Electron delocalization-driven high dielectric electrolyte for solid-state lithium metal batteries

Abstract

Solid-state polymer electrolytes hold significant promise for solid-state lithium metal batteries but have been severely plagued by how to simultaneously achieve high ionic conductivity and construct a stable interphase. Here, we propose an electron delocalization engineering strategy by introducing the conducting polymer with the conjugated system into solid-state polymer electrolytes to solve the above issue. The electron delocalization of conducting polymer such as proton-doped polyaniline can generate tremendous “micro-capacitors” to increase dielectric constant, which significantly facilitate the dissociation of Li salts to enhance the Li⁺ transport, meanwhile they can accelerate the bond-cleavage dynamics of anions to form an inorganic-rich interphase. Solid-state polymer electrolytes with proton-doped polyaniline deliver a high ionic conductivity at a wide temperature range (0.892 mS cm⁻¹ at 25 °C and 0.053 mS cm⁻¹ at −40 °C), which enables Li electrode to achieve a large critical deposition areal capacity of 26.3 mAh cm⁻². The Li | |LiNi_0.8Co_0.1Mn_0.1O₂ cells operate effectively over a wide temperature range from −40 °C to 45 °C and exhibit a life span of over 13000 cycles at 1800 mA g⁻¹, which is highly desired for practical solid-state lithium metal batteries due to the inaccessible design for the current polymer electrolytes.

One of the main challenges in solid-state polymer electrolytes is to simultaneously achieve high ionic conductivity and a stable interphase. Here, authors propose an electron delocalization engineering strategy by introducing the conducting polymer into polymer electrolytes to realise a high ionically conducting and stable polymer electrolyte.

Introduction

Advanced solid-state batteries (SSBs) featuring lithium (Li) metal negative electrodes are widely recognized as promising candidates for next-generation energy storage due to their potential for high energy density and improved safety performance^1,2. Especially, SSBs applied to electric vehicles demand fast charging and operation across a wide temperature range³. A key enabler of such technology is the development of solid-state electrolytes (SSEs) that can ensure both high ionic conductivity and stable electrode-electrolyte interfaces^4,5. Among the various classes of SSEs, solid polymer electrolytes (SPEs) have attracted considerable interest due to their excellent interfacial compatibility with electrodes and compatibility with existing battery manufacturing processes^6–8.

Nevertheless, SPEs continue to face critical challenges, particularly their low ionic conductivity at room temperature, which arises from the limited dissociation of Li salts and the resulting low concentration of mobile Li^+9–14 (Fig. 1a and Supplementary Table 1). Strategies to address this include tailoring the interactions among Li⁺, salt anions, and ligands (polar polymer segments or tiny solvent residues absorbed by the Li salt) to promote salt dissociation and improve ion transport^15–18. In this context, incorporating dielectric ceramic fillers to enhance the effective dielectric constant (ε_r) of the polymer electrolyte has been explored as a way to weaken Li⁺–anion coordination and increase the free ion concentration^19–21. The remarkable issues for use of high ε_r ceramics are related to the aggregation caused by poor interface compatibility and the poor flexibility of composite solid-state electrolytes due to the high concentration ceramic nanoparticles (Fig. 1c), inability to form an ordered and uniform built-in field under the electric field of cells²². A small amount of conducting carbon fillers easily polarized in the polymer matrix can also enhance dielectric property of SPEs²³. However, achieving a uniform distribution and good compatibility of such fillers with the polymer host remains non-trivial and can limit the consistency of their impact on ion transport²⁴.

Fig. 1. Design strategy of solid-state polymer electrolyte.

Evolution of Li⁺ self-diffusion coefficient (a) and anion reduction potential (b) with degree of Li–anion association. c Screening criteria of fillers for the high ε_r SPEs. d Schematic illustration of dissociation of ion pairs under an electric field and their relationship with the dielectric environment. Schematic illustration of forming “micro-capacitors” (e) and expedite the bond-cleavage dynamics of anions (f) via electron delocalization of d-PANI.

Beyond ionic conductivity, stable cycling under high current densities and areal capacities remains another persistent challenge for SPE-based SSBs^25,26. The instability of the solid electrolyte interphase (SEI) under such conditions often leads to dendritic growth and performance degradation²⁷. Recent advances have shown that inorganic-rich SEIs, particularly those containing LiF, can effectively suppress Li dendrite growth and improve interfacial stability^28,29. These interphases can be promoted by engineering anion-dominated coordination environments using fluorinated salts or polymers with designed steric effects^30–32. However, a fundamental trade-off remains: the stronger Li⁺–anion coordination required for interphase formation often compromises Li salt dissociation and reduces ionic conductivity (Fig. 1b).

Addressing this inherent trade-off is essential for the development of high-performance SPEs. In this work, we construct a high-dielectric polymer electrolyte employing electron delocalization engineering via the conducting polymer such as proton-doped polyaniline (d-PANI) to simultaneously achieve high ionic conductivity and form inorganic-rich SEIs. The incorporation of compatible d-PANI can easily generate tremendous “micro-capacitors” within the polymer films to boost ε_r of SPEs to as high as 150, which significantly weakens the binding between Li⁺ and anions, facilitating Li salt dissociation (Fig. 1d, e). More mobile Li⁺ is therefore generated, which ensures a high ionic conductivity of 8.92 × 10⁻⁴ S cm⁻¹ at 25 °C. Meanwhile, the electron delocalization on conjugated π-bonds of d-PANI can expedite the bond-cleavage dynamics of S–F, and S–N bonds in bis(fluorosulfonyl)imide (FSI⁻), resulting in an inorganic-rich SEI, which enhances interfacial kinetics and suppresses Li dendrite growth (Fig. 1f). As a result, the symmetrical Li cells achieve a high critical current density (CCD) of 4.2 mA cm⁻² and stable Li plating/stripping at a high current density of 2 mA cm^–2. Notably, the Li||LiNi_0.8Co_0.1Mn_0.1O₂ (NCM811) cells exhibit a great lifespan of over 13,000 cycles at a high current density of 1800 mA g⁻¹, while operating effectively under a wide temperature range (–40 °C to 45 °C) and achieving a high specific capacity retention of 71% at −30 °C. Our work provides effective insights into solving the conflict of Li salt dissociation and interphase construction in SPEs, highlighting the potential application of SPEs for power-dense SSBs.

