betaFeedback

WorksPhysics and Astronomy

Immiscible two-phase flow in porous media: a statistical mechanics approach

Abstract

The central problem in the physics of immiscible two-phase flow in porous media is to find a proper description of the flow at scales large enough so that the medium may be regarded as a continuum: the scale-up problem. So far, the only workable approach to the multiphase flow scale-up problem has been a set of phenomenological equations that have obvious weaknesses. Attempts at going beyond this relative permeability theory have so far not led to practical applications due to exploding complexity. Edwin T. Jaynes proposed in the fifties a generalization of statistical mechanics to non-thermal systems based on the information theoretical entropy of Shannon. This approach is used to construct a description of immiscible two-phase flow in porous media at the continuum scales, which is directly related to the physics at the pore scale, and at a level of complexity that is manageable. The approach leads to a thermodynamics-like formalism at the continuum scale with all the relations between variables that "normal" thermodynamics has to offer. New emergent variables appear. Among these, the co-moving velocity stands out as a key variable with implications for ordinary thermodynamics. We present here a short review of this approach.

Introduction

Porous media are as the name indicates, full of pores that are typically connected. More formally, one may define them as materials whose surface area is extensive in the volume. This means that surface effects never go away no matter the scale at which they are studied. Central to porous media research, both from a fundamental point of view, but also for applications, is the study of their transport properties. One may study their fluid transport properties at the pore scale. One is then faced with a problem of hydrodynamics in confined spaces. If there is more than one fluid present, all of them Newtonian, and they are immiscible, the problem becomes one of the viscous forces of the fluids competing with the capillary forces stemming from the fluid-fluid and fluid-matrix interfaces [1].

Considering the porous medium at a scale at which the pores are so small that it appears continuous, but not so large that it appears heterogeneous, fluid transport through it takes on a very different character. At this scale, known as the Darcy scale, concepts such as surface tension and interface dynamics no longer appear directly, but are worked into emergent properties that are observable on this scale.

At the Darcy scale, the simultaneous flow of immiscible fluids can be described as the flow of a single non-Newtonian fluid whose local rheological properties are controlled by the transport of two scalar parameters, the saturations SwS_w and SnS_n, which are linked by the relation

Sw+Sn=1.(1)S_w + S_n = 1 . \tag*{(1)}

The saturations are the relative volume densities of the fluids with respect to the pore volume density. The subscripts ww and nn refer to the more wetting (ww) of the two fluids and the less wetting (nn) with respect to the solid matrix. The viscosities of the two fluids μw\mu_w and μn\mu_n are parameters that survive moving from the pore scale to the Darcy scale. There is also a pressure pp and a fluid velocity v⃗p\vec{v}_p, where the subscript “p” stands for “pore”.

Modeling immisicble two-phase flow in porous media at the Darcy scale has a long history in the physics literature. The earliest paper we find is from 1934 by Muskat [2] where we find attempts at formulating constitutive equations for each of the two fluids. Two years later, a complete formulation was in place [3, 4] — an approach which today is known as relative permeability theory. It is a generalization of the fundamental work by Darcy in 1856 for single phase flow [5]. A few years later (1941), Leverett introduced the capillary pressure function, pcp_c [6]. This was necessary as the equations as they stood at that point, could not handle capillary suction (e.g. observed when dipping porous paper into water). With his introduction of the capillary pressure curve, the theory took on its final form. It splits the Darcy scale fluid velocity v⃗p\vec{v}_p into a Darcy scale velocity for each of the two immiscible fluids, v⃗w\vec{v}_w and v⃗n\vec{v}_n,

v⃗p=Swv⃗w+Snv⃗n.(2)\vec{v}_p = S_w\vec{v}_w + S_n\vec{v}_n . \tag*{(2)}

Hence, v⃗p\vec{v}_p is the saturation weighted average over the velocities of the two immiscible fluids. The theory also splits the pressure pp into pressures for each of the two fluids, pwp_w and pnp_n. The difference between the two pressures is the Leverett capillary pressure pcp_c, so that

pc=pn−pw.(3)p_c = p_n - p_w . \tag*{(3)}

The generalized Darcy equations are

v⃗w=−KkrwϕSwμw∇pw,(4)\vec{v}_w = -\frac{K k_{rw}}{\phi S_w \mu_w}\nabla p_w , \tag*{(4)}
v⃗n=−KkrnϕSnμn∇pn.(5)\vec{v}_n = -\frac{K k_{rn}}{\phi S_n \mu_n}\nabla p_n . \tag*{(5)}

Here KK is the permeability of the porous medium, a quantity measuring the conductivity of the porous medium with respect to the flow of a single fluid. The porosity ϕ\phi is the pore volume of an representative elementary volume divided by its volume. Both KK and ϕ\phi are properties of the solid matrix and not the fluids. The two parameters krwk_{rw} and krnk_{rn} are the relative permeabilities. They reduce the effective permeability of the fluids due to the presence of the other fluid.

The central assumption in relative permeability theory is that the three parameters pcp_c, krwk_{rw} and krnk_{rn} only depend on the saturation SwS_w: pc=pc(Sw)p_c = p_c(S_w), krw=krw(Sw)k_{rw} = k_{rw}(S_w) and krn=krn(Sw)k_{rn} = k_{rn}(S_w).

Adding a continuity equation for the saturation,

∂Sw∂t+∇⋅(Swv⃗w)=0(6)\frac{\partial S_w}{\partial t} + \nabla\cdot(S_w\vec{v}_w) = 0 \tag*{(6)}

completes the description and we have a closed set of equations once the three parameters pcp_c, krwk_{rw} and krnk_{rn} have been determined. Note here that the incompressibility of the two fluids is built into this equation. To see this, we note that we have a similar continuity equation for the non-wetting fluid,

∂Sn∂t+∇⋅(Snv⃗n)=0.(7)\frac{\partial S_n}{\partial t} + \nabla\cdot(S_n\vec{v}_n) = 0 . \tag*{(7)}

We then add the two continuity equations and use the definition of the averaged fluid velocity v⃗p\vec{v}_p in Equation (2) to find

∇⋅v⃗p=0(8)\nabla\cdot\vec{v}_p = 0 \tag*{(8)}

i.e., the incompressibility condition.

Equations (3) to (5) are phenomenological, and the strong assumption that the three central parameters depend only on the saturation makes them easy to criticize. Nevertheless, they are completely dominating in all practical applications.

It is the aim of this short review to discuss a recent attempt at constructing a Darcy scale theory for immiscible two-phase flow in porous media that is built on upscaling, i.e. connecting the pore-scale physics to the Darcy scale. An in-depth review of the upscaling problem may be found in Berg et al. [7]. The theory we review here, based on the Jaynes generalization of statistical mechanics [8, 9], have appeared in several papers, see [10, 11, 12] for some of the central ones.

Immiscible two-phase flow in porous media gives rise to a rich spectrum of behaviors. We review these in the next section. In particular, we present a Darcy-scale phase diagram in Figure 3 showing that there is a glassy and a non-glassy phase separated by a critical line [14]. § 3 introduces the generalization of statistical mechanics to the immiscible two-phase flow problem in porous media, which then forms the background for the phase diagram of the previous section.

Figure 3. Phase diagram for immiscible two-phase flow in porous media for a viscosity ratio M=1M = 1. The abscissa shows the non-wetting saturation SnS_n and the ordinate log⁡Ca\log Ca. The color shows the value of the Edwards-Anderson order parameter, revealing a glassy phase and a non-glassy phase. The transition line coincides with the onset of power law behavior, see Equation (9). The figure is adapted from [14].

