You are on page 1of 19

A NUMERICAL STUDY OF SURFACTANT FLUSHING TO REMEDIATE

NAPL (NON-AQUEOUS PHASE LIQUID)

MINHEE LEE
Department of Civil and Environmental Engineering, 116 Everson Hall, University of California at
Davis, Davis, CA 95616, U.S.A.
(e-mail: heelee@ucdavis.edu)

(Received 15 February 1999; accepted 17 September 1999)

Abstract. Mathematical models of flow and transport mechanisms for surfactant flushing to remedi-
ate NAPL (Non-aqueous phase liquid) have been established and solved by numerical simulation. A
column experiment was used as a physical model for the simulation. One percent sorbitan monooleate
was used as a surfactant solution and tetrachloroetylene (PCE) as a NAPL source. The experimental
work was simulated using the commercial computer code, MOTRANS, and its computed result was
compared to the experimental result. The predicted effectiveness of surfactant flushing from the
simulation was better than observed in the experimental work. This behavior was attributed to a
more complicated distribution of PCE in the real column than in the model and kinetic effects. These
effects should be considered in applying surfactant flushing for remediation. The numerical study of
the column experiment might provide a method for prediction of suitability of surfactant flushing as
a remediation technique at the contaminated sites.

Keywords: NAPL, PCE, remediation, simulation, surfactant

1. Introduction

Soil and groundwater contamination by NAPLs (Non-aqueous phase liquids) has


become a major environmental problem (Mackey et al., 1989; Pankow et al., 1996).
NAPLs are immiscible with water and can be a long-term source for contamina-
tion due to the low mass transfer between the NAPL and aqueous phases in the
subsurface (Wilson et al., 1984; Mercer et al., 1990). Drinking water limits for
NAPLs have been set at very low levels, which are more than three orders of
magnitude lower than their solubilities in ground water, suggesting that relatively
small quantities of NAPLs can contaminate a large area (Feenstra et al., 1988).
The conventional clean up method has been the pump-and-treat system, which
pumps groundwater at a high rate through the contaminated site to remove NAPLs
as a dissolved phase. However, the removal of more than 99% of NAPL mass is
probably required to restore a contaminated aquifer to drinking water standards at
most sites, and this level of recovery is impractical when using the conventional
pump-and-treat method alone (Bedient et al., 1996).
As an alternative method, surfactants have been studied to increase the solubility
of NAPL in the aqueous phase and thus to enhance the efficiency of a pump-

Water, Air, and Soil Pollution 121: 289–307, 2000.


© 2000 Kluwer Academic Publishers. Printed in the Netherlands.
290 MINHEE LEE

and-treat method as a remediation technique (Fountain et al., 1996; Brown et al.,


1994; Pennell et al., 1993). Surfactant flushing increases the rate of contaminant
removal by increasing the apparent solubility of the contaminant in the aqueous
phase that improves the mass removal per pore volume in a pump-and-treat system.
Significant researches in the laboratory testing of surfactant-enhanced remediation
methods have been performed in the last decade. Ellis et al. (1986) used a mixture
of ethoxylated alkylphenols and ethoxylated fatty acids as a surfactant solution to
examine the recovery of contaminants from soil. The contaminants included crude
oil, PCBs, and a mixture of chlorophenols. Over 85% of the total contaminant
mass was removed from contaminated columns. Clarke et al. (1991) used sodium
dodecylsulfate (SDS) for the recovery of PCBs from soils. Over 99.7% of PCBs
were removed from the contaminated clay soil with a 50 mM SDS solution. Abdul
et al. (1994) performed pilot-scale tests at a field site contaminated with PCBs and
oils (petroleum products). More than 85% of the PCBs were removed after washing
with 105 pore volumes of an alcohol ethoxylate surfactant solution. Surfactant
flushing was employed by Pennell et al. (1993) to enhance the recovery of resid-
ual dodecane and PCE, using Polyoxyethylene (POE) 20 surfactant solution. Lee
(1998) studied the effect of sorbitan monooleate to remove PCE from the saturated
and the unsaturated zone. Fountain et al. (1996) performed a controlled field test
at the Canadian Forces Base Borden site and concluded that surfactant-enhanced
aquifer remediation could rapidly remove the majority of DNAPL (Dense non-
aqueous phase liquid) using simple modifications of a pump-and-treat system.
However, these studies for surfactant flushing have been limited to experimental
and observational research. The mature understanding of the mechanisms of flow,
transport, and mass transfer between phases for surfactant flushing is essential
to optimize the efficiency of surfactant flushing and to predict the suitability of
surfactant flushing as a remediation technology for field-scale-system. Several the-
oretical studies, related to the NAPL-aqueous phase mass transfer model have been
performed to understand the mechanism of multiphase flow and transport in the
porous media system (Faust, 1985; Hunt et al., 1988; Parker, 1989; Mayer et al.,
1996; Delshad et al., 1999), but the mathematical study for surfactant flushing lags
behind the experimental research dealing with remediation of NAPL.
The objectives of this work were to understand the mechanisms of surfact-
ant flushing mathematically and determine the influence of parameters that affect
the remediation efficiency through numerical simulation. Numerical simulations
for surfactant flushing have been done using the two-dimensional finite element
computer code for multiphase flow and transport, MOTRANS, which was the
commercialized version of MOFAT (Katyal et al., 1991). Results of simulation
were compared with the experimental observations.
A NUMERICAL STUDY OF SURFACTANT FLUSHING 291

