Pith. sign in

REVIEW 2 major objections 4 minor 64 references

Discrete Vector Calculus and Helmholtz Hodge Decomposition for Classical Finite Difference Summation by Parts Operators

T0 review · 2 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read For nullspace consistent tensor product SBP operators, the discrete kernel of curl strictly contains the image of grad (and the kernel of div strictly contains the image of curl), so the discrete Helmholtz Hodge decomposition generally…

desk verdict A clean negative theorem: collocated SBP operators fail a discrete Poincare lemma and Helmholtz-Hodge decomposition, with the obstruction characterized exactly as grid oscillations; worth a serious referee. read the letter →

arxiv 1908.08732 v2 pith:2ADBBFQ5 submitted 2019-08-23 math.NA cs.NA

Hendrik Ranocha , Katharina Ostaszewski , Philip Heinisch This is my paper · ORCID
classification math.NAcs.NA MSC 65N0665M0665N3565M7065Z05
keywords summationbypartsHelmholtzHodgedecompositiongridoscillationsnullspaceconsistencymimeticpropertiesvectorcalculuswavemodeanalysis
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

This paper asks whether the classical representation theorems of vector calculus survive discretization by finite difference summation by parts (SBP) operators: every curl-free vector field is a gradient, every divergence-free field is a curl, and every field splits orthogonally into a gradient plus a curl (the Helmholtz Hodge decomposition). The answer is negative for the classical collocated operators: under nullspace consistency, the discrete kernel of curl is strictly larger than the image of grad, and the kernel of div is strictly larger than the image of curl, in both two and three dimensions. The extra dimensions are spanned by explicit grid oscillation vectors, so some irrotational or solenoidal grid functions have no scalar or vector potential, and the discrete Helmholtz Hodge decomposition generally leaves a nonzero remainder. The paper then shows the obstruction does not prevent the decomposition from being useful: least-squares projections onto the gradient and curl images converge, the remainder shrinks under grid refinement for smooth data, and the machinery separates magnetohydrodynamic wave modes in practice.

What carries the argument

The load-bearing object is the nullspace consistent summation by parts (SBP) derivative operator: a discrete derivative $D$ together with a symmetric positive definite mass matrix $M$ and boundary operator $E$ satisfying $MD + D^T M = E$, with the additional requirement $\ker D = \operatorname{span}\{\mathbb{1}\}$. For such an operator the adjoint $D^* = M^{-1}D^T M$ has a one-dimensional kernel spanned by a grid oscillation vector $\mathbf{osc}$, which alternates between positive and negative values across the grid. In tensor product dimensions, these oscillation vectors placed in single coordinate components generate exactly the gap between $\operatorname{im} \operatorname{grad}$ and $\ker \operatorname{curl}$ (and between $\operatorname{im} \operatorname{curl}$ and $\ker \operatorname{div}$). Two tools carry the proofs: dimension counts relating $\ker D^*$ to $\ker D$, and explicit discrete inverse operators $D^{-1}$ that integrate along coordinate lines to build scalar potentials for fields orthogonal to the oscillations. The same machinery produces the filter operator $F$ that projects grid oscillations out of a field, and the least-squares projection setup used for the numerical decompositions.

What would settle it

Assemble the classical second-order SBP operator of Example 2.2 as a tensor product on a coarse grid such as $N_1 = N_2 = 5$, compute the numerical kernels and ranks of the matrices $\operatorname{grad}$ and $\operatorname{curl}$, and check whether $\dim \ker \operatorname{curl} = N_1N_2 + 1$ while $\dim \operatorname{im} \operatorname{grad} = N_1N_2 - 1$; then take $u = (\mathbf{osc}_{12}, 0)^T$, compute the least-squares projections onto $\operatorname{im} \operatorname{grad}$ and $\operatorname{im} \operatorname{rot}$, and verify that the remainder $r$ does not vanish. If instead $\ker \operatorname{curl} = \operatorname{im} \operatorname{grad}$, the paper's central claim is refuted.

