JID:CRAS2B AID:3513 /SSU [m3G; v1.219; Prn:12/07/2017; 9:22] P.

1 (1-10)
C. R. Mecanique ••• (••••) •••–•••

Contents lists available at ScienceDirect

Comptes Rendus Mecanique

www.sciencedirect.com

A century of fluid mechanics: 1870–1970 / Un siècle de mécanique des fluides : 1870–1970

On the developments of Darcy’s law to include inertial

and slip effects
Didier Lasseux a,∗ , Francisco J. Valdés-Parada b
a
CNRS, Université de Bordeaux, I2M – UMR 5295, Esplanade des Arts-et-Métiers, 33405 Talence cedex, France
b
Universidad Autónoma Metropolitana-Iztapalapa, Departamento de Ingeniería de Procesos e Hidráulica, Av. San Rafael Atlixco 186,
09340 Mexico, D.F., Mexico

a r t i c l e i n f o a b s t r a c t

Article history: The empirical Darcy law describing flow in porous media, whose convincing theoretical
Received 4 December 2016 justification was proposed almost 130 years after its original publication in 1856, has
Accepted 26 March 2017 however been extended to account for particular flow conditions. This article reviews
Available online xxxx
historical developments aimed at including inertial and slip effects (respectively, when the
Reynolds and Knudsen numbers are not exceedingly small compared to unity). Despite the
Keywords:
Porous media early empirical extensions to include inertia and slip effects, it is striking to observe that
Inertial flow clear formal derivations of physical models to account for these effects were reported only
Slip flow recently.
Darcy’s law © 2017 Académie des sciences. Published by Elsevier Masson SAS. This is an open access
Forchheimer correction article under the CC BY-NC-ND license
Klinkenberg correction (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction

Flows in porous media are of interest in numerous applications ranging from hydrology, hydrocarbon recovery, gas and
nuclear waste storage, to drying of wood, transfer in food products or in living tissues to cite but a few. The main char-
acteristic of this particular domain of fluid mechanics lies in the (sometimes extreme) complexity of the geometry of the
channels where the flow takes place. Additionally, in many situations, this geometry is unknown in its very details and
may vary over more or less long distances characteristics of heterogeneities. Within this context, the physical description of
the flow in such materials may appear to be a tremendous challenge.1 This certainly explains why empiricism remained so
strong and lasted longer than in many other fields of fluid mechanics. In many situations, the interest is not in the details of
the flow within the pores but rather in the flux-to-force governing laws at length scales including a large numbers of pores,
although comprehensive analyses at the pore scale remain the corner stone in any progress towards the derivation of gov-
erning laws at larger scales. Clearly, active research in the description of transfer in porous materials was triggered by the
publication of Darcy’s law in the middle of the 19th century and the emergence of a key macroscopic physical characteristic
of a porous medium, namely the ability of a fluid to flow through it, i.e. its permeability.

*
Corresponding author.
E-mail addresses: [email protected] (D. Lasseux), [email protected] (F.J. Valdés-Parada).
1
The one-phase slow flow is probably one of the simplest mechanism one can think about and there is a lot of other tremendously more complex
physical processes in porous media of relevance from both scientific and industrial points of view, including multiphase flows, compressible flows, phase
change, deformable porous media, reactions in multicomponent systems, etc.

http://dx.doi.org/10.1016/j.crme.2017.06.005
1631-0721/© 2017 Académie des sciences. Published by Elsevier Masson SAS. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
JID:CRAS2B AID:3513 /SSU [m3G; v1.219; Prn:12/07/2017; 9:22] P.2 (1-10)
2 D. Lasseux, F.J. Valdés-Parada / C. R. Mecanique ••• (••••) •••–•••

1.1. Darcy’s law as an empirical basis

Ever since its empirical formulation in 1856, Darcy’s law [1] has been a hallmark in modeling momentum transport
through porous media. In this classical publication, there is a section entitled Détermination des lois d’écoulement de l’eau à
travers le sable, dedicated to the study of water flows through a bed of sand where the following relation is proposed (see
page 594 in [1]):
s
q = k (h + e ) (1)
e
where q is the volumetric flow rate, s is the cross section of the sand bed, e is the bed width, h is the pressure (or head)
difference between the surface and the base of the sand bed, and finally k was proposed as a coefficient that depends on
the permeability of the bed2 and on the properties of the fluid. For an excellent review about the origin of Darcy’s law, the
interested reader is referred to the work by Zerner [3]. The use of this simple relation requires that the only resistance to
the flow through the porous medium is due to viscous stresses induced by an isothermal, creeping (or laminar) steady flow
of a Newtonian fluid within an inert, rigid and homogeneous porous medium. However, the lack of a rigorous upscaling
technique prevented a formal derivation of this equation until the late 1960s, as it will be detailed later.
For a very long period of time – around fifty years – this law has been essentially verified experimentally in its global
form, but was not considered in a local differential form nor derived on a theoretical basis. One finds a differential expres-
sion in the analysis of the flows in aquifers by J. Boussinesq [4] as a result of an analogy with heat transfer in a continuum.
This work also reports an extension of the flow-rate-to-head-gradient relationship to non-homogeneous media. A formal
derivation of a 1D local expression of this law obtained from the solution to the Stokes equation for a flow parallel to a
regular array of infinite parallel cylinders (sufficiently apart from each other, i.e. for relatively large porosities) is due to
Emersleben in 1925 [5]. A derivation mainly based on dimensional analysis was later proposed by Muskat and Botset in
1931 [2] for a compressible flow in which the pressure difference is recognized to be replaced by the difference of the
squares of the pressures.
Substantial literature will then appear during the 1950s, in which many different approaches to demonstrate Darcy’s
law will be tested (see for instance [6] and references therein). Although these articles helped progressing into the under-
standing of the applicability of Darcy’s law, almost all of them relied on analogies, hypotheses or postulates that left them
incomplete. The first extension to three-dimensions and to non-isotropic materials was reported by Hall [7], who intro-
duced a permeability tensor, which is also based on some pre-requisites (see in particular Eq. (17) therein and the way the
permeability is identified).
Despite the lack of formal derivation of Darcy’s law, which can be expressed for a 1D flow in the x-direction as [8]

