Main

High-energy Ni-rich NMC (LiNixMnyCo1–x–yO2, x ≥ 0.8), such as LiNi0.8Mn0.1Co0.1O2 (NMC811) or LiNi0.9Mn0.05Co0.05O2 (NMC90), represents one of the most promising cathode materials for advanced lithium-ion battery technologies for electric vehicles. However, polycrystalline NMC811 and NMC90 have not been widely deployed in electric vehicle batteries due to intrinsic challenges such as moisture sensitivity1, cracking2 and gassing3 during cycling, which need to be addressed before large-scale implementation. These issues in Ni-rich NMC polycrystals, which originate from grain boundaries, can potentially be mitigated or eliminated in single-crystal format4, leading the innovation in cathode materials and advancing the value chain towards vehicle electrification5,6.

Two different categories of synthesis routes can be identified for single-crystal Ni-rich NMC: media-directed and solid-state synthesis. The former includes molten-salt5,7, hydrothermal8 or eutectic-assisted synthesis9, while the latter does not utilize reaction media and includes direct solid-state synthesis by ball milling10, multistep lithiation11,12,13 or conversion reaction14. Molten-salt and hydrothermal approaches lead to the formation of well-segregated crystals, which serve as excellent model materials for fundamental studies. Eutectic salts have been proposed to mill and break down transition metal hydroxide (TM(OH)2) precursors to form deagglomerated single crystals9. The additional lithium salts used in the eutectic approach create an environment similar to molten-salt reactions, aiding crystal growth. Solid-state synthesis without reaction media is advantageous from a manufacturing perspective and is developing rapidly. Dry synthesis by ball milling all precursors together, followed by calcination, is time and cost efficient, but the contamination from milling media poses a challenge for impurity control of the prepared NMC. Calcination of TM(OH)2 precursor with LiOH through multiple steps at very high sintering temperatures forms micrometre-sized primary particles, which, however, remain largely agglomerated12. Utilizing transition metal oxide (TMO) derived from TM(OH)2 as the precursor leads to the formation of segregated single crystals, though clusters are still observable14. Additional deagglomeration processes, such as jet milling15, are necessary to crush these aggregates.

One of the root causes of forming agglomerated single crystals is the melting process of lithium salt in the precursors. For Ni-rich cathodes such as NMC811, whether polycrystalline or single crystal, LiOH or LiOH·H2O is usually used as the lithium salt precursor8, while low-cost Li2CO3 is utilized for synthesizing NMC622, NMC532 or NMC111, which have less nickel content. This preference arises because NMC811 (and NMC90) need to be prepared at relatively low temperatures to prevent Ni reduction at high temperatures. To match the reduced synthesis temperature of NMC811 and beyond, LiOH or LiOH·H2O, with a relatively low melting point of 462 °C, is more beneficial than Li2CO3 (melting point 723 °C). Current wisdom also holds that lithium salt needs to melt first to fully wet the TM(OH)2 precursors for the reaction to occur homogeneously. However, the melting process of lithium salts also randomly ‘glues’ the as-formed single crystals into large clusters during calcination.

Li2O, with a high melting point of 1,438 °C, is usually considered unsuitable for cathode material synthesis and typically avoided during NMC growth16. In this work, however, we discuss an unusual sublimation phenomenon of Li2O, which surprisingly promotes the growth and segregation of single crystals surrounded by Li2O vapour. The advantage of Li2O sublimation is further demonstrated by converting spent polycrystalline NMC811 into completely segregated large single crystals. It is experimentally proven that Li2O sublimation occurs at 1 atm of pure oxygen at 870 °C during NMC synthesis. This discovery holds great potential for synthesizing and manufacturing a broad range of single crystals for various applications.

Single crystals from Li2O and fresh or spent precursors

The utilization of Li2O for synthesis of fresh NMC90 and NMC811 single crystals, as well as for the direct conversion of spent polycrystalline NMC811 into single crystals, is displayed in Fig. 1. Both Ni0.9Mn0.05Co0.05(OH)2 (Fig. 1a(i)) and Ni0.8Mn0.1Co0.1(OH)2 (Fig. 1b(i)) precursors, with median diameters (D50) of 3.75 μm and 5.42 μm, respectively (Supplementary Fig. 1), hereafter referred to as TM(OH)2, are prepared using the conventional coprecipitation method. TM(OH)2 is decomposed into TMOs14 (Fig. 1a(ii),b(ii)) before reacting with Li2O. This TM(OH)2-to-TMO conversion is necessary to facilitate deagglomeration of the later-formed crystal clusters when using LiOH salt, as we reported earlier14. To maintain consistency with our previous work, we continue to use TMO as precursor in this study, although the salt precursor has now been changed to Li2O.

Fig. 1: Synthesis and electrochemical evaluation of NMC single crystals using Li2O from various precursors.
figure 1

a(i)–(iii), Scanning electron microscopy (SEM) morphology of Ni0.9Mn0.05Co0.05(OH)2 (i), TMO90 oxide precursor (ii) and NMC90 single crystals (iii). b(i)–(iii), SEM morphology of Ni0.8Mn0.1Co0.1(OH)2 (i), TMO811 oxide precursor (ii) and NMC811 single crystals (iii). c(i)–(iii), SEM morphology of as-received spent NMC811 (i), LixTMO (ii) and NMC811 single crystals from spent NMC811 (iii). NMC90 (a(iii)) and NMC811 (b(iii),c(iii)) single crystals were synthesized from hydroxide precursors TM(OH)2 (a(i),b(i)) and spent NMC811 (c(i)), which were converted to TMO oxide precursors (a(ii),b(ii)) and LixTMO (c(ii)). The synthesis involved a high-temperature anneal with Li2O at 870 °C for NMC90 and 900 °C for NMC811. a(iv),b(iv),c(iv), The voltage profiles from the first cycle of single crystals of NMC90 (a(iv)), NMC811 (b(iv)) and NMC811 derived from spent NMC811 (c(iv)) in Li/NMC half cells, cycled at 0.1 C (where 1 C = 200 mA g−1) between 2.7 and 4.4 V.

