Sequential delithiation behavior and structural rearrangement of a nanoscale composite-structured Li1.2Ni0.2Mn0.6O2 during charge–discharge cycles

Lithium- and manganese-rich layered oxides (LMRs) are promising positive electrode materials for next-generation rechargeable lithium-ion batteries. Herein, the structural evolution of Li1.2Ni0.2Mn0.6O2 during the initial charge–discharge cycle was examined using synchrotron-radiation X-ray diffraction, X-ray absorption spectroscopy, and nuclear magnetic resonance spectroscopy to elucidate the unique delithiation behavior. The pristine material contained a composite layered structure composed of Ni-free and Ni-doped Li2MnO3 and LiMO2 (M = Ni, Mn) nanoscale domains, and Li ions were sequentially and inhomogeneously extracted from the composite structure. Delithiation from the LiMO2 domain was observed in the potential slope region associated with the Ni2+/Ni4+ redox couple. Li ions were then extracted from the Li2MnO3 domain during the potential plateau and remained mostly in the Ni-doped Li2MnO3 domain at 4.8 V. In addition, structural transformation into a spinel-like phase was partly observed, which is associated with oxygen loss and cation migration within the Li2MnO3 domain. During Li intercalation, cation remigration and mixing resulted in a domainless layered structure with a chemical composition similar to that of LiNi0.25Mn0.75O2. After the structural activation, the Li ions were reversibly extracted from the newly formed domainless structure.

www.nature.com/scientificreports www.nature.com/scientificreports/ studies provided no clear indication of which structural model appropriately represents the real structure. Many atomic column observations using advanced scanning transmission electron microscopy (STEM) have provided evidence of two structural domains in the composite structure [16][17][18][19] , while other reports showed a homogeneous atomic column, suggesting a single-phase structure [20][21][22] . Solid-state nuclear magnetic resonance (NMR) spectroscopy is sensitive to cation substitution in the first and second cation coordination shells, allowing the local structure to be examined at a length scale of <5 Å in diameter. Grey et al. reported the nonrandom cation distribution around Li ions in Li[Li (1-2×)/3 M x Mn (2-x)/3 ]O 2 via 6 Li magic-angle spinning (MAS) NMR analyses, implying a composite nature instead of homogeneous solid solution [23][24][25] .
Lithium deintercalation causes continuous changes of the a and c lattice parameters of LMRs [26][27][28][29] . These changes are closely associated with the charge compensation mechanism, which is reflected in the charge-discharge profile. The initial charge process exhibits a characteristic voltage profile, with a potential slope of up to ~4.5 V and subsequent irreversible plateau at ~4.5 V 1, 26 . Based on X-ray absorption spectroscopy (XAS) studies, charge compensation can be achieved by the TM redox couple (e.g., the Ni 2+ /Ni 4+ couple) during the voltage slope 19,[30][31][32] , corresponding to the shrinkage and expansion of the a and c parameters, respectively [26][27][28][29] . In contrast, it is generally accepted that charge compensation can be achieved by oxygen (O 2− /O − couple and/or O 2 release) during the potential plateau, where the lattice parameters remain almost unchanged [26][27][28][29][32][33][34][35][36][37] . Oxygen removal is consistent with an irreversible capacity loss observed during the initial charge-discharge cycle.