Statistical mechanics in the context of the flow problem, scales up the pore-scale physics to the Darcy scale in the form of a thermodynamics-like formalism. This is done in § 4.

We end this small review by some concluding remarks in § 5.

Phase diagrams

Already in 1988, Lenormand et al. introduced a two-dimensional pore-scale phase diagram for shapes that occur when a less wetting fluid is injected into a porous medium saturated by a more wetting fluid with respect to the solid matrix, i.e. drainage [15], see Figure 1. One axis is the logarithm of the non-dimensional viscosity ratio M=μw/μnM = \mu_w/\mu_n. The orthogonal axis is the logarithm of the non-dimensional capillary number Ca=μnv/ϕσCa = \mu_n v/\phi\sigma. Here vv is a typical velocity. Under drainage, vv is the velocity of the injected fluid. Under steady-state flow conditions — to be discussed in a moment — we set v=vpv = v_p, the average velocity. Furthermore, σ\sigma is the interfacial tension between the two fluids. Depending on the values of the two parameters log⁡M\log M and log⁡Ca\log Ca, one would find viscous fingers, capillary finger or stable displacement patters occur. Much of the work done within the physics community at that time in this field was focused on the geometry of these patterns. This was the heyday of fractals [16].

Figure 1. A sketch of the phase diagram that Lenormand et al. [15] suggested for drainage processes, i.e., a less wetting fluid is injected into a porous medium saturated by a more wetting fluid with respect to the solid matrix. The axes are the logarithms of the viscosity ratio MM and the capillary number CaCa, both defined in the text. There are three behaviors: viscous fingering, capillary fingering and stable displacement. There is a crossover region between them marked by the dashed lines.

Drainage, and its counterpart imbibition (when a more wetting fluid is injected into a porous medium saturated with a less wetting fluid) are transients. This is a much more complex situation to deal with compared to steady-state flow. Steady-state flow means continuously injecting simultaneously into the porous medium the two immiscible fluids. One may think of a cylindrical porous sample, injecting the fluids at one end, and letting them leave at the other end by blocking off the side walls. Inside this porous cylinder, away from the two edges, there will be steady-state flow. By this we mean that the Darcy scale variables will fluctuate around well-defined values. In particular, there will be no Darcy-scale gradients in the saturation. It is this situation we will focus on in this review.

Pore scale phases

Avraam and Payatakes studied steady-state immiscible two-phase flow in model porous media [18] at the pore scale. They identified several types of flow patterns depending on the fluid parameters. They defined the following pore-scale phases:

  • Continuous path flow occurs when the flow is so slow that the interfaces between the fluids are held in place by capillary forces. The flow is then along channels that are not obstructed by interfaces.

  • Large ganglion flow. A ganglion is a non-wetting fluid cluster which in this case is much larger than the pore size. These clusters move, break up and coalesce. There is considerable pressure difference internally in these ganglions, causing their motion. This happens at higher capillary number than continuous path flow.

  • Small ganglion flow. At even higher capillary number, the ganglions break down in size to being only somewhat larger than the pores. In this case, the pressure difference internally in the ganglions is small and they move due to the motion of the surrounding fluids.

  • Drop traffic flow. When the capillary number is raised even further, the ganglions break down into droplets inside the pores, leading to a situation similar to traffic flow in cities. Hence, the name.

Darcy scale phases

We now turn to the Darcy-scale phase taxonomy of the same problem: immiscible steady-state two-phase flow in porous media. There are three parameters that define the flow at the Darcy scale, the capillary number CaCa, the viscosity ratio MM and the saturation SwS_w. Keeping MM and SwS_w fixed, Berg et al. [7] propose three phases that they name I, II and III in order of increasing CaCa. This was based on the constitutive equation relating flow velocity vpv_p to the pressure gradient ∣∇p∣\lvert\nabla p\rvert.

In 2009, Tallakstad et al. published two papers investigating experimentally steady-state immiscible two-phase flow in a model porous medium [19, 20]. At the capillary numbers this experiment was run at, they found a power law relation between the flow rate and pressure drop across the porous medium. In terms of velocity vs. pressure gradient we would have

vp=mII∣∇p∣β(9)v_p = m_{\mathrm{II}}\lvert\nabla p\rvert^\beta \tag*{(9)}

where mIIm_{\mathrm{II}} is the conductivity and β=1.85±0.27\beta= 1.85 \pm0.27. This observation was followed up in a number of papers by different groups, see e.g. [21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32].

The following picture has emerged. At very low capillary numbers, the fluid interfaces held in place by the capillary forces essentially do not move or move little (I). This corresponds to the continuous pathway picture of Avraam and Payatakes [18] and gives rise to a linear relationship between velocity and pressure gradient,

vp=mI∣∇p∣.(10)v_p = m_{\mathrm{I}} \lvert\nabla p \rvert. \tag*{(10)}

As the capillary number is increases, the interfaces start moving, thus successively opening more and more flow channels. This causes the power law dependency in Equation (9) (II). At high enough capillary number, all the interfaces that potentially may move, are moving and as a consequence the system reverts to a linear relationship between velocity and pressure gradient (III), and we find again a linear relationship between fluid velocity and pressure gradient,

vp=mIII∣∇p∣.(11)v_p = m_{\mathrm{III}} \lvert\nabla p \rvert. \tag*{(11)}

Based on the work of Armstrong et al. [33], Berg et al. [7] split phase I into a phase Ia and a phase Ib. They noted that before the power law behavior sets in, there is movement of the interfaces. However, fluid transport is still dominated by the channels appearing in continuous path flow. They named this phase Ib, and the phase where the interfaces are stuck, phase Ia. Gao et al. [29] described phase Ib in the following way: “We observe the onset of dynamics, defined as when there is some intermittency but the Darcy law remains valid.” We summarize the phases Ia through III in Figure 2.

Figure 2. We illustrate the three constitutive Equations (9) to (11) and the phases Ia, Ib, II and III here. Phase Ia is characterized by the fluid interfaces being held in place by capillary forces and the fluid transport occurs through open channels. In phase Ib, there is movement of the interfaces, but the flow is still dominated by flow through open channels. The relation between fluid velocity and pressure gradient is linear in phases Ia and Ib, see Equation (10). Regime II is characterized by a power law relation between fluid velocity and pressure gradient (9). This is due to an increasing number of interfaces moving as the pressure gradient is increased. Regime III is again linear, Equation (11), due to all interfaces that can move are now moving.

By mapping a dynamic pore network model [17] onto a spin model, and using Boltzmann machine learning [34], Sinha et al. [14] could identify a glass transition where the Edwards-Anderson order parameter is zero at the high-capillary number side and larger than one on the low side, see Figure 3. Furthermore, the spin-glass susceptibility peaks [35] at the transition. Remarkably, the transition coincides fully with the transition from linear to power law relation between velocity and pressure gradient Ib →\to II, see Figure 2. Phase Ib is therefore a glassy flow state, were there is motion of the interfaces.

Going further down in capillary number past the glassy flow state, the system freezes, identifying phase Ia. This is presumably a kinetic arrest [36] and not a phase transition in the ordinary sense. Its position will depend on the observation time of the system. Likewise, the transition between the power law regime (9) and the linear regime at high capillary numbers, i.e., where II →\to III in Figure 2, is also in the same way presumably not a phase transition in the ordinary sense. Roy et al. [37] have shown that the transition creeps towards the glassy transition when the system size increases.

