Solvation sheath reorganization enables fast ion transfer kinetics in lithium-ion battery

Abstract

The limitations of ion transport kinetics in conventional electrolytes, particularly under extreme operating conditions, arise from suboptimal solvation structures and inefficient charge carrier utilization. Here, we present strategic electrolyte design that reconfigures Li⁺ coordination geometry by modulating intermolecular interactions and solvent molecule volume, fundamentally overcoming these transport constraints. By incorporating an optimized moderator with a low dipole moment and small molecular size, extensive anion aggregation is effectively disrupted into compact ion conduction domains, simultaneously increasing the number of free charge carriers and enhancing ion mobility. Guided by this principle, the designed electrolyte with dichloromethane (85.11 Å, 2.36 Debye) exhibits rapid Li⁺ hopping between adjacent coordination sites (152.3 ps for acetonitrile and 115.7 ps for FSI^-). This electrolyte enables stable cycling of 1.0 Ah 4.5 V graphite (3.13 mAh cm^-2)||LiNi_0.8Mn_0.1Co_0.1O₂ (2.85 mAh cm^-2) pouch cells, delivering 0.87 Ah at −40 °C, surpassing commercial carbonate-based electrolytes, which fail to retain reversible capacity at this temperature. This study establishes fundamental principles for fast ion-transport electrolytes, paving the way for next-generation Li-ion batteries under extreme scenarios.

Slow ion transport limits Li-ion battery performance under fast charging and low temperature. This work establishes an electrolyte design framework based on intermolecular interactions and molecule volume to reshape Li+ solvation structure, enabling Li-ion battery to operate stably at −50 °C and 6 C

Introduction

The rapid expansion of battery applications, ranging from portable electronics to electric vehicles, grid storage, and electrified aviation, has driven an increasing demand for high-performance Li-ion batteries (LIBs)^1–3. To meet these requirements, LIBs must achieve high energy density, support high-rate operation, and maintain stable performance at low temperatures without compromising the cycle life^4–6. To fulfill these criteria, electrolytes must facilitate the formation of robust inorganic interphase on both graphite negative electrode and high-voltage positive electrodes while exhibiting high ionic conductivity and low freezing temperatures—properties governed by the complex interplay among Li⁺, solvents, and anions^7–10. Consequently, rational electrolyte design must strategically modulate these intermolecular interactions to establish an optimized solvation environment, ensuring electrochemical stability under extreme operating conditions^11–13.

The challenge of electrolyte modulation is exacerbated by the intrinsic limitations of conventional electrolyte systems. Carbonate-based electrolytes, for instance, rely on high-dielectric-constant and large-sized solvents to weaken the cation-anion interactions through a shielding effect¹¹. However, this also leads to the formation of rigid solvation structures, as Li⁺ coordinates strongly with carbonyl oxygens, favoring vehicular ion transport^14–16. While such solvation structures with bulky solvent-rich shells contribute to ion transport stability, they also restrict Li⁺ mobility by extending Li⁺ diffusion distances and reducing the density of effective charge carriers, which are defined as mobile ions participating in bulk conduction. This trade-off fundamentally limits the ion transport kinetics of electrolyte systems (Fig. 1a and Supplementary Note 1)^7,16,17. Approaches such as increasing salt concentration^18–21 or incorporating weakly solvating solvents^22–25 enhance the anion integration into the solvation sheaths, leading to the formation of densely packed ion aggregates (AGGs). Although these AGGs facilitate the formation of robust solid electrolyte interphase (SEI) and mitigate interfacial resistance, they simultaneously diminish the availability of free charge carriers^26,27. Moreover, under an applied electric field, the dynamic coordination interactions impose kinetic barriers to ion migration due to time-dependent ligand exchange and asymmetric ion-pair drift (Supplementary Note 2), ultimately suppressing the ionic conductivity to below 5.00 mS cm^{−1 28,29}. Addressing this fundamental trade-off between ionic conductivity and interfacial resistance remains a critical challenge in electrolyte development. Although small-molecule solvents can significantly enhance ionic conductivity, their excessive cost and potential toxicity often constrain large-scale implementation³⁰. Therefore, a molecular engineering approach that precisely regulates intermolecular interactions and solvation structures is essential for next-generation electrolyte design^31–33.

Fig. 1. Moderator screening strategy.

a Variation of Li⁺ transfer energy (E_trans) as a function of the natural logarithm of the total effective charge carriers (Σn_i) in conventional carbonate-based electrolytes and reorganized electrolytes. b Functional relationship between the disintegration of solvation structures (D) and moderator properties [dipole moment (μ_moderator) and molecule volume (V_moderator)]. c μ_moderator and V_moderator for various moderators. The color bar denotes the D value (9.44 × 10⁻⁴ to 4.99 × 10⁻³Å⁻³ Debye⁻¹), with warmer colors corresponding to higher D values and cooler colors corresponding to lower D values. d Transfer energy barriers for the discrete hopping mechanism and structure mechanism in small-sized and large-sized moderator environments, respectively. Source data are provided as a Source data file.

Herein, we present an electrolyte design strategy that reorganizes the solvation structure by incorporating a low-dipole moment, small-sized moderator. This non-coordinating solvent modulates the anion and solvent participation in the solvation sheath via ion-dipole and dipole-dipole interactions. By penetrating the solvation shell, these moderator molecules disrupt large solvation clusters into contact ion pairs (CIPs), reducing Li⁺-anions interaction. This transformation leads to the formation of uniform, compact solvation structures, which significantly increase the population of mobile Li⁺ and reduce the Li⁺ hopping time, thereby enabling fast Li⁺ transfer kinetics in liquid-phase transport and interfacial desolvation through a discrete hopping mechanism^34,35. To guide moderator selection, we introduce a key parameter (D), representing the extent of solvation cluster disintegration, where higher D values favor fast Li⁺ desolvation. Applying this principle, we identify dichloromethane (DCM) as an optimal moderator due to its maximal D value. The resulting electrolyte, 1.5 M lithium bis(fluorosulfonyl)imide (LiFSI) in an acetonitrile (AN)-DCM mixture (1:4, vol) (denoted as FA-DCM electrolyte), demonstrates competitive electrochemical performance, combined with cost-effectiveness (Supplementary Note 3). In 4.5 V graphite||LiNi_0.8Mn_0.1Co_0.1O₂ (NMC811) full cells, the FA-DCM electrolyte enables a capacity of 179.8 mAh g⁻¹ at 6 C. Furthermore, at a low temperature of −50 °C, a 1.0 Ah 4.5 V graphite||NMC811 pouch cell (with an areal capacity of 2.85 mAh cm⁻² for the positive electrode and 3.13 mAh cm⁻² for the negative electrode) exhibits stable cycling, delivering a high capacity of 0.51 Ah. This study establishes an effective strategy for developing electrolytes tailored for fast-charging and low-temperature LIBs, offering valuable guidance for practical high-performance energy storage systems.