Results

Constructing high-dielectric polymer electrolytes with d-PANI

To validate the high-dielectric SPEs-design principles of electron delocalization engineering, the d-PANI was incorporated into poly (vinylidene fluoride) (PVDF)-LiN(SO₂F)₂ (LiFSI) electrolyte with Li_1.4Al_0.4Ti_1.6(PO₄)₃ (LATP) filler (termed PVPL) to fabricate a flexible membrane of composite polymer electrolyte (termed PLPA) (Supplementary Note 1 and Supplementary Figs. 1–9). The SEM images show that the PLPA electrolyte has a more compact structure than the PVPL electrolyte (Fig. 2a and Supplementary Figs. 10 and 11), assisting to reduce the physical contact resistance to suppress Li propagation. The incorporation of d-PANI has a negligible impact on PVDF polymer dynamics (Supplementary Fig. 12). The X-ray computed tomography and atomic force microscopy-based infrared spectroscopy demonstrate that the d-PANI is evenly dispersed in the PVDF-based electrolyte (Fig. 2b–d).

Fig. 2. Properties of high-dielectric polymer electrolytes.

SEM image (a) and X-ray tomography image (b) of PLPA electrolyte. The rendered green region represents the uniformly dispersed d-PANI. AFM image (c) and corresponding nano-infrared overlap of the C=N vibration of d-PANI (d) within PLPA electrolyte. The pink areas indicate a stronger d-PANI intensity. Electrostatic potential (ESP) mappings of d-PANI without (e) and with (f) the electric field. g Statistical comparison of potential energy of d-PANI without and with the electric field obtained from ESP. The center line of each box represents the median, while the box limits correspond to the upper and lower quartiles. The whiskers extend to the most extreme values within 1.5 times the interquartile range (IQR). The individual points represent ESP values calculated at different molecular positions. For d-PANI without and with the applied electric field, n = 98 and n = 100 ESP sampling positions were analyzed, respectively. h Dielectric constant of different films. Hysteresis loop obtained from PFM analysis of PVPL electrolyte (i) and PLPA electrolyte (j). k Electric field simulations inside PLPA electrolyte.

As illustrated in Supplementary Fig. 13, the d-PANI is composed of alternating single and double bonds, forming a conjugated system that enables delocalization of π-electrons across the polymer chains³³. The molecular electrostatic potential (ESP) derived from density functional theory calculations exhibits valuable insights into the charge distribution regions of d-PANI (Fig. 2e), where the red regions represent areas of positive ESP, corresponding to electrophilic centers situated at the nitrogen atoms within the amine (–NH–) groups³³. The localized orbital locator-π mapping demonstrates that the π-electrons are homogeneously distributed within the framework of d-PANI and the delocalized paths extend over the conjugated plane (Supplementary Fig. 14). Notably, the ESP distribution shows that the electron delocalization degree of the d-PANI chain further increases under an applied electric field (Fig. 2f, g), which in turn strengthens the response of the internal electric field. Clearly, the d-PANI with electron delocalization is isolated by the PVDF to create plenty of “micro-capacitors” to form built-in electric field^34,35 (Fig. 1e), where the conductive d-PANI particle dispersed within the insulating PVDF matrix can be viewed as microscopic “floating electrodes”, while the thin PVDF layer surrounding each d-PANI particle serves as the dielectric layer (Supplementary Fig. 15). As a result, the ε_r of PLPA film increases with the increasing content of d-PANI in the PVDF matrix. The ε_r of PLPA film with 3 wt% d-PANI reaches as high as 150, while the PVPL film only possesses ε_r values of 19 at 20 Hz and 25 °C (Fig. 2h and Supplementary Fig. 16). Since the Maxwell‑Wagner‑Sillars polarization is a direct manifestation of this “micro‑capacitor” effect in the frequency domain occurring at the interfaces between heterogeneous phases within the material, the “micro‑capacitors” effect is most pronounced in the low‑frequency region, leading to a significant enhancement in the ε_r of the composite films (Supplementary Fig. 17). The higher dielectric loss tangent (tan δ) of PLPA film further confirms the formation of the “micro‑capacitor” network (Supplementary Fig. 18). The dielectric domains of composite films were investigated by piezoelectric force microscopy³⁶. As shown in Fig. 2i, no hysteresis loop was captured for PVPL film, while the PLPA film delivers a rectangle-like hysteresis loop with a sharply modulated phase, indicating the enhanced dielectric property (Fig. 2j). In addition, finite element simulations show the strength of the internal electric field is enhanced along the d-PANI to form a uniform internal electric field (Fig. 2k and Supplementary Fig. 19). These findings demonstrate that the PLPA film exhibits excellent dielectric property driven by the electron delocalization of d-PANI. It is worth noting that the 3 wt% d-PANI exhibits a negligible influence on the electronic conductivities of the electrolytes, which can meet the requirements of practical batteries (Supplementary Table 2 and Supplementary Fig. 20). Furthermore, the PVPL electrolyte exhibits a notable oxidation current at 4.50 V versus Li⁺/Li, while a negligible response current is observed up to 4.76 V for PLPA electrolyte due to the enhanced coordination between Li⁺ and residual solvent and the Li-ion transference number of PLPA electrolyte is 0.42 (Supplementary Figs. 21–23), demonstrating their better compatibility with high-voltage positive electrodes.

Enhanced dissociation of ion pairs by d-PANI

The G_E (S V⁻¹) is defined to investigate the impact of d-PANI-induced “micro-capacitors” on dissociation kinetics of Li salts, which is denoted as the dissociation rate constant of Li salts per unit voltage (Supplementary Note 2)³⁷. The value of G_E can be determined by analyzing the relation between the ionic conductance of SPEs under different bias voltages (ΔE) and the corresponding applied ΔE, as per Eq. (1):

$${\text{G}}_{\mathrm{\Delta }\text{E}}=\text{f}\left(\mathrm{\Delta }\text{E}\right)={\text{G}}_{\text{E}}\mathrm{\Delta }\text{E}+{\text{G}}_0$$ 1

