Dynamic hydroxyl cycle of zeolite for short and ultra-short chain PFAS free potable water

Abstract

The global shift in per- and polyfluoroalkyl substance (PFAS) crisis towards (ultra-)short-chain congeners highlights the urgent need to design adsorbents with hydrophilic functions. However, current functional group immobilization modifications fail to balance efficiency with regeneration, limiting access to affordable (ultra-)short-chain PFAS-free potable water. Herein, we present a non-immobilized dynamic hydroxyl cycling strategy that uses only water to circularly drive commercial zeolite adsorption-regeneration processes, efficiently and sustainably removing various (ultra-)short-chain PFASs (C-F number: 1-6) from potable water. The enhanced nanopore accessibility and the formation of a “zeolite framework-confined water” dual-binding mode make the modified zeolite one of the highest capacity adsorbents reported (233.82-733.13 mg g^-1). Notably, this methodology offers a potential low-cost, in-situ upgrade for existing purifiers, potentially providing healthy water to underdeveloped regions through 179 days of full-scale purification.

Amid the global PFAS crisis shifting toward (ultra-)short-chain congeners, researchers develop a non-immobilized hydroxyl cycling strategy using water to drive zeolite adsorption regeneration cycles, enabling sustainable water purification upgrades.

Introduction

Per- and polyfluoroalkyl substances (PFASs), often referred to as “forever chemicals”, are extensively utilized across various sectors, making them pervasive in daily life¹. The continuous accumulation of PFASs in the environment, coupled with increasing evidence of their harmful effects on humans even at ultra-trace exposure levels, prompted the European Parliament and Council (EPC) to establish a maximum limit of 500 ng L^–1 for all PFASs in drinking water in 2021^2,3. Although legacy long-chain PFASs (chain length >7), e.g., perfluorooctanoic acid (PFOA), have been phased out in recent years, ultra-short-chain PFASs (chain length <4) and short-chain PFASs (chain length = 4–7) have replaced them and now dominate in potable water^4–6. However, growing evidence suggests that (ultra-)short-chain PFASs may exhibit similar or even greater biological toxicity than long-chain PFASs^5,7. The worse reality is that conventional water treatment methods (e.g., coagulation and advanced oxidation) are ineffective in eliminating (ultra-)short-chain PFASs, leaving individuals are unknowingly exposed to PFAS levels that exceed safe limits through their tap water^4,5. Hence, it is imperative to employ advanced water purification technologies (e.g., reverse osmosis) at the point-of-use. However, even in developed nations, only about 40% of the population can afford these costly solutions⁸.

Adsorption process offers a cost-effective method for removing PFASs through solid-liquid mass transfer methods, thereby enabling affordable point-of-use purification of drinking water. Still, adsorption techniques face significant limitations in the purification of (ultra-)short-chain PFASs: (1) Conventional hydrophobic mass transfer mechanisms leads to a significant decrease in the adsorption affinity of adsorbents for hydrophilic (ultra-)short-chain PFASs as chain length decreases, resulting in diminished or even negative removal efficiency^4,9. To address this issue, the current approaches focus on grafting of hydrophilic functional groups (e.g., -OH or -NH₂)^10,11. Yet, this introduces a second limitation: (2) The hydrophilic functional groups are prone to degradation, becoming inactivated during post-disposal of the PFAS adsorbent (e.g., thermal treatment). Despite substantial efforts, currently proposed solutions tend to optimize either efficiency (e.g., β-cyclodextrin polymer, hollow covalent organic framework, magnetic fluorinated polymer, and metal-organic framework) or reusability (e.g., electrochemical and microwave)^10–17, leading to increased costs over the full adsorption-regeneration life cycle. Thus, the key to providing affordable (ultra-)short-chain PFAS-free potable water lies in resolving the inherent trade-off between adsorption efficiency and reusability.

We present a non-immobilized modification strategy based on dynamic hydroxyl cycling in commercial zeolites for the adsorption and regeneration of (ultra-)short-chain PFASs. Unlike alternative enhancement strategies, the approach described here requires only the introduction of a reagent-free, low-cost water vapor diffusion technique into the conventional adsorption-regeneration process, enabling dynamic cycling of Si-OH and confined water-hydroxyl networks across various commercial zeolites (Fig. 1a, b). This non-immobilized strategy effectively improves the mass transfer process of (ultra-)short-chain PFASs within the nanopores and established a “zeolite framework-confined water” dual-binding mode, thus achieving the highest reported adsorption capacity and stable recycling. More importantly, this protocol is directly applicable for the purification of (ultra-)short-chain PFASs in real potable water, as validated by a full-scale household purifier over a period of 182 days.

Fig. 1. Design principle and optimization of dynamic hydroxyl cycle.

a Potential operation scheme of the dynamic hydroxyl cycle; b potential mechanism of the dynamic hydroxyl cycle; c adsorption efficiency of zeolites with different crystal and Si/Al ratio in the dynamic hydroxyl cycle.

Results

Design of dynamic hydroxyl cycle process

The grand challenge in designing a dynamic adsorption process lies in identifying a non-immobilized enhanced media that balances the efficiency and reusability. High efficiency is required to provide specific binding sites that enable the adsorbent to capture concentrations of (ultra-)short-chain PFASs at concentrations as low as one part per 100 billion in complex environmental systems. Meanwhile, a suitable reusability allows the enhanced media to bind to the adsorbent through non-chemical bonds, facilitating easy cycles of functionalization and de-functionalization while maintaining the structural stability.

