
==== Front
Proc Natl Acad Sci U S A
Proc Natl Acad Sci U S A
PNAS
Proceedings of the National Academy of Sciences of the United States of America
0027-8424
1091-6490
National Academy of Sciences

38261613
202313096
10.1073/pnas.2313096121
research-articleResearch ArticlechemChemistry410
Physical Sciences
Chemistry
Self-terminating, heterogeneous solid–electrolyte interphase enables reversible Li–ether cointercalation in graphite anodes
Xia Dawei a
Jeong Heonjae b c d
Hou Dewen e f
Tao Lei a https://orcid.org/0000-0002-8863-0044

Li Tianyi g https://orcid.org/0000-0002-6234-6096

Knight Kristin a
Hu Anyang a
Kamphaus Ethan P. c
Nordlund Dennis h
Sainio Sami h
Liu Yuzi f
Morris John R. a https://orcid.org/0000-0001-9140-5211

Xu Wenqian g
Huang Haibo i
Li Luxi g
Xiong Hui e https://orcid.org/0000-0003-3126-1476

Cheng Lei leicheng@anl.gov
b c 1 https://orcid.org/0000-0002-3902-1680

Lin Feng fenglin@vt.edu
a j 1 https://orcid.org/0000-0002-3729-3148

aDepartment of Chemistry, Virginia Tech, Blacksburg, VA 24061
bJoint Center for Energy Storage Research, Argonne National Laboratory, Lemont, IL 60439
cMaterials Science Division, Argonne National Laboratory, Lemont, IL 60439
dDepartment of Electronic Engineering, Gachon University, Sujeong-gu, Seongnam-si, Gyeonggi-do 13120, South Korea
eMicron School of Materials Science and Engineering, Boise State University, Boise, ID 83725
fCenter for Nanoscale Materials, Argonne National Laboratory, Lemont, IL 60439
gX-ray Science Division, Argonne National Laboratory, Lemont, IL 60439
hStanford Synchrotron Radiation Lightsource, SLAC National Accelerator Laboratory, Menlo Park, CA 94025
iDepartment of Food Science and Technology, Virginia Tech, Blacksburg, VA 24061
jDepartment of Materials Science and Engineering, Virginia Tech, Blacksburg, VA 24061
1To whom correspondence may be addressed. Email: leicheng@anl.gov or fenglin@vt.edu.
Edited by Alexis Bell, University of California, Berkeley, CA; received August 4, 2023; accepted November 17, 2023

23 1 2024
30 1 2024
23 7 2024
121 5 e231309612104 8 2023
17 11 2023
Copyright © 2024 the Author(s). Published by PNAS.
2024
https://creativecommons.org/licenses/by-nc-nd/4.0/ This article is distributed under Creative Commons Attribution-NonCommercial-NoDerivatives License 4.0 (CC BY-NC-ND).

Significance

Solid–electrolyte interphase (SEI) constitutes a crucial yet intricate component in rechargeable batteries. A traditional SEI facilitating outstanding reversibility in electrodes is required to be thin and homogenous. Herein, we discover that a self-terminating, heterogeneous interphase based on grainy LiF proves to be desirable for operating Li–solvent cointercalation chemistry. Taking advantage of the anisotropic characteristics of natural graphite, we conducted a holistic investigation into the nature, function, and formation of the heterogeneous SEI. Our findings not only shed light on the enigmatic interphase formed by ether electrolytes but also offer critical insights into future electrolyte design for graphite anodes operated under extreme conditions.

Ether solvents are suitable for formulating solid-electrolyte interphase (SEI)-less ion-solvent cointercalation electrolytes in graphite for Na-ion and K-ion batteries. However, ether-based electrolytes have been historically perceived to cause exfoliation of graphite and cell failure in Li-ion batteries. In this study, we develop strategies to achieve reversible Li–solvent cointercalation in graphite through combining appropriate Li salts and ether solvents. Specifically, we design 1M LiBF4 1,2-dimethoxyethane (G1), which enables natural graphite to deliver ~91% initial Coulombic efficiency and >88% capacity retention after 400 cycles. We captured the spatial distribution of LiF at various length scales and quantified its heterogeneity. The electrolyte shows self-terminated reactivity on graphite edge planes and results in a grainy, fluorinated pseudo-SEI. The molecular origin of the pseudo-SEI is elucidated by ab initio molecular dynamics (AIMD) simulations. The operando synchrotron analyses further demonstrate the reversible and monotonous phase transformation of cointercalated graphite. Our findings demonstrate the feasibility of Li cointercalation chemistry in graphite for extreme-condition batteries. The work also paves the foundation for understanding and modulating the interphase generated by ether electrolytes in a broad range of electrodes and batteries.

Li-ion batteries
graphite anode
cointercalation
solid-electrolyte interphase
ether electrolytes
USDA | National Institute of Food and Agriculture (NIFA) 100005825 Not available Feng Lin U.S. Department of Energy (DOE) 100000015 DE-AC02-06CH11357 Hui Xiong
==== Body
pmcGraphite has excellent structural tunability and is an intriguing material for studying guest–host interactions (1). For example, reversible Li-ion intercalation in graphite has spurred the rapid development of Li-ion batteries (LIBs). The early success of LIBs cannot be garnered without a protective solid–electrolyte interphase (SEI) layer formed on graphite cycled in ethylene carbonate (EC)–based electrolytes (2, 3). The operating voltage window of electrode materials can be beyond the stability window of electrolytes, necessitating a uniform, thin, and dense SEI (4, 5). The solvents explored in the earlier stage, including propylene carbonate (PC), 1,2-dimethoxyethane (DME, hereinafter referred to G1), and tetrahydrofuran (THF), were found to induce solvent cointercalation due to the limited SEI formation (6–8). Such a cointercalation process produces ternary graphite-intercalation compounds (t-GICs), which can be structurally stable depending on the type of solvents (9, 10). Compared to PC, acyclic ethers display excellent reductive stability and potential to realize reversible cointercalation (11, 12). Although Li–ether cointercalation has been reported since the 1990s, Li-t-GICs are perceived to be unstable, suffering from more severe degradation at large current densities (1, 13). Therefore, most efforts on exploring the cointercalation mechanism and understanding interphase formation have been focused on beyond-Li chemistries.

