Pith. sign in

REVIEW 4 major objections 4 minor 36 references

Automated Runge-Kutta-Nystr\"om time stepping for finite element methods in Irksome

T0 review · 4 major / 4 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 This paper shows that discretizing second-order time derivatives directly with Runge–Kutta–Nyström methods, rather than rewriting the equation as a first-order system, halves the algebraic unknowns and yields faster finite-element solves…

desk verdict Solid Irksome extension automating RKN time stepping with real runtime savings, but the abstract oversells RKN as 'essentially required' and the mixed-order Biot extension rests on an unproved heuristic about zero initial derivatives. read the letter →

arxiv 2508.20255 v1 pith:AZLF4E7J submitted 2025-08-27 math.NA cs.NA

Robert C. Kirby , Scott P. MacLachlan , Pablo D. Brubeck This is my paper · ORCID
classification math.NAcs.NA MSC 65L0665M2065M60
keywords Runge-Kutta-Nyströmmethodssecond-ordertimederivativesfiniteelementIrksomeUnifiedFormLanguagewaveequationporoelasticityimplicitRunge-Kuttapreconditioning
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 extends the Irksome library so that users can write a finite-element semidiscrete form containing second-order time derivatives and have a Runge–Kutta–Nyström (RKN) method applied automatically. The central claim is that discretizing the second-order form directly, instead of rewriting it as a first-order system and applying an ordinary Runge–Kutta method, produces smaller stage-coupled algebraic systems and better run time. With effective preconditioning, the paper reports, fully implicit RKN methods can match or beat explicit methods for wave equations and are essentially required for wave-type equations with fourth-order spatial derivatives. The paper also reports that the approach handles fully dynamic poroelasticity, a mixed-order system, effectively even near the incompressible limit.

What carries the argument

The key mechanism is the extended Runge-Kutta-Nyström tableau, where the second-derivative stage equation uses $\bar A = A^2$ and the update uses $\bar b = A^T b$; this reduces the stage system to a single set of unknowns. Around that identity, the paper builds an automated UFL preprocessing pass using a DAG traverser that commutes time differentiation with spatial operations, plus two algebraic solver strategies: the Clines block-triangular preconditioner based on an LDU approximation of the stage matrix, and monolithic multigrid with stage-coupled relaxation. The combination turns a user's high-level second-order weak form into a solved stage-coupled system.

What would settle it

Run the Biot manufactured-solution experiment with non-zero smooth initial values for $\boldsymbol{w}_t$ and $p_t$ (for example, $\sin(t)$ evaluated at $t=0$), keeping everything else fixed, and compare the errors and convergence orders; if the errors change at the reported rates, the claim that these artificial initial values do not affect accuracy is false.

Watch

Extended reading notes

Core claim

The paper's central discovery is that RKN stage equations can be written using only one set of stage variables approximating $u_{tt}$, with the solution and velocity updates built from the derived tableau coefficients $\bar A = A^2$ and $\bar b = A^T b$. This avoids auxiliary velocity variables: for a two-stage method on the wave equation, the stage system has $2\times\dim V_h$ unknowns instead of $4\times\dim V_h$. Numerically, a two-stage Gauss-Legendre RKN method takes far fewer time steps than explicit central differences or explicit Nyström, and with the Clines block-triangular preconditioner it is faster than both explicit methods on quadratic hexahedral meshes. On an elastodynamic plate problem with Hsieh-Clough-Tocher elements, RKN runs nearly twice as fast as applying the underlying Runge-Kutta method to the first-order system. For the Biot poroelasticity system, RKN with RadauIIA attains expected convergence orders as the Poisson ratio approaches $0.5$, a regime where the first-order Runge-Kutta formulation with the same solver fails to converge on finer meshes.

Load-bearing premise

The paper assumes that applying the RKN stage replacement to variables whose true time derivative is only first-order, with artificially imposed zero initial values for those derivatives, preserves the expected accuracy; the paper explicitly notes that the available DAE convergence theory does not directly apply to these RKN schemes.

