Reversible densification in nano-Li2MnO3 cation disordered rock-salt Li-ion battery cathodes

The electrochemical activation of Li2O in densified nano-Li2MnO3 is responsible for the large reversible capacity (290 mA h g−1) in this system.


Introduction
The increasing demand for higher energy density batteries continues to foster research on cathode materials with superior capacities to the current state-of-the-art ranging from 140 (in LiCoO 2 ) 1 to 200 mA h g À1 (in LiNi 0.8 Mn 0.1 Co 0.1 O 2 ). 2 Signicant improvement of the cathode specic capacity has been realized in Li-rich layered oxides xLi 2 MnO 3 (1 À x)LiMO 2 , 3 where M is a 3d transition metal (TM), offering extra capacities beyond the limit of conventional TM redox reactions. These composites deliver capacities as high as 250 mA h g À1 (ref. 4) when charged to a high potential ($4.5 V vs. Li/Li + ) that activates charge compensation mechanisms alternative to unfavourable TM redox reaction involving Mn 4+ oct /Mn 5+ oct in the over-lithiated Li 2 MnO 3 component.
Several studies have focused on the performance of Li 2 MnO 3 (ref. 5) to better understand this extra capacity which has been partly attributed to H + for Li + substitution [6][7][8] and more notably to oxygen redox reactions. 9,10 Anionic redox reactions in Li 2 MnO 3 are accompanied by the irreversible evolution of O 2À at the particle surface into gaseous O 2 and superoxide radicals (O 2 c À ). Part of the O 2 released is removed from the cell triggering structural reorganization, such as phase segregation (gradual transformation to spinel with cycling) 11,12 or particle densication with diffusion of transition metal ions from surface to bulk where they occupy vacancies created by Li removal. 13 Highly reactive O 2 c À radicals can react with the carbonate solvent to form Li 2 CO 3 and Li 2 O decomposition products. 14 As a result, while exhibiting excellent initial capacities, Li-rich layered cathodes suffer from voltage fade and poor rate capabilities. In spite of signicant efforts directed to avoid the irreversible evolution of O 2À and capacity fade in Li 2 MnO 3 and xLi 2 MnO 3 (1 À x)LiMO 2 composites 15 through material processing and coating, 16 this remains an issue to be resolved prior to wider application of these systems.
Interestingly, unfavourable O 2 evolution is overcome in isostructural compounds to Li 2 MnO 3 with 4d (Li 2 RuO 3 (ref. 17)) and 5d (Li 2 IrO 3 (ref. 18)) TMs, where reversible anionic redox is hypothesized to be stabilized by the formation of 'peroxo-like' 2.5Å O-O dimers. Reversible oxygen redox have also been proposed for Li 2Àx Ir 1Ày Sn y O 3 based on peroxide 1.4Å O-O dimers bridging Ir and Sn cations, which are closer to the range of bond lengths of typical peroxide species. 19 Within the last few years, superior capacities (200-350 mA h g À1 ) have been observed in Li-rich cubic cation disordered rock-salts owing to the low energetic percolation of lithium diffusion paths associated with a high Li : TM ratio. This emerging new family of cathode materials show no obvious structural changes involving complex phase transformations that could eventually lead to particle cracking and battery failure upon extensive cycling. These promising features have motivated extensive work on these materials and several Li-rich disordered rock-salts have been reported, [20][21][22] where the highest capacities have been observed in Li 1 The versatility of these phases provides further scope to explore multivalent TMs, anionic redox and mixed (O 2À /F À ) systems which may lead to improved performance.
In a previous work, we reported the reduction of the particle size in the range of 10 nm (ref. 26) in layered Li 2 MnO 3 (space group C2/m) by high-energy milling with no change of the space group symmetry. Nanostructured Li 2 MnO 3 (hereaer referred to as nano-Li 2 MnO 3 ) features improved capacity (290 mA h g À1 ) and reduced fading (200 mA h g À1 retained aer over ten cycles) 26 over its micro-structured counterpart which displays a 50% capacity drop over the rst ve cycles. 27 Upon cycling, a phase transformation into a cubic disordered MnO-type rocksalt was determined by ex situ electron diffraction of electrosynthesized Li 0.5 MnO 3 charged to 4.5 V, although no plateau characteristic of a phase transformation could be identied in the electrochemical curve.
