Cited by 704Quantum Computing Algorithms and ArchitectureQuantum Information and CryptographyQuantum many-body systems

The DOI matches its Crossref record; 45 of the 45 references it cites were found in a registry or an index; none of the 28 sources its claims cite is checked for support yet.

Validating quantum computers using randomized model circuits

Open access

Abstract

We introduce a single-number metric, quantum volume, that can be measured using a concrete protocol on near-term quantum computers of modest size ($n\ensuremath{\lesssim}50$), and measure it on several state-of-the-art transmon devices, finding values as high as 16. The quantum volume is linked to system error rates, and is empirically reduced by uncontrolled interactions within the system. It quantifies the largest random circuit of equal width and depth that the computer successfully implements. Quantum computing systems with high-fidelity operations, high connectivity, large calibrated gate sets, and circuit rewriting toolchains are expected to have higher quantum volumes. The quantum volume is a pragmatic way to measure and compare progress toward improved system-wide gate error rates for near-term quantum computation and error-correction experiments.

Overview

AI snapshot

Written from the paper’s full text.

TL;DR

The paper introduces quantum volume, a single-number metric measured by random model circuits of equal width and depth using heavy output generation, and reports measurements on IBM Q transmon devices with values as high as 16. It links quantum volume to system error rates, connectivity, and circuit optimization, and provides simulations for how it scales with two-qubit error and coupling map.

Key findings

  • Quantum volume is defined as \(V_Q = 2^{\min(d,m)}\), measuring the largest square-shaped model circuit a quantum computer successfully implements; measured values reach as high as 16 (\(\log_2 V_Q = 4\)). On the tested devices, \(\log_2 V_Q = 3\) for Tokyo, \(\log_2 V_Q = 2\) for Tenerife, and \(\log_2 V_Q < 2\) for Melbourne.
  • Heavy output probabilities for \(m=d\) circuits include Tokyo \(m=d=2\): 0.718(0.006), \(m=d=3\): 0.682(0.002)*, \(m=d=4\): 0.614(0.003); Johannesburg \(m=d=4\): 0.664(0.004), and optimized \(m=d=4\): 0.699(0.001), against the 2/3 threshold.
  • Circuit optimization improves performance: standard Qiskit Terra transpiler produced 28.1 CX gates on average with \(\hat h_d = 0.614(0.003)\); KAK reduced this to 21.0 CX gates with \(\hat h_d = 0.632(0.005)\); approximate SU(4) with 1% CX error used 17.7 CX gates and achieved \(\hat h_d = 0.649(0.005)\). Noisy depolarizing simulations substantially over-estimated experimental performance.
  • Connectivity simulations show all-to-all connectivity has an advantage: for \(\log_2 V_Q = 12\), the maximum permissible two-qubit error was 0.0032 for all-to-all, 0.0015 for square grid, and 0.0014 for loop. Measurement error reduces permissible two-qubit error for square grid at \(\log_2 V_Q = 12\) from 0.0015 at 0% to 0.00125 at 1% and 0.0002 at 5%.

Methods

  • The protocol uses random model circuits of width \(m\) and depth \(d\), built from layers of random permutations and Haar-random SU(4) two-qubit gates, compiled to the target device’s gate set. Success is evaluated via the heavy output generation problem, requiring heavy output probability \(h_d > 2/3\), with quantum volume defined as \(V_Q = 2^{\min(d,m)}\).
  • Experiments were run on four IBM Q devices: 5-qubit Tenerife, 16-qubit Melbourne, 20-qubit Tokyo, and 20-qubit Johannesburg. For each \(m\), 200 circuits were run for \(d=m\) with \(m \in \{2,3,4\}\); repeats used 5000 circuits for \(m=2\) on Tenerife and \(m=3\) on Tokyo, and 1000 circuits for optimized Johannesburg experiments. Results were compared with depolarizing noise simulations.
  • Simulations of all-to-all, square-grid, and loop coupling maps up to 12 qubits estimated the two-qubit gate error required for \(\log_2 V_Q = 4, 6, 8, 12\), assuming single-qubit error equal to \(\epsilon/10\) and measurement errors of 0%, 1%, and 5% for the square grid.

Limitations

  • The paper states that noisy simulations substantially over-estimate performance, that classical computation of the heavy set scales exponentially with \(m\), and that for error rates as low as \(10^{-4}\) model circuits may still be small enough to simulate classically; lower error rates may require new or modified volume metrics. It also notes that the LOCO optimization did not significantly improve heavy output generation for small numbers of qubits, and that connectivity comparisons involve small total qubit numbers, with asymptotic loop-versus-grid differences potentially larger.
  • Evident from the text: experimental validation is limited to four IBM Q transmon devices and small square circuits up to \(m=d=4\) (with \(m=d=5\) on Johannesburg), so larger-device and architecture conclusions rely on simulations. Device error rates were measured one day before the experiments and can fluctuate during the experiment timescale.

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

Integrity signals

Registry

Crossref · 10.1103/physreva.100.032328

Cites

45 of 45 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

0 of 28 cited sources checked · 0 supported · 0 flagged

Show all 35 claims

  • It may state two separate claims as one. 16
  • Probably not a substantial claim. 23
35 of 35
  • The authors state this as an assumption.

    The approximation of achievable model circuit depth assumes independent stochastic errors.

    “We approximate the achievable model circuit depth ˜ d ( m ) by assuming independent stochastic errors, so that the computation fails with high probability when the model circuit volume (width times depth) satisfies”
  • The authors state this as an assumption.

    In simulations for Table III, the single-qubit gate error is equal to ε/10.

    “assuming the single-qubit gate error is equal to ϵ/ 10 (Table III).”
  • The authors state this as an assumption.

    In the Table III estimates, measurements are perfect.

    “The estimates are based on simulations using a depolarizing noise model with two-qubit error ϵ as given, single-qubit error equal to ϵ/ 10, and perfect measurements.”
3 of 3

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?

Citations

Cites

Flagged first, then by citations

Show all 44 references

Research works printed but not linked: 2

  1. S. Aaronson, 2017, Proceedings of the 32nd Computational Complexity Conference (CCC '17)
  2. T. Häner, 2016, Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis (SC '16)

Cited by

Sorted by citations

Items 201–300 of 701

Metadata

Authors
Cross, A. W., Bishop, L. S., Sheldon, S., Nation, P. D., & Gambetta, J. M.
Affiliations
IBM Research - Thomas J. Watson Research Center (US)
Venue
Physical Review A
Published
2019-09-20
Volume
100
Issue
3
ISSN
  • 0556-2791 (ISSN-L)
  • 2469-9926
  • 2469-9934
OpenAlex
W2903516100
Cited by
704
References
44
Citations by year
2026: 77 · 2025: 94 · 2024: 106 · 2023: 115 · 2022: 109 · 2021: 117 · 2020: 69 · 2019: 17
FWCI
35.58
Type
journal article
Language
en
Open access
Locations
  • Physical Review A (publishedVersion)
  • arXiv (Cornell University) (publishedVersion)
  • arXiv (Cornell University)
Funders
Army Research Office (W911NF-14-1-0124)
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
quantum volume, transmon qubits, gate error rates, near-term quantum devices, 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.

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

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

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