pith. sign in

arxiv: 2604.19555 · v1 · submitted 2026-04-21 · 🧮 math.NA · cs.NA

Local power of approximation in hierarchical spline spaces on weakly admissible meshes

Gustavo A. Fernandez Lezcano , Eduardo M. Garau , B\'arbara Ivaniszyn This is my paper

Pith reviewed 2026-05-10 01:52 UTC · model grok-4.3

classification 🧮 math.NA cs.NA
keywords hierarchical spline spacesweakly admissible mesheslocal approximationquasi-interpolationadaptive refinementtensor-product splinesstabilityisogeometric analysis
0
0 comments X

The pith

Hierarchical spline spaces on weakly admissible meshes deliver local approximation power via stable quasi-interpolants and graded adaptive refinement.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper shows that weakly admissible hierarchical meshes, built from strictly nested cell sets that locally reproduce the full tensor-product spline space, support optimal local approximation. It does so by designing a robust adaptive algorithm that produces locally graded meshes and by proving stability plus approximation bounds for simple quasi-interpolation operators. A reader would care because these results give a clean theoretical basis for using hierarchical splines in adaptive computations without losing the approximation strength of the underlying space.

Core claim

For weakly admissible hierarchical meshes the authors prove that quasi-interpolants are stable in the local mesh size and reproduce the full approximation order of the tensor-product spline space on each active cell; the same meshes are generated by an adaptive refinement procedure that enforces local grading while preserving the nested-cell reproduction property at every level.

What carries the argument

Weakly admissible hierarchical mesh: a collection of strictly nested cell sets, each locally reproducing the full tensor-product spline space, that together define the hierarchical spline space and admit stable quasi-interpolation.

If this is right

  • The adaptive refinement algorithm produces meshes on which local approximation holds at the full spline order.
  • Quasi-interpolants remain stable with respect to the local mesh size on these meshes.
  • Numerical experiments confirm that the resulting spaces outperform standard adaptive strategies in accuracy per degree of freedom.
  • The construction extends the classical tensor-product theory to hierarchical settings without global regularity assumptions.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same mesh construction might allow similar local results for other locally supported bases such as T-splines or LR-splines.
  • In practice one could replace expensive global projectors with these quasi-interpolants inside existing adaptive IGA codes.
  • Error estimates derived here could be combined with a posteriori indicators to drive refinement in time-dependent or nonlinear problems.

Load-bearing premise

The cell sets must be strictly nested and must reproduce the complete tensor-product spline space locally at each level.

What would settle it

A concrete counter-example would be a sequence of weakly admissible meshes on which the quasi-interpolant applied to a smooth test function fails to achieve the expected local convergence rate equal to the spline degree plus one.

Figures

Figures reproduced from arXiv: 2604.19555 by B\'arbara Ivaniszyn, Eduardo M. Garau, Gustavo A. Fernandez Lezcano.

