{"id":"671f550e-c72d-4343-ad99-e9795fbae8b9","arxiv_id":"1908.03399","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A kicked Bose-Hubbard chain at unit filling shows evidence of many-body dynamical localization: Floquet states violate eigenstate thermalization and the system avoids infinite-temperature heating for small hopping strengths.","lead":"A clean, disorder-free kicked Bose-Hubbard chain appears to stop absorbing energy and avoid thermalization for weak hopping, a phase called many-body dynamical localization. The paper presents numerical evidence that this non-thermal phase survives in the large-system limit, in contrast with the generic expectation that driven systems heat to infinite temperature.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Thermodynamic-limit claim rests on finite-size trends that could still be prethermal; the paper's own caveat about a possible downward bend for L>11 is unaddressed.","rationale":"The reader's weakest assumption—that the L≤11 trends represent the thermodynamic limit—is exactly the load-bearing point. The paper itself flags the possibility of a downward bend for L>11, and the truncation for L>10 presupposes localization. No internal inconsistency or numerical error was identified in the presented data; the no-degeneracy remark in Sec. II C is plausible but unproved. The proposed exact-diagonalization check for L=12 and L=13 would either consolidate or falsify the central claim. Since the evidence is suggestive but not conclusive, the CONDITIONAL verdict remains appropriate.","tokens_in":22037,"tokens_out":5509,"duration_ms":60793,"concrete_test":"Perform a targeted exact-diagonalization computation of mid-spectrum Floquet states for L=12 and L=13 in the k=0, P=+1 sector (e.g., with shift-invert Krylov methods) for J=0.01, 0.025, and 0.05, and recompute δ of Eq. (13). If δ continues to bend upward as in Fig. 3, the MBDL interpretation is supported; if δ bends downward or shows non-monotonic behavior, the finite-size trend is a prethermal artifact and the thermodynamic-limit claim does not hold.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that MBDL persists in the thermodynamic limit is based on the upward bend of δ (Eq. 13) for J≤0.05 in Fig. 3, but exact diagonalization stops at L=11 and the authors explicitly state in Sec. III A that they cannot exclude a downward bend from L=12 onward. For L>10, the data rely on a Hilbert-space truncation with ΔT≤4 that presupposes the very localization it is used to probe. The tDMRG results reach L=26 only for the overlap (Fig. 4), not for ϵ or the Vα distribution. This leaves prethermalization as an alternative explanation: the apparent localization for L≤11 could be a long-lived plateau, with thermalization setting in at larger sizes or longer times. Since the strongest claim is about large-size persistence, this is the load-bearing weak point. A secondary technical issue is the unproved no-degeneracy assumption in Sec. II C that underlies the Floquet-diagonal-ensemble average of Eq. (8); if degeneracies occur in the symmetric sector, the infinite-time average and all derived quantities would need revision.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies a periodically kicked Bose-Hubbard chain with on-site interaction V and kicked hopping K. Using exact diagonalization for L≤11, a ΔT-truncated Hilbert-space method and Krylov propagation for larger sizes, and tDMRG for the overlap and entanglement dynamics, the authors present evidence for a many-body dynamically localized (MBDL) phase at small hopping J≲0.05. In this regime, the infinite-time average of the interaction energy remains below the infinite-temperature value, the distribution of Floquet-state expectation values of V stays broad (δ does not decay with L), the overlap with the initial state remains O(1) with revivals at times linear in L, and the half-chain entanglement entropy grows linearly in time before saturating at a subthermal volume-law value. The authors argue that all Floquet states violate eigenstate thermalization, that the phase persists in the thermodynamic limit, and that it is distinct from MBL because of the linear entanglement growth and the delocalized structure of Floquet states within quasi-degenerate eigenspaces of V. A truncated-Wigner calculation shows that the classical counterpart always diffuses, supporting the quantum nature of the effect.","tokens_in":22225,"tokens_out":7241,"duration_ms":80365,"significance":"If the central claim holds, this would be one of the few disorder-free, translation-invariant driven systems exhibiting ergodicity breaking that