A porous tellurium interlayer for high-power and long-cycling garnet-based quasi-solid-state lithium-metal batteries

Abstract

Li-metal batteries with solid oxide electrolytes have garnered increasing attention as promising technologies that can overcome the safety and energy density limits of lithium-ion batteries (LIBs). However, the less than satisfactory long-term stability of Li-metal batteries—a consequence of Li dendrite formation caused by unstable Li plating/stripping at the interface between the Li metal and solid electrolyte—is hampering their commercialisation. Herein, we propose an negative electrode multilayer consisting of porous Te and carbon-based layers that can suppress Li dendrite formation and significantly lower the degree of capacity decay during long-term cycling. A quasi-all-solid-state Li-metal battery fabricated using Li_6.4La₃Zr_1.7Ta_0.3O₁₂ (LLZTO), a Te/Ag-C anodenegative electrode interlayer, and a LiNi_0.8Co_0.1Al_0.1O₂ (NCA811) positive electrode impregnated with an ion-conducting liquid demonstrated high capacity retention (80.1% over 4000 cycles) and Coulombic efficiency (99.7%) when operated at a high current density of 2.2 mA/cm² at 25 °C. Furthermore, we successfully demonstrate 100-mAh-level single cells in a pouch cell configuration using a large-area (36 cm²) LLZTO solid electrolyte and a 3.2-mAh/cm² LiCoO₂ (LCO) positive electrode capable of operating for >400 cycles at a 0.5 C-rate (85 mA/g).

Unsatisfactory long-term stability of solid-state Li-metal batteries is hampering their commercialisation despite their potential high energy density. Here, authors employed a rationally designed porous tellurium interlayer and demonstrated high capacity retention for 4000 cycles.

Introduction

Solid-state Li-metal batteries based on solid electrolytes are considered some of the most promising next-generation batteries because of their potential to enhance safety and overcome the energy density limitations associated with commercial lithium-ion batteries (LIBs)^1–4. Among them, quasi-all-solid-state Li-metal batteries, which incorporate ionic liquids in the positive electrode and solid oxide electrolytes in the negative electrode, have drawn significant attention because of their potential to achieve higher energy densities via the utilisation of high-loading positive electrodes and thin Li-metal negative electrodes^5,6. Furthermore, the ionic liquids used as catholytes in these batteries are generally non-flammable; thus, they offer improved safety compared to the flammable carbonate-based liquid electrolytes in conventional LIBs.

However, several obstacles hinder the commercialisation of quasi-all-solid-state Li-metal batteries. For example, a major technical challenge is the formation and maintenance of a uniform interface between the solid oxide electrolyte and the negative electrode because a damaged interface can decrease the charge/discharge efficiency to potentially cause the cell to short-circuit^7–10. In addition, the presence of physical and chemical defects (e.g., voids, impurities, and secondary phases) at the interface between the Li-metal negative electrode and the solid electrolyte is known to reduce the area containing electrochemically active sites available for charge-transfer and cause an uneven current distribution to ultimately accelerate the formation of lithium dendrites^7,11,12. In particular, at low temperatures, where the diffusion of Li atoms in Li metal is slow, voids can easily form at the interfaces during the Li stripping process. Thus, ensuring conformal contact between the Li-metal negative electrode and the solid electrolyte by suppressing void formation is crucial for inhibiting short circuits caused by lithium dendrites, and for the development of quasi-all-solid-state Li-metal batteries.

To date, various types of inorganic binary compounds, carbons, and metals (e.g., Ag, Au, Si, and Zn) have been used to remedy the problematic negative electrode–solid electrolyte interface^13–21, with carbon-based interlayers being proven particularly effective in maintaining the required conformal contact^5,15–17. For instance, Lee et al. and Kim et al. recently introduced an Ag-C interlayer to effectively regulate Li stripping/plating and suppress short-circuit failure for more than 800 cycles^5,15. Nevertheless, an additional binding material is required for oxide solid electrolytes to establish and maintain conformal contact between carbon and the solid electrolyte owing to the weak electronic interactions between them. Kim et al. additionally applied a thin Ag layer to the surface of LLZTO (Li_6.4La₃Zr_1.7Ta_0.3O₁₂) as a bonding layer, and demonstrated that the quasi-all-solid-state Li-metal battery with a 3 mAh/cm² positive electrode could be reversibly cycled for 800 cycles at 1.6 mA/cm² and 25 °C without the application of external pressure⁵. Although they successfully demonstrated a suitable design strategy for the negative electrode interlayers, the coated Ag gradually disappeared from the LLZTO surface during repeated charge/discharge cycling because of its strong tendency to alloy with Li. This high mobility of Ag toward Li can deteriorate the long-term interface uniformity because loss of the Ag layer at the negative electrode interface could result in void formation and/or delamination of the negative electrode layers from the solid electrolyte. Therefore, the development of materials with high morphological stabilities and the capability to host fast Li diffusion is imperative to fully utilise the potential of the Ag-C interlayer and ensure long-term cycling under high current densities.

In this study, density functional theory (DFT) calculations were employed to explore elements that were not considered likely to diffuse into Li metal. Subsequently, Te was selected as a candidate bonding material between the carbon-based layer and the solid electrolyte, and the Li mobility at the Te surface was compared to that in the bulk. Furthermore, a 40 nm-thick, porous Te layer was prepared on the surface of LLZTO by selectively etching Ge from the sputtered GeTe film, whereupon an Ag-C layer was deposited onto the porous Te layer. Utilising this negative electrode multilayer, a quasi-all-solid-state Li-metal battery cell containing a 20 μm-thick Li-metal negative electrode and a high-loading ionic liquid-based NCA811 positive electrode was operated for more than 4000 cycles at 25 °C at a current density of 2.2 mA/cm². The negative electrode multilayer was additionally applied to a large-area pouch cell with a 36 cm² LLZTO solid electrolyte and a 30 cm² LCO positive electrode, and the resulting single-layer pouch cell was operated over 400 cycles at 25 °C at the 0.5 C-rate (85 mA/g).

