LMTMO materials are regarded as promising high-energy cathodes due to the redox of both TM-cations and O-anion 1,2,3,4. However, the performance of LMTMO is limited by oxygen release, sluggish kinetic of anionic redox, and non-equilibrium Li-ions diffusion5. Also, lattice displacement and nano-strain evolution lead to irreversible structural changes during the electrochemical redox processes6, resulting in voltage decay and disruption of Li-ions transport pathways, further exacerbating the dynamic sluggishness7. Although previous researches have revealed that the degradation of LMTMO is closely related to inherent thermodynamic instability of the host structure8, but the degradation details throughout the particle remain elusive, particularly for localized nanoscale domains.
Since both TMs-cations and O-anion play critical roles in electrochemical performance and structural degradation9, an integrated analysis on these two redox centers is essential to elucidate the complex redox mechanism, along with information on particle heterogeneities that strongly affects the degradation as well. Energy-resolved transmission X-ray microscopy could provide computed phase contrast tomography for comprehensive understanding on the local morphological and chemical mapping with 3D spatial resolution10, including heterogeneous oxidation of cations11 or anions12, cations rearrangement13, chemical composition distribution14,15, and propagation of nano-to micro-sized cracks16. However, it is challenging to integrate such a cation-anion redox determined morphological, structural, chemical and oxidation state evolution analysis, with the association of kinetic and thermodynamic effect. Here, we use transmission X-ray microscopy (TXM) in soft and hard energy region, scanning transmission electron microscopy (STEM) with high angle annular dark field (HAADF) mode and integrated differential phase contrast (iDPC) mode to systematically investigate the degradation heterogeneities from the intact particle down to the nanoscale region. Such an analysis on the representative individual particle can yield comprehensive understandings on the degradation mechanism during the electrochemical process. Our work identified distinct degradation pathways associated with different intra-particle heterogeneous reactions. The degradation of LRTMO at low rate originates from substantial O-defects formation across the particle, which also releases oxygen and triggers progressive phase transformation from the surface into the bulk. Differently, the fast kinetics during ultra-fast Li- (de)intercalation with heterogeneous Li-ion diffusion, lead to O-distortion dominated lattice displacement and couple with TM-ion dissolution and Li-site variation. The revealed degradation mechanisms in this work brings novel insights for the (future) design of LRTMO materials.
Electrochemical properties and atomic visualization on structural degradation
The electrochemical performance of LRTMO is shown in Fig. 1a, b. The O-loss at low current rate (0.1C) causes first-cycle Coulombic inefficiency and voltage decay in prolonged cycles. In contrast, the fast kinetics at a high current rate (10C) exacerbates initial irreversibility, amplifies overpotential, and intensifies capacity fading. To investigate the structural evolution differences with different current rates, cross-sectional STEM was employed to direct probe local ionic structure along the [010] zone axes. As shown in Fig. 1c, d, coupling STEM-HAADF and iDPC image, the ordered atoms are uniformly distributed in TM-O coordination with a well-organized layer structure in pristine sample (Fig. 1e), and Li-columns in the middle of O-slabs correspond to the typical octahedral sites. After 20 cycles at 0.1C, typical layer to rock-salt phase transformation17 on the surface is observed in Fig. 1f, whereas the presence of heterogeneously distributed nanoscale dark areas in bulk lattice are nanovoids formed as results of O-vacancies generations. Prior research has proven that oxidation of O2--anion lowers both the formation energy of O-vacancies and migration barrier of oxidized-O18, thereby providing the energetic driving force for the oxygen gas release from the surface19,20 and O-defects aggregation in the bulk. Nearly 4% of total O-loss occurs in the 1st cycle and remains at consistent level after 20 cycles, as confirmed by neutron powder diffraction (NPD) shown in Fig. S1c and d. Although it is well known that structural irreversible rearrangements aggravate lattice degradation and lead to serious voltage decay, further characterization on the O-chemical diffusivity and its heterogenous distribution originating from the initial electrochemical process (activation) is necessary to elucidate the corresponding mechanism.
