{"id":"48ccbefb-2f03-4d30-a00b-515c53a8b923","arxiv_id":"2507.01583","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"Turbulent puffs in square duct flow have finite lifetimes that scale as (Rec-Re)^(-1/2) below Re ~ 1450 and super-exponentially above, consistent with a noisy saddle-node bifurcation.","lead":"Direct numerical simulations show that turbulent puffs in square duct flow are transient, with mean lifetimes following a square-root scaling law below Reynolds number 1450 and a super-exponential law above it. A noisy saddle-node bifurcation model explains both regimes, suggesting a universal decay mechanism shared with pipe and channel flows.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Pattern-preservation collapse is the load-bearing assumption; the predicted critical point (E_kc=0.355) lies in a high-E_k region where low-Re trajectories are sparse, so Rec and the saddle-node derivation may rest on extrapolated fits.","rationale":"The reader's weakest assumption is the pattern preservation approximation, and I agree that it is the most load-bearing part of the central claim. My reading of the manuscript sharpens this: the approximation is not independently validated, and the specific location of the predicted critical point exacerbates the risk. The paper's own statement that low-Re puffs 'seldom reach high values' implies that the E_k=0.355 region used to extract Rec from Eq. (7) may be under-sampled; the quintic fits in Fig. 4 could therefore be extrapolations. This is not an external disagreement with consensus but an internal validation gap: the theoretical explanation is only as solid as the P_d-D_d collapse, and no quantitative error measure is provided. The paper does have independent support: it uses a well-established solver (Nek5000), includes resolution checks, tests two initial puff conditions, and states insensitivity to the relaminarization threshold. These are real strengths, but they do not resolve the sampling/extrapolation issue. The empirical two-regime lifetime scaling may survive even if the mechanism is not proven, which is why the reader's CONDITIONAL verdict remains appropriate. A direct deterministic ODE integration using the fitted P_d and D_d, or a well-resolved recomputation of the P_d-D_d ratio at high E_k, would settle whether the noisy saddle-node reduction actually explains the observed scaling laws. Until then, the central claim should be viewed as conditional on the pattern-preservation approximation being quantitatively confirmed.","tokens_in":7880,"tokens_out":11992,"duration_ms":142529,"concrete_test":"Recompute P_d(E_k) and D_d(E_k) separately for Re=1380, 1420, and 1440, using only trajectories whose initial E_k exceeds 0.4 so the high-E_k band is adequately sampled; fit each with a quintic and evaluate the minimum of D_d/P_d. If the inferred Rec varies by more than ~30 across these Re, pattern preservation fails. Additionally, report the number of samples per E_k slice near E_k=0.355; if it is below a few hundred, the quintic fit and the predicted Rec should be regarded as extrapolations.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central reduction to the noisy saddle-node equation (Eqs. 8-9) rests entirely on the pattern preservation approximation: deterministic production P_d and dissipation D_d are treated as functions of E_k alone and independent of Re. The evidence is the visual collapse of P_d and D_d curves at different Re in Fig. 4, but the text notes that at low Re the puff energy 'seldom reaches high values', so the high-E_k part of these curves is sparsely populated. The predicted critical parameters (Rec, Ekc) = (1445, 0.355) from Eq. (7) are obtained from quintic fits to these same curves; if E_k=0.355 lies in a region with few samples, the minimum of D_d/P_d and hence Rec could be an extrapolation artifact. The collapse is not quantified (no scatter, no error bars, no tabulated values). If P_d and D_d actually depend on Re or on puff history beyond E_k, the quadratic expansion at E_kc is invalid and the derived square-root and super-exponential scaling laws do not follow. The empirical scaling laws in Fig. 3 could still hold, but the claimed mechanism would be unsupported.