Constructing an anion-capturing interface to achieve Li⁺ cross-phase transport in composite solid electrolytes

Abstract

The ionic conductivity of solid electrolytes is still insufficient to approach performance promises of solid-state Li metal batteries, suffering from their charged interfaces among phase components and movable Li⁺ concentration. Herein, an anion-capturing interface based on FeF₃ is established on Li_6.5La₃Zr_1.5Ta_0.5O₁₂ surface through a sol-gel method. It promotes Li-salt dissociation and formation of anion aggregated layer before blending with polymer. Coulombic interaction of anion on grains boundary showcases multiple merits, including their weakened built-in electric field, restrained charge gradient layer, spontaneous Li⁺ cross-phase migration, and homogenized interfacial charge distribution. As such, the resulting composite solid electrolytes exhibits an ion conductivity of 1.1×10⁻⁴ S/cm² and Li⁺ migration number of 0.75 at 25°C. Its resulting Li symmetrical batteries maintain Li plating/stripping behaviors for over 1300 h and low polarization at 0.1 mA/cm² current density. When being assembled with LiFePO₄ positive electrode in solid-state batteries, it performs a specific capacity of 152.8 mAh/g at 1.0 C (170 mA/g) with 96% retention after 600 cycles. This work prioritizes the promises of interface engineering for solid electrolytes in solid-state Li metal batteries.

Composite solid electrolytes face sluggish inter-phase Li⁺ migration. Here, authors construct a FeF₃-based anion-capturing layer onto Li_6.5La₃Zr_1.5Ta_0.5O₁₂ nanofillers within PVdF solid electrolyte, where the charge gradient layer on the interface promotes Li⁺ cross-phase conductivity.

Introduction

The nonflammable solid electrolytes differ solid-state Li metal batteries (ssLMBs) from counterparts based on liquid electrolytes, considering their merits in safety and energy density^1–4. The candidates include inorganic solid electrolytes (ISEs), polymer solid electrolytes (PSEs) and composite solid electrolytes (CSEs). Among them, CSEs integrate flexibility of PSEs and high ionic conductivity of ISEs⁵. Polyvinylidene fluoride hexafluoropropylene (PVDF-HFP) has been widely investigated as CSEs polymer matrix, while performance barrier still remains towards practical applications of ssLMBs^5,6. One feasible strategy is to incorporate with inorganic conductive fillers, such as Li_6.5La₃Zr_1.5Ta_0.5O₁₂ (LLZTO), aiming to improve their film forming capability, flexibility, and electrochemical window^7,8. However, only continuous chain of LLZTO particles allows Li⁺ transport, and the difference in electronegativity and electrochemical potential between LLZTO and PVDF-HFP result in forming thick interface layer concentrated with charges, i.e., the charge gradient layer (CGL)^9,10. These CGLs epitomize heterogeneous ion-distribution phenomena at LLZTO and PVDF-HFP interfaces, characterized by ion concentration gradients. Specifically, LLZTO act as charge carriers, inducing a TFSI⁻ interfacial layer electrically neutralized by an outward-diffusing Li⁺ layer, resulting in Li deficient region in PVDF-HFP due to the Coulomb repulsion towards Li⁺. The Li-deficient regions does not possess enough charge carriers for efficient ion transport, thereby hindering the migration of Li⁺ across phase boundaries between PVDF-HFP and LLZTO^9–11. Therefore, constraining CGL influence is essential to break through bottleneck in ion conductivity of CSEs.

Interface modification plays a pivotal role in mitigating these CGL issues. Traditionally, high levels of anions in CSEs can be detrimental, as it decreases the efficiency of Li⁺ migration and lead to an increase in interfacial side reactions, without considering the potential benefits of anions^12,13. Fortunately, anions tend to form Coulombic interactions with positive charges at grain boundaries, followed by charge redistribution, thus reduces CGL thickness with the narrowed potential gap^14,15. For example, electronic polarization of ferroelectric materials such as BaTiO₃ to directionally separate positive and negative charges, and rearrange its anion-adsorbing behaviors to weaken CGL effect¹⁴. BaTiO₃ only exhibits electronic polarization when charged, a process that involves overcoming the coordination effect of the polymer and the challenges of interface incompatibility during anion adsorption. F in metal fluoride exhibits electron-deficient feature as spontaneously anion absorbers, owing to its strong electronegativity and electron-withdrawing capability¹⁶. However, evidence is till absent to isolate whether the adsorption of anions by metal fluorides, or the coordination between polymers and anions. Moreover, during CSEs preparation, it is difficult for anions to diffuse in high-viscosity polymer solutions, and the presence of solid-liquid interface energy further hinders adsorption of extensive anions^17–19. If anions are incapable of forming stable aggregates on FeF₃ surface, the formation of strong Coulombic interactions to guide the migration of lithium ions between PVDF-HFP and LLZTO becomes a challenging endeavor.

Herein, CSEs with the anion-capturing layer is developed by compositing FeF₃@LLZTO particles and Li bis(trifluoromethanesulfonyl) imide (LiTFSI) within PVDF-HFP matrix. Prior to film-forming step of PVDF, FeF₃@LLZTO particles are dispersed within LiTFSI solution to facilitate Li salt dissociation and establishes a TFSI⁻-aggregated layer on surface. Then, it was blended with PVDF-HFP solution to retain solution viscosity and FeF₃ adsorption effect on hindering TFSI⁻ detachment. Meanwhile, FeF₃ is also highly compatible within PVDF-HFP matrix through fluorinated architecture⁶. By studying the local atomic and electronic structures of Fe, we confirm the interaction between TFSI⁻ and FeF₃. The Coulomb effect weakens electric field generated by positive charges on interface, reducing CGL thickness and opening Li⁺ transfer channels between polymer and ceramic phases. The resulting CSEs exhibits high conductivity and endows Li symmetrical or ssLMBs stable operation for hundreds of cycles.

Results

CSEs synthesis and characterizations

Figure 1a, b schematically illustrate TFSI⁻ adsorption manners onto FeF₃ fillers. LiTFSI-35wt%FeF₃@LLZTO/PVDF-HFP (L-FL₃₅PH) electrolyte is prepared via a one-pot mixing method, i.e., FeF₃@LLZTO, LiTFSI and PVDF-HFP were blended together in N, N-dimethylformamide (DMF). This method encounters difficulties in TFSI⁻ absorption onto FeF₃ particles, which is primarily due to high viscosity and the presence of solid-liquid interface energy^20,21. The existence of TFSI⁻ species not only impeded ion conductivity but also had the potential to trigger side reactions with electrode, thereby elevating interface resistance. While LiTFSI@35wt%FeF₃@LLZTO/PVDF-HFP (L@FL₃₅PH) CSEs is obtained through stepwise mixing, followed by LiTFSI absorption onto FeF₃@LLZTO surface and its integration with DMF in PVDF-HFP. Prior to this, a series of characterizations demonstrate the coating of FeF₃ onto the LLZTO surface, including high-resolution transmission electron microscope (HR-TEM), energy dispersive spectroscopy, X-ray diffraction (XRD) and X-ray photoelectron spectroscopy (XPS) (Supplementary Note 1 and Supplementary Figs. 1–3). After being assembled into CSEs, FeF₃ capping layer significantly improve compatibility of organic-inorganic interface and relieve the agglomeration phenomenon of LLZTO within PVDF-HFP architecture, as shown in scanning electron microscope (SEM) and Kelvin probe force microscope (KPFM) images (Supplementary Figs. 4–6). In addition, the cross-sectional morphology analysis indicates CSEs with an average thickness of 34 μm, and L@FL₃₅PH electrolyte reveals a discernible absence of particle aggregation phenomena (Supplementary Fig. 5). Particle agglomeration leads to an uneven fracture surface of L-L₃₅PH electrolyte.