As the current increased to 10C, kinetic and thermodynamic effect cause high overpotential associating with inhomogeneous ion diffusion. The occurrence of nanoscale slab twisting and bending results in further distinct structural rearrangement pattern. In detail, the mass-sensitive STEM-HAADF image shown in Fig. 1h demonstrates lattice distortion with varied interlayer distance along c- direction and confirmed TM-deficient in the yellow dashed rectangle region with sharp contrast, which differs from the intermedia phase observed by fast charging LiNi1/3Mn1/3Co1/3O2 material21. This could relate to the early stage of grain boundary formation induced by a highly strained lattice (as shown in Fig. S15, the geometrical phase analysis (GPA)), which promotes the propagation of crack and TMs dissolution22. The corresponding STEM-iDPC image (Fig. 1i) shows that lattice displacement domains are randomly incorporated into the original layered framework while sharing the coherent structure. More details on the atomic configuration are shown in Fig. 1j, k, l. The position of some O-anions and residual Li-ions within dashed rectangles deviates from the original site due to non-uniform Li-ions diffusion, which indicates strongly lattice distortion and Lioct to Litet occupation changes. Since the tetrahedral occupation of Li-site is assumed energetically favorable21, it is possible that some Li-ions are stabilized or trapped in these sites, thus impending their diffusion capability9,23. As a result, a lower proportion of Li-ions can be reversibly inserted/extracted from Li layer in host structure at fast cycling (Fig. S1a, b). The displacement of O-position attributes to a large variation of the TM-O bond length, which is manifested as large-scale slab twisting and bending. Thus, such substantial Li-ion diffusion heterogeneity and gradually accumulated nanoscale tensile strains have a great impact on the structural and chemical evolution in the particle, leading to different degradation pathways of LRTMO.
Visualizing the morphological and chemical-structural evolution with 3D reconstruction
To understand the distinct degradation pathways in LRTMO, synchrotron based soft TXM was carried out to directly visualize the spatial morphological, chemical-structural and oxidation state-related sample particles at a various voltage as circled in Fig.1b. In order to achieve sufficient transparency to soft X-rays and preserve the secondary structure of LRTMO, material was slowly milled to reduce the particle size, and particles located at the top of the composite electrode were selected for further characterization (cf.: Methods in Supplementary Info). As shown in Fig.2a, the colored segments in 3D tomography and cross-sectional slices represent the distribution of chemical species with various signatures, while the categorizing criteria is based on the quantification of TM/O. The reconstructed image is achieved through computed correlation analysis of the absorption coefficient as a function of the X-ray energy (L-edges for TMs and K-edge for O) within imaged nanoscale area (Table S1). For pristine sample, 3D tomographic imaging illustrates that most of the surface areas is shown in blue (the localized region with target composition), while the existence of randomly distributed mixed colors stands for chemical heterogeneities that deviates from the original composition (amount of Co and Ni ions with 0.5%less than standard composition). The representative cross-section image through the 3D volume presumably suggests that a slight nanoscale deficiency of O is only present on the surface, whereas similar behavior could also be observed in the regular NCM material via TMs migration10. Besides that, the presence of randomly distributed red points on the surface indicate oxygen defects, which is originated from the annealing process24.
Through systematic observation of the LRTMO particle after charging with a current rate of 0.1C, we identified that agglomerated O-poor (red) segments initiated simultaneously from both surface and bulk, as the voltage goes up to 4.6V. Appearance of new O-poor segments indicates the existence of domains with higher TMs/O ratio, which highly depends on their location. They could be contributed by O-defects remaining in the lattice, substantial oxygen release from surface, and trapped oxygen in the bulk that may involve in following redox reactions19 and be slowly released from the surrounding secondary structure25. Meanwhile, the overall lattice parameter c shown in Fig. S1e first expands as Li-ions get extracted from the host structure, and then starts to decrease because O-defects release lattice strain. O-loss domains continue to accumulate in the high voltage plateau, which is consistent with the analysis on the O-occupancy by NPD (Fig. S1c) and mRIXS chacterzation19,26. At the end, a dramatic drop of the c value occurs due to the formation of substantial O-defects and TMs-dissolution (mainly for Mn) or in-plane migration that widely distributes throughout the particle (Fig. 2b) with reduced TM/O ratio20. As Li-ions insert back into the particle, the proportion of O-poor segments remain at the same level as with the fully charged state, suggesting that irreversible reactions mainly occur during the charge process and permanently change the structure and composition of LRTMO. Although operando differential electrochemical mass spectrometry (DEMS) has confirmed that oxygen gas release mainly happens during the high voltage plateau of the initial charge process 27, but the distribution of O-defects in the particle and their corresponding activity are not elucidated. With integrated spatial chemical composition analysis, the location of O-defects and insights for the structural evolution pathway are clearly identified. Most of the O-defects are already formed at the end of charge process in 1st cycle, and oxygen gas are released with phase transformation triggered on surface. However, some defects remain in the bulk and continue participating in the redox reaction during following cycles. These locally diffused O-defect in the bulk gradually transformed into the nanovoid that penetrates into the deeper region of the particle18, resulting in an ongoing expansion of the lattice structure as cycles prolonged. Overall, the correlation between morphological, chemical, structural evolution and electrochemical performance of LRTMO is systematically established.
