SPE 166232

Numerical Challenges in Foam Simulation: A Review

W. R. Rossen, Delft University of Technology

Copyright 2013, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Annual Technical Conference and Exhibition held in New Orleans, Louisiana, USA, 30 September–2 October 2013.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not been
reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its
officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohibited. Permission to
reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
We review challenges to accurate simulation of foam enhanced oil recovery, with a focus on numerical issues. Foam responds
in an abrupt, nonlinear way to changes in water saturation, surfactant concentration, and oil saturation, in ways that cause
fluxes to fluctuate in time and space. As the grid is refined these effects have smaller impact on the overall process but
execution of the simulation slows. In addition, in simulations of foam with oil, consecutive grid blocks can lie on opposite
sides of a strong foam/weak foam boundary on the composition diagram. Because by definition foam is an interaction
between gas and water, the naming of phases (gas or oil) in a compositional simulation of a miscible EOR process can have
significant effect on the simulation of a foam flood. Numerical dispersion of surfactant concentration is a problem, but
attempts to minimize its effect can lead to other numerical artifacts. Because foam is so sensitive to water saturation and
capillary pressure, capillary effects are important, especially in finely laminated formations. "Population-balance" foam
simulators, which represent the complex dynamics of bubble creation and destruction along with the effect of foam on gas
mobility, face additional challenges with instability and slow run times, especially for models that represent the multiple
steady states seen in the laboratory.
We collect and review the various numerical challenges to foam simulation. Some of these problems are largely cosmetic,
giving for instance fluctuating fluxes and pressure gradient but no significant effect on final recovery. Others do severely
influence the whole progress of the flood. We discuss the origin of the challenges, how to recognize them, how they can be
mitigated, and whether they arise from a correct representation of foam physics or the unintended result of attempts to solve
other numerical problems.

Introduction
Injected gas (CO2, hydrocarbon gas, N2, or steam) can be very effective at displacing oil in enhanced oil recovery (EOR)
processes, but ultimate recovery suffers from poor sweep efficiency (Lake, 1989). Poor sweep efficiency arises from reservoir
heterogeneity, viscous instability, and gravity override of gas. Foam can address all three causes of poor sweep efficiency
(Schramm, 1995; Rossen, 1996). Foam is a dispersion of gas separated by water films called lamellae that separate the gas
bubbles; the lamellae are stabilized by surfactant. Foam thus requires the presence of gas, water and surfactant. Representing
the foam process accurately with reservoir simulation is essential to design and optimization of foam EOR processes to
specific reservoirs.
When the gas is dispersed in foam, gas mobility is greatly reduced. Darcy's law continues to apply to all phases present.
Many studies find that the viscosity and relative-permeability function for the aqueous phase (which for simplicity we refer to
here as "water") krw(Sw) are unaltered by foam (Bernard et al., 1965; Huh and Handy, 1989; Vassenden and Holt, 2000). The
mobility of gas is greatly reduced, in some cases by a factor of tens of thousands (Cheng et al., 2000; Boeije and Rossen, 2013;
Rossen and Boeije, 2013).
There are two fundamental approaches to representing the effect of foam on gas mobility. Population-balance models
(Falls et al., 1988; Friedman et al., 1991; Kovscek and Radke, 1994; Zitha, 2006; Kam et al., 2007) introduce lamella density
(number of lamellae per unit volume of gas phase) as a separate variable, and perform a balance on lamellae at each location in
the formation, along with material balances on water, gas, surfactant, and oil. Thus there is an additional partial differential
equation to be solved at each location and time step along with those for saturations of the phases. This balance on lamellae
includes convection, generation and destruction (as well as bubble trapping and liberation if those processes are represented
explicitly). The model then represents gas mobility as a function of lamella density as well as other factors (superficial velocity
or capillary number, for instance, if non-Newtonian gas mobility is included).
2 SPE 166232

Local-equilibrium (LE) models (Law et al., 1989; Patzek and Myhill, 1989; Kular et al., 1989; Fisher et al., 1990; Islam
and Farouq-Ali, 1990; Mohammadi and Coombe, 1992; Cheng et al., 2000) assume that the processes of lamella creation and
destruction are always and everywhere at local steady state. It is possible to adapt a population-balance model to LE by setting
the expressions for lamella creation and destruction equal to each other (Myers and Radke, 2000; Kam et al., 2007; Chen et al.,
2010). Most LE models, however, represent the effect of bubble size implicitly in relations for gas mobility as a function water
and oil saturations, surfactant concentration in the aqueous phase, and other factors.
The LE foam model in the STARS simulator (Computer Modeling Group, 2006) is a widely used, and is representative of
the other LE approaches (Cheng et al., 2000). In this approach all the effects of foam on gas mobility are represented for
f
simplicity as a change in gas relative permeability. Gas relative permeability with foam k rg is reduced from its value without
foam krg(Sw) by a factor FM, which is in turn inversely related to a the product of a constant factor fmmob and factors Fi, which
portray various effects that reduce the effect of foam (i.e., increase gas mobility):
krgf  krg (Sw ) FM ................................................................................................................................................................................................ 1

