{"id":"70e9acd1-34b8-42f2-840f-1801f5dd4149","arxiv_id":"1908.02397","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"The authors extend droplet breakup theory to the viscous range of turbulence and couple it with a 15-bin population balance model in LES, validating the relative size distribution against published oil jet experiments.","lead":"This paper develops a large-eddy simulation model that tracks how oil droplets of different sizes break up in turbulent water, using a breakup rate that works for both large and small eddies. The model reproduces measured droplet size distributions in an oil jet in crossflow, which matters for predicting oil spill behavior.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Subgrid closure of the breakup source (§4.1) is the weakest link: with grid scale ≈5.2 mm and η≈13–60 µm, breakup is driven by unresolved eddies, and the assumed closure ~g n ≈ g~ ñ is untested. A resolution/sensitivity study would determine whether the claimed agreement is robust.","rationale":"The central claim is that the LES reproduces the measured relative droplet size distribution and quantifies its variability. The paper is strongest in its transparent, parameter-light breakup kernel (one fitted K*) and in the one-dimensional tests against Li et al. and Eastwood et al., which give some independent support. The LES comparison, however, rests on an untested closure: since Δ≈5.2 mm and η≈13–60 µm in the simulated jet, the eddies that actually break the droplets are subgrid, and the model replaces the filtered breakup source with g evaluated at the filtered dissipation. This is exactly the limitation the authors flag in §4.1. Nonlinearity of g in ε makes the closure error first-order, not a small correction. Figures 17 show ε varying by orders of magnitude at the grid scale, so the Jensen bias can be large. The 1D calibration cannot detect this because it uses experimental mean dissipation, not a filtered field. The validation against Murphy et al. is therefore the only exercise of the closure, and it is weakened further by the reported 1.4–3.7× discrepancies in total oil concentration (which the authors normalize away) and by the K* non-transferability seen in the Eastwood data. None of these issues is fatal by itself; each is addressable. A grid-refinement study, even at one measurement station, would show whether the relative distribution is sensitive to resolution and SGS modeling. If it is not, the central claim is supported; if it is, the claim should remain conditional on a better SGS closure. Thus the reader's conditionality is the right verdict, and our concern aligns with the reader's weakest assumption.","tokens_in":33404,"tokens_out":17698,"duration_ms":189606,"concrete_test":"Re-run the jet-in-crossflow LES at doubled resolution (Δ≈2.6 mm; e.g., 768×300×768 grid, or a locally refined region around the jet near the measurement stations), keeping flow rate, injection condition, K*=0.2 and all other settings fixed. At x=0.76 m and x=1.3 m, compare the relative size distribution N*_i and the d32 statistics with the current Δ=5.2 mm results (Figs. 11–13). If the resolution-induced differences exceed the scatter between the three experimental realizations or the bin-to-bin variation in N*_i, the SGS closure is load-bearing and the claimed agreement is not yet robust; if the differences are small, the closure is adequate for this application.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"In §4.1 the filtered breakup source is closed by assuming ~g n ≈ g~ n and g(ε)≈g(ε~), i.e. subgrid correlations between dissipation and droplet concentration are neglected. This is not a peripheral detail: in the jet-in-crossflow LES the effective resolution is Δ≈5.2 mm while the Kolmogorov scale is η≈13 µm near the nozzle and ≈60 µm downstream (§4.3), so η/Δ≈0.0025–0.012 and virtually all droplet bins (20–1000 µm) are at scales whose breakup is set by subgrid eddies. The paper itself warns that this closure 'must be kept in mind, especially for applications at very high Reynolds numbers when the Kolmogorov scale is much smaller than the grid scale.' Because g depends nonlinearly on ε through Eqs. (2.17)–(2.18) — roughly ε^{1/3} to ε^{1/2} in the inertial and viscous limits, with an additional exponential efficiency factor — Jensen-type bias means ⟨g(ε)n⟩ ≠ g(⟨ε⟩)⟨n⟩, and Fig. 17 shows grid-scale ε fluctuating over many orders of magnitude. The 1D calibration (§3.1) and Eastwood comparison (§3.2) use measured or parameterized mean dissipation and therefore do not exercise this closure; the LES comparison is the only test, and no grid or SGS-model sensitivity is reported. If the bias is significant, the local breakup rates — and hence the relative size distributions in Figs. 11–13 — are not robust predictions.