pith. sign in

arxiv: 2604.14479 · v2 · submitted 2026-04-15 · 🧮 math.NA · cs.NA

Smooth perturbations of diagonally implicit Runge--Kutta methods

John Driscoll , Sigal Gottlieb , Zachary J. Grant , C\'esar Herrera , Tej Sai Kakumanu , Monica Stephens This is my paper

Pith reviewed 2026-05-10 11:51 UTC · model grok-4.3

classification 🧮 math.NA cs.NA
keywords Diagonally implicit Runge-Kuttasmooth perturbationslocal consistency conditionsmixed accuracy methodsorder preservationstability analysisnumerical ODE solverscomputational efficiency
0
0 comments X

The pith

Diagonally implicit Runge-Kutta methods retain high accuracy with smooth lower-cost perturbations when local consistency conditions hold.

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

The paper extends an existing mixed-accuracy approach to Runge-Kutta methods by developing a dedicated accuracy and stability analysis for the case of smooth perturbations. These perturbations come from using cheaper, lower-accuracy models or under-resolved solvers in the implicit stages of DIRK methods. The authors derive additional local consistency conditions that let the overall method keep its designed high order despite the cheaper stages. They then construct new methods that exploit these conditions and verify through numerical tests that the expected accuracy and stability are achieved while computation time drops. A reader would care because many stiff differential equation simulations are limited by the cost of implicit solves, and this supplies a systematic route to reduce that cost without losing convergence rate.

Core claim

For smooth perturbations that arise when the implicit operator in a DIRK method is replaced by a less expensive lower-accuracy version, the method remains high order provided the perturbation satisfies extra local consistency conditions. The analysis extends the prior mixed-accuracy framework directly to this smooth setting, allowing the design of novel perturbed DIRK methods that strategically employ the cheaper operator to cut computational cost while preserving both accuracy and stability.

What carries the argument

Accuracy and stability analysis for smooth perturbations within the mixed-accuracy framework, using additional local consistency conditions on the perturbation to preserve the method order.

If this is right

  • New DIRK methods can be constructed that remain high order specifically for smooth perturbations meeting the local consistency conditions.
  • Computational cost is reduced by replacing the original implicit operator with a lower-accuracy version in selected stages.
  • Stability properties of the perturbed methods can be analyzed and maintained within the same framework used for the unperturbed case.
  • Numerical tests confirm that the designed methods achieve the predicted order and stability for several classes of smooth perturbations.
  • Different types of smooth perturbations produce predictable effects on accuracy and stability that can be controlled by the choice of method.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same local consistency conditions could be checked or enforced when replacing full physics models with reduced-order surrogates inside time-stepping codes.
  • An adaptive strategy might vary the accuracy of the perturbation stage by stage according to local error estimates without losing the global order guarantee.
  • The approach may extend to other families of implicit Runge-Kutta or multistep methods that rely on repeated operator evaluations.

Load-bearing premise

The smooth perturbations must satisfy the extra local consistency conditions that let the overall scheme keep its designed order.

What would settle it

Apply one of the newly designed methods to a test problem whose smooth perturbation violates at least one local consistency condition and observe that the measured order of accuracy falls below the method's design order.

Figures

Figures reproduced from arXiv: 2604.14479 by C\'esar Herrera, John Driscoll, Monica Stephens, Sigal Gottlieb, Tej Sai Kakumanu, Zachary J. Grant.

Figure 4.1
Figure 4.1. Figure 4.1: Burgers’ Equation: Final convergence plots for the diagonally perturbed DIRK [PITH_FULL_IMAGE:figures/full_fig_p008_4_1.png] view at source ↗
Figure 4.2
Figure 4.2. Figure 4.2: Burgers’ equation: final time errors using the novel [PITH_FULL_IMAGE:figures/full_fig_p010_4_2.png] view at source ↗
Figure 4.3
Figure 4.3. Figure 4.3: Shallow water equations: final time errors for dif [PITH_FULL_IMAGE:figures/full_fig_p012_4_3.png] view at source ↗
Figure 4.4
Figure 4.4. Figure 4.4: Porous Medium equation: final time errors for [PITH_FULL_IMAGE:figures/full_fig_p014_4_4.png] view at source ↗
Figure 4.5
Figure 4.5. Figure 4.5: Porous medium equation final time errors with [PITH_FULL_IMAGE:figures/full_fig_p015_4_5.png] view at source ↗
read the original abstract