1
FM  . .......................................................................................................................................... 2
1  fmmob F1 F2 F3 F4 F5 F6
The functions F1, F2, etc., range from 0 to 1 depending on water saturation, oil saturation or composition, pressure gradient,
permeability, surfactant concentration in the water, etc. Here we are interested in the functions that account for the effects of
water saturation, oil saturation, surfactant concentration, which we denote here for simplicity Fw, Fo and Fs. Thus for our
purposes
1
FM  . ........................................................................................................................................................ 3
1  fmmob Fw Fs Fo
Often these functions Fw, Fo and Fs are represented as abrupt, or even discontinuous, jumps between 0 and 1. These abrupt
jumps create challenges for accurate foam simulation, which is the subject of this paper.
Foam can be injected in at least four ways:
1. In co-injection, gas and aqueous surfactant solution are injected simultaneously from a single well. Foam forms in
the surface facilities where the fluids meet, in the tubing, or shortly after the fluids enter the formation.
2. In surfactant-alternating-gas or SAG injection, gas and surfactant solution are injected in separate slugs from a
single well. Foam forms in the formation where gas meets previously injected surfactant solution, or when
surfactant solution meets previously injected gas.
3. It is possible to dissolve some surfactants directly into supercritical CO2 (Le et al., 2008). Then there is no need
to inject aqueous surfactant solution; injected CO2 with dissolved surfactant forms foam as it meets water in the
formation.
4. Surfactant solution and gas can be injected simultaneously, but from different sections of a vertical well (gas
injected below the surfactant solution), or from parallel horizontal wells (gas injected from the lower well) (Stone,
2004; Rossen et al., 2010). As far as we know, this approach has not been tested with foam in the field, and we do
not address it further here.
As shown below, simulating each of these injection methods involved particular challenges.
In this paper we consider whether current simulators accurately solve the equations for foam displacements as they are
currently represented in models. The issue of whether the models themselves are correct is a vital issue but outside the scope
of this study. We focus largely, but not exclusively, on simulating foam processes without oil, because most simulation
research has focused on this challenge. Representing a foam processes without oil is of course a necessary first step toward
representing a foam process with oil. In analyzing challenges to foam simulation we refer frequently below to exactly solutions
obtained with the Method of Characteristics (MOC; also known as fractional-flow theory) for one-dimensional displacements
fitting certain assumptions (incompressible phases, LE, absence of dispersion). The application of the MOC to foam EOR is
described in detail elsewhere (Zhou and Rossen 1995; Shan and Rossen, 2004; Mayberry et al., 2008; Namdar Zanganeh et al.,
2011; Ashoori et al., 2010, 2011a).

Challenges to Foam Simulation

Effect of Water Saturation
Gas, as the nonwetting phase, is at higher pressure than water in the geological formation; this difference is (gas-water)
capillary pressure (Lake, 1989). Capillary pressure draws water out of lamellae; therefore high capillary pressure is destructive
to foam. More specifically, foam collapses abruptly at the "limiting capillary pressure" Pc* (Khatib et al., 1988). The value of
Pc* depends on surfactant formulation, surfactant concentration, temperature, pressure, formation properties, and other factors,
including possibly total superficial velocity. Since in geological formations capillary pressure is linked to water saturation
through the capillary-pressure function Pc(Sw), the abrupt collapse at Pc* is equivalent to a collapse at a water saturation Sw* =
SPE 166232 3

Sw(Pc*). This collapse is represented in the function Fw(Sw) (Eq. 3) as an abrupt change from 1 to 0 as Sw decreases below Sw*.
STARS defines Fw as follows:
arctan( epdry ( S w  fmdry ))
Fw  0.5  . ........................................................................................................................................ 4

Here Fw changes from 0 to 1 for Sw in the vicinity of parameter fmdry, and parameter epdry governs the abruptness of the
transition.
Cheng et al. (2000) choose model parameters that give a very large reduction in gas mobility with foam and abrupt collapse
at Sw*, based on the observation that pressure gradient is nearly independent of gas superficial velocity when Sw is near Sw*.
Figure 1 shows the gas relative-permeability function with a large reduction in gas mobility fmmob and large value of epdry,
using parameter values similar to those of Cheng et al. Ma et al. (2013) show that one can fit behavior at a single total
superficial velocity adequately with a less abrupt change in Fw(Sw) (smaller value of epdry). The effect of an abrupt transition
in Fw on the simulation depends on the method of injection.

Figure 1 (left). Water relative permeability and gas relative permeability with and without foam and for one foam model, from Leeftink
et al. (2013).
Figure 2 (right). Total relative mobility rt as a function of water saturation for the model in Fig. 1. In a SAG process, there is a shock
from the initial state ahead of the shock past the water saturations with lowest mobility. In a finite-difference simulation, each grid
block passes through all of the intermediate water saturations as the foam front enters the grid block.

SAG Injection: Complete Collapse of Foam Depending on Function Fw

A variety of functional forms for Fw have been proposed in different foam models (Dong and Rossen, 2007). Some
functional forms are incapable of representing a SAG process with an effective foam bank, because the functional form gives a
point of tangency at a state of complete foam collapse. One should check the form of the Fw function for this behavior if a
SAG process is to be simulated; see Dong and Rossen (2007) for details.

