| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a Department of Earth Sciences University of North Carolina at Wilmington, Wilmington, NC 28403
b School of Geography and Geology, McMaster University, Hamilton, ON L8S 4L8, Canada
* Corresponding author (henrye{at}uncw.edu)
Received 26 November 2002.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONBACKGROUND: EFFECTS ON...EXPERIMENTAL STUDIES OF...NUMERICAL MODELS OF SURFACTANT...SUMMARYREFERENCES |
|---|
Abbreviations: DNAPL, dense nonaqueous phase liquid • LNAPL, light nonaqueous phase liquid • MA, myristyl alcohol • NAPL, nonaqueous phase liquid • PCE, tetrachloroethylene • TRI, Texas Research Institute • 1D, one-dimensional • 2D, two-dimensional
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONBACKGROUND: EFFECTS ON...EXPERIMENTAL STUDIES OF...NUMERICAL MODELS OF SURFACTANT...SUMMARYREFERENCES |
|---|
The primary impact of surfactants on unsaturated flow in soils is through the dependence of surface tension of the air–water interface on surfactant concentration. Because capillary pressure is a function of surface tension, surfactant-induced reductions in surface tension can cause proportional decreases in capillary pressure. The surfactant-induced changes in capillary pressure result in a scaling of the moisture content–pressure head relationship along the pressure axis (e.g., Salehzadeh and Demond, 1994; Smith and Gillham, 1994, 1999; Lord et al., 1997, 2000; Dury et al., 1998). Furthermore, differences in capillary pressure can exist between surfactant-free and surfactant-contaminated zones that generate associated hydraulic gradients in unsaturated porous media. The magnitude of the surfactant-induced capillary pressure gradient will depend on the surface-active properties of the surfactant, including surfactant type, surfactant concentration, and solution chemistry. The induced hydraulic gradients can cause significant unsaturated flow.
A considerable body of work exists related to the use of surfactants in the petroleum industry for enhanced oil recovery (Pope and Wade, 1995). The use of surfactants for remediation of the saturated zone has also received much attention (e.g., Fountain et al., 1991; 1996; West and Harwell, 1992; Pennell et al., 1993; Martel et al., 1998; Oostrom et al., 1999; Londergan et al., 2001; Ramsburg and Pennell, 2002). Compared with the quantity of data available on surfactant use in the aforementioned areas, relatively little data is available regarding the effects of surfactants on flow and transport in the vadose zone. The sparseness of data related to vadose zone processes may be due to the fact that much of the subsurface surfactant-enhanced remediation research has focused on dense nonaqueous phase liquids (DNAPLs), which are primarily of concern below the water table. In addition, instrumentation and water sampling is often more challenging within the vadose zone than below the water table. However, the use of surfactants to aid subsurface remediation by enhancing the solubility and/or the mobility of organic contaminants need not be confined to the saturated zone. Surfactants can be used within the unsaturated zone and capillary fringe for the remediation of light nonaqueous phase liquids (LNAPLs), residual DNAPL, and sorbed organic compounds. An understanding of surfactant effects on unsaturated flow and transport is critical to the design of systems for the efficient remediation of the vadose zone.
Our objective is to present the state of knowledge regarding the effect of surfactants on the unsaturated flow of water and associated solute transport within the vadose zone. The scope of this paper is limited to experimental data and modeling of unsaturated systems containing surfactants. Although the popularity of surfactant use for the remediation of nonaqueous phase liquid (NAPL)–contaminated regions is one of the justifications for this review, the majority of the available datasets on this topic are from air–water (two fluid phase) systems (i.e., no NAPL phase present). Our primary focus is the effect of surfactants on unsaturated flow and transport, rather than on surfactant use in subsurface remediation. The reader is directed to a number of other papers that provide general information about surfactants and surfactant use in subsurface remediation (e.g., West and Harwell, 1992; Kuhnt, 1993; Edwards et al., 1994; Sabatini et al., 1996, 1998; Mulligan et al., 2001). Further, the review presented here will focus on water-wetted porous media. The topic of surfactant use in hydrophobic soils is addressed elsewhere (e.g., DeBano, 2000; Bauters et al., 2000). This review is presented in four sections: a background discussion of surfactant effects on unsaturated flow, a summary of unsaturated flow experiments involving surfactants (laboratory and field scale), a discussion of flow and transport modeling related to surfactants in the vadose zone, and areas for additional research.
|
BACKGROUND: EFFECTS ON UNSATURATED HYDRAULIC CHARACTERISTICS |
|---|
|
TOPABSTRACTINTRODUCTIONBACKGROUND: EFFECTS ON...EXPERIMENTAL STUDIES OF...NUMERICAL MODELS OF SURFACTANT...SUMMARYREFERENCES |
|---|
–
) and hydraulic conductivity–moisture content (K–) constitutive relationships, where is the moisture content, is the soil water pressure head, and K is the unsaturated hydraulic conductivity. The primary impact of surface-active organic solutes on unsaturated flow is through the dependence of soil water pressure head on surface tension (e.g., Bear, 1972):
 | [1] |
is surface tension, 
is the solution density, g is the gravitational acceleration, 
is the contact angle, and r is the radius of an equivalent circular tube. For a given moisture content and contact angle, Eq. [1] indicates that a decrease in surface tension will result in a proportional decrease in capillary pressure (the negative soil water pressure). The magnitude of the decrease in surface tension will be a function of surfactant type and concentration. Equation [1] also illustrates that if a surfactant alters the contact angle, the soil water pressure head at a given moisture content and surface tension will be modified. Thus, in a homogeneous, uniformly wetted, unsaturated porous medium, gradients in surfactant concentration can result in capillary pressure gradients. The magnitude of the capillary pressure gradient between regions with different surfactant concentrations will depend on the magnitude of the surfactant-induced changes in surface tension and/or contact angle, as well as the moisture content of the regions of interest and the water retention characteristics of the porous medium.
