Introduction

The specific energy density of current state-of-the-art Li-ion batteries (LIBs) is approaching the maximum capacity (300 Wh kg−1) allowed by intercalation chemistry1. Li metal batteries (LMBs) based on Li | |LiNi0.8Co0.1Mn0.1O2 (NCM811) can potentially reach the 500 Wh kg−1 goal set by electric vehicle and electrified aviation applications for a long driving range per charge2,3 due to the high theoretic capacity (3860 mAh g−1)/lowest redox potential (−3.04 V vs. standard hydrogen electrode, SHE) of the Li-metal anode (LMA)4 and the high-voltage/high-capacity NCM811 cathode5. Besides range, requirements on safety at a high temperature during fast charging, and lifespan will heavily inform requirements at the cell level, especially important for electrified aviation6. However, the currently available electrolytes cannot support Li | |NCM811 cells to achieve all the metrics at the same time, especially operation at a high temperature.

At elevated temperatures of >60 oC, the cycling stability of current LIBs quickly degrades due to the increased reactivity of the electrode materials and consequent side reactions, namely severe solid electrolyte interphase (SEI) formation. LiF-rich inorganic SEI and cathode-electrolyte-interface (CEI) enable high energy Li | |NCM811 cells to achieve long cycle life because LiF has the high interfacial energy and weak bonding to Li metal anode and NCM811 cathode4,5,7,8, thus can effectively reduce the stress/strain when electrodes undergo large volume changes during cycling. Recent electrolyte designs for Li | |NCM811 cells such as high-concentration electrolytes (HCE) with hydrofluoroethers (HFEs)9 and localized HCE (LHCE) with antisolvents10,11 focus on the formation of the LiF-rich SEI and CEI by promoting the reduction of fluorinated anion and suppressing the solvent reduction, which significantly enhances the cycling stability of Li | |NCM811 cells at room temperature. However, the poor thermal stability of these hydrofluoroether-based electrolytes limits their safety, especially at high temperatures, even if the LiF interphase is stable at high temperatures. This is because these ether-based solvents are often very volatile and flammable due to the low boiling point and flash point, which even imposes additional challenges to the cell assembly9,12,13,14. Notedly, the flammable or volatile solvents in the electrolytes often lead to thermal runaway15,16 and even explosion if the battery is heated to the flash points of electrolytes17. Even the flame-retardant highly-fluorinated ether-based electrolytes mitigate the fire hazard of batteries in ambiance; however, the battery thermal runaway cannot be suppressed18. In addition, the ether solvents also have a low anodic stability voltage7,12, limiting the cut-off voltage to 4.3 V.

Several factors noted above discourage the use of ether solvents for the safe operation of LMBs, especially at a high temperature10,11. Mixing ether-based electrolytes with high flash point carbonate solvents or flame-retardant additives such as triethyl phosphate (TEP) or trimethyl phosphate (TMP) is a common solution, but the strong solvation capability of phosphate, in turn, promotes the formation of organic-rich SEI, which usually reduces Coulombic efficiency (CE), promotes Li dendrite, resulting in poor cycle life, and severe safety issues19. Since the disadvantage of each solvent is inevitable in conventional “cocktail” electrolytes from mixing multi-solvents, the performance of the current electrolytes is always dominated by the worst component (Supplementary Table 1)20. The requirements for both high safety and high performance of Li | |NCM811 batteries still cannot be simultaneously met. Therefore, additional cooling systems are indispensable for the power battery packs to operate at milder temperatures below 60 °C21,22, which reduces their actual energy density and utilization efficiency23,24. The electrolyte design must compromise these required performance metrics. Therefore, a multifunctional electrolyte design is crucial for LMBs to simultaneously deliver all the required performance metrics.

To simultaneously achieve both high-safety and high performance of Li | |NCM811, a rationally designed solvent via integrating the different function groups, could take full advantage of the functions while shielding their weakness. For example, by adding fluorine group onto ether solvent, the CE of both Li anode and NCM811 cathode can be enhanced21. However, the hydrofluoroethers have low flash points, which are easily ignited in the gaseous state under high temperatures or during the thermal runaway of the battery. In addition, the vapor pressure of hydrofluoroether (about 20–30-fold higher than that of diethyl carbonate) is too high to prevent gasifying, thus greatly increasing the safety risk (Supplementary Tables 2, 3). Therefore, designing an electrolyte that enables LMBs to achieve high safety and high performance in a wide temperature range is highly challenging and has never been reported to date.

In this work, we designed a safe electrolyte using a multifunctional single-solvent of fluorinated phosphate ester (FEBFP) solvent and lithium (fluorosulfonyl) (nonafluorobutanesulfonyl) imide (Li[(FSO2)(n-C4F9SO2)N], LiFNFSI) salt. The LiFNFSI salt in electrolytes has a higher reduction potential than that of the FEBFP solvent, it decomposes first on the surfaces of both Li anode and NCM811 cathode to form robust LiF-rich and CF2CF2-based organics reinforced SEI and CEI layers. In addition, the LiFNFSI salts with a high dissociation capability25,26 increase the salt solubility in the FEBFP solvent while avoiding corrosion with the aluminum (Al) current collector. The 0.8 M LiFNFSI/FEBFP electrolyte with a wide electrochemical stability window up to 5.6 V and ionic conductivity of 3.9 mS cm−1 at room temperature enables Li | |Li symmetric cells to stably charge/discharge for more than 1500 h with the average CE of 99.3%. Since the multifunctional FEBFP solvent has no flash point with a high boiling point and is nonflammable and flame-retardant, the electrolyte is intrinsically safe for full cells to operate even at a high temperature of 90 °C with the super-stable performance. The 3.5-Ah-grade Li | |NCM811 pouch cell delivers a high energy density of 531 Wh kg−1 without any flame and explosion after cycling under extreme conditions. The high-voltage and intrinsically safe electrolyte design provides an avenue to develop and enable high-energy batteries to operate in extreme conditions.

