Fundamentals of metal oxide/oxyfluoride electrodes for Li-/Na-ion batteries

Since its inception in the 1970s¹ and conception in 1987,² rechargeable lithium batteries (so-called Li-ion Batteries) have been widely studied for energy storage applications, and now Li-ion battery technology is highly sophisticated. Nowadays, more than two million electric vehicles equipped with Li-ion batteries as a power source are being sold per year in the global market. Although the transportation system has relied on combustion engines for a long time, the Li-ion batteries are starting to provide an alternative way based on clean and sustainable energy resources without fossil fuels. Li-ion batteries possess 2–3 times greater volumetric energy density in comparison with other commercial batteries (e.g., metal hydride batteries).³ Li-ion batteries consist of two different Li insertion materials, that is, positive and negative electrode materials, with an aprotic electrolyte solution. The main role of Li insertion materials is to reversibly guest both lithium ions and electrons upon charge and discharge processes,⁴ and transition metal oxides containing Li ions, for example, LiMn₂O₄ and LiCo_1/3Ni_1/3Mn_1/3O₂, are generally utilized as positive electrode materials. The first commercialized Li-ion battery by SONY was composed of LiCoO₂⁵ as a positive electrode material and heat-treated petroleum coke as a negative electrode material with aprotic electrolyte. LiCoO₂ has a rock salt-related layered structure with a cubic close-packed (ccp) array of oxide ions. Perpendicular to [111] of the original cubic lattice, Li and Co ions are accommodated at octahedral sites in the ccp array of oxygen and are alternately stacked for each other, leading to the formation of α-NaFeO₂-type layered structure with a rhombohedral crystal lattice (Fig. 1). CoO₆ octahedra in LiCoO₂ share three edges, forming CoO₂ layers, which provide a two-dimensional conduction path for both Li ions and electrons. LiCoO₂ is also classified as O3-type layered structure according to the Delmas' notation.⁶ Here, “O” denotes a coordination environment for Li ions, and Li ions are found at octahedral sites. The number “3” corresponds to the number of CoO₂ layers in a unit cell. The theoretical capacity of LiCoO₂ can be accessed from the following reaction:

Herein, □ denotes a vacant octahedral site generated by the extraction of Li ions from the host structure. Co³⁺ cation in LiCoO₂ is oxidized to Co⁴⁺ (electrons flow to the external circuit), and simultaneously Li ions are removed from the host structure to retain charge neutrality. A Fermi level of LiCoO₂ consists of mainly Co ions in an octahedral environment with t_2g⁶ low spin configuration and a partial contribution of O 2p orbitals. Because oxidation of the Co ion to Co⁴⁺ enhances a covalent nature between Co and O, oxygen 2p orbitals also partially lose electrons as shown in schemes of Fig. 2(a). The theoretical capacity of LiCoO₂ derived from Eq. (1) is 274 mA h g⁻¹ (note that experimentally the reversible capacity is generally restricted to ∼170 mA h g⁻¹ to prevent capacity deterioration). After its inception in 1980,⁵ LiCoO₂ is still widely used in commercial applications, especially for portable electronic devices (its battery size is limited to <100 W h) owing to its high volumetric energy density.

To apply Li-ion batteries to the electric vehicle, much larger batteries with higher energy (20–100 kW h) are required, and therefore, expensive LiCoO₂ cannot be utilized as the electrode material. Instead of LiCoO₂, spinel-type LiMn₂O₄ and its derivatives were originally used as positive electrode materials. In LiMn₂O₄, MnO₆ octahedra share edges, leading to the formation of a three-dimensional Li ion migration path in the structure, while Li ions are located at tetrahedral sites (see Fig. 1). Upon Li extraction, almost all Li ions can be removed without destroying its framework structure. LiMn₂O₄ is operable as a 4-V class positive electrode material, similar to LiCoO₂. Also, Li insertion into LiMn₂O₄ is possible and an additional capacity is obtained, but it is subject to dramatic capacity fading upon cycling due to the Jahn–Teller distortion, which induces large and non-isotropic volume changes.⁷ Therefore, this material has a practically low reversible capacity, 120 mA h g⁻¹, which restricts its energy density to 450–500 W h kg⁻¹ based on the electrochemical potential of metallic Li (and without the consideration of the mass of negative electrode material as the full cell). Seeking Li-ion batteries with higher energy density, Ni-based layered materials are now being used for electric vehicle applications. LiNiO₂ is isostructural with LiCoO₂, and solid solutions are formed in the entire compositional range of LiCo_xNi_1–xO₂ (0 ≤ x ≤ 1).⁸ Stoichiometric LiNiO₂ has a reversible capacity of 200 mA h g⁻¹ and presents a slightly lower operating voltage in comparison with LiCoO₂ and LiMn₂O₄. The experimentally available energy density reaches 700–750 W h kg⁻¹ (based on the electrochemical potential of metallic Li) for Ni-based layered materials. Stoichiometric LiNiO₂ becomes thermally unstable after its delithiation and shows detrimental phase transitions on delithiation.^9–11 For these reasons, Ni ions are generally partially substituted by other transition metal ions. Many voltage plateaus are observed for the stoichiometric LiNiO₂ [Fig. 3(a)], associated with complicated phase transitions. These plateaus are clearly visible in the differential capacity plot in Fig. 3(a). The Co/Al-substituted system, for instance, LiNi_0.8Co_0.15Al_0.05O₂,¹² is already applied for commercial batteries. Another substituted system for positive electrode materials, Ni/Mn, has been widely studied and is practically utilized for electric vehicle (EV) applications.^13–16 Voltage plateaus, which correspond to the phase transition reactions, are less marked for LiNi_0.8Co_0.15Al_0.05O₂ as shown in charge/discharge curves and its differential capacity plot [Fig. 3(a)].

These phase transitions originate from lithium and charge ordering in the layered structure associated with the formation of vacant octahedral sites and two cations with different charges (Ni³⁺ and Ni⁴⁺) on delithiation.¹⁷ Nearly stoichiometric LiNiO₂ is synthesized from a mixture of Ni(OH)₂ and LiOH·H₂O at 650 °C under O₂ flow.¹⁸ Evolution of crystal structures of Li_yNiO₂ examined by in situ x-ray diffraction study is shown in Fig. 4(a). As-prepared LiNiO₂ is isostructural with LiCoO₂ possessing rhombohedral symmetry (denoted as O3 I phase), and this phase changes into a slightly in-plane distorted lattice at y = 0.8 in Li_yNiO₂ with monoclinic symmetry (denoted as O'3 phase, and herein, the prime symbol indicates the presence of in-plane distortion for NiO₂ layers). In-plane distortion results in peak splitting and/or broadening for 10l and 11l lines of the x-ray diffraction pattern. This monoclinic phase is retained and again changes into a non-distorted lattice at y = 0.4 in Li_yNiO₂. A long voltage plateau is observed at 4.2 V in Fig. 3(a), and two phases (O3 II and O3 III phases) with different interlayer distances (a large gap of 7.4%)¹⁸ coexist at this region as visualized in Fig. 4(a). Also, Ni ion migration to neighboring tetrahedral sites on charge has been evidenced by detailed structural analysis.¹⁹ Note that such phase transition behavior is altered by partial Co/Al substitution for Ni. A two-phase reaction and monoclinic phase transition are not observed for LiNi_0.8Co_0.15Al_0.05O₂, and a single-phase reaction is observed for the delithiation process of Li_yNi_0.8Co_0.15Al_0.05O₂, which is consistent with the fact that less voltage plateaus are observed in Fig. 3(a).²⁰ Such metal substitution disturbs the in-plane Li (and vacancy) and charge ordering (Ni³⁺ and Ni⁴⁺ ordering), and thus, the clear difference in the phase transition process, including Ni ion migration, is observed for both Ni-based layered oxides.

