An amorphous Li–V–O–F cathode with tetrahedral coordination and O–O formal redox at low voltage

Abstract

The ever-increasing demand for lithium-ion batteries has necessitated the development of high-performance cathode materials. However, previous studies have predominantly focused on crystal cathodes comprising the octahedral coordination of metal atoms and a well-ordered layered topology. This omits other cathode materials with other structures or coordination that could potentially surpass conventional counterparts in terms of performance. Here, using X-ray diffraction, resonant inelastic X-ray scattering and X-ray absorption near-edge spectra experiments, we investigated an amorphous Li–V–O–F cathode (a-LVOF) with tetrahedral coordination and elucidated an O–O formal redox mechanism at a moderate voltage of 4.1 V, without a conventional octahedral Li–O–Li configuration. The electrochemically amorphized material fosters randomly distributed VO₄ units and scattered dangling oxygen bonds, which facilitated O–O binding. Moreover, a-LVOF demonstrates a high capacity of 230 mAh g⁻¹. Our findings reveal a low-voltage O–O formal redox mechanism in an amorphized cathode material.

Cathode materials for lithium-ion batteries typically possess octahedral coordination, which may exclude other possible solutions to degradation during deep cycling. A series of tetrahedral-framework-based amorphous Li–V–O–F materials are investigated, and shown to demonstrate O–O formal redox at 4.1 V.

Main

The advancement of performance in secondary high-energy-density lithium-ion batteries, where cathode materials play a crucial role in determining the cost and energy density, is urgently required to meet the demand for electronic devices, grid energy storage and electric vehicles¹. The invention and commercialization of LiCoO₂, LiMn₂O₄ and LiFePO₄ have promoted substantial systematic and in-depth research on crystalline cathode materials worldwide^2–4. Among various crystalline compositions, octahedral-based transition metal (TM) oxides, where six oxygen ions are coordinated to the TM ions, have gained the most attention because their specific capacity can be increased via component optimization¹. In particular, researchers have found that the oxygen redox (OR) mechanism in these materials can trigger higher energy density, as reported in Li-rich layered cathodes (LLOs) with high capacities over 250 mAh g⁻¹ and the widely studied Li[Ni_1−x−yCo_xAl_y]O₂ and Li[Ni_1−x−yCo_xMn_y]O₂ cathodes^5–8. However, during deep charging and discharging, lattice strain deteriorates cathode materials and causes oxygen release that destabilizes the material^9–11. To mitigate these disadvantages, researchers have investigated structural engineering involving surface modification and bulk doping, but fundamental solutions to these problems have not been found so far^12,13. Alternative pathways for cathode development should also be pursued.

Over the past few decades, the investigation of diverse crystal structures has opened avenues for research in cathode materials, playing a substantial role in improving the performance. For example, a honeycomb superstructure can enhance the specific capacity of cathode materials¹⁴, whereas the ribbon microstructure can improve structural stability^15,16. A series of crystalline materials with disordered Li and TM atomic arrangements that can achieve higher capacities than those of crystalline layered oxides have been proposed^17,18. In addition to these novel structures, a large class of amorphous materials have been investigated in cathodes, anodes and electrolytes with high battery performance^19–21. Although the functioning mechanism of amorphous cathodes is underinvestigated, an amorphous iron fluorosulfate with a combined intercalation and conversion reaction has been proposed, indicating that amorphous cathodes can possess high performance with a unique charge compensation mechanism²².

In this study, we investigated a tetrahedral-framework-based amorphized Li–V–O–F cathode, denoted as a-LVOF cathode. By studying the structural evolution and charge compensation mechanism, we reveal an OR behaviour in O–O formation at low voltage, which originates from its amorphized nature.