Watch

Extended reading notes

Core claim

The central discovery is a dimension mismatch with explicit generators. For nullspace consistent tensor product SBP operators, $\dim \operatorname{im} \operatorname{grad} = N_1N_2 - 1 < N_1N_2 + 1 = \dim \ker \operatorname{curl}$ in two dimensions, with $\ker \operatorname{curl} = \operatorname{im} \operatorname{grad} \oplus \operatorname{span}\{(\mathbf{osc}_1, 0)^T, (0, \mathbf{osc}_2)^T\}$; in three dimensions $\dim \ker \operatorname{curl} = N_1N_2N_3 + 2$ versus $\dim \operatorname{im} \operatorname{grad} = N_1N_2N_3 - 1$, and the corresponding statements hold for divergence versus curl (Theorems 3.7, 3.8, 3.18, 3.19). Here $\mathbf{osc}_i$ is the one-dimensional grid oscillation vector spanning the kernel of the adjoint derivative $D_i^*$, a vector that alternates in sign across the grid. Because the kernels are strictly larger than the images, a discrete Poincaré lemma fails: there are grid functions $u$ that cannot be written as $\operatorname{grad}\phi + \operatorname{curl} v$, so the discrete Helmholtz Hodge decomposition $u = \operatorname{grad}\phi + \operatorname{curl} v + r$ has a nonzero remainder $r$ lying in the orthogonal complement of both image spaces (Theorem 5.1). The paper's positive claim is that the obstruction is numerically benign: computing the decomposition as mass-matrix-scaled least-norm least-squares projections (LSQR in 2D, LSMR in 3D), the potentials and components converge with the expected order of accuracy, the remainder vanishes under refinement for smooth data, and the method separates MHD wave modes when the projection order is matched to the dominant amplitude.

Load-bearing premise

All of the theorems assume nullspace consistency: each one-dimensional SBP derivative operator $D$ must have kernel exactly $\operatorname{span}\{\mathbb{1}\}$, so that spurious kernel modes are absent; if an operator has extra kernel modes, the dimension counts and the explicit spanning sets no longer follow.

Editorial extensions

If this is right

  • Irrotational discrete vector fields need not be gradients: for nullspace consistent tensor product SBP operators, $\ker \operatorname{curl}$ strictly contains $\operatorname{im} \operatorname{grad}$ in two and three dimensions, so a discrete Poincaré lemma fails.
  • The discrete Helmholtz Hodge decomposition $u = \operatorname{grad}\phi + \operatorname{curl} v + r$ generally has a nonzero remainder $r$, orthogonal to both image spaces; this is a structural property of collocated SBP operators, not a numerical defect.
  • For grid functions orthogonal to the one-dimensional grid oscillations, the classical potential theorems hold, and the paper's filter operator $F$ restores that situation by explicitly removing the oscillations.
  • Because $\operatorname{im} \operatorname{grad}$ and $\operatorname{im} \operatorname{curl}$ are not orthogonal, the two projection orders yield different decompositions, and the MHD tests show the correct order depends on which wave amplitude dominates.
  • In numerical experiments the decomposition converges: potentials and components reach order $p+1$ accuracy for operators of interior order $2p$, the remainder vanishes under refinement for smooth data, and MHD wave-mode separation succeeds when the projection order matches the dominant amplitude.

Reading between the lines

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

  • The same kernel mismatch should appear in any collocated discretization whose derivative adjoint has a nontrivial kernel, including global spectral collocation at Lobatto–Legendre nodes, where the oscillation vector is the highest Legendre mode; the paper notes this class satisfies its assumptions, so the negative result likely transfers verbatim.
  • A practical adaptive rule suggested by the MHD experiments — estimate the amplitudes of the competing wave families and project first onto the dominant one — could be formalized and tested, since the paper only reports the qualitative direction of the effect, not a threshold.
  • A rigorous convergence rate for the remainder $\|r\|_M$ under grid refinement is a natural open problem; the paper reports experimental orders of accuracy ($p+1$, with one component at $4.6$ for the sixth-order operator) but proves no error estimate.
  • Because staggered-grid and mimetic operators do support exact discrete decompositions, as the paper notes, moving to staggered SBP schemes is a concrete, testable remedy for applications that need a discrete Poincaré lemma to hold exactly.
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

2 major / 4 minor

Summary. This paper studies whether classical vector calculus identities behind the Helmholtz-Hodge decomposition survive for collocated tensor-product finite-difference summation-by-parts (SBP) operators. Under the nullspace-consistency assumption that each one-dimensional derivative satisfies ker D = span{1}, the authors prove that the discrete kernels of curl and div are strictly larger than the images of grad and curl/rot, with the missing dimensions explicitly characterized as grid oscillations, i.e., vectors in ker D*. Consequently, a discrete Poincaré lemma fails, and the discrete Helmholtz-Hodge decomposition u = grad φ + curl v + r has a nonvanishing remainder for certain grid functions. The paper also characterizes fields that are both divergence-free and curl-free as gradients of discrete harmonic functions, proposes a least-squares iterative method (LSQR/LSMR) for computing the decomposition, and supports the theory with numerical experiments showing p+1 convergence of the potentials and vanishing of the remainder under refinement, including an application to MHD wave-mode separation.

