{"id":"15ad9784-a151-4fd6-a8f2-cee78fe0c093","arxiv_id":"2411.13338","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A collocation method on C^s-smooth mixed-degree splines solves Poisson and biharmonic equations over multi-patch domains with low degree away from inner edges and extra-ordinary vertices, using two new sets of collocation points.","lead":"The authors modify a mixed-degree spline space so that solving Poisson's and the biharmonic equation by collocation needs high-degree functions only near inner edges and irregular vertices, cutting degrees of freedom. They introduce two new collocation point sets, Greville-type and superconvergent-type, and demonstrate the method on multiple multi-patch geometries.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Mixed degree superconvergent points are selected heuristically; if the clustering breaks superconvergence on untested geometries, the method's accuracy advantage over [29] is lost.","rationale":"The reader's weakest assumption correctly identifies the unproven superconvergence of the clustered point set as the main risk to the paper's accuracy claims. The paper's novelty and headline advantage—solving with mostly low-degree splines while retaining high-order convergence—depend on these points performing as expected. The evidence is purely numerical and limited to the tested domains, so the concern is legitimate and not a matter of disagreeing with current consensus. The basis-construction properties, while also relying on the self-cited preprint [20], are indirectly supported by the solvability and observed convergence of the collocation systems; the superconvergence claim has no analogous support. A CONDITIONAL verdict is appropriate because the method appears promising and the numerical results are consistent, but the lack of derivation or broader testing prevents full acceptance. I therefore see no reason to change the reader's verdict.","tokens_in":34519,"tokens_out":6638,"duration_ms":73616,"concrete_test":"Solve Poisson's equation with a known smooth solution on a multi-patch domain not tested in the paper, e.g., a six-patch domain with an extraordinary vertex of valency 3 and with the same mixed degree space S^{(3,5),2}_h. Compute the L2 and H1 errors using the mixed degree superconvergent points at h=1/16, 1/32, 1/64. If the L2 rate is not approximately O(h^3) (or O(h^4) on a corresponding one-patch domain), the clustering does not preserve superconvergence, and the paper's higher-order claim is not robust. Compare with the mixed degree Greville points to verify the expected gap.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central accuracy claim of the paper rests on the mixed degree superconvergent points introduced in Section 4.2. These points are obtained by taking clustered subsets of the known one-dimensional superconvergent points (14) to match the cardinality of the mixed degree Greville points. The paper provides no proof that this clustering preserves superconvergence for the tensor-product mixed-degree multi-patch spline space; it only observes the expected orders in the numerical examples (Figures 8–11). If the clustering fails on other domain configurations—e.g., different patch valencies, non-bilinear-like G^s mappings, or varying numbers of inner edges per patch—the L2 and H1 convergence orders would degrade to the Greville-order O(h^2) for both Poisson and biharmonic problems, eliminating the headline improvement over the high-degree space [29]. Moreover, the exact clustering rule is not fully specified in the text; it refers to [28, Fig. 3] and [21, Fig. 3] for the point-selection procedure, which makes the method hard to reproduce exactly and leaves the superconvergence claim entirely empirical.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops an isogeometric collocation method for the Poisson and biharmonic equations over planar multi-patch domains, using a C^s-smooth mixed-degree spline space adapted from the authors' earlier construction [20]. The space uses degree p=s+1 in most of the domain and degree p=2s+1 only near inner edges and vertices of valency greater than one. Two sets of collocation points are introduced: mixed-degree Greville points and mixed-degree superconvergent points, the latter obtained by taking clustered subsets of known one-dimensional superconvergent points. Numerical experiments on bilinear and bilinear-like G^s multi-patch domains report convergence orders under h-refinement for the L2 norm and H^m seminorms, and an L-shape two-patch example demonstrates three strategies for selecting superconvergent points, including one that yields a square linear system.","tokens_in":34709,"tokens_out":5496,"duration_ms":61531,"significance":"If the mixed-degree superconvergent points behave as claimed, the method offers a substantial reduction in degrees of freedom relative to the uniformly high-degree space of [29], while preserving or improving the convergence orders observed in earlier collocation work. The paper contains several strengths: the construction of the mixed-degree space is detailed and the five geometric variants are explicitly catalogued in Appendix A; the geometries used in the examples are fully specified, including the control points for the curved bilinear-like G^4 domain in Appendix B; and the numerical results are