A rapid UV/Vis assisted designing of benzodithiophene based polymers by machine learning to predict their light absorption for photovoltaics

As global energy demands escalate, developing high-performance photovoltaic (PV) materials through accelerated design methodologies is imperative. A machine learning (ML) assisted predictive models are used to accelerate the design of benzodithiophene (BDT)-based polymers for their PV applications. The current approach leverages a curated dataset of 191 compounds with experimental UV–Vis spectra, mapped to molecular electronic descriptors via RDKit. Random Forest modeling yields a predictive framework (R² = 0.98) for predicting their maximum absorption (λ_max). After it, their 5000 new designs as novel polymers, identifying top performers with Synthetic Accessibility Likelihood Index scores up to 57, ensuring synthesis feasibility have also been designed. Feature importance analysis highlights MaxPartialCharge and Aromatic rings as crucial descriptors. The designed materials exhibit optimal energy gaps (1.35–2.0 eV), paving the way for efficient PV devices. The computed UV–Vis spectra of best predicted polymers are studied with their λ_max range of 487–987 nm showing a significant redshift behavior. The designed polymers presents and good potential towards and they can be good candidates for organic solar cell applications.