{"id":"89555b9c-08be-4888-9f8c-bdd4be7caa5e","arxiv_id":"2505.12205","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A GPU-accelerated phase-field lattice Boltzmann simulation places the liquid jet dripping-to-jetting transition at We_cr ≈ 2.2, narrowing the experimental range 2 < We < 3, and reproduces measured breakup lengths and drop sizes.","lead":"A CUDA-accelerated lattice Boltzmann framework called CLIP was used to simulate liquid jet breakup, finding the dripping-to-jetting transition at a critical Weber number of about 2.2, in line with experiments that had placed it between 2 and 3. The framework is designed to run high-density-ratio two-phase flow simulations on a standard desktop GPU, which could make multiphase simulation more accessible without HPC clusters.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Claimed We_cr≈2.2 rests on two bracketing simulations with an idealized uniform inlet and no resolution or disturbance sensitivity study; the paper's own caveat (Sec. 6.3) concedes the boundary may shift, so the central quantitative result is not yet supported.","rationale":"The strongest claim of the paper is quantitative: the framework 'closely matches experimental observations' of breakup length and drop size and pins the dripping-to-jetting transition at Wecr≈2.2. Of the possible weaknesses, the most load-bearing is the evidential basis for this value. The bracket [2.17, 2.27] is consistent with the experimental window 2<We<3 but does not by itself justify the stated precision; the abstract and conclusion both assert 'approximately 2.2' as a finding. The absence of a resolution study is especially salient because the jet diameter is only 20 lattice units, so the interface thickness (ξ=4) is 20% of the jet diameter and the breakup dynamics may be under-resolved. Inlet conditions in real nozzles are not exactly uniform, and Rayleigh breakup is sensitive to initial disturbances; the paper expressly excludes artificial disturbances, which could change the effective transition. The authors' own caveat in Sec. 6.3 acknowledges the boundary may vary. These are addressable by targeted numerical experiments. The other issues (the erroneous derivation of Eq. 7, the implausible CPU baseline in Table 2, and the missing repository link) are real but do not directly threaten the jet-physics claim as much as this evidential gap does. The benchmark validations (capillary wave, stationary drop, Poiseuille flow, shear interface, Rayleigh-Taylor) give the framework some independent support, which is why the paper should remain CONDITIONAL rather than be rejected. My read agrees with the Reader's verdict and weakest assumption.","tokens_in":30646,"tokens_out":5498,"duration_ms":54660,"concrete_test":"Rerun the two bracketing jet cases (We=2.17, Re=332 and We=2.27, Re=340) under three conditions: (i) doubled spatial resolution (Dj=40 lu, domain 240x240x600/800); (ii) a parabolic inlet velocity profile in place of the uniform profile; (iii) the uniform profile with a small harmonic perturbation (amplitude ~1% of uj) at the inlet. If the We=2.17 case transitions to jetting or the We=2.27 case returns to dripping under any of these changes, the claimed Wecr≈2.2 is not robust; if both retain their regimes across all three perturbations, the concern is resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central quantitative claim, We_cr≈2.2, is inferred from exactly two simulations: We=2.17 (dripping) and We=2.27 (jetting) in Sec. 6.3. No grid refinement is reported for the jet (the domain is 120x120x300/400 with Dj=20 lu, giving roughly 5 cells across the jet diameter), no inlet perturbation is applied, and the side walls are free-slip. Experiments [53] bracket the transition only as 2<We<3, so any value in that range is consistent; the refined value 2.2 is therefore a fitted midpoint, not a measured outcome. The setup matters: with a perfectly uniform inlet profile and no disturbances, capillary instability growth from numerical noise may be delayed or advanced relative to a realistic nozzle (which has a developed velocity profile and finite perturbations). The manuscript itself states the dripping/jetting boundary 'may not be sharply defined and could vary depending on specific conditions' (Sec. 6.3). If a modest perturbation or a factor-of-2 grid refinement moved either bracketing case across the regime boundary, the headline value would shift; the central claim depends on this not happening.