Abstract
The two-dimensionality (2D) of charge transport significantly affects charge carrier mobility in organic semiconductors, making it a key target for materials discovery and design. Traditional quantum-chemical methods for calculating 2D are resource-intensive, especially for large-scale screening, as they require computing charge transfer integrals for all unique pairs of interacting molecules. We explore the potential of machine learning models to predict whether this parameter will fall within a desirable range without performing any quantum-chemical calculations. Using a large database of molecular semiconductors with known 2D values, we evaluate various machine-learning models using chemical and geometrical descriptors. Our findings demonstrate that the LightGBM outperforms others, achieving 95% accuracy in predictions. These results are expected to facilitate the systematic identification of high-mobility molecular semiconductors.
Original language | English |
---|---|
Pages (from-to) | 3676-3679 |
Number of pages | 4 |
Journal | Chemical Communications |
Volume | 61 |
Issue number | 18 |
Early online date | 31 Jan 2025 |
DOIs | |
Publication status | E-pub ahead of print - 31 Jan 2025 |
Keywords
- high-mobility molecular semiconductors
- machine learning
- charge transport
- carrier mobility