{"id":"6fc9b12b-1091-4cf0-a3b5-aecb33aceadd","arxiv_id":"2507.22982","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Driving a dipolar NV spin ensemble at detunings hzT=4πk freezes the spin magnetization for times far beyond T2, and this freezing is used to build an ac magnetometer with 4.3 dB better sensitivity than periodic dynamical decoupling.","lead":"An interacting ensemble of about 10,000 NV spins in diamond, when driven at specific detuning frequencies, freezes its total magnetization instead of thermalizing, and the magnetization persists more than 30 times longer than the natural coherence time. The same driving scheme is turned into an ac magnetometer that reports a 4.3 dB sensitivity improvement over conventional dynamical decoupling.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"4.3 dB sensing gain is obtained away from the freezing points, so the headline 'dynamical-freezing-enhanced' claim is not supported by the data; the freezing observation itself is credible.","rationale":"The reader already recommended CONDITIONAL, and my concern reinforces that condition rather than moving the verdict. I do not find a fatal flaw in the experimental observation of dynamical freezing: the discrete freezing points, the quantitative agreement of micromotion spectra with the Floquet kick operator, and the Ω³/hz² collapse in Fig. 2d are strong internal evidence for the emergent-conservation mechanism. The alternative of disorder-induced localization is made unlikely by the sharp dependence on hzT and the restoration of thermalization away from freezing points. The load-bearing weakness is in the applied claim: the abstract and title imply that dynamical freezing itself yields the 4.3 dB sensitivity gain, whereas the data show the best sensitivity in the explicitly non-frozen regime. The paper honestly discloses this and defers the mechanism, but the headline remains misleading. A conditional verdict should require either relabeling the protocol as a Floquet-enhanced but not freezing-enhanced method, or providing a mechanistic account of the away-from-freezing sensitivity. The concrete test above would settle whether the freezing-region enhancement alone supports the advertised magnitude; my reading of the reported numbers indicates it does not.","tokens_in":17498,"tokens_out":6868,"duration_ms":87368,"concrete_test":"Recompute the sensitivity comparison using only the blue 'near freezing' regions of Extended Data Fig. 3c,d, taking the minimum over Bac within those shaded intervals for both Ω=Ω0 and Ω=2.4Ω0. If the best freezing-region sensitivity is not at least 4.3 dB better than the PDD optimum, then the headline enhancement cannot be attributed to dynamical freezing, and the sensing claim must be reframed as an unexplained Floquet-coherence effect.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central dual claim couples the observation of dynamical freezing to a sensing enhancement achieved by that mechanism. The freezing observation is well supported: discrete peaks at hzT=4πk, micromotion peaks matching the predicted kick-operator frequencies, and the Ω³/hz² scaling of the early decay. However, the headline 4.3 dB sensitivity improvement is not obtained at the freezing points. In Fig. 3e, the best overall D.F. sensitivity is 6.5(6) nT/√Hz, shown as pink circles explicitly labeled 'away from freezing'; the best sensitivity in the blue 'near freezing' regions is 12.9(6) nT/√Hz, only about 1.5 dB better than the PDD optimum of 18(3) nT/√Hz. The text states that away from freezing 'the system is not frozen' and that the detailed relationship between Floquet parameters, coherence, and slopes in that regime is left to 'future theoretical works.' Consequently, attributing the 4.3 dB gain to dynamical freezing is an over-attribution: the improvement arises from an uncharacterized non-freezing regime of the same Floquet sequence. This does not undermine the freezing observation, but it breaks the causal link in the abstract's claim of a 'dynamical-freezing-enhanced ac magnetometry.'","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports experiments on a dense ensemble of about 10^4 nitrogen-vacancy spins driven by a periodic sequence with alternating detunings. It observes long-lived stroboscopic magnetization at discrete freezing points hzT = 4πk, coherent micromotions whose dominant frequencies match Floquet-Magnus kick-operator predictions, and a collapse of early-time decay curves under the scaling variable Ω^3/hz^2. It then proposes an ac magnetometry protocol based on the same Floquet sequence and reports a 4.3 dB sensitivity improvement over periodic dynamical decoupling (PDD). The freezing observation is