Pith. sign in

REVIEW 6 minor 32 references

Rust Implementation of Finite Element Exterior Calculus on Coordinate-Free Simplicial Complexes

T0 review · 0 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read A Rust library called formoniq solves Hodge-Laplace eigenvalue and source problems on coordinate-free n-dimensional simplicial complexes whose only geometry is edge lengths; the torus's two zero eigenvalues and O(h) convergence validate it.

desk verdict A solid, honest BSc thesis that ships a working arbitrary-dimension, coordinate-free FEEC library in Rust; the math is known, the software is the contribution, and the reader's boundary-trace concern does not land on the actual equations. read the letter →

arxiv 2506.02429 v2 pith:N2SZNBG4 submitted 2025-06-03 math.NA cs.NAmath.APmath.ATmath.DGmath.FA

classification math.NAcs.NAmath.APmath.ATmath.DGmath.FA MSC 65N3058A1065N25
keywords finiteelementexteriorcalculusHodge-LaplaceoperatorWhitneyformscoordinate-freegeometryReggesimplicialcomplexdiscretedifferentialRust
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper reports formoniq, a finite element library written in Rust that solves elliptic Hodge-Laplace eigenvalue and source problems on $n$-dimensional simplicial complexes whose geometry consists only of edge lengths, with no vertex coordinates anywhere. The claim is that a full Finite Element Exterior Calculus pipeline—Whitney $k$-form basis functions, Hodge mass matrices, the discrete exterior derivative, and the mixed weak formulation—works on abstract coordinate-free manifolds, with each cell's Riemannian metric reconstructed intrinsically from its edges via the law of cosines. The supporting evidence is the 1-form spectrum of a torus, where the two zero eigenvalues correctly reproduce the first Betti number $b_1 = 2$, and manufactured-source tests in 2D and 3D showing errors in both the solution and its exterior derivative decreasing at rates close to $O(h)$. A reader should care because the approach cuts the usual dependence of FEEC implementations on an ambient embedding, opening the same solver to curved and high-dimensional domains.

What carries the argument

The load-bearing construction is the pairing of the Whitney complex with an intrinsic metric. A Whitney $k$-form is the piecewise-linear differential form attached to a $k$-simplex, with degrees of freedom given by integrals over $k$-simplices; its local shape functions are built from barycentric coordinates, $\lambda_{i_0\dots i_k} = k!\sum_{l=0}^{k}(-1)^l \lambda_{i_l}\, d\lambda_{i_0}\wedge\cdots\wedge\widehat{d\lambda_{i_l}}\wedge\cdots\wedge d\lambda_{i_k}$. The geometry enters only through each cell's constant metric tensor, recovered from edge lengths by the law of cosines in the spirit of Regge calculus, so the Hodge inner product on forms is computed from the inverse Gramian $G^{-1}$ with the standard determinant rule $\langle e_I, e_J\rangle = \det[\langle dx^{I_i}, dx^{J_j}\rangle]$. All four Galerkin operators needed for the mixed Hodge-Laplace formulation reduce to Hodge mass matrices combined with the purely combinatorial coboundary matrices $d_k = \partial_{k+1}^{\mathsf{T}}$: $D_k = M_k d_{k-1}$, $C_k = d_{k-1}^{\mathsf{T}} M_k$, and $L_k = d_k^{\mathsf{T}} M_{k+1} d_k$. The eigenvalue problem is then the sparse symmetric-indefinite generalized eigenproblem $\left[\begin{smallmatrix} M_{k-1} & -C \\ D & L\end{smallmatrix}\right]\left[\begin{smallmatrix}\sigma\\u\end{smallmatrix}\right] = \lambda \left[\begin{smallmatrix}0&0\\0&M_k\end{smallmatrix}\right]\left[\begin{smallmatrix}\sigma\\u\end{smallmatrix}\right]$, and the source problem augments this system with the harmonic-form block that enforces $u \perp \mathfrak{H}^k$.

What would settle it

Evaluate, by high-order quadrature, the boundary integrals that the variational formulation silently drops: on each face of the cube $[0,\pi]^n$, pull back the exact solution $u$ and the trace of $\star du$ to the face and integrate them against Whitney test forms. Any nonzero result means the assembled right-hand side $\mathbf{b}$ is missing boundary contributions, and the reported $O(h)$ convergence describes a source problem whose data differ from those stated in Eq. (82).

