Pith. sign in

REVIEW 3 major objections 5 minor 59 references

Robust spectral preconditioning for high-P\'{e}clet number convection-diffusion

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read A two-level spectral restricted additive Schwarz preconditioner solves high-Péclet convection-diffusion systems with only a few GMRES iterations, even at vanishing diffusion.

desk verdict Solid MS-GFEM extension to conservative convection-diffusion with strong numerics; the hyperbolic-limit claim rests on one untested regularization constant, but the finite-Péclet core is sound. read the letter →

arxiv 2509.17531 v2 pith:EDMK4JSD submitted 2025-09-22 math.NA cs.NA

Lukas Holbach , Peter Bastian , Robert Scheichl This is my paper · ORCID
classification math.NAcs.NA MSC 65F1065N2265N55
keywords domaindecompositionrestrictedadditiveSchwarzspectralcoarsespaceMS-GFEMconvection-diffusionhighPécletnumberheterogeneousdiffusionGMRES
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 constructs a preconditioner for steady-state convection-diffusion equations whose solutions develop sharp layers when convection dominates diffusion. The coarse space comes from the multiscale spectral generalized finite element method: locally optimal basis functions obtained from generalized eigenproblems on spaces of functions whose convection-diffusion residual vanishes against interior test functions. The authors prove exponential convergence of this multiscale approximation for conservative convection-diffusion, show that the associated restricted additive Schwarz iteration inherits the contraction for divergence-free velocity fields, and then use the iteration as a GMRES preconditioner. Their experiments indicate that iteration counts stay small and coarse spaces stay low-dimensional across grid Péclet numbers, three discretization families, and up to $10^5$ subdomains; the method is also effective, without a supporting theory, for indefinite problems and in the vanishing-diffusion limit.

What carries the argument

The central object is the MS-GFEM coarse space. On each oversampling domain $\omega_j^*$ one solves the local generalized eigenproblem $a_{\omega_j^*}(\chi_j\varphi,\chi_j\psi)=\lambda a_{\omega_j^*}(\varphi,\psi)$, where $a_{\omega_j^*}$ is the diffusion part of the local bilinear form and $\chi_j$ is the partition-of-unity cutoff, and the eigenfunctions are sought among functions whose convection-diffusion residual vanishes against test functions supported in the interior of $\omega_j^*$. Keeping the eigenfunctions with the largest eigenvalues yields locally optimal low-energy modes; because those eigenvalues decay nearly exponentially (Theorem 3.5), a few per subdomain capture the part of the error that a one-level RAS iteration cannot move. The preconditioner applies the one-level restricted additive Schwarz sweep (local solves combined through a partition of unity), then a multiplicative coarse-space correction, and feeds the combined operator into GMRES.

What would settle it

Run the Section 5.4 pure-transport experiment on a fixed grid with the artificial diffusion in the eigenproblem inner product set to $10^{-4}$, $10^{-6}$, and $10^{-8}$, and with different oversampling sizes. If GMRES iteration counts or coarse-space dimensions change substantially across those values, the claimed robustness in the hyperbolic limit depends on an unstated tuning choice rather than following from the exponential eigenvalue decay.

Watch

Extended reading notes

Core claim

The central claim is that, for heterogeneous steady-state convection-diffusion in conservation form, the MS-GFEM coarse space spans the modes that slow down domain-decomposition iterations, so a two-level hybrid restricted additive Schwarz (RAS) preconditioner built from it makes GMRES converge in very few iterations with small coarse spaces. The theory establishes exponential convergence of the local approximation spaces in the number of eigenfunctions per subdomain (Theorem 3.5), and for divergence-free velocity fields it proves that the resulting MS-GFEM iteration contracts monotonically in the energy norm (Theorem 3.10). Numerically, the preconditioner remains robust as the grid Péclet number ranges from $10$ to beyond $10^5$, as the number of subdomains grows past $10^5$, and in the formally infinite Péclet limit, where the method is used without a convergence proof.

Load-bearing premise

The load-bearing premise is that every global and local subproblem has a unique, stable solution and that the spectral coarse spaces capture all modes a one-level Schwarz iteration cannot move; for the vanishing-diffusion experiments, the premise also rests on an artificial diffusion of $10^{-6}$ used only inside the eigenproblem inner product, whose influence is not analyzed.

Editorial extensions

If this is right

  • GMRES with this preconditioner needs only a small, adaptively chosen coarse space whose dimension is almost independent of grid Péclet number and grows mildly with the number of subdomains.
  • The coarse-space construction transfers across discretizations (discontinuous Galerkin, cell-centred finite volume, conforming $Q_1$ elements) and across structured and unstructured meshes, so the preconditioner can be reused rather than re-derived for each solver.
  • Because the coarse space is assembled from local eigenproblems, the method is parallel by construction and stays effective at more than $10^5$ subdomains and roughly $10^8$ degrees of freedom in the tested configuration.
  • For divergence-free velocity fields the MS-GFEM iteration itself contracts monotonically in the energy norm, so the same coarse space can drive a simple fixed-point iteration and not only GMRES.
  • Increasing the oversampling size or the number of eigenfunctions per subdomain yields arbitrarily fast GMRES convergence, at the price of costlier local eigenproblems and coarse solves.