In the present study, average and local structural studies involving combined neutron and X-ray total scattering demonstrate that, contrary to previous claims, 26 a phase transformation occurs during the nanostructuring of Li 2 MnO 3 , forming a disordered cubic MnO-type rock-salt structure analogous to that of Li 4 Mn 2 O 5 , but with nanodomains of Li/Mn layering with a size of $1 nm. The structure and charge evolution of nano-Li 2 MnO 3 during battery cycling was characterized by a combination of techniques performed in operando, including X-ray total scattering and advanced X-ray spectroscopic methods performed at the Mn K-edge, such as High Energy Resolved Fluorescence Detected near-edge X-ray Absorption (HERFD-XANES), Extended X-ray Absorption Fine Structure (EXAFS) and X-ray Emission Spectroscopies (XES) including main (kb 1,3 and kb 0 ) and valence-to-core or V2C (kb 2,5 and kb 00 ) transitions; this is the rst time these XES techniques have been reported for batteries under in operando conditions. X-ray total scattering, HERFD-XANES and EXAFS reveal a solid-solution-type response, where the reversible cycling between charged and discharged states is allowed by the breathing of a cubic Mn-framework, in discordance with a phase transformation from C2/m to Fm 3m from previous studies. 26 Note also that the structural evolution of nano-Li 2 MnO 3 during cycling signicantly differs to that of layered Li 2 MnO 3 in a crystalline state, which contracts along the clattice parameter upon delithiation. The analysis of diffraction data in reciprocal space show a large structural reversibility of nano-Li 2 MnO 3 with a constant structural coherence length of Li/Mn layering nanodomains. In the real space, no distribution of O-O interatomic distances was observed that could be ascribed to the recently reported peroxo-like distances proclaimed in other high capacity cathodes, such as in layered Li 2 MnO 3 -type structures like (Li/Na) 2 RuO 3 (ref. 28 and 29) and Li 2 IrO 3 , 18 Li 3 IrO 4 (ref. 30) and 3D structures like b-Li 2 IrO 3 . 31 The determination of the charge compensation mechanism in nano-Li 2 MnO 3 was possible by the joint analysis of HERFD-XANES and XES data. We characterized the changes of the oxidation state of Mn following the evolution of the kb 1,3 transition, and utilized for the rst time V2C transitions to probe plausible charge compensation mechanisms involving oxygen, like bulk anionic redox or hydroxyl -OH group formation. While so X-rays are typically used to probe the K-edges of light atoms, these are only sensitive to surface effects given their short penetration depth. Unlike so-XAS, the V2C spectra measured with hard X-rays ($6.5 keV) are able to probe the bulk material. Moreover, this technique only detects oxygen bound to manganese, and there is no contamination to the recorded signal from oxygen present in other components of the battery, like the electrolyte. The analysis of the data was aided by ab initio simulations of HERFD-XANES and V2C signals using proposed structural models based on different charge compensation mechanisms, and their confrontation with the measured spectra. Our data and simulations of spectroscopy signals point to a densication mechanism with an effective loss of electrochemically active Li 2 O, which produces electrons and is the main responsible for the exchanged capacity for nano-Li 2 MnO 3 . The oxygen evolved from the Li 2 O electrochemical activation could in turn be stabilized in peroxide species at the cathode surface, or by the trapping of O 2 molecules in vacancy clusters originated during Li-extraction. Interestingly, and unlike previous studies, very small capacity losses (0-3% over the rst 10 cycles) 26 could be associated with the material's densication.

Structural description of pristine nano-Li 2 MnO 3
The atomic structure of nano-Li 2 MnO 3 was investigated by joint analysis of neutron and X-ray total scattering data to analyse both short-and long-range ordering. Rietveld renement was used to determine how well reported crystallographic models matched to the observed average structure. The well-known Li 2 MnO 3 layered model (space group C2/m) failed to produce a satisfactory t to the observed data (R wp > 10%, see Fig. S1 †), which disproves previous work 26 and indicates that a structural transformation occurs during the mechanochemical synthesis of nano-Li 2 MnO 3 by the ball milling of microstructured Li 2 MnO 3 . A model based on a rock-salt MnO-type structure (space group Fm 3m) with modied occupancy values to include Li on the Mn site produced a good t to the neutron data, but several broad reections in the X-ray diffraction pattern could not be indexed with the cubic Fm 3m space group (Fig. S1 †). These reections were attributed to the formation of a shortrange ordered superstructure formed from ordering alternating layers of Li and Mn cations into alternating (111) planes in the rock-salt structure, as previously reported in Li x Ni 2Àx O 2 . 32 The chemical ordering of Li and Mn is described by the rhombohedral space group R 3m, as shown in Fig. 1. The rhombohedral model in Fig. 1 enables the simultaneous description of both cubic Fm 3m reections from the rock salt structure, and the R 3m short-range ordered supercell peaks from cation ordering. 32 In the hexagonal unit cell description, reections deriving from the cubic structure have Miller indices with even values of l (hkl even ), whereas hexagonal reections that are forbidden in the cubic unit cell, but arise due to nanodomains of Li and Mn layering, exhibit odd values of l (hkl odd ).
Both sets of broad and sharper reections were tted using a single structural model, where the hkl-dependent peak broadening was modelled to evaluate the crystallite size of the cubic and hexagonal sublattices. The sample size was calculated using the double-Voigt approach 33 to estimate the volumeweighted column height separated into L hex and L cub components for hkl odd and hkl even reections, respectively. The rened value of L hex of 1.2(6) nm accounts for the volume weighted length of chemical coherence of the layered hexagonal ordering, and L cub ¼ 3.8(4) nm represents the total volume weighted crystallite size of the nanostructured material, which is independent of the layered ordering. This renement approach allowed us to quantify the extent of Li and Mn layering as approximately equal to the length of the c lattice parameter of the R 3m model (c hex ¼ 14.14157(6)Å) in Fig. 1, and is therefore a very short-range phenomena. It is important to note that the growth of this short-range layered ordering may occur along any of the [111] rock-salt directions, with coherent cubic domains between them. This structural difference between nano-Li 2 MnO 3 and its microstructured counterpart may be in part responsible for the increased capacity and reversibility in this material, as opposed to nanostructuring alone.