Result and discussion

Solvation structure reorganized strategy

In high-concentration electrolyte systems, Li⁺ exhibits strong coordination with anions and solvent molecules, resulting in sluggish desolvation dynamics^36,37. Effectively overcoming this limitation requires the disruption of over-coordinated Li⁺-solvent/anion clusters to enhance ion transport. To achieve this, we introduce ion-dipole/dipole-dipole interactions (μ_moderator) and moderator molecular volume (V_moderator) as dual descriptors for selecting ideal moderators to reorganize the solvation sheath. As revealed in Fig. 1b, the extent of solvation cluster disintegration (D) provides a quantitative measure of this regulatory effect (Supplementary Note 4).

$$D=C\cdot \frac{1}{{\mu }_{\text{moderator}}}\cdot \frac{1}{V_{\text{moderator}}}$$ 1

where C is the constant parameter (full derivation and parametric analysis are provided in Supplementary Note 4). The D value quantifies the ability of a moderator molecule to regulate the Li⁺-anions-solvent interactions, which originates from the competitive coordination strength, dictated by the ion-dipole/dipole-dipole interaction energy, and steric accessibility, determined by the molecular volume (Supplementary Note 4). Specifically, D is inversely proportional to both the μ_moderator and V_moderator. A smaller V_moderator facilitates deeper penetration into the Li⁺ solvation sheath, and a low μ_moderator alters the interaction energy balance in favor of moderator-anion coordination over moderator-solvent coupling. This synergistic effect allows the moderator to effectively compete for coordination sites and disrupt the robust Li⁺-anion clusters. This facilitates the disintegration of large clusters into homogenously dispersed, smaller solvation units with reduced local polarity fluctuations. The formation of the compact CIP clusters effectively lowers the desolvation energy barrier and increases the number of effective charge carriers in the electrolyte. The D value is derived from fundamental molecular properties (μ and V), ensuring its generalizability beyond AN-based electrolytes to a wide range of solvent systems.

Among the various moderators examined, DCM exhibits the most favorable properties for solvation structure modulation, possessing the smallest molecular volume (85.11 Å) and the lowest dipole moment (2.36 Debye) (Fig. 1c and Supplementary Table 3). By leveraging its compact molecular dimensions, DCM infiltrates interstitial voids within solvation sheaths, effectively disrupting the bulky solvation structures. Its low dipole moment further weakens the Li⁺-FSI⁻ interaction, facilitating the formation of homogeneous Li⁺ conduction pathways. As a result, the uniformly dispersed CIPs enable discrete hopping transfer through weakly coordinated ligands, promoting faster and long-range Li⁺ migration (Fig. 1d).

In contrast, moderators with large molecular size and high dipole moments exhibit counterproductive effects, as increased steric hindrance and strong dipole interactions tightly encapsulate the solvation sheath peripheries, leading to localized Li⁺ clustering and restricted mobility^36,38. Furthermore, the formation of a rigid ligand coordination network imposes significant kinetic barriers on ionic transport, ultimately reducing overall ionic conductivity and Li⁺ transference number^39–41. Overall, using the D value as a key parameter for screening moderators that enable precise regulation of the solvation environment provides a robust strategy for optimizing electrolyte design to achieve fast Li⁺ conduction.

In order to systematically investigate the influence of different moderators on solvation structures, 1,1,2,2-tetrafluoroethyl-2,2,3,3-tetrafluoropropyl ether (TTE)^42,43, characterized by its large molecular size and high dipole moment, was chosen for comparative analysis. The corresponding electrolyte, 1.5 M LiFSI in an AN-TTE mixture (1:4, vol) (denoted as FA-TTE electrolyte), was designed as the reference electrolyte. The independent gradient model based on Hirshfeld partition (IGMH) was employed to visualize the primary dominant coordination network, revealing specific bonding motifs. The DCM (Cl) interacts with AN (H) via dipole-dipole interactions, while DCM (H) coordinates with both AN (N) and FSI⁻ (O) (Fig. 2a, Supplementary Fig. 1 and Supplementary Data 1). To further elucidate the regulatory role of the moderator in Li⁺ solvation structures, molecular dynamics (MD) simulations were performed. The heatmap of Li⁺ solvation environment (3–5 Å from Li⁺) in the FA-DCM (Fig. 2b and Supplementary Figs. 2 and 3) indicates that Li⁺ is surrounded by an average of 4.54 DCM molecules and 1.66 FSI⁻, resulting in a more dynamic environment compared to FA-TTE (3.05 TTE and 1.84 FSI⁻). The incorporation of DCM breaks the bulky and rigid three-dimensional structures into smaller units, leading to CIPs as the predominant solvation configuration in the FA-DCM electrolyte (Supplementary Fig. 4a). On the contrary, the introduction of TTE moderators strengthens the Li⁺-FSI⁻ coordination, increasing the proportion of AGGs (Supplementary Fig. 4b). The solvated sheath structures with the highest proportion in FA-DCM electrolyte (Li⁺-FSI⁻-2AN) and FA-TTE electrolyte (Li⁺-3FSI⁻-AN) are illustrated in Supplementary Fig. 5 (Supplementary Data 4). Notably, the FA-DCM electrolyte exhibits the highest proportion of CIPs among the thirteen analyzed moderator-based electrolytes (Supplementary Figs. 4c–n and 6). To quantify the compactness of Li⁺ coordination, a distance index was introduced by summing the reciprocal distances between Li⁺ and N (AN)/O (FSI⁻). A higher distance index corresponds to a closer spatial arrangement of solvation components. As shown in Fig. 2c and Supplementary Fig. 7, the distance index for AN (1.2 Å⁻¹) and FSI⁻ (1.8 Å⁻¹) in FA-DCM electrolyte is significantly lower than that of FA-TTE electrolyte (1.2 Å⁻¹ for AN, 4.0 Å⁻¹ for FSI⁻), indicating that DCM facilitates the formation of smaller solvation structures.

Fig. 2. Solvation structures of different electrolytes.

a Sign(λ₂)ρ colored isosurfaces of δg^inter = 0.005 a.u. corresponding to independent gradient model based on Hirshfeld partition analysis for DCM-AN, DCM-FSI⁻ complexes (lime green, Cl atom; red, O atom; yellow, S atom; purple, N atom; silver, F atom; tan, C atom; white, H atom). The color bar represents interaction intensity (a.u.) (−0.002 to 0.004), where red indicates attractive interactions, white denotes van der Waals interactions, and blue corresponds to steric repulsion. b Heatmaps of Li⁺ solvating environment (3–5 Å around Li⁺) in FA-DCM electrolyte, the color bar represents the proportion of the number of this species to the total number of clusters. c Distance index of FSI⁻ and AN in FA-DCM electrolyte. Distance index = Σ1/x_n, where x_n represents the distance between Li⁺ and N (AN)/O (FSI⁻), the color bar represents the density of Li⁺. d Coaxial ⁷Li nuclear magnetic resonance spectra of different electrolytes. e MD simulation snapshots of FA-DCM electrolyte. (lime green, Cl atom; red, O atom; yellow, S atom; purple, N atom; silver, F atom; tan, C atom; white, H atom), the color bar represents the number density of atoms. f Electrical in-situ Fourier-transform infrared spectroscopy of FA-DCM electrolyte assembled in Li||graphite cell during discharging/charging at 0.2 C, 25 °C, where the color bar represents the relative intensity of the infrared absorption signals (left). The corresponding electrochemical profiles of Li||graphite cells (right). 1 C = 372 mA g⁻¹ (specific current for graphite). Source data are provided as a Source data file.