Figure 1
Figure 1. Figure 1: ). Ω0 Ω0 [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Neighborhood NQ of a cell Q, distinguishing the refined region Qb from the highlighted area NQ. The set NQ for a cell Q ∈ Qℓ was previously considered in [Bracco et al., 2018, Definition 5]. This neighborhood of Q will be highly useful in the following section, and its properties are fundamental for characterizing weakly admissible meshes. Proposition 2.14. If C ∈ B∞ p,ℓ−1 , then there exist unique cells c… view at source ↗
Figure 3
Figure 3. Figure 3: Dyadically refining a set C ∈ B∞ p,ℓ−1 (on the left) we obtain 4 cells Q of level ℓ (highlighted on the right) that have the same NQ, which coincides with C, and moreover, the union of their support extensions equals C. To prove (iii), suppose that C ∈ F ∞ p,ℓ−1 . If Q1, Q2, Q3, Q4 ⊂ Ω0, the result follows from (ii). Otherwise, we can write: C ∩ Ω0 =     [ Qi∈core(C), Qi⊂Ω0 Qci ∩ Ω0     ∪     [… view at source ↗
Figure 4
Figure 4. Figure 4: The possible configurations for the cells {Qi} 4 i=1 ⊂ Q∞ ℓ that satisfy (ii) in Proposition 2.14 for which some Qi is not contained in Ω0. Proposition 2.15. If Q ∈ Q∞ ℓ , then parent \(Q) = S {NQ∗ | Q∗ ∈ Q∞ ℓ , Q∗ ⊂ Qb}. Proof. Let Q ∈ Q∞ ℓ . First, note that parent \(Q) is the unique set formed by (2p+1)×(2p+1) cells of level ℓ − 1 centered at the cell parent(Q). On the other hand, by considering Qb, let… view at source ↗
Figure 5
Figure 5. Figure 5: The set parent \Q is represented as the union of cells of level ℓ − 1 (left) and level ℓ (right), where Q denotes the cell of level ℓ shown in yellow. In the right figure, the set Qb is depicted in light gray, while the set A, defined as the union of support extensions of cells of level ℓ in Qb, is shown in dark gray. We complete te proof by noticing that this last set consists of (4p + 2) × (4p + 2) cells… view at source ↗
Figure 6
Figure 6. Figure 6: Left: A three-level WAHM with p = (3, 3) containing an isolated deactivated cell of level 1. Middle and right: Meshes with p = (2, 2). The middle mesh arises from a clustered hierarchy because Ω1 is associated to a p-form, whereas in the right mesh, Ω1 is merely a union of cells of level 0. Isolated areas inside Ωℓ that do not cover the support of at least one B-spline of level ℓ − 1 result in a hierarchic… view at source ↗
Figure 7
Figure 7. Figure 7: Considering p = (3, 3), the initial mesh, on the left, is weakly admissible and contains a marked cell with optimal approximation power. The central figure indicates the cells that must be refined so that the initially marked cell increases its approximation power in the new mesh while ensuring that the mesh remains weakly admissible. Finally, the figure on the right shows the resulting mesh after this ref… view at source ↗
Figure 8
Figure 8. Figure 8: Considering p = (3, 3), the initial mesh, on the left, contains a cell with suboptimal approximation power. The central figure shows the cells that must be refined to increase the approximation power of the marked cell and preserve the weakly admissible property in the new mesh. The figure on the right presents the mesh obtained after this refinement. MSUBOPT (see [PITH_FULL_IMAGE:figures/full_fig_p013_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Cutaway of the mesh at levels n − 1 and n for p = (3, 3). Left: A marked cell Q′ ∈ MOPT n−1 at level n − 1 with its children W∗ n. The set of cells of level n − 1 contained in NQ ∩ Ω0 for each Q ∈ W∗ n is also depicted. Right: After refinement, the cells of W∗ n are successfully included in ω ∗ n. Remark 3.7. Using Method 3.6 yields the following properties for all valid ℓ: (i) Ωℓ ⊂ Ω ∗ ℓ and ωℓ ⊂ ω ∗ ℓ . … view at source ↗
Figure 10
Figure 10. Figure 10: Initial hierarchical mesh and refined meshes obtained via the WA and SA2 methods for Example 5.1. The shaded region indicates the initially marked elements where an improvement of the approximation is sought. The first row of [PITH_FULL_IMAGE:figures/full_fig_p025_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: A initial hierarchical mesh and refined meshes obtained via the WA and SA2 methods for Example 5.2. The shaded region indicates the initially marked elements where an improvement of the approximation is sought. Method DOFs L 2 (M)-error Ieff Ierr – 457 1.47 × 10−5 – – WA 853 7.77 × 10−7 4.71 1.27 SA2 529 4.56 × 10−6 7.99 0.50 (a) Corresponding to the mesh layout shown in [PITH_FULL_IMAGE:figures/full_fig… view at source ↗
Figure 12
Figure 12. Figure 12: A initial hierarchical mesh and refined meshes obtained via the WA and SA2 methods for Example 5.2. The shaded region indicates the initially marked elements where an improvement of the approximation is sought. 5.2 Experimental analysis of the full adaptive cycle Let us consider the Poisson equation with Dirichlet boundary conditions given by ( −∆u = f in Ω u = g on ∂Ω (16) where f and g are given and Ω i… view at source ↗
Figure 13
Figure 13. Figure 13: Decay of the energy norm error versus degrees of freedom for the WA and SA2 methods with element-based error estimators, and for the SA2 method with function-based error estimators. Independent runs of 8 iterations were performed for both methods, yielding an optimal experimental order of convergence. As illustrated in [PITH_FULL_IMAGE:figures/full_fig_p027_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: displays the meshes obtained with both procedures at the iteration where they reach a similar error: iteration 5 for WA and iteration 8 for SA2. We can observe that in the hierarchical mesh obtained with the SA2 method, there are finest-level cells that are somewhat isolated and do not significantly enrich the space. Conversely, with the WA method, all hierarchical subdomains Ωℓ are unions of B-spline sup… view at source ↗
read the original abstract

We study local approximation properties in hierarchical spline spaces through a twofold approach. First, we design and analyze a robust adaptive refinement algorithm to construct locally graded meshes. Second, we establish rigorous stability and approximation results using computationally efficient quasi-interpolation operators. The primary contribution is the analysis of weakly admissible hierarchical meshes. Our framework relies on strictly nested cell sets that locally reproduce the full tensor-product spline space at each level. Theoretical and numerical results demonstrate that this intuitive approach is mathematically elegant and outperforms existing adaptive refinement strategies in various practical scenarios.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript develops an adaptive refinement algorithm to generate locally graded weakly admissible hierarchical meshes for spline spaces and proves local stability and approximation properties via quasi-interpolation operators. It asserts that the framework relies on strictly nested cell sets that locally reproduce the full tensor-product spline space at each level, yielding results independent of hierarchy depth, and claims both theoretical elegance and numerical superiority over existing adaptive strategies.

