Dihydrogen-bonding interactions in ether-based electrolytes to enable high-voltage lithium metal batteries Authors: Meinan Zhao, Kaixiang Ren, Liang Li, Qi Yu, Xinchun Song, Shilin Wu, Zihan Xu, Haiwen Li, Zhipeng Jiang, Fei Wang, Yongtao Li Published: 2026 ABSTRACT AbstractHigh-voltage lithium metal batteries require electrolytes that simultaneously combine oxidative stability with Li metal compatibility, posing a major challenge for conventional ether-based systems, which are typically limited to voltages below 4.0 V. Although conventional electrolyte engineering has been widely employed to enhance oxidative stability, they often compromise ionic conductivity or require complex synthetic routes. Herein, we propose a strategy based on dihydrogen-bonding interactions by introducing 0.05 M LiBH₄ into conventional ether-based electrolytes to construct a dihydrogen-bonded electrolyte. The hydridic H− in BH₄− interacts with the active Hδ⁺ atoms of 1,2-dimethoxyethane to form dihydrogen bonds, thereby weakening the Li⁺–solvent interaction, accelerating Li⁺ de-solvation, and promoting uniform Li deposition. Simultaneously, these dihydrogen-bonding interactions shield the active Hδ⁺ sites of the solvent within the positive electrode interface, thereby significantly suppressing the oxidative decomposition of 1,2-dimethoxyethane. As a result, the oxidative stability of the electrolyte is extended to 5.54 V without compromising ionic conductivity (>16 mS cm−1/30 °C). Lithium metal full cells using this electrolyte exhibit stable cycling at 4.5 V. This study provides a promising pathway for the design of high-voltage ether-based electrolytes.Subject terms: Batteries, Electrochemistry, Batteries Introduction High-voltage lithium metal batteries (LMBs) represent a promising avenue for developing next-generation energy storage technologies with significantly enhanced energy density1,2. The combination of lithium metal anodes (LMAs)—with a high theoretical specific capacity of 3860 mAh g−1—and high-voltage positive electrode materials such as LiNi₀.₈Co₀.₁Mn₀.₁O₂ (NCM811) and LiCoO₂ (LCO) can theoretically deliver energy densities >500 Wh kg−1, far surpassing those of conventional lithium-ion batteries (LIBs)3,4. This renders high-voltage LMBs attractive for a wide range of applications, including electric vehicles, portable electronics, and grid-scale energy storage5,6. However, their practical deployment is hindered by key challenges, including Li dendrite growth, interfacial instability, and electrolyte degradation at elevated voltages—particularly above 4.3 V (potential vs. Li+/Li)7,8. These issues severely compromise the cycling stability and safety of LMBs9,10. To address these limitations, the development of high-voltage electrolytes has become a critical research focus11–13. Conventional carbonate-based electrolytes offer good oxidative stability but tend to form unstable solid electrolyte interphases (SEIs) on LMAs, leading to dendritic Li growth and poor Coulombic efficiency (CE)14,15. In contrast, ether-based electrolytes exhibit interfacial compatibility with LMAs and promote stable SEI formation16,17. However, their oxidative stability—typically limited to below 4.0 V—restricts their applicability in high-voltage LMBs18,19. To address this challenge, electrolyte engineering strategies have been extensively employed to enhance the high-voltage stability of ether-based electrolytes20–22. Specifically, electrolyte properties can be tuned either through direct structural modification of solvent molecules. For example, Wang et al. replaced the α-hydrogen (α-H) in 1,2-dimethoxyethane (DME) with a methyl group, yielding 1,2-diethoxypropane (DEP), thereby preparing a weakly solvating electrolyte (WSE) that enabled stable operation with 4.3 V lithium nickel cobalt aluminum oxide (NCA) positive electrodes23. Similarly, Bao et al. introduced –CF₂– segments into ether molecules to synthesize fluorinated 1,4-dimethoxylbutane (FDMB), thereby obtaining a high-performance WSE with oxidative stability up to 4.4 V24. In addition to designing new solvent molecules, optimizing the solvated molecular clusters in the electrolyte has also emerged as a widely adopted strategy25–27. Zhang et al. proposed localized high-concentration electrolytes (LHCEs), which employ fluorinated diluents to reduce the viscosity of high-concentration electrolytes (HCEs) while maintaining their high-voltage stability28–30. However, the aforementioned conventional electrolyte engineering approaches still face several limitations, including increased synthetic complexity, higher cost, and compromised ionic conductivity caused by extensive solvation tuning31,32. These approaches typically modify bulk solvent molecules or molecular clusters to optimize interfacial chemistry by altering the overall solvation structure, often at the expense of the intrinsic bulk properties of the electrolyte. Therefore, it is important to develop strategies that directly manipulate specific atomic interactions, enabling precise regulation of the interfacial microenvironment while largely preserving the original solvation structure (Fig. 1a). Design principles of ether-based high-voltage electrolytes. a Illustration of conventional electrolyte engineering and dihydrogen interaction strategies for electrolyte design. b, c Schematic comparison of the functional effects of b CEE and c DHBE in high-voltage LMBs. (Red: Li, blue: DME, green: FSI−, yellow: BH4−). In this work, we develop a dihydrogen-bonded electrolyte (DHBE) that simultaneously achieves high-voltage stability and good compatibility with metallic Li. The DHBE is constructed by adding a trace amount (0.05 M) of LiBH₄ to a conventional ether electrolyte (1 M LiFSI in DME, CEE). In this system, the negatively charged hydrogen atoms (H−) in BH₄− form dihydrogen bonds with the positively polarized hydrogens (Hδ⁺) of DME molecules (H−…Hδ⁺). These dihydrogen-bonding (DHB) interactions not only reduce the desolvation energy barrier of Li+ but also regulate the molecular orientation of DME at the positive electrode eletrical double layer (EDL), thereby suppressing the reactivity of active Hδ⁺ atoms under