Master thesis defense by Victoria B. Inselmann

TLS Dynamics Inferred from Sub-second T1 Fluctuations in Superconducting Qubits

Temporal fluctuations in superconducting qubit relaxation times caused by individual two-level system (TLS) defects pose a major challenge for reliable quantum information processing in superconducting devices. In this thesis, we built a TLS-driven hidden Markov model to analyse individual decay channels using full forward-backwards hidden-state estimation of temporal relaxation-time data. Mapping the decay channels helped identify 4 types of TLS interactions: regular and irregular local interacting defects, global drifts, and global irregular shifts. We analysed 121 90-second experimental data windows, chosen for minimal drift and stable switching rates, revealing dwell times ranging from 20 to 1000 milliseconds and switching rates between 1 and 50 Hz for individually estimated decay channels. These results highlight new and improved identification of individual TLS interactions in new short timescales. Using the Bayesian Information Criterion, we estimate approximately 4–6 active decay channels per window. We assess model validity through parametric bootstrapping and assess the goodness of fit for the hidden Markov model on experimental data in a dynamical $\SI{5}{\second}$ resolution, where the hidden Markov model reliably describes the experimental data within a $95\%$, except for outliers such as persistent shifts or unusually long dwell times annotated to the global TLS types of interactions.

This thesis establishes a framework for further analysis of hidden on-chip dynamics inferred from qubit relaxation time, moving from the umbrella term TLS to individual characterisations.