Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

A structure preserving H-curl algebraic multigrid method for the eddy current equations

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The paper introduces an algebraic multigrid method that enforces an edge-gradient commuting identity on every grid level, producing mesh-independent convergence for eddy-current systems.

desk verdict A genuinely new structure-preserving H-curl AMG that cleverly reuses energy minimization, but its advertised generality hinges on an unspecified safeguard in Algorithm 4.1 that needs a concrete fix. read the letter →

arxiv 2506.08284 v1 pith:XN26CWXX submitted 2025-06-09 math.NA cs.NA

Raymond Tuminaro , Christian Glusa This is my paper
classification math.NAcs.NA MSC 65N5565F08
keywords algebraicmultigridH-curleddycurrentequationsenergyminimizationcommutingrelationshipstructurepreservingnull-spacepreservationMaxwell's
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 introduces SpHcurlAMG, an algebraic multigrid method for the linear systems arising from the eddy-current form of Maxwell's equations. Its central claim is that the discrete commuting relationship $P^{(e)} D_H^{(n\to e)} = D_h^{(n\to e)} P^{(n)}$ can be enforced on every multigrid level even when the nodal interpolation $P^{(n)}$ is a general, high-quality operator rather than the piecewise-constant operator that earlier structure-preserving methods required. The paper shows that this relationship can be written as linear constraints inside an energy-minimization construction of the edge prolongator $P^{(e)}$, and that the resulting constrained least-squares problems are tiny because the discrete gradient is sparse. If the claim holds, the method preserves the null space of the curl-curl operator at every level while retaining the interpolation quality needed for mesh-independent convergence.

What carries the argument

The key machinery is constrained energy minimization with the commuting relationship as the constraint set. Starting from a nodal AMG hierarchy for the scalar Poisson-like problem, the algorithm converts the nodal prolongator $P^{(n)}$ to a piecewise-constant operator, uses it with the fine discrete gradient $D_h^{(n\to e)}$ to build coarse edges and the coarse gradient $D_H^{(n\to e)}$, then constructs the edge prolongator by minimizing interpolation basis energy subject to $P^{(e)}D_H^{(n\to e)} = D_h^{(n\to e)}P^{(n)}$ and a sparsity pattern derived from $|D_h^{(n\to e)}||P^{(n)}||D_H^{(n\to e)}|^T$. Because each row's constraint system is a directed graph with a null space spanned by the constant vector, the projection step reduces to independent small QR or normal-equation solves; the algorithm also augments the coarse graph with extra edges when a local subgraph is not irreducible, guaranteeing a feasible solution to the commuting equation.

What would settle it

Construct an irregular 3D mesh and inspect the output of Algorithm 4.1: if any coarse column receives no fine rows, or if Algorithm 4.2 finds a fine-edge row whose coarse-edge subgraph cannot be made irreducible by adding candidate edges, the commuting identity has no solution and the method cannot form $P^{(e)}$.

Watch

Extended reading notes

Core claim

The central discovery is that the commuting identity $P^{(e)} D_H^{(n\to e)} = D_h^{(n\to e)} P^{(n)}$ does not force the nodal interpolation to be piecewise constant. The paper shows this constructively by treating the identity as a constraint in an energy-minimizing AMG construction of $P^{(e)}$: each column of $P^{(e)}$ is chosen to minimize its energy in the $S_h^{(e)}$ semi-norm subject to the commutation constraints and a sparsity pattern built from the coarse-edge graph. The constraint blocks are directed graphs whose Gram matrices are graph Laplacians, so each row's correction is a small, rank-known least-squares problem rather than a costly global solve. On the tested 2D and 3D constant-coefficient problems, the resulting preconditioned CG iteration counts stay flat as the mesh is refined, with operator complexity near 1, and on two strong-material-variation simulation problems the method reduces iteration counts sharply. In certain 2D refinement settings, the energy-minimized edge prolongator is exactly the ideal linear-edge-element interpolation operator.

Load-bearing premise

Algorithm 4.1, which converts a general nodal prolongator into a piecewise-constant one for building the coarse edge graph, is assumed to always produce a valid aggregation with every coarse node receiving at least one fine node; the paper notes that the safeguards guaranteeing this are not shown.

