Square-root growth of operator entanglement in an integrable brickwork circuit

Bal\'azs Pozsgay

REVIEW

Square-root growth of operator entanglement in an integrable brickwork circuit

Balázs Pozsgay

Statistical Mechanics · Submitted 30 Sep 2026

Pith's verdict: A serious counterexample refuting the logarithmic-operator-entanglement expectation — the math is self-contained and sound except for one imported causal-reduction step that deserves a referee check.

T0 · AI review

How this review was made: An AI referee (deepseek/deepseek-v4-flash) read the paper on October 1, 2026 and answered a fixed list of questions about it. T0 is the starting level every paper gets. The levels describe how deeply Pith has checked the paper, never the authors.

T0 AI review An AI read the paper and answered a fixed list of questions.
T1 Audit Automatic checks that the references and files the paper relies on exist and match.
T2 Compute The paper's code or calculations re-run with the same results.
T3 Formal statement The claims written out in a formal math language a computer can read.
T4 Proof Those claims proved, and the proof checked by computer.

How Pith checks papers →

· 1 major objection · 5 minor · Ask about this paper · The compiled record · Read the full review →

Challenge this review Comment (0) Read on arXiv ↗

Cite ▾

arXiv ID2609.39401 (version 2)

Pith IDRZLUIVK3, Pith's permanent record of this paper and its review

CitationBal\'azs Pozsgay. "Square-root growth of operator entanglement in an integrable brickwork circuit." arXiv:2609.39401 (2026). https://pith.science/paper/RZLUIVK3

Cite this reviewBibTeX for Pith's review of this paper

Claim paper

Why Pith says this

Highlight any text to comment on it.

Editor's note

read the letter

This is a real counterexample and worth serious referee time. The headline claim — von Neumann operator entanglement growing as (log 2)/sqrt(pi) sqrt(t) in an integrable brickwork circuit, with an exp(c sqrt(t)) MPO bond-dimension lower bound at fixed accuracy — is new in the cited literature. Prior work reported logarithmic or bounded operator entanglement; this paper breaks that pattern with a mechanism that is actually intelligible: sector labels propagate ballistically, and one-sided non-degenerate color reconstruction generates sqrt(t) typical imbalance, linear Rényi growth below index one, logarithmic above. The gate evades the known involutive-plus-dual-unitary sufficient condition by being involutive but not dual-unitary, and the paper says so plainly.

The mathematics is mostly self-contained and exact. The domain-wall computation is fully proved in the SM: flat Schmidt decompositions, the reordering and common-word-removal lemmas, the bijectivity of the color maps, and the exact telescoping binomial average giving the sqrt(t) prefactor with no fitted constants. The entropy bounds, the Rényi variational optimization, and the MPO retained-weight argument are all rigorous given the operator-state correspondence. The AI-assisted provenance is disclosed up front, which is the right way to do it.

The genuine soft spot is the one the stress-test flags: Eq. (21), the operator-state causal reduction, is imported from the author's companion preprint [15] and not re-derived. SM Sec. II takes it as a starting point. If [15]'s hypotheses don't cover a non-dual-unitary, involutive gate that is the identity on equal sector labels, then the operator-sector result and the MPO bound inherit a gap. I would weigh this as checkable rather than likely-fatal: the property the reduction actually needs, folded-identity stationarity, is verified in Eq. (20) for this gate, and the structural argument the paper proves for the state problem is the same one being imported for the operator case. Still, it should be verified. The commuting-charge integrability claim also imports from [15], but that is minor — the braid relation is checked in the paper and the transfer-matrix construction is standard.

Two smaller absences: no numerical check of Eq. (25), which would independently validate the operator claim without needing [15]; and the D=8 reflection-invariant circuit is clearly labelled a conjecture, which is fine. The Rényi alpha>1 prefactor is honestly left undetermined.

My recommendation: send to a serious referee. The natural ask is that the author either confirm [15]'s hypotheses cover this gate or add a small numerical verification of the sqrt(t) law. With either, this is an accept. I would bring it to the reading group.

My verdict agrees with the reader's CONDITIONAL, but I'd push significance up: a clean refutation of a widely cited expectation, with the mechanism and the proofs laid out, is more than a curiosity.

