Warning: Undefined array key "options" in /nfs/web_tum/www/s/di46los/webserver/htdocs/wp-content/plugins/elementor-pro/modules/theme-builder/widgets/site-logo.php on line 192

Towards a Universal Machine Learning Interatomic Potential for the x Li2S –(100-x) P2S5 Material Class

In this thesis, the aim was to find a method to investigate the structural complexity of the lithium thiophosphate material class and how this influences the lithium-ion conductivity (thus how fast the ions can move through the structure). A machine-learning approach (which replaced a much more computational demanding method) allowed for this. One main finding was that the ionic conductivity in amorphous (thus structures lacking a periodic ordering) is decreased, when many bulky thiophosphate building blocks were present, as they reduce the volume of lithium-ion channels and - other than smaller building blocks - cannot rotate to let the lithium ions pass.

Simplified Abstract

Lithium-ion batteries are so abundant in our daily life, that it is very difficult to imagine a world without them. All portable electric devices are powered by them, and they are by far the most promising technology for the future of automotive mobility. Particularly solid electrolytes are highly investigated in literature right now, as they are inflammable and the battery thus less prone to explosion. Currently, a quest for the best solid electrolyte, which most importantly needs to have a high lithium-ion conductivity, takes place in science. Lithium thiophosphates are a promising material class, containing cheap and earth-abundant materials.

Name:                        Tabea Huss