Skip to content
Quantum.world

The Signal

Q-CTRL reports high-fidelity magic states on IBM Nighthawk

Q-CTRL says it prepared a cultivated magic state on an IBM Nighthawk processor with a logical error probability of 2 × 10⁻⁵. Its preprint reports that lattice-surgery transfer preserved fidelity above 0.99 after 10 error-correction rounds at about 5% shot retention.

Q-CTRL reports high-fidelity magic states on IBM Nighthawk

Q-CTRL says it prepared a cultivated magic state on an IBM Nighthawk processor with a logical error probability of 2 × 10⁻⁵. The company’s results, detailed in an arXiv preprint first submitted November 17, report that lattice-surgery transfer preserved fidelity above 0.99 after 10 rounds of error correction at about 5% shot retention.

The experiment used publicly accessible IBM hardware. Q-CTRL published its account on November 18, describing the work as a demonstration of magic-state cultivation with lattice-surgery escape, a process for preparing a logical state and transferring it into a larger error-correcting code.

Two routes for the cultivated state

Magic states are resource states used to perform T gates, a type of operation needed for universal fault-tolerant quantum computation. The researchers first prepared and refined a T state in a compact color code, then transferred it into a larger code. Runs with unexpected measurement outcomes or detected errors were discarded.

The team tested two transfer methods: grafting, which changes the code, and lattice surgery, which teleports the state into a surface code. The preprint says lattice-surgery escape had not previously been experimentally demonstrated on physical hardware. It describes the study as implementing both methods on a distance-5 code on the IBM processor.

Reported error and fidelity figures

At the T-state injection angle, the cultivated state had a reported logical error probability of 2 × 10⁻⁵. Q-CTRL gives its fidelity as 0.99998 and says 25% of shots were retained; the paper compares the result with the prior experiment, reporting nearly 10 times lower error and about three times the retainment.

After lattice-surgery transfer, Q-CTRL reports fidelity exceeding 0.99 at up to 16% retainment when using decoder-confidence postselection. After 10 full rounds of error correction, the company reports fidelity of 0.991 at 5% retainment. The preprint separately gives fidelity above 0.99 at about 5% retainment after 10 rounds, and reports a logical error rate of about 0.14% per stabilizer round with around 11% average shot retainment.

For comparison, the paper says the matchable-switched grafting variant fell below 0.95 fidelity ten rounds after escape at comparable retainment. Q-CTRL also reports that its grafting implementation reduced the logical error rate per stabilizer round compared with the previous experiment, while retaining a higher share of shots.

What the approach requires

Q-CTRL attributes the performance to custom compilation, qubit-layout selection, interleaved dynamical decoupling and an upgraded decoding procedure. The company says lattice surgery moves the state directly into a surface code using operations that surface-code processors already use. The preprint decodes the lattice-surgery circuit with minimum-weight perfect matching, and the grafting variants with Tesseract or a hybrid Tesseract–matching decoder.

Q-CTRL says the results set three records: magic-state cultivation error, magic-state fidelity immediately after transfer into a surface code, and fidelity after 10 rounds of error correction. The company’s proposed next steps are to scale surface codes to larger distances on processors with more qubits, and to use cultivated states directly in fault-tolerant logical T gates through real-time decoding and active feedforward.

Sources

  1. arXiv
  2. Q-CTRL
  3. arXiv (full text)

Companies in this story

Related stories

Explore

More articles