Oct 3, 2011 - Holtzman and Juanes, 2010; Chevalier et al., 2009; Johnsen et al.,. 73. 2006; Varas et al., ...... C. R. Mecanique 331, 835â. 410. 842. 411. 19 ...
1
Biphasic flow: structure, upscaling, and
2
consequences for macroscopic transport properties
3
R. Toussaint (1), K.J. M˚ aløy (2), Y. M´eheust (3), G. Løvoll (2, 4), M. Jankov (2), G. Sch¨afer (4), J. Schmittbuhl (1) (1) IPGS, CNRS, University of Strasbourg, Strasbourg (France) (2) Physics Department, University of Oslo, Oslo (Norway) (3) G´eosciences Rennes, University of Rennes 1, Rennes (France) (4) Det Norske Veritas AS, Research and Innovation, Høvik (Norway) (5) LHYGES, CNRS, University of Strasbourg, Strasbourg (France)
4
October 3, 2011
5
Abstract
6
In disordered porous media, biphasic immiscible fluid flows orga-
7
nize in patterns that sometimes exhibit fractal geometries over a range
8
of length scales, depending on the capillary, gravitational and viscous
9
forces at play. These forces, as well as the boundary conditions, also
10
determine whether the flow leads to the appearance of fingering path-
11
ways, i.e., unstable flow, or not. We present here a short review of
12
these aspects, summarizing when these flows are expected to be stable
13
or not, what fractal dimensions can be expected, and in which scale
14
ranges. We base our review on experimental studies performed in two-
15
dimensional Hele-Shaw cells, or addressing three dimensional porous
16
media by use of several imaging techniques. We first present configu-
17
rations in which solely capillary forces and gravity play a role. Next,
18
we review configurations in which capillarity and viscosity are the main
1
19
forces at play. Eventually, we examine how this microscopic geome-
20
try affects the macroscopic transport properties: its fractal structure
21
and the way in which the scale range depends on the flux have for
22
instance an impact on the relationship between saturation and the
23
pressure difference between the two phases, at the macroscopic scale.
24
An example of such an upscaling is illustrated in details.
25
1
26
The physics of two phase flows in porous media is a complex and
27
rich topic, with obvious applications to the hydraulics of the vadoze
28
zone, be it water infiltration, its evaporation, or the transport of Dense
29
Non-aqueous Phase Liquids (DNAPL) down to the aquifers (Dridi,
30
2009). Hydrogeologists and soil scientists aim at relating volumetric
31
flow, pressure head, and water content at the Darcy scale, that is,
32
a meso-scale above which the medium and the flow are described by
33
continuous mathematical fields. The basic laws of multiphase flows
34
treated at mesoscopic scale as a continuum require a closure of partial-
35
flow Darcy relations. The key point of this closure is a functional
36
relation between the capillary pressure (in a water-air system) and
37
(water) saturation in the form of retention curves.
Introduction
38
The physics community has been mostly concerned with character-
39
izing and understanding flow structures/patterns from the pore scale
40
and up. These structures and processes have a ajor impact on the
41
retention curves. Notably, viscous fingering may have strong influence
42
on retention curves resulting in dynamic saturationpressure curves in
43
porous media.
44
These flow structures can vary from the compact to the ramified
45
and fractal (Lenormand , 1989; Lenormand et al. , 1989; M˚ aløy et al.,
46
1985; M´eheust et al., 2002; Sandnes et al., 2011; Holtzman and Juanes,
47
2010). One major issue is to simplify this complexity by keeping just
48
enough information to describe the relevant physics at the relevant
2
49
scale for the flow considered, without discarding important informa-
50
tion. The porous body of a piece of soil or rock consists of pores and
51
fracture networks of different length scales and shapes, whose perme-
52
ability is often anisotropic and presents large spatial variations (see e.g.
53
Zimmerman and Bodvarsson (1996); M´eheust and Schmittbuhl (2000,
54
2001); Neuville et al. (2010a,b); Neuville et al. (2011a); Neuville et al.
55
(2011b,c)). In general the soil/rock is a dynamic medium where the
56
porosity can be modified by the fluids involved due to chemical reac-
57
tions and desorption/adsorption mechanisms, in addition to the fluid
58
pressure and the mechanical stress acting on the porous medium. The
59
chemical composition and nano/micro structure of the rock further
60
decides the wetting properties of the fluids which is crucial for the cap-
61
illary front advancement in two-phase flow. For example, when a fluid
62
with high viscosity is displaced in a porous medium by a fluid with a
63
lower viscosity, the displacing fluid tends to channel through the paths
64
of lower flow resistance, thereby forming pronounced fingers. The in-
65
fluence of the physical properties of the fluids plays important practical
66
role on natural flows: e.g., in soil and groundwaters, the identification
67
of pollution sources is difficult due to the fact that organic pollutants
68
can rapidly migrate down to the bottom of the aquifer and/or along
69
paths different from the water (Benremita et al., 2003).
70
In addition, when the porous medium is deformable, branching
71
structures can be observed with transitions to fracturing of the porous
72
medium (Lemaire et al., 1993; Cheng et al., 2008; Sandnes et al., 2011;
73
Holtzman and Juanes, 2010; Chevalier et al., 2009; Johnsen et al.,
74
2006; Varas et al., 2011) or formation of fingers, channels or bubbles
75
in it (Johnsen et al., 2007, 2008; Kong et al., 2011; Vinningland et al.,
76
2007a,b, 2010, 2011; Niebling et al., 2010a,b).
77
In this review we will address the detailed structure and dynam-
78
ics of two phase flow in fixed and disordered porous media, based on
79
pore scale experiments. We will limit the discussion to drainage, i.e.
