The Intrinsic Cost of Quantum Syndrome Extraction

Chia-Tung Chu

REVIEW

The Intrinsic Cost of Quantum Syndrome Extraction

Chia-Tung Chu

Quantum Physics · Submitted 30 Sep 2026

Pith's verdict: The intrinsic cut-cost invariant and the DBT counterexample are the real contributions and hold up; the space-depth tradeoff is conditional on a long chain of external Leverrier-Zemor hypotheses in Appendix D, which is where a referee should spend their effort.

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 →

· 2 major objections · 4 minor · Ask about this paper · The compiled record · Read the full review →

Challenge this review Comment (0) Read on arXiv ↗

Cite ▾

arXiv ID2609.39705 (version 2)

Pith IDOM2I3BEK, Pith's permanent record of this paper and its review

CitationChia-Tung Chu. "The Intrinsic Cost of Quantum Syndrome Extraction." arXiv:2609.39705 (2026). https://pith.science/paper/OM2I3BEK

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

The two things worth taking from this paper are the invariant and the counterexample, and both survive reading. The core is clean: chi_U = dim S - dim S_U - dim S_{U^c}, computed by three binary ranks, is provably the minimum crossing-generator count over all generating sets, it equals the mutual information of the maximally mixed code state, and the Schmidt-rank argument (flat projector spectrum, rank-4 / rank-2 primitive growth) gives tight Bell-pair, CNOT, and crossing-gate costs with self-contained proofs in Theorems 3.5 and 3.8. The gross-code worked example, 60 pairs versus 144, is a nice concrete certificate, and the rank formulas are reproducible.

The DBT counterexample is the most interesting single result. The annular surface code with the expander recombination of vertex checks is genuinely clever, appears to check out against Appendix C, and the paper is careful to scope it to the first arXiv version of DBT in that version's model. That alone justifies the paper.

The soft spot is exactly where the reader and the stress-test put it. The headline space-depth tradeoff, N D_min^2 = Theta(n^2), and the bottleneck bound rest entirely on Theorem 6.1's volume law for one fixed quantum Tanner family, and the proof in Appendix D.1 is a composition of external hypotheses from Leverrier-Zemor: component-code robustness, puncturing resistance, Ramanujan, total no-conjugacy, and the reduction theorem applied in both sectors. The dual-sector application is justified in one sentence. I agree the concern is real, not manufactured: if that step, or any of the constants a_i, kappa_i, fails, the positive tau in Eq. (6.1) is unsupported and the tradeoff collapses. It is a verification burden rather than a known flaw -- LZ's Theorem 17 proof does run the reduction in both sectors -- but the main text does not resupply the constants, and the referee can't check it without the full arXiv:2202.13641v3 in hand.

The paper is honest about everything else: exact noiseless circuits, crossing resources rather than elapsed device time, the volume law claimed only for one fixed family, no fault tolerance, the tradeoff for best rather than prescribed placement. All stated. The bounded-incidence counterexample examples in Appendix C.5, isolating the exact failing inference in DBT's Lemma 4, are a nice diagnostic.

Who gets value: anyone working on qLDPC extraction depth, hardware bottleneck bounds, or entanglement-cost arguments, and anyone who wants a sharp generator-independent measure. It deserves a serious referee; the effort should go to Theorem 6.1 and Appendix D.1 against the cited LZ version, not to the core theorems, which stand on their own.

What would settle it

Evaluate $\chi_U$ exactly on small regions for a concrete good qLDPC family enlarged by $\Theta(n)$ qubits fixed in $|0\rangle$ by weight-one Z checks; the paper argues this gives $\chi_U = 0$ on the added region, directly disproving any claim that rate and distance force the volume law. On the fixed Tanner family, verify the volume law numerically on small instances by computing $\dim S_U$ through the three binary ranks of Lemma 3.2 and checking $\chi_U \ge 2\tau|U|$ for every region with $|U| \le \eta n/2$. For the counterexample, check the annular construction on small $L$ to confirm the contracted Tanner graph of the recombined list actually expands with the claimed constant while the depth-13 circuit still implements the exact instrument.

