{"id":"fa464048-f46f-4cfa-93fb-f0faf41fb6e1","arxiv_id":"1908.06617","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Globally coupled circle maps exhibit 'toric chaos' in the invertible regime: chaotic dynamics coexisting with O(N) zero Lyapunov exponents and a delocalized, slowly wandering first Lyapunov vector.","lead":"This paper studies a model of many coupled oscillators and reports a new kind of chaos that lives on a high-dimensional torus, with many neutral directions and a slowly wandering direction of maximum stretching. The result is mainly interesting to nonlinear dynamics researchers because it fills a gap in the old Landau versus Ruelle-Takens debate about how turbulence can arise from many-frequency quasiperiodic motion.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Toric chaos claim hinges on unverified zero-Lyapunov plateau; footnote 32's finite-time threshold with no convergence tests is insufficient to establish O(N) null exponents.","rationale":"The reader's CONDITIONAL verdict is appropriate. The identified weakest assumption is indeed the zero-Lyapunov threshold, and I agree that this is the load-bearing point because the entire definition of toric chaos, the extensive M, and the scaling laws rest on it. I do not see a more fundamental flaw in the paper's argument: the numerical phenomenology is plausible, the authors honestly hedge the fractalization claim, and they provide some robustness checks across coupling forms. However, without convergence tests in integration time and without a threshold-sweep analysis, the central claim that the plateau is truly null remains unverified. The proposed check would either confirm the plateau or reveal it as a finite-time/threshold artifact. Verdict should remain CONDITIONAL, since the concern is substantial but not a demonstrated refutation.","tokens_in":15182,"tokens_out":5247,"duration_ms":58105,"concrete_test":"Take one toric-chaos parameter set (e.g., N=100, b=0.7, a=0.65) and one standard-chaos set (a=0.98, b=0.5), and recompute the full Lyapunov spectrum at T=10^5, 10^6, 10^7, and 10^8 with the same QR/Benettin algorithm. Plot each candidate null exponent λ_i(T) versus T and check that the plateau exponents converge to 0 as T grows, at a rate faster than the 1/sqrt(T) finite-time fluctuation bound, while positive and negative exponents stabilize. Then sweep the zero threshold from 10^-6 to 10^-3 at fixed T=10^7 and verify that the number M is stable; if M changes by more than about 10%, the O(N) null-exponent count is an artifact of threshold choice. Include a known N-dimensional quasiperiodic torus as a control to calibrate the finite-time noise floor at the same T.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that toric chaos in the invertible regime has O(N) null Lyapunov exponents. The only evidence for this plateau is footnote 32: an exponent is declared zero if it lies in the asymmetric window (-1e-5, 1e-4) after 2.5e6 iterations. This is not a measurement of exact zero; it is a threshold applied to finite-time estimates, and no convergence with integration time T is shown. For near-neutral modes, finite-time Lyapunov exponents fluctuate and converge slowly, so a threshold of 1e-4 is not obviously below the finite-time noise floor. If the M 'null' exponents are actually small nonzero values that do not converge to zero, then the attractor is not on an invariant torus, the 'M-dimensional toric chaos' classification fails, and the extensive-torus-dimension, DL ~ M + 0.3N, and Y2 ~ DL^{-1/3} phenomenology lose their stated interpretation. The asymmetric threshold also makes the count M depend on an arbitrary choice; Fig. 1(f), Fig. 2(a), and Supplemental Fig. S2(a) would need to be re-read with a symmetric, threshold-independent criterion.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript studies N-dimensional globally coupled circle maps with heterogeneous frequencies, combining numerical Lyapunov spectra, inverse participation ratios, and bifurcation diagrams. It reports three findings: (i) N-dimensional tori exist but their fraction decays exponentially with N; (ii) in the invertible regime, chaos with many near-zero Lyapunov exponents, termed toric chaos, occurs for N≥3, with an extensive number M∼O(N) of null exponents, a delocalized first Lyapunov vector obeying Y2∼DL^{-1/3}, and 1/f^{3/2} slow dynamics; (iii) the torus-to-chaos transition appears to proceed through fractalization (strange nonchaotic) of the torus, although the