Coaxial nuclear magnetic resonance (NMR) spectroscopy was employed to further investigate the solvation structures across different electrolytes (Fig. 2d). A downfield shift from −2.09 ppm in 1.0 M LiFSI/AN (denoted as FALCE electrolyte) to −1.82 ppm in 5.0 M LiFSI/AN (denoted as FAHCE electrolyte) suggests a dominant deshielding effect caused by enhanced Li⁺-FSI⁻ coordination, which outweighs the shielding contribution from AN solvent. Notably, the chemical environment of Li⁺ in FA-DCM electrolyte (−2.07 ppm) closely resembles that of FALCE electrolyte, as further confirmed by Raman spectroscopy (AGGs, 18.4%; CIPs, 69.7%; solvent-separated ion pairs (SSIPs), 11.9%; Supplementary Fig. 8). However, the FA-TTE electrolyte, due to the large molecular size and high dipole moment of TTE, fails to effectively regulate the solvation structure, retaining a strong Li⁺-anion coordination environment (−1.82 ppm) similar to FAHCE, which corresponds well with the distribution of AGGs (FA-TTE, 62.0%; FAHCE, 61.6%), CIPs, and SSIPs clusters obtained from the deconvolution of the Raman spectra (Supplementary Fig. 8b and Supplementary Table 4). This results in the formation of large, tightly bound ion clusters, which hinders the Li⁺ desolvation at the interface and imposes a higher energy barrier for structural hopping (Supplementary Fig. 9 and Supplementary Data 2). In contrast, the DCM-based electrolyte, characterized by its small-sized and spatially uniform solvation structures, enables a low-barrier discrete hopping mechanism through optimized ligand-exchange kinetics (Fig. 2e and Supplementary Data 2). To experimentally validate the simulated solvation structures, we performed electrical in-situ Fourier-transform infrared (FTIR) spectroscopy on Li||graphite cells assembled with FA-DCM electrolyte. During discharging/charging, we observed a progressive decrease in the intensity of the coordinated FSI⁻ vibrational band (Ⅰ: 746 cm⁻¹)⁴⁴, accompanied by a simultaneous increase in the intensity of two characteristic bands associated with uncoordinated FSI⁻ (Ⅱ: S-F stretch, 830 cm⁻¹; Ⅲ: SO₂ stretch, 1180 cm⁻¹)⁴⁴. This evolution directly indicates a gradual dissociation of FSI⁻ from Li⁺, consistent with the breakdown of large clusters. This is attributed to the DCM-FSI⁻ interactions, which weaken the FSI⁻-Li⁺ binding, thereby reducing the number of coordinated FSI⁻; otherwise, the peak intensity of FSI⁻ would remain essentially unchanged even in the presence of an electric field⁴⁵. The results of diffusion ordered spectroscopy (DOSY) NMR corroborate the above results; the self-diffusion coefficients obtained by DOSY NMR of Li⁺ in FA-DCM electrolyte (3.330 × 10⁻⁷ cm² s⁻¹) are higher than those of FA-TTE electrolyte (1.133 × 10⁻⁷ cm² s⁻¹), suggesting the faster Li⁺ transport in FA-DCM electrolyte (Supplementary Fig. 10 and Supplementary Table 5). This highlights the effectiveness of DCM in promoting fast Li⁺ transport and enhancing electrolyte performance.

Fast Li⁺ transfer kinetics

To elucidate the modulation effects of DCM on the Li⁺ transport kinetics, the ion migration dynamics was systematically analyzed using three-dimensional lifetime-displacement correlation mapping (Fig. 3a, b and Supplementary Fig. 11). Quantitative trajectory analysis reveals that Li⁺ exhibits a migration distance of 338 pm in the FA-DCM electrolyte within 75 ps, significantly exceeding the 220 pm observed in the FA-TTE electrolyte. Additionally, the Li⁺ solvation clusters in FA-DCM electrolyte exhibit a shorter lifetime (˂200 ps) compared to FA-TTE electrolyte (345 ps), indicating that the rapid hopping of Li⁺ between clusters. The enhanced Li⁺ migration in the FA-DCM electrolyte is attributed to the shift in Li⁺ migration mode induced by DCM. MD simulations were performed to analyze Li⁺ transport behavior, focusing on solvated sheath transport kinetics in the bulk electrolyte and the solvent/anion residence times around the Li cations (Fig. 3c, d, Supplementary Figs. 12–15 and Supplementary Data 4). In the FA-DCM electrolyte, Li⁺ demonstrates a short residence time near solvents and anions, with an average time of 152.3 ps for AN and 115.7 ps for FSI⁻ (Fig. 3c, e). This reduced residence time leads to an elevated Li⁺ hopping frequency between adjacent coordination sites, thereby facilitating rapid ion transport. By contrast, in the FA-TTE electrolyte, the average residence times for Li-AN and Li-FSI⁻ increase to 234.4 and 178.6 ps, respectively (Fig. 3d, e). Owing to its relatively high D value of dichloroethane (DCE), the 1.5 M LiFSI in an AN-DCE mixture (1:4, vol) (denoted as FA-DCE electrolyte) demonstrates a residence time close to that of FA-DCM (177.45 ps for AN and 132.26 ps for FSI⁻) (Supplementary Figs.14a and 15f). In contrast, the 1.5 M LiFSI in an AN-1,1,2,2,3,3,4-heptafluorocyclopentane (HFC) mixture (1:4, vol) (denoted as FA-HFC electrolyte), with an intermediate D value, exhibits a significantly longer residence time (180.71 ps for AN and 163.55 ps for FSI⁻) (Supplementary Figs.14b and 15g). Such a linear inverse relationship between the D value and residence time corroborates the effectiveness of this descriptor in predicting ion transport properties (Supplementary Fig. 16).

Fig. 3. Physical properties of the electrolytes.

Two-dimensional Lifetime/displacement function of AN in FA-DCM electrolyte (a) and FA-TTE electrolyte (b), where the color bar represents the statistical intensity of AN molecules at a given combination of lifetime and displacement. Evolution of Li⁺ solvation environment in FA-DCM electrolyte (c) and FA-TTE electrolyte (d) from MD simulations at 298 K (the vertical axis represents the serial number of different solvents or anions coordinated with the investigated Li⁺ in the simulation time of 2000 ps). e The average residence time of Li⁺ in LiFSI/AN-based electrolytes with different moderators. f Temperature dependence of the ionic conductivity of different electrolytes. g Li⁺ transference numbers of different electrolytes at 25 °C. h Differential scanning calorimetry cooling and heating curves of different electrolytes. Source data are provided as a Source data file.