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript introduces CLIP, a CUDA-accelerated phase-field lattice Boltzmann framework for immiscible two-phase flows with high density and viscosity contrasts. The method uses a conservative Allen-Cahn interface equation, a weighted multi-relaxation-time (WMRT) collision operator, and D3Q19/D2Q9 lattices. The paper validates the solver against capillary wave decay, Laplace-law stationary drops, two-phase Poiseuille flow, shear-driven interface deformation, and 2D/3D Rayleigh-Taylor instability, reporting close agreement with analytical solutions and prior numerical data. The framework is then applied to a water-air liquid jet in conditions matched to the experiment of Suñol and González-Cinca (2015), and the authors report breakup lengths and droplet sizes in agreement with the experiment and claim a dripping-to-jetting transition at a critical Weber number Wecr ≈ 2.2.","tokens_in":30830,"tokens_out":3477,"duration_ms":38857,"significance":"If the claims hold, this is a useful contribution: the paper supplies an open-source GPU-accelerated solver, documents the CUDA implementation, and demonstrates the model on several nontrivial benchmarks, including density ratios of 1000 and 3D Rayleigh-Taylor evolution. The benchmark evidence is credible and reasonably complete: Laplace-law errors are below about 1%, the two-phase Poiseuille flow shows roughly second-order convergence, and the Rayleigh-Taylor positions track published data. The novelty is primarily in the integration and GPU implementation rather than in a new constitutive model. The main scientific claim beyond the benchmarks is the jet transition value Wecr ≈ 2.2, which is a falsifiable, physically meaningful output; however, as discussed in the major comments, the support for this specific value is currently too thin, so the paper needs revision before the central jet claim can be accepted.","major_comments":[{"comment":"The claimed critical Weber number Wecr ≈ 2.2 is inferred from exactly two bracketing simulations: We = 2.17 is classified as dripping and We = 2.27 as jetting. No grid-refinement study is reported for the jet simulations. Given the stated domain of 120 × 120 × 300 (or 400) lattice nodes and the domain size 6Dj × 6Dj × 20Dj, the jet diameter is resolved by only about 20 lattice units, which is marginal for resolving the thin neck and the growth of capillary perturbations. A resolution study on the two bracketing cases (e.g., doubling the lattice resolution while holding dimensionless parameters fixed) is needed to establish that the dripping/jetting classification and the inferred transition value do not shift with resolution. The manuscript's own caveat in §6.3 that the boundary 'may not be sharply defined and could vary depending on specific conditions' underscores that this point is load-bearing.","section":"§6.3, Fig. 22"},{"comment":"The jet simulations use a uniform velocity profile at the inlet, free-slip side walls, and a convective outlet, with the text explicitly stating 'no artificial disturbances'. Liquid jet breakup is highly sensitive to inlet velocity profile, nozzle geometry, and the amplitude of disturbances; a perfectly uniform inlet removes the naturally occurring perturbation spectrum that triggers capillary instability in experiments. The manuscript does not test whether the idealized inlet profile reproduces the nozzle conditions of Ref. [53], nor does it report a perturbation-sensitivity study. Because the dripping/jetting boundary is identified by the behavior of these two simulations, the authors should either add simulations with a controlled inlet perturbation (e.g., small sinusoidal or broadband disturbances of amplitude 0.1–1% of uj) or provide evidence that the uniform inlet gives the same transition as a developed profile. Without this, the claimed Wecr ≈ 2.2 could be an artifact of the idealized setup even if all validation benchmarks are correct.","section":"§6.1, boundary conditions"},{"comment":"The manuscript states that the jet breakup length and droplet size results 'align well' and show 'excellent agreement' with the experiment, but no quantitative error metric is reported. The experimental points in Fig. 21 appear without error bars or a stated measurement uncertainty, and the numerical data in Fig. 21b show noticeable scatter in the jetting regime. The authors should report a quantitative comparison (e.g., mean or maximum relative deviation for Lb/Dj and de/Dj) and comment on the scatter, including the role of coalescence events that are mentioned in the text. A quantitative error statement is needed to substantiate the abstract's claim that the results 'closely match experimental observations.'","section":"§6.2, Fig. 21"}],"minor_comments":[{"comment":"The pseudocode for velocity update reads 'ux← f[index][ex[i]]' and similarly for uy and uz, which is not a valid moment summation; it should presumably be something like 'ux ← ux + f[index][i] * ex[i]' (and analogously for y and z). Please correct the pseudocode.","section":"Algorithm 3, lines 26–28"},{"comment":"The entries in Table 1 are ambiguous (e.g., 'Color-fluid 10 3 > 104 > 104 > 1'), and the column alignment is not clear. Please reformat the table so that each dimensionless number has an explicit value or range, and add a note explaining the provenance of these reported capabilities.","section":"Table 1"},{"comment":"In Eq. (7), the term 'γ 1− 4(ϕ−ϕ0)2 ξ n' is missing parentheses; it should read γ (1 − 4(φ − φ0)^2 / ξ) n, as in the subsequent conservative form in Eq. (8).","section":"Eq. (7)"},{"comment":"Table 6 marks some cases with an asterisk and states that these are chosen to match the target experimental study, but not all compared cases are marked. Please clarify which cases are direct experimental matches and which are additional numerical cases used to map the regime diagram.","section":"Table 6"},{"comment":"Reference [8] is listed with inconsistent volume/year information (volume 27, pages 821–834, year 2012 in the first occurrence and 2017 in the DOI line), and the author name 'Suñol' appears as 'Sunol' in figure captions and text. Please standardize these.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The benchmark portion of the paper is solid and the open-source GPU code is a real asset. The main reservation is that the headline quantitative jet result, Wecr ≈ 2.2, rests on two simulations with an idealized inlet and no resolution or perturbation study; this is a load-bearing claim for the paper's abstract and conclusions. The revision should add a focused sensitivity study, or substantially soften the claim. This is a technical article that fits the journal's scope once the jet-transition evidence is strengthened."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Punchline: this is a competent CUDA phase-field LBM paper whose benchmark validation deserves attention; the headline We_cr ≈ 2.2 should not be quoted as a measured result until they add sensitivity work. The genuinely new engineering piece is a GPU framework assembling published components (conservative Allen-Cahn, WMRT, D3Q19) and applying it to water–air jet breakup. No code is linked and nothing is machine-checked, so the contribution is a numerical framework plus a refined regime boundary.\n\nWhat the paper does well: five benchmark validations, including Laplace-law errors below 1%, second-order convergence in two-phase Poiseuille flow, capillary-wave decay matching Prosperetti, and Rayleigh–Taylor front positions matching prior simulations. The jet breakup length and drop size trends track Suñol and González-Cinca. These give reasonable confidence in the base solver. The critical Weber number is an output, not a fitted constant, which is also good.\n\nSoft spots, in proportion: the derivation in Eq. (7) is garbled, though the final conservative form in Eq. (8) is the standard model. More importantly, We_cr ≈ 2.2 is inferred from exactly two simulations, We = 2.17 (dripping) and We = 2.27 (jetting), with no grid refinement, no inlet perturbation, free-slip side walls, and a convective outlet. The jet diameter is 20 lattice cells, which is coarse for breakup dynamics, and a uniform inlet profile leaves capillary instability to grow from numerical noise. The authors themselves concede the regime boundary may vary. The stress-test note says there are about 5 cells across the jet diameter; that is wrong, Dj = 20 lu gives 20 cells across, but the underlying concern about missing sensitivity studies is correct. The conclusion promises open-source code at a GitHub page, yet no URL appears. The CPU baseline in Table 2, 102 seconds per iteration for a 1M-node D3Q19 domain, looks implausibly slow and needs a fair baseline before the speedup claim carries weight. These are addressable issues, not fatal ones.\n\nBottom line: for readers building GPU LB solvers, the framework details and benchmarks are useful. The jet transition value is plausible within the experimental band 2 < We < 3, but the claimed precision is not supported. It deserves a serious referee, and I would engage with a revised version that adds resolution and perturbation sensitivity for the jet cases and fixes the code link and timing baseline.","headline":"A solid CUDA phase-field LBM benchmark suite, but We_cr ≈ 2.2 is a two-simulation bracket under idealized jet conditions and should not be treated as a sharp result yet.","tokens_in":31456,"tokens_out":2445,"would_cite":false,"duration_ms":27521,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["47.55.df","47.55.nb","47.11.