A holistic data-driven approach to synthesis predictions of colloidal nanocrystal shapes

Abstract The ability to precisely design colloidal nanocrystals (NCs) has far-reaching implications in optoelectronics, catalysis, biomedicine, and beyond. Achieving such control is generally based on a trials-and-errors approach. Data-driven synthesis holds the promise to advance both discovery and mechanistic knowledge. Herein, we contribute to advancing the current state of the art in the chemical synthesis of colloidal NCs by proposing a machine-learning toolbox which operates in a low data regime, yet comprehensive of the most typical parameters relevant for colloidal NC synthesis. The developed toolbox predicts the NC shape given the reaction conditions and proposes reaction conditions given a target NC shape, using Cu NCs as the model system. By classifying NC shapes on a continuous energy scale, we synthesize an unreported shape, which are Cu rhombic dodecahedra. This holistic approach integrates data-driven and computational tools with materials chemistry. Such development is promising to greatly accelerate materials discovery and mechanistic understanding, thus advancing the field of tailored materials with atomic scale precision tunability.