Experimental Instruments: Field Emission Scanning Electron Microscope (Zeiss 500, Carl Zeiss, Germany), Fourier Transform Infrared Spectrometer (Spectrum 100, PerkinElmer, USA), X-ray Diffractometer (D8 Advance, Bruker, Germany), Vibrating Sample Magnetometer (BHV-50HTI, Riken Electronics Corporation, Japan), Electronic Analytical Balance (HC1004, Shanghai Huachao Industrial Co., Ltd., China), Ultrasonic Cleaner (KQ-3200, Kunshan Shumei Ultrasonic Instrument Co., Ltd., China), Vacuum Drying Oven (DZF6020Z, Shaoxing Tongcheng Instrument Manufacturing Co., Ltd., China), magnetic heating stirrer (DF-101S, Henan Yuhua Instrument Co., Ltd., China), ultra-high performance liquid chromatography-tandem quadrupole mass spectrometer (Waters ACQUITY UPLC Xevo TQ, Waters Corporation, USA).

Chemical Reagents: Chitosan (deacetylation degree ≥90%), acetic acid solution (1%, w/w), glutaraldehyde (25%, w/w), PFOA, PFOS, perfluorononanoic acid (PFNA), and perfluorohexanoic acid (PFHxA) (Aladdin Biochemical Technology Co., Ltd.); Ferric chloride hexahydrate, ferrous chloride tetrahydrate, liquid paraffin, ammonia solution (25%), sorbitan oleate, sodium hydroxide, petroleum ether, methacrylic acid (MAA), acetonitrile, ethylene glycol dimethacrylate (EGDMA), azobis(isobutyronitrile) (AIBN), methanol (Shanghai Maclin Biochemical Technology Co., Ltd.). All chemicals were analytical grade reagents. Deionized water was used in all experiments. All experiments were performed in triplicate (n = 3), and the results are expressed as the mean±standard deviation. PFOA concentrations were determined by HPLC-MS/MS (Waters ACQUITY UPLC Xevo TQ) using a seven-point calibration curve with a correlation coefficient (R²) greater than 0.999. Blank experiments were conducted in parallel to correct for any background interference or non-specific adsorption. For complex matrix samples, spiked recovery tests were performed to validate the accuracy of the analytical method.

(a) Magnetic Fe₃O₄ nanoparticles were synthesized via the chemical coprecipitation method [18]. The procedure was as follows: FeCl₂·4H₂O (2.48 g) and FeCl₃·6H₂O (6.5 g) were weighed separately into a 250 mL pre-cleaned beaker. Deionized water (225 mL) was added, and the mixture was sonicated for 10 min. The mixture was then transferred to a 500 mL pre-cleaned three-neck flask. Under nitrogen protection and continuous stirring, dropwise, 90 mL of ammonia solution was slowly added. The mixture was aged at 60°C for 1 h. The reaction product was repeatedly washed with deionized water until neutral, separated magnetically, dried in a 60°C vacuum oven, and then ground to yield black magnetic Fe₃O₄ nanoparticles.

(b) Preparation of Magnetic Chitosan: First, chitosan (0.5 g) was dissolved in 50 mL of acetic acid solution (1%, w/w) by stirring in a 50°C water bath until complete dissolution, followed by ultrasonic degassing for 30 min. Then, Fe₃O₄ nanoparticles (0.2 g) were added and the mixture was stirred until uniformly dispersed to obtain Solution A. Liquid paraffin (40 mL) and Span 80 (2 mL) were placed into a beaker and magnetically stirred at 500 rpm for 1 h to form Solution B. Solutions A and B were thoroughly mixed and emulsified at 30°C for 30 min. The pH was adjusted to 8–9 using 1 mol L⁻¹ NaOH solution. Glutaraldehyde aqueous solution (25%, w/w, 1.5 mL) was slowly added to the mixture, which was then magnetically stirred at 50°C and 800 rpm for 1 h. After reaction, the mixture was cooled to room temperature. The black microspheres were collected by magnetic separation, washed six times with petroleum ether and ethanol, and dried in a vacuum oven at 60°C to obtain magnetic chitosan.

