Cryo-ultramicrotomy enables TEM characterization of global lithium/polymer interfaces

Abstract

Lithium electrochemistry dictates the safety and cycling durability of lithium metal batteries. Although there have been numerous spectroscopy studies of lithium in liquid electrolyte batteries, electron microscopy studies of lithium in solid state batteries are scarce due to the technical difficulty of sample fabrication. In this work, we introduce an innovative specimen preparation methodology dubbed cryo-ultramicrotomy to fabricate large-scale cryo-TEM specimens that permit direct visualization of global anode/electrolyte interfaces and lithium within the polymer electrolyte from the nanometer to millimeter scale. By combining electron energy loss spectroscopy and multivariate least squares fitting, we map the lithium/PEO interface and the deposited lithium clusters, revealing a novel tri-layer SEI|Li|SEI structure on the lithium anode surface, indicating that electrons can tunnel through the SEI layer. The deposited lithium features either isolated or interconnected clusters that are different from the lithium dendrites in liquid electrolytes. The lithium clusters comprise Li⁰ and LiH cores surrounded by inorganic components such as Li₂O and LiOH but with little LiF, which are further enwrapped by the organic components. Based on this understanding, tetrabutylammonium fluoride was introduced into the electrolyte to tailor the SEI and PEO structures to significantly improve the electrochemical performance of lithium metal batteries.

---

Broader context statement

Lithium metal is the optimal choice for anodes in lithium batteries due to its high specific capacity and low electrochemical potential. However, the use of lithium anodes is hindered by uncontrolled dendrite growth and poor interfacial stability with the electrolyte. Therefore, understanding the behavior of lithium and its interaction with the electrolyte is crucial for the development of lithium metal batteries (LMBs). Unfortunately, electron microscopy characterization of the macroscopic lithium/electrolyte interface has been challenging due to the difficulty in sample fabrication. Additionally, the microgrid embedding approach is primarily applicable to liquid electrolyte systems, and its use in solid-state electrolytes has not been demonstrated. In this study, we present the first cryo-TEM analysis of lithium anodes and lithium/electrolyte interfaces in polymer solid-state batteries. We utilized the cryo-ultracryotomy technique to directly section polymer-based batteries. By taking advantage of the freeze-hardening behavior of polymers at low temperatures, we were able to preserve the as-deposited lithium and the lithium/polymer electrolyte interface, allowing for large-scale thin sample preparation. With this technique, we successfully sectioned almost the entire coin cell cross-section, providing a comprehensive view of the lithium electrode and the electrolyte interface for the first time.

Introduction

Lithium metal is the holy grail for the anodes of lithium batteries due to its unparalleled theoretical capacity (3860 mA h g⁻¹) and lowest electrochemical potential (relative to the standard hydrogen electrode: −3.04 V).^1–4 Despite its inherent advantages, the practical implementation of metallic lithium faces notable challenges.^5,6 Among them, the dendritic electrodeposition of lithium is responsible for the major hazards of internal short circuits, thermal runaway, and potential explosions within the battery. Additionally, a critical issue arising during cyclic operation is the formation of inactive lithium species, colloquially termed “dead” lithium, inevitably leading to capacity decay over time.⁷ Indeed, the intricacies of these challenges predominantly stem from the interplay of thermodynamic, kinetic, and mechanical phenomena at the electrochemical interfaces. During cyclic operation, an array of intricate physical and chemical changes occurs at the battery interfaces. These transformations comprise the formation of a solid–electrolyte interphase (SEI),^8–10 dendritic growth,^11–13 the emergence of space charge layers,¹⁴ and fluctuations in interface adhesion.¹⁵ Nevertheless, the heterogenous interface and SEI between the lithium metal anode and electrolyte are still the most important but unclear issues, primarily due to the absence of robust characterization methodologies. Fortunately, recent strides in cryogenic transmission electron microscopy (cryo-TEM) have opened up new possibilities for directly visualizing the lithium dendrite, “dead” lithium, interfaces and SEI,^16–21 although the studies are almost exclusively focused on liquid electrolyte batteries. Cui et al. demonstrated a novel approach for the characterization of SEI on lithium dendrites in liquid electrolytes using TEM by directly depositing lithium onto TEM grids embedded in coin cell batteries.²⁰ Their innovative method, combined with cryogenic electron microscopy, enabled non-destructive examination of the SEI at atomic resolution. Although the advancements in cryo-TEM facilitate non-destructive characterization, the prevailing techniques for TEM characterization of the SEI largely rely on lithium deposition on internally embedded copper grids within the liquid electrolyte battery. While the copper grid deposition method is well-suited for whisker lithium sampling, direct characterization of the macroscopic lithium anode/electrolyte interfaces has not been achieved to date. Furthermore, current investigations primarily focus on liquid electrolyte systems, and TEM studies on solid electrolytes, such as polymer electrolyte systems, are almost unexplored due to the difficulty in sample fabrication. Therefore, there is an urgent need to develop a TEM sample preparation approach capable of non-destructive and extensive sampling of lithium metal batteries. To this end, we propose a novel sample preparation and characterization methodology designed explicitly for solid polymer electrolyte (SPE)-based lithium batteries, which enables us to obtain a comprehensive understanding of the complex Li/SPE interfacial phenomena in SPE-based solid state lithium batteries at the nanometer to millimeter scale.

We employed the cryo-ultramicrotomy technique, commonly applied in the field of biology, to prepare cryo-TEM samples of SPE-based lithium batteries. The SPE assumes a viscoelastic state at elevated temperatures, affording substantial mobility to its macromolecular chains and facilitating lithium-ion conduction. Upon cooling below the glass transition temperature (T_g), the polymer's molecular chain segments undergo a gradual transition into a crystalline state, thereby enhancing its mechanical robustness, and rendering it an ideal candidate for cryo-ultramicrotomy samples. Due to their inherent properties, SPEs can serve as both electrolytes and embedding agents. Therefore, this technique can be widely applied to the study of various polymer-based batteries, including soft alkali metal negative electrode systems, embedded positive electrode active materials in polymers, and diverse composite electrolytes. In this study, the commonly used polymer PEO and metallic lithium were chosen to observe the pristine lithium anode/electrolyte interface and dendritic lithium within the electrolyte via cryo-ultramicrotomy. Additionally, harnessing the sensitivity of electron energy loss spectroscopy (EELS) to discern the distinctive “chemical fingerprints” of various lithium compounds,^22,23 we employed multivariate linear least squares (MLLS) fitting to spatially map the components of the SEI formed on the surface of the lithium anode and deposited lithium.

