{"id":"2c93deb3-bcff-4958-badd-1085623f4a2f","arxiv_id":"2505.05304","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Turbulence-induced intermittency, measured by KL divergence from a Gaussian random field, grows logarithmically with Reynolds number, and dissipation and enstrophy show nearly equal intermittency.","lead":"The paper measures how turbulent flows become intermittent by comparing the probability distributions of energy dissipation and enstrophy with those of a random Gaussian velocity field. It reports that this turbulence-induced intermittency grows logarithmically with Reynolds number and that dissipation and enstrophy acquire nearly the same intermittency.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Key scaling and symmetry claims lack error bars; bootstrap and model-comparison tests are needed to support Eqs. (6)-(7).","rationale":"The paper's Gamma baselines for Gaussian random fields are derived correctly and are independent of the spectrum for isotropic incompressible fields, so the reader's weakest assumption about baseline correctness is less concerning than the empirical support for the headline scalings. The lognormal model provides theoretical backing for logarithmic growth of D(φ), and the derivation of Eq. (4) is internally consistent. However, the two most novel claims—logarithmic growth and ε/Ω symmetry—are fitted to DNS points without uncertainty quantification, and the datasets are heterogeneous. The proposed bootstrap-based test would directly address whether the log fit is genuinely preferred and whether the ε−Ω difference is statistically zero. My concern is specific and actionable, but it does not invalidate the paper; it strengthens the case for a conditional verdict pending replication with error bars. Hence the reader's CONDITIONAL verdict remains appropriate, and no change is needed.","tokens_in":11205,"tokens_out":15309,"duration_ms":156848,"concrete_test":"At each Reλ, resample the velocity fields using block bootstrap over spatial subdomains (or bootstrap over independent snapshots) to obtain N=1000 estimates of D(φ), D(ε), and D(Ω). First, fit both D = a + b log Reλ and D = c Reλ^p using inverse-variance weights from the bootstrap; compare the fits via AIC and an F-test. Second, for each Reλ > 10, compute the bootstrap distribution of ΔD = D(ε) − D(Ω) and test H0: ΔD = 0 at the 2σ level. If the log fit is not significantly preferred over the power-law fit, or if any ΔD is significantly nonzero, the headline claims are not supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claims rest on KL divergences computed from DNS data at 14 Reynolds numbers, yet no error bars, confidence intervals, or statistical tests are reported. Equation (6) is a two-parameter log fit over Reλ ∈ (10, 600), and Eq. (7) is a single log fit asserted to describe both ε and Ω. The ε/Ω symmetry is inferred from visual overlap in Fig. 5, with no test of whether D(ε) and D(Ω) are statistically distinguishable. Because the datasets come from different solvers and forcing schemes, systematic differences between datasets could produce spurious trends or camouflage real differences. Without quantifying sampling and binning uncertainty, the logarithmic versus power-law distinction and the equality of ε and Ω intermittency are not established.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes an information-theoretic characterization of small-scale turbulence intermittency by measuring the Kullback-Leibler divergence of pseudodissipation, dissipation, and enstrophy from analytically derived Gamma-distribution baselines of a Gaussian random velocity field. Using DNS data of forced homogeneous isotropic turbulence over Re_λ ≈ 1–588, the authors report that turbulence-induced intermittency grows logarithmically with Reynolds number (Eqs. 6–7), that dissipation and enstrophy exhibit nearly equal turbulence-induced intermittency, and that the Shannon entropy of pseudodissipation is non-monotonic, with two critical Reynolds numbers near 10 and 100. The paper argues that this logarithmic growth contrasts with the commonly invoked power-law scaling of moments and that the strain/vorticity asymmetry is largely kinematic, not dynamic.","tokens_in":11361,"tokens_out":5015,"duration_ms":57432,"significance":"If the central claims hold, the paper offers a genuinely useful new diagnostic: KL divergence relative to a Gaussian random field separates kinematic intermittency from dynamics-induced intermittency in a single scalar measure, and the analytic Gamma baseline is a clean, parameter-free construction. The reported logarithmic growth and the ε–Ω