Direct visualization of complex mechanisms in lithium tellurium batteries

Abstract

Many studies on lithium-sulfur batteries have been conducted with respect to reaction mechanisms and performance improvements. Despite the high electronic conductivity and volumetric capacity of tellurium, research on lithium-tellurium batteries remains relatively limited. Herein, we shed light on the unique reaction mechanism of lithium-tellurium batteries: a solid-liquid-solid-liquid-solid pathway mediated by the solid Li₂Te₆ intermediate, as revealed by operando analyses, in contrast to the simpler solid-liquid-solid mechanism of lithium-sulfur and lithium-selenium batteries. Moreover, we demonstrate that Li₂Te₆ morphology and size are path-dependent: long-chain lithium polytellurides during discharge yield small cubic-like Li₂Te₆ that dissolve completely, whereas short-chain lithium polytellurides during charge form larger Li₂Te₆ that do not fully dissolve and enable subsequent tellurium growth. Surprisingly, path-dependency causes residual Li₂Te₆ to persist in subsequent cycles, thereby degrading battery performance. To address this limitation, we propose tailored charge protocols to control Li₂Te₆ crystal growth, guided by the reaction mechanisms revealed in this study.

Lithium-tellurium batteries promise high-energy storage, but their reaction pathways remain poorly understood. Here, authors directly visualize a solid Li₂Te₆ intermediate and show that its path-dependent growth and persistence govern electrode morphology, utilization and cycling performance.

Introduction

There is a growing demand for high-energy-density batteries to meet the needs of electric vehicles (EVs) and hybrid electric vehicles (HEVs). Among various candidate materials, chalcogen elements such as sulfur (S), selenium (Se), and tellurium (Te) are attractive owing to their high theoretical capacities (S: 1675 mAh g⁻¹, Se: 675 mAh g⁻¹, Te: 420 mAh g⁻¹) and volumetric capacities (S: 3467 mAh cm⁻³, Se: 3254 mAh cm⁻³, Te: 2621 mAh cm⁻³).^1–3 In particular, lithium-tellurium (Li-Te) batteries have gained attention because of high electrical conductivity of Te (S: 5 × 10⁻²⁸ S m⁻¹, Se: 1 × 10⁻³ S m⁻¹, Te: 2 × 10² S m⁻¹), which enhances reaction kinetics.^4–10 Like other chalcogen systems, Li-Te batteries undergo conversion reactions via soluble lithium polychalcogenides, causing the shuttle effect and associated issues such as self-discharge and lithium negative electrode instability.^11–16 Consequently, most efforts have focused on positive electrode design to improve electrochemical performance.^17–29 However, the reaction mechanism remains poorly understood and under debate.^7,30–36 A precise mechanistic understanding is thus essential for advancing high-performance Li-Te batteries.

Herein, we uncover that Li-Te batteries follow a more complex reaction mechanism than Li-S and Li-Se batteries due to the presence of an additional solid Li₂Te₆ intermediate phase, as demonstrated by real-time monitoring of Li-Te batteries during cycling. Notably, Li₂Te₆ forms during discharge and charge, but its crystal size and morphology are strongly path-dependent, which critically influences the electrochemical performance of the Li-Te batteries. To mitigate the adverse effects of the active material path-dependency, we propose a strategy to control Li₂Te₆ crystal growth by modulating current density during charging based on the mechanistic insight. This approach enhances Te utilization and improves cycling stability. This work provides fundamental insights into the redox chemistry of Li-Te batteries and offers a foundation for the design of high-performance Li-Te batteries.

Results and discussion

Fundamental differences between Li-S and Li-Te reactions: the emergence of Li₂Te₆ as a crystalline intermediate

Voltage profiles of Li-S and Li-Te cells at 0.5 C (837.5 mA g^–1_S and 210 mA g^–1_Te) show voltage polarizations of 0.16 and 0.13 V, respectively, defined as the charge-discharge voltage gap at 50% SOC (Fig. 1a).^37–39 The lower voltage polarization of the Li-Te cell is consistent with the superior electronic conductivity of Te (Fig. 1b).^3,4,7 However, conductivity alone may not fully explain the mechanistic differences. To prove possible thermodynamic origins, we examined the phase diagrams: Li-S comprises only S and Li₂S, whereas Li-Te includes an extra solid phase, called lithium tritelluride (LiTe₃) (Fig. 1c). This raises the key mechanistic question of whether the additional phase appears during cycling and participates electrochemically.

Fig. 1. Comparison between sulfur and tellurium.

a Voltage profiles of Li-S and Li-Te cells during the first discharge and charge at 0.5 C, corresponding to 837.5 mA g^–1_S and 210 mA g^–1_Te, respectively. b Electronic conductivities of sulfur and tellurium. c Phase diagrams of the Li-S and Li-Te systems. d Operando XRD patterns during the first cycle at 0.2 C (84 mA g^–1_Te) and the corresponding galvanostatic profile. e Ex-situ SEM images in BSE mode of Te electrodes at various depths of discharge (DODs) and states of charge (SOCs). Li-Te cells were cycled during the first cycle at 0.2 C (84 mA g^–1_Te) between 1.3 and 2.7 V in a constant-temperature chamber maintained at 25.0 ± 0.2 °C and disassembled at D50, D100, S73 and S100 states for ex-situ SEM analysis. The pristine electrode was analyzed before electrochemical cycling. Source data are provided as a Source Data file.

To address the question, we performed operando X-ray diffraction (XRD) analysis on Li-Te cells to monitor phase evolution and crystallization during cycling (Fig. 1d). The Li-Te cell voltage profile was annotated with depth of discharge (DOD, e.g., D19 = 19%). During the initial discharge plateau, trigonal Te peaks (PDF: 36–1452) diminished, indicating conversion to long-chain lithium polytellurides (Li₂Te_x, 6<x ≤ 8). After D19, new reflections near 20° and 29° identified trigonal Li₂Te₆ (PDF: 31–757), confirming an additional solid phase. Although previous studies have suggested the presence of Li₂Te₆ based on in-situ Raman spectroscopy and TEM analysis, we provide the first operando evidence of crystalline Li₂Te₆ phase formation during cycling in a Li-Te cell using an ether-based electrolyte.^7,32 As the discharge process proceeds, Te peaks disappear after the first plateau and Li₂Te₆ dissolved at D71, while cubic Li₂Te (PDF: 47–1697) appeared at D92 as the final discharge products. On charging, the Li₂Te persisted until reaching 40% state of charge (referred to as S40) and then disappeared gradually due to oxidation to short-chain lithium polytellurides (Li₂Te_x, 2≤ x < 6). These short-chain soluble lithium polytellurides underwent conversion into the Li₂Te₆ crystalline phase. Notably, Li₂Te₆ peaks partially remained even after Te re-formed at full charge.