generated with an external manufactured solution rather than fitted to the method. The convergence rates are consistent across the tested domains and both collocation point sets.","major_comments":[{"comment":"The mixed-degree superconvergent points are constructed by selecting 'clustered subsets' of the univariate superconvergent points in (14), but the exact selection rule is not given: the text refers to [28, Fig. 3] and [21, Fig. 3] and says the redundant points are skipped 'in a clustered way' and that other cases are 'straightforward' modifications. Because the convergence orders reported in Section 5 for these points (e.g., O(h^4) in L2 for one-patch Poisson, O(h^3) for multi-patch) are the basis for the method's accuracy advantage over the Greville points, the selection procedure must be precisely described in this manuscript. Moreover, no proof or analysis is provided that the tensor-product mixed-degree multi-patch space inherits superconvergence at these clustered points; if the clustering breaks superconvergence on other geometries with different valencies or more complex G^s maps, the convergence would degrade to the Greville order O(h^2), eliminating the headline improvement. At minimum, the paper should state that this is a heuristic and provide a reproducible algorithm for the clustering.","section":"Section 4.2, Eq. (14) and Fig. 4"},{"comment":"The abstract and introduction claim that the method solves the PDEs with 'a much lower number of degrees of freedom' compared to the C^s-smooth spline space [29] with degree p=2s+1 everywhere, but no quantitative comparison is presented. The numerical section reports errors and system dimensions for the proposed method only; there is no table or figure comparing accuracy versus degrees of freedom with [29] (or with the earlier collocation papers [28,21]). Since the entire motivation of the mixed-degree construction is the reduction of degrees of freedom, a direct comparison should be included to substantiate the claim.","section":"Section 5, Tables 2 and 3, Figs. 8–11"},{"comment":"The linear independence, partition of unity, and the general dimension formula of the adapted mixed-degree underlying space are asserted with the phrase 'One can show (cf. [20])' and the formula is stated without proof. The adapted space differs from the space in [20] because the set of edges treated as 'inner' depends on the patch configuration (four, three, two adjacent, two opposite, or one inner edge), and the direct sum decomposition (8) is the basis for the collocation system size. If the dimension formula or the basis property fails for any of the five variants, the reported system dimensions in Tables 2 and 3 and the least-squares step would be invalid. The paper should either prove these properties for all five variants or explicitly identify which results in [20] carry over verbatim and which require adaptation.","section":"Section 3.1, Eq. (8) and dimension formula"}],"minor_comments":[{"comment":"Typo: 'dedicted' should be 'dedicated'.","section":"Section 4.1, first sentence"},{"comment":"The phrase 'The modifications to all other cases are straightforward and follow the concept visualized in Fig. 2' is too vague for reproducibility; please provide at least one worked example of a non-four-inner-edge case, or a reference to a precise algorithmic description.","section":"Section 4.2, text after Eq. (14)"},{"comment":"The convergence orders are reported by visual inspection of log-log plots with reference slopes. Since the orders are a central numerical claim, a table listing the estimated convergence rates for each domain, norm, and point set (or a statement of the fitted rates) would be more informative and less ambiguous.","section":"Section 5, Figures 8–11"},{"comment":"The tables give the dimensions of the linear systems but not the dimension of the spline space W^s_h; adding the latter would clarify the degree of overdetermination and allow the reader to verify the 'reduced DOFs' claim directly.","section":"Example 3, Tables 2 and 3"},{"comment":"The paper explicitly states that the extension to extraordinary vertices and the generalization of the square-system strategy are future work; this limitation should be mentioned earlier, e.g., at the end of Section 4 or in the introduction, so that the scope of the present contribution is clear from the outset.","section":"Section 6, Conclusion"},{"comment":"Reference [20] is an arXiv preprint; if the final version is published, the citation should be updated. Also, the superconvergent points in (14) for S^{3,2} and S^{5,2} are stated without derivation; adding the defining polynomials (as is done for the degree-9 case) would make the section more self-contained.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The paper is a reasonable numerical contribution to isogeometric collocation with mixed-degree splines, but the main advantage over existing work is not directly demonstrated and the superconvergent point selection is underspecified. The authors should be asked to provide a precise clustering rule, a quantitative comparison with [29], and either proofs or clear references for the basis properties of the adapted space. The topic fits the journal's scope and the numerical evidence is encouraging, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short take: this is a solid, useful paper for the isogeometric collocation community. The genuinely new pieces are the adapted C^s mixed-degree spline space, which uses high degree p=2s+1 only near inner edges and vertices with valency greater than one and low degree p=s+1 elsewhere, and the two collocation point sets built for it. That extends the authors' mixed-degree construction [20] and reduces degrees of freedom compared with the uniform high-degree space [29].