Confined water–encapsulated activated carbon for capturing short-chain perfluoroalkyl and polyfluoroalkyl substances from drinking water

Significance

Adsorption is the most established method used to block human exposure through drinking water to the notorious perfluoroalkyl and polyfluoroalkyl substances (PFASs), yet is challenged by the increasing short-chain PFAS crisis due to the inherent defects of solid–liquid mass transfer. Herein, by assembling the confined water structure in situ in hydrophobic nanopores, we introduce a dual-drive mode in the activated carbon to completely eliminate the mass transfer barrier and dramatically enhance its adsorption performance for various short-chain PFASs. Significantly, the methodology demonstrated is a potential in situ upgrade of existing adsorption devices. This work will thus revolutionize the understanding of the role of confined water in mass transfer and enable an in situ solution for the short-chain PFAS crisis.

Abstract

The global ecological crisis of perfluoroalkyl and polyfluoroalkyl substances (PFASs) in drinking water has gradually shifted from long-chain to short-chain PFASs; however, the widespread established PFAS adsorption technology cannot cope with the impact of such hydrophilic pollutants given the inherent defects of solid–liquid mass transfer. Herein, we describe a reagent-free and low-cost strategy to reduce the energy state of short-chain PFASs in hydrophobic nanopores by employing an in situ constructed confined water structure in activated carbon (AC). Through direct (driving force) and indirect (assisted slip) effects, the confined water introduced a dual-drive mode in the confined water–encapsulated activated carbon (CW-AC) and completely eliminated the mass transfer barrier (3.27 to 5.66 kcal/mol), which caused the CW-AC to exhibit the highest adsorption capacity for various short-chain PFASs (C-F number: 3-6) among parent AC and other adsorbents reported. Meanwhile, benefiting from the chain length– and functional group–dependent confined water–binding pattern, the affinity of the CW-AC surpassed the traditional hydrophobicity dominance and shifted toward hydrophilic short-chain PFASs that easily escaped treatment. Importantly, the ability of CW-AC functionality to directly transfer to existing adsorption devices was verified, which could treat 21,000 bed volumes of environment-related high-load (~350 ng/L short-chain PFAS each) real drinking water to below the World Health Organization’s standard. Overall, our results provide a green and cost-effective in situ upgrade scheme for existing adsorption devices to address the short-chain PFAS crisis.

Perfluoroalkyl and polyfluoroalkyl substances (PFASs) have been extensively used in manufacturing and consumer goods for decades but are now garnering global interest as top priority pollutants, with >1,400 individual chemicals listed in the Toxic Substances Control Act Inventory (1, 2). As concerns about the ecological persistence, accumulation, and health hazards of PFASs increase, the most widely used long-chain PFASs (chain length >7), perfluorooctane sulfonate and perfluorooctanoate, have been restricted in production and use by the Stockholm Convention (3). Instead, more unrestricted shorter chain PFASs (chain length: 4 to 7) were produced as alternatives to the long-chain PFASs (4, 5), resulting in the dominance of short-chain PFASs in water environments, especially in drinking water (6, 7). To make matters worse, short-chain PFASs cause higher human exposure through drinking water and have longer half-lives than long-chain PFASs (8), which is associated with poor resistance to infectious diseases and may render the public more susceptible to the recent COVID-19 pandemic (9).

Despite intensive efforts, removing short-chain PFASs from drinking water remains a challenge. Traditional chemical and biological methods cannot completely mineralize or defluorinate the strong carbon-fluorine bond (536 kJ/mol), resulting in more short-chain PFAS residues (9, 10). Currently, adsorption is the most widely established PFAS treatment technology, especially activated carbon (AC) adsorption, which presents an opportunity to break the forever PFAS cycle (1, 2, 11). However, it is well known that hydrophilic short-chain PFASs have lower potential energy in water, leading to a preference for remaining in water rather than migrating to adsorbents (12). For instance, the removal efficiency of AC for PFASs in drinking water dropped sharply from ~80% (long-chain) to ~40% (short-chain), and short-chain PFASs escaped rapidly from actual adsorption devices (13, 14). To date, a restricted number of publications have addressed short-chain PFAS adsorption via using novel adsorption materials (e.g., β-cyclodextrin polymer, covalent organic frameworks, and magnetic fluorinated polymer), but these schemes achieved only marginal efficiency gains over AC and cannot be implemented currently due to unsuitable for existing facilities (15–19). Hence, this study focuses on improving the adsorption affinity of AC and in situ upgrading the existing adsorption devices toward solving the short-chain PFAS crisis.

