Safe electrolyte design with isolated solvation nanoclusters for high-energy lithium metal batteries

Abstract

The pursuit of high-energy-density lithium metal batteries is often hampered by severe safety hazards and limited cycle life, leading to increased environmental impact and higher costs. Here, we demonstrate that tuning the underappreciated miscibility between electrolyte components could provide an effective approach. By introducing a non-coordinating, non-flammable cyclic fluorinated diluent, we minimize diluent-ion pair intermolecular forces, thereby enabling the formation of isolated solvation nanoclusters. The designed microstructure could facilitate enhanced Li⁺ transport to suppress dendrite growth and foster an inorganic-rich interphase to enhance interfacial electrochemical and thermal stability. This electrolyte design enables 50 μm Li | |2.5 mAh cm⁻² LiNi_0.8Mn_0.1Co_0.1O₂ cells to retain 80% capacity after 800 cycles and over 95% capacity following an 8-month calendar aging test. Moreover, the Ah-level pouch cells could achieve a specific energy exceeding 500 Wh kg⁻¹ based on the total mass of the pouch cell, while withstanding fully charged nail penetration and 200% overcharge without fire or swelling. Therefore, the nanocluster electrolyte reduces lifetime safety costs and material consumption per unit energy delivered, collectively enhancing practical sustainability for high-energy-density, safe, and long-lasting lithium metal batteries.

Lithium metal batteries face safety and lifespan challenges. Here, authors develop a nanocluster electrolyte that suppresses dendrites, forms stable interphases, and enables >800 cycles in high loading cells and >500 Wh kg-1 in Ah-level pouch cells, with nail penetration and overcharge tolerance.

Introduction

Lithium metal batteries (LMBs), which employ a lithium metal negative electrode in conjunction with a Ni-rich layered positive electrode (LiNi_xMn_yCo_1-x-yO₂, NCM), represent the promising technology for achieving specific energy exceeding 500 Wh kg^−1 1–3. However, their practical implementation is challenged by the severe sustainability issues arising from the high reactivity of both electrodes^4,5. These interfacial parasitic reactions result in dead lithium accumulation, rapid electrolyte consumption, and progressive positive electrode degradation, collectively shortening the battery lifespan, increasing material consumption and replacement, particularly under lean-electrolyte and limited-lithium conditions^6–8. Moreover, the inherent flammability of conventional electrolytes raises safety concerns, jeopardizing the non-negotiable safety and long-term stability of LMBs^9,10. Given these obstacles, electrolyte engineering has emerged as an effective and scalable strategy to address these intertwined challenges by tailoring interfacial chemistry^11–13. Substantial efforts have been devoted to developing advanced electrolyte systems, including high-concentration electrolytes (HCEs)^14,15, localized high-concentration electrolytes (LHCEs)^16–18, liquid gas electrolytes^9,19,20, modifications of salt/solvent functional groups^11,21–25, and optimization of additives^6,26. Among these, LHCEs, derived from HCEs, utilize non-coordinating and non-flammable diluents to mitigate fire risks, thereby enabling reduced electrolyte viscosity, suppressed flammability, and preservation of anion-dominated solvation structures, e.g., contact ion pairs (CIPs) and aggregates (AGGs)²⁷. Moreover, this configuration promotes the formation of inorganic-rich interphases with favorable ionic conductivity and electronic insulation, which prevent further electrolyte decomposition and improve interfacial stability. However, current electrolyte designs struggle to address the performance-cost-safety trilemma, failing to meet the stringent sustainability requirements of next-generation batteries.

In conventional LHCE design, diluents with low donor numbers and dielectric constants are expected to be excluded from Li⁺ solvation sheaths, thereby preserving anion-dominated coordination and minimizing the presence of free solvent molecules²⁸. The solvent is envisioned as a surfactant that mediates between ion pairs and the diluent, reducing interfacial energy and stabilizing localized high-concentration clusters²⁹. However, commonly used diluents, e.g., 1,1,2,2-tetrafluoroethyl-2,2,3,3-tetrafluoropropylether (TTE) and bis(2,2,2-trifluoroethyl) ether (BTFE), exhibit non-negligible coordination with Li⁺ via intermolecular interactions, distorting the intended solvation structure³⁰. More importantly, these diluents interact relatively strongly with both anions and solvents. The diluent-anion interaction disrupts Li⁺-anion coordination and weakens the Li⁺ inductive effect, thereby diminishing the inorganic content in both interfacial layers³¹. Furthermore, although diluent-solvent interactions can lower solvent activity and suppress solvent decomposition, overly strong interactions may stabilize extended ionic aggregates rather than effectively fragment them, thereby limiting Li⁺ transport³². Under electrochemical polarization, these aggregates may undergo dynamic reorganization, increasing the availability of solvent at the electrode interfaces and potentially promoting parasitic reactions and long-term cell aging^33,34. Therefore, it is essential to regulate diluent-aggregate interactions and promote nanoscale segregation without inducing macroscopic phase separation³⁵. Based on Hansen solubility parameters (HSP), mutual solubility is governed by non-polar dispersion forces, permanent dipole-dipole interactions, and hydrogen bonding³⁶. Rational design of non-flammable diluents through modulation of molecular structure, spatial configuration, and functional groups offers an approach to reduce diluent-aggregate affinity while maintaining macroscopic miscibility, thereby minimizing diluent interference with ionic aggregates and promoting the formation of isolated and stable nanoclusters in LHCEs. Therefore, these nanoclusters enable the formation of functional interfacial layers, extending cycle life and reducing material consumption per cycle, while their non-flammable nature could eliminate fire hazards and reduce safety costs, synergistically benefiting the sustainability of energy storage systems³⁷.

Herein, we propose a diluent strategy to mitigate diluent interactions with Li⁺ cations, solvents, anions, and their ionic aggregates. The nonflammable cyclic fluorinated alkane 1,1,2,2,3,3,4-heptafluorocyclopentane (HFC) exhibits a suitable balance of the Hansen dispersion (δD), polar (δP) and hydrogen-bonding (δH) components, allowing it to remain macroscopically miscible with the HCE while potentially favoring a more isolated ionic-cluster environment (Fig. 1 and Supplementary Table 1–2). These nanoclusters facilitate rapid Li⁺ diffusion to suppress dendrite growth, while fostering a stable, compact, and inorganic-rich interphase that enhances high-voltage compatibility and thermal stability. Consequently, this electrolyte mitigates self-discharge and prolongs cycling life, retaining over 95% capacity after 8 months of calendar aging and achieving more than 800 cycles with 80% capacity retention in 50 μm Li | |2.5 mAh cm⁻² NCM811 coin cells. The 6 Ah-level pouch cells could deliver a specific energy over 500 Wh kg⁻¹ (based on the total mass of the pouch cell), while the 2 Ah-level pouch cells maintain 85% capacity after 180 cycles. Furthermore, the electrolyte exhibits enhanced stability under harsh conditions, including elevated temperature (60 °C), high operating voltage (4.5 V), and compatibility with a high-Ni positive electrode (LiNi_0.9Mn_0.05Co_0.05O₂, NCM90). Safety is further demonstrated through nail-penetration and overcharge tests, which show no ignition and tolerance to 200% overcharge without swelling or explosion. This work overcomes critical barriers toward practical high-energy LMBs and provides new insights for developing efficient, safe, and sustainable energy storage systems.