Formation of amorphous structure

The a-LVOF cathode materials were obtained by in situ electrochemical Li insertion. First, crystalline LiVO_3–xF_x (0 ≤ x ≤ 0.025) cathode materials were synthesized, quantified and named F0, F1, F2 and F3 (Supplementary Table 1). The long-scan X-ray diffraction (XRD) patterns (Supplementary Fig. 1) indicate that all LiVO_3–xF_x cathodes are composed of the C2/c lattice structure²³. Except for the lithium fluoride (LiF) residue and the resulting LiV₃O₈ phase in F3, all LiVO_3–xF_x samples contain pure crystal phases. F0, F1 and F2 exhibited very similar galvanostatic curves during the first cycle, implying the same mechanism in these cathodes, whereas F3 showed a discharge profile with LiV₃O₈ impurity²⁴. In particular, F2 exhibited the highest initial capacity (Supplementary Fig. 2), because fluorination can facilitate ion diffusion^25,26 (Supplementary Note 1 and Supplementary Fig. 3). Moreover, the F2 sample has a well-established single-phase structure (Supplementary Note 2, Supplementary Figs. 4–8 and Supplementary Tables 2 and 3). The structural evolution and redox mechanism of F2 were investigated.

Figure 1a and Supplementary Fig. 9 show the first and second galvanostatic charge–discharge and dQ/dV profiles of F2 at a current density of 10 mA g⁻¹ in 2.0–4.8 V. Under direct charging (CH1), the F2 cathode exhibits a capacity of ~40 mAh g⁻¹ with an average voltage of 4.55 V. Two discharge stages at 4.8–2.4 V (DH1) and 2.4 V (DH2) are observed during the discharge process. At DH2, a 2.2-V plateau can be observed, indicating additional Li insertion with a phase change. This overlithiation can lead to Fm$\overline{3}$m rock-salt phase formation and simultaneous structural amorphization according to previous reports^23,27. Since only the DH2 reaction plateau is observed on direct Li intercalation, DH1 should stem from the first charging process (Supplementary Fig. 10). However, during the second cycle, the curves are sloped, indicative of amorphized cathode materials. The second charging stages were divided into CH2 (2.0–3.5 V with the oxidation peak at 3.25 V) and CH3 (3.5–4.8 V). In CH3, a charging plateau is observed at 4.1 V, which is maintained during the subsequent cycles (Supplementary Fig. 11). Moreover, the reduction peak at the DH3 stage was observed at 3.34 V, which is higher than the oxidation peak in CH2, indicating that the species charged in CH3 can be reversibly reduced in the DH3 process. These intriguing electrochemical properties of F2 with a mild reaction at ~4.1 V inspire the investigation of its charge compensation mechanism.

Fig. 1. Electrochemical and structural evolution in a-LVOF cathodes.

a, Charge–discharge profiles of the first two cycles. CH, charge stage; DH, discharge stage. b, Ex situ XRD spectra of the cathodes at different states of charge. C, charge stage; D, discharge stage. c, PDF analysis of the 1C 4.8 V and 2C 4.8 V samples. The peaks at approximately 1.5–2.0 Å are enlarged for observation (right). G(r) is the reduced radial distribution function, representing pair distribution. The vertical dashed lines reflect the peak positions of pair distribution in 1.25–2.25 Å. d, R-space spectra for different states of charge. χ(R) is the result of transforming χ(k), the oscillatory function in wavevector k space into real space R. The orange shading and vertical dashed line reflect the range of V–O(Td) pairs. V–O(Oh), V–O pairs under octahedral coordination; V–O(Td), V–O pairs under tetrahedral coordination.

Source data

Ex situ XRD analyses were conducted to elucidate the lattice structural changes during the first two cycles (Fig. 1b). The unchanged diffraction peaks of F2 during the CH1 and DH1 processes suggest the maintenance of a pristine crystal structure. However, the entire XRD pattern of the C2/c phase, including the main peak at 18.6°, disappears after the DH2 stage, indicating the apparent structural amorphization of the C2/c phase. After the entire structure undergoes amorphization, a very weak peak can be observed at 45° at this stage, indicating the generation of a minor Fm$\overline{3}$m rock-salt phase. This nanocrystalline phase weakens at 2D 2.0 V and nearly disappears at 10D 2.0 V, indicating structural irreversibility (Supplementary Fig. 12). The high-resolution transmission electron microscopy (HR-TEM) images also show the same trend as an activation process (Supplementary Fig. 13), with non-crystalline local structures observed at 1D 2.0 V and 2D 2.0 V (Supplementary Fig. 14). Thus, the overlithiated structure in the initial cycle is a compound of amorphized/rock-salt nanodomains.

