{"id":"3874e2ce-e365-414e-ba8a-f8807cc2dfe5","arxiv_id":"2607.27342","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Constant signal-aligned noise makes asymptotic beyond-SQL quantum sensing impossible for any protocol, including encoded, biased, adaptive, and nonstabilizer schemes.","lead":"Researchers prove that storing a quantum sensor in an error-correcting code cannot rescue Heisenberg-limited metrology when noise acts along the same direction as the signal. They also show that transversal non-Clifford gates force increasingly large stabilizer checks, ruling out concatenation for beyond-standard-quantum-limit sensing.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified","rationale":"The reader's weakest_assumption identifies the constancy of noise rates as the main caveat, and I agree that the no-go is conditional on this. However, the paper explicitly states 'constant-strength signal-aligned noise,' so this is a well-defined boundary rather than an unacknowledged limitation. I found no mathematical error in the proofs: Theorem 3's susceptibility argument is sound, Lemma 9's proof is correct once the ordering of K_a^† and K̇_a is tracked (β=0 implies W†Ẇ=0), and the α_r calculation is plausible and consistent with known single-qubit dephasing results. The only notable discrepancy is the abstract's overstatement of Theorem 2, which the reader correctly flagged. Since this does not affect the main no-go, I recommend keeping the conditional verdict.","tokens_in":28212,"tokens_out":42124,"duration_ms":343928,"concrete_test":"Independently recompute the single-qubit α_r and β_r in Eq. (S112) using the gauge in Eq. (S109) with an explicit symbolic algebra package; verify β_r=0 and α_r = (1−pe)t_r^2 e^{-4γt_r}/(1−(1−pe)e^{-4γt_r}) I. If this fails, the Bures bound in Eq. (7) is invalid.","verdict_should_be":"UNCHANGED","load_bearing_attack":"After detailed review, no internal inconsistency or unsupported step was found in the proof of Theorem 5. The derivation in Appendix S3 is self-consistent: the gauge choice in Eq. (S109) satisfies β=0, the channel-extension QFI bound (Lemma 9) is correctly applied, and the simplified Bures bound (S107) follows from the stated inequalities. The no-go genuinely depends on the explicit assumptions that the dephasing strength γ and the interface erasure probability pe are positive constants independent of n and ε; if these rates are allowed to decay (e.g., γ˜n^{-1}), the third term in Eq. (8) becomes sub-SQL and Heisenberg scaling is not excluded. This is a boundary of the theorem rather than a flaw. The only substantive presentation issue is the abstract's phrase 'any r-level concatenated realization' in Theorem 2, which omits the theorem's condition k_j=1 for j≥2; this does not affect the metrological no-go.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves a set of restrictions on transversal non-Clifford gates in stabilizer codes and applies them to quantum metrology. Theorem 1 shows that a stabilizer code of distance d >= 3 supporting a genuinely level-D transversal logical gate must have a stabilizer generating set containing a check of weight at least 2^D. Theorem 2 bounds the depth r of a concatenated realization by floor(log_2 n / D) under the condition that all inner constituent codes encode one logical qubit (k_j = 1 for j >= 2). Theorem 3 shows that small-angle transversal single-qubit rotations inducing a nontrivial logical action force stabilizer and normalizer weights Omega(s^2/(n|theta|^2)); Theorem 4 expresses this as an irreducibility property of syndrome extraction. Theorem 5, proved in Appendix S3, gives a Bures-distance bound for arbitrary multi-round sensing protocols subject to signal-aligned dephasing and flagged erasure, dB(rho_{omega0}, rho_{omega1}) <= epsilon min{nT, T sqrt(n(1-p_e)/p_e), (1/2) sqrt(nT/gamma)}, and derives the interrogation-time lower bound T = Omega(max{1/(epsilon n), sqrt(p_e/(1-p_e)) 1/(epsilon sqrt n), gamma/(epsilon^2 n)}). The authors argue this rules out any asymptotic beyond-SQL advantage under constant-strength signal-aligned noise without invoking the quantum Cramer-Rao bound or unbiased estimators. The proofs are given in three substantive appendices.","tokens_in":28371,"tokens_out":46274,"duration_ms":443186,"significance":"If the results hold, they close an important gap in the quantum metrology no-go literature: previous HNLS-based results apply only to unbiased estimators under the QCRB, whereas Theorem 5 applies to the distinguishability of output states of arbitrary adaptive multi-round protocols, including biased estimators, QEC, nonstabilizer encodings, and approximate codes. Theorems 1 and 3 also provide quantitative design constraints for transversal non-Clifford gates, complementing Eastin-Knill and Bravyi-Konig. The paper is careful to state its noise model explicitly and to separate the dephasing and interface-erasure assumptions; the technical appendices are detailed and the central derivation in