Moreover, to approximate practical conditions, we applied an electric field (E = 0.1 V Å⁻¹) in the simulations to further clarify the solvation dynamics of the electrolyte under non-equilibrium states. Notably, the interfacial environment remains relatively stable instead of exhibiting increasing kinetic heterogeneity in FA-DCM electrolyte (Supplementary Fig. 17), thus maintaining a more homogeneous solvation environment and markedly accelerated Li⁺ hopping. In contrast, the Li⁺ hopping frequency in FA-TTE is lower than that in the FA-DCM electrolyte (Supplementary Figs. 18–20). These findings highlight the DCM’s ability to fine-tune the interactions between Li⁺ and FSI⁻/AN, leading to fast ion transport in the liquid phase. The ionic conductivity and Li⁺ transference numbers of the different electrolytes were summarized in Fig. 3f, g, respectively. At 25 °C, the FA-DCM electrolyte, which has relatively low viscosity (4.66 mPa s, Supplementary Fig. 21) and moderate density (1.426 g mL⁻¹, Supplementary Table 6), exhibits moderate ionic conductivity (9.15 mS cm⁻¹), significantly higher than that of the large-sized moderator electrolyte (FA-TTE, 3.09 mS cm⁻¹) and the classic electrolyte⁴⁶ of 1.0 M lithium hexafluorophosphate (LiPF₆) in an ethylene carbonate (EC)-methyl propionate (MP)-ethyl methyl carbonate (EMC) mixture (1:3:1, vol) (denoted as BEMP electrolyte, 2.91 mS cm⁻¹). Even at a low temperature of −70 °C, the FA-DCM electrolyte, with its lower activation energy (0.059 eV), maintains an ionic conductivity of 2.01 mS cm⁻¹, ensuring adequate low-temperature ion transfer kinetics²⁶ (Supplementary Fig. 22). In contrast, the commercial carbonate-based electrolyte of 1.0 M LiPF₆ in an EC-dimethyl carbonate (DMC) mixture (1:1, vol) (denoted as BE electrolyte) exhibits slightly higher ionic conductivity at 25 °C (12.00 mS cm⁻¹) but suffers a sharp decline at −38 °C (0.06 mS cm⁻¹) due to its high freezing point and elevated activation energy (1.50 eV), making it unsuitable for low-temperature applications (Fig. 3h). Besides, the FA-DCM electrolyte possesses a high Li⁺ transference number (0.63) (Fig. 3g and Supplementary Fig. 23). The phase transitions of different electrolytes were identified through differential scanning calorimetry (Fig. 3h). The BEMP, FA-TTE, BE electrolytes exhibit observable exothermic peaks at around −88, −74 and −38 °C, respectively. Notably, the FA-DCM electrolyte maintains a stable single-phase state from −100 to 25 °C, ensuring reliable performance even under low temperature conditions.

Low-temperature electrochemical performance of graphite||NMC811 full cells

To evaluate the rate performance of graphite electrodes with different electrolytes, Li||graphite half cells were assembled. The cell using the FA-DCM electrolyte delivers a high specific capacity of 387.9 mAh g⁻¹ at 0.1 C and maintains capacities of 372.9, 361.5, 350.5 mAh g⁻¹ at 1, 5, 10 C, respectively. Moreover, it maintains cycling stability, retaining 94.0% of initial capacity after 1000 cycles at 10 C (Fig. 4a, b and Supplementary Fig. 24). These results indicate that the previously reported incompatibility between AN and graphite has been effectively resolved, and the sluggish Li⁺ desolvation kinetics at the interphases have been significantly improved (Supplementary Fig. 25)^18,47. In addition, the Li⁺ intercalation behaviors in various electrolytes were assessed through electrochemical impedance spectroscopy measurements. The total resistance in the cell primarily comprises bulk resistance (R_b), surface layer resistance (R_SEI), and charge transfer resistance (R_ct) (Fig. 4c, Supplementary Fig. 26 and Table 7). Notably, the FA-DCM electrolyte exhibits much lower R_SEI (1.417 Ω), and R_ct (0.544 Ω) compared to the BE electrolyte (9.180 Ω for R_SEI and 12.85 Ω for R_ct, respectively) at 10 C, highlighting its enhanced interfacial reaction kinetics⁴⁸. The galvanostatic intermittent titration technique (Fig. 4d and Supplementary Figs. 27–29) further supports this, with the total overpotential in the BE electrolyte reaching 192 mV, approximately 2.3 times higher than the 83 mV observed in FA-DCM. A similar trend is observed during delithiation: 240 mV for BE vs. 89 mV for FA-DCM. Moreover, in the low-rate GITT measurements, although the FA-DCM electrolyte exhibits a diffusion coefficient comparable to those of the other three electrolytes when tested with low-loading graphite electrodes, it demonstrates a higher diffusion coefficient under the more demanding high-loading graphite conditions (Supplementary Figs. 28 and 29).

Fig. 4. Electrochemical performance of Li||graphite half cells and graphite||NMC811 full cells at high rates.

a The discharge curves of Li||graphite cells of FA-DCM (top) and BE (bottom) electrolytes under different rates. b Cycling performance of Li||graphite cells with different electrolytes at 10 C. c Electrochemical impedance spectra of Li||graphite cells at different rates for FA-DCM (top) and BE electrolyte (bottom). EIS was obtained at 0% SOC. d Discharge curves of the galvanostatic intermittent titration technique (GITT) measurements conducted after the 60th cycle of Li||graphite cells with different electrolytes at 10 C. The corresponding overpotentials obtained from GITT measurements are presented as box-and-whisker plots. The center line represents the median; the box limits indicate the upper and lower quartiles; whiskers extend to 1.5× the interquartile range; data points correspond to the overpotentials extracted from an individual GITT current pulse. n = 5, 8, 13, 13 for BEMP, BE, FA-TTE, FA-DCM electrolytes, respectively, where n denotes the number of constant-current pulses applied during the GITT measurements. e Linear sweep voltammetry of Li||stainless steel cells of different electrolytes. f Rate performance of graphite||NMC811 cells with different electrolytes in the range of 2.8–4.5 V. g Cycling performance of graphite||NMC811 cells with different electrolytes at 6 C. In all panels with dual y-axes, solid symbols correspond to capacity values (left y-axis), while open symbols correspond to Coulombic efficiency (right y-axis). All electrochemical measurements were carried out at 25 ± 1 °C in a temperature-controlled chamber. 1 C = 372 mA g⁻¹ (specific current for graphite), 1 C = 200 mA g⁻¹ (specific current for NMC811). Source data are provided as a Source data file.