Results and discussion

Design of an optimal material for the interface bonding layer

A molecular dynamics (MD) technique was employed to investigate the alloying behaviours of different elements with Li (Supplementary Data 1). Beginning with a bilayer structure consisting of Li and the candidate element, MD simulations were performed to determine the extent to which the alloying reaction between Li and the candidate element would proceed at the interface. Materials that react rapidly or undergo complete mixing would be likely to alloy with Li, as in the case of Ag^5,15; therefore, the material would not be suitable for use as a bonding layer. However, materials that show limited progress in the alloying reaction could be expected to remain immobile during charge/discharge when employed as a bonding layer; thus, they would be potential candidates. In this design strategy, it is important to accurately observe long-term alloying reactions to obtain reliable results using the MD simulations. For this purpose, an ab initio molecular dynamics (AIMD) simulation was initially employed. Because of the long simulation time (> 100 ps), it was possible to capture noticeable alloying reaction processes; however, it was recognised that a longer simulation time was required to more rigorously understand the alloying behaviour for low-temperature simulations. Considering the low melting temperature of Li (180 °C), a high-temperature simulation, as typically practiced for ceramic ionic conductors^22,23, was not an option. Consequently, the machine-learned interatomic potential (MLIP) was employed to run the MD calculations (see the Methods section for additional details). Among the various elements to be explored, Sn and Te were investigated as a representative metal and metalloid, respectively. However, the alloying behaviour of Li and Ag was initially studied for comparison. As shown in Fig. 1a (on the left), a Li/Ag bilayer model was built in which the atomic ratio of Li to Ag was ~ 10:1. As the MLIP-MD simulation progressed, the alloying reaction between Li and Ag began, and some of the Ag atoms advanced into the Li layer (Fig. 1a (middle)). Further reaction resulted in the almost complete mixing of Li and Ag, indicating that Ag can easily form an alloy with Li and therefore diffuse into the Li layer, corroborating previous experimental observations of Ag loss⁵. The strong tendency of Li-Ag alloying was also supported by the phase diagram, wherein a Li-rich alloy phases exists²⁴. However, alloying between Li and Te saturates after 2 ns (Fig. 1b), and the bilayer structure of separated Li and Li₂Te phases remains, indicating that Te could be immobile toward Li. Sn behaves similarly, with Li-Sn alloying saturating at Li_4.4Sn, without further Li/Sn interdiffusion. These results are supported by Li-Te and Li-Sn phase diagrams, where the biphasic existence of the Li-rich compound and Li was expected for a Li-rich composition. Based on these findings, both Te and Sn were considered applicable as binding layers between LLZTO and the carbon-based interlayers. Among the two, Te was ultimately selected because of its lesser degree of Li uptake (2 Li per Te) compared to Sn (4.4 Li per Sn) during alloying, which could help minimise morphological and volumetric changes during charge/discharge cycling. The sluggish Li diffusion in Te (diffusion coefficient of 10⁻¹⁴–10⁻¹⁵ cm²/s)²⁵ could also be beneficial for alleviating the volume change, as it could limit the rate of Li-Te alloying reaction and possibly restrict the alloying region to the surface of Te, similar to what was recently reported for the lithiation of Si at the solid–solid interface²⁶. DFT calculations also confirmed that Te enhanced the adhesion between LLZTO and the carbon-based interlayer. More specifically, the work of adhesion between LLZTO and carbon was determined to be − 0.33 J/m², whereas that between LLZTO and Te was − 1.34 J/m², and that between and Te and C was − 0.47 J/m² (Supplementary Fig. 1).

Fig. 1. Evaluation of the metal alloying behaviours with Li.

a Li-Ag alloying and (b) Li-Te alloying behaviours during a 10-ns MLIP-MD simulation.

Bonding layer materials should also possess high Li diffusivity to facilitate Li transport and homogenise the Li flux and current distribution. Unfortunately, the diffusivity of Li in Te is reportedly sluggish, with a measured diffusion coefficient in the range of 10⁻¹⁴–10⁻¹⁵cm²/s²⁷. Considering that the diffusion coefficient of Li in Li-Ag alloys is ~ 10⁻⁸cm²/s²⁸, employing Te as a binding layer would be expected to degrade the rate capability of the cell. Nudged elastic band (NEB) calculations were conducted to identify possible alternative paths for fast Li diffusion in Te, to evaluate the activation barrier for Li diffusion at the surface of Li₂Te. The diffusion of Li was fast at the surface of the Li-Te alloy (Supplementary Fig. 2), with only a small activation barrier of 70 meV. Based on these findings, porous Te with a large surface area seemed an ideal interlayer for positioning between the solid electrolyte and the Ag-C layer.

