Revisiting Fluorobenzene as Diluents in Ether-Based Electrolytes for Lithium Metal Batteries

Abstract

The localized high-concentration electrolyte based on the low-cost, low-density, low-viscosity, and low-fluorine-substitution fluorobenzene diluents and 1,2-dimethoxyethane solvents has been successfully demonstrated in high-performance lithium metal batteries. However, it requires high salt-to-solvent molar ratio, which causes high production costs and great environmental burden. Decreasing the salt-to-solvent molar ratio without sacrificing its electrochemical performance remains a challenge. Herein, we reveal that as the salt-to-solvent molar ratio is decreased, the compatibility of the fluorobenzene-diluted 1,2-dimethoxyethane-based electrolyte with lithium metals transitions from compatible to incompatible. We elucidate the degradation mechanism of the fluorobenzene-diluted 1,2-dimethoxyethane-based electrolyte undergoing severe side reactions with lithium metals. Inspired by these findings, we develop a fluorobenzene-diluted dimethyl acetal-based electrolyte, which enhances the stability of the electrolyte under a reduced lithium-salt concentration, making it show good compatibility with lithium metal (Coulombic efficiency: 99.43% at 25 °C, 97.74% at −40 °C). Moreover, the assembled Li | |SPAN battery displays a high capacity retention of 83% after cycling 500 cycles and can operate at −60 °C. Besides, a high specific energy of 334.53 Wh kg⁻¹ (excluding package) can be achieved for the Li | |SPAN pouch cell. This work prompts us to re-examine the applicability of fluorobenzene as diluents in ether-based electrolytes for lithium metal batteries.

Decreasing salt-to-solvent molar ratio without sacrificing electrochemical performance remains a challenge in high-performance lithium metal batteries. Here, authors show that compatibility of fluorobenzene-diluted 1,2-dimethoxyethane-based electrolyte with Li metal is reduced in smaller salt-to-solvent molar ratios and design a fluorobenzene-diluted dimethyl acetal-based electrolyte with significantly enhanced stability towards Li metal under a reduced Li salt concentration.

Introduction

High-performance battery energy storage systems are essential for supporting renewable but intermittent energy sources and advancing sustainable transportation^1–3. Lithium metal batteries (LMBs) have been regarded as the “holy grail” for next-generation energy storage systems due to the extremely high theoretical specific capacity (3860 mAh g⁻¹) and the lowest standard redox potential (−3.04 V vs. the standard hydrogen electrode) of lithium metal negative electrodes^4–6. Unfortunately, the poor reversibility and inferior safety originating from the dendritic Li growth intensely hinder the practical applications of LMBs^7,8. Recently, worldwide efforts have been revived to improve the stability of the lithium metal negative electrodes and enable the practical application of LMBs, including structure modification^9,10 and artificial solid electrolyte interphase (SEI) coating^11,12, as well as electrolyte engineering^13,14.

Electrolyte engineering is a straightforward yet effective strategy for achieving high-performance and practical LMBs^15,16. Representatively, localized high-concentration electrolytes (LHCEs) have made significant advancements in recent years. The high salt-to-solvent molar ratio (SSR) in LHCEs promotes the formation of a large number of contact ion pairs (CIPs) and aggregates (AGGs) in the solvation sheath. This, in turn, generates a robust, inorganic-rich SEI that greatly improves the stability of the Li metal interphase. The commonly used diluents in LHCEs are highly fluorinated ethers or esters (e.g., 1,1,2,2-tetrafluoroethyl-2,2,3,3-tetrafluoropropyl-ether (TTE)¹⁷ and 2,2,2-trifluoroethyl methyl carbonate (FEMC)¹⁸), which usually suffer from high density (≥1.30 g cm⁻³), high cost, and high F content. As a typical fluorinated hydrocarbon, the “4 L” (low-cost, low-density, low-viscosity, and low-fluorine-substitution) fluorobenzene (FB) has been successfully applied in the ether-based LHCEs, and these LHCEs demonstrate good Li metal compatibility^19,20. The development of LHCEs using FB as a diluent has advanced the progression of LMB systems.

However, it is important to note that Li salts, such as LiPF₆ and LiFSI, have consistently been the most expensive component of electrolytes (in commercial ester-based electrolytes, Li salts account for over 40% of the total cost)²¹. Even in the “4 L” FB-based LHCEs, a significant amount of F-containing Li salts is still required. For example, in a 4 M LiFSI in 1,2-dimethoxyethane (DME)/FB (1:1 v/v) LHCE (where “M” denotes molar salt in liter solvent+diluent), LiFSI constitutes over 90% of the cost and 44.2 wt% of the mass. Moreover, the F content of the electrolyte is 252 mg mL⁻¹ and the electrolyte density is as high as 1.32 g cm⁻³. The high F-containing Li salt content significantly increases both the cost of the electrolyte and the environmental burden^22,23. Consequently, decreasing the usage amount of F-containing Li salts without sacrificing the electrochemical performance of LHCEs is significant, yet it remains a huge challenge²⁴.

In this work, we found that even when using a DME-based electrolyte system known for its good compatibility with lithium metal negative electrodes, reducing the Li salt (LiFSI) content in the FB-diluted DME-based LHCE (4 M LiFSI in DME/FB (1:1 v/v)) led to severe degradation reactions of electrolytes with lithium metal negative electrodes. This finding challenges the conventional understanding that FB-diluted DME electrolyte systems are inherently compatible with lithium metal negative electrodes, hindering the development of advanced DME-based localized medium- to low-concentration ether electrolytes. Based on this unexpected phenomenon, we revealed the mechanism by which FB, under the mediation of free DME molecules, undergoes continuous severe side reactions with lithium metal negative electrodes, ultimately leading to electrolyte failure. Therefore, based on the thorough understanding of degradation mechanisms between the FB-diluted DME-based LHCE and lithium metal negative electrode, we designed a FB-diluted dimethyl acetal (DEA)-based localized medium-concentration electrolyte (LMCH: 1 M LiFSI in DEA/FB (1:1 v/v)), which significantly enhances the stability of the electrolyte system under a reduced Li salt concentration. The DEA solvent with a unique “branched” molecular structure reduces the complexation ability of phenyl lithium, inhibits the deprotonation of FB, and suppresses the formation of intermediate products such as biphenyl lithium. This, in turn, significantly enhances the compatibility between the FB-diluted DEA-based LMCH and lithium metal negative electrodes. Moreover, the DEA molecules are preferentially adsorbed on the surface of lithium metal negative electrodes and then block the substitution reaction between FB and Li metal, thereby suppressing the continuous Li corrosion and solvent/diluent decomposition. Meanwhile, benefiting from the large molecular steric hindrance brought by its unique “branched” molecular structure of DEA, more CIPs and AGGs are formed in the electrolytes even at a low Li salt concentration, further improving the stability of lithium metal negative electrodes. As a result, the lithium metal negative electrode displayed high CEs of 99.43% at 25 °C and 97.74% at −40 °C in our LMCH. The assembled Li | |SPAN battery based on our FB-diluted DEA-based LMCH exhibited a high capacity retention of 83% after cycling 500 cycles and can operate at −60 °C. Furthermore, a high specific energy of 334.53 Wh kg⁻¹ can be achieved for the Li | |SPAN pouch cell.

