WorksPhysics and Astronomy
Linearity of the co-moving velocity
Abstract
The co-moving velocity is a new variable in the description of immiscible two-phase flow in porous media. It is the saturation-weighted average over the derivatives of the seepage velocities of the two immiscible fluids with respect to saturation. Based on analysis of relative permeability data and computational modeling, it has been proposed that the co-moving velocity is linear when plotted against the derivative of the average seepage velocity with respect to the saturation, the flow derivative. I show here that it is enough to demand that the co-moving velocity is characterized by an additive parameter in addition to the flow derivative to be linear. This has profound consequences for relative permeability theory as it leads to a differential equation relating the two relative permeabilities describing the flow. I present this equation together with two solutions.
Introduction
The description of immiscible and incompressible two-phase flow in porous media at scales large enough so that the porous medium may be seen as a continuum, has been a challenge for a long time. In 1936 Wyckoff and Botset [1] proposed a generalization of the Darcy equation [2] by splitting it into two independent Darcy-type equations, one for each fluid, based on the idea that each fluid will experience, is the matrix and the other fluid together acting as an effective porous matrix. This reduces the effective permeability for each fluid by a factor, the relative permeability. It is then assumed that the two relative permeabilities depend only on the saturation. If I assume steady-state flow and that there are no continuum-scale gradients in the saturation, the theory is complete as described. This theory has a dominating position in all practical calculations.
The assumption that the relative permeabilities only depends on the saturation is the weak part of the theory. It is fairly accurate in a window of capillary numbers. However, outside this window, the pressure gradient enters and the theory loses its predictive power. For example Tallakstad et al. found that the average flow rate through a model porous medium
PoreLab, Department of Physics, Norwegian University of Science and Technology, NO–7491 Trondheim, Norway.
E-mail: [email protected] over a rather wide range of capillary numbers depends on the pressure drop across it to a power around two [3, 4]. In their experiment, they controlled the flow rates of the two immiscible fluids entering the porous medium, measuring the pressure drop and the saturation. But how to construct two generalized relative permeabilities from this information?
Hansen et al. [5, 6] recognized that the problem is that knowing the flow velocity of each fluid makes it possible to find the combined average flow velocity, but the opposite is not. The combined average flow velocity must be accompanied by a second independent velocity in order to calculate the two separate fluid velocities from the average flow velocity. This would be the co-moving velocity. Roy et al. [7] measured the co-moving velocity from relative permeability data using multiple sources and calculated it in a dynamic pore network model, finding that it would produce a straight line when plotted against the derivative of the average flow velocity with respect to the saturation — the flow derivative in the terminology of Hansen et al. [8] — necessitating only two parameters, one setting the scale and one determining the slope. They observed this linear behavior in the co-moving velocity plots generated from the relative permeability data, When the dynamic pore network model went outside the limits of relative permeability theory and into the non-linear flow regime first observed by Tallakstad et al. [3, 4], the linearity of the co-moving velocity was still in place, but the two parameters would now depended on the pressure gradient. Hansen et al. [8] constructed a statistical mechanics for immiscible two-phase flow in porous media under steady-state conditions that would explain why the derivative of the average flow velocity with respect to the saturation would be the natural variable for the co-moving velocity. Pedersen and Hansen [9] then presented an interpretation of the co-moving velocity by exploring the geometry of the space of variables describing the flow.
The last paper so far addressing the co-moving velocity is Alzubaidi et al. [10] who measured it for a wide range of contact angles using lattice Boltzmann computations on the reconstructed pore spaces of North Sea Sandstone and Bentheimer sandstone, but in the linear flow regime. In addition, they used experimental flow data on Bentheimer sandstone [11]. Their surprising conclusion was that the parameters characterizing the co-moving velocity are insensitive to the contact angle.
Three mysteries can therefore be identified concerning the co-moving velocity: 1. why is it linear in the flow derivative and 2. why is it still linear when the average flow rate becomes non-linear, and 3. why is the value of the parameters insensitive to the contact angle, at least in the linear regime.
The aim of this note is to address the first and second of these points. Is the linearity a result of some underlying mechanism that has to do with the detailed physics of immiscible two-phase flow, or is it rooted in the mathematical structure of the problem [9]? The robustness of the linearity (point two above) hints at the latter as the linear and the non-linear flow regimes are quite different. It is this viewpoint I will take in this note.
