pith. sign in

arxiv: 2605.16995 · v1 · pith:C4WFFQB3new · submitted 2026-05-16 · 🧮 math.NA · cs.NA

Low Stage High Order Explicit Runge--Kutta Methods via Q- and D-Conditions: General Theory and Efficient Recursive Construction

Junyuan He , Jizu Huang This is my paper

Pith reviewed 2026-05-19 19:10 UTC · model grok-4.3

classification 🧮 math.NA cs.NA
keywords explicit Runge-Kutta methodsorder conditionsButcher tableaurecursive constructionnumerical ODE integrationsimplifying assumptionsstage reduction
0
0 comments X

The pith

A Q/D-space reformulation of order conditions yields explicit Runge-Kutta methods of even order p with stage count (p squared minus 2p plus 8) over 4.

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

The paper introduces a framework of sufficient order conditions for explicit Runge-Kutta methods that rests on Q- and D-spaces defined by their residual vectors. These conditions, paired with the classical B(p) requirements, are shown to guarantee overall order p. The same framework supplies a recursive procedure that obtains the Butcher coefficients by solving two structured linear systems at each step. For every even order p at least 4 the procedure produces a method whose stage number equals (p squared minus 2p plus 8) divided by 4, matching the quadratic term of the classical Gragg families while improving the linear term. Retained free parameters inside the construction give a systematic route to tuning stability or short-time accuracy.

Core claim

The Q- and D-space residual conditions together with B(p) form a set of sufficient conditions that guarantee order p for an explicit Runge-Kutta method; the same conditions admit a recursive construction in which the Butcher coefficients are obtained from two structured linear systems, producing, for every even p greater than or equal to 4, a method whose stage count is exactly (p squared minus 2p plus 8) divided by 4.

What carries the argument

The Q- and D-spaces, defined through their residual vectors, which reformulate Butcher's classical simplifying assumptions into a pair of simplified linear spaces whose residuals supply the sufficient order conditions.

If this is right

  • For each even p greater than or equal to 4 an explicit Runge-Kutta method of order p exists with stage count (p squared minus 2p plus 8) divided by 4.
  • The construction is recursive, so higher-order methods can be built systematically from lower-order ones by solving two linear systems per step.
  • Free parameters left by the construction can be chosen to improve linear stability or short-time accuracy without losing the guaranteed order.
  • The Q/D framework supplies a concrete generalization of Butcher's simplifying assumptions that still guarantees the full order when combined with B(p).

Where Pith is reading between the lines

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

  • The improved linear term in the stage count may translate into lower computational cost per step for problems where function evaluations dominate the runtime.
  • The retained free parameters could be optimized for particular classes of problems such as Hamiltonian systems or stiff oscillatory equations.
  • Because the construction is recursive, it may be possible to derive families with additional algebraic properties such as symplecticity or preservation of quadratic invariants by imposing extra linear constraints on the free parameters.

Load-bearing premise

The Q- and D-space residual conditions remain linearly independent and the two structured linear systems at each recursive step remain solvable for the chosen stage count.

What would settle it

Explicitly construct the method for p equals 6 using the recursive procedure, insert the resulting Butcher tableau into a test equation with known exact solution, and verify that the observed order of the local truncation error is exactly 6.

Figures

Figures reproduced from arXiv: 2605.16995 by Jizu Huang, Junyuan He.