Results and discussion

Design and phenomena

Although LHCEs have demonstrated good interfacial stability in LMBs, the extensive use of F-containing Li salts leads to high costs, high density, and high F content, significantly hindering their large-scale practical applications (Fig. 1a, Supplementary Table 1). By reducing the amount of F-containing Li salts, the development of high-performance LMCHs can effectively lower electrolyte costs, reduce electrolyte density, and mitigate the environmental burden associated with high F content. Taking LiFSI-based localized high-concentration ether electrolytes as an example, it can be observed that LiFSI (69.99 $ kg⁻¹) is the most expensive component of the electrolyte (Fig. 1b, Supplementary Tables 2 and 3). In addition, FB (cost: 6.26 $ kg⁻¹) not only offers a lower cost compared to commonly used TTE diluents (23.68 $ kg⁻¹; density: 1.53 g cm⁻³; F content: 1000 mg mL⁻¹), but also has a lower density (1.02 g cm⁻³) and F content (202 mg mL⁻¹). Therefore, constructing FB-based LMCHs can significantly reduce the cost of conventional high-concentration electrolytes (HCEs) and LHCEs while alleviating the environmental burden. As shown in Fig. 1c, our developed electrolyte (1 M LiFSI in DEA/FB (1:1 v/v)) demonstrates advantages in terms of cost (15.27 $ kg⁻¹), density (1.08 g cm⁻³), and F content (139 mg mL⁻¹) compared to conventional HCEs and LHCEs.

Fig. 1. Electrolyte Design and Phenomena.

a Demonstration of the key merits of LMCE over conventional LHCE for practical applications. b Comparison of LiFSI, common solvents, and diluents in terms of price, density, and F content. c Comparison of advanced electrolytes in terms of price, F content, and density. d Schematic illustration of the solvation structure of the LHCE and LMCE and the phenomenon after their exposure to Li metals. Insets are the photos showing the evolution of electrolytes.

The motivation of this work is to reduce the use of F-containing Li salts and develop high-performance ether-based LMCEs in LMBs. However, when we directly decreased the SSR from 1:1.2 (4 M LiFSI in DME/FB (1:1 v/v)) to 1:4.8 (1 M LiFSI in DME/FB (1:1 v/v)), the electrolyte, which was originally stable with Li negative electrode, became unstable with Li negative electrode. After just 3 hours at room temperature (RT, 25 °C), Li metal in the 1 M LiFSI in DME/FB (1:1 v/v) electrolyte completely decomposes, and the electrolyte turned dark brown, indicating severe side reactions between Li metal and the electrolyte (Fig. 1d). This phenomenon challenges the conventional understanding that FB-diluted DME-based electrolytes are intrinsically stable with Li negative electrode.

Mechanism and redesign

Based on the aforementioned anomalous phenomenon, we conducted an in-depth investigation into the reaction mechanisms of the FB-diluted DME-based electrolyte with Li metal. Upon contact with FB, the surface of Li metal gradually turns black while the FB solvent remains unchanged in color (Supplementary Fig. 1). The scanning electron microscope (SEM) image shows that Li metal exposed to FB exhibits a highly uneven morphology, with its surface covered by C and F elements. (Supplementary Fig. 2). X-ray photoelectron spectroscopy (XPS) was employed to further analyze the surface composition of the Li metal. In the C 1 s spectrum, peaks at 284.8 eV and 288.4 eV correspond to C–C/C–H and C–F bonds, respectively, while in the F 1 s spectrum, peaks at 684.8 eV and 687.2 eV are attributed to LiF and C–F bonds (Supplementary Fig. 3). These results suggest that FB undergoes a substitution reaction with Li metal, forming solid phenyl-lithium and LiF (Supplementary Fig. 4)¹⁹. Phenyl-lithium, in turn, readily decomposes at RT, leading to the formation of various carbon-containing species^25,26. Furthermore, nuclear magnetic resonance (NMR) was employed to obtain the ¹³C-spectra of FB solvents both before and after immersion of Li metal. The results indicate the FB is well-preserved, with no detectable by-products generated in the FB (Supplementary Fig. 5).

Upon adding Li metal to the DME/FB (1:1 v/v) mixed solvent, it can be observed that the solvent gradually turns yellow after 3 hours, and the Li metal decomposes, forming precipitates (Fig. 2a). It is attributed to the substitution reaction between Li metal and FB, resulting in the formation of phenyl-lithium (path (1) in Fig. 2l), which can be stably stored in DME as a dimeric aggregate^27,28 (Fig. 2b). The phenyl-lithium acts as a base to deprotonate the fluorobenzene ortho to the F atom to generate ortho-lithiated fluorobenzene under the strong dissociation of DME (paths (2) and (3) in Fig. 2l). Following the elimination of LiF to generate benzyne from ortho-lithiated fluorobenzene (paths (4) and (5) in Fig. 2l), the benzyne intermediate is subsequently trapped by phenyl-lithium through nucleophilic addition, yielding lithiated biphenyl (path (6) in Fig. 2l)^23,29,30. Finally, protonation of the lithiated biphenyl intermediate with DME results in the formation of biphenyl (BP) (path (8) in Fig. 2l), while simultaneously breaking the ether-oxygen bond in DME to yield vinyl methyl ether (VME) and lithium methoxide (LM) (Fig. 2c, d and paths (7) and (9) in Fig. 2l). The chemical degradation processes of FB in reaction with phenyl-lithium were also confirmed by Robert using isotopic labeling (Supplementary Fig. 6)³¹. In addition, benzyne itself can also undergo cycloaddition reactions or capture other intermediates to form a series of benzene derivatives (e.g. Triphenylene) (Supplementary Fig. 7)^32,33. As a result, FB undergoes continuous and severe side reactions with Li metals under the mediation of free DME molecules. Moreover, the DME molecules continuously decompose, ultimately leading to electrolyte failure.