2. Experimental Work as a Physical Model

A column test of surfactant flushing was performed as a physical model for nu-
merical simulation. One percent of sorbitan monooleate was used as a non-ionic
surfactant, based on the solubilization ability for NAPLs, low toxicity, low sorption,
and viscosity of the surfactant. Surfactant flushing may induce the mobilization
of NAPL by lowering the interfacial tension between aqueous and NAPL phase.
Since the mobilization of NAPL may result in the contaminant partially dropping
to greater depth, the surfactant selected for this study had high interfacial tension,
minimizing the risk of mobilization (Lee, 1998). Tetrachloroethylene (PCE) was
used as a NAPL because it is the third most common contaminant at hazardous
waste sites in the United States (Shiau et al., 1994). Sorbitan monooleate increases
PCE solubility to approximately 5453 mg L−1 in a 1% surfactant solution (the
critical micelle concentration: 0.0039%) from an aqueous solubility of 150 mg L−1
(Fountain, 1995). Medium size Ottawa silica sand (0.025–0.035 cm in diameter)
was used for the column test. More detailed information on the column experiments
and properties of sorbitan monooleate are given in Lee (1998).
The glass column (5 cm in diameter and 30 cm in length) was equipped with
porous Teflon end plates and filled with Ottawa sand to a height of 28 cm from
the bottom end plate. A 1 cm layer of fine sand (0.01–0.0125 cm in diameter) was
added to cover the top of the sand, to spread water over the entire surface and to
minimize volatilization. A one cm void space remained between the top end cap
and the top of fine sand. To create an unsaturated condition inside a column, the
Teflon end cap on the top of the column was punctured with a 0.5 cm diameter hole
to provide an air vent. Boiled distilled water was pumped into the top of a column
through a tube connected to the end cap. Water was allowed to exit the column
to create unsaturated and saturated conditions. The water flushing was continued
at a constant pumping rate until no changes of water content in the column were
observed (usually more than 2 pore volumes were needed). Four milliliters of PCE,
dyed red with Sudan IV dye, was injected from the top of the column. The top end
cap was opened and a piece of Teflon tubing was inserted into the center of the
column to a depth of 4 cm. The Teflon tube on the top end cap was attached to a
graduated glass burette that served as the reservoir for the PCE. The burette was
filled with 3 mL of water, then 4 mL of PCE, followed by an additional 3 mL of
water (total 10 mL liquid). The water placed above the PCE in the burette prevented
PCE volatilization during PCE injection. After filling, the stopcock on the burette
was opened to allow PCE to drain into the column by gravity. The column was left
undisturbed for 10–20 min to allow the PCE to redistribute into the pore spaces in
the column. At the start of the surfactant flushing experiment, a Teflon bag (1.5–2 L
in capacity) was attached to the effluent tube and used to capture the effluent for
the mass balance. The surfactant solution was then pumped through the column at
a constant flow rate to dissolve the residual PCE in the column. Samples from the
effluent were taken in 2 mL vials at different time intervals and analyzed on a gas
292 MINHEE LEE

chromatograph for the concentration of PCE in the samples. A plot of effluent PCE
concentration versus pore volumes of surfactant solution injected was generated
to evaluate the remediation efficiency of surfactant flushing. The total mass of
PCE removed from the column was calculated by adding the total mass of PCE
in the bag to the amount of PCE in the 2 mL sample vials. After flushing was
completed, all sand from the column was extruded into a 1000 mL flask partially
filled with methanol. The flask of sand and methanol was shaken on a shaker table
for one hour, and three samples were taken from the flask and analyzed by the gas
chromatograph to determine the mass remaining in the column.

3. Modeling Process

Numerical simulation of surfactant flushing requires mathematical solution for a


multi-phase flow and transport.

