Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

2N-storage Runge-Kutta methods: c-reflection symmetry and factorization of the Butcher tableau

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Every 2N-storage (low-memory) Runge-Kutta method of order at most four has a reflected partner of the same order, and the paper proves the pairing through a factorization of the Butcher tableau.

desk verdict This paper gives a real structural explanation for the observed pairing of 2N-storage Runge-Kutta methods via a new factorization and c-reflection symmetry, but the proof depends on an identity imported without proof from the author's companion paper. read the letter →

arxiv 2506.08444 v1 pith:TXBXGJUD submitted 2025-06-10 math.NA cs.NAhep-latphysics.comp-ph

Alexei Bazavov This is my paper
classification math.NAcs.NAhep-latphysics.comp-ph MSC 65L0665L05
keywords 2N-storageRunge-Kuttamethodslow-storageButchertableaufactorizationc-reflectionsymmetryorderconditionsfourth-orderstabilityconstraintsmethodpairs
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

Low-storage Runge-Kutta methods that keep only two memory registers per solution component are standard tools in large-scale time stepping. The paper proves that for such methods, when they are of global order four or less and their auxiliary parameters are well defined, every method has a mirror-image partner: reflect the node values $c_i$ around $1/2$ and reverse the auxiliary $d_i$, and the new tableau satisfies exactly the same order conditions. The proof runs through a previously unnoticed factorization of the augmented Butcher tableau, $A = F D$, which also explains why published fourth-order low-storage schemes so often appear in pairs. The paper adds new $(5,4)$, $(6,4)$, and $(8,4)$ examples, some with coefficients in radicals or rationals, and shows the symmetry stops at order five.

What carries the argument

The augmented Butcher tableau $A$ and the diagonal node matrix $C$, together with the d-form relation $a_{ij} = d_j(f_{ij} - \sum_{k=j+1}^{i-1} a_{ik})$, where $f_{ij}=c_i-c_j$ for $i>j$ and zero otherwise. This relation turns the tableau into a product $A = F D$: the matrix $F=[C,L]$ encodes node differences and $D$ encodes the $d_i$ in a lower-triangular pattern. The proof machinery is the interplay of four matrices: the antidiagonal transpose operator $T$, the projection and selector matrices $P$ and $Q$, and the inverse factor $G=D^{-1}$, with identities $DP=QD$ and $GCG^{-1}=C-A$ that let reflected products like $\tilde A \tilde C \tilde A \tilde C$ unwind into known order-condition traces.

What would settle it

Choose any fourth-order 2N-storage method with well-defined $d_i$, construct its c-reflected tableau from $\tilde c_i = 1 - c_{s+2-i}$ and $\tilde d_i = d_{s+2-i}$ using high-precision arithmetic, and evaluate the four fourth-order conditions; if any of them deviates from $1/4$, $1/8$, $1/12$, $1/24$, the theorem is false.

Watch

Extended reading notes

Core claim

The central discovery is that the augmented Butcher tableau $A$ of a 2N-storage Runge-Kutta method (the $(s+1)\times(s+1)$ matrix whose bottom row carries the weights and whose last node is $c_{s+1}=1$) factorizes as $A = F D$, where $F = [C,L]$ is the commutator of the node matrix $C$ with the lower-triangular matrix of ones $L$, and $D$ is a lower-triangular matrix built from the parameters $d_i = B_i/(c_{i+1}-c_i)$. From this factorization the paper proves the c-reflection theorem: if the method has global order $p \le 4$ and all $d_i$ are finite with distinct adjacent nodes, then the reflected method defined by $\tilde c_i = 1 - c_{s+2-i}$ and $\tilde d_i = d_{s+2-i}$ satisfies the same order conditions, with augmented tableau $\tilde A = T(G^{-1}AG)^T T$, where $T$ is the antidiagonal unit matrix and $G = D^{-1}$. The proof reduces each reflected order condition to a combination of the original ones via the identities $GCG^{-1} = C - A$ and $DP = QD$; it also shows quantities of the form $\mathrm{Tr}[P A^n C]$ are conserved, so stability-related constraints survive reflection. Numerical solution-branch construction for $(5,4)$ methods confirms the predicted pairing, and the single fifth-order condition that cannot be written in trace form breaks the symmetry.

Load-bearing premise

The result stands on the coefficient formula $a_{ij}=b_j(\sum_{k=1}^j b_k - c_j)[c_i-c_j-\sum_{k=j+1}^{i-1} a_{ik}]$ taken from the companion paper: if that formula does not hold for every 2N-storage method, the factorization and the reflection theorem inherit the error; the theorem also requires distinct adjacent nodes and finite $d_i$, so degenerate cases like the rational $(5,4)$ scheme are excluded.