Details of the short-range structure can be revealed by studying the pair distribution function (PDF) data. Simulation of the short r-range of the PDF (r < 10Å) using the rened average model produced a large mismatch to the observed data in Fig. S2, † indicating a high degree of local atomic rearrangement in this material. The study of the local structure in cation disordered rock-salts is particularly challenging due to the overlap of all cation-cation, oxygen-oxygen partial PDFs and is beyond the scope of this work; a future study will focus on structural modelling via tting of EXAFS, X-ray and neutron total scattering data by Reverse Monte Carlo methods using a modied version of the RMCProle soware, 34 allowing the renement of multiple Bragg data and where the new hkldependent peak shape functions formerly introduced have been implemented. In the present study, the evolution of the local structure of nano-Li 2 MnO 3 during cycling was interpreted by the curve tting and PDF simulations using small clusters of LiO 6 and MnO 6 polyhedra and is introduced in the following section.

Bulk structural evolution during battery cycling
Rietveld renement of total scattering data collected during cycling (Fig. 2) showed a continuous evolution of the lattice parameters that mirrors the shape of the electrochemical curve, indicative of solid solution behaviour. The cell parameter of the cubic sublattice, a cubic , contracted from 4.0741(2) to 4.054(2)Å with the extraction of $1.5 Li during the rst charge to 4.5 V, and re-expanded to 4.101(2)Å during the discharge to 2 V. The solid solution response, contradicting previous claims, 26 is in congruent with the absence of characteristic plateaus in the electrochemical curve associated with phase transformations, and further supports our nding that the structure of pristine nano-Li 2 MnO 3 corresponds to that of a cation disordered rocksalt structure with nanodomains of Li/Mn layering. The isotropic breathing of nano-Li 2 MnO 3 is characteristic of cation disordered rock-salts, and is markedly different to the microstructured compound where the contraction of the cell occurs predominantly along the stacking direction. 35 The evolution of all rened parameters with cycling is given in Fig. S3. † The correlation length for the cubic sublattice, L cub ¼ 3.3(2) nm, remained constant during cell charge/discharge. No signicant differences in the prole of hkl odd reections were observed; however a large overlap between the most intense hkl odd 003 reection and the cell backgrounds (see diffraction data from PFA cell and carbon black additive in Fig. S3 †) meant that the correlation length for the hexagonal sublattice L hex could not be accurately rened and was xed to the value obtained the joint neutron and X-ray renement of pristine nano-Li 2 MnO 3 (L hex ¼ 1.2 nm). The absence of major structural-or microstructural-rearrangements combined with good structural reversibility are both highly desirable features for cycling reversibility and capacity retention of cathode materials.
The evolution of the average structure in nano-Li 2 MnO 3 resembles that of the non-stoichiometric cation disordered rock-salt of composition Li 4 Mn 2 O 5 , 25 in that in both compounds the extraction and insertion of lithium is accompanied by the breathing of the Mn cubic framework. The similarities between the evolution of the local and average structures in Li 4 Mn 2 3+ O 5 and Li 2 Mn 4+ O 3 are remarkable, given that different charge compensation mechanisms are expected in these cathode materials.
X-ray PDF data was used to extract the Mn-O distance during cycling (Fig. 2). Only very minor changes in the Mn-O distance were observed, evolving from 1.907(2)Å for pristine, 1.894(2)Å for charged and 1.908(2)Å for discharged phases, all of which are consistent with the expected r(Mn 4+ VI À O 2À VI ) of 1.93Å. 36 This constant Mn 4+ oxidation state may be expected from ligand eld theory considerations, where the splitting in energy of Mn 5+ dorbitals in an octahedral conguration is substantially larger than for tetrahedral. In absence of Mn redox, and alternative charge compensation mechanism must be found.