Appendix S3 is self-consistent. The main advertised contribution---a no-go for asymptotic advantage under signal-aligned noise---is substantial and would be of significant interest to both the fault-tolerance and quantum metrology communities.","major_comments":[{"comment":"The abstract states 'any r-level concatenated realization satisfies r <= floor(log_2 n / D)' and that concatenation is ruled out for beyond-SQL metrology. Theorem 2, however, assumes that every constituent code C_j for j >= 2 has k_j = 1 (single logical qubit). The proof in Appendix S1 (Lemmas 4 and 6) uses this single-logical-qubit structure to lift genuine level-D action down the concatenation tree. The abstract's unqualified wording is therefore unsupported. Please qualify both the abstract and the discussion in Section II.B to 'any r-level concatenated realization with single-logical-qubit inner constituents' (or supply a proof for general k_j).","section":"Abstract / Theorem 2"},{"comment":"The abstract says 'supporting a transversal logical action in level D of the Clifford hierarchy', but Theorem 1 requires the induced logical unitary to be genuinely level D, i.e. in C_D \\ C_{D-1}. Since the identity and all Clifford gates lie in C_D for D >= 3, the unqualified phrase is false if read as 'in C_D'. The statement should consistently say 'genuinely level D' in the abstract and in the corresponding summary of results.","section":"Abstract / Theorem 1"}],"minor_comments":[{"comment":"The sentence 'The first term is the Heisenberg limit and can dominate only in a nonasymptotic regime...' is unclear. In Eq. (8), the first term is a lower-bound contribution from the coherent Bures bound, not an upper bound that 'dominates'. Please rephrase to avoid confusion between the Bures upper bounds and the resulting T lower bound.","section":"Section II.C, Eq. (8)"},{"comment":"The proof of Lemma 9 is brief and cites Ref. [7] for the adaptive channel-extension bound. The explicit gauge choice (S109) is what sets beta = 0; it would improve readability to state explicitly that this is a unitary change of Kraus representation and not an additional physical assumption.","section":"Appendix S3, Lemma 9"},{"comment":"Figure 1 is referenced in the text but is not present in the supplied manuscript. Please ensure the figure is included in the final version.","section":"Figure 1"},{"comment":"The abstract phrase 'many single-qubit errors commute with every stabilizer or logical Pauli below this weight' could be misread. Theorem 4 shows that for a set J of qubits, the single-qubit errors on J commute with every element of N_{\\le L}(S), i.e. with all normalizer elements of weight at most L, not with every stabilizer or logical Pauli individually. Consider clarifying the wording.","section":"Abstract, Theorem 4"},{"comment":"The master equation (S96) combines the Hamiltonian and the dephasing dissipator. Because the signal generator G and the dephasing operators commute, the solution factors into unitary evolution followed by dephasing; stating this explicitly would help readers understand the Kraus form (S108).","section":"Appendix S3, Eq. (S96)"}],"recommendation":"minor_revision","confidential_remarks":"I found no technical error in the central derivations. The main issue is the abstract's overstatement of Theorem 2: the universal claim about concatenation is not supported by the k_j = 1 assumption in the theorem. Once the abstract is qualified (and the genuinely-level-D wording in the abstract is fixed), the paper is suitable for publication. The metrological no-go in Theorem 5 is the strongest result and its proof in Appendix S3 is sound within the stated noise model."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the paper is worth engaging. It proves new quantitative restrictions on transversal non-Clifford gates and, more importantly, a no-go for beyond-SQL metrology under signal-aligned noise that does not rely on the quantum Cramér-Rao bound. I checked the main theorems and the S3 derivation; the arguments hold together. The abstract needs a small correction, and the noise model is narrower than the prose sometimes suggests, but neither undermines the main result.\n\nWhat's new: Theorem 5 is a distinguishability bound on arbitrary multi-round protocols with signal-aligned dephasing and flagged erasures; it gives T = Ω(max{1/(εn), sqrt(p_e/(1-p_e))/(ε√n), γ/(ε²n)}). That rules out any asymptotic speedup over the SQL under constant-strength aligned noise, including protocols that use QEC, biased estimators, adaptive control, or nonstabilizer codes. Theorems 1-4 are also genuinely new: the 2^D check-weight bound, the concatenation bound, and the syndrome-irreducibility result. The proofs in the appendices are detailed, and I did not find a concrete error. The use of the channel-extension QFI bound is appropriate and the gauge choice β=0 is legitimate.