Pith. sign in

REVIEW 2 major objections 4 minor 28 references

Diffusive Speed Limits for U(1)-Covariant Quantum Error Correction

T0 review · 2 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Charge conservation forces a quadratic time cost on local quantum encoding and sets an $n^{-1/2}$ accuracy floor that random charge-symmetric codes attain.

desk verdict Strong static optimality results for U(1)-covariant erasure codes, with an Ω(n²) dynamical bound that is real but narrower than the abstract suggests. read the letter →

arxiv 2608.04953 v1 pith:GAMM4IXH submitted 2026-08-05 quant-ph math-phmath.MP

Jianqi Sheng This is my paper
classification quant-phmath-phmath.MP
keywords quantumerrorcorrectionU(1)-covariantcodesflaggederasuredecouplingchargeconservationdiffusivespeedlimitrandombrickworkcircuitssymmetricsimpleexclusionprocess
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper establishes that particle-number conservation is not only an obstruction to exact quantum error correction but an operational speed limit for approximate correction. The setting is one logical qubit stored in the two adjacent charge sectors of an odd $n$-cell chain, with a flagged erasure meaning a known-location loss whose erased cells stay inaccessible. For every such encoder, the paper proves a universal floor on the flagged-erasure decoupling error under independent lost-cell patterns: $D_{n,p}(V)\ge M_{\mathrm{eq}}(n,p)=\sqrt{p/(2\pi(1-p))}\,n^{-1/2}+o(n^{-1/2})$, and charge-Haar random codes reach this floor up to exponentially small corrections below half erasure. It then proves a dynamical counterpart: for the canonical boundary code driven by the unbiased local-Haar brickwork ensemble of number-conserving gates, the ensemble-averaged error is at least $p/[48\sqrt{2\sqrt{t-1}+3}]$, so reaching even a fixed factor of the optimal $n^{-1/2}$ endpoint requires $\Omega(n^2)$ complete cycles. The practical point is that symmetry-dictated diffusion, not just scrambling, sets how fast a quantum memory can be prepared.

What carries the argument

The load-bearing objects are the complementary-channel decoupling error $D_E(V)=\frac12\|\rho_{RE}-\rho_R\otimes\rho_E\|_1$, the erased-charge hypergeometric witness, and the diffusive random-interchange representation of the averaged circuit. Because exact erasure correction is equivalent to $D_E(V)=0$, lower bounds on $D_E$ are bounds on any recovery map. The erased-charge witness conditions on the total occupation of $E$; after binomial averaging it becomes $M_{\mathrm{eq}}(n,p)=\sqrt{p/(2\pi(1-p))}n^{-1/2}+o(n^{-1/2})$, and it is independent of the amplitudes inside the charge sectors. For the dynamics, averaging the local-Haar brickwork gates of Eq. (C10) makes the one-copy charge profile evolve by alternating adjacent averaging, the deterministic limit of the symmetric simple exclusion process; the distinguished logical particle performs a reflected random walk of width $\sqrt{t}$, and a negative-correlation counting argument on a window of size $\Theta(\sqrt{t})$ converts the $t^{1/4}$ count fluctuations into the lower bound of Theorem 10. The remaining upper-bound problem is organized by a gate-resolved connected-moment expansion and the exact low-support Pauli-leakage identity $Q_{n,p,b}(B)=\frac14\sum_{\alpha=1}^{3}\sum_{P\in\mathcal{P}_n}\vartheta_{n,p,b}(\mathrm{wt}\,P)|b_{\alpha,P}(B)|^2$, which shows that the unresolved fluctuation is a source-restricted operator-spreading estimate rather than full two-copy mixing.

What would settle it

Simulate the canonical boundary code under the local-Haar brickwork ensemble for large odd $n$ (say $n=101$) and measure the ensemble-averaged flagged count-statistic gap $D_{\mathrm{KS}}$ at depths $t=c n^2$ with $c<1$; observing $D_{\mathrm{KS}}$ decay faster than $\Theta_p(t^{-1/4})$ while $t/n^2\to0$, or full decoupling error reaching a fixed factor of $M_{\mathrm{eq}}(n,p)$ at subquadratic depth, would refute the claimed speed limit.

