INTRODUCTION

A paradigm that oxide cathodes deliver reversible energy based on the redox activity of cations (transition metals) alone has been challenged by the fact that in many oxide electrodes, anions also take part in redox reactions. Li-rich layered materials and Li-excess disordered rocksalt (DRX) materials deliver excess capacity than the calculated capacity from the redox activity of the transition metals alone. 1 Today, it has been widely accepted that the excess capacity can be attributed to the redox activity of oxygen ions in addition to the transition metal ions. 2 The cumulative cationic and anionic redox has enabled the achievement of enhanced capacity and energy density. Hence, the development of materials combining anionic and cationic redox activities has become a focal point of research in Li/Na-ion batteries. 3,4 Layered LiCoO 2 is one of the first commercial cathodes for Li-ion batteries and has enabled the widespread application of Li-ion batteries in many consumer electronics. 5 Further development of these cathode materials involved the partial substitution of Co by Ni and Mn (widely known as NMC materials). 6 NMC materials with high Ni content are desirable because of the high discharge capacity. 7,8 Due to the geopolitical issues with Co mining, a recent incentive is to reduce or completely eliminate Co from cathode materials. 9,10 Li-rich layered materials have attracted particular attention as an alternative to Co-based conventional NMC cathodes because they can provide larger than the theoretical capacity based on the cationic redox. [11][12][13] Initially, it was thought that the irreversible oxygen extraction and structural transformations account for the large irreversible capacity. 11,14 Later on,

Progress and Potential

Oxygen redox plays a critical role in breaking the energy density barrier of oxide cathodes for alkali-ion batteries. The development of advanced cathodes may take advantage of controllable oxygen redox along with conventional cationic redox. However, such development will require a clear understanding of oxygen redox chemistry. With the utilization of complementary multiscale synchrotron characterizations, the understanding of oxygen redox has developed over time. However, oxygen redox can polarize the opinion of experts as the static nature of oxidized oxygen, molecular oxygen, and peroxide formation has been reported. Meanwhile, prominent challenges such as voltage hysteresis and fading have been attributed to transition metal migration, oxygen evolution, and slow oxygen redox kinetics. However, a consensus is yet to be achieved. This review focuses on bridging many mechanisms and issues with oxygen redox so that a convenient understanding can be achieved by the reader. through various studies, the reversible nature of the lattice oxygen redox was investigated. [15][16][17][18] Formulation of Li-excess DRX materials has further thrust oxygen redox chemistry into the limelight. 19,20 These materials deliver high capacity owing to the combined anionic and cationic redox activities without the need for a well-defined structural ordering for Li percolation. Hence, many transition metals can be utilized for designing these materials, which also vastly expands the compositional space for developing cathode materials with cheap transition metals. 21,22 Charging conventional layered oxides to a very high voltage (close to 5.0 V) has been reported to lead to lattice oxygen redox. 23,24 In a conventional layered material, electron holes can be localized in the O 2p molecular orbital, leading to the oxidation of oxygen anions at high states of charge. 25 Meanwhile, Ceder and coworkers showed that the Li-O-Li configurations are important for forming oxygen 2p orbitals prone to oxygen redox, which can be achieved in Li-rich stoichiometry or through structural disordering. 15 While some of the models explain the oxygen redox to be static in nature, 15 other models predict them to be dynamic. 26 For example, some studies claim structural reorganization takes place on the oxidized oxygen to form peroxo-like species 27 as well as transition metal migration that can modify the crystal and electronic structures of cathode materials. 28 Voltage fadings are significant issues in these materials, which are claimed to be caused by irreversible transition metal migration and could be accelerated by oxygen redox, although the exact mechanism remains debatable. [28][29][30][31] It is clear that there are various mechanisms being put forward toward understanding oxygen redox and related stability issues. In this review, we compile a consolidated report on the mechanism of oxygen redox and related issues with irreversibility of oxygen redox. A mechanistic explanation is first presented by providing accounts of both static and dynamic nature of oxygen redox. This is followed by a discussion of Li and Na cathode materials in which oxygen redox was confirmed through spectroscopic and computational studies. Challenges with oxygen redox and mitigation strategies are explained thereafter.

MECHANISMS OF OXYGEN REDOX

Oxygen redox in oxide electrodes mainly refers to the evolution of oxidation states of lattice oxygen that leads to reversible capacity during cycling. 32 The voltage of the conventional layered oxide cathodes is limited by the top of the O 2p bands. In layered oxides, removal of alkali ion during charging strengthens the transition metal (TM)3d-O2p hybridization. However, at deeper delithiation level (e.g., x > 0.55 in Li 1Àx CoO 2 ), the holes created due to the removal of electron gets trapped in the O 2p orbitals, followed by oxygen evolution. 25 This type of oxygen activity is irreversible and does not lead to usable charge/discharge capacity on longterm cycling. However, recent works show that the oxygen redox at high states of charge in conventional layered oxides is not entirely irreversible and can lead to usable capacity. 23,24,33 Reversible oxygen redox in conventional layered oxides is discussed in more detail later.

Reversible oxygen redox is widely studied in Li-rich materials. The work of Ceder and coworkers explored the structural and chemical characteristics of triggering oxygen redox in Li-rich layered oxide cathodes. 15 They performed density functional theory (DFT)-based calculations utilizing the Heyd-Scuseria-Ernzerhof (HSE06) hybrid functional on a series of Li-rich layered and disordered rocksalt materials (Figures 1A-1C). Their calculation showed an increase in partial density of states (pDOS) from the oxygen states between 0 and À2.5 V (Figure 1B). The charge density plot around oxygen confirmed that within the shaded region of the pDOS plot, the charge density resembled an isolated O 2p orbital lying in the direction of the Li-O-Li axis (Figure 1C). The enhanced charge density around the oxygen ion because of the Li-O-Li configuration near the Fermi level indicated that these electrons from oxygen are prone to participate in redox reactions during electrochemical cycling. This can be further visualized by the schematic of the electronic band structure of the Li-rich materials in Figure 1D. The O 2p orbital in the Li-O-Li configuration cannot hybridize with the Li 2s orbital because of the large energy difference. Hence, these O 2p (F) Electronic structure evolution in charge transfer systems upon alkali-ion removal. The O 2p orbital splits into s, p, s*, and p* states with increasing anionic capacity, as highlighted by the red and yellow triangles. Adapted from Ben Yahia et al. 26 orbitals remain unhybridized. The energy of these unhybridized orbitals is higher than that of the bonding orbitals (e.g., t b 1u ) but lower than that of the antibonding metal states (e.g., e* g ). Whether these unhybridized O 2p orbitals or the transition metal cation will oxidize first depends on the energy level of the d states of the transition metal ions. If a transition metal (e.g., Mn 3d) 15 has a higher energy level of the 3d orbital compared with another transition metal (e.g., Ni 3d), 15 the corresponding e g * state will also be higher in energy. This will cause a lesser overlap between the e* g state and the unhybridized O 2p orbital, resulting in better utilization of the transition metal redox. Meanwhile, in certain cases, the oxidized oxygen has been reported to form peroxo-like species. 27 According to the authors, the peroxo-like bond will be formed when the oxygen is bonded to non-transition metals, such as Sb and Sn. The relatively free rotation of the Li-O-Li configuration allows for the formation of a weak sigma bond between two oxygens. Formation of peroxo-like species is difficult for transition metals with partially filled d orbitals because of the strong directionality of the d orbitals (e.g., Mn, Ni, Co). Meanwhile, Doublet and coworkers proposed the charge-transfer gap of Li-rich oxides as an indicator of oxygen redox and its reversibility. 26 The authors classified the oxide cathodes into two categories: Mott-Hubbard system (partially filled metal bands) and charge-transfer systems (empty metal bands). 34 In the Mott-Hubbard-type Li-rich oxides, the partially filled metal bands will empty first, resulting in the destabilization of the O 2p lone pairs and the stabilization of the (M-O)* states (Figure 1E). At the delithiation point beyond the degeneracy level of these two bands, structural distortion must take place to form M-(O-O) species to stabilize the oxidized oxygen. Such coupling will take place with the concomitant reduction of the transition metals, hence the name ''reductive coupling mechanism.'' In contrast, in the charge-transfer type Lirich metal oxides, the anionic band will empty first on delithiation and the oxidized oxygen must recombine to satisfy the octet rule (Figure 1F). This recombination will lead to the splitting of the anionic band (marked by |O 2p in Figure 1F) to the formation of p, p*, s, and s* bands. The separation between the filled anionic band and empty metallic band (M(d)), D CT , and D p O-O are taken as pertinent parameters to predict oxygen redox reversibility. As long as D CT is bigger than D p O-O , the oxygen redox will be reversible and a reductive coupling mechanism can take place. However, when D CT is smaller than D p O-O (with an increase in anionic capacity), oxygen evolution takes place through a ''reductive elimination mechanism.'' Doublet and coworkers argued that the coupling between oxidized oxygens (formation of peroxo-like species) and/or transition metal migration are crucial factors in stabilizing oxygen redox irrespective of the partially filled or empty metal d bands. Further computational work by them showed that formation of peroxo-like species is possible in Li 2-x RuO 3 and that the formation of this species is not linked to the presence of elements such as Sn, Ti, and Mn, 35 much in contrast to other studies that emphasized the importance of non-transition metals for the formation of peroxolike species. 15 A study by Zhou and coworkers claimed to visualize peroxo-like species in Li 1.2 Ni 0.2 Mn 0.6 O 2 through X-ray diffraction (XRD) and Raman spectroscopy. 36 Peroxo-like species are also reported in Li 2 Ru 1Ày Sn y O 3 based on X-ray photoelectron spectroscopy (XPS) study. 37 However, the rationalization of oxygen redox based on XPS was challenged recently by the work of Piper and coworkers. They mention that the O1s peak in the region of 530-531 eV observed at high states of charge can be rationalized in terms of transition metal reduction and electrolyte decomposition rather than lattice oxygen oxidation. 38 Recent studies have predicted that the oxygen redox cannot be explained by a rigid O 2À /O À redox couple and must take into account structural and electronic changes upon oxygen redox. 28,39 In Li/Mn-rich materials (abbreviated as LMR), oxygen oxidation is accompanied by transition metal migration to the Li site. 28,30 Such migration changes the electrostatic environment surrounding the oxidized oxygen and shifts the O 2p states to higher energy compared with the TM-O hybridized states. This reshuffling reduces the redox potential of the bulk oxygen redox couple during electrochemical cycling, which according to the authors cannot be explained by a rigid O 2À /O À redox model. It is evident that the true chemical nature of the oxidized oxygen is still under debate, and the understanding of oxygen redox is likely to evolve over time.

