High-throughput screening of KBX3 halide perovskites: Structural, optoelectronic, and thermoelectric properties via DFT and machine learning with uncertainty quantification
Journal of Physics and Chemistry of Solids, cilt.217, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 217
- Basım Tarihi: 2026
- Doi Numarası: 10.1016/j.jpcs.2026.113798
- Dergi Adı: Journal of Physics and Chemistry of Solids
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Chemical Abstracts Core, Chimica, Compendex, INSPEC
- Anahtar Kelimeler: Bandgap engineering, DFT, Halide perovskites, High-throughput screening, Machine learning, Thermoelectrics
- Gazi Üniversitesi Adresli: Evet
Özet
This study combines density functional theory (DFT) and machine learning (ML) to investigate KBX3 halide perovskites for energy applications. First, explicit DFT calculations using the FP-LAPW method with PBEsol and TB-mBJ functionals were performed for four representative compounds, namely KGeCl3, KGeBr3, KSnCl3, and KSnBr3, which were found to be direct-gap semiconductors with favorable optical absorption and thermoelectric performance. To accelerate discovery beyond these four compounds, machine-learning screening was performed across 20 KBX3 base compositions using a synthetic descriptor-augmented dataset of 100 samples. Five-fold cross-validation yielded R2 = 0.852 ± 0.055 for bandgap prediction and R2 = 0.893 ± 0.120 for ZT prediction, while independent test-set performance gave R2 = 0.615 with RMSE = 0.270 eV for bandgap and R2 = 0.832 with RMSE = 0.104 for ZT. The predicted bandgaps span approximately 0.45–1.77 eV, while the highest predicted ZT values at 600 K are obtained for Pb-containing compositions, especially KPbI3 and KPbBr3. Therefore, the ML results are interpreted as screening-level predictions that identify promising compositional regions for further DFT validation.