PHYSICAL REVIEW E 71, 016311 共2005兲

Deposition in porous media and clogging on the field scale

E. A. Borisova and P. M. Adler
IPGP, 4, Place Jussieu, 75252-Paris-Cedex 05, France
共Received 10 July 2004; revised manuscript received 25 August 2004; published 14 January 2005兲

Deposition is modeled as a first order reaction on the Darcy scale for porous media which are statistically
homogeneous. An elementary analytical solution is derived. A parametric study was done with a three-
dimensional code which is briefly described and checked in media where the solution is known. The role of the
parameters, including the artificial ones, is discussed with an illustrative example. When the Damköhler
number is small, deposition causes smooth changes to the porosity field; the evolution of porosity is well
described by the analytical solution. Very different results are obtained for large Damköhler numbers. The
influence of the correlation of the initial porosity field is studied.

DOI: 10.1103/PhysRevE.71.016311 PACS number共s兲: 47.70.Fw, 47.55.Mh

I. INTRODUCTION caused by hydrodynamic, physico-chemical and chemical in-

Deposition in porous media is a process of high practical teractions between fluids and reservoir rocks. Xu et al. 关1兴
and fundamental interest, since this problem arises in various developed a reactive fluid flow and geochemical transport
branches of industry, particularly in chemical, civil, and pe- numerical model for evaluating long-term CO2 disposal in
troleum engineering. For instance, the deposition at the solid deep geologic formations. This study makes use of an exist-
surface of a reservoir may considerably diminish the oil pro- ing nonisothermal reactive geochemical transport model
duction. Other standard applications include water treatment TOUGHREACT 关10兴. It was applied to a one-dimensional 共1D兲
and environmental studies of transport of chemical contami- radial flow which schematizes flow around a well. No
nants. A recent application is the reduction of the release rate change in the transport properties was taken into account
of carbon dioxide to the atmosphere by mineral trapping in though it is possible.
deep permeable geologic formations 共Xu et al. 关1兴兲. As a rule, the results of the pore scale simulations are
The topic of changes in hydrologic properties due to compared to the experimental measurements on the core
chemical reactions has recently been reviewed by Saripalli et samples which are supposed to be homogeneous. Reservoir
al. 关2兴. Most of the understanding of the reactive transport simulations are applied to fields extending from several
has been obtained on the pore scale. On this scale, the term meters to kilometers. In this case, the geological structure of
deposition can be used in different meanings, referring either the reservoir is taken into account by means of the macro-
to the solute precipitation on the fluid-solid interface, to the
scopic parameters which vary in space and time. Liu et al.
trapping of fine particles at pore throats, or to ion exchange,
关11兴 describe a two-dimensional geochemical simulator,
etc. These processes cause morphological changes of the
CIRF.A, and its application to matrix acidizing analysis and
pore space and can lead to porosity and permeability de-
creases. design. The effects of diagenesis on the properties of sand-
Sahimi et al. 关3兴 reviewed models of fluid-solid interac- stone reservoirs were addressed by Le Gallo et al. 关12兴. They
tions in porous media. Shapiro and Brenner 关4兴 studied dis- developed a three-dimensional, two phase, reaction-transport
persion of a chemically reactive solute in spatially periodic simulator named DIAPHORE which couples kinetically con-
porous media; however, the geometry of the medium was trolled dissolution and precipitation of minerals with equilib-
assumed to be fixed. The precipitation of a reactive solute at rium reactions of chemical species in the water phase. In
the fluid-solid interface in porous media and the subsequent reservoir simulations, the ability of the core-scale data to
morphological changes of the pore space were discussed by represent the macroscopic properties of discretization blocks
Sallès et al. 关5兴. Matthews et al. 关6兴 simulated the subtle is not obvious since in real rocks the macroscopic parameters
changes in void space dimensions which occur during the can vary within a block. This variability must be taken into
artificial deposition of small amounts of illite and kaolinite account and the validity of the permeability-porosity rela-
within sandstones. They compared the mercury intrusion tionships must be verified.
curves of an untreated plug of sandstone and of a similar An appropriate averaging technique requires an investiga-
plug in which the deposit has been precipitated. tion of the deposition phenomenon on the intermediate scale
Deposition in fracture networks is not often addressed in which corresponds to the discretization block. The major
the literature. Mourzenko et al. 关7兴 and Békri et al. 关8兴 stud- purpose of this work is to implement this technique numeri-
ied deposition in a single fracture. They considered deter- cally and to study deposition and the subsequent changes of
ministic and random fractures and the changes in fracture the macroscopic properties in statistically homogeneous po-
morphology were taken into account. Deposition in fracture rous media. To describe such media, a discrete model was
networks is lacking at the moment and the properties have used with uniform porous subdomains possessing various
not been upscaled yet. physical properties. Statistically homogeneous materials are
Chang and Civan 关9兴 developed a model which can satis- viewed as being spatially periodic, i.e., as composed of iden-
factorily simulate the permeability alteration mechanisms tical heterogeneous cells.

1539-3755/2005/71共1兲/016311共19兲/$23.00 016311-1 ©2005 The American Physical Society

E. A. BORISOVA AND P. M. ADLER PHYSICAL REVIEW E 71, 016311 共2005兲

Section II provides a general overview on the deposition K

process of a chemical solute and on subsequent changes of v̄ = − · ⵱ P, 共2.1兲
␮
the macroscopic properties of the porous medium. We con-
sider a three-dimensional transport of the solute in the pres- where K is the permeability tensor, ␮ the fluid viscosity, and
ence of a first order irreversible reaction at a fluid/rock sur- P the pressure.
face. The coupling between the changes of macroscopic The pressure field may be obtained by solving the conti-
parameters and the reactive transport uses a quasisteady hy- nuity equation,
pothesis. In Sec. IV, this assumption is justified and the cor-
responding algorithm is described. The flow and transport ⵱ · v̄ = 0. 共2.2兲
equations are solved numerically by means of a finite volume
A nonzero macroscopic pressure gradient is imposed along a
technique; porosity changes are governed by a balance equa-
given direction over the unit cell,
tion. A brief discussion on the code performance is given at
the end of the section. ⵱ P = const. 共2.3兲
Section III provides the necessary constitutive equations
derived from previous studies on the local scale such as the This direction is arbitrarily chosen to be the x-axis. Note that
relation between porosity and permeability, the local disper- due to the deposition process, the permeability tensor in 共2.1兲
sion tensor and the local deposition rate. It gives also the depends on time.
initial porosity fields whose evolution will be studied. The
numerical algorithm is detailed in Sec. IV. An elementary B. Convection-dispersion
solution for the porosity evolution is derived; it will be sys-
tematically compared to the numerical results. The main di- On the macroscopic scale, the transport of a reactive sol-
mensionless quantities are defined in this section. ute is governed by a three-dimensional convection-
Section V presents the numerical results. Preliminary dispersion equation with reaction
verifications of the code on several configurations are de-
⳵ C␧
tailed. Then, three realizations of statistically equivalent me- + ⵱ · 共Cv̄兲 = ⵱ · 共d̄ · ⵱ C兲 − R, 共2.4兲
dia are generated and very close results are obtained for ⳵t
them. The influence of the unit cell size was studied by com-
paring the results for various ratios of the correlation length where C is the solute concentration, ␧ the porosity, d̄ the
and of the unit cell size. The influence of the discretization macroscopic dispersion tensor and R the reaction term. It is
has been studied; it appears that the size of the elementary assumed that on the macroscopic scale, dispersion is Fickian
cube is small enough that the discretization does not influ- and can be described by a dispersion tensor 关13兴. Moreover,
ence the results. The influence of the Damköhler number is d̄ depends on the local porosity and on the fluid velocity as it
studied; two regimes are shown to exist for small and high will be seen below.
Damköhler numbers. The main statistical parameters of the
initial porosity field are also briefly addressed. Finally, the C. Reaction and porosity changes
influence of the nature of the initial porosity field is studied.
A few concluding remarks end this paper. In the present work, an irreversible first order reaction is
considered. Note that the reaction occurs at the solid-fluid
II. GOVERNING EQUATIONS
interface on the pore scale. On the macroscopic scale, the
heterogeneous reaction is expressed by means of a sink term
Let us start by a general presentation of the problem; ini- R 关14兴 which can be modeled as

