Scaling‑free electro‑membrane crystallization enabled by electric field-assisted organic acid control

Abstract

Mineral scaling, driven by concentration polarization, is an inevitable challenge in electro-membrane technologies for recovering valuable metals from lithium-ion battery (LIB) leachates. Here, we elucidate the synergistic scaling mechanism and propose an electric field-assisted organic acid control (eF-OAC) principle to prevent membrane mineral scaling while enabling metal recovery. The eF-OAC utilizes an electric field to direct organic acids into the membrane matrix, where ligand-metal coordination dissolves hydroxide-rich scale and mobilizes metal species for selective transport and crystallization. Dynamic experimental analysis reveals that concentration polarization enriches metal hydroxides at the membrane interfaces and pores. Static density functional theory calculations further demonstrate that organic ligands are thermodynamically favored over sulfonate groups for coordinating metal hydroxides, enabling targeted dissolution from the membrane. Leveraging this molecular-level selectivity, the eF-OAC converts scale into recoverable crystalline products. In optimized electro-membrane crystallization, this eF-OAC achieves high recovery rates of 97.6% for Li⁺, 95.3% for Ni²⁺, 94.2% for Co²⁺, and 97.8% for Mn²⁺. It further yields high-purity Li₂SO₄ (99.4%), Ni(OH)₂ (99.2%), Co(OH)₂ (99.0%), and Mn₃O₄ (99.3%). This study clarifies the mechanisms of mineral scaling and demonstrates a scaling-free approach that transforms fouling liabilities into valuable resource recovery for spent LIBs.

It is challenging to extract lithium from lithium-ion battery leachates due to mineral scaling. Here the authors report a field assisted organic acid control method to optimize lithium extraction from ionic mixtures.

Introduction

With the rapid expansion of the electronics and electric vehicle industries, lithium-ion batteries (LIBs) have become indispensable energy storage systems owing to their high energy density and lightweight design¹. Global stocks of spent LIBs are projected to exceed 11 million tons worldwide by 2030², creating both an environmental burden and a strategic opportunity³. Leachates generated from spent LIBs contain high concentrations of valuable metals, including lithium (Li), cobalt (Co), nickel (Ni), and manganese (Mn)^4,5, making efficient and sustainable recovery imperative. Conventional recycling routes for LIB primarily rely on pyrometallurgy and hydrometallurgy^6,7, yet these approaches are limited by intensive reagent consumption, substantial waste generation, secondary pollution, and complicated multistep separation schemes^8–10 (Fig. 1a). These drawbacks motivate the urgent development of alternative technologies capable of achieving high metal recovery efficiency while minimizing environmental impact and process complexity^11,12.

Fig. 1. Conventional electro-membrane technologies for LIBs recycling.

a Schematic illustration of conventional pyrometallurgical and hydrometallurgical processes for LIBs recycling. b Representative metal ion concentration profiles across a CM during an electro-driven separation process. c Schematic C-V response of a CM under reverse bias. d Formation of concentration polarization-induced mineral scaling resulting from localized OH^- generation and metal hydroxide precipitation. e Conceptual schematic of the eF-OAC, enabling simultaneous mineral scaling mitigation and enhanced metal crystallization.

Membrane technologies have attracted considerable interest for ion-selective resource recovery, owing to their tailorable functional groups and surface properties^13–16. Among them, electro-membrane technologies such as selective ion-distillation electrodialysis and metal-complexation electrodialysis have demonstrated high permselectivity and controllable ion transport for metal ion recovery^17–21. Selective ion-distillation electrodialysis combines electric-field driven ion migration using monovalent ion-exchange membranes to achieve selective enrichment of monovalent ions from mixed multicomponent solutions^22,23. Metal-complexation electrodialysis relies on chelating ligands to convert targeted metal cations into negatively charged complexes, enabling effective separation of transition metals²⁴. Leveraging these attributes, the concept of electro-membrane crystallization (e-MC) has recently been introduced to directly induce metal ion crystallization via electric-driven ion migration^25,26. By applying a targeted electric field, e-MC drives the directional migration of metal ions across ionexchange membranes, concentrating the target ions in a designated chamber where they can directly form crystalline products. This design not only accelerates the ion transfer process but also ensures that the target metal ions are enriched to a degree sufficient for crystallization. However, when treating metal-rich LIB leachates, the e-MC is severely constrained by concentration polarization-induced mineral scaling, where crystals accumulate on ion exchange membranes, impairing system efficiency and long-term stability^27–30.

In e-MC, concentration polarization manifests as the formation of steep ion concentration gradients at the membrane-solution interface (Fig. 1b). This phenomenon is commonly diagnosed using current-voltage (C-V) curves, which exhibit ohmic, limiting, and overlimiting regions (Fig. 1c)³¹. When the applied current density exceeds the limiting current density (LCD), ion depletion intensifies within the diffusion boundary layer, leading to charge imbalance^32,33. Under these conditions, water molecules undergo forced dissociation to generate H⁺ and OH^- for charge compensation^34,35. The localized accumulation of OH^- near the membrane surface reacts with metal ions, triggering the formation of insoluble hydroxide precursors and ultimately mineral scaling (Fig. 1d)^36,37. Despite its critical importance, the fundamental mechanisms governing concentration polarization-induced scaling in electro-membrane systems remain insufficiently understood^38,39. In particular, the coupled effects of electric fields, ion-membrane interactions, and membrane physicochemical properties have not been systematically elucidated. Moreover, direct visualization of the spatial-temporal evolution of mineral scaling across diffusion boundary layers, membrane interfaces, and pores is still lacking. Although current membrane cleaning methods, such as chemical cleaning with inorganic acids and physical cleaning, can partially restore scaling, they are limited by being inefficient and unsustainable. This is because inorganic acid cannot effectively attack complex metal oxide networks embedded within membrane pores. Moreover, it generates mixed-metal wastewater that not only causes secondary pollution but also leads to the loss of valuable metal resources. These knowledge gaps impede the rational design of effective mineral scaling mitigation and resource-oriented membrane cleaning strategies.

To address this, we systematically investigate the relationship between water hydrolysis and the spatio-temporal evolution of mineral scaling formation in the e-MC system. A series of cation-exchange membranes (CMs) with different properties, including ion exchange capacity and membrane thickness, is fabricated to elucidate the relationship between membrane characteristics and concentration polarization quantitatively. Advanced characterization methods are employed to visualize the evolution of mineral scaling across both surface and internal membrane regions. Building on these mechanistic insights, we propose an electric field-assisted organic acid control (eF-OAC) principle, which integrates electrochemical and coordination-chemistry principles to achieve a scaling-free environment for metal ion crystallization (Fig. 1e and Supplementary Fig. 1). The eF-OAC utilizes the applied electric field to control organic acids through the membrane surface and into its internal pore structure. Leveraging coordination and pH-buffering properties, these organic acids supply protons to attack metal oxide bonds while simultaneously forming water-soluble chelates with the metal ions, thereby enabling efficient membrane cleaning and metal recovery. Static density functional theory (DFT) calculations further confirm the specific molecular interactions between the organic acids and target metal ions. By elucidating the fundamental mechanisms governing mineral scaling and introducing a scalable strategy that transforms fouling challenges into resource recovery opportunities, this work presents a green and sustainable pathway for advanced spent LIBs recycling.