Watch

Extended reading notes

Core claim

The central claim is that the discrete de Rham complex built from first-order Whitney forms can be fully assembled from just two inputs: the combinatorics of a simplicial complex and the length of every edge. Given those lengths, Regge calculus supplies each cell's constant metric tensor $\mathbf{G}$ through the law of cosines, $G_{ij} = \frac{1}{2}(d_{0i}^2 + d_{0j}^2 - d_{ij}^2)$, and the Hodge mass matrices, source vectors, and the mixed Hodge-Laplace system are then computed by quadrature and assembly without ever consulting an embedding. The paper's validation is two-fold: on a torus with major radius $0.5$ and minor radius $0.2$, the computed spectrum of $\Delta_1$ contains two zero eigenvalues, matching the two harmonic 1-forms predicted by Hodge theory and the topology ($b_1 = 2$), followed by eigenvalue groups near the flat-torus values $4$, $16$, and $25$; and for the manufactured 1-form solution $u = \sum_i \sin^2(x_i)\prod_{j\neq i}\cos(x_j)$ on $[0,\pi]^n$, the $\mathrm{L}^2$ errors of $u$ and $du$ both converge at roughly $O(h)$ in $n = 2$ and $n = 3$. The author presents these results as confirmation that intrinsic, coordinate-free discretization preserves both the topological and the geometric content of the Hodge-Laplacian.

Load-bearing premise

The convergence study rests on the assertion that the manufactured solution $u$ and its exterior derivative $du$ have vanishing traces on the boundary of $[0,\pi]^n$, which is what removes all boundary terms from the mixed weak form; the paper states this assertion but never verifies it by computation, so if any of those boundary integrals is actually nonzero, the reported $O(h)$ rates describe a different boundary-value problem than the one stated.

Editorial extensions

If this is right

  • The solver doubles as a homology detector: on a closed manifold, the number of zero eigenvalues of the discrete Hodge-Laplacian on $k$-forms equals the $k$-th Betti number, so the same code that produced $b_1 = 2$ on the torus gives the harmonic-form dimension on any mesh whose edge lengths are supplied.
  • Measured convergence of $\|du - du_h\|_{\mathrm{L}^2}$ at rate near $O(h)$ matches the theoretical expectation for first-order Whitney elements in the $H\Lambda$ norm, which the author takes as evidence that the exterior-derivative discretization and its assembly are correct.
  • Because only edge lengths define the geometry, any simplicial mesh with an admissible (Cayley-Menger-realizable) metric can be solved on, regardless of dimension or curvature, as long as the input comes as abstract simplices plus lengths.
  • The same mixed formulation covers every grade of the de Rham complex—grade 0 reduces to the scalar Laplacian with empty auxiliary blocks—so the implementation is not restricted to the 1-form problems used in the validation.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper's own convergence tables show the $\mathrm{L}^2$ rate stuck at $O(h)$ instead of the $O(h^2)$ an Aubin-Nitsche duality argument suggests; a natural next experiment is refining a curved, non-flat mesh and watching whether the rate stays at $O(h)$, which would implicate the piecewise-flat metric approximation rather than the discrete forms themselves.
  • Because the formulation separates topology from geometry, the same code path could compute harmonic representatives on any closed manifold given only edge lengths, effectively turning the eigensolver into a largely combinatorial cohomology tool once the metric is treated as arbitrary input.
  • The paper leaves the codifferential error $\|\delta(u-u_h)\|_{\mathrm{L}^2}$ unmeasured, so the claimed convergence in the full $H\Lambda$ norm is incomplete; measuring it on the same manufactured solutions is the cheapest direct check of the remaining half of the Hodge-Laplacian.
  • The torus eigenvalue groups sit slightly above the flat-torus predictions (about 4.116 versus 4, and 14.45 versus 16); refining the torus mesh and tracking whether these groups converge downward to the predicted values would separate pure discretization error from the geometric approximation of the curved torus.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

0 major / 6 minor

Summary. The manuscript presents a Rust library, formoniq, implementing Finite Element Exterior Calculus (FEEC) on abstract, coordinate-free simplicial complexes in arbitrary dimensions. The geometry is encoded through edge lengths via a Regge-calculus-derived Riemannian metric. The library provides first-order Whitney basis functions, closed-form element mass matrices, discrete exterior derivatives via incidence matrices, and solvers for Hodge-Laplace eigenvalue and source problems using mixed formulations. Numerical validation includes a 1-form eigenvalue problem on a torus (where the two zero eigenvalues correctly reflect b1 = 2) and a manufactured-solution convergence study for the 1-form source problem in 2D and 3D, showing O(h) convergence for both the solution and its exterior derivative.

