A variational approach to reaction-diffusion equations with forced speed in dimension 1 (Q484411)

The authors analyse the long-time behaviour of solutions to the reaction-diffusion equation \[ u_t=u_{xx}+f(x-ct,u) \] on the real line, together with a bounded, nonnegative and compactly supported initial datum. This equation arises as a model of biological populations subject to the effect of climate change. Above, the parameter \(c\in\mathbb{R}\) corresponds to the speed at which the climate change occurs. The main question is if the population can persist, i.e., keep up with the movement of its favourable environment. In generalization of \textit{H. Berestycki} et al. [Bull. Math. Biol. 71, No. 2, 399--429 (2009; Zbl 1169.92043)], the assumptions on the nonlinearity \(f\) are not of KPP type: Apart from regularity assumptions and \(f(z,0)=0\), it is required that there exist \(M\), \(R\), \(\delta>0\) such that \[ \begin{aligned} f(z,s) &\leq 0\quad\text{for all }s\in\mathbb{R}\setminus (0,M)\text{ and all }z\in\mathbb{R}\quad\text{(prevention of overcrowding),}\\ f(z,s) &\leq -\delta s\quad\text{for all }|z|>R\text{ and all } s\in(0,M)\quad\text{(boundedness of favourable environment).}\end{aligned} \] Lastly, the functional \[ E_c[u]=\int_{\mathbb{R}}e^{cz}\left(\frac{u_z(z)^2}{2}-\int_0^{u(z)} f(z,t)dt\right)dz \] defined on the space \(H^1(\mathbb{R},e^{cz}dz)\) should admit at least one negative value in the case \(c=0\). The first result is about the existence of travelling wave pulses \(u(t,x)=U(x-ct)\), with \(U(\pm\infty)=0\): There exist two threshold values \(\overline{c}\geq \underline{c}>0\) such that for \(0<c<\underline{c}\), (non-trivial) travelling wave solutions exist, whereas for \(c>\overline{c}\), they do not exist. The main idea of the proof is variational: One can show that nonnegative minimizers of the functional \(E_c\) above correspond to travelling wave solutions. By a continuity argument w.r.t. \(c\) and comparison to solutions of the Fisher-KPP equation, it is proven that for \(c>\overline{c}\) the minimizer is identically equal to zero. The authors conjecture that \(\underline{c}=\overline{c}\), which is not proven but indicated by numerical simulations. Second, the long-time behaviour of solutions is investigated for initial data in \(H^2\). As \(t\to\infty\), \(u(t,\cdot-ct)\) converges either to zero or to a travelling wave solution. The proof is again based on a variational approach which is related to that in [\textit{T. Zelenyak}, Differ. Uravn. 4, 34--45 (1968; Zbl 0162.15301)]. Subsequently, the relationship between linear stability of the zero solution and convergence to travelling waves is analysed. As it is the case for the KPP equation, linear instability of \(0\) (in the sense of generalized eigenvalues) yields convergence to a travelling wave solution for each \(c\in(0,c^*)\), where \(c^*\) is a critical speed. However, for general nonlinearities (e.g., bistable) \(f\), the converse does not hold. It is possible that both \(0\) is linearly stable and travelling wave solutions do exist. Then, the choice of the initial datum decides on persistence or extinction. The proofs are based on properties of \(E_c\) in combination with the so-called mountain pass theorem.