Zeolite, a commonly used adsorbent, exerts a hydrophobic effect on the C-F chains of PFASs through its Si-O framework, but lacks a hydrophilic mode of action¹⁸. One of the most promising enhanced media that can meet aforementioned requirements, while remaining compatible with the zeolite framework, is confined water. Due to the surface tension, the liquid water molecules require pressures of >80 MPa to penetrate the hydrophobic nanopores, resulting in cavities within the zeolite¹⁹. However, if water molecules are pre-confined within zeolite nanopores through interfacial interactions, they can potentially form non-immobilized hydrophilic sites for special in-site binding of (ultra-)short-chain PFAS²⁰. Additionally, confined water is non-chemically bound to the nanopore walls, allowing for green and low-cost cycles of functionalization and de-functionalization through diffusion and thermal processes, using only water. Base on this concept, we designed an adsorption-regeneration scheme using the dynamic hydroxyl cycle. This process involves the introduction of water vapor diffusion into the traditional adsorption-thermal regeneration process of zeolites, achieving a three-step “construction-utilization-removal” of the dynamic hydroxyl cycle (Fig. 1a, b).

To verify the feasibility of the dynamic hydroxyl cycling scheme and determine the optimal zeolite substrate, we investigated the adsorption efficiency of a series of commercial zeolites with systematically varied pore sizes (0.52–1.12 nm) and surface hydrophilicity (Si/Al ratio: 25–600, Supplementary Fig. 1 and Table 1) for 9 (ultra-)short-chain PFASs (C-F number = 1–6, Supplementary Table 2) within the dynamic hydroxyl cycle. Notably, we observed an enhanced removal efficiency (4%-88% → 31%-100%) across all commercial zeolites confirming the broad applicability of the dynamic hydroxyl cycle strategy (Fig. 1c). Among the tested zeolites, those with moderate pore size (0.56–0.74 nm, BEA crystal) and high hydrophobicity (Si/Al ratio ≥200) achieved the optimal removal efficiency (>61%) for all (ultra-)short-chain PFASs. These results validate the basic assumption that the dynamic hydroxyl cycle increases adsorption efficiency.

Demonstration of dynamic hydroxyl cycle process

Based on the optimal β200 zeolite (BEA crystal, Si/Al ratio = 200), we further elucidated the change in structural modifications and adsorption properties during the hydroxyl “construction-utilization-removal” cycle. Initially, we thermally treated the original β200 to simulate the structure of the regenerated zeolite (D-β200). After calcination at 800 °C, the Si-OH characteristic peak (954 cm^–1) in the infrared spectrum of D-β200 zeolite disappeared, indicating high-temperature-induced ortho-hydroxyl condensation (Fig. 2a and Supplementary Fig. 3) and a reduction in surface polarity (ζ potential: −43.3 mV → −16.6 mV, Supplementary Fig. 2). Despite these changes, the D-β200 retained a well-preserved Si-O-Si framework (1082 cm^–1 and 800 cm^–1) and maintained its X-ray diffraction crystal facets (Fig. 2b) even under thermal treatment. High-resolution transmission electron microscopy images and pore size analysis of the D-β200 confirmed the preservation of particle morphology and microporous structure (1.894 nm vs. 1.971 nm, Fig. 2c and Supplementary Table 3).

Fig. 2. Demonstration of dynamic hydroxyl cycle.

a Fourier transform infrared spectroscopy (FTIR) patterns of β200 during the dynamic hydroxyl cycle; b X-ray diffraction (XRD) spectroscopy of β200 during the dynamic hydroxyl cycle; c transmission electron microscopy images of β200 and D-β200; ¹H single quantum/double quantum nuclear magnetic resonance (NMR) spectrum of (d) S-D-β200 and e S-D-β200 after PFHpA adsorption; f Comparison of adsorption capacity of on S-D-β200 and other adsorbents reported for (ultra-)short-chain PFASs, the raw data can be obtained from Supplementary Table 6.

To construct the hydroxyl groups, the D-β200 was exposed to a high humidity (90–100%) environment to prepare S-D-β200, facilitating the self-assembly of confined water structures through the vapor diffusion method. After this construction process, the characteristic peaks of the confined water hydrogen-bond network (3750 cm^–1–3000 cm^–1) and Si-OH (954 cm^–1) appeared in the infrared spectrum of S-D-β200, providing evidence for the successful loading of confined water and the rehydration of the unsteady Si-O-Si bonds (ζ potential: −16.6 mV → −41.6 mV, Supplementary Fig. 2 and 3). These results demonstrated that water vapor encapsulation created a confined water structure in the β200 nanopore and achieved orderly regulation of the hydrophilicity/ hydrophobicity (Si-OH and Si-O-Si) of the zeolite surface through the chemical transfer of water molecules (Supplementary Fig. 3). The water balance control (W-β200, Fig. 2a) showed no characteristic peaks, indicating that the water molecules were confined to the nanopore and not attached to the surface. Advanced 2D ¹H-¹H single/double quantum magic angle spinning (MAS) nuclear magnetic resonance (NMR) spectroscopy was used to investigate the confined water hydrogen-bond network and its binding state²¹. In S-D-β200, confined water appeared as water molecular chains within channels (δ = 1.04 ppm) and water clusters at intersections (δ = 5.54 ppm), interacting with Si-OH (δ = 2.13 ppm) and Al-OH (δ = 1.73 ppm, 7.72 ppm) groups (Fig. 2d)²². In terms of proton-proton proximities, only the water cluster exhibited diagonal peaks (11.08, 5.54), with interactions involving other protons displayed non-diagonal peaks ((6.58, 5.54); (7.27, 5.54); (7.67, 5.54)). These results confirm that the confined water modified the zeolite cavity environment in a non-immobilized manner and formed a hydrophilic center of water clusters at the intersections of channels (Supplementary Fig. 4).

