Oxidation behaviour of lattice oxygen in Li-rich manganese-based layered oxide studied by hard X-ray photoelectron spectroscopy

The oxidation/reduction behaviours of lattice oxygen and transition metals in a Li-rich manganese-based layered oxide Li[Li0.25Ni0.20Mn0.55]O1.93 are investigated by using hard X-ray photoelectron spectroscopy (HAX-PES). By making use of its deeper probing depth rather than in-house XPS analyses, we clearly confirm the formation of O ions as bulk oxygen species in the active material. They are formed on the 1 charging process as a charge compensation mechanism for delithiation and decrease on discharging. In particular, the cation–anion dual charge compensation involving Ni and O ions is suggested during the voltage slope region of the charging process. The Ni ions in the material are considered to increase the capacity delivered by a reversible anion redox reaction with the suppression of O2 gas release. On the other hand, we found structural deterioration in the cycled material. The O species are still observed but are electrochemically inactive during the 5 charge–discharge cycle. Also, the oxidation state of Ni ions is divalent and inactive, although that of Mn ions changes reversibly. We believe that this is associated with the structural rearrangement occurring after the activation process during the 1 charging, leading to the formation of spinelor rocksalt-like domains over the sub-surface region of the particles.


Introduction
Rechargeable lithium ion batteries (LIBs) have been widely used as power sources for portable devices and now are available for applications to electric vehicles (EVs), which demand high power, high energy density, long-life, inexpensive, and safe batteries. So far, the most successful positive electrode materials have been layered transition-metal oxides such as LiCoO 2 and its derivatives; however their electrochemical capacities are conventionally limited to ca. 150 mA h g À1 with a potential up to 4.2 V vs. Li/Li + . 1 Lithium-rich manganese-based layered oxides (LLOs), commonly described as a composite xLi 2 MnO 3 $(1 À x) LiMO 2 (M ¼ Ni, Co, etc.), have been extensively studied as a preferred candidate material for next-generation LIB positive electrodes because they show a high reversible capacity beyond 200 mA h g À1 and consist of relatively inexpensive elements such as Mn and Ni. [2][3][4][5][6] It is well known that LLOs show a characteristic irreversible voltage plateau at ca. 4.5 V vs. Li/Li + in the initial charging process, which is considered to come from the activation of the nominally inactive Li 2 MnO 3 component in the materials. 3,5 Much attention has been paid to the origin of this plateau because it is the key behaviour explaining extra capacities in LLOs. The previous papers have related it to O 2 gas release behaviour, which should be accompanied by structural rearrangement. 7,8 The O 2 release was evidenced by the gas chromatography (GC) or differential electrochemical mass spectrometry (DEMS) analyses. [9][10][11][12][13] The oxygen loss from the material is consistent with a large irreversible capacity observed in the initial cycle, but the quantication of the released O 2 is not so straightforward. 9,11,14 As an alternative approach, the oxygen deciency has been shown from X-ray diffraction data analysed by the Rietveld method. 8,15,16 These studies have conrmed that the oxygen is more or less extracted from the structure. On the other hand, the rst-principles calculations have suggested that the oxidation of oxygen anions without O 2 release could explain the plateau prole. 17,18 These theoretical studies indicate that the redox reaction of oxygen in lattice should be taken into account. Recently, we have reported evidence of peroxide formation in Li 1. 16 Ni 0.15 Co 0. 19 Mn 0.50 O 2 at a charged state by using so X-ray absorption spectroscopy (XAS) at the O K-edge. 19 However, the detailed oxidation behaviour of the oxygen in lattice is not yet well understood during the charge-discharge process.
