Selective salt recovery through novel nanofiltration membranes based on 2D-C3N4 materials

Selective salt recovery through novel nanofiltration membranes based on 2D-C 3 N 4 materials P. Iaconis a , D. Fabbri b,* , C. Morgante a , P. Calza b , D. Ziotas c , F. Passarini c , D. Cespi c , L. Ciacci c , V. Boffa a,* a Department of Chemistry and Bioscience, Aalborg University, Fredrik Bajers Vej 7K, Aalborg 9220, Denmark b Department of Chemistry, University of Turin, Via Pietro Giuria 5, Torino 10125, Italy c University of Bologna, Department of Industrial Chemistry “Toso Montanari”, Via Piero Gobetti 85, Bologna 40129, Italy ARTICLE INFO Keywords: 2D material Chitosan Mg Selectivity Carbon nitride Nanofiltration ABSTRACT Nanofiltration (NF) is increasingly explored not only for conventional wastewater treatment but also for the selective recovery of valuable elements from saline sources such as seawater and brine. However, most commercially available NF membranes are negatively charged and produced via interfacial polymerization, which limits their effectiveness in selectively recovering cations like magnesium (Mg²⁺). In this study, a novel positively charged 2D nanofiltration membrane was developed using an environmentally sustainable Mayer rod coating technique. The membrane was fabricated on a commercial ultrafiltration substrate, with an active layer composed of chitosan a natural biopolymer blended with varying percentages of carbon nitride to enhance cation selectivity. Performance was evaluated using single-salt solutions (1 g/L) of MgCl₂, Mg(NO₃)₂, Na₂SO₄, and NaCl. The membrane containing 60 % chitosan and 40 % carbon nitride exhibited the best performance, achieving 87.1 % rejection of MgCl₂ and 83.2 % rejection of Mg(NO₃)₂, outperforming the commercial reference membrane (NF270) in the selective rejection of Mg over sodium ions. Additionally, a preliminary life cycle assessment (LCA) was conducted using a cradle-to-gate approach to evaluate the environmental impact of the fabricated membranes, highlighting that the primary environmental impacts stem from the polysulfone substrate and carbon nitride components. 1. Introduction The growing global concerns over resource scarcity, environmental pollution, and the resulting urgent need for sustainable water management have driven extensive research into advanced water treatment technologies. Among these, membrane-based processes have become well-established for applications in water purification and wastewater treatment, offering efficient, modular, and scalable solutions [1]. In particular, nanofiltration (NF) offers a unique position between Ultrafiltration (UF) and Reverse Osmosis (RO), allowing for the selective separation of solutes based on molecular size and charge. Traditionally employed for water softening and removal of organic micropollutants, NF membranes are now increasingly explored for more specialized applications, including the recovery of valuable multivalent ions from aqueous environments [2,3]. The ion rejection of NF membranes arises from a combination of three separation mechanisms: (i) size exclusion, where ions are physically blocked based on the pore dimension [4], (ii) dielectric exclusion, which results from differences in dielectric constants between the membrane matrix and the bulk aqueous phase, reducing ion partitioning into the membrane [5] and (iii) Donnan exclusion, an electrostatic repulsion mechanism between charged solutes and the surface charge of the membrane [4]. Among the multivalent ions for which NF membranes show a significant high rejection, magnesium (Mg²⁺) represents one such multivalent ion of considerable interest. It plays a key role across various sectors, including metallurgy, agriculture, pharmaceuticals, and electronics [6–8]. Due to its widespread industrial applications and geographically concentrated supply, magnesium has been classified as a critical raw material by the European Union [9]. Its sustainable recovery from water not only aligns with circular economic strategies but also * Corresponding authors. E-mail addresses: [email protected] (P. Iaconis), [email protected] (D. Fabbri), [email protected] (C. Morgante), [email protected] (P. Calza), [email protected] (D. Ziotas), [email protected] (F. Passarini), [email protected] (D. Cespi), [email protected] (L. Ciacci), vb@bio. aau.dk (V. Boffa). Contents lists available at ScienceDirect Journal of Environmental Chemical Engineering journal homepage: www.elsevier.com/locate/jece https://doi.org/10.1016/j.jece.2025.119495 Received 11 July 2025; Received in revised form 22 September 2025; Accepted 23 September 2025 Journal of Environmental Chemical Engineering 13 (2025) 119495 Available online 24 September 2025 2213-3437/© 2025 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ ). contributes to reducing the environmental impacts associated with conventional extraction methods. Seawater in this context represents an optimal opportunity to recover critical raw material such as magnesium, due to its relatively high concentration in this matrix [10]. However, the selective separation of magnesium from complex water matrices remains challenging. Most commercial NF membranes possess a negatively charged surface, which inherently favours the rejection of multivalent anions over cations, due to repulsive electrostatic interactions [11,12]. Therefore, to enhance the rejection of multivalent cations such as Mg²⁺, the development of positively charged membranes with optimized surface chemistry and morphology is crucial. One promising approach to improving NF membrane performance is the incorporation of nanostructured materials, particularly twodimensional (2D) materials. Owing to their intrinsic high porosity and layered architecture characterized by nanoscale interlayer galleries and well-defined nanochannels 2D materials facilitate selective ion and molecule transport [13,14]. Their atomically flat surfaces and tunable surface chemistry also contribute to reduce fouling susceptibility by minimizing foulant adhesion [15,16]. Table 1 summarizes the performance and testing conditions of recently developed positively charged NF membranes based on 2D materials. Among the various 2D materials employed for this purpose, graphitic carbon nitride (g-C₃N₄) stands out due to its high chemical stability, large surface