Quantum computation will rely on quantum error correction to counteract decoherence. Successfully implementing an error correction protocol requires the fidelity of qubit operationsto be well-above error correction thresholds. In superconducting quantum computers, measurement of the qubit state remains the lowest-fidelity operation. For the transmon, a prototypical superconducting qubit, measurement is carried out by scattering a microwave tone off the qubit. Conventionally, the frequency of this tone is of the same order as the transmon frequency. The measurement fidelity in this approach is limited by multi-excitation resonances in the transmon spectrum which are activated at high readout power. These resonances excite the qubit outside of the computational basis, violating the desired quantum non-demolition character of the measurement. Here, we find that strongly detuning the readout frequency from that of the transmon exponentially suppresses the strength of spurious multi-excitation resonances. By increasing the readout frequency up to twelve times the transmon frequency, we achieve a quantum non-demolition measurement fidelity of 99.93% with a residual probability of leakage to non-computational states of only 0.02%.
Hilbert space dimension is a key resource for quantum information processing. A large Hilbert space is not only an essential requirement for quantum error correction, but it can alsobe advantageous for realizing gates and algorithms more efficiently. There has thus been considerable experimental effort in recent years to develop quantum computing platforms using qudits (d-dimensional quantum systems with d>2) as the fundamental unit of quantum information. Just as with qubits, quantum error correction of these qudits will be necessary in the long run, but to date error correction of logical qudits has not been demonstrated experimentally. Here we report the experimental realization of error-corrected logical qutrits (d=3) and ququarts (d=4) by employing the Gottesman-Kitaev-Preskill (GKP) bosonic code in a circuit QED architecture. Using a reinforcement learning agent, we optimize the GKP qutrit (ququart) as a ternary (quaternary) quantum memory and achieve beyond break-even error correction with a gain of 1.82 +/- 0.03 (1.87 +/- 0.03). This work represents a new way of leveraging the large Hilbert space of a harmonic oscillator for hardware-efficient quantum error correction.
Bosonic codes offer a hardware-efficient strategy for quantum error correction by redundantly encoding quantum information in the large Hilbert space of a harmonic oscillator. However,experimental realizations of these codes are often limited by ancilla errors propagating to the encoded logical qubit during syndrome measurements. The Kerr-cat qubit has been proposed as an ancilla for these codes due to its theoretically-exponential noise bias, which would enable fault-tolerant error syndrome measurements, but the coupling required to perform these syndrome measurements has not yet been demonstrated. In this work, we experimentally realize driven parametric coupling of a Kerr-cat qubit to a high-quality-factor microwave cavity and demonstrate a gate set enabling universal quantum control of the cavity. We measure the decoherence of the cavity in the presence of the Kerr-cat and discover excess dephasing due to heating of the Kerr-cat to excited states. By engineering frequency-selective dissipation to counteract this heating, we are able to eliminate this dephasing, thereby demonstrating a high on-off ratio of control. Our results pave the way toward using the Kerr-cat to fault-tolerantly measure error syndromes of bosonic codes.
Efficient quantum control of an oscillator is necessary for many bosonic applications including error-corrected computation, quantum-enhanced sensing, robust quantum communication,and quantum simulation. For these applications, oscillator control is often realized through off-resonant hybridization to a qubit with dispersive shift χ where typical operation times of 2π/χ are routinely assumed. Here, we challenge this assumption by introducing and demonstrating a novel control method with typical operation times over an order of magnitude faster than 2π/χ. Using large auxiliary displacements of the oscillator to enhance gate speed, we introduce a universal gate set with built-in dynamical decoupling consisting of fast conditional displacements and qubit rotations. We demonstrate the method using a superconducting cavity weakly coupled to a transmon qubit in a regime where previously known methods would fail. Our demonstrations include preparation of a single-photon state 30 times faster than 2π/χ with 98±1(%) fidelity and preparation of squeezed vacuum with a squeezing level of 11.1 dB, the largest intracavity squeezing reported in the microwave regime. Finally, we demonstrate fast measurement-free preparation of logical states for the binomial and Gottesman-Kitaev-Preskill (GKP) code, and we identify possible fidelity limiting mechanisms including oscillator dephasing.