A certain amount of lithium bis-trifluoromethanesulfonimide (LiTFSI, J&K^®) was first dissolved in deionized water to obtain a 21 m LiTFSI aqueous electrolyte, where m means molar salts dissolved in per kilogram solvent. Then, the corresponding mass of anhydrous acetonitrile (J&K^®) was added to dilute the 21 m LiTFSI aqueous electrolyte into certain concentrations (e.g., 1, 3, and 5 m). Finally, the zinc trifluoromethanesulfonate (Zn(CF₃SO₃)₂, J&K^®) was added into the diluted electrolyte to reach 1 m based on the total mass of water and acetonitrile, and the water-acetonitrile electrolytes were obtained.

The aMEGO powder was synthesized following the steps reported previously [43]. Briefly, graphite oxide (GO) powder obtained by modified Hummers’ method was reduced in a microwave oven with a power of 800 W for 1 min, to prepare the microwave-exfoliated graphene oxide (MEGO) powder. Then, the MEGO powder was mixed with KOH aqueous solution and heated to evaporate water to get a uniform KOH/MEGO mixture with a mass ratio of 7:1. The mixture was further annealed in Ar at 800 °C for 2 h and naturally cooled down to room temperature. After sequentially washing with dilute HCl solution, deionized water, and ethanol and dried in the oven at 60 °C for 6 h, the aMEGO powder was obtained.

The morphology of electrodes before and after cycling was investigated by a field-emission scanning electron microscope (Philips, FEG-XL30) with an accelerating voltage of 5 kV. The structure analysis was performed by X-ray diffraction (XRD) on Bruker D2 system with Cu K_α irradiation (λ = 1.5418 Å, tube voltage: 30 kV and tube current: 10 mA). The Raman spectra of electrolytes were taken on a Renishaw^® inVia system with a 514 nm laser and 5 × objective. The ¹H nuclear magnetic resonance (NMR) spectra were collected on a Bruker Advance III 400 NMR system, and a co-axial glass tube was used to conduct the experiments with CDCl₃ (99.8 atom% D, contains 0.03% v/v TMS; J&K) as magnetic field locker. The chemical shifts were calibrated with the peak of tetramethylsilane in CDCl₃ as 0 ppm.

The Tafel plots were measured based on a three-electrode system with a scan rate of 10 mV s⁻¹, in which the Pt foil, Zn foil, and Ag/AgCl electrode were used as the counter electrode, working electrode, and reference electrode, respectively. The galvanic corrosion of the Zn foil was evaluated based on a Zn-Ti galvanic cell in a CR2032 coin cell. The Coulombic efficiency of Zn stripping/plating was measured based on a CR2032 coin cell, in which Zn foil (50 μm), Cu foil (20 μm) and glass fiber (Whatman^®) were used as anode, cathode and separator, respectively. The cell was initially discharged for a certain time with a specific current density and then charged to 0.5 V at the same current density for each cycle on a Neware^® battery testing system. To prepare cathode electrode for the evaluation of Zn-ion hybrid supercapacitors, aMEGO powder was added into polytetrafluoroethylene (PTFE, 60 wt% aqueous dispersion, commercial purchased from Aladdin^®) slurry binder at a mass ratio of 5:95, then rolled into a thin film with a thickness around 60-70 μm. Then, the film was punched into disks with a diameter of 10 mm, pressed into a Ti mesh (14 mm diameter) by hydraulic press at 20 MPa for 30 s and dried in a vacuum oven at 100 °C for 6 h. The hybrid supercapacitors were assembled into CR2032 coin-type cells with 16 mm Zn foil, aMEGO and glass fiber (Whatman^®) as anode, cathode and separator, respectively. The electrochemical performance was evaluated by cyclic voltammetry and galvanostatic charge-discharge techniques on an electrochemical workstation and Neware^® battery testing system, respectively. The specific gravimetric capacitance was calculated based on the equation: C =  It/mU, where I is the charge/discharge current, t is the discharge time, m is the mass of the aMEGO, and U the is voltage difference of the discharge curve. The energy density was calculated using the equation: E =  1/2CV², where C is the specific gravimetric capacitance calculated above, and V is the operation voltage of the hybrid supercapacitors. The power density was calculated by the equation: P =  E/t, where E is the energy density and t is the corresponding discharge time [44]. The electrochemical impedance spectroscopy was conducted on an electrochemical workstation with a frequency range from 100 kHz to 100 mHz and an amplitude of 10 mV.

