{"id":"b75a8f50-0aff-46f4-979e-9e65facc4cc8","arxiv_id":"2501.01917","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Dynamical decoupling can recover Heisenberg scaling of quantum Fisher information (QFI ~ t^2) in non-Markovian open quantum systems when the control satisfies specific decoupling conditions.","lead":"Quantum measurements usually lose their best precision when the environment has memory, but this paper shows that carefully timed control pulses can restore the ultimate precision limit. It provides conditions for when such pulses work and tests them on a model atom-in-cavity system.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central QFI scaling is proven for the average-Hamiltonian limit only; finite-pulse corrections are unquantified and may destroy t^2 scaling.","rationale":"The paper's central argument is the factorization of the controlled evolution and the resulting QFI ∝ t^2. Both rest on the average-Hamiltonian approximation, which is only valid in the T→0 limit. The paper acknowledges this but provides no quantitative control of the error; the example's numerics cover only short times and contain a suspicious sign factor in the main-text memory kernel. These issues do not invalidate the idealized claim, but they justify the reader's CONDITIONAL verdict: the proof needs an explicit error bound or a corrected derivation of the finite-T dynamics. The reader's weakest assumption already identified the average-Hamiltonian issue, so we agree.","tokens_in":22568,"tokens_out":29315,"duration_ms":292190,"concrete_test":"Using the recurrence of Appendix C, compute the exact QFI for the damped Jaynes-Cummings model under ideal π pulses at a fixed finite interval (e.g., T=0.05/λ) for times t up to 50/λ, and estimate the local scaling exponent α(t)=d ln F_Q/d ln t. If α(t) is significantly below 2 over a wide time window, or if F_Q/t^2 decays, the finite-T dynamics does not sustain Heisenberg scaling, so the central claim is restricted to the ideal limit.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Eqs. 11–14 replace the time-ordered controlled evolution by e^{-i nT H_eff} (average Hamiltonian). The factorization in Eq. 17 and the QFI result in Eqs. 19–22 are statements about this idealized effective Hamiltonian. For any finite pulse interval T, the exact evolution is a product of exponentials e^{-iT H_k} with non-commuting H_k, so the Magnus expansion has residual terms of order T^2 per interval, proportional to commutators such as [H_S, H_SE]. These accumulate to a system-environment coupling of order tT, which entangles system and bath and can reduce the reduced-state QFI below 4t^2 Var[G_eff]. The paper supplies no error bound showing F_Q ≈ 4t^2 Var[G_eff] for finite T over the relevant time scale; Fig. 3 shows only short times and a visible gap. Additionally, Eq. 31 contains the factor (-1)^{n+τ/T} with τ/T generally non-integer, which is not a well-defined parity factor; this appears inconsistent with the recurrence in Appendix C, so the example's derivation and numerics need independent verification.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies quantum frequency estimation in open quantum systems with non-Markovian noise. The authors propose to apply dynamical decoupling (DDT) to the system, with the aim of removing the system-environment interaction while keeping the parameter-dependent part of the system Hamiltonian active. They formulate Theorem 1, giving necessary and sufficient conditions for the existence of a sequence of unitary controls satisfying two conditions: (i) the averaged interaction takes a trivial c 1_S ⊗ J_E form, and (ii) the averaged signal Hamiltonian remains non-trivial. Under these conditions and an average-Hamiltonian approximation, they show the effective evolution factorizes, the generator of the parameter is approximately t G_eff, and the QFI scales as 4 t^2 Var[G_eff], i.e., Heisenberg scaling, without assuming Markovian dynamics. They illustrate the approach with the damped Jaynes-Cummings model with a Lorentzian spectral density and detuning, reporting that pi-pulse DDT restores near-Heisenberg QFI. A detailed proof of Theorem 1 is given in Appendix A, and the example is solved in Appendices B and C.","tokens_in":22855,"tokens_out":13034,"duration_ms":122804,"significance":"If