1 Introduction
Subsurface porous structures can be saturated with one or more fluids depending on the depth of burial and pressure–temperature conditions (Siddiqui et al.
2014; Ahmed
2018). Coupling between the flow of such fluids and mechanical deformation of the porous structure is critical in subsurface events such as hydrocarbon recovery, hydrology, surface subsidence, seismicity, nuclear waste disposal, carbon dioxide sequestration, etc. (Jha and Juanes
2014). One of the most widely used coupling schemes between fluid flow and geomechanics is the Biot theory of poroelasticity (Biot
1941). This theory initially dealt only with porous materials saturated with a single fluid with later extensions to account for chemical, thermal, and plasticity effects for single fluid-saturated porous materials (Heidug and Wong
1996; Ghassemi and Diek
2003; Roshan and Fahad
2012; Roshan and Oeser
2012; White et al.
2017). However, the challenge persisting is related to simple two-phase poroelasticity and its extension to other effects evident from a dearth of literature (Borja
2006; Chen
2013a; Song and Borja
2014; Choo et al.
2016). Generalization of the theory is important considering the usual existence of two-phase mixtures (gas/water, water/oil, gas/oil) in deeper rocks for geothermal energy extraction, CO
2 geological sequestration, and hydrocarbon production. Also, for shallow rocks close to the atmosphere, pore spaces are often partially saturated, i.e., filled by air and water where a general two-phase poroelasticity theory can be useful.
From a modeling standpoint, constitutive equations for poroelasticity are often used in numerical simulations of coupled flow and geomechanics problems. To develop such constitutive theories, the concept of mixture theories is used. Mixture theory fundamentally describes the interaction of the constituents of a mixture (Truesdell and Toupin
1960; Green and Naghdi
1965; Goodman and Cowin
1972; Passman
1977). In the jargon of continuum mechanics, a mixture exists when each point is co-occupied by two or more constituents (Katsube and Carroll
1987). Based on this assumption the conventional mixture theories exhibit challenges when applied to fluid-saturated porous media because when fluid is removed, a solid material is left instead of a dry porous solid skeleton (Katsube and Carroll
1987). A modification was, thus, introduced to solve this problem. In, what is now referred to as, the modified mixture theory, dry and fluid-saturated porous solid skeletons were replaced by an equivalent homogenous solid and an equivalent homogenous liquid (Katsube and Carroll
1987). The equivalent solid and liquid act as two unique continua occupying the same point. The approach is now commonly used (Coussy
2005) but is not always referred to as the modified mixture theory.
Heidug and Wong (
1996) were among some early researchers who derived, using the modified mixture theory, the single-phase chemo-poroelastic theory based on non-equilibrium thermodynamics and continuum mechanics for porous geomaterials. Their method was picked up and followed by many more constitutive modeling studies. In contrast to the classical mixture theory (Truesdell and Toupin
1960; Truesdell and Bowen
1984), the method of Heidug and Wong (
1996) does not distinguish explicitly between solid and fluid phases. This approach was different from the consolidation theories of Terzaghi (
1943) and Biot (
1941) where many complex constitutive models for real-world applications including unsaturated porous media were based on (Meroi et al.
1995; Sanavia et al.
2002; Li et al.
2006; Seetharam et al.
2007). The issue with consolidation theories is that extending the constitutive equations to incorporate complex chemical fluid–solid interactions becomes challenging (Laloui et al.
2003; Chen and Hicks
2013). This drawback is easily overcome by the modified mixture theory approach (Heidug and Wong
1996). For example, Chen and Hicks (
2010) extended the concept of Heidug and Wong (
1996) to develop constitutive equations for an unsaturated rock with chemical aspects added to the modified mixture theory (Chen and Hicks
2013).
Chemical effects represent an internal coupling phenomenon where the force transfer is not apparent. Although micro-level interactions could be captured with molecular dynamics modeling, such an approach is not feasible to extend the coupling effects to macro-scale continuum models (Chen and Hicks
2013). Here, by macro-scale, it is implied the scale at which the hydro-mechanical coupling occurs. The modified mixture theory (Heidug and Wong
1996) uses continuum mechanics along with non-equilibrium thermodynamics to establish a link between mechanical and chemical coupling by looking into energy dissipation mechanisms and the time evolution of state variables (Heidug and Wong
1996; Chen and Hicks
2010). This approach was used lately to extend the unsaturated poroelastic constitutive equations to incorporate chemical (hydrational swelling, osmosis) and thermal effects (Chen
2013b; Chen et al.
2013,
2016,
2018a;
b). The drawback of these studies and others (Zienkiewicz et al.
1990; Meroi et al.
1995) is that for unsaturated porous media, the gas phase (air) pressure was ignored, i.e., average pore pressure was used for an averaged fluid mixture and chemical-induced damage (microstructural deterioration of the rock fabric) was not considered which can be significant in clay-rich rocks such as shales. Thus, such an approach again may not apply to real-world engineering problems. The issue with assuming gas phase pressure to be atmospheric is that it implies the gas phase can flow freely without any resistance. In fluid flow simulations, the results obtained without such simplification have been found to differ considerably from the assumption of a passive gas phase (Morel‐Seytoux and Billica
1985). This can be significant where a gas phase (especially for gases other than air) is present which can have a considerable change in flow and water swelling properties of medium with changes in pressure and/or temperature. Moreover, neglection of chemical damage implies ignoring the permeability variation that occurs due to the evolution of micro-cracks brought about by osmosis or clay swelling. Therefore, a rigorous modeling framework where gas flow and chemical damage are considered is important especially for clay-rich rocks. Schrefler and Scotta (
2001) made progress with regards to considering an active gas phase where two continuity equations for the water and gas phases were used. Efficient numerical stability was portrayed in their study using the various fluid relationships between saturation, capillary pressure, and relative permeability. More recently, Cheng (
2020) developed phenomenological constitutive equations for unsaturated porous media and presented simplified observable relationships between fluid saturation (in the case of partial saturation) and laboratory-measurable constants (in the case of full saturation). For pure mechanical effects, consistency was found between the phenomenological micromechanical approach and the thermodynamics-based variational energy approach. The solution to the constitutive equations was not presented but the developed model seemed robust. Although such an approach is appealing towards efficient numerical and engineering applications, it is not easy to extend it to couple other multi-physics phenomena including thermal, or reaction mechanisms such as chemical damage (Cheng
2020).
Chemical damage is a critical multi-physics phenomenon that induces the microstructural deterioration of the water-active clay-rich rocks e.g., swelling induced damage in shales due to clay–water reactivity. Rocks, like any solid material, are composed of a unique microstructure. The mechanical behavior of the material such as strength, stiffness, etc. depends on this microstructure (Wei et al.
1989). Observations have shown that a change in microstructure results in the reduction of the mechanical properties (e.g., strength) and can eventually lead to cracking (failure) of the material (Panteleev et al.