The successful construction of the non-immobilized confined water hydrogen-bond network motivated us to conduct adsorption efficiency tests on S-D-β200. S-D-β200 demonstrated rapid, irreversible adsorption of all (ultra-)short-chain PFASs, reaching equilibrium within 2 h (Supplementary Fig. 5). Compared to the original β200, we observed both an increase in equilibrium adsorption capacity and a decrease in the adsorption rate (Supplementary Table 4), indicating the occurrence of enhanced adsorption behavior during rate-limiting steps, such as intra-particle diffusion²³. The maximum adsorption capacity of S-D-β200 for all (ultra-)short-chain PFASs was 1.41-2.26 times higher than that of β200 (Supplementary Fig. 6 and Table 5), suggesting an increase in accessible adsorption sites. The hydroxyl groups (confined water and Si-OH) in S-D-β200 were crucial for providing binding sites (Supplementary Fig. 7), where confined water clusters offered increased adsorption site to selectively bind with the hydrophilic heads of PFASs, leading to deionization (δ = 9.00 ppm, (14.98, 5.98)) and re-exposure of partially hydrated Brönsted-acid sites (δ = 2.91 ppm, Fig. 2e)²². Meanwhile, the shielding effect of confined water also helps to weaken the electrostatic repulsion between (ultra-)short-chain PFASs and S-D-β200, as well as other PFASs²⁴. It is noteworthy that the adsorption capacity of S-D-β200 surpassed that of currently reported optimal adsorbents (Fig. 2f and Supplementary Table 6), such as NU-1000²⁵, and showed a comparative advantage (e.g., preparation costs and energy consumption) when compared to the latest electroassisted adsorption materials (i.e., P(A), P(A98F2), and P(A39T53F8))²⁶. Moreover, S-D-β200 exhibited suitable (ultra-)short-chain PFAS (20 μg L^–1) selectivity with only a 0–10.9% decreased (Supplementary Fig. 8) when exposed to simulate environmental water samples (humic acid = 2 mg L^–1, NaCl = 20 mg L^–1), demonstrating its potential for practical applications.

To initiate the next hydroxyl cycle and degrade PFASs, the saturated S-D-β200 was calcined again via thermal treatment at 800 °C. Four complete dynamic hydroxyl cycles were conducted using simulated environmental water samples (humic acid = 2 mg L^–1, NaCl = 20 mg L^–1), recovering 94.3–101.3% of the adsorption capacity of S-D-β200 for (ultra-)short-chain PFASs (Supplementary Fig. 9). These results demonstrated that high-temperature (800 °C) calcination effectively removes hydroxyl groups of confined water and Si-OH (Fig. 2a and Supplementary Fig. 2), while completely degrading PFASs²⁷, thereby breaking the persistent PFASs cycle in the environment. In addition, the stability of the Si-O-Si framework, XRD crystal facets and pore structure of β200 were further confirmed after cycling (Fig. 2a, b and Supplementary Table 3), while the slightly increased specific surface area (12.5%) promoted the adsorption capacity for certain PFASs to a level that exceeds that of the original S-D-β200.

In summary, the proposed cyclic adsorption-regeneration scheme enables dynamic regulation of the hydrogen-bond network of confined water, addressing the inherent trade-off between efficiency and reusability through a non-immobilized combination strategy. Besides the demonstrated optimal structure (i.e., β200), other zeolites with varying pore sizes and surface hydrophilicity also exhibit similar dynamic hydroxyl cycles during the process (Supplementary Fig. 10), offering broader possibilities for implementing the dynamic hydroxyl cycle strategy and addressing more complex pollution issues.

Enhanced adsorption mechanism

The increased adsorption capacity of S-D-β200 is primarily attributed to the introduction of a confined water structure, as detailed in the previous section. To further clarify the confined water adsorption behavior, we constructed a three-phase full-atom model of “bulk water-confined water-zeolite” with a BEA structure (Si/Al ratio: 200). This model allowed us to demonstrate the structural differences before and after cyclic adsorption in the presence or absence of confined water (Supplementary Fig. 11). We quantified the adsorption free energy (i.e., potential of mean force, PMF) for (ultra-)short-chain PFASs from bulk water into β200 and S-D-β200 along the straight-channels²¹, revealing a characteristic two-stage transition in the mass-transfer energy landscape (Fig. 3a, b and Supplementary Figs. 12–15). Stage (1) involved the surface adsorption of PFAS molecules immobilized by zeolite, where the hydrophobic interactions derived the transfer of C-F chains onto the zeolite framework (Window #10 to #3). During this stage, the adsorption free energy of (ultra-)short-chain PFASs decreased significantly (e.g., −10.70 kcal mol^–1 → −26.14 kcal mol^–1 for perfluoroheptanoic acid (PFHpA) in β200, Fig. 3a), although hydrophilic homologs (e.g., trifluoroacetic acid (TFA)) benefited less from this process (Fig. 3c). Stage (2), involves the (ultra-)short-chain PFASs undergoing the pore diffusion process, during which the C-F chain progressed further along the channel while the hydrophilic head remined exposed to water (Window #3 to #0). Unlike the first stage, the energy gain from the interaction between the zeolite framework and the (ultra-)short-chain PFAS was insufficient to offset the substantial dehydration penalty (e.g., 23.72 to 3.81 for PFHpA, Supplementary Fig. 16), resulting in non-spontaneous diffusion within the nanochannel (e.g., −26.14 kcal mol^–1 → 0 kcal mol^–1 for PFHpA in β200). Interestingly, the pore adsorption energy barriers induced by the chain length of (ultra-)short-chain PFAS were markedly disparate in different range (Fig. 3c); for instance, as the chain length increased (chain length ≥4), the interaction energy raised concomitantly (i.e., perfluorobutane sulfonic acid (PFBS) → PFHpA), whereas as the chain length decreased (chain length ≤5), the steric effect became less pronounced (i.e., perfluoropentanoic acid (PFPeA) → TFA).

Fig. 3. Enhanced adsorption mechanism of dynamic hydroxyl cycle.