Beyond improved graphite performance, the FA-DCM electrolyte also maintains compatibility with high-voltage positive electrodes. Linear sweep voltammetry shows that both FA-DCM and FA-TTE electrolytes maintain stable potential windows exceeding 4.7 V (vs. Li⁺/Li), surpassing that of the BE electrolyte (Fig. 4e). Additionally, the FA-DCM electrolyte exhibits good compatibility with aluminum current collectors (Supplementary Fig. 30). In a full-cell configuration, the small-sized moderator design in FA-DCM significantly enhances electrochemical performance. The graphite||NMC811 full cell with FA-DCM electrolyte delivers a high capacity of 179.8 mAh g⁻¹ at a demanding rate of 6 C—higher than that of FA-DCE (158.3 mAh g⁻¹) and FA-HFC (99.3 mAh g⁻¹) (Supplementary Fig. 31), more than double the capacity of the FA-TTE cell (83.2 mAh g⁻¹), and substantially outperforming cells using BE (66.7 mAh g⁻¹) and BEMP (63.9 mAh g⁻¹) electrolytes (Fig. 4f, g). Comparable performance is observed under high-temperature conditions. (Supplementary Fig. 32). These results validate the effectiveness of the small-moderator-regulated discrete hopping mechanism in enabling fast ion kinetics in liquid-phase transport and interfacial desolvation.

In addition to its fast-charging capability, the FA-DCM electrolyte enables stable low-temperature performance. MD simulations suggest that while correlated Li⁺-anion motions exist, they are less constraining in FA-DCM electrolyte than in FA-TTE electrolyte regardless of temperatures, leading to reduced residence times for FA-DCM electrolyte at 0 °C (193.8 ps for AN, 144.3 ps for FSI⁻) (Supplementary Figs. 33–36 and Supplementary Table 8). Moreover, Li⁺ exhibits a root mean square displacement of 11.67 Å over 3000 ps in the FA-DCM electrolyte, significantly higher than the 2.16 Å observed in the FA-TTE electrolyte at −40 °C (Fig. 5a, Supplementary Fig. 37 and Supplementary Table 8). Additionally, the calculated diffusion coefficient of Li⁺ in the FA-DCM electrolyte reaches 6.6 × 10⁻⁷ cm² s⁻¹ at −40 °C, which is 24.5 times higher than that in the FA-TTE electrolyte, further underscoring the capability of the FA-DCM electrolyte in enabling efficient Li⁺ transport kinetics at low temperatures (Supplementary Fig. 38 and Supplementary Table 8). Electrochemical performances of graphite||NMC811 full cells with different electrolytes at varied temperatures are shown in Fig. 5b. Cells incorporating the BE electrolyte exhibit a fast capacity drop and nearly no capacity at −40 °C, due to electrolyte solidification at −38 °C. In contrast, the cells assembled with FA-DCM electrolyte demonstrate good reversible charge-discharge performance at low temperatures by delivering high capacities of 195.9 mAh g⁻¹ at 0 °C, 185.5 mAh g⁻¹ at −10 °C, 174.5 mAh g⁻¹ at −20 °C, 168.4 mAh g⁻¹ at −30 °C, 153.8 mAh g⁻¹ at −40 °C, 142.1 mAh g⁻¹ at −55 °C, 113.4 mAh g⁻¹ at −65 °C, and 68.3 mAh g⁻¹ even at −75 °C. Such performance is less commonly reported for LIBs, particularly given that many prior studies have primarily focused on discharge behavior under cold conditions^49–51. The corresponding potential profiles are represented in Fig. 5c and Supplementary Fig. 39. Specifically, the graphite||NMC811 full cells with FA-DCM electrolyte retain 66.0% of their capacity at 25 °C when operated at −55 °C. At −40 °C, the full cells exhibit stable and highly reversible cycling performance (excluding minor fluctuations caused by oven temperature variations), delivering a high capacity of 124.7 mAh g⁻¹ with 93.4% retention over 1000 cycles with an average coulomb efficiency (CE) of 99.61% (Fig. 5d). The graphite||NMC811 full cells based on FA-DCE (large D value), FA-HFC (moderate D value), and FA-TTE (small D value) electrolytes reached 113.3, 102.4, and 78.6 mAh g⁻¹, respectively, providing experimental evidence for the universality of adopting the D value as a screening metric. To assess performance under practical conditions, a 4.5 V graphite||NMC811 pouch cell (1.0 Ah, areal capacity of negative/positive electrode: 2.85 mAh cm⁻²/3.13 mAh cm⁻²) using FA-DCM electrolyte (3.0 g Ah⁻¹) was tested. This pouch cell delivers a high capacity of 0.87 Ah with no obvious capacity decay and maintains a high CE of over 99.91% at −40 °C (Fig. 5e). In comparison, the pouch cell with BEMP electrolyte provides only 0.04 Ah and undergoes rapid capacity decay, while the cell with BE electrolyte exhibits virtually no reversible capacity. Even at −50 °C, the FA-DCM-based graphite||NMC811 pouch cell maintains a capacity of 0.51 Ah with an average CE of 99.63% (Fig. 5f and Supplementary Fig. 41). In stark contrast, pouch cells using BE, BEMP, or FA-TTE electrolytes display negligible capacity under the same conditions. Moreover, FA-DCM electrolyte can also effectively stabilize the Si-based negative electrode interface (Supplementary Fig. 42). A control with lithium bis(trifluoromethanesulfonyl)imide replacing LiFSI further confirms the effectiveness of DCM (Supplementary Figs. 43–46). These results mark a significant advancement in enabling the reliable operation of graphite negative electrodes under low temperatures and high-rate conditions, thereby validating the effectiveness of the tailored FA-DCM electrolyte and the underlying small-moderator design strategy for fast Li⁺ chemistries.

Fig. 5. Electrochemical performance of graphite||NMC811 full cells at low temperatures.

a Temperature- and time- dependence of Li⁺ root mean square displacement (RMSD) in FA-DCM electrolyte, color bar represents the magnitude of the Li⁺ RMSD (in Å), with darker colors indicating larger displacements. Specific capacities (b) and the corresponding potential profiles (c) of graphite||NMC811 full cells with FA-DCM electrolyte at 0.1 C under different temperatures. d Cycling performance with different electrolytes in the range 2.8–4.5 V at 0.2 C under −40  °C. Coulombic efficiency plotted on the right y-axis (65–105%) with an axis break between 72 and 85%. e Cycling performance of graphite||NMC811 pouch cells with different electrolyte in the range of 2.8–4.5 V at 0.1 C, −40  °C. Coulombic efficiency plotted on the right y-axis (65–105%) with an axis break between 70 and 80%. f The charge/discharge curves of graphite||NMC811 pouch cell with FA-DCM electrolyte at 0.1 C, −50  °C. In all panels with dual y-axes, solid symbols correspond to capacity values (left y-axis), while open symbols correspond to Coulombic efficiency (right y-axis). 1 C = 200 mA g⁻¹ (specific current for NMC811). Source data are provided as a Source data file.

