Nonionic Surfactant-Enhanced Solubilization and Recovery of Organic

Armstrong Laboratory, 139 Barnes Drive, Suite 2, Tyndall Air Force Base, ... Column effluent breakthrough data are used to evaluate the modeling appro...
0 downloads 0 Views 170KB Size
Download PDF
Environ. Sci. Technol. 1997, 31, 1284-1289

Nonionic Surfactant-Enhanced Solubilization and Recovery of Organic Contaminants from within Cationic Surfactant-Enhanced Sorbent Zones. 2. Numerical Simulations JOEL S. HAYWORTH* Applied Research Associates, P.O. Box 40128, Building 1142, Tyndall Air Force Base, Florida 32403 DAVID R. BURRIS Armstrong Laboratory, 139 Barnes Drive, Suite 2, Tyndall Air Force Base, Florida 32403

A mathematical model is developed to investigate the simultaneous aqueous phase transport and partitioning behavior of a nonionic surfactant and a representative hydrophobic organic contaminant (HOC) in flow-through aquifer material-water systems. Unmodified aquifer material and aquifer material treated with a cationic surfactant are considered. Nonionic surfactant sorption is represented using the equilibrium, nonlinear two-term Langmuir equation and the kinetic, nonlinear Langmuir equation. HOC sorption and solubilization is represented by an expression relating HOC partitioning between the bulk solid phase and the bulk aqueous phase containing monomer and micellar pseudophases. The model is implemented in a onedimensional finite difference numerical model that utilizes Picard iteration to accommodate nonlinearities. Column effluent breakthrough data are used to evaluate the modeling approach. Experimentally determined batch data provided most of the model input parameters. Model simulations show good agreement with measured results when mass transfer limitations for the nonionic surfactant are considered. The model is employed to examine the potential effects of influent nonionic surfactant concentration and flushing rate on the removal of HOCs from within a cationic surfactant-enhanced sorbent zone. The analysis revealed that increasing nonionic surfactant influent concentrations decreased the volume of nonionic surfactant required to recover an HOC pulse and that HOC removal increased with increasing nonionic surfactant flushing rate. It is likely, however, that a maximum flow rate exists above which mass transfer limitations in HOC aqueous-solid phase partitioning will occur.

Introduction In a companion paper, Hayworth and Burris (1) proposed a coupled enhanced sorption/enhanced solubilization groundwater remediation process using cationic and nonionic surfactants to accumulate and remove hydrophobic organic contaminants (HOCs) from contaminated aquifers. Labora* Corresponding author fax: 904-286-6979; e-mail: joel [email protected].

1284

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 31, NO. 5, 1997

tory experiments demonstrating the potential of the coupled process were conducted. A sketch illustrating this process is shown in Figure 1. To apply this remediation strategy, an aqueous cationic surfactant solution is injected into an aquifer to create a permeable enhanced sorption zone for HOCs in situ. The cationic surfactant increases the organic carbon content of the aquifer material primarily through cation exchange of the cationic surfactant monomers with the solid phase (1-5). HOCs migrating through this zone are retarded significantly with respect to their transport through aquifer material not treated with cationic surfactants, thus accumulating within the enhanced sorption zone. By later flushing the zone with a nonionic surfactant solution, sorbed HOCs can be solubilized and pumped to the surface for treatment (1). Solubilization occurs when HOCs preferentially partition to the mobile nonionic surfactant micellar pseudophase (1, 6, 7). Simultaneous transport and partitioning behavior of a nonionic surfactant and an HOC in flow-through aquifer material-water systems are examined in this paper using numerical modeling. Batch and column experimental results presented in the companion paper (1) are utilized in this study. Igepal CO 730 (CO 730) was used as a representative nonionic surfactant. The representative HOC used was 1,2,4trichlorobenzene (TCB). Aquifer materials from Columbus Air Force Base, MS, with and without modification with the cationic surfactant hexadecyltrimethylammonium chloride (HDTMA) were used. The primary objective of the study was to adequately simulate the experimentally observed onedimensional transport of the HOC and surfactants using experimentally determined partitioning and transport parameters. Flow-through system mass transfer limitations, particularly with respect to nonionic surfactant transport, were examined. A sensitivity analysis of the model to the nonionic surfactant flushing rate and influent concentration was performed. Numerical simulation provides a valuable predictive capability in examining aspects of the proposed enhanced sorption/enhanced solubilization HOC remediation scheme.

