Environ. Sci. Technol. 2002, 36, 2082-2087
Density-Modified Displacement for Dense Nonaqueous-Phase Liquid Source-Zone Remediation: Density Conversion Using a Partitioning Alcohol C. ANDREW RAMSBURG AND KURT D. PENNELL* School of Civil and Environmental Engineering, Georgia Institute of Technology, 200 Bobby Dodd Way, Atlanta, Georgia 30332-0512
Entrapped and pooled dense nonaqueous-phase liquids (DNAPLs) often persist in aquifers and serve as a longterm source of groundwater contamination. To address the problematic nature of DNAPL remediation, a surfactantenhanced aquifer remediation (SEAR) technology, densitymodified displacement (DMD), has been developed which significantly reduces the risk of downward migration of displaced DNAPLs. The DMD method is designed to accomplish DNAPL density conversion through the introduction of a partitioning alcohol, n-butanol (BuOH), in a predisplacement flood using conventional horizontal flushing schemes. Subsequent displacement and recovery of the resulting LNAPL is achieved by flushing with a lowinterfacial tension surfactant solution. The objective of this study was to investigate density conversion of two representative DNAPLs, chlorobenzene (CB) and trichloroethene (TCE). A series of batch experiments was performed to assess changes in NAPL composition, density, and phase behavior as a function of BuOH mole fraction. Experimental results were used to develop contaminant/BuOH/water ternary phase diagrams and to elucidate regions of contrasting NAPL density. UNIQUAC calculations are presented to support measured compositional and phase behavior data. Density conversion of CB and TCE, relative to water, occurred at NAPL BuOH mole fractions of 0.38 and 0.50, respectively. Significant incorporation of water into the organic phase was observed at relatively high BuOH mole fractions and was shown to limit changes in NAPL composition and density. Interfacial tensions between CBNAPL and TCE-NAPL and a 6% (by wt) BuOH aqueous solution were found to decrease with increasing NAPL BuOH mole fraction, although in both cases the measured values remained above 2.5 dyn/cm. Total trapping number calculations suggest that, in most aquifer formations, density conversion can be achieved without premature NAPL displacement using a 6% (by wt) BuOH aqueous solution.
Introduction Surfactant-enhanced aquifer remediation (SEAR) describes in situ flushing processes and technologies that incorporate * Corresponding author phone: (404) 894-9365; fax: (404) 8948266; e-mail:
[email protected]. 2082
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 9, 2002
the use of surfactants to overcome many of the limitations associated with conventional pump-and-treat remediation of contaminant source zones. SEAR technologies rely upon two primary mechanisms, micellar solubilization and immiscible displacement, to recover nonaqueous-phase liquid (NAPL) contaminants from the subsurface. Solubilization usually involves the use of micellar surfactant solutions to increase the apparent aqueous solubility of the contaminant in a single-phase miscible displacement flood. While this approach has been shown to be effective in numerous studies (1-7) and may significantly reduce remediation times compared to pump-and-treat, a substantial volume of surfactant solution may still be required to achieve remediation goals (4-7). In addition, rate-limited solubilization may further reduce recovery efficiencies even under favorable site conditions. These two factors may render micellar solubilization prohibitively expensive for large source zones and highly contaminated sites. The second SEAR recovery mechanism involves the immiscible (two phase) displacement of NAPL as free product. The concept of displacing entrapped NAPL through low interfacial tension (IFT) surfactant flooding originated in the field of enhanced oil recovery. Immiscible displacement is often referred to as mobilization, as discrete NAPL ganglia are envisioned to be mobilized from interstitial pore space as a result of a reduction in capillary forces. Mobilization and solubilization are not mutually exclusive processes, and in general, reductions in NAPL/aqueous-phase IFT correspond to increased solubilization capacity. Such behavior has been observed in many solubilization studies where mobilization was not intended (e.g., refs 8 and 9). Immiscible displacement technologies offer the advantage of reduced flushing volumes which may dramatically reduce remediation time and cost. However, the risk of downward migration of dense nonaqueous-phase liquid (DNAPL) free product into uncontaminated regions of aquifers has been the primary limitation for the implementation of immiscible displacement technologies for DNAPL remediation. Partitioning alcohols can reduce, or reverse, the density difference between the organic and aqueous phases, thereby minimizing or eliminating the risk of downward NAPL migration. The use of partitioning alcohols, in combination with low interfacial tension displacement, offers great potential for improving DNAPL displacement technologies. The density-modified displacement (DMD) process relies on partitioning a sufficient quantity of alcohol into a DNAPL to lower the density and effectively convert the DNAPL to a light nonaqueous phase liquid (LNAPL) (10). Following density conversion, NAPL displacement and recovery is achieved by flushing with a low-interfacial tension surfactant solution. The DMD method is designed to accomplish DNAPL density conversion through the introduction of a partitioning alcohol (e.g., n-butanol (BuOH)) in a predisplacement flood (preflood) using conventional horizontal flushing schemes (i.e., flushing with the groundwater gradient). If successfully implemented, the DMD method provides for efficient recovery of displaced DNAPLs, while mitigating the risks associated with downward DNAPL migration. The partitioning of several alcohols at low concentration was shown to be linear in the development of partitioning interwell tracer tests (PITT) (11, 12). However, at higher concentrations, the partitioning of alcohols between aqueous solutions and several chlorinated solvents has been shown to deviate from ideality, producing phase partitioning curves that are nonlinear (13, 14). The deviations observed in these studies indicate that, at equilibrium, more alcohol would be 10.1021/es011357l CCC: $22.00
2002 American Chemical Society Published on Web 03/30/2002
TABLE 1. Relevant Properties of n-Butanol (BuOH), Chlorobenzene (CB), and Trichloroethene (TCE) at 25 °Ca compound
BuOH
CB
TCE
molecular weight (g/mol) solubility in water (g/L) water solubility in organic (wt %) liquid density (g/mL) dynamic viscosity (Pa‚s × 103)
74.12 74.5 20.5 0.806 2.27*
112.56 0.488* 0.033 1.10 0.715*
131.39 1.100 0.320 1.47 0.532
a
All data from Riddick and Bunger (22); (*) 30 °C.
