Oil & Gas Science and Technology – Rev. IFP, Vol. 54 (1999), No. 6, pp.

703-714
Copyright © 1999, Éditions Technip

Mechanical Properties of Rocks:

Pore Pressure and Scale Effects
M. Boutéca1 and Y. Guéguen2
1 Institut français du pétrole, 1 et 4, avenue de Bois-Préau, 92852 Rueil-Malmaison Cedex - France
2 ENS, Laboratoire de géologie, département TAO, 24, rue Lhomond, 75231 Paris Cedex 5 - France
e-mail: maurice.bouteca@IFP.fr - gueguen@geologie.ens.fr

Résumé — Propriétés mécaniques des roches : pression de pore et effets d’échelle — La pression
de pore joue un rôle de première importance dans la considération des propriétés mécaniques des
roches. Dans ce domaine, le concept de contrainte effective est essentiel pour aborder les effets méca-
niques. Toutefois, son utilisation fréquente a conduit à de nombreuses affirmations trompeuses. Compte
tenu des significations diverses accordées à ce concept, nous tentons ici de le clarifier et examinons le
domaine d’application de ses divers emplois dans le cadre du comportement mécanique ou des proprié-
tés des roches. À l’échelle macroscopique, la thermodynamique offre un outil puissant pour cerner ce
concept. La thermodynamique des processus réversibles ou irréversibles fournit des relations générales
d’un intérêt majeur. Mais, compte tenu du fait que les roches sont des systèmes non homogènes, une
approche microscopique est nécessaire pour analyser les propriétés mécaniques à partir d’une descrip-
tion des phénomènes à petite échelle. L’approche microscopique est complémentaire de l’approche
macroscopique thermodynamique, elle conduit au calcul des propriétés effectives du milieu. Dans ce
cadre, la théorie des milieux effectifs est un outil puissant. Les propriétés effectives déduites de l’ana-
lyse à l’échelle microscopique peuvent être combinées aux relations issues de la thermodynamique pour
interpréter les effets de la pression de pore et les effets d’échelle. Le cas des propriétés élastiques des
roches poreuses est plus particulièrement traité à titre d’illustration, compte tenu de l’intérêt qu’il pré-
sente et de son importance du point de vue des applications.
Mots-clés : propriétés mécaniques, pression de pore, contrainte effective, effet d’échelle.

Abstract — Mechanical Properties of Rocks: Pore Pressure and Scale Effects — Pore pressure plays
a major role when considering rocks mechanical properties. In that field, the concept of effective pres-
sure is a key one to deal with fluids mechanical effects. However, its frequent use has been the source of
frequent confusing statements. Because of the various meanings which have been attached to that
concept, an attempt is made in this paper to clarify it and examine the validity of its various uses rela-
tive to rock mechanical behaviour or rock properties. At a macroscopic scale, thermodynamics provides
a powerful tool to investigate this. Reversible or irreversible thermodynamics provide general relation-
ships of great interest. But because real rocks are non homogeneous systems, a microscopic approach
is also required in order to analyze the mechanical properties from a description of the small scale pro-
cesses. The microscopic approach is complementary of the macroscopic thermodynamic one as it leads
to the calculation of the effective properties of the medium. In this last approach, effective medium
theory is a powerful tool. The effective properties as derived from the microscale can be nicely com-
bined to thermodynamic relations to interpret pore fluid pressure effects and scale effects. The example
704 Oil & Gas Science and Technology – Rev. IFP, Vol. 54 (1999), No. 6

of elastic properties of porous rocks is more specifically emphasized to illustrate this because of both its
intrinsic interest and importance as far as applications are concerned.
Keywords: mechanical properties, pore pressure, effective stress, scale effect.

1 MACROSCOPIC SCALE : THE EFFECTIVE We performed experiments on a Tavel limestone (Boutéca

STRESS CONCEPT et al., 1993). Applying an isotropic loading, we increased it
in a stepwise manner: the confining pressure is firstly
In the literature, the effective stress concept is applied to increased, the pore pressure is then increased. The pore
three different problems. It is first used for the constitutive pressure increase (∆pp) is equal to the confining pressure
equations, it is then extended to rock failure and finally to (∆pc) increase. Hence, at the end of a given step, the effective
rock properties. We will review those three categories to Terzaghi’s stress is equal to its initial value (∆pc – ∆pp = 0).
clarify the concept. Such a loading cycle is plotted in Figure 1a and the
theoretical stress-strain plot for this limestone is plotted in
Figure 1b.
1.1 The Effective Stress Concept: Constitutive Law
The data as obtained during the experiment are plotted in
At a macroscopic scale, the effective stress law is defined Figure 1c. The stress-strain line obtained during the confining
from an assumption. However, the assumption can be made at pressure increase is not the same as the stress-strain line
two levels. The first level of assumption consists in defining obtained during the pore pressure increase. Applying several
straightforwardly the stress-pressure relationship. Terzaghi’s cycles as shown in Figure 2a and 2b leads to a behavior that
concept is the best known example. However, this stress-strain strongly departs from Terzaghi’s model. In Figure 2a, we
relationship introduced as an hypothesis must be consistent plotted the applied loading cycles. We loaded and unloaded
with the thermodynamic coherence that derives from the strain along this stress path. In Figure 2b, we plotted the volumetric
energy. It can be shown that Terzaghi’s law can also be strain as a function of Terzaghi’s stress. Aside from a very
derived from energy considerations. slight hysteresis, the loading and unloading cycles are
If the stress-strain relationship is derived from energy, that identical. However in this Terzaghi’s plot, the stress strain
is from thermodynamics, the consistency of the model is relationship is not described by a unique straight line. The
granted. Note that deriving the stress strain relationship from pore pressure does not exactly play the same role as the
thermodynamics means that the level at which the confining pressure.
assumption is made has changed. In other words, the Before moving to another effective stress concept, it is
assumption is now made on the strain energy itself. important to notice that although Terzaghi’s stress has been
The macroscopic laws—both in the elastic and plastic defined straightforwardly, its meaning only appears when
domains—correspond to micromechanical mechanisms considering the rock deformation. That is Terzaghi’s stress is
which are the grain deformation, the cement deformation and the one that works in the strain field. Again this leads us
the grain displacement. towards an identification of the stress based on the energy.

