Abstract Per- and polyfluoroalkyl substances (PFASs) constitute a chemically diverse family of persistent contaminants, the regulation of which is tightening rapidly in Europe and the United States. Granular activated carbon, selective ion exchange, and pressure-driven membranes remove many long-chain PFASs, but their performance is less robust for short-chain and ultrashort species, and all generate concentrated secondary waste streams. Hydrophobic deep eutectic solvents (DESs), including natural deep eutectic solvents (NADESs), have emerged as tunable liquid extractants able to concentrate PFASs into small solvent volumes that can be regenerated or coupled to destruction. This perspective differs from existing DES-PFAS reviews by converting qualitative solvent-selection arguments into a decision framework with explicit acceptance gates: broad PFAS affinity, a component-resolved non-migration specification for treated water, viscosity and mass-transfer limits, regenerability targets, and techno-economic/life-cycle benchmarking against incumbent processes. We refine the bifunctional DES design hypothesis by separating validated regimes from unresolved cases, identifying the reliability limits of COSMO-RS, molecular dynamics, and machine-learning screening, and defining tiered reporting requirements for early-stage studies. The central message is that PFAS-extracting DES should no longer be evaluated only by single-compound removal percentages; they must be judged as integrated, closed-loop treatment materials with solvent losses, regeneration stability, destruction compatibility, cost, and environmental impacts that are quantified from the outset. 1. Introduction and Regulatory Context Regulatory drivers. In the European Union, the recast Drinking Water Directive (EU) 2020/2184 [ 4, 5] introduced two PFAS-specific parameters that entered into force in January 2026: ‘PFAS Total’ (≤0.50 μg/L, encompassing the sum of all measurable PFAS) and ‘Sum of PFAS’ (≤0.10 μg/L, defined as the sum of 20 specified compounds spanning C4–C13 perfluoroalkyl carboxylates and sulfonates). In the United States, EPA’s April 2024 National Primary Drinking Water Regulation [ 6] established individual MCLs of 4 ng/L each for PFOA and PFOS, 10 ng/L each for PFHxS, PFNA, and HFPO-DA (GenX chemicals), and a hazard-index mixture approach for the combined occurrence of four regulated PFAS. These convergent regulatory instruments, coupled with mounting evidence of widespread trifluoroacetic acid (TFA) contamination at μg/L levels in European surface waters [ 7, 8] and the European Chemicals Agency’s proposed broad PFAS restriction covering approximately 10,000 substances [ 9], are compelling utilities to urgently evaluate advanced treatment and source-control strategies that extend well beyond the capabilities of conventional water treatment plants. Incumbent technologies: capabilities and gaps. The three technologies designated as Best Available Technologies by the US EPA for PFAS removal [ 6, 10] are GAC adsorption, PFAS-selective anion exchange (IX), and high-pressure membranes (RO/NF). GAC is the most widely deployed, operating at empty bed contact times of 10–20 min [ 10, 11], but its efficacy declines sharply for short-chain perfluorocarboxylic acids (C4–C6), and competitive adsorption by natural organic matter (NOM) can reduce effective bed life by 30–70% [ 11, 12]. Selective IX resins achieve substantially longer operational capacities (100,000–400,000 bed volumes versus 10,000–50,000 for GAC [ 10, 13]) with correspondingly smaller system footprints, but at an approximately three-fold higher unit media cost, and their regeneration remains technically challenging. RO/NF achieves near-complete PFAS rejection across all chain lengths, but at substantial energy cost (0.5–1.5 kWh/m 3 [ 10]), produces a PFAS-laden concentrate requiring further treatment, and is impractical for many small- and medium-sized utilities. Critically, none of these technologies destroys PFASs; they merely sequester or concentrate these persistent contaminants, thereby deferring the ultimate disposal challenge to downstream processes, typically incineration under conditions that may not achieve complete defluorination [ 14, 15]. Deep eutectic solvents as an alternative paradigm. Deep eutectic solvents are structured liquids formed by complexation of a hydrogen bond acceptor (HBA) with one or more hydrogen bond donors (HBDs), yielding pronounced melting-point depressions and liquids with tuneable polarity, viscosity, and interfacial properties. First introduced using choline chloride–urea mixtures by Abbott et al. in 2003 [ 16], the DES concept was extended to hydrophobic, water-immiscible formulations in 2015 