Significance. If the claims hold, this is a useful contribution as a working reference implementation of coordinate-free FEEC in arbitrary dimensions, with a clean separation between topology and metric geometry. Concrete strengths include a machine-checked unit test for the defining Whitney integral property (Section 4.2.1), closed-form integration of Whitney mass matrices without quadrature (Eq. 70), and external validation targets: an analytic manufactured solution and the topological requirement of two harmonic 1-forms on the torus. The observed first-order convergence is consistent with lowest-order Whitney elements. The main weaknesses are the narrow validation scope (only 1-forms, n = 2,3, trivial topology for the source problem, and only a heuristic flat-torus baseline for the eigenvalue problem) and several presentation inaccuracies, but these do not, in this referee's reading, invalidate the central derivation.

minor comments (6)
  1. [Section 7.2, Eq. (93)] The boundary conditions are written as Tr u = 0 and Tr du = 0, but the strong form in Eq. (82) imposes tr ⋆ u = 0 and tr ⋆ du = 0. Although the manufactured solution in Eq. (88) does satisfy the correct star-trace conditions (the normal component of u and the normal-tangential components of du vanish on the boundary), the notation should be corrected and a brief verification or explicit statement added, since the current text asserts the wrong trace conditions.
  2. [Section 7.2, convergence discussion] The discussion of the L2 convergence rate is misleading. For lowest-order Whitney 1-forms the L2 approximation order is O(h) because the space contains constant 1-forms but not all linear 1-forms; an Aubin-Nitsche duality argument does not generally raise the L2 error to O(h^2) for these elements. The observed O(h) rate for the solution value is therefore expected and not a sign of a flaw; the paragraph should be revised to state this.
  3. [Chapter 6.1, after Eq. (81)] The block matrix in Eq. (81) is described as 'symmetric indefinite', but with D = C^T the off-diagonal blocks are -C and D, which are not transposes of each other. The generalized eigenvalue problem is equivalent to a symmetric problem via the Schur complement (C^T M^{-1} C + L), but the matrix itself is non-symmetric. This statement should be corrected.
  4. [Section 7.1] The comparison to the idealized flat torus spectrum is only heuristic, since the mesh is a curved torus. The first non-zero group (4.116 vs 4) is close, but the second group (about 14.45 vs 16) deviates by roughly 10%. The language 'excellent agreement' and 'closely matches' overstates the agreement; the authors should either soften these claims or compute reference eigenvalues for the specific piecewise-flat discretized torus.
  5. [Section 7.2] The source problem is validated only for the trivial cohomology case (homology_dim = 0 on the cube). The implementation of the harmonic-form constraints in Eq. (83), which is a central part of the mixed formulation for general topology, is not exercised in any numerical test. As the paper explicitly states that verification is partial, this is a limitation rather than an error, but a test on a domain with non-trivial topology would considerably strengthen the central claim about the source problem.
  6. [Throughout] Minor typographical and language issues: 'combing' should be 'combining' (Section 4.2.1), 'irregardless' should be 'regardless' (Section 3.1), and the bibliography lists 'mathemtical' (reference [10]). These do not affect the technical content.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the implementation derives element matrices from defining Whitney formulas and the metric, and its validation targets are external manufactured and analytically known spectra.