Where G_∆E and G₀ are the ionic conductance of SPEs with and without applied ΔE, respectively. The typical linear relationship between G_∆E and ΔE is demonstrated in various SPEs and a lower G_E indicates a higher degree of Li–anion association (Fig. 3a). The SPE containing 15 wt% BaTiO₃ (termed PLPB) was prepared here to compare the effects of natural dipole and “micro-capacitors” on the Li salts dissociation. The G_E of PVPL, PLPB and PLPA electrolytes is 7.73, 19.04 and 32.48 mS V⁻¹, respectively, demonstrating that the Li salt dissociation is significantly improved by the built-in field originating from the “micro-capacitors” in PLPA film under cell operating conditions. Furthermore, molecular dynamics (MD) simulations were performed to investigate the Li⁺ coordination environment (Supplementary Fig. 24 and Supplementary Data 1–8). Compared with PVPL electrolyte, the coordination number of FSI⁻ with Li⁺ (Li–O (FSI^–) and Li–N (FSI^–)) of PLPA electrolyte is decreased under an electric field (Fig. 3b, c and Supplementary Fig. 25), indicating a weaker interaction between FSI^– and Li⁺ to generate more movable Li⁺. Detailed coordination structures of Li⁺ were further analyzed, which exhibit a significant reduction of high-coordination structures for PLPA electrolyte under an electric field (Fig. 3d, e and Supplementary Fig. 26). As a result, the PLPA electrolytes deliver the enhanced ion-conduction capability under an electric field with a higher diffusion coefficient of 2.09 × 10⁻⁹ cm² s⁻¹ than that without an applied electric field (1.69 × 10⁻⁹ cm² s⁻¹), which can be attributed to the increase of Li salts dissociation degree (Fig. 3f).

Fig. 3. Evolution degree of Li–anion association and ion transport with the electric field.

a Fitting plots between G_∆E and ΔE. The relationship between G_∆E and ΔE was fitted by a linear function and G_E was determined by the slope value to exhibit evolution of ion transport with the electric field. Coordination number plots of b Li–O_FSI– pair and c Li–N_FSI– pair in electrolytes with and without the electric field. Proportion of probable coordination structures from MD simulations in the PLPA electrolyte with (d) and without (e) electric field. f Li-ion self-diffusion coefficient of PLPA electrolyte from the MD simulated mean squared displacement with and without the electric field. g Schematic illustration of Raman test. Variation of Raman spectra with the electric field in PLPA (h) and PVPL (i) electrolytes at different voltages (0 V, 1 V, 2 V, 3 V, 4 V) and 25 °C (The intensity was normalized by its maximum value to range from 0 to 1). j Evolution of normalized percentage of free anions, contact ion pairs (CIP) and ion aggregates (AGG) at different electric field stages. k Ionic conductivities of SPEs from 70 °C to −40 °C and Arrhenius plots.

To visualize the real-time Li⁺ coordination environment under battery operation, Raman spectra at different voltages were performed (Fig. 3g). The Li⁺ coordination environment according to S–N–S symmetric stretching vibrational mode of FSI⁻ are composed of free anions (720.9 cm⁻¹, free FSI⁻), contact ion pairs (730.7 cm⁻¹, one FSI⁻ bonding with one Li⁺) and ion aggregates (740.4 cm⁻¹, one FSI⁻ bonding with two or more Li⁺)³⁸ (Supplementary Fig. 27). As shown in Fig. 3h, i, the response of Li⁺–FSI⁻ complexes significantly drifts down as the voltage increases for PLPA electrolyte, while remaining unchanged for PVPL electrolyte. As the voltage rises from 0 V to 4 V, the content of free FSI⁻ in PLPA electrolyte is increased from 19.8 to 56.7% (Fig. 3j), which is attributed to an electrical response at the cell operating voltage to weaken the electrostatic interaction between Li⁺ and the anions, facilitating Li salts dissociation. Consequently, the PLPA electrolyte with 3 wt% d-PANI exhibits a much higher ionic conductivity of 8.92 × 10⁻⁴ S cm⁻¹ (25 °C) than that of PVPL electrolyte (4.97 × 10⁻⁴ S cm⁻¹), with a low activation energy (E_a) of 0.19 eV above 20 °C (Fig. 3k and Supplementary Figs. 28–30). The ionic conductivity of the PLPA electrolyte is competitive with the reported polymer electrolytes with high-dielectric fillers (Supplementary Table 3). It shows that the ionic conductivity of SPEs delivers an inflection point at 10 °C with enhanced E_a due to the restricted movement of ligands at low temperature. Nevertheless, the PLPA electrolyte possesses a lower E_a of 0.28 eV below 10 °C and a high ionic conductivity of 5.34 × 10⁻⁵ S cm⁻¹ even at −40 °C, which is an order of magnitude higher than that of PVPL electrolyte (7.32 × 10⁻⁶ S cm⁻¹). It should be noted that the content of residual DMF for PLPA and PVPL electrolytes was around 15 wt% (Supplementary Figs. 31 and 32), which suggests that the higher ionic conductivity of PLPA electrolyte is not attributed to the residual DMF. Meanwhile, pulsed field gradient nuclear magnetic resonance (PFG-NMR) was employed to measure the self-diffusion coefficients of Li⁺ (D_Li+). The fitted D_Li+ of PLPA electrolyte is 5.91 × 10⁻¹² m² s⁻¹, which is significantly higher than that of PVPL electrolyte (2.67 × 10⁻¹² m² s⁻¹) (Supplementary Fig. 33), further indicating the improved ionic conductivity primarily of PLPA electrolyte originates from accelerated Li⁺ transport. (Supplementary Note 3 and Supplementary Figs. 34–38). These results demonstrate that the PLPA electrolyte displays the fast ion-conduction capability at a wide temperature range, benefiting from the weak Li⁺–anion coordination activated by an electrical response of “micro-capacitors” derived from d-PANI. Furthermore, when the d-PANI was adopted in poly(ethylene oxide)-based and polyacrylonitrile-based electrolytes, the ionic conductivities also exhibit significant increase (Supplementary Note 4 and Supplementary Figs. 39, 40), demonstrating the universality of the electron delocalization strategy in the conductivity enhancement in various polymer electrolytes.