Referee report

Significance

If the uniform volume law is accepted, the paper establishes a sharp, placement-independent interaction cost for syndrome extraction on a fixed qLDPC family and a tight grid tradeoff, while also refuting a published presentation-counting bound. The core cut-cost argument is simple, general, and self-contained: the flat Schmidt spectrum of the syndrome projector in Lemma 3.4, the operator-Schmidt rank growth argument in Theorem 3.5, and the remote-parity attainability protocol in Theorem 3.8 are convincing and are valuable independently of the later geometric results. The annular counterexample is worked out in detail and appears to genuinely satisfy the stated DBT v1 hypotheses. The paper is honest about its scope and limitations, including the fixed-family nature of Theorem 6.1 and the lack of fault-tolerance claims. My reservations concern only the load-bearing external-construction chain in Appendix D.1, not the self-contained core.

Major objections (2)

[Appendix D.1, Eqs. (D.7)–(D.17), Theorem 6.1] The uniform volume law, and hence Corollaries 6.2, 7.1, and 7.3, depends on applying the Leverrier–Zémor reduction theorem to both im Φ_0 and im Φ_1. The application to the second sector is justified in a single sentence: 'Applying the same theorem with the dual component codes and interchanged vertex classes reduces words of the second image using local L1 words, as in their proof of Theorem 17.' This is load-bearing: it is exactly this step that yields the constants a_1, κ_1, the condition 3T_i < σ_i M, the uniqueness/linearity of the local representation, and the positive g_eta and tau in Eqs. (D.16)–(D.21). The manuscript should state precisely which theorem and remark of arXiv:2202.13641v3 are being used and verify that (C_A^⊥, C_B^⊥), together with the interchanged vertex classes, satisfy all their hypotheses, including the robustness and puncturing-resistance bounds and the mixing estimates used in Eq. (D.11). Without that verification, Theorem 6.1 is a conditional statement whose main quantitative consequence (positive tau) is not auditable from the text.
[Appendix D.1, Eqs. (D.18)–(D.21)] The proof of Theorem 6.1 derives explicit formulas for eta_0 and tau_0 in terms of the reduction constants a_i and kappa_i, but those constants are only asserted to follow from 'Remark 15' of [LZ22]. Since the positivity of tau and the validity of Eq. (D.12) depend on the numerical balance between the local distances ℓ_i, the diagonal and union mixing parameters λ̄_i, μ̄, and the reduction-range parameters κ_i, the appendix should state the exact content of the cited reduction result and show how it yields a weight decrease proportional to D in both stabilizer sectors. If the quantitative constants from the external paper are not uniform in the way the appendix assumes, the volume law in Eq. (6.1) and the tradeoff in Corollary 7.3 would not follow. This is a request for verification, not for reproducing the published proof of [LZ22].

Minor objections (4)

[Lemma 3.3] The notation 'ψ_U' in the sentence 'Every pure code state |ψ⟩ then has ψ_U = ω_U' is undefined; it should be written as (|ψ⟩⟨ψ|)_U = ω_U to avoid confusion with a scalar or state vector.
[Section 2, paragraph after the abstract] The phrase 'forcekd2 =O(n)[ BPT10]' appears to be a typesetting or ligature error; it should read 'forces k d^2 = O(n) [BPT10]'.
[Remark 3.6, Eq. (3.11)] The quantities g_L^unitary(β) and g_L^Pauli measurement(β) are used in Eq. (3.11) before being defined; please define them as the numbers of crossing unitary gates and crossing two-qubit Pauli measurements on the trajectory, respectively.
[Theorem 5.1, proof sketch] The sentence 'using four nearest-neighbor CNOT layers for each check type' is clearer if it states that for vertex (X) checks the ancilla is the control of the CNOTs while for face (Z) checks the data qubits are the controls; Appendix C.3 supplies this detail, but the theorem statement should not rely on the appendix for the direction of the gates.