Nevertheless, the visualization of peroxo-like species in Li cathodes with oxygen redox was reported by McCalla et al. based on their work in Li 2 IrO 3 . 27 The oxidation of O 2À was studied through XPS spectra at different states of charge (red peak in Figure 2A), which took place concomitant to Ir 4+ oxidation. The authors argued the oxidized oxygen formed peroxo-like species. The formation of this peroxo species takes place through a local distortion of the oxygen lattice. Such distortion was probed by high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM), annular brightfield scanning transmission electron microscopy (ABF-STEM) (Figures 2B-2E), DFT calculations (Figure 2F), and structural analysis through neutron powder diffraction (Figure 2G). The authors argued that since the material assumed an O1 type structure at the end of charge, the visualization of the oxygen lattice was possible in the c direction without the interference from Ir or Li. Their analysis showed that the symmetry of the IrO 6 octahedron decreased from sixfold to threefold because of the formation of the shortened (black dumbbells in Figure 2D) and lengthened O-O separations (left blank). Such shortening of the O-O distance was attributed to peroxo-like species through DFT calculations (red dumbbells in Figure 2F) and neutron powder diffraction refinement (Figure 2G). The Fukui function (yellow surface in Figure 2F) displays overlapping lobes in the direction of O-O pairs, which the authors attributed to partially empty antibonding s* orbitals from peroxo-like species.

Recently, Bruce and coworkers proposed molecular O 2 formation as one of the models of lattice oxygen redox. 40,41 In a material where alkali ions in the transition metal layer are in honeycomb ordering with the transition metals, all oxygen atoms are coordinated with two transition metals. Upon oxidizing lattice oxygen, alkali ions in the transition metal layer can move to the alkali layer and transition metal migration can break the degeneracy in energy, forming oxidized oxygen coordinated with three to no transition metals. The oxidized oxygen coordinated with no transition metal can stabilize through dimerization with oxygen coordinated with one transition metal. Upon discharging, the O-O bond is cleaved through the reduction of the unoccupied states in O 2 . 41 Molecular oxygen formation on the bulk of the material has been supported by resonant inelastic X-ray scattering (RIXS), which shows energy-loss peaks corresponding to O-O bond vibrations similar to molecular O 2 . 41 The proposed model of molecular oxygen formation has provided an alternative explanation to the oxygen redox chemistry. On the other hand, some studies mention subtle nuances in the RIXS maps of molecular O 2 and oxidized lattice oxygen in battery materials. 42 It is clear that the nature of lattice oxygen redox still remains elusive, and further investigations are required.

In summary, oxygen redox in battery cathodes is closely related to the energy of the TM (n)d bands and the O 2p bands. 25 At deep delithiation/desodiation level, electron holes can be localized on the O 2p bands, making them redox active. Li-rich stoichiometry has enabled tuning of the oxygen redox chemistry in battery cathodes. The work of Ceder and coworkers has established Li-O-Li configurations as a critical structural unit for triggering oxygen redox in Li-rich layered and DRX materials. 15 Unhybridized O 2p orbitals are created along the Li-O-Li configurations which are close to Fermi level, making them potentially redox active. Even though there is general consensus about triggering oxygen redox in Li-rich cathode materials, the account of the consequence of oxygen redox varies significantly. Some studies point out structural reorganization to stabilize oxygen redox with coupling between the oxidized oxygen, irrespective of the material composition. 26,35 Other studies indicate the type of the transition metal element to which the oxidized oxygen is bonded is critical for peroxo-like dimer formation. 15 Peroxo-like species, molecular O 2 formation, and localized electron holes in O 2p orbitals have been reported in the literature. 27,41,43 Indeed, it is critical to truly reveal the chemical nature of oxidized oxygen in the battery cathode materials. With the advancement of characterization techniques for probing oxygen redox, attainment of such understanding might become possible, as discussed in detail later.

CHARACTERIZATION OF OXYGEN REDOX REACTIONS

Many studies probe oxygen redox through XPS. [44][45][46] However, XPS is a surface-sensitive technique that can give information about the redox evolution only on the surface region of particles and electrodes. Hence, XPS cannot provide conclusive evidence as to whether such redox process is a bulk phenomenon or only limited to the surface. It is notable that by utilizing high-energy synchrotron X-rays, the probing depth of XPS can be enhanced to 40 nm at 10.0-keV incident energy. 45 Another challenge facing surfacesensitive XPS is associated with the formation of complex cathode-electrolyte interphase that may interfere with the XPS signal from the lattice oxygen. Fitting of the O 1s XPS spectrum shows a feature in the range of 530-531 eV, which has been attributed to the lattice oxygen redox process. However, Piper and coworkers attributed the peak within this range to transition metal reduction and electrolyte decomposition, 38 which puts into question the reliability of fitting the O1s XPS spectrum to show lattice oxygen redox. O K-edge X-ray absorption spectroscopy (XAS) spectra derived from soft XAS has also been utilized to provide evidence of oxygen redox. However, caution should be taken on interpreting the O K-edge spectrum because it predominantly gives information about the evolution of TM3d-O 2p hybridization on electrochemical cycling rather than probing the localized hole states on O 2p orbitals. 47,48 Nevertheless, since soft XAS directly probes the TM3d-O 2p hybridization, indirect information on oxygen redox reactions can be obtained in some particular cases. On the high-voltage plateau region of materials showing oxygen redox, minimum oxidation of transition metals is reported in many cases. 49 The increase in pre-edge peak area of O K-edge at the high-voltage plateau region can be reasonably attributed to the oxidation of oxygen anions in those cases. 49 Perhaps more direct and convincing spectroscopic evidence of oxygen redox is obtained through RIXS. RIXS has a probing depth similar to that of fluorescence yield of soft XAS and can provide information up to about 100-nm depth of a particle. 50 The reliability of RIXS over XAS for probing oxygen redox reactions stems from the ability to resolve the emission energy of fluorescence photons that is not possible in conventional XAS. Such resolution has enabled discernment of the fingerprint feature of oxidized oxygen, which is typically buried in XAS. An RIXS map (Figure 3A) will consist of features predominantly at around 525-eV emission energy. The excitation energies of 529 eV and 532 eV corresponding to the same emission energy of 525 eV is assigned to TM3d-O 2p hybridization while the excitation energy at >534 eV can be assigned to O 2p bands of O 2À hybridized with TM4s/4p states. Interestingly, for the charged Na 2/3 Mg 1/3 Mn 2/3 O 2 cathode in Figure 3A, an additional feature is observed at 523.7-eV emission energy for corresponding 531-eV excitation energy. Such experimental observation, along with theoretical calculation based on the OCEAN package, of Li 2 O 2 has shown that the feature is generated from the O 2p-O 2p intraband excitation. 51 This interaction requires unoccupied O 2p orbitals, which are absent in O 2À and is a sign of oxygen redox. However, further studies of Li 2 O 2 , O 2 , and CO 2 systems indicate that this RIXS feature interpretation is beyond a simple molecular model with unoccupied O 2p states because the highly covalent CO 2 system does not display such a feature. 42 At this time, theoretical calculation of the oxygen redox signature in RIXS maps remains a major challenge but holds the promise to reveal the fundamental mechanism of oxygen redox. For example, resolving the energy of the emission photons reveals critical information on the nature of the lattice oxygen redox reactions, including subtlety in RIXS signals of Li 2 O 2 , O 2 , and materials showing lattice oxygen redox. 42 This warrants further investigation to reveal the true nature of the lattice oxygen redox. Nonetheless, integrating the red boxed region showing the lattice oxygen redox (Figure 3A) yields a spectrum (red spectrum in Figure 3B) that has clearly different spectral features than when the spectrum is obtained (blue spectrum in Figure 3B) by integrating the blue boxed region (containing the intensity map of pre-edge and O K-edge in Figure 3A). However, integration of the RIXS intensity map on a much wider excitation and emission energy range (green box in Figure 3A) would yield a spectrum wherein this distinction of the oxygen redox signal is completely absent (green spectrum in Figure 3B). The features for this integrated spectrum (TM3d-O 2p hybridization and O 2p bands of O 2À ) match with the energy for the conventional XAS and show why the conventional XAS cannot decipher the true lattice oxygen redox. Furthermore, the bulk sensitivity of RIXS can be enhanced if the surface signal is differentiated from the bulk. 47 Spatially resolved scanning transmission X-ray microscopy has shown a feature at 531-eV excitation energy on the bulk of charged Li/Mn-rich materials. 28 The RIXS map shows a feature at the same excitation energy for the oxidized lattice oxygen (Figure 3), which proves that the oxygen redox is a bulk phenomenon.