Potential of mean force calculated from umbrella trajectories when PFHpA enters (a) β200 and b S-D-β200 nanochannels using weighted histogram analysis method, where the specific sampling window is represented above the step diagram; c pore adsorption energy, surface adsorption energy, and total energy changes of (ultra-)short-chain PFASs entering S-D-β200’s and β200’s nanopores; pore size distributions of (d) S-D-β200 and e β200 before and after adsorption of (ultra-)short-chain PFASs; f the average non-covalent interaction (aNCI) and thermal fluctuation index (TFI) analysis of PFHpA with confined water in S-D-β200 nanopores; g the independent gradient model (IGM) analysis of PFHpA with zeolite framework in S-D-β200 nanopores, where the labeled red frameworks were observed to interact with PFHpA; h intrinsic bond (δg) index between (ultra-)short-chain PFASs and S-D-β200 framework; peak G(r) and coordination number N(r) value of hydrophile head (-COO^- and -SO₃^-) with confined water in the first solvent layer.

After enhancement by the cyclic adsorption scheme (Fig. 3b), introducing confined water reduced dehydration penalties (e.g., 7.93 vs. 1.83 for PFHpA, Supplementary Fig. 16) and surface binding energy state (−11.40 kcal mol^–1 vs. −26.14 kcal mol^–1 for PFHpA), resulting in a smoother overall solid-liquid mass transfer energy. Consequently, the simulation results suggest enhanced nanopore accessibility for all (ultra-)short-chain PFASs in S-D-β200, which aligns with the experimental observations showing an increasing in nanopore occupancy from 0.5 to 22% (Fig. 3d, e)²⁵. Additionally, this enhancement generally improved the mass transfer state of PFASs, regardless of chain length and functional groups (Fig. 3c and Supplementary Figs. 12–15). This result differs from previous studies that indicated a hydrophilic binding pattern dominated by confined water²⁰, suggesting the presence of other cooperative binding mechanisms (e.g., zeolite frameworks) in the S-D-β200 pore adsorption process.

To further elucidate the pattern of interaction in S-D-β200 nanopores, the stress mode of (ultra-)short-chain PFASs was investigated. Analysis of average non-covalent interactions and thermal fluctuations indicated that the confined water had a stable hydrogen bonding interaction with the hydrophilic head (i.e., -COO^- or -SO₃^-) of (ultra-)short-chain PFASs (blue isosurface, Fig. 3f and Supplementary Fig. 17). Meanwhile, the independent gradient model analysis revealed that the C-F chains reside in the zeolite a- and b- straight channels, and binding to the zeolite framework through van der Waals forces (green isosurface, Fig. 3g and Supplementary Fig. 18). These findings confirm that (ultra-)short-chain PFASs bind to both the zeolite framework and confined water, forming a unique double-binding pattern, which aligns with the results of 2D-NMR analysis results for S-D-β200 (Fig. 2e and Supplementary Fig. 19). Previous studies have demonstrated that the compact arrangement of long-chain PFASs within β zeolite nanochannels, along with hydrogen-bonded pairs in the intersections, serve as the foundation for their suitable selective adsorption²¹. In this work, through non-immobilized confined water structures, (ultra-)short-chain PFASs can achieve stable adsorption without hydrophilic head pairing, further expanding the application potential of zeolites. Furthermore, we quantified interaction strength of the zeolite framework and the confined water towards PFASs by the sum δg index, radial distribution function (G(r)), and coordination number (N(r)) (Fig. 3h and Supplementary Fig. 20). Interestingly, the confined water interaction strength (G(r)) was negatively correlated with chain length (e.g., C3 > C4) and functional group (e.g., -COO^- > -SO₃^-). In contrast, the interaction strength with the zeolite framework (δg) was positively correlated with chain length (e.g., C4 > C3) and functional group (e.g., -COO^- <-SO₃^-). The difference between these two binding modes may be the potential reason that the enhanced pore adsorption process of S-D-β200 does not specific respond to the PFAS structure. This was demonstrated by further density functional theory analyses, which revealed that the trend of the binding energy change of (ultra-)short-chain PFASs after double binding, achieved by donating electrons (0.389 e–0.457 e) to both the zeolite framework and the confined water, is consistent with that of PMF (Fig. 3c and Supplementary Figs. 21–23). Meanwhile, we further expanded the molecular dynamics model to the equilibrium state (Supplementary Figs. 24–25). The results demonstrated the dominant role of the dual-binding pattern in the adsorption of (ultra-)short-chain PFASs by S-D-β200, especially for chain lengths less than 6 (pore adsorption ratio: 71.1–83.0%). In nearly saturated S-D-β200 nanopores, the hydrophobic tails of (ultra)short-chain PFASs are bound by zeolite channels, whereas the hydrophilic heads converge at confined water clusters and carry cations to eliminate electrostatic repulsion (charge compensation rate: 92.0–101.6%).

