Ultrafast Anion‐Hopping Conduction in Organic Solvent via Imidazolium‐Grafted Dynamic Ion‐Conducting Spacers for Stable Non‐Aqueous Flow Batteries

ABSTRACT

Non‐aqueous flow batteries (NAFBs) require membranes with superior organic solvent resistance and high ionic conductivity in organic solvents to endure the corrosive effects inherent in non‐aqueous electrolytes. Herein, we designed a highly organic‐solvent‐resistant imidazolium‐grafted anion‐exchange membrane (AEM) for NAFBs, realizing rapid anion hopping conduction in organic solvents. The membrane exhibited an ultrafast ionic conductivity as high as 2.1 mS cm⁻¹ in DMF‐based electrolyte which was much greater than that of the commercial Celgard membrane (only 0.45 mS·cm⁻¹), coupled with exceptional barrier properties evidenced by ultralow active‐species permeability values of 1.6 × 10⁻⁸ cm² s⁻¹ for 2,1,3‐benzothiadiazole anolyte and 3.8 × 10⁻⁹ cm² s⁻¹ for 10‐methylphenothiazine catholyte, respectively. Moreover, the membrane demonstrated outstanding stability in aggressive organic solvent, achieving an average energy efficiency (EE) over 66.1% for over 330 cycles at 5 mA cm⁻² in a NAFB cell test, which far exceeded the commercial Celgard membrane (with only 70 cycles with an average EE of 48.8%). This research presents a strategic advance in next‐generation membrane design for NAFBs with exceptional organic solvent resistance and high ionic conductivity.

We designed an imidazolium‐grafted crosslinked polybenzimidazole (PBI) anion‐exchange membrane (AEM) for NAFB, in which the imidazolium and flexible butyl side chains create an environment conducive to BF₄ ⁻ anion‐hopping conduction.

1. Introduction

Intermittent energy sources, such as solar and wind power, necessitate large‐scale energy storage technologies to balance supply and demand [1, 2, 3]. Flow battery, typically the vanadium flow battery (VFB), is a promising solution for grid‐scale storage [4, 5, 6, 7, 8]. However, the development of VFB is constrained by high material costs and relatively low energy density (typically 25–30 Wh·L⁻¹) [9, 10]. In contrast, non‐aqueous flow battery (NAFB) provides a wider electrochemical stability window (up to 4 V > 1.23 V in water) [11, 12, 13] and more material choices [14, 15, 16], making them highly attractive for high‐energy‐density applications [17].

Membranes, serving as a critical component in flow batteries, have garnered significant research interest due to their essential functions in mitigating active species crossover [18, 19, 20, 21] and facilitating rapid charge carrier transport [22, 23, 24, 25]. However, the development of membranes for NAFBs faces significant challenges, mainly because these membranes must meet multiple stringent criteria including excellent organic solvent resistance, high ionic conductivity in organic solvents and low crossover of active species [26, 27]. Furthermore, these stringent requirements are exacerbated by the intrinsic drawbacks of non‐aqueous electrolytes, specifically their elevated viscosity and suboptimal ionic conductivity [16, 28]. To date, commercial polyolefin‐based membranes such as Celgard [29, 30] and Fumasep [31, 32] are commonly employed in non‐aqueous systems. However, commercial membranes typically exhibit either insufficient ion selectivity or severe swelling in organic solvents. Therefore, the fabrication of membranes with excellent organic solvent resistance, minimal active‐species crossover, and high ionic conductivity is highly crucial for advancing NAFB technology.

In this work, we designed an imidazolium‐grafted crosslinked polybenzimidazole (PBI) anion‐exchange membrane (AEM) for NAFB (Figure 1). The crosslinking process endowed the membrane with robust organic solvent resistance, while the subsequent grafting of imidazolium side chains—functioning as dynamic ion‐conducting spacers—enabled exceptional ionic conductivity in DMF‐based electrolyte. The outstanding battery performance demonstrated well operational stability during prolonged operation in non‐aqueous electrolytes. Combined in‐situ characterization, ex‐situ characterization, and molecular dynamics (MD) simulations proved that the grafted imidazolium cations effectively facilitated anion hopping conduction, providing fundamental insights for NAFB membrane development by elucidating the ion transfer mechanism in non‐aqueous electrolyte solution.

FIGURE 1.

Schematic illustration of the design principle of the high organic‐solvent‐resistant imidazolium‐grafted crosslinked polybenzimidazole (PBI) membrane for non‐aqueous flow battery (NAFB).