Fig. 1. Regulating the internal ion interactions and TFSI⁻ adsorption concentration of CSEs.

a, b Schematic diagram of changes in anion adsorption concentration using one pot mixing method and stepwise mixing method, c FTIR spectra of different samples. d Binding energy between TFSI⁻ and Li⁺ or FeF₃, with the addition of the optimized geometric configurations and electronic potential of FeF₃, LiTFSI, TFSI⁻ and FeF₃-TFSI⁻, e, f C 1s and g, h N 1s spectra of L-FL₃₅PH and L@FL₃₅PH before and after etching.

The synthetic impacts on TFSI⁻ adsorption behaviors have been revealed through Fourier transform infrared spectroscopy spectra (FTIR) spectroscopy (Supplementary Fig. 7). To alleviate the internal integration of CSEs components, L-FL₃₅PH and L@FL₃₅PH electrolytes are fragmented as shown in the insert to Supplementary Fig. 7. Figure 1c exhibits a signal peak of LiTFSI at 771 cm⁻¹ corresponding to its S–N bond, which blue shifts to 766 cm⁻¹ after being adsorbed within LiTFSI@FeF₃@LLZTO particles²². Such S–N bond relaxation is attributed to LiTFSI dissociation with presence of FeF₃, which is also appropriate for L@FL₃₅PH electrolyte^23,24. While its position of L-FL₃₅PH electrolyte maintains at 770 cm⁻¹, indicating the weak interactions between TFSI⁻ and FeF₃. In addition, the S–N peak at 771 cm⁻¹ of LiTFSI/FeF₃/LLZTO remains at its position, representing the low dissociation degree of LiTFSI. In order to gather insights into such interaction, the adsorption energy has been calculated using density functional theory (DFT) as shown in Fig. 1d and Supplementary Data 1. It exhibits an adsorption energy of 10.24 kcal/mol between TFSI⁻ and Li⁺, which elevated towards 29.78 kcal/mol on FeF₃. The electrostatic interaction between Fe and N leads to FeF₃ with preference for capturing TFSI⁻ and promoting LiTFSI dissociation. The electrostatic potential (ESP) calculation provides explanation about TFSI⁻ fragments interaction with Li⁺ or FeF₃, where blue and red electrostatic potential equivalent surfaces represent electrophilic and nucleophilic, respectively (Fig. 1d)²⁵. The electrostatic potential equivalent surface of FeF₃ indicates that positive potential is concentrated near Fe, while the negative potential is near F atom. Before and after LiTFSI dissociation, the surface electrostatic potential changes. Negative potential of TFSI⁻ is mainly located at N and O, which are attracted with positively charged Fe atoms in FeF₃. In addition, we has evaluated against other materials with similar properties, such as MgF₂@LLZTO and AlF₃@LLZTO (Supplementary Fig. 8, Supplementary Data 1 and Note 2).

The full XPS spectra of two electrolytes are shown in Supplementary Figs. 9 and 10. L-FL₃₅PH surface that yields S 2p and N 1s signals, while they are absent in L@FL₃₅PH. Meanwhile, neither of them captures LLZTO features. The intensity of N 1s and S 2p peaks in L@FL₃₅PH electrolyte increases along with measuring depth, whereas L-FL₃₅PH peaks appear reversing trend. In the meantime, the presence of LLZTO is detect in both electrolytes after etching. The origin for this phenomenon is due to the that the L@FL₃₅PH exerted the adsorption effect of FeF₃ to TFSI⁻ through the distribution mixing method, resulting in enrichment of TFSI⁻ on the surface of FeF₃@LLZTO. While the L-FL₃₅PH electrolyte prepared by the one-pot mixing method causes TFSI⁻ to be uniformly dispersed in PVDF-HFP matrix for its weak interaction with FeF₃. Such result aligns with the anticipated outcomes in Fig. 1a, b. The variation in TFSI⁻ concentration gradient from surface to bulk between L-FL₃₅PH and L@FL₃₅PH could be further reflected in high-resolution C 1s spectra (Fig. 1e, f), covering C–F species at 291.8 eV attributed to LiTFSI and PVDF-HFP, C–C/C–H peak at 284.8 eV of PVDF-HFP, and C–S peak at 289.7 eV of TFSI^{− 26}. At the bulk region, O = C-N species appears at 286.4 eV, as evident for residual DMF solvent in L@FL₃₅PH²⁷. Comparing to L-FL₃₅PH, the intensity ratio of C–F/C–C in L@FL₃₅PH increases from surface to bulk, and the increase in C–S/C–C ratio is much larger than that of the L-FL₃₅PH electrolyte. The result validates that the TFSI⁻ is immobilized on FeF₃@LLZTO particles surface via the stepwise mixing method. The C-N peak at 403.7 eV in N 1 s spectra represents the residual DMF solvent, while the peak at 399.3 eV is attributed to TFSI^{− 28,29}. After being etched for 120 s, the Fe-N signal is detected at 400.2 eV for the both electrolyte membranes, resulting in the suppressed TFSI⁻ desorption from FeF₃ layer³⁰. The intensity ratio of TFSI⁻/C-N in L@FL₃₅PH electrolyte is about 2.78, higher than 0.87 of L-FL₃₅PH electrolyte (Supplementary Table 1). It is associated with higher residual DMF and TFSI⁻ concentrations on the FeF₃@LLZTO surface, which coincides with S 2p spectra (Supplementary Fig. 11) and thermogravimetric analysis testing (TGA) results (Supplementary Fig. 12). After Ar etching, the intensity of TFSI⁻ peak at 168.3 eV in S 2p spectra increases for L@FL₃₅PH electrolyte, while it alleviates for L-FL₃₅PH electrolyte. It indicates the TFSI⁻ accumulation on FeF₃@LLZTO surface in L@FL₃₅PH electrolyte. Meanwhile, the slight weight loss of the three CSEs before 140 °C is ascribed to evaporation of moisture and residual DMF solvent between 140 and 200 °C³¹. The differential scanning calorimetry (DSC) curve identifies that the melting temperature (T_m) of PVDF-HFP in L@FL₃₅PH, L-FL₃₅PH and L-L₃₅PH at 147 °C, without additional absorption peak (Supplementary Fig. 13). Therefore, the residual DMF solvent is quantified as 1.4 wt% of L@FL₃₅PH that is similar to 1.7 wt% of L-FL₃₅PH and 1.8 wt% of LiTFSI-35wt.%LLZTO/PVDF-HFP (L-L₃₅PH). FTIR spectroscopic results demonstrate that the presence of DMF-related free –N–C=O bonds is absent in L@FL₃₅PH and L-FL₃₅PH electrolytes, indicating that the residual DMF was all bound to Li⁺ and there were no free DMF molecules(Supplementary Fig. 14)³². The strong Li⁺-DMF interaction hinders complete DMF evaporation³². The signal peak associated with LiTFSI in the crystalline state is observed at 647 cm⁻¹ for L-FL₃₅PH electrolyte, suggesting some LiTFSI residues. In comparison, the weight loss of L@FL₃₅PH in the range of 200–300 °C is smaller, owing to its lower TFSI⁻ composition.