Reducing the potential energy of short-chain PFASs in AC to a level below that in water enables the spontaneous generation of the adsorption process. However, current strategies generally require using a large number of reagents and stringent conditions to graft hydrophilic groups (e.g., fluorination and amination), resulting in unbearable cost multiplication and secondary pollution (12, 20, 21). Thus, the development of greener and low-cost solutions is needed soon. In nature, mimicry predators, such as Venus flytraps, accomplish their trapping behavior by creating an environment that the prey prefers. Inspired by this mimicry capture strategy, we hypothesized that constructing the preferred water environment (mimicry environment) in AC (predator) could reduce the potential energy of short-chain PFASs (prey) in AC and improve its capture ability. This conjecture is based on the following two theoretical foundations: i) Water molecules in nanopores are identified as confined water, which is restricted in mobility and features unique properties over bulk water (e.g., reduced hydrogen bonds and low dielectric constant) depending on the characteristics of the nanopore (22, 23). Consequently, an opportunity to construct a preferred water environment for short-chain PFASs is provided due to the uniqueness and controllability of confined water. ii) AC provides an effective in situ nanospace for the construction of a preferred water environment because there is not enough pressure (>80 MPa) during water treatment for liquid water molecules to occupy the hydrophobic nanopores of AC (24, 25). This is an attempt to exploit the potential of confined water for enhanced adsorption of short-chain PFASs, and the realization of this conjecture will provide means of achieving reagent-free, low-cost modification of adsorbents since only water is used.

Furthermore, the enhanced adsorption mechanism of confined water needs further clarification, because conventional wisdom holds that water mainly acts as a hindrance in regard to mass transfer, and the adsorbent must be intensively dried prior to use (26–28). The function of confined water may overturn the traditional understanding of the solid–liquid mass transfer process, but direct observational technologies are not yet available owing to the reduced dimensions of the pores (23, 29). Recently, computational materials science [e.g., molecular dynamics (MD) simulation] has successfully explained the microstructure and energy state changes of confined water in proton transport at the atomic scale (30, 31). Ma et al. (32) simulated the number of hydrogen bonds formed in different media (i.e., water, methanol, and ethanol) based on MD, arguing that confined water mainly provides hydrogen bond sites to adsorb pollutants. Nevertheless, by simulating only the equilibrium state in an unconfined system, the results cannot eliminate concerns regarding mass transfer obstruction and the lack of key information on the factors influencing confined water application.

Herein, we constructed confined water environments in AC nanopores in situ in a reagent-free and low-cost manner via water vapor diffusion at room temperature and atmospheric pressure, and verified that the confined water–encapsulated activated carbon (CW-AC) adsorption performance was improved for four short-chain PFASs with typical hydrophilicity. MD simulation further provided a detailed molecular-scale view and the energy state of short-chain PFASs entering the AC nanopores, and a confined water–enhancing mechanism and binding patterns were proposed to support the experimental results. Finally, in an in situ upgrade proof-of-concept experiment, the CW-AC column significantly outperformed AC via effective short-chain PFAS removal to below the health advisory level from real drinking water over 150 d of simulated use.

Results and Discussion

Confined Water-Encapsulation for CW-AC.

Based on the fact that AC spontaneously adsorbs water clusters in air (33), we induced water vapor to complete the self-assembly of the confined water structure in the nanopores of AC by exposure to high humidity (Fig. 1A). A commercial AC (Macklin Biochemical Co. Ltd.) with a pure microporous structure was selected as the substrate (Fig. 1 B and C and SI Appendix, Table S1), which provided a better environment for confined water formation relative to mesopores (34). More information is available in the Materials and Methods.

Fig. 1.

Confined water encapsulation for CW-AC. (A) Confined water–encapsulation process for commercial AC. (B) The N₂ adsorption and desorption isotherm and pore size distribution of AC. (C) The scanning electron microscope (SEM) and energy-dispersive X-ray spectroscopy images of AC. (D) The FTIR spectrum (Left) of AC with different treatments and –OH stretching subband (Right) of confined water in the CW-AC (W-AC was AC treated by shaking in water).

To confirm the feasibility of encapsulating confined water in AC nanopores via vapor diffusion at room temperature and atmospheric pressure, the changes in the Fourier transform infrared spectroscopy (FTIR) spectra of CW-AC and AC were investigated (Fig. 1D). The virgin AC spectrum presented characteristic peaks at five locations, including –CH (2,665 cm⁻¹ and 1,180 cm⁻¹), –C=C– or –C≡C– (2,400 to 1,900 cm⁻¹ and 1,552 cm⁻¹), and –C=O (1,780 cm⁻¹) (28, 35). Neither polar nor ionizable functional groups (e.g., –OH or –NH) were found, indicating the strong hydrophobic and electroneutral character of commercial AC, which is also confirmed by the neutral pH and ζ-potentials of AC (SI Appendix, Tables S1). After confined water encapsulation, the spectra of CW-AC showed a new broad –OH band at 3,000 to 3,700 cm⁻¹, which could be assigned to hydrogen bond networks of water (36). Fourier deconvolution revealed that high-wavelength w1 (at 3,620 to 3,660 cm⁻¹) and w2 (at 3,510 to 3,560 cm⁻¹) occupy the largest proportion in the –OH band (SI Appendix, Table S2), indicating that most of the intermolecular hydrogen bonds of water are severely broken in CW-AC without forming a tetrameric structure like bulk water (34, 36). This network of universally broken hydrogen bonds was consistent with the confined water structure in highly confined nanospaces such as hydrogels and phospholipids (34, 36), which further proved that the water molecules in the CW-AC exist in nanopores rather than being attached to the surface. Collectively, these results demonstrated that water molecules were successfully encapsulated into CW-AC nanopores and formed a highly fractured hydrogen bond network of confined water in situ.