high-voltage conditions and further inhibiting the oxidative decomposition of ether-based electrolytes (Fig. 1b, c). This dihydrogen-bond-mediated regulation mechanism extends the oxidative stability of DHBE to 5.54 V, while maintaining a high ionic conductivity over 16 mS cm−1 at 30 °C. Electrochemical evaluations show that DHBE enables stable Li plating/stripping and enhanced CE in Li||Li symmetric cells and Li||Cu half cells, respectively. The Li||NCM811 half cells using DHBE deliver 87% capacity retention over 400 cycles at 25 ± 1 °C and sustain over 300 cycles at 60 °C under a high rate of 30 C. Furthermore, DHBE enables Li||NCM811 full cells (20 mg cm−2, 50 μm Li) to operate stably for over 100 cycles at a cutoff voltage of 4.5 V. These results indicate that the direct manipulation of atomic interactions can regulate interfacial chemistry without substantially altering the bulk solvation structure, providing an effective strategy for designing high-voltage ether-based electrolytes. Results Study on dihydrogen-bonding interactions To elucidate the decomposition behavior of DME, we performed density functional theory (DFT) calculations to analyze the electrostatic potential (ESP) distribution of its constituent atoms33. The DFT results show that the oxygen atoms in DME possess the most negative ESP values (–29.0 kcal mol−1), preferentially coordinating with Li⁺ (Fig. 2a). The α-H1/3 atoms of –CH₂– and Me-H1/4 atoms of –CH₃ exhibit the most positive ESP values (+8.7 kcal mol−1 and +6.6 kcal mol−1, respectively), identifying them as active Hδ⁺ sites. Upon coordination with Li⁺, the Oδ− atoms of DME orient toward the cation, positioning the active Hδ⁺ atoms outward. This configuration enhances the reactivity of the Hδ⁺ sites, making them more susceptible to nucleophilic or oxidative attack and thereby promoting DME decomposition34–36. In stark contrast, the hydrogen atoms in BH₄− exhibit highly negative ESP values (–128.2 kcal mol−1), suggesting their potential to form dihydrogen bonds with the Hδ⁺ sites of DME (Fig. 2b). These interactions are further supported by binding energy and structural optimization results. When DME coordinates with Li⁺, the O atom donates electron density, leaving the active Hδ⁺ atom exposed, with a binding energy of –2.83 eV (Fig. 2c). In contrast, in the DME–BH₄− complex, BH₄− is located near the active Hδ⁺ atoms and stabilizes them via DHB interactions, with a binding energy of –2.00 eV, indicating a thermodynamically favorable formation (Fig. 2d). The dihydrogen-bonding interactions are verified by nuclear magnetic resonance (NMR) spectroscopy (see Supplementary Fig. 1 for sample preparation details). 1H NMR analysis demonstrates that incorporating a small amount of LiBH₄ into the CEE to generate the DHBE leads to distinct splitting of DME proton resonances, along with a minor upfield shift (Fig. 2e). The upfield shift signifies enhanced magnetic shielding and higher local electron density around the DME protons, whereas the splitting reflects the establishment of stable dihydrogen-bonding interactions between BH₄− and DME37,38. The persistence of these split peaks indicates that the DHB interactions are not formed and broken instantaneously, suggesting the BH₄−–DME clusters are relatively long-lived in solution. Notably, the peak splitting and upfield shift are most pronounced at a LiBH₄ concentration of 0.05 M (Supplementary Fig. 2), indicating that DHBE at this concentration may exhibit optimal electrochemical properties; this will be analyzed in detail in subsequent sections. Two-dimensional 1H-1H nuclear Overhauser effect spectroscopy (2D 1H-1H NOESY) provides direct spectroscopic evidence for the proposed intermolecular interactions39,40. In the case of CEE, only diagonal proton signals originating from DME are observed in the NOESY spectrum, indicating weak or negligible specific interactions (Fig. 2f). Upon the introduction of BH₄− to form DHBE, diagonal signals corresponding to both DME and BH₄− protons are observed, accompanied by well-defined cross-peaks between these resonances (Fig. 2g). The emergence of these cross-peaks indicates close spatial proximity and efficient dipolar coupling between BH₄− and DME, thereby confirming the presence of dihydrogen-bonding interactions within DHBE. Characterization of DHB interactions. a, b ESP distributions of individual atoms in DME (a) and BH₄− (b). c, d Calculated binding energies of DME coordinated with Li⁺ (c) and of BH₄− (d) under different interactions. (Red: O, gray: C, white: H, pink: B, purple: Li). e ¹H NMR spectra comparing CEE and DHBE. f, g 2D 1H−1H NOESY spectra of CEE (f) and DHBE (g), illustrating intramolecular interactions. Electrolyte solvation structure The solvation structure of the electrolyte was investigated using molecular dynamics (MD) simulations combined with a series of spectroscopic analyses, providing direct insights into the interactions among Li⁺, BH4−, and DME molecules41–43. In the CEE, a substantial fraction of DME molecules remains uncoordinated, enabling full solvation of Li⁺ and predominantly leading to the formation of solvent-separated ion pairs (SSIPs, Fig. 3a, Supplementary Data 1). Radial distribution function (RDF) analysis further reveals that a portion of FSI− anions can enter the primary solvation shell of Li⁺, resulting in a representative Li⁺–2DME–1FSI− solvation structure (Fig. 3b). Upon the introduction of LiBH₄ to construct the DHBE, MD simulations and the corresponding RDF results indicate that BH₄− anions partially replace FSI− within the primary solvation sheath, yielding a Li⁺–2DME–1BH₄− configuration (Fig. 3d, e, Supplementary Data 2). Meanwhile, the remaining non-coordinating BH₄− anions interact with uncoordinated DME molecules in the outer solvation shell through DHB interactions, thereby reshaping the local solvent environment. The RDF between H− in BH₄− and Hδ⁺ in DME features a broad first peak around 4 Å with a coordination number below 1, indicative of weak but persistent dihydrogen bonding rather than conventional solvation (Supplementary Fig. 3). Remarkably, a single BH₄− anion can engage with two DME molecules via dihydrogen bonds, forming a