found in the aqueous phase than would be expected with ideal behavior. Despite such nonlinear partitioning behavior, the potential exists to carefully develop surfactant-alcohol flushing solutions for in situ density conversion (14). Numerous phase behavior studies of systems containing partitioning alcohol have validated the concept of density reduction (15-20). These studies primarily used ternary phase diagrams in support of column studies. While many of these studies determined the composition of the organic and aqueous phases, none provided a rigorous thermodynamic treatment of these data. Experimental results indicated substantial alteration of NAPL properties with increased alcohol partitioning, including changes in NAPL density, viscosity, and interfacial tension. In previous studies, however, little, if any, attention has been given to the partitioning of water into the NAPL and the resulting impact of such behavior on relevant physical and chemical properties. The purpose of this study was to investigate the effects of BuOH partitioning on the phase behavior, water incorporation, density transformation, IFT, and viscosity of DNAPLs. Two chlorinated solvents of considerably different densities, chlorobenezene (CB) and trichloroethene (TCE), were used as representative organic liquids to assess the DMD method. The selection of BuOH for the partitioning alcohol represents a balance between the ability to deliver the partitioning alcohol in an aqueous solution and the tendency for the alcohol to partition into the NAPL. In addition, the selection of BuOH was based on previous studies (21), in which higher molecular weight alcohols formed gels in the presence of various surfactants. Alcohols lighter than C4 are generally miscible with water at relevant temperatures and therefore do not offer the partitioning necessary for DNAPL density transformation. Further, the straight carbon chain of BuOH provides greater biodegradability than many branched-chain partitioning alcohol alternatives. Should BuOH remain in the aquifer, it may even serve to stimulate postflushing biotransformations of residual contaminants.
Materials and Methods Materials. HPLC-grade CB and TCE were obtained from Aldrich Chemical and dyed with Oil-Red-O (Fisher Scientific), an organic-soluble dye, at a concentration of 4 × 10-4 M for visualization purposes. Previous work has shown no substantial effects on the physical properties (e.g., interfacial tension, density, and viscosity) of these organic liquids when dyed with Oil-Red-O at low concentrations. HPLC-grade BuOH and calcium chloride dihydrate were obtained from Fisher Scientific. All aqueous solutions were prepared with distilled water that was purified using a Nanopure analytical deionization (Barnstead/Thermolyne Corporation) system. Relevant properties (at 25 °C) of CB, TCE, and BuOH are given in Table 1. Phase Behavior Studies. A matrix of batch experiments was conducted to examine the equilibrium distribution of BuOH in the CB/BuOH/water and TCE/BuOH/water systems. Various fractions of CB or TCE, BuOH, and water were added to 10-mL graduated (0.1 mL) conical bottom centrifuge tubes. Samples were gently mixed on Lab Quake shaker trays for
a minimum of 72 h. To ensure that the samples reached equilibrium, phase volumes and concentrations were measured daily until no change was detected between successive observations. After equilibration, the phases were separated by centrifugation at 1100 rpm for 5 min using a Beckman Avanti J-25 centrifuge. Centrifugation was required to ensure the coalescence of all dispersed droplets in each phase. Initial and final volumes as well as relative aqueous and NAPL phase densities were recorded for each sample. To quantify equilibrium alcohol partitioning behavior, aqueous and organic liquid-phase BuOH and contaminant (CB and TCE) concentrations were measured. Analysis of CB, TCE, and BuOH was performed in triplicate, using a Hewlett-Packard model 6890 gas chromatograph (GC) equipped with a flame ionization detector (FID). The separation of BuOH and chlorinated organics was accomplished isothermally (45 °C) on a DB-5 column (J&W Scientific; 0.32 mm i.d., 30 m length). Both aqueous and organic samples were prepared for analysis through dilution (∼5x aqueous phase and ∼40x organic phase) with 2-propanol. A six-point calibration curve