{"id":"4cede3b6-ec9f-4326-b337-50a349698834","arxiv_id":"1908.07166","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A Brownian dynamics algorithm uses triangle meshes and a velocity-folding rule to simulate particles on complex curved surfaces, benchmarked on flat and spherical geometries.","lead":"This paper describes a simulation method that lets colloidal particles move, diffuse, and assemble on surfaces of nearly arbitrary shape, including spheres, tori, and knots. It approximates the surface by a triangle mesh and folds particle directions at triangle edges, which could make it possible to study how curvature and topology shape particle behavior in biology and materials.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Velocity-folding correctness for Brownian displacements is assumed, not established; no mesh/Δt convergence study supports the central claim on complex surfaces.","rationale":"The reader's weakest_assumption isolates the velocity-folding step as the load-bearing premise, and I agree. The paper's physical benchmarks (flat plane, sphere) are consistent with theory, but they only probe a constant-curvature surface at a single resolution; they do not establish convergence of the folded random walk to Laplace–Beltrami diffusion on general meshes. The deterministic geodesic rationale is insufficient because Brownian increments are random and the surface curvature on a mesh is concentrated at vertices—an issue the manuscript never discusses. The torus and knot demonstrations, while visually compelling, lack any error estimate or mesh-refinement study, so the claimed 'captures well' is not yet quantitatively supported for non-trivial topology. This is not an internal contradiction; the scheme is plausible and likely correct via stochastic development, but the evidence is incomplete. Therefore the appropriate verdict remains conditional on a convergence test. My concrete test directly targets the missing evidence: if the sphere error decreases with mesh refinement and the torus matches an independent reference, the central claim is secured; otherwise the folding approximation needs revision.","tokens_in":13642,"tokens_out":7938,"duration_ms":92156,"concrete_test":"Perform a numerical convergence study on a surface with non-constant curvature and a known reference solution. (1) On the unit sphere, compute the L2 error of the angular displacement distribution against the spectral solution (Eq. 19) and the MSAD for at least three mesh resolutions (h = 0.05, 0.02, 0.01 a) and three time steps (Δt = 10^-3, 10^-4, 10^-5 s). If the error does not decrease (ideally O(h) + O(Δt)) as h,Δt→0, the velocity-folding scheme is biased. (2) On a torus with aspect ratio ~5:1, compare the single-particle displacement distribution and MSD against an independent high-accuracy reference (e.g., spectral solution of the torus Laplacian or a fine-mesh finite-element Laplace–Beltrami discretization). If the folded-mesh results converge to the reference, the defect statistics in Fig. 6 are likely not a mesh artifact.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—that the algorithm correctly simulates BD on arbitrary mesh-approximated surfaces—rests on the unproven assumption that the velocity-folding update (Algorithm step 3, §'Algorithm') yields the Laplace–Beltrami diffusion in the continuum limit. The paper justifies folding only by deterministic geodesic motion: 'The critical rationale behind this velocity folding is the fact that a particle on the surface under no force moves along geodesic...' and validates with a single sphere benchmark at one mesh resolution (face length ~0.018a) and one Δt. For a stochastic displacement, folding is a parallel-transport operation—the tangent increment is rotated by the dihedral angle when crossing an edge. This construction plausibly gives the correct Stratonovich development of BM on the piecewise-flat surface, but the paper neither proves it nor addresses the Ito metric drift or the holonomy/angle-defect at vertices (the support of curvature on a mesh). Consequently, the torus/knot results—especially the defect statistics in Fig. 6—could be contaminated by unresolved discretization bias. The claim 'captures well' is thus supported only for the constant-curvature sphere at a fixed resolution; extrapolation to 'extremely complex surfaces' is an assertion, not a demonstrated convergence.