Intercalation and Push‐Out Process with Spinel‐to‐Rocksalt Transition on Mg Insertion into Spinel Oxides in Magnesium Batteries

On the basis of the similarity between spinel and rocksalt structures, it is shown that some spinel oxides (e.g., MgCo2O4, etc) can be cathode materials for Mg rechargeable batteries around 150 °C. The Mg insertion into spinel lattices occurs via “intercalation and push‐out” process to form a rocksalt phase in the spinel mother phase. For example, by utilizing the valence change from Co(III) to Co(II) in MgCo2O4, Mg insertion occurs at a considerably high potential of about 2.9 V vs. Mg2+/Mg, and similarly it occurs around 2.3 V vs. Mg2+/Mg with the valence change from Mn(III) to Mn(II) in MgMn2O4, being comparable to the ab initio calculation. The feasibility of Mg insertion would depend on the phase stability of the counterpart rocksalt XO of MgO in Mg2X2O4 or MgX3O4 (X = Co, Fe, Mn, and Cr). In addition, the normal spinel MgMn2O4 and MgCr2O4 can be demagnesiated to some extent owing to the robust host structure of Mg1−xX2O4, where the Mg extraction/insertion potentials for MgMn2O4 and MgCr2O4 are both about 3.4 V vs. Mg2+/Mg. Especially, the former “intercalation and push‐out” process would provide a safe and stable design of cathode materials for polyvalent cations.

shows the cyclic voltammogram measured for a CsTFSA solvent (TFSA: bis(trifluoromethanesulfonyl)amide, N(CF 3 SO 2 ) 2 − ) using a Pt working electrode at 150 • C. The electrodeposition and stripping of Cs is observed at around −0.2 V vs. Li + /Li. The small cathodic peak at about 0.8 V vs.
Li + /Li corresponds to the reductive decomposition of the solvent, while the CsTFSA solvent is stable to oxidative decomposition up to about 4.5 V vs. Li + /Li. Any anodic and cathodic peaks are not observed in the potential range of 1-4.5 V vs. Li + /Li, suggesting that the electrochemical decomposition of the solvent is irreversible. Incidentally, the cathodic current around 1-1.5 V is considered to correspond to the cathodic decomposition of TFSA anion, as mentioned in the body text. Therefore, it can be concluded that the peak couples seen in Figure 2 correspond to the redox reactions of active materials. In the CV profiles in Figure 2a, we have concluded that these redox behaviors are due to the electrodeposition/stripping of Mg. However, since we used a Mg metal as a working electrode (so as to retard the cathodic decomposition of TFSA anions), there is a possibility that Li can be inserted into Mg metal to form Mg-Li alloy.
Here, we also considered this effect by using a (Li10/Cs90)-TFSA binary ionic liquid; see Figure S2.
The shape of the CV profile for the insertion/extraction of Li into/from the Mg matrix is crucially different from the CV shape for the electrodeposition/stripping phenomenon of Li, because the former reflects the gradual change in the chemical potential of Li (that shifts to the less-noble side), by which the current curve returning back to the anodic side overruns that toward cathodic side, leading to a less-nobler potential of the extraction of Li. This is a typical CV shape for the solid-solution phenomenon. On the other hand, in the case of the electrodeposition/stripping phenomenon, the overpotential is inevitably observed more or less on the electrodeposition, so that the current curve returning toward anodic side inevitably underruns the current curve going toward cathodic side, and it eventually crossovers with each other. Thus, the CV shapes between electrodeposition/stripping and insertion/extraction phe- included in the CV profile more or less, but the upper CV profile mainly indicates the electrodeposition/stripping of Mg.

