Artificial α-amino acid based on cysteine grafted natural aloe-emodin for aqueous organic redox flow batteries

Abstract

Natural redox-active anthraquinone derivatives possess promising attributes for applications in aqueous organic redox flow batteries (AORFBs) due to their environmental friendliness and abundant sources. However, their limited aqueous solubility and electrochemical stability have posed significant challenges to their practical utilization. Herein, inspired by click chemistry, we report the synthesis of an artificial α-amino acid derived from cysteine-functionalized natural aloe-emodin (namely Cys-AE), which exhibits good water-solubility and redox-reversibility, particularly suited for alkaline AORFBs. The bio-inspired Cys-AE molecule exhibits a threefold increase in aqueous solubility compared to pristine aloe-emodin. Furthermore, the AORFB based Cys-AE negolyte with an electron concentration of 1.0 M demonstrates a low capacity fade rate of 0.000948% cycle⁻¹ (equivalent to 0.0438% day⁻¹) during 592 cycles, significantly outperforming the AORFB based on pristine aloe-emodin (0.00446% cycle⁻¹, or 0.908% day⁻¹) during 1564 cycles. Our investigation incorporates time-dependent density functional theory (TDDFT) simulations and detailed spectroscopic analyses reveal the essential role played by the asymmetric distribution of multiple solubilizing groups in enhancing the aqueous solubility and cycling stability of Cys-AE. This study highlights the potential of nature-inspired molecular engineering strategies in creating and crafting redox-reversible organic species poised to revolutionize large-scale and sustainable energy storage applications.

Aqueous organic redox flow batteries face challenges due to the low solubility and stability of natural anthraquinones. Here, authors design a cysteine-functionalized aloe-emodin derivative with enhanced solubility and stability, achieving long-cycle life and high efficiency in alkaline flow batteries.

Introduction

The development of large-scale energy storage systems is of utmost importance to regulate the electricity flow between renewable energy sources and smart grid¹. Redox flow batteries (RFBs), which store significant energy in the electrolyte, offer favourable scalability, variable modular design, and reliable safety, making them a promising solution for large-scale energy storage^2–5. Extensive research has been conducted on aqueous RFBs utilizing transition metal redox species, particularly all-vanadium RFBs^6–8. However, these batteries still encounter long-standing challenges, including high costs and electrolyte crossover issues. It is crucial to explore alternative redox-active species with high solubility, exceptional stability, and low permeability. Redox-active organic compounds have gained significant attention due to their excellent structural adjustability and tunable electrochemical properties^9–15. Generally, anthraquinone compounds possess relatively low potential, high solubility, and good electrochemical stability in alkaline solutions^16,17. In particular, emodin and aloe-emodin, are important natural anthraquinones derivatives commonly found in plants and filamentous fungi¹⁸. However, the limited aqueous solubility of these anthraquinone natural products hinders their utilization in high-energy-density aqueous organic redox flow batteries (AORFBs).

In this study, we present the design and synthesis of a water-solubilizing artificial α-amino acid based on cysteine grafted natural aloe-emodin, namely Cys-AE, which can serve as a highly soluble negolyte material with relatively low potential and excellent redox reversibility in AORFBs. The synthesis pathway for Cys-AE is inspired by the principles of click chemistry, which mimics the natural biosynthesis of complex organic compounds. Under mild reaction conditions, biomimetic Cys-AE is efficiently produced through the formation of carbon-heteroatom bonds (C–S–C), without the need for challenging separation steps. The introduction of a cysteine side-group in Cys-AE brings the zwitterionic α-amino acid structure with hydrophilic amino and carboxyl groups. Similar to natural amino acids, this structure significantly enhances the aqueous solubility due to its amphoteric dissociation equilibrium.

Additionally, the asymmetric distribution of these solubilizing groups on anthraquinone core greatly elevates the solubility limit by competing interactions with solvents and counter ions¹⁹. Compared to the low aqueous solubility of natural aloe-emodin (0.26 M), the hydrophilic modification with cysteine drastically increases the aqueous solubility of Cys-AE by three times. Furthermore, the permeability value of Cys-AE through the Nafion-212 membrane was found to be lower than 1 × 10⁻¹³ cm² s⁻¹, which effectively avoids the crossover contamination. The amino acid functionalization also improves the charge transfer kinetics and long-term stability of AORFBs. The structure of Cys-AE remains highly stable after long-term cycling, with only very few side-chain losses, and these minimal losses do not affect its solubility and redox activity due to the asymmetric distribution of multiple solubilizing groups on the anthraquinone core. As a result, the Cys-AE|| K4Fe(CN)6 AORFBs demonstrate an excellent Coulombic efficiency of nearly 100%, a high-energy efficiency of 75%, and an ultralow capacity fade rate of 0.000948% cycle⁻¹ (equivalent to 0.0438% day⁻¹) over 500 cycles at high electrolyte concentrations. This study highlights the importance of nature-inspired molecular design and the asymmetric distribution of water-solubilizing groups on the redox core for achieving high-performance and high-stability AORFBs.

Results

The Cys-AE sample was synthesized using a two-step process (Fig. 1a). First, aloe-emodin underwent a reaction with phosphorus tribromide (PBr₃) to produce 3-(bromomethyl)-1,8-dihydroxyanthracene-9,10-dione, an intermediate product²⁰. Subsequently, this intermediate was reacted with cysteine in a 1.0 M NaOH solution and followed by acidification, yielding Cys-AE as the final product. The molecular structure and chemical purity of Cys-AE were confirmed using ¹H nuclear magnetic resonance (NMR), ¹³C NMR and high-resolution mass spectrometry (HRMS), as shown in Supplementary Fig. 1−3. The solubility of aloe-emodin in 1.0 M KOH solution was determined to be 0.26 M based on ultraviolet-visible light (UV-Vis) absorption curves at different concentrations (Supplementary Fig. 4). The introduction of the amino acid group in Cys-AE induces an asymmetric charge distribution and higher polarizability, resulting in enhanced intermolecular interactions and increased solubility. Under strong alkaline conditions, the multiple ionic functional groups, i.e., deprotonated hydroxyl and carboxyl groups, in Cys-AE contribute to its higher solubility compared to neutral or weakly alkaline conditions¹⁹. In consequence, compared to aloe-emodin, the solubility of Cys-AE in 1.0 M KOH increased to 0.78 M, which is a roughly threefold increase (Supplementary Fig. 4). The intermolecular interactions between Cys-AE molecules were characterized using ¹H NMR analysis. Supplementary Fig. 5 displays the ¹H NMR spectra of Cys-AE at concentrations of 0.1 M and 0.5 M in 1.0 M KOH solution. The chemical shift for each proton site in the highly soluble Cys-AE exhibits varying degrees of high field shifting (towards lower ppm), indicating the solvation effect and reduced role of magnetic susceptibility²¹.

Fig. 1. Synthesis processes of Cys-AE and schematical illustration of Cys-AE |  | K₄Fe(CN)₆ AORFBs.

a Synthesis route of Cys-AE. b CV curves of aloe-emodin, Cys-AE and K₄Fe(CN)₆ (with 1 mM concentration) measured by a 3 mm glassy carbon electrode in 1.0 M KOH solution at a scanning rate of 100 mV s⁻¹. c Schematic configuration of the Cys-AE |  | K₄Fe(CN)₆ AORFBs based on Cys-AE negolyte, K₄Fe(CN)₆ posolyte and Nafion-212 membrane separator. The voltage is not iR corrected. Source data are provided as a Source Data file.

