You are on page 1of 19

International Journal of Rock Mechanics & Mining Sciences 142 (2021) 104754

Contents lists available at ScienceDirect

International Journal of Rock Mechanics and Mining Sciences


journal homepage: www.elsevier.com/locate/ijrmms

Intrinsic material constants of poroelasticity


Alexander H.D. Cheng
University of Mississippi, Oxford, MS, USA

ARTICLE INFO ABSTRACT

Keywords: Biot and Willis (1957) performed micromechanical analysis on isotropic poroelastic materials and presented
Poroelasticity three material constants, the jacketed compressibility, the unjacketed compressibility, and the coefficient of
Micromechanics fluid content, to characterize the volumetric deformation. Brown and Korringa (1975) and Rice and Cleary
Constitutive equation
(1976) recast them into a drained bulk modulus 𝐾, an unjacketed bulk modulus 𝐾𝑠′ , and an unjacketed pore
Intrinsic material constant
modulus 𝐾𝑠′′ . The constant 𝐾𝑠′′ is tied to the storage capacity of porous medium; hence is highly important in
Unjacketed pore modulus
Effective stress
petroleum and geosequestration applications. It is, however, difficult to measure in the laboratory due to the
Stress dependent constitutive law lack of a direct way to observe pore volume deformation.
In this work, a physics based approach is used to explore these micromechanical constants. Three intrinsic
material constants that isolate the physical mechanisms, an unjacketed solid bulk modulus 𝐾𝑠 , a drained
pore modulus 𝐾𝑝 , and a micro inhomogeneity and anisotropy modulus 𝐾𝜓 , are presented to characterize the
porous material. Bounds on material constants are developed based on thermodynamics and elastic stability
requirements. Past attempts in measuring 𝐾𝑠′′ using direct or indirect approaches are compiled. Most of the
efforts failed to satisfy the theoretical bounds. Only those attempted to measure the change in pore volume
passed the test. The intrinsic constants are used to interpret the bulk material constants, such as the drained and
undrained bulk modulus, and the storage and Skempton pore pressure coefficient. The total volume change and
the storage capacity can be partitioned into those attributed to solid, pore, fluid, and the microinhomogeneity
compressibilities. A published data set on stress dependent Skempton pore pressure coefficient and bulk
modulus is reanalyzed. The data are matched by the calibration of a single of 𝐾𝜓 value.

1. Introduction
porous solid, and 𝑀 is the combined property of solid and fluid.5
To gain insight into the deformation within the porous frame,
The theory of poroelasticity was pioneered by Biot1 for the mod-
Biot and Willis6 further considered the pore volume deformation. The
eling of mechanical responses of porous solid infiltrated by a fluid.
formulated constitutive laws are generally known as the microme-
Together with its dynamics version,2,3 the theory has become widely
chanical analysis. They presented five material constants: the jacketed
accepted in fields of geomechanics and geophysics, as well as other
compressibility 𝜅, the unjacketed compressibility 𝛿, the coefficient of
fields. It has been applied to problems such as consolidation of soil
fluid content 𝛾, the fluid compressibility 𝑐, and the porosity 𝜙. Method
under surcharge, land subsidence due to fluid withdrawal, stability of
for their measurement in the laboratory were suggested. The three
excavated space, landslide, liquefaction, seismicity, CO2 sequestration,
macroscopic constitutive constants, 𝐾, 𝛼, and 𝑀, can be evaluated from
etc.4
these five micromechanical constants.4,5
With practical applications in mind, Biot1 used the phenomenolog-
There were many subsequent analyses7–12 that led to consistent
ical approach to present a continuum model for the solution by partial
result with Biot and Willis.6 In particular, Brown and Korringa7 and
differential equations. He adopted a set of observable dynamic and
Rice and Cleary12 recast the three constants, 𝜅, 𝛿, and 𝛾, into a drained
kinematic variables, such as the total stress, pore pressure, displace-
bulk modulus 𝐾, an unjacketed bulk modulus 𝐾𝑠′ , and an unjacketed
ment of the porous frame, and expelled fluid volume from the frame,
pore modulus 𝐾𝑠′′ . These constants are properties of the solid and its
as the modeling variables. The constitutive equations based on these
pore structure only. Together with porosity and fluid compressibility,
variables admit four independent material constants under material
they form the equivalent set of five constants of Biot and Willis. Also
isotropy, which can be expressed as a drained bulk modulus 𝐾, a shear
of interest is a special model by Gassmann,13 known as the ideal
modulus 𝐺, a Biot effective stress coefficient 𝛼, and a Biot modulus 𝑀.
porous medium model, which assumes that the solid is made of a single
Of these constants, three of them, 𝐾, 𝐺, and 𝛼, are properties of the

E-mail address: acheng@olemiss.edu.

https://doi.org/10.1016/j.ijrmms.2021.104754
Received 13 December 2020; Received in revised form 9 March 2021; Accepted 26 March 2021
Available online 15 April 2021
1365-1609/© 2021 Elsevier Ltd. All rights reserved.
A.H.D. Cheng International Journal of Rock Mechanics and Mining Sciences 142 (2021) 104754

isotropic mineral with bulk modulus 𝐾𝑚 . The assumption led to the The intrinsic constants allow the separation of the bulk material
condition 𝐾𝑠′ = 𝐾𝑠′′ = 𝐾𝑚 , and the reduction of one material constant. constants into identifiable physical mechanisms. The proportion of con-
Many geomaterials, such as sedimentary rocks, are made of several tribution by each mechanism can be assessed. In addition, the partition
minerals, together with an amount of soft clay. Even in a relatively pure allows a rational way to model stress dependent behaviors. The three
mineral, it can contain occluded voids that do not communicate with intrinsic constants should be separately modeled and combined to give
the flow channels. These pores need to be considered as a part of the the composite behavior of the bulk constants.
solid, effectively rendering it inhomogeneous.5 Hence it is important to The work is organized as follows. Section 2 gives a quick overview
use the general model that distinguishes between the two constants, 𝐾𝑠′ of the macroscopic constitutive laws of Biot.1 The Biot and Willis6
and 𝐾𝑠′′ . micromechanical analysis and its subsequent interpretation by others
The unjacketed pore modulus 𝐾𝑠′′ is associated with the pore volume are presented in Section 3. Section 4 discusses the issues in laboratory
change under the unjacketed test condition that the pore pressure is test and compiles the past attempts in measuring 𝐾𝑠′′ . The intrinsic ma-
kept at the same as the confining stress. This constant is tied to the terial constants are derived in Section 5. An abbreviated presentation
storage capacity of porous medium. It has the effect on the produc- of variational work energy principles are given in Section 6. Bounds of
tion of hydrocarbon, formation compaction and surface subsidence, material constants are established in Section 7. The intrinsic constants
injection induced seismicity, and other petroleum and geosequestration are used to explore the bulk constants for physical insights, which are
applications. It is an input parameter for reservoir simulators.14–16 presented in Section 8. Stress dependent model is discussed in Section 9
The measurement of the constants 𝐾 and 𝐾𝑠′ is based on the obser- with the support of a reanalysis of a published data set. Section 10 gives
vation of deformation of porous frame, and is routinely conducted in a summary and conclusion.
the laboratories. The measurement of 𝐾𝑠′′ , on the other hand, is more
difficult, because it requires the knowledge of change in the internal 2. Macroscopic constitutive laws
pore space. One way to obtain pore volume change is to infer it from
the fluid expelled from the test sample. However, when such action Biot1 was the first to establish the constitutive laws of poroelasticity
involves a pore pressure change, the expelled fluid volume is not equal at the macroscopic level. The laws are of phenomenological nature.
to the pore volume change. Calibration is needed for the effect of fluid That is, they are built on properties that are easily observable, and
compressibility, as well as the compliances of the instrument, such the relations among them are generally empirical, not based on the
as the tubing, valve, etc. An inaccurate estimate directly impact the first principles. Biot chose the total stress tensor 𝜎𝑖𝑗 and the pore
result. As a way to side step the issue, a different constant, such as the pressure 𝑝 as the dynamic variables. For their energy conjugates, he
Skempton pore pressure coefficient 𝐵 or the undrained bulk modulus chose the solid frame strain tensor 𝑒𝑖𝑗 , and the variation in fluid content
𝐾𝑢 , which does not require the knowledge of pore space, is measured. 𝜁, that is, the volume of fluid gained per unit volume of frame, as the
The 𝐾𝑠′′ value is obtained by inversion based on known relations among kinematic variables. These variables can be monitored and controlled in
the constants. Such determination, however, is prone to the sensitivity an application setting, and used as boundary conditions in the solution
of inversion to the perturbation (error) of the parameters. In fact, we of partial differential equations.
shall show that all results produced by this approach are questionable, Assuming material isotropy, the constitutive equations can be ex-
because they fail to fulfill the bound requirements of material constants pressed as follows4,5,18 :
derived in this work. ( )
2𝐺
The macroscopic constants, such as 𝐾, 𝛼, 𝑀, 𝐵, 𝐾𝑢 , and the mi- 𝜎𝑖𝑗 = 𝐾 − 𝛿𝑖𝑗 𝑒 + 2𝐺𝑒𝑖𝑗 − 𝛼𝛿𝑖𝑗 𝑝, (1)
3
cromechanical ones, such as 𝐾𝑠′ and 𝐾𝑠′′ , are phenomenological con- 𝑝 = −𝛼𝑀𝑒 + 𝑀𝜁 , (2)
stants. They generally represent a combined response of several phys-
ical mechanisms. For example, the drained bulk modulus 𝐾 is the in which 𝑒 = 𝑒𝑖𝑖 is the volumetric strain, 𝛿𝑖𝑗 the Kronecker delta, 𝐾
combined strength of the solid and the pore structure. The undrained the drained bulk modulus, 𝐺 the shear modulus, 𝛼 the Biot effective
bulk modulus 𝐾𝑢 further involves the strength of the fluid. The mech- stress coefficient, and 𝑀 the Biot modulus. In this work, we focus on
anisms for 𝐾𝑠′ and 𝐾𝑠′′ are not easily described. To gain a deeper the volumetric deformation only. Equation (1) can be contracted to
understanding of these material constants, it is desirable to identify 𝜎
𝑃 = − 𝑖𝑖 = −𝐾𝑒 + 𝛼𝑝, (3)
the basic deformation mechanisms, and tie them to the constants that 3
control the phenomena. where 𝑃 is mean total compressive stress. The set of equations, (2) and
In this work, we aim at performing a micromechanical analysis (3), contain three independent material constants, {𝐾, 𝛼, 𝑀}.
to bring out a clearer physical insight. In addition to modeling the If a drained test is performed, in which 𝑝 is maintained at a constant,
changes in the total (frame) and pore volume, we shall also examine we can measure the drained bulk modulus as
the solid volume and porosity. By this process, we form three intrinsic ( )
1 𝜕𝑒
material constants: an effective bulk modulus of the (composite) solid =− . (4)
𝐾 𝜕𝑃 𝑝
𝐾𝑠 , a drained pore modulus 𝐾𝑝 , and a modulus 𝐾𝜓 characterizing the
microanisotropy and microinhomogeneity of the solid phase. We also Or, we can assemble (2) and (3) into the following forms4
identify the effective stress laws for solid and porosity deformation. A 𝑃 = −𝐾𝑢 𝑒 + 𝛼𝑀𝜁 , (5)
reduction to the ideal porous medium is achieved by setting 𝐾𝜓 → ∞, 𝑀𝐾
with the other two constants intact. Relations between the bulk and 𝑝 = 𝐵𝑃 + 𝜁, (6)
𝐾𝑢
micromechanical constants, and the current intrinsic constants, are
where
established.
To support the micromechanical analysis, a parallel derivation is 𝐾𝑢 = 𝐾 + 𝛼 2 𝑀, (7)
presented using the volume averaging and variational energy
principles.4,17 Correspondences between the two approaches are estab- is the undrained bulk modulus, and
lished. The constitutive constants defining the quadratic form of energy 𝛼𝑀
𝐵= , (8)
allow a thermodynamics and stability analysis, leading to bounds of 𝐾 + 𝛼2 𝑀
material constants. A literature survey is conducted, and yields about is the Skempton pore pressure coefficient. Conducting an undrained
twenty reported 𝐾𝑠′′ values for various rocks. Applying the material test, 𝜁 = 0, we can measure these material constants as follows:
bounds to these values, we find that most of them failed the test. Only ( )
1 𝜕𝑒
five survived and are subjected to further analysis. = − , (9)
𝐾𝑢 𝜕𝑃 𝜁

2
A.H.D. Cheng International Journal of Rock Mechanics and Mining Sciences 142 (2021) 104754
( )
𝜕𝑝 These equations highlight the existence of the effective stresses for the
𝐵 = . (10)
𝜕𝑃 𝜁 volumetric deformation of the frame and pore space, as
As an alternative to the set {𝐾, 𝛼, 𝑀}, we can choose {𝐾, 𝐾𝑢 , 𝐵} as the
𝑃 ′′ = 𝑃 − 𝛼𝑝, (23)
independent constants, and express {𝛼, 𝑀} as follows
𝑃 ′′′ = 𝑃 − 𝛽𝑝, (24)
𝐾𝑢 − 𝐾
𝛼 = , (11)
𝐾𝑢 𝐵 where 𝑃 ′′ is the Biot effective stress. The constants 𝛼 and 𝛽 are the
𝐾𝑢2 𝐵 2 corresponding effective stress coefficients. By setting 𝑝 = 0 in (21), we
𝑀 = . (12) identify 𝐾 as the drained bulk modulus, same as that defined in (4).
𝐾𝑢 − 𝐾
Similarly, 𝐾𝑝 in (22) is the drained pore modulus. From (19), we find
3. Micromechanical constitutive laws 𝜙𝐾
𝐾𝑝 = , (25)
𝛼
We observe that the macroscopic laws treat the porous solid as a and only three of the constants are independent. Comparing (21) and
single continuum, and do not distinguish between the space occupied (22) with (17) and (18), we recognize
by the solid and the pores. When a sample is experiencing a volume
1 1 𝛼 𝛽
change, one part is derived from the solid, and the other from the pore 𝐶𝑏𝑐 = = 𝐶, 𝐶𝑝𝑐 = = 𝐶𝑝 , 𝐶𝑏𝑝 = = 𝜙𝐶𝑝 , 𝐶𝑝𝑝 = . (26)
𝐾 𝐾𝑝 𝐾 𝐾𝑝
space. The knowledge of pore space change is important in many en-
gineering applications, as discussed in the Introduction section. Hence In the above, we have introduced the compressibility notations 𝐶 and
Biot and Willis6 performed a micromechanical analysis that explicitly 𝐶𝑝 as the inverse of 𝐾 and 𝐾𝑝 .
modeled the pore volume. The constitutive equations can also be assembled into the forms
Consider a representative elementary volume (REV) of total volume suggested by Brown and Korringa,7 and Rice and Cleary,12 as1
𝑉 , which can be partitioned into a solid part 𝑉𝑠 , and a pore part 𝑉𝑝 . It 𝛥𝑉 𝑃′ 𝑝
is obvious that = − − ′, (27)
𝑉 𝐾 𝐾𝑠
𝑉 = 𝑉𝑠 + 𝑉𝑝 . (13) 𝛥𝑉𝑝 𝑃′ 𝑝
= − − , (28)
𝑉𝑝 𝐾𝑝 𝐾𝑠′′
We assume that the pore space is fully saturated by a fluid with volume
𝑉𝑓 , and where

𝑃 ′ = 𝑃 − 𝑝, (29)
𝑉𝑝 = 𝑉𝑓 . (14)
is the Terzaghi effective stress. In an unjacketed test (see Section 4), we
A porosity is defined as
have the condition 𝑃 ′ = 0. Hence 𝐾𝑠′ is the unjacketed bulk modulus,
𝑉𝑝 and 𝐾𝑠′′ the unjacketed pore modulus. We also find
𝜙= . (15)
𝑉
𝜙𝐾𝐾𝑠′
We note that all the volumes defined above are referred to their 𝐾𝑝 = , (30)
𝐾𝑠′ − 𝐾
original configuration before deformation, called the Lagrangian frame
by Coussy.19 or
We can express a change in volume as 1
𝐶𝑝 = (𝐶 − 𝐶𝑠′ ), (31)
𝜙
𝛥𝑉 = 𝛥𝑉𝑠 + 𝛥𝑉𝑝 . (16)
in which 𝐶𝑠′ = 1∕𝐾𝑠′ . Comparing (27) and (28) with (21) and (22), we
It is obvious that only two of the three volume changes can be selected observe the following relations
as independent variables; and we choose them as {𝛥𝑉 , 𝛥𝑉𝑝 }. Intro- 𝐶 ′
𝐾
ducing the dynamic variables {𝑃 , 𝑝}, these conjugate pairs allow the 𝛼 = 1− =1− 𝑠, (32)
𝐾𝑠′ 𝐶
construction of two linear constitutive equations20
𝐾𝑝 𝜙𝐶𝑠′′
𝛥𝑉 𝛽 = 1 − ′′ = 1 − . (33)
= −𝐶𝑏𝑐 𝑃 + 𝐶𝑏𝑝 𝑝, (17) 𝐾𝑠 𝐶 − 𝐶𝑠′
𝑉
𝛥𝑉𝑝 with 𝐶𝑠′′ = 1∕𝐾𝑠′′ . Also, from (17) and (18) we find
= −𝐶𝑝𝑐 𝑃 + 𝐶𝑝𝑝 𝑝. (18)
𝑉𝑝
𝐶𝑠′ = 𝐶𝑏𝑐 − 𝐶𝑏𝑝 , (34)
with four constitutive constants. These relations can be made nonlinear
𝐶𝑠′′ = 𝐶𝑝𝑐 − 𝐶𝑝𝑝 . (35)
if these constants are considered as functions of the stresses. For an
elastic (reversible) process, such reciprocity relation Hence we have presented three different ways to express the microme-
( ) ( ) chanical constitutive equations, each containing three independent con-
𝜕𝑉 𝜕𝑉𝑝
=− , (19) stants, selected as {𝐶𝑏𝑐 , 𝐶𝑏𝑝 , 𝐶𝑝𝑝 }, {𝐾, 𝛼, 𝛽}, and {𝐾, 𝐾𝑠′ , 𝐾𝑠′′ }. Relations
𝜕𝑝 𝑃 𝜕𝑃 𝑝
can be found among them.
must be satisfied,7,10 which shows that At this time, it is of interest to examine a special model, the ideal
𝐶𝑏𝑝 porous medium model of Gassmann.13 In such model, it is assumed that
𝐶𝑝𝑐 = . (20) the solid is made of a single isotropic mineral, with the bulk modulus
𝜙
𝐾𝑚 . In an unjacketed test, the material is submersed in a fluid chamber,
Hence only three of the constitutive constants are independent.
and the fluid pressure is raised. The entire surface of the solid, exterior
The physical meaning of these constants becomes clearer if we put
to the sample and interior opening to the pores, is exposed to the same
the above equations into different forms. For example, they can be
assemble into the effective stress forms as Refs. 5, 8:
𝛥𝑉 1 1
The notations in (27) and (28) are adopted from Rice and Cleary.12 Brown
= − (𝑃 − 𝛼 𝑝), (21)
𝑉 𝐾 and Korringa7 denoted 𝜅𝐴 = 1∕𝐾, 𝜅𝑀 = 1∕𝐾𝑠′ , 𝜅𝜙 = 1∕𝐾𝑠′′ , and 𝜅 ′ = 𝜙∕𝐾𝑝 .
𝛥𝑉𝑝 1 Many others21–23 denoted 𝐾𝑠′ as 𝐾𝑠 , and 𝐾𝑠′′ as 𝐾𝜙 . In the present work, 𝐾𝑠
= − (𝑃 − 𝛽𝑝). (22) and 𝐾𝜙 are reserved for other uses.
𝑉𝑝 𝐾𝑝