Results and discussion

Cryo-ultramicrotomy

The sample preparation and slicing procedures are depicted in Fig. 1a. A Li‖SPE‖Li symmetric cell was constructed for the investigation. The SPE was formulated by combining polyethylene oxide (PEO) and lithium bis(trifluoromethanesulfonyl) imide (LiTFSI). To obtain dendritic lithium that causes short circuits, the Li‖SPE‖Li battery underwent cyclic testing under a current density of 1 mA cm⁻² until it reached a soft-short circuit state²⁴ (manifesting as voltage fluctuations) (Fig. 1b). We disassembled the short-circuited battery within an argon gas-filled glovebox. A designated battery segment was extracted and positioned within a specialized slicing holder, and then the sample was transferred to a cryo-ultramicrotomy chamber for precise sectioning at −60 °C. Before sample sectioning, a crucial pre-processing step involved trimming the samples to eliminate any surplus lithium metal, ensuring that only a single-sided lithium/SPE cross-section remained (Fig. S1, Notes S1, ESI†). The sample was sectioned using a diamond knife to obtain lamellas with a thickness below 100 nm (Fig. 1c, the details are described in the Methods section). Fig. 1c shows a continuous lamella with a thickness of 300 nm (deliberately thicker for clarity). The prepared lamellas were affixed onto TEM grids and transferred to a cryo-specimen holder, preserving them in a liquid nitrogen environment to avoid contamination during the transfer process. To provide visual insight into the microstructure, Fig. 1d shows a cryo-TEM image, where a single lamella displays a total width of 132 μm, revealing the alternating lamellas of the lithium anode and SPEs in the longitudinal direction. In Fig. 1e, we present a bright-field cryo-TEM image, which offers a view of the nondestructive lithium anode/SPE interface with a thickness of approximately 200–300 nm. The slicing process induces compression along the cutting direction and leaves knife marks on the lithium surface due to inherent cutting defects of ultrathin sections, which may affect the actual thickness of the slice. The calculated actual thickness of the slice, determined via EELS plasma peaks, was 120.96 nm, which did not impact the experimental outcomes (Fig. S2, Notes S2, ESI†). Moreover, within the SPE matrix, embedded dendrites that had disengaged from the lithium electrode were also observed. We conducted multiple sectioning procedures to highlight the reproducibility of the experimental process (Fig. S3, ESI†).

Fig. 1 Sample preparation process and multi-scale imaging of the Li/SPE interfaces. (a) Schematic for cryo-ultramicrotomy sample preparation. (b) Voltage profiles corresponding to the characterized cells. (c) Serial lamellas obtained by ultrathin sectioning. The thickness of lamellas was standardized at 300 nm to enhance visibility. Scale bar, 0.1 mm. (d) Low-magnification cryo-TEM image of the lamellas, showing alternating distribution of lithium and SPE with short-range continuous lamellas. Scale bar, 40 μm. (e) Bright-field cryo-TEM image (color-enhanced) of the Li/SPE interface, where the light gray region represents the PEO electrolyte, the brown region indicates detached clusters from the negative electrode (white dashed outline), and the green area corresponds to the lithium negative electrode. Scale bar, 2 μm. (f) and (g) SEM images of the cryo-polished battery after soft-short circuiting. The entire cross-section remains flat over a wide area (f), with only slight knife marks visible on the lithium electrode (yellow) and clear visibility of dendrites entering the PEO (red). Zoomed-in image at the interface shows that most of the clusters (blue) have detached from the negative electrode and penetrated the PEO SEI (g). Scale bar, 100 μm, and 10 μm, respectively. (h) The 3D reconstructed morphology of the lithium anode and lithium cluster. For clarity, the PEO structure was removed, and only the lithium morphology was retained. The four perspective views demonstrate that after short-circuiting, substantial lithium clusters disengaged from the anode, infiltrating the electrolyte and manifesting as “dead” lithium.

To ensure the absence of contamination during both the sample preparation and transfer procedures, we conducted a suite of characterization via selected area electron diffraction (SAED) and EELS (Fig. S4, ESI†). The observed diffraction spots in Fig. S4b (ESI†) correspond to the (110) planes of metal Li, confirming the metallic lithium state of the anode. The EELS spectra also show the characteristic plasmon peaks and core-loss “chemical fingerprint” signals of metallic lithium (Fig. S4f, ESI†). This indicates that the metallic lithium anode remained stable during the slicing and transfer operations. The diffraction patterns from the Li–SPE interphase exhibit polycrystalline rings ascribed to Li, Li₂CO₃, and Li₂O components, which are generated by the Li–SPE interface reaction rather than contamination during the sample preparation and transfer (Fig. S4c, ESI†). The diffraction signals obtained from the SPE region display a characteristic broad amorphous halo (Fig. S4d, ESI†) and crystalline PEO domains (Fig. S5, ESI†). This observation highlights the coexistence of glassy and crystalline states in PEO, potentially stemming from the rapid nitrogen-assisted cooling during the slicing procedure. Notably, despite being at −173 °C for cryo-TEM characterization, PEO still displayed significant sensitivity to electron beam irradiation. At a dose of 0.24 e⁻ Å⁻², PEO exhibited discernible mass loss, considerably constraining imaging conditions and impeding the acquisition of high magnification or high-resolution transmission electron microscopy (HRTEM) images within the PEO matrix (Fig. S6a, ESI†). When the dose was lower than this critical value, non-discernible damage was observed in both the lithium and SPE regions (Fig. S6b, ESI†). This stringent imaging approach ensures negligible beam damage to the sample.¹⁶