was obtained prior to each sequence run and was checked with calibration samples every 10 samples for quality assurance. The water content of the organic liquid phase was determined by Karl Fisher (KF) titration. An automated KF titrator (Mettler DL18) was used to measure the water content of each sample in triplicate. The Mettler DL18 requires a single-point calibration of a known mass of water (∼0.05 g) prior to use each day. The calibration was subsequently verified using Hydranal 10 (Riedel-de Haen) standards obtained from Fisher Scientific. Data obtained from the alcohol partitioning experiments allowed for specific NAPL compositions to be prepared by mixing various fractions of either CB or TCE with water and BuOH. Each target NAPL composition was generated in triplicate using 35-mL centrifuge tubes, which allowed for adequate sample volume for composition, density, viscosity, and interfacial tension analyses. Samples were equilibrated and analyzed using the methods described previously. Density, Viscosity, and Interfacial Tension. Density and viscosity measurements were performed to examine the effects of BuOH partitioning on the physical properties of both the aqueous and organic liquid phases. The density of each phase was measured using 2-mL glass pycnometers. The pycnometers were calibrated before each use with Nanopure water. Viscosities of CB-NAPL and TCE-NAPL as a function of BuOH mole fraction were measured using a Haake Rheostress 75 rheometer equipped with a doublegap sensor. Flow curves (shear rate vs shear stress) for each sample were generated at 20 °C using shear rates from 200 to 1000 s-1 to ensure that each sample exhibited Newtonian behavior. Interfacial tension measurements were obtained using the drop volume method of Hool and Schuchardt (23). Initially, NAPL-water interfacial tensions were measured for each pure component NAPL (CB and TCE). The interfacial tension between the organic liquids (CB-NAPL and TCENAPL) and an aqueous BuOH solution was determined as a function of NAPL BuOH mole fraction. Organic liquid samples were equilibrated prior to interfacial tension analysis to obtain the desired BuOH and water compositions. UNIQUAC Calculations. Liquid activity coefficients are related to the molar excess Gibbs energy (gE) through
gE ) RT
∑x ln γ i
(1)
i
i
where R is the gas constant, T is temperature, xi is the mole fraction of component i, and γi is the activity coefficient of component i (24). This allows for the calculation of activity VOL. 36, NO. 9, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2083
FIGURE 1. Ternary phase diagram for the CB/BuOH/water system at 22 ( 2 °C. coefficients, provided a suitable expression for gE is available. One commonly used gE expression is universal quasichemical theory (UNIQUAC) developed by Abrams and Prausnitz (25) and later modified by several others (26-28). The basic UNIQUAC equation consists of both combinatorial and residual parts
( )
gE gE ) RT RT
combinatorial
+
( ) gE RT
residual
(2)
where the combinatorial part describes the entropic contribution and the residual part describes the intermolecular forces leading to the enthalpy change upon mixing (24). For ternary systems, the equations for each part and subsequent equations for activity coefficients become sufficiently complex that the interested reader is referred to refs 24 and 29. With a dependence of calculated activity coefficients on composition, the solution to the liquid-liquid equilibrium problem becomes computationally significant. Thus, the simulation software HYSIS (Hyprotech) was used to perform the necessary isothermal flash calculations. The UNIQUAC model uses binary (pair) interaction parameters that are determined from equilibrium data. For liquid-liquid systems, these interaction parameters are determined from mutual solubility data for partially miscible pairs and from vapor-liquid data for miscible pairs. UNIQUAC was selected for this work over other gE models (e.g., NRTL (30)) based on its ability to successfully calculate the ternary systems of interest. Several studies have employed parameter estimation based on group contribution methods, such as UNIFAC (31), but the presence of existing data allowed for the regression of UNIQUAC binary pair interaction parameters and, thus, did not require UNIFAC estimations (e.g., refs 11 and 31-35).