Pair distribution function (PDF) analysis can probe the local structural rearrangement associated with oxygen redox reactions. 52 By utilizing both Bragg diffraction and diffuse scattering signal, PDF can provide local structural changes and short-range ordering. 53 Neutron PDF holds an advantage over X-ray PDF because it is sensitive toward lighter atoms such as oxygen. Hence, most local structural changes associated with the lattice oxygen framework can be probed by neutron PDF analysis. Through utilizing such a technique, short O-O pair formation similar to peroxo-like dimer has been reported in charged P3-type Na 0.6 [Li 0.2 Mn 0.8 ]O 2 samples, similar to Li 2 IrO 3 . 27,52

MATERIALS SHOWING OXYGEN REDOX

Oxygen Redox in Alkali-Rich Layered Transition Metal Oxides Several research groups have extensively explored the synthesis, structural, and electrochemical characterization of various Li-rich layered oxide materials (Li 1+x TM 1Àx O 2 , where TM is a 3d transition metal or a combination of different transition metals). 11,12,[54][55][56][57][58][59][60][61] A large characteristic voltage plateau and irreversible capacity were observed in these materials at the first cycle (Figure 4). Initial theories of the excess capacity included Mn oxidation to +5 state, irreversible extraction of oxygen on the high-voltage plateau region, 11 and extraction of Li 2 O (lithia) from the Li 2 MnO 3 domain of the composite structured Li-rich cathodes. 58 However, a reversible participation of lattice oxygen was not mentioned. The contribution of oxygen to charge compensation without oxygen loss on Li-rich layered Li[Li 0.15- Ni 0.275Àx Mg x Mn 0.575 ]O 2 was speculated by Amine and coworkers back in 2003 based on their XAS study. 59 An increase in pre-edge peak area was taken as a sign of oxygen redox, and it is now well known that the pre-edge feature is mostly dominated by TM3d-O 2p hybridization and not the localized hole states on the O 2p orbitals, as discussed in the Introduction.

The reversible/irreversible nature of the surface-to-bulk oxygen redox activity in Lirich layered materials was investigated by the works of Delmas and coworkers in 2013 62,63 and later by Bruce and coworkers. 16,64 Delmas' group argued that the observed initial discharge capacity of more than 270 mAh/g in Li 1. 20 Mn 0.54 Co 0.13- Ni 0.13 O 2 can only be explained if both irreversible oxygen redox in the form of oxygen loss from the surface of the particle and reversible oxygen redox in the bulk of the particle are taken into account. 62 Oxygen redox in Li 1. 20 Mn 0.54 Co 0.13 Ni 0.13 O 2 was further probed and quantified by Bruce and coworkers through operando differential electrochemical mass spectrometry (DEMS), Raman spectroscopy, XAS, and RIXS. 16 To estimate the irreversible nature of the oxygen redox, the authors labeled the oxygen in the material with 18 O isotope. Time-of-flight secondary ion mass spectrometry indicated that a total of 15% 18 O enrichment in the lattice oxygen was obtained. Hence, the evolved gases from the irreversible lattice oxygen redox activity during electrochemical cycling must have a certain ratio of 16 O/ 18 O. Their DEMS study revealed that on the 4.5-V plateau region, even though there was no molecular O 2 evolution, a detectable amount of CO 2 was measured (Figure 5C). The CO 2 evolution was also observed at very high states of charge (close to 4.8 V). The total 16 O/ 18 O ratio in the evolved CO 2 was close to that of the expected ratio if all the evolved CO 2 was accounted for only by the reaction of the lattice oxygen with the electrolyte (expected: 20%; experimental: 13%). Even when taking the discrepancy between the expected and experimental percentages into account, it is clear that most of the CO 2 evolution occurred from the contribution of the lattice oxygen. At high states of charge (close to 4.8 V), significant O 2 evolution occurred (Figure 5B) and the ratio of 16 O/ 18 O in O 2 was close to what was expected if the evolved oxygen came only from the contribution of the lattice oxygen (expected: 26%; experimental: 23%). Even though oxygen evolution could account for most of the capacity obtained at the sloping region beyond the 4.5-V plateau (Figure 5A), the irreversible oxygen redox reactions at the plateau region could only account for 9% of the total charge capacity obtained at that region. Hence, a significant portion of the charge capacity was obtained by a reversible oxygen redox, as DEMS can only measure irreversible oxygen evolution. A reversible oxygen redox is expected to create hole states in the O 2p orbitals but those hole states did not create O-O dimers (no peaks similar to peroxide are observed in Figure 5D), unlike Li-rich materials with 4d or 5d transition metals (see Mechanisms of Oxygen Redox for more details). Instead, the authors argued that the hole states were localized in nature as signified by the broadening of the RIXS spectrum and the formation of an elastic peak at around 532 eV (boxed region in Figure 5E). Hence, the analyses carried out by Bruce and (D andE) Probing the chemical nature of the oxidized oxygen through (D) Raman spectroscopy and (E) RIXS. Gas evolution through DEMS probes irreversible oxygen redox, whereas RIXS probes reversible oxygen redox. Adapted from Luo et al. 16 coworkers indicated that the oxygen redox is both reversible and irreversible in nature as opposed to the early model of only irreversible oxygen redox proposed to explain the enhanced capacity delivered by these materials. 11 Li 2 TMO 3 or Li[Li 1/3 TM 2/3 ]O 2 (where TM is either 4d or 5d transition metals) studied in detail by Tarascon and coworkers also delivers excess capacity through oxygen redox. 27,37,65 In case the Li and TM in the transition metal layer are ordered, only one type of oxygen sublattice where all O should have one Li-O-Li configuration is expected. 26 It is argued that the peroxo-like O-O dimer in materials such as Li 2 RuO 3 and Li 2 Ru 1Ày Sn y O 2 forms through a ''reductive coupling mechanism'' (see Mechanisms of Oxygen Redox for more details). 26,35,37 Such structural reorganization stabilizes oxygen redox in these materials. Similar to Li-rich oxides with 3d transition metals, a large high-voltage plateau is observed, and oxygen redox plays a significant role in the charge-compensation mechanism. 27 Na-rich transition metal oxides with 4d and 5d transition metals and the general formula Na 2 TMO 3 also show oxygen redox activity during electrochemical cycling. [66][67][68][69] Similar to oxygen redox chemistry in Li-rich materials, the understanding of the crystal structure of these materials has also evolved over time. Li-rich materials are thought of either as a single-phase solid solution with trigonal structure where all the Li and transition metal ions are distributed in the transition metal layer with long-range ordering 13,70 or as a single-phase monoclinic structure throughout. 71 They are also viewed as a composite of Li 2 MnO 3 -LiTMO 2 with close structural compatibility, with Li and Mn in the Li 2 MnO 3 domains having honeycomb-like ordering. [72][73][74] The actual crystal structure arrangement has sparked a long-standing debate in the research community. Here, we summarize the claims from both ends (single-phase or composite).