Source data

Full size image

NMC90 (Fig. 1a(iii)) and NMC811 (Fig. 1b(iii)) single crystals (see Supplementary Fig. 2 for X-ray diffraction (XRD)) derived from Li2O salt are well segregated, with average crystal sizes of 2–3 μm and 3–4 μm, respectively. It appears that each individual TMO precursor has been fully sintered into a single crystal, as the TMO precursors (Fig. 1a(ii),b(ii)) are also approximately 2–3 μm in size. The first discharge capacity of NMC90 is 221 mAh g−1 (Fig. 1a(iv)), higher than that of NMC811, which has a capacity of 206 mAh g−1, due to the increased amount of Ni in NMC90 (Fig. 1b(iv)). Only one heating process is needed when using Li2O as the salt precursor, whereas the LiOH approach requires multiple heating steps, including preheating all precursors14. Another advantage of using TMO and Li2O precursors for NMC synthesis is that no corrosive water is generated during the heating process, a common concern in NMC manufacturing.

A similar approach is adopted to ‘recycle’ spent NMC811. Surprisingly, Li2O directly converts spent NMC811 polycrystals into single crystals (Fig. 1c(i)–c(iii)) without the need to mill polycrystals, which is usually required in the literature9,17. Commercial NMC811 polycrystals have variable particle sizes ranging from 2 to 10 μm (Fig. 1c(i)), likely to enhance the density at electrode level18. Spent polycrystalline NMC811, scratched from the cycled electrode (Fig. 1c(i)), is heated at 950 °C in O2 to burn out the binders, residual electrolyte and so on in the electrode. More importantly, the heating process emulates the TM(OH)2-to-TMO conversion step. The only difference is that there is an unknown amount of lithium present in the decomposed LixNi0.8Mn0.1Co0.1O2−δ (referred to as LixTMO hereafter). Decomposed LixTMO is also a mixture of spinel and rock-salt phases, similar to TMO14 converted from TM(OH)2 (Supplementary Fig. 3 for XRD), and mostly retains the morphologies (Fig. 1c(ii)) of the original spent NMC811 polycrystals. There is no need to quantify residual lithium in LixTMO because LixTMO will be mixed with additional Li2O to reconstruct the layered lattice structure. After reacting with Li2O, irregularly shaped LixTMO agglomerates (Fig. 1c(ii)) are all converted into well-segregated crystal ‘rocks’ (Fig. 1c(iii)). The NMC811 single crystals (Fig. 1c(iii) and Supplementary Fig. 4) recycled from spent polycrystals are generally larger than freshly made ones (Fig. 1a(iii),b(iii)) because the spent polycrystals from cathodes made of commercial NMC are larger than laboratory-made precursors. The first discharge capacity from recycled single-crystal NMC811 is as high as 209 mAh g−1 (Fig. 1c(iv)).

Note that all the electrochemical tests for the NMC811 crystals in Fig. 1 were conducted in half cells, with lithium metal used as the anode. This ensures that there is no lithium deficiency issue when assessing the achievable discharge capacities of these single crystals. The cutoff voltages of the three cells in Fig. 1 are all set to 4.4 V (versus Li/Li+) to extract as many lithium ions as possible from the layered structure. To evaluate the cycle stability of cathode, graphite-based full cells are still needed, which will be discussed in a later section.

The above findings confirm that Li2O not only can be used as a lithium salt precursor for cathode synthesis below its melting point, but also unexpectedly promotes single-crystal growth and simplifies the synthesis process by reducing the number of heating steps, eliminating corrosive water and even directly converting polycrystal to single crystals. However, there are at least two key questions that remain to be answered to understand the unique role of Li2O in promoting single-crystal growth. First, the reaction temperature in this work is 850–900 °C, much lower than the melting point of Li2O (1,438 °C). How does this solid–solid reaction between Li2O and TMO occur completely, especially for those TMO particles that have no direct contact with Li2O? Second, why do single crystals developed from Li2O exhibit no serious agglomeration, even those derived from spent polycrystalline NMC811?

When compared with TMO precursors, which are approximately 3–4 μm in size, Li2O salts are notably larger, ranging from a few micrometres to 100 μm (Supplementary Fig. 5). Even after mixing with TMO, Li2O remains prominent as ‘giants’, with some TMO particles adhering to the surfaces of Li2O (Supplementary Fig. 6a). Despite these vastly mismatched precursors, after heating at 860 °C, all TMO is transformed into NMC90 single crystals (Supplementary Fig. 6d), even though most of the TMO precursors initially do not have direct contact with Li2O.

On the basis of these observations, it is hypothesized that Li2O sublimates, and it is primarily the Li2O vapour that drives the chemical reaction between TMO and Li2O to proceed well below the melting point of Li2O. This hypothesis explains why the mismatch in particle sizes between TMO and Li2O does not affect the extent of the reaction, as Li2O is present in both solid and gas phases during the chemical lithiation process of TMO. As Li2O vapour permeates, it fills in the pores of TMO agglomerates, sintering individual TMO primary particles into a crystal. This aligns with the observation that NMC811 single crystals are similar in size to their TMO precursors (Fig. 1). Because Li2O never melts during the high-temperature synthesis process, the single crystals formed are not as severely glued together as those with LiOH, resulting in well-segregated single crystals. Although Li2O is only known to form very low-pressure vapour under vacuum conditions19 (for example, 3.0 × 10−8 Pa at 1,000 K with a partial pressure of oxygen (pO2) of 0.2 bar (ref. 20)), here Li2O sublimation at 1 atm of O2 pressure has been utilized to sinter single crystals, presenting potential for broad applications. However, in the absence of a TMO precursor, Li2O itself cannot maintain a continuous sublimation under the same temperature of 870 °C (10 h in an O2 atmosphere). The weight loss of Li2O itself at 870 °C is only 2.70% (Supplementary Fig. 7) because Li2O vapour ‘sinters’ the small Li2O particles into large ones. The sintered large Li2O particles have reduced surface area and vapour pressure, self-limiting the further sublimation of Li2O. In the presence of TMO, the reaction between the TMO and Li2O shifts the solid–vapour equilibrium of Li2O toward its vapour phase, enabling continuous Li2O sublimation to promote single-crystal growth. Under near-vacuum conditions, the presence of TMO further accelerates the sublimation of Li2O, leading to a fast disappearance of large Li2O rocks in the mixtures (see details in Supplementary Video 1). It has also been noticed that the Li2O used in this work has porous structure, which may also facilitate the sublimation of Li2O when reacting with TMO at high temperatures.

