Mineralogical controls on PFAS and anthropogenic anions in subsurface soils and aquifers

Abstract

Per- and polyfluoroalkyl substances (PFAS) migrate into the environment through various means, e.g., soil-amendment impurities and ambient atmospheric deposition, potentially resulting in vegetative uptake and migration to groundwater. Existing approaches for modeling sorption of PFAS commonly treat soil as an undifferentiated homogeneous medium, with distribution constants (e.g., K_d, K_oc) generated empirically using surface soils. Considering the limited mineral variety expected in weathered geologic media, PFAS mobility can be better understood by accounting for predictable mineral assemblages that are ubiquitously distributed in US soils. Here we explore the role of minerals and electrostatic sorption in controlling PFAS mobility in subsurface settings at contaminated agricultural sites by measuring geochemical parameters and PFAS, and calculating pH-dependent mineral surface charges through full soil and aquifer columns. These data suggest subsurface mobility of short-chain PFAS largely is controlled by aluminum-oxide mineral(oid) electrostatic sorption, whereas long-chain PFAS mobility is controlled by organic matter and air-water interfacial area.

Fig. 1. Anthropogenic compounds applied to soils partition between soil solids, water and air-filled pore space.

a Surface soils concentrate in weathering-resistant quartz, organic matter, and low concentrations of incipient amorphous Fe and Al (hydr)-oxides where long-chain PFAS are retained. b Subsurface horizons accumulate authigenic minerals, including clays and crystalline Fe and Al (hydr-)oxides where short-chain PFAS can sorb. c In saturated media, mineral assemblage commonly reflects more extensive weathering in high-flow zones relative to low-flow zones. Al (hydr)oxides and kaolinite can sorb short-chain PFAS regardless of oxidation state, while in reduced settings Fe (hydr)oxides may be subject to reductive dissolution thereby not contributing to the potential exchange complex.

Fig. 2. Major minerals expected to form in soils from weathering and found at the present study sites.

A Weathering dissolution of a wide array of rock minerals produces a limited assemblage of authigenic minerals that are present in almost all soils and weathered settings, the abundance of which evolves with degree of weathering. Here thermodynamically stable minerals are plotted as a function of evolving soil-water composition, with general weathering classifications and USDA Soil Orders depicted. In this example of the global soil-forming process, the common rock-forming mineral albite dissolves, first tending to form 2:1 clays (two silica sheets sandwiching an aluminum sheet). As weathering continues, 2:1 clays dissolve to form 1:1 clays and, in still more weathered zones, the 1:1 clays dissolve leaving (hydr)oxide minerals. B Mineralogic profiles using data from the current study (3-point smoothed), dominated by resistate quartz and authigenic kaolinite and gibbsite, with some vermiculite (a 2:1 clay) as expected for soils of advanced weathering stage. The blue horizontal line depicts the water table.

Fig. 3. Perfluorocarboxylates (PFCAs) in Field 1 soil as a function of depth.

A ∑PFCAs remain highest at the soil surface more than 15 years after sludge application (B) The surface soils are dominated by chain-lengths C10–C14, but subsurface are dominated by shorter chain-lengths (C4–C8). When PFCAs are grouped according to their inferred intermediate precursor compounds, the FTOHs, surface soils have more long chains (C) and subsurface soils have more short chains (D) than typical sidechain fluorotelomer polymers, but summed over the full soil profile, the homologue distribution compares reasonably with the inferred ultimate precursor (E). F Because of relatively higher mobility of shorter-chain PFCAs and perfluorosulfonates (PFSAs), compared to the surface-soil source, groundwater is burdened with more short chain PFCAs than long, with an absence of detections for PFDA (C10) and longer compounds. G Even 1.5 decades after sludge application, groundwater still contains up to two orders-of-magnitude higher PFCAs than the 2024 EPA MCLs of 10 (PFHxS, PFNA) and 4 (PFOA, PFOS) ng/L.

Fig. 4. pH- and ionic-strength –dependent electrostatic charge of authigenic (hydr)oxide Fe and Al mineral(oids).

The electrostatic surface charge of incipient, amorphous hydrous ferric oxide (HFO) and aluminum oxide (HAO), and the crystalline goethite (FeO(OH)) (A) and gibbsite (Al(OH)₃) (B) varies as a function of pH and ionic strength according to Supplementary Eq. 16, positively charged at low pH, diminishing to, and converging at, the zero-point of charge (ZPC). C Using measured concentrations of these mineral phases, and pH and estimated ionic strength (Supplementary Eqs. 17 and 18), the fraction of the total positive electrostatic surface charge is shown in our colluvial study site, showing that the incipient HFO and HAO are considerable in the surface soil, then the crystalline phases become more prominent in the subsurface, with goethite mostly in the vadose zone and gibbsite generally dominant near and below the water table (SI III Extended Geochemical Summary). D, E Summing these surface charges, the sandy surface soil bears relatively little mineral-derived positive electrostatic surface charge and increases to higher capacity in the soil horizons of illuviation, with considerable surface charge maintained through the shallow aquifer for the Al (hydroxides)+kaolinite (E).

Fig. 5. Heat map of nominal correlation coefficients for PFAS concentrations (rows) with geochemical properties (columns).

Geochemical properties are grouped according to close functional relationships. Perfluorocarboxylates (PFCAs) are ordered by increasing CF₂ chain-length, for combined Field 1 and 2 data, and perfluorosulfonates (PFSA) for Field 1 (most PFSAs were undetected in Field 2), above the water table. Increasing intensity of red coloration denotes progressively more significant positive nominal correlations (based on statistical p-values; yellow p ≤ 0.10, orange p ≤ 0.05, dark orange p ≤ 0.01), and blue intensity (light to dark from p ≤ 0.10 to p ≤ 0.01) similarly reflects inverse negative nominal correlations as a consequence of inverse covariance with positively correlated properties in natural soil columns.