3
A.H.D. Cheng International Journal of Rock Mechanics and Mining Sciences 142 (2021) 104754

Fig. 1. Self-similar deformation of a homogeneous and isotropic solid subjected to a


uniform pressure. Fig. 2. Schematic of a jacketed test.

normal stress (see Fig. 1). A trivial solution of this problem is a uniform Although the initial fluid volume 𝑉𝑓 is the same as the pore volume
strain at every point. That is, every part of the solid is deformed by the 𝑉𝑝 , as defined in (14), their changes, 𝛥𝑉𝑝 and 𝛥𝑉𝑓 , are not necessarily
same proportion. This geometrically self-similar deformation leads to equal. Here we shall refer 𝛥𝑉𝑓 as the change of volume of the original
the following relations fluid mass residing in the pores. Based on this volume, we can define
𝛥𝑉𝑠 𝛥𝑉𝑝 𝛥𝑉 the fluid bulk modulus 𝐾𝑓 as
= = . (36) ( )
𝑉𝑠 𝑉𝑝 𝑉 𝜕𝑉𝑓
1 1
=− , (40)
From (27) and (28) with 𝑃 ′ = 0, it is clear that 𝐾𝑓 𝑉𝑓 𝜕𝑝 𝑇

𝐾𝑠′ = 𝐾𝑠′′ = 𝐾𝑚 . (37) where the subscript 𝑇 means keeping the temperature constant. Due to
the imbalance between the fluid and pore deformation, an amount of
The number of independent constants then reduces to two, 𝐾 and 𝐾𝑚 . fluid can enter or leave the frame, given as 𝛥𝑉𝑓′ (positive for leaving).
The constant 𝐾𝑚 can be obtained from the known bulk modulus of We may tie this quantity to the variation in fluid content as
the mineral, and only the measurement of 𝐾 is needed. This model is
𝛥𝑉𝑓′
popular in some applications, such as seismic wave propagation. 𝜁 =− . (41)
As pointed out by Brown and Korringa7 and others, such model 𝑉
is not likely to hold for sedimentary rocks, because they can contain We can write the following balance equation
more than one mineral, and particularly an amount of clay, which is
much softer than the crystals. Even for a relatively pure rock, it can 𝛥𝑉𝑝 = 𝛥𝑉𝑓 − 𝛥𝑉𝑓′ , (42)
contain occluded pores and fissures that are not connected to the flow which allows us to infer the pore volume change from the fluid volume
channels. These pores are not communicating with the fluid for pore changes.
pressure change, and are acting as a part of the solid, making the solid For the measurement of 𝐾𝑝 , we can conduct a jacketed test, in which
effectively inhomogeneous. the pore pressure is kept constant. In such case, we have 𝛥𝑉𝑓 = 0. This
allows us to obtain 𝛥𝑉𝑝 = −𝛥𝑉𝑓′ by monitoring the expelled or injected
4. Laboratory measurement fluid volume. In the same jacketed test, it is possible to increase or
decrease the pore pressure by injecting or withdrawing fluid from the
Biot and Willis6 and Geertsma10 suggested two laboratory tests, a sample. When this is done by carefully adjusting the confining pressure
jacketed and an unjacketed, to measure the micromechanical constants. 𝑃 at a constant, we can measure 𝐶𝑏𝑝 as
In the jacketed test, the sample is wrapped in a flexible membrane ( )
and placed in a fluid chamber. An all around pressure is applied to 1 𝜕𝑉
𝐶𝑏𝑝 = . (43)
the sample by raising the chamber fluid pressure (see Fig. 2). The load 𝑉 𝜕𝑝 𝑃
can be applied in increments. A tube is connected to the inside of the Based on the definition in (26), this test allows us to evaluate the
membrane to allow the pore fluid to drain to the ambient pressure or effective stress coefficient 𝛼, with 𝐾 given.
the initial pore pressure, to perform the drained test. The total volume In an unjacketed test, the sample is submersed in a chamber without
change 𝛥𝑉 can be continuously monitored. Based on (21), the drained a jacket, so the pore pressure 𝑝 is the same as 𝑃 . Actually, the unjack-
bulk modulus can be evaluated as eted test is generally conducted in the jacketed setting by using the
( )
1 1 𝜕𝑉 drainage tube to control the pore pressure, such that 𝑝 = 𝑃 , and 𝑃 ′ = 0.
=− . (38)
𝐾 𝑉 𝜕𝑃 𝑝 In such case, (27) allows the evaluation of 𝐾𝑠′ as
To interpret the drained pore modulus 𝐾𝑝 , defined in (22), and given ( )
1 1 𝜕𝑉
as = − . (44)
( ) 𝐾𝑠′ 𝑉 𝜕𝑝 𝑃 ′
1 1 𝜕𝑉𝑝
=− , (39) We notice that there is a redundancy in measurements in the tests
𝐾𝑝 𝑉𝑝 𝜕𝑃 𝑝
described above, as these constants, 𝐾, 𝐾𝑝 , 𝐾𝑠′ , and 𝐶𝑏𝑝 , are related.
a measurement of the pore volume change 𝛥𝑉𝑝 is needed. As it is not We further observe that these tests determine only two of the three
feasible to directly observe the change in pore space, this is generally independent constants, as none of them leads to the determination of
accomplished by monitoring the fluid exiting or entering the sample. 𝐾𝑠′′ , 𝛽, or 𝐶𝑝𝑝 .

4
A.H.D. Cheng International Journal of Rock Mechanics and Mining Sciences 142 (2021) 104754

In order to measure 𝐾𝑠′′ defined as More careful measurements of 𝛥𝑉𝑝 under changing pore pressure
( ) condition have only been recently carried out.25,34 To avoid the si-
1 1 𝜕𝑉𝑝 multaneous determination of all the calibration factors, the tests were
= − . (45)
𝐾𝑠′′ 𝑉𝑝 𝜕𝑝 𝑃 ′
conducted in two steps. In step 1, only the confining stress 𝑃 was
we again need to determine 𝛥𝑉𝑝 . In this case, however, it is under the changed, and the pore pressure was kept constant. This allowed an
changing pore pressure condition. Unlike the drained case, in which accurate determination of pore volume change by the expelled fluid,
𝛥𝑉𝑝 can be obtained as −𝛥𝑉𝑓′ , this time we need the correction by needing only a minor correction due to an internal tubing. The drained
the amount 𝛥𝑉𝑓 . Although (40) indicates that 𝛥𝑉𝑓 can be estimated pore modulus 𝐾𝑝 can be determined from this step, as indicted by (39).
if we know 𝐾𝑓 , in practice, it is more complicated. With the pressure In step 2, the confining stress 𝑃 was kept constant, and the pore pres-
change, one must also consider the compliances of the instrumentation, sure 𝑝 was raised to equate 𝑃 . The injected fluid volume was corrected
such as the piston, tubing, valve, transducer, and the penetration of the for the fluid compressibility and the system compliances, including the
jacket into the pores.24–26 These non-negligible biases can jeopardize an syringe pump piston, valves, tubes, and pressure transducer. The latter
accurate determination of pore volume change. portion can be separately determined by conducting a calibration test
To side step the issue, most studies attempted to invert 𝐾𝑠′′ from with a hollow metal cylinder. The pore volume change is the sum of
other more easily measured constants, by solving the relations among the two steps. Equation (45) is then used to determine 𝐾𝑠′′ . In addition
them. For example, one can measure the Skempton pore pressure to performing tests on rocks showing different 𝐾𝑠′ and 𝐾𝑠′′ values, a
relatively homogeneous rock, the Dunnville sandstone, was also tested
coefficient 𝐵. Based on the relations
to show 𝐾𝑠′ ≈ 𝐾𝑠′′ .25 These cases are reported in Table 1.
𝐾𝑓 𝐾𝑠′′ (𝐾𝑠′ − 𝐾) We may also mention that in addition to liquid, inert gas such as
𝐵 =
𝐾𝑓 𝐾𝑠′′ (𝐾𝑠′ − 𝐾) + 𝜙𝐾𝐾𝑠′ (𝐾𝑠′′ − 𝐾𝑓 ) helium has been used to measure pore volume change under changing
𝐾𝑓 𝐾𝑠′′ effective stress condition.38 Such technique has the advantage of en-
= , (46) suring saturation, and avoiding the absorption and adsorption effects.
𝐾𝑝 (𝐾𝑠′′ − 𝐾𝑓 ) + 𝐾𝑓 𝐾𝑠′′
However, the issue of calibrating other effects remains. In those tests,38
the constant 𝐾𝑠′′ can be inverted from the above, if {𝐵, 𝐾𝑝 , 𝐾𝑓 }, or the solid was assumed incompressible; hence the measurements erred
{𝐵, 𝜙, 𝐾, 𝐾𝑠′ , 𝐾𝑓 }, are known. There were a number of such by the amount of solid compressibility. It should be mentioned that the
efforts.21,23,27–30 After a literature survey, we compile the laboratory advancement in technology may allow the pore volume to be measured
measurements of 𝐾𝑠′′ for various rocks, using this and other method, using a computerized tomography (CT) scanner.39 Computer simula-
and show them in Table 1. tions on pore compressibility using the digital rock physics workflow
The ability to accurately determine 𝐾𝑠′′ from inversion is dependent have been attempted.40
on its sensitivity to the perturbation of the measured constants. Using
the Dunnville sandstone as an example, Tarokh et al.25 pointed out 5. Intrinsic material constants
that a mere ±0.01 perturbation of 𝐵 around the measured value of
0.49 (±2%) can cause a greater than 100% change in the predicted 𝐾𝑠′′ . In the micromechanical analysis presented in Section 3, the volume
To look into it, we perturb 𝐵 by 𝛥𝐵 in (46) and obtain the following changes 𝛥𝑉 and 𝛥𝑉𝑝 were chosen as the modeling variables based on
formula the laboratory measurement considerations. The resulting constitutive
𝛥𝐾𝑠′′ constants, 𝐾, 𝐾𝑠′ , and 𝐾𝑠′′ , are composite constants that incorporate
1 𝛥𝐵
≈ + 𝑂(𝛥𝐵 2 ). (47) the effects of both solid and pore volume changes. In this section, we
𝐾𝑠′′ (1 − 𝐵) − 𝐵(𝐾𝑝 ∕𝐾𝑓 ) 𝐵
shall choose differently, by using the solid volume change 𝛥𝑉𝑠 and
The multiplication factor in front of 𝛥𝐵∕𝐵 has the value between [1, ∞] porosity change 𝛥𝜙 as the kinematic variables. These variables may not
for 0 ≤ 𝐵 < 𝐾𝑓 ∕(𝐾𝑓 +𝐾𝑝 ), and between [−∞, −𝐾𝑓 ∕𝐾𝑝 ] for 𝐾𝑓 ∕(𝐾𝑓 +𝐾𝑝 ) < be directly measurable, but they can isolate the physical mechanisms.
𝐵 ≤ 1. Using the data of Green and Wang,29 as reported in the first The constitutive constants chosen are each associated with one such
three rows of Table 1, the error made in 𝐵 can be amplified by −28, mechanism, and are called the intrinsic material constants. By relating
−16, and −35 times for the estimate of 𝐾𝑠′′ . So the method based on these constants to the phenomenological ones, we may reveal physical
inversion may not be reliable. insight for these composite constants.
Recent work by Pimienta et al.36 on Bentheimer sandstone used We notice that 𝛥𝑉𝑠 as defined in (16) is related to 𝛥𝑉 and 𝛥𝑉𝑝 . The
pulsating load with various boundary conditions to make seven mea- porosity change is obtained by taking the following variation
surements of various material constants in an effort to determine 𝐾𝑠′′ . ( )
𝑉𝑝 𝛥𝑉𝑝 − 𝜙 𝛥𝑉
Of the methods for inversion, only the one using the undrained bulk 𝛥𝜙 = 𝛥 = , (49)
𝑉 𝑉
compressibility 𝐶𝑢 (= 1∕𝐾𝑢 ) produced acceptable result. The formula
which is also related to the previously define volume changes. So
used was
we are not constructing new laws, but are reassembling the existing
𝐶𝑏𝑝 𝐶𝑝𝑐
𝐶𝑠′′ = 𝐶𝑓 + 𝐶𝑝𝑐 − (48) ones to bring out new physical insight. One reason for such selection
𝐶𝑏𝑐 − 𝐶𝑢 of kinematic variables is that in the more rigorous formulation of
They determined 𝐶𝑠′′ ≈ 0.017 GPa−1 (see Table 1). If we focus on the poroelasticity through variational energy method, the solid strain 𝜖𝑠
first two terms on the right hand side of (48), we notice 𝐶𝑓 = 0.46 and porosity change 𝛥𝜙 are the physical variables chosen to model the
GPa−1 for water, and the measured 𝐶𝑝𝑐 = 0.1 GPa−1 . The determined Helmholtz free energy of the porous medium.4,17,19 In Section 6, we
𝐶𝑠′′ value is less than 1∕30 of 𝐶𝑓 +𝐶𝑝𝑐 . Hence the accuracy of this method shall give a parallel derivation using the variation energy and work
principle, and compare with the present more ad hoc approach.
is also subject to doubt.
Following the lead of (27) and (28), we can build the constitutive
Laurent et al.31 appear to be the first to measure 𝛥𝑉𝑝 , and deter-
laws as
mined the constants 𝐶𝑏𝑐 , 𝐶𝑏𝑝 , 𝐶𝑝𝑐 , and 𝐶𝑝𝑝 , The unjacketed bulk moduli
are evaluated from (34) and (35). The reported 𝐾𝑠′′ values for Lavoux
𝛥𝑉𝑠 𝑃′ 𝑝
= − − = −𝐶𝜎 𝑃 ′ − 𝐶𝑠 𝑝, (50)
and Vilhanneur limestone are respectively 130 and 110 GPa (Table 1). 𝑉𝑠 𝐾𝜎 𝐾𝑠
𝛥𝜙 𝑃 ′ 𝑝
These values are greater than 𝐾𝑚 = 77 GPa of calcite, the major mineral = − − = −𝐶𝜙 𝑃 ′ − 𝐶𝜓 𝑝, (51)
constituent of limestone. The validity of such high values has been 1−𝜙 𝐾𝜙 𝐾𝜓
questioned.23,25 We shall demonstrate in Section 7 that these values, where we have expressed the compressibilities as 𝐶𝜎 = 1∕𝐾𝜎 , 𝐶𝑠 =
though curious, do not violate the bounds for material constants. 1∕𝐾𝑠 , etc. We now seek the interpretation of these constants. First, we

