Opportunities and challenges for machine learning to select combination of donor and acceptor materials for efficient organic solar cells

Organic solar cells (OSCs) have witnessed significant improvement in power conversion efficiency (PCE) in the last decade. The structural flexibility of organic semiconductors provides vast search space for potential candidates of OSCs, but discovering new materials from search space with traditional approaches such as DFT is computationally expensive and time-consuming. Machine learning (ML) is extensively used in OSCs to accelerate productivity and materials discovery. ML is gaining more attention due to the availability of large datasets, improved algorithms, and exponentially growing computational power. In this review, current progress, opportunity, and challenges for ML in OSCs have been identified. Given the rapid advances in this field, impactful techniques that have been useful in extracting meaningful insights are discussed. Finally, we elaborate upon the bottlenecks of the ML workflow with respect to data size, model interpretability, and extrapolation.

Automated summary

The development and challenges of machine learning in organic solar cells are discussed. The flexibility of organic semiconductors offers new possibilities for discovering battery materials. Some progress has been made, and impactful techniques for extracting useful information have been identified. However, there are still limitations in the machine learning workflow.