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a Brownian dynamics (BD) simulation algorithm for particles constrained to curved surfaces represented by triangle meshes. The method computes unconstrained forces and velocities in global coordinates, projects them onto the local tangent plane, and updates particle positions in local barycentric coordinates via an Euler–Maruyama scheme. When a particle crosses a triangle edge, the velocity is 'folded' across the edge, a procedure the authors justify by analogy with deterministic geodesic motion on polyhedral surfaces. The algorithm is benchmarked against analytical solutions for diffusion on a flat plane (2D Gaussian, Eq. (17)) and on a sphere (Legendre expansion, Eq. (19)), with good agreement, and then demonstrated on single- and multi-particle dynamics on torus and knot surfaces, including MSD analysis, field-driven transport, and curvature-induced crystallization with topological defects.","tokens_in":13859,"tokens_out":3605,"duration_ms":37984,"significance":"If the algorithm is correct, it fills a practical gap: existing BD methods for curved surfaces require analytic parametrizations, while the mesh-based hybrid global/local scheme could handle arbitrarily complex geometries, with a C++ implementation made publicly available. The benchmarks are genuinely independent of the algorithm's internal assumptions, so the central results are not circular. The paper also demonstrates potential applications to crystallization and defect formation on non-trivial topologies. However, the central stochastic-transport premise—that velocity folding produces correct Brownian dynamics on the polyhedral surface in the continuum limit—is not rigorously established, and the validation lacks a convergence study. The significance therefore rests on a plausible but partially supported algorithmic claim.","major_comments":[{"comment":"No mesh-convergence or time-step-convergence study is presented. The flat and sphere benchmarks are each run at a single mesh resolution (face length ~0.05a and ~0.018a, respectively) and a single integration time step (Δt = 0.0001 s), and agreement with theory is assessed visually rather than by a quantitative error norm with error bars. Since the abstract claims the method 'captures well' diffusion on extremely complex surfaces, the absence of a resolution/convergence study leaves the central claim under-supported: the torus and knot results in Figs. 4–6 could be subject to uncontrolled discretization bias.","section":"Methods and Results (spherical surface)"},{"comment":"The velocity-folding rule is justified only for deterministic geodesic motion, and its only quantitative validation is the authors' own supplementary geodesic experiment (Fig. S1), which is a self-referential test of the same folding rule. For Brownian displacements, the step is a stochastic parallel-transport operation on a piecewise-flat surface, and the paper does not prove or numerically test that the folded random walk converges to Laplace–Beltrami diffusion as the mesh is refined. In particular, the algorithm does not specify how to treat a particle that reaches a vertex (where the angle defect concentrates curvature and holonomy), nor does it discuss potential Stratonovich/Ito drift corrections on the mesh. This is a load-bearing gap in the stochastic-transport premise.","section":"Algorithm, velocity-folding paragraph"},{"comment":"The torus and knot simulations are not benchmarked against any independent analytical or high-resolution reference solution. The MSAD plateaus in Fig. 4(c,d) and the defect statistics in Fig. 6 are presented as physical results, but without a mesh-resolution or Δt-convergence check on those surfaces they remain qualitative demonstrations. The paper should either provide such checks or explicitly frame these as illustrative, pending validation.","section":"Results: single particle dynamics on complex surfaces"}],"minor_comments":[{"comment":"The spherical displacement distribution is compared with 'Eq. (18)' in both the text and the Fig. 3(c) caption, but Eq. (18) is the flat-plane radial distribution; the correct reference is the Legendre expansion in Eq. (19). Please correct this typo.","section":"Results: spherical surface and Fig. 3(c)"},{"comment":"The variance specification for the Brownian force in Eq. (11) is missing explicit ensemble-averaging brackets and is hard to parse; please rewrite with standard notation such as ⟨F_i^B'(t) F_i^B'(t')⟩.","section":"Eqs. (10)–(11)"},{"comment":"The definition of mean-square angular displacement (MSAD) in Eq. (20) is garbled in the typeset version; please rewrite it in a clear form and define all quantities (e.g., whether r(t) is the lab-coordinate position vector and what the inner product is normalized by).","section":"Eq. (20)"},{"comment":"The caption for Fig. 6 is confusing: it refers to '(b, c) Pair distribution function' but then discusses panels (d) and (e); please renumber the figure panels and rewrite the caption so that each panel is clearly identified.","section":"Fig. 6 caption"},{"comment":"The pseudocode does not explicitly