Significance. The negative result is significant for the SBP community: it shows that classical collocated tensor-product SBP operators cannot exactly mimic the Helmholtz-Hodge decomposition, contrary to what one might expect from their mimetic properties. The obstruction is identified concretely as explicit grid oscillation vectors, and the proofs are elementary finite-dimensional linear algebra. The paper is careful to state its conditional nature, and the numerical experiments with published reproducible code strengthen the contribution. The main limitation is the explicit dependence on nullspace consistency; while the dimension inequality driving the negative result is robust to spurious kernel modes, the exact kernel decompositions would need modification for operators failing Definition 2.3.

major comments (2)
  1. [Section 3.5, Corollary 3.16] Corollary 3.16 is stated without proof, yet it is the step that turns the superset inclusions of Lemmas 3.12 and 3.13 into the exact kernel decompositions of Theorems 3.18 and 3.19. Please add the short argument: by Lemma 3.15, every element of ker curl that is orthogonal to the three vectors (osc1,0,0), (0,osc2,0), (0,0,osc3) lies in im grad; therefore the orthogonal direct sum of im grad with that span has the same dimension as ker curl, namely dim im grad + 3.
  2. [Section 4, Theorem 4.2] The proof of Theorem 4.2 asserts that 'the grid oscillations appearing in ker div are not in ker curl and vice versa' implies ker div ∩ ker curl ⊆ im grad ∩ im rot. This implication is not immediate: an element of the intersection could in principle be a sum of a gradient and an oscillator, even if the oscillator itself is not in the other kernel. Please replace this assertion with a direct argument, for example using the orthogonal decompositions of Theorems 3.7–3.19 to show that any component along the oscillator span cannot satisfy the divergence-free condition, so the intersection consists only of gradients (respectively rotations) whose Laplacian vanishes.
minor comments (4)
  1. [Abstract] The phrase 'the discrete remainder vanishes' could be misread as exact vanishing for finite N, which would contradict Theorem 5.1. Consider rewording to 'the discrete remainder tends to zero under grid refinement'.
  2. [Section 3.2, proof of Theorem 3.7] In the proof, after stating that the span in (14) is contained in ker grad*, it should also be noted explicitly that the two oscillatory vectors belong to ker curl; this is required for the decomposition and is only implicit.
  3. [Section 5.1] The MATLAB-style pseudo-code for the scaled least-squares solves is helpful, but a short prose explanation of the scaling by sqrtM and why it makes the projections orthogonal with respect to the SBP scalar product would improve readability for readers not familiar with the notation.
  4. [Section 6.2] The observed order 4.6 for the vector potential v with the 2p=6 operator differs from the nominal p+1=4. It would be useful to add one sentence clarifying that this is a single test case and no theoretical claim is made about this particular rate.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the main negative theorems follow from rank-nullity arguments and stated nullspace-consistency assumptions, not from fitted inputs or self-citations.

full rationale

The derivation chain is self-contained. The grid oscillation vectors osc_i are introduced in Definition 3.4 as basis vectors of ker D* for a nullspace-consistent derivative operator, and their role in the kernel decompositions is proved rather than assumed. The central dimension identities in Theorems 3.7, 3.8, 3.18, and 3.19 follow from rank-nullity counts (Lemma 3.3), the commutativity of tensor-product derivative operators, and the explicit construction of scalar/vector potentials in Sections 3.3 and 3.5. In particular, the condition u_i in im D_i is derived from the curl/divergence constraints plus orthogonality to the oscillation vectors, and is not imposed to force the theorem. Theorem 5.1 is likewise a direct rank argument: dim(im grad + im curl) <= 3N-3 < 3N, so the claimed nonvanishing remainder is a consequence of finite dimensionality, not an input. The numerical remainder observed in Section 6 is reported as an experimental finding and is not claimed to vanish by construction. The self-citations ([43]-[49]) are background material or code/data provenance; none supplies a load-bearing premise for the main impossibility result. The nullspace-consistency hypothesis is stated explicitly as a condition of the theorems, not fitted to the target conclusion. Thus no step reduces to its own inputs by construction and no load-bearing circularity is present.

Assumptions & free parameters 0 free parameters · 6 assumptions · 0 invented entities

The central claim has essentially no free parameters: grid oscillations are basis vectors of ker D*, constructed from the operator, not fitted. The main additional assumption is nullspace consistency, which is stated and standard for many classical SBP operators. No new physical or mathematical entities are postulated beyond explicit basis vectors.

