Electrolyte chemistry of adaptive hydrogen bonded domains for high voltage lithium metal batteries

Abstract

The practical implementation of lithium metal batteries is hindered by unstable electrode-electrolyte interfaces and sluggish ion transport kinetics. Here, we report a molecular design strategy that remodels electrolyte solvation structures via the formation of hydrogen-bonded domains, thereby enhancing both the thermodynamics and interfacial dynamics of Li⁺ transport. Specifically, we introduce 2-cyano-N-methylacetamide, an electrochemically stable hydrogen bond donor, as a cosolvent to construct stable nanoscale hydrogen-bonded domains ( < 3.5 Å). 2-Cyano-N-methylacetamide generates both classical (H-bond, H^δ⁺–O^δ⁻) and nonclassical (Z-bond, N^δ⁻–H^δ⁺) hydrogen bonding, which disrupts loosely bound solvated clusters and induces tightly coordinated Li⁺ solvation structures. The hydrogen-bonded domains facilitate the formation of oriented fast Li⁺ transport channels. Accordingly, in Li | |LiNi_0.8Co_0.1Mn_0.1O₂ cells cycled under demanding conditions of 4.7 V with a high areal capacity of ~3.0 mAh cm⁻², the electrolyte enables a capacity retention of 78.8% after 400 cycles. In addition, a stable 4.7 V lithium metal pouch cell is demonstrated with a specific energy (based on the mass of all components) of 418.2 Wh kg⁻¹. This work offers a useful electrolyte design principle on solvation chemistry and interfacial engineering for high-voltage lithium metal batteries.

Introduction

Increasing the charging cut-off voltage to 4.6 V or higher has become a core research focus in pursuit of lithium metal batteries (LMBs) with high specific energy exceeding 400 Wh kg⁻¹, a key milestone towards next-generation high-performance energy storage technologies^1,2,3,4. However, the strong affinity between Li⁺ and the solvent in the constrained battery system inhibits the Li⁺ desolvation process, and the disordered mass transfer and unstable interfacial reaction often lead to the degradation of the battery performance^5,6,7. Eminently, the electrolyte is affected by the magnitude and spacing of the inter-substance forces^8,9,10, which plays a crucial role in the battery’s operational process. In response to these challenges, considerable efforts have been made to regulate reaction kinetics by modulating the competitive Li-ion coordination environment among electrolyte components such as anions, solvent molecules, and additives^11,12. For example, high-concentration electrolytes (HCEs) (Fig. 1b) have been employed to replace conventional dilute electrolytes (Fig. 1a), where enhanced ion-ion interactions promote the formation of contact ion pairs (CIPs) and aggregates (AGGs)^13,14,15. Notably, although the large-sized solvation domains formed by abundant AGGs and CIPs are favorable for anion reduction and decomposition, their stable structure restricts Li⁺ movement, increasing the tortuosity of ion migration pathways and reducing the bulk phase diffusion coefficient¹⁶. In contrast, small-sized solvated domains provide continuous ion channels through dynamic ligand bond reorganization, thereby facilitating rapid Li⁺ transport¹⁷. Several studies in recent years have shown that localized high-concentration electrolytes (LHCEs) modulate solvation domains by regulating dipole-dipole interactions between diluents and solvents (Fig. 1c), which in turn form stable interfaces^18,19,20. However, the disordered dipole-dipole interactions introduced by the diluents tend to disrupt the permeable three-dimensional solvation network, thereby preventing Li⁺ hopping^21,22,23. Therefore, it is important to tune electrolyte molecule/ion interactions and regulate their spatial effects on solvation sheaths for realizing stable electrolytes with efficient ion-transport properties²⁴.

Fig. 1: Design strategy for high-voltage electrolytes with hydrogen bonded domains.

Schematics of the solvation structure of (a) Diluted electrolytes (DE), b high-concentration electrolytes (HCE), c localized high-concentration electrolytes (LHCE) and (d) adaptive hydrogen-bonded domains electrolyte. Schemes showing the (e) dipole-dipole interaction, f coordination bond, and (g) hydrogen bond (Color-coded hollow rings denote distinct atomic species, and the white region inside each ring indicates the shift (polarization) of the atomic electron cloud.). h The screening of possible co-solvents with Highest Occupied Molecular Orbital/Lowest Unoccupied Molecular Orbital (HOMO/LUMO) energies. i Electrolyte screening based on the maximum and minimum values of the electrostatic potential (ESP) surfaces. (Ethylene Carbonate (EC), Ethyl methyl carbonate (EMC), Dimethyl Carbonate (DMC), Fluoroethylene carbonate (FEC), 1,3-dioxolane (DOL), Dimethylsulfoxide (DMSO), Vinyl Sulfate (DTD), 1,1,2,2-tetrafluoroethyl-2,2,3,3-tetrafluoropropyl ether (HFE), N-Methyltrifluoroacetamide(NMTF) and 2-cyano-N-methylacetamide (ANM). Gradient isosurfaces of (j) NMTF and EMC clusters, k ANM and EMC clusters, and l FEC and EMC clusters complexes obtained by the RDG method. Source data are provided as a Source Data file.

Existing studies demonstrated that the interactions are usually dominated around smaller dipole-dipole forces (Fig. 1e, 0.1 ~ 5 kJ mol⁻¹) and van der Waals forces (0.4 ~ 4 kJ mol⁻¹)^25,26,27. The key mechanism of how larger ion-dipole interactions (ligand bonding, Fig. 1f, 50 ~ 200 kJ mol⁻¹) are modulated by smaller dipole-dipole forces remains unclear. Considering the enigmatic nature of H-bond interactions in electrolytes, whether relatively strong hydrogen-bond interactions (Figs. 1g, 5 ~ 100 kJ mol⁻¹) can better regulate the strength and topology of Li⁺-solvents has piqued our keen interest^28,29. According to molecular orbital (MO) theory, H-bond donor-acceptor interactions due to charge transfer from the Highest Occupied Molecular Orbital (HOMO) of X to the empty antibonding orbitals of σ*Y-H bonds enhance the stability of the (X^δ−–H^δ+···Y^δ−) cluster^28,29. With the expansion of IUPAC’s definition of the H-bond, X^δ− does not need to be very negatively charged, but only requires X–H to be at least slightly polar³⁰. This requirement includes bonds of C–H, P–H, etc. Antipolar X–H groups can form non-classical hydrogen bonds (Z-bonds), such as C–H···X and C–H···O³¹. The presence of specific co-solvent molecules has been reported in literature to reshape the interfacial molecular interactions by forming Z-bonds with the solvent molecules, which greatly facilitates Li⁺ desolvation dynamics^32,33. Despite these advances, the potential mechanisms of how to rationally regulate the co-solvent-solvent H-bond and Z-bond interactions on the behavior of the electrode-electrolyte interface are still poorly understood.

To fill this gap in understanding, we developed a strategy for high-voltage LMBs to utilize a donor-containing co-solvent to form H-bond (H^δ⁺–O^δ⁻) and Z-bond (N^δ⁻–H^δ⁺)-coupled hydrogen-bonded domains to remodel the solvation structure. The electronegativity of the ester solvent oxygen atom is weakened by H-bond and Z-bond, thus weakening the Li⁺-solvent force and encouraging more anions to enter the Li⁺ inner solvation shell (Fig. 1d), based on which inorganic solid-electrolyte interphase (SEI) and cathode-electrolyte interphase (CEI) dominated by the anionic decomposition products are preferentially formed. In addition, the construction of hydrogen-bonded domains exhibits a dynamic Li⁺-solvent solvation/desolvation process, significantly enhancing the Li⁺ transport kinetics and the interfacial desolvation kinetics. With such an electrolyte design, the Li | |LiNi_0.8Co_0.1Mn_0.1O₂ (NCM811) coin cell cycled at a high-voltage of 4.7 V and 2.98 mAh cm^–2 retained 78.8% capacity after 400 cycles. Furthermore, the 1.6 Ah Li | |NCM811 cell achieved an specific energy of 418.2 Wh kg^–1.