Because moisture retention curves are not commonly measured for systems containing surfactants, methods have been devised to predict the water retention in a medium contaminated with a surfactant on the basis of knowledge of the surface-active behavior of the surfactant and the retention behavior of pure water in the same porous medium. An approach that is commonly used to describe the effect of a surfactant on the – relationship in air–water systems is based on scaling the pressure term for a system that is wetted with a reference fluid (generally pure water) by the ratio of the concentration-dependent surface tension of the solution of interest to the surface tension of the reference fluid (after Leverett, 1941). Scaling of the pressure head term by the relative surface tension can be expressed as
 | [2] |
is the measured pressure head at water content and reference concentration co (co = zero when pure water is the reference solution), and 
is the calculated scaled pressure head at the same water content and at surfactant concentration c. The scaling technique in Eq. [2] has been experimentally shown to be adequate for systems in which the surfactant resides primarily at the air–water interface or in bulk solution.
Surfactants that partition strongly to the water–solid interface may modify the wettability of the system (i.e., the surfactant alters the contact angle), and a more complex scaling approach is required. A second scaling method that has been used to describe surfactant effects on moisture retention considers surfactant-induced changes to surface tension and contact angle:
 | [3] |
- r)/(s - r), with r and s the residual and maximum moisture content, respectively] and is the apparent contact angle. The subscripts 1 and 2 denote the two systems with differing surface tensions and contact angles. The notation used in Eq. [3] is similar to that used by Demond and Roberts (1991), though Demond and Roberts (1991) also included a curvature correction function (Melrose, 1965), which is discussed below. As noted by Rockhold et al. (2002), in practice it may be difficult to distinguish changes in retention behavior that are due to changes in surface tension from those due to changes in contact angle.
On the basis of the work of Demond and Roberts (1991) in organic liquid–water systems, Desai et al. (1992) predicted capillary pressure relationships for an air–water system containing a sorbing surfactant (cetyltrimethylammonium bromide) by scaling the retention curves from a pure water–air system with a scaling factor of the form:
 | [4] |
is surface tension, is the intrinsic contact angle that is measured on a smooth surface (Morrow, 1976), r is the apparent contact angle (calculated by correcting the intrinsic contact angle for roughness; see Desai et al. [1992] for a description of procedure from Morrow [1975]), Z() is the curvature correction function (after Melrose, 1965), is the measured pressure head at effective saturation Se, and is the calculated scaled pressure head at the same effective saturation and at surfactant concentration c. The introduction of the curvature correction function accounts for pore geometry deviations from the simple capillary tube model. These scaling techniques shift the moisture retention curve along the pressure axis. They do not account for surfactant-induced variations in the maximum and/or residual moisture contents that have been observed in some air–water systems containing surfactants (Desai et al., 1992; Salehzadeh and Demond, 1994). The use of effective saturation rather than water content further masks the changes in the residual and saturated moisture contents and makes it difficult to discern the variability in these parameters from the scaled retention curves (Demond and Roberts, 1991).
As an example of the application of retention curve scaling, measured retention curves for Ottawa sand wetted with either pure water or 7% 1-butanol solution are shown in Fig. 1a (Smith and Gillham, 1999). The surface tension of 7% butanol is approximately two-thirds less than that of water, and a proportional shift is seen along the pressure axis in the retention characteristics of the 7% butanol curves relative to the pure water curves. In Fig. 1b (Smith and Gillham, 1999) the 7% butanol retention curves are scaled by the relative surface tension (after Eq. [2]) for comparison with the pure water curves. Scaling of the measured 7% butanol curves was accomplished by multiplying each of the pressure heads measured for the butanol system by the ratio of the surface tension of pure water ( = 72 mN m-1) over the surface tension of the butanol solution ( = 25.8 mN m-1). The scaled pressure heads were then plotted at their respective moisture contents in Fig. 1b. Comparison of the scaled 7% butanol curves with the pure water retention curves in Fig. 1b can be used to evaluate how well Eq. [2] predicts retention behavior in the pure water system using retention data from the 7% butanol system and the surface tensions of the pure water and butanol systems. Visual comparison of the measured pure water curves and scaled 7% butanol curves in Fig. 1b shows that for the silica sand–butanol solution system studied by Smith and Gillham (1999), the surface tension–based scaling technique of Eq. [2] accounts for almost all the difference between the two systems. A summary of research demonstrating various techniques used for the scaling of experimental retention data in air–water systems containing surfactants is given in Table 1.
|
|
The above observations are generalizations. The magnitude of these effects will be system-specific and will depend on the magnitude of the variation between the retention curves for the surfactant-free and surfactant-contaminated regions. This variation will be a function of the degree of surface tension depression in the contaminated region, the retention characteristics of the porous medium, and the wetting–drying history of the system.
