This study utilized a series of laboratory experiments to examine the DNAPL mass removal rate and permeability reduction during ISCO using permanganate (MnO4−). Results show that MnO4− oxidation is effective in removing residual DNAPL from a porous medium. The DNAPL mass removal rate correlated positively with both the hydraulic stress and the oxidant load. A power relationship model of DNAPL mass removal under ISCO was proposed. Results also show that oxidation by-products CO2(g) and Mn oxide can cause pore plugging and flow by-passing. The reduction in hydraulic conductivity due to the Mn oxide precipitates was quantified. Hydraulic conductivity reduction as high as 80% was observed for oxidizing a small quantity of TCE.
This technology has been rushed into field applications to remove source-zone DNAPLs at various sites around the United States [Mott-Smith et al., 2000; McKay et al., 2000]. In some cases, the oxidation was unable to remove DNAPL completely. Concentrations of contaminant rebounded after the treatment was stopped.
 A potential problem is that key reaction products, MnO2 and CO2, can plug pores and divert flow. Schroth et al.  reported permeability reduction with in situ MnO4− oxidation in a column experiment. One of the objectives of this study is to measure directly the permeability reduction due to ISCO using MnO4−. Powers et al.  showed that during TCE dissolution the Sherwood number increased with increasing Reynolds number. Urynowicz and Siegrist  found that the mass transfer rate increased with the oxidation by MnO4−. Another objective of this study is to examine the DNAPL mass transfer rate associated with ISCO using MnO4−.
 Experiments A and B each consisted of five columns, which were individually packed and run as a group at the same time. The setup of 1-D column experiments is summarized in Table 1 and depicted in Figure 1a. Experiments were run with various flushing rates (Experiments A1–A5) and KMnO4 concentrations (Experiments B1–B5), under conditions listed in Table 2. The glass columns were equipped with Teflon end fittings. Medium silica sand (45 micron) (US Silica, Ottawa, IL) filled in the columns. Once the column was packed, it was positioned upside down. TCE was added to the column evenly across the end to create a zone of residual DNAPL saturation. The column was then returned to an upright position. Upward flushing of the columns was accomplished with an Ismatec tubing bed pump (Cole Parmer Instrument Co. Vernon Hills, IL). Effluent samples were collected periodically for chemical analysis. The oxidant flush of each column lasted for several days.
Table 1. Experimental Setups for Column and Flow Tank
Table 2. Conditions for Column Experiments With Various Flushing Rates (A1–A5), and Various KMnO4 Concentrations (B1–B5)
DNAPL (g TCE)
Residence time (hr)
C (KMnO4) (g/L)
Total Volume flushed (L)
Total KMnO4 load (g)
 A falling head permeameter was used to measure the hydraulic conductivity before and after the MnO4− flush, following the method described by Domenico and Schwartz . Each measurement was repeated five times, which usually yielded consistent results. The average of the five measurements was used for comparative purposes.
 An additional 1-D experiment using a larger column (Column C in Table 1) was conducted to quantify the pattern of Mn oxide distribution. DNAPL loading and oxidant flushing procedures remained the same as earlier column experiments. A total of 1 mL TCE was added as residual DNAPL. The experiment lasted for three weeks. After the chemical analysis of effluent indicated that TCE had been removed from the column, the experiment was stopped. Six samples of the porous medium were taken along the column. Samples were treated with thiosulfate to dissolve the Mn oxide. The resulting solution was analyzed by ICP-MS for total Mn concentration.
 Two-dimensional tank experiment (Table 1) was conducted using a thin, small 2-D glass flow tank, outfitted with Teflon end fittings. The flow tank was filled with transparent borosilicate glass beads with a mean grain size of 1 mm. Two PTFE (polytetrafluoroethylene) tubes with an inside diameter of 1.32 mm were installed in the two ends of the tank to function as recharge and discharge wells. Both tubes had ends open at a depth of 0.5 cm from the bottom of the tank. A Spectroflow 400 solvent delivery system (Kratos Analytic, NJ) provided flow into the tank. An Ismatec pump removed fluid at the outlet. With both pumps running at the same steady rate, the ambient flow was horizontal along the length of the tank (Figure 1b). 1 mL of the TCE was added to the tank from the top to form a zone of residual DNAPL saturation across the vertical depth of the tank and a small DNAPL pool at the bottom. The tank was flushed for about two weeks with KMnO4 at 0.2 g/L. Effluent samples were taken three times a day for chemical analysis. Images of the tank were taken with a digital camera to monitor the development of Mn oxide precipitates.