Hence, only the glassy flow transition is a “proper” phase transition — or more precisely, a critical line as defined in thermodynamics.

The glassy state then explains the hysteresis and the large fluctuations seen in phase Ib. The origin of the hysteretic behavior has been a theme for discussion in porous media research, with the presence of first-order transitions or missing variables being alternative explanations, see [7].

Non-thermal statistical mechanics

We stated in the previous section that the glassy flow transition resembles a thermodynamic phase transition. The analysis by Sinha et al. [14] is performed under the assumption that the flow problem can be mapped onto an equilibrium system. But the flow problem is dissipative, and it is therefore not in equilibrium. The way out of this apparent problem is to note that we are not scaling up molecular behavior to form ordinary thermodynamics. Rather, we are scaling up pore-scale hydrodynamics to the Darcy scale. We are not dealing with ordinary molecular thermodynamics. Rather, we will adapt the Jaynes generalized statistical mechanics approach [8, 9] to immiscible two-phase flow in porous media [11]. We follow the presentation in Hansen and Sinha [12] fairly closely in the following.

We now sketch briefly the essence of the Jaynes generalized statistical mechanics. Central to his approach is to replace the Boltzmann statistical interpretation of thermodynamic entropy by the Shannon information entropy concept, being a quantitative measure of what is not known about a system. We consider a stochastic process producing NN possible outcomes, which we order from 1 to NN, assuming a probability pip_i for the iith outcome, which is xix_i. A quantitative measure of our ignorance about this process must be independent of how we group events together. If we know nothing about the process, its information entropy must be at a maximum, since our ignorance is maximal. This maximum is attained when all outcomes are equally probable, i.e., pi=1/Np_i = 1/N. There is only one function that can be constructed from pip_i with these properties, namely

SI=−∑i=1Npiln⁡pi,(12)S_I=-\sum_{i=1}^{N}p_i\ln p_i, \tag*{(12)}

which is then the information or Shannon entropy. What happens if we do know something about the system, e.g., the average

⟨x⟩=1N∑i=1Npixi?(13)\langle x\rangle=\frac{1}{N}\sum_{i=1}^{N}p_i x_i? \tag*{(13)}

Jaynes solution is to maximize the information entropy given the constraint (13). This determines the probabilities pip_i, which take the Gibbs form pi∝exp⁡(−λxi)p_i\propto\exp(-\lambda x_i).

Jaynes provided a set of circumstances for his generalized statistical mechanics to be applicable. Hansen et al. [11] demonstrates that these circumstances are met for steady-state immiscible two-phase flow in porous media, and hence proceeded to construct such a statistical mechanics. Even if molecular entropy is produced, the pore-scale Shannon entropy is not produced under steady-state flow. Hence, we may treat the system as being in equilibrium.

We consider a cylindrical pore sample as shown in Figure 4. Two immiscible fluids enter at the bottom and leave at the top. The side walls are closed off. We assume that the pore structure and chemical composition of the matrix to be statistically uniform throughout the cylinder so that away from the edges, steady-state flow is found. If we make transversal cuts through the sample, the structure of the porous medium and the distribution of fluids in the pore space to be statistically uniform in this region: Comparing one cut to another, we will not be able to determine which is closest to the inlet. Figure 4 shows one such cut. Within this cut, we pick out a Representative Elementary Area (REA), which is large enough to represent the statistics of the sample and small enough to have the flow fluctuate. There is a volumetric flow rate QpQ_p passing through the REA which may be split into a volumetric flow rate of the wetting fluid QwQ_w and the volumetric flow rate of the non-wetting fluid QnQ_n, i.e.,

Figure 4. We define a Representative Elementary Area (REA) within a cut orthogonal to the average flow direction through the cylinder-shaped porous medium sample. There is a wetting fluid flow rate QwQ_w, and a non-wetting fluid flow rate QnQ_n passing through the REA. The total flow rate is QpQ_p. The wetting fluid covers an area AwA_w of the REA and the non-wetting fluid an area AnA_n. The total pore area of the REA is ApA_p. Adapted from [12].

Qp=Qw+Qn.(14)Q_p = Q_w + Q_n . \tag*{(14)}

The area of the REA is AA. An area Ap<AA_p < A cuts through pore space.1 This area may be divided into areas cutting through the pore space filled with the more wetting fluid, AwA_w, and though the pore space filled with the less wetting fluid, AnA_n. We have that

Ap=Aw+An.(15)A_p = A_w + A_n . \tag*{(15)}

At a given time for a given REA, the fluid configuration, characterized by the velocity field v⃗p\vec{v}_p and where the fluids are, is XX. The quantities that describe the REA depend on XX, so that we have Qp(X)Q_p(X), Qw(X)Q_w(X), Qn(X)Q_n(X), Ap(X)A_p(X), Aw(X)A_w(X), and An(X)A_n(X), with Qn(X)=Qp(X)−Qw(X)Q_n(X) = Q_p(X) - Q_w(X) and An(X)=Ap(X)−Aw(X)A_n(X) = A_p(X) - A_w(X). Ap(X)A_p(X) does not dependent on the velocity field, nor the distribution of the two fluids, but only on the shape of pore space. The spatial statistical fluctuations of ApA_p are quenched, whereas the temporal and spatial fluctuations of the fluids are annealed. We denote p(X)p(X) the probability density to find fluid and pore space configuration XX. We define a configurational entropy,

S=−∫dX p(X) ln⁡p(X).(16)S = - \int\mathrm{d}X\, p(X)\,\ln p(X) . \tag*{(16)}

We note that this is a differential entropy as the probability distribution p(X)p(X) is continuous [13].

We assume that we know the averages (in time and position of the REA) of the variables just described, i.e.,

Qp=∫dX p(X) Qp(X),(17)Q_p = \int dX\, p(X)\, Q_p(X), \tag*{(17)}
Aw=∫dX p(X) Aw(X),(18)A_w = \int dX\, p(X)\, A_w(X), \tag*{(18)}
Ap=∫dX p(X) Ap(X),(19)A_p = \int dX\, p(X)\, A_p(X), \tag*{(19)}

are known.

Following the Jaynes maximum entropy principle, we maximize the entropy (16) given the constraints (17) to (19), finding that

p(X;λu,λw,λp)=1Z(λu,λw,λp)exp⁡[−λuQp(X)−λwAw(X)−λpAp(X)],(20)p(X;\lambda_u,\lambda_w,\lambda_p) = \frac{1}{Z(\lambda_u,\lambda_w,\lambda_p)} \exp[-\lambda_u Q_p(X)-\lambda_w A_w(X)-\lambda_p A_p(X)], \tag*{(20)}

where

Z(λu,λw,λp)=∫dX exp⁡[−λuQp(X)−λwAw(X)−λpAp(X)](21)Z(\lambda_u,\lambda_w,\lambda_p) = \int dX\, \exp[-\lambda_u Q_p(X)-\lambda_w A_w(X)-\lambda_p A_p(X)] \tag*{(21)}

is the partition function. Three new variables have appeared, λu\lambda_u, λw\lambda_w, and λp\lambda_p. We rewrite the partition function as

Z(λu,λw,λp)=exp⁡[−λuQz(λu,λw,λp)].(22)Z(\lambda_u,\lambda_w,\lambda_p) = \exp[-\lambda_u Q_z(\lambda_u,\lambda_w,\lambda_p)]. \tag*{(22)}

Ordinary statistical mechanics is an upscaling procedure for taking a molecular description of a system to the continuum scale, i.e., thermodynamics. The generalized statistical mechanics we have just sketched here does the same thing: It takes the flow problem from the pore scale where Qp(X)Q_p(X), Ap(X)A_p(X) and Aw(X)A_w(X) are defined to the Darcy scale, where a formalism similar to thermodynamics occurs.