3.1. T HEORY OF MULTI - PHASE FLOW

3.1.1. Governing Equation for Multi-phase Flow


Fluid flow in porous media is generally described by a Darcy’s Law. Darcy’s
velocities in the p-phase are defined by
 
∂hp
qpi = −Kpij + ρrp uj (1)
∂xj
where Kpij is the p-phase conductivity tensor [L/T ], hp = Pp /gρw is the water
height-equivalent pressure head of phase p where Pp is the p-phase pressure [L],
g is gravity [L/T 2 ], ρp is the density of phase p [M/L3 ], ρrp = ρp /ρw is the p-
phase specific gravity, xi is Cartesian spatial coordinate (i,j = 1, 2, and 3), and uj =
∂z /∂xi is a unit gravitational vector measured positive upwards where z is elevation.
The mass conservation equations for aqueous phase (w: water or surfactant solution
phase), PCE phase (oil phase, o) and air phase (a), assuming an incompressible
porous medium, may be written in summation convention for a two-dimensional
Cartesian domain as
∂Sw ∂qwi Rw
φ =− + (2)
∂t ∂xi ρw
∂So ∂qoi Ro
φ =− + (3)
∂t ∂xi ρo
∂(ρa Sa ) ∂(ρa qai )
φ =− + Ra (4)
∂t ∂xi
where φ is the porosity, Sp is the p-phase saturation, qip is the Darcy velocity of
phase p in the i-direction [L/T ], Rp is the net mass transfer per unit porous media
A NUMERICAL STUDY OF SURFACTANT FLUSHING 293

volume into or out phase p [M/L3 T ]. The phase index p takes the values p = a
(air), o (oil = PCE), or w (aqueous phase). Substituting Darcy’s equation for qp
yields
  
∂Sw ∂ ∂hw Rw
φ = Kwij + ρrw uj + (5)
∂t ∂xi ∂xj ρw
  
∂So ∂ ∂ho Ro
φ = Koij + ρro uj + (6)
∂t ∂xi ∂xj ρo
  
∂ρa Sa ∂ ∂ha
φ = ρa Kaij + ρra uj + Ra . (7)
∂t ∂xi ∂xj

If fluid saturations are treated as functions of the phase pressures, the equations
may be solved directly with phase pressures as primary variables. Kpij can be
derived from the relative permeability of phase p, calculated from the capillary
pressure-saturation-permeability relationships of Parker et al. (1987) and Lenhard
et al. (1987). In many cases, gas pressure gradients may be regarded as negligible
and the gas flow equation may be discarded. In this research, it is assumed that gas
flow has negligible effect on liquid flow due to small gas pressure gradients.

3.1.2. Initial and Boundary Conditions


Initial conditions for each phase p in the flow equations must be stipulated on the
entire flow domain R as

hp (xI , 0) = hp1 (xI ) on R for t = 0 (8)

where hp is pressure heads of p phase. Boundary conditions may be stipulated as


type-1 or type-2 as follows

hp (xI , t) = hp2 (xI , t) on S1 for t > 0 : type-1 (9)

qpi ·nI = qpi (xI , t) on S2 for t > 0 : type-2 (10)

where nI is the outward normal unit vector. Type-1 denotes pressure head on
boundary region S1 . Type-1 will commonly be used for the aqueous phase on
boundaries within a saturated aquifer where positive heads are specified corres-
ponding to assumed vertical hydrostatic conditions and also be used on infiltrating
boundaries for aqueous phase or oil if the infiltration is pressure controlled. Type-2
denotes the normal flux on boundary segment S2 , for phase p with outward normal
unit vector ni . The default flow boundary condition is zero normal flux. Nonzero
flow boundary conditions will be used on boundary regions along which inflow (or
outflow) rates are known.
294 MINHEE LEE

3.2. T HEORY OF MULTI - PHASE TRANSPORT

The interphase mass transfer was assumed to be influenced by PCE partitioning


among the aqueous, PCE and gas phases with local equilibrium interphase mass
transfer. Sorption on the solid phase was ignored based on the results from sorption
experiments that showed negligible sorption (Lee et al., 1999; Lagouski, 1996).

3.2.1. Governing Equation of Multi-Phase Transport


To model PCE transport, continuity and mass flux equations for each phase must
be defined. Mass conservation equation of PCE in the p-phase is described as

∂(Cp Sp ) ∂Jpi
φ =− + Rp + γ p (11)
∂t ∂xi

∂Cp
Jpi = Cp qpi − φSp Dpij (12)
∂xj

where Cp is the concentration of PCE in p-phase [M/L3 ], Jpi is the mass flux
density of PCE in p-phase in the i-direction [M/L2 T ], Rp is the net mass transfer
rate of PCE into or out of the p-phase [M/L3 ], γp is the net production (+) or
decay (–) of PCE due to reactions within phase p [M/L3 ] (γp = −µp Cp , µp :
an apparent first-order decay coefficient), and Dpij is a dispersion tensor [L2 /T ].
Dpij (a dispersion tensor) is given by
 
hyd
Dpij = Dpdif + Dpij (13)

dif
where Dp is the molecular diffusion coefficient of PCE in the p-phase of the
7/3 hyd
porous medium (= φ 1/3 Sp Dpo by model of Millington et al., 1959), Dpij is a
mechanical dispersion coefficient
" #
1 ∗ 1
= AT qp δij + (AL − AT )|qpi qpj | ∗
φSp qp
P 1/2
2
by Bear (1972), qpi is the p phase Darcy velocity in the i direction, q ∗ = qpi
is the absolute magnitude of the phase velocity, AL is the longitudinal dispersivity
[L], AT is the transverse dispersivity [L], Dp0 is the diffusion coefficient of PCE in
bulk p-phase. Employing the bulk p-phase continuity equations from Equations (5,
6 and 7) to Equations (11) and (12), the transport equation for the three phase
system is described as
   