Single-crystal growth without contacting Li2O

To further substantiate Li2O sublimation, we designed an experiment where completely separated Li2O and TMO were heated in a sealed crucible filled with pure O2 (Fig. 2a; see Supplementary Fig. 8 for experimental details). During heating, if Li2O sublimates, its vapour should circulate within the inner crucible containing only TMO (Fig. 2a) and facilitate a reaction. We employed two different TMO precursors to demonstrate the formation of both NMC90 and NMC811 single crystals enabled by Li2O vapour in this sealed environment. Each experiment was repeated at least once. After 10 h of heating at 870 °C (for NMC90) or 900 °C (for NMC811), the crucible was disassembled. Residual Li2O was observed at the bottom of the container (Supplementary Fig. 9a), and the TMO was transformed into large single crystals (Fig. 2b–d). Scanning transmission electron microscopy (STEM; Fig. 2d,g and Supplementary Figs. 10 and 11) and energy-dispersive X-ray spectroscopy (EDS; Supplementary Figs. 12 and 13) further confirmed that the harvested crystals were layered NMC90 (Fig. 2b,c) or NMC811 (Fig. 2e,f), depending on the TMO precursor used. Unlike the reaction with mixed TMO and Li2O precursors, the reaction in this experiment (Fig. 2a) relied entirely on sublimed Li2O vapour without stoichiometry control. Upon cooling, leftover Li2O was also found among the as-formed single crystals (Supplementary Figs. 10c1,c5,c7 and 12a1), further confirming the presence of Li2O vapour. After washing off the residual Li2O, the single crystals became very distinct (Fig. 2c and Supplementary Fig. 12b for NMC90; Fig. 2f and Supplementary Fig. 13b for NMC811). This convincingly proves that Li2O sublimation occurs and can be utilized to form single crystals even without direct contact with other precursors. Moreover, Li2O sublimation facilitates the growth of NMC811 single crystals to a large size of 7–17 µm (Supplementary Fig. 14), mimicking the environment created by molten salts21. Li2O has also been used as a sintering agent for other ceramics such as Li7La3Zr2O12 (refs. 22,23). In those cases, Li2O vapour is generated from melted Li2O heated at very high temperatures (>1,160 °C or 1,300 °C) and utilized to densify the Li7La3Zr2O12 ceramic and reduce the grain boundaries. Even so, Li2O sublimation may co-exist because the high temperatures used in those papers are still lower than the melting point of Li2O. Our finding also explains why Li2O, when used as a sintering agent, can effectively lower the sintering temperatures of ceramics by a few hundred degrees Celsius24,25.

Fig. 2: Vapour study between Li2O and TMO oxide precursors.
figure 2

a, Set-up of the Li2O vapour study. TMO oxide precursor was placed in a small alumina crucible, while Li2O was housed in a larger separate crucible. The system was sealed in O2 at 1 atm using a ceramic ball and adhesive. After curing, the crucibles were heated to 870 °C (NMC90) or 900 °C (NMC811) for 10 h. The NMC product was then removed for SEM/TEM characterization. m.t., melting temperature. b, SEM image of NMC90 single crystals from the vapour study, showing octahedral shape with particle sizes from 1 to 3 µm. Residual Li2O is visible on the surfaces of NMC90 particles. c, After washing with water, the NMC90 surface is clean. d, High-angle annular dark-field (HAADF) STEM image of the interior structure of an unwashed NMC90 single crystal from b with a corresponding electron diffraction pattern (inset), revealing a layered structure. e, SEM image of NMC811 single crystals from the vapour study, showing irregular shapes and particle sizes from 1 to 3 µm, with rough surfaces covered with Li2O. f, After washing with water, the NMC811 surface is clean. g, HAADF-STEM image of the interior structure of an unwashed NMC811 single crystal from e with a corresponding electron diffraction pattern (inset), indicating a layered structure. This vapour study confirms the presence of Li2O vapour in the solid-state reaction between Li2O particles and TMO oxide precursors, even below the melting temperature of Li2O, promoting the growth of NMC single crystals.

Full size image

In situ XRD study and simulation of Li2O sublimation

To elucidate the structural evolution during single-crystal growth, in situ heating XRD was employed to examine the heating process of precursors by monitoring the lattice structures of Li2O/TMO mixtures (Fig. 3a). STEM was also used to analyse the Li2O/TMO mixtures at various heating temperatures.

Fig. 3: Structural evolution of the reaction between TMO811 oxide precursor and Li2O.
figure 3

a, In situ XRD characterization (λ = 0.182 Å) tracks the reaction between TMO811 oxide precursor and Li2O, following a procedure similar to that used for synthesizing NMC811 single crystals. b–g, Ex situ high-resolution TEM characterization of the reaction products between TMO811 oxide precursor and Li2O at critical temperatures indicated in a: at 300 °C (b), at 500 °C (c), at 500 °C after 7 h (d), at 750 °C (e), at 850 °C (f) and at 850 °C after 7 h (g).