However, at the end of CH3, no diffraction peaks are observed in the XRD pattern, preliminarily implying structural amorphization. High-energy synchrotron XRD (SXRD) was further conducted to confirm the precise structure, with an extremely low-wavelength X-ray source to increase the angular resolution (Supplementary Fig. 15). Compared with the 1C 4.8 V sample that shows clear featured peaks of the C2/c lattice phase, a very broad pattern can be detected for the 2C 4.8 V sample, implying that the structure has undergone drastic structural amorphization. Although some weak peaks correspond to the regenerated C2/c lattice phase, the main lattice has disappeared, indicating the highly amorphized structure. HR-TEM was further conducted to detect the local structure in the 2C 4.8 V sample (Supplementary Fig. 16a–f). Despite the dispersion of minor nanoparticles smaller than 10 nm in the bulk (Supplementary Fig. 16a–c), most regions in the bulk are completely amorphous. On the basis of the SXRD and HR-TEM results, the dominant bulk structure of the 2C 4.8 V sample is amorphous.

Extended X-ray absorption fine structure is helpful for detecting the short-range coordination in amorphized structures. The corresponding R-space spectra of the cycled F2 cathode were obtained (Fig. 1c). The peaks at 1.2 Å correspond to the V–O interactions in the VO₄ tetrahedral coordination. This peak remains constant during the first cycle, indicating maintenance of the local tetrahedral coordination. At 1D 2.0 V, a new peak appears at 1.68 Å, representing the emergence of octahedral VO₆ units. This transformation from VO₄ to VO₆ units further proceeds at 1D 1.5 V (Supplementary Fig. 17). The coexistence of VO₄ and VO₆ coordinations is observed in these discharged states. According to the HR-TEM results (Supplementary Fig. 13), the nanocrystal features an octahedral-based rock-salt phase, without a tetrahedral-composed crystalline structure. Therefore, tetrahedral-based components should contribute the amorphous part. During the CH2 and CH3 processes, the local structure completely converts into VO₄ coordination.

The pair distribution function (PDF) was used to analyse the various coordinating pairs in the 1C 4.8 V and 2C 4.8 V samples (Fig. 1d). The function G(r) provides information on bonding pairs, bond strengths and structural ordering²⁸. The G(r) profile of the 1C 4.8 V sample exhibits an intensively ordered coordinating pattern, indicating a well-crystallized structure; however, the 2C 4.8 V sample has much weaker and broadened peaks that indicate structural long-range disordering²⁹. Interestingly, although the peak at 1.73 Å denotes the V–O bonds in tetrahedral units, the peak shifts to a smaller distance at the end of the 2C 4.8 V stage, implying that bonding pairs at distances of 1.3–1.5 Å are newly generated. The charge compensation mechanism of these two stages was subsequently investigated.

Low-voltage O–O formal redox in a-LVOF cathode

Normalized O K-edge and V L-edge X-ray absorption near-edge structure spectra (XANES) were obtained to observe the evolution of the TM–O hybridized electronic states (Fig. 2b). The absorption peaks at 518.1 eV and 525.0 eV represent the V L₃- and L₂-edges³⁰, according to the crystal field theory³¹. The peaks at 529.5 eV and 531.3 eV represent the TMe–O2p/TMt₂–O2p hybridized states in tetrahedral coordination, respectively (Supplementary Fig. 18). In LVOF with almost empty orbitals (because of V⁵⁺ majority), the oxidation/reduction of ions changes the state of the lowest empty e orbitals. Therefore, the ratios of fitted areas 1 and 2 corresponding to peaks 1 and 2, respectively, were calculated, to examine the change in electronic states (Supplementary Fig. 19 and Fig. 2c).

Fig. 2. Charge compensation mechanism.

a, Proceeding states of charge. (i)–(vii) denote pristine, 1C 4.4 V, 1C 4.8 V, 1D 2.4 V, 1D 2.0 V, 2C 3.5 V and 2C 4.8 V states, respectively. The corresponding charge–discharge stages are listed as side labels for clarity. b, Normalized O K-edge XANES data. Areas 1 and 2 represent the TMe–O2p/TMt₂–O2p hybridized states. V L₃ is the indicator of V L₃ absorption peak, and L₂ represents the L₂ peak. The black arrows indicate the position change of V L₂-edge and /TMt₂–O2p absorption peaks. c, Calculated ratios of areas 1 and 2 during the electrochemical process. d, Positions of V L-edge during cycles.