Fig. 1. Weak diluent-ion pairs interactions promote isolated nanocluster formation.

By mitigating the diluents-ionic aggregates interaction and enabling the formation of isolated nanoclusters (n-AGG), cyclic fluorinated alkane diluents offer several benefits: (i) facilitate rapid Li⁺ diffusion; (ii) form a stable, inorganic-rich interphase; (iii) stabilize the electrode-electrolyte interphase to suppress degradation, self-discharge, and safety issues.

Results

Design principle for isolated nanoclusters

To decipher the effects of diluents on solvation environments, we systematically investigated three electrolyte systems: high-concentration electrolytes (HCE), localized high-concentration electrolytes with TTE diluent (LHCE), and sterically regulated high-concentration electrolytes with HFC diluent (SRCE) (See Methods for detailed information). The calculated electrostatic potential and binding energy suggest a strong interaction between the DME solvent and Li⁺ (Fig. 2a, b, and Supplementary Fig. 1). In contrast, the cyclized HFC molecule, which lacks a heteroatom and possesses low polarizability, imposes conformational and electrostatic constraints that diminish its ability to coordinate with Li⁺, yielding a weaker binding energy (−105.6 kJ mol⁻¹) compared to TTE (−141.4 kJ mol⁻¹). Nuclear magnetic resonance (NMR) spectroscopy was performed to further elucidate diluent-induced perturbations in the solvation environment. The negligible chemical shift in the ¹⁹F NMR spectra of both diluents confirms minimal coordination between Li⁺ and diluents, consistent with the previous reports³⁰ (Supplementary Fig. 2). Nevertheless, adding TTE diluent into the HCE induces a pronounced upfield shift in the ¹⁹F NMR of FSI^- (Δδ = −0.75 ppm), consistent with the density functional theory (DFT) calculated hydrogen bond energy, indicating the enhanced electron density around the fluorine atoms due to strong anion-diluent interaction (Fig. 2c, Supplementary Fig. 3, Supplementary Note 1 and Supplementary Data 1). In comparison, the steric hindrance and low polarity of HFC molecules weaken the intermolecular forces interaction with FSI^-, yielding a smaller upfield shift of −0.56 ppm with lower bonding energy. Moreover, while DME solvent as a surfactant engages with both diluents, the HFC exhibits a weaker interaction due to its relatively low dielectric constant (Fig. 2d, Supplementary Fig. 4a, and Supplementary Table 3). Consistent with previous literature, MD simulations suggest that the diluents interact with DME to competitively displace solvent molecules from the Li⁺ solvation sheath and promote anion ingress, whereas the influence of HFC is smaller than that of TTE (Supplementary Fig. 5, Supplementary Table 4, and Supplementary Data 2). Consequently, the HFC interacts weakly with cations, anions, and solvents, largely preserving the anion-rich solvation structure and Li⁺-FSI^- interaction, aligning with the experimental results. Notably, compared to HCE, the ⁷Li NMR spectra of SRCE show negligible variation in the lithium chemical shift (Δδ = −0.007 ppm), indicating that HFC has a minimal impact on the Li⁺ electronic environment (Fig. 2e). This phenomenon may stem from attenuated competing diluent-anion interactions, in which the modestly reduced electron-donating ability and the slightly increased anion coordination counterbalance each other. Conversely, introducing TTE diluents into HCE triggers a distinct downfield shift in ⁷Li NMR spectra for LHCE. This implies that the anion-diluent interaction significantly withdraws electron density from the anions, reducing electron shielding around Li⁺, thereby affecting cation-anion interactions and hampering the Li⁺ inductive effect for further inorganic-rich SEI formation.

Fig. 2. Solvation structures characterization.

a Molecular electrostatic potential (ESP) maps of DME, TTE, and HFC. b Binding energies of Li⁺ with DME, TTE, and HFC molecules. c ¹⁹F NMR spectra of FSI^- in three electrolytes (external reference: 1 M Zn(OTF)₂ in D₂O at −78.84 ppm). d ¹⁷O NMR spectra of DME solvent, DME-TTE mixture, and DME-HFC mixture (external reference: D₂O at 0 ppm). e ⁷Li NMR spectra of three electrolytes (external reference: 1 M LiCl in D₂O at 0 ppm). f Synchrotron wide-angle X-ray scattering profiles of three electrolytes. g Li⁺ transport properties in three electrolytes: Li⁺ diffusion coefficient (D), Li⁺ transference number (t₊), and ionic conductivity (δ).

We further conducted synchrotron wide-angle X-ray scattering (WAXS) to investigate the effect of diluents on ionic aggregates and the mesoscale solvation environment (Fig. 2f and Supplementary Fig. 6). The WAXS results reveal that two characteristic typical solvation clusters of HCE appear at the scattering vector (Q) values around 0.94 and 1.54 Å⁻¹. Based on the Ehrenfest equation (Q·r = 2.46π)^8,38, these features correspond to characteristic real-space correlation distances of approximately 8.22 Å and 5.02 Å, where the high-Q region (1.54 Å⁻¹) reflects direct cation-anion contacts and low-Q regions (0.94 Å⁻¹) stands for medium-range ordered aggregation (formed between CIPs)²⁶. Consistent with the NMR and MD simulation results, the addition of TTE affects Li⁺-FSI^- interactions and the CIPs aggregation peak shifts to a higher Q area with reduced intensity. The slightly stronger interactions of TTE with FSI^- and DME lead to partial but incomplete dissociation of medium-range ionic aggregation, thereby reducing both the size and population of nanoclusters. Notably, the incorporation of HFC induces a pronounced structural reorganization, characterized by the convergence of the two distinct peaks into a single broad feature centered at 1.21 Å⁻¹. In addition to the low $\delta D$, the cyclic structure and absence of coordination sites (oxygen atom) endow the HFC molecule with a lower $\delta P$ and $\delta H$, resulting in a greater difference in HSP distance (R_a) compared to CIPs (Supplementary Fig. 7, Note 2, and Supplementary Table 1–2). Therefore, an optimal R_a could dissociate the interconnections between ion pairs and HFC diluents without inducing phase separation, thereby disassembling the medium‑range CIP aggregates network into smaller, isolated nanoclusters. This separated configuration subsequently enhances cation-anion interactions within the primary solvation sheath. The inductive effect of Li⁺ further promotes FSI^- decomposition, facilitating the formation of an inorganic-rich interphase (Supplementary Fig. 8). Moreover, the isolated nanoclusters are randomly distributed within the diluent network, creating a heterogeneous microstructure with high configurational entropy. Consequently, the WAXS profile exhibits a single, broad peak at 1.2 Å⁻¹, corresponding to the statistical average of scattering from this reconfigured environment. According to the Stokes-Einstein relation^39,40, the smaller nanocluster size and intermolecular forces between ion pairs in SRCE facilitate Li⁺ diffusivity and transference number, thereby enhancing the overall ionic conductivity (Fig. 2g, Supplementary Fig. 9, and Supplementary Table 5).