2. Results and Discussion

2.1. Membrane Preparation and Properties

The original membranes fabricated via non‐solvent induced phase separation (NIPS) were subsequently crosslinked with potassium persulfate (K₂S₂O₈) (Figures 1 and S1–S3), which induced a “healing” effect [33] by tightening the interchain spacing of PBI and yielded membranes with exceptional organic solvent resistance (Figure S4) and superior crossover suppression (Figure S5). The resulting membranes with a hierarchical pore structure (Figure S6) are designated as K‐xPBI, where x corresponds to the volume ratio of ethanol (EtOH) to isopropanol and K denotes crosslinking mediated by K₂S₂O₈. The preparation of K‐xPBI was proven to be successful via Fourier transform infrared spectroscopy (FTIR, Figure S7) and X‐ray photoelectron spectroscopy (XPS, Figure S8). Low field nuclear magnetic resonance (LF‐NMR) revealed the abundant presence of micropores and mesopores within the channels (Figure S9). The optimized 5.2 Å micropores in K‐19PBI corresponded to its highest CO₂ adsorption capacity (Figure S10). Furthermore, the K‐19PBI membrane was selected as the optimal substrate for subsequent studies based on its ionic conductivity (Figure S11), mechanical properties (Figure S12), and battery performance (Figure S13).

To further enhance ionic conductivity in organic electrolyte solutions, different alkylimidazolium (Im) side chains (Figure 2a) including ethylimidazolium (EIm), butylimidazolium (BIm), and hexylimidazolium (HIm) were introduced into K‐19PBI using α,α′ ‐dibromo‐p‐xylene (DBX) as a grafting agent (Figures S14–S19). These imidazolium‐grafted membranes (K‐19PBI‐Im, Figures 2b and S20) maintained the original hierarchical pore structure well (Figures 2c and S21) and exceptional organic solvent resistance (Figure S22). As expected, alkylimidazolium functional groups effectively enhanced uptake of various electrolyte components (Figure 2d), resulting in more rapid ionic transport (Figure 2e). Among K‐19PBI‐Im membranes, K‐19PBI‐Bim membrane demonstrated the most balanced overall performance, as evidenced by its highest ionic conductivity (Figure S23), excellent mechanical properties (Figure S24), low swelling degree (Figure S25), and superior battery efficiency (Figure S26). Therefore, we further studied K‐19PBI‐yBIm membranes with different grafting degrees of BIm in detail, where y denotes the molar concentration of 1‐butylimidazole in 1‐methyl‐2‐pyrrolidinone (NMP) solution.

FIGURE 2.

Characterizations of different alkylimidazolium‐grafted crosslinked PBI membranes (K‐19PBI‐Im). (a) Schematic illustration of ion transport in K‐19PBI‐Im membranes, which were grafted with different imidazolium side chains, including ethylimidazolium (EIm), butylimidazolium (BIm), and hexylimidazolium (HIm) as dynamic ion‐conducting spacers. (b) High‐resolution N 1s X‐ray photoelectron spectroscopy (XPS) spectra for K‐19PBI‐Im membranes. (c) ¹H time‐domain LF‐NMR spectra of K‐19PBI and K‐19PBI‐Im membranes. (d) DMF‐based electrolyte uptake ratios of K‐19PBI and K‐19PBI‐Im membranes. Data are presented as mean ± standard deviation (SD) with n = 3 independent samples. (e) BF₄ ⁻ anion permeability through K‐19PBI and K‐19PBI‐Im membranes in DMF‐based electrolyte. Data are presented as mean ± standard deviation (SD) with n = 3 independent samples.