survives the thermodynamic limit, and the linear-in-time entanglement growth would sharply distinguish this phase from MBL. The paper is carefully executed: it uses multiple complementary diagnostics (ϵ, δ, overlap, entanglement entropy, IPR), reports honest caveats about finite sizes, checks the Hilbert-space truncation with a border projector, and provides error estimates. The main significance is therefore high, provided the thermodynamic-limit extrapolation can be made more robust.","major_comments":[{"comment":"The upward bend of δ for J≲0.05, interpreted as evidence for MBDL, is observed only up to L≈11 in the full exact-diagonalization data, and the larger-L points in Fig. 1 rely on the Hilbert-space truncation of Sec. II B, whose ΔT≤4 cutoff presupposes localization in Fock space. The authors explicitly state that a downward bend from L=12 onward cannot be excluded. Since the paper's central claim is persistence in the thermodynamic limit, this finite-size/prethermal ambiguity is load-bearing. I ask the authors to provide additional evidence against a prethermal plateau, for example long-time Krylov or tDMRG evolutions of ϵ(t) at J≤0.05 for L=12–16 showing no late-time growth toward ϵ=1, or a quantitative scaling analysis of δ that models a possible crossover to a downward bend.","section":"Sec. III A, Fig. 3"},{"comment":"The Floquet-diagonal-ensemble formula for the infinite-time average assumes that no degeneracies occur in the quasienergy spectrum of the symmetric sector. The authors assert this is the case for nonvanishing U and J, but no proof is given; given the translation and parity symmetries, exact degeneracies are not obviously absent. If degeneracies exist, the infinite-time average of V does not reduce to Eq. (8) inside degenerate subspaces, and the quantities Vα, δ, and ϵ become dependent on the choice of eigenbasis. Please either prove the non-degeneracy for the parameter values used, or demonstrate numerically that the main conclusions are unchanged when the degenerate subspaces are diagonalized with a randomly chosen unitary.","section":"Sec. II C, Eq. (8)"},{"comment":"The Hilbert-space truncation for L>10 fixes the maximum Fock-space distance ΔT≤4. This truncation is natural in a localized regime, but it cannot rule out a prethermal scenario in which the wave packet slowly spreads to larger Δ shells at times beyond the simulated window. The border projector Π_B bounds leakage within the truncated space, but only for the simulated time window. I request a convergence check in ΔT at fixed L (e.g., L=12–14) for the MBDL parameters, and, if feasible, an untruncated Krylov simulation at one intermediate size to verify that truncation does not suppress thermalization artificially.","section":"Sec. II B, Sec. III A"}],"minor_comments":[{"comment":"The definition of δ uses ⟨log V⟩; if any Floquet state has Vα=0, the logarithm is singular. Since the uniform initial state is a single Fock state with V=0, the authors should state explicitly whether all Floquet states in the symmetric sector have strictly positive Vα, or restrict the average to Vα>0.","section":"Eq. (13)"},{"comment":"The statement that the system does not generically heat up to infinite temperature is stronger than what is directly tested; the numerics explicitly consider only the uniform state and one excited product state. The all-spectrum distribution of Vα supports the claim, but it would be clearer to state that the conclusion is an inference from the Floquet-state distributions rather than from a survey of initial states.","section":"Abstract"},{"comment":"The interpretation that delocalization in Fock space implies ballistic propagation in real space is an indirect inference; the manuscript does not directly measure a real-space spreading observable such as density-density correlations. Please either add such a measurement or soften the wording.","section":"Sec. IV"},{"comment":"The error bars for the truncated points are described as the average of the border projectior Π_B, but in the figure they are barely visible; please clarify their magnitude relative to the symbol size, or provide a table of values for the larger-L points.","section":"Fig. 1 caption"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is an honest, well-executed numerical study of a potentially important phenomenon. The main obstacle is that the central thermodynamic-limit claim rests on sizes that are too small to rule out prethermalization, and the degeneracy assumption in Sec. II C needs either a proof or a numerical sensitivity check. I would be willing to reconsider after these points are addressed. I also suggest considering whether the paper might be more accurately framed as evidence for a robust prethermal MBDL regime, given that an extrapolation to the thermodynamic limit is not yet conclusive."