stable cluster approximately 12 ~ 14 Å in diameter. Time-resolved molecular dynamics simulations show that these clusters in DHBE remain structurally stable over 2000 ps, whereas free DME molecules in CEE exhibit pronounced size fluctuations (Supplementary Fig. 4). These findings are consistent with the split peaks in ¹H NMR spectra observed in DHBE, evidencing the stability of BH₄−–DME dihydrogen-bonding interactions. De-solvation energy (Eds) calculations further demonstrate that the BH₄−-involved solvation structure in DHBE exhibits a lower de-solvation energy (1.15 eV) than that in CEE (1.31 eV), indicating facilitated DME dissociation from the Li⁺ solvation shell in the presence of BH₄− (Fig. 3c, f). ⁷Li NMR spectroscopy further corroborates these observations. Upon the addition of LiBH₄ to CEE, the ⁷Li chemical shift moves from –1.23 ppm to –0.48 ppm, indicating a decrease in electron density around Li⁺ and a concomitant weakening of Li⁺–DME coordination, thereby facilitating de-solvation (Fig. 3g)44. Remarkably, in DHBE, the ⁷Li NMR spectrum exhibits a distinct splitting at –0.50 ppm, suggesting that the weakened Li⁺–DME coordination is not solely attributable to the partial replacement of DME by BH₄− in the primary solvation shell, but also arises from DHB interactions between free BH₄− and coordinated DME, which further attenuate the Li⁺–DME ion–dipole interaction. To further elucidate the relative strengths of the interactions between Li⁺ and BH₄− versus the DHB interactions between Li⁺ and DME, a control experiment was designed. As shown in Fig. 3h, Supplementary Fig. 5, ¹¹B NMR spectra indicate that the addition of equimolar amounts of Li⁺ and DME to the BH₄− system induces a nearly identical downfield shift in the ¹¹B resonance, suggesting a reduction in electron density around the B atom45,46. This shift may arise from either the strong Li⁺–BH₄− interaction or the formation of DHB interaction between H− and the active Hδ⁺ atom in DME, with these two interactions exhibiting comparable strengths. Therefore, even at low concentrations, LiBH₄ can form dihydrogen-bonding interactions with DME rather than fully participating in the solvation of Li⁺. Based on these observations, the representative solvation structures in CEE and DHBE are depicted in Supplementary Fig. 6. It is important to note that, due to the relatively low concentration of LiBH₄ (0.05 M), its addition did not significantly alter the solvation structure of CEE. Raman spectroscopy confirms this, showing only a slight increase in the proportions of contact ion pairs (CIP) and aggregates (AGG) in DHBE compared with CEE, whereas the proportion of coordinated DME in DHBE increased significantly (Fig. 3i, Supplementary Figs. 7–10)41,47. These observations suggest that the effect of DHBE on electrolyte properties is primarily mediated through dihydrogen-bonding interactions with DME rather than through significant changes in Li+ solvation structure. Investigation of electrolyte solvation structures. a, d MD simulation snapshots of the solvation structures in CEE (a) and DHBE (d). b, e RDFs and corresponding coordination numbers of Li⁺ in CEE (b) and DHBE (e); insets show enlarged schematic illustrations of representative solvation structures. c, f De-solvation energies of Li⁺–2DME–1FSI− (c) and Li⁺–2DME–1BH₄− (f). g ⁷Li NMR spectra of CEE and DHBE. (Red: O, cyan: C, white: H, pink: F, blue: N, purple: B, gray: Li, yellow: S) h ¹¹B NMR spectra of BH₄− under various coordination environments. i Distribution of solvation structures in different electrolytes derived from Raman analysis. Physicochemical properties and compatibility of DHBE with metallic Li To improve the reversibility of the Li negative electrode, DHBE was rationally designed by incorporating LiBH₄ to establish a dynamic dihydrogen-bonding network. This network effectively restricts anion mobility and stabilizes the SEI, thereby enhancing Li plating/stripping efficiency. Because the bulk solvation structure remains largely unaltered, the introduction of LiBH₄ has minimal impact on ionic conductivity. Conductivity measurements indicate that DHBE and CEE exhibit similar values across a range of temperatures (Fig. 4a). Interestingly, the Li⁺ transference number (tLi⁺) increases markedly from 0.34 in CEE to 0.78 in DHBE (Fig. 4b, Supplementary Figs. 11, 12 and Supplementary Table 2)48. This enhancement can be attributed to the formation of a DHB network between BH₄− and DME, which restricts the mobility of free FSI− anions and slightly hinders the collective motion of solvent molecules, as evidenced by the observed increase in viscosity and the 19F NMR signatures, showing upfield shifts and peak broadening (Supplementary Figs. 13, 14)49–51. Despite the modest increase in viscosity, the bulk ionic conductivity remains largely unchanged, indicating that long-range Li⁺ transport is not significantly impeded. The higher Li⁺ transference number suggests that Li⁺ contributes more effectively to overall ionic transport, consistent with a reduction in strong Li⁺–FSI− pairing. Overall, the DHB network simultaneously slows local anion motion, maintains bulk conductivity, and promotes a higher tLi⁺, thereby supporting more uniform Li⁺ flux and mitigating dendrite growth. Electrochemical characterization further demonstrates the mechanistic benefits of DHBE. Arrhenius analysis reveals reduced activation energies for SEI resistance (Ea,SEI = 86.3 ± 2.9 kJ mol−1, R² = 0.99) and charge-transfer resistance (Ea,CT = 59.9 ± 5.0 kJ mol−1, R² = 0.95), indicating accelerated Li⁺ transport through the SEI and faster interfacial de-solvation kinetics (Fig. 4c, Supplementary Figs. 15–18 and Supplementary Table 3)52. Consistently, Tafel measurements reveal that the exchange current density nearly doubles to 10.08 mA cm−2 relative to 5.28 mA cm−2 for CEE, confirming enhanced Li⁺ deposition kinetics (Fig. 4d)53. Li||Li symmetric cells exhibit prolonged Li plating/stripping over 1800 h with low and stable overpotential at 0.5 mA cm-2 and 1 mAh cm-2, whereas CEE shows rapid polarization due to unstable SEI formation (Fig. 4e). Li||Cu half cells with DHBE maintain high CE over 100 cycles at both 