Int[ J[ Solids Structures Vol[ 24\ Nos 23Ð24\ pp[ 3700Ð3732\ 0887

\ Pergamon
Þ 0887 Elsevier Science Ltd[ All rights reserved
Printed in Great Britain
9919Ð6572:87:,*see front matter
PII ] S9919Ð6572"87#99985Ð0

VOLUME AVERAGING\ EFFECTIVE STRESS

RULES\ AND INVERSION FOR MICROSTRUCTURAL
RESPONSE OF MULTICOMPONENT POROUS
MEDIA
JAMES G[ BERRYMAN$
Lawrence Livermore National Laboratory\ P[O[ Box 797 L!199\ Livermore\ CA 83440!8899\
U[S[A[

and

STEVEN R[ PRIDE
Departement de Geomagnetisme et Paleomagnetisme\ Institut de Physique du Globe de Paris\
3\ Place Jussieu\ B78\ Tour 13\ 64141 Paris cedex 94\ France

"Received 19 May 0886 ^ in revised form 02 October 0886#

Abstract*A general volume!averaging technique is used to derive equations satis_ed by the average
scalar stresses and strains in multicomponent porous rock[ The resulting equations are combined
with general thought experiments to produce the e}ective!stress rules that determine the volumetric
changes of the rock induced by changes in the con_ning and ~uid pressures[ The composite porous
material speci_cally treated is an isotropic mixture of two Gassmann materials[ Two distinct cases
are considered depending on whether the grains at the interface between the Gassmann materials
are either "0# welded together "no {{cracks|| can open between the two constituents# or "1# nonwelded
"cracks can open#[ The e}ective!stress laws determine not only the overall volumetric changes of a
given sample "i[e[\ changes in sample volume\ total pore volume\ and ~uid!mass content#\ but
determine as well the changes within each Gassmann component individually[ This additional level
of detail achieved in the analysis is referred to as inversion for the microstructural response[ In the
nonwelded case\ the e}ective!stress law relating the variation of crack porosity with macroscopic
changes in con_ning and ~uid stress can be used to determine optimum strategies for increasing
fracture:crack porosity with applications to reservoir production analysis[ Þ 0887 Elsevier Science
Ltd[ All rights reserved[

0[ INTRODUCTION

Real poroelastic media occurring in nature tend to be very complex materials such as rocks
containing water\ oil\ and:or gas[ These materials are inherently heterogeneous composites
simply because the presence of a solid:~uid mixture\ and the random placement of the
~uids in the pore space[ However\ they will often be multicomponent composites even if
the saturating ~uid is uniform\ because the solid constituents themselves may vary from
quartz to shale to clay\ etc[ Whereas in elastic composites the equations of elasticity are
well accepted and the only issue to be addressed in research on such composites is the
e}ective elastic constants themselves\ the situation is more complicated in such mixed
media[ Because these physical systems exhibit a richer variety of behavior\ di}erent {{e}ec!
tive|| equations must be used to describe their reactions to external stimuli[ Thus\ some
fraction of the e}ort in the _eld of poroelasticity continues to be devoted to derivation and
justi_cation of the equations of motion themselves[ This paper addresses this issue from
the point of view of multiple solid component poroelastic media[
Recent work by Pride et al[ "0881# using volume averaging to derive the form of the
equations of motion for sound traveling through a ~uid!saturated porous medium "Biot\
0851# has been restricted by the assumption that the solid part of the solid:~uid composite
was microhomogeneous\ i[e[\ composed of only a single type of solid constituent[ We call

$ Author to whom correspondence should be addressed[ Tel[ ] 990 814 312 1894[ Fax ] 990 814 312 5896[
E!mail ] berrymanÝs012[es[llnl[gov

3700
3701 J[ G[ Berryman and S[ R[ Pride

such single!solid porous media {{Gassmann materials|| in honor of the famous paper by
Gassmann "0840#\ in which he _rst derived some of the basic and best known results about
such poroelastic media[ In the present paper\ we show how to generalize the results of Pride
et al[ "0881# from Gassmann materials to multicomponent and solid frames and then make
use of these results to establish general e}ective!stress rules for the behavior of such media[
E}ective!stress rules "Nur and Byerlee\ 0860 ^ Berryman\ 0881\ 0882\ 0886# determine the
relative importance of con_ning stress and internal pore!~uid pressure on the response of
rock to changes in the state of stress[ By eliminating the restriction to single solid component
media\ we include models that are more representative of real porous materials and rocks[
Some of the earlier approaches to volume averaging include those of Hill "0852#\
Slattery "0856#\ Whittaker "0858#\ Burridge and Keller "0870#\ and Mei and Auriault
"0878#*among many others[ The approach of Pride et al[ "0881#\ which is also the approach
taken here\ has much in common with the methods of Slattery and Whittaker[ However\
these two authors were studying ~uid ~ow through a rigid solid matrix\ whereas the present
approach necessarily includes the e}ects of solid deformation[ Hill "0852#\ Burridge and
Keller "0870#\ Whittaker "0875#\ and Mei and Auriault "0878# have also treated deformable
media[ One of the major di}erences between the present approach and these related methods
is that we make no assumption of periodicity in the medium and we make relatively few
approximations\ letting the volume averaging do most of the work[ The cell problems that
arise in homogenization methods based on periodicity assumptions are replaced here by
thought experiments and:or real laboratory experiments to evaluate the unknown
coe.cients that necessarily arise in all methods[
There has also been some recent and related work on double!porosity\ dual!per!
meability media by Berryman and Wang "0884#\ Tuncay and Corapcioglu "0884#\ and
Wang and Berryman "0885#[ Tuncay and Corapcioglu "0884# use a volume averaging
approach that is basically the same as that of Pride et al[ "0881# to analyze the situation in
which two distinct types of porosity are present in the system ] one due to fractures or
cracks and the other due to matrix porosity[ Berryman and Wang "0884# and Wang and
Berryman "0885# obtain virtually identical results by analyzing the constitutive equations
and comparing them with quantities measurable in the laboratory[ We will _nd here that
combinations of volume averaging and constitutive equations analysis are required to solve
the complex problems we address[
It will prove important to make connection with the de_nitions of Brown and Korringa
"0864#\ which are themselves based on the well!known jacketed and unjacketed thought
experiments of Biot and Willis "0846#[ Generally similar ideas have also been presented by
Rice "0864# and Rice and Cleary "0865#\ but*for our present purposes*we prefer the
more detailed discussion presented by Brown and Korringa[ If the total volume of the
porous sample is V and the pore volume contained in that sample is Vf "where the porosity
is given by f  Vf:V#\ then Brown and Korringa "0864# de_ne constants so that

dV dpd dpf
−  ¦ "0#
V K Ks

and

dVf dpd dpf

−  ¦ [ "1#
Vf Kp Kf

The independent variables in these formulas are the changes in di}erential pressure dpd and
pore!~uid pressure dpf[ The di}erential pressure is the di}erence between the external
"con_ning# pressure dpc and the ~uid pressure\ so dpd  dpc−dpf[ The coe.cients are written
in terms of the frame "or jacketed# bulk modulus K\ the solid grain "or unjacketed# bulk
modulus Ks\ and the pore "or unjacketed pore# bulk modulus Kf[ The remaining modulus
Kp can be shown to be related to K\ Ks\ and the porosity f by the formula
 Volume averaging\ e}ective stress rules\ and inversion for microstructural response 3702

f 0 0
 − \ "2#
Kp K Ks

assuming only than an energy density for the bulk deformations exists[ Measurements of
Kp have been made by Zimmerman et al[ "0875# for some rocks[ One other important factor
is that\ if the porous solid frame is composed of a single constituent "microhomogeneity#\
then*and only then*Kf  Ks  Km\ where Km is the bulk modulus of the single type of
mineral grain present[ These de_nitions and results were used extensively by Berryman
"0881# to establish general e}ective!stress rules for multicomponent rock[
The next section of the paper presents a brief review of the main results of Pride et al[
"0881#\ and relates those results to the de_nitions of Brown and Korringa "0864#[ Sub!
sequent sections of the paper describe our results on rock models with multiple solid
constituents[ Section 2 introduces models with two constituents and addresses both welded
and nonwelded contacts[ Section 3 shows how to write the general e}ective stress for
materials with welded contact[ Section 4 summarizes the main conclusions[ A series of
appendices expands on some of the details of the analysis left out of the main text[
For simplicity and to allow us to focus on the main ideas\ the analysis in the paper is
restricted to statistically isotropic composite poroelastic media containing isotropic solid
constituents[

1[ REVIEW OF SINGLE SOLID COMPONENT MODEL AND RESULTS

The issue addressed by Pride et al[ "0881# concerns the method by which one arrives
at the equations of motion for sound traveling through a solid:~uid mixture when it is
assumed that the solid is cohesive and porous\ but contains only a single type of mineral[
In the present section and throughout the paper\ the ~uid is homogeneous and completely
_lls the pores[

1[0[ The avera`in` theorem

The averaging theorem to be used is due to Slattery "0856# and is based on the idea
that volume averages of derivatives are closely related to derivatives of volume averages\
but care must be taken to account properly for behavior of the averaged quantities at points
or surfaces where abrupt changes occur[ In particular\ when the quantity to be averaged
exists on one side of an interface and does not exist on the other side\ an interior interface
term will contribute to the volume average of the derivative\ but not to the derivative of
the volume average[
Suppose that Q is a quantity to be averaged ^ Q can be a scalar\ vector\ or tensor[ For
convenience of the discussion\ we will assume that the averaging volume is a _nite sphere
centered at position x\ although other choices are also possible[ We label this volume V"x#
and the surface of this volume is 1V[ The exterior surface has two parts 1V  1E9¦1EQ\
with 1E9 being the part where the quantity of interest Q vanishes identically and 1EQ being
the part where Q  9[ In addition to the exterior surface\ there is also an interior surface
where Q changes abruptly to zero "or is simply unde_ned\ such as the value of the solid
displacement in the interior of a pore# and we label this surface 1IQ\ for interior[ The
exterior surface together with this interior surface is the total bounding surface for the
region we will label VQ\ i[e[\ the region wherein the quantity to be averaged Q is nonzero[
With these de_nitions\ it is straightforward to show that

gV
9Q d2 x 
gVQ
9Q d2 x 
g
1EQ
n¼Q Q dS¦
g 1IQ
n¼Q Q dS\ "3#

where dS is the in_nitesimal of the surface volume element\ and n¼Q is the unit outward
normal vector from the region containing nonzero Q[ The main point of "3# is just that
1EQ¦1IQ is the entire bounding surface of Q in the volume V[ As an example of the meaning
3703 J[ G[ Berryman and S[ R[ Pride

of this result\ consider Q to be a vector quantity\ take the trace of "3#\ and the result is just
a statement of the well!known divergence theorem for vectors[
The second result is that

9
g
V
Q d2 x  9
g
VQ
Q d2 x 
g
1EQ
n¼Q Q dS[ "4#

The resulting Frechet derivative "4# follows from the fact that the volumes V"x# and
V"x¦dx# contain virtually the same internal surface and so these do not contribute to the
gradient\ only the surface average of Q over EQ matters " for example\ the hatched areas in
Fig[ 0#[
Combining these results _nally gives

g1EQ
n¼Q Q dS  9
gV
Q d2 x 
gV
9Q d2 x−
g 1IQ
n¼Q Q dS[ "5#

Dividing by the _xed initial volume V  ÐV d 2x contained in V and de_ning the average as
ðQŁ  V−0 ÐV Q d 2x\ then gives the averaging theorem ]

g
0
9ðQŁ  ð9QŁ− n¼Q Q dS[ "6#
V 1IQ

One further de_nition is required to understand the notation to be used for the single
solid analysis[ The average ðQŁ is an average over the whole volume of V "i[e[ including

