Eshelbys Technique for Analyzing Inhomogeneities in

Geomechanics
J. W. Rudnicki
Northwestern University, Evanston, Illinois, USA
Abstract This chapter describes Eshelbys technique for determining the stress
and strain in regions in an innite elastic body that undergo a change of size or
shape. The technique is also extended to determine the stress and strain in regions
of dierent elastic properties than the surroundings due to loading in the fareld.
The chapter also discusses the relation of this technique to singular solutions in
elasticity and dierent integral forms for the solutions. Example applications of
the technique include determining the eective stress in a narrow fault zone, the
stress and strain in a uid reservoir and subsidence due to uid mass injection or
withdrawal. The approach of Eshelby is also used as a basis for a brief description
of representations of deformation due to slip and seismic source theory.
1
Contents
1 Introduction 3
2 Results from Linear Elasticity 3
3 Results from Eshelby 8
3.1 Thin Inclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.2 Volumetric Transformation Strain . . . . . . . . . . . . . . . . . . . . . . . 12
3.3 Uniform Transformation Strain . . . . . . . . . . . . . . . . . . . . . . . . 13
3.4 Ellipsoidal Inhomogeneity . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.5 Poroelastic Inhomogeneity . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4 Applications 16
4.1 Eective Stress in a Narrow Fault Zone . . . . . . . . . . . . . . . . . . . 16
4.2 Fluid Mass Injection or Withdrawal . . . . . . . . . . . . . . . . . . . . . 17
4.3 Subsidence Due to Fluid Mass Injection or Withdrawal . . . . . . . . . . . 21
4.4 Deformation Due to Slip . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Bibliography 24
2
1 Introduction
The stress and strain in reservoirs, aquifers, intrusions, fault zones, caverns or under-
ground structures can dier signicantly from the regional elds because the material
response, pore pressure or temperature diers from that of the surrounding material.
Often it is of interest to determine the alteration of stress, strain or pore pressure in the
inhomogeneity due to remote loading. Examples of this type of problem are the pore
pressure change in an aquifer due to barometric or tidal loading or the stress induced in
a fault zone by tectonic loading. In other cases, the interest may be in how a change
in the inhomogeneity aects other quantities or perturbs the remote eld. For example,
how does the excavation of an underground storage facility perturb the regional stress?
How does injection or withdrawal of uid mass or a change of pore pressure alter the
stress and strain in a subsurface reservoir? Although it is sometimes assumed that the
pore pressure change occurs with no change in stress or with the constraint of uniax-
ial deformation, the propriety of these limiting cases depends on the geometry of the
reservoir and the dierence in material response from the surroundings.
The well-known solution of Eshelby (1957, 1961) has been widely used to study the
eects of inhomogeneities in materials (e.g., Mura, 1987) and provides a relatively simple
yet reasonably general approach for exploring these types of problems. In addition,
the Eshelby framework can be used as the basis of seismic source theory. Eshelby (1957)
originally posed the following problem: Suppose a region of an elastic body, the inclusion,
undergoes a change of size and shape that could be described by homogeneous strain in
the absence of constraint of the surrounding material; what is the resulting state of stress
and strain in the inclusion and the surroundings, the matrix? Eshelby (1957) did not solve
this problem directly but instead used an ingenious series of conceptual steps involving
cutting, transforming and re-inserting the inclusion. He showed that for an ellipsoidal
inclusion in an innite elastic body, the resulting stress and strain in the inclusion are
uniform. He went on to show that this solution also solves the problem of an elastic
inhomogeneity, that is, the perturbation of a remote stress or strain eld by an ellipsoidal
region of dierent elastic properties.
This chapter will describe Eshelbys approach and show how the results can be applied
to several geomechanical problems. The next section briey recapitulates some results
from elasticity that will be used.
2 Results from Linear Elasticity
In linear elasticity the stress
ij
is related to the innitesimal strain
kl
by

ij
= C
ijkl

kl
(1)
Because of the symmetry of
ij
and
kl
, the modulus tensor C
ijkl
satises
C
ijkl
= C
jikl
= C
ijlk
(2)
In addition, the existence of a strain energy function implies the additional symmetry:
C
ijkl
= C
klij
(3)
3
For an isotropic material, the modulus tensor is given by
C
ijkl
= G(
ik

jl
+
il

jk
) + (K 2G/3)
ij

kl
(4)
where G is the shear modulus, K is the bulk modulus, and
ij
is the Kronecker delta (
ij
=
1, if i = j;
ij
= 0, if i = j). The combination K 2G/3 is one of the Lame constants,
denoted here; the other Lame constant is G. Alternatively, the bulk modulus K can be
expressed in terms of G and Poissons ratio as K = 2G(1 +)/3(1 2). Substituting
(4) into (1) yields

ij
= 2G
ij
+ (K 2G/3)
ij
(5)
where =
kk
is the volume strain. The strain
ij
is related to the gradient of the
displacement u
i
by

ij
= (1/2) (u
i
,
j
+u
j
,
i
) (6)
where (...),
i
denotes (...)/x
i
. For quasi-static processes, in which inertia is negligible,
the stress
ij
must satisfy the equilibrium equation

ij
,
i
+F
j
(x, t) = 0 (7)
where the F
j
(x, t) are the components of the body force per unit volume.
Substituting the strain displacement equation (6) into (5) and the result into equilib-
rium (7) yields
(K + 2G/3),
j
+Gu
j
,
kk
+F
j
(x, t) = 0 (8)
The solution for an arbitrary distribution of body force is given by
u
i
(x) =
_
V
g
ij
(x )F
j
(, t)d
3
(9)
where g
ij
(x) is the displacement at x due to the application of a unit force in the j -
direction at the point (Love, 1944; Sokolniko, 1956). In other words, the solution of
(8) for a singular distribution of body force given by
F
Point
j
(x, t) = P
j

DIRAC
(x ) (10)
is
u
i
(x) = g
ij
(x )P
j
(11)
where
DIRAC
(x) is the Dirac delta function dened by the following property
_
V
f()
DIRAC
(x )dV =
_
f(x) if x is in V
0 if x is not in V
(12)
For later reference, note that the strains due to application of a point force are given by

ij
(x) =
1
2
P
k
_
g
ik
x
j
+
g
jk
x
i
_
(x ) (13)
4
and the stresses are given by

ij
(x) = P
m
C
ijkl
g
km
x
l
(x ) (14)
where we have made use of (2).
In a homogeneous material (the C
ijkl
do not depend on position) the argument of g
ij
depends only on the dierence x . Another property of the function g
ij
(x ) that
will be useful later is
g
ij
(x ) = g
ji
( x) (15)
This relation states that the ith component of the displacement at x due to a unit point
force in the jth direction at is equal to the jth component of the displacement at
due to a unit point force in the ith direction at x. The result follows from the elastic
reciprocal theorem (Sokolniko, 1956). Consider the application of two point forces to
the body: P
A
at
A
and P
B
at
B
with corresponding displacement elds u
A
and u
B
.
The reciprocal theorem requires the work of P
A
on u
B
at x =
A
be equal to the work
of P
B
on u
A
at x =
B
.
If the elastic body is isotropic and unbounded, then the function g
ij
is obtained from
the Kelvin solution (Love, 1944; Sokolniko, 1956) and can be written as
g
ij
=
1
4G
_

ij
r
+
1
4(1 )