Figure 1
Figure 1. Figure 1: Graphical conventions for basic multilinear operations. Vertices represent tensors or [PITH_FULL_IMAGE:figures/full_fig_p009_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: TTN representations of typical expressions appearing in the [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Illustration of the Q-reduction procedure. The following theorem summarizes the sufficient conditions for an explicit RK method to achieve order p within our framework. Theorem 3.2. An explicit Runge–Kutta method has order at least p if the following conditions hold for some m, n with m ≥ n − 1 and m + n + 1 ≥ p: 1. Quadrature conditions for [ k−1 ], denoted B(p): b · c ⊙(k−1) = 1 k , k = 1, . . . , p. 2. … view at source ↗
Figure 4
Figure 4. Figure 4: Schematic tree decomposition and cancellation mechanism in the sufficiency proof for [PITH_FULL_IMAGE:figures/full_fig_p017_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Schematic tree decomposition and cancellation mechanism for the two subcases of case [PITH_FULL_IMAGE:figures/full_fig_p018_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Case 2.1: reduction to a weak QD orthogonality term. dk × 1 I 3 s q ∈ Q m AΦ(t1,1) = dk × 1 I 3 s q AI4 s AΦ(t1,1,3) AΦ(t1,1,2) AΦ(t1,1,1) definition of Wn −−−−−−−−−−→ w × 1 I 4 s AΦ(t1,1,3) AΦ(t1,1,2) AΦ(t1,1,1) + dk AI5 s AΦ q (t1,1,3) AΦ(t1,1,2) AΦ(t1,1,1) where dk AI5 s AΦ q (t1,1,3) AΦ(t1,1,2) AΦ(t1,1,1) dk+1=dk ×1 A =========== dk+1 × 1 I 5 s AΦ q (t1,1,3) AΦ(t1,1,2) AΦ(t1,1,1) [PITH_FULL_IMAGE:figu… view at source ↗
Figure 7
Figure 7. Figure 7: The pivot residual condition P R(n) is used to show that the remainder term in case 2.2 vanishes. The term is split into two terms by the definition of Wn, where the first term vanishes by the P R(n) condition and the second term is reduced to a similar tree with a new root and smaller height. Remark 3.1. The PR condition is a technical condition that is only needed for the proof, and it is not clear wheth… view at source ↗
Figure 8
Figure 8. Figure 8: Stage arrangement for p = 10. The two regions, D-region and Q-region in the figure, correspond to the parameters in matrix A that will be utilized to construct D-type spaces and Q-type spaces, respectively [PITH_FULL_IMAGE:figures/full_fig_p034_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Spatial configuration of the D-system equations on the Butcher tableau A for p = 10. The matrix shows the distribution of different equation types: dk regions and εi,j ×1 A regions. Blank areas represent the entries that are initialized to zero, and shaded areas indicate entries that become zero as a direct consequence of solving the constructed D-system equations. Dashed lines represent the boundary of th… view at source ↗
Figure 10
Figure 10. Figure 10: Spatial configuration of the Q-system equations on the Butcher tableau A for p = 10. The matrix shows the distribution of different equation types: qk regions and Aej regions. Blank areas in the Q-region represent the entries that are initialized to zero, and shaded areas indicate entries that become zero as a direct consequence of solving the constructed Q-system equations. Dashed lines represent the bou… view at source ↗
Figure 11
Figure 11. Figure 11: Sparsity pattern for matrix A ∈ R 32×32 of an order 12 method using our construction. The dotted position indicates nonzero elements [PITH_FULL_IMAGE:figures/full_fig_p063_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Stability regions of the constructed ERK methods for orders [PITH_FULL_IMAGE:figures/full_fig_p065_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Stability regions of the Butcher’s ERK methods for orders [PITH_FULL_IMAGE:figures/full_fig_p065_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Numerical validation of stability regions of the constructed ERK methods for orders [PITH_FULL_IMAGE:figures/full_fig_p069_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Growth of the numerical discrepancy ∥uh(t) − uh/2(t)∥ for the Lorenz–63 system computed with several ERK methods. The discrepancy is measured only at coarse-grid time points. The horizontal dashed line indicates the predictability tolerance ε = 10−6 . Higher-order methods exhibit reduced initial transients and longer predictability time. Numerical results [PITH_FULL_IMAGE:figures/full_fig_p070_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Stability region comparison between optimized and original RK methods. The shaded [PITH_FULL_IMAGE:figures/full_fig_p075_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: Lorenz system global error measurements. ( [PITH_FULL_IMAGE:figures/full_fig_p076_17.png] view at source ↗
Figure 18
Figure 18. Figure 18: Predictability plot showing ∥yh(t) − yh/2(t)∥2 versus time. A horizontal tolerance line captures trajectory separation threshold. The optimized method records a predictability time of 21.62, exhibiting a slight improvement over the original baseline method (21.58). modest increase is consistent with the Lorenz dynamics: once trajectory separation enters the exponential-growth regime, even a noticeable red… view at source ↗
Figure 19
Figure 19. Figure 19: The ordering of the variables in the D-system. We write them explicitly column by column. The RK method to be constructed has s = 22 stages. The equations are listed in [PITH_FULL_IMAGE:figures/full_fig_p081_19.png] view at source ↗
Figure 20
Figure 20. Figure 20: The ordering of the variables in the Q-system. We write them explicitly row by row. The RK method for p = 10 has s = 22 stages. The equations for each variable is given in [PITH_FULL_IMAGE:figures/full_fig_p086_20.png] view at source ↗
read the original abstract