2.2. Ion Transport Behavior and Battery Performance

Along with the preservation of the hierarchical pore structure (Figures 3a and S27), the K‐19PBI‐yBIm membranes underwent a systematic improvement in key properties with increasing BIm grafting degree, including an increased ion exchange capacity (IEC) from 0.49 to 1.23 meq·g⁻¹ (Figure S28), enhanced uptake of various electrolyte components (Figures 3b and S29), an elevated charge carrier transport rate (Figure 3c), and superior mechanical properties (Figure S30). The ionic conductivity results (Table S1) indicated that the K‐19PBI‐0.15BIm membrane achieved the highest ionic conductivity value of 2.14 mS·cm⁻¹ in DMF, which is lower than the bulk electrolyte conductivity (28.9 mS·cm⁻¹, Table S2) and substantially greater than that of K‐19PBI membrane (0.098 mS·cm⁻¹) and the commercial Celgard 2500 membrane (0.45 mS·cm⁻¹). The much lower measured conductivity of Celgard 2500 than its theoretical value is attributed to pore tortuosity and collapse. In addition to high ionic conductivity in DMF‐based electrolyte, these membranes also exhibited extremely low permeability to active‐species (Figure 3d, e). Notably, the K‐19PBI‐0.1BIm membrane enabled significantly superior crossover suppression compared to the commercial Celgard 2500 membrane, with ultralow permeability values of 3.8 × 10⁻⁹ cm² s⁻¹ for 10‐methylphenothiazine (MPT) and 1.6 × 10⁻⁸ cm² s⁻¹ for 2,1,3‐benzothiadiazole (BTD), much lower than typical porous polyolefin membranes (Tables S2 and S3). As summarized in Figure 3f and Table S3, our K‐19PBI‐BIm membrane surpassed most reported membranes for non‐aqueous electrolytes, exhibiting exceptional ionic conductivity and active‐species permeation inhibition ability, resulting in an outstanding selectivity of 1×10⁵ S s cm⁻³ (defined as the ratio of ionic conductivity to active‐species permeability).

FIGURE 3.

Characterizations of K‐19PBI‐yBIm membranes. (a) Cross‐sectional scanning electron microscopy (SEM) image of the K‐19PBI‐0.1BIm membrane, the inset is a sample photo. (b) Electrolyte solution uptake ratios of K‐19PBI‐yBIm membranes. Data are presented as mean ± standard deviation (SD) with n = 3 independent samples. (c) BF₄ ⁻ anion permeability through K‐19PBI‐yBIm membranes in DMF‐based electrolyte. Data are presented as mean ± standard deviation (SD) with n = 3 independent samples. (d) MPT and (e) BTD permeabilities through K‐19PBI‐yBIm membranes and the commercial Celgard 2500 membrane. (f) Comparison of ionic conductivity and active‐species permeability of the reported NAFB membranes evaluated in similar electrolytes. The dashed lines represent the membrane selectivity (the ratio of ionic conductivity to active‐species permeability).

The membrane battery performance was further assessed in a NAFB flow cell under nitrogen atmosphere, utilizing MPT and BTD as the catholyte and anolyte, respectively, and tetraethylammonium tetrafluoroborate (TEABF₄) in DMF as the electrolyte (Figure 4a). Compared with the commercial Celgard and ungrafted membranes, the K‐19PBI‐yBIm series exhibited excellent NAFB performance, as both voltage efficiency (VE) and energy efficiency (EE) increased with grafting degree (Figure 4b, c). The relatively low CE and VE are mainly ascribed to interfacial polarization caused by irreversible byproducts generated from side reactions of redox‐active materials [34, 35, 36, 37], rather than intrinsic membrane resistance, as supported by post‐cycling FTIR (Figure S31), cycled membrane's SEM (Figure S32), and ionic conductivity recovery of the cycled K‐19PBI‐0.1BIm membrane after washing (Table S1). However, the higher grafting degree was accompanied by significant membrane swelling due to excessive electrolyte uptake (Figure S33), which led to a reduction in Coulombic efficiency (CE) (Figure 4d). This observed decline in CE is of critical concern, as maintaining high CE is essential for long‐term cycling stability and represents a key trade‐off in membrane design. Despite this, the swelling ratio of the K‑19PBI‑yBIm membranes remains within a moderate and controllable range (< 20%), which benefits practical stack compression and scalability. Moderate swelling ensures tight interfacial contact and reliable sealing under stack compression, thus avoiding electrolyte leakage and extra ohmic loss [38, 39]. Well‑regulated swelling also enables defect‑free roll‑to‑roll manufacturing and consistent large‑area membrane production [22, 40]. Therefore, even though excessive grafting brings a slight CE penalty, the overall swelling behavior stays within an acceptable window and offers prominent engineering advantages for real‑world flow battery applications. On the basis of its superior battery performance and correct swelling, the K‐19PBI‐0.1BIm membrane was further selected for evaluation in long‐term cycling tests. As shown in Figure 4e, the K‐19PBI‐0.1BIm membrane exhibited outstanding long‐term stability, sustaining an average EE of ∼66.1% over 330 stable cycles at 5 mA cm⁻². Meanwhile, the absolute discharge capacity remained above 200 mAh/L without rapid decay (Figure S34), far superior to that of the commercial Celgard 2500 membrane, which failed after only 70 cycles with a much lower average EE of ∼48.8% and rapid capacity fading (Figures S34 and S35). For further validation, commercial AEMs (AMVN, DSVN, FAP‐450) were evaluated under identical conditions; DSVN and FAP‐450 suffered from severe dissolution or structural collapse in DMF, while AMVN exhibited rapid efficiency decay and poor efficiency. (Figure S34, S36, and S37). These results demonstrated that grafting BIm side chains effectively creates a favorable environment for ion transport, thereby enhancing ionic conductivity in DMF‐based electrolyte. Compared with the reported membranes evaluated in NAFBs applying phenothiazine derivatives or BTD as active‐species, our designed membrane demonstrates superior performance in terms of both cycling stability and EE (Figure 4f and Table S4). Comprehensive post‐cycling characterizations reveal that the K‐19PBI‐0.1BIm membrane retains intact morphology, stable mechanical properties, and well‐preserved imidazolium groups after 330 long‐term cycles (Figures S38–S40).

