Designing active layer of organic solar cells using multi-fidelity molecular simulations and spectral density function

Molecular dynamics simulations have shown substantial promise in the design of organic photovoltaic cells (OPVC). Despite their potential, the utility of molecular dynamics simulations when designing an OPVC is often limited due to their considerable computational cost and their limited prediction accuracy. To address these challenges, we introduce a three-step multi-fidelity design framework that enables a designer to efficiently explore the space of admissible processing conditions, using coarse-grained molecular dynamics (CGMD) simulations, to identify the optimal OPVC design. Using a novel spectral density based approach to reconstruct microstructures of variable size, the framework is able to sequentially search for the globally optimal microstructure using a low-fidelity CGMD simulation with a smaller window size, followed by the optimization of the processing conditions using the high-fidelity simulation. The division in two steps and two fidelities enables the optimization of CGMD simulations at previously intractable lengths and timescales. We validate our results by demonstrating that the CGMD model predictions are consistent with physical experiments reported in the literature and corroborate that the computational complexity is reduced by one order of magnitude.