再 冎
tial configurations of the studied porous media are illustrated
in Fig. 1. Consider some chemical solute contained in under- ␥␧共C − C * 兲 if C ⬎ C * ,
R= 共2.5兲
ground water. While flowing through the porous medium, the 0 if C 艋 C * ,
solute precipitates onto the solid surface of the medium.
Fresh solute is supposed to be brought so that the mean where C* is the saturation concentration on the solid surface
volumetric liquid concentration remains constant all the and ␥ the volumetric reaction rate. ␥ is an unknown function
time. Moreover, we assume that the macroscopic gradient of which depends on deposition on the pore scale. It will be
concentration is negligibly small. This may correspond to a given later.
physical situation where the region of interest is located suf- Due to deposition, porosity decreases with time as de-
ficiently far away from injection/production wells. Deposi- scribed by the balance equation
tion implies porosity, permeability and other macroscopic pa-
⳵
rameter changes. After a very long time, if fresh solute is f 共1 − ␧兲 = R, 共2.6兲
constantly brought by the fluid, the permeability of the me- ⳵t
dium will tend towards zero.
where f is the solid fraction in the deposit 关5兴; for instance,
for a random packing of monodisperse spheres, f is equal to
A. Flow
0.64.
The seepage velocity v̄ inside the porous domain is gov- It is generally assumed in this paper that the amount of
erned by the Darcy equation, precipitating solute is unlimited and that the porosity can

016311-2
DEPOSITION IN POROUS MEDIA AND CLOGGING ON … PHYSICAL REVIEW E 71, 016311 共2005兲

FIG. 1. Examples of initial porous media. Porosity is given by the grey scale 共see the vertical bars兲. Examples are for 共a兲 a lognormal
porosity field with a Gaussian correlation function 共which is studied at length in Sec. V A兲; 共b兲 a uniformly distributed porosity field; 共c兲 a
Gaussian porosity field; 共d兲 a self-affine porosity field. For the three cases 共b兲, 共c兲, and 共d兲, 具␧0典 = 0.2, ␴0 = 0.2, and ␭ / Lc = 8 / 30.

decrease until the porous medium is clogged. This hypoth- ity ␥; note that it has the dimension of the inverse of a time.
esis will be discussed at the end of Sec. V A 2. The average liquid concentration C0 is kept constant with
time and it may be expressed as
D. Dimensionless formulation
In order to derive the dimensionless parameters which
control the problem, the governing equations should be ex-
冕
⍀
C␧d3r
共2.8兲
冕
pressed in dimensionless form. Let us define the dimension- C0 = ,
less variables ␧d3r
⍀
r d̄ C − C*
r⬘ = , d̄⬘ = , t⬘ = ␥0t, C⬘ = ,
L ␥ 0L 2 C0 − C* where ⍀ is the whole domain volume.
Let us define the Damköhler number by the standard re-
v ␥ f lation
v⬘ = , ␥⬘ = , f⬘ = , ⵜ⬘ = L ⵜ ,
V* ␥0 C0 − C*
共2.7兲 ␥ 0L
Da = . 共2.9兲
where L , V* are some characteristic values of length and V*
velocity; it seems natural that L is chosen to be the field
scale. Accordingly, ␥0 is a characteristic value of the reactiv- With the help of 共2.7兲 and 共2.9兲, 共2.4兲 can be written as

016311-3
E. A. BORISOVA AND P. M. ADLER PHYSICAL REVIEW E 71, 016311 共2005兲

⳵ ␧C⬘ 1 quantities. For Peloc 苸 关0.1, 1兴, the values of the coefficients
+ ⵱⬘ · 共C⬘v⬘兲 − ⵱⬘ · 共d⬘ · ⵱⬘C⬘兲 = − ␥⬘␧C⬘ . are of the order of
⳵ t⬘ Da
共2.10兲 ␤储 = 0.175, ␣储 = 0.305, ␤⬜ = 0.127, ␣⬜ = 0.159.
The Damköhler number compares the reaction rate to con- 共3.7a兲
vection. Alternate choices are of course possible for the units For Peloc 苸 关1 , 105兴, they are of the order of
and the dimensionless numbers.
␤储 = 0.175, ␣储 = 1.285, ␤⬜ = 0.127, ␣⬜ = 0.80.
III. THE CONSTITUTIVE EQUATIONS AND THE
POROSITY FIELD 共3.7b兲
A. The constitutive equations Since 共3.4兲 is valid in a local coordinate system, one must
make coordinate changes depending on the velocity orienta-
Constitutive parameters such as the permeability K and tions at each point where the dispersion tensor is needed.
the macroscopic dispersion tensor d̄ depend on the pore ge- The last quantity of interest is the volumetric reaction rate
ometry of the medium. Their determination is a local scale ␥ 共cf. Sec. II C兲. It can be estimated in a very crude way
problem which involves the analysis of the microstructure of from the knowledge of the surface reaction rate constant k p
the porous medium. In this study, a simple model is used for which does not depend on the morphology of the pore space.
K, This can be done as follows. The hydraulic radius m is usu-
ally defined as the ratio of the fluid volume V f to the pore
K = K0I␧␩ , 共3.1兲
surface area S p,
where K0 is a characteristic permeability and I the unit ten-
sor. The exponent ␩ is generally taken equal to 4 关15–17兴. Vf
m= . 共3.8兲
The dispersion tensor can be approximated as a sum of Sp
the macroscopic diffusion tensor d̄0 and of the tensor d̂* V f and S p are relative to a total volume V. The reaction term
which depends on the local Péclet number, R is proportional to the pore surface area per unit volume
d̄ = d̄0 + ␧Dmd̂ * , 共3.2兲 Sp
R = kp 共C − C * 兲 = ␥␧共C − C * 兲. 共3.9兲
where Dm is the molecular fluid diffusion. The macroscopic V
diffusion coefficient will be written as Hence, the effective volumetric reaction rate constant is
d̄0 = DmI␧2 , 共3.3兲 kp
␥= . 共3.10兲
which corresponds to the relation obtained by Coelho et al. m
关17兴 for random packings of spheres in agreement with Ar- The hydraulic radius m decreases during the deposition
chie’s law with a cementation exponent equal to 2. In a co- process. It is difficult to predict the evolution of this param-
ordinate system where the x-axis is parallel to the Darcy eter. There are several useful relations described in the litera-
velocity v̄, the tensor d̂* has a diagonal form ture for the hydraulic radius. Usually, m depends on the mi-