Meanwhile, we excluded the influence of liquid water molecules on the formation of the CW-AC confined water structure, because the –OH characteristic peak was not observed in the FTIR spectra of water-equilibrated AC without water vapor diffusion (Fig. 1D, W-AC). This is consistent with previous reports that liquid water cannot break through the control of surface tension to spontaneously form confined water structures in hydrophobic pores (24, 25). Moreover, the confined water structure is stable at the molecular level due to the presence of molecular congestion mechanisms (25), which is confirmed by long-term (10 ns) MD simulations on the water prefilled nanopore model of CW-AC (SI Appendix, Fig. S2). Therefore, in the water treatment process, the confined water structure would affect the adsorption behavior of CW-AC and distinguish it from virgin AC without confined water in the nanopores.

Adsorption Performance of CW-AC on Short-Chain PFASs.

Motivated by the mimic capture strategies of predators in nature, we examined whether CW-AC could function effectively in the capture of short-chain PFASs from water. Four different short-chain PFAS (C-F number: 3-6) adsorption isotherm tests were performed to evaluate the adsorption capacities of CW-AC (Fig. 2A), including perfluorohexane sulfonic acid (PFHxS), perfluorohexanoic acid (PFHxA), perfluorobutane sulfonic acid (PFBS), and perfluorobutanoic acid (PFBA). These species have different chain lengths (C4 or C6) and functional groups (–COOH or –SO₃H) that represent typical variations in the hydrophilicity (log K_OW) of short-chain PFASs (Table 1). The Langmuir–Freundlich model reasonably described all data, as indicated by the high R² values (0.985 to 0.999, SI Appendix, Table S3). Significantly, the maximum uptake capacity of CW-AC for short-chain PFASs ranged between 374.86 mg/g and 553.08 mg/g, which surpassed other adsorbents reported thus far for short-chain PFAS adsorption; for instance, it was nearly 2.2 to 27.9 times higher than that for the covalent triazine–based framework (37) or that for IRA 67 (38) (Fig. 2B). These results ranked CW-AC among the top candidates for benchmarked short-chain PFAS adsorbents.

Fig. 2.

Adsorption performance of CW-AC for short-chain PFASs. (A) Adsorption isotherms (303 K) for PFHxS, PFHxA, PFBS, and PFBA on CW-AC and AC; the error bars in this figure represent the SD (n =3 independent experiments). (B) The adsorption capacity of short-chain PFASs for CW-AC with other adsorbents, including covalent triazine–based framework (37), AC Calgon (37), GAC F400 (39), IRA 67 (38), and bamboo-derived AC (BAC) (38). (C) Maximal adsorption capacity for short-chain PFASs on CW-AC and AC. (D) The plots of the adsorption capacity increment and the Gibbs free energy variation versus the short-chain PFAS log K_OW. The error bars of Fig. 2 B–D represent the SE of Langmuir–Freundlich model fitting results.

Table 1.

The structure and chemical characterization of four short-chain PFASs in this study

PFASs	Structure*	C-F number†	Chain length	Log Kow‡	pKa‡	Min/max PD‡,§ (nm)
PFHxS		6	6	4.03	−3.32	0.68/1.41
PFHxA		5	6	3.71	−0.78	0.72/1.10
PFBS		4	4	2.63	−3.31	0.70/1.00
PFBA		3	4	2.31	1.07	0.64/0.92

^*Different colored ball represents different atoms, C atom: blue ball, O atom: red ball, S atom: yellow ball, F atom: pink ball.

^†C-F number: number of fluorinated carbons.

^‡Data estimated through MarvinSketch 18.11.0 (ChemAxon Ltd.) (14).

^§PD: projection diameter.

Control experiments with parent AC were also conducted (Fig. 2 A and C and SI Appendix, Table S3), and CW-AC exhibited a higher adsorption capacity (1.4 to 2.7 times) regardless of short-chain PFAS hydrophilicity, indicating the universality of the enhanced effect achieved by confined water encapsulation. However, the enhancement effect was not equal among the different types of short-chain PFASs (Fig. 2D); a lower log K_OW (more hydrophilic) led to a higher increment of adsorption capacity, i.e., PFBA (244.21 mg/g) >PFBS (189.26 mg/g) >PFHxA (179.89 mg/g) >PFHxS (162.92 mg/g). This was different from the widely revealed affinity order dominated by AC hydrophobic interactions (14, 40), and in turn weakened the strong linear relationship between the adsorption capacity and hydrophobicity of short-chain PFASs (SI Appendix, Fig. S3). At the same time, the thermodynamic advantages of hydrophilic short-chain PFASs were also observed (Fig. 2D). The short-chain PFASs with higher hydrophilicity possessed a higher Gibbs free energy decline, suggesting that the confined water environment in CW-AC was more conducive to improving the energy state of hydrophilic short-chain PFASs. It is worth mentioning that the affinity shifted to hydrophilic short-chain PFASs would promote the practical application of CW-AC in real drinking water, where hydrophilic short-chain PFASs tend to escape treatment and result in a higher exposure risk (13, 18, 41).