presented as the first experimental realization of dynamical freezing, and the sensing result is presented as an application enabled by that mechanism.","tokens_in":17670,"tokens_out":6801,"duration_ms":81169,"significance":"If the claims are taken as stated, the paper would constitute a first experimental observation of dynamical freezing in an interacting many-body system, together with a practical sensing application. The freezing observation itself is well supported: the freezing condition and the micromotion frequencies are parameter-free predictions of an independently published theory, the Ω^3/hz^2 collapse of the decay curves is a nontrivial predictive test, and the DTWA simulations plus the data/code availability commitments strengthen the presentation. The sensing claim, however, is not yet supported in the regime that the manuscript identifies with dynamical freezing. The reported 4.3 dB improvement is obtained away from the freezing points, in a regime the authors explicitly state is not theoretically understood. The paper therefore needs a substantial revision of the sensing claim, but the core freezing observation is credible and significant.","major_comments":[{"comment":"The abstract and conclusion attribute the 4.3 dB sensitivity improvement to dynamical freezing, but the optimal sensitivity of 6.5(6) nT/√Hz is obtained at the pink circles, which are explicitly labeled 'away from freezing' in Fig. 3e. The text states that in this regime 'the system is not frozen' and that the detailed relationship between Floquet parameters, coherence, and slopes is left to 'future theoretical works.' The best sensitivity at the blue near-freezing points is 12.9(6) nT/√Hz, corresponding to a much smaller ~1.5 dB improvement over the PDD optimum of 18(3) nT/√Hz. Thus the causal statement 'dynamical-freezing-enhanced ac magnetometry' is an over-attribution. The claim should either be reworded to describe a Floquet-PDD protocol whose high-sensitivity regime lies away from freezing, or the away-from-freezing regime needs a theoretical analysis before the enhancement can be attributed to dynamical freezing.","section":"Dynamical-freezing enhanced magnetometry; Fig. 3e"},{"comment":"The sensitivity comparison uses different data processing for the two protocols: PDD slopes are obtained from sinusoidal fits, while D.F. slopes are obtained by moving-average smoothing over three consecutive points followed by numerical differentiation. This asymmetry can bias the relative sensitivity estimate, particularly in the oscillatory away-from-freezing regions of Fig. 3c. The authors should apply a common fitting and processing procedure to both protocols, or quantify the systematic uncertainty introduced by the smoothing, before the 4.3 dB figure can be taken at face value.","section":"Methods, Experimental sensitivity and Eq. (7)"},{"comment":"The identification of the observed non-thermalizing dynamics with emergent-conservation dynamical freezing, rather than with Floquet many-body localization, is supported by the discrete freezing-point structure and the Ω^3/hz^2 scaling. However, the manuscript does not explicitly discuss whether the residual disordered on-site fields and random couplings could produce a localization-like contribution. A short supplementary analysis, or a direct statement explaining why the freezing-point condition and the Ω^3/hz^2 collapse are incompatible with a disorder-dominated scenario, would make the central interpretation more robust. As written, the claim that the effect is independent of disorder is slightly stronger than what is directly demonstrated.","section":"Observation of dynamical freezing; Conclusion"}],"minor_comments":[{"comment":"The notation Ω0 and h0 is used in the Fig. 3 caption but not defined in the main text; please define these on first use.","section":"Fig. 1 and Fig. 3 captions"},{"comment":"The kick-operator expressions are given in the moving frame, but the transformation between the moving frame and the laboratory frame is not shown; without it, a reader cannot reproduce the micromotion expressions in Eq. (4).","section":"Methods, Eqs. (3)-(4)"},{"comment":"The 4.3 dB improvement is consistent with 10 log10(η_PDD/η_DF) rather than 20 log10; please state explicitly which convention is used.","section":"Sensitivity comparison, Fig. 3e"},{"comment":"The legend labels 'Ω^3/(2π hz^2) 0.9 kHz 3.6 kHz 6.4 kHz' are ambiguous; please specify the exact units and whether these are values of Ω^3/(2π hz^2) or of Ω^3/hz^2.","section":"Fig. 2d"}],"recommendation":"major_revision","confidential_remarks":"The manuscript's own data and text separate the freezing regime from the high-sensitivity regime: the 6.5(6) nT/√Hz optimum is explicitly 'away from freezing.' The headline claim in the abstract and conclusion therefore over-attributes the sensing gain to dynamical freezing. This is correctable, but it is not a purely cosmetic issue: the quantitative 4.3 dB claim would need either a revised interpretation or new theoretical support for the away-from-freezing regime. The freezing observation itself appears solid and could anchor the paper after a careful revision of the sensing narrative."