arXiv:2610.32347
Entangled logical qubits on a superconducting quantum processor
Robin Harper, Xanda C. Kolesnikow, Campbell K. McLauchlan, Georgia M. Nixon, Stephen D. Bartlett, Benjamin J. Brown
Abstract
We demonstrate high-fidelity logical qubit operations using \texttt{ibm\_phoenix}, a square-lattice array of 120 coupled superconducting qubits that are capable of fast reset. Using strategies to exclude underperforming components together with noise informed decoding, we routinely produce surface codes at distance-3 and distance-5 with logical error per round below 1\%. We demonstrate entanglement between two logical qubits, where logical Bell-state fidelities above 93\% are observed for entanglement created via lattice surgery. With three logical qubits and lattice surgery, we perform logical teleportation of a range of states and gates, observing fidelities in the range 78\% - 93\%. Using reset to reduce qubit leakage, we show that repeated rounds of stabilizer measurements continue to exhibit improvements in a stability experiment up to $\sim 2\cdot 10^{-7}$ over 170 measurement rounds.
AI Summary
This paper investigates surface-code memories and lattice-surgery operations on `ibm_phoenix`, a 120-qubit square-lattice superconducting processor, focusing on component variability and leakage. It combines ancilla re-purposing to exclude underperforming components, noise-informed belief-matching decoding, and fast measurement/reset. In memory experiments fitted over approximately 50 syndrome-extraction rounds, defect exclusion reduces the distance-5 logical error per round from $1.9\%$ to $0.9\%$. Distance-5 performance is comparable to the best distance-3 patches, rather than demonstrating clear distance-dependent suppression. Reported distance-5 error rates in the $X$, $Z$, and $Y$ bases are $0.872\%$, $0.855\%$, and $1.690\%$, respectively. Compressed lattice surgery enables operations between adjacent patches and three-logical-qubit teleportation experiments. Two-qubit Bell-state fidelities above $93\%$ are reported using bounds derived from logical $XX$ and $ZZ$ measurements, without $YY$ measurement. These results condition on error-free preparation detectors, retaining about $40\%$ of shots. Three-qubit state and Clifford-gate teleportation yields reported output-state fidelities of $78\%$–$93\%$; remote Bell-pair transfer has fidelity bounded between $85.4\%$ and $91.4\%$, with a $1.2\%$ preparation-selection yield. Gate characterization uses output-state tomography, not process tomography. A separate walking stability experiment resets every participating qubit every two rounds, observing 8 decoding failures in 50 million shots at 170 rounds, approximately $1.6\times10^{-7}$. This is not a logical-memory error rate. The logical-qubit experiments do not use walking circuits, leaving data-qubit leakage unaddressed. Stated remaining requirements include larger, higher-fidelity logical registers and logical non-Clifford operations; injected $T$ states here are prepared non-fault-tolerantly.
Central claim may not hold: The abstract's headline logical Bell-state fidelity omits that the reported high-fidelity result is conditioned on preparation-detector post-selection retaining only 40% ± 2% of shots. The abstract's three-logical-qubit teleportation fidelities omit preparation post-selection, with the Bell-pair-transfer demonstration retaining only 1.2% of shots to obtain fidelity bounds of 85.4%–91.4%.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 22 page, 15 figures, comments welcome