Defect-mediated Jahn-Teller Effect in Layered LiNiO2
Lin Weicheng,Ye Yaokun,Chen Taowen,Jiang Yao,Ouyang Chuying,Pan Feng,Zheng Jiaxin
DOI: https://doi.org/10.1007/s40843-021-1946-9
2022-01-01
Science China Materials
Abstract:Ni-rich layered lithium transition metal oxide materials, such as LiNi1−x−yMnxCoyO2 (NMC) and LiNi1−x−yCoxAlyO2 (NCA), are used as cathode materials in lithium-ion batteries [1–4]. Improvements in the energy density and reductions in Co cost can be achieved by increasing the Ni content so that the compositions of these two materials invariably converge toward LiNiO2 (LNO). Recently, LNO has attracted increased attention for compositional designs of next-generation low-/zero-Co, ultrahigh-Ni layered oxides as alternatives to NMC/NCA. However, the crystallographic and electronic structures of LNO remain subjects of considerable controversy. LNO has been experimentally reported to exist in the R m 3 phase [5–9] and exhibit semiconductor behavior with a small band gap [10–15], while theoretical studies have reported that stoichiometric R m 3 LNO is a conductor and unstable [16–22]. Moreover, the Ni ions in LNO are considered as Jahn-Teller (JT)-active Ni3+ with the electronic configuration of t2geg [8,10,12,23]. Nevertheless, extended X-ray absorption fine structure and neutron diffraction data have indicated that only local Jahn-Teller distortion (LJTD) exist [8,24–26], implying the absence of long-range collective Jahn-Teller distortion (CJTD), which is the ground state of NaNiO2 (NNO), a topologically equivalent compound of LNO [23,27,28]. The discrepancy between the experimental and theoretical studies has been explained based on phases other than R m 3 [22,29–31], building possible orbital ordering patterns [8,21,30,32], defective structures [33–35], time and ordering averages of dynamic stabilization [29,36], and combining theoretical methods [16–20,22,37]; however, few such studies have provided sufficient explanations. The most plausible explanation for this discrepancy was provided by Petit et al. [33]. They established an extra-Ni-based model with a ~1 eV band gap by substituting 25% of Li+ with Ni2+ using the self-interactioncorrected local-spin-density method [33]. However, this band gap is extremely large compared with the experimental data of ~0.5 eV, and no LJTD has been reported [11–14]. Moreover, 25% of the Li sites occupied by Ni2+ appears far-fetched, as (Li1−xNix)NiO2 has been reported with an x of up to 0.14 [5,13,26,38–40]. In this study, we observed that when Ni/Li antisites are considered, the discrepancy between the experimental and theoretical results of R m 3 LNO can be solved. Some theoretical and experimental studies have confirmed the existence of Ni/Li antisites in NMC [41–44] and LNO [7,34], but the focus of these studies was on the removal of Ni/Li antisites and not on the JT effect [45,46]. The defect concentration of (Li1−xNix)(Ni1−xLix)O2 was determined as ~3% by Kanno et al. [13]. Based on hybrid density functionals of Heyd-Scuseria-Ernzerhof (HSE06) calculations, LNO with Ni/Li exhibited a band gap of 0.40 eV with LJTD, which agreed with the experimental results. Further discussion on the Kanno’s model and our computational details can be found in the Supplementary information (SI) Sect. A (Fig. S1) and B. We considered a trigonal structure (R m 3 ) comprising layers of NiO2 slabs, with edge-sharing octahedra, separated by a lithium layer (Fig. 1a). In the ideal structure, Ni ions occupy 3a sites and Li ions occupy 3b sites. Two possible defective LNO types were considered (Fig. 1a): a model with extra Ni (the previous model, (Li26Ni)3b(Ni27)3aO54 [5,7,39,40]) and a model with Ni/Li (the model proposed herein, (Li26Ni)3b(Ni26Li)3aO54). Some theoretical studies have focused on the P21/c phase as it exhibits the lowest energy of LNO among various phases [22,29–31]. However, it should be excluded owing to its inconsistency with experiments. P21/c is significantly different with R m 3 , as reported by prior experiments [5–9]. Besides, defects were not considered in these studies, which conflicted with the fact that ideal LNO has not yet been synthesized [5,10,13,26,47,48]. The calculated parameters of R m 3 LNO (a = b = 2.896 Å, c = 14.397 Å) are closer to the experimental data (a = b = 2.883 Å, c = 14.215 Å) [7,49]. Although the magnetism of LNO is affected by many factors, ferromagnetic (FM) is a common approximation [9]. Therefore, we considered LNO as FM for all calculations. Its lattice parameters match with the experimental data, but NiO6 octahedra in R m 3 LNO have identical Ni–O bond lengths, contradictory to the varying Ni–O bond lengths reported via experiments [8,24,26]. This discrepancy would be explained based on defective models. The density of states (DOS) of the ideal R m 3 LNO indicates that the empty spin-down eg band and half occupied spin-up eg band ensure the electronic configuration of t2geg, corresponding to Ni3+ (Fig. 1b). The spin-up eg band lies on the Fermi level and exhibits a conducting behavior, which is consistent with the previous theoretical studies and contrary to the experimental semiconductor behavior [11–14,16–20]. The extra-Ni model (Li26Ni)Ni27O54 also exhibits a conducting behavior. In the antisite model (Li26Ni)(Ni26Li)O54, the band gap of 0.40 eV is consistent with the experimental data (0.2 eV [13] and 0.5 eV [14]). Models with different Ni/Li concentrations also exhibited