In terms of delithiation behavior of the single-phase structure model, Li ions should be uniformly extracted from the entire structure because the Li environment is homogeneous. In contrast, in the composite structure model, it would be expected that the Li ions are first extracted from the Ni 2+ -bearing LiMO 2 domain at the potential slope, and subsequently from the Li 2 MnO 3 domain at the plateau region 2,3 . However, previous studies have shown that Li ions are deintercalated from the Li 2 MnO 3 -like environment at potential slopes of <4.5 V 24,38 . Recently, well-considered NMR shift assignments were reported for Li 1.2 Ni 0.18 Mn 0.61 Mg 0.01 O 2 , where the local cationic configurations were compatible with the characteristic 7 Li MAS NMR signals while satisfying local electroneutrality constraints 39 . The structural evolution based on the changes in NMR signals during the first electrochemical cycle was also discussed, but the mechanism remains a source of debate due to the complicated spectral deconvolution including spinning sideband (SSB) manifolds in that study. Magic-angle turning phase-adjusted sideband separation (MATPASS) is a recently developed NMR technique and its sheared projection, pj-MATPASS spectrum, can provide well-resolved isotropic signals without the SSB manifolds 40 . Subsequent studies have reported the delithiation/lithiation dynamics of different Li sites in various LMR cathodes (Li 2  These studies suggested that the Li ions in the TM layer were preferentially extracted during the first 20% of the charging and Li ions in both the Li and TM layers were subsequently removed at similar rates 41 . However, the detailed spectral evolutions of each signal component in the pj-MATPASS spectra were not discussed. Herein, synchrotron-radiation XRD, soft XAS, and pj-MATPASS NMR techniques were used to examine the delithiation behavior of a composite-structured Li 1.2 Ni 0.2 Mn 0.6 O 2 in a semi-quantitative manner, and the sequential delithiation from the composite domains during the initial charge was demonstrated. The structural evolution after the initial charge-discharge cycle was also briefly discussed.

Results and discussion
Structural evolution of Li 1.2 ni 0.2 Mn 0.6 o 2 during the 1 st charge-discharge cycle. Figure 1 shows the charge-discharge profiles of Li 1.2 Ni 0.2 Mn 0.6 O 2 for the 1 st , 2 nd , and 20 th cycles (complete profiles and corresponding dQ/dV curves are provided in Fig. S1). Sampling points for the structural analyses are clearly marked. Figure 2(a) shows the SR-XRD profiles for the 1 st charge-discharge cycle. Based on single-phase notation, the pristine material was indexed with a space group of C2/m and the lattice parameters were refined as a = 4.951 Å, b = 8.558 Å, c = 5.028 Å, and β = 109.213° (#1) 42,43 , although the high-angle annular dark field (HAADF)-STEM image of the material showed a composite structure, where the atomic columns of the TM layer of Li 2 MO 3 and LiMO 2 are clearly discernible ( Fig. S2(a)) 42 . Close inspection of the strongest 001 diffraction peak provides clear suggestions regarding average crystal structure modification during the delithiation/lithiation process ( Fig. 2(a), inset). The 001 peak moves to a lower 2θ value during charging to 100 mA h g −1 (potential slope region, #2 and 3). At 200 mA h g -1 , at a midway point of the potential plateau region, the peak width broadens with a position close www.nature.com/scientificreports www.nature.com/scientificreports/ to that at 100 mA h g -1 (#4). Subsequently, the peak splits into two separate peaks at 4.8 V (#5), which then merge into a single peak during the lithiation, indicating that the structural modification is reversible (#6 and 7). A similar type of peak splitting at a high potential has been reported by Koga et al., who attributed the two peaks at lower and higher 2θ to bulk and degraded surface phases, respectively 44 . Herein, the peak at higher 2θ is close to the strongest 111 diffraction peak of the Li-poor cubic spinel Li 0.05 Mn 2 O 4 (λ-MnO 2 , a = 8.0445 Å) 45,46 . The formation of a spinel structure upon charging was substantiated via electron microscopy techniques 16,47,48 . The superlattice peaks characteristic of C2/m were observed for the pristine material in the 2θ range of 6.3°-11.0°. The intensity of these peaks gradually decreased during the delithiation process, which was mostly irreversible (Fig. S3). This indicates that the Li/TM honeycomb ordering in the TM layer was permanently lost by Li extraction from the original structure.
