Impact of control signal phase noise on qubit fidelity

Abstract As qubit decoherence times are increased and readout technologies are improved, nonidealities in the drive signals, such as phase noise, are going to represent a crucial limitation to the fidelity achievable at the end of complex control pulse sequences. While prior studies have addressed the impact of reference oscillator phase noise on qubit performance, its interaction with realistic control pulses and its role in fidelity degradation have not been examined in sufficient detail. Furthermore, previous analyses grounded in the filter-function formalism identify high-frequency spectral components as the dominant source of fidelity loss, a conclusion that has also been used as a compelling marketing feature by some hardware vendors. Here we reevaluate this assertion by means of direct time-domain numerical simulations, in which phase noise realizations with a given power spectral density are applied to the carrier of realistic control pulse sequences, and the resulting qubit evolution is computed with Qiskit-Dynamics and averaged over multiple noise realizations. We show that the claimed dominance of high-frequency components originates from an oversight in interpreting the conversion between the phase-noise and the frequency-noise power spectral densities: at equal power, the components close to the Rabi frequency are by far the most detrimental, while those far from the carrier act only marginally, mainly through residual amplitude modulation. We performed an analysis using both two- and three-level models, the latter capturing leakage outside the computational subspace. We further show that, for realistic phase noise spectra of local oscillators, whose power is concentrated in the low-offset region, the fidelity loss is dominated by such high-power low-frequency components.