1.1.1 Terzaghi’s Effective Stress - Elastic Domain 1.1.2 Biot’s Effective Stress - Elastic Domain

Terzaghi’s effective stress concept, in this paragraph, is Starting from a microscale approach, Biot (Biot, 1941)
restricted to its application in the elastic domain. proposed a potential from which he derived the constitutive
As previously stated, Terzaghi’s stress is defined from an equation in the elastic domain.
assumption on the stresses. It postulates that the pore Here, the assumption is no more at the level of the stress
pressure has no effect on the shear stress and that it decreases but at the level of the thermodynamic potential. The
the effect of compressive stress by an amount equal to the potential is assumed to be quadratic, and, following Coussy
pore pressure. Thus, the effective stress is defined as: (1991), the dual variables are defined as σ, ε, p and m, where
m is the fluid mass content in the rock. Deriving the
σ ijeff = σ ij – p δ ij (1)
potential, the constitutive law for an isotropic elastic rock is
where compressive stresses are taken positive. Expressing obtained as:
the volumetric deformation for an elastic body thus leads to:
σ ij – σ ijo =  Ko –
2G 
trε + 2Gε ij + b ( p – po ) δ ij (3)
ε v = Ko tr (σ eff ) / 3 (2)  3 

For an elastic rock, plotting the volumetric deformation as where Ko is the drained bulk modulus of the rock, G, the
a function of Terzaghi’s effective stress, should lead to a shear modulus and b, the Biot’s coefficient. σoij and po
straight line. are initial values.
 M Boutéca and Y Guéguen / Mechanical Properties of Rocks: Pore Pressure and Scale Effects 705

25 20
18
Confining pressure (MPa)

20 16

Terzaghi's stress (MPa)

14
15 12
10
10 8
6
5 4
2
0 0
0 10 20 30 0 0.0005 0.0010 0.0015
a Pore pressure (MPa) b Volumetric strain

Figure 1a Figure 1b
Isotropic loading cycle (∆pp = ∆pc). Theoretical stress-strain plot.

20
18
16
Terzaghi's sress (MPa)

Confining pressure
14 increase
12
10
8
Pore pressure
6 increase
4
2
0
0 0.0005 0.0010 0.0015
c Volumetric strain

Figure 1c
Experimental results (Tavel limestone).

70 20
18
60
16
Confining pressure (MPa)

Terzaghi's stress (MPa)

50 14
12
40
10
30 8

20 6
4
10 2

0 0
0 20 40 60 80 0 0.0005 0.001 0.0015 0.002
a Pore pressure (MPa)
b Volumetric strain

Figure 2a Figure 2b
Loading cycle. Experimental results (Tavel limestone).
706 Oil & Gas Science and Technology – Rev. IFP, Vol. 54 (1999), No. 6

As a consequence of the constitutive equation, an effective It follows from this equation that the stresses inducing the
stress can be defined: elastic deformation—the ones that insure the mechanical
equilibrium—follow the effective stress concept as explained
σ ijeff = σ ij – b p δ ij (4)
above (see Section 1.1.2). Thus we have:
Formally, Biot’s effective stress is very similar to σ'ijelas = σ ij – b p δ ij σ' ≡ σ eff (7)
Terzaghi’s one, and it can be shown that b varies in the range
[φ, 1], where φ is the rock porosity. When b equals 1, one In the plastic domain, one has to introduce hypotheses on
recognizes Terzaghi’s stress. More precisely the physical the flow rule to define an effective stress concept. Following
meaning of b can be found from Biot’s relation: Coussy (1991), we assume that the plastic porosity φp
Ko —irreversible change of the pore volume per unit initial
b =1– (5) volume—is proportional to the volumetric plastic strain,
Ks
which leads to a new effective stress:
where Ks is the skeleton bulk modulus. If the constitutive
grains are incompressible (Ks → ∞), then Biot’s effective φ p = β p ε p ⇒ σ' ijplas = σ ij – β p pδ ij (8)
stress and Terzaghi’s effective stress are identical.
If the constitutive grains are incompressible then, one
Note that, from micromechanics considerations, many
finds back Terzaghi’s stress.
authors have been deriving an effective stress such that
σ ijeff = σ ij – η pδ ij . One can find an updated review in
1.2 The Effective Stress Concept: Failure Criterion
Lade and de Boer (1997).
For the experiments shown in Figure 2, we determined the A failure criterion deals with a limit in a stress space, which
value of the Biot coefficient, it is equal to 0.65. The implies that the relevant effective stress will be derived from
volumetric strain is plotted as a function of the Biot’s stress a stress-stress relation and no more from a stress-strain
in Figure 3. The results show a straight line in this plot, in relation.
agreement with the theoretical prediction. At the microscopic level, the mechanisms are inter and
intragranular cracking leading to rock splitting. It results that
the effective stress of the constitutive law and the effective
35 stress at failure do not have to coincide. This is illustrated by
our experimental results on the Tavel carbonate (Vincké et
Biot's effective stress (MPa)