Adsorption selectivity and regenerability are critical for practical applications of water management and purification, especially for those with high natural organic matter and ionic strength (42, 43). Thus, a competitive adsorption and desorption experiment involving short-chain PFASs (1 mg/L each) in the presence of simulated drinking water cocontaminants (humic acid = 2 mg/L, ionic strength = 0.01 M) was performed. The adsorption capacity of CW-AC decreased by only 13.0 to 16.6% under the copollution system, exhibiting stronger selectivity and functional stability than their parent AC (23.1 to 31.0%, SI Appendix, Fig. S4). Significantly, the CW-AC could also be readily regenerated by washing with 20% MeOH solution, which recovered 71.6 to 74.7% of the adsorption capacity in the first cycle and remained almost identical for subsequent cycles 2, 3, and 4 (SI Appendix, Fig. S5). The suitable selectivity and regenerability could be attributed to confined water reducing the adsorption of humic acid on AC (8.8 to 13.8%, SI Appendix, Fig. S4), which would help alleviate nanopore blocking (44). Therefore, the selectivity and regeneration benefit of CW-AC suggested its potential feasibility in practical applications.

In addition, to verify the universality of the confined water–encapsulation strategy on different commercial ACs, two other ACs with different pore size distributions (SI Appendix, Fig. S6 and Table S1), AC-87% (micropore distribution ratio: 87%) and AC-79% (micropore distribution ratio: 79%), were employed for confined water encapsulation (SI Appendix, Fig. S7 and Table S2), isothermal adsorption tests (SI Appendix, Figs. S8–S10 and Table S4), competitive adsorption (SI Appendix, Fig. S11), and regeneration experiments (SI Appendix, Fig. S12) as well. To our delight, the adsorption capacity (SI Appendix, Fig. S10), selectivity (SI Appendix, Fig. S11), and regenerability (SI Appendix, Fig. S12) of both two CW-AC for short-chain PFASs were significantly increased regardless of the type of parent AC, and the affinity shifted toward hydrophilic short-chain PFASs (SI Appendix, Fig. S13). These results demonstrated that the confined water–encapsulation strategy could be easily extended to various commercial ACs with a wide range of nanopore structural differences, and provide stable adsorption enhancement and affinity shift effects. We then further discuss the enhanced adsorption and affinity shift mechanism of CW-AC.

Enhanced Mechanisms for the Adsorption Capacity of CW-AC.

To obtain more insights into the mechanisms through which the adsorption of short-chain PFASs is enhanced in confined water and alleviate concerns about the mass transfer hindrance of confined water, we collected the adsorption pathways of short-chain PFASs into CW-AC and AC nanopores via all-atom MD simulation and calculated their binding free energy (i.e., potential of mean force, PMF). A rigid carbon nanotube and graphene sheets were used to mimic the hydrophobic nanopore-wall structure of AC, and the average pore size of the experimental AC was selected as the diameter of the carbon nanotube, i.e., 1.6 nm. The structural differences in CW-AC and AC were reflected using the presence or absence of prefilled water in nanopores (SI Appendix, Fig. S2). More information about the atomistic modeling and computational methodology is available in the Materials and Methods.

The energy state difference between the adsorption paths of CW-AC and AC is clearly reflected, regardless of the short-chain PFAS type (Fig. 3 A and B and SI Appendix, Fig. S14 and Table S5). All short-chain PFAS adsorption pathways of CW-AC and AC could be divided into two stages. Stage (1) exhibited a rapid drop in binding free energy, which is irrelevant for the encapsulation of confined water (e.g., 12.63 kcal/mol → −5.66 kcal/mol for PFBA into CW-AC, Fig. 3A) or not (e.g., 19.03 kcal/mol → 6.61 kcal/mol for PFBA into AC, Fig. 3B). In this stage, the hydrophobic tails (C-F chains) of short-chain PFASs first approach the CW-AC or AC surface (the hydrophilic heads remain buried in bulk water) due to the strong hydrophobic interaction between the CW-AC or AC surface and the hydrophobic tails (10, 42). Stage (2) exhibited a distinct difference: When there was no confined water, the short-chain PFAS hydrophobic tail still first extended into the AC nanopore, whereas the hydrophilic head drove the bulk water molecules to the rear (such as being pulled by a chain of water molecules). This suggested that although AC hydrophobicity continued to drive the short-chain PFAS hydrophobic tail into the nanopore, the strong interaction of bulk water molecules on the hydrophilic head hindered the mass transfer process; therefore, the adsorption pathway showed energy barriers (e.g., −5.66 kcal/mol → 0 kcal/mol for PFBA into AC, Fig. 3B). In contrast, the presence of confined water and CW-AC together formed a more stable double-binding morphology with short-chain PFASs during the adsorption pathway (42), i.e., the hydrophilic head of the short-chain PFASs is buried in the confined water, whereas the hydrophobic tail is close to the nanochannel wall, so its adsorption pathway exhibits a continuous decline in the binding free energy (e.g., 6.4 kcal/mol → 0 kcal/mol for PFBA into CW-AC, Fig. 3A). Collectively, given the global change in PMF free energy along the adsorption pathway (Fig. 3C), the mass transfer of short-chain PFASs into CW-AC nanopores had a universal thermodynamic advantage (−19.02 to −31.95 kcal/mol vs. −12.62 to −28.62 kcal/mol), and the mass transfer energy barriers (3.27 to 5.66 kcal/mol) were completely eliminated.

