Numerical Solution of the Small Dispersion Limit of the Camassa-Holm and Whitham Equations and Multiscale Expansions

TL;DR

This paper investigates the small dispersion limit of the Camassa-Holm equation, providing asymptotic descriptions of oscillations, numerical validation, and a multiscale expansion involving Painlevé II solutions.

Contribution

It introduces a conjecture for the phase of asymptotic solutions, validates it numerically, and develops a multiscale expansion to improve the description near the leading edge.

Findings

Difference between CH and asymptotic solution is of order ε in the interior.

Difference at the trailing edge is of order √ε.

Multiscale expansion accurately describes oscillation amplitude near the leading edge.

Abstract

The small dispersion limit of solutions to the Camassa-Holm (CH) equation is characterized by the appearance of a zone of rapid modulated oscillations. An asymptotic description of these oscillations is given, for short times, by the one-phase solution to the CH equation, where the branch points of the corresponding elliptic curve depend on the physical coordinates via the Whitham equations. We present a conjecture for the phase of the asymptotic solution. A numerical study of this limit for smooth hump-like initial data provides strong evidence for the validity of this conjecture. We present a quantitative numerical comparison between the CH and the asymptotic solution. The dependence on the small dispersion parameter $\epsilon$ is studied in the interior and at the boundaries of the Whitham zone. In the interior of the zone, the difference between CH and asymptotic solution is of the order $\epsilon$, at the trailing edge of the order $\sqrt{\epsilon}$ and at the leading edge of the order $\epsilon^{1/3}$. For the latter we present a multiscale expansion which describes the amplitude of the oscillations in terms of the Hastings-McLeod solution of the Painlev\'e II equation. We show numerically that this multiscale solution provides an enhanced asymptotic description near the leading edge.