2021). When this microstructural change (e.g., micro-cracks or micro-voids) is due to the clay–water reactivity in shales, we refer to this as the chemical damage in this study. The generated micro-cracks need not necessarily always cause measurable permanent deformation of the material but at the same time can lead to a significant loss in strength of the material even without the presence of any mechanical load (i.e., elasto-damage or brittle damage (Khan et al.
2004,
2007; Murakami
2012; Browning et al.
2017; Jia et al.
2020)). It is the accumulation, growth, and localization of this damage that causes the final material to fail. At the scale of the representative elementary volume (REV), a macroscopic continuous damage field is used to represent the microstructural defects. We refer the readers to the pioneering works of continuum damage mechanics for further details (Kachanov
1958; Lemaitre
1985,
1992; Lyakhovsky et al.
1997b; Murakami
2012).
From a thermodynamics point of view, the principle of local state asserts that a point in a continuum can be described by specific state variables. As mentioned, damage refers to the deterioration of the material’s microstructure. This microstructural deterioration at a point in the continuum is an internal state and can be represented by an internal variable called the damage variable, denoted by
\(D\) (Lemaitre
1992; Murakami
2012) which is a scalar assuming the microstructural discontinuities are isotropic. What inspires us to thermodynamically include chemical damage in the chemo-poroelastic constitutive theory is the specific focus of this study—water loss in shale matrix. Shale rock is a porous medium characterized by pores and a solid skeleton where the pores are initially filled with gas (with irreducible water). After hydraulic fracturing, there are main hydraulic fractures and several hydraulically induced micro-fractures where the water resides. During the well shut-in period, water imbibes from the main fractures and the induced micro-fractures into the gas-saturated matrix. Since the induced micro-fractures are hydraulically poorly connected to the main fracture (Dahi-Taleghani and Olson
2011; Nandlal and Weijermars
2019; Siddiqui et al.
2021), the imbibition modes from the main fracture and induced micro-fractures are different. From the induced micro-fractures, spontaneous imbibition is the main mechanism of water uptake by the shale matrix (Tokunaga
2020). The extent of this spontaneous imbibition in causing water loss in shales is the focus of the investigation in this study. As the water imbibes, two-phase flow occurs in the shale matrix and micro-cracks develop in the rock due to water-shale (clay minerals) interactions. These micro-cracks are a timely result of the chemically driven microstructural alteration of the matrix and should not be confused with the above-mentioned hydraulically induced micro-fractures generated during hydraulic fracturing.
It is clear that though advancements have been made to the thermodynamics-based rederivation of the Biot poroelasticity theory to complex porous media scenarios including two-phase flow conditions (e.g., Coussy
2005), attention is still warranted to develop a thermodynamically consistent theory explaining the chemical damage and its association to water loss problem in water-active shales (Roshan et al.
2015; Siddiqui et al.
2018,
2019). The development of such a theory is the focus of this study. Therefore, in this study, using non-equilibrium thermodynamics and continuum mechanics principles, based on the modified mixture theory, we develop a two-phase chemo-poroelastic, damage constitutive theory. The theory builds on previous constitutive theories to provide a new first-order estimate of shale matrix damage due to water uptake in variably saturated conditions. This can help explain the extent of water loss in the shale matrix and serve as a basis for the development of further advanced constitutive theories. The solution to the developed theory was numerically performed to specifically address the shale water loss problem in hydraulic fracturing operations. The details of the derivations are presented in the next section.
2 Theoretical Development Using Thermodynamics
When the first and second laws of thermodynamics are applied to any system, they involve the state variables that characterize the system’s internal energy. Hence, when the evolution of such a system is of interest, an appropriate framework is provided by these laws to develop the respective constitutive equations. In this study, we apply the thermodynamic laws to a porous continuum to account for different possible couplings in the development of the constitutive equations. Such a thermodynamics-based approach takes inspiration from the pioneering works of Biot (Biot
1955,
1977; Biot and Temple
1972) and Sherwood (Sherwood
1993,
1994,
1995), and the comprehensive works of Coussy (Coussy
2005,
2010) and Heidug and Wong (
1996). The extension to the basic poroelasticity theory is performed in this study by accounting for the second fluid phase, clay swelling, and chemical-induced microstructural deterioration, in the overall Helmholtz free energy
\(\left(F\right)\) of the porous solid. In other words, a Helmholtzian-based thermodynamically consistent constitutive theory for a two-phase chemo-poroelastic (hydro-chemo-mechanical) gas–water system is developed.
For this, we consider an isotropic macroscopic region of volume
\(V\) at isothermal conditions within the rock matrix having a boundary
\(R\) which is allowed to deform only infinitesimally. This is henceforth referred to as the representative elementary volume (REV). The REV contains both solid (grains) and voids. Such a continuum-scale representation of the rock microstructure inherently implies that
\(V\) is also significantly larger than the largest grain size of the REV. The REV boundary
\(\left(R\right)\) is considered closed to the flow of solid matter while fluid (and/or solute) inflow and outflow can occur. This further implies that
\(R\) is attached to the external solid (grains) part of
\(V\) such that the deformation of the solid matrix is governed by the evolution of
\(R\) in space (Heidug and Wong
1996). It is also considered that the water phase consists of only one type of solute where water and solute exhibit chemical potential
\({\mu }^{\text{w}}\) and
\({\mu }^{\text{c}}\), respectively (the associated electrolytic behavior is insignificant). It is also imposed that the gas phase is not adsorbed to the matrix but exists as a free phase.
With regards to the microstructural deterioration of the REV, we reiterate that the focus of the theory is reserved for chemical damage. The readers are advised that mechanical damage may become critical during cyclic stress perturbations which is not of concern in the shale water loss context. Furthermore, the water–clay interaction and consequent swelling are considered to be instantaneous to the time frame of the movement of fluids in the rock. This maintains the local physio-chemical equilibrium and the mechanical equilibrium (Heidug and Wong
1996). It is important to note that the theory is developed for strictly monotonic loading (fluid withdrawal from underground reservoirs), i.e., there is no cyclic loading considered nor are there any drying–wetting cycles. To incorporate cyclic stress (or effective stress) perturbations, thermodynamical consistency dictates new internal state variables related to plasticity and viscous effects will need to be added to the free energy potential of the rock skeleton (Arson
2020; Jacquey and Regenauer-Lieb
2021). Moreover, mechanical damage accumulation with cyclic stress perturbation becomes significant and an additional mechanical damage variable needs to be also added (Kuhl et al.
2000). This will give rise to complex damage coupling between chemical and mechanical damage variables that need extensive experimental characterization (Le Bellégo et al.
2003; Comi et al.
2014).