area, and nitrogen-rich functionalities [25,26]. These features make g-C₃N₄ an excellent candidate as a functional filler in positively charged nanofiltration membranes [27]. In parallel, naturally derived polymers are gaining traction as sustainable alternatives to conventional synthetic polymers [28,29]. Chitosan, a biopolymer derived from the deacetylation of chitin, has attracted significant interest due to its biodegradability, low toxicity, and abundance of functional groups particularly primary amines. These groups confer a positive surface charge under mildly acidic or neutral conditions, enabling chitosan-based membranes to exhibit tuneable surface properties and favourable film-forming capabilities [30,31]. The integration of g-C₃N₄ into a chitosan matrix can synergistically enhance membrane properties by improving hydrophilicity, antifouling resistance, and surface charge distribution, while also introducing additional active sites for ion interactions [32]. As a result, chitosan–C₃N₄ composite membranes show significant potential for enhanced ion selectivity, particularly in applications aimed at recovering multivalent cations such as Mg 2+ . Despite these favorable characteristic C 3 N 4 studies have been focused on the antifouling properties and the photocatalytic performances and just few studies have been development for the behavior in salt performances [33–36], This work aims to address this gap by developing and characterizing a series of nanofiltration membranes composed of chitosan and varying ratios of carbon nitride (C₃N₄), with the specific objective of achieving selective rejection of Mg²⁺ over competing ions in solution. Moreover, as highlighted in Table 1, most 2D-material-based membranes reported in the literature are fabricated via interfacial polymerization or vacuum filtration. In contrast, this work introduces a simpler fabrication method using the Mayer rod coating technique, which is commonly used for coating development in the industry. The membranes were evaluated in terms of surface charge behaviour, water permeability, ion rejection performance and ion selectivity. Both single-salt and mixed-salt systems were used to assess membrane selectivity, with particular emphasis on the preferential rejection of magnesium ions. Finally, the environmental sustainability of the membrane production process was evaluated by considering the climate change impact category. This life cycle assessment study was conducted for each membrane formulation, considering the entire production process: from raw material extraction to the final coated membrane. 2. Experimental 2.1. Materials Melamine 99 % (Merck, used as the precursor for the synthesis of carbon nitride (C₃N₄)). Polysulfone ultrafiltration membrane (Alfa Naval, molecular weight cut-off 25 kDa) served as support of the fabricated membrane. Chitosan (low molecular weight, Merck) acetic acid (Merck, purity>99 %) and deionized water were utilized for preparing the coating suspension. NaCl (Merck ACS, ≥99 %), Na 2 SO 4 (Merck, anhydrous, purity ≥99 %), MgCl 2 ⋅6 H 2 O (Carlo Erba, purity 99 %), MgSO 4 ⋅7 H 2 O (Carlo Erba, purity 99 %), Mg(NO 3 ) 2 ⋅6 H 2 O (Merck ACS, purity 99 %), KCl (Merck, purity 99 %), CaCl 2 anhydrous (Merck, purity >93 %), NaHCO 3 (Merck ACS, purity 99.7 %), NaBr (Merck ACS, purity 99 %) were employed to test the membrane performance. Poly (ethylene glycol) (Merck, 35KDa, 25KDa, 20KDa, 6,25KDa, 4KDa, 1Kda, 400 Da) were used for the evaluation of molecular cut off. A commercial membrane (NF270 purchased from DuPont, Wilmington, Delaware) was used for performance comparison purposes. 2.2. Synthesis of carbon nitride (C₃N₄) Carbon nitride (C 3 N 4 ) was synthesized via a conventional thermal condensation process [37]. Specifically, 10 g of melamine was weighed and placed in a ceramic boat, which was then covered with a ceramic lid and positioned at the centre of a tubular furnace. The sample underwent thermal treatment at 550 ◦C for 3 h, achieved with a heating ramp of 5 ◦C min -1 . After cooling to room temperature, the resulting yellow product was ground using an agate mortar. Then the powder was stored for subsequent use. 2.3. Fabrication of C₃N₄/chitosan membranes A total of six membranes were fabricated using the Mayer rod coating technique, each incorporating different weight ratios of carbon nitride (C₃N₄) and chitosan. The composition and corresponding membrane codes are summarized in Table 2. Initially, a suspension of synthesized C₃N₄ was prepared in 5 mL of 2 % (v/v) acetic acid aqueous solution. The suspension was sonicated for 30 min, followed by Table 1 Permeability and rejection performances of newly developed positively charged NF membranes. Ref Paper Fabrication method Permeability (L /m 2 hbar) Feed conc Rejection (%)     MgCl 2 MgSO 4 NaCl Na 2 SO 4 [17] PE/amine reduced GO Vacuum filtration 2.9 1 mM 93.0 88.1 88.2 65.1 [18] “Ion Cage” GO Vacuum filtration 3.5 1 g/L 95.3 83.2 81.8 70.7 [19] Hydrotalcite/GO functionalized IP 3.3 1 g/L 97.0 42.0 42.5 32.5 [20] Mxene (Ti 3 C 2 ) amino functionalized IP 8.3 2 g/L 97.4 98.5 64.6 79.4 [21] PEI/TA/MoS 2 IP 5.8 1 g/L 66.3 93.2 21.5 54.2 [22] IP on PU nanofibrous support Multi step IP 8.8 1 g/L 99.1 82.3 81.0 57.1 [23] Polyethyleneimine-Noria interlayer deposition strategy IP 11.0 2 g/L 95.4 87.5 75.3 63.4 [24] Poly (N, N-dimethyl aminoethyl methacrylate) Quaternization reaction 2.5 1 g/L 95.0 90.0 90.0 65.0 P. Iaconis et al. Journal of Environmental Chemical Engineering 13 (2025) 119495 2 magnetic stirring for an additional 60 min. Chitosan was then added, and the mixture was stirred for 2 h to ensure homogeneity. Polysulfone substrates were cleaned by wiping the surface to remove possible dust and dirtiness from the surface before performing the coating. The coating was applied using a 100 µm Mayer rod. After coating, the membranes were allowed to dry at room temperature and subsequently subjected to thermal treatment at 80 ◦C for 2 h to enhance adhesion of the coating layer (Fig. 1). The membrane was stored in a dry environment until testing. 