 0.05 (Fig. 11.18) using:
the close vicinity of the layer.
A scalar measure of the strength of diffusive

L* 2 1
(11.90) transport is afforded by the ratio of the net rate of
H H2
diffusional volume transport towards the neck
This is a suitable dimensionless group for the per unit depth across a surface at x L/4 to the
present problem. Here, the medium viscosity transport by the perturbing ow through the
(a) D =0 D =10
(b)
(c)
1 5 4
3
2.5
2
log10(L*/H)
0 1.5
1
Fig 11.19 (a) Grids with initially square elements
deformed using only the perturbing velocity without diffusion 0.5
(L*/H 0) and with vigorous diffusion (L*/H3.2). 0.2
(b) Velocity fields for the perturbing flow for the same cases 0.1
1 0.05
as in (a). (c) Ratio of the net diffusional volume flux per unit
depth across the vertical surface at xL/4 to the volume flux 0.01
per unit depth across this surface within the layer from the
perturbing flow only, contoured in (n, L*/H)-space.
0 1 2 3 4
log10(n)
upper half of the necking layer towards the swell As a more concrete example, the sandstone in
(Fig.11.19c). This number may be greater than the frontispiece is made up of larger grains,
unity because the diffusional ux must also chiey of quartz, embedded in a ne-grained
counter the drag of material in the medium matrix rich in phyllosilicates and quartz, two
toward the swell. This suggests that a contribu- components that might be expected to have had
tion of slip at the layermedium contact, which distinct mechanical behavior. The larger grains
would diminish such drag, may enhance the insta- might have behaved as approximately rigid el-
bility further; the effect is not studied here. ements in a weak matrix. Evidently, though, the
large grains may be dissolved. The large grains are
elongate parallel to cleavage. Discontinuous
11.4 Continuum properties of seams depleted in ne-grained quartz, rich in
phyllosilicates, and aligned along cleavage are a
composite materials prominent if volumetrically minor additional
component of the rock. The presence of these and
Excluding volcanic glasses, rocks are made up of the elongate grains argue against isotropy of the
grains or crystals, and contain cracks and pores constitutive behavior, either in rheological or
that in situ will be lled with liquid or gas. At a elastic behavior. In shear parallel to cleavage, the
larger scale, rock masses are heterogeneous, from seams would possibly contribute to the deforma-
the relatively simple case of a layered sedimentary tion to a degree far outweighing their volumetric
rock to the more complex internal structure of a fraction. How might we quantify these features in
strongly deformed metamorphic rock. Treatment producing a model for the bulk behavior?
of these materials as continuous media, the
justication for which we have previously set 11.4.1 Voight and Reuss estimates for
forth, still leaves open questions. How may we esti- the bulk viscosity of a composite
mate the bulk properties of rocks relevant to of two viscous fluids
certain behavior, such as elastic deformation or The simple procedures for estimating the bulk
ow, from the properties of the major component properties of composite materials used here do
materials and their geometric conguration? As not always conform to the methodology so far pre-
the conguration of its mechanical components sented, in which we advocate the formulation and
changes in a ductily deforming rock, can the evol- solution of boundary value problems. In analyz-
ution in macroscopic properties be determined? ing these problems, the judicious use of approxi-
These questions relate to the correlation of mation is useful and often necessary, but
detailed observations of rock composition and conditions of traction continuity and continuity
structure with their constitutive behavior when of displacement or velocity, if appropriate, are
the deformation of interest took place. For honored. In the estimation procedures intro-
example, it is often remarked that the rheological duced here, however, this use of approximation
behavior of a rock in ductile deformation is deter- seems to have been taken too far!
mined by its weakest component, such as quartz As an example, imagine a composite material
in granite. Can we nd support for this idea made up of two isotropic linear viscous uids. For
through mechanical analysis of such a composite deniteness, imagine a conguration like that of
material? Such questions may be answered by sandstone, in which the quartz grains are
direct laboratory experiment. But these are replaced by one viscous uid and the ne-grained
sufciently expensive in time and resources to matrix by the other. Suppose that grain contacts
warrant discovering methods of estimation are welded: no slip or separation occurs along
whose reliability may be tested against particular them. Also, set aside the processes of dissolution
experimental results. Further, we need a concep- and precipitation that are central to the actual
tual basis with which to systematize experimental deformation of this rock. If there were a higher
results and to think about natural deformation. fraction of quartz grains, they would form more
In this section, we examine the simplest methods. complex interconnected bodies of this com-
11.4 CONTINUUM PROPERTIES OF COMPOSITE MATERIALS 447
ponent rather than isolated masses surrounded viscosities 1 and 2 and volume fractions f1 and
by the matrix component. The homogeneous f2 1 f1, the component Dxx will have the values
mechanical elements of the composite will then in the two components:
consist of isolated bodies to rather complex inter- 1 1
digitated bodies consisting of amalgamations of D (1)
xx s , D (2)
xx s (11.91)
21 xx 22 xx
grains composed of one or the other component.
By our hypothesis, that each component may be The macroscopic value of this component is then:
approximated by an isotropic viscous uid, the Dxx f1D (1)
xx f2 D xx
(2)
constitutive relations for the individual bodies f f 1

are given. Boundary conditions on continuity of 1 sxx 2 sxx s (11.92)
2 1 2 2 2R xx
velocity or displacement and of traction at their
interfaces would then be specied. However, the Here, R is the Reuss estimate for the macroscopic
geometry of the composite body is so complex viscosity:

that determining the stress and velocity distribu- f1 f 2
1
tion within it arising from a prescribed bulk R (11.93)

1 2
homogeneous stress or rate of deformation
Note that nothing was said about the congura-
would be difcult to obtain, even by numerical
tion of the composite. We might suppose that the
means. Instead, we consider two simple means
conguration was consistent with the composite
of estimating the bulk behavior that avoid
being isotropic, but the indeniteness has a
difculty.
In the Reuss estimate, the internal stress is further meaning. Not only is the result (11.93) an
taken to be homogeneous and hence equal to the estimate for a composite material made up of the
macroscopic or applied stress. Here a point of view specied volume fractions of the two compo-
espoused by J. N. Goodier is taken. We do not say nents, it is the lowest estimate that one might
that the stress in the composite material is obtain in any way. It is the lower bound on the
assumed homogeneous because it is a priori known bulk viscosity of such a mixture, as Burton Paul
not to be homogeneous, or highly unlikely to be was apparently the rst to prove (Paul, 1960).
so. Rather, we postulate a behavior or condition An estimate of the bulk viscosity of the
for the model, in this case, that the stress is homo- mixture, by a method rst proposed by Voight,
geneous. We are free to postulate anything we like may be obtained by postulating, in this model,
about the model, as that is completely indepen- that the rate of deformation is homogeneous. In
dent of the natural example. Later, we can hope to this case, the velocity eld in the medium is con-
understand to what extent the model corresponds tinuous. The equations of stress equilibrium are
to nature. The equations of stress equilibrium will satised in the sense that the stress is homo-
be satised and the normal and shear stress com- geneous in any volume of one or the other com-
ponents acting on interfaces within the compos- ponents, but the shear and normal tractions will
ite will be continuous as required. However, not generally be equal across the interfaces
because the components of the rate of deforma- between elements of the composite. Likewise, we
tion in the separate constituents are homo- do not specify the conguration of the composite.
geneous, the conditions of velocity continuity at Selecting a component of the deviatoric stress
interfaces will generally not be satised, even in whose mean or macroscopic value is sxx we have:
approximation. Thus, the estimate does not rep- s (1)
xx 21Dxx,s xx 22Dxx (11.94)
(2)
resent an attempt to solve the boundary value
problem for the material of interest specied at But then:
the microscopic scale.
sxx 2V Dxx f1s (1)
xx f2 s xx f121Dxx f222 Dxx (11.95)
(2)
The macroscopic or bulk components of
the rate of deformation are then taken as the aver- Therefore:
ages over the volume distribution of the compo-
nents. If the two-component viscous uids have V f1 1 f2 2 (11.96)
10 n V ,s R (11.97)
9 Our present aim is to construct estimates for an
isotropic composite material, so this association of
Bulk viscosity/Lower viscosity
8
the Voight and Reuss estimates with the principal
Normalized
7 Voight Paul
viscosities of a layered, anisotropic material would
estimate seem no more than fortuitous. However, it does
6 give us a strong clue as to a signicant factor: the
VRH
Paul soft
5 cubical
effect of the two-component geometry of the com-
inclusions posite. By considering the layered conguration,
4 we observe that in layer-parallel shortening or
extension, both layer types support the bulk stress.
3
Paul stiff Thus, if one component has a very much larger vis-
cubical Reuss
2 inclusions cosity, its effect on the principal viscosity n will be
large, if not dominant. The component with
1 higher viscosity in this case may then make up the
0 0.2 0.4 0.6 0.8 1
f1 load-bearing framework, with intervening layers
of much smaller viscosity supporting little load.
Fig 11.20 Estimates of bulk viscosity. VRH refers to the On the other hand, in layer-parallel shear, the resis-
VoightReussHill estimate. tance, or lack of it, may be principally associated
with the low-viscosity component, with sheets of
the high-viscosity component acting as isolated
The Voight estimate is an upper bound on the vis-
inclusions. Indeed, both components support the
cosity of the composite, a result also proved by
same load, in terms of layer-parallel shear stress.
Paul (1960).
Thus, as might have been guessed beforehand, the
A simple way of showing that the two estimates
geometry of the composite may have a great range
are bounds and here we must be careful not to
even if it corresponds to isotropic bulk behavior,
claim we have obtained a proof ! is to consider the
and this will have a major effect on the bulk vis-
case when one component is rigid, or has innite
cosity. As suggested here, two principal types of
viscosity. The Voight estimate indicates that no
conguration exist, those in which the high-vis-
matter how small a volume of this material is con-
cosity component forms a connected load-bearing
tained in the composite, the composite itself will
framework, and those in which it is present as iso-
have innite viscosity. The Reuss estimate indi-
lated inclusions surrounded by a matrix of the low-
cates that no matter how small an amount of the
viscosity component (Handy, 1994). We turn to
material with nite viscosity is present, the vis-
another simple method of estimation that
cosity of the composite will be nite. With regard
accounts for this difference.
to the supposition that the rheological behavior of
a composite material is determined by the low-
viscosity component in it, the Reuss and Voight 11.4.2 Paul estimates for an isotropic
estimates say quite opposite things. The Voight viscous composite
and Reuss bounds for the bulk viscosity are shown The lack of dependence on the geometry of the
in Fig. 11.20 for materials consisting of two viscous conguration inherent in the Voight and Reuss
uids with 2 101 as a function of the volume estimates is not likely to be wholly appealing to
fraction of the uid with the lower viscosity, f1. the structural geologist. A simple method for
The Voight and Reuss bounds on the viscosity obtaining estimates that takes something of the
of an isotropic composite may also be recognized internal conguration into account has been
as the exact principal viscosities of a composite proposed (Paul, 1960). This makes use of the
made up of alternating layers of the two uids, as strength-of-materials plane-sections-remain-plane
used in Chapter 10 to model the gross mechanics approximation used in our initial analysis of
of chevron folds, with: necking. These estimates show the expected asym-
11.4 CONTINUUM PROPERTIES OF COMPOSITE MATERIALS 449
(a) (b)
f 1/3 f 123 (1)
xx (1 f 1 ) xx xx (11.100)
23 (2)
Combining (11.100) and (11.99) we have:
(1f )1/3 Dxx [ f 123(31) (1 f 123)(32)]1xx (11.101)

s xx
The mean rate of extension is the average of these
taken along the x-direction:
1
Dxx (1 f 13
1 ) D
xx f 1 D
13
xx xx (11.102)
3 (1)
P
1 Here (1)
P is a Paul estimate with the 1-component
as an inclusion in a matrix of the 2-component.
Fig 11.21 Central sections of composite cubes used in the Performing the indicated algebra:
Paul estimate for macroscopic viscosity for a volume fraction
f 0.25. (1)
P f 123 (1 f 123)(2 1)
(11.103)
1 f 131 [ f 1 (1 f 1 )(2 1)](1 f 1 )
23 23 13
metry of behaviors for high-viscosity inclusions in

Another estimate is obtained by supposing the
low-viscosity material versus low-viscosity inclu-
material with viscosity 1 surrounds an inclusion
sions in high-viscosity material, an asymmetry not
of viscosity 2, both with the same volume frac-
contained in the Voight and Reuss estimates.
tions as in the previous case (Fig. 11.21b). The
Consider again a composite composed of two
result is given by:
isotropic viscous components. Paul models a com-
posite material in terms of a simple representative
(2)
P g 23
1 (2 1) (1 g 1 )
23
volume element (RVE) of it in the form of a unit
1 g 13 [g 23( ) (1 g 23)](1 g 13)
cube. For deniteness, he also considered specic 1 1 2 1 1 1
simple, if somewhat idealized, congurations of where g1 1 f1 (11.104)

the components within this RVE. The component
of viscosity 1 might be in the form of a cube cen- The result is written to emphasize the symmetry
tered within the unit cube with the surrounding between it and (11.103). These two estimates are
material of viscosity 2 (Fig. 11.21a). The side of the also plotted in Fig. 11.20. They lie within the
cube will have dimension f 113. Let the macroscopic Voight and Reuss bounds, and, as further
stress component xx 0 be applied to a pair of expected, the estimated viscosity for the case that
faces of the cube, and no other tractions applied the stiffer material surrounds inclusions of the
to it. Suppose, as in the necking analysis, that softer, is larger than for the complementary
plane sections remain plane, so that sheets of the conguration.
cube containing only the exterior uid of viscos-
ity 2 undergo a rate of extension: 11.4.3. Discussion
Two more estimates may be constructed from the
Dxx
1 2

2 2 3 xx
1

32 xx
(11.98) two pairs at hand: the Voight and Reuss bounds,
and the two Paul estimates. The average of the
Here sxx xx 13xx 23xx is the deviatoric stress Voight and Reuss bounds is called the
component acting parallel to x. Sheets of uid VoightReussHill estimate after its originator,
that contain an interior portion of uid with vis- Rodney Hill (Hill, 1965):

cosity 1 undergo a rate of extension: VRH 1 V R
(11.105)
1 (1) 1 (2) 1 2 1 1
Dxx
(11.99)
31 xx 32 xx
This estimate may be chosen if we have no infor-
To conform to the macroscopic applied stress, mation on the geometry of the composite, other
the average over the unit area of the section must than that it is consistent with bulk isotropic
be: behavior, or that the sample of interest is selected
randomly from a population with variable geom- anisotropic behavior, given known or postulated
etry but given volume fractions of the two com- behavior of the components. The estimates dis-
ponents. The bounds provide information on the cussed here may also be applied to non-linear
possible range of variation. materials, although computation of the Paul esti-
Another estimate may be obtained by placing mates is then rather complicated. In the special
Paul elements in series, so that the softer material case that the two materials are power-law uids
is neither always an inclusion in the stiffer ma- and the stress exponents are equal, the computa-
terial or vice versa. The latter is more likely if f1 tion becomes quite simple.
is large, so we weight its contribution by f1 and
that of the former by 1f1, obtaining:
P (2) (1) 11.5 Anisotropic fluids and
f1 P (1 f1) P (11.106)
1 1 1 internal instability
All six estimates are plotted in Fig. 11.20 as a
function of the low-viscosity component volume In the treatment of large-scale structures, such as
fraction, f1, for the case 2/1 10. Note that the accretionary wedges, it has been usual to approx-
Voight and Reuss estimates manage to stay imate the rheological behavior, however varied
outside of the tangle of the other four estimates, within the structure, as isotropic. At the scale of
which supports their nature as bounds. The an exposure, when layered and foliated rocks
largest difference in estimates occurs for higher deform in a ductile manner, the resulting struc-
volume fractions of the high-viscosity component, tures generally imply strong anisotropy, but to
1f1 0.5. The difference between the Paul esti- introduce such behavior into models for deforma-
mates, for which explicit congurations of the tion at a much larger scale, it would be necessary
composite are associated, give us an impression of to specify something of the initial disposition of
the joint effect of conguration and the variables layering and to keep track of it during the defor-
f1 and 2/1. For example, a relative bulk viscosity mation. Further, with the development of small-
of 5 corresponds to f1 0.25 for the high-viscosity scale structures, such as folds, which break up
inclusion in a low-viscosity matrix conguration, the layering or foliation within volumes large rel-
but to f1 0.5 for the low-viscosity inclusion in a ative to the scale of folding, the bulk behavior at
high-viscosity matrix conguration, so that a this scale will exhibit a smaller degree of
range in volume fractions may be offset by the anisotropy. Thus, the use of isotropic constitutive
nature of the composite conguration. Again, at relations for deformation at the largest scales may
the same volume fraction, a substantial range in be justied as an approximation.
relative bulk viscosity may be achieved by chang- There is still a strong motivation to study the
ing the conguration. consequences of anisotropic behavior in this and
As a rock consisting of two or more major many other situations. For example, the forma-
mechanical components deforms, its internal tion of structures such as folds and internal boud-
geometry will change. If it began as an assemblage inage in rocks that are foliated, layered, or
of equant grains or mineral aggregates or a more otherwise anisotropic in their bulk behavior may
complex intermeshing conguration still consis- be modeled as an instability in an anisotropic
tent with bulk isotropic behavior, deformation medium. The results may be used to interpret
will tend to produce a material with anisotropic arrays of natural structures, many of which are
properties. Instead of treating idealized congura- quite complex in terms of spatial variation in
tions by the Paul method, the actual congura- length scales, structural morphology, and the
tion of the rock might be sampled and a intensity of structural development in a rock that
plane-sections-remain-plane analysis undertaken, might otherwise have been imagined to be ini-
both in extension parallel to geometrically esti- tially an approximately homogeneous mass of
mated principal directions and in shear. This layered sedimentary or metamorphic rock,
would provide a method of assessing the bulk gneiss, or schist.
11.5 ANISOTROPIC FLUIDS AND INTERNAL INSTABILITY 451
11.5.1 Deformation of an anisotropic we would initially explore the scales and pat-
viscous fluid terns of the deformation and inferred constitu-
Modeling of ductile deformation of a rock volume tive behavior in terms of these components.
is appropriately treated using the constitutive rela-
Consider an anisotropic linear viscous uid
tions for a continuum that captures the bulk
described by relations introduced in Chapter 10:
rheological behavior at the scale of interest.
Understanding that this approximation applies to 1 1
Dxx s ,Dxy s (11.107)
a composite material, we realize that suitable con- 2n xx 2s xy
stitutive relations may be needed to describe an- Here x and y are local axes in the material paral-
isotropic behavior, as for a layered or foliated lel to the principal axes of anisotropy. Consider
material, a composite mass with a strong shape continuous variations in three quantities: , the
fabric, or a polycrystalline material with lattice small deviation of the x-axis from the x-axis of
preferred orientation, or a material in which all of xed reference axes x and y, and the two principal
these elements contribute to rheological ani- viscosities n and s. Write the latter as:
sotropy. Further, all of these factors and others,
such as composition, will contribute to inhomo- n n(1 n),s s(1 s) (11.108)
geneity. In some cases, such inhomogeneity as in a
volume of metamorphic or sedimentary rock may Here n and s are spatially varying fractional devi-
be approximated by a set of a discrete layers or ations from the mean values. Referred to x and y,
bodies of other shape, e.g. inclusions, which are the constitutive relations are:
themselves homogeneous. This is useful when the
4nDxx [(1 m) (1 m) cos 4]sxx (1 m) sin 4 sxy
contrast in mean properties of the discrete bodies
is much greater than that exhibited by internal 4nDxy (1 m) sin 4 sxx [(1 m) (1 m) cos 4]sxy
heterogeneity. Types of structures arising in this (11.109)
case, such as folds and pinch-and-swell structures,
Here m n s.
have been treated earlier in this chapter and in

We now linearize these equations about a basic

Chapter 10. In the present case, we consider the
state with rate of deformation components Dxx, Dxy,
other alternative of a continuous variation in
deviatoric stress components sxx, sxy, orientation
properties at the scale of interest.
0, and principal viscosities n, s. For the per-
We consider a body in which the degree of het-
turbing quantities, we obtain:
erogeneity is small, and may therefore be treated
xx 2s xx (1 m ) 4 sxy2nsxx
~
as a perturbation on a homogeneous material. This 4 n D ~
(11.110)
restriction has four anticipated consequences:
xy 2ms xx (1 m ) 4 sxx 2ms sxx
~
4 n D ~
1. Analysis may be carried out by linearization

Here m n s . We may proceed further without
about a basic state and is therefore tractable.

assigning expressions for the variations of

2. Both in the analysis and in the physical inter-
, n, and s by expressing the deviatoric stress
pretation of the results, any heterogeneous
components in terms of the Airy stress function,
current distribution of properties, the associ-
(x, y) and then substituting (11.110) into the equa-
ated velocity eld, and the relations describing
tion of compatibility (10.64) for the rate of defor-
the evolving deformation and modication in
mation. After expansion, we obtain:
the distributions of properties may be broken
up into independent components, e.g. the 4 4 4
2(2m 1) 2 2 4
Fourier components whose linear superposi- y 4 y x x

tion gives the distribution or eld. 2 2 2
3. Physical intuition is most readily obtained by 4( m 1) 2 sxy 2 s
y 2 x xy xx
focusing on the behavior of the individual com-
ponents.
4. To interpret a naturally deformed rock mass,
2
2n 2n
y 2
2 sxx 2m
x 2s
xy
sxy
(11.111)
The perturbation in principal axis orientation A cylindrical surface to which the principal
and those in the principal viscosities occur in sep- axes are tangent and normal that may be associ-
arate terms, because any term containing a ated with (11.113) is:
product of a perturbing quantity is deleted in the
(x, y) y A cos (x y) (11.114)
linearization. They may thus be thought of as
giving separate effects even though for any Here A and x tan . We may think
volume element in the uid both are required to of the trace of a cylindrical surface (11.114) as a
specify the local rheological behavior. We might foliation surface; its mean height, y, above some
thus imagine one limiting situation in which the reference level, which denes, in part, a material
principal viscosities were perfectly homogeneous surface in the uid, is treated differently from the
but the principal axis orientation varied, and independent coordinate y. An example of a set of
another in which the opposite was true. This sep- foliation surfaces of the form (11.114) is shown in
aration is no longer possible when the dimen- Fig. 11.22 for the case 45.
sionless quantities describing them are both too The solution that we seek starts with the par-
large for the linearized equations to hold. ticular solution to (11.111), where only the rst
In contrast to the many examples of boundary term on the right-hand side is considered.
and initial value problems discussed in this text, Substituting (11.113) into that term, it is then clear
we now consider an initial value problem for an that sin(x y), and we obtain:
unbounded volume of material. The initial value
(1 2)(4m 1)[( 2 1)sxy2 sxx] sin (x y)
description now refers to the functions:
[1 (2m 1) 2 4]
(x, y), n n(x, y), s s(x, y) (11.112) (11.115)
If one wishes to make a full investigation of in-

Given the stresses:
homogeneity of this sort, a sensible rst step is to
consider two simpler cases, one in which the
material is isotropic and all that is considered is a
s
~
xx
1 2 2