1 mA cm−2 and 3 mA cm−2 with a deposition capacity of 1 mAh cm−2, highlighting the role of DHB interactions in improving Li negative electrode reversibility (Fig. 4f, g). Concentration optimization identifies 0.05 M LiBH₄ as optimal; lower concentrations (0.01 M) fail to generate sufficient DHB, while higher concentrations (0.1 M) disrupt the Li⁺ solvation environment (Supplementary Fig. 19). This experimental result is consistent with the previous 1H NMR results. Importantly, DHBE also improves the initial Coulombic efficiency (ICE) of Li||Cu half cells under high current density conditions, which can be attributed to faster de-solvation kinetics that accelerate Li⁺ deposition and reduce parasitic side reactions (Supplementary Fig. 20). Morphological analysis supports this observation: Scanning electron microscope (SEM) top-view images show that Li deposited in CEE is covered with extensive side products, indicative of parasitic reactions, whereas Li deposited in DHBE is smooth and dense (Fig. 4h, k). Cross-sectional SEM images reveal thinner Li deposition in DHBE (20.1 μm) than in CEE (28.2 μm) at the same deposition capacity, confirming denser packing (Fig. 4i, l)54. Surface characterization further demonstrates that DHBE promotes the formation of a stable, anion-derived SEI. Time-of-flight secondary ion mass spectrometry (TOF-SIMS) mapping shows reduced solvent decomposition fragments (C₂HO−, CHO₂−) and increased, homogeneously distributed anion decomposition fragments (LiF₂−, SO₂−, NO₂−) on DHBE-deposited Li surfaces compared with CEE (Fig. 4j, m, Supplementary Figs. 21, 22)55. X-ray photoelectron spectroscopy (XPS) analysis reveals enhanced LiF content and detectable Li₂S on the Li surface, confirming that DHB interactions facilitate anion-derived SEI formation (Supplementary Figs. 23, 24). These results demonstrate that DHBE not only regulates Li⁺ deposition to form a compact and uniform Li layer but also suppresses interfacial side reactions, thereby improving Li plating/stripping efficiency and compatibility with metallic Li. Electrochemical performance of LMA with different electrolytes. a Ionic conductivity of different electrolytes (Data are presented as mean ± SD (n = 3)). b Li⁺ transference number of CEE and DHBE. c Comparison of activation energies of different electrolytes. d Tafel analysis of Li||Li symmetric cells using various electrolytes. e Long-term Li plating/stripping stability of Li||Li symmetric cells in different electrolytes. f, g Cycling performance of Li||Cu half cells at current densities of 1 mA cm−2 (f) and 3 mA cm−2 (g), with a fixed areal capacity of 1 mAh cm−2. h–m Morphological and SEI characterization of deposited Li after cycling in CEE (h–j) and DHBE (k–m), including SEM top-view images (h, k), SEM cross-sectional images (i, l), and TOF-SIMS three-dimensional depth profiles (j, m). (All Li deposition conditions were conducted in Li||Cu half cells at 0.5 mA cm−2 for 3 mAh cm−2, at 25 ± 1 °C). High-voltage stability of DHBE The introduction of LiBH₄ into DHBE not only enhances metallic Li compatibility but also significantly improves the high-voltage stability of ether-based electrolytes by establishing DHB interactions. These interactions suppress the oxidative decomposition of DME and stabilize the positive electrode–electrolyte interface. Linear sweep voltammetry (LSV) shows that the oxidative decomposition potential of DHBE increases from 4.45 V in CEE to 5.54 V (Fig. 5a). Consistently, potentiostatic measurements at 4.4–4.6 V reveal markedly lower leakage currents in DHBE, confirming its oxidative stability (Fig. 5b)56. As a result, Li||NCM811 half cells using DHBE demonstrate enhanced rate performance within 3.0–4.4 V compared to those with CEE (Fig. 5c, Supplementary Figs. 25–27). Long-term cycling tests further highlight the stabilizing effect of DHB interactions. Li||NCM811 half cells with CEE retain only 5.6% of their capacity after 400 cycles at 2 C, reflecting continuous electrolyte degradation (Fig. 5d, Supplementary Fig. 28)57. In contrast, DHBE-based cells maintain 87% capacity with stable CE over the same conditions, indicating that DHB interactions effectively suppress electrolyte decomposition at the positive electrode. The slight fluctuations observed in the voltage curve during the initial cycle may be associated with variations in ambient temperature (Supplementary Figs. 29, 30). To minimize experimental uncertainty and ensure the reproducibility of the electrochemical results, multiple cells were assembled and tested in parallel. Under identical testing conditions, Li||NCM811 half cells based on the DHBE exhibited stable cycling for more than 400 cycles, retaining over 80% of their initial capacity (Supplementary Fig. 31). Since the oxidative decomposition of ether solvents is strongly temperature-dependent, this performance gap becomes more pronounced at elevated temperatures58. At 60 °C and 50 C, DHBE-based Li ||NCM811 half cells deliver a discharge capacity of 140.1 mAh g−1, compared with 118.3 mAh g−1 for CEE-based cells (Fig. 5e, Supplementary Figs. 32–34). Under elevated temperatures, DHBE more effectively suppresses parasitic side reactions, resulting in a markedly enhanced rate performance relative to CEE. Long-term cycling tests show that CEE-based cells exhibit severe capacity fading and CE fluctuations after only 100 cycles, retaining merely 15.2% after 200 cycles, whereas DHBE enables stable cycling over 300 cycles with 70.8% capacity retention (Fig. 5f, Supplementary Figs. 35–37). To evaluate the practical applicability of DHBE, we assembled Li||NCM811 full cells under practical conditions (20 mg cm−2, 50 μm Li) and increased the cutoff voltage to 4.5 V. Under these conditions, CEE undergoes severe decomposition after only 5 cycles, rendering the full cell unable to operate normally (Fig. 5g, Supplementary Fig. 38). In contrast, DHBE enables stable cycling over 100 cycles with a capacity retention of 82.1% and stable CE, confirming its promising potential for high-voltage LMBs (Supplementary Figs. 39, 40). The above high-temperature cycling of the Li||NCM811 half cell and the cycling of the Li||NCM811 full cell were