The linearity of the co-moving velocity when plotted against the flow derivative is a strong statement that has profound consequences. It is the aim of this paper to find a necessary and sufficient criterion for linearity.
I will claim that it is sufficient to demand that two parameters with dimensions of velocity are necessary to characterize the seepage velocity of each fluid, and , where the subscripts refer to the wetting and non-wetting fluids. This does not exclude additional dimensionless parameters. Fix the pressure gradient and do a wetting saturation sweep. When the wetting saturation is at a minimum, which I will assume to be for , the wetting seepage velocity defines the initial level (which may be zero). Likewise, when the non-wetting saturation is a minimum, which I will assume to be for , the non-wetting fluid will be flowing at its initial level (which may be zero). The second parameter I will call the range. It is for the wetting fluid the difference between seepage velocity at minus . I will call it . Likewise, I define for the non-wetting seepage velocity. The two seepage velocities may then be written
where and are dimensionless functions of the saturation. Note that the two parameters and are additive. In the flow regime where one may define the relative permeabilities and , I have
As I am working with the seepage velocities and not the Darcy velocities, I need to divide the relative permeabilities by the saturation.
I will then demonstrate that two parameters with dimensions of velocity being necessary to characterize the seepage velocity of each fluid lead to one parameter with dimension of velocity characterizing the co-moving velocity.
A necessary and sufficient criterion for the co-moving being linear in the flow derivative is then that the parameter with dimension of velocity is additive.
In the next section, I review the central concepts surrounding the co-moving velocity. I provide a concrete example in order to illustrate the concepts in a concrete way, namely that of Corey relative permeabilities with indices both equal to two. I then go on to analyse the average seepage velocity in Section 3. By using the Euler homogeneity theorem combined with the assumption that two parameters with dimensions of velocity in addition to non-dimensional parameters describe the average seepage velocity, I derive a general functional form for it. I continue with the example from Section 2 to illustrate the manipulations. In Section 4 I analyse the co-moving velocity using the method used for the average seepage velocity. Having two parameters with dimensions of velocity needed for the average seepage velocities, this leads to only one dimensional parameter being needed for the co-moving velocity. Assuming then that this parameter is an added constant, a mild assumption, the Euler homogeneity theorem forces the co-moving velocity to be linear. I also here use the previously introduced concrete example to illustrate the theory. In Section 5 I explore the consequences of this linearity for relative permeability theory. In particular, I present an equation combining the two relative permeabilities and provide two solutions of the equation. Finally, I conclude in Section 6.
Central concepts
I assume a porous medium sample having an area transversal to the flow direction and a porosity . The sample is uniform. I inject under steady state conditions a volumetric flow rate of the (more) wetting fluid and a volumetric flow rate of the (less) wetting fluid. I assume the fluids to be well mixed in the sample so that there are no sample-scale gradients in the saturation. The total average flow rate is
I now consider a cut through the porous sample orthogonal to the flow direction. It has an area . The sub-area of the transversal area that cuts through the pores is . The sub-area of the pore area that cuts through the wetting fluid is and the sub-area that cuts through the less wetting fluid is .
I can then define the seepage velocities. The seepage velocity of the wetting fluid is
and the less wetting fluid,
The combined average seepage velocity of the fluids is given by
where I in the last expression have use equations (5)–(7).
The co-moving velocity is defined as [5]
where the subscript means keeping the pressure gradient fixed. By taking the derivative of equation (8), I find that the co-moving velocity may be written
where I have defined .
Equations (8) and (9) constitute the mapping . The inverse mapping, , is
Roy et al. [7] and Alzubaidi et al. [10] observed the co-moving velocity being linear when plotted against ,
It is important for the derivation in the next section to note that is a control parameter in this expression in accordance with [8]. Hence, it does not depend on anything, but other functions will depend on it.
I refrain in the following from writing the pressure gradient explicitly as it does not enter the discussion explicitly before § 5.
An example
In order to illustrate these manipulations, I will consider a concrete example. Suppose
Hence, I have set and in equations (1) and (2). According to equations (3) and (4) this corresponds to using the Corey relative permeabilities with indices both equal to two. The parameters , , , and all have dimensions of velocity.
The average seepage velocity, equation (8), is then
where
and
have dimension of velocity. The parameter
is non-dimensional.
The flow derivative may be written
I invert this expression with respect to , finding
I express in terms of to find
Combining equation (9) with equations (14) and (15) gives the co-moving velocity
which becomes
when expressed in terms of the variable . With reference to equation (13), I find that
Parameter has units of velocity whereas is non-dimensional.