Model Development With reference to Figure 2, TCB partitioning between the bulk aqueous phase and the solid phase can be described by (1)

CH,s ) K*CH,a K* ) Kd

K* ) Kd

(1 + kpm,NSCNS,s)

(1) CNS,a e cmc

(1 + KMNCNS,a)

(2a)

(1 + kpm,NSCNS,s) [1 + KMNcmc + KMC(CNS,a - cmc)] CNS,a > cmc (2b)

where CH,a [MH La-3] and CH,s [MH Mpm-1] are the aqueous and sorbed phase HOC concentrations, respectively; K* [La3 Mpm-1] is the apparent TCB partitioning coefficient; Kd [La3 Mpm-1] is the cationic surfactant-modified soil-water distribution coefficient for TCB in nonionic surfactant-free water; CNS,a [MNS La-3] and CNS,s [MNS Mpm-1] are the aqueous and sorbed CO 730 concentrations, respectively; kpm,NS [Mpm MNS-1] is the partitioning coefficient for TCB between the sorbed CO 730 and the bulk solid phase; KMN [La3 MMN-1] is the TCB partitioning coefficient between the CO 730 monomer psuedophase and the bulk aqueous phase; KMC [La3 MMC-1] is the TCB partitioning coefficient between the CO 730 micellar

S0013-936X(96)00323-9 CCC: $14.00

 1997 American Chemical Society

TABLE 1. Experimentally Determined Model Parameters parameter

HDTMA-treated aquifer material

value

Kd Kd KMN KMC cmc b1 b1 b2 b2 k1 k1 k2 k2

yes no yes yes yes yes no yes no yes no yes no

43.0 mL/g 1.49 mL/g 1.04 L/g 4.4 L/g 0.6 g/L 9.71mg/g 3.61 mg/g 8.03 mg/g 13.85 mg/g 9.51 mL/mg 3.42 mL/mg 0.15 mL/mg 0.03 mL/mg

indirectly using Levenberg-Marquardt (LM) nonlinear parameter estimation (8; see Table 1). The presence of a small aqueous concentration of HDTMA, resulting from treatment of the aquifer material to simulate a cationic surfactantenhanced sorbent zone, causes non-ideal mixed micelle formation with CO 730 (1). Thus, systems incorporating a small aqueous HDTMA concentration were used to determine KMN, KMC, and cmc. To use eqs 1 and 2 to model TCB partitioning in the presence of the nonionic surfactant CO 730, a relationship describing CO 730 partitioning between the bulk aqueous phase and the solid phase is required. Batch sorption experiments performed in ref 1 showed that this partitioning could be approximated by the equilibrium two-term Langmuir equation (9): FIGURE 1. Sketch illustrating the enhanced sorption/enhanced solubilization remediation scheme: (A) cationic surfactant solution is injected into an aquifer to create an in situ enhanced sorbent zone; (B) HOCs migrating in groundwater are accumulated within sorbent zone; (C) nonionic surfactant is flushed through the enhanced sorbent zone to solubilize and remove HOCs.