Interfacial stability of the Li metal negative electrodes

To systematically evaluate the compatibility of three electrolytes with lithium metal negative electrodes, we assessed their electrochemical performance in Li | |Cu asymmetric and Li | |Li symmetric cells. The modified Aurbach method reveals that SRCE possesses the highest Li plating/stripping Coulombic efficiency (CE) of 99.61% at 25 °C (average of 5 trials with standard deviation of ±0.04%) among three systems, thereby favoring the long-term Li plating/stripping stability of LMBs (Fig. 3a). We further probed the rate capability using Li | |Li symmetric cells by increasing the current density from 0.5 to 10 mA cm⁻² with a fixed areal capacity of 2 mAh cm⁻² (Fig. 3b). The HCE cells exhibit significant polarization and fail at ~5 mA cm⁻², while the LHCE cells sustain operation until ~8 mA cm⁻². In contrast, the SRCE cells exhibit smaller polarization and maintain stable Li plating/stripping even at 10 mA cm⁻², owing to fast Li⁺ diffusion kinetics, thereby enhancing resistance to dendrite formation. Scanning electron microscopy (SEM) of Li deposition morphologies (0.5 mA cm⁻², 5 mAh cm⁻²) on Cu foils further corroborates these findings (Fig. 3c–e). The HCE cells produce a porous and dendritic morphology, while the LHCE cells form larger but loose and inhomogeneous Li particles. Conversely, the SRCE cells yield a dense, uniform, and smooth Li deposition morphology, validating the dendrite-suppression property in SRCE. Additionally, the SRCE cells display a more uniform Li deposition morphology after 100 plating/stripping (0.5 mA cm⁻² and 0.5 mAh cm⁻²) and a prolonged Li plating/stripping performance (1 mA cm⁻² and 1 mAh cm⁻²) for over 1600 h without short circuits or overpotential increase (Supplementary Fig. 10).

Fig. 3. Interfacial stability of lithium metal negative electrodes.

a Comparison of Li plating/stripping CE in three electrolytes at 25 °C, measured using the modified Aurbach method⁴⁹. b Rate performance of Li | |Li symmetric cells at 25 °C. SEM images of Li deposition morphologies on Cu foil with c HCE, d LHCE, and e SRCE at 5 mAh cm⁻², 0.5 mA cm^−2, and 25 °C. f F 1 s XPS spectra of SEI layers formed in the three electrolytes after 5 mAh cm⁻² of Li plating at 0.5 mA cm⁻² and 25 °C. g 3D TOF-SIMS reconstruction of the SEI formed in SRCE after 5 mAh cm⁻² of Li plating at 0.5 mA cm⁻² and 25 °C.

Beyond its enhanced Li⁺ diffusion kinetics, the stronger Li⁺-FSI^- interaction in SRCE further promotes the formation of an anion-derived, inorganic-rich solid electrolyte interphase (SEI). X-ray photoelectron spectroscopy (XPS) establishes distinct compositional differences among SEI layers formed in different electrolytes. The high-resolution F 1 s spectra demonstrate a substantially higher proportion of Li-F signal in SRCE compared to the more prominent S-F signal observed in both LHCE and HCE, arising from incomplete decomposition of FSI^- anions (Fig. 3f). Corresponding C 1 s spectra show more intense carbon-related signals in HCE and LHCE, indicating a higher content of organic species in their SEI layers (Supplementary Fig. 11). Unfortunately, these organic composites are likely to originate from solvent decomposition, which subsequently consumes electrolyte, impedes Li⁺ transportation, and accelerates lithium dendrite growth. Moreover, XPS depth profiling reveals an inorganic-rich SEI formed in SRCE, featuring a dominant presence of LiF that intimately coexists with highly enriched Li₂O (Supplementary Fig. 12). Time-of-flight secondary ion mass spectrometry (TOF-SIMS) in negative-ion mode (Ga⁺ sputtering) further demonstrates that the F^- depth profiles peak at sputtering times of 80 s (SRCE), 161 s (LHCE), and 483 s (HCE), providing direct evidence of increasing SEI thickness in the order of SRCE < LHCE < HCE (Supplementary Fig. 13). The 3D reconstruction of TOF-SIMS images confirms that the SEI formed in SRCE exhibits a homogeneous and abundant distribution of LiF with minimal organic carbon, consistent with the XPS results (Fig. 3g). In contrast, the F element distribution in HCE and LHCE is sparse and heterogeneous, reflecting uneven anion reduction and the formation of a non-uniform SEI (Supplementary Figs. 14–15). Moreover, stronger carbon signals are detected in HCE due to constant solvent decomposition, which likely induces irregular lithium deposition and promotes dendrite growth.

Evaluation of full cell electrochemical performance

To evaluate anodic stability, we performed linear sweep voltammetry (LSV) on the three electrolytes (Fig. 4a). The HCE exhibits the lowest decomposition onset potential, accompanied by a pronounced oxidation current, indicating limited oxidative resistance. The SRCE shows high-voltage endurance, with decomposition potentials exceeding 5.0 V. Therefore, leveraging the higher lithium plating/stripping CE and anodic stability of the optimized SRCE, we assembled 50 μm Li | |2.5 mAh cm⁻² NCM811 full cells to comprehensively evaluate their long-term cycling performance. The cells with SRCE could deliver stable cycling over 800 cycles with 80% capacity retention, higher than that of 420 and 540 cycles in cells with HCE and LHCE (Fig. 4b). We further conducted leakage current measurements by holding the voltage at 4.3 V constantly to monitor the interfacial parasitic reactions (Fig. 4c). The SRCE cells experience a rapid current decay and stabilize at a low current of ~1 μA, suggesting the possible formation of a passivating cathode electrolyte interphase (CEI) that effectively passivates the NCM811 surface and mitigates parasitic reactions. Conversely, both HCE and LHCE exhibit higher leakage currents (8 μA and 4 μA), indicating sustained, more severe interfacial side reactions. Moreover, suppressed parasitic reactions could affect the kinetics of interfacial ionic transport. Electrochemical impedance spectroscopy (EIS) analysis during full cell cycling reveals markedly different impedance evolution (Fig. 4d, Supplementary Fig. 16 and Supplementary Table 6). As the cycle number increased from 20 to 400, quantitative analysis shows the interfacial resistance (R_SEI) increased by 290 Ω, 99 Ω, and 68 Ω for HCE, LHCE, and SRCE, respectively. This result underscores the suppression of progressive interfacial side reactions in SRCE, relieving battery aging and performance degradation. Consequently, the high ionic conductivity and the low overall resistance of SRCE promote the rate capability by maintaining 76% of initial capacity (0.2 C) at a high current rate of 4 C, exceeding LHCE (61% retention) and HCE (31% retention), as shown in Fig. 4e and Supplementary Fig. 17. Practical validation under demanding conditions further confirmed these advantages. Using a high-loading 50 μm Li | |4 mAh cm⁻² NCM811 cell, HCE demonstrates rapid capacity fading after ~70 cycles, followed by LHCE after ~130 cycles, while SRCE retained 85% capacity after 250 cycles (Fig. 4f). Furthermore, a 600 mAh-level pouch cell (20 μm Li | |3.6 mAh cm⁻² NCM811) employing SRCE and a lean electrolyte with an electrolyte-to-capacity (E/C) ratio of 1.6 g Ah⁻¹ achieved 87% capacity retention after 150 cycles (Fig. 4g and Supplementary Fig. 18). These results collectively validate our electrolyte design strategy in synergistic stabilization of both negative electrode and positive electrode interfaces, thereby enabling high-energy, long-cycling LMBs.