Because the ionic radius of Na ion (1.16 Å) is larger than that of Li ion (0.90 Å),²¹ the electrostatic interaction of Na ions in layered oxides is clearly pronounced when compared with Li ions. This fact results in the noticeable phase transitions for Na insertion materials, which transcribes into intensified voltage plateau, as observed for NaMnO₂ on the desodiation process [Fig. 3(b)], which is intensified compared with LiNiO₂. The crystal structure of NaMnO₂ is classified as O'3-type layered phase, and in-plane distortion for MnO₂ layers originates from the presence of Jahn–Teller active Mn³⁺ ions with the high spin configuration (t_2g³ e_g¹). Note that Ni³⁺ ions with the low spin configuration (t_2g⁶ e_g¹) are also Jahn–Teller active, and indeed, NaNiO₂ crystallizes into O'3-type layered phase.²² In contrast, in-plane distortion is not observed in LiNiO₂ as mentioned above. However, the distortion of a local structure of NiO₆ in LiNiO₂ is evidenced by extended x-ray absorption fine structure study.²³ In situ x-ray diffraction patterns of Ti-substituted NaMnO₂, Na_0.8Mn_0.8Ti_0.2O₂, which is a mixture of layered oxide (Na_0.88Mn_0.88Ti_0.12O₂) and tunnel-type Na₄Mn₄Ti₅O₁₈, are also displayed in Fig. 4(b). The presence of in-plane distortion is clearly found for Na_0.88Mn_0.88Ti_0.12O₂. This results in the splitting of diffraction lines observed for rhombohedral symmetry. For instance, single 104 line for the O3 phase splits into 20–2 and 111 lines for the monoclinic lattice. Peak splitting is more pronounced for O'3 Na_0.88Mn_0.88Ti_0.12O₂ compared with O'3 Li_yNiO₂. On charge, after sodium extraction and the dilution of Mn³⁺, the O3 phase without in-plane distortion is formed for Na_0.3Mn_0.88Ti_0.12O₂. Further extraction of sodium ions leads to the formation of O1 phase with hexagonal symmetry. Nearly 10% shrinkage is observed for interlayer distances on desodiation. A large gap of interlayer distances before/after Na extraction is brought from the fact that Na ions with the larger ionic radius are extracted from Mn_0.88Ti_0.12O₂ layers. However, similar to LiNiO₂, nearly all Na ions are reversibly extracted from NaMnO₂ without the destruction of the framework structure.

These complicated phase transitions for Na_yMnO₂ are featured from in-/through-plane electronic and magnetic interaction of Mn ions.^24,25 The structural model of Na_5/8MnO₂ is also shown in Fig. 4(c).²⁶ In addition to in-plane Na ions and vacancy ordering, Mn⁴⁺ ions in the MnO₂ layer form a cluster with Na ions. In-/through-plane antiferromagnetic interaction for Na_1/3MnO₂ is also evidenced by neutron scattering study [Fig. 4(d)].²⁷ These magnetic long-range interactions are essential for complicated phase transitions for these layered materials. A practical problem of NaMnO₂ for battery applications is found in the relatively higher solubility of Mn ions into non-aqueous electrolytes including polar solvents.²⁸ Ti substitution in NaMnO₂ effectively improves electrode reversibility because Ti ions are less soluble ions into such electrolytes. Indeed, Na_0.88Mn_0.88Ti_0.12O₂ shows much improved capacity retention as positive electrode materials of Na-ion batteries.²⁸ 

Although the Ni-enriched layered materials are presently used in EV batteries, their theoretical energy density is barred at 750 W h kg⁻¹ when coupled with metallic lithium. Therefore, the development of higher capacity positive electrode materials is imperative to guarantee the ever-needed progress of battery energy density of commercial batteries. For positive electrode materials, the theoretical capacity is based on the amount of Li ions present within the host structure, and therefore to design superior capacity positive electrode materials, cationic transition metal ions needed to be substituted with Li ions. This results in the formation of so-called Li-excess or Li-rich phases, Li_1+xM_1–xO₂ (Fig. 1). In the last ten years, Li-enriched materials, Li₂MO₃-type layered materials (M = Mn⁴⁺, Ru⁴⁺, and other tetravalent transition metal ions), which are also categorized based upon the cation-ordered rock salt-type structure,²⁹ were widely investigated as higher-capacity battery electrode materials.^30–38 Among the Li₂MO₃-type oxides, a Mn system, Li₂MnO₃, and its derivatives were extensively examined as electrode materials. Li₂MnO₃ has a similar structure compared with LiCoO₂ but presents extra Li ions within the transition metal layers. The chemical formula of these materials is generally reformulated from Li₂MnO₃ to Li_4/3Mn_2/3O₂ in order to normalize their oxygen content to 2. The theoretical capacity of Li₂MnO₃ can be accessed from the following reaction:

The reversible extraction of two Li ions and electrons from the host structure results in a larger theoretical capacity (459 mA h g⁻¹) compared to conventional stoichiometric layered oxides. However, until recently, Li₂MnO₃ was not considered as an electrochemically active material due to the complexity of oxidizing Mn ions from tetravalent state to higher oxidation states. Fermi level of Li₂MnO₃ consists of mainly oxygen 2p orbitals and a partial contribution of Mn 3d orbitals [Fig. 2(b)], and therefore, cationic Mn ions are electrochemically inactive and cannot donate electrons on electrochemical oxidation. Nevertheless, the fact is that Li₂MnO₃ is active as the electrode material associated with the electron donation and charge compensation by oxide ions as anionic species on oxidation.^34,36,39 Since 1990, the charge compensation of sulfide-based compounds by non-cationic species had been known,⁴⁰ because sulfide ions in comparison with oxide ions are fairly soft and easily polarizable. Ligand hole stabilization has also been reported in oxides with a perovskite-type structure, for example, Fe⁴⁺ in SrFeO₃.⁴¹ Fe⁴⁺ is strongly hybridized with oxygen 2p orbitals, and therefore, the transfer of charge from oxide ion to transition metal ions results in the formation and stabilization of ligand holes. When ligand holes are produced by electrochemical oxidation in Li₂MnO₃ as shown in Fig. 2(b), this process is classified as “anionic” redox. For Li₂MnO₃, four Li ions and two Mn ions coordinate oxide ions. Because the Li-O bonding character is basically ionic, electronic structures of oxygen in Li₂MnO₃ are different from stoichiometric layered materials, like LiNiO₂ with a relatively higher covalent nature for Ni–O bonds. In Li₂MnO₃, the Li ions transfer additional electrons to oxide ions, leading oxide ions net oxidation number to reach its conventional oxidation number of minus two. The demonstration of this hypothesis has been conducted via density functional theory (DFT) study for the Li-excess oxides.⁴² The results show that a linear Li-O-Li configuration, which is basically unhybridized with transition-metal ions, is contained within Li-excess oxide. Comparatively, with hybridized O 2p states, this unhybridized state is of lower energy, making it easier for oxidation of oxide ions in Li-excess oxides. Recently, such unhybridized orphaned O 2p state for Li₂MnO₃ has been directly evidenced by resonant inelastic x-ray scattering (RIXS) study (Fig. 5).⁴³ 

Following the original publication of Li₂MnO₃,⁴⁴ the reaction mechanisms and electrochemical properties of Li₂MnO₃ have been thoroughly studied and reported in the literature.^{34,36,45–48} The experimental results denote that its electrochemical performances are directly related to its specific surface area, and could reach a discharge capacity of >250 mA h g⁻¹ for nanosized Li₂MnO₃ as shown in Fig. 6. Furthermore, the release of oxygen upon charge, generating Li_2–xMnO_3–y, was also evidenced. Electrode reversibility is rapidly degraded during continuous cycles.³⁶ This fact indicates that holes formed by oxidation of the unhybridized O 2p state are energetically unstable. When the concentrations of holes in oxygen 2p orbitals and vacant octahedral sites formed by Li extraction are increased, chemical bonds between Mn and O are energetically destabilized, leading to the formation of oxygen dimers, namely, peroxide/superoxide ions and/or molecular oxygen. Additionally, a DFT study proposed that following the oxide ions oxidation the dimerization would always cause Mn migration.⁴⁹ Oxygen dimers are chemically active species and would easily react with aprotic solvents. Moreover, oxygen dimers are electrochemically oxidized to O₂ molecules by electrochemical reactions. Recently, the possibility of Mn⁷⁺ ion formation as an intermediate species has been proposed. Chemically unstable Mn⁷⁺ ions formed by electrochemical oxidation migrate into the adjacent tetrahedral site and further oxidize neighboring oxide ions, leading to oxygen dimerization coupled with reduction of Mn ions to the original tetravalent state.⁵⁰ 