A mixed accuracy framework for Runge--Kutta methods presented in [Grant, JSC 2022] has been shown to speed up the computation in diagonally implicit Runge--Kutta (DIRK) methods by using less expensive low accuracy approaches for the implicit stages. This theory included both smooth and nonsmooth perturbations, and subsequent work focused primarily on the case of nonsmooth perturbations that arise from mixed precision simulations. In this work the focus is on smooth perturbations that arise from using less accurate models or under-resolved iterative solvers to simplify the implicit computations. We develop an accuracy and stability analysis based on the framework in [Grant, JSC 2022] to design methods that strategically replace the original operator by a lower accuracy operator to reduce computational cost while mitigating the effect of the perturbations. In particular, we focus on designing novel methods that are high order for smooth perturbations that satisfy additional local consistency conditions. Finally, we verify the performance of the novel perturbed DIRK methods designed in this work and numerically study the impact of different types of smooth perturbations on the accuracy and stability of the methods.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 3 minor

Summary. The manuscript extends the mixed-accuracy framework of Grant (JSC 2022) from nonsmooth to smooth perturbations in diagonally implicit Runge-Kutta (DIRK) methods. It imposes additional local consistency conditions on the perturbations, derives the resulting order conditions and modified stability functions, designs several novel perturbed DIRK schemes that retain high order, and presents numerical experiments verifying accuracy and stability under different smooth perturbation types.

Significance. If the derivations hold, the work supplies a practical route to lower-cost implicit stages in stiff ODE solvers while preserving design order, addressing a gap left by prior emphasis on nonsmooth (mixed-precision) perturbations. Credit is due for the explicit adaptation of the Grant framework, the derivation of order conditions and stability functions for the smooth case, and the inclusion of numerical tests that directly probe the effect of perturbation smoothness on observed order.

major comments (1)
  1. [§3.2] §3.2, the statement of the perturbed order conditions: the additional local consistency conditions (Eq. (15)) are required for the high-order claim, yet the manuscript does not show that these conditions are compatible with the standard DIRK simplifying assumptions (e.g., the row-sum conditions) when the method coefficients are chosen to satisfy both sets simultaneously; this compatibility is load-bearing for the existence of the novel methods.
minor comments (3)
  1. [Figure 3] Figure 3: the stability-region plots for the perturbed methods would benefit from an overlay of the unperturbed DIRK stability region to make the perturbation effect visually quantifiable.
  2. [§5.1] §5.1: the description of the under-resolved iterative solver perturbation lacks a precise definition of the truncation tolerance used; this makes it difficult to reproduce the exact perturbation magnitude in the numerical study.
  3. [Introduction] The introduction could cite one concrete application (e.g., from climate or combustion modeling) where smooth model-reduction perturbations naturally arise, to strengthen the motivation.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their careful reading of the manuscript, positive assessment of its contribution, and recommendation for minor revision. We address the single major comment below.

read point-by-point responses

  1. Referee: [§3.2] §3.2, the statement of the perturbed order conditions: the additional local consistency conditions (Eq. (15)) are required for the high-order claim, yet the manuscript does not show that these conditions are compatible with the standard DIRK simplifying assumptions (e.g., the row-sum conditions) when the method coefficients are chosen to satisfy both sets simultaneously; this compatibility is load-bearing for the existence of the novel methods.

