Lithium Bis(fluorosulfonyl)imide (LiFSI) (≥ 99.8%) salt was purchased from Ningde Guotai Huarong New Material Co., Ltd and dried at 80 ℃ overnight in a vacuum chamber. Trimethyl orthoformate (TMOF) (≥ 98%) and Triethyl orthoformate (TEOF) (≥ 98%) were purchased from Aladdin. 1,1,2,2-Tetrafluoroethyl-2,2,3,3-tetrafluoropropylether (TTE) (≥ 99.5%) was purchased from Shang Fluoro Co., Ltd. Before use, TMOF, TEOF, and TTE solvents were dried with activated molecule sieves for 2 days. The working electrolytes used in this work were LiFSI/TMOF/TTE and LiFSI/TEOF/TTE (1:1.2:3, by molar ratio).

All electrochemical tests in this work were carried out using 2032-type coin cells, which were assembled in the Ar-filled glove box (< 0.1 ppm H₂O and < 0.1 ppm O₂). Li||Cu cells were assembled with one piece of lithium metal foil (450 μm, Φ = 14 mm) and one piece of copper foil (25 μm, Φ = 19.1 mm). Li||Li cells were assembled with two pieces of lithium metal foils. For the long-term cycling of Li||Cu cells, Li metal was firstly deposited on the Cu substrate at a current density of 0.5 mA cm⁻² or 1 mA cm⁻² for 1 mAh cm⁻², then stripped to 1 V. For the preparation of NMC811 electrode, NMC811 cathode materials, super P and poly(vinylidene fluoride) (PVDF) were mixed with a mass ratio of 85:7.5:7.5 using N-methyl-2-pyrrolidone (NMP) as a solvent, the as-prepared mixture was stirred to get a slurry. The slurry was further coated on the carbon-coated Al foil and dried at 60 ℃ overnight under vacuum, then cut into pieces (Φ = 14 mm). Before use, the NMC811 cathode was dried at 110 ℃ overnight and transferred into the glovebox. The loading of cathode material in this work was ~ 3 mAh cm⁻². In Li||NMC811 cells, one piece of thin lithium metal foil (50 μm, Φ = 16 mm) and one piece of NMC811 electrode were used. In Li||LCO cells, one piece of thin lithium metal foil (450 μm, Φ = 15.6 mm) and one piece of LCO electrode were used. In each cell, Celgard 2325 (25 μm, Φ = 19.1 mm) was used as separator and 80 µL of electrolyte was applied. The cells were tested on a Neware battery testing system (CT-3000n). Linear sweep voltammetry (LSV), cyclic voltammetry (CV) and electrochemical impedance spectroscopy (EIS) measurements were tested in Li||Carbon paper cells, Li||Cu cells, and Li||Li cells, respectively. Both LSV, CV, and EIS measurements were conducted on a CHI760e electrochemical station.

The Li deposition morphologies were observed using scanning electron microscopy (SEM, FEI Apreo) operated with a working voltage of 2 kV. The modulus of deposited Li was tested by atomic force microscopy (AFM, Bruker Dimension Icon). The modulus measurements were operated in the peak-force quantitative nano-mechanical mapping (QNM) mode. Before SEM and AFM measurements, the samples were washed with DME for several times to clean the residues and dried naturally. The surface chemistry of the Li deposits was characterized by X-ray photoelectron spectroscopy (XPS, KRATOS AXIS SUPRA⁺). The samples were washed with DME beforehand, and loaded on the XPS transfer holder to avoid exposure to air environment. The depth profiling sputtering was carried out with an Ar-ion gun (5 kV, 2 mm × 2 mm). The sputtering time was 4 min for Li metal samples and 2 min for NMC811 samples, respectively. Raman microscope (Raman, Renishaw inVia) and nuclear magnetic resonance spectrometer (NMR, JNM-ECZ600R/S1) were used to characterize the solvation behaviors in the two electrolytes.

Density function theory (DFT) calculations were performed to optimize molecular geometries of TEOF and TMOF molecules using the Jaguar 8.8 software package at the level of B3LYP level with the 6-311 + G(d,p) basis sets [46]. Dipole moment and electrostatic potential (ESP) were calculated using at B3LYP/6-311 + G(d) level, which has been reported to well predict the formation charge. The binding energy (△E_binding) was calculated as follows:

Where E(sol) and E(sol-Li⁺) are the potential energies of the solvent and Li⁺ together with solvent, respectively. E(Li⁺) represents the energy of the Li⁺.