the results were fully correct, they would be a valuable generalization of dynamical-decoupling-based metrology to non-Markovian environments. The clean separation between the decoupling condition (Eq. 15) and the signal-preservation condition (Eq. 16) is a useful organizing principle, and the sufficient conditions in Corollary 1.1 are concrete and checkable. The paper also correctly stresses the importance of the global QFI bound (Eq. 8), and the Jaynes-Cummings model is an apt non-Markovian testbed. The manuscript is clearly written and the main idea is easy to grasp. However, the central theorem's proof contains an unjustified uniformity assumption, the finite-T corrections to the average-Hamiltonian limit are not quantified, and the example's equations contain sign-factor inconsistencies that call the numerical demonstration into question. These are substantial but potentially repairable issues.","major_comments":[{"comment":"After constructing the channel Φ' = Π ∘ Φ, the proof discretizes the probability density p'(x) into weights m_j/n, and then states 'Assume that p'(x) is the uniform distribution' to conclude m_j/n = 1/n (Eqs. A6-A8). This assumption is not valid for a general mixed unitary channel, and it is exactly what is needed to turn an arbitrary convex combination into the equal-weight average required by the theorem. The discretization step is also only approximate, whereas Theorem 1 is an exact existence statement. As written, the sufficiency direction is not proven. I suggest replacing this step with a multiset construction (repeating a unitary m_j times to obtain an equal-weight average) and, if exact weights are not achievable, stating the theorem with an approximation guarantee; alternatively, restrict the theorem to channels admitting a finite equal-weight representation. The necessity direction appears sound.","section":"Appendix A, proof of Theorem 1 (sufficiency)"},{"comment":"The central QFI scaling F_ω^(Q) ∝ t^2 is derived after replacing the time-ordered product of n interval evolutions by the single exponential e^{-i n T (H_S^eff ⊗ 1_E + 1_S ⊗ H_E + H_SE^eff)}. This is the average-Hamiltonian approximation. For finite pulse spacing T, the Magnus expansion contains corrections of order T per interval (e.g., commutators involving H_SE), and these accumulate over n = t/T intervals, producing a residual system-environment coupling of order t T. The manuscript does not bound this error, and the concluding sentence that increasing the pulse number 'can further restore ideal precision' is not backed by a quantitative statement. To claim restoration of Heisenberg scaling in a real protocol, the authors need a result of the form |F_ω^(Q)(t) - 4 t^2 Var[G_eff]| ≤ ε(t,T) with ε → 0 as T → 0 uniformly on the relevant time scale, or an explicit demonstration that the leading correction does not destroy the t^2 term. Figure 3 shows only short times and a visible gap, so it does not by itself establish asymptotic t^2 scaling.","section":"Section III.A, Eqs. (11)-(14) and (19)-(22)"},{"comment":"The factor (-1)^{n+τ/T} is not well-defined because n = t/T and τ/T are not integers in general; (-1)^x for non-integer x is a complex phase, not a parity sign. Both Eq. (31) and Eq. (32) contain this factor, so the product in the convolution is (-1)^{2n+2τ/T} = e^{2 i π τ/T}, which is not identically 1 and does not yield the standard sign pattern expected from dynamical decoupling. Relatedly, Eq. (33), obtained by differentiating Eq. (31), is independent of n and T, which is inconsistent with the claim that the pulses alter the dynamics; the numerical results in Fig. 3 therefore need to be re-derived and re-verified. The recurrence in Appendix C is the natural place to check the correct piecewise-constant sign factors, but the present formulation does not make this consistent.","section":"Section IV.B, Eqs. (31)-(32)"},{"comment":"For the chosen H_C(t) = (π/2) Σ_{k=0}^{n-1} δ(t-kT) σ_z, the control unitary after k pulses is U_C(kT) = (-i)^k σ_z^k, so the interaction in the k-th interval is (-1)^k (σ_+ B + σ_- B†), up to an irrelevant global phase. Hence the average in Eq. (15) is proportional to (1/n) Σ_{k=0}^{n-1} (-1)^k (σ_+ B + σ_- B†), which vanishes only for even n; for odd n the decoupling condition is not satisfied. The statement that 