In Situ Chemical Oxidation of Nonionic Organic Contaminants: The Effect of Soil Organic Matter and Minerals on Stoichiometric Efficiency

Abstract

Most dissolved organic contaminants readily react with sulfate radicals (SO₄^•-) formed in solution when persulfate is activated by thermolysis or base addition during in situ chemical oxidation (ISCO). However, under conditions encountered in soil and groundwater, hydrophobic organic contaminants tend to be associated with particles. To determine the potential of soils and aquifer solids to protect contaminants from oxidation, we measured the stoichiometric efficiency (i.e., the number of moles of contaminant transformed for each mole of SO₄^•- produced) for a homologous series of chlorinated benzenes using solid-to-water ratios approaching those encountered in the subsurface. Results indicate that sorption to inorganic surfaces (i.e., sand, clay minerals) reduced the stoichiometric efficiency by about three orders of magnitude relative to contaminants that remained in solution. At low initial oxidant concentrations (i.e., 10 mM), adsorbed contaminants were oxidized after they desorbed from inorganic solids to reestablish phase-partitioning equilibrium. In contrast, at higher oxidant concentrations (i.e., 500 mM), contaminant loss was attributable to SO₄^•- that reacted with chlorobenzenes at the inorganic particle surface. Absorption of organic compounds by particulate organic matter (i.e., Pahokee peat) offered greater protection from oxidation. For the most hydrophobic compounds studied (i.e., tetra-, penta-, and hexachlorobenzene), the presence of more than 1.5% organic matter by mass was sufficient to reduce the stoichiometric efficiency by an additional order of magnitude under conditions typical of ISCO. The effect of sorption on the efficacy of persulfate ISCO treatment for nonionic organic compounds can be predicted using information on their affinity for surfaces (i.e., the octanol-water partition coefficient, K_ow), the amount of persulfate applied, and the particulate organic matter content.

Graphical Abstract

INTRODUCTION

Contamination of soils and groundwater with organic chemicals is a worldwide problem.^1,2 Under conditions typically encountered at sites with subsurface pollution, most uncharged organic contaminants are associated with particulate organic matter and inorganic surfaces.. As a result, contaminants accumulate near the source zone and are slowly released into groundwater. Exposure to contaminated soil in the source zone and contaminated groundwater poses long-term risks to human health and biota. Among the strategies used to remediate hazardous waste sites, in situ chemical oxidation (ISCO) with sodium or potassium persulfate (S₂O₈^2-) has become popular due to its relatively low cost and effectiveness.¹ Persulfate ISCO treatment typically involves the injection of concentrated (i.e., 0.5–2 M) solutions of the oxidant into the subsurface, followed by thermal- or base-activation to generate sulfate radicals (SO₄^•-) and hydroxyl radicals (HO^•).¹

Although persulfate ISCO can remediate moderately hydrophobic contaminants, like benzene, toluene, ethylbenzene, xylene (i.e., BTEX compounds), perchloroethylene (PCE), and trichloroethylene (TCE),^3,4 its ability to treat more hydrophobic contaminants (e.g., polychlorinated biphenyls, polynuclear aromatic hydrocarbons, and chlorobenzenes) exhibits considerable variability.^1,4,5 Repeated ISCO treatment also has an inconsistent record with respect to its ability to achieve treatment goals.⁶ Researchers have suggested that these findings are due to the ability of surfaces to shield contaminants from oxidation by SO₄^•- and HO^ߦ.^4,7–9

Efforts to understand the effects of surfaces on treatment efficacy have assumed that most radical production takes place in the solution phase.^3,8–10 Some researchers have hypothesized that the small fraction of the SO₄^•- produced near the particle-water interface are responsible for the observed transformation of particle-associated contaminants,^4,10–12 with rates of transformation being determined by competition for radicals by organic matter and minerals at the surface.^{1,4,7,9,11,13–17} Alternatively, other researchers have presumed that SO₄^•- near the interface is rapidly depleted and that contaminant oxidation is attributable to radicals produced in the bulk solution either diffusing to the surface or reacting with contaminants as they desorb to re-establish phase-partitioning equilibrium.^18,19

To study these phenomena, researchers usually have employed experimental conditions that are substantially different from those encountered during ISCO. For example, many prior studies have used oxidant concentrations around 10 mM, which is one to two orders of magnitude lower than those used in ISCO. For convenience, researchers also have employed solid-to-water ratios that are orders of magnitude lower than those encountered in the subsurface (i.e., 1–10 kg/L). As a result, the fraction of compounds associated with solids tends to be underestimated (i.e., the experimental conditions result in a higher fraction of contaminants in the aqueous phase than conditions typically encountered in the subsurface).^{1,5,6,8,10,14,19,20}

To better understand the effect of partitioning of hydrophobic organic contaminants to aquifer solids on the efficiency of ISCO treatment with persulfate, we subjected six chlorobenzenes spanning a range of hydrophobicities (log K_ow 2.8–5.9; see Table S3) to conditions similar to those encountered during treatment. Relatively high concentrations of chlorobenzenes were used to facilitate precise measurement of phase partitioning and to represent conditions encountered in source zones of hazardous waste sites, where ISCO is often used for remediation. To track the effect of partitioning on the efficacy of ISCO, we measured the stoichiometric efficiency of the process (i.e., the number of moles of contaminant transformed for each mole of SO₄^•- produced) in the presence of varying mixtures of sand, clay minerals (i.e., kaolinite, montmorillonite), and particulate organic matter (i.e., Pahokee peat). By considering the differences between transformation processes on different types of surfaces, we provide new insight into factors affecting the efficacy of treatment and propose a model that can be used to predict treatment efficacy under different conditions.

