{"id":"a6d98d9b-e142-4390-815b-f8ec1247b11a","arxiv_id":"2608.07960","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Squeeze-splash velocity is predicted by adding both local and convective inertia to the disk deceleration and using a momentum-averaged ejection velocity instead of the peak value.","lead":"This paper explains why earlier theory overpredicts the speed of liquid splashing out from between two squeezing disks, and offers a corrected formula. The fix works by including both types of liquid inertia and by using the average, not the fastest, ejection speed to represent the splash.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The measured fastest splash front is identified with the momentum-averaged velocity ve_avg from Eq 5 without direct verification; for low-viscosity liquids the stated viscous-homogenization mechanism is too slow, so the central comparison may be testing a different observable.","rationale":"I read the paper in good faith and found the theoretical machinery internally consistent: Eq 4 is correctly derived from the inertial radial momentum balance with Eq 1, Eq 5 follows algebraically from the momentum-flux definition, and the validation spans a wide parameter range with no fitted parameters. The single most load-bearing link is the identification of the experimentally tracked fastest splash front with the momentum-averaged velocity ve_avg. The reader flagged exactly this assumption, and my stress-test strengthens it with a quantitative timescale argument: for water and the low-viscosity glycerol mixture, viscous diffusion times are orders of magnitude longer than the ejection time, so the stated physical mechanism for homogenization cannot operate in the measurement window. This does not make the paper's claim false, but it makes the central comparison unverified. A two-phase DNS that measures both the tracked front speed and the mass-weighted mean velocity in the same geometry would settle the issue. Since this is the same concern that motivated the conditional verdict, I recommend no change to the reader's verdict.","tokens_in":6524,"tokens_out":9155,"duration_ms":105740,"concrete_test":"Run an axisymmetric two-phase DNS (Basilisk or OpenFOAM interFoam) of the G50W50 case (u0 = 2.41 m/s, R = 30 mm, h0 = 2.48 mm, M = 0.518 kg, mu = 5 mPa-s, rho = 1120 kg/m^3) with no-slip disks, resolving the ejected bulk splash. Measure (i) the fastest front position versus time exactly as in the experiment and (ii) the mass-weighted mean radial velocity of the ejected liquid over the same window. If the front-tracked velocity exceeds ve_avg from Eq 5 by more than the reported roughly 2.5% experimental error, the central observable identification fails; if it matches ve_avg, the concern is resolved. As a cheaper cross-check, recompute both max(ve) and ve_avg from the same ODE and plot them against the Fig 3 data for the water and G50W50 cases, since the two candidates bracketing the measured velocity would reveal which observable the data actually support.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The derivation of Eqs 4 and 5 is internally consistent: Eq 4 follows from plug-flow inertia with Eq 1, and Eq 5 follows from psi = integral(2*pi*R*h*rho*ve^2 dt) divided by m = pi*R^2*h0*rho. The load-bearing step is not the algebra but the claim that the experiment's 'fastest-moving splash front' equals ve_avg. Eq 5 is the center-of-mass velocity of all ejected liquid; a leading edge can be much faster if velocity gradients persist. The paper's homogenization argument is quantified nowhere, and the numbers contradict it for the low-viscosity cases: with nu about 1e-6 m^2/s for water, the viscous diffusion time across even a 1 mm ejected layer is about 1 s, while the whole squeeze/ejection lasts h0/u0 about 1 ms. Internal viscous resistance therefore cannot equalize the layers of the water splash in the measurement window; the same order-of-magnitude problem holds for G50W50, and for G86W14 the homogenization time is only marginal. The residual-film assumption (m = pi*R^2*h0*rho although u approaches zero leaves h > 0) also biases ve_avg, and the paper only hypothesizes that the residual film is pulled into the splash. Because the comparison in Figs 3 and 4 depends on this mapping, the central claim is not yet established.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"Chandra, Perrier, and Brutin revisit the squeeze-film splashing of a viscous liquid between two parallel disks. They argue that previous theory over-predicts the measured splash velocity for two reasons: the inertial resistance of the liquid was incompletely treated, and the peak ejection velocity was used as the representative splash velocity. Their framework couples the kinematic relation Eq. (1), the disk equation of motion Eq. (2), and an inertia term Eq. (4) that includes both local and convective radial acceleration. The proposed observable is a momentum-averaged ejection velocity, Eq. (5), obtained by integrating the radial momentum flux over the entire ejection and dividing by the nominal initial film mass. The model has no fitted parameters and is compared with the authors' high-speed measurements on water and two glycerol-water mixtures (Fig. 3) and with literature data over a wider parameter