3
80
to situations where a non wetting fluid displaces a wetting one – even
81
though imbibition, where a wetting fluid invades a non wetting one, is
82
of equal practical importance. The discussion will be limited to me-
83
dia that are isotropic and homogeneous at large scales, and to cases
84
where chemical reactions and adsorption or desorption between the
85
fluids and the porous medium can be neglected. The structures of
86
clusters of moving fluid and the dynamics of drainage in porous media
87
depend on several parameters like the density difference, the surface
88
tension, the wetting properties, the viscosities and the flow rates of the
89
fluids involved. These different parameters correspond to forces. The
90
forces that dominate on different length scales and their interplay give
91
rise to separate scaling regimes. Up-scaling, which consists in relating
92
the pore scale description to properties defined at the Darcy scale or
93
even at the macroscopic scale, is a central topic within hydrology and
94
oil recovery. Only by understanding the scaling of the structures and
95
dynamics within each regime, and what controls the crossover lengths
96
involved, is it possible to perform up-scaling. The structures involved
97
are typically fractal within some scaling range; their fractal dimension
98
depends on length scales, and often result from fluctuations occur-
99
ring at smaller scales. At the end of this short review we provide an
100
example of upscaling of recent experimental data; the experiments in
101
question were aimed at studying the crossover between capillary and
102
viscous fingering in a quasi two-dimensional porous medium.
103
104
2
Capillary and gravitational effects
105
When drainage is performed in the limit of infinitely slow displacement
106
velocities, the pressure drop accross the porous medium is controlled
107
by the capillary pressure drop accross the interface between the two
108
fluids. The criterium for advancement of the interface into a given pore
109
is that the capillary pressure drop be larger than the capillary pressure
4
110
threshold needed to invade the pore neck that separates that pore from
111
the already-invaded adjacent pore. The value of the capillary pressure
112
threshold fluctuates from pore neck to pore neck, with a distribution
113
function determined by the geometry of the porous medium. In the
114
case of zero gravity or for a horizontal 2D porous medium, the next
115
pore throat/neck to be invaded will be, among the pores that touch
116
the interface, the one whose pore throat has the smallest capillary
117
pressure threshold. This idea is the basis of the invasion percolation
118
algorithm (de Gennes and Guyon, 1978; Chandler et al., 1982; Wilkin-
119
son and Willemsen, 1983), where random numbers representing the
120
capillary pressure threshold values are distributed on a lattice and
121
where the front is moved at each time step at the location along the
122
interface corresponding to the smallest threshold value. The fact that
123
the fluid front always moves at the most easily invaded pore neck and
124
nowhere else is actually not always true in real flows, even if it is a
125
good approximation. What drives the advancement of the front is the
126
capillary pressure build-up in the fluid. The capillary pressure will not
127
relax immediately after invasion of a new pore but is controlled by a
128
back contraction of the fluid interface. This is the reason for the so
129
called Haines jumps which may lead to invasion of several pores in one
130
jump (Haines, 1930; Furuberg et al., 1996; M˚ aløy et al., 1985)
131
When the displaced fluid is incompressible (or lowly compressible),
132
trapping takes place. Trapping is very important in two dimensions
133
(2D) (Wilkinson and Willemsen, 1983) but much less significant in
134
three dimensions (3D). Experiments addressing this so-called capil-
135
lary fingering in 2D model systems were first performed by Lenormand
136
(1989)who found a fractal dimension of the invaded structures equal
137
to D = 1.83, in consistency with the results of numerical simulations
138
based on invasion percolation with trapping (Wilkinson and Willem-
139
sen, 1983). In 3D, several experiments have been performed (Chuoke et
140
al., 1959; Paterson et al., 1984a,b; Chen et al., 1992; Hou et al., 2009;
5
141
Mandava et al., 1990; Nsir et al., 2011; Jianzhao et al., 2011). The
142
fractal dimension found at small scale, between 2 and 2.6 is compat-
143
ible with the dimension D = 2.5 found in three dimensional invasion
144
percolation models (Wilkinson and Willemsen, 1983).
145
Even though the capillary fingering structure is fractal, in practice
146
it is well described by a fractal dimension only within a window of
147
length scales ranging from the pore size up to a crossover length on
148
a larger scale. In the case where the density difference between the
149
two fluids is different from zero, but where viscous forces are small
150
compared to the others, this crossover length corresponds to a scale at
151
which the capillary threshold fluctuations become of the same order of
152
magnitude as the difference in hydrostatic pressure drop between the
153
two fluids. This means that the crossover will always occur when the
154
fluid structures become large enough. When a lighter fluid is displacing
155
a heavier one from above at a slow flow rate resulting in low viscous
156
forces, a stable displacement is observed , that is, the displacing fluid
157
does not finger its way trough the displaced fluid; the crossover length
158
sets the width of the rough interface between the two fluids (Birovljev
159
et al., 1991; M´eheust et al., 2002; Løvoll et al., 2005), parallel to the
160
average flow direction. Gravitational effects can easily be accounted for in the invasion percolation model by mapping the system onto a problem where the capillary threshold values are modified linearly by the hydrostatic pressure difference between the two fluids (Wilkinson, 1984; Birovljev et al., 1991; Auradou et al., 1999). By using this theory of percolation in a gradient , Wilkinson (1984) predicted theoretically the scaling of the front width ξ observed in a gravitational field as −ν
ξ/a ∝ B0ν+1 ,
(1)
161
where the Bond number B0 = ∆ρga2 /γ is the ratio of the difference
162
of the hydrostatic pressure drops in the two fluids on the length scale
6
163
of a single pore to the capillary pressure drop. Here, ∆ρ is the density
164
difference between the fluids, g the gravitational acceleration, a the
165
characteristic pore size and γ the surface tension between the two flu-
166
ids. Gravitationally-stabilized fluid fronts occurring during very slow
167
two-dimensional drainage have been studied both experimentally and
168
by computer simulations (Birovljev et al., 1991). The results were
169
found consistent with the theoretical prediction of Wilkinson (1984).