MATERIALS AND METHODS

Chemicals, sources, and their purities are summarized in Table S1 and Text S1.

All experiments were conducted in 9-mL amber glass vials, used as sacrificial reactors. Reactors with a 1:1 solids-to-water ratio contained 6.5 g of solids. 4.875 mL of a 17.3 mg L⁻¹ solution with only one of the less hydrophobic chlorobenzenes (i.e., mono-, di-, and trichlorobenzene). This was added to the solids and equilibrated for 10–30 min using a wrist-action shaker before oxidant addition. To load the more hydrophobic compounds (i.e., tetra-, penta- and hexachlorobenzene), each chlorobenzene was dissolved in hexane (i.e., 50 mg L⁻¹). A 3.25 mL aliquot of the chlorobenzene stock solution was added to each reactor. After 24 hours, the hexane evaporated and 4.875 mL of deionized water was added to the vial. The reactors were allowed to equilibrate by the same procedure described before. Additional reactors without the addition of water were used to determine the initial mass of chlorobenzenes after the drying step.

Na₂S₂O₈ stock solutions (i.e., 2 M, 800 mM, and 40 mM) were prepared in deionized water. A 1.625 mL aliquot of the stock was added to each reactor (i.e., final persulfate concentration of either10 mM, 200 mM, or 500 mM). Control reactors without persulfate were prepared by adding 1.625 mL of a stock solution of Na₂SO₄ to each control reactor (i.e., 2M, 800 mM, or 40 mM). Compound recovery was determined by comparing the chlorobenzene mass from the dissolved and solid phases in controls with the mass in the initial stock solution for the less hydrophobic compounds or by comparing the initial mass of chlorobenzenes in the hexane to the mass recovered on the particles after the drying step.

In some experiments, reactors with a 4:1 solid-to-water ratio were prepared as above but with 13.9 g of solids. The chlorobenzenes were loaded onto the particles using the methods previously described, but with a smaller volume of the spiking solutions (i.e., 2.55 mL of mono-, di-, and trichlorobenzene solutions, 0.85 mL of Na₂S₂O₈, and Na₂SO₄ solutions).

Each experimental condition was run in triplicate.

The reactors were sealed with Teflon-lined screw caps without headspace and immersed in a water bath at 50^o C for the duration of the experiments. Monitoring of the temperature in sacrificial vials indicated that the temperature reached 50˚C within 10 minutes.

At pre-determined sampling intervals, reactors were removed from the water bath and put into an ice bath to slow the rate of thermal persulfate activation. After 20 minutes in the ice bath, samples were filtered through 0.45 µm glass fiber syringe filters, and the aqueous phase was transferred to a 20-mL glass scintillation vial. Aqueous samples were extracted with 6.5 mL of hexane in the vials by vigorously shaking for 30 seconds. Immediately after extraction, 10 µL of the aqueous phase were pipetted out for the S₂O₈^2- determination as described in a previous method (Text S2, Figure S1).^21,22

The reactor with the remaining particles was washed with 1.5 mL of acetone, followed by two washes of 3.25 mL of hexane. The combined solvent extracts were passed through the used syringe filter before being combined and saved for subsequent analysis by gas chromatography/mass spectrometry (GC/MS).

To determine the maximum desorption rate of chlorobenzenes from sand, N₂ was bubbled into 20 mL glass culture tubes at a rate of 200 mL min⁻¹ to strip away the dissolved chlorinated benzenes. The N₂ was bubbled through a glass frit placed above the solids to avoid turbulence near the surface of the particles. Sacrificial samples were collected at pre-determined intervals to determine desorption rates using the previously described extraction method.

Due to the low water content in reactors with a 4:1 particle-to-water ratio, the extraction method was modified. The entire content of the vial was transferred into a 20-mL scintillation vial without filtering. 6.5 mL of hexane and 1.5 mL of acetone were used to wash the empty reactor. The rinse solvents were transferred into the scintillation vial with the sample to extract the chlorobenzenes from both aqueous and solid phases simultaneously. The mixture was sonicated for 3 minutes. The supernatant organic phase was retained for analysis by GC/MS.

Solvent extracts were diluted by a factor of ten with hexane and analyzed by GC/MS (8890 GC and 5977B MS, Agilent, Santa Clara, CA, US) with a modified version of the U.S. EPA Method 625.1: Base/Neutrals and Acids by GC/MS²³ as described in Text S3. The limit of quantitation (LOQ) for the chlorinated benzenes ranged from 0.21×10⁻³ to 0.54×10⁻³ µM. Example calibration curves and details of the method are included in Text S3, and figures S2-S7.

The pH values in the reactors were measured with an Orion STAR A213 meter coupled with the Orion DB-10 pH meter (Denver Instruments, Bohemia, NY, US).

Total Organic Carbon (TOC) was measured using a Shimadzu TOC analyzer (Shimadzu Scientific Instruments, Pleasanton, CA, US). Details are included in Text S4.

RESULTS AND DISCUSSION

Oxidation of chlorinated benzenes adsorbed onto inorganic surfaces.

In the least complex solid matrix (i.e., sand only) the mono-, di- and trichlorobenzenes were distributed between the water and the particles (i.e., approximately 40%, 60%, and 80% of the mass was associated with the particles throughout the control experiments, respectively), whereas nearly all (i.e., >95%) of the tetra, penta- and hexachlorobenzenes were associated with the sand. The observed partitioning was consistent with predictions made with linear partition coefficients (K_d) measured in separate control experiments at 20˚C (Figure S9, Table S2). The dissolved fraction of mono-, di-, and trichlorobenzene in the persulfate-free control experiments at 50^oC (Figure S9) was not significantly different from those measured in control experiments at 20^oC (i.e., p = 0.34), which was consistent with previously reported data.²⁴ Under conditions used in these experiments, the fraction of chlorobenzenes in the dissolved phase increased by less than 5% when the samples were heated.