Fig[ 0[ Schematic illustration for averaging for solid material B in an inhomogeneous ~uid:solid
mixture containing solids of type!A and B and also containing saturating ~uid F[ When the averaging
sphere "circle with lighter line# is displaced a very small distance\ it produces contributions to the
Frechet derivative of physical quantities associated by B material along the external B interfaces
"labeled by 1EB#[ The internal interfaces "1IA\ 1IF# do not contribute because\ as the sphere dis!
placement becomes in_nitesimally small\ the interfaces contained in both the original sphere and in
the displaced sphere are virtually identical[ For the volume averaging step alone\ no restrictions to
Gassmann porous media needs to be made in the analysis[ That restriction arises in the thought
experiments required to close the system of equations[
 Volume averaging\ e}ective stress rules\ and inversion for microstructural response 3704

the regions where Q is zero by de_nition#\ while we will also want to consider the partial
average Q¹ \ which are related to the full volume average by

ðQŁ  v¹Q Q
Þ[ "7#

In eqn "7#\ v¹Q is the volume fraction of V in which Q is nonzero[

Note that\ although we generally neglect to show this dependence\ all the average
quantities are in fact functions of the particular choice of averaging volume V"x#[ In
principle\ V"x# can be as large as the sample being studied\ or as small as desired[ The
legitimacy of the averaging theorem does not depend on the size of the averaging volume[
However\ some intermediate choice will generally be made for V"x#[ Too small of an
averaging volume implies rapid ~uctuations in the quantities of interest "like the ~uid and
solid dilatations#\ while a very large averaging volume implies all the coe.cients in the
equations are universal constants and therefore prevents us from studying the e}ect of local
inhomogeneities\ which is the main purpose of this paper[ Also\ for a wave problem\ when
the averaging volume becomes as large as the wavelengths\ the oscillatory changes in
particle displacement will tend to average to zero[ Further discussion of the averaging is
presented by Pride and Berryman "0887#[

1[1[ Quasistatic constitutive relations for isotropic materials

Pride et al[ "0881# perform bulk averages on the microscopic stress:strain relations for
a ~uid:solid mixture ^ then\ using the averaging theorem\ they obtain general constitutive
relations for the solid and ~uid stress tensors[ Taking the trace of these equations gives the
following results[ The constitutive relations for dilatations and porosity are

dps df
−  9 = u¹s − "8#
Ks 0−f

and

dpf df
−  9 = u¹f ¦ \ "09#
Kf f

where the partial averages u¹s and u¹f are related to the full volume averages by ðusŁ "0−f#u¹s
and ðufŁ  fu¹f[ One assumption implicit in "8# and "09# is that f changes much more
slowly in space than the displacement variables u¹s and u¹f[ This assumption allows us to
remove the factors involving the porosity from the divergence terms[ Although this assump!
tion is nonessential ðPride and Berryman "0887# derive the general poroelasticity laws
without using this assumptionŁ\ it is expedient for our present purposes[
It is important now to understand the interpretations of all the symbols appearing in
these two equations ðPride and Berryman "0887# derive the exact interpretationsŁ[ First\ the
variable dpf is just the change in the average ~uid pressure throughout the ~uid phase[ The
change in average solid pressure dps is related to the macroscopic con_ning pressure change
dpc by the averaging relation dpc "0−f#dps¦fdpf[ Thus\ dps is just the average change in
solid pressure experienced by the solid[ We view dpc and dpf as the pressures we can control
"i[e[\ the independent variables#\ while the derived quantity "and therefore dependent
variable# dps is the "solid volume# weighted average of the con_ning pressure after sub!
tracting that part of the con_ning pressure supported by the ~uid pressure[ The change in
porosity is given by df[ The porosity change occurs naturally in these expressions because
"see Appendix A#

g g
0 0
df  n¼f = uf dS  − n¼s = us dS[ "00#
V 1If V 1Is

The divergence of the average solid displacement 9 = u¹s is properly interpreted as the
3705 J[ G[ Berryman and S[ R[ Pride

dilatation of the porous solid frame "not the dilatation of the solid alone#[ This interpret!
ation is not obvious\ but it follows from the fact that the term arises from the external
surface integral ðcf eqn "4#Ł

g
0
9 = ðus Ł  n¼s = us dS\ "01#
V 1Es

which is exactly the surface integral needed to de_ne the overall behavior of the porous
solid frame[ Thus\ in terms of the de_nitions of Brown and Korringa "0864#\

dV dpd dpf
9 = u¹s  e− − [ "02#
V K Ks

This interpretation is the same one obtained by Pride et al[ "0881# using a combination of
the standard thought experiments "jacketed and unjacketed# of Biot and Willis "0846#[ To
check that this is so\ we can easily show that

$ %
df 0−f 0
− − "dps −dpf #\ "03#
0−f K Ks

using either approach when a single constituent is present so that the Brown and Korringa
unjacketed constants satisfy Ks  Kf\ as has been assumed for a single solid constituent[
Note that "03# can also be written as

$ %
0−f 0
df  − − dpd \ "04#
K Ks

emphasizing that porosity is constant if di}erential pressure is constant*a general result

for microhomogeneous porous frames\ but not true otherwise[ Thus\ the left!hand side of
"8# is just the solid dilatation dVs:Vs\ while the two terms on the right!hand are
dV:V¦d"0−f#:"0−f#[
Similarly\ it is important to understand that the expression 9 = u¹f is not just a ~uid
dilatation\ but also includes the e}ects of ~uid displacement into and out of the volume[ In
fact\ this is already apparent from "09# since the strict ~uid dilatation satis_es

dVf dpf
−  \ "05#
Vf Kf

yet "09# contains an additional term related to changes in porosity[ The correct physical
interpretation of 9 = u¹f is provided by its relation to the increment of ~uid content

z  f"9 = u¹s −9 = u¹f #\ "06#

where z is de_ned as

0 1
dVf −dVf dV dVf
z f − ¦df "07#
V V Vf

and has the interpretation "Biot\ 0862 ^ Berryman and Thigpen\ 0874# of the relative change
in ~uid mass per unit volume of initial ~uid mass[ Note that "05# and "07# are in agreement
with "06# if the averaging equation "09# is also satis_ed[
 Volume averaging\ e}ective stress rules\ and inversion for microstructural response 3706

The equations "8#\ "09#\ and "02# are su.cient to arrive at the standard form of the
equations relating to e and z to the macroscopic pressures dpc and dpf for a single constituent
porous medium given by

e 0:K 0:Ks −0:K −dpc

0 1 0
−z

0:Ks −0:K 0:K¦f:Kf −"0¦f#:Ks 10 1−dpf
[ "08#

These equations are completely consistent with the results of Pride et al[ "0881# as can be
demonstrated by substituting the de_nitions given above into the formulas "37# and "38#
of Pride et al[ "0881#\ and then doing a straightforward "though somewhat tedious# 1×1
matrix inversion[

2[ MODELS WITH TWO SOLID COMPONENTS

We now consider that the porous sample is a mixture of two di}erent porous materials
denoted by A and B[ "See Fig[ 0[# Both of these A and B components are taken as Gassmann
materials and are assumed to occupy well de_ned "distinct# volumes within the sample so
that VA¦VB  V where V is the total sample "or averaging# volume[ The A and B com!
ponents are su.ciently uniform and of su.cient extent that they can be thought of as
homogeneous porous continua[ The volume fractions vA  VA:V and vB  VB:V are
assumed to be known as are the standard poroelastic properties of the two Gassmann
materials ] these are the porosities "fA\ fB#\ the drained bulk moduli "KA \ KB #\ and the bulk
moduli of the A and B minerals "KA\ KB#[ A single homogeneous ~uid is assumed to saturate
both the A and B components and has a known bulk modulus Kf[ Finally\ the drained bulk
modulus of the total composite sample K is assumed to be known as well[
If increments in con_ning pressure and ~uid pressure are applied to the sample as a
whole\ the central goal in what follows is to de_ne the various volumetric responses of the
sample in terms of the known material properties de_ned above[ For the sample as a whole\
these responses are determined by three Brown and Korringa moduli "K\ Ks\ and Kf#
where K is known but Ks and Kf must be determined[ When the sample is a mixture of
two isotropic Gassmann components\ Berryman and Milton "0880\ 0881# have introduced
uniform strain thought experiments that allow Ks and Kf to be solved for exactly[ We show
here that the volumetric changes of the individual A and B components can also be exactly
determined using the same known properties used to de_ne the overall response[ This is
what we mean by {{inversion for the microstructural response|| and is something that has
not previously been addressed in the literature[
In fact\ we go further and distinguish two cases ] "0# the interface between the A and
B Gassmann components within the sample is welded and "1# the interface is either only
partially welded or nonwelded[ The key distinction between the two cases is that welded
contact between porous constituents implies that no cracks:fractures can open up between
these constituents due to applied temperature or stress\ while cracks:fractures are allowed
with only partially welded contact[ Welded contact may be somewhat easier to analyze\ but
nonwelded contact is expected to be a better model of rocks "Walsh\ 0854#[ In order to
obtain the key results in the case of nonwelded contact\ the response due to temperature
changes will have to be considered[
As a means of simplifying the algebra in the following analysis\ we introduce "see
Appendix A for the de_nition and a discussion# new quantities that we call u½A and u½B\ the
divergences of which are just the dilatations of the corresponding Gassmann components ]
9 = u½A  dVA:VA and 9 = u½B  dVB:VB[ This step is made at this point because it helps to
avoid introducing various terms that would ultimately cancel in the _nal formulas[

2[0[ Fully welded contact

For two porous components A and B in welded contact\ the volume fractions of the
components vA\ vB satisfy
3707 J[ G[ Berryman and S[ R[ Pride

vA ¦vB  0\ "19#

while the overall porosity is given by

f  vA fA ¦vB fB [ "10#

The fractions of the total volume occupied by the solid components are v¹A  vA "0−fA#
and v¹B "0−fB#\ respectively[ The solid components in A and B are individually pure\ so
that each porous constituent may be thought of as a microhomogeneous "or Gassmann#
material[

2[0[0[ Avera`in` equations[ Volume averaging for the three components "two solid and
one ~uid# yields

dpA dfA
−  9 = u½A − \ "11#
KA "0−fA #

dpB dfB
−  9 = u½B − \ "12#
KB "0−fB #

and

dpf df
−  9 = u¹f ¦ [ "13#
Kf f

These three equations should be compared to the single!component results "8# and "09#[
The pressure increment dpA represents the average change in solid pressure throughout the
grains of the porous material A "with an analogous de_nition for dpB#[ To provide some
insight\ notice that the left!hand side of "11# is easily seen to be dðVA "0−fA#Ł:VA "0−fA#
"which is the relative change in type!A solid volume#\ while the right!hand side is
dVA:VA−dfA:"0−fA#[

