arXiv:2610.24077
Quantum error correction well below the color code threshold
Ali Lavasani, Adam P. Reed, Joan M. Dreiling, Annie J. Park, Michael Mills, Matteo Puviani, Gabriel Matos, M. S. Allman, Noah Berthusen, Natalie C. Brown, Victor E. Colussi, Brian Estey, Cameron Foltz, Daniel Gresh, Reza Haghshenas, Akhil Isanaka, Nikhil Kotibhaskar, Anthony Ransford, Ciarán Ryan-Anderson, Brian Neyenhuis, David Hayes, Tzvetan S. Metodi, Andrew C. Potter
Abstract
Quantum error correction enables scalable fault-tolerant quantum computation by suppressing logical errors exponentially with code distance once physical error rates fall below a critical threshold. Here we demonstrate below-threshold operation of triangular color codes with distances $d=3,5,$ and $7$ on Quantinuum's Helios trapped-ion quantum processor. Leveraging both circuit- and hardware- level leakage reduction schemes, we achieve logical error rates as low as $5_{-2}^{+4}\times 10^{-6}$, and demonstrate error suppression by over an order of magnitude, $Λ=12_{-2}^{+3}$, with each successive increase in code distance. Moreover, we demonstrate below threshold scaling of error rates for Clifford two-qubit randomized benchmarking circuits when encoded in $d=3,5$ color codes, achieving logical error rate of $7^{+3}_{-3}\times 10^{-5}$ per logical two-qubit Clifford at distance $5$ with a suppression factor of $Λ=9_{-3}^{+8}$. We compare logical performance for existing near-optimal algorithmic decoders, and both convolutional neural network and transformer-based neural network decoders capable of real-time decoding an order of magnitude faster than the syndrome extraction rate. These results approach the logical error rates required for megaquop-scale quantum computation and indicate that scaled trapped-ion processors could reach application-relevant performance with substantially smaller code blocks than indicated by previous results in other qubit modalities.
AI Summary
The paper investigates how strongly triangular color-code memories and logical Clifford circuits can suppress errors with increasing distance on a trapped-ion processor. Experiments on Helios-1 combine hardware leakage repumping with a SWAP-based syndrome-extraction circuit that periodically replaces data qubits with measured and reset ancillas. Three repump cycles reduce leaked population by $6.5(1)\times$ while overlapping cooling and transport. Memory experiments prepare six logical Pauli eigenstates, perform 20 syndrome-extraction rounds, and measure the logical observable at distances $d=3,5,7$, requiring 13, 37, and 73 physical qubits per block. With the Tesseract decoder, estimated errors per round are $5.5\times10^{-4}$, $3.8\times10^{-5}$, and $5^{+4}_{-2}\times10^{-6}$, respectively. A log-linear distance fit gives suppression $\Lambda=12^{+3}_{-2}$ per distance increment of two. Two-qubit Clifford randomized benchmarking, using transversal gates and syndrome extraction every seven Cliffords, yields errors of approximately $7\times10^{-4}$ and $7\times10^{-5}$ per logical Clifford at distances 3 and 5, with suppression approximately ninefold. Algorithmic and emulator-trained neural decoders are compared on hardware records. Neural decoding transfers without hardware fine-tuning and takes roughly $0.2$–$0.5$ ms per round on an H100 GPU. All decoding is offline; reported processing times do not establish end-to-end feedback latency. The distance-7 memory estimate rests on three failures, and per-round rates are inferred from one circuit length rather than fitted versus duration. Emulator mismatch and coherent-error approximations remain limitations. Distance-7 two-qubit benchmarking exceeds available qubits; universal operation additionally requires high-fidelity magic-state preparation and integrated real-time decoding.
Error found: The abstract's claim of greater-than-ten memory-error suppression with each distance increase is stronger than the individual measured ratios, approximately 14.28 and 7.61, although the common fitted suppression factor Λ=12 remains supported. The lower uncertainty on the logical Clifford suppression factor is reported inconsistently as -3 in the abstract and -4 in the main text.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.