authors state that this is difficult to prove rigorously. The paper is framed as a numerical exploration and discusses possible relevance to neural dynamics and turbulence.","tokens_in":15374,"tokens_out":6868,"duration_ms":74605,"significance":"If confirmed, toric chaos extends the Ruelle-Takens-Newhouse picture by demonstrating a class of high-dimensional chaotic attractors whose neutral subspace dimension grows linearly with system size, and it gives concrete, falsifiable scaling predictions (Y2∼DL^{-1/3}, 1/f^{3/2} spectra). Strengths include long simulations (2.5×10^6 iterations), scaling collapses over N=50–800, and robustness checks with heterogeneous and Kuramoto-type couplings. The central caveat is that the null-exponent count rests on a finite-time threshold rather than on convergence tests, so the extensive-torus-dimension claim is not yet established.","major_comments":[{"comment":"The criterion used to count null Lyapunov exponents, namely a value in (-1e-5, 1e-4) after 2.5×10^6 iterations, is asymmetric and does not establish that the exponents are zero in the infinite-time limit. Near-neutral exponents in high-dimensional systems converge slowly and have finite-time fluctuations that can easily be of order 1e-4; without convergence tests in the integration time T and in N, the claims of O(N) null Lyapunov exponents and M-dimensional toric chaos are not yet supported. I request plots of individual exponents versus T for representative parameters, a threshold-sensitivity analysis (e.g., symmetric thresholds at 1e-3, 1e-4, 1e-5), and a statement of how M/N behaves as T and N increase.","section":"Footnote [32]; Figs. 1(f) and 2(a)"},{"comment":"The extensivity of M rests on the scaled Lyapunov spectra, but the neutral plateau in Fig. 2(a) is displayed only at a fixed threshold, and the inset for N=100–800 does not show whether the number of exponents satisfying the zero criterion converges to a well-defined fraction of N as N grows. Reporting M/N and its dependence on N and T, with error bars over the random frequency ensembles, is needed to substantiate the claim that the torus dimension is extensive.","section":"Fig. 2(a) and the paragraph beginning 'Toric chaos shows the accumulation...'"},{"comment":"The paper identifies toric chaos with the presence of M null Lyapunov exponents, but null exponents can also arise from marginal directions, symmetries, or slow manifolds that are not quasiperiodic tori. The manuscript does not provide a direct check that the neutral directions correspond to angles circulating on an invariant torus, such as bounded phases with nonzero rotation numbers. At minimum, this identification should be stated as an assumption and a concrete numerical test for it should be described.","section":"Definition of toric chaos and the sentence 'such chaos exists on (or in the vicinity of) a torus'"}],"minor_comments":[{"comment":"The word 'appearrance' should be 'appearance'.","section":"Page 3 (text near Fig. 2)"},{"comment":"The statement that Σλ_i≤0 always holds in the invertible regime is not derived and is only confirmed numerically; please either provide a proof or qualify it to the studied parameter range.","section":"Page 3, paragraph 'Note that, in the invertible regime...'"},{"comment":"The box-counting dimension is said to be computed for 240^3 bins; please clarify whether this means 240 bins per dimension and whether the estimate is checked for convergence in box size.","section":"Fig. 3(d) and its caption"},{"comment":"The fitted values p≈0.36 and p≈0.49 are presented without uncertainties or goodness-of-fit measures; since the estimate is explicitly rough, please state that these values are illustrative only.","section":"Footnote [33]"},{"comment":"The convention for torus dimension in the map versus in the flow (M in the map, M+1 in the flow) is easy to confuse; please define the convention explicitly in the captions of Fig. 1(c) and Fig. 1(f).","section":"Figs. 1(c) and 1(f)"},{"comment":"The fractalization route is presented as 'suggested' and 'difficult to prove,' but the concluding discussion treats it as a likely mechanism; please label it explicitly as a conjecture so that it is not read as an established result.","section":"Conclusion, last paragraph"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Things you should know: this is a numerical paper about a new kind of chaotic attractor in N globally coupled circle maps. It claims that invertible high-dimensional maps can support chaos on a torus: one or a few positive Lyapunov exponents plus O(N) neutral directions, with the first Lyapunov vector delocalized as Y2 ~ D_L^{-1/3} and showing 1/f^{3/2} dynamics. That would build a bridge between Landau's and Ruelle-Takens's pictures of turbulence, and it is a genuinely new characterization. Earlier work on three-frequency tori and toroidal chaos did not report extensive null-exponent bands or this scaling law.