Effect of Surfactants on Unsaturated Hydraulic Conductivity
It is well known that surfactants can have a significant impact on hydraulic conductivity in saturated groundwater systems. The effect of surfactants on hydraulic conductivity depends on the physicochemical characteristics of the surfactant and the soil, as well as surfactant concentration and pore water chemistry (e.g., Allred and Brown, 1994). Tumeo (1997) presented a survey of the major mechanisms that may contribute to surfactant-induced changes in hydraulic conductivity. The hydraulic conductivity changes may occur as a result of changes to the intrinsic permeability or fluid characteristics. Tumeo (1997) presented four categories of surfactant-induced hydraulic conductivity changes. The first category involves direct physical effects attributed to surfactants, including decreases in hydraulic conductivity due to surfactant-induced increases in viscosity, increases in hydraulic conductivity due to reductions in contaminant clogging, decreases in hydraulic conductivity due to enhancement of contaminant clogging, decreases in hydraulic conductivity due to the formation of macroemulsions, and hydraulic conductivity decreases due to the adsorption of surfactants on the solid surface. The second category is based on the ability of surfactants to disrupt soil aggregate structure. Surfactant interactions at the soil surface may decrease soil aggregate stability and result in the release of fine soil particles. This may either enhance or degrade hydraulic conductivity, depending on whether the fines are readily transported with advective flow or are trapped in pore spaces, resulting in clogging. The third explanation for surfactant-induced changes to hydraulic conductivity is attributed to surfactant–clay interactions. The effect of surfactants on the diffuse double layer of the clay particles may result in clay swelling and resultant decreases in hydraulic conductivity. Surfactants can also result in the dispersion of clay particles, which may increase or decrease hydraulic conductivity depending on whether the dispersion results in "channeling" or clogging. The fourth explanation involves secondary chemical reactions that may be induced by surfactants. These reactions may cause mineral dissolution and piping (an increase in porosity due to mineral dissolution), processes which increase hydraulic conductivity, or mineral precipitation, which decreases hydraulic conductivity (Tumeo, 1997).
The mechanisms outlined above may all contribute to surfactant-induced hydraulic conductivity changes to some degree (Tumeo, 1997). However, there have been relatively few studies focused on surfactant-induced hydraulic conductivity changes in the vadose zone, so our understanding of these effects, which are based primarily on perceived changes to hydraulic conductivity during surfactant transport experiments, is more speculative than conclusive.
There are a few notable examples of surfactant-induced hydraulic conductivity changes in unsaturated systems. Before their work in unsaturated systems, Allred and Brown (1994) showed that anionic, cationic, nonionic, and amphoteric surfactants were all capable of causing hydraulic conductivity decreases in a sand and a loam under saturated conditions. The ionic surfactants caused a greater reduction in hydraulic conductivity than the nonionic surfactants and the magnitude of the reduction varied as a function of surfactant type and concentration, soil type, soil organic matter content, and electrolyte concentration. The hydraulic conductivity reductions in the sand were as high as 47%, and viscosity effects could account for much of the decrease. Hydraulic conductivity was reduced in the loam by as much as two orders of magnitude, and it was speculated that viscosity effects, surfactant precipitation, and soil structure alteration were responsible for the observed behavior.
Allred and Brown later used horizontal column experiments to study surfactant mobility in unsaturated soils (Allred and Brown, 1995, 1996a, 1996b, 2001; Allred et al., 1996). Surfactant effects on mobility were evaluated, in part, by determining unsaturated diffusivity: D() = K()d/d, where D() is the diffusivity, K() is the unsaturated hydraulic conductivity, is the soil water pressure head, and is the moisture content. The term d/d is also the reciprocal of the soil moisture capacity, which describes the slope of the moisture characteristic curve. Allred and Brown (1995)( 1996b) found that anionic surfactants caused decreases in diffusivity. A similar result was found by Mustafa and Letey (1971) for nonionic surfactants. However, as noted by Allred and Brown (1995), it is difficult to directly relate the observed decreases in diffusivity to decreases in unsaturated hydraulic conductivity since diffusivity depends on d/d as well as K(). As seen in Fig. 1, surfactant-induced changes to surface tension shift the moisture retention curve for pure water along the pressure axis and make the curve steeper. This results in a decrease in d/d with decreasing surface tension. Because moisture content and pressure varied spatially in the horizontal columns, the effects of hydraulic conductivity reduction and d/d reduction could not be separated. Allred et al. (2001) conducted laboratory experiments to evaluate the potential for surfactant pretreatment of soil to minimize the leaching of animal waste due to the ability of surfactants to reduce flow.
Smith and Gillham (1999) considered surfactant effects on unsaturated hydraulic conductivity in a study related to the effect of a surfactant (1-butanol) on one-dimensional unsaturated flow. The 7% butanol (w/w) solution used by Smith and Gillham (1999) had a viscosity approximately 33% greater than that of pure water. Scaling of the unsaturated hydraulic conductivities for Ottawa sand wetted with pure water by the ratio of relative viscosities (viscosity of pure water/viscosity of surfactant solution) predicted measured values of hydraulic conductivity for the surfactant system within experimental error. Thus, differences in the unsaturated hydraulic conductivities for the pure water– and 7% butanol–wetted systems were primarily due to the surfactant concentration-dependent differences in the viscosities of pure water and 7% butanol.