Rietveld refinements with the space group of R-3 m (Fd-3 m for the second phase at 4.8 V) were performed for simplicity, although the refinements with the C2/m structure may yield better Reliability factors (R p and R wp ). For the R-3 m structure, the oxygen occupancy at the 6c Wyckoff position and the TM ion occupancy in the Li and TM layers (3a and 3b positions, respectively) with fixed isotropic atomic displacement parameters were refined in addition to the a and c lattice parameters. For the Fd-3 m structure, only the a lattice parameter was refined and the superlattice peaks in the 2θ range of 6.3°-11.0° were omitted during the refinement procedures. The final R p and R wp values were <4.6% and <7.0%, respectively, for all samples. Figure 2(b) shows the variations of the a and c parameters during the 1 st cycle where a decreases and c increases over the potential slope region. The contraction of the a-axis and elongation of the c-axis are a consequence of Ni oxidation, forming smaller ionic radii and increasing electrostatic repulsion between the close-packed oxygen layers due to Li extraction [26][27][28][29]44 . In contrast, the variations of the lattice parameters were negligible during the potential plateau region, suggesting charge compensation by the oxygen redox mechanism [26][27][28][29]44 . The nanoscale phase separation from the composite layered structure to layered and spinel(-like) structures was observed only at 4.8 V and the mass ratio of the two phases was estimated as 66:34 from the refinement. Recently, it was reported that phase separation at high voltage is dependent on particle size 29 where two-phase behavior was observed in samples with large particles (10 μm), but one-phase behavior was found for samples with sub-micrometer particles. Herein, the sample exhibited a ~5 μm secondary particle size, so phase separation was observed. During the lithiation process, the a parameter largely increases, while the c parameter slightly decreases. The former is associated with reduction of the TM ions, and the latter is likely due to incomplete lithium uptake. Figure 2(c) shows the TM ion occupancy variations in the Li layer for the R-3 m structure, suggesting increased TM migration into the Li layer after activation. Spectroscopic evidence of the migration was also provided by resonant X-ray diffraction spectroscopy (RXDS) 49 . The variation of oxygen occupancy is also shown in Fig. 2(c). Oxygen atoms exhibit relatively weak X-ray scattering power compared to TM atoms, but the high resolution and high flux X-ray beam provided a general trend for oxygen occupancy in the refinement results. The oxygen occupancy remained nearly constant during the potential slope region, but decreased in the plateau region, suggesting oxygen loss from the layered structure. The oxygen loss and TM migration was correlated as the oxygen vacancy may facilitate TM migration 50 . These results indicate that the layered structure retained at 4.8 V is highly defective, which may transform into a spinel-like structure with further delithiation. The oxygen occupancy is partially recovered during lithiation. A similar behavior was previously reported, and it is considered that the oxygen vacancies are partly refilled by extracting the oxygen from surface film components (for example, Li 2 CO 3 ) on the particle, which are caused by electrolyte decomposition 51 . Alternatively, the peroxide moiety formed in the charged material may account for the reversibility of oxygen occupancy 51 . Such O-O dimer species locate at the interstitial positions, which are not explicitly taken into account for structural refinements. Li reinsertion breaks the dimer bond, and the oxygen atoms may go back to the 6c site in the discharged material. A possibility of the formation of peroxide species is discussed in the next section. www.nature.com/scientificreports www.nature.com/scientificreports/ Figure 3 shows the Ni, Mn L II,III -, and O K-edge XAS spectra acquired in partial fluorescence yield (PFY) or inverse partial fluorescence yield (IPFY) mode for the 1 st cycle. The TM L-edge XAS directly probes the electron dipole transition from the 2p core level to 3d unoccupied states and is sensitive to the valence and coordination states. The L-edge spectra split into L III and L II lines due to the spin-orbit interaction of 2p 3/2 and 2p 1/2 core electrons. The Ni L III -edge spectrum of the pristine sample (#1) was similar in shape to those of NiO and LiNiO 2 32,52,53 . The average Ni valence was determined to be 2.3+ from the Ni K-edge XAS spectrum 