(508g) Deep Learning and Atomistic Simulations in High-Throughput Material Discovery | AIChE

(508g) Deep Learning and Atomistic Simulations in High-Throughput Material Discovery


We applied molecular deep learning cheminformatics technology in virtual screening of thousands of candidates for Organic Photovoltaic (OPV) material and processing conditions. We built a predictive model using virtual screening. We have first collected and combined experimental datasets from literature sources. From these datasets, we extracted chemical species and processing conditions and used them to seed exploration of the parameter space. We then identified the most important chemical and experimental condition descriptors to lead fundamental research toward the engineering of molecular-scale features that contribute to higher Power Conversion Efficiencies (PCE) in OPVs. By examining the feature space of the model, we then determined which inputs lead to the largest changes in performance metrics. This feature space might include the chemical identities of the active molecules (i.e. donor and acceptor), processing method (e.g. spin-coating, inkjet printing, roll-to-roll printing), and processing conditions (e.g. solvent identity, annealing temperature and time, additive content). We also generated thousands of hypothetical acceptors and donors and screened them using our model. We performed electronic structure calculations on these hypothetical materials.

In another effort, we combined molecular dynamic simulation (MD), density functional theory (DFT) and deep learning to map the interaction map and bandgap of the solar cell devices. Acceptors and donors interactions at heterojunction interface modulate different bandgaps. The bandgap highly depends on the relative conformation and relative orientation of acceptors and donors. We performed MD simulations to generate thermodynamically stable donor-acceptor couples and computed the bandgaps of each couple using DFT. We used the molecular snapshots as an input and the calculated interaction energy and bandgap as descriptor in our deep learning model. Our model predicts the bandgap and the interaction energy with high accuracy.