Constructing explicit Runge--Kutta (ERK) methods with as few stages as possible for a given order is a classical problem in numerical analysis. In this work, we introduce a $Q$/$D$-space framework of sufficient order conditions for ERK methods. This framework generalizes Butcher's classical simplifying assumptions by reformulating them in terms of simplified $Q$- and $D$-spaces defined through their residual vectors. It yields sufficient conditions which, together with $B(p)$, ensure order $p$. It also leads to a recursive construction procedure for ERK methods of arbitrary even order, in which the Butcher coefficients are obtained from two structured linear systems. For every even order $p\ge 4$, the construction produces ERK methods with stage number $s(p)=(p^2-2p+8)/4$. This stage count has the same leading term as that of the classical Gragg families, while improving the linear term. The free parameters retained by the construction further provide a systematic framework for designing methods with enhanced stability and short-time accuracy.

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 / 2 minor

Summary. The paper introduces a Q/D-space framework that reformulates Butcher's simplifying assumptions via residual vectors to obtain sufficient order conditions for explicit Runge-Kutta methods. Combined with the classical B(p) condition, these yield order p. The framework supports a recursive construction in which Butcher coefficients are obtained from two structured linear systems at each step. For every even order p ≥ 4 the construction produces methods with stage count s(p) = (p² - 2p + 8)/4, which matches the leading term of the classical Gragg families while improving the linear term; free parameters remain for stability or accuracy tuning.

Significance. If the linear-independence claims hold, the work supplies an algebraic route to explicit Runge-Kutta methods whose stage count is asymptotically optimal among known families and whose construction is fully recursive and parameter-free at the order level. The retention of free parameters after the order conditions are satisfied is a concrete strength that enables systematic optimization of stability regions or short-time accuracy without restarting the order derivation.

major comments (1)
  1. [§4] §4, recursive step: the solvability of the two structured linear systems at each recursion level is asserted on the basis of linear independence of the Q- and D-residual vectors for the chosen stage count s(p). The manuscript verifies this for base cases and small even p, but does not supply a general rank argument that the coefficient matrices have full row rank equal to the number of imposed conditions for arbitrary even p. Because the stage-count formula s(p) = (p² - 2p + 8)/4 is derived from this count, a missing general independence proof is load-bearing for the central claim.
minor comments (2)
  1. [§2-4] Notation: the definition of the Q- and D-spaces via residual vectors is introduced in §2 but the precise mapping from residual vectors to the rows of the two linear systems is not restated in §3-4; a short summary table or diagram would improve readability.
  2. [Abstract] The abstract states that the construction 'yields sufficient conditions which, together with B(p), ensure order p'; this should be cross-referenced to the precise theorem number in the main text.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the positive evaluation of the significance of the Q/D-space framework and for the constructive identification of the need for a general rank argument. We address the single major comment below and will revise the manuscript to incorporate the requested strengthening.

read point-by-point responses

  1. Referee: [§4] §4, recursive step: the solvability of the two structured linear systems at each recursion level is asserted on the basis of linear independence of the Q- and D-residual vectors for the chosen stage count s(p). The manuscript verifies this for base cases and small even p, but does not supply a general rank argument that the coefficient matrices have full row rank equal to the number of imposed conditions for arbitrary even p. Because the stage-count formula s(p) = (p² - 2p + 8)/4 is derived from this count, a missing general independence proof is load-bearing for the central claim.