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript reports direct numerical simulations of localized turbulent puffs in square duct flow over a Reynolds number range around 1400–1600. The central claim is that puffs are always transient, with the mean lifetime τ following τ = (a1 Re + b1)^{-1/2} for Re < 1440 and ln[ln(τ)] = a2 Re + b2 for Re > 1450, with a critical Reynolds number Re_c ≈ 1450 separating the two regimes. The authors propose a theoretical explanation based on a pattern-preservation approximation, which reduces the Reynolds–Orr kinetic energy equation to a noisy saddle-node bifurcation equation (Eq. 9). They argue that the square-root law arises from critical slowing down in the subcritical regime and that the super-exponential law arises from noise-activated barrier crossing in the supercritical regime. They also document ensemble-averaged secondary flow structures and conclude that the decay mechanism is geometrically invariant between pipe and duct flows.","tokens_in":8126,"tokens_out":4865,"duration_ms":56645,"significance":"If the claims hold, this is a valuable contribution: it extends the transient-puff picture and the noisy saddle-node framework from pipe flow to duct flow, provides quantitative lifetime scaling laws, and is supported by a substantial DNS campaign of about 1000 individual puff decay simulations. The paper is clearly written, the numerical setup is documented, and the theoretical reduction is conceptually appealing. The main caveat is that the central scaling laws are fits to mean lifetimes without reported uncertainties, and the pattern-preservation collapse that underpins the theory is established only visually from the same DNS data. The significance is therefore conditional on these statistical and methodological gaps being addressed.","major_comments":[{"comment":"The two central scaling laws are fitted to mean lifetimes without any reported uncertainty. With M = 100 and M = 50 per Reynolds number and approximately exponential lifetime distributions, the standard error of each mean is on the order of τ/√M (10–14%), which is large enough to affect the fitted exponents and the inferred Re_c. Please provide bootstrap confidence intervals for the fitted parameters, show the lifetime distributions for representative Re, and demonstrate that the choice of relaminarization threshold E_k = 0.05 does not change the fitted scaling laws quantitatively.","section":"Fig. 3, Eqs. (4)–(5)"},{"comment":"The pattern preservation approximation is the load-bearing assumption, but the collapse of P_d and D_d curves across Re is only visual. The text states that at low Re puff energies 'seldom reach high values', so the high-E_k portion of the curves, where E_kc = 0.355 is located, is sparsely sampled. The quintic fits and the predicted Re_c = 1445 from Eq. (7) may therefore be artifacts of extrapolation. Please quantify the data density in E_k, add scatter or error bars to Fig. 4, test the sensitivity of Re_c and E_kc to the polynomial order and to the fitted E_k range, and show that P_d and D_d are statistically indistinguishable across Re. If P_d and D_d depend on Re or on puff history beyond E_k, the quadratic expansion in Eq. (8) and the derived scaling laws do not follow.","section":"Fig. 4, text before Eq. (7)"},{"comment":"The critical Reynolds number Re_c = 1450 is obtained from the square-root fit (Eq. 4) using data with Re < 1440, and this same Re_c is then used to partition the data into the two scaling regimes, so the boundary is not determined independently. This circularity makes it difficult to evaluate whether the two-regime description is actually required. Please fit both regimes jointly with the breakpoint as a free parameter, or determine Re_c independently from the crossing of the two fits, and report the uncertainty in Re_c. The value Re_c = 1445 from Eq. (7) is a consistency check, not an independent determination.","section":"Eq. (4), Eq. (5), Fig. 3"},{"comment":"The derivation of the super-exponential scaling ln[ln τ] = a2 Re + b2 from the noisy saddle-node equation is asserted through a reference to barrier-crossing theory, rather than derived or quantitatively matched. To support the mechanism claim, the authors should either derive the scaling for Eq. (9) with the noise amplitude estimated from DNS, or compare the predicted barrier height and escape rate with the fitted a2 and b2. As written, the empirical Eq. (5) and the theoretical Eq. (9) are connected only qualitatively.","section":"Eq. (9) and following"}],"minor_comments":[{"comment":"There are several typographical errors, including 'Naiver-Stokes' for 'Navier-Stokes' and missing spaces in phrases such as 'analyzethe' and 'and 𝑅𝑒𝑐'.","section":"Abstract and text"},{"comment":"The reference for Nek5000 lists 'F. Paul, J. W. L. Fischer, S. G. Kerkemeier'; the standard citation is P. F. Fischer, J. W. Lottes, and S. G. Kerkemeier. Please correct the author