Reading between the lines

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

  • A direct way to test whether the hyperbolic-limit effectiveness is intrinsic is to vary the artificial diffusion value ($10^{-6}$ in the paper) used only inside the eigenproblem inner product in the pure-transport experiments; flat iteration counts would support the claim, while visible dependence would expose a tuning parameter the paper does not analyze.
  • The exponential eigenvalue decay suggests the coarse space could be recycled across a sequence of nearby linear systems, for example implicit time steps of unsteady transport or parameter sweeps, which would amortize the setup cost that currently dominates runtime.
  • The paper's extension to non-divergence-free and indefinite problems has no convergence proof; a natural next step is a non-coercive analogue of Theorem 3.10, presumably under a resolution condition on the discarded eigenvalues.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes a two-level hybrid restricted additive Schwarz (RAS) preconditioner for steady-state convection-diffusion equations in conservative form. The coarse space is built from local generalized eigenproblems on operator-harmonic spaces following the MS-GFEM framework of Ma (2025). The authors extend the continuous theory of that framework to convection-diffusion, proving a Caccioppoli-type inequality (Lemma 3.2), a weak approximation property (Lemma 3.3), a local exponential-convergence estimate (Theorem 3.5), and a global quasi-optimal error bound (Theorem 3.8) under coercivity or unquantified resolution conditions. For divergence-free velocity fields they prove monotone convergence of the corresponding iterative method (Theorem 3.10). The preconditioner is implemented in the DUNE framework and tested on structured and unstructured meshes, with up to 10^5 subdomains and around 10^8 degrees of freedom, using DG, finite-volume, and conforming finite-element discretizations. As an extension without theory, the paper also reports numerical behavior near the hyperbolic limit A = 0 and for indefinite problems.

Significance. If the main claims hold, the paper contributes a practically important and parameter-robust spectral preconditioner for high-Péclet convection-diffusion, a class of problems for which classical two-level methods often degrade. The numerical evidence is unusually comprehensive: it includes a 3D case on an unstructured mesh, comparisons with GAMG, BoomerAMG, and MUMPS, and a scalable test with more than one hundred thousand subdomains. The theory is also creditable: the authors verify the hypotheses of an established external framework rather than assuming them, and the exponential-decay result is obtained without fitting parameters to data. The core finite-Péclet contribution appears sound. The main weakness is the vanishing-diffusion section, where an artificial diffusion constant in the local eigenproblem is introduced and tested only at a single value, leaving the robustness claim in that limit unsupported.

major comments (3)
  1. [Section 5.4] The vanishing-diffusion experiment introduces an artificial constant diffusion coefficient of 10^-6 in the restriction operator inner product used in the generalized eigenproblem (3.8), while the PDE itself has A = 0. This parameter fully determines the ordering of eigenfunctions that enter the coarse space, and it is not present in the equation being solved. The paper tests only this single value and gives no analysis or parameter sweep; the reported saturation near Péclet = +∞ (Figure 5) could therefore be an artifact of this particular regularization. Since the abstract explicitly presents the hyperbolic limit as a contribution, the authors should either demonstrate robustness of the coarse-space dimension and GMRES iteration counts over several values of this artificial diffusion, provide a heuristic or theoretical justification for the chosen value, or clearly restrict the claim to the tested setting.
  2. [Theorem 3.8 and Remark 2.1] The global error estimate (3.17) for indefinite problems is stated under the condition that λmax and H*_max are sufficiently small, but these resolution conditions are never quantified, and the proof defers to [36, Corollary 3.27] while describing the resulting constants as overly pessimistic. This leaves the theory for the indefinite case in an unverified state. Furthermore, Remark 2.1 assumes that all subdomain problems are well-posed, an assumption that is not guaranteed for the hyperbolic (A = 0) or indefinite regimes tested in Sections 5.2 and 5.4. The manuscript should either quantify the resolution conditions, state explicitly that the indefinite and hyperbolic extensions are purely empirical, or both, so that readers can distinguish the proven finite-Péclet part from the numerically explored extensions.
  3. [Theorem 3.10 proof] In the proof of Theorem 3.10, the authors claim that the definition of G together with (3.20) immediately yields the local contraction estimate ||(I-G)v||_{a,ω*_j} ≤ ϑ||v||_{a,ω*_j} for all v in H^1(Ω). This does not follow from the global estimate (3.20), because the coarse-space projection π_S introduces global coupling: the restriction of (I-G)v to one oversampling domain depends on v on all subdomains. The global bound ||(I-G)v||_{a,Ω} ≤ ϑ||v||_{a,Ω} would be sufficient to establish the monotone convergence (3.21), so the proof should be corrected by replacing the unjustified local assertion with the global one, or by adding the missing argument that justifies the local bound.
minor comments (5)
  1. [Abstract and Section 5] The phrase 'arbitrarily fast convergence of preconditioned GMRES' in the abstract and at the end of Section 5.1 is stronger than what the numerical experiments establish; the experiments show that convergence rates improve with oversampling and coarse-space dimension, but they do not prove that any prescribed rate can be attained. A more cautious wording such as 'substantially accelerated convergence' would be more accurate.
  2. [Figure 5 and Section 5.4] In Figure 5, the row labelled '+∞' corresponds to A = 0 in the PDE while the eigenproblem uses artificial diffusion 10^-6; this distinction should be made in the caption or in the text to avoid confusion.
  3. [Figure 6] The eigenfunction plots in Figure 6 would be easier to interpret if the corresponding eigenvalues or their ranks were indicated, since the discussion refers to low- versus high-eigenvalue modes.
  4. [Throughout] There are several typographical inconsistencies, including the rendering of 'Péclet' with an unexpected space (e.g., in the abstract and Section 2) and the informal notation 'Péclet = 100' in Table 1; these should be cleaned up.
  5. [Section 5.3] The comparison with GAMG and BoomerAMG would be fairer if the same preconditioner setup (e.g., number of threads or MPI processes) were used for all methods; as reported, the AMG runs use 64 MPI processes while the two-level RAS runs use 64 threads, which can affect runtime comparisons.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the convergence theory is derived from an external framework with verified assumptions, and numerics are benchmarked independently.

