Trajectory arclength reveals chaos

arXiv:2602.07660 · nlin.CD, math.DS · Submitted 2026-02-07 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Nonlinear Sciences. The day in nonlinear science: chaos, patterns, networks and adaptation.

Kai: We're looking at the paper: "Trajectory arclength reveals chaos".

Mira: Here's the gist of the paper. In this paper, they investigate how the time-averaged phase space arclength of a trajectory can be used as a diagnostic to distinguish between regular and chaotic dynamics in Hamiltonian systems.

Kai: The proposed indicator, denoted as omega, is shown to provide different finite-time behaviors for regular and chaotic trajectories despite both converging in the long term (Page 1).

Mira: Indicator Definition and Theoretical Basis The indicator omega is computed by integrating the equations of motion together with a single scalar arclength-accumulation equation, which means it does not require variational equations, neighboring trajectories, or spatial derivatives of a Lagrangian descriptor field (Page 1).

Kai: The approach is based on Birkhoff’s ergodic theorem and the finite-time behavior of the time-averaged trajectory arclength omega (Page 1).

Paper summary: Kai: To recap, we're talking about the paper "Trajectory arclength reveals chaos" and how they use the time averaged phase space arclength of a trajectory to distinguish regular and chaotic motion in Hamiltonian systems.

Mira: The core thesis is that this indicator, omega, provides different finite-time behaviors for regular versus chaotic trajectories, even though both converge in the long run >

Lev: They define omega by integrating the equations of motion with a single scalar arclength accumulation equation, which cleverly avoids needing variational equations or tracking neighboring trajectories >

Kai: That's what makes it attractive computationally. It’s just one extra scalar equation to integrate alongside the dynamics, not a whole new set of complex descriptors >

Mira: They use Birkhoff’s ergodic theorem and look at how this time averaged arclength behaves over finite time, which is the basis for their classification >

Lev: For regular motion, they find that omega t approaches a limit Omega infinity as tau goes to infinity uniformly across all initial phases theta zero >

Kai: The limiting value of that average arclength is given by Omega infinity equals one over two pi times dZ T d g of theta d theta, where g of theta is the norm of the vector field >

Mira: For chaotic motion, they show that the root mean square fluctuation around this limit has a scaling of sigma squared omega tau minus one over two as tau goes to infinity for trajectories distributed according to an ergodic measure >

Lev: This scaling implies that for chaotic systems, you see fluctuations of order sigma omega tau minus one over two as time goes on, which is characteristic of the mixing regime >

Kai: The key distinction they highlight is in those finite time scalings: regular trajectories show deviations decaying as O of tau minus one, while chaotic dynamics show fluctuations scaling at the root mean square rate of tau minus one over two >

Mira: They further explore this using block-scaling analysis to make that distinction clearer between periodic motion and chaos >

Lev: And they tested it on the Hénon–Heiles system and the Fermi–Pasta–Ulam lattice, comparing their results against SALI as a ground-truth indicator of chaos >

Kai: The numerical results showed high accuracy, between ninety-nine point four eight eight percent and ninety-nine point nine nine six percent for the Hénon-Heiles system >

Mira: So they conclude that omega is a useful, complementary single trajectory diagnostic for chaos detection and a practical source of scalar features for analyzing trajectory ensembles >

Conclusion: Kai: The paper "Trajectory arclength reveals chaos" investigates how the time averaged phase space arclength of a trajectory can be used to distinguish between regular and chaotic dynamics in Hamiltonian systems.