K s ∂ p β β
q=− (2)
μ ∂x
the meaning of the permeability and its relationship to the underlying pore structure focused closed attention. In the above
expression, K is the permeability having units of m2 and μ the fluid viscosity. In addition,  p β β is the intrinsically-averaged
pressure in the porous medium, defined as:

1
 p β β = p β dV (3)
Vβ
Vβ

Here, Vβ is the domain (of volume V β ) occupied by the fluid phase β within a representative averaging domain (or REV)
(see Fig. 1), and p β is the pore-scale pressure.
An early estimate of K was inspired by an analysis due to Blake in 1922 [9] of flow over packings of grains of different
shapes and a comparison to flows in capillary tubes that resulted in the following estimate

1 ε3
K= (4)
k0 S 02 (1 − ε )2
where ε is the porous medium porosity, S 0 denotes the specific surface of the particles and it is defined as the ratio of the
area of the particle to its volume. The coefficient S 0 , related to the effective particle diameter, d p , was identified from an
analogy with spherical particles by

6
S0 = (5)
dp

2
In [1], H. Darcy indicated that k “depends on the permeability of the sand layer”. In fact, k is the hydraulic conductivity, having the unit of a length per
unit time. The intrinsic permeability as a physical quantity, denoted by K (or K in tensorial form) in the present article, appeared later in the literature. It
seems that M. Muskat (see for instance [2]) was the first who used this coefficient.
JID:CRAS2B AID:3513 /SSU [m3G; v1.219; Prn:12/07/2017; 9:22] P.3 (1-10)
D. Lasseux, F.J. Valdés-Parada / C. R. Mecanique ••• (••••) •••–••• 3

Fig. 1. Sketch of a porous medium including an averaging domain and the phases involved.

Finally, k0 is an adjustable coefficient that was later known as the Kozeny [10] coefficient. Carman [11] suggested taking
k0 = 5, so that equation (4) can now be expressed as

d2p ε 3
K= (6)
180(1 − ε )2
which is usually known as the Kozeny–Carman equation and which is often used, sometimes abusively, to predict the value
of permeability. Certainly, a precise correlation for any type of pore structure is out of reach.

1.2. Theoretical foundation of Darcy’s law

The lack of convincing formal derivations of Darcy’s law that lasted over more than a century is obviously related to
the lack of clear upscaling methods allowing one to obtain the macroscopic conservation equations form their microscopic
(i.e. pore-scale) counterparts, as pointed out by Zerner [3]. These methods emerged in the 1970s and a first attempt was
proposed by S. Whitaker in 1966 [12], who clearly obtained a generalization of Darcy’s law in the following vectorial form:
K  
vβ  = − · ∇ p β β − ρ g (7)
μ
where ρ is the fluid density, g is the gravity acceleration, K is the permeability tensor and vβ  the seepage velocity, which
is defined as the superficial average:

1
vβ  = vβ dV (8)
V
Vβ

with V being the volume of the REV and vβ the pore-scale velocity vector. The same result was achieved exactly in the
same period by C. Marle [13]. However, in these articles, no clear structural link (i.e. a closure) is provided between the
micro- and the macroscale allowing one to infer the dependence of the permeability upon the pore-scale geometry. It was
not earlier than in the middle of the 1980s that a closed rigorous form was achieved by Whitaker [14] using the volume
averaging method [15], which included an intrinsic closure scheme allowing one to predict the values of the components
of the permeability tensor K. Some derivations of Darcy’s law have used other upscaling techniques such as homogeniza-
tion [16]. These and other techniques depart from the governing equations at the pore scale and, after the application of an
averaging operator (such as the one sketched in Fig. 1) and many mathematical steps, lead to an upscaled model in terms
of effective-medium coefficients that capture the essential (i.e. non-redundant) information from the pore scale. In this way,
the permeability tensor can be viewed as a signature of the porous medium topology at a scale that is larger than the pore
scale.
Over the past century, there have been some modifications brought to Darcy’s law that extended its applicability to
more complicated transport processes than those originally considered by Darcy. Among the extensions to Darcy’s law, a
non-exhaustive list should include: 1) Forchheimer’s modification to allow for the study of non-creeping flow regimes [17];
2) Brinkman’s correction to include macroscopic viscous stress by introducing an effective viscosity coefficient [18];
3) Klinkenberg’s modification of the permeability tensor to study gas slip flows in porous media [19]; 4) application to
heterogeneous media by means of large-scale volume averaging [20]; 5) non-isothermal flow of non-Newtonian fluids in
porous media (cf. [21] for instance).
JID:CRAS2B AID:3513 /SSU [m3G; v1.219; Prn:12/07/2017; 9:22] P.4 (1-10)
4 D. Lasseux, F.J. Valdés-Parada / C. R. Mecanique ••• (••••) •••–•••

In what follows, the focus is laid upon two extensions to Darcy’s law that are of major importance, namely the inertial
one-phase flow and the gas slip flow in homogeneous porous media. Our aim is to carefully review these extensions and draw
some conclusions and perspectives on these subjects.

2. Inertial one-phase flow in porous media

The analogy with flows in ducts has been widely used in the derivation of empirical flow models in porous media and
might have been a source of inspiration for H. Darcy to obtain the filtration law he reported in his book [1]. As mentioned
in a remarkable analysis by Zerner [3], H. Darcy dedicated specific experiments to verify Poiseuille’s law in the context of
slow flow. One of the major motivations was his questioning of the validity of the relationship between the “pressure drop”
 P and the average velocity u in a tube of length L, which was then of common use, i.e.
P
= au + bu 2 (9)
L
a relationship mainly due to du Buat and Gaspard Riche de Prony (see [3]), with a and b being coefficients that had to be
determined experimentally.