3. Results and Discussion
 The concentration of Cl− in the effluent from the column was an indicator of the breakdown of TCE, as described in equation 1. The complete removal of all the DNAPL from the column would result in a breakthrough curve of Cl− concentration, plotted against time. The total volume flushed and total KMnO4 load in Table 2 indicate chemical flush required to achieve breakthrough, although each experiment was continued for additional period of time to ensure that complete breakthrough was achieved. Figure 2a shows chloride breakthrough curves at effluents for the 1-D columns under ISCO with various flushing rates, ranging from 0.038 to 0.24 mL/min. The results indicate that higher flushing rate produced faster breakthrough, which implies a higher mass transfer rate and DNAPL removal efficiency. Figure 2b. shows the chloride breakthrough curves for the effluents in the 1-D columns flushed with various initial KMnO4 concentrations from 0.05 to 0.4 g/L. Results from this experiment clearly demonstrate that higher oxidant concentration leads to earlier breakthrough.
 In studies of the interfacial mass transfer rate, there has been attempt to establish relationship between the dimensionless Sherwood number, representing mass transfer process, and the Reynolds number, representing hydrodynamic conditions. Powers et al.  proposed that modified Sherwood number (Sh′), defined in equation 2, has an power relationship with Reynolds number (Re), as in equation 3, during studies on the dissolution of NAPL. The mass transfer coefficient thus has a power relationship with flow velocity q,
where k′ is the lumped mass transfer coefficient, d is the medium grain diameter, DL is the free liquid diffusivity, ρ is the density of the aqueous phase, and μ is the dynamic viscosity of the aqueous phase. Other researchers [Miller et al., 1990; Imhoff et al., 1994; Zhang and Schwartz, 2000] have used similar power relationship studying the mass transfer process.
 If we define TB as the time required for Cl− to breakthrough, meaning the time required for DNAPL to be completely removed from the column, then 1/TB can be considered as the average rate for the DNAPL removal. Figure 3a. is a plot of 1/TB versus the flushing rate. It seems that the change in 1/TB with flushing rate follows the established power relationship (R2 = 0.971). Using the same notation, the relationship between the average DNAPL removal rate 1/TB with initial oxidant concentration is illustrated in Figure 3b. A power trendline was fitted the data with R2 = 0.991. This result appears to indicate that the mass transfer coefficient follows a power relationship with the initial oxidant concentration (C) during the chemical dissolution of NAPL. Therefore, it seems that a mass transfer relationship under ISCO using MnO4− can be described in a model presented as
where a is the lumped factor defined by the porous medium, NAPL geometry, and oxidant type, m and n are empirical constants.
 A mass balance for each column experiment was calculated based on initial TCE load and total mass of Cl− removed from column as monitored in the effluent. The mass balances for all columns show at least 95% of the initial TCE was destroyed in every column. Thus, virtually all TCE was removed from the column, within a reasonable range of possible error in experimental results. The results demonstrated how flushing with MnO4− was able to remove the residual DNAPL from the contaminated porous medium.
 The hydraulic conductivity (K) of individually packed columns shows variation before the experiment, which should result from difference in column geometry and the process for packing porous medium. Nevertheless, the reduction in K for each column due to ISCO is obvious (Figure 4). The reduction in K ranges from 10% to 80%. However, there seems no recognizable pattern in change of K as a function of the oxidant load and hydraulic regime. The reduction in K appears to be related to the oxidation process because the hydraulic conductivity of one column, which was flushed with Milli-Q water only (Column B-1 in Figure 4), increased by a small quantity. It is obvious that the reduction in K was caused by the pore plugging of Mn oxide precipitates, which also indicated by the dark brown color formed during the oxidant flush. The change in K from the oxidation of a relatively small quantity of DNAPL in a short period of time is significant. Longer flushing with a larger quantity of TCE should lead to a more dramatic reduction in permeability.