-j"],"model":"deepseek-v4-flash","headline":"GPU lattice Boltzmann solver reproduces liquid jet breakup and sets drip-to-jet transition at Weber 2.2.","keywords":["lattice Boltzmann method","phase-field model","Allen-Cahn equation","weighted multi-relaxation time","CUDA","GPU acceleration","liquid jet breakup","dripping and jetting regimes"],"falsifier":"Run the same water-air jet setup at Weber numbers just below and above 2.2 while adding a controlled inlet perturbation amplitude (for example, 1 percent of the jet velocity at a few Rayleigh frequencies) and observe whether the dripping-to-jetting transition shifts noticeably; a large shift would falsify the paper's claim that the transition is an intrinsic property of the capillary jet under the stated conditions.","tokens_in":754,"feed_emoji":"💧","tokens_out":2323,"duration_ms":105025,"temperature":0.7,"pith_summary":"This paper presents a CUDA-accelerated, phase-field lattice Boltzmann framework, named CLIP, for simulating two-phase flows with large density and viscosity contrasts, and claims that it quantitatively reproduces the breakup of a liquid jet in air. The authors validate the solver on five benchmark problems: capillary wave decay, a stationary drop obeying Laplace's law, two-phase Poiseuille flow, a circular interface in shear flow, and Rayleigh-Taylor instability. They then simulate a water jet at a density ratio of 814 with physical properties matched to a published experiment, and find that the dripping-to-jetting transition occurs at a critical Weber number of about $2.2$, inside the range $2<\\mathrm{We}<3$ reported by that experiment. Simulated breakup lengths and droplet sizes follow the same trends as the measurements across both regimes. If the claim holds, a desktop-computable solver can connect interface physics to an engineering-scale regime boundary without a computing cluster.","feed_headline":"Jet transition pinned at Weber 2.2 by GPU lattice Boltzmann","feed_subtitle":"A desktop-computable CUDA framework reproduces droplet sizes and breakup lengths in water-air jets.","key_machinery":"The load-bearing machinery is a pair of lattice Boltzmann equations on a D3Q19 lattice, which has nineteen discrete velocities. One distribution function set solves the incompressible Navier-Stokes equations with surface tension written through a chemical-potential force; the other solves the conservative Allen-Cahn equation for the phase-field variable $\\phi$, which marks the liquid and gas and defines the diffuse interface. Stability at density ratio $814$ comes from the weighted multi-relaxation-time (WMRT) collision operator, a variant of MRT that relaxes weighted, orthogonalized moments rather than simple velocity-space populations. Speed comes from mapping collision, streaming, boundary, and macro-variable steps onto the GPU's parallel memory layout, so one iteration over sixteen million or more lattice nodes runs in fractions of a second. The jet application adds three boundary choices: a uniform velocity inlet, free-slip side walls, and a convective outlet, all intended to keep the domain free of artificial disturbances so that capillary physics alone drives breakup.","core_discovery":"The core discovery the authors argue for is that the combination of the conservative Allen-Cahn interface-tracking equation, the weighted multi-relaxation-time (WMRT) collision operator, and GPU parallelism makes a phase-field lattice Boltzmann model stable and accurate enough to act as a predictive tool for liquid jet breakup. Concretely, their D3Q19 simulation of a water jet in air at $\\mathrm{Oh}=4.4\\times10^{-3}$ and density ratio $814$ places the transition from dripping to jetting at a Weber number of approximately $2.2$, while the matching experiment places it between $2$ and $3$. The same simulations reproduce the observed increase of breakup length with Weber number in both regimes and the decrease of mean droplet size across the transition, including coalescence events in the jetting regime. On the paper's terms, this means the model captures the force balance that governs jet breakup and locates the regime boundary more tightly than the experiment alone did.","pith_inferences":["If $\\mathrm{We}_{\\mathrm{cr}}\\approx 2.2$ is robust, it offers a clean calibration point for other diffuse-interface and volume-of-fluid methods at the same Ohnesorge number; discrepancies would indicate sensitivity to the interface model or boundary treatment rather than to unresolved turbulence.","The paper's jet setup uses an idealized, uniform inlet with no imposed perturbations; the transition value may shift if realistic velocity fluctuations or nozzle geometry are introduced, and a systematic sweep of inlet perturbation amplitude would reveal how much of the result is a property of the ideal capillary