Home Search Collections Journals About Contact us My IOPscience

Liquids in porous media

This article has been downloaded from IOPscience. Please scroll down to see the full text article.

1990 J. Phys.: Condens. Matter 2 SA79

(http://iopscience.iop.org/0953-8984/2/S/008)

View the table of contents for this issue, or go to the journal homepage for more

Download details:
IP Address: 18.7.29.240
The article was downloaded on 10/07/2013 at 20:31

Please note that terms and conditions apply.

J. Phys.: Condens. Matter 2 (1990) SA79-SA88. Printed in the UK

Liquids in porous media

R Lenormand
Institut Franqais du PBtrole, BP 311, 92506 Rueil-Malmaison CCdex, France

Received 23 August 1990

Abstract. The basic mechanisms which take place during the displacement of
immiscible fluids in porous media have been observed in micromodels and have been
modelled. At the pore level, in drainage, the invading fluid chooses the largest throat.
In imbibition, the displacement depends on the local geometry. For a large pore-te
throat ratio (aspect ratio), the main mechanism is the collapse of the invading fluid in
the smallest channel, without entering the pore. For a small aspect ratio, the wetting
fluid invades the pore first, and then the adjacent channels. From observations a t the
pore level, we have modelled the displacement on a large scale in some extreme cases
by using statistical theories. The different behaviours are then displayed as domains
in three phase diagrams: one for drainage and two for imbibition (large and small
aspect ratios). At a high rate, when viscous forces are dominant, all the diagrams
show a stable domain (described by anti-DLA) and a viscous fingering domain ( D L A ) .
In drainage, low capillary numbers lead to capillary fingering represented by invasion
percolation. In imbibition, the capillary domain is described either by a compact
cluster growth (small aspect ratio) or percolation theory (large aspect ratio). In
addition the possibility of flow by film along the roughness of the walls leads to
disconnected structures.

1. Introduction

We are interested in the distribution of two immiscible fluids inside a porous medium,
in order to predict transport properties. In this domain, the lack of detailed infor-
mation on the mechanisms generally leads to the use of empirical parameters. By
using transparent porous media, we have obtained significant insight into the physical
mechanisms and we have been able to develop consistent models based on microscopic
observations.
We have studied the pattern formed by a fluid (2) pushing an immiscible fluid (l),
with slow displacements (no inertial effects), without any gravity effect. The pore size
is large (larger than one micron) and we assume that all the interactions between the
fluids and the solid can be represented by a contact angle. All the physics is then
controlled by two kinds of mechanisms: viscous pressure drop and capillarity. The
relative intensity of these two kinds of forces is quantified by the capillary number,
defined as the ratio of viscous to capillary forces at the pore scale:

where V is a mean velocity of one fluid, p its viscosity and y the interfacial tension.

0953-8984/90/SA0079+10$03.50 @ 1990 IOP Publishing Ltd SA79

SA80 R Lenormand

Another parameter is the viscosity ratio, defined by

The contact angle is near zero and is not a parameter in the study. However, we
must distinguish whether the displacing fluid is the wetting fluid (imbibition) or the
non-wetting fluid (drainage).
This paper is essentially a survey of several former publications where more details
can be found. However, the series of imbibition experiments and the phase diagrams
for imbibition are original.
The first part of this paper gives some examples of displacements in micromodels.
The second part depicts the observations of the displacement of the meniscus a t the
pore scale. Then, the large-scale patterns are explained and modelled for drainage and
imbibition. T h e different types of displacement are considered in the form of phase
diagrams for drainage and imbibition.

