A remark on singular perturbation methods via the Lyapunov-Schmidt reduction (Q1918527)

The paper concerns the singularly perturbed semilinear parabolic boundary value problem \[ \partial_t u= \varepsilon^2 \partial^2_x u+ f(u, x),\quad -1< x< 1,\quad \partial_x u|_{x= \pm 1}= 0.\tag{1} \] It is assumed that the smooth nonlinearity \(f\) satisfies the conditions \(f(0, x)\equiv f(1, x)\equiv 0\), \(\partial_u f(u, x)\leq \text{const}< 0\) for \(u= 0,1\) and \(x\in [- 1, 1]\), \(\int^1_0 f(u, 0) du= 0\), \(\int^1_0 \partial_x f(u, 0) du\neq 0\) and \(\int^v_0 f(u, 0) du< 0\) for all \(v\in (0, 1)\). Let \(U(x, \varepsilon)\) be an approximate equilibrium to (1) that exhibits a transition layer at \(x= 0\), and let \[ L_\varepsilon:= \varepsilon^2 \partial^2_x+ \partial_u f(U(x, \varepsilon), x). \] Then \(L_\varepsilon\) turns out to have exactly one simple eigenvalue (with eigenfunction \(v_1(\varepsilon)\)) that approaches zero as \(\varepsilon\to 0\). Using an idea of \textit{J. K. Hale} and \textit{K. Sakamoto} [Jap. J. Appl. Math. 5, No. 3, 367-405 (1988; Zbl 0669.34027)], the author proceeds a Lyapunov-Schmidt reduction corresponding to a decomposition \(C([- 1, 1])= \text{span}\{v_1(\varepsilon)\}\oplus Y_1\), and, therefore, he has to prove that \(L^{- 1}_\varepsilon: Y_1\to Y_1\) is uniformly bounded for sufficiently small \(\varepsilon> 0\). This is the main issue of the paper.