2.1. Forchheimer’s correction to Darcy’s law

It is striking to observe that the form of Eq. (9) corresponds to the relationship proposed by Forchheimer [17] to account
for “rapid” flows in porous media with the classical analogy on u taken as the filtration or seepage velocity in 1D:

P μ
= u + ρ β u2 (10)
L K
where ρ is the fluid density and β is the Forchheimer coefficient, also called the coefficient of inertial resistance or inertial
resistance factor.
The quadratic correction introduced by Forchheimer about half a century after the empirical validation and publication of
Darcy’s law was obviously inspired from Eq. (9), despite this remarkable time lapse between Forchheimer’s publication and
that of Darcy. In 1931, Muskat and Botset [2] reported experimental results of gas flow through different types of porous
materials, in which they observed that the gradient of the square of the pressure was proportional to a power of the mass
velocity ranging between 1 and 2 (1 when the flow was “completely viscous”, and 2 when it was “completely turbulent”).
This empirical form of Eqs. (9) and (10) was accepted to account for inertial flows in porous media over an additional
half-century during which only few alternative forms were put forth, like for instance [22]

P
= au + bu 2 + cu 3 (11)
L
with a, b and c being adjustable coefficients. During the fifty years following Forchheimer’s publication, some confusion
remained about the physical origin of the quadratic correction to Darcy’s law as it was often attributed to turbulence,
although some references made clear statements on that point [6]. In fact, Irmay [6] argued that there is no reason, in
general, to expect a linear solution to the non-linear Navier–Stokes equations. The Poiseuille solution in straight tubes is an
exception caused by vanishing curvature, which is not the case in real porous media. Agreement has been quite unanimous
on the threshold value of the Reynolds number at which the correction becomes significant, i.e. for 1 ≤ Red ≤ 15, where
ρ  v d
Red = β g
μ is based on the filtration velocity  v β  and the typical grain size d g of the porous material. However, it was
not before 1962 and the publication by Chauveteau and Thirriot [23], in which a flow regime classification was proposed,
that turbulence in this range of Reynolds numbers was dismissed. Turbulence has been confirmed to typically arise for
Red ∼ 100 [24–26].
During this period, and even up to the end of the 1970s, comparisons with experimental results were reported, showing
quite good agreement for various types of porous structures, including packed beds of grains, bundles of capillary tubes or
fibrous media for flows of gases or liquids [27,28]. From a practical point of view, the Forchheimer model has been used for
applications in hydrology, petroleum and chemical engineering [29,30]. From a theoretical point of view, the same period
was marked by various attempts to justify the form of Eq. (9). This was carried out on the basis of different approaches
ranging from simplified derivations from the Navier–Stokes equations or analogies with pressure drop through capillary
orifices [6]. The emergence of more formal upscaling techniques during the 1970s has led to further developments in the
following years that also attempted to justify the quadratic form of the inertial correction to Darcy’s law [13,31–33].

2.2. Refinements on the inertial correction to Darcy’s law

Impulsed by the development of both computational resources and numerical methods, the velocity dependence of
the corrective term on Darcy’s law regained much attention from the early 1990s on. It was during this period that the
quadratic velocity dependence of this correction was questioned and analyzed in depth. Numerical simulations carried out
JID:CRAS2B AID:3513 /SSU [m3G; v1.219; Prn:12/07/2017; 9:22] P.5 (1-10)
D. Lasseux, F.J. Valdés-Parada / C. R. Mecanique ••• (••••) •••–••• 5

through networks of parallel cylinders of circular cross-section for a flow orthogonal to the cylinders showed that the
correction scales as a 3rd power of the filtration velocity [34] instead of a quadratic term. This behavior was theoretically
confirmed quasi simultaneously, regardless the geometry at the pore-scale, from formal derivations based on a rigorous
upscaling procedure using double-scale homogenization with a closure process involving periodic representations of the
porous medium [35–37]. The exponent 3 was confirmed for a Reynolds number, Re p , based on the characteristic pore size
ranging between δ 1/2 and 1, where δ is the micro-to-macroscale ratio. This result was further emphasized later on [38] and,
from its original evidence, led to identify this regime as the so-called “weak inertia regime”, formalized for homogeneous
isotropic and periodic media. Additional numerical works over larger intervals of the Reynolds number for flows in many
different structures confirmed the existence of weak inertia and extended the classification of flows under (at least) three
distinct regimes with crossovers, namely [39] i) the weak inertia regime occurring at the onset of non-linearity in the
flow-rate-to-pressure-drop relationship for δ 1/2 ≤ Re p ≤ 1; ii) the “strong inertia regime” characterized by a correction to
Darcy’s law that scales as the square of the filtration velocity, i.e. leading to a Forchheimer type of model for Reynolds
numbers in the range 1 to 10; iii) the turbulent regime appearing for Reynolds numbers typically of the order of 100.
Nevertheless, a detailed physical explanation of a cubic or quadratic correction to Darcy’s law in order to account for
inertia still leaves much to be desired and remains a widely open question [40], while the description of the non-linearity
is essentially qualitative. Obviously, inertial macroscopic forces cannot be invoked as they remain negligible compared to
viscous forces [41], and this can be proved to hold as long as Re p  δ −1 . Consequently, the nature of the non-linearity in
the relationship between the macroscopic drag force and filtration velocity must certainly be explained from the signature
of viscous and inertial forces at the microscale. On the one hand, several mechanisms suggesting that inertia alone, at the
pore-scale, can explain the macroscale behavior may be put forth such as: i) streamlines bending due to the tortuosity of the
structure and to local converging-diverging flow patterns; ii) backflows and separations resulting from form drag; iii) pore
networks actively involved in the flow that are velocity-dependent as a consequence of i) and ii) yielding variations in the
dissipation of the kinetic energy [42]. On the other hand, microscale viscous drag effects may be considered to contribute
to the non-linearity when boundary layers at the solid–fluid interfaces, which become thicker when the Reynolds number
increases (see an experimental observation in [24]), are taken into account. With this mechanism, the inertial core flow (in
the center of the pores) may be easily understood as being strongly dependent upon the Reynolds number, partly explaining
the different regimes.