2.4. Characterization techniques X-ray diffraction (XRD) analyses were performed using a PANalytical PW3040/60 X’Pert PRO MPD diffractometer (Malver Panalytical, Almelo, Netherlands) operating at 45 kV and 40 mA, with Cu K α radiation (λ =0.15418 nm). Scans were conducted in continuous mode over a 2θ range of 10◦–70. Fourier-transform infrared (FTIR) spectroscopy was carried out using a Bruker TENSOR II spectrometer (Bruker, Karlsruhe, Germany) equipped with a Bruker Platinum Attenuated Total Reflectance (ATR) accessory. Spectra were collected over the range of 4000–400 cm⁻¹ by averaging 64 consecutive scans for each sample. Field emission scanning electron microscopy (FESEM) was conducted using a TESCAN S9000G (Brno, Czech Republic) to examine the morphology and cross-sectional structure of both the synthesized carbon nitride and the fabricated membranes. Membrane thickness was evaluated by measuring cross sections in three different regions of the same sample, and the standard deviation was calculated to account for variability. Imaging was performed using a working voltage of 10 keV. Prior to analysis, all samples were coated with a 7 nm layer of gold to enhance conductivity. Elemental distribution within the membrane coatings was further investigated using energy-dispersive X-ray spectroscopy (EDX) coupled with FESEM. Streaming potential measurements were carried out using an electrokinetic analyzer (SURPASS, Anton Paar, Austria). An area of 20 ×10 mm² was cut from all the membranes and placed it in an adjustable-gap cell (105 ±5 μ m). Measurements were performed at 25◦C using 1 mM KCl solution, while the pH was adjusted between 3 and 6 using HCl or NaOH. Total organic carbon (TOC) was measured using the difference method with the LCK 380 cuvette test (Hach). This approach involves determining both total carbon and total inorganic carbon through complete oxidation to CO₂; TOC is then obtained by subtracting the inorganic fraction from the total. The LCK 380 cuvette provides a measurement range of 2–65 mg/L. Contact angle measurements were performed on all fabricated membranes using the TL101-Dis1 instrument model (Biolin Scientific, Gothenburg, Sweden). A 5 µL droplet of water was deposited onto the membrane surface, and the contact angle was determined from video recordings. Each measurement was repeated three times at different positions on the same membrane to ensure reproducibility. Surface topography was analyzed by atomic force microscopy (AFM) using a Tip-DLC-188 (Kem-En-Tec Nordic A/S, Denmark). Measurements were performed over areas of 10 ×10 µm at a resolution of 128 pixels per line, with three distinct regions examined for each membrane. 2.5. Filtration tests and performance indicators Filtration tests were conducted using a laboratory-scale crossflow filtration setup. Each experiment utilized 1 L of feed solution, with the dual-pump configuration of the system adjusted to maintain a constant transmembrane pressure (TMP) of 4 bar. The crossflow flow was regulated using a flowmeter and kept constant at 0.5 L/s throughout the experiments, as reported in previous work [38]. The pure water permeability of the membranes was assessed using deionized water. The permeate was collected and measured with a digital balance positioned beneath the filtration cell, and the mass was recorded every 5 min over a total duration of 1 h. Permeability (J) was calculated using Eq. 1: J=V P t A (1) where V is the volume of permeate (L), P is the transmembrane pressure Table 2 Composition of the fabricated membranes of this work. Membrane C 3 N 4 [%] Chitosan [%] Loading [%] w/w Chit-CN0 0 100 10 Chit-CN20 20 80 10 Chit-CN40 40 60 10 Chit-CN60 60 40 10 Chit-CN80 80 20 10 CN100 100 0 10 Fig. 1. Schematic process of the membrane fabrication via Mayer rod coating and thermal curing. P. Iaconis et al. Journal of Environmental Chemical Engineering 13 (2025) 119495 3 (bar), t is the filtration time (h), and A is the effective membrane area (m²). To evaluate the rejection performance, single salt solutions (1 g L⁻¹) of NaCl, Na₂SO₄, MgCl₂, MgSO₄, and Mg(NO₃)₂ were tested. A mixed salt solution containing NaCl, Na₂SO₄, MgCl₂, and MgSO₄ was also used. The rejection rate (R) was calculated using Eq. 2: R=(1−Cp Cf)⋅100 (2) where Cₚ and C f are the solute concentrations (g/L) in the permeate and feed, respectively. For single salt experiments, concentrations in both feed and permeate were determined using a conductometer to allow rapid and straightforward measurements. The standard deviation of these measurements was calculated by repeating the analysis on three different batches tested under the same conditions. For the mixed salt system, ion chromatography and ICP-OES were employed. Specifically, sodium (Na⁺) and magnesium (Mg²⁺) ion concentrations were determined via Liberty 100 inductively coupled plasma-optical emission spectrometer (ICP-OES, Varian (Palo Alto, CA, USA)), while anions were quantified by ion chromatography using a DX 500 Ion Chromatograph (IC, Thermo Fisher Scientific, Milan), equipped with IonPac AS9-HC column. The eluent was a solution 9 mM Na₂CO₃ at a flow rate of 1 mL/min. Salt concentrations were determined by calibration curves prepared from standard solutions. The Mg/Na selectivity was then calculated using Eq. 3: ISIon1 Ion2=(1−RIon2 RIon1)(3) Ion selectivity (IS) was calculated based on the rejection values of two ions, R Ion1 (%) (the more highly rejected ion) and R Ion2 (the less rejected ion). An IS value approaching zero indicates no selectivity between the two ions, whereas an IS value approaching one corresponds to perfect selectivity, meaning complete rejection of one ion and no rejection of the other. Pore size distribution was evaluated using polyethylene glycol (PEG) solutions with molecular weights ranging from 400 Da to 35 kDa. Filtration tests were carried out with PEG at a concentration of 50 mg/L, and residual PEG concentrations were quantified via TOC analysis [39, 40]. Finally, the Chit-CN40 membrane was tested in a complex ionic matrix by preparing artificial seawater according to the procedure reported by Carmelo et al. [38] subsequently acidified to pH 5.5 with HCl. The concentrations of cations (Na⁺, Mg²⁺, K⁺, and Ca²⁺) were determined using an atomic absorption spectrometer (PinAAcle 900 F, PerkinElmer, Massachusetts, USA). 