2
r
x
i
x
j
_
(16)
or
g
ij
=
1
16rG(1 )
_
(3 4)
ij
+
(x
i

i
)(x
j

j
)
r
2
_
(17)
where r = |x |. The expression for g
ij
in an elastic half-space can be determined from
the solution of Mindlin (1936; 1955). For small values of r, the expression for g
ij
for the
half-space approaches (16) and (17), but it contains additional, non-singular terms that
are required to make the traction vanish on the surface of the half-space. Okada (1992)
has given a particularly convenient rearrangment of the half-space solution as the sum
of a term corresponding to the Kelvin solution (16, 17), an image term, and two depth
dependent terms, one of which vanishes on the surface of the half-space.
The point force solution can also be used to generate other singular solutions that
will be useful in later discussions. For example, Figure 1a shows the combination of a
point force of magnitude M/ in the x
1
direction at x = + (/2)e
1
, where e
1
is a
unit vector in the x
1
direction, and a force of equal magnitude acting in the opposite
direction at x = (/2)e
1
. The displacement is
u
i
(x) =
M

{g
i1
[x ( + (/2)e
1
)] g
i1
[x ( (/2)e
1
)]} (18)
Taking the limit 0 yields the displacement due to a force dipole or double force of
magnitude M in the x
1
direction
u
i
(x) = M
g
i1

1
(x ) = M
g
i1
x
1
(x ) (19)
5
Combining double forces in three perpendicular directions (Figure 1b) yields the dis-
placement eld of a center of dilatation:
u
i
(x) = M
g
ik

k
(x ) = M
g
ik
x
k
(x ) (20)
M / l M / l
x
x
2
x
1
x
3
l /2 l /2
x
1
x
2
x
3
M / l
M / l
M / l
x
l /2
l /2
l /2
(a) (b)
Figure 1. Illustration of (a) a force dipole with axis in the x
1
direction and (b) the
combination of three perpendicular force dipoles to form a center-of-dilatation.
Because the forces in Figure 1 are arranged in pairs with a common line of action
they combine to give no resultant moment in addition to no resultant force. In contrast,
the combination of forces in Figure 2a results in a counter-clockwise moment about the
x
3
axis of magnitude M. In the limit 0, the displacement eld is
u
i
(x) = M
g
i2

1
(x ) = M
g
i2
x
1
(x ) (21)
The combination of point forces in the positive and negative x
1
directions in Figure 2b
also results in a couple of magnitude M about the x
3
axis but in the clockwise direction.
Combining these two gives the displacement eld
u
i
(x) = M
_
g
i1

2
+
g
i2

1
_
(x ) (22)
Although this combination of forces is called a double couple, in fact, it results in no net
force and moment (Similarly, a double force results in no net force). Later, Eshelbys
result will be used to show that the displacement eld is the same as that due to point
slip at x = . To generalize this procedure, consider the displacement eld due to the
sum of a point force P at + /2 and an oppositely directed point force at /2
6
x
1
x
2
x
3
M / l
M / l
l/2
l/2
(a)
M / l
M / l
l
x
1
x
2
x
3
x
x
(b)
Figure 2. Combination of point forces to give a counterclockwise (a) and clockwise (b)
moment about the x
3
axis.
(Figure 3). Taking the limit || 0, but maintaining the product M
kl
P
k

l
nite
gives
u
i
(x) = M
kl
g
ik

l
(x ) = M
kl
g
ik
x
l
(x ) (23)
P ( + /2) x l
l
P ( - /2) x l
x
x
1
x
2
x
3
Figure 3. Combination of two point forces to form force dipoles and couples.
7
If the forces are arranged in pairs so that there is no net force or moment, then
M
kl
= M
lk
and (23) can be written as
u
i
(x) =
1
2
M
kl
_
g
ik

l
+
g
il

k
_
(x ) (24)
The same expression with a minus sign applies if dierentiation is with respect to x
l
.
3 Results from Eshelby
Now, consider the problem addressed by Eshelby (1957): The elastic eld caused by a
change of size and shape of a region, the inclusion, in an isotropic elastic body. The
treatment here draws heavily on the exposition of seismic source theory by Rice (1980).
He treats the dynamic problem so that the inertia terms are included in (7) and the
function corresponding to g
ij
in (11) also depends on the dierence between the current
time and the time at which the point force is introduced. The development here is for
the quasi-static case which Rice (1980) discusses as a special case of the dynamic result.
The sequence of operations used by Eshelby (1957) to solve this problem is shown in
Figure 4. First, imagine cutting the region V , the inclusion, free from the elastic body.
Since the body is stress-free, this can be done without introducing any body forces or
tractions.
inclusion
matrix
Change size
and shape
f
*
i
e
*
ij
Remove inclusion
from matrix
Reinsert
Restore
size and
shape
Figure 4. Sequence of cutting and welding operations used by Eshelby (1957) to deter-
mine the elastic eld of an inclusion.
8
Next, alter the size and shape of the region V by application of the strain eld

kl
()
(Eshelby (1957) considered this strain to be uniform, a case we will discuss in more detail
below.) Physical examples of

kl
() include a phase change, shrinkage due to cooling of
a region of a thermoelastic solid, or expansion due to injection of uid into a region of a
poroelastic solid. Consequently,

kl
() is called the transformation strain.
The change of size and shape described by the transformation strain occurs without
stress and, hence, both the inclusion and the elastic body remain free from stress. Now,
restore the region V to its original size and shape by the application of a stress eld

ij
() = C
ijkl

kl
() (25)
In order for this stress eld to be in equilibrium, it is necessary to apply simultaneously
a body force eld F

j
() that satises (7):
F

j
() =

k
[C
kjlm

lm
()] (26)
At this point, the strain is zero everywhere and the stress vanishes outside V . Inside V ,
the stress is

ij
() and there is an extraneous body force eld given by F

j
().
The unwanted body force distribution can be removed by applying its negative and
using (9) to calculate the resulting displacement eld:
u
i
(x) =
_
V
g
ij
(x )

k
[C
kjlm

lm
()] d
3
(27)
The desired solution is the sum of the displacement, strain and stress elds resulting
from (27) and those that existed before the removal of the body force eld. Because
the displacement is zero everywhere prior to the removal of the body force, (27) gives
the displacement for the problem originally posed. The strain eld can be obtained by
dierentiation of (27). The stress outside V can be calculated from the strain but, inside
V , it is necessary to add (25). Thus, the complete stress eld is

ij
(x) =
_
C
ijkl
[
kl
(x)

kl
(x)] , inside V
C
ijkl

kl
(x), outside V
(28)
Because the transformation strain

kl
() vanishes outside the nite region V , it is con-
venient to rewrite (27) as the sum of two terms
u
i
(x) =
_
V

k
[g
ij
(x )C
kjlm

lm
()] d
3
+ (29)
_
V
g
ij

k
(x ) [C
kjlm

lm
()] d
3

Using the divergence theorem to evaluate the rst term on a surface just outside V ,
where

kl
= 0, leaves
u
i
(x) =
_
V
g
ij

k
(x ) [C
kjlm

lm
()] d
3
(30)
9
Because of the symmetry (2) of C
kjlm
, the integrand can here be written in the form
(24) with M
kl
= C
klmn

mn
. Thus, the displacement is due to a distribution of double
forces without moment over the volume V . Another intepretation of (30) is obtained by
using (2), (3) and the result from the reciprocal theorem (15) to rewrite (30) as follows:
u
i
(x) =
_
V
_
C
lmjk
g
ji

k
( x)
_

lm
()d
3
(31)
Comparison with (14) reveals that the term in brackets in the integrand is the stress at
due to a unit point force in the ith direction at x:

(i)
lm
( x) = C
lmjk
g
ji

k
( x) (32)
Thus, (31) can be written as
u
i
(x) =
_
V

(i)
lm
( x)

lm
()d
3
(33)
Consequently, the result (31) can be interpreted as following from the equality of work
expressions:
_
V
P
i

DIRAC
( x)u
i
()d
3
=
_
V
P
i

(i)
lm
( x)

lm
()d
3
(34)
The left term is the work of the point force P
i
at x on the displacement at due to
the transformation strain

lm
(). The right side is the work of the stress at due to the
point force P
i
at x on the the transformation strain

lm
(). The left side is dened for the
entire body, not just the volume V and since the transformation strain vanishes outside
V , both integrals can be extended to the entire body. Thus, the symmetry (3) and the
constraint of the reciprocal theorem on the form of g
ij
(15)lead to an intepretation of
the Eshelby result (30) in terms of the reciprocal theorem.
3.1 Thin Inclusions
Often reservoirs or fault zones are very thin, i.e., they have a thickness much less
than their lateral extent. Consequently, it is useful to specialize (30) to this case. For
simplicity, consider a narrow volume V with midsurface perpendicular to the x
3
(
3
)
direction. Because the thickness of the zone is assumed to be small, the displacement
near
3
= 0 can be written as
u

i
() =
b
i
(
1
,
2
)
h(
1
,
2
)

3
(35)
10
where h(
1
,
2
) is the thickness and the b
i
(
1
,
2
) are the net relative displacements across
the zone. The strains are

33
=
b
3
(
1
,
2
)
h(
1
,
2
)
(36a)

3
=
1
2
_
b

(
1
,
2
)
h(
1
,
2
)
+
3

_
b
3
(
1
,
2
)
h(
1
,
2
)
__
(36b)

=
1
2

3
_

_
b

(
1
,
2
)
h(
1
,
2
)
_
+

_
b

(
1
,
2
)
h(
1
,
2
)
__
(36c)
where (, ) = (1, 2). Substituting into (30), expanding other terms in the integrand
about
3
= 0, integrating
3
from h/2 to +h/2 and taking the limit h 0 yields
u
i
(x) =
_

g
ij

k
(x )C
kj3m
b
m
()d
1
d
2
(37)
Generalizing the result to a surface with unit normal n (pointing from the () to the
(+) side of ) yields
u
i
(x) =
_

g
ij

k
(x )C
kjlm
1
2
(n
l
b
m
+ n
m
b
l
)()d
1
d
2
(38)
The same expression can be obtained directly from (30) by writing the transformation
strain as

ij
() =
1
2
(n
i
b
j
+ n
j
b
i
)()
DIRAC
() (39)
where
DIRAC
() is the singular Dirac delta function converting the domain of the
integral from the volume V to the surface .
Comparison of the integrand in (38) with (24) shows that the displacement is due
to a distribution of double forces without moment with strengths C
kjnm
n
n
b
m
over the
surface . The same manipulations that lead from (30) to (31) and use of (2), (3), and
(15) allow us to rewrite (38) as
u
i
(x) =
_

_
C
lmkj
g
ji

k
( x)
_
1
2
(n
l
b
m
+ n
m
b
l
)()d
1
d
2
(40)
Using (32) to again identify the term in square brackets as the stress at due to a unit
point force in the ith direction at x yields the Volterra formula for the displacement eld
due to a distribution of dislocations (displacement discontinuities, b
k
) on (Steketee,
1958):
u
i
(x) =
_

n
l

(i)
lm
( x)b
m
()d
1
d
2
(41)
As noted by Rice (1980), this expression is typically derived beginning with an application
of the reciprocal theorem as expressed in (34).
Specializing (38) to an isotropic material (5) gives
u
i
(x) =
_

_
n
m
b
m
()
g
ik

k
(x ) + Gn
l
b
m
()(
g
il

m
+
g
im

l
)(x )
_
d
1
d
2
(42)
11
If the relative displacement is purely normal to the plane of the zone, opening or overlap,
b
m
() = b
normal
()n
m
and (42) becomes
u
i
(x) =
_

b
normal
()
_

g
ik

k
(x ) + 2Gn
l
n
m
_
g
il

m
_
(x )
_
d
1
d
2
(43)
Thus, the source on is the combination of a center-of-dilatation, weighted by , and
a force dipole in the direction n weighted by 2G. If the displacement is evaluated far
from the surface (compared with a characteristic length of ) then the term in square
brackets in the integrand can be evaluated at any point on and taken outside the
integral. The remaining integration is equal to the average relative normal displacement
over multiplied by the area of . This justies identication of the quantity in square
brackets as a point uniaxial strain (or, less accurately, a tensile) source. If the relative
displacement is perpendicular to the normal, that is, pure slip, n
m
b
m
= 0, and (42)
reduces to
u
i
(x) =
_

Gn
l
b
m
()
_
g
il

m
+
g
im

l
_
(x )d
1
d
2
(44)
Evaluation of the displacement far from leads to a double couple source (without
moment) of strength equal to the product of G, the average slip and the area of .
3.2 Volumetric Transformation Strain
For later reference, we record the forms of the preceding expressions when the trans-
formation strain is purely volumetric

lm
() =
1
3

()
lm
(45)
If the material is isotropic (5), then
C
kjlm

lm
() = K

()
kj
(46)
With (46), (30) becomes
u
i
(x) =
_
V
K

()
g
ik

k
(x )d
3
(47)
Comparison of (47) with (20) demonstrates that the displacement is the result of a
distribution of centers of dilatation of strength K

() over the volume V . When the

transformation strain results from a temperature change, (47) reects the nuclei of strain
method introduced by Goodier (Timoshenko and Goodier, 1970).
Alternatively, substituting (45) into (33) yields
u
i
(x) =
_
V
1
3

(i)
kk
( x)

()d
3
(48)
This is a form of the Maysel integral representation (Maysel, 1941) (also see Nowacki
(1986)) discussed by Lehner et al. (2005). The result was used by Maysel to determine the
12
displacement eld due to a temperature change in a region, but is adapted by Lehner et al.
(2005) to solve an analogous problem of poroelasticity corresponding to uid withdrawal
or injection from a region. Although (48) and (47) are equivalent, Lehner et al. (2005)
note that the former has certain advantages for computations in inhomogeneous media.
They obtain (48) from an adaption of the reciprocal theorem to a poroelastic solid and its
application analogous to (34). Although obtained by a dierent route here, the reciprocal
theorem is used via the property (15).
3.3 Uniform Transformation Strain
In the problem originally considered by Eshelby (1957), the transformation strain

kl
() was taken as uniform,
T
ij
. Consequently, the size and shape of the inclusion (after
transformation and before re-insertion into the matrix) can be restored by a traction
applied to the surface:
t
T
j
= n
k
C
kjlm

T
lm
(49)
Thus, the displacement caused by the removal of this tractions is that due to a distribu-
tion of point forces over the boundary of the inclusion
u
i
(x) = C
kjlm