Editorial extensions

If this is right

  • Any valid 2N-storage method of order at most four yields a second method of the same order for free: reflecting the nodes and reversing the $d_i$ requires no new root finding.
  • Stability-tuning constraints of the form $\mathrm{Tr}[P A^n C]$ are invariant under the transformation, so reflected partners inherit the stability-region properties a developer targeted.
  • The symmetry breaks precisely at order five, in the one order condition that cannot be cast as a trace of $A$ and $C$; hence order-five and higher methods are not generally mirrored.
  • Self-reflected methods exist for even stage counts: the paper constructs self-c-reflected $(6,4)$ and $(8,4)$ schemes, while no $(5,4)$ self-c-reflected method was found.
  • New explicit schemes, including $(5,4)$ methods with rational or radical coefficients and a rational-coefficient $(5,4)$ scheme, verify the predicted fourth-order scaling in numerical tests.

Reading between the lines

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

  • Since the factorization $A=FD$ does not itself use the order, the trace-form invariance might survive in restricted subclasses of higher-order methods that happen to preserve the non-trace fifth-order condition; constructing more order-five schemes would test this directly.
  • The sharp boundary at the non-trace condition suggests the matrix-trace encoding is the natural invariant algebra for 2N-storage methods, and other structural symmetries may be found by looking for conditions that resist this encoding.
  • The dramatic simplification obtained by setting all nontrivial $d_i=2$ indicates a promising ansatz for symbolic searches: many order conditions collapse, and the existence of self-reflected schemes at $(6,4)$ and $(8,4)$ suggests the ansatz will remain tractable at higher even stage counts.
  • Even though c-reflection preserves accuracy and stability constraints, the performance of reflected partners on the tested problems varies by problem; practitioners should regard the reflection as a cheap way to double candidates, not as a guarantee of equal error behavior.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. This paper studies Williamson-type 2N-storage Runge-Kutta methods. It introduces an augmented Butcher tableau of size (s+1)x(s+1), defines a 'd-form' via parameters d_i, proves a factorization A = F D, and shows that under a 'c-reflection' transformation (~c_i = 1 - c_{s+2-i}, ~d_i = d_{s+2-i}), any method satisfying the order conditions for global order p ≤ 4 (with all d_i finite and adjacent nodes distinct) is mapped to another 2N-storage method satisfying the same order conditions. The paper also constructs several new (5,4), (6,4), and (8,4) schemes, provides numerical experiments, and includes a Matlab script that checks the order conditions for several known and c-reflected methods.

Significance. If the central theorem is correct, the paper gives a structural explanation for the previously observed pairings of Carpenter-Kennedy methods and provides an essentially cost-free transformation for generating new methods. The factorization A = F D is novel and potentially useful for designing low-storage schemes. The paper includes concrete analytic and numerical examples, as well as a machine-checkable Matlab script, which strengthens its credibility. However, the central proof depends on an identity imported without proof from the author's companion paper [1], and one stated lemma is incorrect as written, so the standalone correctness of the paper is not fully established.