In 2014, Jache et al. used a diglyme-based electrolyte (1M NaCF3SO3, G2) to achieve a stage-I Na-t-GIC and 1000 cycles with a Coulombic efficiency (CE) of nearly 99.9% using natural graphite (14). Since then, long-cycling cointercalated graphite anode has been broadly shown for Na and K chemistries, which typically utilizes glyme solvents and PF6−/CF3SO3− anions (15–17). In these electrolytes, limited electrolyte decomposition occurs at the initial stage (17, 18). The thin and homogeneous SEI generated in these electrolytes is frequently used to explain the excellent cyclability and rate performance (19–21). On the other hand, some papers report that the surface of cointercalated graphite is fresh or almost fresh (18, 22). Similar controversies about the SEI nature are found in the research of hard carbon and metal stripping/plating when these glyme-based electrolytes are used (17, 23–26).

Despite the extensive attempts to utilize the cointercalation mechanism for rechargeable batteries, several key scientific questions have remained unanswered. First, it is generally believed that the reversibility and kinetics of Li-ion cointercalation are inferior to Na-ion and K-ion counterparts. This is somewhat contradictory to the opinion that cointercalation phenomena are more solvent-dependent (1, 10, 13, 15, 21, 27, 28). Second, the nature of SEI in the cointercalated graphite has been elusive. In previous studies on Na-ion and K-ion systems, the interphase is characterized by transmission electron microscopy (TEM) at the nanoscale (15, 18, 29). The spatial distribution of decomposition products at the statistically relevant scale is underexplored. Third, how ether electrolytes encounter reduction at the surface remains unknown. These scientific questions can be only addressed with a holistic design criterion of electrolytes and multidimensional analysis of the interphase (30).

In this study, we propose that long-cycling graphite anode based on the cointercalation mechanism can be enabled by a heterogeneous interphase. We formulate a new ether electrolyte, 1M LiBF4 G1, which can operate natural graphite with high capacity retention (~90%) after 400 cycles and excellent CE approaching 100%. We highlight that the anion dictates the electrochemical behaviors of graphite anodes in ether electrolytes by reductive stability and the as-generated SEI. Then, 1M LiBF4 G1 contributes to very limited electrolytes reduction. More importantly, the interphase exhibits heterogeneous spatial distribution: Large LiF particles (50 ~ 150 nm) occupy the edge plane and small LiF particles (<5 nm) are distributed sparsely on the basal plane. Computational results depict that the edge plane possesses catalytic effects, allowing for the reduction of anion and solvent molecules. Once the catalytic sites are covered by decomposed products, the reaction is self-terminated. In contrast, 1M lithium bis(trifluoromethanesulfonyl)imide (LiFSI) G1 results in rapid capacity decay for reductive instability of FSI− and homogenous SEI. Besides the understanding of interphase, we reveal the intrinsic excellent structural reversibility of graphite during Li–ether cointercalation by operando synchrotron X-ray diffraction (XRD) and coherent X-ray multicrystal diffraction (CMCD). Due to negligible desolvation and almost nonexistent SEI, the graphite anode shows no capacity difference with increasing current density. Our study has demonstrated the promise of reversibility and kinetics of Li–ether cointercalation chemistry, which have been underestimated. Meanwhile, we utilized the cointercalated graphite to exemplify a multidimensional investigation on the nature, role, and origin of SEI formed by ether electrolytes.

Results and Discussion

To identify the ideal electrolytes for Li–ether cointercalation, we propose a quadrant scheme based on the findings of the previous literature (SI Appendix, Table S1). First, we need to choose a salt that generates limited SEI. PF6−, CF3SO3−, and BF4− have outstanding reductive stability, enabling excellent CE (~99.9%) and cyclability in Na-ion and K-ion anodes, including Na/K-solvent cointercalation in graphite (31). In contrast, DFOB-, TFSI-, and FSI- are good film-forming anions. Thus, these anions are incompatible with the cointercalation mechanism (quadrants II and III in Fig. 1A) (32). Second, we consider the interference of the counter electrode (i.e., Li metal). The electrolytes in quadrant IV fail to provide stable cycling due to rapid degradation of Li metal anode (22). Compared to G2 and G4, G1 delivers much better electrochemical compatibility with Li metal (33). Third, LiBF4 is known to form discontinuous SEI composed of grainy LiF particles, failing to suppress carbonate reduction on graphite surface (34, 35). In principle, Li–ether cointercalation requires discontinuous SEI. Therefore, we propose that a combination of LiBF4 and G1 (quadrant I) is an ideal electrolyte to revisit Li–ether cointercalation in graphite.

Fig. 1. (A) Design strategies for reversible Li–solvent cointercalation in graphite (graphite–Li cells are used). The different combinations of solvent and salt in quadrants II, III, and IV lead to distinct fading mechanisms of the cointercalation chemistry. The limiting factor of battery performance is highlighted by the Strikethrough. (B) CV curves of graphite–Li cells at 0.2 mV/s. The second cycle is shown. The mass loading of graphite is similar (~2.5 mg/cm2). The cointercalation mechanism and desolvation mechanism are exemplified by 1M LiBF4 G1 and 1M LiFSI 1,3-dioxane (DOL), respectively (11). (C) The initial coulombic efficiency (ICE) in different electrode compositions and cycling conditions. XYZ represents the mass ratio between active material, carbon black, and binder, followed by the name of binder (SA: sodium alginate, PVDF: polyvinylidene fluoride). X + Y + Z = 10. The error bars are created based on the SD of at least three repeated cells. (D and E) Charge/discharge voltage profiles of graphite–Li cells using 901SA at 0.2 A/g in different electrolytes. (F) Cycling performance of graphite–Li cells using 901SA. The capacity and CE of the first cycle are not shown. (G) Summary of capacity retention in different electrolytes (22, 36).

