Negative electrode degradation induced by two-stage zinc plating and its recovery in zinc batteries

Abstract

Controlling the morphology of Zinc (Zn) deposits is an effective strategy to produce stable Zn-metal batteries. However, the degradation of Zn negative electrodes and changes in their morphology remain poorly understood. Here, we show that the Zn plating process has two distinct stages. The first being the formation of relatively dense, lumpy Zn, while the second involves the formation of porous mossy Zn on its protrusions, which changes into electrochemically inactive dead Zn during stripping. Based on this, we propose a strategy involving a combination of cationic and anionic reagents to revive the dead Zn. The cations create a positively charged inert region on the negative electrode surface for Zn-ion dispersion that inhibits mossy Zn formation, while the anions act as a redox mediator to revive the dead Zn. Consequently, the Zn negative electrode shows a Coulombic efficiency (CE) of 99.7% and stable Zn plating/stripping over 1400 h (10 mA cm⁻² and 10 mAh cm⁻²). An Ah-scale pouch cell retains 96.2% capacity after 800 cycles. This work provides key insights into mossy Zn formation and proposes a promising approach for stable Zn-metal batteries.

The formation of inactive zinc severely limits zinc battery lifespan. Here, authors reveal its origin from a two-stage plating process and report the use of the combination of cationic and anionic reagents to prevent its formation and recover lost capacity, achieving stable zinc batteries.

Introduction

Rechargeable zinc-metal batteries (ZMBs) are known for their intrinsic safety, low cost, non-toxicity, and compatibility with aqueous electrolytes, and are regarded as a major alternative to lithium batteries^1,2. These distinctive advantages position ZMBs as an ideal candidate for safety-critical uses, including smart grids and residential energy storage systems³. Nevertheless, their practical deployment has some major problems, primarily stemming from non-uniform Zn electrodeposition at the negative electrode, which triggers short circuits and rapid capacity decay^4,5.

Many efforts have focused on elucidating the fundamental mechanisms governing this Zn nucleation and growth^6–10. Current understanding is that the Zn deposition morphology is influenced by many factors, including current density^11,12, ionic transport properties of electrolyte^13,14, interfacial electrochemical environment^15,16, and a lattice mismatch between the substrate and Zn^17,18. To solve the problem of irregular Zn plating on planar substrates, various methods have been used, including, but not limited to, surface modification^19,20, three-dimensional design of the host structure^21,22, and changing the electrolyte formula^23–25. Some recent studies have identified a dominant deposition morphology at low current densities, which is high-surface-area porous Zn comprised of tortuous nanowires or nanosheets, and distinct from classical diffusion-limited dendritic growth^6,26–28. This prevalent but poorly understood structure, often described as mossy Zn rather than conventional dendrites, warrants particular attention due to its correlation with the formation of electrochemically inactive dead Zn that causes capacity loss during stripping. However, the origin, development, and prevention of the formation of this material remain poorly understood.

We now report the reason for the formation of mossy Zn and a strategy for its suppression and the reactivation of dead Zn. Using optical monitoring, we have revealed a two-stage plating mechanism consisting of the initial formation of compact lumpy Zn, followed by a transition to a mossy structure by the development of surface protrusions. During subsequent stripping of Zn, we have shown that the irreversible conversion of mossy Zn to inactive dead Zn is due to the loss of interfacial contact between the substrate and Zn. To deal with this issue, we report the use of cationic and anionic reagents in which the cationic component forms a positively charged inert region on the negative electrode surface to prevent local Zn²⁺ accumulation, while the anionic species acts as a redox mediator to revive the dead Zn. This combination allows prolonged Zn plating/stripping over 1400 h in Zn||Zn symmetric cells at 10 mA cm⁻² and 10 mAh cm⁻² and gives a life of 800 cycles to Ah-scale Zn||I₂ pouch cells.

Results

Two-stage plating and stripping of the Zn negative electrode

To in situ monitor the Zn plating and stripping process, we designed a Zn||Ti cell in a customized micro dish with openings on three sides, as depicted in Fig. 1a and Supplementary Fig. 1. The working electrode (a Ti wire) and counter electrode (a Zn wire) were aligned with ~1 mm spacing in a 1 M zinc trifluoromethanesulfonate (ZnOTF) aqueous electrolyte, accompanied by a Zn reference electrode. Ti was used as the working electrode because it is one of the most common current collectors for Zn negative electrodes and can prevent hydrogen evolution from affecting the observation of Zn plating. When a constant current density of 1 mA cm⁻² was applied, metallic Zn deposits with a characteristic metallic luster nucleated on the Ti surface (Fig. 1c and Supplementary Video 1). The corresponding scanning electron microscopy (SEM) images show that the initial lumpy Zn deposits have a well-defined crystalline structure (Fig. 1d) that matches the thermodynamic Wulff construction of a hexagonal close-packed (hcp) Zn crystal. This smooth Zn surface acts like a mirror, specularly reflecting a large amount of incident light from the microscope’s illumination source directly back into the objective lens, which gives the Zn deposits a luminescent appearance.

Fig. 1. Two-stage plating and stripping of the Zn negative electrode.

a Schematic of the optical cell setup. b Typical voltage-time profile of Zn plating and stripping at a current density of 1 mA cm⁻². c In situ optical microscope photos of the two-stage Zn plating at a current density of 1 mA cm⁻² and a temperature of 25 °C. Scale bar, 500 μm; insets, 100 μm. d Corresponding ex situ SEM images of the Zn deposits after plating 0, 10, 60, 90, and 120 min at a current density of 1 mA cm⁻² and a temperature of 25 °C. Scale bar, 20 μm; insets, 4 μm. COMSOL simulation of (e) current density distribution and (f) electric field distribution near Zn electrode surface at different plating stages. g In situ optical microscope photos of Zn stripping at a current density of 1 mA cm⁻² and a temperature of 25 °C. Scale bar, 500 μm. Source data are provided as a Source data file.

After plating for ~60 min, the lumpy Zn progressively fills intergranular voids while exposing distinct edges. At the same time, a secondary mossy Zn morphology—comprised of a one-dimensional filamentary structure—nucleates preferentially at the protrusions on the lumpy Zn (Fig. 1d) and it is this mossy phase which eventually dominates the plating, whereas the lumpy Zn remains unchanged. The voltage versus time profile showed decreased polarization during mossy Zn formation (Fig. 1b), contrary to classical dendritic behavior where interface depletion typically increases polarization. This fundamental difference suggests distinctly different mechanisms between mossy Zn growth and conventional diffusion-limited dendrite growth.