冢 冣
croscopic geometry of the medium. We suppose that locally
d储 0 0
the deposition process is limited by the surface reaction ki-
d̂ * = 0 d⬜ 0 , 共3.4兲 netics. In that case, a uniform deposition is expected 关7兴.
0 0 d⬜ Since the medium is supposed to be granular, the grains are
coated by a uniform layer of deposited matter. For random
when the porous medium is locally isotropic; d储 is the longi- packings of monodisperse spheres, m can be expressed as
tudinal dispersion and d⬜ the transversal dispersion. They
can be estimated as power functions of the local Péclet num- m共t兲 = mc␧共t兲3/4 , 共3.11兲
ber Peloc,
where mc is some constant. This relation roughly describes
␣储 ␣⬜
d储 = ␤储Peloc , d⬜ = ␤⬜Peloc . 共3.5兲 the hydraulic radius variations for random packings of
monodisperse spheres when the initial porosity is about 0.4.
The local Péclet number is defined as It was shown that the specific surface area of consolidated
V*l media is only slightly affected by a slow deposition process
Peloc = , 共3.6兲 关7,12兴. For small porosities, the hydraulic radius is taken to
Dm
be constant. Note that the relations presented above are only
where l is some local characteristic length. l can be chosen meant to provide reasonable orders of magnitude.
equal to the correlation length of the pore space. In this Since, on the pore scale, the deposition process is as-
work, l is taken equal to 0.5 mm. The molecular diffusion Dm sumed to be reaction-limited, it is useful to provide an ex-
of a reactive solute in water phase is of the order of 10−9 m. pression for the relationship between the Damköhler num-
The approximation of the results obtained by Coelho et al. bers Dap and Da on the pore scale and on the field scale,
关17兴 for random packings of spheres yields the following respectively. Dap is expressed as

016311-4
DEPOSITION IN POROUS MEDIA AND CLOGGING ON … PHYSICAL REVIEW E 71, 016311 共2005兲

kp Ĉ共k兲 ⬀ 关1 + 共␭xkx兲2 + 共␭yky兲2 + 共␭zkz兲2兴−H/2 , 共3.19兲

Dap = , 共3.12兲
vp
where H is the Hurst exponent. The Gaussian correlated field
where v p is the local interstitial velocity. Thus, Dap and Da is generated by means of the Fourier transform method
are related by which is described by Adler and Thovert 关18兴.
The fields generated by these two correlations are very
L vp
Da = Dap . 共3.13兲 different and will be illustrated in Sec. V.
具m0典 V*
Since on the local scale deposition is assumed to be reaction-
IV. PRELIMINARIES TO THE NUMERICAL STUDY
limited, one has
A. Numerical solution
Dap Ⰶ 1. 共3.14兲
The problem is solved in a parallelepipedic cell dis-
Hence, 共3.13兲 shows that because of the length scale ratio, cretized into Ncx ⫻ Ncy ⫻ Ncz elementary cubes. Each elemen-
Da can be very large while Dap is very small. tary cube has given porosity, diffusion, and permeability ten-
sors. The continuity equation, the reactive transport equation,
B. Description of the initial porosity field and the equation for porosity changes were discretized using
a finite volume method. The pressure and concentrations are
The porosity field is generated numerically as a random
evaluated at the cube corners. The pressure field was ob-
correlated or noncorrelated field. Usually, local properties of
tained by using a box integration technique.
the real porous media are described by lognormal distribu-
To solve the transport equations, one is usually interested
tions 关20兴. In this paper, uniform and Gaussian distributions
in a shock-capturing numerical model, which does not pro-
were also used.
duce spurious oscillations near discontinuities and yet re-
The uniform distribution corresponds to porosity values
mains stable. These properties are quite important for simu-
uniformly distributed on some interval, for example, between
lating convection-dispersion phenomena, especially for
0 and 0.4.
situations where convection is dominant. Therefore, our
The probability density of the Gaussian distribution is ex-
main effort was to reduce numerical dispersion. One possible
pressed as
way to achieve it is to apply an operator-splitting technique.

f共x兲 =
1
冑2␲␴G 2
exp −冉 1 共x − ␮G兲2
2 ␴G 2 , 冊 共3.15兲
The idea is to solve the convection part up to a certain time
step ␦t and then to add the dispersive effects in a separate
step. The second order flux limiting scheme was applied to
where ␮G is the mean and ␴G 2
the variance. solve the convection equation 关19兴. For this method, the best
Another class of random distributions is the lognormal accuracy is obtained when the time step is chosen as large as
distribution. If y G is a Gaussian random variable with mean the stability criterion for the convective part allows. Operator
␮G and variance ␴G2 , the variable y LN = exp共y G兲 is lognor- splitting has the advantage that the time step for the convec-
mally distributed. Its mean and variance are given by tion part may be optimized to reduce numerical dispersion.
Then, the dispersive part may be solved implicitly or exlic-
␮LN = exp共␮G + 21 ␴G2 兲, ␴LN
2
= exp共2␮G + 2␴G
2
兲 − ␮LN
2
. itly. Here, the dispersion-reaction part was addressed by a
classical explicit scheme. When the convective shift for ␦t is
共3.16兲 calculated, the dispersion equation is solved n times for time
The spatial organization of the porosity field also needs to step ␦ts = ␦t / n. Usually, convective transport is predominant
be defined. Two major types of structures can be easily in underground formations. Thus, the restrictions on the con-
implemented by means of the correlation function vection time step are more severe than that for dispersion and
n = 1.
具关␧共x + u兲 − ␮␧兴关␧共x兲 − ␮␧兴典 The code was tested on several cases where the analytical
C共u兲 = , 共3.17兲
具共␧共x兲 − ␮␧兲2典 solution is available. A special attention was put on checking
the extreme situations, such as pure convection, convection-
where ␮␧ = 具␧典. dispersion, diffusion-reaction, and convection-reaction. Usu-
The porosity field is said to be Gaussian when its corre- ally, a good agreement between numerical results and ana-
lation C is Gaussian, lytical predictions is observed.

冋 冉 冊 冉 冊 冉 冊册
2 2 2
In Sec. II, the governing equations are introduced into the
ux uy uz framework of a continuous description. To solve these equa-
C共u兲 = ␴G
2
exp − − − , 共3.18兲
␭x ␭y ␭z tions numerically, a quasisteady hypothesis is assumed. The
idea is the following. As long as the porosity changes are not
where ␭x , ␭y , ␭z are the correlation lengths along the x , y, significant, i.e., they do not exceed a value ⌬␧max prescribed
and z directions, respectively. u is the shift with components in advance, we suppose that the macroscopic parameters and
共ux , uy , uz兲. Let us denote by Ĉ共k兲 the Fourier transform of local velocities do not change. Thus, the conservation equa-
the correlation function C共u兲. Self-affine porosity fields are tion for pressure is solved with constant parameters, i.e., for
defined by a fixed situation. Then, asymptotically the concentration field

016311-5
E. A. BORISOVA AND P. M. ADLER PHYSICAL REVIEW E 71, 016311 共2005兲

is expected to decay exponentially with time, but uniformly

in space 关4兴. This corresponds to

C⬘共r⬘,t⬘兲 = Ĉ⬘共r⬘兲exp共− Bt⬘兲, 共4.1兲

where B is some decay rate.
Hence, when this exponential regime takes place, the fol-
lowing ratio over a constant time step ␦t⬘ becomes constant:

冕 ⍀
C⬘共r⬘,t⬘兲d3r⬘
= exp共B␦t⬘兲. 共4.2兲
冕 ⍀
C⬘共r⬘,t⬘ − ␦t⬘兲d r⬘ 3

Thus, denote

冕
冢冕 冣
C⬘共r⬘,t⬘兲d3r⬘
⍀
␣共t⬘兲 = ln . 共4.3兲
C⬘共r⬘,t⬘ − ␦t⬘兲d r⬘
3
⍀

The convergence test consists in checking the absolute value

of the time derivative of ␣共t⬘兲. We assume that the time
necessary to reach such a regime is very small compared to
the deposition time. Hence, the transport equations are iter-
ated until the concentration field reaches the exponential re-
gime.
Then, the porosity changes can be easily obtained as func-
tions of time by solving Eq. 共2.6兲. Thus, it is possible to
calculate the time ␶ during which the maximal porosity FIG. 2. Schematic representation of the algorithm.
change ⌬␧i exceeds the prescribed threshold ⌬␧max in any
elementary cube i, ⳵ ␥⬘
共1 − ␧兲 = ␧C⬘ . 共4.5兲
␶ = 兵␦t⬘:␧i共t⬘0兲 − ␧i共t0⬘ + ␦t⬘兲 ⬇ ⌬␧max其. 共4.4兲
⳵ t⬘ f⬘
Moreover, this relation can be averaged over the cell and it
The new values of the porosity field and of the other
yields an exponential decay for the average porosity 具␧典,
macroscopic parameters are calculated according to the
amount of matter deposited during ␶ in each elementary
cube. Then, the calculations are iterated with these new pa-
rameters.
冉
具␧典 = 具␧0典exp − t⬘
␥⬘
f⬘
冊
C⬘ . 共4.6兲

