[Introduction] Persistent phosphors are materials showing continuous luminescence for long durations from several minutes to hours after optical excitation has ceased. The materials have carrier traps that play the most important role in the persistent luminescence process. Ueda et al. of Kyoto University developed a blue-light-induced persistent luminescence material Ce3+ -Cr3+:Y3Al2Ga3O12 by doping Cr into Ce3+:Y3Al2Ga3O12, where the 3d orbitals of Cr3+ work as trap centers to capture electrons[1]. The persistent luminescence properties are dominated by the position of the energy levels of the trap centers in the band gap and therefore can be controlled by changing the doped transition metal (TM) ion. Generally, it is difficult to calculate the band gap energy by first-principles calculations based on the cluster method due to the poor reproduction of the conduction band. Accordingly, the transition energy from the impurity level to the conduction band cannot be calculated directly. However, the position of the impurity level within the band gap can be estimated by calculating the charge transfer transition energy from the valence band (Ligand to Metal Charge Transfer: LMCT). In this study, in order to obtain the LMCT energy for various trivalent TM ions in α-Al2O3, we performed first-principles calculations using the DVME method[2]based on configuration interaction (CI) method and constructed the theoretical energy diagram of the transition metal 3d levels within the band gap. By comparing the theoretical diagram with the available experimental LMCT energies in detail, the optimized computational condition was also proposed. [Computational Method] We constructed a 7-atom cluster with C3 symmetry consisting of the central aluminum ion and the first-neighbor oxygen ions from the crystal data of α-Al2O3 [3]and then we substituted various trivalent TM (Sc, Ti, V, Cr, Mn, Fe, Co) ions for the central aluminum ion. The effective Madelung potential was considered by setting point charges on the atomic sites around the cluster. In order to improve the computational accuracy, the lattice relaxation (LR) caused by the substitution of TM for aluminum was considered by estimating the change of the bond length based on the Shannon’s crystal radii, and the configuration dependent correction (CDC) was also considered. The LMCT energy was calculated as the transition energy from the top of the valence band corresponding to the highest one of the MOs mainly consisting of oxygen 2p orbital to the MO mainly consisting of TM 3d orbital based on the MO calculation using the DV-Xα method and the CI calculation using the DVME method. [Result] The theoretical LMCT energies calculated with LR and CDC are shown as the diagram in Fig. 1 together with the diagram for the experimental values[4-6]. The figure shows that the experimental trend was successfully reproduced by first-principles calculations. The remaining underestimation for Sc and Ti is probably due to the overestimation of the lattice relaxation arising from the simple approximation. On the other hand, the overestimation for Fe is probably due to the insufficient consideration of electron correlation. [References] [1] J. Ueda, et al,. Appl. Phys. Lett. 104, 101904, (2014). [2] K. Ogasawara, et al., Phys. Rev. B, 64, 115413, (2001). [3] E. N. Maslen, et al., Acta Crystallogr. Sect. B, 49, 973-980 (1993). [4] B. R. Namozov, et al., Physics of the Solid State, 40, 599-600 (1998). [5] H. H. Tippins, Phys. Rev. B, 1, 126-135 (1970). [6] D. S. McClure, J. Phys. Chem., 36, 2757-2779 (1962). Figure 1