    Authors: We agree that the absence of a general proof of full row rank for arbitrary even p constitutes a genuine gap in the current exposition. The manuscript establishes the required linear independence only by direct verification for the base cases p=4 and p=6 together with explicit rank computations for selected larger even orders. In the revised version we will add an inductive argument in §4 showing that the coefficient matrices retain full row rank at every recursion level. The induction proceeds from the block-triangular structure of the Q- and D-residual vectors: each newly introduced stage augments the residual space by a vector whose leading nonzero entry lies outside the span of the preceding residuals, and the specific choice s(p)=(p²-2p+8)/4 guarantees that the number of free parameters is sufficient to maintain this separation. The revised text will include the explicit inductive step together with a short appendix verifying the base of the induction. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation rests on classical B(p) and external linear algebra

full rationale

The paper defines Q- and D-spaces via residual vectors, reformulates Butcher's simplifying assumptions, and states that the resulting conditions plus the classical B(p) are sufficient to guarantee order p. The recursive construction obtains Butcher coefficients by solving two structured linear systems whose dimensions are set by the chosen stage count s(p). No step reduces the order claim or the stage count to a fitted parameter or self-citation by construction; solvability is asserted on the basis of the framework and linear independence of the residual vectors rather than by redefining the target result inside the same equations. The free parameters are explicitly retained for later use and do not enter the order or stage-count claims. The derivation is therefore self-contained against the external benchmarks of Butcher theory and standard linear algebra.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on the linear independence of the Q- and D-residual conditions at each order and on the solvability of the two structured systems; these are domain assumptions rather than new axioms or fitted constants.

axioms (2)
  • domain assumption The Q-space and D-space residual vectors remain linearly independent when the stage count is set to (p²-2p+8)/4
    Invoked to guarantee that the two linear systems determine the Butcher coefficients uniquely up to free parameters.
  • domain assumption B(p) together with the Q/D conditions are sufficient for order p
    Stated in the abstract as the key sufficiency result that replaces the full set of Butcher order conditions.

pith-pipeline@v0.9.0 · 5722 in / 1546 out tokens · 30647 ms · 2026-05-19T19:10:04.663515+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

38 extracted references · 38 canonical work pages · 2 internal anchors

  1. [1]

    Tensor networks and hi- erarchical tensors for the solution of high-dimensional partial differential equations

    Markus Bachmayr, Reinhold Schneider, and André Uschmajew. “Tensor networks and hi- erarchical tensors for the solution of high-dimensional partial differential equations”. In: Foundations of Computational Mathematics 16.6 (2016), pp. 1423–1472

    work page 2016

  2. [2]

    Counting Rooted Trees: The Universal Law t(n) ∼ Cρ −nn−3/2

    Jason P Bell, Stanley N Burris, and Karen A Yeats. “Counting Rooted Trees: The Universal Law t(n) ∼ Cρ −nn−3/2”. In: the electronic journal of combinatorics (2006), R63– R63

    work page 2006

  3. [3]

    A 3(2) pair of Runge-Kutta formulas

    Przemyslaw Bogacki and Lawrence F Shampine. “A 3(2) pair of Runge-Kutta formulas”. In: Applied Mathematics Letters 2.4 (1989), pp. 321–325

    work page 1989

  4. [4]

    Coefficients for the study of Runge-Kutta integration processes

    John C Butcher. “Coefficients for the study of Runge-Kutta integration processes”. In: Journal of the Australian Mathematical Society 3.2 (1963), pp. 185–201

    work page 1963

  5. [5]

    Implicit runge-kutta processes

    John C Butcher. “Implicit runge-kutta processes”. In: Mathematics of computation 18.85 (1964), pp. 50–64

    work page 1964

  6. [6]

    On Runge-Kutta processes of high order

    John C Butcher. “On Runge-Kutta processes of high order”. In: Journal of the Aus- tralian Mathematical Society 4.2 (1964), pp. 179–194

    work page 1964

  7. [7]

    A history of Runge-Kutta methods

    John Charles Butcher. “A history of Runge-Kutta methods”. In: Applied numerical mathematics 20.3 (1996), pp. 247–260

    work page 1996

  8. [8]

    Numerical methods for ordinary differential equations

    John Charles Butcher. Numerical methods for ordinary differential equations . John Wiley & Sons, 2016

    work page 2016

  9. [9]