\n\nThe numerical work is the paper's strong suit. They solve Poisson (s=2) and biharmonic (s=4) over bilinear domains, curved bilinear-like G^s domains, one-patch, multi-patch and L-shape domains, and report consistent h-refinement rates. The manufactured solution cos(x_1)sin(x_2) is external, so the central claims are not fitted. Control points for the more complicated geometries are given in an appendix, which helps reproducibility.\n\nThe soft spots, in proportion. First, the basis properties of the new space variants are not proved in this paper; the text defers to [20] for linear independence, dimension and partition of unity. That is acceptable if the preprint is solid, but it leaves the current paper dependent on an unreviewed reference. Second, the mixed-degree superconvergent points are selected by taking clustered subsets of known one-dimensional superconvergent points, and the clustering rule is described via references to figures in [28] and [21] rather than as a fully specified algorithm. There is no derivation that the clustered tensor-product points retain superconvergence in the multi-patch mixed-degree setting. The numerical evidence on the tested domains is consistent and nontrivial, so this is not a fatal flaw, but it is the place where a referee should push hardest. If the clustering is geometry-dependent, the advertised advantage over [29] could shrink to the Greville-point rate. Third, the DoF reduction is claimed but never shown as an error-versus-DoF comparison; such a plot would make the practical case concrete. The square-system strategy is only demonstrated for a two-patch L-shape, a limitation the authors state openly.\n\nWho this is for: numerical analysts working on isogeometric collocation and smooth multi-patch spline spaces. They will get a clear construction and a solid set of experiments. The paper deserves a serious referee. A good review can ask for a precise clustering specification, a proof or at least a much stronger empirical case for the superconvergent points, and a DoF comparison.","headline":"Solid extension of mixed-degree isogeometric collocation with genuine DoF savings; the superconvergent-point selection is heuristic and partly deferred to referenced figures, but the numerical evidence is consistent and the paper deserves peer review.","tokens_in":35274,"tokens_out":3214,"would_cite":true,"duration_ms":39193,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["65N35","65D17","68U07"],"pacs":[],"model":"deepseek-v4-flash","headline":"A modified C^s-smooth mixed degree spline space confines the high degree p=2s+1 to inner edges and extraordinary vertices, enabling isogeometric collocation for Poisson and biharmonic equations on multi-patch domains with far fewer…","keywords":["isogeometric analysis","collocation","mixed degree spline space","multi-patch domain","Greville points","superconvergent points","Poisson equation","biharmonic equation"],"falsifier":"Solve Poisson's equation on a bilinear multi-patch domain with an extraordinary vertex of valency 5, using the $C^2$-smooth space $W^2_h$ with mixed degree superconvergent points, and measure the $L^2$ error under $h$-refinement with $h=1/8,1/16,1/32,1/64$; if the rate is $O(h^2)$ rather than $O(h^3)$, the claimed superconvergence for multi-patch domains fails. Equivalently, compute the Galerkin residual $D^s(u-u_h)$ at the mixed degree superconvergent points for a known manufactured solution and compare its vanishing rate with that at the mixed degree Greville points.","tokens_in":34302,"feed_emoji":"🧩","tokens_out":15617,"duration_ms":120793,"temperature":0.7,"pith_summary":"The paper aims to make isogeometric collocation practical on planar multi-patch domains by constructing $C^s$-smooth discretization spaces that have the minimal possible degree $p=s+1$ almost everywhere, using the high degree $p=2s+1$ only in small neighborhoods of inner edges and of vertices with patch valency greater than one. It introduces two sets of collocation points for these mixed degree spaces, the mixed degree Greville points and the mixed degree superconvergent points, and uses them to solve Poisson's equation with $s=2$ and the biharmonic equation with $s=4$ in strong form. The numerical experiments on bilinear and bilinear-like $G^s$ multi-patch domains indicate the expected $h$-refinement convergence orders, with superconvergent points reaching up to $O(h^4)$ in the $L^2$ norm on one-patch domains while the space uses far fewer degrees of freedom than a uniform degree $p=2s+1$ space. A sympathetic reader would care because collocation avoids numerical integration, and the mixed degree construction attacks the main cost barrier of smooth multi-patch discretizations, which is the need for high spline degrees everywhere.","feed_headline":"Mixed-degree splines keep high order while cutting degrees of freedom","feed_subtitle":"Poisson