Mira: They propose omega as a diagnostic indicator, showing that regular trajectories have deviations bounded by O of tau minus one, while chaotic motion exhibits fluctuations scaling at the root mean square rate of tau minus one over two >

Lev: It boils down to using the different finite time scalings to separate these regimes, which is what gives this method its practical utility in distinguishing between the two types of dynamics >

Kai: The authors show this works on benchmark systems like Hénon–Heiles and FPU lattice, where they achieved accuracies up to ninety-nine point nine nine six percent when compared against SALI >

Mira: It suggests that the idea isn't just a fancy new indicator, but a way to get a simple physical measure—the trajectory arclength—that reveals chaos >

Lev: The method is computationally simple because it only needs one additional scalar equation after integration, which means it's viable for analyzing large ensembles of trajectories where you need efficiency >

Kai: So, the main implication is that this approach gives us a simple way to classify dynamics based on a measurable property of the trajectory path itself >

Vrije Universiteit Brussel · Departamento de Fnica y Matem꧊s, Facultad de Ciencias, Universidad de Alcalá

nlin.CD, math.DS

Submitted: 2026-02-07

Updated: 2026-11-05

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 76/100

The gist: In this paper, they investigate how the time-averaged phase space arclength of a trajectory can be used as a diagnostic to distinguish between regular and chaotic dynamics in Hamiltonian systems.

Key concepts

Indicator $\Omega$
A diagnostic tool calculated by integrating the equations of motion along with a single scalar equation that accumulates the phase space arclength. It is simple because it avoids needing neighboring trajectories or complex Lagrangian field derivatives, making it useful for high-dimensional systems.
Regular Motion Asymptotics
For trajectories on smooth invariant tori (regular motion), the time-averaged arclength $\Omega(\tau)$ approaches a limit $\Omega\infty$ with deviations that decay proportionally to $O(\tau^{-1})$ as time $\tau$ increases. This predictable, slow decay is characteristic of non-chaotic behavior.
Chaotic Motion Asymptotics
For trajectories in an ergodic chaotic component, the root-mean-square fluctuation of $\Omega(\tau)$ scales with time as $O(\tau^{-1/2})$. This specific scaling is a signature of the mixing chaotic regime and helps identify chaotic dynamics through finite-time analysis.
Finite-Time Classification
The key to distinguishing regimes lies in the scaling of fluctuations: $O(\tau^{-1})$ for regular motion versus $O(\tau^{-1/2})$ for chaotic motion. This difference in how the indicator fluctuates over finite time allows researchers to classify trajectories as either regular or chaotic.

Terminology

Summary

In this paper, they investigate how the time-averaged phase space arclength of a trajectory can be used as a diagnostic to distinguish between regular and chaotic dynamics in Hamiltonian systems. p. 0

Where the paper says it

García-Garrido†2 1Vrije Universiteit Brussel, Brussels, Belgium 2Departamento de Física y Matemáticas, Facultad de Ciencias, Universidad de Alcalá, 28805 Alcalá de Henares, Madrid, Spain. October 20, 2026 Abstract In this paper, we investigate the time-averaged phase space arclength of a trajectory, expressed through a normalized Lagrangian descriptor, as a trajectory-based diagnostic for distinguishing regular and chaotic dynamics.

The proposed indicator, denoted as omega, is shown to provide different finite-time behaviors for regular and chaotic trajectories despite both converging in the long term (Page 1). p. 0
Where the paper says it

The proposed indicator, Ω, is computed by integrating the equations of motion together with a single scalar arclength-accumulation equation and therefore does not require variational equations, neighboring trajectories, or spatial derivatives of a Lagrangian descriptor field. Using Birkhoff’s ergodic theorem together with estimates for quasiperiodic motion and correlation-based arguments for chaotic ergodic dynamics, we show that regular and chaotic trajectories exhibit different finite-time behavior despite the convergence of Ω in both regimes.