2.3. Further developments

Although the existence of the two regimes (weak and strong inertia) has been widely accepted, a lot still needs to be
understood regarding the universal existence and dependence of these regimes (and the associated crossovers) upon many
parameters such as porosity, structural order, anisotropy, etc. In addition, most of the experimental or numerical characteri-
zations of the macroscale inertial correction have been carried out in 1D in a scalar form although many references pointed
out that tensorial coefficients must be involved [31,36,41,32]. Even if the development did not allow for the formal identi-
fication of the above-mentioned regimes, a convincing derivation, relying on rigorous upscaling, of a macroscopic model for
momentum transport of one-phase flowing in homogeneous porous media with inertia is certainly  dueto Whitaker
 [41].
ρ  v β β
This model, subject to time (μt ∗ ρ 2β ), length-scales (β  r0  L) and Reynolds number
β
μ L
 1 constraints
that were clearly formulated, provides the macroscale momentum equation, which reads
H  
vβ  = − · ∇ p β β − ρ g (12)
μ
or equivalently
K  
vβ  = − · ∇ p β β − ρ g − F · vβ  (13)
μ
where vβ  is the filtration (or seepage) velocity as defined in Eq. (8), H is the apparent permeability, K the intrinsic
permeability and F the inertial correction tensor (which is a function of the filtration velocity); ∇ p β β is the macroscopic
pressure gradient,  p β β being defined in Eq. (3). Details on the averaging method employed to obtain this result are given
in [15]. In addition to the macroscopic model, the upscaling provides the means to determine the associated macroscopic
coefficients (K, H and F contained in Eqs. (12) and (13)) from the solution to the ancillary problems (so-called closure
problems). Undoubtedly, while completing the physical description, this contribution concludes some previous derivations
on the same problem performed with homogenization [36,40] and opens new perspectives to more in-depth investigation
of the inertial correction to Darcy’s law.
In this spirit, closure problems were solved in order to compute H and F for different ordered and disordered model
structures so as to investigate the existence of the regimes and their dependence upon the porosity and the microscale pore
structure [43]. The main important conclusions deriving from this analysis can be summarized as follows. The two tensors
H and F are generally dense and non-symmetric for ordered structures, even if the medium is isotropic at the macroscale
in the Darcy regime. The non-symmetry coincides with a macroscopic drag force, which is not necessarily aligned with the
macroscopic velocity. Symmetry is recovered for specific pressure gradient orientations along symmetry axes of the structure
JID:CRAS2B AID:3513 /SSU [m3G; v1.219; Prn:12/07/2017; 9:22] P.6 (1-10)
6 D. Lasseux, F.J. Valdés-Parada / C. R. Mecanique ••• (••••) •••–•••

(when present). Dissymmetry of these tensors decreases when structural disorder increases. Whatever the structure under
concern, the weak inertia regime (a cubic inertial correction to Darcy’s law) is always observed. For ordered structures, the
strong inertia regime (i.e. a quadratic (Forchheimer) correction to Darcy’s law) does not necessarily exist and it is otherwise
restricted to a narrow interval of the Reynolds number. For disordered structures, the Forchheimer type of correction is
a robust approximation over a very significant range of values of the Reynolds number. Moreover, the crossover value of
the Reynolds number at which the quadratic correction becomes relevant decreases with increasing structural disorder, the
weak inertia regime being restricted to a small range of Reynolds numbers where the correction is not very significant. This
certainly explains why this regime is most of the time overlooked in experimental investigations on porous media having a
random pore structure. In any situation, using a model such as that proposed in Eq. (9) implies that the permeability in the
linear term in the filtration velocity differs from the intrinsic permeability. Even if some progress has been achieved in the
physical explanation and the theoretical derivation of formal models to account for inertia effects for one-phase flows in
porous media, much work remains to be done to understand fully the non-linearities associated with this type of process.
For instance, most of the analyses so far were dedicated to the laminar steady regime and the occurrence of unsteadiness
remains widely unexplored (this problem was barely outlined recently in [44]), as well as the turbulent regime; however,
further discussion of these topics is out of the scope of the present review.

3. Slip flow in porous media

Gas flows in porous media differ considerably from liquid-phase flows, in particular for situations in which the pore sizes
are comparable to the mean free path of the gas molecules. This is the case in many practical applications including micro-
and nano-fluidic systems such as MEMS and nano-porous media, transport in fibrous media, gas flow during soil remedia-
tion, long-term nuclear waste disposal, among many others. Due to the current high relevance of this type of transport, the
evolution from the pioneering works performed in the nineteenth century to some of the current developments are briefly
summarized in this section.

3.1. Slip flow background

A one-phase flow in confined systems of dimensions comparable to the mean free path leads to rarefaction effects that
give rise to many interesting contributions in transport phenomena. In his classical study of the stresses that rarefied gases
experience, J.C. Maxwell [45] proposed that, close to the surface of a solid, there should be a sliding of the gas in contact
with the solid in the direction of a tangential stress. Maxwell proposed that the velocity should be proportional to the
tangential stress and inversely proportional to the viscosity of the fluid. Under isothermal conditions, the sliding velocity for
a 1D flow in the x-direction is expressed as:

dv
v=G (14)
dx
where the coefficient G is the coefficient of slipping, defined as
 
2 2
G= − 1 λβ (15)
3 f
with λβ being the mean free path and f the fraction of gas molecules that are diffusely scattered at the surface. Hence, if the
solid surface is wholly absorbent, G = 2/3λβ . The proposal from Maxwell is consistent with a previous study by Navier [46],
and we shall thus refer to Eq. (14) as the Navier–Maxwell equation. It should be noted that Maxwell obtained Eq. (14)
considering a single-component gas; the extension of this equation to multicomponent mixtures is found in Jackson [47].
During the last quarter of the nineteenth century, rarefaction effects were shown to increase significantly the flow rate
with respect to that predicted from Poiseuille’s law [48]. This motivated several investigations during the early years of the
twentieth century, in particular those by M. Knudsen. The scattering of gas molecules from solid walls was fundamental
in Knudsen’s theory [49]. Knudsen studied molecular flows in tubes and determined the dependence on tube dimensions.
He discussed the transition from Poiseuille’s flow in terms of the ratio of the mean free path of the gas molecules to the
characteristic size of the apparatus (say, β ). The reason for considering this ratio is due to the fact that, in the continuum
approach, the slip velocity may be understood as the average flow velocity of the molecules at a distance from the wall
that is equal to the mean free path [50]. Hence, as the mean free path becomes a bigger fraction of the tube diameter,
the slip velocity increases with respect to the bulk velocity. This important ratio between the mean free path and the
tube diameter is nowadays known as the Knudsen number in his honor (i.e. K n = λβ /β ). In the comprehensive review by
Steckelmacher [51], the historical development and relevance of Knudsen’s works are presented in detail.
Later on, Adzumi [52–54], published a series of papers dedicated to the study of gas flow through capillaries. He consid-
ered three cases: 1) when λβ is very small in comparison to the diameter of the capillary (i.e. K n  1); 2) when λβ is large
compared to the diameter (i.e. K n > 1), and 3) when λβ is comparable to the capillary diameter (i.e. K n ∼ 1). In the first
case, he found Poiseuille’s law to be quite suitable, and the flow is inversely proportional to the gas viscosity [52]. In the
second case, the flow rate was found to be independent of the viscosity, but inversely proportional to the square root of the
JID:CRAS2B AID:3513 /SSU [m3G; v1.219; Prn:12/07/2017; 9:22] P.7 (1-10)
D. Lasseux, F.J. Valdés-Parada / C. R. Mecanique ••• (••••) •••–••• 7