2 y2 x2
(11.116)
perturbation in viscosity, and the other in which 2
s
~
xy
the material is anisotropic, but the perturbations yx
in the principal viscosities vanish. To illustrate the We may then substitute into (11.110) and integrate
analysis and results further, we restrict attention to obtain the velocity components:
to the latter case, described only by the perturba- 2
tion in . Since the variations n and s are to be 4(m 1) [2Dxy (1 )Dxx]
v~x A cos (x y)
ignored, it is useful to think of them as zero and [1 2(2m 1) 2 4]
to replace m with m. Having worked through the 2
4(m 1) [2Dxy (1 )Dxx]
analysis with several mathematical forms for the v~y A cos (x y)
[1 2(2m 1) 2 4]
perturbation, we nd that the appropriate el- (11.117)
ementary form for a component in this perturba-
tion is the periodic band-like form: As in previous examples, the evolution of the
interface may be determined from the relation:
(x, y) sin (x y) (11.113)
vy(x, ) vx(x, ) (11.118)
Here is the amplitude or maximum value of the t x
slope and tan , where is the angle made The basic-state ow is:
with the normal to the mean plane of foliation, or vx Dxxx 2Dxy y
the y-axis, with a counterclockwise angle nega- (11.119)
vy Dxx y
tive. Such a perturbation is the only one that
behaves as an independent component, or eigen- Carrying out the expansion of (11.118) to rst
mode, in a general plane deformation. The fol- order in slope, we obtain the evolution equations
lowing analysis will show that this is the case. for the component:
11.5 ANISOTROPIC FLUIDS AND INTERNAL INSTABILITY 453
Fig 11.22 Single component with 45. The Fig 11.23 Symmetric internal boudinage is the sum of two
maximum slope is chosen large enough so that the structure cylindrical perturbations with 45 and45.
is visually distinct.
anisotropic material whose rheological behavior

dA 4(m 1) [2Dxy (1 2)Dxx] is approximated by an anisotropic viscous uid.
Dxx A A
dt [1 2(2m 1) 2 4] As with the folding of a single layer, the behav-
d d dy ior is most simply described by the rate of growth
Dxx, Dxx 2Dxy, Dxx y
dt dt dt of amplitude or slope of a single component.
(11.120) Because two components of the basic-state rate of
deformation may vary arbitrarily, it is useful to
Because this set of equations is complete and self-
scale all relations with the maximum rate of
contained, it demonstrates that the component is
shear, using dimensionless variables:
in fact linearly independent of all others. A per-
turbation that is not a single linearly independent dxx Dxx I2 , dxy Dxy I2
component, in a general plane ow, is the (11.121)
where I 2 (Dxx
2 D2 )
example of internal boudinage shown in Fig. xy
11.23, which combines a perturbation of the sort Focusing on amplitude rather than slope, we then
shown in Fig. 11.22 with another of equal ampli- examine the dimensionless quantity:
tude and phase, but with or of opposite sign,
1 dA
45. In pure extension or shortening, the com- q(; m, dxx)
I2 A dt
ponents will behave in a symmetric manner, but
if there is a component of foliation-parallel shear, 4(m 1)[2 dxy (1 2)dxx]
d xx
they will not and the independent evolution of [1 2(2m 1) 2 4]
the two parts becomes obvious.
(11.122)
The relations (11.120) may be integrated
numerically to follow the evolution of the struc- Because d2xx d2xy 1,
the dependence on the basic
ture, as long as the slope remains small, here not state reduces to the single variable dxx.
on a single interface but throughout a volume In contrast to cases in which a layer of nite
of uid. We now use the above results to illus- thickness H undergoes folding or necking, there is
trate structures that might be produced in an no dependence on an absolute length scale, since
1
0.8
Dxx /(Dxx2 + D 2xy)1/2
0.6
0
0.4 0.05
0.2 0.1
0 0.2
0.2
0.4
0.4
0.6
0.6
0.8
0.8
1
80 60 40 20 0 20 40 60 80
Band orientation (o)
Fig 11.24 Contours of q in (, dxx)-space.
none exists. That is, in principle, structures may

occur at any scale within the material. What is
selected for in the bulk deformation is the most
rapidly amplied band orientation, or, alterna-
tively, that which has received the maximum
cumulative amplication. In reality, in a rock like Fig 11.25 Crenulation simulation m 4, vertical stretch is
schist, the grain scale provides a lower limit on 1.2; initial undeformed area was square.
the scale of coherent structures produced, but
often the schist will also exhibit a compositional component with band orientation or axial plane
lamination, so that structures may occur in orientation normal to the direction of shortening.
response to several scales of microscopic structure Only positive values of q/qmax are shown. q may be
that give anisotropy. Indeed, the largest perturba- negative, indicating decay in the amplitude of a per-
tions are often present at one or both of these turbation. Negative contours are disposed in a
smallest scales at which coherent deformations pattern that has a center of symmetry about the
can take place. Hence, structures will commonly point dxx 0.
occur at these scales, but, in principle, they may For combined states of deformation in which
also occur at larger scales as is often observed positive shearing is added to foliation-parallel
(Price and Cosgrove, 1990). shortening, the maximum shifts to a negative
The expected band-like structures may be repre- value of , representing a fold structure that has a
sented for a given material, specied by the prin- sense of axial plane orientation opposite to that of
cipal viscosity ratio m, as a function of the so-called drag folds. Further, the maximum rate of
basic-state ow. Here we suppose that the com- amplication decreases. The result predicts that
ponents Dxx and Dxy are in constant proportion, or folds may form in foliation-parallel shear alone, dxx
that dxx is constant during the deformation. We 0, but once the axial plane of a component
may then contour q in (, dxx)-space (Fig. 11.24). In rotates to normal to the foliation in the course of
this gure, for m 4, q has been normalized by the a nite deformation, its amplitude will then
maximum value attained, (qd)max 14. This occurs cease growing and begin to decay. Thus, any struc-
for foliation-normal shortening dxx 0 for a tures which form in foliation-parallel shear must
lock-in by a mechanism as yet not identied to orientation for a given bulk shortening might
exist before decay sets in. Potential instability afford some indication of the strength of
exists in combined foliation-parallel extension anisotropy as described by the parameter m the
and shear, but the strength of instability in the basic-state deformation, and the character of the
present case is weak. Continued amplication initial perturbation.
with shear takes place in a basic-state ow with dxx
0.6, and aggregate amplication will outweigh
later decay for any value dxx 0.
By integrating the relations (11.120) through a
nite deformation, we may determine the ampli- This chapter has introduced a few of the factors
tude and orientation history of a component. If an which might be treated in more realistic models
initial set of perturbations with random initial of rock deformation: (i) rheological non-linearity;
slope, A, and phase, and chosen at equal intervals (ii) dilatation at a macroscopic scale mediated by
in orientation to foliation are followed, the nal diffusional transport; (iii) bulk properties of a
form of foliation planes may be computed as a composite material; and (iv) instability in the
summation. An example is shown in Fig. 11.25 for deformation of an anisotropic material. The
foliation-normal shortening. The initial shape of models formulated and analyzed are relatively
the rock element was square. Fold axial planes are simple and to a substantial degree build on ma-
not perfectly normal to the direction of shorten- terial presented in earlier chapters. The rst two
ing and they are nite in extent, even as seen in topics were studied in application to the initia-
this nite segment of the medium. They are tion of pinch-and-swell structures in extension, as
nearly normal to the direction of shortening, as well as to folds and mullions, to suggest the pos-
expected from the rate of shortening maximum sibility of the study of a wide range of behavior in
there (Fig. 11.24). structures such as folds or boudinage. Overall,
At present, little quantitative interpretation though, the presentation is one of brief sketches
of structures in foliated and multi-layer rocks has of restricted subject matter and methods. While
been carried out, and this eld of study remains these sketches often supply means of carrying out
open. Figure 11.24 suggests that the continuity the modeling of naturally deformed rock, they
of axial planes and the distribution in their also provide a means of gaining physical insight.
Chapter 12
Model development and methodology
Black Mesa
Maiden Creek
Oblique aerial photograph of the eastern slope of Mt. and how they work. (2) Then one can nd exactly
Hillers in the Henry Mountains, Utah. The Black Mesa where each part is located in the system in relation to
laccolith and Maiden Creek sill form the black, cliff-forming all the other parts and how they interact with each
outcrops. other. These two approaches are unlikely, by them-
selves, to reveal exactly how the system works. (3) To do
this one must also study the behavior of the system
Three main approaches are needed to unscramble a and its components while interfering very delicately
complicated system. (1) One can take it apart and char- with its various parts, to see what effect such alter-
acterize all the isolated bits what they are made of ations have on behavior at all levels (Crick, 1988).
12.1 IDEALIZATION OF FIELD OBSERVATIONS 457
N
ature, or what we might call natural 7.8). The frontispiece for this chapter is an oblique
reality, can appear to our senses as a very aerial photograph of the eastern ank of Mt.
complicated system when we view geo- Hillers showing laccolithic intrusions at Black
logical structures in outcrop. Francis Crick (1988) Mesa and Trachyte Mesa. Gilbert was one of the
also faced a complicated system when viewing the masters of scientic methodology in the practice
constituents of living cells. He suggests in his of geology, and we have much to learn from study-
book What Mad Pursuit that one should rst char- ing the way he worked. We give an example of the
acterize all the parts of the system and then procedure of idealization, recognizing that this
understand their geometric relationships. This is will not provide one with all the knowledge
what we attempt to do as structural geologists required to implement the procedure. Indeed, to
when mapping structures in the eld. In Chapters perfect this procedure requires considerable ex-
2 and 3 some of the useful tools for mapping were perience mapping structures in the eld, and in-
described and the principles of differential geom- timate familiarity with continuum mechanics.
etry were introduced to provide the fundamental Nevertheless, we introduce the procedure now in
basis for characterization of structures. Unlike the a specic, and perhaps narrowly focused, manner
cell in a test tube, many parts of typical geological using the example of Gilberts discovery of laccol-
structures are inaccessible because of limited ites (now called laccoliths) and his research into
exposure, and usually there are limited data on the origins of the Henry Mountains (Fig. 1.16). It is
the temporal development of structures. Crick assumed that the reader is familiar with the intro-
then suggests one must study the system as a duction to Gilberts research provided in Chapter
whole to understand how it behaves when various 1 and with his conceptual model for laccoliths
parts are perturbed. This step is possible when the (Fig. 1.17).
system is a cell in a test tube, but generally it is
impossible when the system is a rock mass larger 12.1.1 G. K. Gilberts field observations
than a cubic meter. Consequently, we turn to in the Henry Mountains
laboratory and mathematical models of geologic Gilbert arrived on the western edge of the Henry
structures to carry out this step in Cricks pre- Mountains in mid August of 1875 by horseback.
scription for scientic inquiry. Models are con- Intrigued by the range of mountains he saw to the
structed with specic features and attributes of east, Gilbert wondered if they might be a good
the rock mass, and the model system is studied to place to consider the question of whether volcanic
understand how it works. mountains were just piles of lava ows or perhaps
This chapter focuses on the development of the result of doming by injection of magma at
mathematical models of geologic structures. In depth. On August 18, 1875, Gilbert observed the
particular we consider the idealization of observed sedimentary rocks circling the base of Mt. Ells-
structures and the selection of general boundary worth and dipping away from the summit (Fig.
conditions for model development. At the end of 12.1). The next day he speculated that Mt. Hillers
this chapter we present a methodology for the must be capped with trachyte (igneous rock) and
practice of structural geology that summarizes the again observed the sedimentary units dipping
underlying concepts that have been described and steeply around the base of the mountain in con-
utilized throughout this book. trast to their nearly horizontal aspect further
from the mountain.
On August 21st Gilbert approached the base of
Mt. Hillers and recorded in words and a sketch
12.1 Idealization of field (Fig. 12.2) the scene that he observed:
observations
Camp 38 is on the SE base of Hillers. We have found
several minute springs in skirting the mountain and
We begin this chapter by describing how G. K. this one barely sufces us. It cannot be depended on as
Gilbert (1877) idealized the laccolithic intrusions permanent . . . The rock which rises toward Hillers
in the Henry Mountains of southeastern Utah (Fig. from the south is the B cliff (Ferron Sandstone). It is
458 MODEL DEVELOPMENT AND METHODOLOGY
(a) Looking at it from a hill E of camp 38, I am impressed

with the idea that the dikes are radial, diminishing
outward. The dip of the sandstones is not greater at the
center than on the anks. It is just a tumor cracked in
the middle. A main crack (dike) runs S (ab), another
NE (ac), a third W (ad). The S and W dikes show no
ows and the adjacent sandstones are preserved by
their hardness. Nothing below G.M. (Navajo Sandstone)
shows but G.M. goes to the top (Hunt, 1988a).
By the time he wrote these lines the seed for

(b) Gilberts conceptual model had germinated and,
with the use of the word tumor, he was almost
ready to describe it.
The next step was taken when Gilbert pro-
duced the sketch in his notebook shown in Fig.
12.5 and compared his observations at Mt. Hillers
and Mt. Holmes:
The types of Hillers and H.V. (Holmes) are somewhat
different. The radial dikes of the latter are feebly repre-
sented by the thin radials of the former and the con-
centric dikes of Hillers do not appear in H.V. If Hiller be
one extension of the H.V. type, then only the Trias was
Fig 12.1 Photographs of the southwestern flank of Mt. lifted and the Carboniferous either lay below the seat
Hillers. (a) From the ground looking toward the summit of action or below a disturbing reservoir (Hunt, 1988a).
showing the upturned strata around the base of the
mountain. (b) Aerial view showing prominent encircling
This sketch was Gilberts rst attempt to draw the
strata. Photograph by D. D. Pollard. disturbing reservoir that he later termed a lac-
colite. There was no doubt in Gilberts mind that
these mountains were created by the injection of
lost in the debris without increasing the dip (7)
magma at depth to both lift and bend the over-
with which it approaches. But beyond are red and
lying strata. He gathered further evidence on this
white sands inferior rocks tilted almost to the verti-
cal and interspersed with dikes. Moreover these sand- eld trip to support the hypothesis that Hillers
stone hogbacks seem to trend in a curve around the be an extension of the H.V. type and that there
mountain as far as they extend (Hunt, 1988a). exists a progressive evolution of exural ampli-
tudes from Mt. Holmes to Mt. Ellsworth to Mt.
The juxtaposition of the at-lying sedimentary Hillers. It is clear that the concepts embodied in
rocks of the plateau and those somewhat older Fig. 12.5 were used as the basis for the reconstruc-
strata turned up nearly to vertical against the tion used as the frontispiece of Gilberts nal
igneous core of the mountain was a dramatic report (see Chapter 1, frontispiece). The nal step
sight. Clearly these mountains were not a simple that Gilbert took in the idealization procedure
pile of lava ows. The older sedimentary rocks was to replace the mushroom-shaped laccolith
apparently had been pushed up by the invading and domed sedimentary strata (Fig. 12.5) with a
magma. This scene laid the groundwork for cylindrical chamber of magma pushing upward
Gilberts idealization of the mountains that on a rigid piston of at-lying strata (Fig. 1.18).
would come the next day.
The view of Mt. Holmes from near Gilberts 12.1.2 Steps in the idealization
campsite of August 22nd on the southern ank of procedure
Mt. Hillers is reproduced in a modern photograph The structural geologist usually starts with a
(Fig. 12.3). He drew a sketch of this scene in his note- direct eld observation, although photographs
book (Fig. 12.4) and made the following comments: (Figs. 12.1, 12.3) or eld sketches and maps (Figs.
Fig 12.2 Sketch from the field notebooks of G. K. Gilbert

in 1875 showing a cross section through the southwestern earlier stage in the process of uplift and doming.
flank of Mt. Hillers (Hunt, 1988a). Compare this to Fig. 12.1. He noted that the dips of the layers were less at Mt.
Reprinted from Hunt (1988a) with permission of The Holmes, that erosion had cut less deeply into this
Geological Society of America. structure, and that the remaining strata carried
completely over the top of the mountain. In con-
12.2, 12.4) may be used to reproduce the scene for trast the top of Mt. Hillers had apparently been
others. Of course what one sees directly in the stripped of its overlying strata, so only igneous
eld and then records on a photograph, sketch, or rock is exposed there.
map should be recognizably similar, but many fea- Gilbert was engaged in working on the process
tures of the scene are lost, while others are cap- of mountain building, so he sought out scenes that
tured in these procedures. The act of recording exemplied the stages in that process. In doing so
the scene is a form of idealization. The photo- he transformed spatial position and geometry into
graph loses much of the three-dimensional view temporal sequence and motion: the changes in the
provided by ones eyes and may lose much of the geometry of the intrusions and deformed strata
color information as well as features hidden in from Mt. Holmes to Mt. Ellsworth to Mt. Hillers
shadows. Typically, sketches and maps record very recorded snapshots in time of the progressive evo-
little of the totality of a given scene. Yet, with a lution of laccoliths. This transformation of space
few lines and a bit of shading a skillful recorder for time is one of the most important tools of the
such as Gilbert can transform the visible scene structural geologist. Because we rarely are able to
(Fig. 12.3) into a sketch (Fig. 12.4) that includes observe the deformation or ow of rock during the
most of the geologically salient features. development of a structure, nding exposures
An important question is why a particular that exemplify the different stages in such a devel-
scene that one stumbles across in the eld should opment is a critical part of any eld project.
be more attractive than any other. Perhaps Gilbert Having chosen the scene to record, Gilbert
chose to record the southern edge of Mt. Hillers began to throw away many of the things he
(Fig. 12.2) because he believed it would enable him observed directly in the eld, as well as many of the
to address the issue of volcanic mountains raised things we can see in the photographs of Mt. Hillers
by his boss, J. W. Powell. In any case, it is clear that and Mt. Holmes (Figs. 12.1, 12.3). In his sketches
Gilbert had a question in mind when he took the (Figs. 12.2, 12.4) only a few crucial items remain:
time to record the upturned strata and to marvel the gently dipping strata of the plateau; the talus
at this particular scene. He then focused on Mt. on the slopes; the more steeply dipping strata and
Holmes, possibly because it seemed to record an interleaved sills anking the mountains; the
Fig 12.3 Photograph of the northern flank of Mt. Holmes

looking south toward the summit showing the inclined strata
around the base of the mountain. Photograph by
D. D. Pollard

not observe! The dikes and sills that cut and inter-
in 1875 showing a panorama of the northern flank of Mt.
leaved the upturned strata on the anks of the
Holmes (Hunt, 1988a). Compare this to Fig. 12.3. Reprinted
from Hunt (1988a) with permission of The Geological mountains are gone. Presumably, Gilbert decided
Society of America. that these intrusive forms were minor compared
to the laccolith. The strata are completed over the
top of the laccolith, a relationship he observed for
igneous core at Mt. Hillers and the vertical dikes at the G.M. (Navaho Sandstone) at Mt. Holmes, but
Mt. Holmes. No one would mistakenly claim that the strata are completely eroded from the top of
these sketches are photographs because the simple Mt. Hillers. Furthermore, the talus is gone from
lines and patterns are mere representations. On the slope of the mountain. A feeder dike is added
the other hand, anyone with a modicum of geo- at the bottom of the laccolith, presumably
logical experience would have no doubt that the because Gilbert believed that the magma had to
sketch in Fig. 12.4 was made from the same scene ow up from below in some kind of channel. He
recorded in the photograph in Fig. 12.3. Gilbert had not seen such a feeder and had no way of
was a superb craftsman when it came to eld knowing what shape or dimensions it might have.
sketches. They have sufcient lines to capture the Perhaps the most interesting feature that Gilbert
essence of the geological features, while avoiding added, without direct evidence, is the at bottom
those lines that would simply clutter the image. contact between the intruded igneous rock and
In the next step of his idealization (Fig. 12.5) the apparently undisturbed strata below.
Gilbert chose to throw away more things and, In the nal step, the broad exure of the strata
interestingly, he chose to add features that he did is replaced by the highly localized deformation of
pay particular attention to such choices when

engaged in the procedure of idealization.
Apparently the choices made by Gilbert in the
nal step of his idealization procedure (Fig. 1.18)
were dictated by the solutions for boundary and
initial value problems that were available to him,
and the level of complexity that he felt was neces-
sary to address his questions about laccoliths. This
demonstrates the importance of being familiar
in 1875 showing his conceptual model of a laccolith. with solutions to problems in mechanics and also
Reprinted from Hunt (1988a) with permission of The the necessity for pragmatic decisions. Gilbert
Geological Society of America. knew that engineers had developed solutions for
the bending of elastic plates and his conceptual-
ization in Fig. 12.5 includes bending as a promi-
a cylindrical fault, cutting entirely through the nent feature of the deformation, however, he
overburden (Fig. 1.18). This fault forms the periph- chose to ignore bending. Furthermore, he was
ery of the laccolith as the overlying piston of sedi- familiar with solutions for the ow of viscous
mentary rock is pushed upward. The bending of uids, but he chose to ignore the role that ow of
the strata over the laccolith is completely ignored the magma might play in the deformation.
in this idealization. Indeed, there is no deforma- Apparently Gilbert believed that determining the
tion of the strata except along the fault. Because of relationship among the forces acting on a rigid
the geometry of the fault and the piston, the lac- body of simple geometric shape (Fig. 1.18) was
colith has a vertical side and a at top, rather than sufcient to understand the origin of laccoliths.
the mushroom shape of the idealization shown in Working out the mechanics of the piston model
Fig. 12.5. led Gilbert to the hypothesis embodied in Eqn.
At each step in the procedure of idealization (1.2): the laccolith radius is proportional to the
Gilbert made choices about what was important depth of overburden. His eld data proved to be
and what was not, about what should be added consistent with this relationship (Table 1.1).
and what should be deleted. These choices have a Choices made during the idealization pro-
profound effect on the nature of the analysis that cedure may be driven in part by practical consid-
is to follow and determine most of the results that erations. For example, when is the report on this
will be forthcoming. By throwing away the sills problem due on your thesis advisors desk or your
interleaved with the upturned strata, that part of employers desk? How long would it take one to
the intrusive process could not be addressed. By learn the necessary principles and tools for an
transforming the exure of the strata into a fault, unfamiliar area of continuum mechanics in order
the resistance that bending might offer to the to apply these to the particular problem? How
ination of the laccolith was ignored. By replac- much time would it take for the available com-
ing the magma with a static pressure distribution, puter to solve the problem that has been posed for
the ow and heat transport of the magma is rele- a given set of boundary or initial conditions? How
gated to a lesser role. Choosing a fault to provide much would it cost to gather the necessary data to
the resistance to uplift elevated the friction and constrain the model? These and a host of other
strength of the sedimentary rocks to a more considerations can force some purely pragmatic
important role than the resistance to bending. decisions to be made that inuence the outcome
Furthermore, the role of bedding-plane faults of an investigation.
during bending was ignored. Exactly how and why As the idealization procedure continues, par-
Gilbert made these choices is not known, but they ameters and conditions are excluded from the
led him to conclusions about the origin of the denition of the problem, so the behavior of the
Henry Mountains that otherwise would not have model is likely to diverge from the behavior of
been obtained. Sufce it to say that one should the natural system. Therefore, condence in the
model results should diminish. If these parame- possible models for these structures. Here we give
ters and conditions were included, the model may examples, all based upon the structure that
approach the actual behavior of the natural Gilbert set out to investigate over a hundred years
process more closely. On the other hand, the addi- ago in the Henry Mountains, and show how dif-
tional parameters and conditions may be less well ferent questions, posed in the eld about the
constrained by actual data. A master of this kind same structure, lead to different choices for the
of analysis, Art Lachenbruch, describes this context of modeling, that is to different general
dilemma in the following way: boundary conditions.
In selecting a formal model there is a trade-off; as its Because tectonic processes are multifaceted,
complexity increases, we can usually conclude less often involving solid and uid deformation, heat
from it with more condence. In this paper we select a and mass transport, and chemical reactions, rarely
simple model and attempt to conclude a lot from it, is there a single choice of general boundary condi-
while recognizing that the literal application of these tions that enables one to address all the interesting
conclusions is questionable, but that the insight is questions that arise from eld observations.
likely to be useful (Lachenbruch, 1973). Furthermore, we are limited in our understanding
It is relatively easy, given the power of modern of how these various processes are coupled
computers and the sophisticated tools available together, and the solution methods for coupled
for modeling complex mechanical systems, to problems may be difcult to implement. Finally,
create a model of a geological process that is itself the solutions to coupled problems may be so
too complex to understand, or that is too poorly complex that they are difcult to understand and
constrained by data. In hindsight it is clear that therefore are not readily applied to the simple
Gilberts piston laccolith was simple enough to questions asked in the eld. For all of these reasons
give him insight, and practical in the sense that we choose to break up tectonic processes into com-
the necessary eld data could be gathered to test ponent parts following the methodology des-
the hypothesis derived from the model. cribed in the quote at the beginning of this chapter
In general terms the procedure we have (Crick, 1988). Each part is analyzed according to a
described in this section is not new. Indeed, it can particular choice of general boundary conditions.
be traced back to the reductionist methodology Thus, the selection of general boundary conditions
rst articulated by Descartes in 1619 (Davis and is inuenced by practical necessity and guided by
Hersh, 1986). It is one of the cornerstones of an underlying reductionist philosophy of science.
modern scientic research. Once we have made the choice to work within
a particular area of continuum mechanics such as
plate theory, elasticity theory, viscous uid
12.2 Selection of general boundary mechanics, or heat conduction, the most general
differential equations for that area usually are
conditions simplied by eliminating certain variables or
terms from these equations. This elimination may
General boundary conditions dene the context be based on eld observations that justify, for
of a problem, so the appropriate theoretical prin- example, the reduction of a three-dimensional
ciples, fundamental laws, governing equations, structure to a two-dimensional approximation for
and computational tools can be brought to bear that structure. Or, for example, the elimination
on nding a solution. The selection of general may be based on a choice to ignore the initial and
boundary conditions is not a simple matter, and nal phases of deformation during which a ow
no easily written prescription can address all the eld changes in time, so a steady-state approxima-
variations and complexities one is likely to tion can be used to describe the well-developed
encounter. This procedure requires an under- ow eld that existed between these phases. In
standing of both the eld observations that serve other cases dimensional analysis is used to quan-
to characterize the structures, and the physical tify the relative importance of terms in the gover-
principles and scaling relationships that underlie ning equations and eliminate those that are
12.2 SELECTION OF GENERAL BOUNDARY CONDITIONS 463
negligible. In this section we describe a variety of with some power of its depth, but I am unable to say
simplication procedures, and the methods that what power. So far as I am aware, neither mathematical
are employed to make and to justify these choices. analysis nor experimentation has been directed to the
The procedure for idealization, described in the problem in question. According to Rankine the resis-
previous section, and the selection and simpli- tances of exure of similar cross-sections (of elastic
beams) are as their breadths and as the squares of their
cation of general boundary conditions are not
depths (Applied Mechanics, page 316), and it is possi-
independent. The former is described as a eld pro-
ble that the same law applies to the resistances which
cedure and it should, for the most part, be under- continuous strata oppose to the uplifts of domes. But it
taken during mapping or data collection, when the appears more probable that the greater complexity of
decisions can be evaluated by direct observations. the strains developed in the formation of domes causes
The latter is described as an ofce procedure, and it the depth to enter into the formula with a higher power
is likely to be undertaken with a continuum than second (Gilbert, 1877).
mechanics textbook in hand. In fact most struc-
tural geologists, including Gilbert, intertwine To address questions about the shape of the
these two procedures. Apparently Gilbert carried domed strata and their resistance to bending a
an engineering handbook into the eld and pre- number of models have been proposed since
sumably used it to help conceptualize and set up Gilberts report based on a theory for the
problems (Pyne, 1980). An idealization may be deection of thin plates (Johnson, 1970; Pollard
designed specically to take advantage of a known and Johnson, 1973; Koch et al., 1981; Kerr and
solution, for example in elasticity theory or uid Pollard, 1998).
mechanics. Later, having found a solution for a The context, or general boundary conditions,
more realistic set of parameters, certain cons- for the analysis of thin plates is the engineering
traints imposed in the former idealization may be discipline called strength of materials or plate theory.
relaxed. In this way an analysis may progress, for Reference textbooks are available on this subject
example, from one employing a linear and iso- from the engineering literature (Timoshenko,
tropic viscous uid (Chapter 10), to an anisotropic 1958; Timoshenko and Woinowsky-Krieger, 1959).
viscous uid, to a uid with more complex rheology In Section 4.3.1 the dimensionless group for plate
(Chapter 11). This give-and-take relationship bending was derived, (4.45), and showed that dis-
between the idealization of eld observations and placements scale with the fourth power of the
the selection of general boundary conditions for length and inversely with the elastic stiffness and
modeling is vital to the development of effective the third power of the thickness. Thus, bending
models and the analysis of structural data. under distributed loads is most sensitive to the
geometry of the plate.
12.2.1 Plate theory: what controls the To apply plate theory to the laccolith problem
shape of the domed strata over a the sedimentary strata overlying the laccolith are
laccolith? idealized as a stack of n thin plates of thickness, hi
It is clear that Gilbert understood the strata over (Fig. 12.6a). Particular plates may or may not cor-
laccoliths in the Henry Mountains did not behave respond to stratigraphic units, but each behaves
as rigid pistons (Fig. 1.18) because his conceptual as an independent mechanical unit, capable of
model that preceded the piston laccolith shows slipping relative to adjacent mechanical units
the strata bent into a domed shape (Fig. 12.5). He along bedding-plane faults. Unlike Gilberts rigid
noted that resistance to uplift for the piston piston, these plates deform according to linear
model is proportional to the overburden thick- elastic relationships among the stress and inni-
ness and speculated in his report of 1877 that tesimal strain components dened by Hookes
resistance to bending might increase with some Law (Chapter 8). For plates that are isotropic and
greater power of the overburden thickness. homogeneous with respect to elastic properties,
the two relevant material constants for each plate
I am led by the analogy of allied problems in mechanics are Youngs modulus, Ei, and Poissons ratio, i.
to assume that the resistance of the body of strata varies The plates are bent by upward directed forces due
(a) Mechanical Bedding- ution of tangential tractions on these cross sec-