assumptions (6)
  • domain assumption Nullspace consistency: ker D = span of the constant vector for each one-dimensional SBP derivative operator
    Stated in Definition 2.3 and used in Lemma 3.3 and Theorems 3.7, 3.8, 3.18, 3.19 to compute dimensions of kernels and images. Not all SBP operators satisfy it.
  • domain assumption Tensor product structure of multi-dimensional SBP operators on Cartesian grids
    Definitions 2.5 and 2.7; the commutation Dj Di = Di Dj and the split of grid oscillations into products of 1D oscillations rely on this structure.
  • domain assumption Domain is a bounded rectangle or cuboid in R^d, d in {2,3}
    Stated before Theorem 3.1; the integral constructions of potentials in Sections 3.3 and 3.5 assume box-shaped domains with coordinate lines.
  • standard math Classical continuous vector calculus theorems (Poincare lemma, Hodge decomposition) hold on such domains
    Theorems 3.1, 3.2, and 4.1 are cited from references 18, 51, and 27 and used as the continuous benchmark; they are standard results.
  • standard math Finite-dimensional linear algebra facts (rank-nullity, orthogonal complements with respect to the mass-matrix inner product)
    Used throughout Section 3 to compute dim im grad, dim ker curl, and related quantities.
  • domain assumption For numerical tests, diagonal norm SBP operators are used
    Section 2 restricts numerical tests to diagonal mass matrices; this simplifies sqrt(M) scaling in Section 5.1 but the theoretical results do not require it.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Discrete Vector Calculus and Helmholtz Hodge Decomposition for Classical Finite Difference Summation by Parts Operators." pith.science (2026). https://pith.science/paper/2ADBBFQ5

@misc{pith2026190808732,
  author       = {Pith},
  title        = {Pith review of: Discrete Vector Calculus and Helmholtz Hodge Decomposition for Classical Finite Difference Summation by Parts Operators},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2ADBBFQ5}},
  note         = {Machine review of arXiv:1908.08732}
}
read the original abstract

In this article, discrete variants of several results from vector calculus are studied for classical finite difference summation by parts operators in two and three space dimensions. It is shown that existence theorems for scalar/vector potentials of irrotational/solenoidal vector fields cannot hold discretely because of grid oscillations, which are characterised explicitly. This results in a non-vanishing remainder associated to grid oscillations in the discrete Helmholtz Hodge decomposition. Nevertheless, iterative numerical methods based on an interpretation of the Helmholtz Hodge decomposition via orthogonal projections are proposed and applied successfully. In numerical experiments, the discrete remainder vanishes and the potentials converge with the same order of accuracy as usual in other first order partial differential equations. Motivated by the successful application of the Helmholtz Hodge decomposition in theoretical plasma physics, applications to the discrete analysis of magnetohydrodynamic (MHD) wave modes are presented and discussed.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

64 extracted references · 34 canonical work pages

  1. [2]

    Ahusborde, M

    E. Ahusborde, M. Azaïez, J.-P. Caltagirone, M. Gerritsma and A. Lemoine. ‘Discrete Hodge Helmholtz Decomposition’. In:Monografías Matemáticas García de Galdeano39 (2014), pp. 1– 10

    work page 2014

  2. [3]

    In:Mathematical Methods in the Applied Sciences21.9 (1998), pp

    C.Amrouche,C.Bernardi,M.DaugeandV.Girault.‘Vectorpotentialsinthree-dimensional non-smooth domains’. In:Mathematical Methods in the Applied Sciences21.9 (1998), pp. 823–

    work page 1998

  3. [5]

    Beresnyak and A

    A. Beresnyak and A. Lazarian.Turbulence in magnetohydrodynamics. Vol. 12. Studies in Math- ematical Physics. Walter de Gruyter GmbH & Co KG, 2019

    work page 2019

  4. [6]

    In:SIAM Review59.1 (2017), pp

    J.Bezanson,A.Edelman,S.KarpinskiandV.B.Shah.‘Julia:AFreshApproachtoNumerical Computing’. In:SIAM Review59.1 (2017), pp. 65–98./d.sc/o.sc/i.sc: 10.1137/141000671. arXiv:1411. 1607 [cs.MS]

    doi:10.1137/141000671 2017

  5. [7]

    Bhatia, G

    H. Bhatia, G. Norgard, V. Pascucci and P.-T. Bremer. ‘The Helmholtz-Hodge decomposition —ASurvey’.In:IEEETransactionsonVisualizationandComputerGraphics 19.8(2012),pp.1386–

    work page 2012

  6. [9]

    In:SIAM Journal on Matrix Analysis and Applications 40.1 (2019), pp

    R.Estrin,D.OrbanandM.A.Saunders.‘LSLQ:AnIterativeMethodforLinearLeast-Squares with an Error Minimization Property’. In:SIAM Journal on Matrix Analysis and Applications 40.1 (2019), pp. 254–275./d.sc/o.sc/i.sc: 10.1137/17M1113552

    work page doi:10.1137/17m1113552 2019

  7. [10]

    Fernandes and G

    P. Fernandes and G. Gilardi. ‘Magnetostatic and electrostatic problems in inhomogeneous anisotropic media with irregular boundary and mixed boundary conditions’. In:Math- ematical Models and Methods in Applied Sciences7.07 (1997), pp. 957–991./d.sc/o.sc/i.sc: 10 . 1142 / S0218202597000487

    work page 1997

  8. [11]

