Beyond-all-order asymptotics for homoclinic snaking of localised patterns in reaction-transport systems

TL;DR

This paper develops a general asymptotic method to analytically approximate the width of homoclinic snaking regions near Turing bifurcations in reaction-diffusion systems, validated by numerical examples.

Contribution

It extends exponential asymptotics to arbitrary reaction-diffusion systems, providing explicit formulas for the Maxwell point and snaking width near bifurcations.

Findings

Derived algebraic expressions for the Maxwell point.

Obtained exponentially small estimates for snaking width.

Validated theory with numerical simulations of reaction-diffusion models.

Abstract

Spatially localised stationary patterns of arbitrary wide spatial extent emerge from subcritical Turing bifurcations in one-dimensional reaction-diffusion systems. They lie on characteristic bifurcation curves that oscillate around a Maxwell point in a homoclinic snaking phenomenon. Here, a generalisation of the exponential asymptotics method by Chapman \& Kozyreff is developed to provide leading-order expressions for the width of the snaking region close to a Turing bifurcation's super/sub-critical transition in arbitrary $n$-component reaction-diffusion systems. First, general expressions are provided for the regular asymptotic approximation of the Maxwell point, which depends algebraically on the parametric distance from the codimension-two super/sub-critical Turing bifurcation. Then, expressions are derived for the width of the snaking, which is exponentially small in the same distance. The general theory is developed using a vectorised form of regular and exponential asymptotic expansions for localised patterns, which are matched at a Stokes line. Detailed calculations for a particular example are algebraically cumbersome, depend sensitively on the form of the reaction kinetics, and rely on the summation of a weakly convergent series obtained via optimal truncation. Nevertheless, the process can be automated, and code is provided that carries out the calculations automatically, only requiring the user to input the model, and the values of parameters at a codimension-two bifurcation. The general theory also applies to higher-order equations, including those that can be recast as a reaction-diffusion system. The theory is illustrated by comparing numerical computations of localised patterns in two versions of the Swift-Hohenberg equation with different nonlinearities, and versions of activator-inhibitor reaction-diffusion systems.