Editorial extensions

If this is right

  • Irksome users can write second-order semidiscrete forms directly and receive an RKN stage-coupled variational problem automatically, without hand-coding a time-stepping loop.
  • With effective preconditioning, fully implicit RKN methods can outperform explicit central differences and explicit Nyström for wave equations at the tested accuracies.
  • For equations with fourth-order spatial derivatives, implicit RKN avoids the explicit stability restriction $\Delta t = O(h^2)$ and runs in roughly half the time of the equivalent first-order Runge-Kutta formulation.
  • For fully dynamic Biot poroelasticity, the RKN-RadauIIA scheme shows the expected convergence orders as $\nu \to 0.5$, where the first-order RK formulation with the same monolithic solver fails to converge on finer grids.

Reading between the lines

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

  • Extrapolating from the reported plate results, other wave-type PDEs currently integrated in first-order form could see similar runtime savings; a natural test would be nonlinear elasticity or acoustic-wave imaging problems.
  • Because the RKN advantage comes from shrinking the stage system rather than changing spatial discretization, the benefit should grow with the number of stages, so higher-order implicit RKN methods deserve a dedicated preconditioning study.
  • The poroelasticity conclusions rest on an unpublished monolithic multigrid solver; replacing it with a publicly available robust preconditioner for the first-order form would clarify whether the performance gap is intrinsic to RKN or depends on the solver.
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

4 major / 4 minor

Summary. This paper reports an extension of the Irksome/Firedrake time-stepping framework to Runge–Kutta–Nyström (RKN) methods for semidiscrete finite element discretizations of PDEs containing second-order time derivatives. Section 2 derives the RKN stage equations from an underlying RK tableau via A-bar = A^2 and b-bar = A^T b, yielding one set of stage unknowns rather than two. Section 3 presents wave, telegraph, dynamic plate, and fully dynamic Biot poroelasticity formulations; Section 4 surveys stage-coupled preconditioners; Section 5 describes UFL-based automated generation of stage-coupled variational problems and their PETSc solver integration. Numerical experiments compare RKN with explicit central differences and with RK applied to first-order forms, including energy conservation and a factor-of-two speedup for RKN over RK. The paper claims smaller algebraic systems, better runtime, competitiveness with explicit methods for wave equations, and essentiality of RKN for higher-order spatial derivatives.

Significance. The core derivation is standard, parameter-free, and cleanly presented, and the software contribution is potentially valuable: a user can write a UFL form with Dt(u,2) and obtain an implicit RKN discretization with automatically generated stage-coupled variational problems. The energy-conservation benchmark for Gauss-Legendre methods and the reproducible Zenodo artifact are concrete strengths. If the mixed-order poroelasticity extension is rigorously supported, the paper would demonstrate a genuinely broader applicability. At present, however, the central novelty claim for mixed first-/second-order systems rests on an unproved assertion, and the abstract overstates the necessity of RKN, so the contribution is promising but not yet fully established.