'It can be readily verified that the control unitary evolution satisfies the DDT condition in Eq. 15' is therefore true only for a restricted parity of n (or in a suitable limit), and this restriction should be stated and incorporated in the derivation.","section":"Section IV.B, control sequence and Eq. (15)"}],"minor_comments":[{"comment":"The scaling is stated as 'δ2ω ∝ /(N t)2' (twice in the same paragraph); it should read δ²ω ∝ 1/(N t)².","section":"Section I"},{"comment":"The eigenvectors |ψ_k⟩ and eigenvalues λ_k should be those of the initial system state ρ_S(0) (or the state before encoding), not of ρ_S(ω,t); although the final state is related by a unitary commuting with A, this should be stated to avoid ambiguity.","section":"Section III.A, Eq. (20)"},{"comment":"The integrand should contain e^{-i(ω_0-ω_k)τ}, not e^{-i(ω_0-ω_k)t}; as written, the subsequent substitution into Eq. (B7) does not produce the convolution in Eq. (B10).","section":"Appendix B, Eq. (B9)"},{"comment":"There is a typo 'Not that the above expression is nontrivial' which should read 'Note that...'; also the notation U(kT) versus U_C(kT) is inconsistent across the theorem statement and the proof.","section":"Appendix A"},{"comment":"The notation for τ_n, the arguments of the hyperbolic functions in the expression for B_n, and the definitions of η_± are hard to follow; for example, Eq. (C3) contains sinh(d T n / 2), which appears to be a typo for sinh(d (n-1) T / 2). Please clarify and check for typographical errors.","section":"Section IV.B, Eq. (34) and Appendix C"}],"recommendation":"major_revision","confidential_remarks":"The paper addresses a relevant problem and the main idea is plausible, but I have serious reservations about the proof of Theorem 1 and about the correctness of the example's equations. In my view the manuscript needs a careful revision of Appendix A (replacing the uniform-distribution assumption with a valid discretization or a multiset argument) and a complete re-derivation of the Jaynes-Cummings DDT example, including a comparison of the corrected equations with the reported numerics. I would not recommend acceptance without these changes. The framework is worth pursuing, and I do not see a fundamental impossibility, so rejection is not warranted at this stage. The authors should also make the relationship to prior DDT metrology work (e.g., Refs. [66] and [68]) more explicit, since the novelty lies in the general non-Markovian characterization rather than in the use of DDT per se."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this paper is worth refereeing, but it needs substantial revision before I'd trust the central claims. The headline effect—dynamical decoupling restores Heisenberg scaling in noisy metrology—was already demonstrated in Refs. [66] and [68]. What's new here is the necessary-and-sufficient condition theorem for control unitaries that decouple H_SE while keeping the signal Hamiltonian nontrivial, and a detuned Jaynes-Cummings example. The theorem is a useful abstraction, and Corollary 1.1's orthogonality condition is simple and applicable. The example correctly shows the non-Markovian QFI oscillations and the recovery toward t^2 scaling with DD.\n\nThe soft spots are real but fixable.\n\nFirst, the sufficiency proof of Theorem 1 (Appendix A) assumes the probability density p'(x) is uniform when converting the mixed-unitary channel Phi' into an equal-weight finite average. That is not generally valid, and no argument is given for why uniform choice is without loss of generality. The proof can likely be repaired by using rational weights and repeating unitaries, but as written it's a gap in an 'if and only if' claim.\n\nSecond, Eqs. (31)-(32) contain a factor (-1)^{n+tau/T} with n = t/T, so the exponent is generally non-integer. That factor is not well-defined. The intended sign is presumably piecewise constant between pulses, e.g., (-1)^{floor(tau/T)}. This needs to be stated correctly; as written, the integro-differential equation and the numerics built on it are suspect.\n\nThird, the central QFI scaling (Eqs. 17-22) is derived in the average-Hamiltonian limit: ideal delta pulses, T->0. The paper gives no error bound for finite T. Figure 3 shows a visible gap between the DDT curve and the noiseless t^2 line, and the simulation parameters (pulse interval T, number of pulses) aren't fully specified in the text. This makes the example hard to reproduce as-is.