The electrochemical properties of aloe-emodin and Cys-AE were analyzed through cyclic voltammetry (CV) in a 1.0 M KOH aqueous solution (Fig. 1b). Cys-AE exhibited a redox potential of −0.51 V vs. the standard hydrogen electrode (SHE), with a peak separation (ΔE) of 69 mV for 2 electrons. The reversible potential was 60 mV more positive than that of aloe-emodin, which could be attributed to the weakened electron-donating ability of the cysteine-modified side chain. When pairing the Cys-AE negolyte with a potassium ferrocyanide posolyte, an equilibrium cell voltage of ~1.01 V is expected. The structural configuration of the Cys-AE|| K₄Fe(CN)₆ AORFBs sandwiched with a Nafion-212 membrane is illustrated in Fig. 1c. The chemical stability of aloe-emodin and Cys-AE in 1.0 M KOH solution was evaluated using both ¹H NMR and CV. After 14 days of storage, the ¹H NMR spectra of both aloe-emodin and Cys-AE exhibited no noticeable changes (Supplementary Figs. 6 and 7). Similarly, the CV curves of both molecules remained stable over the same period (Supplementary Fig. 8).

Density functional theory (DFT) simulations were performed to study the optimized structures and electrostatic potentials (ESP) of aloe-emodin and Cys-AE. As depicted in Fig. 2a, b, the ESP of the cysteine unit in Cys-AE is more negative than that of the −CH₂OH unit in aloe-emodin. The deprotonated carboxyl group in Cys-AE induces asymmetric charge distribution and higher polarizability¹⁹, leading to enhanced intermolecular interactions and improved solubility of Cys-AE.

Fig. 2. The basic physical and chemical properties of aloe-emodin and Cys-AE.

a, b Optimized structures and electrostatic potentials of a aloe-emodin and b Cys-AE. The electrostatic potentials were calculated at the 6-311 G(d,p) level. c, d UV-Vis absorption spectra of 0.10 mM c aloe-emodin and d Cys-AE at different pH values. e, f Simulated UV-Vis absorption spectra of e aloe-emodin and f Cys-AE calculated by TDDFT method at the PBE0/6-311 G(d,p) level. g Calculated molecular orbitals and energy gaps of aloe-emodin and Cys-AE at oxidized (initial) state and reduced state under the condition of pH 14.

The UV-Vis absorption spectra of aloe-emodin showed two main peaks and a weak shoulder peak ranging from 200 to 700 nm (Fig. 2c). As the pH increased, the absorbance peak of aloe-emodin exhibited a red shift from 430 nm (pH 7) to 504 nm (pH 14), accompanied by a color change from red to yellow. Similarly, the UV-Vis absorption spectra of Cys-AE showed two main peaks and a distinct shoulder peak between 200 and 700 nm (Fig. 2d). With the increase of pH, the absorbance peak of Cys-AE exhibited a red shift from 433 nm (pH 7) to 504 nm (pH 14). Additionally, for simulating the UV-visible absorption spectra of these two molecules, time-dependent density functional theory (TDDFT) calculations were performed to calculate the contributions of various excited states^22,23. The simulated spectra were well matched with the actually measured UV-Vis absorption spectra, as depicted in Fig. 2e, f. Comparatively, the UV-Vis absorption spectrum of Cys-AE exhibited a slight red shift in comparison to that of aloe-emodin, and a distinct shoulder peak at 307.7 nm was observed in the spectrum of Cys-AE, which was not evident in that of aloe-emodin. For aloe-emodin, the absorption peak observed at 432.2 nm is influenced by the S0→S2 excitation, while the peaks below 350 nm are influenced by S0→S6, S15, and S16 excitations. For Cys-AE, the absorption peak observed at 435.2 nm is influenced by S0→S2 and S0→S3 excitations, and the peaks below 350 nm are influenced by S0→S9 and S20 excitations. Molecular conformation plays a critical role in influencing electronic transitions, as changes in conformation can alter the energy distribution of molecular orbitals and the electron structure of molecules. During light absorption, electrons are converted from ground states to excited states, and the energy required for these transitions is determined by the energy gap between the molecular orbitals. Different energy gaps correspond to different absorption characteristics that emerged in the UV-Vis spectra. Specifically, the spectral differences between Cys-AE and aloe-emodin can likely be attributed to the introduction of –NH₂ and –COOH groups in Cys-AE. These functional groups modify the molecular orbital energy levels, thus altering the energy required for electron transitions from ground state to excited states, resulting in a shift in absorption peaks and the appearance of new spectroscopic features. We have summarized the contributions of electronic transitions between different orbitals to the wavelength of aloe-emodin and Cys-AE in Supplementary Tables 1 and 2.

The dissociation behaviors of the ionic functional groups of aloe-emodin and Cys-AE were comparatively investigated²⁴. The dissociation behavior of the hydroxyl groups in aloe-emodin can be reflected by the UV-Vis absorption spectra at different pH values (Fig. 2c and Supplementary Fig. 9a). The absorbance versus pH plots of aloe-emodin at 504 nm revealed only one pKa value (8.2) for aloe-emodin, corresponding to the dissociation of one phenol hydroxyl groups on anthraquinone, indicating that aloe-emodin only loses one proton in slightly alkaline aqueous solutions. The pKa values for the two hydroxyl groups on the shoulder sites of aloe-emodin are indistinguishable. The proximity of the two hydroxyl groups to carbonyl groups makes it easy for intramolecular hydrogen bonds to form, hindering the complete dissociation of both hydroxyl groups. This low charge density distribution on the molecule results in relatively low solubility of aloe-emodin in alkaline solutions (0.26 M). Different from aloe-emodin, Cys-AE, which contains a basic amino group (−NH₂) and an acidic carboxyl group (−COOH), exhibits typical zwitterionic behavior and undergoes multiple dissociation equilibria in aqueous solutions. We propose that only one phenol hydroxyl group in Cys-AE undergoes dissociation, which is comparable to aloe-emodin. Previous literature confirmed that cysteine, as a natural amino acid, has two pKa values of 1.92 (corresponding to the dissociation of −COOH) and 8.18 (corresponding to the dissociation of −NH₂)²⁵. Similarly, Cys-AE also exhibits multiple dissociation equilibria analogous to cysteine. The absorbance versus pH plot of Cys-AE at 502 nm revealed two pKa values (Fig. 2d and Supplementary Fig. 9b), indicating the dissociative behavior of hydroxyl, carboxyl, and amino groups at different pH levels. Based on the reported pKa values of cysteine, the pKa₁ value of Cys-AE was determined to be 4.5, corresponding to the dissociation of carboxyl group, while the pKa₂ value of Cys-AE was calculated to be 8.6, corresponding to the dissociation of both amino and hydroxyl groups. The two pKa values derived from the amino and hydroxyl groups in Cys-AE are too close to be distinguished. The acid-base titration curves reveal that the pKa value of the hydroxyl group in aloe-emodin is 8.4, while the pKa values of the hydroxyl and amino groups in Cys-AE are nearly indistinguishable, ~8.8, as shown in Supplementary Fig. 10. These values closely align with the pKa values measured by UV-Vis spectroscopy.

The highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) of aloe-emodin and Cys-AE at oxidized and reduced states are compared (Fig. 2g)^26,27. The atomic coordinates of the optimized computational models are provided in Supplementary Table 3. The LUMO energy level of Cys-AE (−2.33 eV) is lower than that of aloe-emodin (−2.21 eV), suggesting a higher reduction potential for Cys-AE. Similarly, the HOMO energy level of reduced Cys-AE (re-Cys-AE, −2.37 eV) is lower than that of reduced aloe-emodin (re-aloe-emodin) (−2.21 eV), indicating a higher oxidation potential for Cys-AE. This disparity may be attributed to the introduction of a thioether bond (C–S–C) in Cys-AE. Compared to an oxygen atom, sulfur has a larger atomic radius and higher polarizability, leading to weaker overlap between the lone pair electrons and the π-electron system of the sulfur atom. This reduced conjugation increases electron localization, stabilizing the LUMO and lowering both the HOMO and LUMO energy levels. Consequently, the introduction of sulfur atoms via the thioether bond significantly changes the electronic structure, making the molecule more capable of accepting electrons and thereby increasing the reduction potential. In contrast, the hydroxyl group in aloe-emodin facilitates electron donation via the conjugation or inductive effects, where the lone pair of electrons from the oxygen atom can interact with the π-electron system, increasing the electron density in the HOMO and raising the HOMO energy level. These DFT calculation results corroborate the CV curves of aloe-emodin and Cys-AE (Fig. 1b). The LUMO energy level difference between aloe-emodin and re-aloe-emodin is 2.30 eV, which is similar to that between Cys-AE and re-Cys-AE (2.27 eV), indicating the presence of a comparable energy barrier for the reduction step.

To further investigate the proton and electron transfer behaviors during the redox processes, CV analyses of aloe-emodin and Cys-AE at different pH values were performed to calculate the Pourbaix diagrams, as shown in Supplementary Fig. 11. The Pourbaix diagram of aloe-emodin can be modeled with two distinct segments of curves, showing slopes of 0 mV pH⁻¹ (for pH 8.5–10.0) and −55.9 mV pH⁻¹ (for pH higher than 10.0), corresponding to zero proton/two-electron and two proton/two-electron processes, respectively. Similarly, the Pourbaix diagram of Cys-AE also exhibits two segments of curves with slopes of 0 mV pH⁻¹ (for pH 8.1–10.0) and −56.5 mV pH⁻¹ (for pH higher than 10.0). When the pH value was lower than 8, both aloe-emodin and Cys-AE did not exhibit a reversible CV signal in the buffer solution. The redox potential of both molecules remains constant within the pH range of 8.0–10.0, likely attributed to the rearrangement and transfer of protons through both intra- and intermolecular hydrogen bonds²⁸, which is conducive to facilitating the redox shuttle.

To compare the electrochemical kinetics of aloe-emodin and Cys-AE, we conducted the analyses of diffusion coefficient (D) and kinetic reaction rate constant (k₀) using Nicholson’s method²⁹. The solutions containing 2.0 mM of aloe-emodin or Cys-AE in a 1.0 M KOH solution were utilized to record the CV curves at varying scan rates (10 ~ 5000 mV s⁻¹). The resulting current densities were plotted against the square roots of the scan rates, which revealed a linear relationship (Supplementary Fig. 12). Through the Randles-Sevcik equation, the diffusion coefficient (D) of aloe-emodin and Cys-AE was calculated to be 2.34 × 10⁻⁶ cm² s⁻¹ and 2.78 × 10⁻⁶ cm² s⁻¹ respectively, consistent with most anthraquinone derivatives. Correspondingly, the electron transfer rate constant (k₀) for aloe-emodin and Cys-AE was determined to be 1.53 × 10⁻²cm s⁻¹ and 1.85 × 10⁻²cm s⁻¹ respectively (Supplementary Fig. 13). The diffusion coefficient (D) and electron transfer rate constant (k₀) of Cys-AE are slightly higher than those of aloe-emodin, suggesting that their electrochemical kinetics are closely comparable. This similarity is reflected in their CV curves. D and k₀ jointly determine the molecular transport rapidness and the reaction rate within the system, where higher values of D and k₀ contribute to faster reaction kinetics. The D and k₀ values for Cys-AE and aloe-emodin fall within the same order of magnitude as those of most other anthraquinone derivatives^16,17. Consequently, like typical anthraquinone molecules, these two molecules can maintain rapid reaction rates even under high current densities.

The permeability of aloe-emodin and Cys-AE through a Nafion-212 membrane was compared in a two-compartment H-type cell (Supplementary Fig. 14). Based on UV-Vis absorbance analysis, the permeability value of aloe-emodin is determined to be 1.34 × 10⁻¹¹cm²s⁻¹. However, even after being stored for a long-term period of 12 days, the UV-Vis absorbance spectra measured on the receiving side of H-type cell for Cys-AE still did not exhibit the characteristic peaks of Cys-AE. Based on the previously reported derivation of Fick’s Law³⁰ the permeability value of Cys-AE through the Nafion-212 membrane was found to be lower than 1 × 10⁻¹³ cm² s⁻¹, which is more than two orders of magnitude lower than that of aloe-emodin. In comparison to aloe-emodin, Cys-AE exhibits a larger size and a higher degree of electronegativity, resulting in a more pronounced electrostatic repulsion from the Nafion membrane³¹. This phenomenon is advantageous for reducing the extent of crossover. Further CV tests revealed that crossover between the posolyte and negolyte does not significantly affect the CV curves of these two molecules, as shown in Supplementary Fig. 15. However, the stronger permeation of aloe-emodin can result in a decrease in the number of available redox-active molecules within the negolyte, ultimately leading to capacity degradation. The exceptionally low permeability of Cys-AE assists in preventing cross-contamination of the electrolyte, thereby ensuring stability throughout the extended cycling process of the AORFBs.

The electrochemical performances of aloe-emodin and Cys-AE-based AORFBs at low concentrations were systematically compared in this study (Fig. 3). To ensure that the negolyte serves as the capacity-limiting side, the molar ratio of negolyte-to-posolyte was set at 1:6 for providing an excess amount of ferrocyanide. Electrochemical impedance spectroscopy (EIS) analysis revealed that the resistances of the Nafion-212 were measured to be 0.63 Ω cm² (Supplementary Fig. 16). The aloe-emodin|| K₄Fe(CN)₆ AORFB exhibited an open-circuit voltage (OCV) of 1.08 V at a current density of 100 mA cm⁻², while the Cys-AE|| K₄Fe(CN)₆ AORFB demonstrated an OCV of 1.02 V, both consistent with the predicted values from CV analysis. The aloe-emodin|| K₄Fe(CN)₆ AORFB displayed discharge capacities of 5.164 and 4.473 Ah L⁻¹ at 20 and 100 mA cm⁻², corresponding to capacity utilization ratios of 96.34% and 83.45%, respectively. The Cys-AE|| K₄Fe(CN)₆ AORFBs exhibited discharge capacities of 5.211 and 4.807 Ah L⁻¹ at 20 and 100 mA cm⁻², respectively, accompanied by capacity utilization ratios of 97.22% and 89.68%. These measured capacities affirm that both aloe-emodin and Cys-AE underwent a relatively thorough two-electron reaction process. At 100% state of charge (SOC), the galvanic peak power density for aloe-emodin|| K₄Fe(CN)₆ and Cys-AE|| K₄Fe(CN)₆ AORFBs was measured to be 110 and 118 mW cm⁻², respectively (Supplementary Fig. 17 and Fig. 3d), which are comparable to previously reported anthraquinone derivatives^9,16,17. The rate performances of these two AORFBs are similar, as depicted in Supplementary Fig. 17 and Fig. 3e.