To investigate the electronic structures of different electrolytes, density functional theory (DFT) within the CASTEP package has been applied for all calculations in this work. We have chosen the generalized gradient approximation (GGA) with Perdew-Burke-Ernzerhof (PBE) functions for the descriptions of exchange-correlation interactions. The plane-wave cutoff energy has been set to 380 eV based on ultrafine quality and the ultrasoft pseudopotential scheme. The coarse k-point has been applied for the energy minimization based on the Broyden-Fletcher-Goldfarb-Shannon (BFGS) algorithm. The Hellmann-Feynman forces should be converged to less than 0.001 eV Å⁻¹ and the total energy should be less than 5 × 10^-5 eV atom⁻¹. The convergence criteria maximum stress and atomic displacement have been set to 0.2 GPa and 0.005 Å, respectively.

The molecular dynamic (MD) simulations have been conducted in four electrolytes including 0.5 m + 1 m-H₂O electrolyte, 0.5 m + 1 m-H₂O/AN electrolyte, 0.5 m + 3 m-H₂O/AN electrolyte, and 0.5 m + 5 m-H₂O/AN electrolyte. The molar ratio of these three electrolyte models are Zn (CF₃SO₃)₂: LiTFSI: H₂O = 0.5: 1: 55.6; Zn (CF₃SO₃)₂: LiTFSI: H₂O: AN = 0.5: 1: 2.6: 23.2; Zn(CF₃SO₃)₂: LiTFSI: H₂O: AN = 0.5: 3: 7.9: 20.9; and Zn(CF₃SO₃)₂: LiTFSI: H₂O: AN = 0.5: 5: 13.2: 18.6, respectively. The MD has been carried out under the NVT condition in 298 K. The time step is 1 fs and the total simulation time has been set to 3 ps. We have selected the Noise scheme for the thermostat. After the MD simulations, we further carry out geometry optimizations to investigate the electronic structures and the interaction energies.

The electrochemical performance of Zn metal anodes was evaluated in the water/acetonitrile (H₂O/AN) electrolytes, and the 0.5 m Zn(CF₃SO₃)₂ aqueous electrolyte (denoted as 0.5 m-H₂O) and 0.5 m Zn(CF₃SO₃)₂ + 1 m LiTFSI aqueous electrolyte (denoted as 0.5 m + 1 m-H₂O) as references. From the Tafel plots of Zn metal in the H₂O/AN electrolytes (Figs. 1a and S1), a sharp peak was observed at potentials of -0.77, -0.82, and -0.83 V (vs. Ag/AgCl electrode) in the 0.5 m + 1 m-H₂O/AN, 0.5 m + 3 m-H₂O/AN, and 0.5 m + 5 m-H₂O/AN electrolytes, respectively. By contrast, the peaks at potentials of − 0.99 and − 0.94 V were detected in the 0.5 m-H₂O and 0.5 m + 1 m-H₂O electrolytes, respectively. In general, the peak potential represents the initial potential of Zn plating/stripping processes, and a higher potential is favorable to avoid the hydrogen evolution which is a competitive process accompanied by Zn plating [18,45]. Thus, the higher potentials of Zn plating/stripping processes in the H₂O/AN electrolytes indicate an effectively suppressed hydrogen evolution reaction (HER), which is favorable for improving the Coulombic efficiency of Zn plating/stripping.

Moreover, a galvanic cell was built to investigate the thermodynamic stability of Zn metal anode in the H₂O/AN electrolytes. From the time-dependent corrosion current curves (Figs. 1b and S2), the current densities of the galvanic cells are initially large (~ 40 μA cm⁻²) which are partially ascribed to the capacitive contribution. Then the current densities are quickly dropped by one order of magnitude to µA cm⁻² level. Specifically, the current densities of the galvanic cells are stabilized at 3-5 µA cm⁻² in the H₂O/AN electrolytes. By contrast, the current densities of the galvanic cells fluctuate around 20 µA cm⁻² in the reference electrolytes. The much reduced corrosion currents imply that the galvanic corrosion is drastically reduced in the H₂O/AN electrolytes. Correspondingly, the SEM images (Fig. S3) of the Zn foils after galvanic corrosion maintain a smooth surface in the H₂O/AN electrolytes, whereas severely corroded surfaces are observed in the reference electrolytes. Moreover, the SEM images (Fig. S4) and XRD patterns (Fig. S5) of the Zn foils after soaking in the H₂O/AN and reference electrolytes also demonstrate the anti-corrosion capability of the H₂O/AN electrolytes. These results together reveal that the Zn metal is thermodynamically more stable in the H₂O/AN electrolytes.