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The freezing observation is real and well-supported. Discrete peaks at hzT = 4πk, micromotion spectra matching the predicted kick-operator frequencies, and the Ω³/hz² scaling of the early decay all line up with parameter-free Floquet-Magnus theory, and DTWA simulations agree. That is a genuinely new experimental result and the first clear demonstration of this thermalization-breaking mechanism in a many-body system.\n\nThe soft spot is the sensing claim. The abstract says 'dynamical-freezing-enhanced ac magnetometry' and reports a 4.3 dB improvement over PDD. But the 4.3 dB point is explicitly labeled 'away from freezing,' and the text admits the system is not frozen there and that the mechanism is left to future theory. The method does extend sensing time beyond T2, but attributing the gain to dynamical freezing is an overreach. Near freezing, the best sensitivity is only about 1.5 dB better than PDD. The protocol is a real engineering achievement, but the headline causal claim is not supported by the data.\n\nOther issues are minor: data and code are only promised for future release, so others can't verify yet; the dB convention (field vs power) is not stated; and the disorder in the NV ensemble could in principle confound the interpretation, though the sharp discrete freezing peaks and the quantitative agreement with the emergent-conservation picture argue strongly against a localization explanation.\n\nThe paper is for people working on Floquet thermalization, emergent conservation laws, and NV-based magnetometry. It deserves serious peer review, not desk rejection. A good referee should ask the authors to reword the abstract and the sensing claims so that the non-freezing regime is not presented as freezing-enhanced, and to release the data and code. As it stands, the core freezing result is solid and citeable, but the sensing part needs revision before publication.","headline":"Solid first observation of dynamical freezing, but the headline sensing claim is over-attributed: the best sensitivity is achieved away from the freezing points.","tokens_in":18276,"tokens_out":3131,"would_cite":true,"duration_ms":36328,"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":"This paper reports the first experimental observation of dynamical freezing in a dense ensemble of about $10^4$ interacting nitrogen-vacancy spins, where the total magnetization is stroboscopically conserved at the freezing points $h_z T…","keywords":["dynamical freezing","emergent conservation law","Floquet driving","nitrogen-vacancy centers","quantum magnetometry","thermalization breakdown","many-body spin dynamics"],"falsifier":"Run the same Floquet sequence on a small, exactly solvable system with deterministic couplings, or on a larger system with translation-invariant couplings, and check whether the freezing peaks at $h_z T = 4\\pi k$ appear and persist with micromotion spectra matching $K_F^{(1)}$; if the freezing disappears or shifts, the effect in the NV ensemble is driven by static randomness rather than by the emergent $U(1)$ conservation sector.","tokens_in":17246,"feed_emoji":"🧲","tokens_out":7693,"duration_ms":82771,"temperature":0.7,"pith_summary":"This paper reports the first experimental observation of dynamical freezing, a predicted breakdown of thermalization in periodically driven many-body systems, in a dense ensemble of about $10^4$ interacting nitrogen-vacancy spins in diamond. At specific detunings where $h_z T = 4\\pi k$, the total $z$-magnetization is stroboscopically conserved by an emergent $U(1)$ conservation law, and the paper shows this magnetization persists for more than an order of magnitude longer than the interaction-limited coherence time $T_2$. The same emergent conservation suppresses decoherence well past $T_2$, and the paper uses this to build an ac magnetometry sequence that outperforms conventional dynamical-decoupling magnetometry by 4.3 dB in sensitivity. A sympathetic reader should care because this is a concrete demonstration, in a mesoscopic solid-state system, that interaction-induced thermalization can be tamed not by disorder or high frequency but by a drive-induced symmetry, and that