Under these experimental conditions (i.e., in the absence of a buffer and at an initial [S₂O₈^2-] of 500 mM), the pH dropped to a value below 2 within the first two hours when S₂O₈^2- was thermally activated, which is similar to what is expected to occur in the source zone during ISCO treatment, due to the production of sulfuric acid (Reaction S1).⁸ Under these conditions, S₂O₈^2- exhibited first-order decay, decreasing in concentrations by approximately 40% after 42 hours (Figure S10). The observed persulfate disappearance rates were consistent with values predicted for acid-catalyzed homogeneous activation at 50^oC (i.e., 10^−5.25 s⁻¹),⁸ indicating that the presence of sand or chain reactions initiated during chlorobenzene decomposition did not affect the S₂O₈^2– decay rate.

The three less hydrophobic chlorobenzenes were transformed in less than one hour (i.e., before the first sample was collected) when an initial S₂O₈^2– concentration of 500 mM was employed. Therefore, experiments were conducted at S₂O₈^2– concentrations lower than those used for ISCO (i.e., 10 mM). Under these conditions, the concentrations of dissolved mono-, di-, and trichlorobenzenes decreased to values below the LOQ values within 120 minutes (Figure 1a) while concentrations of the particle-associated mono-, di- and trichlorobenzenes decreased by approximately 30–80% (Figure 1b). Concentrations of the more hydrophobic chlorobenzenes (tetra-, penta- and hexachlorobenzene) on the sand did not change significantly (i.e., p < 0.15) when compared to controls without S₂O₈^2– (Figure S9). These results agree with previous studies^14,15,18,19 in which hydrophobic compounds adsorbed onto inorganic particles did not exhibit substantial losses when exposed to similar amounts of SO₄^•−.

Figure 1.

Transformation of chlorobenzenes in the presence of sand during thermolysis of 10 mM S₂O₈^2- at 50˚C in the a) aqueous and b) particulate phases. Experimental conditions: r_sw = 1 kg/L; total initial mass of chlorobenzenes: mono- = 0.749 µM; di- = 0.574 µM; tri- = 0.465 µM; tetra- = 0.06 µM; penta- = 0.474 µM; and hexa- = 0.479 µM.

When the chlorobenzenes in the aqueous phase were transformed, particle-associated chlorobenzenes should have desorbed to re-establish equilibrium in the phase distribution. The kinetics of desorption of particle-associated compounds depends upon the nature of the interaction as well as the experimental conditions.¹⁹ To determine whether the decrease in concentrations of particle-associated less hydrophobic chlorobenzenes was attributable to desorption, the maximum rate of desorption from sand was measured by N₂ stripping (Figure S11). Results indicate that the desorption of mono- di and trichlorobenzenes from the sand was fast enough to explain the observed losses of particle-associated chlorobenzenes under these conditions.

When a molecule of persulfate is thermally activated, two SO₄^•- are produced (Reaction 1). However, not all the SO₄^•- react with the chlorinated benzenes because a fraction reacts with S₂O₈^2- and water (Reactions 2-3).²⁵ To account for these processes, we used a kinetic model to predict the theoretical stoichiometric yield for the aqueous phase reactions (Reaction S2 and Tables S3-S4). Under the conditions depicted in Figure 2, about 1% of the SO₄^•- should have reacted with S₂O₈^2- while losses to reactions with H₂O were predicted to be negligible. The measured stoichiometric yields of the tri-, di-, and monochlorobenzene were close to the predicted values from the kinetic model (Figure 2a). In an additional experiment with mono-, di and trichlorobenzene present simultaneously, the predicted stoichiometric yields were also comparable to the measured stoichiometric efficiencies (Figure 2b). Furthermore, the observed trend of faster transformation of the least chlorinated benzenes was consistent with predictions made with the kinetic model^12,26 (Tables S3-S4). These results suggest that, under these conditions, the aqueous chlorobenzenes reacted rapidly with SO₄^•-, and the desorption from sand was fast enough to replenish the chlorobenzenes to the aqueous phase without greatly affecting the stoichiometric efficiency.

Figure 2.

Comparison of the measured and predicted stoichiometric yield for the transformation of less hydrophobic chlorobenzenes by SO₄^•- in the presence of sand when a) only a single chlorobenzene was present; b) all three chlorobenzenes are present simultaneously. The experimental conditions were as follows: [S₂O₈^2-]₀ = 10 mM; Temperature = 50^oC; r_sw = 1 kg/L; total initial mass of chlorobenzenes: mono- = 0.749 µM; di- = 0.574 µM; tri- = 0.465 µM; tetra- = 0.06 µM; penta- = 0.474 µM; and hexa- = 0.479 µM ; time = 2 hours.

$$S_2{O}_8^{2-}\stackrel{∆}{\to }{2SO}_4^{•-}$$ Reaction 1

$${SO}_4^{•-}+S_2{O}_8^{2-}\to {SO}_4^{2-}+S_2{O}_8^{•-}k=1.80\times {10}^5{\mathrm{M}}^{-1}{\mathrm{s}}^{-1}$$ Reaction 2¹⁷

$${SO}_4^{•-}+H_{2}O\to {HSO}_4^{-}+{HO}^{•}k=1.80\times {10}^{-3}{\mathrm{M}}^{-1}{\mathrm{s}}^{-1}$$ Reaction 3¹⁷

To assess the stoichiometric efficiency of the more hydrophobic chlorobenzenes, experiments were conducted at an initial S₂O₈^2– concentration comparable to those employed for ISCO (i.e., 500 mM). Under these conditions, mono-, di- and trichlorobenzene were transformed prior to the collection of the first samples whereas concentrations of the more hydrophobic compounds (i.e., tetra-, penta- and hexa-chlorobenzene) on the sand gradually decreased during the 42-hour experiment (Figure 3).

