Unexpectedly Large Contribution of Oxygen to Charge Compensation Triggered by Structural Disordering: Detailed Experimental and Theoretical Study on a Li3NbO4–NiO Binary System

Dependence on lithium-ion batteries for automobile applications is rapidly increasing. The emerging use of anionic redox can boost the energy density of batteries, but the fundamental origin of anionic redox is still under debate. Moreover, to realize anionic redox, many reported electrode materials rely on manganese ions through π-type interactions with oxygen. Here, through a systematic experimental and theoretical study on a binary system of Li3NbO4–NiO, we demonstrate for the first time the unexpectedly large contribution of oxygen to charge compensation for electrochemical oxidation in Ni-based materials. In general, for Ni-based materials, e.g., LiNiO2, charge compensation is achieved mainly by Ni oxidation, with a lower contribution from oxygen. In contrast, for Li3NbO4–NiO, oxygen-based charge compensation is triggered by structural disordering and σ-type interactions with nickel ions, which are associated with a unique environment for oxygen, i.e., a linear Ni–O–Ni configuration in the disordered system. Reversible anionic redox with a small hysteretic behavior was achieved for LiNi2/3Nb1/3O2 with a cation-disordered Li/Ni arrangement. Further Li enrichment in the structure destabilizes anionic redox and leads to irreversible oxygen loss due to the disappearance of the linear Ni–O–Ni configuration and the formation of unstable Ni ions with high oxidation states. On the basis of these results, we discuss the possibility of using σ-type interactions for anionic redox to design advanced electrode materials for high-energy lithium-ion batteries.

S-4 Figure 3a. Sequential Rietveld refinements of the data were carried out based on the structure of LiNi 2/3 Nb 1/3 O 2 adopting orthorhombic symmetry with the space group Fddd. Reliable refinement of atomic structures in the operando data was not deemed feasible and the atomic structure was consequently fixed to the structure obtained from the neutron diffraction data (Table S1) during the sequential refinements. The lattice parameters, scale-factor, zero-shift, and background (Chebychev polynomial) were refined.
The contribution from the aluminum foil was accounted for by implementing an aluminum phase (space group Fm-3m) with a high degree of preferred orientation in the modelling. No additional crystalline phases were observed in the data. The quantitative structural analysis was limited to the relative changes in lattice parameters (which do not rely on peak intensities) and relative variations in the scale factor (see Supporting Figure 9b). The scale factor encompasses 2θ-independent contributions, such as incident beam flux, instrumental geometry, detector efficiency, sample volume, etc. Assuming the instrumental contributions to be constant in time, observed variations in the refined scale factor of a given phase will be proportional to changes in the amount of coherently scattering material of the given phase. For the LiNi 2/3 Nb 1/3 O 2 sample, a small continuous increase (~10%) in the scale factor over the entire two electrochemical cycles is observed (Supporting Figure S9 b-1). No significant changes are observed in the unit cell parameters, suggesting highly reversible structural changes on electrochemical cycles, which is consistent with the ex-situ XRD data. It should be noted that the ex-situ and operando experiments S-5 were run at different current rates.
Supporting Figure S10a shows the background subtracted and normalized in operando XRD contour map of the Li 4/3 Ni 2/9 Nb 4/9 O 2 (x = 0.67) sample. As seen in the figure, no major additional crystalline phases are formed during cycling. However, closer inspection of the reflections shows a gradual decrease in the main peak intensities, which is associated with the appearance of a shoulder on the left side (lower 2θ) of the peaks (see Figure 3a). The shoulder is observed for all reflections, indicating the formation of an isostructural compound with a larger unit cell, which is expected to be a delithiated (or partially delithiated) disordered rocksalt phase formed upon charge.
Notably, the in operando XRD study reveals that the degradation of the main phase is seemingly irreversible as it does not reform upon discharge. Supporting Figure S10b shows an enhancement of the (200) and (220) reflections at selected times throughout the two electrochemical cycles. Closer inspection of the data also reveals that a more complex series of structural transitions takes place, in particular during the initial charge, where the peaks initially shift to higher 2θ values (unit cell decrease) and subsequently back again along with the appearance of a broad shoulder at lower 2θ. A decrease in the overall combined intensity of the reflections (~25%) is also observed during the first charge (see Supporting Figure S10b top left). Sequential Rietveld refinements of the data were carried out based on the disordered rock salt structure of Li 4/3-y Ni 2/9 Nb 4/9 O 2 (space group Fm-3m).
Again, reliable refinement of atomic structure in the operando data was not deemed feasible and the S-6 atomic structure was fixed to the structure obtained from the neutron diffraction data (Table S2) during the sequential refinements. Based on the observation of peak shoulders above, an additional delithiated phase (space group Fm-3m) was implemented in the modelling. Lattice parameters, scalefactor, zero-shift, background (Chebychev polynomial) and one Lorentzian profile parameter associated with isotropic microstrain per phase were refined. The contribution from the aluminum foil was accounted for by implementing an aluminum phase (space group Fm-3m) with a high degree of preferred orientation in the modelling. Representative Rietveld fits at different levels of charge are shown in Supporting Figure S10c. Here, the data has been modelled by the two endmember phases, i.e. the original disordered rock salt structure of Li 4/3 Ni 2/9 Nb 4/9 O 2 and a delithiated disordered rock salt structure. However, the asymmetrical peak shapes likely arise from a non-trivial distribution of lattice parameters caused by gradual extraction of various amounts of Li from the structure and thus non-uniform Li distributions in certain parts of the sample. The implementation of a third highly strained (partially delithiated) rock salt phase in the refinement did not give substantially better fit and furthermore destabilized the refinement due to parameter correlations (see Supporting Figure S10d). The fits to the peak profiles were not entirely robust, and care must be taken not to draw false conclusions based on the Rietveld refined parameters. The refined scale factors suggest a continuous decrease in the amount of the main Li 4/3 Ni 2/9 Nb 4/9 O 2 phase along with a concurrent increase in the secondary range of delithiated phases, which confirms the gradual S-7 transformation of the main phase (also see Figure 3a). The original phase does not reform but rather continues to diminish during discharge indicating an irreversible degradation. This also supports the observations from ex-situ data.  for each composition as shown in Supporting Figure S23b. Note that no oxygen dimers are visible over the entire range of the composition y (0 ≤ y ≤ 1 in Li 1-y Nb 1/3 Ni 2/3 O 2 ), as the minimum bond distance between two oxygen atoms is 2.49 Å (see Supporting Figure S20). The calculated voltages are almost constant ranging from ~4.3 to 4.5 V at y < 2/3, and at y > 2/3 abruptly increase in the voltage