The cryo-ultramicrotomy methodology is also suitable for preparing smooth large-scale cross-section samples of the SPE-based lithium batteries for scanning electron microscopy (SEM) characterization. The SEM images provide an almost complete overview of the entire battery cross-section, effectively capturing details ranging from the millimeter to nanometer scales (Fig. 1f and g). Furthermore, the EDX mappings affirm the absence of contamination during the sample preparation and transfer processes (Fig. S7, ESI†). Distinct from previously reported whisker lithium,²⁵ the lithium in the SPE features a cluster morphology with dimensions ranging from 2 to 5 μm, distributed relatively uniformly in the entire SPE (Fig. 1f, g and Fig. S8, ESI†). The clusters are either isolated from one another and detached from the lithium anode or interconnected, forming isolated “dead” lithium or “dead” lithium clusters, respectively. The cluster morphology is possibly caused by the high mechanical resistance of the PEO electrolyte, which prevents the formation of long whisker-shaped dendritic lithium that is usually seen in liquid electrolyte batteries. In situ optical microscopy (OM) experiments (Movie, Fig. S9, ESI†) show similar cluster morphologies in the SPE. In situ OM experiments also show an electric field-driven flow of PEO that may be attributed to the synchronized movement of ether backbones within the PEO chains, accompanied by Li⁺ migration, thereby inducing the overall flow of the PEO matrix.²⁶

To gain a comprehensive understanding of the three-dimensional (3D) architecture of lithium clusters, we conducted continuous sectioning of the SPE after a short circuit of the cell via a focused ion beam (FIB) and reconstructed the cross-sectional images (Fig. 1h and Fig. S10, ESI†). The resulting 3D spatial distribution indicated that a significant proportion of clusters detached from the lithium surface were isolated in the SPE. Some of these lithium clusters are interconnected and aggregate into sizable dendritic clusters, while others remain discretely dispersed within the SPE matrix. We conjecture that these fractured cluster lithium entities represent a pivotal factor underlying internal soft short-circuit incidents in batteries. Throughout electrochemical cycling, as the SPE experiences migration, these lithium clusters intermittently engage or disengage. Upon their widespread distribution within the SPE structure, they cause voltage perturbations, thereby inducing an internal soft-short circuit state. Notably, conventional sample preparation techniques for lithium anodes, such as cleaning, drying, and TEM grid deposition, only characterize a very localized region. In contrast, the cryo-ultramicrotomy methodology provides a global view of the anode/electrolyte interface, SEI, and electrolyte. Moreover, the chemical environment within the battery was preserved, thereby facilitating a comprehensive exploration of the native state of the SEI and dendritic lithium in the cycled battery.

Li anode interface

The SEI on the lithium anode exhibits sophisticated structure and composition, typically comprising a variety of lithium compounds.^27,28 We employed EELS, including the element distribution and Li-K edge energy loss near edge structure (ELNES), to scrutinize the composition and spatial distribution of SEI components in the lithium anode/SPE interface. The lithium K-edge ELNES spectra of various lithium compounds, which are often the major components of the SEI, were acquired under similar experimental conditions for reference (Fig. S11, ESI†). Metallic lithium features a K-edge peak at 55.9 eV, while the lithium-containing oxygen compounds (Li₂O, LiOH, Li₂CO₃) are characteristically present as various white-line peaks. Fig. 2a and b presents the bright-field images and the associated EELS mapping at the Li/SPE interface. The EELS mapping uncovered an oxygen-enriched layer on the anode side. Intriguingly, a tri-layer configuration of SEI|Li|SEI was observed on the anode surface. The Li-K edge linear scan spectra spanning from the anode's interior to the electrolyte region further unveil the composition of the tri-layer structure (Fig. 2c). Probing from the anode, the initial layer labeled as SEI₁, spanning 0–480 nm, exhibits a composite constituent of both metallic lithium and lithium oxides. The signal intensity of the characteristic loss peak (55.9 eV) of metallic lithium declines while that of the SEI components (59.3 eV) intensifies. In the subsequent layer, which ranges from 480 to 560 nm, the SEI signal declines. At an offset of 560 nm from the anode, a prominent metallic lithium signal reemerges. The third layer labeled SEI₂, persists beyond 560 nm, and is predominantly composed of SEI until the lithium signal from the SPE nearly vanishes.

Fig. 2 EELS characterization of the electrode/SPE interface. (a) A BF image of the negative electrode/PEO interface, where the green area corresponds to the lithium anode electrode. (b) Elemental mapping corresponding to (a). The anode interface exhibits a tri-layered structure. (c) EELS line scan at the interface (marked in a), showing the gradual enhancement of the white line peak as lithium transitions from the anode electrode to the electrolyte, with the reappearance of a metallic lithium signal at 560 nm, consistent with the mapping. (d)–(f) The plasmon signals in various regions, with the volume plasmons of the metallic lithium appearing at 7.5 eV. It is relatively weak in SEI₁ (d), exhibits maximal sharpness in the lithium deposition layer (e), and is nearly absent in SEI₂ (f). (g)–(i) C-K edge signals in different regions, it is absent in SEI₁ (g) and the lithium deposition layer (h) but is prominent in SEI₂, corresponding to Li₂CO₃ (i). (j)–(l), O-K edge signals in different regions, both SEI₁ (j) and SEI₂ (l) exhibit clear peaks corresponding to Li₂O, and this oxygen signal is notably distinguishable from the oxygen signals associated with artificial contamination (j). (m) A schematic diagram illustrating the formation of a tri-layered structure. Potential pathways for lithium ions at the negative electrode: (1) traversing the SEI to reach the negative electrode; (2) combining with tunneling electrons to form metallic lithium; and (3) engaging in solvent reactions to generate the SEI.

The corresponding plasma loss peaks along the line scan corroborate with the Li-K edge results, confirming the tri-layer structured SEI. Fig. 2d–f presents distinct profiles for the plasmon loss peaks in the SEI₁ layer (Fig. 2d), the lithium deposition layer (Fig. 2e), and the SEI₂ layer (Fig. 2f). The lithium deposition layer features a volume plasmon (VP) loss peak for metallic lithium at 7.5 eV, accompanied with diminished intensity from lithium compounds in the SEI₁ and a near absence of the lithium compounds in the SEI₂. Fig. 2g–i displays the C-K edge spectroscopic profiles for each distinct region. C is absent within SEI₁ (Fig. 2g) and the lithium deposition layer (Fig. 2h), while conspicuous C-K peaks emerge in SEI₂ (Fig. 2i), at 294 eV and 305 eV, corresponding to C=O and C–O chemical bonds, respectively, which is consistent with the spectral characteristics of carbon in Li₂CO₃. Fig. 2j–l portrays O-K edge spectra, wherein the oxygen signal within the lithium deposition layer is weak (Fig. 2k). Conversely, both SEI₁ (Fig. 2j) and SEI₂ (Fig. 2l), exhibit more pronounced oxygen signal peaks at 535 eV and 541 eV, respectively, in agreement with the oxygen spectral features in Li₂O. To exclude possible oxygen contamination from the TEM column, comparative oxygen spectra from the metallic lithium bulk and SEI₁ were acquired (Fig. 2j). The results indicate no discernible oxygen peaks in the metallic lithium bulk, in sharp contrast to the prominent oxygen peaks within the SEI. These results indicate that the tri-layer structure on the lithium anode surface is a direct consequence of the electrochemical reaction.