As mentioned above, upon oxidation “pure” Li₂MnO₃ unavoidably and irreversibly lose oxygen; nevertheless, this phenomenon can be partially inhibited via the substitution of some Li⁺ and Mn⁴⁺ by other transition metal ions. For clarity, the Li₂MnO₃ can be reformulated as Li(Li_1/3Mn_2/3)O₂ as if its crystal structure was basically a conventional layered material with a ccp lattice of oxide ions except for the presence of Li ions in the transition metal layers. Therefore, a succession of solid solution phases can be produced from the binary system of x LiMeO₂–(1–x) Li(Li_1/3Mn_2/3)O₂ (Me = transition metal ions). The solid-solution samples prepared between LiNi_1/2Mn_1/2O₂^51,52 and Li(Li_1/3Mn_2/3)O₂ reported by Dahn and co-workers display large capacities as high as 250 mA h g⁻¹ when used as positive electrode material in Li-ion Batteries.^31,32,53 Luo et al. employed soft x-ray absorption spectroscopy (XAS) to examine the anion redox reactions of Li_1.2Co_0.13Ni_0.13Mn_0.54O₂. This study on the solid-solution sample between Li(Li_1/3Mn_2/3)O₂⁵⁴ and LiCo_1/3Ni_1/3Mn_1/3O₂^13,51 has suggested the possibility of hole creation and stabilization on oxide ions.⁵⁴ However, following charge to the voltage plateau, in-plane superlattice order is lost because of partial oxygen loss coupled with cation migration.^55–58 The unhybridized orphaned O 2p state is also observed in Li_1.2Co_0.13Ni_0.13Mn_0.54O₂.⁴³ After oxidation, unhybridized oxygen is lost and the presence of non-localized bonds with a weak π-type interaction between the Mn t_2g and the O 2p orbitals have been evidenced.⁴³ Recently, it has been also proposed that O₂ molecules are formed even in the bulk of particles. Voids formed by in-plane transition metal migration accommodate O₂ molecules, and O₂ are reduced back to oxide ions on discharge.⁵⁹ Because the tetravalent Mn ions partially are reduced to the trivalent state, similar to Li₂MnO₃, when discharge is conducted following a charge above the voltage plateau,^56,58,60 this fact indicates that the anionic redox in the solid solution samples results in partial oxygen loss and causes the irreversibility of this process. Therefore, the electrochemical performance of the solid solution materials is highly dependent on their particle sizes, similar to Li₂MnO₃. In a similar manner, the particle size dramatically influences the oxygen loss degree vs hole stabilization (and/or oxygen dimerization).⁶¹ The loss of oxygen on charging initiates the formation of mobile trivalent Mn ions on the following discharge, resulting in a dramatic phase transition on continuous cycles. The characterization of its structure reveals that the increasing structural symmetry to a cubic phase and the progressive enrichment of trivalent Mn ions lead to the fading of the capacity in the case of nanosized sample.⁶¹ Oxygen loss is observed for continuous cycles, and trivalent Co ions are also reduced to the divalent state, leading to the gradual reduction of operating voltage on cycles.⁶² Although the decrease in operating voltage is unavoidable, clear evidence of anionic redox reaction is found even after 500 cycles as shown in Fig. 7.

Because Mn migration on charge triggers oxygen dimerization/oxygen loss,⁶³ the suppression of Mn migration is an effective strategy to mitigate continuous voltage decay. Upon extraction of Li ions from O3-type layered materials, due to energy stabilization, the migration of transition metal ions is facilitated initially within the Li layers through adjacent face-shared tetrahedral sites and then into octahedral sites. Indeed, as in the Li layers, three transition metal ions share edges with octahedral sites, only a weak cation–cation repulsive interaction can be anticipated [Fig. 8(a)]. On the other hand, an efficient way to inhibit this phase transition would be to apply different layered structures, which present original oxygen stacking, and a promising candidate could be O2-type layered oxides, which are composed of ABCBAB-type oxygen stacking. Indeed, within this layered structure, the octahedral site within the Li layer is directly sharing a face with a transition metal. In this structural configuration, due to the superior proximity between the octahedral site and a transition metal ion, as compared with the O3-type layered materials, it would create a stronger cation–cation repulsive interaction, and thus, the suppression of phase transition is achieved [Fig. 8(a)].⁶⁴ Historically, Mn migration is also known as a critical issue for stoichiometric LiMnO₂, which easily changes within a spinel phase. Such phase transition is, indeed, successfully overcome for LiMnO₂ having an O2-type layered structure.⁶⁵ This structure is a metastable phase that can therefore be synthesized through ion-exchange using a P2-type (ABBA stacking) Na-based layered oxide (also see the section of Na insertion materials). In this notation, the “P” translates the location of the Na ions at the prismatic sites, which induces a larger interlayer distance. Through completion of the Na/Li ion-exchange, the original structure is subjected to the gliding of its metal layers and forms the targeted O2-type layered oxide, which presents a narrower interlayer distance.

Li-excess O2-type layered Li_xMn_3/4O₂ has been prepared from P2-type Na_5/6Li_1/4Mn_3/4O₂ and tested as electrode materials.⁶⁶ Galvanostatic charge/discharge curves of O2 Li_xMn_3/4O₂ are presented in Fig. 9(a). Electrochemical data of an O3-type layered phase with a similar chemical composition are also compared in Fig. 9(b). Voltage decay is effectively suppressed for the O2 phase, but not for the O3 phase. Similar phenomena are also observed for O2 Li_xMn_0.71Ni_0.17O₂.⁶⁷ Recently, the transition metal ion stabilization at the intermediate tetrahedral sites has been visualized for the O2-type Li_1.2Ni_0.2Mn_0.6O₂, by employing high-resolution scanning transmission electron microscopy (STEM).⁶⁴ Furthermore, a theoretical analysis of the repulsive interaction of the edge-shared and face-shared sites has been conducted. The results of this calculation indicate superior site energy of 0.6–1.0 eV for the face-shared of O2-type oxide as compared with the edge-shared site of the O3-type oxide [Fig. 8(b)]. Due to this thermodynamic variation, the O2 phase can effectively suppress the migration of transition metal ions within the face-shared site, and therefore unlock the reversible migration of transition metal between initial octahedral and intermediate tetrahedral sites.

Another approach to resolve this limitation is to employ Ru a 4d transition metal ion instead of 3d transition metal ions because of its superior covalent nature toward oxides ions.^38,68–70 Li₂RuO₃ is essentially isostructural with Li₂MnO₃, and Ru ions with a high oxidation state have excellent stability. Moreover, due to Li₂RuO₃ with superior electronic conductivity, the anionic redox can be easily activated,³⁰ resulting in large reversible anionic redox. This material reversibly delivers more than 90% of its theoretical capacity, reaching a capacity of 300 mA h g⁻¹. To reach this level of capacity, Li₂RuO₃ has to reversibly exchange nearly two moles of Li ions by the cationic/anionic redox. Even at 50 °C this material was able to maintain its performances without capacity fading.³⁸ Therefore, Li₂RuO₃ has higher gravimetric/volumetric energy density than those of Ni-based layered materials, but the material cost of Ru would hinder its use for practical battery applications.

To continue increasing the concentration of Li ions from Li₂MnO₃, transition metal ions with higher oxidation states can be introduced,³⁸ for example, Li₃NbO₄ with pentavalent Nb ions,⁷¹ Li₄MoO₅ with hexavalent Mo ions,⁷² and Li₅ReO₆ with heptavalent Re ions.⁷³ These Li-excess oxides consisting of oxide ions arranged in ccp array are also utilized as host structures of potential electrode materials with high reversible capacity. The effective usage of the anionic redox could unlock a theoretical capacity of 526 mA h g⁻¹ for Li₄MoO₅. Unfortunately, these host materials are mostly insulators and/or semiconductors by nature because they do not possess conductive d electrons, leading to the kinetic inactivation of anionic redox reactions. Li insertion materials as electrode materials necessitate a combination of ionic and electronic conductivity. For these host materials to work, it is necessary to partially substitute transition metals with d electrons.