Results

Adaptive hydrogen-bonded domains electrolyte design strategy

In screening electrolyte co-solvents, multiple metrics were evaluated to ensure an optimal broad electrochemical window. Density functional theory (DFT) calculations of the HOMO and Lowest Unoccupied Molecular Orbital (LUMO) energies served as preliminary indicators of the redox properties of short-chain amides (Fig. 1h and Figs. S1,2). Most of the coordination solvents in conventional solvents such as ethylene carbonate (EC), dimethyl carbonate (DMC), 1,3-dioxolane (DOL), dimethylsulfoxide (DMSO), and vinyl sulfate (DTD) are prone to oxidation due to their high HOMO energy levels. Substituting the hydrogen atom in the amide’s –NH₂ group with a methyl group (e.g., FM, Ac, NMAc, and NNDMAc, the full chemical names of these compounds are provided in Fig S1) enhanced oxidation resistance³⁴. Further modifications, such as replacing the –NH₂ group with isopropyl (–CH₃), trifluoromethyl (–CF₃), or cyanoethyl (–CH₂CN) groups (e.g., TFAc, ANAc, NNDMTFAc, ANNNAc, NMTF, and ANM, the full chemical names of these compounds are provided in Fig. S1), improved compatibility with lithium metal electrodes by increasing polarity³⁵. Moreover, molecules bearing a − CH₂CN group exhibited long-term chemical stability toward both the electrolyte and lithium metal, and no appreciable gas evolution or discoloration was observed (Fig. S3). An initial screening of the magnitude of solvent-Li⁺ forces based on the molecular electrostatic potential (ESP) showed that solvents with large ESP minima (>–30 kcal mol^–1), such as 1,1,2,2-tetrafluoroethyl-2,2,3,3-tetrafluoropropyl ether (HFE), were weakly coordinated to Li⁺, which weakened the electrolyte bulk phase mass transfer (Fig. 1i). In turn, solvents with small ESP minima (>–40 kcal mol^–1) are not conducive to the formation of solvated structures of CIPs and AGGs. Notably, N-Methyltrifluoroacetamide (NMTF, −35.49 kcal mol⁻¹) and 2-cyano-N-methylacetamide (ANM, −31.92 kcal mol⁻¹) exhibited lower ESP minima compared to conventional electrolyte solvents. ANM, in particular, demonstrates a higher ESP maximum (55.30 kcal mol⁻¹) and increased H-bond donor capacity. Reduced density gradient (RDG) analysis revealed significant weak H-bond interactions between amide clusters (N–H) and solvent molecules (O), confirming the existence of H-bond (Fig. 1j-l and Figs. S4,5)³⁶. Fourier transform infrared (FTIR) spectroscopy further supported these findings, showing that the –N–H– stretching frequency of pure NMTF shifts from 3342.0 cm^–1 to 3334.2 cm^–1, and pure ANM shifts from 3353.1 cm^–1 to 3340.4 cm^–1 upon addition of ethyl methyl carbonate (EMC) and fluoroethylene carbonate (FEC), indicating H-bond formation (Fig S6a, b).

The ANM exhibited a more significant blue shift than NMTF, indicating stronger H-bond. The specific atoms involved in these interactions and the strength of the H-bond were elucidated through electron density difference diagrams, revealing that the H-bond formed by ANM with EMC and FEC was considerably stronger than those formed by NMTF (Fig. S7 and Supplementary Data 1–6). Notably, the –C ≡ N (N) group in ANM demonstrated weaker interactions with the –CH₂ (H) group in EMC, which could be attributed to forming the Z-bond. The Z-bond interaction further strengthened the H-bond between ANM and EMC, resulting in a more robust H-bond (− 10.8 kcal mol^–1) compared to that of NMTF ( − 8.8 kcal mol^–1) (Fig. 1j–l). The ¹H nuclear magnetic resonance (¹HNMR) spectra of the electrolytes confirmed the formation of H-bonds between NMTF/ANM and the ester solvents (Fig. S8a). In the ¹H NMR spectra, the formation of H-bonds reduced the shielding of the electron cloud around the hydrogen nucleus, thereby shifting the resonance absorption toward the downfield region. Notably, the −CH₂− group of EMC exhibited an upfield shift in the E-F@NMTF system, whereas it shifted downfield in the E-F@ANM system. This difference arises because ANM can form Z-bond interactions with EMC, while NMTF cannot (Fig. S8b). DFT calculations confirmed the formation of Z-bond (− 2.5 kcal mol^–1) (Table S1). In addition, the H-bond between NMTF/ANM and the (C–F) sites in FEC were comparatively weak, owing to the greater steric hindrance of (C–F) and its lower electron density relative to (C = O). Furthermore, quantum theory of atoms in molecules (QTAIM) calculations were employed to elucidate the detailed characteristics of the hydrogen bonds. Higher electron density (ρ) and a more negative potential energy density (V(r)), generally indicate stronger H-bonds³⁷. In the QTAIM results, all clusters exhibited low ρ and a positive Laplacian (▽²ρ), collectively indicating weak H-bond interactions (Table S2). The ANM and EMC (N–H···O = C) interaction showed the highest ρ, the most negative V(r), and the shortest bond length, indicating that the H-bond between ANM and EMC was the strongest. In addition, relative to the NMTF and FEC cluster, the ANM and FEC cluster displayed weaker interactions. Moreover, a Z-bond (C ≡ N···H–C) was identified in the ANM and EMC cluster, corroborating that Z-bond coupling in the ANM and EMC system further strengthened the H-bond interaction.

The influence of H-bond on the electronic structure was also demonstrated by the reduction in the HOMO energy levels for all cluster when interacting with EMC and FEC clusters (Fig. S9). The strongest H-bonds were formed by ANM and NMTF, leading to the lowest HOMO energy levels for their respective clusters. This outcome was particularly beneficial for enhancing the antioxidant properties of the electrolyte.

To improve the conductivity of the electrolyte, a 1.2 M solution of lithium bis(fluorosulfonyl)imide (LiFSI) in an FEC and EMC mixture in a 3:7 volume ratio (denoted as E-F) was adopted (Fig. S10a–c and Table S3). Addition of NMTF and ANM (denoted as E-F@NMTF and E-F@ANM) at a concentration of 5 M to E-F resulted in an enhancement of electrical conductivity. Notably, the conductivity of the electrolyte exhibits a non-monotonic dependence on the concentration of ANM. Although an increase in co-solvent content typically leads to higher viscosity, the introduction of 5 M ANM results in the highest ionic conductivity among all tested concentrations. This is attributed to the formation of the hydrogen-bonded domains, which enhance Li⁺ transport kinetics despite the increased viscosity of the electrolyte. The electrochemical performance of the Li | |NCM811 coin cell with various amides added to E-F at 5 M was compared in the electrochemical window of 2.8 ~ 4.5 V (Fig. S11). As expected, E-F@ANM demonstrates optimal performance. Similarly, such H-bond interactions were operative across different carbonate solvents. In the FTIR spectra of the EC-EMC-NMTF and EC-EMC-ANM mixed solvents, the N–H stretching vibration showed a pronounced blue shift, unambiguously confirming H-bond (Fig. S12a,b). Moreover, deconvolution of the PF₆^– Raman bands indicated that introducing NMTF or ANM decreased the fraction of solvent-separated ion pairs (SSIP) while increasing the fractions of CIP and AGG, suggesting that H-bond promoted tighter Li⁺–anion association (Fig. S13a–c). In Li | |NCM811 cells employing the E–E@ANM electrolyte, the capacity was retained at 93.3% after 200 cycles, significantly outperforming cells using the E-E electrolyte (Fig. S14).

Electrolyte solvation structures and de-coordination mechanisms

To elucidate the specific role of NMTF/ANM co-solvents in solvent-driven structural remodeling, the solvation structures of the electrolytes were analyzed using molecular dynamics (MD) simulations (Fig. S15a–c). Anion-enriched solvates were observed in E-F@NMTF and E-F@ANM compared to E-F (Fig. S16). In E-F system, the coordination numbers of oxygen in FSI^–, FEC, and EMC were 1.78, 0.21, and 3.6, respectively (Fig. S17a and Supplementary Data 7). In contrast, these values changed to 3.17, 0.18, and 1.51 with NMTF (Fig. S17b and Supplementary Data 8) and to 3.24, 0.26, and 1.46 with ANM (Fig. S17c and Supplementary Data 9), respectively. These changes indicated that NMTF and ANM had minimal involvement in the solvation structure of Li⁺, being primarily distributed around the outer region of the solvation sheath. Because the H-bond interactions between NMTF/ANM and EMC/FEC differ in strength, the introduction of NMTF reduces the proportion of both EMC and FEC molecules participating in the solvation sheath. In contrast, ANM molecules selectively decrease the number of EMC molecules while increasing the proportion of FEC molecules in the solvation sheath (Fig. 2a). Notably, E-F@ANM system further amplified this effect by coupling H-bond and Z-bond interactions, which increases the coordination of more anions with Li⁺. Raman spectroscopy was used to verify the differences in solvated structures. The vibrational peak of the Li⁺–FSI⁻ (720 ~ 740 cm⁻¹) was red-shifted upon adding NMTF/ANM, indicating an increase in aggregation within the electrolyte (Fig. S18a). Concurrently, the addition of NMTF/ANM was accompanied by a simultaneous blue shift of EMC molecules’ vibrational peaks (920 ~ 940 cm⁻¹), indicating a decrease in the coordinated EMC molecules. In contrast to NMTF, the vibrational peaks of FEC molecules (860 ~ 880 cm⁻¹) were red-shifted after the addition of ANM co-solvent, indicating an increase in the coordination of Li with FEC (Fig. S18b). Furthermore, a statistical analysis was conducted on the H-bond distributions between NMTF, ANM and EMC or FEC in the two electrolytes (Fig. 2b, c and Fig. S19a, b). The results demonstrated that the number of H-bond per compartment is significantly higher in the E-F@ANM electrolytes compared to the E-F@NMTF (Fig. S20a–d), with a distribution of N–H ∙ ∙ ∙ O = C bond lengths between 2.5 Å and 3.0 Å, and bond angles between 120° and 180°. The shorter the bond angle and the shorter the bond length were considered the main characteristics of strong H-bonds, this is consistent with previous analysis that E-F@ANM had stronger H-bonds³⁸. In addition, we further quantitatively compared the change in binding energy (E_b) of the solvated structure before and after the NMTF/ANM molecule achieves the drag effect using DFT calculations (Fig. 2d). The Li⁺ dissolution/dissolution energies of EMC, FEC, NMTF and ANM were calculated and compared in Fig. 2e and Fig. S21. Among them, EMC had the highest dissociation energy of −25.17 kcal mol⁻¹, followed by FEC ( − 20.24 kcal mol⁻¹), and the decrease in E_b allowed the coordination reaction to take place thermodynamically. Combining the above results, we showed that ANM could form localized hydrogen-bonded domains within a distance range of less than 3.5 Å from the carbonate molecule, and the formation of hydrogen-bonded domains stabilized the solvent molecule and forms a solvation structure with an anionic inner layer and an outer layer of hydrogen-bonded domains. The formation was confirmed by quantitative analysis of the simulated percentage of MD hydrogen-bonded domains in Fig. S22a, b and peak area fitting calculations of Raman spectra in Fig. S23a–d¹⁷. Furthermore, the solvated cluster size of 8.01 Å for the E-F system (Fig. S24a) was larger than that of 7.07 Å for the E-F@ANM system (Fig. S24b). In the E-F system, Li⁺ forms a large number of SSIPs (Fig. 2f), and the excess solvent ratio in solvation led to parasitic side effects of the ester molecules involved in the desolvation process, causing irreversible damage at the electrolyte-electrode interface. Additionally, the smaller size of Li⁺ clusters could promote faster Li⁺ migration. Unlike LHCEs where the fence effect diluent interrupted the Li⁺ transport hopping mechanism, the unique hydrogen-bonded domains in E-F@ANM were ordered outside the solvated boot layer, resulting in a more relaxed Li⁺ fast channel, which reduced the mass-transport energy barrier in the native phase of the electrolyte (Fig. 2g and Fig. S25)³⁹. In addition, the shortening of the Li⁺-Li⁺ mutual distance also suggested that the construction of adaptive hydrogen-bonded domains forms tighter solvated clusters for Li⁺ structural transport (Fig. S26)⁴⁰.

