The ligand-assisted reprecipitation (LARP)-based synthetic approach has gained attention as a promising method for scalable synthesis of perovskite nanocrystals (PNCs) with outstanding optoelectronic functionalities. However, such distinct synthetic features of the LARP method involve an intrinsic limitation in realizing red-color emissions from I-rich compositions. Herein, we explore the LARP synthesis space of CsPb(BrxI1-x)3 PNCs via a high-throughput robotic synthesis platform integrating machine learning (ML) algorithms, not only allowing for understanding the role of each chemical variable from the multidimensional synthesis space but also refining the bespoke synthesis landscape of PNCs with target functionalities. It is found that ligand ratios as well as the selection of antisolvents dynamically contribute to synthesizing I-rich CsPbX3 PNCs, where their delicate and dedicated adjustments are required depending on the Br-to-I ratios. Furthermore, a disparity between the latent feature in ML-refined synthesis space and the manifested functionality space is identified, where the colloidal nature in the precursor state is found to colligate the bespoke synthesizability and functionality control of the LARP-PNCs. This data-driven approach enables the rational synthetic designs of CsPbX3 PNCs, as well as the fundamental relationship between the synthesis and functionality space.
Kim et al. (2026) studied this question.