Fig. 4: Trade off-free performance scaling.

a Measurement of preparation error as a function of time, ε(t), for different Rabi drive amplitudes of \({{{\Omega }}}_{l,r}^{01}\). b Parametric plot of the steady-state error ε_∞ and the stabilization time constant τ, obtained from fitting the data in (a) showing the expected linear scaling. The minimum ε_∞ and τ are measured at \({{{\Omega }}}_{l,r}^{{{{{{{{\rm{opt}}}}}}}}}\approx 1.5\kappa\). The minimum error \({\varepsilon }_{\infty }^{\min }\) is close to theoretically achievable value for experimental κ (shown as the gray floor), (Supplementary Fig. 11). The dotted line is the error-time scaling obtained from a simulation of Eq. (1) and has slope equal to 1/T_B with T_B = 3.38 μs, independent of drive amplitudes, in accordance with the semi-classical estimate discussed in the “Methods” section. The deviation of experimentally observed slope from 1/T_B is quantitatively explained by including a residual drive detuning and imbalance in the sideband amplitude (Supplementary Fig. 12).