Editorial extensions

If this is right

  • The null-space property $S_H^{(e)} D_H^{(n\to e)} = \Theta_H$ holds on every level, so coarse-grid curl-curl systems have the same kernel structure as the fine grid.
  • Standard high-quality nodal interpolation, such as smoothed aggregation, can be used without sacrificing structure preservation, removing the piecewise-constant limitation of earlier structure-preserving AMG.
  • On constant-coefficient 2D and 3D model problems, preconditioned CG iteration counts are essentially flat under refinement and match geometric multigrid, with operator complexity between about 1.05 and 1.20.
  • The constraint solves are cheap: the largest least-squares systems in the reported tests have column dimension 8 in 3D and 6 in 2D, comparable to the cost of energy-minimization prolongators for elasticity.
  • On 2D uniform-refinement cases with linear nodal interpolation, the energy-minimized edge prolongator coincides with the ideal linear edge element interpolation operator.

Reading between the lines

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

  • I infer that the same constrained-energy-minimization template should transfer to H(div) problems or other de Rham complexes with a commuting diagram, since the constraint matrix is a graph Laplacian rather than a dense mode set.
  • The paper's stated motivation for general nodal interpolation is weak-connection dropping; a testable extension is to run SpHcurlAMG on anisotropic stretched meshes with aggressive dropping and compare its iteration counts against auxiliary-space H-curl solvers.
  • An extension the paper mentions but does not implement is recycling the least-squares factors across a sequence of linear solves that share the same sparsity patterns, which would reduce setup cost in time-stepping or nonlinear simulations.
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

4 major / 6 minor

Summary. This paper proposes SpHcurlAMG, an algebraic multigrid preconditioner for discrete H(curl) eddy current systems. The method extends the Reitzinger and Schoberl structure-preserving AMG by enforcing the commuting relation P^{(e)} D_H = D_h P^{(n)} on every level, but with the edge prolongator obtained by a constrained energy-minimization process (Eq. 4.2) rather than the piecewise-constant nodal interpolation required by the earlier method. A nodal AMG hierarchy is generated for an associated scalar problem; an auxiliary piecewise-constant operator is derived by an ad-hoc conversion algorithm (Algorithm 4.1) to define coarse gradient matrices, and the edge prolongator is constructed via local least-squares/QR projections (Algorithms 4.2 and 4.3). Numerical tests on 2D/3D model problems and on two ALEGRA problems show iteration counts that are roughly constant over mesh refinement and operator complexities near one. The paper also discusses an 'ideal behavior' in 2D and gives statistics on the sizes of the local least-squares systems.

Significance. If the construction can be made rigorous and reproducible, this is a valuable contribution: it removes the main limitation of RSAMG, provides a flexible way to combine H(curl) AMG with arbitrary nodal AMG hierarchies (including those with aggressive dropping), and appears to achieve convergence rates comparable to geometric multigrid at low operator complexity. The constrained energy-minimization formulation (4.2) is elegant, and the observation that the projection subproblems reduce to small graph-Laplacian/QR systems is useful and quantified in Tables 2 and 4. The paper is reasonably honest about open points: Section 5 defers a proof of the ideal behavior, Algorithm 4.1 admits missing safeguards, and the paper states that credible setup timings are not yet available. Since the core heuristic lacks a guarantee and the numerical evidence is not fully clean under tolerance modifications, the contribution is not yet fully established, but the main idea is plausible and worth publishing after substantial revision.