The morphological change of Zn deposits correlates with the increase in local current density at surface protrusions of lumpy Zn, which is attributed to an electric field concentration effect, commonly referred to as the “tip effect”. To support this mechanism, we performed finite element simulations to visualize the current density and electric field distribution of both two Zn morphologies. In this model, regular hexagons and elongated bars represent lumpy Zn and mossy Zn, respectively. As shown in Fig. 1e, f, the gradient colors denote current intensity and electric field. The maximum current density and electric field are intensely concentrated at the tips and edges of the lumpy Zn, triggering the formation of Zn whiskers. Once a mossy filament forms, the concentrated point shifts to the top of the Zn filament, creating a positive feedback loop that drives the rampant growth of mossy Zn.

Parametric studies showed the universality of this two-stage plating mechanism under different conditions. A prolonged deposition time leads to the proliferation of mossy Zn in the space between electrodes, ultimately causing cell short-circuiting without the production of other structures (Supplementary Fig. 2). Increasing or decreasing the current density produces a similar morphology change during Zn plating (Supplementary Figs. 3 and 4). We found that two-stage Zn deposition and stripping occurred on a planar Ti foil electrode during the cycle, indicating that this phenomenon is independent of the electrode’s geometric form (Supplementary Fig. 5). We also discovered that the simple replacement of the Zn salt in the electrolyte with typical ZnSO₄, ZnCl₂, and Zn(ClO₄)₂ electrolytes, did not prevent the initiation and growth of mossy Zn (Supplementary Figs. 6–8). The salt anion might influence the nucleation size and the timing of the transition of Zn morphologies. The use of Zn as the deposition substrate appeared to accelerate the formation of the mossy Zn (Supplementary Fig. 9). In addition, the use of stack pressure compressed the Zn deposits, making it more uniform, but did not change the intrinsic mechanism of the two-stage deposition process (Supplementary Figs. 10 and 11).

During the subsequent stripping process, the mossy Zn rapidly shrank until only black insoluble residues were left on the surface of the lumpy Zn (Fig. 1g), which then to dissolve, accompanied by a small increase in voltage polarization, implying that the lumpy Zn was more electrochemically stable than the mossy Zn. When the cell was charged to a cut-off potential of 0.5 V (vs. Zn/Zn²⁺), substantial Zn remained on the surface of the Ti electrode due to poor interfacial contact, generating electrochemically isolated dead Zn.

Characterization of lumpy and mossy Zn

During the transition from lumpy Zn to mossy Zn, we have investigated the changes in physical and electrochemical properties of the Zn deposits, such as crystallographic features, electrochemical active area, and resistance to side-reactions. As shown in Fig. 2a, diffraction peaks corresponding to (002)_Zn, (101)_Zn, and Ti substrate were detected after plating for 10 min. A key finding is that the (101)_Zn and (100)_Zn peaks intensified as the plating time increased, especially after the formation of the mossy Zn. The corresponding peak intensity ratios I₍₁₀₁₎/I₍₀₀₂₎ and I₍₁₀₀₎/I₍₀₀₂₎ were calculated and are shown in Fig. 2b. The I₍₁₀₁₎/I₍₀₀₂₎ ratio increased dramatically from 1.06 to 1.75 when the deposition time increased from 60 to 120 min, implying (101)_Zn-textured growth occurred in the mossy Zn.

Fig. 2. Characterization of the Zn deposits during the transition to mossy Zn formation.

a XRD patterns of the Zn deposits after 10, 60, 90, and 120 min of electroplating at a current density of 1 mA cm⁻² and a temperature of 25 °C, and b the relative intensities I_(100/002) and I_(101/002) of the Zn deposits. c EDLC of the Zn deposits after 10, 60, 90, and 120 min of electroplating at a current density of 1 mA cm⁻² and a temperature of 25 °C. Error bars represent the standard deviation (SD) of different samples. Data are presented as mean ± SD (n  =  3). d In situ EIS spectra during Zn plating. e Tafel and f hydrogen evolution polarization curves of the Zn deposits after 10, 60, 90, and 120 min of electroplating at a current density of 1 mA cm⁻² and a temperature of 25 °C. The corrosion current (i_corr) densities were obtained from the Tafel plots. g Schematic of the two-stage plating and stripping of the Zn negative electrode. h Schematic of the plating and stripping of the Zn negative electrode when using the cationic and anionic reagents. Source data are provided as a Source data file.

The increased electrochemically active surface area induced by mossy Zn formation was corroborated by electric double-layer capacitance (EDLC) measurements (Fig. 2c and Supplementary Fig. 12). Moreover, the changes in impedance were monitored during the Zn plating. The interfacial impedance remained at ~400 Ω for the first 60 min and then markedly decreased to ~100 Ω, suggesting the formation of mossy Zn (Fig. 2d). This finding is consistent with both the optical microscope observations and EDLC results. The high electrochemical activity of the mossy Zn typically led to decreased anti-corrosion ability and increased hydrogen evolution. As expected, Tafel curves showed a sequential increase in corrosion current from 0.21 to 0.72, 1.77, and 4.12 mA cm⁻² as the plating time increased from 10 to 60, 90, and 120 min (Fig. 2e). Figure 2f showed the linear sweep voltammetry (LSV) curves of different Zn deposits. The formation of the mossy Zn results in a significant increase in hydrogen evolution current for the Zn deposits at a same potential.

Combining the above results, a schematic, which describes the two-stage process of the Zn plating and stripping, is shown in Fig. 2g. The initial plating produces a lumpy Zn structure with a well-defined shape (Stage 1), whose edge growth gradually disrupts the homogeneity of the Zn-ion flux, inducing localized protrusions that develop into one-dimensional Zn filaments, which swiftly evolve into mossy Zn by tip-enhanced deposition (Stage 2). During stripping, preferential dissolution of the mossy Zn generates electrically isolated dead Zn due to the loss of interfacial contact between the filaments. This degradation underscores the critical need to suppress mossy Zn formation and revive dead Zn for battery longevity.