The state of any system can be represented by a fundamental function called thermodynamic potential. Starting from the first and second laws of thermodynamics, the expression for the differential form of the Helmholtz thermodynamic potential for the rock (REV) skeleton can be derived (see Appendix 1 for the detailed derivation). As discussed in the Introduction, the REV-scale (continuum-scale) constitutive modeling is based on the modified mixture theory wherein the porous medium is the superimposition of the skeleton continuum and the fluid continuum. However, the thermodynamics-based approach usually focuses on the skeleton deformation as it is the one that can be observed (e.g., Heidug and Wong
1996; Coussy
2005). Hence, similar to Heidug and Wong (
1996), and Coussy (
2005), we too focus on the skeleton system starting with its Helmholtzian thermodynamic potential:
$$W = W\left( {{\varvec{\varepsilon}}, D,p_{\text{w}} ,p_{\text{g}} ,\mu^{\beta } } \right).$$
(1)
The variables
\({\varvec{\varepsilon}}, D,p_{\text{w}} ,p_{\text{g}} ,\mu^{\beta }\) constitute the state variables characterizing the state of the skeleton system and they, respectively, represent strain, damage variable, water pore pressure, gas pore pressure, and chemical potential of different species (water, gas, solute). According to the postulate of local state, these state variables are macroscopic variables without referring to the microscopic level (Pokorska
2008). We consider the fundamentals of the energy-based approach to constitutive modeling to be well documented. Therefore, even though a major part of the derivation leading to Eq. (
1) is presented in Appendix 1, for a more rigorous step-by-step guide, the readers are referred to other sources (c.f. Appendix 1 in Heidug and Wong (
1996), or Chapter 3 in Coussy (
2005)).
Taking the time derivative of
\(W\) in terms of the state variables, we associate the state variables with their conjugate thermodynamic state variables
\(\left( {{\varvec{\sigma}}, Y_{\text{D}} , v_{\text{w}} , v_{\text{g}} , m_{\text{bound}}^{\beta } } \right)\):
$$\dot{W}\left( {{\varvec{\varepsilon}},D,p_{\text{w}} ,p_{\text{g}} ,\mu^{\beta } } \right) = tr\left( {\varvec{\sigma \dot{\varepsilon }}} \right) + Y_{\text{D}} \dot{D} - v_{\text{w}} \dot{p}_{\text{w}} - v_{\text{g}} \dot{p}_{\text{g}} - \sum \limits_{\beta } m_{\text{bound}}^{\beta } \dot{\mu }^{\beta } ,$$
(2)
where, from Eq. (
2), we can write the state equations as (Algazlan et al.
2022):
$$\begin{gathered} {\varvec{\sigma}} = \left( {\frac{\partial W}{{\partial {\varvec{\varepsilon}}}}} \right)_{{D,p_{\text{w}} ,p_{\text{g}} ,\mu^{\beta } }} ,\;Y_{\text{D}} = \left( {\frac{\partial W}{{\partial D}}} \right)_{{{\varvec{\varepsilon}},p_{\text{w}} ,p_{\text{g}} ,\mu^{\beta } }} ,\;v_{\text{w}} = - \left( {\frac{\partial W}{{\partial p_{\text{w}} }}} \right)_{{{\varvec{\varepsilon}},D,p_{\text{g}} ,\mu^{\beta } }} ,\;v_{\text{g}} = - \left( {\frac{\partial W}{{\partial p_{\text{g}} }}} \right)_{{{\varvec{\varepsilon}},D,p_{\text{w}} ,\mu^{\beta } }} {\text{and}} \hfill \\ m_{\text{bound}}^{\beta } = - \left( {\frac{\partial W}{{\partial \mu^{\beta } }}} \right)_{{{\varvec{\varepsilon}},D,p_{\text{w}} ,\mu^{\prime}\left( { \ne \mu^{k} } \right)}} \hfill \\ \end{gathered}$$
(3)
where
\({\varvec{\sigma}}\) is stress,
\(Y_{\text{D}}\) is damage conjugate variable,
\(v_{\text{w/g}}\) is variation in pore water/gas content, and
\(m_{\text{bound}}^{\beta }\) is the mass concentration of species
\(\beta\) in the skeleton system (e.g., in any occluded porosity). The bold font represents tensor or vector quantities as per the norms of continuum mechanics. Henceforth, indicial notations (without bold font) will be used, e.g.,
\(\sigma_{ij}\) is the second-order stress tensor and
\(\varepsilon_{ij}\) is the corresponding second-order strain tensor; both based on a Cartesian coordinate system
\(\left( {i,j = 1,2,3} \right)\). Hence, Eq. (
2) can be written as:
$$\dot{W} = \left( {\frac{\partial W}{{\partial \varepsilon_{ij} }}} \right)_{{D,p_{\text{w}} ,p_{\text{g}} ,\mu^{\beta } }} \dot{\varepsilon }_{ij} + \left( {\frac{\partial W}{{\partial D}}} \right)_{{\varepsilon_{ij} ,p_{\text{w}} ,p_{\text{g}} ,\mu^{\beta } }} \dot{D} + \left( {\frac{\partial W}{{\partial p_{\text{w}} }}} \right)_{{\varepsilon_{ij} ,D,p_{\text{g}} ,\mu^{\beta } }} \dot{p}_{\text{w}} + \left( {\frac{\partial W}{{\partial p_{\text{g}} }}} \right)_{{\varepsilon_{ij} ,D,p_{\text{w}} ,\mu^{\beta } }} \dot{p}_{\text{g}} + \sum \limits_{\beta } \left( {\frac{\partial W}{{\partial \mu^{\beta } }}} \right)_{{\varepsilon_{ij} ,D,p_{\text{w}} ,\mu^{\prime}\left( { \ne \mu^{k} } \right)}} \dot{\mu }^{\beta } .$$
(4)
At this point, the readers are advised that some materials undergo healing (of the accumulated damage) through different mechanical or chemical processes. A thermodynamically consistent theory of damage and healing mechanics was first proposed by Barbero et al. (Barbero et al.
2005; Arson
2020). For thermodynamically consistent healing mechanics, healing is considered a dissipative process that is characterized by an extra internal variable in addition to the damage variable
\(D\). We chose not to incorporate healing in our model as we develop the theory for fluids withdrawal from underground reservoirs where stress reversals to aid healing are rare. Healing mechanics is a work in progress and the readers are referred to the comprehensive works of the group of Arson et al. (Zhu and Arson
2016; Shen and Arson
2018; Arson
2020) for details.
