Physica A 311 (2002) 320 – 326

www.elsevier.com/locate/physa

Statistical properties for two-dimensional !uid

!ow in percolation porous media
Xiao-Hong Wang ∗ , Zhi-Feng Liu, Qing-Song Wu, Bo Li
Department of Thermal Science and Energy Engineering, University of Science and Technology of
China, Hefei, Anhui 230026, People’s Republic of China

Received 7 March 2002

Abstract
We study the statistical properties of !uid !ows in percolation porous media at very low
Reynolds number by numerical simulations for 20,000 di1erent con2gurations. It is shown that
there are some important di1erences between !uid !ows in macroscopically homogeneous and
fractal porous media. For !uid !ows in macroscopically homogeneous media, the pressure is
de2nite, but the velocity is random and depends on the structural details of porous media. The
permeability k for !uid !ows in fractal changes with the size L as k ∼ L− where  ≈ 1:0,
not approaching the constant expected from Darcy’s law. The statistical distribution of pressure
in fractal is independent of the size L and can be approximated by a function of the triangular
shape. c 2002 Elsevier Science B.V. All rights reserved.

PACS: 47.55.Mh; 47.11.+j; 64.60.Ak

Keywords: Porous media; Percolation; Fractals

Fluid !ows through porous media plays an important role in a wide variety of
environmental and technological processes. Examples include oil recovery, the spread
of hazardous waste in soils and heterogeneous catalysis. The porous rock is highly
heterogeneous and can be separated into high permeability and low permeability. The
spatial distribution is random, in which case this is just the percolation problem [1,2]. In
the recent, Boston’s group performed the numerical simulations for !uid !ows through
the percolation porous structures [3–5]. It was found out that the local kinetic energy
of the !uids has the power-law distribution [4], and the convective e1ects and break-
through time are also studied [6 –9]. In this paper, based upon the Boston’s group

∗ Corresponding author.
E-mail address: [email protected] (X.-H. Wang).

0378-4371/02/$ - see front matter
c 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 3 7 8 - 4 3 7 1 ( 0 2 ) 0 0 8 3 8 - 5
 X.-H. Wang et al. / Physica A 311 (2002) 320 – 326 321

work [3–5], we study the statistical properties of two-dimensional !uid !ows in per-
colation porous media at very low Reynolds numbers by collecting 20,000 di1erent
con2gurations.
The model for the pore connectivity is based on two-dimensional site percolation.
Void spaces occupy the L × L square lattice with the probability p. Nonslip boundary
conditions at the solid–!uid interface is applied. At the inlet and exit in the vertical
direction, the velocity gradient is assumed to be zero. The pressure is set to be P = 1
and 0 at the lower and the upper boundary, so that the pressure drop P between the
inlet and exit is P = 1. In the horizontal direction, the periodical condition is used.
We ignore the convective e1ects to study the !uid !ows in porous media at very low
Reynolds number. SIMPLE method [10] is used to solve Stock’s equation
∇2 v = −∇P; ∇·v=0; (1)
where  is the !uid viscosity. We implement grid elements with length equal to the
solid cell size l. Not losing the generality, we assume that all the quantities are
dimensionless. Due to the linearity of Eq. (1), we can set l= = 1 when establishing
the 2nite-di1erence scheme of Eq. (1). Twenty thousands con2gurations are collected
to get the statistical distributions of !ow velocity and pressure at the central point of
the L × L square lattice and !uid !ux at the inlet. Dealing with the statistical properties
at the central point in the percolation structures, the samples that the central point
occupied by the solid or belonging to an isolated !uid area are cuto1.
Fluid !ows in microscopically disordered and macroscopically homogeneous porous
media for small Reynolds number can be described by Darcy’s law [11,12]
k P
V =− ; (2)
 L̂
where V is the !uid velocity averaged over the total cross-section of the porous sample,
k is the permeability and P is the pressure drop over the sample length L̂(L̂ = Ll).
Darcy’s law is analogous to Ohm’s law for the !ow of electrical current, and k is the
counterpart of the e1ective conductivity. In the percolation structure, the !uid !ows
through the sample-spaning cluster connecting the two opposite edges of the network.
It is well known that the sample-spaning cluster is fractal when p = pc and statistical
homogeneous when p ¿ pc [13]. We do the numerical simulations for !uid !ows
in the percolation porous media with the di1erent sizes L. Averaging over 20,000
con2gurations, at p = 0:75, it is shown that the ensemble average !uid !ux Q at
the inlet approaches constant for the large size L (see Fig. 1), which is equivalent to
Darcy’s law. At p=pc , the !uid pathway has the fractal structure. In fractal substrates,
the obstacles of all sizes up to the size of the system exist. If we increasing the size L,
we increase the size of the obstacles as well, and by this we decrease the conductivity.
As the result, Darcy’s law is breakdown for !uid !ows in fractals. The simulations
give
Q ∼ L−˜ ; (3)
where ˜ ≈ 1:0. It is consistent with the conductivity of the fractal conductors:  ∼ L−˜
where ˜ ≈ 0:9745 [13]. This result may be served to explain the relation of the average
322 X.-H. Wang et al. / Physica A 311 (2002) 320 – 326