T
lm
_
S
n
k
g
ij
(x )d
2
(50)
The strain, obtained by dierentiation, is

rs
(x) = C
kjlm

T
lm
_
S
n
k
1
2
_
g
rj
x
s
+
g
sj
x
r
_
(x )d
2
(51)
The same result is obtained by recognizing that the derivative appearing in the integrand
of (27) is zero everywhere except on the boundary of V . On the boundary, the derivative
is singular as the

ij
() drops from its uniform value,
T
ij
inside V to zero outside and
can be represented formally as

_
C
jkl

T
kl
_
= n

C
jkl

T
kl

DIRAC
(S) (52)
where
DIRAC
is again the Dirac delta function, S is the surface bounding V , and the
negative sign appears because the strain decreases upon crossing S in the direction of the
outward normal with components n
i
. Substitution into (27) changes the volume integral
to a surface integral and reduces this expression to (50).
If the body is unbounded and isotropic and the volume V is an ellipsoid, then Eshelby
(1957) showed that the integrals in (51) can be expressed explicitly in terms of elliptic
integrals. The resulting constrained strain in the inclusion (i.e., that which results when
the transformation strain occurs under the constraint of the surrounding material) is
uniform and can be expressed as

mn
= S
mnkl

T
kl
(53)
The array of factors S
ijkl
depend only on Poissons ratio and the geometry of the ellipsoid.
(Eshelby (1957) also notes that the strain is uniform in an inclusion in an anisotropic
13
elastic solid.) General expressions for the S
ijkl
are given by Eshelby (1957) and by Mura
(1987). In axes coinciding with the principal axes of the ellipsoid, the only nonzero entries
are of the form S
iijj
or S
ijij
and the array is symmetric with respect to interchange of the
rst and last two indicies, but not with respect to interchange of the rst pair and last
pair. Rudnicki (1977) gives expressions and tabulates some results for the S
ijkl
for oblate
and prolate axisymmetric ellipsoids, Rudnicki (2002a) gives extensive graphical results
for oblate axisymmetric ellipsoids, and Mura (1987) gives expressions for axisymmetric
and cylindrical shapes.
3.4 Ellipsoidal Inhomogeneity
For uniform transformation strain of an ellipsoidal region in an innite elastic body,
the constrained strain is given by (53). The stress inside the inclusion, from (28), is

ij
= C
ijkl
_

kl

T
kl

(54)
Eliminating the transformation strain between (53) and (54) yields

ij
= C
ijkl
_

kn

lm
S
1
klmn

mn
(55)
The relation (55) reects the constraint of the elastic material exterior to the inclusion.
Stated dierently, if an ellipsoidal cavity is loaded by a traction n
i

ij
derived from the
uniform stress
ij
, then the resulting strain of the cavity boundary is compatible with
the uniform strain
mn
where
ij
and
mn
are related by (55). Because the strain of the
interior is uniform, the displacement of the boundary is
mn
x
n
.
If the solid is loaded by fareld stresses

ij
and strains

ij
related by

ij
= C
ijkl

kl
(56)
then
ij
and
ij
in (54) and (55) should be replaced by the dierences
I
ij

ij
and

I
ij

ij
where the superscript I denotes the value in the inclusion. Thus, (55) becomes

I
ij

ij
= C
ijkl
_

kn

lm
S
1
klmn
_

I
mn

mn
_
(57)
Note that
I
ij
and
I
mn
are not necessarily related by (1) and that (57) expresses the
constraint of the elastic material surrounding the ellipsoidal region. Consequently, the
elastic inclusion can be replaced any homogeneous material with the stress related to the
strain by
I
ij
= F
ij
_

I
mn

and the stress and strain at the boundary are required to satisfy
(57). In other words, the transformation strain
T
kl
, which has been eliminated from (57)
can always be chosen so that the constitutive relation of the material in the ellipsoidal
region can be satised. For example, (57) solves the problem of the deformation of an
ellipsoidal cavity in an innite elastic solid loaded in the fareld by

ij
. In this case,

I
ij
= 0 since the boundary of the cavity is traction-free and the displacement of the
boundary is given by u
m
=
I
mn
x
n
(and

mn
is related to

ij
by (56)). As another exam-
ple, Rudnicki (1977) has used an incremental version of (57) with a pressure-sensitive,
dilatant, strain-softening elastic-plastic relation in the ellipsoidal region as a model of
14
the evolution of a fault zone due to the slow increase of far-eld strains. In particular,
he examines the relation between the point at which conditions for localization of de-
formation are met in the elliposoidal region and the occurrence of an inertial instability
corresponding to the onset of a seismic event.
3.5 Poroelastic Inhomogeneity
Rudnicki (2002a) has examined in detail the case in which the ellipsoidal region is
poroelastic with a constitutive relation of the form

I
ij
= C
I
ijkl

I
kl
+
I
p
I

ij
(58)
where p is the alteration in pore uid pressure, C
I
ijkl
is the matrix of drained (correspond-
ing to zero change in pore pressure) moduli and
I
is a porous media constant. This
constant is often denoted and is equal to 1 K
I
/K

s
, where K
I
is the bulk modulus
of the porous inclusion material and K

s
is an additional bulk modulus. Under circum-
stances discussed by Rice and Cleary (1976) and Wang (2000), K

s
can be identied with
the bulk modulus of the solid constituents, K
s
, and, more generally, is of the same order.
The identication K

s
with K
s
is equivalent to the requirement that the relative change
in pore volume fraction is equal to the bulk dilatation of the solid phase (see eqn. (85)
of the Chapter by Lehner). Although temperature changes are not considered here, they
could be included by appending the term K
I

ij
to the right hand side of (58), where

s
is the cubical thermal expansion coecient and is the temperature change. Alter-
natively, (58) could be expressed in terms of the undrained moduli (corresponding to the
response for zero change in pore uid mass) and the alteration of uid mass. Because
the ellipsoidal region must be homogeneous, the pore pressure must be uniform. This
will be a good approximation if the ellipsoidal region is much more permeable than the
surroundings. In addition, the relation (57) is based on purely elastic response in the
region surrounding the ellipsoid and, hence, does not account for exchange of uid with
the ellipsoidal region. Thus, the region surrounding the ellipsoid is assumed to deform
under either drained or undrained conditons. These assumptions are, however, reason-
able for application to reservoirs, aquifers, caverns, and, possibly, fault zones. Rudnicki
(2002a) has given extensive graphical results for the eects of geometry and material
mismatch on the stresses induced by uid injection or withdrawal.
It is important to keep in mind that even if the inhomogeneity is ellipsoidal, but
embedded in a bounded region, the constrained strain will not be uniform. In this
case, the simple result (57) relating dierences in stress and strain will not apply. For
many of the applications mentioned in the preceding paragraph, the idealization of an
inhomogeneity embedded in a half-space (rather than an innite space) will be a better
approximation. Nevertheless, the results for the innite space are attractively simple and
will be a good approximation as long as the inhomogeneity is not too close to the surface
of the half-space.
If the ellipsoidal region is poroelastic, then (58) and (56) can be used to eliminate
either the stresses or strains from (57). The result is a relation between the stress (or
strain) and pore pressure (or uid mass change) of the ellipsoidal region in terms of the
15
fareld stress (or strain). For isotropic response, the relation between the strains is