Non-thermal thermodynamics at the Darcy scale

We now explore this non-thermal “thermodynamics” appearing as the end-result of the generalized statistical mechanics described in the previous section.

It is convenient to define new variables based upon the Lagrange multipliers λu\lambda_u, λw\lambda_w and λp\lambda_p,

θ=+1λu,(23)\theta= +\frac{1}{\lambda_u}, \tag*{(23)}
μ=−λwλu,(24)\mu= -\frac{\lambda_w}{\lambda_u}, \tag*{(24)}
π=−λpλu.(25)\pi= -\frac{\lambda_p}{\lambda_u}. \tag*{(25)}

By using the definitions of SS, AwA_w, and ApA_p, Equations (16), (18) and (19), combined with the partition function ZZ in Equation (21) and the flow rate QzQ_z defined in Equation (22), we find

S=−(∂Qz∂θ)μ,π,(26)S = -\left(\frac{\partial Q_z}{\partial\theta}\right)_{\mu,\pi}, \tag*{(26)}
Aw=−(∂Qz∂μ)θ,π,(27)A_w = -\left(\frac{\partial Q_z}{\partial\mu}\right)_{\theta,\pi}, \tag*{(27)}
Ap=−(∂Qz∂π)θ,μ.(28)A_p = -\left(\frac{\partial Q_z}{\partial\pi}\right)_{\theta,\mu}. \tag*{(28)}

The flow rate QpQ_p is given by

Qp(θ,Aw,Ap)=Qz(θ,μ,π)+μAw+πAp.(29)Q_p(\theta,A_w,A_p) = Q_z(\theta,\mu,\pi) + \mu A_w + \pi A_p. \tag*{(29)}

This equation may be recognized as a Legendre transform (assuming convexity of the involved functions) from the variables (θ,μ,π)(\theta,\mu,\pi) to the variables (S,Aw,Ap)(S,A_w,A_p).

Let us also define a flow rate Qu=Qu(θ,Aw,Ap)Q_u = Q_u(\theta,A_w,A_p) given by

Qu(S,Aw,Ap)=Qp(θ,Aw,Ap)−Sθ.(30)Q_u(S,A_w,A_p) = Q_p(\theta,A_w,A_p) - S\theta. \tag*{(30)}

We also define a flow rate Qg(θ,μ,Ap)Q_g(\theta,\mu,A_p) as

Qg(θ,μ,Ap)=Qz(θ,μ,π)+πAp.(31)Q_g(\theta,\mu,A_p) = Q_z(\theta,\mu,\pi) + \pi A_p. \tag*{(31)}

We see that θ\theta (Equation (23)) plays the same role as temperature in ordinary thermodynamics. Hansen et al. [11] named it agiture, which is short hand for “agitation temperature.” The variable μ\mu (Equation (24)) is somewhat reminiscent of chemical potential, was named the flow derivative. The variable π\pi, defined in Equation (25), they called the flow pressure. It has no equivalent in ordinary thermodynamics.

How should we understand the difference between the volumetric flow rates QuQ_u and QpQ_p? The flow rate QpQ_p is that we inject into our porous sample at one end. and which comes out on the other end. It is a conserved quantity throughout the sample as any cut through the sample will have the same QpQ_p passing through it. But locally there are fluctuations in the flow rate due to the competition between the capillary and viscous forces. These fluctuations average to zero by construction for QpQ_p. QpQ_p is based on sampling the flow configurations when the control parameters (θ,Aw,Ap)(\theta, A_w, A_p) are fixed. When our sampling assumes the control parameters (S,Aw,Ap)(S, A_w, A_p) fixed, these fluctuations no longer average to zero as we are averaging in a different way, making QuQ_u different from QpQ_p.

This is completely analogous to the difference between the internal energy (corresponding to QuQ_u) and the Helmholtz free energy (corresponding to QpQ_p) in ordinary thermodynamics. The internal energy contains a fluctuating part due to thermal agitation that is not recoverable in terms of work, whereas the free energy is the energy free to do work — i.e. the part in the internal energy due to fluctuations are subtracted off. In the flow problem, QuQ_u contains both the recoverable flow rate, i.e. the flow rate coming out of the sample and a part due to the fluctuations. QpQ_p is the free flow rate, i.e. the recoverable flow rate.

Returning to the REA shown in Figure 4, we see that there will be fluctuations in the entropy SS and the wetting area AwA_w — but not in the pore area ApA_p. The natural “free” flow rate in this case is then Qg(θ,μ,Ap)Q_g(\theta,\mu,A_p), which includes both of these types of fluctuations. If we now in addition average over the position of the REA, we need to include the fluctuations in ApA_p, making Qz(θ,μ,π)Q_z(\theta,\mu,\pi) the natural flow rate.

The variables QuQ_u, QzQ_z, QpQ_p, AwA_w, ApA_p, and SS are all extensive in the area of the REA. The emergent variables θ\theta, μ\mu, and π\pi are intensive. We rescale QpQ_p, AwA_w, and ApA_p by AA, defining

ϕ=ApA,(32)\phi= \frac{A_p}{A}, \tag*{(32)}
Sw=AwAp=AwAϕ,(33)S_w = \frac{A_w}{A_p} = \frac{A_w}{A\phi}, \tag*{(33)}
vp=QpAp=QpAϕ.(34)v_p = \frac{Q_p}{A_p} = \frac{Q_p}{A\phi}. \tag*{(34)}

In order to keep track of what happens to the variables in, e.g., Qp(θ,Aw,Ap,A)Q_p(\theta,A_w,A_p,A) (where we now write the area AA explicitly), we express the extensivity through the scaling relation

λQp(θ,Aw,Ap,A)=Qp(θ,λAw,λAp,λA).(35)\lambda Q_p(\theta,A_w,A_p,A) = Q_p(\theta,\lambda A_w,\lambda A_p,\lambda A). \tag*{(35)}

We set the scale factor λ=1/Ap\lambda= 1/A_p, finding the average fluid velocity to be

vp(θ,Sw,ϕ)=1ApQp(θ,Aw,Ap,A)=Qp(θ,Sw,1,1ϕ).(36)v_p(\theta,S_w,\phi)=\frac{1}{A_p}Q_p(\theta,A_w,A_p,A)=Q_p\left(\theta,S_w,1,\frac{1}{\phi}\right). \tag*{(36)}

The three emergent variables θ\theta, μ\mu, and π\pi, are the conjugate variables to the flow rate QuQ_u, the wetting area AwA_w, and the pore area ApA_p, respectively. We now go on to discuss further the meaning of these variables.

Agiture

We consider a porous cylinder as in Figure 4, but with the difference that it is divided into two halves A and B having different properties as shown in Figure 5. The difference between the two halves may e.g. be in chemical composition of the matrix, or it may consist of differently structured pore spaces. However, we assume the two halves each to be statistically homogeneous. We now assume that two immiscible and incompressible fluids are simultaneously injected into the cylinder under steady-state conditions.