Fig. 3. Performance comparisons of the aloe-emodin|| K₄Fe(CN)₆ and Cys-AE|| K₄Fe(CN)₆ AORFBs at low concentrations.

a Molecular structures of aloe-emodin and K₄Fe(CN)₆. b Galvanostatic and galvanostatic-potentiostatic cycling performances of the aloe-emodin|| K₄Fe(CN)₆ AORFB for 2630 cycles tested under galvanostatic mode (darker-colored points) and subsequently tested under galvanostatic-potentiostatic mode (lighter-colored points). c Molecular structures of Cys-AE and K₄Fe(CN)₆. d Polarization curves of the Cys-AE|| K₄Fe(CN)₆ AORFB at varied SOCs. e Discharge capacity, Coulombic efficiency and energy efficiency of the Cys-AE|| K₄Fe(CN)₆ AORFB when galvanostatically cycled at current densities of 20, 50, 100, 150, and 200 mA cm⁻², respectively. f Galvanostatic and galvanostatic-potentiostatic cycling performances of the Cys-AE|| K₄Fe(CN)₆ AORFB for 2693 cycles tested under galvanostatic mode (darker-colored points) and subsequently tested under galvanostatic-potentiostatic mode (lighter-colored points). The AORFBs were assembled with either 5 mL of 0.1 M aloe-emodin or Cys-AE in 1.0 M KOH solution (negolyte) and 30 mL of 0.1 M K₄Fe(CN)₆ in 1.0 M KOH solution (posolyte).

As a control experiment, a long-term cycling test of aloe-emodin|| K₄Fe(CN)₆ AORFB at low concentrations was first conducted under a galvanostatic mode at a constant current density of 100 mA cm⁻² with voltage cutoffs of 0.5 V and 1.5 V. The cell exhibited a maximum discharge capacity of 4.496 Ah L⁻¹ and retained a capacity of 4.356 Ah L⁻¹ after 1066 cycles (or 5.2 days). The Coulombic efficiency maintained close to 100%, and the average energy efficiency was 78%. Subsequently, to assess the molecular stability of aloe-emodin at nearly 100% SOC, the AORFBs were tested under a galvanostatic-potentiostatic mode by holding the potential voltage limit (1.5 V for charge and 0.5 V for discharge) until the current density decreased to 4 mA cm⁻², ensuring that all aloe-emodin molecules reached the reduced state as much as possible. Over the duration of 1564 cycles (7.7 days), the capacity decreased from 4.836 Ah L⁻¹ to 4.498 Ah L⁻¹, resulting in a fading rate of 0.00446% cycle⁻¹ or 0.908% day⁻¹. The fade rate (per cycle) in the galvanostatic-potentiostatic test was nearly three times higher than that in the galvanostatic test, indicating that the cycling stability of aloe-emodin was greatly affected by the SOC of the AORFB (Fig. 3b).

In contrast, the specific capacity of Cys-AE|| K₄Fe(CN)₆ AORFB increased rapidly, stabilizing at 4.727 Ah L⁻¹ after 300 cycles. By the end of the galvanostatic test (1090 cycles), the capacity had risen to 4.847 Ah L⁻¹. The Coulombic efficiency kept at nearly 100%, and the average energy efficiency was 78% (Fig. 3f). This continuous capacity increase has also been observed in previous reports^32,33,which may be attributed to membrane activation during testing, leading to reduced impedance, or an increase in ambient temperature, resulting in enhanced capacity. As a result, the apparent discharge capacity may remain stable or even increase over time. To be more reasonable, we re-evaluated the cycling performance of the AORFB based on 0.1 M Cys-AE using a pre-stabilized membrane. As shown in Supplementary Fig. 18, under galvanostatic conditions, the battery capacity continued to increase, reaching a maximum of 4.666 Ah L⁻¹ after 600 cycles and stabilizing at 4.643 Ah L⁻¹ by the end of the test. From the results of both tests, it is evident that the low-concentration AORFBs take a long time to reach a stable maximum capacity under galvanostatic conditions, making it difficult to accurately reflect the actual capacity decay rate³⁴. Furthermore, the galvanostatic testing conditions did not reach 100% SOC, which cannot fully reflect the actual battery capacity. To avoid these issues, we evaluated the capacity fade rate under galvanostatic-potentiostatic mode as a standard to assess molecular stability. During the galvanostatic-potentiostatic cycling test, the discharge capacity of 0.1 M Cys-AE|| K₄Fe(CN)₆ AORFB faded from 5.133 Ah L⁻¹ to 5.115 Ah L⁻¹ after 1630 cycles (or 8.8 days), corresponding to a capacity fade rate of 0.000219% cycle⁻¹ or 0.0398% day⁻¹. To ensure consistency, we tested another identical 0.1 M Cys-AE|| K₄Fe(CN)₆ AORFB, and its capacity fade rate under galvanostatic-potentiostatic test (Supplementary Fig. 18) was calculated to be 0.000342% cycle⁻¹ or 0.0404% day⁻¹ during 1030 cycles, reaffirming the excellent electrochemical stability of Cys-AE under 100% SOC.

Considering the higher solubility and improved electrochemical stability of Cys-AE, we further investigated the electrochemical performances of the Cys-AE|| K₄Fe(CN)₆ AORFBs at a high negolyte concentration of 0.5 M Cys-AE (Fig. 4). The rate performance of the high-concentration Cys-AE|| K₄Fe(CN)₆ AORFB is depicted in Fig. 4a, b. At current densities of 20 and 100 mA cm⁻², the measured discharge capacities were 25.737 and 23.722 Ah L⁻¹, respectively, corresponding to energy efficiency values of 96.03% and 88.51%, respectively. Importantly, all of the Coulombic efficiencies were maintained at nearly 100%. The polarization curves at different SOCs of 20%, 50%, and 100% were recorded using the linear sweep voltammetry (LSV) method (Fig. 4c), with the galvanic peak power density measuring 215 mW cm⁻² at 100% SOC. Figure 4d displays the OCVs of the high-concentration Cys-AE|| K₄Fe(CN)₆ AORFBs measured at different SOCs, which increased linearly from 10% to 90% SOC. The OCV at 50% SOC was measured to be 1.02 V, consistent with the CV results in Fig. 1b. We have summarized a comparison of the performance of various anthraquinone derivatives as presented in Supplementary Table 4. Notably, Cys-AE exhibits distinct advantages in terms of stability^{9,16,17,30,35–39}.