Preparation and characterisation of the porous Te layer

Using a selective etching strategy, we attempted to prepare a porous Te layer on the LLZTO surface. A GeTe film was sputtered onto LLZTO, whereupon Ge was selectively etched from the layer using a simple acid treatment. Consequently, a 40 nm-thick, porous Te layer was successfully prepared (Fig. 2a). The chemical and morphological changes in the porous Te layer during charge/discharge cycling were investigated by performing in situ measurements using electron and ion-beam devices that are commonly employed as neutralisers and excitation sources during X-ray photoelectron spectroscopy (XPS) and Auger electron spectroscopy (AES), respectively. Initially, a Te|LLZTO|Li/Cu structure was fabricated, as illustrated schematically in Fig. 2b. Upon subsequent irradiation with an electron beam (e-beam), Li ions were extracted through LLZTO toward the Te layer, resembling a charge process. Conversely, under ion-beam irradiation, the Li ions moved down through the Te layer, representing the discharge process (Fig. 2b, c). Using this methodology, the in-situ core-level spectrum of each element was measured and separated into various valence states to scrutinise the chemical reactions taking place in the porous Te layer during lithiation and delithiation. Figure 2c shows the Te 3d_5/2 spectral changes and their deconvoluted chemical states under e-beam and ion-beam irradiation conditions. Initially, the LLZTO surface was covered with Te metal (shaded in cyan) and Te oxide (shaded in orange). Upon irradiation with an e-beam, the Li ions began to migrate to the surface and Li-Te alloying proceeded, as evidenced by the increase in the green-shaded area in Fig. 2c. Meanwhile, under ion-beam irradiation, the Li-Te content began to decrease, indicating decomposition of the Li-Te alloy, as further evidenced by the intensification of the Te metal signal. The variation in the relative percentage quantity of each chemical state within in the Te 3d_5/2 spectra was determined using curve fitting as a function of the irradiation time. We note that the O 1 s and Li 1 s spectra could not be utilised to analyse the chemical state, as multiple contributions from various sources were included in the spectra (Supplementary Fig. S3). The Li-Te alloy content increased to 90.8% after 8 h of e-beam irradiation and then decreased to 8.8% as a result of ion-beam irradiation (Fig. 2d), indicating that the Li-Te alloying reaction was almost reversible, although a small amount of Li-Te alloy remained at the surface. Furthermore, identical measurements were performed for the dense Te layer to determine whether the creation of a porous layer facilitates Li transport. The Li/Te atomic ratio—derived from the calculated atomic percentages (at. %) of Li and Te (Fig. 2e) to analyse the variation in the layer composition during irradiation—indicated that this ratio increases more rapidly in the porous than in the dense Te layer. After 8 h of e-beam irradiation, the Li/Te ratio of the porous Te layer increased to 10.3, whereas that of the dense Te layer was 3.4. An identical trend was observed during ion-beam irradiation, where the Li/Te ratio decreased more rapidly in the porous Te layer. These findings therefore indicate that the design strategy described herein for the preparation of a porous Te layer is effective for accelerating Li transport, as corroborated by the theoretical investigation (Supplementary Fig. 2).

Fig. 2. Preparation and characterisation of the porous Te layer.

a Surface SEM images of the GeTe layer before and after acid treatment. The cross-sectional images display a ~ 40 nm-thick coating layer. b Schematic of the measurement process. The electron and ion beams drive Li-ion migration from/to the Te-coated LLZTO. The SEM images show the increased Li content on the Te-coated LLZTO surface over time under e-beam irradiation. The AES elemental mapping images show the distributions of Li and Te on the surface. c Changes in the core-level spectra of the porous Te layer during e-beam and ion-beam irradiation. d Percentages of each chemical state in the Te 3d_5/2 spectrum as a function of the irradiation time. e Changes in the Li/Te atomic ratio during e-beam and ion-beam irradiation.

To further elucidate the role of the porous Te-binding layer, Li|carbon/Te|LLZTO|Te/carbon|Li symmetric cells were fabricated (Supplementary Fig. 4a). As theoretically predicted, the carbon black sheet was uniformly attached to the thin Te-sputtered LLZTO after cold isostatic pressing, indicating that the Te layer could effectively assist in the adhesion between LLZTO and the carbon-based interlayer. The electrochemical impedance spectroscopy (EIS) results (Supplementary Fig. 4b and Supplementary Table 1) revealed that the cell bearing a porous Te layer exhibited a low initial interfacial resistance of 4.98 Ω cm², whereas the interfacial resistance increased to 8.65 Ω cm² for dense Te. The ability of the porous layer to lower the interfacial resistance is attributable to an increase in the electrochemically active area, as confirmed in Fig. 2. Thus, a metal interlayer with a large surface area is a key factor in determining the overall kinetics of the electrochemical reactions taking place in the negative electrode interlayer.

The galvanostatic cycling tests of the Li-metal symmetric cells with increasing current densities (Supplementary Fig. 4c, d) clearly showed that the cell containing the porous Te-coated LLZTO did not short-circuit even at a current density of 3.0 mA/cm², whereas the overpotential of the cell containing the dense Te-coated LLZTO increased significantly to eventually cause the cell to short-circuit at a current density of 1.8 mA/cm². These findings imply that the improved lithium transport kinetics at the porous Te and LLZTO interface contribute to alleviating Li dendrite formation by improving the current uniformity at the interface. Rapid movement of the Li atoms plated on the electrode toward the negative electrode current collector could mitigate the local current concentration to suppress Li dendrite formation and growth.