and biharmonic collocation on multi-patch domains use degree s+1 splines except near seams.","key_machinery":"The machinery is the mixed degree underlying spline space $S^{(p_1,p_2),s}_h([0,1]^2)$, defined as the direct sum $S_1([0,1]^2) \\oplus \\bar{S}_1([0,1]^2) \\oplus S_2([0,1]^2)$, where $p_1=s+1$ and $p_2=2s+1$. The subspace $S_1$ contains degree-$p_1$ B-splines whose derivatives of order up to $s$ vanish on the boundary portions corresponding to inner edges, $S_2$ contains degree-$p_2$ B-splines with nonvanishing derivatives there, and $\\bar{S}_1$ contains truncated degree-$p_1$ B-splines added to restore completeness. The paper's adaptation restricts the high-degree part to inner edges and to vertices of valency greater than one, in contrast to the earlier mixed degree space that used high degree near all edges and vertices. On top of this space sit two sets of collocation points: mixed degree Greville points, which assign one Greville point to every basis function of the mixed degree space, and mixed degree superconvergent points, obtained by taking clustered subsets of the known univariate superconvergent points so that their cardinality matches the Greville count. The global linear systems are formed by inserting these points into the strong form of the PDE; for multi-patch domains the systems are slightly overdetermined and solved by least squares, and for a two-patch L-shape a subset selection of superconvergent points produces a square system.","core_discovery":"The paper's central claim is that the adapted $C^s$-smooth mixed degree isogeometric spline space has minimal possible degree $p=s+1$ everywhere on the multi-patch domain except in a small neighborhood of inner edges and of vertices of patch valency greater than one, where degree $p=2s+1$ is required, and that this space supports isogeometric collocation for Poisson's equation with $s=2$ and the biharmonic equation with $s=4$. The space is built from a modified mixed degree underlying spline space on the unit square, and the construction yields basis functions that are linearly independent, locally supported, nonnegative, and form a partition of unity, together with an explicit dimension formula. For the collocation points, the paper generalizes Greville points to the mixed degree setting and defines mixed degree superconvergent points via clustered subsets of known one-dimensional superconvergent points. The reported numerical results show, for Greville points, convergence orders of $O(h^{p_1-1})=O(h^2)$ in all tested norms for both equations; for superconvergent points, the observed orders rise to $O(h^{p_1+1})=O(h^4)$ in $L^2$ on one-patch domains for Poisson, to $O(h^{p_1-1})=O(h^4)$ in $L^2$, $H^1$, $H^2$ for the biharmonic equation on one-patch domains, and to $O(h^3)$ on the tested multi-patch domains in most norms. The method also extends to bilinear-like $G^s$ parameterizations, allowing curved boundaries.","pith_inferences":["If the clustered-selection premise behind the mixed degree superconvergent points fails on other geometries, the $L^2$ and $H^1$ advantages over Greville points could shrink from $O(h^3)$ back to $O(h^2)$ on multi-patch domains, so the superconvergence claim is the least certain part of the paper.","The square-system subset strategy demonstrated for one L-shape two-patch domain may generalize to multi-patch configurations with extraordinary vertices, but the paper leaves that extension open.","The same mostly-low-degree collocation idea could be applied to Kirchhoff-Love plates and shells or to multi-patch surfaces and volumes, as the authors suggest for future work.","A rigorous error analysis for collocation at mixed degree superconvergent points, rather than numerical evidence alone, would be needed to guarantee the observed convergence orders on general multi-patch domains."],"forward_implications":["Poisson's and the biharmonic equation can be solved by strong-form collocation on planar multi-patch domains using a discretization space that is mostly of degree $p=s+1$ rather than $p=2s+1$ everywhere, cutting the number of degrees of freedom substantially at equal mesh size.","The mixed degree superconvergent points give higher convergence orders than the mixed degree Greville points: up to $O(h^4)$ in the $L^2$ norm on one-patch domains for Poisson's equation, $O(h^4)$ in $L^2$, $H^1$, $H^2$ for the biharmonic equation on one-patch domains, and $O(h^3)$ on the tested multi-patch domains in most norms.","The method works for bilinear-like $G^s$ multi-patch parameterizations, so multi-patch domains with curved boundaries can be handled by the same construction.","On multi-patch domains the collocation system is slightly overdetermined and is solved by least squares; for a two-patch L-shape, a subset of superconvergent points yields a square linear system with the same observed convergence orders.","The high degree $p=2s+1$ is still required near inner edges and at vertices of patch valency greater than one, so the savings are limited to the interior of patches and near ordinary boundary vertices."],"supporting_citations":[{"why":"Supplies the original C^s-smooth mixed degree and regularity spline space that this paper modifies to restrict the high degree to inner edges and extraordinary vertices.","marker":"[20]"},{"why":"Provides the C^s-smooth spline space