gas molecular weight, M, [53]. The flow characteristics in the third case were found by Adzumi to be a combination of the
two first ones [54]. Nowadays, there is some consensus that the following bounds are identifiable: for K n < 10−3 , the laws
of continuum mechanics are safely applicable, and non-slip can be assumed at the solid boundaries; for 10−3 < K n < 10−1 ,
the flow regime corresponds to slip flow, and the Navier–Maxwell equation must be considered at the solid–fluid interface;
for 10−1 < K n < 10, there is a transition regime in which the laws of continuum mechanics are likely to fail because the
continuum hypothesis is no longer satisfied. For porous media applications in the transition regime, Maxwell [55] proposed
that the action of the porous material over the gas was similar to a number of dust particles of the moving system, hence
giving rise to the well-known dusty gas model. This model has the nice feature that the interactions of gas molecules with
the dust molecules simulate their interaction with the rigid porous matrix, thus avoiding the problem of flux variations
across the sections of the pores [47]. Finally, for K n > 10 the gas kinetic theory must be considered because description in
terms of particle–wall collision operators is required. This limiting situation is called molecular streaming or Knudsen flow,
and it is characterized by the fact that the flow takes place by diffusion, instead of viscous, mechanisms [50]. Here, the term
diffusion means that the flow results from creep at the wall rather than from molecule-to-molecule collisions.

3.2. Pore-scale slip flow and its consequence on Darcy’s law

Few years after Adzumi’s works, in his study of the permeability of gas flows in porous media in the slip regime,
Klinkenberg [19] found that this coefficient is almost a linear function of the reciprocal mean pressure,  p β β . Using an
idealized porous medium representation consisting in an array of capillaries, Klinkenberg was able to deduce the following
relation between the apparent permeability, K s , and the intrinsic permeability, K :
 
b
Ks = K 1+ (16)
 p β β
with b being a surface- and gas-dependent constant. Evidently, at sufficiently large gas pressures, K s approaches K . Klinken-
berg used the above equation to predict the values of K s / K in different experimental conditions with reliable accuracy, thus
showing the need to consider slip effects in the determination of the permeability. In this way, the Darcy–Klinkenberg
model is (gravity is omitted here):
 
K b ∂ p β β
v β  = − 1+ (17)
μ  p β β ∂x
for an average one-dimensional flow in the x-direction, while μ is the fluid dynamic viscosity.
An important point of discussion in the recent literature related to gas transport in porous media is about the pertinence
of the linear and first-order interfacial boundary condition given in Eq. (14). Shen et al. [56] derived a first-order slip con-
dition from the Chapman–Enskog solution to the Boltzmann equation. This approach consists in linearizing the Boltzmann
equation using a perturbation expansion for the probability function in terms of the Knudsen number. The resulting condi-
tion includes an additional term due to the pressure gradient along the flow’s direction. The success of first-order models
has been argued to be constrained to slightly rarefied gas flows by Deissler [57]. According to this author, as the pressure
in the gas becomes smaller, the velocity profiles may be nonlinear over a distance from the solid surface corresponding to
the mean free path and the jumps at the interface may be expected to be functions of higher-order normal and tangential
derivatives. Deissler thus proposed to use a second-order boundary condition that matches the Navier–Stokes equations for
slip flows, finding good agreement with experimental results. Another early proposal of second-order models is the scalar
one by Cercignani [58] on the basis of the Bhatnagar–Gross–Krook (BGK) approximation of the Boltzmann equation. How-
ever, these extensions consist of modifications to the scalar equation (14), in other words, they are constrained to simple
geometries where one-dimensional flow is applicable. Unfortunately, extensions to more complicated geometries are not
straightforward and they remain a challenge. Furthermore, since the Navier–Stokes equations are first-order accurate in the
Knudsen number, it is not easy to justify the use of higher-order boundary conditions. Hence, an alternative is to use higher
order momentum transport equations, such as the Burnett equation [59], which are the result of keeping the second-order
terms in the Chapman–Enskog approach to approximate the Boltzmann equation.
As mentioned above, one of the limitations of the Navier–Maxwell boundary condition is that it is restricted to one-
dimensional flows. This limitation is also shared by the Klinkenberg model. To address this issue, Einzel et al. [60] proposed
a generalized version of the slip boundary condition, which can be expressed as follows:
   
2 − σν
vβ = − λβ n · ∇ vβ + ∇ vTβ · (I − nn) (18)
σν

 
ξ

where σν is the tangential-momentum accommodation coefficient, n is the unit normal vector directed from the fluid to-
ward the sold phase and I is the identity tensor. The coefficient σν accounts for the average tangential momentum exchange
between the molecules and the fluid, and can vary from zero to one.
JID:CRAS2B AID:3513 /SSU [m3G; v1.219; Prn:12/07/2017; 9:22] P.8 (1-10)
8 D. Lasseux, F.J. Valdés-Parada / C. R. Mecanique ••• (••••) •••–•••