\n\nOn the positive side, the paper is clearly written, the connection to prior work is honest, and I don't see circular reasoning. The example's qualitative behavior is consistent with what I'd expect from DD in a non-Markovian reservoir.\n\nWho's it for: quantum metrology researchers using control techniques on non-Markovian systems. The theorem, once fixed, could be a useful reference. I would send it to peer review rather than desk reject, but I'd require a corrected proof, a well-defined parity factor, and ideally a finite-T error estimate before accepting.","headline":"Worth refereeing with revisions: the paper has a useful theorem and a clean example, but the sufficiency proof has a gap and the Jaynes-Cummings equation contains an ill-defined sign factor.","tokens_in":23326,"tokens_out":4852,"would_cite":false,"duration_ms":49357,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that carefully chosen dynamical-decoupling pulses restore Heisenberg-limited precision ($t^2$) in open quantum systems even when the environment has memory, without assuming Markovian noise.","keywords":["Heisenberg limit","quantum metrology","dynamical decoupling","non-Markovian noise","quantum Fisher information","open quantum systems","Jaynes-Cummings model","quantum control"],"falsifier":"Simulate or measure frequency estimation in a strongly non-Markovian reservoir (e.g., damped Jaynes-Cummings with $\\gamma_0\\gg\\lambda$) under $\\pi/2$ dynamical-decoupling pulses with fixed finite width and spacing $T$, and compare the QFI to $4t^2\\mathrm{Var}[G_{\\rm eff}]$; if the curve bends away from quadratic growth as $T$ approaches $1/\\lambda$ and the deviation fails to vanish with decreasing $T$, the average-Hamiltonian idealization is the limiting failure point.","tokens_in":22386,"feed_emoji":"🎯","tokens_out":8141,"duration_ms":73710,"temperature":0.7,"pith_summary":"This paper claims that the ultimate quantum precision limit for parameter estimation, Heisenberg scaling ($\\mathrm{QFI}\\propto t^2$), can be recovered in open quantum systems even when the environment has memory, by applying a carefully chosen sequence of dynamical-decoupling pulses. Recovering this limit normally requires coherent, noise-free evolution; decoherence drags precision down to the shot-noise limit. The paper proves necessary and sufficient conditions for a control sequence to erase the system-environment coupling while leaving the parameter-encoding Hamiltonian nontrivial, and shows that under those conditions the quantum Fisher information grows quadratically in time without any Markovian assumption. The mechanism is demonstrated on the damped Jaynes-Cummings model, where the non-Markovian oscillations of the Fisher information are largely suppressed and the controlled curve approaches the noiseless quadratic growth.","feed_headline":"Pulses restore Heisenberg scaling under non-Markovian noise","feed_subtitle":"Dynamical decoupling recovers the t² quantum precision limit even when the environment keeps memory of past interactions.","key_machinery":"The central mechanism is a time-dependent control Hamiltonian $H_C(t)$ acting only on the probe, whose unitaries $U_C(kT)$ average the system-environment coupling $H_{SE}$ to $c\\,1_S \\otimes J_E$ while preserving a nontrivial averaged signal Hamiltonian $H_S^{\\rm eff}(\\omega)$. The argument runs through the average-Hamiltonian approximation: over $n$ short intervals of length $T$, the time-ordered evolution is replaced by $e^{-i nT(H_S^{\\rm eff}\\otimes 1_E+1_S\\otimes H_E^{\\rm eff})}$, and when the dynamical decoupling condition (Eq. 15) holds the total evolution factorizes as $U_S^{\\rm eff}(\\omega,t)\\otimes U_E^{\\rm eff}(t)$. Theorem 1 characterizes when such unitaries exist: a mixed-unitary channel $\\Phi$ that keeps $\\Phi(H_S)$ nontrivial, makes $\\Phi(\\langle \\psi|H_{SE}|\\psi\\rangle_E)$ proportional to the identity, and makes its diagonal entries constant in the eigenbasis of $\\Phi(H_S)$.","core_discovery":"On the paper's own terms, the central result is that Heisenberg scaling is not tied to closed-system evolution: a system-specific control Hamiltonian $H_C(t)$ can dynamically