For this purpose, we have added a two-function reagent with cationic and anionic components to the electrolyte to prevent the formation of mossy Zn and revive the dead Zn (Fig. 2h). The cations migrate to the surface of the negative electrode driven by the electric field, creating a positively charged inert region that homogenizes the Zn-ion distribution, thus deterring the formation of mossy Zn. The anions, with their redox activity, generate oxidative mediators to react with the dead Zn to form electroactive Zn ions that can return for use in the battery cycle. As a proof-of-concept, acetylcholine iodide (ACI), consisting of acetylcholine cations and iodine anions (Supplementary Fig. 13), was introduced into a baseline ZnOTF electrolyte and is subsequently denoted as ZnOTF-ACI. Of note, the combination strategy of cations and anions is not limited to the ACI, as other bi-ionic form analogs exhibit similar effects, which will be discussed later.

Cation-mediated Zn plating without mossy Zn

We first investigated the role of acetylcholine cations in controlling the Zn deposition. To elucidate the adsorption layer of acetylcholine cations at the electrode and electrolyte interface, electrochemical quartz crystal microbalance (EQCM) measurements were performed. The experimental setup is shown in Fig. 3a, where an Au wafer with a quartz crystal sensor serves as the working electrode, and 0.2 M Na₂SO₄ aqueous solution containing NaI, containing ACI, and containing neither were used as electrolytes. When the potential swept from 0.1 V to −0.4 V (vs. Ag/AgCl), the maximum mass increase was ~7.3 and 3.6 ng in the 0.2 M Na₂SO₄ and 0.2 M Na₂SO₄-NaI, respectively (Fig. 3b, c). However, ~60.1 ng was detected when ACI was added. This large increase in mass is mainly attributed to the adsorption of acetylcholine cations rather than Na⁺, I⁻, or H₂O.

Fig. 3. Cation mediated Zn negative electrode interface.

a Schematic of the EQCM measurement setup. b Linear sweep voltammetry curves in an EQCM test at 5 mV s⁻¹ and a temperature of 25 °C. c Mass change curves of different Na₂SO₄ based solutions. d Schematic of the in situ FT-IR measurement setup. e In situ FT-IR spectra during Zn plating in the ZnOTF-ACI electrolyte at a current density of 1 mA cm⁻² and a temperature of 25 °C. f Adsorption energy of the acetylcholine cation (AC⁺), I⁻ anion, Zn ion, and H₂O on the Zn substrate. Corresponding structural formula are shown in the insets, where red, pale pink, silver-gray, purple, brown, blue-gray balls represent O, H, Zn, I, C, and N atoms, respectively. Source data are provided as a Source data file.

To better elucidate the cation-mediated Zn negative electrode interface, we conducted the in situ electrochemical Fourier transform infrared spectroscopy (in situ FT-IR) to monitor the dynamic interfacial evolution at the Zn negative electrode. As shown in Fig. 3d, in situ FT-IR were conducted in a spectra-electrochemical cell, which consisted of a Zn counter/reference electrode, a Ti mesh working electrode, and a ZnOTF-ACI electrolyte. Galvanostatic discharge tests were performed, and changes in the interfacial infrared spectra during the discharge were detected, with the spectra at open-circuit potential serving as the background. As shown in Fig. 3e, the absorption band at ~2980 cm⁻¹ is attributed to the C-H stretching band of the methyl group in acetylcholine cations. The intensity of the C-H stretching band progressively increases during Zn plating, indicating a progressive adsorption process of cations onto the electrode surface. This is also consistent with the results from the electrochemical EQCM measurements.

To better illustrate the interfacial interactions between the Zn negative electrode and ACI, we conducted the density functional theory (DFT) calculations (Supplementary Fig. 14 and Supplementary Data 1). As shown in Fig. 3f, the adsorption energy of the horizontal acetylcholine cations (−3.86 eV) was much higher than that of H₂O (−0.32 eV), Zn²⁺ (− 1.46 eV), iodide anions (−2.79 eV), and the vertical acetylcholine cations (−3.23 eV). This confirms the preferential adsorption between the acetylcholine cations and Zn negative electrode, which creates a positively charged inert layer on the Zn negative electrode to disperse Zn²⁺ ions.

Optical microscopy captured distinct deposition behavior in the ZnOTF-ACI electrolyte (Fig. 4a, Supplementary Fig. 15 and Supplementary Video 2). Specifically, after plating for 10 min, abundant shiny Zn seeds were uniformly distributed on the surface of the Ti electrode. These grew and fused together to produce a uniform coverage of the Ti electrode. Subsequent growth maintained a planar morphology, contrasting sharply with the filamentary growth in the baseline ZnOTF electrolyte. SEM images show a flat, lumpy Zn structure throughout the deposition process, accompanied by the emergence of the (002)_Zn texture with typical microscopic changes (Fig. 4b). X-ray diffraction (XRD) also showed that the intensity of the (002)_Zn peak increased slightly as the plating time increased (Fig. 4c and Supplementary Fig. 16). Notably, the significant improvement in Zn deposition morphology is attributed to acetylcholine cations rather than iodide anions, as the addition of iodide anions alone does not inhibit the formation of mossy Zn (Supplementary Fig. 17).

Fig. 4. Uniform Zn plating and stripping in the ZnOTF-ACI electrolyte.

a In situ optical microscope photos of Zn plating in the ZnOTF-ACI electrolyte at a current density of 1 mA cm⁻² and a temperature of 25 °C. Scale bar, 500 μm; insets, 100 μm. b Corresponding ex situ SEM images of the Zn deposits after plating 0, 10, 60, 90, and 120 min at a current density of 1 mA cm⁻² and a temperature of 25 °C. Scale bar, 20 μm; insets, 4 μm. c XRD patterns of the Zn deposits at after 10, 60, 90, and 120 min of electroplating at a current density of 1 mA cm⁻² and a temperature of 25 °C. d In situ EIS spectra during Zn plating in the ZnOTF-ACI electrolyte. e In situ optical microscope photos of Zn stripping in the ZnOTF-ACI electrolyte at a current density of 1 mA cm⁻² and a temperature of 25 °C. Scale bar, 500 μm. Source data are provided as a Source data file.

