{"id":"51d3d898-d7d3-4366-822e-b715e7eec132","arxiv_id":"2506.14961","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"In a dissipative Bose-Hubbard trimer, classical chaos versus regular dynamics is imprinted in the level statistics of the steady-state density matrix, with positive Lyapunov exponents giving Wigner-Dyson statistics and non-positive ones giving Poissonian statistics.","lead":"This paper studies a three-well Bose-Hubbard model with dissipation and asks whether classical chaotic motion leaves a trace in the quantum steady state. It reports that the sign of the classical Lyapunov exponent matches the level-statistics type of the steady-state density matrix, and introduces a phase-space measure of localization.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The regular-dynamics branch of the central claim rests on Case III, whose level statistics do not convincingly converge to Poisson; without a quantitative convergence test, the Lyapunov-to-level-statistics correspondence is unsupported.","rationale":"The reader's stated weakest assumption is the semiclassical Langevin construction, but that is an explanatory layer rather than the central empirical claim. The strongest claim is that classical Lyapunov exponents govern the steady-state level statistics, and this is supported only by a handful of parameter sets. The exact-diagonalization results for the regular branch are the load-bearing evidence, and they are explicitly inconclusive for Case III. The paper itself marks the finite-size scalings for Cases III and IV with '(?)' in Table I, and admits in Sec. III.B that convergence to Poissonian statistics could not be established. Because the claim is universal, one non-converging regular case is a direct threat. The semiclassical approximation failing would weaken the proposed mechanism, but the empirical correspondence could still survive; conversely, if Case III's level statistics do not converge to Poisson, the central claim fails regardless of the semiclassical framework. The reader's rationale does mention the inconclusive scaling and missing error bars, so there is partial agreement, but the reader's explicit weakest_assumption points to a different concern. My proposed test would settle whether the regular branch actually holds and whether the Lyapunov-sign transition coincides with the level-statistics transition. Until then, conditional acceptance is appropriate, exactly as the reader recommended.","tokens_in":41413,"tokens_out":5322,"duration_ms":55595,"concrete_test":"Compute the mean level-spacing ratio <r> for the exact steady-state H_eff of Case III as a function of N=15,20,25,30,40, using the sparse Liouvillian methods already employed in the paper. Add bootstrap error bars by resampling blocks of the ordered eigenvalue sequence, and compare <r> to the Poisson value 0.386 and the GUE value 0.603 (Atas et al., PRL 110, 084101). If <r> fails to approach 0.386 within error bars as N grows, the 'non-positive lambda to Poisson' branch is falsified. As a second check, perform a parameter sweep in U for fixed dissipative rates (e.g., U from 0.01 to 6) and verify that the parameter value at which the Lyapunov exponent changes sign matches the parameter value at which <r> transitions between the Poisson and GUE plateaus; a mismatch would disprove the claimed governing role of lambda.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim is that the sign of the classical Lyapunov exponent determines the level statistics of the effective Hamiltonian H_eff: non-positive lambda yields Poissonian statistics, while positive lambda yields GUE statistics. This dichotomy is supported by exactly four parameter sets (Cases I-IV). The positive-lambda branch has one clean example (Case II), but the non-positive branch is shaky: Case I is an attractor, and Case III is a limit cycle for which the authors state, in Sec. III.B, that 'we could not conclude that P(r) converges to a Poissonian distribution' despite N up to 30. Since the claim is universal ('non-positive exponents yield Poissonian statistics'), a single regular case that fails to converge invalidates it. Additionally, the level-spacing-ratio histograms in Fig. 6 are single-realization histograms from one steady state: for N=15, H_eff has only d=(N+1)(N+2)/2=136 eigenvalues, so P(r) is built from roughly 135 ratios. No error bars, bootstrap resampling, or goodness-of-fit tests are provided; the labels 'Poissonian' and 'GUE' are assigned by visual inspection. Case IV's mixed distribution is fitted with alpha=0.57 but no uncertainty is given. Thus the central dichotomy is an overgeneralization from a small number of visually classified histograms. The semiclassical rho_C construction is a separate explanatory layer; even if it were correct, the exact-diagonalization evidence for the regular branch is insufficient.