unit i
plane tions of the plate. Thus, the equilibrium of the
fault
free body is determined by the vertical pressure
Ei, vi i =n acting across the plane of the plate, and the
hi bending moments and shear forces acting on the
lateral edges of the plate.
i =1 The vertical displacement of particles along
Magma the middle surface of the plate is referred to as the
2a deflection, w, which is used to characterize the
shape of the deformed plate. The governing equa-
(b) Thin elastic plate tion for the deection distribution, w(x, y), of a
as a free body
z single thin elastic plate subject to a pressure
M acting across the plane of the plate apparently
x was rst derived by Navier in 1820 (Timoshenko,
w
S 1953, p. 121):
P
4w 4w 4w P(x, y)
2 2 2 4 (12.1)
x4 x y y R
(c)
2a The plate is postulated to be in a state of static
2c
h equilibrium so there are no (or negligible) accel-
erations of the middle surface in the z-direction.
In (12.1) the term R Eh3/12(1 2) is the so-called
flexural rigidity of the plate, a measure of the resis-
Fig 12.6 Idealized model for plate bending over a laccolith.
tance to bending. Note that the exural rigidity is
(a) Cross section of multiple mechanical units with bedding-
plane faults. (b) Cross section of a single mechanical unit.
proportional to Youngs modulus and to the cube
(c) Oblique view of plate model for a mechanical unit. of the plate thickness. From the governing equa-
tion we see that the plate problem is two dimen-
sional, that is the deection and the pressure are
to the magma pressure distributed over a length functions of only two spatial coordinates. This
2a, and this upward motion is resisted by the fourth-order partial differential equation is solved
weight of the plates and their elastic stiffness. The for particular choices of boundary conditions to
plates slide over one another as they bend and the determine the distribution of deection. It should
host rock below the level of the laccolith and to be mentioned that this formulation is restricted
either side of the bending plates is treated as rigid. to thin elastic plates, those that have lengths
A particular plate is shown isolated as a free greater than about eight times their thickness,
body (Fig. 12.6b) with the coordinate system ori- 2a/h 8, and to deections that create slopes less
ented so the x-axis and y-axis lie in the plane of the than about 20. For lesser ratios of length to thick-
plate with the z-axis vertical, and the origin is at ness or greater slopes, simplifying postulates
the center of the middle surface of the plate in the about the bending of the plate introduce errors in
undeformed state. The normal component of trac- deection that exceed 10%.
tion on the bottom of the plate is equivalent to a For the sake of this example, consider a stack
distributed pressure, P(x, y), and there are no shear of n plates all with the same elliptical shape in the
components. The top of the plate is traction free (x, y)-plane (Fig. 12.6c). The long dimension of each
and the distal edges are loaded by the bending plate is 2c, and the short dimension is 2a, so the
moment, M, and shear force, S, necessary for static general elliptical plan can vary from circular, c
equilibrium (Timoshenko, 1958; Timoshenko and a, to anticlinal, c a. Gilbert conceived of the
Woinowsky-Krieger, 1959). The bending moment ideal laccolith as being circular in plan shape
accounts for the distribution of the normal trac- (Gilbert, 1877), whereas Hunt (1953) interpreted
tions and the shear force accounts for the distrib- the laccoliths around the anks of the ve Henry
Mountains as being tongue-shaped or roughly 0.045

anticlinal in plan with long axes extending radi- Anticlinal
ally from the intrusive centers. The elliptical plan 0.040 Circular
accounts for every shape from one of these end-
0.035
members to the other. The boundary conditions
Normalized deflection
at the distal edge of each plate are that both the
0.030
deection and the slope are zero.
A uniform distribution of pressure, P con- 0.025
stant, is dened as the difference between the
magma and lithostatic pressures, Pm Pw. For 0.020
plates that are able to slide freely over one another
(no shear tractions across the interfaces), the 0.015
exural rigidity is the sum of the individual
0.010
exural rigidities, and we refer to this as the effec-
tive flexural rigidity, Re (Pollard and Johnson, 1973):
0.005
n Eih3i
Re
i1 12(1 vi )
2
(12.2) 0
0 0.2 0.4 0.6 0.8 1
The plates are numbered from 1 to n starting at Normalized position, x/a
the base of the stack (Fig. 12.6); hi is the thickness Fig 12.7 Plot of normalized vertical displacement
of the ith plate; and Youngs modulus and (deflection) versus position for bending elastic plates with
Poissons ratio for this plate are Ei and i, respec- uniform distributed loads. Because of symmetry only half the
tively. distribution is shown.
The solution to (12.1) for the deection of any
one of the uniformly loaded elliptical plates is
(Love, 1944; Timoshenko and Woinowsky-Krieger, sure and to the fourth power of the half-length, a,
1959): so it is considerably more sensitive to changes in
the length of the plate, than to changes in pres-

2 2 2
x y sure.
1
(Pm Pw) a c The deection is illustrated in Fig. 12.7 where
w (12.3)

8Re 3 2 3 the dimensionless deection is plotted versus the
dimensionless distance from the center of the
a4 a2c2 c4
bent plate for both the circular and anticlinal
The two end-member cases simplify this two- plans. Only positive values of x/a are shown
dimensional problem to one spatial dimension because the solution is symmetric about x/a 0.
where the deection distributions are: All deection distributions for the general ellipti-
cal plan shape would fall between these two so
(Pm Pw) 4
w (a 2a2x2 x4)(anticlinal) the graph illustrates the entire spectrum of
24Re deections for this model. The form of the model
(12.4) deection is double hinged, that is a distal
concave upwards hinge surrounds the edge of the
(Pm Pw) 4
w (a 2a2x2 x4) (circular) (12.5) dome and gradually changes to a limb of nearly
64Re
constant dip where the curvature reverses sign to
The deection distribution along any cross form the concave downward hinge marking the
section in the (x, z)-plane is the same for the anti- center of the dome. This form is qualitatively
clinal plan shape. The deection distribution similar to the form envisioned by Gilbert in one of
along any radial line from the origin is the same his idealizations of a laccolith (Fig. 12.5) and to the
for the circular plan shape. Note that the deec- form illustrated in his restored section of Mt.
tion is directly proportional to the driving pres- Ellsworth (Chapter 1, frontispiece).
For a comparison of the model deection dis- opportunities for the development of bedding-
tributions to eld data we turn to maps and cross plane faults and shear zones within softer layers
sections for the southern Henry Mountains that would act to delaminate the overburden into
(Jackson and Pollard, 1988, 1990). On the geologic thinner mechanical units. Bedding-plane fault
and structural map of Mt. Holmes (Fig. 12.8) note zones are exposed within the sandstones and at
that the sedimentary strata are continuous over formational contacts (Fig. 12.9), indicating that
the top of the dome, and the beds dip more or less the overburden behaved in these places as a stack
in radial directions around the mountain. of mechanical units that could slip over one
Furthermore, the magnitude of the dips is small another. Measurements of bedding-plane faults at
far from the mountain, less than 10; increases to Mt. Holmes demonstrate that they are spaced
over 20 on the steep anks, and decreases to less from 150 to 200 m apart, so this may be a typical
than 10 near the top. The cross section along the thickness of the mechanical units (Jackson and
line AA (Fig. 12.9) is based on the bedding atti- Pollard, 1988). In this way the total overburden
tudes measured at exposures around the moun- thickness of about 4 km could be reduced to an
tain, and the geologic and topographic maps. This effective thickness, he, that was considerably less
cross section reveals a form of bending that is than the diameter of the domes.
similar to the plate model (Fig. 12.7). In addition The effective thickness of a stack of bending
to the doubly hinged structure the cross section plates is equal to the thickness of a single plate
reveals a gently dipping peripheral limb that that would resist bending just as much as the
extends 34 km beyond the lower hinge for all entire stack. For mechanical units with the same
three of the southern Henry Mountains (Jackson elastic properties, E and , that are able to slide
and Pollard, 1990). This does not correlate with freely over one another, the exural rigidity
the simple plate model, and may indicate the pres- reduces to the following simple form:
ence of underlying sills and smaller laccoliths n
E Eh3e
around the anks of the central laccoliths. Re
12(1 2) i1
h 3
i
12(1 2)
(12.6)
The correlations between the plate model
deections and the cross-sectional shapes for Mt. Note that the resistance to bending in this equa-
Ellsworth and Mt. Hillers are more difcult to tion scales with the cube of the effective thick-
assess because the upper parts of these domes are ness. Thus, if the effective thickness is less than
eroded. Furthermore, whereas maximum limb the total thickness, there can be a profound effect
dips of 20 at Mt. Holmes are just within the range on the resistance to bending. For example, if h
permitted for the application of plate theory, the 4 km, then h3 64 km. However, if this overburden
dips are between 50 and 55 at Mt. Ellsworth, and is delaminated into four equal mechanical units
between 75 and 85 at Mt. Hillers. These intrusive by bedding-plane faults, then hi 1 km each, n 4,
structures have developed beyond the stage where and h3e 4. Delamination into four mechanical
elastic plate theory should be applied. However, if units reduces the resistance to bending by a factor
these domes passed through an earlier stage of of 16. Because bedding-plane faults would offer
development when, according to Gilberts con- frictional resistance to sliding and, as pointed out
ceptual model, the sedimentary overburden was in the next section, these faults are unlikely to
domed much like that at Mt. Holmes, plate theory develop over the entire laccolith, this analysis
would be applicable for this earlier stage. over-estimates the reduction in exural rigidity.
The diameters of the southern Henry None-the-less the mechanical effect of bedding-
Mountains domes, 2a 10 to 14 km, are only two plane faults on the development of laccoliths is
to three times the total overburden thickness, h likely to be signicant.
4 km, yet applications of plate theory require a In summary, the plate theory provides insight
ratio of diameter to thickness greater than about concerning the resistance to bending of sedi-
eight. On the other hand the sedimentary section mentary strata over laccoliths, and the solution
is composed of a multitude of sandstone, silt- for uniform loading gives a deection shape that
stone, and shale beds, and this bedding provides is similar to the early stages of doming at
Fig 12.8 Map of Mt. Holmes in the southern Henry Mountains, UT. Black is diorite porphyry; white (N) is Navaho
sandstone. Reprinted from Jackson and Pollard (1988) with permission of The Geological Society of America.
Bedding-plane
faults
A
Peripheral limb Lower Central limb Upper hinge

hinge
Fig 12.9 Cross section of Mt. Holmes identified as AA sliding over one another (Jackson and Pollard,
on Fig. 12.8. Note bedding-plane faults. Reprinted from 1988). Plate theory indicates the importance of
Jackson and Pollard (1988) with permission of The this delamination mechanism in reducing the
Geological Society of America.
effective thickness of the overburden and thereby
reducing the resistance to bending so the laccol-
Mt. Holmes. To answer the question posed at the ith can grow in amplitude. However, these
beginning of this section, the resistance to mechanical units and the boundary conditions
bending apparently exerts an important control between them are prescribed in setting up a
on the shape of the domed strata, and this resis- problem in plate theory, so one cannot address
tance is strongly dependent upon the thickness the questions posed in the title of this section.
of the mechanical units in the sedimentary Instead one can use a model based on Gilberts
sequence. The shape of the domed strata also concept that the earliest stage in the development
depends upon the pressure distribution in the of laccoliths is the insinuation of a thin horizon-
laccolith, the distribution of slip between the tal sill of magma between the strata (Fig. 12.10a).
mechanical units, and the nature of the boundary When the horizontal dimension, 2a, of such a sill
conditions at the distal edge of the laccolith, all of is less than the depth, d, the resistance to bending
which have been investigated using plate theory of the thick overburden is so great that opening of
(Pollard and Johnson, 1973; Koch et al., 1981; Kerr the sill is accommodated primarily by the elastic
and Pollard, 1998). Additional insights have been compression of the surrounding rock, both above
gained by addressing the silllaccolith transition and below the sill. The dashed lines in Fig. 12.10a
using elasticity theory (Zenzri and Keer, 2001). schematically represent how bedding planes
would displace upward and downward to accom-
12.2.2 Elasticity theory: how and where modate opening of the sill. Note that the dis-
do bedding-plane faults form placements decrease away from the sill and are
over sills? negligible at Earths surface.
Field observations of bedding-plane faults within The state of stress on horizontal planes around
strata overlying the laccoliths at the three south- a model sill are calculated to determine where
ern Henry Mountains provide evidence that the bedding-plane faulting might initiate (Jackson
overburden was sub-divided into mechanical and Pollard, 1990). The context, or general bound-
units on the order of 200 m thick and capable of ary conditions, for this analysis is elasticity theory
(a) Hookes Law is used to compute the stress com-

ponents (Chapter 8). These stresses are the supple-
d mentary stress field due to opening of the model sill.
A lithostatic stress eld due to gravity acting on
Sill the body is added to nd the total stress eld.
The coordinate system is oriented so the x- and
2a y-axes are horizontal, the z-axis is vertical, and the
origin is at the traction-free surface (representing
Earths surface) directly above the model sill (Fig.
(b) z 12.10b), which has a square tipline in this three-
Traction-free
y surface dimensional half-space. The sill lies in the z d
x
szz plane over the square area a x a and a
Elastic
half-space szx y a. Before opening, the upper and lower sur-
sxx faces of the sill are an innitesimal distance, ,
above and below the z d plane. The boundary
uz conditions, written in terms of the displace-
ment components on these surfaces, produce an
opening displacement discontinuity in the z-
Opening displacement
direction of magnitude uz dened as:
discontinuity
uz uz(z d ) uz(z d ) (12.8)
Fig 12.10 Idealized model for sill dilation in an elastic half-
space. (a) Dashed lines indicate displacement of horizontal The upper surface of the model sill displaces in
markers. (b) Stress components acting on horizontal markers the positive z-direction, and the lower surface dis-
are used to predict location of bedding-plane faults using the places in the negative z-direction, so the opening
Coulomb criterion.
is uz. Displacements of these two adjacent sur-
faces in x and y are zero. These boundary condi-
(Muskhelishvili, 1954; Timoshenko and Goodier, tions describe what is commonly referred to as a
1970; Barber, 1992). To apply elasticity theory to dislocation surface or a surface of displacement dis-
the problem of bedding-plane faulting the sedi- continuity. The displacement components are con-
mentary units above and below the sill are ideal- tinuous everywhere in the elastic body except for
ized as part of a homogeneous and isotropic paths across the dislocation surface, at which
elastic body (Fig. 12.10b). Thus, differences in the there is an abrupt change in sign with no change
elastic properties of the rocks making up the sed- in magnitude.
imentary sequence are ignored and the two The solution for the rectangular dislocation
isotropic elastic constants for the rock mass are surface (Okada, 1985) is too complicated to repro-
taken as the shear modulus, G, and Lams con- duce here but stress elds computed from this
stant (Chapter 8). solution are illustrated in Fig. 12.11. The opening
Naviers displacement equations of motion of the model sill induces changes in the state of
(7.135)(7.137) written for quasi-static conditions stress everywhere in the surrounding region and
and in the absence of body forces are: the analysis seeks to identify those locations
where stress changes on horizontal planes are
2ui 2uk
G (G ) 0 (12.7) conducive to bedding-plane faulting. The normal
xkxk xixk
and shear stress components acting across hori-
Here the displacement components ui are taken as zontal planes in the model are zz and zx. Recall
ux, uy, and uz and the distinction between spatial that the computed stresses represent the supple-
and material coordinates is ignored. From a solu- mentary stress state, so the stress state due to
tion to (12.7) for the displacement components gravity must be added to the normal components.
the kinematic equations are used to compute The component of shear stress induced on hori-
the innitesimal strain components and then zontal planes, zx, would promote bedding-plane
(a) normal stress induced by opening of the sill (the

supplementary stress state) and the ambient
normal stress due to the weight of overburden
(Andersons standard state). The pore uid pres-
sure from a hydrostatic column of groundwater
with water table at the free surface of the half-
space reduces this normal stress to an effective
normal stress. For the calculations used to produce
Fig. 12.11 the constants are i 0.85 and (r w)
(b) 1.4 103 kg m3.
Vertical distance (km)
Contours of the Coulomb stress are plotted in

Fig. 12.11 for a sill at a depth d 4 km with lengths
2a 1.0, 3.0, and 4.0 km. The solution for the
elastic boundary value problem of a crack with
internal pressure suggests that the opening, uz,
should be proportional to the sill length for con-
stant magma pressure, P. Here the ratio of
opening to length is xed at uz/a 1/200, so the
(c) three plots are for sills that are 5, 15, and 20 m
thick, respectively. Regions of positive Coulomb
stress are stippled to indicate where bedding-
plane faulting is likely to develop. For 2a 1.0 km
(Fig. 12.11a), the Coulomb stress is only positive in
the immediate vicinity of the sill tips and the
regions of stress concentration associated with
sill opening are more or less symmetric above and
below the tip. We would expect any bedding plane
Horizontal distance (km) slip to be localized around the advancing tip of
Fig 12.11 Cross sections through a model sill with the sill. At distances from the center of the sill
Coulomb stress contoured. Stippled region corresponds to greater than its length, the stress eld is domi-
that area where bedding-plane faults are predicted. Reprinted nated by the lithostatic stress of the rock and the
from Jackson and Pollard (1990) with permission of The hydrostatic stress of the groundwater, both of
Geological Society of America. which are linear functions of depth. There, the
contours of Coulomb stress are nearly horizontal
and the values decrease at about 12 MPa km1
faulting, whereas the component of compressive from the surface. Immediately over and under the
normal stress, zz, would resist faulting (Segall and sill the Coulomb stress is negative due to the ver-
Pollard, 1980; Oppenheimer et al., 1988). The rela- tical compression of the rock as it accommodates
tionship that quanties this tendency for faulting the opening of the sill. The presence of the sill is
is based on the Coulomb criterion (see Chapter 9) virtually undetectable at Earths surface.
and is measured as the Coulomb stress, C, which Once the sill has advanced to a length of
is calculated following (9.40): 3.0 km (Fig. 12.11b), the regions of positive
Coulomb stress extend well above and below the
C |zx | i[zz (r w)g*z] (12.9)
sill tips. Also, two regions of positive Coulomb
Here i is the coefcient of internal friction; r and stress have developed at the free surface and
w are the average mass density of the host rock and extend about 500 m downward, toward the model
the groundwater, respectively; and g* is the mag- sill tips. These changes indicate that the sill is
nitude of the near-surface acceleration of gravity. starting to interact mechanically with the trac-
Note that the normal stress is a combination of the tion-free surface. At a length of 4.0 km (Fig.
12.11c), this interaction is signicant and the laccolith (Fig. 4.12). The context, or general bound-
regions of positive Coulomb stress extend from ary conditions, for the analysis of magma ow is
the surface to the sill tips, and these regions that branch of uid mechanics devoted to the ow
extend laterally to a distance of almost 4 km from of uids with mechanical properties that include
the point immediately over the center of the sill. viscosity and strength. Reference textbooks on
The overburden out to this distance is susceptible uid mechanics provide the background for
to the development of bedding-plane faults, model development and include solutions for
except for the region immediately over the center ow of materials with a variety of physical prop-
of the sill where enhanced vertical compression erties in conduits of various shapes (Lamb, 1945;
would prevent frictional slip. Delamination is pre- Schlichting, 1979; Landau and Lifshitz, 1960;
dicted to develop above the distal margin of the White, 1974).
sill and not immediately over its center. This dis- In Section 4.3.2 we used dimensional analysis
tribution of bedding-plane faulting would to identify Reynolds Number (4.55) as the scale
promote the formation of laccoliths with at tops factor for viscous ow in conduits (Reynolds, 1883;
and monoclinal bending over their periphery White, 1974). This dimensionless group is a ratio of
(Koch et al., 1981). inertial to viscous forces in the owing uid and is
From the elasticity theory we understand that proportional to the conduit width, uid density,
the stress perturbation associated with the lateral and characteristic velocity, and inversely propor-
growth of a sill enhances the shear stresses and tional to the Newtonian viscosity. Reynolds lab-
lowers the normal compressive stresses on hori- oratory experiments (Fig. 4.8c) demonstrated that
zontal bedding planes in such a way that faulting the ow regime is laminar for numbers less than
is likely on weak bedding planes above the advanc- about 2000. The apparent viscosities of silicate
ing sill tip. As the sill approaches a length equal to liquids (magma) have been measured for a wide
the overburden thickness, the delamination variety of chemical compositions and water con-
spreads all the way to Earths surface and we would tents over the range of melting temperatures
anticipate a transition to laccolithic bending, (Shaw, 1963, 1969; Shaw et al., 1968; Murase and
accommodated by sliding of the mechanical units McBirney, 1973; McBirney and Murase, 1984; Ryan
over one another along bedding-plane faults. and Blevins, 1987). Given the great viscosity of
Elasticity theory and the Coulomb failure criterion most magmas and the modest velocities for ow in
explain how and where these faults form, and this sills the regime is likely to be laminar.
prediction is consistent with eld observations of For the purpose of a simple example we con-
bedding-plane faults at Mt. Holmes in the Henry sider an isothermal uid, the model magma,
Mountains (Fig. 12.9). owing in a tabular conduit of length 2a and
height 2h, surrounded by rigid host rock (Fig.
12.2.3 Viscous fluid mechanics: how 12.12a). A feeder dike at the center of the sill sup-
rapidly can magma flow into sills? plies the magma. Of course hot magma emplaced
The questions asked in the previous two sections into cold sedimentary rock will loose heat to the
focus attention on host rock deformation during surroundings (Lovering, 1935, 1936; Jaeger, 1957,
sill and laccolith formation. The mechanical role 1964b), but rock is a good insulator, so it is not
of the magma is reduced to providing a pressure unreasonable to postulate that the temperature
on the stack of plates or on the elastic material change is insignicant over the time required for
surrounding the sill. In other words the mechani- the sill to propagate to the transition length for
cal action of the magma is replaced by the appro- laccolith formation. In Section 4.2.2 we reviewed
priate distribution of tractions in the form of a the solution for conductive heat loss from a
boundary condition. This is an effective way to tabular intrusion of magma and showed that
simplify the mechanical system composed of both emplacement times on the order of a few days are
injecting magma and deforming host rock, but it consistent with this postulate (Delaney and
necessarily means that one cannot address ques- Pollard, 1981, 1982). As the previous sections indi-
tions about the rate of development of the sill or cate, the sill grows in length and thickness as
variations in space and time of the pressure and