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper is a serious numerical study of a clean kicked Bose-Hubbard chain, and the central claim is genuinely interesting: for small hopping J ≤ 0.05, the Floquet spectrum shows non-thermal behavior up to L = 11, with entanglement entropy growing linearly in time rather than logarithmically as in MBL. That distinction from MBL is a genuinely useful diagnostic, and the paper does several things well. It checks multiple independent signatures (energy dynamics, overlap revivals, entanglement distributions, IPR scaling), shows the behavior is not restricted to a low-energy corner by testing an excited initial state, and compares against truncated Wigner dynamics to demonstrate that the localization is a quantum coherence effect with no classical analog. The authors are also honest about the limitations of their sizes, which I appreciate.\n\nThe soft spots are exactly where the reader puts them. The thermodynamic-limit claim rests on L ≤ 11 for full exact diagonalization, and for larger L the Hilbert-space truncation with ΔT ≤ 4 presupposes the localization it is used to probe. The paper itself admits it cannot exclude a downward bend in δ from L = 12 onward, which is the prethermal scenario. That is a real weakness, but I would call it a normal limitation for this kind of numerics rather than a fatal flaw; the trends are coherent and the authors do not overclaim beyond their data in the main text. The unproved no-degeneracy assumption in Sec. II C is a smaller technical issue, but it is worth asking them to justify it or at least numerically check it, since the Floquet diagonal ensemble average depends on it. A practical gripe: no code or data are provided, which makes it hard to verify the truncation convergence and the tDMRG results independently.\n\nMy overall take: the paper deserves a serious referee and is a solid candidate for conditional acceptance. The main claim is plausible but not proven, and the authors should be asked to soften the persistence-in-thermodynamic-limit wording, address the degeneracy assumption, and release code and data. I would not desk-reject it, and I would probably cite it if I worked on driven disordered systems.","headline":"A coherent numerical case for many-body dynamical localization in a clean kicked Bose-Hubbard chain; the thermodynamic-limit claim is plausible but rests on sizes too small to rule out a prethermal plateau.","tokens_in":22782,"tokens_out":1575,"would_cite":true,"duration_ms":17851,"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":"A clean, periodically kicked Bose-Hubbard chain can host a many-body dynamically localized phase that persists as the system grows: Floquet states violate eigenstate thermalization and energy absorption stays below infinite temperature.","keywords":["many-body dynamical localization","kicked Bose-Hubbard model","Floquet states","eigenstate thermalization hypothesis","entanglement entropy","ergodicity breaking","periodic driving","infinite temperature"],"falsifier":"Compute $\\delta$ and $\\epsilon$ for $L \\ge 14$ with the same small-$J$ parameters using an approximation whose truncation error is controlled; if $\\delta$ starts bending downward toward zero or $\\epsilon$ creeps upward toward one as $L$ grows, the apparent MBDL is a finite-size or prethermal effect. Likewise, if the revival time stops growing linearly with $L$ or the entanglement growth crosses over from linear to logarithmic at larger sizes, the claimed phase is falsified.","tokens_in":21810,"feed_emoji":"⚛️","tokens_out":15122,"duration_ms":126052,"temperature":0.7,"pith_summary":"The paper tries to establish that a clean, periodically kicked Bose-Hubbard model — with no disorder — can exhibit many-body dynamical localization: after a brief transient the system stops absorbing energy and its local observables never reach the infinite-temperature value, for almost all initial states. The authors argue this is a genuine phase that survives in the thermodynamic limit, involving every Floquet state, and that it is different from many-body localization because the half-chain entanglement entropy grows linearly in time instead of logarithmically. If the claim is right, it would be a rare example of a non-integrable driven many-body system that evades the 'heating to infinite temperature' fate through quantum interference alone, with the Floquet states delocalized in space yet confined to narrow quasi-degenerate subspaces of the undriven interaction energy.","feed_headline":"Clean kicked gas resists heating to infinity","feed_subtitle":"A periodic kick stops energy absorption in a clean Bose-Hubbard chain, breaking ergodicity without disorder.","key_machinery":"The central object is the Floquet-state structure of the unitary step $\\hat U = e^{-i\\hat V \\tau} e^{-i\\hat K}$, where $\\hat V = \\frac{U}{2}\\sum_j \\hat n_j(\\hat n_j - 1)$ is the on-site interaction (diagonal in the Fock basis) and $\\hat K = J \\sum_j (\\hat a^\\dagger_j \\hat a_{j+1} + \\mathrm{h.c.