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

The paper

The paper claims

The true cost of syndrome extraction is one number read off the stabilizer group, and this paper shows it is exact for every hardware cut.

The reading

This paper claims that the interaction a quantum syndrome measurement must pay across a hardware cut is a single quantity computed from the stabilizer group itself, with no dependence on which generating set one happens to write down. That quantity, the intrinsic cut cost $\chi_U$, counts the independent constraints that remain once all stabilizers confined to either side of the cut are removed, and it can be evaluated by three binary ranks on any check matrix. The paper proves that every exact extraction circuit must use at least $\chi_U/2$ two-qubit operations crossing the cut on every nonzero branch, and that with free local processing $\chi_U$ shared Bell pairs, $\chi_U$ crossing CNOTs, or $\lceil\chi_U/2\rceil$ arbitrary crossing gates are both necessary and sufficient. For one fixed family of quantum Tanner codes the cost grows linearly on every sufficiently small region, which forces depth proportional to the ratio of data to boundary edges behind any hardware bottleneck, and a matching adaptive compiler shows the planar space–depth tradeoff $N D_{\min}^2 = \Theta(n^2)$ is tight. The payoff is a way to price a hardware layout before committing to it, and a demonstration that bounds based on listed checks or on global code parameters can both mislead.

What carries the argument

The load-bearing object is the intrinsic cut cost $\chi_U$, the dimension of the quotient $S/(S_U + S_{U^c})$, together with the flat Schmidt decomposition of the syndrome projector $\Pi_s$ across the cut, whose $2^{\chi_U}$ equal coefficients turn exactness into a rank constraint on each branch of an adaptive circuit. The upper bound is carried by the remote-parity gadget, which measures a crossing Pauli with one shared Bell pair and applies $\Pi_s$ on each raw branch, and by the CNOT/SWAP conversions that turn stored Bell pairs into crossing-gate counts. For the volume law, the machinery is the Leverrier–Zémor reduction theorem for quantum Tanner codes: every stabilizer confined to a small region gets a unique short description in terms of local check components, and expansion bounds the dimension of the space of such descriptions, with overlap of boundary squares controlled by the Ramanujan and total no-conjugacy assumptions of the construction.

Extended reading notes

Core claim

For a data region $U$ of an $[[n,k,d]]$ stabilizer code, define $\chi_U(S) = \dim S - \dim S_U - \dim S_{U^c}$, where $S$ is the stabilizer space regarded as a binary vector space and $S_U$, $S_{U^c}$ are the stabilizers supported entirely on one side. This dimension is exactly the minimum number of generators crossing the cut over all generating sets of $S$ (Lemma 3.1), equals the mutual information across the cut in the maximally mixed code state (Lemma 3.3), and can be evaluated from any check matrix by $\chi_U = \operatorname{rank} H_U + \operatorname{rank} H_{U^c} - \operatorname{rank} H$ (Lemma 3.2). The operational content is a Schmidt-rank argument: the vectorized syndrome projector has flat Schmidt spectrum of rank $2^{\chi_U}$ across the cut, a local operation cannot raise the rank, and a crossing two-qubit Kraus operator raises it by at most four, so every exact circuit on every nonzero branch pays at least $\chi_U/2$ crossing operations (Theorem 3.5), with the constant improving to $\chi_U$ when all crossing operations are CNOTs. Remote-parity gadgets show the bound is attainable: $\chi_U$ Bell pairs, $\chi_U$ crossing CNOTs, or $\lceil\chi_U/2\rceil$ arbitrary crossing gates are sufficient and necessary with free local processing (Theorem 3.8). For one fixed quantum Tanner family satisfying the full Leverrier–Zémor construction hypotheses, the paper proves the uniform volume law $\chi_U \ge 2\tau|U|$ on every region with $|U| \le \eta n/2$ (Theorem 6.1), which forces depth at least $\tau u/b$ behind any hardware bottleneck. An adaptive compiler measures any bounded-weight presentation in depth $O(\max\{1, n/\sqrt{N}\})$ on a chosen square-grid placement, making the tradeoff $N D_{\min}^2 = \Theta(n^2)$ tight for that family. As a control on the presentation-dependent view, the paper constructs annular surface codes with expanding bounded-weight presentations that admit depth-13 extraction, contradicting the local-expander corollary of an earlier listed-crossing bound.

