De-solvation of heteroalkali cations enabling stable solid electrolyte interphase for dendrite-free lithium metal batteries

Abstract

Li metal is highly sought after as a negative electrode material due to its high specific capacity and low electrode potential. However, the brittle solid electrolyte interphase and undesirable coulombic efficiency have severely hindered its commercial application. Here we propose a formation mechanism of stable solid electrolyte interphase based on the de-solvation of heteroalkali cations rather than the simple electrostatic shielding effect. Cesium trifluoroacetate (CsTFA), characterized by high electron-donor anion and weak solvent-binding heteroalkali cation, was incorporated into commercial electrolytes to foster an inorganic-rich solid electrolyte interphase. The de-solvation of absorbed Cs⁺ dominates the initial solid electrolyte interphase formation on the Li surface. Simultaneously, the preferential reduction of TFA^- promotes the enrichment of LiF within the solid electrolyte interphase. Owing to the synergistic effect of CsTFA, Li | |Cu half cells deliver a high CE of 99.77%, and Li | |LFP full cells exhibit satisfactory stability over 300 cycles with a 94.3% capacity retention at a negative/positive electrode capacity ratio of 1.36. Moreover, the added CsTFA in conventional ester electrolyte demonstrates improved stability of Li | |NCM811 full cells with an 80% capacity retention over 222 cycles at a negative/positive electrode capacity ratio of 1.

The commercialization of lithium metal batteries has been hampered by intractable safety concerns due to their high reactivity. Here, authors use CsTFA as additive in commercial electrolytes to foster a stable solid electrolyte interphase and improve the electrochemical performance of the Li metal negative electrode.

Introduction

Lithium metal, with high theoretical specific capacity (3860 mAh g⁻¹) and low negative potential (−3.04 V vs. SHE), is an ideal negative electrode material for the ever-growing electrochemical energy storage systems^1–3. However, the commercialization of Li metal batteries (LMBs) is impeded by safety and efficiency issues stemming from the highly reactive nature of Li^4–6. The initially formed solid electrolyte interphase (SEI) is fragile and prone to fracture when undergoing displacements during cycling. This fresh fracture increases the contact surface area of the electrolyte and electrode, leading to the reactive consumption of active Li, which in turn promotes the growth of dendritic Li, ultimately resulting in low coulombic efficiency (CE) during Li plating and stripping⁷. Particularly and recently, many attractive strategies around the basic components of LMBs have been proposed to address the tricky Li negative electrode, including artificial SEI layer^8,9, modified separator^10,11, electrolyte regulation^7,12 and Li-host construction^13,14. While all of these are effective in improving electrochemical performance, there is still the unavoidable sacrifice of excess residual Li during full cell operation to compensate for the depletion of active Li^13–15, leading to a high negative/positive electrode capacity (N/P) ratio, increased cost, and reduced energy density^7,16. In addition, the sharp growth of dendrites and the inactive Li deterioration pose potential safety hazards^17,18. Therefore, developing strategies based on the current industry for achieving stable cycling of lean-Li full cells remains a significant challenge.

As the flowing blood inside LMBs, electrolyte has a great impact on CE¹⁹. Newly reported advanced electrolytes all have high CE values but typically rely on high salt concentrations²⁰, expensive diluents²¹, and hybridized solvent molecules or salts²². Utilizing additive salts in commercial electrolytes at a normal concentration to achieve high CE serves as a more cost-effective approach²³. The reduction-sensitive anions in additive salts were routinely derived into inorganic species such as LiF, Li₃P and Li₂O, enhancing the physicochemical properties and interfacial compatibility of SEI with Li, facilitating dendrite-free Li deposition²⁴. However, the role of the cationic portion in additives has been less explored in depth. Since Zhang et al.²⁵ reported the electrostatic shielding effect of added cations in electrolyte, several later research studies all have followed that^26–30. Particularly, the mechanism underlying this electrostatic shielding was not well-substantiated and remains unclear. Despite ongoing research, the specific influence of cationic additives in the electrolyte remains elusive. Recent studies on cation effects have indicated that smaller cations, possessing more concentrated charges, can form larger solvation shells through the encapsulation by multiple layers of solvent molecules. Thus, as the radii of alkali metal cations increase, their binding energy to solvents and the size of their solvation shells decreases^31,32. Consequently, among the feasible alkali cations (Li⁺, Na⁺, K⁺ and Cs⁺), those with larger radii display more pronounced solvent-phobic characteristics and allow them to approach the electrode surface more closely³³, which brings us a bit of inspiration for the exploration of hetero-cation additives.

Herein, a bifunctional additive with high electron-donor anion and weak solvent-binding hetero cation was introduced into conventional electrolytes to achieve stable cycling and dendrite-free LMBs at ultra-low N/P ratios. CsTFA emerges as a preferred choice due to the weak binding energy of solvated Cs⁺, which results in a reduced coordination number with solvent and a compact solvation shell. This configuration facilitates the incorporation of anions into the internal solvation shell and enhances the absorption of solvated Cs⁺ onto the electrode surface. Unlike the conventional de-solvation process of Li⁺, which often leads to an unstable SEI. By employing electrochemical in-situ infrared spectroscopy (FT-IR) characterization, we propose here that the de-solvation of added Cs⁺ predominantly influences the formation of a stable SEI during the initial reductive process. Concurrently, the re-solvation of Cs⁺ competes with the nearby solvated Li⁺ by dragging solvent and anions from it, leading to an increased release of de-solvated Li⁺. The extracted anions then engage in the solvated Cs⁺, thereby participating in the enhanced formation of SEI. Subsequent Li deposition was based on the electrochemical reduction of de-solvated Li⁺ facilitated by the de-solvating Cs⁺ network (Fig. 1). Besides, the reduction-sensitive TFA^- with a low lowest unoccupied molecular orbital (LUMO) energy level was preferentially decomposed to promote the enrichment of LiF in SEI^34–36. In addition, the salt containing TFA^- anion with low lattice enthalpy exhibits easy solubility in aprotic solvents^37,38. Compositional and physicochemical analysis of the initially formed SEI indicate that the active Li tends to deposit towards larger bulk particles beneath a robust and LiF-rich SEI. Consequently, high Li||Cu coulombic efficiency, CE_Cycle of 99.17% and CE_Aurbach of 99.77%, can be obtained in the optimized electrolyte. Notably, the second-cycle CE shoots straight up to 99.36% with a fast-activated first cycle (initial CE of 93.15%). For practical application, Li||LFP full cell with a N/P of 1.36 demonstrates stable cycling over 300 cycles, with 94.3% capacity retention and 99.79% average CE. Under lean-Li conditions, the optimized electrolyte enables well operation of Li||LFP full cells with a range of N/P ratios from 0.107 to 0.4 (158 cycles with 90% retention and 99.9% average CE for 0.415 N/P, and 44 cycles for 0.107 N/P). Interestingly, the relationship between cycling number and N/P ratio could be linearly fitted for prediction. Moreover, in commercial ester system, improved cycling stability of 4.3 V Li||NCM811 full cells can also be achieved (80% retention for 1 N/P after 222 cycles).

Fig. 1. Schematic diagram of the SEI formation and Li deposition.

a Stable SEI exhibits intact and LiF-enriched, which was mainly derived from the de-solvation of solvated Cs⁺. Black dotted line represents the binding of Cs⁺ with solvent and anion. Red dotted line represents the de-solvating network formed by adsorbed Cs⁺ under electric field. b Unstable SEI exhibits chipped with a messy rippled appearance, which was derived from the de-solvation of solvated Li⁺.