The program is stopped when the macroscopic permeabil- This elementary solution will prove to be very useful.
ity of the porous medium is 103 times smaller than the initial
one. Some additional comments on this point can be found at C. Characteristic quantities, output quantities, and
the end of Sec. V A 2. A schematic representation of the dimensionless parameters
algorithm is given in Fig. 2. The choice of the characteristic parameters is very impor-
Several assumptions used in this problem need to be con- tant since they define the dimensionless parameters control-
firmed by quantitative comparisons with experimental data, ling the process. The characteristic length L is chosen equal
but to the best of our knowledge, no experimental work is to the correlation length ␭ of the porosity field. The charac-
available on this subject. Such a verification is especially teristic volumetric reactivity constant is chosen to be equal to
necessary when the deposition rate is large, with rapid
changes of the geometry of the pore space; in such a case, kp
␥0 = , 共4.7兲
the concentration profile may never reach the exponential 具m0典
regime.
where 具m0典 is the average hydraulic radius at the initial time.
Let us now give the major output quantities. At the field
B. An elementary solution
scale, one is usually interested in the macroscopic properties
An elementary solution can be derived from 共2.6兲 when of the medium. Several macroscopic quantities were system-
the concentration C⬘ and the reactivity ␥⬘ are assumed to be atically measured. The first one is the average macroscopic
constant. Obviously, porosity 具␧典,

016311-6
DEPOSITION IN POROUS MEDIA AND CLOGGING ON … PHYSICAL REVIEW E 71, 016311 共2005兲

FIG. 3. Porosity of a bilayered porous medium for cases A and

B. Flow is parallel to the x-axis. The white and grey zones corre- FIG. 4. Numerical results in a bilayered medium for cases A and
spond to ␧01 = 0.2 and ␧02 = 0.4, respectively. B denoted by 䊊 and ⴱ, respectively. Flow is parallel to the x-axis.
The solid lines represent the analytical solutions.

具␧典t⬘
冕
⍀⬘
␧共t⬘兲d3r⬘ the main pressure gradient and R⬜共v兲 is calculated perpen-
共4.8兲 dicular to the main pressure gradient
冕
= ,
具␧0典
␧共t⬘0兲d3r⬘ 具关␧共x + u兲 − ␮␧兴关␧共x兲 − ␮␧兴典
⍀⬘ R储共u兲 = , 共4.11a兲
具关␧共x兲 − ␮␧兴2典
where ⍀⬘ is the unit cell. The macroscopic permeability is
determined by means of the Darcy law 具关␧共x + v兲 − ␮␧兴关␧共x兲 − ␮␧兴典
R⬜共v兲 = , 共4.11b兲
具关␧共x兲 − ␮␧兴2典
Kt V⬘t⬘ where u and v are spatial shifts parallel and perpendicular to
⬀ , 共4.9兲
K0 V⬘ ⵱P, respectively.
t⬘ 0
It is obviously shown by the previous developments that
the dimensionless parameters are numerous, namely, the ar-
where V⬘t⬘ is the macroscopic seepage velocity at time t⬘.
tificial parameters which are only useful in the course of the
During the simulations, the macroscopic pressure gradient numerical calculations, but which do not possess any physi-
ⵜP is kept constant. cal meaning such as the size of the unit cell Nc and the
The total amount of deposit per unit volume as a function porosity step ⌬␧max, the basic physical parameter Da, the
of time t⬘ is given by parameters which characterize the initial porosity distribu-

冕 ⬘冕
tion.
t⬘ 1
Qdep共t⬘兲 = ␥⬘␧C⬘d3r⬘dt⬘ . 共4.10兲
t0⬘ ⍀ ⍀⬘ V. RESULTS AND DISCUSSION

Two autocorrelation functions will be used to characterize In this work, the evolution of the macroscopic parameters
the evolution of the porosity field. R储共u兲 is calculated along with time is taken into account by means of a quasisteady

016311-7
E. A. BORISOVA AND P. M. ADLER PHYSICAL REVIEW E 71, 016311 共2005兲

FIG. 5. Evolution of a porous medium with an initial lognormal porosity distribution. Each line corresponds to a different time. The left
row is the probability density g共ln ␧兲 of ln共␧共r⬘ , t⬘兲兲; the central one is a cross section of the porosity field parallel to the pressure gradient;
the right one is perpendicular to the pressure gradient. Da= 8 ⫻ 10−6. The porosity convention is given by the vertical bars.

016311-8
DEPOSITION IN POROUS MEDIA AND CLOGGING ON … PHYSICAL REVIEW E 71, 016311 共2005兲

具␧A典 = 具␧B典 = 0.5共␧1 + ␧2兲, 共5.1b兲

where ␧i is the porosity of the ith layer. The calculated mean
porosity values are compared with the analytical solutions in
Fig. 4共a兲. For both cases, the numerical results match the
analytical curves very satisfactorily. Small differences are
only observed at large times when porosity is low. Note that
in the calculations, the hydraulic radius was assumed to be
constant for low porosity values; this simplification explains
the small differences between numerical and analytical re-
sults in this case.
For case A, the macroscopic permeability KA is the sum of
the permeabilities of the layers KA = K1 + K2. For case B, the
macroscopic permeability KB = 共K−1 1 + K2 兲 . Remind that the
−1 −1

permeability of each layer Ki is given by 共3.1兲. Numerically,

the macroscopic permeability is obtained by means of the
Darcy equation. The calculated values are compared to the
analytical solution in Fig. 4共b兲. The agreement is perfect for
both cases.