49 . At the end of the potential slope region, the lower energy peak of the L III -edge decreased in intensity while the higher energy peak increased, with the L II -edge peak shifting to a higher energy, indicating an increase in Ni oxidation state (#3). However, the higher energy peak of the L III -edge decreased in relative intensity, while the L II -edge peak shifted back to a lower energy at the potential plateau, suggesting partial reduction of Ni (#4 and 5). The L III -edge spectrum of the sample discharged to 2.0 V (#7) was almost identical to that of NiO, indicating a valence state close to 2+. From the spectral analysis by the linear combination fitting, in which it was assumed that spectra #3 and #7 are representative of Ni 4+ and Ni 2+ states, respectively, the Ni valence state was estimated to be 3+ at 4.8 V (Table S1). The decreased Ni valence state at high voltages has been previously reported by several researchers 32,54 . Hy et al. suggested that this reduction behavior was associated with the preferential hybridization of Ni with activated oxygen species formed at high potentials 54 . The Mn L III -edge spectrum of the pristine material resembled that of Li 2 MnO 3 54-56 , indicative of the tetravalent state, as expected ( Fig. 3(b)). The oxidation state of Mn remained almost constant during charging, in contrast to previous studies of Li 2 MnO 3 , where the valence state was reduced to 3.5+ at 4.8 V 55,56 . This also departs from the decreased Ni valence mentioned above. Ni has a larger electronegativity and is expected to strongly hybridize with the activated oxygen species preferentially over the Mn ion 54 . In contrast, Mn valence decreased to 3.5+ at 2.0 V, based on the spectral analysis using reference spectra of MnO (Mn 2+ ), Mn 2 O 3 (Mn 3+ ), and Li 2 MnO 3 (Mn 4+ ) ( Table S1). The O K-edge spectra showed pre-edge peaks between 527 and 534 eV and a white line above 534 eV, which come from the O 2p unoccupied states hybridized with TM 3d and 4 s,p states, respectively (Fig. 3(c)). Detailed examination indicated that the spectral intensity of the pre-edge area changes in two steps during charging and its intensity evolution between 525 and 534 eV is plotted in Fig. 3(d). The intensity increased during the potential slope but remained nearly constant over the plateau region. The difference spectra between spectra #1, #3, and #5 are provided in Fig. 3(e). An intensity increase at the lower energy side (~528 eV) of the 529 eV peak was observed until the end of the potential slope 36 . This was also observed in the O K-edge spectrum for Li 1-x MO 2 and was attributed to the increased contribution of the ligand hole in the LiMO 2 domain 54,57-61 . This spectral change is associated with a valence increase of Ni ions www.nature.com/scientificreports www.nature.com/scientificreports/ hybridized with O ions, suggesting a cation-anion dual charge compensation process 43 . The 528 eV component decreased at 4.8 V, and a new peak at ~530 eV increased in intensity. The former behavior is due to the valence decrease of Ni, as shown in the Ni L III -edge spectra, while the latter is unrelated to the valence change of the TM ions in the LiMO 2 domain and is instead related to Li 2 MnO 3 domain structural changes. Previous studies reported a similar intensity increase at 530 eV for delithiated Li 2 MnO 3 55,56 . This is consistent with the assertion that Li ions are extracted from the Li 2 MnO 3 domain over the potential plateau 2,3 . The intensity increase at 530 eV likely corresponds to the formation of peroxide-like oxygen in the Li 2 MnO 3 domain, because the peak position is similar to that of Li 63 . Upon discharge, the spectral features in the pre-edge region were similar to those of pristine material, indicating that the electronic states of O ions were mostly recovered.  64 . The 791 ppm component was tentatively assigned to the 7 Li signal from the Ni-containing LiMO 2 domain due to the lack of the corresponding 7 Li signal from the TM layer. Some assignments differed from those in a recent study 39 , where the peaks at 576 and 946 ppm (corresponding to the peaks at 553 and 791 ppm observed herein) were assigned oppositely. The remaining 352 ppm component was not observed in previous studies 39,64 , and was tentatively ascribed to Li ions in the Li layer of the Li 2 [Ni 2/6 Mn 4/6 ]O 3 -like domain. This contribution is negligible and is merged with the discussion of the 553 ppm peak below. Finally, the sharp signal at 0 ppm arises from impurity phases such as Li 2 CO 3 and organic lithium salts.