Indicator Definition and Theoretical Basis

The indicator omega is computed by integrating the equations of motion together with a single scalar arclength-accumulation equation, which means it does not require variational equations, neighboring trajectories, or spatial derivatives of a Lagrangian descriptor field (Page 1). The approach is based on Birkhoff’s ergodic theorem and the finite-time behavior of the time-averaged trajectory arclength omega (Page 1). p. 1

Where the paper says it

As a result, it reduces both the mathematical complexity and the computational cost of chaos detection, making it one of the simplest geometrical chaos indicators that can be derived from LDs. Our approach is based on Birkhoff’s ergodic theorem [69, 48] and on the finite-time behavior of the time-averaged trajectory arclength, Ω.

p. 1
Where the paper says it

As a result, it reduces both the mathematical complexity and the computational cost of chaos detection, making it one of the simplest geometrical chaos indicators that can be derived from LDs. Our approach is based on Birkhoff’s ergodic theorem [69, 48] and on the finite-time behavior of the time-averaged trajectory arclength, Ω.

Asymptotic Behavior for Regular Motion

For quasiperiodic trajectories evolving on a smooth invariant torus, the time-averaged arclength satisfies omega(τ) = omega∞ + O(τ−1) as τ → +∞ uniformly with respect to the initial phase θ0 (Page 5). The limiting value is given by omega∞ = 1/(2π)dZ Td g(θ) dθ > 0, where g(θ) is the norm of the vector field (Page 5). p. 2

Where the paper says it

(4) The integrand F used to define LDs typically quantifies some geometric or physical property of the trajectory. In this work, we adopt the arclength formulation, in which F is given by the Euclidean norm of the vector field: F(x, t) = ∥x˙ ∥ = ∥∇H(x, t)∥.

Asymptotic Behavior for Chaotic Motion

For trajectories with initial conditions distributed according to the invariant measure of an ergodic chaotic component with absolutely integrable autocorrelation and strictly positive asymptotic variance, the root-mean-square fluctuation satisfies Eµomega(τ) − omega∞ squared = σ2omega τ−1/2 (Page 8). This implies that Stdµ[omega(τ)] ∼ σomega τ−1/2, which is characteristic of the mixing chaotic regime (Page 8). p. 6

Where the paper says it

Rather than the deterministic O(τ −1 ) bound derived above, the deviations of Ω(τ ) from its limiting value exhibit a slower statistical decay, with a rootmean-square magnitude of order τ −1/2 . Such fluctuations are characteristic of the mixing chaotic regime considered here [36, 53, 52].

Finite-Time Classification and Validation

The distinction between the two regimes is made through the different asymptotic scalings of finite-time fluctuations: quasiperiodic trajectories exhibit deviations decaying as O(τ−1), whereas sufficiently mixing chaotic dynamics exhibit root-mean-square fluctuations of order τ−1/2 (Page 11). This is further explored using block-scaling analysis, where the scaling of σblock(∆) provides a distinction between periodic/quasiperiodic motion and chaotic dynamics (Page 13). p. 0

Where the paper says it

For quasiperiodic trajectories with Diophantine frequency vectors, deviations from the limiting average decay as O(τ −1 ), whereas sufficiently mixing chaotic dynamics exhibit root-mean-square fluctuations of order τ −1/2 under suitable correlation assumptions. This difference in finite-time convergence and fluctuation scaling provides the theoretical basis for using Ω to discriminate between the two dynamical regimes.

Numerical Results and Performance

The proposed indicator was validated in the Hénon–Heiles and Fermi–Pasta–Ulam systems, where the resulting classifications were compared with SALI as a reference chaos indicator (Page 15). The results support omega as a complementary single-trajectory diagnostic for chaos detection and as a practical source of scalar features for the analysis of trajectory ensembles (Page 15). p. 0

Where the paper says it

We validate the proposed approach in the Hénon–Heiles and Fermi–Pasta– Ulam systems, comparing the resulting classification of initial conditions with SALI as a reference chaos indicator. The results support Ω as a complementary single-trajectory diagnostic for chaos detection and as a practical source of scalar features for the analysis of trajectory ensembles, particularly in settings involving large numbers of trajectories or high-dimensional systems.

