Optimization and Analysis of a Slow-Release Permanganate Gel for Groundwater Remediation in Porous and Low-Permeability Media

: Dense nonaqueous phase liquids (DNAPLs) like trichloroethylene (TCE) serve as the most common form of groundwater pollution in the world. Potassium permanganate (KMnO 4 ) is a strong oxidant that can quickly destroy DNAPLs into innocuous products. Slow-release permanganate gel (SRPG), a mixture of colloidal silica (CS) and KMnO 4 , has been recently developed as novel treatment option for dilute and large plumes of DNAPLs in groundwater. The objective of this study was to characterize and optimize gelling and release properties of a SRPG solution in saturated porous media. It was hypothesized that CS and KMnO 4 content of the SRPG constrain gelation and release duration. Batch and column tests showed that gelation could be delayed through manipulation of the KMnO 4 content. In column tests, silica content had little effect on the gelation lag stage and release rate but inﬂuenced duration of permanganate release. Flow tank tests comparing Bindzil 1440 (B-40) SRPGs with pure KMnO 4 solutions under varying media conditions demonstrated that the presence of CS enhanced lateral spread and prolonged release duration of the oxidant.


Introduction
Dense nonaqueous phase liquids (DNAPLs) are a class of harmful organic chemical liquids that serve as the most abundant form of groundwater contamination in the world [1]. As DNAPLs are denser than water and relatively insoluble, they can infiltrate the subsurface until an impermeable layer is reached where they then accumulate to form pools and plumes [2]. Trichloroethylene (TCE) is a DNAPL of particular concern as it is the most common organic groundwater contaminant. A maximum contaminant level (MCL) for TCE in drinking water has been set at 5 µg/L [3].
In situ chemical oxidation (ISCO) techniques have become a common combative scheme against DNAPL contamination. In ISCO schemes, a strong chemical oxidant is injected into the subsurface via wells to react with and transform contaminants into environmentally innocuous by-products through a series of reaction steps [4]. Potassium permanganate (KMnO 4 ) is an oxidant commonly used in ISCO schemes, but encounters a major problem, e.g., precipitation of the solid oxidation product manganese dioxide (MnO 2 ) which results from rapid oxidation by permanganate (MnO 4 − ). The MnO 2 plugs pores in the subsurface which causes the oxidant to bypass contaminant zones, leading to a rebound in aqueous concentrations when flushing is terminated [2].
The objective of this study was to characterize and optimize gelling and release properties of a slow-release permanganate gel (SRPG) as a novel low-cost treatment option for large, dilute DNAPL plumes in groundwater. The SRPG is a mixture of KMnO 4 and colloidal silica (CS) which serves as a gelling agent. The SRPG may address problems that persist in other ISCO schemes, e.g., lack of lateral spreading when using slow-release KMnO 4 pellets [5] and pore plugging due to MnO 2 precipitation with KMnO 4 flushing schemes [6]. First, the injection of the SRPG as a liquid can allow MnO 4 − to spread laterally, facilitating mixing between the oxidant and contaminant. As MnO 4 − is released slowly through a gelled CS matrix and oxidize dilute dissolved TCE plume, the SRPG should also produce less MnO 2 precipitation.
Lee and Gupta [7] previously demonstrated gelation lag times up to 6 h in saturated, porous media through manipulation of the KMnO 4 concentration ([KMnO 4 ]) in column flow-through tests with Bindzil 1440 CS. The specific objective of this study was to further optimize SRPG to obtain a more desirable gel lag time, release rate, and release duration in saturated, porous media with flowing water. Ideal conditions were set at a gel lag time ≥3 d, a release rate ≥850 µg/d, and a release duration ≥3 wk. These values were chosen as benchmark variables based on the time and amount of KMnO 4 required to travel 1.5 m from injection wells at a flow rate of 0.5 m/d to degrade a TCE plume of 100 µg/L to concentrations less than the MCL (5 µg/L). It was hypothesized that CS and KMnO 4 content of the SRPG constrain gelation and release duration. A higher silica and KMnO 4 content was expected to combat dilution by providing more opportunities for interparticle interaction. This would promote the formation of siloxane Si-O-Si bonds, and thus yield a stronger gel that would reduce permeability and contact with water per surface area [7,8].
Permanganate can diffuse into low-permeability units and counter-diffuse back into groundwater over time [9][10][11]. Therefore, a portion of this study was also devoted to testing the SRPG solution in porous media with a low-permeability lens. It was hypothesized that the low-permeability media could serve as a secondary slow-release system to prolong the release duration of the SRPG. Fractures in low-permeability layers were expected to further enhance the release duration.

Potassium Permanganate as an Oxidizing Agent
KMnO 4 is a strong oxidant (E o = 1.68 V) and a highly water soluble (64 g/L at 20 • C) inorganic compound. Due to its strong oxidizing capacity, it has become one of the most widely used compounds in contaminated water remediation over the last 20 years [10,12,13]. Ultimately, MnO 4 − oxidation of TCE produces MnO 2 precipitates, chloride, carbon dioxide gas, and acidity [4,14]. The overall net reaction can be written as [2]: C 2 HCl 3 + 2 MnO 4 − → 2 MnO 2 (s) + 2 CO 2 (g) + 3 Cl − + H + (1) KMnO 4 is advantageous as an oxidizing agent as it remains relatively stable in the subsurface, is effective over a wide pH range, and is relatively inexpensive [14]. MnO 4 − also has a relatively slow reaction rate which allows it to travel greater distances in zones with average to high permeability and to diffuse into zones with low permeability [15].

Colloidal Silica as a Gelling Agent
Colloidal silica is a sol, or solution consisting of nanoparticles of amorphous silicon dioxide (SiO 2 ) suspended in aqueous phase in water [7,13]. To prevent premature gelation, CS solutions are typically stabilized during manufacturing by elevating the pH with alkaline solutions such as sodium hydroxide (NaOH). This negatively charges the particles so that they repel each other, thus prohibiting gelation [16,17].
Results from previous studies suggest that CS should serve as an effective gelling agent with KMnO 4 . CS has small particles capable of traveling through porous media, is nontoxic, is biologically and chemically inert, and is durable [18]. In addition, CS has the benefit of a modifiable gelation time for desirable longitudinal, transverse, or vertical transport [7].
Control over gelation rate is important so that the SRPG solution can be modified to travel a desired distance to the contaminant before becoming so viscous that transport is restricted [17]. Sol-gel aggregation is influenced primarily by ionic strength and pH but can also be affected by factors such as temperature, particle size, and silica concentration [8,17]. In general, the gelation time of CS is decreased by increasing the ionic strength [17]. With addition of a salt such as K + , the particles in the CS solution lose their repulsive forces and siloxane bonds can form [16]. The relationship between pH and gelation rate is parabolic. At pH ≥ 8, the particles will bear a slightly negative charge to drive interparticle repulsion. The most rapid gelation rate occurs at a neutral to slightly acidic pH, and when the pH drops below 5, the attraction between particles is lost [8,16,19].

The Slow-Release System
For groundwater remediation, slow-release MnO 4 − serves as a semi-passive option for the in situ remediation of TCE [4]. The SRPG implements a matrix-type slow-release system in situ in which there are multiple pockets of MnO 4 − distributed within a matrix. In a matrix-type slow-release system, oxidant release is governed by diffusive transport through secondary permeability formed within the matrix [5]. As MnO 4 − on the edges of the CS matrix is released, secondary permeability would be created which allows for the slow release of MnO 4 − farther inside the matrix by diffusion through the permeable outer gel matrix. This causes an increase in the diffusion length within the matrix over time. As a result, there would be an initially high spike in [MnO 4 − ] followed by its gradual decline [2]. As solution can diffuse into and counter-diffuse from low-permeability units [9][10][11], part of this study focused on testing the feasibility of producing a secondary slow-release system in situ with the injection of SRPG solution. It was expected that as a gel, the SRPG should be able to offer more lateral area for diffusion of MnO 4 − into a clay matrix compared to a pure KMnO 4 solution mainly due to the higher viscosity of the gel. During counter-diffusion, this should allow the SRPG to behave as a long-term source of MnO 4 − to target DNAPLs. Solpuker et al. [11] investigated the transport of KMnO 4 mixed with ultra-dense silicate solutions in porous media containing low-permeability layers and lenses. Even in the silicate matrix, MnO 4 − was capable of diffusing into a low-permeability clay lens. In the first 2 d, the MnO 4 − infiltrated the clay where it remained emplaced for 11 d. It took~1 month for the MnO 4 − to be depleted from the tank. DNAPLs often gather near low-permeability units, and this tendency could be exploited to extend the release duration of the oxidant in the target zone.  Table 1. SRPG solutions were also manufactured with matrices consisting of 25 mL of a 40 wt.% CS solution mixed with 25 mL of a 50 wt.% CS solution. After mixing, the SRPG solution was deposited into a 50 mL high-density polyethylene container, covered with plastic wrap, and placed under a fume hood to maintain constant temperature. Gel times were estimated for SRPG solutions through visual observations. The pH of CS solutions was measured with a pH meter (IQ Scientific Instruments), and viscosity was measured from 20 g/L SRPG solutions with a viscometer (HAAKE Viscotester 2 Plus, Thermo Fisher Scientific, Waltham, MA, USA).

Column Flow-Through Tests in Porous, Saturated Media
Two-dimensional column flow-through tests served as preliminary tests to select the best candidate SRPG solutions to be used later in larger scale 3-dimensional flow tank tests and to understand the effects of saturated media conditions on different types of SRPGs. Column tests were conducted in KONTES Chromaflex columns with dimensions (L × ID) of 120 cm × 4.8 cm (V = 2.17 L) and 17.5 cm × 5.7 cm (V = 0.45 L). Silica sand (sieve size = 60-100 mesh) served as porous media and was fully saturated with water to an estimated porosity of 0.3. Columns were oriented horizontally with the inflow end slightly elevated and the lid on the outflow end loosened to create a pressure gradient. A peristaltic pump (Masterflex L/S, Cole-Parmer, Vernon Hills, IL, USA) injected the SRPG solution into the column, and deionized water was then pumped into the column at an ambient flow rate of 0.6 mL/min. − ] and photographs were taken to document visual observations of the migration and release patterns. all-porous media, porous media with an intact clay lens, and porous media with a fractured clay lens. Clay lenses were~3 cm thick and~7 cm long with an underlying layer of beads 2 cm thick and top layer of beads~3 cm thick. A peristaltic pump supplied water to the inflow chamber to maintain a background flow rate of 0.6 mL/min throughout testing.

Small Flow Tank Tests in Porous and Low-Permeability Media
Injection and monitoring wells were constructed by shaping stainless steel screens into cylinders and filling them with fiberglass as a filter and Teflon tubing for solution delivery and sample extraction. The injection well was at the upstream end of the tank and~2 cm below the surface. Three multi-level monitoring wells were placed~15 cm downstream from the injection well, spaced~5 cm apart, and collected samples from depths of~3 cm and~6.5 cm. Figure 1A illustrates the set-up for the flow tank. Samples were collected approximately every 12 h for the first day of testing and then daily until MnO 4 − was depleted from the tank. [MnO 4 − ] was measured with the UV-Visible Spectrophotometer (λ = 525 nm) and photographs were taken throughout testing to record MnO 4 − release and migration. After a flow tank test completed, the clay was extracted to observe oxidation behavior and residual MnO 4 − concentrations.

Large Flow Tank Test in Porous Media
A large flow tank test was set up in a fashion similar to the small flow tank tests to monitor the migration and release of 150 mL of a 26 g/L N-40/B-50 mix SRPG in porous media in greater detail. A glass tank (L × W × H = 61.6 cm × 31.8 cm × 42.5 cm, V = 83.25 L) was divided into an inflow chamber, central chamber, and outflow chamber. The central chamber of the flow tank was filled with Mil-10 Ballotini industrial glass beads to a depth of~40 cm, length of~50 cm, and porosity of~0.4. After media were set, a peristaltic pump released water into the inflow chamber at a flow rate of 21.0 mL/min. Solution delivery was performed through an injection well~4.5 cm below the surface.
Multi-level monitoring wells extracted samples from depths of~10 cm,~25 cm, and 38 cm below the media surface. The monitoring wells were spaced~13 cm apart from any side and positioned in parallel rows as depicted in Figure 1B. The first central monitoring well was~11 cm from the injection well. Samples were collected approximately every 12 h and were then tested in the UV-Visible Spectrophotometer (λ = 525 nm) to assess [MnO 4 − ].  Figure 2 depicts this pattern. A two-stage increase in viscosity was also supported by viscosity measurements. For continuity, viscosity tests were performed on SRPG solutions at a [KMnO 4 ] of 20 g/L, and a gelation threshold was set at 1000 cP. Visual observations and gel state data [18] suggest that at this viscosity the SRPG has started to gel and should stop spreading in media. Viscosity tests showed that the 20 g/L B-40 and N-40 SRPGs both underwent lag periods > 2 h.  Table 1. Zeta potential becomes increasingly negative with increasing pH [24]. Around pH 8, the repulsion between particles becomes so strong that they cannot form siloxane bonds. Thus, gelation occurs most rapidly at a neutral to slightly acidic pH when the zeta potential is low and van der Waals attraction can overcome interparticle electrostatic repulsion [8,16]. As the L-40 CS had a lower pH, repulsion between particles was initially weaker. Addition of the salt K + reduced the zeta potential further, allowing for more interparticle interaction and the formation of siloxane bonds at a faster rate.

Results and Discussion
Particle size may have also influenced the rate of gelation. The L-40, B-40, and N-40 CS solutions had particle diameters of 22 nm, 14 nm, and 8 nm, respectively. In viscosity tests, the L-40 SRPG broke the gelation threshold first followed by the B-40 SRPG and finally the N-40 SRPG. Furthermore, the L-40 SRPG had the shortest max gelation lag time while the B-40 SRPG had an intermediate lag time, and the N-40 SRPG had the greatest lag time. This is consistent with observations that increasing particle size is associated with greater opportunities for interparticle interaction for the formation of siloxane bonds that cause gelation [8].

Column Flow-Through Test Results
Compared to batch tests, the overall gelation lag time for the SRPGs was delayed in column flow-through tests. Furthermore, some lower KMnO 4 concentrations were unable to achieve gelation. These outcomes were attributed to dilution effects due to hydrodynamic dispersion.
As Figure 3 shows, the B-40 SRPG solution was unable to achieve gelation at an initial   A summary of the column tests by maximum gel lag time can be found in Table 2

Small Flow Tank Test Results
Small-scale flow tank tests were conducted with two main purposes: to compare the 3-dimensional spread and release characteristics of a B-40 SRPG against a pure KMnO 4 solution of equal concentration and volume and to observe how different media conditions influence the migration and release of the SRPG. Based on batch and column test results, a B-40 SRPG containing 26 g/L KMnO 4 was chosen for its ability to gel in saturated media and for its ease of injection compared to SRPGs with a higher silica concentration. In addition, the B-40 SRPGs achieved higher gel lag times and could hold higher KMnO 4 concentrations than the 50 wt.% CS SRPGs.
In all-porous media, the CS in the solution nearly tripled the distance of lateral dispersion of the oxidant compared to a pure KMnO 4 solution as depicted in Figure 5  The KMnO 4 solution also demonstrated little potential for sustainable release in allporous media as observed in Figure 6A. For simplicity, measurements were averaged between the two depths at which samples could be drawn to focus on lateral spread. The  Figure 6B. As a result, the CS prolonged the release duration to 9.2 d. When placed in porous media with a clay lens, the SRPG was again able to spread more laterally compared to the pure KMnO 4 solution. Surficial observations suggested that the KMnO 4 solution had a lateral extent of~8.5 cm, but [MnO 4 − ] measurements were collected from all three monitoring wells which suggested a minimum lateral extent of 10 cm. With the SRPG, MnO 4 − and oxidation of the clay could be observed on both sides of the tank, indicating that MnO 4 − occupied the tank to the fullest lateral distance permitted,~20 cm.
Measurements of [MnO 4 − ] from monitoring wells suggest that the clay lens did limit MnO 4 − release for both the KMnO 4 solution and the SRPG. Under these media conditions, the two solutions behaved in a manner relatively similar to observations made in the all-porous media. Peak release occurred in the KMnO 4 solution sooner and with higher amounts of the oxidant released resulting in a shorter release duration than the SRPG. As Figure 6C,D demonstrate, the peak mean concentration released from the 26 g/L KMnO 4 solution was reduced from 1491 mg/L to 976 mg/L at the central well when CS was included in the solution. The presence of CS also delayed peak release in the KMnO 4 solution, this time from~21 h after injection to~24 h and elicited a more gradual release of MnO 4 − which extended the release duration from 31 h to 4.6 d. When comparing the SRPG under different media conditions, the presence of clay both contributed to a more dramatic decline in the amount of [MnO 4 − ] in the system and reduced the release duration of the oxidant. As Figure 6B,D,E illustrate, the SRPG in all-porous media experienced a peak release~27 h after injection but a peak release was observed~24 h after injection in the media with a clay lens and~21 h after injection when the clay lens was fractured. Similarly, the release duration of the SRPG was lowered from 9.2 d in the all-porous media to 4.6 d in the presence of the clay lens. When the clay was fractured, the release duration was reduced further to 38 h.
It was expected that MnO 4 − release could be prolonged by diffusion into the clay matrix and a secondary slow-release system could be developed in situ through counterdiffusion. The clay lens appeared to mediate release from the KMnO 4 solution to yield lower MnO 4 − concentrations and a longer release duration. However, the release duration was significantly reduced by the clay lens for the SRPG solutions.
One explanation for the diminished release duration of the SRPG solutions is that the natural oxidant demand (NOD) of the clay media resulted in some MnO 4 − loss. This could also explain the release behavior of the KMnO 4 solution. Rather than mediating MnO 4 − release, the clay lens hindered migration of the KMnO 4 solution over time due to its low permeability, explaining the higher release duration. Meanwhile, NOD led to a reaction between KMnO 4 and the clay to result in lower measurements of [MnO 4 − ]. Organic carbon is one of the largest contributors to NOD in clay materials. The reaction between organic carbon and KMnO 4 can be written as [25]: Marshall et al. [26] analyzed organic carbon content of Wyoming bentonite clay samples and found it composed 0.11 to 0.24 wt.% of the samples. Excavation of the clay lenses supported the notion that MnO 4 − had been consumed through NOD. A significant area of dark MnO 2 precipitates could be observed on the outer edges of the clay that extended~2 to 4 cm inward while no residual MnO 4 − was observed. This also helps explain why release duration was shorter in the fractured clay. Due to the fractures, MnO 4 − contact was not limited to the outer edges but could also infiltrate into the clay. This supplied more surface area for oxidation to occur. Oxidant consumption is greater with increasing initial [MnO 4 − ] [27]. As MnO 4 − was released slowly from the SRPG, the initial concentration in contact with the clay lens was lower than that from the KMnO 4 solution. Consequently, the SRPG solution achieved a much greater release duration than the KMnO 4 solution in media with a clay lens.

Large Flow Tank Test Results
A large flow tank test with a 26 g/L N-40/B-50 mix SRPG was performed to observe the general migration and release patterns of the SRPG on a larger scale. The test showed that overall MnO 4 − from the SRPG sank to the bottom of the tank and demonstrated a potential for vast lateral spreading of the oxidant as depicted in Figure 7. − ] values at the exterior monitoring wells which is indicative of a capability to disperse much farther laterally than the tank permitted ( Figure 7B2).

Seven days after injection, [MnO 4
− ] values had dropped significantly. Moderate concentrations resided deeper in the tank and near the injection point while low concentrations remained at intermediate depths ( Figure 7C1). Lateral spreading of MnO 4 − at this time was relatively uniform, with moderate concentrations contacting each monitoring well past the injection point ( Figure 7C2).

Conclusions
Depending on situational demands, either the 40 wt.% or 50 wt.% CS would make an effective gelling agent for the SRPG solution. SRPGs with a lower silica concentration may be preferred for their ability to hold greater amounts of KMnO 4 while still achieving significant gelation lag times in saturated media and for their slow gelation outside of media which makes low-pressure injection of high volumes of solution easier. However, 50 wt.% CS SRPGs such as the B-50 were found to be more advantageous due to their significantly longer release durations and resistance to dilution.
Flow tank tests demonstrated that regardless of silica wt.% or type of media, the addition of CS to a KMnO 4 solution enhanced lateral dispersion of the oxidant and mediated release to prolong the release duration. Contrary to some of the literature [9][10][11], counter-diffusion of MnO 4 − from a clay lens was not observed due to consumption of the small amount of MnO 4 − by NOD of the clay lens. The release duration of the SRPG was lowered from 9.2 d to 4.6 d in porous media with a clay lens. When the clay lens was fractured, the larger surface area contributed to greater MnO 4 − consumption, reducing the release duration to 38 h. Once the NOD of clay is met, however, the low-permeability media may contribute to prolonged release.
While the SRPG reached well beyond the benchmark release rate ≥850 µg/d, the high release of MnO 4 − and inability to achieve a gel lag time ≥3 d prevented a release duration ≥ 3 wk from being met. Although a desirable SRPG could not be produced in the time allotted for this study, much was learned about the behavior of different CS solutions as gelling agents, the release characteristics of the SRPG vs. pure KMnO 4 solution, and the effects of different media types on the SRPG. If more control can be exercised on the gelation and high release of MnO 4 − , the SRPG can serve as a novel treatment option for dilute or widespread DNAPL plumes. The increased lateral dispersion of the oxidant should facilitate mixing between MnO 4 − and the contaminant and reduce the number of injection wells needed for remediation. Results of this study warrant further investigations focused on assessing the remedial efficiency of the SRPG through more detailed and large-scale flow tank experiments and field applications and developing a numerical model to simulate the gelation, migration, and release patterns of the SRPG solution in heterogeneous saturated porous media. The boundaries of the flow tanks and small number of concentration data in this study did not permit full lateral dispersion of the oxidant, so more research dedicated to observing the migration of the SRPG in a larger space and deriving a dispersion coefficient is warranted.

Conflicts of Interest:
The authors declare no conflict of interest.