Fig. 4. Evaluation of electrolyte anodic stability and full cell electrochemical performance for three electrolytes.

a LSV curves of Li | |Al cells measured at 25 °C with a scan rate of 0.1 mV s⁻¹. b Cycling performance of 50 μm Li | |2.5 mAh cm⁻² NCM811 coin cells at 25 °C with 0.5 C charge/1 C discharge. c Leakage current during a 4.3 V constant-voltage floating test of 50 μm Li | |2.5 mAh cm⁻² NCM811 coin cells after 20 cycles at 25 °C. d Nyquist plots of 50 μm Li | |2.5 mAh cm⁻² NCM811 coin cells after 400 cycles at 25 °C. Inset: the impedance profile of the same cell after 20 cycles. e Rate capability of 50 μm Li | |2.5 mAh cm⁻² NCM811 coin cells at various C-rates at 25 °C. f Cycling performance of high-loading 50 μm Li | |4 mAh cm⁻² NCM811 coin cells at 25 °C (0.2 C charge/0.3 C discharge). g Cycling performance of a 600 mAh pouch cell with SRCE at 25 °C. Note: 1 C is defined as 200 mA g⁻¹ and electrolyte usage for coin cells is 40 μL per cell.

Analysis of positive electrode structural integrity and interfacial stability

To elucidate the fundamental stabilization mechanism of nickel-rich layered positive electrodes enabled by SRCE, we conducted systematic investigations spanning interfacial chemistry to structural evolution under high-voltage operation. XPS analysis reveals that SRCE facilitates the formation of a LiF and Li₂O-dominated CEI, characterized by an intense F 1 s peak at ~684.8 eV⁴¹, alongside O 1 s signal and attenuated C 1 s (Fig. 5a–c and Supplementary Fig. 19). This inorganic-rich interface originates from preferential anion decomposition driven by strong cation-anion coordination, effectively suppressing the solvent-derived organic species observed in HCE and LHCE. This CEI could further ensure effective electrochemical passivation and mechanical properties, e.g., higher density, smoothness, and hardness. Therefore, transmission electron microscopy (TEM) reveals an ultrathin ( < 4 nm), homogeneous, and conformal CEI formed in SRCE (Fig. 5d and Supplementary Fig. 20). Nevertheless, a heterogeneous and thick CEI ( ~ 20 nm) is formed in HCE, indicating persistent side reactions and uncontrolled interfacial growth; meanwhile, LHCE could partially suppress these parasitic reactions, yielding a CEI thickness of ~10 nm (Fig. 5e, f). Moreover, the atomic force microscopy in quantitative nanomechanical mapping (AFM-QNM) demonstrates that the SRCE-derived CEI not only displays a smoother morphology but also exhibits a higher average Derjaguin-Muller-Toporov (DMT) modulus of 9.1 GPa, compared to 4.5 GPa and 5.6 GPa for HCE and LHCE, respectively (Supplementary Fig. 21). This enhanced mechanical robustness contributed significantly to the structural and interfacial stability during high-voltage cycling.

Fig. 5. Analysis of structural evolution and CEI formation.

a–c F 1 s XPS spectra of NCM811 positive electrodes after 50 cycles at 25 °C in Li | |2.5 mAh cm⁻² NCM811 coin cells in three electrolytes. d–f TEM images of the CEI formed on NCM811 positive electrodes after 50 cycles in three electrolytes. g HAADF-STEM image of NCM811 positive electrodes after 200 cycles in SRCE. h ICP-OES measurements for TM dissolution detection after 100 cycles in three electrolytes. i Quantitative NMR analysis of electrolyte consumption of 50 μm Li | |2.5 mAh cm⁻² NCM811 full cells after 200 cycles. Note: All samples were prepared after a certain cycling at 0.5 C charge/1 C discharge (3-4.3 V), where 1 C stands for 200 mA g⁻¹.

The structural protective role of the CEI is further reflected in the microstructure stability of the positive electrode particles. After 200 cycles, high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) demonstrates only an ultrathin and uniform layer (~ 2.6 nm) was transformed into the rock-salt phase in SRCE, indicating that the CEI acts as a barrier to effectively prevent direct contact between the electrolyte and the highly reactive positive electrode surface for electrolyte decomposition, as confirmed by the low leakage current^6,42,43 (Fig. 5g and Supplementary Fig. 22). The CEI protection layer comprises a dense array of embedded inorganic nanocrystals, dominated by LiF and Li₂O (Supplementary Fig. 23). This structure could inhibit electron transfer and nucleophilic attack by electrolytes, thereby suppressing oxygen loss and transition metal (TM) dissolution, which are the culprits behind Li/Ni cation mixing and rock-salt phase formation. Therefore, rietveld refinement of X-ray diffraction patterns from NCM811 positive electrodes after 500 cycles reveals minimal cation mixing (~1.76%) and the dQ/dV profiles exhibit less pronounced peak attenuation and smaller voltage shifts, prolonging the cycling stability (Supplementary Figs. 24–25 and Supplementary Tables 7–9). In contrast, HCE and LHCE exhibit severe peak diminution and cation disordering (3.53% and 2.40%), promoting extensive rock-salt phase growth and accelerating structural degradation. Therefore, the loss of lattice integrity from TM dissolution and phase transformation, combined with the abrupt c-axis contraction during $H_2\to H_3$ phase transition, generates substantial internal stress within NCM particles and initiates microcracks⁴⁴. These cracks facilitate electrolyte penetration and corrosive attack on the grain boundaries, further weakening their cohesion and triggering a self-accelerating degradation. Consequently, after 500 cycles, ion beam-scanning electron microscopy (FIB-SEM) tomography reveals severe particle cracking formed in HCE, whereas LHCE exhibits localized microcracks and voids at the primary particle boundaries (Supplementary Fig. 26). In contrast, SRCE largely preserves the positive electrode particle integrity with no noticeable cracks, underscoring the critical role of a stable CEI in mitigating mechanical degradation.

