Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

Reconstruction locality bounds charge-generated dynamics in quantum codes: an exact accuracy floor and a measurable certificate for it

Zongmin Liu

Published Aug 1, 2026• 0 citations
Publisher

Abstract

Rare correlated error bursts are the reported scaling limit of below-threshold surface-code memo- ries, and existing analyses treat them as events that elevate the error rate across an entire code patch. We show that this assumption fails in the regime every scaled code must enter, where the patch is larger than the burst. There the logical error rate is controlled not by the burst rate but by the tail of the burst footprint distribution: for Pr[Ω > r] ∼ r−α the burst contribution obeys εb(d) ∝ dD−α, so exponential suppression in code distance is replaced by a power law and fault-tolerance over- head becomes polynomial rather than polylogarithmic, with a critical exponent αc = D separating suppressible from unsuppressible environments. Circuit-level simulations over nine code distances, d = 5 to 21, confirm the scaling function and the exponent transfer. We then measure the footprint distribution itself, from 265 radiation bursts resolved on a 60-qubit gap-engineered array, using a forward model of the array that is calibrated against synthetic catalogues before use. The distribu- tion is lognormal, with σ = 0.23 [0.21, 0.26] and median radius 2.9 mm, and is preferred over the best power law by ∆AIC = 71; that comparison is itself calibrated, placing the observed value 9.4σ from the power-law expectation. A lognormal has no tail exponent, so the non-scalable regime α ≤ D is excluded by ten standard deviations and exponential suppression survives in present hardware — refuting a prediction we had made from radiation-transport scaling. The measurement also shows that gap engineering has acted on the shape of the footprint distribution rather than on the burst rate alone, which under the transfer law makes phonon management an asymptotic rather than a constant-factor intervention. Independently, no burst in 5.2 h is large enough to span a distance-7 patch, which accounts for the absence of burst excess in 28 public distance-7 datasets. The largest footprint sampled corresponds to R = 4.9 mm, leaving the distribution beyond that radius — and hence the asymptotics of larger arrays — open to the same measurement on a larger device.

DOI
10.5281/zenodo.22004224

Related papers

Why large-scale quantum computing faces hard physical limits: a source-loss balance tested against measured data

Oleg Arnold

This note applies a single first-order source-loss balance, C dθ/dt = A − ρθ, to systems across many scales — from a planetary surface to a qubit register — and reports, with exact numbers, where the balance reproduces measured values and where it does not. The central quantitative result concerns quantum error correction. Using published surface-code and trapped-ion data (Google Quantum AI, Nature 638, 920 (2024); Quantinuum Helios colour-code results, 2026), the note shows that the logical error per cycle falls by a constant factor Λ only while errors are independent. Correlated bursts act as a common loss channel whose rate does not fall with code distance; the quoted floor is about 3×10⁻¹⁰ per cycle, reached at distance about 46 for Λ = 2.17. For factoring a 2048-bit integer (about 1.4×10³ logical qubits, 6.5×10⁹ Toffoli gates, 4.96 days at a 1 μs cycle), the permitted logical error is 1.7×10⁻¹⁵ per qubit per cycle. At measured suppression factors this requires code distances of about 78 (Λ = 2.17) or about 24 (Λ = 12), while the published design uses 25 — and the correlated-error floor is not removed by increasing distance. This is a quantitative statement about where the scaling of error suppression stops matching the requirements of large computations, not a claim that small-scale quantum devices cannot work. The note also lists, separately and with exact values, the tests the balance fails: quark mass-step prediction (T6), gas/solid classification (T8, p = 0.81), nuclear count with A = 3, 4, prediction SP1 (−11.8% against a 10% threshold), coupling versus number of doors (T10, p = 0.015), width-to-cosmological-scale mapping (q₀ = −0.528 ± 0.011 against 0), human loss-rate estimates (11.1 and 15.3 W/K), critical Q₁₀ = 3.03 against measured 2.0–2.3, Higgs width from the b-quark exit alone (13.7% residual), shell boundaries (13 of 18), and several time constants. Of nine tests where the balance gives a value different from the standard one, one agreed with measurement and eight did not; the rules and thresholds were written before the data were examined. Successful checks include: photosynthesis stage efficiencies (0.487 × 0.094 = 0.0458 against 0.046 reported), a three-stage 2.45 GHz rectenna chain (0.906³ = 0.744), closure of the lunar surface energy budget to a ratio of 1.005, mode quality factors Q = m/Γ for the τ lepton (7.84×10¹¹), the Z boson (36.5) and the Higgs boson (3.38×10⁴), the Higgs lifetime from its width (290.87 fs against 290.3 ± 0.5 fs, a consistency check of standard formulas), and the proton-decay bound ρ < 1.3×10⁻⁴² s⁻¹. The complete theory, tests, scripts and the list of open items are maintained at https://zenodo.org/records/23812417. This note is a standalone publication; readers are invited to report measured values that disagree.