Result

High-voltage safe electrolyte (VSE) design

The current electrolyte design for LMBs mainly focuses on enhancing the electrochemical stability window to support both Li anodes and high-voltage cathodes. To minimize the solvent reduction that forms organic SEI, ether electrolytes were used to enable a high CE for Li deposition/dissolution but a narrow electrochemical stability window (ESW, less than 4 V vs Li+/Li) of ether electrolytes cannot support the Ni-rich NCM811 and other high-voltage cathodes13,27,28. Partial fluorination of ether solvents (e.g., bis(2,2,2-trifluoroethyl) ether, BTFE) can improve oxidative stability, but it also reduces the ionic conductivity and reduction potential25, in turn forming more organic SEI in addition to LiF. Moreover, most ether-based solvents have high volatility and low flash points, which impose additional challenges to cell assembly and safety13,14. Flame-retardant organophosphorus (e.g., TMP and TEP) has been used as either solvents or additives in the electrolyte to reduce the electrolyte flammability. However, the simple mixture of the organophosphorus with ether solvent degrades the cycle performance due to either the instability of SEI on graphite (or Li metal) anodes or CEI on high-voltage cathode17,19. To enable LMBs to simultaneously achieve a high energy density in a wide temperature range, and high safety even in abuse, a multifunction solvent that is intrinsically safe and has a wide stability window should be synthesized by integrating different function groups into a single-solvent molecule, which can take full advantage of the group functions while shielding their weakness.

We designed high-voltage safe electrolytes by synthesizing a fluorinated phosphate ester solvent (Fig. 1a), which integrates all the merits of fire-retardant phosphate esters (-(RO)3-P=O), Li metal compatible and salts soluble ethers (-CH2-CH2-O-CH2CF3), and nonflammable fluorine (-CF3) into a single molecule. By dissolving LiFNFSI salt (Fig. 1b) with a high dissociation capability25,26, high reduction potential, and less corrosion to Al current collector, the single-solvent 0.8 M LiFNFSI/FEBFP electrolyte that is thermally stable and intrinsically safe was prepared. This high-voltage safe electrolyte possesses a wide stable window up to 5.6 V (vs. Li+/Li) and the acceptable ionic conductivity of 3.9 mS cm−1, and enables LMBs with different cathodes to achieve super cycling stability even at 90 °C.

Fig. 1: Design strategy and physical/chemical property of LiFNFSI/FEBFP electrolyte.
figure 1

a Design scheme and molecular structures of the solvents. b Design strategy of Li salts with diversified fluorine. c The optical photograph of the hydrophilicity/hydrophobicity and flammability tests for STD and VSE electrolytes. d Thermogravimetric (TG) analyses for STD, CME, and VSE electrolytes (T05 is defined as the temperature when the sample weight reduces by 5%). e Linear sweep voltammetry of Li||Al half-cells containing STD, CME, and VSE at a scanning rate of 1 mVs−1.

Property of VSE electrolyte

Several fluorinated phosphate ester solvents were synthesized to understand the correlation between functional groups and their properties. The molecular structures and characterization of these phosphate solvents and the LiFNFSI salt are shown in Supplementary Figs. 1, 2. Phosphate solvents are nonflammable, flame-retardant, and weak-volatilized. As shown in Supplementary Fig. 3a, all the phosphate solvents have very high boiling points (>200 °C), which can largely expand the operating temperature and improve the safety of LMBs during cycling, storing, and transporting. Although the viscosities of phosphate solvents are higher than that of propylene carbonate (3.5–8 vs. 2.5 cP), they are much lower than that of EMITFSI (1-ethyl-3-methylimidazolium bis(trifluoromethanesulfonyl)imide) ionic liquid. After dissolving 0.8 M LiFNFSI salt, the room temperature ionic conductivities of these fluorinated phosphate ester electrolytes are in the range of 2.6 to 4.2 mS cm−1 (Supplementary Fig. 3b). The introduction of the ether group into fluorophosphate solvents increases the electrolyte viscosities, it also increases the ionic conductivities, enlarges stability window, and improve the safety of the electrolytes. Since 0.8 M LiFNFSI/FEBFP high-voltage safe electrolyte (denoted as VSE) has better overall properties, it was selected for further investigation. To better demonstrate the unique properties of the VSE over “cocktail” electrolyte, the co-solvent mixed electrolyte of 0.8 M LiFNFSI/TEP-BTFE (1:1 vol.) (denoted as CME) with similar function components to the VSE was used as a control electrolyte for comparison. Although the TFP-BTFE (TFP, tris(2,2,2-trifluoroethyl) phosphate) co-solvent has a closer function component to FEBFP, the salt hardly dissolves in TFP-BTFE co-solvent due to too many fluorinated functional groups, thus cannot form electrolyte for comparison. In addition, the standard 1 M LiPF6/EC-DMC electrolyte (denoted as STD) was also used as a reference electrolyte.

The salt concentration of 0.8 M for VSE and CME is selected due to the highest ionic conductivity of the electrolytes at 0.8 M concentration (Supplementary Fig. 3c). Therefore, 0.8 M CME and VSE electrolytes are selected for further study. The ionic conductivity of 0.8 M VSE electrolytes (3.9 mS cm−1) is lower than that of STD (9.4 mS cm−1) (Supplementary Fig. 3d) but is still significantly higher than that of the reported HCE electrolytes29.

The electrolyte wettability to the separator is another critical parameter controlling the electrolyte permeability. Due to the lower surface energy caused by the rich C-F bonds, the VSE electrolyte has a high electrolyte permeability as demonstrated by the low contact angle (22.3o) on hydrophobic PE/PP separator, which is similar to that (16.1o) of CME but is much smaller than STD (55.5o) electrolyte, demonstrating its high hydrophilicity and oleophobicity due to strong polarity of carbonate group (Supplementary Fig. 4). Furthermore, the transference number tLi+ of VSE electrolyte (0.62) is much larger than 0.54 for CME and 0.25 for STD electrolytes (Supplementary Fig. 5). The high transference number of the VSE is attributed to the reduced volume of Li+ complex caused by: (1) the low association between FNFSI anion and Li+ ion due to the large volume of FNFSI anion; (2) the reduced solvent solvation of Li+ ions due to strong electron attraction of -n-C4F9 function group and large molecular volume of the FEBFP solvent.