Figure 5. A porous cylinder where we inject the two immiscible fluids at the bottom. The cylinder consists of two halves named A and B. The difference between the two halves may e.g. be in chemical composition of the matrix, or they may consist of differently structured pore spaces. However, we assume each of the two halves each to be statistically homogeneous. Adapted from [12].

The flow configurational entropy SS remains constant under such conditions. As shown by Hansen et al. [11], this means that the agiture in sectors A and B, θA\theta^A and θB\theta^B, are equal,

θA=θB.(37)\theta^A=\theta^B. \tag*{(37)}

But, what is the agiture? In the example just given, each of the two zones will each have a pressure gradient ∇pA\nabla p^A and a pressure gradient ∇pB\nabla p^B driving the flow. The average flow direction in both halves will be along the cylinder axis. Hence, we have that

∇pA=∇pB.(38)\nabla p^A=\nabla p^B. \tag*{(38)}

It is natural to assume the relation

θ=c∣∇p∣,(39)\theta=c\lvert\nabla p\rvert, \tag*{(39)}

when comparing Equations (37) and (38). Intuitively, this makes sense as the higher the pressure gradient, the higher the agiture.

We note that Qp=Qp(θ,Aw,Ap,A)Q_p=Q_p(\theta,A_w,A_p,A), defined in Equation (29), may be written as

Qp=Qp(p′,Aw,Ap,A)=Apvp(p′,Sw,ϕ),(40)Q_p=Q_p(p',A_w,A_p,A)=A_pv_p(p',S_w,\phi), \tag*{(40)}

where we have also used Equation (36). We have set p′=∣∇p∣p'=\lvert\nabla p\rvert. Hence, this is a constitutive equation relating average fluid velocity to the local pressure gradient, saturation, and porosity.

Hansen et al. [12] argue that also the flow derivative, Equation (24), is uniform under steady-state flow,

μA=μB.(41)\mu^{A} = \mu^{B}. \tag*{(41)}

No such relation exists, however, for the flow pressure π\pi, defined in Equation (25). As already pointed out, the disorder associated with ApA_p, the conjugate of π\pi, is quenched. The proper way to deal with this disorder is to follow the recipe used for glasses [35]. Sinha et al. [14] is a step in this direction.

Thermodynamic velocities and the co-moving velocity

Hansen et al. [10] defined two thermodynamic velocities

v^w=(∂Qp∂Aw)θ,An,(42)\hat{v}_w = \left(\frac{\partial Q_p}{\partial A_w}\right)_{\theta,A_n}, \tag*{(42)}
v^n=(∂Qp∂An)θ,Aw,(43)\hat{v}_n = \left(\frac{\partial Q_p}{\partial A_n}\right)_{\theta,A_w}, \tag*{(43)}

where the control variables are θ\theta, AwA_w, and AnA_n, defined in (15), making ApA_p a dependent variable. Changing the variables (θ,Aw,An)→(θ,Sw,Ap)(\theta,A_w,A_n) \to(\theta,S_w,A_p), these two equations may be written as

v^w=vp+Sn(∂vp∂Sw)θ,ϕ,(44)\hat{v}_w = v_p + S_n\left(\frac{\partial v_p}{\partial S_w}\right)_{\theta,\phi}, \tag*{(44)}
v^n=vp−Sw(∂vp∂Sw)θ,ϕ.(45)\hat{v}_n = v_p - S_w\left(\frac{\partial v_p}{\partial S_w}\right)_{\theta,\phi}. \tag*{(45)}

The flow derivative μ\mu, which is the conjugate of the wetting area AwA_w, and thereby also the saturation SwS_w, is given by

μ=(∂Qp∂Aw)θ,Ap=(∂vp∂Sw)θ,ϕ.(46)\mu= \left(\frac{\partial Q_p}{\partial A_w}\right)_{\theta,A_p} = \left(\frac{\partial v_p}{\partial S_w}\right)_{\theta,\phi}. \tag*{(46)}

Hence, Equation (45) may be recognized as a Legendre transformation substituting Sw→μS_w \to\mu,

v^n(θ,μ,ϕ)=vp(θ,Sw,ϕ)−Sw(θ,μ,ϕ)μ.(47)\hat{v}_n(\theta,\mu,\phi) = v_p(\theta,S_w,\phi) - S_w(\theta,\mu,\phi)\mu. \tag*{(47)}

In other words, the non-wetting thermodynamic velocity is the Legendre transformation of the average velocity with respect to the saturation.

If we combine Equations (29) and (31), we find

Qg(θ,μ,Ap)=Qp(θ,Aw,Ap)−Awμ.(48)Q_g(\theta,\mu,A_p) = Q_p(\theta,A_w,A_p) - A_w\mu. \tag*{(48)}

By comparing this expression to Equation (47), we see that there is a direct association between the “free” flow rate QgQ_g to the non-wetting thermodynamic velocity v^n\hat{v}_n, namely

Qg=Apv^n.(49)Q_g = A_p\hat{v}_n. \tag*{(49)}

The thermodynamic velocities defined in Equations (42) and (43) are not the average velocities for each of the two fluid species, which we define as

vw=QwAw,(50)v_w = \frac{Q_w}{A_w}, \tag*{(50)}
vn=QnAn.(51)v_n = \frac{Q_n}{A_n}. \tag*{(51)}

They are the velocities that appeared in generalized Darcy Equations (3) and (4), and they are the velocities that are measured in the laboratory, whereas the thermodynamic velocities are not.

We rewrite Equation (14) as

Qp=Qw+Qn=vwAw+vnAn.(52)Q_p = Q_w + Q_n = v_w A_w + v_n A_n . \tag*{(52)}

The saturations in terms of the variables AwA_w and AnA_n are given by

Sw=AwAw+An,(53)S_w = \frac{A_w}{A_w + A_n}, \tag*{(53)}
Sn=AnAw+An,(54)S_n = \frac{A_n}{A_w + A_n}, \tag*{(54)}

leading to Equation (1). We may then write Equation (52) as

vp=vwSw+vnSn,(55)v_p = v_w S_w + v_n S_n , \tag*{(55)}

where vpv_p, the average velocity, is defined in Equation (36).