Chemistry of the interphase induced by d-PANI

Ab initio molecular dynamics (AIMD) simulations were employed to understand the role of the electron delocalization of d-PANI in SEI formation (Fig. 4a, b, Supplementary Fig. 41 and Supplementary Data 9, 10). The preferential reduction of FSI⁻ in PLPA electrolyte is captured and complete breakdown of FSI⁻ into LiF, Li₂O, Li₂S and Li₃N species is observed, corresponding to an eight-electron transfer process (Fig. 4c–e). In contrast, the FSI⁻ in PVPL electrolyte undergoes incomplete decomposition throughout the simulation process, and only the cleavage of the S–F bond is captured. The atomic-charge analyses and bond-distance of FSI⁻ reveal the order of the bond-cleavage processes³⁹. Specifically, in PLPA electrolyte, the complete cleavage of the S–F atom occurs at around 0.25 ps, followed by the breaking of two S–O1 bonds at about 0.40 ps, yielding the decomposition products of N(SO)₂, LiF and Li₂O. Subsequently, the one N–S bond in N(SO)₂ and the two S–O2 bond completely break at approximately 0.68 ps to form Li₂S and Li₂O (Supplementary Fig. 42). Finally, the remaining N–S bond cleavage emerges at around 2.38 ps, leading to the formation of Li₃N and Li₂S, thus implying a complete breakdown of FSI⁻. By contrast, for PVPL electrolyte, the breaking of the first S–F bond is observed at approximately 0.64 ps, accompanied by the decomposition product of N(SO₂)₂F. The secondary release of F atoms from the N(SO₂)₂F anion occurred at 1.20 ps and only the cleavage of S–F bond to form LiF is captured in PVPL electrolyte throughout the simulation with 5 ps, indicating a sluggish and incomplete decomposition pathway. Meanwhile, Fig. 4e displays a more pronounced electron transfer activity from both PLPA electrolyte and Li negative electrode to FSI⁻, compared to PVPL electrolyte, confirming that the electron delocalization of d-PANI accelerates bond-cleavage kinetics of FSI^–. DFT calculations reveal that this strong interaction between d-PANI and FSI⁻ breaks the structural symmetry of the FSI⁻ anion, causing the S–F bond lengths to diverge (1.628 Å vs. 1.665 Å). This structural distortion significantly lowers the reduction free energy of FSI⁻ (Supplementary Fig. 43), rendering the anion thermodynamically more susceptible to reductive decomposition.

Fig. 4. AIMD simulation of FSI⁻ decomposition and the electrolyte/electrode interphase analysis.

The representative snapshots of 0, 0.5, 2, 3 and 5 ps from AIMD simulations in PVPL (a) and PLPA (b) electrolytes. Purple, lithium; blue, nitrogen; red, oxygen; yellow, sulfur; green, fluorine. Temporal evolution of changes in the atomic charge of two different S atoms (c), the bond distance of S–F, S–N and S–O1 bonds (d), and the charge-transfer processes from SPEs and Li metal to FSI⁻ (e) during AIMD simulations. f F and S element atomic concentrations in the SEI formed with electrolytes after 50 times Li plating/stripping at 0.1 mA cm^–2, 0.1 mAh cm^–2 (25 °C) for different Ar⁺ sputtering times. g F 1s XPS of SEI formed with the electrolyte with sputtering time of 0 s. h, i Cryo-TEM images of SEI on deposited Li with a capacity of 0.5 mAh cm⁻² at 0.1 mA cm⁻² (25 °C) in the PLPA electrolyte and corresponding FFT images from different locations of deposited Li (the shaded areas represent the inorganic components). Atomic structure of Li₂O (j) and Li₃N (k) inside the SEI. l EDS maps of the deposited Li.

The cyclic voltammetry (CV) of Li||Cu cell exhibits a higher intensity of cathodic peak corresponding to the reduction of FSI^– for PLPA electrolyte, implying the faster reduction kinetics of FSI^– (Supplementary Fig. 44). Depth-profiling X-ray photoelectron spectroscopy (XPS) was utilized to investigate the chemical composition of the SEI. A higher and depth-consistent signal density of inorganic species (including LiF, Li₂S, etc.) is observed in PLPA-driven SEI and surpasses that with PVPL electrolyte (Fig. 4f, g and Supplementary Figs. 45–47). Furthermore, the atomic ratio of F and S elements in PLPA-driven SEI is higher than that of PVPL-driven SEI (Fig. 4f and Supplementary Fig. 48), reflecting the anion-derived inorganic-rich SEI. The spatial distribution of the SEI was revealed by the time-of-flight secondary-ion mass spectrometry. As shown in Supplementary Fig. 49, a series of inorganic fragments of PLPA-driven SEI (such as Li₂F₃⁻, LiS⁻, LiO⁻, Li₂N⁻) exhibit a higher concentration and a more uniform distribution than PVPL-driven SEI. These findings demonstrate that enhanced FSI⁻ decomposition derived from the electron delocalization of d-PANI in PLPA electrolyte contributes to the formation of a homogeneous and compact SEI predominantly composed of inorganic species, which is consistent with the AIMD results.

Beyond, the high-resolution cryogenic transmission electron microscopy (cryo-TEM) images show PLPA-driven SEI exhibit a typical mosaic structure composed of crystalline nanoparticles (marked by green and pink) embedded in an amorphous phase (Fig. 4h, i and Supplementary Fig. 50). The crystalline nanoparticles, including Li₂O and Li₃N, are confirmed by the corresponding lattice spacing (Fig. 4j, k), where the Li₃N possesses a quite high ionic conductivity to enhance Li⁺ transport of the SEI⁴⁰. Although the crystalline LiF and Li₂S are not observed in high-magnification cryo-TEM images, energy-dispersive X-ray spectroscopy (EDS) maps show the uniform distribution of F and S elements in PLPA-driven SEI (Fig. 4l and Supplementary Fig. 51). This suggests that the inorganic phases of LiF and Li₂S are present in a highly disordered or even amorphous form⁴¹, which can achieve fast Li⁺ transport to enhance interfacial dynamics according to the previous research^42,43.

Improved interfacial stability and dynamics of Li negative electrodes by d-PANI

The interfacial dynamics and feasibility of the SEI were evaluated, which is vital for the Li deposition and stripping. The distribution of relaxation times (DRT) analysis of temperature-dependent electrochemical impedance (EIS) in the Li||Li cell shows that the Li⁺ diffusion across the SEI (R_SEI) with PLPA electrolyte remains quite lower than those observed with PVPL electrolyte across different temperatures (Fig. 5a, b and Supplementary Fig. 52). As the temperature decreases, R_SEI of PVPL electrolyte undergoes a significant increase, suggesting that the process of Li⁺ migration through SEI is critically limited at low temperature. By contrast, only a slight increase of R_SEI using PLPA electrolyte emerges with the reduced temperature, which is also supported by their lower E_a (36.08 kJ mol⁻¹) of Li⁺ migration through SEI than that of PVPL electrolyte (64.13 kJ mol⁻¹), as shown in Fig. 5c and Supplementary Fig. 53. The SEI derived from PLPA electrolyte also exhibits higher exchange current density (j₀) than that PVPL electrolyte at both 25 °C and −10 °C (Fig. 5h and Supplementary Fig. 54). In addition, the 2D exchange spectroscopy solid-state nuclear magnetic resonance (2D-EXSY ssNMR) experiments were carried out to directly detect the spontaneous Li⁺ diffusion between the deposited Li and SEI⁴⁴. The more intense cross-peaks (indicated by the dashed yellow box) associated with PLPA electrolyte indicate more equilibrium Li⁺ exchange between the deposited Li and inorganic-rich SEI (Fig. 5d, e and Supplementary Fig. 55). These results indicate the inorganic-rich SEI derived from PLPA electrolyte provides fast Li⁺ transport kinetics at a wide temperature range. The atomic force microscopy (AFM) was conducted to evaluate the mechanical properties of the SEI. The average Derjaguin–Müller–Toporov (DMT) modulus of the SEI derived from PLPA electrolyte is 13.84 GPa, which is higher than the DMT modulus (4.32 GPa) of PVPL-driven SEI (Fig. 5f).