PFOA (0.0414 g) and MAA (51.65 μL) were weighed into a 50 mL beaker. Acetonitrile (10 mL) was added, and the mixture was stirred for 1 h to form a prepolymer solution. The molar ratio of PFOA:MAA:EGDMA (1:6:25) was selected based on preliminary optimization experiments [19, 20] that aimed to balance template-monomer interaction and polymer stability. Magnetic chitosan (0.2 g) was added to 50 mL of acetonitrile, ultrasonicated for 15 min before being incorporated into the prepolymer solution. EGDMA (0.47 mL) and AIBN (0.02 g) were added, and the mixture was purged with nitrogen for 10 min and sealed in a water bath for 12 h. The resulting polymer was magnetically separated, eluted with methanol/NaOH (1:1, v/v) until no PFOA was detected, and washed with ethanol to neutrality. The final product was dried in a vacuum oven at 60°C to constant weight, yielding the MMIP. The magnetic non-imprinted polymer (MNIP) was prepared following the same procedure in the absence of PFOA. The preparation process of MMIP is illustrated in Fig. 1.

The surface morphologies of MCS and MMIP were investigated by SEM to reveal their microstructural features. SEM images of MCS and MMIP revealed that both exhibit microspherical shapes with particle sizes ranging from 50 to 100 μm. As shown in Fig. 2(a) and Fig. 2(b), MCS surfaces appear smooth with slight protrusions, whereas MMIP surfaces are noticeably rougher, indicating a surface morphology retained after perfluorooctanoic acid elution. Fig. 2(c) and Fig. 2(d) show SEM images of MCS and MMIP after grinding, respectively. Observation reveals that both MCS and MMIP exhibit three-dimensional porous spatial structures internally, featuring large specific surface areas that provide more adsorption sites. Additionally, fine spherical particles are uniformly distributed throughout, indicating successful encapsulation of Fe₃O₄ particles by chitosan. Notably, the internal surface of MMIP is rougher than that of MCS, and this roughness further enhances its specific adsorption capacity for perfluorooctanoic acid, a result that aligns with the improved recognition performance reported for molecularly imprinted polymers with rough and porous surfaces [21].

FTIR spectroscopy was employed to identify the functional groups and chemical bonds in the materials. As shown in Fig. 3(a), the Fe₃O₄ infrared spectrum exhibits an absorption peak at 571 cm⁻¹, characteristic of Fe-O bonds [22]. The infrared spectra of MCS and MMIP reveal broad peaks at 3297 cm⁻¹ and 3346 cm⁻¹, resulting from the overlapping stretching vibrations of N-H and O-H groups in the chitosan structure [23]. In the MCS infrared spectrum, the peaks at 2924 cm⁻¹ and 2855 cm⁻¹ correspond to the asymmetric and symmetric C-H stretching vibrations of the tert-methyl or methylene groups in the sugar residues of chitosan, respectively. The presence of Fe-O absorption peaks at 563 cm⁻¹ and 574 cm⁻¹ indicates effective binding between chitosan and Fe₃O₄.

XRD analysis was performed to determine the crystal structure of MCS and MMIP. The XRD patterns of MCS and MMIP are essentially identical. Fig. 3(b) shows six diffraction peaks (2θ = 30.0°, 35.6°, 43.0°, 53.4°, 57.1°, 62.5°). These peaks are attributed to the cubic spinel structure of Fe₃O₄, corresponding to the crystal planes (220), (311), (400), (422), (511), and (440), respectively [24]. The characteristic peaks match those in the Fe₃O₄ standard card (JCPDS No. 19-0629). The MCS and MMIP XRD patterns are clear with no non-homogeneous peaks, indicating successful Fe₃O₄ incorporation. The cross-linking process did not disrupt the Fe₃O₄ crystal structure, further confirming successful chitosan coating of Fe₃O₄, thereby preserving the adsorbent’s excellent magnetic properties.

The magnetic properties of the composites were evaluated by VSM to confirm their suitability for separation. The hysteresis loops of MCS and MMIP in Fig. 3(c) reveal symmetrically distributed curves without pronounced hysteresis phenomena, indicating that all samples exhibit superparamagnetic behavior. The saturation magnetization for MCS and MMIP are 17.5 and 13.9 emu g⁻¹, respectively. Within the −20,000 to 20,000 Oe range, MMIP exhibits a saturation magnetization of 13.9 emu g⁻¹, representing a 20.5% reduction compared to MCS (17.5 emu g⁻¹). After applying an external magnetic field, MMIP achieved complete separation from the solution within 30 seconds. This separation rate was approximately 40 times faster than that of unmagnetized chitosan, demonstrating MMIP’s outstanding magnetic responsiveness and excellent magnetic separation performance.