Results

Structural determinants of membranes in concentration polarization

The C-V curve provides quantitative insight into the dynamic coupling between ion transport and concentration polarization during electro-membrane operation. An electrochemical system is used to determine the C-V curve by using CMs fabricated from sulfonated polysulfone (SPSF) with different properties (e.g., thicknesses and polymer concentration) (Supplementary Figs. 2-3). Under constant current density, decreasing the concentration of the dilute solution lowers the LCD, thereby promoting the onset of concentration polarization (Fig. 2a and Supplementary Fig. 4a–c). This demonstrates that concentration polarization is intrinsically unavoidable in e-MC systems. Once the replenishment of ions from the bulk solution becomes insufficient to sustain the imposed current, water is dissociated to generate H⁺ and OH^– charge carriers¹⁵. The transport of H⁺ across the CM to the concentrate chamber consequently enriches OH^– in the dilute chamber, accelerating the formation of metal hydroxide precipitates and mineral scaling. Notably, even at initially sufficient ion concentrations, prolonged operation under constant current density can eventually exceed the LCD, leading to unavoidable concentration polarization during extended operation.

Fig. 2. Morphological and structural characterization of mineral scaling.

a Effect of ion concentration on LCD. b Effect of membrane thickness on LCD measured in 0.01 M Ni²⁺, Co²⁺, and Mn²⁺. c Effect of polymer concentration on LCD measured in 0.01 M Ni²⁺, Co²⁺, and Mn²⁺. d Effect of IEC on LCD measured in 0.01 M Ni²⁺, Co²⁺, and Mn²⁺. IEC: ion exchange capacity. e Effect of membrane surface charge on LCD measured in 0.01 M Ni²⁺, Co²⁺, and Mn²⁺. QPSF: quaternized polysulfone. SPSF: sulfonated polysulfone. f Effect of porous membrane structure on LCD measured in 0.01 M Ni²⁺, Co²⁺, and Mn²⁺. The scale bar is 1 cm. g SEM images of SPSF membrane. The scale bar is 100 μm. h TEM images of fresh SPSF membrane and the corresponding EDS mapping for C, O, and S elements. i TEM images of fouled SPSF membrane and the corresponding EDS mapping for Ni, Co, Mn, and O elements. j XRD spectra of the Ni-induced precipitate. k XRD spectra of the Co-induced precipitate. l XRD spectra of the Mn-induced precipitate. m XRD spectra of the mixed Ni, Co, and Mn precipitates. The error bars represent the standard deviation of data from three replicate measurements. Source data are provided as a Source data file.

Membrane thickness and polymer concentration play a relatively minor role in regulating concentration polarization. As membrane thickness increases from 10 to 40 µm, the LCD slightly increases, followed by a decline when the membrane thickness further increases to 100 µm (Fig. 2b and Supplementary Fig. 5a–c). Thinner membranes provide shorter ion transport pathways, which promotes rapid ion depletion at the membrane-solution interface in the dilute chamber, triggering the early onset of concentration polarization. In contrast, thick membranes impose longer ion transport pathways and hinder ion migration. Under these conditions, charge transport increasingly relies on water dissociation, generating H⁺ with a small hydrated radius to sustain charge flux. Consistently, increasing membrane thickness is accompanied by a more pronounced pH decrease in the concentrate solution (Supplementary Fig. 6a–c), indicating that severe concentration polarization occurs in the dilute chamber. Moreover, the elevated polymer concentration directly lowers the LCD, thereby inducing severe concentration polarization (Fig. 2c and Supplementary Fig. 7a–c). This phenomenon induces a rapid pH decline, driven by the intensified water hydrolysis on the membrane matrix (Supplementary Fig. 8a–c).

The CMs with lower ion exchange capacity (IEC) exhibit accumulation of metal ions at the membrane-solution interface²⁷, leading to the formation of an electrical double layer (EDL). The EDL creates an electrostatic repulsion that significantly hinders metal ion transport through the membrane, thereby exacerbating concentration polarization in the diffusion boundary layer (Fig. 2d and Supplementary Fig. 9a–c). Consequently, membranes with lower IEC demonstrate the lower pH in the dilute chamber due to enhanced water hydrolysis (Supplementary Fig. 10a–c). The electrostatic repulsion effect becomes particularly pronounced in the quaternized PSF (QPSF) modified membrane, where an additional positively charged layer exhibits electrostatic repulsion toward metal ions (Fig. 2e and Supplementary Fig. 11). In e-MC, the enhanced repulsion forces extensive water hydrolysis to maintain current, resulting in a more rapid pH decline in the concentrate chamber compared to the SPSF membrane (Supplementary Fig. 12a–c). In addition, mineral scaling behavior is strongly dependent on the surface charge of the foulants⁴⁰. For example, positively charged Ni/Mn oxides exhibit limited deposition due to electrostatic repulsion, whereas negatively charged Co oxides demonstrate significantly higher fouling propensity (Supplementary Fig. 13a-b). Moreover, the implementation of the porous membrane yields a higher LCD value, demonstrating that reduced transport resistance effectively mitigates concentration polarization (Fig. 2f and Supplementary Fig. 14). Accordingly, the porous membrane maintained a lower pH decreasing trend in the dilute chamber compared to the dense membrane (Supplementary Fig. 15a–c), highlighting the importance of structural design in polarization management.

Mineral scaling behavior of the metal ions

Upon the onset of concentration polarization, locally generated OH^- reacts with metal cations to form insoluble precipitates, which initiate mineral scaling. Scanning electron microscopy (SEM) reveals metal precipitation on the membrane surface following e-MC operation (Fig. 2g). The distinct crystal morphologies with elemental distribution are observed for Mn, Ni, and Co, respectively (Supplementary Fig. 16). Transmission electron microscopy (TEM) provides direct insight into the structural analysis of mineral scaling. For the fresh membrane, elemental mapping shows only C, O, and S associated with the polymer backbone and sulfonic acid groups (Fig. 2h). In contrast, after ED operation, TEM images reveal well-defined crystalline domains enriched in Mn, Ni, and Co within the membrane structure (Fig. 2i). Selected area electron diffraction (SAED) patterns confirm the crystalline state of these metal-containing phases.

