Scaling with system size of the Lyapunov exponents for the Hamiltonian Mean Field model

TL;DR

This paper demonstrates that the full Lyapunov spectrum of the Hamiltonian Mean Field model scales as N^{-1/3} with system size, revealing a rapid onset of this scaling and a phase transition from weak to strong chaos near a specific energy threshold.

Contribution

The study extends the known N^{-1/3} scaling of the maximal Lyapunov exponent to the entire Lyapunov spectrum and identifies a sharp transition in chaos strength at a critical energy.

Findings

Full Lyapunov spectrum scales as N^{-1/3}

Scaling is observable at smaller N for the spectrum than for the MLE

A sharp transition from weak to strong chaos occurs near U_t ≈ 0.2

Abstract

The Hamiltonian Mean Field (HMF) model is a prototype for systems with long-range interactions. It describes the motion of $N$ particles moving on a ring, coupled through an infinite-range potential. The model has a second order phase transition at the energy $U_c=3/4$ and its dynamics is exactly described by the Vlasov equation in the $N \to \infty$ limit. Its chaotic properties have been investigated in the past, but the determination of the scaling with $N$ of the Lyapunov Spectrum (LS) of the model remains a challenging open problem. We here show that the $N^{-1/3}$ scaling of the Maximal Lyapunov Exponent (MLE), found in previous numerical and analytical studies, extends to the full LS; not only, scaling is "precocious" for the LS, meaning that it becomes manifest for a much smaller number of particles than the one needed to check the scaling for the MLE. Besides that, the $N^{-1/3}$ scaling appears to be valid not only for $U>U_c$, as suggested by theoretical approaches based on a random matrix approximation, but also below a threshold energy $U_t \approx 0.2$. Using a recently proposed method (GALI) devised to rapidly check the chaotic or regular nature of an orbit, we find that $U_t$ is also the energy at which a sharp transition from {\it weak} to {\it strong} chaos is present in the phase-space of the model. Around this energy the phase of the vector order parameter of the model becomes strongly time dependent, inducing a significant untrapping of particles from a nonlinear resonance.