Enabling high-voltage aqueous dual-ion batteries capable of working at −40 °C in a low-concentration salt electrolyte Authors: Chuan Li, Xintao Ma, Dedi Li, Rong Zhang, Huilin Cui, Shaoce Zhang, Zhaodong Huang, Qingshun Nian, Hui Yang, Chunyi Zhi Published: 2026 ABSTRACT AbstractGraphite positive electrode-based anion-intercalation aqueous dual-ion batteries using bis(trifluoromethanesulfonyl)imide anion intercalations show the potential because of the high discharge plateau (>1.4 V vs. the standard hydrogen electrode at 25 °C ± 2 °C) for anion chemistry. However, currently, the electrochemical performance of aqueous dual-ion batteries based on bis(trifluoromethanesulfonyl)imide anion intercalation can only be achieved in highly concentrated salt electrolytes. This dependence on high anion concentration to activate intercalation inevitably leads to high costs and prevents operation at low temperatures (e.g., < −20 °C). In this context, we explore a hydrogel electrolyte rich in dipole molecules, incorporating only 1 M lithium bis(trifluoromethanesulfonyl)imine and 2 M zinc chloride, with expanded electrochemical stability windows (>4.6 V at −40 °C) to activate bis(trifluoromethanesulfonyl)imide anion insertion into the graphite positive electrode through high-voltage scanning to 3.0 V. The assembled aqueous dual-ion batteries based on bis(trifluoromethanesulfonyl)imide anion intercalation demonstrate high discharge plateaus of 1.9 V at −20 °C and 1.7 V at −40 °C, along with a specific energy of 595 Wh kg−1 (normalized to the 1 mg cm−2 of the graphite positive electrode).Subject terms: Batteries, Batteries Introduction Anion-intercalation aqueous dual-ion batteries (ADIBs) have garnered significant attention due to their environmentally friendly nature, safety features, and electrochemical properties1–5. Generally, bis(trifluoromethanesulfonimide) (TFSI-) intercalation exhibits a higher potential compared to other anions (e.g., F-, NO3-, [ZnCl4]2-) intercalation2,3,6–8. In particular, graphite (G) positive electrode-based ADIBs utilizing TFSI- intercalation show promise owing to their low cost and high discharge plateau (> 1.4 V vs. the standard hydrogen electrode at 25 °C ± 2 °C)6,7. Therefore, developing ADIBs based on TFSI- intercalation into G-positive electrodes presents a compelling option for future energy storage devices. However, activating the electrochemical performance of anion-intercalation ADIBs typically requires a high concentration of TFSI-6,7,9,10. This is because activating anion intercalation into G-positive electrodes necessitates a sufficient anion concentration, which facilitates the desolvation of hydrated anions during cycling and enables high-voltage scanning to activate anion intercalation6,7,9,10. In other words, low-concentration salt electrolytes cannot be used to activate anion intercalation into G positive electrodes due to the strong hydrated anion state and narrow electrochemical stability window (ESW) (Fig. 1). Nonetheless, the high-concentrated salt electrolyte with sufficient anion concentration unavoidably leads to high costs11–14 and a relatively high freezing point15–17 (e.g., > -20 °C, see Fig. S1), which usually results from the recrystallization of salts based on the decreased solubility at low temperatures. Afterwards, the phase separation induces icing based on heterogeneous nucleation in the high-concentrated salt electrolyte (Fig. 1). The hydrated anion promotes anion intercalation chemistry. The schematic diagram illustrates the different hydrated anion states in low-concentration salt electrolytes, high-concentration salt electrolytes, and DM-rich hydrogel electrolytes, as well as the corresponding anion intercalation into the graphite positive electrode electrochemical behaviors at −40 °C. Noted: oxygen atom (red sphere), hydrogen atom (white sphere), carbon atom (brown sphere). The development of intrinsically low-cost dipole molecule (DM)-containing electrolytes, such as those featuring a large number of hydroxyl groups (-OH), demonstrates the capability to simultaneously broaden the ESW and regulate the solvation sheath configuration of charge carriers11,18. Dipole molecules like ethylene glycol (EG) are also commonly used as commercial antifreeze19,20. Based on this, DM-rich hydrogel electrolytes hold the promise to solve the high concentration requirement of TFSI- and improve the low-temperature performances of ADIBs. (Fig. 1) Firstly, the broadened ESW of DM-rich hydrogel electrolytes, particularly at low temperatures, provides an opportunity to activate TFSI- intercalation into G positive electrodes using high-voltage scanning21–23. Secondly, the intrinsic anti-freezing property promotes the desolvation of the hydrated TFSI- anion at low temperatures by increasing the inertness of water molecules21–23. Finally, the abundance of DM reduces the need for a large number of TFSI- anions, thereby reducing the cost of electrolytes11,18. In this study, DM-rich hydrogel electrolytes were synthesized using polyacrylamide-D(+)-glucose monohydrate (PAM-DG), which leverages its hydrophilic properties due to the introduction of DG with