})$ is the hopping delivered by an instantaneous kick. The argument is carried by three diagnostics built from these Floquet states: the heating parameter $\\epsilon$ comparing the infinite-time stroboscopic average of $\\hat V$ with its $T=\\infty$ value; the broadness measure $\\delta$ of the distribution of $V_\\alpha = \\langle \\varphi_\\alpha | \\hat V | \\varphi_\\alpha \\rangle$, which distinguishes ETH from its violation; and the comparison between the inverse participation ratio and its minimal 'delocalized' value $\\mathrm{IPR}^{\\mathrm{deloc}}$, which shows that states fill the degenerate $V$-subspaces they sit in. The half-chain entanglement entropy $S_{L/2}$ completes the picture by separating this phase from MBL through its linear-in-time growth.","core_discovery":"For hopping strength $J \\lesssim 0.05$ (with $U=1$, period $\\tau=1$, unit filling), the stroboscopic dynamics of the kicked Bose-Hubbard chain shows ergodicity breaking that persists up to the largest sizes considered: the infinite-time average of the interaction energy $\\bar V$ stays well below its $T=\\infty$ value, so the normalized heating parameter $\\epsilon$ converges to a number smaller than one as $L$ grows. The Floquet states violate eigenstate thermalization in a size-robust way: the distribution of their $V_\\alpha$ values stays broad, the diagnostic $\\delta = |\\langle V\\rangle - \\exp\\langle \\log V\\rangle|$ bends upward rather than decaying to zero for $J \\lesssim 0.05$, and the same behaviour is found starting from an excited initial state, indicating the whole spectrum is affected. Half-chain entanglement entropy grows linearly in time and saturates to a volume-law value much smaller than the thermal one — the opposite of the logarithmic growth seen when disorder is added to induce many-body localization. Floquet states in this phase are delocalized within quasi-degenerate eigenspaces of $V$ (their inverse participation ratio scales with the 'delocalized' benchmark $\\mathrm{IPR}^{\\mathrm{deloc}}$ raised to a positive power), but they are localized across those subspaces; the authors interpret this as spatial delocalization coexisting with energy localization, which blocks full exploration of the Hilbert space.","pith_inferences":["If the phase survives at larger sizes, it points to an emergent integrable structure: an extensive set of conserved or nearly conserved quantities in the driven system, whose generators are localized in energy rather than in space.","The same mechanism may transfer to other driven lattice models whose undriven part has highly degenerate eigenspaces and whose kick is a hopping operator; a direct test would be to apply the same $\\epsilon$ and $\\delta$ scalings to a kicked Fermi-Hubbard or spin-1/2 chain.","A sharper test of the phase would be to measure correlation spreading: if the picture of delocalization inside degenerate subspaces is right, single-particle correlations should propagate ballistically while energy correlations remain frozen, giving a two-speed light cone.","The finite-size data leave open the possibility of a prethermal plateau for $J$ just below the crossover; pushing the same observables to $L \\gtrsim 20$ with tensor networks, looking for a downturn of $\\delta$ or an upturn of $\\epsilon$, would settle whether the phase extrapolates."],"forward_implications":["Periodically driven clean lattice systems can be protected from heating to infinite temperature without disorder, so driving alone does not force a trivial infinite-temperature state; the transition at $J \\approx 0.06$–$0.1$ is an eigenstate phase transition in the Floquet spectrum.","Because the entanglement-entropy growth is linear rather than logarithmic, an experiment that measures entanglement can directly distinguish many-body dynamical localization from many-body localization in the same platform.","The revival time of the initial-state overlap grows linearly with $L$, giving an observable finite-size echo signature