The thermal stability of the VSE, CME, and STD electrolytes was evaluated using nuclear magnetic resonance (NMR). As shown in the inset of Supplementary Fig. 6, the STD electrolyte becomes black with precipitation in the sealed NMR tube after storing at 50 °C for 2 weeks, demonstrating poor thermal stability, while CME and VSE electrolytes remained transparent and almost unchanged. Many redundant peaks in the NMR spectrum of STD electrolyte after storage at 50 oC also confirmed the instability of STD electrolyte at a high temperature of 50 °C. However, the spectra of CME and VSE electrolytes are almost the same before and after heating (Supplementary Fig. 6).

Another safety-related property of the LMBs is the hydrophobicity of the electrolytes. The high hydrophobic electrolytes can prevent LMA from air-moisture attack even if the cells are exposed to air in abuse. As shown in Fig. 1c, a clear interface between the VSE electrolyte (the lower part) and deionized water (the upper part) was observed, indicating its high hydrophobicity. Li foil could be chemically stable for several hours in the organic phase that is exposed to water. This feature can greatly improve the safety of LMBs under abuse. In contrast, the hydrophilic STD electrolyte is miscible with water showing serious security risks. Additionally, as shown in the flammability tests (Fig. 1c), the separator soaked with VSE electrolyte was nonflammable and exhibited a zero self-extinguishing time after removal of the flame, due to the intrinsic nonflammable and fire-retardant VSE electrolyte with no flash point. Meanwhile, the separator soaked with the flammable STD electrolyte immediately caught fire on the ignition, and then burning continued even after the flame was removed.

The flash points of STD and CME electrolytes are 26.9 and 9.7 °C, respectively, revealing their potential safety risk during operation. The volatility and thermal stability of electrolytes were further evaluated by thermogravimetric analysis (TGA). As shown in Fig. 1d, the weight loss of VSE electrolyte at 220 °C is only 18%, which is much lower than that (79%) of STD electrolyte at the same temperature. Due to the high volatility, the significant weight loss of the CME electrolyte starts at room temperature, and the consequent weight loss with an increase in temperature is not accurate and is shown in the dashed line (Fig. 1d).

The electrochemical anodic stability of the three electrolytes was examined by linear sweep voltammetry (LSV) on Al foil electrodes. The VSE electrolyte shows an oxidation potential of 5.6 V defined at 0.004 mA cm−2 (Fig. 1e)20,30, which is much higher than 4.2 V for the STD electrolyte and 4.7 V for the CME electrolyte. The actual oxidation potential of electrolytes on the NCM cathode may be slightly lower than the measured value using Al foil due to the catalytic and oxidative properties of transition metal oxide materials. The Al foil could not be electrochemically corroded at 4.5 V by the VSE electrolyte, which is proved by the chronoamperometric test and SEM images, as shown in Supplementary Fig. 7a, b. Similar results are also observed by the CME electrolyte (Supplementary Fig. 7c, d), which indicate that LiFNFSI is stable toward Al foil in different solvents even at a high oxidation potential. The electrochemical cathodic stability of electrolytes was examined using stainless steel (Supplementary Fig. 8). The cathodic peaks of VSE electrolyte at 0.73 and 1.46 V corresponding to the reductive decomposition of solvent and LiFNFSI salt, respectively, are observed in the first cycle but almost disappear in the second cycle (Supplementary Fig. 8a, b), indicating that the anion-derived SEI layer formed from the decomposition of LiFNFSI salt in VSE electrolyte effectively avoided the side-reaction, which is different from CME and STD electrolytes (Supplementary Fig. 8c–f) due to the different Li+ solvation structures. Furthermore, the oxidation and reduction stability of the electrolyte components were estimated through simulations using density functional theory (DFT). Supplementary Fig. 9 shows the energy values of the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO) for the pure composition and complexes with Li+ ion. Although the two methods for calculating HOMO-LUMO exhibit value disparities, their trends remain consistent. For the VSE electrolyte, LiFNFSI has a lower LUMO energy and higher HOMO energy than the solvent FEBFP, indicating its stronger tendency to decompose on the surfaces of both electrodes and form an SEI/CEI layer with more inorganic components and higher thermal stability. However, LiFNFSI has a lower LUMO and HOMO energy than the solvent TEP, indicating its weaker tendency to decompose on the surfaces of the cathode. On the other hand, lower LUMO and HOMO energies of Li salt (LiPF6) compared to the solvents (EC and DMC) in the STD electrolyte suggest that LiPF6 is easier to decompose on the anode surface, while these organic solvents are prone to be oxidized on the cathode surface, forming an interphase layer with more organic components and lower thermal stability.

Solvation structure of VSE electrolyte

The Li+ solvation structures of the electrolytes were characterized using NMR and Raman spectra (Fig. 2a–c). When the lone pair electrons in anions or solvents coordinate with Li+, it shields the electronic environment of 7Li nuclei, resulting in chemical shifts in NMR spectra31. The chemical shifts of 7Li nuclei in Li salt molecules decrease in the order of STD > CME > VSE, indicating that the coordination strength between Li+ and anions or solvents follows the STD > CME > VSE trend (Fig. 2a). Figure 2b shows the Raman spectra for CME and VSE. Due to the asymmetric structure of the LiFNFSI salts, the Raman shift of the N–S bending signal deviated into a wide range of ~700–870 cm−132,33. The N–S bending signal in FNFSI anion is classified into three distinctive bands: solvent-separated ion pair (SSIP, at 736 cm−1), contact ion pair (CIP, at 773.2 cm−1), and aggregate (AGG, at 843.5 cm−1). The specific portion of these three species was calculated from the peak area and listed in Fig. 2c. CME electrolyte contains 73.5% of SSIP, 14.1% of CIP, and 12.4% of AGG, indicating that most anions are expelled from the solvation sheath due to the strong solvating power of TEP. On the contrary, VSE electrolyte contains 22.3% of SSIP, 25.1% of CIP, and 52.6% of AGG. The high AGG promotes the reduction of anion-forming LiF-rich SEI.

