Search of electron-rich and electron-deficient building blocks through data mining and library generation for the designing of polymers for organic solar cells

Typical designing of the organic molecules for numerous chemical, biochemical, and electrochemical applications requires extensive human workforce, conceptual intuition, far-reaching academic literature survey, and laborious experimentation. Recently, data-driven approaches, peculiarly machine learning (ML) analysis, have emerged as an alternative methodology to these traditional designing methods. These ML approaches provide robust and quicker way to address the serious restrictions of unavailability of work space, painstaking experimentation, and reduction in the number of experimental trials. In present study, two reference compounds, 5,6difluoro-2-methylbenzo [d] thiazole and benzo [1,2-b: 3,4-b': 6,5-b] trithiophene were selected as standard reference compounds. The ChemSpace database was mined for screening the potential effective building blocks and the new libraries of the organic molecules (having high similarity correlation with the reference compounds) were generated for designing the organic solar cell (OSC) devices. The ML analysis was also performed to predict the reorganization energy. The newly designed molecules were further evaluated with respect to their synthetic viability and electronic distribution characteristics to computationally excess the applicability of these novel organic molecules for the OSCs devices. The work presented in this study provides an alternative methodology for the selection of novel organic molecules for developing high-performance OSCs devices.

Automated summary

This study presents an alternative methodology using machine learning for the selection of novel organic molecules to develop high-performance organic solar cell devices.