symmetry would be notable contributions to the intermittency literature. However, the quantitative support for these claims is currently limited: the scaling laws and the symmetry rest on fits to ensemble DNS data without uncertainty estimates, statistical model comparison, or tests for dataset heterogeneity. The strength of the contribution is therefore conditional on the completion of that statistical work.","major_comments":[{"comment":"The central quantitative claims—logarithmic growth of D_f||f_GRF with Re_λ and the equality of the ε and Ω trends—are based on least-squares fits over Re_λ ∈ (10, 600) with no reported confidence intervals, residuals, or goodness-of-fit statistics. Because the DNS points come from at least three different solvers and forcing schemes (in-house, JHTDB, and Texas A&M), systematic inter-dataset differences could easily corrupt the apparent slopes or the apparent overlap. The authors should provide bootstrap or subsampling estimates of the KL values and their uncertainties, and preferably fit both log and power-law models to the same data points to support the log-law preference.","section":"Intermittency, Eqs. (6) and (7), Figs. 4 and 5"},{"comment":"The claim that dissipation and enstrophy have 'nearly identical' turbulence-induced intermittency is inferred from visual overlap of the two data series. No statistical test is reported for the difference D(ε) − D(Ω) across Re_λ. Since the baseline Gamma distributions have different shape parameters (n = 5 vs n = 3), the near-identity of the KL values is not automatically expected and is a headline result; it therefore requires a quantitative comparison, such as a paired bootstrap on the per-Re_λ differences or a confidence interval for the fitted slope difference.","section":"Intermittency, Eq. (7) and Fig. 5"},{"comment":"The two critical Reynolds numbers, Re_λ^(1) ≈ 10 and Re_λ^(2) ≈ 100, are read from the entropy curve without a stated quantitative criterion (e.g., a change-point test or a threshold on dH/d log Re_λ). Moreover, the absolute entropy H(φ) depends on the histogram bin width through the −log(Δx) term in Eq. (1). The text states that Δx = 0.05 was selected from a range where KL divergence remains approximately constant, but no convergence data or sensitivity analysis for the entropy itself is shown; the location of the maximum could be affected by binning and by finite-sample tail resolution.","section":"Uncertainty, Fig. 3 and Eq. (1)"},{"comment":"The interpretation of D_f||f_GRF as purely 'turbulence-induced' intermittency depends on the assumption that a zero-mean, divergence-free, spatially uncorrelated Gaussian random field is the correct Reynolds-number-independent null for all Re_λ, including the low-Re_λ range 1–10 where the flow is not developed turbulence. This is a reasonable first choice, but it is an assumption about the null, not a proven separation of kinematic and dynamic effects. The authors should at least discuss whether a GRF with the same one-point energy spectrum or a finite-Re_λ Gaussian ensemble would change the conclusion, and should verify that the low-Re_λ 'D ≈ 0' plateau is not an artifact of the baseline being mismatched to the actual low-Re_λ state.","section":"Baseline Distribution, Eq. (3)"}],"minor_comments":[{"comment":"The sentence 'despite continuously growing variability, entropy decays above a certain Reynolds number' is potentially confusing: the 'variability' referred to is presumably the variance of the PDF, while entropy is a different measure. The distinction between variance growth and entropy decay should be stated explicitly in the figure caption or text.","section":"Uncertainty, Fig. 3"},{"comment":"The figure caption does not report the sample size or the number of independent snapshots used for each PDF estimate. Since KL divergence is sensitive to poorly sampled tails, a brief statement of sample sizes and the range of Δx over which the estimate is stable would improve reproducibility.","section":"PDF estimation, Fig. 2"},{"comment":"The claim that KL divergence is 'more robust to sampling errors' than moment-based estimators is cited to references on change detection and soil mapping; a turbulence-specific convergence study or a direct argument would be more appropriate here.","section":"Introduction, refs. [26, 31–33]"}],"recommendation":"major_revision","confidential_remarks":"The paper is well within the scope of the journal and the analytic-baseline idea is attractive, but the headline scaling and symmetry claims currently lack the uncertainty quantification that a quantitative physics letter requires. I would not reject; I would ask the authors for error bars, model-comparison tests, and a discussion of dataset heterogeneity. If the statistical support remains absent, the claims should be downgraded to observations rather than fitted laws."