In addition, the in situ electrochemical impedance spectra (EIS) showed only a slight change in interfacial impedance during the Zn deposition process, indicating a good interfacial stability of the Zn deposits (Fig. 4d). Compared to its counterpart, the Nyquist plots of half-cells with the ZnOTF-ACI electrolyte had a small semicircle at the high-frequency region, possibly corresponding to the acetylcholine cation adsorption forming an additional interface. The Zn deposits with acetylcholine cations had a low surface area, thus providing strong resistance against corrosion and hydrogen evolution even in the baseline ZnOTF electrolyte (Supplementary Figs. 18 and 19). The initial nucleation behavior of Zn ions was investigated using chronoamperometry (CA). It was found that the current response of the ZnOTF-ACI electrolyte stabilized in a short time (~10 s) at a polarization of −150 mV, whereas the baseline ZnOTF electrolyte has a continuously increasing value over 300 s (Supplementary Fig. 20). This indicates that acetylcholine cations restrain the two-dimensional diffusion of Zn ions and promote uniform and flat Zn deposition. These results show that acetylcholine cations effectively prevent the transition from the deposition of lumpy Zn to mossy Zn. As a result, most Zn deposits can be stripped from the Ti substrate during subsequent charging, leaving behind a few particles of Zn debris (Fig. 4e).

Anion-mediated reactivation of dead Zn

In addition to the change in Zn deposition produced by acetylcholine cations, the I⁻ anions of the ACI are believed to revive the dead Zn by generating a certain amount of oxidizing mediator (I₃⁻) during charging. From a thermodynamic viewpoint, the reaction between I₃⁻ and Zn is spontaneous according to Gibbs free energy calculations (Supplementary Fig. 21). Experimental validation involved immersing the collected dead Zn in a Zn(I₃)₂ solution for 5 days (Fig. 5a). The initial orange solution became lighter, indicating that I₃⁻ was partially consumed by the oxidation of dead Zn, as evidenced by the reduced adsorption intensity in the ultraviolet-visible spectroscopy (UV/Vis) spectra (Fig. 5b). Inductively coupled plasma optical emission spectrometer (ICP-OES) quantification test showed the Zn-ion concentration increasing from 2.34 to 7.47 mM, which nearly matched the theoretical yield (Fig. 5c). These results confirm the effectiveness of the iodine in converting dead Zn (Zn⁰) to electroactive Zn²⁺. Furthermore, Zn_x(OTf)_y(OH)_2x−y·nH₂O (ZOTH), a typical by-product of the reaction between the Zn metal negative electrode and the ZnOTF electrolyte, is an additional form of active Zn loss. By soaking a polished Zn foil in the ZnOTF electrolyte for 7 days, ZOTH nanosheets were easily generated (Fig. 5d, e and Supplementary Fig. 22), which greatly increased the electrical resistivity of the Zn foil to ~2.1×10⁵ Ω·cm (Supplementary Fig. 23). The ZOTH nanosheets also increased the Zn-ion transport resistance (Supplementary Fig. 24). Fortunately, the nanosheets can be removed by treating with the iodine mediator to recover a smooth Zn surface (Fig. 5f), as verified by energy-dispersive spectroscopy (EDS), XRD and X-ray photoelectron spectroscopy (XPS) (Supplementary Figs. 25 and 26). Considering that the I₃⁻ might also corrode fresh Zn, the decomposition rates of the I₃⁻ on fresh Zn, dead Zn, and ZOTH were compared. Specifically, we immersed three samples—fresh Zn foil, dead Zn, and pre-synthesized ZOTH—in separate vials containing 50 mL of a 2.5 mM I₃⁻ solution. Note that due to the uncertain molecular weight of ZOTH, all three samples are in excess relative to the I₃⁻. At intervals of 1, 3, and 7 days, the upper supernatant was extracted to quantify the dissolved Zn²⁺ by ICP-OES. The results, presented in the Supplementary Fig. 27, clearly showed that the Zn²⁺ concentration increased in the order: ZOTH > dead Zn > fresh Zn at every time point. This demonstrated that the decomposition rate followed the same order. These different rates might be attributed to the dramatic difference in specific surface area, where dead Zn and ZOTH exhibit porous morphologies with high specific surface areas, while fresh Zn foil is dense and flat and provides a low specific surface area. This kinetic preference ensures that the I₃⁻ mediator will predominantly target the high-surface-area inactive species rather than the bulk Zn electrode. Moreover, dead Zn is more prone to be corroded by the electrolyte to produce ZOTH than fresh Zn (Supplementary Fig. 28).

Fig. 5. The underlying functions of the iodine species.

a Optical photos of dead Zn before and after soaking in a Zn(I₃)₂ aqueous solution for 5 days. b UV-Vis spectra and c ICP-OES of the Zn(I₃)₂ solution before and after treatment with dead Zn. SEM images of (d) polished Zn, e ZOTH@Zn, and f soaked ZOTH@Zn. Scale bar, 20 μm. g Typical voltage profiles of the formation of dead Zn at 1 mA cm⁻². h Voltage profiles of Zn||Ti cells with the ZnOTF-ACI electrolyte at 5 mA cm⁻² after the formation cycle. i Optical photos of Ti electrodes and separators after the Aurbach CE tests. Source data are provided as a Source data file.

To further demonstrate the recovery of dead Zn in a Zn | |Ti cell configuration where dead Zn forms directly on the Ti electrode, we conducted CE measurements with and without ACI. During the formation cycle, 4 mAh cm⁻² Zn was first plated on the Ti substrate and then stripped away with a cut-off voltage of 0.5 V, generating 0.28 and 0.87 mAh cm⁻² of dead Zn on the Ti in ZnOTF electrolytes with and without ACI (Fig. 5g and Supplementary Fig. 29). The Zn||Ti cell was then subjected to cycle at a capacity of 1 mAh cm⁻² for one cycle, and 1.12 mAh cm⁻² of Zn was stripped from the Ti substrate with a corresponding CE of 111.6% (Fig. 5h). The over 100% CE is solid evidence for the ACI-assisted recovery of dead Zn. In contrast, only 0.85 mAh cm⁻² was obtained in the ZnOTF electrolyte, corresponding to a CE of 85.3%.

To decouple the contributions of the cations and anions, other types of bi-ionic form analogs have also been evaluated. Firstly, to elucidate the effect of anions, we kept the acetylcholine cations constant and replaced I⁻ anions with other halides such as bromide (Br⁻) and chloride (Cl⁻). As shown in the Supplementary Figs. 30 and 31, cells with acetylcholine bromide (ACBr) and acetylcholine chloride (ACCl) generated 0.90 and 1.05 mAh cm⁻² of dead Zn in the formation cycle, respectively. In the subsequent cycle, they still produced 0.11 and 0.04 mAh cm⁻² of dead Zn, corresponding to CEs of 89.2% and 95.8%. This indicates that Cl⁻ and Br⁻ cannot effectively reactivate dead Zn, which might be attributed to their high oxidation potential required to form oxidizing mediator. Secondly, to elucidate the effect of cations, we kept the I⁻ anions constant and replaced the acetylcholine cations with other structurally related cations, including choline (C⁺), butyrylcholine (BC⁺), and acetylthiocholine (ATC⁺). All I⁻-based bi-ionic analogs (CI, BCI, ATCI) could reactivate dead Zn to some extent, as evidenced by a CE value exceeding 100% (Supplementary Figs. 32–34). This confirms the primary role of the I⁻ anions in capacity recovery.