Load-bearing premise

The uniform volume law for small regions rests on the complete Leverrier–Zémor construction hypotheses — component codes meeting the stated distance, robustness, and puncturing-resistance bounds, plus the Ramanujan and total no-conjugacy assumptions on the square complex — so if any of those structural inputs fails, the volume law and the tight space–depth tradeoff built on it collapse.

Editorial extensions

If this is right

A circuit designer can compute $\chi_U$ for candidate cuts of a finite code by three binary Gaussian eliminations, obtaining a hardware-dependent lower bound before any circuit is written.
Any placement of the fixed Tanner family on any hardware graph pays depth at least $\tau u/b$ for each small module holding $u$ data qubits behind $b$ boundary edges, so module boundaries have a floor set by the code itself.
The planar scaling $n/\sqrt{N}$ is tight for that family across the whole space range $n \le N \le c n^2$, settling the minimum depth $D_{\min} = \Theta(\max\{1, n/\sqrt{N}\})$ over circuits and fixed placements.
The gap between listed-crossing counts and $\chi_U$ means presentation expansion cannot substitute for an intrinsic stabilizer-space bound, so no local-expander lower bound can be recovered by choosing a better check list.
Concrete resource budgets follow, such as the 60 Bell pairs needed for the half-torus cut of the $[[144,12,12]]$ gross code and the simultaneous attainment of all edge budgets on tree-connected processors.

Reading between the lines

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

The same $\chi_U$ invariant plausibly governs repeated and extended-round extraction, but the paper proves its exact-resource statements only for a single round; checking the invariant's behavior under syndrome accumulation is a natural testable extension left implicit.
Because $\chi_U$ equals a mutual information, the paper's resource counts may glue onto communication-capacity analyses of device wiring, suggesting a benchmark that compares $\chi_U$ against naive remote-CNOT counts on mid-size codes before any hardware is built.
The volume law's dependence on the full Leverrier–Zémor hypotheses suggests a test: search for a good qLDPC family whose products of checks cancel outside small regions, and compute $\chi_U$ exactly on those regions; finding one would delimit exactly which good families admit the uniform small-region theorem.
For networks with cycles, the paper leaves open whether the per-edge cut inequalities characterize a feasible entanglement allocation, so the tree result should be treated as special, not as a general network characterization.

Original abstract

How much time is fundamentally required to extract the syndrome of a stabilizer code on hardware with a given connectivity graph? Previous bounds answer through a chosen list of checks (Delfosse, Beverland, and Tremblay, 2021) or through the global code parameters (Baspin, Fawzi, and Shayeghi, 2023). Here, we show that neither determines the interaction that a particular hardware cut requires. We therefore define the intrinsic cut cost, $\chi$, directly from the stabilizer group, independent of the generating set used to describe the code. It is computable from the check matrix alone, and equals the mutual information across the cut in the maximally mixed code state. We show that every exact extraction circuit, however adaptive, needs at least $\chi/2$ two-qubit operations across the cut, and alternatively with free local processing exactly $\chi$ shared Bell pairs. Guided by this intrinsic cut cost, we construct annular surface codes with constant-depth local extraction that give a counterexample to Corollary 1 of Delfosse, Beverland, and Tremblay. Furthermore, for a fixed quantum Tanner family, $\chi$ grows at least linearly with the size of every sufficiently small region, giving a depth guarantee at each corresponding bottleneck in every layout. This strengthens the global perspective of Baspin, Fawzi, and Shayeghi and, together with our matching square-grid construction, yields a tight space--depth tradeoff. Our results connect the interaction required by syndrome extraction to hardware communication capacity and thus provide a way to assess unavoidable circuit costs before committing to a particular layout.

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, 5,736 input tokens.