range (Fig. 4b).","tokens_in":6882,"tokens_out":6293,"duration_ms":72040,"significance":"If the central claim holds, the paper provides a parameter-free, physically motivated leading-order prediction for squeeze-driven splash velocity, with the practical virtue of requiring only geometric and material inputs. The authors correctly identify the two most plausible sources of the historical discrepancy and give a reproducible derivation of Eqs. (4) and (5). Validation against independent data from Bazilevsky and Rozhkov over two to three orders of magnitude in mass, radius, and viscosity is a genuine strength. The main weakness is that the measured 'fastest-moving splash front' is identified with the momentum-averaged quantity v_e,avg without direct verification, and this identification is doubtful for the low-viscosity cases in which viscous homogenization is too slow. The paper therefore presents a plausible and useful framework, but its central quantitative claim is not yet established.","major_comments":[{"comment":"Equation (2) is dimensionally inconsistent as written: u is defined as the vertical velocity, so M d^2u/dt^2 has units of kg m/s^3 rather than force. The intended first-order equation is M du/dt = Mg - F_v - F_i. Because Eq. (2) is one of the coupled equations solved numerically to produce all theoretical predictions, the manuscript must state the correct form and confirm that the numerics used the first-order version.","section":"Eq. (2)"},{"comment":"The load-bearing identification of the experimentally tracked 'fastest-moving splash front' with the momentum-averaged v_e,avg is not justified. For water (nu ~ 1e-6 m^2/s), the viscous diffusion time across even a 1 mm ejected layer is about 1 s, far longer than the squeeze/ejection time h0/u0 ~ 1 ms; thus internal viscous resistance cannot homogenize the layers of the water splash before the front speed is measured. If the front instead tracks the fastest parcel, the comparison in Figs. 3 and 4b tests a different quantity from Eq. (5). I request a direct test, such as particle tracking or PIV inside the splash, or a quantitative homogenization-time estimate with measured velocity profiles.","section":"Paragraph defining psi and v_e,avg, and Eq. (5)"},{"comment":"The manuscript acknowledges that the film thickness never reaches zero and that 'momentum averaging should be done considering a mass ... smaller than the mass of the initial liquid film,' but then adopts m ~ pi R^2 h0 rho on the hypothesis that residual liquid is pulled into the splash. No evidence is provided for this hypothesis, and it is least plausible for the low-viscosity cases in which the claimed agreement is best. This normalizer directly enters Eq. (5); the authors should quantify the ejected mass, for example from the images or from a mass balance, or bound the resulting error.","section":"Numerical-solution paragraph before Fig. 3"},{"comment":"The validation figures contain no error bars and no point-by-point residuals or quantitative agreement metric. The statement that slope standard errors are below 2.5% addresses only the linear fits used to estimate u0 and v_e,avg, not the spread between theory and experiment, the uncertainty in h0, or the pixel-resolution limit (0.11 mm/pixel at 5000 fps, corresponding to roughly 0.55 m/s per frame in front speed). Without raw data or a table of values, the 'good agreement' claimed in Figs. 3 and 4b cannot be assessed quantitatively.","section":"Figs. 3 and 4b"}],"minor_comments":[{"comment":"The sentence 'We numerically solve equation 5 to compute the momentum-averaged splash velocity' is confusing: Eq. (5) is a quadrature, while the coupled system (1), (2), and (4) is what is actually solved. Please describe the algorithm as evolving the disk motion and accumulating the integral over dh.","section":"Numerical solution paragraph"},{"comment":"Please clarify how the effective mass M is determined from the lever-arm geometry; it is a key input to Eq. (2), and no uncertainty is reported for it.","section":"Experimental methods"},{"comment":"The text refers to 'Figure 1a' for the water splash, but Figure 1 as printed has no visible panel labels; please correct the citation or add the labels.","section":"Figure 1 reference"},{"comment":"There are several typographical and spacing errors, including 'Acknowledgment-This' and missing spaces such as 'ofthesplashvelocity'; these should be corrected throughout.","section":"Typographical issues"},{"comment":"The data availability statement says no data or software support the manuscript; for a validation-focused paper, providing the raw measurement tables and the MATLAB solver would materially improve reproducibility and allow the referee and readers to check the agreement claimed in Figs. 3 and 4b.","section":"Data availability"}],"recommendation":"major_revision","confidential_remarks":"The derivation of Eqs. (4) and (5) is internally consistent, and the paper contains a useful, parameter-free framework. The hinge is the experimental identification of the measured front speed with the momentum-averaged v_e,avg; the current manuscript does not verify this and gives order-of-magnitude reasons to doubt it for water. I would support re-review after the authors either provide direct velocity-field evidence of homogenization or substantially reframe the claim and restrict it to regimes where the mapping is credible. Please also insist on correcting Eq. (2) and on making the validation data quantitative, since the current figures cannot be checked in detail."