Regensburg 2022 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

HL: Fachverband Halbleiterphysik

HL 30: Poster 2

HL 30.45: Poster

Donnerstag, 8. September 2022, 11:00–13:00, P3

Dimensionality Reduction Techniques in Femtosecond Time-Resolved Ellipsometry Data Analysis and Theory — •Noah Stiehm¹, Yixuan Zhang², Erich Runge³, Stefan Krischok¹, Hongbin Zhang², and Rüdiger Schmidt-Grund¹ — ¹Technische Universität Ilmenau, Fachgebiet Technische Physik I, Weimarer Straße 32, 98693 Ilmenau, Germany — ²Technische Universität Darmstadt, Research Group Theory of Magnetic Materials, Otto-Berndt-Straße 3, 64287 Darmstadt — ³Technische Universität Ilmenau, Fachgebiet Theoretische Physik I, Weimarer Straße 32, 98693 Ilmenau, Germany

Physical modeling and interpretation of the transient dielectric function obtained from femtosecond time-resolved spectroscopic ellipsometry [1] poses a significant challenge, as it consists of many temporally and spectrally overlapping processes that need to be reliably separated to obtain stable and physically meaningful fit results. Ab-initio theory can help to separate these processes, but is not available for the whole time scale (≈100 fs to several ns) of the experiment, due to the associated computational costs.

To help overcome these challenges we investigate the use of dimensionality reduction techniques like dynamic mode decomposition and manifold learning methods like locally-linear embeddings and autoencoders to be applied I) in model approximation of the experimental data and II) on ab-initio results from time-dependent density functional theory to cover larger time scales. By this we identify strategies for a reliable modeling pipeline with minimal human intervention.

[1] S. Richter et al., Rev. Sci. Instrum. 92, 033104 (2021).