FIGURE 4.

Battery performance of membranes in non‐aqueous flow battery (NAFB). Data are presented as mean ± standard deviation (SD), and error bars represent the standard deviation calculated from the first 50 cycles (n = 50). (a) Schematic configuration of NAFB employing BTD (anolyte) and MPT (catholyte) as redox couples. (b) Coulombic efficiency (CE), (c) voltage efficiency (VE), and (d) energy efficiency (EE) of Celgard 2500, K‐19PBI, and K‐19PBI‐yBIm membranes at 5 mA cm⁻². (e) Long‐term cycling stability of NAFB applying K‐19PBI‐0.1BIm membrane. (f) Comparison of EE versus cycle number of different membranes in NAFB applying phenothiazine derivatives or BTD as active‐species.

2.3. Mechanism of Imidazolium

We systematically investigated the role of imidazolium side chains in ionic conduction. Quantitative analysis of the relative contributions of cations and anions to ion conduction was achieved through determination of tetraethylammonium cation (TEA⁺) and BF₄ ⁻ transference numbers (Figure 5a), revealing a remarkably high anion transference number (t₋ = 0.9275–0.9942) for K‐19PBI‐BIm membranes. These results unequivocally demonstrated the predominance of BF₄ ⁻ transport in the conduction mechanism. We thought that this phenomenon arises from some strong attractions between BIm and BF₄ ⁻ anions, which effectively facilitate anion transport.

FIGURE 5.

Mechanistic insights of imidazolium. (a) TEA⁺ and BF₄ ⁻ transport numbers through K‐19PBI and K‐19PBI‐BIm membranes. (b) Broad‐frequency dielectric spectra of K‐19PBI‐yBIm membranes. (c) In‐situ FTIR spectra of K‐19PBI‐0.1BIm in TEABF₄/DMF solution (d) Interaction energies (Lennard‐Jones and Coulomb) between BF₄ ⁻ anions and grafted membranes derived from molecular dynamics (MD) simulations. (e) Representative snapshot from MD simulations: BF₄ ⁻ (green), TEA⁺ (red), DMF solvent (semi‑transparent), K‑PBI (purple), and K‑PBI‑BIm (yellow). (f) Diffusion coefficients and mean square displacement (MSD) profiles of membranes.

Broadband dielectric spectroscopy (BDS) was carried out to investigate the attraction between imidazolium and BF₄ ⁻ anions [41, 42]. Variations in the dielectric constant (ε′) directly reflect the attraction of BF₄ ⁻ anions to imidazolium, as well as the resulting enrichment of these anions [43]. As shown in Figure 5b, in the low‐frequency region (10⁻²–10³ Hz), the slowly alternating electric field enables BF₄ ^– anions to have ample time to migrate and accumulate at the electrode interface. With increasing BIm content, stronger anion attraction is achieved, thus leading to a more pronounced ε′ enhancement (Figure S41a). In contrast, the electric field alternation rate in the high‐frequency region (10³–10⁶ Hz) is far faster than the anion migration rate, such that anions cannot migrate directionally in real time, thus inducing a decrease in ε′. The K‐19PBI‐0.15BIm membrane with the highest imidazolium content enriches the largest quantity of BF₄ ⁻ anions, exhibiting the highest decline rate (Figure S41b). These pronounced dielectric response results clearly revealed that the strong attraction between imidazolium and BF₄ ⁻ anions serves as the core mechanism that modulates BF₄ ⁻ anion behaviors within the membrane.

