Convergence to a receding wave in a monostable free boundary problem

We study a monostable reaction-diffusion equation of the form $u_t=du_{xx}+f(u)$ over a semi-infinite spatial domain $[g(t),\infty)$, with $x=g(t)$ the free boundary whose evolution is governed by equations derived from a ``preferred population density'' principle, which postulates that the species with population density $u(t,x)$ and population range $[g(t),\infty)$ maintains a certain density $\delta$ at the habitat edge $x=g(t)$. In the ``high-density'' regime, where $\delta$ exceeds the carrying capacity of the favourable environment represented by a monostable function $f(u)$, it is known (see \cite{DLNS} for the case of a bounded population range $[g(t), h(t)]$) that for large time, the front retreats as time advances. In this work, the unboundedness of the population range $[g(t),\infty)$ allows us to prove that, as time $t$ converges to infinity, the free boundary $x=g(t)$ converges to $\infty$ with a constant asymptotic speed $c(\delta)>0$ determined by an associated semi-wave problem, and the population density $u(t,x)$ has the property that $u(t,x+g(t))$ converges uniformly to $q_{c(\delta)}(x)$, the semi-wave profile function associated with the speed $c(\delta)$. It turns out that in the retreating situation considered here, some key techniques developed for advancing fronts in related free boundary models do not work anymore. This difficulty is overcome here by a ``touching method", which uses a family of lower and upper solutions constructed from semi-waves of some carefully designed auxiliary problems to touch the solution $u(t,x)$ at the moving boundary $x=g(t)$, thereby generating a setting where the comparison principle can be used to obtain the desired estimates for $g'(t)$ and $u(t,x)$. We believe this method will find applications elsewhere.