Other observations of surfactant-induced hydraulic conductivity changes in unsaturated systems have been incidental, rather than primary experimental objectives. During a field-scale study of surfactant effectiveness for remediation of oil contamination within the unsaturated zone and capillary fringe, Bettahar et al. (1999) noted that when surfactant solution (a 1:1 mixture of sodium dioctyl sulfosuccinate and ethoxylated (20OE) sorbitan trioleate) was applied at the soil surface, the infiltration rate decreased with time. This behavior was attributed to a surfactant-induced reduction in hydraulic conductivity. Subsequent laboratory column experiments showed that the flow rate decreased significantly after 10 h, and flow essentially stopped at approximately 20 h (Bettahar et al., 1998). The reduction in hydraulic conductivity was ascribed to the formation of liquid crystals, which had reached a size sufficient to cause plugging after 20 h.
It is also worth noting that there have been several unsaturated systems in which no apparent change in hydraulic conductivity was noted during surfactant flushing. Laboratory-scale studies of surfactant flushing that noted no surfactant-induced change in unsaturated hydraulic conductivity include Walker et al. (1998), Bruell et al. (1997)(1998), Lee (1998), Lee and Fountain (1999), and Powelson and Mills (1998). Since these studies made no affirmative or negative note of surfactant-induced hydraulic conductivity changes, it is unclear whether there actually were no changes or whether changes occurred but were not detected because of the experimental design and instrumentation.
As will be discussed in the next section, surface tension decreases caused by surfactants can result in drainage events within the vadose zone. A decrease in moisture content would result in a decrease in the unsaturated hydraulic conductivity (e.g., Smith and Gillham, 1999; Bruell et al., 1998). Although a decrease in hydraulic conductivity in this fashion is a result of surfactant-induced drainage, it is not a surfactant-induced hydraulic conductivity change induced by the mechanisms described above.
Surfactant-induced hydraulic conductivity changes may be even more difficult to detect at the field scale. For example, Abdul et al. (1992) reported that, during a field test of surfactant application on the soil surface for remediation of a shallow contaminated region, "To our surprise...ponding of the aqueous solution on the surface of the plot resulted." Though they speculated this unexpected ponding was due to the presence of a very compact, clay-cement layer in the upper foot of the surface, it is possible that physicochemical surfactant processes may have contributed to the lower than expected infiltration rate.
The variety of scenarios described above points to the complex nature of surfactant-induced hydraulic conductivity changes. The type and magnitude of the induced changes can be a function of physicochemical surfactant and soil characteristics, surfactant concentration, surfactant-contaminant interactions (e.g., emulsion formation), and solution chemistry. Thus, it may be prudent to conduct soil-, solution-, and surfactant-specific laboratory experiments before field-scale implementation of surfactant-based remediation schemes.
|
EXPERIMENTAL STUDIES OF SURFACTANT-INDUCED FLOW |
|---|
|
TOPABSTRACTINTRODUCTIONBACKGROUND: EFFECTS ON...EXPERIMENTAL STUDIES OF...NUMERICAL MODELS OF SURFACTANT...SUMMARYREFERENCES |
|---|
Karkare et al. (1993) applied the closed horizontal column technique to study the ability of 33 surfactants to move water and found that, in order for surfactants to be effective at moving water, "they must be water-insoluble and must form a condensed solid film at their equilibrium spreading pressure." Tschapek et al. (1984) and Karkare and Fort (2002) reached a similar conclusion. However, Henry et al. (1999) showed that surfactant-induced capillary pressure gradients caused by a high-solubility alcohol, 1-butanol, were sufficient to produce effective water movement in the closed soil columns. Henry et al. (1999) also demonstrated that the nature of surfactant-induced flow caused by a highly soluble compound such as butanol was considerably different than the flow caused by a relatively insoluble surfactant (MA). The differences in the observed flow behavior in the butanol and MA systems was due to the fact that MA only caused large surface tension decreases at concentrations near those required for monolayer coverage and was not transported in sufficient quantities to cause surface tension reduction in the previously clean portion of the column (Karkare and Fort, 1996). As a result, the region of surfactant-induced surface tension reduction remained confined to the original surfactant source zone when a low-solubility surfactant was used. Since flow only occurred at surfactant concentrations equal to or greater than those required for monolayer coverage of the air–water interface in the MA system, the closed column technique was applied to the determination of air–water interfacial areas in unsaturated sand columns (Karkare and Fort, 1996) and glass beads (Silverstein and Fort, 1997) based on the size of MA molecules and the number of molecules required to induce flow.
As discussed above, in a uniformly wetted porous medium, surfactant-induced capillary pressure gradients will induce flow from surfactant-contaminated regions and toward surfactant-free regions. However, the nature of the induced flow will depend, in part, on the relative mobility of the surfactant. Results from closed horizontal column experiments using the low-solubility (i.e., low mobility) surfactant MA are shown in Fig. 2a (Henry et al., 1999). As shown in Fig. 2a, water drained from the surfactant-containing half of the column (the right half of the column) and wetted the half that was originally surfactant free (left half of the column). In contrast, dissolved butanol is highly soluble, can affect surface tension reductions over a large range of concentrations, and was readily transported during surfactant-induced flow, resulting in a propagation of the region of depressed surface tension (Fig. 2b) (Henry et al., 1999). This caused additional drainage in the left-hand side near the center of the column and additional wetting near the left-hand boundary of the column.