X-ray diffraction (XRD) analysis reveals distinct crystalline phases constituting mineral scaling. In the e-MC, operating with Ni²⁺ solution yields characteristic diffraction peaks corresponding to Ni(OH)₂ (PDF#14-0117) (Fig. 2j and Supplementary Fig. 17a). In Co²⁺ solution, both Co(OH)₂ and CoOOH phases (PDF#30-0443) are detected, indicating oxidative conversion of Co(OH)₂ during e-MC operation (Fig. 2k and Supplementary Fig. 17b). Mn²⁺ treatment results in Mn₃O₄ formation (PDF#30-0443), demonstrating a similar oxidative transformation from initially formed Mn(OH)₂ (Fig. 2l and Supplementary Fig. 17c). Such differences in oxidation behavior are governed by redox potential and phase thermodynamic stability. The Mn²⁺ has a lower redox potential and is more susceptible to oxidation under an electric field and ambient oxygen, whereas Ni²⁺ exhibits high redox inertness and rarely undergoes further oxidation. When mixed Ni²⁺, Co²⁺, and Mn²⁺ solutions are processed, XRD confirms the simultaneous presence of Mn₃O₄, Ni(OH)₂, Co(OH)₂, and CoOOH (Fig. 2m). To elucidate the sequence of mineral scaling formation, the dissolution and precipitation zones for different metal ions are calculated by Visual MINTEQ (Supplementary Fig. 18a–c)^41,42. The results indicate that Ni²⁺ precipitates first as Ni(OH)₂, followed by Co²⁺, while Mn²⁺ precipitates last under concentration polarization conditions.

The extent of mineral scaling is quantitatively evaluated by analyzing metal ion loss and permeation rate decline. Increasing membrane thickness intensifies both ion loss and permeation rate reduction due to aggravated concentration polarization (Supplementary Fig. 19a). Similarly, membranes fabricated with higher polymer concentrations show more severe ion permeation rate losses owing to the denser polymeric structure (Supplementary Fig. 19b). A lower IEC further aggravates ion permeation rate loss through enhanced electrostatic repulsion from the EDL, which amplifies concentration polarization (Supplementary Fig. 19c). Surface modification with positively charged layers on the QPSF#SPSF membrane creates particularly strong electrostatic repulsion against metal ions, leading to the most pronounced permeation rate loss among all tested membranes (Supplementary Fig. 19d). However, compared to the SPSF membrane, the as-prepared QPSF#SPSF membrane exhibits reduced foulant adhesion due to electrostatic repulsion from positively charged surfaces. Consequently, a substantial fraction of surface precipitates can be removed by simple deionized water cleaning. Porous membranes with large internal void spaces can enhance ion permeation rates by providing low-resistance pathways for ion transport (Supplementary Fig. 19e). However, water dissociation more readily occurs within the microporous structure due to the confined space. Although some surface deposits are detached during water washing, precipitates remain embedded within the porous structure (Supplementary Fig. 20). These findings systematically demonstrate how membrane structural characteristics govern ion transport by modulating concentration polarization and subsequent scaling behavior.

Mineral scaling mitigation in e-MC via the eF-OAC

Based on the mineral scaling analysis of the prepared membrane, a further exploration of membrane cleaning for the commercial monovalent-selective CIMS membrane in the e-MC system is conducted. The strong electrostatic repulsion towards multivalent cations ensures exceptional Li⁺ selectivity but simultaneously leads to the severe scaling of transition metals (Fig. 3a). From C-V curve analysis, the decreased ion concentration of feed solution leads to a continuous decrease in LCD (Fig. 3b and Supplementary Fig. 21). Thus, significant concentration polarization occurs due to ion depletion in the dilute chamber, which causes water hydrolysis and mineral scaling (Supplementary Fig. 22). Atomic force microscope (AFM) demonstrates that the initially smooth CIMS membrane surface (Ra = 14.5 nm) becomes heavily covered with crystalline deposits after operation, with surface roughness increasing to 53.7 nm (Fig. 3c). The EDS confirms the presence of Ni, Co, and Mn across both the membrane surface and cross-section (Supplementary Figs. 23a–c). Water contact angle measurements further indicate a transition from the hydrophilic nature of fresh CIMS membrane to a fouled surface modified by deposited hydrophilic metal crystals (Supplementary Fig. 24a)⁴³. X-ray photoelectron spectroscopy (XPS) identifies characteristic peaks at 642.3 eV (Mn 2p), 780.42 eV (Co 2p), and 855.48 eV (Ni 2p) (Fig. 3d), confirming the accumulation of mineral scaling species. Attenuated total reflectance-fourier transform infrared (ATR-FTIR) spectra show the prominent and broad O–H stretching band observed at 3453 cm^–1, which is primarily attributed to the formation of metal hydroxide precipitates on the membrane surface and within its pores (Fig. 3e)⁴⁴. The thermogravimetric analysis (TGA) reveals that the fouled membrane exhibits a higher residual mass and a significant reduction in the final degradation stage (to 48%), confirming substantial composition variation due to mineral scaling deposition (Supplementary Fig. 24b). As a result, the membrane area resistance increases sharply to 8.8 Ω cm² after scaling (Supplementary Fig. 15a).

Fig. 3. Evaluation of membrane cleaning effectiveness.

a Photographs of the fouled CIMS (I) membrane facing anode at current density of 2 mA cm^–2, CIMS (II) membrane facing cathode at current density of 2 mA cm^–2, CIMS (III) membrane facing anode at current density of 1 mA cm^–2, and spacer (IV). b C-V curves of the CIMS membrane measured in mixed ion solutions (Li⁺, Ni²⁺, Co²⁺, and Mn²⁺) with different concentrations. c AFM images of fresh membrane and fouled membrane. d XPS spectra. e ATR-FTIR spectra of fresh membrane, fouled membrane, and cleaned membrane. f Changes in membrane area resistance after cleaning with different cleaning agents without using eF-OAC. Cleaning is performed with 0.1 mol L⁻¹ cleaning agent for 1 h. The scale bar is 1 cm. g Changes in membrane area resistance after cleaning with different cleaning agents by using eF-OAC. Cleaning is performed with 0.1 mol L^–1 cleaning agent for 1 h. The scale bar is 1 cm. h Time-dependent evolution of membrane area resistance during citric acid and EDTA-2Na cleaning. The scale bar is 1 cm. i TEM images of cleaned membrane. The error bars represent the standard deviation of data from three replicate measurements. Source data are provided as a Source data file.

