Regensburg 2022 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
CPP: Fachverband Chemische Physik und Polymerphysik
CPP 40: Organic Electronics and Photovoltaics 3
CPP 40.2: Talk
Thursday, September 8, 2022, 15:30–15:45, H38
Reorganization energies of flexible organic molecules as a challenging target for machine learning enhanced virtual screening — •Ke Chen1,2, Christian Kunkel1,2, Karsten Reuter1,2, and Johannes T. Margraf1,2 — 1Technische Universität München, Garching, Germany — 2Fritz-Haber-Institut der Max-Planck-Gesellschaft, Berlin, Germany
The molecular reorganization energy λ strongly influences the charge carrier mobility of organic semiconductors and is therefore an important target for molecular design. Machine learning (ML) has the potential to accelerate this process by providing accurate surrogate models in design space. Unfortunately, λ poses a significant challenge for ML-models as it simultaneously depends on the neutral and ionized potential energy surfaces of the molecule.
In this contribution, we address the questions of how ML models for λ can be improved and what their benefit is in high-throughput virtual screening (HTVS). We find that, while improved predictive accuracy with respect to a semi-empirical baseline model is achieved, the benefits for molecular discovery are actually somewhat marginal. In particular, ML-enhanced HTVS is more effective in identifying promising candidates but leads to a less diverse sample set.