Aurbach CE tests^29,30, which provide a good assessment of Zn plating/stripping efficiency for Zn negative electrodes, also showed the effectiveness of the ACI addition with a significantly higher average CE (98.5%) than with the baseline ZnOTF electrolyte (78.8%, Supplementary Fig. 35). Post-mortem analysis revealed few residual Zn in the ZnOTF-ACI based cell compared to extensive dead Zn accumulation in the baseline electrolyte (Fig. 5i)

Electrochemical performance of the Zn negative electrode

The stability of Zn negative electrodes was assessed in different electrolytes using Zn||Ti asymmetric cells and Zn||Zn symmetric cells. The CE of the Zn negative electrode was first tested to confirm the plating/stripping reversibility. The baseline ZnOTF electrolyte showed instability in its CE during a long-term cycle at 1 mA cm⁻² and 1 mAh cm⁻²; however, the Zn metal CE vastly improved and maintained a high average value of 99.7% over 1800 cycles when using the ZnOTF-ACI electrolyte (Fig. 6a). The ZnOTF-ACI electrolyte retained a low and stable voltage hysteresis of ~34 mV throughout the entire cycling process, whereas that of the baseline ZnOTF electrolyte increased significantly in less than 100 cycles (Fig. 6b, c). Stable Zn plating/stripping was also achieved at a current density of 5 mA cm⁻² and an areal capacity of 1 mAh cm⁻², and even at a high current density of 10 mA cm⁻² and an areal capacity of 5 mAh cm⁻² (Supplementary Fig. 37).

Fig. 6. Zn negative electrode stability in different electrolytes.

a CE of Zn negative electrodes in Zn||Ti cells and b, c the corresponding charge-discharge profiles after different numbers of cycles, tested at 25 °C. d, e Galvanostatic Zn plating/stripping in Zn||Zn coin cells at different cycling conditions, tested at 25 °C. f Comparison of the electrochemical performance between our work and previous studies. Cumulative plating capacity is denoted as CPC. g Optical photos and galvanostatic Zn plating/stripping of a large Zn||Zn pouch cell. h, i SEM images and j, k 3D images of Zn negative electrodes after 100 cycles at 5 mA cm⁻², 5 mAh cm⁻², and a temperature of 25 °C using the different electrolytes. Scale bar, 20 μm; insets, 4 μm. Source data are provided as a Source data file.

The Zn||Zn symmetric cells showed fundamental stability differences between the two electrolytes. As shown in Fig. 6d, while the voltage profiles became unstable after ~100 h for the ZnOTF electrolyte, the voltage remained steady over 6000 h for the ZnOTF-ACI electrolyte at a current density of 1 mA cm⁻² and an areal capacity of 1 mAh cm⁻². When the current and capacity increased to 5 mA cm⁻² and 5 mAh cm⁻², respectively, the cell with the ACI electrolyte, had a much longer lifespan of 2000 h (Supplementary Fig. 38) than the cell without the ACI in which the voltage response suddenly dropped to a low value after ~50 h. Time-dependent EIS tests showed that the failure of the cell with the baseline ZnOTF electrolyte originated from a dendrite-induced “soft short”³¹, a phenomenon absent in cells with the modified electrolyte (Supplementary Figs. 39, 40). Remarkably, under harsh conditions (10 mA cm⁻², 10 mAh cm⁻², and 38% depth of discharge), the cell with the ZnOTF-ACI electrolyte achieved stable Zn plating/stripping over 1400 h with a 7 Ah cm⁻² cumulative plating capacity (Fig. 6e), values that are comparable to, or better than, previously reported designs for Zn negative electrodes (Fig. 6f and Supplementary Table 1)^15,32–46. The increased overpotential observed in the late cycle might be attributed to the gradual decomposition and consumption of the aqueous electrolyte under harsh conditions. The stable Zn plating/stripping obtained using the ZnOTF-ACI electrolyte has also been demonstrated at different current densities (Supplementary Fig. 41).

To ensure reliability, the influence of ACI effects on cell performance was validated using other Zn salts, including ZnSO₄, ZnCl₂, and Zn(ClO₄)₂. After the introduction of ACI into the baseline electrolyte, both the CE and lifespan of the cells were substantially enhanced, demonstrating the applicability of ACI in stabilizing Zn negative electrodes (Supplementary Figs. 42–44). In addition, the stability of Zn plating/stripping, reflected in the lifespan of symmetric cells, showing significant improvement after the addition of other bi-ionic form analogs (Supplementary Figs. 30–34).

To validate the practicality of the ZnOTF-ACI electrolyte, we then scaled up Zn||Zn symmetric coin cells to large-format pouch cells (7 × 7 cm²). As shown in Fig. 6g, a symmetric Zn||Zn pouch cell with the ZnOTF-ACI electrolyte sustainably ran over 1100 h at 245 mA and 245 mAh per cycle, producing a high cumulative plating capacity of 134.75 Ah. In contrast, the baseline ZnOTF electrolyte had a high and unstable overpotential after cycling for ~100 h. The performance of the ZnOTF-ACI verifies the enhanced plating/stripping reversibility of the Zn negative electrode under practical conditions, which is ascribed to the two functions of the ZnOTF-ACI electrolyte in preventing mossy Zn formation and reviving the dead Zn.

We also characterized the morphologies and components of the cycled Zn negative electrodes and separators by SEM, XRD, and a three-dimensional (3D) microscope. As shown in Fig. 6h, i and Supplementary Fig. 45, non-uniform Zn deposits and deep pits were detected after cycling with the ZnOTF electrolyte at 5 mA cm⁻² and 5 mAh cm⁻²; however, the Zn metal negative electrode maintained a smooth and shiny surface when using the ZnOTF-ACI electrolyte. From the corresponding separator with the ZnOTF electrolyte, many black and island-shaped dead Zn particles were attached to its surface and caused a significant loss of active Zn, while that using the ZnOTF-ACI electrolyte had a clean surface (Supplementary Fig. 46). XRD patterns showed that the black residue on the cycled Zn negative electrodes and separators consists of metallic Zn and ZOTH by-products (Supplementary Fig. 47). The height characteristics of the cycled Zn negative electrodes were characterized by a 3D microscope. The Zn negative electrode had a low roughness after cycling in the ZnOTF-ACI electrolyte (Fig. 6j, k). In contrast, a Zn negative electrode cycled in the ZnOTF electrolyte had an uneven surface with a maximum height difference of ~81 μm.