Significance. If the stability constants and approximation orders hold under the stated mesh assumptions, the work could meaningfully advance adaptive isogeometric analysis by permitting more flexible local grading than standard admissibility while preserving efficient quasi-interpolants. The emphasis on computationally efficient operators is a constructive feature.

major comments (2)
  1. [Abstract and framework description] Abstract and framework description: The stability and approximation results rest on the requirement that weakly admissible meshes guarantee strictly nested cell sets with full local tensor-product spline reproduction at each level (essential for quasi-interpolant stability constants independent of depth). The definition of 'weakly admissible' is not shown to enforce this reproduction property automatically, and the adaptive algorithm description does not explicitly verify or enforce it for all generated configurations; without this, the central error estimates may fail to hold.
  2. [Numerical results] Numerical results: The claim of outperformance over existing adaptive refinement strategies is asserted but lacks specific quantitative support (e.g., tabulated error norms, degrees of freedom, or direct comparisons on benchmark problems) that would confirm the advantage across the stated practical scenarios.
minor comments (2)
  1. [Preliminaries] Notation for hierarchical levels and cell sets could be introduced more explicitly in the preliminaries to aid readability.
  2. [References] Ensure the reference list includes all standard works on admissible hierarchical splines for proper context.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and constructive feedback on our manuscript. The comments help clarify key aspects of the framework and strengthen the presentation of the numerical evidence. We address each major comment below.

read point-by-point responses

  1. Referee: [Abstract and framework description] The stability and approximation results rest on the requirement that weakly admissible meshes guarantee strictly nested cell sets with full local tensor-product spline reproduction at each level (essential for quasi-interpolant stability constants independent of depth). The definition of 'weakly admissible' is not shown to enforce this reproduction property automatically, and the adaptive algorithm description does not explicitly verify or enforce it for all generated configurations; without this, the central error estimates may fail to hold.

    Authors: We agree that an explicit link between the definition of weakly admissible meshes and the strictly nested reproduction property strengthens the exposition. The manuscript states that the framework relies on strictly nested cell sets that locally reproduce the full tensor-product spline space at each level, and the adaptive algorithm is constructed to preserve this structure at every step. To address the concern directly, we will insert a short lemma in the revised version proving that every mesh generated by the algorithm satisfies the full local reproduction property, thereby confirming that the quasi-interpolant stability constants remain independent of hierarchy depth. revision: yes

  2. Referee: [Numerical results] The claim of outperformance over existing adaptive refinement strategies is asserted but lacks specific quantitative support (e.g., tabulated error norms, degrees of freedom, or direct comparisons on benchmark problems) that would confirm the advantage across the stated practical scenarios.

    Authors: We acknowledge that the numerical section would benefit from more explicit quantitative comparisons. The current examples illustrate the advantages in mesh grading and approximation efficiency, yet we agree that tabulated data would make the outperformance claim more concrete. In the revised manuscript we will add tables reporting error norms, degrees of freedom, and direct side-by-side comparisons against standard adaptive strategies on the benchmark problems already considered, together with a brief discussion of the observed gains. revision: yes

Circularity Check

0 steps flagged

No circularity detected; derivation self-contained

full rationale

The abstract states the framework relies on strictly nested cell sets reproducing the full tensor-product spline space, then derives stability and approximation results via quasi-interpolants for weakly admissible meshes. No quoted equations or steps reduce a prediction to a fitted input by construction, invoke self-citations as load-bearing uniqueness theorems, or smuggle ansatzes. The central claims rest on explicit assumptions about the mesh hierarchy and standard spline properties, without the derivation looping back to its own inputs. This matches the default case of an independent theoretical development.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claims rest on standard spline theory plus the paper-specific assumption of strictly nested cells reproducing tensor-product spaces locally; no free parameters or invented entities are evident from the abstract.

axioms (1)
  • domain assumption Strictly nested cell sets locally reproduce the full tensor-product spline space at each level.
    Explicitly stated as the basis of the framework in the abstract.

pith-pipeline@v0.9.0 · 5387 in / 1065 out tokens · 30944 ms · 2026-05-10T01:52:42.278432+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

20 extracted references · 20 canonical work pages

  1. [1]