S-9
reaching to > 5V.  Comparison of XRD patterns, SEM images, and electrode performance of LiNi 2/3 Nb 1/3 O 2 synthesized at 1000 o C for 48 and 2 h. Synthesis with longer time results in higher degree of cation ordering, but similar cation ordering is also noted for the sample synthesized for 2 h.

Supporting Tables and Figures
Moreover, much better electrode performance is observed for the sample synthesized for 2 h, likely associated with particle size differences between the samples. Normalized Intensity / a.u.

Photon Energy / eV
As-prepared Half charged Fully charged Normallized Intensity / a.u.

Photon Energy / eV
As-prepared Half charged Fully charged LiNi 0.5 Mn 0.5 O 2

Ni L-edge PFY
Ni L 3 -edge Figure S12. Changes in XAS spectra at Ni L-edge and O K-edge for layered Li 1-y Ni 0.5 Mn 0.5 O 2 . A clear change in a Ni L-edge spectrum is observed for the half charged state (ca. 110 mA h g -1 ) when compared with Li 1-y Ni 2/3 Nb 1/3 O 2 . On further oxidation to 4.8 V, the formation of Ni 4+ is noted, which is not evidenced for Li 1-y Ni 2/3 Nb 1/3 O 2 . Peak shift of O K-edge to a lower energy region is also noted, and this observation is also clearly different from Li 1-y Ni 2/3 Nb 1/3 O 2 . FT magunitude R / Å full-charged discharge 60 mAh g -1 full-discharged Figure S14. Extended X-ray absorption fine structure (EXAFS) spectra of stoichiometric  PLOS ONE, 7, e49182, 2012). Note that a crystalline Ni metal phase was not found by XRD (Figure S8), suggesting that metallic Ni with an amorphous phase or that is nanosized is formed.  Figure S18. Ni L-edge and O K-edge XAS spectra of LiNi 2/3 Nb 1/3 O 2 (x = 0.33) and Li 4/3 Ni 2/9 Nb 4/9 O 2 (x = 0.67) electrodes obtained at PEY mode which is surface sensitive. Holes are stabilized in the bulk of particles as observed in PFY mode (Figure 4c).