Next, differentiating Eq. (
3) with respect to the time, the following constitutive equations for the evolution (time rates) of stress, variation content of water phase, variation content of gas phase, variation in interlayer (bound) water, and damage evolution are obtained:
$$\dot{\sigma }_{ij} = L_{ijkl} \dot{\varepsilon }_{kl} + V_{ij} \dot{Y}_{\text{D}} - M_{ij} \dot{p}_{\text{w}} - N_{ij} \dot{p}_{\text{g}} + S_{\text{w}} \sum \limits_{\beta } S_{ij}^{\beta } \dot{\mu }^{\beta } ,$$
(5)
$$\dot{\upsilon }_{\text{w}} = M_{ij} \dot{\varepsilon }_{ij} + W\dot{Y}_{\text{D}} + \left( {\frac{{\dot{p}_{\text{w}} }}{{M_{w/w} }} + \frac{{\dot{p}_{\text{g}} }}{{M_{w/g} }}} \right) + S_{\text{w}} \sum \limits_{\beta } B^{\beta } \dot{\mu }^{\beta } ,$$
(6)
$$\dot{\upsilon }_{\text{g}} = N_{ij} \dot{\varepsilon }_{ij} + X\dot{Y}_{\text{D}} + \left( {\frac{{\dot{p}_{\text{w}} }}{{M_{w/g} }} + \frac{{\dot{p}_{\text{g}} }}{{M_{g/g} }}} \right) + S_{\text{w}} \sum \limits_{\beta } \Pi^{\beta } \dot{\mu }^{\beta } ,$$
(7)
$$\dot{m}_{\text{bound}}^{\beta } = S_{ij}^{\beta } \dot{\varepsilon }_{ij} + Q\dot{Y}_{\text{D}} + B^{\beta } S_{\text{w}} \dot{p}_{\text{w}} + \Pi^{\beta } S_{\text{g}} \dot{p}_{\text{g}} + S_{\text{w}} \sum \limits_{\beta } Z^{\beta } \dot{\mu }^{\beta } ,$$
(8)
$$\dot{Y}_{\text{D}} = V_{ij} \dot{\varepsilon }_{ij} + Y\dot{D} + WS_{\text{w}} \dot{p}_{\text{w}} + XS_{\text{g}} \dot{p}_{\text{g}} + S_{\text{w}} \sum \limits_{\beta } Q^{\beta } \dot{\mu }^{\beta } ,$$
(9)
where the first terms on the right side of Eqs. (
6)–(
9) are double dot products. The set of constitutive equations (Eqs. (
5)–(
9)) also holds at equilibrium conditions. The local state postulate imposes that during any evolution of a system, it is characterized by the same state variables that characterize the equilibrium states (Coussy
2005). In other words, this implies that Eq. (
2) holds even without time dependency. The set of constitutive equations (Eqs. (
5)–(
9)) are already geometrically linear due to the infinitesimal strain and Cauchy stress assumptions. These equations can be further physically linearized with the understanding that the thermodynamic response coefficients,
\(L_{ijkl}\),
\(M_{ij}\),
\(N_{ij}\),
\(S_{ij}^{\beta }\),
\(B^{\beta }\),
\(\Pi^{\beta }\),
\(\frac{1}{{M_{w/w} }}\),
\(\frac{1}{{M_{w/g} }}\),
\(\frac{1}{{M_{g/g} }}\), and
\(Z^{\beta }\) are material-dependent constants (Heidug and Wong
1996). For isotropic materials,
\(M_{ij}\),
\(N_{ij}\), and
\(S_{ij}^{\beta }\) are diagonal tensors and can be written in terms of scalars
\(\alpha_{\text{w}}\),
\(\alpha_{\text{g}}\), and
\(\omega^{\beta }\), respectively:
$$M_{ij} = \alpha_{\text{w}} \delta_{ij} ,\;N_{ij} = \alpha_{\text{g}} \delta_{ij} \;{\text{and}}\;S_{ij}^{\beta } = \omega^{\beta } \delta_{ij} .$$
(10)
Here,
\(\alpha_{\text{w}}\) and
\(\alpha_{\text{g}}\) are the Biot coefficients for water and gas, respectively,
\(\omega^{\beta }\) is the swelling coefficient of each component, and
\(\delta_{ij}\) is the Kronecker delta. For binary solutions
\(\omega^{\beta } = \omega^{0} M^{\text{c}} {/}RT\) where
\(\omega^{0}\) is the common swelling coefficient representing both solute and solvent,
\(M^{\text{c}}\) is the molar mass of solute,
\(R\) is the universal gas constant, and
\(T\) is the temperature (Ghassemi and Diek
2003). It is further emphasized that
\(\sum \nolimits_{i = w,g} \alpha_{i} = \alpha ,\) where
\(\alpha\) is the Biot coefficient of the saturated porous medium (Jha and Juanes
2014). It is also noted that
\(\alpha_{i}\) are proportional to the respective saturation of each individual phase (
\(\alpha_{i} = S_{i} \alpha\)) where
\(\sum \nolimits_{i = w,g} S_{i} = 1\) (Jha and Juanes
2014) such that in the case of complete saturation with respect to any fluid we will have
\(\alpha_{i} = \alpha\) enabling a smooth transition from partially saturated to fully saturated conditions.
The thermodynamic response coefficient,
\(L_{ijkl}\) then possesses the meaning of the elastic stiffness matrix which is the fourth-order tensor written, for a damaged material, using the strain equivalence theory proposed by Lemaitre (
1992):
$$L_{ijkl} = \left( {1 - D} \right)G\left( {\delta_{ik} \delta_{jl} + \delta_{il} \delta_{jk} } \right) + \left( {1 - D} \right)\left( {K - \frac{2G}{3}} \right)\delta_{ij} \delta_{kl} ,$$
(11)
where
\({ }G\) is the rock’s shear modulus,
\(K\) its bulk modulus, and
\(D\) is the damage variable. According to the equivalent strain concept proposed by Lemaitre (
1992), the virgin material’s elastic constitutive equation can be used for the damaged material by multiplying the elasticity matrix with the damage variable. It is noted that the stiffness tensor formulation is chosen to be in terms of the shear modulus (
\(G\)) as it was observed that under isotropic stresses no microstructural alteration (i.e., damage) occurs (Siddiqui et al.
2020a). Using Eqs. (
10)–(
11), Eqs. (
5)–(
7) can be written as:
$$\dot{\sigma }_{ij} = \left( {1 - D} \right)\left( {K - \frac{2G}{3}} \right)\dot{\varepsilon }_{kk} \delta_{ij} + \left( {1 - D} \right)2G\dot{\varepsilon }_{ij} - \alpha_{\text{w}} \dot{p}_{\text{w}} \delta_{ij} - \alpha_{\text{g}} \dot{p}_{\text{g}} \delta_{ij} + S_{\text{w}} \sum \limits_{\beta } \omega^{\beta } \dot{\mu }^{\beta } \delta_{ij} ,$$
(12)
$$\dot{\upsilon }_{\text{w}} = \alpha_{\text{w}} \dot{\varepsilon }_{kk} + \left( {\frac{{\dot{p}_{\text{w}} }}{{M_{w/w} }} + \frac{{\dot{p}_{\text{g}} }}{{M_{w/g} }}} \right) + S_{\text{w}} \sum \limits_{\beta } B^{\beta } \dot{\mu }^{\beta } ,$$
(13)
$$\dot{\upsilon }_{\text{g}} = \alpha_{\text{g}} \dot{\varepsilon }_{kk} + \left( {\frac{{\dot{p}_{\text{w}} }}{{M_{w/g} }} + \frac{{\dot{p}_{\text{g}} }}{{M_{g/g} }}} \right) + S_{\text{w}} \sum \limits_{\beta } \Pi^{\beta } \dot{\mu }^{\beta } .$$
(14)
When the saturation of the wetting phase (water) increases to attain full saturation, physically there is no second phase present
\(\left( {p_{\text{g}} = 0} \right)\). In such a case, the constitutive theory transitions to a fully saturated flow of a single wetting phase and the constitutive equation for gas (Eq.