CNS,s )

b1k1CNS,a b2k2CNS,a + 1 + k1CNS,a 1 + k2CNS,a

(3)

where b1, k1, b2, and k2 are fitting parameters. Values for these parameters are also given in Table 1. If equilibrium partitioning is valid, then eq 3 allows eq 2 to be rewritten in terms of CNS,a alone. However, there is evidence that in a dynamic groundwater system this partitioning is mass transfer limited (1, 10). Thus, a kinetic form of the Langmuir equation was also employed in our modeling study:

∂CNS,s ) kfCNS,a(b - CNS,s) - kbCNS,s ∂t

(4)

where kf [MNS-1 La3 T-1] and kb [T-1] are fitting parameters describing the forward and reverse sorption rates, respectively, and b [MNS Mpm-1] is a constant related to the maximum sorptive capacity of the aquifer material (11-13). As eq 2 suggests, TCB and CO 730 transport is a coupled process that can be represented in one-dimension by

FIGURE 2. Conceptual representation of HOC partitioning in a soilwater system containing micelle-forming cationic and nonionic surfactants. psuedophase and the bulk aqueous phase; and cmc [MNS La-3] is the bulk aqueous phase critical micelle concentration. The subscripts MN and MC represent surfactant monomers and micelles, respectively; the subscripts NS and CS represent the nonionic and cationic surfactants, respectively. The subscript H represents the HOC. The subscripts a and s represent the bulk aqueous and sorbed phases, respectively; while the subscript pm represents the bulk (saturated) porous medium. In eq 2, K* is a function of CNS,a and CNS,s. Batch and column experiments described in ref 1 allowed for direct measurement of the parameters in eq 2 save kpm,NS, which was determined

∂CH,a Fb ∂CH,s ∂2CH,a ∂CH,a + ) -v + DH ∂t θ ∂t ∂x ∂x2

(5a)

∂CNS,a Fb ∂CNS,s ∂2CNS,a ∂CNS,a + ) -v + DNS ∂t θ ∂t ∂x ∂x2

(5b)

where x [Lpm] and t [T] are spatial and temporal variables, Fb [Mpm Lpm-3] and θ [La3 Lpm-3] are the dry bulk density and saturated porosity of the aquifer material, respectively; v [Lpm T-1] is the average pore water velocity; and DH [Lpm2 T-1] and DNS [Lpm2 T-1] are the hydrodynamic dispersion coefficients for TCB and CO 730, respectively. If it is assumed that diffusion of TCB and CO 730 during transport is negligible, then both DH and DNS can be replaced by RLv, where RL [Lpm] is the longitudinal dispersivity of the aquifer material. The coupling between eq 5a and eq 5b is apparent when eq 1 is differentiated with respect to time:

VOL. 31, NO. 5, 1997 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

1285

∂CH,s ∂CH,a ∂K* ) K* + CH,a ∂t ∂t ∂t

(6)

Thus eq 5a can be rewritten as

Results and Discussion 2

∂ CH,a ∂CH,a FbCH,a ∂K* ∂CH,a + ) -v + R Lv RH ∂t θ ∂t ∂x ∂x2

(7)

where

RH ) RH(CNS,a′CNS,s) ) 1 +

Fb K* θ

(8)

When equilibrium partitioning of CO 730 is assumed, eq 5b can be rewritten as

∂CNS,a ∂2CNS,a ∂CNS,a RNS ) -v + R Lv ∂t ∂x ∂x2

(9)

where RNS is the nonlinear, equilibrium two-term Langmuir retardation coefficient:

RNS ) RNS(CNS,a) ) 1 +

[

]

Fb k2b2 k1b1 + θ (1 + k C )2 (1 + k C )2 1 NS,a 2 NS,a (10)

For kinetic partitioning of CO 730, eq 5b becomes

∂CNS,a ∂2CNS,a ∂CNS,a ) -v + R Lv ∂t ∂x ∂x2 Fb [k C (b - CNS,s) - kbCNS,s] (11) θ f NS,a