Fig. 2: Solvation structure and simulations of desolvation kinetics in an adaptive hydrogen bonded domains electrolyte.

a Schematic diagram of the interaction between Li⁺, EMC, FEC, NMTF and ANM. Comparison of the H-bond number between EMC, NMTF, and ANM with bond angles of 120° to 180° (b) and bond lengths of 2.5 to 3.5 Å (c). d The binding energies of different solvation structures. e The calculated Li⁺ solvation/desolvation energies with different solvents. f Schematic illustration of the conventional solvation structure and interface in the E-F electrolyte. In the FSI⁻ anion, elements are color-coded as follows: N (blue), O (red), F (pink), and S (yellow). g Schematic diagram illustrates Li⁺ migration and subsequent desolvation in the E-F@ANM electrolyte. (Noted that the actual coordination state of Li⁺ should be referred to the data provided in the text). Source data are provided as a Source Data file.

Hydrogen-bonded domains anchored solvent molecules to remodel the solvation structure of electrolytes toward enhancing their antioxidant properties

To evaluate the effect of adaptive hydrogen-bonded domains on the antioxidant properties of electrolytes, the electrochemical stability window (ESW) of various electrolytes was assessed using linear scanning voltammetry (LSV) (Fig. 3a). When constrained by an oxidation current threshold of 5 µA, electrolytes E-F@NMTF and E-F@ANM exhibited oxidation potentials of 4.58 V and 5.15 V, respectively. In contrast, the current in the E-F electrolyte sharply increased around 4.0 V. To further investigate electrolyte interactions with NCM811 at high voltages, leakage current tests were conducted at 4.6 V, 4.7 V and 4.8 V (Fig. 3b). At a constant voltage of 4.8 V, cells utilizing the E-F electrolyte displayed higher leakage currents. Conversely, electrolytes E-F@NMTF and E-F@ANM maintained lower and stable leakage currents. NMTF/ANM co-solvents modulate FSI⁻ and Li⁺ interactions, as evidenced by ⁷Li NMR spectroscopy (Fig. 3c). Additionally, previous research indicated that anion-enriched solvation sheaths can diminish the prevalence of solvent-anion clusters, thereby reducing the negative charge transfer of anions to the solvent and effectively widening the ESW⁴¹. The incorporation of NMTF/ANM co-solvents into the E-F electrolyte resulted in a noticeable upfield shift in the ⁷Li NMR chemical shift, moving from −1.07 ppm in the E-F electrolyte to −0.61ppm and −0.52ppm in the E-F@NMTF/ANM electrolyte, respectively. This low-field shift suggested a reduction in the shielding of Li⁺ by EMC and FEC molecules, attributed to the H-bond interactions between NMTF/ANM and the carbonyl groups of these solvent molecules.

Fig. 3: The adaptive hydrogen bonded domains oxidation stability.

a LSV curves in different electrolytes at a scanning rate of 1 mV s⁻¹. b The leakage currents of Li | |NCM811 cells at 4.6, 4.7 and 4.8 V. c ⁷Li NMR spectra at 25 °C for different electrolytes. In situ FTIR spectra of NCM811 surface layer in the E-F electrolyte d and the E-F@ANM electrolyte e Calculated corresponding ESP changes between different solvents in the electrolyte. f interaction energy (g) between different solvents in the electrolyte. h The calculated oxidation potential of the listed complexes. i Schematic of the ESW. Source data are provided as a Source Data file.

Recognizing that ion and solvent distributions are voltage-dependent, in situ FTIR spectroscopy was employed to monitor the dynamic evolution of the electrode-electrolyte interface under varying voltages during cell charging (Fig. S27). The peaks observed around 1090 ~ 1100 cm⁻¹ and 1140 ~ 1160 cm⁻¹ correspond to the vibrations of free EMC and FSI⁻ coordinated with Li⁺, respectively. In the E-F electrolyte, as the charging voltage increased from 3.5 V to 4.8 V, free EMC molecules began to decompose around 4.1 V, followed by FSI⁻ decomposition starting at ~4.3 V. Subsequently, the intensities of these signals decreased rapidly, indicating significant decomposition of the E-F electrolyte (Fig. 3d). Notably, the oxidative stability of the E-F@ANM electrolyte was enhanced, as evidenced by fewer free EMC molecules (Fig. 3e and Fig. S28a) and delayed FSI⁻ decomposition with the increase of voltage, which commenced at around 4.6 V (Fig. S28b).

To visualize the evolution of free ANM, the hydrogen-bonded domains, and H-bond strength during charge–discharge, we analyzed how the full width at half maximum (FWHM) and integrated intensity of the non H-bond and H-bond N–H peaks in in situ infrared spectra varied with the applied potential (Fig. S29a, b)⁴². First, the free N–H peak was fitted with a single Gaussian function, whose center frequency and FWHM were determined from the FTIR spectrum of ANM. The area of the H-bonded component extracted from the measured spectrum—outside the Gaussian-defined free N–H component—quantitatively characterized the extent of the hydrogen-bonded domains. A larger FWHM together with a larger H-bond peak area indicated a stronger H-bond, and vice versa⁴³. At open-circuit voltage (OCV), H-bond interactions were relatively weak, as reflected by the small FWHM and peak area of the H-bonded feature. Upon applying a potential, both the FWHM and the H-bonded peak area increased, indicating rapid formation of an interfacial hydrogen-bonded domain; this trend persisted up to 4.6 V. Above 4.6 V, both metrics decreased, attributable to partial solvent decomposition that weakened the H-bond. In the early stage of charging, anion decomposition reduced the interfacial negative charge; under these conditions, insufficiently desolvated anions from the bulk were less able to reach the interface, while the high potential triggered partial decomposition of EMC and FEC. This behavior persisted up to 4.8 V. Collectively, these observations confirmed the evolution of the hydrogen-bonded domains during cycling, and their distinctive solvent-dragging effect that promoted preferential anion decomposition at the interface.

This improvement is attributed to the formation of adaptive hydrogen-bonded domains with lowered HOMO energy levels by the ANM and solvent molecules. The presence of hydrogen-bonded domains reduced the contact of free anions with the solvent allowing for a significant broadening of the ESW of the electrolyte. The change in ESP around the carbonyl oxygen atom of the ester molecule in the hydrogen-bonded domains (Fig. 3f) showed an increase in the minimum ESP value, indicating that the generated clusters possess a stronger antioxidant capacity (Fig. S30). Furthermore, DFT calculations of oxidation potentials between various clusters showed that FSI⁻ and EMC-clusters have the lowest oxidation potential. Compared with NMTF, ANM created more stable hydrogen-bonded domains with FEC and EMC molecules, thus slowing down their decomposition (Fig. 3g). Notably, the higher oxidation potentials observed for Li⁺ and FSI⁻ clusters further corroborate that the solvation of the anionic bootstrap layer contributed to the improved antioxidant properties of the electrolyte (Fig. 3h).

Overall, the adaptive hydrogen-bonded domains enhance the oxidative stability of the electrolyte by orienting and anchoring the solvent molecules, thereby increasing the anion-enriched solvation structure. This effectively addresses the oxidation of the solvent after desolvation at the high-voltage positive electrode, slowing down the flow of electrons from the HOMO of the electrolyte to the positive electrode (Fig. 3i), which causes electrolyte degradation and capacity loss.