The dielectric constant of electrolytes, solvation structures and radial distribution function (RDF) were obtained from classical molecular dynamics (cMD) simulations. All cMD simulations were conducted with the GROMACS 2020 program [47]. The simulation of TMOF consists of 100 LiFSI, 120 TTE, and 360 TMOF molecules. The size of the simulation cell of the TMOF system is 109.39 nm³. The simulation of TEOF consists of 100 LiFSI, 120 TTE, and 360 TEOF molecules. The size of the simulation cell of the TEOF system is 142.00 nm³. The electrolyte concentrations were simulated in consistent with experiments. The systems were firstly equilibrated under the isothermal-isobaric (NPT) ensemble. In 2 ns NPT simulations, the temperature was kept at 300 K using a V-rescale thermostat with a relaxation time of 0.1 ps and the pressure was kept at 1 bar using a Berendsen barostat with a relaxation time of 0.5 ps. Thereafter, the systems were simulated in the canonical (NVT) ensemble for 1 ns. In 1 ns NVT simulations, the temperature was kept at 300 K using a V-rescale thermostat with a relaxation time of 0.1 ps.

We selected the commercially available solvent, trimethyl orthoformate (TMOF), as a Foundational solvent, owing to its distinct open-chain structure, as depicted in Fig. 1a. However, the multiple coordination sites result in strong solvation ability of TMOF, which may generate solvent-dominated solvation structures that have poor compatibility with Li metal anode. We expect that increasing the steric hindrance by lengthening the terminal alkyl group could decrease the solvation ability of TMOF. Under this design principle, we introduced triethyl orthoformate (TEOF) as a weakly solvating solvent. Both TMOF and TEOF molecules have three O atoms that can serve as the Li⁺ binding sites, leading to the highly centralized negative charge density center indicated from the DFT-derived electrostatic potentials (ESPs) (Fig. 1b, c). Compared with TMOF, the larger ethoxy groups on TEOF provide a stronger steric hindrance. The steric effects of terminal alkyl groups could hinder the interactions between TEOF and Li⁺, resulting in a lower solvation ability of TEOF than TMOF [30,33]. In order to substantiate this hypothesis, we have constructed the LHCEs utilizing TMOF and TEOF as equivalent solvents, selecting LiFSI as the salt, and employing non-solvation solvent TTE as the diluent.

Molecular dynamics (MD) simulations and Raman spectroscopy were conducted to elucidate the Li⁺ solvation structures in both electrolytes. Diverse solvation structures in the electrolytes can be identified based on the differing coordination states of FSI⁻ anions. These structures can be categorized into separated solvated ion pairs (SSIP), contact ion pairs (CIP), and aggregates (AGG), as illustrated in Fig. S1. In both electrolytes, a typical AGG structure consists of two anions and one solvent molecule directly coordinating with one Li⁺. As depicted in Fig. 2a, b, the increased steric hindrance affected the interactions between Li⁺ and FSI⁻ anions in two electrolytes, which aligns well with our electrolyte design principles. The Raman signals detected in the 1200-1250 cm⁻¹ range correspond to the vibrations of the S = O stretching mode of FSI⁻ anions [48]. Notably, both LiFSI-1.2TMOF-3TTE and LiFSI-1.2TEOF-3TTE electrolytes manifest an overlapping peak within this range, which can be further vested into three characteristic peaks of SSIP, CIP, and AGG (Fig. S2). A noteworthy contrast in the solvation structures is evident when comparing the LiFSI-1.2TMOF-3TTE and LiFSI-1.2TEOF-3TTE electrolytes. The latter exhibits an increased intensity of AGG and a simultaneous decrease in SSIP intensity (Fig. 2c), suggesting that more FSI⁻ anions are incorporated into the solvation sheaths. Statistical calculations further corroborate this observation, indicating that the LiFSI-1.2TEOF-3TTE electrolyte (~ 78%) has a broader distribution of coordinated anions compared to the LiFSI-1.2TMOF-3TTE (~ 73%) (Fig. 2d). This divergence in solvation structures between the two electrolytes can be attributed to the molecular differences between TMOF and TEOF. MD simulations are carried out for a more proFound understanding of these solvation structures (Fig. 2e-f). For the LiFSI-1.2TMOF-3TTE electrolyte, the TMOF solvent primarily resides in the first solvation sheath of Li⁺ (Fig. 2g, Fig. S3). Conversely, the LiFSI-1.2TEOF-3TTE electrolyte shows an enhanced signal of Li⁺-O(FSI⁻) pairs in the first solvation sheaths, indicative of increased interactions between Li⁺ and anions. These anion-dominant solvation structures in LiFSI-1.2TEOF-3TTE may promote a prior decomposition of anions, potentially leading to the formation of an inorganic-rich SEI.