2. Illustrative case for random media

Statistically homogeneous media are considered as three-
dimensional spatially periodic media, composed of identical
unit cells of size Lc = 30 m. The cell is discretized into N3c
= 30⫻ 30⫻ 30 elementary cubes. Since the lognormal distri-
bution is the most common for porous media 关20兴, system-
atic calculations have been performed for log-normally dis-
tributed initial porosity fields with a Gaussian correlation
function. The initial porosity is supposed to be 具␧0典 = 0.2 with
a standard deviation ␴0 = 0.2. The correlation length ␭ is
FIG. 6. Evolution of the correlation functions R储共u兲 共a兲 and equal to 8 m. Because of the finite size of the unit cell, it is
R⬜共u兲 共b兲 for the illustrative case. The dimensionless length u is useful to define the ratio ␭ / Lc which compares the correla-
normalized by the half-size of the unit cell. Da= 8 ⫻ 10−6. tion length to the size of the cell. In this case, ␭ / Lc = 8 / 30.
This particular porous medium is displayed in Fig. 1共a兲.
approximation. A test presented in the following section is The Damköhler number is chosen to be based on the cor-
designed to study the influence of the discontinuities intro- relation length ␭ 共cf. Sec. IV C兲. We assumed Da= 8 ⫻ 10−6.
duced into the code due to the steady state approach. Other The presence of 8 in the Damköhler number is not arbitrary.
verifications are done in order to study the effects of the In fact in numerical simulations, it is more convenient to use
discretization of the porous medium and of the statistical the Damköhler number Danum based on the size of the el-
fluctuations on the output results. ementary cube a. Danum is a purely numerical parameter
which is used in the course of the numerical calculations
since its application provides the dimensionless discretiza-
A. Preliminaries tion steps ⌬r⬘ equal to 1. The corresponding Da is easily
obtained from the relation
1. Simple deterministic media
Consider a two-dimensional porous medium which con- ␭
Da = Danum , 共5.2兲
sists of two homogeneous layers with initial porosities ␧01 a
= 0.2 and ␧02 = 0.4. We suppose that the macroscopic pressure
gradient is imposed along the x direction and is kept constant which means that Danum = 10−6 yields Da= 8 ⫻ 10−6.
with time. The layers can be either parallel or perpendicular Before running the code, it is necessary to determine an
to the x-axis. These two situations will be referred to as cases adequate porosity threshold ⌬␧max for the steady state algo-
A 关Fig. 3共a兲兴 and B 关Fig. 3共b兲兴, respectively. For both cases, rithm described in Sec. IV A. Some numerical tests showed
the evolution of the macroscopic parameters can be predicted that as long as ⌬␧max 艋 10−2␧0, the various curves collapse
analytically. In such systems, the liquid concentration C⬘ is onto a single curve. Hence, ⌬␧max was chosen equal to
expected to be uniform in space. Thus, it is equal to its av- 10−2␧0 in the rest of the simulations. The calculations are run
erage value C⬘ = 1 which is kept constant with time. Then, as long as the porous medium is sufficiently permeable. As it

冉 冊
was mentioned before, we suppose that the porous medium
3 具m0典 4/3
is clogged if its macroscopic permeability is 103 times
␧i = − t⬘ + 共␧0i 兲3/4 , i = 1,2, 共5.1a兲
4 f⬘ smaller than the initial one.

016311-9
E. A. BORISOVA AND P. M. ADLER PHYSICAL REVIEW E 71, 016311 共2005兲

FIG. 7. Macroscopic parameters for various Damköhler numbers for the illustrative case. Data are for Da= 8 ⫻ 10−6 共䊊兲, 0.8 共䉯兲, 8
⫻ 10−2 共ⴱ兲, 8 共〫兲, and 80 共䊐兲. The solid lines represent least-square fits.

The output quantities include the porosity field and the IV B兲 derived for the situation where the liquid concentration
macroscopic parameters as functions of time. Some probabil- and the hydraulic radius are constant. Note that the variations
ity densities and cross sections of the porosity field for suc- of the hydraulic radius implemented in the code are small.
cessive times are shown in Fig. 5. The left cross section is The evolution of the mean porosity as a function of time
taken at the center of the domain parallel to ⵱P and the right is shown in Fig. 7共a兲. Since the curve is linear in a semilog
one perpendicular to ⵱P; in Fig. 5, they are denoted by the plot, the mean porosity is an exponential function of time.
subscripts 储 and ⬜, respectively. This set of three pictures is The mean square approximation gives
plotted for successive times until clogging. These sections
show that porosity changes smoothly. It appears that the re- 具␧典
⬇ exp共− 1.72 t⬘兲. 共5.3兲
gions with low porosity are clogged first. At clogging, re- 具␧0典
gions with large porosity still exist.
The probability densities g共ln ␧兲 are displayed in Fig. 5. The calculated mean porosity is quite close to the elementary
Hence, for the initial lognormally distributed porosity field, solution 共4.6兲. Another important macroscopic parameter is
g共ln ␧兲 is Gaussian. Since lognormally distributed values the total amount of deposit per unit volume Qdep. Figure 7共b兲
vary from zero to infinity, values of porosity larger than 1 gives Qdep as a function of mean porosity. A least-square
were truncated; due to this truncation, some elementary approximation yields
cubes have a porosity equal to 1 and they correspond to the Qdep ⬇ 具␧0典 − 具␧典. 共5.4兲
vertical bars in Fig. 5. During deposition, the probability
density shape gets flat and shifts to the left. At each time, the total amount of deposited matter is propor-
The evolution of the correlation functions is shown in Fig. tional to the change in mean porosity during the correspond-
6. Deposition only slightly affects the correlation function of ing time interval. The evolution of the macroscopic perme-
the porosity field. This uniform reduction of porosity is ex- ability with time and its relation to the mean porosity is a
pected when the liquid concentration does not vary much. In question of particular interest and importance. Figures 7共c兲
this case, it is interesting to compare the obtained macro- and 7共d兲 show the macroscopic permeability as a function of
scopic parameters with the elementary solution 共cf. Sec. time and of the mean porosity, respectively; the variations of

016311-10
DEPOSITION IN POROUS MEDIA AND CLOGGING ON … PHYSICAL REVIEW E 71, 016311 共2005兲

FIG. 8. Evolution of a porous medium with an initial lognormal porosity distribution. Each line corresponds to a different time. The left
row is the probability density g共ln ␧兲 of ln共␧共r⬘ , t⬘兲兲; the central one is a cross section of the porosity field parallel to the pressure gradient;
the right one is perpendicular to the pressure gradient. Da= 8. The porosity convention is given by the vertical bars.

016311-11
E. A. BORISOVA AND P. M. ADLER PHYSICAL REVIEW E 71, 016311 共2005兲

TABLE I. General analysis of the data. Values denoted by ⴱ correspond to simulations where the clogging condition was not obtained due
to numerical difficulties; thus, the time T⬘c corresponds to the last calculated one.

具␧典
⬇ 兺n=0
3
a nt ⬘n
具␧0典

N Initial distribution Da 具␧0典 ␴0 ␭ / Lc a3 a2 a1 a0

K
K0
⬇ exp
共 兲
具␧典 ␩
具␧0典
,␩ Tc⬘ Qdep共Tc⬘兲

1 Lognormal 8 ⫻ 10−6 0.2 0.2 8 / 30 −0.04 0.66 −1.44 1 3.55 1.02 0.160
2 8 ⫻ 10−2
3 0.8 −0.11 0.75 −1.47 1 3.41 1.29 0.167
4 8 −0.86 1.27 −1.29 1 4.18 0.94 0.154
5 80 0.12 −0.47 −0.61 1 4.28 0.81* 0.141*
6 8 ⫻ 10−6 0.4 0.2 0.39 0.18 −1.48 1 3.48 0.78 0.338
7 0.2 0.1 0.38 0.19 −1.48 1 3.55 0.78 0.169
8 0.2 0.3 −0.04 0.66 −1.44 1 3.77 1.18 0.146
9 8 0.4 0.2 −0.86 0.99 −1.44 1 4.18 0.70 0.318
10 0.2 0.1 −0.90 0.99 −1.41 1 4.16 0.71 0.160
11 0.2 0.3 −0.86 1.27 −1.30 1 4.00 1.24 0.148
12 Uniform 8 ⫻ 10−6 0.2 0.115 0 0.04 0.47 −1.43 1 3.20 0.91 0.178
13 Gaussian 0.2 8 / 30 0.04 0.33 −1.22 1 3.58 1.02 0.180
14 Self-affine −0.05 0.33 −1.22 1 3.35 1.06 0.186
15 Uniform 8 0.2 0.115 0 −0.01 0.53 −1.43 1 3.35 0.75 0.156
16 Gaussian 0.2 8 / 30 −0.39 0.69 −1.17 1 4.03 0.95 0.174
17 Self-affine −0.07 0.50 −1.24 1 3.44 0.78 0.151
18 Lognormal 8 ⫻ 10−6 0.2 0.2 8 / 20 −0.02 0.14 −0.37 1.25 3.50 1.02 0.162
19 8 / 40 −0.05 0.44 −1.28 1.88 3.80 0.95 0.155
20 8 8 / 20 −0.18 1.55 −4.00 3.63 4.42 0.93 0.156
22 8 / 40 0.17 −0.17 −1.21 2.21 4.10 0.65* 0.103*