Electrolyte compatibility with lithium metal

Notably, the formation of adaptive hydrogen-bonded domains improved their wettability with lithium metal, facilitating mass transfer (Fig. S31). Importantly, the E-F system forms an unstable SEI leading to excessive dendritic growth and interfacial side reactions, whereas the E-F@ANM promoted uniform Li⁺ fluxes as well as homogeneous Li deposition (Fig. 4a). In situ Raman spectroscopy revealed the dynamic process of electrolyte desolvation on lithium metal surfaces (Fig. S32). In the E-F electrolyte, Raman peaks observed at 890 ~ 900 cm⁻¹ and 930 ~ 940 cm⁻¹ correspond to the vibrations of Li⁺ coordinated with FEC and EMC, respectively (Fig. 4b). The E-F@ANM electrolyte exhibited reduced EMC coordination and increased FEC coordination, which was consistent with prior findings. During discharge, the E-F electrolyte experienced significant decomposition of EMC and FEC peaks, whereas the E-F@ANM electrolyte maintains relatively stable peaks, suggesting that the solvent reduction was mitigated by the pulling effect of ANM on the ester solvent (Fig. 4c). Furthermore, the Raman peaks of FSI⁻ ( ~ 730 cm⁻¹) in the E-F electrolyte red-shifted as the discharge proceeded (Fig. S33a), which was attributed to the fact that the reduction of the E-F electrolyte was mainly solvent-dominated and the decreasing solvent led to the enhancement of Li⁺ and FSI⁻ coordination at the interface. The E-F@ANM electrolyte effectively maintained a stable anion-enriched solvated structure, promoting the formation of an inorganic SEI primarily composed of decomposed FEC and FSI⁻ coordinated with Li⁺ (Fig. S33b).

Fig. 4: Compatibility of different electrolytes with lithium metal.

a Schematic illustration of Li deposition behaviors. In situ Raman spectral maps of (b) E-F and (c) E-F@ANM at different potentials. Depth profiling of F 1 s XPS (d, e) and N 1 s f, g XPS spectra of SEIs formed on Li metal negative electrodes after cycling in E-F and E-F@ANM electrolytes. h Modified Aurbach’s measurement and i cycling of the lithium metal CE in Li | |Cu cells using different electrolytes at 1 mA cm⁻² and 1 mAh cm⁻². j Long-term cycling performance of Li | |Li symmetric cells at 10 mA cm⁻² and 5 mAh cm⁻². Source data are provided as a Source Data file.

This observation was supported by depth-profiled X-ray photoelectron spectroscopy (XPS) analyzes via sputtering conducted after 50 cycles of Li | |Li cells using various electrolytes. In the C 1 s spectra, signals corresponding to C–C/C–H, COOR, C–O, and Li₂CO₃ are observed, with C–O and COOR peaks attributed to EMC and FEC decomposition products (Fig. S34a–c). Notably, the reduced intensity of C–O and COOR peaks in the E-F@NMTF and E-F@ANM electrolytes suggested decreased solvent reduction-generated organic SEI. Conversely, the O 1 s spectra exhibit enhanced Li₂O signals in both E-F@NMTF and E-F@ANM electrolytes, indicating increased inorganic SEI formation (Fig. S34d–f). Additionally, the F 1 s and N 1 s spectra display higher Li₃N and LiF signals in Li metal electrodes cycled with the E-F@NMTF and E-F@ANM electrolytes compared to those cycled with the E-F electrolyte (Fig. 4d–g), aligning with trends observed in FSI⁻ decomposition (Fig. S35a, b). The ratio of clathrate-related signals after sputtering demonstrates that the SEI formed by the E-F system was predominantly organic. In contrast, the SEI formed by the E-F@NMTF and E-F@ANM electrolytes was predominantly inorganic (Fig. S36a–c). Together, these findings suggested that adaptive hydrogen-bonded domains induce the formation of more stable and robust SEI, which could be beneficial for improving the overall performance and stability of the LMBs.

To investigate the effect of adaptive hydrogen-bonded domains on the thermodynamics of mass transport in bulk-phase electrolytes, the Li⁺ transference number (${t}_{{{\mbox{Li}}}^{+}}$) was measured to assess Li⁺ migration within the electrolyte (Fig. S37). The ${t}_{{{\mbox{Li}}}^{+}}$ values were 0.2 for the E-F, 0.47 for the E-F@NMTF, and 0.51 for E-F@ANM electrolytes. Adaptive hydrogen-bonded domains anchoring to solvent molecules as well as constructing Li⁺-directed channels and lowering Li⁺ transference barriers results in high ${t}_{{{\mbox{Li}}}^{+}}$ values.

In addition, the effect of adaptive hydrogen-bonded domains on Li⁺ deposition/stripping kinetics was analyzed. To quantitatively evaluate the kinetics of Li⁺ desolvation and diffusion through the SEI, activation energies were determined from using electrochemical impedance spectroscopy (EIS) spectra collected between −20 and 70 °C. This analysis allowed for the determination of activation energies for charge transfer (E_a,Rct) and Li⁺ transport through the SEI (E_a,RSEI). The values of temperature-dependent charge transfer resistance (R_ct) and SEI resistance (R_SEI) were extracted from the EIS data using the equivalent circuit model depicted in Fig. S38. As shown in Fig. S39a–f, the E_a,Rct of the E-F@NMTF and E-F@ANM are 61.90 kJ mol⁻¹ and 56.53 kJ mol⁻¹, respectively, which are much lower than that of E-F (70.78 kJ mol⁻¹), indicating that the adaptive hydrogen-bonded domains anchoring strategy facilitates enhanced Li⁺ desolvation. Meanwhile, the interfacial Li⁺ diffusion barrier, represented by E_a,RSEI decreased from 49.42 kJ mol⁻¹ in E-F to 43.14 kJ mol⁻¹ in E-F@NMTF and 38.07 kJ mol⁻¹ in E-F@ANM, which was attributed to the formation of an inorganic component-rich SEI with a lower ion mobility barrier. The E-F@NMTF and E-F@ANM were more efficient for lithium plating and stripping due to their anion-rich solvated structures. Fig. S40a–c show the lithium plating morphology of the three electrolytes. Uneven dendritic deposition was observed in the E-F, whereas the deposition microstructure in the E-F@ANM is more uniform and compact compared to that in the E-F@NMTF. The higher coulombic efficiency (CE) values (91.7%/98.9%) of the E-F@NMTF and E-F@ANM (Fig. 4h) compared with that of the E-F (83.43%) can be attributed to the reduction of solvation energy and the inhibition of solvent decomposition in the E-F@ANM. In addition, the lithium nucleation overpotential decreased significantly in the E-F@NMTF and E-F@ANM electrolytes, indicating better interfacial kinetics (Fig. S41). The CE under full plating/stripping conditions was further evaluated at a current density of 1 mA cm⁻² and an areal capacity of 1 mAh cm⁻². Li | |Cu cell cycling with E-F@NMTF and E-F@ANM electrolytes showed stable CE, with average values of 98.1% and 99.0%, respectively, and the E-F@ANM sustained cycling for up to 150 cycles (Fig. 4i). By contrast, the E-F electrolyte barely exceeded 95% CE, and the cell rapidly failed within 30 cycles due to a short circuit caused by lithium dendrites. Comparison of the impedance changes of Li | |Cu cells after plating/stripping cycles revealed that the interfacial impedance was significantly reduced with the E-F@NMTF and E-F@ANM compared to the E-F, indicating the formation of a stable and highly ionic-conducting SEI layer and enhanced Li⁺ charge transport kinetics (Fig. S42). The reductive stability of various electrolytes was evaluated in Li | |Cu cells by monitoring the voltage response over the 0 ~ 1 V window at a current density of 0.2 mA cm⁻² (Fig. S43a). Notably, during the initial discharge to 0 V, corresponding to first-cycle SEI formation, the E-F@ANM system completed the process in the shortest time, indicating fast passivation kinetics and good reductive stability. In addition, a 24 h rest period was introduced during Li | |Cu cycling to assess electrolyte corrosivity toward deposited lithium (Fig. S43b). The results showed that both E-F@ANM and E-F@NMTF mitigated lithium corrosion and retained more reversible active lithium after aging. Meanwhile, the increase in exchange current density confirmed this result (Fig. S44). As shown in Fig. 4j, the symmetric cells cycled at 10 mA cm⁻² and 5 mAh cm⁻² with the E-F@NMTF and E-F@ANM electrolytes exhibited long-term cycling stability above 400 h. In contrast, the polarization voltage of the E-F increased significantly after 10 h. Similar results were observed when current density and capacity were further reduced (Figs. S45–46). The Li | |Li symmetric cells cycling at different current densities also demonstrated the performance of the E-F@ANM electrolyte (Fig. S47). Notably, the cycling stability achieved in this study is consistent with that previously reported for ester-based electrolytes, suggesting the lithium metal–electrolyte interface by adaptive hydrogen-bonded domains.