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a real step forward in measuring small-scale intermittency, but the two headline results—the logarithmic growth law and the dissipation/enstrophy symmetry—are less secure than the text suggests. The missing error bars and the mixed DNS data keep me from endorsing the scalings as they stand.\n\nWhat is actually new: Granero-Belinchón et al. used KL divergence for velocity increments in the inertial range. Here the same machinery is moved to dissipation, enstrophy, and pseudodissipation, with the Gamma baselines derived analytically from a Gaussian random field. That allows the authors to separate kinematic intermittency from turbulence-induced intermittency, which is a clean and useful decomposition. The claim that enstrophy's traditionally heavier tails are mostly kinematic, not dynamic, is an interesting revisionist view that fits with existing work by Gotoh et al. and Tsinober. The lognormal fit provides an independent check on the KL scaling and is a nice touch. Shannon entropy non-monotonicity with a peak near Re_lambda ~ 100 is worth taking seriously because it lines up with other observed convergences of small-scale statistics.\n\nWhere it is soft: the quantitative claims rest on least-squares fits with no error bars, no bootstrap, no convergence checks. Equations (6) and (7) are fits over Re_lambda in (10,600); the low end is sparse and the high end comes from another group's data. The epsilon/Omega symmetry is asserted from overlapped curves in Fig. 5 rather than a statistical test. The bin width dependence of the entropy is mentioned but not documented. The baseline Gamma distribution is derived for a spatially uncorrelated GRF; calling it the 'non-intermittent state' at all Reynolds numbers is reasonable but not proven. None of these are fatal, but they are exactly the things a referee should push on.\n\nCitation pattern is fine. The self-citations to Das and Girimaji are for the in-house solver and prior velocity-gradient analysis, which is appropriate.\n\nWho it's for: turbulence folks interested in intermittency metrics and the kinematic/dynamic decomposition. It deserves a serious referee, not a desk reject. If I were handling it, I'd ask for uncertainty quantification on the KL estimates, a model-comparison test of log vs power-law, a statistical test of the epsilon/Omega equality, and ideally data/code release.","headline":"A genuinely useful kinematic/dynamic decomposition of intermittency, but the headline scalings and symmetry claim need error bars and statistical tests before I'd trust them.","tokens_in":11869,"tokens_out":2499,"would_cite":false,"duration_ms":26217,"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":"Using KL divergence from a Gaussian random field, the paper finds turbulence-induced intermittency grows logarithmically with Reynolds number and is equal for dissipation and enstrophy.","keywords":["turbulence intermittency","Kullback-Leibler divergence","Shannon entropy","Gaussian random field","direct numerical simulation","Reynolds number scaling","dissipation","enstrophy"],"falsifier":"A direct check is to evaluate the KL divergence of dissipation PDFs from the Gamma baseline in a flow whose velocity gradients are Gaussian but with no turbulent dynamics, and to extend DNS of $D_{f\\|f_{\\mathrm{GRF}}}$ beyond $Re_\\lambda=1000$; the central claim fails if the divergence is nonzero without dynamics or if the growth becomes power-law rather than logarithmic.","tokens_in":10992,"feed_emoji":"🌀","tokens_out":11142,"duration_ms":94364,"temperature":0.7,"pith_summary":"The paper claims that small-scale intermittency in turbulence, when measured as the Kullback-Leibler divergence of dissipation, pseudodissipation, or enstrophy from a Gaussian random velocity field, grows logarithmically with the Taylor Reynolds number rather than as the power law reported for individual moments. This means the growth rate of turbulence-induced intermittency slows at higher Reynolds numbers, possibly saturating. The paper also claims that turbulence dynamics generate nearly equal intermittency in dissipation and enstrophy, and that the widely reported difference between them is a kinematic effect already present in a Gaussian random field. Finally, the Shannon entropy of pseudodissipation is non-monotonic, peaking near $Re_\\lambda \\approx 100$, a Reynolds number the