Figure

Description
The Kerr-cat qubit is a superconducting bosonic qubit that encodes quantum information in two coherent-state superpositions of a Kerr nonlinear oscillator stabilized by a resonant two-photon drive. The two computational states are the even and odd Schrödinger cat states , which are confined to a two-dimensional manifold within the oscillator’s infinite-dimensional Hilbert space by the interplay of the Kerr nonlinearity and the parametric drive.
The key feature of the Kerr-cat qubit is exponential noise bias: bit-flip errors (transitions between and ) require tunneling through a phase-space barrier that grows with , and are therefore exponentially suppressed with increasing cat size. Phase-flip errors (relative phase between and ) grow only linearly with . This extreme noise asymmetry can be exploited by tailored quantum error correction codes (e.g., repetition codes for phase flips only), dramatically reducing the overhead for fault-tolerant quantum computing.
Demonstrated by Grimm et al. (2020) at Yale using a SNAIL-based parametric oscillator, and further developed by the Amazon AWS Center for Quantum Computing in collaboration with Yale.
Hamiltonian
The Kerr-cat Hamiltonian in the rotating frame of the oscillator is:
where is the Kerr nonlinearity (self-Kerr), are the ladder operators, and is the two-photon drive amplitude.
The two classical stable states are coherent states with amplitude . The quantum ground states are the even/odd cat states:
The energy gap protecting against bit-flips scales as:
and the bit-flip rate is suppressed as .
Motivation
Standard bosonic codes (GKP, binomial) require complex syndrome extraction. The Kerr-cat qubit achieves autonomous error bias through its Hamiltonian structure alone: the two-photon drive continuously stabilizes the cat-state manifold without active feedback. By concentrating errors into a single channel (phase flips), the Kerr-cat enables a path to fault-tolerant quantum computing using simpler, lower-overhead error correction — potentially a repetition code rather than a full surface code. This could reduce the number of physical qubits needed for a logical qubit by an order of magnitude.
Key Findings
- Bit-flip times exceeding demonstrated with noise bias ratios (bit-flip rate / phase-flip rate) (Lescanne et al. 2020).
- Quantum non-demolition readout of the cat qubit state using dispersive coupling to an ancilla transmon (Grimm et al. 2020).
- Gate set for biased-noise qubits: bias-preserving CNOT and Toffoli gates demonstrated (Puri et al. 2020).
- Combination with repetition code shown to achieve exponential suppression of both error types simultaneously (Guillaud & Mirrahimi 2019).
- Autoparametric stabilization scheme eliminates the need for an external two-photon pump (Réglade et al. 2024).
Key Metrics
| Metric | Value | Notes | Fidelity reference |
|---|---|---|---|
| Bit-flip time | >1 ms | Exponentially suppressed with | Lescanne et al. 2020 |
| Phase-flip time | 1–10 μs | Grows linearly with ; dominant error | Grimm et al. 2020 |
| Noise bias ratio | – | Lescanne et al. 2020 | |
| Cat size | 2–8 photons | Larger increases bias but also phase-flip rate | Grimm et al. 2020 |
| Kerr nonlinearity | 1–10 MHz | Set by SNAIL or transmon nonlinearity | Grimm et al. 2020 |
| 1Q gate fidelity (Z-rotation) | 99%+ | Bias-preserving via detuning of drive | Puri et al. 2020 |
| 2Q gate fidelity (ZZ-CNOT) | 98–99% | Bias-preserving CNOT | Puri et al. 2020 |
| Operating temperature | 10–20 mK | Dilution refrigerator | — |