PDF data can also provide information on the local oxygen environment and so potential for anionic redox mechanisms involving oxygen were investigated. Reversible anionic redox have been attributed to the sharing of one electron hole between two oxygen atoms via formation of peroxo-like (O 2 ) 3À species in Li 2 RuO 3 and Li 2 IrO 3 , which are isostructural with the crystalline phase of Li 2 MnO 3 (space group C2/m). O-O dimers approximately 2.3-2.5Å apart have been proposed based on Ru K-edge EXAFS, 28 Bragg diffraction, 31 and transmission electron microscopy (TEM). Unlike PDF analysis, these techniques are not direct probes of O-O distances. (i) EXAFS is a model dependent technique where interatomic distances appear shifted due to the scattering phase-shi, (ii) Bragg diffraction quanties the distance between mean atomic positions d hOi-hOi different to the actual distribution of interatomic distances d hO-Oi , and (iii) projection limitations exist by TEM. Therefore, large uncertainties surround the estimated O-O distances in peroxolike dimers. In the absence of Mn redox, we consider whether a charge compensation mechanism accounting for $1.5Li during charge may be solely attributed to anionic redox. Such a compensation mechanism would change the oxidation state of all oxygen anions from 2 À to 1.5 À to give Li 0.5 Mn 4+ O 3

1.5À
, where all the oxygen atoms could become dimerized following the charge compensation mechanisms hypothesized for Li 2 RuO 3 and Li 2 IrO 3 . At such a high concentration of O-O dimers, their contribution to the total XPDF signal is not negligible. To demonstrate the impact of dimer formation, XPDF data was simulated for MnO 6 containing three dimerized O-O pairs at 2.3Å and 1.4Å for peroxo-like and peroxo bridges 19 Fig. 2 In operando nano-Li 2 MnO 3 evolution with cycling characterized by X-ray total scattering. From left to right: electrochemical performance (solid line) overlaid with lattice parameters from a sequential Rietveld refinement (grey empty circles with error bars), reciprocal and real space (PDF) total scattering data. The PDF of carbon black additive to aid electronic conductivity of the electrode is provided at the bottom of the figure. See Fig. S3 † for the evolution of all refined parameters and fits. Charge data is given in blue (from dark to light) and discharge in red (from light to dark). respectively (Fig. S4 †). The formation of 'peroxo-like' dimers signicantly changes the O-O partials within MnO 6 which in turn gives rise to the formation of new peaks with detectable/ measurable intensities at the new O-O distances (at 2.3 and 1.4Å in Fig. S4 †), and also affects the width and intensity of the peak at $2.9Å comprising overlapped cation-cation and O-O distances.
The 1.4-2.5Å range in the observed in operando PDF data were carefully examined (Fig. 3) and the evolution of peak width and intensity were characterized by the tting of ve Gaussian functions (Fig. S5 †). The superposition of PDFs for the peaks at $1.4 and 2.5Å attributed to C-C distances from the carbon black additive, rules out the O-O evolution at these distances. The intensity of the Li-O peak at $2.3Å decreases during charge and increases during discharge, as expected for the reversible extraction-reincorporation of lithium. No increase in the intensity of this peak at the 4-4.5 V range characteristic of anionic redox activity was observed that could be attributed to the formation of O-O dimers in signicant concentrations. For the Mn-O peak at $1.9Å, a decrease in width was observed upon charging, corresponding to a more ordered local structure, but the integrated intensity remained constant. The absence of peroxo-like or peroxo bridge distances is further supported by the constant rened width of the peak at $2.9Å attributed to cation-cation and O-O distances overlapped in the rock-salt structure. As previously shown by our simulations, the O-O pairs contribute to 20% of the intensity of this peak, and dimer formation would entail a signicant change of its width and intensity, which was not observed experimentally. It can be concluded that PDF data does not support anionic redox process stabilized by the formation of peroxo-like or peroxo bridges in nano-Li 2 MnO 3 .

Local structure around Mn during cycling
The evolution of the coordination environment around MnO 6 in nano-Li 2 MnO 3 was further characterized by in operando Mn Kedge HERFD-XANES, which is highly sensitive to structural changes around Mn cations. As established for known Mn standards, 37 different coordination environments of 3d transition metals show very different pre-edges, and these features were used to characterize changes in the coordination number and centrosymmetricity of Mn in nano-Li 2 MnO 3 .
The evolution of the HERFD-XANES data measured in operando during the rst charge to 4.5 V and discharge to 1.2 V are shown in Fig. 4. The most signicant change in the spectra correspond to the progressive fading of the shoulder at $6550 eV. First principle calculations were performed using the proposed R 3m structural model and an Fm 3m models where suitable Li and Mn compositions were randomly distributed on the same site (Fig. 5); these calculations demonstrated that the intensity of the shoulder feature correlates with the amount of Li on the Mn lattice, as determined previously by rst principle calculations on Li 4 Mn 2 O 5 . 37 The absence of this shoulder simulated at a charged composition of Li 0.5 MnO 3 is reected in the observed data at 4.5 V and the reappearance of this feature at the same energy during discharge indicates a high reversibility, in agreement with in operando total scattering data. Note that the apparent shi of the edge to higher energy values during charge is not caused by an increased Mn oxidation state, but instead by the expansion of the spectra on the energy scale due to the contraction of the lattice in accordance with Natoli's rule, and in agreement with the smaller lattice parameter observed from Rietveld renement.
Signicant changes of the pre-edge region of the XANES spectra are shown in Fig. 5. In the pristine material, two peaks can be distinguished at 6541 and 6543 eV, which are $2 eV apart as expected for the energy gap between the t 2g and e g states of a regular octahedral environment. Upon charging, the overall intensity of the pre-edge increases, and the relative intensity between the two peaks gradually changes until they become approximately equal intensity in the charged state. Such an observation points to a reduction in centrosymmetricity of the Mn centres, which become more distorted/irregular during charge. This reduction is centrosymmetricity is completely reversible upon discharge. Since no meaningful differences were apparent between the calculated spectra using R 3m and Fm 3m models from XANES simulations, only the Fm 3m model, which allowed more facile composition changes to be investigated, was retained for further spectroscopic analyses in the following sections.
