Finding the Right Bricks for Molecular Lego: A Data Mining Approach to Organic Semiconductor Design

Abstract

Improving charge carrier mobilities in organic semiconductors is a challenging task that has hitherto primarily been tackled by empirical structural tuning of promising core compounds. Knowledge-based methods can greatly accelerate such local exploration, while a systematic analysis of large chemical databases can point towards promising design strategies. Here, we demonstrate such data mining by clustering an in-house database of >64.000 organic molecular crystals for which two charge-transport descriptors, the electronic coupling and the reorganization energy, have been calculated from first principles. The clustering is performed according to the Bemis-Murcko scaffolds of the constituting molecules and according to the sidegroups with which these molecular backbones are functionalized. In both cases, we obtain statistically significant structure-property relationships with certain scaffolds (sidegroups) consistently leading to favorable charge-transport properties. Functionalizing promising scaffolds with favorable sidegroups results in engineered molecular crystals for which we indeed compute improved charge-transport properties.