We first performed cyclic voltammetry (CV) for graphite–Li cells (Fig. 1B). 1M LiBF4 G1 gives rise to multiple peaks between 0.3 V to 0.8 V, associated with the cointercalation behaviors and formation of ternary GICs (13, 37, 38). In contrast, desolvation-based mechanism is shown by three peaks below 0.1 V, suggesting the formation of binary GICs (39). The difference in the current density also implies the distinct storage capacity of two mechanisms (39). In constant-current tests, 1M LiBF4 G1 contributes to stable cycling of graphite anode with a ~93% capacity retention after 200 cycles at 1 A/g between 0.01 V and 2 V (SI Appendix, Fig. S1). In contrast, 1M LiFSI G1 triggers rapid capacity fading. To address the low ICE in LiBF4 G1, we then systematically investigated the influence of electrode composition and discharge cut-off voltage on ICE (Fig. 1C). We conclude that removing carbon black, replacing PVDF by SA, and increasing the lower cut-off voltage from 0.01 V to 0.1 V collectively enhance the ICE to >90%. Such an ICE is among the highest reported for cointercalated graphite in Li chemistry. The use of fluorine-free SA binder allows for fewer physically trapped Li-ions (40) and no fluorine interference during the surface analyses. The optimized anode delivers limited irreversibility during the first cycle and no capacity decay at 0.2 A/g (Fig. 1D and SI Appendix, Fig. S2). The potential-rebounding phenomenon in the first discharge when LiFSI is used (Fig. 1E) signifies an SEI nucleation process, which will be discussed later (41). In LiBF4 G1, the capacity retention reached ~92%, ~88%, and ~96% after 400 cycles at 0.5 A/g, 1 A/g, and 2 A/g, respectively (Fig. 1F). The average CE after the first cycle approaches 100%.

In brief, Li–G1 cointercalation can provide excellent cyclability at larger current densities, demonstrating the effectiveness of the Quadrant I electrolytes (Fig. 1G). SI Appendix, Fig. S3 supports that Li metal is not responsible for the drastic capacity decay of graphite–Li cells in 1M LiFSI G1. To further evaluate the effectiveness of the quadrant scheme, G2 and G4 electrolytes using LiBF4 (Quadrant IV) are measured (SI Appendix, Fig. S4). There was continuous capacity degradation due to the counter electrode failure: Li metal in 1M LiBF4 G2 electrolytes exhibits extreme instability (SI Appendix, Fig. S5). Though 1M LiFSI G2 has a higher CE, its behavior toward graphite anode is similar to that of LiFSI G1 because of the reductive instability of FSI− (36).

Structural and Compositional Understanding of the Cointercalated Graphite.

Next, we examined the morphological change of cycled graphite. After the first cycle, irreversible exfoliation happens (Fig. 2A and SI Appendix, Fig. S6) because of the Li–G1 cointercalation. XRD patterns (Fig. 2B) confirm that the Li–G1 cointercalation generally reduces crystallinity of the electrodes, as reflected by the broadened and weakened (002) peak. The graphite is not destroyed, different from the results in electrolytes (PC and DOL/DME mixture) inducing runaway solvent reduction (22, 42). The D/G ratio in Raman spectra (Fig. 2C) substantiates that LiBF4 G1 leads to higher disorder than 1M LiFSI G1. Such irreversible structural and morphological changes are not observed in graphite with the desolvation-based chemistry (SI Appendix, Fig. S7). In summary, LiBF4 G1 rapidly increases the bulk structural disorder in graphite but still enables long cycle life and high CE. When using LiFSI, the structural change seems to be retarded, agreeing with lower accumulative charge capacity. These data imply that the interphase properties directly impact the operational possibility of Li–G1 cointercalation in Li batteries.

Fig. 2. (A) SEM images of pristine graphite powder and graphite electrodes in 1M LiBF4 G1 and 1M LiFSI G1 after the first cycle. Ch stands for charged to 2.0 V. The scale bar of SEM images is 10 μm. (B) XRD patterns of cycled graphite at the charged state. Cu peaks are used for calibration. (C) Raman spectra of cycled graphite at the charged state. XPS spectra of cycled graphite in 1M LiBF4 G1 and 1M LiFSI G1: (D) Atomic concentration, where the error bars are created based on the SD of independent measurements on five different locations on the electrode. (E) High-resolution spectra-F 1 s. (F) Soft XAS spectra (C K-edge) of cycled graphite in 1M LiBF4 G1 and 1M LiFSI G1. 901SA was cycled at 0.2 A/g between 0.1 V and 2.0 V for 10 cycles (discharged state) and used for soft XAS and XPS.

We used X-ray photoelectron spectroscopy (XPS) to investigate the chemical compositions of the SEIs on graphite cycled using different salts. Atomic concentration (Fig. 2D) demonstrates a higher F content (11.6 ±   0.9% over 2.3 ±   0.2%) in SEI derived from 1M LiFSI G1 over 1M LiBF4 G1. It corroborates the strong reduction tendency of FSI−. Since BF4− is oxygen-free, the higher O content in the as-formed SEI can be attributed to the cointercalated G1 close to the surface. From high-resolution F spectra (Fig. 2E and SI Appendix, Fig. S8), we observe a strong Li–F peak located at 684.5 eV and a feature from B-F and S-F, consistent with the literature (43, 44). F K-edge soft X-ray absorption (soft XAS) spectra (SI Appendix, Fig. S9) show the characteristic peaks at ~690 eV and ~700 eV and the edge shift to higher energy, indicating the existence of LiF (11). In conclusion, XPS and soft XAS data show an interphase enriched with inorganic species in LiBF4 G1. This explains the ~10% irreversible capacity in the first cycle. Other suggestive data are the C K-edge (Fig. 2F) in total electron yield (TEY) mode with a 5–10 nm probing depth. The peak at 285.5 eV signifies the transition of C 1s to π* (sp2 bond) and the edge at 292.3 eV stems from the transition of C 1s to σ* (sp3 bond). The π* and σ* intensities follow the trend of pristine>1M LiBF4 G1>1M LiFSI G1, showing that graphite cycled in 1M LiFSI G1 is covered by a thicker SEI. A σ* resonance at 287.2 eV can be attributed to the C–O states originating from solvated ether between graphene layers (45). To conclude, trace electrolyte decomposition in LiBF4 G1 is related to full utilization of Li–ether cointercalation. However, XPS and soft XAS are ensemble-average techniques, thus the spatial distribution of reduction products is unclear.