2.6. Life cycle analysis (LCA) To complement the technical evaluation of the fabricated chitosanC₃N₄ membranes, this section provides a Life Cycle Assessment (LCA) focusing specifically on their potential climate change impact. The environmental sustainability of these membranes is evaluated alongside their performance. This analysis quantifies the potential greenhouse gas emissions associated with the production of the five primary membranes discussed in this study: Chit-CN0, Chit-CN20, Chit-CN40, Chit-CN60, and Chit-CN80. The assessment follows a cradle-to-gate approach, encompassing processes from raw material extraction through membrane fabrication, but excluding downstream stages like membrane use and disposal. The functional unit for this assessment is defined as the production of one coated membrane. The Environmental Footprint 3.1 impact assessment method was applied, utilizing the Ecoinvent 3 database for life cycle inventory data [51,52,54]. Environmental Footprint 3.1 was used as it is the official EU-recommended method for product environmental assessments, ensuring consistency with European Commission guidelines [50]. The APOS (Allocation at the Point of Substitution), Unit (U) system model was selected for this study due to its capacity to provide a consequential modeling perspective [53]. The Unit (U) system model ensures that all environmental flows are normalized per functional unit. SimaPro software was employed for modelling and calculations [55]. 3. Results and discussion 3.1. Chit-CN composite membranes To confirm the successful incorporation of carbon nitride within the chitosan matrix, ATR-FTIR spectra were collected for the individual membrane components: chitosan, polysulfone, and carbon nitride, as well as for the composite membrane containing 40 % carbon nitride (Chit-CN40), as shown in Fig. 2. The Chit-CN40 membrane was selected as a representative sample for the composite series, as spectral differences among the various compositions were minimal. The FTIR spectrum of polysulfone (PSf) exhibits the characteristic CH₂ stretching vibrations at 2904 and 2850 cm⁻¹, along with a distinctive aromatic C – – C band at 1462 cm⁻¹cm -1 [41,42]. The spectrum of chitosan displays characteristic vibrational bands associated with C–O stretching at 1065 and 1016 cm⁻¹ [43]. The spectrum of the Chit-CN40 composite membrane reveals the presence of all key functional groups from the individual components. Notably, the triazine ring bending (800 cm⁻¹), C – – O and C–N stretching bands, as well as the CH₂ stretching vibrations originating from the underlying polysulfone support, are all clearly visible. These findings confirm the successful integration of carbon nitride into the chitosan matrix. 3.2. Membrane morphology and surface characteristics The synthesized carbon nitride (C₃N₄) exhibited a morphology consisting of large particles measuring several micrometers in size (Fig. 3h). This morphology significantly influenced the characteristics of the membrane coatings. As shown in Fig. 3a-c, increasing the C₃N₄ content within the membranes led to a progressive increase in surface roughness. The membrane composed solely of chitosan (Chit-CN0) displayed a smooth and homogeneous surface (Fig. 3a), whereas membranes incorporating higher concentrations of C₃N₄ exhibited visibly rougher textures due to the presence of the large carbon nitride particles (Fig. 3bc). Fig. 2. ATR-FTIR spectra for the three individual components (PSf support, chitosan and carbon nitride), as well as for the composite membrane Chit-CN40. P. Iaconis et al. Journal of Environmental Chemical Engineering 13 (2025) 119495 4 Cross-sectional analysis showed that the membrane composed solely of chitosan exhibited the greatest thickness, measuring approximately 5.1 µm. For membranes with varying chitosan/C₃N₄ ratios, the overall thickness remained relatively consistent exhibiting a thickness around 2µm, as illustrated in Fig. 3g. However, an increase in the standard deviation of thickness measurements was observed with higher carbon nitride content. This variability is attributed to the presence of larger C₃N₄ particles, particularly near the edges of the coating layer, which resulted in a more uneven and irregular cross-sectional profile. However, a homogeneous distribution of C₃N₄ within the chitosan matrix can be observed through EDX elemental mapping of the membrane surface. Fig. 4. shows the spatial distribution of elements within the coating layer, highlighting the nitrogen signal, originating from the carbon nitride material, uniformly distributed across the surface. Fig. 3. FESEM images showing the surface morphology of membranes: (a) Chit-CN0, (b) Chit-CN40 and (c) Chit-CN80 at 1kx magnification, and the corresponding cross-sectional views of the same membranes (d–f). (g) Plot of the coating thickness of the fabricated membranes, (h) FESEM image of C 3 N 4 at 10Kx of magnification. Fig. 4. FESEM images of the Chit-CN0 (a) and Chit-CN40 (b) coupled with EDX analysis for elemental map. P. Iaconis et al. Journal of Environmental Chemical Engineering 13 (2025) 119495 5 3.3. Streaming potential analyses and molecular weight cut off An essential characteristic of nanofiltration membranes is their surface charge at different pH values. In this study, the fabricated chitosan–C₃N₄ membranes were compared to a widely used commercial nanofiltration membrane, NF270, which features a thin-film polypiperazine active layer and is commonly applied in municipal water treatment. The NF270 membrane consistently exhibited a negative surface charge across the entire pH range (Fig. 5a). In contrast, the fabricated membranes showed a pH-dependent surface charge: they showed a negative charge at pH 10 and gradually became positively charged as the pH decreased below 6.5. Specifically, Chit-CN40 shifted to a positive charge around pH 6.5, while Chit-CN0 and Chit-CN80 exhibited this transition at approximately pH 5.8. This behavior aligns with the known properties of chitosan, which contains both amino (–NH₂) and hydroxyl (–OH) groups. At pH values above 6.5, the deprotonation of these groups leads to a net negative charge. Below this threshold, the protonation of amino groups results in a net positive charge. Furthermore, the addition of C₃N₄ appears to enhance the positive surface charge of the membranes. This effect is attributed to the increased presence of –NH groups and the relative reduction in –OH groups on the membrane surface as the C₃N₄ content increases. Surface charge plays a critical role in separation performance, especially in applications targeting the recovery of magnesium or other multivalent cations. A positively charged membrane favors the selective rejection of highly charged cations over those with lower charge, primarily due to electrostatic repulsion mechanisms such as the Donnan effect. To complement the membranes characterization, the pore size distribution of the membranes was evaluated through molecular weight cut off (MWCO) experiments using polyethylene glycol (PEG) as a model solute. Fig. 5b. shows the rejection performance of the Chit-CN40 membrane compared with the polysulfone support used in its fabrication. The polysulfone support exhibited an MWCO of approximately 25 kDa, while the Chit-CN40 membrane achieved a significantly lower MWCO of around 9.5 kDa, confirming the tighter structure introduced by the chitosan–C₃N₄ selective layer. 