14) becomes redundant and all terms associated with the variable
\(p_{\text{g}}\) vanish in other constitutive equations. The attention of the readers is sought here in realizing that Eqs. (
12)–(
13) simplify to the classical Biot poroelasticity theory when the presence of gas phase
\(\left( {\dot{\upsilon }_{\text{g}} } \right)\) and chemical reaction
\(\left( {\dot{\mu }^{\beta } } \right)\) are neglected. Moreover, neglecting only the chemical term
\(\left( {\dot{\mu }^{\beta } } \right)\) in Eqs. (
12)–(
14), results in the same unsaturated poroelastic constitutive equations derived by Coussy (c.f. Section 7.3 in (Coussy
2010)). The approach to deriving thermodynamics-based constitutive equations in this study follows Heidug and Wong (
1996), which is slightly different from the approach of Coussy (
2005,
2010). Nonetheless, the fact that similar consistent constitutive equations emerge from both approaches provides confidence in the accuracy of the developed theory. It is noted that since gas does not cause any swelling,
\(\Pi^{\beta }\) will remain zero.
Furthermore, as was investigated in the authors’ previous study (Siddiqui et al.
2020a), the effect of water/gas pore pressures is negligible in damage evolution, i.e.,
\(W = X = 0\) and invoking the strain equivalence concept
\(V_{ij} = 0\). Hence, we have:
$$\dot{Y}_{\text{D}} = Y\dot{D}.$$
(15)
2.1 Field Equations
Before deriving the field equations, the chemical potential
\(\left( {\mu^{\beta } } \right)\) in the constitutive equations (Eqs. (
12)–(
14)) is replaced by postulating an ideal solution (dilute solutions) with a single solute where the solute chemical potential can be estimated as (Ghassemi and Diek
2002,
2003):
$$\dot{\mu }^{{\text{c}}} \approx \frac{{RT}}{{M^{{\text{c}}} }} \left( {{\text{ln}}\dot{C}^{{\text{c}}} } \right)$$
(16)
where
\(R\) is the gas constant,
\(T\) is the system temperature,
\(M^{\text{c}}\) is the solute molar mass, and
\(C^{\text{c}}\) is the mass fraction of the solute which is related to the water mass fraction
\(\left( {C^{\text{w}} = 1 - C^{\text{c}} } \right)\). The linear approximation of Eq. (
16) yields (Ghassemi and Diek
2002,
2003):
$$\dot{\mu }^{\text{c}} \approx \frac{RT}{{M^{\text{c}} }}\frac{{\dot{C}^{\text{c}} }}{{\overline{C}^{\text{c}} }},$$
(17)
where
\(\overline{C }^{\text{c}}\) is the average solute mass fraction over the range of interest. Rearranging the Gibbs–Duhem equation, the chemical potential of water is then expanded (Heidug and Wong
1996; Chen
2013a; Chen et al.
2018a):
$$\dot{\mu }^{\text{w}} \approx \frac{1}{{\overline{\rho }^{\text{w}} }}\left( {\dot{p}_{\text{w}} - \overline{\rho }^{\text{c}} \dot{\mu }^{\text{c}} } \right),$$
(18)
where
\(\overline{\rho }^{\text{w}}\) and
\(\overline{\rho }^{\text{c}}\) are the density of water and solute relative to the unit fluid volume, respectively.
2.1.1 Momentum Balance for Solid Phase
To arrive at the final field equation for the solid phase, we consider the mechanical equilibrium condition in the absence of the body, gravity, and dynamic forces (Coussy
2005):
$$\frac{{\partial \sigma_{ij} }}{{\partial x_{i} }} = 0,$$
(19)
and incorporating Eqs. (
17)–(
18) into Eq. (
12), yields the field equation based on the strain:
$$\left( {1 - D} \right)\left( {K - \frac{2G}{3}} \right)\dot{\varepsilon }_{kk} \delta_{ij} + \left( {1 - D} \right)2G\dot{\varepsilon }_{ij} - \alpha S_{\text{g}} \nabla \dot{p}_{\text{g}} \delta_{ij} - S_{\text{w}} \left( {\alpha - \frac{{\omega^{0} M^{\text{c}} }}{{RT\overline{\rho }^{\text{w}} }}} \right)\nabla \dot{p}_{\text{w}} \delta_{ij} + \frac{{S_{\text{w}} \omega^{0} }}{{\overline{C}^{\text{c}} }}\left( {1 - \frac{{\overline{\rho }^{\text{c}} }}{{\overline{\rho }^{\text{w}} }}} \right)\nabla \dot{C}^{\text{c}} \delta_{ij} = 0.$$
(20)
It is noted that the last term in Eq.
20 represents the time increment of the swelling stress component
\(\left( {\sigma_{\text{swell}} } \right)\). Equation (
20) can be written in terms of the displacements (Ghassemi and Diek
2003):
$$\left( {1 - D} \right)G\nabla^{2} \dot{u}_{i} + \left( {1 - D} \right)\frac{G}{1 - 2\upsilon }\nabla \left( {\nabla \cdot \dot{u}_{i} } \right) - \alpha S_{\text{g}} \nabla \dot{p}_{\text{g}} \delta_{ij} - S_{\text{w}} \left( {\alpha - \frac{{\omega^{0} M^{\text{c}} }}{{RT\overline{\rho }^{\text{w}} }}} \right)\nabla \dot{p}_{\text{w}} \delta_{ij} + \frac{{S_{\text{w}} \omega^{0} }}{{\overline{C}^{\text{c}} }}\left( {1 - \frac{{\overline{\rho }^{\text{c}} }}{{\overline{\rho }^{\text{w}} }}} \right)\nabla \dot{C}^{\text{c}} \delta_{ij} = 0,$$
(21)
where
\({\upupsilon }\) is the rock’s Poisson’s ratio and the solid displacements
\(\left( {u_{i} } \right)\) are related to the strain through:
$$\varepsilon_{ij} = \frac{1}{2}\left( {\frac{{\partial u_{i} }}{{\partial x_{j} }} + \frac{{\partial u_{j} }}{{\partial x_{i} }}} \right).$$
(22)
2.1.2 Transport Laws and Mass Balance for Water, Gas, and Solute
It is recalled that during the theoretical development (Appendix 1), it was stipulated that no interfacial exchange or mixing occurs between water–gas phases. The dissipation function (Eq. (94)) is then used to derive the phenomenological equations with the addition of the thermodynamic driving forces and corresponding fluxes using Onsager’s theory of irreversible processes (Onsager
