Interpretable Machine Learning for Solvent‐Dependent Carrier Mobility in Solution‐Processed Organic Thin Films

Understanding how solvent properties influence charge transport in solution‐processed organic films is essential for the rational optimization of processing conditions. In this study, we employed a transfer learning‐based multiscale simulation framework to predict the carrier mobility of quadruple thiophene (QT) films prepared using thirteen different solvents. By combining iterative leave‐one‐out cross‐validation with interpretable machine learning (ML), we developed a quantitative model capable of predicting carrier mobility based on solvent properties, achieving an R 2 value of up to 0.80 in independent tests. Shapley value analysis revealed that the Hansen solubility parameters δD and δH, as well as the solvent molar volume, are the key factors determining film performance. Specifically, strong dispersion interactions (δD), moderate hydrogen bonding (δH), that closely match and are compatible with the QT, and a small molar volume were found to enhance structural relaxation and the formation of percolated transport networks. These findings provide a systematic approach for solvent selection in the fabrication of high‐mobility organic semiconductors and demonstrate the potential of ML in constructing reliable physical models from limited data.