The HAADF-STEM images visualized that the atomic columns of the TM layer characteristic of Li 2 MO 3 and LiMO 2 -like structure were alternately stacked over several nm (Fig. S2(a)) 42   www.nature.com/scientificreports www.nature.com/scientificreports/ the Ni ions are tetravalent in the Ni-doped Li 2 MnO 3 -like domain with Ni:Mn = 1:5 (see also Supplementary Information). The estimated phase ratio is comparable to the composite notation of Li  39 . The composition and ratio of these domains differed from those estimated in ref. 39 with a similar bulk composition, Li 1.2 Ni 0.18 Mn 0.61 Mg 0.01 O 2 , mostly due to different peak assignments. It should be noted that the composite structure of differently composed domains in Li 1.2 Ni 0.2 Mn 0.6 O 2 may introduce chemical inhomogeneity within a nanoscale particle. The inhomogeneous Ni distribution relative to Mn and O is highlighted in the energy-dispersive X-ray spectrometry (EDS) mapping images of a single particle (Fig. S2(b)). The LiNi 0.41 Mn 0.57 O 2 domain may be concentrated in the Ni-rich region.
The 7 Li pj-MATPASS and its difference spectra indicate that Li extraction from each signal component is not constant (Figs. S5 and S7). Figure 4(b) shows the integrated intensity variation of 7  . This abrupt intensity reduction during the early stages of delithiation arises from the localized nature of the electrons on paramagnetic ions and nearby Li + ions, causing rapid relaxation of all nearby 7 Li nuclear spins before signal acquisition. In contrast, the Li content in the sample at 4.8 V was slightly overestimated, indicating that the delivered charging capacity includes some contribution from side reactions such as electrolyte decomposition at high potentials. During discharging, the Li contents increased almost linearly. Figure 4(b) shows that the Li ions were extracted from the Li and TM layers of Li [Li 0.17 Ni 0.21 Mn 0.59 ]O 2 . The Li TM /Li Li ratio remained almost constant during charging, but decreased gradually during discharging (Fig. S9). This suggests that the Li ions cannot easily be reinserted into the TM layer of the discharged material.
The Li content in each domain during the 1 st charge-discharge cycle is plotted in Fig. 4(c). The LiMO 2 domain content is drastically reduced at the 50 mA h g -1 charge, leading to the abrupt intensity reduction discussed above ( Fig. 4(b)). Therefore, it is clear that Li ions are preferably extracted from the LiMO 2 domain over the potential slope region. From the XAS results ( Fig. 3(a) and ref. 49 ), Li extraction is electronically compensated by the valence increase of Ni in the LiMO 2 domain. It should be noted that some 7 Li signals were lost in the Ni-free and Ni-doped Li 2 MnO 3 domains over the potential slope region (Figs. 4(a) and S7). This signal loss is likely fictitious, resulting from the strong interaction with the localized paramagnetic spins in the nearby LiMO 2 domain. Li ions are subsequently deintercalated from the Li 2 MnO 3 domain during the potential plateau with charge compensation via oxidation of lattice oxygen and O 2 release, as described above. At 4.8 V, a broad 7 Li signal remained at ~600 ppm ( Fig. 4(a), #5), suggesting that Li ions are mainly retained in the disordered Li layer of the Ni-doped Li 2 MnO 3 domain. It should be noted that some researchers have attributed the broad signal at 500-600 ppm to Li ions at tetrahedral sites 38,64 . Considering the phase separation observed in the XRD profile at 4.8 V (Fig. 2(a) The 7 Li signal at 1514 ppm is relatively sharp and well separated from other components, allowing the peak position changes to be confirmed with high confidence (Fig. S10). Over the potential slope region, the peak position shifted towards lower frequencies corresponding to the Ni valence increase. The valence state of Ni in the LiMO 2 domain changed from Ni 2+ (S = 1) to Ni 4+ (S = 0), reducing the Fermi contact interaction and shifting the peak position to lower frequencies. In contrast, no peak shift was observed over the potential plateau region, indicating that the TM ions were not involved in charge compensation. During the discharging process, the peak position returns, to some extent, to higher frequencies due to TM ion reduction. These changes are qualitatively consistent with that of the a lattice parameter.