|
Henry and Smith (2002) used two-dimensional (2D) sandbox experiments to investigate the infiltration of a surfactant solution from a point source located on the soil surface above an unconfined aquifer with ambient groundwater flow. They used the same type of surfactant solution (7% butanol) used in one-dimensional (1D) work of Smith and Gillham (1999) and Henry et al. (1999). The surfactant solution contained dye (FD&C Blue #1) as a visual tracer. Surfactant effects on unsaturated flow were monitored visually through a glass plate on the front of the sandbox, as well as through the in situ measurement of soil water pressure head and moisture content. As in the 1D work of Smith and Gillham (1999), large changes in moisture content and pressure head were associated with the advance of the butanol solution in the 2D system, and capillary fringe depression was observed as butanol solution displaced pure water within the vadose zone below the point source. An aspect of unsaturated flow not captured by the earlier 1D work was the manner in which the surfactant-induced drainage of the vadose zone propagated within the 2D sandbox. In addition to observed drainage of the vadose zone and depression of the capillary fringe directly below the point source, surfactant-induced drainage of the vadose zone propagated across the sandbox as a wedge-shaped zone of reduced moisture contents (referred to as a "drainage wedge" by Henry and Smith, 2002). This drainage wedge was positioned within the original capillary fringe. The toe of the drainage wedge pointed in the down-gradient direction and the bottom of the drainage wedge was horizontal and bounded by the top of the depressed capillary fringe. The surfactant-induced drainage of the 2D flow cell after 202.5 h of surfactant solution application at the point source is seen in Fig. 3 (Henry and Smith, 2002). The wedge-shaped propagation of the drainage front was attributed to the transport of butanol in the depressed capillary fringe, which was rapid relative to butanol transport in the unsaturated zone. The drainage wedge, wetted with low surface tension butanol solution, was overlain by a saturated region (the undrained region of the original capillary fringe) containing pure water. Henry and Smith (2002) speculated that surfactant-induced capillary pressure gradients between the drainage wedge and the overlying tension-saturated zone prevented the downward movement of pure water from the saturated region through the drained zone. Horizontal propagation of surfactant-induced drainage in the direction of ambient groundwater flow was also observed in 2D gasoline remediation experiments conducted by the Texas Research Institute (TRI) (1979). However, no mention of a wedge-shaped drainage front was made. This may be due to the fact that the fine sand used by Henry and Smith (2002) had an uncontaminated capillary fringe height of approximately 55 to 60 cm, while the sand used by TRI (1979) had an uncontaminated capillary fringe height of 12 to 15 cm.
|
Since capillary pressure is a function of the effective pore radius, soil texture can have an impact on surfactant-induced flow perturbations. Walker et al. (1998) investigated surfactant use for NAPL remediation in the vadose zone by applying surfactant solution (4% Triton X-100; Promega Corp., Madison, WI) on the soil surface of a 2D flow cell that had a fine soil layer packed in the center of a coarse medium. As the surfactant plume moved downward from the unsaturated coarse sand into the wetter fine sand layer "spontaneous" spreading of the surfactant solution occurred within the fine layer and resulted in drainage of the fine layer. The enhanced spreading of surfactant solution within the fine layer, relative to spreading in the coarse layer, was primarily due to the fact that capillary effects are greater in finer-grained materials. The effect of soil texture on surfactant-induced flow was also shown by Karkare and Fort (1993) in their 1D experiments.
Most of the experimental observations presented above provide evidence of flow caused by surfactant-induced capillary pressure gradients. A second aspect of surfactant occurrence within the vadose zone is related to the effect of surfactants on the retention properties of a porous medium. For example, Gradwell (1958) investigated the use of a surfactant (PR51) to improve drainage of pasture lands. While he concluded that this approach was impractical, he did observe reduced water retention. Zartman and Bartsch (1990) surveyed the ability of 17 surfactants to enhance drainage from dewatered soil columns. The addition of 150 mL of surfactant solution increased gravity-induced outflow from the columns (i.e., the volume of fluid recovered following surfactant application was greater than the volume of surfactant solution applied). Since surface tension is a function of surfactant concentration, the volume of fluid recovered from the columns was correlated to surfactant concentration. Surfactant-enhanced drainage was also seen in the 1D NAPL remediation experiments of Bruell et al. (1997)(1998). They found that as surfactant solution was applied, column dewatering occurred and the effluent flow rate subsequently decreased due to "reductions in interconnected water filled pores through which flow occurred, leading to a reduction in the effective permeability" (Bruell et al., 1998). Allred and Brown (1996a) also noted surfactant-induced modifications in retention behavior. They observed a notch in the moisture content profiles during 1D, unsaturated, surfactant transport experiments. The notch coincided with the surfactant solute front and was attributed to variation in the water retention relationship in the transition zone between high and low surfactant concentrations. The effect of surfactants on moisture retention was also shown by Law (1964), who noted that at a given pressure head, both a nonionic surfactant and a fatty alcohol reduced the water-holding capacity of a sandy loam.