5
A.H.D. Cheng International Journal of Rock Mechanics and Mining Sciences 142 (2021) 104754

Table 1
Reported unjacketed bulk modulus (𝐾𝑠′ ) and unjacketed pore modulus (𝐾𝑠′′ ) from various sources. Cases with 𝐵 value listed: 𝐾𝑠′′ determined by
inversion of 𝐵; with *: determined by inversion of 𝐾𝑢 ; and with †: pore volume change measured.
Rock type [Ref] 𝑃′ 𝜙 𝐵 𝐾 𝐾𝑠′ 𝐾𝑠′′ 𝐾𝑠′′ > 𝜙𝐾𝑠′
(MPa) (GPa) (GPa) (GPa)
0 0.2 0.99 0.31 45 1.7 No
Berea sandstone29 0.9 0.2 0.95 1.0 45 3.1 No
2.0 0.2 0.87 1.64 45 11 Yes
Fused glass beads21,27 0.11 0.39 0.98 0.22 25 4.5 No
Lavoux limestone31 2.0 0.23 † 8.3 44 130 Yes
Vilhanneur limestone31 2.0 0.14 † 20 56 110 Yes
Berea sandstone23 10 0.19 0.75 6.6 28.9 4.4 No
Indiana limestone23 20 0.13 0.46 20.5 70.7 7.5 No
14.5 0.2 0.48 9.7 31 12 Yes
Berea sandstone30
25.3 0.2 0.42 12.8 36 12 Yes
Tarbert sandstone (unconsolidated)32 8 0.37 † 1.5 37 1.2 No
Tarbert sandstone (consolidated)32 8 0.37 † 5.3 37 4.0 No
0.8 0.125 0.994 1.3 45 4.6 No
Fletchtinger sandstone28
3.6 0.125 0.892 2.1 45 16 Yes
0.9 0.236 0.94 2.2 45 3.2 No
Bentheimer sandstone28
7.3 0.236 0.60 5.7 45 12 Yes
Berea sandstone33 0.5 0.23 † 10 29 12 Yes
Opalinus clay (shale)34,35 0.5 0.13 † 1.85 8.9 7 Yes
Bentheimer sandstone36 5–30 0.24 * 14 37 59 Yes
Dunnville sandstone25,37 5–20 0.295 † 10.3 35 35 Yes
Silica polymorph25,37 5 0.474 † 2.1 24 24 Yes
Berea sandstone25,37 5–20 0.23 † 14 29 22 Yes

𝐾𝜙
observe from (50) that in an unjacketed test, 𝐾𝑠 is the bulk modulus of 𝜒 = 1− . (59)
solid defined as 𝐾𝜓
( )
1 1 𝜕𝑉𝑠 We note that the effective stress laws (54) and (55), and coefficients
=− . (52)
𝐾𝑠 𝑉𝑠 𝜕𝑝 𝑃 ′ (58) and (59), were introduced by Berryman,41 and also partially by
Carroll,8 without introducing the moduli 𝐾𝜎 , 𝐾𝜙 , and 𝐾𝜓 as constitu-
The solid here refers to the solid phase made of heterogeneous miner-
tive constants.
als, the cementing materials, and the occluded pores with or without
Eqs. (54) and (55) allow us to define
entrapped fluid in them. The occluded fluid is not communicating ( )
with the fluid in flow channels, where a change in pore pressure 1 1 𝜕𝑉𝑠
= − , (60)
can be transmitted. (See Fig. 3 for an illustration.) Assuming that the 𝐾𝜎 𝑉𝑠 𝜕𝑃 𝑝
heterogeneities are randomly distributed to give macroscopic isotropy, ( )
1 1 𝜕𝜙
it is proven in Appendix A that 𝐾𝑠 is the effective bulk modulus of = − , (61)
𝐾𝜙 1 − 𝜙 𝜕𝑃 𝑝
the composite solid. If the solid is made of a single isotropic mineral
without occluded pores, then 𝐾𝑠 = 𝐾𝑚 , the bulk modulus of the where 𝐾𝜎 is the drained bulk modulus of the solid, and 𝐾𝜙 the drained
mineral. porosity modulus. We note in the above that only three of the four
From (51) we observe that constants can be independent. We choose them as {𝐾𝑠 , 𝐾𝜙 , 𝐾𝜓 }, and
( ) find
1 1 𝜕𝜙
=− , (53) (1 − 𝜙)𝐾𝑠 𝐾𝜓
𝐾𝜓 1 − 𝜙 𝜕𝑝 𝑃 ′ 𝐾𝜎 = , (62)
𝐾𝑠 + 𝐾𝜓
so 𝐾𝜓 is the unjacketed porosity modulus. To understand its signifi-
and in compressibility notations
cance, we consider an ideal porous medium. As discussed at the end of
Section 3, when such a material is subjected to an unjacketed test, it 𝐶𝑠 + 𝐶𝜓
𝐶𝜎 = . (63)
deforms in a self-similar fashion. Based on (36) and (49), it is clear 1−𝜙
that 𝛥𝜙 = 0, and 𝐾𝜓 → ∞ or 𝐶𝜓 = 0. So 𝐾𝜓 is a measure of Hence the solid drained compressibility 𝐶𝜎 is a composite constant of
the disproportionate deformation due to the microinhomogeneity and 𝐶𝑠 and 𝐶𝜓 .
microanisotropy. At this point we are ready to select a set of intrinsic constants for
Following (21) and (22), we can express (50) and (51) into the their isolated physical mechanisms. Although we have demonstrated
effective stress laws: the physical significance of the set {𝐶𝑠 , 𝐶𝜙 , 𝐶𝜓 }, we shall replace 𝐶𝜙 ,
𝛥𝑉𝑠 1 the drained porosity modulus, by 𝐶𝑝 , the drained pore modulus defined
= − (𝑃 − 𝜎 𝑝), (54)
𝑉𝑠 𝐾𝜎 in (22), to form the set {𝐶𝑠 , 𝐶𝑝 , 𝐶𝜓 }. As discussed in the preceding
𝛥𝜙 1 section, 𝐶𝑝 can be measured in the laboratory. Also, for application
= − (𝑃 − 𝜒𝑝), (55)
1−𝜙 𝐾𝜙 purposes, the change in pore space is of more interest than the change
in which we identify the respective effective stresses in porosity; hence 𝐶𝑝 is a more natural choice.
Once the intrinsic constants are selected, we can compare (50) and
𝑃 ′′′′ = 𝑃 − 𝜎𝑝, (56) (51) with (27) and (28), and use the definitions (16) and (49), to solve
𝑃 ′′′′′ = 𝑃 − 𝜒𝑝, (57) for the following relations
𝐶𝜙 𝐶𝑠 𝐶𝜓
and the effective stress coefficients 𝐶𝑝 = + + , (64)
𝜙 1−𝜙 1−𝜙
𝐾
𝜎 = 1− 𝜎, (58) 𝐶 = 𝐶𝑠 + 𝜙 𝐶 𝑝 + 𝐶𝜓 , (65)
𝐾𝑠

6
A.H.D. Cheng International Journal of Rock Mechanics and Mining Sciences 142 (2021) 104754

Fig. 3. Schematic sketch of a representative elementary volume (REV) of a fluid saturated porous medium. Solid phase 𝛺𝑠 includes the heterogeneous solid constituents and
occluded pores with trapped fluid. Interconnected pores that allow the fluid flow and the transmission of pore pressure are the fluid space 𝛺𝑓 . In this illustration, the sample is
subjected to an unjacketed test with 𝑃 = 𝑝; hence the entire solid surface 𝛴𝑠 is under a uniform pressure 𝑝.

𝐶𝑠′ = 𝐶𝑠 + 𝐶𝜓 , (66) 𝜙𝐾(𝐾𝑠′ − 𝐾𝑠′′ ) 𝜙(𝐶𝑠′′ − 𝐶𝑠′ )


𝜒 = 1− =1− . (75)
𝐶𝜓 𝐾𝑠′′ [(1 − 𝜙)𝐾𝑠′ − 𝐾] (1 − 𝜙)𝐶 − 𝐶𝑠′
𝐶𝑠′′ = 𝐶𝑠 + . (67)
𝜙 which are the same as those given in Berryman.41
We also find
6. Variational work energy formulation
𝐶𝑠 + 𝐶𝜓
𝛼 = 1− , (68)
𝐶𝑠 + 𝜙 𝐶𝑝 + 𝐶𝜓 To offer a theoretical support to the constitutive equations presented
𝜙 𝐶 𝑠 + 𝐶𝜓 in Sections 3 and 5, we shall give an alternative derivation using the
𝛽 = 1− , (69)
𝜙 𝐶𝑝 variational energy principles. In Section 7 we also use the results for
(1 − 𝜙)𝐶𝑠 a stability analysis, leading to bound criteria for material constants.
𝜎 = 1− , (70)
𝐶𝑠 + 𝐶𝜓 Due to limited space, only an abbreviated presentation is given. A
(1 − 𝜙)𝐶𝜓 more detailed derivation based on volume averaging and variational
𝜒 = 1− [ ]. (71) principles can be found in Lopatnikov and Cheng,17 and Cheng.4
𝜙 (1 − 𝜙)𝐶𝑝 − 𝐶𝑠 − 𝐶𝜓
In the presentation below, only the volumetric deformation is con-
We observe from the above that the intrinsic constants help to identify sidered. Assume an REV as in Fig. 3, which is further conceptualized
the physical mechanisms. For example, the drained bulk compressibil- into Fig. 4. It is subjected to a uniform compressive normal stress 𝑝𝑠
ity 𝐶 is a composite constant of the three mechanisms, that due to applied to the external part of solid surface, and a pore pressure 𝑝 to
the solid compressibility 𝐶𝑠 , the drained pore compressibility 𝐶𝑝 , and the external part of fluid surface, giving the external surface averaged
the unjacketed porosity compressibility 𝐶𝜓 associated with the solid total stress as
microinhomogeneity. These mechanisms work like a series spring. The
𝑃 = (1 − 𝜙)𝑝𝑠 + 𝜙𝑝. (76)
constants 𝐶𝑠′ and 𝐶𝑠′′ , on the other hand, involve two mechanisms only,
those due to 𝐶𝑠 and 𝐶𝜓 . Equations (64)–(67) can be expressed in the The total work performed to the system by these stresses can be
forms of elastic moduli, 𝐾, 𝐾𝑠′ , etc., as (B.1)–(B.4) in Appendix B. partitioned into a part through the solid external surface, and a part
To be able to use the intrinsic constants, we need their values. through the fluid external surface, as:
Although 𝐾𝑠 and 𝐾𝜓 cannot be directly measured, we can infer them
𝑊 = (1 − 𝜙)𝑊𝑠 + 𝜙𝑊𝑓 , (77)
from the measurable ones. Such relations are easily found from (66)
and (67) as where 𝑊 , 𝑊𝑠 , and 𝑊𝑓 are the work densities (work per unit volume),
with
𝐶𝑠′ − 𝜙𝐶𝑠′′
𝐶𝑠 = , (72)
1−𝜙 𝑊𝑠 = −𝑝𝑠 𝑒𝑠 , (78)
′′ ′
𝜙(𝐶𝑠 − 𝐶𝑠 )
𝐶𝜓 = . (73) 𝑊𝑓 = −𝑝 𝑒𝑓 , (79)
1−𝜙
in which 𝑒𝑠 and 𝑒𝑓 are the solid and fluid external strain (deformation
Relations for other constants, 𝐶𝜙 and 𝐶𝜎 , are given in Appendix B, and
observed from the external surface). We note that 𝑊𝑠 and 𝑊𝑓 are
𝐶𝑝 in (31). We observe that 𝐶𝑠 and 𝐶𝜓 are the weighted difference
not the total work performed to the solid and fluid phases, as they
between 𝐶𝑠′ and 𝐶𝑠′′ . If 𝐶𝑠′ , and 𝐶𝑠′′ can be reliably determined in the do not include the internal works performed to each other through
laboratory, and their values are significantly different from each other the interface (see Fig. 4(a)). These works cancel each other in the
(see Table 1), then 𝐶𝑠 and 𝐶𝜓 can be determined with equal reliability. accounting of the total work.
If 𝐶𝑠′ and 𝐶𝑠′′ are close to each other, then 𝐶𝑠 ≈ (𝐶𝑠′ , 𝐶𝑠′′ ), and 𝐶𝜓 ≈ 0. As illustrated in the figure, the solid external strain 𝑒𝑠 is equal to
We also find the effective stress coefficients in (58) and (59) as the total strain 𝑒 through the relation
𝜙𝐾𝑠′ 𝜙𝐶𝑠′′ (1 − 𝜙)𝛥𝑉 𝛥𝑉
𝜎 = = , (74) 𝑒𝑠 = = = 𝑒. (80)
𝐾𝑠′′ 𝐶𝑠′ 𝑉𝑠 𝑉

7
A.H.D. Cheng International Journal of Rock Mechanics and Mining Sciences 142 (2021) 104754

Fig. 4. Conceptualization of work performed to a porous medium and the corresponding volume changes. (a) Work done to the solid, and (b) work done to the fluid.

We note that 𝑒𝑠 ≠ 𝜖𝑠 , the solid internal strain (the true solid strain), as Based on the above, we can take the variation of the strain energy as
there exists internal deformation of the solid. 𝜕𝑈𝑓 (𝜖𝑓 )
𝜕𝑈𝑠 (𝜖𝑠 , 𝜙) 𝜕𝑈 (𝜖 , 𝜙)
The fluid external strain needs to be more carefully defined. As 𝛿𝑈 = (1 − 𝜙) 𝛿𝜖𝑠 + (1 − 𝜙) 𝑠 𝑠 𝛿𝜙 + 𝜙 𝛿𝜖𝑓 . (89)
𝜕𝜖𝑠 𝜕𝜙 𝜕𝜖𝑓
observed in Fig. 4(b), fluid can enter or leave the frame; hence the fluid
material surface does not coincide with the frame surface. We hence The work energy principle states that the work done to a system is equal
need to express the fluid strain associated with the external work as to the change in its internal energy; hence
𝜙𝛥𝑉 + 𝛥𝑉𝑓′ 𝛥𝑉 𝛥𝑉𝑓 1 − 𝜙 𝛥𝑉𝑝 𝛥𝑉𝑓 𝛥𝜙 𝛿𝑊 = 𝛿𝑈 , (90)
𝑒𝑓 = = + − = − , (81)
𝜙𝑉 𝑉 𝑉𝑓 𝜙 𝑉𝑝 𝑉𝑓 𝜙
where
in which 𝛥𝑉𝑓′ is the fluid volume escaping from the specimen, as defined
𝛿𝑊 = −(1 − 𝜙)𝑝𝑠 𝛿𝑒𝑠 − 𝜙𝑝 𝛿𝑒𝑓 . (91)
in (41) and (42), and we have utilized (49).
Next we consider the total strain energy density function 𝑈 , parti- In order to compare (89) with (91), we replace the internal strains in
tioned into a solid part 𝑈𝑠 and a fluid part 𝑈𝑓 , as (89) by the external ones using (85) and (86), to become
𝑈 = (1 − 𝜙)𝑈𝑠 + 𝜙𝑈𝑓 . (82) 𝜕𝑈𝑠 (𝜖𝑠 , 𝜙) 𝜕𝑈𝑓 (𝜖𝑓 )
𝛿𝑈 = (1 − 𝜙) 𝛿𝑒𝑠 + 𝜙 𝛿𝑒𝑓
𝜕𝜖𝑠 𝜕𝜖𝑓
These strain energy densities are dependent on the internal (true) solid [ ]
𝜕𝑈 (𝜖 , 𝜙) 𝜕𝑈𝑠 (𝜖𝑠 , 𝜙) 𝜕𝑈𝑓 (𝜖𝑓 )
and fluid strains, defined as + (1 − 𝜙) 𝑠 𝑠 − + 𝛿𝜙. (92)
𝜕𝜙 𝜕𝜖𝑠 𝜕𝜖𝑓
𝛥𝑉𝑠 𝛥𝑉 − 𝛥𝑉𝑝 𝛥𝑉 𝛥𝜙
𝜖𝑠 = = = − , (83) By equating (91) and (92), we obtain the following definitions of
𝑉𝑠 (1 − 𝜙)𝑉 𝑉 1−𝜙
𝛥𝑉𝑓 stresses
𝜖𝑓 = , (84) 𝜕𝑈 (𝜖 , 𝜙)
𝑉𝑓 𝑝𝑠 = − 𝑠 𝑠 , (93)
𝜕𝜖𝑠
Observing from (80), (81), (83), and (84), we find 𝜕𝑈𝑓 (𝜖𝑓 )
𝛥𝜙 𝑝 = − . (94)
𝜖𝑠 = 𝑒𝑠 − , (85) 𝜕𝜖𝑓
1−𝜙
𝛥𝜙 As (91) does not contain a 𝛿𝜙 term, we must set the term in (92) to
𝜖𝑓 = 𝑒𝑓 + , (86) zero, and obtain the following equation as a constraint:
𝜙
which tie the internal strains to external strains through porosity 𝜕𝑈𝑠 (𝜖𝑠 , 𝜙)
𝑝𝑠 − 𝑝 = −(1 − 𝜙) , (95)
changes. 𝜕𝜙
The next step is to establish the functional dependence of the strain in which we have substituted in (93) and (94).
energy to the deformation state of the material phases. For a non- Next, we introduce the quadratic forms for the strain energy density
porous elastic solid, the strain energy density is a function of the solid functions as
volumetric strain, 𝑈𝑠 = 𝑈𝑠 (𝜖𝑠 ) (recalling that only volumetric deforma- 1 1
tion is considered). The situation, however, is more complicated for a 𝑈𝑠 (𝜖𝑠 , 𝜙) = 𝐾𝛼 𝜖𝑠2 + 𝐾𝛽 𝛥𝜙2 + 𝐾𝛾 𝜖𝑠 𝛥𝜙, (96)
2 2
porous solid—the strain energy is also dependent on the change in pore 1
𝑈𝑓 (𝜖𝑓 ) = 𝐾𝑓 𝜖𝑓2 , (97)
structure, although there is no mass contained in it. Hence in principle 2
we need to write 𝑈𝑠 = 𝑈𝑠 (𝑒𝑠 , 𝛷), where 𝛷 represents the pore geometric in which 𝐾𝛼 , 𝐾𝛽 , 𝐾𝛾 , and 𝐾𝑓 are constitutive constants.2 Utilizing
factors. As each geometry is unique, from practical considerations, we the definitions (93) and (94), we obtain the following constitutive
can only choose some bulk quantities to characterize the pore space, equations
which may include the porosity, specific surface, tortuosity, and other
pore fabrics. For the simplest model, we shall write 𝑝𝑠 = −𝐾𝛼 𝜖𝑠 − 𝐾𝛾 𝛥𝜙, (98)

𝑈𝑠 = 𝑈𝑠 (𝜖𝑠 , 𝜙). (87)


2
For the fluid under isothermal condition, we have Note on change in notation: The current 𝐾𝛼 , 𝐾𝛽 , and 𝐾𝛾 were denoted
as 𝐾𝛼 , 𝐾𝜙 , and 𝐾𝜓 in Cheng.4 The current 𝐾𝜙 in (61) corresponds to 𝐾𝜑 in
𝑈𝑓 = 𝑈𝑓 (𝜖𝑓 ). (88) Cheng,4 and 𝐾𝜓 in (53) is a new definition.

