Machine learning intermolecular electronic couplings: Opportunities and challenges

Intermolecular electronic couplings (IECs) play a critical role in theories of charge and excitation transfer. However, the considerable computational cost of evaluating IECs limits the exploration of large molecular systems and long timescales key to numerous molecular and materials applications. Recent advances in machine learning (ML) offer a means of accelerating IEC predictions, enabling previously intractable calculations for large disordered systems and their optoelectronic properties. In this pedagogical review, we summarize recent progress in applying ML to predict IECs and leverage a case study of IEC prediction applied to a canonical molecular dopant and organic semiconductor system to illustrate conformational data sampling, feature engineering, and ML algorithms. For each step, we discuss the existing challenges and outline future research directions to improve the accuracy, interpretability, and transferability of ML models for IEC predictions.