One of the first claims of the composite nature of the Li-rich materials was made by Thackeray and coworkers. 14,75 Experimental verifications of the composite nature of Li 1.2 Mn 0.55 Ni 0.15 Co 0.10 O 2 (commercially known as TODA HE5050) were provided through neutron diffraction and magnetic susceptibility studies. 72 Two models were chosen by the authors for neutron diffraction refinement: a single monoclinic phase throughout with a C2/m space group, and a composite of a monoclinic phase with a C2/m space group and a trigonal phase with an R3m space group. Reasonable fitting was obtained for both models. For example, a fitting model with the composite structure of a monoclinic phase with only Li and Mn and a trigonal phase with all three transition metals and Li is shown in Figure 6A. The Li-and Mn-containing phase had Li/Mn ordering, which explained the superstructure peaks in the pattern. Moreover, including a 3% Li/Ni site exchange increased the reliability of the fitting. To conclusively choose between the composite or single-phase models, the authors further performed a temperature-dependent magnetic susceptibility study (Figure 6B). From the magnetic susceptibility study, it can be seen that the field-cooling (FC) and the zero-field-cooling (ZFC) curves follow each other up to 100 K, which the authors attributed to the disordering of the transition metals in the trigonal phase (one component of the composite structure). At temperature lower than 100 K, the curves diverge from each other due to a magnetic ordering in the material, which the authors attributed to the Li/Mn ordering in the Li 2 MnO 3 phase of the material (the other component of the composite structure). Even though some studies claim the composite nature of the Li-rich materials based on transmission electron microscopy, 73 such multiphase composite structure is highly debated. Dahn and coworkers argued that certain transition-metal-rich domains such as the Li 2 MnO 3 are unlikely because the high-temperature synthesis should not allow the clustering of a particular transition metal from the entropic perspective. 13 The counterclaim against the composite nature of the Li-rich materials is now discussed.

It is necessary to provide visual representations of the phases at the single-particle level in addition to structural refinement analysis to prevent any structural ambiguity. High-resolution TEM (HRTEM) imaging can be one such way, but the small field of view of the HRTEM imaging cannot unanimously prove whether a material is a single phase or of composite nature. Shukla et al. provided a comprehensive TEM analysis on the nature of the phases present in Li 1. 20 Mn 0.54 Co 0.13 Ni 0.13 O 2 by imaging the particles of this material at a much larger field of view (Figure 7). 71 The needle-like particles with small thickness enabled imaging of the whole primary particle with atomic resolution (Figures 7C and7G). HAADF images showed that there was no sign of composite nature in the material. Instead, the crystal structure of the material consisted of three different variants of the monoclinic phases alternating between each other randomly (colored boxes in Figure 7A and colored rows in Figure 7B). The three variants of the monoclinic phases are defined by the projections along the [100], [110], and [110] directions, as shown in Figures 7D-7F. Meanwhile, imaging at a much larger field of view did not show any presence of the composite nature of the material (Figures 7C and7G). The authors made similar observations in commercial Li-rich TODA HE5050 material, in contrast to the study summarized earlier. 72 It must be mentioned that whether the structural ambiguity with the Li-rich materials was raised because of the different synthesis method and composition used by different groups or from the misinterpretation of the TEM data is not clearly understood. Shukla et al. provided one such example that because of the different stacking of more than one monoclinic variant in the thicker regions of the particle, those regions may assume hexagonal ordering when viewed from a certain direction. This may lead to the erroneous assignment of trigonal phases to these regions. More study is needed to clear up the confusion regarding the composite versus singlephase debate in the Li-rich materials. However, it seems that the oxygen redox chemistry takes place in these materials independent of the debate.

Oxygen Redox in Conventional Layered Transition Metal Oxides

The contribution of oxygen to the charge-compensation mechanism in conventional layered LiTMO 2 with no Li excess (e.g., LiCoO 2 , LiNi 0.5 Co 0.5 O 2 ) was claimed in the early 2000s, [76][77][78] even before the general acceptance of significant reversible oxygen redox in Li-rich materials. However, these studies were based on observations 72 made from O K-edge and TM L-edges from soft XAS, which, as mentioned before, cannot probe the localized hole states on the O 2p orbitals. Hence, a conclusive claim on oxygen redox in conventional layered materials could not be reached. However, recent studies showed that oxygen redox could be achieved in conventional layered materials at very high states of charge (SOCs). 24,33,79 In Ni-based layered oxides, Ni in +3 oxidation state has an electronic configuration of 3d 7 (t 2g 6 e g 1 ) with

partially filled e g orbital, which means electron removal will first take place from the e g orbital. Hence, electron removal from O 2p orbital will take place later at high SOCs. 80,81 This might give an indication of the high-voltage behavior of many technologically significant Ni-rich layered NMC materials (LiNi x- Co y Mn 1ÀxÀy O 2 ), since LiNiO 2 can be taken as the parent material for this series of materials. Besides, high-voltage charging pushes the O 2p states closer to the Fermi level, making them redox active. 82 Utilizing RIXS on these materials at high SOCs allowed deciphering of the localized hole states in the O 2p orbitals contributing to the capacity at high voltages.

Oxygen Redox in Alkali-Poor Layered Transition Metal Oxides

The work of Ceder and coworkers has established the Li-O-Li configuration as an important structural descriptor to trigger oxygen redox in Li cathodes. 15 Such a descriptor can be obtained in large quantity in Li-rich layered oxides and DRX materials. In Li-rich DRX materials there are different oxygen sublattices, some of which have the Li-O-Li configuration because of the disordering of the transition metal cations, whereas in Li-rich layered materials the excess Li sitting in the transition metal layer enables the formation of the Li-O-Li configurations. 15 Sodium layered cathodes (G) HAADF image of the entire primary particle shows that only one phase is present. The marked region in red and the corresponding inset show the occasional transition-metal-rich defects observed in the particle. Adapted from Shukla et al. 71 provide a unique scenario with respect to oxygen redox and the alkali stoichiometry.