major comments (4)
  1. [Section 4.1, Algorithm 4.1] The construction of P_const is load-bearing for the whole approach, but the algorithm contains only the caveat 'Safeguards (not shown) needed to guarantee each column has some row assignments.' If a column of P_const is empty in a practical run, then the corresponding coarse node has no incident edge in the coarse gradient D_H built via Eq. (4.3), while the right-hand side D_h^{(n->e)} P^{(n)} of the commuting relation can have nonzero entries in that column. Then no edge prolongator P^{(e)} can satisfy Eq. (2.3), and the structure-preserving property fails. Algorithm 4.2's make-irreducible step can add edges to a local subgraph but cannot create a missing coarse node. Because all numerical experiments in Tables 1, 3, 5, and 6 use this conversion, the central claim that SpHcurlAMG 'can leverage fairly general and more sophisticated nodal interpolation operators' is not supported without a guarantee that the conversion produces a valid aggregation. Please specify the missing safeguards, prove or numerically demonstrate (over a nontrivial ensemble, including classical AMG hierarchies with dropped weak connections) that every aggregate is nonempty and that the resulting coarse edge graph is adequate for the commuting constraints.
  2. [Section 6.1, Tables 1 and 3; Section 6.2, Tables 5 and 6] The reported iteration counts are obtained under problem-dependent tolerance scalings based on condition number estimates. For example, the text states that for the sigma=1e-2 triangular mesh the tolerance is relaxed by factors ranging from roughly 12 to 17,000 as the mesh is refined, and the ALEGRA tables use tolerances experimentally set to kappa(D^{-1}A)/(5e17). With such tolerances, a fixed number of CG iterations does not represent convergence to a common residual reduction, and comparisons with geometric multigrid (e.g., '8 iterations or 6 iterations' in Section 6.1) are not meaningful. Please report the actual residual reductions achieved, run all cases to a common tolerance, or report per-V-cycle convergence factors; otherwise the mesh-independence claim rests on a looser-than- apparent stopping criterion.
  3. [Section 5] The 'ideal behavior in two dimensions' is presented as a theoretical property of SpHcurlAMG, but no theorem is stated or proved; the text says 'We defer a detailed proof to a future monograph.' The argument around Eq. (5.1) assumes, among other things, that a feasible initial guess is the ideal eddy-current interpolant and that the saddle-point system has a unique solution, yet the authors themselves note non-uniqueness because S_h^{(e)} is singular. Since this section is one of the paper's advertised contributions, either turn it into a clearly stated theorem with a proof, or explicitly label it as a numerical observation and remove it from the technical claims. As written, the proof sketch does not establish the stated ideal behavior.
  4. [Section 4.2, Eq. (4.4) and Algorithm 4.2] The consistency of the local commutator sub-systems is asserted to follow from the nodal prolongator preserving constants, but the paper then says that sparsity-pattern entries might be missed due to cancellation and that this is avoided 'by adding a small random perturbation' to D_h^{(n->e)} or P^{(n)} when constructing the sparsity pattern. Algorithm 4.2 instead uses absolute values. These two mechanisms are not equivalent, and neither is given a correctness argument. Accidental or engineered cancellation is a real possibility for nonsymmetric or varying-sign nodal interpolation, and a failure here means the local least-squares system is inconsistent, so Eq. (2.3) is not enforced. Please provide a deterministic construction of the sparsity pattern that provably contains all coarse nodes contributing to r_{i:}, or prove that the absolute-value/perturbation procedure always yields a consistent sub-system.
minor comments (6)
  1. [Section 4, Eq. (4.2)] The factor '1/2' appears outside the arg min, so the equation as written evaluates to half of the minimizer; the factor should be inside the objective function as in Eq. (3.1).
  2. [Section 3, Eq. (3.2)] The notation 'y=(Y' and '⌢p' is garbled in the typeset version; please use standard vec and inverse-vec notation or spell out the conversions in words.
  3. [Section 4.1, Eq. (4.3)] The notation switches between \bar P^{(n)} in Eq. (4.3) and P_const in Algorithm 4.1; please unify the notation and state explicitly that the same operator is meant.
  4. [Algorithm 4.2] The definition of the diagonal matrix W is typeset with unreadable symbols ('# »3− ...'); please rewrite the formula clearly, for example W_ii = 2 for edges with one Dirichlet endpoint and W_ii = 1 otherwise.
  5. [Section 6.2] There is a typo 'It it possible' in the paragraph discussing the relaxation-only solves for the snake problem.
  6. [Section 6.2] The comparison in Table 6 between a 3-level AMG solve and two Hiptmair relaxation sweeps is not a like-for-like work comparison, since a V-cycle contains more relaxation and coarse-grid work; please clarify what the iteration-count ratio is intended to show.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the commuting relation is enforced by construction and convergence is empirically benchmarked, not derived from fitted inputs or self-citation chains.