Therefore, it can be concluded that the irreversible potential plateau in the initial charge-discharge cycle corresponds to a structural change (i.e., structural activation) from a composite structure of Li 2 MO 3 and LiMO 2 domains to a single-phase (domainless) structure. This is associated with cation mixing (i.e., migration of TM ions) during the formation of a spinel-like structure at a high potential. Upon back-transformation to a layered structure during discharging, the TM ions move back to positions different from their original positions in the TM layer, and the TM ions are uniformly dispersed, resulting in a domainless structure. Structural evolution after the initial charge-discharge cycle. The structure of Li 1.2 Ni 0.2 Mn 0.6 O 2 after the initial charge-discharge cycle will be briefly discussed. The charge-discharge profiles for the 2 nd and 20 th www.nature.com/scientificreports www.nature.com/scientificreports/ cycles are shown in Fig. 1. The capacity fade was small but gradual voltage fading was observed up to the 20 th cycle (Fig. S1). The charging and discharging capacities at the 20 th cycle were 218.8 and 216.8 mA h g -1 , respectively.
The SR-XRD profiles of the fully charged and discharged samples at the 1 st , 2 nd , and 20 th cycles are shown in Fig. 5(a). The diffraction profiles of the 2 nd charged and discharged samples (#9 and 10) are quite similar to those of the 1 st cycle (#5 and 7). Even at the 20 th charge-discharge cycle, the 001 diffraction peak of the charged sample was split (see Fig. 5(a), inset), merging into a single peak for the discharged sample (#11 and 12). We note that each peak at lower and higher 2θ of the split 001 diffraction peak in the 20 th charged sample was broader and narrower, respectively, than the corresponding peaks in the 1 st and 2 nd charged samples. This may indicate the decrease in domain size (or crystallinity) of layered structure phase and the increase in domain size (or crystallinity) of spinel phase. In addition, a small shoulder at the lower 2θ side of the 001 diffraction peak in the 20 th discharged sample may be a remnant peak of the layered structure observed in the charged state. These results indicate that the structural changes are reversible after structural activation during the initial cycle. This structural stability is significantly different from that of Li 2 MnO 3 , which changes from a layered structure to a highly disordered spinel-like structure in 20 cycles 56 . It should be noted that the spinel-like phase was observed at 4.8 V in the 20 th cycle, although the Li 2 MnO 3 domain completely disappeared, as discussed below. Therefore, this spinel-like structure results from the domainless layered structure at the high voltages 68 . Figure 5(b) shows the lattice parameters changes of the pristine and discharged samples at the 1 st , 2 nd , and 20 th cycles. The domain structure in the pristine material changes to a domainless structure after the 1 st discharge and both the a-and c-axes were correspondingly expanded. The lattice parameters further increased for the 2 nd and 20 th discharged materials. The extent of TM ion migration (to the Li layer) and oxygen site occupancy remained nearly constant after the 1 st cycle (Fig. 5(c)). Therefore, the increases in a-and c-axes are likely caused by the decreased average oxidation state of TM ions and decreased Li content, respectively. Figure 6 shows the Ni and Mn L II,III -edge XAS spectra for the 1 st , 2 nd , and 20 th charged and discharged samples. The Ni valence was divalent for all discharged materials. The Ni valence was estimated to be 3.0+ and 2.6+ in the 2 nd and 20 th charged materials, respectively (Table S1). The average Mn valence in the pristine and 1 st , 2 nd , and 20 th discharged states gradually decreased from 4.0+ to 3.5+, 3.4+, and 3.3+, respectively (Table S1), which is consistent with increased a-axis, indicating that Mn ions are redox active after the initial cycle. This is consistent with a previous study using hard X-ray photoelectron