state that when a particle crosses an edge within a time step, the remaining part of the interval is used to continue integration on the adjacent face. This event-driven sub-stepping is described in the text but should appear in the pseudocode for reproducibility.","section":"Algorithm pseudocode"},{"comment":"The mesh statistics are presented as an unnumbered text list; converting this into a proper table would improve readability.","section":"Methods, mesh information"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the scope of the journal and the code availability is a plus. The main concern is that the central algorithmic claim (velocity folding yields correct Brownian dynamics on polyhedral surfaces) is not backed by a convergence study or a stochastic analysis, and the only validation of the folding rule is the authors' own supplementary experiment. I would urge the editor to require a convergence study on at least one non-trivial surface and a quantitative benchmark (e.g., an independent spectral or geodesic random walk solution) before publication. The Eq. (18)/(19) mix-up in the spherical benchmark should also be corrected, since it currently undermines the credibility of the benchmark description."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my read on Yang and Li. The genuinely new thing is the hybrid scheme: forces and velocities in global coordinates, positions updated in local barycentric coordinates on a triangle mesh, with velocity folding at edges. That combination isn't in the BD literature they cite—Castro-Villarreal et al. and Apaza/Sandoval stick to analytic local coordinates, and Paquay/Kusters use projection MD. As an enabling tool for experimentally reconstructed surfaces, this is a real step forward, and the flat and sphere benchmarks against analytic solutions are clean. The code is on GitHub, and the descriptions are concrete enough to reimplement.\n\nThe soft spot is exactly what the stress-test note flags. The velocity-folding rule is justified by deterministic geodesic motion, but the update is stochastic. Folding a random tangent displacement across an edge is plausible as a parallel transport operation on a piecewise-flat surface, but the paper doesn't prove it converges to Laplace–Beltrami diffusion, doesn't discuss the Ito drift or the holonomy at vertices, and validates it only on a single sphere at one mesh resolution and one time step. There are no error bars, no mesh-refinement or Δt-convergence study. So the torus and knot simulations, especially the defect statistics in Fig. 6, are illustrations rather than quantitatively validated predictions. I don't think this sinks the paper—the algorithm is likely correct in the continuum limit—but the claim that it 'captures well' diffusion on extremely complex surfaces is not yet backed by evidence.\n\nMinor issues: the text refers to Eq. (18) for the sphere benchmark when it means Eq. (19), and there is no discussion of how the time step or face size should be chosen in practice. The supplementary geodesic test is self-referential and doesn't address stochastic transport.\n\nWho is this for? Researchers who need particle simulations on realistic curved surfaces—soft matter, biophysics, colloidal assembly on complex substrates. It deserves a serious referee: the idea is novel and the benchmarks give a solid foundation. I would ask for a convergence study on a non-constant-curvature surface, error bars on the sphere MSAD, a comment on vertex/edge singularities, and a commit-hashed code snapshot before publication. With those, this becomes a widely used method paper.","headline":"A practical and novel mesh-based Brownian dynamics algorithm whose flat/sphere benchmarks are solid, but whose central velocity-folding step lacks convergence evidence on complex surfaces.","tokens_in":657,"tokens_out":1668,"would_cite":true,"duration_ms":29458,"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":"This paper claims that Brownian dynamics on curved surfaces can be simulated by triangulating the surface and folding velocities at triangle edges.","keywords":["Brownian dynamics","curved surfaces","triangle mesh","velocity folding","colloidal assembly","surface diffusion","topological defects","geodesic motion"],"falsifier":"Run the algorithm on a spherical surface at several mesh refinements and compare the mean-square angular displacement to the exact short-time law $4Dt/R^2$ and the long-time plateau $(\\pi^2-4)/2$. If the folded random walk does not converge to these limits as the mesh is refined, velocity folding is introducing a spurious stochastic drift.","tokens_in":13397,"feed_emoji":"🔬","tokens_out":8555,"duration_ms":83396,"temperature":0.7,"pith_summary":"Brownian dynamics of colloidal particles has