Electrochemical performance of full cells

To explore the practical use of the ZnOTF-ACI electrolyte, we coupled Zn metal negative electrodes with iodine positive electrodes in both coin and pouch cell configurations. Cyclic voltammetry (CV) curves of assembled Zn||I₂ cells showed a typical I⁻/I₂ redox pair, regardless of the incorporation of ACI in the electrolyte (Fig. 7a). In addition, a higher peak current density in the ZnOTF-ACI electrolyte indicated that the iodine anions of ACI also participate in the charge-discharge process. Coupled with the contribution of the activated carbon to the capacity, the assembled Zn||I₂ cells had a high specific capacity exceeding the theoretical value (211 mA h g⁻¹, Supplementary Fig. 48). Specifically, those using the ZnOTF-ACI electrolyte had a higher specific capacity of 289.5 mAh g⁻¹ at 1 A g⁻¹ (based on a preloaded active I₂ mass, ~5.3 mg cm⁻²) compared to the ZnOTF electrolyte (208.9 mAh g⁻¹, Supplementary Fig. 49). As the current increased to 10 A g⁻¹, a high specific capacity of 211.0 mAh g⁻¹ and a capacity retention of 72.9% were maintained in the ZnOTF-ACI, surpassing the values using the ZnOTF electrolyte (135.7 mAh g⁻¹, 65.0%), indicating its good rate performance and reaction kinetics (Supplementary Fig. 50). Figure 7b showed the cycling stability of the Zn||I₂ cells at 2 A g⁻¹. The one with the ZnOTF-ACI electrolyte maintained a high discharge capacity of ~230.6 mA h g⁻¹ after 20,000 cycles. However, the Zn||I₂ cell with the ZnOTF electrolyte showed a fast capacity decay after ~1600 cycles.

Fig. 7. Electrochemical performance of Zn||I₂ full cells.

a CV and b cycling performance of Zn||I₂ coin cells using different electrolytes, tested at 25 °C. Schematic and cycling performance of (c) single-layer and (d) bi-layer Zn||I₂ pouch cell under harsh conditions, including a low capacity ratio of the negative electrode to the positive electrode (N/P) and a low ratio of electrolyte amount to cell capacity (E/C), tested at 25 °C. Source data are provided as a Source data file.

We fabricated a single-layer Zn||I₂ pouch cell (7 × 7 cm²) consisting of a Zn foil negative electrode (45 µm, 26.3 mAh cm⁻²) and a high mass loaded I₂ positive electrode (~28 mg cm⁻², 5.6 mAh cm⁻²), which corresponds to a capacity ratio of the negative electrode to the positive electrode (N/P) of 4.7 (Fig. 7c). The cell had a high capacity of ~264.4 mAh after 20 cycles which remained at ~245.0 mAh after 2500 cycles. The gradual increase in capacity during the initial cycles represents a progressive wetting and activation process, which has been commonly reported in previous studies^47,48. We also scaled up the cell to an Ah level by using a double side-coated positive electrode, increasing both the I₂ loading (~42 mg cm⁻² per side) and electrode size (7.5 × 9.5 cm²), as illustrated in Fig. 7d. The critical parameters, including positive electrode areal capacity (~8.6 mAh cm⁻² per side), N/P (~3.0), and the ratio of electrolyte amount to cell capacity (E/C, ~10 μL mAh⁻¹), were kept at practical levels. The energy density of this pouch cell was calculated as ~70.0 Wh L⁻¹ (see Supplementary Table 3 for the details). As shown in Fig. 7d and Supplementary Fig. 51, the cell with the ZnOTF-ACI electrolyte had a high discharge capacity of 1230 mAh at ~0.45 C and showed a cycling stability for 800 cycles with a high-capacity retention of 96.2% at 1.8 C. Of note, the C-rate is defined as the reciprocal of the time required for the cell to complete one full cycle. The discharge curves at different numbers of cycles reflect the significantly improved cycling stability. Our Zn||I₂ pouch cell is at the advanced level of Ah-scale Zn-metal batteries, as demonstrated by the comparison of its performance to other cells given in Supplementary Table 2^49–58.

Discussion

We have described a critical morphology transition in Zn electrodeposition and its mechanism, revealing that mossy Zn formation constitutes a secondary stage arising from localized ion aggregation on primary lumpy Zn deposits. The subsequent generation of dead Zn because of interfacial contact loss, is the fundamental way the Zn negative electrode in Zn-metal batteries degrades. On this basis, we introduced an anion-cationic reagent, which formed an inert cation-rich region on the Zn negative electrode surface to homogenize the Zn-ion flux and prevent the formation of mossy Zn on the one hand, and operated as a redox mediator to revive the dead Zn on the other. Consequently, the Zn negative electrode had a high reversibility, including an average CE of 99.7% and a long lifespan of more than 1400 h at a high current and capacity. A practical Zn||I₂ pouch cell achieved cycling stability over 800 cycles. This work provides insights to study the mechanism of Zn deposition and to produce stable Zn-metal batteries.

Methods

Materials

Zn(CF₃SO₃)₂ (purity ≥99.5%) was purchased from Duoduo Chem. ZnSO₄·7H₂O (purity ≥99%), ZnCl₂ (purity ≥99.99% metals basis), Zn(ClO₄)₂·6H₂O (purity ≥99.995% metals basis), Na₂SO₄ (purity ≥99%), NaI (purity ≥99.5%), ZnI₂ (purity ≥98%), I₂ (purity ≥99.8%), ethanol (purity ≥99.5%), Acetylcholine bromide (purity ≥98%), Acetylcholine chloride (purity ≥99%), and Acetylthiocholine iodide (purity ≥98%) were procured from Aladdin. Choline iodide (purity ≥98%) and Butyrylcholine iodide (purity ≥99%) were bought from Macklin. Acetylcholine iodide (purity >98%) were obtained from TCI. Activated carbon (AC, YP50F) and polytetrafluoroethylene (PTFE, PTFE-770) was purchased from Shenzhen Kejing Star Technology Co., Ltd. Zn foil (purity ≥99.99%), Ti foil (purity ≥99.99%), Zn wire (purity ≥99.99%) and Ti wire (purity ≥99.99%) were purchased from Wuhan Luojia Yanxin Technology Co., Ltd.