∂Cw ∂ ∂Cw ∂Cw Rw
φSw = φSw Dwij − qwi + Rw − µw + Cw (14)
∂t ∂xi ∂xj ∂xi ρw
A NUMERICAL STUDY OF SURFACTANT FLUSHING 295
   
∂Co ∂ ∂Co ∂Co Ro
φSo = φSo Doij − qoi + Ro − µo + Co (15)
∂t ∂xi ∂xj ∂xi ρo
   
∂Ca ∂ ∂Ca ∂Ca Ra
φSa = φSa Daij − qai + Ra − µa + Ca (16)
∂t ∂xi ∂xj ∂xi ρa

where w is the aqueous phase, o is the PCE phase, and a is the air phase.

3.2.2. Phase-Summed Equation for Local Equilibrium Transport


Linear partitioning of PCE among phases based on the equilibrium condition may
be described as

Co = 0o Cw , Ca = 0a Cw (17)

where Cp is the concentration of PCE in the p phase, 0o is the equilibrium partition


coefficient for PCE between aqueous solution and NAPL phase, and 0a is the
equilibrium partition coefficient for PCE between aqueous solution and gas phase
(Henry’s constant). Using the equilibrium relations, the transport equations can be
rewritten in terms of a single phase concentration. For aqueous phase-wet systems,
it is logical to retain the aqueous phase concentration, since water will always be
present in the system. The phase-summed transport equation is shown in terms of
aqueous phase concentrations:
 
∗ ∂Cw ∂ ∂Cw ∂Cw
φ = Dij∗ − qi∗ − µ∗ Cw (18)
∂t ∂xi ∂xj ∂xi

where

φ ∗ = φSw + φSo 0o + φSa 0a + 0s (19)

Dij∗ = φSw Dwij + φSo Doij 0o + φSa Daij 0a (20)

qi∗ = qwi + qoi 0o + qai 0a (21)

Rw Ro 0o Ra 0a
µ∗ = µw + µo 0o + µa 0a + µs 0s + + + (22)
ρw ρo ρo

3.2.3. Non-Equilibrium Mass Transfer


The phase-summed formulation of the transport equations may be generalized
to account for non-equilibrium phase partitioning condition by introducing the
concept of ‘apparent partition coefficients’ (Katyal et al., 1991). Under non-equi-
librium conditions, at a given location in time and space, actual phase concentration
296 MINHEE LEE

ratios may differ from the equilibrium ratios defined in the previous section. Thus
apparent partition coefficients could be defined as

Co = 0o∗ Cw (23)

Ca = 0a∗ Cw (24)

where 0p∗ is the apparent partition coefficient for PCE (they will vary in time and
space), and p = o (PCE), a = (air), w = (aqueous phase).
For any two phases, the rate of mass transfer will be described by first-order
mass transfer functions of the form

Row = kow (Cwmax − Cw ) (25)

where Row is the rate of mass transfer of PCE from PCE phase to aqueous solution
phase, Cw is the actual concentration of PCE in aqueous solution phase, Cwmax is the
concentration of PCE that would occur in aqueous phase if it were in equilibrium
with PCE phase, and kow is a mass transfer rate coefficient [T −1 ]. For the case of
mass transfer between oil and aqueous phases when both phases exist at a point in
time and space, the rate of mass transfer is derived from Equations (23) and (25).

Row = kow (0o Cw − 0o∗ Cw ) (26)

where 0o is the equilibrium partition coefficient for PCE between aqueous solution
and PCE phase. This equation may be solved in terms of 0o∗ as
Row
0o∗ = 0o − (27)
kow Cw
which indicates that the apparent partition coefficient may be expressed as the
actual partition coefficient, the actual mass transfer rate, and the concentration in
the water phase. Mass transfer will also occur between aqueous solution and gas
phases. Proceeding in the same manner as for oil-gas mass transfer yields
Rwa
0a∗ = 0a − . (28)
kwa Cw
In order to describe transport with non-equilibrium phase partitioning, the indi-
vidual phase transport equations may be summed as described in Equation (18) but
using apparent partition coefficients. Non-linearity is introduced into the result-
ing phase-summed transport equation due to the dependence of apparent partition
coefficients on concentration as well as mass transfer rates.
A NUMERICAL STUDY OF SURFACTANT FLUSHING 297

3.2.4. Initial and Boundary Conditions


Since the transport equation is written in the phase-summed form, initial conditions
and boundary conditions must be specified in terms of aqueous phase concentra-
tions. Initial conditions for PCE are given as

Cw (xi , t = 0) = Cwo (xi ) on R for t = 0 (29)

where Cwo (xi ) is the initial aqueous phase concentration of PCE at location xi .
Boundary conditions are stipulated by