The in situ EIS results for the Li | |Li symmetric cell indicated that during the initial cycles of the E-F system, the R_ct value remained in a relatively high range, suggesting uneven lithium deposition and ongoing degradation at the electrode–electrolyte interface (Fig. S48a). In contrast, the E-F@NMTF and E-F@ANM systems promoted rapid stabilization of R_ct during the initial cycles, which was attributed to the formation of a more stable SEI, thereby ensuring long-term interface stability (Fig. S48b, c). Additionally, by decomposing the high- and low-frequency components of the EIS, the relaxation time distribution (DRT) was obtained. The DRT analysis decomposed the overall impedance into six distinct relaxation time peaks (p1 ~ p6): p1 corresponded to low contact resistance (R_c); p2 and p3 corresponded to SEI resistance (R_SEI); p4 and p5 corresponded to charge transfer resistance (R_ct); and p6 corresponded to Ohmic resistance (R_b)⁴⁴. The significant p2 and p3 features observed in the E-F system indicated instability in the electrode–electrolyte interface properties. In contrast, the nearly negligible R_ct and R_SEI in the E-F@ANM system suggested a faster Li⁺ desolvation process and faster Li⁺ transport through the SEI.

NCM811 electrolyte interface stability

In addition to being LMA compatible, the electrolyte must also be compatible with high-voltage positive electrode materials to increase specific energy. Unfortunately, the oxidation state of Ni in NCM811 increases (e.g., Ni²⁺ to Ni⁴⁺) when the cell voltage increases. The high oxidation state of Ni⁴⁺ will catalyze the decomposition of the electrolyte, leading to problems such as overgrowth of the CEI and transition metal loss. To further quantify the effect of the adaptive hydrogen-bonded domains on the oxidation resistance of the solvent, the hydrogen transfer reaction energy between the solvent and the charging positive electrode (using a LiNiO₂ substrate) was calculated using DFT. Comparing Fig. 5a, b, the hydrogen transfer energies of EMC ( − 4.88 eV) and FEC ( − 4.34 eV) were not only lower than those of the cosolvent ANM ( − 4.27 eV) but also lower than that of the ANM-EMC cluster ( − 4.11 eV). In addition, the cross-section of the absorbed solvent molecule relaxation structure at the positive electrode was analyzed for charge density. After the H-transfer reaction, the charge density of O atoms on the absorbing sites changed to a spherical distribution, and the change in O charge density revealed the extent of charge transfer feasibility⁴⁵. As shown in Fig. 5c, the EMC molecules absorbed at the positive electrode surface exhibit stronger charge transfer activity. In contrast, the FEC molecules had a slower dehydrogenation reaction due to the spatial site resistance effect. Moreover, the anchoring of EMC (C = O) by ANM (N–H) effectively attenuated the charge density of O (Fig. 5d), which demonstrated that the introduction of ANM could effectively alleviate the high activity of EMC in the dehydrogenation reaction of NCM811. X-ray absorption near-edge structure (XANES) studies were performed to further investigate the effect of high-voltage electrolytes on the stability of the bulk positive electrode structure (Fig. S49a, b). The Ni K-edge spectra of NCM811 at a cut-off voltage of 4.7 V exhibited higher edge energies for electrodes in E-F@NMTF and E-F@ANM than for those in E-F, indicating a higher Ni oxidation state. In contrast, the lower oxidation state in the E-F electrolyte was attributed to the continuous side reactions between the ester solvent and the dissolved metal ions, which disrupted the crystal structure of NCM811 thereby hindering Ni oxidation to higher states. The extended X-ray absorption fine structure (EXAFS) spectra and k³-weighted wavelet-transformed (WT) EXAFS spectra were used to probe changes in the local coordination environment of the electrodes after extended cycling in different electrolytes^{46,47,48,49,50}. Compared to fresh NCM811, wavelet-transformed EXAFS for electrodes cycled in E-F@ANM, relative to those cycled in E-F, reveal an increase in Ni–O bond strength, indicative of greater structural order, while the Ni–O bond length increased by 0.01 Å, suppressing the Jahn-Teller effect (Fig. 5e, f and Fig. S50)^51,52. The synchrotron X-ray diffraction (SXRD) Rietveld refinement was performed in GSAS-II using a single rhombohedral R-3m phase (Fig. 5g, h and Fig. S51)⁵³. The results showed that the E-F@ANM system (2.58%) exhibited a lower degree of Li/Ni cation mixing relative to the E-F system (3.73%) (Tables S4–6). The E-F@ANM electrolyte enhanced the structural integrity of the NCM811 positive electrode by promoting cation ordering within its layered structure. This improvement was evidenced by a higher I₍₀₀₃₎/I₍₁₀₄₎ intensity ratio, indicative of reduced cation mixing and a more ordered crystal lattice.

Fig. 5: Electrolyte dehydrogenation reaction barriers and positive electrode structure/interfacial stability.

Relaxed structure of a solvent molecule absorbed on the de-lithiated positive electrode and the H-transfer reaction energy: EMC and FEC (a) ANM and ANM-EMC (b). Atoms are color-coded as follows: H (white), Li (green), C (brown), N (blue-gray), O (red), F (blue), and Ni (gray). Charge density analysis for the cross-sectional profile for the structure of a solvent absorbed on the de-lithiated positive electrode: EMC and FEC (c) ANM and ANM-FEC (d). The color map of charge density ranges from 0 to 0.5 e bohr⁻³. Ni K-edge wavelet transformed (WT) EXAFS spectra of the NCM811 after 200 cycles of 4.7 V high-voltage cycling with (e) E-F and (f) E-F@ANM electrolytes. Rietveld refinement results for SXRD of NCM811 after 200 cycles using E-F (g) and E-F@ANM (h) electrolytes (The single-phase model used by Rietveld was refined and used to significantly explore the (003) characteristic peak as well as the (104) characteristic peak change.). 3D visualization of TOF-SIMS of selected secondary ion fragments (i) C₂H₃O⁻, (j) C₂HO⁻, (k) LiF₂⁻, l LiO⁻ and m CN⁻. Source data are provided as a Source Data file.

To investigate the morphological and compositional changes of the CEI induced by the E-F@NMTF and E-F@ANM electrolytes, the NCM811 positive electrodes were analyzed using scanning electron microscopy (SEM), high-resolution transmission electron microscopy (HRTEM), XPS, and time-of-flight secondary-ion mass spectrometry (TOF-SIMS). Positive electrode morphology after cycling was examined by SEM. As shown in Fig. S52a–d and Fig. S53a–c, the positive electrode particles cycled in the E-F electrolyte were encapsulated by entangled decomposition products from side reactions. They exhibited significant fractures, which could be attributed to the collapse of the positive electrode structure and dissolution of the transition metal elements (TMEs, including Ni, Co, and Mn). In contrast, the positive electrode surfaces in the E-F@NMTF and E-F@ANM electrolytes exhibited intact secondary particles after cycling. Energy-dispersive X-ray spectroscopy (EDS) maps (Figs. S54–57 and Table S7) indicated that the introduction of NMTF reduced the content of TMEs on the positive electrode surface, particularly in the E-F@ANM system. In addition, ICP-MS of the post-cycled electrolytes showed that the contents of TMEs in the E-F@ANM system were much lower than those in the E-F system (Fig. S58). These results strongly suggested that the CEI layer derived from the electrolyte of the H-bond-anchored solvation structure strategy inhibits the solubilization of TMEs ions and effectively protects the positive electrode/electrolyte interface. In addition, with the E-F@ANM electrolyte, a thin and uniform CEI of ~3.5 nm formed on the surface of NCM811, whereas NCM811 cycled in the E-F electrolyte was encapsulated by a thick and non-uniform CEI of 12.8 nm (Fig. S59a–c). The structure and composition of the CEI formed on the surface of NCM811 in the E-F@ANM electrolyte were further analyzed by HRTEM images (Fig. S60a). LiF and Li₂O nanoparticles were distributed in the outer layer of the CEI, while Li₃N dominated the inner layer. This result was further verified by the corresponding HRTEM and inverse fast Fourier transform (FFT) images (Fig. S60b–d). We analyzed the composition of CEI in NCM811 after 100 cycles using depth-resolved XPS. As shown in the O 1 s spectra of NCM811 in Fig. S61a–c, the RCO₂Li and C = O peaks assigned to carbonate decomposition products decreased progressively from E–F, E–F@NMTF, and E–F@ANM. This indicated that the thickness of the organic CEI layer formed on these positive electrodes decreases accordingly. Meanwhile, a transition metal oxide peak (TM-O) appeared in the E-F system, indicating transition metal dissolution on the NCM811 surface. In addition, the increase in the intensities of the LiF (Fig. S62a–c) and Li₃N peaks (Fig. S63a–c) indicated that the decomposition of FSI⁻ anions on the surface of NCM811 was facilitated, which might contribute to the formation of an inorganic-rich CEI. The intensities of C–F peaks decreased in the E-F@NMTF and E-F@ANM systems, whereas LiF peaks increased. The C–F-containing products may be generated by FSI⁻ decomposition or attributed to the polyvinylidene fluoride (PVDF) binder. Given that the NCM811 electrodes were prepared under the same conditions, the contribution of PVDF to the C–F signal should not be significantly different in these three samples. Therefore, it could be inferred that the increase in LiF peak intensity should be attributed to the decomposition of the FSI⁻ (Fig. S64a–c).