major comments (4)
  1. [Section 3 (Biot system, Eq. (3.16)) and Section 6.3] The extension of the RKN stage replacement to the mixed first-/second-order Biot system is asserted rather than proved. The text states that introducing non-physical initial values for w_t and p_t and taking them to be zero does not affect accuracy, but no order or stability analysis is given for the resulting mixed-order partitioned scheme, and Section 6.3 explicitly concedes that the cited DAE convergence theory "does not directly apply to the Runge–Kutta–Nyström schemes used here." This is load-bearing because the abstract and Section 2 advertise coupled systems mixing first- and second-order time derivatives, and the poroelasticity experiments are a central validation. The authors should supply either a convergence analysis for this RKN treatment of first-order variables or numerical evidence that isolates temporal errors, for example temporal refinement at a fixed fine spatial mesh and a comparison of zero versus consistent non-zero initial values for w_t and p_t.
  2. [Abstract and Section 6.2] The abstract's claim that RKN methods are "(essentially) required to discretize wave-type equations with higher-order spatial derivatives" is contradicted by the paper's own results: Figure 9 shows that the equivalent first-order system with RK-GL(2) also solves the plate problem, with RKN being roughly twice as fast but not uniquely capable. The unconditional stability of Gauss-Legendre is a property of the underlying RK method and is available in the first-order formulation as well. The wording should be softened to a comparative claim, such as "substantially more efficient in this setting," rather than a claim of necessity.
  3. [Section 6.3, Figures 10 and 11] The observed temporal convergence rates in the Biot experiments are not decisive, because the study refines time and space simultaneously with n_t = n_x/4, and the product norm has at most second-order spatial accuracy for the chosen elements, as the text itself acknowledges. A first-order temporal defect could be masked by a second-order spatial error floor. A temporal-only refinement study at a fixed fine spatial mesh would make the claimed RKN orders (second order for Radau, first order for Gauss) convincing and would also directly test the sensitivity to the non-physical initial values for w_t and p_t.
  4. [Section 6.3] The poroelasticity convergence rates and solver performance rest on the monolithic multigrid solver and DAE theory of [Adler et al. 2025], which is cited as "in preparation." This work is used both to solve the linear systems and to justify the expected convergence rates, so the manuscript is not self-contained on a load-bearing point. The authors should either include sufficient detail on the solver and the applicable convergence theory or clearly mark these results as preliminary and provide alternative supporting evidence.
minor comments (4)
  1. [Equation (3.13)] The clamped boundary conditions are written as u|_Omega = ∂u/∂n|_Omega = 0; the restriction should be to ∂Omega, the boundary of the domain, rather than to Omega itself.
  2. [References] The reference to "Clough and Toucher 1965" should read "Clough and Tocher 1965" to match the standard name of the Hsieh-Clough-Tocher macroelement.
  3. [Section 6.3, product norm definition] The product norm contains a term 10^{-6} h_t ‖w‖^2, where h_t is a time step; the mixed length/time scaling of this term should be clarified or the notation adjusted to avoid dimensional confusion.
  4. [Section 6.1, Figure 6b] The sentence noting that GL(2) "produces some inaccuracy with Q2" on coarse meshes would benefit from a one-sentence explanation, since the phenomenon is not reflected in the asymptotic discussion that follows.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the RKN stage equations are exact algebraic reductions of the underlying RK tableau, and the unresolved Biot assumptions are missing proofs, not constructed equivalences.

full rationale

The central derivation is self-contained. The Runge–Kutta–Nyström stage system in (2.13)–(2.15) is obtained by direct substitution into the first-order RK stage equations (2.11)–(2.12), yielding A-bar = A^2 and b-bar = A^T b in (2.18); there are no fitted constants and no quantity is renamed as a prediction. Gauss–Legendre energy conservation is cited to the standard Hairer–Wanner theory, not to the authors' own work, and the wave-equation comparisons are self-contained against explicit central differences and Nyström's tableau. The mixed-order Biot extension in Section 3 is an explicit ansatz: the paper states “this requires non-physical initial values for w_t and p_t, which we take to be zero” and asserts these values “do not impact the resulting accuracy,” while Section 6.3 concedes that the cited DAE convergence estimates “do not directly apply to the Runge–Kutta–Nyström schemes used here.” That is a missing proof or correctness risk, not a circular reduction by construction. Similarly, the poroelasticity solver performance relies on the co-authored, in-preparation reference [Adler et al. 2025], but the paper's own numerical experiments are the evidence and the RKN derivation does not reduce to that citation. No equation in the paper equals its input by construction, and no fitted input is presented as a prediction.

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

No fitted parameters are used to force the central claims. The explicit time-step sizes in Section 6 are chosen empirically for stability and fairness, but they are reported experimental settings, not hidden parameters used to derive the conclusions. The Biot physical parameters are taken from the cited literature. No new physical entities are introduced; the dDAE boundary condition and extended tableau are notational devices, not entities with independent falsifiable handles.

assumptions (4)
  • domain assumption The UFL DAGTraverser correctly commutes Dt with spatial derivatives, tensor and indexing operations, so that Dt(div(y)) becomes div(Dt(y)) and Dt(sin(t)) becomes cos(t).
    Section 5.1; the entire automated stage assembly depends on this symbolic preprocessing being correct for higher-order derivatives and compound function spaces.
  • ad hoc to paper Runge-Kutta-Nyström stage substitutions can be applied to mixed-order systems where only first-order derivatives appear, with zero non-physical initial derivatives for w_t and p_t not affecting accuracy.
    Section 3 (Biot model); the paper asserts consistency but cites Adler et al. 2025 (in preparation) which the authors admit does not directly cover RKN schemes.
  • domain assumption The stage-coupled algebraic system produced by the substitutions is invertible and the monolithic or Clines preconditioners converge as reported for the tested ranges.
    Sections 4-6; numerical convergence is demonstrated but not proven for all cases, especially near-incompressible poroelasticity.
  • standard math Gauss-Legendre collocation preserves the quadratic energy (3.6) and (3.14) exactly through symplecticity.
    Section 3; used to interpret Figure 7 as exact conservation in practice.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Automated Runge-Kutta-Nystr\"om time stepping for finite element methods in Irksome." pith.science (2026). https://pith.science/paper/AZLF4E7J