3.3. Theoretical macroscopic slip flow models in porous media

In the same line of thoughts as those mentioned for inertial flows, the Klinkenberg model, although widely used in
the literature, was formally derived only more than half a century after its publication. Using the homogenization method,
Skjetne and Auriault [61] considered the steady-state, low-velocity Navier–Stokes equations for compressible flows in the
slip regime, i.e. Re  K n  1 (for Re = O(δ = / L ), with  and L being the characteristic length scales at the microscale and
macroscale, respectively). In this work, for the first time, a vectorial form of the Klinkenberg model was rigorously deduced
for conditions in which both local compressibility and inertial effects are smaller than the wall-slip effects. The apparent
permeability tensor was found to be positive-definite and non-symmetric, in general. This study was subsequently expanded
by Chastanet et al. [62] to derive the corresponding upscaled model for low-pressure gas flows in dual-porosity media
including fractures. Their study highlighted that the Knudsen number should be considered in addition to the separation of
characteristic length scales in the system in order to assess the domains of validity of upscaled models for gas flow.
The derivation of the effective-medium equation corresponding to slightly-compressible slip-flow conditions using the
volume averaging method has been carried out recently by Lasseux et al. [63]. This work completes the previous derivations
by Skjetne and Auriault on the following points: 1) the compressibility effects are taken into account in the framework of
slightly compressible flows restricted to small Reynolds numbers and small frequency number; 2) the vectorial form of the
slip boundary condition is considered in its complete form, i.e. including the complete shear-rate at the fluid-solid interface
as shown in Eq. (18); 3) it is derived for a barotropic fluid without any assumption on the equation of state of the gas. For
an ideal gas, the resulting upscaled model is
⎛  ⎞
1 ξμ π R  T β β
 v β  = − K · ⎝I + S⎠ (19)
μ  p β β 2M

which involves two tensors, namely, the intrinsic permeability tensor, K, and a slip-flow correction tensor, S. The parallelism
between Eqs. (17) and (19) is obvious. The ancillary closure problem required to predict the values of the effective-medium
coefficients was derived and formally solved for simple porous medium geometries in two- and three-dimensional unit
cells. Furthermore, the dependence of s (S = sI) on K (K = K I) was found to obey a power-law relationship, with the
value of the exponent depending on the geometrical configuration. In a subsequent work by Lasseux et al. [64], the slip
correction was more accurately described by considering an expansion in the Knudsen number at the closure level. This
leads to a reformulation of the closure problem as a differential (instead of an integro-differential) boundary-value problem,
which was solved in more complicated unit cells in order to predict the apparent permeability tensor. Furthermore, with
this expansion, the slip-flow correction tensor was shown to be the sum of slip corrections at the successive orders of the
Knudsen number. The consideration of the complete form of the boundary condition at the solid–fluid interface was found
to be crucial for the prediction of the slip corrections at the different orders of K n. Their analysis evidenced a nonlinear
relationship between the apparent permeability and the Knudsen number. This relationship motivates further theoretical
and experimental research on the subject, in particular for highly porous structures.

4. Conclusions

This work has been dedicated to the analysis of the evolution of two major modifications to Darcy’s law, namely the
inclusion of inertial and slip effects. The review carried out for both extensions suggests the following conclusions and
prospects.

– The empirical introduction of a quadratic correction in terms of the filtration velocity to Darcy’s law by Forchheimer in
1901 for 1D flows was certainly inspired by a model that was commonly used for flows in pipes prior to the publication
of Darcy’s law. This correction has been widely supported by empiricism for more than 90 years, a period after which a
first theoretical 3D model was achieved by upscaling (homogenization), showing that the onset of deviation from Dar-
cy’s law due to inertia involves a correction that rather scales as a 3rd power of the filtration velocity in a weak-inertia
regime. For larger Reynolds number values, a so-called strong inertia regime, where the Forchheimer correction is due
to hold, was accepted. An upscaled 3D complete model, obtained by volume averaging five years later, was used to high-
light the fact that, if the weak inertia regime always exists, the quadratic correction does not hold in some particular
situations of pore-scale ordered structures and that, however, in the presence of disorder, the Forchheimer-type correc-
tion is a robust one. At this point, a quite different observation from that indicated below for slip flows may be pointed
out regarding corrections made to Darcy’s law in order to include inertial effects. Indeed, if the underlying physics at
the pore scale does not rise any particular question, issues are mainly related to the understanding of the different flow
regimes at the macroscale, the physical mechanisms that trigger the transition from one regime to another and the
associated range of Reynolds numbers, together with the possible occurrence of unsteadiness.
– The original identification of the slip effect for gas flows in porous media, when the Knudsen number is not exceedingly
small compared to unity, by Klinkenberg (1941), which led this author to propose a correction to the intrinsic perme-
ability (in the 1D case) inversely proportional to the mean pressure, has been accepted with empirical justifications for
almost 60 years. After this period, a first theoretical 3D model was derived, followed by more refined ones 15 years later
JID:CRAS2B AID:3513 /SSU [m3G; v1.219; Prn:12/07/2017; 9:22] P.9 (1-10)
D. Lasseux, F.J. Valdés-Parada / C. R. Mecanique ••• (••••) •••–••• 9

using rigorous upscaling techniques. While these upscaling tools are now at hand for such theoretical developments,
it clearly appears that the main issues lie in an appropriate pore-scale description of the physics in that case, and in
particular, in the slip-boundary condition, which still requires some efforts to better capture, and possibly extend, its
domain of application in terms of the Knudsen number. In the same spirit, despite some attempts, macroscale models
for transitional and strongly rarefied flow regimes still require important efforts. In parallel, numerical and experimental
investigations are necessary to highlight the understanding of the detailed physics.

As a prospect, one may finally conclude with the following. From the evolution of two modifications to Darcy’s law inves-
tigated in the present work, it is important to remark that, after one-dimensional corrections were proposed, followed by
extensive experimental analyses, there was a long time period before rigorous deductions were presented. This delay can be
explained by the fact that the different necessary theoretical frameworks for performing upscaling from the pore-scale to the
macroscale are relatively new (about 30–40 years old) and also by the fact that, in many situations, the one-dimensional ver-
sions remained relatively satisfactory. However, with recent applications directed to micro- and nanofluidic devices, among
others, there is a need for more accurate macroscale models, that can be validated through comparison with reliable ex-
perimental data and direct numerical simulations. In this last issue, there is a strong tendency in current research towards
understanding macroscale phenomena through imaging and direct simulations. This has been made possible by recent ad-
vances in computational capabilities. With this perspective in mind, it is relevant to pose the question of which direction
future advances of flows in porous media will take. Probably, experimental and numerical studies would still evolve very
significantly. However, strong efforts should certainly be dedicated to more sophisticated upscaling approaches applicable to
more challenging situations than those studied in the current state of the art.

