The processing conditions during solvent-based fabrication of thin film organic electronics significantly determine the ensuing microstructure. The microstructure, in turn, is one of the key determinants of device performance. In recent years, one of the foci in organic electronics has been to identify processing conditions for enhanced performance. This has traditionally involved either trial-and-error exploration, or a parametric sweep of a large space of processing conditions, both of which are time and resource intensive. This is especially the case when the process → structure and structure → property simulators are computationally expensive to evaluate.

In this work, we integrate an adaptive-sampling based, gradient-free, Bayesian optimization routine with a phase-field morphology evolution framework that models solvent-based fabrication of thin film polymer blends (process → structure simulator) and a graph-based morphology characterization framework that evaluates the photovoltaic performance of a given morphology (structure → property simulator). The Bayesian optimization routine adaptively adjusts the processing parameters to rapidly identify optimal processing configurations, thus reducing the computational effort in process → structure → property explorations. This serves as a modular, parallel ‘wrapper’ framework that facilitates swapping-in other process simulators and device simulators for general process → structure → property optimization. We showcase this framework by identifying two processing parameters, the solvent evaporation rate and the substrate patterning wavelength, in a model system that results in a device with enhanced photovoltaic performance evaluated as the short-circuit current of the device. The methodology presented here provides a modular, scalable and extensible approach towards the rational design of tailored microstructures with enhanced functionalities.