Sustainable design of organic solar cells utilized machine and deep learning

Abstract

Abstract In this work, an Organic Solar Cell (OSC) with a structure of ITO/PEDOT: PSS/PBDB-T: IT-M/PFN-Br/Al is extensively simulated and optimized. The impact of layer thicknesses and materials on device performance is simulated using a one-dimensional solar cell simulator (SCAPS-1D). The simulation model is first validated using experimental data, and it shows a high degree of alignment. Among the various Electron Transport Layers (ETLs) that are investigated, PFN-Br has the highest Power Conversion Efficiency ( PCE ) of 12.04%. The PFN-Br thickness is shown to be most effective at 5 nm. A simulated PCE of 19.50% results from the active layer reaching its optimum efficiency at 300 nm. PEDOT: PSS is the most effective Hole Transport Layer (HTL) with reliable performance at thicknesses ranging from 30 to 100 nm. Due to optical interference, the short-circuit current density ( $$\:{J}_{sc}$$ ) slightly increases. Additionally, based on structural parameters, PCE and open-circuit voltage ( $$\:{V}_{oc}$$ ) are predicted by using Artificial Intelligence (AI) models, including Convolutional Neural Networks (CNN) and Support Vector Regression (SVR). CNN achieves the highest prediction accuracy in modelling PCE , demonstrating its ability to model nonuniform photovoltaic behavior. This combined approach using both detailed simulations and AI‑based prediction doesn’t just make organic solar cells more efficient. It also supports the larger goal of making clean energy more accessible and sustainable. By fine‑tuning device designs, cutting down material waste, and simplifying fabrication, this work can help move the world closer to achieving key Sustainable Development Goals (SDG 7, 9, 12, and 13). In this way, advances in solar technology can play a meaningful role in addressing climate and energy challenges.