Cw (xi,t ) = Cwi (xi,t ) on S1 for t > 0 : type-1 , (30)

∂Cw
=0 on S2 for t > 0 : type-2 , (31)
∂xn

Co = qwn Cw ∂Cw
− Dij on S3 for t > 0 : type-3 . (32)
Co = qwn Cw ∂xj

Type-1 boundary condition is the specified PCE concentration Cw1 (xi,t ) within the
porous medium at the boundary region S1 and type-2 boundary condition is the zero
normal concentration gradient on region S2 . The zero gradient condition indicates
zero normal dispersive flux at the boundary. This boundary condition is commonly
applied at boundaries, which have fluid fluxes to permit constituent transport across
the boundaries. If the concentration at the boundary is nonzero due to presence
of chemical near the surface, the type-2 condition will stipulate an inward mass
flux density equal to the aqueous phase concentration at the boundary times the
velocity. In the case for aqueous phase infiltration with zero concentration, a type-3
boundary condition should be specified to force the desired zero-flux condition.

3.3. S IMULATOR DESCRIPTION

Fluid flow equations are highly coupled due to the interdependence of fluid per-
meabilities, saturations and pressures, mandating a simultaneous solution approach
as discussed in the previous section. The transport equation is also highly depend-
ent on the solution of the flow equations due to the occurrence of fluid velocity and
phase saturation terms directly in the transport equation and in the functional forms
for the dispersion coefficients. Thus, the flow equation must be solved concurrently
with or prior to evaluating transport. Due to the weak back-coupling, an efficient
approach is to solve the transport equation serially with the flow equations. The
governing equations for multiphase flow and transport are solved using an efficient
upstream-weighted finite element scheme developed by Huyakorn et al. (1979).
The result of the column experiment with 1% sorbitan monooleate was sim-
ulated using the computer code of MOTRANS for the modeling study and the
solution approach for MOTRANS code is shown in Figure 1. MOTRANS is a
298 MINHEE LEE

Figure 1. Flow chart to solve the coupled-flow and transport equation with MOTRANS.

commercial code of MOFAT developed by Katyal et al. (1991) to simulate the flow
of water, non-aqueous phase liquid and air, and transport of up to five partition-
able species in two-dimensional vertical sections through saturated and unsaturated
zones. The domain for simulation of 1% sorbitan monooleate surfactant flushing
was a 30 × 5 cm section with a water table at a depth of 28 cm, and was described
by five columns of one hundred fifty 1 × 1 cm elements using 180 nodes (Fig-
ure 2). Properties of the porous medium and parameters for modeling are given in
Table I. These values were derived from the experimental results or referred from
literatures. The surfactant flushing was simulated in two stages. PCE was added
under a constant head of ho at the two nodes until the total PCE infiltration was
about 0.8 cm3 (Stage one: Constant head PCE infiltration). All boundaries were no
flow boundaries for the PCE phase. Dissolved phase transport was not considered
during stage one since the total duration of the infiltration was less than 0.005 day.
Stage two involved PCE transport and surfactant flushing that was initiated at the
upper surface at a constant flux of qw = 5 cm day−1 . Surfactant solution flows out
from the bottom boundary of the column at the same velocity. Zero PCE flow was
assumed at all boundaries.
A NUMERICAL STUDY OF SURFACTANT FLUSHING 299

Figure 2. Model grid and element for column simulation.

4. Simulation Results

4.1. R ESULTS OF COLUMN EXPERIMENT

Figure 3 shows the PCE mass removal of surfactant flushing and water flushing
in the column. The concentration of PCE from the effluent of column was about
100 ppm (mg L−1 ) for water flushing and the concentration remained constant
even after 30 pore volumes, suggesting that more than 300 pore volumes would
be needed to remove PCE in the column with water flushing. The use of 1%
sorbitan monooleate solution resulted in an increase in the concentration of PCE
in the effluent of more than 40 times compared to water flushing. Mass balance
was measured by collecting the effluent of column in the Teflon bag and the mass
remaining in the column after surfactant flushing. Ninety one percent of the original
PCE was removed within 12 pore volumes and 5.3 mg (0.08% of original mass)
of PCE remained in the column. The discrepancy between the mass removal from
the effluent solution and original injected PCE mass came from the volatilization
of PCE through the air vent during the column experiment. Approximately 9% of
the original mass was volatilized within five days of surfactant flushing (Lee et al.,
1999).
300 MINHEE LEE