30
al., 1998).
25 To define the effective stress at failure, we performed
three sets of triaxial experiments with three different pore
b = 0.65

20
pressures (1 MPa, 10 MPa and 20 MPa). Assuming that the
15 effective stress can be defined as:
10 σ ijeff = σ ij – β p δ ij (9)
5
then, by plotting the value at failure in a q (deviatoric
0 stress)-p (total mean stress) plot, one should obtain one
0 0.0005 0.001 0.0015 0.002
failure curve per pore pressure level. Furthermore, β∆pp
Volumetric strain should be the translation vector between a failure curve
Figure 3 obtained for a given pore pressure pp and a failure curve
Volumetric strain vs. Biot’s stress - Tavel limestone (after obtained for a pore pressure equal to pp + ∆pp (Fig. 4).
Boutéca and Sarda, 1995). Using this procedure, we determined β for the Tavel
limestone and obtained β = 1. This is shown in Figure 5
1.1.3 Plastic Domain where we plotted at failure the deviatoric stress as a function
of the mean effective stress (p – β∆pp = p – ∆pp) for the
The constitutive equations are: three different pore pressures.

σ ij – σ ijo =  Ko –
2G  It is interesting to remind the preceeding results for the
trε elas + 2Gε ijelas + b ( p – po ) δ ij
 3  Tavel limestone:
(6) – constitutive equation: Biot’s effective stress
with ε = ε elas + ε plas σ ijeff = σ ij – 0.65 pδ ij ;
where the superscript elas stands for the elastic part and the
– failure criterion: effective stress σ ij = σ ij – pδ ij .
eff
superscript plas stands for the plastic part.
 M Boutéca and Y Guéguen / Mechanical Properties of Rocks: Pore Pressure and Scale Effects 707

240
Failure curve at pp
220
q: deviatoric stress

200
180
β∆pp

q (MPa)
160
Linear reg. pp = 1 MPa
140 Linear reg. pp = 10 MPa
Linear reg. pp = 20 MPa
120 pp = 1 MPa
Failure curve at pp + ∆pp pp = 10 MPa
100 pp = 20 MPa
80
20 70 120
p' (MPa) (β = 1)
p: total mean stress

Figure 4 Figure 5
Determination of β-failure criterion. Failure criterion (after Vincké et al.)

1.3 The Effective Stress Concept: Rock Properties result is often interpreted in the literature by defining a
relation between the bulk modulus and an effective stress.
The effective stress concept is also applied in the literature to Thus, nonlinear elasticity—as it is the case here—is treated
rock properties, both mechanical properties and transport as incremental linear elasticity with properties depending on
properties. We will first review its meaning in the case of an effective stress.
mechanical properties before moving towards transport Equation (3) is thus rewritten:
properties.
∆σ ijeff = ∆σ ij – b∆pδ ij =  K˜ o –
2G 
tr∆ε + 2G∆ε ij
 3  (10)
1.3.1 Mechanical Properties
K˜ o = f ( trσ ij – κp)
To illustrate this, let us consider the experimental data shown
in Figure 6 (Boutéca et al., 1994). The drained bulk modulus where the new “effective stress” in f has nothing to do with
increases with the confining pressure. Similarly, there is a the one previously introduced for the constitutive law.
shift in the curves obtained with a pore pressure equal to Instead of relating stresses and strains it relates a coefficient
1 MPa and a pore pressure equal to 51 MPa. This kind of of the constitutive law with stresses and pressure.

HP - 19.8% HP - 19.8%
10000 120

8000 100

80
6000 pp = 51 MPa
Ko

Pc

60
4000 pp = 1 MPa pp = 51 MPa
40
2000
20 pp = 1 MPa
0 0
0 50 100 0 0.005 0.010 0.015
pc (MPa) dV/V

Figure 6 Figure 7
Bulk modulus under drained conditions Reservoir sandstone Volumetric strain as a function of the confining pressure
of initial porosity equal to 19.8% (after Boutéca et al., 1994). (after Boutéca et al., 1994).
708 Oil & Gas Science and Technology – Rev. IFP, Vol. 54 (1999), No. 6

