Feature Engineering for Microstructure–Property Mapping in Organic Photovoltaics

Abstract

Linking the highly complex morphology of organic photovoltaic (OPV) thin films to their charge transport properties is critical for achieving high performance material systems that facilitate cost-efficient energy harvesting. In this paper, the current Materials Knowledge Systems (MKS) framework was extended so that it was able to establish reduced-order high-fidelity structure–property linkages for OPV films. Specifically, the following extensions were needed: (i) the proper application of digital image processing algorithms to identify the salient local material states in OPV microstructures controlling the charge transport phenomenon, (ii) computationally efficient feature engineering that not only utilized 2-point spatial correlations and principal component analysis, but also two new distance-based metrics, and (iii) the successful application of a localized version of the Gaussian process (laGP) together with an active learning Cohn (ALC) for building the desired surrogate models linking the OPV microstructures to their short-circuit currents. It is demonstrated that the extended MKS framework can produce high-fidelity structure–property linkages for OPV films.