Fig. 3.

Enhanced mechanisms for the adsorption capacity of CW-AC. PMF profiles computed from umbrella sampling trajectories using the WHAM for PFBA entry to (A) CW-AC and (B) AC; snapshots highlighted by the dashed box represent five specific observed windows. (C) PMF free energy and energy barriers for short-chain PFASs entry to CW-AC and AC. (D) Average Z-force energy for short-chain PFASs at key window #4 on CW-AC and AC. (E) The average noncovalent interaction analysis of the interactions between PFBA and water at key window #4 on CW-AC (Left) and AC (Right).

The decline in mass transfer energy states indicated that confined water introduced a driving mode for short-chain PFASs into CW-AC; therefore, we selected key nodes on the adsorption pathway to perform force analysis of short-chain PFASs. At the observation window 4 (the nanopore interface), the morphologies of the PFASs in CW-AC and AC showed a significant difference (vertical to parallel), suggesting a dramatic change in the force mode introduced by confined water. Therefore, the forces in the z direction exerted by water molecules (bulk water and confined water) and adsorbents on short-chain PFASs in window #4 were collected in an unbiased sampling manner (SI Appendix, Fig. S15), and the average and resultant forces were calculated (Fig. 3D and SI Appendix, Table S6). For virgin AC, adsorbent and water played the traditional mass transfer modes of driving (−0.95 to −1.64 kcal/mol Å) and hindering (2.45 to 2.86 kcal/mol Å), respectively, so the resultant force was manifested as resistance (1.23 to 1.60 kcal/mol Å) (10, 45). In CW-AC, the hindering force of water molecules on short-chain PFASs was significantly reduced (1.51 to 2.50 kcal/mol Å) relative to AC, indicating that confined water directly participates in the mass transfer drive to counteract the hindering force of bulk water. The average noncovalent interactions and thermal fluctuation index analysis by Multiwfn further confirmed that confined water acted simultaneously on the hydrophilic head and hydrophobic tail of short-chain PFASs with fluctuant van der Waals forces (blue isosurface and spike) and stable hydrogen bonds (green isosurface and spike) (46, 47), respectively, to directly generate driving forces (Fig. 3E and SI Appendix, Fig. S16). This also explained why the short-chain PFAS head and tail underwent significant rotation parallel to the interface (Fig. 3A, window #4). In addition, it is quite unexpected to find that the driving force of the adsorbent was improved in CW-AC (−1.08 to −1.87 kcal/mol Å) because the confined water encapsulation had no effect on the surface properties of the CW-AC model (SI Appendix, Fig. S2). This might be related to the widespread revelation of confined water–assisted slip effects (48, 49); that is, confined water reduced the surface energy of the hydrophobic nanopore and caused the short-chain PFASs to slip frictionless along the hydrogen bond chain (SI Appendix, Fig. S17). It should be mentioned that this effect strongly depends on the highly fractured hydrogen bond network of confined water at the micropore scale (48), so it only played a driving role (SI Appendix, Figs. S19−S21 and Table S8) without having the assisted slip effect (SI Appendix, Figs. S17 and S20 and Table S8) in the simulation of the typical mesopore (2.7 nm) all-atomic model (SI Appendix, Fig. S18 and Table S7) (50), which was consistent with the result that pure microporous CW-AC exhibited the highest adsorption capacity (SI Appendix, Fig. S10). Overall, the mass transfer process of short-chain PFASs in CW-AC exhibits an adsorbent-confined water dual-driving mode, which greatly reduces the resistance (0.43 to 0.84 kcal/mol Å) compared to AC.

In summary, the enhanced adsorption performance of CW-AC could be attributed to the fact that the confined water directly (driving force) and indirectly (assisted slip) participated in the mass transfer process of short-chain PFASs into the nanopores, which led to the improvement of the energy states as well as the complete elimination of the mass transfer energy barrier (3.27 to 5.66 kcal/mol).

Shift Mechanisms for the Adsorption Affinity of CW-AC.

The aforementioned mechanisms revealed the enhanced adsorption of CW-AC well, but the affinity shift was still not well understood (Fig. 2D). We therefore further investigated the binding patterns of different PFASs with confined water by computing the radial distribution functions and coordination numbers at the CW-AC nanopores (SI Appendix, Figs. S22–S25). It is evident from Fig. 4A and SI Appendix, Table S9 that the chain length and functional group of short-chain PFASs controlled the binding mode of water molecules to their hydrophobic tails and hydrophilic heads, respectively, and both preferred the more hydrophilic species. For instance, irrespective of the length of the short-chain PFAS molecule, the interaction intensity [G(r) peak] and coordination numbers of the confined water around the carboxyl groups of PFBA (5.30 & 3.45) and PFHxA (4.99 & 3.70) were greater than those of the sulfonic acid groups of PFBS (4.76 & 2.64) and PFHxS (3.49 & 2.39). Similarly, the interaction intensity and coordination numbers of confined water on the hydrophobic tail showed that the C4 (PFBA: 1.62 & 9.70 and PFBS: 1.24 & 9.27) was higher than the C6 (PFHxA: 1.15 & 7.92 and PFHxS: 0.93 &7.57), regardless of the functional group type. As a result, the confined water formed a stronger association with more hydrophilic short-chain PFASs, and the hydrogen bond formation probability (Fig. 4B) exhibited the order of PFBA (0.518)> PFBS (0.503)> PFHxA (0.481)> PFHxS (0.460), which was consistent with our experimental results (Fig. 2D).