    D. C. D. R. Fernández, P. D. Boom, M. H. Carpenter and D. W. Zingg. ‘Extension of Tensor- Product Generalized and Dense-Norm Summation-by-Parts Operators to Curvilinear Co- ordinates’.In:JournalofScientificComputing (),pp.1–40. /d.sc/o.sc/i.sc: 10.1007/s10915-019-01011-3

    work page doi:10.1007/s10915-019-01011-3

Show all 64 references
  1. [12]

    D. C. D. R. Fernández, P. D. Boom and D. W. Zingg. ‘A generalized framework for nodal firstderivativesummation-by-partsoperators’.In:JournalofComputationalPhysics 266(2014), pp. 214–239./d.sc/o.sc/i.sc: 10.1016/j.jcp.2014.01.038

    2014 doi

  2. [13]

    In:Computers & Fluids95 (2014), pp

    D.C.D.R.Fernández,J.E.HickenandD.W.Zingg.‘Reviewofsummation-by-partsoperators with simultaneous approximation terms for the numerical solution of partial differential equations’. In:Computers & Fluids95 (2014), pp. 171–196./d.sc/o.sc/i.sc: 10.1016/j.compfluid.2014. 02.016

    2014 doi

  3. [14]

    T. C. Fisher and M. H. Carpenter. ‘High-order entropy stable finite difference schemes for nonlinear conservation laws: Finite domains’. In:Journal of Computational Physics252 (2013), pp. 518–557./d.sc/o.sc/i.sc: 10.1016/j.jcp.2013.06.014

    2013 doi

  4. [15]

    D. C.-L. Fong and M. A. Saunders. ‘LSMR: An Iterative Algorithm for Sparse Least-Squares Problems’.In:SIAMJournalonScientificComputing 33.5(2011),pp.2950–2971. /d.sc/o.sc/i.sc: 10.1137/ 10079687X

    2011

  5. [16]

    L. Gao, D. C. D. R. Fernández, M. Carpenter and D. Keyes. ‘SBP–SAT finite difference discretizationofacousticwaveequationsonstaggeredblock-wiseuniformgrids’.In: Journal ofComputationalandAppliedMathematics 348(2019),pp.421–444. /d.sc/o.sc/i.sc: 10.1016/j.cam.2018. 08.040

    2019 doi

  6. [17]

    G. J. Gassner. ‘A Skew-Symmetric Discontinuous Galerkin Spectral Element Discretization and Its Relation to SBP-SAT Finite Difference Methods’. In:SIAM Journal on Scientific Com- puting 35.3 (2013), A1233–A1253./d.sc/o.sc/i.sc: 10.1137/120890144

    2013 doi

  7. [18]

    Girault and P.-A

    V. Girault and P.-A. Raviart.Finite Element Methods for Navier-Stokes Equations: Theory and Al- gorithms.Vol.5.SpringerSeriesinComputationalMathematics.BerlinHeidelberg:Springer Science & Business Media, 2012./d.sc/o.sc/i.sc: 10.1007/978-3-642-61623-5

    2012 doi

  8. [19]

    In:Planetary and Space Science36.8 (1988), pp

    K.-H.Glaßmeier.‘Reconstructionoftheionosphericinfluenceonground-basedobservations of a short-duration ULF pulsation event’. In:Planetary and Space Science36.8 (1988), pp. 801–

    1988

  9. [20]

    Glaßmeier

    K.-H. Glaßmeier. ‘Reflection of MHD-waves in the Pc4-5 period range at ionospheres with non-uniform conductivity distributions’. In:Geophysical Research Letters10.8 (1983), pp. 678–

    1983

  10. [21]

    Glaßmeier, C

    K.-H. Glaßmeier, C. Othmer, R. Cramm, M. Stellmacher and M. Engebretson. ‘Magneto- spheric Field Line Resonances: A Comparative Planetology Approach’. In:Surveys in Geo- physics20.1 (1999), pp. 61–109./d.sc/o.sc/i.sc: 10.1023/A:1006659717963

    1999 doi

  11. [22]

    Glaßmeier

    K.-H. Glaßmeier. ‘On the influence of ionospheres with non-uniform conductivity distribu- tion on hydromagnetic waves’. In:Journal of Geophysics54 (1984), pp. 125–137

    1984

  12. [23]

    J. E. Hicken and D. W. Zingg. ‘Summation-by-parts operators and high-order quadrature’. In: Journal of Computational and Applied Mathematics237.1 (2013), pp. 111–125./d.sc/o.sc/i.sc: 10.1016/ j.cam.2012.07.015. 26

    2013

  13. [24]

    H. T. Huynh. ‘A Flux Reconstruction Approach to High-Order Schemes Including Discon- tinuousGalerkinMethods’.In:18thAIAAComputationalFluidDynamicsConference .American Institute of Aeronautics and Astronautics. 2007./d.sc/o.sc/i.sc: 10.2514/6.2007-4079

