Enhancing predictive modeling of photovoltaic materials’ solar power conversion efficiency using explainable AI

Abstract

We present a study on Explainable AI-based prediction of power conversion efficiency (PCE) of organic solar cells, conducted on a dataset of 566 small-molecule organic solar cell materials samples with varying donor and acceptor species combinations. This research uncovers an interesting phenomenon, the first of its kind to be reported, of PCE quantization, where the PCE values increase in steps with the increase in feature values. Our findings have significant implications for the development of efficient organic solar cells, as they provide a better understanding of the factors that influence PCE, and highlight the feature value ranges for which more efficient PCE would be achieved. Our study demonstrates the power of XAI techniques in uncovering hidden patterns in scientific datasets and highlights the importance of interdisciplinary research in the field of materials science.