Figure 2. Experimental time-division multiplexing for qubit stabilization against dephasing.

(a) Schematics showing the key aspects of the implementation. Noise is continuously injected into the system. Measurements are taken up to t₀ and processed in real-time to predict future evolution of the qubit phase many time steps ahead, φ^P(t_k). From t₀ measurement-free compensation based on φ^P(t_k) is applied during each discrete time step (light green arrows) up to t_k when a diagnostic measurement is performed to verify the accuracy of the prediction/correction process. Full details appear in Supplementary Methods. (b) Probe and stabilization cycles of a time-division multiplexed measurement using n=100 past measurements and prediction/correction up to k=50 time steps ahead. Green shading indicates reduced residual phase errors. (c) r.m.s. results from time-division multiplexed measurements for different t_k and n compared against traditional feedback and averaged over 50 unique stabilization periods. Data are normalized to the r.m.s. of φ^A indicated by the horizontal dashed line. The other dotted/dashed lines are simulations. Markers represent the averaged results of diagnostic measurements. For these data ω_s=40ω_c.