{"id":"3e7106c7-467c-4c22-a205-80ed7c45322f","arxiv_id":"2611.17056","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Subspace twirling can simultaneously suppress time-dependent noise and achieve super-Heisenberg scaling in quantum metrology without adaptive controls.","lead":"By averaging random unitaries over a subspace rather than the whole Hilbert space, this paper shows how to suppress environmental noise while preserving a time-dependent signal, and to reach super-Heisenberg scaling without adaptive controls. The result is a concrete design criterion for practical, noise-robust quantum sensors.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek/deepseek-v4-flash","headline":"Theorem 2's sufficiency proof assumes a single fixed vector |v⟩ can witness the norm growth ||K||=Ω(t^k), but the stated hypothesis does not imply this; the maximizing eigenvector can rotate with t, so the 'if' direction is false as written.","rationale":"The central claim b29 depends on Theorem 2's sufficiency direction. The reader's weakest assumption (bang-bang control) is explicitly acknowledged by the paper and is a practical limitation, not the weakest point of the argument. The unresolved point is a quantifier gap inside the proof of Theorem 2: the theorem's stated condition ||K(θ,t)||=Ω(t^k) only bounds the operator norm, while the proof needs a fixed vector |v⟩ with |⟨v|K(θ,t)|v⟩|=Ω(t^k) for all t. A simple rotating-field qubit Hamiltonian satisfies the norm condition but has no such fixed vector, so the claimed 'if and only if' fails under the natural fixed-subspace reading. The paper's own numerical example is unaffected because its K=(t²/2)σ_x has a time-independent eigenvector; this is why the gap is not visible in Fig. 1. Since the framework is likely repairable by adding an explicit fixed-witness condition (or by declaring V_t per total time), the appropriate disposition is conditional acceptance with a corrected Theorem 2 rather than outright rejection.","tokens_in":19277,"tokens_out":24237,"duration_ms":167392,"concrete_test":"Symbolically test Theorem 2 for a qubit with H_S(θ,t)=θ t(cos t σ_z+sin t σ_x) and no noise. Compute K(θ,t)=∫_0^t s(cos s σ_z+sin s σ_x)ds; verify ||K(θ,t)||=t+O(1)=Ω(t), so the norm condition holds with k=1. For any fixed one-dimensional V with basis {|v⟩,|v⊥⟩}, compute T[HS]=a(t)σ_z with a(t)=θ t ⟨v|cos t σ_z+sin t σ_x|v⟩. Then the optimal QFI for this fixed V is F_V(t)=4(∫_0^t ∂_θ a(s)ds)^2. Show that for every fixed |v⟩ the integral is t(n_z sin t−n_x cos t)+O(1), which crosses zero at arbitrarily large t, so max_V F_V(t) is not Ω(t²). If confirmed, the fixed-V reading of Theorem 2 is false; if the authors instead intend a family V_t selected per total time, the theorem statement and proof need to be amended accordingly.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"Supplement §III, proof of Theorem 2, asserts: 'let |v⟩ be a normalized vector such that ||K(θ,t)||=|⟨v|K(θ,t)|v⟩|=Ω(t^k)' and then fixes V=span{|v⟩} (Eqs. S33–S36). The theorem's hypothesis only guarantees that the operator norm grows; it does not guarantee that one fixed vector |v⟩ can witness this growth for all t. The maximizing eigenvector of K(θ,t) can rotate with t. Concrete counterexample on a qubit: take H_S(θ,t)=θ t(cos t σ_z + sin t σ_x). Then K(θ,t)=∫_0^t s(cos s σ_z+sin s σ_x)ds has ||K(θ,t)||≈t=Ω(t), so the norm condition holds with k=1. But for any fixed one-dimensional V (the only strict subspace of a qubit), the twirled Hamiltonian is diagonal in a fixed basis with diagonal entry a(t)=θ t(n_z cos t+n_x sin t), and the achievable QFI is 4(∫_0^t ∂_θ a(s)ds)^2. This integral equals t(n_z sin t−n_x cos t)+O(1), which is not Ω(t): it has zeros at arbitrarily large t. Hence no fixed V yields F=Ω(t²), contradicting the stated 'if' direction. If the authors intend V to be chosen separately for each total interrogation time t, that is a different statement and must be said explicitly; as written, the proof uses one fixed |v⟩, and the advertised 'check the operator norm' criterion is not sufficient. The qubit example in the paper has K=(t²/2)σ_x with a fixed witness vector, so it is not affected; but the general iff in Theorem 2, and the headline claim b29, are not supported as written.