THE FERMI-PASTA-ULAM PROBLEM REVISITED - STOCHASTICITY THRESHOLDS IN NONLINEAR HAMILTONIAN-SYSTEMS

Citation
L. Casetti et al., THE FERMI-PASTA-ULAM PROBLEM REVISITED - STOCHASTICITY THRESHOLDS IN NONLINEAR HAMILTONIAN-SYSTEMS, Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topics, 55(6), 1997, pp. 6566-6574
Citations number
38
Categorie Soggetti
Physycs, Mathematical","Phsycs, Fluid & Plasmas
ISSN journal
1063651X
Volume
55
Issue
6
Year of publication
1997
Part
A
Pages
6566 - 6574
Database
ISI
SICI code
1063-651X(1997)55:6<6566:TFPR-S>2.0.ZU;2-5
Abstract
The Fermi-Pasta-Ulam alpha model of harmonic oscillators with cubic an harmonic interactions is studied from a statistical mechanical point o f view. Systems of N = 32 to 128 oscillators appear to be large enough to suggest statistical mechanical behavior. A key element has been a comparison of the maximum Lyapunov coefficient lambda(max) of the EPU alpha model and that of the Toda lattice. For generic initial conditio ns, lambda(max)(t) is indistinguishable for the two models up to times that increase with decreasing energy (at fixed N). Then suddenly a bi furcation appears, which can be discussed in relation to the breakup o f regular, solitonlike structures. After this bifurcation, the lambda( max) of the FPU model appears to approach a constant, while the lambda (max) of the Toda lattice appears to approach zero, consistent with it s integrability. This suggests that for generic initial conditions the FPU alpha model is chaotic and will therefore approach equilibrium an d equipartition of energy. There is, however, a threshold energy densi ty epsilon(c)(N) similar to 1/N-2, below which trapping occurs; here t he dynamics appears to be regular, solitonlike, and the approach to eq uilibrium-if any-takes longer than observable on any available compute r. Above this threshold the system appears to behave in accordance wit h statistical mechanics, exhibiting an approach to equilibrium in phys ically reasonable times. The initial conditions chosen by Fermi, Pasta , and Ulam were not generic and below threshold and would have require d possibly an infinite time to reach equilibrium. The picture obtained on the basis of lambda(max) suggests that neither the KAM nor the Nek horoshev theorems in their present form are directly relevant for a di scussion of the phenomenology of the FPU a model presented here.