Full size image

In the original TMO precursor, the diffraction peaks at 2.4, 3.7 and 6.4° are characteristic of Li2O. A trace amount of LiOH is also identified, probably due to Li2O’s exposure to air during sample preparation. The TMO’s diffraction peaks, located at 4.3, 5.0 and 7.1°, correspond to mixed spinel and rock-salt phases, as previously reported14. With the temperature increasing from 30 to 300 °C, there are no noticeable changes in the X-ray peaks (Fig. 3a). The selected-area diffraction pattern also indicates the coexistence of rock-salt, spinel, Li2O and LiOH phases (inset of Fig. 3b), confirming that it remains a mixture of precursors at 300 °C.

XRD patterns indicate that the lithiation process does not commence until 500 °C (Fig. 3a), with the lithiated precursors existing in a very primitive form (Fig. 3c), showing serious Li/TM mixing. Following the isothermal process at 500 °C, characteristic diffraction peaks of NMC811 at 2.2, 4.3, 5.2, 7.2, 8.5 and 8.7° emerge, and the crystallinity of the lithiated mixtures gradually increases (Fig. 3a,d). After 7 h of heating at 500 °C, a few Li2O grains are still visible within the lithiated TMO particles (Fig. 3d). The Li2O and TMO precursors are initially mixed by a mild vibration process, which does not break down large chunks of Li2O. Thus, the nanodomain of Li2O seen in Fig. 3d probably results from the diffusion of Li2O vapour among primary TMO particles, further evidencing Li2O sublimation. Continuous chemical lithiation of TMO leads to complete layered lattice structures by about 750 °C, as confirmed by both XRD patterns (Fig. 3a) and TEM images (Fig. 3e). Li2O residue, along with rock-salt and spinel phases, is still detected at 750 °C, indicating an incomplete reaction at this temperature. As the temperature reaches 850 °C, the layered α-NaFeO2 phase becomes dominant (Fig. 3a), although spinel and rock-salt phases are still visible (Fig. 3f). No Li2O is identifiable in the mixtures at 850 °C (Fig. 3f), but the chemical lithiation of TMO continues during the isothermal process at 850 °C because the XRD peaks at 5.4, 6.5 and 7.2°, corresponding to the (105), (107) and (113) planes of NCM811, appear (Fig. 3a). This is consistent with the proposed Li2O sublimation mechanism, as at 850 °C Li2O vapour is now the main participant in the chemical reaction. After 7 h of heating at 850 °C, a pure phase of NMC811 readily forms (Fig. 3a) with very good crystallinity (Fig. 3g). During the cooling process, all the intensities of XRD peaks (Fig. 3a) show a slight shift towards a higher angle, indicating a slight shrinkage of lattice volume. Detailed changes in the a and c lattice parameters and Li/TM cation mixing during in situ heating XRD characterization are provided in Supplementary Fig. 15. In addition to the morphological and structural changes during the synthesis process, the evolution of the valence state of transition metals, particularly Ni, is critical for understanding the reaction mechanism. As shown in the L2,3-edge electron energy-loss spectra of transition metals in Supplementary Fig. 16, the L3/L2 ratio of Ni in the initial TMO precursor at 30 °C is very high, approximately 4.8, indicating a low valence state of Ni. This is observed even after heating TMO811 from TM(OH)2811 at 900 °C for 15 h in pure O2. As the temperature of the TMO and Li2O mixtures increases, the L3/L2 ratio of the transition metals gradually decreases and stabilizes around 700 °C. According to in situ XRD studies, the layered lattice structure begins to form from TMO at approximately 700 °C, and during this stage a corresponding increase in the oxidation state of Ni is also observed.

Previous studies have investigated Li2O vapour formed from melted Li2O (melting point 1,438 °C) rather than from solid Li2O, in terms of gaseous equilibria26,27,28. However, in this work the reaction between Li2O vapour and TMO occurs at 870 °C through a sublimation process in a pure oxygen atmosphere. Previous reports of Li2O sublimation primarily focused on vacuum conditions26,28, whereas the synthesis in this study is conducted at 1 atm of pure oxygen.

Both theoretical and experimental observations have demonstrated that the sublimation of certain metal or metal oxide particles is size dependent29,30,31,32,33. In this study the Kelvin equation is employed to estimate the critical size of Li2O particles during the sublimation process.

$${\rm{ln}}\left(\frac{{P}_{r}}{{P}_{\infty }}\right)=\left(\frac{M}{T{R}\rho }\right)\left(\frac{{2\gamma }_{t}}{r}\right)$$
(1)

where M is the molecular weight, R is the ideal gas constant, ρ is the density of the solid, T is the temperature, Pr is the vapour pressure above a particle of radius r, P∞ is the equilibrium vapour pressure of the sublimating species over a flat surface, r is the radius of the Li2O particle and γt is the weighted average surface free energy associated with the facets (Supplementary Fig. 17a) of the Li2O solid. It is assumed that the solid is isotropic, and γt can be calculated via density functional theory for specific facets at various temperatures and pressures using equation (2) (details of the simulation can be found in Methods and Supplementary Note 1).

The simulation results (Supplementary Fig. 17b) show a noteworthy increase in the critical size of the Li2O particle as the vapour pressure ratio decreases. Initially (Supplementary Fig. 17c), the reaction of Li2O vapour with the NMC precursor leads to a decrease in the average vapour pressure of Li2O. This causes the vapour pressure surrounding the Li2O particle to approach the pressure of a flat Li2O surface. The critical size of the Li2O particle reaches tens of micrometres (Supplementary Fig. 17b) when the vapour pressure ratio slightly exceeds 1.0, according to equation (1). Notably, the particle size distribution of Li2O used in this study has a D50 of 32.9 μm, suggesting that a few tens of micrometres of Li2O can fully convert TMO precursors into NMC via sublimated Li2O vapour surrounding the TMO. As the reaction between Li2O (vapour) and TMO progresses, Li2O vapour is continually replenished from solid Li2O to sustain the reaction.