X-ray photoelectron spectroscopy (XPS) is one of the unique techniques for obtaining chemical state information on a specic element in the battery materials. This technique is especially surface-sensitive, which usually has a probing depth of <ca. 5 nm using a conventional X-ray source (Al Ka radiation of 1.49 keV, for example), and is suitable for the chemical characterisation of the solid electrolyte interphase (SEI) comprising light elements (Li, C, O, and F) deposited on the electrode surface. [20][21][22] Recently, Tarascon and coworkers used this technique to identify the oxidation of oxygen anions in positive electrode materials. 23 25,26 However, because of its surface-sensitive nature, it is oen the case that the O 1s XPS spectra were largely affected by electrolyte degradation products such as Li 2 CO 3 , which makes it difficult to validate the formation of peroxide species in the bulk. XPS measurements using synchrotron X-ray radiation are advantageous to probe the chemical state information from the deeper inside of the material (the probing depth of <ca. 50 nm), 27 which enables us to focus on the lattice structure of the LIB active materials buried beneath the surface-deposited products. In this study, we have investigated the oxidation/reduction behaviours of lattice oxygen and transition metals in a lithium-rich manganese-based layered oxide Li[Li 0.25 Ni 0.20 Mn 0.55 ]O 1.93 during the irreversible initial charge-discharge cycle by using hard X-ray photoelectron spectroscopy (HAX-PES). A possible structural modication occurring during the charge-discharge cycles has also been examined for the material disassembled on the 5 th cycle. was determined from the inductively coupled plasma-atomic emission spectrometry (ICP-AES, ICPS-8100, Shimadzu) and the iodometric titration measurements that estimate the average oxidation states of the transition metals using 0.03 M Na 2 S 2 O 3 solution. The materials were also characterised by X-ray diffraction and indexed to C2/m (Fig. S1 †).

Experimental
A mixture of the active material (Li[Li 0.25 Ni 0.20 Mn 0.55 ]O 1.93 or Li 2 MnO 3 ), acetylene black (Denki Kagaku Kogyo) and polyvinylidene diuoride (PVDF, Kureha) with a weight ratio of 80 : 10 : 10 was spread onto an aluminium foil with 1-methyl-2pyrrolidone (NMP) and then dried at 80 C under vacuum overnight to constitute a positive electrode. The electrode was pressed to a typical thickness of 30-35 mm. A foil of metallic lithium (0.2 mm in thickness, >99.9%, Honjo Metal) was used as counter and reference electrodes. These components were assembled together with a Celgard 2500 separator and soaked in the electrolyte solution in an Ar-lled glove box, which were sealed in an aluminium-coated laminate-type cell. The electrolyte used in this study was 1 M LiPF 6 , which was dissolved in anhydrous ethylene carbonate (EC) and ethylmethyl carbonate (EMC) with a volumetric ratio of 3 : 7 (Kishida Chemical).
The electrochemical measurements were performed at room temperature on an automatic cycling and data recording system (HJ1001SD8, Hokuto Denko). The cells were galvanostatically cycled between 4.8 and 2.0 V vs. Li/Li + at a current rate of 10 mA g À1 . A series of delithiated/relithiated samples were prepared for HAX-PES measurements by disassembling the cell at desired states of charge and discharge (Tables S1 and S2 †). They were carefully disassembled in the glove box and rinsed with dimethyl carbonate (DMC) to remove the electrolyte residue.
The HAX-PES spectra were acquired at room temperature at BL28XU at SPring-8 (Hyogo, Japan). The electrode samples were placed onto a sample holder in the glove box and then transferred to an ultra-high vacuum sample chamber (<5 Â 10 À6 Pa) without exposing them to air. The incident photon energy was 7.94 keV. The detection depth was estimated to be ca. 50 nm from thickness-controlled oxidised Si wafers. The photoelectron analyser was a VG Scienta EW4000. The pass energy was set to 200 eV. The binding energies of all the spectra were calibrated with respect to the lattice oxygen peak of Li[Li 0.25 Ni 0.20 Mn 0.55 ] O 1.93 (529.4 eV relative to the C 1s signal from acetylene black at 284.5 eV) and Li 2 MnO 3 (529.5 eV) in O 1s spectra, respectively. These values were carefully determined by applying a dual beam charge neutraliser (low energy electrons and Ar + ions) on an inhouse photoelectron spectrometer PHI5000 VersaProbe II (monochromated Al Ka radiation; 1486.6 eV, ULVAC-PHI.Inc). The spectra were analysed aer the Shirley-type background subtraction.