At the microscopic scale, the evolution of the rock Hence the nonlinearity is not to be defined through an a
properties corresponds to Hertz contacts—localization of the priori evolution of the bulk modulus, elaborated from
∼
stress at the grain contact—and elastic behavior of small experimental curves in some (Kt = k o (σ, p)) plot. It has to be
cracks in the grain or in the cement. Herein, the model is derived from the enrichment of the thermodynamic potential.
purely elastic, as opposed to Hertz-Mindlin model where
friction is involved at the micro scale (see Section 2.1.3). 1.3.2 Transport Properties
This departing from the constant value of Ko, is reflected in
When measuring transport properties for rocks submitted to
the stress strain curve which is no longer a straight line and
pressure and stress evolution (David et al., 1994; Sarda et al.,
becomes non-linear as shown in Figure 7.
1998), an effective stress is often defined. In fact it appears
The obvious meaning of Figure 7 is that the hypotheses
that a description at the microscale level is needed.
that led to a linear relationship between stresses and strains as
There is a clear definition of the effective stress for the
expressed in Equation (3) are no longer valid. Then, one has
constitutive law. Hence there is a clear definition of an
to come back to the origin of this linear relationship which is
effective stress for the strains. The next step is to move from
the quadratic form of the strain energy. Instead of assuming a
strains to permeability which implies a knowledge of the
quadratic form, let us assume a third order one. Obviously,
structural evolution at the pore scale. At that scale, cracks and
when deriving with respect to the strain, this will introduce
grain deformation or displacement, will play very different
terms of the second order in strain wich will produce non-
roles and their contribution will have to be accounted for in
linearity.
specific ways.
Based on a microscopic description, Biot (1973) defined
such a third order potential and we applied it (Boutéca et al.,
1994) to the sandstone shown in Figures 6 and 7. It results in 2 MICROSCOPIC SCALE: EFFECTIVE PROPERTIES
a nonlinear tangent drained modulus and a nonlinear tangent
Biot coefficient that depend (linearly) on εv + p/Ks where εv is Microscopic analysis allows to identify the relevant defor-
the volumetric strain: mation micromechanisms. It leads to express macroscopic
thermodynamic parameters in terms of microvariables which
dσ = 2Gdε + 3(λ + λnl ) dε + 3(b + b nl ) dp p
can be obtained from various observations. Such an analysis

[ ]
= 2G + 3(λ + λnl ) dε + 3(b + b nl ) dp p
(11)
is a required step in order to discuss possible scale effects on a
sound basis. It may be also a crucial point in preventing any
extrapolation of a macroscopic law beyond its range of
( 4 F + 2 D)  pp  ( 4 F + 2 D)  pp 
validity. We examine in the following the mechanical
λnl = ε +  b nl = – ε + 
3  Ks  3K s  Ks  behavior and more specifically the elastic properties. This
refers to Sections 1.1.2 and 1.3.1 above. We do not consider
dσ  dσ ≡ p 
= Kt dε + bt dp p plastic behavior. The microscopic analysis of failure and that
3  3 c

of transport properties is also beyond the scope of this paper.
where dσ is the incremental volumetric stress, F and D are Our restrictive choice is dictated by the fact that elastic wave
constant rock properties. velocities play a major role for crustal exploration and that our
The experimental evidence is shown in Figure 8 where we present understanding of elastic behavior and properties is
∼
plotted Kt (≡ Ko) as a function of – (εv + p/Ks). much more advanced than for the other issues considered in
Section 1.
15000
Linear regression
2.1 Elastic Moduli of High Porosity Rocks

10000 At a microscopic scale, different descriptions of the rocks are

possible, and, depending on the rock nature, some are more
appropriate than others. High porosity, poorly consolidated
Kt

5000 and unconsolidated rocks can be adequately modeled as

granular assemblages where contact stiffnesses play the major
role. The calculation of elastic constants from micromodels is
0 in that case a two steps procedure: the first step consists in
-0.02 -0.015 -0.01 -0.005 0
deriving the contact stiffnesses of two grains, and the second
- (dV/V + p/Ks) one is an averaging process over a random packing of
Figure 8 identical spherical particles. As far as elastic properties are
concerned, the rock can be viewed as a network of elastic
Tangential bulk modulus (after Boutéca et al., 1994).
springs, each spring being a grain contact (Fig. 9).
 M Boutéca and Y Guéguen / Mechanical Properties of Rocks: Pore Pressure and Scale Effects 709

2.1.1 Contact Stiffnesses different from the above Hertz-Mindlin ones but his
averaging procedure constitutes an independent step which
All such models rely on the Hertz’s and Mindlin’s contact allows to express the effective bulk and shear moduli as
theory (Johnson, 1985). In the Hertz model, two identical functions of Dn and Dt, whatever the precise values of the
spherical grains of radius R (Fig. 9) are deformed by a normal stiffnesses are. With the Hertz-Mindlin stiffnesses, one gets
force N. The radius of the contact area is: the following expressions for the effective bulk and shear
1/ 3 modulus Keff and the Geff :
 3 NR (1 – ν s ) 
a=  
 8Gs  C(1 – Φ) C(1 – Φ)
Keff = Dn Geff = ( Dn + 1.5 Dt )
where νs and Gs are respectively the Poisson’s ratio and the 12 πR 20 πR
shear modulus of the grain. The normal displacement of a where C is the coordination number (average number of
sphere center relative to the contact area center is un = a2/R contacts per sphere, which is close to 9 for dense random
and the normal stiffness Dn = ∂N/∂un is: pack).

4Gs a 2.1.3 Predictions of the Model

Dn =
1– ν s
A simple test of the above model can be obtained by
Mindlin model considers an additional tangential load T comparing predicted Poisson’s ratio values to measured
superimposed afterwards on the previous system. Assuming values on glass beads and high porosity sandstones
that there is no partial slip in the contact area, the tangential (Domenico, 1977). As shown by Winkler (1983) the above
stiffness Dt = ∂T/∂ut is: model predicts values (<0.05) much lower than those
observed (0.15) as shown on Figure 10. Although surface
8Gs a roughness may explain some of these discrepancies at low
Dt =
2 – νs pressures (Palciauskas, 1992; Manificat and Guéguen, 1998),
it is likely that other effects have to be accounted for. In the
where ut is the tangential displacement of a sphere center case of sandstones, the presence of cement at grain contacts
relative to the contact area center. modify strongly the elastic properties (Dvorkin et al., 1994).
Soft cements increase Poisson’s ratio so that even a small
2.1.2 Elastic Moduli
amount of cement could explain the above discrepancies.
Assuming a random isotropic packing, Digby (1981) derived Another test is provided by the pressure dependence. It
the effective elastic constants from the contact stiffnesses. He results from the contact model that the spring constants
used relationships for stiffnesses which are somewhat depend on stress so that the model is a non linear elastic one.