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript studies an open SU(3) Bose-Hubbard trimer with number-conserving incoherent hopping, using four representative parameter sets to compare noiseless mean-field classical dynamics (Lyapunov exponents, Poincaré sections) with the exact Liouvillian spectrum and the steady-state density matrix, and with a semiclassical density matrix built from Langevin trajectories. The central claim is that the sign of the largest classical Lyapunov exponent determines the level statistics of the effective Hamiltonian H_eff defined by rho_0 = exp(-H_eff): non-positive exponents give Poissonian statistics and positive exponents give GUE statistics. The paper also introduces a phase-space inverse participation ratio IPR_phi whose scaling N^{-D/2} defines an effective dimension D, reports S proportional to ln N^D, and shows that the stochastic coherent-state mixture rho_C reproduces these spectral and structural features.","tokens_in":41682,"tokens_out":6053,"duration_ms":59142,"significance":"If established, the claimed correspondence would extend the Prosen–Žnidarič steady-state integrability dichotomy to a collective bosonic model with a well-defined classical limit, connecting the sign of the classical Lyapunov exponent to random-matrix statistics of a steady-state density matrix. The paper has real strengths: the Lyapunov exponents are computed from the noiseless mean-field equations independently of the quantum steady state; the Liouvillian is diagonalized exactly up to N=15 and treated with sparse methods at N=20,30; Appendix E gives an analytic attractor spectrum showing a Poisson-like two-grid structure; and the Langevin dynamics is derived from the Keldysh action rather than fitted. The authors are also transparent about inconclusive scalings and about the non-convergence of Case III. These strengths make the paper a plausible and potentially useful contribution, but the advertised universal dichotomy is currently supported by a small number of visually classified histograms, and the regular-dynamics branch lacks quantitative convergence evidence.","major_comments":[{"comment":"For Case III, the authors state explicitly that 'we could not conclude that P(r) converges to a Poissonian distribution' even at N=30. Since Case III is the only non-attractor regular-dynamics example supporting the non-positive-Lyapunov branch of the central dichotomy, the claim that non-positive exponents yield Poissonian level statistics is not presently established for limit-cycle dynamics. The manuscript should provide a quantitative convergence test for Case III, such as the scaling of the mean and variance of r with N, a Kolmogorov–Smirnov or chi-square statistic against the Poisson and GUE ratio distributions, and bootstrap error bars; visual inspection and the absence of level repulsion are not sufficient to identify the distribution as Poissonian.","section":"§III.B, Fig. 6"},{"comment":"The P(r) histograms in Fig. 6 are single-realization histograms: for N=15 the Hilbert-space dimension is d=(N+1)(N+2)/2=136, so each histogram is built from roughly 135 level-spacing ratios. No error bars, bootstrap resampling, or goodness-of-fit measures are provided, and the labels 'Poissonian' and 'GUE' are assigned by visual comparison with the limiting curves. With this sample size, distinguishing Poissonian statistics from weakly correlated distributions is not reliable, especially because the binning and the treatment of degenerate or nearly degenerate eigenvalues are not specified. Please add quantitative statistical measures and, where possible, averages over realizations or symmetry sectors.","section":"§II.C, §III.B, Fig. 6"},{"comment":"For Case IV, the mixed distribution is fitted with alpha=0.57 without any reported uncertainty, and the finite-size scalings of IPR_phi and S are marked inconclusive in Table I. The statement in §IV.C that 'Across all regimes explored in Table I, we find that the entropy consistently scales as S∝ln N^D' is therefore stronger than the presented evidence: for Case IV the scaling is expressly not determined. The authors should either supply a scaling analysis with uncertainty for Case IV or restrict the entropy-scaling claim to the cases for which the scaling is actually established.","section":"§III.B, Table I, Fig. 7"},{"comment":"The semiclassical density matrix rho_C is constructed from Langevin trajectories obtained by truncating the Keldysh action at quadratic order and using the smooth-path hypothesis of Eq. (C6). The paper validates rho_C against exact results mainly through observable averages and through the N=15 level-statistics insets of Fig. 6; it does not test whether rho_C approximates rho_0 in a norm that controls eigenvalue statistics at long times. Since the rho_C insets are presented as evidence that the semiclassical framework reproduces spectral correlations, please add a systematic check, for example the scaling of a trace distance or spectral distance between rho_C and rho_0 with N, and convergence in sampling time and in the number of noise realizations M_S, or explicitly label this as a heuristic