␧ are close to a power law which can be approximated as = 0.8, suppose that Qdep is at most equal to 0.1; Fig. 7共b兲
K
K0
⬇冉 冊
具␧典
具␧0典
␩
, 共5.5兲
implies that t⬘c ⬇ 0.5. Therefore, for t⬘ ⬎ t⬘c , Qdep is equal
to 0.1.

with ␩ = 3.55. Note that locally an analogous power-law

3. Statistical fluctuations
permeability-porosity relation was used with an exponent
equal to 4 关see Eq. 共3.1兲兴. The elementary solution is ob-
tained by applying this relation to the macroscopic perme- Since the initial configurations are produced at random,
ability as a function of 具␧典, one should average over many of them in order to obtain

冉 冊
meaningful statistical quantities. However, such an approach
Ke 具␧典e 4
is very expensive in terms of computer time. Instead, we
⬇ , 共5.6兲
K0 具␧0典 provide here an analysis of the output results for different
where the index e denotes the elementary solution. The cal- realizations of the initial random porosity field.
culated macroscopic permeability is very close to this el- In addition to the illustrative case, two groups of calcula-
ementary solution. tions were done for the same set of parameters. The porosity
When deposition begins, the porosity decreases according field is lognormal; the mean porosity 具␧0典 is equal to 0.2, the
to the elementary solution which is derived for a constant standard deviation ␴0 to 0.2 and the ratio ␭ / Lc is equal to
liquid concentration. Then, the rate of the porosity reduction 8 / 30. The calculations were done for Da= 8 ⫻ 10−6 and Da
accelerates relatively to the elementary solution. = 8. The agreement between the realizations is very good for
As mentioned in Sec. II C, the amount of precipitating both regimes, except may be in the period close to clogging.
solute can be limited. Such a limitation would imply that the Therefore, the calculated macroscopic quantities are not sig-
proposed curves are valid for a time t⬘ smaller than a certain nificantly influenced by the statistical fluctuations. Thus, the
critical time t⬘c ; for larger times, deposition stops because tedious averaging procedure described above seems to be
there is nothing to deposit anymore. For instance, for Da useless.

016311-12
DEPOSITION IN POROUS MEDIA AND CLOGGING ON … PHYSICAL REVIEW E 71, 016311 共2005兲

sented tests have shown that the ratio ␭ / Lc = 8 / 30 would be

an optimal choice.

5. Influence of the discretization

It should be reminded that the porosity field is discretized.
In principle, the size of the elementary cubes should be very
small compared to the correlation length ␭ in the practically
important case of correlated media. The test included three
calculations for porous media discretized into 120
⫻ 120, 60⫻ 60, and 30⫻ 30 elementary cubes. In all three
cases, the porosity field is lognormally distributed with 具␧典
= 0.4, ␴ = 0.2, and ␭ / Lc = 8 / 30. The calculations were per-
formed for Da= 32⫻ 10−6 and Da= 32.
For small Damköhler number, the results are not influ-
enced by the discretization. For large Damköhler number,
only the macroscopic permeability for Nc = 120 is slightly
different at the beginning of the deposition. Thus, discretiza-
tion is shown to play a very small role.

B. Influence of the Damköhler number

As it was mentioned before, the basic physical parameter
is the Damköhler number which compares the reaction rate
to convection. For the small Damköhler number equal to 8
⫻ 10−6 considered in the illustrative case, convection is very
large compared to reaction. As it was shown in this case,
deposition causes smooth changes to the porosity field. The
evolution of the mean porosity can be predicted by the el-
ementary solution derived for a situation where the concen-
tration C⬘ and the reactivity ␥⬘ are constant. Another behav-
FIG. 9. Evolution of the correlation functions R储共u兲 共a兲 and ior is expected when the Damköhler number is high. The
R⬜共u兲 共b兲. The dimensionless length u is normalized by the half- domination of reaction over convection can cause significant
size of the unit cell. Da= 8. variations of the concentration field and affect the macro-
scopic properties. In order to investigate the role of the
4. The influence of the size of the unit cell Damköhler number, calculations were done for Da= 8
⫻ 10−2, 0.8, 8, and 80. The initial porosity field corresponds
The size Lc of the unit cell is one of the artificial param- to the one used in the illustrative case. Note that the
eters introduced in the simulations. For correlated media, it is Damköhler number is based on initial quantities such as the
usually supposed that Lc is large enough so that the unit cell macroscopic velocity V* and the reactivity ␥0. Both will vary
can contain many “independent” samples of the porous me- during deposition and the actual Damköhler number will in-
dium. If this condition is not satisfied, the discrete model crease.
fails to represent the random character of a real medium. The macroscopic parameters for the various Da are com-
Thus, it is likely that an improper choice of the cell size can pared to the illustrative case in Fig. 7. The curves for Da
cause significant errors in the estimation of the macroscopic = 8 ⫻ 10−2 collapse onto the curves of the illustrative case
parameters of the random medium. The ratio ␭ / Lc was used where Da= 8 ⫻ 10−6 关Fig. 7共a兲兴. On the contrary, for Da= 8
to characterize the cell size. In the illustrative case, ␭ / Lc is and 80, the behavior of the mean porosity differs consider-
equal to 8 / 30. Two more calculations were done for ␭ / Lc ably from the illustrative case. The mean porosity curve is
equal to 8 / 20 and 8 / 40. not exponential anymore. At the beginning of deposition, the
For all cases, the mean porosity value 具␧典 is equal to 0.2 mean porosity diminishes more slowly than for the illustra-
and the standard deviation ␴0 = 0.2. For all calculations, we tive case and after a certain time it decreases sharply. Clog-
kept the ratio ␭ / a constant and equal to 8 as for the illustra- ging occurs earlier than for the illustrative case. For Da
tive case. In order to change the ratio ␭ / Lc the size of the = 0.8, the mean porosity values match well the results for the
unit cell was varied. The mean porosity value and the total illustrative case except for the times close to clogging. Clog-
amount of deposit were shown not to be influenced by ␭ / Lc. ging occurs a bit later than for the illustrative case.
On the contrary, the macroscopic permeability is a bit sensi- The mean porosity evolution can be approximated by a
tive to the variations of ␭ / Lc especially for small Damköhler polynomial of degree 3 in all cases 共Table I兲. The evolution
numbers. Note that correct results are only obtained for suf- of the total amount of deposit per unit volume is proportional
ficiently large unit cells, but the computational time increases to the mean porosity change. Qdep is shown as a function of
significantly with the increase of the domain size. The pre- time in Fig. 7共b兲. Finally, at clogging, the total amount of

016311-13
E. A. BORISOVA AND P. M. ADLER PHYSICAL REVIEW E 71, 016311 共2005兲

FIG. 10. Evolution of a porous medium with an initial lognormal porosity distribution. Each line corresponds to a different time. The left
row is the probability density g共ln ␧兲 of ln共␧共r⬘ , t⬘兲兲; the central one is a cross section of the porosity field parallel to the pressure gradient;
the right one is perpendicular to the pressure gradient. Da= 80. The porosity convention is given by the vertical bars.

