Machine Learning Predictions of Electronic Couplings for Charge Transport Calculations - Code
| Main Authors: | Miller, Evan, Jones, Matthew, Jankowski, Eric |
|---|---|
| Format: | info software |
| Terbitan: |
, 2019
|
| Subjects: | |
| Online Access: |
https://zenodo.org/record/2635495 |
Daftar Isi:
- Developing the next generation of inexpensive solution processed organic photovoltaics requires understanding how self-assembled morphology dictates device charge transport. Kinetic Monte Carlo simulations are able to link individual molecules to bulk morphology charge transport properties for systems self-assembled with molecular dynamics. However, this linking comes at the cost of computing electronic couplings for every neighboring pair of molecules using expensive quantum chemical calculations (QCC). Here we present the code for training, validating and implementing random forest predicted couplings that offer a speed-up of 2-3 orders of magnitude.