Quantinuum researchers report a sharp reduction in quantum memory errors on their Helios trapped-ion processor, a step toward more reliable quantum computation. Their preprint appears in arXiv’s new listings for October 20, following the company’s October 19 announcement from Broomfield, Colorado. The reported memory fidelity is 99.9995%, equivalent to about five logical failures per million error-correction rounds.
The study by Ali Lavasani and colleagues tests how effectively extra physical qubits can protect a logical qubit, the encoded unit that stores quantum information. According to the researchers, error correction becomes useful when hardware noise is low enough that adding protection reduces errors instead of introducing more of them. This is a central requirement for running longer quantum calculations.
The team tested triangular color codes with distances of three, five and seven, progressively increasing the protection. Logical memory errors declined by an average factor of about 12 at each step, according to the preprint. In separate two-qubit Clifford gate tests, increasing the code distance from three to five reduced errors by a factor of about nine. Memory and gate results measure different aspects of performance.
Quantinuum’s technical account says the strongest memory estimate came from three observed failures across 29,000 experimental shots, each containing 20 correction rounds. That totals 580,000 rounds. The preprint gives an estimated rate of five errors per million, with uncertainty extending from roughly three to nine. The figure describes protected memory per round, rather than an entire algorithm succeeding with that accuracy.
The company says the memory error rate improves tenfold on its September result using the same Helios hardware with a different code. It links the progress to its planned Apollo fault-tolerant computer for 2029. That date remains a company target, and the newly listed research is a preprint, rather than evidence of an already available universal fault-tolerant machine.
Further integration remains necessary. Quantinuum acknowledges that below-threshold scaling of magic-state preparation, a resource needed for universal computation in this approach, has not yet been demonstrated. Its stated next step is to combine the color-code components into larger, more reliable calculations. The results therefore support progress in error suppression while leaving the full computing system to be demonstrated.
Comments