The evolution of the environment around manganese that could be responsible for the changes in the shape and intensity of the HERFD-XANES pre-edge were investigated ex situ using EXAFS collected at the start (pristine) and end of the rst charge cycle (charged to 4.5 V). A comparison with in operando data collections is provided in Fig. S6, † showing an excellent agreement. Fig. S7 † compares the c(k) and Fourier transform (FT) c(r) signals for the pristine and charged to 4.5 V compositions with a signicant damping of the higher order shells (>3Å) in c(r), a common feature in nanostructured materials (e.g., Li 4 Mn 2 O 5 , 37 a-MnO 2 (ref. 38)) and attributed to a high degree of structural disorder.
The average structures from Rietveld renements were used as starting congurations for the least-square tting of EXAFS data, to quantitatively evaluate short-range order structural parameters. Three single scattering paths including O and Mn shells were rened, while Li shells were not included in the renement due to the small scattering power of this atom. Interatomic distances (R) and mean square displacements (s 2 ) were rened for all paths. However, due to the high degree of correlation amongst path degeneracies (N degen ) and s 2 , N degen were xed to known values, which produced a good t to the data for pristine nano-Li 2 MnO 3 (R wp ¼ 2.8%, Fig. 6). However, for charged nano-Li 2 MnO 3 , N degen were xed to known values for Mn-O and Mn-O2 shells (see Table S1 †), while those of Mn-Mn were allowed to vary. The interested reader is referred to Page S14 † for a full discussion of the tting parameters. The renement of Mn-Mn path degeneracy was key for the obtainment of a good t for the charged sample. The t improved from R wp ¼ 4.6 to 2.9%, when the rened mean path degeneracy varied from 2 in the Rietveld model to 5 AE 2. This result could be correlated to the migration of Mn cations to the vacancies generated by the extraction of $1.5 Li per formula unit amounting to 50% of the cations in nano-Li 2 MnO 3 . Such an increase in the Mn-Mn path degeneracy will imply the occupation of 25% of the generated cation vacancies by Mn. Cation rearrangement and densication processes are not uncommon amongst positive electrode materials, 39    In summary, the evolution of HERDF-XANES and total scattering data measured in operando demonstrated an excellent structural reversibility of the nano-Li 2 MnO 3 cathode during the rst (dis)charge cycle with a solid-solution type response allowed by the breathing of a cubic Mn-framework, in agreement with in operando total scattering data evolution. The evolution of the short r-range structure studied by EXAFS, and previously by XPDF, did not show a shortening of Mn-O bonds characteristic of peroxide species with smaller radii, or signicant changes in the local structure associated to the formation of 'peroxo-like', or peroxo-bridge species during the stabilization of anionic redox processes, but pointed instead to the densication of the cation sublattice. During densication, the migration of Mn cations between two octahedral sites will involve the formation of intermediate stages with Mn in different coordination environments, which will in turn explain the increased pre-edge intensity in Fig. 5. The participating charge compensation mechanisms were studied in depth by X-ray emission spectroscopies, and are discussed in the following section.

Electronic structure evolution during cycling
Note that while the position of the absorption edge or "white line" is commonly used as an indication of the oxidation state, its shape is strongly inuenced by structural effects leading to large inaccuracies. We have instead determine the oxidation state of Mn with greater precision from the position of the kb 1,3 transition, as explained in the following section.
Redox processes were investigated through in operando XES by the combined interpretation of main (kb 0 and kb 1,3 ) and V2C (kb 00 and kb 2,3 ) transitions (Fig. 7). The kb emission of rst row TMs can be explained as a two-step process represented by the schematic diagram in Fig. 7a. The different chemical selectivity of main kb and V2C to manganese and oxygen respectively, makes XES a powerful tool to investigate the electrochemical activity of Li-rich cathode materials with capacities beyond those expected from TM redox. The use of hard X-rays in XES or Resonant Inelastic X-ray Scattering (RIXS) provides bulk structural information, unlike techniques employing so X-rays (e.g., XPS, or light elements K-absorption edges) which are only sensitive to surface atoms and require high vacuum compatible sample environments. In addition, V2C combines both bulk sensitivity and selectivity to the detect exclusively the O bound to Mn in the active material; unlike RIXS, V2C measurements eliminate possible contamination of the spectroscopy signal with oxygen in other components of the battery, like the cell case or electrolyte. In operando XES data was collected using a conventional cell for XAS experiments equipped with a beryllium window. 37 While advanced spectroscopic measurements Previously we demonstrated the use of main kb transitions are a more accurate indication of the Mn oxidation state than XANES 1s-4p energy transitions. 37 The 3p orbitals lay deeper within the electronic shells, and as such are less dependent on structural effects than the 1s-4p transition probed at the Mn absorption K-edge. In main kb XES, the number of unpaired 3d electrons can be indirectly inferred from magnitude of 3p-3d exchange interactions (proportional to the number of unpaired 3d electrons), given by the difference between kb 0 and kb 1,3 transitions. We use the comparison of the most intense kb 1,3 transition with those of Mn oxides with well-known oxidation states in Fig. S8 and S9 † as a gauge of Mn oxidation state at different states of charge.
