Per-and polyfluoroalkyl substance (PFAS) retention by colloidal activated carbon (CAC) using dynamic column experiments ☆

.

Several soil treatment approaches to counteract the PFAS soil contamination issue have become available in recent years.Often, they involve excavation of the soil and treatment ex-situ or disposal in landfills (dig and dump) (Igor et al., 2021).However, these approaches can be problematic because they entail high costs and result in shifting the contamination from one place to another (Igor et al., 2021;Ross et al., 2018).Conversely, in-situ soil remediation at the source zones has been described as a sustainable and cost-effective treatment approach (Høisaeter et al., 2021).In the case of PFAS, in-situ treatment includes destruction (e.g. chemical oxidation (Dombrowski et al., 2018), biodegradation (Huang and Jaffé, 2019)), removal (e.g.electrochemical removal (Niarchos et al., 2022;Sörengård et al., 2019), phytoremediation (Huang et al., 2021), soil flushing (Senevirathna et al., 2021)), and stabilisation or solidification (e.g. using biochar (Inyang and Dickenson, 2017), activated carbons (Sörengård et al., 2020), stabilisation/solidification (S/S) (Sörengård et al., 2021)) techniques.Stabilisation methods have been identified as the most mature and feasible techniques to-date (Ross et al., 2018).They typically involve the addition of fixation agents in the subsurface to prevent leaching from soils into groundwater.Stabilisation has also been shown to decrease the bioavailability of PFAS (Hearon et al., 2022), reducing their uptake in plants and other living organisms (Bolan et al., 2021).
Different materials have been used as fixation agents for PFAS treatment, such as natural clays and clay-biochar composites (Mukhopadhyay et al., 2021;Yao et al., 2014), various biochars (Kupryianchyk et al., 2016;Sørmo et al., 2021), and activated carbons (Du et al., 2014;Sorengard et al., 2019).Activated carbons come in various forms based on grain size, typically as granular (GAC, 0.4-1.2mm) or powdered activated carbon (PAC, <1 mm).Colloidal activated carbon (CAC) is a novel material consisting of very fine particles (1-2 μm) and, consequently, a very large surface area.It has the added benefit that carbon particles are suspended in solution and can be injected in the subsurface, thus exhibiting ease of application in the field (McGregor, 2018).
Sorption of PFAS to activated carbons and their partitioning behaviour depend on their diverse molecular structure (polymeric, nonpolymeric; neutral, anionic, cationic, zwitterionic; short-chain, longchain).The main mechanisms for PFAS sorption include hydrophobic and oleophobic effects, electrostatic interactions, and interfacial behaviours, such as inner-sphere complexes (Du et al., 2014).The perfluorocarbon chain length is reportedly one of the dominant factors for PFAS sorption, with long-chain PFAS referring to perfluoroalkyl carboxylic acids (PFCAs) with C ≥ 8, perfluoroalkyl sulfonates (PFSAs) with C ≥ 6, and all other PFAS with C ≥ 7 (Buck et al., 2011).Adsorption capacity reportedly increases with increasing perfluoroalkyl chain length (Campos Pereira et al., 2018;Higgins and Luthy, 2006;Sepulvado et al., 2011); therefore, short-chain PFAS have a higher susceptibility to leaching.PFAS partitioning studies have so far mainly been conducted through batch tests, i.e. shaking tests to reach adsorption equilibrium (Qian et al., 2017;Son et al., 2020;Sorengard et al., 2019;Sørmo et al., 2021).Recently, the sorption of perfluorooctanoic acid (PFOA) and perfluorooctane sulfonic acid (PFOS) to soils was compared between batch and column tests, with both methods producing similar results (Van Glubt et al., 2021).However, the reported advantages of column tests were recognised, including the simulation of more realistic conditions and their usefulness to detect secondary transport phenomena (Van Glubt et al., 2021).Additionally, batch tests are unsuitable for investigating desorption phenomena and long-term retention.Maizel et al. (2021) studied the leaching of over one hundred PFAS from AFFF-impacted surface soils using column experiments and identified nonideal adsorption behaviour, denoting the complexity of PFAS transport.Other column studies have also exhibited non-equilibrium transport (Guelfo et al., 2020) and the significance of infiltration rates (Høisaeter et al., 2019) in various soils.Still, the PFAS transport under CAC soil treatment remains largely unexplored.
In this context, the objective of this study was to investigate the transport behaviour of 12 PFAS in a soil treatment scenario with CAC, with the use of 1-D flow dynamic column experiments for the first time.The target PFAS included legacy perfluoroalkyl acids (PFAAs) and fluorotelomer sulfonates (FTSAs), which are highly relevant for AFFF-impacted sites.The sorption of PFAS to soil and CAC was examined, followed by a desorption phase to assess the reversibility of the process.Furthermore, the study aimed to increase the understanding of CAC's behaviour in the subsurface by investigating its washing-out potential and distribution along the soil column.

Soil sampling and preparation
Soil was sampled at a depth of approximately 3 m in central Sweden (59 • 23 ′ 13.7 ′′ N, 15 • 53 ′ 48.2 ′′ E).The sampling location was adjacent to a PFAS-contaminated site of interest.Based on initial screening, most of the target PFAS were below detection limits (Table S4 in SI).After sampling, the soil was air-dried for 10 days at 40 • C in a room sheltered from dust deposition, then ground with a mortar and pestle, sieved to remove large particles (>2 mm), and homogenised by shaking.The soil was a silt loam, with a relatively high amount of clay and fine silt particles, while it was low in organic carbon and PFAS (for details, see Tables S5 and S6 in SI).Before packing, the soil was mixed with premium #40/50 silica sand (AGSCO, USA) at 1:5 (field soil):(silica sand) ratio to increase the hydraulic conductivity of the columns.For the treated soil, CAC was added to the mixture at a final concentration of 0.031% w/w, based on manufacturer suggestions (Regenesis, personal communication, 2020).

Column design and packing
Column tests were conducted in custom-made transparent PVC pipes (soil chamber L = 15 cm, d = 3.6 cm) having a length to diameter ratio >4 to simulate field conditions and minimise transverse dispersivity (Banzhaf and Hebig, 2016;Gibert et al., 2012;Lewis and Sjöstrom, 2010).The flow was directed upward to facilitate the escape of entrapped air bubbles.Two chambers of gravel in the inlet and outlet of the columns were included to ensure uniform flow in the soil column (L = 2 cm, d = 3.6 cm), by increasing dispersion and avoiding flow convergence at the entrance and exit points.Nylon mesh (pore diameter = 50 μm) was placed at the inlet and outlet and secured in place with perforated PVC plates.
Packing of the columns aimed at a uniform matrix, saturated and free of air bubbles.In total, four column experiments were conducted, with treated (T1, T2) and reference soil (R1, R2) tested in duplicates (Table 1).Preparation of soil columns followed the guidelines of Lewis and Sjöstrom (2010); thin layers of dry soil were deposited into the column from a few centimetres height and soil was saturated from the inflow while mixing the soil and vibrating the column to remove entrapped air.It has been reported that this technique can produce the highest uniformity of packing (Lewis and Sjöstrom, 2010).In total, 306 ± 1.7 g of soil was packed in each column (n = 4).After packing, columns were equilibrated by pumping artificial groundwater through them for 48 h using a multichannel peristaltic pump (ISM931C, Simatic® IPC, Germany).

Tracer tests
Estimations of hydrological parameters in each column were conducted using tracer tests after equilibration of the columns and before switching to PFAS-spiked water (Table S7 in SI).Sodium chloride (NaCl) was chosen as a conservative tracer at a starting molar concentration of 0.43 M.After equilibration of the columns, the inflow was switched to the tracer solution, which was pumped for ~24 h before the influent changed back to artificial groundwater.The flow rate was monitored through the tracer test and was steadily 288 mL d − 1 .Effluent samples were taken every 50 min, starting at the same time as the tracer pulse, with the use of a fraction collector (CF-2 Fraction Collector, 115 V, Repligen), and the electrical conductivity of the aliquots was measured (Cond 3310 & TetraCon 325, WTW).Conductivity measurements were converted to molar concentrations of NaCl, using a 5-point calibration curve (R 2 > 0.99) (Figure S1 in SI).After sample collection and analysis, results from the tracer tests with NaCl were processed with the CXTFIT 2.0 code, embedded in Excel, which incorporates a solution of the 1-D convection-dispersion equation (CDE) as the transport model (Toride et al., 1995).

Preparation of feed solution and sample collection
Spiked artificial groundwater was prepared with solution chemistry resembling the sampling location's groundwater based on field measurements (Tables S8 and S9 in SI).Using the field data, a geochemical equilibrium model in VisualMINTEQ was developed to define the solution chemistry of the artificial feed water.The pH of the solution was adjusted to 7.7 with the addition of HCO 3 − , to simulate field conditions.
The salts added to the feed solution and their concentrations can be seen in Table S10 in SI.
The feed solution was spiked with a total of 12 PFAS, including C 4 -C 10 perfluorocarboxylic acids (PFCAs) (PFPeA, PFHxA, PFHpA, PFOA, PFNA, PFDA, PFUnDA), C 4 , C 6 , C 8 perfluorosulfonic acids (PFSAs) (PFBS, PFHxS, PFOS), and 6:2 and 8:2 fluorotelomer sulfonic acids (FTSAs) (6:2 FTSA, 8:2 FTSA).The final total concentration was 0.87 mg L − 1 for ∑ 12 PFAS, evenly distributed among individual PFAS (C = 67 μg L − 1 for individual PFAS), corresponding to the high end of environmentally relevant concentrations for contaminated soils (Brusseau et al., 2020).The spiked feed solution was pumped through the columns at a flow rate of Q = 288 mL d − 1 , ensuring a continuous upward flow and keeping saturated conditions.After approximately 400 pore volumes (PV), the inflow was switched from spiked to non-contaminated artificial groundwater to test the desorption potential of PFAS from the soil.The duration of the whole experiment was approximately 150 days, or 596 PV.PFAS concentrations were quantified for water outflow samples, as well as in segments of the soil columns.Sacrificial samples of the outflow water were taken at various times during the duration of the experiments (approximately every three days).After completion of the experiment, the soil columns were cut into five layers, to analyse soil concentration of PFAS for mass balance calculation.
Water samples were directly injected after methanol and IS addition, centrifugation (3000 rpm, 15 min), and filtration through a 0.45 μm recycled cellulose (RC) syringe filter fitted directly onto the syringe, producing aliquots of 1:1 methanol:water.Soil samples of 3 g were extracted twice with methanol and sodium hydroxide (NaOH), centrifuged, concentrated with N 2 , and filtered through a 0.45 μm RC filter.
The IS in soil samples was added before solvent extraction, to account for extraction losses.Negative blanks (n = 2) and positive control samples (n = 2) were included in the analysis.The positive controls were used as a reference to starting concentrations.The average recovery of ISs was 104% ± 5.2% for the water samples and 91% ± 23% for the soil samples (Tables S11 and S12 in SI).The method detection limit (MDL) for each target PFAS was calculated as MDL = mean blanks + 3*σ blanks when concentrations were found in the method negative blanks, or using the lowest point of a 9-point calibration curve (0.01-100 ng mL − 1 ) in the absence of blank concentrations (Table S13 in SI).The positive control analysis showed a deviation of − 9.60% from the theoretical spiked PFAS concentrations, which can be attributed to losses in the preparation and storage of spiking solutions or sorption of PFAS to the tube walls and other equipment that the inlet solution came in contact with.

Organic carbon analysis
Organic carbon content was determined for reference and treated soil along the five layers of the soil columns to evaluate the redistribution and washing out of CAC in the columns.Soil samples of each layer were analysed for each column (Table S14 in SI).The carbon content and nitrogen content in the soils were determined through dry combustion according to ISO10694 (1995) and ISO 13878 (1998), respectively, using an elemental analyser for macrosamples (TruMac® CN, Leco corp, S:t Joseph, MI, USA).In short, the samples were combusted in a stream of pure O 2 at a temperature of 1350 • C, converting all carbon and nitrogen present in the sample to CO 2 and NO x , respectively.In two separate runs, each sample's total carbon (TC) and total inorganic carbon (TIC) contents were determined.The total organic carbon (TOC) content was then calculated as TOC (%) = TC (%) -TIC (%).

Tracer tests
Breakthrough curves (BTCs) of experimental and modelled data are shown in Fig. 1.The shape of the BTCs and the calculated parameters were indicative of uniform packing of all columns.A complete breakthrough of NaCl was noticed slightly before 2 PV for all columns.The BTCs showed a relatively symmetrical shape, denoting an equilibrium G. Niarchos et al. state in all the columns.Hydrological parameters were calculated based on the CXTFIT code and are presented in Table S7 in SI.Longitudinal dispersivities were estimated to be λ = 2.72 ± 0.87 cm (n = 4).Peclet numbers were estimated to be less than 10 for all columns.The curves' fit had an average R 2 = 0.979 ± 0.0041 (n = 4).The average pore water velocity was estimated by taking the time to reach C/C 0 = 50% for the tracer, as v = 2.97 ± 0.07 cm hr − 1 (n = 4).

Column experiments
The leaching behaviour of PFAS differed between CAC-treated and reference soil columns (Figs. 2 and 3).Specifically, the breakthrough in treated soil columns (T1, T2) appeared much slower for many PFAS (Fig. 3) compared to reference soil columns (Fig. 2).Complete breakthrough for nearly all PFAS apart from PFNA and PFUnDA was noticed much faster in the reference soil (~30 PV) than in the treated soil (~233 PV), indicating an approximately eight times higher retardation for ∑ PFAS in the treated soil.The tail of the BTCs showed similar behaviour for most PFAS.However, the drop of concentrations appeared steeper for the treated soil columns, indicating stronger adherence of PFAS to CAC than soil (Figure S2 in SI).In some instances, and especially in the starting samples for columns R1 and R2 (Fig. 2), effluent concentrations of PFAS were higher than the influent concentrations, resulting in C/C 0 > 1, thus showing an "overshoot" of concentrations.This effect has been previously described by other studies (McCleaf et al., 2017;Park et al., 2020) and can be attributed to chromatographic effects (i.e.PFAS leaching out at different time points) and the replacement of PFAS stored in the column when more hydrophobic PFAS (typically longer chain) arrive and compete for adsorption sites.Overshoot was more noticeable in PFCAs than PFSAs, confirming this theory, as PFCAs have lower hydrophobicity than their homologues (Higgins and Luthy, 2006).This is important to consider in field-scale applications since long-chain PFAS plumes can transport more slowly and thus induce a release of short-chain PFAS if they arrive at the treatment zone at a slower pace and compete for sorption sites.
Marked differences were observed for individual PFAS depending principally on their perfluorocarbon chain length (BTCs for ∑ 12 PFAS and some individual PFAS can be seen in Figures S3 and S4 in SI).In the reference soil, an almost complete breakthrough (90%) was noticed immediately at the first sample (~3.5 PV) for C < 8 chain length PFAS, while PFUnDA (C = 10) exhibited the highest retardation, with a complete breakthrough at ~42.2 PV.PFOA showed a complete breakthrough at ~149 PV in treated soil, compared to ~30 PV in reference soil.In treated soil, some long-chain PFAS showed no complete breakthrough.Specifically, the longest chain PFCAs, PFDA and PFUnDA, had a maximum of 86% and 65% breakthrough, respectively, after 422 PV.For PFUnDA, there was also a delay in reaching maximum concentrations, with maximum concentrations being reached at 22 PV after the rest of the PFAS and while concentrations for the rest of PFAS were decreasing (Fig. 3).Therefore, there can be an increase in concentrations of long-chain PFAS even after switching to non-contaminated groundwater, indicating a delayed response to desorption.This phenomenon can be attributed to the lower transport rate of long-chain PFAS due to their lower solubility or to competition effects among PFAS for sorption sites, which can increase available sites for longer-chain PFAS when short-chain PFAS are not present in the solution (McCleaf et al., 2017).Micelle breakdown could also lead to increased concentrations for long-chain PFAS in the dissolved phase, however, the concentrations tested in this study were well below the critical micelle concentrations (Table S3 in SI) and micelle formation was unlikely.Conversely, short-chain PFAS exhibited a much faster breakthrough and a steeper decrease in concentrations after switching to non-contaminated groundwater (~400 PV).In field-scale scenarios flowrate fluctuations would also occur (e.g.due to heavy rainfall), resulting in desorption of shorter chain PFAS, as has been observed by Maizel et al. (2021), which would likely accentuate the increase of their concentration in the aqueous phase.The faster breakthrough of short-chain PFAS is important to consider in future treatment applications, due to the production Comparing functional groups, PFCAs had slightly faster breakthroughs than both PFSAs and FTSAs, indicating stronger sorption for PFSAs and FTSAs than PFCAs.A similar trend has been noticed by Sorengard et al. (2019), who calculated K d values for CAC based on batch tests.Differences based on the functional group were more noticeable in long-chain PFAS.Specifically, PFOS had a complete breakthrough at 400 PV, much later than its analogous PFNA, which showed a complete breakthrough at 234 PV.Stronger sorption of PFSAs than PFCAs has similarly been reported before for sorption to soils (Campos Pereira et al., 2018) and activated carbons (Park et al., 2020;Son et al., 2020).This can be explained by the stronger hydrophobicity of PFSAs than their analogous PFCAs due to the larger size of the sulphonate moiety (Higgins and Luthy, 2006).Higher experimental variability was noticed for PFCAs in treated soil, especially for short-chain ones (Fig. 3).This could be attributed to competition effects among PFAS, especially when the more hydrophobic longer-chain PFAS arrive and compete for sorption sites.Despite these experimental variations, the aforementioned trends are still noticeable.

Partitioning coefficients
Retardation of PFAS was calculated by comparing the BTCs of the tracer (Fig. 1) to the BTCs for individual PFAS (Figs. 2 and 3).Due to the fast breakthrough of some PFAS, a more accurate estimation of the retardation could be obtained by comparing the times for 80% breakthrough for PFAS to that of the tracer, as compared to doing this for 50% breakthrough times.Therefore, the time of 80% breakthrough was chosen, and the retardation factors were determined from R f = t 80.NaCl t80.PFAS .Assuming linear adsorption, K d values could subsequently be calculated as K d = (Rf − 1)*n eff / ρ , where Rf is the retardation factor calculated by the time required to reach C/C 0 = 0.80, n eff is the effective porosity, and ρ is the bulk density of the packed soil (USEPA, 1999).The K d coefficients ranged from ~10 3 to 10 5 L kg − 1 , with a positive correlation between perfluorocarbon chain length and K d for PFCAs, PFSAs, and FTSAs alike, increasing one order of magnitude from C 4 to C 10 PFCAs (Fig. 4).The log K d increased by 0.26, 0.32 and 0.20 log units for each CF 3 moiety for PFCAs, PFSAs and FTSAs, respectively.Comparing functional groups and the same perfluorocarbon chain length, the K d for PFSAs was on average 0.05 log units higher compared to PFCAs (e.g.PFNA had in SI).For those that K d could be determined, values for CAC appeared to be 3-5 orders of magnitude higher than for the non-treated soil, which is in agreement when comparing with values reported in the literature for G. Niarchos et al. natural soils (Milinovic et al., 2015).This indicates much stronger sorption to CAC than to soil, which can be attributed to the material's high surface area (900 m 2 g − 1 ) and small particle size (1-2 μm).The strong sorption despite the low ζ potential values of CAC (− 80 to − 2 mV) indicates that hydrophobic interactions likely govern the adsorption.This phenomenon has also been described by Du et al. (2014), who found it more pronounced for long-chain PFAS.However, interactions based on electrostatic effects cannot be excluded.Hansen et al. (2010) reported K d values for GAC and PAC based on shaking tests, on the order of 10 3 -10 5 for PAC and 10-10 2 for GAC, therefore the values calculated herein appear to be comparable to PAC and higher than GAC.The materials tested by Hansen et al. (2010) had similar iodine numbers and surface area, however, they consisted of larger particles (particle size 80% < 45 μm for PAC and 226 μm for GAC), which likely explains the noted differences.Specifically, a smaller particle size would lead to a larger micropore surface area, which can increase the adsorption capacity (Park et al., 2020).Sorengard et al. (2019) also derived K d values for CAC-treated soil, which ranged between approximately 2-3.5 log units.The K d values in this study are higher since they are based only on the CAC fraction of the treated soil (Fig. 4).

PFAS mass balancesoil concentrations
According to the PFAS analysis in the soil sections, treated soil columns retained a total of 13,000 μg kg − 1 for ∑ PFAS, versus 38 μg kg − 1 for ∑ PFAS for the reference soil, corresponding to 99.7% higher retention of PFAS in the treated soil columns after completion of the experiments.Additionally, the PFAS mass in the CAC-treated soil corresponds to 37 ± 5.04% of the total mass of ∑ PFAS that was flushed through the columns during the experimental duration, indicating 63 ± 5.04% of ∑ PFAS leached through the columns T1 and T2.For the treated soil columns (T1, T2), the total mass of PFAS was primarily distributed within 6-15 cm from the inlet (82% ± 4.9%) and 33% ± 0.75% was concentrated at the last 3 cm of the column.The opposite was noticed for the reference columns (R1, R2), where they were mainly distributed close to the entrance point of the columns (42 ± 7.4% at 0-3 cm) (Fig. 5).Looking at the individual PFAS distribution in the treated soil (Figure S4 in SI), the same trend was noticed for most compounds, apart from PFUnDA, which mainly accumulated in the middle, and 6:2 FTSA, which was evenly distributed.In the same graph, it is also notable that the compound that accumulated the most is PFOS, along with long-chain PFCAs, i.e.PFUnDA, PFDA, and PFNA, while short-chain PFAS are present at much lower, almost undetectable concentrations.This is in accordance with the calculated K d values, which show that the adsorption coefficient is similar for these four PFAS (Fig. 4).The challenges for stabilisation of short-chain PFAS have been exhibited in other stabilisation applications, such as in pilot-scale S/S treatment (Sörengård et al., 2021) and treatment of drinking water with GAC filters (McCleaf et al., 2017).The effect of higher concentrations in the outlet for columns T1 and T2 was attributed to the transport of CAC within the soil matrix during the experiments, which is discussed in Section 3.5.Specifically, the low PFAS mass close to the inlet of the treated columns was expected to be due to depletion of CAC from the inlet and its deposition close to the outlet.In the reference soil columns (R1, R2), the cumulative mass gives a mechanistic understanding of PFAS fate in soils, indicating that they will transition to the next available site when available sorption sites are occupied and that the adsorption kinetics are relatively fast, considering the high flowrates of the experiments.

CAC distribution
During the equilibration of columns at the beginning of the experiments, CAC washout was noticed from the treated soil columns (T1, T2).This washout effect was quantified using TOC content analysis to estimate the lost carbon mass in treated soil (Fig. 6, Table S14 in SI).According to the TOC analysis, it was estimated that the treatment concentration of CAC was 0.031% in the treated soil.Further, the results suggested that 22% ± 8.4% of added CAC was eluted from the columns during the experiments, resulting in a final concentration of 0.024% CAC dry weight.It is worth noting that the transport of carbon particles was likely favoured by the relatively high pore water velocity and the short length of the columns.Lower velocities that would be found in realistic scenarios could enhance filtration of CAC, therefore leading to more carbon being attached to soil particles.
Complete depletion of CAC was noticed in the soil closer to the inlet (0-3 cm), whereas accumulation of CAC was detected in the soil closer to the outlet (12-15 cm from the inlet) (Fig. 6).This validated the results of the PFAS mass balance that indicated higher concentrations close to the outlet for columns T1, and T2 (Fig. 5).The calculations in this study, e.g. for K d values, were based on the initial concentration of 0.031% since the timing of the CAC elution was uncertain.

Conclusions
The effectiveness of using colloidal activated carbon for PFAS soil treatment was demonstrated through column tests.Treatment of soil with an initial concentration of 0.031% w/w CAC resulted in an average of 8 times higher retardation of ∑ PFAS transport compared to reference soil, despite 22% of CAC washing out during the experiments.For some long-chain PFAS (PFDA, PFUnDA), retardation was much higher, and breakthrough was never reached in treated soil during the experimental duration.Treatment efficiency was positively correlated with chain   length, with long-chain PFAS exhibiting higher K d values and functional group (PFSAs > PFCAs).The K d values for CAC were in the range of 10 3 -10 5 L kg − 1 , and thus 3-5 orders of magnitude higher than the ones reported in the literature for soil.The mass balance analysis indicated a 37% retention of ∑ PFAS of the total amount that was flushed through the treated columns.CAC washout was noticed during the experiments and, based on carbon analysis, was estimated to be 22% of the total carbon added to system.A certain degree of CAC mobility is considered necessary, to make sure it is well-distributed through all the flow channels of a soil matrix; however, CAC washout can also result in less effective treatment and even transport of PFAS to downstream locations.Further studies are therefore needed to fully understand the fate and transport of CAC in the subsurface.
Based on these data, treatment with CAC can result in significant retardation of PFAS, especially long-chain ones, and hinder their transport in the subsurface.While the treatment is not expected to last forever, it offers a promising solution when it comes to stabilisation efforts, chiefly due to its ease of application through injection in the subsurface.Re-injection of CAC post-treatment might be considered necessary when breakthrough is noticed in groundwater.Future research should focus on investigating further the transport of CAC and developing methods to improve its stability, avoiding elution of the material from the treated soil.Lastly, future experiments could focus on simulating more realistic conditions, for instance with the use of flowinterruption experiments.

Fig. 2 .
Fig. 2. Column breakthrough curves (BTCs) for a) PFCAs, b) PFSAs, and c) FTSAs in reference soil (columns R1, R2).C 0 corresponds to positive control concentrations.Error bars indicate methodological error as the standard deviation (n = 2).The red dashed line represents the time of switching to non-PFAS-spiked artificial groundwater.(For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 3 .Fig. 4 .
Fig. 3. Column breakthrough curves (BTCs) for a) PFCAs, b) PFSAs, and c) FTSAs in CAC-treated soil (columns T1, T2).C 0 corresponds to positive control concentrations.Error bars indicate methodological error as the standard deviation (n = 2) The red dashed line represents the time of switching to non-PFAS-spiked artificial groundwater.(For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 6 .
Fig. 6.Organic carbon distribution in the treated soil layers after completion of the experiments.

G
.Niarchos et al.

Table 1
Overview of experiments and operating conditions.