(a)
the velocity components. Flow through a sill is
idealized in Fig. 12.12b as ow through a conduit
2a with straight sides parallel to the x-axis and sepa-
Tip
rated by a constant height, 2h. Because the
2h conduit is in the horizontal (x, y)-plane, the only
non-zero component of gravitational acceleration
Sill is gz g*. We postulate that the ow is entirely in
the x-coordinate direction and that it is steady
Feeder
Rigid host rock state, so the components of velocity are:
vx f(z), vy 0 vz (12.10)
(b) z
Because vy and vz are both zero, the continuity
equation (Section 7.3.2) insures that vx does not
vary in the x-direction. Because we postulate that
Pf vx x Pt all cross sections parallel to the (x, z)-plane are
2h
identical, vx does not vary in the y-direction.
Furthermore, the steady-state condition requires
h
that vx is not a function of time.
With these constraints on the gravitational
a
and velocity vectors, the NavierStokes equations
for the pressure and velocity distributions reduce
Fig 12.12 Idealized model for flow of viscous magma in a
sill. (a) Geometry of sill conduit with feeder dike. (b) Velocity to:
profile in model sill. p d2v p p
2x 0, 0, g* 0
x dz y z
magma is injected from the feeder dike. Here the (12.11)
ow is determined for conduits with particular
The rst equation governs the rate of ow in the x-
ratios of thickness to length, representing differ-
direction and the distribution of this velocity
ent moments during development of the sill. The
from the top to the bottom of the model sill. The
solution does not describe the progressive devel-
second and third equations, respectively, require
opment from an early stage to a later stage in this
the pressure to be constant in the y-coordinate
process and neglects the vertical component of
direction, and the pressure to vary linearly in the
velocity. We also neglect any consideration of the
vertical z-direction in proportion to the unit
propagation mechanism or ow near the tip of
weight of the magma. The distribution of velocity
the conduit, which is postulated to advance with
from the bottom to the top of the model sill is
the same velocity as the magma.
found by integration with a no-slip boundary con-
For the purpose of this example we postulate
dition at the conduit walls:
that the model magma has a constant density, ,
and behaves like an isotropic Newtonian viscous
uid with constant viscosity, . In other words
this uid obeys the Stokes condition described in
vx
h2 p
2 x z2
z2
1 2 vx(max) 1 2
h h (12.12)
Section 7.4.3 and the constitutive law given in The maximum velocity is at the center, z 0, and
(7.166). The viscosity of magma increases dramati- is proportional to the square of the conduit thick-
cally as temperature decreases, but we have ness and the pressure gradient in the ow direc-
already postulated an isothermal ow, so it is con- tion, and inversely proportional to the
sistent to ignore changes in viscosity. Under these Newtonian viscosity. The distribution of velocity
conditions Cauchys Laws of Motion reduce to the across the conduit is symmetric and parabolic
NavierStokes equations (7.170) which describe (Fig. 12.12b).
We evaluate the velocity of magma ow in a 1.E+00
Magma velocity (m s1)

sill, preceding the development of a laccolith, by 1.E-01
putting (12.12) in the following form:
1.E-02
vx(max)
2
h2 P t Pf
a (12.13) 1.E-03
1.E-04
Here 2h and 2a are the thickness and length of the 1 MPa s
sill, whereas Pt and Pf are the magma pressures at 1.E-05 100 MPa s
the tip and the feeder. For the Henry Mountains 10 000 MPa s
1.E-06
0 200 400 600 800 1000 1200 1400
examples, the sills that developed into the larger
laccoliths formed at about 4 km depth where the Sill length (m)
lithostatic pressure would be about Pw 100 MPa. Fig 12.13 Plot of magma velocity versus sill length for
The magma pressure in excess of the lithostatic three different viscosities.
pressure at this depth has been estimated to range
from 30 to 70 MPa, based on the density of the
magma, the density stratication of the host rock, would range between about 105 and 1 m s1. This
and the depth to the source of the magma implies that sills would take between about 103 s
(Johnson and Pollard, 1973). Taking 50 MPa as rep- (25 minutes) and 108 s (5 years) to develop to
resentative of this excess pressure, the total the transition stage. While the time scale is not
magma pressure at the feeder of the sill would be very well constrained, it clearly is a very short
Pf 150 MPa. We postulate that the magma pres- time compared to geologic eons.
sure would decrease to the lithostatic pressure at The results we have just obtained ignore the
the tip of the sill, so Pt 100 MPa. To be consistent deformation of the rock into which the magma is
with the previous section, we use a thickness to injected. A number of papers have treated the
length ratio of h/a 1/200. If the viscosity is taken coupled problem of host rock deformation and
as 102 MPa s, a value representative of silica- magma ow during dike and sill emplacement
rich magma (Johnson and Pollard, 1973), a rela- (Spence and Turcotte, 1985; Lister, 1990; Lister and
tively short sill, a 200 m, would have a maximum Kerr, 1991; Rubin, 1995). Comparing the energy
velocity of about 0.001 ms1 (Fig. 12.13, square consumed by fracturing, which is taken as inde-
symbols). For this sill nearing the transition to lac- pendent of the intrusion length, and that con-
colithic bending of the overburden, a 1500 m, sumed by viscous ow, which is taken as
the lesser pressure gradient would tend to increasing linearly with intrusion length, it is
decrease the velocity, but this is more than com- clear that viscous dissipation will dominate
pensated for by the greater sill thickness, so the beyond some critical length. For typical laboratory
maximum velocity would be about 0.01 m s1. values of fracture energy this length is on the
Magma viscosity is very sensitive to tempera- order of 1 m. However, it has been suggested that
ture and water content, and the constitutive the fracture energy may not be constant and that
properties of the magma are likely to be non- the region of inelastic deformation at the tip of a
Newtonian (Johnson and Pollard, 1973). These dike or sill may increase in size with the length of
factors are not well constrained for the intrusions the intrusion (Rubin, 1993). Under these condi-
in the Henry Mountains, so the value of viscosity tions the fracture energy should not necessarily be
chosen above could be in error by a couple orders neglected, particularly where geological evidence
of magnitude. Based on the linear relationship supports the development of such large regions of
between viscosity and velocity found above, this inelastic deformation (Delaney et al., 1986).
would imply differences in velocity of a couple As sills propagate laterally and then bulge
orders of magnitude. To illustrate this range of upward to form laccoliths, magma must be con-
behaviors we plot velocities for 10 and for 104 tinually injected from below. This injection of
MPas in Fig. 12.13. From this plot we conclude that magma results in the transport of heat from a
the maximum ow velocity for a viscous magma source at greater depth into and throughout the
shallow intrusion. To address questions about the 8.15). The outcome of a structural investigation is
cooling of the magma in tabular intrusions, and judged to be successful if there is a compelling
heating of the host rock, a number of models have correspondence between the views provided by
been investigated, some that consider heat ow observation and modeling.
from a stagnant magma (Lovering, 1935, 1936; At times these different views of structural
Jaeger, 1957, 1964b; Irvine, 1970), and others that geology seem too disparate to be reconciled. The
consider the transport of heat within and from a one is the world of boots and backpacks, rock saws
owing magma (Delaney and Pollard, 1982; and microscopes, maps and photographs, com-
Habert and De Saint-Blanquat, 2004). The context, passes and measuring tapes. The other is the
or general boundary conditions, for these analy- world of vectors and tensors, material continua
ses is the subject of heat transfer, and there are and differential equations, keyboards and com-
many useful textbooks on this subject (Carslaw puters, graphs and numbers. For some practition-
and Jaeger, 1959; Bird et al., 1960). ers of structural geology a choice is made at an
early stage in their education that closes the door
on one of these worlds in favor of the other. One
12.3 A methodology for the objective of this textbook is to encourage struc-
tural geologists to integrate these two worlds and
practice of structural geology discover the benets of both.
A methodology for the practice of structural
Newtons axiomatic framework allowed him to pursue geology that integrates observations and modeling
a strategy in which he could construct a simplied,
is illustrated in Fig. 12.14 as a set of ten stepping
idealized mathematical model of the physical system
stones on a path with the suggestion that this be
he wanted to probe in this case, the solar system.
Using mathematics, Newton could work out the conse-
traversed in a counterclockwise sense starting at
quences of certain actions and compare them with the top. Of course scientic investigations are
measurements and empirical observations. That com- rarely this well organized or rationalized. Instead
parison, in turn, would suggest ways in which the of beginning with a eld observation one may be
model could be adjusted and rened to achieve even inspired to investigate a new problem while
greater realism. In essence, this strategy of maintain- reading a textbook on uid mechanics or contem-
ing a right interplay between mathematical analysis plating the solution to an elastic boundary value
and physical experience afforded a marvellously pro- problem displayed as velocity vectors on a com-
ductive way of using mathematics to explain the work- puter screen. Regardless of the source of inspira-
ings of nature. Revolutionary in Newtons time, this
tion one could argue that seeking the map or
kind of approach is taken for granted in modern
photograph or measurement in the eld to conrm
research (Peterson, 1993).
that the phenomenon in question occurs naturally
The material presented in this textbook reveals in Earths crust is a pre-requisite to launch an inves-
two distinct views of structural geology. On the tigation. Thus, we nd ourselves at the top of the
one hand we have described observations of struc- diagram and suggest that this step of seeking data
tures: for example views through a microscope of in the eld should benet from the techniques of
tiny spherical objects that were deformed into eld-based structural geology described in Chapter
ellipsoidal shapes as rocks were contorted into the 2. Also, for the quantitative characterization of
South Mountain fold (Fig. 5.4), or photographs of structures we advocate the use of differential
exposures that show successive stages in the devel- geometry, some of which is described in Chapter 3.
opment of strike slip faults in granitic rock of the Once launched along the path (Fig. 12.14) it is
Sierra Nevada (Fig. 9.37). On the other hand we common to step off and move directly to a step-
have described models of structures: for example ping stone out of sequence, or even to backtrack
a plot of principal stress trajectories associated along the path. None-the-less this simple diagram
with a pattern of dikes around the Spanish Peaks provides a way to organize the primary scientic
(Fig. 6.37), or a plot of displacement vectors asso- procedures employed by a structural geologist
ciated with the 1999 Hector Mine earthquake (Fig. and to understand how they might relate to one
12.3 A METHODOLOGY FOR THE PRACTICE OF STRUCTURAL GEOLOGY 475
Field observations
Maps
Photographs
Measurements
Descriptions
Idealizations Rules of correspondence

Deleting irrelevant data Conceptual correlation
Adding necessary features Numerical comparison
Simplifying essential features Statistical inference
Postulating first-order causes Inverse theory
Visualization techniques
General boundary conditions
Graphs
Solid mechanics
Contour maps
Fluid mechanics
3D projections
Heat transport
Animations
Solutions
Fundamental relationships Stress
Conservation laws Displacement
Laws of motion Velocity
Constitutive laws Temperature
Governing equations Solution methods

Biharmonic (elastic stress) Analytical
NavierStokes (viscous flow) Numerical
Fourier (heat conduction) FDM, FEM, BEM
Specific boundary conditions

Model geometry
Boundary conditions
Initial conditions
Fig 12.14 A methodology for the practice of structural arguably, the most challenging steps to describe
geology. and, therefore, to teach. It is generally not possi-
ble to include all of the features of a structure in
another. For example, in Chapter 12 we have a model and those that are included usually are
focused on two of the stepping stones near the idealized. Idealization is given as the second step
beginning of the path: idealization of observed along the path and is a transition from observa-
structures and selection of general boundary con- tions to a model. It is largely done in the eld
ditions for model development. These are, during the process of mapping and measurement.
Here one postulates rst-order cause-and-effect viscous ow problem. Then the specic boundary
relationships and uses these to exclude observa- and/or initial conditions are selected to provide
tions that are deemed to be irrelevant. the constants of integration necessary to solve the
The third step involves the selection of the governing equations. Sometimes, the specic
branch of continuum mechanics that provides boundary conditions may be constrained by eld
the general boundary conditions for modeling. measurements, thereby establishing a direct link
Here the scaling arguments introduced in with the stepping stone at the top of the diagram.
Chapter 4 should prove helpful, and the concepts In other cases the specic boundary conditions
of strain, rate of deformation, traction, and stress are chosen arbitrarily, just to see what the
described in Chapters 5 and 6 must be understood. outcome might be, and thereby learn about the
Although some sub-disciplines of continuum behavior of the model structure. We have illus-
mechanics are spelled out on the third stepping trated both of these approaches to modeling with
stone, this list is not meant to be exclusive or to examples throughout the text. In Chapters 8, 9,
imply that coupled problems, for example of uid 10, and 11 we have employed methods that
ow and solid deformation, are unimportant. achieve analytical solutions to particular prob-
Indeed, the level of understanding of coupled lems in elasticity theory or uid mechanics. In the
problems among structural geologists and the simpler cases these solutions are derived, but
modeling resources now available in terms of others are taken from the vast literature on these
both hardware and software suggest that this will subjects without derivation to illustrate a partic-
be a very fruitful area of investigation in the ular point or concept in the context of the
twenty-rst century. seventh stepping stone. In a few cases solutions
After selecting the theoretical context for an were obtained by numerical methods but these
investigation one should focus on the fundamen- methods are not even introduced here, much less
tal relationships, two of which are described in developed from rst principles. The nite-differ-
Chapter 7 as the conservation laws of mass and ence method (FDM), nite-element method (FEM),
momentum. This fourth step along the path com- and boundary element method (BEM) have been
monly is overlooked, but it provides a touchstone exploited by engineers and scientists to nd
that makes clear how the Laws of Motion have numerical solutions to boundary value problems
been formulated and how the constitutive laws (Timoshenko and Goodier, 1970; Crouch and
are employed to reduce these to a specic set of Stareld, 1983; Hughes, 1987).
governing equations identied with the fth step Whether found using analytical or numerical
along the path. At this point one has made an methods, the solutions themselves are the eighth
explicit commitment to a material behavior and step in this methodology and they provide the
perhaps simplied the equations of motion to eld quantities such as stress, displacement,
exclude temporal or spatial variations in material velocity, and temperature as functions of the
properties. In this textbook we give considerable spatial coordinates and time. Only in rare
attention to the linear elastic material as instances are the distributions in space and time
described in Chapter 8 and the linear viscous of these quantities so simple as to be understand-
material as described in Chapter 10. One of the able from an analytical equation. Therefore, one
most commonly employed governing equations in employs visualization techniques, the ninth step-
the elastic context is the biharmonic equation for ping stone, that range from two-dimensional
the stress function in two dimensions. In the graphs and contour plots to three-dimensional
viscous context the NavierStokes equations are projections and animations. Throughout this text
the benchmark. we have used Matlab to aid in visualization.
In the sixth step one chooses a particular Because the scripts that produce these illustra-
model geometry, which may reduce the govern- tions are available, visualization is a dynamic
ing equations to two or even one spatial dimen- process in which the reader can adjust the bound-
sion. Here one may eliminate time altogether, as ary conditions or values of the parameters and see
in a quasi-static elastic problem or steady-state the response.
The last stepping stone along the path of Fig.