full rationale

The central theoretical claim, exponential convergence of the MS-GFEM approximation (Theorem 3.5), is not assumed: the paper proves a Caccioppoli-type inequality (Lemma 3.2) and a weak approximation property (Lemma 3.3) for the conservative convection-diffusion bilinear form, verifies the abstract assumptions of the external framework [36], and then applies [36, Theorem 3.8]. The local error estimate (Theorem 3.4) and global estimate (Theorem 3.8) are consequences of these verified assumptions plus the spectral construction; the eigenvalue decay is proved, not fitted or defined into existence. The iterative/RAS and GMRES claims follow from the error estimates and the definition of the operator G (Theorem 3.10, Section 4); no parameter is fitted to data and no predicted quantity is a renamed fit. The numerical sections benchmark against external solvers (GAMG, BoomerAMG, MUMPS) and against a partition-of-unity coarse space, providing independent support. The vanishing-diffusion experiments in Section 5.4 use an artificial constant diffusion of 10^-6 in the eigenproblem only and are explicitly labelled by the authors as "an extension, for which we do not have theory yet"; this is a parameter-robustness concern, not circularity. Although the paper is part of the MS-GFEM research program and cites related work by Scheichl and collaborators ([37], [38], [50]), the load-bearing framework [36] is a separate published paper by a different author and is used as an external mathematical result with all of its assumptions verified rather than as a self-justifying citation.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central mathematical result is an extension of an external framework (Ma 2025), with the paper's main new work being the verification of Caccioppoli and weak-approximation properties for conservative convection-diffusion. The practical claims rely on several user-chosen algorithmic parameters (oversampling, thresholds, penalty) and one ad hoc artificial-diffusion choice for the hyperbolic limit; none of these are fitted to match a predetermined prediction, but they are free choices that affect performance and should be reported and tested for sensitivity.

free parameters (6)
  • Oversampling layers (ell) = 2 (default)
    Number of element layers added beyond the overlap to build the oversampling domains; fixed to 2 in most experiments, varied in Figure 2. Directly controls eigenvalue decay rate and setup cost (Sections 4.3, 5.1).
  • Coarse-space threshold (lambda_max) = 2 (Section 5.1), 1 and 0.5 (Section 5.2)
    Adaptive criterion selecting all eigenfunctions with eigenvalue above the threshold; chosen by hand per experiment to balance coarse-space dimension and iteration count.
  • Eigenfunctions per subdomain (n_sd) = 2 in 3D test; 6 to 64 in convergence test
    Fixed-size coarse space variant used in Figure 2 and Section 5.3; user-selected.
  • DG penalty parameter (alpha) = 3
    Parameter in the symmetric weighted interior penalty discontinuous Galerkin discretization (Section 4.2), chosen following benchmark [9]; polynomial degree p = 1 throughout.
  • Overlap layers = 2
    Number of element layers used to create the overlapping subdomains omega_j; fixed throughout the experiments.
  • Artificial diffusion in restriction inner product = 10^-6
    Used only to define the a-inner product for the operator-harmonic space in the vanishing-diffusion limit (Section 5.4); ad hoc choice with no sensitivity study.
assumptions (5)
  • domain assumption All global and subdomain problems are well-posed.
    Remark 2.1 states this is assumed throughout. It is not guaranteed in the indefinite and vanishing-diffusion regimes tested numerically; if it fails, the projections and local solves in the preconditioner are not defined.
  • domain assumption A is symmetric and uniformly elliptic with 0 < amin <= amax < infinity, b in W^{1,infinity}, and the outflow boundary satisfies Gamma_O subset of {b dot nu >= 0}.
    Equation (2.2) and Section 2; standard assumptions for the weak formulation and the Caccioppoli inequality.
  • ad hoc to paper The mixed generalized eigenproblem (3.8) is equivalent to (3.6) for non-zero eigenvalues and is non-defective.
    Section 3 states this without complete proof; the discretized version is assumed to inherit the property. The Lagrange-multiplier formulation is used in all experiments.
  • standard math The unified MS-GFEM framework of Ma (2025) applies, with Assumptions 2.13, 3.1, and 3.4 satisfied via the new Lemmas 3.2 and 3.3.
    Theorems 3.4, 3.5, and 3.8 import results from Ma (2025) after the authors verify the hypotheses. This external theory is the main analytical pillar.
  • domain assumption The continuous convergence results carry over to the discrete iterations used in the experiments.
    Remark 4.1 states the discrete analysis is not performed for the DG, CCFV, and CG discretizations and suggests the continuous result should carry over based on prior elliptic and Helmholtz work. This is an unproved transfer of theory to the actual numerical method.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Robust spectral preconditioning for high-P\'{e}clet number convection-diffusion." pith.science (2026). https://pith.science/paper/EDMK4JSD