major comments (3)
  1. [Sec. 2, Eq. (11)] Equation (11) is taken from the companion paper [1] without proof, and all subsequent results — the d-form (Eq. (32)), the factorization (Eq. (44)), and Theorem 3 — rely on it. Since the manuscript is presented as a follow-up, please either prove Eq. (11) for general 2N-storage methods (or at least for the class with distinct adjacent nodes and finite d_i), or state it as an explicit assumption whose validity is verified independently. As written, a sign or indexing error in Eq. (11) would propagate through the entire derivation.
  2. [Sec. 4.2, Lemma 1] The proof of Lemma 1 is given as 'Continuation of this process' after checking the first two columns. This is a sketch rather than a complete induction. Because the factorization A = F D is load-bearing for all later results, please provide a formal inductive proof (or a precise algorithm with a proof) for the general s-stage case.
  3. [Sec. 5.1, Lemma 5] Lemma 5 is stated incorrectly. With the definition (M^τ)_{ij} = M_{s+2-j,s+2-i}, the correct identity is M^τ = T M^T T, not T M T^T. As written, the lemma is false for non-symmetric M, and the proof contains a corresponding index error: (T M T^T)_{ij} equals M_{s+2-i,s+2-j}, not M_{s+2-j,s+2-i}. The later Eq. (74) and the proof of Theorem 3 use the correct form T(G^{-1}AG)^T T, but the lemma itself should be corrected, and the derivation in Sec. 5.2 should be re-examined for consistency.
minor comments (5)
  1. [Abstract] The abstract states that '2N-storage methods of the order of global accuracy less than five possess a symmetry' without qualification, but Theorem 3 requires all d_i finite and adjacent c_i distinct. Please qualify the claim to match the theorem's hypotheses, since methods such as (5,4)_5 in Table 7 are excluded.
  2. [Sec. 2, Eq. (13)] Equation (13) contains a typographical error: 'Bi = bi iX k=1 bk − ci (ci+1 − ci), , (13)' should read 'B_i = b_i (sum_{k=1}^i b_k - c_i)/(c_{i+1}-c_i)', with a properly placed denominator.
  3. [Sec. 5.3, Eq. (79)] In the sentence following Eq. (79), 'C T = C' should be 'CT = C' (the diagonal matrix C commutes with T only in the sense used there); please clarify the notation to avoid confusion.
  4. [References] Reference [1] is an arXiv preprint (arXiv:2506.07359) and its status (submitted, under review, or published) is not indicated. Since Eq. (11) is taken from that preprint, please either include the proof or clearly state the reference's status so the reader can assess the dependency.
  5. [Sec. 6.1, Table 5 and Table 6] The decimal values in Tables 5 and 6 are given to only a few digits; for reproducibility, it would be helpful to state the number of digits used in the computations or provide the exact expressions in a supplementary file.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: Theorem 3 is a genuine algebraic derivation from stated assumptions, not a fitted or self-referential prediction.

full rationale

The central result, Theorem 3, is proved by explicit matrix manipulation of the augmented Butcher tableau, not by importing the conclusion. The c-reflected tableau is defined through Eq. (74), and each order condition is then verified individually using trace identities, Eqs. (76)-(90), that reduce the reflected expressions to the original order conditions. The factorization A = FD and the d-form are derived from Eq. (11), which is imported from the author's companion paper [1] without re-derivation in this manuscript. While this is a load-bearing self-citation and a verification gap, it is not circular reasoning: Eq. (11) is a general algebraic identity about 2N-storage coefficients, it is not the theorem being proved, it involves no fitted parameters or data, and it can be checked independently for any method. The transparent exclusion of special cases (adjacent equal nodes and vanishing d_i) in Theorem 3 is a stated domain condition, not a smuggled assumption. The numerical experiments and Matlab script are illustrative checks, and the proof does not depend on them. No step in the derivation reduces the conclusion to its own input by construction, so the paper is not circular.

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

The central derivation rests on linear algebra identities and on the standard order conditions. It inherits Eq. (11) from the author's previous paper without a proof here, and it assumes the method is in general position so that the d-form exists. No free parameters are fitted and no new entities are introduced.

assumptions (4)
  • domain assumption Equation (11) from Ref. [1]: aij = bj( sum_{k=1}^j b_k - c_j )[ c_i - c_j - sum_{k=j+1}^{i-1} a_ik ]
    Central starting point for the d-form and factorization; cited from the author's own companion paper without proof in this manuscript.
  • domain assumption The 2N-storage method is in general position: c_1=0, c_{s+1}=1, adjacent nodes distinct, and all denominators d_i = B_i/(c_{i+1}-c_i) finite and nonzero.
    Stated in Sec. 4.1; Theorem 3 applies only when the d-form exists. Special cases with repeated nodes are excluded.
  • standard math Standard order conditions for explicit Runge-Kutta methods up to order four (Eqs. A.1-A.8) are complete and correct.
    Used in Theorem 3 and in the numerical checks.
  • standard math The row-sum condition c_i = sum_j a_ij holds for all considered explicit Runge-Kutta methods.
    Assumed in Sec. 2 and used in the second simplifying rule around Eq. (77).

how reviews work

0 comments
Cite this review

Pith. "Pith review of 2N-storage Runge-Kutta methods: c-reflection symmetry and factorization of the Butcher tableau." pith.science (2026). https://pith.science/paper/TXBXGJUD

@misc{pith2026250608444,
  author       = {Pith},
  title        = {Pith review of: 2N-storage Runge-Kutta methods: c-reflection symmetry and factorization of the Butcher tableau},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TXBXGJUD}},
  note         = {Machine review of arXiv:2506.08444}
}
read the original abstract