@misc{pith2026250820255,
  author       = {Pith},
  title        = {Pith review of: Automated Runge-Kutta-Nystr\"om time stepping for finite element methods in Irksome},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AZLF4E7J}},
  note         = {Machine review of arXiv:2508.20255}
}
read the original abstract

Irksome is a library based on the Unified Form Language (UFL) that automates the application of Runge-Kutta time-stepping methods for finite element spatial discretizations of partial differential equations (PDEs). This paper describes recent updates to Irksome that allow users to express semidiscrete forms of PDEs that contain second-order temporal derivatives, whence it generates stage-coupled variational problems to be solved at each time step for Runge-Kutta-Nystr\"om methods. Firedrake then generates code for these variational problems and provides a rich interface to PETSc for solving them. Directly discretizing second-order time derivatives with Runge-Kutta-Nystr\"om methods provides several advantages relative to discretizing a rewritten first-order system with a standard Runge-Kutta method. Besides working with an interface closer to the problem formulation in UFL, avoiding these auxiliary variables means that Runge-Kutta-Nystr\"om methods lead to smaller algebraic systems and better run-time. Our numerical results indicate that, with effective preconditioning, fully implicit Runge-Kutta-Nystr\"om methods can be made competitive with more traditional explicit methods for wave equations. They are also (essentially) required to discretize wave-type equations with higher-order spatial derivatives. We also provide numerical experiments for fully dynamic poroelasticity, a system of mixed temporal order, where our time-stepping and algebraic solvers perform effectively even as we approach the incompressible limit.

Figures

Figures reproduced from arXiv: 2508.20255 by the authors.

Figure 1
Figure 1. Firedrake listing for the Poisson equation. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Firedrake/Irksome listing for the heat equation ✞ ☎ from firedrake import * from irksome import Dt , NystromTimeStepper , GaussLegendre msh = UnitSquareMesh (16 , 16) x , y = SpatialCoordinate ( msh ) t = Constant (0) f = cos ( pi * x ) * sin (3 * pi * y ) * cos ( t ) V = FunctionSpace ( msh , 'CG ', 2) u = Function ( V ) v = TestFunction ( V ) F = ( inner ( Dt (u , 2) , v ) * dx + inner ( grad ( u ) , grad ( v )) *… view at source ↗
Figure 3
Figure 3. Firedrake/Irksome listing for the wave equation In the rest of the paper, we introduce Runge–Kutta and Runge–Kutta–Nyström methods in Section 2. In addition, we describe the treatment of essential boundary conditions when these methods are applied to finite element discretizations of PDEs. Then, in Section 3, we describe a handful of particular models. These include the standard wave equation, a biharmonic wave equa… view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: UFL representation of the variational problem for a 2-stage implicit Runge–Kutta–Nyström method. [PITH_FULL_IMAGE:figures/full_fig_p013_4.png]
Figure 5
Figure 5. Figure 5: Typical options dictionary for implementing the Clines preconditioner for the wave equation [PITH_FULL_IMAGE:figures/full_fig_p015_5.png]
Figure 6
Figure 6. Figure 6: Comparing accuracy and run-time of two-stage Gauss-Legendre Runge–Kutta–Nyström to explicit central differences and the [PITH_FULL_IMAGE:figures/full_fig_p017_6.png]
Figure 7
Figure 7. Figure 7: Discrete energy over one period of oscillation on a [PITH_FULL_IMAGE:figures/full_fig_p018_7.png]
Figure 8
Figure 8. Figure 8: Error norms computed at the final time for the elastodynamic plate discretized with cubic HCT macroelements and two-stage [PITH_FULL_IMAGE:figures/full_fig_p019_8.png]
Figure 9
Figure 9. Figure 9: Performance comparison of explicit central differences and Runge–Kutta–Nyström and Runge–Kutta methods for the dynamic [PITH_FULL_IMAGE:figures/full_fig_p019_9.png]
Figure 10
Figure 10. Figure 10: Performance comparison of Runge–Kutta–Nyström and Runge–Kutta methods for the Biot manufactured solution problem. [PITH_FULL_IMAGE:figures/full_fig_p021_10.png]
Figure 11
Figure 11. Figure 11: Performance of Runge–Kutta–Nyström method for the Biot manufactured solution problem with Radau time steppers and [PITH_FULL_IMAGE:figures/full_fig_p022_11.png]
Figure 12
Figure 12. Figure 12: Schematic for Biot footing problem. A downward traction is posed on the segmented below the arrows. [PITH_FULL_IMAGE:figures/full_fig_p022_12.png]
Figure 13
Figure 13. Figure 13: Performance of Runge–Kutta–Nyström and Runge–Kutta methods for the Biot footing problem with 2-stage Radau time [PITH_FULL_IMAGE:figures/full_fig_p023_13.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 27 canonical work pages