by van Osch et al. [ 17]. Only recently has this conceptual framework been explicitly translated to PFAS removal. The trajectory of progress has been rapid: proof-of-concept work using a menthol:acetic acid NADES achieved extraction of perfluoroheptanoic acid (PFHpA) [ 18] but simultaneously revealed a critical limitation—partial leaching of the acidic component into the treated water; COSMO-RS-guided multi-criteria screening subsequently identified natural hydrophobic DES candidates for PFOA and PFOS [ 19, 20] with improved phase stability; integrated experimental–computational work then demonstrated near-quantitative PFOA removal with excellent reusability for a TOPO:lauric acid DES [ 21]; and most recently, an energy-based screening strategy achieved [ 22] >99% removal efficiency across multiple PFAS classes—including perfluorocarboxylic acids (PFCAs), perfluorosulfonic acids (PFSAs), and perfluoroalkyl amides—in real water matrices [ 22, 23]. In an important conceptual advance, hydrophobic DES platforms have now been demonstrated to be capable of integrated PFAS capture and in situ mild-condition mineralization [ 24], representing a paradigm shift from separation-only to separation-plus-destruction architectures. Scope of this perspective. This perspective is organized around four non-negotiable design targets: (1) high extraction affinity across both long-chain and short-chain PFAS (including the full EU PFAS-20 list and emerging ultrashort species such as TFA); (2) quantitative non-migration specifications enforced at environmentally relevant concentrations (ng/L–μg/L); (3) exclusive reliance on natural, benign, biodegradable, and low-cost components compatible with large-scale industrial supply chains; and (4) demonstrated process scalability at total treatment costs competitive with or below those of incumbent technologies ( 0.8), such as tertiary amine oxides, long-chain amides, or phosphine oxides. The demonstration of PFAS extraction by TOPO-based DES elegantly illustrates the potency of strong HBAs [ 21]; however, TOPO is not a ‘natural’ component under current Safe-and-Sustainable-by-Design frameworks, motivating the search for bio-based functional analogues, including matrine (a tetracyclic quinolizidine alkaloid from Sophora flavescens [ 28]), phospholipid-derived phosphine oxides, or sterol-derived amine oxides. The bifunctional concept is strongly supported by recent ab initio molecular dynamics (AIMD) simulations [ 23], which demonstrate that DES components self-organize to form a complementary, flexible non-covalent interaction network around diverse PFAS targets, with spatially distinct solvation shells for the fluorinated tail and the polar headgroup operating simultaneously within the nanostructured liquid. Experimental anchoring of the design principle. A fully matched pair comparing a purely hydrophobic DES and an otherwise identical bifunctional DES for a complete PFAS mixture has not yet been reported; this is now stated explicitly as an important data gap. Nevertheless, the existing literature provides directional support: simple menthol/short-acid NADES demonstrated the feasibility of hydrophobic extraction but also revealed leaching limitations [ 18], whereas TOPO:lauric acid and ionizable DES platforms showed stronger PFOA or multi-PFAS uptake when a high-basicity headgroup-binding motif was present [ 21, 22, 23]. The revised text, therefore, presents the bifunctional concept as a testable design hypothesis supported by convergent evidence, rather than a universally proven mechanism. Validated and unresolved regimes. The bifunctional hypothesis is best supported for long-chain and mid-chain PFCAs/PFSAs (approximately C6–C12), for which hydrophobic tail transfer provides a large thermodynamic driving force and strong HBA/basic motifs can stabilize the polar headgroup. It is partially supported for C4–C6 short-chain acids, neutral precursors, and amide-like PFAS, where headgroup-specific interactions become more decisive, and the role of ionic strength increases. It remains a working hypothesis, not an established mechanism, for ultrashort PFAS such as TFA and PFPrA, the hydration free energies, small hydrophobic volume, and high mobility of which make extraction much less likely to be governed by the same tail-solvation term. These domains are now treated separately throughout the screening workflow. 