0.15

0.10
ν

0.05

h 0.0
0.0 0.1 0.2 0.3 0.4 0.5
νs

Figure 9 Figure 10
Granular model. The rock is described as a random packing Poisson’s ratio of a random packing of spheres (solid line)
of identical spheres of radius R. Normal displacement at from Digby’s model as compared to high porosity sandstones
granular contact h = 2un. Tangential displacement is not data (shaded area).
shown.
710 Oil & Gas Science and Technology – Rev. IFP, Vol. 54 (1999), No. 6

At zero stress, the elastic moduli should vanish and they are schemes have been used by different researchers in order to
predicted to vary as (P)1/3 from the above stiffnesses achieve this. They have their roots in the effective media
expressions. The results reported by Domenico (1977) for theories of physics. This diversity may be confusing and we
unconsolidated glass beads follow more closely a (P)1/2 attempt here to clarify the problem. The calculation of elastic
dependence, whereas those of Toksöz et al. (1979) for Berea constants from micro-models is obtained through the choice
sandstone show a much weaker dependence. The soft shell of a specific scheme and of a precise assumption on the
model proposed by de Gennes (1996) explains a (P)1/2 geometrical inclusion shape. In the case of pores, the moduli
dependence but is not appropriate for sandstone. Again the are obtained as functions of porosity Φ or more generally as
presence of cement is likely to affect strongly the pressure functions of the ratio Φ/α for ellipsoidal pores of aspect
dependence. As shown by Wong and Wu (1995), normal ratio α (Guéguen and Palciauskas, 1994). In the case of
stiffness is not sensitive to pressure in a cemented system. cracks (which can be considered as the limiting case of zero
Finally, due to the path dependent nature of contact volume inclusions), the moduli are functions of the crack
forces, effective elastic constants are also path dependent. At density parameter defined in 2-D as ρ = 1/A ∑ (li)2 where
the microscopic scale, each contact is assumed to be rough. each rectilinear crack has a length li and A is the area of the
This is why a no slip assumption is made. But this means considered element.
also that contact forces are not purely elastic since friction is
involved. Contact forces are not derivable from a 2.2.1 Effective Media Schemes
thermodynamic potential. Because of this fundamental
complexity of contact models, it is not possible to establish a The elastic moduli are calculated by replacing the
unique result for elastic constants independently of the stress heterogeneous rock by a simple homogeneous equivalent
path trajectory. The medium is not an hyperelastic medium medium. The properties of this equivalent medium are the
but an hypoelastic one. effective properties. The use of the word effective here should
not be confused with that of Section 1 where effective stress
has been discussed. The effective stress concept assumes that
2.2 Elastic Moduli of Low Porosity Rocks
the medium is statistically homogeneous and replaces the real
The above model breaks down when porosity is below some rock by an equivalent homogenized medium. The purpose of
threshold. Vernik (1997) estimated the threshold value to be Section 2 is precisely to examine how the elastic properties of
approximately 30%. Low porosity rocks can be described by this equivalent (effective) medium are derived.
using inclusion-based models that view the rock as a solid The simplest model is the one which assumes negligible
matrix with randomly embedded inclusions representing interactions between inclusions. It may be called a first order
individual pores and cracks (Fig. 11). Several approximate perturbation scheme. In that case, the medium is a simple

1.0

0.8

0.6
σ oij = constant
KIKs

0.4

0.2
Ω D-Ω
0.0
0.0 0.2 0.4 0.6 0.8 1.0

Figure 11 Figure 12
Inclusion model: a pore or crack is described as an inclusion Normalized bulk modulus as a function of porosity for the
of specific shape (Ω) included in the rock matrix (D – Ω). inclusion model. The assumptions are: non interacting inclu-
sions and spherical pores. The matrix bulk modulus is Ks =
38 GPa, the matrix shear modulus is Gs = 22.8 GPa. The pores
are saturated with a fluid of bulk modulus Kf = 2.2 Gpa.
 M Boutéca and Y Guéguen / Mechanical Properties of Rocks: Pore Pressure and Scale Effects 711

