The Cat Qubit: A Qubit Built to Ignore One Kind of Error
Most qubits suffer two kinds of error and need a large grid of checks to fix both. A cat qubit is designed so that one kind almost never happens, leaving a simpler job.
The trick is to remove photons in pairs. That pins the field into two stable states, which oscillate exactly out of step.1 One state is the 0, the other the 1. Despite the name, the “cat states” of many entangled atoms or ions, used in some architectures, are a different thing from cat qubits.
Two errors, one held off
Qubits suffer two kinds of error. A bit flip swaps 0 and 1. A phase flipAn error that leaves 0 and 1 alone but flips the sign between them in a superposition, which classical bits never suffer. leaves them in place but scrambles the delicate relation between them.
In a cat qubit, a bit flip means the field jumping all the way from one state to the other. Adding more photons pushes the two states farther apart, and the bit-flip rate falls exponentially.2 The cost is gentle: phase flips grow only in proportion to the photon number. The result is biased noiseNoise in which one kind of error is much rarer than the other..
In one 2025 experiment, with about two photons, bit flips took over a millisecond to appear, while phase flips arrived in about 30 microseconds.5
A line of checks, not a grid
Error correctionTechniques that combine many shaky physical qubits into fewer reliable ones, so a long calculation stays correct. normally guards against both errors, which is why the surface codeAn error-correcting code that spreads one protected qubit over a square grid of physical qubits. needs a two-dimensional grid. If bit flips are held off, only phase flips are left, and a single line of qubits with checks between neighbors can catch them.3 A distance-5 surface code uses 49 qubits in all. A distance-5 line uses five cat qubits plus a few helpers that run the checks.
This does not remove error correction; it moves part of the job into the hardware. Removing photons in pairs keeps correcting bit flips on its own, continuously, inside each qubit.
What experiments show, and the catch
A 2024 experiment pushed bit-flip times beyond 10 seconds, ten thousand times better than before, but its phase-flip times were just over 490 nanoseconds.4 And every gate must keep the bias, or bit flips sneak back in.
As of October 2026, two companies build cat-qubit chips: AWS and Alice & Bob.57
A 2023 design study estimated that about 126,000 cat qubits could run one code-breaking calculation in 9 hours, if its hardware assumptions are met.6 When you read such a claim, ask how many photons are used, how long phase flips take, and whether gates kept the bias.
Figure 1: Why cat qubits need fewer checks. A: the two states sit far apart, so a bit flip means a jump all the way across. B: adding photons makes bit flips rapidly rarer, while phase flips get only a little more frequent. Shapes are schematic. C: with bit flips held off, a single line of checks can be enough to catch the phase flips.
Sources
M. Mirrahimi et al., “Dynamically protected cat-qubits: a new paradigm for universal quantum computation,” New J. Phys. 16, 045014 (2014). ↩
R. Lescanne et al., “Exponential suppression of bit-flips in a qubit encoded in an oscillator,” Nat. Phys. (2020). ↩
J. Guillaud and M. Mirrahimi, “Repetition Cat Qubits for Fault-Tolerant Quantum Computation,” Phys. Rev. X 9, 041053 (2019). ↩
U. Réglade et al., “Quantum control of a cat-qubit with bit-flip times exceeding ten seconds,” Nature 629, 778 (2024). ↩
H. Putterman et al. (AWS), “Hardware-efficient quantum error correction via concatenated bosonic qubits,” Nature 638, 927 (2025). ↩
É. Gouzien et al., “Performance Analysis of a Repetition Cat Code Architecture: Computing 256-bit Elliptic Curve Logarithm in 9 Hours with 126133 Cat Qubits,” Phys. Rev. Lett. 131, 040602 (2023). ↩