016311-14
DEPOSITION IN POROUS MEDIA AND CLOGGING ON … PHYSICAL REVIEW E 71, 016311 共2005兲

ability density g共ln ␧兲 is less spread than for small

Damköhler numbers. There are no regions with high poros-
ity. Thus, the situations where Da⬇ 1 can be considered as
intermediate between the regime where convection is domi-
nant Da⬍ 1 and the regime where reaction is dominant
Da⬎ 1.
It is necessary to add that the clogging time Tc⬘ has no
particular importance since clogging has been somewhat ar-
bitrarily defined. Hence, attention should be focused on the
evolution of the macroscopic parameters with time.
In the following, each calculation will be done for a small
and a large Damköhler number, since the system is expected
to behave differently in these conditions. More precisely, we
mean the evolution of the probability density g共ln ␧兲 and the
relation between the macroscopic permeability and the mean
porosity. As for the mean porosity evolution, it will be shown
later that the exponential decrease for small Damköhler num-
bers only occurs for lognormally distributed porosity fields.
Polynomial approximations of the mean porosity values pro-
vide a very good fit for all the results described in this work.
Note that they are only valid for times t⬘ 苸 关0 , T⬘c 兴. The data
are gathered in Table I.
C. Influence of the parameters characterizing the initial
porosity distribution
This section studies the influence of the mean 具␧0典 and of
the standard deviation ␴0 of the initial porosity field on the
evolution of the macroscopic parameters during deposition.
The initial porosity field is assumed to be lognormally dis-
tributed. The ratio ␭ / Lc is equal to 8 / 30.
The first test addresses the effect of the mean porosity
FIG. 11. Evolution of the correlation functions R储共u兲 共a兲 and value. 具␧0典 is taken to be equal to 0.4 and ␴0 = 0.2. The sec-
R⬜共u兲 共b兲. The dimensionless length u is normalized by the half- ond and third calculations are done for 具␧0典 = 0.2 and ␴0
size of the unit cell. Da= 80. equal to 0.1 and 0.3. The calculations were performed for
Da= 8 ⫻ 10−6 共Fig. 14兲 and Da= 8. The calculated mean po-
deposit per unit volume is nearly the same for all cases. The rosity values are compared to the analytical solution derived
macroscopic permeability as a function of the dimensionless for a homogeneous medium. As it can be seen, for situations
time is shown in Fig. 7共c兲, the results for Da= 8 ⫻ 10−6 , 8 where the ratio ␴0 / 具␧0典 is small, the mean porosity is close to
⫻ 10−2, 0.8 and 8 are close to each other; only Da= 80 differs the analytical solution for a homogeneous medium. Note that
considerably from the others. Thus, it is remarkable that the the mean porosity decreases faster for the homogeneous me-
macroscopic permeabilities for Da= 8 and Da= 80 collapse dium than for the heterogeneous ones. The clogging time
onto a single curve as a function of the mean porosity. For all increases with ␴0 / 具␧0典. Moreover, it is obvious now that the
cases, K is a power of 具␧典 with an exponent ␩s = 3.55 for exponential decrease of the mean porosity value takes place
small Damköhler numbers 共Da= 8 ⫻ 10−6 and 8 ⫻ 10−2兲 and only for heterogeneous systems with ␴0 / 具␧0典 sufficiently
␩h = 4.22 for high Damköhler numbers 共Da= 8 and 80兲. large.
Thus, two different regimes are observed for small and The amount of deposit necessary to clog the medium is
large Damköhler numbers. The evolution of the porosity field higher for smaller ␴0, but the variations are not significant.
for high Da can be observed in Fig. 8 for Da= 8 and in Fig. The initial porosity value strongly affects Qdep 关Fig. 14共b兲兴.
9 for Da= 80. Contrarily to small Da, a large proportion of The higher the initial mean porosity, the larger the deposit
matter is deposited in the regions where porosity is high. amount necessary to clog the medium.
This modifies the correlations displayed in Fig. 10 for Da The macroscopic permeability as a function of the mean
= 8 and in Fig. 11 for Da= 80. At clogging, the porosity field porosity does not depend on the average statistical properties
is nearly uniform. of the initial field. The results collapse onto a single curve for
The case where Da= 0.8 is different from the rest of the both Damköhler numbers 关Fig. 14共d兲兴. The macroscopic per-
results. At the beginning of deposition, the evolution of the meability evolutions are different, because of different mean
porosity field is analogous to the illustrative case where Da is porosity values.
small. The porosity field sections and the probability densi-
ties are displayed in Fig. 12. The evolution of the correlation D. Uniform, Gaussian, and self-affine porosity distributions
functions is shown in Fig. 13. At clogging, the porosity field The random porosity field is characterized by the porosity
is not uniform as for large Damköhler numbers, but the prob- probability density and the spatial organization of the porous

016311-15
E. A. BORISOVA AND P. M. ADLER PHYSICAL REVIEW E 71, 016311 共2005兲

FIG. 12. Evolution of a porous medium with an initial lognormal porosity distribution. Each line corresponds to a different time. The left
row is the probability density g共ln ␧兲 of ln共␧共r⬘ , t⬘兲兲; the central one is a cross section of the porosity field parallel to the pressure gradient;
the right one is perpendicular to the pressure gradient. Da= 0.8. The porosity convention is given by the vertical bars.

016311-16
DEPOSITION IN POROUS MEDIA AND CLOGGING ON … PHYSICAL REVIEW E 71, 016311 共2005兲

FIG. 14. Influence of the mean initial porosity and of its stan-
dard deviation on the macroscopic properties. Data are for
具␧0典 = 0.4 and ␴0 = 0.2共䊊兲 , 具␧0典 = 0.2, ␴0 = 0.1共ⴱ兲 , 具␧0典 = 0.2,
␴0 = 0.3共〫兲, illustrative case 共䊐兲. The solid line represents a least-
square data fit. The elementary solution 共4.6兲 is shown by the
dashed-dotted lines. The dashed line corresponds to the analytical
solution for a homogeneous medium with ␧ = 0.2. ␭ / Lc = 8 / 30. Da
= 8 ⫻ 10−6.

rosity and low porosity are clearly distinguishable. The

aspect of the lognormal porosity field is similar to the Gauss-
ian one, but the variations are more pronounced, i.e., the
porosity is either very small or very high. The self-affine
porosity field displays a more complex spatial structure since
FIG. 13. Evolution of the correlation functions R储共u兲 共a兲 and both sharp and smooth changes occur around a single el-
R⬜共u兲 共b兲. The dimensionless length u is normalized by the half- ementary cube.
size of the unit cell. Da= 0.8. The macroscopic parameters derived for these initial
fields are shown in Figs. 15 and 16. As expected, the case of
material; this latter aspect can be described by correlation the uniformly distributed porosity field is distinct from the
functions such as 共3.18兲. Four different porosity fields were others. For both Damköhler numbers, the mean porosity
used in this study, namely Gaussian, lognormal, self-affine, value decreases faster than for the other cases. For small
and uniform. The detailed description is given in Sec. III B. Damköhler numbers, results for Gaussian and self-affine
Gaussian and lognormal fields are characterized by the same fields collapse onto a single curve. For large Damköhler
Gaussian correlation function; they are useful to study the numbers, they are close to each other, yet different. Remark-
influence of the porosity variability on the output results. On ably, the exponential decrease of the mean porosity for small
the other hand, Gaussian and self-affine fields have the same Damköhler numbers is only valid for the lognormal distribu-
Gaussian distribution of the porosity values, but their spatial tion. For the other distributions, the variations can be ap-
organization is very different due to different correlation proximated by a polynomial of degree 3. The least-square
functions. By comparing these two cases, one can estimate approximation coefficients are given in Table I.
the influence of the spatial correlation on the results. The evolution of the mean porosity depends much on the
All the correlated fields are generated with 具␧0典 = 0.2 and spatial correlation of the porosity field as well as on the
variability of the porosity values. At the beginning of depo-
␴0 = 0.2. For the self-affine distribution, H was taken equal to
sition, the total amount of deposit per unit volume as a func-
2. The uniform distribution corresponds to porosities uni-
tion of time is nearly the same for all the cases. Close to
formly distributed over the interval 关0,0.4兴. The uniform dis-
clogging, the results for the lognormal porosity field differ
tribution is distinct from the others since it is not correlated
from the others with less deposited matter.
and since ␴0 is determined by the interval of variations; here For all Damköhler numbers, the macroscopic permeabili-
␴u0 ⬇ 0.115. ties are of the form 共5.5兲. For small Da, the curves of the
The porosity fields are shown in Fig. 1 and their aspects macroscopic permeability as a function of the mean porosity
are seen to be very different. In the uniformly distributed are very close to each other for all cases. They are slightly
porous medium, porosity changes abruptly because of its un- different only for times close to clogging.
correlated character. The Gaussian porosity field is character- For large Da, the relationship between the macroscopic
ized by smooth porosity changes, and regions with high po- permeability and the mean porosity depends on the spatial

