Google Quantum AI researchers have achieved a logical error probability of ≤ 3. 20× 10^(-8) at 60 cycles on a 105-qubit processor, a result limited only by bursts of ionizing radiation. This performance, two orders of magnitude below standard constructions, stems from the development of a “walking stability circuit” designed to combat temporally correlated errors. The team’s work extends beyond circuit design, identifying data-qubit heating to the third excited state |3⟩ as a key source of these errors and demonstrating a 38% reduction in spacetime logical error through adaptive syndrome extraction.
105-Qubit Willow Processor Enables Surface Code Stability Experiments
Scaling quantum error correction experiments to larger systems revealed a timelike block suppression factor of = 1. 47 ± 0. 25, demonstrating the ability to maintain temporal error suppression as the number of qubits increases on the 105-qubit Willow processor. This result, achieved by extending surface code stability experiments from 3×3 to 7×7 patch sizes, shows how to build fault-tolerant quantum computers by suppressing errors not just spatially, but also over time.
The team’s work moves beyond traditional memory experiments, which primarily address spatial faults, by focusing on the more challenging issue of temporal error chains. This heating effect, where qubits absorb energy and transition to higher energy levels, introduces errors that persist over time and are difficult to correct with standard methods.
Researchers fully captured this effect with a parameter-free, device-level leakage model incorporating higher-state transport during syndrome extraction cycles, providing a detailed understanding of the underlying physics. To address this error mechanism, the team developed and implemented a “walking stability circuit,” achieving a logical error probability floor of ≤ 3. Further refinement of the system through adaptive termination of syndrome extraction, based on decoder confidence in an offline analysis, reduced spacetime logical error by 38% and time overhead by 34% compared to the optimal fixed-cycle strategy.
Scaling Surface Code Stability to 7×7 Patches on Willow
Scaling to a 7×7 patch size on the Willow processor allowed researchers to probe temporal error floors down to , establishing a practical metrology tool for identifying correlated error mechanisms before scaling to larger quantum processors. This level of sensitivity is achieved because the fault distance in a stability experiment is determined by the number of cycles, rather than the number of physical qubits, enabling even small patches to reveal temporal limitations.
The stability experiment, alongside memory experiments, introduces and benchmarks the spacetime error, which is sensitive to both temporal and spatial errors. The work highlights the need to address low-probability correlated errors, such as those caused by high-energy events, through hardware mitigation strategies before further scaling up. The stability experiment should be treated as a benchmark equal in importance to memory experiments for scaling logical operations, providing a complementary view of device performance.
Data-Qubit Heating to |3⟩ Drives Error Suppression Behavior
Scaling from patch sizes of 3× 3 to 7× 7, the researchers realized a timelike block suppression factor of = 1.47 ± 0.25, demonstrating sustained temporal error suppression even as the processor size increased. The observed sum-of-exponentials behavior is not uniform across patch sizes; the exponential suppression after the knee remains consistent regardless of patch width, suggesting a shared underlying mechanism limits suppression beyond that point for all patches tested.
Walking Stability Circuit Reduces Logical Error to ≤ 3.20× 10^(-8)
A logical error probability of ≤ 3. 20× was achieved using a newly developed walking stability circuit, representing a two-order-of-magnitude improvement over standard surface code compilation techniques. This performance, realized on a 105-qubit processor, demonstrates a significant reduction in error rates critical for scaling quantum computation, though current limitations stem from bursts of ionizing radiation. Device-informed models revealed the relative importance of various physical error channels, highlighting the substantial impact of readout and reset processes within the stability experiment.
Researchers then used the stability experiment as a metrology tool, uncovering low-probability errors linked to temporal correlations, specifically, the heating of data qubits. These results demonstrate a pathway toward more robust and scalable quantum error correction, despite the remaining challenges posed by external radiation sources.
Decoder Analysis Reduces Spacetime Logical Error by 38%
This finding clarifies a key source of noise previously unaccounted for in standard error models. Researchers quantified the impact of this temporal error using a spacetime error probability, , calculated from both memory and stability experiments on a 105-qubit processor, confirming the ability to maintain error suppression as qubit numbers increase.
Stability Experiment Complements Memory Experiments for QEC
By extending the stability experiment for many cycles, the researchers uncover a sum-of-exponentials behavior in error suppression caused by a temporally correlated error: data-qubit heating to the third excited state |3⟩. The team found that, similar to repetition codes, the “walking stability circuit” is limited by ionizing radiation bursts. This optimization demonstrates the potential for intelligent error correction protocols that dynamically adjust to changing conditions, improving efficiency and reliability.
👉 More information
🗞 Resolving and Suppressing Temporally Correlated Noise in Quantum Error Correction
✍️ Alec Eickbusch et al.
🧠 ArXiv: https://arxiv.org/abs/2611.21578




See today’s quantum computing news on Quantum Zeitgeist for the latest breakthroughs in qubits, hardware, algorithms, and industry deals.