Full-scale application

In our scenario, the dynamic hydroxyl cycle scheme enables direct installation of S-D-β200 in existing domestic purifiers, providing a low-cost and in-site resolution to the crisis associated with (ultra-)short-chain PFASs. In a proof-of-concept experiment, we substituted the compressed activated carbon (CAC) adsorption column in the final stage of a full-scale three-stage domestic water purifier with a custom-designed S-D-β200 adsorption column (Fig. 4a, b, and Supplementary Fig. 26). Whole-process simulation experiments were conducted over a continuous period of 182 days, simulating real household water consumption patterns (8 L day^–1, Fig. 4c)²⁸ and using the highest reported concentrations (Supplementary Table 7) of (ultra-)short-chain PFASs in potable water as documented in various studies^29,30. In this full-scale simulation (Supplementary Fig. 27), the dynamic hydroxyl purifier achieved an average removal efficiency for (ultra-)short-chain PFASs ranging from 73 to 95%, which surpasses the performance of the commercial purifier (44–90%) and shows promising potential as a viable alternative to the most efficient reverse osmosis purifiers (75–99%)³¹. The dynamic hydroxyl purifier effectively reduced concentrations of (ultra-)short-chain PFASs in real potable water to meet the proposed EPC health standard (estimated at 450 ng L^–1 based on 90% of reported general abundance levels) within approximately 6 months (179 days)^5,6. This reduction significantly mitigated the potential risk of PFAS exposure associated with the utilization of the commercial purifier (0-day, Fig. 4d). Notably, the evaluation result obtained under the highest PFAS load and showed no effect on other decontamination capacities, such as removal of dissolved organic matter and heavy metal (Supplementary Table 8), indicating its potential to effectively address most potable water treatment demands. Furthermore, the proposed dynamic hydroxyl cycle scheme introduces only water as an enhanced medium in the traditional adsorption-desorption process (Fig. 1a, b) and offers suitable reusability (Supplementary Fig. 9), significantly reducing the overall cost of the adsorption process.

Fig. 4. Full-scale application of dynamic hydroxyl cycle.

a Practical application scenario and optimization scheme of the dynamic hydroxyl cycle; b the full-scale potable water purifier device used in this study; c home simulation operation pattern of the full-scale water purifier in this study, the width of the colored regions represent the duration of the process; d dynamic hydroxyl purifier and commercial purifier for total (ultra-)short-chain PFASs concentration breakthrough curve; the dashed line represents 90% (estimated based on concentration proportion) of the European Parliament and Council (EPC) health limit for total PFAS concentrations in drinking water. Figure 4d was generated with three replicates, and the average values along with their standard deviations were plotted.

Discussion

The cyclic adsorption strategy that we have developed harnesses the dynamic cycle of hydroxyl groups in zeolite to break the forever cycle of (ultra-)short-chain PFASs in real potable water, demonstrated using a series of commercial zeolites, typically β200. The experimental and computational results unequivocally demonstrate the successful construction of non-immobilized hydroxyl groups that facilitate mass transfer and establish a dual-binding mode in pore adsorption. As a result, S-D-β200 achieves an adsorption capacity for (ultra-)short-chain PFASs (C-F number: 1–6) ranging from 233.82 to 733.13 mg g^–1. The reagent-free modification, which relies exclusively on water, effectively preserves the stability of the zeolite’s crystal structure, ensuring suitable selectivity and regeneration capabilities under simulated co-pollution systems. More importantly, the full-scale water purifier utilizing S-D-β200 can reduce the highest level of environmentally relevant (ultra-)short-chain PFAS pollution to levels meeting EPC health standards within approximately 6 months of continuous operation. Therefore, our findings present significant opportunities for affordable, (ultra-)short-chain PFAS-free potable water, especially for individuals in underdeveloped regions.

Methods

Materials

To evaluate the adsorption efficiency of dynamic hydroxyl cycling schemes, 8 typical ultra-short-chain PFASs (C-F number: 1-6), trifluoroacetic acid (TFA, 99%, Aladdin, Shanghai, China), perfluoropropionic acid (PFPrA, 97%, Aladdin, Shanghai, China), perfluorobutanoic acid (PFBA, 98%, Macklin, Shanghai, China), perfluorobutane sulfonic acid (PFBS, 98%, Macklin, Shanghai, China), perfluoropentanoic acid (PFPeA, 97%, Macklin, Shanghai, China), perfluorohexanoic acid (PFHxA, 98%, Macklin, Shanghai, China), perfluorohexane sulfonic acid (PFHxS, 95%, Macklin, Shanghai, China), and perfluoroheptanoic acid (PFHpA, 95%, Macklin, Shanghai, China), were selected in this study. These (ultra-)short-chain PFASs exhibit a range of structural variations, including differences in chain length and functional groups (Supplementary Table 2), which allow for a comprehensive evaluation of the adsorption efficiency for dynamic hydroxyl cycle and the establishment of a fundamental response law governing the structural characteristics of (ultra-)short-chain PFASs. Concurrently, these ultra-short-chain PFAS are markedly prevalent in the environment^4–6 and constitute a substantial proportion of the total PFAS pollution⁶, which is of considerable importance for the assessment of the practical applicability of dynamic hydroxyl cycling.

To evaluate the effect of crystal configuration on dynamic hydroxyl cycling schemes, three commercial zeolites (Zhuoran Environmental Protection Co., Ltd, Dalian, China) with the same Si/Al ratio (200) but different crystal types (Z200 of MFI crystal, β200 of BEA crystal, and Y200 of FAU crystal) was selected to conduct relevant experiments. The pore sizes (supercage) of the three zeolites were 0.51–0.56 nm, 0.56–0.77 nm, and 0.74–1.12 nm, respectively (Supplementary Table 1), creating nanospaces of different sizes for confined water³². Meanwhile, the surface hydrophilicity of zeolite is controlled by the ratio of Si and Al atoms in its framework; the higher the Si/Al ratios, the more hydrophobic the zeolite³³. Therefore, five BEA zeolites with different Si/Al ratios (β25, β68, β150, β200, and β600) were selected for the relevant experiments (Supplementary Table 1), which provide a platform for studying the response of confined water enhanced effect to zeolites’ hydrophilicity (varying Si/Al ratio in the range of 25–600).

Other reagents were purchased in reagent grade from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China) and used without further purification, unless otherwise described.

Dynamic hydroxyl cycling

To simulate the structure of the regenerated zeolites, the commercial zeolites were thermally treated as follows. First, commercial zeolite was stirred in ultra-pure water at room temperature for 4 h, repeated three times, and then dried in a blast drying oven at 70 °C for 12 h, in a vacuum drying oven at −1.0 MPa, dried at 105 °C for 6 h, and finally calcined in a muffle furnace (Thermo Scientific, USA) at 800 °C for 6 h (10 °C min^–1). Defunctionalized zeolites add “D-“ before their names, for example “D-β200”.