170
When a lighter fluid is injected into a heavier fluid from below (Frette
171
et al., 1992; Birovljev et al., 1995; Wagner et al., 1997), gravitational
172
fingering of the displacing fluid through the displaced fluid occurs; a
173
scaling behavior consistent with Eq. (1) has has also been found in
174
this unstable case: the characteristic length scale ξ then corresponds
175
to the width of the unstable gravitational fingers, perpendicularly to
176
the average flow direction. The same simple mapping to invasion per-
177
colation as described above can also be performed in the case of slow
178
displacement in a rough fracture joint filled with particles (Auradou et
179
al., 1999).
180
When comparing systems with different capillary pressure threshold
181
distributions, Eq.(1) needs to be modified. From the phenomenology of
182
the invasion process that we have explained above, it is quite intuitive
183
that a gravity-stabilized front in a porous medium presenting a narrow
184
capillary noise (i.e. a narrow distribution of the capillary thresholds)
185
will give a flatter front than a porous medium with a wide capillary
186
noise. Instead of equation Eq. (1), it has therefore been suggested to
187
take into account a dimensionless fluctuation number F = ∆ρga/Wt ,
188
in which capillary fluctuations are accounted for in terms of the width
189
Wt of the capillary pressure distribution (Auradou et al., 1999; M´eheust
190
et al., 2002). Experiments to check the dependence of the displace-
191
ment process on the capillary noise are difficult, because controlling
192
the capillary noise in a systematic way is not straightforward. These
193
experiments therefore remain to be done.
7
194
3
Capillary and viscous effects
The crossover between capillary fingering and regimes for which viscous effects are dominant was first studied in the pioneering work of Lenormand (1989). He classified the different flow structures in a phase diagram depending on the viscosity contrast M = µi /µd between the fluids and the capillary number Ca = ∆Pvisc /∆Pcap , which is the ratio of the characteristic viscous pressure drop at the pore scale to the capillary pressure drop. Here µi and µd are the viscosities of the injected and displaced fluid, respectively. From the Darcy equation, the capillary number can be evaluated as
Ca =
∆Pvisc µa2 v a∇Pvisc = . = ∆Pcap γ/a κγ
(2)
195
where γ is the surface tension, a the characteristic pore size, κ the in-
196
trinsic permeability of the medium, v the seepage velocity associated to
197
the imposed flow rate of the displaced fluid, µ the viscosity of the most
198
viscous fluid, and γ/a the typical capillary pressure drop accross the
199
interface. Lenormand identified three flow regimes: (i) stable displace-
200
ment, for which the interface roughness is not larger than one linear
201
pore size, (ii) capillary fingering, which we have discussed in section 2,
202
and (iii) viscous fingering, which occurs when large scale fingers of the
203
displacing fluid develop inside a more viscous defending fluid, resulting
204
in a much faster breakthrough of the displacing fluid. It is important
205
to keep in mind that the observed structures will depend on the length
206
scale considered. For large systems it is therefore not meaningful to
207
talk about a sharp transition in a phase diagram between capillary
208
and viscous fingering, because one will always have both structures
209
present, i.e. capillary fingering on small length scales, and either vis-
210
cous fingering or stable displacement on large length scales. When the
211
two fluids involved have different viscosities, the viscous pressure drop
212
between two points along the fluid interface will typically be differ-
8
213
ent in the two fluids. This viscosity contrast will produce a change
214
in the capillary pressure along the fluid interface, therefore playing a
215
role similar to that of density contrasts in the presence of a grav-
216
itational field (see section 2). At sufficiently large length scales, the
217
difference in viscous pressure drop between the two sides of the in-
218
terface will become larger than the typical fluctuations in capillary
219
pressure threshold. This means that at sufficiently large length scales,
220
and thus for a sufficient large system , viscous pressure drops , rather
221
than capillary forces associated to random capillary thresholds, de-
222
termine the most likely invaded pores; consequently, viscous fingering
223
will always dominate at sufficiently large scales when a viscous fluid
224
is injected into another more viscous fluid. At these large scales, and
225
in the absence of a stabilizing gravitational effects, two-dimensional
226
flows exhibit tree-like branched displacement structures with a fractal
227
dimension Dv = 1.6 (M˚ aløy et al., 1985). M´eheust et al. (2002) have introduced an generalized fluctuation number F =
∆ρga − Wt
aµv κ
(3)
which is the ratio of the typical total pressure drop in the fluids over one pore , including both viscous and gravitational pressure drops, to the width of the capillary pressure threshold distribution Wt . The experiments by M´eheust et al. (2002), identical to those by Birovljev et al. (1991) but performed at larger flow rates and therefore under significant viscous effects, showed that the characteristic width of the rough interface parallel to the macroscopic flow could be characterized with the fluctuation number according to an equation analog to Eq.(1): −ν
ξ/a ∝ F ν+1
(4)
228
where the exponent ν/(1 + ν) = 4/7 is consistent with percolation the-
229
ory (ν = 4/3) in 2D. Since the viscous pressure field is not homogeneous
9
230
like the gravitational field, this result is not obvious. Note that in this
231
case ξ can be interpreted as the length scale at which the sum of the
232
viscous and gravitational pressure drop becomes of the same order of
233
magnitude as the spatial fluctuations of the capillary pressure thresh-
234
old. In terms of fluid-fluid interface, ξ corresponds to the crossover
235
scale between capillary fingering structures at small scale and the sta-
236
bilized structure, which is linear (dimension 1) at large scales. When
237
the displacement is large enough for viscous forces to play a role, the
238
fractal dimension typical of viscous fingering structures is also seen at
239
intermediate scales (M´eheust et al., 2002). Even in the case where the
240
two fluids involved have the same viscosity, the width of the front was
241
found to be consistent with Eq.(4) (Frette et al., 1992). As observed in
242
the experiments of Frette et al. (1992) the effect of trapping turns out,
243
at least in two dimensions, to be of central importance. The trapped
244
islands result in a decrease in the relative permeability of the invaded
245
fluid, which is equivalent to having a fluid with a higher viscosity. This
246
is the effect responsible for the well-known decreasing dependence of
247
a soil’s matric potential on its water content. This result is consistent
248
with the scaling relation Eq.(4), which is expected from theoretical
249
arguments for percolation in a stabilizing gradient (Xu et al., 1998;
250
Wilkinson, 1984; Lenormand , 1989). Note thate other scaling rela-
251
tions have been derived theoretically and observed experimentally by
252
other authors, as reported by Xu et al. (1998). As mentioned previously, when a viscous fluid is injected into a 2D medium filled with a more viscous fluid, viscous fingers occur. The scaling of the finger width was studied experimentally by Løvoll et al. (2004) and Toussaint et al. (2005). The measurements were found to be consistent with a scaling law in the form ξ/a ∝ Ca−1
253
(5)
This result is different from the scaling laws that can be explained
10
254
from the theory of percolation in a gradient, as observed with stabi-
255
lizing viscous forces.