TABLE I
Input parameters for simulation
∗ Hydraulic conductivity: Kswx = Kswz = 0.001 cm s−1 for a Ottawa medium sand
∗ Porosity: 0.4
PCE injected: about 0.85 mL
Residual water saturation: Sm = 0.1
Maximum residual oil saturation: S max o = 0.1
a Parameter α of VG model for soil: 0.08 cm
a Parameter n of VG model for soil: 2
b Longitudinal dispersivity: 0.1 cm
b Transverse dispersivity: 0.03 cm
a Surface tension of PCE, σ : 31.3 dynes cm−1
o
c Surface tension of 1% sorbitan monooleate solution, σ : 44.5 dynes cm−1
w
c Interfacial tension between PCE and 1% sorbitan monooleate, σ : 11.08 dynes cm−1
ow
d Henry’s law constant, K : 1.46 × 10−2 atm m3 mol−1
H
e Diffusion coefficients in bulk water, Dw = 0.656 cm2 d−1
a Diffusion coefficients in oil, Do = 0.7289 cm2 d−1
a Diffusion coefficients in air, Da = 6400 cm2 d−1

∗ From experimental data.


a from Katyal et al. (1991).
b from Pickens et al. (1981).
c from Fountain (1995).
d from Montgomery (1991).
e from Lyman et al. (1982).

Figure 3. Extraction results for surfactant flushing and water flushing.


A NUMERICAL STUDY OF SURFACTANT FLUSHING 301

Figure 4. Initial saturation of surfactant solution for column modeling.

Figure 5. PCE infiltration at two nodes.

4.2. R ESULTS OF SIMULATION

The result of the column experiment with 1% sorbitan monooleate solution was
simulated using MOTRANS. Local equilibrium mass transfer between PCE and
aqueous phases was assumed. Figure 4 shows the initial aqueous phase satura-
tion condition, which ranged from 40–100% based on the result of the column
experiment. 1358 mg (0.8 mL) of PCE was injected at the two injection points, 1%
302 MINHEE LEE

Figure 6. PCE distribution after 5 days of surfactant flushing.

Figure 7. PCE distribution after 21 days of surfactant flushing.

surfactant solution was added at the top at a constant rate of 5 cm−1 day and flowed
out from the bottom (Figure 5). Figures 6, 7 and 8 show the PCE distribution in the
pore space during surfactant flushing. PCE moved downward and most of the PCE
was located above the saturation boundary within 5 days (Figure 6). After 30 days,
the maximum saturation of PCE was less than 1% and most of PCE phase (over
99%) had been extracted (Figure 8).
A NUMERICAL STUDY OF SURFACTANT FLUSHING 303

Figure 8. PCE distribution after 30 days of surfactant flushing.

4.3. H ETEROGENEITY AND KINETIC EFFECTS

Figure 9 compares the results of mass removal from the simulation and from the
column experiment. The trend of the experimental and simulated results is similar,
but the simulation shows faster removal. Some portion of this discrepancy may
be explained by the effect of a more complicated PCE distribution in the actual
columns than in the simulation. In the numerical modeling, the PCE distribu-
tion was specified with a two-dimensional Cartesian coordinates system (X and
Z coordinates). Because the column was modeled as homogeneous, PCE spread
homogeneously across the total width of the column. Thus, all surfactant solution
flowed through the PCE and the effluent concentration from the column reached
the maximum solubility of PCE, assuming local equilibrium conditions. However,
the actual distribution of PCE in the column was probably far more irregular than
in the simulation due to small-scale heterogeneities and thus some portion of the
surfactant solution bypassed the PCE. As a result, even if local equilibrium was
attained, the effluent concentration was less than the maximum solubility. The
resulting dilution effect, arising from the fact that not all water pumped from a
contaminated zone encountered NAPL, is commonly seen in pump-and-treat sys-
tems, and tends to increase the time and pore volumes of surfactant solution needed
to clean a site.
Mass transfer limitations may also be expected to result in lower effluent con-
centrations from those predicted by a local equilibrium simulation. To date, very
few papers have been published pertaining to the modeling of enhanced solubiliz-
ation. In early studies, local equilibrium between the aqueous phase and the NAPL
phase was assumed (Wilson et al., 1991; Clarke et al., 1991; Powers, 1992). How-
304 MINHEE LEE

Figure 9. Comparison of mass fraction between experimental and modeling results.

ever, recent laboratory experiments determined that the rate-limited mass transfer
may control dissolution of NAPL (Bahr, 1990; Pennell et al., 1993; Powers, 1994;
Imhoff, 1994; Mayer et al., 1996). Pennell et al. (1994) investigated the rate-limited
solubilization of entrapped dodecane and PCE in the Ottawa sand and sorbitan
monooleate surfactant system. These data were used to derive the mass transfer
rate coefficient of dodecane in a sorbitan monooleate surfactant solution. A value
of 0.0294/hr was reported (Abriola et al., 1993). A very similar rate (0.034 hr−1 )
was calculated from our experiments from Equation (25) based on the highest
concentration of PCE in the effluent (about 4500 ppm compared to a solubility
of PCE (Cwmax ) of 5483 ppm in the 1% sorbitan monooleate surfactant solution).
In order to evaluate the effect of rate-limited mass transfer on the mass removal,
a non-equilibrium model with a mass transfer coefficient between PCE and surfact-
ant solution phases was implemented for the same conditions described previously.
The result with the non-equilibrium model was also shown in Figure 9. The mass
removal with the non-equilibrium condition was close to the experimental result
with a r-square of 0.97, suggesting that a non-equilibrium condition should be
considered in the analysis of surfactant flushing technology for remediation. Non-
equilibrium solubilization of residual NAPLs would influence remediation time,
optimal injection rate strategies, and the volume of surfactant solution needed to
remove NAPLs from a site. The results from this numerical study demonstrated the
ability of surfactant flushing to enhance the remediation of residual NAPLs in the
unsaturated zone even under rate-limited solubilization conditions.
A NUMERICAL STUDY OF SURFACTANT FLUSHING 305