12.14 links the output of a modeling investigation
to eld observations through rules of correspon-
dence (Margenau, 1977). These come in various In summary, we advocate a methodology that
forms and levels of mathematical and statistical applies the extensive machinery of the physics and
rigor. In the simplest form a concept, such as the chemistry of the processes involved, most centrally
linear relationship between depth and diameter rock deformation, both to suggest observations
of laccoliths as conceived by Gilbert (Fig. 1.18), is and to analyze, visualize, and interpret them. In
tested by estimating the stratigraphic position of this way, a coherent and self-consistent, if ideal-
laccoliths in the Henry Mountains (Table 1.1). In a ized, and non-unique, re-construction of the devel-
more quantitative evaluation the calculated slip opment of a geological structure may be achieved.
directions for intersecting normal faults in an We concede that tectonic processes and their prod-
elastic model are compared to eld measure- ucts will not be completely described by mechani-
ments of slickenline orientations from the cal models, but advocate the position that the
Chimney Rock area (Fig. 2.31). In another example, mechanics must be complete. Thus, for prescribed
linear inverse theory is used to calculate the slip initial and boundary conditions a forward model is
distribution on a fault model for the 1999 Hector generated that may produce likenesses of some of
Mine earthquake (Fig. 8.15) as constrained by the observed geological structures and fabrics, to
surface displacement data from satellite radar some satisfactory degree of approximation. If the
images. The calculated slip distribution is then forward model fails to produce satisfactory like-
used as a boundary condition in a forward elastic nesses we learn that one or more of the postulates
model and the resulting surface displacements is inappropriate and must be excluded or modied.
compared to those from the radar image. In all of One cannot, however, exclude or modify the fun-
these examples the correspondence proved damental laws upon which the mechanical model
sufciently compelling to convince the structural is based. The general method we advocate is the
geologist that the investigation was on the right construction of a sequence of quantitative models,
track. Discrepancies, on the other hand, motivate graduated in their degree of detail, and succes-
further eld work and modeling. sively providing an improved understanding.
References
Achenbach, J. D., 1973, Wave Propagation in Elastic Solids. Microstructure of deformation bands in porous
New York: American Elsevier. sandstones at Arches National Park, Utah. Journal of
Ackermann, R. V., Schlische, R. W., and Withjack, M. O., Structural Geology 16: 94159.
2001, The geometric and statistical evolution of Archuleta, R., 1984, A faulting model for the 1979
normal fault systems: an experimental study of the Imperial Valley earthquake. Journal of Geophysical
effects of mechanical layer thickness on scaling Research 89: 455985.
laws. Journal of Structural Geology 23: 180319. Argand, E., 1911, Les nappes de recouvrement des Alpes
Aczel, A. D., 2000, Mystery of the Aleph: Mathematics, the Pennines et leurs prolongements structuraux.
Kabbalah, and the Search for Infinity. New York: Four Beitrge zur geologischen Karte der Schweiz 31: 125.
Walls, Eight Windows. Arrowsmith, J. R., Pollard, D. D., and Rhodes, D. D.,
Adams, F. and Nicholson, J., 1901, Experimental investi- 1996, Hillslope development in areas of active tec-
gation into the flow of marble: Philosophical tonics. Journal of Geophysical Research, B, Solid Earth and
Transactions of the Royal Society of London 195: 363401. Planets 101(3): 625575.
Aki, K. and Richards, P. G., 1980, Quantitative Seismology: Arrowsmith, J. R., Rhodes, D. D., and Pollard, D. D.,
Theory and Methods. San Francisco: W. H. Freeman. 1998, Morphologic dating of scarps formed by
Albritton, C. C., Jr., 1980, The Abyss of Time: Changing repeated slip events along the San Andreas Fault,
Conceptions of the Earths Antiquity after the Sixteenth Carrizo Plain, California. Journal of Geophysical
Century. San Francisco: Freeman, Cooper and Co. Research, B, Solid Earth and Planets 103(5): 10 14160.
Alpha, T. R., Vigil, J. R., and Buchholz, L., 1988, Map Atkinson, B. K., 1987, Fracture Mechanics of Rock. London:
Projections Via Hyper Card, Open File Report 88364. Academic Press.
Menlo, CA: US Geological Survey. Atkinson, B. K. and Meredith, P. G., 1987, Experimental
Amadei, B. and Stephansson, O., 1997, Rock Stress and its fracture mechanics data for rocks and minerals. In
Measurement. London: Chapman and Hall. B. K. Atkinson, ed., Fracture Mechanics of Rock. London:
Anderson, D., 1975, Chemical plumes in the mantle. Academic Press, pp. 477525.
Geological Society of America Bulletin 86: 15931600. Aydin, A., 1977, Faulting in sandstone. Unpublished
Anderson, E. M., 1936, The dynamics of the formation Ph.D. thesis, Stanford University.
of cone-sheets, ring-dykes, and caldron-subsidences. Aydin, A., 1978, Small faults formed as deformation
Proceedings of the Royal Society of Edinburgh 56: 12856. bands in sandstone. Pure and Applied Geophysics 116:
Anderson, E. M., 1951, The Dynamics of Faulting and Dyke 91330.
Formation with Application to Britain. Edinburgh: Aydin, A., 2000, Fractures, faults, and hydrocarbon
Oliver and Boyd. entrapment, migration and flow. Marine and
Anderson, E. M., 1972, The Dynamics of Faulting and Dyke Petroleum Geology 17(7): 797814.
Formation with Applications to Britain. New York: Aydin, A. and DeGraff, J. M., 1988, Evolution of polyg-
Hafner. onal fracture patterns in lava flows. Science 239:
Anderson, T. L., 1995, Fracture Mechanics. Boca Raton: 4716.
CRC Press. Aydin, A. and Eyal, Y., 2002, Anatomy of a normal fault
Andrews, D. J. and Hanks, T. C., 1985, Scarp degrada- with shale smear: implications for fault seal.
tion by linear diffusion: inverse solution for age. American Association of Petroleum Geologists Bulletin
Journal of Geophysical Research 90: 10 193208. 86(8): 136781.
Antonellini, M. and Aydin, A., 1994, Effect of faulting Aydin, A. and Johnson, A. M., 1978, Development of
on fluid flow in porous sandstones: petrophysical faults as zones of deformation bands and as slip sur-
properties. American Association of Petroleum Geologists faces in sandstone. Pure and Applied Geophysics 116:
Bulletin 78: 33577. 93142.
Antonellini, M. and Aydin, A., 1995, Effect of faulting Aydin, A. and Schultz, R. A., 1990, Effect of mechanical
on fluid flow in porous sandstones: geometry and interaction on the development of strike-slip faults
spatial distribution. American Association of Petroleum with echelon patterns. Journal of Structural Geology 12:
Geologists Bulletin 79: 64271. 1239.
Antonellini, M. A., Aydin, A., and Pollard, D. D., 1994, Badoux, H., 1963, Les blemnites tronones de
REFERENCES 479
Leytron (Valais). Bulletin de Gologie, Minralogie, Bieniawski, Z. T., 1984, Rock Mechanics Design in Mining
Gophysique et du Muse Gologique de lUniversit de and Tunneling. Rotterdam: A. A. Balkema.
Lausanne 138: 17. Bieniawski, Z. T. and Bernede, M. J., 1979, Suggested
Bahat, D., 1991, Tectonofractography. New York: Springer- methods for determining the uniaxial compresssive
Verlag. strength and deformability of rock materials.
Bai, T. and Pollard, D. D., 2000a, Closely spaced frac- International Journal of Rock Mechanics and Mining
tures in layered rocks: initiation mechanism and Sciences and Geomechanics Abstracts 16(2): 13540.
propagation kinematics. Journal of Structural Geology Billings, M. P., 1972, Structural Geology. Englewood
22: 140925. Cliffs, NJ: Prentice-Hall.
Bai, T. and Pollard, D. D., 2000b, Fracture spacing in Biot, M. A., 1961, Theory of folding of stratified visco-
layered rocks: a new explanation based on the stress elastic media and its implications in tectonics and
transition. Journal of Structural Geology 22: 4357. orogenesis. Geological Society of America Bulletin 72:
Bai, T., Pollard, D. D., and Gao, H., 2000, Explanation 1595620.
for fracture spacing in layered materials. Nature 403: Biot, M. A., 1965, Mechanics of Incremental Deformation.
7536. New York: John Wiley and Sons.
Barber, J. R., 1992, Elasticity. Dordrecht: Kluwer Birch, F., 1966, Compressibility: elastic constants. In
Academic. S. P. Clark, ed., Handbook of Physical Constants. New
Barton, N., 1973, Review of a new shear-strength cri- York: The Geological Society of America,
terion for rock joints. Engineering Geology 7: 287332. pp. 97173.
Barus, C., 1891, The viscosity of solids. US Geological Bird, R. B., Stewart, W. E., and Lightfoot, E. N., 1960,
Survey Bulletin 73. Transport Phenomena. New York: John Wiley and Sons.
Batchelor, G. K., 1967, An Introduction to Fluid Mechanics. Bott, M. H. P., 1959, The mechanics of oblique slip fault-
Cambridge: Cambridge University Press. ing. Geological Magazine 96: 10917.
Batt, G. E., Brandon, M. T., Farley, K. A., and Roden-Tice, Bourne, S. J., 2003, Contrast of elastic properties
M., 2001, Tectonic synthesis of the Olympic between rock layers as a mechanism for the initia-
Mountains segment of the Cascadia wedge, using tion and orientation of tensile failure under remote
two-dimensional thermal and kinematic modeling compression. Journal of Geophysical Research 108:
of thermochronological ages. Journal of Geophysical 10.1029/2001JB001725.
Research 106: 26 73146. Bourne, S. J. and Willemse, E. J. M., 2001, Elastic stress
Beckmann, P., 1971, A History of (pi). New York: Dorset control on the pattern of tensile fracturing around a
Press. small fault network at Nash Point, UK. Journal of
Bell, J. S. and Babcock, E. A., 1986, The stress regime of Structural Geology 23: 175370.
the western Canadian Basin and implications for Bowden, F. P. and Tabor, D., 1950, The Friction and
hydrocarbon production. Bulletin of the Canadian Lubrication of Solids. Oxford: Clarendon Press.
Petroleum Geologists 34: 36478. Brace, W. F., 1964, Brittle fracture of rocks. In W. R.
Bell, R. J. T., 1920, An Elementary Treatise on Coordinate Judd, ed., State of Stress in the Earths Crust. New York:
Geometry of Three Dimensions. London: Macmillan. Elsevier, pp. 11174.
Bergbauer, S., 2002, The use of curvature for the analy- Brace, W. F. and Bombolakis, E. G., 1963, A note on
ses of folding and fracturing with application to the brittle crack growth in compression. Journal of
Emigrant Gap Anticline, Wyoming. Unpublished Geophysical Research 68: 370913.
Ph.D. thesis, Stanford University. Brace, W. F. and Byerlee, J. D., 1966, Stick slip as a mech-
Bergbauer, S. and Martel, S. J., 1999, Formation of anism for earthquakes. Science 153: 9902.
joints in cooling plutons. Journal of Structural Geology Brandon, M. T., 2004, The Cascadia subduction wedge:
21(7): 82135. the role of accretion, uplift, and erosion. In B. A. van
Bergbauer, S. and Pollard, D. D., 2003, How to calculate der Pluijm and S. Marshak, eds., Earth Structure: An
normal curvatures of sampled geological surfaces. Introduction to Structural Geology and Tectonics. New
Journal of Structural Geology 25: 27789. York: W. W. Norton, pp. 56674.
Bergbauer, S. and Pollard, D. D., 2004, A new concep- Braunstein, J. and OBrien, G. D., 1968, Diapirism and
tual foldfracture model including prefolding Diapirs: Memoir. Tulsa, OK: American Association of
joints, based on the Emigrant Gap anticline, Petroleum Geologists.
Wyoming. Geological Society of America Bulletin 116: Bredehoeft, J. D., Wolff, R. G., Keys, W. S., and Shuter, E.,
294307. 1976, Hydraulic fracturing to determine the regional
480 REFERENCES
in situ stress field, Piceance Basin, Colorado. Chandrasekhar, S., 1979, Beauty and the quest for
Geological Society of America Bulletin 87(2): 2508. beauty in science. Physics Today July: 2530.
Bridgman, P. W., 1931, Dimensional Analysis. New Haven, Chapple, W. M., 1968, A mathematical theory of finite-
CT: Yale University Press. amplitude rock-folding. Geological Society of America
Brodkey, R. S. and Hershey, H. C., 1988, Transport Bulletin 79: 4768.
Phenomena: A Unified Approach. New York: McGraw- Chapple, W. M., 1978, Mechanics of thin-skinned fold-
Hill. and-thrust belts. Geological Society of America Bulletin
Brown, E. T. and Hoek, E., 1978, Trends in relationships 89: 118998.
between measured in situ stresses and depth. Chinnery, M. A., 1961, The deformation of the ground
International Journal of Rock Mechanics and Mining around surface faults. Bulletin of the Seismological
Science 15: 21115. Society of America 51: 35572.
Buckingham, E., 1915, Model experiments and the Chinnery, M. A., 1963, The stress changes that accom-
forms of empirical equations. Transactions of the pany strike-slip faulting. Bulletin of the Seismological
American Society of Mechanical Engineers 37: 26392. Society of America 53: 92132.
Brgmann, R. and Pollard, D. D., 1992, Influence of the Chinnery, M. A., 1966, Secondary faulting: I and II.
state of stress on the brittleductile transition in Canadian Journal of Earth Sciences 3: 16390.
granitic rock: evidence from fault steps in the Sierra Choukroune, P. and Seguret, M., 1973, Tectonics of the
Nevada, California. Geology 20: 64548. Pyrenees: role of compression and gravity. In K. A. De
Brgmann, R. and Pollard, D. D., 1994, Strain accom- Jong and R. Scholten, eds., Gravity and Tectonics. New
modation about strikeslip fault discontinuities in York: John Wiley and Sons, pp. 14156.
granitic rock under brittle-to-ductile conditions. Clark, S. P. J., 1966, Handbook of Physical Constants. New
Journal of Structural Geology 16: 165574. York: The Geological Society of America.
Brgmann, R., Pollard, D. D., and Martel, S. J., 1994, Slip Cloos, E., 1946, Lineation: A Critical Review and Annotated
distributions on faults: effects of stress gradients, Bibliography, Geological Society of America, Memoir 18.
inelastic deformation, heterogeneous host-rock stiff- Ann Arbor, MI: Edwards Brothers.
ness, and fault interaction. Journal of Structural Cloos, E., 1947, Oolite deformation in the South
Geology 16: 167590. Mountain Fold, Maryland. Geological Society of America
Byerlee, J., 1978, Friction of rocks. Pure and Applied Bulletin 58: 843918.
Geophysics 116: 61526. Cloos, E., 1971, Microtectonics along the western edge
Calladine, C. R. and Drucker, D. C., 1962, Nested sur- of the Blue Ridge, Maryland and Virginia. Baltimore,
faces of constant rate of dissipation in creep. Journal MD: The Johns Hopkins University Press.
of Applied Mathematics 29: 7984. Cobbold, P. R., Cosgrove, J. W., and Summers, J. M., 1971,
Carena, S. and Suppe, J., 2002, Three-dimensional Development of internal structures in deformed
imaging of active structures using earthquake after- anisotropic rocks. Tectonophysics 12: 2353.
shocks: the Northridge Thrust, California. Journal of Cobbold, P. R., Durand, S., and Mourgues, R., 2001,
Structural Geology 24(4): 887904. Sandbox modelling of thrust wedges with fluid-
Carslaw, H. S. and Jaeger, J. C., 1959, Conduction of Heat assisted detachments. Tectonophysics 334: 24558.
in Solids. Oxford: Clarendon Press. Cocco, M. and Rice, J. R., 2002, Pore pressure and poro-
Carter, N. L., Friedman, M., Logan, J. M., and Stearns, elasticity effects in Coulomb stress analysis of earth-
D. W., 1981, Mechanical Behavior of Crustal Rocks (The quake interactions. Journal of Geophysical Research 107:
Handin Volume), Geophysical Monograph. Washington, 117.
DC: American Geophysical Union, p. 326. Collins, G. C., Head, J. W., III, and Pappalardo, R. T.,
Cartwright, J. A., Trudgill, B. D., and Mansfield, C. S., 1998, The role of extensional instability in creating
1995, Fault growth by segment linkage: an explana- Ganymede grooved terrain: insights from Galileo
tion for scatter in maximum displacement and trace high-resolution stereo imaging. Geophysical Research
length data from the Canyonlands grabens of SE Letters 25: 2336.
Utah. Journal of Structural Geology 17: 131926. Committee on the Future of the Global Positioning
Cathles, L. M., 1975, The Viscosity of the Earths Mantle. System, 1995, The Global Positioning System: A Shared
Princeton, New Jersey: Princeton University Press. National Asset. Recommendations for Technical
Cathles, L. M., 1977, An analysis of the cooling of intru- Improvements and Enhancements. Washington, DC:
sions by ground-water convection which includes Aeronautics and Space Engineering Board, National
boiling. Economic Geology 72: 80426. Research Council, National Academy Press.
REFERENCES 481
Compton, R. R., 1962, Manual of Field Geology. New York: echelon normal faults. Journal of Geophysical Research,
John Wiley and Sons. B, Solid Earth and Planets 103(10): 24 37391.
Cooke, M., Mollema, P., Pollard, D. D., and Aydin, A., Crouch, S. L. and Starfield, A. M., 1983, Boundary Element
2000, Interlayer slip and joint localization in East Methods in Solid Mechanics: with Applications in Rock
Kaibab Monocline, Utah: field evidence and results Mechanics and Geological Engineering. London: George
from numerical modeling. In J. W. Cosgrove and Allen and Unwin.
M. S. Ameen, eds., Forced Folds and Fractures, Cruikshank, K. M. and Aydin, A., 1995, Unweaving the
Special Publication. London: Geological Society, joints in Entrada Sandstone, Arches National Park,
pp. 2349. Utah, USA. Journal of Structural Geology 17: 40921.
Cooke, M. and Pollard, D. D., 1996, Fracture propaga- Culling, W. E. H., 1960, Analytical theory of erosion.
tion paths under mixed mode loading within rectan- Journal of Geology 68: 33644.
gular blocks of polymethyl methacrylate. Journal of Dana, J. D., 1847a, Geological results of the Earths con-
Geophysical Research 101: 3387400. traction. American Journal of Science 3: 17688.
Cooke, M. L., 1997, Fracture localization along faults Dana, J. D., 1847b, Geological results of the Earths con-
with spatially varying friction. Journal of Geophysical traction. American Journal of Science 4: 8892.
Research, B, Solid Earth and Planets 102(10): 22 42534. Dahlen, F. A., Suppe, J., and Davis, D., 1984, Mechanics
Cooke, M. L. and Underwood, C. A., 2001, Fracture ter- of fold-and-thrust belts and accretionary wedges:
mination and step-over at bedding interfaces due to cohesive Coulomb theory. Journal of Geophysical
frictional slip and interface opening. Journal of Research 89: 10 087101.
Structural Geology 23(23): 22338. Daly, R. A., Manger, G. E., and Clark, S. P. J., 1966,
Cotterell, B. and Rice, J. R., 1980, Slightly curved or Density of rocks. In S. P. J. Clark, ed., Handbook of
kinked cracks. International Journal of Fracture 16: Physical Constants. New York: Geological Society of
15569. America, pp. 1926.
Coulomb, C. A., 1773, Sur une application des rgles de Davatzes, N. C. and Aydin, A., 2003, Overprinting fault-
maximis et minimis quelques problmes de sta- ing mechanisms in high porosity sandstones of SE
tique relatifs larchitecture. Acad. Roy. des Sciences, Utah. Journal of Structural Geology 25: 1795813.
Mmoires de math. et de physique par divers savants 7: Davies, R. K. and Pollard, D. D., 1986, Relations between
34382. left-lateral strike-slip faults and right-lateral mono-
Couples, G., 1977, Stress and shear fracture (fault) pat- clinal kink bands in granodiorite, Mt. Abbot
terns resulting from a suite of complicated bound- Quadrangle, Sierra Nevada, California. Pure and
ary conditions with applications to the Wind River Applied Geophysics 124: 177201.
Mountains. Pure and Applied Geophysics 115: 11333. Davis, D., Suppe, J., and Dahlen, F. A., 1983, Mechanics
Couples, G. D. and Lewis, H., 1998, Lateral variations of of fold-and-thrust belts and accretionary wedges.
strain in experimental forced folds. Tectonophysics Journal of Geophysical Research 88: 115372.
295: 7991. Davis, E. E. and Lister, C. R. B., 1974, Fundamentals of
Cowan, D., 1992, Chairpersons message. Structural ridge crest topography. Earth and Planetary Science
Geology and Tectonics Division Newsletter. Boulder CO: Letters 21(4): 40513.
Geological Society of America, p. 1. Davis, G. H. and Reynolds, S. J., 1996, Structural Geology
Cowie, P. A. and Scholz, C. H., 1992, Growth of faults by of Rocks and Regions. New York: John Wiley and Sons.
accumulation of seismic slip. Journal of Geophysical Davis, J. C., 1986, Statistics and Data Analysis in Geology.
Research 97: 11 08595. New York: John Wiley and Sons.
Cowie, P. A. and Shipton, Z. K., 1998, Fault tip displace- Davis, P. J. and Hersh, R., 1986, Descartes Dream. Boston,
ment gradients and process zone dimensions. MA: Houghton Mifflin.
Journal of Structural Geology 20: 98397. Dawers, N. H., Anders, M. H., and Scholz, C. H., 1993,
Crick, F., 1988, What Mad Pursuit: A Personal View of Growth of normal faults: displacement-length
Scientific Discovery. New York: Basic Books. scaling. Geology 21: 110710.
Crider, J. G., 2001, Oblique slip and the geometry of De Sitter, L. U., 1964, Structural Geology. New York:
normal-fault linkage: mechanics and a case study McGraw-Hill Book Company.
from the Basin and Range in Oregon. Journal of DeGraff, J. M. and Aydin, A., 1987, Surface morphology
Structural Geology 23(12): 19972009. of columnar joints and its significance to mechanics
Crider, J. G. and Pollard, D. D., 1998, Fault linkage: and directions of joint growth. Geological Society of
three-dimensional mechanical interaction between America Bulletin 99: 60517.
482 REFERENCES
Delaney, P. T., 1990, Deep magma body beneath the equipment, model construction techniques and rhe-
summit and rift zones of Kilauea Volcano, Hawaii. ology of model properties. Journal of Structural Geology
Science 247: 1265372. 7: 83102.
Delaney, P. T., Denlinger, R. P., Lisowski, M., et al., 1998, Dombard, A. J. and McKinnon, W. B., 2001, Formation
Volcanic spreading at Kilauea, 19761996. Journal of of grooved terrain on Ganymede: extensional insta-
Geophysical Research 103: 18 00323. bility mediated by cold superplastic creep. Icarus 154:
Delaney, P. T., Miklius, A., Arnadottir, T., Okamura, 32136.
A. T., and Sako, M. K., 1993, Motion of Kilauea Donath, F. A., 1961, Experimental study of shear failure
Volcano during sustained eruption from the Puu Oo in anisotropic rocks. Geological Society of America
and Kupaianaha vents, 19831991. Journal of Bulletin 72: 98590.
Geophysical Research 98: 17 80120. Donath, F. A. and Parker, R. B., 1974, Folds and folding.
Delaney, P. T. and Pollard, D. D., 1981, Deformation of Geological Society of America Bulletin 75: 4562.
host rocks and flow of magma during growth of Duba, A. G., Durham, W. B., Handin, J. W., and Wang,
minette dikes and breccia-bearing intrusions near H. F., 1990, The BrittleDuctile Transition in Rocks (The
Ship Rock, New Mexico. US Geological Survey Heard Volume), Geophysical Monograph. Washington,
Professional Paper 1202, pp. 161. DC: American Geophysical Union, p. 243.
Delaney, P. T. and Pollard, D. D., 1982, Solidification of Emerman, S. and Turcotte, D. L., 1983, A fluid model
basaltic magma during flow in a dike. American for the shape of accretionary wedges. Earth and
Journal of Science 282: 85685. Planetary Science Letters 63: 37984.
Delaney, P. T., Pollard, D. D., Ziony, J. I., and McKee, Enever, J. R., Cornet, F., and Roegiers, J. C., 1992, ISRM
E. H., 1986, Field relations between dikes and joints: commission on interpretation of hydraulic fracture
emplacement processes and paleostress analysis. records. International Journal of Rock Mechanics and
Journal of Geophysical Research 91(B5): 492038. Mining Science 29: 6972.
Dennis, J. G., 1987, Structural Geology: An Introduction. Engelder, T., 1993, Stress Regimes in the Lithosphere.
Dubuque, IA: William C. Brown. Princeton, NJ: Princeton University Press.
Dieterich, J. H., 1969, Origin of cleavage in folded Engelder, T., and Peacock, D. C. P., 2001, Joint develop-
rocks. American Journal of Science 267: 15565. ment normal to regional compression during flex-
Dieterich, J. H., 1979a, Modeling of rock friction: 1. ural-flow folding: the Lilstock buttress anticline,
Experimental results and constitutive equations. Somerset, England. Journal of Structural Geology 23:
Journal of Geophysical Research 84: 21618. 25977.
Dieterich, J. H., 1979b, Modeling of rock friction: 2. Erdogan, F. and Sih, G. C., 1963, On the crack extension
Simulation of preseismic slip. Journal of Geophysical in plates under plane loading and transverse shear.
Research 84(B5): 216975. Journal of Basic Engineering, Transactions ASME 85:
Dieterich, J. H., 1981, Constitutive properties of faults 51927.
with simulated gouge. In N. L. Carter, M. Friedman, Eshbach, O. W., 1961, Handbook of Engineering
J. M. Logan and D. W. Stearns, eds., Mechanical Fundamentals, Handbook Series. New York: John Wiley
Behavior of Crustal Rocks: the Handin Volume. and Sons.
Washington, DC: American Geophysical Union, Eshelby, J. D., 1957, The determination of the elastic
pp. 10320. field of an ellipsoidal inclusion, and related prob-
Dieterich, J. H., 1986, A model for the nucleation of lems. Proceedings of the Royal Society of London Series A
earthquake slip. Geophysical Monograph 37: 3747. 241: 37696.
Dieterich, J. H., 1994, A constitutive law for rate of Evans, B. and Kohlstedt, D. L., 1995, Rheology of rocks.
earthquake production and its application to earth- In T. J. Ahrens, ed., Handbook of Physical Properties of
quake clustering. Journal of Geophysical Research, B, Rocks. Washington, DC: American Geophysical
Solid Earth and Planets 99(2): 260118. Union, pp. 14865.
Dieterich, J. H. and Carter, N. L., 1969, Stress-history of Faerseth, R. B., Knudsen, B.-E., Liljedahl, T., Midbe,
folding. American Journal of Science, 267: 12954. P. S., and Sderstrm, B., 1997, Oblique rifting and
Dieterich, J. H. and Onat, E. T., 1969, Slow finite defor- sequential faulting in the Jurassic development of
mations of viscous solids. Journal of Geophysical the northern North Sea. Journal of Structural Geology
Research 74: 20818. 19(10): 1285302.
Dixon, J. M. and Summers, J. M., 1985, Recent develop- Fairhurst, C., 1964, Measurement of in situ rock
ments in centrifuge modelling of tectonic processes: stresses with particular references to hydraulic frac-
REFERENCES 483
turing. Rock Mechanics and Engineering Geology 2: Flodin, E. and Aydin, A., 2004, Faults with asymmetric
12947. damage zones in sandstone, Valley of Fire State Park,
Farvolden, R. N. and Cherry, J. A., 1991, Opinion: are southern Nevada. Journal of Structural Geology 26:
geology departments prepared for the 21st century? 9838.
Geology May: 419. Flugge, W., 1967, Viscoelasticity. London: Blaisdell.
Feehan, J. G. and Brandon, M. T., 1999, Contribution of Fossen, H. and Gabrielsen, R. H., 1996, Experimental
ductile flow to exhumation of low Thigh P meta- modeling of extensional fault systems by use of
morphic rocks: San JuanCascade Nappes, NW plaster. Journal of Structural Geology 18: 67387.
Washington State. Journal of Geophysical Research 104: Freund, L. B., 1979, The mechanics of dynamic shear
10 883901. crack propagation. Journal of Geophysical Research 84:
Feshbach, H. and Weisskopf, V. F., 1988, Ask a foolish 2199209.
question. Physics Today October: 9. Friedel, J., 1964, Dislocations. Oxford: Pergamon Press.
Fialko, Y. A., Khazan, Y., and Simons, M., 2001, Friedman, M., Handin, J., and Alani, G., 1972, Fracture-
Deformation due to a pressurized horizontal circu- surface energy of rocks. International Journal of Rock
lar crack in an elastic half-space, with applications Mechanics and Mining Science 9: 75766.
to volcano geodesy. Geophysical Journal International Frocht, M. M., 1948, Photoelasticity. New York: John
146: 18190. Wiley and Sons.
Fialko, Y. A. and Rubin, A. M., 1999, What controls the Fung, Y. C., 1965, Foundations of Solid Mechanics, Prentice-
along-strike slopes of volcanic soft zones? Journal of Hall International Series in Dynamics. Englewood Cliffs,
Geophysical Research 104: 20 00720. NJ: Prentice-Hall.
Fialko, Y. A., Sandwell, D. T., Agrew, D., Simons, M., Fung, Y. C., 1969, A First Course in Continuum Mechanics.
Shearer, P., and Minster, B., 2002, Deformation on Englewood Cliffs, NJ: Prentice-Hall.
nearby faults induced by the 1999 Hector Mine Gallagher, J. J., Friedman, M., Handin, J., and Sowers,
earthquake. Science 297: 185862. G. M., 1974, Experimental studies relating to
Fink, J. H. and Fletcher, R. C., 1981, A Mechanical Analysis microfracture in sandstone. Tectonophysics 21:
of Extensional Instability on Ganymede, Technical 20347.
Memorandum 84211. Washington, DC: National Gauss, K. F., 1827, Karl Friedrich Gauss, general investi-
Aeronautics and Space Administration, pp. 513. gations of curved surfaces. In J. C. Morehead, ed.,
Fisher, M. P. and Wiklerson, M. S., 2000, Predicting the Investigations of Curved Surfaces of 1827 and 1825 (pub-
orientation of joints from fold shape: results of lished in 1902). Princeton, NJ: Princeton University
pseudo-three-dimensional modeling and curvature Library.
analysis. Geology 28: 1518. Ge, H. and Jackson, M. P. A., 1998, Physical modeling
Fisher, R. A., 1953, Dispersion of a sphere. Proceedings of of structures formed by salt withdrawal: implica-
the Royal Society of London Series A 217: 295. tions for deformation caused by slat dissolution.
Fletcher, R. C., 1974, Wavelength selection in the American Association of Petroleum Geologists Bulletin
folding of a single layer with power-law rheology. 82: 22850.
American Journal of Science 274: 102943. Gere, J. M. and Weaver, W. J., 1965, Matrix Algebra for
Fletcher, R. C., 1982, Coupling of diffusional mass Engineers. New York: Van Nostrand Reinhold.
transport and deformation in a tight rock. Gilbert, G. K., 1877, Report on the Geology of the Henry
Tectonophysics 83: 27591. Mountains, US Geographical and Geological Survey of the
Fletcher, R. C. and Hallet, B., 1983, Unstable extension Rocky Mountain Region. Washington, DC: US
of lithosphere: a mechanical model for basin-and- Government Printing Office.
range structure. Journal of Geophysical Research 83: Gleick, J., 1987, Chaos: Making a New Science. New York:
27591. Viking Penguin.
Fletcher, R. C. and Pollard, D. D., 1981, Anticrack model Goodman, R. E., 1980, Introduction to Rock Mechanics.
for pressure solution surfaces. Geology 9: 41924. New York: John Wiley and Sons.
Fletcher, R. C. and Pollard, D. D., 1999, Can we under- Goodman, R. E. and Shi, G.-H., 1985, Block Theory and Its
stand structural and tectonic processes and their Application to Rock Engineering. Englewood Cliffs, NJ:
products without a complete mechanics? Journal of Prentice-Hall.
Structural Geology 21: 107188. Gordon, J. E., 1976, The New Science of Strong Materials.
Fleuty, M. J., 1964, The description of folds. Proceedings Princeton, NJ: Princeton University Press.
of the Geologists Association 75(4): 46192. Grachev, A. F., Magnitsky, V. A., Mukhamediev, S. A., and
484 REFERENCES
Nikolaev, V. A., 2001, The effect of neotectonic move- extensional diapirs overprinted by compression.
ments on the gradients and curvatures of the north- American Association of Petroleum Geologists Bulletin 84:
ern Eurasian lithosphere surface. IzvestiyaRussian 1095108.
Academy of Sciences, Physics of the Solid Earth, 37: Guiton, M. L. E., Sassi, W., Leroy, Y. M., and Gauthier,
89106. B. D. M., 2003, Mechanical constraints on the
Green, H. W. and Burnley, P. C., 1989, The mechanism chronology of fracture activation in folded Devonian
of failure responsible for deep-focus earthquakes. sandstone of the western Moroccan Anti-Atlas.
Nature 341: 7337. Journal of Structural Geology 25: 131730.
Gregory, B., 1990, Inventing Reality: Physics as Language. Gutscher, M., Kukowski, N., Malavielle, J., and
New York: John Wiley and Sons. Lallemand, S., 1998, Material transfer in accretion-
Griffith, A. A., 1921, The phenomena of rupture and ary wedges from analysis of a systematic series of
flow in solids. Philosophical Transactions of the Royal analog experiments. Journal of Structural Geology 20:
Society of London Series A 221: 16398. 40716.
Griffith, A. A., 1924, The theory of rupture. In C. B. Habert, G. and De Saint-Blanquat, M., 2004, Rate of
Biezeno and J. M. Burgers, eds., First International construction of the Black Mesa bysmalith, Henry
Congress on Applied Mechanics. Delft: J. Waltman, Mountains, Utah. In C. Breitkreuz and N. Petford,
pp. 5563. eds., Physical Geology of High-Level Magmatic Systems.
Griggs, D. T., 1939, Creep of rocks. Journal of Geology 47: London: Geological Society Special Publications 234,
22551. pp. 163174.
Griggs, D. T. and Handin, J., 1960a, Observations on Hafner, W., 1951, Stress distributions and faulting.
fracture and an hypothesis of earthquakes. In D.T. Geological Society of America Bulletin 62: 37398.
Griggs and J. Handin, eds., Rock Deformation, Memoir Hhner, G. and Spencer, N., 1998, Rubbing and scrub-
79. New York: Geological Society of America, bing. Physics Today 51(9): 227.
pp. 34764. Haimson, B. C., 1968, Hydraulic fracturing in porous
Griggs, D. T. and Handin, J., 1960b, Rock Deformation, and nonporous rock and its potential for determin-
Memoir 79. New York: Geological Society of America, ing in situ stresses at great depth. Unpublished Ph.D.
p. 382. thesis, University of Wisconsin.
Grosfils, E. B. and Head, J. W., 1994, The global distribu- Haimson, B. C. and Fairhurst, C., 1967, Initiation and
tion of giant radiating dike swarms on Venus: impli- extension of hydraulic fractures in rocks. Society of
cations for the global stress state. Geophysical Research Petroleum Engineers Journal September: 31018.
Letters 21: 7014. Handin, J. and Hager, R. V., 1957, Experimental deforma-
Groshong, R. H. Jr., 1975, Strain, fractures, and pressure tion of sedimentary rocks under confining pressure:
solution in natural single layer folds. Geological tests at room temperature on dry samples. American
Society of America Bulletin 86: 136376. Association of Petroleum Geologists Bulletin 41: 150.
Groshong, R. H. Jr., 1999, 3-D Structural Geology: A Handin, J. and Hager, R. V., 1958, Experimental defor-
Practical Guide to Surface and Subsurface Map mation of sedimentary rocks under confining pres-
Interpretation. Berlin: Springer-Verlag. sure: tests at high temperature. American Association
Gross, M. R., 1993, The origin and spacing of cross of Petroleum Geologists Bulletin 42: 2892934.
joints: examples from Monterey Formation Santa Handin, J., Hager, R. V., Friedman, M., and Feather, J. N.,
Barbara Coastline, California. Journal of Structural 1963, Experimental deformation of sedimentary
Geology 15: 77351. rocks under confining pressure: pore pressure tests.
Gross, M. R., Fischer, M. P., Engelder, T., and Greenfield, American Association of Petroleum Geologists Bulletin 47:
R. J., 1995, Factors controlling joint spacing in 71755.
interbedded sedimentary rocks: integrating numeri- Handy, M. R., 1994, Flow laws for rocks containing
cal models with field observations from the two non-linear viscous phases: a phenomenological
Monterey Formation, USA. In M. S. Ameen, ed., approach. Journal of Structural Geology 16: 287301.
Fractography: Fracture Topography as a Tool in Fracture Hanks, T. C., 1977, Earthquake stress drops, ambient
Mechanics and Stress Analysis. London: Geological tectonic stresses and stresses that drive plate
Society, pp. 21533. motions. Pure and Applied Geophysics 115: 44158.
Grout, F. F., 1945, Scale models of structures related to Happel, J. and Brenner, H., 1965, Low Reynolds Number
batholiths. American Journal of Science 111: 26084. Hydrodynamics: With Special Applications to Particulate
Guglielmo, G., Vendeville, B. C., and Jackson, M. P. A., Media. Englewood Cliffs, NJ: Prentice-Hall.
2000, 3-D visualization and isochore analysis of Harris, R. A. and Segall, P., 1987, Detection of a locked
REFERENCES 485
zone at depth on the Parkfield, California, segment Pretoria: Natural Mechanical Engineering Research
of the San Andreas fault. Journal of Geophysical Institute, CSIR.
Research 92(B8): 794562. Hoek, E. and Bieniawski, Z. T., 1965, Brittle fracture
Harris, R. A. and Simpson, R. W., 1998, Suppression of propagation in rock under compression. International
large earthquakes by stress shadows: a comparison Journal of Fracture Mechanics 1: 13755.
of Coulomb and rate-and-state failure. Journal of Hofmann-Wellenhof, B., Lichtenegger, H., and Collins,
Geophysical Research 103: 24 43951. J., 1997, Global Positioning System. New York: Springer-
Haskell, N. H., 1937, The viscosity of the aestheno- Verlag.
sphere. American Journal of Science 33: 228. Hoshino, K., Koide, H., Inami, K., Iwamura, S., and
Hausman, D. M., 1998, Causal Asymmetries. Cambridge: Mitsui, S., 1972, Mechanical Properties of Japanese Tertiary
Cambridge University Press. Sedimentary Rocks Under High Confining Pressures, Report
Head, J. W., Crumpler, L. S., Aubele, J. C., Guest, J. E., and 244. Kawasaki: Geological Survey of Japan.
Saunders, R. S., 1992, Venus volcanism: classification Hossain, K. M., 1979, Determination of strain from
of volcanic features and structures, associations, and stretched belemnites. Tectonophysics 60: 27988.
global distributions from Magellan data. Journal of Howarth, R. J., 1999, Measurement, portrayal and
Geophysical Research 97: 13 15397. analysis of orientation data and the origins of early
Heard, H. C., 1960, Transition from brittle fracture to modern structural geology (16701967). Proceedings of
ductile flow in Solenhofen limestone as a function the Geologists Association 110: 273309.
of temperature, confining pressure, and interstitial Hubbert, M. K., 1937, Theory of scale models as applied
fluid pressure. In D. T. Griggs and J. Handin, eds., to the study of geologic structures. Geological Society
Rock Deformation. Boulder, CO: Geological Society of of America Bulletin 48: 1459520.
America, pp. 193226. Hubbert, M. K., 1945, Strength of the Earth. American
Hearmon, R. F. S., 1961, An Introduction to Applied Association of Petroleum Geologists Bulletin 29: 163053.
Anisotropic Elasticity. Oxford: Oxford University Press. Hubbert, M. K., 1951, Mechanical basis for certain
Herget, G., 1993, Rock stresses and rock stress monitor- familiar geologic structures. Geological Society of
ing in Canada. In J. A. Hudson, ed., Rock Testing and America Bulletin 62: 35572.
Site Characterization: Comprehensive Rock Engineering. Hubbert, M. K., 1972, Structural Geology. New York:
Oxford: Pergamon Press, pp. 47396. Hafner.
Hill, R., 1965, A self-consistent mechanics of composite Hubbert, M. K., 1974, Is being quantitative sufficient?
materials. Journal of the Mechanics and Physics of Solids In D. F. Merriam, ed., The Impact of Quantification on
13: 21322. Geology. Syracuse, NY: Syracuse University, pp. 2749.
Hillerman, T., 1990, Coyote Waits. New York: Harper and Hubbert, M. K. and Willis, D. G., 1957, Mechanics of
Row. hydraulic fracturing. Petroleum, Transactions AIME 210:
Hilley, G. E. and Strecker, M. R., 2004, Steady-state 15366.
erosion of critical Coulomb wedges with applica- Hudleston, P. J., 1973, Fold morphology and some geo-
tions to Taiwan and the Himalaya. Journal of metrical implications of theories of fold develop-
Geophysical Research 109: doi:10.1029/2002JB002284. ment. Tectonophysics 16(12): 146.
Hilley, G. E., Strecker, M. R., and Ramos, V. A., 2004, Hudleston, P. J., 1992, A comparison between glacial
Growth and erosion of fold-and-thrust belts, with an movement and thrust sheet or nappe emplacement
application to the Aconcagua Fold-and-Thrust Belt, and associated structures. In S. Mitra and G. W.
Argentina. Journal of Geophysical Research 109: Fisher, eds., Structural Geology of Fold and Thrust Belts.
doi:10.1029/2002JB002282. Baltimore, MD: The Johns Hopkins University Press,
Hirth, J. P. and Lothe, J., 1982, Theory of Dislocations. New pp. 8191.
York: John Wiley and Sons. Hudson, J. A., Crouch, S. L., and Fairhurst, C., 1972,
Hoagland, R. G., Hahn, G. T., and Rosenfield, A. R., 1973, Soft, stiff and servo-controlled testing machines: a
Influence of microstructure on fracture propagation review with reference to rock failure. Engineering
in rock. Rock Mechanics 5: 77106. Geology 6: 1559.
Hobbs, D. W., 1962, The strength of coal under biaxial Hughes, T. J. R., 1987, The Finite Element Method.
compression. Colliery Engineering 39: 28590. Englewood Cliffs, NJ: Prentice-Hall.
Hodgson, R. A., 1961, Classification of structures on Hunt, C. B., 1953, Geology and Geography of the Henry
joint surfaces. American Journal of Science 259: Mountains Region, Utah, US Geological Survey Professional
493502. Paper. Washington, DC: US Government Printing
Hoek, E., 1965, Rock Fracture Under Static Stress Conditions. Office.
486 REFERENCES
Hunt, C. B., 1988a, Geology of the Henry Mountains, Utah, failure in rings of rock loaded in diametral tension
as Recorded in the Notebooks of G. K. Gilbert, 187576. or compression. British Journal of Applied Physics 17:
Boulder, CO: The Geological Society of America. 68592.
Hunt, C. B., 1988b, The laccolith-stock controversy: new Johnson, A. M., 1970, Physical Processes in Geology. San
results from the southern Henry Mountains, Utah. Francisco, CA: Freeman, Cooper and Co.
Discussion and reply. Geological Society of America Johnson, A. M. and Fletcher, R. C., 1994, Folding of
Bulletin 100: 16579. Viscous Layers. New York: Columbia University Press.
Huntoon, P. W., 1982, The Meander anticline, Johnson, A. M. and Pollard, D. D., 1973, Mechanics of
Canyonlands, Utah: an unloading structure result- growth of some laccolithic intrusions in the Henry
ing from horizontal gliding on salt. Geological Society Mtns, Utah: Part I, Field observations, Gilberts
of America Bulletin 93: 94150. model, physical properties and flow of magma.
Hurlbut, C. S. and Griggs, D. T., 1939, Igneous rocks of Tectonophysics 18: 261309.
the Highwood Mountains, Montana. Geological Society Johnson, C. E., Rohahn, C., and Sharp, R. V., 1982,
of America Bulletin 50: 1032112. Introduction, The Imperial Valley, California,
Hytch, M. J., Putaux, J.-L., and Penisson, J.-M., 2003, Earthquake of October 15, 1979, Professional Paper
Measurement of the displacement field of disloca- 1254. Washington, DC: United States Geological
tions to 0.03 by electron microscopy. Nature 423(15 Survey, pp. 13.
May): 2703. Johnson, K. M. and Johnson, A. M., 2000, Localization of
Inglis, C. E., 1913, Stresses in a plate due to the pres- layer-parallel faults in San Rafael swell, Utah, and
ence of cracks and sharp corners. Royal Institute of other monoclinal folds. Journal of Structural Geology
Naval Architects Transactions 55: 21930. 22: 145568.
Irvine, T. N., 1970, Heat transfer during solidification Johnson, R. B., 1961, Patterns and origin of radial dike
of layered intrusions: I. Sheets and sills. Canadian swarms associated with West Spanish Peak and Dike
Journal of Earth Science 7: 103161. Mountain, south-central Colorado. Geological Society
Irwin, G. R., 1957, Analysis of stresses and strains near of America Bulletin 72: 57990.
the end of a crack traversing a plate. Journal of Johnson, R. B., 1968, Geology of the Igneous Rocks of the
Applied Mechanics 24: 3614. Spanish Peaks Region, Colorado, Professional Paper
Irwin, G. R., 1958, Fracture. In S. Flugge, ed., 594G. Menlo, CA: US Geological Survey, pp. 147.
Encyclopedia of Physics. Berlin: Springer-Verlag, Jones, G., Fisher, Q. J., and Knipe, R. J., 1998, Faulting,
pp. 55190. Fault Sealing and Fluid Flow in Hydrocarbon Reservoirs,
Jackson, M. D. and Pollard, D. D., 1988, The laccolith- Special Publication 147. London: Geological Society,
stock controversy: new results from the southern p. 319.
Henry Mountains, Utah. Geological Society of America Jonsson, S., Zebker, H. A., Segall, P., and Amelung, F.,
Bulletin 100: 11739. 2002, Fault slip distribution of the 1999 M-w 7.1
Jackson, M. D. and Pollard, D. D., 1990, Flexure and Hector Mine, California, earthquake, estimated from
faulting of sedimentary host rocks during growth of satellite radar and GPS measurements. Bulletin of the
igneous domes, Henry Mountains, Utah. Journal of Seismological Society of America 92: 137789.
Structural Geology 12: 185206. Kamb, W. B., 1959a, Theory of preferred crystal orienta-
Jaeger, J. C., 1957, The temperature in the neighbor- tion developed by crystallization under stress.
hood of a cooling intrusive sheet. American Journal of Journal of Geology 67: 15370.
Science 255: 30618. Kamb, W. B., 1959b, Ice petrofabric observations from
Jaeger, J. C., 1964a, Elasticity, Fracture and Flow with Blue Glacier, Washington, in relation to theory and
Engineering and Geological Applications. London: experiment. Journal of Geophysical Research 64:
Methuen. 1891909.
Jaeger, J. C., 1964b, Temperatures outside a cooling Kanninen, M. F. and Popelar, C. H., 1985, Advanced
intrusive sheet. American Journal of Science 257: 4454. Fracture Mechanics. New York: Oxford University Press.
Jaeger, J. C. and Cook, N. G. W., 1979, Fundamentals of Kattenhorn, S. A., Aydin, A., and Pollard, D. D., 2000,
Rock Mechanics. London: Chapman and Hall. Joints at high angles to normal fault strike: an expla-
Jaeger, J. C. and Hoskins, E. R., 1966a, Rock failure nation using 3-D numerical models of fault-per-
under the confined Brazilian test. Journal of turbed stress fields. Journal of Structural Geology 22(1):
Geophysical Research 71: 26519. 123.
Jaeger, J. C. and Hoskins, E. R., 1966b, Stresses and Kattenhorn, S. A. and Pollard, D. D., 2001, Integrating
REFERENCES 487
3-D seismic data, field analogs, and mechanical for oceanic spreading centers. Journal of Geophysical
models in the analysis of segmented normal faults Research 78: 3395417.
in the Wytch Farm oil field, southern England, Lamb, H., 1945, Hydrodynamics. New York: Dover.
United Kingdom. American Association of Petroleum Landau, L. D. and Lifshitz, E. M., 1960, Fluid Mechanics.
Geologists Bulletin 85(7): 1183210. Reading: Addison-Wesley.
Kelly, V. C., 1955, Monoclines of the Colorado Plateau. Lawn, B. R. and Wilshaw, T. R., 1975, Fracture of Brittle
Geological Society of America Bulletin 66: 789804. Solids. Cambridge: Cambridge University Press.
Kerr, A. D. and Pollard, D. D., 1998, Toward more realis- Lawson, A. C., 1908, The California Earthquake of April, 18,
tic formulations for the analysis of laccoliths. Journal 1906, Report of the State Earthquake Investigation
of Structural Geology 20: 178393. Committee. Washington, DC: Carnegie Institute of
Kilgore, B. D., Blanpied, M. L., and Dieterich, J. H., 1993, Washington, p. 641.
Velocity dependent friction of granite over a wide Lee, T.-C. and Delaney, P. T., 1987, Frictional heating
range of conditions. Geophysical Research Letters 20(10): and pore pressure rise due to fault slip. Geophysical
9036. Journal of the Royal Astronomical Society 88: 56991.
Kind, D. and Quinn, T., 1999, Metrology: quo vadis? Lekhnitskii, S. G., 1963, Theory of Elasticity of an
Physics Today (Sixteenth Annual Buyers Guide) 52: Anisotropic Elastic Body. San Francisco: Holden-Day.
BG13BG15. Li, V. C., 1987, Mechanics of shear rupture applied to
King, G. C. P., Stein, R. S., and Lin, J., 1994, Static stress earthquake zones. In B. K. Atkinson, ed., Fracture
changes and the triggering of earthquakes. Bulletin of Mechanics of Rock. San Diego, CA: Academic Press.
the Seismological Society of America 84: 93553. Lide, D. R., 2004, CRC Handbook of Chemistry and Physics.
King, W., 1875, Report on the superinduced divisional Boca Raton, FL: Chemical Rubber Company Press, Inc.
structure of rocks, called jointing, and its relation to Linker, M. F. and Dieterich, J. H., 1992, Effects of vari-
slaty cleavage. Royal Irish Academy Transactions 25: able normal stress on rock friction: observations and
60562. constitutive equations. Journal of Geophysical Research,
Kirby, S. H. and Kronenberg, A. K., 1987, Rheology of the B, Solid Earth and Planets 97(4): 492340.
lithosphere: selected topics. Reviews of Geophysics 25: Lipschutz, M. M., 1969, Theory and Problems of Differential
121944. Geometry, Schaums Outline. New York: McGraw-Hill.
Kirsch, G., 1898, Die Theorie der Elastizitt und die Lisle, R. J., 1994, Detection of zones of abnormal strains
Bedrfnisse der Festigkeitslehre. Zeitschrift des Verlines in structures using Gaussian curvature analysis.
Deutscher Ingenieure 42: 797807. American Association of Petroleum Geologists Bulletin 78:
Koch, F. G., Johnson, A. M., and Pollard, D. D., 1981, 181119.
Monoclinal bending of strata over laccolithic intru- Lisle, R. J., 2000, Predicting patterns of strain from
sions. Tectonophysics 74: T21T31. three-dimensional fold geometries: neutral surface
Koenig, E. and Pollard, D. D., 1998, Mapping and mod- folds and forced folds. In J. W. Cosgrove and M. S.
eling of radial fracture patterns on Venus. Journal of Ameen, eds., Forced Folds and Fractures, Special
Geophysical Research 103: 15 183202. Publication 169. London: Geological Society
Kostrov, B. V. and Das, S., 1988, Principles of Earthquake pp. 21321.
Source Mechanics. Cambridge: Cambridge University Lister, J. R., 1990, Buoyancy-driven fluid fracture.
Press. Journal of Fluid Mechanics 210: 26380.
Krantz, W. R., 1988, Multiple fault sets and three- Lister, J. R. and Kerr, R. C., 1991, Fluid-mechanical
dimensional strain: theory and application. Journal models of crack propagation and their application
of Structural Geology 10: 22537. to magma transport in dykes. Journal of Geophysical
Kranz, R. L., 1983, Microcracks in rocks: a review. Research 96: 10 04977.
Tectonophysics 100: 44980. Liu, S. and Dixon, J. M., 1990, Centrifuge modelling of
Krumbein, W. C. and Graybill, F. A., 1965, An thrust faulting: strain partitioning and the sequence
Introduction to Statistical Models in Geology. New York: of thrusting in duplex structures. In R. J. Knipe, ed.,
McGraw-Hill. Deformation Mechanisms, Rheology and Tectonics, Special
Labuz, J. F., Shah, S. P., and Dowding, C. H., 1987, The Publication 54. London: Geological Society,
fracture process zone in granite: evidence and effect. pp. 43144.
International Journal of Rock Mechanics and Mining Liu, S. and Dixon, J. M., 1991, Centrifuge modelling of
Science 24: 23546. thrust faulting: structural variation along strike in
Lachenbruch, A. H., 1973, A simple mechanical model fold-thrust belts. Tectonophysics 188: 3962.
488 REFERENCES
Lockner, D. A. and Beeler, N. M., 2003, Rock failure and Marshak, S. and Mitra, G., 1988, Basic Methods of
earthquakes. In W. H. K. Lee, H. Kanamori, P. C. Structural Geology. Englewood Cliffs, NJ: Prentice-Hall.
Jennings, and C. Kisslinger, eds., International Martel, S. and Boger, W. A., 1998, Geometry and
Handbook of Earthquake and Engineering Seismology, mechanics of secondary fracturing around small
Part A. Amsterdam: Academic Press, pp. 50538. three-dimensional faults in granitic rock. Journal of
Love, A. E. H., 1944, A Treatise on the Mathematical Geophysical Research 103: 21 299314.
Theory of Elasticity. New York: Dover. Martel, S., Pollard, D. D., and Segall, P., 1988,
Lovering, T. S., 1935, Theory of heat conduction applied Development of simple strike-slip fault zones in
to geological problems. Geological Society of America granitic rock, Mount Abbot quadrangle, Sierra
Bulletin 46: 6994. Nevada, California. Geological Society of America
Lovering, T. S., 1936, Heat conduction in dissimilar Bulletin 99: 145165.
rocks and the use of thermal models. Geological Martel, S. J., 1990, Formation of compound strike-slip
Society of America Bulletin 47: 87100. fault zones, Mount Abbot Quadrangle, California.
Maerten, F., Resor, P., Maerten, L., and Pollard, D. D., Journal of Structural Geology 12(7): 86982.
2005, Inverting for slip on three-dimensional fault Martel, S. J. and Pollard, D. D., 1989, Mechanisms of slip
surfaces using angular dislocations. Bulletin of the and fracture along small faults and simple strike-slip
Seismological Society of America in press. fault zones in granitic rock. Journal of Geophysical
Maerten, L., 2000, Variation in slip on intersecting Research 94: 941728.
normal faults: implications for paleostress inversion. Massonnet, D., Rossi, M., Carmona, C., et al., 1993, The
Journal of Geophysical Research 105: 25 55365. displacement field of the Landers earthquake
Maerten, L., Gillespie, P., and Pollard, D. D., 2002, mapped by radar interferometry. Nature 364: 13842.
Effects of local stress perturbation on secondary McBirney, A. R. and Murase, T., 1984, Rheological prop-
fault development. Journal of Structural Geology 24: erties of magmas. Annual Reviews of Earth and
14553. Planetary Sciences 12: 33757.
Maerten, L., Pollard, D. D., and Karpuz, R., 2000, How to McClay, K. and Bonora, M., 2001, Analog models of
constrain 3D fault continuity and linkage using restraining stepovers in strike-slip fault systems.
reflection seismic data: a geomechanical approach. American Association of Petroleum Geologists Bulletin 85:
American Association of Petroleum Geologists Bulletin 84: 23360.
131124. McClay, K. R., Waltham, D. A., Scott, A. D., and
Maerten, L., Pollard, D. D., and Maerten, F., 2001, Abousetta, A. A., 1991, Physical and seismic model-
Digital mapping of three-dimensional structures of ing of listric fault geometries. In A. M. Roberts, G.
the Chimney Rock fault system, central Utah. Journal Yielding, and B. Freeman, eds., The Geometry of Normal
of Structural Geology 23: 58592. Faults, Special Publication 56. London: Geological
Maerten, L., Willemse, E. J. M., Pollard, D. D., and Society, pp. 2319.
Rawnsley, K., 1999, Slip distributions on intersecting McClay, K. R. and White, M., 1995, Analogue models of
normal faults. Journal of Strurctural Geology 21: 25971. orthogonal and oblique rifting. Marine and Petroleum
Mallet, J. L., 2002, Geomodeling. Oxford: Oxford Geology 12: 13751.
University Press. McClintock, F. A. and Walsh, J. B., 1962, Friction on
Malvern, L. E., 1969, Introduction to the Mechanics of a Griffith cracks in rocks under pressure. In Fourth US
Continuous Medium. Upper Saddle River, NJ: Prentice- Congressional Applied Mechanics, Proceedings. Berkeley,
Hall. CA: American Society of Mechanical Engineers,
Manduit, T. and Brun, J. P., 1998, Growth fault/rollover pp. 101521.
systems: birth, growth, and decay. Journal of McConaughy, D. T. and Engelder, T., 2001, Joint initia-
Geophysical Research 103: 18 11936. tion in bedded clastic rocks. Journal of Structural
Mardon, D., 1988, Localization of pressure solution Geology 23(23): 20321.
and the formation of discrete solution seams. McGarr, A. and Gay, N. C., 1978, State of stress in the
Unpublished Ph.D. thesis, Texas A&M University. Earths crust. Annual Reviews of Earth and Planetary
Margenau, H., 1977, The Nature of Physical Reality: A Sciences 6: 40536.
Philosophy of Modern Physics. Woodbridge, CT: Ox Bow McGill, G. E. and Stromquist, A. W., 1979, The grabens
Press. of Canyonlands National Park, Utah: Geometry,
Marrett, R. and Peacock, D. C. P., 1999, Strain and stress. mechanics, and kinematics. Journal of Geophysical
Journal of Structural Geology 21: 105763. Research 84: 454763.
REFERENCES 489
McIntyre, D. B. and Weiss, L. E., 1956, Construction of the North Anatolian Fault with static stress changes.
block diagrams to scale in orthographic projection. Geophysical Journal International 154(2): 47182.
Proceedings of the Geological Association 67: 14255. Muller, O. H. and Pollard, D. D., 1977, The stress state
McKenzie, D. P., 1969, The relation between fault plane near Spanish Peaks, Colorado, determined from a
solutions for earthquakes and the directions of the dike pattern. Pure and Applied Geophysics 115: 6985.
principal stresses. Bulletin of the Seismological Society of Murase, T. and McBirney, A. R., 1973, Properties of
America 50: 595601. some common igneous rocks and their melts at high
Means, W. D., 1976, Stress and Strain. New York: temperatures. Geological Society of America Bulletin 84:
Springer-Verlag. 356392.
Mechtly, E. A., 1973, The International System of Units: Murray, J. R. and Segall, P., 2002, Testing the time-pre-
Physical Constants and Conversion Factors, Special dictable earthquake recurrence model by direct
Publication SP-7012. Washington, DC: National measurement of strain accumulation and release.
Aeronautics and Space Administration. Nature 28: 28791.
Meier, M. F., Kamb, B., Allen, C. R., and Sharp, R. P., Murray, J. R., Segall, P., Cervelli, P., Prescott, W., and
1974, Flow of Blue Glacier, Olympic Mountains, Svarc, J., 2001, Inversion of GPS data for spatially
Washington, USA. Journal of Glaciology 13: 187212. variable slip-rate on the San Andreas fault near
Melosh, H. J., 1989, Impact Cratering: A Geologic Process. Parkfield, CA. Geophysical Research Letters 28: 35962.
New York: Oxford University Press. Muskhelishvili, N. I., 1954, Some Basic Problems of the
Meriaux, C. and Lister, J. R., 2002, Calculation of dike Mathematical Theory of Elasticity, tr. J. R. M. Radok,
trajectories from volcanic centers. Journal of Leyden: Noordhoff.
Geophysical Research 107(B4): doi: 10.1029/2001JB000 Nakamura, K., 1977, Volcanoes as possible indicators of
436. tectonic stress orientation: principal and proposal.
Meyers, R. D. and Aydin, A., 2004, The evolution of Journal of Volcanology and Geothermal Research 2: 116.
faults formed by shearing across joint zones in sand- Nakamura, K., Jacob, K. H., and Davies, J. N., 1977,
stone. Journal of Structural Geology 26(5): 94766. Volcanoes as possible indicators of tectonic stress
Michell, J. H., 1899, On the direct determination of orientation: Aleutians and Alaska. Pure and Applied
stress in an elastic solid, with application to the Geophysics 115: 87112.
theory of plates. Proceedings of the London Mathematical Narr, W. and Suppe, J., 1991, Joint spacing in sedimen-
Society 31: 10024. tary rocks. Journal of Structural Geology 13(9): 103748.
Mogi, K., 1958, Relations between the eruptions of Nelson, R. A., 2003, Guide for metric practice. Physics
various volcanoes and the deformations of the Today 56(8): BG15BG16.
ground surfaces around them. Bulletin of Earthquake Nemat-Nasser, S. and Horii, H., 1982, Compression-
Research Institute, University of Tokyo 36: 99134. induced nonplanar crack extension with application
Mohr, O., 1882, ber die Darstellung des to splitting, exfoliation, and rockburst. Journal of
Spannungszustandes eines Krperelements. Geophysical Research 87: 680521.
Civilingenieure 28: 11356. Netz, R., 2000, The origins of mathematical physics:
Mohr, P. J. and Taylor, B. N., 2003, The fundamental new light on an old question. Physics Today June: 327.
physical constants. Physics Today 56(8): BG6BG13. Neurath, C. and Smith, R. B., 1982, The effect of ma-
Mollema, P. N. and Antonellini, M. A., 1996, terial properties on growth rates of folding and
Compaction bands: a structural analog for anti- boudinage: experiments with wax models. Journal of
mode I cracks in aeolian sandstone. Tectonophysics Structural Geology 4: 21529.
267: 20928. Newton, I., 1999, The Principia. Berkeley, CA: University
Moore, J. M. and Schultz, R. A., 1999, Processes of fault- of California Press. (Originally published 1687,
ing in jointed rocks of Canyonlands National Park, Cambridge: Cambridge University Press.)
Utah. Geological Society of America Bulletin 111: Nicholson, R. and Ejiofor, I. B., 1987, The three-
80822. dimensional morphology of arrays of echelon and
Morris, A. P., Ferrill, D. A., and Henderson, D. B., 1996, sigmoidal, mineral-filled fractures: data from north
Slip tendency analysis and fault reactivation. Geology Cornwall. Journal of the Geological Society, London 144:
24: 2758. 7983.
Muller, J. R., Aydin, A., and Maerten, F., 2003, Nur, A. and Byerlee, J. D., 1971, An exact effective stress
Investigating the transition between the 1967 law for elastic deformation of rock with fluids.
Mudurnu Valley and 1999 Izmit earthquakes along Journal of Geophysical Research 76: 641419.
490 REFERENCES
Nur, A., Ron, H., and Scotti, O., 1986, Fault mechanics Topographic wavelengths of Ganymede groove lanes
and kinematics of block rotation. Geology 14: 7469. from Fourier analysis of Galileo images. Journal of
Nye, J. F., 1953, The flow law of ice from measurements Geophysical Research, E, Planets 104: 24 05774.
made in glacier tunnels, laboratory experiments, Paterson, M. S., 1978, Experimental Rock Deformation: the
and the Jungfrau borehole experiment. Proceedings of Brittle Field. New York: Springer-Verlag.
the Royal Society of London, Series A 219: 47789. Patton, T. L., Logan, J. M., and Friedman, M., 1998,
Nye, J. F., 1985, Physical Properties of Crystals. Oxford: Experimentally generated normal faults in single-
Clarendon Press. layer and multilayer limestone specimens at confin-
Obert, L. and Duvall, W. I., 1967, Rock Mechanics and the ing pressure. Tectonophysics 295: 5377.
Design of Structures in Rock. New York: John Wiley and Paul, B., 1960, Prediction of elastic constants of multi-
Sons. phase materials. Transactions of the Metallurgical
OConnor, J. J. and Robertson, E. F., 2004, The MacTutor Society of AIME 218: 3641.
History of Mathematics Archive. St. Andrews, Fife: Paul, B., 1961, Modification of the CoulombMohr
School of Mathematics and Statistics, University of theory of fracture. Journal of Applied Mechanics 28:
St. Andrews. 25968.
Od, H., 1957, Mechanical analysis of the dike pattern Peacock, D. C. P., 1991, Displacement and segment
of the Spanish Peaks area, Colorado. Geological Society linkage strike-slip fault zones. Journal of Structural
of America Bulletin 68: 56776. Geology 13: 102535.
Okada, Y., 1985, Surface deformation due to shear and Peacock, D. C. P. and Marrett, R., 2000, Strain and stress:
tensile faults in a half-space. Bulletin of the reply. Journal of Structural Geology 22: 136978.
Seismological Society of America 75: 113554. Peacock, D. C. P. and Sanderson, D. J., 1991,
Olson, J. and Pollard, D. D., 1989, Inferring palaeo- Displacements, segment linkage and relay ramps in
stresses from natural fracture patterns: a new normal fault zones. Journal of Structural Geology 13:
method. Geology 17: 3458. 72133.
Olson, J. E., 1993, Joint pattern development: effects of Peacock, D. C. P. and Sanderson, D. J., 1994, Geometry
subcritical crack growth and mechanical crack inter- and development of relay ramps in normal fault
action. Journal of Geophysical Research 98(B7): systems. American Association of Petroleum Geologists
12 25165. Bulletin 78: 14765.
Olson, J. E. and Pollard, D. D., 1991, The initiation and Pearl, J., 2000, Causality: Models, Reasoning, and Inference.
growth of en chelon veins. Journal of Structural Geology Cambridge: Cambridge University Press.
13(5): 595608. Peck, L., Barton, C. C., and Gordon, R. B., 1985,
Oppenheimer, D. H., Reasenberg, P. A., and Simpson, Microstructure and the resistance of rock to tensile
R. W., 1988, Fault plane solutions for the 1984 fracture. Journal of Geophysical Research 90: 11 53346.
Morgan Hill, California, earthquake sequence: evi- Peng, S. and Johnson, A. M., 1972, Crack growth and
dence for the state of stress on the Calaveras fault. faulting in cylindrical specimens of Chelmsford
Journal of Geophysical Research 93: 900726. granite. International Journal of Rock Mechanics and
Owen, S., Segall, P., Freymueller, J., et al., 1995, Rapid Mining Science 9: 3786.
deformation of the south flank of Kilauea Volcano, Peng, S. D., 1971, Stresses within elastic circular cylin-
Hawaii. Science 267: 132832. ders loaded uniaxially and triaxially. International
Ozkaya, S. I., 2002, QUADRO: a program to estimate Journal of Rock Mechanics and Mining Science 8:
principal curvatures of folds. Computers and 339432.
Geosciences 28: 46772. Peterson, I., 1993, Newtons Clock: Chaos in the Solar
Pachell, M. A., Evans, J. P., and Taylor, W. L., 2003, System. New York: Freeman.
Kilometer-scale kinking of crystalline rocks in a Petit, J.-P. and Barquins, M., 1988, Can natural faults
transpressive convergent setting, Central Sierra propagate under mode II conditions? Tectonics 7:
Nevada, California. Geological Society of America 124356.
Bulletin 115: 81731. Petit, J.-P. and Mattauer, M., 1995, Palaeostress superim-
Pan, E., Amadei, B., and Savage, W. Z., 1995, position deduced from mesoscale structures in lime-
Gravitational and tectonic stresses in anisotropic stone: the Matelles exposure, Languedoc, France.
rock with irregular topography. International Journal Journal of Structural Geology 17(2): 24556.
of Rock Mechanics and Mining Science 32: 20114. Phillips, F. C., 1954, The Use of Stereographic Projection in
Patel, J., Pappalardo, R. T., Head, J. W., III, et al., 1999, Structural Geology. London: Edward Arnold.
REFERENCES 491
Pinto da Cunha, A., 1990, Scale Effects in Rock Masses. Prager, W., 1961, Introduction to the Mechanics of Continua.
Rotterdam: A. A. Balkema. Boston, MA: Ginn and Co.
Poirier, J.-P., 1985, Creep of Crystals: High-Temperature Prescott, W., 1993, Seeing earthquakes from afar.
Deformation Processes in Metals, Ceramics and Minerals. Nature 364: 1001.
London: Cambridge University Press. Price, E. J. and Brgmann, R., 2002, Interactions
Poliakov, A. N. B., Dmowska, R., and Rice, J. R., 2002, between the Landers and Hector Mine, California,
Dynamic shear rupture interactions with fault earthquakes from space geodesy, boundary element
bends and off-axis secondary faulting. Journal of modeling, and time-dependent friction. Bulletin of the
Geophysical Research 107(B11): no. 2295, doi:10. Seismological Society of America 92: 145069.
1029/2001JB000572. Price, E. J. and Sandwell, D. T., 1998, Small-scale defor-
Pollard, D. D., 2000, Strain and stress: discussion. mations associated with the 1992 Landers,
Journal of Structural Geology 22: 135967. California, earthquake mapped by synthetic aper-
Pollard, D. D. and Aydin, A., 1984, Propagation and ture radar interferometry phase gradients. Journal of
linkage of oceanic ridge segments. Journal of Geophysical Research 103: 27 00116.
Geophysical Research 89(B12): 10 01728. Price, N., 1966, Fault and Joint Development in Brittle and
Pollard, D. D. and Aydin, A., 1988, Progress in under- Semi-Brittle Rock. Oxford: Pergamon Press.
standing jointing over the past century. Geological Price, N. J. and Cosgrove, J. W., 1990, Analysis of
Society of America Bulletin 100: 1181204. Geological Structures. Cambridge: Cambridge
Pollard, D. D., Bergbauer, S., and Mynatt, I., 2004, Using University Press.
differential geometry to characterize and analyze Price, R. A., 1973, Large-scale gravitational flow of
the morphology of joints. In J. W. Cosgrove and T. supracrustal rocks, southern Canadian Rockies. In
Engelder, eds., The Initiation, Propagation, and Arrest of K. A. De Jong and R. Scholten, eds., Gravity and
Joints and Other Fractures, Special Publication 231. Tectonics. New York: John Wiley and Sons,
London: Geological Society, pp. 15382. pp. 491502.
Pollard, D. D., Delaney, P. T., Duffield, W. A., Endo, E. T., Pyne, S. J., 1980, Grove Karl Gilbert: A Great Engine of
and Okamura, A. T., 1983, Surface deformation in Research. Austin, TX: University of Texas Press.
volcanic rift zones. Tectonophysics 94: 54184. Ragan, D. M., 1985, Structural Geology: An Introduction to
Pollard, D. D. and Johnson, A. M., 1973, Mechanics of Geometrical Techniques. New York: John Wiley and
growth of some laccolithic intrusions in the Henry Sons.
Mountains, Utah, II. Tectonophysics 18: 31154. Ramberg, H., 1960, Relationships between length of
Pollard, D. D. and Muller, O. H., 1976, The effect of gra- arc and thickness of ptygmatically folded veins.
dients in regional stress and magma pressure on the American Journal of Science 258: 3646.
form of sheet intrusions in cross section. Journal of Ramberg, H., 1963, Fluid dynamics of viscous buckling
Geophysical Research 81: 97584. applicable to folding of layered rocks. American
Pollard, D. D., Muller, O. H., and Dockstader, D. R., 1975, Association of Petroleum Geologists Bulletin 47: 484505.
The form and growth of fingered sheet intrusions. Ramberg, H., 1967, Gravity, Deformation and the Earths
Geological Society of America Bulletin 86: 35163. Crust. London: Academic Press.
Pollard, D. D., 2000, Strain and stress: Discussion. Ramsay, J. G., 1967, Folding and Fracturing of Rocks. New
Journal of Structural Geology 22: 135967. York: McGraw-Hill.
Pollard, D. D. and Segall, P., 1987, Theoretical displace- Ramsay, J. G. and Huber, M. I., 1987, The Techniques of
ments and stresses near fractures in rock: with Modern Structural Geology, Vol. 2, Folds and Fractures.
applications to faults, joints, veins, dikes, and solu- London: Academic Press.
tion surfaces. In B. K. Atkinson, ed., Fracture Mechanics Ramsay, J. G., 1989, Emplacement kinematics of a
of Rock. London: Academic Press, pp. 277349. granite diapir: the Chindamora batholith,
Pollard, D. D., Segall, P., and Delaney, P. T., 1982, Zimbabwe. Journal of Structural Geology 11: 191210.
Formation and interpretation of dilatant echelon Ramsay, J. G. and Huber, M. I., 1983, The Techniques of
cracks. Geological Society of America Bulletin 93: Modern Structural Geology, Vol. 1, Strain Analysis.
1291303. London: Academic Press.
Potter, D. B. and McGill, G. E., 1978, Valley anticlines Ranalli, G., 1987, Rheology of the Earth: Deformation and
of the Needles District, Canyonlands National Flow Processes in Geophysics and Geodynamics. London:
Park, Utah. Geological Society of America Bulletin 89: Allen and Unwin.
95260. Rawnsley, K. D., Peacock, D. C. P., Rives, T., and
492 REFERENCES
Petit, J.-P., 1998, Joints in the Mesozoic sediments ics of fracture. In H. Liebowitz, ed., Fracture: An
around the Bristol Channel Basin. Journal of Advanced Treatise, Vol. 2. New York: Academic Press,
Structural Geology 20: 164161. pp. 191311.
Raymond, C. F., 1971, Flow in a transverse section of Rice, J. R., 1980, The mechanics of earthquake rupture.
Athabasca glacier, Alberta, Canada. Journal of In A. M. Dziewonski and E. Boschi, eds., Physics of the
Glaciology 10: 558. Earths Interior. Delft: North Holland, pp. 555649.
Reches, Z., 1978, Development of monoclines: Part I. Rice, J. R., 1992, Fault stress states, pre-pressure distrib-
Structure of the Palisades Creek branch of the East utions, and the weakness of the San Andreas fault.
Kaibab monocline, Grand Canyon, Arizona. Geological In B. Evans and T. F. Wong, eds., Fault Mechanics and
Society of America Memoir 151: 23571. Transport Properties of Rocks. New York: Academic
Reches, Z., 1983, Faulting of rocks in three-dimen- Press, pp. 43559.
sional strain fields: II. Theoretical analysis. Richardus, P. and Adler, R. K., 1972, Map Projections For
Tectonophysics 95: 13356. Geodesists, Cartographers and Geographers. Amsterdam:
Reches, Z. and Dieterich, J. H., 1983, Faulting of rocks North-Holland.
in three-dimensional strain fields: I. Failure of rocks Rispoli, R., 1981, Stress fields about strike-slip faults
in polyaxial, servo-control experiments. inferred from stylolites and tension gashes.
Tectonophysics 95: 11132. Tectonophysics 75: T29T36.
Reches, Z. and Johnson, A. M., 1978, Development of Roberts, A., 2001, Curvature attributes and their appli-
monoclines: Part II. Theoretical analysis of mono- cation to 3D interpreted horizons. First Break 19:
clines. Geological Society of America Memoir 151: 85100.
273311. Robin, P.-Y. and Cruden, A. R., 1994, Strain and vorticity
Renshaw, C. E., 2000, Fracture spatial density and the patterns in ideally ductile transpression zones.
anisotropic connectivity of fracture networks. Journal of Structural Geology 16: 44766.
Geophysical Monograph 122: 20311. Robinson, L. H., 1959, The effect of pore and confining
Renshaw, C. E. and Harvey, C. F., 1994, Propagation pressure on the failure process in sedimentary rock.
velocity of a natural hydraulic fracture in a poroelas- Quarterly of the Colorado School of Mines 54: 17799.
tic medium. Journal of Geophysical Research, B, Solid Rubin, A. M., 1990, A comparison of rift-zone tectonics
Earth and Planets 99(11): 21 65977. in Iceland and Hawaii. Bulletin of Volcanology 52(4):
Renshaw, C. E. and Park, J. C., 1997, Effect of mechani- 30219.
cal interactions on the scaling of fracture length Rubin, A. M., 1993, Tensile fracture of rock at high con-
and aperture. Nature 386(6624): 4824. fining pressure: implications for dike propagation.
Renshaw, C. E. and Pollard, D. D., 1994, Numerical sim- Journal of Geophysical Research 98: 15 91935.
ulation of fracture set formation: a fracture mechan- Rubin, A. M., 1995, Propagation of magma-filled cracks.
ics model consistent with experimental Annual Reviews of Earth and Planetary Sciences 23:
observations. Journal of Geophysical Research, B, Solid 287336.
Earth and Planets 99(5): 935972. Rubin, A. M. and Pollard, D. D., 1988, Dike-induced
Renshaw, C. E. and Pollard, D. D., 1995, An experimen- faulting in rift zones of Iceland and Afar. Geology 16:
tally verified criterion for propagation across 41317.
unbonded frictional interfaces in brittle, linear Rudnicki, J. W., 1977, The inception of faulting in a
elastic materials. International Journal of Rock rock mass with a weakened zone. Journal of
Mechanics and Mining Science 32(3): 23749. Geophysical Research 82: 84454.
Resnick, R. and Halliday, D., 1977, Physics, part 1, 3rd Rudnicki, J. W., 1979, Rotation of principal stress axes
edn. New York: John Wiley and Sons. caused by faulting. Geophysical Research Letters 6:
Reynolds, O., 1883, An experimental investigation of 1358.
the circumstances which determine whether the Rudnicki, J. W., 1980, Fracture mechanics applied to
motion of water shall be direct or sinuous, and the the Earths crust. Annual Reviews of Earth and
laws of resistance in parallel channels. Philosophical Planetary Sciences 8: 489525.
Transactions of the Royal Society of London 174: 93582. Rummel, F., 1987, Fracture mechanics approach to
Ribe, N. M. and Christensen, U. R., 1999, The dynamical hydraulic fracturing stress measurements. In B. K.
origin of Hawaiian volcanism. Earth and Planetary Atkinson, ed., Fracture Mechanics of Rock, Academic Press
Science Letters 171: 51731. Geology Series. London: Academic Press, pp. 21739.
Rice, J. R., 1968, Mathematical analysis in the mechan- Rutter, E. H., 1983, Pressure solution in nature, theory
REFERENCES 493
and experiment. Journal of the Geological Society of Shaw, H. R., 1963, ObsidianH2O viscosities at 1000 and
London 140: 72540. 2000 bars in the temperature range 700 to 900.
Ryan, M. P. and Blevins, J. Y. K., 1987, The Viscosity of Journal of Geophysical Research 68: 633743.
Synthetic and Natural Silicate Melts and Glasses at High Shaw, H. R., 1969, Rheology of basalt in the melting
Temperatures and 1 bar (105 Pascals) Pressure and at range. Journal of Petrology 10: 51035.
Higher Pressures, Bulletin 1764. Denver, CO: US Shaw, H. R., Peck, D. L., Wright, T. L., and Okamura, R.,
Geological Survey. 1968, The viscosity of basaltic magma: an analysis of
Ryan, R. J. and Smith, P. K., 1998, A review of the field measurements in Makaopuhi lave lake, Hawaii.
mesothermal gold deposits of the Meguma Group, American Journal of Science 266: 22564.
Nova Scotia, Canada. Ore Geology Reviews 13: 15383. Sheriff, R. E. and Geldart, L. P., 1995, Exploration
Sander, B., 1970, An Introduction to the Study of Fabrics of Seismology. New York: Cambridge University Press.
Geological Bodies. Oxford: Pergamon Press. Sherwin, J.-A. and Chapple, W. M., 1968, Wavelengths
Sanford, A. R., 1959, Analytical and experimental study of single layer folds: a comparison between theory
of simple geologic structures. Geological Society of and observation. American Journal of Science 266:
America Bulletin 70: 1952. 16779.
Savage, J. C., Prescott, W. H., Lisowski, M., and King, N., Shipton, Z. K. and Cowie, P. A., 2001, Damage zone and
1979, Deformation across the Salton trough, slip-surface evolution over micron to km scales in
California, 19731977. Journal of Geophysical Research high-porosity Navajo sandstone, Utah. Journal of
84: 306979. Structural Geology 23: 182544.
Savage, W. Z. and Swolfs, H. S., 1986, Tectonic and gravi- Sibson, R. H., 1986, Earthquakes and rock deformation
tational stresses in long symmetric ridges and in crustal fault zones. Annual Reviews of Earth and
valleys. Journal of Geophysical Research 91: 367785. Planetary Sciences 14: 14975.
Savage, W. Z., Swolfs, H. S., and Powers, P. S., 1985, Sibson, R. H., 1987, Earthquake rupturing as a mineral-
Gravitational stresses in long symmetric ridges and izing agent in hydrothermal systems. Geology 15:
valleys. International Journal of Rock Mechanics and 7014.
Mining Science 22: 291302. Sibson, R. H., 1989, Earthquake faulting as a structural
Scheidegger, A. E., 1962, Stresses in the Earths crust as process. Journal of Structural Geology 11(1/2): 114.
determined from hydraulic fracturing data. Geologie Sieh, K., Jones, L., Hauksson, E., et al., 1993, Near-field
und Bauwesen 27: 4553. investigations of the Landers earthquake sequence,
Schlichting, H., 1979, Boundary-Layer Theory. New York: April to July 1992. Science 260: 1716.
McGraw-Hill. Sih, G. C., 1973, Handbook of Stress Intensity Factors.
Scholz, C. H., 1990, The Mechanics of Earthquakes and Bethlehem, PA: Institute of Fracture and Solid
Faulting. Cambridge: Cambridge University Press. Mechanics, Lehigh University.
Segall, P. and Harris, R. A., 1986, Slip deficit on the San Smith, R. B., 1975, Unified theory of the onset of
Andreas Fault at Parkfield, California, as revealed by folding, boudinage, and mullion structure. Geological
inversion of geodetic data. Science 233: 140913. Society of America Bulletin 86: 16019.
Segall, P., McKee, E. H., Martel, S. J., and Turrin, B. D., Smith, R. B., 1977, Formation of folds, boudinage, and
1990, Cretaceous age of fractures in the Sierra mullions in non-Newtonian materials. Geological
Nevada batholith, California. Geology 18: 124851. Society of America Bulletin 88: 31220.
Segall, P. and Pollard, D. D., 1980, Mechanics of discon- Smith, R. B., 1979, The folding of a strongly non-
tinuous faults. Journal of Geophysical Research 85: Newtonian layer. American Journal of Science 279:
433750. 27287.
Segall, P. and Pollard, D. D., 1983a, Joint formation in Smoluchowski, M., 1909, ber ein gewisses
granitic rock of the Sierra Nevada. Geological Society of Stabilittsproblem der Elastizittslehre und dessen
America Bulletin 94: 56375. Beziehung zur Entstehung von Faltengebirgen. Anz.
Segall, P. and Pollard, D. D., 1983b, Nucleation and Akad. Wiss. Krakau, Math. Naturw. 2: 3 and 727.
growth of strike slip faults in granite. Journal of Sokolnikoff, I. S., 1956, Mathematical Theory of Elasticity.
Geophysical Research 88(B1): 55568. New York: McGraw-Hill.
Segall, P. and Simpson, C., 1986, Nucleation of ductile Sokoutis, D., 1987, Finite strain effects in experimental
shear zones on dilatant fractures. Geology 14: 569. mullions. Journal of Structural Geology 9: 23342.
Selby, S. M., 1975, Standard Mathematical Tables. Sommer, E., 1969, Formation of fracture lances in
Cleveland, OH: CRC Press, p. 756. glass. Engineering Fracture Mechanics 1: 53946.
494 REFERENCES
Sosa, E. and Tooley, M., 1993, Causation. Oxford: Oxford echelon fractures in rock: implications from labora-
University Press, p. 252. tory and numerical experiments. Journal of Structural
Spence, D. A. and Turcotte, D. L., 1985, Magma-driven Geology 15: 32334.
propagation of cracks. Journal of Geophysical Research Thomson, W., 1891, Electrical units of measurement:
90: 57580. lecture delivered at the Institution of Civil Engineers
Stein, R. S., Dieterich, J. H., and Barka, A. A., 1996, Role on May 3, 1883. Nature Series, Popular Lectures and
of stress triggering in earthquake migration on the Addresses, Vol. 1, Constitution of Matter. London:
North Anatolian Fault. Physics and Chemistry of the MacMillan, pp. 80134.
Earth 21: 22530. Tikoff, B. and S. F. Wojtal, 1999, Displacement of
Stein, R. S. and Lisowski, M., 1983, The 1979 Homestead control of geologic structures. Journal of Structural
Valley earthquake sequence, California: control of Geology 21: 95967.
aftershocks and postseismic deformation. Journal of Timoshenko, S., 1953, History of Strength of Materials.
Geophysical Research 88: 647790. New York: McGraw-Hill.
Stephansson, O., 1974, Stress-induced diffusion during Timoshenko, S., 1958, Strength of Materials. I. Elementary
folding. Tectonophysics 22: 23351. Theory and Problems. Princeton, NJ: Van Nostrand.
Stephansson, O., 1993, Rock stress in the Timoshenko, S. and Goodier, J. N., 1970, Theory of
Fennoscandian shield. In J. A. Hudson, ed., Elasticity. New York: McGraw-Hill.
Comprehensive Rock Engineering. Oxford: Pergamon Timoshenko, S. and Woinowsky-Krieger, S., 1959, Theory
Press, pp. 44559. of Plates and Shells. New York: McGraw-Hill.
Sternlof, K. R., Chapin, J. R., Pollard, D. D., and Timoshenko, S. and Young, D. H., 1968, Elements of the
Durlofsky, L. J., 2004, Permeability effects of defor- Strength of Materials. Princeton, NJ: Van Nostrand.
mation band arrays in sandstone. American Treagus, S. H. and Lisle, R. J., 1997, Do principal sur-
Association of Petroleum Geologists Bulletin 88(9): faces of stress and strain always exist? Journal of
131529. Structural Geology 19: 9971010.
Stoker, J. J., 1969, Differential Geometry. New York: Wiley- Treagus, S. H. and Treagus, J. E., 2002, Studies of strain
Interscience. and rheology of conglomerates. Journal of Structural
Struik, D. J., 1961, Lectures on Classical Differential Geology 24: 154167.
Geometry. Reading, MA: Addison-Wesley. Treiman, J., Kendrick, K., Bryant, W., Rockwell, T., and
Tada, H., Paris, P. C., and Irwin, G. R., 1973, The Stress McGill, S., 2002, Primary surface rupture associated
Analysis of Cracks Handbook. Hellertown, PA: Del with the Mw 7.1 16 October 1999 Hector Mine earth-
Research Corp. quake, San Bernardinno County, California. Bulletin
Tapponnier, P. and Brace, W. F., 1976, Development of of the Seismological Society of America 92: 1171191.
stress-induced microcracks in Westerly granite. Truesdell, C., 1961, Stages in the development of the
International Journal of Rock Mechanics and Mining concept of stress. In Problems of Continuum Mechanics.
Science 13: 10312. Philadelphia, PA: Society for Industrial and Applied
Taylor, W. L. and Pollard, D. D., 2000, Estimation of in Mathematics, pp. 55664.
situ permeability of deformation bands in porous Truesdell, C. and Noll, W., 1965, The non-linear field
sandstone, Valley of Fire, Nevada. Water Resources theories of mechanics. In S. Flugge, ed., Encyclopedia
Research 36(9): 2595606. of Physics, Vol. 3, No. 1. Berlin: Springer-Verlag.
Taylor, W. L., Pollard, D. D., and Aydin, A., 1999, Fluid Trusheim, F., 1960, Mechanism of salt migration in
flow in discrete joint sets: field observations and northern Germany. American Association of Petroleum
numerical simulations. Journal of Geophysical Research, Geologists Bulletin 44: 151940.
B, Solid Earth and Planets 104(12): 28 98329 006. Tullis, T. E., 1988, Rock friction constitutive behavior
Terzaghi, K., 1943, Theoretical Soil Mechanics. New York: from laboratory experiments and its implications
John Wiley and Sons. for an earthquake prediction field monitoring
Thatcher, W. and Bonilla, M., 1989, Earthquake fault program. Pure and Applied Geophysics 126: 55588.
slip estimation from geologic, geodetic and seismo- Tullis, T. E. and Tullis, J., 1986, Experimental rock
logical observations: implications for earthquake deformation techniques. In B. E. Hobbs and H. C.
mechanics and fault segmentation, US Geological Heard, eds., Mineral and Rock Deformation: Laboratory
Survey, Open File Report 89315. Menlo Park, CA: Studies, The Paterson Volume, Geophysical Monograph.
pp. 38699. Washington, DC: American Geophysical Union,
Thomas, A. L. and Pollard, D. D., 1993, The geometry of pp. 297324.
REFERENCES 495
Turcotte, D. L. and Schubert, G., 1982, Geodynamics: tation and patterns of wing cracks and solution sur-
Applications of Continuum Physics to Geological Problems. faces at the tips of a sliding flaw or fault. Journal of
New York: John Wiley and Sons. Geophysical Research 103: 242738.
Turner, F. J. and Weiss, L. E., 1963, Structural Analysis of Willemse, E. J. M., Pollard, D. D., and Aydin, A., 1996,
Metamorphic Tectonites. New York: McGraw-Hill. Three-dimensional analyses of slip distributions on
Twiss, R. J. and Moores, E. M., 1992, Structural Geology. normal fault arrays with consequences for fault
New York: W. H. Freeman and Co. scaling. Journal of Structural Geology 18(23): 295309.
Van Dyke, M., 1982, An Album of Fluid Motion. Stanford, Williams, M. L., 1957, On the stress distribution at the
CA: The Parabolic Press. base of a stationary crack. Journal of Applied
Varberg, D. and Purcell, E. J., 1992, Calculus with Mechanics, Transactions ASME 24: 10914.
Analytical Geometry. Englewood Cliffs, NJ: Prentice- Wilson, G. and Cosgrove, J. W., 1982, Introduction to
Hall. Small-Scale Geological Structures. London: George Allen
Vendeville, B. C., Ge, H., and Jackson, M. P. A., 1995, and Unwin.
Scale models of salt tectonics during basement- Wiltschko, D. V. and Sutton, S. J., 1982, Deformation by
involved extension. Petroleum Geosciences 1: 17983. overburden of a coarse quartzite conglomerate.
Venkatraman, B. and Patel, S. A., 1970, Structural Journal of Geology 90: 72533.
Mechanics with Introduction to Elasticity and Plasticity. Wong, T.-F., 1982a, Effects of temperature and pressure
New York: McGraw-Hill. on failure and post-failure behavior of Westerly
Wang, W.-H. and Davis, D. M., 1996, Sandbox model granite. Mechanics of Materials 1: 317.
simulation of forearc evolution and noncritical Wong, T.-F., 1982b, Micromechanics of faulting in
wedges. Journal of Geophysical Research 101: 11 32939. Westerly granite. International Journal of Rock
Wallace, R. E., 1951, Geometry of shearing stress and Mechanics and Mining Sciences 19: 4964.
relation to faulting. The Journal of Geology 59: 11830. Wood, D. S. and Oertel, G., 1980, Deformation in the
Weertman, J., 1996, Dislocation Based Fracture Mechanics. Cambrian slate belt of Wales. Journal of Geology 88:
River Edge, NJ: World Scientific. 30926.
Weertman, J. and Weertman, J. R., 1964, Elementary Woodford, A. O., 1956, What is Geologic Truth? Journal
Dislocation Theory. New York: Macmillan. of Geological Education 4(1): 58.
Weinberger, R., 2001, Joint nucleation in layered rocks Woodworth, J. B., 1896, On the fracture system of
with non-uniform distribution of cavities. Journal of joints, with remarks on certain great fractures.
Structural Geology 23: 124154. Boston Society of Natural Historical Proceedings 27:
Weyl, H., 1987, The Continuum: A Critical Examination of 16383.
the Foundation of Analysis. Kirksville, MO: The Thomas Wosser, T. D., Campi, D. E., Fovince, M. A., and Smith,
Jefferson University Press, p. 94. W. H., 1982, Damage to Engineered Structures in
White, F. M., 1974, Viscous Fluid Flow. New York: California, The Imperial Valley, California, Earthquake of
McGraw-Hill. October 15, 1979, Professional Paper 1254. Washington,
Whitehead, J. A. Jr. and Luther D. S., 1975, Dynamics of DC: US Geological Survey.
laboratory diapir and plume models. Journal of Wu, H. and Pollard, D. D., 1995, An experimental study
Geophysical Research 80: 70517. of the relationship between joint spacing and layer
Whitney, C. S., 1943, Discussion on paper by V. P. thickness. Journal of Structural Geology 17(6): 887905.
Jensen. Journal of the American Concrete Institute 39: Yeats, R. S., Sieh, K., and Allen, C. R., 1997, The Geology of
548. Earthquakes. New York: Oxford University Press.
Wilczek, F., 2004, Whence the force F ma? II: Zakarevicius, A., 2000, Skaitmeniniai vertikaliuju
Rationalizations. Physics Today V: 1011. zemes plutos deformaciju rodikliai. Geodezija ir
Willemse, E. J. M., 1997, Segmented normal faults: cor- Kartografija [Digital values of vertical crustal defor-
respondence between three-dimensional mechanical mations. Geodesy and Cartography] 26: 579.
models and field data. Journal of Geophysical Research Zebker, H. A., Rosen, P. A., Goldstein, R. M., Gabriel, A.,
102(1): 67592. and Werner, C. L., 1994, On the derivation of co-
Willemse, E. J. M., Peacock, D. C. P., and Aydin, A., 1997, seismic displacement fields using differential radar
Nucleation and growth of strike-slip faults in lime- interferometry: the Landers earthquake. Journal of
stones from Somerset, UK. Journal of Structural Geology Geophysical Research 99: 19 61743.
19: 146177. Zenzri, H. and Keer, L. M., 2001, Mechanical analyses of
Willemse, E. J. M. and Pollard, D. D., 1998, On the orien- the emplacement of laccoliths and lopoliths. Journal
496 REFERENCES
of Geophysical Research, B, Solid Earth and Planets 106(7): New evidence on the state of stress of the San
13 78192. Andreas fault system. Science 238: 110511.
Zheng, Z., Kemeny, J., and Cook, N. G. W., 1989, Zoback, M. L., 1992, First- and second-order patterns of
Analysis of borehole breakouts. Journal of Geophysical stress in the lithosphere: the world stress map
Research 94: 717182. project. Journal of Geophysical Research 97: 1170328.
Ziv, A. and Rubin, A. M., 2000, Stability of dike intru- Zoback, M. L., Zoback, M. D., Adams, J., et al., 1989,
sion along preexisting fractures. Journal of Geophysical Global patterns of tectonic stress. Nature 341: 2918.
Research 105(3): 594761. Zuber, M. T., Parmentier, E. M., and Fletcher, R. C., 1986,
Zoback, M. D., 1985, Well bore breakouts and in-situ Extension of continental lithosphere: a model for
stress. Journal of Geophysical Research 90: 552330. two scales of basin and range deformation. Journal of
Zoback, M. D., Zoback, M. L., Mount, V. S., et al., 1987, Geophysical Research 91: 482638.
Index
acceleration convention dip