The suppression of mechanical degradation further decelerates TM dissolution and electrolyte depletion. To evaluate TM dissolution during cycling, inductively coupled plasma-optical emission spectrometry (ICP-OES) was applied to quantify the amounts of Ni, Mn, and Co dissolved from the NCM811 electrodes after 100 cycles (see Methods for detailed information). The content of dissolved TMs in SRCE is only 0.39, 0.015, and 0.01 ppm for Ni, Mn, and Co, respectively (Fig. 5h). In sharp contrast, the positive electrodes from HCE and LHCE exhibit greater TM dissolution, especially for Ni ( > 2 times) and Mn ( > 10 times), further jeopardizing the Li metal negative electrode. (Supplementary Fig. 27). Moreover, the dissociated TM cations and crack-induced activation sites may further catalyze electrolyte decomposition. To link interfacial behavior to electrolyte evolution, we evaluated electrolyte consumption during cycling using quantitative ¹⁹F and ¹H NMR (see Methods for detailed information). After 200 cycles under lean electrolyte conditions, HCE and LHCE consume 53.9% and 41.2% of LiFSI salt, respectively (Fig. 5i and Supplementary Figs. 28–31). SRCE suppresses salt decomposition to 25.5%, together with minimized diluent and solvent consumption, confirming that effective CEI sufficiently mitigates failure modes associated with electrolyte exhaustion, thereby enhancing practical cycling performance. These multiscale analyses establish that the SRCE electrolyte promotes the formation of an inorganic-rich, mechanically coherent, and chemically stable CEI, which effectively decelerates both structural degradation of the NCM811 positive electrode and electrolyte decomposition, leading to significantly enhanced cycling performance in high-energy LMBs. Moreover, cross-swap experiments were conducted to disentangle the contributions of bulk transport and interphase chemistry to electrochemical performance (see Methods for detailed information). As shown in the Supplementary Fig. 32, the bulk electrolyte enhances rate capability by governing ion transport and interfacial desolvation, while the interphase extends cycle life by suppressing parasitic reactions and ensuring interfacial stability, together underpinning the enhanced electrochemical performance of SRCE.

Performance validation under harsh conditions and abuse testing

We have further established a multidimensional evaluation framework to assess the commercialization potential of LMBs under extreme operating conditions. For high-voltage operation (4.5 V), SRCE demonstrates enhanced cycling stability with 80% capacity retention after 500 cycles, longer than those in HCE (260 cycles) and LHCE (320 cycles) (Fig. 6a and Supplementary Fig. 33a–c). For elevated temperature (60°C), HCE suffers premature failure, whereas LHCE maintains effective operation for only 217 cycles (Fig. 6b and Supplementary Fig. 33d–f). In contrast, SRCE retains 80% capacity after 380 cycles with higher stability for both electrodes, demonstrating its enhanced thermal resilience and effectively interphase passivation against high-temperature degradation (Supplementary Fig. 34). For high specific energy demands, the Ni-richer positive electrode (LiNi_0.9Co_0.05Mn_0.05O₂, NCM90) with SRCE achieves 91% capacity retention after 430 cycles, longer than both HCE and LHCE (Fig. 6c and Supplementary Fig. 33g–i). These results collectively confirm the efficacy of SRCE in stabilizing positive electrode-electrolyte interfaces through reinforced interfacial protection, thus ensuring high specific energy under practical conditions.

Fig. 6. Performance validation under harsh conditions and abuse testing.

a Cycling performance of 50 μm Li | |2.5 mAh cm⁻² NCM811 coin cells at 25 °C in the voltage range of 3-4.5 V. b High-temperature (60 °C) cycling of 50 μm Li | |2.5 mAh cm⁻² NCM811 coin cells at 3–4.3 V. c Cycling performance at 25 °C with 50 μm Li | |2 mAh cm⁻² NCM90 coin cells at 3-4.3 V. d Calendar aging test of 50 μm Li | |2.5 mAh cm⁻² NCM811 coin cells at 25 °C, with each charge followed by a 24-hour rest interval before discharge. e Cycle testing of 2.2 Ah Li metal pouch cells at 25 °C (Inset: Photo of the pouch cell). f Nail penetration test on 1 Ah Li metal pouch cell: Optical and thermal images during penetration. g Voltage and temperature profiles during nail penetration (Inset: Schematic of temperature probe placement). h Overcharge voltage and temperature profiles of Li metal pouch cell (Inset: thermal image at 5.7 V). Note: The electrolyte usage for coin cells and pouch cells is 40 μL per cell and 1.6 g Ah⁻¹, respectively, where 1 C stands for 200 mA g⁻¹.

To comprehensively evaluate the electrolyte’s long-term stability in practical scenarios, particularly during static storage, we conducted calendar aging tests simulating prolonged idle conditions. SRCE-based 50 μm Li | |2.5 mAh cm⁻² NCM811 full cells retain 96% capacity after 8 months (180 cycles), with a monthly self-discharge rate below 0.5% (Fig. 6d, Supplementary Fig. 33j–l, and 35). In comparison, HCE exhibits severe self-discharge and rapid capacity decay within 2 months. LHCE shows partial improvement, where the capacity decays to 80% in less than 6 months (120 cycles). These findings highlight the long-term durability of SRCE, enabled by functional interfacial layers on both the positive and negative electrodes. Subsequently, we assembled Ah-level lithium metal pouch cells under lean electrolyte conditions (1.6 g Ah⁻¹). SRCE-based 2.2 Ah pouch cells could achieve a cycle life of 180 cycles with 85% capacity retention and rate performance, yielding a specific energy of 405 Wh kg⁻¹ based on the total mass of the pouch cell (Fig. 6e and Supplementary Figs. 36–37, and Supplementary Table 10). Moreover, with the LiNi_0.92Co_0.03Mn_0.05O₂ (NCM92) positive electrodes, the 2.6 Ah pouch cell delivers a specific energy of 475 Wh kg⁻¹, with a capacity retention of 91% after 100 cycles, while the 6 Ah-level pouch cell could achieve a higher specific energy over 500 Wh kg⁻¹ with leaner electrolyte of 0.93 g Ah⁻¹ (Supplementary Figs. 38–39, and Supplementary Table 11–12).

Acknowledging the potential for battery misuse in real-world applications, we further assessed the safety performance of SRCE-based pouch cells. In nail penetration tests on fully charged Ah-level pouch cells, the SRCE-based pouch cell displays no sparks or explosion, with only a 30°C temperature rise (less than 60°C) followed by rapid stabilization (Fig. 6f, g, Supplementary Fig. 40 and Movie 1). The fire-extinguishing property of the SRCE-based pouch cell derives not only from the nonflammable electrolyte but also from the high thermal stability of the CEI (Supplementary Figs. 41–42). Differential scanning calorimetry (DSC) analysis of fully charged NCM811 positive electrodes after 20 cycles shows that SRCE system generates minimal heat (131.9 J g⁻¹), compared with 832.6 J g⁻¹ for HCE and 303.6 J g⁻¹ for LHCE. Moreover, SRCE system exhibits a delayed onset temperature for initial heat release (182.1°C), higher than LHCE (171.9 °C) and HCE (143.6 °C), indicating that the SRCE-derived CEI could delay the initiation of electrolytes and substantially enhance thermal stability. Additionally, overcharge abuse testing was performed by charging an Ah-level pouch cell to 200% state of charge (SOC) for 2 h, during which the voltage increased to 5.7 V. The cell’s temperature increases by less than 10°C at 5.7 V and remains stable during sustained charging, with no observable swelling or leaking (Fig. 6h and Supplementary Fig. 43). These enhanced safety outcomes stem from the passivating electrode interfaces and intrinsic physicochemical stability of SRCE, which collectively mitigate thermal runaway under abusive conditions.