A clear manifestation of the effect of surfactants on retention behavior is the depression of the capillary fringe that occurs when pure water is displaced by lower surface tension surfactant solution. In addition to the examples given above (TRI, 1979; Walker et al., 1998; Smith and Gillham, 1999; Henry and Smith, 2002), capillary fringe depression has been documented in several studies related to surfactant-based remediation techniques. In the 2D sandbox experiments of Jawitz et al. (1998), 70% ethanol solution was applied at the up-gradient boundary of a 2D flow cell with an ambient horizontal groundwater flow. As the solution migrated across the box and displaced pure water, a transient depression of the capillary fringe from an initial height of 9 cm to a final height of 5 cm was observed, as expected, based on the ratio of the surface tension of the surfactant solution (40 mN m-1) to that of water (72 mN m-1). The effect of the capillary fringe depression on transport was seen as a less-prolonged tail in the breakthrough curve for a dye tracer (Jawitz et al., 1998).
Capillary fringe depression can be caused by dissolved contaminants as well as by surfactants used for remediation. Chevalier et al. (1998) studied NAPL lens configurations by applying unleaded gasoline through an injection well located in the unsaturated zone of a 2D flow cell with a fixed and level water table (i.e., no horizontal groundwater flow). As the gasoline reached the top of the capillary fringe, the height of the capillary fringe was depressed from 14.1 to 10.7 cm across the entire width of the box, despite the fact that the NAPL lens remained confined to the middle of the box. The capillary fringe depression was attributed to "capillary pollution" (after Schiegg, 1984) caused by dissolved gasoline. The subsequent application of surfactant through the injection well further reduced the capillary fringe height to 5 cm. In a separate experiment, the application of surfactant was not preceded by gasoline and resulted in a capillary fringe depression from 15.4 to 7.4 cm (Chevalier et al., 1998).
Although there have been several field-scale surfactant-based remediation experiments (Jafvert, 1996), data collected from the vadose zone during such experiments are limited, and evidence of surfactant-induced flow perturbations is even sparser. There appear to be several reasons for the lack of data from the vadose zone. First, many of the field experiments were designed for the remediation of DNAPL contamination. Because DNAPLs reside primarily below the water table, remedial system design and monitoring logically focus on the saturated zone. Second, monitoring of the vadose zone, which may require the use of suction samplers, tensiometers, and in situ moisture content measurements, is somewhat more complex and more instrument-intensive than monitoring of the saturated zone. Finally, the use of surfactants in remedial schemes is still somewhat more of an art than engineering (Selvakumar et al., 1995, as referenced by Tumeo, 1997). Thus, in some field-scale surfactant applications the end may be more important than the means, resulting in only nominal consideration of the actual mechanisms of the remediation process. In such an approach, monitoring is generally limited to the measurement of contaminant and surfactant concentrations in soil and water samples before and after surfactant application. This type of monitoring scheme does not capture information on transient surfactant flow and transport behavior in the vadose zone.
Field-scale implementations of surfactant flushing, which may potentially yield data regarding surfactant-induced flow phenomena, may involve the application of surfactant solution at the soil surface or within the vadose zone for the remediation of LNAPL contamination. For example, Abdul et al. (1992) applied surfactant solution at the soil surface to remediate a vadose zone and shallow saturated zone (depth to the water table was approximately 1.2 m) contaminated with oils and polychlorinated biphenyls. Though Abdul et al. (1992) monitored moisture content within the vadose zone occasionally during their study, no surfactant-related flow effects were reported. Bettahar et al. (1999) investigated the effect of applying surfactant solution at the soil surface, as well as through a well located at the top of the capillary fringe, on the remediation of shallow (water table approximately 2.5–3 m below soil surface) oil-contaminated soil. Aside from the surfactant-induced hydraulic conductivity reduction, which was discussed above, no surfactant-induced flow perturbations were reported.
Without in situ measurements of pressure or moisture content, surfactant-induced flow perturbations would be difficult to discern in large-scale studies. Using standard saturated zone monitoring techniques, the most likely way for surfactant effects to be detected would be as a transient increase in water table elevation due to surfactant-induced drainage of the unsaturated zone and capillary fringe. In a large-scale (3 by 3 by 1.2 m deep sandbox) investigation of gasoline contamination, TRI (1985) reported that the application of surfactant solution on the soil surface using a rainmaker resulted in a dramatic increase in the elevation of the water column in observation wells. This behavior was ascribed to capillary fringe depression caused by the surfactant. A related effect was a "rapid outflux of gasoline from the observation wells into the soil area which previously contained the capillary zone" (TRI, 1985). The magnitude and duration of water table fluctuations caused by surfactant-induced drainage will depend in part on the surfactant application rate, depth to the water table, the antecedent moisture content in the unsaturated zone, and the height of the capillary fringe. Because the transient nature of surfactant-induced flow perturbations can occur on relatively short time and spatial scales (Smith and Gillham, 1999; Henry and Smith, 2002), frequent water level measurements may be necessary to identify surfactant-induced vadose zone drainage. Further, because of the difficulty in distinguishing water table increases caused by surfactant-induced drainage from those that are simply due to the addition of fluid to the subsurface, careful mass balances on water and surfactant solution may aid in identifying surfactant effects on unsaturated flow.
Though the experimental results discussed above provide evidence that surfactants can affect unsaturated flow and transport, it is worth noting that there have been studies involving surfactant use in unsaturated porous media that did not report surfactant-induced flow perturbations (e.g., vadose zone drainage). These include the field-scale studies of Abdul et al. (1992) and Bettahar et al. (1999). In the field-scale studies, flow perturbations could easily have been missed due to the difficulties outlined above. Laboratory-scale unsaturated flow studies that did not report any apparent surfactant affects on flow are also found in the literature (e.g., Powelson and Mills, 1998; Lee, 1998; Lee and Fountain, 1999). It is unclear whether surfactant-induced flow perturbations were either very minor or absent in these experiments, or whether the experimental design was such that surfactant effects on flow and transport were either minimized or could not be detected.