I
mn
+ S
mnkk
e
I
_
K
I
K

G
I
G
_
+
_
G
I
G
1
_
S
mnkl

I
kl
=

mn

1
3

I
p
I
K
S
mnkk
(59)
where
I
=
I
kk
, G
I
and K
I
are the shear modulus and drained bulk modulus of the
inclusion and the same symbols without subscripts are the moduli of the surrounding
material. If the strains are eliminated, the result can be written as the following two
equations:
s
I
mn
+
_
G
I
G
1
__
S
mnkl

S
ppkl
[(K
I
/K 1) S
mnqq
+
mn
]
3 [1 + (K
I
/K 1)]
_
s
I
kl
(60)
=
G
I
G
s

mn

2G
I
3K
[S
mnkk

mn
]
[1 + (K
I
/K 1)]
_

I
p
I
+
_
K
I
K
1
__
and

I
_
1 +
_
K
I
K
1
__
+
_
G
I
G
1
_
K
I
2G
I
S
mnkl
s
I
kl
=
K
I
K

+ (1 )
I
p
I
(61)
where the stress has been separated into spherical and deviatoric parts
ij
= s
ij
+(/3)
ij
and 3 = S
ppqq
= (1 +

)/(1

).
4 Applications
4.1 Eective Stress in a Narrow Fault Zone
Cocco and Rice (2002) have examined whether the pore pressure induced in a narrow,
at fault zone for undrained conditions (no change in uid mass content) is determined
by the mean stress or the normal stress. Here, we reproduce their result from the above
analysis in the limit of a at, ellipsoidal inhomogeneity with normal in the x
3
direction.
In this limit, the only non-zero values of the array S
ijkl
are
S
3333
= 1, S
1313
= S
2323
= 1/2
S
3311
= S
3322
=

/(1

)
(Rudnicki (2002a) gives the expansions including the rst order term in the aspect ratio
of the inclusion.) With these values it is possible to determine from (57) or from the
limits of (59), (60) and (61) that
I
ij
=

ij
whenever i or j is 3 and that
I
ij
=

ij
when
neither i nor j is 3. Cocco and Rice (2002) deduce these results directly from equilibrium
and compatibility. The requirement that
I
33
=

33
, which appears as the third of (19)
in Rudnicki (2002a) can be written in the more compact form
M
I

I
33
= M

33
+ (

11
+

22
)(

I
) +
I
p
I
(62)
where M = 2G(1 )/(1 2) is the elastic modulus for one-dimensional strain and
= 2G/(1 2) is a Lame constant. Specializing (61) for a at axisymmetric ellip-
soid, setting p
I
= 0, consistent with taking the elastic moduli of the inhomogeneity to
16
be undrained values, and using
I
33
=

33
yield the mean stress in the inhomogeneity

I
kk
/3 in terms of the remote normal stress

33
and the remote mean stress

kk
/3. For
undrained conditions, the pore pressure induced in the fault zone is equal to B
I

I
kk
/3,
where B
I
is Skemptons coecient in the fault zone (Cocco and Rice, 2002):
p
I
= B
I
K
I
M
I
_
G
I
G

kk
/3 +
_
1
G
I
G

33
_
(63)
As noted by Cocco and Rice (2002), the pore pressure in the fault zone is determined
by the normal stress

33
if the shear modulus of the fault zone is much less than that
of the surroundings (G
I
G

). On the other hand, the pore pressure is determined

by the mean normal stress

kk
/3, if the shear moduli are nearly the same (G
I
G

).
Cocco and Rice (2002) also note that for anisotropy so pronounced that uid pumping
causes strain only in the x
3
direction when the fault is held at constant stress, the pore
pressure is determined solely by the normal stress.
4.2 Fluid Mass Injection or Withdrawal
Determining the stress and strain changes due to the injection or withdrawal of uid
mass is one of the fundamental problems of aquifer or reservoir management. Rudnicki
(2002a) has used the Eshelby approach to examine in detail the eects of geometry and
elastic mismatch on the stresses and strains induced by pressure or uid mass changes
in an ellipsoidal region embedded in an unbounded linear elastic solid. He assumes that
the uid mass or pressure changes occur on a time scale much shorter than that for
changes of regional stress or strain and, hence,

ij
and

ij
may be taken as zero. For
an axisymmetric ellipsoidal region with the x
3
direction distinguished, Rudnicki (2002a)
gives the ratio of lateral strain change to axial strain change, R =
I
11
/
I
33
, due to a
pressure change p
I
, as
R =
3 S
33kk
+ (G
I
/G1) {3S
3333
S
33kk
S
pp33
}
2S
33kk
+ (G
I
/G1) {3S
3333
S
33kk
S
pp33
}
(64)
where the expressions for the relevant S
ijkl
in this case are given by
= (1/3)S
kkjj
= (1 +

)/3(1

)
S
kk33
= 1
(1 2

)
(1

)
I(e)
S
3333
=
1
2
(1 + S
kk33
)
e
2
(2 3I(e))
2(1

)(1 e
2
)
S
33kk
= 3(1 I(e))
e = c/a is the aspect ratio (ratio of shorter principal axis to longer), and
I(e) =
e
(1 e
2
)
3/2
_
arccos(e) e(1 e
2
)
1/2
_
(65)
The expression (65) corrects Equation (22) of Rudnicki (2002a), which has e
2
in the
numerator of the rst term outside the bracket. (I am extremely grateful to Colin Sayers,
17
Schlumberger, Houston, USA for tracking down this misprint.) Figure 5 (after Figure
4 of Rudnicki (2002a)) plots the strain ratio R against the aspect ratio of the reservoir,
e, for three values of the ratio of the reservoir shear modulus to matrix shear modulus
G
I
/G, 0.5, 1.0, and 2.0. (The plot is for = 0.2, but the dependence on is weak).
Although the strain state induced in the reservoir is often assumed to be uniaxial, Teufel
et al. (1991) have questioned the validity of this assumption. The plot demonstrates
that uniaxial strain (R = 0) is a good assumption for thin reservoirs (e 1), but the
approach to uniaxial strain is slower for larger values of G
I
/G.
0.0 0.2 0.4 0.6 0.8 1.0 1.2
e , Inhomogeneity aspect ratio
-0.1
0.1
0.3
0.5
0.7
0.9
R

=

1
1
/

3
3
G
I
/G

= 0.5
G
I
/G

= 1.0
G
I
/G

= 2.0
Figure 5. Ratio of lateral to axial strain induced by a pressure change in a reservoir
embedded in an innite elastic solid (after Rudnicki, 2002a, Fig. 4).
The axial strain change can be written as (Rudnicki, 2002a, eqns. (26) and (27))

I
33
=
I
p
I
E
3
/K
I
(66)
where
E
3
=
(K
I
/K)
(1 + 2R) [1 + (K
I
/K 1)] + (G
I
/G1) (1 R)(S
kk33
)
(67)
Equations (64), (66) and (67) can be used to calculate the slope of the stress path caused
by injection or withdrawal. In particular, Figure 6 shows a diagram of q versus s
where q = (
3