  1. [2]

    Monolithic multigrid for a reduced-quadrature discretization of poroelasticity. SIAM J. Sci. Comput. 45, 3 (2023), S54–S81. James H. Adler, Xiaozhe Hu, Scott MacLachlan, and Selvabavitha Vijendiran

    work page 2023

  2. [5]

    Communications in Applied Mathematics and Computational Science 18, 1 (2023), 1–28

    Design of DIRK schemes with high weak stage order. Communications in Applied Mathematics and Computational Science 18, 1 (2023), 1–28. J.W. Both, N.A. Barnafi, F.A. Radu, P. Zunino, and A. Quarteroni

    work page 2023

  3. [8]

    arXiv preprint arXiv:2501.14599 (2025)

    FIAT: enabling classical and modern macroelements. arXiv preprint arXiv:2501.14599 (2025). Mark H Carpenter, David Gottlieb, Saul Abarbanel, and Wai-Sun Don

    arXiv 2025

  4. [11]

    Efficient Order-Optimal Preconditioners for Implicit Runge-Kutta and Runge-Kutta-Nystr\"om Methods Applicable to a Large Class of Parabolic and Hyperbolic PDEs

    Efficient Order-Optimal Preconditioners for Implicit Runge-Kutta and Runge-Kutta- Nyström Methods Applicable to a Large Class of Parabolic and Hyperbolic PDEs. arXiv preprint arXiv:2206.08991 (2022). Ray W. Clough and J. L. Toucher

    work page Pith review arXiv 2022

  5. [18]

    Imperial College London and University of Oxford and Baylor University and University of Washington

    Firedrake User Manual (first edition ed.). Imperial College London and University of Oxford and Baylor University and University of Washington. https://doi.org/10.25561/104839 Van Emden Henson and Ulrike Meyer Yang

    doi:10.25561/104839

  6. [19]

    Applied Numerical Mathematics 41 (2002), 155–177

    BoomerAMG: a Parallel Algebraic Multigrid Solver and Preconditioner. Applied Numerical Mathematics 41 (2002), 155–177. Volker John, Alexander Linke, Christian Merdon, Michael Neilan, and Leo G Rebholz

    work page 2002

  7. [22]

    ACM Transactions on Mathematical Software (TOMS) (2025)

    Extending Irksome: improvements in automated Runge–Kutta time stepping for finite element methods. ACM Transactions on Mathematical Software (TOMS) (2025). https://doi.org/10.1145/3759245 To appear. Robert C. Kirby and Lawrence Mitchell

    doi:10.1145/3759245 2025

  8. [24]

    Time-continuous strongly conservative space-time finite element methods for the dynamic Biot model. (2025). arXiv:2507.19955 [math.NA] https://arxiv.org/abs/2507.19955 Anders Logg, Kent-Andre Mardal, and Garth N. Wells (Eds.)

    arXiv 2025

Show all 36 references
  1. [30]

    Stage-parallel fully implicit Runge–Kutta solvers for discontinuous Galerkin fluid simulations. J. Comput. Phys. 335 (2017), 700–717. Md Masud Rana, Victoria E Howle, Katharine Long, Ashley Meek, and William Milestone

    2017

  2. [32]

    SIAM Journal on Scientific Computing (2025), S183–S206

    Efficient solution of fully implicit Runge–Kutta methods for linear wave equations. SIAM Journal on Scientific Computing (2025), S183–S206. Florian Rathgeber, David A. Ham, Lawrence Mitchell, Michael Lange, Fabio Luporini, Andrew T. T. McRae, Gheorghe-Teodor Bercea, Graham R. ...