    Authors: We agree that explicit verification of compatibility is necessary to support the existence of the novel high-order methods. The coefficients of the perturbed DIRK schemes presented in the paper were in fact chosen to satisfy both the standard DIRK row-sum conditions and the additional local consistency conditions (15) simultaneously. To address the referee's point, we will revise §3.2 by adding a short verification paragraph (or table) that confirms this joint satisfaction for each constructed method, thereby making the compatibility explicit and reinforcing the validity of the order claims. revision: yes

Circularity Check

0 steps flagged

No circularity: extension of prior framework with independent derivations

full rationale

The manuscript cites the Grant JSC 2022 mixed-accuracy framework as the foundation but then derives new local consistency conditions, order conditions, stability functions, and numerical results specifically for smooth perturbations of DIRK methods. These steps add content that does not reduce to the cited framework by definition or by fitting; the central claims rest on explicit new analysis and verification rather than self-referential closure. Self-citation is present but not load-bearing for the novel contributions.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on extending the mixed-accuracy framework of Grant 2022 to smooth perturbations and on the existence of local consistency conditions that restore high order; no new free parameters or invented entities are introduced in the abstract.

axioms (1)
  • domain assumption The mixed accuracy framework presented in Grant JSC 2022 applies to smooth perturbations arising from less accurate models or under-resolved solvers.
    The paper states it develops the analysis based on that framework.

pith-pipeline@v0.9.0 · 5503 in / 1242 out tokens · 27491 ms · 2026-05-10T11:51:38.487973+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

11 extracted references · 11 canonical work pages

  1. [1]

    Broyden , A class of methods for solving nonlinear simultaneous equations , Mathematics of Computation, 19 (1965), pp

    C. Broyden , A class of methods for solving nonlinear simultaneous equations , Mathematics of Computation, 19 (1965), pp. 577--593

    work page 1965

  2. [2]

    Burnett, S

    B. Burnett, S. Gottlieb, and Z. J. Grant , Stability analysis and performance evaluation of additive mixed-precision R unge- K utta methods , Commun. Appl. Math. Comput., 6 (2024), p. 705–738

    work page 2024

  3. [3]

    Burnett, S

    B. Burnett, S. Gottlieb, Z. J. Grant, and A. Heryudono , Performance evaluation of mixed-precision R unge- K utta methods , in 2021 IEEE High Performance Extreme Computing Conference (HPEC), 2021, pp. 1--6

    work page 2021

  4. [4]

    Crouzeix , Sur la b-stabilité des méthodes de R unge- K utta , Numerische Mathematik, 32 (1979), pp

    M. Crouzeix , Sur la b-stabilité des méthodes de R unge- K utta , Numerische Mathematik, 32 (1979), pp. 75--82

    work page 1979

  5. [5]

    Driscoll, S

    J. Driscoll, S. Gottlieb, Z.J. Grant, C. Herrera, T.S. Kakumanu, M. H. Sawicki, and M. Stephens , Stable corrections for perturbed diagonally implicit Runge-Kutta methods , Submitted (2026). https://arxiv.org/abs/2603.24451

    work page arXiv 2026

  6. [6]

    Driscoll, S

    J. Driscoll, S. Gottlieb, Z.J. Grant, C. Herrera, T.S. Kakumanu Smooth Perturbations DIRK methods (2026) https://github.com/Mixed-Precision/SmoothPerturbationsDIRKmethods

    work page 2026

  7. [7]

    Z. J. Grant , Perturbed Runge-Kutta methods for mixed precision applications , Journal of Scientific Computing, 92 (2022), pp. 1--20

    work page 2022

  8. [8]

    Hairer and G

    E. Hairer and G. Wanner , Solving Ordinary Differential Equations II: Stiff and Differential-Algebraic Problems , vol. 14 of Springer Series in Computational Mathematics, Springer-Verlag, Berlin, Heidelberg, 2nd ed., 1996

    work page 1996

  9. [9]

    C. A. Kennedy and M. H. Carpenter , Diagonally implicit Runge-Kutta methods for ordinary differential equations. a review , NASA Technical Report, NASA/TM–2016–219173 (2016)

    work page 2016

  10. [10]

    Kurdi , Stable High Order Methods for Time Discretization of Stiff Differential Equations , Ph.D

    M.A. Kurdi , Stable High Order Methods for Time Discretization of Stiff Differential Equations , Ph.D. Thesis, Applied Mathematics, Univ. of California, Berkeley, Berkeley (1974)

    work page 1974

  11. [11]

    Nørsett , Semi explicit Runge-Kutta methods , Math

    S. Nørsett , Semi explicit Runge-Kutta methods , Math. and Comp. Rpt. 6/74 Dept. of Math., Univ. Trondheim, (1974)

    work page 1974