Discussion

In summary, we have designed a steric-regulation strategy for high-concentration electrolytes aimed at increasing the HSP distance between diluent and ion pairs. By leveraging HSP principles, a non-flammable and sterically hindered cyclic fluorinated alkane was introduced to weaken intermolecular interactions and disrupt ion pair clustering, thereby forming isolated and moderately aggregated ionic nanoclusters. These isolated nanoclusters facilitate rapid Li⁺ transport, suppress lithium dendrite growth, and strengthen cation-anion interactions, thereby promoting the formation of inorganic-rich interphases with enhanced chemical, mechanical, and thermal stability. Therefore, both electrodes demonstrate enhanced electrochemical performance in SRCE, where dendrite-free Li plating/stripping achieves a high CE of 99.61% and the 50 μm Li | |2.5 mAh cm⁻² NCM811 full cell delivers stable operation for over 800 cycles. At the Ah-level, 6 Ah pouch cells deliver a high specific energy exceeding 500 Wh kg⁻¹, while 2 Ah pouch cells could retain 85% capacity after 180 cycles. Moreover, the LiF-rich CEI endows the enhanced resilience under harsh conditions, including high operating voltage (4.5 V), elevated temperature (60 °C), compatibility with high-Ni positive electrodes (NCM90), and extended calendar aging (8 months). Safety tests under abuse conditions are also confirmed, with nail penetration tests showing no ignition and overcharge tests demonstrating tolerance up to 200% state of charge without swelling or explosion. This work addresses critical barriers toward practical high-energy LMBs and establishes a foundational design principle for next-generation electrolytes, contributing to their deployment in high-performance electric vehicles and large-scale grid energy storage for a sustainable society.

Methods

Materials

LiFSI ( ≥ 99.9%) and TTE ( ≥ 99.9%) were obtained from DodoChem. DME ( ≥ 99.5%) was purchased from Sigma-Aldrich and used as received. HFC ( ≥ 98%) was purchased from TCI and dehydrated with 4 Å molecular sieves (Sigma-Aldrich) for a week in the glove box before use. Dimethyl sulfoxide (DMSO-D₆, ≥ 99.9 atom % D), deuterium oxide (D₂O, ≥ 99.9 atom % D), hexafluorobenzene (C₆F₆, ≥ 99.0%), acetonitrile (CH₃CN, ≥ 99.9%), N-methyl-2-pyrrolidone (NMP, ≥ 99%), and LiCl ( ≥ 99.0%) were purchased from Aladdin. Li chip (16 mm $\times$ 0.6 mm, 99.95%), separator (2325-type, trilayer microporous membrane, 25 μm thickness, 39% porosity), Al foil (12 μm, battery grade), Cu foil (12 μm, 99.95%), Super-P (conductive carbon black, battery grade), polyvinylidene fluoride (PVDF, battery grade), and CR2032 cell cases were ordered from Canrd. 50 μm Li foil (99.95%) was purchased from Changsha Xinkang New Material Co. NCM811 ( ≥ 99.5%) and NCM90 ( ≥ 99.5%) were purchased from Canrd and MTI, respectively. Industrial-grade dry Li | |NCM pouch cells were provided by Chengdu Yanchuang Juli New Energy Technology Co., Ltd. (2.2 Ah pouch cells with NCM811 positive electrode, and 6 Ah pouch cells with NCM95 positive electrode) and Li-Fun Technology (2.6 Ah pouch cells with NCM92).

Electrolyte preparation

All electrolytes were prepared in an argon-filled glove box. The HCE was formulated with a molar ratio of LiFSI: DME = 1:1. The LHCE and SRCE were prepared by further introducing TTE and HFC diluents, respectively, into the base HCE, each achieving a final molar ratio of LiFSI: DME: Diluent=1:1:1.5. This composition was selected as the optimal electrolyte, offering a better balance of cost and Li⁺ reservoir, higher conductivity and reversibility, and enhanced full-cell cycling performance (Supplementary Fig. 44). All mixtures were stirred until homogeneous and allowed to equilibrate before use.

Electrode fabrication

The positive electrode was prepared by coating a slurry of NCM811 (purchased from Canrd) or NCM90 (MTI), Super-P, and PVDF (Mw=600000 g mol⁻¹) in a mass ratio of 94:3:3 in NMP. The slurry was coated onto a 12 μm Al foil using an MTI automatic film coater (model MSK-AFA-EI300) and dried at 80 °C for 12 h in a vacuum oven. After roll-pressing, the electrodes were punched into 8-mm-diameter discs using a manual disc cutter (MSK-2200-H5). The resulting active material loadings were approximately 12.5 mg cm⁻² for an areal capacity of 2.5 mAh cm⁻² NCM811, 21.5 mg cm⁻² for 4 mAh cm⁻² NCM811, and 10.5 mg cm⁻² for 2 mAh cm⁻² NCM90. The mass loading of the 50 μm Li foil is approximately 2.67 mg cm⁻² (based on lithium density of 0.534 g cm⁻³) and the 0.6 mm Li chip has a mass loading of approximately 32 mg cm⁻².

For coin cells, the N/P ratios (negative-to-positive electrode areal capacity ratios) were approximately 4.1, 2.6, and 5.2 for a 50 μm Li | |2.5 mAh cm⁻² NCM811 cell, a 50 μm Li | |4 mAh cm⁻² NCM811 cell, and a 50 μm Li | |2 mAh cm⁻² NCM90 cell, respectively. The N/P ratios for each pouch-cell configuration are provided in Supplementary Tables 10–12.

Cell assembly details

CR2032 coin cells were assembled with stainless steel 316 cases, a wave spring, a 16 mm Li negative electrode, an 8 mm NCM811 positive electrode (single-side coated) or an 18 mm Cu/Al foil, and 40 μL electrolyte. Pouch cells were injected with electrolyte, vacuum-sealed, and rested for 12 h at room temperature before testing. A constant external pressure of 100 kPa was applied during pouch-cell cycling. The E/C ratio was calculated based on the nominal (design) capacity of each pouch cell. Detailed pouch-cell parameters are provided in Supplementary Tables 10–12. All electrochemical measurements were performed in CR2032 coin cells unless otherwise stated (e.g., pouch cells, which are clearly indicated in the corresponding figure or table captions).