2[0[1[ Other relationships[ It is useful to think of eqns "11# and "12# as equations for
the changes in the constituent porosities dfA and dfB[ To relate these values to the overall
response\ we need another pair of equations[ First\ note that from "10#

df  vA dfA ¦vB dfB ¦dvA "fA −fB #\ "14#

so we need an expression for the change in vA[ For welded contact\ we obtain such an
expression by noting that by de_nition

VA "0¦9 = u½A #
vA ¦dvA  \ "15#
VA "0¦9 = u½A #¦VB "0¦9 = u½B #

which upon expansion and neglect of second order terms yields

dvA  vA vB "9 = u½A −9 = u½B #\ "16#

and furthermore for welded contact dvB  −dvA[ Note that\ if A and B expand or contract
at the same rate so that 9 = u½A  9 = u½B\ then dvA  9 as expected[
We also want to view the combined solid volume Vs  VA "0−fA#¦VB "0−fB# as a
whole in order to recover Biot|s "or Brown and Korringa|s# macroscopic equations for the
inhomogeneous material[ Then\ it is important to recognize that the solid dilatations must
satisfy
 Volume averaging\ e}ective stress rules\ and inversion for microstructural response 3708

9 = u¹s  vA 9 = u½A ¦vB 9 = u½B \ "17#

and\ similarly\ the solid pressures must satisfy

"0−f#dps  v¹A dpA ¦v¹B dpB [ "18#

Relation "17# may be easily derived by considering the denominator of the right!hand side
of "15#\ whereas "18# is just a statement of force conservation across a material boundary[
Alternatively stated\ both "17# and "18# are direct consequences of the de_nitions of the
volume averages[

2[0[2[ De_nitions of microstructural coef_cients[ Now the issue to be addressed is how

the volume\ porosity\ and average solid pressure of the individual A and B porous materials
change when increments in con_ning and ~uid pressure are applied to the sample as a
whole[ We assume the system is isotropic and linear\ so all variables can be assumed to
depend linearly on small changes in the applied stresses[ Thus\ the changes of the average
solid pressure within the A and B components are taken to be linear functions of the form

"0−fA #dpA  QA "dpc −gA dpf # "29#

and

"0−fB #dpB  QB "dpc −gB dpf #[ "20#

Due to the e}ect of microstructure\ these average solid pressures will di}er\ in general\ from
the average solid pressure throughout the sample as a whole[ Expressing QA\ QB\ gA and gB
in terms of the Brown and Korringa moduli and other known properties is really the key
to this analysis because it will be possible to express the volumetric response of the A and
B components entirely in terms of them and the known properties of the Gassmann
materials[
The A and B dilatations are de_ned by linear response laws of the form

−9 = u½A  SA "dpc −a¹ A dpf # "21#

and

−9 = u½B  SB "dpc −a¹ B dpf #\ "22#

while the A and B porosity changes are de_ned by

−dfA  WA "dpc −xA dpf # "23#

and

−dfB  WB "dpc −xB dpf #[ "24#

Such linear response is always expected "Hill\ 0852 ^ Berryman and Berge\ 0885#\ but not
always possible to determine explicitly in inhomogeneous systems[ Here SA and SB\ for
example\ are constant compliances\ while a¹ A and a¹ B are e}ective stress coe.cients*anal!
ogous to the BiotÐWillis parameter but di}ering from it in general because of the e}ects of
microgeometry in the composite ^ i[e[\ because the average pressure increments throughout
the A and B components will be di}erent\ in general\ from the con_ning pressure dpc applied
to the sample as a whole[ We will call constants such as SA\ SB\ a¹ A\ and a¹ B\ and the analogous
quantities for pressure and porosity changes the {{microstructural response coe.cients||[
3719 J[ G[ Berryman and S[ R[ Pride

There are 01 such microstructural coe.cients to be determined "QA\ QB\ gA\ gB\ SA\ SB\
a¹ A\ a¹ B\ WA\ WB\ xA\ xB#[ Our strategy in what follows is to ] "0# de_ne the A and B micro!
pressure coe.cients in terms of the Brown and Korringa moduli "K\ Ks\ Kf# and other
known constants by combining the volume!averaging results with the overall response laws
of Brown and Korringa ^ "1# introduce the known micro!pressure laws into the A and B
Gassmann laws in order to de_ne the coe.cients in the dilatation and porosity response
laws of the A and B components ^ and\ lastly "2# de_ne the moduli Ks and Kf by means of
the Berryman and Milton "0880# thought experiment[

2[0[3[ General relations for microstructural response[ We now use the averaging equa!
tions to express the micro!pressure coe.cients in terms of the Brown and Korringa moduli[
Equation "18# shows that

vA QA "dpc −gA dpf #¦vB QB "dpc −gB dpf #  dpc −fdpf [ "25#

This equality is completely general for arbitrary values of dpc and dpf\ which implies that
the coe.cients of each pressure change must satisfy the equality separately " _rst set dpf  9\
then set dpc  9#[ The resulting equations are

vA QA ¦vB QB  0 "26#

and

vA QA gA ¦vB QB gB  f[ "27#

Equations "26# and "27# give us two equations for these four unknown coe.cients[
To obtain more equations relating the coe.cients to known quantities\ we consider
the volume averaging eqns "11# and "12#[ Multiplying "11# by v¹A and "12# by v¹B\ adding\
and substituting "14# and "17#\ we have the equation relating the constituent changes in
average pressure to the overall solid dilatation

dpA dpB
−v¹A −v¹B "0−f#9 = u¹s −df[ "28#
KA KB

The right hand side of "28# can be written in terms of macroscopic coe.cients by recalling
that the de_nitions of Brown and Korringa show the porosity change is of the form d"Vf:V#\
which is

0 1
a−f 0 0
−df  "dpc −dpf #¦f − dpf [ "39#
K Kf Ks

Thus\ using "02# and "39#\ we _nd that

dpA dpB dpc fdpf

−v¹A −v¹B − ¦ [ "30#
KA KB Ks Kf

This equation leads to two more equations for the Qs and the gs ]

vA QA vB QB 0
¦  \ "31#
KA KB Ks

and
 Volume averaging\ e}ective stress rules\ and inversion for microstructural response 3710

vA QA gA vB QB gB f
¦  [ "32#
KA KB Kf

Solving the pair of eqns "26# and "31# for the Qs gives

v A QA 0:KB −0:Ks
0 1 0 1
0
 "33#
vB QB 0:KB −0:KA 0:Ks −0:KA

and solving the pair of eqns "27# and "32# for the gs gives

vA QA gA 0:KB −0:Kf
0 1 0 1
f
 \ "34#
vB QB gB 0:KB −0:KA 0:Kf −0:KA

showing that

0:KB −0:Kf
gA  f "35#
0:KB −0:Ks

with a similar expression for gB[ Since Kf  Ks in general\ we see that both gs di}er somewhat
from f[ Thus\ the average pressure response of the solid components due to changes in
external pressure has been completely expressed in terms of the Brown and Korringa
moduli[ "We will _nd the same result holds as well for nonwelded contact[# These are the
key results that when combined with the Gassmann laws of the next section allow all
microstructural coe.cients to be expressed in terms of K\ Ks and Kf and known properties
of the Gassmann materials[
Although not essential to our general program here\ the averaging equations do imply
certain other relations involving the microstructure coe.cients[ For example\ "17# shows
that for the sample as a whole

0
vA SA "dpc −a¹ A dpf #¦vB SB "dpc −a¹ B dpf #  "dpc −adpf #\ "36#
K

where a  0−K:Ks is the BiotÐWillis parameter[ From "36# it follows that\ since dpc and
dpf are independent variables\

0
vA SA ¦vB SB  "37#
K

and

a
vA SA a¹ A ¦vB SB a¹ B  [ "38#
K

Equations "37# and "38# give two equations in these four unknowns[
Now to get relations involving the porosity changes\ we consider eqns "11# and "12#[
Substituting from the de_nitions "21# and "24#\ we obtain

QA
WA "0−fA #SA − "49#
KA

and
3711 J[ G[ Berryman and S[ R[ Pride

QA gA
WA xA "0−fA #SA a¹ A − \ "40#
KA

with similar expressions for WB and xB[

Not all of these results can be independent because equation "28# was also obtained
by taking a linear combination of "11# and "12#[ Indeed\ substituting "49# and "40# into the
porosity eqn "14# just reproduces "37# and "38#[ Considering "37# and "38#\ we have one
system of two equations in four unknowns[ Thus\ the implications of the averaging equa!
tions alone has brought us to the point where just two more equations will allow all the
microstructural coe.cients to be expressed in terms of K\ Ks\ and Kf[

2[0[4[ Gassmann laws for components A and B[ The central assumptions made about
materials A and B to this point have been that they respond linearly and isotropically to
the applied pressure increments[ Furthermore\ in the averaging eqns "11#Ð"12#\ we have
assumed that the minerals comprising the frames of materials A and B are homogeneous[
Thus\ even before introducing the Gassmann laws here\ the analysis has been restricted to
a treatment requiring the A and B components to be isotropic monomineral porous
materials "i[e[\ Gassmann materials#[
Pride and Berryman "0887# have shown that the volumetric response of an isotropic
"but otherwise arbitrary# porous material does not depend on the details of how stress is
applied to the surface of the sample[ The volumetric response depends only on the average
~uid and solid pressures generated throughout the sample[ Since our A and B Gassmann
components are each isotropic\ this says that even though they are entangled within the
sample and have nonuniform stress _elds distributed over their surfaces in general\ their
{{isolated state|| compressibility laws still hold so long as we use the micro!con_ning
pressures dpcA and dpcB de_ned as the average total pressures throughout the A and B
regions\ respectively[ Such micro!con_ning pressures are thus given by

dpcA  "0−fA #dpA ¦fA dpf "41#

 QA dpc ¦"fA −QA gA #dpf "42#

where the microstructural pressure law "29# has been used[ Analagous expressions hold for
dpcB[
To close the system of microstructural coe.cients\ the Gassmann compressibility laws
for the A and B components are now given[ Instead of working with these laws in the form
of "08#\ it is more convenient to write them as

0
−9 = u½A  "dpcA −aA dpf # "43#
KA

$ %
QA "aA −fA ¦QA gA
 dpc − dpf "44#
KA QA #

aA −fA
−dfA  "dpcA −dpf # "45#
KA

$ %
"aA −fA #QA "0−fA ¦QA gA #
 dpc − dpf "46#
KA QA

where aA is again the BiotÐWillis constant aA  0−KA :KA [ Analogous laws hold for the B
component[
We now simply read o} the correspondence between the microstructural laws and the
Gassmann laws "44# and "46# to _nd
 Volume averaging\ e}ective stress rules\ and inversion for microstructural response 3712

QA
SA  "47#
KA

aA −fA
aA  g A ¦ "48#
QA

"aA −fA #QA

WA  "59#
KA

0−fA
xA  gA ¦ "50#
QA

with identical expressions for the B coe.cients[ Since the Qs and gs have already been
obtained\ we now have all microstructural coe.cients uniquely "and exactly# expressed in
terms of K\ Ks\ and Kf[
Note that eqns "37#Ð"40# provide alternative expressions for some of these coe.cients[
In particular\ it is easy to show that

