{"id":"6662f1e5-65eb-4ddc-be2d-e325dd24ed99","arxiv_id":"2504.16597","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Numerical simulations predict that thermocapillary Marangoni flow sorts small and large particles in evaporating droplets with a moving contact line, producing a small-particle core surrounded by large particles.","lead":"This paper simulates evaporating ethanol droplets containing two sizes of particles and finds that temperature-driven Marangoni flow makes small particles gather at the droplet's apex while large particles form a shell around them. A generalist might read it because particle sorting in drying droplets matters for inkjet printing, coatings, and medical diagnostics, and the paper identifies which flows cause or suppress the sorting.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The predicted small-core/large-shell particle sorting rests on an unvalidated, under-resolved stage-2 model: simplified interface forces (Eqs. 23-24) with a 1-MPa Young's modulus on a ~21 μm mesh for 3-5 μm particles, with no sensitivity, convergence, or experimental check.","rationale":"The reader's conditional verdict is sound. The flow-side results are the paper's strongest element: volume evolution against the analytical solution (Fig. 2), contact-radius evolution against Zhu and Shi (Fig. 3), and the Appendix A mesh study for the Marangoni case all support stage 1. My concern targets the particle stage, where all the conditions needed for the central claim must hold but are least secure. The sorting pattern is produced by Eqs. 22-24 with parameters that are neither measured nor swept, at a resolution (Rc/24 ≈ 21 μm) several times the particle diameter, so the interface forces that drive adsorption and lateral attraction are effectively sub-grid. The paper's own text in Section 6.3 describes the sorting as arising from particle interactions and gap filling, which makes the unvalidated contact and capillary models load-bearing rather than incidental. I also note a quantitative inconsistency worth correcting: Table 2 lists 607 particles and a 0.5 volume fraction, while Section 3.1 states 0.1% per size; for the stated droplet size these cannot all be true. The Eq. 14 dimensional issue (K_n with units m^3/N) is likely a typo for the standard Hertzian prefactor (4/3)E*√R*, but it signals that the particle-stage equations have not been checked to the same standard as the flow stage. Appendix B is honest and useful: it shows the vortex is fragile to 20-50 μm slip lengths, but physical ethanol slip lengths are typically far below 1 μm and the no-slip simulation does produce the vortex, so I weigh this as secondary to the particle-model gap. The proposed test (finer mesh, physical stiffness, parameter sweeps, multiple realizations) would settle whether the morphology is robust; until then CONDITIONAL is correct and my read does not change it.","tokens_in":21642,"tokens_out":22443,"duration_ms":200355,"concrete_test":"Rerun stage 2 for C2S and C4S at Rc/48 and Rc/96, with E=70 GPa, Δγ (Eq. 22) and f (Eq. 23) each swept over one decade, and five independent random particle initializations per configuration; compare the radial adsorbed/deposited concentration profiles at t50 and t25 (Figs. 11-12). If the small-core/large-shell morphology persists with peak-to-peak changes below ~20%, the sorting is robust and the concern is retired; if the pattern disappears, inverts, or varies beyond the small-vs-large separation run to run, the headline claim is a model artifact. A decisive experimental complement: evaporate ~1 mm ethanol droplets seeded with 3-5 μm silica particles on a heated (50 °C) hydrophobic substrate and image interface particle positions during the CCA phase to check whether small particles form an apex core.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim — that Marangoni flow sorts interface-adsorbed particles into a small-particle core with a large-particle shell (Section 7) — is generated entirely by the stage-2 particle model: the capillary adsorption force (Eq. 23), whose amplitude parameter f is never specified; the simplified Cheerios attraction (Eq. 24); DMT adhesion (Eq. 22) with uncalibrated Δγ; and Hertz contact with a Young's modulus of 1 MPa (Table 2), about five orders of magnitude below silica's ~70 GPa. Stage 2 has no validation, no mesh-convergence study, and no parameter sensitivity; the only robustness probe shown is the particle-density variation. The