Different from O-loss dominated structural evolution that occurred at 0.1C, fast charging of LRTMO induced an inhomogeneous electrochemical kinetic and non-equilibrium ionic diffusion dynamic, resulting in a distinct degradation pathway. Gray color in Fig 2.a stands for O-rich segment, which indicates the TMs/O ratio is lower than target composition. This is possibly contributed by the combined factors of TMs-dissolution, O-distortion, and Li-O sliding, as demonstrated in Fig.1h and i. Although only a few numbers of O-rich nano domains appear at the subsurface when the voltage reach to 4.4V, it indicates that lattice displacements have already started due to the heterogeneous reactions, and may also relate to the diffusive O species (such as more mobile O22- or O- than O2-). Inhomogeneous Li-ions diffusion arises large overpotential throughout nanoscale particles, which associates with the formation of O-rich segments, trapped Li-ions in tetrahedral sites, and hinders further Li-ions diffusion. Therefore, it provides solid evidence to support the conclusion that fast charging could lead to early activation of O-anion heterogeneously9. As the voltage goes up to 4.8V, widely distributed nanoscale O-rich domains become interconnected with larger size, occupying around 17% space throughout the particle (Fig.2c). The confined coherent lattice displacement gradually expands and then acquires cumulated tensile strain in delithiated LRTMO, which corresponds to a higher c value without any oxygen release to lower lattice strain. After fully lithiation, a portion of O-rich domains could return to their original composition. The recovery of O-rich domains suggests that they are primarily caused by lattice distortion, and thus can be partly recovered by reversible lithiation. However, a fraction of smaller size O-rich domains resists, which indicates the occurrence of irreversible TMs-dissolution, and diffusion of O species or with the mixture of residual O-displacement. Such phenomenon further influences the reversible lithiation process, which is quantified by NPD (Fig. S1a). These results indicate that the degradation of fast cycled LRTMO, from initial to prolonged cycles, is determined by the dynamic governed inhomogeneous Li-ions diffusion and regional lattice displacement-imposed strain.