5. Conclusion

The efficiency of surfactant flushing using the solubility increase for NAPL re-
mediation has been simulated by the mathematical model, based on the flow and
transport mechanism. The column experiment for surfactant flushing was used as
a physical model for the numerical study. Surfactant solution was pumped through
a sand-packed glass column contaminated with PCE. The effluent concentration
was monitored to measure the mass balance. Numerical study results with equi-
librium and non-equilibrium assumptions were compared with the experimental
results. The simulated result with the non-equilibrium condition to consider kinetic
effects was similar to the experimental result with a r-square of 0.97. The results
of mass removal solved by numerical methods suggest that the fitting of the data
and model predictions are sufficiently good to allow quantitative evaluations of
surfactant flushing as a remediation procedure in real contaminated sites. Values of
some parameters which were used for the simulation were obtained from published
papers. These values should be confirmed by experimental works for more precise
numerical study. Irregular distribution of PCE in the column and kinetic effects
decreased the efficiency of surfactant flushing in the experiments. Knowledge of
the distribution of NAPL and the reasonable pumping rate to control the kinetic
effects were necessary to increase the availability of surfactant flushing for NAPL
remediation. The simulation was performed with the assumption that the medium
of the column was homogeneous and thus no preferential flow occurred. Therefore,
the results from the simulation had limitations to describe the small amount of PCE
remaining in the column after flushing. Micro-scale heterogeneity of the medium
should be considered for more accurate simulation. The efficiency of surfactant
flushing may decrease with the heterogeneous conditions of real contaminated
sites, which derive unpredictable NAPL distribution and increase the time and cost
of surfactant flushing. This experimental and numerical study has been performed
only with homogeneous column conditions. Further work is needed to evaluate the
effect of heterogeneity on surfactant flushing in numerical study. The recycling of
surfactant and the treatment of surfactant sorbed on the medium after surfactant
flushing are further issues to be addressed for more effective remediation.

Appendix I. Nomenclature

AL = Longitudinal dispersivity [L]


AT = Transverse dispersivity [L]
Cwmax = Concentration of PCE in aqueous phase with equilibrium [M L−3 ]
Cp = Concentration of PCE in p-phase [M L−3 ]
306 MINHEE LEE

dif
Dp = Molecular diffusion coefficient of PCE in the p-phase of the porous
medium
hyd
Dpij = Mechanical dispersion coefficient of p-phase in i and j direction
Dpo = Diffusion coefficient of PCE in bulk p-phase
g = Gravity [L/T 2 ]
hp = Water height-equivalent pressure head of phase p [L]
Jpi = Mass flux density of PCE in p-phase in the i direction [M/L2 T ]
Kpij = The p-phase conductivity tensor [L/T ]
kow = Mass transfer rate coefficient between PCE and aqueous phase [T −1 ]
ni = Outward normal unit vector
Pp = The p-phase pressure [L]
qip = Darcy velocity of phase p in the i direction [L/T ]
Rp = Net mass transfer per unit porous media volume into or out phase p
[M/L3 T ]
Sp = p-phase saturation
uj = Unit gravitational vector measured positive upwards where z is elevation
xi = Cartesian spatial coordinate (i, j = 1, 2, and 3)
w = Aqueous phase (water or surfactant solution phase)
p = Phase index (a: air phase, o: PCE phase, and w: aqueous phase)
ρp = Density of phase p [M/L3 ]
ρrp = (= ρp /ρw ) is the p-phase specific gravity
γp = Net production (+) or decay (–) of PCE due to reactions within phase p
[M/L3 ]
µp = Apparent first-order decay coefficient
φ = The porosity
0o = Equilibrium partition coefficient for PCE between aqueous and PCE
phase
0a = Equilibrium partition coefficient for PCE between aqueous and gas
phase (Henry’s constant)
0p∗ = Apparent partition coefficient for PCE

References

Abriola, L. M., Dekker, T. J. and Pennell, K. D.: 1993, Environ. Sci. Technol. 27(12), 2341.
Abdul, A. S. and Ang, C. C.: 1994, Ground Water 32(5), 727.
Bear, J.: 1972, Dynamics of Fluids in Porous Media, American Elsevier, New York, pp. 101–132.
Bedient, P. B., Rifai, H. S. and Newell, C. J.: 1996, ‘Ground Water Contamination: Transport and
Remediation’, Prentice Hall PTR, Englewood Cliffls, New Jersey, pp. 349–386.
A NUMERICAL STUDY OF SURFACTANT FLUSHING 307