mesh selected in Section 4 (Rc/24 ≈ 21 μm) is 4-7 times the particle diameter (3-5 μm), so the interface position entering r_l in Eq. 23 is known to no better than about one cell — several particle diameters — making the adsorption criterion (r_l < R) and the sorting forces effectively sub-grid quantities. The mechanism stated in Section 6.3 ('gaps... preferentially filled by the small particles') is a packing outcome that depends precisely on these contact and capillary models; a different force balance, e.g., the uncalibrated DMT adhesion scaling with R*, could suppress or invert the pattern. Eq. 14 for K_n is dimensionally inconsistent (units of m^3/N rather than N/m^(3/2)), a verification concern in the same particle-stage equations. Appendix B additionally shows the Marangoni vortex behind the sorting is suppressed for slip lengths of 20-50 μm, so the flow-side mechanism is not fully independent of the contact-line treatment either. With a single random realization and no error bars on the concentration profiles (Figs. 10-12), the predicted morphology is a plausible but untested hypothesis: either it is a robust consequence of the vortex and survives the sensitivity suite below, or it is an artifact of the unvalidated model.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a two-stage numerical study of bidisperse particle dispersion in an evaporating ethanol sessile droplet on a heated substrate, evaporating in constant-contact-angle (CCA) mode. Stage 1 solves the gas-liquid flow with a VOF method, including heat and vapor transport and thermocapillary Marangoni stress; stage 2 tracks 607 particles (radii 1.5 and 2.5 μm) with a DEM/Lagrangian model including drag, DMT adhesion, a capillary adsorption force, and a simplified Cheerios force. Six first-stage cases vary the contact angle (60°, 90°, 120°) and the presence/absence of Marangoni stress, with silica and neutrally buoyant particles in stage 2. The central claims are: (i) in the absence of Marangoni flow, the receding-contact-line flow dominates and carries both particle sizes toward the apex without size sorting; (ii) with Marangoni flow, a single vortex creates an interface-parallel flow near the apex that produces size-based self-sorting, with small particles forming a core at the apex and large particles surrounding them.","tokens_in":22008,"tokens_out":3514,"duration_ms":27094,"significance":"If the sorting result holds, the paper would support a useful and falsifiable design rule: in CCA-mode evaporation with Marangoni stresses, the thermocapillary flow is the control knob for producing a small-particle-core/large-particle-shell ordered deposit. Strengths of the manuscript: the stage-1 flow solver is validated against an analytical d²-law volume evolution and against the Zhu and Shi experiments at three mesh resolutions; the mesh-independence checks are done for both evaporative volume evolution and the Marangoni-affected velocity/temperature fields; the code and data are openly available with a DOI; and the particle model is not fitted to the sorting outcome, so the sorting claim is a genuine prediction rather than a curve fit. The major weakness is that the sorting prediction rests entirely on the stage-2 particle-interface force model and particle-contact model, neither of which is validated, resolved, or subjected to sensitivity analysis.","major_comments":[{"comment":"The central sorting claim (small-particle core, large-particle shell) is produced entirely by the stage-2 force model, yet stage 2 has no validation, no mesh or time-step convergence study, and no sensitivity analysis. The mesh selected in Section 4 resolves the initial contact radius with 24 cells, i.e., approximately 21 μm per cell, while the particle radii are 1.5 and 2.5 μm; the interface position entering r_l in Eq. (23) is therefore known to no better than about one cell, which is several particle diameters. Because the adsorption criterion r_l < R and the capillary/Cheerios forces are sub-grid quantities, the predicted sorting is fragile and needs at least a resolved-interface or sub-grid-interface assessment and a parameter sensitivity study before the Section 7 conclusion can be supported.","section":"Sections 6.2–6.3 and Eqs. (23)–(24)"},{"comment":"The amplitude f of the capillary adsorption force is never specified, although this force is the mechanism that keeps particles at the interface and thereby enables the sorting pattern. Since the