\n\nSoft spots: The abstract says 'any r-level concatenated realization' but Theorem 2 requires k_j=1 for j≥2. That should be fixed; the theorem is still strong but not as universal as advertised. The no-go is explicitly conditional on γ and p_e being constants independent of n and ε; if you allow them to decay, the SQL terms shrink and the conclusion can fail. The authors state this, so it's a boundary, but readers should not walk away thinking all noise forever. The physical argument that interface erasure is unavoidable leans on the authors' own prior work (Refs. 32-34); that doesn't affect the math, but a referee may want a more independent justification before accepting the 'physically unavoidable' claim. The proofs are long, and the imported bounds from Kubica-Demkowicz-Dobrzański and Taddei et al. are not re-derived; that's normal, but it means confidence is moderate rather than high.\n\nWho it's for: anyone working on error-corrected metrology or fault-tolerant non-Clifford gates. It deserves a serious referee. I would recommend sending it to peer review, with a request to correct the abstract and to state the noise-model boundaries in the main text.","headline":"New no-go for beyond-SQL metrology under signal-aligned noise, with solid proofs; minor abstract overstatement in Theorem 2.","tokens_in":28923,"tokens_out":4263,"would_cite":true,"duration_ms":40903,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["03.67.Pp","03.65.Ta"],"model":"deepseek-v4-flash","headline":"Transversal non-Clifford gates force high-weight stabilizer checks, and constant signal-aligned noise pushes all quantum sensing back to the standard quantum limit.","keywords":["quantum metrology","standard quantum limit","Heisenberg limit","transversal gates","Clifford hierarchy","stabilizer codes","signal-aligned noise","fault tolerance"],"falsifier":"A concrete falsifier would be a family of [[n,k,3]] stabilizer codes supporting a transversal T gate (level 3) whose stabilizer generating set has maximum check weight < 8; Theorem 1 says no such family exists. Alternatively, a DC sensing experiment with n qubits under constant signal-aligned dephasing gamma and interface erasure probability p_e that distinguishes two frequencies separated by epsilon in time T = o(1/(epsilon sqrt(n))) with constant success probability would falsify Theorem 5.","tokens_in":28071,"feed_emoji":"🔬","tokens_out":8270,"duration_ms":70106,"temperature":0.7,"pith_summary":"This paper tries to close the last known route to noise-robust quantum advantage in metrology: encoding sensors in quantum error-correcting codes so that the signal acts as a transversal logical gate. It proves two families of results. First, codes that transversally implement genuinely non-Clifford logical gates—required for small-angle rotations—must have stabilizer checks whose weight grows with the Clifford-hierarchy level and with the inverse square of the rotation angle, so fault-tolerant syndrome extraction becomes increasingly nonlocal. Second, a model-independent no-go theorem shows that under constant-strength dephasing aligned with the signal generator (plus an interface erasure floor), any multi-round sensing protocol—including biased estimators, nonstabilizer or approximate encodings, memory, adaptivity, and intermediate measurements—has output distinguishability bounded by epsilon times a SQL-scale quantity, forcing interrogation time to scale at least as the standard quantum limit. If correct, these results rule out asymptotic Heisenberg scaling in canonical DC and AC single-parameter sensing under realistic noise, while leaving finite-size and bosonic or multi-parameter settings open.","feed_headline":"Aligned noise rules out Heisenberg-limited quantum sensing","feed_subtitle":"Even with error correction, biased estimators, and adaptivity, interrogation time must scale as the standard quantum limit.","key_machinery":"The central devices are the minimax stabilizer weight lambda_S (the smallest possible largest weight of a stabilizer generator) and the codespace signal susceptibility Xi_G = inf_c ||(G - cI)Pi||^2_op, which measures how much a signal generator G can act nonclassically on the codespace. The proof of the code restrictions uses a dyadic normal form (reducing any transversal unitary to local Cliffords and a diagonal gate with dyadic phases) and a descent argument showing that if all checks have weight < 2^D, the logical gate drops to level D-1 in the Clifford hierarchy. The metrological no-go uses a quantum-Fisher-information bound for adaptive channel-extension protocols (Lemma 9) applied to t","core_discovery":"For a stabilizer code of distance d >= 3, the authors show that a transversal product of single-qubit unitaries that induces a genuinely level-D logical gate forces every generating set of the stabilizer to include a check of weight at least 2^D; concatenation cannot evade this, since any r-level concatenated realization must have r <= floor(log_2 n / D). They further show that a transversal rotation by angle theta inducing a nontrivial logical action requires