Figure 1. Situation of the fluids in a channel of the micromodel

2. Experiments

The micromodels used in the experiments are made of transparent resin cast in a
photographically etched mold [l].We will call the cylindrical ca.pillaries channels, and
we will call the volumes of the intersections of the etched network pores. The cross
section of each channel is rectangular (figure 1) with a constant depth (e = 1 mm)
and a width d that varies from pore to pore with a given distribution and a random
location. Various sizes of networks are used for the experiments with the largest one
containing more than 250000 channels. Figure 2 shows some situations of fluids in
the pores after a displacement: ( a ) air and oil (gray) after drainage (air injection), ( b )
after imbibition; (c) mercury (black) and mercury vapor after imbibition (withdrawal).
Now we give some examples of fluid injection at a large scale, in a 40 000 channel
network (130 x 150 mm). Each series of experiments is performed with the same fluids
(same viscosity ratio M ) at various flow rate (or C a ) . In drainage, the non-wetting
fluid is injected on the left. In imbibition, the wetting fluid is injected a t the centre.
In this case, the capillary number is calculated for radial injection at a distance of
30 mm from the centre.
(i) Stable drainage (figure 3): mercury displacing air ( M = 80).
(ii) Unstable drainage (figure 4): air displacing oil ( M = 2 x
(iii) Stable imbibition in a square network with a narrow distribution of channel
width (figure 5 ( a ) ) : oil displacing air ( M = 2.9); same imbibition in a triangular
 Liquids in porous media SA81

Figure 2. Close-up of various of displacements in micromodels: (a)drainage, air

displacing oil (gray); (b)imhihition at very low rate (oil-air); (c) mercury withdrawal
(liquid mercury in black).

Figure 3. Drainage: injection of mercury Figure 4. Drainage: injection of air displacing

(black) displacing air at various capillary nuiii- oil (black) at various capillary numbers.
bers.
network (figure 5 ( b ) ) ;same imbibition in a square network with a wide distribution of
channel width (from 0.16 to 0.64 mm): figure 5(c).
All these experiments are examples of displacements we want to int.erpret. For
this purpose, we will first study the mechanisms at the pore level.
SA82 R Lenormand

Figure 5. Imbibition. ( U ) Injection of oil (black) displacing air in a squarc network

with a narrow pore size distribution. ( b ) Injection of oil (black) displacinb air in a
triangdar net.work with a narrow pore size distribution. (c) Injection of oil (black)
displacing air in a square network wit.h a wide pore size distribution.

3. Physical mechanisms at tlie pore level

Experiments show that both fluids can flow simultaneously in the same channel with
different velocities, the wetting fluid remaining in the extreme corners of the cross
section and roughness of the walls (figure 1). This effect explains the richness of
mechanisms involved during displacements.
A displacement can be divided into three parts: (i) flow of the injected fluid from
the entrance towards the moving meniscus, (ii) displacement of the meniscus, and (iii)
flow of the displaced fluid towards the exit.

3.1. Menisciis displacements

T h e mechanisms are different in drainage and imbibition.
 Liquids in porous media SA83

In drainage, the non-wetting fluid is stopped by the throat at the entrance of a

channel until the pressure exceeds the pressure in the wetting fluid by a value equal to
the capillary pressure, After passsing this throat, the non-wetting fluid spontaneously
invades the adjacent pore (intersection), which is larger than the channel.
In imbibition, there are two additional effects due to pore geometry and also to
the flow by film along the roughness of the solid. The effect of geometry, which can
introduce a selection of imbibition mechanisms, have been dealt with in detail in a
previous paper [2]. Similar results have also been observed in micromodels by Wardlaw
and Li [3].
We found two main mechanisms.
(i) Pore invasion (figure 6). During invasion of the wetting fluid, the capillary
pressure decreases and consequently the radius of curvature of the meniscus increases
(figurs 6(a) and ( b ) . At a given pressure PI(figure 6 c ) , the meniscus touches the wall
and instantaneously the wetting fluid invades the pore and the adjacent channels. This
displacement shows some analogy with a contact line moving on a two-dimensional
rough surface [4,5].
(ii) Collapse in a channel (figure 7). The wetting fluid can flow along the roughness
of the walls and surround the solid grains (figure 7 ( a ) ) . At a given pressure Pz, the fluid
which accumulates on the walls of a channel becomes unstable and fills the channel
( 7 ( b ) ) . This mechanism has been described under the name of pinch-off by Vizika
and Payatakes [6].

a) b)

C) d)
Figure 6. Mechanism of pore invasion.
I

Figure 7. Mechanism of fluid collapse in a channel.

