Berlin 2024 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
CPP: Fachverband Chemische Physik und Polymerphysik
CPP 34: Emerging Topics in Chemical and Polymer Physics, New Instruments and Methods I
CPP 34.7: Talk
Thursday, March 21, 2024, 11:30–11:45, H 0106
Using Neural Network Potentials to Predict Thermal Isomerization Barriers for Spiropyran Derivatives — •Robert Strothmann1, Johannes T Margraf1,2, and Karsten Reuter1 — 1Fritz-Haber-Institut der MPG, Berlin — 2Universität Bayreuth, Bayreuth
First principles methods like density-functional theory can be used to study reaction barriers and give insights into the influence of chemical modifications on barrier heights and mechanisms. However, their large computational cost hinders their usage in high-throughput settings, which prevents large scale studies including thousands of structures. Surrogate models with reduced costs are therefore of great interest. In this context, neural network potentials (NNPs) have attracted much attention, as their uncertainty can be systematically decreased by increasing the amount and quality of the training data.
In this talk, we will highlight the usage of NNPs to assist in the prediction of thermal isomerization barriers of spiropyran photoswitches. This barrier governs the half-life of the thermal back-reaction, which is one of the key properties in photoswitch design. By utilizing transferability, NNPs trained on only a few spiropyran molecules and fine tuned in active-learning cycles allow the prediction of isomerization barriers for thousands of chemically modified spiropyrans. This not only gives a ranking of suitable photoswitches for a specific application, but also enables a more systematic study on how to tune the thermal half-life via chemical modification.
Keywords: Machine learning; Neural network potentials; Photoswitches