Result

Bulk electrolyte solvation chemistry

Both cation and anion of alkali-metal-salt additives have a significant impact on the electrochemical performance of electrolyte. Notably, the cation’s radius was inversely proportional to its binding energy with solvent molecules; thus, a larger radius corresponds to a weaker solvating capability^{31,32,39–41}. Among available alkali metals (Li, Na, K and Cs), the large Cs⁺ of 1.67 Å exhibits the lowest binding energy to solvent and the greatest de-solvating ability (Fig. 2a and Supplementary data 1). This suggests that the solvated Cs⁺ adsorbed on the Li negative electrode surface by electric field can attract less solvent but more anions to participate in the reductive reactions during discharge, thus dominating the original inorganic-rich SEI formation. The Gutmann donor number (DN) of anion indicates its solvating capacity, and those with high DN values were prone to attract electron-deficient cations^42,43. TFA^- with a high DN value of 34.0 kcal mol⁻¹ (Fig. 2b and Supplementary data 1) and good solubility (Supplementary Figs. 1–3 and Supplementary Table 1) emerges as a potential anion candidate. The strong coordination ability of TFA^- with cations reduces the coordination number of solvent molecules involved in the inner solvation shell, thereby preferentially participating in the reduction reaction and exhibiting a higher CE (Supplementary Figs. 4, 5)^38,44. Synergistically, CsTFA was introduced as an electrolyte additive into commercial electrolytes to modulate the solvation structure for achieving stable cycling of lean-Li full cells at ultra-low N/P ratios. The concentration of the CsTFA additive in the baseline electrolyte (BE) was optimized initially. According to the Li | |Cu protocol (Supplementary Fig. 6), an additive concentration of 0.05 M CsTFA in BE electrolyte demonstrates the most extended cyclic stability and the highest CE. Therefore, 0.05 M has been set as the benchmark for assessing other available alkali-metal-cation trifluoroacetate additives. Commercial 1 M LiTFSI in 1,2-dimethoxyethane (DME)/1,3-Dioxolane (DOL) electrolyte, devoid of the commonly used LiNO₃ additive, was selected as the sieving electrolyte (SE) to amplify the impact of additives considering the significant cyclic fluctuation and low CE within 30 cycles (0.5 mA cm⁻² with a fixed capacity of 1 mAh cm⁻², Supplementary Fig. 7). The electrolytes with the addition of 0.05 M LiTFA, NaTFA, KTFA and CsTFA exhibit improved properties. The relatively worse performance of SE + Na and SE + K electrolytes may be attributed to the co-deposition during Li plating, as suggested by the Nernst equation (Supplementary Table 2). Due to the interfacial de-solvation of Cs⁺, SE + Cs demonstrates greater cyclic performance and higher Aurbach CE compared to SE + Li (Supplementary Figs. 7, 8). The influence of different additives on the Li negative electrode follows a similar tendency in the commercial BE electrolyte supplemented with additional 0.1 M LiNO₃ (Supplementary Figs. 9, 10). Consequently, the subsequent research all focuses on the optimized BE electrolyte with the incorporation of 0.05 M CsTFA.

Fig. 2. Theoretical calculations and solvation chemistry.

a Radii of alkali metal cations and their binding energies to DME and DOL. b Donor number of commonly used anions and their binding energies to Li⁺. Snapshot of MD simulation trajectories (40 ps) of BE + Cs (c) and BE (e) electrolytes. Schematic illustration of the representative solvation shell of Li ions in BE+Cs (d) and BE (f) electrolytes. Radial distribution functions and coordination numbers of Li⁺-O_DOL, Li⁺-O_DME, Li⁺-OTFSI^-, Li⁺-ONO₃^- and Li⁺-OTFA^- pairs in BE + Cs (g) and BE (h) electrolytes. i Raman spectra of different electrolytes. j ⁷Li NMR spectra of different electrolytes. k Binding energy and electrostatic potential of the Li⁺-solvent/anion complex. l Calculated HOMO and LUMO of the solvent, Li salts and additives.

Classical molecular dynamics (MD) simulations were conducted to clarify the effect of additive on the solvation structure of bulk BE + Cs and BE electrolytes (Fig. 2c–h and Supplementary data 3). Analysis of the radial distribution functions (RDF, denoted as g(r)), particularly for the Li⁺-O(TFA^-) and Li⁺-O(NO₃^-) interactions within the BE+Cs electrolyte (Fig. 2g), reveals that anions TFA^- and NO₃^- were more likely to associate with Li⁺ over solvent molecules in the innermost solvation shell. This preference was quantified by the calculated coordination numbers of Li⁺ with various species in the BE+Cs electrolyte, including 0.378 for DOL, 1.061 for DME, 0.480 for TFSI^-, 0.085 for NO₃^-, and 0.048 for TFA^- (detailed in Supplementary Table 3). Moreover, Supplementary Fig. 11 and Supplementary Table 4 show the calculated solvation structure of Cs⁺ in BE+Cs. The RDF (g(r)) of Cs⁺-O(TFA^-) and Cs⁺-O(TFSI^-) interactions demonstrate a more pronounced and closer association with Cs⁺ (around 2.7 Å) over solvent molecules. And the calculated coordination numbers of Cs⁺ with solvent, 0.228 for DOL and 0.335 for DME, were smaller than those for Li⁺. The peak located at −35.9 ppm in ¹³³Cs nuclear magnetic resonance (NMR) spectra demonstrates the successful involvement of Cs in solvation structure (Supplementary Fig. 12)^45,46. These results indicate that fewer solvent molecules but more anions (TFSI^- and TFA^-) tend to coordinate with Cs⁺ in the innermost solvation shell. The reduced involvement of solvent molecules during Li plating suggests the formation of a more inorganic-rich and stiffer SEI. Various characterization techniques were performed to investigate the chemical properties of the electrolyte. Ex-situ FT-IR spectra show the reactions between Li⁺, anions, and solvent molecules (Supplementary Figs. 13–17). With CsTFA addition, TFA^- partially participates into the solvent shell of Li⁺ once dissociation. The spectra of BE + Cs and SE + Cs electrolytes both show characteristic peaks at 1705 and 1720 cm⁻¹ (C=O asymmetric stretching frequency), corresponding to the strong interaction between Li⁺ and carbonyl of TFA^-47. Due to the strong binding energy of Li⁺ to TFA^- and NO₃^- (Fig. 2k and Supplementary data 1), ⁷Li nuclear magnetic resonance (NMR) spectra shift to low field with a higher chemical shift value (Fig. 2j). The de-shielding effect in ⁷Li NMR and the apparent blueshift at the characteristic peak of TFSI^- in Raman spectra (located at 735–755 cm⁻¹, S-N-S stretching, Fig. 2i) were rationally expected with the addition of CsTFA. Density functional theory (DFT) calculation and frontier molecular orbital theory were implemented to estimate the solvation structure and redox reactivity of the electrolytes. As shown in Fig. 2l and Supplementary data 2, both CsTFA and LiTFA possess relatively high highest occupied molecular orbital (HOMO) energy levels, suggesting that TFA^- in the innermost solvation shell was prone to oxidatively decompose on the positive electrode surface, forming a fluorine-rich positive electrode electrolyte interphase³⁴. Furthermore, attributed to the strongly polar CF₃^- group, TFA^- owns relatively low LUMO and high reduction potential (Supplementary Fig. 18 and Supplementary data 2, 4) compared to other solvent and anionic components, indicating that TFA^- near the negative electrode surface was more sensitive to electrochemical reduction, thus facilitating the generation of high-quality SEI⁴¹, which was demonstrated by the distinct peak at 1.0–1.2 V in linear sweep voltammetry analysis (LSV, Supplementary Fig. 19)^36,48.