assumption whose failure would not affect the exact-diagonalization results.","section":"§II.B, Appendix C, Eq. (10), Eq. (C6)"}],"minor_comments":[{"comment":"In the first paragraph of the introduction, 'The reminder of Section I' should read 'The remainder of Section I'.","section":"§I"},{"comment":"In Section II.E, 'stochasitc evolution' is a typo for 'stochastic evolution'; the Fig. 4 caption also contains 'parmaeters' instead of 'parameters'.","section":"§II.E"},{"comment":"In the caption of Fig. 6, 'GUI' should be 'GUE', and in the main text 'si well described' should be 'is well described'.","section":"§III.B, Fig. 6"},{"comment":"The statement that a strong symmetry 'necessarily leads to regular Liouvillian dynamics' in the trimer would benefit from a one-sentence justification that autonomous flows on a two-dimensional manifold cannot produce chaos, since the paper currently leaves this as an implicit dimensional argument.","section":"§IV.B"},{"comment":"The table uses empty entries to denote zeroes, but the note 'Empty entries correspond to zeroes' is easy to miss; please consider printing explicit 0 values or a clearer legend.","section":"Table I"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a serious contribution and the central idea is worth publishing, but the universal Lyapunov-to-level-statistics claim needs stronger quantitative support. The most important issue is Case III: the authors themselves admit non-convergence to Poisson statistics, yet that case is the only clean regular non-attractor example. If the authors can provide a convincing statistical analysis and either establish or soften the regular-dynamics claim, this could become acceptable. There is no concern about novelty or attribution; the related-work discussion is appropriate."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth a serious look. The paper's real contribution is the semiclassical machinery: a Keldysh derivation of Langevin dynamics with multiplicative 1/sqrt(N) noise for the SU(3) trimer, a classical mixture of coherent states rho_C that reproduces observables and spectral statistics, and a new phase-space IPR_phi with an effective-dimension interpretation. These are concrete, reproducible tools. The analytic attractor spectrum in Appendix E is also nice and explains why the attractor case shows Poisson statistics.\n\nThe central claim—non-positive Lyapunov exponent gives Poisson, positive gives GUE—is plausible but not as solid as the abstract suggests. The chaotic branch has one clean case (II) and the attractor branch has one clean case (I) plus the analytic argument. The regular limit-cycle branch (III) is explicitly not converged: the authors say they cannot conclude P(r) approaches Poisson even at N=30. They do show no level repulsion, which is suggestive, but the dichotomy as stated is not established for regular non-stationary dynamics. Case IV is marked inconclusive in the table. The histograms are single realizations, with no bootstrap or goodness-of-fit. That said, the paper is honest about these limits; the abstract is a bit ahead of the evidence.\n\nI don't think the stress-test's stronger claim—that the correspondence is unsupported—holds. The regular branch is weak but not absent: no repulsion is visible, and the semiclassical rho_C shows the same distributions, which would be a strange coincidence if the link were wrong. The paper deserves review and publication after revision, not rejection.\n\nWho is it for? People working on dissipative quantum chaos, open many-body systems, and semiclassical methods. The IPR_phi is likely to be reused. I'd send it to a competent referee; ask for quantitative convergence tests (e.g., K-S or chi-square, bootstrap over initial conditions), more regular cases, and a softer statement of the conjecture. The formal parts are the strongest; the numerical evidence just needs to match the boldness of the claim.","headline":"A valuable semiclassical framework and a new phase-space diagnostic, but the central Lyapunov-to-level-statistics claim rests on thin numerical evidence in the regular branch.","tokens_in":42228,"tokens_out":1866,"would_cite":true,"duration_ms":20113,"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 sign of the classical Lyapunov exponent controls the level statistics of the steady-state density matrix in an open many-boson system, with regular motion giving Poissonian statistics and chaotic motion giving Wigner-Dyson statistics.","keywords":["dissipative quantum chaos","Bose-Hubbard trimer","steady state","level statistics","Lyapunov exponent","semiclassical Langevin dynamics","phase-space inverse participation ratio","open quantum systems"],"falsifier":"Take a parameter set of the trimer with no strong symmetry and an unambiguously positive leading Lyapunov exponent, exactly diagonalize the Lindblad equation at $N=30$, and compute the level-spacing ratio distribution of $H_{\\mathrm{eff}}$; if that distribution fails to converge toward the GUE prediction, the