Fig. 2. Reaction Mechanisms and Solvent Design.

a Digital photographs showing the evolution of Li metal and mixed solvent when Li metal added to the DME/FB (1:1 v/v) mixed solvent. b Schematic of dimer formation by dissolution of phenyl-lithium in DME. c Reaction paths of FB and phenyl-lithium in DME. d Cleavage reaction of DME in the presence of lithiated biphenyl intermediate. Mass spectra of (e) vinyl methyl ether and (f) biphenyl in the DME/FB (1:1 v/v) mixed solvent after the reaction. g Schematic showing the structure of DME and DEA solvents. h Digital photographs showing the evolution of Li metal and mixed solvent when Li metal is added to the DEA/FB (1:1 v/v) mixed solvent. i NMR ¹³C spectra of DEA/FB (1:1 v/v) mixed solvents before and after the addition of Li metal. j DFT calculations of the dissociation energy of phenyl-lithium (PHLi) in DME and DEA. k DFT calculation of the cleavage energy of DME and DEA. l Schematic showing the reaction paths of DME/FB (1:1 v/v) and DEA/FB (1:1 v/v) mixed solvents with Li metal.

Mass spectrometry (MS) was employed to analyze the products formed in the DME/FB (1:1 v/v) mixed solvents after a 3-hour reaction with Li metal, in order to verify the aforementioned reaction mechanism. The m/z signals at 58.1 and 154.2 confirm the formation of VME and BP, respectively (Fig. 2e, f). Additionally, signals corresponding to other benzene derivative products, such as tribiphenyl, were also detected (Supplementary Fig. 8). These results well verify the above-mentioned reaction mechanisms of the FB with Li metal under the mediation of DME. When Li metal was immersed in the 4 M LiFSI in DME/FB (1:1 v/v) LHCE for 24 hours, no sign of decomposition of the electrolyte was observed (Supplementary Fig. 9). This phenomenon is primarily attributed to the abundant AGGs and CIPs in the electrolyte, which are preferentially reduced to form a robust LiF-rich passivating layer on the surface of Li metal to inhibit the further side reactions (Supplementary Fig. 10). Furthermore, most DME molecules coordinate with Li⁺ in the 4 M LiFSI in DME/FB (1:1 v/v) LHCE. No free DME molecules are available to form dimeric aggregates through complexation that stabilizes phenyl-lithium in the electrolyte, thereby hindering the subsequent series of decomposition reactions of FB and DME. (Supplementary Fig. 11).

Based on the above findings, we find that the DME solvent suffers from significant drawbacks in constructing FB-diluted local medium/low concentration electrolytes. And we recognize that the mediation of the solvent, particularly the complexation and dissociation of phenyl-lithium, as well as the stability of the solvent molecules themselves, is the key factor in triggering the continuous parastics reaction between FB diluent and Li metal. To better design a local medium/low-concentration electrolyte compatible with the low-cost, low-density, low-viscosity, and low-fluorine-substitution FB diluent, by shortening the carbon chain in the central segment of ethylene glycol in DME and introducing an electron-donating methyl group, we screened a solvent molecule, DEA, with a “branched” structure (Fig. 2g). The unique “branched” molecular structure of DEA decreases its complexation ability with phenyl-lithium, thereby hindering the dissociation of phenyl-lithium and the deprotonation of FB, while simultaneously enhancing the stability of the ether-oxygen bond. After immersing Li metal in the DEA/FB (1:1 v/v) mixed solvents for 24 hours, no visible changes were observed in either the electrolyte or the Li metal (Fig. 2h). NMR ¹³C spectra indicated no new by-products were detected in the DEA/FB (1:1 v/v) mixed solvents, suggesting good stability of Li metal with FB in the DEA solvent (Fig. 2i).

Density functional theory (DFT) calculations revealed that phenyl-lithium has higher dissociation energy in DEA (20.89 kcal mol⁻¹) compared to DME (20.38 kcal mol⁻¹) (Fig. 2j and path (11) in Fig. 2l). Furthermore, the Gibbs Free Energy (ΔG) required to break the ether-oxygen bond of DME and DEA to produce the intermediates of VME and methanol (MA) increases from 0.45 eV in DME to 0.62 eV in DEA (Fig. 2k, path (12) in Fig. 2l, and Supplementary Fig. 12). Molecular dynamics (MD) simulations were further conducted to gain deeper insight into how the “branched” structure of DEA molecules contributes to the suppression of side reactions (Supplementary Fig. 13). Phenyl-lithium as a Li salt model was simulated in DEA and DME environments (Solvent/Phenyl-lithium molar ratio=10:1), respectively. The radial distribution function (RDF) and average coordination number (CN) of Li⁺ in different solvents reveal that the proportion of DME molecules (CN = 1.38) occupying the first solvation shell of Li⁺ is significantly higher than that of the DEA system (CN = 0.57), indicating that phenyl-lithium dissociates more fully in DME solvent than in DEA solvent. The Potential of Mean Force (PMF) results indicate stronger interaction energies between Li⁺ and solvent oxygen atoms in DME, suggesting that phenyl-lithium more readily dissociates in DME. In contrast, the shallower PMF profile observed in the DEA system reflects a weaker driving force for dissociation, implying that phenyl-lithium remains more predominantly in its molecular form. These results suggest that the strong coordination ability in DME promotes the dissociation and solvation of phenyl-lithium, whereas in DEA, the steric hindrance introduced by methyl branches significantly reduces the coordination ability of Li⁺, thereby suppressing phenyl-lithium dissociation. These results further indicate the favorable compatibility of Li metal with FB in the presence of DEA. Moreover, due to the preferential adsorption of DEA on the surface of Li metal (path (10) in Fig. 2l and Supplementary Fig. 16), the deep substitution reaction between FB and Li metal is inhibited³⁴ and since DEA solvent also cannot stabilize phenyl-lithium through complexation, a thin LiF-rich interfacial layer forms on the surface of Li metal (path (14) in Fig. 2l and Supplementary Figs. 14 and 15).

Physicochemical properties and solvation structures of electrolytes

The physical parameters of DEA and DME are listed in Supplementary Table 4. The low freezing point (−113 ^oC) and low dielectric constant (3.49) of DEA make it an ideal solvent for low-temperature electrolytes. Although the electrostatic potential (ESP) distribution shows that the DEA molecule exhibits a higher electron cloud density around the O atoms compared to DME (Fig. 3a), the binding energy of DME-Li⁺ (−3.01 eV) is higher than that of DEA-Li⁺ (−2.59 eV) due to the unique five-membered ring chelate structure of DME-Li⁺ (Fig. 3b). The low binding energy of DEA-Li⁺ facilitates DEA solvent molecule desolvation at the electrode interface and improves the kinetic properties of the DEA-based electrolyte at low temperatures³⁵. It should be pointed out that although the RT ionic conductivity of the 1 M LiFSI in DEA/FB (1:1 v/v) LMCE is lower than the 4 M LiFSI in DEA/FB (1:1 v/v) LHCE, it shows higher low-temperature ionic conductivity than that of the LHCE (Supplementary Fig. 17, Table 5). Viscosity is a critical factor influencing the electrolyte wettability. The viscosity of the 1 M LiFSI in DEA/FB (1:1 v/v) LMCE is lower than that of the 4 M LiFSI in DME/FB (1:1 v/v) LHCE, 1 M LiFSI in DEA, and 1 M LiFSI in DME electrolytes (Fig. 3c). The results suggest the 1 M LiFSI in DEA/FB (1:1 v/v) LMCE has a favorable wettability, which is verified by the contact angle tests (Fig. 3d).