a large number of -OH groups. Then, EG was added to the 1 M LiTFSI + 2 M ZnCl2 solution electrolyte (TFSI-Cl) to prepare the hydrogel electrolyte labeled EG-PAM-DG-TFSI-Cl. The EG-PAM-DG-TFSI-Cl demonstrated an ESW of >4.6 V at −40 °C for activating TFSI- intercalation into G positive electrodes of ADIBs with a 3,4,9,10-perylenetetracarboxylic diimide (PTCDI) organic negative electrode by high-voltage scanning to 3.0 V, resulting in a 1.7 V discharge plateau and a specific capacity of 200 mAh g−1 for >470 hours at 80 mA g−1, and 350 mAh g−1 at 60 mA g−1. Additionally, the assembled ADIBs delivered a 1.9 V discharge plateau and a capacity of 200 mAh g−1 regardless of the mass loading of G (1 or 3 mg cm−2) when operating at −20 °C. Results Fabrication and characterization The synthesis of the EG-PAM-DG-TFSI-Cl utilizing radical polymerization is depicted in Fig. S2. DG, rich in -OH groups, is incorporated into PAM chains to produce PAM-DG. Furthermore, EG-PAM-DG is created by using EG-H2O as the mixed solvent, providing an additional source of -OH groups. When EG is mixed with TFSI-Cl as a solvent, EG-PAM-DG-TFSI-Cl is fabricated, as verified by attenuated-total-reflection Fourier transform infrared (ATR-FTIR) spectra (Fig. S3a). Peaks at 1455 and 1320 cm−1 are ascribed to the stretching vibrations of PAM polymer chains24, while peaks at 1640 cm−1 belong to DG25. The characteristic signal peaks at 1200 and 1140 cm−1 of LiTFSI are observed in the ATR-FTIR spectrum of the EG-PAM-DG-TFSI-Cl, but are not present in that of the EG-PAM-DG, indicating the introduction of LiTFSI in the EG-PAM-DG-TFSI-Cl (Fig. S3a). Similarly, the Raman spectra show the same results as the ATR-FTIR spectra. Specifically, the characteristic signal peak at 745 cm−1, corresponding to the TFSI-hydrated ion26, is observed in the Raman spectrum of the EG-PAM-DG-TFSI-Cl but is absent in the EG-PAM-DG. Additionally, the peak at 284 cm−1 belonging to Zn(H2O)2Cl42- can be found in the Raman spectrum of the EG-PAM-DG-TFSI-Cl but not in the EG-PAM-DG16, indicating the successful incorporation of ZnCl2 in the EG-PAM-DG-TFSI-Cl (Fig. S3b). Of note, EG-PAM-DG-TFSI-Cl shows the water content of 35 wt% (Table S1). The increased bound water and decreased free water of the EG-PAM-DG-TFSI-Cl can be discerned through changes in hydrogen bond (H-bond) strength11. Specifically, the ATR-FTIR spectra of the EG-PAM-DG-TFSI-Cl and TFSI-Cl are depicted in Fig. 2a. The H-O bending (1640 cm−1) of H2O in the TFSI-Cl shifts to a higher wavenumber after the addition of the EG-PAM-DG (from 1640 cm-1 in the TFSI-Cl to 1664 cm−1 in the EG-PAM-DG-TFSI-Cl). The signal peak at 1602 cm−1 in the EG-PAM-DG-TFSI-Cl corresponds to the H-O (DG) bending peak27. Raman spectroscopy was employed to analyze the H-O stretching modes and the relative amounts of different water molecule states. The vibration frequencies of H bonds offer critical insights into their structural characteristics. Strong H bonds in a tetrahedral coordination vibrate at frequencies centered around 3242 cm−1, whereas weak H bonds in an incomplete tetrahedral structure are centered near 3422 cm−1 (Fig. 2b)28. These two signal peaks at 3242 and 3422 cm−1 are admitted as the ice-like and liquid-like modes29. The I3242/I3422 ratio in the EG-PAM-DG-TFSI-Cl (1.02) is larger than that (0.97) of TFSI-Cl, suggesting that the stronger H bonds form in the EG-PAM-DG-TFSI-Cl11,30,31. This can be attributed to the incorporation of large numbers of DM, which strengthens the H-bond interactions between polymer chains and water molecules. Fabrication and characterization of the EG-PAM-DG-TFSI-Cl. a ATR-FTIR spectra of various electrolytes. b Raman spectra of various electrolytes. c DSC profiles of various electrolytes. d LSV profiles of EG-PAM-DG-TFSI-Cl at 5 mV/s ranging from 20 to −40 °C. Noted: The hydrogen evolution reaction (HER) onset potential was selected as −0.3 V vs Zn²⁺/Zn at 20, 0, and −40 °C. At 20 °C, the HER was so severe that it produced a very large current response on the curve. To facilitate a better comparison between curves, the y‑axis range was shortened, which prevented the curve from being fully displayed. Additionally, for the measurement at −20 °C, the onset potential was chosen as −0.4 V vs Zn²⁺/Zn to ensure a clear comparison of the HER potential with the other curves. The oxidative stability window was defined at a constant current density of 0.02 mA cm⁻² for each temperature. This current density value was derived as follows: first, the tangent to the LSV curve at 20 °C at the oxide termination end was drawn, and its intersection with the LSV curve at the point where the current density equals zero was identified. Then, the current density on the LSV curve corresponding to the voltage at that intersection point was taken as the desired value. Noted: au arbitrary units. Differential scanning calorimetry (DSC) was utilized to differentiate the states of water in the electrolytes11,16,32. The sharp peak at -22 °C corresponds to the freezing point of water in the TFSI-Cl (Fig. 2c). The EG-PAM-DG exhibited a lower freezing point of water at −32 °C. Notably, after adding