Driven by the chemical reaction to form NMC single crystals, Li2O gas continuously interacts with TMO solids through gas–solid interactions, leading to a unique sintering process. In this process, Li2O penetrates all voids within the TMO to react, effectively sintering the single crystals. This reaction involves no melted lithium salts, so the resulting crystals do not glue together as they might with LiOH salt. The outcome is well-segregated single crystals.

Single crystals after 1,000 cycles

The cycling stability of Li2O-derived single crystals is further evaluated in graphite-based full cells (Fig. 4a(i),b(i),c(i)). The cutoff voltage of all cells in Fig. 4 is set at 4.2 V versus graphite (Gr), which corresponds to approximately 4.3 V versus Li/Li+. It is well documented that NMC with high nickel content becomes unstable at elevated voltages. For instance, LiNi0.95Mn0.04Co0.01O2 remains stable only at a restricted voltage of 4.04 V versus Gr34. Therefore, a cutoff voltage of 4.2 V is chosen to cycle NMC811 and NMC90 in full cells for a consistent comparison. The cycling performances of all single crystals charged to an elevated voltage of 4.3 V versus Gr (equivalent to 4.4 V versus Li/Li+) are also presented in Supplementary Fig. 18.

Fig. 4: Electrochemical evaluation and postmortem analysis of as-prepared NMC single crystals.
figure 4

a(i),b(i),c(i), Voltage profiles of the first formation cycle for as-prepared single crystals: NMC90 (a(i)), NMC811 (b(i)) and NMC811 derived from spent NMC811 (c(i)) in Gr/NMC full cells cycled at 0.1 C (1 C = 200 mA g−1 for NMC90 and NMC811 from spent NMC811; 1 C = 190 mA g−1 for NMC811) between 2.6 and 4.2 V. a(ii),b(ii),c(ii), Cycling performance of Gr/NMC full cells at 1 C with single crystals: NMC90 (a(ii)), NMC811 (b(ii)) and NMC811 from spent NMC811 (c(ii)). All cells were cycled up to the 1,000th cycle, after which the cycled NMC cathode electrodes were harvested for postmortem analysis. a(iii),b(iii),c(iii), SEM morphology of the surface of the NMC single-crystal electrodes after 1,000 cycles: NMC90 (a(iii)), NMC811 (b(iii)) and NMC811 from spent NMC811 (c(iii)). a(iv),b(iv),c(iv), HAADF-STEM images of the NMC single crystals after 1,000 cycles: NMC90 (a(iv)), NMC811 (b(iv)) and NMC811 from spent NMC811 (c(iv)).

Source data

Full size image

When cycled between 2.6 and 4.2 V in a full cell, the initial discharge capacities from single-crystal NMC90 (Fig. 4a(i)), NMC811 (Fig. 4b(i)) and NMC811 converted from spent polycrystal cathode (Fig. 4c(i)) all decrease slightly when compared with their respective half-cell performances, which utilize lithium metal as the counter-electrode (Fig. 1a(iv),b(iv),c(iv)). This decrease is attributed to the reduced electrochemical window and some irreversible consumption of Li+ ions on the anode side, such as solid–electrolyte interphase formation on graphite during the initial cycling. Notably, in full cells Li+ ions are only stored on the cathode side, whereas in half cells they are stored in both the cathode and the lithium metal anode. Despite these conditions, the trend in usable capacity is consistent across the three types of single crystal, regardless of the anode materials used. Specifically, NMC90 delivers a high capacity of 195.0 mAh g−1 when compared with NMC811’s 188.6 mAh g−1, although its initial Coulombic efficiency of 88.6% is slightly lower than NMC811’s 89.6%, suggesting more potential for enhancing NMC90’s reversibility. Remarkably, NMC811 single crystals reformed from spent polycrystal NMC811 demonstrate even higher capacity of 197.1 mAh g−1 with a good Coulombic efficiency of 88.8%. By the second formation cycle, the capacities of NMC90 and NMC811 single crystals increase to 200.8 mAh g−1 and 194.5 mAh g−1, respectively (Supplementary Fig. 19). After formation cycles (Fig. 4a(i),b(i),c(i)), the three cells were tested at an accelerated rate of 1 C for both charging and discharging (see Supplementary Fig. 20 for the charge–discharge curve of the first cycle at 1 C). Among these, NMC90 exhibits the fastest decay, which is expected given its higher Ni content, achieving 80% capacity retention after 426 cycles (Fig. 4a(ii)). In contrast, both single-crystal NMC811 (Fig. 4b(ii)) and the remade ‘NMC811’ (Fig. 4c(ii)) from spent polycrystals show extraordinary cycling stability, maintaining over 80% capacity even after 1,000 cycles.

The extensively cycled single crystals exhibit varying degrees of cracking and gliding, predominantly in the larger crystals (Fig. 4a(iii),b(iii),c(iii)). Generally, most single crystals maintain their integrity without obvious damage (Supplementary Fig. 21), whereas their polycrystalline counterparts often show deterioration after long-term cycling35. Of particular interest is the evolution of cation mixing and surface layers on the three different single crystals. Before cycling, fresh NMC90 displays a thin, approximately 0.5 nm, amorphous surface film (Supplementary Fig. 22a1,a2). In contrast, the surfaces of both freshly made and remade NMC811 single crystals from spent polycrystals are exceptionally clean, with no discernible surface film (Supplementary Fig. 22b1,b2,c1,c2). However, single-crystal NMC811 derived from spent polycrystals exhibits a clear cation mixing layer about 5 nm thick before cycling (Supplementary Fig. 22c1,c2), probably a result of the previous cycling of the used polycrystalline NMC811. Freshly synthesized NMC90 (Supplementary Fig. 22a1) and NMC811 single crystals (Supplementary Fig. 22b1) show no apparent cation mixing layer.