Fig. 2: Experimental and theoretical studies of electrolyte solvation structures.
figure 2

a Natural abundance 7Li NMR spectra of STD, CME, and VSE electrolytes. b Raman spectra of CME and VSE electrolytes. c The ratio of different solution structures in various solvents calculated from (b). df Snapshots of the MD simulation for the representative solvation structures. Colors for different elements: Li, green; S, yellow; P, light pink; O, red; N, blue; F, light blue; H, light gray; and C, dark gray. gi RDFs g(r) of Li+-F and Li+-O pairs of STD, CME, and VSE electrolytes, respectively.

The solvation structures of the three electrolytes were investigated using molecular dynamics (MD) simulations with a canonical ensemble (NVT) carried out for various systems. The initial structure of each Li salt/solvent mixture system was set up by randomly placing the numbers of Li salts and solvent molecules based on experimental densities and molar ratios (concentration). The snapshots of the three salt/solvent mixture systems from MD summations are shown in Fig. 2d–f and Supplementary Fig. 10. Li+ ions exhibit a preference for proximity to O atoms in double bonds, such as C=O or P=O due to their lower electronegativity and more attractive to Li+ ions. The MD simulation trajectories were used to calculate the radial distribution functions (RDFs) of Li+-F pairs highlighted in Fig. 2d and Li+-O highlighted in Fig.2d–f. The Li-O peaks of EC, TEP, FNFSI, and FEBFP in electrolytes correspond to the distances of 2.4, 2.5, 2.7, and 2.9 Å, respectively (Fig. 2g–i). The shortest distance of the Li+-O(EC) peak (Li+-O(EC) < Li+-O(DMC) <Li+-F(PF6)) reveals that Li+ ion is strongly solvated by EC molecule in STD, which is consistent with the solvation structure of the STD electrolyte obtained from Raman spectra in Fig. 2b, c. High Li+ ion solvation is also observed in the CME electrolyte. However, the Li+ ion is preferentially coordinated with the FNFSI anion rather than the solvent FEBFP molecular in the VSE electrolyte. The weakly solvating structures of VSE lead to the prior electrochemical reduction of LiFNFSI salt to the solvent and form inorganic SEI/CEI28,34.

Li plating/stripping in VSE electrolyte

The cycling stability of Li plating/stripping in different electrolytes was evaluated using a symmetric Li | |Li cell. We first pre-evaluated all the synthesized fluorinated phosphate ester electrolytes. Supplementary Fig. 11a shows the time-overvoltage curves of the Li||Li cells with VSE electrolyte. The Li electrodes can be stably cycled at 0.5 mA cm−2 for about 1200 h with smooth and small polarization voltages in the VSE electrolyte, which is much more stable than other single-solvent electrolytes under the same conditions (Supplementary Fig. 11b–f).

Supplementary Fig. 12 shows the Li plating/stripping profiles of the Li||Li symmetric cells using STD, CME, and VSE electrolytes at different currents, capacities, and temperatures. At a current of 0.5 mA cm−2, a capacity of 0.5 mAh cm−2, and a temperature of 30 °C (Supplementary Fig. 12a–c), the Li | |Li symmetric cell with STD electrolyte shows large voltage polarization, strong voltage fluctuation, and shorter cycle life (Supplementary Fig. 12a). The Li plating/stripping overvoltage in CME electrolytes gradually decreased and then quickly increase due to the interaction between different solvents (Supplementary Fig. 12b). In contrast, the Li plating/stripping in VSE electrolyte is stable for ~1200 h with slightly rising polarization voltages (Supplementary Fig. 12c). The higher overvoltage than those in CME and STD electrolytes at the initial cycling stage is mainly due to its relatively lower ionic conductivity and higher viscosity. At a high temperature of 50 °C, a high current of 1.0 mA cm−2, and a high capacity of 2.0 mAh cm−2 (Supplementary Fig. 12d–f), Li symmetric cells with STD and CME electrolytes show reduced life-spans of less than 300 and 500 h, respectively, but stable cycling was maintained for more than 1500 h in VSE electrolyte (Supplementary Fig. 12d–f). At a high temperature of 90 °C and 1 mA cm−2/1.0 mAh cm−2 (Supplementary Fig. 12g, h), the Li electrode shows exceptional cycle life in the VSE electrolyte, which is about ten times longer than that in the STD electrolyte.

Li plating/stripping CE in three electrolytes at a current of 0.5 mA cm−2 and capacity of 0.5 mAh cm−2 was compared using the Li | |Cu half cells. Li CE in VSE at both 30 and 50 °C can reach >99.3% using complete Li plating/stripping protocol (Fig. 3a, b) or partial Li plating/stripping (Fig. 3c). The average Li plating/stripping Coulombic efficiency (CEavg) at the partial Li utilization is only 94.8% for CME, and 89.6% for STD (Fig. 3d, e)35. The Li | |Cu cells can be charged/discharged for 300 cycles, which is much longer than those in CME and STD electrolytes. The results prove the excellent thermal stability and compatibility of VSE electrolyte to Li metal electrode.

Fig. 3: The CEs of Li cycling in different electrolytes.
figure 3

a, b CEs vs. cycle number for VSE, CME, and STD electrolytes by Li||Cu half-cells at a 30 °C and b 50 °C, respectively. ce Li CE in c STD, d CME, and e VSE electrolytes under 0.5 mA cm−2 current density at 30 °C for average cycling efficiency calculation.