Fig. 5. Li negative electrode stability and interfacial dynamics.

Temperature variable DRT analysis for the Li||Li cells with PVPL (a) and PLPA (b) electrolytes. c Arrhenius behavior of the R_SEI and the calculated activation energy for Li⁺ diffusion through SEI. 2D ⁷Li-⁷Li exchange spectroscopy of Li and its corresponding SEI induced by PVPL (d) and PLPA (e) electrolytes at 25 °C. f Distribution of DMT modulus of the cycled Li metal negative electrode after 50 times Li plating/stripping at 0.1 mA cm^–2, 0.1 mAh cm^–2 (25 °C) using the electrolytes. g Galvanostatic discharge/charge voltage profiles of the Li||Cu cells at 0.1 mA cm⁻² with the electrolytes and SEM images of Li deposited on Cu foils at 0.1 mA cm⁻² (25 °C). h Exchange current density obtained from Tafel plots of Li||Li cells using the electrolytes at 25 °C and −10 °C. Galvanostatic Li plating/stripping curves of the Li||Li cells using the electrolytes at 2 mA cm⁻², 1 mAh cm⁻² and 25 °C (i) and 0.1 mA cm⁻², 0.1 mAh cm⁻² and −30 °C (j).

The critical deposition areal capacity (CDC) was tested by Li||Cu cells to evaluate the potential applications at a large areal capacity (Fig. 5g). A high CDC of 26.3 mAh cm⁻² is achieved for PLPA electrolyte, with compact ordered columnar structures of deposited Li and a thickness of 105 μm even at 20 mAh cm⁻². In contrast, the PVPL electrolyte shows a much lower CDC of 4 mAh cm⁻² with a highly porous structure of the deposited Li, accompanied by uncontrolled side reaction products, which is consistent with the results of X-ray computed tomography (Supplementary Fig. 56 and Supplementary Note 5). Besides, the effect of the inorganic-rich SEI on the stability of Li negative electrodes was evaluated by measuring CCD and cycling stability in Li||Li cells. For PVPL electrolyte, short-circuiting of the Li||Li cells quickly emerges at 1.4 mA cm⁻² due to inorganic-less SEI. In sharp contrast, the PLPA electrolyte exhibits a high CCD of 4.2 mA cm⁻², enabling the operation of cells at high current density (Supplementary Fig. 57). Meanwhile, the results of the CCD indicate that under high current densities, a dense structure of polymer electrolyte only provides partial physical blocking to Li propagation, while not achieving a homogeneous Li deposition that demands a robust SEI with fast transport kinetics to enable interfacial stability against Li metal (Supplementary Note 6 and Supplementary Fig. 58). The Li||Li cells with PLPA electrolyte enable stable Li plating/stripping for 3200 h at 0.1 mA cm⁻² and 25 °C, while the Li|PVPL|Li cells show short-circuit after 1570 h of cycling (Supplementary Fig. 59). When Li plating/stripping at a high current of 2 mA cm⁻², the Li|PVPL|Li cells fail within 70 h owing to rapid overpotential increase (Fig. 5i), while the Li|PLPA|Li cells stably Li plating/stripping for 1200 h at 25 °C. Furthermore, the Li|PLPA|Li cells still enable Li plating/stripping for 120 h at 3 mA cm⁻² (Supplementary Fig. 60). Even at a large capacity of 5 mAh cm⁻², the Li|PLPA|Li cells maintain consistent Li plating/stripping performance over 1500 h, rivaling those of many reported SPEs (Supplementary Fig. 61 and Supplementary Table 4). More excitingly, the Li|PLPA|Li cells are capable of Li plating/stripping for 1450 h at a low temperature of −30 °C, while the Li|PVPL|Li cells can hardly sustain a Li plating/stripping (Fig. 5j). These results exhibit the enhancements for Li negative electrode with PLPA electrolyte, elucidating the vital role of the electron delocalization-driven SEI in regulating Li morphology and interfacial kinetics, which contributes to the Li plating/stripping stability of solid-state batteries at extreme conditions. The d-PANI and residual solvents coordinated with Li⁺ in PLPA electrolyte have the negligible side reactions with Li negative electrode. However, compared with the state-of-the-art ether-based electrolytes, Li reversibility should be further improved by fundamentally solving the intrinsic nature of the composite electrolyte containing residual DMF (Supplementary Note 7 and Supplementary Figs. 62–66).

Performance of Li||NCM811 cells with d-PANI at various conditions

Operando XRD measurements show that the NCM811 paired with PLPA electrolyte demonstrates the typical H1-M-H2-H3 phase transition characteristics with sufficient reversibility⁴⁵, while the incomplete phase transition and coexistence of various phases are observed with PVPL electrolyte due to sluggish Li⁺ transfer kinetics (Fig. 6a, b). This is also confirmed by the CV experiments at different scanning rates (Supplementary Fig. 67). As a result, the Li|PLPA|NCM811 cells display an enhanced rate performance, delivering a high specific capacity of 128.3 mAh g⁻¹ at 900 mA g^–1 (Fig. 6c), which is attributed to their small impedances (Supplementary Fig. 68). Long-term cycling of Li||NCM811 cells with electrolytes were evaluated under multiple harsh conditions. The Li|PLPA|NCM811 cells exhibit a capacity retention of 64% over 4000 cycles at 900 mA g^–1, whereas the Li|PVPL|NCM811 cells suffer from rapid capacity decay and complete failure after 2000 cycles (Fig. 6d). Notably, the Li|PLPA|NCM811 cells deliver a capacity of 110.7 mAh g⁻¹ at 1800 mA g^–1 and present a lifespan of 13,000 cycles. In contrast, the Li||NCM811 cells with PVPL electrolyte show a severe capacity drop under the same conditions (Fig. 6f). Based on two criteria, including charging current density and cycle life, the performance of the Li||NCM811 cells with PLPA electrolyte is competitive with that of the cells with reported SPEs (Fig. 6g and Supplementary Table 5)^{8,13,14,20,27,31,46–49}, illustrating the effectiveness of our SPEs-design concept in improving the fast-charging performance of SSBs. Furthermore, the structure and composition of the LATP and d-PANI can be well retained in the electrolytes after cycling (Supplementary Figs. 69 and 70), and the electronic conductivity of electrolyte shows a negligible increase after cycling (Supplementary Fig. 71).