In-situ visualization of de-solvation

To explore the specific mechanism by which CsTFA addition influences the de-solvation process and the initial SEI formation, electrochemical in-situ FT-IR spectroscopy was employed to monitor the dynamic variation of the interfacial electrolyte configuration during real-time Li deposition (Fig. 3a)^49,50. The in-situ raw FT-IR spectra and the corresponding differential A(t_n)-A(t₀) spectra for both BE+Cs and BE electrolytes were presented in Fig. 3d and Supplementary Fig. 20. As de-solvation proceeds, de-solvated solvents were released and gradually accumulate at the interphase. Synchronously, anions were repelled to maintain charge balance once Li⁺ begins to plate. Therefore, the absorption bands corresponding to free solvent (+, marked light green) appear as upward peaks, whereas those associated with Li⁺-solvent (-, marked light yellow) and anion (-, marked light gray) species exhibit downward dips (Supplementary Table 5 for peaks assignments). The intensities of the corresponding peaks change in tandem. During Li plating in the BE electrolyte (Fig. 3d, bottom), a clear shift in relative absorbances from Li⁺-solvent (-) to free-solvent (+) was observed, along with significant changes in TFSI (-) absorbances, indicating a pronounced solvent-related reaction and a decreased content of anions at the interphase. In contrast, the BE+Cs electrolyte (Fig. 3d, top) shows weaker and less variable solvent-related absorbances, with smaller and more moderate changes in TFSI (-) absorbances, consistent with the data from Raman and⁷ Li NMR spectra, suggesting that there were less solvent and more anions at the interphase in the presence of Cs⁺. Besides, the appearance of TFA (-) absorbance in the raw spectra and the related variable peak intensities in the differential spectra demonstrate its role in enhancing SEI enriched in LiF. Consequently, the initial plating process in BE electrolyte follows the conventional de-solvation and electroreduction of solvated Li⁺, leading to the formation of an anion-lean and solvent-rich SEI (Fig. 3c). However, integrating the weak solvent-binding behavior of Cs⁺ and the different results from in-situ FT-IR characterization, it’s proposed that the initial anion-rich and solvent-lean SEI was primarily shaped by the de-solvation of adsorbed Cs⁺, which involves a higher concentration of anions and a reduced solvent content (Fig. 3b). The de-solvated but undeposited Cs⁺ induces an increased release of de-solvated Li⁺ by dragging solvent from solvated Li⁺, and the subsequent de-solvation and electroreduction of de-solvated Li⁺ facilitate Li deposition.

Fig. 3. Capturing the dynamic variation of the interfacial electrolyte configuration during real-time Li deposition via in-situ FT-IR spectroscopy.

a, The schematic of the ATR-IR that in-situ monitoring electrolyte configuration at electrode/electrolyte interphase. Schematic illustration of b the formation of anion-rich solvent-lean SEI via de-solvated Li⁺ deposition across the de-solvating network formed by adsorbed Cs⁺ and c the formation of anion-lean solvent-rich SEI via solvated Li⁺ deposition. d In-situ raw FT-IR spectra and corresponding differential spectra of BE + Cs (spectra color from dark red to light pink) and BE (spectra color from dark blue to light cyan) electrolytes collected during potentiostatic Li plating at −0.1 V for 1200 s, testing temperature of 30 °C. Note that the A(t₀) baseline spectra collected at open circuit voltage without current as background to be deducted, and the A(t_n)-A(t₀) relative absorbance evolution collected with current applied. The bipolar “peak-dip” pattern in differential spectra manifests the dynamic variation of interfacial electrolyte configurations.

Electrode compatibility and interfacial kinetics

Li||Cu half cells were assembled to assess the Li plating/stripping efficiency. As shown in Fig. 4a, the commercial LiNO₃-containing BE electrolyte displays an inconspicuous average CE of ~98% and fails within 120 cycles. In contrast, improved cycling stability and a high CE_Cycle of about 99.17% over 300 cycles could be obtained in BE + Cs with the addition of bifunctional CsTFA. This enhancement was attributed to the high-quality SEI formed by the de-solvation of weak solvent-binding Cs⁺ during the initial cycle. Specifically, the second-cycle CE of BE + Cs shoots straight up to 99.36%, and the average CE for the following four cycles reaches 99.32% (Fig. 4b). Furthermore, the voltage-capacity profiles (Supplementary Fig. 21), in conjunction with the anodic cyclic voltammetry (CV, Fig. 4d and Supplementary Fig. 22), consistently indicate that the plated Li in BE + Cs electrolyte contains a much higher content of active Li.

Fig. 4. Electrochemical stability and interfacial kinetics of different electrolytes.

Cycling (a) and Aurbach measurement (c) of Li metal CE assessed by Li||Cu half cells in different electrolytes at 0.5 mA cm⁻² with a fixed capacity of 1.0 mAh cm⁻². b Voltage profiles and CE for the initial five cycling laps. d CVs of Li||Cu half cells 1.0 V to −0.2 V with a fixed scan rate of 0.5 mV s⁻¹. e Cyclic stability of Li||Li symmetric cells at 5.0 mA cm⁻² with a fixed capacity of 1.0 mAh cm⁻². f Tafel plots in different electrolytes obtained from Li||Li symmetric cells. Testing temperature was set to 30 °C except for specific test. Activation energies (g) and the corresponding Nyquist plots at various temperatures (h) obtained from Li||Li symmetric cells in BE+Cs and BE electrolyte. i Comparisons of CE_Cyling and CE_Aurbach we obtained and other previously reported electrolytes in the form of normal concentration electrolytes (NCE), high concentration (HCE) or locally high concentration electrolyte (LHCE) systems (Detailed information can be referenced in Supplementary Tables 6,7).