1931). The ‘flow’ of damage is not included in the formulation of transport phenomenological equations. Such phenomenological equations relate the thermodynamic driving forces to their consequent fluxes (Heidug and Wong
1996; Ghassemi and Diek
2003)—in this case, the water, gas, and solute fluxes. Hence, the phenomenological equations including gas transport can be written as:
$${\varvec{u}}^{\text{w}} = - \frac{{L^{11} }}{{\overline{\rho }_{\text{w}} }}\nabla p_{\text{w}} - \frac{{L^{12} }}{{\overline{\rho }_{\text{w}} }}\nabla p_{\text{g}} - L^{13} \nabla \left( {\mu^{\text{c}} - \mu^{\text{w}} } \right),$$
(23)
$${\varvec{u}}^{g} = - \frac{{L^{21} }}{{\overline{\rho }_{\text{g}} }}\nabla p_{\text{w}} - \frac{{L^{22} }}{{\overline{\rho }_{\text{g}} }}\nabla p_{\text{g}} - L^{23} \nabla \left( {\mu^{\text{c}} - \mu^{\text{w}} } \right),$$
(24)
$${\varvec{J}}^{\text{c}} = - \frac{{L^{31} }}{{\overline{\rho }_{\text{w}} }}\nabla p_{\text{w}} - \frac{{L^{32} }}{{\overline{\rho }_{\text{w}} }}\nabla p_{\text{g}} - L^{33} \nabla \left( {\mu^{\text{c}} - \mu^{\text{w}} } \right).$$
(25)
Here, the thermodynamic driving forces are
\(\nabla p_{\text{w}}\),
\(\nabla p_{\text{g}}\), and
\(\nabla \left( {\mu^{\text{c}} - \mu^{\text{w}} } \right)\). It is commonly considered that the cross-permeability coefficients can be neglected
\(\left( {L^{12} = L^{21} = 0} \right)\) implying that the pressure gradient of a fluid phase can only generate a flux of that phase. The readers might be interested to read the view of others (Kalaydjian
1990; Mannseth
1991; Avraam and Payatakes
1999) who argue that the cross-permeability coefficients are nonnegligible even though their magnitudes are much lower than the diagonal permeability coefficients
\(\left( {L^{11} , L^{22} } \right)\). The significance of the cross-permeability coefficients has been widely debated (Avraam and Payatakes
1995). The cross-permeability coefficients represent the viscous coupling exerted between fluid phases which is a strong function of the size of the fluid/fluid interface and the ratio of viscosities of the two fluids. With the pore sizes commonly found in shales (Villamor Lora et al.
2016) and a wide contrast in water/gas viscosities, we ignore
\(L^{12} , L^{21}\) in this study. It is also obvious that there will be generally no solute present in the gas phase; hence
\(L^{23} = 0\). The effect of water and gas pressure on solute transport is also minimal allowing
\(L^{31} = L^{32} = 0\) (Roshan and Rahman
2013; Marbach and Bocquet
2019). It is noted that Onsager’s symmetry (Onsager
1931) is not inherently invoked here as is mostly done in the literature. Whether the symmetry holds or not is subject to specifically designed experiments which will not be discussed here for now. It is also noted that for dilute solutions which can be considered to be ideal, the following approximation is valid (Ghassemi and Diek
2003):
$$\nabla \left( {\mu^{\text{c}} - \mu^{\text{w}} } \right) \approx \frac{RT}{{\overline{C}^{\text{w}} \overline{C}^{\text{c}} M^{s} }}\nabla C^{\text{c}} ,$$
(26)
where
\(\overline{C}^{\text{c}}\) and
\(\overline{C}^{\text{w}}\) are the average mass fraction of solute and water, respectively. Therefore, the transport equations for water, gas, and solute phases are derived as:
$${\varvec{u}}^{\text{w}} = - \frac{{kk_{\text{rw}} }}{{\tau_{\text{w}} }}\left( {\nabla p_{\text{w}} + {\Re }\frac{RT}{{M^{S} \overline{C}^{\text{w}} \overline{C}^{\text{c}} }}\nabla C^{\text{c}} } \right),$$
(27)
$${\varvec{u}}^{g} = - \frac{{kk_{\text{rg}} }}{{\tau_{\text{g}} }}\nabla p_{\text{g}} ,$$
(28)
$${\varvec{J}}^{\text{c}} = - \left( {1 - {\Re }} \right)D_{\text{e}} \nabla C^{\text{c}} ,$$
(29)
where
\({\varvec{u}}^{\text{w}}\),
\({\varvec{u}}^{g}\), and
\({\varvec{J}}^{\text{c}}\) are the water, gas, and solute flux vectors, respectively,
\(D_{\text{e}}\) is the effective solute diffusion coefficient,
\(k\) is the absolute permeability of the porous media,
\(k_{\text{rw}}\) and
\(k_{\text{rg}}\) are the water and gas relative permeability, respectively. Also,
\({\Re }\) is the reflection coefficient or the membrane efficiency of the shale which varies from 0 to 1 and
\(\tau_{\text{w}}\) and
\(\tau_{\text{g}}\) are the dynamic viscosities of water and gas, respectively. The relative permeability to be used in the transport laws (Eqs. (
27)–(
28)) is estimated from the fractal analysis technique which has proven to be efficient in clay-rich rocks (Zhang et al.
2018b,
a; Siddiqui et al.
2020b):
$$k_{{{\text{rw}}}} = S^{{\prime}{\frac{{11 - D_{\text{f}} }}{{3 - D_{\text{f}} }}}} ,$$
(30)
$$k_{{{\text{rnw}}}} = (1 - S^{\prime})^{2} \left( {1 - S^{{\prime}{\frac{{5 - D_{\text{f}} }}{{3 - D_{\text{f}} }}}} } \right),$$
(31)
$$S^{\prime} = \frac{{S_{\text{w}} - S_{wi} }}{{1 - S_{wi} }},$$
(32)
where
\(S_{\text{w}}\) is the water saturation. The readers are advised that there are other familiar relative permeability models such as van Genuchten (
1980), Brooks and Corey (
1966), Purcell (
1949), etc. which may be used if applicable to the specific shale under consideration. A good comparative view of these and other models can be found in Li and Horne (
2006).