SAG Injection: Fluctuating Injectivity During Gas Injection

Figure 2 shows total relative mobility rt with foam (the sum of water and gas mobilities kri/i) as a function of water
saturation Sw for the model in Fig. 1. During gas injection in a SAG process, there is a shock front at the leading edge of the
gas bank from the initial state of the reservoir to a point of very low water fractional flow, where foam has partially collapsed.
Thus the water saturations that would have the greatest mobility reduction with foam are not experienced in the reservoir
except within the shock front of negligible width. (For more discussion of the traveling wave, see Rossen and Bruining, 2007;
Ashoori et al., 2011a). In a finite-difference simulation, however, each grid block at the foam front passes through all those
saturations, and thus the mobility reduction at the foam front is overestimated. Figure 3 illustrates this effect: there is at least
one grid block at the leading edge of the foam bank with mobility computed to be below that anywhere in the foam bank itself.
As this grid block passes through the minimum in mobility in Fig. 2, the effect on overall injectivity can be large, as discussed
further below. In extreme cases where the foam collapses completely behind this shock (as discussed in the preceding
paragraph) this one grid block can mask the problem and give diversion of gas flow in the simulation that would not occur in
reality; see Fig. 4. The fluctuating injectivity as foam advances and then dries out in each grid block is one possible cause of
multiple local optima found in a study of adjoint-based optimization of surfactant-slug size in SAG flooding (Namdar
Zanganeh et al., 2012).
This problem reflects the finite-difference formulation of the foam displacement, not an inability of the simulator to solve
the finite-difference equations. The fluctuation in injectivity is an accurate solution of the foam model equations on a finite-
difference grid, where each new grid block must pass through the minimum in mobility illustrated in Fig. 2. The fluctuations
are not eliminated, for instance, by reducing time-step size or tightening the convergence criterion in the simulation.

SAG Injection: Calculating Injectivity

The grid block with the injection well must likewise pass through the minimum in mobility in Fig. 2. Injectivity is
conventionally calculated assuming a uniform saturation and mobility in the injection-well grid block using the Peaceman
4 SPE 166232

Equation. Therefore, for a time, injectivity in a simulation of a SAG process is extremely poor. In reality, the near-well region
crucial to injectivity rapidly dries out and injectivity is much greater than estimated in a finite-difference simulation. Figures
5 and 6 compare injectivity as it would be calculated in a finite-difference simulation with that calculated more accurately
using the Method of Characteristics (Leeftink et al., 2013). The rise in injection well pressure in the finite-difference
simulation, at its maximum, is over 100 times any that would actually be experienced in the near-wellbore region. If injection
pressure were limiting rather than injection rate, it would take about 40 times as long to inject one grid-block pore volume of
gas in the simulation as it should, assuming injectivity is dominated by the injection-well grid block.

Figure 3 (left). Total relative mobility during gas injection in a SAG process, from Shan and Rossen (2004).
Figure 4 (right). Example from Shan and Rossen (2004) where the form of the Fw function gives complete collapse of foam behind the
shock front at the leading edge of the foam bank (cf. Dong and Rossen, 2007) - in other words, no foam anywhere. Here water
saturation is plotted at the same time in three simulations with different grids. The low mobility in the grid block at the leading edge of
the gas bank (cf. Fig. 3) is enough to prevent gravity override of gas, even as the grid is refined.

1
0.9
50‐m grid block
PV gas injected

0.8
0.7
5‐m grid block
0.6
0.5
0.4
0.3
Peaceman Equation
0.2
0.1
0
0 0.5 time 1 1.5 2

Figure 5 (left). Dimensionless injection pressure as a function of dimensionless time for gas injection in SAG, from Leeftink et al.
(2013). Dimensionless injection pressures is the rise in injection-well pressure divided by that for injecting water with Sw = 1 in the
reservoir; dimensionless time is grid-block pore volumes injected. Two grid-block sizes are shown: 50 m radius (dotted line) and 5 m
radius (dashed line). For the simulation, the grid block size does not matter in this dimensionless plot.
Figure 6 (right). Cumulative injection in terms of grid-block pore volume of gas for the case in Fig. 5, but with fixed injection pressure
rather than fixed injection rate: MOC solution for grid blocks of 5-m and 50-m radii and Peaceman equation (which gives same
solution for either grid-block size). One unit of time here represents the time it would take to inject one grid-block pore volume of
water with 100% water saturation in the grid block at the fixed injection pressure. It takes over 40 time units to injection one pore
volume of gas as computed by the Peaceman Equation.

Related to this issue of low injectivity caused by the finite-difference solution of flow equations is the issue of foam
mobility near the well. The functional form of Fw in STARS (Eq. 4) does not give complete foam collapse (Fw = 0) at any
value of Sw. It is not clear what foam mobility should be at the wellbore, but if foam does collapse there, injectivity calculated
using Eq. 4 could be substantially in error. Model fits to lab data at larger water saturations using Eq. 4 can predict that gas
mobility is reduced by a factor of 1000 even at irreducible water saturation Swr at the wellbore (Rossen and Boeije, 2013).

Fine-Scale Capillary Effects

Since foam is so sensitive to water saturation and capillary pressure, capillary effects on foam near abrupt transitions in
permeability can be significant. Falls et al. (1988) showed how foam is created as gas flows across an abrupt transition from
low to high permeability; see also Yortsos and Chang (1990), van Lingen et al., (1996), Rossen (1999) and Tanzil et al.
(2002). Similarly, in flow parallel to an abrupt layer boundary, capillary pressure can draw water from the higher-permeability
layer and weaken or destroy foam in a narrow zone near the boundary (Rossen and Lu, 1997; Cheng et al., 2000). In both cases
SPE 166232 5

the affected region is small in width but influences flow over a much larger scale. Accurate modeling of the effects with larger
grid blocks remains a challenge.

Foam Injection: Fluctuating Pressure Gradient and Non-Newtonian Mobility

In principle, the same issues of fluctuating pressure gradient in the grid block at the leading edge of the foam bank would
affect simulations of foam injection as well as SAG, but the effects are less severe because mobility reduction is lower
throughout the foam bank, all the way back to the injection well. The minimum in mobility shown in Fig. 2 corresponds to the
boundary between the "low-quality" and "high-quality" flow regimes (Cheng et al., 2000; Boeije and Rossen, 2013; cf. also
Ma et al., 2013, who assume a less abrupt transition between regimes). The injected foam quality then in general could lie on
either side of the minimum in mobility. If it lies in the high-quality regime (to the left of the minimum in Fig. 2), there is a
shock from the initial state, past the minimum in mobility to the injected foam state. There would be some fluctuations in
mobility at the foam front as it advances, but not as severe as in a SAG flood.
Foam in the low-quality regime is strongly shear-thinning (Alvarez et al., 2001); in such a case using the Peaceman
equation in the injection-well grid block can lead to underestimation of injectivity; see Leeftink et al. (2013).