largely been confined to surfaces that can be written down analytically, such as spheres, cylinders, and ellipsoids. This paper claims that the restriction is unnecessary: any curved surface, however complex, can be approximated by a triangle mesh, and a Brownian simulation can be run directly on that mesh. The key move is to compute forces and velocities in global coordinates but update positions in the local barycentric frame of each triangle, folding the velocity at edges so particles continue along geodesic paths. Benchmarks on a flat plane and a sphere match analytic diffusion predictions, and demonstrations on a torus and a knot show free diffusion, field-driven transport, and crystallization into hexagonal packings with pentagon-heptagon defects. If the folding step is correct, the algorithm opens arbitrary biological, chemical, and engineered geometries to particle-scale simulation.","feed_headline":"One folding rule extends Brownian dynamics to any curved surface","feed_subtitle":"Triangulate the surface, compute forces globally, update positions locally, and diffusion works on tori and knots.","key_machinery":"The central object is the triangle-mesh surface, with particle positions stored both in 3D lab coordinates and in per-triangle barycentric coordinates tied to the Jacobian matrix $J$ and its pseudo-inverse $J^*$. The central mechanism is velocity folding: whenever a particle's local update would cross an edge, the remaining velocity is rotated about that edge into the neighboring triangle, as though the two faces were unfolded into a plane, preserving speed while changing direction. This is what lets the mesh stand in for a smooth surface: the particle always remains on the mesh without constraint forces, and constant-velocity motion plus edge folding is argued to reproduce geodesic motion. The projection operator $P = I - nn^T$ keeps velocities tangent, and the edge-hitting times in Eq. (13) decide which face to fold into.","core_discovery":"The paper establishes a practical algorithm: triangulate a curved surface; at each step compute the unconstrained velocity from deterministic and Brownian forces in 3D lab coordinates; project it onto the triangle's tangent plane; convert to barycentric local coordinates; advance the barycentric position; and when the step would cross an edge, fold the remaining velocity into the neighboring triangle as if the two triangles were hinged flat, preserving speed and rotating the direction. This velocity-folding device keeps particles on the mesh without constraint forces and makes the random walk follow geodesic paths across the mesh. It verifies the device by reproducing the exact Gaussian distribution on a flat mesh, and on a sphere by matching the Legendre-polynomial diffusion kernel and the mean-square angular displacement in both short- and long-time limits. On tori and knots it shows MSD plateaus reflecting global geometry and topology, field-biased transport, and crystalline packing with topological defects.","pith_inferences":["Because velocity folding treats a random tangent velocity exactly like a deterministic one, the same algorithm should work for self-propelled active particles; the paper lists active systems as an outlook but does not test them.","A natural open question is convergence: on a fixed smooth surface, refining the mesh should drive the folded walk toward true Laplace-Beltrami diffusion, and the rate of that convergence can be measured from the sphere benchmarks.","The similar confinement plateaus on the much larger knot and smaller torus suggest that the plateau MSD is controlled by intrinsic geodesic diameter or graph structure of the mesh, not by Euclidean size; this could be tested by varying tube geometry at fixed area.","Because the algorithm is built from pre-computable rotation and projection matrices, it is straightforward to parallelize across particles and could be embedded in existing mesh-processing pipelines for organ- or scaffold-scale geometries."],"forward_implications":["Any surface representable as a triangle mesh becomes a valid arena for Brownian dynamics, with no analytic parametrization required.","On flat and spherical meshes the simulated displacement distributions and mean-square angular displacements match closed-form theory, so the method reproduces known diffusion in those limits.","On tori and knots, single-particle MSD shows the expected short-time $4Dt$ behavior, curvature-induced softening, and a long-time plateau set by the surface's global geometry and topology.","With a uniform external field, particles descend and equilibrate at the bottom of a torus or knot, and their descent path is sensitively controlled by Brownian fluctuations.","Multi-particle simulations with depletion attraction