The time evolution of variation of compressible/incompressible fluid contents
\(\left( \zeta \right)\) is expressed as (Ghassemi and Diek
2003):
$$\dot{\zeta }^{i} = \frac{{\dot{m}_{i} }}{{\overline{\rho }_{i} }} \approx \dot{v}_{i} + \frac{{S_{i} \phi \dot{p}_{i} }}{{K_{f,i} }},$$
(33)
where
\(i = w,g\) and
\(K_{f,i = w,g}\) is the fluid bulk modulus for water and gas, respectively. For incompressible water, as is obvious, the second term on the right of Eq. (
33) becomes negligible due to its large fluid bulk modulus. Now, invoking the continuity equation (mass balance for water, gas, and solute) yields:
$$\left( {\rho_{\text{w}} \zeta^{\text{w}} } \right)^{.} + \nabla \cdot \left( {\rho_{\text{w}} {\varvec{u}}^{\text{w}} } \right) = 0,$$
(34)
$$\left( {\rho_{\text{g}} \zeta^{g} } \right)^{.} + \nabla \cdot \left( {\rho_{\text{g}} {\varvec{u}}^{g} } \right) = 0,$$
(35)
$$\rho_{\text{c}} S_{\text{w}} \dot{C}^{\text{c}} + S_{\text{w}} \left( {\rho_{c} {\varvec{u}}^{\text{w}} \nabla C^{\text{c}} + \nabla \cdot \left( {\rho_{c} {\varvec{J}}^{\text{c}} } \right)} \right) = 0,$$
(36)
where
\({\varvec{u}}^{\text{w}}\),
\({\varvec{u}}^{g}\) are the fluxes of pore water and gas, respectively. Assuming water is the wetting phase, the capillary pressure,
\(p_{c}\) is defined as:
$$p_{c} \left( {S_{\text{w}} } \right) = p_{\text{g}} - p_{\text{w}} ,$$
(37)
where
$$S_{\text{w}} + S_{\text{g}} = 1.$$
(38)
Substituting Eqs. (
13) and (
14) into Eq. (
33), and using
\(B^{\beta } = \frac{1}{K}\left( {\alpha - 1} \right)\omega^{\beta }\) proposed by Heidug and Wong (
1996) (
\(K\) here is the bulk modulus of porous medium), the time evolution of variations in water and gas contents are obtained as:
$$\dot{\zeta }^{\text{w}} = \alpha_{\text{w}} \dot{\varepsilon }_{kk} + \left( {\left[ {\frac{1}{{M_{w/w} }} + \frac{{S_{\text{w}} \phi }}{{K_{f,w} }} + \frac{{S_{\text{w}} \omega^{o} \left( {\alpha - 1} \right)M^{\text{c}} }}{{KRT\overline{\rho }^{\text{w}} }}} \right]\dot{p}_{\text{w}} + \frac{1}{{M_{w/g} }}\dot{p}_{\text{g}} } \right) + \frac{{S_{\text{w}} \omega^{o} \left( {\alpha - 1} \right)}}{{K\overline{C}^{\text{c}} }}\left( {1 - \frac{{\overline{\rho }^{S} }}{{\overline{\rho }^{\text{w}} }}} \right)\dot{C}^{\text{c}} ,$$
(39)
$$\dot{\zeta }^{g} = \alpha_{\text{g}} \dot{\varepsilon }_{kk} + \left( {\left[ {\frac{1}{{M_{w/g} }}} \right]\dot{p}_{\text{w}} + \left[ {\frac{{S_{\text{g}} \phi }}{{K_{f,g} }} + \frac{1}{{M_{g/g} }}} \right]\dot{p}_{\text{g}} } \right).$$
(40)
Substituting the water and gas fluxes (Eqs. (
27) and (
28)), and the time change of water and gas contents (Eqs. (
39)–(
40)), into the water and gas continuity equation (Eqs. (
34)–(
35)), yields the coupled water and gas two-phase field equations:
$$\alpha_{\text{w}} \dot{\varepsilon }_{kk} + \left( {\left[ {\frac{1}{{M_{w/w} }} + \frac{{S_{\text{w}} \phi }}{{K_{f,w} }} + \frac{{S_{\text{w}} \omega^{o} \left( {\alpha - 1} \right)M^{\text{c}} }}{{KRT\overline{\rho }^{\text{w}} }}} \right]\dot{p}_{\text{w}} + \frac{1}{{M_{w/g} }}\dot{p}_{\text{g}} } \right) + \frac{{S_{\text{w}} \omega^{o} \left( {\alpha - 1} \right)}}{{K\overline{C}^{\text{c}} }}\left( {1 - \frac{{\overline{\rho }^{\text{c}} }}{{\overline{\rho }^{\text{w}} }}} \right)\dot{C}^{\text{c}} = \nabla \cdot \left( {\frac{{kk_{\text{rw}} }}{{\tau_{\text{w}} }}\nabla p_{\text{w}} } \right) + {\Re }\frac{RT}{{M^{S} \overline{C}^{\text{w}} \overline{C}^{\text{c}} }}\nabla^{2} C^{\text{c}} ,$$
(41)
$$\alpha_{\text{g}} \dot{\varepsilon }_{kk} + \left( {\left[ {\frac{1}{{M_{w/g} }}} \right]\dot{p}_{\text{w}} + \left[ {\frac{{S_{\text{g}} \phi }}{{K_{f,g} }} + \frac{1}{{M_{g/g} }}} \right]\dot{p}_{\text{g}} } \right) = \nabla \cdot \left( {\frac{{kk_{\text{rg}} }}{{\tau_{\text{g}} }}\nabla p_{\text{g}} } \right).$$
(42)
Finally, the coupling coefficients,
\(M_{\alpha /\beta }\) may have various forms but for now can be defined as (Cheng
2020):
$$\frac{1}{{M_{g/g} }} = \frac{1}{{M_{g/w} }} = \frac{1}{{M_{w/w} }} = \frac{\phi }{{K_{C} }},$$
(43)
where
\(K_{c} = - \frac{{{\text{d}}S_{\text{w}} }}{{{\text{d}}p_{c} }}\) is the inverse capillary pressure derivative defined using the fractal concepts (Siddiqui et al.
2020b):
$$\frac{{{\text{d}}S_{\text{e}} }}{{{\text{d}}P_{{\text{c}}} }} = \left( {D_{{\text{f}}} - 3} \right)P_{{{\text{cmin}}}}^{{3 - D_{{\text{f}}} }} P_{{\text{c}}}^{{D_{{\text{f}}} - 4}} ,$$
(44)
$$K_{c} = - \frac{1}{{\left( {D_{\text{f}} - 3} \right)}}P_{{{\text{cmin}}}}^{{D_{\text{f}} - 3}} P_{c}^{{4 - D_{\text{f}} }} .$$
(45)
The solute transfer equation can be obtained by substituting Eq. (
29) into Eq. (
36). Keeping the convective term
\(\left( {\rho_{\text{c}} {\varvec{u}}^{\text{w}} \nabla C^{\text{c}} } \right)\) in Eq. (
36), we arrive at:
$$\rho_{\text{c}} S_{\text{w}} \dot{C}^{\text{c}} + S_{\text{w}} \left( {\rho_{c} {\varvec{u}}^{\text{w}} \nabla C^{\text{c}} + \nabla \cdot \left( { - \rho_{c} \left( {1 - {\Re }} \right)D_{\text{e}} \nabla C^{\text{c}} } \right)} \right) = 0.$$
(46)
2.1.3 Damage Evolution
Next, to close the set of field equations, we need to characterize the complementary damage force evolution law (Eq.
15). This requires the knowledge of the thermodynamic response coefficient
\(\left( Y \right)\). Hence, we alternatively use the entropy production (dissipation) due to microstructural deterioration to directly characterize damage evolution
\(\left( {\dot{D}} \right)\). For chemical damage due to the combined action of water–solute–clay interactions, we can formulate the evolution of
\(D\), based on the stress constitutive equation (Eq.