spectroscopy (HAXPES) 43 , while these soft XAS spectra include some contributions from the active material surface degradation. The present results suggest that Ni ions become redox inactive at higher voltages (> ~3.2 V for Ni 2+ /Ni 4+ redox couple), and Mn ions become redox active at lower voltages (<~3.2 V for Mn 3+ /Mn 4+ redox couple) during repeated cycles. This is involved in voltage fading behavior shown in Fig. S1(b). Figure 7(a) shows the 7 Li pj-MATPASS spectra of the 1 st , 2 nd , and 20 th charged and discharged samples, where comparison of the charged (#5, 9, and 11) and discharged samples (#7, 10, and 12) shows no significant differences in the spectra, except for consumption of the residual signals of the Li 2 MnO 3 component at the 20 th cycle. Therefore, the Li ions were reversibly extracted and inserted into the domainless layered structure after the initial cycle. Broad signals centered at 585 and 630 ppm were observed in the 20 th charged and discharged materials, respectively. The higher frequency isotropic shift corresponds to the lower average oxidation state of TM ions, with higher spin numbers of S = 2 and 1 (for Mn 3+ and Ni 2+ , respectively) compared to S = 3/2 and 0 (for Mn 4+ and Ni 4+ , respectively). The Li contents estimated from the signal intensities of the 20 th charged and discharged samples were slightly overestimated relative to that calculated from the charging and discharging capacities www.nature.com/scientificreports www.nature.com/scientificreports/ ( Fig. 7(b)). Li ions were observed in LiNi 0.25 Mn 0.75 O 2 in the 20 th charge (#11), although the Li content calculated from the capacity at 4.8 V should be empty. These results indicate that the long-cycled charge-discharge capacity includes significant contributions from accumulative side reactions and that NMR spectroscopy is a reliable tool for directly measuring the Li content in active materials.

conclusions
Based on the composite structure concept of LMRs, the delithiation behavior of Li 1.2 Ni 0.2 Mn 0.6 O 2 was reexamined during charge-discharge cycles. It was confirmed that the pristine material contains a composite layered structure composed of Li 2 MO 3 and LiMO 2 nanoscale domains using electron microscopy (HAADF-STEM). Solid-state NMR further provided a quantitative information about the composite structure (i.e., composition and ratio of the domains), which consists of pure Li    www.nature.com/scientificreports www.nature.com/scientificreports/ associated with TM ion migration from the TM layer into the vacant Li layer in the Li 2 MnO 3 domain. During Li intercalation into the structure, cation remigration and mixing formed a domainless layered structure, and the active material became a single Li[Ni 0.25 Mn~0 .75 ]O 2 phase with a small content of remnant Li 2 MnO 3 domain. After the 1 st charge-discharge cycle, when irreversible structural activation occurred, the bulk structure showed reversible Li extraction accompanied by reversible lattice changes and phase separation into layered and spinel-like phases at high potentials. These results provide a new insight into the delithiation/lithiation behavior of nanoscale composite-structured LMRs, which further serve as understanding of the capacity and voltage fading mechanism relating to structural degradation of these materials. experimental methods Li 1.2 Ni 0.2 Mn 0.6 O 2 was synthesized via solid-state reaction at 900 °C for 12 h with the starting materials, LiOH · H 2 O, Ni 2 CO 3 (Wako Pure Chemical Industries) and MnCO 3 (Kojundo Chemical Laboratory) 42,43 . From scanning electron microscopy (SEM) observations, the average particle size was estimated to be ~5 μm, composed of an aggregate of small crystallites (~200 nm in diameter). The chemical composition of the obtained material was determined to be Li 1.25 Ni 0.20 Mn 0.55 O 2-δ (δ ≈ 0.07) via inductively coupled plasma-atomic emission spectrometry (ICP-AES; ICPS-8100, Shimadzu) and iodometric titration measurement that estimates the average oxidation state of the transition metals.