Dimensional analysis of the average velocity
My assumption is that the seepage velocity of the two fluids and each need two parameters with dimensions of velocity, level and range, see equations (1) and (2). Combining equation (8) with these two equations will generate two parameters with dimensions of velocity and that will be functions of the four parameters , , , and .
The average seepage velocity must be invariant under change of units. This means that if I divide the two parameters and , and the flow derivative by a velocity scale , the relation between them stays unchanged. In other words, they obey
This defines as an Euler homogeneous function of order one in . I take the derivative with respect to and set it equal to one. This gives
I set
and
so that equation (28) becomes
Both and are homogeneous functions of order zero in . The parameter sets the level. It is therefore an added constant. This means that
From equation (31) we have that
where I have used equation (32). This gives
for all choices of and . I conclude that neither nor depends on , and we may write as
where and both are functions of the dimensionless single variable .
I note that the dimensional analysis just performed does not generate any constraints on the two functions and apart from both being functions of the non-dimensional variable .
The example: average velocity
Here I calculate and , equations (29) and (30) for the example given in § 2.1. By using equation (22) I find
and
Combining these two expressions with equation (31) recovers equation (22). I note that even if I set , is still non-zero.
Dimensional analysis of the co-moving velocity
The three velocities, , and each contain two dimensional parameters of which one is an added constant. The definition of the co-moving velocity (9) contains the derivatives of and with respect to the saturation. Hence, the two parameters and , defined in equations (1) and (2) disappear when the co-moving velocity is expressed as a function of the saturation . Only the parameters and appear, i.e.,
I now express the co-moving velocity using the flow derivative rather than the saturation . As has the dimensions of velocity, there will now only be one parameter with this dimensionality, the parameter . Hence, we have
The co-moving velocity must be invariant with respect to a change of velocity scale . This was e.g., used in the analysis in Roy et al. [7], as it was the basis for using the relative permeability data. This means that obeys the scaling relation
Hence, is also an Euler homogeneous function of order one in the parameter and the variable . Taking the derivative with respect to and setting gives
I define
where and are homogeneous functions of degree zero. Hence,
I must have that
I now introduce a variable defined as
so that equation (45) is reduced to the equation
I note that if the parameter is additive, the function is
i.e., a constant. From equation (47) I then have that the function is also a constant. Combining this result with equation (33) gives the final result,
Hence, the co-moving velocity is linear in if and only if is an additive constant. The assumption that is an additive constant makes it possible to identify it. From equation (10) I have that
This identification is only possible if falls within the physical range of the parameters; in other words, has a minimum in the range . If there is no such minimum, it is still possible to define , but it will not have the simple physical interpretation given in equation (50).
The example: co-moving velocity
The co-moving velocity in the example is given in equation (23). In order to calculate according to equation (46), I set in the equation (21) giving the corresponding saturation . I use the equations for and , (14) and (15) together with the definition of , equation (19),
We see that and appear in the expression in spite of equation (38). This is so since switching from to as control variable reintroduces them.
Consequences for relative permeability theory
The generalized Darcy equations that forms the core of relative permeability theory are
where and are viscosities, the permeability and and are the relative permeabilities. The saturations appear in the denominators since and are seepage velocities.
Following Roy et al. [7], I define a velocity scale
which is the seepage velocity of the wetting fluid at saturation . I may then write the generalized Darcy equations as
where I have defined the viscosity ratio . The average velocity, equation (8), is then given by
I now combine equations (55) and (56), with the co-moving velocity, equation (10) using equation (13). I also make a change in notation by setting , as in Roy et al. [7]. The resulting equation is
A simplified version of this equation was used in Alzubaidi et al. [10] to determine the non-wetting relative permeability when the , and are known.
It is also possible to find analytic solutions to equation (58). I give an example in the following. Picchi and Battiato [12] suggested the semi-empirical form (note that their viscosity ratio is the inverse of the definition I use)
where and are the end point relative permeabilities. Setting
and
solves equation (58).
By solving for the saturation for which , using equations (59) and (60), I find that
According to equation (28) this is equal to , as found solving equation (58) directly, leading to (62).
The Corey empirical relative permeabilities
are, on the other hand, not solutions to equation (58), except when , they are solutions, but with . In that case I find
The example I gave in § 2.1 was based on the Corey relative permeabilities with exponents . The non-zero that was found, equation (25), came from the assumption that and were non-zero. The generalized Darcy equations (52) and (53) assumes and are both zero.