TFSI-Cl into the EG-PAM-DG, the melting point was not observed due to the limitation of the testing range of the used instrument (below −70 °C), indicating a lower freezing point of water in the EG-PAM-DG-TFSI-Cl compared to that of TFSI-Cl and EG-PAM-DG. To further explore ESWs of the hydrogel electrolyte, linear sweep voltammetry (LSV) profiles of the EG-PAM-DG-TFSI-Cl at different temperatures (20, 0, -20, −40 °C) were scrutinized (Fig. 2d). The ESWs of the EG-PAM-DG-TFSI-Cl ranged from 3.1 to 4.6 V at different temperatures. The expanded ESW of 4.6 V at −40 °C is attributed to the reduced activity of water molecules resulting from cooling21. Of note, the broader ESW of EG-PAM-DG-TFSI-Cl at low temperatures compared to that at 25 °C ± 2 °C can be attributed to both the higher redox inertness of H₂O and the lower solvation energies of OH⁻ and H⁺ at reduced temperatures, which also applies to other aqueous electrolytes (Fig. S4)21,33. Noted that, compared to low-concentration salt electrolytes, the broader ESW of ~3.1 V of EG-PAM-DG-TFSI-Cl at 25 °C ± 2 °C can be attributed to the formed bound water and the reduced free water based on the introduced large amounts of DMs (e.g., DG and EG)11,34. The costs of the EG-PAM-DG-TFSI-Cl and typical “WIS” electrolyte are summarized in Table S2, with a detailed calculation process exhibited in the supporting information. The specific cost breakdown and the substantial difference in cost between the two types of electrolytes provide a compelling argument for the cost-effectiveness of the EG-PAM-DG-TFSI-Cl. The cost of the EG-PAM-DG-TFSI-Cl is approximately 0.182 dollars per mL. In contrast, the costs of “WIS” electrolytes are an order of magnitude higher, ranging from $ 1.795 to $ 3.358 per mL. This stark contrast highlights the cost-effectiveness of the EG-PAM-DG-TFSI-Cl. Mechanism Evolution in the PTCDI||G Batteries The expanded ESWs are a prerequisite for activating TFSI- anion intercalation into the G positive electrode through high-voltage scanning. Additionally, the introduced ZnCl2 salt exhibited a stronger interaction with water molecules than that of LiTFSI, as evidenced by the lower freezing point of 2 M ZnCl2 (Fig. S5), facilitating the desolvation of hydrated TFSI- anions. Subsequently, we examined the ionic conductivities of EG-PAM-DG incorporating different salts at various temperatures to investigate their ionic behavior. As depicted in Figs. 3a, S6, and Table S3, we initially measured the ionic conductivities of the EG-PAM-DG incorporating 1 M LiTFSI (EG-PAM-DG-TFSI), 2 M ZnCl2 (EG-PAM-DG-Cl), and EG-PAM-DG-TFSI-Cl at different temperatures (5, 10, 15, 20 °C) based on electrochemical impedance spectroscopy (EIS) spectra. At 20 °C, the ionic conductivities of EG-PAM-DG-TFSI-Cl closely resembled those of EG-PAM-DG-Cl, indicating similar behavior of the anion (Cl-) in these two hydrogel electrolytes. Upon cooling to 5 °C, the ionic conductivity of the EG-PAM-DG-TFSI-Cl gradually diverged from that of the EG-PAM-DG-Cl, but approached that of the EG-PAM-DG-TFSI. Activating TFSI- intercalation into the G positive electrode. The ionic conductivities at varied temperatures. a > 0 °C. b <0 °C. Noted: The red dashed circle highlights the condition under which, at a specific temperature, the corresponding y‑axis values of two different samples become more closely aligned. c–e CV profiles of PTCDI | | G batteries assembled by various hydrogel electrolytes at different temperatures. Noted: I represents the specific current. (c) 25 °C. (d) -20 °C. (e) −40 °C. Charge-discharge profiles of PTCDI||G batteries assembled by various hydrogel electrolytes at different temperatures. f 25 °C. g -20 °C. h −40 °C. Interestingly, at -20 °C, the ionic conductivity of the EG-PAM-DG-TFSI-Cl closely resembled that of the EG-PAM-DG-TFSI (Figs. 3b, S7, and Table S4), indicating similar behavior of the anion (TFSI-) in them. Therefore, it is reasonable to assume that the lower the temperature, the easier it is for TFSI- to insert into the G positive electrode due to easier desolvation of hydrated TFSI-. Furthermore, the activation energy of the EG-PAM-DG incorporating various salts, calculated using the Arrhenius equation at temperatures above 0 °C, was lower than that obtained at temperatures below 0 °C, confirming the increased inertness of water molecules in these hydrogel electrolytes (Fig. 3a, b). Of note, the mS/cm-scale ionic conductivity exhibited by the EG-PAM-DG-TFSI-Cl hydrogel electrolyte (Tables S3 and S4) is higher than that of the frozen aqueous electrolytes on the order of μS/cm, which might be attributed to the restricted ion mobility within the frozen electrolyte’s lattice structure. Subsequently, we conducted in situ cyclic voltammetry (CV) tests at 5 mV/s to verify the aforementioned hypothesis. Initially, we recorded CV profiles of PTCDI | | G batteries assembled using EG-PAM-DG-TFSI, EG-PAM-DG-Cl, and EG-PAM-DG-TFSI-Cl at 25 °C, respectively (Fig. 3c). Clearly, the PTCDI | | G battery with EG-PAM-DG-TFSI showed no redox peaks in the corresponding CV profile. Notably, the PTCDI | | G battery with EG-PAM-DG-Cl exhibited a reduction