{"id":"9f23f4d3-c9a5-4142-ab0e-924af89606bf","arxiv_id":"2608.06537","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A two-step heteroclinic bifurcation mechanism converts a simple exact coherent structure into a chaotic edge state and a fractal basin boundary in subcritical Taylor-Couette flow.","lead":"This paper shows how a simple travelling-wave edge state in minimal Taylor-Couette flow turns into a chaotic edge state through two successive heteroclinic tangencies, first with a chaotic saddle that makes the basin boundary fractal, then through a heteroclinic cycle that makes the edge state chaotic. It provides a concrete mechanism for the sensitive dependence on initial conditions observed in subcritical shear flow transition, a long-standing puzzle in fluid dynamics.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Cantor-set and Poisson-escape claims for C3 rest entirely on a spline-fitted 1D return map; transverse dynamics are unquantified, so the fractal basin-boundary conclusion could be a projection artifact.","rationale":"The reader's weakest_assumption identifies exactly this: the one-dimensional reduction underlies the fractal, escape-rate, and heteroclinic-tangency conclusions. My reading of the manuscript confirms that the Cantor-set construction in §3.3 is explicitly built on a spline-fitted scalar return map, and the paper's own admission in §4.2 that the dynamics near R'_c3 cannot be fully captured by that map sharpens the concern. No machine-checked proof, parameter-free derivation, or reproducible code is provided to independently support the reduction. The concern is not an external disagreement with consensus; it is an internal, quantifiable risk in the argument's foundation. If a concrete transverse-thickness check or direct DNS survival-statistics test were performed and passed, the central claim would be substantially strengthened; if it fails, the fractal-basin-boundary conclusion would be a projection artifact. Since the reader already judged the paper CONDITIONAL on this basis, my stress-test does not change the verdict.","tokens_in":21713,"tokens_out":4597,"duration_ms":44742,"concrete_test":"At R=395.564, apply local PCA to the full DNS states on Σ within the shaded square of Fig. 10d (embedded, after method-of-slices reduction, in the full phase space) and compare the transverse spread (s.d. normal to the best-fit spline manifold) with the smallest gap between the 64 subintervals of S_3 = F^{-3}(S_0). If the transverse spread is not small relative to that gap, the Cantor-set conclusion is not supported. Independently, compute survival probabilities directly from DNS (no spline map) for N_0 = 10^8 initial conditions in the square: if the fraction surviving to iterate n does not follow the exponential law predicted from F, then the Poisson-escape claim fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that C3 is a fractal chaotic saddle and that the basin boundary inherits that fractality depends on §3.3's construction of C3 as a Cantor set via S_n = F^{-n}(S_0), where F is a spline interpolation of the torque return map on the Poincaré section Σ. The paper asserts in §3 that the discrete dynamics on Σ evolve on a nearly one-dimensional manifold of negligible thickness, but it provides no quantitative estimate of that thickness and, in §4.2, explicitly states that the dynamics near R'_c3 cannot be fully grasped with the one-dimensional discrete map approximation. If the neglected transverse directions possess any non-negligible contraction rate, the preimage sets S_n are not one-dimensional intervals; they acquire transverse thickness, and the 4^n subinterval counting, the exponential (Poisson) survival statistics, and the topological-transitivity argument invoked in §3.3 and §4.1 to extend local tangency to the whole saddle would all be artifacts of the projection. The paper's further claim of a 'first direct demonstration of the fractal nature of a chaotic saddle in a shear flow' therefore rests on an unverified dimensional reduction. This is the load-bearing assumption because both the escape-rate analysis and the inference of a heteroclinic tangency at R'_c3 are conducted through this one-dimensional map, rather than through direct manifold computations in the full phase space.