Discussion and Conclusion
I have in this paper shown that the co-moving velocity is of the form given by equation (13) given that two conditions are in place: 1. the average seepage velocity is characterized by two parameters with dimensions of velocity, allowing for additional dimensionless parameters, and 2. the necessity of an additive velocity scale in the co-moving velocity. Both assumptions are weak and does not hinge on any further restrictions on the constitutive law for , linear or non-linear. Equation (58), which is the combination of the relative permeability constitutive laws (52) and (53) with the linear equation for , (10), leads to a relation between the two relative permeabilities: knowing one, gives the other one.
The parameter , equation (28) may be given a straight-forward physical interpretation when falls within the physical range of the parameters. However, the second parameter, is not. It typically takes on a value around 0.6–0.8 and seems quite insensitive to the contact angle between the fluids and the matrix as shown by [10]. It is an open question why.
Declaration — The author declares no conflict of interest.
Acknowledgement — The author thanks R. T. Armstrong, C. F. Berg, S. Berg, E. G. Flekkøy, J. E. McClure, B. Hafskjold, H. Pedersen, T. Ramstad, S. Roy, and S. Sinha for countless discussions. This work was partly supported by the Research Council of Norway through its Centres of Excellence funding scheme, project number 262644.
References
- [1]R. Wyckoff, H. Botset, The flow of gas-liquid mixtures through unconsolidated sands, Physics 7(9), 325 (1936). DOI 10.1063/1.1745402DOI
- [2]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. 1 (Victor Dalmont, 1856)
- [3]K.T. Tallakstad, H.A. Knudsen, T. Ramstad, G. Løvøll, K.J. Måløy, R. Toussaint, E.G. Flekkøy, Steady-state two-phase flow in porous media: statistics and transport properties, Physical review letters 102(7), 074502 (2009). DOI 10.1103/PhysRevLett.102.074502DOI
- [4]K.T. Tallakstad, G. Løvøll, H.A. Knudsen, T. Ramstad, E.G. Flekkøy, K.J. Måløy, Steady-state, simultaneous two-phase flow in porous media: An experimental study, Physical Review E 80(3), 036308 (2009). DOI 10.1103/PhysRevE.80.036308DOI
- [5]A. Hansen, S. Sinha, D. Bedeaux, S. Kjelstrup, M.A. Gjennestad, M. Vassvik, Relation between seepage velocities in immiscible, incompressible two-phase flow in porous media, Transport in Porous Media 125, 565 (2018). DOI 10.1007/s11242-018-1139-6DOI
- [6]J. Feder, E.G. Flekkøy, A. Hansen, Physics of flow in porous media (Cambridge University Press, 2022)DOI
- [7]S. Roy, H. Pedersen, S. Sinha, A. Hansen, The co-moving velocity in immiscible two-phase flow in porous media, Transport in Porous Media 143(1), 69 (2022). DOI 10.1007/s11242-022-01783-7DOI
- [8]A. Hansen, E.G. Flekkøy, S. Sinha, P.A. Slotte, A statistical mechanics framework for immiscible and incompressible two-phase flow in porous media, Advances in Water Resources 171, 104336 (2023). DOI 10.1016/j.advwatres.2022.104336DOI
- [9]H. Pedersen, A. Hansen, Parameterizations of immiscible two-phase flow in porous media, Frontiers in Physics 11, 116 (2023). DOI 10.3389/fphy.2023.1127345DOI
- [10]F. Alzubaidi, J.E. McClure, H. Pedersen, A. Hansen, C.F. Berg, P. Mostaghimi, R.T. Armstrong, The impact of wettability on the co-moving velocity of two-fluid flow in porous media, arXiv preprint arXiv:2309.00362 (2023). DOI 10.48550/arXiv.2309.00362
- [11]S. Zou, R.T. Armstrong, J.Y. Arns, C.H. Arns, F. Hussain, Experimental and theoretical evidence for increased ganglion dynamics during fractional flow in mixed-wet porous media, Water Resources Research 54(5), 3277 (2018). DOI 10.1029/2017WR022433DOI
- [12]D. Picchi, I. Battiato, Relative permeability scaling from pore-scale flow regimes, Water Resources Research 55(4), 3215 (2019). DOI 10.1029/2018WR024251DOI