Electrochemical testing

All battery assemblies in this study were conducted in an argon-filled glove box, where oxygen and water content were below 1 ppm and 0.1 ppm, respectively. All room‑temperature electrochemical tests were conducted in a laboratory environment without a climatic chamber, with the ambient temperature recorded as 25 ± 3°C. High‑temperature tests (60 °C) were performed in a temperature‑controlled chamber with a temperature stability of ±1 °C.

The CEs of Li | |Cu cells were evaluated by a modified Aurbach method, in which the CE was calculated as the average of 5 trials. First, 5 mAh cm⁻² of Li was deposited onto a copper foil at 0.5 mA cm⁻², followed by stripping to 1 V to form a stable SEI. Then, 5 mAh cm⁻² of Li was redeposited as the lithium source. Following this, 1 mAh cm⁻² was deposited/stripped at 0.5 mA cm⁻² for 10 cycles, and the cell was finally stripped to 1 V. The CE was calculated by dividing the total stripping capacity by the total plating capacity. The rate performance of Li | |Li symmetric cells was evaluated by plating/stripping 2 mAh cm⁻² of areal capacity at different current densities, with five plating/stripping cycles performed at each current density. LSV measurements were performed using Li | |Al cells containing 40 μL of different electrolytes, with a voltage sweep from the open-circuit voltage to 6 V at a scan rate of 0.1 mV s⁻¹. The galvanostatic charge-discharge tests for Li | |NCM cells were conducted over a voltage range of 3–4.3 V. The procedure involves two formation cycles at 0.2 C, followed by a constant current (CC) charge at 0.5 C to 4.3 V, a constant voltage (CV) holds at 4.3 V until the current drops below 0.1 C, and finally a 1 C discharge (1 C = 200 mA g⁻¹). The leakage current test was conducted using Li | |2.5 mAh cm⁻² NCM811 coin cells with 40 µL of different electrolytes. After 20 cycles, the cells were charged to 4.3 V, followed by a 24-hour constant-voltage hold at 4.3 V while monitoring current decay. Electrochemical impedance spectroscopy (EIS) (Biologic SAS, France) analyses were performed using a Li | |NCM811 coin cell cycled for different cycles, followed by EIS tests in the frequency range of 1 MHz to 1 mHz and alternating voltage amplitude of 5 mV on a VMP3 multichannel electrochemical station (potentiostatic mode, 10 points per decade). Before each measurement, the cells were rested at open-circuit voltage for 10 min to reach equilibrium. Calendar life testing was performed on 50 μm Li | |2.5 mAh cm⁻² NCM811 cells with 40 μL electrolyte. Following 2 cycles at 0.2 C, the cells were charged to 4.3 V at 0.3 C. Thereafter, they were left to rest for 24 h prior to being discharged at 0.3 C. All lithium metal pouch cells were vacuum-sealed inside an argon-filled glovebox after electrolyte injection. Subsequently, they were subjected to a constant pressure of 100 kPa using a fixture and allowed to rest for 12 h prior to electrochemical testing. The Li⁺ transference numbers ($t_{{Li}^{+}}\left)\right)$ were determined by the following equation:

$$t_{{Li}^{+}}=\frac{I_s\left(\mathrm{\Delta }V-I_0{R}_0\right)}{I_0\left(\mathrm{\Delta }V-I_s{R}_s\right)}$$ 1

where I₀ and R₀ denote the initial current and interface resistance, while I_s and R_s represent their respective steady-state values.

Physical characterizations

The solvation structures of the electrolytes were characterized by NMR spectroscopy on a Bruker AVANCE III 600 MHz instrument. To maintain the native solvation environment without perturbation, the coaxial NMR tube configuration was used with the external reference in the inner tube and test electrolytes in the outer tube (Supplementary Fig. 45). This configuration could ensure the absolute chemical shift accuracy and eliminate baseline drift. External references were 1 M Zn(OTf)₂ in D₂O (−78.84 ppm) for ¹⁹F NMR, 1 M LiCl in D₂O (0 ppm) for ⁷Li NMR, and D₂O (0 ppm) for ¹⁷O NMR.

Quantitative NMR spectroscopy was employed to analyse electrolyte evolution after 200 cycles (Supplementary Fig. 46 and Supplementary Note 3). 50 μm Li | |2.5 mAh cm⁻² NCM811 full cells with precisely recorded electrolytes were disassembled in an argon-filled glovebox after cycling. All internal components were immediately submerged into a centrifuge tube pre‑filled with 2 mL DMSO‑D₆. The tube was then sealed without delay, vigorously vortexed, and left to rest for 12 h. For internal referencing, precisely weighed amounts of acetonitrile and hexafluorobenzene were added for ¹H and ¹⁹F analyses, respectively. The NMR spectra were acquired on a Bruker AVANCE NEO 400 MHz spectrometer and four independent parallel cells were tested for each condition to ensure statistical reliability.

The self-diffusion coefficients of Li⁺ ($D_{{Li}^{+}}$) in the three electrolytes were measured using pulsed-field gradient NMR (Bruker AVANCE III 600 MHz). The values for $D_{{Li}^{+}}$ were derived from fitting the attenuation of the Li⁺ signal with the Stejskal-Tanner equation:

$$\ln \left(\frac{S}{S_0}\right)=-{\gamma }^2{g}^2{\delta }^{2}D\left(\mathrm{\Delta }-\frac{\delta }{3}\right)$$ 2

Where γ is the gyromagnetic ratio, g is the gradient strength, δ is the gradient-pulse duration, Δ is the diffusion time, and D is the self-diffusion coefficient.

Synchrotron-based wide-angle X-ray scattering (WAXS) measurements were conducted at the WAXS beamline of SPring-8. The experiments utilized a monochromatic X-ray beam with an energy of 10 keV and a beam size of 75 × 75 μm². The sample-to-detector distance was set at 84.8 mm. Prior to measurement, the electrolyte was sealed in a Hilgenberg glass capillary with a diameter of 1.5 mm. The obtained 2D scattering patterns were processed using the FIT2D software, and all data were corrected by subtracting the background signal from an empty capillary and air.

X-ray photoelectron spectroscopy (XPS, Thermo Scientific NEXSA) was employed to analyze the interface chemistry of both the cycled Li metal negative electrode and the NCM811 positive electrode. Electrode samples were rinsed with DME and dried under vacuum. For the SEI of Li metal negative electrode, 5 mAh cm⁻² of Li was deposited onto a Cu foil at 0.5 mA cm⁻² before disassembly. The harvested Li electrode was then rinsed with DME and vacuum-dried. To prevent oxidation, all samples were transferred and measured using a vacuum transfer vessel. For the CEI of the positive electrode, the NCM811 electrode after 50 cycles was similarly washed with DME and dried under vacuum. All XPS spectra were calibrated by setting the adventitious carbon C 1 s peak to 284.8 eV.

TOF-SIMS analysis in negative-ion mode (Ga⁺ sputtering source) was conducted on a cycled Li metal surface using a TESCAN C-TOF system. An area of 50 × 50 μm was etched with a 0.15 nA, 30 kV ion beam. The Li metal electrode was prepared using the same procedure as described for the XPS SEI analysis. The same instrument was also employed for FIB-SEM cross-sectional analysis to examine particle cracking in the NCM811 positive electrode after 500 cycles, using a 10 nA ion beam for material removal and imaging.