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies the subcritical regime of Taylor-Couette flow in a minimal parallelogram-annular domain, varying the inner Reynolds number R while keeping R_o=-1200. Building on previously published results, it constructs an extended bifurcation diagram involving the DRW travelling wave, the P2/P3 periodic-orbit families, and the chaotic sets C2 and C3. The central claim is a two-step mechanism for the emergence of a chaotic edge state: first, at R'_c3, the unstable manifold of the chaotic saddle C3 becomes tangent to the stable manifold of DRW_L, subsuming C3 into the laminar/turbulent basin boundary while DRW_L remains the edge state (type ii); second, at R_het, a heteroclinic cycle between DRW_L and P3_L forms, so that the chaotic saddle replaces DRW_L as the edge state (type iii). The paper further claims, via a spline-fitted one-dimensional return map F on the Poincaré section Σ, that C3 is a Cantor set and that its escape statistics are Poissonian, which would be the first direct demonstration of fractal structure of a chaotic saddle in a shear flow.","tokens_in":22005,"tokens_out":9594,"duration_ms":83000,"significance":"The proposed mechanism is significant: type (ii) and type (iii) chaotic saddles have not previously been explained in shear flows, and the paper connects them to heteroclinic tangencies and to the Poisson relaminarization statistics observed in pipe and plane-Couette flows. The numerical work is substantial and internally consistent: exact coherent states are tracked with a Poincaré-Newton-Krylov method, edge tracking is used to identify the edge state, and the bifurcation scenario is supported by multiple DNS runs at closely spaced Reynolds numbers. The reliance on previously published results (Wang et al. 2022, 2025a,b) is appropriate because those results are external and the new analysis is based on new return-map and manifold computations. However, the two load-bearing novelties—the Cantor-set structure of C3 and the heteroclinic tangency at R'_c3—rest on a dimensional reduction whose accuracy is not quantified. If those issues are resolved, this would be a valuable contribution to the dynamical-systems understanding of subcritical transition.","major_comments":[{"comment":"The claim that C3 is a Cantor set, and the accompanying escape-rate and Poisson-statistics analysis, rest entirely on the spline-interpolated one-dimensional return map F. The premise stated at the start of §3 is that the discrete dynamics on Σ evolve on a nearly one-dimensional manifold of negligible thickness, but no quantitative estimate of that thickness is given, nor is there a comparison between the preimage sets S_n=F^{-n}(S_0) and sets obtained by direct integration of the DNS. If the neglected transverse directions have non-negligible contraction rates, the sets S_n are not unions of intervals and the 4^n counting, the Cantor-set conclusion, and the survival statistics in Fig. 11 would be artifacts of the projection. Please quantify the transverse thickness (e.g., the variance of DNS points about the fitted manifold) and show that S_1 and S_2 are robust to the spline interpolation and to small transverse perturbations.","section":"§3.3, Fig. 11"},{"comment":"The identification of R'_c3 as a heteroclinic tangency between W^u(C3) and W^s(DRW_L) is inferred from return-map cobwebs and from the observation that cobweb trajectories land near DRW_L. The authors themselves note at the beginning of §4 that the one-dimensional map cannot fully capture the dynamics near R'_c3, so this inference is not sufficient for the load-bearing type (ii) claim. Please provide direct manifold evidence, for example by computing W^s(DRW_L) via edge tracking in the full phase space, measuring its distance to W^u(C3) as a function of R, and showing that this distance vanishes at R'_c3; otherwise the tangency and the resulting fractal basin boundary remain plausible but undemonstrated.","section":"§4.1–4.2, Figs. 10 and 13"},{"comment":"The critical value R'_c3 is inconsistent: §2.2 states R'_c3≈395.563 (consistent with Fig. 6c and Fig. 10 at R=395.564), while the opening paragraph of §4 gives R'_c3≈395.5401, a value that elsewhere is attributed to R_c3≈395.540. Since the paper's central narrative distinguishes R_c3 (boundary crisis of C3) from R'_c3 (heteroclinic tangency with DRW_L), this discrepancy must be corrected and the values cross-checked against figure 3.","section":"§2.2 vs §4 (opening paragraph)"},{"comment":"The topological-transitivity argument used both to justify initializing DNS near P3_L and to extend the local tangency to the whole saddle requires that the invariant set of the four-branch map F