2.3. Non-Migration as a First-Class Design Objective Solvent loss to the aqueous phase is the central technical barrier for translating DES extraction into potable-water or discharge-quality treatment. From a regulatory and risk perspective, it is wholly insufficient to demonstrate ‘phase separation’ macroscopically or to report a low percent mass loss: the operationally relevant targets are component-resolved concentrations at ng/L–µg/L levels in the treated water, assessed using analytical methods with appropriate limits of quantification. We therefore advocate for a formal ‘non-migration specification’ analogous to the specific migration limits established for food-contact polymers under EU Regulation (EU) No 10/2011 [ 29], encompassing three quantitative criteria: (a) each individual DES component must exhibit an aqueous solubility below a defined threshold (e.g., 3) rather than introduced as a water-soluble salt or small polar molecule. 2.4. Quantitative Design Criteria Table 1 consolidates the minimum quantitative criteria that a PFAS-extracting DES must satisfy, organized according to the four design targets of this perspective, as well as illustrative examples in Table 2. The values should be read as transparent go/no-go gates, not as claims that have already been achieved simultaneously. The table therefore distinguishes three evidence classes: criteria demonstrated in at least one literature system, criteria partially demonstrated for restricted PFAS classes or cycle numbers, and prospective engineering targets derived from incumbent-technology benchmarking, migration-control logic, or TEA/LCA constraints. 2.5. Predictive Modelling: Three Sequential Computational Gates Gate 1—DES formation and phase behaviour. Gate 1 predicts the solid–liquid equilibrium to identify compositions that yield a stable liquid phase at the target operating temperature (T 3000 population equivalents—a target that appears achievable on paper but remains to be demonstrated at pilot scale. Social factors. Public perception of water treatment additives presents a distinctive challenge for DES-based extraction that does not affect adsorption-based technologies. Whereas GAC and IX are perceived as ‘filtering’ technologies that remove contaminants, liquid–liquid extraction introduces a chemical agent into intimate contact with drinking water—a distinction that may trigger public concern regardless of the actual risk profile. The use of food-grade, GRAS-listed (Generally Recognized as Safe) components—many of which are already present in consumer food products as flavourings or preservatives—provides the strongest available foundation for social acceptance. Transparent publication of third-party toxicological data, aquatic ecotoxicity assessments, and biodegradability results in open-access venues is essential for building and maintaining social licence. Technological factors. Several unresolved technological challenges define the current development bottleneck. Mass-transfer limitations imposed by the characteristically high viscosities of many DES formulations (50–500 mPa·s at 25 °C, compared to 0.5–2 mPa·s for conventional organic solvents) may require elevated operating temperatures (40–60 °C), prolonged contact times, or intensive mechanical agitation—all of which add to energy costs and process complexity. Fouling of membrane contactors by NOM, colloidal matter, and biofilm formation in real water matrices remains uncharacterized for DES systems. Validated regeneration protocols demonstrating stable extraction performance over industrially relevant cycle numbers (≥50 cycles, ideally >200) are entirely lacking for PFAS-extracting DES. And the compatibility of loaded DES with various PFAS destruction technologies (electrochemical, plasma, and SCWO) has been explored only in preliminary studies with limited analytical characterization of degradation by-products. Legal factors. The regulatory treatment of DES components in drinking water falls into an uncharted category. Any detectable solvent residue in treated water will face stringent scrutiny from drinking-water regulators accustomed to evaluating inorganic and organic contaminants with established toxicological databases. For DES components without existing regulatory limits (which is the case for most terpene–fatty acid combinations), regulators may apply conservative default thresholds—potentially as low as the analytical LOQ—until substance-specific assessments are completed. Under REACH (Registration, Evaluation, Authorisation and Restriction of Chemicals), manufacturing or importing DES at quantities exceeding 1 tonne per year triggers registration obligations, including submission of physicochemical, toxicological, and ecotoxicological data packages. Proactive engagement with regulatory agencies, ideally through pre-submission consultations, and development of migration-testing protocols explicitly modelled on the food-contact materials framework (EU Regulation 10/2011) will be critical for establishing a viable regulatory pathway. Environmental