pith. sign in

arxiv: 2512.23002 · v2 · submitted 2025-12-28 · 🌊 nlin.CD · math-ph· math.DS· math.MP

On the efficient numerical computation of covariant Lyapunov vectors

Jean-Jacq du Plessis , Malcolm Hillebrand , Charalampos Skokos This is my paper

Pith reviewed 2026-05-16 20:07 UTC · model grok-4.3

classification 🌊 nlin.CD math-phmath.DSmath.MP
keywords covariant Lyapunov vectorsCLVHamiltonian systemsnumerical computationcenter subspaceGinelli algorithmtangent dynamicsaccuracy
0
0 comments X

The pith

Adapting the CLV algorithm to prevent alignment in the center subspace improves long-time computation accuracy in Hamiltonian systems.

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

The paper examines how to choose the right durations for the transient phases in the standard algorithm for computing covariant Lyapunov vectors in chaotic conservative systems. It identifies that long backward integration periods cause the vectors in the two-dimensional center subspace to align or anti-align, which degrades accuracy. The authors test two ways to decide when to stop these phases on the Hénon-Heiles system and a three-oscillator chain, and propose a simple change to the procedure that stops this alignment and restores accuracy even for extended runs.

Core claim

In the tangent dynamics of the center subspace, CLVs tend to align or anti-align during long backward evolutions, reducing the accuracy of two-dimensional center subspace computations. An adaptation of the Ginelli algorithm that prevents this alignment/anti-alignment improves the accuracy of such computations over long times.

What carries the argument

The proposed adaptation to the Ginelli et al. algorithm that monitors and prevents alignment/anti-alignment of CLVs in the center subspace during backward evolution stages.

If this is right

  • Two termination methods give similar results, so the more efficient one is recommended.
  • Accuracy reduction is significant for long backward intervals in center subspaces.
  • The alignment behavior is observed similarly in both tested systems.
  • The adaptation maintains accuracy without restricting to short time intervals.

Where Pith is reading between the lines

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

  • The method could extend to non-Hamiltonian chaotic systems where similar subspace alignments occur.
  • It may allow more efficient computation by permitting longer stable integration windows.
  • Applications in stability analysis of multi-body systems could benefit from sustained accuracy.

Load-bearing premise

The alignment or anti-alignment of CLVs in the center subspace is the primary cause of accuracy loss during long backward phases.

What would settle it

If applying the proposed adaptation to the algorithm on the tested systems shows no improvement in CLV accuracy for long backward intervals, or if alignment persists, the claim would be falsified.

Figures

Figures reproduced from arXiv: 2512.23002 by Charalampos Skokos, Jean-Jacq du Plessis, Malcolm Hillebrand.

Figure 1
Figure 1. Figure 1: FIG. 1. Diagram depicting the numerical integrations required [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. The time evolution of the distance ∆ (7) between esti [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. The backward time evolution of the distance ∆ between [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. The backward time evolution of the distance ∆ between [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Similar to Fig. 3, where panels (a) and (b) correspond [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6. Similar to Fig. 3, except the GC algorithm has been [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7. Similar to Fig. 6, except the figure is in log-log scale [PITH_FULL_IMAGE:figures/full_fig_p007_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: FIG. 8. Similar to Fig. 2, but for system (9) with initial con [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: FIG. 9. Similar to Fig. 3, but for system (9): the backward [PITH_FULL_IMAGE:figures/full_fig_p009_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: FIG. 10. Similar to Fig. 9, where panels (a) and (b) correspond [PITH_FULL_IMAGE:figures/full_fig_p009_10.png] view at source ↗
read the original abstract

Covariant Lyapunov vectors (CLVs) are useful in multiple applications, but the optimal time windows needed to accurately compute these vectors are yet unclear. To remedy this, we investigate two methods for determining when to safely terminate the forward and backward transient phases of the CLV computation algorithm by Ginelli et al.~\cite{GinelliEtAl2007} when applied to chaotic orbits of conservative Hamiltonian systems. We perform this investigation for two prototypical Hamiltonian systems, namely the well-known H\'enon-Heiles system of two degrees of freedom and a system of three nonlinearly coupled harmonic oscillators having three degrees of freedom, finding very similar results for the two methods and thus recommending the more efficient one. We find that the accuracy of two-dimensional center subspace computations is significantly reduced when the backward evolution stages of the algorithm are performed over long time intervals. We explain this observation by examining the tangent dynamics of the center subspace wherein CLVs tend to align/anti-align, and we propose an adaptation of the algorithm that improves the accuracy of such computations over long times by preventing this alignment/anti-alignment of CLVs in the center subspace.