3.4. Single salt rejection and permeability The rejection performance of the fabricated membranes was evaluated for all compositions, with the exception of the membrane composed entirely of carbon nitride (100 % C₃N₄), as this coating lacked sufficient stability without the presence of chitosan and was easily washed away during preliminary tests under the adopted experimental conditions. All filtration experiments were conducted at a constant pressure of 4 bar and a crossflow velocity of 0.5 L/s. The pH of the feed solutions was not modified, and it stabilized naturally at 5.5, corresponding to the equilibrium pH with atmospheric CO₂. The solutions tested each contained 1 g/L of following salts: NaCl, Na₂SO₄, MgCl₂, MgSO₄, and Mg(NO₃)₂, as commonly used in other works [44,45]. As shown in Fig. 6b, the rejection behaviour of the fabricated membranes displayed a trend that was opposite to the one observed for NF270. At pH 5.5, NF270 exhibits a negative surface charge, whereas the chitosan-C₃N₄ composite membranes possess a positive surface charge. The observed rejection order for the fabricated membranes was MgCl₂ >Mg(NO₃)₂ >NaCl >MgSO₄ >Na₂SO₄. This trend is characteristic of positively charged nanofiltration membranes and suggests that electrostatic interactions, specifically, the Donnan effect, are the dominant rejection mechanism. In particular, the positively charged membrane surface preferentially rejects multivalent cations such as Mg²⁺ over monovalent ions like Na⁺. It is also evident that the nature of the counterion influences membrane performance; when sulphate ions (SO₄²⁻) are present, the rejection decreases, this is due to the higher attraction of the bivalent anions with the positive surface respect to the Cl - anion. This phenomenon is characteristic of positively charged membranes and is well-explained by the Donnan effect, which suggests that salts with higher cation-to-anion valence ratios (Z⁺/Z⁻) are rejected more efficiently. As a result, chloride-based salts are typically retained more effectively than sulfate-based salts [46,47]. Among all tested compositions, the Chit-CN40 membrane (composed by 40 % C₃N₄ and 60 % chitosan) demonstrated the best overall performance, achieving rejection rates of 83.2 % for Mg(NO₃)₂, 87.1 % for MgCl₂, 50.6 % for NaCl, 22.1 % for MgSO₄, and 17.7 % for Na₂SO₄ (Fig. 6b). Regarding water permeability, the membranes exhibited relatively consistent values across different carbon nitride loadings (Fig. 6a). This result suggests that the incorporation of C 3 N 4 inside the film doesn’t impede water transport through the membrane. Specifically, the ChitCN40 membrane showed a permeability of 5.6 L/m²hbar. Additional test was performed using a mixed salt solution containing NaCl (250 mg/L), Na₂SO₄ (250 mg/L), MgCl₂ (250 mg/L), and MgSO₄ (250 mg/L) in 1 L of water, to verify the selectivity of Mg/Na for the chitosan-C 3 N 4 membrane. The Chit-CN40 membrane was selected for this test, as it had previously exhibited the highest rejection performance for magnesium salts. The concentrations of individual ions were analyzed using ion chromatography and ICP-OES to determine both Mg/ Na and SO₄²⁻/Cl⁻ selectivity. Chit-CN40 membrane demonstrated an Mg/Na selectivity ratio of approximately 0.9, which is significantly higher than the 0.5 selectivity observed for the NF270 membrane. In contrast, NF270 displayed superior selectivity for sulphate ions compared to chloride ions. These results are coherent with the Fig. 5. a) ζ potential of NF270, PSf and the fabricated membrane in the pH range from 4 to 10. b) molecular weight cut off test performed on PSf and ChitCN40 membrane. P. Iaconis et al. Journal of Environmental Chemical Engineering 13 (2025) 119495 6 mechanism of electrostatic repulsion already described previously. Therefore, the high cation selectivity demonstrated by Chit-CN40 suggests its potential applicability in separation processes involving multivalent ions. Compared to the commercial NF270 membrane, ChitCN40 exhibits superior performance in selectively rejecting specific cations. This enhanced selectivity makes it a promising candidate for the targeted recovery of magnesium from magnesium-rich sources such as seawater, desalination brines, and mining effluents. [48–50]. As demonstrated by the streaming potential analysis, the surface charge of the membrane is strongly influenced by the pH of the solution. Since surface charge plays a critical role in separation performance, the effect of pH on the Chit-CN40 membrane was systematically investigated through filtration tests conducted at three representative pH values: 3 (acidic), 5.5 (slightly acidic), and 8 (basic). Previous streaming potential measurements indicated that the surface of Chit-CN40 becomes positively charged at pH values below 6.5. Therefore, at pH 3 and 5.5 the membrane surface is expected to be positively charged, while at pH 8 it becomes negatively charged. Fig. 7. reports the rejection of MgCl₂ at the three pH conditions. A slight increase in rejection was observed when the pH decreased from 5.5 to 3, although the difference was not significant. In contrast, at pH 8 the rejection performance of Chit-CN40 dropped drastically to only 9.4 %. This behavior is consistent with the streaming potential results: above pH 6.5 the membrane acquires a negative charge, which eliminates electrostatic repulsion with divalent cations such as Mg²⁺, allowing them to pass more easily through the membrane. The monovalent counterion (Cl⁻) is also able to permeate quite freely in this case, due to low electrostatic repulsion. This is analogous to the low rejection observed for Na₂SO₄ at pH 5.5, where the counterion also experiences minimal electrostatic hindrance. These findings confirm that the dominant separation mechanism in the ChitCN40 membrane is governed by the Donnan effect (electrostatic interactions). As soon as the surface charge of the membrane shifts from positive to negative, the rejection of Mg²⁺ collapses, highlighting the crucial role of surface charge in defining membrane selectivity. 