entire small-core/large-shell outcome depends on particles being adsorbed, the value of f (and its relation to the surface free energy of the liquid-particle interface) must be stated and varied.","section":"Section 2.2.1, Eq. (23)"},{"comment":"Equation (14) defines K_n = 4R*/(3E*), which has units of m³/N, but it is used in Eq. (13) as though the normal force were K_n δ_n^{3/2}, which requires units of N/m^{3/2}. This dimensional inconsistency is a verification concern in exactly the contact-force equations that determine the packing of the sorted core. The formula should be the Hertzian coefficient K_n = (4/3) E* sqrt(R*) (or the paper should otherwise clarify its normalization).","section":"Section 2.2.1, Eq. (14)"},{"comment":"Appendix B shows that imposing a 20–50 μm slip length suppresses the Marangoni vortex that is identified (Section 6.3) as the sorting mechanism, and the authors justify the no-slip choice by appealing to experimental evidence that the apex is the coldest region. However, the VOF method itself carries an implicit numerical slip of order one mesh cell (about 21 μm), as the authors note in Section 3.2, so the effective slip of the simulation is not actually zero. Given that the sorting flow is suppressed for slip lengths of 20–50 μm, the paper should demonstrate that the sorting result is robust to the effective numerical slip, or provide a quantitative argument that the implicit slip is below the threshold at which the vortex is suppressed.","section":"Appendix B"},{"comment":"The DMT adhesion force uses an uncalibrated work of adhesion Δγ, and the Hertz contact model uses a Young's modulus of 1 MPa (Table 2), about five orders of magnitude below the bulk value for silica. The small-particle core and large-particle shell pattern is described in Section 6.3 as arising from packing in gaps between particles, so the pattern depends precisely on these contact and adhesion parameters. The paper needs a sensitivity analysis with respect to E*, Δγ, and f to establish that the sorting pattern is not an artifact of these choices.","section":"Section 2.2.1, Eq. (22) and Table 2"}],"minor_comments":[{"comment":"The surface-concentration curves of the individual cases are reported without error bars, repeated realizations, or a discussion of statistical sampling fluctuations; since only 607 particles are simulated, the concentration peaks may not be statistically robust and a repetition with different random initial placements should be reported or at least discussed.","section":"Section 6.1, Figures 10–12"},{"comment":"The table lists 'Volume fraction of particles 0.5' while the text in Section 3.1 states that the volume fraction associated with each particle size is 0.1%; this is confusing and should be reconciled (e.g., 0.5% total vs. 0.1% per size).","section":"Table 2"},{"comment":"The caption of Table 3 states that cases C1–C6 have 'variation in substrate temperature,' but the substrate temperature is fixed at 50 °C for all cases; the caption should instead refer to variation in contact angle and Marangoni stresses.","section":"Table 3"},{"comment":"The sub-figure label in the caption of Figure 16 repeats '(d)' for the 50 μm slip case and the figure text 'drops' should read 'droplet'; these should be corrected.","section":"Appendix B, Figure 16"},{"comment":"The drag coefficient correlation in Eq. (20) uses the criterion Re < 1000 versus Re > 1000, but the second branch is typically used for Re > 1000; the 1000 value is a conventional crossover and should be cited or justified, and the use of the Wen–Yu correlation in a dilute one-way-coupled setting should be briefly justified.","section":"Section 2.2.1, Eq. (20)"}],"recommendation":"major_revision","confidential_remarks":"The paper is a candidate for publication after the stage-2 particle/interface-force model is properly supported. The stage-1 flow predictions are genuinely validated and useful on their own. The main concern is that the sorting claim, which is the title and conclusion, is generated by an under-resolved and unvalidated particle-interface model with several unspecified parameters, and the Appendix B sensitivity suggests the mechanism may be fragile. I would ask for the same standard of validation and sensitivity for stage 2 as the authors apply to stage 1. I also note that the paper's claim of no fitted parameters for sorting is roughly correct, which is a point in its favor."