256
Toussaint et al. (2005), the destabilizing field is highly inhomogeneous
257
and screened around the fingers, which may explain why the behaviour
258
expected from percolation in a gradient is not observed, but a rather
259
simpler one instead. Note however that scaling laws based on the the-
260
ory of percolation in a gradient are still expected for some types of
261
unstables flows (Xu et al., 1998).
In the experiments of Løvoll et al. (2004) and
262
The scaling law (5) can be explained from a simple mean field ar-
263
gument: consider an approximation for the pressure field for which
264
the pressure gradient is homogeneous around the mobile sites at the
265
boundary between the two fluids; the pressure difference between two
266
points at a distance l is l∇P ∼ lCaγ/a2 . This relation holds not only
267
when the viscosity of one fluid can be neglected with respect to that
268
of the other one (for example, for air and water), in which case the
269
definition of the capillary number if given by Eq. (2); it also holds in
270
the general case of two viscous fluids, in which the capillary number
271
can be defined from Eq. (2) by replacing the viscous pressure drop by
272
a the difference in viscous pressure drop between the two fluids. If the
273
viscous pressure drop between these two points exceeds the character-
274
istic random fluctuations of the capillary pressure threshold from one
275
pore throat to another along the interface, then the viscous pressure
276
field is dominant in determining at which of the two points considered
277
new pores are going to be invaded. On the contrary, if the random
278
fluctuations of capillary threshold exceed the viscous pressure drop be-
279
tween the two points, then this random pressure difference component
280
dominates. Its magnitude is of the same order as the average capillary
281
pressure value, γ/a. Hence, capillary effects are expected to dominate
282
for scales l such that lCaγ/a2 < γ/a, whereas viscous effects will dom-
283
inate for larger scales, such that lCaγ/a2 > γ/a: this explains the
284
observed cross over scale ξ = a/Ca between the structures character-
11
285
istic of capillary fingering and those characteristic of viscous fingering
286
.
287
4
288
to viscous fingering
289
In many situations of non-miscible biphasic flow in porous media, the
290
flow gets organized in fingering structures (preferential paths); in these
291
unstable configurations, the fluids arrange in fractal geometries with
292
nontrivial fractal dimensions depending on the observation scale, and
293
the scale range over which each dimension is observed depends on the
294
imposed boundary conditions (such as the macroscopic fluxes). This
295
microscopic structuration has far reaching consequences for the up-
296
scaled relationships between, for example, saturation and the pressure
297
difference between the two phases.
298
An example of upscaling from capillary
An example of such a situation has been mentioned briefly earlier
299
in this review:
if gravity is negligible, only capillarity and viscosity
300
play a role on the flow ; when a lowly viscous fluid displaces a more
301
viscous one, the fluid-fluid interface is unstable due to viscous pressure
302
gradients increasing at the most advanced parts of the invader, so that
303
fingers naturally arise. The development of such an interface instabil-
304
ity occuring during drainage was studied optically in Hele Shaw cells,
305
at several controlled injection rates (Løvoll et al., 2004; Toussaint et
306
al., 2005). Air was injected into an artifical porous medium composed
307
of a monolayer of immobile glass beads sandwiched between two glass
308
plates, and initially filled with a wetting dyed glycerol-water solution.
309
The Hele-Shaw cell dimensions were denoted L × W × H, H = 1 mm
310
being the cell thickness as well as the typical glass bead diameter; flow
311
was imposed along the length L, with impermeable lateralboundaries
312
defining a channel of width W (see Fig. 1). It was shown that the path-
313
way of the air, defined as the locations where the occupancy probability
12
Ca = 0.022
inlet
outlet
x
L-x Ca = 0.033
λW
outlet
inlet
a/Ca
x
W
λW
a/Ca
x
outlet
inlet
Ca = 0.062
L Figure 1: Invasion structure of a lowly viscous fluid (white) into a much more viscous one (dark grey) during drainage in an artificial 2D porous medium of width W and extent L, at three different extraction speeds. The position of the invasion tip is denoted x. Characteristic crossover scales between fractal regimes, λW and a/Ca, separate a straight finger structure, a viscous fingering geometry, and a capillary fingering geometry, down to the pore scale a. 314
exceeds half its maximum value, was a finger of width λW positioned
315
in the centre of the channel, with λ = 0.4. This was attributed to a
316
similarity between the process of selection of the the pore throats to be
317
invaded and a Dielectric Breakdown Model with η = 2, as was shown
318
by computing the average invasion speed taking into account the width
319
of the capillary threshold distribution (Toussaint et al., 2005). This
320
structure is illustrated in Fig. 1.
321
From approximations on the shape of the pressure around this fin-
322
ger, mostly controlled by the viscous pressure drop, one can derive an
323
upscaled pressure-saturation relation (Løvoll et al., 2010).