After 1,000 cycles, the cathode–electrolyte interphase (CEI) growth on NMC90 crystals results in a thicker passivation layer, now 4 nm, accompanied by the formation of an approximately 1.2 nm cation mixing layer (Fig. 4a(iv) and Supplementary Fig. 22a3). For NMC811, the CEI layer remains thin at about 2.2 nm after 1,000 cycles, with a cation mixing layer of approximately 1.6 nm (Fig. 4b(iv) and Supplementary Fig. 22b3), indicating the lattice stability of the as-prepared single crystals. For the remade NMC811 single crystals derived from spent polycrystals, not only is the surface layer the thickest at 7 nm (Fig. 4c(iv) and Supplementary Fig. 22c3) among the cycled crystals, but the cation mixing layer has also tripled to 15 nm after 1,000 cycles (Fig. 4c(iv)). It is important to note that the remade single-crystal NMC811 had a pre-existing cation mixing layer of 5 nm from the beginning (Supplementary Fig. 22c1,c2). Given that the remade NMC811 demonstrates the superior cycling stability but also the thickest cation mixing layer among the three crystals after 1,000 cycles, it is possible that the cation mixing layer needs to reach a certain critical thickness to play a dominant role in dictating cathode stability. Conversely, the surface layer or CEI has a more considerable impact on cathode cycling, as evidenced by this work and others34, which may trigger more substantial impedance growth. Although the CEI of NMC90 (Fig. 4a(iv)) is thinner than that of remade NMC811 (Fig. 4c(iv)), NMC90 decays much faster than both remade NMC811 and freshly made NMC811. Cathode impedance growth becomes more severe with increased Ni at the interface between the cathode and electrolyte34,36. Functional electrolyte and element doping or coating are necessary for NMC90 and higher to tune CEI properties and further enhance their cycle stability.

Conclusions

The sublimation of Li2O has been utilized to synthesize single-crystal Ni-rich cathode materials, effectively mimicking a molten-salt environment without melting any salts. Owing to the rapid diffusion of Li2O vapour during high-temperature reactions, large chunks of Li2O salt precursors can be used directly, eliminating the need for premilling and simplifying the scale-up process of single crystals. Li2O sublimation is further employed to directly convert spent polycrystalline NMC811 into well-segregated single crystals through an effective sintering process enabled by Li2O vapour. Ni-rich single-crystal cathodes developed from Li2O demonstrate excellent cycling stability after 1,000 cycles. More interestingly, even for large single crystals of NMC811 remade from spent polycrystals, 82.9% capacity retention is maintained after 1,000 cycles. Postmortem analysis of extensively cycled single crystals suggests that the surface passivation layer, rather than the original cation mixing layer, plays a critical role in determining the cathode stability. Li2O sublimation, driven by chemical reactions at high temperatures, provides fundamentally new insights into making single crystals with controllable properties, potentially applicable across various fields.

Methods

Single-crystal synthesis

The NMC precursors, Ni0.9Mn0.05Co0.05(OH)2 and Ni0.8Mn0.1Co0.1(OH)2 (denoted as TM(OH)2), were synthesized using the coprecipitation method, following the same procedures as described in our previous work14. A two-step method was employed for synthesizing NMC single crystals, involving (1) converting hydroxide precursors TM(OH)2 to oxide precursors TMO and (2) heating TMO with lithium salt, where Li2O (Albemarle Corporation) was selected as the lithium source instead of the LiOH used in previous studies. Approximately 8 g of oxide precursor (either Ni0.9Mn0.05Co0.05O or Ni0.8Mn0.1Co0.1O, denoted as TMO) was produced by heating 10 g of the corresponding TM(OH)2 in an alumina crucible at either 500 °C (TM(OH)290) or 700 °C (TM(OH)2811) for 15 h in pure oxygen at a flow rate of 200 cm3 min−1 in a tube furnace (GSL-1100x, MTI Corporation). The heating and cooling rate was maintained at 10 °C min−1. Oxide precursor (4 g of either TMO90 or TMO811) was then mixed with 1.2 g of Li2O and 10 g of plastic balls in a round aluminium container within an Ar-filled glovebox (MBRAUN). The Li/TM ratio was 1.4–1.5. The mixture was further homogenized using a rolling mill (VMR-3R Mix Rotor) for 2 h. Single crystals of NMC90 or NMC811 were prepared by heating 2 g of the TMO/Li2O mixture in an alumina crucible at 870 °C (TMO90) or 900 °C (TMO811) for 10 h in pure oxygen at the same flow rate. The heating and cooling rate was reduced to 5 °C min−1. After completion, the product was transferred to a glovebox, ground manually with a mortar and pestle, and sieved through a 400-mesh screen. The powder was then washed with deionized water, filtered and dried at 60 °C under vacuum. To repair the surface of NMC single crystals during washing, a post-annealing step at 580 °C for 4 h was necessary. About 1.5 g of NMC90 or NMC811 single crystals was finally obtained. These crystals, sometimes found as aggregates, were characterized using SEM, and for improved electrochemical performance they were gently segregated by mild ball milling using a THINKY mixer at 800 r.p.m. for 10 min. Spent polycrystalline NMC811 from a cycled 350 Wh kg−1 Li/NMC811 pouch cell provided by the Batt500 project underwent the same procedure for conversion to single crystals. Before mixing with Li2O, the spent NMC811 was heated at 950 °C for 15 h under pure oxygen to remove completely carbon additive, polyvinylidene difluoride binder and electrolyte residue. Characterization of Li2O and NMC single crystals was performed using powder XRD (Rigaku MiniFlex II, Cu Kα (λ = 1.54 Å)) with a scan speed of 2° min−1 and a step size of 0.05°, covering angular ranges of 20°–70° for Li2O and 10°–90° for NMC single crystals. Particle size distributions of Li2O and TM(OH)2 precursors were measured using a laser diffraction particle size analyser (Malvern Panalytical, Mastersizer 3000E), with isopropyl alcohol as the dispersant and sonication for 5 min.