EELS further reveals that the inorganic constituents of SEI₁ consist predominantly of Li⁰ and Li₂O. Beyond SEI₁ lies the lithium deposition layer, characterized by the presence of metallic lithium. Finally, the SEI₂ layer, in contact with the SPE, is composed primarily of Li₂O and Li₂CO₃.

The electron diffraction results further confirm this finding (Fig. S12, ESI†). We conjecture that the initial coexistence of Li and Li₂O within the SEI₁ may arise from the lithium–SPE interfacial reaction during electrochemical cycling. This stage is characterized by SEI gradient enhancement concurrently with the presence of partially unreacted metallic lithium. Ideally, Li⁺ traverses through the SEI₁ to reach the anode electrode; however, due to the limited ionic conductivity of the SEI layer, the transport of Li⁺ through the SEI to the anode is hindered. However, electrons may tunnel through the SEI layer and combine with Li⁺ to form a thin lithium layer. The freshly deposited lithium continues to react with the SPE, forming the constituents of SEI₂ (Fig. 2m). The prevailing belief regarding lithium dendrite growth is that the rupture of the loosely packed SEI layer at the negative electrode interface leads to dendritic growth.¹ However, our observations indicate that in the case of a SEI layer with poor ionic conductivity, lithium can be directly deposited atop the SEI layer, possibly due to the weak electronic conductivity of the SEI layer. Recent experiments by Balbuena et al. have indeed demonstrated the electronic conductivity of the SEI, confirming our findings.²⁹ Thus, in addition to the mechanical properties of the SEI, its ionic conductivity appears to be more critical for the uniform deposition of lithium.

Lithium clusters

Fig. 3a shows an annular dark-field (ADF) image of the morphology of the lithium clusters detached from the electrode surface (white dashed line). EELS line-scan analysis of an individual cluster reveals its composition (Fig. 3b): the Li⁰ region (A1), the Li⁰-side SEI region (A2), and the SPE-side SEI region (A3). Fig. 3c illustrates the distinct Li-K edge ELNES spectra for these three regions. In the A1 region, discernible peaks manifest the presence of metallic Li. In the A2 region, the Li-K edge spectra feature an additional white-line peak, which reflects the inorganic SEI components. In the A3 region, the Li signals diminish, suggesting the plausible presence of primarily carbon-based organic lithium compounds and other species. Fig. 3d presents the K-edge EELS results for Li, C, and O. In the A1 and A2 regions, Li elemental signals are evident, supported by their respective Li-K ELNES, associating them with Li⁰ and lithium compounds. Meanwhile, the A3 region demonstrates prominent C element signals, indicative of organic SEI components. Concurrently, the O elements are predominantly concentrated in the A2 region, which are consistent with the prevalence of inorganic SEI components rich in lithium oxides. To mitigate uncertainty, we obtained EELS element mappings from various regions within the sample, revealing a notable oxygen-enriched shell encircling the Li component (Fig. S13, ESI†). Through EELS line scan analysis of the lithium clusters, we suggest that the inorganic components form a compact SEI layer, primarily composed of oxygen-based compounds. Simultaneously, the organic components are situated at the surface inorganic layer, creating a transitional zone between the SEI and the SPE. This structure is similar to that of dendrites in liquid electrolyte environments.³⁰

Fig. 3 EELS characterization of the lithium clusters. (a) An ADF image of a lithium cluster in the electrolyte. (b) An ADF image of a single cluster used for EELS analysis, with the pink arrow denoting the line scan region and direction. (c) Li-K edge signals from three regions, all subjected to deconvolution to eliminate multiple scattering effects and background subtraction. (d) Li, C, and O K-edge signals in the line scan region. All scale bars in the figures represent 1 μm. (e) An ADF image of the fitting region and the overall morphology of the lithium cluster. Scale bar, 2 μm. (f) MLLS fitting results and corresponding fitting residuals for the six lithium compounds. Li and LiH are situated in the central region, while Li₂CO₃, Li₂O, and LiOH collectively constitute the SEI, with LiF predominantly distributed in the outer region. (g) MLLS fitted spectra, with blue shading representing the acquired spectra and black curve denoting the overall fitting results. (h) Statistical results of the fifth-order fitting coefficients for the six chemical species obtained from five separate measurements. (i) Schematic diagram of the structure of lithium cluster.