•Nov 13, 2026•0 citations

Correlated Error Bursts from Ionising Impacts: A Rate Budget for Millimetre-Scale Surface-Code Patches

William Alexander Aldred

Ionising radiation, from cosmic rays and from trace radioactivity in the laboratory itself, deposits energy in a superconducting chip in bursts that momentarily raise the quasiparticle density across millimetre scales and corrupt many qubits at once. Published monitoring of a large superconducting processor found such events recurring on a timescale of roughly 10 seconds. This paper converts that recurrence into the units of error correction. At a one-microsecond syndrome cycle, bursts arrive every 1.0e7 cycles somewhere on the chip; a distance-twenty-seven patch occupies about 29 per cent of a square-centimetre die, so a thousand-logical-qubit machine spread over many dies should expect a burst inside some patch every few seconds of operation. Correlated bursts are the error model's worst citizen: they violate the independence assumption on which distance-based suppression rests, and the paper budgets the mitigations, gap engineering, phonon trapping, shielding and burst-aware decoding, against the residual rate.

•Nov 15, 2026•0 citations

Quantum error correction below the surface code threshold

Google Quantum AI and Collaborators, Rajeev Acharya, Dmitry A. Abanin +97 more

Quantum error correction [1, 2, 3, 4] provides a path to reach practical quantum computing by combining multiple physical qubits into a logical qubit, where the logical error rate is suppressed exponentially as more qubits are added. However, this exponential suppression only occurs if the physical error rate is below a critical threshold. Here, we present two below-threshold surface code memories on our newest generation of superconducting processors, Willow: a distance-7 code, and a distance-5 code integrated with a real-time decoder. The logical error rate of our larger quantum memory is suppressed by a factor of $$\varLambda \mathrm{=2.14}\pm 0.02$$ when increasing the code distance by two, culminating in a 101-qubit distance-7 code with 0.143% ± 0.003% error per cycle of error correction. This logical memory is also beyond break-even, exceeding its best physical qubit’s lifetime by a factor of $$2.4\pm 0.3$$ . Our system maintains below-threshold performance when decoding in real time, achieving an average decoder latency of 63 μs at distance-5 up to a million cycles, with a cycle time of 1.1 μs. We also run repetition codes up to distance-29 and find that logical performance is limited by rare correlated error events occurring approximately once every hour, or $$3$$ $$\times $$ $${10}^{9}$$ cycles. Our results present device performance that, if scaled, could realize the operational requirements of large scale fault-tolerant quantum algorithms.

arXiv:2408.13687•Nature•Dec 9, 2024•1,314 citations

Exponential Error Confinement in Random Measurement-Enhanced Quantum Circuits

S. N. PRAJAPATI

Abstract For three decades, the fault-tolerance threshold for quantum error correction has remained approximately one percent under realistic noise models. This barrier has defined the trajectory of quantum hardware development, demanding ever-higher gate fidelities and millikelvin operating temperatures. Here we report the discovery of a universal mechanism—exponential error confinement—that elevates the threshold to pth = 25.7±0.3% under circuit-level depolarizing noise. The mechanism arises from the interplay of random unitary scrambling and projective measurements, which together generate an effective attractive potential in error space, inducing Anderson localization of error excitations. We prove rigorously that the logical error rate scales as pL ∝exp−αd2 with code distance d, in contrast to the standard pL ∝ exp(−αd) scaling of all previously known codes. This quadratic exponential suppression reduces the physical qubit overhead to 18 qubits per logical qubit—an 833-fold improvement over the surface code. Wevalidate all claims through large-scale stabilizer simulations of up to 5×103 qubits and 106 random circuit instances, executed entirely on a standard laptop computer. Our results establish that fault-tolerant quantum computation is achievable with near-term hardware operating at error rates exceeding 20%, eliminating the principal barrier to scalable quantum computing.

