Fluctuations of finite-time stability exponents in the standard map and the detection of small islands

TL;DR

This paper analyzes the statistical properties of finite-time stability exponents in the standard map, revealing their sensitivity to small regular islands and proposing a local approximation method for efficient detection.

Contribution

It introduces a local approximation to the stability matrix trace that simplifies calculations and enhances detection of tiny islands in phase space.

Findings

Higher cumulants of stability exponents detect dynamical correlations.

Variance scaling reveals presence of small islands as small as 0.01%.

Inverse trace average is highly sensitive and exhibits fractal behavior.

Abstract

Some statistical properties of finite-time stability exponents in the standard map can be estimated analytically. The mean exponent averaged over the entire phase space behaves quite differently from all the other cumulants. Whereas the mean carries information about the strength of the interaction, and only indirect information about dynamical correlations, the higher cumulants carry information about dynamical correlations and essentially no information about the interaction strength. In particular, the variance and higher cumulants of the exponent are very sensitive to dynamical correlations and easily detect the presence of very small islands of regular motion via their anomalous time-scalings. The average of the stability matrix' inverse trace is even more sensitive to the presence of small islands and has a seemingly fractal behavior in the standard map parameter. The usual accelerator modes and the small islands created through double saddle node bifurcations, which come halfway between the positions in interaction strength of the usual accelerator modes, are clearly visible in the variance, whose time scaling is capable of detecting the presence of islands as small as 0.01% of the phase space. We study these quantities with a local approximation to the trace of the stability matrix which significantly simplifies the numerical calculations as well as allows for generalization of these methods to higher dimensions. We also discuss the nature of this local approximation in some detail.