Watch

Extended reading notes

Core claim

The central discovery is a sharp static optimum paired with a diffusive dynamic cost for $U(1)$-covariant encoding of one logical qubit in adjacent charge sectors of $n=2q+1$ cells. For any isometry whose codewords live in sectors $q$ and $q+1$, the erased total charge is a universal witness: uniformly averaged over erasure sets of fixed size $k=\alpha n$, the decoupling error is at least $D_{n,k}$, and averaging over Bernoulli-pattern erasures gives $D_{n,p}(V)\ge M_{\mathrm{eq}}(n,p)$. Charge-Haar codewords saturate this floor up to $e^{-\Omega(n)}$ corrections for $p<1/2$, yielding the exact $n^{-1/2}$ extensive-erasure law and the sharp $(0,3/8,3/4)$ one-half-erasure transition. Dynamically, the one-copy averaged charge profile of the canonical boundary code under the unbiased local-Haar brickwork ensemble evolves by alternating nearest-neighbor averaging, i.e. as the random-interchange process, so the logical charge forms a cloud of width $\sqrt{t}$ and the erased background count in a diffusive window fluctuates on the $t^{1/4}$ scale; the paper shows this forces the ensemble-averaged decoupling error to be at least $p/[48\sqrt{2\sqrt{t-1}+3}]$. Consequently, no circuit in this ensemble can reach a fixed fraction of the optimum before $\Omega(n^2)$ cycles, while the matching $O(n^2)$ upper bound remains open and is reduced to a source-restricted low-support operator-spreading estimate.

Load-bearing premise

The quadratic lower bound rests on the unbiased local-Haar brickwork gate ensemble, whose averaged one-copy dynamics is exactly diffusive; if the gates are biased, deterministic, or drawn from a different structured number-conserving ensemble, the $t^{-1/4}$ count-statistic bound does not follow and the $\Omega(n^2)$ obstruction could fail.

Editorial extensions

If this is right

  • Every adjacent-charge encoder of one logical qubit has flagged decoupling error at least $\sim\sqrt{p/(2\pi(1-p))}\,n^{-1/2}$ under iid erasure, so no choice of amplitudes or phases inside the two charge sectors can beat the erased-charge floor.
  • Charge-Haar codes attain that floor up to exponentially small corrections below half erasure, so the exact $n^{-1/2}$ law and the $(0,3/8,3/4)$ transition describe the information-theoretic optimum of the adjacent-charge model.
  • Under the unbiased local-Haar brickwork ensemble, the canonical boundary code needs $\Omega(n^2)$ complete cycles to reach any fixed factor of the optimal accuracy; this is a time-to-optimality obstruction, not a comparison with a convenient benchmark.
  • The complete flagged output is exactly rigid (total variation $p$) until holes from the background diffuse into the distinguished charge cloud, so faster full-channel decay is impossible in the pre-boundary causal regime.
  • The classical insertion-tolerance part of the error reaches its equilibrium $O(n^{-1/2})$ scale in $O(n^3)$ cycles, and the remaining full-channel bound is exactly the low-support operator-spreading estimate that would close the formation time at $\Theta(n^2)$.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Read as a transport statement, the result suggests that any one-dimensional number-conserving encoder in the same diffusive universality class inherits the quadratic preparation cost; Appendix E already removes the special role of the canonical background, so the speed limit is plausibly generic within that class.
  • A cheap experimental probe is to measure the two occupation-count cumulative distributions on a diffusive window under the stated gate ensemble: the Kolmogorov–Smirnov gap should decay as $\Theta_p(t^{-1/4})$ while $t/n^2\to0$, which is visible without full process tomography.
  • If the missing balanced two-copy estimate is supplied, the formation time should lock in at $\Theta(n^2)$, and diffusion would then play the role for covariant encoding that causality plays for local circuit depth in general scrambling bounds.
  • The static floor is proved for one logical qubit in adjacent charge sectors; the natural next question is whether multi-qubit or separated-charge encoders obey analogous polynomial floors with constants set by the charge gap, rather than escaping the $n^{-1/2}$ scaling altogether.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 4 minor