be transitive. For a piecewise-monotone non-invertible map on a Cantor set this is not automatic; please verify transitivity numerically (e.g., by constructing a connecting orbit between the four branches or by checking that the symbolic dynamics on the 4^n preimages is a full shift) or restrict the claim to the branch on which transitivity can be demonstrated.","section":"§3.3 and §4.1"}],"minor_comments":[{"comment":"Typo: 'Relaminarision' should be 'relaminarization'.","section":"§2.3"},{"comment":"The label P2_L in the caption is not introduced in the text; please define it (presumably the lower branch of P2) and use it consistently.","section":"Fig. 12c"},{"comment":"The uncertainty estimate R_het≈395.64686±2×10^{-5} is presented without stating the number of independent trajectories or the criterion used to define approach to P3_L; please specify the procedure that yields this uncertainty.","section":"§5.1"},{"comment":"The Poincaré section condition dτ_i/dt > dτ_o/dt is natural, but it would help to state explicitly near (2.4) that travelling-wave solutions such as DRW_L lie on Σ and that the Poincaré map is singular there, since this point is only discussed in §4.3.","section":"Eq. (2.4)"}],"recommendation":"major_revision","confidential_remarks":"To the editor: the paper is a strong numerical study, but its headline claims ('first direct demonstration of fractal structure') exceed what the current evidence supports because the one-dimensional reduction is asserted rather than measured. The inconsistency in R'_c3 should be fixed before acceptance. The novelty relative to the authors' prior papers is real but modest; the new contribution is the global-bifurcation classification and the return-map analysis. I recommend major revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. The central mechanism—a type (i) chaotic saddle being subsumed into the basin boundary via a heteroclinic tangle with the edge state, then converted into a chaotic edge state via a heteroclinic cycle—is new and plausible. It gives a concrete route from simple ECS edge states to chaotic edge states in subcritical shear flow. The paper supports it with extensive DNS, continuation, and edge tracking; the bifurcation diagram is careful and the evidence is internally consistent. This is a real contribution.\n\nThe softer part is the fractal claim. The Cantor-set construction for C3 and the Poisson escape statistics come from a spline-fitted 1D return map on the Poincaré section. The paper asserts the dynamics there are nearly one-dimensional with negligible thickness, but gives no quantitative estimate. It even says in §4.2 that the 1D map cannot fully grasp dynamics near R'_c3. That tension matters: if the transverse directions are dynamically important, the S_n = F^{-n}(S_0) counting, the topological transitivity argument, and the \"first direct demonstration\" of fractality could be artifacts of the projection. The heteroclinic tangency itself is inferred from trajectory outcomes, so the type (ii) conversion is on firmer ground than the Cantor set. But the strong fractal headline should be toned down or backed by direct manifold computations or at least an estimate of the transverse contraction rate.\n\nAlso, no code or data is provided; the spline fits and escape statistics are not reproducible from the paper alone. And the Poisson claim is computed from the map, not from DNS lifetimes, so the link to memoryless decay remains inferential.\n\nThat said, the mechanism is the core of the paper and it holds up. The 1D reduction is a legitimate concern, but it is a fixable one—quantify the thickness, or compute heteroclinic tangencies directly. This paper deserves a serious referee. I would send it out, with a request that the fractal claims be either hardened or softened, and that the data/code be made available if possible.","headline":"A genuinely new two-step heteroclinic mechanism for chaotic edge states in subcritical shear flow, with a solid numerical core but a fractal claim that leans on an unquantified 1D reduction.","tokens_in":22525,"tokens_out":3537,"would_cite":true,"duration_ms":35629,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["37D45","37C29","76F06"],"pacs":["47.20.Ft","47.27.Cn","05.45.