Fig. 4.

Shift mechanisms for the adsorption affinity of CW-AC. (A) The G(r) peak value and coordination numbers for the hydrophobic tail (–SO₃⁻ or –COO⁻) and the hydrophobic tail (CF₃) of short-chain PFASs and confined water at the first and second solvent layers in CW-AC. (B) Hydrogen bonding formation probability of short-chain PFASs in CW-AC.

Based on the above discussion, our results revealed an interesting microscopic picture of the mechanism through with confined water affects CW-AC adsorption for short-chain PFASs (Fig. 5). Compared with AC, the enhanced adsorption performance contributed to the participation of confined water in short-chain PFAS mass transfer, i.e., the direct drive of hydrogen bonding and van der Waals interactions and the indirect drive of assisted slip. Consequently, the short-chain PFAS adsorption of CW-AC exhibited an adsorbent-confined water dual-drive mode, which led its adsorption performance to be far superior to that of other reported adsorbents (Fig. 2B). Meanwhile, the presence of confined water facilitated the adsorption of the more hydrophilic short-chain PFAS with a higher stable binding pattern in the CW-AC, which agreed with the shift in the adsorption affinity of the CW-AC to hydrophilic (Fig. 2D).

Fig. 5.

The underlying mechanism of short-chain PFAS adsorption by virgin AC and CW-AC (the numbers in this figure represent the corresponding PMF; the isosurfaces around short-chain PFASs represent the interaction between short-chain PFASs and water molecules, blue isosurface: hydrogen bonds, green isosurface: van der Waals interactions).

Real Drinking Water Purification by CW-AC.

Based on the mild confined water–encapsulation method, we envision that CW-AC could be prepared on-site and packed into existing adsorption devices in a full-scale drinking water plant to achieve an in situ upgrade. To confirm the direct transfer ability of CW-AC functionality to existing full-scale adsorption devices, we performed a rapid small-scale column test (RSSCT) breakthrough experiment with real drinking water as a proof-of-concept experiment. The constant diffusion model was employed to scale down the operational information of the full-scale AC column into the RSSCT, and comparative studies were conducted using the CW-AC and its parent AC (SI Appendix, Table S10). The influent was composed of real drinking water (Jiangsu, China) spiked with four short-chain PFASs (~350 ng/L each) to represent the environmentally related high-level of contamination in human-exposed water (7) (SI Appendix, Table S11). Based on the World Health Organization (WHO) health advisory level (PFAS < 100 ng/L each) (12), all short-chain PFASs exhibited rapid breakthroughs in the AC column, especially PFBA, which broke the limit when processing approximately 2,500 bed volumes that represented only ~18 d of operation in full-scale devices (Fig. 6A and SI Appendix, Table S10). This was in line with previous reports that PFBA, as the most hydrophilic short-chain PFAS, was the first to break through in adsorption facilities (12, 14). In contrast, the breakthrough of all short-chain PFASs in the CW-AC column was up to 21,000 bed volumes (8.4 times that of the parent AC), which corresponds to ~150 d of simulated continuous operation in a full-scale treatment plant (Fig. 6B and SI Appendix, Table S10). This is quite significant given that few novel short-chain PFAS adsorbents have been reported to enrich environment-related concentrations (ng/L-level) of short-chain PFASs in real drinking water and lack an actual operation effect (7, 17). Notably, the short-chain PFAS concentrations used in the experiments represent the environmentally related high-level PFAS load (7), and CW-AC replacement has the potential to improve the broad-spectrum removal of other non-PFAS organic pollutants (e.g., dissolved organic matter, SI Appendix, Fig. S26) by the original adsorption devices. These results, therefore, further highlighted that the CW-AC could be oriented toward in situ upgrading of existing adsorption devices and addressing the majority of short-chain PFAS health exposure risks in the current water environment.

Fig. 6.

Real drinking water purification by CW-AC. Breakthrough curves of all short-chain PFASs in actual drinking water for (A) AC and (B) CW-AC; the parallel dashed line represents the human health advisory level of short-chain PFASs in drinking water recommended by the WHO (PFAS < 100 ng/L each). The error bars in this figure represent the SD (n = 3 independent experiments).