Summary. The paper studies U(1)-covariant quantum error correction for one logical qubit encoded in adjacent charge sectors of n cells under flagged erasure. It proves a universal static lower bound on the decoupling error for every adjacent-charge encoder, shows that charge-Haar codes attain this bound up to exponentially small corrections below half erasure (exact n^{-1/2} extensive-erasure law and a sharp (0,3/8,3/4) transition at half erasure), and establishes an ensemble-averaged lower bound of order t^{-1/4} for the canonical boundary code under the unbiased local-Haar brickwork circuit ensemble, giving an Ω(n^2) cycle requirement to reach the optimal n^{-1/2} scale. It also proves an unconditional O(n^3) mixing bound for the classical insertion-tolerance component and reduces the remaining full-channel upper bound to a source-restricted low-support operator-spreading problem, left open in Problem O1.

Significance. The static results, if correct, provide the first exact extensive-erasure decoupling law for symmetry-constrained codes and pinpoint the information-theoretic optimum of the adjacent-charge model. The derivations are exact combinatorial and central-limit arguments with no fitted parameters; Theorem 7's universal floor and Theorem 8's worst-pattern concentration are particularly clean. The dynamic lower bound is rigorous for its stated hypotheses, and the paper is exemplary in labeling conditional milestones (Theorems 16-17) and open scope (Problem O2, Remark 14). However, the advertised diffusion speed limit is not a no-go for arbitrary local number-conserving circuits or for individual circuits; it is an ensemble- and encoding-class-specific statement. That distinction is important for the paper's main claim.

major comments (2)
  1. [Abstract; Section IV] The abstract's claim that 'for local number-conserving brickwork circuits, diffusion of the logical charge enforces an Ω(n^2) encoding-time lower bound' and Section IV's statement that 'reaching even a fixed factor of the optimum therefore requires Ω(n^2) cycles' overstate the proved result. Theorem 10 (Eq. C192) and Corollary 8 bound the ensemble-averaged error E_{V,E}D_E(V) under the specific unbiased local-Haar brickwork measure of Eq. (C10), and Appendix E extends this only to computational-basis product encodings sharing a deterministic background and differing by one distinguished particle. Multi-particle and multi-hole encodings are explicitly open (Problem O2, Remark 14), and the proof does not cover other gate measures or structured circuit families. The abstract and conclusion should be qualified accordingly.
  2. [Theorem 10; Corollary 8] The lower bound is an average over circuit realizations and erasure patterns. As an average, it does not exclude the existence of individual depth-o(n^2) circuits in the support of the measure with substantially lower error, nor does it constrain deterministic or structured encoder families. Consequently the operational statement that 'reaching even a fixed factor of the optimum requires Ω(n^2) cycles' is justified only for the ensemble average, not for every local number-conserving circuit. The paper's own Remark 11 and Problem O2 acknowledge this; the abstract and Section IV should carry the same qualification, since the current wording presents the result as a general diffusive speed limit.
minor comments (4)
  1. [Section III] The phrase 'iid Bernoulli-perasure patterns' should read 'iid Bernoulli-erasure patterns' or 'Bernoulli(p) erasure patterns'; 'perasure' appears to be a typo.
  2. [Remarks 10 and 11] The expression 'exp[Θ(n16n)]' in Remark 11 is garbled; it should read 'exp[Θ(n 16^n)]' or similar, and 'O(n5 + n4 log(1/ε))' in Remark 10 should be typeset with superscripts.
  3. [Introduction / Section I] The introduction presents the dynamic lower bound for the canonical boundary code without mentioning until Appendix E that the result extends only to single-distinguished-particle product encodings; adding a sentence in Section I or III would help readers see the scope of the claim.
  4. [Eq. (C10)] In Eq. (C10) the gate ensemble is defined with an independent phase e^{iα} per gate, but this independence is not stated explicitly; it would improve clarity to say so.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the static floor and dynamic lower bound are derived from explicit independent calculations, not from fitted inputs or self-citations.