peak at approximately 1.8 V, attributed to the reduction reaction of Cl2 to Cl- and an obvious chlorine evolution reaction (ClER) was also observed. However, the PTCDI | | G battery with EG-PAM-DG-TFSI-Cl displayed a higher reduction peak at 1.9 V, corresponding to the two reduction reactions of Cl2 to Cl- as well as TFSI- intercalation into the G positive electrode. Upon cooling to -20 °C, the PTCDI | | G battery with EG-PAM-DG-TFSI still showed no redox peaks in the corresponding CV profile (Fig. 3d). Interestingly, the PTCDI | | G battery with EG-PAM-DG-TFSI-Cl demonstrated a lower reduction peak compared to the reduction peak of Cl2 to Cl- achieved by the PTCDI | | G battery with EG-PAM-DG-Cl, attributed to the voltage drop at low temperatures due to increased internal resistance and slow kinetics. It is important to note that Cl2 remains a gas at −20 °C, resulting in high activity and a similar reduction peak to that at 25 °C. Upon cooling to −40 °C, the reduction peak of the PTCDI | | G battery with EG-PAM-DG-TFSI-Cl further decreased to 1.7 V, while the reduction peak of the PTCDI | | G battery with EG-PAM-DG-Cl hardly exhibited the voltage drop due to the high activity of Cl2 at −40 °C (Fig. 3e)22,23. Of note, the poorly defined oxidation plateaus can be attributed to the multistage intercalation process of TFSI- and the inherent high operational voltage, coupled with trace amounts of gas generation during charging, which can disrupt the interfacial process6. Notably, an analogous phenomenon has also been reported in organic graphite-based dual-ion batteries35. Finally, galvanostatic charge-discharge (GCD) curves of PTCDI | | G batteries assembled by EG-PAM-DG with various salts were also recorded to verify the aforementioned assumption. As shown in Fig. 3f, at 25 °C, PTCDI | | G batteries assembled by EG-PAM-DG-Cl and EG-PAM-DG-TFSI-Cl exhibited an obvious ClER when the voltage was swept to 2.2 V, and the discharge plateaus were consistent with the CV results (Fig. 3c). At -20 °C, when sweeping to 2.5 V, PTCDI | | G batteries assembled employing EG-PAM-DG-TFSI-Cl exhibited TFSI- intercalation into the G positive electrode but were unable to reach 2.5 V (only to 2.2 V) due to the ClER of the PTCDI | | G battery assembled using EG-PAM-DG-Cl (Fig. 3g). The discharge plateaus of these two ADIBs were also consistent with the CV results (Fig. 3d). Of note, when high-voltage scanning to 2.3 V, PTCDI | | G battery assembled by EG-PAM-DG-TFSI-Cl exhibited no TFSI- intercalation into the G positive electrode as well as the redox of Cl2/Cl- (Figure S8), implying that high-voltage scanning is a necessary condition for activating TFSI- anions intercalation chemistry. This can be attributed to the thorough desolvation of hydrated TFSI- under a strong electric field, which facilitates more efficient adsorption and subsequent intercalation into the G positive electrode, ultimately yielding a higher discharge capacity and an elevated discharge plateau, which is consistent with the conversion-type electrolytic aqueous Zn | |MnO₂ battery (Fig. S9)36. Upon cooling to -40 °C, when sweeping to 3.0 V, the PTCDI | | G battery assembled by EG-PAM-DG-TFSI-Cl showed TFSI- intercalation into the G positive electrode but was unable to reach 3.0 V (only to 2.5 V) due to the ClER of the PTCDI | | G battery assembled using EG-PAM-DG-Cl (Fig. 3h). The discharge plateau of the ADIB with EG-PAM-DG-TFSI-Cl was 1.7 V, while that of the ADIB with EG-PAM-DG-Cl was 1.9 V, thus confirming the TFSI- intercalation into the G positive electrode at -40 °C instead of the redox reaction of Cl2/Cl- in the PTCDI | | G battery assembled by EG-PAM-DG-TFSI-Cl. Of note, the liquid state of Cl₂ at -40 °C and the host property of graphite prevent the loss of Cl₂ from the system, enabling its participation in the electrochemical reaction22. In brief, ClER occurs preferentially over TFSI intercalation during charging at 25 °C, provided that Cl- is present in the electrolyte. In contrast, at −20 °C, TFSI- intercalation occurs preferentially over ClER. The TFSI- deintercalation shows a discharge plateau at 1.9 V, close to the reduction potential of Cl2/Cl-. At −40 °C, the successful deintercalation of TFSI- from G shows a discharge plateau of 1.7 V due to the poorer kinetics, but the reduction potential of Cl2/Cl- is still at 1.9 V. This difference is because of the less negative effects of low temperature on the redox of Cl-/Cl2, considering the gaseous nature of Cl2 at > −35 °C. That is, at −40 °C, we observe only a 1.7 V discharge plateau, indicating that the deintercalation reaction of TFSI- from the G positive electrode occurs. The study of TFSI- insertion behavior into the G positive electrode using the EG-PAM-DG-TFSI-Cl was conducted through ex situ Raman spectroscopy of the G positive electrode in PTCDI | | G batteries at −20 °C. Figure 4a illustrates a typical GCD profile of a PTCDI | | G battery operating at −20 °C, coupled with ex situ Raman spectra of the G positive electrode recorded at various voltage stages. The pristine electrode of graphite displayed a characteristic G band near 1580 cm−1, corresponding