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

2 major / 2 minor

Summary. The paper examines two methods for selecting termination times of the forward and backward transient phases in the Ginelli et al. (2007) algorithm for computing covariant Lyapunov vectors (CLVs) on chaotic orbits of conservative Hamiltonian systems. Numerical tests on the Hénon-Heiles system (2 dof) and a 3-oscillator chain (3 dof) yield consistent results, leading the authors to recommend the more efficient method. They observe that accuracy in the two-dimensional center subspace degrades for long backward integrations, attribute this to alignment/anti-alignment of CLVs in the tangent dynamics, and propose an adaptation that prevents such alignment to maintain accuracy over extended times.

Significance. If the adaptation is shown to be robust, the work would provide a practical improvement for reliable long-time CLV computations in Hamiltonian systems, which are central to studies of chaos in conservative dynamics. The mechanistic explanation of center-subspace degradation adds insight into tangent-space behavior. The significance is currently limited by the narrow scope of validation and absence of quantitative error analysis.

major comments (2)
  1. [§4] §4 (numerical experiments on Hénon-Heiles and 3-oscillator systems): the adaptation's accuracy improvement is demonstrated only for these two Hamiltonian flows; no tests on dissipative, non-Hamiltonian, or higher-dimensional systems are reported, leaving the central claim of improved long-time accuracy without evidence of broader applicability.
  2. [§4.2] §4.2 (tangent dynamics of the center subspace): the assertion that CLV alignment/anti-alignment is the dominant error mechanism is supported only by visual inspection of trajectories; no quantitative decomposition isolating this effect from integration errors or other sources is provided, weakening the justification for the adaptation.
minor comments (2)
  1. [§3] The efficiency comparison between the two termination methods lacks explicit runtime benchmarks or operation counts, making the recommendation for the 'more efficient' method qualitative rather than quantitative.
  2. [Figures 4-6] Figure captions for the alignment plots should explicitly state the integration time intervals and the precise definition of the alignment metric used.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading of the manuscript and the constructive comments. We address each major comment below.

read point-by-point responses

  1. Referee: [§4] §4 (numerical experiments on Hénon-Heiles and 3-oscillator systems): the adaptation's accuracy improvement is demonstrated only for these two Hamiltonian flows; no tests on dissipative, non-Hamiltonian, or higher-dimensional systems are reported, leaving the central claim of improved long-time accuracy without evidence of broader applicability.

    Authors: The manuscript is explicitly focused on conservative Hamiltonian systems, as stated in the title, abstract, and choice of test cases. The adaptation addresses alignment issues specific to the tangent dynamics in such systems. We do not claim general applicability beyond this class. We will revise the conclusions to explicitly note this scope limitation and suggest extensions to dissipative or non-Hamiltonian systems as future work. revision: partial

  2. Referee: [§4.2] §4.2 (tangent dynamics of the center subspace): the assertion that CLV alignment/anti-alignment is the dominant error mechanism is supported only by visual inspection of trajectories; no quantitative decomposition isolating this effect from integration errors or other sources is provided, weakening the justification for the adaptation.

    Authors: We agree that the current support relies on visual inspection of the tangent-space trajectories. To strengthen the justification, we will add quantitative measures in the revised §4.2, such as the time series of angles between the relevant CLVs and a comparison of error growth with and without the adaptation, to better isolate the alignment effect from numerical integration errors. revision: yes

Circularity Check

0 steps flagged

No significant circularity; claims rest on independent numerical experiments

full rationale

The paper's central claims and proposed adaptation are derived from direct numerical simulations on two standard benchmark Hamiltonian systems (Hénon-Heiles and three coupled oscillators). No equations or results are defined in terms of parameters fitted to the target data, no predictions reduce by construction to inputs, and no load-bearing self-citations or uniqueness theorems from the authors' prior work are invoked. The derivation chain consists of empirical observations of tangent dynamics followed by an algorithmic modification, which remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The work relies on the pre-existing Ginelli algorithm and standard properties of Hamiltonian flows; no new free parameters are introduced or fitted, and no new physical entities are postulated.

axioms (1)
  • domain assumption The systems possess chaotic orbits with well-defined tangent dynamics under the standard Hamiltonian flow.
    Invoked when applying the Ginelli procedure to the Hénon-Heiles and coupled-oscillator models.