Ex-situ scanning electron microscopy (SEM) images of the Te electrode at specific DODs and SOCs were consistent with the XRD results (Fig. 1e). Initially, the Te electrode appeared as multi-plate lumps. At D50, Te was fully dissolved into long-chain polytellurides (LiPTe I), leaving only cubic-like Li₂Te₆ crystals. At D100, Li₂Te formed as micrometer-sized particles. During charging, Li₂Te transformed into short-chain lithium polytellurides (LiPTe II) and recrystallized into Li₂Te₆ at S73. At S100, Li₂Te₆ converted into multi-plate Te via LiPTe I. Supplementary Figs.1 and 2 further illustrate Te dissolution and Li₂Te₆ formation. Collectively, these results establish Li₂Te₆ as an electrochemically active solid intermediate unique to Li-Te batteries, marking a fundamental mechanistic deviation from Li-S and Li-Se systems.

Reaction mechanisms of Li-Te batteries under discharge and charge conditions

Operando transmission X-ray microscopy (TXM) imaging (Supplementary Video 1) tracked Te electrodes during cycling, showing the morphology of Te with high intensities (yellow) and the thread-like features of carbon-fiber current collector with relatively low intensities (blue) (Fig. 2a–c). Te crystals dissolved during the first plateau, and Li₂Te₆ particles (red circles) appeared at new sites by D19, consistent with SEM images. The Li₂Te₆ particles later converted to soluble lithium polytellurides, with full discharge yielding small Li₂Te. At higher current density (0.5 C, 210 mA g^–1_Te), the overall sequence was similar to that at 0.2 C (84 mA g^–1_Te), but Li₂Te₆ formation shifted to D32 and its dissolution was incomplete, showing that faster kinetics hinder phase transitions (Fig. 2 and Supplementary Video 2).

Fig. 2. Investigation of discharge and charge processes in Li-Te batteries with direct observation of Li₂Te₆ intermediates.

a Voltage profile of the Li-Te cell during the discharging process at 0.2 C (84 mA g^–1_Te) for the initial cycle. b Operando TXM images of the Te electrode at various DODs at 0.2 C (84 mA g^–1_Te) and c the corresponding enlarged area of the red box. d Operando TXM images of the Te electrode at various DODs at 0.5 C (210 mA g^–1_Te). e Operando optical images of the Te electrode at different DODs. f Voltage profile of the Li-Te cell during the charging process at 0.2 C (84 mA g^–1_Te) for the initial cycle. g Operando TXM images of the Te electrode at various SOCs and h the corresponding enlarged area of the red box. The color scales in b–d, g and h represent normalized TXM intensity [arb.units], with lower and upper limits of 0 and 1, respectively. Red circles indicate Li₂Te₆ particles and white dashed outlines indicate the regions highlighted for comparison in the TXM images. i Ex-situ SEM images in a BSE mode, ex-situ TEM images, and the corresponding SAED patterns of Te electrodes at S73 and at S100 states. Li-Te cells were cycled during the first cycle at 0.2 C (84 mA g^–1_Te) between 1.3 and 2.7 V in a constant-temperature chamber maintained at 25.0 ± 0.2 °C. The cells were stopped at S73 and S100 and disassembled in an Ar-filled glovebox (O₂ and H₂O < 1 ppm) for ex-situ SEM, TEM and SAED analyses. j 3D X-ray CT images of the Te electrodes at S73 and S100. TXM, transmission X-ray microscopy; BSE, backscattered electron; TEM, transmission electron microscopy; SAED, selected-area electron diffraction. Source data are provided as a Source Data file.

Operando optical imaging was conducted (Fig. 2e, Supplementary Fig. 2 and Supplementary Video 3). At D19, the separator turned red-brown as polytellurides dissolved, and a boundary probably indicating the presence of the solid Li₂Te₆ emerged between long-chain and short-chain lithium polytellurides.⁴⁰ After D71, lighter red-brown short-chain lithium polytellurides diffused out, leaving the separator pale at full discharge. These direct visualization analyses during the discharge process provide insight into the dissolution of lithium polytellurides and confirm Li₂Te₆ as a redox-active crystalline intermediate, distinguishing Li-Te from Li-S.

During charging, operando TXM analysis showed Li₂Te₆ nucleation after S56 and growth until S73 (Fig. 2f–h). Interestingly, the Li₂Te₆ crystals formed during charging appeared significantly larger than those formed during discharge. As the charge progressed further, the Li₂Te₆ crystals gradually dissolved; however, some remained undissolved even after the growth of Te crystals. Meanwhile, the thread-like yellow fibers, which indicate lithium polytellurides, evolved into a clear blue background upon oxidation. Notably, some Te crystals formed independently, while others grew on residual Li₂Te₆ crystals, indicating incomplete conversion.

Ex-situ SEM images corroborated these findings (Fig. 2i and Supplementary Fig. 3). At S73, large cubic Li₂Te₆ crystals were re-formed, and at S100 Te appeared as stacked triangular plates. Interestingly, some Li₂Te₆ crystals (marked by a red arrow) had not dissolved even at S100. To further characterize the crystals, ex-situ transmission electron microscopy (TEM) and selected area electron diffraction (SAED) analyses were conducted on the SEM samples (Fig. 2i). At S73, the SAED pattern obtained from an angular ~10 μm particle was indexed to single-crystalline Li₂Te₆. At S100, the SAED pattern was indexed to crystalline Te, with the (-1 1 -2) and (1 0 -1) plane traces aligned along the long and short axes, respectively. These results support the imaging analysis, confirming the phase identities and crystallographic orientations of the Li₂Te₆ and Te crystals. 3D nano-computed tomography (CT) was also performed to visualize the large-scale crystal morphology (Fig. 2j). At S73, only cube-like Li₂Te₆ was observed. In contrast, the fully recharged positive electrodes exhibited a coexistence of Li₂Te₆ and elongated Te crystals. Together, these results reveal the path-dependent formation, persistence, and morphological divergence of Li₂Te₆ as a defining feature of Li-Te chemistry.