0
SA  "51#
KA "xA −a¹ A #

vA "xA −xB #:KA

a¹ A  xA − \ "52#
"a−xB #:K¦vA :KA ¦vB :KB

with analogous expressions for the B coe.cients " found always by interchanging subscripts
A−B#[ We now proceed to the _nal step which is to obtain Ks and Kf in terms of known
quantities[

2[0[5[ Uniform expansion:contraction thou`ht experiment[ The well!known jacketed

experiment of Biot and Willis "0846# has been implicitly used to de_ne both the overall
frame modulus K and the frame moduli KA and KB of the two Gassmann components[
The unjacketed thought experiment of Biot and Willis "in which pore pressure and con_ning
pressure increments are the same# has also been implicitly used to obtain the Gassmann
laws for dfA and dfB[ However\ the unjacketed experiment is not useful in determining Kf
and Ks when the solid framework of grains consists of two "or more# distinct mineral types[
Fortunately\ Berryman and Milton "0880# have presented a di}erent thought experiment
for two!component media that allows both Ks and Kf to be determined exactly[ ðNorris
"0881# has con_rmed and extended these results using an analogy between the equations
of poroelasticity and thermoelasticity[Ł Due to the terseness with which Berryman and
Milton stated their thought experiment\ the key ideas are reviewed here[
Our A and B components can be viewed as porous continua subject to the laws of
poroelasticity[ Consider a body that is a composite of di}erent porous continua and that
is immersed in a reservoir of uniform con_ning pressure[ After a su.cient time\ the ~uid
pressure throughout the body is a constant[ Both the equilibrium laws of poroelasticity and
the boundary conditions on the body|s surface can be satis_ed by a stress distribution in
which the con_ning stress at all points of the composite is constant and given by the pressure
of the reservoir[ However\ such a uniform stress _eld will only be the solution of the actual
poroelasticity problem if the corresponding strain compatibility equations are satis_ed
along with the welded interface conditions[ With uniform stress\ the strain within each
component of the body is also uniform and thus the compatibility equations are satis_ed[
However\ because the strain in di}erent components will\ in general\ be di}erent\ there is
no way to keep the interfaces welded except for very special geometric arrangements of
the components "e[g[\ concentric layers#[ Thus\ for an arbitrary distribution of di}erent
components\ the total stress _eld will not\ in general\ be uniform throughout the composite
3713 J[ G[ Berryman and S[ R[ Pride

body even though such uniform stress always satis_es both the equilibrium law and the
external boundary conditions[
What Berryman and Milton "0880# noticed is that if there are only two isotropic
components "A and B#\ one can always select the con_ning pressure of the reservoir "with
~uid pressure _xed# such that the total strain of the two components is equal ^ i[e[\ such
that dVA:VA  dVB:VB[ One possible way to achieve this macroscopic condition is if the
poroelastic strain at each and every point throughout the A and B continua is constant[
Such a uniform strain _eld guarantees that the welded interface conditions are satis_ed
and\ in so doing\ corresponds to a total stress _eld that is also uniform[ Since these uniform
strain and stress _elds satisfy all conditions for uniqueness of a solution\ they must be the
actual _elds throughout the body[ Thus\ for arbitrary distributions of the A and B
components\ one _nds that when "and only when# 9 = u½A  9 = u½B\ then dpcA  dpcB  dpc
exactly[ Furthermore\ under these conditions\ the poroelastic stress and strain throughout
the body are constant[ These are the key results of the Berryman and Milton thought
experiment[
The special con_ning pressure at which the strain becomes everywhere uniform is
denoted here as dpc  udpf[ An expression for u is obtained by equating the Gassmann laws
for 9 = u½A and 9 = u½B when dpcA  dpcB  dpc to give

aA :KA −aB :KB

u [ "53#
0:KA −0:KB

The constant Ks "or\ equivalently\ the BiotÐWillis constant a  0−K:Ks# is determined

from the requirement that 9 = u¹s  9 = u½A " 9 = u½B# when dpc  udpf so that

0 1
aA :KA −a:K aB :KB −a:K
u  [ "54#
0:KA −0:K 0:KB −0:K

The combination of "53# and "54# may appear to be too many equations for the same thing\
but in fact any two of the three implies the third\ so the count is correct[
The constant Kf is determined by noting that dvA  9 during uniform strain so that
"14# becomes

df  vA dfA ¦vB dfB "55#

$ %
vA "aA −fA # vB "aB −fB #
 −"u−0# ¦ dpf \ "56#
KA KB

where the Gassmann law "59# was used for dfA and dfB with dpcA  dpcB  udpf[ A second
expression for df is obtained by introducing the uniform strain condition dpc  udpf into
"39# to give

g $ %
f "u−0#"0−f# f¦u−0
− f ¦ − dpf [ "57#
Kf K Ks

These two expressions for df combine to give an exact expression for Kf entirely in terms
of known quantities[
Finally\ the uniform!strain conditions provide alternative expressions for the SA and
SB microstructural coe.cients

u−a u−a
SA  and SB  [ "58#
K"u−a¹ A # K"u−a¹ B #

Substituting "58# into "37# gives

 Volume averaging\ e}ective stress rules\ and inversion for microstructural response 3714

0 vA vB
 ¦ \ "69#
u−a u−a¹ A u−a¹ B

which provides an additional equality "not needed for closure# relating the values of the
microstructural e}ective stress coe.cients a¹ A and a¹ B[ "See also Appendices B\ C and D[#

2[0[6[ Summary of welded contact results[ For the reader|s convenience\ the key results
of the above analysis are now summarized[ The e}ective stress!laws governing the overall
response of a multicomponent porous rock are "Brown and Korringa\ 0864#

0
−9 = u¹s  "dp −adpf # "60#
K c

a−f
−df  "dpc −xdpf # "61#
K

where the e}ective!stress coe.cients a and x are de_ned by

a  0−K:Ks "62#

f:Kf −f:Ks
x  0− [ "63#
"0−f#:K−0:Ks

For the speci_c case where the rock is an isotropic mixture of two Gassmann porous
materials in welded contact\ the Brown and Korringa coe.cient Ks is expressed "Berryman
and Milton\ 0880#

0 0−u u−aA
 ¦ \ "64#
Ks K KA

while the coe.cient Kf can be written "Berryman and Milton\ 0880#

$ %
f f vA "aA −fA # vB "aB −fB # "a−f#
 ¦"u−0# ¦ − [ "65#
Kf Ks KA KB K

In both of these expressions\ the coe.cient

aA :KA −aB :KB

u "66#
0:KA −0:KB

de_nes the pressure ratio dpc  udpf that results in uniform strain throughout the A and B
porous continua[
The microstructural response laws of the Gassmann components are

−9 = u½A  SA "dpc −a¹ A dpf # "67#

−dfA  WA "dpc −xA dpf # "68#

"0−fA #dpA  QA "dpc −gA dpf # "79#

where the coe.cients are

3715 J[ G[ Berryman and S[ R[ Pride

0:KB −0:Ks
vA QA  "70#
0:KB −0:KA

f:KB −f:Kf
gA  "71#
0:KB −0:Ks

QA
SA  "72#
KA

aA −fA
a¹ A  gA ¦ "73#
QA

"aA −fA #QA

WA  "74#
KA

0−fA
xA  g A ¦ [ "75#
QA

Similar expressions for the B component response are obtained by replacing the subscript
A with B throughout these expressions[ Alternative expressions for these coe.cients are
presented in Appendices B and C as well as in the previous sections[
To summarize the approach\ we have ] "0# averaged the ~uid and solid response
throughout the two component body ^ "1# combined the averaging equations with the
Brown and Korringa laws to obtain the micropressure laws for dpA and dpB ^ "2# introduced
the micropressure laws into the A and B Gassmann laws in order to identify the other
microcoe.cients ^ and "3# used the Berryman and Milton "0880# uniform strain thought
experiment in order to _nd the moduli Ks and Kf[

2[1[ Partially welded or nonwelded contact

For two porous components A and B in partially welded or nonwelded contact\ the
volume fractions of the components vA\ vB satisfy

vA ¦vB ¦vC  0\ "76#

where vC is the crack or fracture volume fraction which may possibly be zero prior to the
application of a change in con_ning or pore pressure[ The porosity is now given by

f  vA fA ¦vB fB ¦vC \ "77#

where fA\ fB are again\ respectively\ the porosities of the porous constituents A and B\ but
now a third crack "or fracture# phase is present\ and being pure void it therefore contributes
its entire volume fraction vC to the porosity[
In order to close the system of equations for a composite porous medium including
cracks\ we need three applied _elds ðthis fact is clear from the earlier work of Berryman
and Milton "0881#Ł and therefore must generalize the equations of Brown and Korringa
"0864# to include the e}ects of a temperature _eld[ The resulting equations are ]

dV dpd dpf
−  ¦ −bdT\ "78#
V K Ks

dVf dpd dpf

−  ¦ −bf dT\ "89#
Vf Kp Kf

and
 Volume averaging\ e}ective stress rules\ and inversion for microstructural response 3716

dVf dpf
−  −bf dT[ "80#
Vf Kf

The new coe.cients are the volume thermal expansions b\ bf\ and bf for the overall
composite volume\ the pore volume\ and the pore ~uid\ respectively[ The change in tem!
perature "assumed uniform throughout the solid and the ~uid# is dT[ The constant bf has
also been introduced previously by McTigue "0875#[
A similar set of equations is valid for each porous constituent with the exception that
the unjacketed bulk moduli "Ks  Kf  Km# are equal for Gassmann materials\ as are the
overall volume and pore volume thermal expansions "b  bf  bm# ðand\ of course\ the
di}erential pressures must be expressed in terms of the micro!con_ning pressures "e[g[\
dpcA−dpf# when considering the Gassmann material embedded within the AÐB compositeŁ[
Thus\ each Gassmann material has only one thermal expansion parameter associated with
it\ labeled\ respectively\ bA and bB[
We assume that bA\ bB\ and b are all known from experiments[ We will _nd that in
order to close the system of coe.cients\ either bf or Kf must also be measured[ With
external stress constant\ if the increment in ~uid content zp due to a temperature deviation
dT is measured\ then bf is given as

zp
b f  bf ¦ [ "81#
fdT

For example\ if a jacketed sample with a tube piercing the jacket is immersed in a tem!
perature reservoir\ bf can be measured by monitoring the ~uid volume changes in the tube[
If the tube is made of material with a coe.cient of thermal expansion that is dramatically
lower than the sample\ this measurement may be easier to make than the equivalent
measurement of Kf[ In what follows\ bf will be considered as known a priori and Kf as
unknown[

2[1[0[ Volume avera`in` with temperature chan`es[ The volume averaging results for
the constituents are easily generalized from "11#Ð"13# to

dpA dfA
− ¦bA dT  9 = u½A − \ "82#
KA "0−fA #

dpB dfB
− ¦bB dT  9 = u½B − \ "83#
KB "0−fB #

and

dpf df
− ¦bf dT  9 = u¹f ¦ [ "84#
Kf f

The meanings of all the constants are the same as in "11#Ð"13#\ but the thermal expansion
terms have been added to the left!hand side[
We see that the basic averaging equations for partially welded or nonwelded contact
are almost the same as "11#Ð"13# for welded contact\ but now temperature changes are
included and the interpretation of some terms is a little di}erent[ The porous solid volumes
VA and VB do not complete the whole averaging volume V\ so df may now include some
void space "due to cracks or fractures# outside of these volumes but still inside V[ Also\ we
will _nd that the average pressure changes "dpA and dpB# are now functions of temperature
change as well as being functions of both con_ning and ~uid pressure changes[
3717 J[ G[ Berryman and S[ R[ Pride