list.","section":"Reference [25]"},{"comment":"The ordinate labels should explicitly state the transformed variables (τ^{-2} and ln[ln τ]) and should include error bars or a statement explaining why they are omitted. The legend would benefit from clarifying the meaning of Re_init = 1500 and 1510.","section":"Fig. 3"},{"comment":"The statement that the data are 'available within the article' is not accurate for the full lifetime dataset; please deposit the individual lifetimes and the P_d and D_d curves in a public repository so that the fits can be reproduced.","section":"Data Availability"}],"recommendation":"major_revision","confidential_remarks":"This manuscript is within scope and the DNS effort is substantial. The main risk is that the theoretical framework, which closely follows the authors' pipe-flow paper (Ref. 24), is applied here with insufficient statistical support. If the authors can provide error bars, quantify the pattern-preservation collapse, and de-circularize the Re_c determination, I would support publication. I would also ask the editor to consider whether the novelty relative to Ref. [24] is sufficiently articulated in the introduction and conclusions."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read this. It's the first DNS lifetime statistics for puffs in square duct flow, and the two scaling laws (square-root below Re_c~1450, super-exponential above) look real. The theoretical reduction to a noisy saddle-node equation is not new—it's carried over from the authors' pipe-flow paper (ref 24) and Zhang & Tao (ref 29)—but the consistency check between the predicted Rec from the production/dissipation curves (1445) and the scaling-law fit (1450) is decent evidence that the mechanism transfers.\n\nWhat it does well: the DNS resolution checks are adequate, the relaminarization threshold is checked for sensitivity, and they ran 1000 cases total. The duct geometry is a genuine gap in the literature, and the comparison to pipe/channel makes the geometric-invariance claim concrete.\n\nSoft spots, in order. First, no error bars on the mean lifetimes. With 50-100 runs per Re and exponential lifetime distributions, the standard error is about 10-14% of the mean. That is not negligible when you're distinguishing a square-root from a super-exponential fit. They need to tabulate the data and at least give standard errors. Second, the pattern preservation approximation is load-bearing, and the evidence for it is visual. The stress-test note is right: at low Re, puffs rarely reach high E_k, so the region around E_kc=0.355 is sparsely sampled. The quintic fits that give Rec=1445 are extrapolations into thin data. The match to 1450 is suggestive, but the authors should report how many trajectories actually contribute to that part of the P_d and D_d curves, and show the prediction is robust to, say, quartic vs. sextic fits. Third, the super-exponential regime is described by a line through what looks like a small number of points; the slope and intercept need uncertainties.\n\nOverall: the empirical scaling laws are likely correct, and the theory is a plausible explanation, not a proof. This is a perfectly good letter for PRL or PoF, provided the authors add error bars and address the extrapolation concern. A serious referee should engage with it, not desk reject. I'd want the missing statistics before citing the scaling exponents, but I'd bring it to reading group for the geometry comparison.","headline":"A solid first DNS study of duct-puff lifetime scalings with a plausible but under-validated noisy saddle-node explanation; deserves review once error bars and extrapolation checks are added.","tokens_in":8682,"tokens_out":3613,"would_cite":true,"duration_ms":41299,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["47.27.Cn"],"model":"deepseek-v4-flash","headline":"Localized turbulent puffs in square ducts are transient structures whose mean lifetime obeys a square-root law below Re≈1450 and a super-exponential law above it.","keywords":["turbulent puff","duct flow","lifetime scaling","saddle-node bifurcation","relaminarization","Reynolds-Orr equation","pattern preservation","transient turbulence"],"falsifier":"Run DNS at Re=1420, 1430, 1440, 1450, 1460, and 1470 with at least 100 independent initial conditions each, and plot both $τ^{-2}$ and ln(ln τ) against Re; the square-root law requires $τ^{-2}$ to be strictly linear in Re below Re_c, and the super-exponential law requires ln(ln τ) to be linear above Re_c. A visible bend or curvature at either end, or a shift of the transition away from 1450 