The Coulombic efficiency of Zn plating/stripping was further tested based on a Zn-Cu coin cell with the H₂O/AN electrolytes. As shown in Fig. S6, the cells with H₂O/AN electrolytes show stable and significantly higher Coulombic efficiency (average > 99%) after 100 cycles as expected, while the cells with aqueous electrolyte (0.5 m + 1 m-H₂O) demonstrate gradually decreasing and fluctuating Coulombic efficiency (from 90 to 70% after 50 cycles) under the same condition. Moreover, for the H₂O/AN electrolytes with different concentrations of LiTFSI (1-5 m), the initial Coulombic efficiency decreases with the increase of LiTFSI concentration, and the average Coulombic efficiency of Zn plating/stripping is the highest in 0.5 m + 1 m-H₂O/AN electrolyte. From the SEM images of Zn foils in the Zn-Cu cells after specific cycles (Fig. S7), a great number of macro-pores are observed on the Zn foils after the 10th, 20th, and 50th cycles in the aqueous electrolytes (0.5 m-H₂O and 0.5 m + 1 m-H₂O), which indicates an uneven Zn plating/stripping behavior and severe Zn corrosion during cycling. Nevertheless, a compact and uniform surface is revealed on the Zn foils cycled in 0.5 m + 1 m-H₂O/AN electrolyte, confirming an excellent Zn plating/stripping reversibility in 0.5 m + 1 m-H₂O/AN electrolyte. At enlarged current densities up to 10 mA cm⁻², a high Coulombic efficiency (97.3% on average) is still maintained in 0.5 m + 1 m-H₂O/AN electrolyte, in contrast to its rapid decrease and fluctuation in the aqueous electrolyte (0.5 m + 1 m-H₂O), as shown in Figs. 1c and S8. However, for the H₂O/AN electrolytes with further increased LiTFSI concentrations (0.5 m + 3 m-H₂O/AN and 0.5 m + 5 m-H₂O/AN), a decreased Coulombic efficiency and increased voltage hysteresis are observed at higher current densities (e.g., > 5 mA cm⁻², Fig. S9), indicating that an excessive concentration of LiTFSI is unfavorable to the reversibility of Zn plating/stripping at high current densities. Correspondingly, the symmetric Zn-Zn cells with 0.5 m + 1 m-H₂O/AN electrolyte only show slightly increased voltage hysteresis with enlarged current densities (Fig. 1d), while a dramatically increased voltage hysteresis followed by short-circuit is observed for 0.5 m + 3 m-H₂O/AN and 0.5 m + 5 m-H₂O/AN electrolytes (Fig. S10). After cycling at enlarged current densities (e.g., 5 mA cm⁻²), the Zn foil maintains uniform and compact morphology in 0.5 m + 1 m-H₂O/AN electrolyte, in accordance with a low voltage hysteresis even at high current densities. In contrast, the Zn foils cycled in 0.5 m + 3 m-H₂O/AN and 0.5 m + 5 m-H₂O/AN electrolytes show a rough surface, revealing an uneven plating behavior at increased current densities (Fig. S11). These observations demonstrate that excessive LiTFSI in electrolytes is unfavorable to the transportation of Zn²⁺ ions and deteriorates the rate performance of Zn metal anode, which is in agreement with the larger charge transfer resistances in the Zn-Zn symmetric cells with increased LiTFSI concentrations (Fig. S12). The long-term cyclic stability of the Zn metal anode in the 0.5 m + 1 m-H₂O/AN electrolyte was further verified by the Zn-Cu and Zn-Zn cells. A high average Coulombic efficiency of 97.3% is achieved in the 0.5 m + 1 m-H₂O/AN electrolyte for more than 3,000 cycles (Fig. 1e), and no obvious changes are observed in the charge-discharge curves of specific cycles (the inset of Fig. 1e). Correspondingly, the Zn-Zn cells with the 0.5 m + 1 m-H₂O/AN electrolyte also show stable voltage hysteresis with prolonged cycles, while a short-circuit quickly occurs in the Zn-Zn cells with the reference electrolyte (0.5 m + 1 m-H₂O), as shown in Fig. 1f. These results further verify the superior stability of the Zn metal anode in the 0.5 m + 1 m-H₂O/AN electrolyte.