-a"],"model":"deepseek-v4-flash","headline":"This paper shows that a simple edge state in Taylor-Couette flow becomes chaotic when a chaotic saddle is first folded into the basin boundary by a heteroclinic tangle and then promoted to the edge state by a heteroclinic cycle.","keywords":["chaotic saddle","basin boundary","edge state","subcritical transition","Taylor-Couette flow","heteroclinic bifurcation","Cantor set","period-doubling cascade"],"falsifier":"At $R=395.564$, one could take the spline map $F$, construct $S_n = F^{-n}(S_0)$ up to $n=6$, and check whether transient orbits whose first iterates lie in $S_n$ but whose transverse coordinates are perturbed at order $10^{-8}$ escape at the rate predicted by the one-dimensional map; if the escape statistics or the $4^n$ interval count are not reproduced, the Cantor-set and Poisson claims are artifacts of the projection.","tokens_in":21533,"feed_emoji":"🌊","tokens_out":6239,"duration_ms":53233,"temperature":0.7,"pith_summary":"The paper aims to show how the edge state, the invariant object that decides whether a perturbed shear flow decays to laminar flow or becomes turbulent, can turn chaotic. In a small periodic domain of counter-rotating Taylor-Couette flow, the authors track one chaotic set, C3, through three roles: a chaotic attractor, a type (i) chaotic saddle away from the basin boundary, a type (ii) saddle embedded in the basin boundary, and finally a type (iii) saddle that is itself the edge state. The type changes are caused by two global bifurcations, a heteroclinic tangle with the travelling-wave edge state DRW_L followed by a heteroclinic cycle connecting DRW_L to a period-3 orbit at the boundary of C3. A sympathetic reader would care because this supplies a concrete mechanism for the sensitive dependence on initial conditions and Poisson-distributed lifetimes observed in subcritical shear flows, and because it shows that observing a simple non-chaotic edge state at one Reynolds number gives no guarantee that the same situation holds at nearby values.","feed_headline":"Chaos takes over a shear flow's laminar-turbulent boundary","feed_subtitle":"Two heteroclinic bifurcations turn a simple edge state into a chaotic one, explaining turbulence's memoryless decay.","key_machinery":"The central objects are the nearly one-dimensional return map $F$ on a Poincar\\'e section $\\Sigma$ defined by equal torques, and the Cantor-set construction $S_n = F^{-n}(S_0)$, whose four branches produce $4^n$ preimage intervals whose limit is the chaotic saddle. The return map lets the authors measure escape rates and prove the Cantor structure, while heteroclinic tangency theory between the stable and unstable manifolds of DRW_L, C3, and P3_L supplies the two global bifurcations that move the saddle from the interior of a basin to the basin boundary and finally to the edge-state role.","core_discovery":"The central claim is that the chaotic saddle C3 is subsumed into the basin boundary and becomes type (ii) in a heteroclinic bifurcation at $R'_{c3}$ involving the edge state DRW_L, and that a second heteroclinic bifurcation at $R_{\\mathrm{het}}$ generates a heteroclinic loop between DRW_L and P3_L that transforms the saddle into type (iii), whereupon the chaotic saddle replaces DRW_L as the edge state. The paper further claims that the chaotic saddle is a Cantor set, constructed as $S_\\infty = \\lim_{n\\to\\infty} F^{-n}(S_0)$ with each $S_n$ consisting of $4^n$ subintervals, and that this is the first direct demonstration of the fractal structure of a chaotic saddle in a shear flow. It claims that the fractal structure and the near-constant escape rate of the one-dimensional return map account for the memoryless, Poisson-like escape statistics seen from such saddles.","pith_inferences":["If this two-step scenario, a heteroclinic tangle followed by a heteroclinic cycle, is generic, similar type (ii) and type (iii) saddle conversions should appear in plane Couette and pipe flow; the paper notes that the plane Couette case already shows one tangency of each kind.","The Cantor-set construction suggests that the fractal dimension of the basin boundary and the escape rate are related through the return map $F$; computing that dimension and comparing it with lifetime statistics in longer domains would test the framework beyond the minimal box.","The edge-state replacement at $R_{\\mathrm{het}}$ implies that control strategies based on a simple lower-branch exact coherent structure will fail across such global bifurcations, because a controller that assumes DRW_L remains the edge state would be invalid above $R_{\\mathrm{het}}$.","The apparent cascade of heteroclinic bifurcations accumulating at $R_{\\mathrm{hom}}$, noted but