In addition, the porous structure of MMIP was characterized by nitrogen adsorption-desorption measurements. As shown in Fig. 3(d), the nitrogen adsorption-desorption isotherm of MMIP exhibits a typical Type IV isotherm with an H3 hysteresis loop at relative pressures (P/P₀) of 0.5–1.0, indicating the presence of abundant mesoporous structures. The H3 hysteresis loop suggests that the mesopores are primarily cylindrical pores formed by particle aggregation, which facilitates PFOA diffusion and mass transfer.

The BET surface area of MMIP was calculated to be 28.07 m² g⁻¹, providing abundant active sites for PFOA adsorption. The BJH pore size distribution (Fig. 3(d)) reveals a relatively concentrated pore size distribution with an average pore diameter of 5.50 nm, and a distinct pore volume peak in the 2–10 nm mesoporous range. These mesoporous features reduce mass transfer resistance and enable rapid adsorption kinetics.

Under conditions of 35°C, pH = 3, a PFOA solution volume of 10 mL, and an initial concentration of 50 mg L⁻¹, adsorption was conducted for 24 hours. The adsorption efficiency was evaluated at adsorbent dosages ranging from 0.5 to 3 g L⁻¹. As shown in Fig. 4(a), the PFOA removal rate increased from 91% to 94% with rising dosage. Beyond 10 mg L⁻¹, the removal rate stabilized around 97% with no further significant improvement. This indicates that PFOA molecules in the solution are fully adsorbed at a certain dosage level, and further increasing the dosage does not markedly enhance the removal rate. This may occur because PFOA concentrations in the solution approach adsorption equilibrium relatively quickly at low dosages. The residual PFOA concentration is low at high dosages, making it difficult for additional adsorbents to significantly enhance removal efficiency. Contrary to the removal rate trend, Q_e gradually decreased with increasing dosage, dropping from 91 mg g⁻¹ to 16.27 mg g⁻¹. At higher dosages, the number of active sites provided by the adsorbent exceeds the number of PFOA molecules, leading to underutilization of some sites and a consequent decrease in equilibrium adsorption capacity. This phenomenon has been reported in multiple studies [25, 26]. In summary, a 1–1.5 mg L⁻¹ dosage achieves high removal efficiency while maintaining good adsorption capacity.

At 35°C, with a PFOA solution volume of 10 mL and an initial concentration of 50 mg L⁻¹, the adsorption time was set to 24 hours. The effect of different pH values on MMIP adsorption of PFOA was investigated. As shown in Fig. 4(b), at pH 2–3, removal rates exceeded 95%, with Q_e maintained around 32 mg g⁻¹, demonstrating high and stable adsorption performance. When pH increased to 4–5, removal rates decreased to 93–96%, and Q_e dropped to approximately 31 mg g⁻¹, though adsorption capacity remained at a relatively high level. However, at pH ≥ 6, the adsorption performance significantly declined. Particularly at pH 9, Q_e dropped to only about 13.8 mg g⁻¹, with removal efficiency falling to approximately 41%. This trend is closely related to PFOA’s dissociation behavior and the surface charge characteristics of the adsorbent. PFOA’s pK_a value ranges between 1.3 and 2.8. At pH values below pK_a, it primarily exists in molecular form, enabling effective binding to the adsorbent’s active sites through hydrophobic interactions and hydrogen bonding [27]. At pH values above pK_a, PFOA predominantly exists as an anion. Under acidic conditions, the protonated -NH₂ groups on MMIP surfaces carry a positive charge, favoring the adsorption of PFOA anions. Under neutral or alkaline conditions, the number of protonated sites on MMIP decreases, reducing the surface positive charge and thereby enhancing electrostatic repulsion with PFOA anions. Increased pH may also cause OH⁻ ions in water to compete with PFOA anions for adsorption sites, ultimately diminishing adsorption performance [28, 29].