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper develops a population balance model for polydisperse droplet evolution in an LES framework. The breakup frequency is modeled as a droplet–eddy collision integral with a Batchelor structure function that smoothly bridges inertial and viscous ranges, and is parameterized by algebraic fits. The one free constant K* is calibrated against the Li et al. (2017) breaking-wave experiment, tested against Eastwood et al. (2004) jet data, and used in a 3D LES of a jet in crossflow. The LES predictions are compared with Murphy et al. (2016) data using volume-normalized relative size/volume distributions. The paper also reports intermittency statistics of Sauter mean diameter, total surface area, and breakup source terms.","tokens_in":33843,"tokens_out":5306,"duration_ms":57151,"significance":"The viscous-range extension via a structure function is a useful physical improvement over standard inertial-range kernels, and the algebraic fits make LES application affordable. The paper is honest about limitations: it notes the subgrid-closure assumption, the limited statistical convergence of Murphy et al., and tests sensitivity to the injection distribution and daughter-probability model. If the remaining validation issues are resolved, the model would be a practical tool for oil-spill and spray applications, where polydisperse droplet evolution matters. The current evidence supports the relative-shape predictions in the specific Murphy et al. configuration, but not yet a generally transferable quantitative prediction.","major_comments":[{"comment":"The closure \\widetilde{g n}\\approx\\tilde{g}\\tilde{n} and g(\\epsilon)\\approx g(\\tilde{\\epsilon}) is load-bearing, since the grid scale is 5.2 mm while \\eta\\approx 13–60 \\mu m (§4.3), so breakup rates are set by unresolved eddies. The breakup frequency depends nonlinearly on \\epsilon via Eqs. (2.17)–(2.18), and Fig. 17 shows grid-scale \\epsilon varying over orders of magnitude, so Jensen-type bias is expected. The 1D calibrations (§3.1, §3.2) use mean or parameterized dissipation and do not exercise this closure. Because no grid-resolution or SGS-model sensitivity is reported, the LES comparisons in Figs. 11–13 cannot yet be regarded as a clean test of the model; the authors should quantify the closure error or at least demonstrate resolution insensitivity.","section":"§4.1, Eq. (4.3)"},{"comment":"The claim that K*=0.2 is robust is weakened by the Eastwood comparison: for heptane and 10 cSt silicone oil the model with K*=0.2 overpredicts decay, and K*=0.1 and 0.15 are needed. Since no parameter dependence is identified, the fitted constant is not shown to be transferable across fluids. Please either explain why these discrepancies are within model uncertainty or restrict the transferability claim and assess the impact on the Murphy comparison in §4.","section":"§3.2, Fig. 6"},{"comment":"The total oil concentrations differ between simulation and experiment by factors of 1.4 and 3.7, and the comparisons are normalized by total volume concentration. Therefore the abstract's claim of 'good agreement for the relative size distribution' is accurate only for the shape, not the absolute concentration. The model also predicts absolute concentration, so this discrepancy is a substantive part of the validation and should be discussed quantitatively (e.g., statistical convergence of Murphy et al., measurement volume, effective injection in Appendix B) rather than only used to justify a shape-only comparison.","section":"§4.3, Figs. 11–13"}],"minor_comments":[{"comment":"The caption states the comparison is 'at x = 1.3 m', but the text describes the total size distribution over the window x = 0.76–1.66 m; please align the caption with the text.","section":"Figure 13 caption"},{"comment":"Using the symbol e both as Euler's number and as the coefficient e(y) in the fit is confusing; consider renaming the coefficient (for example, f or h).","section":"Eqs. (2.20)–(2.21)"},{"comment":"The error measure E uses the squared logarithmic difference but does not state whether absolute values or signed differences are intended; please clarify.","section":"Eq. (3.4)"},{"comment":"The notation log10[-c(y)] = ... is unusual because c(y) appears in the fit G = a x^b + c x^d - e; please clarify the signs and the domain of validity of the fit.","section":"Appendix A"},{"comment":"The word 'Acknowledegements' is misspelled; it should be 'Acknowledgements'.","section":"Acknowledgements heading"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within the scope of JFM; the main concern is not novelty but whether the validation supports the transferability claim. I would encourage the authors to add a resolution/SGS sensitivity study or a quantitative subgrid-closure estimate, and to soften the transferability claim. The paper may be acceptable after major revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Three things you should know. The genuinely new piece is the Batchelor-blending structure function for the breakup kernel, which extends droplet-eddy collision models to viscous-range eddies in physical space, much simpler than Solsvik and Jakobsen's spectral expressions. Coupling a 15-bin Eulerian population balance with an LES of an oil jet in crossflow and comparing relative size distributions to Murphy et al. is a legitimate new combination. The paper is transparent about its machinery.