    2025

  3. [33]

    ACM Trans

    Firedrake: automating the finite element method by composing abstractions. ACM Trans. Math. Software 43, 3 (2016), Manuscript submitted to ACM 26 Robert C. Kirby, Scott P. MacLachlan, and Pablo D. Brubeck 24:1–24:27. https://doi.org/10.1145/2998441 arXiv:1501.01809 Ben S. Sout...

    2016 arXiv

  4. [36]

    The SMAI Journal of computational mathematics 9 (2023), 61–93

    Theoretical and practical aspects of space-time DG-SEM implementations. The SMAI Journal of computational mathematics 9 (2023), 61–93. Gerhard Wanner and Ernst Hairer

    2023

  5. [84]

    https://doi.org/10.1007/978-3-642-23099-8 Lawrence Mitchell and Eike Hermann Müller

    Springer. https://doi.org/10.1007/978-3-642-23099-8 Lawrence Mitchell and Eike Hermann Müller

    doi

  6. [1925]

    Acta Soc

    Über die numerische Integration von Differentialgleichungen. Acta Soc. Sci. Fenn. 50, 13 (1925), 1–54. D Pathria

    1925

  7. [1956]

    Theory of Propagation of Elastic Waves in a Fluid-Saturated Porous Solid. I. Low-Frequency Range. The Journal of the Acoustical Society of America 28, 2 (03 1956), 168–178. https://doi.org/10.1121/1.1908239 Abhijit Biswas, David I. Ketcheson, Benjamin Seibold, and David Shirokoff

    1956 doi

  8. [1963]

    BIT Numerical Mathematics 3, 1 (1963), 27–43

    A special stability problem for linear multistep methods. BIT Numerical Mathematics 3, 1 (1963), 27–43. Andreas Dedner, Robert Kloefkorn, and Martin Nolte

    1963

  9. [1985]

    Order results for implicit Runge–Kutta methods applied to stiff systems. SIAM J. Numer. Anal. 22, 3 (1985), 515–534. Guosheng Fu

    1985

  10. [1995]

    SIAM Journal on Scientific Computing 16, 6 (1995), 1241–1252

    The theoretical accuracy of Runge–Kutta time discretizations for the initial boundary value problem: a study of the boundary error. SIAM Journal on Scientific Computing 16, 6 (1995), 1241–1252. Manuscript submitted to ACM Automated Runge–Kutta–Nyström time stepping for finite ...

    1995

  11. [1997]

    SIAM Journal on Scientific Computing 18, 5 (1997), 1255–1266

    The correct formulation of intermediate boundary conditions for Runge–Kutta time integration of initial boundary value problems. SIAM Journal on Scientific Computing 18, 5 (1997), 1255–1266. Will Pazner and Per-Olof Persson

    1997

  12. [1998]

    Computer Methods in Applied Mechanics and Engineering 163, 1-4 (1998), 193–204

    Projection techniques for iterative solution of Ax= b with successive right-hand sides. Computer Methods in Applied Mechanics and Engineering 163, 1-4 (1998), 193–204. Reinhard Frank, Josef Schneid, and Christoph W. Ueberhuber

    1998

  13. [1999]

    Journal of Engineering Mathematics 35 (1999), 405–426

    Higher-order triangular and tetrahedral finite elements with mass lumping for solving the wave equation. Journal of Engineering Mathematics 35 (1999), 405–426. Michael R Clines, Victoria E Howle, and Katharine R Long

    1999

  14. [2002]

    In Computational Science—ICCS 2002: International Conference Amsterdam, The Netherlands, April 21–24, 2002 Proceedings, Part III

    hypre: A library of high performance preconditioners. In Computational Science—ICCS 2002: International Conference Amsterdam, The Netherlands, April 21–24, 2002 Proceedings, Part III . Springer, 632–641. Patrick E. Farrell, Robert C. Kirby, and Jorge Marchena-Menendez. 2021a. ...