To assess the practical applicability of H₂O/AN electrolytes, a prototype of a Zn-ion hybrid supercapacitor was fabricated, in which porous graphene (aMEGO) with high specific capacitance and excellent cyclic stability was used as the cathode material [46], and Zn foil and glass fiber were used as anode and separator, respectively. First, the operation voltages of the hybrid supercapacitors were assessed in 0.5 m + 1 m-H₂O/AN electrolyte by the galvanostatic charge-discharge and cyclic voltammetry techniques. As shown in Fig. 2a, the charge-discharge curves of the Zn-aMEGO hybrid supercapacitors maintain linear and symmetric within the voltage up to 2.2 V, and a plateau is observed when the voltage is above 2.2 V. Correspondingly, the cyclic curves (Fig. S13) exhibit a quasi-rectangular shape in the voltage range of 0-2.2 V, and have an obvious current leap at voltages beyond 2.2 V. These results suggest an optimal electrochemical window of 0-2.2 V. In sharp contrast, only a narrow operation voltage range of 0.2-1.8 V can be achieved in the reference electrolyte (0.5 m + 1 m-H₂O) under the same test condition (Fig. S14). In addition, the linear scan voltammetry curves of the Zn-Ti cells also verify the superior cathodic stability of 0.5 m + 1 m-H₂O/AN electrolyte (Fig. S15).

Furthermore, the rate performance of the Zn-aMEGO hybrid supercapacitors with 0.5 m + 1 m-H₂O/AN electrolyte was evaluated within the voltage range of 0-2.2 V. The cyclic voltammetry curves (Fig. 2b) remain a quasi-rectangular shape with little distortion as the scan rates increase from 5 to 100 mV s⁻¹, and a high capacitance retention up to 64% is also achieved as the current density increases by 40 times from 0.5 to 20 A g⁻¹ (Figs. 2c and S16). By contrast, the cyclic voltammetry curves become obviously distorted with decreasing integral areas at the enlarged scan rates in 0.5 m + 3 m-H₂O/AN and 0.5 m + 5 m-H₂O/AN electrolytes (Fig. S17), which indicates an inferior rate performance. These results demonstrate that the Zn-aMEGO hybrid supercapacitors have an excellent rate performance in 0.5 m + 1 m-H₂O/AN electrolyte as summarized in a Ragone plot (Fig. S18), benefiting from its high ionic conductivity (23 mS cm⁻¹) and small charge transfer resistance (Fig. S19). The cyclic stability of the Zn-aMEGO hybrid supercapacitors with 0.5 m + 1 m-H₂O/AN electrolyte was further tested, and a high capacitance retention up to 88% was maintained even after 120,000 cycles with a high operation voltage range of 0-2.2 V (Fig. 2d, e). A clear and smooth surface of the aMEGO electrode was observed from the SEM images (Fig. S20a, b), and no byproducts were detected from the XRD diffraction pattern (Fig. S20c). These results together verify that the 0.5 m + 1 m-H₂O/AN electrolyte has superior electrochemical stability and as well the capability to stabilize Zn metal anodes.