    Tensor networks for dimensionality reduction and large-scale op- timization: Part 1 low-rank tensor decompositions

    Andrzej Cichocki et al. “Tensor networks for dimensionality reduction and large-scale op- timization: Part 1 low-rank tensor decompositions”. In: Foundations and Trends in Machine Learning 9.4–5 (2016), pp. 249–429

    work page 2016

  10. [10]

    Some Explicit Runge–Kutta Methods of High Order

    GJ Cooper and JH Verner. “Some Explicit Runge–Kutta Methods of High Order”. In: SIAM Journal on Numerical Analysis 9.3 (1972), pp. 389–405. REFERENCES 94

    work page 1972

  11. [11]

    A family of embedded Runge–Kutta formulae

    John R Dormand and Peter J Prince. “A family of embedded Runge–Kutta formulae”. In: Journal of Computational and Applied Mathematics 6.1 (1980), pp. 19–26

    work page 1980

  12. [12]

    Vandermonde systems on gauss–lobatto chebyshev nodes

    Alfredo Eisinberg and Giuseppe Fedele. “Vandermonde systems on gauss–lobatto chebyshev nodes”. In: Applied mathematics and computation 170.1 (2005), pp. 633–647

    work page 2005

  13. [13]

    Vandermonde systems on equidis- tant nodes in [0, 1]: accurate computation

    Alfredo Eisinberg, Giuseppe Fedele, and C Imbrogno. “Vandermonde systems on equidis- tant nodes in [0, 1]: accurate computation”. In: Applied mathematics and computation 172.2 (2006), pp. 971–984

    work page 2006

  14. [14]

    High-order explicit Runge-Kutta methods using m-symmetry

    Terry Feagin. “High-order explicit Runge-Kutta methods using m-symmetry”. In: Neural, Parallel & Scientific Computations 20.3-4 (2012), pp. 437–458

    work page 2012

  15. [15]

    Classical fifth-, sixth-, seventh-, and eighth-order Runge-Kutta formulas with stepsize control

    Erwin Fehlberg. Classical fifth-, sixth-, seventh-, and eighth-order Runge-Kutta formulas with stepsize control . Vol. 287. National Aeronautics and Space Administra- tion, 1968

    work page 1968

  16. [16]

    Low-order classical Runge–Kutta formulas with stepsize control and their application to some heat transfer problems

    Erwin Fehlberg. “Low-order classical Runge–Kutta formulas with stepsize control and their application to some heat transfer problems”. In: NASA TR R-315 (1969)

    work page 1969

  17. [17]

    Mathematical constants

    Steven R Finch. Mathematical constants. Cambridge university press, 2003, p. 296

    work page 2003

  18. [18]

    The condition of Vandermonde-like matrices involving orthogonal poly- nomials

    Walter Gautschi. “The condition of Vandermonde-like matrices involving orthogonal poly- nomials”. In: Linear algebra and its applications 52 (1983), pp. 293–300

    work page 1983

  19. [19]

    On extrapolation algorithms for ordinary initial value problems

    William B Gragg. “On extrapolation algorithms for ordinary initial value problems”. In: Journal of the Society for Industrial and Applied Mathematics, Series B: Nu- merical Analysis 2.3 (1965), pp. 384–403

    work page 1965

  20. [20]

    Geometric Numerical Inte- gration: Structure-Preserving Algorithms for Ordinary Differential Equations

    Ernst Hairer, Gerhard Wanner, and Christian Lubich. Geometric Numerical Inte- gration: Structure-Preserving Algorithms for Ordinary Differential Equations . Springer, 2006

    work page 2006

  21. [21]

    Solving ordinary differential equations I: Nonstiff problems

    Ernst Hairer, Gerhard Wanner, and Syvert P Nørsett. Solving ordinary differential equations I: Nonstiff problems . Springer, 1993

    work page 1993

  22. [22]

    Do numerical orbits of chaotic dy- namical processes represent true orbits?

