Machine‐Learning Aided First‐Principles Prediction of Earth‐Abundant Pnictogen Chalcohalide Solid Solutions for Solar‐Cell Devices

Abstract Discovering novel families of materials composed of earth‐abundant elements and characterized by non‐toxicity, high thermodynamic stability, and simple low‐temperature synthesis processes, is paramount for the advancement of urgently needed energy storage and conversion technologies. Pnictogen chalcohalides, represented by the general formula ABC ( A = Bi, Sb; B = S, Se; C = I, Br), emerge as a promising class of energy materials particularly well‐suited for photovoltaic applications. However, the compositional landscape of Bi _x Sb _{1 − x} S _y Se _{1 − y} I _z Br _{1 − z} is vast and remains largely unexplored, with traditional experimental and theoretical exploration techniques facing limitations in covering the entire solid‐solution range due to their labor‐intensive and time‐consuming nature. Here, an integrated bottom‐up approach that combines first‐principles calculations, machine learning models, experiments, and device optimizations is introduced to provide a comprehensive fundamental understanding of pnictogen chalcohalides with arbitrary composition and to expedite the design of high‐performance multi‐junction solar cells. The synergistic investigations unveil a broad and continuous spectrum of bandgaps and optical absorption coefficients ranging from 1.2 to 2.1 eV and from 2.5 · 10 ⁵ to 6.6 · 10 ⁵ cm ⁻¹ , respectively, across a wide variety of thermodynamically stable compounds. Additionally, a tandem BiSBr–BiSeI device is identified as an optimal multi‐junction solar cell, exhibiting a maximum short‐circuit current density of 18.65 mA cm ⁻² under intensity‐matching conditions. The introduced bottom‐up materials design approach may facilitate an unprecedented and rapid translation of basic knowledge into the most demanded solar cell applications.