Spatial Distribution of the Fluorinated Interphase at Different Length Scales.

We collected high-resolution TEM (HRTEM) images for cycled graphite after 10 cycles. LiBF4 G1 endows the cointercalated graphite with a clean surface and the noticeable lattice fringe of exfoliated graphite (Fig. 3A and SI Appendix, Fig. S10). In some regions (Fig. 3B and SI Appendix, Fig. S11), trace ultrasmall LiF particles (<5 nm) can be seen. In contrast, the graphite cycled in 1M LiFSI G1 is covered by a thick SEI (Fig. 3C and SI Appendix, Fig. S12). From the fast Fourier transform (FFT) image of the graphite domain, no clear lattice fringe is observed. This further corroborates that SEI vastly covers the surface of graphene layers. In some other regions (Fig. 3D), graphite (002) can be detected in FFT. At the 10th cycle, there is still 30 ~ 50% capacity retention in 1M LiFSI G1 cells because SEI has not covered the entire surface. We further conducted a TEM measurement for graphite cycled in 1M LiFSI G1 for one cycle and found that the surface is still not as clean as the graphite surface cycled for 10 cycles in 1M LiBF4 G1 (SI Appendix, Fig. S13 vs. Fig. 3 A and B).

Fig. 3. HRTEM images of graphite cycled in 1M LiBF4 G1 with inset FFT (fast Fourier transform) patterns: (A) Representative SEI-less surface. (B) Representative surface with LiF nanoparticles. HRTEM images of graphite cycled in 1M LiFSI G1: (C) Representative SEI-covered surface. (D) Representative surface with thick LiF coverage. The scale bar of HRTEM images is 10 nm. All FFT images are processed using the entire HRTEM image. STEM-EDS mapping of cycled graphite along the ab plane (basal plane): (E) 1M LiFSI G1. (F) 1M LiBF4 G1. (G) STEM-EDS mapping of cycled graphite along the c axis (i.e., nearly perpendicular to the ab plane) in 1M LiBF4 G1. The scale bar of images in E is 300 nm. The scale bar of images in F and G is 2 μm. (H) Color-split (Red) EDS mapping. The color bar shows the color gradation and represents the intensity of F. The scale bar is 200 nm. (I) Quantification of F image in H with x axis as intensity and y axis as the counts. The Inset image shows the Kurt value of color distribution in both electrolytes. The error bar in the 1M LiBF4 G1 is the SD generated by three different F-mapping images. Graphite is cycled at 0.2 A/g for 10 cycles (charged state).

To date, most studies characterize electrode surfaces at the nanoscale, decomposition products distribution at mesoscopic scale is also important. We obtained scanning-TEM (STEM)–energy dispersive X-ray spectroscopy (EDS) for the cycled graphite particles along the ab plane. Expanded graphite features can be seen from the high-angle annular dark-field (HAADF) images. F signal is widely dispersed throughout the graphite cycled in 1M LiFSI G1 (Fig. 3E). In stark contrast, cycled in LiBF4 G1 (Fig. 3F), LiF nanoparticles (50 ~ 150 nm) distribute heterogeneously on graphite. When observing along the c direction, we find that LiF nanoparticles mostly locate at the edge plane (Fig. 3G and SI Appendix, Fig. S14). A set of further processed EDS maps (Red channel only) is presented in Fig. 3H, more clearly demonstrating the agglomeration extent of LiF. To quantitively understand the LiF distribution, we plot the intensity distribution for the EDS map (Fig. 3I). We calculated the statistical parameter Kurtosis (Kurt) value (the Inset in Fig. 3I). A higher Kurt value corresponds to a larger extremity of deviations. The Kurt value of the former is much higher than the latter (11.7 vs. 0.3), further confirming the heterogeneity at the micron length scale. In summary, the LiF signals found in XPS and soft XAS originate primarily from these heterogeneously distributed LiF nanoparticles.

After examining the characteristics of the decomposition products on cointercalated graphite, we recommend that “pseudo-SEI” is a better term to describe the protective layer formed on graphite cycled in 1M LiBF4 G1. The reasons are shown in SI Appendix, Discussion.

Pseudo-SEI Formation Mechanisms and Implications.

Here, we discuss the formation mechanism of such a pseudo-SEI. We first applied the potentiostatic intermittent titration technique (PITT) to study the electrochemical reduction of different anions (Fig. 4A). In 1M LiFSI G1, the anion reduction (SEI nucleation) starts at the potential of 0.9 V. During FSI− reduction, each nucleus experiences two-dimensional growth and finally overlaps with others, similar to the trend reported elsewhere (41). In contrast, 1M LiBF4 G1 is free of the above characteristics. To further investigate the reductive stability of the electrolytes, we calculated the LUMO level of solvation complexes of LiBF4 and LiFSI with different numbers of solvents using DFT (Fig. 4B). With one G1 solvent, the contact ion pair (CIP) structure is more stable than the solvent-separated ion pair (SSIP) configuration. With 2 or 3 G1 solvents, the SSIP is more stable. The relaxed geometries on these electrolyte complexes are shown in SI Appendix, Fig. S15 and we report the LUMO of the most stable configurations in Fig. 4B. The LUMO level of LiBF4 in different structures is higher than LiFSI counterparts, suggesting the higher intrinsic reductive stability of the former.