8
A.H.D. Cheng International Journal of Rock Mechanics and Mining Sciences 142 (2021) 104754

Table 2
Calculated 𝐾𝑠 , 𝐾𝑝 , and 𝐾𝜓 values from 𝐾, 𝐾𝑠′ . 𝐾𝑠′′ , and 𝜙 in Table 1. 𝜐: Volume fraction. –: No data. *Additional component: feldspar 20%
(𝐾𝑚 = 38 GPa). † Additional components: calcite 24%, quartz 20%.
Rock type [Ref] 𝑃′ 𝐾𝑠 𝐾𝑝 𝐾𝜓 Dominant 𝜐 𝐾𝑚 𝐾𝑠 < 𝐾𝑚
(MPa) (GPa) (GPa) (GPa) Mineral (%) (GPa)
Berea sandstone29 2.0 200 0.34 59 quartz 94 37 No
Lavoux limestone31 2.0 37 2.4 −222 calcite 100 77 Yes
Vilhanneur limestone31 2.0 51 4.4 −683 calcite – 77 Yes
14.5 53 2.8 75 quartz 80 37 No
Berea sandstone30
25.3 73 4.0 70 quartz 80 37 No
Fletchtinger sandstone28 3.6 62 0.28 167 quartz 65∗ 37 No
Bentheimer sandstone28 7.3 51 1.7 67 quartz 95 37 No
Berea sandstone33 0.5 50 3.5 69 quartz 90 37 No
Opalinus clay34,35 0.5 9.3 0.30 219 clay 50† – ?
Bentheimer sandstone36 5–30 33 5.4 −317 quartz 95 37 Yes
Berea sandstone25,37 5–20 32 6.2 305 quartz 85 37 Yes

𝑝 = −𝐾𝑓 𝜖𝑓 . (99) 7. Bounds of material constants

In addition, (95) yields 7.1. General model


𝑝𝑠 − 𝑝 = −(1 − 𝜙)𝐾𝛽 𝛥𝜙 − (1 − 𝜙)𝐾𝛾 𝜖𝑠 . (100)
In this section we shall explore the bounds of the various materi-
With 𝜖𝑠 defined in (83), equations (98) and (100) are equivalent to (50) als constants. Based on the thermodynamics consideration of particle
and (51), except that 𝑝𝑠 and 𝑝, instead of 𝑃 ′ and 𝑝, are used as the interactions,42–44 the compressibility 𝐶𝑠 of the solid phase, which in-
dynamic variables. clude the entrapped fluid in the occluded pores, must be positive. Hence
we can establish that
It is of interest to examine the physical meaning of these constitutive
constants. First, we notice 𝐶𝑠 > 0, 𝐾𝑠 > 0. (107)
𝑃′ The same, however, cannot be concluded for the bulk compressibility
𝑝𝑠 − 𝑝 = . (101)
1−𝜙 𝐶 of the porous material, which contains void space.45 A negative
Under the unjacketed test, 𝑃 ′ = 0, and (100) gives the proportion compressibility is possible and has been observed in biological and
of deformation between 𝛥𝑉𝑠 ∕𝑉𝑠 and 𝛥𝜙. As discussed earlier, for an man-made materials.46,47 A statement of 𝐶 > 0 and 𝐾 > 0 requires
ideal porous medium under such test, 𝛥𝜙 = 0. Equation (100) can be further qualification.
balanced only if 𝐾𝛾 = 0. Hence 𝐾𝛾 is tied to microinhomogeneity and Given the positiveness of 𝐶𝑠 , (72) lead to the condition
microanisotropy of the material. The condition 𝐾𝛾 = 0 also leads to the 𝐶𝑠′
𝐶𝑠′′ < . (108)
uncoupling of 𝜖𝑠 and 𝛥𝜙 in the quadratic form (96), and the reduction 𝜙
of (98) and (100). For the other two constants, they can be interpreted We note that this condition does not state the positiveness of either 𝐶𝑠′
as or 𝐶𝑠′′ , and they can be negative. If both of them are positive, as for the
( )
𝜕𝜖𝑠 cases in Table 1, we can express (108) as
𝐶𝛼 = − , (102)
𝜕𝑝𝑠 𝛥𝜙
( ) 𝐾𝑠′′ > 𝜙𝐾𝑠′ . (109)
𝜕𝜙
𝐶𝛽 = −(1 − 𝜙)2 , (103) We observe that this condition is not met by a number of the cases in
𝜕𝑃 ′ 𝜖
𝑠
Table 1, which are marked with a ‘‘No’’ in the last column. As (107) is
where 𝐶𝛼 = 1∕𝐾𝛼 , and 𝐶𝛽 = 1∕𝐾𝛽 . Their physical meanings, however, a thermodynamics condition, these reported 𝐾𝑠′′ are disqualified.
are more obscure, as it is difficult to keep the solid strain and porosity For the surviving cases, we use the data in Table 1 to evaluate the
constant. Hence we shall use {𝐾𝑠 , 𝐾𝑝 , 𝐾𝜓 } as the intrinsic material intrinsic constants, and report them in Table 2. (Two cases, Dunnville
constants.3 sandstone and silica polymorph, with 𝐾𝑠′ = 𝐾𝑠′′ , are not included.) In
Comparing (98) and (100) with (50) and (51), and converting them the same table, we also show the dominant mineral component for
to the same variables, we can establish the following relations each rock, its volume fraction 𝜐, as well as the bulk modulus of the
mineral 𝐾𝑚 obtained from Carmichael.48 The Voigt bound states that
𝐾𝛼 𝐾𝛽 − 𝐾𝛾2 the effective bulk modulus of a heterogeneous material has the upper
𝐾𝑠 = , (104)
𝐾𝛽 bound
𝜙(1 − 𝜙)3 (𝐾𝛼 𝐾𝛽 − 𝐾𝛾2 ) 𝐾𝑠 ≤ 𝜐1 𝐾𝑚1 + 𝜐2 𝐾𝑚2 + 𝜐3 𝐾𝑚3 + ⋯ , (110)
𝐾𝑝 = , (105)
𝐾𝛼 + (1 − 𝜙)[𝜙(1 − 𝜙)𝐾𝛽 − (1 + 𝜙)𝐾𝛾 ]
where 𝐾𝑚1 , 𝐾𝑚2 , … are the bulk modulus of mineral 1, 2, etc., and
(1 − 𝜙)(𝐾𝛼 𝐾𝛽 − 𝐾𝛾2 ) 𝜐1 , 𝜐2 , … are their volume fraction. For all the rocks in Table 2, with
𝐾𝜓 = − . (106)
𝐾𝛾 the exception of Opalinus clay, the dominant mineral is also the stiffest
among the constituents; hence we can establish the upper bound for
Their compressibility forms are shown in Appendix B as (B.7)–(B.9).
these rocks as
We hence have established the full equivalence between the microme-
chanical and the variational energy analysis. . 𝐾𝑠 < 𝐾𝑚 , (111)

where 𝐾𝑚 is the bulk modulus of the dominant mineral. We observe in


Table 2 that this condition is violated by most of the cases, which are
3
In Cheng,4 {𝐾𝑠 , 𝐾𝛽 , 𝐾𝛾 } (in current notations) were chosen as the intrinsic marked by a ‘‘No’’ in the last column. We notice that all cases based on
constants. Based on the analysis in Section 5, the current choice of {𝐾𝑠 , 𝐾𝑝 , 𝐾𝜓 } the inversion of 𝐵 failed this and the last test; hence the reliability of
is preferred. this method is questionable. For the case of Bentheimer sandstone,36

9
A.H.D. Cheng International Journal of Rock Mechanics and Mining Sciences 142 (2021) 104754

𝐾𝑠′′ was determined by inversion of 𝐾𝑢 . Although its accuracy was Table 3


questioned in (48), it nevertheless passed the test. All other cases that Checking the lower bound criterion Eq. (128).

survived the tests are based on pore volume measurement. Rock type [Ref] 𝜙 𝐶 𝐶𝑠′ 𝐶𝑠′′ Satisfy
(GPa−1 ) (GPa−1 ) (GPa−1 ) Eq. (128)
We can write a similar equation as (110) for 𝐾, with the solid and
pore space as two components on the right hand side. Considering that Lavoux limestone31 0.23 0.12 0.023 0.0075 Yes
Vilhanneur limestone31 0.14 0.050 0.018 0.0090 Yes
the void has no strength, we may build an upper bound for 𝐾 as Ref. 22 Opalinus clay34,35 0.13 0.54 0.11 0.14 Yes
Bentheimer sandstone36 0.24 0.071 0.027 0.017 Yes
𝐾 < (1 − 𝜙)𝐾𝑠 . (112)
Berea sandstone25,37 0.23 0.071 0.035 0.046 Yes
Such condition is met by all the surviving cases in Table 2. The
above condition, however, does not guarantee the positiveness of 𝐾,
as mentioned above.
Their compressibility forms are given as (B.10)–(B.12). Based on the
We next examine the elastic stability condition. Such examination
second relation of (121), we can determine a lower bound for the
requires the work energy relations presented in Section 6. To remove
drained bulk modulus as
the influence of the fluid, we consider the drained condition; hence only
the strain energy density of the solid defined in (96) is considered. We 𝐾 > 0, 𝐶 > 0. (124)
notice that (96) is a quadratic equation in terms of the solid strain 𝜖𝑠
and 𝛥𝜙, which can be expressed as Hence, although the positiveness of 𝐾 is not required by the thermo-
dynamics condition, it is required for the stability of the material.
𝐴𝜖𝑠2 + 𝐵𝜖𝑠 𝛥𝜙 + 𝐶𝛥𝜙2 = 𝐹 , (113) We can also explore the stability condition (119). Substituting
with (B.13)–(B.15) into (119) and simplifying, we obtain
𝐾𝛼 𝐾𝛽 𝐾𝐾𝑠′ 2 𝐾𝑠′′
𝐴= , 𝐵 = 𝐾𝛾 , 𝐶 = , 𝐹 = 𝑈𝑠 . (114) > 0. (125)
2 2
(1 − 𝜙)2 (𝜙𝐾𝑠′ 2 − 𝐾𝐾𝑠′′ − 𝐾𝑠′ 𝐾𝑠′′ )
When a sample is subjected to an external load of finite magnitude,
(113) must represent an ellipse, or else the deformation can be un- Since 𝐾 > 0, we can multiply (125) by (1 − 𝜙)2 , divide it by 𝐾, and flip
bounded. Hence the mechanical stability requires the following con- the numerator and denominator, to obtain the following
ditions ( )
1 𝐾 𝜙 𝛼 𝜙

1 − ′ − ′′ = ′ − ′′ > 0. (126)
𝐵 2 − 4𝐴𝐶 < 0, (115) 𝐾𝑠 𝐾𝑠 𝐾𝑠 𝐾𝑠 𝐾𝑠
𝐴𝐹 < 0. (116) The above equation is exactly the stability condition obtained by Berry-
man and Milton22,49 using a different strain-energy functional. Using
Since the strain energy 𝑈𝑠 > 0, we can deduce from the above equations
(126), Berryman41 showed the inequality among the effective stress
that
coefficients:
𝐾𝛼 > 0, 𝐶𝛼 > 0; (117)
𝛽 > 𝛼 > 𝜎. (127)
𝐾𝛽 > 0, 𝐶𝛽 > 0; (118) Converting (126) to compressibility notations, we obtain
and 𝐶𝑠′ 𝐶𝑠′ 2
𝐶𝑠′′ < − . (128)
𝐾𝛼 𝐾𝛽 > 𝐾𝛾2 , 𝐶𝛾2 > 𝐶𝛼 𝐶𝛽 . (119) 𝜙 𝜙𝐶

We notice from the above that while 𝐾𝛼 and 𝐾𝛽 must be positive, there The above equation can be compared to (108). As the second term on
is no constraint on the sign of 𝐾𝛾 . the right hand side is positive, (128) is a more stringent condition, and
Based on the above, we may examine the positiveness of the in- it supersedes (108). In Table 3, we list all the surviving cases from
trinsic constants. First, Eq. (104) clearly indicates that 𝐾𝑠 > 0, which Table 2, and their 𝐶, 𝐶𝑠′ , and 𝐶𝑠′′ values. All of them pass the test
is a condition already established in (107) using the thermodynamics of (128). We can also translate (128) into a bound statement for 𝐶𝜓 .
argument. Also, we can use the inequalities to show that Substituting (65)–(67) into the above, we find
[ √ ]
1
𝐾𝛼 > 𝐾𝑠 , 𝐶𝑠 > 𝐶𝛼 . (120) −(1 + 𝜙)𝐶𝑠 + (1 − 𝜙)𝐶𝑠 [(1 − 𝜙)𝐶𝑠 + 4𝜙𝐶𝑝 ] > 𝐶𝜓
2
[ √ ]
Eqs. (106) shows that 𝐾𝜓 carries the opposite sign of 𝐾𝛾 , so they 1
> −(1 + 𝜙)𝐶𝑠 − (1 − 𝜙)𝐶𝑠 [(1 − 𝜙)𝐶𝑠 + 4𝜙𝐶𝑝 ] . (129)
both do not have sign constraint. Referring to (51), we find that in an 2
unjacketed test with a compressive load, 𝑝 > 0, the sample can have The expression in the radical is guaranteed to be positive if the original
a reduction in porosity if 𝐾𝜓 > 0, or an increase if 𝐾𝜓 < 0. Similarly, inequality is satisfied.
(105) shows that we cannot establish the positiveness of 𝐾𝑝 . Hence, Although some of the above inequalities involved 𝐾𝑠′ and 𝐾𝑠′′ , we
under the drained condition, (22) indicates that the pore space can have so far not determined whether these moduli can become negative.
increase or decrease subjected to a compressive load. Examination of (122) and (123), or (B.11) and (B.12), cannot establish
We can express the micromechanical constants {𝐾, 𝐾𝑠′ , 𝐾𝑠′′ } in terms the positiveness of 𝐾𝑠′ and 𝐾𝑠′′ . This situation is more clearly observed
of {𝐾𝛼 , 𝐾𝛽 , 𝐾𝛾 } as follows in (66) and (67). A negative 𝐶𝜓 with |𝐶𝜓 | > 𝐶𝑠 causes both 𝐶𝑠′ < 0
(1 − 𝜙)3 (𝐾𝛼 𝐾𝛽 − 𝐾𝛾2 ) and 𝐶𝑠′′ < 0. For 𝐶𝜓 < 0 and 𝐶𝑠 > |𝐶𝜓 | > 𝜙𝐶𝑠 , we find 𝐶𝑠′ > 0
𝐾 = and 𝐶𝑠′′ < 0. These conditions do not violate the bound requirement in
𝐾𝛼 + (1 − 𝜙)[(1 − 𝜙)𝐾𝛽 − 2𝐾𝛾 ] (129), so 𝐶𝑠′ and 𝐶𝑠′′ are not constrained to be positive. Indeed, using a
(1 − 𝜙)3 𝐾𝑠 𝐾𝛽2 model of particles made of a hard core of mineral, and a soft coating of
= , (121) clay, Berge and Berryman49 demonstrated that 𝐾𝑠′′ can be negative. If
𝐾𝑠 𝐾𝛽 + [(1 − 𝜙)𝐾𝛽 − 𝐾𝛾 ]2
such a material is not found in nature, it can certainly be constructed.
(1 − 𝜙)(𝐾𝛼 𝐾𝛽 − 𝐾𝛾2 )
𝐾𝑠′ = , (122) With the above, we also notice that despite the inequality (127), the
(1 − 𝜙)𝐾𝛽 − 𝐾𝛾 effective stress coefficients are not bounded. They can be greater than 1,
𝜙(1 − 𝜙)(𝐾𝛼 𝐾𝛽 − 𝐾𝛾2 ) or negative. Additional assumptions are needed to have them bounded
𝐾𝑠′′ = . (123) between these limits.
𝜙(1 − 𝜙)𝐾𝛽 − 𝐾𝛾