not resolved in the paper, raises the possibility that the edge state switches repeatedly, which could help explain why identifying a sharp critical Reynolds number is notoriously difficult in longer pipes."],"forward_implications":["Below $R'_{c3}$, the chaotic saddle C3 is type (i): all nearby trajectories leak to the non-trivial attractor P2 and none reach the laminar state; above $R'_{c3}$, C3 lies in the basin boundary separating P2 and CCF, so the basin boundary inherits fractal structure while DRW_L remains the edge state.","Above $R_{\\mathrm{het}}$, trajectories from DRW_L can no longer reach P2; edge tracking converges to a chaotic state that spends most of its time near P3_L and occasionally visits DRW_L, so the chaotic saddle replaces DRW_L as the edge state.","Because global bifurcations change the nature of the edge state as the Reynolds number increases, observing a simple non-chaotic edge state at one parameter value does not guarantee that the same state is the edge state at nearby values.","Fractal structure in the chaotic saddle gives a precise rationale for the memoryless, Poisson-distributed turbulence lifetimes seen in subcritical shear flows."],"supporting_citations":[{"why":"Supplies the Taylor-Couette setup, the DRW branch, and the parallelogram-annular domain used throughout the study.","marker":"Wang et al. (2022)"},{"why":"Established the Feigenbaum period-doubling cascade producing the chaotic attractors C2 and C3 and provided the near-one-dimensional return map.","marker":"Wang et al. (2025a)"},{"why":"Identified the P3 orbit and the chaotic set C3 that are central to the crisis and heteroclinic scenarios.","marker":"Wang et al. (2025b)"},{"why":"Provides the canonical lower-branch edge-state picture that this paper generalizes.","marker":"Wang et al. (2007)"},{"why":"Gives the plane-Couette tangency scenario used to argue genericity and to contrast a travelling-wave edge state with a periodic-orbit edge state.","marker":"Lustro et al. (2019)"},{"why":"Supplies the boundary-crisis concept used to classify the first transformation of C3 into a chaotic saddle.","marker":"Grebogi et al. (1983, 1986)"},{"why":"Provides the Smale-Birkhoff theorem linking homoclinic and heteroclinic tangles to chaos on Cantor sets.","marker":"Smale (1965)"},{"why":"Supplies an analogous type (i) chaotic saddle in pipe flow and its geometric characterization.","marker":"Budanur et al. (2019)"}],"fun_headline_variants":["Heteroclinic tangles explain chaos in shear flow's basin boundary","Two bifurcations turn a simple edge state into a chaotic one","Fractal basin boundaries emerge from chaotic saddles in shear flow","Why shear-flow turbulence decay is memoryless: edge state chaos"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole fractal and escape-rate analysis assumes that the dynamics on the Poincar\\'e section evolve on a nearly one-dimensional manifold of negligible thickness, so that the spline return map $F$ together with its branch-selection rules fully captures the phase-space structure around C3.","fun_headline_variants_meta":{"raw":{"variants":["Heteroclinic tangles explain chaos in shear flow's basin boundary","Two bifurcations turn a simple edge state into a chaotic one","Fractal basin boundaries emerge from chaotic saddles in shear flow","Why shear-flow turbulence decay is memoryless: edge state chaos"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000238,"raw_usage":{"total_tokens":1548,"prompt_tokens":1017,"completion_tokens":531,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":633,"completion_tokens_details":{"reasoning_tokens":455}},"tokens_in":633,"tokens_out":531,"duration_ms":5551,"temperature":1.0,"reasoning_tokens":455,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T04:16:41.756035+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"At $R=395.564$, one could take the spline map $F$, construct $S_n = F^{-n}(S_0)$ up to $n=6$, and check whether transient orbits whose first iterates lie in $S_n$ but whose transverse coordinates are perturbed at order $10^{-8}$ escape at the rate predicted by the one-dimensional map; if the escape statistics or the $4^n$ interval count are not reproduced, the Cantor-set and Poisson claims are artifacts of the projection.","supporting_citations":[{"cited_title":"and Ayats, R","cited_arxiv_id":null,"evidence_quote":"Supplies the Taylor-Couette setup, the DRW branch, and the parallelogram-annular domain used throughout the study."}],"review_version":1}