YFarmX logoYFarmX

Quantum News

Quantinuum reports five logical-memory errors per million rounds

Quantinuum reports a logical-memory error rate of about five in a million correction rounds on Helios, corresponding to 99.9995% fidelity per round. Its distance-7 colour-code result and error-suppression measurements support the company's fault-tolerance research programme.

Editorial collage based on Quantinuum’s official Helios chip image, with illustrative violet ion light and the reported 99.9995% logical-memory fidelity per correction round.

Listen to this articleListen

Podcasts

Quantinuum has reported a logical-memory error rate of roughly five per million error-correction rounds on its Helios trapped-ion processor. The result corresponds to 99.9995% logical-memory fidelity per round, according to the company’s 19 October account.

The experiment measures how reliably protected quantum information survives repeated checks for errors. Its unit is a syndrome-extraction cycle: one round of collecting the information needed to identify errors in the protected state.

Stronger protection reduced memory errors

The scientific paper tests triangular colour codes with distances of three, five and seven. Code distance describes the strength of the protection, with larger codes using more physical qubits to protect the logical state.

The reported memory-error estimates are 5.5 × 10⁻⁴ at distance three, 3.8 × 10⁻⁵ at distance five and 5 × 10⁻⁶ at distance seven. The last estimate has a reported range of +4/−2 × 10⁻⁶. Quantinuum’s paper describes the uncertainties as one-standard-deviation Wilson intervals unless specified otherwise.

Reported logical-memory errors per million syndrome-extraction cycles, with the paper’s uncertainty intervals. Bar lengths use a linear scale. Source: Quantinuum.
Reported logical-memory errors per million syndrome-extraction cycles, with the paper’s uncertainty intervals. Bar lengths use a linear scale. Source: Quantinuum. Open for full size.

The fitted memory-error suppression factor is approximately 12 as the tested code distance increases. This scaling result is significant because error correction becomes useful when added protection reduces the remaining logical failures despite requiring additional hardware operations.

Independent surface-code experiments by Google researchers have also demonstrated error suppression as code distance increases. Those results use a different processor and code; the shared engineering goal is to make a larger encoded memory more reliable.

Memory and gate experiments have their own results

The team separately tested logical two-qubit Clifford operations at distances three and five. The paper reports below-threshold scaling for those experiments too, with a suppression factor of roughly nine.

The 99.9995% headline describes logical memory per correction round. The logical-gate experiments provide their own error rates and scaling evidence.

Conceptual error-correction cycle. The distance-7 memory experiment uses 37 data qubits and 73 physical qubits in total including ancillas. Source: Quantinuum paper.
Conceptual error-correction cycle. The distance-7 memory experiment uses 37 data qubits and 73 physical qubits in total including ancillas. Source: Quantinuum paper. Open for full size.

Helios informs the Apollo programme

Quantinuum says the memory result reduces errors tenfold compared with its September record. The two demonstrations used different error-correcting codes on the same Helios hardware, allowing the researchers to investigate different balances of protection and qubit use.

The company connects that work with Apollo, its planned universal fault-tolerant system for 2029. That remains a development target. Extending reliable memory and logical operations into larger computations is the programme those measurements are intended to support.

Illustrative protection symbols with estimated Helios memory errors per million correction rounds: 550, 38 and 5 at code distances 3, 5 and 7.
Illustrative protection symbols and ion-trap hardware. Reported logical-memory estimates correspond to 550, 38 and 5 errors per million correction rounds at code distances 3, 5 and 7. The paper also reports uncertainty intervals. Source: Quantinuum. Open for full size.

Sources

  1. Quantinuum: logical-memory result and experimental contextquantinuum.com
  2. Scientific paper: quantum error correction below the colour-code thresholdquantinuum.com
  3. Quantinuum: 19 October announcementir.quantinuum.com
  4. Independent context: quantum error correction below the surface-code thresholdarxiv.org

How we use AI