References

[1] H. Darcy, Les Fontaines Publiques de la Ville de Dijon, Dalmont, Paris, 1856.
[2] M. Muskat, H.G. Botset, Flow of gas through porous materials, J. Appl. Phys. 27 (1931) 27–47, http://dx.doi.org/10.1063/1.1744983.
[3] M. Zerner, Aux origines de la loi de darcy (1856), Docs. hist. tech., http://dht.revues.org/1625.
[4] J. Boussinesq, Recherches théoriques sur l’écoulement des nappes d’eau infiltrées dans le sol et sur le débit des sources, J. Math. Pures Appl. 10 (1904)
5–78.
[5] O. Emersleben, The Darcy filter formula, Phys. Z 26 (1925) 601–610.
[6] S. Irmay, On the theoretical derivation of Darcy and Forchheimer formulas, J. Geophys. Res. 39 (1958) 702–707.
[7] W.A. Hall, An analytical derivation of the Darcy equation, Trans. Am. Geophys. Union 37 (1956) 185–188.
[8] M. Muskat, Flow of Homogeneous Fluids through Porous Media, McGraw–Hill, 1937.
[9] F. Blake, The resistance of packing to fluid flow, Trans. AIChE 14 (1922) 415–421.
[10] J. Kozeny, Ueber kapillare leitung des wassers im boden, Sitz.ber. – Akad. Wiss. Wien 136 (1927) 271–306.
[11] P.C. Carman, Flow of Gases Through Porous Media, Academic Press, New York, 1956.
[12] S. Whitaker, The equations of motion in porous media, Chem. Eng. Sci. 21 (3) (1966) 291–300, http://dx.doi.org/10.1016/0009-2509(66)85020-0.
[13] C. Marle, Écoulements monophasiques en milieu poreux, Rev. Inst. Fr. Pét. XXII (1967) 1471–1509.
[14] S. Whitaker, Flow in porous media I: a theoretical derivation of Darcy’s law, Transp. Porous Media 1 (1) (1986) 3–25, http://dx.doi.org/10.1007/
BF01036523.
[15] S. Whitaker, The Method of Volume Averaging, Kluwer Academic Publishers, 1999.
[16] J.L. Auriault, Nonsaturated deformable porous media: quasistatics, Transp. Porous Media 2 (1987) 405–464.
[17] P. Forchheimer, Wasserbewegung durch boden, Z. Ver. Dtsch. Ing. XXXXV (49) (1901) 1781–1788.
[18] H.C. Brinkman, A calculation of the viscous force exerted by a flowing fluid on a dense swarm of particles, Flow Turbul. Combust. 1 (1) (1949) 27–34,
http://dx.doi.org/10.1007/BF02120313.
[19] L.J. Klinkenberg, The Permeability of Porous Media to Liquids and Gases, Amer. Pet. Inst., 1941, pp. 200–213, https://www.onepetro.org/conference-
paper/API-41-200.
[20] M. Quintard, S. Whitaker, Écoulement monophasique en milieu poreux: effet des hétérogénéités locales, J. Méc. Théor. Appl. 6 (1987) 691–726.
[21] R.R. Kairi, P.V.S.N. Murthy, Effect of viscous dissipation on natural convection heat and mass transfer from vertical cone in a non-Newtonian fluid
saturated non-Darcy porous medium, Appl. Math. Comput. 217 (20) (2011) 8100–8114, http://dx.doi.org/10.1016/j.amc.2011.03.013.
[22] P.Y. Polubarinova-Kochina, Theory of Ground Water Motion, Goss. Izdat. Tekh.-Teoret. Lit., Moscow, 1952, see also a translation by J.M. Roger de Wiest,
Princeton University Press, 1962.
[23] G. Chauveteau, C. Thirriot, Régimes d’écoulement en milieu poreux et limite de la loi de Darcy, Houille Blanche 2 (1967) 141–148.
[24] A. Dybbs, R.V. Edwards, A new look at porous media fluid mechanics – Darcy to turbulent, in: J. Bear, M.Y. Corapcioglu (Eds.), Fundamentals of
Transport Phenomena in Porous Media, Springer Nature, Dordrecht, The Netherlands, 1984, pp. 199–256.
[25] C.K. Ghaddar, On the permeability of unidirectional fibrous media: a parallel computational approach, Phys. Fluids 7 (11) (1995) 2563–2586.
[26] J.B. Koch, J.C. Ladd, Moderate Reynolds number flows through periodic and random arrays of aligned cylinders, J. Fluid Mech. 340 (1997) 31–66.
[27] S. Ergun, Fluid flow through packed columns, Chem. Eng. Prog. 48 (1952) 89–94.
[28] F.A.L. Dullien, M.I.S. Azzam, Flow rate-pressure gradient measurement in periodically nonuniform capillary tubes, AIChE J. 19 (1973) 222–229.
[29] J. Bear, Dynamics of Fluids in Porous Media, Dover, New York, 1972.
[30] J. Geertsma, Estimating the coefficient of inertial resistance in fluid flow through porous media, SPE J. (1974) 445–450.
[31] S.M. Hassanizadeh, W.G. Gray, High velocity flow in porous media, Transp. Porous Media 2 (1987) 521–531.
[32] T. Giorgi, Derivation of the Forchheimer law via matched asymptotic expansions, Transp. Porous Media 29 (1997) 191–206.
[33] Z. Chen, S.L. Lyons, G. Qin, Derivation of the Forchheimer law via homogenization, Transp. Porous Media 44 (2) (2001) 325–335.
[34] J. Barrère, Modélisation des écoulement de Stokes et de Navier–Stokes en milieu poreux, Ph.D. thesis, University of Bordeaux, France, 1990.
[35] J.C. Wodie, T. Levy, Correction non linéaire de la loi de Darcy, C. R. Acad. Sci. Paris, Ser. II (1991) 157–161.
[36] C.C. Mei, J.L. Auriault, The effect of weak inertia on flow through a porous medium, J. Fluid Mech. 222 (1991) 647–663.
[37] M. Rasoloarijaona, J.L. Auriault, Nonlinear seepage flow through a rigid porous medium, Eur. J. Mech. B, Fluids 13 (1994) 177–195.
[38] E. Skjetne, A. Hansen, J.S. Gudmundsson, High velocity flow in a rough fracture, J. Fluid Mech. 383 (1999) 1–28.
[39] E. Skjetne, J.L. Auriault, New insights on steady, non-linear flow in porous media, Eur. J. Mech. B, Fluids 18 (1) (1999) 131–145.
JID:CRAS2B AID:3513 /SSU [m3G; v1.219; Prn:12/07/2017; 9:22] P.10 (1-10)
10 D. Lasseux, F.J. Valdés-Parada / C. R. Mecanique ••• (••••) •••–•••