full rationale

The paper's central claims do not reduce to their inputs by construction. The static universal floor is obtained in Theorem 7 by applying a fixed charge measurement and classical data processing to an arbitrary adjacent-charge encoder, giving D_{n,k}(V) >= D_{n,k} and D_{n,p}(V) >= M_eq(n,p) (Eqs. C77-C79); this bound depends only on the erased total occupation, not on any code amplitude. Charge-Haar optimality is then established independently in Theorem 6 and Corollaries 3-4 by exact hypergeometric calculations of the charge-Haar expectation, so the optimality statement does not presuppose the lower bound. The dynamic Omega(n^2) speed limit is likewise derived, not fitted: Theorem 10 starts from the explicit unbiased local-Haar brickwork ensemble of Eq. (C10), uses the exact random-interchange coupling of Eq. (C198), proves the required negative-correlation property by induction (Eqs. C202-C204), and lower-bounds the quantum decoupling error through a classical count statistic via Eq. (C210-C211). No free parameter is introduced, and the lower bound is not the same object as the n^{-1/2} endpoint it is compared against. The O(n^3) classical mixing bound in Theorem 18 relies on the external Caputo-Liggett-Richthammer spectral-gap theorem and the detectability lemma, not on a self-citation chain. The paper contains no self-citations by the present author. The only qualification is that the abstract's broad phrasing "diffusion enforces an Omega(n^2) encoding-time lower bound" is stronger than the proved ensemble-averaged statement for single-distinguished-particle product encodings under one specified unbiased local-Haar measure, a scope restriction the paper itself flags explicitly in Problem O2 and Remark 14; this is an overreach of presentation, not circular reasoning.

Assumptions & free parameters 0 free parameters · 10 assumptions · 0 invented entities

No fitted constants are introduced; the model assumptions are explicit and flagged; the paper introduces no new particles, forces, dimensions, or other invented entities.

assumptions (10)
  • domain assumption Strict U(1) covariance with a fixed background charge Q0 (Eq. 2).
    Defines the resource model; the entire paper operates inside this constraint.
  • domain assumption Flagged erasure: loss locations are known and erased cells are inaccessible, with pattern-conditioned number-conserving recovery.
    The classical flag is part of the channel; all error metrics and witnesses use it.
  • domain assumption Logical qubit encoded in adjacent charge sectors H_q and H_{q+1} for odd n.
    The universal floor and charge-Haar optimality are proved only for this encoding class.
  • domain assumption Unbiased local-Haar brickwork gate ensemble (Eq. C10): each bond gate is 1 xor U_1 xor e^{i alpha} with U_1 Haar and alpha uniform.
    The dynamic lower bound is for this measure; its one-copy average is exactly the SSEP random-interchange process.
  • domain assumption Single-distinguished-particle product encodings: the two codewords share a deterministic background and differ by one occupied site (Appendix E).
    The generalized quadratic speed limit is proved only within this class; multi-particle differences are left open in Remark 14.
  • standard math Caputo-Liggett-Richthammer spectral gap theorem for the interchange process on a path.
    Used in Lemma 20 and Theorem 18 to obtain the exact Omega(n^{-2}) gap of the tagged exclusion process.
  • standard math Strong Rayleigh property and Gaussian concentration for homogeneous strong Rayleigh measures.
    Used in Theorem 13 and Lemma 22 for negative correlation and purity burn-in.
  • standard math Detectability lemma for frustration-free local Hamiltonians on a line.
    Used in Lemma 21 to convert the Hamiltonian gap into a brickwork-cycle compression gap.
  • domain assumption Generated-group second moment equals the charge-Haar moment for the local U(1) gate group, cited from Refs. [4,5].
    Used in Theorem 5 to identify the fixed space with charge-Haar; not used in the main static or lower-bound results.
  • ad hoc to paper Premise Eq. (C388), the CQA spectral-gap assumption of Ref. [6], is taken as an explicitly unproved premise in conditional Theorem 17.
    The paper audits Ref. [6] and flags a correction; this premise is not used in any unconditional conclusion.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Diffusive Speed Limits for U(1)-Covariant Quantum Error Correction." pith.science (2026). https://pith.science/paper/GAMM4IXH

@misc{pith2026260804953,
  author       = {Pith},
  title        = {Pith review of: Diffusive Speed Limits for U(1)-Covariant Quantum Error Correction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GAMM4IXH}},
  note         = {Machine review of arXiv:2608.04953}
}
abstract