Fabrication and electrochemical performance of the quasi-all-solid-state Li-metal battery

Based on the above understanding of the lithiation/delithiation behaviour of LLZTO coated with porous Te, a quasi-all-solid-state Li-metal cell was fabricated to assess its electrochemical performance. Pouch cells were assembled using a 140-μm-thick tape-cast solid electrolyte, a 4.4 mAh/cm² NCA811 positive electrode, and a 20 μm-thick Li-metal negative electrode with a porous Te/Ag-C interlayer (Fig. 3a). The impedance spectrum of the cell (Fig. 3b) exhibits the typical impedance behaviour observed for LLZTO-based quasi-all-solid-state cells, comprising two depressed arcs in the high- and middle-frequency regions, in addition to Warburg impedance and low-frequency capacitive lines. The overall interfacial resistance, which corresponds to the resistances of the positive electrode- and negative electrode-LLZTO interfaces, was calculated to be ~ 12 Ω cm². This value is lower than the initial resistance (32.5 Ω cm²) reported previously for an Ag/Ag-C interlayer⁵, indicating that the interface charge-transfer reaction kinetics for Li plating/stripping is faster in the presence of the porous Te/Ag-C interlayer. The long-term stability of the quasi-all-solid-state Li-metal battery was evaluated by conducting cycling tests at current densities of 2.2 mA/cm² (4.4 V cutoff) and 3 mA/cm² (4.3 V cutoff). The charge/discharge curves (Fig. 3c) demonstrate that the prepared cell can operate stably without significant voltage drops or short circuits for up to 4000 cycles. These results are consistent with the high critical current density (> 3 mA/cm²) observed in Supplementary Fig. 4d for a symmetrical Li cell containing a porous Te/carbon interlayer. Notably, the long-term cycling results (Fig. 3d) show that the cell containing porous Te/Ag-C negative electrode interlayers exhibited capacity retention of 80.1% after 4000 cycles at 25 °C. This superior cycling performance can be attributed to the facile Li transport through the negative electrode layer and the enhanced interface stability originating from the immobility of Te after repeated cycles. In fact, the scanning electron microscopy (SEM)/energy-dispersive X-ray spectroscopy (EDS) results (Fig. 3e, f) reveal the presence of the Te layer at the solid electrolyte/Ag-C interface even after 4000 cycles, where this layer stabilised the interface to prevent the formation of voids or delamination of the negative electrode layer from LLZTO. Compared with state-of-the-art quasi-all-solid-state cells with a garnet-type solid electrolyte reported thus far^{5,6,16,29–34}, the performance of our cell in terms of areal capacity, current density, and long-term capacity retention is notable. In our previous study, when we adopted Li/Ag-C/Ag configuration as a negative electrode⁵, we could cycle the cell for 800 cycles at the current density of 1.6 mA/cm². By contrast, when we replace interface-binding material from Ag to Te, such that the negative electrode configuration is Li/Ag-C/Te, the cycling stability is highly improved (4000 cycles at the current density of 2.2 mA/cm²), clearly indicating that Te is considerably better than Ag as a binding layer. This clearly verifies the described design strategy for the exploration of interface-binding materials, namely to identify an element with a low mobility towards Li (Fig. 1). We would like to additionally note that binding layers at the interface between solid electrolyte and the interlayer are only required for oxide solid electrolytes, as explained in the introduction of this manuscript. In Fig. 5c of our previous paper⁵, the cell using LLZTO and Ag-C interlayer without the binding layer shorted after 2 cycles, whereas the sulfide cell showed high stability without the binding layer¹⁵.

Fig. 3. Fabrication and electrochemical performance of the quasi-all-solid-state Li-metal battery.

a Schematic of the assembled Li|Ag-C/Te|LLZTO | NCA811-EMIFSI (2 M LiFSI) cell (1 C-rate = 205 mA/g). b Nyquist plots of the AC-impedance spectra obtained from the cell. c Galvanostatic voltage profiles recorded for the prepared cell. d Cycling performance of the quasi-all-solid-state Li-metal battery. e Cross-sectional SEM image of the cell in a fully discharged state (capacity: 3.163 mAh/cm²) after 4000 cycles. f SEM/EDS image of the cell at the LLZTO/Te interface in a fully discharged state (capacity: 3.163 mAh/cm²) after 4000 cycles. g Cross-sectional SEM image of the cell at the Li/Cu interface in a fully discharged state (capacity: 3.163 mAh/cm²) after 4000 cycles. h SEM image and (i) EDS mapping images of the NCA positive electrode in a fully discharged state (capacity: 3.163 mAh/cm²) after 4000 cycles. The cycling was conducted at 25 °C with a current density of 2.2 mA/cm² (4.4V-cutoff).

Interestingly, the partial fragmentation and agglomeration of the Cu particles near the interface of the Cu current collector with Li-metal (Fig. 3g) is presumably attributable to the strain/stress induced by volume changes during repetitive Li plating/stripping. These results suggest that Li plating occurs primarily at the Ag-C/Cu interface rather than at the Ag-C/solid electrolyte interface, a key requirement for the long-term cycling stability of solid-state Li-metal batteries⁵. The degradation of Cu could possibly affect further Li plating behaviour; for instance, the uniformity of Li plating and stripping could be undermined due to the irregular morphology of Cu, a possible reason for the slow capacity fade after 2000 cycles (Fig. 3d). However, the partial fragmentation of Cu could be considered unlikely to change the preferred site of Li plating/stripping (Ag-C/Cu interface) as long as the Ag-C/solid electrolyte interface is tightly bonded. As demonstrated in previous studies^5,35, Li tends to be plated at the Ag-C/Cu interface when the Ag-C/solid electrolyte bonding is stronger than that at the Ag-C/Cu interface. The degradation of Cu could only weaken the bonding between Ag-C/Cu; therefore, Li would still prefer plating at the Ag-C/Cu interface.