TFSI⁻ interactions with FeF₃@LLZTO surfaces

The impact of TFSI⁻ adsorption on CSEs surface potential is reflected through KPFM tests as depicted in Fig. 2 and Supplementary Fig. 15a, b. When comparing to L-L₃₅PH, uniform potential distribution on L@FL₃₅PH and L-FL₃₅PH electrolytes is revealed, illustrating FeF₃ encapsulation on LLZTO improves interface compatibility between organic and inorganic components. The charge-carrying feature of FeF₃@LLZTO particles surface could be recognized by the surface zeta potential that is measured at 441 mV (Supplementary Fig. 15c). As a reference, the surface zeta potential of LLZTO particles is measured to be 59 mV, indication an isolated FeF₃ coating as the charge aggregated surface. However, the surface zeta potential of L@FL₃₅PH electrolyte is measured at −340 ± 10 mV, which is much lower than those of L-FL₃₅PH and L-L₃₅PH. It is induced by its higher concentration of TFSI⁻, which decreases potential difference at the organic-inorganic interface, the reduced CGL thickness facilitates Li⁺ transport crossing phase boundaries³³. The L-FL₃₅PH exhibits minimal Coulomb interaction on the equilibrium electric field (Fig. 2e), resulting in thick CGL that impedes Li⁺ transport. It hinders Li⁺ migration from the PVDF-HFP matrix towards FeF₃@LLZTO particles within CSEs. While the stepwise mixing method endows L@FL₃₅PH with the concentrated TFSI⁻ on FeF₃@LLZTO surface (Fig. 2f). The organic-inorganic interface Coulomb interactions is established to decrease CGL thickness. Then, slow unidirectional pathway for Li⁺ transport is transformed into fast bidirectional pathway, and ion transport efficiency is improved. CGL is the charge distribution caused by the electric field intensity at the interface and has an electric bilayer effect^34,35. The result indicated that the double-layer capacitance of stainless steel (SS)∣celgard∣L@FL₃₅PH is 9.4 μF/cm², which is bigger than that of SS∣celgard∣L-FL₃₅PH (6.5 μF/cm², Supplementary Fig. 16). The result is ascribed to the enrichment of TFSI⁻ that enhances the surface charge density of FeF₃, promoting its closer arrangement, thereby compressing the charge layer spacing and simultaneously increasing the capacitance of the double-layer capacitance³⁵. In order to further validate CGL thickness, DFT models with various TFSI⁻ concentrations from 0.05 to 0.95 M have been constructed to simulate Li⁺ concentration, surface potential and charge density on FeF₃@LLZTO surface. The interfacial charge density increases along with TFSI⁻ concentration, resulting in decreasing surface potential and repelling effect for Li⁺ (Supplementary Fig. 17). Similarly, the introduction of anions leads to the redistribution of charge at the interface, resulting in a decrease in CGL thickness with increasing TFSI⁻ concentration (Supplementary Fig. 18). This theoretical simulation result is consistent with the KPFM and electric double layer capacity analysis.

Fig. 2. The effect of different TFSI⁻ concentrations on CGL.

a–d Interface potential distribution images of L-FL₃₅PH and L@FL₃₅PH electrolytes. e, f The schematic diagram of the effect of CGL thickness on lithium ion transport channels, where c_Li⁺ is the Li⁺ concentration and λ is the CGL thickness, g Fe K-edge XANES (min–max normalized), inset: a enlarged view, h fourier transforms Fe K-edge EXAFS spectra of Fe foil, Fe₂O₃, FeF₃ and FeF₃-TFSI⁻ and i wavelet transform of FeF₃ and FeF₃-TFSI⁻, respectively.

To verify TFSI⁻ absorption onto FeF₃ particles, Fe K-edge X-ray adsorption near-edge structure spectra has been collected. As shown in Fig. 2g, FeF₃ shows higher valent state than Fe³⁺ in Fe₂O₃, and Fe edge further redshift towards higher energy in LiTFSI@FeF₃. It reveals the electrons delocalization induced by TFSI⁻ adsorption^36,37. Its coordination environment is also investigated using k³-weighted Fourier-transformed extended X-ray absorption fine structure (EXAFS) spectra in Fig. 2h. FeF₃ shows Fe-F interatomic scattering at 1.51 Å, and it blueshifts to lower distance at 1.44 Å in FeF₃-TFSI^− 38. This trend is further reflected by contours distortion in their wavelet-transformed (WT) spectra as compared in Fig. 2i³⁷. These results prove that F in FeF₃ with strong electronegativity and electron-withdrawing capability provides Fe with electron-deficient feature as TFSI⁻ absorbers. TFSI⁻ adsorption further distorts coordination environment of Fe sites through forming asymmetric N-Fe/F-Fe scatterings (Fig. 2i and Supplementary Fig. 19). Raman spectroscopy is also utilized to characterize TFSI⁻ species in CSEs. Its concentration is calculated as 59% in FeF₃-treated L-FL₃₅PH, which is 11% higher than fresh L-L₃₅PH (Supplementary Fig. 20a, b)¹⁹. In contrast, L@FL₃₅PH exhibits TFSI⁻ concentration up to 71% as indicated in Supplementary Fig. 20c. These findings suggest that the stepwise mixing technique not only improves the adsorption of TFSI⁻ by FeF₃ and facilitates the dissociation of LiTFSI, but also decreases the obstruction of CGL to Li cross-phase migration.

Li⁺ transport mechanism

The influence of TFSI⁻ adsorption concentration on the electrochemical performance of CSEs is investigated using Nyquist plots. L@FL₃₅PH with 35 wt% FeF₃@LLTZO content is measured with ion conductivity of 1.1 × 10⁻⁴ S/cm at 25 °C, which is an order higher than 5.3 × 10⁻⁵ S/cm of L@FL₂₀PH (20 wt% FeF₃@LLTZO content), 6.3 × 10⁻⁵ S/cm of L@FL₅₀PH (50 wt% FeF₃@LLTZO content), 2.9 × 10⁻⁵ S/cm of L-FL₃₅PH and 1.5 × 10⁻⁵ S/cm of L-L₃₅PH (Supplementary Fig. 21). Further analysis on CSEs using electrochemical impedance spectroscopy at various temperature from 25 to 50 °C is conducted for calculating Li⁺ diffusion energy barrier (E_a, Supplementary Figs. 22–24). E_a value of L@FL₃₅PH is measured as 0.35 eV (Fig. 3a), which is lower than 0.73 eV of L-FL₃₅PH and 1.04 eV of L-L₃₅PH (Supplementary Fig. 25). Furthermore, solid-state electrolytes function as mix conductors of ions and electrons. With Li metal as negative electrode, even minor electrical conductivity results in electron leakage, leading to Li dendrite growth within CSEs and increasing the risk of battery short circuits^39,40. Therefore, lower electronic conductivity played a crucial role in the interface stability and cycling stability of CSEs. As shown in Fig. 3b, the steady-state current plots against the applied voltage at 25 °C for CSEs using the Hebb–Wagner polarization method. The average electronic conductivities calculate within the voltage range of 0.5–2.5 V for L@FL₃₅PH, L-FL₃₅PH and L-L₃₅PH electrolytes were 4.3 × 10⁻⁹, 1.7 × 10 ⁻⁹ and 0.9 × 10⁻⁹ S/cm, respectively (Supplementary Figs. 26–28). It is observed that having more TFSI⁻ immobilizes on the surface of FeF₃@LLTZO results in an increase in electronic conductivity, which could potentially lead to electron leakage. However, these values are lower than the values of ion conductivity and can be considered negligible¹⁸. Additionally, with an increase in input voltage, the steady-state current amplitude of L@FL₃₅PH shows increase, further indicating its higher ion conductivity (Supplementary Fig. 29). Since the transport of electrons within the electrolyte is driven by the concentration gradient of electron charge carriers, following Fick’s diffusion law, using Ohm’s law for measuring and calculating electronic conductivity may introduce errors; however, if the input voltage was below the electrolyte’s decomposition potential, the measured electronic conductivity values are still valuable⁴¹. The decomposition potentials of L@FL₃₅PH, L-FL₃₅PH and L-L₃₅PH (Supplementary Fig. 30) electrolytes obtains through linear sweep voltammetry (LSV) measurements are 5.0, 4.8 and 4.6 V, respectively, thus confirming the reliability of the average electronic conductivity calculated using Ohm’s law.