claimed Lyapunov-to-statistics correspondence is falsified.","tokens_in":41191,"feed_emoji":"🌀","tokens_out":10470,"duration_ms":88306,"temperature":0.7,"pith_summary":"The paper targets a long-standing question: when a dissipative quantum system has a well-defined classical limit, do the classical dynamics—attractor, limit cycle, or chaos—show up in the quantum steady state? Using a three-well Bose-Hubbard model with particle-conserving dissipation, it argues yes. The sign of the leading Lyapunov exponent of the mean-field trajectories determines the level statistics of the effective Hamiltonian $H_{\\mathrm{eff}}$ defined by $\\rho_0 \\propto e^{-H_{\\mathrm{eff}}}$: non-positive exponents give Poissonian statistics, positive exponents give Wigner-Dyson (GUE) statistics. The paper adds a phase-space inverse participation ratio whose scaling defines an effective dimension $D$ of the steady state's support, with the entropy obeying $S \\propto \\ln N^D$. If correct, this gives a semiclassical explanation for the integrability–chaos dichotomy in open-system steady states and a practical diagnostic for quantum ergodicity in dissipative systems.","feed_headline":"Classical chaos dictates steady-state spectral statistics","feed_subtitle":"In a dissipative trimer, regular classical orbits leave Poissonian steady-state spectra; chaotic ones leave Wigner-Dyson.","key_machinery":"The central object is the effective Hamiltonian $H_{\\mathrm{eff}}$, defined by writing the steady-state density matrix as $\\rho_0 \\propto e^{-H_{\\mathrm{eff}}}$; its level-spacing ratio distribution is the paper's quantum-chaos diagnostic. The argument is carried by two further pieces. First, a semiclassical density matrix $\\rho_C$ is built by adding multiplicative $1/\\sqrt{N}$ stochastic noise to the classical mean-field equations (the noise comes from truncating a path-integral expansion of the dissipative dynamics at quadratic order), evolving many samples, and averaging the resulting coherent states; this converts classical trajectory ensembles into a candidate steady state. Second, a phase-space inverse participation ratio $\\mathrm{IPR}_\\phi$ computed from the Husimi distribution defines an effective dimension $D$ through $\\mathrm{IPR}_\\phi\\sim N^{-D/2}$, linking phase-space localization to the entropy scaling $S\\propto \\ln N^D$. The sign of the leading Lyapunov exponent of the deterministic flow is the classifier that separates regular ($\\lambda\\le 0$) from chaotic ($\\lambda>0$) classical behavior.","core_discovery":"The central discovery is a clean correspondence between classical mean-field dynamics and the structure of the quantum steady state in the $\\mathrm{SU}(3)$ Bose-Hubbard trimer. When the classical equations of motion relax to a fixed-point attractor, the Liouvillian is gapped and the steady state is localized, with Poissonian level spacing in $H_{\\mathrm{eff}}$ and entropy of order one. When dissipation instead drives cyclic motion, the Liouvillian gap closes as $N\\to\\infty$, and the steady state's spectral statistics follow the classical Lyapunov exponent: limit cycles and periodic orbits keep $H_{\\mathrm{eff}}$ Poissonian, whereas chaotic dynamics produce GUE statistics, a delocalized Husimi distribution with $\\mathrm{IPR}_\\phi\\propto N^{-2}$, and entropy scaling $S\\propto \\ln N^2$. The paper further establishes that a semiclassical density matrix built by averaging Langevin trajectories over noise samples reproduces all of these features, including the level statistics, to next-to-leading order in $1/N$.","pith_inferences":["An untested extension of the correspondence is that $H_{\\mathrm{eff}}$ level statistics could serve as an experimental probe of the underlying classical attractor in platforms like triple-well Bose-Einstein condensates, where direct measurement of chaotic trajectories is difficult but steady-state correlation functions are accessible.","The $\\mathrm{IPR}_\\phi$-based effective dimension $D$ suggests a general principle for open systems: in mixed phase spaces, the component with the smallest effective dimension dominates the IPR, so the asymptotic scaling of entropy and localization is set by the most localized sector rather than the most chaotic one.","Because $\\rho_C$ captures GUE statistics without any quantum coherence, the paper implies that Wigner-Dyson statistics in dissipative steady states is compatible with a fully incoherent, classical-mixture interpretation; a similar mechanism may explain apparent random-matrix signatures in other driven-dissipative systems whose classical limit is chaotic.","The role of strong symmetries suggests a testable prediction beyond the trimer: in a four-well (or higher-dimensional) dissipative system, strong symmetries should reduce $D$ but still allow chaos when the remaining phase-space dimension exceeds two."],"forward_implications":["In any large-$N$ dissipative system with a classical limit, the steady-state level statistics of $H_{\\mathrm{eff}}$ can be predicted from the classical Lyapunov spectrum alone: Poissonian for non-positive exponents, GUE for positive exponents.","The Liouvillian spectral gap distinguishes stationary from non-stationary classical regimes (gapped for fixed-point attractors, closing as $N^{-1}$ for limit cycles and chaos), but does not by itself identify chaos; the steady-state statistics of $H_{\\mathrm{eff}}$ are needed for that.","The entropy of the steady state obeys the universal scaling $S\\propto \\ln N^D$, with $D$ read off from $\\mathrm{IPR}_\\phi$: $D=0$ at a fixed-point attractor, $D=4$ for full phase-space delocalization, and reduced $D$ under strong symmetries or dynamical confinement.","Strong symmetries constrain the dynamics to lower-dimensional manifolds and suppress chaos (in this $D=4$ system, any strong symmetry forces regular dynamics), while weak symmetries leave the phase-space structure and chaotic behavior intact.","The semiclassical density matrix $\\rho_C$ reproduces not only expectation values but also entropy, localization, and level statistics, so for large $N$ the Lindblad steady state is effectively a classical mixture of coherent states."],"supporting_citations":[{"why":"Provides the original steady-state effective-Hamiltonian level-statistics criterion that this paper gives a semiclassical explanation for.","marker":"[41]"},{"why":"Establishes the dissipative Bose-Hubbard model as a minimal system with classical chaos in the absence of driving, the testbed of this work.","marker":"[7]"},{"why":"Supplies the field-theoretic framework from which the stochastic Langevin correction to the classical equations is derived.","marker":"[45]"},{"why":"Supplies the complex spacing ratio diagnostic used to classify the Liouvillian spectrum and to contrast it with steady-state $H_{\\mathrm{eff}}$ statistics.","marker":"[29]"},{"why":"States the conjecture that regular classical motion leads to Poissonian level statistics, the benchmark used for regular cases.","marker":"[84]"},{"why":"States the conjecture that chaotic classical motion leads to Wigner-Dyson statistics, the benchmark used for chaotic cases.","marker":"[85]"},{"why":"Provides the participation-ratio concept that the paper generalizes to phase space as $\\mathrm{IPR}_\\phi$.","marker":"[87]"}],"fun_headline_variants":["Lyapunov exponent sets steady-state statistics in dissipative trimer","Quantum steady state mirrors classical chaos in Bose-Hubbard trimer","Chaos imprints Wigner-Dyson, regularity leaves Poisson in open trimer","Dissipative chaos leaves Wigner-Dyson in steady-state entropy","Langevin trajectories reproduce quantum steady-state spectrum"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole correspondence rests on the assumption that quantum fluctuations around the classical trajectory are accurately represented by adding Gaussian noise to the classical equations, and that this quadratic-order approximation remains valid at arbitrarily long times.","fun_headline_variants_meta":{"raw":{"variants":["Lyapunov exponent sets steady-state statistics in dissipative trimer","Quantum steady state mirrors classical chaos in Bose-Hubbard trimer","Chaos imprints Wigner-Dyson, regularity leaves Poisson in open trimer","Dissipative chaos leaves Wigner-Dyson in steady-state entropy","Langevin trajectories reproduce quantum steady-state spectrum"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00092,"raw_usage":{"total_tokens":4005,"prompt_tokens":1064,"completion_tokens":2941,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":680,"completion_tokens_details":{"reasoning_tokens":2851}},"tokens_in":680,"tokens_out":2941,"duration_ms":19515,"temperature":1.0,"reasoning_tokens":2851,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T00:10:05.395179+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a parameter set of the trimer with no strong symmetry and an unambiguously positive leading Lyapunov exponent, exactly diagonalize the Lindblad equation at $N=30$, and compute the level-spacing ratio distribution of $H_{\\mathrm{eff}}$; if that distribution fails to converge toward the GUE prediction, the claimed Lyapunov-to-statistics correspondence is falsified.","supporting_citations":[{"cited_title":"Equilib- rium fluctuations in maximally noisy extended quantum systems.SciPost Phys., 6:045, 2019","cited_arxiv_id":null,"evidence_quote":"States the conjecture that regular classical motion leads to Poissonian level statistics, the benchmark used for regular cases."},{"cited_title":"Generalized coherent states and their ap- plications.Springer-Verlag New York Inc., 1 1986","cited_arxiv_id":null,"evidence_quote":"States the conjecture that chaotic classical motion leads to Wigner-Dyson statistics, the benchmark used for chaotic cases."},{"cited_title":"Tabor, and John Michael Zi- man","cited_arxiv_id":null,"evidence_quote":"Provides the participation-ratio concept that the paper generalizes to phase space as $\\mathrm{IPR}_\\phi$."}],"review_version":1}