The BE + Cs electrolyte also outperforms in CE_Cycle across various deposition amounts and rates (Supplementary Figs. 23–25). Using a modified Aurbach protocol method³, the modified BE+Cs electrolyte yields a high CE_Aurbach reaching up to 99.77% (Fig. 4b). In both CE test configurations conducted (CE_Cycle and CE_Aurbach), the BE + Cs electrolyte demonstrates excellent Li-utilization efficiency consistently, which stands out compared to other electrolyte systems reported in the literature, including both conventional normal concentration and high concentration/locally high concentration electrolyte systems (Fig. 4i and Supplementary Tables 6, 7). Furthermore, the compatibility of the electrolytes with Li metal electrode was evaluated through galvanostatic and rate performance measurements of Li||Li symmetric cells. As shown in Fig. 4e and Supplementary Fig. 26a, the incorporation of CsTFA significantly enhances the cycling stability and reduces the overpotential under specific cycling protocols (3 mA cm⁻², 1 mAh cm⁻² and 5 mA cm⁻², 1 mAh cm⁻²). Notably, the BE+Cs electrolyte can sustain a stable cyclic calendar lifetime of 2000 h and 600 h under these conditions, respectively. Meanwhile, the Li||Li symmetric cells containing the BE+Cs electrolyte exhibited improved rate performance and diminished room-temperature electrochemical impedance both before and after the rate test, underscoring the great capabilities of the BE + Cs electrolyte (Supplementary Fig. 26b–d). To further explore the interfacial kinetics between the electrolyte and Li negative electrode, the exchange current density (j₀) and activation energy (E_a) were measured using the Li||Li symmetric cell model⁵¹. Figure 4f illustrates the Tafel plots and the calculated j₀, and Supplementary Fig. 27 displays the corresponding voltage-current curves. The j₀ value fitted to the curve slope for BE + Cs electrolyte (1.82 mA cm⁻²) was much higher than that for BE electrolyte (1.65 mA cm⁻²), indicating that the processes of charge transfer were accelerated at the interphase with the addition of CsTFA. The E_a obtained from temperature-dependent electrochemical impedance spectroscopy (EIS) measurements by linear fitting of ln(1/R) vs. 1/T was also investigated to support this conclusion (Fig. 4g, h and Supplementary Fig. 28)⁵². The calculated E_a value of BE+Cs (41.66 kJ/mol) was lower than that of BE (46.93 kJ/mol), implying a lower energy barrier for Li deposition and faster electrochemical kinetics in the BE+Cs electrolyte.

Since the CE of Li plating/stripping was correlated with the electrodeposition state, scanning electron microscopy (SEM) was recorded to investigate the morphology and thickness of freshly deposited Li in different electrolytes (Supplementary Fig. 29). With a specific areal capacity of 5 mAh cm⁻² under a current density of 0.5 mA cm⁻²,

the Li||Cu cells were terminated at 0 V prior to disassembly and postmortem electrode measurements. The top-sectional view reveals that Li deposition in the SE electrolyte appears as fragmented, bean-like particles with conspicuous gaps, which promote the generation of “dead” Li and exacerbate the irreversible decomposition of the electrolyte due to the increased contact surface area. Besides, loose dendritic Li deposition can be observed in BE electrolyte with additional 0.1 M LiNO₃. In contrast, the SE + Cs electrolyte, after the addition of CsTFA, results in Li growing towards larger bulk particles with reduced voids. Notably, the optimized BE+Cs electrolyte facilitates dense Li deposition sizing in the teens of microns range and virtually void-free. This ideal deposition morphology mitigates side reactions caused by over-exposure of freshly active Li metal to electrolyte and inhibits the formation of “dead” Li during stripping, thus contributing to a desired high CE value⁵³. In conjunction with the corresponding electrochemical deposition curves (Supplementary Fig. 30), the initial nucleation overpotential upon galvanostatic plating decrease from 82.2 mV in SE and 65.3 mV in BE to 36.2 mV in SE + Cs and 31.4 mV in BE+Cs, respectively. Cross-sectional SEM images with a fixed capacity of 1.0, 2.0, 3.0 and 5.0 mAh cm⁻² were also conducted to investigate the Li densification relative to the reaction surface (Supplementary Fig. 31). It was revealed that the thickness of deposited Li decreases from 38.8 µm in SE and 29.3 µm in BE to 35.7 µm in SE + Cs and 25.1 µm in BE + Cs, respectively. These results clearly demonstrate that the addition of CsTFA leads to improved deposition morphology and reduced nucleation barriers.

Interfacial physicochemistry of SEI

Depth-profiling X-ray photoelectron spectroscopy (XPS) was further carried out on freshly deposited Li to investigate the chemical composition of different SEI across the thickness direction. From the curves of elemental atomic concentration versus sputtering time and the high-resolution XPS depth profiles of F 1s, O 1s, C 1s Li 1s and N 1s peaks (Fig. 5a–d and Supplementary Figs. 32, 33), there were a decreased content of C-F, RCOOLi, C-C/C-H and an increased amount of LiF, Li₂O, LiN_xO_y with the addition of CsTFA (comparing BE + Cs to BE, and SE + Cs to SE)^7,54. This indicates that the original SEI was enriched with fluorine and oxygen while being depleted of organic components due to the interfacial de-solvation of absorbed Cs⁺, which inhibits solvent decomposition and promotes effective anions decomposition, including TFA^-, TFSI^- and NO₃^-55.

Fig. 5. Ex-situ postmortem physicochemical characterizations of SEI.

Depth-profiling XPS spectra of F 1s and O 1s spectra of SEI formed in BE + Cs (a, b) and BE (c, d) electrolytes. e–h Cryo-TEM images and the morphology of Li deposited on a Cu grid in BE+Cs (e, g) and BE (f, h) electrolytes (The corresponding local fast Fourier transform images can be referenced in Supplementary Figs. 21, 22). i The depth profiles of TOF-SIMS for several interested secondary ion fragments. j The 3D ionic distributions of Li^-, LiF^-, LiO^- and Li₃N^- obtained by sputtering and TOF-SIMS analysis of SEI in BE+Cs. The analysis area was 100 × 100 μm² and 900 s etching time. The AFM surface height of deposited Li in BE + Cs (k) and BE (l) electrolytes and the detected average Young’ modulus (m). The samples sent for XPS, TOF-SIMS and AFM tests were obtained by depositing 1.5 mAh cm⁻² at 0.5 mA cm⁻² on Cu after cycling at 1 mA cm⁻² and 1 mAh cm⁻² for 5 cycles at 30 °C.

Cryogenic transmission electron microscopy (Cryo-TEM) was employed to directly visualize the microstructure and chemical composition of the freshly formed SEI. Figure 5g, h exhibits the low magnification cryo-TEM images of plated Li in both BE+Cs and BE electrolytes. In the BE electrolyte, small-sized and inhomogeneous Li aggregates

can be observed, with apparent ruptures depicted by yellow dashed circles in Fig. 5h. In contrast, the optimized BE+Cs electrolyte promotes large and chunky Li deposition on the Cu mesh, consistent with the SEM image results. High-resolution cryo-TEM images reveal distinct SEI with visible boundaries in both electrolytes (areas with brighter diffraction lining). The SEI thickness in BE electrolyte was approximately 25 nm, slightly thicker than the 17 nm observed in BE+Cs electrolyte. Specifically, the marginal region of SEI in BE electrolyte shows a messy, rippled appearance composed of small Li flakes (depicted by translucent blue line in Fig. 5f), whereas the BE + Cs electrolyte shows a flat SEI edge line (depicted by translucent red line in Fig. 5e). High-resolution TEM (HR-TEM) images and fast Fourier transform (FFT) patterns (Supplementary Figs. 34, 35) confirm the crystalline phases in the inner SEI regions of the BE+Cs electrolyte as LiF, Li₂O, Li₃N and LiN_xO_y, with lattice fringes of 2.3, 2.66, 2.76 and 3.06 Å, respectively^24,54. A homogeneous SEI with an amorphous organic outer layer and an inorganic inner layer exhibits great mechanical stability and ideal ionic conductivity, which were crucial for alleviating volume expansion during cycling. However, the formed SEI in BE electrolyte presents a disordered component distribution, with crystalline Li₂CO₃ (lattice fringes of 2.83 Å) found near the messy fringe area. This Li₂CO₃ originates from solvent decomposition and exhibits poor ionic conductivity, thus leading to uneven Li⁺ flux and escalating dendritic growth of Li⁵⁶.