Complex reaction mechanisms in Li-Te batteries (1): Path-dependent Li₂Te₆ formation

To understand why the Li₂Te₆ crystals remain partially undissolved during charging, despite full dissolution during discharge, we examined the operando TXM images. Unlike Li-S or Li-Se batteries, which exhibit a simple solid-liquid-solid phase transition, Li-Te batteries feature an additional solid-state intermediate, Li₂Te₆, as revealed by XRD and TXM (Fig. 3a, b). There is an additional intermediate solid phase, Li₂Te₆, and the size of the Li₂Te₆ crystals varies depending on the reaction path. To provide precise information, we extracted TXM images at specific points where Li₂Te₆ formed and marked the Li₂Te₆ crystals with arrows (Fig. 3c, d). Li₂Te₆ crystals formed via two distinct pathways: reduction of long-chain lithium polytellurides (cyan) and oxidation of short-chain lithium polytellurides (red). The Li₂Te₆ particles formed from long-chain lithium polytellurides grew along the current collector fiber and were relatively smaller than those formed from short-chain lithium polytellurides. The larger size of the Li₂Te₆ formed from short-chain lithium polytellurides makes it difficult to fully oxidize into the Te phase, resulting in a significant portion remaining in the Li₂Te₆ phase. This phenomenon could aggravate not only the utilization of Te but also the overall performance of Li-Te batteries.

Fig. 3. Path-dependent Li₂Te₆ growth in Li-Te batteries.

Schematic illustration of reaction pathways in a Li-S or Li-Se batteries and b Li-Te batteries. TXM images of the Te electrode and the corresponding enlarged views of c sky-blue boxes (during discharge at D50) and d red boxes (during charge at S73), accompanied by schematic illustrations of the Li-Te reaction pathway. Cyan and red denote long-chain and short-chain lithium polytelluride pathways, respectively. Arrows indicate the direction of Li₂Te₆ formation and conversion. Colored boxes indicate the enlarged regions shown in the corresponding panels. Source data are provided as a Source Data file.

We also conducted additional operando TXM imaging on a Te positive electrode using Al current collector to demonstrate that this phenomenon arises from the chemo-mechanical properties of Te, and it is not influenced by the carbon current collector (Supplementary Video 4). The morphology of Te was depicted with high intensities (pale blue), and the background was illustrated with relatively low intensities in black (Supplementary Fig. 4). We compare the TXM images of specific DOD and SOC, which show only Li₂Te₆ crystals. These measurements clearly show that Li₂Te₆ decreases in size during discharge and increases during charge, regardless of the type of current collector used.

Complex reaction mechanisms in Li-Te batteries (2): Seed role of residual Li₂Te₆

To investigate the morphological behavior of Te and Li₂Te₆ crystals in detail, higher magnification TXM analysis revealed that Li₂Te₆ crystals partially dissolved as Te crystals grew during the first charge, as indicated by white dotted lines (Fig. 4a, b). Subsequently, after the Te crystal dissolved, the Li₂Te₆ crystals re-formed and again dissolved during the second discharge process. This growth and dissolution behavior of the Li₂Te₆ crystals was more clearly observed during the second charge up to the third discharge process. It is noteworthy that during the re-discharge process, Li₂Te₆ tends to grow on the pre-existing Li₂Te₆ crystals formed during the previous charge process. To further quantify the seed effect of residual Li₂Te₆ on subsequent cycles, we calculated the overlap fraction of Li₂Te₆-dominant regions between each state and its immediately preceding Li₂Te₆ dominant state over two consecutive cycles (Supplementary Fig. 5). The overlap fraction between 1st D45 and 1st S73 was 16.28%, indicating limited spatial coincidence between Li₂Te₆ domains formed during discharge and charge in the first cycle. In contrast, the overlap fraction markedly increased to 78.82% between 1st S73 and 2nd D47, suggesting that Li₂Te₆ formed during the second discharge preferentially reappeared at locations previously occupied during the preceding charge. The overlap fraction between 2nd D47 and 2nd S73 remained 46.76%, which was still substantially higher than that between 1st D45 and 1st S73. These results imply that the remaining Li₂Te₆ crystals could influence the subsequent chemo-mechanical growth mechanism. To explore its morphological implication, we compared the surface structure of Li₂Te₆ at full charge and S73 (Fig. 4c, d). Te was observed to form directly on Li₂Te₆ surfaces (Case I), to grow atop Li₂Te₆ (Case II), or to nucleate independently (Case III), consistent with the aforementioned TXM imaging results (Fig. 4e). This growth behavior highlights the complex growth behavior in the Li-Te batteries during the charging process. A schematic in Supplementary Fig. 6 summarizes the structural evolution and spatial relationship between Li₂Te₆ and Te during charging.

Fig. 4. Investigation of morphological behavior of Li₂Te₆ and Te.

a Voltage profile of the Li-Te cell at 0.5 C (210 mA g^–1_Te). b High resolution operando TXM images of the Te electrode at various DODs and SOCs. White and green dashed outlines in TXM images mark the tracked Li₂Te₆ and Te regions, respectively. c Voltage profile of the Li-Te cell during charging process at 0.2 C (84 mA g^–1_Te). d Operando TXM images of the Te electrode at various SOCs. Orange circles indicate Li₂Te₆ seeds, white circles indicate Te growth, and the red boxes identify the regions corresponding to Cases I–III shown in the ex-situ SEM images in panel e. e Ex-situ SEM images in a BSE mode of Te electrodes at S73 and S100 states. Li-Te cells were cycled during the first cycle at 0.2 C (84 mA g^–1_Te) between 1.3 and 2.7 V in a constant-temperature chamber maintained at 25.0 ± 0.2 °C. The cells were stopped at S73 and S100 and disassembled in an Ar-filled glovebox (O₂ and H₂O < 1 ppm) for ex-situ SEM analysis. Case I, Case II and Case III indicate Te formation directly on Li₂Te₆ surfaces, Te growth atop Li₂Te₆ and independent Te nucleation, respectively. Source data are provided as a Source Data file.