The surface morphologies of Li electrodes cycled in different electrolytes at 50 °C have been examined using scanning electron microscopy (SEM). The morphology of Li deposit in the STD electrolyte is dendritic, spongy, and severely pulverized (Fig. 4a), which could result in unstable interfacial properties, short cycle life, or even short circuits. In contrast to the rode Li with different sizes observed in CME electrolyte (Fig. 4b), it is even and compact with micro-sized and rounded particle shape when using VSE electrolyte (Fig. 4c). Furthermore, the cross-section images of Li electrodes after 100 cycles also reveal that the thickness of active Li layer rises from 22 μm in VSE electrolyte to 97 μm in CME and 145 μm in STD electrolyte (Fig. 4d–f) demonstrating that the deposited Li in VSE electrolyte is much denser than those in CME and STD electrolytes, which is an agreement with the high CE.

Fig. 4: SEM images and proposed schematic of the cycled Li electrodes at 50 °C.
figure 4

af SEM images of the Li anodes cycled at 1 mA cm−2/2 mAh cm−2 for 100 times in a, d STD, b, e CME, and c, f VSE electrolytes. ac and df are the top-view and cross-sectional SEM images, respectively. g The schematic of speculated Li electrode structure evolutions during cycling in different electrolytes.

The high CE and dense deposited Li in VSE are attributed to robust SEI. As demonstrated by electrochemical impedance spectroscopy (EIS) of Li | |Li symmetric cells in Supplementary Fig. 13, after five cycles, the first semicircle in the high-frequency region, representing the interface resistance, in EIS of Li | |Li using VSE is smaller and more stable than that in CME or STD. This could be explained by the stable LiF-rich SEI structure formed in VSE electrolyte to decrease the interfacial resistance and improve the Li deposition morphology. On the other hand, the continuously reduced semicircle during cycling of Li | |Li cells using STD and CME electrolytes is associated with the increased specific area due to Li dendrite growth along with the continued formation of organic-inorganic SEI. According to the SEM observation and electrochemical properties, the Li plating/stripping mechanism in STD, CME, and VSE electrolytes are summarized in Fig. 4g. The lithiophobic LiF-rich SEI was formed in the VSE electrolyte. The LiF in SEI exhibits the highest interface energy of 73.28 meV/Å2 and the weakest bonding to the Li anode compared to other SEI components such as Li2O, Li2S, and Li2CO3, thereby helping to alleviate the stress from electrode volume expansion. Furthermore, the LiF-rich SEI exhibits high mechanical strength with a bulk modulus of 70 GPa and a γE value of 5129 eV/Å2 MPa36,37, effectively suppressing vertical Li growth into the LiF-rich SEI as Li dendrites. Additionally, the low electronic conductivity of LiF in the SEI serves to suppress side reactions of the electrolyte decomposition, even under elevated temperatures37,38.

SEI and CEI in VSE electrolyte

As mentioned above, the composition of the SEI layer plays a critical role in its structural stability and electrochemical performance. The chemical compositions of SEI on cycled Li were investigated by X-ray photoelectron spectroscopy (XPS). As shown in Supplementary Figs. 14, 15, LiF is detected on all the SEI layers of the cycled LMA. Meanwhile, different groups are also observed, such as the P−F group in STD electrolytes, and the CF2CF2 and S−F group on the SEI surface in CME and VSE electrolytes (Supplementary Fig. 14c). The LiF peak in SEI of CME and VSE electrolyte is derived from the reduction of LiFNFSI or fluorinated solvents in electrolytes (such as BTFE solvent in CME and FEBFP solvent in VSE), while S-F and CF2-CF2 are derived solely from the decomposition of LiFNFSI salts. The VSE electrolytes result in more S-F (Supplementary Fig. 14d) and total fluorinated bonds (Supplementary Fig. 14e) in SEI compared with the CME electrolytes, indicating the prone LiFNFSI salts decomposition to form LiF-rich inorganic-SEI layer in VSE electrolytes. Additionally, the CF2CF2 group can increase the exchange current due to its high dielectric constant and reduced overpotentials39, and decrease the swelling degree of SEI, thereby inhibiting the continuous consumption of electrolytes, particularly at high temperatures. In contrast, the SEI layer formed in STD electrolyte is relatively inhomogeneous and mainly composed of organics (i.e., ROCO2Li and ROLi, R=alkyl) (top row in Supplementary Fig. 14b), resulting in Li dendrite growth40. In CME electrolyte, TEP is more readily reduced than LiFNFSI, generating a large amount of P group and organics (i.e., RLi and ROLi, R=alkyl) in the formed SEI. Moreover, the SEI formed in CME electrolyte was also LiF-rich, but contained more RfOLi and RfLi (Rf partially fluorinated alkyl) (middle row in Supplementary Fig. 14b, C-O and C-F), which could increase the swelling degree of SEI. The organic components in SEI promote Li dendrite growth.

To further investigate the construction and composition of SEI in VSE electrolytes, an in-depth XPS analysis was conducted with continuous Ar gas sputtering from the surface to the depth. Supplementary Fig. 15 displays the evolution of SEI composition in the VSE electrolyte during a sputtering duration of 700 s. The main species are the inorganics (i.e., LiF, Li2S, and Li2SOx), which remain at a roughly constant high concentration through the entire SEI. The amounts of CF2CF2-based organics gradually reduced from the SEI surface to the SEI/Li interface. The organic CF2CF2-based component on the SEI out layer and inorganic LiF-rich inner layer can accommodate the large volume change during Li plating/stripping, improving the cycling stability41.