@misc{pith2026250917531,
  author       = {Pith},
  title        = {Pith review of: Robust spectral preconditioning for high-P\'eclet number convection-diffusion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EDMK4JSD}},
  note         = {Machine review of arXiv:2509.17531}
}
abstract

We introduce a two-level hybrid restricted additive Schwarz (RAS) preconditioner for heterogeneous steady-state convection-diffusion equations at high P\'{e}clet numbers. Our construction builds on the multiscale spectral generalized finite element method (MS-GFEM), wherein the coarse space is spanned by locally optimal basis functions obtained from local generalized eigenproblems on operator-harmonic spaces. Extending the theory of Ma (2025) to convection-diffusion problems in conservation form, we establish exponential convergence of the MS-GFEM approximation with respect to the dimension of the local approximation space. Rewriting MS-GFEM as a RAS-type iteration, we show for coercive problems that this exponential convergence property is inherited by the RAS-type iterative method (at least in the continuous setting). Employed as a preconditioner within the generalized minimal residual method (GMRES), the resulting method requires only a few iterations for high accuracy even with low-dimensional coarse spaces. Through extensive numerical experiments on problems with high-contrast diffusion and non-divergence-free, rotating velocity fields, we demonstrate robustness with respect to the grid P\'{e}clet number and the number of subdomains (tested up to $10^5$ subdomains), while coarse-space dimensions remain small as grid P\'{e}clet numbers increase. By adapting the coarse space and oversampling size, we are able to achieve arbitrarily fast convergence of preconditioned GMRES. As an extension, for which we do not have theory yet, we show effectiveness of the method even for indefinite problems and in the vanishing-diffusion limit.

Figures

Figures reproduced from arXiv: 2509.17531 by the authors.

Figure 1
Figure 1. Model setup for Section 5.1. Left: Diffusion coefficient and velocity field. The [PITH_FULL_IMAGE:figures/full_fig_p015_1.png] view at source ↗
Figure 2
Figure 2. Convergence test for the model of Section 5.1 with [PITH_FULL_IMAGE:figures/full_fig_p016_2.png] view at source ↗
Figure 3
Figure 3. Robustness with respect to grid P´eclet number P´e [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Model setup for the 3D example from Section 5.3. (a) Computational domain. [PITH_FULL_IMAGE:figures/full_fig_p019_4.png]
Figure 5
Figure 5. Figure 5: Robustness in the pure-transport limit for the model of Section 5.4, using [PITH_FULL_IMAGE:figures/full_fig_p020_5.png]
Figure 6
Figure 6. Figure 6: Selected eigenfunctions φj,k in an interior subdomain ωj of the homogeneous example from Section 5.4. The model parameters are A ≡ 10−3 , b ≡ 0 (P´eh = 0) in the first row, and b from (5.1) with A ≡ 10−3 (P´eh ≈ 2) and A ≡ 10−6 (P´eh ≈ 2 000) in the second and third ro…

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

59 extracted references · 36 canonical work pages

  1. [1]

    A domain decomposition precondi- tioner for an advection–diffusion problem

    Y. Achdou, P. Le Tallec, F. Nataf, and M. Vidrascu. “A domain decomposition precondi- tioner for an advection–diffusion problem”. In: Computer Methods in Applied Mechanics and Engineering 184.2 (2000), pp. 145–170. doi: 10.1016/S0045-7825(99)00227-3

    work page doi:10.1016/s0045-7825(99)00227-3 2000

  2. [2]