Unlike their Li counterparts which largely form O3-type structures, sodium layered cathodes can be synthesized in different polymorphs (such as O3, P2, and P3). 83 O3 structure is empirically stabilized when the Na stoichiometry is close to 1 in Na x TMO 2 . 84,85 The P2 and P3 polymorphs are stabilized in sodium-poor stoichiometry (x is less than 1 in Na x TMO 2 ). 86 Na-rich stoichiometry is attainable with 4d (e.g., Ru) 67 and 5d (e.g., Ir) 69 transition metals because of the smaller size difference between Na ion, and Ru or Ir ions. Hence, the Na-O-Na configuration can be found in sodium transition metal oxides with 4d and 5d transition metals, and oxygen redox has been reported in these materials. 68,69,87 However, for the case of sodium layered oxides with 3d transition metals (e.g., Ti, Cr, Mn, Fe, Co, Ni, Cu), the larger size difference between the transition metal ion and the Na ion does not allow the Na ion to stably occupy the transition metal layer. 83,88 Hence, attainment of the Na-O-Na configuration is not favored in sodium layered oxides with 3d transition metals. However, the almost similar size of Li ion and Mg ion compared with the transition metal ions allows them to replace some 3d transition metal ions in the transition metal layer. 43,49,89 Hence, the formation of Na-O-Li or Na-O-Mg configuration becomes possible. O 2p orbitals in these configurations are also expected to be unhybridized and, thus, redox active. 90 Additionally, cation vacancies are reported to increase the O 2p lone pair electrons and can trigger oxygen redox in alkali-poor sodium layered materials, 26 Structural characterization such as XRD, neutron diffraction, and Li-nuclear magnetic resonance (NMR) can give direct evidence of the specific lattice sites Li and Mg can occupy in these materials. [93][94][95] Combined XRD and neutron diffraction analysis with Rietveld refinement in P2-Na 0.66 Li 0.18 Fe 0.12 Mn 0.7 O 2 indicated the occupancy of Li in the transition metal layer (Figures 8A and8B). An in-plane Li/TM ordering can be defined by a ffiffiffi 3 p a 3 ffiffiffi 3 p a supercell in this material. 89 The ordering is similar to typical honeycomb ordering of Li/Mn observed in Li-rich materials. 96,97 The honeycomb ordering can be assumed by alkali/alkaline-ion substitution in the transition metal layer, as shown in Figure 8D. Li-NMR can provide information of Li occupancy in the transition metal layer, e.g., in Li-substituted P2-Na 0.78 Li 0.25 Mn 0.75 O 2 by a characteristic shift at 1,850 ppm (Figure 8C). 49 Direct participation of oxygen to redox reactions has been documented in Na x [Li y TM 1Ày ]O 2 and Na x [Mg y TM 1Ày ]O 2 through various spectroscopic techniques. A reversible oxygen redox was observed in Na 0.6 [Li 0.2 Mn 0.8 ]O 2 through RIXS by a characteristic increase in the intensity at 531-eV excitation energy and a corresponding 523.7-eV emission energy (boxed region in Figure 9A). 32 Bruce and coworkers reported a change in spectral weight in O K-edge at high SOCs in Na 2/3 [Mg 0.28 Mn 0.72 ]O 2 , indicative of oxygen redox (Figures 9B and9C). 43 The O K-edge spectral evolution studied in alkali-poor cathode materials is similar to oxygen redox active Li-rich cathode materials reported in the literature. 28,30 Recently, Yang and coworkers reported the contribution of oxygen redox in charge compensation of layered Na 2/3 Ni 1/3 Mn 2/3 O 2 during cycling. 23 The oxidation of oxygen anion is particularly observed during the 4.2-V plateau on charging. This oxygen redox activity is fundamentally distinct from the other sodium layered oxides where transition metals are substituted with Mg and Li. The oxidation of oxygen anion at high voltage can be similar to the oxygen redox activity observed in conventional stoichiometric layered oxide materials (see Mechanisms of Oxygen Redox and Oxygen Redox in Conventional Layered Transition Metal Oxides for more details).

Oxygen Redox in Disordered Rocksalt Oxides Li-rich cation-DRX materials have opened up a vast compositional space to explore new Ni/Co free high-energy cathode materials for Li-ion batteries. These materials can utilize combined cationic and anionic redox chemistry to deliver large discharge capacity and energy density. DRX materials have a face-centered cubic structure. Normally, these materials do not have an open percolation network for Li insertion/de-insertion. However, having a Li-rich stoichiometry opens up a continuous percolation network for Li ions and allows for electrochemical cycling of Li ions. 19 In DRX materials, the Li percolation takes place by hopping of the Li ion from one octahedral site to another octahedral site through an intermediate tetrahedral site (Figure 10A). Li ion in the tetrahedral site is regarded as an active Li for the hopping. In DRX materials, since many oxygen sublattices exist, the intermediate tetrahedral site for the Li hopping can be face sharing with no-TM (0-TM in Fig- ure 10B), 1-TM (Figure 10C), or 2-TM (Figure 10D) octahedral sites. 19 The electrostatic repulsion between a Li in the tetrahedral site and the surrounding octahedral TM can indicate whether one tetrahedral site will be feasible for Li percolation. The height of the tetrahedral site (about 2.35-2.45 A ˚) can be taken as a measure of the activation energy required for Li hopping. Owing to the small height of the tetrahedral site, the 1-TM channel in DRX materials has an activation energy of Li hopping of more than 500 meV and is usually inactive for Li hopping. However, these 1-TM channels are active for Li hopping in ordered layered transition metal oxides because of the larger interslab distance. Hence, only 0-TM channels are active for Li percolation in DRX materials. For macroscopic Li-ion conduction, a continuous network of these 0-TM channels is required. Monte Carlo simulations showed the critical Li concentration for the formation of sufficient 0-TM channels in DRX 43 materials for continuous macroscopic Li diffusion. 98 Figure 10E shows the periodical wrapping probability for various combinations of percolation channels in DRX materials. A cluster of sites are periodically wrapping if the cluster has different periodic images of the same Li site. Since 0-TM channels are likely to be the main percolation channels for DRX materials, it can be seen from Figure 10E that at least 10% Li-excess composition is required for the 0-TM channels to percolate Li ions (red dashed line). Furthermore, the fraction of accessible Li ions from the DRX materials for electrochemical cycling increases with increasing Li composition. For at least one accessible Li per formula unit, an excess of 25% Li in the composition is required (red dashed line and black horizontal line in Figure 10F). Hence, this shows that excess Li composition is required for DRX materials to electrochemically cycle Li ions. The rule of thumb of excess Li composition can be applied to most DRX materials and is independent of the type of the transition metal ions. Disordering of the transition metals in the rocksalt structure is facilitated with transition metals of d 0 electronic configuration while transition metals with d 6 electronic configuration prohibit disordering. 99 Cation disordering can affect the properties of cathode materials such as Li-ion conductivity. For example, in Li Adapted from Wu et al. 32 and Maitra et al. 43 Because of the disordered nature of the cation distribution in DRX materials, there can be significant numbers of Li-O-Li configurations, 15 and some oxygens may have more than one Li-O-Li configuration (e.g., the local Li-rich environment such as that in 0-TM channels). 98 Hence, oxygen redox activity is expected in these materials. 101 Indeed, Yabuuchi et al. reported disordered Li 1.3 Nb 0.3 Mn 0.4 O 2 with large discharge capacity due to the combined cationic and anionic redox activities. 20 The authors of this study performed partial density of states (pDOS) calculation on the material with the general formula Li (4Ày)/3 Mn 1/3 Nb 1/3 O 2 (0 % y % 4) to understand the participation of oxygen on charge compensation (Figure 11). Charge compensation in this material at the initial y values (0-1) takes place mainly from the Mn e g orbitals of the MnO 6 octahedra according to the calculated density of states from the top of the valence band at y = 0 and the bottom of the conduction band at y = 1 (green circles on the right panel of Figure 11 for the y values 0-1). For the y values 1-3, O 2p orbitals mainly take part in redox reaction as can be seen from the top of the valence band at y = 1 (blue circle in the right panel of Figure 11) and the bottom of the conduction band at y = 2, 3 (blue circles on the right panel of Figure 11), and they are mainly composed of dumbbell-shaped O 2p orbitals. The drastic change in the density of states on the Mn e g orbitals at higher SOCs (y = 4) indicates that further Mn oxidation can take place at deeper delithiation level. Lattice oxygen redox in DRX materials has also been substantiated by experimental evidence through techniques such as RIXS. [102][103][104][105] However, quite often the feature indicating lattice oxygen oxidation appears as only a shoulder because of the strong overlap of TM-4d states in second-row transition metal-containing DRX materials. 103 In addition, it has been reported that the lattice oxygen redox does not lead to distortion of the lattice oxygen framework with reduced O-O separation. 102 Due to the random distribution of the cations in DRX materials, the formation probability of coplanar unhybridized O 2p orbitals greatly reduces, leading to the lesser interaction between the oxidized oxygens.

Utilizing a vast array of compositions, various DRX materials have been reported with large capacity and energy density. 20,[106][107][108][109] Fluorine substitution strategy and incorporating d 0 transition metals with high valence states can keep the valence states of the redox active species low, thus enabling facile redox reactions of the redox active transition metal ions. 4,[110][111][112] For example, the lower valence state of Mn achieved in Li 2 Mn 2/3 Nb 1/3 O 2 F through the incorporation of high-valence-state Nb 5+ and F substitution has enabled the use of Mn 2+ /Mn 4+ redox couple in this material. 113 In addition to a small amount of oxygen redox, this material can deliver more than 300 mAh/g discharge capacity and almost 1,000 Wh/kg discharge energy, although the lower cutoff voltage has to reach 1.5 V versus Li/Li + . Due to the absence of specific ordering in DRX materials, these materials can utilize a wide variety of transition metals, which can be crucial for developing a cost-effective alternative to more conventional ordered lithium layered oxide materials with expensive Ni/Co as the transition metals. 4 In summary, various cathode materials can utilize oxygen redox and transition metal redox to deliver high capacity. For the convenience of the reader, we have summarized below (Table 1) the capacity contribution of oxygen and transition metal redox during initial cycling when the data were reported in the literature.