Fig. 3. Physicochemical properties and solvation structures of electrolytes.

a Electrostatic potential of DME and DEA solvent molecules. b The binding energies of Li⁺-solvent complex. c Viscosity of different electrolytes. d Contact angle of different electrolytes on the Celgard2500 separator. e ⁷Li NMR spectra of different electrolytes. f–h Raman spectra of different electrolytes (f) 1 M LiFSI in DME (g) 1 M LiFSI in DEA (h) 1 M LiFSI in DEA/FB (1:1 v/v). i Snapshots derived from MD simulation and corresponding proportions of different solvation structures in 1 M LiFSI in DME and 1 M LiFSI in DEA/FB (1:1 v/v) electrolytes respectively. j Ratios of solvation structures in which the number of FSI^- ≥ 4. k ¹H NMR spectra of different solvents.

Raman spectroscopy was employed to investigate the solvation structures of different electrolytes. In 1 M LiFSI in DME electrolyte, the S-N-S bending peak of FSI^- exhibits a significant red shift compared to pure LiFSI (Fig. 3f and Supplementary Fig. 18), indicating a strong dissociation of the Li⁺/FSI^- interactions when dissolved in the DME solvent. Distinct fitting peaks at 723 cm⁻¹ and 733 cm⁻¹ represent free FSI^- and coordinated FSI^-, indicating the presence of abundant solvent-separated ion pairs (SSIPs) and CIPs in the 1 M LiFSI in DME electrolyte³⁶. In contrast, a smaller red shift is observed in the 1 M LiFSI in DEA electrolyte, with characteristic peaks at 733 cm⁻¹, 741 cm^−1, and 754 cm⁻¹, corresponding to CIP, AGG I, and AGG II, respectively (Fig. 3g)³⁷. Upon the introduction of FB, the redshift decreases further, and the AGG content increases in the 1 M LiFSI in DEA/FB (1:1 v/v) LMCE (Fig. 3h). The addition of FB facilitates more anion recruitment into the Li⁺ solvation sheath. Additionally, ⁷Li NMR directly confirms the coordination state of Li⁺ with different solvents (Fig. 3e). The ⁷Li chemical shift displays a downfield displacement in 1 M LiFS in DEA compared to 1 M LiFSI in DME, suggesting a weaker Li⁺-solvent interaction for DEA than for DME³⁸. Upon introducing FB, the ⁷Li chemical shift in 1 M LiFSI in DEA/FB (1:1 v/v) exhibits an upward shift than that in 1 M LiFSI in DEA, which is attributed to the increased ion pairing in the solvation sheath³⁹.

MD simulations were further utilized to explore the solvation structures of different electrolytes. The snapshots from MD simulations indicate that the FSI^- is evenly dispersed in the 1 M LiFSI in the DME electrolyte, suggesting that LiFSI is fully dissociated. In contrast, bulky aggregate networks are observed in the 1 M LiFSI in DEA electrolyte and 1 M LiFSI in DEA/FB (1:1 v/v) LMCE, indicating the presence of a significant amount of undissociated ion pairs. For 1 M LiFSI in DME, 1 M LiFSI in DEA, and 1 M LiFSI in DEA/FB (1:1 v/v) LMCE, the proportion of solvation structures with four or more FSI^- anions in the solvation sheath increases sequentially, with values of 42.7%, 52.9%, and 55.1%, respectively (Fig. 3i, j and Supplementary Fig. 19). The ¹H NMR reveals that the FB diluent molecule form hydrogen bond with DEA, which regulates the solvation structure and induces more anions to enter the solvation structure (Fig. 3k)⁴⁰. The radial distribution function (RDF) and average coordination number (CN) of Li⁺ for different electrolytes are displayed in Supplementary Fig. 20. The higher frequency (the absolute value of g(r)) and higher CN (3.5) of Li-OFSI^- in 1 M LiFSI in DEA/FB (1:1 v/v) LMCE suggest that more FSI^- can enter the first solvation sheath. The FSI^--rich complexes in the electrolyte promote the formation of a robust LiF-rich SEI layer on lithium metal negative electrodes, resulting in improved reversibility.

Performance of the Li negative electrode

Li | |Li symmetric cells were employed to evaluate the compatibility of different electrolytes with Li metal (Fig. 4a). The Li | |Li symmetric cell with the 1 M LiFSI in DME/FB (1:1 v/v) electrolyte exhibits significant potential fluctuations, contributed to the violent chemical side reaction between the electrolyte and Li metal. The cell with the 4 M LiFSI in DME/FB (1:1 v/v) LHCE operated for only 750 h before the overpotential sharply increased. In contrast, the cell with the 1 M LiFSI in DEA/FB (1:1 v/v) LMCH demonstrated a stable cycle life exceeding 1000 h. This performance is also more stable compared to cells with the 1 M LiFSI in DME and 1 M LiFSI in DEA electrolytes (Supplementary Fig. 21). To assess the reversibility of Li metal in different electrolytes, we conducted Li metal CE tests at 25 °C using the Aurbach method (Fig. 4b)⁵. The CE for Li | |Cu cells with the 1 M LiFSI in DME and 1 M LiFSI in DEA electrolytes is 98.32% and 98.93%, respectively. In contrast, the CE of the Li | |Cu cell with the 1 M LiFSI in DEA/FB (1:1 v/v) LMCE increases to 99.43%, which is higher than that of the 4 M LiFSI in DME/FB (1:1 v/v) LHCE (99.23%), 1 M LiFSI in DME/TTE (1:1 v/v) LMCE (98.92%) and 1 M LiFSI in DEA/TTE (1:1 v/v) LMCE (99.01%) (Supplementary Figs. 22, 23).

Fig. 4. Electrochemical performance of Li negative electrode.

a Cycling performance of Li | |Li symmetric cells in different electrolytes. b CE tests of Li | |Cu cells in different electrolytes using the Aurbach method at 25 °C. c Long cycle performance of Li | |Cu cells in different electrolytes at 25 °C. Inset: Localized enlarged figures from the 100^th to 200^th cycles. d Performance comparison of Li negative electrodes in terms of Li salt concentration, cycle life, and average CE (The source of the literature data shown in this figure can be found in Supplementary Table 6). e–g Voltage capacity curves in different electrolytes after different cycles: (e) 1 M LiFSI in DME; (f) 1 M LiFSI in DEA; (g) 1 M LiFSI in DEA/FB (1:1 v/v). Inset: enlarged view of the stripping process. h CE test of Li | |Cu cells with 1 M LiFSI in DEA/FB (1:1 v/v) LMCE using the Aurbach method at −20 °C and −40 °C. i Long cycle performance of Li | |Cu cells in 1 M LiFSI in DEA/FB (1:1 v/v) LMCE at −20 °C. Inset: voltage capacity curves after different cycles and the localized enlarged figure from the 300^th to 500^th cycles.