also evaluated in parallel to ensure the reproducibility of DHBE in stabilizing high-voltage LMBs (Supplementary Figs. 41, 42). To uncover the underlying mechanisms, we performed a systematic failure analysis of cycled NCM811 positive electrodes. Focused ion beam scanning electron microscopy (FIB-SEM) imaging shows extensive intragranular cracking in CEE-cycled positive electrodes, indicative of structural degradation, whereas DHBE-cycled particles retain smooth morphology and structural integrity (Figs. 5h, i, Supplementary Figs. 43, 44)59. High-resolution transmission electron microscopy (HR-TEM) analysis reveals that CEE promotes persistent DME decomposition, forming a thick and disordered cathode electrolyte interphase (CEI, ~42.7 nm, Fig. 5j). In contrast, the DHB interactions in DHBE inhibit active Hδ⁺ atom oxidation, resulting in a compact CEI (~10.1 nm) that better preserves the NCM811 electrode (Fig. 5k)60. Surface characterization further confirms these observations. TOF-SIMS analyses show that Li− and Co− signals are more uniformly distributed in DHBE, indicating a homogeneous CEI (Fig. 5l, Supplementary Figs. 45, 46). Signals associated with DME oxidative decomposition (CO₃−) are significantly reduced, consistent with XPS results, confirming that DHB interactions effectively suppress solvent oxidation (Supplementary Figs. 47, 48). Interestingly, the intensities of anion-derived fragments (LiF₂−, SO₂−) are similar in both electrolytes, suggesting that DHBE stabilization at the positive electrode does not rely on promoting anion decomposition, unlike its effect at the Li metal negative electrode. Control experiments using LiDFOB—a conventional CEI-forming additive—show inferior performance compared with LiBH₄, highlighting that the high-voltage stability enhancement of DHBE arises from DHB interactions rather than alterations in positive electrode interfacial composition (Supplementary Fig. 49)61. These results demonstrate that DHBE not only extends the electrochemical stability window of ether-based electrolytes but also maintains positive electrode structural integrity and forms a compact, uniform CEI, thereby enabling high-voltage, high-temperature, and high-rate operation of LMBs. Electrochemical performance of Li||NCM811 cells with different electrolytes. a LSV curves of Li||Al cells in CEE and DHBE. b Potentiostatic stability tests of Li||NCM811 half cells using different electrolytes. c, d Rate (c) and cycling performance (d) of Li||NCM811 half cells at 25 °C (3–4.4 V, 2 C; 1 C = 180 mA g−1; Supplementary Figs. 25, 30). e, f High-temperature rate (e) and cycling performance (f) of Li ||NCM811 half cells at 60 °C (3–4.4 V, 30 C; 1 C = 180 mA g−1; Supplementary Figs. 32, 37). g Cycling performance of Li||NCM811 full cells employing commercial NCM811 electrodes (20 mg cm−2, 50 μm Li, 3–4.5 V, 1 mA cm−2 charge and 2 mA cm−2 discharge; Supplementary Fig. 40). h, i FIB-SEM images and j, k HR-TEM images of NCM811 positive electrodes after cycling in CEE (h, j) and DHBE (i, k). l TOF-SIMS 3D depth profiles of the CEI formed in different electrolytes. (Cycling tests were conducted using Li||NCM811 half cells at 3–4.4 V and 2 C at 25 ± 1 °C for 100 cycles; 1 C = 180 mA g−1). Positive EDL behavior To understand the enhanced oxidative stability of DHBE at the positive electrode, we investigated the structure of the positive EDL using MD simulations. Under an applied potential, Li⁺ rapidly migrates away from the high-potential region, while anions and solvent molecules accumulate near the electrode to form the inner Helmholtz plane (IHP, Fig. 6a, Supplementary Videos 1, 2)62. In CEE, the IHP is primarily composed of FSI− and DME, with DME molecules oriented such that the Oδ− atoms face the electrode surface and the active Hδ⁺ atoms point away (Fig. 6b, Supplementary Fig. 50 and Supplementary Data 3, 4). This orientation exposes the Hδ⁺ sites to the high-potential field, making DME susceptible to oxidative dehydrogenation and continuous degradation, the presence of BH₄− in DHBE alters the EDL structure. Analysis of relative molecular concentration profiles shows that BH₄− preferentially resides at the innermost IHP, where it forms dynamic dihydrogen-bonding interactions with the exposed Hδ⁺ atoms of DME (Fig. 6c, Supplementary Fig. 51 and Supplementary Data 5, 6). These DHB interactions shield the vulnerable Hδ⁺ sites and induce a reorientation of DME molecules, effectively reducing their susceptibility to active Hδ⁺ removal and improving oxidative stability. Surface-enhanced Raman spectroscopy (SERS) provides further evidence for the DHB interaction mechanism in the positive EDL63. Compared to CEE, DHBE exhibits a significantly enhanced Raman signal corresponding to active Hδ⁺ atoms, particularly the α–H atoms on –CH₂– groups. These signals include bending vibrations (1426–1516 cm−1), asymmetric stretching vibrations (2800–2870 cm−1), and symmetric stretching vibrations (2910–2950 cm−1), indicating a pronounced modification of the local chemical environment surrounding the α–H atoms (Fig. 6d)64,65. Notably, the symmetric stretching vibration of –CH₂– exhibits a blue shift from 2843 to 2852 cm−1 in DHBE, suggesting that the C–H bond in –CH₂– becomes shorter and stronger due to the DHB interaction, thereby rendering the DME molecule more compact and less prone to α-H removal reactions66,67. DFT calculations further corroborate this protective effect. The activation energy for α-H removal in DME increases from 4.24 eV to 4.85 eV upon BH₄− addition, confirming that DHB interactions kinetically inhibit DME oxidation (Fig. 6e, f). These results demonstrate that dihydrogen-bonding interactions can shield vulnerable Hδ⁺ sites in the positive IHP, thereby enhancing the high-voltage stability of DHBE at the positive electrode interface. Effect of DHB interaction at the positive EDL. a Evolution of molecular orientations at the positive EDL in different electrolytes. b, c Relative molecular concentration profiles along the z-axis normal to the NCM811 surface in CEE (b) and DHBE (c). Insets show the corresponding distribution of DME molecules within the EDL. d SERS spectra of different electrolytes