\n\nThe 1D calibration against Li et al. is done properly: they minimize a log-error over K*, find K*=0.2 as the best fit, and test sensitivity to daughter-size distribution and to the upper cutoff of the eddy integral. The Eastwood comparison is honestly reported: K*=0.2 over-predicts decay for heptane and 10 cSt silicone oil, and they say so before proceeding with 0.2 because it works for the other two oils. The LES setup is careful about matching the jet's centerline decay and spreading with a body-force injection, and they check robustness to the initial injection distribution.\n\nSoft spots, in proportion. The biggest is the SGS closure for the breakup source. In section 4.1 they explicitly assume that the filtered source equals the product of filtered breakup rate and filtered concentration, and they warn it must be kept in mind at high Reynolds numbers. But the numbers make that warning central: grid scale is about 5.2 mm, Kolmogorov scale is about 13 um near the nozzle and 60 um downstream, so essentially all droplet bins are in the viscous range and breakup is set by unresolved eddies. Since the breakup frequency depends nonlinearly on dissipation, the closure is biased, and the paper itself shows epsilon fluctuating over many orders of magnitude. No grid-refinement or SGS-model sensitivity is reported. That is the weakest link.\n\nSecond, the Murphy et al. comparison is shape-only. They normalize both distributions by total volume concentration and report that total oil concentration differs by factors of 1.4 and 3.7. The word relative is load-bearing. Third, K* is not universal: the Eastwood data show a factor-of-two spread depending on the oil, so the model is not yet predictive without per-fluid calibration. Appendix A fits are curve fits to the model's own integral, which is a computational convenience rather than a flaw.\n\nBottom line: for multiphase flow modelers working on oil spills or sprays, this is a useful, citable paper that deserves a serious referee. It does not prove the SGS closure is adequate, and K* transferability is unresolved, but it ships open data and states its limitations. I would send it to review.","headline":"The viscous-range kernel extension and LES-PBE application are genuinely new and useful; the validation is shape-only and the SGS closure is untested, but the paper deserves review.","tokens_in":34306,"tokens_out":4975,"would_cite":true,"duration_ms":48649,"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":"A breakup model extends droplet-size prediction into the viscous range and matches jet-in-crossflow data.","keywords":["droplet breakup","population balance equation","large eddy simulation","polydisperse droplets","viscous range","Batchelor structure function","jet in crossflow","oil spill modeling"],"falsifier":"Run the same LES breakup model at a much finer grid resolution (or in a DNS-resolved setting) for the identical jet-in-crossflow case and compare the predicted size distributions; if the coarse-grid and fine-grid results diverge by more than the reported experimental scatter, the subgrid closure and the fitted K* are not transferable across resolutions.","tokens_in":33222,"feed_emoji":"💧","tokens_out":1272,"duration_ms":16617,"temperature":0.7,"pith_summary":"This paper develops a population balance model for polydisperse droplet evolution that is built to run inside large eddy simulations, and tests it against lab experiments of oil droplets in turbulence. The central claim is that a breakup kernel based on droplet-eddy collisions, extended from the inertial to the viscous range with a Batchelor structure function and one fitted constant, can predict the relative droplet size distribution of a turbulent jet in crossflow. If true, the model gives engineers and oceanographers a practical way to compute how turbulent transport and breakup reshape oil droplet distributions, including the intermittent variability that mean-flow models miss.","feed_headline":"One breakup model spans both turbulent scales and fits jet oil data","feed_subtitle":"A viscous-range extension of droplet-eddy collisions predicts size distributions and their variability in a jet in crossflow.","key_machinery":"The central object is the breakup frequency integral g(di) = K ∫ π/4(di+de)^2 u_e(de) Ω(di,de) dn_e(de), evaluated with droplet-eddy collisions treated kinetically. The key extension is replacing the inertial-range eddy velocity with the Batchelor structure function S2(r) = C2 $ε^{{2/3}}$ $r^{{2/3}}$ [1 + (r/(γ2 η))^{-2}]^{-2/3}, which transitions smoothly to the viscous range; the parameter γ2 = (15 C2)^{3/4} ≈ 13 sets the crossover scale. A fast analytic fit for the integral makes the breakup rate cheap enough to evaluate at every LES grid point and timestep.","core_discovery":"The paper claims that turbulent breakup of droplets in the viscous range, where the Kolmogorov scale is larger than the droplet, can be captured with a single breakup-frequency model that smoothly blends inertial and viscous scalings. Using a Batchelor blending function for the eddy velocity, the breakup frequency becomes a function of a local Reynolds number, Ohnesorge number, and density-viscosity ratio, with one fitted constant K* = 0.2 tuned on a breaking-wave experiment. Applied to a three-dimensional