To circumvent the hindrance posed by liquid surface tension, a water vapor diffusion method was employed to construct confined water structures at the zeolite nanopores in situ³⁴. The specific steps were as follows: the defunctionalized zeolite was placed in a vacuum drying oven for vacuum degassing for 12 h and then immediately placed in a closed space with high humidity (relative humidity of 90% to 100%) for water vapor diffusion for 24 h to achieve confined water self-assembly. Subsequently, the sample was transferred to pure water and left to stand overnight to stabilize the confined water structure. Finally, excess surface moisture was evaporated at 40 °C for 20–30 min. Confined water-encapsulated zeolites are preceded by the addition of “S-”, for example “S-D-β200”.

Confined water-encapsulated zeolites were evaluated using different adsorption experiments to determine the adsorption performance and behavior for (ultra-)short-chain PFASs, including removal efficiency, adsorption kinetics, adsorption isotherms, and competitive adsorption experiments. Details of experiments and (ultra-)short-chain PFAS measurements are listed in the Supplementary method 1–4.

Saturated zeolites were heated in a muffle furnace at a rate of 10 °C min^–1, until the temperature reached 800 °C. The temperature was maintained at a constant level for 6 h to ensure full degradation of the PFASs. Subsequently, the zeolite samples were allowed to cool foe an adequate period, resulting in a temperature that was consistent with that of the surrounding environment. Finally, related adsorption experiments were performed to determine the removal efficiency of (ultra-)short-chain PFASs.

During the process, the materials were characterized at several stages to identify chemical and other characteristics. The details of the material characterization are available in Supplementary method 5.

Molecular dynamics simulation

To understand the mass energy transfer state and binding mode of (ultra-)short-chain PFASs in zeolite, a “zeolite-bulk water-confined water” three-phase periodic model was constructed, as shown in Supplementary Fig. 11. Specifically, a supercell consisting of 3 × 3 × 3 BEA zeolite crystal cells was placed in the middle of the simulated box, cut along the 010 and 001 crystal faces, and compensated for unsaturated silicon-oxygen bonds with hydroxyl (-OH) groups³⁵. Based on the ζ potential and the Gouy-Chapman formula (Eqs.( 1) and (2)), the surface charge densities of β200 and S-D-β200 were calculated to be –0.11 C m^–2 and -0.105 C m^–2, respectively³⁶. Therefore, two Si-OH groups were randomly deprotonated, resulting in a negatively charged zeolite surface.

$$\mathrm{\sigma }=\text{-}\mathrm{\epsilon }\mathrm{\kappa }\mathrm{\xi }\frac{\sin \mathrm{h}\left(\frac{\mathrm{F}\mathrm{\xi }}{\mathrm{2RT}}\right)}{\frac{\mathrm{F}\mathrm{\xi }}{\mathrm{2RT}}}$$ 1

$${\mathrm{\kappa }}^{\text{-}1}={\left(\frac{\mathrm{\epsilon }\mathrm{RT}}{{\mathrm{2F}}^2{\mathrm{C}}_0}\right)}^{1/2}$$ 2

where ε (6.933 × 10^–10F m^–1) is the permittivity, κ^–1 (m) is the Debye length, $\mathrm{\xi }$ (V) is the ζ potential, F (96485 C mol^–1) is the Faraday constant, R (8.3145 J (mol × K)^–1) is the gas constant, and C₀ (mol m^–3) is the concentration of the electrolyte.

The zeolite structure was constructed according to the BEA zeolite in the International Zeolite Association database, and a certain number of Si atoms was randomly replaced by Al atoms to obtain a β200 molecular sieve model with a Si/Al ratio of 200. The distribution of Si and Al is assigned according to two rules: Lowenstein’s rule requires avoiding Al-O-Al bonds³⁷, and Dempsey’s rule requires the maximum separation of Al atoms given the Si/Al ratio³⁸. After replacement, the generated negative charge is compensated by hydrogen protons, and the specific standard is implemented in ref. ³⁵. Under the periodic boundary condition, water molecules were placed at both ends of the zeolite cutting surface, and their density met the standard of 303 K and 1 atm.

The CVFF model was used to describe the zeolite structure. The (ultra-)short-chain PFAS molecules were constructed utilizing the OPLS-AA model, and Na⁺ was incorporated to neutralize the system charge. The SPC/E model was used to describe the water molecules. The LAMMPS software package was used for molecular dynamics simulation. 12-6 Lennard-Jones empirical force field and geometric mixing rules were used to calculate the van der Waals interactions between the non-bonded atoms. The specific equations are shown in (3), (4), and (5).

$$\mathrm{U}\left({\mathrm{r}}_{\mathrm{ij}}\right){=4\mathrm{\epsilon }}_{\mathrm{ij}}\left[{\left(\frac{{\mathrm{\sigma }}_{\mathrm{ij}}}{{\mathrm{r}}_{\mathrm{ij}}}\right)}^{12}-{\left(\frac{{\mathrm{\sigma }}_{\mathrm{ij}}}{{\mathrm{r}}_{\mathrm{ij}}}\right)}^6\right]$$ 3

$${\mathrm{\epsilon }}_{\mathrm{ij}}=\sqrt{{\mathrm{\epsilon }}_{\mathrm{i}}{\mathrm{\epsilon }}_{\mathrm{j}}}$$ 4

$${\mathrm{\sigma }}_{\mathrm{ij}}=\sqrt{{\mathrm{\sigma }}_{\mathrm{i}}{\mathrm{\sigma }}_{\mathrm{j}}}$$ 5

where r_ij is the distance between particles, σ is the radius of the minimum potential energy, and ε is the depth of the potential well.