A positive electrode was prepared using a mixture of the active material, acetylene black (Denki Kagaku Kogyo), and polyvinylidene difluoride (PVDF, Kureha) at a weight ratio of 80:10:10 spread onto an aluminum foil with 1-methyl-2-pyrrolidone (NMP) and dried at 80 °C under vacuum overnight. 43,56 The electrode was pressed to a thickness of 30-35 μm and foils of metallic lithium (0.2 mm in thickness, >99.9%, Honjo Metal) were used as counter and reference electrodes. These components were assembled with the Celgard 2500 separator and soaked in an electrolyte solution, and sealed in an aluminum-coated laminate-type cell in an Ar-filled glove box. The electrolyte solution used herein was 1 M LiPF 6 solution dissolved in anhydrous ethylene carbonate (EC) and ethylmethyl carbonate (EMC) at a volumetric ratio of 3:7 (Kishida Chemical).
The electrochemical measurements were performed at 50 °C on an automatic cycling and data recording system (HJ1001SD8, Hokuto Denko). The cells were galvanostatically cycled between 2.0 and 4.8 V vs. Li/Li + at a current rate of 20.5 mA g -1 56 . The cells were carefully disassembled at the desired charge/discharge states in the glove box and rinsed with dimethyl carbonate (DMC) to remove any electrolyte solution residue.
Synchrotron-radiation XRD (SR-XRD) measurements were performed using powdered samples sealed in 0.5 mmφ borosilicate glass capillaries in the 2θ range of 4.0-52.7° with Δθ of 0.0025° at the BL28XU beamline in SPring-8 (Hyogo, Japan). An incident radiation of λ = 0.4997 Å was used, which was calibrated with the lattice parameters of CeO 2 . The Rietveld refinements of the crystal structure parameters were performed using Jana2006 69 .
Soft XAS measurements were performed at beamline BL-11 in the SR Center, Ritsumeikan University (Shiga, Japan). The samples were placed into a sample holder in the glove box and transferred using a transfer vessel to the high vacuum sample chamber without exposure to the air. The Ni, Mn L II,III -edge, and O K-edge spectra were acquired in partial fluorescence yield (PFY) mode. Mn L II,III -edge spectra were also acquired in inverse partial fluorescence yield (IPFY) mode to account for self-absorption effects 70 . The PFY/IPFY spectra are (relatively) bulk-sensitive with a probing depth of up to ~200 nm 71 . 7 Li MAS NMR spectra were acquired using an ECA-600 spectrometer (JEOL RESONANCE Inc.) at a magnetic field of 14.1 T with a wide-bore T3 MAS probe (Agilent Technologies). The powder samples were packed into 1.6 mmφ MAS ZrO 2 rotors with airtight caps in a glove box and spun at a spinning rate of 30 kHz during the experiments. The practical temperatures of the spinning samples at 30 kHz were estimated to be ~60 °C due to frictional heating. To obtain high resolution spectra, the pj-MATPASS technique 40 was used with a π/2 pulse width of 1.1 μs and a relaxation delay of 0.1 s. All spectra were referenced to a 1 M LiCl solution at 0.0 ppm. The signal intensities were normalized by the sample weight in the rotors.