For example, pure Li₃NbO₄ is electrochemically inactive, but partially substituting 3d transition metal ions for Nb and Li ions provides conductive electrons to the structure.^74,75 These 3d transition metal ions, by accepting electrons from oxide ions, result in the activation of the anionic and cationic redox as electrode materials. When subject to metal substitution, the cation ordering of pure Li₃NbO₄ disappears and is replaced by a rock salt-type structure. Ti⁴⁺, Nb⁵⁺, and Mo⁶⁺ ions with d0 configuration are beneficial to stabilize the cation-disordered rock salt structure because these ions accommodate large octahedral distortions.⁷⁶ Ta⁵⁺ ions are also utilized for this purpose.⁷⁷ Historically, due to the absence of migration path for the Li ions within the cation-disordered rock salt structure, these host structures were often considered inactive.^78–80 However, for Li-excess cation-disordered rock salt (Li_1+xMe_1–xO₂), the facile migration of Li ions becomes feasible through the self-generation of a percolative network within the host structure.^81–83 Li ions at octahedral sites in rock salt oxides migrate through adjacent and face-shared tetrahedral sites to reach the neighboring octahedral sites, that is, o-t-o migration [Fig. 10(a)]. Migration barriers for Li ions depend on the local environment for intermediate tetrahedral sites. When tetrahedral sites share faces with transition metal ions with higher oxidation states, large electrostatic repulsive interaction is anticipated between transition metal ions and Li ions [1-TM channel and 2-TM channel in Fig. 10(a)]. In contrast, if tetrahedral sites share only Li ions [0-TM channel shown in Fig. 10(a)], repulsive interaction is effectively reduced, leading to facile migration for Li ions. Such 0-TM channel is formed for Li-excess materials, resulting in the activation of percolative migration for Li ions in the rock salt oxides [Fig. 10(b)].

Metal substitution by Mn³⁺ and Fe³⁺ ions in Li₃NbO₄ results in the formation of cation-disordered rock salt oxides, Li_1.3Nb_0.3Mn_0.4O₂ and Li_1.3Nb_0.3Fe_0.4O₂; galvanostatic charge/discharge curves are compared in Fig. 11(a).⁷⁴ On the initial cycle, both samples display a voltage plateau at 4.1–4.3 V and reach a charge capacity of ∼350 mA h g⁻¹. However, a drastic change is observed upon discharging process. For Li_1.3–xNb_0.3Fe_0.4O₂, the first discharge is subject to a larger polarization, and the following charge curve completely differs from the initial one, which results in losing the voltage plateau initially observed. On the other hand, for Li_1.3–xNb_0.3Mn_0.4O₂ almost no differences are observed from the first and to the second cycle where a sloping region can be visualized from 3 to 4 V and followed by a voltage plateau at 4.3 V. For the Mn substituted sample, a large reversible capacity of 300 mA h g⁻¹ is visualized at 50 °C while displaying quite high discharging voltage. The displayed reversible capacities are superior to the theoretical capacity calculated based on the Mn³⁺/Mn⁴⁺ cationic redox, which therefore clearly demonstrates the participation of reversible anionic redox reactions for Li_1.3–xNb_0.3Mn_0.4O₂. Similar to Li₃NbO₄, Li₂TiO₃ has been employed as an anionic redox-based host structure, and rock salt phases are synthesized with Li₂TiO₃ and Mn³⁺/Fe³⁺,^84,85 and similar electrochemical properties are observed with Nb-based oxides. Studies by soft x-ray absorption spectroscopy (XAS) at O K-edge in Li_1.2–xTi_0.4Mn_0.4O₂ and Li_1.3–xNb_0.3Mn_0.4O₂ reveal that the large reversible capacity originates from reversible redox reaction of anionic species.^84,86 Similar variations in O K-edge XAS spectra are also noted for Li₂MnO₃-based electrode materials;^60,87 however, larger differences have been observed for both samples. The formation of π-like bonds between Mn⁴⁺ with t_2g3 configuration and O 2p orbitals and its rehybridization⁸⁸ are also expected for the rock salt phase.

In contrast to the Mn system, the Fe substituted samples, Li_yNb_0.3Fe_0.4O₂ and Li_yTi_0.4Fe_0.4O₂, show completely original results. Those conclusions indicate that on the initial charge process irreversible phase transitions occur. The appearance of news peaks of O K-edge and Fe L-edge XAS spectra on the Li_1.3–yNb_0.3Fe_0.4O₂ is observed after charge to 120 mA h g⁻¹,⁸⁴ corresponding to the extraction of approximately 0.3 mol of Li ions. This fact indicates that oxygen redox with partial oxidation of Fe to tetravalent state at the early stage of the initial charge. However, continuing charging up to 4.8 V leads to the disappearance of those peaks, which suggests the energetic instability of these species formed at the early stage of charge, leading to oxygen loss as an irreversible process. Such irreversible loss of oxygen leads to the modification of the particle morphologies, reducing its particle size to nanometer, and this fact is clearly visualized in TEM images shown in Fig. 11(b). These findings are consistent with a recent publication that demonstrates that oxygen molecules are formed even inside particles,⁵⁹ and oxygen loss is the dominant process for the Fe system. Consecutive to the loss of oxygen upon charging, Fe³⁺ is subjected to electrochemical reduction into Fe²⁺ on the initial discharge,⁸⁴ whereas for the Mn system the reversible change in cationic and anionic redox is observed.⁸⁶ Furthermore, the fading of anionic redox activity of Fe systems induces enormous polarization as shown in Fig. 11. Comparably, Li_1.33Sb_0.33Fe_0.33O₂ system is also subject to oxygen loss related to instabilities of anionic redox.⁸⁹ 

When oxides are constituted of Fe³⁺, strong covalency exists between O and Fe, which leads to a Fermi level composed of both Fe e_g and O 2p orbitals (Fig. 12).⁸⁴ After charge, both Fe and O ions are, therefore, oxidized. However, Fe⁴⁺ is chemically unstable, and therefore, Fe ions are reduced by electron donation from oxygen. Thus, oxygen progressively loses electrons, causing the dimerization and the release of oxygen. Contrastingly, when oxides are constituted of Mn⁴⁺, a high ionic character exists between O and Mn, and Mn⁴⁺ presents superior chemical stability when compared with Fe⁴⁺. In addition, in comparison with the late transition metal ions, like Ni and Co, Ti⁴⁺ and Nb⁵⁺ ions are highly ionized ions. The chemical bonds nature between Ti⁴⁺/Nb⁵⁺ without valence electrons and oxide ions is essentially non-bonding,⁸⁸ and O 2p orbitals and metal d orbitals mixing is negligible. Hence, a more ionic character is obtained for oxide ions, which is a similar situation to the Li enrichment in the host structure as discussed above related to Li₂MnO₃. This point is profitable to the activity of anionic redox, which induces lowering electrochemical potential. The comparison of Li_1.2Co_0.13Ni_0.13Mn_0.54O₂ and Li_1.2Ti_0.4Mn_0.4O₂ shows analogous voltage plateaus for initial charge/discharge profiles in Li cells. However, the voltage related to anionic redox is 0.15 V lower for Li_1.2–yTi_0.4Mn_0.4O₂ than that of Li_1.2–xCo_0.13Ni_0.13Mn_0.54O₂.³⁹ Indeed, in comparison with Li-excess oxides composed of late transition metal ions (d-block elements from the groups 8 to 11), the presence of Nb⁵⁺ and Ti⁴⁺ facilitates and activates the anionic redox at a lower voltage. Recently, anionic redox reaction of Li_1–yTi_0.4Mn_0.4O₂ has been experimentally visualized by high-energy x-ray Compton measurements.⁹⁰ 