The main kb spectra of pristine nano-Li 2 MnO 3 , as the HERFD-XANES discussed above and the V2C spectra below, are in good agreement with the MnO 2 data with the same oxidation state of manganese. During charge, a shi of kb 1,3 transition towards lower energy values is observed that is fully reversed during discharge, in agreement with the good structural reversibility observed for the in operando total scattering and HERFD-XANES data. Note the subtle variation of kb 1,3 energy is the only difference between the spectra, and the integrated intensities for kb 0 and kb 1,3 , and the kb 0 energy transition remain constant during battery cycling (see Fig. S10 †). As well as the number of unpaired 3d electrons, the magnitude of the 3p-3d exchange interaction depends also on the spin state and the covalency/charge-transfer contribution effects which are expected to change when chemical environment around MnO 6 centres is modied during cell (dis)charge. Due to the lower magnitude of the kb 1,3 shi, in comparison to the more signicant differences observed when the oxidation state of Mn is varied across the different standards, this subtle energy variation was attributed to changes in the Mn-O covalency that accompany the slight shortening of Mn-O bonds during charge by $0.02Å observed by PDF. Therefore, this analysis indicates manganese remains at a constant state of Mn 4+ , but changes of the Mn-O covalency occur during charge and discharge processes.
Mn-O interactions were further investigated by V2C spectroscopy (including lower kb 00 and higher kb 2,5 energy transitions) which has a high sensitivity to the chemical environment around Mn centres. The energy separation between kb 00 and kb 2,5 transitions is determined by the binding energy difference between ligand-s and -p orbitals, and their intensities reect the degree of hybridization between ligand and TM p states. 41 The lower energy kb 00 transition, which originates from ligand s molecular orbitals to TM p states, is a valuable tool for ligand identication. The energy of this kb 00 transition depends on the difference between the ligand-2s and TM-1s binding energies, and reects the spatial overlap between ligand 2s and TM 4p/3d molecular orbitals. The kb 00 intensity is affected by the TM-ligand distance and by the coordination number of the TM. Shorter distances (e.g. from higher bond order or deprotonation) result in increased kb 00 intensity, with an approximate exponential dependence on distance. 42 Thus, the evolution of the kb 00 transition during cycling is able to detect changes in the charge density on the metal (Mn), ligand (O) protonation, and changes in the coordination environment of the ligand. The higher energy and more intense kb 2,5 originates from the transition of electrons in an occupied valence molecular orbital with mainly ligand-p and metal-p contributions (see Fig. S11 †). The evolution of this signal during cycling signies changes in the valence orbitals and TM-ligand chemical bond nature.
The evolution of kb 2,5 and kb 00 transitions with electrochemical cycling of nano-LiMn 2 O 3 are given in Fig. 7. Very small changes of the peak position and intensities were observed upon cycling, which were particularly small in comparison with the much larger changes between the spectra of standard compounds of Mn in different oxidation states. During charge, the kb 2,5 transition increased in energy by $0.5 eV with a concurrent increase of 5% in the integrated intensity; these changes was fully reversed during discharge. No signicant change in the kb 00 transition was identied by curve-tting (Fig. S10 †).
The V2C spectra were simulated using a 2 Â 2 Â 2 supercell of composition Li 21 Mn 11 O 32 built from the expansion of the proposed average model from Rietveld where the lattice parameters have been rened with a high accuracy (a cubic ¼ 4.0741(2)Å), and Mn-O distances were relaxed to $1.9Å to match the values proposed from XPDF and EXAFS analyses (Fig. S12 †). These simulations reproduced the main V2C spectral features for pristine nano-Li 2 MnO 3 . Minor changes were observed in the V2C spectra during cycling, which we evaluate along with input from other observations in the following section.

Evaluating possible charge compensation mechanisms
The combination of observations from the above techniques allows for possible charge compensation mechanisms to be compared and appraised as an alternative to the unfavourable evolution of octahedrally coordinated manganese from Mn 4+ to Mn 5+/7+ . 43 Here we evaluate four proposed hypotheses of plausible charge compensation mechanisms in nano-Li 2 MnO 3 : (i) lattice densication, (ii) formation of hydroxyl groups (-OH), (iii) trapping of positively charged hydronium molecules (H 3 O + ), and (iv) anionic redox.
