Biased-noise cat qubits and their gates

Short answer: Cat qubits are bosonic qubits whose noise is strongly biased toward one kind of error, which can make error correction cheaper. A PNAS paper on a detuned Kerr-cat qubit reports a noise bias approaching 250 with a phase-flip time of 4 microseconds, using gate set tomography and randomized benchmarking. A SciPost preprint chapter explains why the Kerr-cat qubit's noise structure is preserved during gates. These are laboratory characterizations and a tutorial, not a fault-tolerant machine.

Why biased noise matters

The PNAS abstract says ubiquitous noise remains a key obstacle to quantum computers and calls for error correction. It adds that codes tailored to noise-biased systems have been shown to offer high fault-tolerance thresholds and reduced hardware overhead, which makes noise-biased qubits promising candidates. The catch, the authors say, is that operations on these platforms remain challenging and their noise structures have not been benchmarked as rigorously as conventional hardware.

The Kerr-cat qubit

The SciPost chapter introduces the bosonic Kerr-cat qubit, encoded in a parametrically driven nonlinear oscillator. The author says the qubit couples anisotropically with its environment, which produces a structure or bias in its noise channel. The structure of the noise is preserved even during non-trivial gate operations, and the chapter says this can be exploited to design more efficient fault-tolerant error correcting codes. It is a tutorial chapter rather than an experiment.

The PNAS benchmark

In the PNAS paper the authors develop a quantum control toolbox for a scalable noise-biased qubit, the detuned Kerr-cat qubit. It covers initialization, universal single-qubit gates and quantum nondemolition readout. They characterize the noise of these operations with gate set tomography and dihedral randomized benchmarking. They report a Z(pi/2) gate fidelity of 99.18% with an uncertainty of 0.066, and an X(pi/2) gate fidelity of 92.5% with an uncertainty of 0.23. The gap between the two shows the gates are not equally good.

They report a noise bias approaching 250 with a phase-flip time of 4 microseconds, which they say outperforms resonant Kerr-cat qubits reported previously. They also say their results reveal a critical overestimation of the operational noise bias inferred from bit-flip and phase-flip times alone. The abstract continues beyond that point, but the rest could not be read here and is not described.

A third source

A 2023 PRX Quantum paper on Zeno gates for dissipative cat qubits was listed as a source, but its abstract was not readable, so it is not described.

Limits

The benchmarks come from a single type of device in one laboratory and concern single-qubit operations. A bias near 250 is a measured property of that qubit, and the abstracts do not show a logical qubit or an error-corrected computation.

FAQ

What is noise bias?

A strong imbalance between error types, such as phase flips being much more likely than bit flips, which tailored codes can exploit.

How biased was the detuned Kerr-cat qubit?

The authors report a bias approaching 250 with a phase-flip time of 4 microseconds.

Were all the gates equally good?

No. The Z(pi/2) fidelity was 99.18% and the X(pi/2) fidelity was 92.5%.

Why not just use bit-flip and phase-flip times?

The authors say that overestimates the operational bias, so they benchmark the operations directly.

Is this a fault-tolerant qubit?

No. The abstracts report single-qubit characterization.