13
324
Indeed, the pressure presents to first order a linear viscous pressure
325
drop from the tip of the invasion cluster, at position x, to the outlet
326
of the system, at position L. Over the rest of the system, the pressure
327
gradient is screened by the finger, rendering the pressure in the wetting
328
fluid essentially constant at a value close to the sum of the air pressure
329
and the entrance pressure γ/a. Hence, the pressure difference between
330
the two phases, with a pressure in the wetting fluid Pw measured at
331
the outlet, the one in the non wetting phase equal to the atmospheric
332
pressure Pn.w. , and a correction to this entrance pressure, writes as ∆P ∗
=
Pw. − Pn.w. − γ/a = (L − x)∇P
(6)
=
(L − x)∆Pvisc /a
(7)
=
(L − x)∆Pcap Ca/a
(8)
=
(L − x)
γCa a2
(9)
333
where ∆Pvisc and ∆Pcap are considered at the pore scale. Thus, there
334
is a linear relationship between the viscous pressure drop accross the
335
cell and the distance between the finger tip and the outlet. Besides, the relationship between saturation and capillary number can be inferred from the fractal structure of the non-wetting invading fluid. At scales above the width λW , the finger is a linear structure of dimension 1. Between the scale λW and a/Ca, the structure has a viscous fingering geometry of fractal dimension Dv ∼ 1.5 (M˚ aløy et al., 1985; Løvoll et al., 2004). Between the crossover scale a/Ca and the pore scale a, the structure has a capillary fingering geometry of fractal dimension Dc = 1.83. Hence, the total number of pores invaded by the non viscous fluid can be evaluated as a function of these fractal dimensions, the ratio of the finger length to its width, x/(λW ), and the ratios of the latter length to the two others lengths, the crossover
14
length and the pore size. This leads to:
Nn.w.
x = λW
λW aCa−1
Dv
aCa−1 a
Dc (10)
Together with the relationship between the total number of pores and the characteristic model dimensions,
Ntot =
LW a2
(11)
and the relationship between the wetting phase and non-wetting phase saturations, Sn.w. = 1 − Sw. = 336
Nn.w. , Ntot
(12)
Eq. (10) leads to
1 − Sw. = Sn.w. = λ
Dv −1
a 2−Dv a2 ∆P ∗ Dv −Dc 1− Ca W γLCa
(13)
337
This relationship allows to collapse all the pressure difference
338
curves measured as a function of saturation in the set of experiments
339
performed by Løvoll et al. (2010) onto a unique master curve, for
340
capillary numbers ranging from around 0.008 to 0.12. Since P ∗ =
341
a2 ∆P ∗ /(γLCa) is nothing else than the capillary pressure measured
342
at the scale L of the experimental model, Eq. (13) in effect relates that
343
capillary pressure to the imposed seepage velocity. In other words, it
344
defines the dependence on Darcy flow velocity of what is commonly de-
345
noted dynamic capillary pressure, measured at scale L. This example
346
shows how both viscous and capillary effects play a role in constrain-
347
ing the geometry of the invasion structures, resulting in a dynamic
348
capillary pressure, as it is traditionally called (Hassanizadeh, 2002),
349
that is simply due to the upscaling of the invasion structure, with only
350
capillary and viscous effects seen at the REV scale, and without any
351
dynamic capillary/wetting effects occuring at the pore scale.
15
Ca = 0.009 Ca = 0.022 Ca = 0.029 Ca = 0.033 Ca = 0.054 Ca = 0.058 Ca = 0.060 Ca = 0.062 Ca = 0.079 Ca = 0.110
4000
Δ P*(Pa)
3000 2000 1000 0 0
0.05
0.1
S n.w.
0.15
0.2
0.25
Figure 2: Dependence of the pressure difference between the two phases and the saturation of the invading fluid, at different injection speeds. Figures 2 and 3 illustrate respectively the raw measurements at several injection speeds and how Eq. (13) allows to collapse these curves of saturation versus pressure The viscous pressure drop accross the cell drops linearly as the finger progresses into the cell, from a maximum value at the beginning of the invasion of γLCa/a2 , down to 0 at breakthrough of the invasion finger. In the previous equation, the wetting saturation is indeed initially 1 as it should be at initial total saturation, but we also obtain the final and maximum value of the residual wetting saturation as 1 − Sw.r. = Sn.w.r. = λDv −1
a 2−Dv CaDv −Dc . W
(14)
352
This relation between the residual saturation and the capillary number
353
is indeed consistent with the observed residual saturations, as shown
16
Snw(tbt)
1.25
1
0.2 0.1 0
0.75
0
0.1
0.2
P
*
Ca
0.5
Ca = 0.009 Ca = 0.022 Ca = 0.029 Ca = 0.033 Ca = 0.054 Ca = 0.058 Ca = 0.060 Ca = 0.062 Ca = 0.079 Ca = 0.110 * * P =1-S
0.25
0 0
0.2
0.4
0.8
0.6
1
1.2
*
S
Figure 3: The collapse of the relationship between the reduced pressure difference (between the two phases), P ∗ = a2 ∆P ∗ /(γLCa), and the reduced a −2+Dv saturation of the invading fluid, S∗ = λ1−Dv W CaDc −Dv Sn.w. , at different injection speeds, shows the influence of the structure on the upscaling. Dotted curve: prediction. Inset: Residual saturation at breakthrough.
17
354
in the inset of Fig.3.
355
Besides the relation between saturation and the macroscopic pres-
356
sure difference between the phases, other macroscopic relations can
357
be obtained via upscaling, as e.g., in some situations, the relative per-
358
meability. For example, in other experiments where both fluids were
359
injected at the same time, the trapped structures of wetting fluids
360
were observed to be fractal up to a certain cutoff depending on the
361
imposed flux (Tallakstad et al., 2009). The upscaling explaining the
362
cutoff and the structures allowed to explain the measured scaling law
363
of the relative permeability of the viscous phase as a function of the
364
imposed flux, κrel ∼ Ca−1/2 .