Figure 4c shows the long-cycle performance of the Li | |Cu cells in different electrolytes (Fig. 4e–g). The Li | |Cu cell with the 1 M LiFSI in DME electrolyte fails after 150 cycles, with an average CE of 97.63%. In contrast, the reversibility of the cells with the 1 M LiFSI in DEA/FB (1:1 v/v) LMCE can be stably operated for 800 cycles with a high average CE of 99.11% (300^th-800^th: 99.24%), which is also higher than the cells with the 1 M LiFSI in DEA electrolyte (380 cycles; average CE: 98.45%). Even in 4 M LiFSI in DME/FB (1:1 v/v) LHCE, the CE of the Li | |Cu cell decreases after 700 cycles, with an average CE of 98.64%, which is lower than that of the cell with 1 M LiFSI in DEA/FB (1:1 v/v) LMCE (Supplementary Fig. 24). The Li metal in our 1 M LiFSI in DEA/FB (1:1 v/v) LMCE shows good stability and reversibility, outperforming some relevant recent reports in terms of Li salt concentration and cycle life (Fig. 4d, and Supplementary Table 6)^{35,39,41–53}. Furthermore, the Li | |Cu cells with our 1 M LiFSI in DEA/FB (1:1 v/v) LMCE showed high CEs of 98.41% and 97.74% at temperatures as low as −20 °C and −40 °C, respectively (Fig. 4h). The Li | |Cu cell can cycle stably for 500 cycles at −20 °C, with an average CE of 98.08% (Fig. 4i). The Li | |Li symmetric cell can also cycle stably over 2000 hours at −20 °C without an increase in potential or short circuit, further demonstrating the stable Li stripping and deposition (Supplementary Fig. 25).

SEI characterization of the Li negative electrode

XPS was employed to analyze the detailed chemical composition of SEI on the cycled Li metal in different electrolytes. Figure 5a, b show the C 1 s and F 1 s XPS spectra of cycled Li metals in different electrolytes with the etching time from 0 s and 60 s, and the detailed surface C and F element evolution versus sputtering time is summarized in Fig. 5c, d. It can be observed that the contents of organic species and LiF in the SEI on the cycled Li metals in different electrolytes follow a similar trend with increasing etching time. However, the increase in LiF and the decrease in organic compositions of the SEI on the cycled Li metal in the 1 M LiFSI in DEA/FB (1:1 v/v) LMCE are more pronounced, indicating the formation of a unique organic-inorganic gradient structure. Additionally, the C and F elements in the SEI on the cycled Li metal in the 1 M LiFSI in DEA/FB (1:1 v/v) LMCE always exhibit lower and higher proportions, respectively, compared to those in the 1 M LiFSI in DME and 1 M LiFSI in DEA electrolytes. This is attributed to the suppression of solvent decomposition and the preferential decomposition of anions. The inorganic-rich SEI components contribute to suppressing Li dendrite growth and promoting stable Li metal deposition/stripping.

Fig. 5. Characterization of SEI components on Li metal surfaces.

a, b Depth-profiling XPS spectra of the Li metal surface in different electrolytes after cycling in the Li | |Li symmetric cell: (a) C 1 s spectra; (b) F 1 s spectra. c, d Elemental ratio evolution of Li metal surface at different etching times: (c) C atom, (d) F atom. e, f 3D reconstructions of the TOF-SIMS images of cycled Li metal surfaces in different electrolytes. g, h Depth profiles of different segments on cycled Li metal surfaces: (g) C₂HO^-; (h) LiF₂^- and LiO^-. i Schematic diagram of SEIs derived from the 1 M LiFSI in DME and 1 M LiFSI in DEA/FB (1:1 v/v) LMCE on Li metal surfaces. j SEM images of the surface and cross-section of Li metal in different electrolytes after cycling in the Li | |Li symmetric cell (After 10 cycles with the current density of 0.5 mA cm^-2 and the areal capacity of 5 mAh cm^-2).

Time-of-flight secondary ion mass spectrometry (TOF-SIMS) was further employed to investigate the SEI components on the cycled Li negative electrode in different electrolytes (Fig. 5e, f). The depth profiles of representative secondary ion fragments show that the Li metal cycled in the 1 M LiFSI in DEA/FB (1:1 v/v) LMCH exhibits a stronger signal of LiF₂^- and LiO^-, and a weaker signal of C₂HO^-, compared to the Li metal cycled in the 1 M LiFSI in DME electrolyte. Moreover, the content of inorganic ion fragments in the SEI of Li metal cycled in the 1 M LiFSI in DEA/FB (1:1 v/v) LMCH is consistently higher than that in the 1 M LiFSI in DME electrolyte throughout the entire sputtering process (Fig. 5g, h), which is consistent with XPS results. The anion-rich solvation sheath induced by the “branched” DEA and FB molecules in the 1 M LiFSI in DEA/FB (1:1 v/v) LMCH promotes the formation of an inorganic-rich organic-inorganic gradient SEI layer (Fig. 5i and Supplementary Fig. 26). The postmortem morphology analysis of cycled Li metals in different electrolytes was conducted using SEM (Fig. 5j). Li metal cycled in the 1 M LiFSI in DME electrolyte shows an uneven, cracked surface morphology, with the porous deposition resulting in a cross-sectional thickness of 34 μm. In contrast, when cycled in the 1 M LiFSI in DEA electrolyte, the Li deposition appears uniform, reducing the cross-sectional thickness to 24 μm. For the Li metal cycling in the 1 M LiFSI in DEA/FB (1:1 v/v) LMCE, the cross-sectional thickness further decreases to 10 μm, with the surface becoming even denser. This improvement is primarily attributed to the formation of a thin, uniform, and robust inorganic-rich SEI layer (Supplementary Fig. 27).