    Bracco, C., Giannelli, C., and V \'a zquez, R. (2018). Refinement algorithms for adaptive isogeometric methods with hierarchical splines. Axioms , 7(3):43

    work page 2018

  2. [2]

    and Scott, R

    Brenner, S. and Scott, R. (2002). The Mathematical Theory of Finite Element Methods . Springer, 3rd edition

    work page 2002

  3. [3]

    and Garau, E

    Buffa, A. and Garau, E. M. (2017). Refinable spaces and local approximation estimates for hierarchical splines. IMA J. Numer. Anal. , 37(3):1125--1149

    work page 2017

  4. [4]

    and Garau, E

    Buffa, A. and Garau, E. M. (2018). A posteriori error estimators for hierarchical B -spline discretizations. Math. Models Methods Appl. Sci. , 28(8):1453--1480

    work page 2018

  5. [5]

    and Giannelli, C

    Buffa, A. and Giannelli, C. (2016). Adaptive isogeometric methods with hierarchical splines: Error estimator and convergence. Math. Models Methods Appl. Sci. , 26(01):1--25

    work page 2016

  6. [6]

    Buffa, A., Giannelli, C., Morgenstern, P., and Peterseim, D. (2016). Complexity of hierarchical refinement for a class of admissible mesh configurations. Comput. Aided Geom. Design , 47:83--92

    work page 2016

  7. [7]

    Gantner, G., Haberlik, D., and Praetorius, D. (2017). Adaptive IGAFEM with optimal convergence rates: hierarchical B -splines. Math. Models Methods Appl. Sci. , 27(14):2631--2674

    work page 2017

  8. [8]

    Giannelli, C., J \"u ttler, B., and Speleers, H. (2012). THB -splines: The truncated basis for hierarchical splines. Comput. Aided Geom. Design. , 29(7):485 -- 498

    work page 2012

  9. [9]

    Giannelli, C., J \"u ttler, B., and Speleers, H. (2014). Strongly stable bases for adaptively refined multilevel spline spaces. Adv. Comput. Math. , 40(2):459--490

    work page 2014

  10. [10]

    Kraft, R. (1997). Adaptive and linearly independent multilevel B -splines. In Surface fitting and multiresolution methods ( C hamonix-- M ont- B lanc, 1996) , pages 209--218. Vanderbilt Univ. Press, Nashville, TN

    work page 1997

  11. [11]

    a ngige multilevel B-Splines und ihre Anwendungen . PhD thesis, Universit \

    Kraft, R. (1998). Adaptive und linear unabh \"a ngige multilevel B-Splines und ihre Anwendungen . PhD thesis, Universit \"a t Stuttgart

    work page 1998

  12. [12]

    Lee, B.-G., Lyche, T., and M rken, K. (2001). Some examples of quasi-interpolants constructed from local spline projectors. In Mathematical methods for curves and surfaces ( O slo, 2000) , Innov. Appl. Math., pages 243--252. Vanderbilt Univ. Press, Nashville, TN

    work page 2001

  13. [13]

    Lyche, T., Manni, C., and Speleers, H. (2018). Foundations of spline theory: B-splines, spline approximation, and hierarchical refinement. In Lyche, T., Manni, C., and Speleers, H., editors, Splines and PDEs: From Approximation Theory to Numerical Linear Algebra , volume 2219 of Lecture Notes in Mathematics . Springer, Cham

    work page 2018

  14. [14]

    H., Siebert, K

    Nochetto, R. H., Siebert, K. G., and Veeser, A. (2009). Theory of adaptive finite element methods: an introduction. In Multiscale, nonlinear and adaptive approximation , pages 409--542. Springer, Berlin

    work page 2009

  15. [15]

    Nochetto, R. H. and Veeser, A. (2012). Primer of adaptive finite element methods. In Multiscale and adaptivity: modeling, numerics and applications , volume 2040 of Lecture Notes in Math. , pages 125--225. Springer, Heidelberg

    work page 2012

  16. [16]

    Schumaker, L. L. (2007). Spline functions: basic theory . Cambridge Mathematical Library. Cambridge University Press, Cambridge, third edition

    work page 2007

  17. [17]

    Speleers, H. (2017). Hierarchical spline spaces: quasi-interpolants and local approximation estimates. Advances in Computational Mathematics , 43:235--255

    work page 2017

  18. [18]

    and Manni, C

    Speleers, H. and Manni, C. (2015). Effortless quasi-interpolation in hierarchical spaces. Numer. Math. , pages 1--30

    work page 2015

  19. [19]

    V\'azquez, R. (2016). A new design for the implementation of isogeometric analysis in O ctave and M atlab: G eo PDE s 3.0. Comput. Math. Appl. , 72(3):523--554

    work page 2016

  20. [20]

    Vuong, A.-V., Giannelli, C., J\"uttler, B., and Simeon, B. (2011). A hierarchical approach to adaptive local refinement in isogeometric analysis. Comput. Methods Appl. Mech. Engrg. , 200(49-52):3554--3567

    work page 2011