Fast preparation of quantum error-correcting codes is essential for scalable quantum memories, but geometric locality and $U(1)$ charge conservation impose an unavoidable transport constraint. We combine exact complementary-channel geometry, charge-sector Haar analysis, and a gate-resolved connected-moment expansion to study one-dimensional covariant encoders under flagged erasure. Charge-Haar codes attain the universal adjacent-charge lower bound up to exponentially small corrections, yielding an exact $n^{-1/2}$ extensive-erasure law and a sharp half-erasure transition. For local number-conserving brickwork circuits, diffusion of the logical charge enforces an $\Omega(n^2)$ encoding-time lower bound; we also prove an $O(n^3)$ mixing bound for the classical component and reduce the remaining full-channel upper bound to a source-restricted low-support operator-spreading problem. These results identify diffusion as an operational limit on symmetry-constrained quantum coding and establish a route to its exact formation time.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

28 extracted references · 22 canonical work pages

  1. [1]

    Conditional milestones (premises open) Theorem 16(Conditional diffusion-gap achievability (premise unproved)).Fix0< p <1/2. Suppose the largest nontrivial singular valueλ n of one complete bal- anced two-copy brickwork moment cycle restricted to the task module satisfies the unproved bound 1−λ n ≥ c n2 (C383) for a constantc >0(i.e. the task-sector gap ma...

    work page

  2. [2]

    Applying it to every source after the first cycle and repeating the telescoping argument of Lemma 13 proves Eq

    Since∥C 11∥2 = √ 12, the second identity follows. Applying it to every source after the first cycle and repeating the telescoping argument of Lemma 13 proves Eq. (C310). Lemma 16(Collision Casimir and exact erasure extrac- tion).On the charge-one subspace of a physical bond, de- fine Jx =|01⟩ ⟨10|+|10⟩ ⟨01|,(C314) Jy =−i|01⟩ ⟨10|+i|10⟩ ⟨01|,(C315) Jz =|01...

    work page

  3. [3]

    The logical product code- words are |ψ0⟩=|B 0⟩,|ψ 1⟩=|B 0 ∪ {s0}⟩.(E1) Under the common random interchange permutation, Bt = Πt(B0), S t = Πt(s0), C t =B t ∪ {St}

    Setup and the distinguished-particle coupling LetB 0 ⊂[n] be an arbitrary deterministic set ofqoc- cupied sites and lets 0 /∈B0. The logical product code- words are |ψ0⟩=|B 0⟩,|ψ 1⟩=|B 0 ∪ {s0}⟩.(E1) Under the common random interchange permutation, Bt = Πt(B0), S t = Πt(s0), C t =B t ∪ {St}. (E2) This exact coupling is the only encoding-specific input nee...

    work page

  4. [4]

    (C204) for the canonical initialization

    Universality of negative correlation The proof of the quadratic speed limit in Theorem 10 hinges on the pairwise negative correlation of the back- ground occupation indicatorsη i =1 {i∈Bt}, established in Eq. (C204) for the canonical initialization. This property is in fact universal. Lemma 22(Universal negative correlation).LetB 0 ⊆ [n]beanydeterministic...

    work page

  5. [5]

    Define ¯δ(s0) i (t) = [(OE) tes0 ]i = Pr{St =i|S 0 =s 0},(E4) thet-cycle position distribution of a single particle un- dergoing the lazy reflected random walk on{0,

    Generalized diffusive window and speed limit The distinguished-particle profile generalizes directly. Define ¯δ(s0) i (t) = [(OE) tes0 ]i = Pr{St =i|S 0 =s 0},(E4) thet-cycle position distribution of a single particle un- dergoing the lazy reflected random walk on{0, . . . , n−1} with initial positions 0. The cell-walk analysis of Theo- rem 9 applies with...

    work page

  6. [6]

    Generalized causal plateau The causal plateau of Theorem 11 also generalizes: its length is controlled by theprotected radius—the distance from the distinguished site to the nearest hole inC 0. Theorem 20(Generalized causal plateau).For a single- distinguished-particle product encodingC 0 =B 0 ∪ {s0}, define the protected radius r0 = min d(s0, i) :i /∈C0 ...

    work page

  7. [7]

    Eastin and E

    B. Eastin and E. Knill, Physical Review Letters102, 110502 (2009)

    2009

  8. [8]