composite which behaves as if it was made of a single takes approximately into account the inclusions inter-
inclusion within a matrix. The elastic compliances are derived actions. The simplest effective field is the volumetric
by submitting this medium to a constant stress at infinity, and average (Mori-Tanaka method).
the compliances are found to be linear functions of porosity Φ
or crack density ρ. Elastic stiffnesses are thus non linear 2.2.2 Predictions of the Model
functions of Φ or ρ (Fig. 12). In the limit of a diluted
concentration of inclusions, linearization of these expressions Effective medium theory has been successfully used to
is however possible (Le Ravalec and Guéguen, 1996a). predict the properties of many heterogeneous systems. The
More sophisticated models account for inclusions various possible schemes have each some specific advan-
interactions in some approximate ways. Following Kachanov tages and drawbacks which are discussed below.
(1993), we can consider that most of these schemes calculate The self-consistent (effective matrix) scheme predicts a
the effective elastic constants by analyzing one isolated strong decrease of moduli down to zero at some cut-off
inclusion in either an effective matrix or an effective field. porosity value, or crack density value (Fig. 13). For that
This corresponds to a basic subdivision of effective media reason, this model has been attractive and widely used. It can
schemes in two groups: be shown however that this cut-off does not have a physical
– The effective matrix scheme assumes that the inclusion is meaning and that it corresponds to an extrapolation of the
placed within a matrix which has effective moduli. If the model beyond its range of validity as explained below
inclusions are placed in a single step, this scheme is known (Kachanov, 1993, Guéguen et al., 1997). The self-consistent
as the self-consistent one. The interactions between model overestimates the moduli decrease.
inclusions are approximately taken into account by The differential self-consistent (effective matrix) scheme
replacing the background medium with the as-yet- does not predict any cut-off and exhibits a less severe
unknown effective medium. The self-consistent scheme is decrease of moduli with pores/cracks content. It predicts
an implicit scheme which assumes the solution known in remarkably well the porosity dependence, at least for
order to calculate it. If the inclusions are placed in spherical inclusions (Fig. 13a). Since the pores and cracks are
infinitesimal increments, the scheme is an iterative one added incrementally, this model can be extended to consider
and is known as the differential self-consistent scheme various families of pores and cracks, introducing for instance
(Le Ravalec and Guéguen, 1996a). a range of aspect ratios. But the elastic moduli depend on the
– The method of effective field approximates inclusions order in which the incremental additions are done. This is a
interactions by assuming that an inclusion is submitted to drawback specific to that method. A possible way to go
an effective stress field. This field is not the one applied to around this difficulty is to realize several numerical
the medium but some average of the real stress field which simulations with the same input data but with a different

50 1.0
Effective bulk modulus (GPa)

40 0.8

30 0.6 First order

E/E0

perturbation
First order
20 perturbation 0.4

D
10 D 0.2
SC SC
0 0.0
0 20 40 60 80 100 0.0 0.2 0.4 0.6
a Porosity (%) b ρ

Figure 13a Figure 13b

Effective Bulk moduli for glass foam samples as a function Effective Young moduli for randomly oriented cracks as a
of porosity (Walsh et al., 1965). The predictions are those of function of crack density. Numerical simulations are shown
the first order perturbation model (noninteracting cracks), the as vertical bars (Kachanov, 1993). The predictions are the
self-consistent model and the differential self-consistent same as above.
model.
712 Oil & Gas Science and Technology – Rev. IFP, Vol. 54 (1999), No. 6

order of introduction of inclusions. Averaging over a large because they provide a nice example in which combining
number of numerical simulations provides then valid results. micro- and macroanalysis proves to be very fruitful.
In practice, it has been shown that one hundred simulations is
sufficient to get stable solutions (Le Ravalec and Guéguen, 2.3.1 Poroelasticity
1996a).
Poroelasticity describes the mechanical behaviour of a
The effective field method appears to be a better saturated porous rock at moderate pressure and temperature
approximation in general than the effective matrix method. conditions (Biot, 1941; Rice and Cleary, 1976). As discussed
An important and interesting result (Kachanov, 1993) is that, in Section 1.1.2, it is a thermodynamic theory which
in the case of randomly oriented cracks, the effective moduli expresses strains as linear functions of effective stresses. It
predicted with the Mori-Tanaka method are identical to those can be extended to nonlinear behaviour (Biot, 1973; Boutéca
obtained from the approximation of noninteracting cracks et al., 1996) and to anisotropic media (Brown and Korringa,
(i.e., from first order perturbation calculations). Kachanov 1975) but we restrict here to the simple case of linearity and
pointed out clearly that this is not a coincidence but results isotropy. A basic result of this theory is that the drained bulk
from the fact that introducing zero volume cracks does not modulus Kd is related to the undrained bulk modulus Ku
change the average stress in a solid if tractions are prescribed through the following equation (Gassmann equation):
on the boundary. Consequently, the average effective field
b2
in the Mori-Tanaka approximation is exactly the original Ku = K d +
Φ (b – Φ)
field and thus the calculation is that of a noninteracting +
approximation. Numerical simulations show that the non- Kf Ks
interacting cracks approximation is the best one in that case.
(Kachanov, 1993). If cracks are described as flat ellipsoids of Both shear moduli, drained and undrained, are identical.
nonzero volume, the differential self consistent results and The b coefficient has been defined in Section 1.1.2.
the noninteracting results remain very close (Le Ravalec and Poroelasticity does not provide of course any information
Guéguen, 1996b). on the above moduli. But the micromodels described in
Section 2 do. Both drained and undrained moduli are
Up to what degree of heterogeneity can effective media effective moduli. The drained modulus corresponds to the
theory be used? This theory is an approximation which relies situation where fluid can flow at constant pressure. It can be
on the assumption that the medium is statistically homoge- obtained from effective media theory by considering dry
neous. As stated above, the critical thresholds or cut-off inclusions, i.e. empty cracks and pores (low porosity rocks)
derived from the self consistent scheme do not fit with real or a dry granular assemblage (high porosity rocks). If the
data. This is not surprising since the existence of such drained modulus is known, using the above equation
thresholds is linked to clustering effects which become provides then the undrained modulus. It is this last one which
important in a strongly nonhomogeneous medium (Guéguen has to be used for elastic waves velocities.
et al., 1997). Using the previous scheme near the threshold is
clearly extrapolating it beyond its range of validity. The 2.3.2 Scale and Frequency Effects
differential scheme should be preferred as it does not exhibit
such thresholds and accordingly does not overestimate so Pushing the analysis one step further, it is possible to
much the decrease of moduli. The differential scheme itself examine scale and frequency effects. The fluctuating stresses
has also a limited range of validity however. In the case of a caused by the passing of a seismic wave in a porous saturated
strongly heterogeneous medium, percolation theory is the rock induce pore pressure gradients on the scale of individual
appropriate tool since clustering effects are a central issue of pores and cracks. At high frequencies, the gradients are
percolation theory. It can be shown (Guéguen et al., 1997) unrelaxed and the rock is stiffer than at low frequencies
that various thresholds have to be distinguished depending on where pore pressures are equilibrated through the pore space.
the properties of interest. In particular, the mechanical Gassmann equation applies at low frequencies (<100Hz)
threshold and the permeability threshold are completely because in that case there is sufficient time for fluid to flow
different. For cracked rocks, the first one is reached at high so that fluid pressure is equilibrated. Laboratory ultrasonic
crack density (of the order of 1) but the second one is reached measurements (106 Hz) are likely to be high frequency
at much lower values (of the order of 0.1). measurements in the above sense of an unrelaxed pore fluid
state. How can we calculate the appropriate elastic moduli in
such a situation? The answer is given by using effective
2.3 Pore Pressure and Scale Effects: Combining media models once more, but with fluid inclusions and no
Micro- and Macroanalysis longer with dry inclusions. For instance the differential self-
consistent scheme considers that the inclusions are isolated.
We focus in the following on the elastic properties because of The pore fluid pressure within each inclusion is consequently
their importance for underground exploration and also variable from one to the other. This is an unrelaxed state and
 M Boutéca and Y Guéguen / Mechanical Properties of Rocks: Pore Pressure and Scale Effects 713