Figure 3.

Transformation of chlorobenzenes in the presence of sand expressed in terms of the overall mass of each compound. The experimental conditions were as follows: [S₂O₈^2-]₀ = 500 mM; Temperature = 50^oC; r_sw = 1 kg/L; total initial mass of chlorobenzenes: mono- = 0.749 µM; di- = 0.574 µM; tri- = 0.465 µM; tetra- = 0.06 µM; penta- = 0.474 µM; and hexa- = 0.479 µM.

The observed rate of transformation of more hydrophobic chlorobenzenes at the high [S₂O₈^2–] was about three times faster than the maximum rate of desorption rate observed in the nitrogen stripping experiments (Figure S11), suggesting that SO₄^•- reacted with the particle-associated compounds. For this to have taken place, the radicals would need to have diffused from the bulk to the surface before they reacted with other solutes (e.g., SO₄^•- loss through reaction with H₂O and S₂O₈^2– under these conditions results in a half-life of approximately 8 μsec; Equation S3).¹⁰ Using a diffusion coefficient (D_coeff) of 1065 nm²/μsec⁵, SO₄^•- was predicted to diffuse about 92 nm in 30 μsec (Equation S4) which is considerably further than the estimated thickness of the double layer (i.e., 0.25 nm; Equation S5) under these high ionic strength conditions (i.e., I = 1.5 M).²⁷ The stoichiometric efficiency measured for the tetra-, penta-, and hexachlorobenzene in the only sand experiments (Figure 4) suggests that less than one-in-a-thousand of the SO₄^•- produced in the bulk diffused to the surface and reacted with adsorbed compounds. It is unclear if there was a SO₄^•- sink at the surface of particles that competed for the radical with the particle-associated chlorobenzenes.

Figure 4.

Measured stoichiometric efficiencies for sand amended with clay or particulate organic matter. The experimental conditions were as follows: [S₂O₈^2-]₀ = 500 mM; Temperature = 50^oC; r_sw = 1 kg/L; total initial mass of chlorobenzenes: mono- = 0.749 µM; di- = 0.574 µM; tri- = 0.465 µM; tetra- = 0.06 µM; penta- = 0.474 µM; hexa- = 0.479 µM; time = 24 hours.

Under these high oxidant concentration conditions, which are similar to those that occur in the source zone during ISCO treatment, the pH was below the pH_pzc for silica (i.e., the surface of the sand was positively charged).²⁸ Thus, S₂O₈^2- and SO₄^-• concentrations were not affected by electrostatic repulsion from the particle surfaces.

Results from repetition of the experiment at an intermediate concentration of S₂O₈^2- (i.e., 200 mM) were consistent with a transition from desorption-limited aqueous phase reactions for tetrachlorobenzene at 200 mM to a combination of desorption and on the surface of the sand becoming important at 500 mM (Figure S11). For penta- and hexachlorobenzene, reactions on the surface explained most of the observed transformation at 200 and 500 mM persulfate.

Kinetic models that included second-order rate constants for dissolved solutes were used to predict the rate of transformation of dissolved chlorobenzenes by SO₄^•- (Reaction S8). Results from these models showed good agreement with the observed behavior of mono-, di-, and trichlorobenzene (i.e., R² >9.0) (Figure S11). However, these models were unable to describe the transformation of the more hydrophobic chlorobenzenes. As described in the results and discussion section, tetra-, penta-, and hexachlorobenzene associated with inorganic surfaces were oxidized by sulfate radicals through a process that cannot be modeled readily with kinetic equations for dissolved solutes.

The experiment depicted in Figure 3 was repeated with a small amount of clay minerals (i.e., 0.015 gm clay/gm sand) added to the sand. This particle mixture was representative of conditions in aquifers, where the water-bearing formations tend to have clay contents well below 10%..⁴ When kaolinite was present, the solution phase concentration of the least chlorinated benzenes did not change appreciably. Adding montmorillonite—a higher surface area clay mineral due to its 2:1 layered structure—instead of kaolinite lowered the dissolved chlorobenzene concentrations, with more than 90% of the mono-, di-, and trichlorobenzene associated with the solids in the control experiments. The partitioning of hydrophobic compounds could be described by a two-phase model that included the different solids (i.e., sand and clay minerals; see Equations S6-S8).^5,29 Using the partitioning experimental results from control experiments and the two-phase model, the distribution coefficient (K_min) for the clays was estimated (Tables S5-S6). These data indicated that more than half of the tetra-, penta-, and hexachlorobenzene were associated with the clay fraction for kaolinite and montmorillonite (Text S5 and Figure S12).

The presence of kaolinite and montmorillonite increased the specific surface area of the solids by a factor of 3 and 75, respectively. Despite the higher surface area and the altered partitioning of the chlorobenzenes, neither the rate of persulfate activation nor the fraction of the three hydrophobic chlorobenzenes transformed was altered by the presence of clay. As a result, the measured stoichiometric efficiency was identical for chlorobenzenes associated with sand and with both types of clays (First three bars in Figure 4). Mono-, di-, and trichlorobenzene were removed completely under these experimental conditions within two hours, which was consistent with stoichiometric efficiencies greater than 10%. The rate of transformation of the least chlorinated compounds was faster than the maximum desorption rate (Figure S11), thus, the transformation rate was not limited by desorption. Under these conditions, about 40% of the SO₄^-• produced by reaction 1 was predicted to react with S₂O₈^2–.