Numerical Method Simulations investigating the transport behavior of CO 730 (eqs 9-11), TCB (eqs 7 and 8, with K* equal to Kd and the partial time derivative of K* equal to zero), and coupled CO 730/TCB (eqs 2 and 7-11) in one-dimensional columns of HDTMA-treated and untreated Columbus, MS, aquifer material were performed using the experimental data presented in ref 1. Conditions for these simulations are summarized in Table 2. In all cases, the finite-difference, Crank-Nicholson numerical technique was used. Picard iteration (14) was used when simulating nonlinear systems (CO 730 and coupled CO 730/TCB transport; TCB transport equations were solved simultaneously). All simulations utilized continuous mass flux inflow boundary conditions and continuous aqueous concentration outflow boundary conditions. In all cases, initial concentrations throughout the simulation domain (columns) were equal to zero. Numerical dispersion and oscillations were controlled by discretizing the problem so that the grid Peclet and Courant number criteria were satisfied (14). Numerical accuracy of simulation output was evaluated based on comparison with analytical solutions (for the equilibrium partitioning models) and on mass balance considerations (for both equilibrium and kinetic partitioning models). Mass balance calculations for both TCB and CO 730 were made by equating the total mass entering and leaving the column to the total mass stored within the column at a given time. The equilibrium partitioning model simulations compared well with analytical solutions; in all simulations the relative mass balance error was below 3%. In Table 2, simulations A, B, and E correspond to column transport experiments described in ref 1; simulations C and D are numerical experiments performed to illustrate the expected behavior of TCB in natural aquifer material and

1286

9

within a cationic surfactant-enhanced sorbent zone (C and D, respectively), without nonionic surfactant-enhanced solubilization.

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 31, NO. 5, 1997

CO 730 Transport. Effluent breakthrough curves for simulations A (untreated porous media) and B (HDTMA-treated porous media), utilizing equilibrium two-term Langmuir (eq 3) and kinetic Langmuir (eq 4) partitioning relationships, are shown in Figure 3, panels a and b, respectively. Measured effluent data from ref 1 are also shown in these figures. In these simulations, a pulse of 50 mg/mL CO 730 solution (CNS,o) was input into the column for 5 h at a volumetric flow rate (QL) of 0.25 mL/min, followed by continuous flushing with CO 730-free water for 20 h at the same value of QL. For the curve representing the kinetic model simulation, the parameters kf, kb, and b were obtained by LM parameter estimation. Values for these parameters are given in Table 3. The curves are plotted as a function of total column pore volumes (vt/L). In Figure 3a,b, the simulated breakthrough curves obtained using the equilibrium two-term Langmuir partitioning relationship do not adequately describe the measured data, suggesting that during these experiments local equilibrium was not attained. Previous researchers (15) have noted that failure to attain local equilibrium can depend on a number of factors (e.g., pore water velocity, dispersion coefficients, boundary conditions, form of partitioning relationship); in this case it is most likely a result of the relatively high pore water velocities used in our experiments. It should be noted, however, that the pore water velocities used in our experiments are within the range of those observed at the Columbus, MS, field site (16). Additionally, significant increases in nearfield pore water velocities would be expected when the surfactant remediation method is applied at a field site. Thus, the expectation of kinetic effects during nonionic surfactantenhanced solubilization is reasonable. It is clear from Figure 3a,b that the kinetic model provides a reasonable fit to the measured data, whereas the equilibrium model does not; this is also illustrated by comparison of the R 2 values between the measured and simulated curves given in Figure 3. However, because the kinetic simulations required indirect estimation of the parameter b and the mass transfer terms kf and kb, it is difficult to assess the overall validity of our choice for a kinetic partitioning relationship. Nevertheless, Figure 3a,b strongly indicates that kinetic effects dominated the transport process and, as employed eq 4, adequately represent these effects. TCB Transport. Effluent breakthrough curves for simulations C and D (Table 2) are shown in Figure 4, panels a and b, respectively. In these simulations, a pulse of 16.44 mg/L TCB (CH,o) was input into the column for 5 h, followed by continuous flushing with clean water for 20 h (all at QL ) 0.25 mL/min). For simulation C, TCB transport was modeled through untreated porous media [Kd ) 1.49 mL/g; 3]; for simulation D, TCB transport was modeled through HDTMAtreated porous media [Kd ) 43 mL/g; 1]. Both simulations assumed physical porous media characteristics corresponding to measured values in a TCB solubilization experiment discussed in ref 1 (simulation E, this paper). As these figures show, HDTMA-treated porous media is a much more favorable partitioning medium than untreated porous media. In the untreated porous media, the initial TCB breakthrough (Ti) begins after approximately 1.5 pore volumes have been eluted; in the HDTMA-treated porous media, this occurs after approximately 35 pore volumes. Similarly, approximately 19 column pore volumes (pv) are required to elute essentially all of the TCB in the untreated porous media; in the HDTMAtreated porous media, approximately 450 column pore volumes are required. These observations are consistent with expected results based on a simple comparison of the