On the NCM811 cathode side, more inorganic compounds, such as LiF, existed in the CEI formed in VSE electrolyte than that in CME and STD electrolytes (Supplementary Fig. 16). The LiF-rich CEI layer shows high anodic stability, thin thickness, and good conductivity42. Similar results are also confirmed by the time-of-flight secondary ion mass spectrometry (TOF-SIMS) (Supplementary Fig. 17). Strong F signal is found within the top surface layer (10–20 nm) formed in VSE electrolyte, similar to the distribution of the S and Li elements, indicating that a thin and hybrid Li, F, S-rich CEI layer is formed on the NCM811 particle surface. However, the element content of P on the depth profile is relatively constant and low. This result proves once again that the FEBFP solvent rarely participates in the CEI film-forming process. The inorganic components in CEI mainly come from the reduction of LiFNFSI salt in VSE electrolyte. The proposed composition evolution of VSE electrolyte in the Li||NCM811 cell is illustrated in Supplementary Fig. 18.

Performance of Li metal full-cell with VSE electrolyte

The VSE electrolyte was evaluated in Li metal cells using three cathodes: high capacity NCM811, high voltage (5.0 V) LiNi0.5Mn1.5O4 (LNMO), and thermal robust LiFePO4 cathodes. Typically, the Li||NCM811 cell shows almost no capacity decay at 0.2 C for 500 cycles at 30 °C up to 4.3 V (Fig. 5a). Its capacity retention reaches 92% over 500 cycles at a high temperature of 50 °C (Fig. 5b). In contrast, the capacity declines quickly after 200 cycles with CME and STD electrolytes at both 30 and 50 °C. Notably, the VSE electrolyte enables high voltage with an upper limit rise to 4.5 V. In situ XRD characterization demonstrated that the NCM811 electrode undergoes a reversible phase change in the voltage range from 2.7 to 4.5 V (Supplementary Fig. 19). The 4.5 V Li||NCM811 cells with VSE can be stably cycled for 300 cycles at 50 °C, delivering intriguing high capacity of 210 mAh g−1 (Fig. 5c), higher initial and average CEs than that in CME and STD (Supplementary Fig. 20a and Fig. 5c) and no discharge medium voltages decline (Supplementary Fig. 20b). To assess practical feasibility, the Li||NCM811 cell with VSE was tested at 0.5 and 30 °C as demonstrated in Supplementary Fig. 21a. Compared with the Li||NCM811 cell that operated at 50 °C, the good electrochemical reversibility is maintained. The high-temperature performance of the Li | |NCM811 cells in VSE at 50 °C was further compared with that in the current state-of-the-art LHCE electrolyte (LiPF6/FEC-FEMC-HFE)8. The LHCE electrolyte only allows Li | |NCM811 cells to achieve 74% capacity retention after 250 cycles at 0.5 and at 50 °C (Supplementary Fig. 21b), which is significantly lower than the capacity retention of 91% for VSE electrolyte. The impact of designed LiFNFSI salt on full-cell performance was investigated by comparing the 0.8 M LiFNFSI/FEBFP electrolyte to the 0.8 M LiTFSI/FEBFP electrolyte. As shown in Supplementary Fig. 21c–f, the Li | |NCM811 cell exhibits poor cycle life in the 0.8 M LiTFSI/FEBFP electrolyte due to the corrosive effect of LiTFSI on the aluminum current collector at a high voltage.

Fig. 5: Electrochemical performances of the Li||NCM811 and Li||LiFePO4 cells.
figure 5

ac Cycle performance of Li||NCM811 cells under different test conditions. df Fabrication of LiNCM811 pouch cell with VSE electrolyte and its practical test. d Optical image. e The discharge curve of the LiNCM811 pouch cell. The inset in (e) is the optical image of the pouch cell’s weight. f Cycle performance of Li||NCM811 pouch cells using STD and VSE electrolyte. gi Electrochemical performance of Li||LiFePO4 cells using different electrolytes at 90 °C. g Initial Columbic efficiencies at 0.5 C. h Cycle performance. i Rate performance. All the related cells would be activated by a small current (0.1 C) in the initial two cycles. The areal mass loadings of cathodes for coin cells and pouch cells are 6–7 and ~28 mg cm−2, respectively.

In VSE electrolytes, the formation of LiF-rich CEI prevents cracking in NCM811 due to the weak bonding between LiF and NCM811 and the high mechanical strength of LiF, which helps alleviate stress from the volume changes of NCM811 cathode. Additionally, the high anti-oxidation stability of LiF CEI prevents the side reactions of the liquid electrolytes. Compared with the pristine NCM811 cathode (Supplementary Fig. 22a–c), the NCM811 particles cracked after cycling in STD and CME electrolytes (Supplementary Fig. 22d, g, e, h) due to the unstable organic-rich CEI at a high voltage of 4.5 V and the strains caused by the strong bonding of organic CEI to NCM811, resulting in the continued side reaction8,43,44,45,46. The side reaction between the electrolyte and cracked NCM811 accelerates the dissolution of transmission metal in the electrolyte, oxygen evolution and CEI growth will degrade the electronic contact of the active material and further accelerates capacity decay. In sharp contrast, the VSE electrolyte maintains intact particles even after 200 deep cycles up to 4.5 V (Supplementary Fig. 22f, i). The XRD patterns in Supplementary Fig. 23 reveal that the crystal structure of NCM811 material is degraded after cycled in STD electrolyte at 50 °C 200 times, along with shifting, broadening, and disappearance of some peaks. However, no obvious difference is observed for the NCM811 electrode before and after cycling in the VSE electrolyte, confirming the good structure stability in the VSE electrolyte. Based on the above-mentioned analysis, the proposed electrolyte function and electrode stabilization mechanism can be demonstrated in Supplementary Fig. 2443,44,45. The high chemical/electrochemical stability of the VSE electrolyte improves the interfacial property, maintains the structure and morphology of NCM811, and finally leads to excellent electrochemical performance. The high anodic stability potential of 5.6 V can support high voltage 5 V LiNi0.5Mn1.5O4 (denoted as LNMO) cathode, as demonstrated by the amazing capacity retention of 93.0% after 200 cycles with an average CE of 99.5% (Supplementary Fig. 25), versus capacity retention of 48.5% after 120 cycles for STD electrolyte. All reported electrolytes stabilizing Li anode, including high-concentration ether electrolytes, local high-concentration ether electrolytes, and partially fluorinated ether electrolytes, can only support 4.3 V NCM811 cathode, but not 5.0 V LiNi0.5Mn1.5O4 cathodes. To demonstrate the practical application of VSE electrolyte for commercialization, 3.5-Ah-grade Li | |NCM811 pouch cells were fabricated with a low n/p ratio of 1.5 and lean electrolyte of 1.7 g Ah−1cathode (Fig. 5d and Supplementary Table 4). The pouch cell using VSE electrolyte could provide an energy density of 531 Wh kg-1 (Fig. 5e) and good cycle performance (Fig. 5f).