factors. From an environmental perspective, DES extraction must demonstrate life-cycle impacts that are at minimum equivalent to—and ideally better than—those of GAC and IX across multiple impact categories, as documented through the LCA framework proposed in Life-Cycle Assessment: Requirements, Hotspots, and Reporting Standards Section. A particular vulnerability is the sourcing of bio-based components: large-scale demand for menthol, thymol, or specific fatty acids could, if met through dedicated agricultural cultivation rather than waste-stream valorization, create land-use, water-use, and biodiversity trade-offs that undermine the ‘green’ narrative. Sourcing DES components from existing industrial waste streams (e.g., terpenes from citrus peel processing and fatty acids from vegetable oil refining residues) provides a far more environmentally defensible supply chain and should be prioritized wherever component purity specifications can be met. 4.2. SWOT Analysis Figure 5 presents a structured SWOT analysis that synthesizes the internal strengths and weaknesses of DES-based PFAS extraction technology alongside the external opportunities and threats identified through the PESTLE assessment. We elaborate on the most strategically significant elements of each quadrant below. Weaknesses. The most consequential weakness is the inadequately characterized leaching behaviour of DES components at environmentally relevant concentrations. The literature to date reports extraction efficiencies and phase-separation characteristics but almost universally fails to report component-resolved aqueous-phase residuals at the ng/L–μg/L level using appropriately sensitive analytical methods. This information gap is not merely an academic lacuna—it represents the single largest technical barrier to regulatory acceptance. The characteristically high viscosities of many DES formulations (often 50–500 mPa·s at ambient temperature) impose mass-transfer limitations that may negate the thermodynamic advantages of high K D values. Additionally, the absence of validated, multi-cycle regeneration data means that the practical cost of solvent replacement—and therefore the economic viability of the entire approach—remains essentially unknown beyond approximately 5–10 cycles. Opportunities. The regulatory environment presents the most powerful external opportunity. With the EU PFAS-20 limit of 0.10 μg/L now legally binding and the US EPA MCLs entering their monitoring-and-compliance phase, the demand for effective, affordable PFAS treatment is not speculative but immediate and growing. DES extraction is particularly well-positioned to serve the emerging market for hybrid treatment trains—systems that combine pre-concentration (by DES-LLE) with downstream electrochemical or SCWO destruction—addressing the fundamental criticism that adsorption-based methods merely transfer PFAS from water to a solid waste stream. The rapid maturation of machine-learning approaches for molecular property prediction, combined with the vast combinatorial space of potential DES formulations (>10 5 feasible binary pairs from bio-based component libraries), creates an unprecedented opportunity for accelerated computational screening that was not available even five years ago. Threats. The most significant competitive threat is the continued incremental improvement of established technologies—particularly next-generation PFAS-selective ion-exchange resins that combine long operational lifetimes (>300,000 bed volumes) with single-use disposal to high-temperature incineration, eliminating the regeneration challenge entirely. If selective IX costs continue to decline through manufacturing scale-up, the economic window for DES-based alternatives may narrow. A second serious threat is the potential for negative public perception: the concept of adding a liquid chemical to drinking water, even one composed of food-grade ingredients, may encounter resistance in communities already sensitized to chemical contamination. Finally, the diversity of the PFAS contamination challenge itself—spanning long-chain legacy compounds, short-chain substitutes, ultrashort species like TFA, and emerging precursors—means that no single DES formulation may achieve universal applicability, potentially limiting market penetration to specific contamination profiles. 