10
A.H.D. Cheng International Journal of Rock Mechanics and Mining Sciences 142 (2021) 104754

At this point, we can give a summary of the above findings: • Given the two assumptions in (130), and the earlier established
bounds, the following moduli, {𝐾, 𝐾𝑠′ , 𝐾𝑠′′ , 𝐾𝑝 , 𝐾𝑠 , 𝐾𝜙 , 𝐾𝜎 , 𝐾𝛼 , 𝐾𝛽 },
• The moduli 𝐾𝑚 , 𝐾𝑠 , 𝐾𝛼 , 𝐾𝛽 and 𝐾 are positive based on thermo-
are all positive, leaving only {𝐾𝜓 , 𝐾𝛾 } to take either sign.
dynamics and elastic stability conditions.
• It is established that 𝐾𝑠′ > 𝐾 > 𝐾𝑝 > 𝜙𝐾, 𝐾𝑠′′ > 𝐾𝑝 , and 𝐾𝑠′′ > 𝜙𝐾𝑠′ .
• The moduli 𝐾𝑠′ , 𝐾𝑠′′ , 𝐾𝑝 , 𝐾𝜙 , 𝐾𝜓 , and 𝐾𝜎 are not necessarily
positive. The inequality between 𝐾𝑠′′ and 𝐾𝑠′ , and 𝐾𝑠′′ and 𝐾, cannot be
• The effective stress coefficients 𝛼, 𝛽, 𝜎, and 𝜒 are not necessarily established.
bounded between 0 and 1. • 𝐾𝑠′ and 𝐾𝑠′′ can be greater than or less than 𝐾𝑠 , depending on the
• The positive moduli have the following inequalities: sign of 𝐾𝜓 .
• The combination of (127) and (136) gives
𝐾𝑚 > 𝐾𝑠 and 𝐾𝛼 > 𝐾𝑠 > 𝐾∕(1 − 𝜙) > 𝐾 > 0
1 > 𝛽 > 𝛼 > 𝜎 > 0.
• The stability condition (119) leads to the inequality (127) for
effective stress coefficients.
• The same stability condition leads to the constraints for 𝐶𝑠′′ and 7.3. Ideal porous medium
𝐶𝜓 in (128) and (129).
We can carry the analysis further by assuming ideal porous medium,
7.2. Special model
which is defined by
The above bounds are based on thermodynamics and elastic stability 𝐶𝜓 = 0, 𝐾𝜓 → ∞, (138)
considerations. They can be too wide for practical purposes, as they
allow some ‘‘strange’’ behaviors to exist, which are not observed in the or
laboratory for common materials. For example, (22) indicates that a
negative 𝐾𝑝 will cause the pore volume to increase under a increment of 𝐶𝛾 → ∞, 𝐾𝛾 = 0. (139)
compressive stress in drained condition. As 𝐾 is positive, the total vol- Equations (65)–(67), and (104) become
ume will decrease, so the stability condition is not violated. Similarly,
based on (28), a negative 𝐾𝑠′′ has the same effect under the unjacketed 𝐶 = 𝐶𝑠 + 𝜙 𝐶 𝑝 , (140)
test. A negative modulus can cause the effective stress coefficients to be
𝐶𝑠′ = 𝐶𝑠′′ = 𝐶𝑠 = 𝐶𝛼 = 𝐶𝑚 . (141)
greater than 1; thus the pore pressure can overcompensate the applied
total stress. Also of interest is the reduction of effective stress coefficients. From
In the laboratory, the modulus 𝐾𝑠′ has been routinely measured for (59) and (70) we have:
decades for various rocks. In addition to those reported in Table 1, we
can mention the earlier attempts,11,20,50–52 as well as many recent ones 𝜎 = 𝜙, (142)
reviewed in Tarokh & Makhnenko.53 There was no indication that 𝐾𝑠′
𝜒 = 1. (143)
could be negative. The measurement of 𝐾𝑠′′ , on the other hand, was
much less attempted. As demonstrated in Tables 1 and 2, the confidence These limits were also shown by Carroll8 and Berryman.41
of its measurement has not been fully built. Eqs. (55) and (143) indicate that the porosity change 𝛥𝜙 is con-
Nevertheless, we shall make the following assumptions for practical trolled by the Terzaghi effective stress 𝑃 ′ . There were plenty of ex-
materials, perimental evidences on sandstones,54–57 limestone,58 and granite,59
𝐶𝑠′′ > 0, 𝐶𝜙 > 0. (130) which showed that the ductility and ultimate strength of porous rock
are unique functions of the Terzaghi effective stress. These observations
Based on (28) and (55), the above conditions mean that in an un-
suggest that the failure process is controlled by the porosity change, and
jacketed test, a positive 𝑃 = 𝑝 will not cause the pore volume to
not the pore volume change (with the effective stress coefficient 𝛽). We
increase, and in a drained test, a positive 𝑃 will not cause the porosity
to increase. Following (64), (66), (108), and (B.6), we find may consider that the initial part of yield leading to failure is caused
by the deformation and destruction of pore structure, and not by the
𝐶𝑠′ > 0, 𝐶𝜎 > 0, 𝐶𝑝 > 0. (131) failure of the solid phase. The change of pore volume consists of two
Hence all compressibilities are positive, except for 𝐶𝜓 and 𝐶𝛾 , which parts: one part is the self-similar deformation of solid and pore space,
can take either sign. Furthermore, from (65) and (66) we find that and the other is the disproportionate change. The former maintains
the integrity of the pore structure, and does not weaken it. The latter
𝐶 > 𝐶𝑠′ , 𝐾𝑠′ > 𝐾. (132) deforms the shape of pore space, leading to yield and failure. The latter
And from (65) and (64), we can establish mechanism causes porosity change, and is governed by the effective
stress 𝑃 ′′′′′ , which can be approximated by the Terzaghi effective stress
𝐶𝑝 > 𝐶 > 𝜙𝐶𝑝 , 𝐾 > 𝐾𝑝 > 𝜙𝐾. (133)
𝑃 ′ , based on (143).
It is proven in Appendix C that The ideal porous medium assumption allows us to use the com-
𝐶𝑝 > 𝐶𝑠′′ , 𝐾𝑠′′ > 𝐾𝑝 . (134) posite material bound criteria, such as the Hashin–Shtrikman bound.60
By considering that the porous medium is made of two materials, a
From (72) and (B.6), we can show that homogeneous and isotropic solid, and a pore space, we can obtain the
𝐾𝑠 > 𝐾𝜎 . (135) following bounds4 :

The positiveness of the compressibilities, and the inequalities (132), 2(1 − 𝜙)


𝐾< 𝐾 , (144)
(134), and (135), allow us to establish from (32), (33), and (58) that 2+𝜙 𝑚
2(1 − 𝜙)
0 < {𝛼, 𝛽, 𝜎} < 1. (136) 𝐾𝑝 < 𝐾𝑚 , (145)
3
3𝜙
For 𝜒, we can only show that < 𝛼 < 1, (146)
2+𝜙
𝜒 > 0, (137) 1 + 2𝜙
< 𝛽 < 1. (147)
and indeed it can be greater than 1. 3
With the above, a summary for this special model is given: Additional analysis can be found in Mavko et al.61 and Sevostianov.62

11
A.H.D. Cheng International Journal of Rock Mechanics and Mining Sciences 142 (2021) 104754

Table 4
Compressibilities and effective stress coefficients evaluated from the measured 𝐾, 𝐾𝑠′ and 𝐾𝑠′′ values for cases in Table 3.
Rock type 𝜙 𝐶𝑠 𝐶𝑝 𝐶𝜙 𝐶𝜓 𝐶 𝛼 𝛽 𝜎 𝜒
(GPa−1 ) (GPa−1 ) (GPa−1 ) (GPa−1 ) (GPa−1 )
Lavoux limestone 0.23 0.027 0.43 0.091 −0.0045 0.12 0.81 0.98 0.076 1.05
Vilhanneur limestone 0.14 0.019 0.23 0.029 −0.0015 0.050 0.64 0.96 0.070 1.05
Opalinus clay 0.13 0.11 3.3 0.41 0.0046 0.54 0.79 0.96 0.17 0.99
Bentheimer sandstone 0.24 0.03 0.19 0.036 −0.0032 0.071 0.62 0.91 0.15 1.09
Berea sandstone 0.23 0.031 0.16 0.027 0.0033 0.071 0.52 0.72 0.30 0.88

8. Physical insight for bulk material constants where


𝐾𝑝
𝛾=𝛽+ . (154)
The macroscopic material constants, such as 𝐾, 𝐾𝑢 , 𝐵, 𝑀, 𝛼, 𝛽, 𝐾𝑓
𝐶𝑏𝑐 , 𝐶𝑝𝑐 , etc., are bulk constants that incorporate several of the above
With the definition (41), (153) is equivalent to (6), but written in an
discussed physical mechanisms. In this section we shall explore their effective stress form for the fluid expelled from the frame, with 𝛾 as
relations with the intrinsic constants in order to isolate the mechanisms. its effective stress coefficient.41 Assuming the positiveness of 𝐾𝑝 , it is
The experiments that passed the various tests, shown in Table 3, are obvious that
used for illustration. In Table 4, the data in Table 3 are used to evaluate
the compressibilities, as well as the effective stress coefficients, based 𝛾 > 𝛽. (155)
on the various relations already established. In an undrained test, 𝛥𝑉𝑓′ = 0, and (153) reduces to
We observe from the effective stress coefficients that 𝜒, correspond-
ing to porosity change, is the largest, and can be greater than 1. The 𝑝 = 𝐵𝑃 , (156)
coefficient 𝜎, for solid volume change, is the smallest, and close to zero
where 𝐵 is the Skempton pore pressure coefficient, and
for limestones. The coefficients 𝛽, for pore volume change, and 𝜒 are
( )
close to each other. 1 𝜕𝑝 𝐾𝑓 𝐾𝑠′′ (𝐾𝑠′ − 𝐾)
𝐵= = = . (157)
Next, we examine the contributions of the various mechanisms to 𝛾 𝜕𝑃 𝑉 ′ 𝜙𝐾𝐾𝑠 𝐾𝑠 + 𝐾𝑓 (𝐾𝑠′ 𝐾𝑠′′ − 𝐾𝐾𝑠′′ − 𝜙𝐾𝐾𝑠′ )
′ ′′
𝑓
the volume changes of the frame and the pore space. We can write the
constitutive equations (21), (22), (54), and (55) into the following: Or, in compressibility form, we have

𝛥𝑉 ( ) ( ) 𝐶 − 𝐶𝑠′
= − 𝐶𝑠 + 𝜙𝐶𝑝 + 𝐶𝜓 𝑃 ′′ = − 𝐶𝑠 + 𝜙𝐶𝑝 + 𝐶𝜓 𝑃 + 𝜙 𝐶𝑝 𝑝, (148) 𝐵 =
𝜙𝐶𝑓 + 𝐶 − 𝐶𝑠′ − 𝜙𝐶𝑠′′
.
𝑉
( )
𝛥𝑉𝑝 𝐶𝜓 𝜙𝐶𝑝
= −𝐶𝑝 𝑃 ′′′ = −𝐶𝑝 𝑃 + 𝐶𝑝 − 𝐶𝑠 − 𝑝, (149) = . (158)
𝑉𝑝 𝜙 𝜙(𝐶𝑓 + 𝐶𝑝 − 𝐶𝑠 ) − 𝐶𝜓
( )
𝛥𝑉𝑠 𝐶𝑠 𝐶𝜓
= − + 𝑃 ′′′′ , (150) By the inequalities (132) and (C.4), it is easy to show that
𝑉𝑠 1−𝜙 1−𝜙
( ) 𝐵 > 0. (159)
𝛥𝜙 𝜙 𝐶𝑠 𝜙 𝐶𝜓
= −𝐶𝜙 𝑃 ′′′′′ = − 𝜙 𝐶𝑝 − − 𝑃 ′′′′′ . (151)
1−𝜙 1−𝜙 1−𝜙 The constant, however, is not bounded by 1. We can show from (158)
utilizing relations in (32), (33), (58), (59), and (65)–(69). We observe that
from the above that these volume changes are controlled by their
𝐵 > 1, if 𝐶𝑠′′ > 𝐶𝑓 or 𝐾𝑓 > 𝐾𝑠′′ . (160)
respective effective stresses multiplied by the partitioned compressibil-
ities. For 𝛥𝑉 and 𝛥𝑉𝑝 , we also expanded the effective stresses to show That is, if the solid is more compressible than the fluid. A greater than
the individual contributions attributed to 𝑃 and 𝑝. 1 value for 𝐵 was indeed measured for polyurethane foam saturated
In Table 5, we show the percentage contribution to 𝛥𝑉 , 𝛥𝑉𝑠 , 𝛥𝑉𝑝 , with a silicon fluid.66
and 𝛥𝜙, attributed to each of the three mechanisms. For the total To find 𝐾𝑢 , we substitute (156) into (21) to eliminate 𝑝, and obtain
volume change 𝛥𝑉 , we observe that both the solid and the pore
compressibilities are important, and the microinhomogeneity effect 𝛥𝑉 𝑃
=− . (161)
is relatively small, less than 5% for these rocks. For pore volume 𝑉 𝐾𝑢
change 𝛥𝑉𝑝 , it is entirely derived from the pore compressibility 𝐶𝑝 . The where
solid volume change 𝛥𝑉𝑠 is dependent on both 𝐶𝑠 and 𝐶𝜓 , with non- ( )
1 𝜕𝑉 𝐾
𝐾𝑢 = − = . (162)
negligible effect from microinhomogeneity. For the change in porosity 𝑉 𝜕𝑃 𝑉𝑓′ 1 − 𝛼𝐵
𝛥𝜙, we find that the influence of 𝐶𝜓 is small, less than 4%.
We can also express 𝐾𝑢 as
The micromechanical analysis also allows the incorporation of the
fluid effect. Of particular interest is the undrained bulk modulus 𝐾𝑢 𝐾𝑓 𝐾𝑠′′ (𝐾𝑠′ − 𝐾)2
𝐾𝑢 = 𝐾 + , (163)
defined in (5) and (7). The Gassmann equation13,63 is an important 𝜙𝐾𝑠′ 2 𝐾𝑠′′ + 𝐾𝑓 [𝐾𝑠′ 𝐾𝑠′′ − 𝐾𝐾𝑠′′ − 𝜙𝐾𝑠′ 2 ]
formulas in seismicity64,65 that allows the compressive wave velocity
or in compressibility form
in saturated rock to be evaluated from the laboratory measurement of
dry rock. The analysis can be performed as follows. (𝐶 − 𝐶𝑠′ )2
𝐶𝑢 = 𝐶 −
Eqs. (14), (40) and (42) allow 𝛥𝑉𝑝 to be expressed as 𝜙𝐶𝑓 + 𝐶 − 𝐶𝑠′ − 𝜙𝐶𝑠′′
𝑝𝑉𝑝 𝜙2 𝐶𝑝2
𝛥𝑉𝑝 = − − 𝛥𝑉𝑓′ . (152) = 𝐶− (164)
𝐾𝑓 𝜙𝐶𝑓 + 𝜙(𝐶𝑝 − 𝐶𝑠 ) − 𝐶𝜓

Substituting the above into (22), we obtain which is the equation given by Brown and Korringa.7 As it is obvious
that
𝛥𝑉𝑓′ 1
= (𝑃 − 𝛾𝑝), (153) 𝐾𝑓 > 0, 𝐶𝑓 > 0, (165)
𝑉𝑓 𝐾𝑝

12
A.H.D. Cheng International Journal of Rock Mechanics and Mining Sciences 142 (2021) 104754

Table 5
Partition of volumetric deformation due to effective stress into the solid, pore, and microinhomogeneity components.
Rock type 𝛥𝑉 𝛥𝑉𝑝 𝛥𝑉𝑠 𝛥𝜙
𝐶𝑠 𝐶𝑝 𝐶𝜓 𝐶𝑝 𝐶𝑠 𝐶𝜓 𝐶𝑠 𝐶𝑝 𝐶𝜓
(%) (%) (%) (%) (%) (%) (%) (%) (%)
Lavoux limestone 23 81 −3.7 100 120 −20 −8.9 107 1.5
Vilhanneur limestone 39 64 −2.9 100 108 −8.1 −11 110 0.8
Opalinus clay 20 79 0.8 100 96 4.1 −3.9 104 −0.2
Bentheimer sandstone 42 62 −4.4 100 112 −12 −27 124 2.8
Berea sandstone 44 52 4.6 100 91 9.5 −35 139 −3.7

Table 6
Comparison of 𝐾𝑠 , 𝐾𝑠′ , 𝐾𝑠′′ , 𝐾, and 𝐾𝑚 for cases in Table 3, except for Opalinus clay, whose 𝐾𝑚 is not available. 𝐾𝑢 is
evaluated from Eq. (163), and 𝐾𝑢∗ from Eq. (167).
Rock type [Ref] 𝐾𝑠 𝐾𝑠′ 𝐾𝑠′′ 𝐾 𝐾𝑚 𝐾𝑢 𝐾𝑢∗
(GPa) (GPa) (GPa) (GPa) (GPa) (GPa) (GPa)
Lavoux limestone31 37 44 130 8.3 77 14 15
Vilhanneur limestone31 51 56 110 2.0 77 26 28
Bentheimer sandstone36 33 37 59 14 37 17 17
Berea sandstone25,37 32 29 22 14 37 16 17

we can prove in Appendix C that In this case, there is no clear partition. We also notice the relation18