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the paper. The genuinely new result is a computational prediction: in a CCA-mode ethanol droplet on a heated substrate, the Marangoni vortex sorts interface-adsorbed bidisperse particles into a small-particle core with a large-particle shell. Alongside that, the authors report that the receding-contact-line flow dominates over capillary flow at all three contact angles they tested, which corrects Bhardwaj's angle-dependent picture. The flow-side story is credible. The droplet volume evolution matches the analytical solution and the contact radius matches Zhu and Shi's experiments, with mesh convergence at 24/32/40 cells per radius. They also run an honest slip-length study in Appendix B, showing that 20-50 µm slip suppresses the Marangoni vortex, and they justify the no-slip choice by comparison with observations.\n\nThe sorting result is where I part ways. It is produced entirely by the stage-2 particle model, and that model has real problems. Eq. 14 gives K_n = 4R*/(3E*), which with the δ^{3/2} factor does not produce a force; the units are m^3/N. If that is a transcription error, fine, but it needs to be corrected in the actual code. The particle Young's modulus is 1 MPa, five orders of magnitude below silica, which changes the contact mechanics behind the claimed 'small particles fill the gaps' packing. The capillary force amplitude f in Eq. 23 and the Cheerios parameters in Eq. 24 are never given values, so the sorting simulation is not reproducible from the paper alone. And the particles are 3-5 µm while the mesh cell is about 21 µm, so the interface position entering Eq. 23 is a sub-grid quantity. Add in a single random realization with no error bars, and the sorting pattern is a plausible hypothesis, not a demonstrated result.\n\nNone of this kills the paper. The flow field that drives the proposed mechanism is well resolved and validated, and the sorting mechanism itself is stated clearly enough to be tested. I would send this to peer review and ask for major revision: fix Eq. 14, specify or calibrate the particle force parameters, add a sensitivity sweep, and ideally an experimental target for the core-shell pattern. The flow reversal finding alone is worth refereeing. Data are on Zenodo, which helps.","headline":"A credible flow-side study that corrects the CCA-mode scaling picture, alongside a plausible but under-validated particle-sorting prediction that needs a serious referee.","tokens_in":22613,"tokens_out":4192,"would_cite":true,"duration_ms":41246,"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":"In an evaporating ethanol droplet with a moving contact line, thermocapillary Marangoni flow sorts bidisperse particles by size: small particles gather at the apex while large particles surround them.","keywords":["sessile droplet evaporation","Marangoni flow","particle self-sorting","bidisperse particles","moving contact line","VOF-DEM simulation","constant contact angle mode","core-shell deposition"],"falsifier":"Measure the final deposit and in-drop flow of a 60-degree contact-angle ethanol droplet containing 1.5 and 2.5 micrometer particles on a 50-degree-Celsius substrate evaporating in constant-contact-angle mode: the claim predicts a small-particle core at the apex ringed by large particles whenever the thermocapillary gradient is active, and predicts that suppressing the Marangoni vortex, for instance with a surfactant or a slip-promoting coating, removes the sorting entirely.","tokens_in":21401,"feed_emoji":"💧","tokens_out":7322,"duration_ms":67666,"temperature":0.7,"pith_summary":"This paper claims that thermocapillary Marangoni flow is the mechanism that sorts particles by size inside an evaporating ethanol droplet whose contact line is free to move. Using two-stage simulations that first compute the gas-liquid flow, heat, and vapor fields, then track 1.5 and 2.5 micrometer particles with a Lagrangian method, the authors find that without Marangoni stress the flow driven by the receding contact line dominates and sweeps both particle sizes to the droplet apex, so no size separation appears. With Marangoni stress, an interface-parallel vortex forms and drives adsorbed particles apart enough that small particles preferentially fill the gaps and collect at the apex, while large particles assemble around them. The pattern is reported for initial contact angles of 60, 90, and 120 degrees and is unaffected by whether the particles are denser than the