The detailed composition and structure of SEI in BE+Cs and BE electrolytes were further investigated by time-of-flight secondary ion mass spectrometer (TOF-SIMS) analysis (Fig. 5i, j and Supplementary Figs. 36–40). Detected secondary ion fragments such as LiF^- and LiF₂^-, LiO^- and LiO₂^-, Li₃N^- and LiN^-, CO₃^-, C₂HO^- during negative mode sputtering, were characteristic of LiF, Li₂O, Li₃N, Li₂CO₃ and organic species, respectively. Figure 5i and Supplementary Fig. 36 display the trace compositional variability in quantity along with sputtering time. Notably, in BE+Cs electrolyte, the intensity of LiF^- was much higher than that of other species throughout the entire

sputtering process, indicating an enrichment of inorganic LiF in the inner SEI, as proved by depth-profiling XPS spectra. The intensity signal of C₂HO^- fades gradually after an etching depth of 100 s, suggesting that organic components were primarily located in the outermost SEI. Additionally, a significant Li-enriched layer was detected at an etching depth of about 200 s. By comparing these results with those of BE electrolyte (Supplementary Fig. 36c–h), we observed that the SEI in BE + Cs electrolyte contains a higher concentration of inorganic components (LiF, Li₂O and Li₃N) and exhibits more active Li beneath the SEI layer. In contrast, the SEI in BE electrolyte was enriched with organic components and a higher amount of Li₂CO₃, which was generally considered detrimental to SEI stability. The specific hierarchical structure of SEI was more clearly shown by 3D depth profiles (Fig. 4j and Supplementary Figs. 37–40). These TOF-SIMS results were consistent with those obtained from cryo-TEM and XPS, implying that CsTFA induces a higher LiF content in the SEI.

Atomic force microscopy (AFM) was utilized to assess the mechanical properties of SEI (Fig. 5k–m and Supplementary Fig. 41). The surface height mode reveals a flatter and smoother SEI surface in the BE+Cs electrolyte with a roughness (Rq) of 213 nm, contrasting with the pronounced undulations of 377 nm Rq observed in the BE electrolyte. Adequate mechanical strength of the SEI was pivotal in preventing interfacial breakage during cycling⁵⁷. The average interfacial mechanical strength significantly increased from 3.36 GPa in BE to 10.75 GPa in BE+Cs due to the addition of CsTFA, signifying a robust SEI.

Stringent electrochemical evaluations of full cells

In order to verify the high compatibility of BE+Cs electrolyte with Li negative electrode, Li||LFP full-cell protocols (2.8–4.0 V) with an N/P ratio of 1.36 were assembled for testing by pairing a quantized Li foil negative electrode (20 µm in thickness, 4 mAh cm⁻²) with a quantized LFP positive electrode (17.3 mg cm⁻², 2.94 mAh cm⁻²) and injecting a precise dosage of 30 µl electrolyte per cell to ensure a lean electrolyte condition (10.2 µl mAh⁻¹). Figure 6a,c exhibits the cycling performance (0.2 C formation then 1 C cycle) and rate performance (0.2, 0.5, 1, 2, 3 and 1 C) of the Li||LFP full cells, respectively, and the corresponding charge/discharge curves were recorded in the Supplementary Figs. 42, 43. The capacity of full cells in BE electrolyte decays after 100 cycles and rapidly declines after 160 cycles, attributing to the depletion of active Li or electrolyte. However, the full cells with BE+Cs electrolyte maintain stable operation over 300 cycles, with a high reversible capacity of 146 mAh g⁻¹ at 1 C, a capacity retention of 94.3%, and an average CE of 99.79% over the whole full-cell calendar lifespan (Fig. 6a, b). Moreover, the full cells with BE + Cs electrolyte exhibit great rate capacities of 158, 154, 148, 139 and 128 mAh g⁻¹ at 0.2, 0.5, 1, 2 and 3 C, respectively (Fig. 6c, d). Even after reverting the current density back to 1 C, the cells could maintain stable cycling for an extended lifespan. The reduced internal resistance and improved interfacial kinetics by the added CsTFA were evidenced by the results of EIS tests during battery cycling process (Supplementary Fig. 44). The benefit of synergistic CsTFA additive was further demonstrated in lean-Li||LFP full cells with ultra-low N/P ratios ranging from 0.1 to 0.4. The limited Li negative electrodes were prepared by pre-depositing specific amounts of Li (0.4, 0.8 and 1.5 mAh cm⁻²) on Cu substrates in BE + Cs electrolyte. The corresponding plating curves, optical photographs and SEM images of the pre-deposited Li-Cu electrodes were shown in Supplementary Fig. 45. Then the harvested Li-Cu electrodes were paired with LFP positive electrodes of varying mass loadings to meet the range of ultra-low N/P ratios (Supplementary Table 8). Notably, the Li||LFP full cells with a N/P ratio of 0.415 exhibit great cycling stability, maintaining operation over 158 cycles with 90% capacity retention and an average CE above 99.9% during the initial 147 cycles (Fig. 6e). Even at much lower N/P ratios, the Li||LFP full cells also operate stably (Supplementary Figs. 46–49). In particular, at a squeezed N/P ratio of 0.107, the full cells still sustain a high CE of 99.9% over 28 cycles (Fig. 6f). The calculated Li inventory loss rate (LILR) values of the lean-Li||LFP full cells range from 0.138% to 0.395%, demonstrating the high stability and compatibility of the electrolyte with Li negative electrode⁵⁸. In the anode-free protocol (no residual Li on the current collector), the Cu||LFP full cells possess a capacity retention of 40.5% after 200 cycles with a high average CE of 99.4% and an average Li inventory retention rate (LIRR) of 99.436% (Fig. 6g and Supplementary Fig. 50)⁵⁹. After analyzing the cycling results of the lean-Li metal full cells (Fig. 6i, j), we establish two linear correlations between the cycle number (CN) and the N/P ratio, corresponding to the points when the capacity retention (CR) drops to 90% (CN^CR≥90%) and when the coulomb efficiency (CE) falls to 99.7% (CN^CE≥99.7%). The relationships were described by the following equations:

$${\mathrm{CN}}^{\mathrm{CR}\ge 90\%}=356\times \mathrm{N}/\mathrm{P}+4.49$$ 1

$${\mathrm{CN}}^{\mathrm{CR}\ge 99.7\%}=360.2\times \mathrm{N}/\mathrm{P}-16.2$$ 2

Fig. 6. Electrochemical performance of Li||LFP full cells in different electrolytes.

30 µl dosage of electrolyte per cell, testing temperature of 30 °C, cyclic voltage window of 2.8–4.0 V, 1 C = 170 mA g⁻¹ for LFP. a Long-term cycling performance of Li||LFP cells at 0.2 C formation and 1 C cycling. c Rate performance of Li||LFP cells at 0.2 C/0.5 C/1 C/2 C/3 C/1 C. Corresponding cycling (b) and rate (d) charge-discharge curves of Li||LFP cells. e, f Long-term cycling performance and calculated LILR of Li||LFP cells with ultra-low N/P ratios at 0.2 C formation and 0.5 C cycling. g Long-term cycling performance of anode-free Cu||LFP cells at 0.3 C charge and 0.5 C discharge. h, i The fitted linear relationship between cycle number and N/P ratios at retention ≥90% and CE ≥ 99.7% of full cells.

