Executive Summary : | This proposal aims to develop a machine learning model to understand and predict energy loss in polymer:non-fullerene based organic solar cells, demonstrating 20% device efficiency. The model will investigate the role of device architecture, molecular design, and intrinsic material properties such as carrier mobility, HOMO-LUMO offsets, morphology, D:A ratio, exciton dynamics, and interface layers in photovoltaic effect formation. The project will be divided into two work packages (WPs). WP1 involves developing, training, and validating the ML model by creating a big data set for non-fullerene based organic solar cells. Detailed statistical analysis will be carried out to understand the effects of different parameters on device performances and their interdependencies. Different machine learning models will be developed and tested to accurately predict energy loss for these novel material systems. WP2 focuses on developing 20% efficient NFA based devices by pursuing three major approaches: Donor:non-fullerene acceptor bulk homojunction, Donor/non-fullerene acceptor pseudo-bilayer, and Donor:acceptor:non-fullerene acceptor ternary bulk homojunction. Devices will be rationally optimized by controlling material properties, active layer film morphology, and device engineering. Various techniques will be employed for detailed characterization of materials and devices, including Uv-Vis absorption studies, XPS, PL, electrochemistry, XRD, AFM, FESEM, and TEM. Device photovoltaic properties will be investigated using I-V, and C-V measurements of the cells. These properties will be used as inputs to the ML algorithm for further improvement. By combining these detailed studies, a deep understanding of the structure property relation between film morphology and photophysical properties will enable fabrication of highly efficient (~20%) long-term stable NFA based photovoltaic devices. |