To mitigate mineral scaling during e-MC, we test the efficacy of inorganic acids (HCl, HNO₃, and H₂SO₄) and organic acids (formic acid, oxalic acid, citric acid, ethylenediaminetetraacetic acid disodium (EDTA-2Na), ethylenediaminetetraacetic acid tetrasodium (EDTA-4Na)) on the scaling of the CIMS membrane. The inorganic acids show limited cleaning effectiveness (Fig. 3f), as membrane resistance persists at a higher level even at acid concentrations up to 1 mol L^–1 (Supplementary Fig. 26). More notably, inorganic acids fail to form stable complexes with metal ions. The dissolved metal ions easily re-deposit on the membrane matrix, leading to secondary pollution and the loss of valuable metal resources. In contrast, organic chelating agents significantly improve fouling removal. The EDTA-2Na outperforms EDTA-4Na. At its natural pH of ~4.4, EDTA-2Na provides both chelating capacity and a mildly acidic environment, which supplies H⁺ ions to dissolve metal oxyhydroxide on the membrane surface directly. In contrast, EDTA-4Na solutions are strongly alkaline (pH ~11.2). This high pH inhibits the protonation and dissolution of metal oxyhydroxides, limiting their ability to break down the metal oxyhydroxides. Among the organic chelating agents, the synergistic combination of EDTA-2Na and citric acid proves particularly effective. Cleaning with EDTA-2Na and citric acid induces a clear color change of the fouled membrane, indicating effective dissolution of metal oxyhydroxides. Citric acid facilitates dissolution of Mn, Co, and Ni oxyhydroxides on the membrane surface⁴⁵, while the EDTA-2Na strongly chelates the released metal ions, extracting the metal ions from the membrane⁴⁶. This dual mechanism restores the membrane resistance to a lower value.

Despite the effectiveness of organic acids, static cleaning without an electric field still shows limited cleaning efficiency (Fig. 3f and Supplementary Tables 1–2). This is because the cleaning agents are in a disordered dispersion state in the solution and cannot be effectively utilized for cleaning the mineral scaling inside the membrane matrix. The application of an electric field addresses the shortcomings by driving the directional migration of cleaning agents, significantly enhancing cleaning efficiency (Fig. 3g). By introducing organic acids into the feed chamber and modulating the direction of the applied electric field to control their migration toward the membrane surface and into its internal pores, targeted removal of mineral scaling deposits is achieved (Supplementary Table 3 and Supplementary Fig. 27). Therefore, the eF-OAC is proposed to completely solve the mineral scaling in the e-MC system. By leveraging the applied electric field in the e-MC system to direct the migration of organic acids (EDTA-2Na and citric acid) towards the membrane, the eF-OAC achieves efficient removal of both surface and internal scale. Compared with static cleaning without an electric field, the application of an electric field significantly enhances cleaning efficiency by driving the directional migration of cleaning agents into membrane pores (Fig. 3h). The SEM further corroborates the eF-OAC cleaning performance. EDS analysis shows a substantial decrease in Ni, Co, and Mn signals after individual citric acid/EDTA-2Na treatments, with complete removal achieved using the combined strategy (Supplementary Fig. 28). AFM images also reveal that large surface bulges and pore-blocking foulants present on the fouled membrane are replaced by minor surface features after cleaning (Supplementary Fig. 29). XPS spectra confirm the disappearance of Mn 2p, Co 2p, and Ni 2p signals (Supplementary Fig. 30), and water contact angle measurements return to values comparable to the fresh membrane. ATR-FTIR spectra show no alteration in membrane functional groups, and TGA curves of the cleaned membrane overlap completely with those of the fresh sample, providing strong evidence for complete foulant removal. TEM confirms the effective removal of internal precipitates after eF-OAC treatment. Elemental mapping shows the complete elimination of Mn, Ni, and Co from the membrane (Fig. 3i), while the SAED patterns indicate the absence of crystalline phases within the membrane.

Mechanism of the eF-OAC toward scaling-free operation

The phenomenon of concentration polarization, which serves as the core trigger for mineral scaling, exhibits distinct mechanisms depending on membrane transport properties. For high-permeability membranes, characterized by reduced thickness or increased IEC, polarization predominantly originates from rapid ion depletion within the diffusion boundary layer (Supplementary Fig. 31a). Conversely, low-permeability membranes with increased thickness and reduced IEC exhibit concentration polarization through EDL formation, where ion accumulation near the membrane surface generates strong electrostatic repulsion, suppressing ion transport and intensifying concentration polarization (Supplementary Fig. 31b). This effect is particularly pronounced in monovalent-selective CM (e.g., CIMS membrane), where the positively charged surface layer establishes an ion depletion zone that exacerbates concentration polarization (Fig. 4a). Under polarization conditions, water hydrolysis generates H⁺ and OH^- ions in the diffusion boundary layer. The OH^- readily precipitates dissolved Ni²⁺, Co²⁺, and Mn²⁺ as metal oxides (Supplementary Fig. 32 and Supplementary Movie 1). Simultaneously, within the membrane matrix, transitional ions become strongly bound to sulfonic acid groups through electrostatic interaction, blocking ion-exchange sites. The internal water hydrolysis induced by polarization produces OH^- ions that react with metal ions, resulting in the direct growth of metal precipitates within membrane pores Eqs. (1)–(5), (Fig. 4b). This coupled interfacial and intrapore precipitation mechanism accounts for the severe and persistent scaling observed in monovalent-selective membranes. To rationalize the strong tendency for metal accumulation at sulfonated membranes, DFT simulations are performed to evaluate the adsorption of metal ions and their hydroxide species on sulfonate groups under aqueous conditions. As shown in Fig. 4c, Ni²⁺, Co²⁺, and Mn²⁺ exhibit moderately strong adsorption on sulfonate, with calculated Eads values of –2.76, –2.65, and –2.35 eV/ion, respectively. In contrast, Li⁺ exhibits much weaker interaction with sulfonate (-0.35 eV/ion in Supplementary Fig. 33a), consistent with its preferential transport through the membrane. A similar trend is observed for hydroxide species that Li–OH shows negligible binding to sulfonate (–0.24 eV/ion in Supplementary Fig. 33b), whereas Ni–OH, Co–OH, and Mn–OH bind strongly (Fig. 4d), reflecting the experimentally observed hydroxide-assisted metal accumulation under polarization conditions.

Fig. 4. Mineral scaling mechanism analysis.

a Mechanism of positively charged modified membrane-induced concentration polarization. b Mechanism of concentration polarization-induced mineral scaling. c Calculated adsorption energies of metal ions (Ni²⁺, Co²⁺, and Mn²⁺) on sulfonic acid and citric acid groups. d Calculated adsorption energies of metal hydroxide species (metal–OH) on sulfonic acid and citric acid groups. e Displacement energies for metal ions and metal–OH species from sulfonic acid to citric acid. f Representative structures, M–O bond lengths, and charge density difference plots for metal species adsorbed on sulfonic acid and citric acid groups. The isosurface level is set to 0.15 e/ų. Sulf: sulfonate. Cit: citrate. g Mechanism of mineral scaling cleaning using citric acid and EDTA-2Na.