to the stretch vibration of carbon-carbon bonds, indicative of the degree of graphitization6,37. The G band displayed a blueshift as TFSI- continued to intercalate during the charging process, ultimately reaching near 1650 cm−1 once charged to 2.50 V. This shift demonstrates that TFSI- anions intercalated changed the chemical surroundings of the adjacent G layers, aligning with previous reports6,38. During the discharging process, the Raman signal peaks exhibited a redshift. As expected, they reverted to the initial position of G once the discharge is complete, further confirming the high reversibility of TFSI- (de)intercalation in the G positive electrode (Fig. 4b). Ex situ X-ray diffraction (XRD) patterns of the G positive electrode experiencing the charge/discharge process in PTCDI | | G batteries at 0.1 mA cm-2 were also obtained. When charging, the staging mechanism of TFSI- insertion into G was evidenced according to the splitting of the (002) reflection, indicating the generation of graphite intercalation compounds6. During the discharging process, the split peaks gradually went back to the initial position of 26.4° as the anions were steadily extracted from the G, showcasing the high reversibility of anion (de)intercalation (Fig. 4c)6. Mechanism analysis. a GCD profile of PTCDI | | G battery with EG-PAM-DG-TFSI-Cl electrolyte working at -20 °C at 0.1 mA cm−2. Ex situ Raman spectra (b) and ex situ XRD pattern (c) of G in PTCDI | | G battery at different charging/discharging states. Noted: Ex situ measurements were recorded at specific voltages during charge (C) and discharge (D), represented by the blue and red data points, respectively, in (a). Noted: au arbitrary units. The testing conditions: the testing cell will cycle for 90 cycles at 91 mA g−1, and then be disassembled in air upon reaching the designated voltages. The harvested electrode samples were collected at 25 °C ± 2 °C, cleaned with deionized (DI) water, and processed without the use of an inert atmosphere. G loading mass: 1.1 mg cm−2. PTCDI loading mass: 1.5 mg cm−2. d The schematic illustration of the intercalation process of TFSI- insertion into the G positive electrode. Noted: oxygen atom (red sphere), hydrogen atom (white sphere), carbon atom (brown sphere), fluorine atom (orange sphere), sulfur atom (purple sphere), nitrogen atom (light pink sphere). Based on the mechanism evolution observed in the PTCDI | | G batteries, we summarize two necessary conditions in electrolyte engineering for achieving TFSI- intercalation into the G positive electrode: the expanded ESWs for high-voltage scanning and easy desolvation of anions (Fig. 4d). Noted that, the intercalation of anions into graphite can be divided into four distinct stages (Stage 0 to Stage 3)6,39. Specifically, Stage 0 represents the edge-plane adsorption state prior to intercalation, while Stages 1-3 correspond to the sequential insertion of anions into every four, two, and one graphite layers, respectively6,39. Of note, the high output voltages of aqueous graphite-based dual-ion batteries are governed by both the anion intercalation energy and the hydration (solvation) energy (Fig. S10)6,40. Therefore, under conditions of low concentration of TFSI- anions, the introduced DM in the EG-PAM-DG-TFSI-Cl can simultaneously meet these two demands. Additionally, the intrinsic anti-icing property of EG helps the EG-PAM-DG-TFSI-Cl operate at low temperatures. Electrochemical performance By exploring the electrochemical performance of aqueous PTCDI | | G batteries at low temperatures, we initially recorded the cycle performances of aqueous PTCDI | | G batteries at −20 °C. As shown in Figure S11, the specific capacity of 200 mAh g−1 was observed in the PTCDI | | G battery with 1 mg cm−2 of G, running for approximately 240 hours, which was consistent with the electrochemical result of the PTCDI | | G battery with a loading mass of 3 mg cm−2 of G (Figure S12). Of note, after cycling at −20 °C, the battery subsequently runs at 25 °C ± 2 °C and shows a 2.2 V discharge plateau (Fig. S13). Subsequently, the rate performance of PTCDI | | G in the EG-PAM-DG-TFSI-Cl at -40 °C is shown in Fig. 5a, b, delivering specific capacities of 350, 250, 160, 120, and 82 mAh g−1 at specific currents of 60, 70, 80, 90, and 100 mA g−1, respectively. These relatively large changes in specific capacity are attributed to the low ionic conductivity (at the level of 10−4 S cm−1) at low temperatures (−40 °C). Of note, it can be attributed to three factors: thermodynamic more inert, the decreased diffusion coefficient, and interface process deterioration (Figure S14)41–43. Noted that, the capacity degradation observed in the PTCDI | | G cell originates primarily from irreversible G expansion/exfoliation based on anion intercalation and electrolyte depletion (Fig. S15)39,44. Therefore, we selected a suitable specific current of 80 mA g−1, based on specific capacity, to explore the cycle performance of aqueous PTCDI | | G batteries. As shown in Fig. 5c, the GCD profiles of various cycles, from the 1st to the 30th, demonstrate a stable discharge plateau of 1.7 V, with a gradually increasing