Anda di halaman 1dari 6

The Body Force Equivalent to an Earthquake: A Tutorial

Jos Pujol
CERI
As noted in the Introduction to the classic paper by Burridge and Knopoff (1964) (this issue), this paper played a critical role in the development of the theory of the earthquake source. Because the mathematical development presupposes a level of knowledge beyond that of the seismologist or seismology student not actively engaged in theoretical studies, i will discuss a number of basic results that will help understand the paper as well as to put it into a broader perspective. I hope this tutorial will entice those untheoreticaUy oriented to read the Burridge and Knopoff paper, so that they can appreciate the essential elegance (as opposed to a brute-force approach) of the arguments used to derive the expression for the body force equivalent. To make the tutorial self-contained a few excerpts from the Introduction are repeated here. For a homogeneous isotropic elastic medium the equation that governs the propagation of seismic waves can be written as (Miklowitz, 1978) and in addition introduced what he called "double forces", which are pairs of parallel forces in opposite directions (Love, 1927). If the forces have different lines of action the pair is known as a couple; otherwise it is known as a dipole. Assuming that the distance between the forces is infinitesimal, the solution of Equation 1 for a double force can be obtained without much effort when f is a concentrated (or point) force. Let f = g(t) 8(x - ~)e., where 6 is Dirac's delta, ~: indicates the location of the force,:1and ej is a unit vector in the :9 direction, and let uij (x, ~, ~) be the corresponding ith component of the vector u. Here t is the time measured from the time at which the source began to act. Expressions for u/j and their derivations can be found in Aki and Richards (1980), Miklowitz (1978), and Pujol (2003), among others. For our purposes, however, we need to know only that such expressions exist. To find the displacement due to a couple consider now a concentrated force f having time-dependence g(t)/h acting at = (0, O, h/2) in the x 1 direction, and another force of equal magnitude acting at ~r = (0, 0,-h/2) in the negative x 1 direction (Figure 1). The second force can be written as -f' and assumed to be in the positive x I direction. Then the corresponding displacements are given by +u i.~ (x, t; ~)/h, where U i.r] is the displacement due to the force with time-dependence g(t). Finally, let h go to zero. Under these conditions, the displacement due to this couple can be written as

pii(x, t)= (;t, + 2bt)graddiv tt(x, t) -/~ curl curl u(x, t) + f(x, t)

(1)

where u is the vector that describes the displacement of the particles in the medium, x = (x I, x 2, x3) denotes the observation point, such as a seismic station, the double dots indicate second derivative with respect to time, ~. and/,t are the Lamd's parameters, p is density, and f is the body force per unit volume that is the source of the waves (e.g., Aid and Richards, 1980). Although solving Equation 1 for a given f is a difficult problem, solutions for simple forces have been known since 1849. The problem that seismologists faced when they began to study the generation of waves by an earthquake was to find the body force to be used in Equation 1. As noted below, the relation between faults and earthquakes was well established before a model for the earthquake source was available and the problem that seismologists addressed was, briefly stated, the following: What is the equivalent body force that in the absence of the fault will cause exactly the same displacement field as slip on the fault? To see how this problem was approached first we must summarize the work done toward solving Equation 1. In 1849 Stokes presented the solution for the case of a spatially concentrated force directed along a coordinate axis. Love confirmed this solution in 1904 using a different approach

u/,,;/., ,; o) - 'ira [ u''('';/~176


0~3

u,,(,,,;/o.o.-

where 0 = (0, 0, 0) and the superscript stands for single couple (after Miklowitz, 1978). The right-hand side of Equation 2 can also be obtained by adding the displacements written in Taylor expansions to first order (Udfas, 1999). This derivative offers a practical and general way to obtain the displacement field due to a couple, but one can actually start with the two displacements indicated in Equation 2, introduce a number of approximations, and then take the limit. This approach was used by Lay and Wallace (1995) to determine the far field, but it is much simpler to use the derivative (see Pujol and Herrmann, 1990).

SeismologicalResearchLetters March/April2003 Volume74,Number2 163

(A)

~-3 ~2

= au,,(=,,;o)

(3)

h/2
_f~

f
-h/2

;f

The effect of this couple is a clockwise rotation about the x 2 axis and the sign of its moment is opposite to that of the previous couple. Therefore, if the two couples are combined into a single system of forces the result is a double couple without moment with displacement equal to the sum of the displacements given in Equations 2 and 3:

UiSI(X, )so ,;O U/1S(X, )so t;O ~_ aUil(X't;O) _[_ aUiJ(X' t;O)
_

(4)

a~ 3

a~l

(B)
~2

-h/2

h/2

-f

,~, Figure 1. Graphical representationof single couples with forces +_.fin (A) the x~ direction and (B)the x3 direction.
Let us note some features of this single couple. First, the absolute value of g(t)lh approaches infinity as h approaches zero. Second, the arm of the couple is in the x3 direction. Third, the couple tends to produce a counterclockwise rotation about the x2 axis. Thus, the couple has a net moment, which from mechanics we know it is equal to force times distance, or h x g/h = g. Also note that the three subindices of u on the left-hand side of Equation 2 denote the component of the displacement vector, the direction of the force, and the direction of the arm of the couple, respectively. Now consider a problem similar to the previous one but let the forces be in the x3 direction and the arm of the couple i n the x 1 direction, with the negative force acting at /2,0,0) (Figure 1). In this case the displacement is given by

(-h

The first equality in Equation 4 arises because we can change the order of the two terms on the right-hand side. Note that the fact that the double couple does not have a net mechanical moment does not mean that it does not have strength as a seismic source. For example, the double couple equivalent to an earthquake has zero mechanical moment, but its size is given by the scalar seismic moment. The debate over the mechanism of earthquakes spanned several decades. It began in the 1920's and lasted through the early 1960's and is well described by Stauder (1962). As a result of observational studies that showed a quadrantal pattern of P-wave dilatations and compressions and the recognition of the close relation between earthquakes and faults, two leading models, based on the couples described above, were proposed. One of them, favored by California seismologists, was a single couple, while Japanese seismologists preferred a double couple. A problem with the single couple was that it had a net moment, which was an undesirable feature because it implied a nonequilibrium condition inside the Earth. In principle, the observations should have allowed discrimination between the two models, but this was not possible for two reasons. First, the single and double couples have similar P-wave radiation patterns (Figures 2 and 3). Second, although the corresponding S-wave radiation patterns are quite different (Figures 2 and 3), the quality of the data was inadequate to establish uniquely which model was more appropriate. In addition, even if the observations had allowed the identification of the right model, the fact would have remained that no theoretical justification existed for either of them. As this debate continued, during the i950's another theoretical tool was brought to bear, namely dislocation theory. This theory originated in the work of a number of Italian mathematicians, particularly Volterra, who used the word distorsione. Dislocation is Love's translation (Love, 1927). A dislocation can be visualized through the following thought experiment, based on Steketee (1958). Consider a cut made over a surface E within an elastic body. After the cut has been made there are two surfaces, indicated with s and E-, which will be deformed differently by application of some force distribution. If the combined system of forces is in static equilib-

164 Seismological Research Letters Volume 74. Number 2 March/April 2003

SV

x1

x2

x1

x2

(A)

(B)

,& Figure 2. P- and SV-wave radiation patterns for the single couple of Figure lB. In these plots only the orientation of the observation location with respect to the source location, not the distance, is considered. From Pujol (2003). This source does not generate SH waves. For the equations used to generate the plots see also Pujol and Herrmann (1990). x3 P

X1

x2

x1

x2

(A)

(B)

x3

x3

sv

s.

_1

(C)

(D)

,i, Fi0ure 3. P-, S-, $V-, and SH-wave radiation patterns for the double couple obtained by superposition of the two single couples of Figure 1. The S pattern corresponds to the absolute value of the S-wave displacement. From Pujol (2003). For the equations used to generate the plots see also Pujol and Herrmann (1990). The only difference between this P radiation pattern and that in Figure 2 is a factor of 2.

Seismological Research Letters March/April 2003

Volume 74, Number 2

165

rium, the body will remain in the original equilibrium state. The result of this operation is a discontinuity in the displacement across E, known as a dislocation, which is accommodated by deformation within the body. This description should be compared to our model for a tectonic earthquake, which is represented by slip on a fault plane. When an earthquake occurs, the two sides of the fault suffer a sudden relative displacement with respect to each other, and this discontinuity in the displacement across the fault is the source of the displacement elsewhere in the medium. A good discussion of the relation between earthquakes and dislocations is provided by Steketee (1958), although his analysis was restricted to the static case. Contributions from other authors are described by Stauder (1962). The debate ended when Maruyama (1963), Haskell (1964), and Burridge and Knopoff (1964) demonstrated that the body force equivalent was a double couple. In the three cases the derivations were based on a number of results derived in the context of theoretical elasticity and wave propagation. While the first two authors addressed the case of homogeneous isotropic media, what distinguishes Burridge and Knopoff's paper is its generality, as their results apply to heterogeneous anisotropic media. To achieve this generality the authors relied heavily on the powerful formal manipulation tools provided by Dirac's delta (see below). Because the analysis in Burridge and Knopoff (1964; hereinafter equation numbers from this paper will be preceded by BK) is completely general, the starting point is the equation

Aki and Richards, 1980). For readers not familiar with indicial notation, when Equation 5 is written in full it becomes

(10)

(s

q(X,t)),j--~O(X)Ui(X,t)=--fi(x,r )