Fig. 4. (A) PITT curves of graphite–Li cells in two electrolytes. The voltage range for 1M LiBF4 G1 and 1M LiFSI G1 are 1 to 0.75 V. The step of voltage is 0.05 V. Each voltage is held for 2 h. (B) LUMO energy of different Li–G1 complexes with two anions of interest. AIMD simulation results of 1M LiBF4 G1 in SSIP configuration on −1 charged basal plane: (C) snapshots of AIMD simulations in the sequence of 0 ps, 1 ps, 2 ps, and 10 ps; (D) charge transfer plots of different electrolyte species. AIMD simulation results of 1M LiBF4 G1 in SSIP configuration on neutral edge plane: (E) snapshots of AIMD simulations in the sequence of 0 ps, 1 ps, 2 ps, and 10 ps with highlighted anion reduction; (F) charge transfer plots of different electrolyte species. AIMD simulation results of 1M LiBF4 G1 in SSIP configuration on −1 charged edge plane: (G) snapshots of AIMD simulations in the sequence of 0 ps, 1 ps, 2 ps, and 10 ps with highlighted solvent reduction; (H) charge transfer plots of different electrolyte species.

To investigate the molecular origin for the observed heterogeneous distribution of LiF, we carried out ab initio molecular dynamics (AIMD) simulations of the electrolytes in both CIP and SSIP configurations on basal and edge planes in neutral and charged conditions. Note that the edge plane model used in our simulations is terminated with ketonic groups, which was previously shown to be the dominant termination for graphite edge surface (46). In the charged surface, an additional electron (−1 charge) was introduced to represent a higher discharge state or a lower potential. We found that electrolyte with salt in CIP configuration is stable on both the basal and edge planes in both neutral and charged states as there is no reaction observed within 10 ps of simulations of these systems. When the initial electrolyte configuration is SSIP, no reaction was observed on the basal planes under either neutral (SI Appendix, Fig. S16) or charged (Fig. 4 C and D) conditions within the 10 ps simulation time. In contrast, we observed interface reactions of the SSIP electrolyte on the edge planes under both neutral (Fig. 4 E and F) and charged (Fig. 4 G and H) conditions. On the neutral edge plane, the anion goes through cleavage of the B–F bond (Fig. 4 E and F, BF4− breaks into BF3 + F−) and the charge transfers from the graphite surface to the electrolyte species during the simulation time of 0 to 1 ps. Between 1 and 2 ps, the decomposed BF3 absorbs on the oxygen site of the graphite surface. Subsequently, a proton (H+) dissociates from one of the G1 solvents and combines with F- to form HF species. After 2 ps, the formed BF3 and HF molecules remain on the graphite surface without further reactions. On the charged edge plane, as shown in Fig. 4 G and H, G1 loses a proton that binds to the oxygen site of the graphite surface during the simulation time of 0 to 1 ps. After 1 ps, no further reactions were observed at the interface.

The relative reductive stability of electrolyte calculated by DFT and the surface-specific reactivity simulated by AIMD are consistent with the experimental findings that LiFSI is more reactive than LiBF4 and the edge surface is more active toward LiBF4 G1 decomposition than basal plane. When the catalytic site on the edge is passivated by pseudo-SEI, the reduction can be attenuated. Such a self-terminated mechanism explains the EDS mapping data and high CE (>99.5%) after a few cycles.

Structural Transformations and Reversibility of Graphite during Cointercalation.

To further elucidate the reversibility of cointercalated graphite cycled in 1M LiBF4 G1 and provide detailed phase transformation mechanisms, we conducted synchrotron operando XRD measurements. The contour plot and the voltage profile are displayed in Fig. 5 A and B, respectively. The structural change of cointercalated graphite in Li chemistry is complex but indicates similarity to Na systems (38). Initially, (002) fades away, corresponding to the exfoliation process (a). In section (b), a stage-3 GIC formation is observed supported by peak splitting of (002) to (005) and (006). Then, stage-3 GIC gradually changes to stage-2 GIC (c). The remarkable plateau at ~0.75 V in the voltage profiles is associated with the transformation of stage-2 GIC to stage-1 GIC (d). Fig. 5C demonstrates the two-phase transition clearly. Stage-1 GIC with high lithiation extent is featured by (001) and (002) peaks at lower angle as well as a very intense (003) peak. Next, further lithiation into stage-1 GIC happens (e). At the end of lithiation (f), more weak peaks are found between (001) and (002) as well as (002) and (003), which can be assigned to in-plane reconstruction (38). During delithiation, the recovery of expanded interlayer can be substantiated by the single (002) peak at the original location.

Fig. 5. (A) Operando synchrotron XRD patterns of graphite–Li cells cycled in 1M LiBF4 G1. The voltage range is 0 ~ 3 V. 1C is used (1C = 100 mAh/g). The X-ray wavelength is 0.024 nm. (B) The corresponding voltage profile (the first cycle) and different phase transformation steps are specified. (C) XRD patterns extracted from the operando measurements at specific potentials (the first lithiation). (D) Selected patterns of operando CMCD measurements. The X-ray wavelength is 0.103 nm. Superior graphite is used in operando XRD and CMCD measurements. (E) Schematic of the evolution of graphite based on cointercalation including the bulk and surface.

While XRD provides ensemble-averaged characterization for the electrode, CMCD probes crystal structure changes for a small number of particles, with each particle contributing to a bright spot on each diffraction ring. Therefore, the technique allows us to determine whether these particles have concurrent charging/discharging reactions. We obtained operando CMCD for the first two cycles of graphite–Li cells (Fig. 5D and SI Appendix, Fig. S17). The initial CMCD pattern displays bright diffraction spots, corresponding to the (002) planes of graphite particles. The bright spots gradually disappear when the voltage reduces to below 1.0 V, indicating a disorder formation. Upon both lithiation and delithiation, the evolution of CMCD rings is consistent with the observations in XRD patterns. Since all bright spots evolve in the same fashion, we conclude that the cointercalation reaction is concurrent between different graphite particles.