full rationale

The paper's central claim is that SpHcurlAMG enforces the discrete commuting relationship P^{(e)} D_H = D_h P^{(n)} while using general nodal interpolation. This is presented as a construction, not as a prediction derived from the inputs. D_H is assembled from a piecewise-constant projection of P^{(n)} via equation (4.3), and P^{(e)} is then computed subject to the commuting constraint in the energy-minimization formulation (4.2). The structure-preserving property S_H D_H = 0 is an algebraic consequence shown in equations (2.3)-(2.4), using the fine-level identity S_h D_h = 0. No fitted parameter is renamed as a prediction: the numerical results in Tables 1, 3, 5, and 6 are empirical convergence comparisons against geometric multigrid and relaxation-only preconditioners. The paper does cite prior work by one of its authors, notably the energy-minimization framework of reference [27], but the relevant equations and algorithmic steps are reproduced in the manuscript itself, and the citation is used as background methodology rather than as an unverified premise that forces the conclusion. The self-citations are not used to invoke a uniqueness theorem or to forbid alternative constructions. Algorithm 4.1 does contain a heuristic for converting a general nodal prolongator into a piecewise-constant one, with the paper explicitly noting that safeguards are needed, but an unspecified implementation detail is a robustness gap, not a circular derivation. Overall, the derivation chain is self-contained: the claimed structure preservation is built into the algorithm by definition and the convergence behavior is measured against external benchmarks.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The method depends on the nodal AMG hierarchy preserving constants, the edge element null-space property, and a heuristic piecewise-constant conversion that defines the coarse edge graph. No new physical entities are introduced.

free parameters (5)
  • damping parameter omega for energy minimization Jacobi smoothing = 0.5 (model problems), 0.3 (ALEGRA problems)
    Chosen by hand; affects convergence rate and is not derived from theory.
  • number of energy minimization iterations per level = 1 (occasionally 2)
    Chosen by hand; more iterations can reduce iteration count by one in some cases.
  • drop tolerance for nodal smoothed aggregation = 0 (model), 0.0025 (snake), 0.04 (coax)
    Chosen by hand to respect material boundaries; affects coarse grid quality and convergence.
  • tolerance scaling factor for CG stopping criterion = kappa/(5e17) for ALEGRA problems
    Explicitly described as 'experimentally determined as a reasonable threshold'; affects reported iteration counts.
  • targetNzPerCol and nPasses in Algorithm 4.1 = not specified
    Free design choices that influence the constructed piecewise-constant prolongator; no values or formulas are given.
assumptions (4)
  • standard math The discrete gradient null-space property S_h^{(e)} D_h^{(n->e)} = 0 holds for first-order edge element discretizations.
    Invoked in Section 2, equation (1.3) and used to prove coarse-level structure preservation.
  • domain assumption The nodal prolongator P^{(n)} is modified so that all row sums equal one (preserves constants), making the commuting equation consistent.
    Stated in Section 4.2; required for feasibility of the commutator constraints.
  • domain assumption The coarse edge graph generated by Algorithm 4.2 can always be made irreducible by adding edges, yielding a feasible solution to the commuting constraints.
    Section 4.2 and Algorithm 4.2; the authors say this situation is 'very rare' but provide no proof of existence or of convergence impact.
  • ad hoc to paper The 'ideal behavior' in 2D (Section 5) is assumed or observed, not proven; the paper defers a detailed proof to a future monograph.
    Underpins the claim that the method can produce ideal interpolation operators in special cases, but this is not load-bearing for the main numerical results.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A structure preserving H-curl algebraic multigrid method for the eddy current equations." pith.science (2026). https://pith.science/paper/XN26CWXX

@misc{pith2026250608284,
  author       = {Pith},
  title        = {Pith review of: A structure preserving H-curl algebraic multigrid method for the eddy current equations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XN26CWXX}},
  note         = {Machine review of arXiv:2506.08284}
}
read the original abstract

