Fig. 5: ML-enabled search of alloyed perovskite systems.
From: Machine learning-enabled chemical space exploration of all-inorganic perovskites for photovoltaics
a CGCNN-predicted \(\Delta {H}_{{\rm{decomp}}}-T\Delta {S}_{{\rm{mix}}}\) of a CsGexSn1-xBr3 and b CsGexHgySn1-x-yCl3 systems in comparison with the training and DFT data (80 atoms). The variance of CGCNN data is represented as a line at each composition. In b, only the lowest \(\Delta {H}_{{\rm{decomp}}}-T\Delta {S}_{{\rm{mix}}}\) value is plotted at each composition for clearance. In b, the validation and CGCNN data at the same composition (Ge: Hg: Sn = 0.5625: 0.25: 0.1875) are −107.60 and −103.79 meV atom−1, respectively.
