Purity benchmarking study of error coherence in a single Xmon qubit

In this study, we employ purity benchmarking (PB) to explore the dynamics of gate noise in a superconducting qubit system. Over 1110 hours of observations on an Xmon qubit, we simultaneously measure the coherence noise budget across two different operational frequencies. We find that incoherent errors, which predominate in overall error rates, exhibit minimal frequency dependence, suggesting they are primarily due to wide-band, diffusive incoherent error sources. In contrast, coherent errors, although less prevalent, show significant sensitivity to operational frequency variations and telegraphic noise. We speculate that this sensitivity is due to interactions with a single strongly coupled environmental defect - modeled as a two-level system - which influences qubit control parameters and causes coherent calibration errors. Our results also demonstrate that PB offers improved sensitivity, capturing additional dynamics that conventional relaxation time measurements cannot detect, thus presenting a more comprehensive method for capturing dynamic interactions within quantum systems. The intricate nature of these coherence dynamics underscores the need for further research.