    Efficient Algebraic Two-Level Schwarz Preconditioner for Sparse Matrices

    H. Al Daas, P. Jolivet, and T. Rees. “Efficient Algebraic Two-Level Schwarz Preconditioner for Sparse Matrices”. In: SIAM Journal on Scientific Computing 45.3 (2023), A1199–A1213. doi: 10.1137/22M1469833

    doi:10.1137/22m1469833 2023

  3. [3]

    Alber, P

    C. Alber, P. Bastian, M. Hauck, and R. Scheichl. Optimal Spectral Approximation in the Overlaps for Generalized Finite Element Methods . 2025. arXiv: 2507.12226 [math.NA]

    arXiv 2025

  4. [4]

    Constrained Local Approximate Ideal Restriction for Advection-Diffusion Problems

    A. Ali, J. J. Brannick, K. Kahl, O. A. Krzysik, J. B. Schroder, and B. S. Southworth. “Constrained Local Approximate Ideal Restriction for Advection-Diffusion Problems”. In: SIAM Journal on Scientific Computing 46.5 (2024), S96–S122. doi: 10.1137/23M1583442

    work page doi:10.1137/23m1583442 2024

  5. [5]

    A Fully Asynchronous Multifrontal Solver Using Distributed Dynamic Scheduling

    P. Amestoy, I. S. Duff, J. Koster, and J.-Y. L’Excellent. “A Fully Asynchronous Multifrontal Solver Using Distributed Dynamic Scheduling”. In: SIAM Journal on Matrix Analysis and Applications 23.1 (2001), pp. 15–41. doi: 10.1137/S0895479899358194

    doi:10.1137/s0895479899358194 2001

  6. [6]

    Optimal Local Approximation Spaces for Generalized Finite Element Methods with Application to Multiscale Problems

    I. Babuska and R. Lipton. “Optimal Local Approximation Spaces for Generalized Finite Element Methods with Application to Multiscale Problems”. In: Multiscale Modeling & Simulation 9.1 (2011), pp. 373–406. doi: 10.1137/100791051

    doi:10.1137/100791051 2011

  7. [7]

    Balay et al

    S. Balay et al. PETSc/TAO Users Manual. Tech. rep. ANL-21/39 - Revision 3.23. Argonne National Laboratory, 2025. doi: 10.2172/2476320

    doi:10.2172/2476320 2025

  8. [8]

    A generic grid interface for parallel and adaptive scientific computing. Part II: implementation and tests in DUNE

    P. Bastian, M. Blatt, A. Dedner, C. Engwer, R. Kl¨ ofkorn, R. Kornhuber, M. Ohlberger, and O. Sander. “A generic grid interface for parallel and adaptive scientific computing. Part II: implementation and tests in DUNE”. In: Computing 82.2–3 (2008), 121–138. doi: 10.1007/s00607-008-0004-9 . 22

    doi:10.1007/s00607-008-0004-9 2008

Show all 59 references
  1. [9]

    Benchmark 3d: Symmetric weighted interior penalty discontinuous galerkin scheme

    P. Bastian. “Benchmark 3d: Symmetric weighted interior penalty discontinuous galerkin scheme”. In: Finite Volumes for Complex Applications VI Problems & Perspectives: FVCA 6, International Symposium, Prague, June 6-10, 2011 . Springer. 2011, pp. 949–959. doi: 10.1007/978-3-642...

    2011 doi

  2. [10]

    The Dune framework: Basic con- cepts and recent developments

    P. Bastian, M. Blatt, A. Dedner, N.-A. Dreier, C. Engwer, R. Fritze, C. Gr¨ aser, C. Gr¨ uninger, D. Kempf, R. Kl¨ ofkorn, M. Ohlberger, and O. Sander. “The Dune framework: Basic con- cepts and recent developments”. In: Computers & Mathematics with Applications 81 (2021). Deve...

    2021 doi

  3. [11]

    Downwind numbering: robust multigrid for convection-diffusion problems

    J. Bey and G. Wittum. “Downwind numbering: robust multigrid for convection-diffusion problems”. In: Applied Numerical Mathematics 23.1 (1997). Multilevel Methods, pp. 177–

    1997

  4. [12]

    Super-localized orthogonal decomposition for convection-dominated diffusion problems

    F. Bonizzoni, P. Freese, and D. Peterseim. “Super-localized orthogonal decomposition for convection-dominated diffusion problems”. In:BIT Numerical Mathematics 64.3 (2024). doi: 10.1007/s10543-024-01035-8

    2024 doi

  5. [13]

    Overlapping Schwarz methods with GenEO coarse spaces for indefinite and nonself-adjoint problems

    N. Bootland, V. Dolean, I. G. Graham, C. Ma, and R. Scheichl. “Overlapping Schwarz methods with GenEO coarse spaces for indefinite and nonself-adjoint problems”. In: IMA Journal of Numerical Analysis 43.4 (2022), pp. 1899–1936. doi: 10.1093/imanum/drac036