A number of difficulties associated with detecting surfactant-induced flow perturbations in field studies have been noted above. There are also several factors that may obscure flow perturbations caused by surfactant-induced capillary pressure gradients in laboratory systems. For example, soil dewatering during 1D miscible displacement experiments can occur when the invading fluid (surfactant solution) has a higher viscosity than the resident fluid (Smith, 1995). Surfactant-induced flow and transport behavior in 1D unsaturated miscible displacement experiments will also depend on boundary condition type (Smith and Gillham, 1994). Under constant flux inlet boundary conditions in 1D systems, the moisture content profiles before surfactant application and following transient surfactant-induced flow perturbations would be expected to be the same if the invading and resident fluids have the same viscosity. However, if the viscosity of the surfactant solution is greater than that of pure water, the moisture content of the soil following the transient flow perturbation will be greater than the initial moisture content to account for the decrease in hydraulic conductivity caused by the more viscous fluid (Smith and Gillham, 1999). Though the initial and final moisture contents may be similar when constant flux inlet conditions are used, the soil water pressure heads within the soil may differ considerably depending on whether the soil is wetted with pure water or surfactant solution (Fig. 1a). Figure 1a also shows that at a given pressure head, moisture content can differ considerably depending on whether the porous medium is wetted with pure water or surfactant solution. Therefore, if constant head boundary conditions are used for surfactant transport experiments, the initial and final pressure heads in the soil will be the same, but the initial and final moisture contents will vary.
|
NUMERICAL MODELS OF SURFACTANT-INDUCED FLOW |
|---|
|
TOPABSTRACTINTRODUCTIONBACKGROUND: EFFECTS ON...EXPERIMENTAL STUDIES OF...NUMERICAL MODELS OF SURFACTANT...SUMMARYREFERENCES |
|---|
Depending on the degree to which the various surfactant-induced flow effects described above are present in a given system, successful simulation of the system may require the incorporation of surfactant concentration-dependent changes to moisture retention (i.e., changes in surface tension and/or contact angle) and hydraulic conductivity. Smith and Gillham (1994) presented a nonhysteretic, 1D unsaturated flow and transport model that accounted for surfactant concentration-dependent surface tension effects. Surfactant effects on moisture retention were incorporated into the model by scaling the pressure term of the van Genuchten (1980) moisture retention function by the relative surface tension, after Eq. [2]:
 | [5] |
, m, and n are soil-specific curve fitting parameters and the other terms are as previously defined. The scaling parameter, o/, was calculated as a function of surfactant concentration (after Adamson, 1990):
 | [6] |
is the surface tension at concentration c, and o is the surface tension at the reference concentration, co (e.g., o = 72 mN m-1 at co = 0.0%). Concentration-dependent viscosity effects on hydraulic conductivity were later added by Smith and Gillham (1999) who used the model to simulate their 7% butanol column experiments described above. Smith and Gillham (1994)(1999) simulated the effects caused by dissolved butanol by solving both the flow and transport equations within the iterative loop during each time step. Discrepancies between simulation and experimental results were largely accounted for by the fact that the model did not incorporate hysteresis. Most unsaturated flow and transport models do not incorporate concentration-dependent surface tension and viscosity effects and would typically be incapable of simulating surfactant-induced flow events. However, Henry et al. (1999) used the 1D, unsaturated flow and transport model HYDRUS 5.0 (Vogel et al., 1996) to simulate surfactant-induced flow caused by the presence of the relatively insoluble surfactant, MA, in closed horizontal column experiments. Because MA is relatively insoluble, it is essentially immobile, and the zone of surfactant-induced surface tension reduction remained confined to the original surfactant source zone in the experiments. Thus, each half of the experimental columns essentially acted as a separate porous medium, with different retention properties. Though HYDRUS 5.0, like most numerical models, does not incorporate concentration-dependent surface tension or viscosity, Henry et al. (1999) simulated surfactant effects in the myristyl alcohol system by assigning separate moisture retention functions to the surfactant-free and surfactant-containing halves of the column. The effects of viscosity and hysteresis were not included in the transient simulations. This approach was successful for simulating flow in columns containing the insoluble surfactant MA, but was not successful for simulating similar experiments with flow induced by the soluble surfactant butanol.
Henry et al. (2001) modified HYDRUS 5.0 to account for concentration-dependent surface tension and viscosity effects on hysteretic unsaturated hydraulic properties. HYDRUS 5.0, like most flow and transport models, solves first for flow within an iterative loop and then uses the flow solution to calculate solute transport. When solute concentration-dependent hydraulic properties are introduced to such a scheme the hydraulic properties are upstream weighted in that the solute concentrations from the previous time step are used to scale the hydraulic functions. Because the time steps required to solve these coupled models are very small, it was found that the modified HYDRUS 5.0 model was capable of simulating the 1D dissolved butanol experiments of Henry et al. (1999). In addition, the model was used to evaluate the importance of hysteresis and dispersivity on surfactant-induced flow perturbations in unsaturated porous media. As suggested by Smith and Gillham (1999), both hysteresis and dispersivity were found to be important factors for accurately modeling the flow behavior induced by the solute-dependent surface tension changes.