Raman spectra were acquired on a MonoVista CRS system with 532 nm excitation to probe the solvation structure in various electrolytes. Sample handling was performed in an argon-filled glovebox, where electrolytes were loaded and sealed in capillaries for measurement.

The surface morphologies of the cycled negative electrodes and positive electrodes were characterized using a Hitachi SU8020 scanning electron microscope (SEM). Electrode samples were rinsed with DME and dried under vacuum.

TEM was conducted on a JEM-2200FS instrument operated at an accelerating voltage of 200 kV. The NCM811 positive electrode material was scraped from the current collector after 50 cycles, ultrasonically dispersed in ethanol, and drop-cast onto a copper grid for observation.

HAADF-STEM was performed using a JEM-ARM300F microscope equipped with a probe aberration corrector (JEOL) at an accelerating voltage of 300 kV. Prior to the analysis, the NCM811 positive electrode after 200 cycles was thinned to below 100 nm using a focused ion beam system (FEI Scios 2 HiVac).

ICP-OES (Agilent 725) was used to determine the dissolution of transition metals in the positive electrode material of a Li | |2.5 mAh cm⁻² NCM811 coin cell after 100 cycles (Supplementary Fig. 47). To accurately determine the respective concentrations of individual transition metals (Ni, Co, Mn) retained in the residual electrolyte and deposited on the Li negative electrode, the cycled cell was disassembled in an argon-filled glovebox. The positive electrode chip was rinsed with 5 mL of DME, and the washings were collected. The remaining battery components, including the separator, lithium foil, spacer, contact spring, and positive electrode and negative electrode shells, were immersed in the same wash liquid for 24 h. The Li chip was subsequently removed and air-oxidized for 48 h. The wash liquid was evaporated at 80 °C in a fume hood, and the resulting residue was dissolved in 2 g H₂O and 1 g nitric acid. The oxidized Li metal was separately dissolved in a mixture of 7 g H₂O and 1 g nitric acid. Finally, these two digestion solutions, one containing TMs from the electrolyte and surface adsorption and the other containing TMs deposited on Li metal negative electrodes, were combined and filtered for ICP-OES analysis.

AFM measurements were performed on a Bruker Dimension Icon system using Quantitative Nanomechanical Mapping (QNM) mode to characterize the mechanical properties of the positive electrode-electrolyte interphase formed on NCM811 positive electrodes after 50 cycles. Electrode samples were rinsed with DME and dried under vacuum.

Safety characterizations

The thermal stability of both the electrolyte and the fully delithiated NCM811 was evaluated using differential scanning calorimetry (DSC, TA Q2000). Li | |2.5 mAh cm⁻² NCM811 cells after 20 cycles were charged to 4.3 V at 0.1 C to achieve the highly delithiated NCM811. The positive electrode material was then retrieved and mixed with the electrolyte at a mass ratio of 5:2. The mixture was heated from 50 °C to 350 °C at a scanning rate of 5 °C min⁻¹ under a N₂ atmosphere.

Nail penetration tests were conducted using a programmable compression and penetration tester (GAOXING, GX-5067-CSM20T). A 1 Ah lithium metal pouch cell (E/C = 1.6 g Ah⁻¹) was first activated with one formation cycle at 0.2 C and charged to 4.2 V prior to testing. The test was performed using a 5 mm diameter penetration needle at a speed of 25 mm/s. During the test, voltage and temperature probes were connected to the pouch cell to monitor its behavior, while the entire process was simultaneously recorded using a smartphone and a thermal imaging camera (HIKMICRO, K20) to capture visual and thermal responses. The overcharge test involved charging a pouch cell to 6.45 V (1.5 times the normal cutoff voltage of 4.3 V) with a 2-hour hold period. However, the voltage plateaued at 5.7 V due to electrolyte decomposition. Cell behavior was monitored using voltage/temperature probes and simultaneously recorded with a thermal camera and smartphone.

Theoretical calculations

Classical molecular dynamics (MD) simulations were performed with GROMACS⁴⁵, employing the OPLS-AA/L force field to describe the intermolecular interactions of the solvent molecules. The initial configuration was constructed via PACKMOL to ensure homogeneous spatial distribution⁴⁶. The system first underwent energy minimization through the steepest descent algorithm until the maximum force on any atom was below 100 kJ mol⁻¹nm⁻¹, with a 1 fs integration step over a 10 ps relaxation period. Subsequently, an annealing process was applied, gradually cooling the system from 700 K to 300 K under constant-volume conditions to remove residual strain. Thereafter, isothermal-isobaric (NPT) equilibration was conducted for 10 ns using the V-rescale thermostat coupled with a Berendsen barostat to obtain the target density. The equilibrated configuration was then subjected to an additional 10 ns canonical (NVT) simulation for structural stabilization. Finally, the radial distribution functions (RDFs) were calculated from the equilibrated trajectories with an in-house Python analysis script (Supplementary Data 2).

Density functional theory (DFT) calculations were carried out with the Gaussian 16 software package⁴⁷. The initial geometries, obtained from molecular dynamics simulations, were fully optimized at the B3LYP-D3(BJ) level of theory employing the 6-311 + + G (d, p) basis set (Supplementary Data 1). To ensure that each structure corresponds to a true local minimum on the potential energy surface, vibrational frequency analyses were subsequently conducted, confirming the absence of imaginary frequencies. The nature of hydrogen-bond interactions was analyzed based on the Atoms in Molecules (AIM) theory using the topological analysis module implemented in the Multiwfn program⁴⁸.

The calculations are intended to provide qualitative and comparative insights. DFT considers selected static molecular configurations, whereas classical MD employs finite-size, non-reactive models over limited timescales without explicit electrode interfaces, applied electric fields, or interphase-forming reactions. Therefore, the calculated binding energies, RDFs, and coordination numbers should not be regarded as exact descriptions of the electrolyte under operating conditions and are interpreted together with the experimental results.

Author contributions

J.T., H.J. and F.D. conceived the original idea and designed the experiments. Z.C. contributed to the theoretical calculations and simulations. J.T. carried out the majority of the experiments and material characterizations. C.W. assisted with NMR testing and data analysis. Z.Z.W. and W.Z. performed the STEM-HAADF characterization and data analysis. J.T., H.J., Z.X.W., G.Z. and F.D. analyzed the physical characterization and electrochemical data. J.T. and H.J. co-wrote the manuscript. All authors participated in discussions and contributed to the final version of the manuscript.

Peer review

Peer review information

Nature Communications thanks Ruhong Li, Hayoung Park, and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This work is supported by the National Natural Science Foundation of China, Grant No. 92372113 (F.D.), 22309059 (H.J.), and the Department of Science and Technology of Jilin Province (20250205069GH, H.J.).

Data availability

The data that support the findings of this study are available 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/s41467-026-77589-8.