Fig. 3. Conductivity and ion migration numbers for CSEs obtained via various methods.

a Arrhenius plots of the ionic conductivities of the CSEs at 25–50 °C. b Variation of current with time during polarization of Li|CSEs|Li symmetric batteries configuration; the inset figure shows the Nyquist plots before and after polarization. c Measurement of critical current density for Li|CSEs|Li symmetric batteries. d Li plating/stripping curves of Li|CSEs|Li symmetric batteries with CSEs at 0.1 mA/cm² current density. e, f The ⁶Li high-resolution Li isotope SSNMR spectra of the L-FL₃₅PH and L@FL₃₅PH electrolytes before and after ⁶Li symmetric batteries Li plating/stripping for 60 h, and g corresponding peak proportions of ⁶Li⁺ with different transport paths.

In general, the permeation of Li⁺ in the electrolyte is usually anisotropic, leading to heterogeneous nucleation on the surface of lithium metal and potential dendritic growth. The initial time of dendrite growth is called the Sand’s time (τ_Sand), and extending the Sand’s time delays the generation of Li dendrites. The calculation formula is as follows Eq. 1 (Eq.1)⁴²:

$${\tau }_{\mathrm{Sand}}=\pi D\frac{e{C}_0{\left(t_{\mathrm{a}}+t_{{\mathrm{Li}}^{+}}\right)}^2}{{2jt}_{\mathrm{a}}}$$ 1

Where t_a and t_Li⁺ represent the migration rates of anions and Li⁺, respectively. D, e, C₀ and j represent the bipolar diffusion coefficient, the charge, the initial concentration of the electrolyte and the current density, respectively.

According to ion migration behavior, the sum of t_a and t_Li⁺ is a constant parameter equaling to 1. It is determined through Eq. 1, in which the limiting anion movement and increasing t_Li⁺ prolong τ_Sand and delay Li dendrites formation. To verify this rule, t_Li⁺ of L@FL₃₅PH, L-FL₃₅PH and L-L₃₅PH electrolytes are collected using the Li|CSEs|Li symmetric batteries, as shown in Fig. 3b. It is calculated as 0.39 for L-L₃₅PH electrolyte, reflecting its shortened τ_Sand and infeasibility to inhibit Li dendrites growth (Supplementary Fig. 31a). With the FeF₃-coating LLZTO, the migration number increases to 0.53 of L-FL₃₅PH electrolyte (Fig. 3b and Supplementary Table 2.), while it is still struggling in dendrites suppression at high current densities, such as 1.3 mA/cm² current density with increasing TFSI⁻ concentration fixed onto FeF₃@LLZTO particles, LiTFSI dissociation is promoted within L@FL₃₅PH electrolyte, followed by establishing the cross-phase ion transfer channels. Its Li⁺ migration number is measured as 0.75 with the extended τ_Sand. This result also coincide with trend in their Li⁺ conductivities as compared in Supplementary Fig. 31b. Additionally, the critical permeation threshold is also investigated as the key factor for the interface charge transfer. It is identified through recording the impedance changes with adding various quantity of liquid electrolytes as the interface infiltration agent. As illustrated in Supplementary Fig. 32, the impedance value undergoes a slight change around 305 Ω below 5 μL volume. When increasing to over 8 μL, it is reduced down to 177 Ω, as evident for the wetting-percolation threshold for liquid electrolytes onto CSEs. Given so, below the threshold of 8 μL, the electrolyte-electrode interactions are dominated by interfacial wetting effects, rather than bulk-phase ionic percolation.

The critical current density (CCD) before short circuit is another benchmark for solid-state electrolytes. The CCD values of L-FL₃₅PH and L@FL₃₅PH electrolytes are determined via galvanostatic charge/discharge measurements. Once the CCD is reached, the overpotential stopped increasing and suddenly drops sharply due to Li dendrite formation. The Li plating/stripping curves are obtained by increasing the current density by 0.3 mA/cm² current density every 2.0 h, as shown in Fig. 3c. L@FL₃₅PH electrolyte has a higher CCD value of 2.5 mA/cm² current density compared to L-FL₃₅PH electrolyte (CCD = 1.3 mA/cm² current density). Additionally, as the current density increases from 0.1 mA/cm² current density, the overpotential increase for the L@FL₃₅PH electrolyte is significantly smaller than for the L-FL₃₅PH electrolyte. This is primarily due to the less active TFSI⁻ being anchored on FeF₃ layer surface, making it harder to reach the electrode/electrolyte interface and causing concentration polarization, thus inhibitting the growth of Li dendrites and side reactions⁴³. Li|CSEs|Li symmetrical batteries have been assembled to study compatibility between solid-state electrolyte and Li foil. As illustrated in Fig. 3d, the Li|L-L₃₅PH|Li battery shows the highest overpotential of 114 mV and fails at 40 h. In comparison, Li|L@FL₃₅PH|Li and Li|L-FL₃₅PH|Li batteries both exhibit overpotentials of 28 mV. While Li|L-FL₃₅PH|Li battery gradually increases after 40 h, due to the increased interface resistance induced by side reactions of Li negative electrode with TFSI^− 44. While Li|L@FL₃₅PH|Li battery maintains its overpotential for above 1300 h Li plating/stripping. Even elevating current density to 1.0 mA/cm² with an area capacitance of 1.0 mAh/cm², the Li|L@FL₃₅PH|Li battery still operates for over 600 h (Supplementary Fig. 33). The post-cycling SEM images reveal obvious Li dendrites in Li|L-L₃₅PH|Li battery, while their growth is constrained in Li|L-FL₃₅PH|Li battery (Supplementary Fig. 34).

In order to reveal Li⁺ cross-phase transport mechanism, ⁶Li|CSEs|⁶Li symmetric batteries with L-FL₃₅PH and L@FL₃₅PH electrolytes were cycled for 60 h at 0.1 mA/cm² current density. Partial ⁷Li⁺ in CSEs is substituted by ⁶Li⁺ during Li plating/stripping, and then Li⁺ transport channels are identified using high-resolution ⁶Li isotope solid state nuclear magnetic resonance (SSNMR)¹⁴. The fitting result in Fig. 3e, f indicate three types of Li⁺ environments, i.e., Li⁺ transport within LLZTO (light red peak), interface of FeF₃@LLZTO and PVDF-HFP (blue peak), and PVDF-HFP chains (orange peak)⁴⁵. The fitting results of the ⁶Li SSNMR spectra before Li plating/stripping illustrate that the relative content difference of ⁶Li⁺ between the interface and PVDF-HFP in L@FL₃₅PH electrolyte is 1%, which is much smaller than that of L-FL₃₅PH electrolyte (9%), demonstrating the influence of the charge gradient layers on anchoring TFSI⁻ distribution (Fig. 3g). When Li plating/stripping with ⁶Li negative electrode, ⁶Li⁺ content within LLZTO for L@FL₃₅PH electrolyte increases from 73% to 87%, while its interfacial feature decreases from 13 to 3%, reflecting Li⁺ transport within LLZTO as the primary mechanism. In comparison, ⁶Li⁺ content within LLZTO changes from 67 to 26% for L-FL₃₅PH electrolyte. Therefore, due to the alleviated CGL effects, the concentrated TFSI⁻ in FeF₃ layer promotes Li⁺ migration through LLZTO particles. Such Li⁺ transport mechanism is further confirmed by the exchange current density of Li|L@FL₃₅PH|Li battery (0.41 mA/cm² current density) that is two-fold of the Li|L-FL₃₅PH|Li and Li|L-L₃₅PH|Li batteries (Supplementary Fig. 35).