3.5. Filtration test on artificial seawater The final test was carried out using artificial seawater, acidified to pH 5.5 with hydrochloric acid to ensure the membrane surface remained positively charged. As shown in Fig. 8, the Chit-CN40 membrane exhibited high ion selectivity, achieving an IS of 0.9 for Mg/Na and 0.8 for Mg/K. These results confirm that the membrane can effectively discriminate between divalent and monovalent cations even in complex matrices, a key feature for practical applications such as magnesium recovery from seawater or brines. Fig. 8. also reports the selectivity for Ca²⁺, the other divalent cation present, which followed a similar trend to Mg²⁺ but with slightly lower selectivity. Finally, the Mg/Ca selectivity was plotted, showing a relatively low value (~0.3). This outcome is expected, as both ions carry the same divalent charge, making them Fig. 6. Pure water permeability and rejections of single salt solutions of the support, NF270 (commercial NF membrane) and the fabricated chitosanC 3 N 4 membranes. Fig. 7. pH influence on the rejection performance of Chit-CN40 towards MgCl 2 salt. Fig. 8. Selective rejection of Mg²⁺ and Ca²⁺ compared to other ions during seawater filtration using Chit-CN40. P. Iaconis et al. Journal of Environmental Chemical Engineering 13 (2025) 119495 7 difficult to separate solely through electrostatic (Donnan) interactions. 4. Life cycle assessment (LCA) - climate change impact 4.1. Overview of contributors The assessment indicates that electricity consumption and the commercial polysulfone support membrane are significant contributors to overall environmental impacts. Polysulfone’s production is known to carry a notable environmental burden, especially concerning climate change. Electricity usage in this context primarily stems from laboratory equipment, including the oven for thermal treatment, the magnetic stirrer, and the sonicator used during suspension preparation. The total electricity demand for the coating procedure is estimated at approximately 805 Wh. 4.2. Climate change impact analysis Focusing on the climate change impact category, we examine how varying the proportions of carbon nitride (C₃N₄) and chitosan affects the total emissions, measured in kg CO₂ equivalents (kg CO₂ eq). This category is crucial for understanding the atmospheric impact of the membrane production process. Table 3 presents the climate change impacts for each membrane formulation. The ’Grand Total’ includes all inputs (electricity, support membrane, chemicals), while ’Carbon Nitride’ and ’Chitosan’ show the impacts solely from these two components, and ’Total’ represents their combined contribution. The data from Table 3 clearly indicates that increasing the C₃N₄ content in the membrane coating leads to a higher climate change impact. When comparing the contributions of C₃N₄ and chitosan directly, C₃N₄ exhibits a significantly larger environmental footprint on a per-unit basis. For instance, the C₃N₄ impact rises from 0 kg CO₂ eq in Chit-CN0–0.0028 kg CO₂ eq in Chit-CN80, while the chitosan impact decreases concurrently. This demonstrates that C₃N₄ contributes substantially more to greenhouse gas emissions than chitosan at equivalent mass fractions. This higher impact is attributed mainly to the use of melamine which is required for C₃N₄ production. The ’Grand Total’ also reflects this trend, increasing steadily as the C₃N₄ proportion rises. From a climate change perspective, the incorporation of C₃N₄ into the chitosan matrix increases the environmental burden associated with membrane production. Carbon nitride exhibits a higher impact compared to chitosan. While C₃N₄ enhances certain performance aspects, its contribution to greenhouse gas emissions must be considered. The Chit-CN40 membrane, which demonstrated the best overall performance in this study, represents a mid-point in terms of climate change impact among the tested formulations. This suggests it may offer a balanced option when considering both technical efficacy and climate impact. Future research could aim to reduce the impact by optimizing C₃N₄ synthesis, exploring lower-impact alternatives, and improving energy efficiency in the fabrication process, particularly concerning electricity consumption. 5. Conclusions This work focused on the selective magnesium recovery from rich sources, such as seawater, mining effluent and brines. For this purpose, a series of novel chitosan-C 3 N 4 membranes were synthesized to exploit the peculiar properties of these two materials. The incorporation of carbon nitride in the chitosan matrix attributes a positive charge to the membrane, fundamental feature to discriminate multivalent ions over monovalent ones. Several compositions were tested to identify the optimal balance between permeability and ion rejection. The results indicate that membrane permeability remained relatively constant across different compositions, averaging around 5.6 L/m²hbar. However, significant differences were observed in ion rejection: while both the bare support and the membrane composed solely of chitosan exhibited negligible rejection, the optimized formulation (Chit-CN40) achieved rejection rates of 83.2 % for Mg(NO₃)₂ and 87.1 % for MgCl₂. When tested in a mixed salt solution, Chit-CN40 exhibited excellent magnesium selectivity over sodium, achieving an ion selectivity (IS) of 0.9; significantly higher than the 0.5 observed for the commercial NF270 membrane. Nevertheless, further optimization is necessary, particularly to enhance permeability, which remains lower than that of commercial benchmarks. Finally, the Life Cycle Assessment (LCA) conducted in this study provided critical insight into the climate change impact associated with the production of chitosan–carbon nitride (C₃N₄) nanofiltration membranes. Notably, increasing