Machine learning assisted designing of small molecule acceptors with multiple terminal electron-deficient groups and performance prediction for next-generation photovoltaics

Small molecule acceptors (SMAs) are extensively used in organic photovoltaics (OPVs) because of their capacity to efficiently receive electrons and enable charge separation. The performance of organic photovoltaic (OPV) devices can be enhanced by designing SMAs with several terminal electron-deficient groups, which will increase their electron-accepting capacity. However, it takes a lot of effort and time to rationally build such complicated molecules. In this work, a strategy for designing novel SMAs with numerous terminal electron-deficient groups is introduced. In order to predict the performance of newly designed SMAs, ML models have been employed that have been trained. Data on the power conversion efficiency (PCE) of 124 solar cell devices is collected. PCE's maximum value is taken into account. To compute molecular descriptors, utilize RDkit. A library of about 200 descriptors has been generated. The chemical similarity of the designed SMAs is studied using cluster plot and heatmap. For this purpose, chemical fingerprints are used. Using this method, thirty SMAs with highest PCE (%) ranges from 9.99 to 9.65 % have been selected. Correlation between power conversion efficiency PCE (%) and top descriptors in our study of machine learning. • Small molecule acceptors (SMAs) have been studied by using machine learning. • Electron-deficient groups have increased the electron-accepting capacity. • Thirty finest SMAs with highest PCE (%) ranges from 9.99 to 9.65 % have been selected.