Electrode preparation and cell assembly

The NMC single-crystal cathode electrode was prepared by mixing a slurry consisting of as-prepared NMC90 (or NMC811) single crystals, carbon additive (C-NERGY C65) and polyvinylidene difluoride (Kynar HSV1800) in a weight ratio of 96:2:2, with N-methyl-2-pyrrolidone (MSE) as solvent. This mixture was coated onto 12-µm-thick aluminium foil using a doctor blade in a dry room with 0.1–0.3% humidity. After drying overnight at 60 °C under vacuum, the NMC electrode was calendared to ~3.0 g cm−3 and punched into 12.7-mm-diameter round discs. The loading of most NMC electrodes was 10–14 mg cm−2, corresponding to an areal capacity of 2–3 mAh cm−2. For half-cell assembly, 16-mm-diameter round discs of lithium metal foil were punched from 50 µm lithium foil laminated onto 10 µm Cu foil (China Energy Lithium Co., Ltd). For full-cell evaluation, a graphite electrode was used as anode, prepared similarly by coating a slurry comprising of graphite (Superior Graphite), carbon additive and polyvinylidene difluoride in a weight ratio of 95:1:4 onto 10-µm-thick Cu foil. This was also dried overnight at 60 °C under vacuum, then calendared to ~1.5 g cm−3 and punched into 14-mm-diameter round discs. The loading of the graphite electrode was 7–9 mg cm−2, with a negative/positive capacity ratio controlled to ~1.1. Both the NMC and graphite slurries were mixed using a THINKY mixer (ARE-310). The assembly of Li/NMC half cells and Gr/NMC full cells was performed in an Ar-filled glovebox (O2 < 0.1 ppm, H2O < 0.1 ppm), using 2032 coin cells with an aluminium-coated positive can (MTI Corp, USA) and 19-mm-diameter round discs of polyethylene separator (20 µm thick, Energy Technology Solution). The electrolyte used was a mixture of lithium bis(fluorosulfonyl)imide (Gotion), ethyl propionate (Sigma-Aldrich), ethylene carbonate and 1,1,2,2-tetrafluoroethyl-2,2,3,3-tetrafluoropropyl ether in a molar ratio of 1:2.8:0.2:1, with an addition of 1 wt% lithium difluorophosphate37. Each coin cell was injected with 70 µl of this electrolyte.

Electrochemical testing

The specific capacity and voltage profile of NMC single crystals were assessed using Li/NMC 2032 coin cells on a Landt battery tester (CT2001A). The cells were charged to 4.4 V at a constant-current rate of 0.1 C (1 C = 200 mA g−1) and discharged to 2.7 V at 25 °C in a battery testing chamber (TE-TEC1, TestEquity). To evaluate the cycling stability, Gr/NMC 2032-type full cells were utilized. These cells underwent three formation cycles at 0.1 C between 2.6 and 4.2 (or 4.3) V before the cycling test started. After formation, the cells were cycled at either 1 C between 2.6 and 4.2 V or 0.33 C between 2.6 and 4.3 V at 25 °C in the chamber. A constant-voltage charge at 4.2 (or 4.3) V after the constant-current charge was applied until the current dropped to 0.05 C. After 1,000 cycles at 1 C, the cells were disassembled in an Ar-filled glovebox. The harvested NMC cathodes were rinsed with dimethoxyethane and characterized with SEM and (S)TEM after drying.

Vapour experiment

To validate the presence of Li2O vapour during the synthesis of NMC single crystals, a vapour study as outlined in Supplementary Fig. 8 was conducted. The small alumina crucible was cut from a large alumina crucible (NETZSCH) primarily used for thermal testing. The small crucible containing either 3.6 mg TMO90 or 9.1 mg TMO811 was placed inside a large crucible with either 44.3 mg Li2O or 55.6 mg Li2O, respectively, in an Ar-filled glovebox. Although the Li/TM ratio was not maintained at 1.4–1.5, a substantial amount of Li2O was used to ensure sufficient vapour generation. An alumina support made from an alumina foam block (MTI Corporation) was employed to hold the crucibles. The crucibles with the TMO and Li2O on the alumina support were then loaded into a tube furnace and flushed with O2 at a flow rate of 200 cm3 min−1 for 2 h. Once saturated with O2, the crucibles were removed, and a ceramic ball was placed atop the larger crucible to seal in the oxygen at 1 atm. Ceramic sealant (Ceramabond 503 alumina adhesive, AREMCO) was applied around the junction of the ceramic ball and the larger crucible under a continuous flow of O2. The sealant was cured at temperatures of 93 °C, 260 °C and 371 °C each for 2 h under an O2 flow of 200 cm3 min−1 in the tube furnace. The sealed crucible was then heated to 870 °C for NMC90 or 900 °C for NMC811 single crystals for 10 h, with a ramp rate of 5 °C min−1 and an O2 flow of 200 cm3 min−1 in the tube furnace, mirroring the conditions used in the synthesis of these crystals. After the heating was completed, the ceramic ball, sealant and alumina support were removed in a dry room. The harvested samples from the small crucible and Li2O residue in the large crucible were characterized with SEM/EDS/(S)TEM to confirm the formation of single crystals.