produce hexagonally packed crystals whose $g(r)$ peaks broaden with curvature, and pentagon-heptagon defects appear as required by the surface topology."],"supporting_citations":[{"why":"Supplies the Brownian dynamics formulation that the paper adapts to curved surfaces.","marker":"[22]"},{"why":"Supplies the local-coordinate BD scheme on curved manifolds and the spherical analytic benchmarks used for verification.","marker":"[18]"},{"why":"Represents a prior method for molecular dynamics on curved surfaces that is limited to parameterizable geometries.","marker":"[17]"},{"why":"Provides a local-coordinate active-matter scheme that motivates the hybrid global/local update strategy.","marker":"[21]"},{"why":"Gives the constrained Brownian dynamics equations and constraint-force treatment that the paper projects onto tangent planes.","marker":"[34]"},{"why":"Supplies the Euler-Maruyama integration scheme used for the local position updates.","marker":"[35]"},{"why":"Provides the classical-mechanics geodesic rationale for the velocity-folding rule.","marker":"[29]"},{"why":"Supplies the depletion interaction potentials used in the multi-particle crystallization simulations.","marker":"[37]"},{"why":"Provides the flat-plane compressed crystal reference used for the pair-correlation comparison.","marker":"[41]"}],"fun_headline_variants":["Velocity folding: Brownian dynamics on any triangular mesh","Fold the velocity, run Brownian dynamics on complex surfaces","From tori to knots: mesh-based Brownian dynamics","Fold and go: Brownian dynamics on arbitrary surfaces"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The method's accuracy rests on the assumption that folding a particle's random tangent step across triangle edges reproduces the same diffusion as on the smooth curved surface; the paper supports this with benchmarks but gives no proof of convergence.","fun_headline_variants_meta":{"raw":{"variants":["Velocity folding: Brownian dynamics on any triangular mesh","Fold the velocity, run Brownian dynamics on complex surfaces","From tori to knots: mesh-based Brownian dynamics","Fold and go: Brownian dynamics on arbitrary surfaces"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001522,"raw_usage":{"total_tokens":6075,"prompt_tokens":900,"completion_tokens":5175,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":516,"completion_tokens_details":{"reasoning_tokens":5110}},"tokens_in":516,"tokens_out":5175,"duration_ms":32760,"temperature":1.0,"reasoning_tokens":5110,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:26:19.281311+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the algorithm on a spherical surface at several mesh refinements and compare the mean-square angular displacement to the exact short-time law $4Dt/R^2$ and the long-time plateau $(\\pi^2-4)/2$. If the folded random walk does not converge to these limits as the mesh is refined, velocity folding is introducing a spurious stochastic drift.","supporting_citations":[{"cited_title":"L.; McCammon, J","cited_arxiv_id":null,"evidence_quote":"Supplies the Brownian dynamics formulation that the paper adapts to curved surfaces."},{"cited_title":"M.; Castañeda-Priego, R.; Estrada- 27 Jiménez, S., A Brownian dynamics algorithm for colloids in curved manifolds","cited_arxiv_id":null,"evidence_quote":"Supplies the local-coordinate BD scheme on curved manifolds and the spherical analytic benchmarks used for verification."},{"cited_title":"Biophysical journal 2016, 110 (6), 1226-1233","cited_arxiv_id":null,"evidence_quote":"Represents a prior method for molecular dynamics on curved surfaces that is limited to parameterizable geometries."},{"cited_title":"Soft Matter 2018, 14 (48), 9928- 9936","cited_arxiv_id":null,"evidence_quote":"Provides a local-coordinate active-matter scheme that motivates the hybrid global/local update strategy."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the constrained Brownian dynamics equations and constraint-force treatment that the paper projects onto tangent planes."},{"cited_title":"SIAM Review 2001, 43 (3), 525-546","cited_arxiv_id":null,"evidence_quote":"Supplies the Euler-Maruyama integration scheme used for the local position updates."},{"cited_title":"I., Mathematical methods of classical mechanics","cited_arxiv_id":null,"evidence_quote":"Provides the classical-mechanics geodesic rationale for the velocity-folding rule."},{"cited_title":"D.; Bevan, M","cited_arxiv_id":null,"evidence_quote":"Supplies the depletion interaction potentials used in the multi-particle crystallization simulations."},{"cited_title":"M.; Bevan, M","cited_arxiv_id":null,"evidence_quote":"Provides the flat-plane compressed crystal reference used for the pair-correlation comparison."}],"review_version":1}