Fig. 6. Electrochemical performance of solid-state Li||NCM811 cells.

Operando XRD characterization for the NCM811 positive electrode during the initial and second cycle with PVPL (a) and PLPA (b) electrolytes at 18 mA g^–1 (25 °C). c Charge-discharge curves of the Li||NCM811 cells at different rates for PVPL (top) and PLPA electrolyte (bottom) at 25 °C within 2.8 to 4.3 V. Cycling performance of the Li||NCM811 cells with different electrolytes at 25 °C (each cycled for three cycles at 36 mA g^–1) and then 900 mA g^–1 (d) and 1800 mA g^–1 (f) for long cycling within 2.8 to 4.3 V; long cycling with high positive electrode loading within 2.8 to 4.3 V (e). g Comparison of solid-state Li metal cells based on SPEs in previous reports and this work (NCM811, 1C = 180 mA g⁻¹; LFP, 1C = 145 mA g⁻¹; SPAN, 1C = 811 mA g⁻¹; The source of the literature data shown in this figure can be found in Supplementary Information, Table 5). h Cycling performance of the Li||NCM811 pouch cells at 180 mA g^–1 and 25 °C within 2.8 to 4.3 V; i Cycling performance of the Li||NCM811 cells with different electrolytes at 18 mA g^–1 and −30 °C within 2.8 to 4.5 V.

To further demonstrate the practical application of electrolyte, high-loading NCM811 positive electrodes were used to assemble the Li metal batteries. As shown in Fig. 6e, the Li|PLPA|NCM811 cells with high positive electrode loading of 1.8 mAh cm⁻² deliver a capacity retention of 80% after 170 cycles at 90 mA g^–1. Even when positive electrode loadings increase to 3.2 mAh cm⁻², the Li|PLPA|NCM811 cells show stable operation over 170 cycles. The pouch cells with PLPA electrolyte maintain a capacity retention of 93% after 200 cycles at 90 mA g^–1 and show a stable cycling performance with a capacity retention of 82% after 360 cycles at 180 mA g^–1 (Fig. 6h and Supplementary Fig. 72). The Li||NCM811 cells were further evaluated in a wide temperature window ranging from −40 to 45 °C. The Li|PLPA|NCM811 cells exhibit stable cycling performance with a capacity retention of 64% after 450 cycles at 45 °C, benefiting from the robust interphase (Supplementary Fig. 73). Meanwhile, the Li|PLPA|NCM811 cells demonstrate the enhanced low-temperature performance, exhibiting great rate capability and cycling performance at −30 °C compared to the PVPL electrolyte (Fig. 6i and Supplementary Fig. 74a). Remarkably, even at −40 °C, the Li|PLPA|NCM811 cells still exhibit reversible capacities (Supplementary Fig. 74b). These results demonstrate the advancement of PLPA electrolyte for Li||NCM811 cells under high rate and low temperature, thus confirming the effectiveness of electron delocalization engineering derived from d-PANI to enable the SPEs simultaneously possess fast Li⁺ conduction and inorganic-rich interphase.

Discussion

In summary, we proposed a robust strategy of electron delocalization engineering to address the longstanding contradiction between high salt dissociation and the formation of inorganic-rich interphase in SPEs. This strategy involves promoting the dissociation of Li salts to produce more movable Li⁺ in SPEs through an electrical response of “micro-capacitors” activated by electron delocalization within conducting polymer such as d-PANI, and simultaneously realizing the formation of inorganic-rich SEI owing to enhanced FSI⁻ decomposition dynamics driven by electron delocalization of conducting polymer. Based on this principle, a unique SPE with the d-PANI was designed to achieve fast Li⁺ conduction kinetics of bulk electrolyte and a stable electrode/electrolyte interface, enabling enhanced performance of cells under fast charging and low temperature conditions. As a result, the PLPA electrolyte achieves a high CCD and CDC of 4.2 mA cm⁻² and 26.3 mAh cm⁻² on Li negative electrodes, respectively. The Li||NCM811 cells with PLPA electrolyte exhibit long cycling lifespans of 4000 cycles at 900 mA g⁻¹ and 13,000 cycles at 1800 mA g⁻¹, and deliver a great electrochemical performance at a wide operating temperature range from −40 to 45 °C. This work describes an encouraging avenue towards the application engineering of SPEs for advanced solid-state batteries operating at extreme conditions.

Methods

Preparation of solid-state polymer electrolytes

PVDF powder (Mw = 600,000, Kynar 761, 99%) and polyaniline (sulfonate acid-doped PANI (d-PANI), Shanghai Aladdin Biochemical Technology Co., Ltd., 98%) were dried for 24 h in a vacuum oven prior to use. The precursor slurry of the PVPL electrolyte was made by mixing 400 mg PVDF, 267 mg LiFSI (Guangdong Canrd New Energy Technology Co., Ltd., 99.9%), 60 mg LATP (Guangdong Canrd New Energy Technology Co., Ltd., 99%) and 15 mL DMF (Aladdin, 99.9%) under magnetic stirring for 6 h and the PVPL slurry was dried at 55 °C for 24 h to form PVPL electrolyte. The precursor slurry of PLPA electrolyte was mixed by ball milling. A certain amount of d-PANI and LATP were first placed in an agate jar (100 mL) along with 50 g agate balls (diameters of 5 mm and 8 mm) and ball milled at 450 rpm for 3 h, followed by the sequential addition of 400 mg PVDF powder and 15 mL DMF, and the mixture was further milled for 3 h. Then, 267 mg LiFSI was added to the precursor solution of ball milling, which was magnetically stirred for 2 h. The PLPA slurry was dried at 55 °C for 24 h to form PLPA electrolyte. The manufacturer of ball milling equipment is Nanjing Nan Da Instrument Co., Ltd (QM-3SP-CL).