of gravity 251 on-in 209 angle of 55
linear 246 right-hand 54, 90, 99, 216, 247 apparent 55
particle 263 coordinates direction of 54
vector 245 Eulerian 263 direction angle 38, 42, 68
accretionary wedge 393, 410 Lagrangian 262 direction cosine 43 ,68
Airy stress function 310, 313 material 262 direction number 112
Amontons law 351 right-handed 36, 39 dislocation
Andersons standard state 229 spatial 263 core 302
anisotropy 325, 416 coordinate system edge 301
elastic 325 Cartesian 34, 36 glide plane 302
viscous 416 cylindrical 44 line 301
annum 124 elliptical 47 screw 302
anticrack 17 polar 45 displacement
arc length 82 spherical 48 discontinuity of 469
azimuth 54 correspondence principle 387 gradient of 185
Coulomb vector 160, 185, 191
basin and range 436 criterion 358 divergence 268
binomial distribution 67 critical angle 360 divergence theorem 275
boundary condition 46, 132, 214 critical stress 361 ductile state 338
boundary value problem 10, 236, stress 359 dynamics
279, 285, 299, 308, 388 curvature particle 245
boudinage 423 Gaussian 114 rigid body 248
brittle state 334, 338 mean principal normal 114
brittle-ductile transition 7 principal normal 110, 113 effective spring constant 202
Buckingham theorem 140 principal directions of normal 111, effective thickness 466
Burgers vector 302 113 eigenvalue 219
Byerlees law 354 radius of 86 eigenvector 219
scalar 85 elastic modulus
cartographic criterion vector 84 bulk 298
equidistance, conformality, cylindrical structure 299 shear 278, 298
equivalency 30 Lames 278, 299
Cauchys tetrahedron 203, 213, damage zone 355 Poissons ratio 295
215 datum 27 tangent 294
Cauchys formula 213 sea level 29 Youngs 294
center of gravity 252 deformation elastic solid
center of mass 249 gradient 185, 190 linear 292
cleavage 169 Green tensor of 190 non-linear 293
compatibility 310, 314 homogeneous 154, 189 orthotropic 329
compliance 326 inelastic 293 transversally isotropic
composite material 446 progressive 177 328
compressibility 295 degree Celsius 124 elastic heterogeneity 323
conservation law diapir 137 element
of angular momentum 252, 274 differential geometry 75 linear 52
of linear momentum 250, 269 dike 9, 40, 47, 100 planar 52
of mass 266 dimensional analysis 127 epicenter 7
constitutive law 276, 278, 283, 326, direct method of 133, 135 equation of state 283
416 Rayleigh method of 137 equilibrium
continuity 266, 268 dimensional homogeneity 128 quasi-static 280, 309
continuum 124 dimensionless group 132 static 251, 253
498 INDEX
Euclidean space 37 propagation of 354 lineation