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read it. The new piece is the combination of a corrected inertia term (local plus convective, already in Lang et al.) with the momentum-averaged ejection velocity as the splash observable. That averaging idea is simple and, as far as the citations show, genuinely absent. The ODE system is parameter-free, Eq 5 follows from momentum conservation, and the collated dataset spans a useful range. Credit where due: this is a clean, testable claim.\n\nSoft spots. The biggest is the bridge from theory to experiment. They measure the fastest-moving splash front, then compare to ve_avg, the center-of-mass speed of all ejected liquid. Unless internal stresses wipe out velocity gradients fast, the leading edge moves faster than the average. The paper's homogenization argument is qualitative and, for water, the viscous diffusion time across a millimeter layer is about a second, while the squeeze lasts a millisecond. That order-of-magnitude mismatch is not addressed. It is possible the front is actually slowed by a different mechanism—fast parcels catching up and merging with slow fluid—but the paper doesn't say that. The residual-film assumption (all mass m = πR² h0 ρ goes into the splash) is also asserted, not measured. Eq 2 has a dimensional typo: d²u/dt² for u a velocity should be du/dt. No raw data or error bars are given, just a claim of <2.5% slope error.\n\nNone of this kills the paper. The empirical agreement might still be real, but the comparison is testing an unverified map between observable and theory. A referee should ask for direct evidence: e.g., particle tracking or simulation showing the front speed equals ve_avg, and a measurement of the residual film.\n\nRecommendation: send to peer review. It's a serious, parameter-free attempt with a plausible central idea, and the subfield (blood spatter, squeeze-flow) would benefit from an independent check. Reading group: yes, it'll generate good discussion.","headline":"A clean parameter-free theory for squeeze-film splash velocity, but the mapping from measured front speed to the momentum-average is asserted, not shown, and that is the load-bearing step.","tokens_in":7330,"tokens_out":5923,"would_cite":true,"duration_ms":65930,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A viscous liquid squeezed between parallel disks ejects a peripheral splash whose bulk speed is set by the momentum-averaged ejection velocity, not the peak velocity, and including both local and convective inertia makes theory match…","keywords":["squeeze flow","splash velocity","momentum-averaged ejection velocity","liquid inertia","parallel disks","viscous liquid","high-speed imaging","blood spatter"],"falsifier":"Measure the radial velocity profile across the thickness of the ejected splash at several instants and compare the fastest front speed to both $\\max(v_e)$ and $v_{e,\\mathrm{avg}}$; if the front tracks the fastest parcel, or if a measurable liquid film remains trapped between the disks so that $m \\approx \\pi R^2 h_0 \\rho$ is the wrong normalizer, the proposed agreement would fail.","tokens_in":6351,"feed_emoji":"💧","tokens_out":5805,"duration_ms":65562,"temperature":0.7,"pith_summary":"The paper tries to establish why existing theory overpredicts the speed of the splash thrown out when a disk falls onto a liquid film, and to supply a corrected prediction. It argues that earlier theory left out part of the liquid's inertia and compared its prediction to the wrong observable: the peak ejection velocity. Once both local and convective inertia are kept in the disk's equation of motion, and once the splash speed is represented by a momentum-averaged ejection velocity, the model matches high-speed experiments for water and glycerol-water mixtures and independent data spanning wide parameter ranges.","feed_headline":"Momentum averaging fixes squeeze-film splash velocity predictions","feed_subtitle":"Both inertia terms kept, then ejection speed averaged; matches experiments across viscosities and scales.","key_machinery":"The central object is the momentum-averaged ejection velocity (Eq. 5), $v_{e,\\mathrm{avg}} = (1/h_0)\\int_0^{h_0} v_e\\, dh$, derived from the rate of radial momentum ejection $\\dot{\\psi} = 2\\pi R h \\rho v_e^2$ divided by the total splash mass $m \\approx \\pi R^2 h_0 \\rho$. It replaces the peak value $\\max(v_e)$ as the observable for bulk splash speed. It is fed by the coupled system: kinematic mass conservation $v_r = u r/(2h)$, the top-disk equation of motion $M\\,d^2u/dt^2 = Mg - F_v - F_i$, and the corrected inertial resistance $F_i = \\rho\\pi R^4\\left(3u^2/(16h^2) + \\dot{u}/(8h)\\right)$, with $F_v = 3\\pi\\mu u R^4/(2h^3)$. The ODE system is solved numerically, and $v_{e,\\mathrm{avg}}$ is computed by integrating $v_e$ over $h$ from $0$ to $h_0$.","core_discovery":"The bulk splash velocity of a squeezed viscous film is not the maximum radial velocity reached at the disk edge during squeezing. The relevant speed is