Electrode preparation and solid-state cell assembly

LiNi_0.8Co_0.1Mn_0.1O₂ (NCM811, Guangdong Canrd New Energy Technology Co., Ltd), PVDF 5130 (Mw = 1,200,000, Arkema, 99%) and Super P (Guangdong Canrd New Energy Technology Co., Ltd.) were dried at 60 °C for 24 h in the vacuum oven prior to use. The positive electrode was prepared by mixing NCM811 (0.8 g), Super P (0.1 g), PVDF 5130 binder (0.1 g) in a weight ratio of 8:1:1 in 2.7 mL N-methyl-2-pyrrolidone (NMP, Shanghai Aladdin Biochemical Technology Co., Ltd., 99.9%), followed by casting the slurry on an aluminum foil (10 μm thickness, purchased from Guangdong Canrd New Energy Technology Co., Ltd.) and then it was dried at 60 °C under vacuum to remove the solvent and bound water. The porosity of the NCM811 electrodes is about 54.57%. Here, the loading of the NCM811 material is 1.0 mg cm^–2. The other positive electrodes with high loadings were also prepared using NCM811, Super P, PVDF 5130 with a 9:0.5:0.5 weight ratio. The CR2032 solid-state cells were assembled in an Ar-filled glove box (O₂ < 0.1 ppm, H₂O < 0.1 ppm). The Li||NCM811 cell was assembled by placing PLPA or PVPL electrolyte (around 100 μm thickness) between a large Li metal piece (the diameter of 15.6 mm, 450 μm in thickness, China Energy Lithium Co., Ltd.) and a positive electrode (diameter of 12.0 mm). The symmetric Li cell used a large Li metal piece on one side and a small Li metal piece with the diameter of 12.5 mm (China Energy Lithium Co., Ltd.) on the other side. Li metal pieces were directly used without any pre-activation or pre-treatment. The Li||Cu cell used a copper foil (10 μm thickness, purchased from Guangdong Canrd New Energy Technology Co., Ltd.) with the diameter of 14 mm on one side and a small Li metal piece on the other side. The pouch cells were prepared in a dry room with the dew point at −60 °C. The positive electrode (3 × 3.5 cm) and the negative electrode (4 × 4.5 cm, 50 μm Li coated on 10 μm Cu, China Energy Lithium Co., Ltd.) were stacked on the opposite sides of electrolyte (5 × 5.5 cm). The proper pressure of 700 kg (4.9 × 10⁶Pa) was applied to ensure excellent interfacial contact between electrolyte and electrode.

Electrochemical measurements

Electrochemical impedance measurements were conducted on a VMP3 multichannel workstation (Bio-Logic Science Instruments, France) to determine the ionic conductivity of the electrolytes. The spectra were collected from 7 MHz to 0.5 Hz under a sinusoidal voltage amplitude of 10 mV and 6 frequency points per decade. Symmetric stainless-steel (SS) cells were employed, with a small electrolyte positioned between two blocking electrodes of SS (the diameter of 15.8 mm, 1 mm in thickness). The ionic conductivities (σ) were calculated following Eq. (2):

$$\sigma =\frac{\text{L}}{\text{R}\text{S}}$$ 2

where L is the thickness of electrolytes, R is the ohm impedance obtained by EIS measurement, and S is the area of stainless steel. For ionic conductance under different voltages, the EIS of SS||SS cells were performed with different bias voltages (0, 0.02, 0.04, 0.06, 0.08, 0.10, 0.12, 0.14, 0.16, 0.18, and 0.20 V). The activation energy was determined by fitting the ionic conductivities at different temperatures according to the Arrhenius law Eq. (3):

$$\sigma ={\sigma }_0\exp \left(-\frac{E_a}{\text{RT}}\right)$$ 3

where σ is the ionic conductivities at different temperatures, σ₀ is the pre-exponential factor, E_a is the activation energy of ions transport, R is the standard gas constant and T is the absolute temperature.

The Chronoamperometry method was used by setting a certain control voltage-U (such as 1 V) and recording the response current curve with time. After the current value is stable, the current value of I_s (contributed by electron transport fully) was recorded, then the resistance was calculated

$$R=\frac{U}{I_s}$$ 4

The R was substituted into the equation below to calculate the electronic conductivity of the electrolyte

$${\sigma }_{\text{e}}=\frac{L}{RS}$$ 5

Where σ_e, L, and S is the electronic conductivity, thickness, and cross-sectional area of the solid-state electrolyte, respectively.

The electrochemical stability window was examined by the linear sweep voltammetry (LSV) using a VMP3 multichannel electrochemical station, with the potential scanned from 0 to 6 V versus Li/Li⁺ at 0.1 mV s⁻¹. The cyclic voltammetry (CV) tests were carried out in Li||NCM811 cells between 2.8 and 4.3 V and the successive CV cycles were collected at each scan rate of 0.05, 0.1, 0.2, 0.3, 0.4, and 0.5 mV s⁻¹. Galvanostatic charge and discharge tests of cells were performed on a LAND battery testing system and NEWARE battery test system with a voltage range from 2.8 to 4.3 V at 45 and 25 °C. For cells at extreme temperature, the Li||NCM811 cells were cycled at 36 mA g⁻¹ for the three cycles at 25 °C, followed by long cycling at 18 mA g⁻¹ with a voltage range from 2.8 to 4.5 V at low temperatures (−30 °C and −40 °C). The cycled cells were transferred into a glovebox and disassembled for further examination. The deposited Li and cycled Li metal negative electrode were transferred into a chamber with a sealed Ar-filled vessel for SEM and XPS examination. The presented results are based on measurements of individual cells.