Brown, C. L., Pope, G. A., Abriola, L. M. and Sepehrnoor, K.: 1994, Water Resources Res. 30(11),
2959.
Clarke A. N., Plumb, T. D., Subramanyan, T. K. and Wilson, D. J.: 1991, Sep. Sci. Technol. 26(3),
301.
Delshad, M., Lenhard, R. J., Oostrom, Mart, Pope, G. A. and Yang, S.: 1999, ‘Mixed-Wet Hysteretic
Relative Permeability and Capillary Pressure Model in a Chemical Compositional Reservoir
Simulator, Proceedings of the SPE Symposium on Reservoir Simulation, Soc. Pet. Eng. (SPE),
Richardson, TX, U.S.A., SPE 51891 pp. 135–146.
Ellis, W. D., Morgan, D. R. and Ranjithan, S. R.: 1986, EPA/600/2-85/129, Hazardous Waste
Engineering Research Laboratory.
Faust, C. R.: 1985, Water Resources Res. 21, 587.
Fountain, J. C.: 1995, ‘Bench Tests For a Treatability Study of Surfactant Enhanced Aquifer Re-
mediation at Hill Air Force Base, Utah’, Final report, Buffalo, State University of New York at
Buffalo, pp. 235.
Fountain, J. C., Starr, R. C., Middleton, T. M., Beikirch, M. G., Tayler, C. and Hodge, D.: 1996,
Ground Water 34(5), 910.
Hunt, J. R., Sitar, N. and Udell, K. S.: 1988, Water Resources Res. 24(8), 1247.
Huyakorn, P. S. and Nilkuha, K.: 1978, Applied Mathematical Modeling, 3, 7.
Imhoff, P. T., Jaffe, P. R. and Pinder, G. C.: 1994, Water Resources Res. 30(2), 307.
Katyal, A. K., Kaluarachchi, J. J. and Parker, J. C.: 1991, ‘MOFAT: A Two-Dimensional Finite
Element Program for Multiphase Flow and Multicomponent Transport’, EPA/600/2-91/020.
Lagouski, A. A.: 1996, ‘Surfactant Adsorption on Porous media as a Function of Surfactant Type
and Soil Type’, Master thesis, Department of Geology, State University of New York at Buffalo,
New York (unpublished).
Lee, M.: 1998, Eviron. Technol. 19(11), 1073.
Lee, M. and Fountain, J. C.: 1999, J. Soil Contamination 8(1), 39.
Lenhard, R. J. and Parker, J. C.: 1987, Water Resources Res. 23(12), 2197.
Lyman, W. J., Reehl, W. F. and Rosenblatt, D. H.: 1982, ‘Handbook of Chemical Property Estimation
Methods’, McGraw-Hill, New York.
Mackay, D. M. and Cherry, J. A.: 1989, Environ. Sci. Technol. 23(6), 630.
Mayer, A. S. and Miller, C. T.: 1996, Water Resources Res. 32(6), 1151.
Mercer, J. W. and Cohen, R. M.: 1990, J. Contam. Hydrol. 6, 107.
Millington, R. J. and Quirk, J. P.: 1959, Science 130, 100.
Montgomery, J. H.: 1991, Groundwater Chemicals Field Guide, Lewis Publishers.
Pankow, J. F. and Cherry, J. A.: 1996, ‘Dense Chlorinated Solvents and other DNAPLs in
Groundwater’, Waterloo Press, pp. 3–28.
Parker, J. C.: 1989, Reviews of Geophysics 27(3), 311.
Parker, J. C. and Lenhard, R. J.: 1987, Water Resources Res. 23(12), 2187.
Pennell, K. D., Abriola, L. M. and Weber Jr., W. J.: 1993, Environ. Sci. Technol. 27(12), 2332.
Pennell, K. D., Jin, M., Abriola, L. M. and Pope, G. A.: 1994, J. Contam. Hydrol. 16, 35.
Picken, J. F. and Grisak, G. E.: 1981, Water Resources Res. 17, 1191.
Powers, S. E., Abriola, L. M. and Weber, W. J.: 1994, Water Resources Res. 30, 321.
Shiau, B. J., Sabatini, D. A. and Harwell, J. H.: 1994, Ground Water 32(4), 561.
Wilson, D. J. and Clarke, A. N.: 1991, Sep. Sci. Technol. 26(9), 1177.
Wilson, J. L. and Conrad, S. H.: 1984, ‘Is Physical Displacement of Residual Hydrocarbons a Real-
istic Posibility in Aquifer Restoration?’, Proceedings of the NWWA/API Conference on Petroleum
Hydrocarbons and Organic Chemicals in Groundwater – Prevention, Detection and Restoration,
Water Well Journal Publishing Company, p. 274.

You might also like