pith-pipeline@v0.9.0 · 5509 in / 1284 out tokens · 32243 ms · 2026-05-16T20:07:26.665116+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

39 extracted references · 39 canonical work pages · 1 internal anchor

  1. [1]

    The existence of these LEs is typically guaranteed by Oseledets’ multiplicative ergodic theorem [2]

    INTRODUCTION Lyapunov exponents (LEs) describe the exponential growth rates of perturbations (the so-called deviation vec- tors) to a trajectory of a dynamical system in various di- rections of the tangent space. The existence of these LEs is typically guaranteed by Oseledets’ multiplicative ergodic theorem [2]. A so-called covariant Lyapunov vector (CLV)...

    work page

  2. [2]

    THEORETICAL ASPECTS AND NUMERICAL METHODS In this paper, we focus our attention on continuous time dynamical systems with time invertible dynamics in order to guarantee the existence of the so-called Oseledets split- ting (see Sect. 2.1). Considering such a dynamical system with anN-dimensional (N-D) phase space, we denote by xa point in this space. A so-...

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  3. [3]

    We first present our results for the H´ enon-Heiles system (8) and then for the system with three degrees of freedom (9)

    NUMERICAL RESULTS In our investigations, we examine several chaotic orbits in both systems, but show results for only one represen- tative orbit for each Hamiltonian model. We first present our results for the H´ enon-Heiles system (8) and then for the system with three degrees of freedom (9). For our nu- merical integrations, we implement the ABA864 symp...

    work page

  4. [4]

    SUMMARY AND CONCLUSIONS After briefly reviewing the relevant theory of CLVs and discussing their computation via the GC algorithm, we used the distance ∆ (7) to measure the convergence be- tween relevant subspaces in both the H´ enon-Heiles system (8) and a Hamiltonian system with three degrees of free- dom (9). We proposed two methods, a direct one and a...

    work page

  5. [5]

    Ginelli, P

    F. Ginelli, P. Poggi, A. Turchi, H. Chat´ e, R. Livi, and A. Politi, Phys. Rev. Lett.99, 130601 (2007)

    work page 2007

  6. [6]

    V. I. Oseledets, Trans. Mosc. Math. Soc.19, 197 (1968)

    work page 1968

  7. [7]

    C. L. Wolfe and R. M. Samelson, Tellus A59, 355 (2007)

    work page 2007

  8. [8]

    P. V. Kuptsov and U. Parlitz, J. Nonlinear Sci.22, 727 (2012)

    work page 2012

  9. [9]

    Yang and G

    H.-L. Yang and G. Radons, Phys. Rev. E82, 046204 (2010)

    work page 2010

  10. [10]

    Yang and G

    H.-L. Yang and G. Radons, inNonequilibrium Statistical Physics of Small Systems, Rev. Nonlinear Dyn. Complex., edited by R. Klages, W. Just, and C. Jarzynski (Wiley,

    work page

  11. [11]

    Yang and G

    H.-L. Yang and G. Radons, J. Phys. A46, 254015 (2013)

    work page 2013

  12. [12]

    Kamiyama, M

    K. Kamiyama, M. Komuro, T. Endo, and K. Aihara, Int. J. Bifurc. Chaos24, 1430034 (2014)

    work page 2014

  13. [13]

    J. S. Frederiksen, Entropy25, 244 (2023)

    work page 2023

  14. [14]

    Paz´ o, M

    D. Paz´ o, M. A. Rodr´ ıguez, and J. M. L´ opez, Tellus A62, 10 (2010)

    work page 2010

  15. [15]

    Herrera, D

    S. Herrera, D. Paz´ o, J. Fern´ andez, and M. A. Rodr´ ıguez, Tellus A63, 978 (2011)

    work page 2011

  16. [16]

    Quinn, D

    C. Quinn, D. Harries, and T. J. O’Kane, J. Atmos. Sci. 78, 1647 (2021)

    work page 2021

  17. [17]

    Ginelli, H

    F. Ginelli, H. Chat´ e, R. Livi, and A. Politi, J. Phys. A46, 254005 (2013)

    work page 2013

  18. [18]

    Noethen, Physica D396, 18 (2019)

    F. Noethen, Physica D396, 18 (2019)

    work page 2019

  19. [19]

    Noethen,Computing Covariant Lyapunov Vectors – A Convergence Analysis of Ginelli’s Algorithm, Ph.D