A practical problem for Li_1.2Ti_0.4Mn_0.4O₂ is found in electrode kinetics, which originates from low electronic conductivity coupled with slow Li migration after the oxidation of Mn ions to tetravalent state. The Fermi level is mostly composed of O 2p orbitals with a partial contribution of Mn t_2g orbitals (Fig. 12). Therefore, the large reversible capacity is obtained only at elevated temperatures (50–60 °C) because higher temperature simply results in higher electron conductivity and Li mobility, and probably better hole conductivity. However, applying this electrode material at elevated temperature dramatically induces damage especially for anionic redox (e.g., loss of oxygen and side reaction with electrolyte, etc.), which leads to the continuous capacity fading. To solve these problems, nanosized samples are synthesized by mechanical milling. Also, electrode reversibility is deteriorated due to excess enrichment of holes on oxidation,⁹¹ and therefore, the Mn content is slightly increased within Li_1.2Ti_0.4Mn_0.4O₂ to reduce the concentration of holes. A, thus, designed and synthesized nanosized sample, Li_7/6Ti_1/3Mn_1/2O₂, has been tested as positive electrode materials for Li cells and delivered an excellent reversible capacity of nearly 300 mA h g⁻¹ at room temperature (Fig. 13).⁹² Instead of Ti⁴⁺, the sample with Mn⁴⁺ has been also prepared, and a metastable and nanosized oxide, Li_7/6Mn⁴⁺_1/3Mn³⁺_1/2O₂, has been successfully synthesized by mechanical milling. Details about the synthesis of nanosized materials by mechanical milling are discussed in the literature.⁹³ Upon initial charge, Li_7/6Mn⁴⁺_1/3Mn³⁺_1/2O₂ delivers an enormous capacity of ∼350 mA h g⁻¹, but the deterioration of capacity retention is observed. The comparison of experimental data for Li_7/6Ti_1/3Mn_1/2O₂ and Li_7/6Mn⁴⁺_1/3Mn³⁺_1/2O₂ indicates that the high ionic character of Ti ions participates to stabilize anionic redox reaction as discussed above, and anionic redox for Li-excess rock salt oxides cannot be stabilized only by Mn ions with weak π-type interaction to oxygen.

The utilization of electrochemically inactive sites for oxygen is another strategy to suppress the irreversible oxygen dimerization and loss. The integration of Li₃PO₄ in rock salt oxides improves the reversibility of anionic redox because oxygen coordinated to phosphorus ions is not oxidized and thus suppresses oxygen dimerization.⁹² The integration of Li₃PO₄ to LiMnO₂ has been conducted by mechanical milling. Through this process, the initial high crystallinity LiMnO₂ diffraction pattern is replaced by phases with subsequently lower crystallinity and displaying the cation-disordered structure. To clarify detailed nanostructures, thus obtained sample with a chemical composition of Li_7/6P_1/6Mn_2/3O₂ was analyzed by combining scanning transmission electron microscopy (STEM) and electron energy loss spectroscopy (EELS) [Fig. 14(a)]. This combined measurement technique, by extrapolating data from the intensity of the EELS spectra, gives the atomic distribution of P and Mn ions with a spatial resolution of 1.5 nm. The results denote the uniform integration and distribution of P and Mn ions at the nanoscale and the formation of a nanometer-sized crystal below 5 nm [Fig. 14(b)], which is consistent with XRD observation.⁹² Additionally, the profile of P L-edge EELS spectra in Fig. 14(c) confirms the P ions incorporation within the crystalline LiMnO₂ phase. A clear difference is observed for P L-edge EELS spectra measured between the original Li₃PO₄ and Li₃PO₄ with LiMnO₂ integration. Comparable trends in P ions are noted by other characterization techniques, and these results suggest that P ions are successfully integrated and dissolved within LiMnO₂.⁹² 

Electrochemical reversibility of Li_7/6P_1/6Mn_2/3O₂ in Li cells was assessed as positive electrode materials. The sample delivers a capacity superior to 300 mA h g⁻¹ [Figs. 15(a) and 15(b)], which is largely over the theoretical capacity where only Mn³⁺/Mn⁴⁺ redox participates (218 mA h g⁻¹), therefore suggesting the contribution of anionic redox. As compared with its rock salt LiMnO₂ counterpart without Li₃PO₄, Li_7/6P_1/6Mn_2/3O₂ displays improved rate capability performances maintaining a capacity of 190 mA h g⁻¹ at 640 mA g⁻¹ [Fig. 15(a)].⁹⁴ By raising the cutoff voltage to 5.2 V a capacity of 320 mA h g⁻¹ is reached [Fig. 15(b)]. Nevertheless, this boost of capacity is obtained through the detriment of the cyclability, leading to 50% capacity loss after only 30 cycles. On the other hand, by selecting a mild cutoff voltage of 4.5 V allows good capacity retention, maintaining 80% of its initial capacity after 30 cycles.

The direct evidence of anionic redox in Li_7/6–yP_1/6Mn_2/3O₂ has been observed by operando XAS study, and the detailed setup for measurement is described in Fig. 15(c). As the charge capacity and voltage increased, visible variations in the O K-edge XAS spectra were observed, as well as an increase in the peak intensity near 532 eV. A similar tendency is observed for Li-rich Mn-based oxides.^43,74,84 The origin of this has been hypothesized to be related to the formation and stabilization of holes by a π-type interaction with Mn t_2g orbital. Furthermore, theoretical calculation shows that only oxygen bonded with Mn ions is oxidized; on the other hand, oxygen bonded with phosphorus is not responsible for charge compensation. Additionally, the fully charged sample did not show any oxygen dimerization, and this even after structural relaxation.⁹² The superior stability of anionic redox of Li_7/6–yP_1/6Mn_2/3O₂ is understood to originate from the P ions presence, which tends to suppress the excessive oxidation of oxide ions; therefore, limiting the probability of oxygen dimerization bonded to P ions.⁹² Similar to Li₃PO₄, anionic redox has been also applied to the non-rock salt system, for example, Li₄Mn₂O₅,^95,96 Li₂O (Refs. 97–101), etc. Further comparative studies on these electrode materials for Li-excess layered, rock salt, and non-rock salt oxides are necessary to design stable anionic redox high-capacity and long cycle life electrode materials.

Another group of Li-excess rock salt oxides is based on V and Mo chemistry. As mentioned above, inevitable competition between anionic and cationic redox generally causes the loss of oxygen for late 3D transition metal ions. Due to the lower energy of d electrons for V and Mo ions as compared with oxygen p-electrons, the cationic/anionic redox competition is effectively suppressed, and therefore, solely cationic redox is utilized (Fig. 16).⁷⁵ Electrochemical properties of Li_1.25Nb_0.25V_0.5O₂, which is found in Li₃NbO₄–LiVO₂ binary system, are displayed in Fig. 16. The structural analysis of Li_1.25Nb_0.25V_0.5O₂ denotes the formation of a cation-disordered rock salt structure where the cations can be found at the same octahedral 4a sites after electrochemical cycles.¹⁰² Electrochemical evaluation of this electrode material in Li cell reveals excellent capacity retention. X-ray absorption spectroscopy demonstrates the utilization of stable cationic two-electron redox of V ions (Fig. 16). Moreover, at elevated temperatures, its reversible capacity is further enhanced, reaching at 50 °C a capacity as high as 300 mA h g⁻¹, corresponding to nearly the theoretical capacity based on V ion two-electron redox (Fig. 16). It results in delivering an energy density of 770 W h kg⁻¹ (based on the electrochemical potential of metallic Li), exceeding conventional layered oxides and phosphates. Even though lower rate capability is usually awaited from cation-disordered rock salt phases, Li_1.25Nb_0.25V_0.5O₂ displays remarkable rate capability, and this most probably originating from the presence of V ions with high electronic conductivity.

Further enhancement theoretical capacity based on cationic redox is obtained if three-electron redox is utilized. Three-electron redox of Cr³⁺/Cr⁶⁺ is reported in Li_1.2Cr_0.4Mn_0.4O₂.¹⁰³ Nevertheless, chemical compounds with hexavalent Cr ions are toxic and hindered in use. Instead, nontoxic Mo³⁺/Mo⁶⁺ three-electron redox is a promising candidate for practical battery applications. Because LiMo³⁺O₂ with a conventional layered structure has one mole of Li in the structure, only Mo³⁺/Mo⁴⁺ one electron redox is activated.¹⁰⁴ Therefore, Mo³⁺ is diluted in Li₃NbO₄ and tested as model electrode materials. A high theoretical capacity of 317 mA h g⁻¹ redox is expected for Li_9/7Nb_2/7Mo_3/7O₂ on the basis of three-electron redox of Mo ions. However, Li_9/7Nb_2/7Mo_3/7O₂ cannot be synthesized by conventional calcination and was therefore not considered thermodynamically stable.⁹³ Accordingly, an alternative mechanical milling synthesis route was employed for the preparation of a metastable phase.⁹³ Mechanical milling provides mechanical energy to materials, including shear stress, leading to solid-state phase transition.¹⁰⁵ Nanosized particles with a size of less than 10 nm are generally synthesized by mechanical milling, but nanosized particles are agglomerated, and thus, the surface area of secondary particles is relatively small unlike non-agglomerated nanosized samples.¹⁰⁶ By employing mechanical milling, nanosized and low crystallinity Li_9/7Nb_2/7Mo_3/7O₂ was synthesized. As-prepared sample shows a reversible capacity superior to 280 mA h g⁻¹. This capacity performance for Li_9/7Nb_2/7Mo_3/7O₂ reaches nearly 85% of theoretical capacity on the basis of the three-electron Mo cationic redox.