To further support the XPS analysis and to observe the homogeneity of Li₃N/LiF in CEI, we conducted TOF-SIMS on the cycled NCM811 electrodes. As in Fig. 5i, j, C₂H₃O⁻ and C₂HO⁻ fragments correspond to the decomposition products of EMC and FEC. It was observed that the organic components of the E-F system were more extensively distributed along the CEI depth than the E-F@ANM system. In addition, LiO⁻ signal (corresponding to Li₂O) was more abundant in the E-F@ANM system than the E-F system (Fig. 5l), which is attributed to decomposition of the FSI⁻ anion. Furthermore, LiF⁻ (Fig. 5k) and CN⁻ fragments (Fig. 5m) (products of LiFSI) were uniformly distributed throughout the interface (Figs. S65, 66). When LiF was concentrated in the outer layer, it protected the positive electrode interface and reduced side reactions. However, when LiF penetrated deeper into the CEI, it indicated that side reactions were occurring at the positive electrode/electrolyte interface, continuously generating LiF as a decomposition product. This aggravated collapse of the positive electrode structure and led to decrease high-voltage cycling performance. Similarly, in the E-F@ANM system, the peaks of TMEs ions were lower (Fig. S67a–d), and the dissolution of TMEs ions was reduced (Figs. S68, 69). Furthermore, to quantitatively analyze the various components in the CEI, the integrated yields of secondary-ion fragments, including LiF₂⁻ (a marker of the thickest CEI layer), from the surface to 50% depth were compared across electrolytes, and these data were used for relative quantification of degradation products (Fig. S70)⁵⁴. The solvent decomposition products in the E-F@ANM system were lower than those in the E-F system, and the inorganic CEI content arising from anion decomposition products was higher in E-F. This indicated that adaptive hydrogen-bonded domains anchoring strategy reduced electrolyte decomposition by slowing down solvent decomposition under high voltage and forming a CEI rich in Li₃N/LiF with high ionic conductivity and good electronic insulating properties.

The effect of the E-F@ANM induced CEI layer on the crystal structure of the protected NCM811 at a high-voltage of 4.7 V was further investigated by using in situ X-ray diffraction (XRD) to track the shifts of the (003) and (104) peaks. During the initial charging process, the (003)_H1 peak gradually moved to a lower 2θ angle and finally transforms to (003)_H2, indicating the lattice extension along the c-axis. When the charging voltage reached about 4.25 V, the (003)_H2 peak shifted back to a higher 2θ angle, indicating the phase transition from H2 to H3⁵⁵. A significant anisotropic volume change accompanies the H2-H3 phase transition, and the irreversible H2-H3 phase transition can lead to the formation of microcracks at the phase boundaries of the NCM811 material⁵⁶. The (003)_H2-H3 peak shift at the positive electrode in the E-F system was 1.76° (Fig. 6a and Fig. S71), which was higher than the peak shift in the E-F@ANM system (1.71°) (Fig. 6b and Fig. S72). During charging, the peak of (104) shifted to a higher angle, indicating lattice contraction along the a- and c-axes. Similarly, the peak shift in the E-F@ANM system decreased from 2.03° to 1.87°. More importantly, the narrower voltage range of the H2 to H3 phase transition in the E-F@ANM demonstrates that the E-F@ANM system effectively mitigates the irreversible phase transition and improved the structural stability of the positive electrode of the NCM811 during charging and discharging. The dQ/dV contour plots during cycling further confirmed that the E-F@ANM system suppressed the irreversible phase transitions of H2-H3, H1 + M and M + H2 (Fig. 6c, d)⁵⁷. The above results demonstrated that the E-F@ANM derived CEI film effectively suppressed the irreversible phase transition in the positive electrode, thereby enhancing the stability of the NCM811 active material during the cycling process.

Fig. 6: In situ characterization of positive electrode phase transitions and interfacial kinetics.

a, b Charge-discharge curves during the initial cycle at 20 mA g^–1 between 2.8 V and 4.7 V and in situ XRD plots of (003) and (104) reflections of Li | |NCM811 cells using E-F and E-F@ANM electrolytes. dQ/dV curves of Li | |NCM811 cells with E-F (c) and E-F@ANM (d) electrolytes cycled for 400 cycles. In situ EIS plots of Li | |NMC811 cells using E-F (e) and E-F@ANM electrolyte (f) after 50 cycles at 4.7 V, 0.2 C. In situ DRT data representing different SOCs for Li | |NCM811 cells at 25 °C using E-F g and E-F@ANM electrolytes (h). Source data are provided as a Source Data file.

In situ EIS was used to detect the impedance changes during delithiation and lithiation in a high-voltage environment. At each equal voltage, the cell impedance of the E-F@ANM system was smaller than that of the cell using the E-F system (Fig. 6e, f). This indicated that the CEI layer of E-F@ANM has fast and stable ionic conduction efficiency. The effect of the E-F@ANM system on Li⁺ migration kinetics was further confirmed by relaxation time in situ DRT analysis. By mathematically deconvolving the electrochemical impedance data in the relaxation time-space, DRT can accurately distinguish between the interface resistance (R_f; R_f = R_CEI + R_SEI) and charge transfer resistance (R_ct). In this case, the R_ct of the cell was located in the low-frequency region (10¹ > τ > 10⁻¹), which corresponded to the kinetics of the Faraday process associated with the dissolution and desolvation of Li⁺ occurring at both electrodes. The mid-frequency peaks (10⁻¹ > τ > 10⁻³) can be attributed to the R_f⁵⁸. The R_f decreases slightly during Li⁺ desolvation, corresponding to CEI formation, and then rises during discharge, reflecting CEI stripping⁵⁹. The smaller R_f and R_ct in the E-F@ANM compared to the cell using the E-F system suggested that Li⁺ transported more readily through the CEI and the Li⁺ desolvation was much faster (Fig. 6g, h and Fig. S73a, b). And the R_ct of E-F was very large near 4.7 V, indicating that Li⁺ is difficult to desolvate at high-voltages. This result further demonstrated that the improved CEI, induced by the formation of adaptive hydrogen-bonded domains, promotes Li⁺ desolvation.

Electrochemical properties of electrolytes

To demonstrate the feasibility of the adaptive hydrogen-bonded domains anchoring strategy in high-voltage LMBs, we conducted a comparative study of the electrochemical performance of Li | |NCM811 cells using different electrolytes. The cycling of Li | |NCM811 cells with the E-F@NMTF and E-F@ANM electrolytes were much stable at a cut-off voltage of 4.7 V, exhibiting high capacity retention of 66.5% (127.4 mAh g⁻¹) and 78.8% (152.4 mAh g⁻¹) after 400 cycles with E-F@NMTF and E-F@ANM electrolytes, respectively (Fig. 7a). More importantly, when charged to 4.7 V, the E-F@ANM electrolyte enabled an initial coulombic efficiency of 89.6% for NCM811 (Fig. 7b). In contrast, the cell’s capacity with the E-F electrolyte dropped drastically to only 66 mAh g⁻¹ after 400 cycles. In addition, the coulombic efficiency of the E-F@ANM electrolyte was stably above 99.8%, which was much higher than the average coulombic efficiency of 99.1% for the E-F electrolyte, proving that the oxidative decomposition of the E-F@ANM electrolyte was minimal (Fig. S74). Moreover, the E-F@ANM system exhibited stable rate performance (Fig. 7d), with a discharge-specific capacity of 146.6 mAh g⁻¹ at a high C-rate of 3 A g⁻¹. The electrolyte side-reaction rate was accelerated at high temperatures of 60 °C. The battery with the E-F@ANM electrolyte retained 79.5% of its capacity after 250 cycles (Fig. 7c and Fig. S75). In addition to the 4.7 V NCM811, we had also demonstrated the high-voltage stability of the E-F@ANM electrolyte in 4.8 V Li-rich layered oxide positive electrodes Li_1.2Ni_0.13Co_0.13Mn_0.54O₂ (LRLOs) and 4.6 V LiCoO₂ (LCO) (Fig. S76a–d, Fig. S77a–d and S74), and showed the stable rate performance in Li | |LCO (Fig. S79). Furthermore, we assembled lithium metal pouch cells under harsh real-world conditions. To maximize the specific energy, we used a highly loaded negative electrode (20 mg cm⁻²), a lean electrolyte (2.4 g Ah⁻¹), and a thin lithium negative electrode (80 μm, N/P ratio of 1.7). Here, the specific energy of the 1.6 Ah pouch cell with the E-F@ANM electrolyte was as high as 418.2 Wh kg⁻¹ (Fig. 7e, Fig. S80 and Tables S8,9). The voltage profile of the Li | |NCM811 pouch cell with E-F@ANM electrolyte was shown in Fig. S81, where the first cycle at 20 mA g⁻¹ produces an initial coulombic efficiency of up to 89.8%. The battery then maintained 80% capacity retention after 70 cycles of stable cycling at 20 mA g⁻¹ charging/40 mA g⁻¹ discharging. The capacity of the pouch cell with the E-F electrolyte decayed much faster, and the capacity dived after the 22^th charging process, leading to battery failure. This indicated that severe electrolyte side reactions, negative electrode structure collapse and lithium dendrite growth occur in the E-F electrolyte under high-voltage conditions, making stabilizing LMBs for stable cycling at such high mass loading challenging. This also confirmed that the adaptive hydrogen-bonded domains anchoring strategy is conducive to improving the intrinsic stability of the electrolyte, decreasing Li⁺ desolvation energy, effectively forming electrode-electrolyte interfaces with low interfacial impedance and high ionic conductivity. The Ah-grade specific energy presented in this work is comparable to that reported in recent relevant studies (Fig. 7f), demonstrating potential for practical applications^{60,61,62,63,64,65,66,67,68,69,70,71,72,73,74}. To further validate the practical value of our E-F@ANM system, a 2 Ah-class Gr | |NCM811 full cell was tested under 4.5 V conditions (Fig. S82). With the E-F system, the capacity retention declined to 62.5% after 300 cycles. In contrast, the E-F@ANM system exhibited slow capacity decay, retaining 91.3% of its capacity after cycling, demonstrating its effective suppression of electrolyte decomposition, transition metal dissolution, and impedance growth under high-voltage conditions, while maintaining a stable interface layer. These results indicate that the E-F@ANM system is not only compatible with LMBs systems but also holds significant potential for high-voltage operation in commercial lithium-ion batteries.