The Li deposition morphologies in these two electrolytes were observed by scanning electron microscopy (SEM). At a deposition capacity of 0.5 mAh cm⁻², the LiFSI-1.2TMOF-3TTE electrolyte exhibits the growth of dendritic Li (Fig. 3a, Fig. S4a-c). With a deposition capacity of 1 mAh cm⁻², dendritic Li with lengths ranging from 5 to 10 μm can be observed (Fig. 3b, Fig. S4d-i). As shown in the cross-sectional SEM images, the Li deposits exhibit loose structures, with a thickness of ~ 12 μm (Fig. 3c). The dendritic Li and loose structure with high specific surface areas may cause serious parasitic reactions, accelerating the consumption of electrolytes and decreasing the cycling stability of Li metal anode. In contrast, the Li deposits in LiFSI-1.2TEOF-3TTE show closely packed and dendrite-free morphologies (Fig. 3d, e, Fig. S5). With a deposition of 1 mAh cm⁻², the deposited Li in LiFSI-1.2TEOF-3TTE exhibits a thickness of ~ 7.2 μm (Fig. 3f), indicating a better compatibility between the electrolyte and Li metal anode. Atomic force microscopy (AFM) characterizations were further conducted to investigate the mechanical properties of SEI formed on the Li metal surface. As shown by the quantitative nanomechanical mapping-AFM images, the modulus distribution of the SEI formed in LiFSI-1.2TMOF-3TTE is inhomogeneous (Fig. 3g), which potentially leads to Li dendrite growth. By contrast, the SEI formed in LiFSI-1.2TEOF-3TTE show a smooth and uniform distribution of modulus (Fig. 3h), leading to an average modulus of ~ 6.28 GPa, which is much higher than that formed in LiFSI-1.2TMOF-3TTE (~ 4.39 GPa, Fig. 3i). The higher modulus of SEI formed in LiFSI-1.2TEOF-3TTE verifies its inorganic-rich feature, which is beneficial in suppressing the Li dendrite growth and improving the reversibility of Li metal anode [49].

To evaluate the compatibility of two electrolytes with Li metal anode, Li||Cu cells were assembled to determine the Coulombic efficiencies (CEs) of Li plating/stripping processes. As shown in Fig. 4a, the LiFSI-1.2TMOF-3TTE electrolyte exhibits an average CE of 99.1% for 380 cycles at 0.5 mA cm⁻² and 1 mAh cm⁻². However, the CEs after 380 cycles show significant fluctuations (Fig. S6a). In contrast, the Li||Cu cell cycled in LiFSI-1.2TEOF-3TTE shows improved cycling stability, with an average CE of 99.1% for 450 cycles without short circuits (Fig. S6b). Furthermore, when cycled at 1 mA cm⁻² and 1 mAh cm⁻², the LiFSI-1.2TEOF-3TTE electrolyte also exhibits an enhanced cycling stability of Li||Cu cells than LiFSI-1.2TMOF-3TTE (Fig. 4b, Fig. S7). The former can enable stable cycling of Li||Cu cells for 450 cycles while only 320 cycles for the latter. Although the two electrolytes show similar values of CEs (Fig. S8), the improved kinetics of Li plating/stripping processes and the enhanced lifespan of long-term cycling both verify the higher compatibility of LiFSI-1.2TEOF-3TTE electrolyte with Li metal anode (Fig. S9). Furthermore, electrochemical impedance spectra (EIS) measurements were conducted in Li||Li symmetric cells to compare the electrochemical kinetics of the Li metal anode in two electrolytes. The EIS results of Li||Li cells after 1 cycle and 10 cycles were shown in Fig. 4c and d, respectively. Two semicircles can be observed from the EIS spectra, where the semicircle at high frequency can be assigned to the SEI resistance (R_SEI) and the semicircle at middle frequency represents the charge transfer resistance (R_ct). After 1 cycle, the R_SEI and R_ct of the Li||Li cell cycled in LiFSI-1.2TEOF-3TTE (8 Ω and 18 Ω) were lower than that cycled in LiFSI-1.2TMOF-3TTE (12 Ω and 20 Ω). The R_SEI of Li||Li cell cycled in LiFSI-1.2TEOF-3TTE decreased to only 3 Ω after 10 cycles, while the LiFSI-1.2TMOF-3TTE shows a much higher value of R_SEI (18 Ω), which suggests that the former has better electrochemical kinetics [50].