    2007 doi

  14. [25]

    J. M. Hyman and M. Shashkov. ‘Natural discretizations for the divergence, gradient, and curlonlogicallyrectangulargrids’.In: Computers&MathematicswithApplications 33.4(1997), pp. 81–104./d.sc/o.sc/i.sc: 10.1016/S0898-1221(97)00009-6

    1997 doi

  15. [26]

    J. M. Hyman and M. Shashkov. ‘The orthogonal decomposition theorems for mimetic finite difference methods’. In:SIAM Journal on Numerical Analysis36.3 (1999), pp. 788–818./d.sc/o.sc/i.sc: 10.1137/S0036142996314044

    1999 doi

  16. [27]

    HomogenizationofDifferentialOperatorsandIntegral Functionals.BerlinHeidelberg:SpringerScience&BusinessMedia,1994

    V.V.Jikov,S.M.KozlovandO.A.Oleinik. HomogenizationofDifferentialOperatorsandIntegral Functionals.BerlinHeidelberg:SpringerScience&BusinessMedia,1994. /d.sc/o.sc/i.sc: 10.1007/978- 3-642-84659-5

    1994 doi

  17. [28]

    In:The Astrophysical Journal720.1 (2010), p

    G.KowalandA.Lazarian.‘Velocityfieldofcompressiblemagnetohydrodynamicturbulence: wavelet decomposition and mode scalings’. In:The Astrophysical Journal720.1 (2010), p. 742. /d.sc/o.sc/i.sc: 10.1088/0004-637X/720/1/742

    2010 doi

  18. [29]

    Kreiss and G

    H.-O. Kreiss and G. Scherer. ‘Finite Element and Finite Difference Methods for Hyperbolic PartialDifferentialEquations’.In:MathematicalAspectsofFiniteElementsinPartialDifferential Equations. Ed. by C. de Boor. New York: Academic Press, 1974, pp. 195–212

    1974

  19. [30]

    In:Journal of Scientific Computing65.1 (2015), pp

    A.Lemoine,J.-P.Caltagirone,M.AzaïezandS.Vincent.‘DiscreteHelmholtz–Hodgedecom- position on polyhedral meshes using compatible discrete operators’. In:Journal of Scientific Computing65.1 (2015), pp. 34–53./d.sc/o.sc/i.sc: 10.1007/s10915-014-9952-8

    2015 doi

  20. [31]

    In:SIAM Journal on Numerical Analysis56.2 (2018), pp

    V.Linders,T.LundquistandJ.Nordström.‘OntheorderofAccuracyofFiniteDifferenceOp- erators on Diagonal Norm Based Summation-By-Parts Form’. In:SIAM Journal on Numerical Analysis56.2 (2018), pp. 1048–1063./d.sc/o.sc/i.sc: 10.1137/17M1139333

    2018 doi

  21. [32]

    Linders, J

    V. Linders, J. Nordström and S. H. Frankel.Convergence and stability properties of summation- by-parts in time. Technical Report LiTH-MAT-R, ISSN 0348-2960; 2019:4. Linköping, Sweden: Linköping University, Apr. 2019

    2019

  22. [33]

    Lipnikov, G

    K. Lipnikov, G. Manzini and M. Shashkov. ‘Mimetic finite difference method’. In:Journal of Computational Physics257 (2014), pp. 1163–1227./d.sc/o.sc/i.sc: 10.1016/j.jcp.2013.07.031

    2014 doi

  23. [34]

    91–111./d.sc/o.sc/i.sc: 10.1016/j.jcp.2013.12.041

    K.Mattsson,M.AlmquistandM.H.Carpenter.‘Optimaldiagonal-normSBPoperators’.In: Journal of Computational Physics264 (2014), pp. 91–111./d.sc/o.sc/i.sc: 10.1016/j.jcp.2013.12.041

    2014 doi

  24. [35]

    Mattsson, M

    K. Mattsson, M. Almquist and E. van der Weide. ‘Boundary optimized diagonal-norm SBP operators’.In:Journalofcomputationalphysics 374(2018),pp.1261–1266. /d.sc/o.sc/i.sc: 10.1016/j.jcp. 2018.06.010

    2018 doi

  25. [36]

    Mattsson and J

    K. Mattsson and J. Nordström. ‘Summation by parts operators for finite difference approx- imationsofsecondderivatives’.In: JournalofComputationalPhysics 199.2(2004),pp.503–540. /d.sc/o.sc/i.sc: 10.1016/j.jcp.2004.03.001

    2004 doi

  26. [37]

    Mattsson and O

    K. Mattsson and O. O’Reilly. ‘Compatible diagonal-norm staggered and upwind SBP oper- ators’. In:Journal of Computational Physics352 (2018), pp. 52–75./d.sc/o.sc/i.sc: 10.1016/j.jcp.2017. 09.044