    F. Noethen,Computing Covariant Lyapunov Vectors – A Convergence Analysis of Ginelli’s Algorithm, Ph.D. thesis, Universit¨ at Hamburg (2019)

    work page 2019

  20. [20]

    Skokos, inDynamics of Small Solar System Bodies and Exoplanets, Lect

    Ch. Skokos, inDynamics of Small Solar System Bodies and Exoplanets, Lect. Notes Phys., Vol. 790, edited by J. J. Souchay and R. Dvorak (Springer, Berlin, Heidelberg,

    work page

  21. [21]

    Pikovsky and A

    A. Pikovsky and A. Politi,Lyapunov Exponents: A Tool to Explore Complex Dynamics(Cambridge University Press, Cambridge, 2016)

    work page 2016

  22. [22]

    Benettin, L

    G. Benettin, L. Galgani, A. Giorgilli, and J.-M. Strelcyn, Meccanica15, 21 (1980)

    work page 1980

  23. [23]

    Benettin, L

    G. Benettin, L. Galgani, A. Giorgilli, and J.-M. Strelcyn, Meccanica15, 9 (1980)

    work page 1980

  24. [24]

    G. H. Golub and C. F. Van Loan,Matrix Computations, 3rd ed., Johns Hopkins Studies in the Mathematical Sci- ences (Johns Hopkins University Press, Baltimore, 1996)

    work page 1996

  25. [25]

    Ye and L.-H

    K. Ye and L.-H. Lim, SIAM J. Matrix Anal. Appl.37, 1176 (2016)

    work page 2016

  26. [26]

    N. I. Akhiezer and I. M. Glazman,Theory of Linear Op- erators in Hilbert Space(Dover Publications, New York, 1993)

    work page 1993

  27. [27]

    Deza and M

    E. Deza and M. M. Deza,Encyclopedia of Distances (Springer, Berlin, Heidelberg, 2009)

    work page 2009

  28. [28]

    Bj¨ orck and G

    ˚A. Bj¨ orck and G. H. Golub, Math. Comput.27, 579 (1973)

    work page 1973

  29. [29]

    A. V. Knyazev and M. E. Argentati, SIAM J. Sci. Comput. 23, 2008 (2002)

    work page 2008

  30. [30]

    du Plessis,Investigating Hamiltonian Dynamics by the Method of Covariant Lyapunov Vectors, MSc Thesis, Uni- versity of Cape Town (2024)

    J.-J. du Plessis,Investigating Hamiltonian Dynamics by the Method of Covariant Lyapunov Vectors, MSc Thesis, Uni- versity of Cape Town (2024)

    work page 2024

  31. [31]

    Skokos, J

    Ch. Skokos, J. Phys. A34, 10029 (2001)

    work page 2001

  32. [32]

    Skokos, Ch

    Ch. Skokos, Ch. Antonopoulos, T. C. Bountis, and M. N. Vrahatis, Prog. Theor. Phys. Suppl.150, 439 (2003)

    work page 2003

  33. [33]

    Skokos, Ch

    Ch. Skokos, Ch. Antonopoulos, T. C. Bountis, and M. N. Vrahatis, J. Phys. A37, 6269 (2004)

    work page 2004

  34. [34]

    Skokos and T

    Ch. Skokos and T. Manos, inChaos Detection and Predictability, Lect. Notes Phys., Vol. 915, edited by Ch. Skokos, G. A. Gottwald, and J. Laskar (Springer, Berlin, Heidelberg, 2016) pp. 129–181

    work page 2016

  35. [35]

    H´ enon and C

    M. H´ enon and C. Heiles, Astron. J.69, 73 (1964)

    work page 1964

  36. [36]

    Contopoulos, L

    G. Contopoulos, L. Galgani, and A. Giorgilli, Phys. Rev. A18, 1183 (1978)

    work page 1978

  37. [37]

    Blanes, F

    S. Blanes, F. Casas, A. Farr´ es, J. Laskar, J. Makazaga, and A. Murua, Appl. Numer. Math.68, 58 (2013)

    work page 2013

  38. [38]

    Danieli, B

    C. Danieli, B. Many Manda, T. Mithun, and C. Skokos, Math. Eng.1, 447 (2019)

    work page 2019

  39. [39]

    Skokos, T

    Ch. Skokos, T. C. Bountis, and Ch. Antonopoulos, Physica D231, 30 (2007)

    work page 2007