Based on this scaling mechanism, the eF-OAC is proposed (Supplementary Fig. 34). The eF-OAC utilizes the applied electric field to directionally drive organic acids toward the membrane surface and into its internal pores (Supplementary Fig. 35). The electric field not only enhances the local concentration of chelating agents at scaling-prone sites but also accelerates the collision frequency between chelating agents and scaling species. Citric acid, containing multiple carboxylate groups, can dissolve metal precipitates via proton-assisted pathways while simultaneously coordinating with metal centers through multidentate chelation^45,47. Structural analysis from DFT reveals a clear coordination-mode contrast between sulfonate and citrate groups. The displacement energy is defined as ΔEdisp = Ecitrate–Esulfonate, where Ecitrate and Esulfonate represent the total energies of the metal species coordinated to citrate and sulfonate sites within the same simulation cell, respectively. As shown in Fig. 4e, all ΔEdisp values are negative for both metal ions and metal–OH species, demonstrating a clear energetic preference for citrate coordination over sulfonate binding. Metal ions and metal–OH species bind to sulfonate predominantly through a single M–O contact, whereas citrate consistently forms bidentate coordination involving two M–O bonds for all metal species examined. For example, Co²⁺ binds to sulfonate through a single Co–O bond (1.90 Å), while citrate coordinates Co²⁺ via two Co–O bonds with distances of 2.00 and 2.17 Å (Fig. 4f). Charge density difference maps show pronounced electron accumulation along the Co–O bonds and around the metal centers upon citrate coordination (Supplementary Fig. 33c-d), consistent with the formation of multiple metal–oxygen bonding interactions. Correspondingly, charge transfer (Δq) analysis indicates consistently greater electron redistribution for citrate than for sulfonate coordination. The charge transfer from metal ions to ligands increases from 0.40–0.48 e on sulfonate to 0.79–0.83 e on citrate, and from 0.93–1.32 e to 1.19–1.46 e for metal–OH species.

Similarly, EDTA-2Na, a polyprotic organic acid, enables rapid removal of metal-containing scales from both the membrane surface and internal pores. Upon dissolution, EDTA-2Na predominantly exists as [EDTA]⁴⁻ in solution. The [EDTA]⁴⁻ ligand possesses six strong donor atoms (i.e., two nitrogens and four oxygens), enabling octahedral coordination with metal cations. Before the bulk H⁺ attack, organic anions preferentially adsorb onto coordinatively metal sites on the oxide scale surface, forming incipient surface complexes (Supplementary Figs. 36-37). This pre-coordination triggers an electronic redistribution at the metal center, thereby weakening the adjacent metal-oxygen lattice bonds. Consequently, the activated metal-oxygen bonds exhibit markedly reduced resistance to subsequent protonolysis. The H⁺ can easily attack the pre-weakened metal-oxygen bonds, releasing metal ions into solution. These liberated metal ions subsequently coordinate with organic acids to form stable chelate complexes. This chelation-driven sequestration shifts the dissolution equilibrium forward, enabling substantially more efficient removal of mineral scaling than that of the inorganic acids (Eqs.( 6)–(12) and S11-19). The chelation mechanism of EDTA⁴⁻ is further supported by DFT calculations. A representative Co[EDTA]²⁻ complex is examined (Supplementary Fig. 33e), and the calculated adsorption energy and charge transfer closely resemble those of the Co@citrate system, indicating a similar multidentate coordination mode. This result provides molecular-level support for the experimentally observed efficiency of citric acid and EDTA-assisted membrane cleaning and highlights the general role of multidentate organic ligands in removing mineral scaling species (Fig. 4g and Supplementary Figs. 38-39). For traditional offline acid soaking, conventional cleaning agents are homogeneously dispersed in the bulk solution without targeted enrichment at the membrane interface, resulting in low local effective concentration and limited cleaning performance. By contrast, the eF-OAC employs an applied electric field to drive organic acid anions to migrate directionally and accumulate on the membrane surface. This targeted enrichment enables efficient in situ chelation of metal ions and suppresses scale deposition, while converting scaling precursors into recyclable crystalline products. As a result, through the application of eF-OAC, the total cleaning time is reduced to 50 min, a reduction of approximately 58% compared to the 120 min required without electric field assistance (Supplementary Movies 2-3).

Sustainable recovery of metal ions in the scaling-free e-MC system

Guided by the eF-OAC principle, the e-MC system combines selective ion-distillation electrodialysis with metal-complexation electrodialysis to enable LIB recycling and mitigate scaling (Fig. 5a and Supplementary Fig. 40). In this integrated system, Li⁺ is first selectively recovered via ion-distillation electrodialysis, followed by the extraction of metals from mineral scaling using organic acids. Specifically, the applied electric field precisely controls the directional migration of organic acids, enabling targeted membrane cleaning and scaling recovery. The metal-rich effluents generated during cleaning are subsequently routed to the metal-complexation electrodialysis, where Ni²⁺, Co²⁺, and Mn²⁺ are selectively separated and crystallized. This electrochemical approach eliminates the need for extreme temperatures, strong acids/alkalis, and organic solvents while achieving comparable metal recovery rates.

Fig. 5. The proposed eF-OAC for metal ions recovery.

a Schematic diagram of the scaling-free e-MC system. b Temporal evolution of ion concentration during selective ion-distillation electrodialysis at a current density of 1 mA cm^–2. c Temporal evolution of ion concentration during membrane cleaning by using EDTA-2Na and citric acid. d Temporal evolution of ion concentration during metal-complexation electrodialysis at a molar ratio of 1.05. e XRD spectra of the metal products. f Performance analysis of selective ion-distillation electrodialysis at current density of 1 mA cm^–2. g Performance analysis of metal-complexation electrodialysis at molar ratio of 1.05. h Performance analysis of chemical precipitation process. i Comparison of metal ion recovery rate with/without using eF-OAC. j Techno-economic analysis of metal products recovery. The error bars represent the standard deviation of data from three replicate measurements. Source data are provided as a Source data file.