016311-17
E. A. BORISOVA AND P. M. ADLER PHYSICAL REVIEW E 71, 016311 共2005兲

FIG. 15. Evolution of the macroscopic parameters for differ-

FIG. 16. Evolution of the macroscopic parameters for differ-
ently distributed porosity fields. Data are for Gaussian porosity dis-
ently distributed porosity fields. Data are for Gaussian porosity dis-
tribution 共䊊兲, self-affine 共ⴱ兲, uniform porosity 共〫兲, lognormal dis-
tribution 共䊊兲, self-affine 共ⴱ兲, uniform porosity 共〫兲, lognormal dis-
tribution 共䊐兲. The solid line represents a least-square data fit. The
tribution 共䊐兲. The solid line represents a least-square data fit. The
elementary solution 共4.6兲 is shown by the dashed-dotted lines. The
elementary solution 共4.6兲 is shown by the dashed-dotted lines. The
dashed line corresponds to the analytical solution for a homoge-
dashed line corresponds to the analytical solution for a homoge-
neous medium with ␧ = 0.2. ␭ / Lc = 8 / 30. Da= 8 ⫻ 10−6.
neous medium with ␧ = 0.2. ␭ / Lc = 8 / 30. Da= 8.

organization of the porous medium. For Gaussian and log- the porous medium properties depending on the Damköhler
normal porosity fields, the curves of the macroscopic perme- number which compares the effects of reaction and convec-
ability as a function of the mean porosity are very close to tion. When convection is dominant 共Da⬍ 1兲, the porous me-
each other. Note that these porosity fields have the same
dium properties change smoothly. At clogging, the porosity
Gaussian correlation function 共3.18兲. On the other hand, the
field is analogous to the initial one. The regions where po-
corresponding results for the self-affine porosity field are
rosity is high, still exist. When reaction is dominant 共Da
very close to those for the uniformly distributed porosity
⬎ 1兲, the porous medium is nearly uniform at clogging. For
field.
Damköhler numbers close to 1, the evolution of the porous
medium is similar to the case Da⬍ 1 at the beginning of
VI. CONCLUDING REMARKS
deposition, and similar to the case Da⬎ 1 for the times close
Deposition in statistically homogeneous media and the to clogging.
subsequent clogging have been studied. A numerical code The heterogeneity of the medium was observed to have a
has been developed which simulates the fluid flow, the trans- strong influence on the evolution of the mean parameters.
port of a reactive solute in heterogeneous porous media as The evolution of the mean porosity value depends on the
well as the alterations of the macroscopic properties of the statistical characteristics of the initial porosity field while the
porous medium due to the deposition of the solute onto the macroscopic permeability as a function of the mean porosity
pore walls. The code has been tested on several situations was found to depend mostly on the spatial organization of
where the analytical solution is known and a good agreement the porous material.
was usually observed. Statistically homogeneous media were This study is a first step towards the understanding of
approximated by spatially periodic media. Since these mod- deposition phenomena on the macroscopic scale. It would be
els are discrete, it is always necessary to validate the choice interesting to extend this work to other cases, particularly to
of several unphysical parameters such as the size of the unit nonuniform local deposition and different constitutive rela-
cell or the space discretization step, etc., applied in the simu- tionships.
lations. Appropriate tests have been performed. The results
have shown that the parameter set was chosen correctly. ACKNOWLEDGMENTS
The code was systematically applied to predict the evolu-
tion of the macroscopic properties of the porous medium Most computations were performed at CINES 共Montpel-
during the deposition of a chemically reactive solute. Some lier兲, subsidized by the MENESR, whose support is grate-
important features were revealed by this study. Starting from fully acknowledged. This work was partly supported by the
the same porous medium, one obtains different evolutions of European grant Arisston ENK6-CT-2000-00052.

016311-18
DEPOSITION IN POROUS MEDIA AND CLOGGING ON … PHYSICAL REVIEW E 71, 016311 共2005兲

关1兴 T. Xu, J. A. Apps, and K. Pruess, J. Geophys. Res. 108共B2兲, 38, Berkeley, California, 1998.
2071 共2003兲. 关11兴 X. Liu, A. Ormond, K. Bartko, Y. Li, and P. Ortoleva, J. Pet.
关2兴 K. P. Saripalli, P. D. Meyer, D. H. Bacon, and V. L. Freedman, Sci. Eng. 17, 181 共1997兲.
Crit. Rev. Envir. Control, CRC 31, 311 共2001兲. 关12兴 Y. Le Gallo, O. Bildstein, and E. Brosse, J. Hydrol. 209, 366
关3兴 M. Sahimi, G. R. Gavalas, and T. T. Tsotsis, Chem. Eng. Sci. 共1998兲.
45, 1443 共1990兲.
关13兴 P. M. Adler, Porous Media. Geometry and Transports
关4兴 M. Shapiro and M. Brenner, Chem. Eng. Sci. 43, 551 共1988兲.
共Butterworth-Heinemann, Washington, DC, 1992兲.
关5兴 J. Sallés, J. F. Thovert, and P. M. Adler, Chem. Eng. Sci. 48,
关14兴 E. E. Petersen, Chemical Reaction Analysis 共Prentice-Hall,
2839 共1993兲.
关6兴 G. P. Matthews, C. J. Ridgway, and J. S. Small, Mar. Pet. Geol. Englewood Cliffs, N J, 1965兲.
13, 581 共1996兲. 关15兴 C. G. Jacquin, Rev. Inst. Fr. Pet. Ann. Combust. Liq. 19, 921
关7兴 V. V. Mourzenko, J. F. Thovert, and P. M. Adler, Chem. Eng. 共1964兲.
Sci. 43, 551 共1996兲. 关16兴 D. Guillot, Ph.D. thesis, ENSM, Paris, 1982.
关8兴 S. Bekri, J. F. Thovert, and P. M. Adler, Eng. Geol. 关17兴 D. Coelho, J. F. Thovert, and P. M. Adler, Phys. Rev. E 55,
共Amsterdam兲 48, 283 共1997兲. 1959 共1997兲.
关9兴 F. F. Chang and F. Civan, J. Pet. Sci. Eng. 17, 123 共1997兲. 关18兴 P. M. Adler and J. F. Thovert, Fractures and Fractures Net-
关10兴 T. Xu and K. Pruess, “Coupled modeling of non-isothermal works 共Kluwer-Academic, Berlin, 1992兲.
multiphase flow, solute transport and reactive chemistry in po- 关19兴 P. K. Sweby, Lect. Appl. Math. 22, 289 共1985兲.
rous and fractured media, 1, Model development and valida- 关20兴 G. Dagan, Flow and Transport in Porous Formations
tion,” Lawrence Berkeley Natl. Lab. Rep. LBNL-42050, pp. 共Springer-Verlag, Berlin, New York, 1989兲.

016311-19