TABLE 2. Simulation Parameters simulations parameter

A

HDTMA-treated aquifer material column cross-sectional area column length (L) porosity (θ) dry bulk density (Fb) longitudinal dispersivity (RL) source condition CO 730 TCB source concn CO 730 TCB volumetric flow rate vol of source fluid added CO 730 TCB av pore water velocity (v) a

B

C and D

E

no 3.5 cm2 25 cm 0.35 1.65 g/cm3 2.22 cm

yes 3.5 cm2 25 cm 0.40 1.66 g/cm3 2.22 cm

no (C); yes (D) 3.5 cm2 25 cm 0.46 1.52 g/cm3 2.22 cm

yes 3.5 cm2 25 cm 0.46 1.52 g/cm3 2.22 cm

pulse N/Aa

pulse N/A

N/A pulse

continuous pulse

50 g/L N/A 0.25 mL/min

50 g/L N/A 0.25 mL/min

N/A 16.44 mg/L 0.25 mL/min

50 g/L 16.44 mg/L 0.25 mL/min

75 mL N/A 12.4 cm/h

75 mL N/A 10.91 cm/h

N/A 75 mL 9.42 cm/h

6L 75 mL 9.42 cm/h

Not Applicable.

FIGURE 3. Measured (1) and simulated (equilibrium and kinetic partitioning) CO 730 effluent breakthrough using (a) untreated porous media and (b) HDTMA-treated porous media.

TABLE 3. Kinetic Langmuir Parameters Used in CO 730 Transport Simulations soil condition

b (mg/g)

kf (mL/mg h)

kb (1/h)

HDTMA-treated untreated

7.4 9.2

1.8 × 10-1 9.0 × 10-3

1.8 × 10-2 3.6 × 10-2

retardation coefficients (R ) 1 + (Fb/θ)Kd) for TCB in these media: (Rtreated/Runtreated) ) (Ti,treated/Ti,untreated) ) (pvtreated/ pvuntreated) ≈ 24. Coupled CO 730/TCB Transport. Coupled TCB and CO 730 transport representing simulation E (Table 2) is shown in Figure 5. Measured concentration data from an experiment discussed in ref 1 are also shown in this figure for comparison. In these simulations, a pulse of 16.44 mg/L TCB was input into the column for 5 h, followed by flushing with clean water for 65 h, and then flushing with water containing 50 g/L CO 730 for 30 h (all at QL ) 0.25 mL/min). In Figure 5, Co represents the input TCB and CO 730 concentrations for the respective curves. The curve representing CO 730 transport was obtained using the kinetic Langmuir model (eq 11; b, kf, and kb for HDTMA-treated porous media [Table 3]) and