this taming can be used directly for a practical sensing task.","feed_headline":"10,000 interacting spins freeze far beyond their coherence limit","feed_subtitle":"At drive detunings matching the freezing points, magnetization survives past T2 and ac magnetometry gains 4.3 dB.","key_machinery":"The load-bearing object is the strong-drive Floquet-Magnus expansion in a moving frame, applied to the alternating Hamiltonian $H(t) = H_0 + \\Omega \\hat{S}_x \\pm h_z \\hat{S}_z$. The zeroth-order effective Floquet Hamiltonian $H_F^{(0)}$ is $H_0$ plus a term proportional to $(2\\Omega/h_z T)(\\sin(h_z T/2)\\hat{S}_x - (1 - \\cos(h_z T/2))\\hat{S}_y)$, the first-order correction vanishes by the reflection symmetry of the drive, and at the freezing condition $h_z T = 4\\pi k$ the remaining leading symmetry-breaking term is proportional to $\\Omega^3/(4h_z^2)\\hat{S}_x$. This structure produces an emergent approximate conservation of $\\langle \\hat{S}_z \\rangle$ that fractures the Hilbert space into sectors, while the intra-period micromotions are governed by the first-order Floquet kick operator $K_F^{(1)}(t)$, whose predicted frequencies match the observed spectra. The same expansion gives the field-response formula $H_F^{(0)} \\approx H_0 + (\\Omega/h_z)(2\\gamma_{\\mathrm{NV}} B_{\\mathrm{ac}}/\\pi)\\hat{S}_x$ used to predict and optimize the sensing slopes.","core_discovery":"The paper's central claim is that a strongly driven, interacting ensemble of about $10^4$ NV spins exhibits dynamical freezing: at detunings satisfying $h_z T = 4\\pi k$ ($k = \\pm 1, \\pm 2, \\ldots$), the stroboscopic dynamics conserve $\\langle \\hat{S}_z \\rangle$ even though the undriven system is interacting and would thermalize. The evidence is threefold: the time-averaged $\\langle \\hat{S}_z \\rangle$ shows sharp peaks at exactly these freezing points and remains near unity for hundreds of driving cycles; the intra-period micromotions of $\\langle \\hat{S}_x \\rangle$, $\\langle \\hat{S}_y \\rangle$, and $\\langle \\hat{S}_z \\rangle$ have spectral peaks at the frequencies predicted by the leading Floquet kick operator; and the early-time relaxation rates for different drive parameters collapse when plotted against $\\Omega^3/h_z^2$, the coefficient of the leading symmetry-breaking term in the effective Floquet Hamiltonian. The paper further claims that this frozen dynamics can be converted into an ac magnetometry protocol: at sensing times far beyond $T_2$, the dynamical-freezing sequence retains a steep field response, yielding a best sensitivity of $6.5(6)\\ \\mathrm{nT}/\\sqrt{\\mathrm{Hz}}$, a 4.3 dB improvement over conventional PDD magnetometry under identical conditions.","pith_inferences":["A natural test beyond the paper would be to repeat the freezing measurement in an array with deterministic, translation-invariant couplings, for example a one-dimensional chain, where disorder-induced localization cannot occur; if the freezing peaks at $h_z T = 4\\pi k$ persist, the emergent-conservation interpretation would be directly confirmed.","The paper's sensitivity analysis is based on slopes extracted from $\\langle \\hat{S}_z \\rangle$ versus $B_{\\mathrm{ac}}$; a full quantum-Fisher-information optimization over Floquet parameters and pulse sequences could reveal a larger performance gap than the reported 4.3 dB.","Because freezing points appear at every even multiple of the driving frequency, the same protocol could be adapted for multi-frequency ac magnetometry or noise spectroscopy, selecting different harmonics by changing $h_z$ rather than by changing the $\\pi$-pulse train.","The paper's outlook connects dynamical freezing to Floquet time crystals and symmetry-protected topological phases; an explicit bridge would be to test whether the frozen sector survives quasiperiodic or amplitude-modulated drives, probing the robustness of the emergent conservation law beyond strict periodicity."],"forward_implications":["Freezing is not tied to a specially prepared state: the paper observes conserved magnetization and state-dependent micromotion amplitudes for initial states with different polar angles, so the effect is a property of the drive, not of a scar subspace.","The emergent conservation law protects $\\langle \\hat{S}_z \\rangle$ for over 200 driving cycles while $\\langle \\hat{S}_x \\rangle$ and $\\langle \\hat{S}_y \\rangle$ decay to their thermal values, making the system a long-lived single-axis quantum memory or sensor axis.","The two-stage decay at freezing points, with an early-time rate governed by $\\Omega^3/h_z^2$ and a late-time decay driven by residual disorder, gives experimental control knobs for how long and how rigidly the frozen