Fig. 4. Electrochemical performances of the Cys-AE|| K₄Fe(CN)₆ AORFBs at high concentrations.

a Galvanostatic charge-discharge curves of the Cys-AE|| K₄Fe(CN)₆ AORFB at current densities of 20, 50, 100, 150, and 200 mA cm⁻², respectively. b Discharge capacity, Coulombic efficiency, and energy efficiency of Cys-AE|| K₄Fe(CN)₆ AORFB at varied current densities. c Polarization curves of the Cys-AE|| K₄Fe(CN)₆ AORFB at varied SOCs. d OCV versus SOC curve of the Cys-AE|| K₄Fe(CN)₆ AORFB. e Self-discharge test at 100% SOC. The battery was rested for 24 h after being fully charged at the 810th cycle. f Corresponding charge−discharge curves from the 809th to the 812th cycle. g Long-term cycling performance of the Cys-AE|| K₄Fe(CN)₆ at a current density of 100 mA cm⁻² for the entire 804 cycles. h Long-term cycling performance of the Cys-AE|| K₄Fe(CN)₆ under galvanostatic-potentiostatic mode for the entire 592 cycles. The Cys-AE|| K₄Fe(CN)₆ AORFBs were assembled with 6 mL of 0.5 M Cys-AE in 1.0 M KOH solution (negolyte) and 30 mL of 0.4 M K₄Fe(CN)₆ in 1.0 M KOH solution (posolyte).

To study the self-discharge behavior of the high-concentration Cys-AE|| K₄Fe(CN)₆ AORFB, the battery was charged to 1.5 V at 100 mA cm⁻² (~100% SOC) and then rested for 24 h. This resulted in a drop in OCV from 1.12 to 1.09 V and a capacity fade from 23.642 Ah L⁻¹ (charge capacity) to 22.280 Ah L⁻¹ (discharge capacity) in the subsequent discharge step, corresponding to a low self-discharge ratio of 5.76% (Fig. 4e, f). Notably, the charge and discharge capacities fully recovered to 23.562 Ah L⁻¹ at the 811th cycle. These results demonstrate the excellent chemical and electrochemical stability of Cys-AE in its reduced state. The self-discharge behavior after resting for 24 h at 100% SOC may be attributed to the instability of the K₄Fe(CN)₆ posolyte, rather than the Cys-AE negolyte, possibly due to insufficient conversion of ferricyanide to ferrocyanide⁴⁰.

The long-term cycling performance of the high-concentration Cys-AE|| K₄Fe(CN)₆ AORFB at 100 mA cm⁻² is shown in Fig. 4g. The initial capacity was 23.781 Ah L⁻¹, and the capacity retention after 800 cycles (in 19.1 days) was 99.748% (23.722 Ah L⁻¹ remained), corresponding to a low fading rate of 0.000310% cycle⁻¹, equivalent to 0.0130% day⁻¹. To evaluate the intrinsic stability of the active molecule at high concentrations and near 100% SOC, we performed galvanostatic-potentiostatic test on a 0.5 M Cys-AE|| K₄Fe(CN)₆ AORFB, as shown in Fig. 4h. The initial capacity was 24.779 Ah L⁻¹, and the capacity retention after 592 cycles (in 12.8 days) was 99.44% (24.640 Ah L⁻¹ remained), corresponding to a low fading rate of 0.000948% cycle⁻¹, equivalent to 0.0438% day⁻¹. Even at ~100% (SOC), the battery capacity remained highly stable, thereby validating the outstanding electrochemical stability of Cys-AE negolyte.

To analyze the redox and degradation mechanisms of aloe-emodin and Cys-AE during the cycling processes of AORFBs at pH 14, the cycled electrolytes were investigated using CV, UV-Vis absorption, EPR, and NMR spectroscopies. The CV tests were performed on the negolyte and posolyte reservoirs without dilution (Supplementary Fig. 19), showing no crossover signs of ferrocyanide into the negolyte or Cys-AE into the posolyte. However, the penetration of aloe-emodin through the Nafion membrane could not be completely prevented. For the aloe-emodin|| K₄Fe(CN)₆ AORFB, a pair of new redox peaks appeared in the CV curve of K₄Fe(CN)₆ posolyte after galvanostatic cycling, with potentials close to that of aloe-emodin. This suggests the permeation of aloe-emodin through the Nafion membrane, which could be an important reason for the capacity attenuation of the aloe-emodin|| K₄Fe(CN)₆ AORFB.

The EPR signals of aloe-emodin and Cys-AE at different charge states were consistent with the oxidation process of anthraquinone molecules, as shown in Fig. 5a, d. The EPR resonance of aloe-emodin was centered at a G-factor of 3512.5 (corresponding to a g-factor of 2.0037). Similarly, the EPR resonance of Cys-AE was centered at a G-factor of 3512.0 (corresponding to a g-factor of 2.0034), which was assigned to the Cys-AE^3−• radical anion. As the charging process continued, the EPR signal increased in intensity and broadened, reaching its maximum intensity at 50% SOC, after which it started to decrease in intensity and sharpen. The broadening of EPR signals is caused by the Heisenberg spin exchange, specifically the “flip-flop” dipolar-driven (zero-quantum) spin exchange between two unpaired electrons⁴¹. In paramagnetic molecules, the unpaired electrons have magnetic interactions with nearby magnetic nuclei, known as the hyperfine interaction. This interaction results in the splitting of the original single EPR lines into multiple lines⁴². The different substituent side chains of aloe-emodin and Cys-AE result in variations in the electron density distribution, the distance between unpaired electrons and atomic nuclei, as well as the nuclear spin. Consequently, these variations exert a profound influence on the spin interactions occurring between unpaired electrons and their surrounding atomic nuclei. These interactions manifest as distinctive hyperfine splitting and hyperfine structures within the EPR spectra. Specifically, when compared to an oxygen atom in ether bonds (R–O–R’), the sulfur atom exhibits greater polarizability owing to its more diffuse electron cloud, thereby enhancing its propensity to accept electrons. Conversely, the hydroxyl group (–OH) within the –CH₂OH moiety of aloe-emodin primarily functions as an electron-donating group, facilitated by a conjugation effect. The distinct electron-withdrawing or electron-donating characteristics of these side chains modify the distribution of unpaired electrons within the molecules, resulting in varying degrees of electron density localization. Subsequently, this localized electron density gives rise to differential EPR signal splitting and hyperfine structures⁴³.

Fig. 5. Redox and degradation mechanism studies of aloe-emodin and Cys-AE.

a EPR spectra of aloe-emodin negolyte under different charge states. b, c In situ UV-Vis absorption spectra of aloe-emodin during b charging and c discharging processes. d EPR spectra of Cys-AE negolyte under different charge states. e, f In situ UV-Vis absorption spectra of Cys-AE during (e) charging and f discharging processes. g ¹H NMR spectra of 0.1 M aloe-emodin negolyte at initial state, after fully discharged in the 50th, 100th, and 200th cycle at a current density of 100 mA cm⁻², respectively. h ¹H NMR spectra of 0.1 M Cys-AE negolyte at initial state, after fully discharged in the 50th, 100th, and 200th cycle at a current density of 100 mA cm⁻², respectively. i Possible decomposition pathways and byproducts of Cys-AE.