•May 19, 2026•0 citations

Resilience of the surface code to error bursts

Shi Jie Samuel Tan, Christopher A. Pattison, Matt McEwen +1 more

Quantum error correction works effectively only if the error rate of gate operations is sufficiently low. However, some rare physical mechanisms can cause a temporary increase in the error rate that affects many qubits; examples include ionizing radiation in superconducting hardware and large deviations in the global control of atomic systems. We refer to such rare transient spikes in the gate error rate as error bursts. In this work, we investigate the resilience of the rotated surface code to generic error bursts. We assume that, after appropriate mitigation strategies, the spike in the error rate lasts for only a single syndrome-extraction cycle; we also assume that the enhanced error rate is uniform across the code block. Under these assumptions, and for a circuit-level depolarizing noise model, we perform Monte Carlo simulations to determine the regime in burst error rate and background error rate for which the memory time becomes arbitrarily long as the code block size grows. Our results indicate that suitable hardware mitigation methods combined with standard decoding methods may suffice to protect against transient error bursts in the rotated surface code.

arXiv:2406.18897•Physical Review A•Jan 5, 2026•18 citations

Diffusion-Operator Qubit Bounds and an Error-Burst Qubit Ceiling for Oracle-Based Quantum Search: A Conditional Resource Estimate

Ying Liu

Grover’s algorithm searches an unstructured database of N = 2n items in O( N ) sequential oracle-and-diffusion rounds, a bound already proven optimal. This paper asks a different question: given realistic fault-tolerant hardware constraints, how large an unstructured search can actually be carried to completion? Using a spacetime-volume resource model—accounting for circuit depth, active and idle qubit exposure, and surface-code error suppression—we first derive an oracle-independent upper bound: the diffusion operator alone, together with the round count fixed by the N lower bound, bounds the number of index qubits that a completed search can use to roughly 22 to 53, depending on code distance, a bound that no oracle implementation can loosen, only tighten. We then identify a second, distinct constraint: error bursts, rare correlated fault events with a specific physical origin in cosmic-ray and ambient-radiation exposure, measured directly on current superconducting hardware, are not suppressed by increasing code distance the way ordinary logical error is. This produces an asymptotic ceiling of approximately 50 index qubits that persists regardless of code distance, since the limiting factor becomes a fault mechanism that ordinary error correction does not address. This finding is scoped deliberately: it is a conditional resource estimate tied to one class of hardware’s currently measured, largely unmitigated behavior, not a fundamental limit on quantum computing or a demonstration that this failure mode is physically irreducible—active mitigation research and recent theoretical work indicate the ceiling may relax as hardware and decoding strategies mature. Finally, we show the oracle-independent bound degrades gracefully rather than catastrophically as realistic oracle cost is introduced, via a single-parameter sensitivity analysis.