[40] E. Skjetne, J.L. Auriault, High-velocity laminar and turbulent flow in porous media, Transp. Porous Media 36 (1999) 131–147.
[41] S. Whitaker, The Forchheimer equation: a theoretical development, Transp. Porous Media 25 (1996) 27–61.
[42] J.S. Andrade, U.M.S. Costa, M.P. Almeida, H.A. Makse, H.E. Stanley, Inertial effects on fluid flow through disordered porous media, Phys. Rev. Lett. 82 (26)
(1999) 5249–5252.
[43] D. Lasseux, A.A. Abbasian Arani, A. Ahmadi, On the stationary macroscopic inertial effects for one phase flow in ordered and disordered porous media,
Phys. Fluids 23 (7) (2011) 073103.
[44] M. Agnaou, D. Lasseux, A. Ahmadi, From steady to unsteady laminar flow in model porous structures: an investigation of the first Hopf bifurcation,
Comput. Fluids 136 (2016) 67–82, http://dx.doi.org/10.1016/j.compfluid.2016.05.030.
[45] J.C. Maxwell, On stresses in rarified gases arising from inequalities of temperature, Philos. Trans. R. Soc. Lond. 170 (1879) 231–256, http://dx.doi.org/
10.1098/rstl.1879.0067.
[46] M. Navier, Mémoire sur les lois du mouvemet des fluides, Academie royale des sciences de l’institut de France, 1822.
[47] R. Jackson, Transport in Porous Catalysts, Elsevier, 1977.
[48] A. Kundt, E. Warburg, On friction and heat-conduction in rarefied gases, Philos. Mag. Ser. 4 50 (328) (1875) 53–62, http://www.tandfonline.com/
doi/abs/10.1080/14786447508641259?journalCode=tphm15.
[49] M. Knudsen, Die gesetze der molekularstromung und der inneren reibungsströmung der gase durch röhren, Ann. Phys. 28 (1909) 75–130.
[50] F.A.L. Dullien, Porous Media. Fluid Transport and Pore Structure, 2nd ed., Academic Press, 1992.
[51] W. Steckelmacher, Knudsen flow 75 years on: the current state of the art for flow of rarefied gases in tubes and systems, Rep. Prog. Phys. 49 (10)
(1986) 1083–1107, http://dx.doi.org/10.1088/0034-4885/49/10/001.
[52] H. Adzumi, Studies on the flow of gaseous mixtures through capillaries. I. The viscosity of binary gaseous mixtures, Bull. Chem. Soc. Jpn. 12 (5) (1937)
199–226, http://dx.doi.org/10.1246/bcsj.12.199.
[53] H. Adzumi, Studies on the flow of gaseous mixtures through capillaries. II. The molecular flow of gaseous mixtures, Bull. Chem. Soc. Jpn. 12 (6) (1937)
285–291, http://dx.doi.org/10.1246/bcsj.12.285.
[54] H. Adzumi, Studies on the flow of gaseous mixtures through capillaries. III. The flow of gaseous mixtures at medium pressures, Bull. Chem. Soc. Jpn.
12 (6) (1937) 292–303, http://dx.doi.org/10.1246/bcsj.12.292.
[55] J.C. Maxwell, II. Illustrations of the dynamical theory of gases, Philos. Mag. Ser. 4 20 (130) (1860) 21–37, http://www.tandfonline.com/doi/abs/
10.1080/14786446008642902.
[56] S. Shen, G. Chen, R.M. Crone, M. Anaya-Dufresne, A kinetic-theory based first order slip boundary condition for gas flow, Phys. Fluids 19 (8) (2007)
086101, http://dx.doi.org/10.1063/1.2754373.
[57] R.G. Deissler, An analysis of second-order slip flow and temperature-jump boundary conditions for rarefied gases, Int. J. Heat Mass Transf. 7 (6) (1964)
681–694, http://dx.doi.org/10.1016/0017-9310(64)90161-9.
[58] C. Cercignani, Higher Order Slip According to the Linearized Boltzmann Equation, Tech. rep., Institute of Engineering Research Report AS-64-19, Uni-
versity of California, Berkeley, 1964.
[59] L. García-Colín, R. Velasco, F. Uribe, Beyond the Navier–Stokes equations: Burnett hydrodynamics, Phys. Rep. 465 (4) (2008) 149–189, http://dx.doi.org/
10.1016/j.physrep.2008.04.010.
[60] D. Einzel, P. Panzer, M. Liu, Boundary condition for fluid flow: curved or rough surface, Phys. Rev. Lett. 64 (19) (1990) 2269–2272.
[61] E. Skjetne, J.L. Auriault, Homogenization of wall-slip gas flow through porous media, Transp. Porous Media 36 (1999) 293–306.
[62] J. Chastanet, P. Royer, J.-L. Auriault, Flow of low pressure gas through dual-porosity media, Transp. Porous Media 66 (3) (2007) 457–479,
http://dx.doi.org/10.1007/s11242-006-0023-y.
[63] D. Lasseux, F.J. Valdés-Parada, J.A. Ochoa-Tapia, B. Goyeau, A macroscopic model for slightly compressible gas slip-flow in homogeneous porous media,
Phys. Fluids 26 (2014) 053102.
[64] D. Lasseux, F.J. Valdés-Parada, M. Porter, An improved macroscale model for gas slip flow in porous media, J. Fluid Mech. 805 (2016) 118–146.