The adsorption kinetics of PFOA by MMIP were investigated at 35°C and an initial concentration of 50 mg L⁻¹. The adsorption process primarily underwent two distinct phases: During the initial reaction period (0–200 min), PFOA was rapidly adsorbed, with a significant increase in adsorption capacity. This was attributed to the abundance of available adsorption sites on the MMIP surface and its specific imprinting sites, enabling rapid binding of PFOA molecules. As time progressed, adsorption sites gradually saturated and electrostatic interactions weakened, causing the adsorption rate to decline. Equilibrium was ultimately reached at approximately 300 min, with an equilibrium adsorption capacity of about 30–33 mg g⁻¹.

The experimental data were fitted using PFO and PSO kinetic models, with results shown in Fig. 5(a) and Table S1 in the supplementary materials. Both models adequately describe the adsorption process, but the PSO kinetic model exhibits superior goodness-of-fit (R²= 0.995) compared to the PFO model (R² = 0.979). Moreover, its calculated equilibrium adsorption capacity (33.167 mg g⁻¹) aligns more closely with experimental values. This indicates that the adsorption of PFOA onto the MMIP follows the PSO kinetic model more closely, suggesting that chemisorption is the dominant mechanism. This is primarily attributed to specific interactions between PFOA and functional groups on the MMIP surface: the carboxylate group (-COO⁻) of PFOA can form hydrogen bonds with the amino (-NH₂) and hydroxyl (-OH) groups of chitosan, as well as with carboxyl groups from methacrylic acid, while also undergoing strong electrostatic attraction with protonated amino groups (-NH^{3 +}).

The adsorption isotherm characteristics of MMIP for PFOA were investigated at different temperatures of 298.15K, 308.15K, and 318.15K. Results indicate that the adsorption process of MMIP exhibits corresponding changes with temperature. At 298.15K, the equilibrium adsorption capacity of MMIP was 46.21 mg g⁻¹, while increasing to 48.63 mg g⁻¹ and 50.58 mg g⁻¹ at 308.15K and 318.15K, respectively. This indicates that elevated temperatures promote the adsorption process, as higher temperatures enhance the interaction between the MMIP surface and PFOA, thereby improving adsorption efficiency. The experimental data were fitted using the Langmuir and the Freundlich adsorption models. The isotherm fits are shown in Fig. 5(b) and (c), with model parameters listed in Table S2. The fitting results indicate that both models adequately describe the experimental data. At temperatures of 298.15K, 308.15K, and 318.15K, the maximum adsorption capacity Q_m from the Langmuir model closely matched the equilibrium adsorption amount Q_e. Furthermore, the Langmuir model exhibited a superior correlation coefficient R² compared to the Freundlich model, indicating that PFOA adsorption onto the MMIP adsorbent surface primarily occurs as monolayer adsorption. Meanwhile, the adsorption index n in the Freundlich model reflects the strength of interaction between the adsorbent and the adsorbate. With n ranging between 2 and 10, it is concluded that the adsorption of PFOA onto the adsorbent occurs relatively readily. This result indicates that a large number of structurally similar and uniform recognition sites may have formed on the material surface, thereby exhibiting adsorption behavior consistent with the Langmuir model. This phenomenon serves as evidence for the successful construction of the MMIP surface imprint layer. It should be noted that the Freundlich model still provides a good fit, suggesting that a small number of non-specific adsorption sites also exist in MMIP. Overall, the Langmuir model better fits the adsorption behavior of MMIP, indicating that the adsorption process of PFOA onto MMIP is primarily dominated by chemisorption in a monolayer.

Through adsorption selectivity experiments, the adsorption performance of MMIP and MNIP toward PFOA and structurally similar PFASs (PFOS, PFNA, PFHxA) was investigated. As shown in Fig. 6(a), MMIP exhibited the highest adsorption capacity for PFOA while demonstrating specific adsorption capabilities for PFOS, PFNA, and PFHxA. Among these, the adsorption capacity for PFOS was relatively higher due to its higher structural similarity with PFOA—both compounds share the same 8-carbon perfluoroalkyl chain and differ only in their terminal functional groups. PFNA and PFHxA, with carbon chain lengths of 9 and 6, respectively, show greater structural deviation from PFOA. Consequently, MMIP’s adsorption capacity for these compounds progressively decreases, likely attributable to reduced spatial matching with the imprinted cavities of MMIP caused by differences in carbon chain length. In contrast to MMIP, MNIP exhibited lower adsorption capacity for PFOA and showed minor differences in adsorption toward the other three PFASs, as MNIP primarily adsorbs PFASs through non-selective mechanisms.