Effect of Water Saturation: Mitigation

The sensitivity of foam to capillary pressure and water saturation introduces two problems: individual grid blocks
experience mobilities and saturations not present in the displacement, except within the shock (Fig. 2); abrupt changes in
mobility in individual grid blocks introduce oscillations that are unsightly and may challenge the stability of the simulator.
Reducing the magnitude of the gas-mobility reduction with foam (e.g., fmmob) reduces both problems, if a smaller value is
justified by the data. A less-abrupt change in gas mobility near Sw* (smaller epdry) reduces the oscillations and speeds the
execution of the simulator, again if justified by the data, but can give very poor estimates of injectivity (Leeftink et al., 2013).
Rossen et al. (1999) introduced a small physical dispersivity to their simulations to mitigate the effects of a sharp change in
mobility with water saturation. The artifacts at the leading edge of the foam bank are reduced in significance as the grid is
refined near this front, though again this may slow execution of the simulator.
Similarly, the overall impact of the artificially low computed injectivity decreases as the size of the injection-well grid
block is reduced. Dimensionless time in Fig. 5 is based on the pore volume of the injection-well grid block; hence the period
of artificially low injectivity passes more quickly as the grid is refined there. Also, the fraction of overall injection-well
pressure dissipated in that grid block decreases as well (though more slowly) as grid-block size is reduced.
Finally, gas compressibility reduces the impact of fluctuating mobility at the foam front on injection-well pressure,
especially as the size of the foam bank increases. (Several of the examples above were computed using the MOC, assuming
zero compressibility). Figure 7, from Namdar Zanganeh and Rossen (2013), illustrated the effects of grid resolution and gas
compressibility in a linear 1D SAG simulation. Fluctuations in injection rate at fixed injection pressure are huge even with 100
grid blocks, but they decrease in magnitude as the gas bank advances. Refining the grid does not eliminate the artifact; each
grid block must still pass through the minimum in mobility in Fig. 2, but refining the grid does reduce its impact on overall
injectivity.

Figure 7 (left). Injection rate at fixed injection pressure during gas injection in a SAG process in a 100-grid-block 1D reservoir;
comparison to 1000-grid-block reservoir. Gas injection here starts on Day 66. The abrupt rise in injectivity in about day 78 is the
breakthrough of gas to the production well. From Namdar Zanganeh et al. (2013).
Figure 8 (right). Schematic of reasoning behind step-function in Fs, to reduce effect of numerical dispersion on surfactant
concentration. From Cheng (2002).

Effect of Surfactant Concentration

The effect of surfactant concentration Cs on gas mobility in foam is nonlinear, with most of the effect coming at low
concentrations (Sanchez et al., 1986; Kuhlman et al., 1992; Apaydin and Kovscek, 2001). In the absence of dispersion
(physical and numerical), two surfactant concentrations primarily matter to a foam process: the initial concentration in the
6 SPE 166232

reservoir (i.e., zero surfactant concentration) and the injected concentration. In the absence of significant physical dispersion,
numerical dispersion rapidly spreads the surfactant front from its initial sharp profile over many grid blocks, though surfactant
adsorption (if modeled with a Langmuir isotherm) helps to sharpen the front (Lake et al., 2003).
If the true effect of surfactant on foam mobility is combined with numerical dispersion of the surfactant front, the foam
bank is extended beyond where it should lie, toward the leading edge of the dispersed front in Cs, and the arrival of foam at
any location is earlier and more gradual than it should be. In order to restore the dispersion-free sharp front for the foam bank,
Rossen et al. (1999) proposed a step function from 0 to 1 for Fs(Cs) at ½ the injected surfactant concentration. The idea is
illustrated in Fig. 8: if the dispersed profile of Cs is symmetric (shaded area I in figure is equal to shaded area II), then the step-
function with an abrupt onset of foam at ½ the injected surfactant concentration puts the foam front at the same location in
which it would lie if there were no dispersion. STARS allows for a nonlinear function Fs(Cs) that can approximate either a step
change at any given concentration or a nonlinear increase with most of the effect at smaller values of Cs.
Cheng (2002) investigated the implications of using a step change. With surfatactant adsorption, the profile of Cs is not
symmetric, and the position of the front is distorted a small amount. Moreover, in simulations of foam injection the abrupt
change in gas mobility as Cs passes through its threshold value in consecutive grid blocks gives a sudden change in mobility
and fractional flow there. Figure 9 illustrates the effect, using a plot of water fractional flow based on 1D simulation results.
Foam is injected into a 1D porous medium saturated with water (without surfactant). The gray-black boundary is the leading
edge of the gas bank ahead of the foam. The foam front advances slowly in this example: only five grid blocks have reached
the threshold value of Cs in the time shown here. As each grid block reaches the threshold, gas flow stops and water pours out
of the grid block. Later, as each grid block dries out, its water saturation approaches Sw*, Fw decreases, and gas flow resumes.
The fast-moving, (white) wave with increased water fractional flow disperses somewhat as it advances toward the front of the
gas bank.
Cheng (2002) argues that, although unsightly, this artifact has little effect on the overall simulation. On average the foam
front and gas bank advance at the correct velocities. It is a likely cause of multiple local optima in a study of gradient-based
optimization of surfactant-slug size in SAG flooding (Namdar Zanganeh et al., 2012). Since the effect of increasing surfactant-
slug size is not continuous, but occurs in one-grid-block increments, the approach to the global optimum slug size would not
be expected to be smooth. This could explain the multiple local optima near the global optimum found in that study.