Now, in a micromodel the capillary pressure decreases when the fluid is injected.
Consequently the dominant mechanism corresponds to the highest threshold pressure
SA04 R Lenormand

(it occurs first). These pressures PI and P2 can be calculated as functions of the
local geometry and also of the contact angle [2]. Generally, because of the pore size
distribution, both mechanisms occur during imbibition. However, we can distinguish
two extreme cases, depending on the pore-to-throat ratio (aspect ratio).
(i) Small aspect ratio. When the size of the pore is small compared with channel
diameters, the meniscus touches the wall very quickly (figure 6(c)) and Pl is higher
than P2. Imbibition takes place by a succession of pore invasion. This leads to filling
of the network line by line as shown in figure 8.
(ii) Large aspect ratio. In the opposite case, P2 is greater than Pl and the collapse
inside a channel is the dominant mechanism (figure 7).
These two mechanisms lead to different patterns a t a large scale: flat interfaces
for a small aspect ratio and ramified structures for a large aspect ratio.

Figure 8. Different stages of the growth of a compact cluster by pore invasion.

3.2. Flow of fluids

The non-wetting (NW) fluid can flow only in the bulk of the channels. So flow occurs
only if a continuous path of channels or pores filled with this phase exists towards
either the entrance (during drainage) or the exit (during imbibition) of the network.
Otherwise the NW fluid is trapped.
T h e flow of the wetting fluid is more complicated. We observe three kinds of flow
of the wetting fluid: (i) flow in the bulk of the channels, (ii) along the corners when
the NW fluid fills the central part of the channel and (iii) by film along the roughness
of the walls, but only for strong wettability and very low flow rate.

4. Drainage

Now we observe the patterns formed by the injected fluid during invasion at the scale
of the network.
In drainage, we have identified three main regimes for the displacement of a wetting
fluid by a NW fluid. This approach leads to a n original display of the corresponding
domains on a general diagram with axes representing the capillary number and the
viscosity ratio (figure 9). These three regimes correspond to the limits when two of the
three kinds of forces involved during displacement are negligible: (i) capillary fingering,
when the injection rate is very low and viscous forces are negligible in both fluids; (ii)
stable displacement, a t high rate when capillarity forces are low and with negligible
 Liquids an porous media SA85

DRAINAGE
Figure 9. Phasediagram for drainage.

pressure drop in the displaced phase (small M); and (iii) unstable displacement, in
the opposite case, when the pressure drop is negligible in the injected fluid (large M).
In addition we have shown [7] that these three limit regimes can be represented
by statistical models.
(i) Invasion percolation for capillary fingering [8,9]. The capillary threshold across
the meniscus allows only a fraction of channels (or bonds) to be open to the flow. A
computer simulation of invasion percolation is based on the following rules: a random
number (between 0 and 1) is given at each bond of the network, and this number is
kept fixed during the entire simulation. At each step, the interface is moved to invade
the bond adjacent to the interface with the largest number.
(ii) Diffusion limited aggregation (DLA)for viscous fingering [lo]. This mechanism
is linked to a growing interface in a Laplacian field, which is the case for the pressure
field during displacement in porous media. DLA can be represented by the following
process [ll]: a seed particle or a line is placed on a lattice. Another particle is lauched
from a faraway source and moves at random. It is trapped when it reaches the seed or
the line. Another particle is then launched, and so on. This diffusive process produces
very ramified clusters similar to the injection pattern shown in figure 4-1.
(iii) Anti-DLA for stable displacement. This model is the reverse of DLA. It consists
in releasing antiparticles near a compact aggregate. The antiparticle moves a t random
and, when it reaches a particle, both are removed.
Now, the two series of drainage experiments (figures 3 and 4) can be understood
through this phase diagram. They represent the evolution with the capillary number
when the viscosity ratio is kept constant: figure 3 at M = 80 from a stable displace-
ment in the anti-DLA domain (1) toward capillary fingering in the percolation domain
(4);figure 4 at M = 2 x from an unstable displacement in the DLA domain (1)
toward capillary fingering in the percolation domain (4).
In drainage, this phase diagram has been validated by a large number of experi-
ments together with numerical simulation for a network of capillaries [7].
SA86 R Lenormand

We will now examine the case of imbibition, which is much more complicated
because of the effect of the geometry of the solid and the flow by film a t a very low
flow rate.