First-principles calculations were performed based on the spin-polarised density functional theory (DFT) using the CASTEP code 28 to obtain essential information about the electronic structure of LLOs from a theoretical point of view. The generalised gradient approximation with the Hubbard U correction (GGA+U) for Ni and Mn 3d states (U eff ¼ 4.5 eV) was employed. The Perdew-Burke-Ernzerhof potential was used for the exchange-correlation term. 29 The plane-wave cut-off energy was set to 550 eV. The k point sampling for the Brillouin zone integration was done according to the Monkhorst-Pack scheme with a grid spacing of <0.04Å À1 . 30 The model structure was constructed to reect a minimal structural domain in the real material, which was prepared from a 2 Â 1 Â 2 supercell of the Li 2 MnO 3 (C2/m) structure with 4Li + ions in a transition-metal layer (LiMn 2 layer) substituted by 2Ni 2+ ions, making the composition Li 28 Ni 2 Mn 16 O 48 in the cell (the chemical composition in the simulation is different from that in the experiments for simplication). Atomic positions and cell parameters were fully relaxed until the residual forces and stresses acting on each atom were less than 0.03 eVÅ À1 and 0.05 GPa, respectively. Delithiated structures (Li 24   On the 1 st charging process, a voltage slope is observed below 4.5 V, and a plateau is above 4.5 V, which is characteristic of the charging prole of LLOs. 2,4 This plateau is irreversible and is not observed on the 1 st discharging process and subsequent charge-discharge cycles. The charging and discharging capacities are 309 and 172 mA h g À1 for the 1 st cycle and 218 and 215 mA h g À1 for the 5 th cycle, respectively. These capacities are much higher than that of Li 2 MnO 3 prepared under the same conditions (Fig. S2 †).  Fig. 2b shows the Ni 2p 3/2 photoelectron spectra for the 1 st cycle. The pristine sample consists of divalent and trivalent Ni ions (Sp. #1), although the chemical shis for the divalent, trivalent, and tetravalent Ni references were very close to each other. 38 The shoulder at the higher energy side increases in intensity during the voltage slope. This means the increase in the amount of tetravalent Ni ions up to 4.5 V (Sps. #2-4). In the plateau region, the peak shape is almost unchanged, indicating that Ni ions do not contribute to the charge compensation at this stage (Sp. #6). The high valence Ni ions are reduced to be divalent down to 3.5 V (Sps. #7 and 8). The peak position is  unchanged in the voltage window between 3.5 and 2.0 V (Sp. #9). Similar behaviours were also reported in the previous XPS and XAS studies. 37,[39][40][41] We note that the Li 1s spectra show a monotonous decrease in intensity during the charging process up to 4.8 V (Fig. S5a †), indicating that the above spectral changes in Mn and Ni ions cannot explain the charge compensation in the voltage plateau region.