Nano-probing SOC heterogeneity
As coupling the full-field imaging with the function of energy tunability from the soft X-ray scan (Table S2), two-dimensional (2D) resolved spectroscopic signals with sufficient sensitivities can be extracted from corresponding measurements28. Then the morphological and chemical features related oxidation states mapping are obtained through quantifying the local spectroscopic fingerprints. The color distribution mapping shown in Fig 3a stands for the oxidation states of TMs cations and anion, whereas the particles at representative SOCs are consistent with analysis presented in Fig.2. The color bar represents the valence states evolution of corresponding elements (conducted by overlapping reference spectra shown in Fig.S16). An inhomogeneous mixture of multi-valence states for both Mn and Ni in the pristine particle could be attributed to the nanodomain boundaries on the surface with chemical phase heterogeneity. Most TMs within the particle already reaches to their highest charged state at 4.4V with the charging rate of 0.1C. However, multi-heterogenous oxidation domains are still widely distributed at 4.6V. Although no obvious further oxidation of TMs can be observed from XAS once the voltage reaches 4.4V9,29, but the dynamic charge-compensation continues within the particle as voltage goes higher, including the anionic activation plateau. These findings indicate that non-uniform charge transfer is hard to be prohibited even at slow reaction rates. At the end of charging process (4.8V), Co undergoes uniformly oxidization throughout the particle, whereas hierarchically reduction of Mn and Ni only appears on surface, which is correlated with irreversible oxygen release30,31. Both TMs-cation and O-anion experience heterogeneous electrochemical reactions, but the oxidation pattern of O2- demonstrates a close relationship with particle geometry and size. There is an obvious valence-states gradience from surface to the deeper bulk, which may be contributed to the observed thickness dependence. The outer layer of higher oxidized O-anion exhibits stronger chemical diffusivity, and thus responsible for the oxygen-release on the particle surface. Meanwhile, the inner parts may undergo further partial transformation or activation as the form of trapped oxygen or peroxo-like species in the following cycles20, as indicated by the O-poor segments in Fig.2a. The inward growth of highly oxidized-O and reduced Mn/Ni at same location provide strong evidence for revealing the structural degradation mechanism. TMs migration is considered to be one of the consequences related to oxygen-release, and it coupled with transformation from Mn4+ to Mn3+ on the surface. During discharge process, these structural rearrangement induce Jahn-Teller distortion and stimulate the disproportionation reaction32. Interestingly, besides the reduced Mn on the surface, we also notice a small area of nano domains with the co-existence of Mn2.7+(calculated on average result) and Ni3.4+ in the same location, indicating that the electron equilibrium of TMs on particle surface may experience alternative distortion route, leaving Ni4+ or mixed with Mn4+ in lattice and dissolved Mn2+ on the surface. Apparently, sporadic distributed oxidized-O could still be identified along the particle edge, which may be related to surface reconstruction progressing into the bulk, residual non-fully reduced O due to heterogenous reaction, or formation of SEI layer via side reactions between extracted oxygen species and electrolyte33. These oxidation-state maps allow us to quantitatively evaluate or track redox reaction occurred in first cycle, and provide visible evidence that oxygen-release dominates irreversible phase transformation of LRTMO.
Regarding kinetic-dependent charge compensation mechanism, systemically analysis on the interplay of TMs cation and O-anion is also performed by collecting the corresponding oxidation state mappings under the current of 10C, correlating with investigations on the evolution mentioned in previous chapters. Both Mn and Co demonstrate similar oxidizing behavior with the charging current of 0.1C and 10C at same SOCs. However, Ni presents obvious sluggish oxidation rate and demonstrates obvious heterogeneous reaction pattern. As shown in the color-coded quantification mapping at 4.4V and 4.6V, the heterogeneous oxidation of Ni neither follows geometrical gradience nor the path of nanodomain differences. It forms distinguishable interconnected large phase areas with highly deviation on the valence-state, reflecting the strong kinetic effect on reaction heterogeneity. Ni is not completely oxidized even when the voltage reaches to 4.8V, but there is no substantial reduction happens along the edge of particle, and only a small area on the top surface is reduced with Mn. Concerning the oxidation of O-anion with charging rate of 10C, it shows lower proportion of highly oxidized O2- on the surface of particle at 4.8V. Therefore, the capacity contributed by O-anion oxidation at high voltage is substantially limited. Nevertheless, combined with the observation from the oxidation states mapping of Mn and Ni at same voltage, it can be concluded that there is neglectable phase transformation occurring on the electrode surface at the charging rate of 10C.
Turning to the end of discharge process, none of the redox involved elements return to their initial oxidation state, which are responsible for the highly irreversible capacity shown in Fig.1b. Though slight amount of Mn located on particle surface is reduced to provide extra discharge capacity, the reduction of Ni and Co is still far from their original state. A simple charge balance calculation suggests that irreversible redox of Ni and Co need to compensate for the capacity of 39 mAh g-1, where the rest irreversible capacity could be mainly attributed to the formation of new SEI or other side reactions from anode. Notably, the highly reversible reduction of O-anion indicates that the impact of O-distortion on the redox property is limited. However, the sluggish kinetic reactivity and heterogenous charge compensation further aggravate as cycle continues, associated with inhomogeneous dynamics of Li-ions to increase inner strain. Therefore, the distribution of highly irreversible valence states and the formation of aggregated Li ions are responsible for a lower first cycle Coulombic efficiency of LRTMO, and the further accumulated strain intensifies the extent structural degradation. Even though less layer-to-spinel phase transformation at the electrode surface (Fig 3.b) and the redox centers still maintain high activity (as shown in Fig 3b.c) after 20th cycles, the capacity fading of LRTMO is still more obvious when cycled at higher rates.