jet.","The same CUDA-based framework could be used to produce a detailed Ohnesorge-Reynolds regime diagram for a wider range of liquid-gas pairs, and the critical-Weber line could be tested against the classical Ohnesorge correlation, which is an extension the paper does not explicitly compute."],"forward_implications":["The dripping-to-jetting transition in a water-air system at $\\mathrm{Oh}=4.4\\times10^{-3}$ is pinned to $\\mathrm{We}\\approx 2.2$, which is a sharper bound than the experiment's $2<\\mathrm{We}<3$ range.","A solver built on a desktop GPU can perform three-dimensional, high-density-ratio, high-viscosity-ratio multiphase simulations that previously required a high-performance computing cluster.","The validated benchmarks establish a baseline accuracy for the phase-field WMRT approach: capillary wave decay, Laplace's law, two-phase Poiseuille flow (about second-order convergence), shear-deformed interface recovery, and Rayleigh-Taylor spike and bubble evolution.","Because the phase-field formulation keeps the interface diffuse but locally tracked, the approach can be extended to other low-Weber-number interfacial processes such as drop formation on demand or microfluidic droplet generation.","The reported speedups, about 48 times for a two-dimensional case and 86 times for a three-dimensional case on a modest desktop GPU, indicate that parameter sweeps over dimensionless groups are now practical on standard hardware."],"supporting_citations":[{"why":"The experimental liquid jet breakup data for breakup length, droplet size, and transition range that the jet simulations are matched against.","marker":"[53]"},{"why":"The phase-field-based lattice Boltzmann model for immiscible fluids with density and viscosity contrasts that the present approach builds on.","marker":"[19]"},{"why":"Prosperetti's analytical solution for capillary wave decay used to validate the transient interfacial dynamics.","marker":"[47]"},{"why":"The conservative phase-field lattice Boltzmann equation for interface tracking that the paper adapts for the phase-field distribution function.","marker":"[43]"},{"why":"The cumulant and weighted multi-relaxation-time lattice Boltzmann formulation that supplies the basis for the WMRT collision operator.","marker":"[42]"},{"why":"The three-dimensional Rayleigh-Taylor instability data used as a benchmark reference for the 3D validation cases.","marker":"[21]"}],"fun_headline_variants":["Weber 2.2 pins dripping-to-jetting transition","Desktop GPU model predicts jet breakup at Weber 2.2","CUDA lattice Boltzmann nails jet transition near Weber 2","Lattice Boltzmann on GPU reproduces jet breakup regimes","GPU phase-field model locates jet regime shift at Weber 2.2"],"cache_read_input_tokens":33536,"weakest_assumption_plain":"The jet simulations assume an idealized setup: a uniform velocity at the nozzle inlet, free-slip side walls, and a convective outlet with no artificial disturbances, and the claim that $\\mathrm{We}_{\\mathrm{cr}}\\approx 2.2$ depends on these conditions being physically representative of the experiment.","fun_headline_variants_meta":{"raw":{"variants":["Weber 2.2 pins dripping-to-jetting transition","Desktop GPU model predicts jet breakup at Weber 2.2","CUDA lattice Boltzmann nails jet transition near Weber 2","Lattice Boltzmann on GPU reproduces jet breakup regimes","GPU phase-field model locates jet regime shift at Weber 2.2"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000218,"raw_usage":{"total_tokens":1443,"prompt_tokens":954,"completion_tokens":489,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":570,"completion_tokens_details":{"reasoning_tokens":402}},"tokens_in":570,"tokens_out":489,"duration_ms":4400,"temperature":1.0,"reasoning_tokens":402,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T20:39:45.466412+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same water-air jet setup at Weber numbers just below and above 2.2 while adding a controlled inlet perturbation amplitude (for example, 1 percent of the jet velocity at a few Rayleigh frequencies) and observe whether the dripping-to-jetting transition shifts noticeably; a large shift would falsify the paper's claim that the transition is an intrinsic property of the capillary jet under the stated conditions.","supporting_citations":[{"cited_title":"Su˜ nol, R","cited_arxiv_id":null,"evidence_quote":"The experimental liquid jet breakup data for breakup length, droplet size, and transition range that the jet simulations are matched against."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The three-dimensional Rayleigh-Taylor instability data used as a benchmark reference for the 3D validation cases."}],"review_version":1}