Source data

During the CH1 stage (Fig. 2a,b), a successive V reduction is observed with a low-energy shift in the V L-edges (Fig. 2d). This phenomenon also emerges in the V X-ray absorption spectroscopy (XAS) results (Supplementary Fig. 20). When V ions are reduced, the ratios (Fig. 2c) show an increase at (ii)–(iii), indicating that O oxidation exclusively creates more empty electronic states. Owing to the redox-coupling mechanism, the O oxidation can lead to the reduction of high-valence TM ions via electron back-donation, increasing the stability of the structure^32,33. O1s X-ray photoelectron spectroscopy (XPS) fitted results (Supplementary Fig. 21) further confirm the emergence of high-valence O^(2–n)– species in the CH1 and CH3 stages. The XPS results also exhibit the formation of cathode electrolyte interphase species with some carbonates at the surface in the first charge, without evolution in the second charge^34,35. Although the cathode electrolyte interphase formation can also cause surface V reduction on the first charge, the V reduction in the bulk should originate from the redox-coupling mechanism. The combined evolution of XANES and XPS data elucidates the participation of O in the CH1 and CH3 stages.

Resonance inelastic X-ray scattering (RIXS) was performed to further probe the delicate chemical and electronic states of bulk O in the CH1 and CH3 stages^36,37. Figure 3a–c and Supplementary Fig. 22 present the O K-edge RIXS two-dimensional maps as functions of the emission energy (x axis) and excitation energy (y axis). All the RIXS maps display an emission feature at approximately 524.6 eV, corresponding to the TM–O hybridization states, alongside an elastic peak denoted by a white line. In particular, during the first cycle, no oxidized O feature (Supplementary Fig. 22a, dotted circle) is observed at 1C 4.8 V and the pristine state (Fig. 3a) and not at 1D 2.0 V (Supplementary Fig. 22b)³⁸. This result suggests that during the first cycle, the O reaction predominantly forms electron holes without generating O–O bonding. However, on subsequent charging to 4.8 V, the O K-edge mapping of RIXS spectra exhibit O–O-bonding characteristics, with excitation energies near 529.8–532.1 eV and emission energy of 523.7 eV (Fig. 3b,d, black arrows), which is consistent with previous work³⁷. This fingerprint of O–O bonding disappears after discharging to 2.0 V (Fig. 3c), indicating the reversible redox activity of oxygen in the second cycle. To further identify the oxidized oxygen state, both PDF and Raman spectroscopy were used. The PDF results reveal O–O bonding at approximately 1.43 Å (Fig. 1d), which is consistent with the O–O dimer range and longer than the O–O bond in O₂ molecules (<1.3 Å). Raman spectroscopy (Supplementary Fig. 23) further confirms the characteristics of the O–O peroxo-species^39,40, with a clear peak at 847 cm⁻¹. Therefore, the OR mechanism is realized by generating pure electron holes in the crystal structure and the formation of O–O dimers in the amorphized electrode.

Fig. 3. O K-edge RIXS map collected on electrodes with different electrochemical states.

a, Pristine (PRI) state. b, At the second charging state (2C 4.8 V). The arrows indicate the fingerprint feature of the oxidized oxygen state. c, At the second discharging state (2D 2.0 V). d, RIXS cut for an excitation energy of 530.8 eV on each sample.