This operando XRD and CMCD data provide direct evidence for the very first time to demonstrate the structural reversibility of Li–G1 cointercalation in graphite and the homogenous reaction among primary particles. The structural change and interphase design based on 1M LiBF4 G1 are demonstrated in Fig. 5E, suggesting the path to full utilization of cointercalation capacity.

Rapid Interphase Kinetics Based on Reversible Li–ether Cointercalation.

We further tested the rate performance of cointercalated graphite. As the current density increases from 0.1 A/g to 4 A/g, there is almost no capacity fading by using 1M LiBF4 G1 (Fig. 6A). When cycled in LP57, the specific capacity drops notably and becomes lower than that in 1M LiBF4 G1 cells starting from 0.5 A/g. According to the voltage profiles (Fig. 6 B and C), cointercalated graphite shows minimal overpotentials and outstanding capacity retention. Furthermore, the kinetic of Li–ether cointercalation is comparable to that of the Na–ether cointercalation (Fig. 6D and SI Appendix, Fig. S18).

Fig. 6. (A) Rate capability test from 0.1 A/g to 4 A/g in 1M LiPF6 ethylene carbonate (EC)/ethyl methyl carbonate (EMC) 3:7 by weight (LP57) and 1M LiBF4 G1. Five cycles are performed for each current density. The capacity shown here is discharge capacity. Corresponding voltage profiles: (B) 1M LiBF4 G1. (C) LP57. (D) Summary of reversible cointercalation capacity in Na cells and Li cells at different current densities. The most representative Na-based electrolyte 1M NaPF6 G2 (diglyme) is used. (E) CV data of natural graphite cycled in 1M LiBF4 G1 at different scan rates. (F) Peak current dependence on the scan rate derived from the CV profiles. (G) The Nyquist plots of cycled cells in different electrolytes after 1st lithiation. (H) The cycling performance of a higher loading graphite anode (SA901) in 1M LiBF4 G1. (I) Schematic of the cointercalation process in graphite.

To understand the diffusion behavior of solvated-Li at different potentials, we collected CV data with multiple scan rates from 0.2 to 2 mV/s, as seen in Fig. 6E. There are three stages for the cointercalation (i–iii). The scan rate and peak current follow a power-law equation i = a vb, where i is the measured peak current, v is the voltage sweep rate, and a and b are adjustable parameters (Fig. 6F). A b value of 0.5 typically indicates a diffusion-controlled reaction while a b value of 1 suggests a capacitive reaction. A high b value of 0.94 is reported for 1.2 V to 0.8 V (iii), indicative of the supercapacitive behavior. In this region, interlayer spacing starts to increase, which can be viewed as in situ formation of slit micropore. The capacity is based on the absorption of solvated species onto the surface of micropores (37). (i) and (ii) have b values of 0.61 and 0.68, respectively, which means that they have mixed behaviors of adsorption and diffusion. Nyquist plots (Fig. 6G) show a 90-degree curve symbolizing the pseudocapacitive nature of cointercalation (47). The overall pseudocapacitive characteristic is consistent with the excellent fast-charge capability, which is linked to reduced graphite–cation interaction and repulsion among cations. Then, we increased the mass loading of graphite to examine if cointercalation maintains its excellent kinetics. LiBF4 has a lower dissociation extent, and the ionic conductivity of as-formed electrolytes is mediocre (SI Appendix, Fig. S19) (48). The graphite electrodes with higher loading (>7 mg/cm2) still exhibit similar capacity at high current density (98 mAh/g at 0.5 A/g and 90 mAh/g at 1 A/g, Fig. 6H). From the perspective of interface, the cointercalation has outstanding rate capability because it only needs to remove 0.5 G1 molecules from the first coordination shell (Li–2.5G1 to Li–2G1), depicted in Fig. 6I (49). This process corresponds to a low activation energy of 25 kJ/mol, which is much lower than that of EC-based electrolytes (60 to 70 kJ/mol) (50). Our results here demonstrate that, in contrast to the conventional wisdom, the Li–ether cointercalation chemistry can be reversibly operated even at large current densities. The finding here lays the foundation for LIBs working under extreme conditions via the cointercalation mechanism. Since Li cathodes offer higher energy density than Na and K cathodes, operating cointercalation chemistry in Li batteries is more meaningful than that in Na or K batteries.

Conclusion

Our study presents effective electrolyte and interphase design toward the discovery of reversible, fast Li–ether cointercalation in graphite. We quantitively revealed the nature, function, and formation of the interphase on graphite when cycled in 1M LiBF4 G1. We summarize crucial findings as below:1. The Quadrant scheme has illustrated why the reversibility of cointercalation in graphite was previously underestimated and guided the exploration of cointercalation for Li-ion batteries.

2. 1M LiBF4 G1 enables excellent cycling stability of cointercalated graphite anode, which complements our recently proposed 1M LiFSI DOL (11). Our studies collectively demonstrate that ether electrolytes using 1M salts can operate the graphite anode in Li-ion batteries by matching the reductive stability of electrolyte ingredients.

3. 1M LiFSI G1 cannot operate graphite due to the instability of FSI− and formation of SEI enriched with homogenously distributed LiF.

4. Unlike traditional continuous, thin, or uniform SEI, heterogeneous SEI covers a limited area of surface. It still enables the reversible graphite operation basically because of intrinsic thermodynamic stability of electrolytes.

5. Pseudo-SEI generated in 1M LiBF4 G1 prevents continuous electrolyte reductions. Pseudo-SEI may be extended to other systems using electrochemically stable electrolytes, for example, hard carbon in Na-ion batteries (51).

6. We experimentally and theoretically show that the electrolyte components are stable on the basal plane but become destabilized on the catalytic edge plane. The destabilization will self-terminate when the catalytic sites are covered by LiF.

7. We suggest that future efforts can focus on investigating the decomposition mechanism of anion species in ether electrolytes. More sophisticated models can be established in the future. For example, other functional groups and different types of defects can be introduced (52).

Supplementary Material

Appendix 01 (PDF)

Click here for additional data file.