Selective ion-distillation electrodialysis demonstrates effective Li⁺ separation (Supplementary Fig. 41). The polyamino compound layer on CIMS membranes provides exceptional monovalent selectivity^25,44, achieving high rejection of 99.8% for Ni²⁺, 99.7% for Co²⁺, and 99.9% for Mn²⁺ while allowing Li⁺ migration to the recovery chamber with Li⁺ concentrating to 2.7 g L^–1 and 97.6% recovery (Fig. 5b and Supplementary Fig. 42a). Increasing the current density from 1 to 2 mA cm^–2 increases ion permeation rates but slightly compromises selectivity (Supplementary Fig. 42b), with lower multivalent metal rejections and decreased Li⁺ purity from 99.4 to 98.2% (Supplementary Table 4 and Supplementary Fig. 42c-d). However, this high selectivity inherently promotes severe mineral scaling. Mass balance analysis reveals substantial ion losses at 1 mA cm^–2 (i.e., 14.7% Ni²⁺, 12.1% Co²⁺, and 10.7% Mn²⁺), and these ion losses are more severe at a higher current density of 2 mA cm^–2 (Supplementary Fig. 43-44). Concurrently, energy consumption rises markedly due to elevated electrical stack resistance (Supplementary Fig. 45). Interestingly, an increase in the permselectivity for Li⁺/Ni²⁺, Li⁺/Co²⁺, and Li⁺/Mn²⁺ is observed at the later stage of the experiment (Supplementary Fig. 46). This observed enhancement in permselectivity is due to Ni²⁺, Co²⁺, and Mn²⁺ precipitation in the dilute chamber. Instead, Li⁺ transports through CM to the concentrate chamber, thus resulting in a scaling-induced amplification of Li⁺ permselectivity.

Following Li⁺ crystallization by evaporation and drying of the solution in the recovery chamber (Supplementary Fig. 47), the EDTA-2Na and citric acid are introduced into the feed chamber to enable the extraction of metal precipitates on the membrane. Under the electric field, organic acids effectively dissolve and extract metal ions from the membrane surface and pores, which are then transported through the anion exchange membrane into the recovery chamber. Thus, the concentrations of Ni²⁺, Co²⁺, and Mn²⁺ increase continuously in the recovery chamber (Fig. 5c), mainly due to the scale dissolution. Post-operation disassembly of the membrane stack reveals scaling-free membranes and spacers, providing direct visual evidence of effective scale removal (Supplementary Fig. 48).

Metal-complexation electrodialysis further enables efficient separation and recovery of divalent metal ions through organic acid-mediated complexation. This system leverages the distinct coordination affinities of Mn²⁺, Ni²⁺, and Co²⁺ toward EDTA to achieve their selective separation. Ni²⁺ and Co²⁺ form highly stable anionic complexes with EDTA (K_Ni = 2.5 × 10¹⁸ and K_Co = 2.8 × 10¹⁶) while Mn²⁺ exhibits a much weaker interaction (K_Mn = 7.8 × 10¹³) (Supplementary Fig. 49). Consequently, the anionic Ni[EDTA]^2- and Co[EDTA]^2- complexes migrate toward the anode, whereas free Mn²⁺ migrates to the cathode chamber (Fig. 5d), where it is precipitated as Mn(OH)₂ and is subsequently oxidized to Mn₃O₄ (PDF#38-0715) (Fig. 5e). Mn²⁺ recovery is critically dependent on the EDTA:(Ni+Co) molar ratio. At a ratio of 1:1, incomplete complexation leads to partial co-migration of Ni²⁺ and Co²⁺ with Mn²⁺ (Supplementary Figs. 50-51a). Increasing the ratio to 1.05 ensures complete complexation of Ni²⁺ and Co²⁺ (Supplementary Fig. 51b), thereby maximizing Mn₃O₄ purity. Comparative analysis highlights the effectiveness of the eF-OAC in mitigating mineral scaling. In the absence of mineral scaling recovery, the concentrations of Ni²⁺, Co²⁺, and Mn²⁺ in the recovery chamber are much lower compared to those achieved with the eF-OAC (Supplementary Fig. 51c).

Subsequent selective precipitation enables further purification of the residual solution containing Ni²⁺ with Co²⁺. XRD and SEM analysis confirm the production of Ni(OH)₂ (PDF#14-0117) and Co(OH)₂ (PDF#30-0443) (Supplementary Figs. 52-53). Overall, this scaling-free e-MC system achieves a recovery of 97.6% for Li⁺ (Figs. 5f), 97.8% for Mn²⁺ (Figs. 5g), 95.3% for Ni²⁺, and 94.2% for Co²⁺ (Fig. 5h), yielding purities of 99.4% for Li₂SO₄, 99.2% for Ni(OH)₂, 99.0% for Co(OH)₂, and 99.3% for Mn₃O₄, fully meeting battery-grade material requirements (Supplementary Fig. 54a and Supplementary Tables 5–6). The scaling-free e-MC yields supplementary recovery of 16.1% Ni²⁺, 12.4% Co²⁺, and 14.3% Mn²⁺, further boosting the overall metal recovery rate (Fig. 5i). Economic evaluation of the selective ion-distillation electrodialysis shows operating expense (OPEX) of 0.54 $ kg^–1, capital expenditure (CAPEX) of 3.85 $ kg^–1, and total cost of 4.39 $ kg^–1 (Supplementary Fig. 54b and Supplementary Tables 7-8)⁴⁸. Similarly, metal-complexation electrodialysis exhibits the highest purity, and the total cost of 5.16 $ kg^–1 can be achieved simultaneously (Supplementary Fig. 54c and Supplementary Table 9). The combined scaling-free e-MC system delivers substantial profitability, generating 26.1 $ kg^–1 for Li₂SO₄, 3.8 $ kg^–1 for Mn₃O₄, 13.7 $ kg^–1 for Ni(OH)₂, and 17.9 $ kg^–1 for Co(OH)₂ (Fig. 5j and Supplementary Table 10). Meanwhile, Ni(OH)₂ and Co(OH)₂ generate substantial profit margins owing to their low extraction costs and high market valuation, whereas Li₂SO₄ offsets the primary operational expenditures of the system. Although Mn₃O₄ yields the lowest margin due to its lower market value, the collective revenue from these high-purity products confirms the robust economic viability of the e-MC process.

Discussion

By integrating principles of coordination chemistry and electro-driven separation, this work develops a scaling-free electro-membrane crystallization (e-MC) system enabled by an electric field-assisted organic acid control (eF-OAC) principle for sustainable LIBs recycling. Combined dynamic experimental analysis and DFT calculations reveal that concentration polarization induces localized water hydrolysis, generating hydroxyl ions that subsequently react with metal ions to form mineral scaling. In addition, the concentration polarization-induced mineral scaling is governed by the physicochemical properties of the membrane interface. The eF-OAC represents a paradigm shift in membrane operation by transforming the conventional perception of mineral scaling from an operational challenge into a valuable metal recovery opportunity. Organic agents (e.g., citric acid and EDTA-2Na) can achieve protonation of metal-oxygen bonds and extract metal ions through the formation of stable complexes. The applied electric field drives the migration of cleaning agents, enabling targeted delivery not only to the membrane surface but also deep into its internal membrane pores. Thus, the eF-OAC establishes a scaling-free environment for e-MC operation while simultaneously enhancing crystallization. The design concept and validation methodology established in this work transform the potential mineral scaling crisis into a continuous resource recovery opportunity, enhancing the sustainability of the electro-membrane for spent LIB leachate recovery. Future work will upgrade the current batch operation to a semi-continuous cascaded process by adopting serially connected identical ion-distillation electrodialysis, metal-complexation electrodialysis, and chemical precipitation modules. This strategy requires no major modification to the original electrodialysis stack configuration, and can well maintain the high metal recovery rate and purity of the e-MC system.