Mechanistic Insights into Interphase Formation with FA-DCM Electrolyte

The interphase formation process at the graphite negative electrode surface was investigated using ab initio molecular dynamics (AIMD). As shown in Fig. 6a (Supplementary Data 3), regions exhibiting more intense blue coloration correspond to higher electron localization function values, signifying enhanced electron exchange between the electrolyte components and the graphite substrate. These localized electrons facilitate the reduction of electrolyte constituents, particularly the FSI⁻ anions and DCM molecules in the FA-DCM electrolyte, which are rapidly reduced to form a predominantly inorganic SEI (Fig. 6a, Supplementary Figs. 47 and 48 and Supplementary Data 3). In contrast, the SEI formed by BE electrolyte is dominated by organic-rich species, resulting in high interfacial resistance and the elevated Li⁺ desolvation energy^4,5. To further investigate the dynamic and time-dependent process, we employed in-situ electrochemical quartz crystal microbalance with dissipation monitoring (EQCM-D) coupled with cyclic voltammetry The simultaneous frequency and dissipation responses reveal that, for BE, the mass of the quartz sensor increases sharply within the first 6,000 s and continues to rise over 20,000 s (Fig. 6b and Supplementary Fig. 49), indicating prolonged accumulation of interphase products. In contrast, cells with FA-DCM electrolyte exhibit a significantly smaller and more stable mass change, suggesting the formation of a thinner, more compact, and mechanically robust SEI. These observations are further supported by transmission electron microscopy (TEM). After 50 cycles, a uniform and thin SEI layer (~3 nm) is observed on the graphite negative electrode, with a similarly thin (~2 nm) interphase on the NMC811 positive electrode (Fig. 6c, Supplementary Figs. 50 and 51), affirming the formation of a highly stable interphase capable of supporting fast Li⁺ transport kinetics⁵². In sharp contrast, the BE forms substantially thicker interphases on both electrode surfaces, triggering the positive electrode structural degradation and severe side reactions (Fig. 6d, Supplementary Figs. 52 and 53 and Table 9).

Fig. 6. Electrolyte/Electrode interphase analysis.

a The initial decomposition reaction of FA-DCM electrolyte (left) and BE electrolyte (right) on the graphite interphase calculated by ab initio molecular dynamics (AIMD) (lime green, Cl atom; red, O atom; yellow, S atom; purple, N atom; silver, F atom; tan, C atom; white, H atom). The color bar represents electron localization function (ELF, 0.0–1.0), where regions with more intense blue coloration indicate higher ELF values, reflecting enhanced electron localization and stronger electron exchange between the electrolyte components and the graphite substrate. b Electrochemical quartz crystal microbalance with dissipation monitoring (EQCM-D) profiles of mass change (Δ_Mass) versus time. (Graphite-sputtered Cu chip resonator is used as the working electrode (WE), and Li metal is used as the counter electrode (CE) and reference electrode (RE). Cells are scanned at a potential window from the open circuit voltage to 0.01 V and then back to 1.50 V (vs. Li⁺/Li) at a scan rate of 5 mV s⁻¹, at 25 °C for 3 cycles. Frequency is taken at the 3rd overtone (n = 3). Transmission electron microscopy (TEM) images of graphite negative electrodes using FA-DCM (c) and BE electrolyte (d). Scale bars, 10 nm. X-ray photoelectron spectroscopy (XPS) results of the SEI on cycled graphite negative electrodes using FA-DCM electrolyte (e) and BE electrolyte (f). 3D-rendering images of F⁻ (g), S⁻ (h), Cl⁻ (i) and SNO⁻ (j) species on the graphite negative electrode using FA-DCM electrolytes and F⁻ (k), PO₂⁻ (l), OH⁻ (m) and C₂HO⁻ (n) species on the graphite negative electrode using BE electrolytes, the color bar represents the range of signal intensity corresponding to the pixel colors in the image. The analysis area was 100 × 100 μm² and 820 s etching time. The graphite negative electrodes used for TEM, XPS, and TOF-SIMS tests were disassembled from graphite||NMC811 full cells cycled 100 cycles at 0.5 C, 25 °C, and rinsed with anhydrous DMC three times to remove the residuals (inside the glovebox). 1 C = 200 mA g⁻¹ (specific current for NMC811). Source data are provided as a Source data file.

Time of flight secondary ion mass spectrometry (TOF-SIMS) and in-depth X-ray photoelectron spectroscopy (XPS) analysis were conducted to comprehensively explore the chemical composition of electrode-electrolyte interphase (Fig. 6e–n, Supplementary Figs. 54–58 and Supplementary Tables 10 and 11). In the C 1s spectra (Supplementary Fig. 54a), the FA-DCM electrolyte initially forms carbonyl-containing organic compounds (C = O), which diminish with sputtering depth, whereas the LiF signal (Fig. 6f) and N- and S-containing derivatives (Supplementary Fig. 54d, e) increase, indicating preferential decomposition of FSI⁻ anions at graphite interphase. In addition, Cl 2p spectra further reveal the incorporation of LiCl species (Supplementary Fig. 54f), implying the formation of a dual-layer SEI architecture consisting of an inner LiF-LiCl-rich inorganic layer and an outer organic matrix^53–55. 3D-rendering images obtained from TOF-SIMS further confirm the formation of this bilayer structure (Fig. 6g–j and Supplementary Fig. 57a), which facilitates efficient charge-transfer kinetics and mechanical robustness. By contrast, the BE electrolyte forms an SEI dominated by alkyl lithium carbonate species and PF₆⁻ decomposition derivatives (Fig. 6k–n and Supplementary Figs. 55 and 57b), resulting in a loosely packed, porous interface with high impedance. Such a structure, also observed on the NMC811 surface (Supplementary Fig. 58), fails to effectively mediate the Li⁺ intercalation/deintercalation, accounting for the sluggish interfacial kinetics observed in graphite-based LIBs employing conventional carbonate-based electrolytes.

It is worth noting that the toxicity and volatility of DCM limit its broader applicability, but the sealed cells using DCM-based electrolyte still demonstrated decent stability and capacity retention under extended cycling or storage, as the swelling of the pouch cells after cycling at ambient/low temperature is limited and even after being stored at 45 °C for 7 days, the pouch cell retained 90.7% of its initial capacity (Supplementary Figs. 59–61). More importantly, as a moderator with a small molecular size and a low dipole moment, DCM successfully tunes the solvation structure, creating a more uniformly distributed and compact solvation cluster, which is essential for low-temperature operation and fast charging. This success validates the feasibility of the concept and paves the way for future electrolytes designed on these principles to be both high-performing and environmentally benign.