that is absent in the thermalizing regime.","The truncated-Wigner comparison predicts that any classical or decohered version of the same dynamics heats diffusively, so the localization should be destroyed by loss of quantum coherence — a testable prediction for experiments."],"supporting_citations":[{"why":"Defines the eigenstate thermalization hypothesis that the MBDL phase is claimed to violate.","marker":"[1]"},{"why":"Provides the Floquet diagonal ensemble formalism and the $\\epsilon$ criterion used to quantify deviation from infinite-temperature thermalization.","marker":"[19]"},{"why":"The many-body kicked rotor in which dynamical localization appeared to vanish in the thermodynamic limit, serving as the key contrast for the new phase.","marker":"[22]"},{"why":"Earlier example of many-body dynamical localization in the thermodynamic limit for a system integrable at high energies, which this paper extends to a whole-spectrum phase.","marker":"[23]"},{"why":"Kicked Lieb-Liniger result showing MBDL for small-energy initial states; the paper extends this to all initial states in a non-integrable lattice model.","marker":"[24]"},{"why":"Establishes linear entanglement growth in clean integrable driven systems, the signature used to distinguish MBDL from MBL.","marker":"[26]"},{"why":"Establishes logarithmic entanglement growth in MBL, the benchmark that the disorder comparison in the paper reproduces.","marker":"[52]"},{"why":"Supplies the truncated-Wigner approximation whose diffusive heating shows the localization is absent in the classical limit.","marker":"[65]"}],"fun_headline_variants":["Clean kicked chain stays far from hot","Kicked Bose-Hubbard: ergodicity broken cleanly","No disorder, no heat: kicks localize dynamics","Kicked chain defies infinite temperature","Clean kicks beat thermalization"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the behaviour seen for $J \\lesssim 0.05$ at sizes up to about nineteen sites is the thermodynamic-limit behaviour: that the upward bend of $\\delta$ and the saturation of $\\epsilon$ below one are signs of a true localized phase, not a prethermal plateau that would give way to heating at larger sizes or later times.","fun_headline_variants_meta":{"raw":{"variants":["Clean kicked chain stays far from hot","Kicked Bose-Hubbard: ergodicity broken cleanly","No disorder, no heat: kicks localize dynamics","Kicked chain defies infinite temperature","Clean kicks beat thermalization"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000338,"raw_usage":{"total_tokens":1888,"prompt_tokens":984,"completion_tokens":904,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":600,"completion_tokens_details":{"reasoning_tokens":836}},"tokens_in":600,"tokens_out":904,"duration_ms":9529,"temperature":1.0,"reasoning_tokens":836,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:15:15.849870+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute $\\delta$ and $\\epsilon$ for $L \\ge 14$ with the same small-$J$ parameters using an approximation whose truncation error is controlled; if $\\delta$ starts bending downward toward zero or $\\epsilon$ creeps upward toward one as $L$ grows, the apparent MBDL is a finite-size or prethermal effect. Likewise, if the revival time stops growing linearly with $L$ or the entanglement growth crosses over from linear to logarithmic at larger sizes, the claimed phase is falsified.","supporting_citations":[{"cited_title":"Fock basis","cited_arxiv_id":null,"evidence_quote":"Defines the eigenstate thermalization hypothesis that the MBDL phase is claimed to violate."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The many-body kicked rotor in which dynamical localization appeared to vanish in the thermodynamic limit, serving as the key contrast for the new phase."},{"cited_title":"Notarnicola, F","cited_arxiv_id":null,"evidence_quote":"Earlier example of many-body dynamical localization in the thermodynamic limit for a system integrable at high energies, which this paper extends to a whole-spectrum phase."},{"cited_title":"Parameswaran, A","cited_arxiv_id":null,"evidence_quote":"Establishes linear entanglement growth in clean integrable driven systems, the signature used to distinguish MBDL from MBL."},{"cited_title":"ˇZnidariˇ c, T","cited_arxiv_id":null,"evidence_quote":"Establishes logarithmic entanglement growth in MBL, the benchmark that the disorder comparison in the paper reproduces."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the truncated-Wigner approximation whose diffusive heating shows the localization is absent in the classical limit."}],"review_version":1}