12), by defining stress
\(\left( \sigma \right)\) in
\(\sigma - \mu\) space instead of
\(\sigma - \varepsilon\) space:
$$\sigma_{ij} = S_{\text{w}} \omega^{\text{c}} \mu^{\text{c}} \delta_{ij} .$$
(47)
Similar to Eq. (
105) (see Appendix 2 for details), in
\(\sigma - \mu\) space, we have the following expression for the dissipated energy due to an infinitesimal increase of chemical damage at constant strain and fluid pressures:
$${\text{d}}w_{\text{e}} = S_{\text{w}} \omega^{\text{c}} \mu^{\text{c}} {\text{d}}\mu^{\text{c}} .$$
(48)
Integrating Eq. (
48) and invoking the definition of
\(\omega^{\text{c}}\) (
\(\omega^{\text{c}} = \omega^{0} M^{\text{c}} /RT\)), and the Gibbs–Duhem relationship (Eq.
17), the following expression for
\(\overline{Y}_{\text{D}}\) is obtained after substitution in Eq.
107 (see Appendix 2 for details on
\(\overline{Y}_{\text{D}}\) which is equated to
\(- Y_{\text{D}}\) to work with positive quantities):
$$\overline{Y}_{\text{D}} = \frac{{S_{\text{w}} \omega^{0} RT\left( {C^{\text{c}} } \right)^{2} }}{{2\left( {1 - D} \right)M^{\text{c}} \left( {\overline{C}^{\text{c}} } \right)^{2} }}.$$
(49)
Next, the entropy production (dissipation) due to microstructural deterioration is equated to the dissipation by chemical interactions (Amiri and Modarres
2014) to arrive at:
$$\bar{Y}_{D} \dot{D} = \mu ^{c} \dot{C}^{c} .$$
(50)
Solving for
\(\dot{D}\) using the Gibbs–Duhem relationship (Eq.
17) on the right-hand side for
\(\mu^{\text{c}}\) gives the following expression:
$$\dot{D} = \frac{{2RT\left( {1 - D} \right)\overline{C}^{\text{c}} \dot{C}^{\text{c}} }}{{S_{\text{w}} \omega^{0} C^{\text{c}} }},$$
(51)
where
\(RT\) converts
\(\omega^{0}\) to molar density units. Hence, Eqs.
20,
41,
42,
46, and
51 form a set of constitutive equations that describe the dynamic evolution of thermodynamic state variables of bulk displacement, gas, and water pressure, solute transfer, and damage in a porous system saturated with water and gas. The set of material constants and other inputs needed to solve the field equations were obtained through the authors’ previous experimental studies or the literature (more details in Sect.
3). The readers are advised here that the direct entropy-based chemical damage characterization relies on the time-dependent increase in continuum damage to weaken the material (Amiri and Modarres
2014). The actual force causing damage (
\(Y\) in Eq.
15) is not measured which requires in situ monitoring and advanced sensors (Amiri and Modarres
2014). Such an approach is considered practically sufficient for metals, though much research is warranted for other materials.
The damage (micro-cracks) initiation due to clay-fluid interactions is reflected in the absolute permeability on the continuum scale. Based on the experimental observations from the authors’ previous study (Siddiqui et al.
2020a); chemically induced micro-fractures only appear considerably in presence of anisotropic stresses. It is therefore assumed that the permeability alteration by these micro-fractures occurs only during deviatoric loading or shearing. This alteration is given by the commonly used Eq. (
52) below (Choinska et al.
2007; Zhang et al.
2007; Zhu and Wei
2011; Qu et al.
2019), which is considered only when anisotropic stresses exist:
$$k = k_{0} \left( {\frac{\phi }{{\phi_{0} }}} \right)^{3} e^{{\alpha_{\text{D}} *D}} ,$$
(52)
where
\(\alpha_{\text{D}}\) is a constant,
\(k_{0}\) and
\(\phi_{0}\) are the initial permeability and porosity, respectively. Hence, with the evolution of
\(D\), the absolute permeability of the matrix (REV) is updated according to Eq. (
52). The readers are advised that nonlinear auxiliary relationships like Eq.
52 and the like (e.g., Eqs.
30,
31,
44,
45) do not necessarily cause internal consistency problems in the context of engineering applications such as in this study and others (Laloui et al.
2003; Gens et al.
2006; Cheng
2020). It is fair to state here that in the absence of chemical damage
\(\left( {D = 0} \right)\), there will be no change in the permeability of the system. It is noted that the basis for Eq. (
52) are experimental observations of permeability and porosity increase or decrease with damage accumulation or healing, respectively (Tenthorey et al.
2003; Lyakhovsky and Hamiel
2007). Such an expression is physically motivated by the logarithmic response of the static friction coefficient with time (Lyakhovsky and Hamiel
2007). The damage evolution measured with acoustic emissions has also been shown to exponentially increase with time (Lyakhovsky et al.
1997a). The porosity
\(\left( \phi \right)\) after microstructural alteration is given by (Peng et al.
2015; Qu et al.
2019):
$$\left( {\frac{\phi }{{\phi_{0} }}} \right)^{3} = \left( {1 + \frac{{\Delta \varepsilon_{v} - \Delta \varepsilon_{{{\text{swell}}}} }}{{\phi_{0} }}} \right)^{3} ,$$
(53)
where the
\(\Delta\) symbol represents the difference between the current and initial value of a parameter,
\(\varepsilon_{v}\) is the total poroelastic volumetric strain, and
\(\varepsilon_{{{\text{swell}}}}\) is the swelling strain. It is assumed that the swelling strain is recoverable. The readers are advised that different forms of Eq.
53 can be found in the literature (Peng et al.
2015,
2018), but in this study, the matrix compressive strain component is included in
\(\Delta \varepsilon_{v}\) (c.f. (Aghighi et al.
2021)). The swelling strain to be used in Eq. (
53) is derived from the stress constitutive equation by dividing the swelling stress component by the bulk modulus (
\(K\)):
$$\varepsilon_{{{\text{swell}}}} = \frac{{S_{\text{w}} \omega^{0} \Delta C^{c} }}{{\overline{C}^{\text{c}} K}}.$$
(54)
The formulation of swelling strain, like the chemical damage variable, is a function of the coupled chemical reaction of water and solute, i.e.,
\(f(S_{\text{w}} ,C^{c} )\). In Eq. (
54), it is intuitive to say that the contribution of water saturation to swelling is dominant, i.e., even if the imbibing water has lower salinity compared to in situ water, due to the increasing
\(S_{\text{w}}\) with imbibition, there will still be swelling. Although both water and solute actions are responsible for clay swelling, it would be ideal to have their unique damage contributions through having separation functions, i.e.,
\(f_{1} (S_{\text{w}} ) + f_{2} \left( {C^{c} } \right)\). This, however, requires further experimental characterizations that are a focus of our future studies. The following section pursues the numerical solution of the developed constitutive equations using the finite element method based COMSOL Multiphysics platform.