5. Imbibition

First of all, we assume that all the viscous domains are the same as for drainage.
Experimentally we have verified only the stable case (figure 5 ) a t the highest rates.
In the unstable case, it seems that an important effect is the increase of the contact
angle with velocity. At high rate, the injected fluid invades the central part of the
channels as in drainage.
When capillary forces are great, we must account for the geometry of the solid.
We will examine the two extreme cases observed at the microscopic scale and related
to the ratio of pore-to-throat size (aspect ratio).

5.1. Large aspect ratio

The injected wetting fluid invades the network by a succession of collapses in the
channels. Due to pressure effects, the smallest channel is filled first. Without flow by
film, the filled channel is adjacent to the interface. With flow by film, this channel can
be anywhere in the network. Consequently, we are filling bonds a t random locations in
the network, and this mechanism is related to a bond percolation problem. However,
the topology is not the same as in drainage. As shown in figure 7 ( b ) , the collapse of
the wetting fluid in the channel links the two centres of the grains. The centres of
these grains form the dual network.
To summarize, a t intermediate flow rate (Ca R 10-5 [ 2 ] ) ,we should observe
capillary fingering described by invasion percolation. At a lower rate (Ca < lov7,
[ 2 ] ) ,we expect ordinary percolation.

5.2. Small aspect ratio

In this case, the filling of the network line by line (figure 8) leads to a kind of faceted
crystal growing a t the injection point. The compact shape of the crystal is related to
the network mesh: rectangular (figure 5 ( a ) , log Ca = -6) or triangular (figure 5(b)).
For flow by film, the same mechanism occurs anywhere in the network, without
any apparent connectivity with the injection. The result is an ensemble of compact
clusters. We have studied the statistics of this process and have shown that, for an
infinite network, there is no threshold in pressure, as opposed to percolation [12].
Figures 2(b) and 2(c) are examples of such disconnected clusters in a square network
with oil (2b) and mercury (2c).

5.3. Phase diagram for imbibition

All the above-described mechanisms can be displayed in the two diagrams in fig-
ure lO(a) for a large aspect ratio and 10(b) for a small aspect ratio. In both cases, we
have the unstable and stable viscous cases at high capillary number, an intermediate
 Liquids in porous media SA87

Viscous
domains

Continuous
CaDillarv domains

Discontinuous
capillary domains

83 large
aspect-ratio
s mall
aspect-ratio

Imbibition
Figure 10. Phase diagrams for imbibition: left-hand figure, large aspect ratio;
right-hand figure, small aspect ratio.

zone for the continuous capillary domain (without flow by film), and at very low Ca,
the discontinuous capillary domain (with flow by film).
In fact, these two diagrams are extreme (and theoretical) cases. In experiments
in micromodels with a wide channel width distribution the clusters are more compact
than in standard percolation. In fact, the geometry is intermediate between a large
and a small aspect ratio.
Let us return to the series of experiments. Figures 5 ( a ) and ( b ) are two cases of a
small aspect ratio. The narrow distribution prevents any overlap of the mechanisms.
These series start from a faceted crystal (respectively rectangular and triangular),
without flow by film at low Ca. Increasing the injection rate progressively leads to
stable circular displacements. Fig 5(c) is intermediate between a large and small
aspect ratio, with a branched cluster close to percolation in the continuous capillary
domain. Increasing the flow rate leads to a stable circular displacement.
We have no experiments showing the transition between unstable flow and faceted
crystals. However, displacements of viscous oil by water have been performed by
Stokes et a1 [15] in glass bead packs, which behave like small aspect ratio media. The
main result is an increase in the width of the fingers when Ca decreases. In other
domains of physics, similar cross-overs from ramified to compact patterns have been
experimentally obtained by Skjeltorp for aggregation of uniformly sized microspheres
[13] and have been simulated by Schwartz Sgirensen et a1 [14] for solidification.
For imbibition, we cannot rely on numerical simulations to improve and validate
these diagrams. Except cellular automata on a very small scale (a few pores [16]),
there are no simulators taking into account both imbibition mechanisms and viscous
effects. Cieplak and Robbins (171 simulated imbibition with various contact angles.
They have taken into account the meniscus displacement, but not the flow by film
(leading to the collapse mechanism) and viscous effects. Payatakes [18] improved a
SA88 R Lenormand