    2018 doi

  27. [38]

    In:Applied Numerical Mathematics38.3 (2001), pp

    J.NordströmandM.Björck.‘Finitevolumeapproximationsandstrictstabilityforhyperbolic problems’. In:Applied Numerical Mathematics38.3 (2001), pp. 237–255./d.sc/o.sc/i.sc: 10.1016/S0168- 9274(01)00027-7

    2001 doi

  28. [39]

    Nordström, K

    J. Nordström, K. Forsberg, C. Adamsson and P. Eliasson. ‘Finite volume methods, unstruc- tured meshes and strict stability for hyperbolic problems’. In:Applied Numerical Mathematics 45.4 (2003), pp. 453–473./d.sc/o.sc/i.sc: 10.1016/S0168-9274(02)00239-8. 27

    2003 doi

  29. [40]

    O’Reilly, T

    O. O’Reilly, T. Lundquist, E. M. Dunham and J. Nordström. ‘Energy stable and high-order- accurate finite difference methods on staggered grids’. In:Journal of Computational Physics 346 (2017), pp. 572–589./d.sc/o.sc/i.sc: 10.1016/j.jcp.2017.06.030

    2017 doi

  30. [41]

    C. C. Paige and M. A. Saunders. ‘Algorithm 583 LSQR: Sparse Linear Equations and Least SquaresProblems’.In:ACMTransactionsonMathematicalSoftware(TOMS) 8.2(1982),pp.195–

    1982

  31. [42]

    C. C. Paige and M. A. Saunders. ‘LSQR: An Algorithm for Sparse Linear Equations and Sparse Least Squares’. In:ACM Transactions on Mathematical Software (TOMS)8.1 (1982), pp. 43–71./d.sc/o.sc/i.sc: 10.1145/355984.355989

    1982

  32. [43]

    H. Ranocha. ‘Mimetic Properties of Difference Operators: Product and Chain Rules as for Functions of Bounded Variation and Entropy Stability of Second Derivatives’. In:BIT Nu- merical Mathematics59.2 (June 2019), pp. 547–563./d.sc/o.sc/i.sc: 10.1007/s10543-018-0736-7. arXiv: 18...

    2019 arXiv

  33. [44]

    H. Ranocha. ‘Shallow water equations: Split-form, entropy stable, well-balanced, and posit- ivity preserving numerical methods’. In:GEM – International Journal on Geomathematics8.1 (Apr. 2017), pp. 85–133./d.sc/o.sc/i.sc: 10.1007/s13137-016-0089-9. arXiv:1609.08029 [math.NA]

    2017 arXiv

  34. [45]

    H. Ranocha. ‘Some Notes on Summation by Parts Time Integration Methods’. In:Results in Applied Mathematics1 (June 2019), p. 100004./d.sc/o.sc/i.sc: 10.1016/j.rinam.2019.100004 . arXiv: 1901.08377 [math.NA]

    2019

  35. [46]

    H.Ranocha,P.ÖffnerandT.Sonar.‘Extendedskew-symmetricformforsummation-by-parts operatorsandvaryingJacobians’.In: JournalofComputationalPhysics 342(Apr.2017),pp.13–

    2017

  36. [47]

    In:Journal of Computational Physics311 (Apr

    H.Ranocha,P.ÖffnerandT.Sonar.‘Summation-by-partsoperatorsforcorrectionprocedure via reconstruction’. In:Journal of Computational Physics311 (Apr. 2016), pp. 299–328./d.sc/o.sc/i.sc: 10.1016/j.jcp.2016.02.009. arXiv:1511.02052 [math.NA]

    2016 arXiv

  37. [48]

    2019_SBP_vector_calculus_REPRO.DiscreteVec- torCalculusandHelmholtzHodgeDecompositionforClassicalFiniteDifferenceSummationbyParts Operators

    H.Ranocha,K.OstaszewskiandP.Heinisch. 2019_SBP_vector_calculus_REPRO.DiscreteVec- torCalculusandHelmholtzHodgeDecompositionforClassicalFiniteDifferenceSummationbyParts Operators. https://github.com/IANW-Projects/2019_SBP_vector_calculus_REPRO.Aug

  38. [49]

    Ranocha, K

    H. Ranocha, K. Ostaszewski and P. Heinisch.Numerical Methods for the Magnetic Induction EquationwithHallEffectandProjectionsontoDivergence-FreeVectorFields .Submitted.Oct.2018. arXiv: 1810.01397 [math.NA]

    2018 arXiv

  39. [50]

    D. D. Schnack. Lectures in Magnetohydrodynamics With an Appendix on Extended MHD. Berlin Heidelberg: Springer, 2009./d.sc/o.sc/i.sc: 10.1007/978-3-642-00688-3

    2009 doi

  40. [51]

    Schweizer

    B. Schweizer. ‘On Friedrichs inequality, Helmholtz decomposition, vector potentials, and thediv-curllemma’.In:TrendsinApplicationsofMathematicstoMechanics .Ed.byE.Rocca,U. Stefanelli,L.TruskinovskyandA.Visintin.Vol.27.SpringerINdAMSeries.Cham:Springer, 2018, pp. 65–79./d.sc/o....