at the NCM811 positive electrode interface; the inset provides a magnified view of the 2800–2880 cm−1 region. e, f Calculated reaction energy barriers for α-H removal from DME in the absence (e) and presence (f) of BH₄−. (Red: O, gray: C, white: H, pink: B, dark blue: N, light blue: F, purple: Li, yellow: S, green: Ni). Discussion In summary, we developed a dihydrogen-bonded electrolyte (DHBE) by introducing a trace amount (0.05 M) of LiBH₄ into conventional ether-based electrolytes. MD simulations and Raman spectroscopy show that this minor addition does not significantly alter the solvation structure, preserving high ionic conductivity (>16 mS cm−1 at 30 °C). Moreover, 1H, 7Li, and 11B NMR, together with 2D 1H–1H NOESY spectra, confirm that the hydridic H− in BH₄− forms dihydrogen bonds (H−···Hδ⁺) with Hδ⁺ in DME. These interactions weaken Li⁺–DME ion–dipole coordination, accelerate Li⁺ de-solvation at the negative electrode, and promote formation of anion-derived SEI, thereby significantly enhancing negative electrode compatibility. Furthermore, MD simulations combined with SERS at the positive electrode–electrolyte interface reveal that these dihydrogen-bonding interactions regulate the orientation of DME molecules within the positive EDL, passivate reactive α-H sites in the IHP, and suppress DME decomposition under high-voltage conditions, thereby extending the electrolyte’s oxidative stability to 5.54 V. Together, these synergistic effects enable practical Li||NCM811 full cells (20 mg cm−2, 50 μm Li) using DHBE to operate stably with a 4.5 V cutoff. These results demonstrate that direct manipulation of atomic interactions can enhance electrolyte performance without substantially altering the bulk solvation structure, providing a strategy complementary to conventional solvation-based design approaches. By leveraging dihydrogen-bonding interactions between the anion and solvent, DHBE improves oxidative stability while maintaining high ionic conductivity, overcoming limitations of conventional electrolyte engineering that often compromise transport properties68–70. Notably, trace amounts of LiBH₄ (0.05 M) effectively establish a DHB network, achieving high-voltage stability comparable to that of other additives such as LiDFOB and LiNO₃, which typically require higher concentrations or combinational formulations20,41,62,71–73. The concept of stabilizing solvent molecules via H−-mediated dihydrogen bonding may also be applicable to other metal hydrides (e.g., LiAlH₄, MgH₂), offering a versatile approach for future electrolyte development74. This design strategy, based on dihydrogen-bonding interactions, provides a promising pathway for rationally enhancing the performance of high-voltage LMBs. Methods Materials Lithium bis(fluorosulfonyl)imide (LiFSI, 99.9%), 1,2-dimethoxyethane (DME, 99.9%), and polyvinylidene fluoride (PVDF) were purchased from Dodo Chem. Lithium borohydride (LiBH₄, 95%) was obtained from Acros Organics, and N-methylpyrrolidone (NMP, 99%) was sourced from Aladdin. Tetrabutylammonium tetrahydroborate (TBABH₄, 98%) was purchased from Shanghai Haohong Biopharmaceutical Technology Co., Ltd. No additional drying was performed on DME, LiFSI, or LiBH₄. All chemicals were used as received from the suppliers and handled exclusively inside an Ar‑filled glovebox (H₂O < 0.1 ppm), which prevents moisture uptake. Lithium foil (diameter: 14 mm; thickness: 50 μm or 400 μm) was supplied by China Energy Lithium Co., Ltd. Copper foil (16 μm) was provided by Shenzhen Jingliang Copper Industry Co., Ltd. All metal electrodes (Li foil, Cu foil, and Al foil) were used as received from their respective suppliers without any additional cleaning or surface treatment. All handling and cell assembly were performed inside an Ar-filled glovebox (H₂O < 0.1 ppm, O₂<0.1 ppm). Aluminum foil (14 μm), carbon-coated aluminum foil (15 + 2 μm), LiNi₀.₈Co₀.₁Mn₀.₁O₂ (NCM811) powder, Super P conductive carbon black, and double-sided ceramic separators (12 + 2 + 2 μm) were obtained from Guangdong Canrd New Energy Technology Co., Ltd. The separator was a polyethylene (PE) membrane coated with an Al₂O₃ ceramic layer. Electrolyte preparations The conventional ether electrolyte (CEE) was prepared by dissolving 1 M LiFSI in DME and stirring until fully dissolved, yielding an electrolyte with a concentration of 1 mol L−1. To obtain the DHBE, 0.05 M LiBH₄ was further added to the CEE formulation. All electrolyte preparations were conducted in an argon-filled glovebox, with oxygen and moisture levels maintained below 0.1 ppm. Electrochemical measurements NCM811, Super P, and PVDF were mixed in a weight ratio of 8:1:1 in a mortar. After adding the NMP solution, the mixture was ground for 30 min. The resulting slurry was coated onto carbon-coated Al foil and dried in a vacuum oven at 80 °C for 10 h, yielding a low-loading positive electrode. Li||Cu cells, Li||NCM811 cells, and Li||Al cells were assembled in CR2032-type coin cells (stainless steel cases, positive and negative caps, and stainless-steel wave springs, thickness 0.3 mm, Shenzhen Kejing Star Technology Co., Ltd.) inside a glovebox filled with argon (H₂O < 0.1 ppm, O₂ <0.1 ppm), approximately 40 μL of the corresponding electrolyte was added before sealing. The Li||Cu half cells were prepared using Li foil (φ14 mm, 400 μm) and Cu foil (φ16 mm). For Li||Cu half-cells, the precycling plating process was terminated at a cutoff capacity of 1 mAh cm−2 or 3 mAh cm−2 (current density: 1 mA cm−2, ten cycles). The Li||NCM811 half cells (3.0–4.4 V) were constructed with Li foil (φ14 mm, 400 μm) and NCM811 positive electrodes (φ8 mm, 2.0 mg cm−2) using the studied electrolytes for rate and cycling performance tests. For Li||NCM811 half cells, the precycling charging step was terminated at the upper cutoff voltage of 4.4 V, which corresponds to a typical capacity of ca. 180 mAh g−1 at 0.1 C (one cycle). The Li||NCM811 full cells (3.0–4.5 V) were assembled with Li foil (φ14 mm, 50 μm) and high-loading NCM811 positive electrodes (φ8 mm, 20 mg cm−2) using a different electrolyte for cycling stability tests. Before the long-term cycling test, the Li||NCM811 full cells were pre-activated