Methods

Preparation of membranes

The dense SPSF membranes were fabricated via the thermally induced phase separation (TIPS) method (Supplementary Fig. 55). The SPSF casting solution was prepared by dissolving SPSF in DMSO and stirring for 2 h. The different thicknesses of SPSF membranes were cast by the SPSF solution on a dry glass plate at different knife gaps (250 μm, 500 μm, and 1000 μm), then dried at 60 °C for 6 h. The different polymer concentrations (15 wt.%, 17.5 wt.%, and 20 wt.%) of SPSF membranes were cast by the SPSF solution on a dry glass plate at knife gaps of 1000 μm, then dried at 60 °C for 6 hours. The different IECs of SPSF membranes were cast by the mixed PSF and SPSF solution on a dry glass plate at knife gaps of 1000 μm, then dried at 60 °C for 6 h. The porous SPSF membrane was prepared via the non-solvent induced phase separation (NIPS) method. The chloromethylated polysulfone (CMPSF) was prepared by chloromethylation of polysulfone according to the reported method⁴⁹. The CMPSF was dissolved in DMSO to prepare a casting solution with a concentration of 25 wt.%. The QPSF was coated on the SPSF membrane using a film knife with a thickness of 100 μm and dried at 60 °C for 6 h. The QPSF#SPSF membrane was immersed in 33 wt.% trimethylamine solution at 40 °C for 48 h. Then the membrane was immersed in 5 wt.% HCl for 12 h to remove the excess trimethylamine and washed with deionized water.

Characterization methods

The SEM (Sirion 200, FEI, USA) analysis was carried out to evaluate the membrane structure. Qualitative characterization of the elements was achieved through EDS (INCA X-Act, USA). The internal structure of membranes was analyzed by TEM (FEI Tecnai G2F 20, USA). The membrane roughness was measured using an AFM (MFP-3D, Asylum Research, USA). The groups of the membranes were characterized using ATR-FTIR (Nicolet 6700, USA) and XPS (AXIS UltraDLD, Shimadzu, Japan). The membrane surface hydrophilicity was evaluated by measuring the water contact angle (WCA, Dropmeter A-200, MAIST Vision Inspection & Measurement Co., China). The inductively coupled plasma optical emission spectroscopy (ICP-OES, Avio 500, PerkinElmer, USA) was employed for determining the cation concentrations. The thermal stability of the membranes was measured by TGA (TGA 8000, PerkinElmer, USA). The structure of the precipitates was identified by XRD (XRD-6100, Shimadzu, Japan). The zeta potential of the membrane was analyzed by an electrokinetic analyzer (SurPASS, Anton-Paar GmbH, Austria).

Configuration of the scaling-free e-MC system

In selective ion-distillation electrodialysis, the membranes were stacked between anode and cathode in sequence: a bipolar membrane (BM, BP-1E), an AM (AMX), an AM (AMX), an AM (AMX), a monovalent CM (CIMS), a monovalent CM (CIMS), a monovalent CM (CIMS), and a BM (BP-1E) (Supplementary Fig. 56a and Supplementary Table 11). Correspondingly, a feed chamber, a 1st stage chamber, a 2nd stage chamber, and a recovery stage chamber were designed in series. The initial volumes were 2 L of feed wastewater for the feed chamber (Supplementary Tables 12-13), and 0.5 L of deionized water each for the 1st stage chamber, 2nd stage chamber, and recovery chamber. The current of the selective ion-distillation electrodialysis stack was supplied by a power supply at 1 mA cm^–2 and 2 mA cm^–2, respectively. However, concentration polarization-induced water hydrolysis caused serious mineral scaling on the CMs.

$${\mathrm{Ni}}^{2+}+{2{\mathrm{OH}}^{-}\to \mathrm{Ni}\left(\mathrm{OH}\right)}_2\downarrow$$ 1

$${\mathrm{Co}}^{2+}+{2\mathrm{OH}}^{-}\to {\mathrm{Co}\left(\mathrm{OH}\right)}_2\downarrow$$ 2

$${4\mathrm{Co}\left(\mathrm{OH}\right)}_2+{\mathrm{O}}_2\to 4\mathrm{CoOOH}+{2\mathrm{H}}_2\mathrm{O}$$ 3

$${\mathrm{Mn}}^{2+}+{2\mathrm{OH}}^{-}\to {\mathrm{Mn}\left(\mathrm{OH}\right)}_2\downarrow$$ 4

$${6\mathrm{Mn}\left(\mathrm{OH}\right)}_2+{\mathrm{O}}_2\to {2\mathrm{Mn}}_3{\mathrm{O}}_4+{6\mathrm{H}}_2\mathrm{O}$$ 5

Following the selective recovery and crystallization of Li₂SO₄ in the recovery chamber, 0.083 mol L^–1 EDTA-2Na and citric acid are added to the feed chamber for mineral scaling recovery. By regulating the current density at the optimized value of 0.2 mA cm^–2, the applied electric field drives directional migration and interfacial enrichment of organic acids at scaling-prone membrane regions. The organic acids effectively extract metal ions from the mineral scaling, which are then transported through the anion exchange membrane into the subsequent stage.

$${\mathrm{Ni}}^{2+}+{\left[{\mathrm{C}}_6{\mathrm{H}}_5{\mathrm{O}}_7\right]}^{3-}\to {\left[{\mathrm{NiC}}_6{\mathrm{H}}_5{\mathrm{O}}_7\right]}^{-}$$ 6

$${\mathrm{Co}}^{2+}+{\left[{\mathrm{C}}_6{\mathrm{H}}_5{\mathrm{O}}_7\right]}^{3-}\to {\left[{\mathrm{CoC}}_6{\mathrm{H}}_5{\mathrm{O}}_7\right]}^{-}$$ 7

$${\mathrm{Mn}}^{2+}+{\left[{\mathrm{C}}_6{\mathrm{H}}_5{\mathrm{O}}_7\right]}^{3-}\to {\left[{\mathrm{MnC}}_6{\mathrm{H}}_5{\mathrm{O}}_7\right]}^{-}$$ 8

$${\mathrm{Ni}}^{2+}+{\left[\mathrm{EDTA}\right]}^{4-}\to {\mathrm{Ni}\left[\mathrm{EDTA}\right]}^{2-}$$ 9

$${\mathrm{Co}}^{2+}+{\left[\mathrm{EDTA}\right]}^{4-}\to {\mathrm{Co}\left[\mathrm{EDTA}\right]}^{2-}$$ 10

$${\mathrm{Co}}^{3+}+{\left[\mathrm{EDTA}\right]}^{4-}\to {\mathrm{Co}\left[\mathrm{EDTA}\right]}^{-}$$ 11

$${\mathrm{Mn}}^{2+}+{\left[\mathrm{EDTA}\right]}^{4-}\to {\mathrm{Mn}\left[\mathrm{EDTA}\right]}^{2-}$$ 12