The redox behaviours of anthraquinone cores in aloe-emodin and Cys-AE during cycling processes were further monitored by in situ UV-Vis absorption analyses. Prior to charging, the diluted aloe-emodin in 1.0 M KOH negolyte exhibited a red color, which transformed into a yellow color upon charging to its reduced form. As the charging process progressed, the absorption peak at λ_504nm gradually diminished, and a new peak emerges at λ_431nm, with its intensity gradually increasing, indicating the transformation from the C=O to the C–O⁻. These reversible changes in absorption peaks during the charge/discharge process are clearly observed in both aloe-emodin (Fig. 5b, c) and Cys-AE (Fig. 5e, f), demonstrating the robust redox activity and stability of anthraquinone cores in both molecules.

The NMR spectra of aloe-emodin and Cys-AE negolytes after different cycles are shown in Fig. 5g, h. Due to the simpler molecular structure of aloe-emodin, the NMR spectra did not change significantly after long-term cycling. The methylene side-chain remained stable, while the peak splitting in the aromatic region was not as clear as in the initial state. A common degradation mechanism observed for anthraquinones is the formation of anthrone and/or anthrone dimer⁴⁴. The presence of anthrone and/or anthrone dimer would not be detected in the above post-cycling NMR and HRMS analyses due to the presence of an air atmosphere, which could oxidize any anthrone and/or anthrone dimer that might have formed²⁸. To investigate the formation of anthrone derived from aloe-emodin, the negolyte was electrochemically reduced to approximately 100% SOC and then rested for 14 days in an N₂ atmosphere. The aged aloe-emodin negolyte was acidified to protonate the anthraquinone derivatives (AQ) for precipitation, and the precipitates were analyzed using NMR and HRMS. The NMR spectra (Supplementary Fig. 20) in combination with the HRMS data (Supplementary Fig. 21) from the treated aloe-emodin support the formation of a trace amount of anthrone. Some new peaks emerged in the ¹H NMR spectrum (marked by “✽“ in Supplementary Fig. 20), especially in the chemical shift region of the aliphatic protons, suggesting that the degradation mechanism of aloe-emodin involves the formation of anthrone. Nevertheless, the relative abundances of HRMS peaks demonstrated that the proportion of anthrone formation is extremely small, and no evident peaks of anthrone dimer byproduct were found in the HRMS results. However, it is notable that the capacity fade rate of the aloe-emodin|| K₄Fe(CN)₆ AORFB was significantly higher than that of Cys-AE|| K₄Fe(CN)₆ AORFB, which could be mainly attributed to the higher permeability of aloe-emodin (as depicted by the CV results in Supplementary Fig. 19), rather than the anthrone transformation of aloe-emodin.

To accurately assess the stability of Cys-AE, NMR spectra were measured after cycling for different numbers of cycles at current densities of 20 mA cm⁻² and 100 mA cm⁻², respectively (Supplementary Figs. 22 and 23). Regardless of whether at low or high current density, the NMR signals of Cys-AE remained essentially consistent with the initial state during the early period of the corresponding cycling test. However, as the test duration increased, the cycled Cys-AE negolyte exhibited new peaks in the aliphatic region, and the peaks in the aromatic region did not show clear splitting. The new aliphatic peaks could be ascribed to the loss of cysteine side-group (Fig. 5i). Nevertheless, stable long-cycling performance demonstrated that a small proportion of side-group loss in Cys-AE did not obviously impact the capacity stability of the Cys-AE|| K₄Fe(CN)₆ AORFB. The presence of a soluble hydroxyl group on the anthraquinone core prevents the molecule from becoming insoluble and precipitating even with a small amount of side-group loss. The Cys-AE negolyte after long-term cycling was treated in the same manner as the aloe-emodin negolyte and characterized using NMR and HRMS. Some new peaks emerged in the ¹H NMR spectrum (marked by “✽“ in Supplementary Fig. 24) of Cys-AE negolyte after the cycling test, particularly in the chemical shift region of the aliphatic proton, indicating the cleavage of the cysteine group from the anthraquinone core. According to the HRMS spectrum (Supplementary Fig. 25), we deduce that the byproducts were 1,8-dihydroxy-3-(mercaptomethyl)anthracene-9,10-dione (byproduct 1) and (9,10-dihydro-1,8-dihydroxy-9,10-dioxoanthracen-6-yl)methanesulfonic acid (byproduct 2), as illustrated in Fig. 5i. The relative abundances of HRMS peaks indicated that only a trace amount of Cys-AE underwent side-group loss, with the major byproduct being (9,10-dihydro-1,8-dihydroxy-9,10-dioxoanthracen-6-yl)methanesulfonic acid. Additionally, no significant peaks of anthrone derivative or anthrone dimer derivative byproducts were observed in the HRMS results.

We have developed a brief techno-economic analysis model to evaluate the economic feasibility and environmental friendliness of Cys-AE, presenting the synthesis costs and carbon emissions compared with 4,4’-((9,10-anthraquinone-2,6-diyl)dioxy)dibutyrate (2,6-DBEAQ)¹⁶, which is synthesized from fossil fuel-based precursors^45–47. As shown in Supplementary Tables 5 and 6, and Supplementary Fig. 26, by utilizing renewable raw materials such as aloe-emodin and cysteine, the synthesis pathway of Cys-AE significantly reduces the dependence on non-renewable fossil resources, enhancing the overall sustainability of the AORFBs. Although the cost of aloe-emodin is determined by the current plant extraction and purification technology, the overall synthesis pathway of Cys-AE possesses lower solvent demand, lower energy consumption and reduced overall carbon emissions compared to 2,6-DBEAQ, demonstrating strong potential for practical applications.

While reaction conditions involving C–S–C linkages are milder and allow for easier separation, mechanistic analysis has revealed the possible breakage of C–S–C linkages during extended electrochemical cycling. To address this issue, here we propose several possible optimization strategies to further reduce costs and improve stability, as shown in Supplementary Fig. 27. These strategies include selecting the more affordable emodin as precursor and replacing C–S–C linkages with more stable C–O–C linkages. However, compared to cysteine, whether tyrosine and salicylic acid with a hydrophobic benzene ring might negatively impact the molecule’s solubility still requires detailed experimental validation. We plan to further optimize the synthesis methods and evaluate the performance of these derivatives in future studies.

To evaluate the stability of Cys-AE in the air environment, the Cys-AE|| K₄Fe(CN)₆ AORFB was cycled without the protection of an inert atmosphere (Supplementary Fig. 28). The capacity continuously declined despite the Coulombic efficiency remaining close to 100%. We suggest that this is due to the oxidation of reduced anthraquinone units by O₂ in the air, which hinders the reduction of [Fe(CN)₆]³⁻ and leads to capacity loss. However, by replacing the posolyte with a fresh K₄Fe(CN)₆ solution, the battery capacity can be fully restored. To further understand the influence of air exposure on the stability of Cys-AE, we conducted NMR testing on the charged negolyte. Due to interference from free radicals, the reduced form of Cys-AE did not yield a clear NMR spectrum. However, after being exposed to air atmosphere for a period of 12 h, the reduced Cys-AE was oxidized by O₂ in the air, and its NMR spectrum (Supplementary Fig. 29) still aligned well with that of the initial Cys-AE negolyte, indicating that both the oxidized and reduced forms of Cys-AE exhibit good chemical stability in air.