Characterizations

X-ray diffraction (XRD) patterns were recorded using a Bruker D8 Advance diffractometer equipped with Cu Kα radiation at a scan rate of 10 deg min⁻¹ and step size of 0.01 deg. For operando measurements, XRD patterns were continuously acquired during charging and discharging at 18 mA g⁻¹ and 25 °C with a scan rate of 8 deg min⁻¹. The morphology was analyzed by scanning electron microscope (SEM, HITACH S4800) at 5 kV with energy dispersive spectroscopy. Field emission transmission electron microscopy (FE-TEM, FEI Tecnai F30) was employed for structural characterization. For cryogenic imaging, a Thermo Fisher Scientific Spectra 300 aberration-corrected microscope with a cold field-emission gun was operated at 200 kV. The samples were transferred and maintained under cryogenic conditions using a Fischione M2550 vacuum-transfer holder, with the specimen temperature kept at −176 °C. TOF-SIMS measurements were carried out on a Nano TOF-2 instrument (ULVACPHI, Japan) equipped with a Bi3⁺⁺ beam (30 kV) cluster primary-ion gun. X-ray photoelectron spectroscopy (XPS) measurement was collected on a PHI 5000 VersaProbe II instrument via a sealed transfer device to avoid air exposure, equipped with a monochromatic Al-K_α X-ray source and a beam size of 200 μm. TGA was performed using a Netzsch STA 449F3 thermal analyzer from 30 to 600 °C at a heating rate of 10 °C min^–1 in N₂ atmosphere. The AFM-nano-infrared examinations were performed with a Bruker Anasys nanoIR2-fs instrument. The Raman spectra were obtained by HORIBA LabRAM HR800. AFM system (Bruker Dimension Icon) was employed to observe the roughness and Young’s modulus of the cycled Li metal in Ar-filled glove box. The dielectric property of the electrolytes was investigated by AFM with piezoelectric force microscopy (PFM). The solid-state nuclear magnetic resonance (NMR) was performed in a 9.4 T magnetic field with a Bruker 400 MHz AVANCE III spectrometer. A Bruker HFX 3.2 mm magic-angle spinning (MAS) probe was used, and the MAS spin rate was set to 5 kHz.

DFT calculations

The stable structure of LiFSI, DMF, d-PANI and PVDF molecules was optimized by the quantum chemistry calculation with the GAUSSIAN16 software package⁵⁰ at the b3lyp/6-311++G (d, p) level of theory⁵¹. For the calculations involving interaction energies and reduction free energies, an implicit solvent model (IEFPCM) was further employed using DMF as the solvent with ε_r = 36.7. A vibrational frequency analysis was then performed at the same basis set to ensure that the optimized structure is a true energy minimum. All complex structures were initially explored by conformational searches using the Molclus program. The obtained conformers were then ranked by their relative energies, and the lowest-energy structure was chosen for further DFT calculations. To achieve better computational efficiency, the PVDF chain is represented by a short oligomer consisting of 6 CH₂CF₂ monomers, and the d-PANI is represented by 4 C₆H₇N²⁺ plus 2 p-toluenesulfonate anions. The surface ESP was analyzed and visualized using the Visual Molecular Dynamics (VMD) software⁵².

Finite element simulation

The local distribution of electric field and current density simulation was calculated by creating a 2D model with the electrostatic potential. The domain was created with the size shown in Fig. 2k and Supplementary Fig. 19, the top line was at a 0 V electric potential, and the bottom line was at a 3 V electric potential. Particle areas were set as charge conservation domain with dielectric constant of 150. The remaining areas were set as another charge conservation domain with dielectric constant of 19.

Classic molecular dynamics simulations

All classic molecular dynamics simulations were conducted with the Gromacs 2021 program. The General Amber Force Field (GAFF) was used⁵³. The charges of Li⁺/FSI⁻ are scaled by 0.8 to describe the ion–ion and ion–dipole interactions⁵⁴. The atomic charges for electrolyte molecules are assigned with RESP charges, which are calculated based on the quantum chemistry calculation by the Multiwfn software⁵⁵. Four modeling systems were constructed in the simulation boxes (size of 60 × 60 × 60 Å), and the detailed simulation system compositions are listed in Supplementary Table 6. For each system, the initial energy minimization (energy and force tolerance of 10^–5) was performed to obtain the ground-state structure. After minimization, the simulation systems were equilibrated in an isothermal-isobaric (NPT) ensemble for 2 ns at 500 K, for 3 ns at 330 K, and 20 ns at 298 K. Then the systems were equilibrated in a constant volume and temperature dynamics (NVT ensemble) for 10 ns at 298 K and followed by a final production simulation of 20 ns with NVT ensemble at 298 K. The static electric field of two systems was achieved using the GROMACS electric field module. The snapshots of MD structures were obtained using the VMD software. The coordination structure was analyzed using the MDAnalysis Python package⁵⁶.

Ab initio molecular dynamics simulations

Ab initio molecular dynamics calculations were conducted using CP2K/Quickstep⁵⁷. The electronic wave functions were described using the Gaussian and plane-wave (GPW) scheme in combination with pseudopotentials. The exchange–correlation energy was treated at the generalized-gradient-approximation level using the Perdew–Burke–Ernzerhof (PBE) functional. All elements were treated with Goedecker−Teter−Hutter (GTH) pseudopotentials and the double‑ζ MOLOPT basis set (DZVP − MOLOPT − SR − GTH)^58–60. A kinetic energy cut-off of 500 Ry has been used. The long-range dispersion was accounted for using the DFT-D3 corrections. An electronic energy convergence criterion of 1 × 10⁻⁵ eV was used for simulations. The Li-metal slab was first optimized with a vacuum layer of above 30 Å. Initial geometries of a Li-metal slab and the electrolyte were created using the Gromacs program and then were optimized by energy minimization using molecular mechanics. Ab initio molecular dynamics simulations were then performed at a temperature of 300 K using a canonical ensemble (NVT) for 5 ps with a time step of 1.0 fs.

Author contributions

Y.-B.H. and X.A. conceived the idea. Y.-B.H. and P.S. supervised the project. Y.-B.H., X.A., L.C., and K.Y. designed the experiments. X.A. performed the experiments with help from K.Y., L.C., Y.L., J.M., G.X., Z.H., Y.M., S.G., P.S., and M.L. L.C. performed the theoretical calculations. W.C. performed the cryo-TEM experiments. All authors discussed the results in the manuscript. X.A., Y.-B.H., K.Y., L.C., D.L., and F.K. wrote and revised the initial paper, which was approved by all authors.

Peer review

Peer review information

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

Funding

Y.-B.H. discloses support for the research of this work from National Science Fund for Distinguished Young Scholars [grant number 52325206], the National Key Research and Development Program of China [grant number 2021YFF0500600], Energy Revolution S&T Program of Yulin Innovation Institute of Clean Energy [grant number E511060817], Shenzhen Technical Plan Project [grant numbers JCYJ20241202124002004 and JCYJ20220818101003007], Shenzhen Pengrui Young Faculty Program Research Plan [grant number SZPR2023006]. M.L. discloses support for the research of this work from the National Natural Science Foundation of China [grant number 52203298], Shenzhen Technical Plan Project [grant number JCYJ20220530143012027]. K.Y. discloses support for the research of this work from the National Natural Science Foundation of China [grant number 523B2022], Shenzhen Science and Technology Program [WDZC20231126160733001].

Data availability

All other relevant data that support the findings of this study can be obtained from the corresponding authors upon request. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77650-6.