full rationale

The derivation chain is self-contained against external mathematical and numerical benchmarks. The Hodge-Laplace mixed system is assembled from mass matrices computed from the Whitney basis definitions and the edge-length-derived metric (Section 5 and Eq. 69–70), with no fitted parameters and no post-hoc exclusions. The eigenvalue validation on the torus compares the computed spectrum to analytically known flat-torus eigenvalues and to the topologically protected count b1 = 2 of zero eigenvalues, which is an external, non-parametric target. The source-problem convergence study uses a manufactured solution whose right-hand side is the analytical Laplacian of the exact solution; the solution and its exterior derivative are then measured against the exact fields, so the prediction is not forced by construction. The asserted boundary-trace conditions in Eq. 93 are a property of the manufactured solution, not an input-equivalent construction; even if the assertion were incorrect, it would affect the validity of the test, not the circularity of the derivation. Self-citations are standard references to FEEC and Whitney forms and are not load-bearing in the sense of supplying the derived result itself. I find no step where a fitted parameter is renamed as a prediction, where an ansatz is smuggled in via citation, or where a uniqueness theorem from the authors is invoked to force a choice.

Assumptions & free parameters 0 free parameters · 7 assumptions · 0 invented entities

The library introduces no fitted constants and no new entities: all inputs are mesh topology plus edge lengths (or vertex coordinates converted to edge lengths), and the element matrices are computed from the defining formulas of Whitney forms and the metric. The listed axioms are the standard mathematical framework the implementation relies on, plus two test-specific assumptions that carry most of the validation risk.

assumptions (7)
  • standard math Arnold-Falk-Winther FEEC theory: the mixed formulation of the Hodge Laplacian is stable and convergent on the Whitney subcomplex
    Invoked throughout Chapters 5-6 to justify the mixed weak form (Eq. 77) and the convergence rates cited in Section 7.2.
  • standard math de Rham theorem and Hodge theory: the kernel of the Hodge Laplacian on k-forms has dimension equal to the k-th Betti number
    Used in Section 7.1 to interpret the two zero eigenvalues on the torus as harmonic 1-forms and the b1 = 2 topology.
  • standard math Whitney forms satisfy the integral duality property and form a subcomplex of the de Rham complex
    Basis of the discretization in Chapter 4 (Eq. 58) and of the discrete exterior derivative as the transpose incidence matrix (Eq. 51); tested, not proved, in the whitney_basis_property test.
  • domain assumption Piecewise-flat (Regge) edge-length geometry is an admissible variational crime for first-order Whitney methods
    Assumed in Sections 2.8-2.9 via Holst-Stern; needed so that mesh geometry error does not spoil the measured convergence rates on curved domains such as the torus.
  • domain assumption Barycentric quadrature (degree-1 exact) suffices for cochain projection (Eq. 44) and source assembly (Eq. 73)
    The thesis states this is admissible for first-order FEEC; the quadrature error is asserted, not quantified.
  • domain assumption The manufactured solution u in Eq. 88 has zero trace and zero trace of its exterior derivative on the cube boundary
    Eq. 93 in Section 7.2; asserted rather than demonstrated, and load-bearing for the source-problem validation.
  • ad hoc to paper The analytic spectrum of the idealized flat torus with circumferences pi and 0.4 pi approximates the spectrum of the curved torus mesh well enough to serve as a validation baseline
    Section 7.1 uses lambda = 4m^2 + 25n^2 to validate eigenvalues computed on a curved torus; the approximation is justified only qualitatively.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Rust Implementation of Finite Element Exterior Calculus on Coordinate-Free Simplicial Complexes." pith.science (2026). https://pith.science/paper/N2SZNBG4

@misc{pith2026250602429,
  author       = {Pith},
  title        = {Pith review of: Rust Implementation of Finite Element Exterior Calculus on Coordinate-Free Simplicial Complexes},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/N2SZNBG4}},
  note         = {Machine review of arXiv:2506.02429}
}
read the original abstract

This thesis presents the development of a novel finite element library in Rust based on the principles of Finite Element Exterior Calculus (FEEC). The library solves partial differential equations formulated using differential forms on abstract, coordinate-free simplicial complexes in arbitrary dimensions, employing an intrinsic Riemannian metric derived from edge lengths via Regge Calculus. We focus on solving elliptic Hodge-Laplace eigenvalue and source problems on the nD de Rham complex. We restrict ourselves to first-order Whitney basis functions. The implementation is partially verified through convergence studies.

Figures

Figures reproduced from arXiv: 2506.02429 by the authors.

