The use of machine learning, density functional theory, and molecular dynamics simulations for the designing and screening of efficient small molecule donors for organic solar cells

Abstract

Indeed, a proper understanding of materials is necessary to get the full benefit from them. For this purpose, multiscale computational modeling is the ultimate need. For machine learning analysis, data is collected from the literature. Machine learning analysis is performed using molecular descriptors as independent parameters and power conversion efficiency (PCE) as dependent property. Various machine learning models are tried. The support vector machine (SVM) model has outperformed others. New donor materials that are small molecules are designed using both well-known and new building blocks. Their PCE is predicted using a SVM model. The top 10 small molecule donors are further studied using density functional theory calculations. Their electronic behavior is studied. Reorganization energy, exciton binding energy and transfer integral are also calculated. Finally, the best three small molecule donors are selected for molecular dynamics simulations. Molecular packing and mixing of active layer materials is studied using radial distribution function. Our proposed framework has the ability to design potential donor materials in short time with marginal computational cost.