The O 1s photoelectron spectra are shown in Fig. 3. The prominent peak at 529.4 eV is assigned to the lattice oxide ions (O 2À ) in the pristine material (Sp. #1). The broader signal centred at 532.7 eV comes from the surface-deposited oxygen species. On delithiation, a shoulder signal arises at ca. 530.6 eV, the higher energy side of the oxide peak (Sps. #2-6). According to the previous studies, this signal corresponds to the more oxidised oxygen state than O 2À ions in the material, i.e., peroxide-like O À ions (here, the term "peroxide" does not necessarily mean the molecular unit O 2 2À as in Li 2 O 2 ). 23,[42][43][44] They are also visible in the Li 2 MnO 3 electrodes (Sps. #2 and 3 in Fig. S6 †). We here emphasise that this O À signal is clearly visible and enhanced in the HAX-PES spectra rather than in the "surface-sensitive" in-house XPS spectra used in the previous studies, 23-26 which strongly suggests that they are formed in the lattice structure and are not from the surface-deposited oxygenated product such as Li 2 O 2 . We note that the in-house XPS O 1s spectrum with Ar + sputtering also gives clear evidence of the O À species in the bulk (Fig. S7 †). Fig. 4 1.93 , the initial Ni oxidation state is +2.1, and therefore, it can deliver a capacity of ca. 130 mA h g À1 when it is oxidised to +4 during the charging up to 4.8 V. At the end of charge, 33% of the lattice oxygen (i.e., 0.63O) change to the O À species (Fig. 4, Sp. #6), corresponding to an additional charge of 208 mA h g À1 by considering the O 2À /O À couple. Therefore, the total capacity can be 338 mA h g À1 . This is slightly higher but reasonably close to the actual charging capacity of 309 mA h g À1 . At the end of discharge, the capacities recovered by Ni 2+ /Ni 4+ and O 2À /O À couples are ca. 130 and 76 mA h g À1 , respectively, the sum of which is again slightly higher than the actual discharging capacity of 172 mA h g À1 including the partial contributions from the Mn reduction. The small differences (overestimates) partly come from the assumption that all the Ni 2+ ions change to Ni 4+ without any electric isolation of active material particles and/or from the uncertainty in peak decomposition for the O 1s spectra. It is also suggested that the irreversible capacity during the 1 st charge-discharge cycle is caused by the irreversible O 2À /O À redox reaction (Fig. 4).
To gain some insight into the charge compensation mechanism in the LLO system, we calculated the electronic density of states (DOS) of a Li 2 MnO 3 -Li(Ni,Mn)O 2 composite-type model structure. The model structure was prepared in a simple fashion from a 2 Â 1 Â 2 supercell of the Li 2 MnO 3 (C2/m) structure with 4Li + ions in a LiMn 2 layer substituted by 2Ni 2+ ions (and 2 vacancies) being a Ni 1/2 Mn 2 layer (Fig. 5). This model gives atomic column images of -Li-Mn-Mnarrangement as in Li 2 MnO 3 and -M-M-Mas in LiMO 2 along the b axis, which were observed in the previous studies using scanning  Because qualitative or quantitative gas detection measurements such as DEMS were not carried out, it is still unclear whether or not there is some participation of oxygen loss for charge compensation in this material. On the other hand, the present study shows the signicant role of the redox reaction of the lattice oxygen (O 2À /O À ), suggesting that the Ni ions in the structure suppress the O 2 gas release at the voltage plateau. According to the previous studies, it is expected that a more electronegative cation is preferably hybridised to the ligand, the activated oxygen species. 23,24,36 Here, the Ni 4+ ion at charged states has a larger electronegativity and has a preference to hybridise with the activated oxygen species over Mn 4+ ions. 36 The present study supports the suggestion given by the previous studies that strong hybridisation of Ni 4+ and O À species would stabilise the lattice structure preventing the O 2 release as evidenced by the fact that the Mn oxidation state remains tetravalent during the voltage plateau, while Li 2 MnO 3 shows Mn valence reduction originating from the substantial O 2 release from the structure. Next, we examine the structural stability on charge-discharge cycling. Fig. 6a shows the Mn 2p 3/2 HAX-PES spectra for the 5 th  charge-discharge cycle. The sample disassembled at 2.0 V aer the 4 th cycle (Sp. #10) shows a similar spectrum to that disassembled at 2.0 V aer the 1 st cycle (Sp. #9 in Fig. 2a). On the charging process, the decrease of lower-valence Mn 2+ and Mn 3+ components and concomitant increase of Mn 4+ state are observed above 3.5 V (Sps. #12 and 13). This spectral change is reversible (Sps. #14-16). The increased discharging capacity below 3.5 V for the 5 th cycle compared with that for the 1 st cycle (Fig. 1) is attributable to the increasing participation of the Mn redox reaction for charge compensation. On the other hand, the oxidation state of Ni ions remains divalent during the 5 th charge-discharge cycle (Fig. 6b). Fig. 7 shows the O 1s spectra for the 5 th cycle. Unfortunately, the signal at higher energy (>ca. 531.0 eV), mainly attributable to the surface-deposited lithium carbonate Li 2 CO 3 , is signicant for the samples disassembled below 3.5 V. Based on the peak decomposition (Fig. S9 †), we found that the O À species still exist in the material, but they are almost constant in amount on the 5 th cycle (Fig. 8).