FIGURE 4. Simulated TCB effluent breakthrough using (a) untreated porous media and (b) HDTMA-treated porous media. appears to match the experimental data well. The difference in the plateau values (1.0 for the simulation vs ∼0.95 for the experimental) is a measure of the error in the CO 730 analysis (∼5%; 1). For the TCB transport, two simulation curves are shown: in one, the value for kpm,NS (0.59 g/mg) used was determined by LM parameter estimation using eq 2 and the batch data described in ref 1; in the other, kpm,NS (0.27 g/mg) was determined using LM parameter estimation and the dynamic model (eqs 2 and 7-11) described in this paper. From these curves, it is apparent that the value for kpm,NS determined using the dynamic model provides a closer fit to the experimental data than does the kpm,NS determined using the static (batch) approach, although the dynamic model overestimates the maximum TCB effluent concentration and underestimates the TCB concentration of the tailing portion of the curve. The exact reasons for this are unclear, although it is likely that uncertainty associated with using a large number of experimental and empirical parameters in our model contributes to this effect. The sensitivity of the model to kpm,NS suggests that treating this parameter as a constant may not be appropriate over the entire range of CO 730 concentrations. As noted by other researchers (17, 18), it is reasonable to expect that, at low aqueous CO 730 concentra-

VOL. 31, NO. 5, 1997 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

1287

FIGURE 5. Measured (1) and simulated CO 730 and TCB effluent breakthrough using HDTMA-treated porous media. Simulated CO 730 breakthrough obtained using kinetic Langmuir equation; simulated TCB breakthrough obtained for two different kpm,NS values. tions, sorbed CO 730 (mostly as monomers) would not appreciably increase the TCB sorptive capacity of the solid phase; thus, kpm,NS would be relatively small. As the aqueous CO 730 concentration exceeds the cmc, sorption of both monomers and micelles would likely increase the TCB sorptive capacity of the solid phase, increasing the value of kpm,NS. Another possible explanation for the differences between the measured and simulated data in Figure 5 follows from the representation of TCB solubility in eq 2. TCB partitioning into the monomer and micellar psuedophases is modeled as two distinctly different linear processes above and below the cmc, with the partitioning coefficients (KMN and KMC in eq 2) determined from the slopes of the linear portions of a plot of TCB solubility vs aqueous CO 730 concentration (1). However, it is more likely that TCB solubility enhancement changes gradually near the cmc. In this regard, previous researchers (17) have defined a transitional partitioning coefficient to account for this effect. These researchers have applied this transitional model to batch data with good results, although this approach requires an additional fitting parameter. In the case of our dynamic model, it is questionable whether an additional fitting parameter is justified, considering the previously noted uncertainty in the measured and empirical parameters and the adequate representation of the experimental data by the model. Sensitivity Analysis. Sensitivity of the model to influent nonionic surfactant concentrations is shown in Figure 6. All simulations assumed physical porous media characteristics corresponding to simulation E (Table 2). To initiate a simulation, a 5-h pulse of TCB (16.44 ppm) was introduced at the column inlet at a constant volumetric flow rate (0.25 mL/min). This was followed by a constant input (0.25 mL/ min) of aqueous CO 730 solution for the duration of the simulation. CO 730 concentrations were varied between 25 and 100 g/L. Breakthrough curves for CO 730 at the column outlet are shown in Figure 6a. The shift to earlier breakthrough at higher influent CO 730 concentrations indicates a decrease in the apparent retardation of the surfactant. This is a result of the Langmuir-type rate-limited sorption isotherm: as influent surfactant concentrations increase, the mass fraction of sorbed surfactant decreases, thereby decreasing surfactant retardation. All of the breakthrough curves in Figure 6a are characterized by an abrupt initial rise, followed by a more gentle approach to a maximum value. This reduction in apparent dispersion is also a result of Langmuir-