Initiation and growth of tensile strain
Local atomic structures have an influence on determining the macroscopic material properties, and this helps to understand material properties across various scales. In order to evaluate the internal evolution of large secondary particles on the basis of chemical composition, the ionic distribution of Li-ions and O-anion in the particle is demonstrated in Fig. 4a-c by FIB-SEM secondary electron (SE) images and corresponding TOF-SIMS maps. The cross-sectional image of the pristine sample exhibits compacted void structures at the core, which are formed during the synthesis process. TOF-SIMS quantitative analysis reveals that Li-ions and O-anion are quite uniformly distributed throughout the particle in the pristine state. Ionic concentration after operation at 0.1C is shown in Fig. S18a (4.8V in 1st cycle) and Fig. 4b (4.8V of 20th cycle), Li-ions and Mn-ions remain homogeneously distributed through the particle except for the topographic effect related intensity differences at crack edges34. This suggests relatively uniform Li-ions extraction/insertion processes within the host structure and no Mn dissolution or migration occurred in the bulk area. However, the lighter red or even white patches, which stands for lower concentration of O compared to adjacent domains, are identified from the same sample. The observed heterogeneous O-loss throughout the particle is consistent with the soft TXM observation of an O-poor area (Fig. 2a) at high voltage. Fig 4b shows clear evidence that the size of the patches increases and the contrast to neighbor components become more pronounced upon 20 cycles. Since gas is evolved as a result of chemical diffusion of O-defect in bulk, which directly leads to the formation of nano- to micro sized pores, thus the stress-level and elemental occupations are easily changed, creating and stimulating cracks at multi uneven points during the following cycles.
As the cycling current increased to 10C, Li, O and Mn demonstrate inhomogeneous distribution patterns (Fig. 4c), which are highly related to the chemical heterogeneities shown in Fig 2a. Inhomogeneous Li aggregation could be clearly identified, but the distribution pattern does not follow Fick's first law35 to form a concentration gradience along the diffusion pathway. Li heterogeneity is already formed within the first cycle (Fig. S18b), which is related to internal inhomogeneous dynamic Li-ions diffusions. The Li-site transformation aggravates the accumulation of residual “trapped” Li within the host structure, in agreement with NPD characterizations of the electrode after 20 cycles (Fig. S1b). Fewer Li-ions can be reversibly extracted and inserted during fast cycling. In addition, these regions partly overlap with O enrichment area, which could possibly be associated with higher proportion of O-distortion or Li-O intraformational sliding. Furthermore, the large O-rich areas also cover region with lower concentration of Mn, which is contributed to Mn-ions dislocation and dissolution22. EELS spectral analysis focused on two representative regions in the TOF-SIMS maps (as illustrated in Fig. 4d), and the O-rich domains can be easily identified. The pre-edge peaks in the Li K-edge spectrum are highly sensitive to local Li-ion occupancy and migration between different sites. In addition, the intensity is found to be strongly dependent on Li-ion concentration36. The pre-peak intensity differences agree well with the quantification of local Li atoms for the selected region as shown in TOF-SIMS mapping. Features in the O K-edge spectrum correspond to the fine structure of coordinated environment, especially the pre-peak at 530 eV arising from the hybridization of O 2p with TM 3d orbitals37 and local Li amount38. O-distortion alters the TM-O distance and strongly influences the Li-ions diffusion. Although the fact that lower intensity of the pre-edge feature stands for the presence of more electrons occupying the TM 3d orbitals, less Li-ions in region B leads to higher Mn valence state in the system than that of in O-rich domains. Therefore, the structural evolution is inconsistent with the regular layer to spinel phase transition behavior4.