A new algebraic multigrid method (AMG) is presented for solving the linear systems associated with the eddy current approximation to the Maxwell equations. This AMG method extends an idea proposed by Reitzinger and Schoberl. The main feature of the Reitzinger and Schoberl algorithm (RSAMG) is that it maintains null-space properties of the Curl-Curl operator throughout all levels of the AMG hierarchy. It does this by enforcing a commuting relationship involving grid transfers and the discrete gradient operator. This null-space preservation property is critical to the algorithm's success, however enforcing this commuting relationship is non-trivial except in the special case where one leverages a piece-wise constant nodal interpolation operator. For this reason, mesh independent convergence rates are generally not observed for RSAMG due to its reliance on sub-optimal piece-wise constant interpolation. We present a new AMG algorithm that enforces the same commuting relationship. The main advance is that the new structure preserving H-curl algorithm (SpHcurlAMG) does not rely on piece-wise constant interpolation and can leverage fairly general and more sophisticated nodal interpolation operators. The key idea is to employ energy minimization AMG (EAMG) to construct edge interpolation grid transfers and to enforce the commuting relationship by embedding it as constraints within an EAMG procedure. While it might appear that solving such a constrained energy minimization is costly, we illustrate how this is not the case in our context. Numerical results are then given demonstrating mesh independent convergence over a range of test problems.

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. A black-box, multilevel algebraic preconditioning framework for conforming finite elements

    math.NA 2026-07 conditional novelty 7.0 of 10

    Conforming finite-element assembly naturally induces element-local Gram representations, enabling black-box multilevel algebraic preconditioning for H(div), H^2, and elasticity problems where classical AMG fails.

Reference graph

Works this paper leans on

37 extracted references · 33 canonical work pages · cited by 1 Pith paper

  1. [1]

    D. N. Arnold, R. S. Falk, and R. Winther , Multigrid in h( div ) and h( curl ) , Numer. Math., 85 (2000), pp. 197--217

    work page 2000

  2. [2]

    Bell and L

    N. Bell and L. N. Olson , Algebraic multigrid for k-form laplacians , Numerical Linear Algebra with Applications, 15 (2008), pp. 165--185, https://doi.org/https://doi.org/10.1002/nla.577, https://onlinelibrary.wiley.com/doi/abs/10.1002/nla.577, https://arxiv.org/abs/https://onlinelibrary.wiley.com/doi/pdf/10.1002/nla.577

    work page doi:10.1002/nla.577 2008

  3. [3]

    Bochev, C

    P. Bochev, C. Garasi, J. Hu, A. Robinson, and R. Tuminaro , An improved algebraic multigrid method for solving M axwell's equations , SIAM J. Sci. Comput., 25 (2003), pp. 623--642

    work page 2003

  4. [4]

    Bochev, J

    P. Bochev, J. Hu, A. Robinson, and R. Tuminaro , Towards robust 3 D Z -pinch simulations: discretization and fast solvers for magnetic diffusion in heterogeneous conductors , Elec. Trans. Numer. Anal., 15 (2003)

    work page 2003

  5. [5]

    Bochev, J

    P. Bochev, J. Hu, C. Siefert, and R. Tuminaro , An algebraic multigrid approach based on a compatible gauge reformulation of M axwell's equations , SIAM J. Sci. Comput., 31 (2008), pp. 557--583, https://doi.org/http://dx.doi.org/10.1137/070685932

    work page doi:10.1137/070685932 2008

  6. [6]

    Brandt , General highly accurate algebraic coarsening , Electronic Trans

    A. Brandt , General highly accurate algebraic coarsening , Electronic Trans. Num. Anal, 10 (2000), pp. 1--20

    work page 2000

  7. [7]

    Brandt , Multiscale scientific computation: Review 2001 , in Multiscale and Multiresolution Methods, Springer Verlag, 2001, pp

    A. Brandt , Multiscale scientific computation: Review 2001 , in Multiscale and Multiresolution Methods, Springer Verlag, 2001, pp. 1--96

    work page 2001

  8. [8]

    Brandt, J

    A. Brandt, J. Brannick, K. Kahl, and I. Livshits , Bootstrap AMG , SIAM Journal on Scientific Computing, 33 (2011), pp. 612--632

    work page 2011

Show all 37 references
  1. [9]

    Brandt, J

    A. Brandt, J. Brannick, K. Kahl, and I. Livshits , Bootstrap algebraic multigrid: Status report, open problems, and outlook , Numerical Mathematics: Theory, Methods and Applications, 8 (2015), pp. 112--135, https://doi.org/https://doi.org/10.4208/nmtma.2015.w06si, https://glob...