(i) Densication. A densication mechanism was suggested by EXAFS observations. Two models have been proposed for charged compositions that have been originated from the nano-Li 2 MnO 3 supercell with updated lattice parameters from Rietveld renement (a cubic ¼ 4.054(2)Å), where lithium atoms have been replaced by cation vacancies to obtain a Li : Mn ratio of 1 : 2 based on the ratio proposed by electrochemistry and ICP analyses. The V2C simulation of such model (Li 5 Mn 10 O 32 ), with $50% cation vacancies is given in Fig. S12. † Additionally, 23% of the cation vacancies were removed from Li 5 Mn 10 O 32 in a second 'densied' model (Li 6 Mn 13 O 32 ) matching the Mn-Mn path degeneracies suggested by EXAFS. Such a structural modication will imply the effective loss of 25% of oxygen in nano-Li 2 MnO 3 with the effective extraction of Li 2 O following the reaction: where ',' represents a cation vacancy. The comparison between V2C simulations of 'undensied' Li 5 Mn 10 O 32 and densied Li 6 Mn 13 O 32 in Fig. S12 † shows a shi of the kb 00 and kb 2,5 transitions to higher energies for the 'unden-sied' composition by approximately 0.5 and 1 eV. This trend is also observed for Mn standards, where kb 00 and kb 2,5 transitions shi also by 0.5-1 eV to higher energies with the increased oxidation state of manganese, and is consistent with the higher oxidation state of Mn expected for the undensied model vs. the densied one. The different energy transitions between pristine and the undensied models are not experimentally observed and a much better agreement is obtained with the densied model based on EXAFS rened path degeneracies. Additionally, the lower intensity of the kb 00 transition in the undensied model correlated to the lower number of Mn neighbours around oxygen in this model, is opposite to the experimental observation with a small 5% increase of the kb 00 intensity.
(ii) Mn-OH formation. Protonation during cycling and formation of a -OH hydroxyl group has previously reported for spinel electrodes 44  In fact, the kb 00 region of the V2C spectra has been used as a probe to distinguish the protonation state of TM bound ligands in a previous study. 47 The V2C spectra were simulated for two models with a cubic disorder rock-salt structure with compositions $H 1.5 -Li 0.5 MnO 3 (H 15 Li 5 Mn 11 O 32 ) and Li 0.5 MnO 3 (Li 5 Mn 11 O 32 ), where Mn-OH and Mn-O distances were xed at a constant value, as previously determined by XPDF and EXAFS renements (see Fig. S13 †). Even in these unphysical models with identical Mn-OH and Mn-O distances, our simulations show a shi of 2 eV towards lower energies occurs for both kb 00 and kb 2,5 transitions upon protonation of 50% of the oxygen ligands, as well as signicant changes in the shape and width of the kb 00 transition. Hence, the protonation of oxygen involving the formation of hydroxyl groups is contradicted by both total scattering and spectroscopy data.
(iii) Trapped H 3 O + . Protonation and trapping of H 3 O + molecules has previously been reported for a-MnO 2 (or Mn 8 O 16 ) type materials. 49,50 a-MnO 2 has a hollandite-type structure with space group I4/m, which consists of corner and edge sharing MnO 6 octahedra with mixed oxidation states of Mn 3+ and Mn 4+ cations. The structure is composed of open tunnels that can accommodate the presence of cations and small molecules including K + , Ag + , Na + , Mg 2+ , NH 4 + and H 3 O + . [49][50][51] This has led to tunnel structured manganese oxides being investigated in the elds of catalysis, adsorption-separation, and ion exchange. The V2C spectral signature of trapped H 3 O + molecule was evaluated by the simulation of V2C spectra from structural models for a-MnO 2 reported in the literature with/without H 3 O + in Fig. S14. † These simulations show a small increase in of the kb 00 intensity, compatible with the 5% increase of the integrated intensity observed for this transition during cycling of nano-Li 2 MnO 3 . Similarly-sized one-dimensional tunnels to those present for a-MnO 2 2 Â 2 (0.46 Â 0.46 nm 2 ), with only a slightly smaller surface of 0.44 Â 0.44 nm 2 , are formed as a result of the extraction of lithium in nano-Li 2 MnO 3 ; the surface of these voids could be further enlarged by the displacement of Mn cations away from their average positions. The large concentration of voids in charged nano-Li 2 MnO 3 generated from the extraction of lithium could potentially host H 3 O + molecules originated from electrolyte decomposition, where trapped H 3 O + molecules compensate part of the charge exchanged upon charging following the reaction in eqn (2).
The participation of eqn (2) in the charge compensation of nano-Li 2 MnO 3 (if any) is expected to be small, based on the reported concentration of these molecules previously reached in a-MnO 2 with a H 3 O + to Mn ratio of 0.15 : 1. This amounts to only 10% of the 1.5 electrons per formula unit extracted over the course of the rst charge. The corroboration and quantication of this charge compensation mechanism to the overall capacity could not be accurately determined by the techniques used in this work, and requires follow-up H-NMR studies.
(iv) Anionic redox. V2C spectroscopy has been a powerful tool to detect the presence of O-O bonds, 41 oxo bridges 52 and Mnperoxo units 53 of Mn oxygen-evolving complex that revealed different spectroscopic signatures of kb 00 transitions. While signicant changes between V2C transitions of Mn-O 2À and Mn-(O 2 ) 2À bonds have been reported which showed that the kb 00 transition is split into two peaks when Mn (or other TMs) is bound to peroxo species, 41,52 the V2C spectral signature of peroxolike species is currently unknown. Having previously ruled out the sole participation of peroxo-like dimers as a charge compensation route in nano-Li 2 MnO 3 by XPDF and EXAFS, we simulated the V2C spectra for charged nano-Li 2 MnO 3 structural models (Li 6 Mn 13 O 32 ) with O-O dimers at 2.3Å, formerly introduced in Fig. S4b, † to (i) identify the V2C spectral signature of such defects and (ii) determine the extent to which V2C can be sensitive to a small concentration of such defects. The comparison of Li 6 Mn 13 O 32 structural models with dimerized and regular MnO 6 shows large similarities between the two practically superimposable V2C spectra (Fig. S15 †). Moreover, the integrated charge for oxygen atoms in regular and dimerized MnO 6 models is also comparable, with the charge of oxygen atoms in dimerized MnO 6 models actually 8% smaller (more reduced) than for oxygen in the regular octahedral environment, or in other words, a charge variation 'Dq' in oxygen of 0.16. Thus, our calculations indicate not only that the longer O-O peroxo-like dimers at 2.3Å could not be detected by V2C even at the largest concentrations, but also that the presence of such defects does not signicantly change the integrated charge for oxygen.