requiring degree 2s+1 everywhere, the baseline for the degrees-of-freedom comparison, and the gluing-function framework for edge and vertex subspaces.","marker":"[29]"},{"why":"Sets up the multi-patch isogeometric collocation method for Poisson's equation and gives the superconvergent points for S^{3,2} and S^{5,2} used in the mixed degree construction.","marker":"[28]"},{"why":"Sets up the isogeometric collocation method for the biharmonic equation and gives the superconvergent points for S^{5,4} and S^{9,4}.","marker":"[21]"},{"why":"Defines bilinear-like G^s multi-patch parameterizations and the gluing functions used to extend the method to curved boundaries.","marker":"[26]"},{"why":"Provides the characterization of superconvergent points as roots of the Galerkin residual used to compute the one-dimensional point sets.","marker":"[16]"},{"why":"Supplies the superconvergent points for spline spaces with maximal regularity used in the square-system subset selection for the two-patch L-shape example.","marker":"[33]"},{"why":"Supplies the separation of collocation points into inner and boundary sets used to handle the biharmonic equation.","marker":"[17]"}],"fun_headline_variants":["Mixed-degree splines slash degrees of freedom in isogeometric collocation","Smoother splines, fewer unknowns: collocation for Poisson and biharmonic","Low-degree splines near seams cut cost of multi-patch isogeometric PDEs","Isogeometric collocation gets efficient with mixed-degree smooth splines","Multi-patch PDEs: mixed-degree splines keep accuracy, drop degrees"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The higher convergence orders claimed for the mixed degree superconvergent points rest on the unproven premise that clustered subsets of one-dimensional superconvergent points, selected to match the mixed degree Greville cardinality, retain their superconvergence for the tensor-product mixed-degree multi-patch spaces; the paper only tests this numerically on the chosen domains.","fun_headline_variants_meta":{"raw":{"variants":["Mixed-degree splines slash degrees of freedom in isogeometric collocation","Smoother splines, fewer unknowns: collocation for Poisson and biharmonic","Low-degree splines near seams cut cost of multi-patch isogeometric PDEs","Isogeometric collocation gets efficient with mixed-degree smooth splines","Multi-patch PDEs: mixed-degree splines keep accuracy, drop degrees"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000823,"raw_usage":{"total_tokens":3708,"prompt_tokens":1159,"completion_tokens":2549,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":775,"completion_tokens_details":{"reasoning_tokens":2448}},"tokens_in":775,"tokens_out":2549,"duration_ms":17576,"temperature":1.0,"reasoning_tokens":2448,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T16:32:29.710351+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Solve Poisson's equation on a bilinear multi-patch domain with an extraordinary vertex of valency 5, using the $C^2$-smooth space $W^2_h$ with mixed degree superconvergent points, and measure the $L^2$ error under $h$-refinement with $h=1/8,1/16,1/32,1/64$; if the rate is $O(h^2)$ rather than $O(h^3)$, the claimed superconvergence for multi-patch domains fails. Equivalently, compute the Galerkin residual $D^s(u-u_h)$ at the mixed degree superconvergent points for a known manufactured solution and compare its vanishing rate with that at the mixed degree Greville points.","supporting_citations":[{"cited_title":"A C^s-smooth mixed degree and regularity isogeometric spline space over planar multi-patch domains","cited_arxiv_id":"2407.17046","evidence_quote":"Supplies the original C^s-smooth mixed degree and regularity spline space that this paper modifies to restrict the high degree to inner edges and extraordinary vertices."},{"cited_title":"Kapl and V","cited_arxiv_id":null,"evidence_quote":"Provides the C^s-smooth spline space requiring degree 2s+1 everywhere, the baseline for the degrees-of-freedom comparison, and the gluing-function framework for edge and vertex subspaces."},{"cited_title":"Kapl and V","cited_arxiv_id":null,"evidence_quote":"Sets up the multi-patch isogeometric collocation method for Poisson's equation and gives the superconvergent points for S^{3,2} and S^{5,2} used in the mixed degree construction."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Sets up the isogeometric collocation method for the biharmonic equation and gives the superconvergent points for S^{5,4} and S^{9,4}."},{"cited_title":"Kapl and V","cited_arxiv_id":null,"evidence_quote":"Defines bilinear-like G^s multi-patch parameterizations and the gluing functions used to extend the method to curved boundaries."},{"cited_title":"Gomez and L","cited_arxiv_id":null,"evidence_quote":"Provides the characterization of superconvergent points as roots of the Galerkin residual used to compute the one-dimensional point sets."},{"cited_title":"Maurin, F","cited_arxiv_id":null,"evidence_quote":"Supplies the superconvergent points for spline spaces with maximal regularity used in the square-system subset selection for the two-patch L-shape example."},{"cited_title":"Gomez, A","cited_arxiv_id":null,"evidence_quote":"Supplies the separation of collocation points into inner and boundary sets used to handle the biharmonic equation."}],"review_version":1}