1288

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 31, NO. 5, 1997

FIGURE 6. Effect of influent CO 730 concentration on (a) CO 730 and (b) TCB effluent breakthrough. type sorption: the leading front of the surfactant solution is more strongly sorbed, resulting in a reduction in dispersion in the initial portion of the breakthrough curves. Similar trends in nonionic surfactant breakthrough curves have been noted by previous researchers (19, 20). In Figure 6b, TCB breakthrough curves corresponding to the respective influent surfactant concentrations are shown. TCB breakthrough occurs at earlier times, and apparent dispersion of TCB decreases with increased influent surfactant concentrations, similar to CO 730 breakthrough behavior. Increased influent CO 730 concentrations result in a more rapid removal of TCB, requiring less flushing volume to achieve a given level of removal. Also, as influent CO 730 concentration increases, the effluent to influent TCB ratio (CH,a/CH,o) exceeds 1. This can be explained by a “concentrate and strip” mechanism: TCB is concentrated within the HDTMA-treated sorbent zone, but is stripped when CO 730 flushing occurs. At higher CO 730 concentrations, there are more micelles available; therefore, TCB partitioning into the micellar pseudophase is more efficient and the concentrating effect is more pronounced. A similar mechanism was proposed by Wagner et al. (21) to explain effluent/influent ratios greater than 1 for aqueous dichlorobenzene eluting from a column packed with aquifer material treated with the cationic surfactant dodecylpyridinium. Simulations examining the model sensitivity to surfactant flushing rates were performed using physical porous media characteristics corresponding to simulation E (Table 2). To initiate a simulation, a 5-h pulse of TCB (16.44 ppm) was introduced at the column inlet at a constant volumetric flow rate (0.25 mL/min). This was followed by a continuous input of aqueous CO 730 solution (50 g/L) at a different (but constant) volumetric flow rate for the duration of the simulation. Flushing rates were varied between 0.05 and 0.25 mL/min, corresponding to average pore water velocities between 1.9 and 9.4 cm/h. Breakthrough curves for CO 730 and TCB at the column outlet are shown in Figure 7, panels a and b, respectively. For both CO 730 and TCB, breakthrough occurs at earlier times for higher flow rates. Also, the apparent CO 730 dispersion decreases as flushing rates increase, indicated by the more rapid rise and roll-off to the maximum value for increasing values of QL (Figure 7a). Both effects are the result of Langmuir-type sorption (eq 11) and an increase in advective transport relative to dispersive transport (eq 5a) at higher flow rates. In Figure 7b, the normalized TCB concentration (CH,a/ CH,o) is plotted as a function of time rather than the

FIGURE 7. Effect of CO 730 flushing rate on (a) CO 730 and (b) TCB effluent breakthrough.

limitations would likely have a detrimental effect on HOC removal: larger volumes of aqueous nonionic surfactant solution would be required to achieve a given level of HOC removal at higher pore water velocities. Thus, it is likely that there is an ideal nonionic surfactant flushing rate that corresponds to the maximum flushing rate under which the equilibrium assumption for HOC/solid phase partitioning is still valid. Flushing rates that exceed this ideal value will not produce added HOC removal efficiency. The limitations on influent aqueous nonionic surfactant concentrations are both physical and practical: at large aqueous concentrations, the unique properties of surfactants (foaming, etc.) make them difficult to work with. Also, increasing surfactant concentration corresponds to an increase in material costs, which may not be beneficial when compared to the increase in HOC removal efficiency. These considerations suggest that optimization of the enhanced sorption/enhanced solubilization remediation scheme at a field site will require, in addition to laboratory batch studies to determine partitioning relationships as described in ref 1, laboratory experimentation coupled with numerical simulation to determine whether HOC/solid phase mass transfer limitations will occur at the anticipated flow rate. The numerical model developed here can be a useful tool or starting point for this assessment.

