Multiscale Modeling of Organic-Inorganic Semiconductor Materials: Opportunities and Challenges

J. Stevenson, J. Saathoff, P. Clancy
Cornell University,
United States

Keywords: multiscale modeling, reactive force fields, organic semiconductors


Many of the most interesting new materials that show promise for solar photovoltaics, involve atomic species, like lead and selenium, that are non-traditional from the point of view of computational modeling. This poses a problem in several ways: First, many traditional "force fields" do not handle these atom types, especially in a mixed organic-inorganic system. Second, since many of these new materials, e.g., hybrid organic-inorganic perovskites (HOIPs), are solution-processed and involve complexation and/or reactions that current reactive force fields are equally exposed in terms of difficulty of parameterization and availability. Thirdly, the number of moieties in the system is typically large enough that the complexity of the system makes it overwhelming to undertake given the large combinatorial burden to describe all possible pairs of interacting entities. What is needed is a fresh approach to model such systems. We suggest a possible solution to this through a new reactive force field that is far simpler than existing approaches, is accurate enough to reproduce ab initio data, and is relatively easy to parameterize. We show a proof-of-concept for this Simple Reactive Molecular Force Field (SMRFF) by giving results for PbS nanoparticles that rival the ab initio database we created. We show how this approach can be extended to perovskite systems. As examples, we show how a multiscale approach, incorporating ab initio calculations and Molecular Dynamics, can provide unique insight into the complexation that occurs from precursor solutions that give rise to HOIP thin films. As a second test case, we will show how solution processing of lead chalcogenide reactive systems can be optimized using a multiscale computational modeling approach. Finally, we illustrate an example of the advantage of using a Bayesian optimization search to speed up the identification of organic semiconductor polymorphs. The choice of solvent also affected the experimental results and we will show how simple models can capture at least some of these effects and explain the mechanistic underpinnings of this effect.