Source data

Ab initio molecular dynamics (AIMD) calculations were performed to simulate bond formation in the first and second charges. To match the experimental results, for delithiated a-Li₄₂V₄₈O₁₄₁F₃, a1 and a2 were started from overlithiated a-LVOF with tetrahedral structure and crystalline rock-salt structures, respectively. Delithiated c-Li₄₂V₄₈O₁₄₁F₃ was obtained from c-LVOF with the C2/c phase. The calculation details are shown in Supplementary Note 3 and Supplementary Figs. 24 and 25. The entire structural evolution is given in Supplementary Videos 1–3. The rationality of the AIMD calculations is shown in Supplementary Note 4. After being heated to 1,000 K and cooled to 300 K, obvious spatial disordering is observed in both a1 and a2 structures, with randomly oriented VO₄ tetrahedral units (Fig. 4a,b). The G(r) values of a1 and a2 over 5 Å demonstrate a uniform distribution of the chemical bonds without apparent fluctuations, indicating an amorphized state with the absence of long-range ordering (Supplementary Fig. 26). Although a2 from the crystalline rock-salt structure can reach the amorphized state, delithiated c-LVOF still maintains its crystallinity under the same heating–quenching process (Fig. 4c). This result means that amorphization from overlithiated structures is thermodynamically more favourable than that directly from c-LVOF, which is consistent with the experimental structures at 2C 4.8 V (amorphized) and 1C 4.8 V (crystalline).

Fig. 4. Resulting simulated delithiated structures under AIMD.

a, a1 structure. b, a2 structure. c, Delithiated c-LVOF. d–f, Corresponding simulated O PDFs for a1 (d), a2 (e) and delithiated c-LVOF (f) structures. The arrows in d and e indicate the O–O dimers in the a1 and a2 structures, respectively. g, Schematic of the oxidized O in c-LVOF. h, Schematic of O–O dimerization in a1 and a2 structures.

Source data

In particular, an O–O distance of 1.4–1.6 Å appears in both a1 and a2 structures, which also emerges in the O–O G(r) profiles, suggesting that O–O dimerization occurs in delithiated a-LVOF (Fig. 4d,e). Comparatively, no such short O–O interaction is found in the delithiated c-LVOF structure (Fig. 4f). This simulated emergence of the O–O dimers is consistent with the experimental PDF results, whereas more O–O dimers can be found in the actual structure, indicating that structural amorphization can trigger an OR mechanism for O–O dimerization (Supplementary Fig. 27).

For the O ions that lose electrons, the O–O dimer formation can decrease the systematic energy⁴¹; therefore, these dimers are more stable species that can facilitate the OR process at lower voltages⁴². In c-LVOF, the entire structure and basic units undergo limited variations under AIMD simulation; the O atoms are restricted to the crystalline skeleton and cannot form more stable O–O dimers (Fig. 4g and Supplementary Fig. 28). Therefore, c-LVOF has much greater difficulty overcoming the energy barrier to form O–O dimers. In comparison, the a1 and a2 structures undergo drastic changes during the heating process (Supplementary Videos 1–3), passing through various thermodynamic states and ultimately forming more stable O–O dimers. The mechanism is illustrated in Supplementary Fig. 29. Moreover, two different surrounding environments are found in the a1 and a2 structures (Fig. 4h), which originate from the structural diversity of amorphized structures. Overall, the formation of O–O dimers in the AIMD simulation is in good accordance with the experimental results.

Electrochemical performance of a-LVOF cathode

Amorphized structures can facilitate Li transport by creating considerable nanoscale diffusion channels and inducing pseudocapacitance among many TM oxides²². Hence, cyclic voltammetry (CV) tests were performed at different scan rates to determine the Li diffusion kinetics in a-LVOF cathodes. Figure 5a shows the CV scans of a-LVOF at various rates in the 0.1–0.5 mV s⁻¹ range after the first cycle. The oxidation/reduction peaks A/A′ denote the origin of the main capacity from the V redox process. The relationship between the peak current i and the scan rate v in the CV experiment is expressed as follows:

$$i=a{v}^b,$$ 1

where b indicates whether the reaction involves surface capacitive behaviour (b ≈ 1) or bulk insertion diffusion (b ≈ 0.5)⁴³. The fitted slopes of A and A′, denoted by b, in the ln[i]/ln[v] plots are 1.04 and 0.86, respectively, indicating major capacitive behaviour (Fig. 5b). The capacitive and diffusion contributions to the capacity were further resolved using the following equation:

$$i=k_{1}v+k_2{v}^{1/2}.$$ 2

Fig. 5. Electrochemical performance of the LVOF cathode.

a, CV profiles at gradient scan rates. The oxidation/reduction peaks A/A′ denote the origin of the main capacity from the V redox process. b, Fitted lines and slopes of electrochemical A and A′ processes based on the CV tests. The value of b represents the slopes. c, Cycling performance of the cathode under 200 mA g⁻¹ at 1.5–4.8 V and 2.0–4.8 V. d, DEMS results of the first two cycles. log[P (torr)] represents the log function of the gas partial pressure. The line breaks on the right and left y axis are from −11.5 to −10.725. The red dashed line reflects the base of CO₂ gas. e, Average voltage during 100 cycles.