Performance of the SPAN positive electrode

Sulfur-based positive electrodes, known for their high theoretical specific capacity and abundant availability in the Earth’s crust, are considered environmentally friendly candidates for high-specific-energy LMBs^54,55. SPAN used in this work was synthesized by heating sulfur and polyacrylonitrile⁵⁶, with further material details provided in Supplementary Table 7. The cycling stability of SPAN in various electrolytes was initially evaluated at a current density of 300 mA g⁻¹ (1 C = 600 mAh g⁻¹) (Fig. 6a and Supplementary Figs. 28, 29). The Li | |SPAN cell with the 1 M LiFSI in DEA/FB (1:1 v/v) LMCE shows stable cycling performance for 500 cycles with a high capacity retention of 83% and the average CE reaches to 99.83% and there is no capacity decay for the first 100 cycles. The Li | |SPAN cells with the 1 M LiFSI in DME/TTE (1:1 v/v) LMCE and 1 M LiFSI in DEA/TTE (1:1 v/v) LMCE only achieve the capacity retention of 78.4% and 92.6% after 100 cycles, respectively. For the Li | |SPAN cell with the 1 M LiFSI in DEA electrolyte, a capacity retention of 80% is achieved after 341 cycles, with an average CE of 99.74%. In contrast, the Li | |SPAN cell cycled in the 1 M LiFSI in DME electrolyte suffers from a significant capacity decay and with an average CE significantly lower than 100% (Supplementary Fig. 30). This is primarily attributed to the shuttle effect induced by the extensive dissolution of lithium polysulfides (LiPSs). Lithium polysulfide dissolution tests confirm that Li₂S₆ shows a much higher solubility in DME solvent than in DEA solvent (Fig. 6b). DFT calculations were employed to further study the interactions of LiPSs (Li₂S₄, Li₂S₆, Li₂S₈) with different solvents (Fig. 6c). A lower binding energy is observed in the LiPS-DEA complexes, indicating a weaker affinity of DEA toward LiPSs. This approach of comparing DFT binding energies across different solvents emphasizes relative interaction trends, enabling a clearer assessment of solvent selectivity toward LiPSs.

Fig. 6. Performance and characterization of SPAN positive electrodes.

a Long cycle performance of Li | |SPAN half cells in different electrolytes. b Dissolution photograph of 0.25 mol Li₂S₆ in different solvents after 24 h. c Optimized structures of Li₂S_x-solvent (x = 4, 6, and 8) complex and binding energy of LiPSs with DEA and DME. d CV curves of Li | |SPAN half cells in different electrolytes (scan rate: 0.1 mV s⁻¹). e, f XPS spectra of the SPAN cycled in the 1 M LiFSI in DME electrolyte and 1 M LiFSI in DEA/FB (1:1 v/v) LMCE. g Elemental ratios of SPAN after cycling in different electrolytes. h Schematic diagram of Li | |SPAN batteries operation in different electrolytes.

Cyclic voltammetry (CV) tests were employed to reveal the detailed transformations of SPAN in different electrolytes (Fig. 6d and Supplementary Fig. 31). The Li | |SPAN cells with the 1 M LiFSI in DEA electrolyte and 1 M LiFSI in DEA/FB (1:1 v/v) LMCE show typical “solid-solid” transformations. Specifically, the reduction peaks at 1.7 V and 2.0 V correspond to the reduction of solid S to solid Li₂S/Li₂S₂, while the oxidation peak around 2.4 V signifies the reversible transformation of discharge products⁵⁷. Aside from the peaks corresponding to the “solid-solid” transition, the Li | |SPAN cell cycled in the 1 M LiFSI in DME electrolyte exhibits typical solid-liquid conversions, indicating the existence of significant LiPSs in the electrolyte. To comprehend the distinctions in the positive electrode electrolyte interphase (CEI) of SPAN, XPS analysis was conducted on the cycled SPAN surface components in different electrolytes (Fig. 6e–g and Supplementary Fig. 32). Characteristic peaks of LiF and SO_x are identified in SPAN after cycling in different electrolytes, mainly resulting from the decomposition of FSI^-. LiF in the CEI layer can effectively inhibit polysulfide diffusion⁵⁸. Compared to the 1M LiFSI in DME electrolyte, the SPAN cycled in the 1M LiFSI in DEA electrolyte and 1M LiFSI in DEA/FB (1:1v/v) LMCE show a higher content of inorganic LiF, which is primarily a result of the FSI^- anion-rich solvation sheath. Furthermore, the signals of C-S/S-S on the SPAN surface after cycling in the 1 M LiFSI in DEA/FB (1:1 v/v) LMCE are weaker than those in the 1 M LiFSI in DME electrolyte, indicating the formation of dense and homogeneous CEIs on the SPAN positive electrode in the 1 M LiFSI in DEA/FB (1:1 v/v) LMCE. The unique solvation structure formed by the “branched” DEA molecules effectively inhibits the dissolution of LiPSs and promotes the formation of robust SEI and CEI layers, which synergistically enhance the long-term stable operation of the Li | |SPAN battery (Fig. 6h).

Li||SPAN batteries in extreme conditions

To evaluate the performance of our designed electrolyte under extremely low-temperature (LT) and practical scenarios, we conducted tests on Li | |SPAN batteries with the 1 M LiFSI in DEA/FB (1:1 v/v) LMCE in different temperatures. The Li | |SPAN cell cycled at −20 °C is still able to maintain a specific capacity of 401.7 mAh g⁻¹ (4^th cycle) at 300 mA g⁻¹, with a capacity retention rate of 103.1% after 400 cycles (Compared to 4^th cycle) and an average CE of 99.86% (Fig. 7a, b). Even at −40 °C, the Li | |SPAN cell cycled at 120 mA g⁻¹ still sustains a specific capacity of 279.3 mAh g⁻¹ (26^th cycle). The specific capacity of 281.6 mAh g⁻¹ is maintained after 150 cycles, with a capacity retention rate of 100.8% (Compared to 26^th cycle) and an average CE of 99.91% (Fig. 7c). Additionally, the cell with the 1 M LiFSI in DEA/FB (1:1 v/v) LMCE can operate stably even at −60 °C (89.1 mAh g⁻¹) and, after returning to 25 °C, its specific capacity can recover to 542.9 mAh g⁻¹, exhibiting good temperature adaptability (Fig. 7d). The temperature distribution map of China on Jan. 1^st, 2024, shows that more than half of the country is experiencing temperatures below 0 °C, with some areas in northern Inner Mongolia and Heilongjiang being below −40 °C (Supplementary Fig. 33). In such extreme conditions, traditional commercial electrolytes struggle to function, highlighting the advantages of our designed electrolyte and its promising application prospects.

Fig. 7. Electrochemical performance of Li | |SPAN batteries at low temperatures and in practical conditions.

a Long cycle performance of Li | |SPAN half cells in the 1 M LiFSI in DEA/FB (1:1 v/v) LMCE at −20 °C. b Capacity voltage curves of the Li | |SPAN cell cycled in the 1 M LiFSI in DEA/FB (1:1 v/v) electrolyte at −20 °C. c Long cycle performance of Li | |SPAN half cells in the 1 M LiFSI in DEA/FB (1:1 v/v) LMCE at −40 °C. d Electrochemical performance of Li | |SPAN half cells cycled at temperatures ranging from 25 °C to −60 °C. e Long cycle performance of Li | |SPAN full cells in the 1 M LiFSI in DEA/FB (1:1 v/v) LMCE at 25 °C. f Long cycle performance of Li | |SPAN pouch cells in the 1 M LiFSI in DEA/FB (1:1 v/v) LMCE at 25 °C. Inset: Digital photo of Li | |SPAN pouch cells and schematic diagram of pouch cell structure. g Weight ratio of each component in the pouch cell. h Specific parameters of the Li | |SPAN pouch cell. i Comparison of the performance of Li | |SPAN pouch cells in this study with other advanced Li | |SPAN pouch cells (The source of the literature data shown in this figure can be found in Supplementary Table 9).

