Quantum Efficient Measurement of a Transmon via a High Saturation Power Josephson Parametric Amplifier Part 2

Achieving high-fidelity, quantum non-demolition qubit readout is a key challenge for the future of superconducting quantum computing. Readout pulses must be strong enough to produce a detectable dispersive frequency shift yet weak enough to avoid excess qubit dephasing and unwanted qubit-resonator energy transitions. In this talk, we present a set of methods for optimizing transmon measurement fidelity using a high-saturation power, quantum-efficient, rf-SQUID-based parametric amplifier. Given our amplifier's performance, we find that the primary limitation is the transmon's degradation of coherence when measured with strong pulses. To optimize fidelity, we vary the pulse shape, amplitude, and duration of our readout to find the optimal compromise between qubit decoherence and information output. Furthermore, we classify the results with a machine learning algorithm, which can autonomously learn the weight functions that extract the maximum information from each readout shot. With this method we achieve maximum total classification fidelity of 99.5% with phase-sensitive measurement.