Building on the dissolution and re-formation dynamics revealed in Fig. 4, we next examined how residual Li₂Te₆ crystals influence subsequent nucleation and growth, providing mechanistic insight into their role as persistent seeds. Figure 5 presents the nucleation site distribution map, tracing Li₂Te₆ and Te crystals at notable points during the first charge and subsequent discharge. The emerging Li₂Te₆ nuclei are highlighted with orange circles, whereas Te nuclei are indicated with white circles. Te crystals began to grow from S81, while some Li₂Te₆ crystals dissolved. After the first charge, the dissolution of Li₂Te₆ and Te was tracked using orange and white dotted lines, respectively (Fig. 5a, b). Tracking specific Li₂Te₆ crystals during cycling clearly shows that some remained in the same position until S100, while others (marked by red circles) dissolved (Fig. 5c). During the dissolution of Te crystals, Li₂Te₆ growth was observed on pre-existing Li₂Te₆, demonstrating that residual Li₂Te₆ crystals can act as seeds for subsequent processes. To quantify the seed effect of Li₂Te₆ during charging and the subsequent discharge process, we analyzed the overlap fraction in Supplementary Fig. 7. For Supplementary Fig. 7a, the Li₂Te₆-dominant reference state was S70 during charging, whereas for Supplementary Fig. 7b, the Li₂Te₆-dominant reference state was S64 during the following discharge. During charging, both the overlap fraction and the seed associated Te fraction gradually decreased but remained substantial throughout the process (Supplementary Fig. 7a). The results indicate that a considerable portion of Te formation occurred on or near pre-existing Li₂Te₆ domains. In addition, Li₂Te₆ related domains identified at S64 showed strong spatial persistence up to S100, with an S64 derived domain persistence fraction of 96.38% (Supplementary Fig. 7b). Together with direct tracking and morphological observations in Fig. 5, this strong spatial persistence is consistent with incomplete conversion of charge-formed Li₂Te₆ derived domains at the end of charge. During the subsequent discharge, both the overlap fraction and the seed associated Li₂Te₆ fraction increased, indicating that these residual Li₂Te₆ domains were associated with renewed Li₂Te₆ formation in later cycles. To further evaluate the morphological consequence of this persistent seed behavior, we quantified the average equivalent diameter and standard deviation of Li₂Te₆ domains formed during the first charge and the subsequent discharge (Supplementary Fig. 8). Li₂Te₆ formed during the second discharge exhibited both a larger average particle size and a broader size distribution than Li₂Te₆ formed during the first charge. Consistent with this interpretation, the fractions of particles with equivalent diameters exceeding 5 and 8 µm were also higher during the subsequent discharge. These results suggest that residual Li₂Te₆ is associated with the formation of larger and more heterogeneous Li₂Te₆ particles in later cycles. Operando X-ray imaging at a higher C-rate of 0.5 C (210 mA g^–1_Te) further supported this behavior (Supplementary Fig. 9). At higher current density, the conversion reaction was accelerated, leading to incomplete dissolution of Li₂Te₆ after the second discharge, which then served as seeds for the subsequent charge. By the third discharge, these crystals no longer dissolved and became electrochemically inactive Te. Collectively, these results establish Li₂Te₆ as a highly stable intermediate whose persistence enables seed-mediated growth but ultimately hinders complete Te utilization.

Fig. 5. Incomplete dissolution of Li₂Te₆ as nucleation seeds for later cycles.

a Voltage profile of the Li-Te cell during charging process at 0.2 C (84 mA g^–1_Te) for the first charging and second discharging process. b Operando TXM images of the Te electrode at various DODs and SOCs. c TXM images tracking the Li₂Te₆ crystals at specific points. Orange circles indicate newly formed Li₂Te₆ nuclei, white circles indicate Te nuclei and red circles indicate Li₂Te₆ crystals that dissolve during cycling. Orange and white dashed lines track the dissolution of Li₂Te₆ and Te crystals, respectively. Source data are provided as a Source Data file.

Kinetic origin and control of Li₂Te₆ crystals for improved Li-Te cell performance

To provide mechanistic insight into the distinct Li₂Te₆ particle sizes (~30 μm during charging vs. ~10 μm during discharging) in terms of thermodynamic and kinetic properties, we performed combined density functional theory (DFT) and kinetic Monte Carlo (kMC) simulations. Free energy calculations suggested that Li₂Te₆ formation during discharge faces relatively low barriers (0.26 eV from Li₂Te₇ and 0.18 eV toward Li₂Te₅), which could facilitate rapid conversion and suppress particle growth, whereas charging faces higher energy requirements (0.37 eV from Li₂Te₅ and 0.47 eV toward Li₂Te₇), which may contribute to the persistence of Li₂Te₆ in its solid state (Fig. 6a–c and Supplementary Fig. 10). Detailed free energy and activation energies for Li₂Te_x interconversion are summarized in Supplementary Table 1. To complement the static DFT calculations, we conducted kMC simulations that provide a dynamic perspective, assessing the transient nature of Li₂Te₆ during discharge and its prolonged persistence during charge. The kMC results are qualitatively consistent with this kinetic asymmetry: Li₂Te₆ behaves as a transient intermediate during discharge but accumulates and persists during charge, consistent with the experimentally observed difference in Li₂Te₆ persistence (Fig. 6d, e and Supplementary Fig. 11). Consequently, by integrating DFT-derived activation energies into kMC simulations, we qualitatively captured the time-dependent interconversion of Li₂Te_x species, supporting the interpretation that reaction kinetics play an important role in the distinct formation and persistence of Li₂Te₆ during charge and discharge. The calculations employ an idealized graphene substrate with limited periodicity, an implicit-solvent treatment, and a simplified interfacial environment, and thus may not fully reproduce all structural, kinetic, and electrolyte effects present in an operating cell. Nevertheless, the consistent trends across DFT, kMC, and operando measurements provide physically grounded mechanistic support for the experimentally observed kinetic asymmetry.