𝐾𝑢 > 𝐾, 𝐶𝑢 < 𝐶. (166) 𝑆𝑒 𝐾


= . (172)
𝑆𝜎 𝐾𝑢
For ideal porous medium, (163) reduces to the Gassmann equation13
In Table 7, we show the calculated 𝐵, 𝑆𝜎 and 𝑆𝑒 values, by assuming
𝐾𝑓 (𝐾𝑚 − 𝐾)2 that rocks are saturated with water. In the same table, we also show the
𝐾𝑢 = 𝐾 + . (167)
𝜙𝐾𝑚 2 + (1 − 𝜙)𝐾𝑚 𝐾𝑓 − 𝐾𝐾𝑓 partition of 𝑆𝜎 into the contributions by the solid, pore, microinhomo-
geneity, and fluid compressibility effects. We notice that for these rocks,
The advantage of the above equation is that the value of 𝐾𝑚 can be
the contributions by 𝐶𝑠 and 𝐶𝜓 are relatively small. These relations also
found from the published value for the mineral, and only a test on dry
show that if the rock is permeated by a gas, instead of a liquid, then
rock for 𝐾 is needed to apply the formula.
𝐶𝑓 will be much larger than other compressibilities. Then the storage
In Table 6 we summarize the values of 𝐾𝑠 , 𝐾𝑠′ , 𝐾𝑠′′ , and 𝐾 for a
coefficient reduces to
comparison with 𝐾𝑚 , for the cases shown in Table 3. We observe that
these values can be very different. We then use both (163) and (167) 𝑆 𝜎 ≈ 𝜙 𝐶𝑓 . (173)
to calculate 𝐾𝑢 , by assuming that rocks are saturated by water with
𝐾𝑓 = 2.2 GPa. We notice that the resultant 𝐾𝑢 values are close to each 9. Stress dependent response
other, despite that 𝐾𝑚 does not represent 𝐾𝑠′ and 𝐾𝑠′′ at all. The reason
is that 𝐾𝑢 is dominated by the values of 𝐾 and 𝐾𝑓 , and has relative The constitutive relations of porous rock are generally nonlinear.
small dependence on the solid moduli, 𝐾𝑠′ and 𝐾𝑠′′ . Hence the Gassmann The material ‘‘constants’’ discussed above are in fact variables depen-
formula can give a good estimate of seismic wave velocity in saturated dent on the deformation state. As strains are more difficult to measure
rocks. and control, it is a common practice to express constitutive constants
Another fluid related constant of interest is the storage coefficient. as functions of stresses. The proposed relations are generally empirical,
The constant gives an indication of the amount of fluid that can be fitted using laboratory data. The past investigations have produced a
released from (or absorbed into) the porous frame with a change in number of such observations. In this section, we shall examine some of
confining stress or pore pressure. The storage coefficient can be defined the reported results, with an attempt to explain their behaviors using
a number of ways,4,18 depending on the applications. For example, we the intrinsic constants.
can define a constant stress storage coefficient 𝑆𝜎 as the volume of fluid Tarokh and Makhnenko53 have conducted a comprehensive survey
gained per unit volume of frame and per increment of pore pressure, of past efforts. We shall comment on some of the cases reported therein.
under constant confining stress condition, as Among the earliest efforts were those by Adams and Williamson67 and
( ′)
1 𝜕𝑉𝑓 𝜙𝛾 Zisman.68 Adams and Williamson67 found that for basalt and diabase,
𝑆𝜎 = − = , (168) 𝐾𝑠′ > 𝐾 (𝐶 > 𝐶𝑠′ ), and for marble and granite, 𝐾𝑠′ ≈ 𝐾.4 These findings
𝑉 𝜕𝑝 𝐾𝑝
𝑃 are consistent with the material bound (132). Based on (65) and (66),
based on (153). We an also express 𝑆𝜎 as we find that the difference between 𝐶 and 𝐶𝑠′ is 𝜙𝐶𝑝 . The finding that
𝐶𝑠′ ≈ 𝐶 implies that 𝐶 ≫ 𝜙𝐶𝑝 . Based on (133), 𝐶𝑝 is larger than 𝐶; and
𝑆𝜎 = 𝐶 − 𝐶𝑠′ − 𝜙𝐶𝑠′′ + 𝜙𝐶𝑓
Table 5 shows that it can be much larger. For the above condition to
= −𝜙𝐶𝑠 + 𝜙𝐶𝑝 − 𝐶𝜓 + 𝜙𝐶𝑓 . (169) be true, it requires a very small porosity 𝜙. Indeed, the porosities for
marble and Washington granite reported in Adams and Williamson67
The above equation shows the partition of the storage coefficient into
were 0.008 and 0.006, respectively; so the condition can be satisfied.
the various compressibilities.
What puzzling, however, is that the reported porosities for Palisade
Another storage coefficient is one that is under constant strain,
given as diabase and basalt are also small, 0.0006 and 0.009, respectively. For
( ′) Palisade diabase, 𝐶 = 0.0181 GPa−1 and 𝐶𝑠′ = 0.0144 GPa−1 . These
1 𝜕𝑉𝑓 1
𝑆𝑒 = − = , (170)
𝑉 𝜕𝑝 𝑀
𝑒 4
In Adams and Williamson,67 the enclosed and outer compressibility is
in which we have utilized (2). We find the current drained bulk compressibility 𝐶, and the unenclosed and inner
( )
𝐶′ compressibility is the current unjacketed bulk compressibility 𝐶𝑠′ . The unit of
𝑆𝑒 = 𝐶𝑠′ 1 − 𝑠 − 𝜙𝐶𝑠′′ + 𝜙𝐶𝑓 . (171) megabar defined in footnote 3 on p. 478 should be bar.
𝐶

13
A.H.D. Cheng International Journal of Rock Mechanics and Mining Sciences 142 (2021) 104754

Table 7
Evaluated 𝐵, 𝐾𝑢 , 𝑆𝜎 , and 𝑆𝑒 values, and the partition of 𝑆𝜎 into compressibility components. Fluid is assume to be water,
with 𝐾𝑓 = 2.2 GPa.
Rock type [Ref] 𝐵 𝑆𝜎 𝑆𝑒 −𝜙𝐶𝑠 ∕𝑆𝜎 𝜙𝐶𝑝 ∕𝑆𝜎 −𝐶𝜓 ∕𝑆𝜎 𝜙𝐶𝑓 ∕𝑆𝜎
(GPa−1 ) (GPa−1 ) (%) (%) (%) (%)
Lavoux limestone31 0.49 0.20 0.12 −3.1 49 2.2 52
Vilhanneur limestone31 0.34 0.094 0.074 −2.9 34 1.6 67
Opalinus clay34,35 0.91 0.47 0.13 −3.0 91 −1.0 13
Bentheimer sandstone36 0.30 0.15 0.12 −4.8 30 2.1 73
Berea sandstone25,37 0.28 0.13 0.11 −5.5 28 −2.5 80

measurements lead to an estimate of 𝐶𝑝 = 6.2 GPa−1 , which is too above experiments do not necessarily violate that condition. As shown
large. Possible explanations could be experimental error, or that the in (72), to interpret 𝐾𝑠 , we need to measure both 𝐾𝑠′ and 𝐾𝑠′′ . The latter
reported porosities were too small. Porosity reported for diabase from was missing in those experiments.
other sources are typically more than 10 times larger. We argue in this work that the development of a nonlinear model
Zisman68 conducted experiments on several hard (e.g., granite) should be guided by physical insights, and not merely based on data
and soft (e.g., sandstone) rocks. The measurements showed that the fitting of measured constant, and particularly not on the bulk material
uncovered bulk modulus (𝐾𝑠′ ) is larger than the covered bulk modulus constants. For example, we may observe the stress dependency in the
(𝐾). In addition, Zisman68 observed that 𝐾 is noticeably dependent on bulk constants such as 𝐾, 𝐾𝑢 , 𝐵, etc. Rather than building separate
the applied hydrostatic pressure 𝑃 , whereas 𝐾𝑠′ is not. Based on (65) empirical laws for each of them, we should seek the underlying physical
and (66), the above observations can be interpreted as that the solid mechanism that ties them together, and model that mechanism. In
bulk compressibility 𝐶𝑠 is independent of the loading 𝑃 , while the pore other words, the nonlinear laws should be built upon the intrinsic
compressibility 𝐶𝑝 is dependent. Hence the stress dependency of 𝐶 is material constants.
derived from 𝐶𝑝 , and not 𝐶𝑠 . Based on the theoretical analysis and laboratory evidences, the first
Nur and Byerlee11 conducted unjacketed and drained experiments order importance in building a nonlinear law is to model the pore
on Westerly granite (𝜙 ≈ 0.01). They found that 𝐾𝑠′ is not sensitive to modulus 𝐾𝑝 . As pores are being consolidated, the more compliant
the confining stress 𝑃 for the range up to 250 MPa. They also found that ones will deform first, and possibly close. The progressive change in
𝐾 is dependent on the confining stress 𝑃 . At low 𝑃 , they found 𝐾 < 𝐾𝑠′ . geometry will change the modulus. Table 5 shows that the pore volume
When 𝑃 was increased to above 100 MPa, most cracks were closed, change generally contributes the most to the total volume change. Next
the material became stiffer, and 𝐾 approached 𝐾𝑠′ . These behaviors are we may consider the solid bulk modulus 𝐾𝑠 . The solid phase is generally
consistent with the above interpretation using intrinsic constants. That stiffer, and its structural change may take place at a higher stress
is, within the experimental range of confining stress, 𝐾𝑠 for granite is range. However, if the solid phase is more heterogeneous, containing
a constant. The pore modulus 𝐾𝑝 is initially small (𝐶𝑝 large) due to the hard and soft constituents, such as the case for some sandstones, it
presence of cracks. After the cracks are closed, 𝐶𝑝 decreases to make can also behave nonlinearly. Finally, we may consider the nonlinear
the term 𝜙𝐶𝑝 in (65) negligible; hence 𝐾 approaches 𝐾𝑠′ . behavior of 𝐾𝜓 . At this time, 𝐾𝜓 is more difficult to determine in
Andersen and Jones69 conducted tests on Berea sandstone, and the laboratory. As observed in Tables 5 and 7, its contribution to
found that 𝐾 and 𝐾𝑝 increase with increasing 𝑃 . They also measured 𝐾𝑠′ the deformation is relatively small; so we may leave it as a constant.
under variable 𝑃 but constant 𝑃 ′ , and found that 𝐾𝑠′ was a constant. Once the nonlinear behaviors of these constants are established, the
Warpinski and Teufel70 tested several chalks and one sandstone, and behaviors of the bulk constants will automatically emerge from their
found 𝐾𝑠′ not sensitive to the stress. Many more such observations were relation with the intrinsic constants.
made and surveyed in Tarokh and Makhnenko,53 and are not repeated One more piece of information is needed to build a nonlinear
here. model—the constant’s dependence on an effective stress. As there are
There are, however, recent experiments that indicated 𝐾𝑠′ could be two independent loading components, 𝑃 and 𝑝, a unique relation with
dependent on the effective stress 𝑃 ′ .28,71 For this, we may argue that if a single effective stress is crucial in the development of a model.
the rock is made of a single mineral, it is unlikely that 𝐾𝑚 can be depen- Past efforts in demonstrating the stress dependency of bulk material
dent on the stress. However, for a microscopically inhomogeneous rock constants, such as 𝐾, 𝐾𝑢 , and 𝐵, either did not identify an effective
made of several linear constituents with certain microstructure, it is stress, or used the Terzaghi effective stress 𝑃 ′ . Here we argue that the
possible that the composite response behaves nonlinearly. Furthermore, identification of an effective stress should follow the physics. Observing
we should be aware that 𝐾𝑠′ is not 𝐾𝑠 , though 𝐾𝑠 was often used as from the effective stress relation (22) or (149), we anticipate 𝐶𝑝 to
the notation in the literature. Based on (66), we also need to consider be a function of 𝑃 ′′′ , which is associated with the effective stress
the contribution from 𝐾𝜓 . As 𝐾𝑝 can behave nonlinearly, so can 𝐾𝜓 . coefficient 𝛽. There is, however, a practical difficulty in using 𝑃 ′′′ ,
However, at this time, very little is known about the behavior of 𝐾𝜓 . because according to (69), 𝛽 is dependent on 𝐶𝑝 . Hence an iterative
In the experiments by Warpinski and Teufel,70 they found that the procedure may be needed. Equation (69) also shows that 𝛽 is close to
measured 𝐾𝑠′ for chalk was significantly lower than 𝐾𝑚 of calcite. Hart one if 𝐶𝑝 ≫ 𝐶𝑠 . As shown in Table 4, this is indeed the situation for
and Wang23 performed unjacketed experiments on Berea sandstone most rocks; so we may simply adopt the Terzaghi effective stress 𝑃 ′ as
and Indiana limestone. They found that 𝐾𝑠′ for Berea sandstone is a close approximation to build the nonlinear relation.
noticeably smaller than 𝐾𝑚 of quartz, likely due to the presence of In the following, we shall give the above rationale a test, using
kaolinite in its composition. They also found that for Indiana limestone, a published data set. Hart and Wang30 performed experiments on
𝐾𝑠′ is close to 𝐾𝑚 of calcite. Fabre and Gustkiewicz72 tested several Berea sandstone over a range of confining stresses and pore pressures.
limestones and sandstones. Mixed results were reported. For example, They measured the drained bulk compressibility 𝐶, unjacketed bulk
𝐾𝑠′ for several sandstones were much lower than 𝐾𝑚 , and for others, compressibility 𝐶𝑠′ , undrained bulk compressibility 𝐶𝑢 , and Skempton
they could be very close. There are recent experiments that reported pore pressure coefficient 𝐵. The measured 𝐵 was used in (158) to invert
opposite observation, with 𝐾𝑠′ > 𝐾𝑚 . For example, Ingraham et al.71 for the unjacketed pore compressibility 𝐶𝑠′′ . Two selected results, at the
examined the unjacketed response of Castlegate sandstone, and found stress level 𝑃 ′ = 14.5 and 25.3 MPa, were reported in Tables 1 and 2
𝐾𝑠′ significantly higher than 𝐾𝑚 of quartz. Mitra et al.73 also reported to check their reliability. The interpreted 𝐾𝑠′′ values, however, led to
a 𝐾𝑠′ greater than 𝐾𝑚 for Berea sandstone. In Section 7, we established unreasonable estimates of 𝐾𝑠 , as shown in Table 2, and these data were
the condition 𝐾𝑚 > 𝐾𝑠 . We note, however, 𝐾𝑠′ is not 𝐾𝑠 ; hence the rejected. As discussed in Section 4, such discrepancy could be caused

14
A.H.D. Cheng International Journal of Rock Mechanics and Mining Sciences 142 (2021) 104754

Table 8
Measured 𝐶, 𝐶𝑠′ , 𝐵, and 𝐶𝑢 values for Berea sandstone under various stress and pressure conditions, extracted from Table 1 of Hart and Wang30 ; and calculated
𝐶𝑠 , 𝐶𝑝 , 𝐶𝜙 , 𝐶𝜓 , 𝐶𝑠′′ , 𝐵, and 𝐶𝑢 values. 𝐶𝑓 is based on water at 20 ◦ C.
𝑃′ 𝑝 𝜙 𝐶𝑓 𝐶 𝐶𝑠′ 𝐵 𝐶𝑢 𝐶𝑝 𝐶𝜙 𝐶𝜓 𝐶𝑠 𝐶𝑠′′ 𝐵 𝐶𝑢
MPa MPa GPa−1 GPa−1 GPa−1 GPa−1 GPa−1 GPa−1 GPa−1 GPa−1 GPa−1 GPa−1
0.4 0.7 0.19 0.459 0.281 0.036 0.87 0.071 1.29 0.237 0.009 0.027 0.074 0.77 0.092
0.6 7.3 0.19 0.458 0.263 0.043 0.89 0.059 1.16 0.210 0.009 0.034 0.081 0.75 0.097
1.2 17.1 0.19 0.458 0.309 0.040 0.83 0.065 1.42 0.260 0.009 0.031 0.078 0.79 0.097
2.5 7.9 0.19 0.456 0.299 0.034 0.75 0.090 1.39 0.257 0.009 0.025 0.072 0.78 0.091
2.7 32.1 0.19 0.456 0.239 0.035 0.75 0.070 1.07 0.196 0.009 0.026 0.073 0.74 0.089
2.8 6.9 0.19 0.456 0.218 0.029 0.75 0.080 0.99 0.182 0.009 0.020 0.067 0.72 0.082
3.0 2.8 0.19 0.456 0.225 0.037 0.74 0.085 0.99 0.179 0.009 0.028 0.075 0.72 0.089
4.0 16.7 0.19 0.455 0.230 0.031 0.68 0.087 1.05 0.192 0.009 0.022 0.069 0.73 0.085
4.1 15.9 0.19 0.454 0.195 0.030 0.72 0.079 0.87 0.158 0.009 0.021 0.068 0.69 0.081
5.5 11.7 0.19 0.453 0.174 0.032 0.61 0.081 0.75 0.134 0.009 0.023 0.070 0.66 0.080
5.5 20 0.19 0.453 0.175 0.029 0.67 0.079 0.77 0.139 0.009 0.020 0.067 0.67 0.078
5.7 6.7 0.19 0.453 0.172 0.034 0.61 0.084 0.73 0.130 0.009 0.025 0.072 0.66 0.081
8.9 13.3 0.19 0.449 0.138 0.032 0.53 0.076 0.56 0.098 0.009 0.023 0.070 0.60 0.075
9.0 25.5 0.19 0.449 0.137 0.031 0.59 0.074 0.56 0.099 0.009 0.022 0.069 0.60 0.074
9.2 24.8 0.19 0.449 0.131 0.030 0.55 0.074 0.53 0.094 0.009 0.021 0.068 0.58 0.072
9.7 10.4 0.19 0.448 0.135 0.031 0.54 0.075 0.55 0.097 0.009 0.022 0.069 0.59 0.074
11.1 9.6 0.19 0.447 0.124 0.029 0.53 0.075 0.50 0.088 0.009 0.020 0.067 0.57 0.070
12.6 21.9 0.19 0.445 0.106 0.031 0.49 0.071 0.39 0.068 0.009 0.022 0.069 0.51 0.068
14.5 16.6 0.19 0.443 0.103 0.032 0.48 0.067 0.37 0.063 0.009 0.023 0.070 0.50 0.067
18.6 22.1 0.19 0.438 0.090 0.029 0.46 0.062 0.32 0.054 0.009 0.020 0.067 0.46 0.062
21.3 9.8 0.19 0.436 0.083 0.029 0.44 0.061 0.28 0.047 0.009 0.020 0.067 0.44 0.059
25.3 16.1 0.19 0.431 0.078 0.028 0.42 0.059 0.26 0.043 0.009 0.019 0.066 0.42 0.057
31.5 9.9 0.19 0.424 0.072 0.028 0.40 0.057 0.23 0.037 0.009 0.019 0.066 0.39 0.055
33.1 7.6 0.19 0.423 0.075 0.029 0.40 0.059 0.24 0.039 0.009 0.020 0.067 0.41 0.056
35.7 5.7 0.19 0.420 0.070 0.028 0.39 0.056 0.22 0.035 0.009 0.019 0.066 0.38 0.054

by the high sensitivity of 𝐾𝑠′′ to the perturbation of 𝐵 in the inversion. Table 9