Referee report

Significance

If the operator-sector claims hold, this is a genuine and striking counterexample: it would falsify the widely cited expectation that Yang–Baxter integrability forces at most logarithmic operator-entanglement growth, and it would simultaneously obstruct polynomial-bond-dimension MPO simulation at fixed accuracy. The paper has real strengths that deserve explicit credit: the leading √t prefactor is parameter-free, coming from exact binomial averages (SM Eqs. S31–S33); the flat Schmidt spectra of the residual cores are proved rather than assumed (SM Secs. I.D and II.B); the Rényi rates follow from an optimization with a unique interior maximum, with the correct limits v_α → log 2 as α→0+; and the MPO lower bound is obtained by a clean Eckart–Young / retained-weight argument. The domain-wall half of the computation is essentially self-contained. The overall verdict therefore hinges on a single load-bearing imported step, the operator–state correspondence of Eq. (21), discussed in the major comments.

Major objections (1)

[Main text, 'From states to operators', Eq. (21); SM Sec. II, Eq. (S8)] The operator-sector results — Eq. (25) for the von Neumann entropy, Eq. (26) for Rényi entropies, and the MPO lower bound (S58) — all inherit the assertion that the Schmidt spectrum of the Heisenberg-evolved operator across cut 0|1 equals that of the folded rectangle state |Ω_t⟩ of Eq. (21). The main text introduces Eq. (21) with only 'the causal reduction using (20) leads to the auxiliary state [15]', and SM Sec. II explicitly takes Eq. (S8) as its starting point rather than proving it. Because Ref. [15] is a same-author preprint whose hypotheses are not restated or verified for the present gate (involutive, non-dual-unitary, acting as the identity on equal-sector inputs), the operator √t law and the exp(c√t) MPO bound are conditional on an unverified external result. This is fixable within the manuscript's scope: given that Eq. (20) (invariance of the folded identity under R̂) is already proved, the light-cone reduction to the rectangle Ŵ_t is a short argument and should be supplied in the text (one paragraph plus a figure), or, alternatively, the hypotheses of Ref. [15] should be stated and checked explicitly against this gate. Until this is done, the central operator-entanglement claim is not self-contained.

Minor objections (5)

[References, footnote 24] Ref. [15] is cited for three distinct purposes: the causal reduction of Eq. (21), the transfer-matrix construction of local conserved charges, and the existence of ballistic gliders and the superintegrable family. Since it is a companion preprint, the authors should add a sentence in the Discussion stating which of its results are assumed, and ideally note its publication status.
[Discussion, paragraph on the counterexample claim] The wording 'our main result is that Yang–Baxter integrability does not enforce logarithmic growth' is slightly stronger than what is established: strictly, the manuscript shows this for the operator-entanglement observable under the specific Heisenberg convention of footnote [17] and for the specific charge construction of Ref. [15]. A sentence qualifying the scope would prevent over-reading.
[Main text, Fig. 2 and surrounding text] The two-step reduction that 'the remaining network is W_t' in the domain-wall problem is stated in a paragraph rather than pinned down formally. Since the operator problem inherits the same rectangle geometry, a precise statement (which specific permutation of the 2t sites survives after the light-cone cut) would make the foundation of both computations more transparent.
[SM Eq. (S22)] The branch-entropy formula g(d) = |d − 1/2| − 1/2 renders poorly in the SM (missing spacing and odd absolute-value bars); please fix the typesetting.
[SM Sec. V.B] The central-limit step selecting c_ε from the continuity of the limiting half-Gaussian at zero is correct, but the logic would be clearer if the sentence explicitly stated that P(G_t/√t ≤ a) tends to the limiting half-Gaussian cumulative value, so that any ε<1 leaves room to choose a small enough; as written the reader must re-derive this.

How Pith checks papers → · Full audit of this paper →

The paper

The paper claims

A Yang–Baxter gate breaks the logarithmic bound on operator entanglement

The reading

The paper constructs a four-state brickwork circuit whose gate is an involutive permutation matrix satisfying the constant Yang–Baxter equation, and shows that the von Neumann entanglement of a local operator under Heisenberg evolution grows as $(\log 2)/\sqrt{\pi}\,\sqrt{t} + O(\log t)$, which is square-root growth rather than the logarithmic growth that had been widely expected for integrable systems. It is a counterexample to the expectation that integrability enforces logarithmic or bounded operator-entanglement growth. The paper also finds that fixed-index Rényi entropies of the operator grow linearly for index below one and logarithmically for index above one, and that any matrix-product-operator simulation at fixed relative Hilbert–Schmidt error requires a bond dimension at least $\exp(c\sqrt t)$ for some $c>0$. The mechanism is concrete: sector labels propagate ballistically while color degrees of freedom become maximally entangled when the sector labels differ, and the square-root law is set by the $\sqrt t$ scale on which the count imbalance between the two packets fluctuates.