(5)

(see BK. 1, page 154), which has Equation 1 as a special case. Equation 5 applies to anisotropic heterogeneous media and can be obtained from the following relations:

To go from this equation to Equation 5 the following rules are used: Two repeated indices imply summation and the indices to the right of a comma indicate derivatives with respect to the corresponding component of x. A basic result in Burridge and Knopoff (1964) is a socalled representation theorem, which is an expression involving integrals that relates a pair of forces and their corresponding displacements, with one of the forces being a Dirac's delta in the space and time domains. The displacement corresponding to this concentrated force is known as Green's function. Given this force-displacement pair, the displacement generated by a second force is given by BK.9, page 156. This equation needs some explanation. The fault plane on which the earthquake occurs is represented by the surface E in Figure 1 of Burridge and Knopoff (1964), which in principle need not be planar. This surface is embedded in a volume V of an elastic medium bounded by a surface S. To introduce E into the problem two steps are needed. First, one of the volume integrals in BK.4, page 155, is converted to a surface integral using Gauss' (or divergence) theorem. This gives BK.5, page 155, which in turn was written for the case of a force gi(x, t) concentrated both in space and time and assuming that the displacement and Green's function satisfy homogeneous boundary conditions on the surface S. The corresponding result is BK.8, page 155, but it cannot be applied directly to the earthquake problem because the displacement has a discontinuity on ~, which is within the integration volume V. The second step is to account for this situation, which requires modification of Gauss' theorem as follows. Given a vector-valued function b(x, t), continuous within a volume V, the theorem states that
~r V . b d V = ~s b . ndS

~2Ui = PUi rij, j + f - P at i

(6)

(11)

where P indicates density and ~ is the ith component of the force f, and
1

r/s = copqEpq = c,jpq u~,q


Cog p = %q

(8)
(9)

The integrand on the left side is the divergence of b, which using indicial notation is written as bi, i (a summation over i is implied), and n is an outward unit vector normal to S. In indicial notation the dot product b .n is written as bin i. Equation 1 t was used to obtain BK.5, page 155. When b is discontinuous across a surface E within V the theorem becomes
Jv V " b d V = Js b "ndS - fx [ b ] "vdE

(12)

Equation 6 is Cauchy's equation of motion, T~ and epq indicate the stress and strain tensors, u corresponds to displacement, and the first equality in Equation 8 is Hooke's law (e.g.,

where v is a unit vector normal to E and [b] is the difference in the values of b across E (e.g., Kraut, 1967; Pujol, 2003). The proof of this result requires the following steps: (a) Sur-

166 SeismologicalResearchLetters Volume74, Number2 March/April2003

(15)

i Figure 4. Geometry for Gauss' theorem in the presence of a surface of discontinuity (E). Y_., + and E- indicate the two sides of Y_,,.C+ and 6'- combined form a cavity Cthat surrounds Z;. v + and v-are unit normal vectors. The vector v in Equation 12 is equal to v-.

(e.g., Aki and Richards, 1980) where 8/j is Kronecker's delta, which is equal to one when I = j and equal to zero otherwise. If, in addition, it is assumed that the traction vector has no discontinuity, the second term in BK. 10 vanishes. Continuity of the traction is a reasonable assumption for an earthquake source because it corresponds to Newton's law of action and reaction and is expected unless external forces are applied to E. Burridge and Knopoff (1964) discusses several specific examples of faulting. One of them has E on the (x 1, x 2) plane, so that the normal vector v is in the x3 direction, while the discontinuity in displacement is nonzero in the x 1 direction only. In this case
(16) Under these conditions Equation 10 becomes

round E by a cavity having surface C (Figure 4), (b) introduce a volume V' having external surface S and internal surface C, i.e., V' is equal to Vwith the cavity subtracted, (c) apply Gauss' theorem to V', within which b is continuous, and (d) use a limiting argument to extend the theorem to the whole V. Combining BK.8, page 155, and Equation 12 gives BK.9, page 156, which in turn must be rewritten to derive the expression for the body force equivalent. This requires introducing volume integrals into the second term on the right-hand side of BK.9, page 156, which is done using ~(x- ~), with ~: a vector that identifies points on s In Burridge and Knopoff (i 964) this delta is written as 8(x;~). Then, for a given function f ( x ) , the following applies:

CI3pq
(17)

/(q>) = l-v

(13)

To indicate the integration variable the subscript x was used. This explains the first equation in the Equivalent Forces section of Burridge and Knopoff (1964), page 156. The second equation is similar, but now an expression involving the derivative of the delta is used, namely