\n\nWhat is good: the simulations are long (2.5e6 iterations after 5e5 transient) and the scaling plots are consistent across N, parameter ranges, and different coupling forms in the supplement. The authors are honest about the fractalization claim, explicitly saying it is difficult to prove numerically. The exponential rarity of N-tori is presented as a rough estimate with a fitted locking probability, and they call it rough. The citation pattern is fine: Kaneko's earlier work is contextual, not definitionally load-bearing.\n\nThe soft spots are real but not fatal on their own. The central number M is defined by a threshold in footnote 32: a Lyapunov exponent is counted as zero if it is less than 1e-4 and greater than -1e-5 after 2.5e6 iterations. That threshold is asymmetric, hand-chosen, and there is no convergence test with integration time or N. If the neutral plateau consists of genuinely zero exponents in the infinite-time limit, the torus interpretation is sound; if those exponents are merely small finite-time values, the \"M-dimensional toric chaos\" classification and the Y2 ~ M^{-1/3} story lose their footing. That is the main issue. There are also no error bars on fractions or exponents, and no code or data release. Those are common in numerical letters, but they matter more here because the whole classification hinges on distinguishing zero from small nonzero.\n\nMy take: I agree with the reader's conditional verdict, and the stress-test critique lands on the right spot. But I would not reject the paper over it. The phenomenon is plausible, the paper is careful in most places, and the authors flag their own uncertainties. A serious referee should ask for a symmetric or threshold-free criterion for null exponents, a convergence check showing the count stabilizes with T and N, error bars, and code or data. If those come out, the result would be a solid advance for nonlinear dynamics.\n\nThis is for nonlinear dynamics readers working on coupled oscillator networks, Lyapunov spectra, or routes to turbulence. It deserves peer review rather than desk rejection.","headline":"Toric chaos is a plausible new attractor class but the O(N) null-exponent count rests on a single asymmetric finite-time threshold; worth serious referee attention with code and convergence checks.","tokens_in":15955,"tokens_out":2250,"would_cite":true,"duration_ms":24157,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["37D45","37C55"],"pacs":[],"model":"deepseek-v4-flash","headline":"In coupled circle maps with three or more phases, chaos can coexist with a growing number of exactly neutral directions, forming what the authors call toric chaos.","keywords":["toric chaos","high-dimensional torus","Lyapunov exponents","globally coupled circle maps","quasiperiodicity","inverse participation ratio","chaotic itinerancy","fractal torus"],"falsifier":"Extend the Lyapunov computation to longer integration times at fixed parameters and check whether the number $M$ of exponents below the $10^{-4}$ threshold stays constant or shrinks; if $M$ decreases systematically with time, the neutral plateau is a finite-time artifact. Separately, vary $N$ while holding the Lyapunov dimension $D_L$ fixed and see whether the inverse participation ratio follows $D_L^{-1/3}$; if $Y_2$ instead stays $O(1)$, the claimed delocalization is not a property of the torus dimension.","tokens_in":14917,"feed_emoji":"🌀","tokens_out":13052,"duration_ms":118091,"temperature":0.7,"pith_summary":"This paper asks whether quasiperiodic motion with many frequencies can pass into chaos while keeping its high-dimensional torus skeleton. Using $N$ globally coupled circle maps with heterogeneous natural frequencies, it shows that for $N \\ge 3$ chaos can appear even when the map is invertible, and this 'toric chaos' carries not just one or a few positive Lyapunov exponents but also a number of null (neutral) exponents that grows linearly with $N$. The first Lyapunov vector of toric chaos is neither fully delocalized nor localized: its inverse participation ratio $Y_2$ decays as the inverse cube root of the Lyapunov dimension $D_L$, and its power spectrum shows $1/f^{3/2}$ slow itinerancy. If true, the result means that high-dimensional tori are not transient curiosities; they host a recurring type of chaos with an extensive neutral dimension, relevant to systems in which many quasiperiodic modes coexist with disorder.","feed_headline":"Chaos can ride a high-dimensional torus","feed_subtitle":"In coupled circle maps, chaotic motion coexists with many zero Lyapunov exponents and slow, delocalized dynamics.","key_machinery":"The load-bearing object is the $N$-dimensional globally coupled circle map together with its Lyapunov spectrum, especially the count $M$ of null Lyapunov exponents and the inverse participation ratio $Y_2$ of the first Lyapunov vector. The map combines heterogeneous natural frequencies with global sinusoidal coupling and can be invertible when the nonlinearity $a$ lies below a critical value. The diagnostic that carries the argument is the plateau of $O(N)$ exponents that are exactly or nearly zero; chaos with such a plateau is identified as chaos 'on a torus.' The scaling $Y_2 \\sim D_L^{-1/3}$ and the $1/f^{3/2}$ power spectrum serve as signatures of the first Lyapunov vector's slow, itinerant spread across the torus.","core_discovery":"The central discovery is the existence and prevalence of toric chaos in $N$-dimensional globally coupled circle maps with $N \\ge 3$. For invertible parameter regimes ($a$ below the critical value $a_c$), where the one-dimensional circle map cannot be chaotic, chaotic attractors appear that sit on or near an $M$-dimensional torus: their Lyapunov spectra show one or a few positive exponents, $O(N)$ exponents indistinguishable from zero, and compensating negative exponents. The number $M$ of neutral directions is extensive, growing linearly with $N$, and the first Lyapunov vector's inverse participation ratio $Y_2$ scales as $D_L^{-1/3}$, indicating that the unstable direction spreads over the torus but with partial localization, while its power spectrum shows $1/f^{3/2}$ fluctuations. The paper also reports that full $N$-dimensional tori exist but become exponentially rare with $N$, and that the torus-to-chaos transition passes through fractalized tori, interpreted as strange nonchaotic attractors.","pith_inferences":["If the paper's supplement relation $D_L \\simeq M + 0.3N$ is taken literally, toric chaos in the large-$N$ limit would carry a fixed density of positive-expansion directions, so the chaotic component itself would be extensive.","A sharper definition of the neutral plateau, based on convergence with integration time rather than the fixed $10^{-4}$ threshold, would let the claim $M = O(N)$ be tested as an asymptotic property rather than a finite-time count.","Because the paper reports the same qualitative behavior with heterogeneous couplings and with Kuramoto-type sine coupling, one can expect toric chaos in other large coupled-oscillator networks, where the same $Y_2 \\sim D_L^{-1/3}$ and $1/f^{3/2}$ signatures could be sought.","The $1/f^{3/2}$ slow dynamics of the first Lyapunov vector resembles intermittency in shell models of turbulence; checking whether the same exponent controls energy-transfer fluctuations in such models would connect toric chaos to the Landau picture of turbulence."],"forward_implications":["For systems with $N \\ge 3$ phases in the invertible regime, toric chaos is a possible attractor and its frequency of occurrence grows with $N$ before saturating, so high-dimensional tori are not automatically destroyed by weak nonlinearity.","The neutral dimension $M$ is extensive, so in the large-$N$ limit toric chaos carries a macroscopic number of near-zero Lyapunov exponents, making its effective attractor dimension scale with system size.","Unlike standard chaos, where the first Lyapunov vector is localized, toric chaos has a delocalized first Lyapunov vector with $Y_2$ approximately proportional to $D_L^{-1/3}$ and slow $1/f^{3/2}$ dynamics.","The transition from a torus to toric chaos is accompanied by fractalization of the torus, suggesting that autonomous strange nonchaotic attractors can appear at this transition.","Toric chaos provides a possible dynamical picture for chaotic itinerancy and for systems such as EEG or turbulence that combine broad frequency spectra with quasiperiodic components."],"supporting_citations":[{"why":"Supplies the theorem that chaos can occur in any neighborhood of a