Conclusion

We described a strategy of constructing a confined water environment in CW-AC for the removal of various short-chain PFASs from contaminated water, as demonstrated in the context of commercial AC. Based on both experiments and MD simulations, a dual-driving mode in CW-AC was well demonstrated, which completely eliminated the mass transfer barriers (3.27 to 5.66 kcal/mol) and made CW-AC the top adsorption material to short-chain PFASs among the parent AC (1.4 to 2.7 times) and other adsorbents reported (2.2 to 27.9 times). Meanwhile, the confined water–binding pattern dominated by the chain length and functional groups changed the hydrophobic-dependent affinity between CW-AC and short-chain PFASs, which enabled CW-AC to remove hydrophilic species that could easily escape treatment (e.g., PFBA). Moreover, CW-AC demonstrated suitable selectivity and regenerability (at least four cycles) in a cocontaminants system, and could be readily extended to other commercial ACs with different structures. Importantly, the concept transfer of CW-AC functionality to a full-scale drinking water plant was verified, the short-chain PFAS in real drinking water were removed to below the WHO health advisory level at an environment-related high-load concentration (~350 ng/L each), and the treated volume (21,000 bed volumes) was 8.4 times that of its parent AC. Overall, this work provides a highly efficient, reagent-free, and low-cost solution to the short-chain PFAS crisis, and an in situ upgrade of existing adsorption devices was demonstrated to be achievable. More broadly, since such a strategy involving a confined water environment can be readily extended to other hydrophobic adsorbents (e.g., graphene and carbon nanotubes) and hydrophilic contaminants, this study is a valuable starting point to solve the universal mass transfer problem of hydrophilic pollutants in the adsorption process.

Materials and Methods

Chemicals and Materials.

Short-chain PFASs exhibit different hydrophilicities (log K_OW) ranging from 2.31 to 4.15 due to variations in chain length and terminal functional groups, which leads to differences in adsorption behavior (14, 40). To evaluate the broad spectrum removal ability of CW-AC for various short-chain PFASs, the following four short-chain PFAS reagents (≥98%, Macklin Biochemical Co., Ltd.) with typical log K_OW values (Table 1) were jointly employed to investigate the adsorption performance of CW-AC: PFHxS (4.03), PFHxA (3.71), PFBS (2.63), and PFBA (2.31). Importantly, the log K_OW values of these four short-chain PFASs decrease regularly due to the different chain lengths (C4 or C6) and functional groups (–COOH or –SO₃H), which would help us to further understand the underlying effect mechanism of hydrophilicity on CW-AC adsorption of short-chain PFASs (4, 7). It should also be mentioned that different studies define short-chain PFASs differently; in this paper, short-chain PFASs are those with chain lengths of 4 to 7 (4, 5). They are widely used as long-chain substitutes (8), which are detected at high frequencies in the water environment (4, 7) and easily break through adsorption devices (5, 51).

The application of different ACs will lead to a difference in the size distribution of hydrophobic nanopores, which is mainly reflected in the microporosity and mesoporosity (26, 44). It has been widely demonstrated that micropores can provide a higher confinement degree and form more confined water than mesopores (34, 50). Therefore, a pure microporous AC, Mac (Macklin Biochemical Co., Ltd., raw materials: coconut shell), was used as the substrate for the confined water encapsulation to achieve the highest degree of confined water configuration for CW-AC (SI Appendix, Fig. S7). The nitrogen adsorption–desorption curves and the density functional model proved that micropores constitute the entire internal surface area of this AC (100%, SI Appendix, Table S1). In addition, two different commercial ACs, AC-87% (Aladdin Co., Ltd., raw materials: wood) and AC-79% (Meryer Co., Ltd., raw materials: coal), were also subjected to the confined water–encapsulation experiment to verify the universality of the confined water–encapsulation strategy. The properties of the three ACs are available in SI Appendix, Table S1.

All other reagents used were purchased from Sinopharm Chemical Reagent Co., Ltd. Information on material characterization and short-chain PFAS measurement is available in SI Appendix, Text S1–S3.

Confined Water Encapsulation.

The confined water encapsulation of AC nanopores was accomplished by water vapor diffusion. Previous studies have found that carbonaceous materials adsorb water vapor into nanopores as clusters, due to the low water–carbon interactions and the strong intermolecular interactions (20 kJ/mol) (33). Therefore, we placed the AC under near-saturated water vapor to allow water molecules to self-assemble within the AC nanopores to form confined water structures, and the specific process is as follows (Fig. 1A). First, the AC was vacuum degassed to remove adsorbed water. Then, the AC was immediately placed in a closed environment and exposed to high humidity (maintained 90 to 100%) for 24 h, followed by the water uptake (weight increase) until a constant weight was reached. Subsequently, the AC was placed in pure water and equilibrated with shock for 24 h to promote the self-assembly process and stabilization of the confined water structure. Finally, the excess water adsorbed on the exterior of the AC was evaporated at 45 °C for 15 to 30 min.

To rule out the possibility of liquid water molecules invading the AC nanopores and prove the necessity of water vapor diffusion for the formation of a confined water structure, we performed a control experiment involving liquid water equilibrium on AC. The preparation process was consistent with that of CW-AC except that no water vapor diffusion was performed.

Batch Adsorption Experiments.

CW-AC and AC were evaluated with different adsorption experiments to extract the adsorption performance and behavior of CW-AC for short-chain PFASs. Briefly, adsorption isotherm tests were conducted by challenging CW-AC and AC with four typical short-chain PFAS solutions of a wide range from 20 to 250 mg/L for 24 h. All isotherm data were then fitted with the Langmuir–Freundlich model, and the Gibbs free energy of short-chain PFAS adsorption was determined. Furthermore, the adsorption selectivity and regeneration capacity of CW-AC in competitive systems were also evaluated. Details of the experiments and calculations are available in SI Appendix, Text S4.