Journal of Quantum Information Science•Jan 1, 2026•0 citations
Rare correlated error bursts are the reported scaling limit of below-threshold surface-code memo- ries, and existing analyses treat them as events that elevate the error rate across an entire code patch. We show that this assumption fails in the regime every scaled code must enter, where the patch is larger than the burst. There the logical error rate is controlled not by the burst rate but by the tail of the burst footprint distribution: for Pr[Ω > r] ∼ r−α the burst contribution obeys εb(d) ∝ dD−α, so exponential suppression in code distance is replaced by a power law and fault-tolerance over- head becomes polynomial rather than polylogarithmic, with a critical exponent αc = D separating suppressible from unsuppressible environments. Circuit-level simulations over nine code distances, d = 5 to 21, confirm the scaling function and the exponent transfer. We then measure the footprint distribution itself, from 265 radiation bursts resolved on a 60-qubit gap-engineered array, using a forward model of the array that is calibrated against synthetic catalogues before use. The distribu- tion is lognormal, with σ = 0.23 [0.21, 0.26] and median radius 2.9 mm, and is preferred over the best power law by ∆AIC = 71; that comparison is itself calibrated, placing the observed value 9.4σ from the power-law expectation. A lognormal has no tail exponent, so the non-scalable regime α ≤ D is excluded by ten standard deviations and exponential suppression survives in present hardware — refuting a prediction we had made from radiation-transport scaling. The measurement also shows that gap engineering has acted on the shape of the footprint distribution rather than on the burst rate alone, which under the transfer law makes phonon management an asymptotic rather than a constant-factor intervention. Independently, no burst in 5.2 h is large enough to span a distance-7 patch, which accounts for the absence of burst excess in 28 public distance-7 datasets. The largest footprint sampled corresponds to R = 4.9 mm, leaving the distribution beyond that radius — and hence the asymptotics of larger arrays — open to the same measurement on a larger device.
This note applies a single first-order source-loss balance, C dθ/dt = A − ρθ, to systems across many scales — from a planetary surface to a qubit register — and reports, with exact numbers, where the balance reproduces measured values and where it does not. The central quantitative result concerns quantum error correction. Using published surface-code and trapped-ion data (Google Quantum AI, Nature 638, 920 (2024); Quantinuum Helios colour-code results, 2026), the note shows that the logical error per cycle falls by a constant factor Λ only while errors are independent. Correlated bursts act as a common loss channel whose rate does not fall with code distance; the quoted floor is about 3×10⁻¹⁰ per cycle, reached at distance about 46 for Λ = 2.17. For factoring a 2048-bit integer (about 1.4×10³ logical qubits, 6.5×10⁹ Toffoli gates, 4.96 days at a 1 μs cycle), the permitted logical error is 1.7×10⁻¹⁵ per qubit per cycle. At measured suppression factors this requires code distances of about 78 (Λ = 2.17) or about 24 (Λ = 12), while the published design uses 25 — and the correlated-error floor is not removed by increasing distance. This is a quantitative statement about where the scaling of error suppression stops matching the requirements of large computations, not a claim that small-scale quantum devices cannot work. The note also lists, separately and with exact values, the tests the balance fails: quark mass-step prediction (T6), gas/solid classification (T8, p = 0.81), nuclear count with A = 3, 4, prediction SP1 (−11.8% against a 10% threshold), coupling versus number of doors (T10, p = 0.015), width-to-cosmological-scale mapping (q₀ = −0.528 ± 0.011 against 0), human loss-rate estimates (11.1 and 15.3 W/K), critical Q₁₀ = 3.03 against measured 2.0–2.3, Higgs width from the b-quark exit alone (13.7% residual), shell boundaries (13 of 18), and several time constants. Of nine tests where the balance gives a value different from the standard one, one agreed with measurement and eight did not; the rules and thresholds were written before the data were examined. Successful checks include: photosynthesis stage efficiencies (0.487 × 0.094 = 0.0458 against 0.046 reported), a three-stage 2.45 GHz rectenna chain (0.906³ = 0.744), closure of the lunar surface energy budget to a ratio of 1.005, mode quality factors Q = m/Γ for the τ lepton (7.84×10¹¹), the Z boson (36.5) and the Higgs boson (3.38×10⁴), the Higgs lifetime from its width (290.87 fs against 290.3 ± 0.5 fs, a consistency check of standard formulas), and the proton-decay bound ρ < 1.3×10⁻⁴² s⁻¹. The complete theory, tests, scripts and the list of open items are maintained at https://zenodo.org/records/23812417. This note is a standalone publication; readers are invited to report measured values that disagree.
Ionising radiation, from cosmic rays and from trace radioactivity in the laboratory itself, deposits energy in a superconducting chip in bursts that momentarily raise the quasiparticle density across millimetre scales and corrupt many qubits at once. Published monitoring of a large superconducting processor found such events recurring on a timescale of roughly 10 seconds. This paper converts that recurrence into the units of error correction. At a one-microsecond syndrome cycle, bursts arrive every 1.0e7 cycles somewhere on the chip; a distance-twenty-seven patch occupies about 29 per cent of a square-centimetre die, so a thousand-logical-qubit machine spread over many dies should expect a burst inside some patch every few seconds of operation. Correlated bursts are the error model's worst citizen: they violate the independence assumption on which distance-based suppression rests, and the paper budgets the mitigations, gap engineering, phonon trapping, shielding and burst-aware decoding, against the residual rate.