Rietveld analysis before/after electrochemical tests
The crystal structure parameters were determined by Rietveld refinement with the program RIETAN-FP[2] using the XRD profiles of the as-synthesized MgCo 2 O 4 sample and one after partial discharge (120 mAh g −1 ) in the Mg battery system that corresponds to Figure 3a. Although we mentioned in the body text that the rocksalt phase would have vacancies after discharge, here we assume that the discharged rocksalt structure does not have any vacancies. The cation ratio of the spinel structure was fixed at Mg/Co = 1/2, and no constraint was imposed on the cation ratio in the rocksalt structure, i.e., the discharge amount (120 mAh g −1 ) was not taken into account for the Rietveld analysis.
The fitting results and refined crystal structure parameters are shown in Figure S3 and Table S1, respectively. The Rietveld refinement revealed that the as-synthesized MgCo 2 O 4 has a disordered spinel structure, and the structure of the partially discharged Mg 1+x Co 2 O 4 contains spinel and rocksalt phases, whose molar fraction (or volume fraction in this case) is determined to be spinel (27%) and rocksalt (73%) structures, indicating that about 70 % of the discharge process proceeds in terms of the present structure analysis. Considering the fact that the discharge amount was less than half of the full capacity (120 mAh g −1 / 260 mAh g −1 ), we need to consider the presence of vacancies in the rocksalt crystal. Thus, the insertion of one Mg atom induces the spinel to rocksalt transition in a larger region than  Figure 2a) and that for the insertion/extraction of Li into/from the Mg matrix (lower). The redox potential in each case is located closer, but the CV shapes are crucially different from each other.
one unit cell of the rocksalt structure.

Ab initio calculations
Determination of spin and magnetic configurations in spinel MgCo 2 O 4 We here discuss the stable spinel structure of MgCo 2 O 4 before inserting Mg or Li cations. Since MgCo 2 O 4 is a partially disordered spinel, the model has three degrees of freedom, i.e., the cation configuration, spin configuration, and magnetic configuration. Regarding the cation configuration, we considered all possible configurations in the primitive unit cell comprising 14 atoms, i.e., 6 C 2 = 15 configurations, including symmetrically equivalent ones. For each atomic configuration, we searched for the most stable spin and magnetic configuration within the collinear magnetic configuration in the unit cell. In MgCo 2 O 4 , Co is trivalent, and can take high-or low-spin states at the tetrahedral or octahedral sites. Especially, it is well known that the energies of Co(III) high-and low-spin states at the octahedral sites, the latter of which has no local magnetic moment, are so close that both can exist depending on temperature in some materials. On the other hand, the high-spin state would be favorable for Co(III) at the tetrahedral sites. Therefore, we considered S1: Crystal structure parameters of MgCo2O4 as-synthesized and after partial discharge (120 mAh g −1 ) in the binary ionic liquid of Mg(TFSA)2:CsTFSA=1:9 at 150 • C. a The figure in parenthesis represents the error, e.g., 8.1378(3) means 8.1378±0.0003. b The notation g(Mg@8a) means the occupancy of Mg at 8a site. As to the other similar notations, the same rule is applied. c The notation B(Mg@8a) denotes the Debye-Waller factor for the Mg at 8a. As to the other similar notations, the same rule is applied.
both high-and low-spin states for the octahedral sites, whereas only the high-spin state for the tetrahedral sites. In total, we calculated, for each atomic configuration, 10, 11, or 12 spin and magnetic configurations without considering symmetry, which includes one non-spin state, 2 3 = 8 magnetic configurations for the high-spin state, and 0, 1, or 2 magnetic configurations for the mixed-spin state comprising low spins for the octahedral sites. As a result, it is found that the high-spin states are more stable than the other spin states except in one structure (No. 6 in Figure S4), where the mixed-spin state is 0.2 eV lower in energy than the most stable high-spin state. cations to give rise to the rocksalt structure as

Mg insertion/extraction into/from spinel MgCo
where e(> 0) is the elementary charge and E denote the total energies of respective phases. Mg cations were placed at the 16c sites in the primitive unit cell for each atomic configuration, and then the atomic positions and cell parameters were fully optimized. We started the structure optimization from two sets of initial atomic positions, i.e., with and without displacement of the 8a site cations into the 16c site. We considered only one cation configuration for each in the former set, since preliminary calculations indicated that the cation configuration in the rocksalt phase makes little impact on the redox potentials. Without the displacement, it has been found that some 8a cations spontaneously move to 16c sites but others do not, showing much higher energies compared to the rocksalt structures with the displacement. Subsequently, the most stable magnetic configuration was again determined for each atomic configuration. Since Co(II) is known to take the high-spin state at both of the tetrahedral and octahedral sites, we considered the high-spin state only.
Since it has been clarified that in most of configurations the total energies are substantially equal, we only considered the normal spinel structure for other spinel oxides MgX 2 O 4 , especially, for the extraction process. Based on the Nernst equation, we calculated the redox potential for Mg extraction [2] F. Izumi, K. Momma, Solid State Phenom. 2007, 130, 15-20.