Fig. 6. DFT and kMC mechanistic insights into charging-induced Li₂Te₆ particle enlargement and its mitigation by charging protocols.

a Calculated activation energies required for Li₂Te_x interconversion during charging and discharging processes. Free energy diagrams (FEDs) illustrating Li₂Te₆ formation during b discharging and c charging processes under DOL/DME solvated conditions. kMC simulation prediction of time evolution of the number of molecules for various Li-Te intermediates for the d discharging process and e charging process. Operando XRD patterns and the corresponding galvanostatic profile during the first charging process at 0.2 C (84 mA g^–1_Te) f with voltage holding and g with rest process. Operando XRD patterns and the corresponding galvanostatic profile during the first charging process h at 0.2 C (84 mA g^–1_Te) and i at 0.5 C (210 mA g^–1_Te). DFT, density functional theory; kMC, kinetic Monte Carlo; FED, free energy diagram. In the atomistic models, pink, orange and gray spheres represent Li, Te and C atoms, respectively. Source data are provided as a Source Data file.

To verify the chemo-mechanical stability of the Li₂Te₆ crystals, operando XRD analysis was performed (Fig. 6f, g). Two cells were compared under different conditions: one held at a voltage of 2.7 V after one cycle, while the other remained at rest. The peaks of the Li₂Te₆ became smaller and eventually disappeared when the voltage was held at 2.7 V. However, even after more than 5 h of rest, some peaks of the crystals remained partially visible. From this result, Li₂Te₆ was found to be a very stable species that did not convert into Te instantly, and if a higher driving force was applied, Li₂Te₆ could convert into Te. The XRD analysis was compared to the conversion rate of Li₂Te₆ by controlling current density. In Fig. 6h, the full-width at half maximum (FWHM) of the Li₂Te₆ peak (0.396) was larger than that of the Te peak (0.201). However, in Fig. 6i, the FWHM of the Te peak (0.232) was larger than that of the Li₂Te₆ peak (0.184). Although these FWHM trends suggest different peak-broadening behaviors at different current rates, FWHM comparison alone does not directly quantify conversion kinetics, as it reflects the convolution of multiple factors, including instrumental broadening, microstrain, preferred orientation, and crystallite size. To further support this interpretation, we analyzed the SOC-resolved evolution of the integrated peak area during charge (Supplementary Fig. 12). The Li₂Te₆ signal persists to higher SOC at 0.5 C (210 mA g^–1_Te) than at 0.2 C (84 mA g^–1_Te), while the conversion of the Te signal is delayed and remains substantially lower at 0.5 C (210 mA g^–1_Te). The results indicate that Li₂Te₆ to Te conversion is kinetically hindered at a higher current rate. Encouraged by the promising results obtained thus far, we investigated the electrochemical performance to compare the impact of the Li₂Te₆ crystal conversion (Supplementary Fig. 13). The Li-Te cell, depicted in yellow, was discharged and charged at 1 C in the voltage range of 1.3 – 2.7 V. For comparison, another cell, depicted in sky-blue, was discharged at 1 C (420 mA g^–1_Te) and charged at 1 C (420 mA g^–1_Te) until reaching 1.9 V, then charged at 0.1 C (42 mA g^–1_Te) to induce a greater conversion of Li₂Te₆ crystals (Supplementary Fig. 13a). The experimental cell exhibited a higher capacity of 191.1 mAh g⁻¹ over 30 cycles. In contrast, the cell, which cycled at 1 C (420 mA g^–1_Te), demonstrated a lower discharge capacity of 20.9 mAh g⁻¹ at the 30th cycle. We also implemented a cycling test in Li-S batteries, which have a chemically similar battery system to Li-Te batteries (Supplementary Fig. 13d). The Li-S cell, depicted in purple, was discharged and charged at 1 C (1675 mA g^–1_S) in the voltage range of 1.7 – 2.8 V. For comparison, Li-S cell depicted in pale blue, was discharged at 1 C (1675 mA g^–1_S) and charged at 1 C (1675 mA g^–1_S) until reaching 2.50 V, then charged at 0.1 C (167.5 mA g^–1_S). In the Li-S case, there is no improvement in cycle performance, indicating that the enhancement observed in Li-Te cells is due to the oxidation of the new solid phase, Li₂Te₆ crystals. These findings provide compelling evidence that our proposed approach, inducing the conversion of Li₂Te₆ crystals, significantly enhances the active material utilization in Li-Te battery performance.

In this study, we elucidated the reaction mechanism of Li-Te batteries and identified Li₂Te₆ as the only solid-phase intermediate, which is not seen in Li-S and Li-Se batteries, with crucial implications for their performance and efficiency. Notably, Li₂Te₆ formation is strongly path-dependent: during discharge, small and transient Li₂Te₆ crystals dissolve completely, whereas during charge, larger and persistent crystals remain as seeds for subsequent Te growth. The DFT/kMC results support a underlying kinetic basis for this asymmetry, while the detailed manifestation of this mechanism in an operating cell may be further influenced by structural, interfacial, and electrolyte complexity. By tailoring charging protocols to suppress uncontrolled residual Li₂Te₆ crystallization, we achieved a high capacity of 191.1 mAh g⁻¹ after 30 cycles at 1 C (420 mA g^–1_Te), compared with only 20.9 mAh g⁻¹ without such treatment, demonstrating a practical route to overcome this mechanistic bottleneck. These findings redefine the fundamental redox chemistry of Li-Te batteries and provide important guidance for advancing Te-based batteries toward high-energy and long-life applications.