365
5
366
We have discussed the local flow structures that are observed ex-
367
perimentally during drainage in a disordered porous medium. They
368
are fractal, with a fractal dimension that depends on the observation
369
scale. At small scales, capillary fingering exhibits a fractal dimension
370
of 1.8 for two-dimensional media, and between 2 and 2.6 for three-
371
dimensional media. At larger scales a branched structure characteris-
372
tic of viscous fingering is seen, with a fractal dimension 1.6 for two-
373
dimensional systems. The crossover between the two behaviors occurs
374
at a length scale for which the differential viscous pressure drop equals
375
the typical capillary pressure threshold in the medium. This means
376
that for horizontal flow, unstable viscous fingering is always seen at
377
large enough scales, even if the medium exhibits no permeability het-
378
erogeneities at the Darcy scale. From the definition of the crossover
379
length, it follows that it scales as the inverse of the capillary number,
380
which explains why experiments performed at a given experimental
381
scale and at very slow flow rates have evidenced capillary fingering,
382
while those performed at very large flow rates have evidenced viscous
383
fingering. As for the effect of gravity, it can be to either destabilize
Conclusion
18
384
or stabilize the interface, depending on which fluid is the densest. In
385
the latter case, it acts against capillary effects and, when the displac-
386
ing fluid is the most viscous, against the destabilizing viscous forces,
387
resulting in an amplitude of the interface roughness that scales as a
388
power law of the generalized fluctuation number (or generalized Bond
389
number). In horizontal two-dimensional flows, viscous fingering is ob-
390
served to occur up to another characteristic length that is a fixed frac-
391
tion of the width of the medium. Upscaling of the local flow structures
392
is possible once one knows the fractal dimensions typical of the flow
393
regimes, and the relevant length scale range for each of them. We have
394
given an example of how the measured capillary pressure can be re-
395
lated theoretically to water saturation, a relation that is confirmed by
396
measurements. In that example, the capillary pressure measured at
397
the scale of the experimental setup exhibits dynamic features, i. e., a
398
dependence on the flow rate, that is fully explained by the geometry
399
of the upscaling, without any dynamic effects in the physical capillary
400
pressure as defined at the pore/interface scale.
401
6
402
This work was supported by the CNRS through a french-norwegian
403
PICS grant, the Alsace region through the REALISE program, and
404
the Norwegian NFR.
405
References
406
Auradou H., K.J. M˚ aløy, J. Schmittbuhl, A. Hansen, and D. Bideau.
407
1999. Competition between correlated Buoyancy and uncorrelated
408
capillary effects during drainage, Phys. Rev. E. 60:7224-7234.
Acknowledgements
409
Benremita H., G. Sch¨ afer. 2003. Transfert du trichlor´ethyl`ene en milieu
410
poreux ` a partir dun panache de vapeurs. C. R. Mecanique 331, 835–
411
842.
19
412
Birovljev A., L. Furuberg, J. Feder, T. Jssang, K.J. M˚ aløy, and A.
413
Aharony. 1991. Gravity invasion percolation in two dimensions: ex-
414
periment and simulation. Phys. Rev. Lett. 67(5):584-587.
415
Birovljev A., G. Wagner, P. Meakin, J. Feder, and T. Jøssang. 1995.
416
Migration and fragmentation of invasion percolation clusters in two-
417
dimensional porous media. Phys. Rev. E 51(6):5911-5915.
418
Chandler R., J. Koplik, K. Lerman, and J.F. Willemsen. 1982. Cap-
419
illary diplacement and percolation in porous media. J. Fluid Mech,
420
119:249-267.
421
Chen S., K.H. Kim, F. Qin, and A. T. Watson. 1992. Quantitative
422
NMR imaging of multiphase flow in porous media. Magnetic Reso-
423
nance Imaging 10(5):815-826.
424
Cheng, X., L. Xu, A. Patterson, H.M. Jaeger, and S.R. Nagel. 2008.
425
Towards the zero-surface-tension limit in granular fingering instabil-
426
ity. Nature Physics. 4(3):234-237. doi:10.1038/nphys834.
427
Chevalier, C., A. Lindner, M. Leroux, and E. Cl´ement. 2009. Mor-
428
phodynamics during air injection into a confined granular sus-
429
pension. Journal of Non-newtonian fluid mechanics 158(1-3):63-72.
430
doi:10.1016/j.jnnfm.2008.07.007.
431
Chuoke R. L., P. Van Meurs, and C. Van der Poei. 1959. The instability
432
of slow, immiscible, viscous liquid-liquid displacement in permeable
433
media, Trans. AIME 216:188-194.
434
de Gennes P.G., and E. Guyon. 1978. Lois generales pour l’injection
435
d’un fluide dans un milieu poreux aleatoire. J. Mec (France), 17:403-
436
432.
437
Dridi, L. and Pollet, I. and Razakarisoa, O. and Schfer, G. 2009.
438
Characterisation of a DNAPL source zone in a porous aquifer us-
439
ing the Partitioning Interwell Tracer Test and an inverse modelling
440
approach, J Contamin. Hyd.
20
441
Frette V., J. Feder, T. Jøssang, P. Meakin, and K.J. M˚ aløy. 1994. Fast,
442
immiscible fluid-fluid displacement in three-dimensional porous me-
443
dia at finite viscosity contrast. Phys. Rev. E 50:2881.
444 445
Frette V., J. Feder, T. Jøssang, and P. Meakin. 1992. Buoyancydriven fluid migration in porous media. Phys. Rev. Lett. 68:3164-3167.
446
Frette O.I., K.J. M˚ aløy, J. Schmittbuhl, and A. Hansen. 1997. Immis-
447
cible Displacement in 2D Porous media With a Viscosity Contrast
448
Equal to One. Phys. Rev. E. 55:2969-2975.
