1
Abst r act A t heoret i cal exami nat i on of t he f r act ur i ng
of rocks by means of pressure appl i ed i n borehol es
l eads t o t he concl usi on t hat , regardl ess of whet her t he
f r act ur i ng f l ui d is of the penet r at i ng or nonpenet r at i ng
t ype, t he fractures pr oduced shoul d be appr oxi mat el y
per pendi cul ar t o t he axi s of least stress. The gener al
state of stress under gr ound is t hat i n whi ch the t hree
pr i nci pal stresses are unequal . For t ect oni cal l y r el axed
areas charact eri zed by nor mal f aul t i ng, the least stress
shoul d be hor i zont al ; the fractures pr oduced shoul d be
ver t i cal , and t he i nj ect i on pressure shoul d be less t han
t hat of t he over bur den. In areas of acti ve tectoni c
compressi on, the least stress shoul d be vert i cal and
equal t o the pressure of t he over bur den; the fractures
shoul d be hor i zont al , and i nj ect i on pressures shoul d be
equal t o, or gr eat er t han, t he pressure of the over bur den.
Hor i zont al fractures cannot be pr oduced by hydr aul i c
pressures less t han t he t ot al pressure of t he over bur den.
These concl usi ons are compat i bl e wi t h f i el d exper i ence
i n f r act ur i ng and wi t h t he results of l abor at or y ex-
per i ment at i on.
INTRODUCTION
The hydraulic-fracturing technique of well
stimulation is one of the major developments in
petroleum engineering of the last decade. The
technique was introduced to the petroleum in-
dustry in a paper by J. B. Clark of the Stano-
lind Oil and Gas Company in 1949. Since then,
its use has progressively expanded so that, by
the end of 1955, more than 100,000 individual
treatments had been performed.
The technique itself is mechanically related
to three other phenomena concerning which an
extensive literature had previously developed.
They are: (1) pressure parting in water-injec-
tion wells in secondary-recovery operations,
(2) lost circulation during drilling, and (3) the
breakdown of formations during squeeze-ce-
menting operationsall of which appear to in-
volve the formation of open fractures by pres-
sure applied in a well bore. The most popular
interpretation of this mechanism has been that
the pressure had parted the formation along a
1
Manuscript received, July 4, 1972. Published, in
revised form, with permission of Society of Petroleum
Engineers of AIME; originally published in Transac-
tions of Society of Petroleum Engineers of AIME, 1957,
v. 210, p. 153-168.
2
U.S. Geological Survey.
3
Consultant.
M. KI NG HUBBERT
2
and DAVI D G. WILLIS'
Washi ngt on, D.C., and At her t on, Cal i f or ni a
bedding plane and lifted the overburden, not-
withstanding the fact that, in the great majority
of cases where pressures were known, they
were significantly less than those due to the to-
tal weight of the overburden as determined
from its density.
Prior to 1948, this prevalent opinion had al-
ready been queried by Dickey and Andresen
(1945) in a study of pressure parting and by
Walker (1946, 1949), who, in studies of
squeeze cementing, pointed out that the pres-
sures required were generally less than those of
the overburden, and inferred that the fractures
should be vertical. J. B. Clark (1949), in his
paper introducing hydraulic fracturing, and
later Howard and Fast (1950) and Scott et al.
(1953), postulated that the entire weight of the
overburden need not be lifted in producing hor-
izontal fractures, but that it is only necessary to
lift an "effective overburden," requiring a cor-
respondingly lower pressure. Hubbert (1953a)
pointed out that the normal state of stress un-
derground is one of unequal principal stresses
and that, in tectonically relaxed areas charac-
terized by normal faults, the least stress should
be horizontal. Therefore, in most cases, fractur-
ing should be possible with pressure less than
that of the overburden; moreover, such frac-
tures should be vertical. Harrison et al. (1954),
also on the expectation that the least principal
stress should be horizontal, argued strongly in
favor of vertical fracturing.
Scott et al. (1953) observed that, by the use
of penetrating fluids, hollow, cylindrical cores
could be ruptured at less than half the pres-
sures required if nonpenetrating fluids were
used. They also observed that, with penetrating
fluids, the fractures occurred parallel with the
bedding, irrespective of the orientation of the
bedding with respect to the axis of the core;
whereas, with nonpenetrating fluids, the frac-
tures tended to be parallel with the axis of the
core. Reynolds et al. (1954) described addi-
tional experiments confirming the earlier work
of Scott et al. (1953) and concluded that it
should be possible to produce horizontal frac-
tures with penetrating fluids and vertical frac-
tures with nonpenetrating fluids.
2 3 9
240 M. King Hubbert and David G. Willis
L E A S T P R I N C I P A L
< S T R E S S
k.
FIG. 1Stress element and preferred plane of fracture.
During the last 2 years, the writers have been
engaged in a critical reexamination of this
problem, and, because the results obtained have
sustained the conjecture offered earlier by Hub-
bert (1953), the principal content of this paper
is an elaboration of that view.
STATE OF STRESS UNDERGROUND
The approach commonly made to the
the problem of underground stresses is to as-
sume that the stress field is hydrostatic or
nearly hydrostatic, the three principal stresses
being approximately equal to one another and
to the pressure of the overburden. That this as-
sumption cannot generally be true is apparent
from the fact that, over long periods of geo-
logic time, the earth has exhibited a high degree
of mobility whereby the rocks have been re-
peatedly deformed to the limit of failure by
faulting and folding. In order for such defor-
mation to occur, substantial differences be-
tween the principal stresses are required.
The general stress condition underground is
therefore one in which the three mutually per-
pendicular principal stresses are unequal. If
fluid pressure were applied locally within rocks
in this condition, and if the pressure were in-
creased until rupture or parting of the rocks re-
sulted, that plane along which fracture or part-
ing would first be possible would be the one
perpendicular to the least principal stress. It is
here postulated that this plane is also the one
along which parting is most likely to occur
(Fig. 1).
In order, therefore, to have a mechanical ba-
sis for anticipating the fracture behavior of the
rocks in various localities, it is necessary to
know something concerning the stress states
that can be expected. The best available evi-
dence bearing upon these stress conditions is
the failure of the rocks themselves, either by
faulting or by folding.
The manner in which the approximate state
of stress accompanying various types of geo-
logic deformation may be deduced was shown
in a paper by Hubbert (1951); the rest of this
section is a paraphrase of that paper.
Figures 2 and 3 show a box having a glass
front and containing ordinary sand. In the mid-
dle, there is a partition which can be moved
from left to right by turning a handscrew. The
white lines are markers of powdered plaster of
paris which have no mechanical significance.
As the partition is moved to the right, a normal
fault with a dip of about 60 develops in the
left-hand compartment (Fig. 2) . With further
movement, a series of thrust faults with dips of
about 30 develops in the right-hand compart-
ment (Fig. 3) .
The general nature of the stresses which ac-
company the failure of the sand can be seen in
Figure 4. Adopting the usual convention of
designating the greatest, intermediate, and least
principal stresses by <T
1(
<J
2
, and <r
3
, respectively
(here taken as compressive), in the left-hand
compartment <r
3
will be the horizontal stress,
which is reduced as the partition is moved to
FIG. 2Sand-box experiment showing normal fault.
LV
FIG. 3Sand-box experiment showing thrust fault.
Mechanics of Hydraulic Fracturing 241
the right, and o^ will be the vertical stress,
which is equal to the pressure of the overlying
material. In the right-hand compartment, how-
ever, <J