The electrochemical performance of Li | |SPAN full cell with the 1 M LiFSI in DEA/FB (1:1 v/v) LMCE under practical conditions was further investigated. Specifically, a high-loading SPAN electrode (discharge areal capacity: 2.5 mAh cm^-2) paired with 80 μm Li foil and 45 μL electrolyte was charged and discharged at 120 mA g⁻¹ (Fig. 7e and Supplementary Fig. 34). 90.25 % of the initial discharge capacity was retained after 85 cycles, with an average CE of 99.27 %. The 54-mAh Li | |SPAN pouch cell was assembled and tested at 12 mA g⁻¹, using a 3.3 mAh cm^-2 SPAN electrode and 50 μm Li foil as the positive electrode and negative electrode, respectively, with a lean electrolyte (E/C = 3 g Ah⁻¹) (Fig. 7f, g, Supplementary Fig. 35 and Supplementary Table 8). The Li | |SPAN pouch cell can be stably operated over 50 cycles with a capacity retention of 37.3 mAh. Moreover, the Li | |SPAN pouch cell achieves a gravimetric energy of 334.53 Wh kg⁻¹ (excluding package) (Fig. 7h). The Li | |SPAN batteries based on our 1 M LiFSI in DEA/FB (1:1 v/v) LMCE show relative improvement compared to some previously relevant reported Li | |SPAN batteries with different advanced electrolytes in terms of lithium salt concentration, minimum operating temperature, Li negative electrode CEs at RT, SPAN cycle numbers at LT and SPAN mass loading (Fig. 7i and Supplementary Table 9)^{35,38,42,44,59–62}.

In summary, we found that reducing the LiFSI content in the 4M LiFSI in DME/FB (1:1 v/v) electrolyte resulted in severe degradation reactions of the electrolyte with lithium metal negative electrodes. This finding challenges the conventional understanding that FB-diluted DME-based electrolyte systems are inherently compatible with lithium metal negative electrodes and hinders the development of advanced DME-based electrolytes. Based on this unexpected phenomenon, we elucidated the degradation mechanism of the FB-diluted DME-based electrolyte undergoing continuous severe side reactions with lithium metal negative electrodes mediated by free DME solvent molecules. Inspired by the understanding of degradation mechanisms between the FB-diluted DME-based LHCE and lithium metal negative electrode, we designed a FB-diluted DEA-based localized medium-concentration electrolyte (LMCH: 1 M LiFSI in DEA/FB (1:1 v/v)), which can enhance the stability of the electrolyte system under a relatively low Li salt concentration. The “branched” DEA solvent molecules can effectively suppress the complexation of phenyl-lithium, further inhibit the deprotonation of FB, and suppress the formation of intermediate products, enhancing the compatibility between the 1 M LiFSI in DEA/FB (1:1 v/v) LMCE and lithium metal negative electrodes. Furthermore, the DEA molecules are preferentially adsorbed on the surface of lithium metal negative electrodes, blocking the substitution reaction between FB and Li metal, which suppresses continuous Li corrosion and the decomposition of the solvent and diluent. Benefiting from the large steric hindrance provided by its unique ‘branched’ molecular structure, DEA facilitates the formation of more CIPs and AGGs in the electrolyte, even at low Li salt concentrations, thereby enhancing the stability of Li lithium metal negative electrodes. As a result, the lithium metal negative electrode displayed a high CE of 99.43% at 25 °C and 97.74% at −40 °C. The Li | |SPAN cells based on our 1 M LiFSI in DEA/FB (1:1 v/v) LMCE show a high capacity retention of 103.1% after 400 cycles at −20 °C and 100.8% after 150 cycles at −40 °C, respectively. The elaboration of the applicability of FB as a diluent in ether-based electrolytes is expected to spur further exploration of low-cost, sustainable, high-performance FB-diluted ether-based electrolytes for LMBs.

Methods

Raw materials

The lithium bis(fluorosulfonyl)imide (LiFSI, 99.9%) and 1,2-dimethoxyethane (DME, 99.9%) were purchased from DoDoChem. The dimethyl acetal (DEA, ≥ 98%), fluorobenzene (FB, ≥ 98%), and polyacrylonitrile (PAN, Mw: 149,000-151,000) were purchased from Aladdin. The sulfur (S, ≥ 99.99%) powder was purchased from Canrd Technology Co. Ltd. The lithium foils (thickness: 50 μm, 450 um, Li content ≥99.9%) were obtained from Canrd Technology Co. Ltd. All the materials were used directly after purchase, with no further purification.

Synthesis of SPAN

Sulfur powder and PAN powder were firstly mixed under a weight ratio of 4:1 and ball milled under 500 rmp for 1 h. The obtained mixture was then placed in a horizontal quartz tube which was installed in a furnace. The furnace was heated to 450 °C with a heating rate of 2 °C min⁻¹ and kept at 450 °C for 6.0 h under an argon atmosphere. SPAN was obtained after the furnace was cooled to room temperature with continuous argon flow.

Preparation of electrodes

A common coating process was used to prepare SPAN electrodes. First, SPAN, Super P, and Poly(acrylic acid) (PAA) are manually mixed in an agate mortar at a mass ratio of 8:1:1, for the high-loading SPAN electrode, the mass ratio of SPAN, Super P, and PAA is 9.0:0.5:0.5. Add deionized water to the above mixture, and then mix the slurry evenly through a high-speed homogenizer. Next, the uniformly mixed slurry is evenly coated on the aluminum foil (thickness: 16 μm, no further treatment) through a doctor blade. Finally, the solvent was removed in an 80 °C vacuum drying oven for 12 h to obtain the SPAN electrode. By adjusting the thickness of the coating, electrodes with different surface capacities can be obtained. The electrode sheets were then cut to the corresponding size using a battery slitter (MSK-520, Shenzhen Kejing Star Technology Company).

