Ionic Transport in Doped Solid Electrolytes by Means of DFT Modeling and ML Approaches: A Case Study of Ti-Doped KFeO₂

We present a comprehensive study on the influence of Ti doping on K⁺ migration in the K₁–ₓFe₁–ₓTiₓO₂ solid electrolyte. A novel approach is proposed which is based on modeling of configurational spaces (CSs) and full sets of inequivalent migration pathways by means of density functional theory (DFT) calculations and machine learning (ML) techniques. A 2 × 1 × 1 supercell (32 formula units) of a low-temperature polymorph of the KFeO₂ compound with space group symmetry Pbca was used. For the three lowest Ti contents (x = 0.03, 0.06, and 0.09), all symmetrically inequivalent configurations of atomic arrangements (CSs) and K⁺ migration pathways (total numbers: 128, 59520, and 8630400) were generated. With the DFT-derived energetics of K⁺ migration at the lowest doping level (x = 0.03), the ML models were trained to predict ionic transport properties by using geometrical descriptors for the pathway-dopant arrangement. The trained ML models were then used to evaluate the K⁺ migration properties for pathways at higher doping levels. The computational results obtained are in good agreement with the results of a previous experimental study of the title compound. This demonstrates the applicability of the proposed approach for modeling and predicting effects of doping in crystalline solids, such as solid electrolytes and intercalation cathodes. Brief recommendations are given on the application of the proposed combined approach.