    Faist, S

    P. Faist, S. Nezami, V. V. Albert, G. Salton, F. Pastawski, P. Hayden, and J. Preskill, Physical Re- view X10, 041018 (2020), arXiv:1902.07714 [quant-ph]

    arXiv 2020

Show all 28 references
  1. [9]

    Kong and Z.-W

    L. Kong and Z.-W. Liu, PRX Quantum3, 020314 (2022), arXiv:2112.01498 [quant-ph]

    2022 arXiv

  2. [10]

    S. N. Hearth, M. O. Flynn, A. Chandran, and C. R. Laumann, Physical Review X15, 021022 (2025)

    2025

  3. [11]

    Mitsuhashi, R

    Y. Mitsuhashi, R. Suzuki, T. Soejima, and N. Yoshioka, Physical Review Letters134, 180404 (2025)

    2025

  4. [12]

    Z. Li, H. Zheng, and Z.-W. Liu, Efficient quan- tum pseudorandomness under conservation laws (2024), arXiv:2411.04893 [quant-ph]

    2024 arXiv

  5. [13]

    Knill and R

    E. Knill and R. Laflamme, Physical Review Letters84, 2525 (2000)

    2000

  6. [14]

    Hayden and J

    P. Hayden and J. Preskill, Journal of High Energy Physics2007, 120 (2007)

    2007

  7. [15]

    T. M. Liggett,Interacting Particle Systems(Springer, New York, 1985). 40

    1985

  8. [16]

    Aldous and P

    D. Aldous and P. Diaconis, American Mathematical Monthly93, 333 (1986)

    1986

  9. [17]

    Spohn,Large Scale Dynamics of Interacting Particles (Springer, Berlin, 1991)

    H. Spohn,Large Scale Dynamics of Interacting Particles (Springer, Berlin, 1991)

    1991

  10. [18]

    Caputo, T

    P. Caputo, T. M. Liggett, and T. Richthammer, Journal of the American Mathematical Society23, 831 (2010)

    2010

  11. [19]

    D. N. Page, Physical Review Letters71, 1291 (1993)

    1993

  12. [20]

    Ledoux,The Concentration of Measure Phenomenon, Mathematical Surveys and Monographs, Vol

    M. Ledoux,The Concentration of Measure Phenomenon, Mathematical Surveys and Monographs, Vol. 89 (Amer- ican Mathematical Society, Providence, RI, 2001)

    2001

  13. [21]

    Kretschmann, D

    D. Kretschmann, D. Schlingemann, and R. F. Werner, Journal of Functional Analysis255, 1889 (2008)

    2008

  14. [22]

    Borcea, P

    J. Borcea, P. Br¨ and´ en, and T. M. Liggett, Journal of the American Mathematical Society22, 521 (2009), arXiv:0707.2340 [math.PR]

    2009 arXiv

  15. [23]

    Pemantle and Y

    R. Pemantle and Y. Peres, Combinatorics, Probabil- ity and Computing23, 140 (2014), arXiv:1108.0687 [math.PR]

    2014 arXiv

  16. [24]

    S. N. Hearth, M. O. Flynn, A. Chandran, and C. R. Laumann, Physical Review X16, 019901 (2026)

    2026

  17. [25]

    Alon and D

    G. Alon and D. Puder, Aldous-type spectral gaps in uni- tary groups (2026), arXiv:2603.00353 [math.PR]

    2026

  18. [26]

    Rakovszky, F

    T. Rakovszky, F. Pollmann, and C. W. von Keyser- lingk, Physical Review Letters122, 250602 (2019), arXiv:1901.10502 [cond-mat.stat-mech]

    2019 arXiv

  19. [27]

    Aharonov, I

    D. Aharonov, I. Arad, Z. Landau, and U. Vazirani, in Proceedings of the 41st Annual ACM Symposium on The- ory of Computing (STOC)(ACM, 2009) pp. 417–426, arXiv:0811.3412 [quant-ph]

    2009 arXiv

  20. [28]

    Anshu, I

    A. Anshu, I. Arad, and T. Vidick, Physical Review B93, 205142 (2016), arXiv:1602.01210 [quant-ph]

    2016 arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.