fluid or neutrally buoyant. The practical upshot is that the internal flow becomes a controllable handle for arranging colloids in drying droplets, which matters for printing, coating, and particle deposition.","feed_headline":"Marangoni flow, not contact-line drag, sorts particles by size","feed_subtitle":"Thermocapillary stress gathers small particles at the droplet apex and wraps large ones around them.","key_machinery":"The load-bearing object is the temperature-dependent surface tension and its tangential gradient, $\\nabla_s\\sigma = \\sigma_T[\\nabla T - (\\nabla T\\cdot \\mathbf{n})\\mathbf{n}]$, which produces the Marangoni stress that drives the interface-parallel vortex. In the particle stage, the decisive interactions are the capillary force pinning a particle to the gas-liquid interface and the simplified Cheerios pair force between two adsorbed particles (Eqs. 23 and 24); these keep the particles at the interface where the tangential Marangoni flow can act on them, and the pair attraction is what large particles rely on to surround the small-particle core. The three-way flow competition among receding-contact-line flow, capillary flow, and Marangoni flow is the organizing framework that explains when sorting does and does not occur.","core_discovery":"The paper's central discovery is that in constant-contact-angle evaporation with a moving contact line, the direction of the internal flow is decided by a three-way competition: evaporation-driven capillary flow toward the contact line, the receding contact line dragging fluid toward the apex, and thermocapillary Marangoni flow along the interface. For the ethanol droplets studied, the receding-contact-line flow beats the capillary flow at every contact angle considered, so particles are carried from the bulk to the apex in a size-neutral way when Marangoni stress is absent. When Marangoni stress is present, it creates a single vortex whose near-interface branch moves adsorbed particles along the surface; this tangential motion lets small particles slip into gaps and build a compact core at the apex stagnation point, with large particles arranged around that core. The authors identify this Marangoni-driven flow, not wettability or particle density, as the mechanism responsible for the observed self-sorting.","pith_inferences":["Not claimed in the paper: if the vortex topology is the only requirement, any system generating the same interface-parallel circulation, whether from surfactants, solutal Marangoni stress, or external forcing, should produce the same small-core/large-shell pattern in other solvents.","Not claimed in the paper: the simplified Cheerios force lumps interface-deformation details into a few parameters, so the sharpness of the core-shell boundary should be sensitive to the particle's contact angle at the interface; a testable prediction is that changing particle hydrophobicity changes how cleanly small and large particles separate.","Not claimed in the paper: the mechanism suggests a miniature size-classification device, a drying droplet whose Marangoni strength sets the cutoff radius between particles that go to the core and particles that stay in the shell; the cutoff should scale with the ratio of capillary attraction to Marangoni drag."],"forward_implications":["In constant-contact-angle evaporation without Marangoni stress, size-based particle separation should not occur; suppressing thermocapillary convection is enough to suppress sorting.","Substrate heating and surface wettability can be used as knobs: turning the thermocapillary vortex on or off should switch between a mixed apex deposit and a small-core/large-shell pattern.","Particle density is not a controlling parameter for interfacial sorting; the same shell pattern should appear for heavy and neutrally buoyant particles.","The sorting should persist across contact angles from 60 to 120 degrees as long as the Marangoni flow dominates the receding-contact-line and capillary flows.","Because Marangoni mixing accelerates evaporation in shallow droplets but slows it in tall ones, the time window for sorting changes with contact angle and heating."],"supporting_citations":[{"why":"Supplies the validated two-stage VOF-DEM framework and the evaporation model that this study extends to moving contact lines and bidisperse particles.","marker":"[30]"},{"why":"Provides the prior scaling analysis claiming contact-line-to-apex flow only for contact angles above 90 