Deciphering conductivity in PEDOT guided by machine learning: From solvent baths to charge paths

PEDOT:Tos is a promising conducting polymer for electronic and bioelectronic applications, yet its charge transport is affected by various factors and remains challenging to optimize. This study investigates the impact of solvent posttreatment on PEDOT:Tos thin films, exploring its influence on morphology and electrical conductivity. A combined experimental-theoretical approach is employed, integrating molecular dynamics, density functional theory, and transport calculations on one hand and conductivity, GIWAXS and XPS measurements on the other hand. Moreover, we developed a machine learning (ML) framework based on convolutional neural networks with the Coulomb matrix as the predictor, and transfer integrals for multiscale transport calculations as targets. Our results reveal that solvent-induced morphological changes strongly affect charge transport, with the ML model effectively reproducing observed conductivity. The developed ML model dramatically boosts the speed of mobility calculations, enabling the analysis of large-scale polymer films that were previously beyond computational reach.