The lifespan of the prepared cell provides meaningful insights not only into the durability of the negative electrode interface, but also into the long-term stability of the NCA811 positive electrode material. The SEM/EDS cross-sectional images of the positive electrode (Fig. 3h, i) reveal the presence of microcracks and the partial infiltration of ionic liquids among the NCA811 particles after 4000 cycles. This phenomenon is commonly associated with the microcracking degradation behaviour observed in NCA and NCM (LiNi_1-x-yCo_xMn_yO₂) electrodes, originating from the transformation of the surface phases into spinel and rock salt phases. This micro-cracking degradation of the positive electrode can affect not only the capacity retention but also the electrochemical kinetics of the cell. Indeed, observation of the charging capacity in the CV region across cycles reveals an increase after 2000 cycles (Supplementary Fig. 5). This signifies an increase in overpotential under constant current conditions and is presumably due to degradation caused by micro-crack formation in the positive electrode and/or degradation at the positive electrode-ionic liquid interface. Interestingly, the capacity retention of 80.1% after 4000 cycles for the prepared quasi-all-solid-state Li-metal battery compares the reported cycling performances of the high-Ni NCA or NCM electrodes in LIBs containing carbonate-based liquid electrolytes^36–39. This strongly suggests that the degradation behaviours at the positive electrode/liquid interfaces, including transition metal dissolution, surface reconstruction of the positive electrode, and electrolyte decomposition, occur much less frequently for ionic liquids (2 M LiFSI in EMIFSI) with high oxidation potentials than for carbonate-based liquid electrolytes.

Subsequently, a quasi-all-solid-state cell containing a porous Te/C interlayer was fabricated and evaluated to determine the effect of Ag on the performance of the carbon-based interlayer (Supplementary Fig. 6a). The interfacial resistance was calculated to be 23 Ω cm² from the EIS measurements presented in Supplementary Fig. 6b, which is a slightly higher value than that of the cell containing the Te/Ag-C interlayer. The capacity retention of ~ 83% after 500 cycles (Supplementary Fig. 6c, d) is relatively inferior to that of the porous Te/Ag-C cell; however, stable operation was achieved up to a maximum of 500 cycles without any significant capacity drop. This indicates that the carbon interlayer can effectively bind to the porous Te metal layer, and that Li dendrite formation is suppressed up to a current density of 2.2 mA/cm². Nevertheless, Ag in the interlayer certainly plays an important role, as the capacity retention of the Te/C cell (Supplementary Fig. 6d) is inferior to that of the Te/Ag-C cell (Fig. 3d). The identical tendency was also reported for the sulphide electrolyte system, where the incorporation of Ag in the C interlayer resulted in improved cycling stability¹⁷. Considering that the critical current density is not highly dependent on the presence of Ag in the negative electrode interlayer, Ag is expected to have a greater impact on the kinetics of Li plating/stripping than on the suppression of Li dendrite formation. The low Li nucleation barrier and high Li diffusivity in Li-Ag alloys^18,28 presumably indicate that Ag in the C interlayer facilitates Li reduction and enhances the Li transport inside the interlayer. However, carbon was considered to play a pivotal role in preventing the formation of Li dendrites¹⁶.

Assuming the cell configuration shown in Fig. 3a as a representative unit cell for battery pack integration, the calculated volumetric energy density is 519 Wh L⁻¹ (Supplementary Table 2). To achieve an energy density exceeding 700 Wh L⁻¹ (comparable to that of state-of-the-art commercial LIBs), both the thickness of the solid electrolyte and the lithium metal negative electrode must be reduced from their current levels. Utilising Corning’s fast-firing roll-to-roll ribbon ceramic processing (https://www.corning.com/media/worldwide/Innovation/documents/ribbonceramics/1_CGMSeptember2020.pdf), the thickness of the LLZTO solid electrolyte can be reduced to below 60 μm, while Applied Materials’ sputtering technology allows the lithium metal thickness to be decreased to below 5 μm (https://www.osti.gov/servlets/purl/2341379). In fact, Corning’s ribbon ceramic process enabled us to successfully fabricate a 30 μm-thick solid electrolyte (Supplementary Fig. 7). However, such thin solid electrolytes tend to be brittle, and therefore, mechanical reinforcement is necessary for practical application. Addressing this issue remains a critical focus of our ongoing and future research efforts.

To further examine the applicability of the cell for use in commercial devices, a 100 mAh-capacity cell was fabricated and evaluated using a large-area tape solid electrolyte (36 cm²), a Te/C interlayer, and a 3.2 mAh/cm²-LCO positive electrode (Fig. 4a). The impedance spectrum (Fig. 4b) indicates that the large-area cell exhibits a low initial interfacial resistance of < 10 Ω cm². The lower interfacial resistance of the large-area cell compared to that of the small-area cell shown in Fig. 3b is likely due to the increased surface area of the large-area solid electrolyte resulting from the inhomogeneous distribution of Li on the surface and subsequent increased porosity during acid treatment. Because large-area cells are more susceptible to micro-short circuits due to the increased formation of surface defects and localised current densities as compared to small-area cells, the long-term cycling performance was evaluated at 45 °C, at which the enhanced lithium mobility can mitigate failure due to micro-short circuits. The charge/discharge curves and capacity retention after cycling this cell at the 0.5 C rate for up to 400 cycles (Fig. 4c, d) demonstrate successful operation without short circuits and that the capacity was maintained at ~ 95 mAh. Notably, this corresponds to the design capacity of the positive electrode. The relatively lower capacity retention compared to that of the small-sized cells shown in Fig. 3 and Supplementary Fig. 4 may be attributed to the surface inhomogeneity of the large-area LLZTO electrolyte. As the area of the solid electrolyte increases, Li evaporation may occur unevenly during the high-temperature sintering process at 1100 °C, resulting in non-uniform Li distribution and/or localised secondary phase formation on the LLZTO surface, which could lead to poor cell cyclability. Nevertheless, the Te/C interlayer can effectively prevent the formation of Li dendrites, even in a practically meaningful large-area LLZTO solid electrolyte. Importantly, the demonstration of 100 mAh-level single cell performance confirms the potential of these quasi-all-solid-state Li-metal batteries for application in high-capacity batteries for electronic devices and/or electric vehicles.