We further employed in‐situ Fourier transform infrared spectroscopy (in situ FTIR) to investigate the attraction in detail. Specifically, as shown in Figure 5c, over time, the original anions attracted by the imidazolium within the membrane are gradually replaced by BF₄ ⁻ anions, accompanied by distinct spectral changes relative to the initial baseline spectrum (Figure S42). The B–F stretching vibrations [44, 45] (1049 cm⁻¹ and 787 cm⁻¹) show progressively increasing intensities, while the characteristic vibrations of imidazolium rings [46] (1655 to 1660 cm⁻¹, Figure S43) exhibit blueshifts along with intensity variations. These spectral changes are attributed to interactions between imidazolium and BF₄ ⁻ anions, which drive the redistribution of electron density within the imidazolium ring. Finally, spectral stabilization after ∼120 min (Figure S44) indicates that the attraction‐induced electron redistribution has reached an equilibrium state.

To further elucidate the underlying interaction, molecular dynamics (MD) simulations were performed to examine the interaction energies and diffusion coefficients after grafting imidazolium. The interaction energies between imidazolium and BF₄ ⁻ anions for the grafted membrane showed strongly negative values with pronounced temporal fluctuations (Figure 5d), revealing a Coulomb‐dominated attraction [47] that serves as the fundamental force for BF₄ ⁻ hopping between imidazolium side chains. The inherent flexibility of the butyl side chains promotes this ion transport process by enabling the dynamic switching of BF₄ ⁻ anions between imidazolium side chains. Time‐evolved MD snapshots (Figures 5e and S45) capture the progressive accumulation and hopping transfer of BF₄ ⁻ anions near the side chains throughout the simulation period, offering direct visual evidence for this dynamic process. The combined contribution of Coulomb attraction and butyl side chain promotion is directly reflected in the mean squared displacement (MSD) profiles (Figure 5f), which demonstrate an increase in diffusion coefficients from (8.37 ± 0.4) × 10⁻⁷ to (10.69 ± 0.96) × 10⁻⁷ cm² s⁻¹. This combined action as dynamic ion‐conducting spacers expedites BF₄ ⁻ anion hopping transfer between imidazolium side chains in DMF‐based electrolyte (Figure S46), which ultimately underpins the superior performance of the membranes in NAFBs.

3. Conclusion

In summary, this work demonstrates an integrated fabrication strategy to construct high organic‐solvent‐resistant AEMs for NAFB, achieved by synergistically combining NIPS, chemical cross‐linking, and imidazolium grafting. Control over the channel structure and surface properties enabled the formation of an optimized ion transport network. The resulting membranes exhibited an exceptional combination of properties, including remarkably low permeability to active species, a high ionic conductivity of up to 2.14 mS·cm⁻¹ in DMF‐based electrolyte, and outstanding organic solvent resistance. In a DMF‐based electrolyte NAFB, the designed membranes significantly outperformed the commercial Celgard 2500, delivering ∼35% higher average EE and ∼13% higher average CE while maintaining stable operation for 330 cycles. We further elucidated that the performance enhancement originates from an anion‐hopping conduction mechanism, where imidazolium and flexible butyl side chains create an environment conducive to BF₄ ⁻ anion‐hopping conduction. Collectively, these findings establish our membrane design as a highly promising candidate for next‐generation NAFBs, offering a comprehensive solution for advanced energy storage applications.

Author Contributions

Pengzhu Gai: conceptualization, data curation, investigation, methodology, writing ‐ original draft. Hangyuan Wang: investigation, methodology, software. Qiang Chen: formal analysis, investigation, software. Hengqi Shi: methodology, validation, investigation. Hongyan Cao: methodology, data curation, validation, supervision. Yixing Wang: formal analysis, methodology, supervision. Ying Yu: visualization, supervision, resources. Huidong Qian: supervision, visualization. Kang Huang: conceptualization, resources, project administration, supervision, writing — review and editing. Zhi Xu: conceptualization, writing — review and editing, resources, project administration, supervision.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File: anie72618‐sup‐0001‐SuppMat.docx.

Gai P., Wang H., and Chen Q., et al., Angewandte Chemie International Edition. 65, no. 34 (2026): e6160285. 10.1002/anie.6160285

Data Availability Statement

The data that supports the findings of this study are available in the supplementary material of this article.