Note that the synthesis of nanosized samples unlocks a new way for the design of high-capacity electrode materials. Stoichiometric LiVO₂ is electrochemically inactive because of V migration on delithiation and the formation of rock salt phase.⁷⁸ The percolation network for Li ions with 0-TM channel (Fig. 10) is not formed for stoichiometric rock salt oxides. Instead, 2-TM sites, at which strong repulsive interaction for the migration process is anticipated, are enriched for the stoichiometric layered oxides. Nevertheless, nanosized and rock salt LiVO₂ synthesized by mechanical milling is electrochemically active.¹⁰⁷ Li ions can probably migrate through the network formed by 1-TM channels, which has relatively low repulsive interaction from transition metal ions. Sacrifice on electrode kinetics is unavoidable, but nanosizing particles make the Li migration possible for the stoichiometric rock salt oxides associated with a shorter migration distance. Similar phenomena are also observed for LiMnO₂ and NaMnO₂.⁹⁴ Moreover, better capacity retention is achieved for these stoichiometric rock salt oxides without the use of anionic redox.

Another important factor for the rock salt oxides is found in voltage profiles. Cation disorder raises the voltage slope on charge/discharge by increasing the statistical distributions of transition metal environments next to Li sites. Moreover, the voltage slope increase for the rock salt oxides is proposed to be relatively smaller for high voltage transition metals because of a more efficient screening of electrostatic interaction between transition metal and Li ions by the presence of oxide ions.¹⁰⁸ 

By nature, fluoride ion is the most electronegative element, and therefore, it induces a rise of the electrochemical potential. In the case of metal fluoride, a higher operating voltage is generally obtained, such as FeF₃ with Fe²⁺/Fe³⁺ redox couple.¹⁰⁹ However, for FeF₃ without Li ions in the host structure, the preparation of a functional full cell with FeF₃-positive electrode and graphite negative electrode is challenging. The use of metal fluorides, which contain Li ions in the structure, is beneficial for practical battery applications. Analogous to previously studied layered LiMeO₂, LiMeF₂ is being considered as a potential electrode material. However, its material design is restricted because only monovalent transition metals ions are utilized. Inferior electronic conductivity is another practical issue in the case of a pure fluoride system. One strategy to solve these practical problems is a mixed-anion system, as LiMeOF, where both oxide and fluoride ions are employed in the structure, which allows the introduction of divalent transition metal ions. LiFe²⁺OF is successfully obtained by Li ion insertion into Fe³⁺OF.¹¹⁰ Direct synthesis of LiFe²⁺OF is also possible by mechanical milling¹¹¹ and enables the Fe²⁺/Fe³⁺ redox to reversibly extract 1 mol of Li ions from its crystal lattice. In addition, the formation of Li-excess metal oxyfluorides can be obtained following the same mechanical milling procedure. Li₂VO₂F, which is metastable, was the first Li-excess metal oxyfluoride reported.¹¹² Partial fluorination of metal oxides leads to lowering the oxidation states of transition metal ions and thus effectively modifies its balance between cationic and anionic redox on electrochemical reactions.^38,113 Tetravalent transition metal ions are needed in Li₂MeO₃ whereas trivalent transition metal ions can be used in Li₂MeO₂F.

Electrochemical properties of Li-excess Mo-based oxide and oxyfluoride, Li₂MoO₃⁹³ and Li₂MoO₂F,¹¹⁴ are compared in Fig. 17. Both nanosized oxides and oxyfluorides were synthesized through mechanical milling. Li₂MoO₂F with a cation-/anion-disordered rock salt structure is synthesized from a mixture of LiF and LiMoO₂. In a similar manner, mechanical milling was employed to prepare nanosized and rock salt Li₂MoO₃ from a mixture of LiMoO₂ and Li₄MoO₅. Both Li-excess Mo oxide and oxyfluoride reversibly deliver large capacities superior to 320 mA h g⁻¹ through reversible Mo cationic redox. Oxidation numbers of Mo ions are tetravalent and trivalent states for Li₂MoO₃ and Li₂MoO₂F, respectively, and the energy of Mo 4d orbitals is anticipated to be higher for Li₂MoO₃ with Mo⁴⁺ ions. However, the d orbital energy shift is brought by the presence of fluoride ions with high electronegativity as illustrated in Fig. 17. As a result, the average voltage is approximately 0.3 V higher for Li₂MoO₂F with Mo³⁺ ions. Nevertheless, better rate capability is observed for Li₂MoO₃ without fluoride ions. A practical issue of oxyfluorides is found in the higher solubility of oxyfluoride into electrolyte solutions. Ti substitution into Li₂MoO₂F effectively improves the electrode reversibility associated with the suppression of dissolution into electrolytes.¹¹⁵ Reversibility is further improved by the use of concentrated electrolyte solutions without uncoordinated solvents to Li salts because transition metal ions are dissolved by uncoordinated solvents as shown in Figs. 17(d) and 17(e).

The electrochemical performance at room temperature of other Li-excess metal oxyfluorides with the anion-/cation-disordered structure, including Li₂VO₂F,¹¹² Li₂CrO₂F,¹¹⁶ and Li₂MnO₂F,¹¹⁷ is displayed in Fig. 18. The data of Li₂MoO₂F are also plotted for comparison. These metal oxyfluorides were prepared by ball milling, similarly to Li₂MoO₂F. As compared with Li₂MoO₂F, Li₂VO₂F displays a small increase in operating voltage. For Li₂CrO₂F and Li₂MnO₂F, an even larger increase in the operating voltage is observed; nevertheless, this increase is accompanied by a dramatic rise of the polarization upon discharging. As an increase in the operating voltage, a larger contribution of anionic redox is anticipated. Recently, the formation of trapped oxygen molecules inside particles on charge is observed in Li₂MnO₂F.¹¹⁸ It has been also reported that capacity performance in Li₂Mn_2/3Nb_1/3O₂F reversibly reaches nearly 300 mA h g⁻¹ in lithium cells through partial anionic redox and cationic Mn²⁺/Mn⁴⁺ redox.^113,119 Also, metal oxyfluorides with high crystallinity have been directly prepared by a high-pressure technique,¹²⁰ and such research progresses further accelerate the development of metastable electrode materials for practical battery applications.