the C₃N₄ content in the membrane formulation resulted in a proportional rise in greenhouse gas emissions, primarily due to the energy-intensive synthesis of C₃N₄ from melamine. Despite this, the Chit-CN40 membrane emerged as a promising compromise, offering high separation performance while maintaining a moderate climate impact. These findings underscore the importance of balancing membrane performance with environmental sustainability and highlight potential avenues for reducing the footprint through greener material sourcing and process optimization. CRediT authorship contribution statement Paolo Iaconis: Conceptualization, Data curation, Investigation, Writing – original draft. Debora Fabbri: Supervision, Data curation, review & editing. Carmelo Morgante: Visualization, review & editing. Vittorio Boffa: Supervision, review & editing. P. Calza: Funding acquisition. D. Ziotas: Investigation, Writing – original draft. F. Passarini: Supervision, Data curation. D. Cespi: Supervision, Data curation. L. Ciacci: Supervision, Data curation. List of abbreviations UF ultrafiltration NF nanofiltration RO reverse osmosis IP interfacial polymerization PU polyurethane PE polyelectrolyte GO graphene oxide PEI Polyethyleneimine TA Tannic Acid Chit-CN Chitosan-carbon nitride XRD X-ray diffraction FTIR Fourier transform infrared ATR Attenuated Total Reflectance FESEM Field emission scanning electron microscopy EDX energy dispersive X-ray spectroscopy LCA Life Cycle Assessment Table 3 Total climate change impact (kg CO₂ eq) from the production of the five studied membranes, including a comparison between carbon nitride and chitosan, and their combined total. Results were generated using SimaPro. Membrane Grand Total Carbon Nitride Chitosan Total Unit Chit-CN0 0.1557 0.0000 0.0011 0.0011 kg CO 2 eq Chit-CN20 0.1562 0.0007 0.0009 0.0016 kg CO 2 eq Chit-CN40 0.1567 0.0014 0.0007 0.0021 kg CO 2 eq Chit-CN60 0.1571 0.0021 0.0004 0.0025 kg CO 2 eq Chit-CN80 0.1576 0.0028 0.0002 0.0030 kg CO 2 eq P. Iaconis et al. Journal of Environmental Chemical Engineering 13 (2025) 119495 8 APOS Allocation at the Point of Substitution Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Acknowledgments This project has received funding from the European Union under the Horizon Europe Marie Skłodowska-Curie Actions (MSCA) Doctoral Network program (IN2AQUAS-grant agreement number 101119555). Data availability Data will be made available on request. References [1] X. Peng, Y. Jiang, Z. Chen, A.I. Osman, M. Farghali, D.W. Rooney, P.S. Yap, Recycling municipal, agricultural and industrial waste into energy, fertilizers, food and construction materials, and economic feasibility: a review, Environ. Chem. Lett. 21 (2023) 765–801, https://doi.org/10.1007/s10311-022-01551-5. [2] Z.R. Usha, C. Liu, S. Zhang, Z. Wang, A comprehensive study of recent advances, challenges and factors of nanofiltration membrane in various industrial applications, Desalination 599 (2025), https://doi.org/10.1016/j. desal.2024.118461. [3] M.A. Abdel-Fatah, Nanofiltration systems and applications in wastewater treatment: review article, Ain Shams Eng. J. 9 (2018) 3077–3092, https://doi.org/ 10.1016/j.asej.2018.08.001. [4] N.S. Suhalim, N. Kasim, E. Mahmoudi, I.J. Shamsudin, A.W. Mohammad, F. M. Zuki, N.L.A. Jamari, Rejection mechanism of ionic solute removal by nanofiltration membranes: an overview, Nanomaterials 12 (2022), https://doi.org/ 10.3390/nano12030437. [5] D.L. Oatley, L. Llenas, N.H.M. Aljohani, P.M. Williams, X. Martínez-Llad´ o, M. Rovira, J. de Pablo, Investigation of the dielectric properties of nanofiltration membranes, Desalination 315 (2013) 100–106, https://doi.org/10.1016/j. desal.2012.09.013. [6] J. Bai, Y. Yang, C. Wen, J. Chen, G. Zhou, B. Jiang, X. Peng, F. Pan, Applications of magnesium alloys for aerospace: a review, J. Magnes. Alloy. 11 (2023) 3609–3619, https://doi.org/10.1016/j.jma.2023.09.015. [7] J. Walker, S. Shadanbaz, T.B.F. Woodfield, M.P. Staiger, G.J. Dias, Magnesium biomaterials for orthopedic application: a review from a biological perspective, J. Biomed. Mater. Res B Appl. Biomater. 102 (2014) 1316–1331, https://doi.org/ 10.1002/jbm.b.33113. [8] D.J. Hemanth, V.D.Ambeth Kumar, ⋅ S. Malathi, O. Castillo, B. Patrut, Emerging Trends in Computing and Expert Technology Lecture Notes on Data Engineering and Communications Technologies 35, n.d. 〈http://www.springer.com/serie s/15362〉. [9] COMMITTEE AND THE COMMITTEE OF THE REGIONS Critical Raw Materials Resilience: Charting a Path towards greater Security and Sustainability 1, n.d. 〈http ://info.worldbank.org/governance/wgi/〉. [10] C.A. Quist-Jensen, M.K. Jørgensen, M.L. Christensen, Treated seawater as a magnesium source for phosphorous recovery from wastewater—a feasibility and cost analysis, Membranes 6 (2016), https://doi.org/10.3390/membranes6040054. [11] S. Hermans, H. Mari¨ en, C. Van Goethem, I.F. Vankelecom, Recent developments in thin film (nano)composite membranes for solvent resistant nanofiltration, Curr. Opin. Chem. Eng. 8 (2015) 45–54, https://doi.org/10.1016/j.coche.2015.01.009. [12] Y. He, J. Miao, S. Chen, R. Zhang, L. Zhang, H. Tang, H. Yang, Preparation and characterization of a novel positively charged composite hollow fiber nanofiltration membrane based on chitosan lactate, RSC Adv. 9 (2019) 4361–4369, https://doi.org/10.1039/C8RA09855G. [13] S. Xing, S. Du, Y. Huang, X. Qi, M. Sui, Advances in High-Performance nanofiltration membranes facilitated by Two-Dimensional materials, Water 16 (2024), https://doi.org/10.3390/w16070988. [14] Y. Yu, X. Zhang, P. Lu, D. He, L. Shen, Y. Li, Enhanced separation performance of polyamide Thin-Film nanocomposite membranes with interlayer by constructed Two-Dimensional nanomaterials: a critical review, Membr. (Basel) 12 (2022), https://doi.org/10.3390/membranes12121250. [15] G. Saeed, A. Alam, R. Vinodh, D. Kumar KM, C.V.V. Muralee Gopi, K. Ho Kim, B. Tadesse, Two-dimensional (2D) material nanofiltration membranes for effective recovery of lithium, J. Ind. Eng. Chem. (2025), https://doi.org/10.1016/j. jiec.2025.03.004. [16] S. Elakkiya, G. Arthanareeswaran, D.B. Das, Embedding low–cost 1D and 2D iron pillared nanoclay to enhance the stability of polyethersulfone membranes for the removal of bisphenol a from water, Sep Purif. Technol. 