Multivariate linear least squares (MLLS) fitting was conducted to further illustrate the spatial distribution of various inorganic SEI components. Six standard core-loss spectra, corresponding to Li⁰, LiH, Li₂O, Li₂CO₃, LiOH, and LiF, were used as reference spectra for fitting. Fig. 3e exhibits the ADF image of the lithium cluster, which detached from the anode (a part of the area shown in Fig. 1f). In Fig. 3f, the MLLS analysis results for the EELS-acquired regions are illustrated. The analysis reveals the coexistence of the Li⁰ and LiH phases in the cluster. The SEI layer surrounding the cluster comprises LiOH, Li₂O, and Li₂CO₃. Moreover, partial detachment of LiOH and Li₂CO₃ from the SEI layer is observed. The presence of LiF is minimal, and it is distributed at the outskirts of the SEI layer. In Fig. 3g, we present the spectral profiles obtained through MLLS fitting. The fitting was carried out in the energy range of 50–75 eV, with background removal and multiple scattering corrections for all spectra. The blue-shaded region corresponds to the EELS spectra acquired during the experiment, and the black curve represents the overall fitted curve. The results reveal a close correspondence between the fitted curves and the experimental spectra for the entire acquisition region, with residuals remaining below 5%. The fitting coefficients in MLLS provide insights into the relative contributions of each component. By summing and normalizing these coefficients for each component, we obtained the relative proportions of SEI components across the region of interest. An additional set of four spectral images from distinct regions (regions 2–5) was obtained to confirm the reproducibility of the experiments (Fig. S14–S17, ESI†). In Fig. 3h, we present the normalized fitting coefficients for the different SEI components. Our statistical analysis highlights LiOH and Li₂O as the predominant components, reaching maximum proportions of 64.3% and 43.4% in the five acquisitions, with average values of 26.8% and 32.6%, respectively. In contrast, LiF shows the lowest content, consistently below 5% in all cases, with an average value of 3% (Table S1, ESI†). Based on the combined analysis of proportions and spatial distributions, it is apparent that the SEI structure primarily consists of Li₂O, LiOH, and Li₂CO₃ as the main inorganic components (Fig. 3i). Among these, Li₂O forms a closely adhering layer on the Li⁰ surface, and LiOH shows the highest content. Conversely, the LiF content remains relatively low and is not densely distributed within the SEI layer. Furthermore, we confirmed the presence of LiH in the SPE-based lithium batteries. Notably, prior literature has already demonstrated the existence of LiH in the SEI of liquid electrolytes,¹⁶ where it plays a pivotal role in influencing the electrochemical performance.³¹ The emergence of LiH is conventionally linked to electrolyte decomposition, leading to hydrogen gas generation that permeates the SEI and triggers corrosion of the Li⁰. Our findings confirm this hypothesis in SPE-based lithium batteries, as LiH is observed to be homogeneously dispersed within the SEI layer, coexisting with Li⁰, signifying the occurrence of hydrogen gas permeation through the SEI, thereby inducing Li⁰ corrosion. These observations provide compelling evidence that polymer electrolytes are susceptible to partial degradation, resulting in the evolution of hydrogen gas even at low voltage conditions.

Moreover, apart from its presence within the lithium cluster regions, we also identified a conspicuously enhanced distribution of LiH on the Li anode, further validating the degradation of PEO (Fig. S18, ESI†). Prior investigations have elucidated that PEO undergoes degradation under high voltage conditions at the cathode,^32–35 while lithium salts preferentially acquire electrons under low voltage levels at the anode,^36,37 thereby engaging in interface reactions that lead to their degradation. In this study, however, the inorganic components within the SEI predominantly manifest as LiOH, Li₂CO₃, and Li₂O species. The presence of OH⁻ in the SEI should be traced back to the chain termini of PEO rather than LiTFSI. Moreover, the formation of LiH is solely reliant on the degradation of PEO, which induces hydrogen evolution. Although it remains plausible that Li₂CO₃ and Li₂O arise as degradation products of lithium salts, the lack of large quantities of LiF in the SEI does not appear to support this hypothesis. Therefore, the SEI components stem primarily from the degradation of the PEO macromolecular chains.

Distribution of LiF

The formation of SEI on the lithium dendrites involves the premature degradation of PEO, resulting in the production of SEI components like LiOH and Li₂CO₃, which are considered unfavorable for battery cycling,^18,28 while the formation of more favorable SEI products such as LiF^38,39 is restricted. To elucidate the distribution and role of LiF, we used tetrabutylammonium fluoride (TBAF) as a potent nucleophilic fluorinating reagent, which can act as a viable source of F⁻ ions to generate additional LiF. Upon introducing 5% TBAF into SPE, we constructed a symmetrical cell and performed cycling to reach a state of soft-short circuit under a current density of 0.5 mA cm⁻² (Fig. S19, ESI†). To observe the generation and distribution of LiF while preventing rapid battery short-circuiting, we employed lower current densities (0.5 mA cm⁻²) during cycling. Subsequently, the samples were prepared using identical methodologies as described above. Fig. 4a portrays the mapping of the anode/SPE interface and the lithium clusters employing the MLLS technique after cycling with 5% TBAF. Notably, besides the predominant components of Li₂O, LiOH, and Li₂CO₃ within the interface and SEI, the spatial distribution of LiF was observed to significantly increase; however, it remained absent in the vicinity of the lithium anode or dendrites. Instead, it exists independently in the form of polycrystalline particles within the SPE matrix (Fig. S20, ESI†). The EELS analysis further corroborates this observation, revealing a robust peak at approximately 62 eV in the SEI region (Fig. 4b), which is characteristic of the LiF fingerprint. To strengthen our observations, we assessed the content of six inorganic species following TBAF incorporation. By sampling two additional regions (regions II and III) and calculating their respective mean values (Fig. S21 and S22, ESI†), the reliability and accuracy of our findings are reinforced (Fig. 4c). The findings indicate that following the introduction of TBAF, there was a substantial increase in the mean LiF content from 2.9% to 12.2%. Simultaneously, the proportions of Li₂O, LiOH, and LiH experienced marginal declines, decreasing by 5.8%, 3.5%, and 6.9% respectively (Table S2, ESI†). This result signifies the impact of TBAF addition, which serves as an effective F⁻ source, fostering the formation of LiF. The consequence of the improved SPE on the cell performance is graphically presented in Fig. 4d, wherein the Li/Li symmetric cells' cyclic performance is depicted before and after TBAF introduction. Under the sustained cycling conditions of 0.1 mA cm⁻² for 900 hours (Here, the areal capacity is different from the areal capacities of 1 mA h cm⁻² and 0.5 mA h cm⁻² used during characterization. The reason for using a larger areal capacity in the characterization is that it generates a sufficient quantity of dendrites throughout the cycling process for ease of sample preparation), the cell without TBAF experienced a progressive increase of overpotential, which was attributed to ever-increasing interface impedance due to the formation of an unsatisfactory SEI. This demonstrates the unavoidable interface deterioration. In stark contrast, the cells treated with TBAF demonstrate stable cycling. Despite the significant improvement in battery stability with increasing LiF content, the TEM results indicate that LiF does not exist within the dense layer of the SEI but rather exists independently within the SPE matrix. This should not be considered a contributor to the SEI composition, a view consistent with previous research.³⁸ Therefore, the role of LiF cannot be simply attributed to SEI contributions, suggesting a potentially more complex underlying mechanism.