ISSUES WITH OXYGEN REDOX Voltage Hysteresis and Fading

Even though oxide cathodes showing oxygen redox are a promising group of materials due to their high discharge capacity, they present significant challenges toward 20 materials development, which has hindered their practical applications. 114 Voltage hysteresis and fading, and low initial Coulombic efficiency (Figure 12) are some of these challenges, which require further understanding and mitigation strategies. 115 A large voltage hysteresis, which refers to the difference in voltage between charge and discharge cycles, is especially observed in the first cycle. Voltage hysteresis leads to a decrease in energy efficiency and is persistent throughout the electrochemical cycling of most of the cathode materials with oxygen redox. The continuous decrease of the average discharge voltage referred to as voltage fading renders subpar energy output even when the capacity is not fading drastically. 115 Meanwhile, the low first-cycle Coulombic efficiency indicates irreversible structural and chemical transformations. 116 In general, some of the reasons behind voltage fading include but are not limited to (1) irreversible transition metal migration, 32,39,117,118 (2) emergence of new redox couples, 30 and (3) oxygen evolution. 26,101 Often these processes are interconnected; for example, irreversible oxygen evolution can cause irreversible transition metal migration and associated structural changes. 101,119,120 Meanwhile, some of the reasons behind the particularly severe voltage hysteresis between the first charge and first discharge have been attributed to the following: (1) charge-transfer band gap, 121 (2) asymmetric redox couple evolution, 28 (3) sluggish nature of oxygen redox, 29 and (4) type of ordering of alkali-ion/transition metal ion in the transition metal layer. 41 Tarascon and coworkers reported out-of-plane transition metal migration to the Li layer in Li 2 Ru 0.75 Ti 0.25 O 3 . 117 The migration can take place from one octahedral site to another through an intermediate tetrahedral site. Such migration is reported to be reversible to some degree, but the migrated transition metal can get entrapped in the tetrahedral site on repeated cycling. The entrapment of the transition metal ions in the tetrahedral site has been attributed to the continuous voltage fading of Li 2 Ru 0.75 Ti 0.25 O 3 . 117 The study by Chueh and coworkers showed that transition metal migration to the Li layer is necessary for triggering and stabilizing oxygen redox in materials such as Li 2Àx Ir 1Ày Sn y O 3 , even though such migration may lead to voltage hysteresis. 39 An out-of-plane migration of Sn to the Li site creates a point defect on the transition metal layer and significantly changes the electronic structure of the material (Figure 13). Meanwhile, the migration of Sn is favored over Ir because of the stabilization effect of Sn on the delithiated structure (stabilization of 1.34-1.36 eV by Sn migration against a stabilization of 0.02 eV by Ir migration). The authors predicted that upon the out-of-plane migration of Sn, one of the two outcomes can take place depending on the type of local structure from which Sn is migrating. 30 The emergence of these new redox couples raises the Fermi level and lowers the operating voltage because of the smaller difference between the Fermi level and the Li 0 /Li + energy level (Figure 14). Such lowering of the operating voltage contributes to the voltage fading of the material. Similar to this observation, Yang and coworkers attributed the voltage fading of P2-Na 0.6 [Li 0.2 Mn 0.8 ] O 2 to the increasing contribution of Mn redox on cycling. 32 Assat et al. evaluated the electrochemical reaction mechanism of Li 1.2 Ni 0.13 Mn 0.54- Co 0.13 O 2 in detail and revealed that the oxygen redox reaction considerably slows down the electrochemical kinetics. 29 A detailed charge-compensation study revealed that small anionic activity and Mn 3+ /Mn 4+ redox is mainly observed on low potential charging (blue shaded region in Figure 15). Meanwhile, the capacity at charging to 4.1 V is predominantly obtained by the redox activity of Ni 2+/3+/4+ and Co 2+/3+ redox active species (gray shaded region in Figure 15). Charging to higher voltage (4.8 V), the charge compensation primarily takes place from the oxidation of O 2À (red shaded region in Figure 15). On subsequent discharge, O nÀ (n < 2) reduces over two voltage regions: the high-voltage region (green shaded area in Figure 15), and low-voltage region (red shaded area with 3.2-V reduction peak in Figure 15). Such asymmetric oxygen redox evolution leads to voltage hysteresis. Tarascon and coworkers proposed charge-transfer band gap as an indicator of voltage hysteresis. 121 The density of states calculation on the disordered Li was attributed to the differences in Li/Mn ordering in the transition metal layer of these two materials (Figures 16E and16F). 41 Na 0.75 [Li 0.25 Mn 0.75 ]O 2 shows honeycomb-type ordering, which is not effectively maintained on cycling (Figure 16E). Na 0.6 [Li 0.2 Mn 0.8 ]O 2 shows ribbon-type ordering that is well maintained (Figure 16F), leading to less voltage hysteresis.

It is clear that a host of reasons have been related to the voltage hysteresis and voltage fading of materials showing oxygen redox. It is thus most reasonable to mention the underlying reasons behind voltage hysteresis/fading as materials specific. However, there seems to exist a strong correlation between transition metal migration and voltage hysteresis and fading. Reversible transition metal migration may lead to voltage hysteresis, but irreversibility in transition metal migration will lead to voltage fading. 122 In some materials, oxygen redox itself due to slow redox kinetics and asymmetric redox evolution has been attributed to voltage hysteresis, even when no appreciable transition metal migration is observed. 43 Indeed, voltage hysteresis has been observed to increase when anionic redox chemistry is involved. 29 Oxygen redox is usually kinetically slow and causes voltage hysteresis on electrochemical cycling. The structural and electronic reorganization that ensued after stabilization of the oxygen redox may also be responsible for the voltage hysteresis. 118,123 Such reorganization caused by the oxygen redox may alter the sequence of cationic and anionic redox reactions (e.g., cationic-to-anionic oxidation on charge followed by cationic-to-anionic reduction on discharge) and cause voltage hysteresis. 30 Meanwhile, irreversible oxygen redox in the form of oxygen evolution may lead to a cation-densified state as a result of irreversible transition metal migration. The accumulation of the irreversible transition metal migration may lead to voltage fading. Irreversible oxygen redox can also lead to transition metal reduction, which causes emergence of new redox couples. Contribution of some of these redox couples such as Co 3+ /Co 4+ and Mn 3+ /Mn 4+ can lower the output voltage during cycling, causing voltage fading. 30 Irreversible Oxygen Redox and Oxygen Evolution Irreversible oxygen redox and oxygen evolution are commonly observed because of the reactive nature of the oxidized oxygen. The irreversibility associated with the O 2À /O À redox couple might lead to permanent structural modifications, 118,[124][125][126] oxygen gas evolution, [127][128][129][130][131] and accelerated electrolyte decomposition through the nucleophilic attack of the oxidized oxygen with the electrolyte. 29,103 Oxide ion in the oxidized state can either form O-O dimers through structural reorganization, or the hole states can be localized in the unhybridized O 2p orbitals. 23,26,43,132 The O-O bonds are kinetically labile toward oxygen evolution. 133 Meanwhile, if the hole formation in the O 2p orbitals is not stabilized, it can lead to oxygen release and accelerated electrolyte decomposition. Indeed, the migration barrier of the oxidized oxygen is found to be significantly lower than that of O 2À , which may contribute to the irreversibility of the oxygen redox. 134,135 Meanwhile, in conventional Li/Na layered materials the oxygen stability is closely related to the SOC. 120,136,137 At higher SOCs, O 2 and CO 2 evolution is observed in LiCoO 2 . 138,139 The thermal stability of the delithiated Li x CoO 2 (x < 1), which can indicate the oxygen stability, decreases significantly in comparison with the pristine LiCoO 2 . 139 The thermal stability is also compromised in delithiated Li-rich materials. 140 Ni-rich layered materials are also prone to irreversible oxygen redox and associated structural transformations. 141 Ni can form facile antisite defects with Li in layered materials, and the process is accelerated at high SOCs. 9,10 Such migration can lead to structural transformation (layered to spinel/rocksalt) and lead to oxygen release. 119 The instability of the Ni ions at high valence states is particularly severe because of the thermodynamic instability of the Ni e g electrons. 139 Various spectroscopic techniques such as RIXS, 24,48,50,142,143 XAS, 48,144,145 and XPS are utilized to probe the reversibility of oxygen redox. 27,94 Meanwhile, DEMS has allowed the direct study of the nature of irreversible oxygen redox in cathode materials (see Oxygen Redox in Alkali-Rich Layered Transition Metal Oxides for more details). [146][147][148] A combined analysis with DEMS and spectroscopic techniques such as RIXS, XAS, or XPS allows the study of the whole range of oxygen redox whether reversible or irreversible. Gasteiger and coworkers showed the evolution of singlet oxygen from Li-rich NMC materials at high SOCs through emission spectroscopy and CO and CO 2 evolution through on-line mass spectrometry. 149 This further signifies the reversible and irreversible nature of oxygen redox observed in these materials.