capacity to 200 mAh g−1. Even after cycling for over 470 hours, the PTCDI | | G battery still delivered a specific capacity of more than 100 mAh g−1 (Fig. 5d). Of note, the activation process (e.g., low desolvation activation energy) based on the staging mechanism is a key factor contributing to the observed increase in capacity during the initial stage of battery operation (Fig. S10)39,44. The Coulombic efficiency (CE) of the PTCDI | | G battery in the EG-PAM-DG-TFSI-Cl is around 95% upon long-term cycling. It should be noted that the inferior CE is a common problem of G-based ADIBs. It is primarily attributed to the high voltage of G-based ADIBs and trace amounts of gas production6. Of note, the multistage intercalation mechanism, the extended activation process, and the use of a Ti current collector (Fig. S16) collectively reduce the material’s reaction efficiency during the initial cycles, thereby lowering the CE6,39. The electrochemical behaviors of PTCDI | | G batteries at -40 °C. Discharge curves of (a) PTCDI | | G batteries at different specific current working at −40 °C and (b) the corresponding rate performance. Mass loading of G is 1 mg cm−2. The detailed cell parameters used can be obtained in Table S5. GCD profiles (c) and the corresponding cycling performance (d) at 80 mA g−1. Noted: Mass loading of G is 1.2 mg cm−2. The detailed cell parameters used can be obtained in Table S5. e, The Comparison of the specific energy of this work and other aqueous dual-ion batteries (the calculation is based on the 1 mg cm−2 of G mass loading in this work). Note: Fc/C and LTO are the abbreviations of the compound of ferrocene into the nanopores of activated carbon and Li4Ti5O12, respectively. Noted: The source of the literature data shown in this figure can be found in Supplementary Information, Table S6. f Comparison of aqueous dual-ion batteries (based on anion intercalation G electrode) in current electrolyte systems in terms of salt concentration, high specific energy, low cost, and low temperature. Note: “Water in Salts” (“WIS”) electrolytes belong to highly concentrated salt electrolytes that are the precondition of anion intercalation chemistry for achieving high potential (e.g., TFSI- intercalation). Non-TFSI- intercalation electrolytes represent electrolytes consisting of other anions (e.g., [ZnCl4]2-, NO3-, etc.), which cannot support the high potential of anion intercalation chemistry. Anion intercalation chemistry in aqueous batteries is difficult to achieve and is rarely reported. Noted: The source of the literature data shown in this figure can be found in Supplementary Information, Table S6. Most of the discharge plateaus of ADIBs were lower than 1.5 V due to the narrow ESW of aqueous electrolytes. After the concept of WIS was proposed, many high-voltage aqueous batteries, including ADIBs, have been explored. However, these WIS-based ADIBs, especially 30 M ZnCl216, 21 M LiTFSI17, et al., cannot work in harsh environments, for example, at −40 °C, and also have high costs. Here, compared with the specific energies of various WIS-based ADIBs achieved at 25 °C ± 2 °C, our developed PTCDI | | G with 1 M LiTFSI + 2 M ZnCl2 delivered the higher specific energy of 595 Wh kg-1 (normalized to the 1 mg cm−2 of the G positive electrode) even at −40 °C, as exhibited in Fig. 5e2,6,9,10,45. Overall, in terms of salt concentration, cost, specific energy, and the ability to tolerate low temperatures, the PTCDI | | G battery based on the EG-PAM-DG-TFSI-Cl exhibits the potential for practical application compared to TFSI- and non-TFSI- intercalation electrolyte-based aqueous PTCDI | | G batteries (Fig. 5f)2,6,9,10,45. The development of aqueous electrolytes with wide ESWs and the advancement of high-capacity, robust G-positive electrode materials represent critical and convergent research frontiers for the future of ADIBs. Discussion Our work studies a low-cost DM electrolyte with an expanded ESW of >4.6 V at −40 °C (>4.2 V at −20 °C). This advancement has enabled the intercalation of TFSI- into the G positive electrode of the assembled ADIBs by high-voltage scanning to 3.0 V at −40 °C (2.5 V at -20 °C), despite employing a low concentration of TFSI- that was typically insufficient for activating TFSI- intercalation. The resulting ADIBs, based on the effective intercalation of TFSI-, exhibited a high discharge plateau of 1.9 V at −20 °C and 1.7 V at −40 °C, as well as a large specific capacity of approximately 200 mAh g−1. These results can be attributed to three aspects: (1) the expanded ESW of EG-PAM-DG-TFSI-Cl at low temperatures, which aids in activating TFSI- insertion into the G positive electrode by high-voltage scanning; (2) the intrinsic anti-freezing property that promotes the desolvation of the hydrated TFSI- anion at low temperatures by enhancing confinement to water molecules; (3) the large number of -OH groups resulting from the precursors of DG and EG makes the intercalation of low concentration TFSI- anions into G positive electrode. This work provides valuable insights into the design of high-voltage ADIBs with low-cost operations at low temperatures. Methods Materials Acrylamide (AM, AR, 99%), N,N’-Methylenebisacrylamide (MBAA, 99%), D(+)-Glucose monohydrate (DG, 98%), photoinitiator 2959 (IRGACURE 2959, 98%), ethylene glycol (EG, 98%), and lithium bistrifluoromethanesulfonimidate (LiTFSI, 99%) were purchased from Aladdin and used without further purification. Zinc chloride (ZnCl2, anhydrous, 99.95%) was purchased from Thermos Scientific and used directly. 