Furthermore, the thermostability of Cys-AE negolyte under alkaline conditions at relatively high temperatures was evaluated. The NMR spectrum of Cys-AE in 1.0 M KOH solution after heating for 7 days at 60 °C is presented in Supplementary Fig. 30, indicating that long-term heating in an alkaline environment could lead to the loss of cysteine side-group in Cys-AE. Therefore, in the practical application of Cys-AE-based AORFBs, it is recommended to avoid high operating temperatures and oxygen exposure.

Discussion

In summary, we report the successful synthesis of Cys-AE, an α-amino acid grafted anthraquinone derivative, by functionalized aloe-emodin natural product with cysteine. The introduction of hydrophilic amino acid side chains in Cys-AE resulted in high solubility, excellent electrochemical reversibility, and fast electron transfer kinetics. The performances of AORFBs based on Cys-AE negolyte paired with ferro/ferricyanide catholyte were extensively investigated at pH 14 under different cut-off voltage windows and concentrations (0.1 M and 0.5 M). The Cys-AE negolyte demonstrated significantly superior cycling stability (with a fade rate of 0.0438% day⁻¹ during 592 cycles) compared to that of pristine aloe-emodin (with a fade rate of 0.908% day⁻¹ during 1564 cycles). The capacity degradation of pristine aloe-emodin was primarily attributed to transmembrane crossover and the formation of electrochemically inactive anthrone derivative, as verified by ¹H NMR, CV, and HRMS studies. On the other hand, the main degradation mechanism of Cys-AE was ascribed to a small ratio of side-chain loss, with no obvious formation of anthrone derivative or anthrone dimer derivative. However, despite a small portion of side-chain loss, the small quantities of Cys-AE decomposition byproducts still remain soluble due to the phenolic hydroxyl group grafted onto the anthraquinone core. This solubility ensures that the electrochemical performance of the AORFB remains almost unaffected, demonstrating the robustness and durability of Cys-AE negolyte. This study highlights the structural engineering strategy of highly soluble and redox-reversible organic molecules derived from green and resource-rich natural products to enhance the overall performances of AORFBs and establish sustainable large-scale energy storage systems.

Methods

Chemicals and materials

Naturally produced aloe-emodin (95%), L-cysteine (99%), PBr₃ (95%), and carbon tetrachloride (CCl₄, 99%) were purchased from Shanghai Bide Bio-Chem Technology Co., Ltd. Potassium ferrocyanide (K₄Fe(CN)₆, 99%), potassium ferricyanide (K₃Fe(CN)₆, 99%), HCl (36%), and KOH (99%) were purchased from Sinopharm Chemical Reagent Co., Ltd. All the reagents were used without further purification. Graphite felts and graphite plates were purchased from SGL Corp., Germany. Nafion-212 membranes were purchased from DuPont Corp., USA.

Synthesis of cysteine-functionalized aloe-emodin (Cys-AE)

A mixture of aloe-emodin (1.35 g, 5 mmol) and CCl₄ (150 mL) was added into a round-bottom flask. Subsequently, PBr₃ (5 mL) was added dropwise to the mixture at room temperature. The resulting mixture was then heated and evaporated at 65 °C for 48 h, yielding a dark yellow solid. This solid was purified through flash column chromatography with silica gel using a mixture of dichloromethane and petroleum ether in a volume ratio of 2:1. The desired product, 3-(bromomethyl)-1,8-dihydroxyanthracene-9,10-dione, was obtained in a yield of 1.23 g (70%). Then, the obtained intermediate product (1 g, 3 mmol) was mixed with cysteine (0.73 g, 6 mmol) in a 1.0 M NaOH solution (30 mL). The resulting mixture was stirred at room temperature for 12 h and then acidified with hydrochloric acid. The final Cys-AE product was obtained by filtration with a yield of 1.06 g (95%).

Materials characterizations

The as-synthesized Cys-AE powder was dissolved in D₂O with 1.0 M KOH, and the NMR spectra were recorded using a Bruker DPX 400 MHz spectrometer. The chemical shifts (δ) were reported in ppm with respect to TMS. 0.1 mL of negolyte at fully discharged states was extracted from the reservoir and diluted in D₂O containing 1.0 M KOH at a fixed volume ratio of 1:4 to prepare the NMR samples. The HRMS analysis of Cys-AE was conducted on a time-of-flight (TOF) MS analyzer (micrOTOF-Q III MS, Bruker) with electrospray ionization (ESI) source and recorded in the positive ionization mode. The Electron paramagnetic resonance (EPR) spectra were acquired utilizing a Bruker EMX plus-6/1 variable-temperature apparatus. The aloe-emodin or Cys-AE based negolyte at different charge-discharge states was hermetically sealed in a quartz capillary, and promptly evaluated at ambient temperature.

UV-Vis absorption characterizations

For the solubility tests at pH 14, standard aloe-emodin or Cys-AE solutions in 1.0 M KOH were prepared within 25 mL volumetric flasks. The UV-Vis absorbance measurements of these solutions were obtained using a UV-Vis spectrometer (UV-2600, Shimadzu Scientific Instrument) in a wavelength range of 200 to 700 nm, with a 0.5 nm interval. Working curves of the standard solutions were generated based on their peak absorbance at 504 nm. Saturated solutions were created by adding aloe-emodin or Cys-AE into 1.0 mL of 1.0 M KOH solution until a small quantity of precipitate formed. Subsequently, 0.1 mL of the saturated solution was diluted 5,000-fold (aloe-emodin) or 10,000-fold (Cys-AE) with 1.0 M KOH solution using a pipette. The maximum solubilities were determined by plotting UV-Vis absorbance against concentration.

To investigate the structural transition of aloe-emodin and Cys-AE, the UV-Vis absorption spectra of 0.10 mM aloe-emodin or Cys-AE solution were recorded at varying pH values, ranging from pH 14 to pH 0, and adjusted using diluted HCl and KH₂PO₄ solutions. The pH measurements were obtained using a pH meter (pHS-3C, INESA, China) calibrated with pH 4.00 and 6.86 buffer solutions.

DFT calculations

The geometry optimizations and frequency calculations were performed using M06-2X functional with 6-311 G(d,p) basis set. The implicit conductor-polarizable continuum model (CPCM) was used to represent the solvent effect of water molecules. The structures of aloe-emodin (AE) and Cys-AE ground state were optimized at PBE0/6-311 G(d,p) level. After structural optimization, TDDFT calculations were conducted on the optimized structures at the same level, taking the lowest 20 excited states into account. The CPCM solvent model with a dielectric constant of water was utilized to account for the solvent effect. The UV-Vis spectra were then drawn using the Multiwfn program. The molecular orbitals and energy gaps of aloe-emodin and Cys-AE were calculated using the Multiwfn program.

Pretreatment of membranes and permeability tests

The Nafion-212 ion-exchange membranes were pre-treated in an aqueous solution containing 5% H₂O₂ and 1.0 M KOH solution and then heated for 30 minutes. Then, the Nafion-212 membranes were rinsed with deionized water, soaked in deionized water and heated for another 30 minutes. Finally, the membrane was soaked in an aqueous solution containing 1.0 M KOH solution and boiled for another 30 minutes. After cooling, the processed Nafion-212 membranes were stored in deionized water for later use.