−1
10

p=0.75

<Q>
slope=1.0
−2
10

p=pc

−3
10 1 2
10 10
L

Fig. 1. Plot of the ensemble average !uid !ux Q as a function of the size L for p = pc = 0:5927 (2lled
triangle up) and p = 0:75 ¿ pc (2lled diamond).

O ∼ t 0:38 in the paper imbibition experiment of HorvPath and

height with the time: h(t)
Stanley [14] if assuming that the paper imbibition is !uid !ow in fractal substracts.
In this paper, we investigate the statistical distributions of the !uid !ux and the
horizontal velocity, vertical velocity and pressure at the central point belonging to the
sample-spaning cluster. When p = 0:75 ¿ pc , numerical simulations show that the hor-
izontal velocity vx has the exponential distribution (see Fig. 2(a)). Fig. 2(b) shows that
the probability density function (pdf) f(vy ) of the vertical velocity vy is asymmetrical.
Its negative part is very steep and its positive part presents an exponential decay, which
is in agreement with the experimental measurements by nuclear magnetic resonance
(NMR) imaging [15 –18]. The statistical distributions of the !uid !ux Q and the pres-
sure P of the central point are Gaussian (see Fig. 2(c) and (d)). At p=pc , although the
horizontal and vertical velocities seem to have the exponential decay for the large value,
Fig. 3(a) shows that the statistical distributions for the small value are complicated.
Investigating the pdfs of the horizontal velocity
magnitude |vx |, the vertical velocity
magnitude |vy | and the velocity magnitude v = vx2 + vy2 , we 2nd that they have the
power-law behaviors for the intermediate values and the scaling exponent is about 0.63
(see Fig. 3(b)). We note that the velocity magnitude in the percolation porous media of
one con2guration also has the power-law behavior and the scaling exponent is 0.71 [7].
It is interesting that the pdf of the !uid !ux Q is not Gaussian and can be 2tted very
well by 2 distribution: f(Q) = 6 × 108 (Q − 0:001)2:0 e−1061(Q−0:001) . For the pressure
at the central point which belongs to the sample-spaning cluster, we surprisingly 2nd
that its pdf can be approximated very well by the function: f(P) = 3:43(P + 0:04) if
−0:04 ¡ P 6 0:5 and f(P) = 3:43(1:04 − P) if 0:5 6 P ¡ 1:04 apart from the neigh-
borhoods of P = 0 and P = 1. We 2nd that the pressure P of the central point can
be P = Pmin = −0:04 ¡ 0 and P = Pmax = 1:04 ¿ 1 with the small probability due to
the e1ects of the local !ow in the opposite direction of the pressure gradient. We note
that the statistical distribution of the pressure is invariant with the di1erent sizes L.
 X.-H. Wang et al. / Physica A 311 (2002) 320 – 326 323

3 3
10 10

2
10

f(vy)
f(vx)

2 1
10 10

0
10

1 −1
10 10
− 0.004 − 0.002 0 0.002 0.004 − 0.002 0 0.002 0.004 0.006
(a) vx (b)
vy
60
6

40
4
f(Q)

f(P)

20
2

0 0
0.03 0.05 0.07 0.09 0.11 0.1 0.3 0.5 0.7 0.9
(c) Q (d) P

Fig. 2. At p = 0:75 ¿ pc , the pdfs of (a) the vertical velocity vx at the central point, (b) the horizontal
velocity vy at the central point, (c) the !uid !ux Q at the inlet where circle corresponds to the Gaussian
distribution and (d) the pressure P at the central point where circle corresponds to the Gaussian distribution.