quasiperiodic flow.","marker":"[21]"},{"why":"Extends the theorem to strange Axiom A attractors near quasiperiodic flows on $T^m$ with $m \\ge 3$, justifying chaos near high-dimensional tori.","marker":"[22]"},{"why":"Provides numerical evidence on three-frequency quasiperiodic orbits and the competition between tori and chaos.","marker":"[24]"},{"why":"Gives the earlier three-coupled-oscillator example of mode-locking and toroidal chaos that the present work generalizes.","marker":"[26]"},{"why":"Defines the globally coupled circle map model class and the context of collective dynamics and chaotic itinerancy.","marker":"[30]"},{"why":"Supplies the inverse participation ratio as the localization measure used to quantify the first Lyapunov vector.","marker":"[34]"},{"why":"Provides the Lyapunov analysis method for coupled map systems.","marker":"[35]"},{"why":"Introduces strange nonchaotic attractors, the concept used to interpret fractalized tori at the transition.","marker":"[38]"}],"fun_headline_variants":["Toric chaos: chaos on high-dimensional tori","Extensive neutral modes mark toric chaos","Chaos coexists with many zero Lyapunov exponents","Torus chaos: delocalized slow dynamics","Invertible maps can still show chaos on tori"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The classification of toric chaos depends on counting a Lyapunov exponent as exactly zero whenever it falls between $-10^{-5}$ and $10^{-4}$ on a finite simulation; if those exponents actually converge to small nonzero values as the simulation time grows, the claim of $O(N)$ null exponents and the torus dimension would fail.","fun_headline_variants_meta":{"raw":{"variants":["Toric chaos: chaos on high-dimensional tori","Extensive neutral modes mark toric chaos","Chaos coexists with many zero Lyapunov exponents","Torus chaos: delocalized slow dynamics","Invertible maps can still show chaos on tori"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000203,"raw_usage":{"total_tokens":1356,"prompt_tokens":883,"completion_tokens":473,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":499,"completion_tokens_details":{"reasoning_tokens":398}},"tokens_in":499,"tokens_out":473,"duration_ms":5496,"temperature":1.0,"reasoning_tokens":398,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:39:47.881731+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Extend the Lyapunov computation to longer integration times at fixed parameters and check whether the number $M$ of exponents below the $10^{-4}$ threshold stays constant or shrinks; if $M$ decreases systematically with time, the neutral plateau is a finite-time artifact. Separately, vary $N$ while holding the Lyapunov dimension $D_L$ fixed and see whether the inverse participation ratio follows $D_L^{-1/3}$; if $Y_2$ instead stays $O(1)$, the claimed delocalization is not a property of the torus dimension.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the theorem that chaos can occur in any neighborhood of a quasiperiodic flow."},{"cited_title":"On the nature of turbulence,","cited_arxiv_id":null,"evidence_quote":"Extends the theorem to strange Axiom A attractors near quasiperiodic flows on $T^m$ with $m \\ge 3$, justifying chaos near high-dimensional tori."},{"cited_title":"Turaev, Maps Close to Identity and Universal Maps in the Newhouse Domain, Commun","cited_arxiv_id":null,"evidence_quote":"Provides numerical evidence on three-frequency quasiperiodic orbits and the competition between tori and chaos."},{"cited_title":"Kaneko, Fates of Three-Torus I – Double Devil’s Stair- cases in Lockings, Prog","cited_arxiv_id":null,"evidence_quote":"Gives the earlier three-coupled-oscillator example of mode-locking and toroidal chaos that the present work generalizes."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the globally coupled circle map model class and the context of collective dynamics and chaotic itinerancy."},{"cited_title":"Accordingly, the decrease ofT N andT 0 can be fitted byp≃ 0.36 for a = 0.7 and p≃ 0.49 for a = 0.8 [see also Supplemental Material [29], Fig","cited_arxiv_id":null,"evidence_quote":"Supplies the inverse participation ratio as the localization measure used to quantify the first Lyapunov vector."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the Lyapunov analysis method for coupled map systems."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces strange nonchaotic attractors, the concept used to interpret fractalized tori at the transition."}],"review_version":1}