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek/deepseek-v4-flash","summary":"The paper proposes using subspace-twirled dynamical decoupling to achieve super-Heisenberg scaling (SHS) in time-dependent quantum metrology while simultaneously suppressing noise and removing the need for adaptive controls. The main results are two theorems: Theorem 1 gives necessary and sufficient conditions for subspace twirling to suppress noise while preserving the signal; Theorem 2 gives necessary and sufficient conditions for achieving Heisenberg or super-Heisenberg scaling without adaptive controls via subspace twirling, based on the operator norm growth of the integrated derivative Hamiltonian. The paper illustrates these results with a qubit example (rotating field with dephasing noise) and numerical simulations showing convergence to the theoretical SHS bound as the number of control cycles increases.","tokens_in":19806,"tokens_out":2529,"duration_ms":157648,"significance":"If correct, the results provide a general framework for achieving super-Heisenberg scaling in noisy time-dependent quantum metrology without adaptive controls, which is a significant advance. The theorems are stated in a checkable form (vector condition for noise suppression, operator norm condition for SHS), which could guide experimental design. The paper also provides a concrete qubit example and numerical support.","major_comments":[{"comment":"The sufficiency proof of Theorem 2 assumes that a single fixed vector |v⟩ can witness the operator norm growth ‖K(θ,t)‖ = |⟨v|K(θ,t)|v⟩| = Ω(t^k). However, the theorem's hypothesis only guarantees that the operator norm grows; it does not guarantee that one fixed vector can achieve this for all t. The maximizing eigenvector of K(θ,t) can rotate with t. The proof then fixes V = span{|v⟩} and constructs the twirl based on this single vector. This is insufficient to prove the 'if' direction as stated. A concrete counterexample on a qubit: take H_S(θ,t) = θ t (cos t σ_z + sin t σ_x). Then K(θ,t) = ∫_0^t s (cos s σ_z + sin s σ_x) ds has ‖K(θ,t)‖ ≈ t = Ω(t), so the norm condition holds with k=1. But for any fixed one-dimensional V (the only strict subspace of a qubit), the twirled Hamiltonian is diagonal in a fixed basis with diagonal entry a(t) = θ t (n_z cos t + n_x sin t), and the achievable QFI is 4(∫_0^t ∂_θ a(s) ds)^2. This integral equals t(n_z sin t - n_x cos t) + O(1), which is not Ω(t): it has zeros at arbitrarily large t. Hence no fixed V yields F = Ω(t^2), contradicting the stated 'if' direction. The qubit example in the paper has K = (t^2/2)σ_x with a fixed witness vector, so it is not affected; but the general iff in Theorem 2 is not supported as written. The authors must either revise the theorem statement to include an additional condition (e.g., that the maximizing eigenvector is constant in time) or modify the proof to handle the rotating case (e.g., by allowing V to depend on t, which would break the non-adaptive claim). This is a load-bearing issue for the central claim that the operator norm condition is sufficient.","section":"Supplement §III, proof of Theorem 2 (sufficiency)","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"Related to the above, the proof of necessity (Supplement §III) uses the inequality |∑_j ⟨e_j|K|e_j⟩| ≤ m‖K‖ to conclude ‖K‖ = Ω(t^k) from the assumption that SHS is achievable. This direction is correct. However, the sufficiency direction is the one that is problematic. The paper's headline claim (b29) that 'subspace twirling can simultaneously (i) suppress time-dependent noise, and (ii) remove the requirement for adaptive controls while reaching SHS' relies on the sufficiency of the norm condition. If the sufficiency proof is flawed, the central claim is not established in full generality.","section":"Theorem 2 statement and proof","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The proof of necessity for Theorem 1 constructs a vector |v⟩ in V⊥ such that ⟨v|S_i|v⟩ = 0 and ⟨v|H_S(θ,t)|v⟩ is not constant. This part appears sound. However, the sufficiency proof constructs V⊥ = span{|v⟩} and V as its orthogonal complement. The resulting twirl T_V is then applied. The proof claims that T_V(S_i) = 0 because ⟨v|S_i|v⟩ = 0. But Eq. S11 shows