goto contents

Thomas Gerhard Heinrich Stielow

Analyzing and controlling large nanosystems with physics-trained neural networks

Universität Rostock, 2022

Abstract: This thesis investigates the possible application of neural networks in accelerating the evaluation of physical experiments while minimizing the required simulation effort. Neural networks are capable of inferring universal reconstruction rules for reconstructing silver nanoclusters from single wide-angle scattering patterns from a small set of simulated data and when trained directly on scattering theory reaching unmatched accuracy. A dynamic excitation for giant dipole states of Rydberg excitons in cuprous oxide is derived through deep reinforcement learning interacting and simulation data.

doctoral thesis   free access    


OPACGVKDataCite Commons


all rights reserved

This work may only be used under the terms of the German Copyright Law (Urheberrechtsgesetz).