Fig. 4. Electrochemical performance of the Li | C/Te | LLZTO | LCO-EMIFSI (2 M LiFSI) single cell.

a The LLZTO tape and the LCO positive electrode used to fabricate the 100 mAh Li|C/Te|LLZTO | LCO single cell (1 C-rate = 170 mA/g). b Nyquist plots of the AC-impedance spectra obtained from the prepared cell. c Galvanostatic voltage profiles recorded for the cell. d Cycling performance of the quasi-all-solid-state Li-metal battery.

Finally, to demonstrate the safety of the prepared quasi-all-solid-state Li-metal batteries, flammability tests were conducted by directly exposing the fully charged positive electrode to a flame of a butane gas torch. The LCO electrode and the LCO powder scraped off from the charged cell did not ignite, even when directly exposed to the flame for several tens of seconds (Supplementary Fig. 8). These results clearly highlight the flame-retardant property of the ionic liquid. Moreover, nail penetration tests on single-pouch cells demonstrated stable voltage maintenance without producing smoke, undergoing combustion, or experiencing voltage drops, even after penetration (Supplementary Video 1). These results suggest the potential advantages of quasi-all-solid-state Li-metal batteries over conventional LIBs, not only in terms of their high energy densities, but also in terms of their safety profiles due to the flame-retardant properties of ionic liquids.

In summary, an intermediate layer design strategy was developed to address Li dendrite formation at the interface between an oxide solid electrolyte and a Li-metal electrode with the aim of securing long-term stability for Li-metal batteries. Assisted by careful computational design, the application of porous Te was proposed as a bonding layer between LLZTO (Li_6.4La₃Zr_1.7Ta_0.3O₁₂) and a carbon-based interlayer. Rigorous analysis revealed that the porous Te coating not only helped maintain strong adhesion between the LLZTO solid electrolyte and the carbon-based interlayer, but that it also facilitated Li transport. This design strategy was further validated by the high capacity retention (80.1% after 4000 cycles) and Coulombic efficiency (99.7%) of a quasi-all-solid-state Li-metal battery operated at a high current density of 2.2 mA/cm² at 25 °C. Moreover, this design was confirmed to enable a 100-mAh-level single cell to be stably operated, suggesting its potential applicability to commercial devices. Finally, the safety of the quasi-all-solid-state Li-metal battery was attributed to its fire-retardant capability conferred by the ionic liquid. These findings provide important insights into building and maintaining a conformal interface between the inorganic solid electrolyte and Li-metal negative electrode, and offer guidance for developing Li-metal batteries not only with high energy densities, but also with long lifespans.

Methods

Materials

The LLZTO (Li_6.4La₃Zr_1.7Ta_0.3O₁₂) powder was synthesised from a precursor mixture of Li₂CO₃ (> 99.0%, ChemPoint, USA), La₂O₃ (98.6%, MolyCorp, USA), Ta₂O₅ (99.99%, Sigma-Aldrich, USA), and ZrO₂ (98%, Zircoa Inc., USA) using a solid-state reaction method. The synthesised LLZTO powder was then ball-milled in air with zirconia balls for 10 min at 300 rpm using a planetary ball mill (Pulverisette 7, Fritsch, Germany) prior to fabrication of the pellet- and tape-type electrolytes.

The initial LLZTO powder (100 g) was compacted into a high-density pellet by hot-pressing in a graphite die, applying a pressure of 3 kpsi. The compacted material then underwent sintering at 1100 °C for 2 h in an Ar environment, following a ramp rate of 300 °C/h. The resulting pellet (5.12 g/cm³, 97.5% relative density with respect to the theoretical density of LLZTO) was subsequently micro-machined using a laser cutter in an air environment, establishing the final dimensions as 14 mm in diameter and 360 µm in thickness. Surface treatment involved a 10 min ultrasonic cleaning in hexane, immediately followed by heat treatment in dry air at 800 °C for 1 h. Final planarization was executed via a mechanical polishing process with a LaboForce-3 (Struers) machine, reducing the pellet thickness to a target value of ~ 250 µm.

A large-area (36 cm²) electrolyte film with a thickness of 110 µm was prepared via tape casting. For this purpose, LLZTO and Li₂CO₃ powders (9:1 LLZTO/Li₂CO₃ mass ratio) were added in the same mass to a mixed solvent of propyl propionate and butyl propionate (1:1 volume ratio) with a pre-dissolved Dispersbyk 2155 dispersant (BYK) and stirred at 25 °C for 24 h. Elvacite 2046 binder (ChemPoint, USA) and PL029 plasticiser (Polymer Innovations, USA) were then added to the slurry as the binder and the dispersant, respectively, and mixed for 24 h. The resulting slurry was cast into a thin film on a Mylar® sheet, and then dried at 60 °C. Subsequently, the resulting solid electrolyte tape was sintered at 1100 °C for 1 h. XRD analysis revealed that the 36 cm² LLZTO tape possessed a cubic garnet structure. The density of the solid electrolyte tape was measured to be 4.81 g/cm³, which corresponds to a relative density of 91.65% compared to its theoretical density (5.25 g/cm³).

To remove any impurity phases (e.g., Li₂CO₃) from the LLZTO surface, the LLZTO electrolyte was simply immersed in a glass bottle containing a 1 M HCl solution (prepared in distilled water) in a dry room (dew point, − 60 °C) at 25 °C for 20 min (1:10 electrolyte/acid solution mass ratio). Continuous agitation of the container was maintained using a bottle roller, operating at a rotational speed of ~ 60 rpm during the entire treatment period. This rolling action was critical to suppress local variations in the Li concentration within the acid solution, which could otherwise result from Li release. Furthermore, it precluded direct, static contact between the electrolyte and the glass container walls. The electrolyte disks were removed from the solution, followed by washing with ethanol and drying at 25 °C in a dry room.