Intrinsic safety of full-cells with VSE electrolyte

The VSE electrolyte with nonflammable and no flash point can operate at an extremely high temperature up to 90 °C. At such a high temperature, high power is easily achieved. NCM811 has a low thermostability due to accelerated dissolution of transmission metal and evolution of oxygen at a high temperature. Therefore, LiFePO4 cathodes were selected to match the Li anode to demonstrate the high-temperature operation of the VSE electrolyte at 90 °C (Fig. 5g–i). The overvoltage between the charge and discharge platform is below 50 mV at 90 °C (Fig. 5g), which improves the energy conversion efficiency. In VSE electrolyte, almost no capacity decay is observed for Li||LiFePO4 cells up to 200 cycles at 90 °C (Fig. 5h). In striking contrast, the cell using STD electrolyte fails rapidly within the initial 20 cycles with a capacity retention of 42.8%. The fast capacity fade of Li | |LiFePO4 in STD electrolytes is mainly due to the dissolution of organic-rich SEI/CEI and the decomposition of electrolytes during charging. The use of the VSE electrolyte leads to not only excellent cyclability but also amazing rate performance at high temperatures. Little capacity changes are shown in capacity from 0.5 to 5 C due to the tremendous improvement in the cell kinetics at 90 °C (Fig. 5i), demonstrating the capability for high-power devices.

The high-temperature thermal stability of various electrolytes in combination with cycled Li anode and fully charged NCM811 cathode powders was investigated using differential scanning calorimetry (DSC). The results are summarized and compared in Supplementary Figs. S26, S27. Heat flow curves exhibit two critical temperatures: the onset temperature, indicating the start of exothermic heat generation, and the peak temperature, marking the moment when the intense thermal runaway reaction occurs. Due to the high reactivity of cycled Li and fully charged NCM811, all three electrolytes showed apparent exothermal peaks. However, the VSE electrolyte demonstrated exceptional thermal stability with the highest heat release peak temperatures and the smallest integrated heat flow for both cathode and anode. This value is also far superior to the advanced hydrofluoroethers FDMB-based electrolyte (e.g., NCM cathode + electrolyte, 751 vs ~1300 J g−1)47. The differences mainly stem from the thermal stability of the electrolyte itself and the side reactions occurring at high temperatures with highly reactive electrodes. The SEI/CEI formed in VSE electrolyte is composed of more inorganic species like LiF and small amounts CF2CF2-based organics, which could alleviate the side reactions and exotherms between electrolyte and electrodes. Moreover, VSE itself has high thermal stability. These factors contribute to the excellent thermodynamic testing results.

Furthermore, due to the unique synergistic effect of fluorine and phosphate function groups of the VSE solvent, the nonflammable and no flash point VSE electrolyte endows Li||NCM811 cell with intrinsic high safety under extreme conditions. As shown in Fig. 6a, c and Supplementary Movie 1, the fully charged pouch cell using commercial STD electrolyte underwent drastic explosion and combustion during the nail penetration test. Moreover, Ouyang et al. reported the EC solvent in STD electrolytes could act as an oxygen consumption protagonist and trigger thermal runaway in the Ni-rich battery system48. In contrast, the VSE-based pouch cell can keep a good shape without any flame and expansion after the test (Fig. 6b, d and Supplementary Movie 2). To better understand the thermal runaway features, the accelerating rate calorimetry (ARC) tests were conducted to evaluate the thermal safety of the fully charged 3.5-Ah Li | |NCM811 pouch cell, in which T1, T2, and T3 are marked in Fig. 6e, f. T1 is defined as the point at which Dt/dt = 0.02 °C min−1, indicating the onset of self-exothermic reactions mainly due to the decomposition of SEI and electrolyte reduction. The following self-progressive temperature increase is driven by a chain of side reactions toward T2. T2 is defined as the point at which Dt/dt = 1 °C min−1, indicating that the cell enters an uncontrollable state, with underlying strong exothermic reactions. T3 is the maximum temperature of the ARC tracking point which appeared in the stage of severe thermal runaway. Both the T1 and T2 temperatures of VSE-cell are about 15 °C higher than those of STD-cell, but the T3 temperature has decreased by 700 °C, thus significantly enhancing the thermal safety of the cell.

Fig. 6: Safety test of the LiNCM811 pouch cells.
figure 6

ad Nail penetration tests of fully charged LiNCM811 pouch cells with a, c STD and b, d VSE electrolytes. a, b The changes of voltage and temperature during nail penetration tests. c, d The optical images of pouch cells after nail penetration. e, f Thermal runaway temperature plots of the fully charged LiNCM811 pouch cells with STD and VSE electrolyte measured by accelerating rate calorimetry (ARC) tests. e STD electrolyte; f VSE electrolyte.

Compared with the current commercially available carbonate solvents, which suffer from serious gas release (Supplementary Fig. 28a) and severe safety issues especially at elevated temperatures49, the LiF-rich and CF2CF2-reinforced SEI/CEI in situ formed in VSE electrolyte can eliminate side reactions and stop continuous electrolyte decomposition thus no gas was released in LMBs even after cycled at a high temperature of 90 °C (Supplementary Fig. 28b).