CGL effects on interface stability of CSEs

The surface composition of CSEs after 40 h Li plating/stripping is analyzed by XPS. As shown in Fig. 4a, F 1s spectra appear a signal peak at 687.9 eV corresponding to C–F bond for PVDF-HFP and LiTFSI in both L@FL₃₅PH and L-FL₃₅PH, while the LiF peak at 684.8 eV is primarily associated with TFSI⁻ decomposition⁴⁶. It is evident that is hindered for decomposition on positive electrodes or negative electrodes in the lower LiF content of L@FL₃₅PH, which is attributed to the adsorbed TFSI⁻. In N 1s spectra (Fig. 4b), two peaks ascribed to TFSI⁻ and N-SO₄ are detected at 399.7 and 398.8 eV, as the result of TFSI⁻ decomposition²⁸. Similarly, two peaks at 167.2 eV and 165.4 eV in S 2p spectra are related to TFSI⁻ and N-SO₄ (Fig. 4c), respectively. While the weakening intensity of N-SO₄ peak also reveals TFSI⁻ decomposition^47,48. The TFSI⁻ signal of L@FL₃₅PH electrolyte is absent on surface, due to its adsorption on the FeF₃@LLZTO surface. With continuously decomposing TFSI⁻, its concentration at the Li negative electrode surface decreases. The peak at 162.1 eV ascribed to Li₂S is collected after decomposing TFSI^− 16. Such result reflects that a thicker SEI layer is established between L-FL₃₅PH electrolyte and Li negative electrode, resulting in interface polarization, lithium dendrite amplification, leading to secondary reactions and cycle life shortening^49–51. Thus, fixing TFSI⁻ is a feasible strategy to suppress Li dendrites.

Fig. 4. Composition analysis of CSEs and Li anodes after Li plating/stripping for 40 h in Li|CSEs|Li cells.

XPS profiles of a F 1s, b N 1s and c S 2p, d FTIR spectra e, f ToF-SIMS depth profiles (min–max normalized), with the 3D views of LiF₂⁻, LiS⁻ and NSO₄⁻ in L-FL₃₅PH and L@FL₃₅PH electrolytes, Li plating/stripping (0.1 mA/cm² current density) for 40 h at 25 °C.

After 40 h Li plating/stripping, blueshift of S–N bond from 770 to 776 cm⁻¹ is observed in FTIR spectra of L-FL₃₅PH, reflecting its enhanced bonding strength. It is ascribed to be aggregation of N-SO₄ after TFSI⁻ decomposition. In contrast, the S–N bond almost maintains its position in L@FL₃₅PH, followed by anions migration towards the Li negative electrode surface. The time-of-flight secondary ion mass spectrometry (ToF-SIMS) analysis reveals chemical evolution of the negative electrode during SEI formation (Fig. 4e, f and Supplementary Fig. 36). LiF₂⁻ and LiS⁻ signals in L@FL₃₅PH electrolyte accumulated on the intermediate stage after 25 s sputtering, whereas it is within 100 s time for L-FL₃₅PH. The results indicate that the SEI thickness formed on the Li negative electrode matches with L@FL₃₅PH electrolyte and is thinner than that with L-FL₃₅PH electrolyte. Furthermore, the intensity of the LiS⁻ signal also reflects the aggregated TFSI⁻ decomposition in L-FL₃₅PH electrolyte. The three-dimensional (3D) reconstruction in Fig. 4d reveals the depth distribution of LiF₂⁻, LiS⁻ and NSO₄⁻ fragments. The dense and uniform SEI components is demonstrated on the Li negative electrode for the two electrolytes, while L@FL₃₅PH enriches with LiF₂⁻ and LiS⁻. NSO₄⁻ fragments in comparison to L-FL₃₅PH. It further confirms the aggregation of TFSI⁻ species Li negative electrode. The ionic conductivity of LiF at 25 °C is measured as 6 × 10⁻⁸ S/cm calculated by the Nyquist plot (Supplementary Fig. 37), indicating that LiF species and thick SEI hinder ion transport.

The ssLMBs behaviors

The electrochemical performance of Li|CSEs|LiFePO₄ (LFP) solid-state batteries is collected. The charge-discharge curves of three ssLMBs at 0.5 C (85 mA/g), as illustrated in Fig. 5a. The discharge specific capacity of Li|L@FL₃₅PH|LFP battery reaches 159.4 mAh/g, closed to the theoretical specific capacity of LiFePO₄, and higher than the value 147.3 mAh/g of Li|L-FL₃₅PH|LFP battery and the 139.5 mAh/g of Li|L-L₃₅PH|LFP battery. As demonstrated in Fig. 5b, the Li|L-L₃₅PH|LFP battery exhibited a larger polarization potential, attributing to the presence of Li₂CO₃ on the LLZTO surface, which is in consistent with Li plating/stripping results. Figure 5c, d displayed the rate charge-discharge curves and rate performance of Li|CSEs|LFP batteries. At 0.5, 1.0, 1.5, 2.0 and 3.0 C (1.0 C = 170 mA/g) rates, the reversible capacity values of Li|L@FL₃₅PH|LFP battery were 159.4, 152.8, 140.3, 133.5 and 121.4 mAh/g, respectively, significantly higher than Li|L-FL₃₅PH|LFP and Li|L-L₃₅PH|LFP batteries. With increasing rates, the discharge specific capacity of Li|L@FL₃₅PH|LFP and Li|L-FL₃₅PH|LFP batteries exhibited similar attenuation amplitudes. But when the rate was decreased back to 0.5 C (85 mA/g), the discharge specific capacity of Li|L-FL₃₅PH|LFP battery could not recover its initial level. This result indicated that the Li|L@FL₃₅PH|LFP battery possesses better rate performance. After 600 cycles, compared to the initial capacity of 152.8 mAh/g, the Li|L@FL₃₅PH|LFP battery retains 96% capacity, much higher than 71% of the Li|L-FL₃₅PH|LFP battery. Furthermore, the average Coulombic efficient of Li|L@FL₃₅PH|LFP battery was around 99.98%. To further validate the feasibility of the practical application of L@FL₃₅PH electrolyte, we increased the loading of LFP to 3.3 mg/cm² (0.6 mAh/cm²), and the initial capacity of Li|L@FL₃₅PH|LFP battery is 152.0 mAh/g at 1.0 C (170 mA/g). After 600 cycles, compared to the initial capacity, retains 95% capacity (Supplementary Fig. 38). When the LFP loading is further increased up to 12.1 mg/cm² (2.0 mAh/cm²), the Li|L@FL₃₅PH|LFP battery still gives a specific capacity of 144.6 mAh/g at 1.0 C (170 mA/g), and the capacity retention rate maintains 98% after 200 cycles (Supplementary Figs. 39 and 40). At the same time, after subjected to high rate (3.0 C) charging and discharging, it exhibits still a good rate recovery performance (Supplementary Figs. 41 and 42). Due to uneven particle dispersion and poor interface compatibility, the Li|L-L₃₅PH|LFP battery quickly experiences a short circuit. In summary, the results suggested that fixing more TFSI⁻ on the FeF₃@LLZTO surface could not only enhance the efficiency of Li⁺ migration but also improve the discharge specific capacity and reversibility of solid-state batteries. On one hand, restricting TFSI⁻ mobility could prevent excessive decomposition and alleviate the increase in interface resistance. On the other hand, the Coulomb interaction between TFSI⁻ and positive charges reduced the thickness of CGL, providing a pathway for Li⁺ transfer across interfacial boundaries. As shown in Fig. 5f, the Li|L@FL₃₅PH|LFP pouch battery could light up multiple LED bulbs, even after folding and cutting, providing reliable evidence of its high safety performance.