Low-storage Runge-Kutta schemes of Williamson's type, so-called 2N-storage schemes, are further examined as a follow-up to the recent work. It is found that the augmented Butcher tableau factorizes into a product of matrices with special properties. Those properties reveal that the 2N-storage methods of the order of global accuracy less than five possess a symmetry, called c-reflection symmetry, i.e. most methods exist in pairs. A transformation that relates the Butcher tableaux of the pairs is found and the fact that the c-reflected method satisfies the same order conditions as the original one is proven. Numerical evidence that validates the analytic results is presented. Branches of solutions for (5,4) methods, first explored by Carpenter and Kennedy, are constructed numerically. Four new (5,4) schemes with coefficients expressed in radicals and one with rational coefficients are examined for illustration. Eight new (6,4) schemes, some of which can be expressed in rationals or radicals, and one (8,4) scheme, are studied to understand the practical implications of the c-reflection symmetry for methods with higher number of stages. In the absence of closed-form analytic solutions for 2N-storage Runge-Kutta methods of order four and above, the general symmetry properties, as well as some specific analytic solutions presented here, may help in development and optimization of 2N-storage schemes.

Figures

Figures reproduced from arXiv: 2506.08444 by the authors.

Figure 1
Figure 1. The three branches of solutions of Eq. ( [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. The distance from the exact solution d(h) for the three test problems (arranged horizontally) for the five fourth-order 2N-storage methods (arranged vertically), taken from literature and described in the text, shown in blue, and the corresponding c-reflected methods, shown in red. 20 [PITH_FULL_IMAGE:figures/full_fig_p020_2.png] view at source ↗
Figure 3
Figure 3. An analog of the Williamson curves for (3,3) methods, Fig. [PITH_FULL_IMAGE:figures/full_fig_p021_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Scaling of the new (5,4)i, i = 1, . . . 5 methods, shown in color, and the four (5,4) methods of Ref. [3], shown in gray, for test problems 1 (left) and 2 (right). By exploring special values of the coefficients ci , i.e., 0, 1/2, 1, one (5,4) 2N-storage scheme with ra…
Figure 5
Figure 5. Figure 5: Scaling of the new (5,4)i, i = 1, . . . 5 methods, shown in color, and the four (5,4) methods of Ref. [3], shown in gray, for test problem 3 (left). Stability regions for all (5,4) methods considered (right). In general, it appears that the new tools developed in Ref. …
Figure 6
Figure 6. Figure 6: Scaling of the new (6,4)i, i = 1, . . . 8 methods, shown in color, and the HALERK64 method of Ref. [11], shown in black and the RK46-NL method of Ref. [10], shown in gray, for test problems 1 (left) and 2 (right). methods: for a general method of order of accuracy p ⩽ …
Figure 7
Figure 7. Figure 7: Scaling of the new (6,4)i, i = 1, . . . 8 methods, shown in color, and the HALERK64 method of Ref. [11], shown in black and the RK46-NL method of Ref. [10], shown in gray, for test problem 3 (left). Stability regions for all (6,4) methods considered (right). The stabil…

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

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

  1. CosmoLattice 2.0

    astro-ph.CO 2026-07 conditional novelty 6.0 of 10

    CosmoLattice v2.0 extends lattice cosmology simulations with non-minimal scalars, ALP–gauge couplings, defect networks, low-storage RK integrators, optimized GWs, and O(10) GPU speedups.

Reference graph

Works this paper leans on

17 extracted references · 11 canonical work pages · cited by 1 Pith paper

  1. [1]

    Bazavov, 2N-storage Runge-Kutta methods: Order conditions, general properties and some analytic solutions (2025)

    A. Bazavov, 2N-storage Runge-Kutta methods: Order conditions, general properties and some analytic solutions (2025). arXiv:2506.07359. 39 URL https://arxiv.org/abs/2506.07359

    arXiv 2025

  2. [3]

    Carpenter, C

    M. Carpenter, C. Kennedy, Fourth-order 2N-storage Runge-Kutta schemes, Tech. Rep. NASA- TM-109112, NASA (1994)

    work page 1994

  3. [4]

    Hairer, S

    E. Hairer, S. Nørsett, G. Wanner, Solving Ordinary Differential Equations I Nonstiff problems, 2nd Edition, Springer, Berlin, 2000

    work page 2000

  4. [5]

    Butcher, Numerical Methods for Ordinary Differential Equations, 3rd Edition, Wiley, 2016

    J. Butcher, Numerical Methods for Ordinary Differential Equations, 3rd Edition, Wiley, 2016

    work page 2016

  5. [6]

    Bazavov, Commutator-free Lie group methods with minimum storage requirements and reuse of exponentials, BIT Numerical Mathematics 62 (3) (2021) 745–771