449 450
451 452
Furuberg L., K.J. M˚ aløy, and J. Feder. 1996. Intermittent behavior in slow drainage. Phys. Rev. E 53:966-977. Haines W.B. 1930. Studies in the physical properties of soil. J. Agric. Sci. 20:97-116.
453
Hassanizadeh, S.M. and Celia, M.A. and Dahle, H.K. 2002. Dynamic
454
effect in the capillary pressuresaturation relationship and its impacts
455
on unsaturated flow. Vadose Zone J. 1, 3857.
456
Holtzman, R. and R. Juanes. 2010. Crossover from fingering to frac-
457
turing in deformable disordered media. Phys. Rev. E 82:046305.
458
Hou J., Z.Q. Li, S.K. Zhang, X.L. Cao, Q.J. Du, and X.W. Song. 2009.
459
Computerized tomography study of the microscopic flow mechanism
460
of polymer flooding, Transp. Porous Med. 79(3):407-418.
461
Jianzhao Y., X. Luo, W. Wang, F. Chen, R. Toussaint, J. Schmit-
462
tbuhl, G. Vasseur, Y. Alan, L. Zhang. 2011. An Experimental Study
463
of Oil Secondary Migration in a Three Dimensional Porous Space.
464
A.A.P.G. Bulletin. accepted.
465
Johnsen, Ø., R. Toussaint, K.J. M˚ aløy and E.G. Flekkøy. 2006. Pat-
466
tern formation during central air injection into granular materi-
467
als confined in a circular Hele-Shaw cell. Phys. Rev. E 74:011301.
468
doi:10.1103/PhysRevE.74.011301
21
469
Johnsen, Ø., R. Toussaint, K.J. M˚ aløy, E.G. Flekkøy and J. Schmit-
470
tbuhl. 2007. Coupled air/granular flow in a linear Hele-Shaw cell”,
471
Phys Rev E 77:011301. doi:10.1103/PhysRevE.77.011301
472
Johnsen, Ø., C. Chevalier, A. Lindner, R. Toussaint, E. Cl´ement,
473
K.J. M˚ aløy, E.G. Flekkøy and J. Schmittbuhl. 2008. Decompaction
474
and fluidization of a saturated and confined granular medium by
475
injection of a viscous liquid or a gas. Phys. Rev. E 78:051302.
476
doi:10.1103/PhysRevE.78.051302
477
Kong, X.Z. M. Holzner, F. Stauffer, and W. Kinzelbach. 2011. Time-
478
resolved 3D visualization of air injection in a liquid-saturated
479
refractive-index-matched porous medium. Experiments in fluids
480
50(6):1659-1670.doi:10.1007/s00348-010-1018-6.
481
Lemaire, E., Y. O. M. Abdelhaye, J. Larue, R. Benoit, P. Levitz, and H.
482
van Damme. Pattern formation in noncohesive and cohesive granular
483
media. Fractals 1: 968-976. doi:10.1142/S0218348X93001040.
484 485
Lenormand, R. 1989. Flow Through Porous Media: Limits of Fractal Patterns, Proc. R. Soc. London, Ser. A 423:159-168.
486
Lenormand R, E. Touboul, and C. Zarcone. 1988. Numerical mod-
487
els and experiments on immiscible displacements in porous media.
488
Journal of Fluid Mech. 189:165-187.
489
Løvoll, G., Y. M´eheust, K.J. M˚ aløy, E. Aker and J. Schmittbuhl. 2005.
490
Competition of gravity, capillary and viscous forces during drainage
491
in a two-dimensional porous medium, a pore scale study. Energy
492
30:861872.
493
Løvoll, G., M. Jankov, K.J. M˚ aløy, R. Toussaint, J. Schmittbuhl, G.
494
Schaefer and Y. M´eheust. 2010. Influence of viscous fingering on
495
dynamic saturation-pressure curves in porous media. Transport in
496
Porous Media 86(1):305-324. doi:10.1007/s11242-010-9622-8
22
497
Løvoll, G., Y. M´eheust, R. Toussaint, J. Schmittbuhl and K.J. M˚ aløy.
498
Growth activity during fingering in a porous Hele Shaw cell. Phys.
499
Rev. E 70:026301. doi:10.1103/PhysRevE.70.026301
500
Mandava, S. S., A. Ted Watson, and M.E. Carl. 1990. NMR imaging
501
of saturation during immiscible displacements, AIChE 36(11):1680
502
1686.
503 504
M´eheust, Y., and J. Schmittbuhl. 2000. Flow enhancement of a rough fracture. Geophysical Research Letters 27: 29892992.
505
M´eheust Y., and J. Schmittbuhl. 2001. Geometrical heterogeneities and
506
permeability anisotropy of rough fractures. Journal of Geophysical
507
Research 106(B2): 20892012.
508
M´eheust, Y., G. Løvoll, K.J. M˚ aløy, and J. Schmittbuhl. 2002. Inter-
509
face scaling in a two-dimensional porous medium under combined
510
viscous, gravity, and capillary effects. Phys. Rev. E 66:051603.
511 512
513 514
515
M˚ aløy K.J., J. Feder, and T. Jøssang. 1985. Vicous fingering fractals in porous media, Phys. Rev. Lett. 55:2688-2691 M˚ aløy K.J., L. Furuberg, J. Feder, and T. Jøssang. 1992. Dynamics of Slow Drainage in Porous Media. Phys. Rev. Lett. 68:2161-2164 Neuville A., R. Toussaint and J. Schmittbuhl. 2010. Fracture
516
roughness and thermal exchange:
A case study at Soultz-
517
sous-forts. 2010. Comptes Rendus Geoscience 342(78):616625.
518
doi:10.1016/j.crte.2009.03.006.
519
Neuville A., R. Toussaint, and J. Schmittbuhl. 2010. Hydro-thermal
520
flows in a self-affine rough fracture. 82: 036317. doi:10.1103/Phys-
521
RevE.82.036317.