In this work, we establish a universal screening framework through introducing a key parameter (D), defined by two fundamental molecular descriptors: the ion-dipole/dipole-dipole interactions (μ_moderator) and the moderator molecular volume (V_moderator). This descriptor-based metric quantitatively captures the cluster-disintegration capacity of a given moderator, with D inversely proportional to both μ_moderator and V_moderator. An optimal balance is achieved with DCM (85.11 Å, 2.36 D), which effectively disrupts the over-coordinated Li⁺ networks by weakening Li⁺-anion interactions, thereby reducing solvation cluster size and promoting the formation of compact, highly conductive ion domains. This moderation strategy enhances the number of effective free charge carriers and facilitates rapid Li⁺ transport via efficient ligand exchange across a uniformly distributed solvation sheath. Leveraging this principle, the rationally designed FA-DCM electrolyte enables scaled-up 1.0 Ah graphite (3.13 mAh cm⁻²)||NMC811 (2.85 mAh cm⁻²) pouch cells to deliver high reversible capacity of 0.87 Ah and 0.51 Ah at −40 and −50 °C, respectively, with negligible capacity fade over prolonged cycling. This work provides mechanistic insights into electrolyte design and establishes a generalizable strategy for tailoring solvation structures, thereby advancing the practical deployment of high-performance LIBs under high-rate and low temperature conditions.

Methods

Materials

Li metal electrodes (99.9%, thickness: 450 μm, diameter: 15.6 mm) were supplied by Tianjin China Energy Lithium Co., Ltd. LiNi_0.8Mn_0.1Co_0.1 (NMC811, 99.5%) powder and conductive carbon (Super C45, 99.5%) were purchased from Hefei Keijing Co., Ltd. N-Methyl pyrrolidone (NMP, 99.5%), polyvinylidene fluoride (PVDF, molecular weight: ~50 W, 99.5%) was obtained from Duoduo Co., Ltd and natural graphite (99.5%, average diameters: 2.5 μm) was purchased from Beijing Xinjincheng Science and Trade Co., Ltd. Lithium bis(fluorosulfonyl)azanide (LiFSI, 99.96%), lithium hexafluorophosphate (LiPF₆, 99.97%), lithium bis(trifluoromethanesulphonyl)imide (LiTFSI, 99.96%), ethylene carbonate (EC, 99.993%), ethyl methyl carbonate (EMC, 99.98%), dimethyl carbonate (DMC, 99.995%) were purchased from Changde Dadu New Material Co., Ltd. Acetonitrile (AN, 99.9%), dichloroethane (DCE, 99.8%), methyl propionate (MP, 99%) and PAALi (solid content, 6%) were purchased from Sigma-Aldrich. 1,1,2,2-tetrafluoroethyl-2,2,3,3-tetrafluoropropyl ether (TTE, 98%), dichloromethane (DCM, 99.9%), and 1,1,2,2,3,3,4-heptafluorocyclopentane (HFC, 98%) were purchased from Macklin Inc.

Preparation of electrolytes and electrodes

All electrolyte systems were formulated by dissolving lithium salts (LiPF₆, LiFSI, or LiTFSI) into designated solvent mixtures. Specifically, the BE, BEMP, FA-TTE, FA-DCM, FA-DCE, FA-HFC electrolyte and TA-DCM electrolytes were 1.0 M LiPF₆/EC-DMC (1:1, vol), 1.0 M LiPF₆/EC-EMC-MP (1:3:1, vol), 1.5 M LiFSI/AN-TTE (1:4, vol), 1.5 M LiFSI/AN-DCM (1:4, vol), 1.5 M LiFSI/AN-DCE (1:4, vol), 1.5 M LiFSI/AN-HFC (1:4, vol) and 1.5 M LiTFSI/AN-DCM (1:4, vol), respectively. Prior to electrolyte preparation, all solvents were dehydrated by storage over activated 4 Å molecular sieves (Sigma-Aldrich) for 24 h. Electrolyte formulation was conducted inside an argon-filled glovebox (Mikrouna, China), where both moisture and oxygen concentrations were rigorously maintained below 0.01 ppm.

Negative electrodes were fabricated by casting a water-based slurry containing natural graphite (80 wt%), Super C45 (10 wt%), and PAALi (10 wt%) onto copper foil. The cathode slurry was prepared by dispersing 80 wt% NMC811, 10 wt% Super C45, and 10 wt% PVDF in NMP, followed by coating onto aluminum foil. Both slurries were homogenized using a planetary mixer (AR-100, THINKY Co.) through three consecutive mixing cycles of 8 min each. Coating was performed on one side of the current collectors using a coating machine (MSK-AFA-HDS-150, Hefei Kejing), followed by vacuum drying at 80 °C overnight. The electrodes were punched into discs with diameters of 14 mm and 12 mm for negative electrodes and positive electrodes, respectively, followed by an additional vacuum-drying step at 80 °C for 12 h. The areal mass loadings of graphite and NMC811 were approximately 2.13 and 3.29 mg cm⁻², respectively.

CR2032-type coin cells (Guangdong Canrd New Energy Technology Co.) were assembled using 316 stainless-steel cases and springs (15.4 × 1.1 mm). Celgard 2325 separators (Thickness: 25 μm; Lateral dimension: 100 mm; Porosity: 40%; Average pore size: 0.028 μm) were punched into 19 mm discs prior to use. Each coin cell was filled with 100 μL of the as-prepared electrolyte and assembled with either NMC811 or graphite as the working electrode. For full-cell configurations, the negative-to-positive (N/P) capacity ratio was fixed at 1.2, calculated based on theoretical specific capacities of 372 mAh g⁻¹ for graphite and 200 mAh g⁻¹ for NMC811.

Double-sided-coated pouch cells (1.0 Ah) employing graphite (3.13 mAh cm⁻²) and NMC811 (2.85 mAh cm⁻²) electrodes were supplied by Li-FUN Technology Co., Ltd., and filled with electrolyte at a dosage of 3.0 g Ah⁻¹. The assembled pouch cells underwent two formation cycles at 0.1 C, followed by degassing prior to long-term electrochemical evaluation. A constant stack pressure of approximately 100 kPa was applied during testing.

Electrolyte viscosity was determined using a rotational rheometer (Anton Paar MCR102e). For thermal analysis, electrolyte samples were hermetically sealed in stainless-steel pans inside an argon atmosphere glovebox and subjected to differential scanning calorimetry measurements (NEXTA DSC600, Hitachi, Ltd.). All the experiments were conducted at a uniform heating and cooling rate of 5 °C min⁻¹.

Electrochemical measurements

All electrochemical tests were carried out with charging and discharging at the same temperature over a range of −75 to 25 °C. This protocol differs from earlier approaches that employed charging at 25 °C followed by low-temperature discharge. Battery cycling was carried out using a LAND battery testing system (Wuhan LAND Electronics Co., Ltd.). For coin-type graphite‖NMC811 cells evaluated at low temperatures, an initial formation process was performed by cycling at 0.5 C for ten cycles at 25 ± 1 °C in a temperature-controlled chamber. Subsequently, long-term cycling tests were conducted at -40 °C under a current rate of 0.2 C. In the case of pouch-type graphite‖NMC811 cells, formation cycling was carried out at 0.2 C for the first ten cycles at 25 ± 1 °C, followed by prolonged cycling at 0.1 C under subzero conditions (−40 °C and −50 °C).