Literature Cited

FIGURE 8. Simulated TCB effluent breakthrough as a function of CO 730 flushing volume and flushing rate. dimensionless parameter vt/L (which is a function of the CO 730 flushing rate). In Figure 8, CH,a/CH,o is plotted as a function of the dimensionless CO 730 flushing volume, QF/Vp (QF [La3] is the volume of CO 730 solution flushed, Vp [La3] is the column pore volume). As this figure illustrates, the volume of CO 730 required to remove the sorbed TCB is independent of the flushing rate. This is a result of the equilibrium form of the TCB partitioning relationship (eq 1). Based on Figures 6b and 8, it is tempting to conclude that the apparent dependence of TCB removal efficiency on surfactant concentration and flushing rates could be exploited when applying the enhanced sorption/enhanced solubilization remediation scheme in a field situation by flushing high concentration nonionic surfactant solutions at large pore water velocities through the enhanced sorbent zone. However, at abnormally high pore water velocities, mass transfer limitations in HOC partitioning may occur, invalidating the assumption of equilibrium partitioning used here (eqs 1 and 2). In this regard, there is evidence (19) that mass transfer

(1) Hayworth, J. S.; Burris, D. R. Environ. Sci. Technol. 1997, 31, 1277-1283. (2) Boyd, S. A.; Jaynes, W. F.; Ross, B. S. In Organic Substances and Sediments in Water. Volume 1, Humics and Soils; Baker, R. A., Ed.; Lewis Publishers: Chelsea, MI, 1991; pp 181-200. (3) Brown, M. J. Enhancement of Organic Contaminant Retardation by the Modification of Aquifer Material with Cationic Surfactants. M.S. Thesis, University of Waterloo,Waterloo, Ontario, Canada, 1993. (4) Burris, D. R.; Antworth, C. P. J. Contam. Hydrol. 1992, 25, 325337. (5) Wagner, J.; Chen, H.; Brownawell, B. J.; Westall, J. C. Environ. Sci. Technol. 1994, 28, 231-237. (6) Kile, D. E.; Chiou, C. T. Environ. Sci. Technol. 1989, 23, 832-838. (7) Edwards, D. A.; Liu, Z.; Luthy, R. G. J. Environ. Eng. 1994, 120 (1), 23-41. (8) Marquardt, D. W. J. Soc. Ind. Appl. Math. 1963, 11, 431-441. (9) Sposito, G. Soil Sci. Soc. Am. J. 1982, 46, 1147-1152. (10) Adeel, Z.; Luthy, R. G. Environ. Sci. Technol. 1995, 29, 10321042. (11) Hayworth, J. S.; Burris, D. R. Ground Water 1995, 34 (2), 274282. (12) Fetter, C. W. Contaminant Hydrogeology; Macmillan Publishing Co.: New York, 1993. (13) Travis, C. C.; Etnier, E. L. J. Environ. Qual. 1981, 10 (1), 8-17. (14) Huyakorn, P. S.; Pinder, G. F. Computational Methods in Subsurface Flow; Academic Press, Inc.; New York, 1983. (15) Bahr, J. B.; Rubin, J. Water Resour. Res. 1987, 23 (3), 438-452. (16) Boggs, J. M.; Beard, L. M.; Waldrop, W. R.; Stauffer, T. M.; Macintyre, W. G. RP-2485-05. final report; Electric Power Research Institute; Palo Alto, CA, 1993. (17) Sun, S.; Inskeep, W. P.; Boyd, S. A. Environ. Sci. Technol. 1995, 29, 903-913. (18) Gu, T.; Zhu, B.-Y; Rupprecht, H. Prog. Colloid Polym. Sci. 1992, 88, 74-85. (19) Abriola, L. M.; Dekker, T. J.; Pennell, K. D. Environ. Sci. Technol. 1993, 27, 2341-2351. (20) Abdul, A. S.; Gibson, T. L. Environ. Sci. Technol. 1991, 25, 665671. (21) Wagner, J.; Chen, H.; Brownawell, B. J.; Westall, J. C. Environ. Sci. Technol. 1994, 28, 231-237.

Received for review April 9, 1996. Revised manuscript received December 23, 1996. Accepted January 2, 1997.X ES960323O X

Abstract published in Advance ACS Abstracts, March 1, 1997.

VOL. 31, NO. 5, 1997 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

1289