522
Neuville, A., R. Toussaint and J. Schmittbuhl. 2011. Hydraulic trans-
523
missivity and heat exchange efficiency of rough fractures: a spec-
23
524
tral approach, Geoph. J. Int. 186:10641072. doi: 10.1111/j.1365-
525
246X.2011.05126.x
526
Neuville A., R. Toussaint, J. Schmittbuhl, D. Koehn, and J. Schwarz.
527
2011. Characterization of major discontinuities from borehole cores
528
of the black consolidated marl formation of Draix (French Alps).
529
Hydrological Processes in press, doi: 10.1002/hyp.7984.
530
Neuville A., R. Toussaint, and J. Schmittbuhl. 2011. Fracture aperture
531
reconstruction and determination of hydrological properties: a case
532
study at Draix (French Alps). Hydrological Processes in press, doi:
533
10.1002/hyp.7985.
534
Niebling, M.J., E.G. Flekkøy, K.J. M˚ aløy, R. Toussaint. 2010. Mixing
535
of a granular layer falling through a fluid. Phys. Rev. E 82:011301.
536
doi:10.1103/PhysRevE.82.011301
537
Niebling, M.J., E.G. Flekkøy, K.J. M˚ aløy, R. Toussaint. 2010. Sedimen-
538
tation instabilities: impact of the fluid compressibility and viscosity.
539
Phys. Rev. E 82:051302. doi: 10.1103/PhysRevE.82.051302
540
Nsir, K., G. Sch¨ afer, R. Di Chiara Roupert, O. Razakarisoa, R. Tou-
541
ssaint. 2011. DNAPL gravity fingering in water-saturated porous
542
media: laboratory experiments and pore scale modelling, Internat.
543
J. Multiphase Flow, under revision.
544
Paterson, L., V. Hornof, and G. Neale. 1984. Visualization of a surfac-
545
tant flood of an oil-saturated porous medium, S.P.E. J, 325-327.
546
Paterson, L., V. Hornof and G. Neale. 1984. Water fingering into an oil-
547
wet porous medium saturated with oil at connate water saturation.
548
Rev. Inst. Franc. Pert. 39:517-522.
549
Sandnes, B., E.G. Flekky, H.A. Knudsen, K.J. Mly, and H. See. 2011.
550
Patterns and flow in frictional fluid dynamics, Nat. Commun. 2:288,
551
doi:10.1038/comms1289
24
552
Tallakstad, K.T., H.A. Knudsen, T. Ramstad, G. Løvoll, K.J. M˚ aløy,
553
R. Toussaint and E.G. Flekkøy. 2009. Steady-state two-phase flow in
554
porous media: statistics and transport properties. Phys. Rev. Lett.
555
102:074502.
556
Tallakstad, K.T., G. Løvoll, H. A. Knudsen, T. Ramstad, E. G.
557
Flekkøy, and K. J. M˚ aløy. 2009. Steady-state, simultaneous two-
558
phase flow in porous media: An experimental study. Phys. Rev. E
559
80:036308.
560
Toussaint, R., G. Løvoll, Y. M´eheust, K.J. M˚ aløy and J. Schmittbuhl.
561
2005. Influence of pore-scale disorder on viscous fingering during
562
drainage. Europhys. Lett. 71: 583. doi:10.1209/epl/i2005-10136-9
563
Varas, G., V. Vidal, and J.C. Geminard. 2011. Morphology of air in-
564
vasion in an immersed granular layer. Phys. Rev. E 83(6):061302.
565
doi:10.1103/PhysRevE.83.061302
566
Vinningland, J.L., Ø Johnsen, E.G. Flekkøy, R. Toussaint and
567
K.J. M˚ aløy. 2007. A granular Rayleigh-Taylor instability:
568
periments and simulations”,
569
doi:10.1103/PhysRevLett.99.048001
Phys. Rev. Lett.,
99,
ex-
048001.
570
Vinningland, J.L., Ø. Johnsen, E.G. Flekkoy, R. Toussaint and
571
K.J. M˚ alø. 2007. Experiments and Simulations of a gravita-
572
tional granular flow instability, Phys. Rev. E, 76, 051306.
573
doi:10.1103/PhysRevE.76.051306
574
Vinningland, J.L., Ø. Johnsen, E.G. Flekkøy, R. Toussaint and K.J.
575
M˚ aløy. 2010. Influence of particle size in Rayleigh Taylor granu-
576
lar flow instability, Phys. Rev. E 81, 041308. doi: 10.1103/Phys-
577
RevE.81.041308
578
Vinningland, J.L., R. Toussaint, M. Niebling, E.G. Flekkøy, and K.J.
579
M˚ aløy. 2011. Family Viscek scaling of detachment fronts in Granu-
25
580
lar Rayleigh-Taylor instabilities during sedimentating granular/fluid
581
flows, European Physical Journal Special Topics. submitted
582
Wagner G., A. Birovljev, J. Feder, and T. Jøssang. 1997. Fragmen-
583
tation and migration of invasion percolation clusters: Experiments
584
and simulations, Phys. Rev. E 55:7015-7029.
585 586
587 588
589 590
591 592
593 594
Wilkinson D. 1984. Percolation model of immiscible displacement in the presence of buoyancy forces. Phys. Rev. A 30:520-531. Wilkinson D. 1986. Percolation effects in immiscible displacement, Phys. Rev. A 34:1380-1391. Wilkinson D., and J.F. Willemsen, 1983. Invasion percolation: a new form of percolation theoryJ. Phys. A 16:3365-3376. Xu, B., Y.C. Yortsos, and D. Salin. 1998. Invasion percolation with viscous forces. Phys. Rev. E 57:739-751. Zimmerman R., and G. Bodvarsson. 1996. Hydraulic conductivity of rock fractures. Transport in Porous Media 23(1): 130.
26