5. Actionable Recommendations and Screening Workflow (1) Treat non-migration as a quantitative performance metric from day one: Every publication reporting DES-based PFAS extraction should include component-resolved leaching data obtained by LC-MS/MS with LOQ ≤ 50 ng/L, TOC drift (ΔTOC after extraction), and pH drift (ΔpH) measured after extraction and after a minimum of 10 reuse cycles. Studies that report only percent mass recovery of the DES phase or visual confirmation of phase separation are insufficient for technology translation and should be clearly identified as preliminary. (2) Screen against PFAS mixtures, not single compounds: The EU PFAS-20 list (C4–C13 perfluoroalkyl carboxylates and sulfonates) should serve as the minimum target compound set, supplemented by at least one neutral precursor or amide and, where analytically feasible, one ultrashort PFAS (TFA or PFPrA). All screening experiments should include realistic ionic strength (5–10 mM NaCl) and NOM (≥5 mg/L TOC as Suwannee River humic/fulvic acid or equivalent standardized reference material). (3) Adopt multi-objective optimization as the default design paradigm: Simultaneous optimization can be achieved across ≥5 objectives (KD, viscosity, aqueous solubility, toxicity, component cost) using Pareto-front analysis. Publish the complete Pareto set, not only the single best-performing formulation, to enable community-wide learning and meta-analysis. Machine-learning models trained on curated DES physicochemical property databases can accelerate this screening by orders of magnitude beyond what is achievable by sequential experimental evaluation. (4) Prefer process architectures that physically retain the DES phase@ For potable-water applications, membrane contactors or supported liquid phases should be the default contacting mode, given their inherent advantage in suppressing solvent carryover to diffusion-limited levels. Dispersive contactors (mixer–settlers and centrifugal extractors) may be acceptable for industrial side-stream or wastewater applications where residual DES concentration tolerances are orders of magnitude higher than for drinking water. (5) Couple extraction with an explicit end-of-life strategy from the outset: Regeneration of the DES followed by destruction of a small, highly concentrated PFAS stream should be designed as an integral part of the process from the earliest stages, not deferred to ‘future work’. Candidate destruction methods (electrochemical oxidation, SCWO, and non-thermal plasma) must be evaluated for chemical compatibility with the DES matrix components and for the verified absence of toxic or persistent by-products. (6) Establish and enforce minimum reporting standards All experimental studies should report complete DES composition (including residual water content, identity and concentration of impurities), density and dynamic viscosity at a minimum of two temperatures, phase behaviour characterization (visual + DSC), thermal and oxidative stability data, and reuse performance over ≥10 extraction–regeneration cycles with component-resolved analytical quality control at each cycle. Tiered reporting set. To make the recommendations actionable at TRL 2–3, we now separate minimum Tier 1 metrics from more demanding Tier 2 indicators. Tier 1 should be reported by every early-stage study: PFAS removal/KD for a defined mixture, component-resolved DES leaching or a justified LOQ-limited surrogate, viscosity/water content after contact, and reuse over at least ten extraction–regeneration cycles. Tier 2 indicators, required before pilot advancement, include full PFAS-20 mixture testing, NOM/salinity stress tests, ≥50-cycle solvent-quality control, preliminary TEA/LCA inventories, by-product analysis after destruction, and membrane-contactor or centrifugal-contactor operation with real matrices. (7) Benchmark cost and environmental impacts early using transparent, simplified TEA/LCA. Even at TRL 2–3, researchers can and should compile mass and energy balances and estimate unit treatment costs ( $/m 3) and key environmental impacts (kg CO 2-eq/m 3, CED in MJ/m 3) using the framework presented in Table 3 and Section Life-Cycle Assessment: Requirements, Hotspots, and Reporting Standards. Solvent make-up cost, analytical monitoring cost, and waste-handling cost must all be included. Studies that claim ‘green’ or ‘sustainable’ credentials without any quantitative LCA or TEA data should be viewed with appropriate scientific scepticism. 6. Conclusions and Outlook Hydrophobic deep eutectic solvents and NADESs have crossed the threshold [ 18, 19, 20, 21, 22, 23, 24] from conceptual ‘green solvents’ to credible PFAS extractants, with recent demonstrations achieving > 99% removal efficiencies for multiple PFAS classes [ 22, 24] in real-water matrices, mechanistic understanding at the atomistic level through advanced computational methods (COSMO-RS, AIMD) [ 19, 20, 23, 30], and initial—though still preliminary—explorations of integrated capture-and-destroy platforms. Yet the pathway from these encouraging laboratory results to real-world