 The additional larger column was flushed with MnO4− for 365 hours. A mass calculation indicated that 96.9% of the initial TCE was removed from the column. The distribution of Mn oxide in the column at the end of the experiment is shown in Figure 5. The majority of the Mn oxide was located at or closely adjacent to the original DNAPL zone.
 The results from the 2-D flow tank experiment (Figure 6) were similar to those from the column experiment, except that the removal efficiency was noticeably lower. After 313 hours of flushing with MnO4−, only 34.9% of the initial TCE was removed from the tank. Half of the TCE was removed in the first 25% of the elapsed time. Figure 6 shows that at the end of the experiment the TCE concentration rebounded, once the injection of MnO4− stopped. With increased TCE concentration caused by the remaining DNAPL, the residual MnO4− in the flow tank reacted and produced an increase in Cl− concentration at the end of the experiment.
 The injected TCE formed a zone of residual DNAPL saturation across the vertical depth of the tank and a small DNAPL pool at the bottom (Figure 7a). Digital images (Figure 7) indicated that Mn oxide precipitation started once MnO4− came in contact with dissolved TCE, which occurred in close proximity to the zone with pure TCE DNAPL. A zone of Mn oxide precipitation first became evident with a light brown color, located down gradient of where the residual DNAPL occurred (Figure 7b). There was a tendency for the MnO4− flood to bypass the zone with the highest DNAPL saturation, moving instead through a much less saturated zone in the upper portion of the tank (Figure 7c). With time, the precipitation of Mn oxide reduced the permeability across most of the tank. Mn oxide rapidly formed a rind of precipitates above the small DNAPL pool, which was located near the bottom of the tank (Figures 7d–7f). A greater injection pressure was required to maintain the flow near the end of the experiment. The experiment was halted when Mn oxide plugged the tank nearly completely and MnO4− could no longer be injected. Careful examination of the tank after the experiment indicated the presence of tiny CO2 bubbles, produced from the oxidation reaction. The gas bubbles likely played a role in reducing the permeability and the flow in the system. Although CO2 bubbles in the subsurface can dissolve into the aqueous phase according to equation 5, the rapid accumulation of CO2 early in the flushing led to the production of the bubbles. Flow bypassing also reduced the possibility for water to contact and dissolve the bubbles. Thus, the effect of the bubbles in reducing the permeability can be persistent.
 Flushing with MnO4− appears effective in removing residual DNAPL. However, much of the original volume of the pool of DNAPL at the bottom of the tank was evident at the end of the experiment. Similar observations have been reported by other researchers [MacKinnon and Thomson, 2002].
 This study utilized a series of laboratory experiments to explore the mass removal rate and permeability reduction during ISCO using MnO4− to clean up DNAPL contamination. The column and flow tank experiments continue to indicate that the oxidation of chlorinated ethylenes by MnO4− is a useful remediation technology. The DNAPL mass removal rate seems to follow a power relationship with flow rate and initial oxidant concentration. However, Mn oxide precipitates and CO2 gas bubbles present significant delivery problems. The rapid buildup of these reaction products in zones with high DNAPL saturation leads to pore plugging, which lowers the hydraulic conductivity, and causes flow bypassing. The reduction of K can be as high as 80% due to oxidation of a small quantity of TCE in column. Without controlling this problem, the flushing efficiency will be reduced. It is anticipated that in actual field settings, the issue of flushing efficiency needs to be considered in the design. For this technology to be practical, there is a critical need to find a solution for the precipitation of Mn oxide.
 This research is supported by the U.S. Dept. of Energy through Environmental Management Science Program under contract No. FG07-02ER63487.