266 (2021), https://doi. org/10.1016/j.seppur.2021.118560. [17] X. Song, R.S. Zambare, S. Qi, B.N.I.L. Sowrirajalu, A.P. James Selvaraj, C.Y. Tang, C. Gao, Charge-Gated ion transport through polyelectrolyte intercalated amine reduced graphene oxide membranes, ACS Appl. Mater. Interfaces 9 (2017) 41482–41995, https://doi.org/10.1021/acsami.7b13724. [18] S. Wang, R. Yi, Y. Huang, H. Zuo, Y. Zhang, Y. Huang, L. Chen, S. Liang, Ion-cage structure of graphene oxide membranes with stable interlayer spacings towards efficient desalination, Sep Purif. Technol. 336 (2024), https://doi.org/10.1016/j. seppur.2024.126296. [19] X. Wang, H. Wang, Y. Wang, J. Gao, J. Liu, Y. Zhang, Hydrotalcite/graphene oxide hybrid nanosheets functionalized nanofiltration membrane for desalination, Desalination (2019) 209–218, https://doi.org/10.1016/j.desal.2017.05.012. [20] Q. Li, T. Zhang, Z. Dai, F. Su, X. Xia, P. Dong, J. Zhang, A novel positively charged nanofiltration membrane stimulated by amino-functionalized MXene Ti3C2Tx for high rejection of water hardness ions, J. Memb. Sci. 671 (2023), https://doi.org/ 10.1016/j.memsci.2023.121385. [21] Y. Feng, X. Meng, Z. Wu, J. Chen, C. Sun, S. Huo, A novel polyurea nanofiltration membrane constructed by PEI/TA-MoS2 for efficient removal of heavy metal ions, Sep Purif. Technol. 300 (2022), https://doi.org/10.1016/j.seppur.2022.121785. [22] M. Moslehi, Positively charged thin film composite nanofiltration membrane via layer-by-layer interfacial polymerization based on PU nanofibrous support, J. Environ. Chem. Eng. 10 (2022), https://doi.org/10.1016/j.jece.2022.108769. [23] M. Wang, W. Dong, Y. Guo, Z. Zhai, Z. Feng, Y. Hou, P. Li, Q.J. Niu, Positively charged nanofiltration membranes mediated by a facile polyethyleneimine-Noria interlayer deposition strategy, Desalination 513 (2021), https://doi.org/10.1016/j. desal.2020.114836. [24] J.R. Du, F. Du, K. Yang, J. Zheng, K. Ma, C. Du, H. Yang, Screening of highefficiency crosslinking agents for poly(N,N-dimethylaminoethyl methacrylate) and performance of its nanofiltration membranes, J. Memb. Sci. 683 (2023), https:// doi.org/10.1016/j.memsci.2023.121868. [25] G. Akonkwa Mulungulungu, T. Mao, K. Han, Two-dimensional graphitic carbon nitride-based membranes for filtration process: progresses and challenges, Chem. Eng. J. 427 (2022), https://doi.org/10.1016/j.cej.2021.130955. [26] M.A. Qamar, M. Javed, S. Shahid, M. Shariq, M.M. Fadhali, S.K. Ali, M.S. Khan, Synthesis and applications of graphitic carbon nitride (g-C3N4) based membranes for wastewater treatment: a critical review, Heliyon 9 (2023), https://doi.org/ 10.1016/j.heliyon.2022.e12685. [27] T. Zheng, X. Zou, M. Li, S. Zhou, Y. Zhao, Z. Zhong, Two-dimensional graphitic carbon nitride for membrane separation, Chin. J. Chem. Eng. 42 (2022) 297–311, https://doi.org/10.1016/j.cjche.2021.01.011. [28] S. Sarode, P. Upadhyay, M.A. Khosa, T. Mak, A. Shakir, S. Song, A. Ullah, Overview of wastewater treatment methods with special focus on biopolymer chitin-chitosan, Int. J. Biol. Macromol. 121 (2019) 1086–1100, https://doi.org/10.1016/j. ijbiomac.2018.10.089. [29] E. S, A. G, I. A F, G. P S, L.T. Y, Review on characteristics of biomaterial and nanomaterials based polymeric nanocomposite membranes for seawater treatment application, Environ. Res. 197 (2021) 111177, https://doi.org/10.1016/j. envres.2021.111177. [30] Q. Guo, X. Wu, Y. Ji, Y. Hao, S. Liao, Z. Cui, J. Li, M. Younas, B. He, pH-responsive nanofiltration membrane containing chitosan for dye separation, J. Memb. Sci. 635 (2021), https://doi.org/10.1016/j.memsci.2021.119445. [31] S. Rauthan, A. Singh, Exploring trends of wastewater treatment by using nanomaterials and their composites with bio-polymer, Environ. Conserv J. 25 (2024) 1252–1256, https://doi.org/10.36953/ECJ.30492024. [32] N. Pandey, S.K. Shukla, N.B. Singh, Water purification by polymer nanocomposites: an overview, Nanocomposites 3 (2017) 47–66, https://doi.org/10.1080/ 20550324.2017.1329983. [33] Z. Wang, Z. Yang, X. Shi, Q. Zhang, W. Ma, Q. Bi, Y. Lin, Enhancement of Mg2+/Li +selectivity and water permeability in nanofiltration membranes using dopaminefunctionalized graphitic carbon nitride, J. Memb. Sci. 731 (2025), https://doi.org/ 10.1016/j.memsci.2025.124247. [34] L. Yang, F. Jia, Z. Juan, D. Yu, L. Sun, Y. Song, Y. Wang, L. Huang, J. Tang, Highpermeable graphene oxide/graphitic carbon nitride composite nanofiltration membrane for selective separation of dye and desalination, J. Environ. Chem. Eng. 11 (2023), https://doi.org/10.1016/j.jece.2023.109306. [35] Z. Xu, Z. Mai, Y. Wu, X. Li, X. Zeng, C. Meng, G. Li, H. Zhu, Investigation on the rejection of mixed salt solution by g-C3N4 functionalized nanofiltration membrane, Desalination 593 (2025), https://doi.org/10.1016/j. desal.2024.118226. [36] Y. Wang, J. Gong, J. Li, F. Sang, S. Fang, H. Zhou, L. Tang, Q. Niu, Double-charged self-assembled rGO/g-C3N4 membrane prepared by “functional group substitution” for heavy metal ions rejection at low pressure, Sci. Total Environ. 865 (2023), https://doi.org/10.1016/j.scitotenv.2022.161234. [37] Z. Jin, X. Jiang, Q. Zhang, S. Huang, L. Zhang, L. Huang, T. He, H. Zhang, T. Ohno, S. Ruan, Y.J. Zeng, Infrared response in photocatalytic polymeric carbon nitride for water splitting via an upconversion mechanism, Commun. Mater. 1 (2020), https://doi.org/10.1038/s43246-020-00093-z. [38] C. Morgante, X. Ma, X. Chen, D. Wang, V. Boffa, V. Stathopoulos, J. Lopez, J. L. Cortina, A. Cipollina, A. Tamburini, G. Micale, Metal-Organic-Framework-based nanofiltration membranes for selective multi-cationic recovery from seawater and brines, J. Memb. Sci. 685 (2023), https://doi.org/10.1016/j.memsci.2023.121941. [39] X. Zhu, Z. Sun, F. Tan, F. Chen, X. Luo, F. Wang, D. Wu, H. Liang, D. Xu, X. Cheng, Tailoring high-performance polyester loose nanofiltration membrane for selective separation of salt/dyes: the equilibrium of condensation and hydrolysis, Sep Purif. Technol. 333 (2024), https://doi.org/10.1016/j.seppur.2023.125848. [40] H. Li, S. Xiao, X. Zhao, J. Yuan, S. Yu, Preparation of high-flux loose nanofiltration membranes for efficient dye/salt separation by controlling interface polymerization through physical and chemical dual constraints, Sep Purif. Technol. 362 (2025), https://doi.org/10.1016/j.seppur.2025.131720. P. Iaconis et al. Journal of Environmental Chemical Engineering 13 (2025) 119495 9