Due to the linearity of Eq. (1), the size L → ∞ is equivalent to the cell size
l → 0. Larger size L corresponds to more re2ned percolation network. 1 Fig. 4(a)
shows that the !uctuations of the !uid !ux RQ have RQ ∼ L− where  ≈ 1:1,
when p = 0:75 ¿ pc . When the size L is 2nite, for the macroscopically homogeneous
porous media (p ¿ pc ), Darcy’s law can be recovered by averaging over many dif-
ferent con2gurations of the percolation structures. Only the size L is large (L → ∞),
the structural details of the porous media can be ignored. Fig. 4(a) also shows that
the !uctuations of the pressure has RP ∼ L−0:82 . This means that the pressure in the
sample cluster is de2nite when the cell size l → 0. This may be explained as that the
pressure is determined by its location and the connectivity to the two opposite boundary
and the structure details of the percolation porous media can be ignored. However, the

1 Assume that the velocity v is the solution of equation (1) with the pressure drop P = 1 for the size L
i i
where the cell size is l. We rescale the spatial length so that Li l = 1. From Eq. (1), we have vi = Li lvi
where vi is the solution in the rescaling space. The pressure Pi (m; n)=Pi (m; n) where Pi (m; n) is the pressure
at the point (ml ; nl ) in the rescaling space and Pi (m; n) is the pressure at the point (ml; nl).
324 X.-H. Wang et al. / Physica A 311 (2002) 320 – 326

4
10
3
10

f(|vx|)(f(|vy|),f(|v|)
slope=− 0.63
3
f(vx)(f(vy))

10
2
10

2
1 10
10

0 1
10 10
−5 −4 −3
− 0.002 0 0.002 10 10 10
(a) vx (vy) |vx|(|vy|,v)
(b)

300 2

200
f(Q)

f(P)

1
100

0 0
0.001 0.006 0.011 − 0.2 0 0.2 0.4 0.6 0.8 1 1.2
(c) (d)
Q P

Fig. 3. At p = pc = 0:5927, the pdfs of (a) the vertical velocity vx (solid line) and the horizontal velocity vy
(dotted line) at the central point, (b) the

magnitude of the vertical velocity, horizontal velocity and velocity
at the central point: |vx |; |vy | and |v| = vx2 + vy2 , (c) the !uid !ux Q at the inlet where circle corresponds to
the 2
distribution: f(Q) = 6 × 10 (Q − 0:001)2:0 e−1061(Q−0:001) and (d) the pressure P at the central point
8

where circle corresponds to the function: f(P)=3:43(P +0:04) if −0:4 ¡ P 6 0:5 and f(P)=3:43(1:04−P)
if 0:5 6 P ¡ 1:04.

!uctuations of the horizontal velocity and the vertical velocity approach constants (See
Fig. 4(b)). As a result, the velocity in the statistical homogeneous porous media is
random and depends on the structure details, for example, the local pore size. When
p = pc , the sample-spaning cluster is fractal. We have the scaling relation: LRvx ∼
L−0:4 ; LRvy ∼ L−0:53 and RQ ∼ L−1:04 (see Fig. 4(c)). At p = pc , the velocity v → 0
when the cell size l → 0(L → ∞), which is in agreement with that the permeability
k → 0 at the threshold p = pc [4]. However, the !uctuations of the pressure RP is
constant for the di1erent sizes L at p = pc . This means that the pressure at the central
point is di1erent for the di1erent con2gurations. Although there exists the pressure
drop between the central point and the inlet or exit, the velocity in the sample-spaning
should tend to zero because the length of the path from the central point to the inlet
or exit tends to in2nity in fractal structure.
 X.-H. Wang et al. / Physica A 311 (2002) 320 – 326 325

p=0.75 p=0.75

SLOPE=− 0.82
The fluctuations

−1 −1
10 10
∆P L∆vy

SLOPE=− 1.1 L∆vx

−2 −2
10 10
∆Q

1 2 1 2
10 10 10 10
(a) L (b) L

0
10
p=pc
∆P

−1
10 slope=− 0.53

L∆vy

−2
10 slope=− 0.4 L∆vx

∆Q
slope=− 1.04
−3
10 1 2
10 10
(c) L

Fig. 4. The !uctuations as a function of the size L for (a) the !uid !ux (2lled circle) and the pressure (2lled
square) when p = 0:75, (b) the horizontal velocity (2lled diamond) and vertical velocity (2lled triangle)
when p = 0:75 and (c) the horizontal velocity (2lled circle), the vertical velocity (2lled square), the !uid
!ux (2lled diamond) and pressure (2lled triangle) when p = pc .