    2015 doi

  2. [10]

    Brandt, S

    A. Brandt, S. McCormick, and J. Ruge , Algebraic multigrid ( AMG ) for sparse matrix equations , in Sparsity and Its Applications, D. J. Evans, ed., Cambridge University Press, Cambridge, 1984

    1984

  3. [11]

    Brannick and L

    J. Brannick and L. Zikatanov , Algebraic multigrid methods based on compatible relaxation and energy minimization , in Domain Decomposition Methods in Science and Engineering XVI, O. Widlund and D. E. Keyes, eds., vol. 55 of Lecture Notes in Computational Science and Engineeri...

    2007

  4. [12]

    Brezina, R

    M. Brezina, R. Falgout, S. MacLachlan, T. Manteuffel, S. McCormick, and J. Ruge , Adaptive smoothed aggregation ( SA ) multigrid , SIAM Rev., 47 (2005), pp. 317--346 (electronic)

    2005

  5. [13]

    D'ambra, S

    P. D'ambra, S. Filippone, and P. S. Vassilevski , Bootcmatch: A software package for bootstrap amg based on graph weighted matching , ACM Trans. Math. Softw., 44 (2018), https://doi.org/10.1145/3190647, https://doi.org/10.1145/3190647

    2018 doi

  6. [14]

    De Sterck, R

    H. De Sterck, R. D. Falgout, J. W. Nolting, and U. M. Yang , Distance-two interpolation for parallel algebraic multigrid , Numerical Linear Algebra with Applications, 15 (2008), pp. 115--139, http://doi.wiley.com/10.1002/nla.559 (accessed 2021-06-22)

    2008 doi

  7. [15]

    Hiptmair , Multigrid method for H( div ) in three dimensions , Electron

    R. Hiptmair , Multigrid method for H( div ) in three dimensions , Electron. Trans. Numer. Anal., 6 (1997), pp. 133--152. Special issue on multilevel methods (Copper Mountain, CO, 1997)

    1997

  8. [16]

    Hiptmair , Multigrid method for M axwell's equations , SIAM J

    R. Hiptmair , Multigrid method for M axwell's equations , SIAM J. Numer. Anal., 36 (1998), pp. 204--225

    1998

  9. [17]

    Hiptmair and J

    R. Hiptmair and J. Xu , Nodal auxiliary space preconditioning in H(curl) and H(div) spaces , SIAM J. on Numer. Anal., 45 (2007), pp. 2483--2509

    2007

  10. [18]

    J. Hu, R. Tuminaro, P. Bochev, C. Garasi, and A. Robinson , Toward an h-independent algebraic multigrid method for M axwell's equations , SIAM J. Sci. Comput., 27 (2006), pp. 1669--1688

    2006

  11. [19]

    Janna, A

    C. Janna, A. Franceschini, J. B. Schroder, and L. Olson , Parallel energy-minimization prolongation for algebraic multigrid , SIAM Journal on Scientific Computing, 45 (2023), pp. A2561--A2584, https://doi.org/10.1137/22M1513794, https://doi.org/10.1137/22M1513794, https://arxi...

    2023 doi

  12. [20]

    Jones and B

    J. Jones and B. Lee , A multigrid method for variable coefficient M axwell's equations , SIAM J. Sci. Comput., 27 (2006), pp. 1689--1708

    2006

  13. [21]

    Kolev, J

    T. Kolev, J. Pasciak, and P. Vassilevski , H (curl) auxiliary mesh preconditioning , Numerical Linear Algebra with Applications, 15 (2008), pp. 455--471

    2008

  14. [22]

    T. V. Kolev and P. S. Vassilevski , A MG by element agglomeration and constrained energy minimization interpolation , Numer. Linear Algebra Appl., 13 (2006), pp. 771--788