The low temperature tolerance of the H₂O/AN electrolytes was also investigated to assess their applicability under harsh conditions. From Fig. 3a, in 0.5 m + 1 m-H₂O/AN electrolyte, a high Coulombic efficiency and slightly increased voltage hysteresis are observed with ever-increasing current densities based on the Zn-Cu cells at a reduced temperature (− 30 °C). Consistently, the Zn-Zn symmetric cells also demonstrate a slightly increased yet stable voltage hysteresis at enlarged current densities (Fig. 3b). Moreover, a high average Coulombic efficiency (99.4%) and excellent cyclic stability (over 500 cycles) are achieved in 0.5 m + 1 m-H₂O/AN electrolyte at − 30 °C based on the Zn-Cu cells (Fig. 3c). These observations indicate that a high reversibility and rapid kinetics of Zn plating/stripping are retained in 0.5 m + 1 m-H₂O/AN electrolyte even at a harsh temperature condition. As a result, the Zn-aMEGO supercapacitors demonstrate an excellent rate performance within a wide temperature range in the 0.5 m + 1 m-H₂O/AN electrolyte, as verified by the little distortion and slightly decreased integral areas of cyclic voltammetry curves (Fig. 3d). In contrast, in the 0.5 m + 1 m-H₂O electrolyte, the cyclic voltammetry curves become distorted with drastically decreased integral areas at lower testing temperatures (Fig. 3e). The Zn-aMEGO supercapacitors with 0.5 m + 1 m-H₂O/AN electrolyte could stably operate under − 30 °C for more than 2,000 cycles (Fig. 3f). As the concentration of LiTFSI increases, however, inferior tolerance to the low temperatures with obviously distorted cyclic voltammetry curves were observed for 0.5 m + 3 m-H₂O/AN and 0.5 m + 5 m-H₂O/AN electrolytes (Fig. S21). The excellent low temperature performance of the 0.5 m + 1 m-H₂O/AN electrolyte is believed to be associated with its high ionic conductivities at low temperatures (7 mS cm⁻¹ at − 30 °C, Fig. S22).

To understand the electrolyte performances, the solvation structures of H₂O/AN electrolytes were comprehensively studied. From the Raman spectra (Fig. S23), as compared with reference electrolytes (0.5 m-H₂O and 0.5 m + 1 m-H₂O), the intensity of the broad peak (3,100-3,700 cm⁻¹) ascribed to hydrogen bonds among water molecules in the H₂O/AN electrolytes is remarkably decreased, implying that most of the water molecules are solvated. Correspondingly, the Fourier Transform Infrared (FT-IR) spectra (Fig. S24) reveal valuable insights into the molecular interactions in the electrolytes. It is observed that the introduction of acetonitrile significantly reduces the intensity and width of the absorption peaks (2,800-3,700 cm⁻¹) associated with the stretching vibrations of -OH groups in water. This suggests that acetonitrile disrupts the hydrogen bonding among water molecules. Furthermore, nuclear magnetic resonance (NMR) results (Fig. S25) demonstrate that the chemical shifts of the ¹H in water shift to higher fields, indicating an increase in electron density as the concentration of LiTFSI increases. These results suggest that both acetonitrile and LiTFSI exhibit strong interactions with water molecules, thereby expanding the electrochemical window of water.

We also carried out density functional theory (DFT) calculations of H₂O/AN electrolytes with different concentrations through the molecular dynamic (MD) simulations. For the 0.5 m + 1 m-H₂O electrolyte, the bonding orbitals and anti-bonding orbitals near the Fermi level are mostly near the Zn(CF₃SO₃)₂ and TFSI⁻, respectively, where water molecules only limited contributed to the anti-bonding orbitals (Fig. 4a). The relatively weak orbital couplings between Zn(CF₃SO₃)₂ and water molecules indicates that original solvation structure is not stable, which potentially leads to the side reactions and lower the overall performances. As the acetonitrile molecules are introduced in the electrolyte, the s, p orbitals of acetonitrile have strong contributions to both bonding and anti-bonding orbitals, especially near the C≡N bonds, implying that the acetonitrile dominates the interactions with both Zn(CF₃SO₃)₂ and LiTFSI in 0.5 m + 1 m-H₂O/AN, which is accompanied with improved orbital couplings (Fig. 4b). However, as the concentration of LiTFSI increases to 3 m, the bonding and anti-bonding orbitals are more dominated by LiTFSI, while the contributions of water and acetonitrile are alleviated (Fig. 4c). In this case, the energy barrier for electron transfer is increased. With the further increase of LiTFSI concentration to 5 m, the contributions of water molecules become much stronger in the bonding orbitals, which increases the possibility of HER (Fig. 4d). Meanwhile, the further weakened orbital couplings result in increased barriers for electron transfer, which is consistent with the enlarged resistance of experimental characterizations induced by the excessive LiTFSI.