    Stephen M Hammel, James A Yorke, and Celso Grebogi. “Do numerical orbits of chaotic dy- namical processes represent true orbits?” In: Journal of Complexity 3.2 (1987), pp. 136– 145

    work page 1987

  23. [23]

    The CMA Evolution Strategy: A Tutorial

    Nikolaus Hansen. “The CMA evolution strategy: A tutorial”. In: arXiv preprint arXiv:1604.00772 (2016)

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  24. [24]

    Twenty-step algorithm for deter- mining the asymptotic number of trees of various speces

    Frank Harary, Robert W Robinson, and Allen J Schwenk. “Twenty-step algorithm for deter- mining the asymptotic number of trees of various speces”. In: Journal of the Australian Mathematical Society 20.4 (1975), pp. 483–503

    work page 1975

  25. [25]

    qdwithclustersminimalv1-rk-tableaux

    Junyuan He. qdwithclustersminimalv1-rk-tableaux. https://github.com/JunyuanHe/ qdwithclustersminimalv1-rk-tableaux . Accessed: 2026-04-09. 2026. REFERENCES 95

    work page 2026

  26. [26]

    Derivation of Runge--Kutta Order Conditions via Functional Tree Tensor Networks

    Junyuan He, Zhonghao Sun, and Jizu Huang. “Derivatives of tree tensor networks and its applications in Runge–Kutta methods”. In: arXiv preprint arXiv:2504.15516 (2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  27. [27]

    Fundamental tensor operations for large-scale data analysis using tensor network formats

    Namgil Lee and Andrzej Cichocki. “Fundamental tensor operations for large-scale data analysis using tensor network formats”. In: Multidimensional Systems and Signal Processing 29.3 (2018), pp. 921–960

    work page 2018

  28. [28]

    Deterministic Nonperiodic Flow

    Edward N Lorenz. “Deterministic Nonperiodic Flow.” In: Journal of the Atmospheric Sciences 20.2 (1963), pp. 130–148

    work page 1963

  29. [29]

    Minkowski operations and vector spaces

    Juliette Mattioli. “Minkowski operations and vector spaces”. In: Set-valued analysis 3.1 (1995), pp. 33–50

    work page 1995

  30. [30]

    On the 25 stage 12th order explicit Runge-Kutta method

    Hiroshi Ono. “On the 25 stage 12th order explicit Runge-Kutta method”. In: Transactions- Japan Society for Industrial and Applied Mathematics 16.3 (2006), p. 177

    work page 2006

  31. [31]

    The number of trees

    Richard Otter. “The number of trees”. In: Annals of Mathematics 49.3 (1948), pp. 583– 599

    work page 1948

  32. [32]

    High order embedded Runge–Kutta formulae

    Peter J Prince and John R Dormand. “High order embedded Runge–Kutta formulae”. In: Journal of Computational and Applied Mathematics 7.1 (1981), pp. 67–75

    work page 1981

  33. [33]

    How long do numerical chaotic solutions remain valid?

    Tim Sauer, Celso Grebogi, and James A Yorke. “How long do numerical chaotic solutions remain valid?” In: Physical Review Letters 79.1 (1997), p. 59

    work page 1997

  34. [34]

    The MATLAB ODE suite

    Lawrence F Shampine and Mark W Reichelt. “The MATLAB ODE suite”. In: SIAM Journal on Scientific Computing 18.1 (1997), pp. 1–22

    work page 1997

  35. [35]

    Completely imbedded Runge–Kutta pairs

    PW Sharp and JH Verner. “Completely imbedded Runge–Kutta pairs”. In: SIAM journal on numerical analysis 31.4 (1994), pp. 1169–1190

    work page 1994

  36. [36]

    On Runge-Kutta methods of order 10

    Misha Stepanov. “On Runge-Kutta methods of order 10”. In: arXiv preprint arXiv:2504.17329 (2025)

    work page arXiv 2025

  37. [37]

    Explicit Runge–Kutta pairs with lower stage-order

    James H Verner. “Explicit Runge–Kutta pairs with lower stage-order”. In: Numerical Algorithms 65.3 (2014), pp. 555–577

    work page 2014

  38. [38]

    Strategies for deriving new explicit Runge-Kutta pairs

    JH Verner. “Strategies for deriving new explicit Runge-Kutta pairs”. In: Ann. Numer. Math 1 (1994), pp. 225–244

    work page 1994