authors associate with fully developed small-scale turbulence. These findings matter because they change the predicted high-Reynolds-number behavior of intermittency and reframe a long-standing strain-versus-vorticity asymmetry.","feed_headline":"Turbulence intermittency grows logarithmically, not by power law","feed_subtitle":"A KL-divergence measure from Gaussian random fields also finds dissipation and enstrophy equally intermittent.","key_machinery":"The load-bearing object is the Gamma-distribution baseline PDF for a Gaussian random velocity field from Eq. (3): $f_{\\mathrm{GRF}}(x) = \\frac{(n/2)^{n/2}}{\\Gamma(n/2)} x^{n/2-1} e^{-nx/2}$, with $n=8$ for pseudodissipation, $n=5$ for dissipation, and $n=3$ for enstrophy. Because mean-normalized quadratic forms of Gaussian gradients follow Gamma distributions, this baseline encodes the kinematic intermittency inherent in the definitions of $\\phi$, $\\epsilon$, and $\\Omega$; any deviation from it is attributed to turbulence dynamics. KL divergence (Eq. 2) measures that deviation, and the lognormal model for $\\phi$ (Eq. 5) provides an analytic approximation whose KL divergence scales as $2\\sigma_\\theta^2 - \\tfrac{1}{2}\\log\\sigma_\\theta^2$, explaining the logarithmic growth.","core_discovery":"The central discovery is that KL divergence from a Gamma-distributed Gaussian random field baseline gives a clean separation between kinematic and dynamic intermittency, and that the dynamic part obeys simple log laws. For pseudodissipation, $D_{f\\|f_{\\mathrm{GRF}}}(\\phi) \\approx -0.925 + 0.376 \\log Re_\\lambda$ for $10 < Re_\\lambda < 600$, while dissipation and enstrophy both follow $D_{f\\|f_{\\mathrm{GRF}}}(X) \\approx -0.459 + 0.173 \\log Re_\\lambda$. The Shannon entropy of pseudodissipation increases below $Re_\\lambda\\approx100$ and decreases above it, producing a maximum near $Re_\\lambda \\approx 100$. The authors interpret two critical Reynolds numbers: $Re_\\lambda \\approx 10$, where turbulence-induced intermittency begins, and $Re_\\lambda \\approx 100$, where small-scale statistics become fully developed and the uncertainty of the field starts to decline. These results contradict the usual power-law scaling of moments and the common inference that enstrophy is more intermittent than dissipation.","pith_inferences":["A direct extension would test whether the logarithmic growth saturates by pushing DNS to $Re_\\lambda \\gtrsim 10^4$; if the divergence plateaus, the hypothesis of finite maximum intermittency would be supported.","The $\\epsilon$-$\\Omega$ symmetry may also appear in the joint statistics of strain-rate and rotation-rate invariants, which could be checked in existing high-$Re$ DNS databases.","If the two critical Reynolds numbers are universal, the same entropy and KL analysis applied to shear flows, boundary layers, or convective turbulence would show the same $Re_\\lambda\\approx10$ and $Re_\\lambda\\approx100$ markers."],"forward_implications":["At higher Reynolds numbers, intermittency grows more slowly than individual moment scaling suggests, and if the log trend continues, the KL divergence may approach a finite asymptotic value.","The near-equal intermittency of dissipation and enstrophy implies that small-scale models should treat strain and rotation dynamics as equally intermittent rather than encoding a built-in asymmetry.","The entropy maximum at $Re_\\lambda\\approx100$ provides a quantitative marker for the onset of fully developed small-scale turbulence, coinciding with the dissipative anomaly and velocity-gradient partitioning.","The Reynolds-number-independent Gamma baseline offers a single reference PDF, so intermittency can be compared across different flows and Reynolds numbers without re-fitting the null.","The lognormal-derived approximation reproduces the log growth, tying the scaling to the established logarithmic growth of the log-gradient variance $\\sigma_\\theta^2$."],"supporting_citations":[{"why":"Derives the Gamma-distribution baseline PDFs for pseudodissipation, dissipation, and enstrophy in a Gaussian random velocity field; Eq. (3) relies on it.","marker":"[7]"},{"why":"Introduces KL divergence as an intermittency measure based on the full PDF, the core metric the paper applies.","marker":"[26]"},{"why":"Provides DNS datasets and the power-law moment scaling and $Re_\\lambda\\approx10$ transition that the logarithmic growth is contrasted with.","marker":"[12]"},{"why":"Extends the anomalous-exponent framework whose power-law scaling the paper's log law directly challenges.","marker":"[13]"},{"why":"Documents scaling of dissipation and enstrophy