Figure 9 (left). Water fractional flow (lighter gray = increasing water fractional flow) in a 1D linear simulation of foam injection. These
simulation results are plotted in the manner of a "time-distance diagram" (Lake, 1989) from fractional-flow theory: the vertical axis is
position and the horizontal axis time: the slope of fronts represents their velocity in the medium. The faster moving, white front is the
gas bank ahead of foam; the burst of white from five grid blocks represents the advance of foam one grid block at a time. The region
ahead of the gas bank is shown as black because it is off-scale (water fractional flow of 1). From Cheng (2002).

Effect of Surfactant Concentration: Mitigation

As with water saturation, the less abrupt the rise in gas mobility with decreasing surfactant concentration, the smaller the
effects described here. In this case the artifacts arise from an abrupt change not required to match data but as the unintended
consequence of trying to mitigate another problem, the effect of dispersion of surfactant concentration. Refining the simulation
grid both reduces the magnitude of the effect and, possibly, by reducing the numerical dispersion of surfactant concentration,
the need to introduce the abrupt change in Fs at all.

Effect of Oil Saturation and Composition

The effect of oil saturation and composition on foam is complex and not fully understood, and there are many outstanding
questions on how to represent the effect of oil on foam (Farajzadeh et al., 2012). The effect of oil on foam is sometimes
represented in the function Fo as an abrupt change as a function of oil saturation. This leads potentially to the same sort of
complications observed with water saturation, as oil saturation in grid blocks at a foam front pass through oil saturations where
foam properties change abruptly. The presence of a third mobile phase introduces an additional, different complication:
composition paths on the phase diagram may attempt to travel along a boundary in which foam properties change abruptly.
SPE 166232 7

Maintaining the correct course in such cases is extremely difficult. Namdar Zanganeh et al. (2008) illustrate such a case,
reproduced in Figs. 10 to 12.

Figure 10 (left). Water saturation from a simulation of gas injection in a SAG foam displacement with oil; different symbols represent
different dimensionless times (pore volumes gas injected). Note alternating grid blocks have strong and weak foam.
Figure 11 (center). Same displacement as Fig. 10, with grid-block saturations plotted as a composition path on the phase diagram.
Excluded regions to the upper right and left represent residual oil saturation and irreducible water saturation. All three phases are
completely immiscible. Oil destroys foam here for oil saturation So > 0.2. Alternating grid blocks fall on opposite sides of this
boundary, with either strong or collapsed foam. Blue and red curves are "slow" and "fast" composition paths on the phase diagram
used in fractional-flow modeling (Lake, 1989).
Figure 12 (right). Analytical solution (heavy black line) for displacement using MOC: from injection condition J to state IJ and then to
initial condition I. The correct solution skirts the boundary where oil saturation kills foam. Arrowheads here indicate directions of
increasing velocity along characteristics, not grid-block saturations.
Figures 10-12 from Namdar Zanganeh et al. (2008).

Effect of Oil: Phase Names

Simulators define foam as an interaction between aqueous surfactant solution and gas. In a first-contact or multiple-contact
miscible displacement, the identity of the nonaqueous phase changes from "oil" to "gas" as the miscible front passes a given
location, and there are only two phases present at any location: water and gas or water and oil. As Liu et al. (2011) point out,
therefore, by definition, there is no effect of oil on foam: if oil is present, there is no "gas" and therefore no foam; if gas is
present, there is no "oil" to affect the foam. In fact, the effect of oil on foam in such a case is to abruptly create foam (if
surfactant is present) when the simulator changes the name of the nonaqueous phase from "oil" to "gas." If the miscible front is
ahead of the surfactant front, then this effect in unimportant, but in injection of lower-quality foam, or in a SAG flood, where
surfactant injection precedes gas injection and the miscible front lags the surfactant front, at least at first, this would control the
propagation rate of foam. The effect would also be important in 2D or 3D displacements, where extreme viscous instability
between the gas bank ahead of foam and the oil bank ahead of it (cf. Ashoori et al., 2010) would cause fingering or override of
the gas and could leave oil in contact with the foam.

Effect of Oil: Displacement Front

In an immiscible (or partially miscible) displacement, if oil saturation ahead of the foam bank is great enough to kill foam,
foam may still displace oil if oil drains out the grid block at the leading edge of the displacement front fast enough for foam to
build. Liu et al. (2011) give an example (Fig. 13). Whether this would occur in reality depends on the "traveling wave" at the
displacement front (cf. Ashoori et al., 2011a), where oil is displaced as foam arrives. Whether this would happen in the
simulation depends on how well the simulator, in a single grid block, represents the traveling wave.

Figure 13. Simulation of foam in an immiscible displacement where oil saturation ahead of the foam bank is great enough to kill foam,
but it is displaced ahead of the foam to a lower saturation that is not harmful to foam. From Liu et al. (2011).
8 SPE 166232

Population-Balance Simulation
Population-balance simulators incorporate all the factors discussed above, and in addition solve dynamically for foam
texture throughout the displacement. The dynamics of lamella creation and destruction are fast compared to the time scale of
the displacement (Ashoori et al., 2011b), which makes the set of equations "stiff." There are many variants of the population-
balance approach (Falls et al., 1988; Friedman et al., 1991; Kovscek and Radke, 1994; Zitha, 2006; Kam et al., 2007), differing
most in how they treat lamella creation, and each approach faces its own numerical challenges. In the versions based on foam
generation triggered by pressure gradient (Kam et al., 2007; Ashoori et al. 2011a,b, 2012a), numerical dispersion can alter the
stability of the strong-foam state (Ashoori et al., 2012b), and the equations are extremely stiff.