irreducible stabilizers of weight Omega(1/(n theta^2)), and that the syndromes of many single-qubit errors can only be reconstructed by measuring such high-weight checks. Finally, in a noise model with constant-strength dephasing align","pith_inferences":["If the noise parameters gamma or p_e are allowed to decay with n or epsilon, the no-go bound weakens; this suggests a possible crossover regime where sufficiently fast-decaying noise might allow Heisenberg-like scaling at finite size, and the paper's bounds quantify the threshold.","The syndrome-irreducibility result has implications beyond sensing: any fault-tolerant scheme that uses small-angle transversal rotations (for instance, certain approaches to non-Clifford gate synthesis) will need high-weight stabilizer measurements, so low-weight syndrome-extraction tricks cannot work regardless of the logical gate.","The bounds are tight for Clifford-hierarchy level (codes achieve lambda_S = 2^D), so the obstruction is not code existence but fault-tolerant measurement; this shifts design effort toward weight-reduction or subsystem encodings, which the paper leaves open.","A natural next test is to extend the no-go theorem to bosonic or continuous-variable encodings, or to multi-parameter and nonlinear sensing tasks; the paper leaves open whether those settings can sustain noise-robust quantum advantage through mechanisms distinct from Heisenberg scaling."],"forward_implications":["No family of stabilizer codes with bounded-weight checks can support transversal sensing beyond the SQL: for interrogation time O(1/(epsilon n^alpha)) with alpha > 1/2, the required minimax stabilizer weight grows as n^{2alpha-1} and diverges with n.","Concatenation cannot restore beyond-SQL transversal sensing: for alpha > 1/2, any r-level concatenated code family has r <= 1 for sufficiently large n.","Under constant-strength signal-aligned dephasing plus interface erasure, arbitrary multi-round protocols (with QEC, nonstabilizer codes, approximate encodings, adaptivity, intermediate measurements, biased estimators) require T = Omega(max{1/(epsilon n), sqrt(p_e/(1-p_e))/(epsilon sqrt(n)), gamma/(epsilon^2 n)}), so no asymptotic advantage over the SQL survives.","Transversal sensing codes cannot correct signal-aligned noise with vanishing logical error: in the high-precision regime, the optimal recovery error remains bounded away from zero as n grows.","The same no-go applies to AC amplitude sensing with known envelope, via the toggling-frame argument, so the result covers both canonical DC and AC phase estimation.","The bounds quantify a tradeoff between noise strength and sensing efficiency, so they also serve as design constraints for finite-size quantum sensors: any preasymptotic advantage must come from tuning code size, noise rates, and target precision within the derived crossover regimes."],"fun_headline_variants":["Aligned noise prevents Heisenberg-limited sensing","No quantum advantage from aligned noise","Beyond-SQL metrology impossible under aligned noise","Constant aligned noise kills Heisenberg scaling","Error correction cannot restore quantum sensing advantage"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The no-go theorem rests on the assumption of a constant-strength noise floor aligned with the signal: independent single-qubit dephasing of fixed strength gamma during every interrogation, plus an independent flagged erasure of fixed probability p_e at every processing interface. If either rate decays with n or epsilon, or the noise is not aligned with the signal generator, the bound shrinks and the Heisenberg-limited window can reopen; the concatenation result in Theorem 2 a","fun_headline_variants_meta":{"raw":{"variants":["Aligned noise prevents Heisenberg-limited sensing","No quantum advantage from aligned noise","Beyond-SQL metrology impossible under aligned noise","Constant aligned noise kills Heisenberg scaling","Error correction cannot restore quantum sensing advantage"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000696,"raw_usage":{"total_tokens":3065,"prompt_tokens":905,"completion_tokens":2160,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":649,"completion_tokens_details":{"reasoning_tokens":2107}},"tokens_in":649,"tokens_out":2160,"duration_ms":14464,"temperature":1.0,"reasoning_tokens":2107,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T09:07:42.930375+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A concrete falsifier would be a family of [[n,k,3]] stabilizer codes supporting a transversal T gate (level 3) whose stabilizer generating set has maximum check weight < 8; Theorem 1 says no such family exists. Alternatively, a DC sensing experiment with n qubits under constant signal-aligned dephasing gamma and interface erasure probability p_e that distinguishes two frequencies separated by epsilon in time T = o(1/(epsilon sqrt(n))) with constant success probability would falsify Theorem 5.","supporting_citations":[],"review_version":1}