Fig. 5. The Li|CSEs|LFP ssLMBs performance.

a Galvanostatic discharge/charge potential profiles of Li|CSEs|LFP batteries with b local amplification curve. c Galvanostatic discharge potential profile and d cycling performances of Li|CSEs|LFP batteries at different charging rates. e Capacity retention performance of Li|CSEs|LFP batteries at 1.0 C (170 mA/g). f The flexibility and security tests of pouch graphite|L@FL₃₅PH|LFP battery.

Li|CSEs|LiNi_0.8Co_0.1Mn_0.1O₂ (NCM811) batteries are also assembled to compare the feasibility of L@FL₃₅PH and L-FL₃₅PH electrolytes matched with high-potential positive electrodes. As shown in Fig. 6a, the discharge specific capacity of the Li|L@FL₃₅PH|NCM811 battery at 0.5 C (100 mA/g) is 187.3 mAh/g, higher than 174.6 mAh/g of the Li|L-FL₃₅PH|NCM811 battery. Following the rate increase, the decrease in discharge specific capacity of Li|L@FL₃₅PH|NCM811 battery is smaller than that of Li|L-FL₃₅PH|NCM811 battery. As shown in Fig. 6b, Li|L@FL₃₅PH|NCM811 batteries exhibit the smaller overpotentials in the charge-discharge curves at different rates, indicating that fixing TFSI⁻ is beneficial for accelerating Li⁺ migration kinetics⁵². Additionally, it shows specific capacity of 184.2 mAh/g at 1.0 C (200 mA/g) with a capacity retention above 82% after 200 cycles. It is 16% higher than Li|L-FL₃₅PH|NCM811 battery (Fig. 6c), and comparable to the values in literatures as listed in Supplementary Table 4. It could be clearly seen from the charge-discharge curves at various cycles, indicating the specific capacity attenuation amplitudes of Li|L@FL₃₅PH|NCM811 batteries are smaller than Li|L-FL₃₅PH|NCM811 batteries (Supplementary Figs. 43 and 44). Furthermore, the average CE of Li|L@FL₃₅PH|NCM811 battery is around 99.96%. It is evident that Li|L@FL₃₅PH|NCM811 battery exhibits superior performance when comparing to Li|L-FL₃₅PH|NCM811 battery, in terms of rate performance, overpotential and cycling stability.

Fig. 6. The Li|CSEs|NCM811 ssLMBs performance and influence of CGL on cathode structure evolution.

a Cycling performances and b galvanostatic discharge potential profiles of Li|CSEs|NCM811 batteries at different charging rates. c Capacity retention performance. d, e The Nyquist plots before and after cycling and f representative plots of imaginary impedance as a function of the log f (Debye plots) for Li|CSEs|NCM811 batteries at 1.0 C (200 mA/g). The vertical lines indicate the Debye peak positions. Where R_b, R_i, CPE₁ and W₀ is the battery resistance, interface resistance, constant phase element and warburg resistance, respectively. g, j In situ XRD patterns during the first charge-discharge cycle of Li|CSEs|NCM811 batteries. The charge-discharge curves were shown at the left and the enlarged views of the (003), (101) and (104) peaks were shown at the middle of each XRD pattern. h, k The (003) peaks position at the initial charge and full charge. i, l The (003) peak fitting result of the XRD pattern at the end of charge.

To analyze the underlying chemistries, the battery impedance before and after cycling tests is analyzed and fitted (Fig. 6d, e and Supplementary Table 3). L@FL₃₅PH electrolyte showcases the lower charge-transfer impedance, indicating reduction in the contribution of CGL to interface resistance. The plot of imaginary axis impedance against the logarithm of frequency (Nyquist plot) is fitted with the Lorentz function (Fig. 6f) to determine the relaxation time related to the overall ion migration. The peak frequency of the Debye peak is corresponding to the characteristic frequency of ion conduction, which could be determined by the reciprocal of the conductivity relaxation time (τ_σ) or the overall conductivity (σ_All) of the battery (Eq.2)¹⁶:

$${2\mathrm{\pi }f}_{\mathrm{Max}}=\mathrm{\omega }={\left({\tau }_{\mathrm{\sigma }}\right)}^{-1}={\sigma }_{\mathrm{All}}{\left(e_0{\epsilon }^{\prime }\right)}^{-1}$$ 2

Where ε‘ (real component of permittivity) and e₀ (8.854 × 10⁻¹⁴ F/cm) is the permittivity of frequency-independent and free space, respectively.

It could be observed that the peak position of the Li|L@FL₃₅PH|NCM811 battery shifts to higher frequency, resulting in a faster relaxation process. This result indicated that the reduction of CGL improved the ion migration efficiency of the Li|L@FL₃₅PH|NCM811 battery, consistent with the calculated results of ion conductivity. To further analyze the effect of the adsorbed concentration of TFSI⁻ on ion migration, in situ XRD is used to compare the structural evolution of NCM811 in Li|L@FL₃₅PH|NCM811 battery with that in Li|L-FL₃₅PH|NCM811 batteries. As shown in Fig. 6g, j, the diffraction peaks corresponding to crystal planes were labeled at the top of each XRD spectrum. Its peaks corresponding to (003), (101) and (104) planes exhibit similar evolution during the first cycle, with conventional polycrystalline NCM811 electrodes. The (003) peak gradually shifted to lower diffraction angles as Li⁺ deintercalates, indicating c-axis expansion; when the potential reaches 4.0 V, its position shifts to higher diffraction angles, indicating contraction of the c-axis. After that, (003) peak shifts by Δ = 0.2° in Li|L@FL₃₅PH|NCM811 battery (in Fig. 6h, k), lower than Li|L-FL₃₅PH|NCM811 battery of 0.4°, indicating that reducing CGL thickness improve battery reversibility. The comparison of the results of the peaks corresponding to the (003) crystal plane at the end of charging for Li|L-FL₃₅PH|NCM811 and Li|L@FL₃₅PH|NCM811 batteries is shown in Fig. 6i, l, respectively. Unlike the Li|L@FL₃₅PH|NCM811 battery, the Li|L-FL₃₅PH|NCM811 battery is fitted with three components at the end of charging from the peak corresponding to (003) crystal plane, namely the sluggish phase, intermediate phase and active phase. The appearance of the lag phase during the first cycle of charging indicated fatigue behavior of NCM811, leading to decrease in Li⁺ deintercalation and subsequent degradation in cycling stability⁵³. In stark contrast, Li|L@FL₃₅PH|NCM811 batteries don’t show any sluggish phase during the first charging cycle. The main factors causing hysteresis include obstructed Li⁺ transport channels, large kinetic barriers for Li⁺ release and unstable interface side reactions. The L@FL₃₅PH electrolyte not only had lower ion migration barrier and high interface stability, but also lower surface potential that enables Li⁺ to transport across phase boundaries. Undoubtedly, fixing anions on ceramic surfaces was an effective strategy for achieving high-potential composite solid-state batteries: on one hand, it could protect the positive electrode from further side reactions and degradation under high potential; on the other hand, it provided additional ion migration pathways for CSEs.

