Machine Learning-Based Prediction of Optical Activity in Chiral Materials

Chiral systems span a wide range of materials from single molecules to hierarchically assembled nanoparticles and metamaterials. They possess many unique physicochemical features, including circular dichroism, circularly polarized photoluminescence, nonlinear optics, ferroelectricity, and spintronics. However, the origin of their chirality and related optical activity have not been unveiled comprehensively. In this work, we train machine learning algorithms with data sets of chiral ligand-protected metal nano clusters and chiral hybrid organic–inorganic perovskites to characterize and predict the optical activity. Our datasets encompass critical parameters such as crystal structure data, ligand properties, UV-visible absorption spectra, and circular dichroism signatures. The preliminary results reveal a strong consistency within the dataset, demonstrating a notable level of classification among the elements which promise the prediction the optical activity. Our results can facilitate the design and evaluation of chiral materials with intriguing electronic, magnetic, and optical effects.