In metal-complexation electrodialysis, effluents sourced from the feed chamber, 1st stage, and 2nd stage of the selective ion-distillation electrodialysis, along with the recovery chamber from the membrane cleaning step, were used to separate Mn²⁺ from Ni²⁺ and Co²⁺. The membranes were stacked between anode and cathode in sequence: a CM (CMX), an AM (AMX), a CM (CMX) and an AM (AMX) (Supplementary Fig. 56b). Correspondingly, a feed chamber, a recovery chamber, and a complexation chamber were designed in series. The EDTA-4Na was used to form complex anions such as Ni[EDTA]^2- and Co[EDTA]^2-. At the current density of 4 mA cm^–2, anions were transported toward the complexation compartment through the AM, whereas Mn²⁺ was transported toward the recovery compartment through the CM. To increase the recovery rate of Mn²⁺, different molar ratios of EDTA-4Na to metal ion (Ni²⁺ and Co²⁺) (1.0 and 1.05) were investigated. The Mn²⁺ was crystallized as Mn₃O₄ by adding 2.0 mol L^-1 NaOH solution.

The decomplexation of EDTA from Ni²⁺ and Co²⁺ was achieved by adding H₂SO₄ solution until the pH was below 0.5. Subsequently, the EDTA solid was recovered through filtration and can be reused for EDTA complexation again. The remaining liquor containing Ni²⁺ and Co²⁺ was further treated by chemical precipitation. The NaClO was added at pH 3 to precipitate Co²⁺ as Co₂O₃⁵⁰. After filtering out Co₂O₃, Ni²⁺ in the remaining filtrate was crystallized as Ni(OH)₂ by adding NaOH to increase the pH above 11. Finally, Co₂O₃ was crystallized as Co(OH)₂ by adding H₂SO₃ solution and NaOH solution.

Evaluation methods

The C-V curve of the membranes is tested by Li⁺, Ni²⁺, Co²⁺ and Mn²⁺ solutions in an electrodialysis stack. Current density (I_s, mA cm^–2) is calculated by:

$$I_s=\frac{I}{s}$$ 13

where I (mA) is the current through the membranes, and S (7.065 cm²) is the effective membrane area.

Membrane area resistance (R_S, Ω cm²) is measured in 0.5 M NaCl and calculated by (Supplementary Fig. 57):

$$R_S=\frac{U-U_0}{I^{\prime }}\times S$$ 14

where U is the measured voltage when CM is assembled in the device, U₀ (V) is the voltage without CM, and I’ (0.04 A) is the constant current.

The recovery rate ($\delta$, %) is calculated by:

$$\delta =\frac{C_0{V}_0-C_t{V}_t}{C_0{V}_0}\times 100\%$$ 15

where C₀ (mol L^-1) and C_t (mol L^–1) are the ion concentrations at initial time and time t in the dilute chamber, V₀ (L) and V_t (L) are the volumes of solution at initial time and time t in the dilute chamber.

The Li⁺/M²⁺ permselectivity ($P_{M^{2+}}^{L{i}^{+}}$, %) of the membrane is calculated by:

$$P_{M^{2+}}^{L{i}^{+}}=\frac{J_{L{i}^{+}}\cdot C_{M^{2+}}}{J_{M^{2+}}\cdot C_{L{i}^{+}}}\times 100\%$$ 16

where J (mol m^–2 h^–1) is the ion permeation rate, $C_{M^{2+}}$ (mg L^–1) and $C_{{Li}^{+}}$ (mg L^–1) are the concentrations of multivalent ions (Ni²⁺, Co²⁺ and Mn²⁺) and Li⁺ in the feed chamber and recovery chamber, respectively.

Density functional theory calculations

Spin-polarized density functional theory (DFT) calculations are carried out by using the Vienna Ab initio Simulation Package (VASP, version 5.4.4)^51–53. The exchange-correlation energy was described using the Perdew-Burke-Ernzerhof (PBE) functional within the generalized gradient approximation (GGA), and the interactions between valence electrons and ionic cores were treated using the projector augmented-wave (PAW) method^54–56. The valence electrons were treated explicitly as follows: Co (3 d⁸4s¹), Ni (3d⁹4s¹), and Mn (3p⁶3d⁶4s¹). On-site Coulomb interactions for transition-metal 3d orbitals were corrected using the DFT + U approach within the Dudarev scheme⁵⁷, with U values of 5.9 eV for Ni, 3.52 eV for Co, and 2.98 eV for Mn, following established literature parameters^58,59. A vacuum spacing of 15 Å was introduced in all directions to avoid spurious interactions between periodic images. Solvent effects were included to approximate aqueous conditions. All atomic structures were fully relaxed until the residual forces were smaller than 0.01 eV Å^-1. A plane-wave basis set with a kinetic energy cutoff of 500 eV was employed. Brillouin-zone integrations were performed using a Γ-centered 1 × 1 × 1 k-point mesh throughout the calculations.

Author contributions

Y. Qiu and L. Xia performed experiments. X. Yan and Z. Liu conducted the DFT simulation. Y. Zhao and C. Tang provided resources for project implementation. Y. Zhao, C. Tang, and K. Shih supported experimental results. Y. Qiu wrote the manuscript. All the authors discussed the results, provided comments, and revised the manuscript.

Peer review

Peer review information

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

Funding

This work was supported by the RGC General Research Fund of the Hong Kong Special Administration Region (No. 17203924 and No. 17205325), the Innovation and Technology Support Program Seed of the Hong Kong Special Administration Region (NO. ITS/332/23), the National Natural Science Foundation of China (No. 22308288), and the Guangdong Basic and Applied Basic Research Foundation (No. 2025A1515010751). X. Yan and J.Z. Liu acknowledge grant support from the Australian Research Council (DP210103888). L. Xia would like to acknowledge the support of Fonds Wetenschappelijk Onderzoek-Vlaanderen (FWO) (No. 12B1E24N), and FWO travel grant for a long stay abroad (No. V417625N).

Data availability

The data generated in this study are provided in the article and its Supplementary Information. Visualization and figure preparation are carried out using Origin 2018, Microsoft PowerPoint 2021, and Microsoft Excel 2021. The optimized DFT structures generated in this study have been deposited in Zenodo and are available at 10.5281/zenodo.20571206. All data are available from the corresponding author upon request. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-75277-1.