3,4,9,10-perylenetetracarboxylic diimide (PTCDI, 99%) was purchased from MACKLIN and used directly. High-purity flake graphite (G, 1000 mesh, 99.5 wt%) was purchased from XFNANO. Deionized (DI) water was used in all the experiments. Synthesis of EG-PAM-DG-TFSI-Cl and PAM hydrogel electrolyte The DM-rich hydrogel electrolyte was created using a one-step atomic radical polymerization process11. To begin, 0.01 mol of DG, 0.07 mol of AM, and 0.1 g of IRGACURE 2959 were dissolved in 20 mL of a mixed solution containing 40% EG and 60% (1 M LiTFSI and 2 M ZnCl2) water solution (volume ratio). Of note, prior to fabricating the hydrogel electrolytes, an aqueous electrolyte consisting of 1 M LiTFSI and 2 M ZnCl₂ was prepared. Specifically, 28.71 g of LiTFSI and 27.26 g of ZnCl₂ were dissolved in 100 mL of DI water to obtain this mixture. The resulting aqueous solution was subsequently mixed with EG at a volume ratio of 60%:40%. After stirring the mixture for 10 minutes, it was transferred to a customized polytetrafluoroethylene mold with dimensions of 1 cm × 1 cm × 0.3 cm and exposed to UV light (365 nm, 400 mJ cm−2) for 40 minutes. This process led to the synthesis of the DM-rich hydrogel electrolyte, denoted as EG-PAM-DG-TFSI-Cl. Additionally, the mass percent composition of the EG-PAM-DG-TFSI-Cl, calculated based on the charge ratio. The water content of the hydrogel electrolyte was calculated following the equation below: Where λ represents water content of hydrogel electrolyte (wt%), Ww represents the weight of water in the hydrogel electrolyte, and Wt represents total weight of the hydrogel electrolyte. The detailed calculation results are shown in Table S1. In a typical synthesis, the polyacrylamide (PAM) hydrogel electrolyte was prepared by dissolving 3.2 g of AM, 0.007 g of MBAA, and 0.1 g of IRGACURE 2959 in 20 mL of a (1 M LiTFSI and 2 M ZnCl₂) mixed solution. The mixture was stirred for 10 min, then transferred to a custom mold and polymerized under UV light (365 nm, 400 mJ cm⁻², 1000 L, UVP) for 1 h to form the PAM hydrogel electrolyte. Of note, hydrogel electrolytes typically possess three-dimensional network structures that swell but do not dissolve, which generally precludes the determination of their relative molecular weight. Preparation of the G-positive electrode and the PTCDI negative electrode In the fabrication of the G positive electrode, a mixture of G powder, Ketjenblack (KB, EC300J, Canrd Co.), and polyvinylidene fluoride (PVDF, 900, Arkema) in a weight ratio of 7:2:1 was vigorously blended and ground using Quartz Mortar. Subsequently, N-methyl pyrrolidone (NMP, 99%, Aladdin) solvent was added to the mixture, and a homogeneous slurry was obtained through stirring. The slurry was then carefully coated onto a titanium (Ti) mesh (80 mesh, >99.99%, 100 μm, Kangdeli Metal) using a doctor blade and dried in an electric constant temperature drying oven (DHG-9030A, Shanghai Jinghong Experimental Equipment Co., Ltd.) at 60 °C for 10 hours to fabricate the G positive electrode. Similarly, the PTCDI negative electrode can be successfully fabricated using the same process, apart from Ti foil as a substrate. Of note, Ti mesh was used directly without any prior processing, and electrodes were generally cut with scissors unless otherwise specified. Calculation of the costs of EG-PAM-DG-TFSI-Cl The costs of the EG-PAM-DG-TFSI-Cl and typical WIS electrolytes are summarized, and the detailed calculation process for these electrolytes is presented below. 1). The unit price of pure hydrogel (without salts) can be calculated using the following equation: Where P1 represents the unit price of pure hydrogel, Pt represents the total price of all components of the hydrogel and Wt represents total weight of the hydrogel electrolyte. 2). The unit price of hydrogel electrolyte (incorporating salts) can be calculated following the equation: Where P2 represents the unit price of hydrogel electrolyte, and Ps represents the total price of all salts. The detailed values can be found in Table S2. Electrochemical measurements The cyclic voltammetry (CV) and linear sweep voltammetry (LSV) profiles, as well as the AC impedance spectra, were recorded using an electrochemical workstation (CHI 760e). The rate and cycle performance were evaluated using the LAND test system. All tests were conducted using pouch cells. It is noted that pouch cells were used for electrochemical measurements under one standard atmosphere, unless otherwise specified. Specifically, pouch cells were assembled in a single-layer configuration by stacking a G positive electrode, a hydrogel electrolyte, and a PTCDI negative electrode. A Ti mesh (80 mesh, >99.99%, 100 μm, Kangdeli Metal) served as the current collector for both