the modulus derived from this model is an unrelaxed one. There, the effective stress concept, as pointed out by Coussy
Combining porelasticity and effective media theory allows (1991) is a consequence of the constitutive law. At the
thus to derive high and low frequency moduli and to predict microscopic level, the corresponding mechanisms to
elastic wave dispersion (Fig. 14) (Le Ravalec et al., 1996a be considered are the grain deformation, the cement
and 1996b). deformation and the grain displacement.
The concept is then applied to criteria (shear failure, limit
of the elastic domain, plasticity criterion) and the effective
stress concept involves a stress-stress relation. At the
microscopic level, the mechanisms are inter and intra-
granular cracking leading to rock splitting.
Pocket and local Local flow No flow
The concept is then applied to rock properties changes.
For mechanical properties the relation between mechanical
Velocity

properties, stresses and pressure should be derived from

energy considerations. For transport properties, it should be
derived from microscopic considerations, linking structural
changes with pressure and macroscopic stresses.

Effective stress concept

fCG fC L Frequency
σ, p
Figure 14
Constitutive Criteria Rock properties
Velocity dispersion resulting from local and pocket flow.
Low frequency velocity is derived from Gassmann equation law: σ’, ε σ’, σ’ changes
and effective media theory. High frequency velocity is
derived from effective media theory.
Mechanical Transport
properties properties

A realistic situation is that of partial saturation. We have Link with

structural
so far only considered dry or totally saturated rocks. The changes
saturation may be variable from one point to the other. In that
case, one can describe the rock as made of pockets (or
Figure 15
patches) of saturation S1 in a matrix of saturation S2. Both the
The effective stress concept in saturated media.
pockets and the matrix are made of a mixture of dry and
saturated inclusions. Using for instance the differential self-
consistent scheme of effective media theory, it is possible to At a microscopic scale, use of specific descriptions of
derive the unrelaxed and relaxed moduli of a pocket and of the rock microstructure allows to calculate the effective
the matrix. If the frequency is low enough for the fluid properties of the rock. Such a microscopic description is
pressure to be equilibrated at the pocket scale, but high complementary of the macroscopic one. Combining both
enough to be unequilibrated at the matrix scale, fluid pressure allows to investigate scale effects. The example of elastic
gradients exist at the pocket scale and velocity dispersion is properties has been developed and it has been shown that the
again expected (Fig. 14). frequency dependence of elastic waves in saturated rocks is
The dispersion mechanisms at local scale and at pocket in fact a scale effect which can be accounted for within the
scale are frequency and scale effects which have to be taken framework of poroelasticity and effective medium theory
into account in many different situations: extrapolation from
laboratory data to field data, interpretation of saturation
hysteresis, interpretation of seismic waves anomalies. ACKNOWLEDGMENTS

The authors would like to thank their colleagues Jean-

CONCLUSION Paul Sarda, Pascal Longuemare, Olivier Vincké and Atef
Onaisi for fruitful discussions. Olivier Coussy and Luc
The concept of effective stress in rock mechanics covers very Dormieux have contributed to the process of clarifying the
different notions, as summarized in Figure 15. It firstly ideas herein presented. We also benefited from discussions
applies to the relationship between stresses and strains. with Pr. Teng-Fong Wong and Pr. Jim Rice.
714 Oil & Gas Science and Technology – Rev. IFP, Vol. 54 (1999), No. 6