What carries the argument

The central object is the permutation gate $R|s,a;\tau,b\rangle = |\tau,\, a+(s+\tau)b;\; s,\, b\rangle$ acting on sector labels $s,\tau\in\{A,B\}$ and color bits $a,b\in\mathbb{F}_2$, where binary arithmetic is understood. It is real, unitary, involutive, and satisfies the braid relation $R_{12}R_{23}R_{12}=R_{23}R_{12}R_{23}$, giving a unitary representation of the symmetric group on the $n$-site Hilbert space; it acts as the identity when the sector labels coincide and applies a color shear $h(a,b)=a+b$ otherwise. The argument proceeds by expressing the evolved state or folded operator as a coherent sum over fixed-count branches, using a common-word removal lemma to strip shared sector content and leaving a residual core whose Schmidt spectrum is flat with rank $2^{|d|}$, and a count-sector entropy bound then shows the von Neumann entropy lies between the mean branch entropy and the mean plus the Shannon entropy of the count labels, which is $O(\log t)$. The binomial average of the count imbalance $|k-\ell|$ yields $\sqrt{t}/\sqrt{\pi}$, giving the square-root law. For the operator problem the same machinery acts on the doubled Hilbert space via the folding trick and a Hermitian-traceless source with the same normalized Schmidt spectrum.

Extended reading notes

Core claim

Yang–Baxter integrability does not by itself enforce logarithmic growth of operator entanglement. For a one-site operator in the $D=4$ permutation-gate circuit, the von Neumann operator entropy is $S_1^{\mathrm{op}}(t) = (\log 2)/\sqrt{\pi}\,\sqrt{t} + O(\log t)$ (Eq. 25), and the MPO bond dimension needed for fixed relative Hilbert–Schmidt error $\varepsilon<1$ obeys $\chi_\varepsilon(t) \ge \exp(c_\varepsilon \sqrt t)$ for sufficiently large $t$. The gate is involutive but not dual-unitary, which places it outside the sufficient conditions for logarithmic growth identified in prior work on Yang–Baxter gates. The same square-root hierarchy holds for a simpler Schrödinger-picture domain-wall quench, which the paper uses as a template for the operator calculation.

Load-bearing premise

The load-bearing premise is that the causal-reduction and folding trick of the companion preprint apply to this involutive, non-dual-unitary gate, so that the Schmidt spectrum of the folded auxiliary state at Eq. (21) equals the operator-Schmidt spectrum of the Heisenberg-evolved operator; if those hypotheses exclude the gate, the square-root operator law does not follow even though the domain-wall computation is self-contained.

Editorial extensions

If this is right

Yang–Baxter integrability, even accompanied by an exponentially large family of ballistic gliders, does not enforce logarithmic operator-entanglement growth.
The von Neumann operator entropy in the infinite volume grows as $(\log 2)/\sqrt{\pi}\,\sqrt{t} + O(\log t)$, so the typical Schmidt spectrum across the cut is dominated by branches whose count imbalance is of order $\sqrt{t}$.
Fixed-index Rényi entropies grow linearly in $t$ below index one and logarithmically above index one, exposing an exponentially large number of small Schmidt probabilities carried by rare, heavily imbalanced branches.
Matrix-product-operator simulations at fixed relative Hilbert–Schmidt error $\varepsilon<1$ require bond dimension at least $\exp(c_\varepsilon\sqrt t)$; any polynomial bond dimension captures an asymptotically vanishing fraction of the operator's Hilbert–Schmidt weight.
The same square-root growth hierarchy holds for a Schrödinger-picture domain-wall quench, so the mechanism is not special to the folded operator construction.

Reading between the lines

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

Beyond the paper: the same count-sector bound should apply to any integrable gate that propagates a ballistic label while conditionally entangling a second degree of freedom, so the square-root law may be characteristic of decomposable Yang–Baxter gates; the flatness of the residual-core spectrum with rank $2^{g(|k-\ell|)}$ is a testable signature.
Beyond the paper: the exponential rank growth implies a fast crossover — the time at which the required bond dimension exceeds a fixed practical budget should scale as $(\log\chi / c_\varepsilon)^2$, which is within numerical reach for moderate $t$.
Beyond the paper: the $D=8$ reflection-invariant circuit conjectured in Eq. (S59) of the Supplemental Material is a direct falsification target; if it fails to reproduce the same scaling, the mechanism would be tied to the specific asymmetry of the $D=4$ gate.
Beyond the paper: because chaotic circuits typically show linear operator entanglement, the distinct spectrum shape (flat branches with $2^{|k-\ell|}$ equal Schmidt coefficients) provides a clean diagnostic separating decomposable integrable dynamics from chaos.