Fig. 4 Cryo-TEM characterization and electrochemical testing after the addition of 5% TBAF. (a) EELS-MLLS fitting of the negative electrode after the addition of 5% TBAF to PEO. The generated LiF is not found in proximity to lithium. Scale bar, 500 nm. (b) MLLS fitting spectra of the SEI constituents, exhibiting a distinct LiF signal peak. (c) Variation in the content of individual SEI components before and after the addition of TBAF, indicating an increase in the LiF content and a decrease in the LiH, Li₂O, and LiOH content. (d) Li–Li symmetric cell cycling performance before and after the of addition TBAF.

Conclusion

A cryo-ultramicrotomy technique was implemented to prepare large-scale cryo-TEM samples for SPE-based solid state lithium batteries. We unveil the existence of a tri-layered architecture at the lithium anode surface, characterized as SEI₁|Li|SEI₂, showing direct evidence that lithium can be deposited atop the SEI layer, indicating that electrons can tunnel through the SEI layer. Additionally, we show that the deposited lithium in SPE features an isolated or interconnected cluster morphology, which is distinctly different from the whisker-shaped dendritic lithium observed in liquid electrolyte batteries. The lithium clusters form a multilayer structure comprising LiH and Li⁰ species, predominantly occupying the innermost layer, followed by Li₂O in the succeeding layer, whereas the outermost layer consists of LiOH and Li₂CO₃ species. Based on this understanding, we introduced TBAF into the electrolyte to facilitate the formation of nanocrystalline LiF within the SEI and electrolyte, which improved the battery performance significantly. Our results not only reveal new lithium electrochemistry in solid state batteries but also introduce a versatile methodology for cryo-TEM sample preparation in SPE-based solid state batteries, thus providing an unprecedented opportunity to understand the electrochemistry of SPE-based solid state batteries from the atomic to the macroscopic scale.

Methods

Li‖SPE‖Li symmetrical cell preparation

Solid polymer electrolytes (SPEs) were fabricated by blending polyethylene oxide (PEO, MW = 6 × 10⁵ g mol⁻¹, Sigma-Aldrich) with lithium bis(trifluoromethanesulfonyl)imide (LiTFSI, 99.9%, Canrd), after thorough pre-drying. Anhydrous acetonitrile (Aladdin) was employed as the sacrificial solvent. PEO and LiTFSI were mixed in 10 ml of acetonitrile solvent at a predetermined molar ratio (EO/Li⁺ = 10 : 1), followed by vigorous agitation for 24 hours to yield a homogenized slurry. In the optimized scenarios, tetrabutylammonium fluoride (TBAF, 99.9%, Canrd) was introduced into the slurry at a mass ratio of 5%. Subsequently, the resultant mixture was cast onto a PTFE substrate and air-dried at ambient conditions for 24 hours, leading to the formation of the electrolyte membrane. For all cell configurations, lithium metal symmetric cells were assembled in the form of CR2032 coin cells. The entire cell assembly procedure was conducted within an oxygen- and moisture-controlled glovebox environment with concentrations maintained below 0.01 ppm.

Electrochemical testing

The constant current charge–discharge behaviors of all cell configurations were evaluated using an electrochemical test platform (Land CT3001A, LANHE Ltd). The electrochemical assessment was conducted at a current density of 0.1 mA cm⁻² coupled with an areal capacity of 0.1 mA h cm⁻², as demonstrated in Fig. 4d. For characterization, the cells were subjected to cycling conditions, leading to a soft-short circuit. Specifically, this entailed employing a current density of 1 mA cm⁻² with an areal capacity of 1 mA h cm⁻² (as depicted in Fig. 1b), as well as utilizing 0.5 mA cm⁻² current density paired with a 0.5 mA h cm⁻² areal capacity (as illustrated in Fig. S19, ESI†). All experimental evaluations were conducted under controlled temperature conditions of 60 °C.

Cryo-ultramicrotomy

After a short circuit event, the battery was disassembled in a glovebox and a piece of sample was extracted utilizing a surgical blade, which was subsequently transferred to a dedicated fixture. The extracted specimen was subsequently hermetically transferred to an ultramicrotome instrument (Leica EM UC7). A 20 minute pre-cooling procedure was carried out within the chamber before sample introduction. Liquid nitrogen gas was introduced to attain an internal chamber temperature of −60 °C, which is below the crystallization temperature threshold of PEO to fortify the mechanical robustness of SPE as an embedding medium for lithium dendrites. Moreover, the deliberate release of nitrogen gas facilitated the efficient purging of oxygen within the sample chamber, thus protecting the sample from potential contamination by moisture or oxygen during the slicing process. The entire slicing process involved a sequence of actions: initial trimming, subsequent sectioning, and ultimate sample transfer (as visually depicted in Fig. S1, ESI†). The initial trimming operation was accomplished utilizing a diamond trimming knife (DiATOME trim 45 DTB45) operating at a cutting speed of 100 mm s⁻¹ and a feed distance of 700 nm. This specialized process yielded a precisely defined trapezoidal area measuring 130 × 130 μm², meticulously centered around the critical Li-PEO interface region. Thereafter, the critical sectioning operation was conducted, employing a diamond knife (DiATOME cryo 45 DCO4530) engineered to ensure optimum sectioning precision. The operational parameters of this step included a carefully calibrated cutting speed of 0.1 mm s⁻¹ and a feed step of 90 nm. It is noteworthy that the entire sectioning methodology was conducted in a dry environment without any liquid being involved. Upon successful sectioning, the resulting ultra-thin sections were transferred to a TEM grid utilizing a fine brush. In the final phase, the TEM grid housing the sample was securely placed within a dedicated grid holder, which was immersed in liquid nitrogen. Finally, the prepared TEM grid was mounted onto a cryo-TEM specimen holder for observation.