    2022 doi

  6. [14]

    Domain Decomposition Algorithms for Indefinite Elliptic Problems

    X.-C. Cai and O. B. Widlund. “Domain Decomposition Algorithms for Indefinite Elliptic Problems”. In: SIAM Journal on Scientific and Statistical Computing 13.1 (1992), pp. 243–

    1992

  7. [15]

    Multiplicative Schwarz Algorithms for Some Nonsymmetric and Indefinite Problems

    X.-C. Cai and O. B. Widlund. “Multiplicative Schwarz Algorithms for Some Nonsymmetric and Indefinite Problems”. In: SIAM Journal on Numerical Analysis 30.4 (1993), pp. 936–

    1993

  8. [16]

    Multiscale stabilization for convection-dominated diffusion in heterogeneous media

    V. M. Calo, E. T. Chung, Y. Efendiev, and W. T. Leung. “Multiscale stabilization for convection-dominated diffusion in heterogeneous media”. In: Computer Methods in Applied Mechanics and Engineering 304 (2016), pp. 359–377. doi: 10.1016/j.cma.2016.02.014

    2016 doi

  9. [17]

    Steady-State Simulation of Semiconductor Devices Using Discon- tinuous Galerkin Methods

    L. Chen and H. Bagci. “Steady-State Simulation of Semiconductor Devices Using Discon- tinuous Galerkin Methods”. In: IEEE Access 8 (2020), 16203–16215. doi: 10.1109/access .2020.2967125

    2020

  10. [18]

    Sensitivity of temporal moments calculated by the adjoint-state method and joint inversing of head and tracer data

    O. A. Cirpka and P. K. Kitanidis. “Sensitivity of temporal moments calculated by the adjoint-state method and joint inversing of head and tracer data”. In: Advances in Water Resources 24.1 (2000), pp. 89–103. doi: 10.1016/S0309-1708(00)00007-5

    2000 doi

  11. [19]

    Algorithm 832: UMFPACK V4.3—an unsymmetric-pattern multifrontal method

    T. A. Davis. “Algorithm 832: UMFPACK V4.3—an unsymmetric-pattern multifrontal method”. In: ACM Trans. Math. Softw. 30.2 (2004), 196–199. doi: 10.1145/992200.992206

    2004

  12. [20]

    D. A. Di Pietro and A. Ern. Mathematical aspects of discontinuous Galerkin methods . en. 2012th ed. Math´ ematiques et Applications. Berlin, Germany: Springer, 2011. isbn: 978-3- 642-22980-0

    2011

  13. [21]

    Dolean, P

    V. Dolean, P. Jolivet, and F. Nataf. An Introduction to Domain Decomposition Methods: Algorithms, Theory, and Parallel Implementation . Society for Industrial and Applied Math- ematics, 2015. isbn: 9781611974065

    2015

  14. [22]

    Estimation of penalty parameters for symmetric interior penalty Galerkin methods

    Y. Epshteyn and B. Rivi` ere. “Estimation of penalty parameters for symmetric interior penalty Galerkin methods”. In: Journal of Computational and Applied Mathematics 206.2 (2007), pp. 843–872. doi: 10.1016/j.cam.2006.08.029

    2007 doi

  15. [23]

    Ern and J.-L

    A. Ern and J.-L. Guermond. Theory and Practice of Finite Elements . Springer New York,

  16. [24]

    A discontinuous Galerkin method with weighted averages for advection–diffusion equations with locally small and anisotropic diffusivity

    A. Ern, A. F. Stephansen, and P. Zunino. “A discontinuous Galerkin method with weighted averages for advection–diffusion equations with locally small and anisotropic diffusivity”. In: IMA Journal of Numerical Analysis 29.2 (2008), pp. 235–256. doi: 10.1093/imanum/drm0 50

    2008 doi

  17. [25]

    MONKES: a fast neoclassical code for the evaluation of monoenergetic transport coefficients in stellarator plasmas

    F. Escoto, J. Velasco, I. Calvo, M. Landreman, and F. Parra. “MONKES: a fast neoclassical code for the evaluation of monoenergetic transport coefficients in stellarator plasmas”. In: Nuclear Fusion 64.7 (2024), p. 076030. doi: 10.1088/1741-4326/ad3fc9. 23

    2024 doi

  18. [26]

    Wavelet-Based Edge Multiscale Finite Element Methods for Singularly Perturbed Convection-Diffusion Equations

    S. Fu, E. Chung, and G. Li. “Wavelet-Based Edge Multiscale Finite Element Methods for Singularly Perturbed Convection-Diffusion Equations”. In: Multiscale Modeling & Simula- tion 23.1 (2025), pp. 431–457. doi: 10.1137/24M1659017

    2025 doi

  19. [27]

    Domain Decomposition Preconditioners for Multiscale Flows in High-Contrast Media

    J. Galvis and Y. Efendiev. “Domain Decomposition Preconditioners for Multiscale Flows in High-Contrast Media”. In: Multiscale Modeling & Simulation 8.4 (2010), pp. 1461–1483. doi: 10.1137/090751190