The permeability of aloe-emodin or Cys-AE across the Nafion-212 membrane was assessed using a standard two-compartment H-type cell and a UV-Vis spectrometer (UV-2600, Shimadzu Scientific Instrument). The left compartment of the H-type cell was filled with a 0.2 M aloe-emodin or 0.5 M Cys-AE solution in 1.0 M KOH, while the right compartment contained an isovolumic 1.0 M KOH solution. Throughout the tests, both compartments of the H-type cell were stirred continuously. The concentration of the solution in the right compartment was determined based on a standard light absorbance-concentration curve of known concentrations at various time intervals. Subsequently, the permeability of aloe-emodin or Cys AQ was calculated using Fick’s diffusion law, as described in the following equation:

$$P=\frac{\mathrm{\Delta }\ln \left(1-\frac{2C_t}{C_0}\right)\left(\frac{V_{0}l}{2A}\right)}{\mathrm{\Delta }t}$$ 1

where P is the permeability of active species (cm² s⁻¹), A is the effective area of the membrane (2.01 cm²), C_t (mol L⁻¹) is the concentration of active species in the right compartment at time t, C₀ is the concentration of active species in the left compartment at time zero (0.5 mol L⁻¹), V₀ is the volume of the solution in either reservoir (15 cm³), l is the thickness of the membrane (50 μm for the Nafion-212 membrane, respectively). The measurements were only performed once. The data are collected and directly used.

Electrochemical characterizations

CV measurements were conducted using a Chenhua CHI-760E electrochemical workstation. A standard three-electrode configuration was employed, comprising a glassy carbon electrode (GC 130, China) with a diameter of 3 mm as the working electrode, a saturated Ag/AgCl electrode (0.199 V vs. SHE) as the reference electrode, and a platinum electrode as the counter electrode. 5 mL of 1 mM aloe-emodin or Cys-AE solution in 1.0 M NaOH were used. The CV curves of 1 mM aloe-emodin or Cys-AE solution were measured at various pH values ranging from pH 14 to pH 7, which were adjusted using diluted HCl and KH₂PO₄ solutions, respectively. The pH values were determined using a pH meter (pHS-3C, INESA, China) calibrated with pH 4.00 and 6.86 buffer solutions. All electrochemical tests were conducted at room temperature (~25 °C) under ambient atmospheric conditions, with no additional environmental control measures applied. The diffusion coefficients (D) were measured with 1 mM aloe-emodin or Cys-AE in 1.0 M KOH solution under different scan rates and calculated according to the Randles-Sevcik equation:

$$ip=268600\times n^{2/3}{AD}^{1/2}{cv}^{1/2}$$ 2

where ip is the peak current in the CV curve, n is the number of electrons transferred in the redox reaction, A is the area of the graphite felt, v is the scan rate, and c is the mole concentration of the active material.

The electron transfer rate constant (k₀) was calculated by Nicholson’s method:

$$\psi =k_0\times {\left(\frac{\pi DnF}{RT}\right)}^{-1/2}{v}^{-1/2}$$ 3

$$\psi =\frac{-0.6288+0.0021n\Delta Ep}{1-0.017n\Delta Ep}$$ 4

where D and n are defined in Eq. (2), F is the Faraday’s constant (96485 C mol⁻¹), R is the universal gas constant (8.314 J K⁻¹), T is the temperature (293.15 K), ν represents the scan rate, and ΔEp is the peak-to-peak separation between the cathodic and anodic peak potentials. CV measurements were conducted in triplicate, and the data from the second cycle was selected for analysis.

The EIS measurement of AORFBs was performed before cycling test. The sinusoidal voltage oscillations with an amplitude of 5 mV were applied in a frequency range of 100 kHz to 0.1 Hz.

RFB tests

Graphite felts were pressed onto the graphite flow fields on each side of the flow cell and sealed with PTFE rubbers. The SGL graphite felts, with a thickness of 3 mm and an area of 5 cm², were preheated in a muffle furnace at 400 °C for 24 h. The graphite felts were separated by a Nafion-212 membrane (50 µm thick, 6 cm × 6 cm in area). The electrolytes were introduced into the flow battery at a rate of 70 mL min⁻¹ using a two-channel peristaltic pump. The flow battery was tested by a battery test station (CT3002A, Wuhan Land Instruments, China). The SOCcurve was obtained by recording the OCVs after charging the flow cell to different states at a current density of 100 mA cm⁻² and resting for 30 s. Approximately 100% SOC was achieved by charging the cell with the voltage held at 1.5 V and maintaining the voltage until the current density decreased to 4 mA cm⁻². The polarization curves were measured using the LSV method at 20%, 50%, and 100% SOCs with a scan rate of 50 mV s⁻¹, employing a CS-350H electrochemical workstation with a current booster (Wuhan Corrtest Instruments, China).

For the galvanostatic cycling tests involving 0.1 M aloe-emodin or Cys-AE, the negolyte consisted of 5 mL of 0.1 M aloe-emodin or Cys-AE in 1.0 M KOH, while the posolyte contained 30 mL of 0.1 M K₄Fe(CN)₆ and 0.01 M K₃Fe(CN)₆ in 1.0 M KOH solution. Firstly, the AORFBs were galvanostatically cycled at 100 mA cm⁻¹ between 1.5 V and 0.5 V. Subsequently, the AORFBs were cycled with a galvanostatic-potentiostatic mode, i.e., the AORFBs were firstly charged to 1.5 V at a current density of 100 mA cm⁻², and then the voltage was maintained at 1.5 V until the current density decreased to 4 mA cm⁻², followed by discharging to 0.5 V, and the voltage was maintained at 0.5 V until the current density decreased to 4 mA cm⁻².

For the cycling tests involving 0.5 M Cys-AE, the negolyte comprised 5 mL of 0.5 M Cys-AE in 1.0 M KOH, and the posolyte included 40 mL of 0.4 M K₄Fe(CN)₆ and 0.04 M K₃Fe(CN)₆ in 1.0 M KOH solution. The AORFBs were cycled with a galvanostatic or galvanostatic-potentiostatic mode. Unless otherwise specified, all the tests of AORFBs were conducted in a nitrogen environment and at room temperature (~25 °C). We reported all battery test results in the main text, with only the battery based on 0.1 M Cys-AE undergoing duplicate testing.

Author contributions

Z.J. conceived the idea of this study. Z.J. and Y.Z.L. designed the experiments. Y.Z.L., Z.A.W., P.B.Z., and J.W. performed the sample synthesis, electrochemical measurements, cell tests, and data analysis. Y.Z.L, P.B.Z., J.J.L., and J.M. did the theoretical calculations and analysis. Y.Z.L., Z.A.W., P.B.Z., J.W., H.Z.W., S.W., J.C.L., Y.K.C, T.F.D., Z.X.T., X.Z.W., and Z.J. analyzed the data and discussed the results. Y.Z.L. and Z.J. co-wrote and revised the manuscript. Z.J. supervised the project.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Data availability

The authors declare that all data supporting the finding of this study are available within the paper and its Supplementary Information. 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/s41467-025-58165-y.