Eulers theorem 111 shear 334 discrete 77
sliding 372 penetrative 64, 77, 80
fabric splitting 334 metamorphic 64
diagram of 64 tearing 372 superficial 77
failure criterion fracture toughness 357 local rate of change 265
Griffith 369 friction lubrication theory 394
Coulomb 358 coefficient of 353
failure surface 345 coefficient of internal 359 mass 245
fault dynamic 352 material continuum 124
normal 70 static 351 material line 154, 185
reverse 70 frictional strength 353 material time derivative 266
strike-slip 72 fundamental form matrix
termination of 13 first 103 column 50
thrust 256 coefficients of 104, 108 identity 219
tipline of 13 second 108 row 50
fault-plane solution 228 Froude Number 147 square 51
fluid metrology 122
anisotropic viscous 481 Gausss theorem 275 minimal surface 114
power-law 389, 423, 426 geographic coordinates Mohr diagram 223
linear viscous 282, 385, (easting, northing, elevation) 31 momentum
fold (latitude, longitude, elevation) 27 angular 247, 248
antiformal 114 geoid 29 linear 246
axial surface of 92 geometric model 158 principle of 273
axis of 80, 92 Global Positioning System (GPS) 29 motion
chevron 168, 418 graben 9 equation of 269, 272, 279, 285
cylindrical 81, 92 great circle 59 referential description of 260
dominant wavelength of 407 spatial description of 263
hinge of 92 homogeneous equation 219 moving trihedron 89
single layer 403 homogeneous material property 293 mullion 397
synformal 114 Hookes law 297, 325
foliation 64 hyperbolic function 47 nappe 155, 394
forward problem 153 hypocenter 8 neotectonics 159
force net
body 196 incompressible 269, 284, 295, 387 Wulff (meridional stereographic)
resultant 245, 250 index (dummy, free) 49 57
surface 196 indicial notation 49 Schmidt 63
flow inertial frame of reference 245, 247, Newtons First Law 246
antiplane 388 249 Newtons Second Law 204, 246
barotropic 183 inflection point 85 Newtons Third Law 200
laminar 137 inverse problem 153 normal (pole) to a planar element
plane 388 isochemical 244 62
steady 138, 165, 265, 386, 388 isothermal 244, 260, 293
turbulent 137 isotropy oid 153
flux elastic 297 order of magnitude 123
mass 267 viscous 387 osculating plane 108, 110
momentum 270
fracture kinematic model 158 parametric representation 79, 93
extension 334 kinematics 158, 160 passive folding 175
hydraulic 234 kinetics 244 perfect confinement 234
mechanics of 356 Kronecker delta 50 photoelasticity 194, 226, 356
modes of 371,372 physical dimension 127
opening 237, 239, 354, 371 laccolith 21, 133, 461 plane curve 91
potential shear 358, 360 Lams constant 278, 299 plate theory 463
INDEX 499
plunge sliding principal normal 216,