Figure 1
Figure 1. Vector proxies of Reference Local Shape Functions [PITH_FULL_IMAGE:figures/full_fig_p043_1.png] view at source ↗
Figure 2
Figure 2. Vector proxies of Global Shape Functions [PITH_FULL_IMAGE:figures/full_fig_p044_2.png] view at source ↗
Figure 3
Figure 3. Vector proxies of some example Whitney forms on equilateral triangle mesh [PITH_FULL_IMAGE:figures/full_fig_p044_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: The two harmonic forms on the torus, representing the 1-cohomology spaces. [PITH_FULL_IMAGE:figures/full_fig_p061_4.png]
Figure 5
Figure 5. Figure 5: Finite element solution to manufactured source problem in 2D at refinement level 4. [PITH_FULL_IMAGE:figures/full_fig_p065_5.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

32 extracted references · 25 canonical work pages

  1. [1]

    Numerical Methods for (Partial) Differential Equations,

    R. Hiptmair, “Numerical Methods for (Partial) Differential Equations,” 2025. Accessed: Apr. 24, 2025. [Online]. Available: https://www.sam.math.ethz.ch/~grsam/NUMPDEFL/NUMPDE.pdf

  2. [2]

    Higher order Whitney forms,

    R. Hiptmair, “Higher order Whitney forms,” Geometric Methods for Computational Electromagnetics , vol

  3. [3]

    Finite element exterior calculus, homological techniques, and applications,

    D. Arnold, R. Falk, and R. Winther, “Finite element exterior calculus, homological techniques, and applications,” Acta Numerica, vol. 15, pp. 1–155, 2006, doi: 10.1017/S0962492906210018

  4. [4]

    D. N. Arnold, Finite Element Exterior Calculus , 0th ed. Philadelphia, PA: Society for Industrial, Applied Mathematics, 2018. doi: 10.1137/1.9781611975543

  5. [5]

    Frankel, The Geometry of Physics: An Introduction, 3rd ed

    T. Frankel, The Geometry of Physics: An Introduction, 3rd ed. Cambridge University Press, 2011

  6. [6]

    Hatcher, Algebraic Topology

    A. Hatcher, Algebraic Topology. in Algebraic Topology. Cambridge University Press, 2002. [Online]. Available: https://books.google.ch/books?id=BjKs86kosqgC

  7. [7]

    The Rust Programming Language

    The Rust Programming Language Developers, “The Rust Programming Language.” 2025

  8. [8]

    Digital Geometry Processing with Discrete Exterior Calculus,

    M. D. P. S. Keenan Crane Fernando de Goes, “Digital Geometry Processing with Discrete Exterior Calculus,” in ACM SIGGRAPH 2013 courses, in SIGGRAPH '13. Anaheim, California: ACM, 2013

Show all 32 references
  1. [9]

    General relativity without coordinates,

    T. Regge, “General relativity without coordinates,” Il Nuovo Cimento (1955-1965), vol. 19, no. 3, pp. 558–571, Feb. 1961, doi: 10.1007/BF02733251

  2. [10]

    Whitney, Geometric integration theory

    H. Whitney, Geometric integration theory. in Princeton mathemtical series. Univ.Pr., 1957. [Online]. Available: https://books.google.ch/books?id=o-5szQEACAAJ

  3. [11]

    Geometric Variational Crimes: Hilbert Complexes, Finite Element Exterior Calculus, and Problems on Hypersurfaces,

    M. Holst and A. Stern, “Geometric Variational Crimes: Hilbert Complexes, Finite Element Exterior Calculus, and Problems on Hypersurfaces,” Foundations of Computational Mathematics , vol. 12, no. 3, pp. 263–293, 2012, doi: 10.1007/s10208-012-9119-7

  4. [12]

    Klabnik and C

    S. Klabnik and C. Nichols, The Rust Programming Language. No Starch Press, 2018

  5. [13]

    RustBelt: Securing the Foundations of the Rust Programming Language,

    R. Jung, J.-H. Jourdan, and R. Krebbers, “RustBelt: Securing the Foundations of the Rust Programming Language,” Proceedings of the ACM on Programming Languages, 2017

  6. [14]

    nalgebra: Linear algebra library for the Rust programming language

    S. Crozet and others, “nalgebra: Linear algebra library for the Rust programming language.” 2025

  7. [15]