Electrolyte preparation

For the 1 M ZnOTF electrolyte, 36.353 g Zinc trifluoromethanesulfonate was completely dissolved in ultrapure water to form a 100 mL solution with a volumetric flask. The ZnOTF-ACI electrolyte was prepared similarly, except that a certain amount of acetylcholine iodide powder was introduced during dissolution. The electrolyte preparation was conducted in an ambient atmosphere at 25 ± 2 °C. Our experiments indicate that the optimum concentration of ACI is ~0.5 M (Supplementary Fig. 36).

Electrode fabrication

AC electrode. The AC electrode was prepared by a dry process. AC powder and PTFE powder were manually mixed at a 9:1 weight ratio until uniform, then transferred to the chamber of a high-speed disperser (MSK-SFM-D300, Shenzhen Kejing Star Technology Co., Ltd). After rotating at 9000 rpm for 4 min and repeating this process three times, the PTFE powder transformed into fibers and created a robust network to bind the AC powder. The mixture was then evenly spread onto a cleaned aluminium foil and converted into a seamless, self-supporting film by a hot calendering machine (MSK-H2150-E, Shenzhen Kejing Star Technology Co., Ltd) at 60 ± 2 °C. The thickness of the AC film was controlled by adjusting the gap between the two rolls of the calendering machine. By sandwiching a titanium mesh collector between two AC films and consolidating them using a mechanical press (PC-30, Tianjin Jingtuo Instrument Technology Co., Ltd), the AC electrodes were obtained. They were then resized according to the requirements of the different cells.

I₂ positive electrode. The aforementioned AC electrode was used as the host for the I₂ to produce a composite electrode. A low-loaded I₂ positive electrode (~5.3 mg cm⁻², for a coin cell) was prepared by a drop-coating method, which involved dropping an appropriate amount of I₂ containing ethanol solution onto a thin AC electrode (~6 mg cm⁻²). High-loaded I₂ positive electrodes (~28 and ~84 mg cm⁻²) were prepared by electrodepositing. Specifically, thick AC electrodes and Zn foils were used as the positive and negative electrodes, respectively. These electrodes were placed on opposite sides of an acrylic container, with a 0.5 M ZnI₂ aqueous electrolyte filling the space between them. During galvanostatic charging, iodine ions in the electrolyte were electroplated onto the AC electrodes to form solid I₂. An areal iodine loading of ~28 mg cm⁻² was achieved by charging at a current density of 150 mA for 2 h on a 7 × 7 cm² AC electrode. A higher loading of ~84 mg cm⁻² (for the bi-layer pouch cells) was achieved by charging at 600 mA for 2 h on a 7.5 × 9.5 cm² AC electrode. The iodine positive electrode was obtained by washing several times with ultrapure water and drying for 24 h in an oven at 40 °C.

Pouch cell assembly

Zn foils were mechanically polished and rinsed in ethanol before use. The Zn||Zn pouch cell was assembled using two polished Zn foils (7 × 7 cm²) and one Whatman GF/B glass fiber separator (7.2 × 7.2 cm²), with 5 mL of electrolyte injected. The thickness of the Whatman GF/B separator after compression is ~500 μm, with an average pore size of 1.0 μm. For a single-layer pouch cell, one I₂ positive electrode (7 × 7 cm²) and one Zn negative electrode (7 × 7 cm²) were placed on either side of the separator. An Ah-scale Zn||I₂ pouch cell was produced by a bi-layer structure of two Zn negative electrodes (7.5 × 9.5 cm²), one I₂ positive electrode (7.5 × 9.5 cm²), and two separators (7.7 × 9.7 cm²). These assembled stacks were accommodated in Al-plastic bag and filled with electrolyte. The injection of the electrolyte for single-layer Zn-I₂ pouch cells and Ah-scale Zn||I₂ pouch cells was controlled at approximately 21 and 10 μL mAh⁻¹, respectively. After degassing and heat sealing, these pouch cells rested for 2 h to ensure complete wetting. Galvanostatic cycling stability tests were conducted under a uniform pressure (~2 kPa) to ensure satisfactory component contact for these cells.

Material characterization

The morphologies, microstructure, and corresponding elemental maps were obtained using a scanning electron microscope (SEM, Thermo Scientific Apreo 2S) equipped with energy-dispersive spectroscopy (EDS). The crystallographic phase and chemical composition were analyzed by (XRD, Rigaku MiniFlex), X-ray photoelectron spectroscopy (XPS, PHI 5000 Versa Probe II), and inductively coupled plasma optical emission spectrometer (ICP-OES, Arcos II MV). The surface roughness and 3D morphologies of Zn negative electrodes were examined using the same 3D microscope. The cycled Zn electrode is extracted from the cell and cleaned with ultrapure water and ethanol. Following vacuum drying at 40 °C for 2 h, it is subjected to SEM, XRD, and other tests. The UV/Vis absorption spectra of iodine-containing solutions were tested by a UV/Vis spectrometer (LAMBDA 365) after diluting with ultrapure water by a factor of 50.

In situ optical observation

The optical microscopy observations were conducted using a homemade liquid cell, where a Ti wire (30 mm in length and 0.5 mm in diameter), a Zn wire (30 mm in length and 1 mm in diameter), and a Zn wire (10 mm in length and 1 mm in diameter) served as the working, counter, and reference electrodes, respectively. Prior to the assembly of the optical cell, the Ti wire and Zn wire were polished with 10,000 mesh sandpaper to remove surface oxides and thoroughly cleaned with deionized water and ethanol. The cell was sealed with waterproof tape to prevent electrolyte leakage. The optical cell was connected to a battery tester to perform a cycle of Zn plating and stripping. Corresponding voltage response was also recorded. Meanwhile, an ultra-depth three-dimensional microscope (DeltaPix DPX M6000) was used to capture a series of changes in Zn plating/stripping morphology over time during the cycle.