2[1[1[ Other relations[ Equations "82#Ð"84# do not explicitly show dfC because it is the
mis_t porosity and\ therefore\ de_ned by dfC  df−dfA−dfB[ To see what this quantity
must be\ consider the general expressions for the change in volume fractions ]

VA "0¦9 = u½A #
vA ¦dvA  \ "85#
VA "0¦9 = u½A #¦VB "0¦9 = u½B #¦VC ¦dVC

VB "0¦9 = u½B #
vB ¦dvB  \ "86#
VA "0¦9 = u½A #¦VB "0¦9 = u½B #¦VC ¦dVC

and

VC ¦dVC
vC ¦dvC  [ "87#
VA "0¦9 = u½A #¦VB "0¦9 = u½B #¦VC ¦dVC

The resulting formulas for the changes in the porosities "neglecting terms higher than _rst
order are#

dvA  vA ð"vB ¦vC #9 = u½A −vB 9 = u½B −dVC :VŁ\ "88#

dvB  vB ð"vA ¦vC #9 = u½B −vA 9 = u½A −dVC :VŁ\ "099#

and

dfC  dvC "0−vC #dVC :V−vC "vA 9 = u½A ¦vB 9 = u½B #\ "090#

satisfying the constraint for volume fractions that dvA¦dvB¦dvC  9[

Addition of a temperature _eld does not change the pressure relation "18#\ but the
overall dilatation is now

dVC
9 = u¹s  vA 9 = u½A ¦vB 9 = u½B ¦ \ "091#
V

including the added volume introduced by the presence of cracks[

2[1[2[ De_nitions of microstructural coef_cients[ The micro!pressure laws "29# and "20#
must be changed when a temperature _eld is added because local average pressures are
dependent on temperature[ Thus\ "29# becomes

"0−fA #dpA  QA "dpc −gA dpf #¦oA dT\ "092#

with a similar expression for the B material[ We will see that the Qs and gs have exactly the
same algebraic de_nitions as in the welded case[
In the presence of temperature change\ the expressions for response of the dilatations
"21#Ð"22# become

−9 = u½A  SA "dpc −a¹ A dpf #−b¹ A dT\ "093#

with a similar relation for material B[ We must also add a new relation for the crack
dilatation
 Volume averaging\ e}ective stress rules\ and inversion for microstructural response 3718

dVC
−  SC "dpc −a¹ C dpf #−b¹ C dT[ "094#
V

Cracks can be opened or closed by either pressure or temperature changes\ but we assumed
that welded contacts prevented this from happening in Section 2[0[
Modi_ed equations for the changes in the porosities are

−dfA  WA "dpc −xA dpf #−kA dT\ "095#

with a similar relation for B[ Again\ we must add a new law for the cracks

−dfC  WC "dpc −xC dpf #−kC dT[ "096#

This completes the de_nitions of the microstructural coe.cients[

There are 13 coe.cients for partially welded or nonwelded contact\ compared to 01
for welded contact[ Eight of these coe.cients are new thermal expansion coe.cients\ and
the other four new ones are related to pressure dependence of the crack phase[ The overall
strategy used to _nd these coe.cients will be similar to that in the welded case[ We will ]
"0# relate the averaging equations to the temperature!generalized Brown and Korringa
laws ^ "1# introduce the micropressure laws into the A and B Gassmann laws ^ and "2# use
the Berryman and Milton "0881# thought experiment "involving the temperature _eld this
time#[ It will be possible to express all coe.cients exactly in terms of K\ b\ bf and the
same known properties of the A and B Gassmann materials used in the welded case[

2[1[3[ General relations for microstructural response[ Now we _nd from "091# that

0
vA SA ¦vB SB ¦SC  \ "097#
K

a
vA SA a¹ A ¦vB SB a¹ B ¦SC a¹ C  \ "098#
K

and

vA b¹ A ¦vB b¹ B ¦b¹ C  b[ "009#

All three equations follow from the direct generalization of eqn "36#[ These equations are
what allow the crack coe.cients SC\ a¹ C\ and b¹ C to be determined[
They also allow us to eliminate the coe.cients in "096#\ by _rst equating "090#Ð"096#
and then using "097#Ð"009# to simplify the results to

vC
WC  SC − \ "000#
K

vC a
WC xC  SC a¹ C − \ "001#
K

and

kC  b¹ C −vC b[ "002#

These three equations are what allow WC\ xC and kC to be determined[

The addition of the temperature _eld does not change "18# or "25# and therefore does
not change either "26# or "27#\ with the exception that vA¦vB ¾ 0 now[ Thus\ the _rst two
equations for the Qs and gs do not change for nonwelded contact[ The temperature
3729 J[ G[ Berryman and S[ R[ Pride

dependence on the right!hand side of "18# must average to zero\ since the applied pressures
are independent variables\ so

vA oA ¦vB oB  9[ "003#

Equation "30# changes to

0 1 0 1
dpA dpB
−v¹A −bA dT −v¹B −bB dT "0−f#9 = u¹s −df\ "004#
KA KB

ðsome algebra was needed using "88#Ð"091# to verify that the right!hand side is correctŁ
where now the porosity change is given by

0 1
a−f 0 0
−df  "dpc −dpf #¦f − dpf −f"bf −b#dT[ "005#
K Kf Ks

It follows that there is no change in eqns "31# and "32# for the Qs and gs\ but that we now
have a condition for the micro!pressure temperature coe.cients

v A oA v B oB
v¹A bA ¦v¹B bB − −  b−fbf [ "006#
KA KB

The e}ects of the crack volume "and in fact of all the porosity# have cancelled on the right!
hand side of "004#\ which is why there is no explicit C dependence in either "004# or "006#[
Solving "003# and "006# for the os gives

b−fbf −v¹A VA −v¹B bB

vA oA  "007#
0:KB −0:KA

vB oB  −vA oA [ "008#

Thus\ just as in the welded case\ we have succeeded in expressing all the micropressure
coe.cients in terms of the macroscopic moduli and other known properties[
Although not central to our program\ we can also make the following observations at
this point[ Considering the Qs and gs as known\ we can return to "82# and "83# to determine
the Ws and xs[ Substituting "093# and "095# "along with their equivalent B expressions#
into "82# and "83# and equating coe.cients produces exactly the same eqns "49#Ð"40#
relating the Ws and xs to the Ss and a¹ s and known quantities[ The expressions relating the
thermal expansion coe.cients turns out to be redundant when compared to others we will
obtain next\ so we will not show these here[ Also\ substituting into
df  vAdfA¦vBdfB¦dfC¦fAdvA¦fBdvB\ and equating to "005# just reproduces "097#Ð
"009#[ All information obtainable from the averaging equations alone has been exhausted[

2[1[4[ Gassmann laws for the A and B components[ In the presence of an applied
temperature change\ the micro!con_ning pressure for Gassmann material A given by
dpcA "0−fA#dpA¦fAdpf is de_ned

$ %
"QA gA −fA #
dpcA  QA dpc − dpf ¦oA dT "019#
QA

with an analagous expression for the B component[ The Gassmann laws thus take the form
 Volume averaging\ e}ective stress rules\ and inversion for microstructural response 3720

$ % 0 1
QA "aA −fA ¦QA gA # oA
−9 = u½A  dpc − dpf − bA − dT "010#
KA QA KA

$ %
"aA −fA #QA "0−fA ¦QA gA # "aA −fA #oA
−dfA  dpc − dpf ¦ dT "011#
KA QA KA

with analagous expressions for the B component[

By comparing to the microstructural laws\ we see that because the Qs and gs have the
same functional de_nition as in the welded case\ then so do the Ss\ Ws\ a¹ s and xs[ The
temperature coe.cients of the microstructural laws are de_ned by

oA
b¹ A  bA − "012#
KA

"fA −aA #oA

kA  "013#
KA

with analagous results for the B coe.cients[

At this point\ just as in the welded case\ the only remaining coe.cients that must be
determined are Ks and Kf "since bf is assumed known#[

2[1[5[ Uniform expansion or contraction for partially welded contact[ In the presence of
cracks\ it is not possible to guarantee that the con_ning pressure is uniform throughout the
composite porous material "Berryman and Milton\ 0881#[ Therefore\ the thought exper!
iment of Berryman and Milton "0880# for welded contact described in the last section is
not valid for this case[ The key idea behind the success of that approach was that there
existed a circumstance in which two independently controlled _elds could be used to cause
the uniform expansion or contraction of the composite[ Having introduced the temperature
_eld "which is assumed here to be uniform throughout the sample given enough time#\ we
can again construct a situation wherein the sample expands or contracts uniformly\ but
now the two required applied _elds are ~uid pore pressure and temperature[ It is also
necessary to set the changes in external con_ning pressure equal to those of the ~uid
pressure\ so that all di}erential pressure changes vanish[
Supposing that it is possible to have 9 = u½A  9 = u½B  9 = u¹s\ we _nd the corresponding
dilatations are given by

dpf dpf dpf

−bA dT  −bB dT  −bdT  −9 = u¹s [ "014#
KA KB Ks

It follows that these equalities are satis_ed if the changes of the _elds occur in the ratio

dT 0:KA −0:KB
0t "015#
dpf bA −bB

0 1
0:Ks −0:KA 0:Ks −0:KB
  \ "016#
b−bA b−bB

showing that a ratio t does indeed exist and is expressible in terms of known quantities[
Furthermore\ we see that Ks has been determined and is related to the known thermal!
expansion coe.cients[ A similar result has been known in the theory of thermoelasticity
since the work of Levin "0856# and Cribb "0857#\ but these authors implicitly assumed
welded contact between constituents\ and they were not considering porous media[
Changes in ~uid pressure and temperature do not induce changes in porosity if the
change in di}erential pressure vanishes while the composite expands or contracts uniformly[
3721 J[ G[ Berryman and S[ R[ Pride

Thus\ we obtain another valid relation for the uniform expansion or contraction scenario
by setting df  9[ The result is

0 0
 ¦"bf −b#t "017#
Kf Ks

and was obtained earlier by Berryman and Milton "0881#[ With bf a known quantity\ Kf
is determined by this equation[ Thus\ the entire system of coe.cients has been closed at
this point and our problem is solved[
One might think that all possible identities have not yet been extracted and that\ in
particular\ bf might yet be determined "rather than measured#[ Not only is the macroscopic
porosity unchanged when dT:dpf  t\ but furthermore all of the component porosity chan!
ges vanish "dfA  dfB  dfC  9#\ and so do the volume fraction changes
"dvA  dvB  dvC  9#[ If we set the left!hand side of "82# equal to "014#\ we _nd that

−oA t  QA "0−gA #−"0−fA #\ "018#

and a similar expression for oB holds[ The result may be written as

$ %
0 v¹A v¹B 0 f
v A oA t  ¦ − ¦ [ "029#
0:KA −0:KB KA KB Ks Kf

It is easy to check that these expressions satisfy "003#\ since changing A to B in "029# only
changes the sign of the right hand side[ Also\ see Appendix B[
Equation "029# shows that "006# becomes