Considering that the Li 1s spectra for the 5 th cycle still show the reversible delithiation/relithiation behaviour (Fig. S5b †), the charge compensation is satised only by the Mn redox reaction on the 5 th cycle. It is a very confusing result, because the previous XAS studies (bulk-sensitive) reported that the Ni ions act as a charge compensator even aer the 1 st cycle. 19,40,41 Therefore, we believe that the Ni ions become inactive over (at least) 50 nm thickness from the surface, which is a probing depth of HAX-PES measurements, while they are still active in the deeper inside of the particles. This is probably related to structural rearrangement (deterioration) on the surface and sub-surface region. Recent STEM observations have revealed the formation of spinel-or rocksalt-like structural domains on the particle surface of LLOs aer the charge-discharge cycles. 46,48-51 Therefore, the Ni 2p HAX-PES spectra of the cycled materials represent the electrochemically inactive Ni ions stabilised in the sub-surface Lipoor spinel-or rocksalt-like domains, which do not act as a charge-compensating ion on the charge-discharge cycles. The oxygen anions also have no more contribution to charge compensation in the sub-surface region, although it is possible that they are redox active in the bulk. 19,39,41 It is noteworthy that the formation of the spinel-or rocksalt-like domains suggests the oxygen loss from the surface. This indicates that both the   mechanisms (oxygen loss and oxidation of the lattice oxygen) are important for the delithiation of the LLO positive electrode materials, and their contributions would be dependent on the chemical composition. The former mechanism would be dominant in Li 2 MnO 3 , but the latter becomes more important in Ni-containing LLOs.

Conclusions
We have investigated the oxidation/reduction behaviours of oxygen and transition metals in a Li-rich manganese-based layered oxide Li[Li 0.25 Ni 0.20 Mn 0.55 ]O 1.93 by using hard X-ray photoelectron spectroscopy. The Ni oxidation state increased in the voltage slope region of the 1 st charging process, while it remained unchanged in the plateau region. In the relithiated material, the Ni oxidation state reverted to divalent. Tetravalent Mn ions did not contribute to the charge compensation for the 1 st charging process but were partially reduced on the discharging to 2.0 V. Then, the low-valence Mn ions participated in the delithiation/relithiation process on the subsequent cycles. It was clearly shown that the O À species were formed in the material. The cation-anion dual charge compensation was suggested in the voltage slope region with the aid of theoretical calculations. We emphasise that HAX-PES is an excellent technique to understand the contribution of the lattice oxygen in charge compensation, because it enables us to get chemical state information on the lattice structure buried beneath the surface-oxygenated products due to its deeper probing depth than in-house XPS.
The Ni ions in the LLO lattice structure are considered to increase the capacity delivered by a (partially) reversible O 2À /O À redox couple. This is related to the suppression of O 2 release due to the strong covalent bonding between Ni 4+ and O À ions in the LiMO 2 -and Li 2 MO 3 -like domains. As a result, the discharging capacity exceeds 200 mA h g À1 . The O 2 release, if any, would come from the Li 2 MnO 3 domain in the LLO lattice structure. Therefore, the homogeneous Ni incorporation into the lattice, that is a shi from the "composite" to "solid solution" structure, may be a key factor to design a fully reversible anion redox system. Furthermore, we suggested the structural rearrangement occurring in the sub-surface region in the cycled material. During the 5 th charge-discharge cycle, the O À species were found to be electrochemically inactive. The divalent Ni ions were also inactive, but the oxidation state of Mn ions reversibly changed. It is very confusing because Ni and O ions were important redox-active elements in the 1 st cycle. We therefore consider that structural deterioration that occurred in the cycled material is related to the formation of spinel-or rocksaltlike domains over the sub-surface region (ca. 50 nm thickness) of the particles.