Discussion

In summary, we developed PVDF-HFP-based CSEs with TFSI⁻ fixation, high Li⁺ throughput and high ionic conductivity through a stepwise mixing method. LiTFSI coating onto FeF₃@LLZTO particles promotes TFSI⁻ adsorption as the anion-capturing layer. It not only triggers LiTFSI dissociation, but also decreases CGL thickness between FeF₃@LLZTO and PVDF-HFP, thus achieving spontaneous Li⁺ diffusion cross phase boundary. As a result, L@FL₃₅PH exhibits ion conductivity of 1.1 × 10⁻⁴ S/cm² and t_Li⁺ of 0.75 at 25 °C, respectively. Furthermore, TFSI⁻ layer also homogenizes the interfacial potential of L@FL₃₅PH and facilitating uniform Li⁺ transport behaviors with low Li plating/stripping polarization for over 1300 h. The Li|L@FL₃₅PH|LFP ssLMBs achieve a specific capacity of 152.8 mAh/g at 1.0 C with 96% retention after 600 cycles. Therefore, an interface engineering strategy is proposed and validated with promises to advance the development of solid electrolyte for high-performance and practical ssLMBs.

Methods

Materials

LiTFSI (99.99%), PVDF-HFP (the molecular weight is 254000), DMF (AR), ammonia solution (25%), Fe(NO₃)₃·9H₂O (99.9%) and NH₄F (AR) were purchased from DoDoChem. All the chemicals were analytical grade. Commercial LFP (Hefei Kejing Material Technology Co., Ltd.), commercial NCM811 (Hefei Kejing Material Technology Co., Ltd.) and commercial LLZTO (Shenzhen Kejing Zhida Technology Co.) were all used as received.

Surface functionalization of LLZTO

Disperse 500 mg of LLZTO in 20 mL of distilled water, and added 54 mg of Fe(NO₃)₃·9H₂O and 14 mg of NH₄F during magnetic stirring. The liquid color changes from white to reddish brown. Then adjusted the pH to 9 with ammonia solution and stirred at 90 °C until dry. Finally, transferring to a muffle furnace and calcining at 400 °C for 4 h to obtain a reddish brown powder, which was FeF₃@LLZTO powder.

Preparation of composite solid electrolyte

CSEs by the stepwise mixing method. Firstly, adding 90 mg (20 wt%), 190 mg (35 wt%), and 360 mg (50 wt%) FeF₃@LLZTO powder was respectively mixed with 60 mg LiTFSI in 2 mL DMF, and dried at 60 °C to obtain LiTFSI@FeF₃@LLZTO powders; then dispersed into a mixed solution of 300 mg PVDF-HFP and 6 mL DMF with magnetic stirring for 12 h. The LiTFSI@FeF₃@LLZTO/PVDF-HFP composite solid electrolyte was obtained by pouring onto a glass plate and dried in a vacuum oven at 60 °C for 12 h. Then the films were named L@FL₂₀PH, L@FL₃₅PH and L@FL₅₀PH electrolytes, respectively, according to the different FeF₃@LLZTO powder contents.

CSEs by the one-pot mixing method. One hundred ninety milligrams FeF₃@LLZTO powder, 60 mg LiTFSI, 300 mg PVDF-HFP and 6 mL DMF were mixed simultaneously to prepare LiTFSI/FeF₃@LLZTO/PVDF-HFP composite solid electrolyte membrane as a control group named L-FL₃₅PH electrolyte. Consistent with the above method, LiTFSI/LLZTO/PVDF-HFP composite solid electrolyte was prepared as a blank control group named L-L₃₅PH electrolyte.

Characterization

The crystal structure of the powder sample was obtained by using an XRD (Ultima-IV) in steps of 0.02° in the 2θ range 10°–80° at a scan rate of 10°/min. FTIR were measured using a Thermo Scientific Nicolet iS50 spectrometer in an attenuated total reflectance setup. The morphology of the powder sample characterized by using HR-TEM (JEOL JEM 2100). Raman spectra were obtained by using a Raman spectrometer (IDSPEC Arctic). TGA (Instruments Q500) was measured from 25 to 800 °C at a heating rate of 10 °C/min under an argon atmosphere. The surface zeta potential of powder was obtained by using a Zetasizer Ultra (Nano-ZS MPT-2). DSC measurements were performed on a METTLER DSC 3 instrument with a heating rate of 10 °C/min from 50 to 230 °C under an argon atmosphere. The interface potential and surface roughness were obtained by KPFM (MultiMode8). The Fe K-edge extended EXAFS spectra were collected under fluorescence yield mode at the Very Sensitive Elemental and Structural Probe Employing Radiation from a Synchrotron (VESPERS, operated at 2.5 GeV and 250 mA) beamline of Canadian Light Source, combined with the RapidXAFS HE Ultra to gain insights into its chemical state at the atomic level. To prevent air exposure, samples were mounted on a copper sample holder with conductive carbon tape within an argon-filled glovebox. They were then transferred via an argon environment transport container to the X-ray absorption antechamber, maintaining an inert atmosphere throughout the process. XAS data were processed using the Athena and Artemis software with FEFF code for energy calibration and normalization, while the WT analysis was simulated according to the Morlet wavelet^54,55. The ⁶Li SSNMR spectra of CSEs before and after Li plating/stripping were collected using a AVANCE Ⅲ HD 400 MHz spectrometer (19.6 T, ω0/2π  =  122.14 MHz for ⁶Li), with a 3.2 mm probe spinning at 10 kHz for ⁶Li(single-pulse sequence, and the recycle delay of 10 s for ⁶Li). Li⁺ transport contributions can be quantified according to the ⁶Li peak ratio change via integrating the peaks for various components with and without ⁶Li plating/stripping. The morphologies of the CSEs and the Li metal negative electrodes were observed by SEM (HITACHI S-4800) at a beam voltage of 15.0 kV. XPS (Thermo Scientific ESCALAB Xi+) were used to investigate the compositions of the CSEs. Cycled batteries were disassembled in an argon-filled glovebox to collect the CSEs and Li metal negative electrodes. The samples were rinsed with 1,3-dioxolane (DOL) three times to remove residual electrolytes and then dried under vacuum in a transition cabinet overnight⁵⁶. For Li metal negative electrodes, a vacuum transition cabinet was employed to transfer them from the glovebox into the chambers of SEM, XPS, or ToF-SIMS to prevent degradation due to air exposure. The ToF-SIMS analysis was performed using the IONTOF M6 instrument with a primary ion energy of 30 keV, an ion current of 2 nA, and the use of Bi3++ ions for analysis. It was used to investigate the compositions of the SEI layer on the Li negative electrodes after 40 h in the Li|CSEs|Li symmetric batteries. The structural evolution of NCM811 in Li|CSEs|NCM811 batteries during charge/discharge were observed by in situ X-ray diffractometer (in situ XRD, Bruker D8 discover).

Adsorption energy calculation

DFT calculations were performed using the DMol3 module in MaterialStudio 2020. The electron-ion (nucleus) and Van der Waals interactions were described using the projector augmented wave and the DFT-D3 dispersion correction methods, respectively. For the exchange-correlation energy calculations, we employed the generalized gradient approximation framework with the Perdew–Burke–Ernzerhof functional. The valence electron wavefunctions were expand via a plane-wave basis set with the kinetic energy cutoff of fine. The structural optimizations were conducted using a 2 × 2 × 2 Monkhorst–Pack k-point mesh for sampling the Brillouin zone, while a denser 4 × 4 × 4 mesh was adopted for electronic structure computations. The total energy of convergence criteria and the ionic forces during geometry optimization were set to 10⁻⁵eV and 0.02 eV Å⁻¹, respectively. Along the z-direction to eliminate interactions from periodic boundary conditions. The free energies of the isolated FeF₃ (denoted as E₁), TFSI⁻ anion (E₂), and the FeF₃–TFSI⁻ adsorption complex (E₃) were obtained after geometry optimization. The adsorption energy (E_ads) was calculated according to the equation (Eq.3):

$$E_{\mathrm{abs}}=E_3-\left(E_1+E_2\right)$$ 3

The adsorption energies between CSEs components Li⁺ and TFSI⁻ were obtained by the same calculation method.