Original abstract

It is widely expected that the von Neumann entanglement of a local operator grows at most logarithmically in integrable many-body systems in infinite volume. We give a counterexample in a four-state brickwork circuit whose gate is a permutation matrix solving the constant Yang--Baxter equation. The von Neumann operator entropy grows as $(\log2)\sqrt{t/\pi}+O(\log t)$, whereas fixed-index R\'enyi entropies grow linearly below index one and logarithmically above it. At fixed relative Hilbert--Schmidt error below one, matrix product operator simulations require a bond dimension growing at least as $\exp(c\sqrt t)$ for some $c>0$.

The compiled record

The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-10-01: 1 request, 3,698 input tokens.

What the abstract itself states, as probabilities: claims first 0.09 · states limitation 0.30 · human participants 0.01 · controlled comparison 0.11 · randomized assignment 0.01 · states quantitative result 0.98 · compares with prior results 0.92 · mentions code or data availability 0.02.

The abstract, sentence by sentence

arXiv listing abstract v1. Results and the main claim are underlined; show every role.

It is widely expected that the von Neumann entanglement of a local operator grows at most logarithmically in integrable many-body systems in infinite volume. We give a counterexample in a four-state brickwork circuit whose gate is a permutation matrix solving the constant Yang--Baxter equation. The von Neumann operator entropy grows as $(\log2)\sqrt{t/\pi}+O(\log t)$, whereas fixed-index R\'enyi entropies grow linearly below index one and logarithmically above it. At fixed relative Hilbert--Schmidt error below one, matrix product operator simulations require a bond dimension growing at least as $\exp(c\sqrt t)$ for some $c>0$.

background contribution method result interpretation main claim

What the classifier reports for each sentence. The number is the probability it gave the label.
#	sentence	role	the authors' own work	states the main conclusion	strength
s0	It is widely expected that the von Neumann entanglement of a local operator grows at most logarithmically in integrable many-body systems in infinite volume.	background 1.00	0.03	0.04	hedged possibility 0.79
s1	We give a counterexample in a four-state brickwork circuit whose gate is a permutation matrix solving the constant Yang--Baxter equation.	contribution 0.94	0.95	0.13	asserted 1.00
s2	The von Neumann operator entropy grows as $(\log2)\sqrt{t/\pi}+O(\log t)$, whereas fixed-index R\'enyi entropies grow linearly below index one and logarithmically above it.	result 1.00	0.93	0.77	asserted 1.00
s3	At fixed relative Hilbert--Schmidt error below one, matrix product operator simulations require a bond dimension growing at least as $\exp(c\sqrt t)$ for some $c>0$.	result 0.94	0.84	0.12	asserted 0.99

What kind of paper the classifier reports this is: theoretical result 1.00 · evidence from proof or derivation 0.89.

What was examined, and what was not

sources read

arXiv listing abstract v1

not examined

full text
figures and tables
references and citing sentences
relations to other papers

levels run

P abstract

not run

S sections
T body sentences
C citations
R cross-paper relations

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. Every number in the measurements table is the authors', checked by code. Record as JSON

Figures

**Figure 1.** FIG. 1. Braid representation of the circuit and its local gate.

**Figure 2.** FIG. 2. Entangling rectangles for

Figures reproduced from arXiv:2609.39401 by the authors.

References

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Free fermion models with an operator entanglement growing as $O(\sqrt{t})$
quant-ph 2026-10 conditional novelty 7.0 of 10

In random free fermion models, OSEE of Jordan-Wigner string operators and of the evolution operator grows as O(sqrt(t)) for all Renyi indices, with exact prefactors obtained from an SSEP large deviation function.
Operator-Space Entanglement Entropy in Random Fermionic Gaussian Dynamics
quant-ph 2026-10 conditional novelty 6.0 of 10

Operator-space entanglement entropy of Majorana strings under Brownian SYK2 dynamics is solved exactly in the thermodynamic limit via the real β=1 free Jacobi process, yielding Gaussian Page curves distinct from Haar-...

Discussion

Disagree with the review, or know something it missed? Leave a comment, highlight any passage above to comment on it, or challenge this review formally.