electrodes, corresponding to a negative-to-positive capacity ratio (N/P) of 1.5. The specific energy at −40 °C, based on the positive electrode, was determined by integrating the discharge potential (e.g., 1.7 V vs PTCDI at 80 mA g⁻¹) and subsequently normalized by the mass of the active material (e.g., 160 mAh g⁻¹ at 80 mA g⁻¹). A thermoplastic polyurethane hot-melt adhesive film (double-sided, Shanghai Qindan Industrial Co., Ltd.) was used as the sealing material. The hydrogel electrolytes served as separators without the need for electrolyte injection or gas release. The dimensions of the hydrogel electrolytes were 1 cm × 1 cm × (0.2–0.3) cm. The areal mass loading of the positive electrodes was maintained at 1–3 mg cm⁻². It is worth noting that after the pouch cells were assembled at 25 °C ± 2 °C under one standard atmospheric pressure, they were allowed to rest for 1 hour. Subsequently, they were placed directly into a programmable temperature and humidity test chamber (YH-36R, Dongguan Tuode Environmental Testing Equipment Co., Ltd.) for testing. Electrochemical impedance spectroscopy (EIS) measurements were performed using pouch cells assembled in a single-layer configuration. The stack consisted of a Ti foil positive electrode (>99.99%, 100 μm, Kangdeli Metal), a hydrogel electrolyte, and a Ti foil negative electrode (>99.99%, 100 μm, Kangdeli Metal). The amplitude of the applied signal was 0.005 V, and the frequency range was 0.01 Hz to 10⁴ Hz. The measurements were conducted in potentiostatic mode, with 12 points recorded per decade of frequency. The open-circuit voltage prior to the EIS measurement was maintained at approximately 0 V. Of note, the ionic conductivities can be calculated according to the equation as below: where d is the thickness of the hydrogel electrolyte, A is the area of the hydrogel electrolyte, and R is the impedance obtained by the EIS spectra. Characterizations Fourier-transform infrared spectroscopy (FTIR, Nicolet iN10, Thermo Fisher Scientific) was employed to characterize the structural properties of the hydrogel electrolytes and the state of water molecules within them. The spectra were recorded at a resolution of 4 cm⁻¹ with 32 scans over a wavenumber range of 400–4000 cm⁻¹. Raman spectroscopy (532 nm laser, WiTec RAMAN alpha 300 R) was also conducted in the range of 40–4000 cm⁻¹. Differential scanning calorimetry (DSC 25, TA Instruments) was used to measure the enthalpy changes of the hydrogel electrolytes as a function of temperature, aiming to investigate the state of water molecules within the samples. The DSC measurements were performed using the following thermal program: first, the sample was cooled from 20 °C to –50 °C at a rate of 10 °C/min and held at –50 °C for 3 min. Subsequently, it was heated to 20 °C at a rate of 10 °C/min and maintained at 20 °C for approximately 3 min. This cycle was repeated twice. Ex situ X-ray diffraction (XRD, D2 PHASER) with Cu Kα radiation (λ = 1.5406 Å, 40 kV, 40 mA) was performed to determine the crystal structure of the G electrode during cycling, with patterns recorded over the 2θ range of 20–35°. Additionally, ex situ Raman spectroscopy (532 nm laser, range 1500–1700 cm⁻¹, Horiba LabRam HR Evolution) was employed to characterize the state of the G positive electrode following TFSI⁻ intercalation during cycling. For the ex situ characterization, the aqueous cells were disassembled in air upon reaching the designated voltages. The harvested electrode samples were collected at 25 °C ± 2 °C, cleaned with deionized (DI) water, and processed without the use of an inert atmosphere. Supplementary information Supplementary Information Transparent Peer Review file Source data Source Data Author contributions C. L. and C. Z. conceived the idea, designed and performed the experiments, and wrote the manuscript. X. M., D. L., R. Z., and H. C. contributed to data analysis and discussion. S. Z., Z. H., and Q. N. assisted with data interpretation. H. Y. and C. Z. supervised the project and revised the manuscript. All authors discussed the results and approved the final manuscript. Peer review Peer review information Nature Communications thanks Yue Gao, Minghao Yu, and Qiang Zhang for their contribution to the peer review of this work. A peer review file is available. Funding This work is supported by a grant from the Research Grants Council of the Hong Kong Special Administrative Region, China [Project No. CityU 11314822]. Data availability All the data generated in this study are provided in the Supplementary Information/Source Data file. Source data are provided with this paper. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Hui Yang, Email: yanghui17@chinatowercom.cn. Chunyi Zhi, Email: cyzhi@hku.hk. Supplementary information The online version contains supplementary material available at 10.1038/s41467-026-74628-2. References Associated Data Supplementary Materials Supplementary Information Transparent Peer Review file Source Data Data Availability Statement All the data generated in this study are provided in the Supplementary Information/Source Data file. Source data are provided with this paper.