    2010 doi

  20. [28]

    Domain Decomposition Preconditioners for Multiscale Flows in High Contrast Media: Reduced Dimension Coarse Spaces

    J. Galvis and Y. Efendiev. “Domain Decomposition Preconditioners for Multiscale Flows in High Contrast Media: Reduced Dimension Coarse Spaces”. In: Multiscale Modeling & Simulation 8.5 (2010), pp. 1621–1644. doi: 10.1137/100790112

    2010 doi

  21. [29]

    Downwind Gauß-Seidel Smoothing for Convection Domi- nated Problems

    W. Hackbusch and T. Probst. “Downwind Gauß-Seidel Smoothing for Convection Domi- nated Problems”. In: Numerical Linear Algebra with Applications 4.2 (1997), pp. 85–102. doi: 10.1002/(SICI)1099-1506(199703/04)4:2<85::AID-NLA100>3.0.CO;2-2

    1997 doi

  22. [30]

    An optimal order interior penalty discontinuous Galerkin discretization of the compressible Navier–Stokes equations

    R. Hartmann and P. Houston. “An optimal order interior penalty discontinuous Galerkin discretization of the compressible Navier–Stokes equations”. In: Journal of Computational Physics 227.22 (2008), pp. 9670–9685. doi: 10.1016/j.jcp.2008.07.015

    2008 doi

  23. [31]

    Reduced dimension GDSW coarse spaces for monolithic Schwarz domain decomposition methods for incompressible fluid flow problems

    A. Heinlein, C. Hochmuth, and A. Klawonn. “Reduced dimension GDSW coarse spaces for monolithic Schwarz domain decomposition methods for incompressible fluid flow problems”. In: International Journal for Numerical Methods in Engineering 121.6 (2020), pp. 1101–

    2020

  24. [32]

    https://llnl.gov/casc/hypre

    hypre: High Performance Preconditioners. https://llnl.gov/casc/hypre

  25. [33]

    METIS and ParMETIS

    G. Karypis. “METIS and ParMETIS”. In: Encyclopedia of parallel computing . Springer, 2011, pp. 1117–1124

    2011

  26. [34]

    Finite Volume Methods for Convection- Diffusion Problems

    R. D. Lazarov, I. D. Mishev, and P. S. Vassilevski. “Finite Volume Methods for Convection- Diffusion Problems”. In: SIAM Journal on Numerical Analysis 33.1 (1996), pp. 31–55. doi: 10.1137/0733003

    1996 doi

  27. [35]

    Error analysis of a variational multiscale stabi- lization for convection-dominated diffusion equations in two dimensions

    G. Li, D. Peterseim, and M. Schedensack. “Error analysis of a variational multiscale stabi- lization for convection-dominated diffusion equations in two dimensions”. In: IMA Journal of Numerical Analysis 38.3 (2017), 1229–1253. doi: 10.1093/imanum/drx027

    2017 doi

  28. [36]

    A Unified Framework for Multiscale Spectral Generalized FEMs and Low-Rank Ap- proximations to Multiscale PDEs

    C. Ma. “A Unified Framework for Multiscale Spectral Generalized FEMs and Low-Rank Ap- proximations to Multiscale PDEs”. In: Foundations of Computational Mathematics (2025). doi: 10.1007/s10208-025-09711-z

    2025 doi

  29. [37]

    C. Ma, C. Alber, R. Scheichl, and Y. Zhang. Two-level Restricted Additive Schwarz precondi- tioner based on Multiscale Spectral Generalized FEM for Heterogeneous Helmholtz Problems

  30. [38]

    Novel Design and Analysis of Generalized Finite Element Methods Based on Locally Optimal Spectral Approximations

    C. Ma, R. Scheichl, and T. Dodwell. “Novel Design and Analysis of Generalized Finite Element Methods Based on Locally Optimal Spectral Approximations”. In: SIAM Journal on Numerical Analysis 60.1 (2022), pp. 244–273. doi: 10.1137/21M1406179

    2022 doi

  31. [39]

    Adaptive selection of face coarse degrees of freedom in the BDDC and the FETI-DP iterative substructuring methods

    J. Mandel and B. Soused´ ık. “Adaptive selection of face coarse degrees of freedom in the BDDC and the FETI-DP iterative substructuring methods”. In: Computer Methods in Ap- plied Mechanics and Engineering 196.8 (2007). Domain Decomposition Methods: recent ad- vances and new c...