Discussion
Discussion here has focused on simulating foam injection and SAG processes. SAG processes are most sensitive to water-
saturation artifacts, because of the extreme difference between mobilities within the shock front and the mobility in the gas
bank (Fig. 2) and the difficulty in resolving the high-mobility region near the well and therefore calculating injectivity
correctly (Figs. 6 and 7). Foam-injection processes show fluctuations in pressure gradient as the surfactant front advances, if
the surfactant effect is represented as an abrupt jump. Processes with surfactant dissolved in the "gas" (i.e. supercritical fluid)
(Le et al., 2008) are subject to both problems, because there is a shock to low water saturation as in a SAG process and also a
surfactant front that lags the gas front (Ashoori et al., 2010).
All these issues arise from representing the effects of water saturation, oil saturation and surfactant concentration as abrupt
changes in gas mobility with respect to the given variable. In reality it is not just the abruptness of the transition but also the
magnitude of reduction in gas mobility that matter. Thus the effects can be reduced by choosing a lesser reduction in gas
mobility with foam. The magnitude of the reduction gas mobility by foam in laboratory studies (under admittedly idealized
conditions) can be enormous, however (Cheng, 2000). The issue then is whether the choice of model parameters forced by a
compromise for numerical efficiency is still an accurate representation of the process.

Conclusions
This text focuses primarily on effects of abrupt changes in foam properties with water saturation, surfactant concentration, and
oil saturation. Our conclusions are as follows:
1. Water saturation. There are several effects. First, an abrupt change in gas mobility with water saturation is most
problematic of SAG foam processes, where it can introduce fluctuations in mobility and injectivity. Second, and more
fundamentally, the fact that in finite-difference simulations each grid block must pass through saturations and
mobilities not seen in the actual displacement artificially reduces injectivity early in gas injection and introduces
fluctuations in injectivity and mobility; in some cases the overall course of the process is altered. The magnitude of
this second effect is reduced upon grid refinement near the well and at the foam front. The functional form the
simulator uses to represent the effect of water saturation is important: it can preclude a successful SAG process and/or
lead to underestimation of injectivity throughout a SAG process. Since capillary effects are so important to foam,
representing foam accurately in finely laminated formations with grid blocks much larger than the scale of the
laminations remains a challenge.
2. Surfactant concentration. If numerical dispersion of the surfactant front is significant, and the foam model correctly
represents the effect of surfactant concentration on foam strength, the foam bank is accelerated somewhat. But
replacing this dependence on surfactant concentration with a step function introduces fluctuations in fluxes as each
successive grid block reaches the threshold for foam formation. Refining the grid would reduce the problem in both
cases.
3. Oil saturation. With a third mobile phase present, a new numerical challenge appears: successive grid blocks may lie
on opposite sides of a boundary between strong and weak foam. Whether foam can displace oil ahead of it depends in
part on how well the simulator resolves the narrow traveling wave where oil is displaced. In miscible processes, the
very naming of the phases in the simulator can affect foam propagation, since foam is defined as an interaction
between gas and water phases.

References
Alvarez, J. M., Rivas, H., and Rossen, W.R., "A Unified Model for Steady-State Foam Behavior at High and Low Foam Qualities," SPE
Journal 6, 325-333 (2001).
Apaydin, O. G., and Kovscek, A. R., "Surfactant Concentration and End Effects on Foam Flow in Porous Media," Transport Porous Media
43, 511-536 (2001).
Ashoori, E., van der Heijden, T.L.M., and Rossen, W.R., "Fractional-Flow Theory of Foam Displacements with Oil," SPE Journal 15 260-
273 (2010).
Ashoori, E., Marchesin, D., and Rossen, W.R., "Multiple Foam States and Long-Distance Foam Propagation in Porous Media," SPE Journal
17, 1231-45 (2012a).
Ashoori, E., Marchesin, D., and Rossen, W.R., "Stability Analysis of Uniform Equilibrium Foam states for EOR Processes," Transport
Porous Media 92, 573-595 (2012b).
Ashoori, E., Marchesin, D., and Rossen, W.R., "Roles of Transient and Local Equilibrium Foam Behavior in Porous Media: Traveling
Wave," Colloids and Surfaces A: Physicochem. Eng. Aspects. 377, 228–242 (2011a).
SPE 166232 9