The subindices 1 and 3 come from [u l] and v3, respectively, which are the only components of u and v different from zero. Equation 17 shows that the possible 27 terms in BK.10 reduce to one because the only nonzero contributions come from p = 1 and q = 3 and p = 3 and q = 1. In both cases the result is p. Finally, under the assumption that E is limited to the origin (point source approximation), the three components of the body force are given by the expressions on page 157 of Burridge and Knopoff (1964). To interpret e1, approximate the derivative in e, (including the minus sign) as follows"

]
Oxq x+=r

=
x-7
(14)

~3 8(x3 )___h 8 x 3 -

-8 x3 +

(18)

(e.g., Pujol, 2003). Using Equations 13 and 14, BK.9, page


156, can be written as in the equation preceding BK. 10, page 156, which leads directly to the expression for the body force equivalent, given by BK. 10. For a given value ofp, that equation involves up to 27 terms coming from the sums over the indices i, j, and q (i.e., 33 terms). BK.10 is completely general. To specialize it to the case of isotropic media we use

Since e1 is in the x 1 direction, the left-hand side of Equation 18 can be represented by a couple in the x, direction with an arm in the xe direction, as shown in Figure 5. in a similar way it can be shown that u 3 corresponds to a couple in the x 3 direction with an arm in the x 1 direction (Figure 5). Therefore, the body force equivalent is a double couple without moment. Finally, it must be noted that the importance of Burridge and Knopoff's paper goes much farther than having demonstrated that the body force equivalent is a double couple. In fact, the paper was used by Aki (1966) to derive the well known relation M o =/x~S, where M 0 is the scalar seismic moment, K is the average of the absolute value of [u], and S is the area of E. In addition, BK.9, page 156, is extremely important in its own right because it is the basis of the

Seismological Research Letters

March/April 2003

Volume 74, Number 2

167

(A)

X3

ACKNOWLEDGMENT
The material presented here is based on Pujol (2003). CERI contribution No. 458.

REFERENCES
hi2
- 5(x 3 + h/2) " = 6(x3- h/2)
..... >X 1

-h/2

(B)

X3

(x3- h/2)

-h/2
. . . . .

h/2

-> X 1

-~(x~+

h/2)

Aki, K. (1966). Generation and propagation of G waves from the Niigata earthquake of June 16, 1964, Part 2: Estimation of earthquake moment, released energy, and stress-strain drop from the G wave spectrum, Bulletin of the Earthquake Research Institute 44, 73-88. Aki, K. and P. Richards (1980). Quantitative Seismolog~ Vol. I, San Francisco: W. H. Freeman and Co. Burridge, R. and L. Knopoff (1964). Body force equivalents for seismic dislocations, Bulletin of the SeismologicalSociety of America 54, 1,875-1,888. Reprinted in Seismological Research Letters 74, 154-162. Haskell, N. (1964). Total energy and energy spectral density of elastic wave radiation from propagating faults, Bulletin of the Seismological Society ofAmerica 54, 1,811-1,841. Kraut, E. (1967). Fundamentals of Mathematical Physics, New York: McGraw-Hill. Lay, T. and T. Wallace (1995). Modern Global Seismolog)4 San Diego, CA: Academic Press. Love, A. (1927). A Treatise on the Mathematical Theory of Elasticity, Cambridge: Cambridge University Press. (Reprinted New York: Dover, 1944.) Maruyama. T. (1963). On the force equivalents of dynamical elastic dislocations with reference to the earthquake mechanism, Bulletin of the Earthquake ResearchInstitute 41,467-486. Pujol, J. and R. Herrmann (1990). A student's guide to point sources in homogeneous media, SeismologicalResearchLetters61,209-224. Pujol, J. (2003). Elastic WavePropagationand Generation in Seismolog~ Cambridge University Press (in press). Spudich, P. and R. Archuleta (1987). Techniques for earthquake ground-motion calculation with applications to source parameterization of finite faults, in B. Bolt (editor), Seismic Strong Motion Synthetics, Academic Press, 205-265. Stauder, W. (1962). The focal mechanism of earthquakes, in H. Landsberg and J. Van Mieghem (editors), Advances in Geophysics9, Academic Press, 1-76. Steketee, J. (1958). Some geophysical applications of the elasticity theory of dislocations, CanadianJournal of Physics36, 1,168-1,198. Udfas, A. (1999). Principlesof Seismolog~ Cambridge: Cambridge University Press.

,A, Figure 5. Graphicalrepresentationof (A) ~x3)iOx 3 and (B)


-a~Xl)laXl .

moment tensor density and moment tensor discussed in Aki and Richards (1980) and because it allows the computation of the displacement caused by an earthquake when Green's function is known. For its application to strong motion studies see, e.g., Spudich and Archuleta (1987). El

CERI The University of Memphis Memphis, TN38152 pujol @ceri.memphis.edu

168 Seismological ResearchLetters Volume74,Number2 March/April2003

Anda mungkin juga menyukai