Structural and Chemical Evolution in Irreversible Oxygen Redox

A gradual structural and chemical transformation triggered by irreversible oxygen redox is usually observed in materials where oxygen plays a significant role in redox reactions. 118 In the literature where Li-rich materials are reported as a composite of Li 2 MnO 3 and layered LiMO 2 , Li 2 O (lithia)-like structural units are claimed to be extracted at high-voltage charging. 14,54 Such irreversible Li and oxygen extraction is accompanied by structural transformations from layered to cationdense structures (spinel and rocksalt). The origin of structural transformations may be involved with irreversible transition metal migration to the Li site, led by oxygen extraction. 150 A transition metal ion surrounded by five oxygens instead of six due to the oxygen release will be destabilized and move toward the vacant octahedral site of the Li layer. 151 Structural transformation in Li-rich materials from layered to spinel to rocksalt is observed, which is more severe at the surface region of the particle. 17). 151 The pristine material can be defined as a well-ordered layered structure with a homogeneous distribution of the transition metals from the surface to the bulk (Figure 17A). However, on prolonged cycling (50 cycles at C/10 rate), a thick reconstructed surface layer emerges, which is composed of defective spinel structure with an outer layer of rocksalt (Figure 17B). Moreover, segregation of transition metal is observed with a higher ratio of Ni at the surface region of the particle in comparison with the bulk (Figure 17B). Such segregation indicates that the surface reconstruction takes place mostly through the migration of the Ni ions. Ni at higher oxidation state (close to +4) is thermodynamically unstable, triggering the migration of Ni ions to the vacant Li sites. 139 Meanwhile, a gradual loss of the superstructure reflection on cycling in the XRD pattern (Figure 17C) indicates progressive disordering of Li and transition metal ions, consistent with the transition metal migration observed in STEM images. Usually the migrated transition metal ions in the Li site are in the reduced state compared with the transition metal ions in the transition metal site (Figure 17D). 119,153 Since the reconstructed layer is a few nanometers thick and mostly limited to the surface, the oxidation state of the transition metals underneath the surface reconstruction layer is higher (Figure 17D).

Oxygen release can cause mechanical degradation of cathode particles. 120 The authors performed an in situ TEM study at 230 C of the delithiated NMC particles in order to accelerate the oxygen release and associated chemomechanical evolutions. Z-contrast STEM images were recorded to observe the nucleation and propagation of cracks at the primary particle level (Figure 18). After a short incubation period of about 7 min, several microcracks grew quickly in length up to 15 min and grew slowly afterward (about 90 nm after 40 min). Such chemomechanical degradation in practical electrodes can lead to several other phenomena such as detachment of the cathode particle from the binder matrix, 155 transition metal dissolution, accelerated cathodeelectrolyte side reactions, 141 and poor electronic conductivity. 156 All of these processes combined can step up the overall degradation of the cathode material.

STABILIZATION OF OXYGEN REDOX AND MITIGATION OF THE STRUCTURAL/CHEMICAL IRREVERSIBILITY Suppressing Voltage Hysteresis and Fading

Voltage hysteresis and fading are major roadblocks to the development of high-energy Li-rich layered materials. Many mitigation strategies to suppress voltage hysteresis and fading have been reported. These methods include cation doping, 157 tuning crystal structure, 158 controlling redox couple evolution, 32 compositional control, 159,160 gradient Li distribution, 161 and tuning the non-electroactive cathode component. 162 Here we summarize these strategies.

Tuning the crystal structure of Li x (Li 0.2 Ni 0.2 Mn 0.6 )O 2 from an O3 to an O2 type has resulted in suppression of voltage fading on long-term cycling and minimum initial (D) STEM image highlighting the regions (in circles) in which EELS spectra were acquired (left) and Mn L-edge EELS spectra on the marked regions of the STEM image (right). Adapted from Lin et al. 119 voltage hysteresis. 158 For the O2-type structure, TMO 6 octahedra in the transition metal layer is face shared with the LiO 6 octahedra in the Li layer (Figure 19B). Transition metal migration from an intermediate tetrahedral site to an octahedral site of the Li layer is not favored because of the electrostatic repulsion between the cations in the face-shared octahedral sites (Figure 19B). The authors argue that the large electrostatic repulsion encourages the transition metal ions in the tetrahedral site to return to the octahedral site of the transition metal layer and thus minimize the initial voltage hysteresis and voltage fading on long-term cycling. 158 Similar to crystal structure tuning, careful compositional control has resulted in suppression of voltage fading in Li-rich materials. For example, Kang and coworkers showed that increasing the Ni content from 20% to 40% in Li 1.2 Ni x Mn 0.8-x O 2 can improve the voltage stability. 159 The underlying mechanism is the introduction of Ni 3+ species with increasing Ni content in the material. Irreversible oxygen redox necessitates the triggering of another redox couple in the form of Mn 3+ /Mn 4+ in Li/Mn-rich materials. Formation of Mn 3+ is detrimental to the voltage stability because Mn 3+ can migrate to the Li site and form a spinel phase. However, Ni 3+ is more likely to be reduced than Mn 4+ in Li 1.2 Ni 0.4 Mn 0.4 O 2 because the empty Ni 2+ /Ni 3+ states lie below the Mn 3+ /Mn 4+ states (Figures 19D and19F). Hence, the Mn 3+ /Mn 4+ redox couple is not triggered for charge compensation (Figures 19E and19F). Controlling redox couple evolution such as suppressing the activation of Mn 3+ /Mn 4+ redox couple in charge compensation can also mitigate voltage fading in Mn-based sodium layered oxide materials. 32 157 On K + free Li 1.2 Mn 0.54 Co 0.13- Ni 0.13 O 2 , the Mn 4+ /Mn 3+ redox continuously shifts to lower potential, contributing to voltage decay, along with a continuous growth of the spinel phase. Meanwhile, the K + doping is argued to suppress spinel phase formation and the polarization of Mn 4+ /Mn 3+ redox coupling, resulting in comparatively lesser voltage decay in the material. Meanwhile, a gradient distribution of Li (Li-poor surface and Li-rich bulk) has been proposed as a strategy to stabilize Li-rich materials. 161 The Li-poor surface is claimed to be immune to irreversible oxygen redox while the bulk can still maintain the high capacity from oxygen redox. Interestingly, Pan and coworkers reported that utilizing a novel non-electroactive cathode component, e.g., sodium carboxymethyl cellulose (CMC) binder, can enhance the cycling performance of 174 These surface coatings are argued to suppress oxygen release by acting as a barrier between the cathode and the electrolyte, preventing oxygen loss, 175 electrolyte oxidation, 163 transition metal dissolution, 169,170 and structural transformations. 163,165,167 Often these benefits are accompanied by the sacrifice of initial discharge capacity because of the utilization of electrochemically inactive coating materials. 175 Electronic and ionic conductivities are also compromised because of the structural mismatch between the coating and the cathode particle. 176 Hence, phase compatibility becomes an important factor in maintaining the electronic and ionic conductivities of the coated materials. 176 175 The prevention of oxygen loss resulted in better cycling stability and less voltage fading on the coated material compared with the uncoated one.