Ashoori, E., Marchesin, D., and Rossen, W.R., "Dynamic Foam Behavior in the Entrance Region of a Porous Medium," Colloids and
Surfaces A: Physicochem. Eng. Aspects 377, 217–227 (2011b).
Bernard, G. G., Holm, L. W., and Jacobs, W. L., "Effect of Foam on Trapped Gas Saturation and on Permeability of Porous Media to
Water," SPE Journal 5, 295-300 (1965).
Boeije, C. S., and Rossen, W. R., "Fitting Foam Simulation Model Parameters to Data," presented at the 17th European Symposium on
Improved Oil Recovery, St. Petersburg, Russia, 16-18 April 2013.
Chen, Q., Gerritsen, M. G., and Kovsck, A. R, "Modeling Foam Displacement With the Local-Equilibrium Approximation: Theory and
Experimental Verification," SPE Journal 15, 171-183 (2010).
Cheng, L., “Modeling and Simulation Studies of Foam Processes in Improved Oil Recovery and Acid-Diversion,” PhD dissertation, The
University of Texas at Austin, 2002.
Cheng, L., Reme, A. B., Shan, D., Coombe, D. A. and Rossen, W. R., "Simulating Foam Processes at High and Low Foam Qualities," paper
SPE 59287 presented at the 2000 SPE/DOE Symposium on Improved Oil Recovery, Tulsa, OK, 3-5 April.
Computer Modeling Group, STARS User's Guide, Version 2006, Calgary, Alberta, Canada.
Dong, Y., and Rossen, W. R., "Insights from Fractional-Flow Theory for Models for Foam IOR," presented at the 14th European
Symposium on Improved Oil Recovery, Cairo, Egypt, 22-24 April, 2007.
Falls, A. H., Hirasaki, G. J., Patzek, T. W., Gauglitz, D. A., Miller, D. D., and Ratulowski, J, "Development of a Mechanistic Foam
Simulator: The Population Balance and Generation by Snap-Off," SPE Reser. Eng. 3, 884-892 (1988).
Farajzadeh, R., Andrianov, A., Krastev, R., Hirasaki, G. J., and Rossen, W. R., "Foam-Oil Interaction in Porous Media: Implications for
Foam Assisted Enhanced Oil Recovery," Adv. Colloid Interface Sci., 183-184, 1-13 (2012).
Fisher, A. W., Foulser, R. W. S., and Goodyear, S. G., "Mathematical Modeling of Foam Flooding," paper SPE/DOE 20195 presented at the
1990 SPE/DOE Symposium on Enhanced Oil Recovery, Tulsa, OK, Apr. 22-25.
Friedmann, F., Chen, W. H., and Gauglitz, P. A., "Experimental and Simulation Study of High-Temperature Foam Displacement in Porous
Media," SPE Reser. Eng. 6, 37-75 (1991).
Huh, D. G. and Handy, L. L.,"Comparison of Steady and Unsteady-State Flow of Gas and Foaming Solution in Porous Media," SPE Reserv.
Eng. 4, 77-84 (1989).
Islam, M. R. and Farouq-Ali, S. M., "Numerical Simulation of Foam Flow in Porous Media," J. Canadian. Pet. Tech. (July-Aug. 1990) 47-
51.
Kam, S. I. , and Rossen, W.R., "A Model for Foam Generation in Homogeneous Porous Media, " SPE Journal 8, 417-425 (2007)
Kam, S. I., Nguyen, Q. P., Li, Q., and Rossen, W. R., "Dynamic Simulations With an Improved Model for Foam Generation," SPE Journal
12, 35-48 (2007).
Khatib, Z. I., Hirasaki, G. J. and Falls, A. H., “Effects of Capillary Pressure on Coalescence and Phase Mobilities in Foams Flowing through
Porous Media,” SPE Reser. Eng. 3, 919-926, (1988).
Kovscek, A. R. and Radke, C. J., “Fundamentals of Foam Transport in Porous Media,” in Foams: Fundamentals and Applications in the
Petroleum Industry, L.L. Schramm (ed.) ACS Advances in Chemistry Series, Am. Chem. Soc., Washington, D.C. (1994) 3, No. 242.
Kuhlman, M. I., Falls, A. M., Hara, S. K., Monger-McClure, T. G., and Borchardt, J.K., "CO2 Foam With Surfactants Used Below Their
Critical Micelle Concentrations," SPE Reserv. Eng. 7, 445-452 (1992).
Kular, G. S., Lowe, K., and Coombe, D., "Foam Application in an Oil Sands Steam Flood Process," paper SPE 19690 presented at the 1989
SPE Annual Tech. Conf. and Exhibition, San Antonio, TX, Oct. 8-11.
Lake, L. W., Enhanced Oil Recovery, Prentice Hall, Englewood Cliffs, New Jersey, USA (1989).
Lake, L. W., Bryant, S. L., and Araque-Martinez, A. A., Geochemistry and Fluid Flow, Elsevier, Amsterdam (2003).
Law, D. H., Yang, Z.-M., and Stone, T., "Effect of Presence of Oil on Foam Performance: A Field Simulation Study," paper SPE 18721
presented at the 1989 SPE Symposium on Reservoir Simulation, Houston, TX, Feb. 6-8.
Le, V. Q., Nguyen, Q. P. and Sanders, A. W., "A Novel Foam Concept with CO2 Dissolved Surfactants," paper SPE 113370 presented at the
2008 SPE/DOE Symposium on Improved Oil Recovery, Tulsa, Oklahoma, USA, 20-23 April.
Leeftink, T. N., Latooij, C. A., and Rossen, W. R., "Injectivity Errors in Simulation of Foam EOR," presented at the 17th European
Symposium on Improved Oil Recovery, St. Petersburg, Russia, 16-18 April 2013.
Liu, M. K., Andrianov, A. I., and Rossen, W. R., "Sweep Efficiency in CO2 Foam Simulations with Oil," paper SPE 142999 presented at the
SPE EUROPEC/EAGE Annual Conference and Exhibition, Vienna, Austria, 23–26 May 2011.
Ma, K., Biswal. S. L., and Hirasaki, G. J., "Experimental and Simulation Studies of Foam in Porous Media at Steady State," presented at the
2012 AIChE Spring Meeting, Houston, TX, 1-5 April.
Mayberry, D. J., Afsharpoor, A., and Kam, S. I.,"The Use of Fractional Flow Theory for Foam Displacement in Presence of Oil," SPE
Reserv. Eval. Eng. 11, 707-718 (2008).
Mohammadi, S., and Coombe, D. A., "Characteristics of Steam-Foam Drive Process in Massive Multi-Zone and Thin Single-Zone
Reservoirs," paper SPE 24030, presented at the 1992 SPE Western Regional Meeting, Bakersfield, CA, March 30-April 1.
Myers, T. J., and Radke, C. J., "Transient Foam Displacement in the Presence of Residual Oil: Experiment and Simulation Using a
Population-Balance Model," Ind. Eng. Chem. Res. 39, 2725-2741 (2000).
Namdar Zanganeh, M., Jansen, J. D., Kraaijevanger, J. F. B. M., Buurman, H. W., and Rossen, W. R., "Adjoint-Based Optimization of a
Foam EOR Process," presented at the European Conference on the Mathematics of Oil Recovery, Biarritz, France, 10 - 13 September
2012.
Namdar Zanganeh, M., Kam, S. I., LaForce, T. C., and Rossen, W.R., "The Method of Characteristics Applied to Oil Displacement by
Foam," SPE Journal 16, 8-23 (2011).
Namdar-Zanganeh, M., La Force, T., Kam, S. I., van der Heijden, T. L. M., and Rossen, W. R., "Fractional-Flow Theory of Foam
Displacements with Oil," presented at 11th European Conference on Mathematics of Oil Recovery, Bergen, Norway, 8-11 Sept. 2008.
Namdar Zanganeh, M., and Rossen, W. R., "Optimization of Foam EOR: Balancing Sweep and Injectivity," accepted for publication in SPE
Reservoir Evaluation and Engineering.
Patzek, T. W. and Myhill, N. A., "Simulation of the Bishop Steam Foam Pilot," paper SPE 18786 presented at the 1989 SPE California
10 SPE 166232