Source data

The fitted capacitance/capacity ratio is shown in Supplementary Fig. 30, in which the capacitive contribution is ~71.7% at 0.2 mV s⁻¹ (Supplementary Fig. 31). The rate capability of a-LVOF is also provided in Supplementary Fig. 32, where 133 mAh g⁻¹ can still be reached at 1,000 mA g⁻¹. Therefore, facile Li transportation at high rates has been shown in a-LVOF.

The a-LVOF cathode was subsequently cycled at 200 mA g⁻¹ for 100 cycles (Fig. 5c). After the activation process, a capacity of ~230 mAh g⁻¹ and stable cycling are observed at 2.0–4.8 V. The activation process can be attributed to the incomplete particle refinement during the subsequent 2.0-V cycles (Supplementary Fig. 33). On further Li intercalation to 1.5 V, over 300 mAh g⁻¹ and ~80% retention are observed after 100 cycles, which is comparable to the capacities of current cathode materials. The a-LVOF can also exhibit stable long cycling at 4,000 mA g⁻¹ (Supplementary Fig. 34) and a cycling capability in full-cell testing with graphite as the anode (Supplementary Fig. 35). Differential electrochemical mass spectrometry (DEMS) was used to systematically evaluate the stability by testing the gas production (Fig. 5d). No O₂ release is observed during the CH1 and CH3 stages, indicating that the OR mechanism in the LVOF cathode is relatively stable compared with that in LLOs¹⁰. This result also echoes the RIXS, PDF, Raman and AIMD consequences. Minor CO₂ generation (~0.22 nmol) was detected at high voltages, which can be attributed to electrolyte decomposition. Although oxygen release and irreversible TM migration can lead to severe voltage decay in LLOs, the absence of these two factors in the a-LVOF results in a stable voltage during cycling (Fig. 5e). Hence, a-LVOF can achieve a high capacity with a stable OR mechanism.

Correlation of O redox stability and amorphized structures

In delithiated c-LVOF, the fixed structural distribution of VO₄ tetrahedra and O lone pairs increase the difficulty of forming spatial O–O dimers; thus, electrons are lost from the O holes (Fig. 4g). However, the O holes are regarded as high-energy intermediates based on previous works, leading to a high-voltage reaction in c-LVOF^44,45. Instead, the oxidized O ions are prone to form O–O dimers with systematic energy decrease⁴¹, which are more stable species and can undergo the OR process at lower voltages⁴². Therefore, a-LVOF cathode with O–O dimers can demonstrate more stability than the c-LVOF cathode.

Although conventional OR active cathodes, for example, LLOs, can also generate O–O dimers, the layered structural degradation due to TM migration can cause sequential O₂ molecule generation, aggravating instability of the entire structure. On the basis of the previous report⁴⁶, O₂ molecule formation can occur once at 900-K AIMD in Li-rich Mn-based layered oxides with structural degradation. In comparison, even under 1,000-K simulation, no O₂ molecules are observed in the a-LVOF models, as well as in the PDF, Raman and DEMS experiments, demonstrating the intrinsic OR stability of a-LVOF. Therefore, amorphized structures prevent lattice structural degradation, making them advantageous over crystal cathodes.

Overall, the OR stability is based on two main factors: (1) the intrinsic thermodynamic stability of O–O dimers with a lower redox voltage in a-LVOF than O hole generation in c-LVOF and (2) the structural maintenance in a-LVOF without sequential structural degradation and O₂ generation compared with LLOs.