Further investigations on the morphological difference were performed with full-field transmission X-ray phase contrast nano-tomography with a voxel value of 20nm. The 3D reconstruction shown in Fig.5a proves the existence of compact void structure at the core, whereas the samples cycled at different rates show two forms of internal pore volume. As depicted in Fig. 5b, a higher proportion of void space can be observed within the particle when the electrode is cycled at 0.1C, which could be formed by the accumulation of released gas. A visible crack propagates through the nano and micro-pores with a distribution gradience along the radial direction. Less in-quantity but larger in-size irregularly shaped pores can be observed in the 10C sample (Fig. 5c). The pores are interconnected and developed with the spreading of cracks through the particles. The cracks can be clearly observed on particles cycled at different currents, but are likely developed by two formation routes. Through modeling the constant current dis/charge process in a single LRTMO particle with micro-size pores which located close to the core (Fig. S8 and S9), the stress distribution can be simulated. As illustrated in Fig.5d and e, the depth of discharge (DOD) progresses lead to a notable tensile of hoop stress occurred at the bridge point, which makes the crack prone to nucleate at the internal void area from the inside. Although a higher number of pores distributed within the material cycled at 0.1C is observed, the stress at the internal void area of the high-rate cycling sample is even greater. The hoop stress reaches up to 3.17x108 N m-2 when discharged at 10C with 20% DOD, which is at least two orders of magnitude higher than electrode discharged with 0.1C to the same stage of DOD stage. Furthermore, the calculation on the von-Mises stress (Fig. 5f and g) demonstrates that Li-ions insertion-extraction induce large volume change, and the particle would experience a dramatic variation depending on the stress concentration, particularly for the case of high-rate cycling. The heterogenous Li-ions diffusion and non-uniform lattice distortion at high current rate increases the possibility of crack initiation and accelerate the crack growth rate through the particle.
In contrast to the fact that non-equilibrium structural dynamics induced lattice displacement/strain is considered as main reason to trigger oxygen release and TMs migration, which then aggravates degradation of LRTMO6, the mechanism of nano-strain related chemical diffusion as well as O-loss through the particle remains unclear, and there is no discussion on the kinetic and dynamic effects of structural degradation neither. What is the real degradation mechanism on the particle level, and whether there is only one route for the degradation of LRTMO remain unanswered. Herein, our work comprehensively details the intricate evolution for the degradation of LRTMO in terms of morphology, chemistry, structure and oxidation state. Also, we elucidate the interrelationships among all involved elements. With integrated analysis on TMs-cations, the O-anion and diffusible Li-ions by multi-dimensional analytical tools, heterogeneous activation, non-equilibrium irreversible reactions, and their corresponding distinct degradation pathways were identified. The scenario of O-defect, phase transformation, O-distortion, TMs-dissolution, and Li-site transformation does heterogeneously occur during the electrochemical process, but kinetic and dynamic effects are found to be the main factor determining degradation pathways. Our findings suggest that strategies to inhibit degradation of LRTMO should focus on the reaction homogeneity, lattice stability and ionic diffusivity in the host structure.
In essence, soft X-ray tomography enables visualization of consolidated chemical states in LRTMO particle with high spatial resolution, and helps to reveal insights of heterogeneous chemical reaction at representative SOCs during the electrochemical cycling process. We identified substantial quantities of O-defects or O-distortions in the initial cycle, which represents distinct degradation pathways of LRTMO. The occurrence of O-defects at slow electrochemical process dominates the degradation in the form of lattice structure transformation and nano-void injection, which are the consequences of oxygen release at the surface and local diffusion in bulk, and are mainly responsible for voltage decay. In sharp contrast with the case of O-defect driven degradation, the primary cause of capacity fading at fast cycling rates is the accumulation of lattice displacement/strain and limited ionic diffusion resulting from the O-distortion, TM-dissolution, Li-site transformation, and sluggish kinetics. The findings presented in this report are critical for accurately revealing the degradation mechanism of LRTMO, and identify the important role of homogeneous reaction kinetics and ionic diffusion in maintaining the compositional and structural stability of host materials. The insights in this study will inspire new thoughts for designing high-performance cathode with stable and efficient cationic and anionic redox process.