The corresponding CN increases linearly with the N/P ratio under lean-Li conditions, suggesting that the lifespan of full cells with pre-deposited Li can be predicted ahead. The electrochemical performance of electrolyte was reflected in the slope of the linear fit. Notably, the Y-axis intercept in Eq. (1) represents the best-effort cycling lifespan of the anode-free protocol, and the X-axis intercept in Eq. (2) can be referenced as the minimum amount of irreversible Li necessary for the cell operation. Our study outperforms recent works on Li metal batteries with ultra-low N/P ratios (<4), as shown in Supplementary Fig. 51 and Supplementary Table 9.

Interestingly, the application of CsTFA in ester system for high-voltage positive electrode Li metal batteries was also evaluated. As displayed in Supplementary Fig. 52, the CsTFA has a saturated solubility of 0.025 M in the commercial EC/DEC/FEC system, nevertheless, the formulated electrolyte with the addition of 0.05 M remains opaque after 24 h. The difference at characteristic peak (1705–1720 cm⁻¹) in ex-situ FT-IR spectroscopy confirms the dissociation of CsTFA in the electrolyte (Supplementary Fig. 53). DFT calculation for the LiPF₆-ester system indicates that large Cs⁺ also has a small binding energy to ester solvent molecules, and TFA^- owns the lowest LUMO energy level and the strongest binding energy with Li⁺ (Supplementary Fig. 54). The improved Li||Cu and Li||Li half-cell performance and reduced EIS impedance in BCE+Cs electrolyte demonstrate its improved compatibility with Li negative electrode (Supplementary Figs. 55–57). High-voltage full-cell protocols (3.0–4.3 V, N/P of 1) were assembled by pairing Li foil negative electrode (20 µm in thickness, 4 mAh cm⁻²) with NCM811 positive electrode (20.7 mg cm^-2, 4 mAh cm^-2) and injecting a precise dosage of 30 µl electrolyte per cell (7.5 µl mAh⁻¹). The Li||NCM811 full cells with BCE+Cs electrolyte exhibit stable cycling over 222 cycles with 80% capacity retention and good rate performance even at large mass loadings (Supplementary Fig. 58). These outcomes reinforce the viability of incorporating CsTFA into commercial ester-based electrolytes.

Discussion

This work introduces a concept about the SEI formation based on the de-solvation of heteroalkali cations to enable stable Li negative electrode. CsTFA, with high electron-donor anion and weak solvent-binding cation, was selected as an additive for commercial electrolytes to achieve high compatibility with Li metal. A series of theoretical tools and combinatorial characterizations, such as MD, in-situ electrochemical FT-IR spectroscopy and cryo-TEM, have been co-applied to demonstrate that the original SEI, involving more anions and less solvent, was derived from the preferential de-solvation process of absorbed Cs⁺ on the surface. The decomposition of reduction-sensitive TFA^- promotes the enrichment of LiF in SEI. The optimized BE+Cs electrolyte exhibits high CE, CE_Cycle of 99.17% and CE_Aurbach of 99.77%, respectively. Practically, the Li||LFP full cells with an N/P of 1.36 display stable cycling performance (94.3% capacity retention after 300 cycles with 99.79% average CE at 1 C) and improved rate performance. Even at ultra-low N/P ratios, the lean-Li||LFP full cells also maintain attractive cycling stability (90% retention, 99.9% average CE for 0.415 N/P after 158 cycles and for 0.107 N/P after 44 cycles), and it was found that the cycle number can be predicted by a linear relationship with the N/P ratio. Moreover, in commercial ester electrolyte, improved cycling stability of 4.3 V Li||NCM811 full cells can be obtained with additional CsTFA (80% retention for 1 N/P after 222 cycles). Our work paves a synergistic mechanism involving hetero cations and anions in electrolyte additives for dendrite-free, long-term cycling Li metal batteries, potentially applicable to other hetero-cation additives and alkali-metal batteries.

Methods

Materials and preparation of electrolytes

LiPF₆ (99%), LiTFSI (99%), LiBF₄ (99%), LiClO₄ (99%), LiFSI (99%), LiOTF (99%), LiNO₃ (99%), LiTFA (99%), NaTFA (99%), KTFA (99%) and CsTFA (99%) were all purchased from Adamas Reagent. The ether electrolytes (1 M LiTFSI in DOL/DME = 1:1 vol and 1 M LiTFSI in DOL/DME = 1:1 vol with 0.1 M LiNO₃) were obtained from Zhangjiagang Guotai-Huarong New Chemical Materials Corporation. The ester electrolyte (1 M LiPF₆ in EC/DEC = 1:1 vol and 1 M LiPF₆ in EC/DEC/FEC = 42.5:42.5:5 vol) were obtained from Dodochem. All the additives and electrolytes were stored and used in the glove box (O₂ < 0.01 ppm, H₂O < 0.01 ppm) without any purification. The LiFePO₄ (LFP) powder, LiNi_0.8Co_0.1Mn_0.1 (NCM811) powder, polyvinylidene fluoride binder (PVDF, HSV900), Super-P carbon black, copper foil (9 µm in thickness) and aluminum foil (16 µm in thickness) were obtained from Shenzhen-Kejing Corporation. N-methylpyrrolidone (NMP, >99.0%) was purchased from Aladdin. Li foils (20 μm and 450 μm in thickness, Li content ≥99.9%) were purchased from China Energy Lithium Co. Celgard 2500 (porosity: 41%; average pore size: 0.043 μm; thickness: 25 μm), 2032-type coin-cell cases, springs and spacers were obtained from Canrd Technology Co. Ltd. LFP positive electrodes with a series of loadings (from 13.9 to 22.5 mg cm⁻², inextricably) were prepared by casting the slurry consisting of LFP, Super-P and PVDF with a weight ratio of 85:7.5:7.5 in NMP onto aluminium foil. NCM811 positive electrodes (20.7 mg cm⁻²) were similar to LFP, except that the mass ratio was changed to 92:4:4.

Positive electrodes preparation and electrochemical measurements

CR2032 coin cells were assembled in an Ar-filled glovebox (O₂ < 0.01 ppm, H₂O < 0.01 ppm) to test the basic electrochemical properties of the electrolytes. All electrolyte preparations were carried out in the same glovebox. To prepare the additive-modified ether electrolyte, 0.05 M trifluoroacetate salts (LiTFA, NaTFA, KTFA or CsTFA) of 0.05 M were dissolved into SE or BE, denoted as SE + Li, SE + Na, SE + K, SE + Cs, BE + Li, BE + Na, BE + K and BE + Cs, respectively. Moreover, a range of concentrations of CsTFA were dissolved into BE, denoted as BE + Cs 0.005, BE + Cs 0.01 and BE + Cs 0.1. 0.05 M different anionic salts (LiPF₆, LiTFSI, LiBF₄, LiClO₄, LiFSI, LiOTF, LiNO₃ and LiTFA) were dissolved into SE or BE, denoted as SE + LiPF₆, SE + LiTFSI, SE + LiBF₄, SE + LiClO₄, SE + LiFSI, SE + LiOTF, SE + LiNO₃, SE + LiTFA, BE + LiPF₆, BE + LiTFSI, BE + LiBF₄, BE + LiClO₄, BE + LiFSI, BE + LiOTF, BE + LiNO₃ and BE + LiTFA. To prepare the additive-modified carbonate electrolyte, 0.05 M CsTFA was dissolved into SCS or BCE, denoted as SCE + Cs, BCE + Cs. The amount of electrolyte per cell was controlled at 30 µL, and Celgard 2500 was used as the separator.