Discussion

In this study, an O–O formal redox was revealed in an amorphous structure at a mild voltage. This finding implies that the OR mechanism can be triggered in amorphous cathodes, showing more structural stability over conventional cathodes and high capacity. These findings can be extended to and further investigated in other inexpensive/low-toxic oxide cathode materials, including Mn-based oxides and a wide class of disordered rock-salt cathodes. This study highlights the substantial importance of amorphous structures as future cathode materials.

Methods

Synthesis of LiVO_3–xF_x cathode materials and F quantification

Synthesis

The designed LiVO_3–xF_x (x = 0, 0.02, 0.04, 0.06) cathode materials, named F0, F1, F2 and F3, respectively, were synthesized by high-temperature sintering and high-energy ball-milling processes. First, stoichiometric amounts of LiOH·H₂O (AR, >99.0%), NH₄VO₃ (Macklin, >99.9%) and LiF (AR, >99.0%) are mixed in deionized water. The solution was stirred at 80 °C until it became completely dry. The mixture was then ground and sintered at 450 °C for 12 h under an O₂ atmosphere. Subsequently, the precursor was mixed with Super P and multiwalled carbon nanotubes at a mass ratio of 7:1.5:0.5 using a mortar and pestle and sealed in a ball-milling tank with isovolumic mill balls. The blend was stirred thoroughly in air at 400 rpm for 3 h to obtain the desired product.

Ion chromatography

Since LiF can be partially hydrolysed into HF and evaporates, the ion chromatography to quantify the final F content was conducted on a Thermo ICS-600 instrument. The chromatographic column is from Dionex IonPac AS14. Each sample was placed in a clean polyester plastic bottle, ultrapure water was added to a volume of 10 ml, sealed with a lid and ultrasonic extraction was performed for final testing.

Electrochemical tests of LiVO_3–xF_x cathodes

Half-cell testing

A homogeneous slurry of the corresponding cathode material and poly(vinylidene fluoride) at a mass ratio of 9:1 was prepared. An electrode loading of 1–2 mg was slid into discs of ф = 14 mm. Then, the electrodes were vacuum dried overnight at 110 °C and placed into 2032-type coin cells in an argon glovebox with H₂O and O₂ levels at <0.1 ppm. The anode was a lithium metal ribbon, and a high-voltage electrolyte (LANTE) was used. The galvanostatic charge–discharge of the coin cells was analysed using a NEWARE tester at room temperature. The cycling cells at 200 mA g⁻¹ (1C) and 4,000 mA g⁻¹ were first discharged to 1.5 V for activation.

Full-cell testing

To fabricate the full cells, the LVOF electrodes were first discharged to 1.5 V at 10 mA g⁻¹ in half-cells for enough Li storage into the structure. Graphite was used as the anode and precycled for two cycles in half-cells to suppress the first-cycle capacity loss. Then, the cathode and anode were fabricated into 2032-type coin cells.

DEMS measurements

DEMS measurements were conducted to identify the gas species during the cycling process. For DEMS, an in situ cell consisting of a lithium metal ribbon as the anode and a Celgard 2320 film as the separator was assembled in an argon glovebox. The cell was then connected to a Hiden Analytical spectrometer, forming gas pathways with the Ar carrier gas flowing at a constant pressure.

CV measurements

CV measurements were performed using an electrochemical workstation (Chenghua).

Structural characterization

HR-TEM

The morphology and lattice structure of the as-prepared samples were characterized using a JEOL JEM-F200 transmission electron microscope with probe size mode 3 and convergence angle mode 5.

XRD analysis

The general XRD analyses were performed on a Bruker D8 Advance diffractometer equipped with a Cu Kα radiation source (λ = 1.5406 Å). The SXRD experiment was conducted at the BL14B1 station of the Shanghai Synchrotron Radiation Facility (SSRF) under an incident energy of 18 keV (0.686 Å). The high-energy SXRD was conducted at beamline 12SW in SSRF with an incident energy of 88.6 keV (0.139 Å). The refinement of SXRD was conducted using Le Bail refinement in the GSAS-II software.