MD Simulation.

All-atom MD simulations were performed to obtain detailed molecular-scale views of short-chain PFAS adsorption on CW-AC and AC. An all-atom three-dimensional pore-wall model was established to simulate the AC structure (SI Appendix, Fig. S1), i.e., an uncapped armchair single-wall carbon nanotube (SWCNT) with a length of 5.0 nm was embedded in two graphene sheets along the z direction. The SWCNT and graphene sheets represent the highly graphitized nanopores and surface of AC (SI Appendix, Fig. S1A). The SWCNT of 1.6 nm (12, 12) was selected to represent AC nanopore structures, according to the average pore size calculated from 4V/A by Brunauer-Emmett-Teller (SI Appendix, Table S1). The water molecules were placed on the open side of the pore-wall model, and the average density was adjusted close to bulk water at 303 K and 0.1 MPa. The simulated box size dimensions were 4.912 nm × 5.105 nm × 10.175 nm, and periodic boundary conditions were imposed in all directions. All the models were built using the Moltemplate software package, and visual analysis was performed using VMD software.

MD simulations were performed by the LAMMPS software package. The nonbond interatomic van der Waals interaction was performed using the 12-6 Lennard–Jones empirical forcefield with the geometric mix rule, as follows (44):

$$U\left(r_{\mathit{ij}}\right)=4{\epsilon }_{\mathit{ij}}\left[{\left(\frac{{\sigma }_{ij}}{r_{ij}}\right)}^{12}-{\left(\frac{{\sigma }_{ij}}{r_{ij}}\right)}^6\right],$$

$${\epsilon }_{\mathit{ij}}=\sqrt{{\epsilon }_i{\epsilon }_j},$$

$${\sigma }_{\mathit{ij}}=\sqrt{{\sigma }_i{\sigma }_j},$$

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

The cutoff distance of the van der Waals interaction was selected as 10 Å, and the long-range Coulombic interaction was evaluated by particle-particle-particle-mesh with a root-mean-square accuracy of 10⁻⁴ kcal/mol. For the SWCNT and graphite sheets, a rigid model was adopted that remained fixed during the simulation. The short-chain PFAS molecules were built by the optimized potentials for liquid simulations-all atom model, and Na⁺ was added to neutralize the system charge. The most popular extended simple point charge potential model was employed to describe the water molecules, which captures relevant properties of liquid water close to ambient conditions and is widely used to study confined water (25, 50) and water–PFAS interactions (45, 52). The canonical ensemble (constant temperature, volume, and particle number) was used to perform all simulations at 303 K by a Nose–Hoover thermostat. The system was first energy minimized and then equilibrated for 1 ns with a timestep of 2 fs. Information on the average noncovalent interactions, radial distribution functions, coordination numbers, and hydrogen bond calculation are available in SI Appendix, Text S5–S8.

PMF Calculation.

To further extract the change in the binding free energy of the adsorption pathway of short-chain PFASs entry into nanopores, the PMF was calculated by using the umbrella sampling method (30, 42, 53). The adsorption process of PFASs from bulk water into the nanopores was divided into eight reaction windows on average, and the spacing of each window in the Z direction was 2 Å. After the system was balanced, the steer MD method with a harmonic potential (spring constant: 25 kcal/mol Å and velocity of pulling: 0.001 Å/fs) was used to drag the short-chain PFAS to the specific reaction window. Then, a weak spring constant (2 kcal/mol Å) harmonic potential was used to fix the short-chain PFAS at the reaction window of 100 ps for balancing. Finally, the simulation was continued for 50 ps, and the PFAS reaction coordinates were recorded every 100 fs. The reaction coordinate data were converted to PMF free energy by the weighted histogram analysis method (WHAM) (54); the specific formula is as follows (42, 53):

$$W_{\mathit{window}}=-K_b\times T\times ln{P}_{\mathit{window}},$$

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

Proof-of-Concept for the In Situ Upgrade of Existing Adsorption Devices by CW-AC.

Real drinking water purification using AC for household or municipal applications is most often conducted in a flow-through packed bed geometry, where contaminated water is passed through an AC-containing column (20). We thus performed real drinking water purification with RSSCTs to assess the feasibility of transferring CW-AC functions to existing full-scale adsorption devices. The RSSCT is based on the concept of similitude to scale down the full-scale adsorption process using dimensionless parameters developed from the dispersed-flow pore and surface diffusion model. In this work, the RSSCT design was based on actual full-scale AC column parameters and the constant diffusivity model, which has been proven to effectively simulate the adsorption behavior of PFASs in full-scale adsorption devices (14, 40). The specific design formula and parameters can be found in SI Appendix, Text S9 and Table S10, and for more information on column operations, see SI Appendix, Text S10.

The real drinking water used in the experiment was collected from a horizontal sedimentation tank at a full-scale drinking water plant (Jiangsu, China). A 24-h continuous automatic sampler was used to ensure stable and representative water quality. Immediately after collection, the water sample was filtered through a 0.45 polyethersulfone filter membrane and stored at 4 °C before use. The collection and storage containers were all made of polystyrene to avoid adsorption.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.