Machine learning assisted designing of low bandgap indacenodithiophene (IDT)-based organic semi-conductors

Abstract

Development of the suitable organic semiconductor materials (OSMs) that can exhibit lower band gap values for organic-electronic devices (OEDs) applications is tedious and laborious task. Data-driven methodologies, particularly machine learning (ML) approach, have been explored as an alternative effective and rapid approach for getting the acquired results. Here, the ML based approach has utilized for generating the chemical database of OSMs having low band gap values. At first, 40 ML models were trained on the acquired dataset. The Extra Trees Regressor (ETR) model was statistically selected for the predictive analysis. The RDKit tool was utilized to generate the molecular descriptors associated with OSMs and the descriptors having high feature relevance to band gap was selected for further use. The new OSMs database having 10,000 data points was generated. t-Distributed Stochastic Neighbor Embedding (TSNE) plot were used for reducing the dimensions of the generated chemical database. The top 30 contenders were selected from the chemical database on the basis of the lower values of the band gap. The study presented here utilize the ML modeling as an effective tool for identifying the optimal OSMs exhibiting the band gap value of 1.42 eV (even lower than the previously documented band gap for OSMs) for number of OEDs applications.