Solid-state nuclear magnetic resonance spectroscopy

¹⁹F solid-state nuclear magnetic resonance spectra were acquired for both precursor and as-synthesized powder samples at room temperature. Data acquisition was performed using a Bruker 400-MHz WB solid-state NMR spectrometer (AVANCE III) with a 20-kHz magic-angle spinning setup. To establish accurate chemical shift references, LiF powder was used as the standard with a known chemical shift of diso(¹⁹F) = –204 ppm.

PDF

The PDF data were obtained at beamline 12SW in SSRF with an incident energy of 88.6 keV (0.139 Å). The experimental setup was calibrated using CeO₂. The exposure time was 300 s, for each sample, to guarantee the total signal intensity. The G(r) function was processed by the PDFgetX3 software using the Lorch function to obtain a low signal-to-noise ratio. The scattering data were truncated at and normalized at Q_max = 18.0 Å⁻¹. The resolution is 0.185 Å.

XAS

The V XAS spectra were collected in the transmission mode at beamline 1W1B using a Si(111) double-crystal monochromator at the Beijing Synchrotron Radiation Facility. O K-edge XANES data were obtained in the total-electron-yield mode at beamline 4B7B at the Beijing Synchrotron Radiation Facility. The data were normalized into unity jumps by setting the pre-edge at 510 eV to zero and the post-edge at 560 eV to unity. This method can exclude any disturbance from the absorption peaks, and is consistent with previous published works⁴⁷. The peak at 529.5 eV was set as the reference.

RIXS

The RIXS data were collected at the BL20U2 beamline of the SSRF. The mapping of RIXS was conducted by collecting emission signals within the excitation energy range of 527–532 eV at 0.2-eV intervals. The monochromator facility at the beamline is equipped with an 800 l mm⁻¹ high-line-density grating this time, and the accumulating time is long enough—1,800 s. Unlike the 300 l mm⁻¹ grating used previously, the 800 l mm⁻¹ grating used this time offers a superior beamline resolution (ΔE_beamline)⁴⁸, reducing the broadening of the elastic peak to 0.5 eV, which is the high-level resolving ability in SSRF.

XPS

The XPS spectra were measured using an AXIS Ultra instrument from Kratos Analytical (Al Kα radiation, hν = 1,486.6 eV). The data were corrected by setting the binding energy of the adventitious carbon (C1s) to 284.8 eV. The fitting process was conducted using Advantage software 6.8.1.

Raman spectra

The Raman spectra were conducted under 633-nm wavelength. The power is 1 mW and accumulation time is 100 s. For data parallelism, three areas of the electrode were collected.

Computational details

AIMD

We computed the potential energies and forces by using density functional theory (DFT) in conjunction with the Quickstep code⁴⁹ available in CP2K⁵⁰. For this purpose, we used the Perdew–Burke–Ernzerhof functional constructed using the Goedecker–Teter–Hutter pseudopotential⁵¹ and the double-zeta-polarized MOLOPT basis set⁵². The Goedecker–Teter–Hutter pseudopotential represents the atomic nucleus and core electrons, whereas the double-zeta-polarized MOLOPT basis set captures the behaviour of valence electrons. Our calculations focused exclusively on the Γ point within the reciprocal space.

DFT + U

Spin-polarized DFT calculations for LVO and LVOF were performed using the GGA + U method based on the Vienna ab initio simulation package code. The U value is determined to be 3.25 eV according to the Materials Project⁵³. The core electron states were represented by the projector-augmented wave method, whereas the Perdew–Burke–Ernzerhof exchange correlation and a plane-wave basis set with a cut-off energy of 500 eV were used for structural optimization. A Monkhorst–Pack k-point mesh of 3 × 3 × 6 was used.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41563-025-02293-9.

Author contributions

K.Z. and D.X. synthesized and designed the samples and carried out the experiments. D.X. provided guidance for the whole work. K.Z. and T.Y. analysed the data. K.Z. and D.X. composed the paper. K.Z. carried out the DFT calculation. K.Z. and T.C. conducted the full-cell experiment. K.Z., Z.J. and Y.Y. conducted the XAS experiments. C.G., Y.Z. and W.X. provided guidance for structure characterization.

Peer review

Peer review information

Nature Materials thanks Maria Hahlin and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Data availability

All data generated during this study are included in the Article and its Supplementary Information. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41563-025-02293-9.