sector survives.","The 4.3 dB sensitivity gain over PDD is achieved with only global microwave control, so the protocol transfers to other spin ensembles or platforms without single-site addressing.","Operating away from the exact freezing points yields even lower sensitivity than at the freezing points while still extending coherence, suggesting the useful sensing regime is broader than the strict freezing condition."],"supporting_citations":[{"why":"Supplies the theoretical prediction of dynamical freezing via emergent conservation laws in strongly driven Floquet matter that this experiment tests.","marker":"[15]"},{"why":"Extends the dynamical-freezing theory toward the thermodynamic limit, supporting applicability to a mesoscopic ensemble of roughly $10^4$ spins.","marker":"[16]"},{"why":"Provides the Floquet-Magnus expansion used to derive the effective Floquet Hamiltonian, the freezing condition, and the first-order kick operator for micromotions.","marker":"[44]"},{"why":"Gives the discrete truncated Wigner approximation used for numerical simulations of the interacting NV ensemble dynamics.","marker":"[45]"},{"why":"Supplies the quantum sensing framework and sensitivity definition used to benchmark the dynamical-freezing protocol against PDD.","marker":"[46]"},{"why":"Provides the PDD sensitivity formula and experimental parameter characterization against which the 4.3 dB improvement is measured.","marker":"[51]"}],"fun_headline_variants":["Interacting spins freeze beyond T2, boosting magnetometry by 4.3 dB","Dynamical freezing lets NV spins beat T2 and sense 4.3 dB better","Frozen spin dynamics extend magnetometry beyond T2 by 4.3 dB","10,000 interacting spins freeze, enabling 4.3 dB better magnetometry","Driven NV spins freeze, sensing far beyond T2 with 4.3 dB gain"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The interpretation stands on the assumption that the observed long-lived magnetization is produced by the predicted emergent conservation law of dynamical freezing, and not by disorder-induced localization or another prethermalization mechanism that the measurements cannot fully rule out.","fun_headline_variants_meta":{"raw":{"variants":["Interacting spins freeze beyond T2, boosting magnetometry by 4.3 dB","Dynamical freezing lets NV spins beat T2 and sense 4.3 dB better","Frozen spin dynamics extend magnetometry beyond T2 by 4.3 dB","10,000 interacting spins freeze, enabling 4.3 dB better magnetometry","Driven NV spins freeze, sensing far beyond T2 with 4.3 dB gain"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000751,"raw_usage":{"total_tokens":3416,"prompt_tokens":1093,"completion_tokens":2323,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":709,"completion_tokens_details":{"reasoning_tokens":2212}},"tokens_in":709,"tokens_out":2323,"duration_ms":18924,"temperature":1.0,"reasoning_tokens":2212,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T11:11:18.565810+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same Floquet sequence on a small, exactly solvable system with deterministic couplings, or on a larger system with translation-invariant couplings, and check whether the freezing peaks at $h_z T = 4\\pi k$ appear and persist with micromotion spectra matching $K_F^{(1)}$; if the freezing disappears or shifts, the effect in the NV ensemble is driven by static randomness rather than by the emergent $U(1)$ conservation sector.","supporting_citations":[{"cited_title":"Adler, D","cited_arxiv_id":null,"evidence_quote":"Supplies the theoretical prediction of dynamical freezing via emergent conservation laws in strongly driven Floquet matter that this experiment tests."},{"cited_title":"Haldar, D","cited_arxiv_id":null,"evidence_quote":"Extends the dynamical-freezing theory toward the thermodynamic limit, supporting applicability to a mesoscopic ensemble of roughly $10^4$ spins."},{"cited_title":"De Lange, Z.-H","cited_arxiv_id":null,"evidence_quote":"Provides the Floquet-Magnus expansion used to derive the effective Floquet Hamiltonian, the freezing condition, and the first-order kick operator for micromotions."},{"cited_title":"Bukov, L","cited_arxiv_id":null,"evidence_quote":"Gives the discrete truncated Wigner approximation used for numerical simulations of the interacting NV ensemble dynamics."},{"cited_title":"Schachenmayer, A","cited_arxiv_id":null,"evidence_quote":"Supplies the quantum sensing framework and sensitivity definition used to benchmark the dynamical-freezing protocol against PDD."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the PDD sensitivity formula and experimental parameter characterization against which the 4.3 dB improvement is measured."}],"review_version":1}