The imprinting factor (IF) and selectivity coefficient (SC) were calculated, with results in Table S3. MMIP demonstrated an IF value of 2.65 for PFOA, significantly higher than those for other structural analogs, indicating its high selectivity toward PFOA. The selectivity coefficient further confirmed MMIP’s preference for PFOA. The SC values of PFOA relative to PFOS, PFNA, and PFHxA were 1.55, 1.73, and 2.08, respectively, demonstrating that MMIP maintains strong selective adsorption capacity even when confronted with structurally similar PFASs. Overall, MMIP exhibits remarkable specificity in selectively adsorbing PFOA and its analogs, showing superior adsorption capacity and selectivity for the target compound compared to MNIP.

Nevertheless, While this study focused on selectivity against structurally similar PFASs, natural waters contain additional competing substances such as inorganic ions and natural organic matter. Future studies should assess the effects of ionic strength and coexisting organic compounds to further validate the practical applicability of MMIP in real water environments.

To investigate the regenerative capacity of the adsorbent, five adsorption-desorption cycles were performed using 5 mg of MMIP on 10 mL of a PFOA solution with an initial concentration of 50 mg L⁻¹. During desorption, a methanol/NaOH (1:1, v/v) solution was employed to remove PFOA adsorbed by MMIP by disrupting the electrostatic and hydrophobic interactions between PFOA and the functional groups. The experimental results are shown in Fig. 6(b). Although the adsorption capacity of the material decreased with increasing adsorption-desorption cycles, it remained above 90% after the fifth recovery cycle. This indicates that MMIP possesses excellent structural stability and high regenerative capacity.

A comparative analysis of the adsorption performance between the developed MMIP and other reported MIP-based adsorbents is presented in Table 1. Conventional MIPs generally exhibit limited PFOA/PFOS adsorption capacities and low imprinting factors. Although the use of advanced supports (e.g., carbon microspheres) can significantly improve capacity, it often comes at the cost of reduced selectivity. In contrast, the MMIP in this work demonstrates a remarkable balance of high adsorption capacity (48.28 mg g⁻¹ for PFOA) and superior selectivity (IF = 2.65), outperforming most previously reported materials.

Based on the collective experimental evidence presented above, the adsorption mechanism of PFOA onto the MMIP is proposed and summarized in Fig. 7. The PSO kinetics model, which best fits the experimental data (R² = 0.995), suggests that chemisorption dominates the adsorption process, indicating strong interactions between PFOA and MMIP’s functional groups. The Langmuir isotherm model further supports monolayer adsorption, with a maximum capacity of 51.28 mg g⁻¹, confirming the presence of homogeneous recognition sites on MMIP. The high selectivity, evidenced by the imprinting factor (IF = 2.65, SC = 3.4), confirms the critical role of the specific recognition cavities formed during imprinting. The adsorption behavior, which was highly sensitive to the initial solution pH and best described by the PSO kinetic model, points to electrostatic attraction and chemisorption as the primary rate-controlling steps. The presence of functional groups critical for adsorption, such as -NH₂ and -OH, was established by FTIR analysis. Concurrently, the successful formation and incorporation of the magnetic Fe₃O₄ core were confirmed by XRD and VSM, respectively. The proposed mechanism, as depicted in Fig. 7, is a mixed adsorption system dominated by chemisorption; involves hydrophobic interactions between the fluorocarbon chain of PFOA and the polymer matrix, electrostatic attraction between the protonated amino groups on MMIP and the anionic carboxyl group of PFOA, and hydrogen bonding between the hydroxyl and amino groups on chitosan and the carboxyl group of PFOA. The superparamagnetic Fe₃O₄ nanoparticles in MMIP enable easy separation from the solution after adsorption. These results validate the mechanism and demonstrate MMIP’s efficiency in PFOA removal.