$ %
0 0 f v¹A v¹B
b−fbf  v¹A bA ¦v¹B bB ¦ − − − \ "020#
t Ks Kf KA KB

which is a general identity for partially welded\ nonwelded\ or welded contact\ and which
may be viewed as another formula relating Kf to bf[ However\ "020# is a linear combination
of "016# and "017# and thus does not provide a separate equation for bf[ We thus conclude
that all possible information has been extracted from this thought experiment and that
either bf or Kf must be measured in order to close the system of coe.cients[

2[1[6[ The special case of a rubblized bed[ In special cases\ it may be possible to avoid
the measurement of bf[ "See also Appendix E[# Consider the extreme case of fully nonwelded
contact[ By this term\ we mean that we have two porous constituents A and B jumbled
together so that wherever these two types of materials touch there are no constraints
requiring them to remain in contact during a deformation caused by stress or temperature
change[ This scenario implies that A and B materials form a {{rubblized bed|| such that
even some pieces of A material are not welded to other pieces of A material\ and similarly
for the B material[ This model is special in that results can be obtained for it when only
temperature is changed or when only the pore pressure is changed while di}erential pressure
is constant[
Consider the situation when a change in temperature occurs in the fully nonwelded or
rubblized material[ With no constraints to maintain contact with the surrounding materials\
there should be no feedback between temperature changes and average pressure changes in
the solid[ Thus\ we expect to _nd

oA  oB  kA  kB  9[ "021#

Similarly\ the changes in solid volume in this situation are controlled simply by the pure
material volume thermal expansion coe.cients\ so
 Volume averaging\ e}ective stress rules\ and inversion for microstructural response 3722

b¹ A  bA and b¹ B  bB \ "022#

which in turn implies that

b¹ C  b−vA bA −vB bB [ "023#

And since kC  b¹ C−vCb from "002#\ all the thermal constants are completely determined[
In particular\ it follows from "006# that

fbf  b−v¹A bA −v¹B bB \ "024#

showing that the thermal expansion coe.cient for the pore space is completely determined
from more easily measured quantities in this special case of fully nonwelded contact[

2[1[7[ Summary for partially welded and nonwelded contact[ As was done in the welded
case\ we now summarize the key results for the case of partially welded and nonwelded
contact[ The e}ective stress laws governing the overall response of a multicomponent
material in the presence of temperature changes are de_ned by

0
−9 = u¹s  "dpc −adpf #−bdT "025#
K

a−f
−df  "dpc −xdpf #−f"bf −b#dT\ "026#
K

where the e}ective!stress coe.cients a and x are algebraically de_ned as in the welded case[
The thermal expansion moduli b and bf are assumed known[ For the case of two Gassmann
materials in partially welded or nonwelded contact\ the moduli Ks and Kf "that are within
a and x# are de_ned by

0 0
 ¦t"b−bA # "027#
Ks KA

0 0
 ¦t"bf −b#[ "028#
Kf Ks

The coe.cient

0:KA −0:KB
t "039#
bA −bB

is the special ratio of temperature increment to ~uid pressure increment that results in
uniform strain[
The microstructural laws are de_ned by

−9 = u½A  SA "dpc −a¹ A dpf #−b¹ A dT "030#

dVC
−  SC "dpc −a¹ C dpf #−b¹ C dT "031#
V

−dfA  WA "dpc −xA dpf #−kA dT "032#

−dfC  WC "dpc −xC dpf #−kC dT "033#

"0−fA #dpA  QA "dpc −gA dpf #¦oA dT "034#

3723 J[ G[ Berryman and S[ R[ Pride

with analagous expressions for the B component[ The coe.cients SA\ a¹ A\ WA\ xA\ QA\ and
gA and their corresponding B expressions have the exact algebraic de_nition given already
in the welded case "of course\ because Ks and Kf are di}erent from those in the welded
case\ the numerical values of these coe.cients will be di}erent in the two cases#[ The new
coe.cients\ that are associated with crack openings and the temperature dependence\ are
de_ned by

b−fbf −v¹A bA −v¹B bB

vA oA  "035#
0:KB −0:KA

oA
b¹ A  bA − "036#
KA

oB
b¹ B  bB − "037#
KB

b¹ C  b−vA b¹ A −vB b¹ B "038#

0
SC  −vA SA −vB SB "049#
K

$ %
0 a
a¹ C  −vA SA a¹ A −vB SB a¹ B "040#
SC K

kC "0−vC #b−vA bA −vB bB "041#

0−vC
WC  −vA SA −vB SB "042#
K

$ %
0 "0−v#a
xC  −vA SB a¹ A −vB SB a¹ B [ "043#
WC K

Although\ in general\ bf is assumed to be a known "measured# constant\ for the special

case of fully nonwelded contact "a so!called rubblized bed#\ it takes the special form

fbf  b−v¹A bA −v¹B bB "044#

and\ thus\ need not be measured independently[

3[ GENERAL EFFECTIVE STRESS EQUATION FOR MATERIALS WITH WELDED

CONTACT

Welded contact may be treated as a special case of partially welded contact[ Everything
in the analysis goes through\ but now vC  dvC  dfC  9\ and therefore SC  a¹ C  9[ Both
thought experiments considered so far apply simultaneously\ so we again have "58# and
"69#[ Thus\ we are again in the position of having one eqn "69# in two unknowns "i[e[\ a¹ A
and a¹ B#[
There is still another legitimate thought experiment for this problem[ Since welded
contact implies the possibility of constant con_ning pressure throughout the sample "Berry!
man and Milton\ 0881#\ we can consider an experiment with dpf  9\ and try to _nd a ratio
of dpc and dT such that 9 = u½A  9 = u½B  9 = u¹s[ Then\ from "78#\ we have
 Volume averaging\ e}ective stress rules\ and inversion for microstructural response 3724

dpc dpc dpc

−bA dT  −bB dT  −bdT  9 = u¹s \ "045#
KA KB K

showing that

dT 0:KA −0:KB 0:K−0:KA 0:K−0:KB

0v   [ "046#
dpc aA −bB b−bA b−bB

These results are completely equivalent to those of Levin "0856# and Cribb "0857#\ indicating
necessary relations between K and b for welded contact\ but those authors were treating
only thermoelastic materials\ not thermoporoelastic such as we are treating here[
If we try to carry through the analysis of this thought experiment "even though it is
perfectly legitimate#\ we will learn nothing useful[ The reason for this is that the equations
resulting from the three thought experiments are not linearly independent[ It is easy to see
that

t
 0−u[ "047#
v

Thus\ we cannot obtain any additional constraints on the coe.cients this way[ Furthermore\
we _nd that the region in which SA is constant is actually a plane in the three!dimensional
space of the _eld variables and that the functional form of SA is

SA  SA ðdT−v"dpc −udpf #Ł\ "048#

for welded contact[ It is easy to check that this is so by writing out the expressions for dvA
and dvB\ and noting that the condition for dvA  −dvB  9 is the same as the condition for
the argument of "048# to be zero[ When the various equalities satis_ed by v\ u\ and t have
been accounted for\ we _nd for example that

dvA  vA "b¹ A −b#ðdT−v"dpc −udpf #Ł[ "059#

But\ no matter how we choose to write this expression\ it always has one parameter
undetermined\ and therefore requires one additional measurement[
It is also important to recognize that we have not exhausted the identities[ Considering
entropy "which is an extensive quantity#\ we can obtain another identity in the isothermal
thought experiment[ The full matrix equation for the thermoporoelastic problem is

2 32 32 3
de 0:K −a:K b −dpc
−dz  −a:K a:K¦f"0:Kf −0:Kf # f"bf −bf # × −dpf \ "050#
ds b f"bf −bf # ðfcf ¦"0−f#cp Ł:T9 dT

where the new terms are the change in the entropy ds\ and the heat capacities at constant
pressure for the ~uid cf and for the composite frame cp [ The ambient temperature is T9[
Then\ it follows immediately from the isothermal thought experiment with dpc:dpf  u that
the change in total entropy is the sum of the changes in the two components\ so*making
use of the bottom row of the matrix*we _nd

b−fbf  v¹A bA ¦v¹B bB ¦"u−0#ðvA bA ¦vB bB −bŁ\ "051#

which is a new identity for bf that should be compared with the discussion of Appendix B[
This formula is true only for materials subject to welded contact between constituents[
3725 J[ G[ Berryman and S[ R[ Pride
4[ CONCLUSIONS AND DISCUSSIONS

For the case of an isotropic composite of two Gassmann materials "A and B#\ we have
determined the e}ective stress laws controlling not only the overall volumetric response but
also the response of the individual A and B components[ In addition to the basic properties
of the individual "isolated# A and B materials\ we have expressed the laws using only ]
"0# the moduli K and b in the case of fully welded or fully nonwelded contact ^ and
"1# the moduli K\ b\ and bf in the case of partially welded contact[ Having such e}ective
stress laws is key to understanding how the physical properties of the earth|s crust vary as
a function of depth[
The present results are limited by assumptions of isotropy for the constituents\ stat!
istical isotropy for the overall poroelastic medium\ and by the necessity of assuming the
constituents are lumped together as Gassmann materials[ Nevertheless\ even with these
limitations\ the subset of real materials covered by this analysis still includes many inter!
esting materials ðsuch as quartz grains embedded in interstitial clay*see Berryman and
Milton "0880# for a discussionŁ that cannot be treated by previous methods[ We therefore
view the present work as one step toward a more sophisticated analysis that will be required
to treat more general porous media[

Acknowled`ements*The work of J[G[B[ was performed under the auspices of the U[S[ Department of Energy by
the Lawrence Livermore National laboratory under contract No[ W!6394!ENG!37\ in part while on sabbatical at
the Institut de Physique du Globe de Paris[ Partial support of the French government for both authors is also
gratefully acknowledged[

