Toward the Rational Design of Mid‐Infrared Nonlinear Optical Materials with Targeted Properties via a Multi‐Level Data‐Driven Approach

Abstract Design and exploratory synthesis of new mid‐infrared (mid‐IR) nonlinear optical (NLO) materials are urgently needed for modern laser science and technology because the widely used IR NLO crystals still suffer from their inextricable drawbacks. Herein, a multi‐level data‐driven approach to realize fast and efficient structure prediction for the exploration of promising mid‐IR NLO materials is proposed. Techniques based on machine learning, crystal structure prediction, high‐throughput calculation and screening, database building, and experimental verification are tightly combined for creating pathways from chemical compositions, crystal structures to rational synthesis. Through this data‐driven approach, not only are all known structures successfully predicted but also five thermodynamically stable and 50 metastable new selenides in A I B III Se 2 systems ( A I = Li, Na, K, Rb, and Cs; B III = Al and Ga) are found, among which eight outstanding compounds with wide bandgaps ( > 2.70 eV) and large SHG responses ( > 10 pm V −1 ) are suggested. Moreover, the predicted compounds I 2 d ‐LiGaSe 2 and I 4/ mcm ‐KAlSe 2 are successfully obtained experimentally. In particular, LiGaSe 2 exhibits a robust SHG response ( ≈ 2 × AGS) and long IR absorption edge that can cover two atmospheric windows (3–5, 8–12 µ m). Simultaneously, this new research paradigm is also applicative for discovering new materials in other fields.