We now turn to the thermodynamic velocities, Equations (42) and (43). The average flow rate QpQ_p obeys the Euler scaling relation,

λQp(p′,Aw,An)=Qp(p′,λAw,λAn).(56)\lambda Q_p(p', A_w, A_n) = Q_p(p', \lambda A_w, \lambda A_n) . \tag*{(56)}

We then have from the Euler theorem for homogeneous functions,

Qp=(∂Qp∂Aw)An,ϕ,p′Aw+(∂Qp∂An)Aw,ϕ,p′An=v^wAw+v^nAn.(57)Q_p = \left(\frac{\partial Q_p}{\partial A_w}\right)_{A_n,\phi,p'} A_w + \left(\frac{\partial Q_p}{\partial A_n}\right)_{A_w,\phi,p'} A_n = \hat{v}_w A_w + \hat{v}_n A_n . \tag*{(57)}

Dividing this equation by ApA_p gives

vp=v^wSw+v^nSn.(58)v_p = \hat{v}_w S_w + \hat{v}_n S_n . \tag*{(58)}

Comparing Equations (55) and (58) we find

vp=vwSw+vnSn=v^wSw+v^nSn.(59)v_p = v_w S_w + v_n S_n = \hat{v}_w S_w + \hat{v}_n S_n . \tag*{(59)}

This equation does not imply that vw=v^wv_w=\hat{v}_w and vn=v^nv_n=\hat{v}_n. Rather, the most general relation between the thermodynamic and seeping velocities are

vw=v^w−Snvm,(60)v_w=\hat{v}_w-S_n v_m , \tag*{(60)}
vn=v^n+Swvm,(61)v_n=\hat{v}_n+S_w v_m , \tag*{(61)}

where vmv_m is the co-moving velocity [7, 10, 38, 39, 40, 41, 42]. We may combine these two equations with Equations (44) and (45) to find

vw=vp+Sn[(∂vp∂Sw)ϕ,p′−vm],(62)v_w=v_p+S_n\left[\left(\frac{\partial v_p}{\partial S_w}\right)_{\phi,p'}-v_m\right] , \tag*{(62)}
vn=vp−Sw[(∂vp∂Sw)ϕ,p′−vm].(63)v_n=v_p-S_w\left[\left(\frac{\partial v_p}{\partial S_w}\right)_{\phi,p'}-v_m\right] . \tag*{(63)}

By taking the derivative of Equation (59) with respect to SwS_w, and using Equations (60) and (61), we find

vm=(∂vp∂Sw)ϕ,p′−vw+vn=Sw(∂vw∂Sw)ϕ,p′+Sn(∂vn∂Sw)ϕ,p′.(64)v_m=\left(\frac{\partial v_p}{\partial S_w}\right)_{\phi,p'}-v_w+v_n=S_w\left(\frac{\partial v_w}{\partial S_w}\right)_{\phi,p'}+S_n\left(\frac{\partial v_n}{\partial S_w}\right)_{\phi,p'} . \tag*{(64)}

We now return to Equation (47), which may now be expressed in terms of the velocity vnv_n, rather than the thermodynamic velocity v^n\hat{v}_n,

vn(θ,μ,ϕ)−Sw(θ,μ,ϕ)vm(θ,μ,ϕ)=vp(θ,Sw,ϕ)−Sw(θ,μ,ϕ)μ,(65)v_n(\theta,\mu,\phi)-S_w(\theta,\mu,\phi)v_m(\theta,\mu,\phi)=v_p(\theta,S_w,\phi)-S_w(\theta,\mu,\phi)\mu, \tag*{(65)}

signifying that (θ,μ,ϕ)(\theta,\mu,\phi) are the natural variables for the co-moving velocity. Measurements of the co-moving velocity suggest that it has the simple functional form [7, 39, 41, 42],

vm(θ,μ,ϕ)=a(θ,ϕ)+b(θ,ϕ)μ,(66)v_m(\theta,\mu,\phi)=a(\theta,\phi)+b(\theta,\phi)\mu, \tag*{(66)}

based on analysis of relative permeability data, on dynamic pore network modeling, and on lattice Boltzmann simulations on reconstructed sandstones. We show Figure 6 the co-moving velocity vs. μ\mu based on a large number of relative permeability data sets.

Measured relative permeability curves and corresponding co-moving velocity

Figure 6. Measured relative permeability curves, krw(Sw)k_{rw}(S_w) and krn(Sw)k_{rn}(S_w) plotted against SwS_w, and the corresponding co-moving velocity vM/v0v_M/v_0 plotted against (1/v0)dvp/dSw=μ/v0(1/v_0)\mathrm{d}v_p/\mathrm{d}S_w = \mu/v_0, where v0v_0 is a velocity scale. From [7].

The thermodynamic velocities (v^w,v^n)(\hat{v}_w,\hat{v}_n) can be found by knowing vpv_p alone; see Equations (44) and (45). The velocities, on the other hand, requires additional knowledge, i.e., vpv_p and vmv_m, which are given by Equations (62) and (63). They

provide a two-way mapping,

\begin{pmatrix} v_w \\ v_n \end{pmatrix} \leftrightarrow\begin{pmatrix} v_p \\ v_m \end{pmatrix}. \tag*{(67)}\end{pmatrix}

Is there an equivalent to the co-moving velocity in ordinary thermodynamics?

The answer to the question posed in the title of this section is “yes.” In fact, for any extensive variable, i.e., additive variable for two-component mixtures, there will be a corresponding co-“something” function.2

Olsen et al. [43] studied the intrinsic volumes of fluid mixtures using molecular dynamics. It is well known that mixing a volume V1V_1 of water with a volume V2V_2 of ethanol produces a total volume V<V1+V2V < V_1 + V_2. If there is N1N_1 mol of species 1 and N2N_2 mol of species 2, the total number of mol are N=N1+N2N = N_1 + N_2. The mol fractions are defined as x1=N1/Nx_1 = N_1/N and x2=N2/Nx_2 = N_2/N, so that

x1+x2=1,(68)x_1 + x_2 = 1 , \tag*{(68)}

and the partial molar volumes are

ν^1=(∂V∂N1)T,p,N2=ν+x2(∂ν∂x1)T,p,(69)\hat{\nu}_1 = \left(\frac{\partial V}{\partial N_1}\right)_{T,p,N_2} = \nu+ x_2\left(\frac{\partial\nu}{\partial x_1}\right)_{T,p}, \tag*{(69)}
ν^2=(∂V∂N2)T,p,N1=ν−x1(∂ν∂x1)T,p,(70)\hat{\nu}_2 = \left(\frac{\partial V}{\partial N_2}\right)_{T,p,N_1} = \nu- x_1\left(\frac{\partial\nu}{\partial x_1}\right)_{T,p}, \tag*{(70)}

where ν=V/N\nu= V/N, and we have that

ν=x1ν^1+x2ν^2.(71)\nu= x_1\hat{\nu}_1 + x_2\hat{\nu}_2 . \tag*{(71)}

We may also define intrinsic molar volumes

ν1=V1N,(72)\nu_1 = \frac{V_1}{N}, \tag*{(72)}
ν2=V2N.(73)\nu_2 = \frac{V_2}{N}. \tag*{(73)}

Here V1V_1 and V2V_2 are the volumes that each species actually occupies. It may e.g., be found by constructing a Voronoi tessellation of the volume VV. Each molecule is then assigned a volume, and ViV_i is then the sum of the volume around each molecule of type ii. We have that

ν=x1ν1+x2ν2.(74)\nu= x_1\nu_1 + x_2\nu_2 . \tag*{(74)}

Comparing Equations (71) and (74), we may now define a co-molar volume νm\nu_m by the relations

ν1=ν^1−x2νm,(75)\nu_1 = \hat{\nu}_1 - x_2\nu_m , \tag*{(75)}
ν2=ν^2+x1νm.(76)\nu_2 = \hat{\nu}_2 + x_1\nu_m . \tag*{(76)}

We show in Figure 7 the co-molar volume measured in molecular dynamics simulations for three mixtures interacting in different ways. We introduce the notation3

Co-molar volume as a function of partial molar volume difference

Figure 7. Co-molar volume (vmv_m) as a function of partial molar volume difference (w=dv/dx1w = d v/d x_1) for the mixtures. Mixture 1 is stable, whereas mixtures 2 and 3 are unstable in that they have a tendency to form homogeneous clusters. From [43].

w=(∂ν∂x1)T,p.(77)w = \left(\frac{\partial\nu}{\partial x_1}\right)_{T,p}. \tag*{(77)}

For the same reasons that μ\mu was the natural variable for νm\nu_m in the porous media problem, ww is the natural variable for νm\nu_m together with TT and pp. We may write

νm(T,p,w)=a(T,p)+b(T,p)w+∑k=2∞bk(T,p)wk.(78)\nu_m(T,p,w) = a(T,p) + b(T,p)w + \sum_{k=2}^{\infty} b_k(T,p)w^k . \tag*{(78)}

Analyzing the data in Figure 7, Olsen et al. [43] find that the three mixtures are close to having the parameters bkb_k close to zero — but for mixtures 2 and 3 not equal to zero. We note that mixture 1 is stable whereas mixtures 2 and 3 are not. It is remarkable that the behavior of the co-moving velocity shown in Figure 6 and the co-molar volume in Figure 7 are so similar. It is encouraging that a notion imported from the non-thermal thermodynamics of immiscible two-phase flow in porous media carries over to ordinary thermodynamics and not only the other way.