Cryo-TEM and EELS measurements

The investigation was performed in a cryo TEM sample holder (Fischione Model 2550) attached to a Cs-corrected transmission electron microscope (TEM) (FEI, Titan G2, 300 kV). The samples were maintained at an ultra-low temperature of −173 °C during the TEM observations. The electron energy-loss spectroscopy (EELS) data were acquired using a Gatan Quantum ER spectrometer, configured with a collection angle of 12 mrad and a 5 mm entrance aperture. The established energy resolution, quantified via the full width at half maximum (FWHM) of the zero-loss peak (ZLP), was about 1.1 eV. The acquisition of EELS spectral images featured a pixel size of 42 × 42 nm, accompanied by a dwell time of 40 ms. These specifications corresponded to an image size of 111 × 60 pixels. All acquired low-loss spectra in the experiment were maintained at a dispersion of 0.1 eV per ch, covering a range of −20 to 180 eV. Meanwhile, core-loss spectra were captured with a dispersion of 0.5 eV per ch, spanning a range from 100 to 1124 eV.

MLLS data processing

Upon acquiring spectral images, the initial step involved employing multivariate statistical analysis (MSA) – principal component analysis (PCA) for denoising and retaining vital spectral attributes while mitigating extraneous noise (Fig. S23, ESI†). Sequentially, deconvolution was applied to each pixel to address multiple scattering effects, followed by background signal subtraction through power law model fitting within the 46 eV to 53 eV range. Then, a deconvolution approach was applied to each pixel to ensure the mitigation of the complexities introduced by multiple scattering phenomena. Subsequently, multivariate linear least squares (MLLS) regression was used to decompose the Li core loss spectra for each pixel. Notably, spectra with distinct differentiation and robust signals, such as Li and LiH, directly utilized experimentally acquired spectra as reference standards. In contrast, the spectra with lower differentiation encompass Li₂O, Li₂CO₃, and LiOH. Standard spectra were collected from the original phase powders under the same conditions, which served as reference spectra. Employing these references, MLLS fitting was executed within the Li core loss spectral range (50–75 eV), rendering each pixel's spectrum as a linear amalgamation of the reference spectra, characterized by minimal deviations from the actual spectra. This iterative process generated distribution maps for substances, reflecting their relative allotments in each pixel. Summing and normalizing these coefficients across pixels offered insights into substance content distribution across the spatial domain, as depicted through color-coded MLLS distribution maps.

SEM characterization and FIB-3D reconstruction

To comprehensively characterize the battery's cross-section, the disassembled symmetric cells were placed directly into fixtures within a glovebox without any prior treatment. These cells were hermetically transported to an ultrathin-slicing machine chamber for cryogenic polishing. Working at −60 °C, a trimming blade operated at a 100 mm s⁻¹ cutting rate and a 700 nm feed distance, establishing a flat profile, and subsequently refined by a diamond blade operating at a 50 mm s⁻¹ cutting rate and a 300 nm feed distance. The fixture and polished battery were directly introduced into a liquid nitrogen cup for cryopreservation and subsequently transferred into a glove box. After retrieval within the glovebox, the sample was sealed and then transferred to the FIB chamber using a vacuum transfer apparatus. SEM imaging was performed using a FEI Helios 5 DualBeam FIB/SEM instrument, with image acquisition parameters set at 5 kV and 0.17 nA. To achieve a three-dimensional anode surface reconstruction, a sample tilt of 52° aligned the surface normal to the ion beam direction. By employing Ga ions at 30 kV and 2.5 nA, 200 discrete images were collected and processed using Avizo software (Thermo Fisher Scientific).

Author contributions

Jianyu Huang and Yongfu Tang conceived and designed the project. Xuedong Zhang and Fangyuan Li fabricated the samples and measured the electrochemical data. Qiao Huang and Huan Hu designed the optimization scheme for SEI. Ziang Guo performed FIB milling and 3D reconstruction of the sample, Xin Li and Xuedong Zhang designed in situ optical experiments and data analysis. Qiunan Liu and Xuedong Zhang analyzed and processed EELS data. Xuedong Zhang performed microscopic characterization and data analysis. Xuedong Zhang, Yongfu Tang, and Jianyu Huang co-wrote the paper. Liqiang Zhang, Yongfu Tang, and Jianyu Huang supervised the experiments. All the authors discussed the results and commented on the manuscript.

Conflicts of interest

The authors declare no competing interests.

Acknowledgements

This work was financially supported by the National Natural Science Foundation of China (No. 22279112, 51772262, 5197010923, 51971245, 52002346, 52022088, U20A20336, 21935009), Natural Science Foundation of Hebei Province (No. B2023203031, B2022203018, B2018203297, B2020203037, F2021203097), Fok Ying-Tong Education Foundation of China (No. 171064), Hebei One Hundred Talent Program (No. 4570028), High-Level Talents Research Program of the Yanshan University (No. 00500021502, 005000201), Hebei Key Laboratory of Applied Chemistry after Operation Performance (No. 22567616H), Science and Technology Innovation Program of Hunan Province (No. 2021RC3109).