1
)/2, s = (
3
+
1
)/2,
3
and
1
are the maximum and minimum
principal stresses (positive in compression) and
i
=
i
p is the form of the eective
18
stress appropriate for failure (Rice, 1977; Paterson and Wong, 2005). If injection simply
increased the pore pressure without altering the other stress components, injection would
cause the stress point to move horizontally to the left (s < 0) by an amount equal to
the change in pore pressure. Similarly, withdrawal would cause the stress point to
move horizontally to the right (s > 0). Because the change of pore pressure generally
does alter the other stress components, q = 0 (and, also, s is not simply equal to the
change in pore pressure). The slope q/s is given by (Rudnicki, 2002a, eqn. (31))
q
s
=

I
E
3
(G
I
/K
I
)(1 R)
1
I
{1 [3E
3
/2(1 +
I
)] [1 + R(1 + 2
I
)]}
(68)
In the limit of a very narrow zone (e 1 and R = 0) this ratio becomes
q
s
=
_
2(1
I
)

I
(1 2
I
)
1
_
1
(69)
(This expression corrects Eqn. (32) of Rudnicki (2002a) which contains several misprints.
I am grateful to F. Mulders, Delft Technical University for identifying the errors in this
expression.) Because q and s have the same sign, injection causes the stress point
to move down and to the left and withdrawal causes it to move up and to the right.
Figure 7 (after Fig. 10 of Rudnicki (2002a)) plots values of the slope q/s against the
Poissons ratio of the reservoir for various aspect ratios, G
I
/G = 1.0 and
I
= 0.9. The
line for e = 0 shows the slopes that occur from uniaxial strain (69), but, as Figure 7
shows, the slopes are smaller for reservoirs of nite aspect ratio. The values of all the
slopes decrease with increasing values of the Poissons ratio of the inhomogeneity.
Mulders (2003) has shown that the results of the ellipsoidal inhomogeneity model
agree well with nite element calculations and has used the nite element method to
explore more complex geometries. Holt et al. (2004) has used the model to examine the
eects of stress path, in particular, deviations from the condition of uniaxial strain, on
compaction and pore compressibility within the poroelastic approximation. Holt et al.
(2004) use a discrete element model to explore non-elastic eects.
If a failure criterion is adopted, then the expressions (68) or (69) can be used to assess
in a simple way whether the pressure changes due to injection or withdrawal will move
the stress state of the reservoir closer to or further from failure. Figure 6 illustrates
the situation for a linear Mohr-Coulomb failure envelope. For an inhomogeneity with
short axis in the vertical direction, a normal stress state with a vertical stress
v
=
3
that exceeds the horizontal stress
h
=
1
corresponds to positive q and plots in the rst
quadrant. Since (68) shows that q has the same sign as s, withdrawal will cause the
stress state will move closer to the failure envelope if
q
s
> sin (70)
where is the friction angle. According to this condition, if = 30

, values of the
slope above the line corresponding to sin = 0.5 shown in Figure 7 would be unstable
for withdrawal in the sense of moving the stress toward the failure line; values below
would be stable for injection. If the current stress state is typical of a compressional
19
Failure surface
(injection)
(withdrawal)
Current stress state
s s s = >
v 3 1
s s =
h 3
|Ds|
| | Dq
1
sin f
sin f
1
| | Dq
|Ds|
Extensional regime
Figure 6. Illustration of stress paths and approach to failure by uid injection or with-
drawal.
environment, that is, the horizontal stress
h
=
1
exceeds the vertical stress
v
=
3
(again, for an inhomogeneity with short axis in the vertical direction), then q < 0 for the
current stress and the current stress state plots in the fourth quadrant. The slope of the
failure line in this quadrant is sin . Consequently, withdrawal is always stable and
injection is always unstable. Of course, this analysis assumes linear poroelasticity and is
of little consequence if the current stress is far from the failure envelope by comparison to
the change in stress state. In addition, it is well to reemphasize that the results assume
the region of injection or withdrawal is embedded in an innite space.
4.3 Subsidence Due to Fluid Mass Injection or Withdrawal
In a series of papers, Geertsma (1966, 1973,a) addressed the problem of land subsi-
dence due to uid withdrawal from subsurface reservoirs using a Greens function ap-
proach based on an analogy of poroelasticity and thermoelasticity (Geertsma, 1957).
The approach of Geertsma can be reconstructed by substituting (58), specialized to an
isotropic material, into (7). Because the gradient of the pore pressure now appears as
a body force, the representation (9) can be used to calculate the displacement, strains
and stress. Because the gradient of the pore pressure appears and the alteration in pore
pressure vanishes outside a nite volume, the same manipulations used in going from
20
0.0 0.1 0.2 0.3 0.4 0.5

I
, Inhomogeneity Poisson's Ratio
0.0
0.2
0.4
0.6
0.8
1.0
S
l
o
p
e

o
f

S
t
r
e
s
s

P
a
t
h
,

q
/

s
e = 0.0
e = 0.1
e = 0.2
e = 0.3
e = 0.5
Stable for injection
Stable for withdrawal
sin = 0.5
Figure 7. Plots of the stress path q/s against the inhomogeneity aspect ratio
I
for
various inclusion aspect ratios. For a friction angle = 30

and a normal stress regime,

values above (below) the horizontal line, corresponding to sin = 0.5, move the stress
away from (toward) the failure line.
(27) to (30) can be used to express the displacement as an integration of a distribution
of centers of dilatation given by the pore pressure change. In the corresponding thermo-
elastic result, the temperature plays the role of the pore pressure and for many materials
of practical interest, the constant corresponding to is nearly unity. Segall (1985) and
Mossop and Segall (1997) have also considered subsidence due to uid withdrawal using
the approach of Geertsma and, in addition, have calculated the induced stress changes
and used them to assess eects on seismicity (Segall, 1985, 1989, 1992; Segall et al., 1994).
Both Geertsma and Segall work out results for a at, tabular reservoir geometry and use
the g
ij
for an innite half space (Mindlin, 1936, 1955; Mindlin and Cheng, 1950; Okada,
1992).
Not surprisingly, the same results(though limited to an innite space) are obtained
from the Eshelby framework. Specializing (58) to the isotropic case and taking the trace
yields
= K p (71)
where = (1/3)
kk
is the mean stress, =
kk
is the volume strain, and the superscript
I denoting the inclusion material has been dropped. Setting the stress to zero denes the
21
stress-free strain or transformation strain

resulting from an alteration of pore pressure

= p/K (72)
Substituting into (47) gives the same result as Geertsma and shows that the displace-
ment is the result of a distribution of centers-of-dilatation. (As noted above, 1 in
Geertsmas analysis based on the thermoelastic analogy.)
Recently, Walsh (2002) has reexamined the problem using Eshelbys approach. In
particular, he begins by assuming that the reservoir undergoes a change of uid mass,
and then procedes to follow the Eshelby procedure of transformation, application of
tractions to restore the size and shape, re-insertion into the matrix, and removal of
tractions. As shown by Walsh (2002) and discussed by Rudnicki (2002b), this procedure
arrives at the same result as Geertsmas analysis if the reservoir (inclusion) is poroelastic
and the changes in pore pressure in each step are accounted for. The result, in either
case is that the displacements are given by (47) although a novel aspect of Walsh (2002)s
analysis is that he uses the reciprocal theorem to evaluate expressions for the strain of
the reservoir and the surface subsidence.
Lehner et al. (2005) generalize Walsh (2002)s use of the reciprocal theorem to es-
tablish a reciprocal theorem for a poroelastic solid. They use it to obtain an integral
representation for the displacement due to uid injection or withdrawal and recognize
its analogy to Maysel (1941)s solution of the corresponding thermoelastic problem. As
noted by Lehner et al. (2005) and mentioned earlier in this chapter, the result is equiv-
alent to that obtained by Goodiers (Timoshenko and Goodier, 1970) nucleus of strain
method which has been the predominant one for subsidence problems (and, also, for a
variety of other geophysical problems such as fault slip and magma injection (Segall,
2009)). Lehner et al. (2005) explain, however, that the Maysel representation can be ap-
plied more easily to problems of more general geometries and inhomogeneous materials.
In particular, it oers signicant computational advantages.
The computational advantages of the Lehner et al. (2005) Maysel formulation can be
understood by comparison of (47) and (48). In (47), the term g
ik
(x )/
k
is the
solution for the displacement at x due to a unit center-of-dilatation at , where is the
variable to be integrated over the volume for which the transformation strain is non-
zero. Consequently, it is necessary to re-compute this term whenever the distribution
of transformation strain or the volume of integration changes. In contrast, the term