Discussion and conclusion

This short and incomplete review was written to rid the reader of the notion that porous media, and especially flow associated with such structures, is messy and best left to engineers. In fact, immiscible two-phase flow in porous media seems to be an excellent example of a non-thermal system that borrows the mathematical structure of thermodynamics. There are other examples of non-thermal thermodynamics, however, such as the non-thermal thermodynamics of powders developed by Edwards and coworkers, [44], see also [45].

As for developing the theory for steady-state immiscible two-phase flow in porous media further, there are many directions to follow. The glassy flow phase discussed in section 2.2 is one direction to pursue. This is the flow phase that is most relevant for practical applications, and its properties should be investigated. From a fundamental point of view, the system offers a clean problem where experimental investigations may closely pair with theoretical and computational investigations.

Another direction that we see as important is to move away from steady state flow. In terms of the thermodynamics-like formalism developed, the next step is to invoke non-equilibrium thermodynamics [46]. Rather than dividing the pressure pp into a wetting and non-wetting pressure, pwp_w and pnp_n as in Equation (3), the driving forces would then be gradients in the agiture (and thus of ∣∇p∣|\nabla p|) and the flow derivative μ\mu. There would be no need to introduce a capillary pressure function pcp_c. This is ongoing work.

As a last remark, we note that this short review is a first presentation of this material to the statistical physics community. Previous work along these lines have been published in journals typically not read by this community. Our hope is that we have managed to illustrate here that the problem of immiscible two-phase flow in porous media is well approachable as a problem within the realm of statistical mechanics.

Acknowledgements

We thank José Soares Andrade Jr., Saman Aryana, Dick Bedeaux, Carl Fredrik Berg, Steffen Berg, Humberto Carmona, Erika Eiser, Eirik G. Flekkøy, Daan Frenkel, Bjørn Hafskjold, Signe Kjelstrup, Anders Lervik, Marcel Moura, Håkon Pedersen, Thomas Ramstad, Subhadeep Roy and Per Arne Slotte for numerous discussions.

This work was partly supported by the Research Council of Norway through its Centers of Excellence funding scheme, project number 262644, and by the European Research Council ERC (Grant Agreement 101141323 AGIPORE).

