(304b) Characterizing and Refining Coarse-Grained Models of Aromatic Macromolecules for Organic Photovoltaics Using Experimental Scattering Data

Engineering the morphology of organic photovoltaic active layers is crucial for the performance of a power-producing device: The shape and surface area of the interface between electron donors and electron acceptors influences the generation of charges, and the structure of the regions between the interface and the electrodes influences the transport of charges out of the active layer.  In order to engineer this morphology through the choice of active layer components and the conditions under which they are processed, we must understand how the chemical features of molecules influences their thermodynamically-driven self-assembly.  In practice, the conjugated polymers and aromatic macromolecules used in organic photovoltaics self-assemble partially-disordered morphologies that have short-range ordering on the order of 4 Angstroms, and large-scale structural features on the order of 20 Angstroms.  Computer simulations allow detailed investigations into how changing certain chemical features or processing conditions can change morphology, but simulating the volumes needed to observe the micro-scale ordering of 20-Angstrom features demands massively parallel computing power and/or coarse-grained models.

In this work we evaluate coarse-grained models of benzodithiophene-based copolymers (BDT) mixed with phenyl-C₆₁-butyric acid methy ester (PCBM).  We perform molecular dynamics simulations on graphics processing units (GPUs) that permit 10-nm features to be observed over 100-ns time scales.  We compare the morphologies predicted from our simulations against experimentally characterized films using grazing-incidence wide angle X-Ray scattering (GIWAXS).  We find that coarse-grained models of BDT copolymers and PCBM permit qualitative prediction of experimentally-observed films in a few days of simulation time.  Further, we find that modeling the aromatic rings of these molecules as rigid bodies provides not only a performance boost, but also more accurate morphologies than flexible rings.  Finally, we identify discrepancies between the simulated and experimental morphologies and discuss methods for iteratively refining the coarse-grained models using this information.