(i)
kk
()/3 in (48) is the mean stress at due to a point force at x. Since the purpose
is to calculate the displacement at x this term need be computed only once for a given
geometry regardless of changes in the distribution of transformation strain or the volume
over which it occurs.
4.4 Deformation Due to Slip
As noted earlier, the treatment of Eshelbys analysis follows the treatment of Rice
(1980) of seismic source theory. Conversely, the Eshelby framework also provides a
convenient entree to seismic source theory (Aki and Richards, 1980; Rice, 1980; Kostrov
and Das, 1988) and here we briey recapitulate some of the results of Rice (1980). As
noted in connection with (30), the displacement can be regarded as due to the integration
22
of force dipoles and double couples of density C
ijkl

kl
. This motivates the terminology
of seismic moment density tensor for the negative of the right hand side of (25)
m
ij
(x, t) = C
ijkl

kl
(73)
where the transformation strain

kl
is now a function of time in addition to position.
Equation (26) then expresses the body force equivalence of seismic sources. The neces-
sary Greens function, denoted G
ij
(x , t t

), is the solution to (8) with the inertia

terms included on the right hand side for an instantaneous (delta function time depen-
dence) body force given by (10). The dynamic version of (30) includes an additional
superposition over time
u
i
(x, t) =
_
V
_
t

G
ij

k
(x , t t

)m
kj
(, t

)dt

d
3
(74)
As in the static case, the derivatives G
ij
/
k
represent force dipoles and couples. Be-
cause of (2), the seismic moment tensor is also symmetric. As a result (74) can be
written as
u
i
(x, t) =
_
V
_
t

1
2
_
G
ij

k
(x , t t

) +
G
ik

i
(x , t t

)
_
(75)
m
kj
(, t

)dt

d
3

and the sources represented in the integrand are double forces and double couples without
moment.
Although the form of G
ij
(x , t t

) is known for an innite, isotropic elastic

solid and other simple cases, the Earth is strongly nonhomogeneous, anisotropic and,
at least, partially non-elastic. Indeed, the challenge of seismology, stated succintly,
is to use observations of the left side of (75) at points on the surface of the Earth
to infer information about its composition and the nature of the source generating the
displacement. A further complication is that the instrument measuring the displacement
may be accurate only in a range of magnitudes and frequencies.
Despite the absence of sources with a net moment in the representation, the following
quantity is dened as the seismic moment tensor:
M
ij
(t) =
_
V
m
ij
(, t)d
3
(76)
The seismic moment tensor plays an important role in observations of earthquake both
from seismograms and from measurements of surface displacements after earthquakes. If
the material is isotropic and the source is conned to a surface, e.g., a fault, then (39)
and (73) lead to
M
ij
(t) =
_

{n
m
b
m
(, t)
ij
+ G[n
i
b
j
(, t) + n
j
b
i
(, t)]} d
1
d
2
(77)
For shear faulting (no opening) n
m
b
m
= 0; if, in addition, the fault surface is at with
normal in the x
3
direction and slip u = b
1
(
1
,
2
, t) only in the x
1
direction, the only
23
nonzero components of M
ij
(t) are M
13
= M
31
= M where
M(t) = G
_

u(
1
,
2
, t)d
1
d
2
(78)
In an isotropic material, the displacement eld (75) is the sum of u
d
i
(x, t) and u
s
i
(x, t),
the contributions from the dilatational and shear waves, respectively. Expressions for
u
d
i
(x, t) and u
s
i
(x, t) in the fareld are
u
d
i
=

i

k
4rc
3
d
_
V ()
m
kj
(, t r/c
d
+ /c
d
)d
3
) (79)
and
u
s
i
=
1
4rc
3
s
_
1
2
(
ij

k
+
ik

j
)
i

k
_ _
V ()
m
kj
(, t r/c
s
+ /c
s
)d
3
(80)
where r is the distance from a point in the source region to the observation point,
= (x )/r, is the density, c
d
=
_
( + 2G) / is the dilatational wave speed,
c
s
=
_
G/ is the shear wave speed, and the superposed dot denotes the time derivative.
The fareld approximation is accurate for distances r that are much larger than the
characteristic source dimension and 2c/, the wavelength of radiation associated with
the frequency . As discussed by Rice (1980), Kostrov (1970) has shown that the source
can be approximated as a point if the time dierences /c between dierent points
of the source region can be neglected (This imposes a limit on the upper frequency for
which the representation is accurate). In this case, the integrals in (79) and (80) dene
the time dependence of the moment and the displacements can be written as
u
d
i
=

i

k
4r
0
c
3
d

M
kj
(, t r
0
/c
d
) (81)
u
s
i
=
1
4r
0
c
3
s
_
1
2
(
ij

k
+
ik

j
)
i

k
_

M
kj
(, t r
0
/c
s
) (82)
The Fourier transforms of the displacements (81) and (82 are proportional to terms
of the form
exp(r
0
/c)
_

M
kj
()
_
where

M
kj
() is the Fourier transform of M
kj
(t). As explained by Rice (1980), the
short term, seismic motion of an earthquake can regarded as essentially complete for
long periods. Consequently,