References

1. D. Lin, Y. Liu and Y. Cui, Nat. Nanotechnol., 2017, 12, 194–206.
2. B. Liu, J.-G. Zhang and W. Xu, Joule, 2018, 2, 833–845.
3. J. Liu, Z. Bao, Y. Cui, E. J. Dufek, J. B. Goodenough, P. Khalifah, Q. Li, B. Y. Liaw, P. Liu, A. Manthiram, Y. S. Meng, V. R. Subramanian, M. F. Toney, V. V. Viswanathan, M. S. Whittingham, J. Xiao, W. Xu, J. Yang, X.-Q. Yang and J.-G. Zhang, Nat. Energy, 2019, 4, 180–186.
4. W. Xu, J. Wang, F. Ding, X. Chen, E. Nasybulin, Y. Zhang and J.-G. Zhang, Energy Environ. Sci., 2014, 7, 513–537.
5. Z. A. Ghazi, Z. Sun, C. Sun, F. Qi, B. An, F. Li and H.-M. Cheng, Small, 2019, 15, 1900687.
6. C. Fang, X. Wang and Y. S. Meng, Trends Chem., 2019, 1, 152–158.
7. K.-H. Chen, K. N. Wood, E. Kazyak, W. S. LePage, A. L. Davis, A. J. Sanchez and N. P. Dasgupta, J. Mater. Chem. A, 2017, 5, 11671–11681.
8. E. Peled and S. Menkin, J. Electrochem. Soc., 2017, 164, A1703.
9. E. Peled, J. Electrochem. Soc., 1979, 126, 2047.
10. D. Aurbach, J. Power Sources, 2000, 89, 206–218.
11. M. D. Tikekar, S. Choudhury, Z. Tu and L. A. Archer, Nat. Energy, 2016, 1, 16114.
12. X.-B. Cheng, R. Zhang, C.-Z. Zhao, F. Wei, J.-G. Zhang and Q. Zhang, Adv. Sci., 2016, 3, 1500213.
13. J. Steiger, D. Kramer and R. Mönig, J. Power Sources, 2014, 261, 112–119.
14. N. J. J. de Klerk and M. Wagemaker, ACS Appl. Energy Mater., 2018, 1, 5609–5618.
15. B. Son, M.-H. Ryou, J. Choi, T. Lee, H. K. Yu, J. H. Kim and Y. M. Lee, ACS Appl. Mater. Interfaces, 2014, 6, 526–531.
16. M. J. Zachman, Z. Tu, S. Choudhury, L. A. Archer and L. F. Kourkoutis, Nature, 2018, 560, 345–349.
17. X.-C. Ren, X.-Q. Zhang, R. Xu, J.-Q. Huang and Q. Zhang, Adv. Mater., 2020, 32, 1908293.
18. O. Sheng, J. Zheng, Z. Ju, C. Jin, Y. Wang, M. Chen, J. Nai, T. Liu, W. Zhang, Y. Liu and X. Tao, Adv. Mater., 2020, 32, 2000223.
19. Y. Liu, Z. Ju, B. Zhang, Y. Wang, J. Nai, T. Liu and X. Tao, Acc. Chem. Res., 2021, 54, 2088–2099.
20. Y. Li, Y. Li, A. Pei, K. Yan, Y. Sun, C.-L. Wu, L.-M. Joubert, R. Chin, A. L. Koh, Y. Yu, J. Perrino, B. Butz, S. Chu and Y. Cui, Science, 2017, 358, 506–510.
21. D. Cheng, T. A. Wynn, X. Wang, S. Wang, M. Zhang, R. Shimizu, S. Bai, H. Nguyen, C. Fang, M.-C. Kim, W. Li, B. Lu, S. J. Kim and Y. S. Meng, Joule, 2020, 4, 2484–2500.
22. J. J. Lodico, M. Mecklenburg, H. L. Chan, Y. Chen, X. Y. Ling and B. C. Regan, Sci. Adv., 2023, 9(28), eadg5135.
23. L. Yu, M. Li, J. Wen, K. Amine and J. Lu, Mater. Chem. Front., 2021, 5, 5186–5193.
24. S. Menkin, J. B. Fritzke, R. Larner, C. de Leeuw, Y. Choi, A. B. Gunnarsdóttir and C. P. Grey, Faraday Discuss., 2024 10.1039/D3FD00101F.
25. Z. Zhang, Y. Li, R. Xu, W. Zhou, Y. Li, S. T. Oyakhire, Y. Wu, J. Xu, H. Wang, Z. Yu, D. T. Boyle, W. Huang, Y. Ye, H. Chen, J. Wan, Z. Bao, W. Chiu and Y. Cui, Science, 2022, 375, 66–70.
26. Y. Su, X. Zhang, C. Du, Y. Luo, J. Chen, J. Yan, D. Zhu, L. Geng, S. Liu, J. Zhao, Y. Li, Z. Rong, Q. Huang, L. Zhang, Y. Tang and J. Huang, Small, 2022, 18, 2202069.
27. Y. Li, Z. Gao, F. Hu, X. Lin, Y. Wei, J. Peng, J. Yang, Z. Li, Y. Huang and H. Ding, Small Methods, 2020, 4, 2000111.
28. S. Tan, J.-M. Kim, A. Corrao, S. Ghose, H. Zhong, N. Rui, X. Wang, S. Senanayake, B. J. Polzin, P. Khalifah, J. Xiao, J. Liu, K. Xu, X.-Q. Yang, X. Cao and E. Hu, Nat. Nanotechnol., 2023, 18, 243–249.
29. Y. Xu, H. Jia, P. Gao, D. E. Galvez-Aranda, S. P. Beltran, X. Cao, P. M. L. Le, J. Liu, M. H. Engelhard, S. Li, G. Ren, J. M. Seminario, P. B. Balbuena, J.-G. Zhang, W. Xu and C. Wang, Nat. Energy, 2023, 8, 1345–1354.
30. A. Wang, S. Kadam, H. Li, S. Shi and Y. Qi, npj Comput. Mater., 2018, 4, 15.
31. R. A. Vilá, D. T. Boyle, A. Dai, W. Zhang, P. Sayavong, Y. Ye, Y. Yang, J. A. Dionne and Y. Cui, Sci. Adv., 2023, 9(12), eadf3609.
32. P. de Sainte Claire, Macromolecules, 2009, 42, 3469–3482.
33. Y. Yusim, E. Trevisanello, R. Ruess, F. H. Richter, A. Mayer, D. Bresser, S. Passerini, J. Janek and A. Henss, Angew. Chem., Int. Ed., 2023, 62, e202218316.
34. M. Yi, J. Li, M. Wang, X. Fan, B. Hong, Z. Zhang, Z. Zhang, H. Jiang, A. Wang and Y. Lai, Energy Storage Mater., 2023, 54, 579–588.
35. Q. Zhou, J. Ma, S. Dong, X. Li and G. Cui, Adv. Mater., 2019, 31, 1902029.
36. J. Li, H. Hu, W. Fang, J. Ding, D. Yuan, S. Luo, H. Zhang and X. Ji, Adv. Funct. Mater., 2023, n/a, 2303718.
37. Y. Zhang, W. Feng, Y. Zhen, P. Zhao, X. Wang and L. Li, Ionics, 2022, 28, 2751–2758.
38. Z. Shadike, H. Lee, O. Borodin, X. Cao, X. Fan, X. Wang, R. Lin, S.-M. Bak, S. Ghose, K. Xu, C. Wang, J. Liu, J. Xiao, X.-Q. Yang and E. Hu, Nat. Nanotechnol., 2021, 16, 549–554.
39. J. Tan, J. Matz, P. Dong, J. Shen and M. Ye, Adv. Energy Mater., 2021, 11, 2100046.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3ee04189a

---