Methods

Li-S and Li-Te phase diagrams

The Li-S and Li-Te phase diagrams were replotted from previously reported literature phase diagrams by setting the temperature range to 0–1400 °C and redrawing each diagram to dimensions of 83 mm in width and 44 mm in height while maintaining the original aspect ratio.^41,42

Operando XRD measurements

We designed a custom coin cell with a central hole and sealed it with Kapton film. The electrode slurry was prepared with Te (99.8%, Sigma Aldrich), Super P carbon (estimated carbon content ≥99.5 wt%, average particle size: 400 nm, Timcal), and polyvinylidene fluoride (PVDF) (Solef 5130, Mw:1,000,000-1,100,000 g mol^–1, PVDF content > 99.9%, Solvay) in an 8:1:1 weight ratio. The mixtures were separately dispersed in N-methyl-2-pyrrolidone (NMP; HPLC grade, ≥99%, Sigma-Aldrich) to form homogeneous slurries and cast onto AvCarb P50 carbon paper (Ballard, thickness: 184 µm, density: 0.40 g cm^–3, compressibility: 14.0%, mean pore size: 10 µm) using a doctor blade (SeKTQ150, MTI Korea). The coated electrodes were dried under vacuum at 60 °C for 8 h. The Te mass loadings were ~3 mg cm⁻². The coated material was then punched into 11 mm diameter disks to serve as the positive electrodes using an electrode punch (P02, ROTECH, Korea). In an argon-filled glovebox (O₂, H₂O < 1 ppm), we assembled a CR2032-type custom coin cell. The cell contained Te disks as the working electrode, with a Li metal foil (99.95%, Honjo, thickness: 350 µm) cut into 15 mm diameter disks used as both the counter and reference electrodes using a 15-mm-diameter electrode punch (P02, ROTECH, Korea). A glass fiber membrane (Whatman GF/B; catalog number: 1821-110; punched diameter: 19 mm; thickness: 0.675–0.68 mm; nominal pore size: 1.0 µm; estimated porosity: ~90%) served as the separator. The electrolyte was a 160 µL solution of 1.0 M lithium bis(trifluoromethanesulfonyl)imide (LiTFSI) (99.95%, Sigma Aldrich) in a 1:1 volume ratio mixture of 1,3-dioxolane (DOL) (anhydrous, 99.8%, Sigma Aldrich) and 1,2-dimethoxyethane (DME) (anhydrous, 99.5%, Sigma Aldrich). Operando XRD analysis was carried out at beamline 3D of the Pohang Accelerator Laboratory (PAL), with 2D MAR345 diffraction images processed using the Fit2D program and calibrated with a LaB₆ diffraction pattern. Galvanostatic charge/discharge tests were performed using a potentiostat (ZIVE SP1, WonA Tech, Korea) in the experimental hutch at the Pohang Accelerator Laboratory without an environmental chamber. The measurements were conducted under ambient room-temperature conditions, nominally at 25 °C, with the cells cycled between 1.3 and 2.7 V at a rate of 0.2 C (84 mA g^–1_Te) and 0.5 C (210 mA g^–1_Te).

Operando TXM measurements

We designed a custom coin cell with a central hole and sealed it with Kapton film. The electrode slurry was prepared with Te (99.8%, Sigma Aldrich), Super P carbon (estimated carbon content ≥99.5 wt%, average particle size: 400 nm, Timcal), and polyvinylidene fluoride (PVDF) (Solef 5130, Mw:1,000,000-1,100,000 g mol^–1, PVDF content > 99.9%, Solvay) in an 8:1:1 weight ratio. The mixtures were separately dispersed in N-methyl-2-pyrrolidone (NMP; HPLC grade, ≥99%, Sigma-Aldrich) to form homogeneous slurries and cast onto AvCarb P50 carbon paper (Ballard, thickness: 184 µm, density: 0.40 g cm^–3, compressibility: 14.0%, mean pore size: 10 µm) using a doctor blade (SeKTQ150, MTI Korea). The coated electrodes were dried under vacuum at 60 °C for 8 h. The Te mass loadings were ~3 mg cm⁻². The dried electrodes were punched into 11-mm-diameter disks using an electrode punch (P02, ROTECH, Korea). For the additional current-collector control experiment shown in Supplementary Fig. 4 and Supplementary Video 4, a slurry containing Te, Super P carbon and PVDF at an 8:1:1 mass ratio was dispersed in NMP and cast onto Al foil (99.99%, Welcos; thickness, 20 µm) using a doctor blade (SeKTQ150, MTI Korea). The coated electrodes were dried under vacuum at 60 °C for 8 h, with a Te mass loading of ~3 mg cm⁻² and punched into 11-mm-diameter disks using an electrode punch (P02, ROTECH, Korea). In an argon-filled glovebox (O₂, H₂O < 1 ppm), we assembled a CR2032-type custom coin cell. The Te positive electrode was used as the working electrode, and a 15-mm-diameter Li metal foil (99.95%, Honjo; thickness, 350 µm) was used as the counter/reference electrode. The Li metal foil was punched using a 15-mm-diameter electrode punch (P02, ROTECH, Korea). A glass fiber membrane (Whatman GF/B; catalog number: 1821-110; punched diameter: 19 mm; thickness: 0.675–0.68 mm; nominal pore size: 1.0 µm; estimated porosity: ~90%) served as the separator. The electrolyte was a 160 µL solution of 1.0 M LiTFSI in a 1:1 volume ratio mixture of DOL and DME. Operando TXM analysis was performed at beamline 6C for operando transmission X-ray imaging and beamline 7C for high-resolution operando transmission X-ray imaging of the Pohang Accelerator Laboratory (PAL), using an incident X-ray energy of 10 keV. This energy level allowed the X-rays to penetrate through the entire cell, including the positive electrode, Li negative electrode, carbon paper or Al-foil current collector, and electrolyte. TXM images were captured every 2 min to ensure accurate representation of the electrode morphology throughout the cycling process. The images were acquired with a Pco. Edge 5.5 sCMOS camera. Galvanostatic charge/discharge measurements were performed using a potentiostat (ZIVE SP1, WonA Tech, Korea) in the experimental hutch at the Pohang Accelerator Laboratory without an environmental chamber. The measurements were conducted under ambient room-temperature conditions, nominally at 25 °C, with the cells cycled between 1.3 and 2.7 V at a rate of 0.2 C (84 mA g^–1_Te) or 0.5 C (210 mA g^–1_Te).