moments, supplying the prior expectation that enstrophy is more intermittent.","marker":"[18]"},{"why":"Gives Reynolds-number-dependent enstrophy and dissipation moment scaling used to frame the $\\epsilon$-$\\Omega$ symmetry result.","marker":"[25]"},{"why":"Shows kinematic effects on dissipation and enstrophy PDFs, supporting the Gamma baseline as the correct non-intermittent reference.","marker":"[42]"},{"why":"Supplies the public turbulence database DNS fields used for the $Re_\\lambda$ sweep.","marker":"[49]"},{"why":"Reports the logarithmic growth of log-gradient variance $\\sigma_\\theta^2$ used in the lognormal KL fit.","marker":"[53]"},{"why":"Establishes the near-lognormal distribution of dissipation that justifies the lognormal model for $\\phi$.","marker":"[54]"}],"fun_headline_variants":["Turbulence intermittency scales logarithmically, not as power law","Equal intermittency of dissipation and enstrophy revealed by KL divergence","Logarithmic growth of intermittency challenges power-law scaling","Non-monotonic entropy marks a shift in small-scale statistics"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The Gamma distributions of Eq. (3) are assumed to be the exact and Reynolds-number-independent representation of the non-intermittent state, so if the kinematic baseline changes with $Re_\\lambda$, the measured KL divergences would not isolate turbulence dynamics alone.","fun_headline_variants_meta":{"raw":{"variants":["Turbulence intermittency scales logarithmically, not as power law","Equal intermittency of dissipation and enstrophy revealed by KL divergence","Logarithmic growth of intermittency challenges power-law scaling","Non-monotonic entropy marks a shift in small-scale statistics"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000441,"raw_usage":{"total_tokens":2252,"prompt_tokens":981,"completion_tokens":1271,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":597,"completion_tokens_details":{"reasoning_tokens":1199}},"tokens_in":597,"tokens_out":1271,"duration_ms":8620,"temperature":1.0,"reasoning_tokens":1199,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T23:07:48.042506+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct check is to evaluate the KL divergence of dissipation PDFs from the Gamma baseline in a flow whose velocity gradients are Gaussian but with no turbulent dynamics, and to extend DNS of $D_{f\\|f_{\\mathrm{GRF}}}$ beyond $Re_\\lambda=1000$; the central claim fails if the divergence is nonzero without dynamics or if the growth becomes power-law rather than logarithmic.","supporting_citations":[{"cited_title":"Gotoh and J","cited_arxiv_id":null,"evidence_quote":"Derives the Gamma-distribution baseline PDFs for pseudodissipation, dissipation, and enstrophy in a Gaussian random velocity field; Eq. (3) relies on it."},{"cited_title":"Buaria and A","cited_arxiv_id":null,"evidence_quote":"Introduces KL divergence as an intermittency measure based on the full PDF, the core metric the paper applies."},{"cited_title":"Yakhot and D","cited_arxiv_id":null,"evidence_quote":"Provides DNS datasets and the power-law moment scaling and $Re_\\lambda\\approx10$ transition that the logarithmic growth is contrasted with."},{"cited_title":"Yakhot and D","cited_arxiv_id":null,"evidence_quote":"Extends the anomalous-exponent framework whose power-law scaling the paper's log law directly challenges."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Documents scaling of dissipation and enstrophy moments, supplying the prior expectation that enstrophy is more intermittent."},{"cited_title":"Elsinga, T","cited_arxiv_id":null,"evidence_quote":"Gives Reynolds-number-dependent enstrophy and dissipation moment scaling used to frame the $\\epsilon$-$\\Omega$ symmetry result."},{"cited_title":"Buaria and K","cited_arxiv_id":null,"evidence_quote":"Shows kinematic effects on dissipation and enstrophy PDFs, supporting the Gamma baseline as the correct non-intermittent reference."},{"cited_title":"Pekurovsky, P3dfft: A framework for parallel compu- tations of fourier transforms in three dimensions, SIAM Journal on Scientific Computing 34, C192 (2012)","cited_arxiv_id":null,"evidence_quote":"Supplies the public turbulence database DNS fields used for the $Re_\\lambda$ sweep."},{"cited_title":"Das and S","cited_arxiv_id":null,"evidence_quote":"Reports the logarithmic growth of log-gradient variance $\\sigma_\\theta^2$ used in the lognormal KL fit."},{"cited_title":"Lin and H","cited_arxiv_id":null,"evidence_quote":"Establishes the near-lognormal distribution of dissipation that justifies the lognormal model for $\\phi$."}],"review_version":1}