A pseudo two-phase model of colloid transport in porous media

T. Ilina, M. Panfilov & M. Bus

Laboratoire Environnement, Gomcanique & Ouvrages ENSG INPL, Vandoeuvre-ls-Nancy, France

ABSTRACT: We suggest a new mathematical model for the groundwater transport of colloid particles through
porous media which is able to describe some significant effects observed experimentally but not captured within
the framework of the classical approach. Our basic idea is to consider pure water and colloid suspension as
two thermodynamic phases. Using the network models of porous media, we simulated numerically the transport
process at the pore scale. By averaging the result derived, we have obtained the relative permeabilities for both
phases, the percolation threshold for suspension, and the effective suspension viscosity. Due to specific laws
of colloid particles repartition between various classes of pores, the relative permeability of suspension occurs
to be a highly non-linear function of saturation, very far away from the diagonal straight line. This determines
a difference between the macroscale phase velocities. The suspension velocity is shown to be higher than that
of the water in major cases, if only the colloid concentration is not too large. Otherwise the increased effective
suspension velocity significantly reduces its mobility.
The suggested model predicts and describes in a closed form the effect of colloid facilitation.

1 PROBLEM FORMULATION where = (x, t) is the volume of retained or immobile

colloid particles per unit medium volume; = (x, t)
Within the classic approach the colloid particles trans- is the variable porosity; 0 is the original porosity;
ported by water in porous media are examined as is the compaction factor (nonnegative constant); is
infinitely small and so are treated as a component the filtration coefficient; F is an empirical nonnega-
of a single-phase water solution. The transport of tive and nonincreasing function describing the rate of
colloid concentration C is then described by the trapping. If the porosity is invariable, then we obtain
convection-diffusion equation: model (1).
Both these models are unable to contain only one
velocity which assumes that colloid particles have the
same velocity that mean water velocity. But exper-
imentally with nuclide transport it was observed an
where D is the diffusion/dispersion coefficient; V is increasing of colloid/radoinuclide velocity.
the Darcys velocity; is the porosity; f is the source
term which expresses any mechanisms of mass vari-
ation (trapping, adsorption and so one); x and t are 2 SUGGESTED PSEUDO TWO-PHASE MODEL
respectively spatial coordinate and time.
A modification of this model is sometimes applied, 2.1 Statement of model
which partially takes in consideration the finite par- Let now consider the suggested model of colloid trans-
ticle size. Due to this, the medium porosity port. The behavior of colloid particles is independent
may change in time and in space if some particles and essentially differs from that of solution. Indeed,
are trapped. The corresponding mathematical model the model accounts for colloids occupying only these
assuming a linear law for porosity variation is (Logan, pores whose size is larger than that of the particles,
2001): by forming an individual colloid cluster characterized
by an individual permeability. Moreover, the effective
viscosity of the colloid suspension is characterized by
its own effective viscosity which is higher than that
of pure water. Finally, the macroscopic boundary of
the colloid cluster, notwithstanding that it cannot be
observed at a microscopic scale, plays the role of an
interface separating two different liquids: one is the

431
Copyright 2005 Taylor & Francis Group plc, London, UK
pure water, while the other is the aqueous colloid sus-
pension. So, in fluids study the colloid transport should
be represented by a kind of two-phase flow.
At the same time, this flow cannot be considered as
totally two-phase, when it is compared with a true two-
phase system (oil-water). Significant differences exist
between them. First of all, in the examined mixture the
interface between liquids may be only observed at the
macroscopic scale; at the scale of a single pore no inter-
face exists between the pure water and the water carry-
ing solid particles. Secondly, the domain occupied by
each phase remains invariable, which is not the case of
a generalized two-phase flow. Due to these differences,
we call this kind of flow as pseudo two-phase. Figure 1. Empiric parameter A versus a/b.
In the present paper we develop the pseudo two-
phase model for colloid transport through porous The effective viscosity of suspension is determined
media, which has the following form: by the Einsteins formula (Landau & Lifshitz, 1988):

where a, b are half-axis of a particle; A is an empiric

parameter which depends on particle form. It is shown
in Figure 1.
The relative permeability is determined as a result
of the averaging the two-phase flow over an represen-
tative elementary volume (REV) of porous medium.
This procedure leads to a cell-problem with respect to
an auxiliary function which determines the structure of
relative permeabilities as a function of saturation. This
problem was solved numerically using a stochastic net-
where the two first equations are the masse balance work model of porous media described in Panfilova
equations for everyone pseudo-phase and two second (2003). As porous medium we used a second-order
the two momentum balance equations; the subscripts w stationary stochastic medium which is a model in
and s denote the pure water and the colloid suspension good agreement with a real medium. This stochastic
respectively; w and w are the density and viscosity medium structure was generated by the turning bands
of pure water; s and s are the density and viscos- method (Mantoglou & Wilson, 1982). For determine
ity of colloid suspension; S is the saturation of colloid the domains (connected clusters) occupied with each
suspension; is the medium porosity; Vw and Vs are pseudo-phase a percolation method has been used (von
respectively the Darcys velocities of pure water and Neumann, 1966).
colloid suspension; P is the pressure; z is the coor-
dinate axe directed inversely to the gravity force; K
is the absolute permeability of the porous medium; 3 NUMERICAL SOLUTION FOR THE
kw (S) and ks (S) are respectively the relative perme- RELATIVE PERMEABILITY OF
abilities of pure water and colloid suspension; C is the COLLOID SUSPENSION
volume colloid concentration in colloid suspension;
a(C) characterizes the colloid adsorption mechanism. 3.1 Presentation of porous media
2.2 Closure of system To perform a numerical solution the porous medium is
For closure the model it is also necessary to determine presented as a network of cylindrical capillaries no
some parameters such as a density, viscosity and rel- crossed. First of all it is necessary to generate the
ative permeability of colloid suspension. We suggest porous medium which was chosen to be second-order
following closure relations. The effective density of stationary. For this purpose the turning bands method
suspension is determined as an arithmetic mean: is used. This method has been chosen simply as one of
possibilities a random field generation with required
properties. Investigation of methods of random field
construction being not our objective, we required only
where the subscript c stand for colloid particles. some method of such random fields construction.