decouple the probe from its environment while preserving the signal, regardless of whether the noise is Markovian or non-Markovian. Theorem 1 gives necessary and sufficient conditions in terms of a mixed-unitary channel $\\Phi$: $\\Phi(H_S)$ must be non-trivial, the averaged environment-induced term $\\Phi(\\langle \\psi|H_{SE}|\\psi\\rangle_E)$ must be orthogonal to the traceless part of $\\Phi(H_S)$, and its diagonal entries in the eigenbasis of $\\Phi(H_S)$ must be constant. When these hold, the effective evolution factorizes as $\\tilde{U}_{\\rm tot}\\approx U_S^{\\rm eff}(\\omega,t)\\otimes U_E^{\\rm eff}(t)$, so for $H_S(\\omega)=\\omega G_S$ the quantum Fisher information becomes $F_\\omega^{(Q)}(\\rho_S(\\omega,t))\\approx 4t^2\\,\\mathrm{Var}[G_{\\rm eff}]$, i.e. the Heisenberg $t^2$ growth. The same framework covers discrete, non-periodic control sequences, generalizing earlier integral decoupling conditions, and the damped Jaynes-Cummings example confirms the predicted recovery of quadratic scaling under a Lorentzian non-Markovian reservoir.","pith_inferences":["A practical design rule suggested by the proof is to implement a control sequence whose average is a pinching channel onto the eigenbasis of the effective signal, which cancels noise whose environment-averaged matrix is diagonal-constant in that basis; the paper gives a two-qubit illustration of this principle but not a general pulse-construction algorithm.","A natural next test, already implied by the example's finite-pulse gap, is to simulate finite-width pulses and compare with the average-Hamiltonian prediction; such simulations would define the practical upper bound on pulse interval for each noise spectrum.","If the framework extends to $N$ correlated probes—a question the paper explicitly leaves open—then dynamical decoupling could restore both $t^2$ and $N^2$ scaling simultaneously, which would matter for entangled sensors in memory-bearing environments.","Dynamical decoupling here acts like error correction without syndrome measurement, so it may complement quantum error correction in settings where the Markovian assumptions behind standard QEC break down."],"forward_implications":["For any $H_S(\\omega)=\\omega G_S$ satisfying the theorem's conditions, the quantum Fisher information grows as $4t^2 \\mathrm{Var}[G_{\\rm eff}]$, matching the noiseless Heisenberg limit rather than the shot-noise limit.","The necessary and sufficient conditions give a concrete test for whether a given system-environment Hamiltonian can be decoupled while keeping the signal: the environment-averaged coupling must be trace-orthogonal to the signal part and constant on diagonals in the signal eigenbasis.","The discrete summation condition extends prior integral decoupling conditions to non-periodic control sequences, so the control needs only to average over intervals, not to repeat periodically.","In the damped Jaynes-Cummings example with a Lorentzian, strongly non-Markovian reservoir ($\\gamma_0 \\gg \\lambda$), the controlled quantum Fisher information approaches the noiseless quadratic curve.","Because no complete-positivity or Markovian assumption enters, the result applies to general open-system dynamics, including correlated initial system-environment states."],"supporting_citations":[{"why":"Provides the earlier integral decoupling condition that the paper generalizes to discrete, non-periodic pulse sequences.","marker":"[82]"},{"why":"Shows that dynamical-decoupling pulses enhance parameter-estimation precision in noisy systems, the technique the paper extends to non-Markovian metrology.","marker":"[66]"},{"why":"Relates quantum Fisher information flow to non-Markovianity, underpinning the QFI-oscillation analysis in the example.","marker":"[85]"},{"why":"Supplies the damped Jaynes-Cummings/Lorentzian reservoir theory and decay-rate formalism used in the worked example.","marker":"[36]"},{"why":"Gives the QFI variance formula and optimal-probe construction on which the $4t^2\\mathrm{Var}[G]$ derivation rests.","marker":"[79]"},{"why":"Establishes the decoherence-induced shot-noise limit that the paper aims to beat, defining the baseline.","marker":"[21]"},{"why":"Introduces dynamical decoupling of open quantum systems, the control toolbox the paper uses.","marker":"[57]"},{"why":"Shows