The work was supported by Institute for Critical Technology and Applied Science at Virginia Tech and the Sun Grant program of the National Institute of Food and Agriculture, USDA, USA. The computational research was supported by the Joint Center for Energy Storage Research (JCESR), a U.S. Department of Energy (DOE), Energy Innovation Hub. We gratefully acknowledge use of the Bebop or Swing or Blues cluster in the Laboratory Computing Resource Center at Argonne National Laboratory. D.H. and H.X. acknowledge the support by the U.S. DOE, Office of Science, Office of Basic Energy Sciences program under Award Number DE-SC0019121. Work performed at the Center for Nanoscale Materials and Advanced Photon Source, both U.S. DOE Office of Science User Facilities, was supported by the U.S. DOE, Office of Basic Energy Sciences, under Contract No. DE-AC02-06CH11357. Graphite was produced at the U.S. DOE CAMP (Cell Analysis, Modeling and Prototyping) Facility, Argonne National Laboratory, which is fully supported by the DOE Vehicle Technologies Program within the core funding of the Applied Battery Research for Transportation Program.

Author contributions

D.X. and F.L. designed research; D.X., H.J., D.H., L.T., T.L., K.K., A.H., E.P.K., D.N., S.S., Y.L., J.R.M., W.X., H.H., L.L., H.X., L.C., and F.L. performed research; D.X., L.C., and F.L. analyzed data; and D.X., H.J., L.C., and F.L. wrote the paper.

Competing interests

A provisional patent containing some of the data presented in this work has been filed. Portions of the findings presented in this manuscript have been incorporated into a patent application submitted by Virginia Tech.

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.