Cationic and Anionic Doping/Substitution to Stabilize Oxygen Redox

The incorporation of foreign elements into the crystal lattice can effectively modify the crystal and electronic structure of cathode materials, which can be beneficial for stabilizing oxygen redox. Many different dopants have been studied, and the incorporation of these dopants to a specific lattice site can be carefully controlled. For example, Mg and Zn can be incorporated into the alkali layer and/or the transition metal site simultaneously. 177,178 Meanwhile, elements such as Ti, Nb, Zr, Cr, Al, and Fe are doped into the transition metal site of the layered structure. [179][180][181][182][183][184][185][186] An anionic dopant/substituent such as F can replace some O in the oxide framework of the crystal lattice. 108,187 F can enhance the structural stability by decreasing the Jahn-Teller distortion due to Mn 3+ . 108 F incorporation into the host lattice has been reported to enhance high-voltage stability of DRX materials. 4 Often, the stability is rationalized based on greater contribution of transition metal cations to redox reactions with respect to the oxygen anions. F substitution can reduce the effective anionic charge, thus lowering the positive charge on the transition metals. Lower positive charge on the transition metals results in larger capacity from the redox activity of the transition metals, which can limit the contribution of oxygen to redox reactions. 113 A strategy of F substitution and high-valence-state transition metal ion incorporation has enabled triggering of the facile redox of Mn 2+ /Mn 4+ in Mn-based DRX materials. 113 Thus, high capacity can be obtained without triggering extensive oxygen redox. Consequently, the degradation of cathode materials related to irreversible oxygen redox can be minimized. However, the effect of F Review doping/substitution can extend beyond to modifying the electronic structure of transition metal cations and oxygen anions. For example, unlike the oxide version of the DRX material, simultaneous redox of Mn and O can be observed in Li 2 Mn 2/ 3 Nb 1/3 O 2 F because of the overlap between Mn e g * states and unhybridized O 2p orbitals. The overlap can take place because of the Mn-F bonds and longer separation between Mn and the anions. 113 Meanwhile, cationic dopants such as Ti 4+ can stabilize the oxygen environment by strengthening the bonding between metal and oxygen. 10,188,189 Mg 2+ is reported to act as a pillar when sitting in the alkali layer, thus stabilizing the structure at high SOCs. 178 Mg 2+ can also improve the electrochemical stability by breaking the alkali-ion/vacancy ordering. 10 Zn 2+ doping can reduce Mn 3+ -induced Jahn-Teller distortion and can enhance phase stability. 177 Intentional incorporation of proton in the Li layer of the Li-rich oxides is reported to improve the initial Coulombic efficiency and discharge capacity. 190 Protons can be incorporated in the host structure through solution treatment of the pristine material at different pH. 190 A moderate coupling between the oxidized lattice oxygen and the inserted proton is observed, which is claimed to improve the stability of the oxygen redox activity.

Some dopants such as Ti, 10,188 Zr, 185 and W 191 are observed to segregate preferentially on the surface of the cathode particles. The segregation on the surface can form a protective layer between the cathode and the electrolyte, similar to the coating on a cathode particle. For example, Ti when segregated to the surface of the cathode particle can stabilize the interface because of the strong Ti-O bonding, and minimize the interfacial degradation phenomena. 10,188 Doping with multiple elements with unique distribution from the surface to the bulk can enhance the surface-to-bulk structural and chemical stability of conventional Li layered cathodes, 192 and the strategy can be adapted for stabilization of Li-rich layered oxides as well. 148,183 Shin et al. conducted a comprehensive study of many dopants to theoretically ascertain the tendency of the dopants to segregate on the surface of the particle. 97 The authors calculated the dopant segregation energy at the low index surface facets (namely two surface facets (001) and (010), and three subsurface facets (100), (110), and ( 111)) of Li 2 MnO 3 , which is summarized in Figure 20. The study effectively summarized the segregation behavior of the dopants and provided a reference point for identifying dopants to impart surface stability on the Li-rich cathodes. These calculations are based on thermodynamics. During the practical synthesis, kinetic parameters, such as temperature profile, will likely govern how dopants redistribute in cathode particles. 193

CONCLUSIONS AND PERSPECTIVE

Oxygen redox chemistry in alkali-ion batteries has opened up new opportunities for designing materials with high capacity and energy density. The large capacity obtained by the Li-rich materials has intrigued battery scientists in figuring out the chemistry behind the excess capacity. Early explanations included irreversible structural transformations and Li and O extraction. 11,14 However, from various pioneering works, 15,16,26,37,43,62,63,117,194 it is now widely accepted that the excess capacity can be obtained by the reversible redox chemistry of lattice oxygen. In conventional layered oxides, holes created by electron removal at a deeper delithiation/desodiation level can be trapped at the O 2p orbitals, causing oxygen redox. 25 This type of oxygen redox is often deemed irreversible. However, recent studies on the reversible nature of high-voltage lattice oxygen redox in conventional layered oxides warrants further investigation. 24,33 Li-O-Li configurations have been identified as critical structural units for triggering reversible oxygen redox in Li-rich materials. 15 However, the true nature of oxidized oxygen is still under debate. Some studies indicate localized holes in the O 2p orbitals, 43 whereas others claim significant structural transformations with the formation of peroxolike bonding along with ligand (oxidized oxygen) to transition metal ion charge transfer. 26,28 Recently, molecular oxygen entrapment in the lattice has also been presented as a possible model of lattice oxygen redox. 41 In this review, we have summarized these models of oxygen redox to develop a convenient understanding for readers. While many techniques such as XPS, [44][45][46] DEMS, [195][196][197] and electron paramagnetic resonance 63 are utilized to probe the chemical nature of oxygen redox, the recent development of RIXS has unequivocally proved the reversible nature of the lattice oxygen redox. 51,143 Through resolving the emission energy against the excitation energy of the fluorescence photons, the true signature of oxidized oxygen has been resolved. Such resolution has also enabled deciphering of the chemical nature of the oxidized lattice oxygen and has provided critical information on the bulk process of the lattice oxygen redox. 28,42 However, we note that the theoretical interpretation of the RIXS results of the transition metal oxide-based system remains elusive, although it may provide the most direct modeling related to the fundamental mechanism of oxygen redox reactions. We are aware of the current debate over the most legitimate analytical tools to probe the oxygen redox chemistry. We believe that a complete understanding of the highly controversial anionic redox requires a comprehensive diagnostic toolbox that can probe oxygen redox chemistry at multiple length and time scales.

We have also summarized the challenges with oxygen redox that prohibit the commercialization of the Li-rich cathode materials. Transition metal migration to the alkali metal-ion layer is widely observed in materials with oxygen redox. 117 Such migration has contributed to voltage fading on consecutive cycling. This migration is often accompanied by oxygen evolution and structural transformations. 30 The loss of oxygen induces the reduction of some transition metals, which can form new redox couples during subsequent cycling. The evolution of Mn-and Co-based redox The background rainbow color scheme of the element labeling indicates the radius of the elements. The segregation preference of the dopants is represented by a pie chart, and each segment of the pie chart belongs to a different facet. The segments are colored differently based on the segregation behavior of a dopant, with blue color indicating that the dopant would tend to segregate on the surface of that facet while the red color means that the dopant would prefer to stay in the bulk. Adapted from Shin et al. 97 couples has been found to contribute to the voltage fading of Li-and Mn-rich cathode materials. 30 Transition metal migration can alter the electronic structure of cathode material and change the redox sequence, causing redox asymmetry (cationicto-anionic oxidation on charge and cationic-to-anionic reduction on discharge). 28 Voltage hysteresis and fading have also been reported to originate from such asymmetric redox evolution. A recent study of oxygen redox in Na 2/3 Ni 1/3 Mn 2/3 O 2 raises further questions regarding the mechanistic understanding of voltage hysteresis and oxygen redox. 23 This material can utilize oxygen redox for charge compensation but with very little voltage hysteresis (<0.1 V), i.e., almost perfect electrochemical kinetics. No doubt a complicated relationship exists between oxygen redox, transition metal migration, transition metal reduction, and voltage hysteresis and fading. Further development of oxygen redox chemistry will involve the formulation of a unified understanding of oxygen redox to that of transition metal migration and voltage fading. Meanwhile, the critical challenges inhibiting the practical application of oxygen redox, such as voltage hysteresis, voltage fading, and slow oxygen redox kinetics, call for materials development that can address these issues. Future development of cathode materials with oxygen redox may focus on engineering the crystal structure and exploring the compositional space in order to tackle these issues.

Enhancing the reversibility of transition metal migration can be promoted by tuning the crystal structure (e.g., O2-type structure over O3-type structure) that can suppress voltage fading on electrochemical cycling. 158 Controlling the long-range ordering of transition metals also provides a means of reducing voltage hysteresis. 41 Meanwhile, close composition control of transition metals can suppress irreversible oxygen evolution. 159 Exploring the compositional space also provides an opportunity for improving oxygen redox kinetics. For example, incorporating Co in Na 0.6 [Mg 0.2 Mn 0.8Àx Co x ]O 2 can reduce the band-gap energy and enable facile electron transfer from Co 3d and O 2p states. This results in improved rate capability of the material. 198 Oxygen redox allows the development of high-energy cathode materials as an alternative to Li-ion cathodes. Sodium layered cathodes constitute such examples. In this review we have summarized that oxygen redox can be triggered in sodium layered cathodes, which can improve the capacity and energy density. 94,179 Crystal structures of sodium layered cathodes are more diverse than those of Li counterparts, and the compositional space is also vastly expanded. 83 Thus, along with utilizing oxygen redox, sodium layered cathodes can be synthesized with cheap and abundant transition metals, improving the cost-effectiveness of these materials. 199 Moreover, the broad range of crystal structures can provide an opportunity to minimize transition metal migration, thus suppressing voltage fading.