the momentum-averaged value $v_{e,\\mathrm{avg}} = (1/h_0)\\int_0^{h_0} v_e\\, dh$, obtained by conserving linear momentum of the coherent splash. The paper shows that correcting the inertial resistance term by including both local and convective acceleration (Eq. 4) and then averaging the ejection velocity over the film thickness brings theory into good agreement with their experiments and with literature data.","pith_inferences":["If the homogenization argument is right, there should be a dimensionless group comparing the viscous homogenization time of the ejected layer to the squeeze time; the transition from dome-shaped splashes (water) to disk-shaped splashes (viscous liquids) should be encoded in that group, which the paper leaves implicit.","The early spray-like ejecta, excluded here, may set initial conditions for fragmentation; coupling the bulk $v_{e,\\mathrm{avg}}$ with an ejecta velocity model could predict droplet size distributions in blood spatter.","Applying the momentum-averaging logic to non-Newtonian liquids would replace $F_v$ with a viscoelastic force, and the resulting prediction should be testable with existing squeeze-flow rheometry data.","The framework's use of the entire film mass $m \\approx \\pi R^2 h_0 \\rho$ could be checked directly by weighing the residual film left between the disks after impact; if the residual is large, $v_{e,\\mathrm{avg}}$ should be renormalized by the ejected mass only."],"forward_implications":["The splash velocity is fixed by the coupled ODEs (Eqs. 1, 2, 4) plus Eq. 5, with no fitted parameters, so any new squeeze experiment can be compared to a unique prediction.","The relevant observable for splash propagation is the momentum-averaged ejection velocity, so future analyses should stop using peak ejection velocity as the proxy for splash speed.","Both local and convective inertia must enter the inertial resistance; dropping either one reintroduces a large overestimate.","The corrected model works across roughly three orders of magnitude in viscosity and one order in disk radius, including independent data, so it can serve as a leading-order baseline for splash morphology and fragmentation.","The result gives a quantitative input for bloodstain-spatter modelling, where bulk splash speed is the leading-order quantity."],"supporting_citations":[{"why":"Bazilevsky and Rozhkov's impact experiment and theory is the main baseline whose peak-velocity overprediction this paper explains and whose image sequences are re-analyzed.","marker":"[22]"},{"why":"Viscous-liquid round-splash data provide independent validation across much higher viscosities and different disk geometry.","marker":"[23]"},{"why":"Transient squeezing-flow solutions supply the corrected inertial term that includes local acceleration alongside convective acceleration.","marker":"[18]"}],"fun_headline_variants":["Splash speed is not the peak: momentum averaging wins","Averaging ejection speed fixes squeeze-film splash predictions","Squeeze-film splash: peak velocity overestimated; average is key","Momentum-averaged ejection speed: the missing factor in splash theory"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The measured fastest-moving bulk splash front moves at the momentum-averaged velocity, which holds only if internal viscous resistance homogenizes velocity across ejected layers and nearly all the initial film mass is pulled into the splash before fragmentation.","fun_headline_variants_meta":{"raw":{"variants":["Splash speed is not the peak: momentum averaging wins","Averaging ejection speed fixes squeeze-film splash predictions","Squeeze-film splash: peak velocity overestimated; average is key","Momentum-averaged ejection speed: the missing factor in splash theory"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000432,"raw_usage":{"total_tokens":2107,"prompt_tokens":750,"completion_tokens":1357,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":366,"completion_tokens_details":{"reasoning_tokens":1284}},"tokens_in":366,"tokens_out":1357,"duration_ms":13175,"temperature":1.0,"reasoning_tokens":1284,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T00:37:12.619961+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the radial velocity profile across the thickness of the ejected splash at several instants and compare the fastest front speed to both $\\max(v_e)$ and $v_{e,\\mathrm{avg}}$; if the front tracks the fastest parcel, or if a measurable liquid film remains trapped between the disks so that $m \\approx \\pi R^2 h_0 \\rho$ is the wrong normalizer, the proposed agreement would fail.","supporting_citations":[{"cited_title":"Bazilevsky and A","cited_arxiv_id":null,"evidence_quote":"Bazilevsky and Rozhkov's impact experiment and theory is the main baseline whose peak-velocity overprediction this paper explains and whose image sequences are re-analyzed."},{"cited_title":"Bazilevskii and A","cited_arxiv_id":null,"evidence_quote":"Viscous-liquid round-splash data provide independent validation across much higher viscosities and different disk geometry."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Transient squeezing-flow solutions supply the corrected inertial term that includes local acceleration alongside convective acceleration."}],"review_version":1}