by more than the data scatter, would falsify the claim. Alternatively, compute P_d(E_k) and D_d(E_k) separately for puffs initialized at Re=1500 and Re=1510 and check whether the curves coincide after rescaling by E_k; if they do not, the pattern preservation approximation fails.","tokens_in":7632,"feed_emoji":"🌀","tokens_out":6576,"duration_ms":69848,"temperature":0.7,"pith_summary":"The paper asks whether turbulent puffs in square-duct flow are self-sustained or transient, and what controls their lifetimes. Using direct numerical simulations and the Reynolds-Orr energy equation, it shows that mean puff lifetime follows a square-root law below Re≈1450 and a super-exponential (double-log linear) law above it. The two regimes are explained by reducing the energy dynamics to a noisy saddle-node bifurcation: deterministic critical slowing down below the critical Reynolds number, stochastic barrier crossing above it. The authors conclude that duct puffs never self-sustain in this range and that the decay mechanism is the same as in pipe and channel flows despite the duct's secondary flows.","feed_headline":"Duct puffs are transient: two lifetime laws split at Re 1450","feed_subtitle":"Same decay mechanism as pipe flow: deterministic slowing below Re 1450, stochastic barrier crossing above.","key_machinery":"The load-bearing object is the noisy saddle-node reduction of the Reynolds-Orr equation. With deterministic production and dissipation written as P_d(E_k) and D_d(E_k), the deterministic growth rate σ_d(E_k)=P_d(E_k)-D_d(E_k)/Re is expanded around the critical energy E_kc where dσ_d/dE_k=0. This yields d(E_k-E_kc)/dt ≈ D_d(E_kc)(Re-Re_c)/$Re_c^{2}$ - (A/2)(E_k-E_kc)^2 + σ_s, the standard stochastic saddle-node normal form. The pattern preservation approximation—that P_d and D_d are functions of E_k alone, nearly independent of Re—is what turns a PDE problem into this one-dimensional normal form and allows the two lifetime scalings to be derived.","core_discovery":"The central claim is that localized turbulent puffs in a square duct are transient structures whose mean lifetime τ obeys τ=(a1 Re+b1)^(-1/2) for Re<1440, implying a divergence at Re_c=-b1/a1=1450, and ln[ln(τ)]=a2 Re+b2 for Re>1450, a super-exponential growth. These scalings are reproduced by a stochastic saddle-node normal form obtained from the Reynolds-Orr equation under a pattern preservation approximation in which the ensemble-averaged production P_d and dissipation D_d depend only on the disturbance kinetic energy E_k. The critical Reynolds number predicted from the ratio D_d/P_d, namely 1445, closely matches the lifetime-scaling value 1450. Hence the puff's fate is decided by the same noisy saddle-node mechanism previously found in pipe and channel flows: subcritical puffs decay through critical slowing down, while supercritical puffs are metastable and decay by fluctuation-activated barrier crossing.","pith_inferences":["A direct testable extension: measure the full lifetime distribution, not just the mean. The saddle-node model predicts a specific non-exponential distribution below Re_c and an exponential, memoryless distribution above Re_c; a mismatch would rule out the reduction.","The pattern preservation approximation may break down at higher Re where puff splitting occurs; then the single-variable normal form would need extra dimensions, possibly changing the scaling laws.","One could check whether the same two-regime lifetime scaling appears for rectangular ducts of different aspect ratios, using the aspect ratio as a continuous parameter to see how Re_c shifts and whether the universal form survives.","If the critical energy E_kc≈0.355 corresponds to an edge state, the theory connects directly to the dynamical-systems view of the laminar-turbulent boundary; verifying that the unstable saddle branch is an edge state would link lifetime scaling to the edge manifold."],"forward_implications":["If correct, duct puffs are transient for all Re covered here; there is no self-sustained puff below Re≈1450, only a metastable one above it.","The square-root scaling below Re_c follows from critical slowing down in the ghost of the saddle-node, so lifetime data can be used to locate Re_c independently of the DNS fit.","The super-exponential regime implies that above Re_c the mean lifetime grows extremely fast with Re, so laboratory or DNS experiments must use very long runs to observe decay.","The same mechanism as in pipe and channel flow suggests a universal description of localized turbulence decay in wall-bounded shear flows, despite structural