simulator (with viscous effects) to represent the flow by film, but mechanisms at the
pore level seem closer to drainage than imbibition.

6. Conclusion

With the help of visualizations in transparent micromodels, we have been able to

describe and to model the mechanisms that take place during the displacement of
immiscible fluids in porous media.
At the pore level we have observed the displacement of the meniscus. In drainage
the invading fluid chooses the largest throat. In imbibition the displacement depends
on the local geometry. For a large aspect ratio, the main mechanism is the collapse
of the invading fluid in the smallest channel, without entering the pore. For a small
aspect ratio, the wetting fluid invades the pore first, and then the adjacent channels.
From observations a t the pore level, we have been able to model the displacement
at a large scale by using statistical theories. The different domains have been displayed
in three phase diagrams: one for drainage and two for imbibition (for a large and small
aspect ratio).
At a high rate, when viscous forces are dominant, all the diagrams show a stable
domain (described by anti-DLA) and a viscous fingering domain (DLA).
In drainage, low capillary numbers lead to capillary fingering represented by in-
vasion percolation. In imbibition, the capillary dornain can be decribed either by
faceted crystals (small aspect ratio) or percolation (large aspect ratio). In addition
the possibility of flow by film along the roughness of the walls leads to disconnected
structures .

References

Lenormand R, Zarcone C and Sarr A 1983 J . Fluid Mech. 135 637

Lenormand R and Zarcone C 1984 Paper presented at 59th Ann. Meeting. of Soc. Petr. Eng.,
Houston paper 13264
Wardlaw N c and Li Y 1988 Transport in porous media 3 17-3.1
Jansons K M 1985 J. Fluid Mech. 154 1
Robbins M 0 and Joanny J F 1987 Europhys. Lett. 3 729
Vizika 0 and Payatakes A C 1989 PCH Physicocheniical Hydrodyn 11 187-204
Lenormand R, Touboul E and Zarcone C 1988 J . Fluid Mech. 189 165587
Lenormand R and Zarcone C 1985 Phys. Rev. Lett. 54 2226-9
Wilkinson D and Willemsen J F 1983 J. Phys. A : Math. Gen. 16 3365
Paterson L 1984 Phys. Rev. Lett. 52 1621-4
Witten T A and Sander L M 1983 Phys. Rev. B 27 5686-97
Lenormand R and Zarcone C 1984 in Kinetics of Aggregation and Gelation ed F Family and D
P Landau (Amsterdam: Elsevier)
Skjeltorp A T 1987 Phys. Rev. Lett. 58 1444-7
Schartz Sorensen E, Fogedby C and Mouritsen Ole G 1988 Phys. Rev. Lett. 61 2770-3
Stokes J P, Weit D A , Gollub J P, Dougherty A , Robbins M 0, Chaikin P M and Lindsay H
M 1986 Phys. Rev. Lett. 57 1718
Cieplak M and Robbins M 0 1988 Phys. Rev. Lett. 60 2042-5
Payatakes A 1990 Fundamentals of Fluid Transport in P O T O UMedia
S (fides, France 1 9 9 0 )
Rothman D H 1989 Discrete Kinetic Theory, Lattiee-gas Dynamics, and Foundations of Hy-
drodynamics ed R Monaco (Singapore: World Scientific) pp 286-99