    A. Bazavov, Commutator-free Lie group methods with minimum storage requirements and reuse of exponentials, BIT Numerical Mathematics 62 (3) (2021) 745–771. doi:10.1007/ s10543-021-00892-x . URL http://dx.doi.org/10.1007/s10543-021-00892-x

    doi:10.1007/s10543-021-00892-x 2021

  6. [7]

    C.-W. Shu, S. Osher, Efficient implementation of essentially non-oscillatory shock-capturing schemes, Journal of Computational Physics 77 (2) (1988) 439–471. doi:https://doi.org/10. 1016/0021-9991(88)90177-5. URL https://www.sciencedirect.com/science/article/pii/0021999188901775

    arXiv 1988

  7. [8]

    an Yan, Low-storage Runge-Kutta method for simulating time-dependent quantum dynamics, Chinese Journal of Chemical Physics 30 (3) (2017) 277 – 286

    Y. an Yan, Low-storage Runge-Kutta method for simulating time-dependent quantum dynamics, Chinese Journal of Chemical Physics 30 (3) (2017) 277 – 286

    work page 2017

  8. [9]

    Berland, C

    J. Berland, C. Bogey, C. Bailly, Optimized Explicit Schemes: Matching and Boundary Schemes, and 4th-order Runge-Kutta Algorithm, 2012. arXiv:https://arc.aiaa.org/doi/pdf/10. 2514/6.2004-2814, doi:10.2514/6.2004-2814. URL https://arc.aiaa.org/doi/abs/10.2514/6.2004-2814

    work page doi:10.2514/6.2004-2814 2012

Show all 17 references
  1. [10]

    Berland, C

    J. Berland, C. Bogey, C. Bailly, Low-dissipation and low-dispersion fourth-order Runge-Kutta algorithm, Computers and Fluids 35 (10) (2006) 1459 – 1463. doi:https://doi.org/10.1016/ j.compfluid.2005.04.003. URL http://www.sciencedirect.com/science/article/pii/S0045793005000575

    2006

  2. [11]

    Allampalli, R

    V. Allampalli, R. Hixon, M. Nallasamy, S. D. Sawyer, High-accuracy large-step explicit Runge- Kutta (HALE-RK) schemes for computational aeroacoustics, Journal of Computational Physics 228 (10) (2009) 3837 – 3850. doi:https://doi.org/10.1016/j.jcp.2009.02.015. URL http://www.sc...

    2009 doi

  3. [12]

    Toulorge, W

    T. Toulorge, W. Desmet, Optimal Runge-Kutta schemes for discontinuous Galerkin space dis- cretizations applied to wave propagation problems, Journal of Computational Physics 231 (4) (2012) 2067 – 2091. doi:https://doi.org/10.1016/j.jcp.2011.11.024. URL http://www.sciencedirect...

    2012 doi

  4. [13]

    Niegemann, R

    J. Niegemann, R. Diehl, K. Busch, Efficient low-storage Runge-Kutta schemes with optimized stability regions, Journal of Computational Physics 231 (2) (2012) 364 – 372. doi:https://doi. org/10.1016/j.jcp.2011.09.003. URL http://www.sciencedirect.com/science/article/pii/S002199...

    2012 doi

  5. [14]

    W. H. Press, W. T. Vetterling, S. A. Teukolsky, B. P. Flannery, Numerical Recipes in C++: the art of scientific computing, 2nd Edition, Cambridge University Press, USA, 2001. 40

    2001

  6. [15]

    D. J. Bates, J. D. Hauenstein, A. J. Sommese, C. W. Wampler, Bertini: Soft- ware for numerical algebraic geometry, Available at bertini.nd.edu with permanent doi: dx.doi.org/10.7274/R0H41PB5

    doi

  7. [16]

    Fousse, G

    L. Fousse, G. Hanrot, V. Lef` evre, P. P´ elissier, P. Zimmermann, Mpfr: A multiple-precision binary floating-point library with correct rounding, ACM Trans. Math. Softw. 33 (2) (2007) 13–es. doi:10.1145/1236463.1236468. URL https://doi.org/10.1145/1236463.1236468

    2007

  8. [17]

    Decker, G.-M

    W. Decker, G.-M. Greuel, G. Pfister, H. Sch¨ onemann, Singular 4-3-2 — A computer algebra system for polynomial computations, http://www.singular.uni-kl.de (2023)

    2023

  9. [18]

    W. R. Inc., Mathematica, Version 14.2, champaign, IL, 2024. URL https://www.wolfram.com/mathematica 41

    2024

Pith tools

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