High-fidelity detection of information encoded in bosonic modes: Part II

Single-shot qubit measurement is vital for universal quantum computation. In the field of superconducting qubits, much progress has been made in the readout and amplification chain; nonetheless, state-of-the-art measurement fidelities are limited by relaxation and detector inefficiency to about 99%. We present an experimental demonstration of a recent proposal [Hann et al, Phys. Rev. A 98, 022305] to improve measurement fidelities by orders of magnitude. By combining repeated QND measurements with error-tolerant encodings, we suppress the effects of both relaxation and detector inefficiency. The results are compared with theoretical predictions and achievable limits are described.