Supplementary Fig. 29 compares and summarizes the physical and electrochemical properties of the studied STD, CME, and VSE electrolytes and the corresponding Li metal cells. To the best of our knowledge, VSE is the only single-solvent electrolyte that can stably operate 5.0 V Li||LNMO cells at 50 °C13,14,15, Li | |LiFePO4 at 90 °C and a high rate of 5 C, and 4.5 V Li||NCM811 at 25–50 °C. Moreover, the use of VSE realizes the intrinsic high safety even in abuse, which is the ideal power source for aviation and electric vehicles.

Discussion

In conclusion, we have designed a unique single-solvent electrolyte based on the high-fluorine content of LiFNFSI salt and an ether-functionalized fluorophosphate VSE solvent. Benefiting from the multifunctional function groups, including flame-retardant phosphate, high Li-stabilizing ethers, and nonflammable fluorine, the VSE solvent exhibits comprehensive properties, such as high boiling point, noninflammability, no flash point, and admirable hydrophobicity, which lead to the intrinsic high safety of LMBs. Additionally, LiFNFSI is more prone to electrochemical reaction on the electrode surface than FEBFP solvent in VSE electrolyte, leading to the anion-derived and LiF-rich SEI and CEI with high compactness and thermostability. These unique advantages result in the extraordinary electrochemical performances of Li||NCM811 cell, including 92% capacity retention after 500 cycles and ultrahigh energy density of 531 Wh kg−1 for 3.5-Ah-grade pouch cell. Furthermore, the VSE electrolyte is also well adaptable to 5V-class LNMO and the commercial LiFePO4 cathode even at 90 °C. Most strikingly, the pouch cell that uses our electrolytes passed the nail penetration test without any flame and expansion and decreased the maximum thermal runaway temperature by 700 °C during the ARC tests. We hope that this single-solvent EC-free electrolyte design concept and the exceptional results will greatly promote the development and practical application of LMBs with high safety and high energy density.

Methods

Preparation of the electrolytes

Lithium hexafluorophosphate (LiPF6), dimethyl carbonate (DMC), ethylene carbonate triethyl phosphate (TEP), and bis(2,2,2-trifluoroethyl) ether (BTFE) (all in battery-grade purity) were obtained from Capchem. or Shangfluoro. and used as-received. The designed phosphate solvents and LiFNFSI salt were synthesized and purified. Before preparing the electrolytes, these solvents were dried over an activated molecular sieve (4 Å). The electrolytes of 1 M LiPF6 in EC-DMC (1:1, by volume) and LiFNFSI with different concentrations in the single or mixed solvents were obtained by blending adequate amounts of salts and solvents in a PP flask, and further stirring for 12 h at room temperature. A Karl Fisher titrator (Metrohm 831 Coulometer) was utilized to measure the water content of the electrolytes, which was lower than 10 ppm in all prepared electrolytes. The experimental work was carried out, and the materials were stored in an Ar-filled dry glove box (MIKROUNA) containing less than 0.5 ppm H2O and O2.

Characterization of the electrolytes

For the measurements of impurities that existed in the electrolyte, the water content was determined by Karl-Fischer titration (Metrohm KF 831 Coulometer), and the content of F or Cl was measured by ion chromatography (DIONEX ICS-1000). The electrolyte viscosity was measured on a programmable rheometer (DV-III Ultra, Brookfield) at room temperature. The ionic conductivity of the electrolyte solutions was measured by a conductivity meter (DDS-301A, Shanghai INESA Scientific Instrument Co., Ltd.). The electrolyte flammability was tested by burning an electrolyte-soaked PE Celgard separator. The thermogravimetric tests were carried out using a Thermogravimetric Analyzer (TA Instruments, Q5000IR) with a heating rate of 10 °C min−1 under Ar protection. The flash point was investigated using a SETA closed-cup flash-point tester following the ASTM D3278 standard method. The Raman spectrometer was LabRAM HR800 (HORIBA Jobin Yvon Ltd.) with an exciting laser of 633 nm.

Assembly of cells

The as-prepared electrodes were assembled into 2032-coin cells of a two-electrode configuration with a Celgard polyethylene separator in an Ar-filled dry glove box (MIKROUNA) containing less than 0.5 ppm water and O2. The electrolyte contained LiPF6 or LiFNFSI salt in the commercial carbonate solvents or synthesized phosphate solvents. The pouch cells were fabricated in Shanghai Institute of Space Power-Sources, and the used NCM811 cathode and Li metal anode were purchased from Canrd Technology Co., Ltd.

Electrochemical measurements

Cyclic voltammetric studies of the electrolyte solutions were conducted in a half-cell using a CHI760A electrochemical workstation. Stainless steel or Al foil served as working electrodes, and Li metal foil was used as both the counter and reference electrodes. The Li | |Cu half-cells were used to investigate the cycling stability of Li plating/stripping in different electrolytes. The Coulombic efficiency (CE) of Li plating and stripping was calculated from the ratio of the Li removed from the Cu substrate to that deposited in the same cycle. The detailed procedure was as follows: ten pre-cycles between 0 and 1 V at 0.1 mA cm−2 were initialized to clean the Cu electrode surface, and then cycling was done by depositing 0.5 mAh cm−2 of Li onto the Cu electrode followed by stripping to 1 V under 0.5 mA cm−2 current density. The average CE is calculated by dividing the total stripping capacity by the total deposition capacity after the formation cycle. For the average CE test21,35, a standard protocol was followed: (1) perform one initial formation cycle with Li deposition of 5 mAh cm−2 on Cu under 0.5 mA cm−2 current density and stripping to 1 V; (2) deposit 5 mAh cm−2 Li on Cu under 0.5 mA cm−2 as a Li reservoir; (3) repeatedly strip/deposit Li of 1 mAh cm−2 under 0.5 mA cm−2 for 10 cycles; (4) strip all Li to 1 V. The Li-ion transference number (tLi+) was measured and calculated in accordance with the Bruce-Vincent-Evans equation50.