In situ EQCM

In situ EQCM (DyneChem DC-QMC3100) measurements were performed in a three-electrode cell consisting of an Au wafer working electrode, Ti counter electrode, and Ag/AgCl reference electrode. The assembled cell was examined with a LSV test at 5 mV s⁻¹. The change in resonant frequency (∆f) of the quartz crystal during the LSV test was recorded. Therefore, the mass change (∆m) of the quartz crystal electrode can be calculated using the following Sauerbrey equation:

$$\mathrm{\Delta }m=-\frac{A\sqrt{\rho \mu }}{2f_0^2}\mathrm{\Delta }f$$ 1

Where A is the piezoelectrically active crystal area (0.28274 cm²), ρ is the density of quartz (2.648 g cm⁻³), μ is the shear modulus of quartz (2.947 × 10¹¹g cm⁻¹s⁻²), f₀ is the resonant frequency of the quartz crystal (9 MHz), and ∆f is the frequency change.

In situ FT-IR

In situ FT-IR measurements were performed using a Thermo Fisher Scientific Nicolet iS50 spectrometer equipped with a liquid nitrogen-cooled Mercury Cadmium Telluride (MCT) detector and a silicon attenuated total reflection (ATR) accessory. Prior to use, the substrate plane of the Si prism was mechanically polished using aluminum oxide suspensions of different particle sizes until a mirror-like finish was achieved, followed by rinsing with deionized water. In the in situ FTIR cell, a Ti mesh served as the working electrode, while Zn foil functioned as both the counter electrode and reference electrode. The incidence angle was set at 60°, with a spectral resolution of 4 cm⁻¹. Each spectral data was collected at an interval of 60 s. The FT-IR spectroscopy combined with the galvanostatic discharge method was used to monitor the Zn deposition process on the electrode surface.

Electrochemical measurements

The Zn||Zn symmetrical coin cells (CR2032) were assembled with two identical polished Zn foils (~45 μm in thickness and 12 mm in diameter), one Whatman GF/B glass fiber separator (16 mm in diameter), and 120 μL of the electrolyte. Galvanostatic Zn plating/stripping was performed in different cycling conditions. For Zn||Ti asymmetrical coin cells, Ti foil (~30 μm in thickness and 12 mm in diameter) was used as the working electrode, and the charge cutoff potential was set at 0.5 V (vs. Zn/Zn²⁺). A “reservoir” protocol was used to assess the average CE of Zn plating/stripping. First, the assembled asymmetrical Zn-Ti cell was subjected to one cycle at a capacity of 5 mAh cm⁻² to remove any impurities on the Ti substrate. Then, a given amount of charge (Q_T), e.g., 5 mAh cm⁻², was deposited on the Ti substrate as a Zn “reservoir”. Afterwards, repeated Zn stripping/plating was conducted (10 cycles) under a areal capacity (Q_C) of 1 mAh cm⁻². Finally, the residual Zn was completely stripped at a cut-off potential of 0.5 V (vs. Zn/Zn²⁺), providing the final stripping capacity (Q_S). The average CE was calculated using the following equation:

$${CE}_{avg}=\frac{Q_S+10\times Q_C}{Q_T+10\times Q_C}$$ 2

The voltage window for the Zn||I₂ cells was 0.3–1.8 V. The hydrogen evolution performance was investigated in a 1 M Na₂SO₄ aqueous solution with a three-electrode system using Zn deposits as the working electrode, platinum as the counter electrode, and Ag/AgCl as the reference electrode. In situ EIS was conducted every 10 min during Zn plating using an electrochemical workstation (Biologic VSP-3e). The EIS spectra were collected over the frequency range of 10⁵ to 0.05 Hz under open-circuit potential conditions with a potentiostatic mode (10 mV amplitude, 10 points per decade). For resistivity tests, the voltage response was measured at a constant current of 0.5 mA on a sample sandwiched between two stainless steel electrodes. The resistivity (ρ) was calculated using the following equation:

$$\rho =\frac{U\times S}{I\times L}$$ 3

where U is the average response voltage, S is the contact area (1.13 cm²), I is the applied current (0.5 mA), and L is the thickness of the sample (~45 μm). The electrochemical experiments were conducted in an air-conditioned laboratory environment with an average temperature of 25  ±  2 °C.

Finite element simulations

Finite element method simulations were performed using COMSOL Multiphysics. A two-dimensional model was developed to simulate the current density and electric field distribution at the negative electrode/electrolyte interface. In this model, the length of two electrodes was 15 μm and the distance between them was 10 μm. The lumpy Zn was represented by three hexagonal protrusions, each with a side length of 2 μm. Additionally, rectangular Zn with a height of 1 μm and width of 0.1 μm were incorporated on the hexagons to simulate mossy Zn growth. The ionic conductivity of the 1 M ZnOTF electrolyte filling the interelectrode gap was set to 3.2 S m⁻¹⁵⁹. The positive electrode and negative electrode potentials were held at 0.05 and 0 V (vs. Zn/Zn²⁺), respectively.

DFT calculations

First-principles DFT calculations were performed in the Vienna Ab initio Simulation Package (VASP)⁶⁰. The exchange-correlation interactions were described by the PBE functional within the generalized gradient approximation, with a plane-wave cutoff energy of 500 eV⁶¹. The electronic self-consistency criterion was set to 10⁻⁵ eV, and the force on each atom was converged below 0.03 eV Å⁻¹. The Zn substrate was modeled as a four-layer 6 × 6 supercell and a 25 Å vacuum layer was constucted. The dispersion corrections were included by the DFT-D3 scheme. Geometry optimizations were performed using a Γ-only k-mesh for the Brillouin zone sampling. The geometry relaxation was applied for the top two layers of the Zn slab⁶². The adsorption energy (E_ads) was calculated using the following equation:

$$E_{ads}=E_{total}-E_{sub}-E_{Zn}$$ 4

where E_total is the total energy of the adsorption system, E_sub is the energy of the adsorbate, E_Zn is the energy of a free Zn(002)-slab.

Author contributions

H.-M.C. and C.P.H. conceived and directed the project. H.G. designed and performed the experiments. D.J.L. and Y.Z. contributed to SEM and XRD characterizations. H.G. wrote the manuscript, and H.-M.C. and C.P.H. revised it. All authors discussed the results and commented on the manuscript.

Peer review

Peer review information

Nature Communications thanks Shaojian Zhan, Yunkai Xu, and Jinhong Lee for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data generated in this study are provided in the manuscript, Supplementary Information, and Source Data file. Source data are provided with this paper.

Competing interests

H.-M.C., C.P.H., and H.G have one China patent application related to this work (CN202510673370.X). The other authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-68844-z.