differences like corner vortices.","The predicted Re_c≈1445 from D_d/P_d can be cross-checked against lifetime measurements, providing a direct test of the theory."],"supporting_citations":[{"why":"The pipe-flow model that this work extends to ducts; supplies the noisy saddle-node derivation and the two-regime scaling.","marker":"[24]"},{"why":"Source of the pattern preservation approximation that P_d and D_d are functions of E_k alone, nearly independent of Re.","marker":"[29]"},{"why":"Established the super-exponential lifetime scaling in pipe flow that is reproduced here in ducts.","marker":"[19]"},{"why":"Provided the pipe-flow onset scenario and puff splitting that motivates the metastability interpretation at higher Re.","marker":"[20]"},{"why":"Identified localized turbulent puffs as the characteristic transitional structures in ducts and described their front dynamics.","marker":"[12]"},{"why":"The theory of noisy saddle-node barrier crossing used to explain the super-exponential regime.","marker":"[30]"}],"fun_headline_variants":["Duct puff lifetimes split at Re 1450: two distinct scaling laws","Puff decay laws in ducts mirror pipe flow, Re 1450 critical","Two lifetime scaling laws for duct puffs, critical Re=1450","Puffs in ducts: square-root then super-exponential lifetime growth","Duct puff decay: deterministic below Re 1450, stochastic above"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The pattern preservation approximation—that the ensemble-averaged production and dissipation of disturbance kinetic energy depend only on the current level of disturbance energy and not on Reynolds number or on the puff's history—is the load-bearing premise, because if P_d and D_d vary with Re or with the puff's internal dynamics beyond E_k, the quadratic expansion and the two scaling laws do not follow.","fun_headline_variants_meta":{"raw":{"variants":["Duct puff lifetimes split at Re 1450: two distinct scaling laws","Puff decay laws in ducts mirror pipe flow, Re 1450 critical","Two lifetime scaling laws for duct puffs, critical Re=1450","Puffs in ducts: square-root then super-exponential lifetime growth","Duct puff decay: deterministic below Re 1450, stochastic above"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000769,"raw_usage":{"total_tokens":3409,"prompt_tokens":951,"completion_tokens":2458,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":567,"completion_tokens_details":{"reasoning_tokens":2361}},"tokens_in":567,"tokens_out":2458,"duration_ms":16103,"temperature":1.0,"reasoning_tokens":2361,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T20:47:43.759205+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run DNS at Re=1420, 1430, 1440, 1450, 1460, and 1470 with at least 100 independent initial conditions each, and plot both $τ^{-2}$ and ln(ln τ) against Re; the square-root law requires $τ^{-2}$ to be strictly linear in Re below Re_c, and the super-exponential law requires ln(ln τ) to be linear above Re_c. A visible bend or curvature at either end, or a shift of the transition away from 1450 by more than the data scatter, would falsify the claim. Alternatively, compute P_d(E_k) and D_d(E_k) separately for puffs initialized at Re=1500 and Re=1510 and check whether the curves coincide after rescaling by E_k; if they do not, the pattern preservation approximation fails.","supporting_citations":[{"cited_title":"The Onset of Metastable Turbulence in Pipe Flow","cited_arxiv_id":"2504.14465","evidence_quote":"The pipe-flow model that this work extends to ducts; supplies the noisy saddle-node derivation and the two-regime scaling."},{"cited_title":"Zhang \\ and\\ author J","cited_arxiv_id":null,"evidence_quote":"Source of the pattern preservation approximation that P_d and D_d are functions of E_k alone, nearly independent of Re."},{"cited_title":"Hof , author A","cited_arxiv_id":null,"evidence_quote":"Established the super-exponential lifetime scaling in pipe flow that is reproduced here in ducts."},{"cited_title":"Avila , author D","cited_arxiv_id":null,"evidence_quote":"Provided the pipe-flow onset scenario and puff splitting that motivates the metastability interpretation at higher Re."},{"cited_title":"Barkley , author B","cited_arxiv_id":null,"evidence_quote":"Identified localized turbulent puffs as the characteristic transitional structures in ducts and described their front dynamics."},{"cited_title":"Hathcock \\ and\\ author J","cited_arxiv_id":null,"evidence_quote":"The theory of noisy saddle-node barrier crossing used to explain the super-exponential regime."}],"review_version":1}