Ultrafast flux-assisted non-dispersive readout of a fluxonium qubit

Maximising readout speed while simultaneously preserving the purity (QNDness) of the measured qubit is critical to designing an efficient readout protocol. Existing efforts have focused mainly on readout in the dispersive regime, where QNDness is high, but readout time is limited by the large qubit-resonator detuning. There, the readout time t_r~Δ/g²√N, where Δ and g are the qubit-resonator detuning and coupling constants, respectively, and N the initial number of photons in the readout resonator, is typically in the hundreds of nanoseconds. We propose instead to take advantage of the non-dispersive regime, where readout time is sped up to t_r~1/g, or just a few nanoseconds. Remarkably, the purity remains approximately one up to Nt_r after measurement is performed, as long as the qubit is initialised in one of two pairs of "magic angles'', where it is (anti-)parallel to the Rabi vector of the resonator. We further find that adding a second, classical, flux drive further speeds up the qubit readout while simultaneously improving the qubit purity, both at order O(1/N). Finally, we suggest that the large anharmonicity of fluxonium combined with the choice of non-dispersive regime and low initial number of photons protects the protocol from parasitic measurement-induced phase transitions (PMIST).