基于第一性原理与机器学习方法研究双钙钛矿Li2CuBiX6

Abstract The development of efficient and stable lead-free materials is essential for advancing next-generation photovoltaic technologies. In this study, we investigate Li 2 CuBiX 6 (X = Br, I) double perovskites as promising absorber materials, using first-principles calculations and machine learning techniques. Density functional theory (DFT) results show indirect band gaps of 1.7 eV (Br) and 1.3 eV (I), suitable for solar energy conversion. Key optical properties, including absorption coefficient, reflectivity, refractive index and dielectric function, confirm their strong ability to capture light. A solar cell architecture FTO/ETL/Li 2 CuBiX 6 /HTL/Mo was modeled in SCAPS-1D, evaluating various electron and hole transport layers. SnS 2 and Cu 2 O were identified as the best ETL and HTL, respectively, producing high energy conversion efficiencies of 27.24 % (Li 2 CuBiBr 6 ) and 31.80 % (Li 2 CuBiI 6 ). We also analyzed the effects of interfacial defects, doping concentration, absorber thickness and temperature on device performance. To predict efficiency trends and optimize configurations, we applied machine learning models (XGBoost, Random Forest, SVR). XGBoost achieved the highest accuracy, with R 2 = 99.87 % and a low RMSE. This work highlights the potential of Li 2 CuBiX 6 as an efficient, lead-free solar absorber and demonstrates the value of combining first-principles simulations with machine learning for photovoltaic design.