T_V(S_i) has a block proportional to 11_V with coefficient -⟨v|S_i|v⟩/(d-1) and a block proportional to |v⟩⟨v| with coefficient ⟨v|S_i|v⟩. If ⟨v|S_i|v⟩ = 0, then indeed T_V(S_i) = 0. This is correct. However, the proof also needs to ensure that the twirl over V (the (d-1)-dimensional subspace) is a valid subspace twirl (i.e., averaging over all unitaries on V). The paper states that for any finite-dimensional subspace there exists a unitary 1-design, so this is fine. The sufficiency proof seems valid. No major issue here.","section":"Theorem 1 proof (Supplement §I)","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The numerical simulation uses a single bosonic bath mode with a specific dephasing model. The authors acknowledge that the bang-bang limit is idealized and that finite control cycles lead to deviation from the theoretical bound. This is appropriate. However, the simulation only demonstrates the qubit example, which is a special case where the witness vector is fixed (|+⟩ for σ_x). It does not test the general sufficiency condition of Theorem 2 in cases where the maximizing eigenvector rotates. The paper would be strengthened by a numerical example where the norm condition holds but the witness vector is time-dependent, to illustrate whether the theorem's claim still holds or fails. As it stands, the numerical evidence does not address the identified gap.","section":"Numerical simulation (Fig. 1)","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The paper assumes bang-bang control (instantaneous, error-free pulses) and that the parameter θ is contained entirely within the signal Hamiltonian. These are stated explicitly. The paper also assumes finite-dimensional Hilbert space. These are reasonable for a theoretical proposal. The main concern is the logical gap in Theorem 2's sufficiency, not the assumptions per se.","section":"General scope and assumptions","authors_acknowledge":null,"authors_acknowledge_p":null}],"minor_comments":[{"comment":"The abstract states 'We first prove necessary and sufficient conditions for subspace twirling to preserve the encoding signal while simultaneously removing environmental noise.' This refers to Theorem 1, which appears correct. The second claim about SHS without adaptive controls refers to Theorem 2, which has the issue noted above. The abstract should be updated if the theorem is revised.","section":"Abstract and Introduction","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The block matrix form in Eq. (10) writes T[H_S(θ,t)] with entries μ(θ,t)/Tr(11_V) 11_V and -μ(θ,t)/Tr(11_{V⊥}) 11_{V⊥}. The notation Tr(11_V) is the dimension of V. This is fine. However, the text says 'where μ(θ,t) is some function of θ, t.' It would be helpful to explicitly relate μ to the original Hamiltonian, e.g., μ(θ,t) = (1/dim(V)) Tr_V(H_S(θ,t))? Actually from the derivation, μ(θ,t) = (1/dim(V)) Tr(Π_V H_S(θ,t) Π_V) = (1/dim(V)) ∑_{i=1}^{dim(V)} ⟨e_i|H_S|e_i⟩. This is clear from the supplement but could be stated in the main text for clarity.","section":"Eq. (10) and surrounding text","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The paper says 'Suppose we are estimating a low frequency θ, so we are primarily interested in the measurement sensitivity locally around θ = 0.' This is fine, but the subsequent evaluation of K(θ,t)|_{θ=0} uses the derivative with respect to θ and then sets θ=0. This is standard for local estimation. However, the operator norm condition ‖K(θ,t)‖ = Ω(t^k) must hold for the true value of θ (which is unknown). The paper evaluates at θ=0 as an example. This is acceptable for illustration, but the theorem requires the condition to hold at the true parameter value. The authors should clarify that in practice one would evaluate the condition at a guessed value or over an interval.","section":"Page 4, sentence b107","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The expression for ∂_θ T[H_S(θ,t)] in block matrix form uses β(θ,t)/m 11_V and -β(θ,t)/(d-m) 11_{V⊥}. This assumes that the trace of ∂_θ H_S over V is β(θ,t) and over V⊥ is -β(θ,t) (since total trace is zero). This is correct if H_S is traceless. The paper earlier assumes Gi are traceless, so H_S is traceless. Good.","section":"Supplement