Operando optical imaging measurements

We designed a custom coin cell with a central hole and sealed it with a glass window. The electrode slurry was prepared with Te (99.8%, Sigma Aldrich), Super P carbon (estimated carbon content ≥99.5 wt%, average particle size: 400 nm, Timcal), and polyvinylidene fluoride (PVDF) (Solef 5130, Mw:1,000,000-1,100,000 g mol^–1, PVDF content > 99.9%, Solvay) in an 8:1:1 weight ratio. The mixtures were separately dispersed in N-methyl-2-pyrrolidone (NMP; HPLC grade, ≥99%, Sigma-Aldrich) to form homogeneous slurries and cast onto Al foil (99.99%, Welcos, 20 µm) using a doctor blade (SeKTQ150, MTI Korea). The coated electrodes were dried under vacuum at 60 °C for 8 h. The Te mass loadings were ~3 mg cm⁻². The dried electrodes were punched into 11-mm-diameter disks using an electrode punch (P02, ROTECH, Korea). In an argon-filled glovebox (O₂, H₂O < 1 ppm), we assembled a CR2032-type custom coin cell. The Te positive electrode was used as the working electrode, and a 15-mm-diameter Li metal foil (99.95%, Honjo; thickness, 350 µm) was used as the counter/reference electrode. The Li metal foil was punched using a 15-mm-diameter electrode punch (P02, ROTECH, Korea). A glass fiber membrane (Whatman GF/B; catalog number: 1821-110; punched diameter: 19 mm; thickness: 0.675–0.68 mm; nominal pore size: 1.0 µm; estimated porosity: ~90%) served as the separator. The electrolyte was a 160 µL solution of 1.0 M LiTFSI in a 1:1 volume ratio mixture of DOL and DME. The optical observation was visualized using an optical microscopy (×50 lens, ECLIPSE LV150N, Nikon Corp). Galvanostatic charge/discharge measurements were performed using a potentiostat (ZIVE SP1, WonA Tech, Korea) without an environmental chamber, under ambient laboratory conditions nominally at 25 °C, with the cells cycled between 1.3 and 2.7 V at a rate of 0.2 C (84 mA g^–1_Te).

Ex-situ physicochemical measurements

For the SEM and TEM measurements, we assembled a CR2032 coin cell with a Te positive electrode as the working electrode and Li metal disks as both the counter and reference electrodes in a glovebox (O₂, H₂O < 1 ppm) under an argon (Ar) atmosphere. The electrode slurry was prepared with Te (99.8%, Sigma Aldrich), Super P carbon (estimated carbon content ≥99.5 wt%, average particle size: 400 nm, Timcal), and polyvinylidene fluoride (PVDF) (Solef 5130, Mw:1,000,000-1,100,000 g mol^–1, PVDF content > 99.9%, Solvay) in an 8:1:1 weight ratio. The mixtures were separately dispersed in N-methyl-2-pyrrolidone (NMP; HPLC grade, ≥99%, Sigma-Aldrich) to form homogeneous slurries and cast onto AvCarb P50 carbon paper (Ballard, thickness: 184 µm, density: 0.40 g cm^–3, compressibility: 14.0%, mean pore size: 10 µm) using a doctor blade (SeKTQ150, MTI Korea). The coated electrodes were dried under vacuum at 60 °C for 8 h. The Te mass loadings were ~3 mg cm⁻². The dried electrodes were punched into 11-mm-diameter disks using an electrode punch (P02, ROTECH, Korea). The Te positive electrode was used as the working electrode, and a 15-mm-diameter Li metal foil (99.95%, Honjo; thickness, 350 µm) was used as the counter/reference electrode. The Li metal foil was punched using a 15-mm-diameter electrode punch (P02, ROTECH, Korea). A glass fiber membrane (Whatman GF/B; catalog number: 1821-110; punched diameter: 19 mm; thickness: 0.675–0.68 mm; nominal pore size: 1.0 µm; estimated porosity: ~90%) was used as the separator. The electrolyte comprised 160 µL of a 1.0 M LiTFSI solution in a 1:1 volume ratio of DOL and DME. Galvanostatic charge/discharge measurements were performed using a battery cycler (WBCS3000, WonA Tech, Korea), with the cells placed in a constant-temperature chamber (HG-I150, Hangil Biotech, Korea) maintained at 25.0 ± 0.2 °C. The cells were cycled between 1.3 and 2.7 V at 0.2 C (84 mA g^–1_Te). After cycling, the coin cells were disassembled in a glovebox (O₂, H₂O < 1 ppm) under an argon (Ar) atmosphere to retrieve the positive electrodes. The positive electrodes were then washed multiple times with DME to remove residuals. For ex-situ SEM measurements, we used a vacuum transfer holder in the glovebox (O₂, H₂O < 1 ppm) to prevent exposure to air. Top-view scanning electron microscope (FE-SEM, Regulus8230) images and corresponding EDS elemental maps were acquired at 15 kV. TEM measurements were carried out to verify the crystallographic orientation and crystal structure using a JEOL JEM-2100F field-emission TEM operated at 200kV.

