Machine Learning Approaches for Determining Molecular Packing of Organic Semiconductors: Toward Accurate Crystal Structure Prediction

Organic semiconductors (OSCs) with π-electron skeletons (π-cores) have garnered significant attention. The development of innovative molecules with high carrier mobility necessitates strategic molecular design. One critical property affecting the carrier mobility of π-conjugated OSCs is the molecular arrangement, particularly the two-dimensional (2D) molecular packing of the π-cores, such as π-stacking, herringbone (HB) packing, and brickwork (BW) packing. These molecular packing structures, which are similar 2D packings, have not been theoretically predicted, leading chemists to design new OSC molecules based on empirical knowledge. Therefore, computational science and informatics are crucial for the strategic design of OSC molecules with unprecedented properties and functions. In this study, we introduce a machine learning method to determine with high accuracy whether an OSC forms the HB packing, a 2D molecular packing known to enhance carrier mobility. We also present a computational method to predict the crystal structure of an OSC from its chemical structure using molecular mechanics (MM) calculations and molecular dynamics (MD) simulations, coupled with our proposed machine learning model to classify the type of 2D molecular packing.