Electrochemical measurements

The electrochemical performance was tested by CR2032-type coin cells (diameter: 20 mm, height: 3.2 mm, material: 304 stainless steel) on the battery testing systems (NEWARE Battery Test system, Shenzhen, China). All the electrolytes were prepared in a glove box filled with argon gas (H₂O and O₂ ≤ 0.01 ppm) at 25 °C and used immediately. All the cells were also assembled in the glove box filled with argon gas. A two-electrode system was used to test Li | |Cu, Li | |Li, and Li | |SPAN cells. Li metal foil (diameter: 10 mm) serves as one electrode of the cells, and copper foil (diameter: 14 mm), Li metal foils (diameter: 10 mm), or SPAN electrodes (diameter: 8 mm) serve as the other electrodes and with 60 μL electrolytes. For Li | |SPAN full cells and pouch cells, thin Li metal foils (thickness: 50 μm) were used as the negative electrode, and high-loading SPAN electrodes were used as the positive electrode, all SPAN cells were cycled in the voltage range of 1–3 V, and the energy of pouch cell is based on the stack (excluding package). The pouch cell was tested with an initial pressure of about 100 kPa. All cells used Celgard 2500 (thickness: 25 μm; porosity: 55%; average pore size: 0.064 μm) as a separator. It is worth mentioning that before the long cycle at −40°C, Li||SPAN cells were activated by 0.05 °C for 25 cycles. The cyclic voltammetry curves of Li | |SPAN cells with electrolytes were measured by CHI660E electrochemical workstation at the voltage range of 1–3 V and with the scan rate of 0.1 mV s⁻¹. The ionic conductivity of different electrolytes was measured by testing the electrochemical impedance of symmetric coin cells with stainless steel electrodes on the CHI660E (Shanghai Chenhua Instrument Co Ltd.) electrochemical workstation in the frequency range of 100 kHz to 0.1 Hz with an amplitude of 5 mV. The specific value was calculated by the following formula:

$$\sigma =\frac{L}{RA}$$ 1

Where $\sigma$ is the Ionic conductivity, L represents the thickness of the separator, A is the area of stainless steel, and R is the resistance.

Characterizations

The micromorphology of all samples was characterized by field-emission scanning electron microscopy (SEM, FEI Nova NanoSem 450). The solvation structure of different electrolytes was measured by a Raman spectrometer (LabRAM HR Evolution) equipped with a green diode laser (λ = 532 nm). The chemical components of samples were analyzed by an X-ray photoelectron spectrometer (XPS, Thermo Scientific K-Alpha) and Time of Flight Secondary Ion Mass Spectrometry (TOF-SIMS, PHI NanoⅡ TOF). The nuclear magnetic resonance (NMR) spectra of solvents and electrolytes were acquired by an NMR spectrometer (AVANCEⅢ 400 MHz). The products of the Li metal react with DME/FB (1:1 v/v) mixed solvents were analyzed by Gas chromatography-mass spectrometry (GC-MS, Agilent 5977 A), and the mixed solvents were diluted with ethyl ether before the test. The thickness of the SEIs was characterized by cryogenic transmission electron microscopy (cryo-TEM, Titan Krios G3). The viscosity of different electrolytes was obtained using a rotational rheometer (HAAKE MARS III). The contact angles of the different electrolytes to the Celgard 2500 separator were obtained by a contact angle measuring instrument (Dataphys OCA15EC).

Quantum chemistry calculations

Quantum chemistry calculations for the pure molecule’s parts were performed using the Gaussian 16 package⁶³. The structures of molecules were optimized at the B3LYP/6-311 + G(d,p) level. All the visualizations of quantum chemistry calculations were implemented by the GaussView software.

First-principle calculations

First-principle simulations were performed on the basis of density functional theory calculations with the plane-wave technique, which is implemented in the Vienna ab initio simulation package⁶⁴. The gradient corrected exchange-correlation functional of Perdew, Burke, and Ernzerhof models were used under the projector augmented wave method⁶⁵, with a cut-off kinetic energy of 500 eV for plane wave basis. Additionally, the DFT-D3(BJ) functional method was considered. For the bulk structure optimization, Monkhorst-Pack k-points mesh was set to 11×11×11 for the conventional cells of Li. The slab model of Li (110) surface contains three atomic layers in thickness with the bottom layer fixed, and a 4 × 3 supercell in the lateral plane is adopted with 72 atoms, the k-points samplings were reset to 3 × 3 × 1 for surface structure optimization and 15 vacuum along the vertical direction was used for erasing the effect of periodic condition. Binding energy ($E_b$) between the Li surface and molecules is defined as:

$$E_b=E_{complex}-E_{surface}-E_{molecule}$$ 2

In which $E_{complex}$, $E_{surface}$ and $E_{molecule}$ were the total energy of the adsorption system, Li surface and investigated molecules, respectively. The convergence criterion of the total energy was set up to be within 1 × 10^-5eV, while all the atoms and geometries were optimized until the residual forces became less than 1 × 10^-2eV/Å. All the visualizations of first principle calculations were implemented by VESTA software.

Materials dynamic simulations

Dynamic simulations of all materials were conducted using the GROMACS 2022.5 simulation package, paired with the amber99sb-ildn force fields^66,67. Atomic charges for the solvent and anions were generated via the restrained electrostatic potential fitting (RESP) procedure, utilizing the Multifwn software⁶⁸. Both equilibrium and production simulations were carried out in the NPT ensemble, under conditions of constant pressure (1.01325 bar) and temperature (298.15 K), within a cubic box that applied periodic boundary conditions across all XYZ Cartesian directions. For the equilibrium process, temperature was sustained through V-rescale coupling with a time constant of 0.2 ps, while pressure control was achieved using the Berendsen barostat (coupling constant: 0.5 ps). This equilibrium simulation lasted for 5 ns. The production simulation employed V-rescale coupling for temperature regulation and the Parrinello-Rahman barostat for pressure control; it ran for 20 ns to acquire the equilibrium Li⁺ coordination shell structure of electrolytes at room temperature. Electrostatic interactions were addressed using the Particle-Mesh-Ewald (PME) method⁶⁹, and all visualizations of the materials dynamic simulations were executed via the Virtual Molecular Dynamics software. To further explore the dissociation energy of PHLi in various solvents, Potential of Mean Force (PMF) analysis based on the radial distribution function was performed:

$$PMF\left(r\right)=-k_B\cdot T\cdot lng\left(r\right)$$ 3

Where k_B is the Boltzmann Constant, T is temperature and g(r) is the radial distribution function.

The atomic coordinates of the 1 M LiFSI in DEA/FB (1:1 v/v) electrolyte obtained from the first and last frames of molecular dynamics simulation are provided in Supplementary Data 1.

Author contributions

Y. Li. conceived the project, and directed and supervised the work. Y.Li., H.P., and T.W. conducted the concept design. Y. Jiang provided material characterizations. H.P. and J.O. prepared the electrode materials and conducted the experiments. Z. Wang and T. Wang conducted the DFT calculations and MD simulations. H. Pan and X. Li analyzed the results and drafted the manuscript with L.Ch., C.S., N.Ch., Q.Y., and S.Wu. H.P. and Y.Li. wrote the paper. X.Li., T.W., and Y.Li. revised the manuscript. All authors commented on the final manuscript.

Peer review

Peer review information

Nature Communications thanks the 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.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-64784-2.