    2018 doi

  41. [52]

    arXiv:1511.08408 [math.NA]

    /d.sc/o.sc/i.sc: 10.1016/j.jcp.2017.04.044. arXiv:1511.08408 [math.NA]

    2017 arXiv

  42. [53]

    J. Sims, M. Giorgi, M. Oliveira, J. Meneghetti and M. Gutierrez. ‘Directional analysis of car- diac motion field from gated fluorodeoxyglucose PET images using the Discrete Helmholtz Hodge Decomposition’. In:Computerized Medical Imaging and Graphics65 (2018), pp. 69–78. /d.sc/o....

    2018 doi

  43. [54]

    Sjögreen, H

    B. Sjögreen, H. C. Yee and D. Kotov. ‘Skew-symmetric splitting and stability of high order central schemes’. In:Journal of Physics: Conference Series. Vol. 837. 1. IOP Publishing. 2017, p. 012019./d.sc/o.sc/i.sc: 10.1088/1742-6596/837/1/012019. 28

    2017 doi

  44. [55]

    47–67./d.sc/o.sc/i.sc: 10.1006/jcph.1994.1005

    B.Strand.‘SummationbyPartsforFiniteDifferenceApproximationsfor d/dx’.In:Journalof Computational Physics110.1 (1994), pp. 47–67./d.sc/o.sc/i.sc: 10.1006/jcph.1994.1005

    1994

  45. [56]

    823–830./d.sc/o.sc/i.sc: 10.1007/s10543-014-0471-7

    M.Svärd.‘Anoteon L∞ boundsandconvergenceratesofsummation-by-partsschemes’.In: BIT Numerical Mathematics54.3 (2014), pp. 823–830./d.sc/o.sc/i.sc: 10.1007/s10543-014-0471-7

    2014 doi

  46. [57]

    29–42./d.sc/o.sc/i.sc: 10.1023/A:1025881528802

    M.Svärd.‘OnCoordinateTransformationsforSummation-by-PartsOperators’.In: Journalof Scientific Computing20.1 (2004), pp. 29–42./d.sc/o.sc/i.sc: 10.1023/A:1025881528802

    2004 doi

  47. [58]

    Svärd and J

    M. Svärd and J. Nordström.On the convergence rates of energy-stable finite-difference schemes. TechnicalReportLiTH-MAT-R–2017/14–SE.Linköping,Sweden:LinköpingUniversity,Oct. 2017

    2017

  48. [59]

    Z. J. Silberman, T. R. Adams, J. A. Faber, Z. B. Etienne and I. Ruchlin. ‘Numerical generation of vector potentials from specified magnetic fields’. In:Journal of Computational Physics379 (2019), pp. 421–437./d.sc/o.sc/i.sc: 10.1016/j.jcp.2018.12.006

    2019 doi

  49. [60]

    Svärd and J

    M. Svärd and J. Nordström. ‘Review of summation-by-parts schemes for initial-boundary- value problems’. In:Journal of Computational Physics268 (2014), pp. 17–38./d.sc/o.sc/i.sc: 10.1016/j. jcp.2014.02.031. 29

    2014 doi

  50. [66]

    /d.sc/o.sc/i.sc: 10.1016/j.jcp.2006.02.014

    M.SvärdandJ.Nordström.‘Ontheorderofaccuracyfordifferenceapproximationsofinitial- boundaryvalueproblems’.In:JournalofComputationalPhysics 218.1(2006),pp.333–352. /d.sc/o.sc/i.sc: 10.1016/j.jcp.2006.02.014

    2006 doi

  51. [209]

    /d.sc/o.sc/i.sc: 10.1145/355993.356000

  52. [681]

    /d.sc/o.sc/i.sc: 10.1029/GL010i008p00678

    doi

  53. [817]

    /d.sc/o.sc/i.sc: 10.1016/0032-0633(88)90086-4

    doi

  54. [864]

    /d.sc/o.sc/i.sc: 10.1002/(SICI)1099-1476(199806)21:9<823::AID-MMA976>3.0.CO;2-B

    doi

  55. [1404]

    /d.sc/o.sc/i.sc: 10.1109/TVCG.2012.316

    2012 doi

  56. [2019]

    /d.sc/o.sc/i.sc: 10.5281/zenodo.3375170

    doi

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.