Calculated material constants for Berea sandstone at various stress levels.
Here, assuming that the original data set was reliable, we shall give a
different analysis based on the intrinsic constants. 𝑃′ 𝑆𝜎 𝑀 𝛼 𝛽 𝜎 𝜒
MPa GPa−1 GPa
In Table 8, the data of 𝑃 ′ , 𝑝, 𝐶, 𝐶𝑠′ , 𝐵, and 𝐶𝑢 are extracted from
2.5 0.34 9.7 0.89 0.95 0.40 0.96
Hart and Wang’s Table 1, and displayed as columns 1, 2, and 5–8. In
5.5 0.21 10.1 0.82 0.91 0.42 0.93
Figs. 5(a) and (b), we plot the 𝐶 and 𝐶𝑠′ values versus 𝑃 ′ . The plots show 9.7 0.18 10.4 0.77 0.87 0.43 0.91
that there is a large reduction in the compressibility 𝐶 with increasing 14.5 0.14 10.8 0.69 0.81 0.42 0.86
effective stress. On the other hand, the change in 𝐶𝑠′ is relatively small. 21.3 0.12 11.3 0.65 0.76 0.44 0.81
We also observe that the data fluctuate in the low effective stress range, 25.3 0.12 11.5 0.64 0.75 0.45 0.79
31.5 0.11 11.7 0.61 0.71 0.45 0.76
and are likely unreliable. No attempt was made to smooth the data. 35.7 0.11 11.9 0.60 0.70 0.45 0.75
In columns 3 and 4 of Table 8, we fill in the values of porosity 𝜙
and compressibility of water 𝐶𝑓 . The initial porosity is set at 0.19. Using
(51), we estimate that the change in porosity 𝛥𝜙 over the stress range
is of the order 0.001. Hence the initial porosity is used throughout. The above practice gives the confidence that the proposed model
For water at 20◦ C, its compressibility varies linearly from 0.459 GPa−1 can predict the behaviors of the composite constants through the iso-
at 𝑝 = 0 MPa to 0.415 GPa−1 at 𝑝 = 40 MPa;74 and 𝐶𝑓 is evaluated lated physical mechanisms. Using the set of intrinsic constants, and the
accordingly. In columns 9 and 10, 𝐶𝑝 and 𝐶𝜙 are evaluated using (30) various relations presented above, we calculate the constants 𝑆𝜎 , 𝑀, 𝛼,
and (B.5). 𝛽, 𝜎, and 𝜒 for a few of selected stresses, and present them in Table 9.
At this point, we are missing one piece of information, either 𝐶𝑠′′ We observe that the constant stress storage coefficient 𝑆𝜎 reduces to
or 𝐶𝜓 , to complete the model. Rather than performing an inversion less than 1/3 of its initial value with the increasing stress. The constant
based on the measured values of 𝐵 or 𝐶𝑢 , we shall assign a 𝐶𝜓 value by strain storage coefficient 𝑆𝑒 = 1∕𝑀, on the other hand, reduces for
trial and error. A constant value 𝐶𝜓 = 0.009 GPa−1 , which satisfies the about 20%. We also find that the effective stress coefficients 𝛼, 𝛽, and
stability criterion (129), is entered into columns 11, without an effort 𝜒 are sensitive to the stress loading. The coefficient 𝜎, associated with
the solid deformation, is less sensitive. These coefficients satisfy the
to adjust for the different stress level. With the set of data, 𝐶𝑠 , 𝐶𝑝 , 𝐶𝜓 ,
inequalities (127) and (136).
𝐶𝑓 , and 𝜙, the model is now complete, and all other constants can be
evaluated. Using (66) and (67), 𝐶𝑠 and 𝐶𝑠′′ are calculated and shown
10. Summary and conclusion
in the next two columns. By (158) and (164), the last two columns of
Table 8 give the calculated 𝐵 and 𝐶𝑢 values. The original Biot theory of poroelasticity1 was modeled at the
For a comparison, the measured and the predicted values of 𝐵 and macroscopic level and considered the externally observed quantities,
𝐶𝑢 are plotted in Figs. 6(a) and (b). The match was performed by such as deformation of the frame, and the fluid escaping from it.
selecting a single 𝐶𝜓 value that gives the best prediction of 𝐵 in the The material constants, such as the drained bulk modulus 𝐾, effective
higher 𝑃 ′ range, and we observe in Fig. 6(a) a near perfect match in stress coefficient 𝛼, and Skempton pore pressure coefficient 𝐵, are bulk
the range of 𝑃 ′ > 15 MPa. In the low 𝑃 ′ range, we find similar trend. constants that incorporate the physical properties of the solid, the pore
The fluctuation of the curve and the deviation from measured data space, and the fluid. The Biot and Willis micromechanical analysis6
are caused by the poor quality data in the 𝑃 ′ < 10 MPa range (see modeled the internal deformation using the total and pore volume as
Fig. 5). In Fig. 6(b), we observe that the effort of matching 𝐵 led to the variables. The resulting modeling constants, such as 𝐾, 𝐾𝑠′ , and 𝐾𝑠′′ , are
automatic matching of 𝐶𝑢 , without additional fitting. The comparison still composite constants that mixed physical mechanisms in them.
in the 𝑃 ′ > 5 MPa range is good. Below 5 MPa, the simulated results In the present work, we offer an alternative micromechanical analy-
are more reasonable than the measured ones. sis using solid volume and porosity as variables. It is aimed at isolating

15
A.H.D. Cheng International Journal of Rock Mechanics and Mining Sciences 142 (2021) 104754

Fig. 5. Plot of stress dependent compressibilities of Berea sandstone. Data from Table 1 of Hart and Wang.30 (a) 𝐶 vs. 𝑃 ′ , and (b) 𝐶𝑠′ vs. 𝑃 ′ .

the different physical mechanisms, hence the material constants. Three than individually fitting the stress dependent experimental data of the
constants, an unjacketed solid bulk modulus 𝐾𝑠 , a drained pore modu- various bulk constants, their underlying physical mechanisms should be
lus 𝐾𝑝 , and an unjacketed porosity modulus 𝐾𝜓 , are presented. They uncovered. The nonlinear model should be built on the corresponding
are called the intrinsic material constants, because they respectively intrinsic constants. It is recommended that a nonlinear model should
represent the effective rigidity of the solid, the pore space, and the non- first be built on 𝐾𝑝 , for it has the largest contribution to deformation,
self-similar deformation of the material under unjacketed test, due to and is more dependent on the changes in pore geometry. If necessary,
the microinhomogeneity and microanisotropy of the solid phase. a nonlinear model can be introduced for 𝐾𝑠 . The constant 𝐾𝜓 repre-
A parallel derivation based on the variational energy principles are senting the microinhomogeneity effect has the smallest contribution. It
presented, using the solid strain and porosity to represent the free can be left as a constant.
energy of the system. The links between the two approaches are es- To demonstrate such idea, a published data set is reanalyzed. Based
tablished. Relations among the macroscopic and microscopic constants, on the measured stress dependent data of 𝐾 and 𝐾𝑠′ , a selection of a
and the constitutive constants of the free energy, are derived. The constant 𝐾𝜓 is made to match the measured Skempton pore pressure
variational energy approach is used to perform a stability analysis of the coefficient 𝐵. The undrained bulk modulus 𝐾𝑢 evaluated from the data
system. The bounds of material constants are presented for the general set then automatically gives a good fit to the measured data. This is a
case, a practical case, and an ideal porous medium case. confirmation of the correct modeling of the underlying physics.
The bulk and micromechanical constants are routinely measured in For the closing, we may add a few comments on future research. The
the laboratory, with the exception of 𝐾𝑠′′ , which requires the measure- current analysis assumes that rock is a macroscopically isotropic mate-
ment of pore volume change under varying pore pressure. Only about rial, though microanisotropy is allowed. For macroscopically
twenty measurements were conducted in the past, which are compiled anisotropic rocks, micromechanical analysis has been performed by
in this work. By performing a check using the material bounds, we find Cheng.75 The analysis requires the introduction of the pore strain tensor
that most of the reported 𝐾𝑠′′ values failed the criteria. together with its constitutive constants as a fourth rank tensor. While
Using the sets of data that survived the bound criteria, the intrinsic the frame deformation of porous material can be measured the same
constants are evaluated. These constants allow the identification of con- way as that of an anisotropic elastic solid, it is generally not feasible
tribution to deformation from each physical mechanism. For example, to measure the pore deformation as a tensor in order to resolve its
the total volume change under drained test can be partitioned into anisotropic material constants. The inability for such measurements
percentage contributions derived from the solid compressibility, the makes it difficult to assess the microanisotropy effects; hence it is more
pore compressibility, and the microinhomogeneity effect. The same is sensible to adopt the microisotropy assumption. It was demonstrated in
true for the fluid storage effect. Cheng75 that the 28 macroscopic anisotropy material constants can be
Using the intrinsic constants, a rational modeling of the nonlinear fully defined within the 21 elastic constants of the frame, plus a solid
behavior of material under increasing stress loading is proposed. Rather bulk modulus 𝐾𝑠 , and the Biot modulus 𝑀.

16
A.H.D. Cheng International Journal of Rock Mechanics and Mining Sciences 142 (2021) 104754

Fig. 6. Comparison of measured and predicted 𝐵 and 𝐶𝑢 of Berea sandstone at various effective stresses 𝑃 ′ . Measured data are from Table 1 of Hart and Wang.30 (a) 𝐵 vs. 𝑃 ′ ,
and (b) 𝐶𝑢 vs. 𝑃 ′ .

( )
1 1 1 𝑠
Another area that requires attention is the double or multiple poros- 𝜖 𝑠𝑖𝑗 = 𝑒𝑠 𝑑𝒙 = 𝑢 𝑛 + 𝑢𝑠𝑗 𝑛𝑖 𝑑𝒙, (A.2)
ity nature of rocks. In most of the confining stress dependent measure- 𝛺𝑠 ∫𝛺𝑠 𝑖𝑗 𝛺𝑠 ∫𝛴𝑠 2 𝑖 𝑗
ments reviewed in this work, the rock deformation exhibits two stages. in which 𝑡𝑠𝑖 = 𝜎𝑖𝑗𝑠 𝑛𝑗 is the boundary traction, with 𝑛𝑗 the components of
Starting from zero or low confining stress, the rock is more compliant. the outward normal of the surface 𝛴𝑠 , and 𝛺𝑠 is the domain of the
Beyond certain stress threshold, it rapidly stiffens. This is an indication
solid phase, inclusive of all the solid constituents and the occluded
that the rock contains at least two types of pores, a slit or fracture type,
pores (see Fig. 3). We note that in the above we have utilized the Gauss
and a rounded type. The more compliant fracture type rapidly closes
(divergence) theorem and the equilibrium equation
after an initial loading, and the mechanical response of the second type
manifests. The experiments and the current analysis were focused on 𝑠
𝜎𝑖𝑗,𝑗 = 0, (A.3)
the second stage of deformation. To analyze the full range, the double-
porosity dual-permeability model of poroelasticity76–78 can be utilized. to convert from the volume to surface integral.
This is, however, left for a potential future investigation. As we are seeking the volumetric deformation of a macroscopically
isotropic material, we can contract the above equations to
Declaration of competing interest
1 𝑝
𝑠𝑠𝑘𝑘 = 𝑡𝑠 𝑥 𝑑𝒙 = − 𝑛 𝑥 𝑑𝒙 = −3𝑝 (A.4)
The authors declare that they have no known competing finan- 𝛺𝑠 ∫𝛴𝑠 𝑘 𝑘 𝛺𝑠 ∫𝛴𝑠 𝑘 𝑘
cial interests or personal relationships that could have appeared to 1 𝛥𝛺𝑠 𝛥𝑉𝑠
influence the work reported in this paper. 𝜖 𝑠𝑘𝑘 = 𝑢𝑠 𝑛 𝑑𝒙 = = (A.5)
𝛺𝑠 ∫𝛴𝑠 𝑘 𝑘 𝛺𝑠 𝑉𝑠

Appendix A. Proof of 𝑲𝒔 as effective bulk modulus of solid phase We note that in the above we have made the substitution of 𝑡𝑠𝑘 = 𝑝 𝑛𝑘
under the unjacketed test condition, and utilized the identity
Following the composite materials theory,79,80 we can perform
volume averaging on stresses and strains in a representative elementary 1
𝑥 𝑛 𝑑𝒙 = 3. (A.6)
volume (REV) of a heterogeneous material, and build constitutive 𝛺𝑠 ∫𝛴𝑠 𝑖 𝑖
relations between the two sets. The resultant constitutive constants are Hence we can construct the constitutive equation
the effective elastic moduli.
Here we shall focus only on the solid phase. We define the average
𝛥𝑉𝑠 𝑝 ||
=− , (A.7)
solid stress tensor 𝑠𝑠𝑖𝑗 and the average solid strain tensor 𝜖 𝑠𝑖𝑗 as follows4 𝑉𝑠 𝐾𝑠 ||𝑃 =𝑝

1 1 with 𝐾𝑠 identified as the effective bulk modulus of the solid phase that
𝑠𝑠𝑖𝑗 = 𝜎 𝑠 𝑑𝒙 = 𝑡𝑠 𝑥 𝑑𝒙, (A.1)
𝛺𝑠 ∫𝛺𝑠 𝑖𝑗 𝛺𝑠 ∫𝛴𝑠 𝑖 𝑗 relates the volume averaged stress to the volume averaged strain.