LES of a jet in crossflow with 15 droplet size bins, the model reproduces the measured relative size distribution and reveals that the Sauter mean diameter and total surface area fluctuate strongly and non-Gaussianly in time and space.","pith_inferences":["If the subgrid closure holds, the same kernel could be applied to bubble plumes and spray atomization downstream of primary breakup, where coalescence remains negligible.","The fitted K* may absorb the uncertainty in the upper integration cutoff and daughter-size probability; a systematic study varying both could separate physical breakup physics from closure tuning.","The strong non-Gaussian variability of d32 and total surface area implies that single-point mean predictions of oil fate may understate the risk of locally high surface area available for biodegradation."],"forward_implications":["The LES model can predict polydisperse droplet transport and breakup in flows where the Kolmogorov scale exceeds the droplet size, which earlier inertial-range-only models overestimate.","The fitted constant K* = 0.2 transfers from a breaking-wave experiment to a round jet and then to a jet in crossflow, suggesting the breakup kernel is flow-independent to first order.","LES outputs yield temporal and spatial statistics of droplet surface area and Sauter mean diameter, quantities that control biodegradation rates and optical properties of oil plumes.","The model's insensitivity to the assumed initial droplet injection size (monodisperse vs bidisperse) suggests far-field size distributions are set by breakup dynamics rather than nozzle details."],"supporting_citations":[{"why":"Provides the breaking-wave oil droplet size distribution data used to fit K* = 0.2 and the time-dependent diffusion and dissipation model used in the 1D calibration.","marker":"Li et al. (2017)"},{"why":"Provides the crude oil jet-in-crossflow experimental size distribution data against which the 3D LES results are compared.","marker":"Murphy et al. (2016)"},{"why":"Provides the axisymmetric turbulent jet oil droplet breakup data used to test the model's transferability to different flow conditions and oils.","marker":"Eastwood et al. (2004)"},{"why":"Supplies the phenomenological daughter-droplet probability model and the collision-based breakup frequency framework extended here.","marker":"Tsouris & Tavlarides (1994)"},{"why":"Supplies the blending-function form of the second-order structure function that connects inertial and viscous range eddy velocities.","marker":"Batchelor (1951)"},{"why":"Provides the breakup source term formulation and resistive-energy modeling for viscous droplets that the present kernel builds upon.","marker":"Zhao et al. (2014b)"}],"fun_headline_variants":["LES breakup model bridges inertial and viscous scales","Single model predicts oil droplet sizes in jet flow","Droplet breakup model captures non-Gaussian variability","Viscous-aware breakup kernel fits jet oil data"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The paper assumes that subgrid fluctuations of dissipation rate and droplet concentration are uncorrelated, so the filtered breakup source can be computed from filtered quantities alone; if these correlations matter at high Reynolds numbers, the predicted breakup rates will be biased.","fun_headline_variants_meta":{"raw":{"variants":["LES breakup model bridges inertial and viscous scales","Single model predicts oil droplet sizes in jet flow","Droplet breakup model captures non-Gaussian variability","Viscous-aware breakup kernel fits jet oil data"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00058,"raw_usage":{"total_tokens":2753,"prompt_tokens":986,"completion_tokens":1767,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":602,"completion_tokens_details":{"reasoning_tokens":1707}},"tokens_in":602,"tokens_out":1767,"duration_ms":13289,"temperature":1.0,"reasoning_tokens":1707,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:45:24.302923+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same LES breakup model at a much finer grid resolution (or in a DNS-resolved setting) for the identical jet-in-crossflow case and compare the predicted size distributions; if the coarse-grid and fine-grid results diverge by more than the reported experimental scatter, the subgrid closure and the fitted K* are not transferable across resolutions.","supporting_citations":[{"cited_title":", Xue, X","cited_arxiv_id":null,"evidence_quote":"Provides the crude oil jet-in-crossflow experimental size distribution data against which the 3D LES results are compared."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the axisymmetric turbulent jet oil droplet breakup data used to test the model's transferability to different flow conditions and oils."},{"cited_title":"& Tavlarides, L","cited_arxiv_id":null,"evidence_quote":"Supplies the phenomenological daughter-droplet probability model and the collision-based breakup frequency framework extended here."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the blending-function form of the second-order structure function that connects inertial and viscous range eddy velocities."}],"review_version":1}