Diffusion of uncharged organic compounds into the interlayer of the montmorillonite occurs over a timescale of weeks to years.^30–33 To assess the potential for diffusion of chlorobenzenes into the clay interlayer to protect the contaminants from oxidation, samples containing 10% montmorillonite and 90% sand were prepared at a r_sw of 1 kg/L and stored for 1 month at room temperature to gain insight into the possible effects of a longer equilibration period under laboratory conditions.. No differences in the stoichiometric efficiency for oxidation of the chlorobenzenes were observed (Figure 4). The results suggest that the protective effect of 2:1 clays was small for these compounds or that additional time is required for diffusion; nevertheless, this phenomenon merits additional study because longer aging (e.g., > 12 months) and/or the presence of more hydrophobic compounds might result in more migration of contaminants to the interlayer, which could have a greater impact on stoichiometric efficiency.

Our results show that for uncharged compounds, the stoichiometric efficiency for compounds associated with inorganic particles was between 2 to 3 orders of magnitude lower than the efficiencies for compounds that were present in solution. Although the second-order rate constants for reactions of chlorobenzenes with SO₄^•- decreases as the number of chlorine substituents increases (Table S3), most of the decrease in the stoichiometric efficiency was attributable to the lower reactivity of compounds adsorbed to the inorganic surface.

It is worth noting that redox-active minerals, like pyrolusite, ferrihydrite, and goethite can increase the rate of S₂O₈^2– activation through reactions taking place on the mineral surface.^11,17 Additional research is needed to assess the effect of these redox-active minerals on the transformation of particle-associated contaminants during ISCO. Furthermore, alumina has been reported to scavenge SO₄^•-, potentially affecting the stoichiometric yield for contaminant transformation in the presence of minerals.¹⁵ The effect of these reactive minerals on the stoichiometric efficiency under conditions observed in aquifers (e.g., < 10%)⁴ requires further investigation.

Oxidation of chlorinated benzenes absorbed into particulate organic matter.

To assess the impact of absorption of the chlorobenzenes by particulate organic matter, experiments were conducted in sand amended with particulate organic matter at two different concentrations (i.e., 0.15% and 1.5% Pahokee peat by weight). For the experiment amended with 0.15% of Pahokee peat, the partitioning of the mono-, di-, and trichlorobenzenes was comparable to the sand only experiments. We predicted that approximately equal amounts of the tetra- and pentachlorobenzene would be adsorbed onto the sand and absorbed within the particulate organic matter. In contrast, over 60% of the hexachlorobenzene should have been absorbed into the particulate organic matter under these conditions (Table S7-S9). Consistent with these partitioning predictions, under typical ISCO conditions (i.e., [S₂O₈^2-]_initial = 500 mM, T = 50^oC), the rates of transformation of mono-, di-, tri-, tetra- and pentachlorobenzene were similar to those observed in the absence of particulate organic matter, whereas hexachlorobenzene transformation slowed (Figure S13). These results suggest that for each of the chlorobenzenes there is a threshold particulate organic matter concentration below which particulate organic matter has a negligible effect on the rate of contaminant transformation because most of the compound remains adsorbed onto inorganic surfaces where it could be oxidized by SO₄^•−. This threshold depends upon partitioning of the chlorobenzenes between the inorganic surfaces and the particulate organic matter. At low particulate organic matter concentrations, the majority of the hydrophobic compounds will be adsorbed onto inorganic surfaces, where they are subject to oxidation by SO₄^•−. As the particulate organic matter content increases, a larger fraction of the contaminants will be absorbed by the organic particles, leading to a drastic decrease in transformation due to the protective effect of the organic matrix.

When the sand was amended with 1.5% Pahokee peat by weight, a larger fraction of the mono-, di- and tri-chlorobenzenes were associated with the solid phase in the controls (i.e., between 5–10% more compared to the particulate organic matter free experiments depicted in Figure 1). Predictions based on partitioning coefficient (K_d) values (Table S8) indicated that more than half of the more hydrophobic chlorobenzenes were associated with particulate organic matter in this case. After exposing the system to the same treatment conditions, the mono-, and dichlorobenzene were transformed completely within 24 hours, while 20% of the total mass of trichlorobenzene remained (Figure S13). In contrast, the extent of transformation of tetra-, penta-, and hexachlorobenzene was almost 90% lower than in the experiments with 0.15% or no particulate organic matter (Figure S13). Because most of the chlorobenzenes in this case were absorbed in the particulate organic matter instead of adsorbed onto the sand, these results suggest that the particulate organic matter protected the hydrophobic contaminants from SO₄^•−.

The presence of particulate organic matter (i.e., 1.5% Pahokee peat by weight) also increased the rate of S₂O₈^2- disappearance by about 50% relative to rates observed in experiments conducted with sand and clay-amended sand. The observed rate of transformation was consistent with direct oxidation reactions between S₂O₈^2- and electron-rich moieties in the particulate organic matter or the initiation of radical chain reactions when SO₄^•- reacted with organic matter.^16,34,35 Regarding the faster observed S₂O₈^2- decomposition rate, results from previous studies indicate that the presence of particulate organic matter (i.e., at concentrations of particulate organic matter ranging from 0.06–3.5%) also increased the rate of S₂O₈^2- decomposition.^4,34,35 In addition to enhancing the rate of S₂O₈^2- disappearance, the addition of particulate organic matter to the system resulted in the release of polymeric organic matter to the solution phase where it could react with SO₄^•−. However, under the conditions employed in our experiments, the impact of particulate organic matter was relatively small: scavenging of SO₄^•- by polymeric organic matter released from the Pahokee peat was estimated to account for <1% of the SO₄^•- loss (Text S6). Both the increase in S₂O₈^2- activation by reactions with the particulate organic matter and the reduction of the mass of the chlorobenzenes transformed by SO₄^•- resulted in a drastic reduction of the stoichiometric efficiency for the most chlorinated benzenes compared to experiments with little or no organic matter content (Figure 4). These results suggest that the organic matter content in soils and aquifers directly affect the stoichiometric efficiency of the treatment, with higher organic matter content dramatically lowering the stoichiometric efficiency.

One possible explanation for the protective effect of particulate organic matter is the scavenging of SO₄^•- that diffuses to the particle-water interface by the organic matter. Results from prior studies indicate that SO₄^•- reacts with electron-rich substituents in dissolved organic matter with second-order reaction rates in the order of 10⁶ to 10⁷ M⁻¹ s⁻¹ (i.e., as moles of C).¹ Furthermore, it has been shown that SO₄^•- hydroxylates organic moieties from particulate organic matter.³⁶ Because the mass of particulate organic matter is much greater than that of the absorbed chlorobenzenes (i.e., in these experiments chlorobenzenes mass accounted for less than 0.17 % of the total mass of organic matter present in the system), if the chlorobenzenes were evenly distributed throughout the particulate organic matter, they would be unlikely to encounter SO₄^•- before it reacted with an electron-rich moiety within the organic matter (i.e., there are almost a 1000 carbons associated to the particulate organic matter for every carbon associated with the chlorobenzenes). In contrast, the chlorobenzenes adsorbed onto inorganic particles are directly exposed to the SO₄^•- that diffuse to the interface.³⁶

Another possible explanation for the protective effect of organic matter could be related to the ability of dissolved organic matter to act as an inhibitor for pollutant transformation by reducing the radical intermediates generated after one-electron oxidation by SO₄^•-, thereby reversing the initial transformation step.³⁷ It is possible that this mechanism can also happen on surfaces, but more mechanistic studies are needed.

To assess the effect of particulate organic matter on the rate of transformation of chlorobenzenes, the experiment was repeated with a higher concentration of particulate organic matter (i.e., 2.3% Pahokee Peat by weight) for three consecutive cycles of S₂O₈^2- re-application and thermal activation (i.e., [S₂O₈^2-]_initial = 500 mM, T = 50^oC, 48 hours/cycle). Under these conditions, the fraction of mono-, di-, and tri-chlorobenzenes associated with the solid phase in the controls was around 10% greater than for the experiments without Pahokee Peat. Predictions based on partitioning coefficient (K_d) values (Table S8) suggested that between 50–60% of the most hydrophobic chlorobenzenes were associated with particulate organic matter in this case. After the first cycle of treatment, mono- and di-chlorobenzene were transformed completely, whereas around 60% of the trichlorobenzene was transformed (Figure S14). The mass of the more chlorinated benzenes that remained after the treatment ranged from 40 to 70% (Figure S14). Nevertheless, the mass of the remaining compounds did not change significantly after the second and third persulfate exposure (i.e., p > 0.12). These results suggest that the SO₄^•- transformed a small fraction of chlorobenzenes that was adsorbed onto the sand during the first treatment, but the absorbed fraction inside the particulate organic remained protected during all three treatments or transformation was very low. The stoichiometric efficiency for the tri-, tetra-, penta-, and hexachlorobenzene were in the order of 10⁻³% (Figure 5), which is consistent with the previous results in the presence of particulate organic matter (Figure 4). For the second and third cycles, the drop in stoichiometric efficiency indicates that once the fraction of chlorobenzenes that remained associated with inorganic particles was removed, the compounds associated with particulate organic matter were protected (Figure 5). In some cases, it was not possible to distinguish stoichiometric efficiencies in the order of 10⁻³-10⁻⁴% from zero. Further investigation is needed to assess the effect of multiple cycles on stoichiometric efficiencies for soils with higher f_oc.

Figure 5.

Measured stoichiometric efficiencies for sand amended with particulate organic matter. The experimental conditions were as follows: 0.023 gm Pahokee peat/gm sand; [S₂O₈^2-]₀ = 500 mM; Temperature = 50^oC; r_sw = 1 kg/L; total initial mass of chlorobenzenes: he- = 0.479 µM; pe- = 0.474 µM; te- = 0.06 µM; tr- = 0.465 µM; di- = 0.574 µM; and mono- = 0.749 µM.

Reapplication of Na₂S₂O₈ is one of the main strategies suggested for treating hydrophobic compounds and managing the concentration rebound of contaminants (i.e., the gradual increase of concentration of pollutants in groundwater after ISCO treatment).^1,6 The efficacy of this approach for moderately hydrophobic contaminants may depend upon the effect of the oxidant on particulate organic matter. Oxidation of particulate organic matter by SO₄^•- could alter its structure by adding -COOH or -OH groups to surface moieties which would increase the polarity of the organic matter. It also could release small molecules to solution, where they would be exposed to high concentrations of SO₄^•- and possibly undergo mineralization. Despite the fact that oxidation of the particulate organic matter might make it easier to oxidize the more hydrophobic chlorobenzenes, we estimate that around only one of every thousand SO₄^•- generated during ISCO would be likely to reach the surface of solid particles, which means that each treatment with 0.5 M Na₂S₂O₈ would only be capable of mineralizing less than 0.01% of the total mass of organic matter present in a shallow aquifer (i.e., r_sw = 4 kg/L and a f_oc of 1%) (Text S6 and Reaction S9).³⁸

To gain further insight into the relevance of these experiments to field conditions, the particle-to-water ratio was increased by a factor of four while maintaining the same mass of particulate organic matter in the system (i.e., to a r_sw value of 4 kg/L; Figure S15). The additional surface area from the added sand shifted the distribution of the chlorobenzenes on the solids. Under these conditions, we predict that over 50% of the tetra-, penta-, and hexachlorobenzenes were associated with the sand. In agreement with this prediction, the measured stoichiometric efficiencies were comparable to those observed at lower r_sw values in the absence of particulate organic matter (Figure S16). Using results of the partitioning model (Equations S6-S8), we predict that the protective effect of particulate organic matter will lower the stoichiometric efficiency of the process at values between 0.15% and 1.5% organic matter by weight (i.e., f_oc values between 0.075% and 0.75% approximately) for compounds with log K_ow values between 4.6 and 5.1. This drop in the stoichiometric efficiency was first observed at the f_oc value at which partitioning to the organic matter was an order of magnitude greater than partitioning to inorganic particles (Figure S17).

Predicting the efficacy of ISCO treatment of nonionic organic contaminants.

According to our findings, the oxidation of nonionic organic compounds under conditions relevant to ISCO treatment is affected by four phenomena: (1) reactions of the contaminants with SO₄^•- in solution; (2) reactions of the contaminants with SO₄^•- on the surfaces of inorganic particles; (3) protection of contaminants from SO₄^•- by absorption into particulate organic matter; and (4) desorption of contaminants from particles (Figure S20). The relative importance of each of these phenomena will depend mainly on the properties of the contaminants and the composition of the solids. Reactions of contaminants in solution with SO₄^•- will predominate for less hydrophobic contaminants and for moderately hydrophobic compounds in the presence of sandy, low organic matter content aquifer solids. These conditions will result in the highest treatment efficiencies. In contrast, hydrophobic contaminants and organic matter-rich soils and shallow aquifers (i.e., f_oc values greater than 1%) will lead to low treatment efficiency, due to the protective effect of absorption into particulate organic matter.

For those compounds that were mainly associated with particles (i.e., chlorobenzenes with three or more chlorine substitutents), f_oc was the main factor determining the stoichiometric efficiency and hence the extent of transformation when persulfate was activated. In the absence of particulate organic matter, the stoichiometric efficiency was between 0.01 and 0.04%. When a sufficient amount of particulate organic matter was added to shift the sorption to a point at which more than about 90% of the compound was associated with the particulate organic carbon the stoichiometric efficiency decreased by about an order of magnitude (Figure S17 and S18). These experimental observations were used to model how the stoichiometric efficiency would change as the f_oc increased, assuming that the fraction adsorbed onto inorganic particles can be transformed more efficiently than the fraction absorbed into particulate organic matter (Equations 1, 2 and 3).

$$\varepsilon ={\varepsilon }_{inorg}.f_{inorg}+{\varepsilon }_{pom}.f_{pom}$$ Equation 1

$$f_{inorg}=\frac{f_{sand}.{Kd}_{sand}+{\sum }_i^n{f}_{clayi}.{Kmin}_{claii}.A_i}{f_{sand}.{Kd}_{sand}+{\sum }_i^n{f}_{clayi}.{Kmin}_{claii}.A_i+foc.Koc}$$ Equation 2

$$f_{pom}=\frac{foc.Koc}{f_{sand}.{Kd}_{sand}+{\sum }_i^n{f}_{clayi}.{Kmin}_{claii}.A_i+foc.Koc}$$ Equation 3

Where:

ɛ Stoichiometric efficiency at any given condition [%];

ɛ_inorg Stoichiometric efficiency for compounds sorbed to inorganic particles [%];

ɛ_pom Stoichiometric efficiency for compounds sorbed to organic particles [%];

f_inorg Fraction of compounds associated with inorganic particles [-];

f_pom Fraction of compounds associated with organic particles [-];

f_sand Fraction of sand in the solids;

f_clay Fraction of clay in the solids;

f_oc Fraction of particulate organic matter;

K_d,sand Distribution coefficient for sand size particles [L/kg];

K_min Distribution coefficient for clay i [L/m²];

A_i Specific surface area mineral i [m²/m³];

K_oc Organic carbon normalized distribution coefficient [L/Kg];

To visualize the model outputs, we illustrate predictions made for five hypothetical compounds (i.e., log K_ow values ranging from 2 to 6) that react with sulfate radicals in a manner similar to what we observed for the chlorobenzenes (Figure 6). Based on our experimental data, we assumed the stoichiometric efficiency for sorbed compounds in the absence of particulate organic matter to be: 0.10%, 0.08%, 0.06%, 0.04%, 0.02% for Log K_ow values of 2, 3, 4, 5, and 6, respectively. Similarly, we used our data to estimate values for the K_d-sand of 0.5, 2, 5, 11, and 12 L/kg, and Log K_oc values of 1.5, 2.4, 3.3, 4.3, and 5.2 for compounds with Log K_ow of 2, 3, 4, 5, and 6, respectively. We used an upper limit for the stoichiometric efficiency of reactions of compounds associated with particulate organic matter of 0.001% (i.e., ɛ_pom) based on experiments with hexachlorobenzene in which over 99.9% of the compound was associated with the particulate organic matter and no measurable transformation was observed. for these experiments, we estimated that we would be unable to detect the transformation of less than 5% of the particle-associated hexachlorobenzene due to the precision of our analytical method.

Figure 6.

Predicted stoichiometric efficiencies of SO₄^•- for nonionic organic contaminants as a function of the fraction of organic carbon of solids.

During persulfate ISCO, the contaminants in the dissolved phase (i.e., compounds with a very low affinity for soils and aquifer solids, typically log K_ow values much less than 2) are rapidly oxidized or are displaced by the injected oxidant solution. If they are pumped into a recovery well, they may be oxidized when more oxidant is added and the solution is reinjected (i.e., by reactions with solution phase SO₄^•- produced by thermolysis or base-activation).^4,7 For these compounds the stoichiometric efficiencies should range between 10 and 100% depending upon concentrations of S₂O₈^2-, other contaminants, and solutes capable of scavenging SO₄^•- (e.g., dissolved organic matter, HCO₃^-, CO₃^2-, Cl^-).^4,7,9 Typically, activation of 500 mM Na₂S₂O₈ should be sufficient to transform contaminants if they are present at concentrations below 1 mM. This prediction is consistent with field observations (e.g., a single treatment with 500 mM Na₂S₂O₈ leads to decreases in concentrations of polar non-ionic contaminants like MTBE).³⁹

For moderately hydrophobic compounds (i.e., log K_ow between 2 and 4), the fraction organic carbon of the soil or aquifer solids should affect treatment efficacy, with a decrease in stoichiometric efficacy of a little less than an order of magnitude for compounds with log K_ow values of 3 when the f_oc value increases from 1 to 5% (Figure 6). For compounds with log K_ow values of 4, an order of magnitude decrease in stoichiometric efficiency will take place when f_oc increases from 1 to 2%. For more hydrophobic compounds (i.e., log K_ow > 4), a decrease in efficiency of ISCO treatment will take place at even lower f_oc values. Contaminated sites vary considerably with respect to the fraction organic carbon.^1,4 Shallow aquifers and surficial soils (i.e., those with depths ranging from zero to five meters) typically exhibit f_oc values between 1 and 5%. ²⁸ In contrast, aquifers composed mainly of sand and gravel tend to exhibit considerably lower f_oc (e.g., 0.01–0.2%).³⁸ Based on the model depicted in Figure 6, for surficial soils and shallow aquifers, it will be challenging to remediate nonionic contaminants with log K_ow values above 4. For aquifers with lower f_oc values, it may be possible to treat these more hydrophobic contaminants, depending upon the organic matter content of the aquifer solids and the concentrations of solutes that react with SO₄^•- in the aqueous phase.

The presence of clay is an additional factor that could be considered in the model. The adsorption to clay could compete with the absorption of compounds by particulate organic matter. As a result, soils with the same content of particulate organic matter but differing clay content might have different stoichiometric efficiencies (Figure S18). An increase in clay content can increase the stoichiometric efficiency for soils under these conditions. Using the model described in Equation S7, it is possible to compare the fractional K_d associated with the sand, clay, and organic fractions to recalculate the model predictions using the same criteria described above . The variability of adsorption behavior of nonionic hydrophobic compounds in the presence of different types of clays constitutes one major challenge in the prediction of ISCO efficacy for real soils. ^29,40

Environmental Implications

Results from previous studies indicated that the efficacy of persulfate ISCO in treating hydrophobic contaminants varied in a manner that was difficult to anticipate.^1,4–9 Our results suggest that some of the variability is attributable to the absorption of contaminants by particulate organic matter. The model that we developed allows engineers to predict the stoichiometric efficiency of the treatment process with readily available parameters (i.e., log K_ow and f_oc). The predicted stoichiometric efficiency can be used to estimate the minimum amount of oxidant needed for remediation (i.e., under field conditions preferential flow paths for the oxidant and heterogeneity will also affect the efficacy of treatment). Despite these promising results, additional research is needed to assess model predictions for authentic soil samples, where additional constituents, such as iron oxides, black carbon, and aluminum oxides might affect phase partitioning or alter the radical chemistry at the solid-water interface.. Similarly, the use of analytical techniques like Fourier Transform ion cyclotron resonance mass spectrometry could be used to characterize the POM before and after oxidation to gain insight into the tendency of SO₄^•- to be scavenged by particulate organic matter.

The ability of particulate organic matter to substantially slow or prevent the oxidation of more hydrophobic compounds poses a limitation to efforts to use ISCO for site remediation. Our results suggest that oxidizing the particulate organic matter with multiple applications of S₂O₈^2- is an ineffective strategy for overcoming this limitation. Pre-treatment of aquifer solids with H₂O₂ might have a greater impact on the organic matter,¹ which could result in a higher stoichiometric yield in subsequent S₂O₈^2- treatment steps.

Alternatively, additives that solubilize hydrophobic contaminants before oxidative treatment could overcome the protective effect of particulate organic matter. The coupling of ISCO with surfactant washing has been rejected because surfactants would compete with the contaminants for SO₄^•−.^1,4 Nevertheless, for hydrophobic contaminants that are otherwise resistant to ISCO treatment, even a small change in their partitioning could increase the stoichiometric efficiency by several orders of magnitude, allowing them to be treated, especially if a surfactant could be found that was less reactive with SO₄^•−.⁴¹ Additional research is needed to assess the potential for using surfactants coupled with ISCO to treat hydrophobic contaminants, considering the treatment benefits and their potential negative effects on the environment.

SYNOPSIS:

A model is developed to predict the impact of sorption to surfaces on the efficacy of in situ chemical oxidation with persulfate.