Quantum error correction [1, 2, 3, 4] provides a path to reach practical quantum computing by combining multiple physical qubits into a logical qubit, where the logical error rate is suppressed exponentially as more qubits are added. However, this exponential suppression only occurs if the physical error rate is below a critical threshold. Here, we present two below-threshold surface code memories on our newest generation of superconducting processors, Willow: a distance-7 code, and a distance-5 code integrated with a real-time decoder. The logical error rate of our larger quantum memory is suppressed by a factor of Λ=2.14±0.02\varLambda \mathrm{=2.14}\pm 0.02Λ=2.14±0.02 when increasing the code distance by two, culminating in a 101-qubit distance-7 code with 0.143% ± 0.003% error per cycle of error correction. This logical memory is also beyond break-even, exceeding its best physical qubit’s lifetime by a factor of 2.4±0.32.4\pm 0.32.4±0.3 . Our system maintains below-threshold performance when decoding in real time, achieving an average decoder latency of 63 μs at distance-5 up to a million cycles, with a cycle time of 1.1 μs. We also run repetition codes up to distance-29 and find that logical performance is limited by rare correlated error events occurring approximately once every hour, or 333 ×\times × 109{10}^{9}109 cycles. Our results present device performance that, if scaled, could realize the operational requirements of large scale fault-tolerant quantum algorithms.
Abstract For three decades, the fault-tolerance threshold for quantum error correction has remained approximately one percent under realistic noise models. This barrier has defined the trajectory of quantum hardware development, demanding ever-higher gate fidelities and millikelvin operating temperatures. Here we report the discovery of a universal mechanism—exponential error confinement—that elevates the threshold to pth = 25.7±0.3% under circuit-level depolarizing noise. The mechanism arises from the interplay of random unitary scrambling and projective measurements, which together generate an effective attractive potential in error space, inducing Anderson localization of error excitations. We prove rigorously that the logical error rate scales as pL ∝exp−αd2 with code distance d, in contrast to the standard pL ∝ exp(−αd) scaling of all previously known codes. This quadratic exponential suppression reduces the physical qubit overhead to 18 qubits per logical qubit—an 833-fold improvement over the surface code. Wevalidate all claims through large-scale stabilizer simulations of up to 5×103 qubits and 106 random circuit instances, executed entirely on a standard laptop computer. Our results establish that fault-tolerant quantum computation is achievable with near-term hardware operating at error rates exceeding 20%, eliminating the principal barrier to scalable quantum computing.
Quantum error correction works effectively only if the error rate of gate operations is sufficiently low. However, some rare physical mechanisms can cause a temporary increase in the error rate that affects many qubits; examples include ionizing radiation in superconducting hardware and large deviations in the global control of atomic systems. We refer to such rare transient spikes in the gate error rate as error bursts. In this work, we investigate the resilience of the rotated surface code to generic error bursts. We assume that, after appropriate mitigation strategies, the spike in the error rate lasts for only a single syndrome-extraction cycle; we also assume that the enhanced error rate is uniform across the code block. Under these assumptions, and for a circuit-level depolarizing noise model, we perform Monte Carlo simulations to determine the regime in burst error rate and background error rate for which the memory time becomes arbitrarily long as the code block size grows. Our results indicate that suitable hardware mitigation methods combined with standard decoding methods may suffice to protect against transient error bursts in the rotated surface code.
Grover’s algorithm searches an unstructured database of N = 2n items in O( N ) sequential oracle-and-diffusion rounds, a bound already proven optimal. This paper asks a different question: given realistic fault-tolerant hardware constraints, how large an unstructured search can actually be carried to completion? Using a spacetime-volume resource model—accounting for circuit depth, active and idle qubit exposure, and surface-code error suppression—we first derive an oracle-independent upper bound: the diffusion operator alone, together with the round count fixed by the N lower bound, bounds the number of index qubits that a completed search can use to roughly 22 to 53, depending on code distance, a bound that no oracle implementation can loosen, only tighten. We then identify a second, distinct constraint: error bursts, rare correlated fault events with a specific physical origin in cosmic-ray and ambient-radiation exposure, measured directly on current superconducting hardware, are not suppressed by increasing code distance the way ordinary logical error is. This produces an asymptotic ceiling of approximately 50 index qubits that persists regardless of code distance, since the limiting factor becomes a fault mechanism that ordinary error correction does not address. This finding is scoped deliberately: it is a conditional resource estimate tied to one class of hardware’s currently measured, largely unmitigated behavior, not a fundamental limit on quantum computing or a demonstration that this failure mode is physically irreducible—active mitigation research and recent theoretical work indicate the ceiling may relax as hardware and decoding strategies mature. Finally, we show the oracle-independent bound degrades gracefully rather than catastrophically as realistic oracle cost is introduced, via a single-parameter sensitivity analysis.