    PETSc/TAO Users Manual,

    S. Balay et al. , “PETSc/TAO Users Manual,” 2024. [Online]. Available: https://petsc.org/release/docs/ manual.pdf

  8. [16]

    Efficient Management of Parallelism in Object Oriented Numerical Software Libraries,

    S. Balay, W. D. Gropp, L. C. McInnes, and B. F. Smith, “Efficient Management of Parallelism in Object Oriented Numerical Software Libraries,” 1997. [Online]. Available: https://library.anl.gov/anl-external/technical-reports/ 1997/ANL-97-1.pdf

  9. [17]

    SLEPc: A scalable and flexible toolkit for the solution of eigenvalue problems,

    V. Hernandez, J. E. Roman, and V. Vidal, “SLEPc: A scalable and flexible toolkit for the solution of eigenvalue problems,” ACM Transactions on Mathematical Software (TOMS) , vol. 31, no. 3, pp. 351–362, 2005, doi: 10.1145/1089014.1089019

  10. [18]

    itertools: Extra iterator adaptors, iterator methods, free functions, and macros

    I. Authors, “itertools: Extra iterator adaptors, iterator methods, free functions, and macros.” 2025

  11. [19]

    indexmap: A hash table with consistent order and fast iteration; access items by key or sequence index

    bluss and J. Stone, “indexmap: A hash table with consistent order and fast iteration; access items by key or sequence index.” 2024

  12. [20]

    Rayon: A data parallelism library for Rust

    N. Matsakis and others, “Rayon: A data parallelism library for Rust.” 2025

  13. [21]

    Gmsh: A 3-D finite element mesh generator with built-in pre- and post- processing facilities,

    C. Geuzaine and J.-F. Remacle, “Gmsh: A 3-D finite element mesh generator with built-in pre- and post- processing facilities,” International Journal for Numerical Methods in Engineering , vol. 79, no. 11, pp. 1309– 1331, 2009, doi: 10.1002/nme.2579. i

  14. [22]

    mshio: Parser library for the Gmsh MSH file format

    F. L. (w1th0utnam3), “mshio: Parser library for the Gmsh MSH file format.” 2020

  15. [23]

    Numerical Methods for Computational Science and Engineering,

    R. Hiptmair, P. Arbenz, and V. Gradinaru, “Numerical Methods for Computational Science and Engineering,”

  16. [24]

    L. M. Blumenthal, Theory and Applications of Distance Geometry . in Oxford Mathematical Monographs. Oxford: Oxford University Press, 1953

  17. [25]

    Some Combinatorial Lemmas in Topology,

    H. W. Kuhn, “Some Combinatorial Lemmas in Topology,” IBM Journal of Research and Development, vol. 4, no. 5, pp. 518–524, 1960, doi: 10.1147/rd.45.0518

  18. [26]

    Finite elements in computational electromagnetism,

    R. Hiptmair, “Finite elements in computational electromagnetism,” Acta Numerica, vol. 11, pp. 237–339, 2002, doi: 10.1017/S0962492902000041

  19. [27]

    PyDEC: A Python Library for Discretizations of Exterior Calculus

    A. N. Hirani, “PyDEC: A Python Library for Discretizations of Exterior Calculus.” 2011

  20. [28]

    FEEC++ (formerly simplefem): A C++ library for Finite Element Exterior Calculus

    M. Licht, “FEEC++ (formerly simplefem): A C++ library for Finite Element Exterior Calculus.” 2017

  21. [29]

    DDF.jl: A Julia package for discretizing differential forms via finite elements

    E. Schnetter, “DDF.jl: A Julia package for discretizing differential forms via finite elements.” 2021

  22. [30]

    dexterior: A Rust toolkit for Discrete Exterior Calculus

    M. Myyrä, “dexterior: A Rust toolkit for Discrete Exterior Calculus.” 2024. ii

  23. [32]

    271–299, 2001

    EMW Publishing, Cambridge, MA, pp. 271–299, 2001. [Online]. Available: http://hdl.handle.net/20.500. 11850/158477

  24. [2025]

    24, 2025

    Accessed: Apr. 24, 2025. [Online]. Available: https://www.sam.math.ethz.ch/~grsam/NumMeth/ NumCSE_Lecture_Document.pdf

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.