by a formation cycle at 0.05 mA between 3.0 and 4.5 V to stabilize the SEI on the LMA. After pre-activation, the cells were cycled under the conditions described above. The capacity retention of Li||NCM811 cells in this study was calculated from the 2nd cycle to the final cycle, defined as the capacity after X cycles divided by the capacity in the 2nd cycle. The Li||Al cell configuration consisted of Li foil (φ14 mm, 400 μm) and Al foil (φ16 mm, 14 μm). All coin cells use a double-sided ceramic separator. Electrochemical testing was carried out using a LAND test system at 25 °C or 60 °C. Linear sweep voltammetry (LSV) was performed on Li||Al cells using a CHI660E electrochemical workstation at a scan rate of 1 mV s−1. Tafel analysis was conducted on Li||Li symmetric cells over a voltage range from −0.2 V to 0.2 V at a sweep rate of 1 mV s−1 using the same workstation. Electrochemical impedance spectra of Li||Li symmetric cells were measured on the CHI660E with a voltage amplitude of 5 mV over a frequency range of 100 kHz to 0.1 Hz. The potentiostatic stability was evaluated by holding the Li||NCM811 half cells at 4.4–4.6 V (vs. Li+/Li) for 10 h at 25 ± 1 °C. The Li+ transference number was determined at 25 °C using Li||Li symmetric cells and the Bruce–Vincent method75. The equation used is: First, the electrochemical impedance spectrum was recorded (R₀) over a frequency range of 10⁵ to 10−1 Hz. Next, a DC polarization test was performed by applying a 10 mV potential for 3600 s, during which the initial current (I₀) and the steady-state current (Iss) were recorded. Finally, the electrochemical impedance was measured again to obtain the steady-state resistance (Rss). Temperature-dependent interfacial kinetics were evaluated using Li||Li symmetric cells. EIS measurements were conducted over a temperature range of −20 °C to 60 °C, from which the SEI resistance (RSEI) and charge transfer impedance (RCT) were extracted based on the Nyquist plots. The corresponding activation energy (Eₐ) was determined using the Arrhenius relationship76,77. Materials characterization Raman spectra of electrolytes were recorded using a 785 nm laser with a HORIBA LabRAM HR Evolution spectrometer. For surface-enhanced Raman spectroscopy (SERS), a suspension of Au@SiO₂ nanoparticles in ethanol was drop-cast onto the NCM811 positive electrode surface and subsequently dried at 80 °C to immobilize the nanoparticles78. A Raman electrochemical cell was assembled using the modified NCM811 foil, Li foil, and electrolyte, and charged to 4.4 V. At this voltage, a 532nm laser was directed onto the positive electrode surface to monitor the electrolyte response at the electrical double layer (EDL). The electrolyte conductivities were measured using a DDS-307A conductivity tester over a temperature range of −20°C to 60 °C, with measurements taken at 10 °C intervals. X-ray photoelectron spectroscopy (XPS) was conducted using a Thermo Fisher Scientific K-Alpha instrument. Samples were transferred from the glovebox to the analyzer using an Ar‑purged vessel without air exposure. Spectra were calibrated to C 1s at 284.8 eV. Monochromatic AlKα (1486.6 eV) was used as the X‑ray source. The morphology of lithium deposition was examined with a Zeiss Sigma 300 scanning electron microscope (SEM). The acceleration voltage (beam voltage) was 3 kV, with a working distance of 8 mm and an aperture of 30 μm. Images were acquired using the in‑lens detector. Focused ion beam scanning electron microscopy (FIB-SEM) was performed using an FEI Scios 2HiVac instrument. Nuclear magnetic resonance (NMR) spectroscopy was carried out on an AVANCE NEO 400 MHz instrument. The viscosity of the electrolyte at different temperatures was measured using a HAAKE MARS60 rheometer. High-resolution transmission electron microscopy (HR-TEM) was performed using an FEI Talos F200S to observe the CEI on the positive electrode. Samples for HR-TEM were prepared by ultrasonically immersing electrode sheets in DME, collecting the supernatant, drop-casting it onto lacey-carbon-coated Cu grids, and drying under ambient conditions. Time-of-flight secondary ion mass spectrometry (TOF-SIMS) analysis was carried out using a TOF-SIMS IV instrument (ION-TOF GmbH) with a sputtering duration of 1000 s. Samples were transferred in an Ar‑filled sealed container. Analysis was performed with a Bi₃⁺ primary ion beam (30 keV) and depth profiling with a Cs⁺ sputter beam (1 keV, Cs⁺ chemistry). The internal structure of NCM811 after cycling was examined using a focused ion beam scanning electron microscope (FIB-SEM, Crossbeam 350). All disassembled cells were handled in an inert gas glove box filled with argon (H₂O < 0.1 ppm, O₂ < 0.1 ppm). Electrodes were rinsed three times with 1 mL DME, dried at 25 ± 1 °C for 30 min, and then sealed in a vacuum bag. For XPS and TOF-SIMS analyses, containers were purged with argon during sample transfer to avoid contact with air. All electrochemical tests and sample handling were performed at 25 ± 1 °C. Theoretical calculations Quantum chemical (QC) simulations All quantum chemical (QC) calculations were performed using the ORCA 4 software package. The geometries of all electrolyte components were fully optimized at the B3LYP level of density functional theory, employing the 6–31 + G(d,p) basis set. Based on the optimized structures, molecular electrostatic potential (MESP) values were generated and analyzed using the Multiwfn program79–82. The bonding and de-solvation energies between various electrolyte species were determined by evaluating the interaction energies between individual components and their corresponding electrostatic potential distributions. The bonding energies were calculated as: The de-solvation energies were calculated as: Transition state (TS) searches were also carried out at the same theoretical level to explore possible reaction pathways and associated energy barriers. Frequency calculations were performed to confirm the nature of stationary points, ensuring that energy minima exhibited no imaginary frequencies, while transition states had a single imaginary frequency corresponding to the reaction coordinate. Molecular structures