Regional Meeting, Bakersfield, Apr. 5-7.

Rossen, W. R., "Foam Generation at Layer Boundaries in Porous Media," SPE Journal 4, 409-412 (Dec. 1999).
Rossen, W. R., "Foams in Enhanced Oil Recovery," in R. K. Prud'homme and S. Khan, ed., Foams: Theory, Measurements and
Applications, Marcel Dekker, New York (1996), pp. 413-464.
Rossen, W. R., and Boeije, C. S., "Fitting Foam Simulation Model Parameters for SAG Foam Applications," paper SPE 165282 presented at
the 2013 SPE Enhanced Oil Recovery Conference, Kuala Lumpur, Malaysia, 2-4 July.
Rossen, W. R., and Bruining, J., "Foam Displacements With Multiple Steady States," SPE Journal 12, 5-18 (2007).
Rossen, W. R. and Lu, Q., "Effect of Capillary Crossflow on Foam Improved Oil Recovery," paper SPE 38319 presented at the 1997 SPE
Western Regional Meeting, Long Beach, CA, June 25-27.
Rossen, W. R., van Duijn, C. J., Nguyen, Q. P., Shen, C., and Vikingstad, A. K., "Injection Strategies to Overcome Gravity Segregation in
Simultaneous Gas and Water Injection Into Homogeneous Reservoirs," SPE Journal 15, 76-90 (2010).
Rossen, W. R., Zeilinger, S. C., Shi, J.-X., and Lim, M. T., "Simplified Mechanistic Simulation of Foam Processes in Porous Media," SPE
Journal 4, 279-287 (1999).
Schramm, L. L. (ed.) Foams: Fundamentals and Applications in the Petroleum Industry, ACS Advances in Chemistry Series No. 242, Am.
Chem. Soc., Washington, DC (1994).
Sanchez, J. M., Schechter, R. S., and Monsalve, A., "The Effect of Trace Quantities of Surfactant on Nitrogen/Water Relative
Permeabilities," paper SPE 15446 presented at the 1986 Annual Technical Conference. of the SPE., New Orleans.
Shan, D. and Rossen, W. R., “Optimal Injection Strategies for Foam IOR,” SPE Journal 9, 132-150 (2004).
Stone, H. L.: "A Simultaneous Water and Gas Flood Design with Extraordinary Vertical Gas Sweep," paper SPE 91724 presented at the
2004 SPE International Petroleum Conference in Mexico, Puebla Pue., Mexico, 7-9 Nov.
Tanzil, D., Hirasaki, G. J., and Miller, C. A.: “Conditions for Foam Generation in Homogeneous Porous Media,” paper SPE 75176 presented
at the 2002 SPE/DOE Symposium on Improved Oil Recovery, Tulsa, OK, 13-17 April.
van Lingen, P. P., Bruining, J., and van Kruijsdijk, C. P. J. W., "Capillary Entrapment Caused by Small-Scale Wettability Heterogeneities,"
SPE Reserv. Eng. 11, 93-99 (1996).
Vassenden, F. and Holt, T.,"Experimental Foundation for Relative Permeability Modeling of Foam" SPE Reserv. Eval. Eng. 3: 179-185
(2000).
Yortsos, Y. C., and Chang, J., "Capillary Effects in Steady-State Flow in Heterogeneous Cores," Transport Porous Media 5, 399-420
(1990).
Zhou, Z. H. and Rossen, W. R,. "Applying Fractional-Flow Theory to Foam Processes at the 'Limiting Capillary Pressure'," SPE Adv.
Technol. 3, 154-162 (1995).
Zitha, P. L. J., "A New Stochastic Bubble Population Model for Foam in Porous Media," paper SPE 98976 presented at the 2006 SPE/DOE
Symposium on Improved Oil Recovery, Tulsa, Oklahoma, U.S.A., 22–26 April.