CoChat

Quantum computing with realistically noisy devices

Open access
Knill, E. · 2005 · Nature

The DOI matches its Crossref record; 60 of the 60 references it cites were found in a registry or an index; 16 of the 55 sources its claims cite were checked for support.

Abstract

In theory, quantum computers offer a means of solving problems that would be intractable on conventional computers. Assuming that a quantum computer could be constructed, it would in practice be required to function with noisy devices called ‘gates’. These gates cause decoherence of the fragile quantum states that are central to the computer's operation. The goal of so-called ‘fault-tolerant quantum computing’ is therefore to compute accurately even when the error probability per gate (EPG) is high. Here we report a simple architecture for fault-tolerant quantum computing, providing evidence that accurate quantum computing is possible for EPGs as high as three per cent. Such EPGs have been experimentally demonstrated, but to avoid excessive resource overheads required by the necessary architecture, lower EPGs are needed. Assuming the availability of quantum resources comparable to the digital resources available in today's computers, we show that non-trivial quantum computations at EPGs of as high as one per cent could be implemented.

Overview

CoChat Summary

TL;DR

Knill presents the C4/C6 fault-tolerant architecture and evidence that scalable quantum computing may be possible at gate error probabilities above 3% per gate, though resource costs are excessive at such high error rates. At lower error probabilities around 1%, non-trivial computations could be implemented with resources comparable in scale to classical computing.

Key findings

  • Postselected C4/C6 computing: heuristic analysis gave thresholds above γ = 0.03; direct simulation up to two levels suggested a threshold above γ = 0.06 by extrapolation, and at γ = 0.03 logical preparation and measurement errors were consistent with being below threshold.
  • Error-correcting C4/C6 architecture: threshold above 0.01. At γ = 0.01, logical CNOT detected errors were 2.4 ± 0.0 × 10⁻² at level 3 and 2.4 ± 1.0 × 10⁻³ at level 4; conditional errors were 6.4 ± 0.6 × 10⁻⁴ at level 3 and 0.0 ± 4.4 × 10⁻⁴ at level 4. State-injection conditional logical error was 8.6 ± 0.6 × 10⁻³ at level 3 and 1.1 ± 0.1 × 10⁻² at level 4.
  • Resource example: at γ = 0.01, a computation with 100 logical qubits and up to 1000 |π/8⟩-preparations requires 1.23 × 10¹⁴ physical CNOTs, with conditional probability of incorrect output ≈ 0.02.
  • Comparison at γ = 10⁻⁴: Steane’s architecture achieves effective logical error ≈ 7 × 10⁻¹² using ≈ 420 physical CNOTs per qubit per gate; C4/C6 achieves detected errors of 5.5 × 10⁻⁹ at level 3 or 6 × 10⁻¹⁴ at level 4 using ≈ 2100 or 2.6 × 10⁴ physical CNOTs per qubit, respectively.

Methods

  • Architecture: concatenated error-detecting stabilizer codes C4 (four-qubit code) and C6 (three qubit-pair code), with error-correcting teleportation, transversal Clifford gates, logical Bell state preparation, postselected computing, and state injection/purification for universality.
  • Error model: independent Pauli errors with a depolarizing model per gate, setting e_c = γ, e_m = e_p = 4γ/15, and e_h = 4γ/5; additional assumptions include no classical computation error or speed constraint and arbitrary two-qubit gates without delay or additional error.
  • Analysis: computer-assisted heuristic threshold analysis, direct stabilizer/Clifford simulation in Octave up to multiple encoding levels, and resource estimates using naive models with Fibonacci-sequence error scaling.

Limitations

  • Stated: the error model is idealized and does not match physical qubits and gates, notably coherent rotations, correlations between nearby errors, and leakage; memory error and measurement time are not specified; gates other than measurements are assumed to take the same time; classical computation is assumed error-free and unconstrained; arbitrary two-qubit gates are assumed available without delay or additional error, which the paper calls unrealistic.
  • Stated: high-EPG resource requirements are excessive, postselection overheads are extreme, simulations do not constitute full statistical proof because of implicit correlations from the random-number generator, resource models are naive and most reliable for levels < 4, and the threshold ≈ 0.028 obtained from recursion is conjectural.
  • Evident: the high-threshold claims rely on heuristic analysis, simulation, and extrapolation rather than mathematical proof, and the resource estimates at high γ depend on unoptimized |π/8⟩-preparation and may be impractical.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/nature03350

Cites

60 of 60 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

10 Valid · 22 Not valid · 20 Not checked · 203 Own results and interpretations

147 more claims · 6 with their cited sources checked

Create a free account to see all 255 claims and to verify the 39 references still open.

See all claims — free account
  • It may state two separate claims as one. 67
  • Probably not a substantial claim. 47
  • The quote only partly states this claim. 4

Create a free account to see these claims.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

3 of 3

Citations

Cites

Show all 60 references

Cited by

Items 201–300 of 1,003

Metadata

Authors
Knill, E.
Affiliations
National Institute of Standards and Technology (US)
Corresponding author
Knill, E.
Venue
Nature
Published
2005-03
Volume
434
Issue
7029
Pages
39-44
ISSN
  • 0028-0836 (ISSN-L)
  • 1476-4687
OpenAlex
W2032147770
Cited by
1,184
References
54
Citations by year
2026: 46 · 2025: 102 · 2024: 78 · 2023: 66 · 2022: 69 · 2021: 81 · 2020: 74 · 2019: 48 · 2018: 40 · 2017: 49 · 2016: 52 · 2015: 39 · 2014: 44 · 2013: 41 · 2012: 36
FWCI
54.93
Type
journal article
Language
en
Open access
Licence
www.springer.com/tdm (tdm)
Locations
  • Nature (publishedVersion)
  • PubMed (publishedVersion)
  • arXiv (Cornell University) (submittedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
  • CERN Document Server (European Organization for Nuclear Research) (submittedVersion)
  • arXiv (Cornell University)
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › Quantum Computing Algorithms and Architecture
  • Physical Sciences › Physics and Astronomy › Atomic and Molecular Physics, and Optics › Quantum Information and Cryptography
  • Physical Sciences › Physics and Astronomy › Atomic and Molecular Physics, and Optics › Quantum many-body systems
Keywords
fault-tolerant quantum computing, decoherence, quantum computing architecture, resource overhead, quantum error correction

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-07, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-07 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (qwen3.8-omni-flash and gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.