    2006

  15. [23]

    T. V. Kolev and P. S. Vassilevski , Parallel auxiliary space amg for h(curl) problems , Journal of Computational Mathematics, 27 (2009), pp. 604--623, http://www.jstor.org/stable/43693530 (accessed 2025-04-30)

    2009

  16. [24]

    T. V. Kolev and P. S. Vassilevski , Parallel auxiliary space amg solver for \ h(div)\ problems , SIAM Journal on Scientific Computing, 34 (2012), pp. A3079--A3098, https://doi.org/10.1137/110859361, https://doi.org/10.1137/110859361, https://arxiv.org/abs/https://doi.org/10.11...

    2012 doi

  17. [25]

    Mandel, M

    J. Mandel, M. Brezina, and P. Van e k , Energy optimization of algebraic multigrid bases , Computing, 62 (1999), pp. 205--228

    1999

  18. [26]

    J. H. Niederhaus, S. W. Bova, J. B. Carleton, J. H. Carpenter, K. R. Cochrane, M. M. Crockatt, W. Dong, T. J. Fuller, B. N. Granzow, D. A. Ibanez, S. R. Kennon, C. B. Luchini, R. J. Moral, C. J. O'Brien, M. J. Powell, A. C. Robinson, A. E. Rodriguez, J. J. Sanchez, W. A. Scott...

    2023

  19. [27]

    Olson, J

    L. Olson, J. Schroder, and R. Tuminaro , A general interpolation strategy for algebraic multigrid using energy minimization , SIAM Journal on Scientific Computing, 33 (2011), p. 966

    2011

  20. [28]

    Reitzinger and J

    S. Reitzinger and J. Sch\" o berl , An algebraic multigrid method for finite element discretizations with edge elements , Numer. Linear Algebra Appl., 9 (2002), pp. 223--238

    2002

  21. [29]

    A. C. Robinson, R. R. Drake, C. B. Luchini, R. J. Moral, J. H. Niederhaus, and S. V. Petney , An MPMD approach coupling electromagnetic continuum mechanics approximations in ALEGRA , Computer Methods in Applied Mechanics and Engineering, 429 (2024), p. 117164, https://doi.org/...

    2024

  22. [30]

    A. C. Robinson and et. al . , ALEGRA : An arbitrary L agrangian- E ulerian multimaterial, multiphysics code , in AIAA 2008-1235 46th AIAA Aerospace Sciences Meeting and Exhibit, Reno, NV, 2008

    2008

  23. [31]

    Ruge and K

    J. Ruge and K. St \"u ben , Algebraic multigrid ( AMG ) , in Multigrid Methods, S. F. McCormick, ed., vol. 3 of Frontiers in Applied Mathematics, SIAM, Philadelphia, PA, 1987, pp. 73--130

    1987

  24. [32]

    Van e k, M

    P. Van e k, M. Brezina, and J. Mandel , Convergence of algebraic multigrid based on smoothed aggregation , Numerische Mathematik, 88 (2001), pp. 559--579

    2001

  25. [33]

    Van e k, J

    P. Van e k, J. Mandel, and M. Brezina , Algebraic multigrid by smoothed aggregation for second and fourth order elliptic problems , Computing, 56 (1996), pp. 179--196

    1996

  26. [34]

    P. S. Vassilevski , General constrained energy minimization interpolation mappings for AMG , SIAM J. Sci. Comput., 32 (2010), pp. 1--13

    2010

  27. [35]

    Wagner , On the algebraic construction of multilevel transfer operators , Computing, 65 (2000), pp

    C. Wagner , On the algebraic construction of multilevel transfer operators , Computing, 65 (2000), pp. 73--95

    2000

  28. [36]

    W. L. Wan, T. F. Chan, and B. Smith , An energy-minimizing interpolation for robust multigrid methods , SIAM J. Sci. Comput., 21 (2000), pp. 1632--1649

    2000

  29. [37]

    Xu and L

    J. Xu and L. Zikatanov , On an energy minimizing basis for algebraic multigrid methods , Computing and Visualization in Science, 7 (October 2004), pp. 121--127(7)

    2004

Pith tools

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