REFERENCES Lade, P.V. and de Boer, R. (1997) The Concept of Effective

Stress for Soil, Concrete and Rock. Géotechnique, 47, 1, 61-78.
Biot, M.A. (1941) General Theory of Three Dimensional Con- Le Ravalec, M., and Guéguen, Y. (1996a) High and Low
solidation. J. Appl. Phys., 12, 155-164. Frequency Elastic Moduli for a Saturated Porous/Cracked Rock-
Biot, M.A. (1973) Nonlinear and Semilinear Rheology of Porous Differential Self-Consistent and Poroelastic Theories. Geophysics,
Solids. J. Geoph. Res., 78, 23, 4924-4937. 61, 1080-1094.
Boutéca, M.J., Bary, D. and Maury, V. (1993) An Experimental Le Ravalec, M. and Guéguen, Y. (1996b) Comment on “The
Study of the Contribution of the Pore Volume Change to Elastic Modulus of Media Containing Strongly Interacting
Production Within the Framework of Biot’s Theory. Proceedings Antiplane Cracks” by Paul M. Davis and Leon Knopoff.
of the 1993 Society of Core Analysts Technical Conference. Geophys. Res. Lett., 101, 25373-25375.
Boutéca, M.J., Bary, D., Piau, J.M., Kessler, N., Boisson, M. and Le Ravalec, M., Guéguen, Y. and Chelidze, T. (1996a) Elastic
Fourmaintraux D. (1994) Contribution of Poroelasticity to Waves Velocities in Partially Saturated Rocks: Saturation
Reservoir Engineering: Lab Experiments, Application to Core Hysteresis. J. Geophys. Res., 101, 837-844.
Decompression and Implication in HP-HT Reservoirs Depletion. Le Ravalec, M., Guéguen, Y. and Chelidze, T. (1996b) The
Paper SPE/ISRM 28093. Magnitude of Velocity Anomalies Prior to Earthquakes. J.
Geophys. Res., 11, 217-223.
Boutéca, M.J. and Sarda, J.P. (1995) Experimental Measurements
of Thermoporoelastic Coefficients. Mechanics in Porous Media, Manificat, G., and Guéguen, Y. (1998) What Does Control Vp/Vs
Charlez, P. (ed.), Balkema, 31-41. in Granular Rocks? Geophys. Res. Lett., 25, 381-384.
Brown, R. and Korringa, J. (1975) On the Dependence of the Palciauskas, V. (1992) Compressional to Shear Wave Ratio of
Elastic Properties of a Porous Rock on the Compressibility of the Granular Rocks: Role of Rough Grain Contacts. Geophys. Res.
Pore Fluid. Geophysics, 40, 608-616. Lett., 19, 1683-1686.
Coussy, O. (1991) Mécanique des milieux poreux, Éditions Rice, J.R., and Cleary, M. (1976) Some Basic Stress Diffusion
Technip, Paris. Solutions for Fluid Saturated Porous Media with Compressible
Constituents. Rev. Geophys. Space Phys., 14, 227-241.
David, C., Wong, T.F., Zhu, W. and Zhang, J. (1994) Laboratory
Measurement of Compaction - Induced Permeability Change in Sarda J.P., Ferfera, F.M.R., Vincké‚ O., Boutéca‚ M. and
Porous Rocks: Implications for the Generation and Maintenance Longuemare, P. (1998) Experimental Study of the Stress Paths
of Pore Pressure Excess in the Crust. PAGEOPH, 143, 1/2/3. Influence on Monophasic Permeability Evolution. SCA 9827.
De Gennes, P.G. (1996) Static Compression of a Granular Toksoz, M.N., Johnston, D.H. and Timur, A. (1979) Attenuation
Medium: the Soft Shell Model. Europhysics Letters, 35, 145-149. of Seismic Waves in Dry Saturated Rocks: I Laboratory
Measurements. Geophysics, 44, 681-690.
Digby, P.J. (1981) The Effective Elastic Moduli of Porous
Granular Rocks. J. Appl.Mech., 48, 803-808. Vernick, L. (1997) Predicting Porosity from Acoustic Velocities
in Siliciclastics: A New Look. Geophysics, 62, 1, 118-128.
Domenico, S.N. (1977) Elastic Properties of Unconsolidated Vincké, O., Boutéca, M.J., Piau, J.M. and Fourmaintraux, D.
Porous Sand Reservoirs. Geophysics, 42, 1339-1368. (1998) Study of the Effective Stress at Failure. Proceedings of the
Dvorkin, J., Nur, A. and Yin, H. (1994) Effective Properties of Biot Conference on Poromechanics.
Cemented Granular Materials. Mech. Mater., 18, 351-366. Walsh, J.B., Brace, W.F. and England, A.W. (1965) Effect of
Guéguen, Y., Chelidze, T. and Le Ravalec, M. (1997) Porosity on Compressibility of Glass. J. Am. Ceram. Soc., 48,
Microstructures, Percolation Thresholds, and Rock Physical 605-608.
Properties. Tectonophysics, 279, 23-35. Winkler, K.W. (1983) Contact Stiffness in Granular Porous
Guéguen, Y. and Palciauskas, V. (1994) Introduction to the Materials: Comparison between Theory and Experiment. Geophys.
Physics of Rocks, Princeton University Press, Princeton, NJ. Res. Lett., 10, 1073-1076.
Johnson, K.L. (1985) Contact Mechanics, Cambridge University Wong, T.F., and Wu, L.C. (1995) Tensile Stress Concentration and
Press. Compressive Failure in Cemented Granular Material. Geophys.
Res. Lett., 22, 1649-1652.
Kachanov, M. (1993) Elastic Solids with Many Cracks and
Related Problems. Adv. Appl. Mech., 30, 259-445. Final manuscript received in July 1999