Fig. 7: Analysis of LMA and cycling performance in different electrolytes.

a Long-term cycling performance at 100 mA g⁻¹ of Li | |NCM811 cells with different electrolytes, and b corresponding voltage curves. c Cycling performance of Li | |NCM811 cells using different electrolytes at 66 mA g⁻¹ at 60 °C. d Rate performances of Li | |NCM811 cells using different electrolytes. e Cycling performance of Li | |NCM811 pouch cells at 20 mA g⁻¹ charge/40 mA g⁻¹ C discharge under charging cut-off voltage of 4.7 V. f Comparison of high voltage cycling stability and specific energy of lithium metal pouch cells with previous reports. Source data are provided as a Source Data file.

Discussion

We demonstrate that stronger H-Bond and Z-Bond interactions are essential for modulating the high-pressure performance of LMBs. After carefully screening H-bond donor co-solvents by DFT calculations, the designed co-solvent ANM forms an anion-rich solvation layer by constructing hydrogen-bonded domains, reducing the solvent’s severe side reactions and accelerating the electrolyte’s mass transfer kinetics. This behavior will lead to the preferential decomposition of FSI⁻ and contribute to the formation of homogeneous anion-derived SEI/CEI. The performance of the Li | |NMC811 cell at a high positive electrode loading of ~3 mAh cm⁻², 60 °C and 4.7 V charging cutoff demonstrates the effectiveness of this design strategy. Our work provides a promising approach for high-voltage, wide-temperature domains, high-specific-energy LMBs and contributes to studying H-bond chemistry in electrolytes.

Materials and methods

Material

Battery-grade LiFSI, EMC, and FEC were obtained from Zhuhai Saiwei Electronic Materials Co. and used as-received. Polyvinylidene difluoride (PVDF) and N-methyl pyrrolidone (NMP) were obtained from DodoChem. Formamide (FM), Acetamide (Ac), Trifluoroacetamide (TFAc), Cyanoacetamide (ANAc), N-Methylacetamide (NMAc), N-N-Dimethylacetamide (NNDMAc), N, N-Dimethyltrifluoroacetamide (NNDMTFAc), 2-Cyano-N, N-dimethylacetamide (ANNNAc), N-Methyltrifluoroacetamide(NMTF) and 2-Cyano-N-methyl-acetamide (ANM) (all in battery-grade purity) were purchased from Aladdin Biochemical Technology Co (Aladdin). Thick lithium metal negative electrodes (15.6 mm diameter and 450 μm thick) and thin Li metal negative electrodes (15.6 mm diameter and 50 μm thick) were received from China Energy Lithium. Polycrystalline NCM811 particles, carbon black (Super P), Celgard 2400 battery separator, Cu collectors, etched Al collectors, 2032-type coin-cell cases, springs and spacers were purchased from Kelude Co, Ltd. positive electrodes (NCM811 positive electrode, LCO positive electrode, LRLOs positive electrode) were prepared by mixing active materials, Super P (Timcal), and polyvinylidene fluoride (PVDF5130, MTI) in N-methyl-2-pyrrolidone (NMP, Aladdin) in a ratio of 8:1:1 to form a slurry, which was then coated onto aluminum foils. The positive electrodes were initially dried at 80 °C for 4 h and subsequently in a vacuum oven at 120 °C for 12 h. Post-dried positive electrodes were punched into circular positive electrodes disks with a diameter of 13 mm, and then transferred to an argon-filled glovebox (water and oxygen levels maintained below 0.01 ppm) for subsequent use. The mass loadings of the three as-prepared positive electrodes (NCM811, LCO, and LRLOs) were measured to be 14.9 mg cm⁻², 14.7 mg cm⁻², and 8.84 mg cm⁻², respectively, with all positive electrodes adopting a single-side coating configuration. Electrolytes were prepared by dissolving the lithium salt LiFSI in the chosen solvents in glass sample vials equipped with polytetrafluoroethylene (PTFE) magnetic stir bars, within a glovebox filled with inert argon gas (water and oxygen levels maintained below 0.01 ppm). The mixtures were stirred at 28 °C for 12 h to yield the electrolytes.

Electrochemical measurements

CR2032 coin cells were assembled in an argon-filled glovebox for electrochemical tests, with 50 μL of electrolyte added before crimping. The as-assembled coin cells were rested at 28 °C for 6 h prior to conducting constant current charge-discharge cycling tests. The charge/discharge intervals of coin cell Li | |LCO, Li | |NCM811, and Li | |LRLOs are 2.8 V ~ 4.6 V, 2.8 V ~ 4.7 V, and 2 V ~ 4.8 V, respectively. Li | |NCM811, Li | |LCO, and Li | |LRLOs coin cells were first pre-cycled twice at 20, 18, and 25 mA g⁻¹, respectively, followed by continuous cycling at 200, 180, and 250 mA g⁻¹ (1 C = 200, 180, and 250 mA g⁻¹ for NCM811, LCO, and LRLOs, respectively). For Li | |NCM811 pouch cells, charging was performed at a constant current-constant voltage (CC-CV) of 0.1 C, and discharging at a constant current (CC) of 0.2 C (NEWARE-CT4008 battery testing system). Linear sweep voltammetry (LSV) and cyclic voltammetry (CV) tests were conducted using a BioLogic instrument, while cycling tests were carried out using a Neware battery cycler (CT-4008) at 28 °C. The LSV curves were obtained using a three-electrode system at a sweep rate of 1.0 mV s^–1. Platinum wire (Φ = 0.1 mm, 99%) and two lithium pieces were used as the working counter, and reference electrodes, respectively. The voltage ranges from 3.5 ~ 6.0 V vs Li/Li⁺ for the LSV measurement. For Li | |Li symmetric cells, thick lithium metal negative electrode, separator, and thick lithium metal negative electrode were stacked together, while Cu replaced the lithium metal in Li | |Cu cells.

The Tafel test was conducted on an AutoLab (PGSTAT302N) electrochemical workstation with the Li | |Li symmetric cells. The electrochemical impedance spectra (EIS) measurements were conducted on an AutoLab (PGSTAT302N) electrochemical workstation over a frequency range from 100 kHz to 0.1 Hz with an amplitude of 5 mV. The measurement was conducted with 5 points per decade in logarithmic spacing.

The Li⁺ transference number was determined with chronoamperometry at a constant step potential of 10 mV, which is based on Li | |Li symmetric cells after several cycles to form a stable SEI. Li⁺ transference number was calculated according to the following Eq. (1):

$${t}_{{{Li}}^{+}}=\frac{{I}_{S}(\Delta V-{I}_{0}{R}_{0})}{{I}_{0}(\Delta V-{I}_{s}{R}_{s})}$$

(1)

Where t_Li+ is the transference number, I₀ and I_s are the initial and steady state current; R₀ and R_s are the initial and steady state resistances of the passivating layers, V is a constant applied polarizing potential.

To obtain the activation energies of Li⁺, the EIS tests of Li | |Li symmetric cells under various temperatures were carried out. The following Eq. (2, 3) can obtain the activation energy:

$$\frac{1}{{R}_{{{\rm{SEI}}}}}=A{e}^{-{E}_{a1}/{RT}}$$

(2)

$$\frac{1}{{R}_{{{\rm{ct}}}}}=A{e}^{-{E}_{a1}/{RT}}$$

(3)

Where R_SEI, R_ct, A and R represent the Li⁺ transport through the SEI resistance, charge–transfer resistance, pre–exponential constant and the standard gas constant, respectively.

The exchange current density was obtained from the Tafel Eq. (4):

$${{\rm{\eta }}}={{\rm{A}}}\times (\log i-\log {i}_{0})$$

(4)

where η and A represent overpotential and the kinetic constant in the diffusion–controlled surface dynamic range, respectively. And ${i}_{0}$ represents the exchange current density. The value of ${i}_{0}$ can be obtained from the intersection of the extrapolated linear part of the $\log i$ versus η plot with the μ = 0 line.