Electrochemical measurements

The S and Te positive-electrode slurries were composed of sulfur (−324 mesh, 99.5%, Alfa Aesar) or Te (99.8%, Sigma Aldrich), Super P carbon (estimated carbon content ≥99.5 wt%; average particle size, 400 nm; Timcal), and polyvinylidene fluoride (PVDF; Solef 5130, M_w = 1,000,000^–1,100,000 g mol⁻¹, PVDF content >99.9%, Solvay) at mass ratios of 6:3:1 and 8:1:1, respectively. The mixtures were separately dispersed in N-methyl-2-pyrrolidone (NMP; HPLC grade, ≥99%, Sigma-Aldrich) to form homogeneous slurries and cast onto AvCarb P50 carbon paper (Ballard, thickness: 184 µm, density: 0.40 g cm^–3, compressibility: 14.0%, mean pore size: 10 µm) using a doctor blade (SeKTQ150, MTI Korea). The coated electrodes were dried under vacuum at 60 °C for 8 h. The S and Te mass loadings were ~2 and 3 mg cm⁻², respectively. The dried electrodes were punched into 11-mm-diameter disks using an electrode punch (P02, ROTECH, Korea). For the electrochemical performance tests, conventional CR2032-type coin cells without a central window were assembled in an Ar-filled glovebox (O₂ and H₂O < 1 ppm). The S or Te positive electrode was used as the working electrode, and a 15-mm-diameter Li metal foil (99.95%, Honjo; thickness, 350 µm) was used as the counter/reference electrode. The Li metal foil was punched using a 15-mm-diameter electrode punch (P02, ROTECH, Korea). Li-S cells used a polypropylene separator (Celgard 2400; punched diameter, 19 mm; thickness, 25 µm; porosity, 41%; nominal pore size, 0.043 µm), whereas Li-Te cells used a glass-fiber separator (Whatman GF/B; catalog number, 1821-110; punched diameter, 19 mm; thickness, 0.675–0.68 mm; nominal pore size, 1.0 µm; estimated porosity, ~90%). Each Li–S cell contained 60 µl of an electrolyte comprising 1.0 M LiTFSI and 0.2 M LiNO₃ (99.99%, Sigma-Aldrich) dissolved in a 1:1 volume ratio mixture of DOL and DME, whereas each Li-Te cell contained 160 µl of 1.0 M LiTFSI dissolved in the same solvent mixture. Galvanostatic charge/discharge measurements were performed using a battery cycler (WBCS3000, WonA Tech, Korea), with the cells placed in a constant-temperature chamber (HG-I150, Hangil Biotech, Korea) maintained at 25.0 ± 0.2 °C. Li-S and Li-Te cells were cycled over voltage ranges of 1.7–2.8 V and 1.3–2.7 V, respectively. The first-cycle voltage profiles were recorded at 0.5 C, corresponding to specific currents of 837.5 mA g⁻¹ based on the mass of S and 210 mA g⁻¹ based on the mass of Te for the Li-S and Li-Te cells, respectively. Cycling-performance tests were conducted at 1 C, corresponding to specific currents of 1675 mA g⁻¹ based on the mass of S and 420 mA g⁻¹ based on the mass of Te, respectively. For the tailored charging tests, selected Li-S and Li-Te cells were initially charged at 1 C to 2.5 and 1.9 V, respectively, and subsequently charged at 0.1 C to their respective upper cut-off voltages of 2.8 and 2.7 V. The corresponding specific currents at 0.1 C were 167.5 mA g⁻¹ based on the mass of S and 42 mA g⁻¹ based on the mass of Te, respectively.

Computational calculation

All ab initio calculations were performed with the Vienna Ab initio Simulation Package (VASP 5.4.4).^43–46 We used the projector augmented wave (PAW) method^47,48 with the generalized gradient approximation based on the Perdew-Burke-Ernzerhof (PBE)⁴⁹ functional. A plane-wave cut-off energy of 500 eV was used. Lattice constants and internal atomic positions were fully optimized until the residual forces were less than 0.02 eV/ Å. The vacuum slab space of a unit cell in the z-direction was set to 15 Å to avoid interactions between layers. Spin polarization and dipole correction were also included to decouple the electrostatic interaction between periodically repeated surface systems.

For the DFT model, Li₂Te_x species (x = 1 ~ 8) were constructed on an idealized graphene sheet with limited periodicity to reduce computational cost and enable consistent evaluation of relative kinetic trends for subsequent kMC simulations. The optimized CIF structures of the graphene-supported Li₂Te_x intermediates used in the DFT calculations are provided in Supplementary Data 1. The surrounding electrolyte within the interspace region was described by an implicit solvent using the polarizable continuum model (PCM) as implemented in VASPsol.⁵⁰ The dielectric constant of a 50:50 DOL/DME mixture (${ϵ}_b$ = 7.101) was used to model the reaction environment for Li₂Te_x interconversion under charging and discharging processes.

The kinetic Monte Carlo (kMC)^51,52 simulation was designed to model the dynamic evolution of Li-Te intermediates during battery operation by incorporating 16 elementary reactions that capture both reduction and oxidation processes. For each reaction, the rate constant is calculated using the Arrhenius equation, $k_i=A_i\exp \left(-E_{a,i}/k_{B}T\right)$ where the activation energies $\left(E_a\right)$ are derived from our DFT calculations and the pre-exponential factors $\left(A_i\right)$ are estimated for both the forward (reduction) and reverse (oxidation) directions. The simulation tracks species concentrations on a two-dimensional lattice representing the graphene substrate, with reaction events executed stochastically via Gillespie’s algorithm.⁵³ At every simulation step, reaction rates are computed based on the current species concentrations and the corresponding kinetic parameters, and the overall reaction propensity is determined by summing these individual rates. Two uniformly distributed random numbers are then employed: one to calculate the time increment until the next event, and the other to select which reaction occurs, thereby updating the species populations. This iterative process continues until the total simulation time is reached, enabling the capture of time-dependent phenomena for solid–liquid phase transitions and particle growth. Ultimately, this stochastic approach reveals the interplay between reaction kinetics and electrochemical conditions, providing critical insights into the mechanisms underlying the observed differences in Li₂Te₆ particle sizes during charging versus discharging.

Figure preparation and image processing

TXM images were processed using MATLAB (MathWorks). The image intensities were normalized to a range of 0–1, and the corresponding color maps were applied using MATLAB. All schematic illustrations were drawn from scratch by the authors using Adobe Illustrator (Adobe Inc.) and did not incorporate any third-party materials, stock images, online illustration-platform assets or generative artificial intelligence. Original experimental images acquired by the authors were imported into Adobe Illustrator only for cropping, panel assembly, labeling and the addition of graphical annotations, including boxes and arrows. No image features were added, removed or selectively altered.

Author contributions

H.-Y Choi and M. Kim (Mihyun Kim) conducted the writing the draft and experiments for battery electrochemistry and operando XRD and operando TXM. C. H. Lee and S. Nagpal performed DFT and kMC simulations to interpret the reaction kinetics and drafted the theoretical section. M. Kim (Minki Kim), J.-H. Lim, J. Lim, S. H. Kim and S.-Y. Lee assisted with the TEM, SEM and TXM measurements. J. S. Kwon is a co-principal investigator and supervised the theoretical work and contributed to the writing. S.-H. Yu. is the chief principal investigator of the project and supervised whole process of experiments and writing. All authors discussed the results and commented on the manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous, reviewers for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea Government (MSIT) (No. RS-2024-00422387 and RS-2025-00518953).

Data availability

The data supporting the findings of this study are available within the paper, its Supplementary Information and Source Data files. Additional raw imaging and video data are available from the corresponding authors upon request. Source data are provided with this paper.

Competing interests

The authors declare no competing interest.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-76636-8.