Fig. 3: Measuring qubit stability using time-resolved GST.

The results of two time-resolved GST experiments using the gates G_i, G_x, and G_y, with drift compensation added for the second experiment. a, b The evidence for instability in each circuit in the first experiment, quantified by \({\lambda }_{{\rm{p}}}=-{\mathrm{log}\,}_{10}({\rm{p}})\) where p is the p-value of the largest power in the spectrum for that circuit. A pixel is colored green when λ_p is large enough to be 5% statistically significant, otherwise, it is greyscale. Each circuit consists of repeating a “germ” sequence in between six initialization and pre-measurement sequences. The data is arranged by germ and approximate circuit length L, and then separated into the 6 × 6 different preparation and measurement sequence pairs, as shown on the axes of B (“{}” denotes the null sequence). Only long circuits containing repeated applications of G_i exhibit evidence of drift. In the second experiment, none of the λ_p are statistically significant (data are not shown). c, d Time-resolved tomographic reconstructions of the gates in each experiment, summarized by the diamond distance error of each gate, and the decomposition of the coherent errors in G_i into rotation angles around \(\hat{x}\), \(\hat{y}\) and \(\hat{z}\) (\({t}_{\max }\approx 5.5\) h and \({t}_{\max }\approx 40\) h for the first and second experiment, respectively). e The power spectrum for each experiment obtained by averaging the individual power spectra for the different circuits, with filled points denoting power above the 5% significance thresholds (the thresholds are not shown).