Layered materials with Li ions generally crystallize into the O3-type layered structure with the stoichiometric composition, for example, LiCoO₂ and LiNiO₂. Similarly, O3-type and stoichiometric layered materials with Na ions, NaCoO₂, etc., can be synthesized and were tested as electrode materials in 1980s.¹²¹ In addition, larger Na ions stabilize Na-deficient and non-stoichiometric layered phases, and P2- and P3-type layered oxides can be synthesized. For both phases, Na ions are located at prismatic sites, but numbers of MeO₂ layers in the unit cells are different, 2 or 3, as shown in Fig. 19. Oxygen–oxygen electrostatic interaction between MeO₂ layers is larger for prismatic sites compared with octahedral sites, leading to wider interlayer distances for P2 and P3 phases. Facile Na diffusion with lower activation energy is thus achieved. For P2 and P3 phases, four different Na sites are found, which have different face- and edge-sharing manners with MeO₂ layers as shown in Fig. 19. In O3 phase, Li/Na ions at octahedral sites share three edges with MeO₂ layers, and moreover, Li/Na ion migration to neighboring octahedral sites is achieved via narrow intermediate tetrahedral sites,¹²² potentially leading to relatively high activation energy. When Na ions are partially extracted from O3 phase, Na ions are energetically preferred to occupy prismatic sites, often resulting in phase transition to P3 phase.¹²³ Such phase transition is also unique for the Na system and is not observed in the Li system because small Li ions cannot be stabilized at large prismatic sites. Na-deficient P2 and P3 phases can be sometimes used as both positive and negative electrode materials, while stoichiometric O3 phase can only be used as positive electrode materials.^124,125 In recent years, following the path drawn by the three decades of experiences from Li-ion battery research, layered Na_xMeO₂ materials with different compositions and layered structures have been again targeted as electrode materials for Na-ions batteries.^124–138 

To increase the reversible capacity of positive electrode materials, anionic redox has been also extensively studied for Na insertion materials. Potentially, Na₂MnO₃ is a straightforward candidate for Na anionic redox as it is analogous to intensively studied Li₂MnO₃. However, the preparation of Na₂MnO₃ (Na(Na_1/3Mn_2/3)O₂) remains quite challenging through traditional preparation methods due to the great size difference between Mn⁴⁺ (0.67 Å) and Na⁺ (1.16 Å) ions. To overcome this obstacle, an alternative route, the synthesis of Na(Li_1/3Mn_2/3)O₂, rather than Na(Na_1/3Mn_2/3)O₂ has been introduced,⁶⁶ and recently, successfully synthesis of O3-type layered structure Na(Li_1/3Mn_2/3)O₂ has been achieved.¹³⁹ This material displays a well-defined voltage plateau in Na cells, which is similar to Li₂MnO₃ in Li cells. Additionally, the electrode reversibility of Na(Li_1/3Mn_2/3)O₂ is much better than Li(Li_1/3Mn_2/3)O₂ [Fig. 20(a)].¹³⁹ Mn migration for O3 Na(Li_1/3Mn_2/3)O₂ would be suppressed because Mn ions are too small to occupy the vacant octahedral sites formed in Na layers for O3 Na_y(Li_1/3Mn_2/3)O₂, leading to superior reversibility as electrode materials even though oxygen loss is observed in the initial charge process. Also, the synthesis of Na_5/6(Li_1/4Mn_3/4)O₂, a non-stoichiometric Na-deficient phase, with the P2-type layered structure was reported and its application to Na cells displayed excellent reversible capacity based on cationic and anionic redox.⁶⁶ Equivalently, P2 Na_0.6Li_0.6Ni_0.25Mn_0.75O_y has been reported to reversibly deliver a large capacity.¹⁴⁰ Li ions can be replaced by Mg ions in the transition metal layers, resulting in Na_2/3(Mg_0.28Mn_0.72)O₂ with the P2-type layered structure. This material is able to reversibly reach a capacity superior to 200 mA h g⁻¹ [Fig. 20(b)],¹³⁶ and subsequently, the reversible anionic redox contribution was reported.¹⁴¹ Mg ions also provide the creation of and unhybridized orphaned O 2p states, similar to the excess Li ions in the host structure. However, on charge/discharge, these layered oxides display wider voltage hysteresis with high reversible capacities through anionic redox.

The suppression of such voltage hysteresis for anionic redox is recently evidenced for Na₂Mn₃O₇ [Fig. 20(c)].¹⁴² Na₂Mn₃O₇ is reformulated to Na_5/9□_1/6Mn_5/6O₂, and herein, □ denotes vacant octahedral sites in transition metal layers.¹⁴³ The presence of vacant sites also results in the formation of undercoordinated oxygen and thus unhybridized O 2p states. In a similar way, a reversible voltage plateau at 4.2 V in Na cell with small hysteresis is also observed for Na_0.6(Li_0.2Mn_0.8)O₂ with the P3-type layered structure [Fig. 20(d)].¹⁴⁴ It has been proposed that small voltage hysteresis correlates with stabilization of holes and suppression of oxygen dimerization as shown in Fig. 21.¹⁴² Nevertheless, the enrichment of holes in O 2p orbitals inevitably results in its destabilization and dimerization, and thus, small voltage hysteresis would be achieved for a limited reversible capacity, <100 mA h g⁻¹.

Na-excess Mn-based oxides synthesis, for example, Na_1.3Nb_0.3Mn_0.4O₂¹⁴⁵ Na_1.14Ti_0.29Mn_0.57O₂,¹⁴⁶ has been also reported. Even if the size of Na⁺ ions is much larger than those of Nb⁵⁺, Ti⁴⁺, and Mn³⁺ ions, metastable and low crystallinity oxides with the cation-disordered rock salt structure are synthesized by mechanical milling. However, inferior electrode reversibility is observed when compared with stoichiometric and nanosize NaMnO₂ prepared by mechanical milling,⁹⁴ indicating that anionic redox is less stable for these Na-excess rock salt oxides with low crystallinity and larger surface area prepared by mechanical milling.

Another important chemistry for Na insertion materials with anionic redox is found in NaFeO₂^147,148 and other layered materials containing trivalent Fe ions.^{127,132,138,149} Although oxidation of Fe³⁺ to Fe⁴⁺ is evidenced for these compounds by Mössbauer spectroscopy,¹³⁷ a recent study on NaFeO₂ reveals that the contribution of anionic redox cannot be ignored.¹⁵⁰ Electrode performance of NaFeO₂ in a Na cell is displayed in Fig. 22(a), and excellent capacity stability is obtained when the cutoff voltage is lower than 3.5 V vs Na.¹⁴⁷ A small energy shift is detected in Fe K-edge XAS spectra on charge [Fig. 22(b)], while a clear variation is noted for O K-edge XAS spectra during Na extraction for NaFeO₂ [Fig. 22(c)]. Activation of anionic redox results in destabilization of the particle surface, which triggers chemical/electrochemical reaction with electrolyte. The degradation of electrode reversibility with higher cutoff voltage partially derives from the anionic redox instability. Anionic redox is also evidenced for NaVO₃ with pentavalent V ions. This material is able to reversibly deliver a capacity superior to 200 mA h g⁻¹ through anionic and cation redox of V ions.¹⁵¹ 

In Li-ion batteries, graphite is conventionally used as a negative electrode material because of its low price, abundance, and excellent ability to reversibly store Li ions. Nevertheless, graphite is ineffective for electrochemical Na ions storage. Indeed, within aprotic electrolyte solutions containing Na salts, its intercalation solely results in the formation of a high stage graphite intercalation compound.¹⁵² Alternatively, hard carbon, which presents a different morphology than graphite,¹⁵³ is employed as a negative electrode for Na cells.^154,155 This material is able to absorb and store Na ions within its pores by forming quasi-metallic Na.²⁴ However, hard carbon electrochemically absorbs Na ions at a potential near to metallic Na plating voltage (<0.1 V), making fast charging full cell truly challenging.¹⁵⁶ 