References

  1. [1]J. Feder, E. G. Flekkøy and A. Hansen, Physics of Flow in Porous Media (Cambridge U. Press, Cambridge, 2022).DOI
  2. [2]M. Muskat, Two fluid systems in porous media. The encroachment of water into an oil sand, 1934, Physics 5, 250; DOI:10.1063/1.1745259.DOI
  3. [3]R. D. Wyckoff and H. G. Botset, The flow of gas-liquid mixtures through unconsolidated sands, 1936, Physics 7, 325; DOI:10.1063/1.1745402.DOI
  4. [4]M. Muskat and M. W. Meres, The flow of heterogeneous fluids through porous media, 1936, Physics 7, 346; DOI:10.1063/1.1745403.DOI
  5. [5]H. Darcy, Les fontaines publiques de la ville de Dijon: Exposition et application des principes à suivre et des formules à employer dans les questions de distribution d’eau, Vol. 2. (Victor Dalmont, Paris, 1856).
  6. [6]M. C. Leverett, Capillary behavior in porous solids, 1941, Transactions of the AIME, 142, 152; DOI:10.2118/941152-G.DOI
  7. [7]S. Berg, R. T. Armstrong, M. Rücker, A. Hansen, S. Kjelstrup and D. Bedeaux, From interface dynamics to Darcy scale description of multiphase flow in porous media, 2026, J. Coll. Interf. Sci. 351, 103791; DOI:10.1016/j.jcis.2026.103791.DOI
  8. [8]E. T. Jaynes, Information theory and statistical mechanics, 1957, Physical Review 106, 620; DOI:10.1103/PhysRev.106.620.DOI
  9. [9]E. T. Jaynes, Information theory and statistical mechanics II, 1957, Physical Review 108, 171; DOI:10.1103/PhysRev.108.171.DOI
  10. [10]A. Hansen, S. Sinha, D. Bedeaux, S. Kjelstrup, M. Aa. Gjennestad and M. Vassvik, Relations between seepage velocities in immiscible, incompressible two-phase flow in porous media, 2018, Transp. Porous Media 125, 565; DOI:10.1007/s11242-018-1139-6.DOI
  11. [11]A. Hansen, E. G. Flekkøy, S. Sinha and P. A. Slotte, A statistical mechanics framework for immiscible and incompressible two-phase flow in porous media, 2023, Adv. Water Res. 171, 104336; DOI:10.1016/j.advwatres.2022.104336.DOI
  12. [12]A. Hansen and S. Sinha, Thermodynamics-like formalism for immiscible and incompressible two-phase flow in porous media, 2025, Entropy 27, 121; DOI:10.3390/e27020121.DOI
  13. [13]S. Floerchinger and T. Haas, Thermodynamics from relative entropy, Phys. Rev. E 102, 052117 (2020); DOI:10.1103/PhysRevE.102.052117.DOI
  14. [14]S. Sinha, H. Carmona, J. S. Andrade Jr. and A. Hansen, Glassy phase transition in immiscible steady-state two-phase flow in porous media, arXiv:2603.08586 (2026); DOI:10.48550/arXiv.2603.08586.DOI
  15. [15]R. Lenormand, E. Touboul and C. Zarcone, Numerical models and experiments on immiscible displacements in porous media, 1988, J. Fluid Mech. 189, 165; DOI:10.1017/S0022112088000953.DOI
  16. [16]J. Feder, Fractals (Plenum, New York, 1988).DOI
  17. [17]S. Sinha, M. Aa. Gjennestad, M. Vassvik and A. Hansen, Fluid meniscus algorithms for dynamic pore-network modeling of immiscible two-phase flow in porous media, 2021, Front. Phys. 8, 548497; DOI:10.3389/fphy.2020.548497.DOI
  18. [18]D. G. Avraam and A. C. Payatakes, Flow regimes and relative permeabilities during steady-state two-phase flow in porous media, 2006, J. of Fluid Mech. 293, 207; DOI:10.1017/S0022112095001698.
  19. [19]K. T. Tallakstad, H. A. Knudsen, T. Ramstad, G. Løvoll, K. J. Måløy, R. Toussaint and E. G. Flekkøy, Steady-state two-phase flow in porous media: Statistics and transport properties, 2009, Phys. Rev. Lett. 102, 074502; DOI:10.1103/PhysRevLett.102.074502.DOI
  20. [20]K. T. Tallakstad, G. Løvoll, H. A. Knudsen, T. Ramstad, E. G. Flekkøy and K. J. Måløy, Steady-state simultaneous two-phase flow in porous media: an experimental study, 2009, Phys. Rev. E 80, 036308; DOI:10.1103/PhysRevE.80.036308.
  21. [21]E. M. Rassi, S. L. Codd and J. D. Seymour, Nuclear magnetic resonance characterization of the stationary dynamics of partially saturated media during steady-state infiltration flow, 2011, New J. Phys. 13, 015007; DOI:10.1088/1367-2630/13/1/015007.DOI
  22. [22]S. Sinha and A. Hansen, Effective rheology of immiscible two-phase flow in porous media, 2012, EPL 99, 44004; DOI:10.1209/0295-5075/99/44004.DOI
  23. [23]A. G. Yiotis, L. Talon, and D. Salin, Blob population dynamics during immiscible two-phase flows in reconstructed porous media, 2013, Phys. Rev. E 87, 033001; DOI:10.1103/PhysRevE.87.033001.DOI
  24. [24]O. Aursjø, M. Erpelding, K. T. Tallakstad, E. G. Flekkøy, A. Hansen and K. J. Måløy, Film flow dominated simultaneous flow of two viscous incompressible fluids through a porous medium, 2014, Front. Phys. 2, 63; DOI:10.3389/fphy.2014.00063.DOI
  25. [25]S. Sinha, A. T. Bender, M. Danczyk, K. Keepseagle, C. A. Prather, J. M. Bray, L. W. Thrane, J. D. Seymour, S. L. Codd and A. Hansen, Effective rheology of two-phase flow in three-dimensional porous media: Experiment and simulation, 2017, Transp. Porous Med. 119, 77; DOI:10.1007/s11242-017-0874-4.DOI
  26. [26]Y. Gao, Q. Lin, B. Bijeljic and M. J. Blunt, X-ray microtomography of intermittency in multiphase flow at steady state using a differential imaging method, 2017, Water Res. Res. 53, 10274; DOI:10.1002/2017WR021736.DOI
  27. [27]A. G. Yiotis, A. Dollari, M. E. Kainourgakis, D. Salin and L. Talon, Nonlinear Darcy flow dynamics during ganglia stranding and mobilization in heterogeneous porous domains, 2019, Phys. Rev. Fluids 4, 114302; DOI:10.1103/PhysRevFluids.4.114302.DOI
  28. [28]S. Roy, A. Hansen and S. Sinha, Effective rheology of two-phase flow in a capillary fiber bundle model, 2019, Front. Phys. 7, 92; DOI:10.3389/fphy.2019.00092.DOI
  29. [29]Y. Gao, Q. Lin, B. Bijeljic and M. J. Blunt, Pore-scale dynamics and the multiphase Darcy law, 2020, Phys. Rev. Fluids 5, 013801; DOI:10.1103/PhysRevFluids.5.013801.DOI
  30. [30]Y. Zhang, B. Bijeljic, Y. Gao, Q. Lin and M. J. Blunt, Quantification of non-linear multiphase flow in porous media, 2020, Geophys. Res. Lett. 48, e2020GL090477; DOI:10.1029/2020GL090477.DOI
  31. [31]H. Fyhn, S. Sinha, S. Roy and A. Hansen, Rheology of immiscible two-phase flow in mixed wet porous media: Dynamic pore network model and capillary fiber bundle model results, 2021, Trans. in Porous Med. 139, 491; DOI:10.1007/s11242-021-01674-3.DOI
  32. [32]Y. Zhang, B. Bijeljic and M. J. Blunt, Nonlinear multiphase flow in hydrophobic porous media, 2022, J. Fluid Mech. 934, R3; DOI:10.1017/jfm.2021.1148.DOI
  33. [33]R. T. Armstrong, J. E. McClure, M. A. Berrill, M. Rücker, S. Schlüter and S. Berg, Beyond Darcy’s law: the role of phase topology and ganglion dynamics for two-fluid flow, 2016, Phys. Rev. E, 94, 043113; DOI:10.1103/PhysRevE.94.043113.DOI
  34. [34]L. Meshulam and W. Bialek, Statistical mechanics for networks of real neurons, 2025, Rev. Mod. Phys. 97, 045002; DOI:10.1103/jcrn-3nrc.DOI
  35. [35]K. Binder and A. P. Young, Spin glasses: Experimental facts, theoretical concepts, and open questions, 1986, Rev. Mod. Phys. 58, 801; DOI:10.1103/RevModPhys.58.801.DOI
  36. [36]G. Foffi, E. Zaccarelli, F. Sciortino, P. Tartaglia and K. A. Dawson, Kinetic Arrest Originating in Competition Between Attractive Interaction and Packing Force, 2000, J. Stat. Phys. 100, 363; DOI:10.1023/A:1018660200206.DOI
  37. [37]S. Roy, S. Sinha and A. Hansen, Effective rheology in the continuum limit, 2024, Transp. Porous Med. 151, 1295; DOI:10.1007/s11242-024-02073-0.
  38. [38]S. Roy, S. Sinha and A. Hansen, Flow-area relations in immiscible two-phase flow in porous media, 2020, Front. Phys. 8, 4; DOI:10.3389/fphy.2020.00004.DOI
  39. [39]S. Roy, H. Pedersen, S. Sinha and A. Hansen, The co-moving velocity in immiscible two-phase flow in porous media, 2022, Transp. in Porous Media, 143, 69; DOI:10.1007/s11242-022-01783-7.DOI
  40. [40]H. Pedersen and A. Hansen, Parameterizations of immiscible two-phase flow in porous media, 2023, Front. Phys. 2023, 11, 1127345; DOI:10.3389/fphy.2023.1127345.DOI
  41. [41]A. Alzubaidi, J. E. McClure, H. Pedersen, A. Hansen, C. F. Berg, P. Mostaghimi and R. T. Armstrong, The impact of wettability on the co-moving velocity of two-fluid flow in porous media, 2024 Transp. Porous Med. 151, 1967; DOI:10.1007/s11242-024-02102-y.DOI
  42. [42]A. Hansen, Linearity of the co-moving velocity, 2024 Transp. Porous Med. 151, 2477; DOI:10.1007/s11242-024-02121-9.DOI
  43. [43]K. P. Olsen, B. Hafskjold, A. Lervik and A. Hansen, A new thermodynamic function for binary mixtures: The co-molar volume, 2025, J. Chem. Phys. 163, 184504; DOI:10.1063/5.0032106.
  44. [44]S. F. Edwards and R. B. S. Oakeshott, Theory of powders, 1989, Physica A 157, 1080; DOI:10.1016/0378-4371(89)90034-4.DOI
  45. [45]A. Baule, F. Morone, H. J. Herrmann and Hernán A. Makse, Edwards statistical mechanics for jammed granular matter, 2018, Rev. Mod. Phys. 90, 015006; DOI:10.1103/RevModPhys.90.015006.DOI
  46. [46]S. Kjelstrup, D. Bedeaux, E. Johannesen and J. Gross, Non-Equilibrium Thermodynamics for Engineers, Second Ed. (World Scientific, Singapore, 2017).

Paper details

vlnc:W_4up9p66w
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents