Transport of PFOS in Aquifer Sediment: Transport Behavior and a Distributed-Sorption Model

Abstract

The objectives of this research were to examine the transport of perfluorooctane sulfonic acid (PFOS) in aquifer sediment comprising different geochemical properties, and to compare the behavior to that observed for PFOS transport in soil and sand. PFOS retardation was relatively low for transport in all media. The PFOS breakthrough curves were asymmetrical and exhibited extensive concentration tailing, indicating that sorption/desorption was significantly nonideal. The results of model simulations indicated that rate-limited sorption/desorption was the primary cause of the nonideal PFOS transport. Comparison of PFOS transport in aquifer media to data reported for PFOS transport in two soils and a quartz sand showed that PFOS exhibited more extensive elution tailing for the soils, likely reflecting differences in the relative contributions of various media constituents to sorption. A three-component distributed-sorption model was developed that accounted for contributions from soil organic carbon, metal oxides, and silt+clay fraction. The model produced very good predictions of K_d for the five media with lower soil organic-carbon contents (≤0.1%). Soil organic carbon was estimated to contribute 19–42% of the total sorption for all media except the sand, to which it contributed ~100%. The contribution of silt+clay ranged from 51–80% for all media except the sand. The only medium for which the contribution of metal-oxides was significant is Hanford, with an estimated contribution of 15%. Overall, the results of the study indicate that sorption of PFOS by these aquifer media comprised contributions from multiple soil constituents.

Graphical Abstract

1. Introduction

Per- and polyfluoroalkyl substances (PFAS) have been detected in the atmosphere, surface water, sediments, soil, groundwater, and wastewater globally due to their use in a wide range of consumer products, manufacturing processes, and fire-fighting foams. Their presence in soil and groundwater is of particular concern with respect to potential human-health risks via multiple exposure pathways. A sound understanding of the retention and transport behavior of PFAS in the subsurface is necessary for accurate characterization of risk and the design of effective remediation methods. The standard approach for investigating retention and transport behavior in the laboratory is to conduct miscible-displacement column experiments. To date, relatively few such transport studies have been conducted for PFAS.

McKenzie et al. (2015) investigated the transport of a multicomponent PFAS solution containing perfluorooctanoic acid (PFOA), perfluorooctane sulfonic acid (PFOS), and 9 other PFAS in a loamy sand. Brusseau and colleagues (Lyu et al., 2018; Brusseau et al., 2019a; Brusseau et al., 2019b; Brusseau, 2020; Lyu and Brusseau, 2020; Van Glubt et al., 2021a, 2021b) investigated the transport of PFOA and PFOS in two natural quartz sands and two soils. Aly et al. (2018) conducted column experiments for PFOS and PFOA transport in a natural quartz sand. The transport of PFOS, PFOA, and 4 other PFAS in a soil was measured individually by Aly et al. (2019). The transport of PFOA in 3 soils was investigated by Lyu et al. (2019). Guelfo et al. (2020) investigated the transport of a multicomponent PFAS solution containing PFOA, PFOS, and 8 other PFAS in a natural quartz sand and 3 soils. Yan et al. (2020) investigated the transport of GenX in a natural quartz sand and 4 soils.

In aggregate, the studies cited above include 14 different sands or soils. Conversely, there have been no reports of experiments investigating PFAS transport specifically in aquifer media. It is well recognized that the physical and geochemical properties of soils and aquifer media vary greatly. For example, the majority of the soils used in the studies cited above have relatively high organic carbon (OC) contents, with only two of the soils having OC ≤ 0.1%. Conversely, many aquifer media typically contain very low organic-carbon contents. Due to the presence of both ionic and nonpolar functional groups, it is anticipated that sorption of PFAS may be more complex for aquifer media in that metal oxides, clay minerals, and other constituents may exert a more prominent impact on sorption compared to systems with higher OC contents. Furthermore, pure quartz sand is generally not representative of actual aquifer media. Given the great concern with PFAS contamination of groundwater and the need to understand transport behavior to characterize contaminant-plume mobility for risk assessments and mitigation, it is critical to investigate the transport behavior of PFAS in aquifer media. As the prior laboratory-based transport studies conducted for PFAS have employed soils or quartz sands, and not aquifer materials, there is clearly a need for experiments conducted specifically with aquifer media.

The objective of this study is to investigate the transport behavior of PFOS, the selected representative PFAS, in aquifer media. The contributions of soil constituents, including soil organic carbon, metal oxides, and the silt and clay fraction, to the magnitude of sorption is examined. Miscible-displacement experiments are conducted to measure the sorption and transport of PFOS in three aquifer materials with different geochemical properties. Additional experiments are conducted with media treated to remove soil organic carbon and at different input concentrations. Particular attention is placed on extended elution tailing of breakthrough curves. A mathematical model incorporating nonlinear sorption/desorption kinetics is used to simulate the measured data sets. The results of the experiments are compared to data reported in prior studies for PFOS and PFOA transport in soils and sands. This study is to our knowledge the first to focus specifically on transport of PFAS in aquifer media.

2. Materials and Methods

2.1. Materials

Three aquifer media of similar median grain diameter but different geochemical properties were used--- AFP44, Borden, and Hanford. Select properties of the media are presented in Table 1, including metal-oxide (Fe, Mn, Al) and organic-carbon contents. The AFP44 aquifer medium was collected from a location within the Tucson International Airport Area Superfund site in Tucson, AZ. It has low organic carbon and metal-oxide contents. The Borden aquifer medium was collected from the Canadian AFB Ontario, which is a well-known site for contaminant hydrology field studies. It also has low organic carbon and metal-oxide contents, but an appreciable carbonate content. The Hanford aquifer medium was collected from the Hanford, WA Department of Energy facility. It has a relatively high metal-oxide content, with Fe-oxides comprising ~73% of the total, and low organic carbon. No known PFAS contamination is present at the sample-collection locations. Data reported for PFOS transport in Eustis soil, Vinton soil, and a natural quartz sand (Brusseau et al., 2019a, 2019b; Brusseau, 2020; Van Glubt et al., 2021a) are used for comparison.

Table 1.

Properties of porous media.

Medium	TOCa (%)	Fe-Mn-Aloxides (mg/kg)	Sand-Silt-Clay (%)	Median Grain Diameter (mm)	Bulk Density (g/cm3)	Porosity (−)
AFP44	0.06	<100	97.5–1.3–1.2	0.33	1.79	0.34
Borden	0.03	~500b	96.2–2.0–1.8	0.21	1.80	0.30
Hanford	0.06	19960	97.3–1.6–1.1	0.25	1.78	0.38
Sand	0.04	28	100	0.35	1.63	0.35

^aTOC = total organic carbon

^bdata from Reinhard et al., 1990.

PFOS (CAS# 1763–23-1, 98% purity) was purchased from Sigma-Aldrich. Pentafluorobenzoic acid (PFBA) (99%, Strem Chemicals) was used as the nonreactive tracer. This tracer was selected in part because it has a very similar aqueous diffusion coefficient to that of PFOS, which provides a more robust characterization of diffusive mass-transfer processes specific to PFOS (Brusseau et al., 2019b). A synthetic groundwater solution was prepared by dissolving select salts into distilled, deionized water as the background electrolyte. Major ions are listed in Table 2. The pH and ionic strength (IS) are 7.7 and 0.01 M, respectively. The columns used for the experiments are constructed of acrylic or stainless steel to minimize interactions with PFOS, with a length of 7–20 cm and an inner diameter of 2–2.5 cm.

Table 2.

Constituents of synthetic groundwater.

Major cations	Concentration (mg/L)
Na +1	50
Ca +2	36
Mg +2	25

Major anions	Concentration (mg/L)
NO3 −1	6
Cl −1	60
CO3 −2 /HCO3 −1	133
SO4 −2	99

2.2. Miscible-displacement Experiments

Several sets of miscible-displacement experiments were conducted (see Table 3). The columns were packed to obtain uniform bulk densities. Synthetic groundwater solution was injected into the column from the bottom at a low flow rate until complete water saturation. A solution containing the nonreactive tracer, pentafluorobenzoic acid (120 mg/L, dissolved in synthetic groundwater), was then injected into the column for 3–4 pore volumes, followed by another ~10 pore volumes of synthetic groundwater solution. After this, ~20 pore volumes of PFOS solution (PFOS dissolved in synthetic groundwater) was injected into the column, followed by another ~50 pore volumes of synthetic groundwater solution. Effluent samples were continuously collected using a fraction collector. Each sample was weighed prior to analysis to determine volumetric discharge. Experiments were conducted with two nominal PFOS input concentrations, 10 mg/L and 100 μg/L. The higher input concentration was selected to represent the higher range of concentrations observed in groundwater in proximity to aqueous film forming foam source areas (e.g., Anderson et al., 2019; Brusseau et al., 2020). Replicate experiments were conducted for Borden and Hanford media.

Table 3.

PFOS retardation factors (R) and sorption coefficients (K_d) measured with miscible-displacement column experiments from this and prior studies.

Experiment	Medium	Input Concentration (mg/L)	Retardation Factor	Column Kd (cm3/g)	Log Koc
1	AFP44	10	2.2	0.23	2.6
2	Borden	10	2.0	0.18	2.8
3	Bordena	10	2.0	0.14	2.7
4	Borden	0.1	5.8	0.89	3.5
5	Hanford	10	4.1	0.64	3.0
6	Hanforda	10	4.0	0.64	3.0
7	Hanford	0.04	6.0	0.99	3.2
8	Hanford (treated)b	10	1.6	0.12	-
-	Sandc	10		0.10	2.4
-	Sandd	0.1		0.23	2.8
-	Vinton soile	0.1		1.47	3.2
-	Eustis soile	0.1		1.86	2.7

^aReplicate experiment

^bMedia treated to remove soil organic carbon

^cData from Brusseau et al., 2019b and Brusseau, 2020

^dData from Van Glubt et al., 2021a

^eData from Brusseau et al., 2019b and Van Glubt et al., 2021a

An additional experiment was conducted for the Hanford medium to further investigate sorption behavior. For this experiment, the Hanford medium was treated to remove all organic carbon. This was done using a standard method of H₂O₂ digestion at elevated temperature. The experiment was conducted with an input concentration of 10 mg/L to match the baseline high-concentration experiments. The impact of the treatment on soil properties was tested by comparing measurements for treated and non-treated samples. The OC content was reduced to below detection. The metal-oxide content was reduced by 20%, with most of the reduction observed for Fe-oxide. Potential impacts on clay minerals and other constituents were not assessed. There was no measurable impact of the treatment on bulk density or porosity.

Additional details of the experiment methods are provided in Brusseau et al. (2019b). These methods have been shown to produce robust characterization and quantification of PFAS sorption and transport (Brusseau et al., 2019b; Van Glubt et al., 2021a).

2.3. Analytical Methods

The samples were analyzed for pentafluorobenzoic acid using ultraviolet-visible spectrophotometry (model 1601, Shimadzu, Corp). This method has a quantitative detection limit (QDL) of approximately 1 mg/L. Two methods were employed to analyze PFOS. The methylene blue active substances (MBAS) assay (Chitikela et al., 1995) was the first method, which was used for analysis of the higher-concentration samples for AFP44 and Borden experiments. This standard method has been used successfully for analysis of single-solute PFAS (Brusseau et al., 2019a, b). The QDL is ~0.4 mg/L. High performance liquid chromatography tandem mass spectrometry was the second analytical method, used for the lower-concentration samples for AFP44 and Borden and for all Hanford experiments. The system comprised a Waters Alliance 2695 LC coupled to a Micromass Quattro Ultima Triple Quadrupole MS system, using negative electrospray ionization (ESI-). The QDL is <0.05 μg/L. Relevant QA/QC methods were used. The two methods have been demonstrated to provide consistent results (Brusseau et al., 2019b). Additional details of the methods are available in Brusseau et al. (2019a, 2019b).

2.4. Data Analysis

The results of the miscible-displacement experiments were used to construct breakthrough curves. The effluent concentrations were divided by the input concentration to determine relative concentration (C/C₀). The cumulative volume of solution discharged for each effluent sample was divided by the resident pore volume of the column to determine pore volumes discharged. The retardation factor (R) and equilibrium sorption coefficient (K_d) were calculated by moment analysis of the breakthrough curves (e.g., Brusseau et al., 2019b; Van Glubt et al., 2021a). The organic-carbon normalized sorption coefficients (K_oc) were determined as K_oc = K_d/f_oc, where f_oc is the fraction of soil organic carbon.

A one-dimensional advection-dispersion model incorporating nonlinear, rate-limited sorption was used to simulate the breakthrough curves obtained from the experiments. Sorption kinetics is represented with the standard two-domain approach, wherein sorption is assumed to reach equilibrium instantaneously for part of the domain and to be rate-limited for the other domain. The Freundlich isotherm is used to describe nonlinear sorption.

The model requires input for five nondimensional parameters. The Peclet numbers (and associated dispersivities) were obtained from analysis of the nonreactive tracer breakthrough curves. The retardation factors were obtained from moment analysis as noted above. The volumes of the input pulses were measured. Two non-dimensional parameters were optimized during the curve-fitting of the model to the measured breakthrough curves. These are β, the fraction of retardation that is effectively instantaneous, and ω, the nondimensional sorption rate constant (a Damkohler Number). Details of the equations, solution, and parameter determination are presented in our prior works (Brusseau et al., 2019b; Brusseau, 2020).

3. Results and Discussion

3.1. Transport Results

The breakthrough curves for the nonreactive tracer are ideal, with sharp arrival and elution fronts, and retardation factors of 1, indicating that the columns were homogeneously packed and water flow was uniform. Breakthrough curves for PFOS transport in the three aquifer materials– AFP44, Borden, and Hanford are displayed in Figures 1–3, respectively. The curves are asymmetrical, with extended concentration tailing for both arrival and elution fronts. Effluent mass recoveries for PFOS averaged 99.3%. This indicates no measurable mass loss via transformation or irreversible sorption.

Figure 1.

Measured and simulated breakthrough curves for transport of PFOS in AFP44 aquifer material. The simulations are produced using the two-domain sorption-kinetics model, including both linear (LS) and nonlinear (NLS) sorption. Optimized parameter values: β = 0.7, ω = 0.2.

Figure 3.

Measured and simulated breakthrough curves for transport of PFOS in Hanford aquifer material. The simulations are produced using the two-domain sorption-kinetics model. Optimized parameter values: β = 0.45, ω = 1.1.

The extended elution-front tailing is illustrated in Figures 4 and 5. The pore volume (x-axis) for Figure 5 is normalized by the respective retardation factor for each solute and medium to eliminate the effect of differences in the magnitudes of retardation on the extent of tailing. Inspection of Figure 5 shows that the extents of elution-front tailing for PFOS in all three media are much greater than that observed for the nonreactive tracer. Furthermore, the minimal tailing observed for the nonreactive tracer indicates the absence of significant nonideal transport associated with physical heterogeneity factors (e.g., diffusive mass transfer associated with immobile water domains). These results indicate that the nonideal transport observed for PFOS is caused by some combination of nonlinear and rate-limited sorption/desorption.

Figure 4.

Comparison of PFOS breakthrough curves for transport in untreated versus treated Hanford aquifer material. The treatment removed all measurable soil organic carbon. The same input concentration was used for both experiments.

Figure 5.

Normalized elution fronts for PFOS transport in six porous media. Also included is a representative data set for the nonreactive tracer (NRT); data reported for AFP44. The elution pore volumes are divided by the respective retardation factors for each medium.

The retardation factors and K_d values determined from moment analysis of the breakthrough curves are presented in Table 3. The values reported for the replicate experiments are very similar, indicating good replication. Inspection of Table 3 reveals that the retardation factors and K_d values measured for the lower-C₀ experiments are larger than for the higher-C₀ experiments. This disparity indicates that sorption is nonlinear.

The K_d values for all three aquifer media are larger than the values reported for PFOS sorption by a quartz sand in our previous study (Table 3). Inspection of Table 1 reveals that the four media have similar organic-carbon contents. Conversely, the three aquifer media have measurable metal-oxide and silt+clay contents compared to negligible amounts for the sand. These results suggest that inorganic soil constituents may be contributing substantially to PFOS sorption for the aquifer media.

To test this hypothesis, the Hanford medium was treated to remove all organic carbon. A column experiment was then conducted to measure R and K_d (experiment 8 in Table 3). The K_d obtained for this experiment is 0.12, which is ~5 times lower than the value obtained from the transport experiments (# 5 and 6) for the untreated medium (0.64). Note that all three experiments were conducted with the same input concentration. These results indicate that the inorganic constituents do provide some degree of sorption of PFOS. They also indicate that soil organic carbon provides a significant contribution overall. Assessing the actual contributions of the organic and inorganic components based on these limited results is problematic for multiple reasons. First, H₂O₂ treatment is known to impact clay minerals and other inorganic soil constituents (e.g., Mikutta et al.,, 2005). Second, the treatment was observed to reduce the metal-oxide content by 20%. Alternations to the amount and/or nature of the inorganic constituents may impact their sorption capacity for PFOS, which would affect their relative contributions.

The simulations produced with the transport model provide reasonable fits to the breakthrough curves (See Figures 1–3). This is consistent with the results reported previously for PFOS transport in two soils (Brusseau et al. 2019b). Simulations of the breakthrough curves including and excluding nonlinear sorption were similar for all media (see Figure 1 for an example), indicating that nonlinear sorption contributed minimally to the observed nonideal transport (asymmetry and tailing). The optimized values for β and ω are presented in the respective Figure captions. These translate to values of 0.47 (0.41–0.52), 0.40 (0.34–0.46), and 0.26 (0.22–0.30) for F, the fraction of sorbent for which sorption is instantaneous, and 2.1 (1.4–3.0), 2.1 (1.4–3.1), 1.6 (1.2–1.9) h⁻¹ for k₂ the first-order desorption rate coefficient, for the AFP44, Borden, and Hanford media, respectively. The values presented in parentheses are the 95% confidence intervals. The k₂ values are very similar to those reported for PFOS sorption by the two soils in the Brusseau et al. (2019b) study. Overall, these results indicate that the observed nonideal transport behavior is due primarily to rate-limited sorption/desorption. This is consistent with the results reported by Brusseau et al. (2019b) for PFOS transport in soil.

Inspection of Figure 4 shows that while asymptotic tailing is observed for PFOS transport in the untreated Hanford aquifer material, there is minimal arrival- and elution-front tailing observed for PFOS transport in the treated media. This provides a strong indication that the rate-limited sorption/desorption behavior observed for the untreated media is associated with PFOS interaction with the soil organic carbon. Furthermore, these results indicate that PFOS interaction with the inorganic constituents is rapid with respect to residence time. This latter observation is consistent with a likely sorption mechanism for this domain comprising electrostatic interactions, which are known to be rapid.

3.2. K_oc Values, Distributed-Sorption Models, and Sorbent-Constituent Contributions

While the use of organic-carbon normalized sorption coefficients (K_oc) is subject to uncertainty for PFAS, they can be calculated to help illuminate the potential contributions of sorbent constituents. Log K_oc values are reported in Table 3 for the three aquifer media, the two soils, and the sand. The measured values can be compared to a literature-based value of 2.8, which was determined from batch sorption measurements reported for 23 media compiled from seven separate studies (Brusseau, 2019). The organic-carbon contents of the soils and sediments employed in the literature studies ranged from 0.2–39%. PFAS interaction with soil organic carbon was identified by the original authors as the primary sorption mechanism for these media.

The log K_oc values determined for the lower-concentration experiments for the sand and Eustis soil are very close to the literature value of 2.8. Conversely, the log K_oc values of 3.2–3.5 for the Borden, Hanford, and Vinton media exceed the literature value. The application of the K_oc model to systems for which multiple soil constituents contribute significantly to sorption results in aggregate values that no longer represent the sole influence of organic-carbon content. Hence, the larger than predicted log K_oc value observed for the three media supports the contention that inorganic soil constituents are contributing to PFOS sorption to some extent.

The results discussed in this and the preceding section indicate that inorganic soil constituents are contributing to PFOS sorption by the aquifer media. Hence, the standard K_oc model would not be appropriate for characterizing PFOS sorption by these media. Distributed-sorption models can be developed for systems that comprise contributions of multiple soil constituents to sorption. Such models have been developed for PFAS sorption that account for contributions from soil organic carbon and inorganic soil components (Higgins and Luthy, 2007; Knight et al., 2017). For example, Knight et al. (2019) developed a three-component model incorporating OC, silt+clay fraction, and soil pH to characterize PFOA sorption by a large number of soils. These two studies focused on media with larger soil organic-carbon contents (>0.1%).

A simplified approach can be employed for distributed-sorption models based on the use of the K_d concept (e.g., McCarty et al., 1981; Rebhun et al., 1992; Schwarzenbach et al., 2003). In this approach, the composite K_d is defined as the sum of individual terms representing the amount of each constituent and the associated constituent-specific distribution coefficient $\left[{\text{K}}_{\text{d}}={\sum }_{i=1}^n{f}_i\cdot K_i\right]$. A three-component model was developed for the present study accounting for contributions from soil organic carbon, metal oxides, and silt+clay fraction, with the latter term used as a surrogate for clay-mineral content and other non-metal-oxide inorganic constituents. The K_d is defined as: K_d = f_oc K_oc + f_mo K_mo + f_sc K_sc, where K_mo is the metal-oxide normalized sorption coefficient, f_mo is the fraction of metal-oxide content, K_sc is the silt+clay normalized sorption coefficient, and f_sc is the fraction of silt+clay content. The values for f_oc, f_mo and f_sc are taken from Table 1 for the aquifer media and the sand, and from Brusseau et al. (2019b) for Eustis and Vinton soils. The model analysis was conducted using the K_d data generated for the lower input-concentration experiments.

The sand has no silt+clay (f_sc=0) and a very small f_mo. Therefore, PFOS sorption by the sand is assumed to be controlled solely by soil organic carbon. Hence, the sand data can be used to determine the value for K_oc a priori, which is determined to be 568. Notably, this value is very similar to K_oc values of 575 and 536 measured for sorption of PFOS by humin (Zhao et al., 2014) and peat (Zhi and Liu, 2018), respectively. Given their composition, the values measured for these organic materials may be considered as the most representative of true K_oc values. The concordance of the K_oc value determined for the sand to those measured for humin and peat provides validation for the former and for the assumption that sorption of PFOS by the sand is controlled by the soil organic carbon component.

The comparison of predicted and measured K_d values is presented in Figure 8. The model provides very good predictions for all media except the Eustis soil. This medium has the largest soil organic-carbon content (0.38%), and is the only one of the six that exceeds 0.1%. Prior research has indicated that PFOS sorption by the Eustis soil is predominated by soil organic carbon (Brusseau et al., 2019b). Hence, the failure of the three-component model for Eustis soil is not unexpected. A three-component model with the silt+clay fraction replaced by clay mineral content was also tested. It did not perform as well as the original model. A best-fit value of 7 was determined for K_mo, which is 80-times smaller than the K_oc value of 568, whereas a best-fit value of 19 was determined for K_sc (which is 30-times smaller than the K_oc). These results indicate that soil organic carbon has much greater sorptivity for PFOS compared to the inorganic constituents.

Figure 8.

Comparison of predicted K_d values determined from a three-component distributed-sorption model to measured values. Note that the data point for the Eustis soil is not included as it is off the scale.

The percent-contributions of each of the three soil constituents to total sorption based on the modeling analysis are presented in Table 4 for the five low-OC media. Soil organic carbon is estimated to contribute 19–42% of the total sorption for all media except the sand. The contribution of silt+clay ranges from 51–80% for all media except the sand. The only medium for which the contribution of metal-oxides is determined to be significant is Hanford, with an estimated contribution of 15%. The contribution from soil organic carbon is estimated at 34% for the Hanford medium, which is lower than one would anticipate based on the difference in K_d values obtained for the experiment conducted with the OC removed versus those for the untreated media. Comparison of the two sets of analyses is complicated by the use of different concentrations, with 10 mg/L used for the treatment-effect analysis and 0.1 mg/L employed for the model analysis. In addition, as discussed previously, potential impacts of the soil treatment on the contributions of inorganic constituents to PFOS sorption imparts uncertainty to the treatment-effect analysis.

Table 4.

Relative contributions of soil constituents to PFOS sorption

Medium	%-Contribution OC	%-Contribution MO	%-Contribution S+C
AFP44	42	0	58
Borden	19	1	80
Hanford	34	15	51
Sand	100	0	0
Vinton	39	2	59

Note: The K_d values used for the analysis are determined for a nominal concentration of 0.1 mg/L. The K_d for AFP44 was estimated using a mean Freundlich n value determined from the other media.

It is important to note the limitations to this modeling application. One limitation is that the model coefficients and outcomes will be a function of concentration for cases with nonlinear sorption. The results may vary for analyses conducted for significantly different concentrations, particularly if the degree of sorption nonlinearity varies greatly among the different media. Additionally, it is assumed that the same K_oc applies to all media, whereas it is possible that K_oc may vary as a function of the geochemical properties of soil organic carbon. Another limitation is the small data set; more robust analysis requires a much larger database.

3.3. Comparison to Literature Data

As noted in the Introduction, the few miscible-displacement studies conducted to date have employed a range of media. However, multiple studies have investigated PFOA and PFOS transport in natural quartz sands (Aly et al., 2018; Lyu et al., 2018; Brusseau et al., 2019a; Guelfo et al., 2020). The results of these studies can be compared to evaluate consistency and the impacts of system conditions on transport. Arrival fronts for PFOA transport in sand from these studies are presented in Figure 6. While there is some degree of noise, the data are reasonably consistent. Normalized elution fronts for PFOS transport in sand from two of the studies are shown in Figure 7. The pore volume (x-axis) is normalized by the respective retardation factor for each medium to eliminate the effect of differences in the magnitudes of retardation on the extent of tailing. The curves are reasonably consistent.

Figure 6.

Arrival fronts for PFOA transport in quartz sands. GUE = Guelfo et al., 2020 [S and F refer to experiments conducted with slower or faster pore-water velocities, respectively]; LYU = Lyu et al., 2018; Aly = Aly et al., 2018. The C₀ (input concentration) values are in μg/L.

Figure 7.

Normalized elution fronts for PFOS transport in quartz sand. The elution pore volumes are divided by the respective retardation factors for each experiment. Guelfo = Guelfo et al. (2020); Brusseau = Brusseau et al. (2019a). The input concentration for the Guelfo et al. experiment is 5 μg/L; the input concentration for the Brusseau et al. experiment is 1 mg/L.

The consistency between the two data sets presented in Figure 7 is of particular significance given the 200-fold difference in input concentrations (5 μg/L vs 1 mg/L) used for the two experiments. Another interesting point is that the PFOA and PFOS data for the Guelfo et al. (2020) study represent a multiple-component PFAS system, whereas the data for the other studies (Aly et al., 2018; Lyu et al., 2018; Brusseau et al., 2019a) represent-single-solute systems. The consistency of the transport data among the studies suggests that there was minimal impact of multiple components on sorption and transport in the sand for the conditions of that study.

Figure 5 presents a comparison of the extended elution tailing of PFOS in the three aquifer media compared to the two soils reported in Brusseau et al. (2019b) and the sand reported in Brusseau (2020). As noted above, the x-axis is normalized by the respective retardation factors. Additionally, an example data set is presented for the nonreactive tracer. Inspection of Figure 5 shows that PFOS exhibits greater tailing in the two soils than in the aquifer materials and sand. Notably, the two soils have larger soil OC contents than the aquifer media and the sand. This supports the hypothesis that the rate-limited sorption/desorption behavior observed for PFOS transport is related primarily to interaction with the soil organic-carbon components of the media. This is consistent with a prior study conducted by Brusseau and colleagues investigating the elution tailing of trichloroethene in several soils and aquifer materials (Russo et al., 2010), some of which are the same media as used in this study. Sorption of trichloroethene by the media is controlled by interaction with the soil organic carbon, and this interaction is responsible for the observed rate-limited sorption/desorption behavior.

4. Conclusion

The transport of PFOS in aquifer sediment comprising low organic-carbon contents was investigated and compared to transport in soils and sands. PFOS retardation was relatively low for transport in all aquifer media. The breakthrough curves measured for PFOS transport were asymmetrical and exhibited extensive concentration tailing, indicating that sorption/desorption was significantly nonideal. The results of model simulations indicated that sorption/desorption was rate-limited, which was determined to be the primary cause of the observed nonideal PFOS transport. Sorption was nonlinear, which influenced the magnitude of retardation but had minimal impact on the asymmetry and tailing of the breakthrough curves.

The magnitude of sorption was greater for the three aquifer media compared to the sand. Whereas the four media have similar OC contents, the aquifer media have appreciable metal-oxide and silt+clay contents compared to the sand. This indicates that inorganic soil constituents are contributing to PFOS sorption by the aquifer media. This was confirmed by the observation that sorption was reduced but not eliminated for a Hanford sample that was treated to remove soil organic carbon. The impact of rate-limited sorption was greatly reduced for the treated medium, indicating that sorption/desorption kinetics is associated with PFOS interaction with soil organic carbon. Comparison of PFOS transport in aquifer media to data reported for PFOS transport in two soils showed that PFOS exhibited more extensive elution tailing for the soils, likely reflecting the greater OC contents of the soils and concomitant differences in the relative contributions of various media constituents to sorption.

A three-component distributed-sorption model was developed that accounted for contributions from soil organic carbon, metal oxides, and silt+clay fraction. The model produced very good predictions of K_d for the media with low soil organic-carbon contents (≤0.1%), which comprise all of the media except Eustis soil. Overall, the results of the study indicate that sorption of PFOS by these media comprises contributions from multiple soil constituents. The results of this study provide insight into the contribution of aquifer-media constituents to sorption of PFOS under transport conditions, and have implications for characterizing and simulating PFAS transport in groundwater at the field scale.

Figure 2.

Measured and simulated breakthrough curves for transport of PFOS in Borden aquifer material. The simulations are produced using the two-domain sorption-kinetics model. Optimized parameter values: β = 0.7, ω = 0.2.

Highlights.

PFOS transport in aquifer media is examined

Transport is observed to be nonideal with asymptotic tailing

A distributed-sorption model is developed to predict K_d

Multiple soil constituents appear to contribute to PFOS sorption