We discuss the stability of a Hamiltonian system by comparing the standard Lyapunov error (LE) with the forward error (FE) due to a small random perturbation. We introduce also the reversibility error (RE) where the evolution is computed forward up to time t and backwards to t = 0 in presence of noise. This procedure has been investigated in the case of symplectic maps, but it turns out that the results are simpler in the case of a noisy flow, in the limit of zero noise amplitude. Indeed the stochastic processes defined by the displacement of the noisy orbit at time t for FE, or at time 0 for RE after the evolution up to time t, satisfy linear Langevin equations, are Gaussian processes, and the errors are just their root mean square deviations. All the errors are expressed in terms of the fundamental matrix L(t) of the tangent flow and can be evaluated numerically using a symplectic integrator. Letting e(L)(t) be the Lyapunov error and e(R)(t) be the reversibility error a very simple relation holds e(R)(2)(t) = integral(t)(0) e(L)(2)(s) ds. The integral relation is quite natural since the local errors due to a random perturbations accumulate during the evolution whereas for the Lyapunov case the error is introduced only at time zero and propagated. The plot of errors for initial conditions in a Poincare section reflects the phase portrait, whereas in the action plane it allows to single out the resonance strips. We have applied the method to a 3D Hamiltonian model H = H-0(J) + lambda V (Theta), where analytic estimates can be obtained for the single resonances from perturbation theory. This allows to inspect the double resonance structure where the single resonance strips intersect. We have also considered the Henon-Heiles Hamiltonian to show numerically the equivalence of the errors apart from a shift of 1/2 in the power law exponent in the case of regular orbits. The reversibility error method (REM), previously introduced as the error due to round off in the symplectic integration, appears to be comparable with RE also for the models considered here. (C) 2018 Elsevier Ltd. All rights reserved.
机构:
Hong Kong Polytech Univ, Dept Appl Math, Kowloon, Hong Kong, Peoples R ChinaHong Kong Polytech Univ, Dept Appl Math, Kowloon, Hong Kong, Peoples R China
Cui, Jianbo
Liu, Shu
论文数: 0引用数: 0
h-index: 0
机构:
UCLA, Dept Math, Los Angeles, CA 90095 USAHong Kong Polytech Univ, Dept Appl Math, Kowloon, Hong Kong, Peoples R China
Liu, Shu
Zhou, Haomin
论文数: 0引用数: 0
h-index: 0
机构:
Georgia Tech, Sch Math, Atlanta, GA 30332 USAHong Kong Polytech Univ, Dept Appl Math, Kowloon, Hong Kong, Peoples R China