In galvanostatic intermittent titration technique measurements, after 60 cycles at 10 C/1 C charge and discharge, the Li||graphite cells were cycled at a rate of 10 C/1 C and 3 h rest time. The ionic conductivities of the electrolytes were determined by alternating current (AC) impedance spectroscopy. Two platinum black plate electrodes (12 × 120 mm) were immersed in 6 mL of electrolyte, and impedance measurements were carried out using an electrochemical workstation (Ivium-n-Stat, Ivium Technologies BV Co., Ltd.) with an AC perturbation amplitude of 10 mV over a frequency range of 1 Hz to 1 MHz. Measurements were performed at various temperatures and electrolyte concentrations.

Electrochemical impedance spectroscopy of coin cells in the fully discharged state was conducted using the same workstation over a frequency window from 10 kHz to 10 mHz with an excitation amplitude of 10 mV. Each spectrum was recorded after a 30 s open-circuit stabilization period, and 12 data points were collected per decade of frequency. The lithium-ion transference number ($t_{{\text{Li}}^{+}}$) was determined by using direct current potentiostatic polarization (ΔV = 10 mV) and alternating current (AC) impedance spectra with a Li||Li symmetrical cell. Chronoamperometry was used to obtain the initial current (I₀) and steady-state current (I_ss), while the interfacial resistances before (R₀) and after (R_ss) polarization were extracted from impedance spectra acquired prior to and following polarization. And $t_{{\mathrm{Li}}^{+}}$ can be calculated by the following equation⁵⁶:

$$t_{{\text{Li}}^{+}}=\frac{I_{\text{ss}}\left(\mathrm{\Delta }tV-{tI}_0{tR}_0\right)}{I_0\left(\mathrm{\Delta }tV-{tI}_{\text{ss}}{tR}_{\text{ss}}\right)}$$ 2

where I_ss is the steady-state current, I₀ the initial current, ∆V the applied potential, and R_ss and R₀ the electrode resistances after and before the polarization, respectively.

Characterization

The solvation structures of different electrolytes were measured by ⁷Li coaxial NMR spectroscopy (Bruker 600 MHz), Raman spectrum (Horiba LabRAM HR Evolution), and electrical in-situ FTIR spectroscopy (Thermo IS50). The self-diffusion coefficients of different electrolytes were measured by ¹H, ⁷Li, and ¹⁹F DOSY NMR (Bruker 600 MHz). The positive electrodes and negative electrodes were disassembled from graphite||NMC811 full cells cycled 100 cycles and rinsed with anhydrous DMC three times to remove the residuals (inside the glovebox), followed by characterizations of scanning electron microscopy (SU-70, Hitachi Ltd.), TEM (Tecnai G2 F20), XPS (Thermo Scientific ESCALAB 250Xi) with a sputtering rate of 0.02 nm s⁻¹ and TOF-SIMS (ToF.SIMS 5-100 instrument IONTOF GmbH). Samples were transferred directly to the XPS sample chamber to avoid interference from the external rings. X-ray diffraction (X-pert Powder, PANalytical B.V.) experiment was used with Cu Kα radiation from 5° to 60° at a scanning speed of 5° min⁻¹. For the inductive coupled plasma optical emission spectrometer characterizations, the fully-charged-state NMC811 disassembled from graphite||NMC811 cells was soaked in the electrolyte (10 mL) for 7 days, and then removed into the Agilent 720ES instrument to quantify the content of Ni, Co, and Mn.

Computational methods

Gaussian 16 software was implemented to optimize molecular geometries of the ground state with the m062x/def2TZVP method and the GD3 dispersion model⁵⁷. The dipole moment and volume of molecules were calculated by the Multiwfn program⁵⁸. Independent gradient model based on Hirshfeld partition (IGMH)⁵⁹ was used for the visual analysis of interactions between molecules.

Molecular dynamics (MD) simulations were performed in LAMMPS using the all-atom optimized potentials for liquid simulations (OPLS-AA) force field. The force field of solvents was obtained from the LigParGen web server⁶⁰. The electrolyte systems were set up initially with the salt and solvent molecules distributed in the simulation boxes using Moltemplate (http://www.moltemplate.org/). For each system, an initial energy minimization at 0 K was performed to obtain the ground-state structure. After that, the models were simulated in NPT ensemble (350 K, 1 bar) for 0.5 ns; Then simulations were performed from 350 to 300 K for 0.5 ns; Subsequently, the systems were simulated at 300 K in NPT ensemble for 2 ns; Finally, 300 K simulations for 5 ns in NVT ensemble (constant volume) were conducted. To investigate the effects of electric field and temperature, simulations were conducted under various electric field strengths and temperatures, while all other simulation parameters remained consistent with the procedures described above. Coordination structure of electrolyte, evolution of Li-ion solvation environment, and root mean square displacement of atoms were analyzed using the Visual Molecular Dynamics software. Lifetime/displacement functions were obtained by Trajectory Analyzer and Visualizer^61,62.

AIMD were carried out using the Vienna Ab initio Simulation Package^63–65. The projector augmented wave method was used, accompanied by the Perdew–Burke–Ernzerhof exchange-correlation functional in the Generalized Gradient Approximation. To simulate the reaction between electrolyte and graphite, the graphite and electrolyte models were placed in the same box. The graphite was composed of three layers of fully discharged graphite (LiC₆) with armchair edges, which were entirely functionalized with oxygen atoms. Electrolyte models were constructed using PACKMOL and simulated in the NVT ensemble using a Nosé-Hoover thermostat. Each model was simulated for 5 ps with a time step of 1 fs at 300 K. The electron localization function (ELF) was simulated to study the electronic structure of the graphite–electrolyte interface.

Author contributions

M.L., D.L., J.W., and X.F. conceived the idea and designed the experiments. M.L. and D.L. conducted the electrochemical experiments and the material characterizations, with the assistance of S.Z., L. Lv, B.M., H.Z., L.Li., J.Q., L.C., and X.F. S.Y. and Y.W. performed DSC under the guidance of L.F. Z.L. performed the viscosity test. J.W. provided the theoretical calculations. M.L., D.L., J.W., S.Z., R.L., T.D., and X.F. prepared the manuscript, with input from all the co-authors. X.F. supervised all the studies.

Peer review

Peer review information

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

Data availability

All data that support the findings of this study are presented in the manuscript and Supplementary Information, or are available from the corresponding author upon request. Source data are provided with this paper. The atomic coordinates of the optimized geometries, along with related information, can be found in the Supplementary Data. 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/s41467-026-70570-5.