In summary, based upon the direct numerical solutions of Stock’s equation, it is

shown that !uid !ows in fractal porous media is quite di1erent from !uid !ows in
microscopically disordered and macroscopically homogeneous porous media. Darcy’s
law is only correct for !uid !ows in macroscopically homogeneous porous media,
not for !uid !ows in fractals. For !uid !ows in macroscopically homogeneous porous
media, our calculations indicate that, in contrast to being de2nite for pressure, the
velocity is random and depends on the structural details of the porous media. However,
setting the pressure drop between the two opposite boundaries, the pressure in fractal
substrate is random and depends on the structure details of fractal porous media though
the velocity tends to zero. We also 2nd that the 2nite size e1ect has little in!uence
upon the pressure 2eld in fractal porous media.

This work is supported by the National Natural Science Foundation of China (Grant
No.10002019) and the Opening Foundation of the State Key Laboratory of Oil=Gas
Reservoir Geology and Exploitation (Grant No. PLN0101).
326 X.-H. Wang et al. / Physica A 311 (2002) 320 – 326

References

[1] H.E. Stanley, J.S. Andrade Jr., S. Havlin, H.A. Makse, B. Suki, Physica A 266 (1999) 5.
[2] P.R. King, J.S. Andrade Jr., S.V. Buldyrev, N. Dokholyan, Y. Lee, S. Havlin, H.E. Stanley, Physica A
266 (1999) 107.
[3] J.S. Andrade Jr., M.P. Almeida, J.M. Filho, S. Havlin, B. Suki, H.E. Stanley, Phys. Rev. Lett. 79 (1997)
3901.
[4] J.S. Andrade Jr., D.A. Street, T. Shinohara, Y. Shibusa, Y. Arai, Phys. Rev. E 51 (1995) 5725.
[5] J.S. Andrade Jr., U.M.S. Costa, M.P. Almeida, H.A. Makse, H.E. Stanley, Phys. Rev. Lett. 82 (1997)
5249.
[6] Y. Lee, J.S. Andrade Jr., S.V. Buldyrev, N.V. Dokholyan, S. Havlin, P.R. King, G. Paul, H.E. Stanley,
Phys. Rev. E 60 (1999) 3425.
[7] H.A. Makse, J.S. Andrade Jr., H.E. Stanley, Phys. Rev. E 61 (2000) 583.
[8] J.S. Andrade Jr., S.V. Buldyrev, N.V. Dokholyan, S. Havlin, P.R. King, Y. Lee, G. Paul, H.E. Stanley,
Phys. Rev. E 62 (2000) 8270.
[9] J.S. Andrade Jr., A.D. AraPujo, S.V. Buldyrev, S. Havlin, H.E. Stanley, Phys. Rev. E 63 (2000) 051403.
[10] S.V. Patankar, Numerical Heat Transfer and Fluid Flow, Hemisphere, Washington, DC, 1980.
[11] P.M. Alder, Porous Media: Geometry and Transports, Butterworth–Heinemann, Boston, 1992.
[12] M. Sahimi, Flow and Transport in Porous Media and Fractured Rock, VCH, Boston, 1995.
[13] A. Bunde, S. Havlin, in: A. Bunde, S. Havlin (Eds.), Fractals and Disordered Systems, Springer,
Heidelberg, 1991.
[14] V.K. HorvPath, H.E. Stanley, Phys. Rev. E 52 (1995) 5166.
[15] M. Shattuck, R. Behringer, J. Georgiadis, H.T. Davis, in: O. Hern, R. Gore (Eds.), Experimental
Techniques in Multiphase Flows, ASME Proceedings, ASME, New York, 1991.
[16] J. Georgiadis, R. Behringer, M. Shattuck, G. Johnson, Ninth Symposium of Energy Engineering
Sciences, Fluid and Dynamical Systems, Argonne National Lab., Argonne, IL, 1991.
[17] L. Lebon, L. Oger, J. Leblond, J.P. Hulin, N.S. Martys, L.M. Schwartz, Phys. Fluids 8 (1996) 293.
[18] Y.E. Kutsovsky, L.E. Scriven, H.T. Davis, B.E. Hammer, Phys. Fluids 8 (1996) 863.