Eq. S21","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"Reference [32] is to the Supplemental Material, which is fine. Reference [26] is to an arXiv preprint (Zeng et al., 2026). This is acceptable as a citation to related work. No issues.","section":"References","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The figure shows QFI vs time for various n. The legend says 'no control', 'n=5', 'n=15', 'n=25', and 'Theoretical t^4'. The curves for finite n show deviation from t^4 at larger t. The authors note that for large t relative to n, the bang-bang limit is not approached. This is a known limitation of finite control cycles. The figure is clear.","section":"Figure 1","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The abstract contains a typo: 'super-Heisenberg scaling in time-dependent q uantum sensing' has an extra space. Also in the main text, 'magnatic' should be 'magnetic' (Fig. 1 caption). These are minor.","section":"Typographical issues","authors_acknowledge":null,"authors_acknowledge_p":null}],"recommendation":"major_revision"},"author_rebuttal":null,"desk_editor":{"model":"deepseek/deepseek-v4-flash","letter":"The paper has a real idea: use subspace twirling to suppress noise and remove the need for adaptive controls in time-dependent metrology, and it gives a clean necessary condition (Theorem 1) for when noise can be suppressed while preserving the signal. The qubit example is concrete and the numerics show the expected t^4 scaling as the number of control cycles increases. The authors are upfront about the bang-bang assumption and the finite-cycle limitation.\n\nBut there is a load-bearing problem in Theorem 2. The sufficiency proof picks a single fixed vector |v⟩ that witnesses the operator norm growth ‖K(θ,t)‖ = Ω(t^k) and then fixes V = span{|v⟩}. The theorem's hypothesis only guarantees that the operator norm grows; it does not guarantee that one fixed vector can witness that growth for all t. The maximizing eigenvector of K(θ,t) can rotate with t. The stress-test counterexample on a qubit (H_S = θ t (cos t σ_z + sin t σ_x)) satisfies the norm condition with k=1, but no fixed one-dimensional V yields QFI = Ω(t^2). The paper's own qubit example has K ∝ t^2 σ_x with a fixed witness, so it is not affected, but the general 'if' direction of Theorem 2 is false as written. The authors would need to either prove that the witness vector can be chosen independently of t (which the hypothesis does not imply) or change the statement to allow V to depend on t, which would be a different claim.\n\nThis is not a minor gap. The headline claim b29—that subspace twirling can simultaneously suppress noise and achieve SHS without adaptive controls—rests on Theorem 2. If the sufficiency direction is unsupported, the paper's central advertised result is not established in general. The necessity direction (if SHS is achievable then the norm condition holds) is fine.\n\nThe paper is otherwise well-structured, the proofs in the supplement are clear where they do not rely on this step, and the authors state their assumptions honestly. But the flaw in Theorem 2's sufficiency is structural. I would not send this to peer review as is; the authors need to fix the theorem statement or the proof before it is ready for referees.","headline":"The paper's central claim—that a single fixed vector can witness operator-norm growth for all t—is not supported by the proof of Theorem 2, which makes the 'if' direction unsound as written. The qubit example works, but the general theorem overreaches.","tokens_in":20268,"tokens_out":588,"duration_ms":45326},"pith_extraction":null,"created_at":"2026-11-13T10:42:30.189980+00:00","model_set":{"reader":"deepseek/deepseek-v4-flash"},"pipeline_version":"pith-pipeline@v1.4.1-daily-deepseek-or-medium","input_kind":null,"input_chars":null,"input_total_chars":null,"passes_missing":null,"published_standing":null,"published_confidence":null,"objection_audit":null,"run_record":null,"basis":{"kind":"not_recorded","source":null,"chars_given":null,"chars_total":null},"receipt":null,"falsifier":null,"supporting_citations":[],"lane_accuracy":{"status":"not yet measured","lane":{"model":"deepseek/deepseek-v4-flash","input_kind":null},"url":"https://pith.science/accuracy"},"review_version":1}