Galvanostatic charge/discharge tests were carried out using a Neware battery cycler (CT-4008) at 30 °C in a climatic chamber. For Li||Cu coin cells, the cycle CE was measured by plating/stripping 1, 2, 3 or 5 mAh cm⁻² Li repeatedly at 0.5 mA cm⁻² with a charge cut-off voltage of 2 V. The modified Aurbach CE test was as follows: (1) perform one initial formation cycle with Li deposition of 5 mAh cm⁻² on Cu under 0.5 mA cm⁻² current density and stripping to 2 V; (2) subsequently deposit 5 mAh cm⁻² Li as a total Li reservoir (Q_T); (3) repeatedly strip/deposit Li of 1 mAh cm⁻² (Q_C) under 0.5 mA cm⁻² for 10 cycles; (4) strip all the residual Li (Q_S) to 2 V under 0.5 mA cm⁻² exhaustively. The CE_Aurbach was calculated according to the following equation:

$${\mathrm{CE}}_{\mathrm{Aurbach}}=\left(10\times {\mathrm{Q}}_{\text{C}}+{\mathrm{Q}}_{\mathrm{S}}\right)/\left(10\times {\mathrm{Q}}_{\text{C}}+{\mathrm{Q}}_{\mathrm{T}}\right)$$

Cyclic voltammetry (CV) and linear sweep voltammetry (LSV) measurements of Li||Cu half cells were conducted by a CHI 760E electrochemical workstation with different voltage ranges respectively, the testing temperature was set at 30 °C.

Li||Li symmetrical cells were assembled with two identical Li foils (450 µm in thickness, 10 mm in diameter) and cycled at specified current density and capacity in each half cycle. Cyclic capabilities of symmetric batteries were measured at 3 and 5 mA cm⁻² current density under a fixed capacity of 1 mAh cm⁻². Rate capabilities of symmetric batteries were tested at 0.5, 1, 2, 3, 5 and 1 mA cm⁻² of 1 mAh cm⁻². Tafel plots of Li||Li half cells were obtained from a CHI 760E electrochemical workstation by linear sweep voltammetry measurement at a scanning rate of 10 mV s⁻¹ within the voltage range of −0.12−0.12 V, the testing temperature was set at 30 °C. The exchange current density j₀ was calculated from the Tafel equation:

$$\mathrm{\eta }=\mathrm{A}\times \left(\log \mathrm{j}-\log {\mathrm{j}}_0\right)$$

where η and j were the overpotential and current, respectively, and A represents the kinetic constant in the diffusion-controlled surface dynamic range that could be acquired after fitting the data. The value of j₀ can be obtained from the intersection of the extrapolated linear part of the log j versus η plot with the μ = 0 line.

Electrochemical impedance spectroscopy (EIS) measurements of Li||Li half cells and Li||LFP full cells were conducted by a CHI 760E electrochemical workstation from 100 kHz to 100 mHz at the open-circuit potential with an amplitude of 10 mV after 30 min of stabilisation at open-circuit voltage. Activation energy measurements were tested by using Li||Li symmetrical cells through EIS measurements in an incubator under different temperatures from 5 to 30 °C with 5 °C separation. The values of Li deposition activation energy (E_a) were calculated by the following equation that could be extracted from the slope plot of log R_SEI vs. inverse temperature (1/T).

$${\mathrm{T}/\mathrm{R}}_{\text{SEI}}=\mathrm{A}\exp \left(-{\mathrm{E}}_{\text{a}}/\mathrm{RT}\right)$$

where E_a was the activation energy, T was the absolute temperature, R was the gas constant (8.314 J·mol⁻¹ K⁻¹), R_SEI was the interfacial Li⁺ transfer resistance obtained from the EIS plots, and A was the pre-exponential factor.

The Li||LFP full cells were assembled by matching LFP positive electrode (17.3 mg cm⁻²) with Li foil (20 µm in thickness), followed by activating at 0.2 C (current density: 1 C = 3.3 mA cm⁻², 170 mAh g⁻¹ for LFP) for 2 cycles and charging/discharging at 1.0 C in subsequent cycles. Each Li||LFP full cell was ejected with 30 µl of ether electrolyte and the voltage window was set within 2.8–4.0 V. The Li||NCM811 full cells were assembled by matching NCM811 positive electrode (20.7 mg cm⁻²) with Li foil (20 µm in thickness), followed by activating at 0.2 C (current density: 1 C = 4.1 mA cm⁻², 200 mAh g⁻¹ for NCM811 at 4.3 V) for 2 cycles, then charged at 0.33 C and discharged at 1.0 C in subsequent cycles. Each Li||NCM811 full cell was ejected with 30 µl of carbonate electrolyte and the voltage window was set within 3.0–4.3 V. The rate capabilities of LFP or NCM811 full cells were performed by charging/discharging 6 cycles at progressively increasing current densities, respectively.

For the assembly of Li||LFP full cells at ultra-low N/P ratios, some quantities of 0.4, 0.8 and 1.5 mAh cm⁻² Li were pre-electrodeposited onto the bare Cu foil (14 mm in diameter) through the Li||Cu cell protocol in advance, which were paired with LFP positive electrodes of different mass loadings (from 13.9 to 22.5 mg cm⁻², the specific ratios were shown in Supplementary Table 6) to meet the set N/P ratios. Li||LFP full cells at ultra-low N/P ratios were activated at 0.2 C for 2 cycles, followed by charging/discharging at 0.5 C in subsequent cycles. The electrolyte dosage for each cell was controlled at 30 µl and the voltage window was set within 2.8–4.0 V.

The Li inventory loss rate (LILR) in Li||LFP full cells was calculated by the following equation⁵⁸:

$$\mathrm{LILR}=100\times \left(1-\sqrt[\mathrm{n}\text{-}\mathrm{t}]{{\mathrm{Q}}_{\mathrm{n}}/{\mathrm{Q}}_{\mathrm{t}}}\right)$$

where where Q_n was the discharge capacity after n cycles, Q_t was the discharge capacity at the transition point (cycle number = t) and LILR was an averaged percentage loss of Li inventory per cycle.

The Li inventory retention rate (LIRR) in Cu||LFP full cells was calculated by the following equation⁵⁹:

$$\mathrm{LIRR}=100\times {\left({\mathrm{Q}}_{\mathrm{n}}/{\mathrm{Q}}_{\mathrm{int}}\right)}^{1/\mathrm{n}}$$

where where Q_n was the discharge capacity after n cycles, Q_int was the discharge capacity of the first cycle and LILR was an averaged percentage retention of Li inventory per cycle.