Supporting Information

This article is a PNAS Direct Submission.
==== Refs
1 Y. Li, Y. Lu, P. Adelhelm, M.-M. Titirici, Y.-S. Hu, Intercalation chemistry of graphite: Alkali metal ions and beyond. Chem. Soc. Rev. 48 , 4655–4687 (2019).31294739
2 W. Liu, P. Liu, D. Mitlin, Review of emerging concepts in SEI analysis and artificial SEI membranes for lithium, sodium, and potassium metal battery anodes. Adv. Energy Mater. 10 , 2002297 (2020).
3 S. K. Heiskanen, J. Kim, B. L. Lucht, Generation and evolution of the solid electrolyte interphase of lithium-ion batteries. Joule 3 , 2322–2333 (2019).
4 Y. S. Meng, V. Srinivasan, K. Xu, Designing better electrolytes. Science 378 , eabq3750 (2022).36480630
5 N. Qin , Over-potential tailored thin and dense lithium carbonate growth in solid electrolyte interphase for advanced lithium ion batteries. Adv. Energy Mater. 12 , 2103402 (2022).
6 A. Naji, J. Ghanbaja, P. Willmann, D. Billaud, Electrochemical reduction of graphite in LiClO4-propylene carbonate electrolyte: Influence of the nature of the surface protective layer. Carbon 35 , 845–852 (1997).
7 J. O. Besenhard, M. Winter, J. Yang, W. Biberacher, Filming mechanism of lithium-carbon anodes in organic and inorganic electrolytes. J. Power Sources 54 , 228–231 (1995).
8 R. Yazami, Surface chemistry and lithium storage capability of the graphite–lithium electrode. Electrochim. Acta 45 , 87–97 (1999).
9 G. Yoon, H. Kim, I. Park, K. Kang, Conditions for reversible Na intercalation in graphite: Theoretical studies on the interplay among guest ions, solvent, and graphite host. Adv. Energy Mater. 7 , 1601519 (2017).
10 L. Li , Understanding high-rate K+-solvent co-intercalation in natural graphite for potassium-ion batteries. Angew. Chem. Int. Ed. Engl. 59 , 12917–12924 (2020).32298024
11 D. Xia , Design criteria of dilute ether electrolytes toward reversible and fast intercalation chemistry of graphite anode in li-ion batteries. ACS Energy Lett. 8 , 1379–1389 (2023).
12 M. Goktas , Stable and unstable diglyme-based electrolytes for batteries with sodium or graphite as electrode. ACS Appl. Mater. Interfaces 11 , 32844–32855 (2019).31397560
13 M. L. Divya, Y.-S. Lee, V. Aravindan, Solvent co-intercalation: An emerging mechanism in Li-, Na-, and K-ion capacitors. ACS Energy Lett. 6 , 4228–4244 (2021).
14 B. Jache, P. Adelhelm, Use of graphite as a highly reversible electrode with superior cycle life for sodium-ion batteries by making use of co-intercalation phenomena. Angew. Chem. Int. Ed. Engl. 53 , 10169–10173 (2014).25056756
15 H. Kim , Sodium storage behavior in natural graphite using ether-based electrolyte systems. Adv. Funct. Mater. 25 , 534–541 (2015).
16 K. Subramanyan, V. Aravindan, Towards commercialization of graphite as an anode for Na-ion batteries: Evolution, virtues, and snags of solvent cointercalation. ACS Energy Lett. 8 , 436–446 (2023).
17 Y. Li , Ether-based electrolytes for sodium ion batteries. Chem. Soc. Rev. 51 , 4484–4536 (2022).35543354
18 M. Goktas , Graphite as cointercalation electrode for sodium-ion batteries: Electrode dynamics and the missing solid electrolyte interphase (SEI). Adv. Energy Mater. 8 , 1702724 (2018).
19 Q. Wei , An ultrahigh-power mesocarbon microbeads|Na+-diglyme|Na3V2(PO4)3 sodium-ion battery. Adv. Mater. 34 , 2108304 (2022).
20 J. Chen , A desolvation-free sodium dual-ion chemistry for high power density and extremely low temperature. Angew. Chem. Int. Ed. Engl. 60 , 23858–23862 (2021).34463020
21 Z.-L. Xu , Tailoring sodium intercalation in graphite for high energy and power sodium ion batteries. Nat. Commun. 10 , 2598 (2019).31197187
22 H. Kim , Exploiting lithium-ether co-intercalation in graphite for high-power lithium-ion batteries. Adv. Energy Mater. 7 , 1700418 (2017).
23 L. Tao , Tailoring solid-electrolyte interphase and solvation structure for subzero temperature, fast-charging, and long-cycle-life sodium-ion batteries. Energy Storage Mater. 55 , 826–835 (2023).
24 H. S. Hirsh , Role of electrolyte in stabilizing hard carbon as an anode for rechargeable sodium-ion batteries with long cycle life. Energy Storage Mater. 42 , 78–87 (2021).
25 Z. W. Seh, J. Sun, Y. Sun, Y. Cui, A highly reversible room-temperature sodium metal anode. ACS Central Sci. 1 , 449–455 (2015).
26 B. Ma, Y. Lee, P. Bai, Dynamic interfacial stability confirmed by microscopic optical operando experiments enables high-retention-rate anode-free Na metal full cells. Adv. Sci. 8 , 2005006 (2021).
27 A. Inoo , Molecular structural influence of glymes on co-intercalation behavior of solvated Li+ in graphite electrodes. J. Electrochemical Soc. 168 , 060525 (2021).
28 J. Park, Z.-L. Xu, K. Kang, Solvated ion intercalation in graphite: Sodium and beyond. Front. Chem. 8 , 432 (2020).32509735
29 L. Wang , Graphite as a potassium ion battery anode in carbonate-based electrolyte and ether-based electrolyte. J. Power Sources 409 , 24–30 (2019).
30 D. Hou , Spatial and temporal analysis of sodium-ion batteries. ACS Energy Lett. 6 , 4023–4054 (2021).34805527
31 L. Zhou , Engineering sodium-ion solvation structure to stabilize sodium anodes: Universal strategy for fast-charging and safer sodium-ion batteries. Nano Lett. 20 , 3247–3254 (2020).32319776
32 R. Younesi, G. M. Veith, P. Johansson, K. Edström, T. Vegge, Lithium salts for advanced lithium batteries: Li–metal, Li–O2, and Li–S. Energy Environ. Sci. 8 , 1905–1922 (2015).
33 G. M. Hobold , Moving beyond 99.9% Coulombic efficiency for lithium anodes in liquid electrolytes. Nat. Energy 6 , 951–960 (2021).
34 Y. Aoki , Effective approach by computational chemical prediction and experimental verification to elucidate SEI formation mechanism in LiPF6-, LiFSI-, and LiBF4-containing electrolyte solutions. J. Phys. Chem. C 127 , 69–77 (2023).
35 C. Gong , The role of an elastic interphase in suppressing gas evolution and promoting uniform electroplating in sodium metal anodes. Energy Environ. Sci. 16 , 535–545 (2023).
36 P. Ma , Co-intercalation-free ether electrolytes for graphitic anodes in lithium-ion batteries. Energy Environ. Sci. 15 , 4823–4835 (2022).
37 G. Åvall , In situ pore formation in graphite through solvent co-intercalation: A new model for the formation of ternary graphite intercalation compounds bridging batteries and supercapacitors. Adv. Energy Mater. 13 , 2301944 (2023).
38 H. Kim , Sodium intercalation chemistry in graphite. Energy Environ. Sci. 8 , 2963–2969 (2015).
39 Y. Yamada, M. Yaegashi, T. Abe, A. Yamada, A superconcentrated ether electrolyte for fast-charging Li-ion batteries. Chem. Commun. 49 , 11194–11196 (2013).
40 M. Liu , Deciphering the paradox between the Co-intercalation of sodium-solvent into graphite and its irreversible capacity. Energy Storage Mater. 26 , 32–39 (2020).
41 C. Yan , Nucleation and growth mechanism of anion-derived solid electrolyte interphase in rechargeable batteries. Angew. Chem. Int. Ed. Engl. 60 , 8521–8525 (2021).33496038
42 J. Ming , New insights on graphite anode stability in rechargeable batteries: Li ion coordination structures prevail over solid electrolyte interphases. ACS Energy Lett. 3 , 335–340 (2018).
43 B. S. Parimalam, B. L. Lucht, Reduction reactions of electrolyte salts for lithium ion batteries: LiPF6, LiBF4, LiDFOB, LiBOB, and LiTFSI. J. Electrochem. Soc. 165 , A251 (2018).
44 M. Nie, B. L. Lucht, Role of lithium salt on solid electrolyte interface (SEI) formation and structure in lithium ion batteries. J. Electrochem. Soc. 161 , A1001 (2014).
45 V. A. Coleman , Defect formation in graphene nanosheets by acid treatment: An x-ray absorption spectroscopy and density functional theory study. J. Phys. D Appl. Phys. 41 , 062001 (2008).
46 N. N. Intan, J. Pfaendtner, Composition of oxygen functional groups on graphite surfaces. J. Phys. Chem. C 126 , 10653–10667 (2022).
47 T. S. Mathis , Energy storage data reporting in perspective—Guidelines for interpreting the performance of electrochemical energy storage systems. Adv. Energy Mater. 9 , 1902007 (2019).
48 T. Yoon, N. Chapman, D. M. Seo, B. L. Lucht, Lithium salt effects on silicon electrode performance and solid electrolyte interphase (SEI) structure, role of solution structure on SEI formation. J. Electrochem. Soc. 164 , A2082 (2017).
49 Y. Zhang , Investigation of ion-solvent interactions in nonaqueous electrolytes using in situ liquid SIMS. Anal. Chem. 90 , 3341–3348 (2018).29405699
50 T. Abe, H. Fukuda, Y. Iriyama, Z. Ogumi, Solvated Li-ion transfer at interface between graphite and electrolyte. J. Electrochem. Soc. 151 , A1120 (2004).
51 F. Zeng , Innovative discontinuous-SEI constructed in ether-based electrolyte to maximize the capacity of hard carbon anode. J. Energy Chem. 79 , 459–467 (2023).
52 H. Zhu , A comparison of solid electrolyte interphase formation and evolution on highly oriented pyrolytic and disordered graphite negative electrodes in lithium-ion batteries. Small 17 , 2105292 (2021).