432
Copyright 2005 Taylor & Francis Group plc, London, UK
 The turning bands method enables to construct a 1
second-order stationary random field proceeding from 0,9 kw(S)
both mean and covariance function given. The main 0,8 kc(S)
idea of this method consists in producing realizations 0,7
on the one-dimensional lines instead of producing
0,6
multidimensional realizations directly.
0,5
0,4
3.2 Determination of connected clusters
0,3
After that it is necessary to determine the domains 0,2
occupied by colloid suspension and by pure water.
0,1
They are the connected clusters. For this purpose the
0
percolation method is used which is an example of 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1
cellular automaton (von Neumann, 1966). S

3.3 Pseudo-discrete formulation of the Figure 2. Relative permeabilities of water (kw(S)) and col-
loid suspension (kc(S)) as function of suspension saturation.
microscopic problem
As mentioned above, the porous medium is presented 1
as a network of cylindrical capillaries no crossed. 0,9
F(S)
Within each capillary the velocity is expressed by
0,8
the Poiseuilles law which establishes a relationship
between velocity and gradient of pressure and which 0,7
is a result of integration of Navier-Stokes equations 0,6 =1.015
through the cross section of tube: 0,5 =1.07
=1.149
0,4

0,3

0,2

0,1
where V is a velocity; P is a pressure drop; lp is a
capillary length; r is a capillary radius; is a parameter 0
0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1
dependent on capillary form; is a viscosity. S
In each cross-point (node) we have the following
equation: Figure 3. Fractional flow.

two-phase model instead of single-phase, even in the

case where the colloid suspension has not significant
viscosity, i.e. essentially different viscosity compared
to that of pure water.
where summation occurs on all capillaries incoming in
A comparison between the suggested model and the
the node. This formula expresses the equality between
traditional one shows a qualitatively different colloid
incoming and outgoing fluxes in the node.
transport behavior in both cases, with formation of
non linear traveling waves predicted by the two-phase
model.
4 THE NUMERICAL RESULTS

4.1 Relative permeabilities and fractional flow 4.2 Facilitated colloid transport
According to the numerical analysis, the obtained rela- Moreover, the proposed model as one of results done
tive permeabilities are sufficiently non linear functions an increase of colloid suspension velocity compared
of saturation, as shown in Figure 2. to that of pure water. This effect was experimentally
And also the obtained fractional flow function is observed during radionuclide contaminating transport
non linear (Fig. 3). that is a very important problem presently.
Below a threshold saturation which is of order of In the Table 1 the ratio between the velocity of col-
0.37, the colloid phase is entirely immobile, because loid suspension and the velocity of pure water along
the cluster of large pores occupied by this phase the direction of flow is presented.
remains disconnected. Such non-linear percolation According to the submitted results one can see an
effects are sufficient and necessary criteria to apply the increasing of velocities ratio to one and a half time

433
Copyright 2005 Taylor & Francis Group plc, London, UK
 Table 1. Ratio between colloid and The obtained numerically mean velocity of colloid
pure water velocities depending on suspension is higher than that of flow, what is observed
colloid viscosity. experimentally for the colloid/radionuclids facilitated
transport and what does not reflect into classical one-
ms Vs /Vw
phase model.
1.015 1.45
1.070 1.37
1.149 1.28 REFERENCES

Landau L.D. & Lifshitz E.M. 1988. Hydrodynamics.

Moscow: Nauka (in Russian) (Engl. Transl.: 1993. Fluid
Mechanics. Oxford: Pergamon Press).
depending on colloid concentration (because of colloid
Logan J. & David. 2001. Transport modeling in hydrogeo-
viscosity depends on it). chemical systems. Interdisciplinary applied mathematics.
New-York : Springer-Verlag, New-York Inc.
Mantoglou A. & Wilson J.L. 1982. The turning bands method
5 CONCLUSIONS for simulation of random fields using line generation by
a spectral method. Water Resources Research vol. 18 no.
The proposed model of colloid transport is carried 5: pages 13791394.
out when the colloid suspension is examined as a one Panfilova I. 2003. Ecoulements diphasiques en milieux
pseudo-phase flowing with another pseudo-phase as poreux: modle de mnisque. Thse, Institut National
Polytechnique de Lorraine. Nancy.
pure water.
von Neumann J. 1966. Theory of self-reproducing automata.
From the numerical results the relative permeabil- Urbana: (A.W. Burks, Editor) University of Illinois
ities are obtained sufficiently non linear functions of Press, IL.
suspension saturation, that is due to a specific phase Wonho Oh. 1998. Random field simulation and an applica-
repartition in pores and that is sufficient criteria to tion of kriging to image thresholding. New York: State
apply the two-phase model instead of single-phase one. University of New York.

434
Copyright 2005 Taylor & Francis Group plc, London, UK