Results and Discussion Phase Behavior. The results of the CB and TCE phase behavior studies are best described using ternary phase diagrams (Figures 1 and 2, respectively). These diagrams indicate that both the CB/BuOH/water and TCE/BuOH/water systems exhibit type II phase behavior (36) resulting from two partially miscible pairs (CB or TCE/water and BuOH/ 2084
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 9, 2002
FIGURE 2. Ternary phase diagram for the TCE/BuOH/water system at 22 ( 2 °C. water) and one miscible pair (CB or TCE/BuOH) in each system. It is important to note that both the CB and TCE diagrams contain two single-phase regions, although the single aqueous-phase region (lower right corner) is very small and this boundary is not distinguishable. Phase behavior data previously reported by others (37, 38) were found to be in close agreement with those collected in this study (Figures 1 and 2). Tie lines (lines of constant composition within the two-phase envelope) slope downward toward the water apex (lower right) for both CB and TCE, indicating a substantial partitioning of BuOH into the organic liquid phase. Although gE models typically have difficulty predicting ternary liquidliquid equilibrium, calculated UNIQUAC phase behavior (boundaries and tie line slopes) were found to be in relatively close agreement with experimental data (Figures 1 and 2). The UNIQUAC-calculated tie lines shown in both phase diagrams provide an indication of the predicted tie line slope in the proximity of each experimental tie line. The agreement demonstrated between measured and calculated data arises, in part, from the observed Type II phase behavior, which allowed the calculated phase boundaries to be anchored on two axes in the absence of a plait point (29). The tie lines represent lines of constant composition within the two-phase envelope. Any mixture composition residing on a tie line will separate into two phases with compositions described by the end points of the tie line. A significant application of this concept, in regard to DMD, is that there exists one tie line that represents DNAPL-to-LNAPL density conversion. Regions in which equilibrated sample mixtures contained two phases are labeled as DNAPL and LNAPL and are separated by a tie line corresponding to density conversion (indicated with a double line in Figures 1 and 2). The location of the DNAPL-LNAPL boundary is a function of contaminant density and extent of BuOH and water partitioning. While the exact location of this boundary may be altered by changes in aqueous-phase density (e.g., the addition of salts (39, 40)), the intent here was to assess the density conversion resulting from BuOH partitioning. The location of the density transition boundary and, consequently, the area of the LNAPL region are governed largely by the initial (pure component) NAPL density. For CB, the DNAPL-LNAPL boundary coincided with smaller mole
FIGURE 3. Aqueous-phase BuOH concentrations in equilibrium with ternary (CB/BuOH/water) NAPL mixtures at 22 ( 2 °C.
FIGURE 4. Aqueous-phase BuOH concentrations in equilibrium with ternary (TCE/BuOH/water) NAPL mixtures at 22 ( 2 °C. fractions of BuOH, resulting in a relatively larger LNAPL region (Figure 1). In contrast, the DNAPL-LNAPL boundary for TCE occurred at a higher mole fraction of BuOH, resulting in a much smaller LNAPL region (Figure 2). These results illustrate the increased difficulty associated with density conversion of relatively dense NAPLs. In practice, relatively larger amounts of BuOH will be required to modify the density of DNAPLs, such as TCE or PCE. Preflood solutions for these relatively heavy DNAPLs must contain high concentrations of BuOH (i.e., near the solubility limit), and several pore volumes may be required to achieve in situ density conversion. Alcohol Partitioning. The equilibrium partitioning of BuOH in the ternary system consisting of CB/BuOH/water and TCE/BuOH/water was investigated over the entire range of organic-phase BuOH mole fractions. Experimental and calculated (UNIQUAC) aqueous-phase BuOH concentrations (g/L) are plotted against organic liquid BuOH mole fractions (XBuOH) for both CB-NAPL and TCE-NAPL (Figures 3 and 4, respectively). Partitioning of BuOH in these systems was observed to be nonlinear (nonideal), which is consistent with the findings reported in previous studies (13, 14). Strong nonideal behavior results from the greater BuOH selfassociations (like pair interactions) in the organic phase relative to hydrogen bonding (unlike pair interactions) occurring in the aqueous phase (41). While the incorporation of water into the organic phase increases the unlike pair interactions, the relative effect remains unchanged. In binary systems, this type of behavior is often referred to as a positive deviation from Raoult’s law and likely contributes to the
relatively poor agreement between the UNIQUAC predictions and experimental data shown in Figure 4 (24). UNIQUAC calculations were likely affected by the interaction parameters for the miscible contaminant/BuOH pairs being regressed from binary vapor-liquid equilibrium (VLE). The effects of BuOH self-associations and parameter regression from VLE only become apparent when the data are presented as alcohol partitioning curves, as the ternary diagrams are predicted well by the UNIQUAC model. In several studies, a Raoult’s law convention has been employed to describe liquid-liquid equilibrium (e.g., refs 33 and 42-44). This type of approach is represented by the straight lines in Figures 3 and 4 and is based on the assumption of an ideal organic phase. The assumption of organic-phase ideality breaks down for the systems described herein because BuOH may self-associate or even associate with water in the organic phase, resulting in activity coefficients larger than unity. It should be noted that Raoult’s law conventions, which may be useful as first-order approximations in nonassociating systems, rely upon simplifications that are not valid for partitioning in associated systems. The sigmoidal trend of the BuOH partitioning data and lower range (