Detailed electronic structures are revealed by the projected partial density of states (PDOSs). In 0.5 m + 1 m-H₂O electrolyte, the Zn-3d orbitals show an evident peak near E_V − 4.6 eV (E_V = 0 eV) (Fig. 4e). Notably, the s, p orbitals from LiTFSI and Zn(CF₃SO₃)₂ have dominant contributions near both valence band maximum (VBM) and conduction band minimum (CBM), which are consistent with the electronic distributions. Meanwhile, it is noticed that the s, p orbitals of H₂O have considerable overlapping with both LiTFSI and Zn(CF₃SO₃)₂, supporting the formation of the solvation structures. For such an electrolyte, the relatively large bandgap (3.0 eV) between VBM and CBM limits the overall electrical conductivity. As acetonitrile is introduced in the electrolyte (0.5 m + 1 m-H₂O/AN), a slight downshifting of Zn-3d orbitals due to the interactions with acetonitrile (Fig. 4f). More importantly, the s, p orbitals of acetonitrile not only dominate the CBM and VBM but also lead to the formation of gap states, which significantly alleviates the energy barrier for electron transfer and improves the electrochemical performances. The orbital overlapping between acetonitrile-s, p, and Zn-3d is improved while the overlapping between H₂O-s, p, and Zn-3d is decreased. Compared with the results of 0.5 m + 1 m-H₂O electrolyte, the interactions between H₂O with Zn(CF₃SO₃)₂ and LiTFSI are drastically weakened in the 0.5 m + 1 m-H₂O/AN electrolyte, demonstrating the change of solvation structure due to the interruption by acetonitrile molecules. On the other hand, as the concentration of LiTFSI increases, a significant downshift of Zn-3d orbitals is observed, revealing a further suppression of the interactions between water molecules with both Zn²⁺ cations and acetonitrile molecules (Fig. 4g). Meanwhile, the H₂O-s, p orbitals, instead of the acetonitrile-s, p orbitals, become dominant near the Fermi level and the barrier for electron transfer is increased compared to that of 0.5 m + 1 m-H₂O/AN electrolyte. With the further increase of LiTFSI concentrations, though Zn-3d orbitals are upshifted to E_V − 5.0 eV, the energy gap between CBM and VBM is enlarged to 4.0 eV, resulting in an even higher barrier for electron transfer (Fig. 4h). In such a circumstance, the interactions between Zn(CF₃SO₃)₂ and LiTFSI are enhanced. However, both Zn(CF₃SO₃)₂ and LiTFSI show weakened interactions with H₂O and acetonitrile, leading to poor electrochemical performances.

The radial distribution functions (RDFs) of different electrolytes from the MD simulations are also compared. In the 0.5 m + 1 m-H₂O electrolyte, the main bonding contributions come from the O-H bonds in water molecules and the corresponding hydrogen bonds between water molecules also show an important contribution with the second peak within 1.5-1.6 Å (Fig. 4i). We notice a broad peak after 2.6 Å, which corresponds to the coordination shell formed between Zn ions and water molecules. In contrast, in the water/acetonitrile electrolyte, the peak of water molecules almost disappears, where the stronger peak at 1.1 Å is ascribed to C-H, C≡N bonding of the acetonitrile (Fig. 4j). Compared to the 0.5 m + 1 m-H₂O electrolyte, the hydrogen bonding network of water is strongly perturbed with no evident peak presented. The weak hydrogen bond and interactions between acetonitrile molecules lead to a peak at 1.8 Å. The coordination shell between Zn ions and acetonitrile molecules is formed at the distance of 2.1 Å, which is much smaller than that with water, indicating the modulations of solvation structure. Meanwhile, the interactions between Zn and water molecules are strongly suppressed, which potentially lowers the possibility during HER. Figure 4k depicts the interaction energies between ions and solvent molecules in different electrolytes. As compared with 0.5 m + 1 m-H₂O, the interaction energies in the 0.5 m + 1 m-H₂O/AN electrolyte are reduced, suggesting stronger interactions. As the concentration of LiTFSI increases, the interaction energies increase accordingly due to the weakened interactions, which is consistent with the NMR results. In addition, the mean square displacement (MSD) demonstrates the highest diffusion is induced by the acetonitrile in the 0.5 m + 1 m-H₂O/AN electrolyte (Fig. 4l). The stronger ion interactions in water/acetonitrile electrolyte, together with higher migration of ions and molecules than those in pristine aqueous electrolyte, not only promote the conductivity of electrolyte but also improve the operation voltage of the supercapacitor.