SCL thickness calculation

We performed steady-state numerical simulations of the Poisson–Nernst–Planck equations to model the transport of solvated ionic species. The simulations were implemented in COMSOL Multiphysics, with a computational domain spanning from the FeF₃ surface to the PVDF-HFP film and an electrolyte region length of 500 nm. The governing equations describe ion transport through coupled diffusion and electromigration processes (Eqs. 4–6).

$$J_{\mathrm{i}}=-D\mathrm{i}\left(\nabla c_{\mathrm{i}}+\frac{c_{\mathrm{i}}{n}_{\mathrm{i}}F}{RT}\nabla \mathrm{\phi }\right)$$ 4

$$\frac{\partial c_{\mathrm{i}}}{\partial t}=-\nabla J_{\mathrm{i}}=0$$ 5

$$E=-\nabla \mathrm{\phi }$$ 6

Where c_i, D_i, and n_i denote the concentration, the diffusion coefficient and the charge of species i (with i = Li⁺, cation with positive charges and anion with negative charges), respectively. R and T correspond to the ideal gas constant and the temperature, respectively. φ denotes for the potential and E denotes for electric field. A concentration of 0.05 M is set for Li⁺ with 0.055 M for the anion and 0.005 for the cation.

We assume a constant surface charge density of 0.01 C/m² on the PVDF-HFP film, considering its nonconductive properties as an organic material. Afterwards, we modulated the surface potential of FeF₃ by applying a potential sweep from −0.4 to −0.1 V, and evaluated the compressive behavior of the charge gradient layer.

Electrochemical measurements

The ionic conductivity (σ) of the composite membranes was determined by alternating current (AC) impedance using a Princeton 2263 electrochemical workstation. Measurements were performed over a frequency range of 1–10⁶ Hz and 10 mV voltage, while the membranes were sandwiched between two stainless-steel plates (1 mm thickness). The σ was calculated using Eq.7.

$$\sigma =\frac{L}{RS}$$ 7

where L and S corresponding to the thickness (thickness: 94 μm) and the area of CSEs, respectively. R is the impedance.

In contrast to the σ calculation, the average electronic conductivity contributions from electrons and holes in the CSEs were determined by applying Ohm’s law to steady-state current measurements across polarization voltages (Eq.8).

$$\overline{{\sigma }_{\mathrm{e}+\mathrm{h}}}=\frac{L}{S}\times \frac{I}{E}$$ 8

Here, E and I are the polarization voltage and the steady-state current of the battery, respectively.

Li⁺ transference number (t_Li⁺) was determined in a Li||Li symmetric battery (Li foil thickness: 0.5 mm, diameter: 16 mm) by combining direct-current polarization and AC impedance measurements. Resistance and current values were acquired using a Princeton 2263 electrochemical workstation at 25 °C. The t_Li⁺ value was calculated based on Eq. 9.

$$t_{{\mathrm{Li}}^{+}}=\frac{I_{\mathrm{s}}\left(\mathrm{\Delta }V-I_0{R}_0\right)}{I_0\left(\mathrm{\Delta }V-I_{\mathrm{s}}{R}_{\mathrm{s}}\right)}$$ 9

Where ΔV corresponding to the constant polarization voltage (10 mV), I₀ and I_s represent the current before and after polarization, respectively. R₀ and R_s represent the resistance before and after polarization, respectively.

The electrochemical stability windows of the CSEs were evaluated using LSV on a Princeton 2263 electrochemical workstation. Measurements were performed at 25 °C with a scan rate of 10 mV/s over a potential range of 0 to 7 V (vs. Li/Li⁺). Prior to testing, each membrane was sandwiched between a SS electrode and a Li foil counter electrode (Li foil thickness: 0.5 mm, diameter: 16 mm). Additionally, the current/time relationship curves (0.5–2.5 V) and Tafel curves of CSEs were acquired using the same workstation.

Cyclic voltammetry measurements were conducted over a potential range of 0.5 to 0.7 V at various scan rates (5, 10, 20, 30, 40, 50, and 60 mV/s) using the Princeton 2263 electrochemical workstation. The average current density at 0.6 V was selected and plot the scanning rate and current density. The slope of the linear fit to these data corresponds to the double-layer capacitance.

For battery assembly, working electrodes were prepared from L-FL₃₅PH and L@FL₃₅PH (CSEs thickness: 34 μm). A Celgard 2400 membrane was used as the separator, and a SS served as the counter electrode. To ensure adequate electrode wetting, 2 μL electrolyte solution (1 M LiTFSI in fluoroethylene carbonate) was applied to the interface between the CSEs and the separator.

Full battery assembly and performance measurements

NMC811 and LFP positive electrodes were prepared using a slurry casting technique. In detail, the cathode slurry was prepared by homogeneously mixing NMC811 or LFP powder, super P carbon, and polyvinylidene fluoride at a weight ratio of 8:1:1 in N‑methyl‑2‑pyrrolidone. This mixture was uniformly coated onto an Al foil current collector and subsequently dried in a vacuum oven at 110 °C for 12 h. The resulting LFP cathode exhibited an active material loading of 2.0 mg/cm². The active loading densities of the NMC811 positive electrode were 2.2 mg/cm². The positive electrode was cut into circular discs with a diameter of 12 mm, and placed at the center of the positive electrode shell. Then sequentially place the composite solid electrolyte membrane, Li foil with thickness of 0.5 mm and diameter of 16 mm, gasket, spring plate, and negative electrode shell. All coin cells were assembled in an argon-filled glove box. During the assembly process, 2 μL (1 M LiTFSI in FEC) was added to the interface between the solid electrolyte and electrodes to ensure sufficient electrode wetting. All the cells were tested by the LAND measurement system within a potential range of 2.8–4.1 V (LFP) or 2.8–4.3 V (NCM811) at 25 °C. All the cells were tested by a LAND electrochemical workstation with a potential range of 2.8–4.1 V (LFP) or 2.8–4.3 V (NCM811) at 25 °C.

Pouch cell assembly

The pouch cell was assembled by sequentially stacking an LFP positive electrode (36.3 mg/cm² load), CSEs, graphite negative electrode (18.0 mg/cm² load). The stack was encapsulated in an aluminum plastic pouch under vacuum conditions to prevent electrolyte contamination. To ensure interfacial contact between the electrode and CSEs, the pouch cell was positioned between two polished iron plates and applied 50 MPa of uniform pressure through screw fastening. Subsequently, the pouch cell charging was conducted using a LAND battery testing system at 25 °C, with a voltage window of 2.8–3.65 V (vs. Li/Li⁺).

Author contributions

J.L., Y.Z., and H.P. conceived and designed the research. J.L. and Y.Z. performed materials synthesis and characterization, and data analysis. Y.Z., N.H. and H.P. conducted battery preparation and electrochemical measurements. S.-Z.L. and Z.-D.M. conducted DFT simulation calculations. L.H., S.-G.S., and Y.-P.D. supervised the project. J.L. wrote the manuscript. L.H., S.-G.S., and Y.-P.D. revised the manuscript. All authors contributed to the interpretation, conclusions and preparation of the manuscript.

Peer review

Peer review information

Nature Communications thanks Lingyun Zhu and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The detailed data supporting findings from this work are available within this paper and the Supplementary Information. All other relevant data supporting findings are available from the corresponding author on 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 10.1038/s41467-025-67065-0.