The negative electrode interlayer consisted of a layer of carbon or Ag-C (1:3 Ag/C mass ratio) coated on 10 μm-thick stainless steel (SUS) foil. Carbon black powder (99.7%, average particle size = 38 nm, Asahi Carbon) was used for both the carbon and the Ag-C layers. Ag nanoparticles (D₅₀ = 60 nm) were employed to prepare the Ag-C layer. The powders were mixed in N-methyl-2-pyrrolidone (Sigma-Aldrich) solvent containing 7 wt% poly(vinylidene fluoride) binder (Solvay) under constant stirring (1000 rpm) for 30 min using a mixer (Thinky Corporation, AR-100). The prepared slurries were applied to SUS foil through a screen printing technique. The initial drying step was conducted in ambient air at 80 °C for 20 min. Following this, the coated layers underwent a thorough secondary drying process by subjecting them to a vacuum environment at 100 °C for 12 h.

Characterisation and in situ spectroscopy

The cross-sectional microstructure of the interface between the negative electrode and LLZTO electrolyte was examined using field-emission SEM (FE-SEM; Hitachi SU-8030) coupled with EDS at an accelerating voltage of 5 kV and a working distance of 8 mm. For sample preparation, the cycled cells were disassembled in an Ar-filled glove box, and after the removal of the positive electrode, the cross-sections of LLZTO in contact with the Te/Ag-C/Li negative electrode were transferred and secured onto the sample holder within a load-lock chamber environment to rigorously prevent any exposure to ambient air. To mitigate potential thermal and structure damage, the samples were etched with a 5 kV Ar-ion beam for 8 h while the temperature was maintained below − 140 °C. The cross-sectional polishing procedure was optimised using trial-and-error iteration.

A porous or dense Te layer was deposited on top of the LLZTO via sputtering, followed by etching with HCl. Li/Cu was attached to the bottom of the structure to act as the Li source. Exposure to ambient air was minimised by preparing the samples in a glove box filled with pure Ar gas, after which they were transferred to the XPS chamber using a sealed vessel for measurement. The core-level spectra were measured using monochromatic Al-Kα radiation as the excitation source (1486.6 eV beam with a diameter of 200 μm). When an e-beam is used to irradiate the surface of the interlayer, the Li ions move through the electrolyte and the interlayer because of the negatively charged surface potential. Conversely, the use of an ion beam (i.e., neutraliser) to spray positive Ar ions onto the surface renders the surface positively charged, which causes Li ions to migrate away from the surface of the interlayer^40–42. The beam size of the e-beam was approximately 10 mm, while that of the ion beam exceeded 10 mm. The XPS analysis area had a diameter of 200 μm. When measured using the 800 μm Faraday cup holder provided by the manufacturer (PHI), the e-beam neutralisation mode yielded a current exceeding 20 nA, whereas the ion beam neutralisation mode yielded approximately 400 pA. The corresponding target currents were approximately 2 μA and 30 nA, respectively. Each chemical state in the spectra was calculated using curve fitting as a function of the irradiation time. In situ AES measurements were also performed in an ultrahigh-vacuum (UHV, 1 × 10⁻⁸ Pa) chamber at 25 °C using a cylindrical mirror analyser (CMA) and an FE electron gun at 5 keV and 1 nA.

Cell assembly and electrochemical measurements

Quasi-all-solid-state cells were fabricated to perform electrochemical performance evaluations. Initially, a GeTe alloy film with a thickness of ~ 100 nm was coated onto the LLZTO surface via radio-frequency sputtering (SNTEK, 16-SN-055; 200 W for 1800 s; plasma gas: N₂ (8 mTorr)) using a GeTe target (99.99% purity). The mixed atomic ratio of Ge and Te in the GeTe alloy film was 2:1. The GeTe-coated LLZTO was treated with acid to prepare a porous Te layer by the selective etching of Ge. The sample was immersed in a 1 M aqueous HCl solution for 10 s, followed by washing with ethanol. The Ag-C layer was attached as a negative electrode interlayer to the acid-treated LLZTO surface via cold isostatic pressing (CIP) at 250 MPa. Subsequently, Li-metal foil (99.99%, thickness = 20 µm, Honjo Metal Co., Ltd.) was attached to the Ag-C surface via CIP, again at 250 MPa. The 20 µm-thick lithium metal was used for pre-lithiation to improve the initial irreversibility of the Te/Ag-C interlayer.

Commercially available LiNi_0.8Co_0.1Al_0.1O₂ (NCA811, loading capacity: 4.4 mAh/cm² (22.3 mg/cm²) with a specific capacity of 205 mAh g⁻¹; active material: 97 wt%; poly(vinylidene fluoride) (PVDF): 1.0 wt%, carbon black:1.3 wt%; thickness: 62 µm; Samsung SDI) and LiCoO₂ (LCO, loading capacity: 3.2 mAh/cm² (18.8 mg/cm²) with a specific capacity of 170 mA g⁻¹; active material: 97 wt%; poly(vinylidene fluoride) (PVDF): 1.0 wt%, carbon black:1.3 wt%; thickness: 46 µm; Samsung SDI) were employed as the positive electrode active materials. As a current collector, Al foil (9 μm foil, Nippon Foil Mfg. Co., Ltd.) was used as received. The catholyte was prepared by incorporating with 2 M lithium bis(fluorosulfonyl)imide (LiFSI, 99.9%, water content <10 ppm) salt into 1-Ethyl-3-methylimidazolium bis(fluorosulfonyl)imide (EMIFSI, 99.9%, water content < 20 ppm, Kanto Chemical Co. Inc.). The catholyte solution was then infiltrated into the positive electrode, with the solution mass fixed at 25 wt% relative to the electrode mass. All infiltration procedures were conducted inside a dry room maintained at a strict dew point of − 60 °C, and the positive electrode was placed under vacuum for 2 h. Following removal of the residual solution on the positive electrode surface using disposable wipers (Kimwipes, Kimberly-Clark Co. Inc.), the solution uptake by the positive electrode was ~ 10 wt%. The positive electrode infiltrated with the ionic liquid was then placed on the positive electrode side of the LLZTO in a single-layer pouch cell, and the cell was sealed under vacuum (750 Torr).