and ESP surface were visualized and examined using Gaussian View. Molecular dynamics (MD) simulations Classical molecular dynamics (MD) simulations were conducted using the Forcite module in Materials Studio (MS) 2019 to investigate both bulk-phase electrolyte behavior and EDL formation. The Universal force field was employed to describe all atomic interactions. The simulation workflow comprised three sequential stages: energy minimization, NPT equilibration, and NVT production. The simulation procedure began with energy minimization using the steepest descent algorithm, in which atomic positions were iteratively adjusted until convergence criteria were met, with a total energy tolerance of 0.01 kJ·mol‒1 and a maximum force threshold of 1000 kJ·(mol·nm)‒1. This was followed by equilibration under NPT conditions using a Berendsen thermostat (set at 298 K) and a Berendsen barostat (set at 0.0001 GPa) to control temperature and pressure, respectively, thereby stabilizing the system’s density. The final production simulations were performed under the NVT ensemble at 298 K for a total duration of 2 ns, with a time step of 1 fs. Periodic boundary conditions were applied in all three spatial directions throughout the simulations. Bulk-phase molecular dynamics (MD) simulations were performed using a cubic simulation box with dimensions of 45 × 45 × 45 ų. Two electrolyte formulations were considered for simplicity: (i) Li⁺: FSI−: DME = 50: 50: 480, corresponding to 1 M LiFSI in DME (representing the simulated solvation structure of CEE), and (ii) Li⁺: FSI−: DME: BH4− = 54: 50: 480: 4, corresponding to 1 M LiFSI + 0.08 M LiBH4 in DME (representing the simulated solvation structure of DHBE). Note that the simulated concentrations slightly deviate from the experimental values due to computational cost, but the deviation is minor and within an acceptable range, thus not affecting the conclusions. Charge neutrality was maintained in both systems. To investigate interfacial behavior, EDL simulations were conducted in a rectangular simulation box of dimensions 45 × 45 × 210 ų. A 140 Å vacuum region was introduced along the z-axis to emulate the electrolyte–vacuum interface, with the electrolyte slab placed adjacent to this vacuum region. This configuration enabled the formation of interfacial double-layer structures under confined geometry. Non-bonded interactions, including van der Waals and electrostatic forces, were treated with a cutoff radius of rc = 12.5 Å. The van der Waals interactions were described using the Lennard-Jones (LJ) potential: Electrostatic interactions were modeled using the Coulomb potential: To characterize the local structural environment, radial distribution functions (RDFs), g(r), were calculated between selected atom pairs. The RDF represents the probability of finding a particle at a distance r from a reference atom, normalized by the average particle density. The RDF presented in this work is calculated by taking the hydrogen (H) atoms in the BH4‒ anion as the reference center, the oxygen (O) atoms in DME molecules as the reference center, and the oxygen (O) atoms in the FSI‒ anion as the reference center. The position of the first peak in g(r) indicates the most probable nearest-neighbor distance. The coordination number (CN) was computed by integrating the RDF up to the first minimum according to the following relation: where ρ is the number density of atoms in the system. These structural analyses were based on trajectory data obtained from MD simulations. Simulation outputs were further analyzed and visualized using both the built-in tools in Materials Studio, VMD and the open-source software OVITO (Open Visualization Tool), which enabled trajectory visualization, coordination analysis, and time-resolved structural evolution83,84. Supplementary information Supplementary Information Description of Additional Supplementary Files Supplementary Data 1-6 Supplementary Video 1 Supplementary Video 2 Transparent Peer Review file Source data Source Data Acknowledgments The authors acknowledge Dr. Xianyu Song from Chongqing Three Gorges University for help with theoretical calculations. The authors also acknowledge Scientific Compass (www.shiyanjia.com) for the XPS and NMR measurements. Author contributions Z.J. and M.Z. conceived the idea and designed the experiments. M.Z., L.L., Q.Y., X.S., S.W. and Z.X. carried out the electrochemical measurements, materials characterization, data collection, and manuscript preparation. K.R. performed theoretical calculations. Z.J., H.L., F.W. and Y.L. supervised the research. All authors contributed to the analysis of the experimental data, discussion of the results, and preparation of the manuscript. Peer review Peer review information Nature Communications thanks Jian Pen and the other anonymous reviewers for their contribution to the peer review of this work. A peer review file is available. Funding This work was financially supported by the National Natural Science Foundation of China (22308003 to Z.J. and 52471027 to Y.L.) and the Natural Science Foundation of Anhui Provincial Education Department (2022AH020033 and 2022AH010025 to Y.L.). Data availability All the data supporting the findings of this study are available within the paper and its supplementary information files. Source data are provided with this paper. Competing interests There are no conflicts to declare. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Meinan Zhao, Kaixiang Ren, Liang Li, Qi Yu. Contributor Information Zhipeng Jiang, Email: jzp1994@ahut.edu.cn. Fei Wang, Email: feiw@fudan.edu.cn. Yongtao Li, Email: liyongtao@ahut.edu.cn. Supplementary information The online version contains supplementary material available at 10.1038/s41467-026-74013-z. References Associated Data Supplementary Materials Supplementary Information Description of Additional Supplementary Files Supplementary Data 1-6 Supplementary Video 1 Supplementary Video 2 Transparent Peer Review file Source Data Data Availability Statement All the data supporting the findings of this study are available within the paper and its supplementary information files. Source data are provided with this paper.