Numerical simulation of the surfactant-induced effects observed in a 2D experimental system was conducted by Henry et al. (2002). They modified HYDRUS-2D (Simunek et al., 1999) to include surfactant concentration-dependent effects on surface tension and viscosity. The modified HYDRUS-2D model was used to simulate the 2D experimental system of Henry and Smith (2002). Though HYDRUS-2D can account for hysteretic moisture retention, Henry and Smith (2002) were unable to obtain successful hysteretic simulations of 2D surfactant-induced flow. Therefore, since the surfactant-induced flow observed in the experiment resulted in drainage of large areas of the vadose zone (see Fig. 3), nonhysteretic simulations were performed using the retention properties of the drainage moisture characteristic curve. Comparison of simulated moisture content and dye tracer contour plots to experimental results indicated that the model captured the major processes of the experimental system. As in the 1D, nonhysteretic simulations of Smith and Gillham (1999), differences between simulated and experimental moisture contents and pressure heads could largely be accounted for by the fact that hysteresis was not included in the 2D simulations.
To our knowledge, the modeling exercises noted above are the only studies in which transient numerical simulations have been compared with experimentally observed surfactant-induced unsaturated flow perturbations. Schneid et al. (2000) presented a variably saturated flow and transport model that includes surfactant effects on surface tension and hydraulic conductivity variations due to surfactant sorption. Simulations for flow in a system with surfactant-induced permeability changes were also given, but the simulation results were not verified by comparison with experimental data. Other models also exist which may be capable of simulating surfactant effects on flow and transport in unsaturated porous media. For example, the theoretical models of Bear and Bensabat (1989) and Nassar and Horton (1999), as well as the numerical model of Nassar et al. (1999), include the effects of solute-induced surface tension changes on unsaturated flow. There are also existing numerical models that can simulate surfactant flushing in multiphase systems (e.g., STOMP [White and Oostrom, 1996, 1998] and UTCHEM UTCHEM [Delshad et al., 1996]). Assuming that surfactant concentration–dependent constitutive relationships can be defined for the system of interest, these types of models should be capable of simulating surfactant-induced unsaturated flow perturbations.
Because models that account for surfactant effects on unsaturated flow are relatively uncommon, unsaturated flow models that do not include surfactant concentration-dependent hydraulic functions have been used for simulating surfactant use in unsaturated systems. Numerical modeling was used by Abdul et al. (1992) to determine start-up conditions for their field-scale remediation study, but it does not appear that surfactant effects on hydraulic properties were included in the model. When the surfactant application rate determined from the numerical modeling was utilized in the field, unexpected ponding occurred. It is unclear whether the use of a model that accounts for surfactant effects on flow would have provided improved results. Lee and Fountain (1999) and Lee (2000) used the multiphase flow and transport code MOTRANS (Environmental Systems and Technologies, Blacksburg, VA), a commercial version of MOFAT (Katyal et al., 1991), to simulate surfactant flushing of tetrachloroethylene (PCE) in homogeneous unsaturated soil columns. Although the simulations did not specifically account for concentration-dependent surfactant effects on flow, the initial surface tension of the surfactant solution and the interfacial tension between the surfactant solution and PCE were used for scaling of the capillary pressure curves during the flushing period. Comparison of simulation and experimental results showed generally good agreement, and differences were attributed to volatilization, a more complex distribution of PCE in the experimental columns than in the simulations, and kinetic effects (Lee and Fountain, 1999). Since this approach accounts for surfactant effects on moisture retention, including capillary fringe depression, it should be adequate for simulating systems in which surfactant is conservative (i.e., surface tension does not vary with concentration) and transient surfactant-induced flow perturbations are of secondary or minimal concern.
|
SUMMARY |
|---|
|
TOPABSTRACTINTRODUCTIONBACKGROUND: EFFECTS ON...EXPERIMENTAL STUDIES OF...NUMERICAL MODELS OF SURFACTANT...SUMMARYREFERENCES |
|---|
There is insufficient data on surfactant effects on flow and transport within the vadose zone to predict the behavior of many systems with an acceptable degree of certainty. The need exists to expand our data and knowledge to a wider range of surfactants, a wider range of surfactant concentrations (both above and below their critical micelle concentrations), and to field studies where geochemical and biogeochemical conditions are expected to play roles. However, there is enough evidence to prove that in some cases surfactants can cause flow and transport in the vadose zone to deviate substantially from constant surface tension systems. Experiments investigating surfactant systems require higher frequency data acquisition and more sample points than corresponding constant surface tension systems. However, standard vadose zone methods, such as time domain reflectometry and pressure transducer equipped tensiometers, have proven sufficient to capture the induced flow behaviors. The simulations performed with numerical models that include solute dependent effects on capillary flow require finer spatial grids and smaller time steps, and tend to take more iterations to converge. These models have proven capable of capturing the primary flow effects induced by surfactants. Surfactant effects on vadose zone flow may be much more prevalent than currently generally assumed in studies of the fate of organic compounds in the vadose zone. Consideration of these effects in our conceptual models, experimental design, and data analyses for solving problems of contamination by organic compounds in the vadose zone is warranted.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONBACKGROUND: EFFECTS ON...EXPERIMENTAL STUDIES OF...NUMERICAL MODELS OF SURFACTANT...SUMMARYREFERENCES |
|---|