    2007 doi

  32. [40]

    Nonsymmetric Reduction- Based Algebraic Multigrid

    T. A. Manteuffel, S. M¨ unzenmaier, J. Ruge, and B. Southworth. “Nonsymmetric Reduction- Based Algebraic Multigrid”. In:SIAM Journal on Scientific Computing 41.5 (2019), S242–S268. doi: 10.1137/18m1193761

    2019 doi

  33. [41]

    Nonsymmetric Algebraic Multigrid Based on Local Approximate Ideal Restriction (ℓAIR)

    T. A. Manteuffel, J. Ruge, and B. S. Southworth. “Nonsymmetric Algebraic Multigrid Based on Local Approximate Ideal Restriction (ℓAIR)”. In: SIAM Journal on Scientific Computing 40.6 (2018), A4105–A4130. doi: 10.1137/17m1144350

    2018 doi

  34. [42]

    Nataf and E

    F. Nataf and E. Parolin. Coarse spaces for non-symmetric two-level preconditioners based on local extended generalized eigenproblems. 2025. arXiv: 2404.02758 [math.NA]

    2025 arXiv

  35. [43]

    Numerical solution of steady-state groundwater flow and solute transport problems: Discontinuous Galerkin based methods compared to the Stream- line Diffusion approach

    A. Ngo, P. Bastian, and O. Ippisch. “Numerical solution of steady-state groundwater flow and solute transport problems: Discontinuous Galerkin based methods compared to the Stream- line Diffusion approach”. In: Computer Methods in Applied Mechanics and Engineering 294 (2015), ...

    2015 doi

  36. [44]

    Semiconductor de- vice modelling from the numerical point of view

    S. J. Polak, C. Den Heijer, W. H. A. Schilders, and P. Markowich. “Semiconductor de- vice modelling from the numerical point of view”. In: International Journal for Numerical Methods in Engineering 24.4 (1987), pp. 763–838. doi: 10.1002/nme.1620240408

    1987 doi

  37. [45]

    Y. Qiu. Spectra: C++ Library For Large Scale Eigenvalue Problems . https://spectralib .org. 2015

    2015

  38. [46]

    H.-G. Roos, M. Stynes, and L. Tobiska. Robust numerical methods for singularly perturbed differential equations. en. Springer Series in Computational Mathematics. Berlin, Germany: Springer, 2010. isbn: 978-3-540-34467-4

    2010

  39. [47]

    Algebraic Multigrid

    J. W. Ruge and K. St¨ uben. “Algebraic Multigrid”. In: Multigrid Methods. Society for Indus- trial and Applied Mathematics, 1987, 73–130. isbn: 9781611971057

    1987

  40. [48]

    Partition of unity coarse spaces: enhanced versions, discontinuous coefficients and applications to elasticity

    M. Sarkis. “Partition of unity coarse spaces: enhanced versions, discontinuous coefficients and applications to elasticity”. In: Proceedings of the 14th International Conference on Domain Decomposition Methods (2003), pp. 149–158

    2003

  41. [49]

    Abstract ro- bust coarse spaces for systems of PDEs via generalized eigenproblems in the overlaps

    N. Spillane, V. Dolean, P. Hauret, F. Nataf, C. Pechstein, and R. Scheichl. “Abstract ro- bust coarse spaces for systems of PDEs via generalized eigenproblems in the overlaps”. In: Numerische Mathematik 126.4 (2013), 741–770. doi: 10.1007/s00211-013-0576-y

    2013 doi

  42. [50]

    Strehlow, C

    A. Strehlow, C. Ma, and R. Scheichl. Fast-convergent two-level restricted additive Schwarz methods based on optimal local approximation spaces . 2024. arXiv: 2408.16282 [math.NA]

    2024 arXiv

  43. [51]

    Steady-state convection-diffusion problems

    M. Stynes. “Steady-state convection-diffusion problems”. In: Acta Numerica 14 (2005), 445–508. doi: 10.1017/S0962492904000261

    2005 doi

  44. [52]

    FETI domain decomposition methods for scalar advection–diffusion problems

    A. Toselli. “FETI domain decomposition methods for scalar advection–diffusion problems”. In: Computer Methods in Applied Mechanics and Engineering 190.43 (2001), pp. 5759–5776. doi: 10.1016/S0045-7825(01)00195-5

    2001 doi

  45. [53]

    Toselli and O

    A. Toselli and O. B. Widlund. Domain Decomposition Methods — Algorithms and Theory . Springer Berlin Heidelberg, 2005. isbn: 9783540266624

    2005

  46. [54]

    A balancing domain decomposition method by constraints for advection- diffusion problems

    X. Tu and J. Li. “A balancing domain decomposition method by constraints for advection- diffusion problems”. In: Communications in Applied Mathematics and Computational Sci- ence 3.1 (2008), 25–60. doi: 10.2140/camcos.2008.3.25. 25

    2008 doi

  47. [192]

    doi: 10.1016/S0168-9274(96)00067-0

    doi

  48. [258]

    doi: 10.1137/0913013

    doi

  49. [952]

    doi: 10.1137/0730049

    doi

  50. [1119]

    doi: 10.1002/nme.6258

    doi

  51. [2025]

    arXiv: 2409.06533 [math.NA]

    arXiv

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.