REFERENCES

Berryman\ J[ G[ "0881# E}ective stress for transport properties of inhomogeneous porous rock[ J[ Geophys[ Res[
86\ 06\398Ð06\313[
Berryman\ J[ G[ "0882# E}ective!stress rules for pore!~uid transport in rocks containing two minerals[ Int[ J[ Rock
Mech[ Min[ Sci[ and Geomech[ Abst[ 29\ 0054Ð0057[
Berryman\ J[ G[ "0886# Generalization of Eshelby|s formula for a single ellipsoidal elastic inclusion to poroelasticity
and thermoelasticity[ Physics Review Letters 68\ 0031Ð0034[
Berryman\ J[ G[ and Berge\ P[ A[ "0885# Critique of two explicit schemes for estimating elastic properties of
multiphase composites[ Mech[ Materials 11\ 038Ð053[
Berryman\ J[ G[ and Milton\ G[ W[ "0880# Exact results for generalized Gassmann|s equations in composite
porous media with two constituents[ Geophysics 45\ 0849Ð0859[
Berryman\ J[ G[ and Milton\ G[ W[ "0881# Exact results in linear thermomechanics of ~uid!saturated porous
media[ Applied Physics Letters 50\ 1929Ð1921[
Berryman\ J[ G[ and Thigpen\ L[ "0874# Nonlinear and semilinear dynamic poroelasticity with microstructure[
Journal of the Mechanics and Physics of Solids 22\ 86Ð005[
Berryman\ J[ G[ and Wang\ H[ F[ "0884# The elastic coe.cients of double!porosity models for ~uid transport in
jointed rock[ J[ Geophys[ Res[ 099\ 13\500Ð13\516[
Biot\ M[ A[ "0851# Mechanics of deformation and acoustic propagation in porous media[ Journal of Applied
Physics 22\ 0371Ð0387[
Biot\ M[ A[ "0862# Nonlinear and semilinear rheology of porous solids[ J[ Geophys[ Res[ 67\ 3813Ð3826[
Biot\ M[ A[ and Willis\ D[ G[ "0846# The elastic coe.cients of the theory of consolidation[ ASME Journal of
Applied Mechanics 13\ 483Ð590[
Brown\ R[ J[ S[ and Korringa\ J[ "0864# On the dependence of the elastic properties of a porous rock on the
compressibility of a pore ~uid[ Geophysics 39\ 597Ð505[
Burridge\ R[ and Keller\ J[ B[ "0870# Poroelasticity equations derived from microstructure[ Journal of the Acoustic
Society of America 69\ 0039Ð0035[
Cribb\ J[ L[ "0857# Shrinkage and thermal expansion of a two phase material[ Nature 119\ 465Ð466[
Gassmann\ F[ "0840# U Ý ber die elastizitat poroser medien[ Veirteljahrsschrift der Naturforschenden Gesellschaft in
Zurich 85\ 0Ð12[
Hill\ R[ "0852# Elastic properties of reinforced solids ] some theoretical principles[ Journal of the Mechanics Physics
of Solids 00\ 246Ð261[
Landau\ L[ D[ and Lifshitz\ E[ M[ "0875# Theory of Elasticity\ pp[ 2Ð6[ Pergamon\ Oxford[
Levin\ V[ M[ "0856# Thermal expansion coe.cients of heterogeneous materials[ Mech[ Solids 1\ 47Ð50[
McTigue\ D[ F[ "0875# Thermoelastic response of ~uid!saturated porous rock[ J[ Geophys[ Res[ 80\ 8422Ð8431[
Mei\ C[ C[ and Auriault\ J[ L[ "0878# Mechanics of heterogeneous porous media with several spatial scales[
Proceedin`s of the Royal Society of London A 315\ 280Ð312[
Norris\ A[ N[ "0881# On the correspondence between poroelasticity and thermoelasticity[ Journal of Applied
Physics 60\ 0027Ð0030[
Nur\ A[ and Byerlee\ J[ D[ "0860# An exact e}ective stress law for elastic deformation of rock with ~uids[ J[
Geophys[ Res[ 65\ 5303Ð5308[
Pride\ S[ R[ and Berryman\ J[ G[ "0887# Connecting theory to experiment in poroelasticity[ Journal of the
Mechanics and Physics of Solids 35\ 608Ð636[
 Volume averaging\ e}ective stress rules\ and inversion for microstructural response 3726
Pride\ S[ R[\ Gangi\ A[ F[ and Morgan\ F D[ "0881# Deriving the equations of motion for porous isotropic media[
Journal of the Acoustic Society of America 81\ 2167Ð2189[
Rice\ J[ R[ "0864# On the stability of dilatant hardening for saturated rock masses[ J[ Geophys[ Res[ 79\ 0420Ð
0425[
Rice\ J[ R[ and Cleary\ M[ P[ "0865# Some basic stress di}usion solutions for ~uid!saturated elastic porous media
with compressible constituents[ Rev[ Geophys[ Space Phys[ 03\ 116Ð130[
Slattery\ J[ C[ "0856# Flow of viscoelastic ~uids through porous media[ Am[ Inst[ Chem[ En`[ J[ 02\ 0955Ð0960[
Tuncay\ K[ and Corapcioglu\ M[ Y[ "0884# E}ective stress principle for saturated fractured porous media[ Water
Resources Res[ 20\ 2092Ð2095[
Walsh\ J[ B[ "0854# The e}ect of cracks on the compressibility of rock[ J[ Geophys[ Res[ 69\ 270Ð278[
Wang\ H[ F[ and Berryman\ J[ G[ "0875# On constitutive equation and e}ective stress principles for deformable\
double!porosity media[ Water Resources Res[ 21\ 2510Ð2511[
Whitaker\ S[ "0858# Advances in the theory of ~uid motion in porous media[ Ind[ En`[ Chem[ 01\ 03Ð17[
Whitaker\ S[ "0875# Flow in porous media III ] deformable media[ Transp[ Porous Media 0\ 016Ð043[
Zimmerman\ R[ W[\ Myer\ L[ R[ and Cook\ N[ G[ W[ "0883# Grain and void compression in fractured and porous
rocks[ Int[ J[ Rock Mech[ Min[ Sci[ Geomech[ Abstr[ 20\ 068Ð073[
Zimmerman\ R[ W[\ Somerton\ M[ H[ and King\ M[ S[ "0875# Compressibility of porous rocks[ J[ Geophys[ Res[
80\ 01\654Ð01\666[

APPENDIX A ] VOLUME AVERAGING WITH TWO SOLID CONSTITUENTS

When two solid constituents are present\ the averaging theorem gives some signi_cantly di}erent results that
we will present and discuss here[ For simplicity\ we will limit the discussion to averages of divergence of
displacement and of the displacement itself[ We make one assumption implicitly here\ that the averaging volume
V is large enough so that statistical di}erences between the bulk porosity in the volume and the outcrop of porosity
at the surface of the volume are negligible[ When this assumption is not made\ gradients of porosity appear in the
_nal equations as is shown in Pride and Berryman "0887#[

A[0[ Results for all solids

If there are two solids present as in Fig[ 0\ for purposes of averaging we may lump them together and act as
if they are both {{solids||\ using subscript {{s|| to refer to the regions occupied by these solids[ The averaging
theorem "or in this case just the normal divergence theorem# states that

g g g
0 0 0
ð9 = us Ł  9 = us d2 x  n¼s = us dS¦ n¼s = us dS\ "052#
V V"x# V 1E V 1I

where V is the averaging volume and V  ÐV d 2x\ with 1E being the external boundary and 1I being the internal
or pore boundary[ The divergence of the average " found by taking the Frechet derivative with respect to the
averaging volume# is

g
0
9 = ðus Ł  n¼s = us dS[ "053#
V 1E

The internal surface integral is easily interpreted as the negative of the change in porosity df\ since the displacement
integrated along the bounding surface produces a volume change that is then normalized by the total volume V
in the averaging volume V"x#\ so

g
0
n¼s = us dS  −df[ "054#
V 1I

The left!hand side of "052# is also easily interpreted as the total change in solid volume dðV"0−f#Ł divided by the
averaging volume V[ Comparing these expressions shows that

dV
9 = ð us Ł "0−f# [ "055#
V

Note that there has been no change in the averaging volume V\ but there has been movement of solid in or out of
volume and:or a change in state of compression of the solid[ In this regard\ our notation is trying to re~ect the
fact that experimentally we normally start with a _xed volume of material V and then measure changes dV in that
volume[
Accounting for the volume occupied by the solid initially\ we have ðusŁ 0"0−f#u¹s\ so

dV u¹s = 9f
9 = u¹s  ¦ [ "056#
V 0−f

We normally neglect the second term on the right!hand side of "056#\ since we assume that the scales of variation
of the displacement _eld are much smaller than those for the porosity\ so that =u¹s = 9f= ð =9 = u¹s =[ However\ a fully
general analysis must account for the presence of this term ðsee Pride and Berryman "0887#Ł[

A[1[ Results for constituents A and B with welded contact

When we want to distinguish the constituent solids A and B as in Fig[ 0\ we can break up the averaging
volume into two pieces such that
3727 J[ G[ Berryman and S[ R[ Pride
ð9 = us Ł  ð9 = uA Ł¦ð9 = uB Ł\ "057#

which follows immediately from the fact that material A and B occupy disjoint parts of the averaging volume[
The averaging theorem for the divergence of uA alone then states that

g g
0 0
ð9 = uA Ł  9 = ðuA Ł¦ n¼A = uA dS¦ n¼A = uA dS\ "058#
V 1If V 1IB

where we have explicitly noted that the interior interface has two parts ] one boundary 1If with the ~uid "or pore
space# and one 1IB with the other solid "B#[
Since

g
0
9 = ðuA Ł  n¼A = uA dS\ "069#
V 1EA

and a similar expression for 9 = uB\ we _nd easily from the identity

g g g
0 0 0
n¼s = us dS  n¼A = uA dS¦ n¼B = uB dS "060#
V 1Es V 1EA V 1EB

that
9 = ðus Ł  9 = ðuA Ł¦9 = ðuB Ł[ "061#

In order to specify the physical signi_cance of 9 = ðuAŁ\ we need to repeat the analysis for all solids\ taking
into account the fact that when there are two or more solids there must also be additional interior interfaces
between these various constituents[ Of the four terms in "058#\ each requires some interpretation[ First\ the left!
hand side has an interpretation similar to that of the left!hand side of "052#[ Thus\ we have the volume average
of the dilatation of A material must be

dðVA "0−fA #Ł
ð9 = uA Ł  \ "062#
V

where VA is the total porous volume of A material and VA "0−fA# is the total solid volume of A material[
The two integrals on the right!hand side of "058# are more di.cult to interpret because they involve the
contact region of two Gassmann materials having possibly di}erent porosities[ "See Fig[ 1[# Statistically the A
material should have solid material at this interface occupying the fraction 0−fA of the total interface area and
A!pore the remaining fraction fA of the total[ The B material has corresponding proportions[ Now these continuous
surfaces may be statistically correlated or uncorrelated[ If they are uncorrelated\ we can easily compute the
coe.cients we will need[ But if they are correlated\ we introduce new interface constants with the following
properties ] hAB  hBA is the fraction of the interface on which solid A touches solid B\ hAf is the fraction of the
interface on which solid A touches the ~uid in B\ and similarly hBf is the fraction on which the solid B touches the
~uid in A[ With our assumption of statistical homogeneity\ these constants should obey the general sum rules
hAB¦hAf  0−fA and hBA¦hBf  0−fB[ One immediate general result is that the di}erence hAf−hBf  fB−fA[
The solid:solid contact area should be proportional to hAB\ which may be very small or it can be as large as the
minimum of the two solid fractions "0−fA#\ "0−fB#[ For uncorrelated surfaces\ we expect hAB "0−fA#"0−fB#\