The sealed pouch cell was subjected to a uniform pressure of 0.3 MPa using a lab-made pressure jig for the cell performance evaluation. Prior to conducting the potentiostatic electrochemical impedance spectroscopy (PEIS) measurements, the cells were maintained at the open-circuit potential for 10 min.

The PEIS was conducted using a system that couples a frequency response analyser (Solartron, SI 1255 FRA, AMETEK Inc., USA) with a potentiostat (Solartron, SI 1287 ECI, AMETEK Inc., USA). All measurements were performed at 25 °C under open circuit potential in potentiostatic mode. The applied AC perturbation was 10 mV, and the frequency range swept from 0.1 Hz to 10 kHz. Data acquisition density was strictly controlled, recording 10 points per frequency decade.

The charge/discharge performance of the quasi-all-solid-state cells was evaluated at a temperature of 25 °C using a battery cycler (Toscat-3100, Toyo System Co., LTD., Japan). The established cycling routine involved a cycling protocol implemented a constant current (CC)–constant voltage (CV) method for charging and a CC method for discharging. Distinct potential windows were applied based on the cathode material: 2.8–4.4 (vs. Li/Li⁺) for NCA811and 2.8–4.5 V (vs. Li/Li⁺) for LCO.

All the electrochemical test was performed within the environmental chamber of constant temperature (temperature error: ± 1 °C, Shin Corporation). Five cells from each sample were evaluated to determine the data reliability. The median value from these measurements was then presented.

Computational details

DFT calculations were performed using the Vienna Ab initio Simulation Package (VASP)⁴³. The Perdew–Burke–Ernzerhof parameterisation of the generalised gradient approximation was employed as an exchange-correlation functional⁴⁴, while a projector-augmented wave pseudopotential was employed for all calculations, as provided by VASP⁴⁵. To describe the van der Waals interactions when calculating the work of adhesion for models including Te or C, the DFT-D3 dispersion correction was applied⁴⁶. All input parameters for the DFT and AIMD calculations were set by pymatgen⁴⁷, using the pymatgen.io.vasp.sets module.

NEB calculations were performed to investigate the activation barrier for Li diffusion on the (111) surface of Li₂Te⁴⁸. Seven intermediate images were created along the diffusion path to evaluate the energy profile of Li diffusion.

To build the interface models, the method described by Kim et al.⁵ was adopted, and the pymatgen Python package was used to generate interface models with thick vacuum slabs. The work of adhesion was derived by comparing the energies of the interfacial and isolated slab structures⁵.

MD calculations were performed to study the alloying behaviour. Initially, AIMD simulations were carried out to accurately describe the alloying reaction using the interface slab model. Li/M (M = Ag, Sn, or Te) bilayer slab models were constructed using the method described above based on an atomic ratio of ~ 10:1 for Li:M. The system was heated to the target temperature (300, 400, or 500 K) in 2 ps, and the temperature was maintained for at least 100 ps in the NVT ensemble to explore the reaction progress. The resulting AIMD trajectories were used to train the MLIP using deepmd-kit⁴⁹. We ensured that the root mean square error of the trained MLIP is less than 1 meV per atom. Since the AIMD calculations were performed over a wide range of temperatures for long periods of time, the various intermediate stages that occur during alloying were considered to be well covered in the training sets, along with the pristine Li and M structures.

Finally, MLIP-MD simulations were conducted using a large-scale atomic/molecular massive parallel simulator (LAMMPS)⁵⁰. Similar to the AIMD simulations, interface slab models were built with a Li:M atomic ratio of ~ 10:1. However, a larger model was built to scrutinise the alloying reaction, wherein the total number of atoms was ~ 3000. The system was heated to 400 K, and the temperature was maintained for > 10 ns in the NPT ensemble to thoroughly observe the reaction. A simulation time of 10 ns was not sufficient to capture the entire reaction at 300 K; therefore, to accelerate the reaction, the simulation temperature was set to 400 K, which is below the melting temperatures of Li and the Li-M alloys.

Author contributions

J.-S.K., G.Y., Y.S.K. and S.H. conceived and designed the overall experiments, analysed the data and wrote the manuscript. J.-S.K., T.-H.K., S.K., J.K., J.L., R.K. and M.-J.L. performed all electrochemical experiments. G.Y. performed the DFT and AIMD calculations on the interface metal nanolayer. N.Y., S.S. and T.A. prepared the carbon-based negative electrode interlayers. Y.S.K. performed the X-ray photoelectron spectroscopy (XPS) analysis. S.H. performed the Auger electron spectroscopy (AES) analysis. Y.S.K. and T.-H.K. examined the cross-sectional microstructure of the interface between the LLZTO and the negative electrode interlayer. M.B. and Z.S. prepared the tape-cast LLZTO electrolyte. T.-H.K. performed flammability tests for the positive electrodes with ionic liquid. J.K. performed the nail penetration tests on the single-pouch cells. All the authors participated in the discussion and provided constructive advice for experimental design.

Peer review

Peer review information

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

Data availability

The data that support the findings of this study are available in the Supplementary Material of this article. Source data are provided in this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-66308-4.