angle of 55 stable 351 principal trajectory of 220
direction of 55 stick-slip 351 shadow 325
Poissons ratio 295 slip shear component of 210
preferred orientation 64 dip 3 supplementary 469
pressure left-, right-lateral 18 tensile 208
breakdown 235, 237 strike 72 stress tensor 211, 214
confining 335, 345 small circle 59 Cartesian components of
driving 290 Smoluchowski Number 148 cylindrical components of 212
effective confining 349 solution seam (surface) 17 deviation 346
mean normal 283 spring constant 202, 290 invariant of 221, 345
pore 346, 350 Stokes condition 284 stress-strain curve 338
re-opening 238 Stokes Number 148 stretch 154, 186
shut-in 235 strain strike
static 283 axial 338 azimuth of 54
thermodynamic 283 bulk 172 line of 54
principal value problem 219 plain 299 substantial derivative 266
projection strain-rate thinning 423 summation convention 49
gnomonic 30 strain softening 423
Lambert (equal area) 63 strain tensor tangent plane 97
map 30 infinitesimal 186, 190, 278, 281, tensor
stereographic (equal angle) 56 300 fourth-rank (order) 326
Universal Transverse Mercator Lagrangian 190 second-rank (order) 185, 214, 226
(UTM) 30 three-dimensional 190, 191 test
prototype 144 two-dimensional 186 compression 319, 335
stream function 165 extension 334
rake 56 streamline 165 triaxial 347
Ramberg Number 148 strength 337 topographic contour 27
rate of twist 101 differential 347 torsion 90
reductionism 24, 456, 462 inherent shear 359 traction
reference circle 56 intrinsic tensile 366 director surface 205
reference sphere 56 maximum shear 358 ellipsoid 205
reflection 39 octahedral shear 346 free surface 231, 310, 315
representative elementary area triaxial compressive 347 vector 198
203 uniaxial tensile, compressive transformation
Reynolds transport theorem 275 342 coordinate (rotation) 39
Reynolds Number 137 stress homogeneous linear 174
rigidity axial 338 translation (rigid) 313
effective flexural 465 compressive 210 transpression 391, 393
flexural 464 concentration of 227, 237, 357
rotation (pure) 187 differential 347 umbilical point 113
rotation (rigid) 313 effective 348 unit vector
hydrostatic 229, 298 binormal 89
scale factor 134, 137 intensity of 357, 372 normal 99, 107
scientific notation 123 in situ 234 principal normal 88
seismic reflection survey 11 isotropic state of 228 tangent 82
shear zone 334 lithostatic state of 230 unit of measurement 122
SI units 122 maximum shear 221, 358 UTM grid 33
significant figures 123 mean normal 298
similarity near-tip 356, 372 vector
geometric 145 normal component of 208 base 36
kinematic 145 octahedral normal 346 cross product 87
dynamic 146 octahedral shear 346 dyadic product 270
slickenline 53 polyaxial state of 345 fixed 37
500 INDEX
vector (cont.) particle 262 Newtonian 386

null 219 rupture 8 non-Newtonian 422
orthonormal basis 36 vector 247, 262, 264
position 34, 37 viscosity wellbore breakout 239
scalar component of 36 bulk 284 work
scalar product 42 Newtonian 284, 386 principle of virtual 341
triple scalar product 90 Paul estimate of 449
vein 17 Reuss estimate of 447 xenolith 161
velocity Voight estimate of 448
local 264 viscous fluid Youngs modulus 294

Fundamentals of Structural Geology (Pollard D.D., Fletcher R.C.)

Diunggah oleh

Informasi Dokumen

Judul Asli

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

Fundamentals of Structural Geology (Pollard D.D., Fletcher R.C.)

Diunggah oleh

Hak Cipta:

Format Tersedia

Fundamentals of Structural Geology

Cambridge University Press is part of the University of Cambridge.

Chapter 1 Motivations and opportunities 1

Chapter 2 Structural mapping techniques and tools 25

Chapter 3 Characterizing structures using differential

Chapter 4 Physical quantities, fields, dimensions, and

Chapter 5 Deformation and flow 152

Chapter 6 Force, traction, and stress 194

Chapter 7 Conservation of mass and momentum 243

Chapter 8 Elastic deformation 287

Chapter 9 Brittle behavior 333

Chapter 10 Viscous flow 384

Chapter 11 Rheological behavior 421

11.4 Continuum properties of composite materials 446

Chapter 12 Model development and methodology 456

Motivations and opportunities

on the fault offset the two small drainage chan-

(a) stable monuments. In some cases the instruments

directly record fault slip, even at the surface.

the fault as a set of segments, each with a constant

Models of the sub-surface fault geometry are

placements associated with earthquakes was 86

region is imaged at two different times, for 32o 45'

Cross sectional view Shear

because it occurred within a dense array of geo- 6

212o 213o 214o 215o 216o E

1.2 Radar lineaments on Venus

(a) the reectors in the pile of sedimentary layers.

NORWAY Specic sedimentary horizons are interpreted

in Fig. 1.9b including the top of the Brent Group,

the top of the Heather Formation, and the base of

Bergen the Cretaceous. Interpretations of the normal

Shetland faults (thick steeply inclined white lines) were

Oslo drawn and labeled by Faerseth and his colleagues

from the Norwegian oil company Norsk Hydro

ASA. The faults strike approximately northsouth

so seismic sections along eastwest trends such as

volume of rock. The seismic section in Fig. 1.9 has 2 km

closed contours distributed along the fault at

Interpreted slip from 3D seismic data

(b) Computed slip from model 1

(c) Computed slip from model 2

Fig 1.12 Fault maps and slip distributions on fault A

oriented parallel to the bottom of the photograph.

1.5 Mountain building on the

Grove Karl Gilbert, to investigate the Wasatch (a)

strata, would nd it easier to spread laterally at 2a

Table 1.1. Gilberts field data on laccolite diameter.

Zone Formations Laccolite name Diameter (miles)

Upper Blue Gate Shale Sentinel 0.7

Structural mapping techniques and tools

identify the positions of outcrops or sampling (a) Latitude N

Latitude: 38 45 N (a) Rotation axis GRS-80

Proje dene the Cartesian coordinates (X, Y) of projected

(a) (c) Zone 10 84o N

Fig 2.4 The Transverse Mercator (TM) projection

horizontal plane from a local origin at the instru- (a)

The precision of these coordinates is about 30 m

in an azimuth of 056, which is approximately and

Fig 2.7 Structural map of several segments of the

new coordinates, P(x, y), which are found using (a) in a

The inverse transformation for the components of (a) z

a table of directional cosines for rotation of posi-

Latitude: 38 45 N (a) Rotation axis GRS-80

The precision of these coordinates is about 30 m

in an azimuth of 056, which is approximately and

new coordinates, P(x, y), which are found using (a) in a

px py pz Km

p pxex pyey pzez The direction cosines of (2.28) are referred to as

Angle ACP 180 90 p 90 p 2 Projection x P

Because AC CP R, the triangle ACPA is isosceles W A E, x

Because the angle ZPC is , the angle ZPA is a right

c(t) a (cos t)ex a( sin t)ey btez

cx(t) a(cos t), cy(t) a(sin t), cz(t) bt,

c(t) a(cos t)ex a(sin t)ey (circle) (3.4)

In the limit, as s goes to zero the secant becomes