To compare the high-voltage tolerance of these two electrolytes, linear sweep voltammetry (LSV) measurements were performed in Li||carbon paper cells to investigate the oxidation stability. Compared with LiFSI-1.2TMOF-3TTE, the LiFSI-1.2TEOF-3TTE electrolyte exhibits a broadened electrochemical stability window (> 4.5 V vs. Li/Li⁺) (Fig. S10), indicating its enhanced anti-oxidation ability. Furthermore, Li (50 μm)||NMC811 (3 mAh cm⁻²) full cells were assembled to evaluate the cycling performances under high-voltage conditions. The cells were tested within the voltage range between 2.6 and 4.3 V, with a charge rate of 0.2 C and a discharge rate of 0.33 C after three formation cycles of 0.1 C. As shown in Fig. 5, the Li||NMC811 cell cycled in LiFSI-1.2TMOF-3TTE released an initial discharge capacity of ~ 175 mAh g⁻¹, which rapidly decreased to 150 mAh g⁻¹ after 70 cycles, followed by a drop almost to no capacity release after 120 cycles (Fig. 5c). By contrast, the LiFSI-1.2TEOF-3TTE electrolyte exhibits improved cycling stability of Li||NMC811 cells. With a high initial discharge capacity of ~ 182 mAh g⁻¹, the cell remains a stable discharge capacity release of ~ 180 mAh g⁻¹ in the following cycles, with a capacity retention of ~ 95.6% after 150 cycles (Fig. 5b). More importantly, the cell cycled in LiFSI-1.2TMOF-3TTE exhibits an obvious reduction of the CE after 90 cycles, which indicates serious side reactions, while the CE of cell cycled in LiFSI-1.2TEOF-3TTE maintains stable during cycling. In addition, the LiFSI-1.2TEOF-3TTE electrolyte also shows improved cycling stability of Li (450 μm)||LCO (1.4 mAh cm⁻²) cells cycled within 2.7-4.4 V (Fig. S11). These results suggest that the anion-dominant solvation structures in the LiFSI-1.2TEOF-3TTE electrolyte show improved compatibility with the high-voltage cathodes compared to LiFSI-1.2TMOF-3TTE.

SEM and X-ray photoelectron spectroscopy (XPS) characterizations were conducted to investigate the morphology evolution and interfacial chemistry of the Li metal anode cycled in Li||NMC811 cells after 50 cycles. The cycled Li metal in the LiFSI-1.2TMOF-3TTE electrolyte presents extensive cracks (Fig. 6a, Fig. S12), which results in the continuous penetration of electrolyte to occur in continuous side reactions with fresh Li metal. By contrast, the Li metal anode cycled in LiFSI-1.2TEOF-3TTE exhibits a uniform and smooth surface without any cracks (Fig. 6b). The distinct morphologies of the cycled Li metal origin from the different properties of SEI, which were regulated by the different solvation structures in two electrolytes [51,52,53]. XPS characterizations were further applied to investigate the compositions of SEI. As shown in Fig. 6c, although the similar atomic ratio of C element can be observed in the outer layer of SEI formed in two electrolytes (30% for LiFSI-1.2TMOF-3TTE, 31% for LiFSI-1.2TEOF-3TTE), after argon ion cluster sputtering for 4 min, more content of C can be detected in the inner layer of SEI formed in LiFSI-1.2TMOF-3TTE (21%) than LiFSI-1.2TEOF-3TTE (13%). Furthermore, apparent signals of C-C (~ 284.8 eV), C-O (~ 286 eV), C = O (~ 288.5 eV), and CO₃²⁻ (~ 289.8 eV) can be observed in the SEI formed in two electrolytes (Fig. 6d h). The increased content of C-containing organic species in the SEI formed in LiFSI-1.2TMOF-3TTE indicates the major contribution of solvent molecules on interfacial reactions, which matches well with the results of solvation structures. More importantly, the Li metal anode cycled in the LiFSI-1.2TEOF-3TTE (7% and 14%) electrolyte shows an increase in the atomic ratio of F element in the outer and inner layer of SEI than that cycled in LiFSI-1.2TMOF-3TTE (4% and 8%). In the F 1s spectra, higher intensity of LiF signal (~ 684.7 eV) and S-F species (~ 687.8 eV) can be detected in the SEI formed in LiFSI-1.2TEOF-3TTE (Fig. 6e and i), which matches with the trend of LiF signal in the Li 1s spectra (Fig. 6f and j). After 4 min sputtering, the Li₂O (~ 528.2 eV) peak intensity in the O 1s spectra in LiFSI-1.2TEFO-3TTE is slightly higher than that in LiFSI-1.2TMOF-3TTE (Fig. 6g and k). The enrichments of LiF and Li₂O are critical to improving the compatibility of LiFSI-1.2TEOF-3TTE towards the Li metal anode. In addition, the interfacial chemistry of cathode cycled in Li||NMC811 cells after 50 cycles were also characterized (Figs. S13, S14). The cathode cycled in LiFSI-1.2TEOF-3TTE shows an increased peak intensity of the LiF signal, which improves the cycling stability of the high-voltage NMC cathode. Combined with the steric hindrance effect, we Found that the anion-dominant solvation structures in the LiFSI-1.2TEOF-3TTE electrolyte contribute to the formation of inorganic-rich SEI and CEI, which are beneficial in improving the cycling stability of high-voltage Li||NMC811 cells.