degrees, which this study's all-angle result directly addresses.","marker":"[5]"},{"why":"Offers the numerical study of depinned droplets used to compare Marangoni vortex behavior and evaporative cooling trends at large contact angles.","marker":"[40]"},{"why":"Supplies experimental constant-contact-angle-mode hexane droplet data used to validate the moving-contact-line evaporation model.","marker":"[55]"},{"why":"Source of the capillary pair-force law whose simplified form in Eq. (24) governs the Cheerios interactions responsible for surrounding the small-particle core.","marker":"[45]"},{"why":"Experimental observation of size-dependent particle classification in drying bidisperse droplets, the phenomenon for which the paper proposes a flow-based mechanism.","marker":"[9]"},{"why":"Identifies the contact-line singularity that motivates the slip-length analysis in Appendix B, where the no-slip choice is load-bearing for the Marangoni vortex.","marker":"[28]"}],"fun_headline_variants":["Marangoni flow sorts particles in drying droplets","Contact-line drag blocks sorting, Marangoni unlocks it","Thermocapillary swirl separates particle sizes in droplets","Size sorting in droplets driven by Marangoni vortex","Drying droplet: Marangoni flow, not drag, sorts particles"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The size-sorting result rests on the simplified surface forces that pin particles to the interface and pull pairs of adsorbed particles together, combined with the no-slip contact-line condition; if real droplets have even modest slip at the contact line, the Marangoni vortex that does the sorting may not form.","fun_headline_variants_meta":{"raw":{"variants":["Marangoni flow sorts particles in drying droplets","Contact-line drag blocks sorting, Marangoni unlocks it","Thermocapillary swirl separates particle sizes in droplets","Size sorting in droplets driven by Marangoni vortex","Drying droplet: Marangoni flow, not drag, sorts particles"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000668,"raw_usage":{"total_tokens":3047,"prompt_tokens":946,"completion_tokens":2101,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":562,"completion_tokens_details":{"reasoning_tokens":2020}},"tokens_in":562,"tokens_out":2101,"duration_ms":14335,"temperature":1.0,"reasoning_tokens":2020,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T10:59:28.299726+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the final deposit and in-drop flow of a 60-degree contact-angle ethanol droplet containing 1.5 and 2.5 micrometer particles on a 50-degree-Celsius substrate evaporating in constant-contact-angle mode: the claim predicts a small-particle core at the apex ringed by large particles whenever the thermocapillary gradient is active, and predicts that suppressing the Marangoni vortex, for instance with a surfactant or a slip-promoting coating, removes the sorting entirely.","supporting_citations":[{"cited_title":"Analysis of an Evaporating Sessile Droplet on a Non-Wetted Surface","cited_arxiv_id":null,"evidence_quote":"Provides the prior scaling analysis claiming contact-line-to-apex flow only for contact angles above 90 degrees, which this study's all-angle result directly addresses."},{"cited_title":"Transients of Marangoni and Stefan advection dynamics during generic sessile droplet evaporation","cited_arxiv_id":null,"evidence_quote":"Offers the numerical study of depinned droplets used to compare Marangoni vortex behavior and evaporative cooling trends at large contact angles."},{"cited_title":"Capillary Forces between Spherical Particles Floating at a Liquid-Liquid Interface","cited_arxiv_id":null,"evidence_quote":"Source of the capillary pair-force law whose simplified form in Eq. (24) governs the Cheerios interactions responsible for surrounding the small-particle core."},{"cited_title":"Interaction of bi-dispersed particles with contact line in an evaporating colloidal drop","cited_arxiv_id":null,"evidence_quote":"Experimental observation of size-dependent particle classification in drying bidisperse droplets, the phenomenon for which the paper proposes a flow-based mechanism."},{"cited_title":"Hydrodynamic model of steady movement of a solid/liquid/fluid contact line","cited_arxiv_id":null,"evidence_quote":"Identifies the contact-line singularity that motivates the slip-length analysis in Appendix B, where the no-slip choice is load-bearing for the Marangoni vortex."}],"review_version":1}