The classification achieved accuracies between 99. p. 1
Where the paper says it

Given this broad relevance, the accurate identification of regular and chaotic regimes has become a key requirement in modern computational applications, including the training of symplectic Machine Learning models for Hamiltonian dynamics [24, 5] and the development of classifiers capable of efficiently characterizing large ensembles of initial conditions [31]. The quantitative distinction between regular and chaotic motion in phase space is commonly achieved through chaos indicators, that is, functions designed to exhibit markedly different asymptotic behavior for regular and chaotic trajectories.

488 % and 99.996 % across all four ensembles for the Hénon-Heiles system (Page 21). p. 0
Where the paper says it

We validate the proposed approach in the Hénon–Heiles and Fermi–Pasta– Ulam systems, comparing the resulting classification of initial conditions with SALI as a reference chaos indicator. The results support Ω as a complementary single-trajectory diagnostic for chaos detection and as a practical source of scalar features for the analysis of trajectory ensembles, particularly in settings involving large numbers of trajectories or high-dimensional systems.

Conclusion and Future Directions

The different finite-time scaling of omega(τ) provides a mechanism for distinguishing regular and chaotic trajectories, with quasiperiodic motion exhibiting deviations bounded by O(τ−1) and chaotic motion exhibiting root-mean-square fluctuations of order τ−1/2 (Page 11). The method is simple computationally, requiring only one additional scalar arclength variable after integration, making it attractive for high-dimensional systems (Page 15). p. 0

Where the paper says it

We validate the proposed approach in the Hénon–Heiles and Fermi–Pasta– Ulam systems, comparing the resulting classification of initial conditions with SALI as a reference chaos indicator. The results support Ω as a complementary single-trajectory diagnostic for chaos detection and as a practical source of scalar features for the analysis of trajectory ensembles, particularly in settings involving large numbers of trajectories or high-dimensional systems.

Future work could extend this method to systems exhibiting slow transport or stickiness (Page 4). p. 2
Where the paper says it

1 Lagrangian descriptors Originally developed as a mathematical framework to analyze transport and mixing processes in geophysical flows [40, 47], Lagrangian descriptors have emerged as a powerful, trajectory-based scalar diagnostic tool for revealing phase space structures in general dynamical systems [77]. In this work we will focus on applying this method in the context of continuous Hamiltonian systems, although this approach can be used similarly for discrete systems (maps) [37].

The scalar features derived from the omega time series also provide a natural basis for constructing large datasets of regular and chaotic trajectories for machine-learning approaches (Page 15). p. 1
Where the paper says it

Given this broad relevance, the accurate identification of regular and chaotic regimes has become a key requirement in modern computational applications, including the training of symplectic Machine Learning models for Hamiltonian dynamics [24, 5] and the development of classifiers capable of efficiently characterizing large ensembles of initial conditions [31]. The quantitative distinction between regular and chaotic motion in phase space is commonly achieved through chaos indicators, that is, functions designed to exhibit markedly different asymptotic behavior for regular and chaotic trajectories.

Pruning redundant reasoning tokens

The indicator is obtained by integrating the equations of motion together with a single scalar equation that accumulates the phase space arclength (Page 1). The approach is based on Birkhoff’s ergodic theorem and the finite-time behavior of the time-averaged trajectory arclength omega (Page 1).

How the training signal is built

The indicator requires only one additional scalar equation to be integrated together with the equations of motion, avoiding neighbor trajectories or deviation vectors (Page 1). The analysis exploits two complementary procedures based on the finite-time evolution of omega: block-scaling analysis and residual fluctuation analysis (Page 13). p. 0

Where the paper says it

For quasiperiodic trajectories with Diophantine frequency vectors, deviations from the limiting average decay as O(τ −1 ), whereas sufficiently mixing chaotic dynamics exhibit root-mean-square fluctuations of order τ −1/2 under suitable correlation assumptions. This difference in finite-time convergence and fluctuation scaling provides the theoretical basis for using Ω to discriminate between the two dynamical regimes.