M
kj
() can be regarded as the transform of a constant,

M
kj
() M
kj
(t
r
)/, where t
r
is the time at which rupture is completed. Therefore,
the Fourier spectrum of the displacement becomes at at low frequency and its magnitude
is characterized by the moment at the end of rupture. For a double couple source, the
magnitude is the product of the shear modulus, the average slip and the area of the slip
surface (44).
24
Acknowledgement
Thanks to the organizers Yves Leroy and Florian Lehner for inviting me to take
part in the CISM course and to Jean Rudnicki for preparing some of the gures. I am
also grateful to Florian for bringing to my attention the Maysel representation. The
U.S. Department of Energy, Oce of Basic Energy Sciences provided partial nancial
support of some of my own research reported in this chapter.
Bibliography
Aki, K. and P.G. Richards (1980). Quantitative Seismology Vol. 1. W. H. Freeman and
Company, San Francisco.
Cocco, M. and J.R. Rice (2002). Pore pressure and poroelasticity eects in Coulomb stress
analysis of earthquake interactions. J. Geophys. Res. 107(B2), 10.1029/2000JB000138.
Eshelby, J.D. (1957). The determination of the elastic eld of an ellipsoidal inclusion
and related problems. Proceedings of the Royal Society of London A 241, 376396.
Eshelby, J.D. (1961). Elastic inclusions and inhomogeneities. In Progress in Solid Me-
chanics Vol. 2, edited by I.N. Sneddon and R. Hill, pp. 87140, North-Holland, Ams-
terdam.
Geertsma, J. (1957). A remark on the analogy between thermoelasticity and the elasticity
of saturated porous media. J. Mech. Phys. Solids 6, 1316.
Geertsma, J. (1966). Problems of rock mechanics in petroleum engineering. In Proc. 1st
Congr. Int. Soc. Rock Mechanics Vol. 1, 585594, National Laboratory of Civil Engi-
neering, Lisbon.
Geertsma, J. (1973). Land subsidence above compacting oil and gas reservoirs.
J. Petrol. Technology 25, 734744.
Geertsma, J. (1973a). A basic theory of subsidence due to reservoir compaction: The
homogeneous case. Verh. Kon. Ned. Geol. Mijnbouwk. Gen. 28, 4362.
Holt, R. M., O. Flornes, L. Li, and E. Fjr (2004). Consequences of depletion-induced
stress changes on reservoir compaction and recovery. In Rock Mechanics Across Bor-
ders and Disciplines, Proceedings of Gulf Rocks 2004, the 6th North American Rock
Mechanics Symposium, Paper No. 589.
Kostrov, B.V. (1970). The theory of the focus for tectonic earthquakes. Izv. Earth
Physics 4, 84101.
Kostrov, B.V. and S. Das (1988). Principles of Earthquake Source Mechanics. Cambridge
Monographs on Mechanics and Applied Mathematics. Cambridge University Press,
Cambridge.
Lehner, F. K., J. K. Knoglinger, and F. D. Fischer (2005). Use of a Maysel integral
representation for solving poroelastic inclusion problems. In Thermal Stresses 2005,
Proc. 6th Int. Congr. on Thermal Stresses, May 2629, 2005, Vienna, Austria, edited
by F. Ziegler, R. Heuer, and C. Adam, pp. 77-80, Schriftenreihe der TU Wien.
Love, A.E.H. (1944). A Treatise on the Mathematical Theory of Elasticity, 4th edition.
Dover Publications, New York. Republication of 4th (1927) edition published by
25
Cambridge University Press.
Maysel, V.M. (1941). A generalization of the Betti-Maxwell theorem to the case of
thermal stresses and some of its applications (in Russian). Dokl. Acad. Nauk. USSR
30, 115118.
Mindlin, R.D. (1955). Force at a point in the interior of a semi-innite solid. In Proc. 1st
Midwestern Conf. Solid Mechanics, pp. 5659.
Mindlin, R.D. (1936). Force at a point in the interior of a semi-innite solid. Physics 7,
195202.
Mindlin, R.D. and D.H. Cheng (1950). Nuclei of strain in the semi-innite solid.
J. Appl. Physics 21, 926933.
Mossop, A. and P. Segall (1997). Subsidence at The Geysers geothermal eld, N. Cal-
ifornia from a comparison of GPS and leveling surveys. Geophys. Res. Lett. 24(14),
18391842.
Mulders, F.M.M. (2003). Modelling of Stress Development and Fault Slip in and Around
a Producing Gas Reservoir. PhD thesis, Delft University of Technology, Delft, The
Netherlands.
Mura, T. (1987). Micromechanics of Defects in Solids, 2nd revised edition. Kluwer
Academic Publishers, Norwell.
Nowacki, W. (1986). Thermoelasticity, 2nd edition. PWN-Polish Scientic Publishers,
Warsaw; Pergamon Press, Oxford.
Okada, Y. (1992). Internal deformation due to shear and tensile faults in a half-space.
Bull. Seismological Society of America 82, 10181040.
Paterson, M. and T.-f. Wong (2005). Experimental Rock Deformation: The Brittle Field,
2nd edition. Springer-Verlag, Berlin etc.
Rice, J.R. (1977). Pore pressure eects in inelastic constitutive formulations for s-
sured rock masses. In Advances in Civil Engineering through Engineering Mechanics,
pp. 360363. American Society of Civil Engineers, New York.
Rice, J.R. (1980). The mechanics of earthquake rupture. In Physics of the Earths
Interior (International School of Physics Enrico Fermi, Course 78, 1979) edited by
A.M. Dziewonski and E. Boschi, pp. 555649, North Holland Publishing Company,
Amsterdam.
Rice, J.R. and M.P. Cleary (1976). Some basic stress diusion solutions for uid-
saturated elastic porous media with compressible constituents. Reviews of Geophysics
and Space Physics 14, 227241.
Rudnicki, J.W. (1977). The inception of faulting in a rock mass with a weakened zone.
J. Geophys. Res. 82, 844854.
Rudnicki, J.W. (2002a). Alteration of regional stress by reservoirs and other inhomo-
geneities: Stabilizing or destabilizing? In Proc. 9th Int. Congr. Rock Mechanics
Vol. 3, Paris, Aug. 25-29, 1999, edited by G. Vouille and P. Berest, pp. 16291637,
Balkema, Rotterdam.
Rudnicki, J.W. (2002b). Eshelby transformations, pore pressure and uid mass changes,
and subsidence. In Poromechanics II, Proc. 2nd Biot Conf. on Poromechanics,
26
Grenoble, Aug. 26-28, edited by J.-L. Auriault, C. Geindreau, P. Royer, J.-F. Bloch,
C. Boutin, and J. Lewandowska, pp. 307312, Balkema, Rotterdam.
Segall, P. (1985). Stress and subsidence resulting from subsurface uid withdrawal in the
epicentral region of the 1983 Coalinga earthquake. J. Geophys. Res. 90, 68016816.
Segall, P. (1989). Earthquakes triggered by uid extraction. Geology 17, 942946.
Segall, P. (1992). Induced stresses due to uid extraction from axisymmetric reservoirs.
Pure and Applied Geophysics 139, 535560.
Segall, P., J.-R. Grasso, and A. Mossop (1994). Poroelastic stressing and induced seismic-
ity near the Lacq gas eld, southwestern France. J. Geophys. Res. 99, 15,42315,438.
Segall, P. (2009). Earthquake and Volcano Deformation. Princeton University Press,
Princeton. To appear.
Sokolniko, S. (1956). Mathematical Theory of Elasticity, 2nd edition, McGraw-Hill,
New York.
Steketee, J. A. (1958). Some geophysical applications of the elasticity theory of disloca-
tions. Canadian Journal of Physics 36, 11681198.
Teufel, L.W., D. W. Rhett, and H. E. Farrell (1991). Eect of reservoir depletion and pore
pressure drawdown on in situ stress and deformation in the Ekosk eld, North Sea.
In Proc. 32nd U.S. Symposium on Rock Mechanics, pp. 6372, Balkema, Rotterdam.
Timoshenko, S.P. and J. N. Goodier. Theory of Elasticity, 3rd edition, McGraw-Hill,
New York.
Walsh, J. B. (2002). Subsidence above a planar reservoir. J. Geophys. Res. 107(B9),
doi:10.1029/2001JB000606.
Wang, H. F. (2000). Theory of Linear Poroelasticity with Applications to Geomechanics
and Hydrology, Princeton University Press, Princeton.
27