From the results above, the best agreement between proposed charge compensation mechanism and observed experimental data (XPDF, EXAFS and V2C) is given by the densied model, as shown in Fig. 8 (and Table S2 †). The proposed densication of the Mn sublattice implies the loss of Li 2 O during the cathode charge, and has been previously reported in other cathode materials within the LiMnO system. 54,55 However, unlike previous studies, the cation densication could not be linked to unfavourable anionic redox that result in the formation of evolving O 2(g) or Oc À radicals that in turn react with the electrolyte to produce LiOH, Li 2 O 2 or Li 2 CO 3 secondary phases associated with capacity drop and structural instability. The absence of undesirable secondary phases in nano-Li 2 MnO 3 could be the origin of the large structural reversibility and reversible charge capacity of over 290 mA h g À1 .

Discussion
This comprehensive study utilising multiple techniques with different chemical and lenghtscale sensitivities have provided an unequivocal answer: no oxygen or manganese redox reactions in the bulk material or structural transitions are responsible for the charge compensation in nano-Li 2 MnO 3 . Instead, the observation of lattice densication points to the reversible extraction and reinsertion of Li 2 O in nano-Li 2 MnO 3 . The absence of a Li 2 O secondary phase in our data is indicative of the electrochemical activation of this compound as the main charge compensation mechanism responsible for the reversible capacity in nano-Li 2 MnO 3 (see Fig. 9). This observation is in agreement with the recently reported ability of nanostructured Li-Mn-O disordered rock-salts to catalyse the electrochemical activation of Li 2 O in Li 2 O:Li-Mn-O nanocomposites 56 with superior rst charge capacities (>1100 mA h g À1 ) following the reactions: The reaction of highly reactive oxygen radicals formed in eqn (3)  Our results showed that, contrary to the general assumption that extra capacities of cation-disordered rock salts are mediated by bulk anionic redox reactions, nano-Li 2 MnO 3 acts as an activator to catalyse the reversible formation of electrochemically active Li 2 O. The use of full structural (and not just average structural) probes, as well as bulk probes of anionic redox sensitive to only the oxygen bounds to Mn, avoids issues of surface sensitivity of XPS and so-XAS, which are also sensitive to any other oxygen containing phases like the electrolyte. Based on the success of these techniques to characterize in operando and with excellent time resolution and data quality these bulk structural phenomena, the introduced methods should be used to revisit the characterisation of other high capacity cathodes with bulk anionic redox process have been based on TEM, TM K-edge EXAFS, average structure, XPS or DFT calculations, to nd conclusive evidence of such reactions.
We further explore the similarities between nano-Li 2 MnO 3 and other nanostructured Li-Mn-O cathodes, namely Li 4 Mn 2 O 5 . In operando Bragg data of the Li 4 Mn 2 O 5 cathode charged to 4.5 V revealed the appearance of broad reections unindexed by the Fm 3m space group (see Fig. S16 †) that were attributed to unidentied impurities in previous work. 66 The comparison between Fig. 1 and S16 † indicates the origin of the diffuse reections in charged Li 4 Mn 2 O 5 could be instead linked to the formation of a short-range ordered superstructure during cycling. In addition, previous studies of Li 4 Mn 2 O 5 combining main kb transitions, XANES and EXAFS showed Mn 3+ irreversibly oxidizes to Mn 4+ during the rst charge. 37 Therefore, Li 4 Mn 2 O 5 could be transformed into a 'nano-Li 2 MnO 3 like' material (i.e., with nanodomains of Li/Mn layering and Mn 4+ ) that cycles reversibly between charged and discharged states with a constant oxidation state of Mn 4+ and charge compensation could be allowed by the reversible formation of Li 2 O. Even though Li-rich cation disordered rock-salts are generally associated with the absence of structural transformation during cycling, the chemical ordering observed during charging of Li 4

Conclusion
This work has determined the true average structure of nano-Li 2 MnO 3 as a cubic superstructure with short-range layered ordering in space group R 3m. This ordering has been observed  The full understanding of the charge compensation in nano-Li 2 MnO 3 was possible thanks to hard X-rays techniques in XES that provide bulk structural information in operando; techniques employing so X-rays (e.g. XPS, or light elements Kabsorption edges) may not provide such rich information due to their surface-sensitivity and the requirement for high vacuum environments which may compromise the observed electrochemistry. [70][71][72][73]

Conflicts of interest
There are no conicts to declare.