Materials characterization

SEM characterization is conducted on a Hitachi SU8010 scanning electron microscope to obtain the morphologies of the samples. Raman tests are carried out on NTEGRA Spectra AFM Raman Confocal SNOM instrument with laser wavelength of 633 nm. Ex-situ FT-IR spectra is collected on a Spectrum Two LiTa spectrometer FT-IR instrument (PerkinElmer, America), covering 4000–450 cm⁻¹ at 1.0 cm⁻¹ resolution. Atomic force microscopy (AFM) is carried out on a Bruker Dimension Fastscan microscope in peak force quantitative nanomechanics mode, and Young’s moduli are determined using the Derjaguin-Muller-Toporov model, the samples sent for AFM tests were obtained by depositing 1.5 mAh cm⁻² at 0.5 mA cm⁻² on Cu after cycling at 1 mA cm⁻² and 1 mAh cm^-2 for 5 cycles at 30 °C. In-situ FT-IR spectra is collected on a Nicolet iS50 spectrometer FT-IR instrument (Thermo Fisher Scientific, America) fitted with a single-reflection diamond attenuated total reflection (ATR) cell. A liquid-nitrogen-cooled MCT detector is employed at a 60° incidence angle with 4 cm⁻¹ spectral resolution. A lithium rod is positioned flush against the diamond window as the working electrode, while a lithium ring and lithium wire are served as the working electrode and reference electrode, respectively. A CHI660E electrochemistry workstation is used for potential control and current measurement. Li is plated at −0.1 V for 1200 s with tested electrolyte potentiostatically. ⁷Li and ¹³³Cs nuclear magnetic resonance (NMR) spectra are obtained on a BRUKER AVANCEIIIHD 500 Nuclear Magnetic Resonance spectrometer at 30 °C. Cryogenic-transmission electron microscopy (Cryo-TEM): the samples were obtained by depositing 0.2 mAh cm⁻² at 1 mA cm⁻² on copper grids after cycling at 1 mA cm⁻² and 1 mAh cm⁻² for 5 cycles. The Li-deposited copper grids are placed into a four hole TEM sample cassette and sealed under argon gas. It is then immersed in liquid nitrogen and disassembled (ensuring isolation from water and oxygen) and transferred to a TEM column for testing. The information resolution of the instrument is 0.23 nm and the acceleration voltage is 300 kV. Argon ion etching-assisted XPS spectra are received using a Thermo Fisher Scientific Model K-Alpha spectrometer equipped with Al Kα radiation (1486.6 eV). Time-of-flight secondary-ion mass spectrometry (TOF-SIMS) is performed on a PHI nano TOF III instrument. A 3 kV, 2 nA Bi₃⁺ beam rastered over 100 µm is served as the analysis probe, and a 2 kV, 100 nA Ar⁺ beam covering 400 × 400 µm² erode the sample for depth profiling. The erosion rate is calibrated at ~9.16 nm min⁻¹ on SiO₂. The samples sent for XPS and TOF-SIMS tests were obtained by depositing 1.5 mAh cm⁻² at 0.5 mA cm⁻² on Cu after cycling at 1 mA cm⁻² and 1 mAh cm⁻² for 5 cycles at 30 °C.

Nernst equation:

$${\mathrm{E}}_{\mathrm{reduction}}={\mathrm{E}}_{\mathrm{reduction}}^0-\frac{\text{RT}}{\text{tF}}\ln \frac{{\alpha }_{\text{reduction}}}{{\alpha }_{\text{oxidation}}}$$

where R denotes the universal gas constant (8.314 J K⁻¹ mol⁻¹), T stands for the absolute temperature, α represents the chemical activity of each species (α_reduction for the reduced form, α_oxidation for the oxidized form). F is the Faraday constant (9.6485 × 10⁴J V⁻¹ mol⁻¹), t is the number of moles of electrons exchanged. For dilute solutions, activity α can be approximated by concentration c, yielding the simplified form at 298.15 K:

$${\mathrm{E}}_{\mathrm{reduction}}={\mathrm{E}}_{\mathrm{reduction}}^0-\frac{0.0592}{\text{t}}{\log }_{10}\frac{1}{{\alpha }_{\text{oxidation}}}$$

Theoretical calculations

Classical molecular dynamics (MD) calculations are performed by using Forcite package in Materials Studio software 2020. The electrolyte modes of BE+Cs and BE are simulated by adopting amorphous cells with 56.2 × 56.2 × 56.2 ų and 56.4 × 56.4 × 56.4 ų linear dimensions, corresponding to salt-solvent ratios of 100:10:5:470:470 for 1.0 M LiTFSI + 0.1 M LiNO₃ + 0.05 M CsTFA-DOL/DME, and 100:10:470:470 for 1.0 M LiTFSI + 0.1 M LiNO₃-DOL/DME, respectively. With a fixed time step of 1.0 fs (femtosecond), the statistical averages are calculated from trajectories of at least 50 ps (picosecond) in length after 0.1 ps equilibration steps. The system temperature is set to 298 K with Nose thermostat. All production simulations are carried out in isothermal-isovolumetric (NVT) ensemble using the COMPASS III force field. DFT calculations are performed by DMol3 module Materials Studio software 2020. Geometries optimizations and electrostatic potential are performed with parameter hybrid method using the LEE-Yang-Parr correlation functional (B3LYP). Convergence is enforced as the maximum force falls below 2.0 × 10⁻³Ha Å⁻¹ (1 Ha = 27.21 138 eV), the total-energy change below 1.0 × 10⁻⁵Ha, and the largest displacement below 5.0 × 10⁻³Å. An all-electron DNP basis set was employed together with a 0.005 au smearing to accelerate SCF convergence. Calculations are conducted with an all-electron DNP basis set with a 0.005 au smear for faster convergence.The values of binding energies (E_binding) of Li⁺ with anion or solvent are calculated by the following formula:

$${\mathrm{E}}_{\mathrm{binding}}={\mathrm{E}}_{\mathrm{total}}-{\mathrm{E}}_1-{\mathrm{E}}_2$$

where E_n is the energy of the Li⁺-solvent complex, single solvent, respectively.

Reduction potentials are performed by Gaussian 16 package from M05-2X calculations within the SMD implicit solvation model (acetone, ε = 20.49). The Li⁺-complex geometry optimizations are performed with the pre-equilibrium structure at the B3LYP density functional and 6-31 + G** basis set. The reduction potentials (E_reduction) of complex within the SMD implicit solvation are calculated by the following formula^60–62:

$${\mathrm{E}}_{\mathrm{red}\mathrm{uction}}=-\left[\mathrm{\Delta }{\mathrm{G}}_{\mathrm{gas}}\left({\mathrm{M}}^{-}\right)-\mathrm{\Delta }{\mathrm{G}}_{\mathrm{gas}}\left({\mathrm{M}}^0\right)+\mathrm{\Delta }{\mathrm{G}}_{\mathrm{sol}}\left({\mathrm{M}}^{-}\right)-\mathrm{\Delta }{\mathrm{G}}_{\mathrm{sol}}\left({\mathrm{M}}^0\right)\right]/\mathrm{F}-1.4$$

where ΔG_gas(M^-), ΔG_gas(M⁰) are the free energies of reduced and neutral complexes in gas-phase at 298.15 K. ΔG_sol(M^-), ΔG_sol(M⁰) are the free energies of reduced and neutral complexes within the SMD implicit solvation model. F is the Faraday constant (9.6485 × 10⁴J V⁻¹ mol⁻¹).

Author contributions

H.Z. and Y.Q. conceived the idea and supervised the research. A.C. conducted the experiments. J.W. conducted the in-situ FT-IR characterization. L.W. conducted the TOF-SIMS characterization. A.C. and H.Z. analyzed the experiment results and wrote the manuscript. W.Y., P.H. and L.Z. discussed the results and commented on the manuscript.

Peer review

Peer review information

Nature Communications thanks Amruth Bhargav and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The datasets generated and analyzed in this work are included in this article and Supplementary Information. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-66197-7.