A promising alternative candidate for Na storage negative electrode materials are titanium-based oxides, such as Na₂Ti₃O₇,¹⁵⁷ Na₂Ti₆O₁₃,¹⁵⁸ Na₄Ti₅O₁₂,¹⁵⁹ TiO₂,¹⁶⁰ etc.¹⁶¹ Within the literature, P2-type Na_0.6Cr_0.6Ti_0.4O_2,¹²⁴ and P3-type Na_0.58Cr_0.58Ti_0.42O₂,¹²⁵ two Cr substituted Ti-based layered oxides, display excellent reversible electrochemical performances as negative electrode for Na ions storage. Also, Na-deficient and Cr-free O3-type Na_0.67Mg_0.33Ti_0.67O₂ was reported.¹⁶² Through Ti³⁺/Ti⁴⁺ cationic redox reaction, these Ti-based layered oxide materials display a slightly lower operating potential within 1.1 to 0.5 V vs metallic Na. Furthermore, these Ti-based layered oxides demonstrate excellent capacity retention and rate capability.¹³⁴ Indeed remarkably, these materials present lower operating potential than their Li analogous, such as Li_4+xTi₅O₁₂, which has a spinel-type structure and displays an operating voltage of 1.55 vs metallic Li,¹⁶³ as presented in Fig. 23. Ti-based negative electrode materials for Li-ion batteries are now able to display advantageous safety, cyclability, and rate capability in comparison with conventional graphite materials due to their reduced volume expansion and ability to prevent dendritic growth of metallic Li.¹⁶⁴ However, this latter point is enabled by the higher operating voltage of Ti-based materials, which inevitably causes energy density loss as compared with graphite used in practical Li-ion batteries and therefore limits their competitiveness. On the other hand, Ti-based layered oxide materials for Na-ions batteries, owing to their lower operating potential, represent an excellent choice for the negative electrode, which combines moderate energy density, excellent rate capability, safety, and cyclability. Because the difference in standard electrode potential for metallic Li and Na is only of 0.33 V, the large potential difference (0.75 V on average) for Li_4+yTi₅O₁₂ and P3 Na_0.58+yCr_0.58Ti_0.42O₂ cannot be explained only from standard electrode potential. This observation originates from a lower covalent bonding nature for Na ions to oxide ions. Therefore, as the concentration of Na ions in solid increases on reduction, the voltage difference becomes more visible for both systems. Also, lower covalency is the important character to facilitate the mobility of Na⁺ ions in solid. A neutron diffraction pattern of Na_0.67Cr_0.33Mg_0.17Ti_0.50O₂ with the P2-type layered structure, which is partially Mg-substituted sample of Na_0.6Cr_0.6Ti_0.4O₂, is shown in Fig. 24.¹³⁴ Structural analysis shows that, as it is often the case for the P2-type structure, Na ions are located at two different prismatic sites, 2b and 2d sites.¹³⁸ For the Na ions located at the 2d prismatic sites, three edges are shared with octahedral sites from the transition metal layer and are therefore expected to be energetically stable sites. On the other hand, at the 2b prismatic sites, Na ion shares a face with an octahedral site from the transition metal layer, which is therefore expected to generate a strong electrostatic interaction. Nevertheless, when such P2-layered phases have a wide inter-layer distance, the adverse interaction at the face-shared site is suppressed, allowing Na ions to be located at the 2b prismatic sites. This experimental observation also suggests that self-diffusion of Na ions is expected to be fast in these layered materials.

Electrochemical properties of Na_0.67Cr_0.33Mg_0.17Ti_0.5O₂ and Na_0.67Cr_0.66Ti_0.37O₂ in Na cells are compared in Fig. 24(b). Indeed, since the mass of Mg²⁺ is lower than Cr³⁺, substituting Cr³⁺ by Mg²⁺ directly increases the theoretical capacity. As expected, the Mg-substituted sample displays superior reversible capacity. At 10 mA g⁻¹, a reversible capacity of 96 mA h g⁻¹ is obtained which almost corresponds to the theoretical capacity of Na_0.67+yCr_0.33Mg_0.17Ti_0.5O₂ where 1/3 mol of Na ions is reversibly exchanged. Both samples display similar voltage profiles [Fig. 24(b)]; however, the Mg-containing sample is noted to have a broadened differential capacity profile [Fig. 24(c)]. The substitution of Cr³⁺ by Mg²⁺ also affects the phase transition process, leading to the loss of the voltage plateaus at 0.55 and 0.60 V. Additionally, Na_0.67Cr_0.33Mg_0.17Ti_0.5O₂ displays a remarkable rate capability response reaching a capacity superior to 60 mA h g⁻¹ at 2560 mA g⁻¹ as shown in Fig. 24(d). Furthermore, this material demonstrates outstanding capacity retention without any capacity loss over 450 cycles at 100 mA g⁻¹ as shown in Fig. 24(e).¹³⁴ 

Although available energy density as battery applications is inevitably smaller for the Na chemistry once compared with the Li chemistry with small-sized ions coupled with lower standard electrode potential, higher mobility of Na⁺ ions are beneficial to design not only cost-effective, but also high-performance batteries (i.e., high rate capability, safe, and long cycle life batteries) with these Ti-based negative electrode materials. Gravimetric capacity (mA h g⁻¹) is smaller compared with hard carbon, but its density is relatively higher (3.6 g cm⁻³ for Na_0.67Cr_0.33Mg_0.17Ti_0.5O₂) than hard carbon (<2.0 g cm⁻³),¹⁵⁵ leading to smaller differences in volumetric capacity (mA h cm⁻³). Higher density (g cm⁻³) of Ti-based oxides compared with carbonaceous materials is another benefit for battery applications.

Currently, derivatives of layered LiCoO₂ and LiNiO₂ are widely utilized for commercial battery applications. Ni-/Co-free and high energy electrode materials are necessary to further reduce the battery cost, and Li-excess oxide/oxyfluorides with rock salt structures are the attractive candidate for this purpose. In order to improve the theoretical capacity of positive electrode materials, increasing the Li⁺/Na⁺ contents while employing less transition-metal ions has been shown to be an effective approach. Indeed, in the past decades, several original rock salt structures based on Li-excess positive electrode materials have been reported. Numerous potential materials are achievable including various chemistry based on both cationic and anionic redox as multi-electron redox reaction. Recently, the use of Li-excess rock salt oxides with V ions has been suggested as a low-voltage negative electrode material on the basis of V²⁺/V³⁺ redox.^165–168 A practical problem for anionic redox of oxide ions with Mn ions is found in cyclability, and the suppression of degradation of electrode reversibility, related to oxygen loss on continuous cycles, is required. Further studies on destabilization mechanisms for anionic redox are encouraged to solve the cyclability problem. The excess enrichment of holes in oxygen on charge results in irreversible oxygen dimerization,⁹¹ thus leading to oxygen loss.⁸⁶ The suppression of Mn migration, which triggers oxygen loss and phase transitions, is also effective to improve reversibility.^63,64,66,139 Further experimental and theoretical understanding of electrode materials with reversible anionic/cationic redox will ensure the development of affordable and higher-energy battery systems on the basis of these emerging electrode materials. Highly reversible anionic redox is realized for Ru ions because Ru ions are chemically stable even for higher oxidation states as 4d transition metal ions. Charge compensation by anionic species is achieved, for the oxide/oxyfluoride system, as well as for the Li-excess metal sulfide system, for example, Li₂TiS₃.^169,170 Anionic redox is easily activated for soft sulfide ions with a higher polarizable character when compared with hard oxide ions. The research progress on theoretical understanding of anionic redox is expected to contribute to the further development of Li-/Na-excess electrode materials with higher theoretical capacities. Such high-capacity positive electrode materials containing Li/Na ions are also necessary for “Li-free”^171–173 and/or “Na-free”^174,175 battery manufacturing with higher energy density as proposed in recent publications. These negative electrode-free batteries utilize in situ formed Li/Na metal as negative electrode materials. Therefore, energy density of the batteries mostly depends on Li/Na ion concentrations in positive electrode materials, thus leading to the development of higher energy density as rechargeable battery applications when compared with commercial Li-ion batteries.

In addition to such potential high-energy density batteries, the development of cost-effective batteries is urgent to guarantee a fair and global energy production transition toward sustainability. For this purpose, commercialization of Na-ions batteries made from abundant elements, without Li, Ni, and Co ions, is on the rise thanks to progress in research in the past decade. Indeed, even if by nature Na is heavier than Li, indicating that energy density of Li-ion batteries will always be superior to Na-ion batteries, this latter technology is now realistic as it achieves better rate capability and similar cyclability with Li-ion batteries. Furthermore, metallic Na does not alloy with Al, unlike Li, which enables the use of Al current collector rather than expensive Cu current collector increasing its overall competitiveness. However, in general, layered oxides with Na ions are hygroscopic,¹⁷⁶ similar to Ni-enriched layered oxides with Li ions.¹⁷⁷ This fact inevitably results in the increase in battery production cost. Therefore, the development of effective technology for surface coating, etc., is needed, by which electrode materials would be handled in ambient atmosphere without the battery assembling in a dry room.

N.Y. acknowledges the partial support from JSPS, Grant-in-Aid for Scientific Research (Grant Nos. 15H05701, 19H05816, and 21H04698), and MEXT program “Elements Strategy Initiative to Form Core Research Center (No. JPMXP0112101003),” MEXT (Ministry of Education, Culture, Sports, Science and Technology), Japan.

The authors declare no conflicts of interest to disclose.

The data that support the plots in Figs. 3, 4(a), 18, and 23 are available from the corresponding author upon reasonable request.