SEM/TEM characterization

The morphology of as-prepared NMC precursors (TM(OH)2 and TMO), spent NMC, single crystals and as-received Li2O was captured using SEM (JSM-IT200LA, JEOL)14. The element analysis of as-received Li2O, Li2O residue and NMC single crystals from vapour study was conducted through EDS mapping using the same SEM system. After completing SEM/EDS analysis, samples adhered to carbon tape atop the SEM sample stage were repeatedly soaked with a few drops of water until the pH reached 7, then dried at 60 °C under vacuum. These washed samples were re-imaged using SEM/EDS to assess the presence of Li2O and to further analyse the morphology and elemental composition. Additional SEM/EDS data were obtained using a Thermo Scientific Helios 5 Hydra DualBeam (plasma focused ion beam SEM). TEM samples were prepared with the FEI Helios NanoLab 600 DualBeam system. Selected cathode particles underwent a lift-out process, followed by thinning at 30 kV, 5 kV and 2 kV to polish and clean the surface. TEM observations were conducted using a 300 kV FEI Titan monochromated (S)TEM equipped with a probe aberration corrector and a Spectra Ultra STEM featuring an Ultra-X EDS system. EDS elemental mapping involved scanning the same region multiple times with a dwell time of 0.1 ms and an electron probe current of approximately 40 pA. Electron energy-loss spectroscopy experiments were carried out using a Gatan GIF-Continuum spectrometer.

In situ heating SEM characterization

When compared with conventional gas-assisted sample treatment carried out in the whole SEM chamber, the MicroReactor developed by Thermo Fisher allows for a well-controlled gas environment in the sample area up to a temperature of 1,200 °C obtained via a microelectromechanical system heating device38. Due to the pressure-limiting aperture in the reactor cap, the sample area is not influenced by residual gases (air, water) present in the SEM chamber, and local overpressures up to 500 Pa can be achieved in the reactor with low inlet flows via the gas inlet line. Feeds are premixed via a gas feeding system outside the microscope, giving flexibility in mixture compositions. The aperture in the cap enables SEM imaging inside the reactor during the operation of the MicroReactor. Dry powder samples (TMO precursor and Li2O) were placed on the microelectromechanical system heating chip. The sample was subjected to heating ramps at 10 °C s−1 under a variety of partial pressures and atmospheres while monitoring the surface morphology evolution.

In situ XRD test

The in situ time-resolved XRD experiment for single-crystal NMC811 synthesis was conducted at beamline 28-ID-2 (XPD) of the National Synchrotron Light Source II, using an image plate detector in transmission mode14. The experiment utilized a wavelength of 0.182 Å. Approximately 3.5–4.0 mg of the TM(OH)2811 and Li2O precursors was loaded into a quartz capillary with an inner diameter of 0.7 mm and open ends on both sides. One end of the capillary was connected to an oxygen carrier gas system and the other to an exhaust. A flow meter controlled the oxygen gas flow rate, and quartz wool was positioned at each end of the sample to stabilize it during gas flow. The data collection duration for each XRD pattern was around 1 min. The heating process involved (1) a ramp of 5 °C min−1 from room temperature to 500 °C, (2) a hold at 500 °C for 7 h, (3) a ramp of 5 °C min−1 to 850 °C, (4) a hold at 850 °C for 7 h and (5) cooling at 5 °C min−1 back to room temperature. The collected in situ XRD patterns of the NMC811 evolution during heating were sequentially refined using the Rietveld method with the software TOPAS.

Density functional theory modelling

For a specific Li2O facet at a certain temperature (T) and pressure (P), the surface free energy is determined by39,40

$$\gamma \left(T,P\right)=\frac{1}{2A}\left[{G}_{\mathrm{slab}}\left(T,P,{N}_{\mathrm{Li}},{N}_{\mathrm{O}}\right)-{G}_{\mathrm{bulk}}\left(T,P\right)-\left(\frac{1}{2}{N}_{\mathrm{Li}}-{N}_{\mathrm{O}}\right){\mu }_{\mathrm{O}}\left(T,P\right)\right]$$
(2)

where γ(T, P) represents the surface free energy at specified T and P, A is the area of the Li2O facet, Gslab is the free energy of the Li2O facet slab and Gbulk is the free energy of bulk Li2O. NLi and NO are the numbers of lithium and oxygen atoms in the slab, respectively. μO(T, P) is the chemical potential of oxygen, calculated using41

$${\mu }_{\mathrm{O}}\left(T,P\right)={\mu }_{\mathrm{O}}\left(T,{P}^{0}\right)+1/2{k}_{\mathrm{B}}T\,\mathrm{ln}\left(\frac{P}{{P}^{0}}\right).$$
(3)

Here, kB represents the Boltzmann constant, and P0 is the standard atmosphere pressure. Max\([\mu _{\mathrm{O}}\left(T,{P}^{0}\right)]=1/2{E}_{{\mathrm{O}}_{2}}^{\mathrm{total}}\). Assuming that the proportion of facets remains constant in the particle, the surface free energy of the particle is estimated by summing the weighted average surface free energies of the facets. Details of the facets and the weighting of these facets were derived from experimental XRD data (Supplementary Fig. 5c).

To determine the internal energies at 0 K, density functional theory calculations were executed using the Vienna Ab Initio Simulation Package (VASP). Electron–ion interactions were modelled using projector-augmented wave pseudopotentials42, with a cutoff energy set at 400 eV. The exchange–correlation functional was represented using the Perdew–Burke–Ernzerhof generalized gradient approximation43. The exchange–correlation functional with a Gaussian smearing width term of 0.05 eV was used. The convergence criteria for the electronic self-consistent iterations were set to 1 × 10−7 eV and 1 × 10−5 eV Å−1. The Monkhorst–Pack k-point mesh was selected, adhering to the condition ka ≥ 25 Å, where a represents the length of the basis vector and k the number of k-points in any given direction. For the simulation of various Li2O facets, slab models consisting of six to eight layers were utilized, on the basis of the convergence test detailed in Supplementary Fig. 5c. The central two layers of the slabs were kept fixed during simulations to ensure stability. To eliminate interactions between the slab atoms and their periodic images, a vacuum thickness of 16 Å was maintained in the z direction.

The free-energy calculations were conducted using the quasiharmonic approximation through the phonopy software, which utilizes finite displacements44. In all phonon calculations, a non-analytic correction was applied to the dynamical matrix via Ewald’s method, incorporating the dielectric tensor and Born effective charges derived from density functional perturbation theory calculations. Further details of model validation and additional discussions are available in Supplementary Note 1.