Fig[ 1[ At an internal interface between Gassmann porous materials A and B\ there will be regions
where ~uid from the A side interfaces with ~uid from the B side\ where ~uid from the A side
interfaces with solid from the B side\ where solid from the A side interfaces with ~uid from the B
side\ and where solids from both sides touch "not seen in this particular blowup of a two!dimensional
cross section of a three!dimensional medium#[
 Volume averaging\ e}ective stress rules\ and inversion for microstructural response 3728
hAf "0−fA#fB\ and hBf "0−fB#fA[ These identities are easily shown to satisfy the statistical sum rules for these
coe.cients[ For correlated interfaces\ we may view hAB as a new microstructural parameter that characterizes the
internal "to the averaging volume# solid:solid interface[
The _rst integral on the right!hand side of "058# is the surface integral of displacement along the ~uid
boundary[ This term has the same signi_cance as the corresponding one for the whole solid ^ it is the change in
porosity associated with type!A material[ The total pore volume associated with A is VAfA\ so the change in pore
volume must be a change in this quantity[ However\ the surface integral is strictly over the original boundary of
the A material "prior to the displacements uA#\ so the correct expression for this change in the absence of other
solids is clearly VAdfA[ But\ in the presence of other solids\ we must account for the possibility of changes in
overall porosity due to changes in volume fraction[ Thus\ the full contribution of this term is

g1If
n¼A = uA dS  −VA dfA ¦VhAf dvA \ "063#

using one of the interface constants introduced in the preceding paragraph[ When the volume fraction does not
change\ as in the case when the averaging volume happens to contain only A material\ we see that this expression
reduces correctly to "054#[ When we write the corresponding relation for the B phase and then consider that by
de_nition it must be true that

g1If
n¼s = us dS 
g1If
n¼A = uA dS¦
g
1If
n¼B = uB dS\ "064#

we see that the extra terms proportional to change in volume fraction are exactly what were needed to guarantee
that "064# is equivalent to df  vAdfA¦vBdfB¦"fA−fB#dvA[
The second integral on the right!hand side of "058# is the surface integral over the AB solid:solid interface[
This term also has the important characteristic that it must be exactly the negative of the corresponding term for
the B material[ So however we interpret it\ the expression should be easily identi_ed by the fact that interchanging
A and B should change the sign of the term[ We make the identi_cation

g
0
n¼A = uA dS 0 hAB dvA \ "065#
V 1IB

where dvA is the change in volume fraction of porous constituent A[ This interpretation is reasonable ] if a _ctitious
continuous surface is drawn between the porous constituents and the corresponding surface integral taken\ then
the result would be exactly dvA[ Since the true AB interface occupies only a fraction hAB of this total interface area\
we see that "065# follows[
The remaining term in "058# needing interpretation is proportional to the surface integral of the A component
displacement[ Combining the previous results\ this average must be given by

g1EA
n¼A = uA dS "0−fA #ðdVA −VdvA Ł "0−fA #vA dV\ "066#

with a matching expression for the B phase[ That these two integrals must satisfy the sum rule in "060# together
with "055# implies that their sum must be equal to "0−f#dV which is easily seen to be true "since vA¦vB  0 and
vAfA¦vBfB  f# for the case of welded contact[
Comparing all these expressions\ we _nally determine that

dVA
9 = ðuA Ł  v"0−fA # −"0−fA #dvA \ "067#
VA

which is the desired expression for divergence of the average displacement of A[

Although the divergence of ðuAŁ is given rigorously by "067#\ the quantities that actually appear in the
quasistatic equations of motion are simply the dilatations of the constituents\ so we will de_ne a new quantity u½A
satisfying

dVA
9 = u½A 0 \ "068#
VA

which is related to u¹A by

dvA
9 = u¹A  9 = u½A − \ "079#
vA

where the change in volume fraction is itself related to u½A and the corresponding expression of B by
dvA  vAvB "9 = u½A−9 = u½B#[ These de_nitions and interrelations will be important for analysis of wave propagation
issues for multicomponent rocks[
3739 J[ G[ Berryman and S[ R[ Pride
APPENDIX B ] PRESSURE COEFFICIENT IDENTITY
B[0[ Welded contact
Using the results from the uniform expansion:contraction thought experiment for con_ning and ~uid pressure
changes under isothermal conditions\ we can obtain a somewhat simpli_ed expression for the combination of
constants ðsee "33# and "34#Ł

0 1
0 0 f 0 f
vA QA "0−gA #  − ¦ [ "070#
0:KB −0:KA KB Ks Kf

First recall the identities ðsee "54# or Berryman and Milton "0880#Ł

0 1 0 1
0 0 0 0 0 0 0
 ¦"u−0# −  ¦"u−0# − \ "071#
Ks KA KA K KB KB K

and

$ 0 1 0 1 %
f f aA −fA aB −fB a−f
− "u−0# vA ¦vB − [ "072#
Kf Ks KA KB K

By volume averaging the two expressions in "071#\ we obtain the intermediate result that

0 1
0−f v¹A v¹B v¹A v¹B 0−f
 ¦ ¦"u−0# ¦ − [ "073#
Ks KA KB KA KB K

Then\ subtracting "072# from "073#\ we obtain

0 1
0 f v¹A v¹B vA vB 0
−  ¦ ¦"u−0# ¦ − [ "074#
Ks Kf KA KB KA KB Ks

Substituting "074# into "070# produces the useful identity

0 1
u−0 0 vA vB
vA QA "0−gA #  v¹A ¦ − − \ "075#
0:KB −0:KA Ks KA KB

together with a corresponding result for the B component[

B[1[ Partially welded contact

Using the results from the uniform expansion:contraction thought experiment for temperature and pressure
changes at zero di}erential pressure change\ we can also obtain a di}erent simpli_ed expression for the combination
of constants in "070#[ First\ recall ðsee "016# or Berryman and Milton "0881#Ł that

0 0 0
 ¦t"b−bA #  ¦t"b−bB #\ "076#
Ks KA KB

and that

0 0 0
 ¦t"bf −b#  ¦t"bf −bA #[ "077#
Kf K s KA

The constant t may be determined by solving for it in the second equation of "076#[ The second equality in eqn
"077# follows by substituting for Ks from "076#[
From these equalities\ it is straightforward to obtain

0 f v¹A v¹B
−  ¦ ¦"b−fbf −v¹A bA −v¹B bB #t[ "078#
Ks Kf KA KB

Combining these results with the de_nition of t and substituting into "070# produces the _nal result

0
vA QA "0−gA #  v¹A ¦ ðv¹ b ¦v¹ b −"b−fbf #Ł[ "089#
bB −bA A A B B

The corresponding result with A t B also holds[ Equation "089# should also be compared to "075#[

APPENDIX C ] SPECIAL EXAMPLE FOR WELDED CONTACT

In a few cases the algebra for these calculations simpli_es[ One particularly interesting and easy case assumes
that one of the constituents "say B# is purely solid\ i[e[\ fB  9 initially and remains so "dfB  9#[ Then\ the
number of equations that must be solved is dramatically reduced[ In particular\ we know from the outset that
 Volume averaging\ e}ective stress rules\ and inversion for microstructural response 3730
KB  KB [ As long as the other constituent|s porosity remains connected\ this example is still a valid limit of the
equations and has already been considered by Berryman and Milton "0880#[
Following the analysis of Section 2[0\ it is easily shown that the ratio of con_ning pressure to pore pressure
that gives rise to uniform expansion or contraction is

dpc 0:KA −0:KA

u [ "080#
dpf 0:KA −0:KB

We also _nd easily that

0 1
QB KA 0 0
SB   − "081#
KB vB "KA −KB # Ks KA

and that

0 1
0:Kf −0:KA
a¹ B  gB  f [ "082#
0:Ks −0:KA

Substituting these results into "37# and "38#\ we obtain

0 1
0 KA 0 0
vA SA  − − \ "083#
K KA −KB Ks KA

and

0 1
a fKA 0 0
vA SA a¹ A  − − [ "084#
K KA −KB Kf KA

For porosity variations\ WB  9 and xB is arbitrary "and therefore may be taken as zero#\ while WA and xA
are determined by "49# and "40#\ respectively[
Including the e}ects of thermal expansion in this model\ we _d that the microstructural constants for thermal
expansion take the form

$ 0 1%
0 vB KA 0−f 0 f
b¹ A  b¦ ¦ − ¦ "085#
vA t Ks KA −KB KA Ks Kf

and

$ 0 1 %
0 KA 0 f 0−f vB
b¹ B  b¦ − − − \ "086#
vB t KA −KB Ks Kf KA Ks

which correctly satisfy the constraint vAb¹ A¦vBb¹ B  b[

APPENDIX D ] EFFECTIVE STRESS RULES FOR MICROSTRUCTURAL PARAMETERS

An implicit assumption in the analysis of welded contact is that during the uniform expansion:contraction
experiment the coe.cients of the equations for the material dilatations are invariant[ So we have

dVA
−  SA "dpc −a¹ A dpf #\ "087#
VA

where SA and a¹ A are constant along the line dpc  udpf\ corresponding to uniform expansion or contraction[ A
similar result holds for the B material[ If we want to consider possible deviations from linearity in the overall
analysis\ we need to maintain the constancy of these coe.cients along this line but can suppose that they have
dpc\ dpf dependence of the form

SA  SA "dpc −udpf # and a¹ A "dpc −udpf #\ "088#

where u is a constant[ Thus\ we are introducing an e}ective stress rule for these coe.cients[
Now it will be useful to consider an integrability condition arising from the fact that

b
1 ln VA
SA 0 − "199#
1pc dpf 9

and that
3731 J[ G[ Berryman and S[ R[ Pride

b
1 ln VA
a¹ A SA 0 [ "190#
1pf dpc 9

It follows that the cross derivatives must be equal\ so

1SA 1"a¹ A SA #
− \ "191#
1pf 1pc

this being the standard condition for integrability of such equations[

Using the e}ective stress rule "088#\ we _nd from the chain rule that

1SA 1SA
 −u [ "192#
1pf 1pc

It then follows easily from "191# that

1SA 1a¹ A
"u−a¹ A #  SA \ "193#
1pc 1pc

which may be rewritten as

1
ln ð"u−a¹ A #SA Ł  9[ "194#
1pc

This form of the result shows that

constant"pc #
SA  \ "195#
u−a¹ A

which should be compared to the welded contact result "58# ] SA "u−a#:K"u−a¹ A#[ Using e}ective stress
arguments\ it is not hard to show that the factor "u−a#:K is indeed a constant as dpc varies\ as was done
previously by Berryman "0881#[

APPENDIX E ] SPECIAL EXAMPLE FOR PARTIALLY WELDED CONTACT

There exists one example of partially welded contact in which the equations can be completely solved[ We
consider a granular composite\ having two types of solid\ nonporous grains labeled A and B[ The entire pore
phase C consists of the gaps and mis_ts between and among the grains of both types[ The pore phase must be
continuous "connected# in order for pore pressure to equilibrate[ This model is included in the analysis of the
partially welded porous composite by taking the porosities fA  fB  9\ and assuming that dfA  dfB  9 for
any state of temperature and stress achieved[ The frame moduli satisfy KA  KA and KB  KB [
For partially welded contact\ it is anticipated that the local average pressures experienced by the grains may
be in~uenced by both the con_ning pressure and the pore pressure[ Therefore\ we make no assumptions about
the values of the a¹ s and gs[
The values of the coe.cients Q and g are as always given by "33# and "34#[ Since changes in porosity for
constituents A and B cannot occur in this model\ WA  WB  9\ so

QA
SA  "196#
KA

and

0:KB −0:Kf
a¹  gA  f \ "197#
0:KB −0:Ks

and similarly for the B grains[ Using "097#\ we _nd

a
SC  \ "198#
K

while "098# shows that

fK
a¹ C  0− "109#
aKf

follows from "098#[

 Volume averaging\ e}ective stress rules\ and inversion for microstructural response 3732
The thermal expansion for the A grains is given by

$ 0 1 %
0 KB 0−f 0 f vA
b¹ A  b¦ − ¦ − \ "100#
tvA KA −KB KB Ks K f Ks

and a similar expression for the B grains ðsee "086#\ which happens to be of the same formŁ[ The thermal expansion
for the pores then becomes

$ %
0 f vC
b¹ C  vC b¦ − [ "101#
t Kf Ks

These formulas also satisfy the constraint vAb¹ A¦vBb¹ B¦b¹ C  b[