    2021

  15. [2005]

    SIAM Journal on Scientific Computing 27, 1 (2005), 67–92

    Multigrid methods for implicit Runge–Kutta and boundary value method discretizations of parabolic PDEs. SIAM Journal on Scientific Computing 27, 1 (2005), 67–92. https://doi.org/10.1137/030601144 Lea Miko Versbach, Viktor Linders, Robert Klöfkorn, and Philipp Birken

    2005 doi

  16. [2006]

    Modeling, Identification, and Control 27, 1 (2006), 109–123

    Preconditioning of fully implicit Runge-Kutta schemes for parabolic PDEs. Modeling, Identification, and Control 27, 1 (2006), 109–123. Jan Van Lent and Stefan Vandewalle

    2006

  17. [2011]

    Numerical Methods for Partial Differential Equations 27, 5 (2011), 1290–1312

    Order optimal preconditioners for fully implicit Runge-Kutta schemes applied to the bidomain equations. Numerical Methods for Partial Differential Equations 27, 5 (2011), 1290–1312. E. J. Nyström

    2011

  18. [2012]

    In 2012 11th International Symposium on Parallel and Distributed Computing

    Composable linear solvers for multiphysics. In 2012 11th International Symposium on Parallel and Distributed Computing . IEEE, 55–62. Pablo D. Brubeck and Robert C. Kirby

    2012

  19. [2016]

    High level implementation of geometric multigrid solvers for finite element problems: Applications in atmospheric modelling. J. Comput. Phys. 327 (2016), 1–18. Trygve K Nilssen, Gunnar A Staff, and Kent-Andre Mardal

    2016

  20. [2017]

    SIAM Rev

    On the divergence constraint in mixed finite element methods for incompressible flows. SIAM Rev. 59, 3 (2017), 492–544. Robert C Kirby

    2017

  21. [2018]

    SIAM Journal on Scientific Computing 40, 1 (2018), C76–C98

    Solver composition across the PDE/linear algebra barrier. SIAM Journal on Scientific Computing 40, 1 (2018), C76–C98. https://doi.org/10.1137/17M1133208 Johannes Kraus, Maria Lymbery, and Kevin Osthues

    2018 doi

  22. [2019]

    Computers & Mathematics with Applications 77 (2019), 237–252

    A high-order HDG method for the Biot’s consolidation model. Computers & Mathematics with Applications 77 (2019), 237–252. Michael W. Gee, Chris M. Siefert, Jonathan J. Hu, Ray S. Tuminaro, and Marzio G. Sala

    2019

  23. [2020]

    Python bindings for the DUNE-FEM module. Zenodo. doi 10 (2020). Todd Dupont. 1973.𝐿2-estimates for Galerkin methods for second order hyperbolic equations. SIAM J. Numer. Anal. 10, 5 (1973), 880–889. Robert D Falgout and Ulrike Meier Yang

    2020

  24. [2021]

    SIAM Journal on Scientific Computing 43, 5 (2021), S475–S495

    A new block preconditioner for implicit Runge–Kutta methods for parabolic PDE problems. SIAM Journal on Scientific Computing 43, 5 (2021), S475–S495. Aman Rani, Pieter Ghysels, Victoria Howle, Katharine Long, and Michal Outrata

    2021

  25. [2022]

    Computer Methods in Applied Mechanics and Engineering 388 (2022), 114183

    Iterative splitting schemes for a soft material poromechanics model. Computer Methods in Applied Mechanics and Engineering 388 (2022), 114183. https://doi.org/10.1016/j.cma.2021.114183 Jed Brown, Matthew G. Knepley, David A. May, Lois Curfman McInnes, and Barry Smith

    2022

  26. [2023]

    Monolithic multigrid for implicit Runge-Kutta discretizations of incompressible fluid flow. J. Comput. Phys. 478 (2023), 111961. https://doi.org/10.1016/j.jcp.2023.111961 James H. Adler, Yunhui He, Xiaozhe Hu, Scott MacLachlan, and Peter Ohm

    2023

  27. [2024]

    SIAM Journal on Scientific Computing (2024)

    On the convergence of monolithic multigrid for implicit Runge-Kutta time stepping of finite element problems. SIAM Journal on Scientific Computing (2024). To appear. Robert C Kirby and Scott P MacLachlan

    2024

  28. [2025]

    Robust monolithic multigrid solvers for Runge–Kutta discretizations of Biot elasticity. (2025). In preparation. Maurice A. Biot

    2025

Pith tools

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