For the van der Waals interaction, a 10 Å cutoff distance was implemented. The particle-particle-particle-mesh method was employed to evaluate the long-range Coulomb interaction, with a 10^–4kcal mol^–1 convergence tolerance. All simulations were conducted under NVT conditions, with temperature maintained at 303 K using the Nose-Hoover thermostat. Prior to running the simulation, the system’s energy was minimized. Subsequently, a 1 ns equilibration was performed using a 1 fs time step.

Potential of mean force (PMF) calculation

To further obtain the binding free energy variation of the adsorption pathway of (ultra-)short-chain PFASs into nanopores, the PMF was calculated using the umbrella sampling method^39–41. The process of (ultra-)short-chain PFASs adsorbing from bulk water onto zeolite nanopores was segmented into an average of 11 reaction windows, with each window separated by 2 Å along the Z axis. Once equilibrium was achieved in the system, a steered molecular dynamics technique employing a harmonic potential (spring constant of 25 kcal mol · Å^–1, tension velocity of 0.001 Å fs^-1) was utilized to drag the (ultra-)short-chain PFASs into a designated reaction window. Subsequently, the (ultra-)short-chain PFASs were stabilized at the reaction window for a 300 ps equilibrium period using a harmonic potential with a low spring constant (2 kcal (mol × Å)^–1). The simulation was then extended for an additional 100 ps, during which the PFAS reaction coordinates were documented every 100 fs. The data from the reaction coordinate was transformed into PMF free energy by weighted histogram analysis⁴¹:

$${\mathrm{W}}_{\mathrm{window}}=-{\mathrm{K}}_{\mathrm{b}}\times \mathrm{T}\times {\mathrm{lnP}}_{\mathrm{window}}$$ 6

where K_b is the Boltzmann constant, T is the temperature, and P_window is the probability of finding (ultra-)short-chain PFASs at the observation window.

Average non-covalent interaction (aNCI) calculation

The aNCI method proposed by Wu et al.⁴² was used to analyze the average non-covalent interactions of (ultra-)short-chain PFASs in dynamic systems. The principle of the aNCI method is to show the weak interaction region by the reducing density gradient (RDG) function isosurface and then project the sign(λ₂)rho function onto the RDG isosurface of different colors to show the type of weak interaction⁴³. The electron density, electron density gradient, and electron density Hessian matrix used to calculate the RDG and sign(λ₂)rho functions were obtained by averaging the multi-frame trajectory structure⁴³. Blue, green, and red represent hydrogen bond, van der Waals interactions, and steric effects, respectively. The average RDG was calculated using the following Eq. (7):

$$\overline{\mathrm{RDG}\left(\mathrm{v}\right)}=\frac{1}{2{\left({3\mathrm{\pi }}^2\right)}^{1/3}}\frac{\overline{\mid \nabla \mathrm{\rho }\left(\mathrm{r}\right)\mid }}{{\overline{\mathrm{\rho }\left(\mathrm{r}\right)}}^{4/3}}$$ 7

where $\rho \left(r\right)$ is the average electron density, $\nabla \rho \left(r\right)$ is the average electron density gradient. Additionally, the average RDG method defines a thermal fluctuation index (TFI) to represent the stability of weak interactions in different regions⁴². TFI is mapped to the average RDG equivalence surface using different colors, and the stability of the weak interaction in other regions can be judged by the colors⁴³. A bluer isosurface indicates a more stable interaction, whereas a redder isosurface indicates an unstable one. TFI was calculated using Eq. (8):

$$\mathrm{TFI}=\frac{\mathrm{std}\left(\left\{{\mathrm{\rho }}_{\mathrm{i}\left(\mathrm{r}\right)}\right\}\right)}{\mathrm{a}\mathrm{\nu }\mathrm{g}\left(\left\{{\mathrm{\rho }}_{\mathrm{i}\left(\mathrm{r}\right)}\right\}\right)}$$ 8

where avg({ρ_i(r)}) and std({ρ_i(r)}) are the mean and variance, respectively, of the electron density of the entire trajectory. The optimal structure for aNCI analysis was obtained by molecular dynamics simulations. The (ultra-)short-chain PFAS molecules were fixed, and equilibrium simulations were performed for 1 ns. The aNCI analysis used structures from the last 2000 frame trajectories. All the aNCI and TFI calculations were performed using the Multiwfn package (version 3.8).

Independent gradient model (IGM) calculation

In this study, we used the IGM method proposed by Hénon et al.⁴⁴ to analyze the interactions between (ultra-)short-chain PFASs and zeolite framework. In IGM analysis, the δg^inter isosurface is often used to visualize the interaction region. Meanwhile, the sign(λ₂)ρ function is usually mapped in different colors on the isosurface to intuitively characterize the nature of the interaction. The sign(λ₂)ρ represents the product of the sign(λ₂), the second-largest eigenvalue of the Hessian matrix of electron density ρ, and ρ. The ρ of the interaction region is generally positively correlated with the interaction strength, and the sign(λ₂) has the ability to distinguish whether the interaction is attractive or repulsive⁴⁵. The color criterion of the IGM analysis was consistent with aNCI. δg^inter was calculated using Eqs. (9), (10), and (11):

$${\mathrm{g}}^{\mathrm{inter}}\left(\mathrm{r}\right)=∣\sum _{\mathrm{A}}{\sum }_{\mathrm{i}\in \mathrm{A}}\nabla {\mathrm{\rho }}_{\mathrm{i}}^{\mathrm{free}}\left(\mathrm{r}\right)∣$$ 9

$${\mathrm{g}}^{\mathrm{IGM},\mathrm{inter}}\left(\mathrm{r}\right)={\sum }_{\mathrm{A}}∣{\sum }_{\mathrm{i}\in \mathrm{A}}\nabla {\mathrm{\rho }}_{\mathrm{i}}^{\mathrm{free}}\left(\mathrm{r}\right)∣$$ 10

$${\mathrm{\delta }\mathrm{g}}^{\mathrm{inter}}\left(\mathrm{r}\right){=\mathrm{g}}^{\mathrm{IGM},\mathrm{inter}}\left(\mathrm{r}\right)\text{-}{\mathrm{g}}^{\mathrm{inter}}\left(\mathrm{r}\right)$$ 11

where g^inter is the size of superposition of all debris density gradients, g^{IGM, inter} is the sum of the size of all debris density gradients, A loops through all fragments, and i loops through all atoms in the corresponding fragment. IGM calculations were performed using the Multiwfn package (version 3.8).