The modified Aurbach method was carried out in Li | |Cu half cells to evaluate the average CE of different negative electrodes. Modified Aurbach’s method was applied as follows. First, 5 mAh cm^–2 Li was plated on Cu and charged until 0.5 V at 0.5 mA cm^–2 to passivate the Cu collector. Second, plated 5 mAh cm^–2 Li at 0.5 mA cm^–2 on Cu again. Third, charged to 1 mAh cm^–2 at 0.5 mA cm^–2 and then discharged to 1 mAh cm^–2 at 0.5 mA cm^–2 for 10 cycles. Last, charged to 0.5 V at 0.5 mA cm^–2 and stripped all remaining Li on Cu. Then, a pre-plating capacity of Li (Q_T) was used as the Li source and a fixed capacity Li (Q_C) was used between the Li working and counter electrodes. After n cycles, the remaining Li reservoir was completely stripped to the cut-off voltage. By measuring the amount of Li remaining after cycling, the average CE over n cycles can be calculated by the Eq. (5):

$${{\rm{CE}}}=\frac{n{Q}_{C}+{Q}_{S}}{n{Q}_{c}+{Q}_{T}}$$

(5)

Where Q_S corresponds to the final stripped capacity or the amount of Li remaining after n cycles. Herein, the value of Q_T of the pre-plating Li was 5 mAh cm^–2, and the value of Q_C was 1 mAh cm⁻². Finally, Q_S was measured after 10 cycles and stripping to 0.5 V. at 1 mA cm⁻². The current density of the above operations was 0.5 mA cm⁻².

Characterizations

At 28 °C, the ionic conductivity of the electrolyte was measured using an 856 conductivity module (from Sweden). At 28 °C, the ionic conductivity of the electrolyte was measured using an 856 conductivity module (from Sweden). For physical and chemical characterization, coin cells before and after cycling were transferred to an argon-filled glove box and disassembled. The as-prepared electrode samples were placed in glass sample vials, vacuum-sealed for storage, and maintained in this state until transported to the testing site. X-ray photoelectron spectroscopy (XPS, Thermo Fisher, Nexsa G2) was used to analyze the Li electrode after 10 cycles. A 1 keV Ar ion gun sputtered In-depth XPS (Thermo Fisher, Escalab 250Xi, the sputtering area was 3 mm × 3 mm, and the sputtering rate was calibrated to 1.5 nm/min by thermally oxidated SiO₂ (standard). Sputtering time is 400 s (10 nm), 800 s (20 nm). The field emission scanning electron microscope (FESEM, ZEISS sigma 500) coupled with an energy dispersive spectroscopy (EDS, BRUKER XFlash 6130) was used to analyze the morphology of NCM811 and sodium deposition after cycling. The CEI of the solid electrolyte membrane of the NCM811 positive electrode after cycling was measured using a transmission electron microscope (TEM) operated at 100.0 kV using HT7700 (Hitachi, Tokyo, Japan). Raman spectra were obtained with inVia^TM confocal Raman microscope (Qontor, RENISHAW) using a 532 nm laser source. ¹H,⁷Li NMR of the electrolytes were conducted on 600 MHZ NMR (AVANCE III HD). TOF-SIMS (ION-TOF.SIMS5) investigated SEI components on Li metal negative electrode. The depth profile was obtained by sputtering etching using Bi₃⁺ beam (30 keV) cluster primary-ion gun. The area of analysis is 100 μm × 100 μm, while the sputtering area is 400 μm × 400 μm. All sample preparation was done in an argon atmosphere glove box. Using in situ Fourier transform infrared (FTIR) spectroscopy (Thermo Scientific Nicolet 6700 spectrometer) in the attenuated total reflection (ATR) mode with a wave number range of 500 ~ 4000 cm^–1 and a resolution of 1.0 cm^–1, the dynamic conditions of the electrolyte interfacial reactions. An in situ X-ray diffraction XRD measurement was performed using a Bruker D8 advance equipped with a Neware battery testing system. The measurements were conducted within a 2θ range of 10° to 80° with a step rate of 10° min^–1. The amount of transition metal ions (Mn and Ni) dissolved from the cycled NCM811 positive electrode was detected by Inductively Coupled Plasma-Mass Spectrometry (ICP-OES, PerkinElmer Optima8000, USA).

As for the local structure, the hard X-ray absorption spectroscopy (XAS) experiments with transmission mode were performed using the BL13SSW beamline in the SSRF at Shanghai, China. Ni-foils were measured simultaneously as the standard comparison sample. The ATHENA software package⁷⁵ was used to process and fit the X-ray absorption near-edge structure (XANES) and extended X-ray absorption fine structure (EXAFS) data.

Synchrotron X-ray diffraction (SXRD) patterns were obtained at the beamline BL14B of the Shanghai Synchrotron Radiation Facility (SSRF) in which the X-ray wavelength was 0.6887 Å.

MD simulation

GROMACS 2021.6 package⁷⁶ for molecular dynamics simulations and the Packmol package for creating the initial simulation box⁷⁷. The OPLSAA force field parameters were used to describe the components of the system. The charges of the anions and cations were scaled down by a factor of 0.8 to account for polarization and charge transfer effects⁷⁸. The FSI^– anions were taken from the literature^79,80, while the solute molecules used the 1.2 CM5 charge obtained at the B3LYP -D3(BJ)/def2-TZVPP computational level. The system was first subjected to a 3000-step energy minimization using the steepest gradient method to eliminate unreasonable overlaps between atoms, followed by 100 ps relaxation under NVT and NPT tethered traces, respectively, and finally the whole system was simulated by sampling the whole system under NPT tethered traces for 50 ns. In the MD simulations, periodic boundary conditions were used in all three directions, and all hydrogen-containing chemical bonds (C-H, O-H) bond lengths were constrained using the LINCS algorithm with a time step of 2.0 fs. The temperature was maintained at 298.15 K, by means of a V-rescale thermostat with a coupling time of 1.0 ps. The pressure was controlled at 1 bar using a Parrinello - Rahman thermostat with a coupling time of 2.0 ps. Long-range electrostatic interactions were handled using the Particle Mesh Ewald (PME) method, and short-range electrostatic and vdW interactions were calculated using a cut-off of 1.2 nm. Trajectories were analyzed and processed and visualized using the Gromacs and VMD 1.9.3 software packages⁸¹, respectively.

Density functional theory (DFT) calculation

Density Functional Theory (DFT) calculations were used to obtain further information on the hydrogen bonding bond energies and binding energies between the electrolyte components. DFT theoretical calculations were performed using the built-in B3LYP two-hybridization generalization of Gaussian 16 software⁸² in combination with DFT-D3 dispersion corrections with damping^83,84. Geometry optimization of the solutes and frequency analysis was carried out using a B3LYP/def2SVP calculation setup to obtain an optimized structure and ensure that it is an energy minimum point. Finally, the B3LYP-D3(BJ)/def2TZVPP calculation level was used to obtain more accurate electron densities and energies. Calculation of hydrogen bonding bond energies based on Eq. (6):

$${E}_{{\rm{bind}}}={H}_{{AB}}{{\rm{\hbox{-}}}}{H}_{A}{{\rm{\hbox{-}}}}{H}_{B}$$

(6)

Where E_bind is the bonding energy and H_AB, H_A and H_B are the enthalpies of the AB complex and the A and B molecules, respectively.

In addition, to explore the mechanism of hydrogen bond formation, Multiwfn 3.8 software⁸⁵ was used to derive the full electron density diagram of the system before and after the formation of the hydrogen bond.

The structural optimization and frequency analysis were performed by the Gaussian 16 A.03 package. The density functional theory (DFT) function based on the 6-311 + G (d, p) level basis set with the use of an implicit solvent model was applied to calculate LUMO/HOMO, ESP. The binding sites were investigated by the molecular electrostatic potential (MESP) method using Multiwfn 3.8 based on wavefunction statistics. All MESP isosurface maps are extracted by visual molecular dynamics (VMD) 1.9.3 program based on the outputs of the Multiwfn analysis. The adsorption energy was calculated as following Eq. (7):

$${E}_{{ads}}={E}_{{AB}}-{E}_{A}-{E}_{B}$$

(7)

Where ${E}_{{AB}}$, ${E}_{A}$, ${E}_{B}$ are the total energy of AB combination, the total energy of A and the total energy of B, respectively.

Theoretical calculation of the solvent dehydrogenation reaction on the surface of NCM811. All computations were carried out with VASP, where electron–ion interactions were represented using the PAW formalism. The exchange–correlation term was described by the PBE functional within the GGA framework. The plane-wave basis set was truncated at 500 eV. Self-consistent field cycles were terminated once the total-energy change dropped below 10^–5 eV. Reciprocal-space integration used a 3 × 2 × 1 k-point sampling. Geometry optimizations were performed until the remaining atomic forces were reduced to under 0.03 eV Å^–1 for every atom.

Solvent contributions to the oxidation potential were incorporated via a dielectric continuum description. Aniline was taken as the solvent medium with a dielectric constant of 7.3, and E_ox referenced to Li/Li⁺ was computed following Eq. (8):

$${E}_{{OX}}=\frac{G({M}^{+})-G(M)}{F}-1.4V$$

(8)

At 298.15 K, the Gibbs free energies of the neutral and oxidized states are denoted as G(M) and G(M⁺) respectively, and F is the Faraday constant. Deformation charge density analysis was conducted with DMol3 (Materials Studio).