Where the gains come from

The different finite-time scaling of these deviations provides the theoretical basis for using omega as a practical chaos indicator (Page 11). The numerical experiments show that these quantities provide an effective separation between regular and chaotic trajectories over the phase space regions investigated (Page 3). p. 6

Where the paper says it

In the present work, however, we retain the standard unweighted time average because the finite-time convergence and fluctuation properties of Ω(τ ) themselves provide the information used to distinguish regular from chaotic motion. 2 Time-averaged arclength in ergodic chaotic regions In what follows, we consider trajectories evolving in an invariant region that is ergodic with respect to the microcanonical measure.

The simple computational structure is a major advantage when large ensembles of trajectories must be analyzed or when the phase space dimension becomes large (Page 15). p. 1
Where the paper says it

Given this broad relevance, the accurate identification of regular and chaotic regimes has become a key requirement in modern computational applications, including the training of symplectic Machine Learning models for Hamiltonian dynamics [24, 5] and the development of classifiers capable of efficiently characterizing large ensembles of initial conditions [31]. The quantitative distinction between regular and chaotic motion in phase space is commonly achieved through chaos indicators, that is, functions designed to exhibit markedly different asymptotic behavior for regular and chaotic trajectories.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Develop a single trajectory-based diagnostic for distinguishing regular and chaotic dynamics by computing the time-averaged phase space arclength, Omega. This indicator requires only "integrating the equations of motion together with a single scalar arclength-accumulation equation and therefore does not require variational equations, neighboring trajectories, or spatial derivatives of a Lagrangian descriptor field."

  2. Implement a classification strategy based on the finite-time fluctuation scaling of Omega to differentiate dynamical regimes. Specifically, use block-scaling analysis where for sufficiently mixing chaotic dynamics with an integrable correlation function, the central limit theorem implies Var h Omega (Delta) k i ∼ Delta−1 =⇒ σblock(Delta) ∼ Delta−1/2.

  3. Create a complementary classification metric based on the normalized fluctuation amplitude of Omega. This allows for a distinction between regimes where "for quasiperiodic motion there exists a constant C > 0, independent of the initial phase, such that Omega(tau) − Omega∞ ≤ C τ, and where for trajectories with initial conditions distributed according to the invariant measure of an ergodic chaotic component with absolutely integrable autocorrelation and strictly positive asymptotic variance: Eµ Omega(τ) − Omega∞ squared = σ2omega τ + o(τ−1)."

  4. Enhance ensemble classification robustness by using a Matthews correlation coefficient (MCC) threshold maximization. This allows testing the robustness of both methodologies for the study of initial conditions ensembles that may contain both regular and chaotic trajectories to determine the optimal threshold for each energy ensemble.

Abstract

In this paper, we investigate the time-averaged phase space arclength of a trajectory, expressed through a normalized Lagrangian descriptor, as a trajectory-based diagnostic for distinguishing regular and chaotic dynamics. The proposed indicator, Ω, is computed by integrating the equations of motion together with a single scalar arclength-accumulation equation and therefore does not require variational equations, neighboring trajectories, or spatial derivatives of a Lagrangian descriptor field. Using Birkhoff's ergodic theorem together with estimates for quasiperiodic motion and correlation-based arguments for chaotic ergodic dynamics, we show that regular and chaotic trajectories exhibit different finite-time behavior despite the convergence of Ω in both regimes. For quasiperiodic trajectories with Diophantine frequency vectors, deviations from the limiting average decay as O(τ-1), whereas sufficiently mixing chaotic dynamics exhibit root-mean-square fluctuations of order τ-1/2 under suitable correlation assumptions. This difference in finite-time convergence and fluctuation scaling provides the theoretical basis for using Ω to discriminate between the two dynamical regimes. We validate the proposed approach in the Hénon--Heiles and Fermi--Pasta--Ulam systems, comparing the resulting classification of initial conditions with SALI as a reference chaos indicator. The results support Ω as a complementary single-trajectory diagnostic for chaos detection and as a practical source of scalar features for the analysis of trajectory ensembles, particularly in settings involving large numbers of trajectories or high-dimensional systems.

Sources

Related papers