17
A.H.D. Cheng International Journal of Rock Mechanics and Mining Sciences 142 (2021) 104754

Appendix B. Relations among material constants First, we shall examine the negative root, which corresponds to a
positive 𝐶𝜓 . Substituting (C.1) into the second relation of (B.8) and
Equations (64)–(67) can be expressed in terms of moduli as (B.12) to eliminate 𝐶𝛾 , we find the difference between 𝐶𝑝 and 𝐶𝑠′′ as
𝜙(1 − 𝜙)𝐾𝑠 𝐾𝜙 𝐾𝜓
𝐾𝑝 = , (B.1) √ √
𝜙𝐾𝜙 𝐾𝜓 + 𝐾𝑠 [𝜙𝐾𝜙 + (1 − 𝜙)𝐾𝜓 ] 2𝜙(1 − 𝜙) 𝐶𝑠 𝐶𝛼 𝐶𝛽 𝐶𝑠 − 𝐶𝛼 + 𝐶𝑠 [𝐶𝛽 + 𝜙2 (1 − 𝜙)2 𝐶𝛼 ]
𝐶𝑝 − 𝐶𝑠′′ = . (C.2)
𝐾𝑠 𝐾𝑝 𝐾𝜓 𝜙(1 − 𝜙)3
𝐾 = , (B.2)
𝐾𝑠 𝐾𝑝 + 𝜙𝐾𝑠 𝐾𝜓 + 𝐾𝑝 𝐾𝜓 As all terms on the right hand side are positive, we have
𝐾𝑠 𝐾𝜓
𝐾𝑠′ = , (B.3) 𝐶𝑝 − 𝐶𝑠′′ > 0. (C.3)
𝐾𝑠 + 𝐾𝜓
𝜙𝐾𝑠 𝐾𝜓 Hence (134) is proven. For the case of the positive root in (C.1), 𝐶𝜓
𝐾𝑠′′ = . (B.4) is negative. Equations (66) and (67) indicate that 𝐶𝑠′ > 𝐶𝑠′′ . In view of
𝐾𝑠 + 𝜙𝐾𝜓
(132) and (133), the inequality (134) is again proven.
Similar to (72) and (73), we present
To prove 𝐾𝑢 > 𝐾, we need the quantity contained in the paren-
𝐶𝑠′ theses in the denominator of (164) to be greater than zero. This is
𝐶𝜙 = 𝐶 − , (B.5)
1−𝜙 accomplished by substituting in the constitutive constants {𝐶𝑠 , 𝐶𝛽 , 𝐶𝛾 }
𝐶𝑠′ to obtain
𝐶𝜎 = , (B.6)
1−𝜙 𝐶𝛽 𝐶𝛾2 + 𝐶𝑠 [𝐶𝛽 − 𝜙(1 − 𝜙)𝐶𝛾 ]2
𝐶 − 𝐶𝑠′ − 𝜙𝐶𝑠′′ = >0 (C.4)
Eqs. (104)–(106) in compressibility forms are (1 − 𝜙)3 𝐶𝛾2
𝐶𝛼 𝐶𝛾2 Hence (166) is proven.
𝐶𝑠 = , (B.7)
𝐶𝛾2 − 𝐶𝛼 𝐶𝛽
[ ] References
𝐶𝛾 𝐶𝛽 𝐶𝛾 + 𝜙(1 − 𝜙)2 𝐶𝛼 𝐶𝛾 − (1 − 𝜙2 )𝐶𝛼 𝐶𝛽
𝐶𝑝 =
𝜙(1 − 𝜙)3 (𝐶𝛾2
− 𝐶𝛼 𝐶𝛽 ) 1. Biot MA. General theory of three-dimensional consolidation. J Appl Phys.
[ ] 1941;12(2):155–164.
𝐶𝛽 𝐶𝛾2 2
+ 𝐶𝑠 𝐶𝛽 + 𝜙(1 − 𝜙)2 𝐶𝛾2 − (1 − 𝜙2 )𝐶𝛽 𝐶𝛾 2. Biot MA. Theory of propagation of elastic waves in a fluid-saturated porous solid.
= (B.8) 1. Low-frequency range. J Acoust Soc Am. 1956;28(2):168–178.
𝜙(1 − 𝜙)3 𝐶𝛾2
3. Biot MA. Theory of propagation of elastic waves in a fluid-saturated porous solid.
𝐶𝛼 𝐶𝛽 𝐶𝛾 2. Higher frequency range. J Acoust Soc Am. 1956;28(2):179–191.
𝐶𝜓 = − . (B.9) 4. Cheng AHD. Poroelasticity. Springer; 2016.
(1 − 𝜙)(𝐶𝛾2 − 𝐶𝛼 𝐶𝛽 )
5. Detournay E, Cheng AHD. Fundamentals of poroelasticity. In: Fairhurst C, ed.
Eqs. (121)–(123) can be expressed in terms of compressibilities as Comprehensive Rock Engineering: Principles, Practice and Projects, Vol. II, Analysis
and Design Method. Oxford/New York: Pergamon Press; 1993:113–171.
follows:
6. Biot MA, Willis DG. The elastic coefficients of the theory of consolidation. ASME
𝐶𝛾 {𝐶𝛽 𝐶𝛾 + (1 − 𝜙)𝐶𝛼 [(1 − 𝜙)𝐶𝛾 − 2𝐶𝛽 ]} J Appl Mech. 1957;24:594–601.
𝐶 = ,
(1 − 𝜙)3 (𝐶𝛾2 − 𝐶𝛼 𝐶𝛽 ) 7. Brown RJS, Korringa J. Dependence of elastic properties of a porous rock on
compressibility of pore fluid. Geophysics. 1975;40(4):608–616.
𝐶𝛽 𝐶𝛾2 + 𝐶𝑠 [(1 − 𝜙)𝐶𝛾 − 𝐶𝛽 ]2 8. Carroll MM. Mechanical response of fluid-saturated porous materials. In: Rimrott
= , (B.10) FPJ, Tabarrok B. eds. 15th Int. Cong. Theoretical and Appl. Mech. Toronto;
(1 − 𝜙)3 𝐶𝛾2
1980:251–262.
𝐶𝛼 𝐶𝛾 [(1 − 𝜙)𝐶𝛾 − 𝐶𝛽 ] 9. Cornet FH, Fairhurst C. Influence of pore pressure on the deformation behavior
𝐶𝑠′ = of saturated rocks. In: Proceedings of the Third Congress of the International Society
(1 − 𝜙)(𝐶𝛾2 − 𝐶𝛼 𝐶𝛽 )
for Rock Mechanics, vol. 1. Washington, DC: National Academy of Sciences;
𝐶𝑠 𝐶𝛽 1974:638–644.
= 𝐶𝑠 − , (B.11) 10. Geertsma J. The effect of fluid pressure decline on volumetric changes of porous
(1 − 𝜙)𝐶𝛾
rocks. AIME Petrol Trans. 1957;210(12):331–340.
𝐶𝛼 𝐶𝛾 [𝜙(1 − 𝜙)𝐶𝛾 − 𝐶𝛽 ] 11. Nur A, Byerlee JD. Exact effective stress law for elastic deformation of rock with
𝐶𝑠′′ =
𝜙(1 − 𝜙)(𝐶𝛾2 − 𝐶𝛼 𝐶𝛽 ) fluids. J Geophys Res. 1971;76(26):6414–6419.
12. Rice JR, Cleary MP. Some basic stress diffusion solutions for fluid-saturated elastic
𝐶𝑠 𝐶𝛽 porous media with compressible constituents. Rev Geophys. 1976;14(2):227–241.
= 𝐶𝑠 − , (B.12)
𝜙(1 − 𝜙)𝐶𝛾 13. Gassmann F. Über die Elastizität poröser Medien (On elasticity of porous media).
Veirteljahrsschrift Nat Ges Zü. 1951;96:1–23.
where we can refer 𝐶𝑠 from (B.7). These equations can be inverted to 14. Davudov D, Moghanloo RG, Lan Y, Vice D. Investigation of shale pore compress-
give {𝐾𝛼 , 𝐾𝛽 , 𝐾𝛾 } as functions of {𝐾, 𝐾𝑠′ , 𝐾𝑠′′ }, as ibility impact on production with reservoir simulation. In: SPE Unconventional
Resources Conference. Calgary, Canada; 2017:16.
𝐾𝑠′ {(1 − 𝜙)2 𝐾𝑠′ 𝐾𝑠′′ − 𝐾[𝜙𝐾𝑠′ + (1 − 2𝜙)𝐾𝑠′′ ]} 15. Gutierrez M, Lewis RW, Masters I. Petroleum reservoir simulation coupling fluid
𝐾𝛼 = (B.13)
(1 − 𝜙)[𝐾𝑠′ (𝐾𝑠′′ − 𝜙𝐾𝑠′ ) − 𝐾𝐾𝑠′′ ] flow and geomechanics. SPE Reserv Eval Eng. 2001;4(03):164–172.
16. Mainguy M, Longuemare P. Coupling fluid flow and rock mechanics: Formulations
𝐾𝐾𝑠′ (𝐾𝑠′′ − 𝜙𝐾𝑠′ ) of the partial coupling between reservoir and geomechanical simulators. Oil Gas
𝐾𝛽 = (B.14)
(1 − 𝜙)3 [𝐾𝑠′ (𝐾𝑠′′ − 𝜙𝐾𝑠′ ) − 𝐾𝐾𝑠′′ ] Sci Technol. 2002;57(4):355–367.
17. Lopatnikov SL, Cheng AHD. Variational formulation of fluid infiltrated porous
𝜙𝐾𝐾𝑠′ (𝐾𝑠′′ − 𝐾𝑠′ )
𝐾𝛾 = (B.15) material in thermal and mechanical equilibrium. Mech Mater. 2002;34(11):685–
(1 − 𝜙)2 [𝐾𝑠′ (𝐾𝑠′′ − 𝜙𝐾𝑠′ ) − 𝐾𝐾𝑠′′ ] 704.
18. Wang HF. Theory of Linear Poroelasticity: With Applications to Geomechanics and
Appendix C. Proof of inequalities Hydrogeology. Princeton University Press; 2000.
19. Coussy O. Mechanics and Physics of Porous Solids. Chichester/Hoboken: Wiley;
2010.
To prove (134), 𝐶𝑝 > 𝐶𝑠′′ , we examine the two relations, (B.8)
20. Zimmerman RW, Somerton WH. Compressibility of porous rocks. J Geophys
and (B.12). Their relative magnitude is not clear due to the sign Res–Solid Earth Planets. 1986;91(B12):2765–2777.
indeterminacy of 𝐶𝛾 . So we seek to eliminate it. Equation (B.7) allows 21. Berge PA, Wang HF, Bonner BP. Pore pressure buildup coefficient in synthetic
us to solve for and natural sandstones. Int J Rock Mech Min Sci. 1993;30(7):1135–1141.
√ 22. Berryman JG, Milton GW. Exact results for generalized Gassmann’s equations in
𝐶𝑠 𝐶𝛼 𝐶𝛽 composite porous media with two constituents. Geophysics. 1991;56(12):1950–
𝐶𝛾 = ± (C.1)
𝐶𝑠 − 𝐶𝛼 1960.

18
A.H.D. Cheng International Journal of Rock Mechanics and Mining Sciences 142 (2021) 104754

23. Hart DJ, Wang HF. Laboratory measurements of a complete set of poroelastic 51. Fatt I. The Biot-willis elastic coefficients for a sandstone. J Appl Mech.
moduli for berea sandstone and Indiana limestone. J Geophys Res–Solid Earth. 1959;26:296–297.
1995;100(B9):17741–17751. 52. Van der Knaap W. Nonlinear behavior of elastic porous media. AIME Petrol Trans.
24. Ghabezloo S, Sulem J. Effect of the volume of the drainage system on the 1959;216:179–186.
measurement of undrained thermo-poro-elastic parameters. Int J Rock Mech Min 53. Tarokh A, Makhnenko RY. Remarks on the solid and bulk responses of fluid-filled
Sci. 2010;47(1):60–68. porous rock. Geophysics. 2019;84(4):WA83–WA95.
25. Tarokh A, Detournay E, Labuz J. Direct measurement of the unjacketed 54. Dropek RK, Johnson JN, Walsh JB. The influence of pore pressure on
pore modulus of porous solids. Proc R Soc Lond Ser A Math Phys Eng Sci. the mechanical properties of Kayenta sandstone. J Geophys Res: Solid Earth.
2018;474(2219):20180602. 1978;83(6):2817–2824.
26. Wissa AEZ. Pore pressure measurement in saturated stiff soils. ASCE J Soil Mech 55. Garg SK, Nur A. Effective stress laws for fluid-saturated porous rocks. J Geophys
Found Div. 1969;95(SM4):1063–1073. Res. 1973;78(26):5911–5921.
27. Berge PA. Pore compressibility in rocks. In: Thimus JF, Abousleiman Y, 56. Handin J, Hager RV, Friedman M, Feather JN. Experimental deformation of
Cheng AHD, Coussy O, Detournay E, eds. Poromchanics: A Tribute to Maurice A. sedimentary rock under confining pressure: Pore pressure effects. AAPG Bull.
Biot. Balkema; 1998:351–356. 1963;47:717–755.
28. Blöcher G, Reinsch T, Hassanzadegan A, Milsch H, Zimmermann G. Direct and 57. Wong TF, David C, Zhu WL. The transition from brittle faulting to cataclastic
indirect laboratory measurements of poroelastic properties of two consolidated flow in porous sandstones: Mechanical deformation. J Geophys Res-Solid Earth.
sandstones. Int J Rock Mech Min Sci. 2014;67:191–201. 1997;102(2):3009–3025.
29. Green DH, Wang HF. Fluid pressure response to undrained compression in 58. Vincké O, Boutéca MJ, Piau JM, Fourmaintraux D. Study of the effective stress at
saturated sedimentary rock. Geophysics. 1986;51(4):948–956. failure. In: Thimus JF, Abousleiman Y, Cheng AHD, Coussy O, Detournay E, eds.
30. Hart DJ, Wang HF. Variation of unjacketed pore compressibility using Gassmann’s Poromechanics–A Tribute to Maurice A. Biot. Balkema: 1998:635–640.
equation and an overdetermined set of volumetric poroelastic measurements. 59. Schmitt DR, Zoback MD. Laboratory tests of the effects of pore pressure on tensile
Geophysics. 2010;75(1):N9–N18. failure. In: Maury V, Fourmaintraux D, eds. Rock at Great Depth, ISRM International
31. Laurent J, Bouteca MJ, Sarda JP, Bary D. Pore-pressure influence in the Symposium. Balkema; 1989:883–889.
poroelastic behavior of rocks—experimental studies and results. SPE Form Eval. 60. Hashin Z, Shtrikman S. Note on a variational approach to the theory of composite
1993;8(2):117–122. elastic materials. J Franklin Inst—Eng Appl Math. 1961;271(4):336–341.
32. Hettema MHH, Raaen AM, Naumann M. Design and interpretation of laboratory 61. Mavko G, Mukerji T, Dvorkin J. The Rock Physics Handbook: Tools for Seismic
experiments to determine the pore volume compressibility of sandstone. In: 47th Analysis of Porous Media. 2nd ed. Cambridge University Press; 2009.
U.S. Rock Mechanics/Geomechanics Symposium. San Francisco; 2013:9. 62. Sevostianov I. Gassmann equation and replacement relations in micromechanics:
33. Makhnenko RY, Labuz JF. Elastic and inelastic deformation of fluid-saturated rock. A review. Internat J Engrg Sci. 2020;154:103344.
Phil Trans R Soc A. 2016;374(2078):20150422. 63. Gassmann F. Elastic waves through a packing of spheres. Geophysics.
34. Makhnenko RY, Tarokh A, Podladchikov YY. On the unjacketed moduli of 1951;16(4):673–685.
sedimentary rock. In: Vandamme M, Dangla P, Pereira J-M, Ghabezloo S, 64. Castagna JP, Batzle ML, Eastwood RL. Relationships between compressional-wave
eds. Proceedings of the Sixth Biot Conference on Poromechanics. Paris: ASCE; and shear-wave velocities in clastic silicate rocks. Geophysics. 1985;50(4):571–581.
2017:897–904. 65. White JE. Computed seismic speeds and attenuation in rocks with partial gas
35. Makhnenko RY, Podladchikov YY. Experimental poroviscoelasticity of common saturation. Geophysics. 1975;40(2):224–232.
sedimentary rocks. J Geophys Res: Solid Earth. 2018;123(9):7586–7603. 66. Kim YK, Kingsbury HB. Dynamic characterization of poroelastic materials. Exp
36. Pimienta L, Fortin J, Guéguen Y. New method for measuring compressibility and Mech. 1979;19(7):252–258.
poroelasticity coefficients in porous and permeable rocks. J Geophys Res: Solid 67. Adams LH, Williamson ED. On the compressibility of minerals and rocks at high
Earth. 2017;122(4):2670–2689. pressures. J Franklin Inst B. 1923;195(4):475–529.
37. Tarokh A. Poroelastic Response of Saturated Rock [Ph.D. Dissertation]. University 68. Zisman WA. Compressibility and anisotropy of rocks at and near the earth’s
of Minnesota; 2016. surface. Proc Natl Acad Sci. 1933;19(7):666–679.
38. Oliveira GLPd, Ceia MAR, Missagia RM, et al. Pore volume compressibilities of 69. Andersen MA, Jones FO. A comparison of hydrostatic-stress and uniaxial-
sandstones and carbonates from Helium porosimetry measurements. J Petrol Sci strain pore-volume compressibilities using nonlinear elastic theory. In: 26th U.S.
Eng. 2016;137:185–201. Symposium on Rock Mechanics; 1985:403–410.
39. Siddiqui S, Funk JJ. Use of X-ray CT to measure pore volume compressibility of 70. Warpinski NR, Teufel LW. Laboratory measurements of the effective-stress law
Shaybah carbonates. SPE Reserv Eval Eng. 2010;13(01):155–164. for carbonate rocks under deformation. Int J Rock Mech Min Sci Geomech Abstr.
40. Das V, Saxena N, Hofmann R. Compressibility predictions using digital thin-section 1993;30(7):1169–1172.
images of rocks. Comput Geosci. 2020;139:104482. 71. Ingraham MD, Bauer SJ, Issen KA, Dewers TA. Evolution of permeability and Biot
41. Berryman JG. Effective stress for transport properties of inhomogeneous porous coefficient at high mean stresses in high porosity sandstone. Int J Rock Mech Min
rock. J Geophys Res–Solid Earth. 1992;97(B12):17409–17424. Sci. 2017;96:1–10.
42. Kubo R. Thermodynamics: An Advanced Course with Problems and Solutions. 72. Fabre D, Gustkiewicz J. Poroelastic properties of limestones and sandstones under
North-Holland; 1968. hydrostatic conditions. Int J Rock Mech Min Sci. 1997;34(1):127–134.
43. Münster A, Kirkwood JG. Statistical Thermodynamics. Springer; 1969. 73. Mitra A, Govindarajan S, PaiAngle M, Aldin M. Measurement of grain compress-
44. Wall DC. Thermodynamics of Crystals. Dover; 1998. ibility of fine-grained rock. In: 50th U.S. Rock Mechanics/Geomechanics Symposium.
45. Lakes R, Wojciechowski KW. Negative compressibility, negative Poisson’s ratio, Houston, Texas; 2016:7.
and stability. Phys. Status Solidi b. 2008;245(3):545–551. 74. Fine RA, Millero FJ. Compressibility of water as a function of temperature and
46. Baughman RH, Stafström S, Cui C, Dantas SO. Materials with negative pressure. J Chem Phys. 1973;59(10):5529–5536.
compressibilities in one or more dimensions. Science. 1998;279(5356):1522–1524. 75. Cheng AHD. Material coefficients of anisotropic poroelasticity. Int J Rock Mech
47. Yu X, Zhou J, Liang H, Jiang Z, Wu L. Mechanical metamaterials associated Min Sci. 1997;34(2):199–205.
with stiffness, rigidity and compressibility: A brief review. Prog Mater Sci. 76. Berryman JG, Wang HF. Elastic wave propagation and attenuation in
2018;94:114–173. a double-porosity dual-permeability medium. Int J Rock Mech Min Sci.
48. Carmichael RS. Practical Handbook of Physical Properties of Rocks & Minerals. CRC 2000;37(1–2):63–78.
Press; 1989. 77. Pride SR, Berryman JG. Linear dynamics of double-porosity dual-permeability
49. Berge PA, Berryman JG. Realizability of negative pore compressibility in materials. I. Governing equations and acoustic attenuation. Phys Rev E.
poroelastic composites. J Appl Mech. 1995;62(4):1053–1062. 2003;68(3):036603.
50. Fatt I. Pore volume compressibilities of sandstone reservoir rocks. J Petrol Technol. 78. Pride SR, Berryman JG. Linear dynamics of double-porosity dual-permeability
1958;6:4–66. materials. II. Fluid transport equations. Phys Rev E. 2003;68(3):036604.
79. Mura T. Micromechanics of Defects in Solids. Dordrecht: Springer; 1987.
80. Nemat-Nasser S, Hori M. Micromechanics: Overall Properties of Heterogeneous
Materials. 2nd ed. North-Holland; 1999.

19

You might also like