Radial distribution function (RDF) calculation

To investigate the intensity of the interaction between the confined water and (ultra-)short-chain PFASs, the RDF between their atomic pairs were calculated^46,47. The RDF estimates the density of a pair of atoms at a given distance using Eq. (12):

$$\mathrm{G}\left(r\right)=\mathrm{V}\mathrm{\rho }\left(r\right)/4\mathrm{N}\mathrm{\pi }{r}^2\Delta r$$ 12

where V is the system volume, ρ(r) is the number of atomic pairs found in the range of r to r + ∆r shell, and N is the total number of selected atomic pairs. The G(r) peak indicates the frequency of the occurrence of the atom pair and can be used to infer the interaction strength between the two atoms.

Coordination number calculation

The coordination number represents the number of corresponding molecules within each spherical coordination range of the selected atom^46,47. Specifically, integrating G(r) in spherical coordinates with the first minimum value of the RDF yields the coordination number of the corresponding molecule. The specific Eq. (13) for calculating the coordination number is as follows.

$$\mathrm{N}\left(r^{\prime }\right)=4\pi \mathrm{\rho }\underset{0}{\overset{r\prime }{\int }}\mathrm{G}\left(r\right)r^2\mathrm{d}r$$ 13

where r’ is the cutoff distance of the solvent layer, ρ is the system density, and G(r) is the radial distribution function of the selected atomic pair.

Density functional theory (DFT) calculation

The DFT calculations were carried out using DMol3 modules of Material Studio. The generalized gradient approximation (GGA) method with Perdew-Burke-Ernzerhof (PBE) function was employed to describe the interactions between core and electrons. Spin-polarization with DFT semi-core Pseudopots was included in all calculations. It is essential to take into account the van der Waals forces for adsorption. Hence, the Tkatchenko-Scheffler (TS) method for DFT-D correction was used for this purpose. Double numerical plus polarization (DNP) functions were adopted as the basis set. The force and energy convergence criterion were set to 0.002 Ha Å^-1 and 10^-5Ha, respectively. When the optimization was completed, the binding energy (ΔE) was calculated as Eq. (14).

$$\mathrm{\Delta }\mathrm{E}=27.212\times \left({\mathrm{E}}_{\mathrm{total}}\text{-}{\mathrm{E}}_1\text{-}{\mathrm{E}}_2\right)$$ 14

where the E_total is the energy of the optimized system; E₁ is the energy of the S-D-β200; E₂ is the energy of (ultra-)short-chain PFASs.

Full-scale water purification

In this study, a full-scale water purifier was constructed to evaluate the removal efficiency of (ultra-)short-chain PFASs from potable water using S-D-β200. The full-scale water purifier employed a three-stage treatment process (size: 36 cm × 33 cm × 13 cm), including polypropylene (PP) cotton filtration (filtration accuracy: 1 μm, size: 25.4 cm×6 cm, Anhui Letou Water Purification Technology Co., Ltd., China), granular activated carbon (size: 25 cm × 7 cm, Anhui Letou Water Purification Technology Co., Ltd., China), and the confined water enhanced zeolite (i.e., S-D-β200, size: 25 cm × 7 cm). A commercial household water purifier was also tested as a control, including PP cotton filtration, granular activated carbon, and compressed activated carbon.

The experimental operation mode of the full-scale water purifier was based on the water intake pattern of an ordinary 4-member family, as shown in Fig. 4c. Potable water treatment was carried out at breakfast, lunch, and dinner times every day for 0.5 h, 2 h, and 1.5 h respectively. The influent flow rate was controlled at 2 L h^-1 to obtain the recommended drinking water intake of 8 L per day for 4 people²⁸. In addition, a 10-min backwash was carried out in the early morning hours of each day to prevent adsorption column clogging and biofilm formation. To prevent nanosized zeolite particles from causing excessive pressure drop in the adsorption column, commercial β200 spherical zeolite (particle size: 0.4–0.9 mm, Dalian Zhuoran Environmental Protection Technology Co., Ltd., China) was used in the full-scale water purifier.

Author contributions

Y.J.S., H.D.H., and H.Q.R. conceived the study. Y.J.S. and H.D.H. designed experiments. H.Q.R. conducted and supervised all experiments and data collection. Y.J.S. prepared materials, performed cyclic adsorption experiments, molecular dynamics simulation, and density functional theory analysis. M.H.Y. and H.X.M. assisted with material characterization, cyclic adsorption experiments, and (ultra-)short-chain PFAS measurement. Y.J.S. and H.D.H. reviewed and interpreted the results. Y.J.S. and H.D.H. wrote the original draft. All authors reviewed and commented on the paper.

Peer review

Peer review information

Nature Communications thanks Dhara Trivedi, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data are available in the main text or the supporting information. Should any raw data files be needed in another format, they are available from the corresponding authors upon request. Source data have been deposited in the Figshare database (10.6084/m9.figshare.31123411)⁴⁸.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-70507-y.