A Network Approach to Atomic Spectra

In this thesis we investigate atomic spectra using network theory. The spectroscopic data of an atom is mapped onto a network by identifying nodes with energy levels and links with optical transitions. These so-called spectroscopic networks are almost bipartite, which is ascribed to the parity symmetry. We apply community detection to these networks and show that the communities found correspond to known quantum numbers. For thorium II we demonstrate that the states of additional communities detected form one or two clusters in the energy domain. These clusters correlate with the dominant configurations of the states. We test the ability to use the network to predict quantum mechanical properties of individual states and show that for thorium II parity and J can be predicted with 100% and 95% accuracy respectively. We show that new transitions can be predicted using state of the art methods of link prediction. We benchmark this by a random dropout and demonstrate that depending on the atom 20% to 40% of the transitions can be recovered with few errors using the structural perturbation method. We develop a method that can predict the transitions to new states, for which structural information is given.

Level thesis:degree_level

master

Grantor dc:publisher

Universität Heidelberg

Year

2018

Author dc:creator

• Wellnitz, David

Contributors dc:contributor

• Weidemüller, Matthias

Repository record source_url

http://www.ub.uni-heidelberg.de/archiv/30157

OAI identifier oai:identifier

oai:archiv.ub.uni-heidelberg.de:30157