quantum error correction can achieve the Heisenberg limit under Markovian noise, the contrast case motivating DDT's broader applicability.","marker":"[25]"}],"fun_headline_variants":["Heisenberg limit returns via decoupling in non-Markovian noise","Dynamical decoupling restores t² precision under memory noise","Quantum control beats non-Markovian noise for Heisenberg scaling","Heisenberg scaling survives non-Markovian baths via decoupling","Decoupling recovers Heisenberg limit for non-Markovian metrology"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument's load-bearing idealization is the average-Hamiltonian approximation: pulses are taken to be instantaneous delta functions and intervals so short that the Hamiltonian is effectively time-independent, so the replacement $e^{-iT\\sum \\tilde H(kT)}$ is exact only in the limit of infinitely fast control; finite pulses or intervals comparable to the environment correlation time degrade the quadratic scaling.","fun_headline_variants_meta":{"raw":{"variants":["Heisenberg limit returns via decoupling in non-Markovian noise","Dynamical decoupling restores t² precision under memory noise","Quantum control beats non-Markovian noise for Heisenberg scaling","Heisenberg scaling survives non-Markovian baths via decoupling","Decoupling recovers Heisenberg limit for non-Markovian metrology"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000602,"raw_usage":{"total_tokens":2861,"prompt_tokens":1048,"completion_tokens":1813,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":664,"completion_tokens_details":{"reasoning_tokens":1719}},"tokens_in":664,"tokens_out":1813,"duration_ms":12646,"temperature":1.0,"reasoning_tokens":1719,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T22:15:37.446547+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate or measure frequency estimation in a strongly non-Markovian reservoir (e.g., damped Jaynes-Cummings with $\\gamma_0\\gg\\lambda$) under $\\pi/2$ dynamical-decoupling pulses with fixed finite width and spacing $T$, and compare the QFI to $4t^2\\mathrm{Var}[G_{\\rm eff}]$; if the curve bends away from quadratic growth as $T$ approaches $1/\\lambda$ and the deviation fails to vanish with decreasing $T$, the average-Hamiltonian idealization is the limiting failure point.","supporting_citations":[{"cited_title":"Control of decoherence: Analysis and comparison of three different strategies,","cited_arxiv_id":null,"evidence_quote":"Provides the earlier integral decoupling condition that the paper generalizes to discrete, non-periodic pulse sequences."},{"cited_title":"En- hancement of parameter-estimation precision in noisy systems by dynamical decoupling pulses,","cited_arxiv_id":null,"evidence_quote":"Shows that dynamical-decoupling pulses enhance parameter-estimation precision in noisy systems, the technique the paper extends to non-Markovian metrology."},{"cited_title":"Quantum Fisher informa- tion flow and non-Markovian processes of open systems,","cited_arxiv_id":null,"evidence_quote":"Relates quantum Fisher information flow to non-Markovianity, underpinning the QFI-oscillation analysis in the example."},{"cited_title":"The theory of open quantum systems,","cited_arxiv_id":null,"evidence_quote":"Supplies the damped Jaynes-Cummings/Lorentzian reservoir theory and decay-rate formalism used in the worked example."},{"cited_title":"Statistical distance and the geometry of quantum states,","cited_arxiv_id":null,"evidence_quote":"Gives the QFI variance formula and optimal-probe construction on which the $4t^2\\mathrm{Var}[G]$ derivation rests."},{"cited_title":"Improvement of frequency standards with quantum entanglement,","cited_arxiv_id":null,"evidence_quote":"Establishes the decoherence-induced shot-noise limit that the paper aims to beat, defining the baseline."},{"cited_title":"Dynamical decoupling of open quantum systems,","cited_arxiv_id":null,"evidence_quote":"Introduces dynamical decoupling of open quantum systems, the control toolbox the paper uses."},{"cited_title":"Achieving the Heisenberg limit in quantum metrology using quantum error correction,","cited_arxiv_id":null,"evidence_quote":"Shows quantum error correction can achieve the Heisenberg limit under Markovian noise, the contrast case motivating DDT's broader applicability."}],"review_version":1}