What the abstract itself states, as probabilities: claims first 0.08 · states limitation 0.24 · human participants 0.01 · controlled comparison 0.08 · randomized assignment 0.01 · states quantitative result 0.54 · compares with prior results 0.97 · mentions code or data availability 0.03.

The abstract, sentence by sentence

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

How much time is fundamentally required to extract the syndrome of a stabilizer code on hardware with a given connectivity graph? Previous work by Delfosse--Beverland--Tremblay (2021) and Baspin--Fawzi--Shayeghi (2023) established bounds connecting syndrome extraction to locality constraints on quantum hardware. Here, we address this question by first formulating an intrinsic measure of the interaction required for syndrome extraction directly from the stabilizer group, independent of the generating set used to describe the code. We then show that this measure yields hardware-dependent lower bounds on exact syndrome-extraction depth for arbitrary stabilizer codes. Guided by this measure, we also construct annular surface-code examples with constant-depth local syndrome extraction that provide a counterexample to Corollary 1 of DBT21. Furthermore, for a fixed quantum Tanner family, this measure strengthens the depth bound explicitly stated in BFS23 and, together with our matching construction, establishes an asymptotically tight space--depth tradeoff. In theory, our results connect the information required by syndrome extraction to the communication capacity of a device; while in practice, they provide a way to assess unavoidable circuit costs before committing to a hardware layout.

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	How much time is fundamentally required to extract the syndrome of a stabilizer code on hardware with a given connectivity graph?	background 0.82	0.05	0.02	unclear 0.97
s1	Previous work by Delfosse--Beverland--Tremblay (2021) and Baspin--Fawzi--Shayeghi (2023) established bounds connecting syndrome extraction to locality constraints on quantum hardware.	background 1.00	0.02	0.02	asserted 1.00
s2	Here, we address this question by first formulating an intrinsic measure of the interaction required for syndrome extraction directly from the stabilizer group, independent of the generating set used to describe the code.	method 0.65	0.92	0.09	asserted 0.98
s3	We then show that this measure yields hardware-dependent lower bounds on exact syndrome-extraction depth for arbitrary stabilizer codes.	result 0.93	0.96	0.44	asserted 1.00
s4	Guided by this measure, we also construct annular surface-code examples with constant-depth local syndrome extraction that provide a counterexample to Corollary 1 of DBT21.	result 0.75	0.96	0.13	asserted 1.00
s5	Furthermore, for a fixed quantum Tanner family, this measure strengthens the depth bound explicitly stated in BFS23 and, together with our matching construction, establishes an asymptotically tight space--depth tradeoff.	result 0.95	0.94	0.31	asserted 1.00
s6	In theory, our results connect the information required by syndrome extraction to the communication capacity of a device; while in practice, they provide a way to assess unavoidable circuit costs before committing to a hardware layout.	interpretation 1.00	0.75	0.21	asserted 0.94

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

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.** Equivalent checks and a physical boundary. In the top panels, the four independent checks Si = ZℓiZr are replaced by T1 = S1 and Ti = Si−1Si for i = 2, 3, 4. Three of the new checks lie entirely in U = {ℓ1, . . . , ℓ4}, leaving the intrinsic crossing count χU = 1. The lines in these panels describe check support. The bottom panel instead shows hardware edges, where the vertex region L contains data and ancillas while U = Q ∩ L contains only its data. Only quantum operations across the physical boundary contribute to the cut cost, and changing generators changes neither that boundary…

**Figure 2.** The adaptive grid compiler. (a) Data sites are chosen in 4 × 4 tiles and balanced over tile rows. A batch of terminal pairs uses disjoint endpoint rows, and pairs in different rows use distinct routing columns. (b) The tile has data at (0, 0), terminal port T = (0, 1), and directional ports N = (0, 2), E = (2, 3), S = (3, 2), W = (2, 0), with coordinates ordered as row and column. Its ancilla grid prepares each requested port pair locally, serializing up to two pair preparations in a shared tile while restoring all spectator states. (c) Bell measurements on disjoint facing ports…

Figures reproduced from arXiv:2609.39705 by the authors.

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.