Concentrating bounded states for a class of singularly perturbed Kirchhoff type equations with a general nonlinearity (Q324104)

The goal of the paper is to establish the existence, and to study properties, of a positive solution \(0<u \in H^1(\mathbb{R}^3)\) of the Kirchhoff-type equation \[ -\left(\varepsilon^2 a + \varepsilon b \int_{\mathbb{R}^3}|\nabla u|^2\right)\Delta u + V(x)u = f(u)\tag{1} \] in \(\mathbb{R}^3\) for constants \(a,b>0\) and a sufficiently small parameter \(\varepsilon>0\), under suitable assumptions on the continuous potential \(V:\mathbb{R}^3 \to \mathbb{R}\) and, in particular, novel assumptions on the continuous nonlinearity \(f: \mathbb{R} \to \mathbb{R}\). More precisely, suppose that the potential has a strictly positive infimum over \(\mathbb{R}^3\) and has a well somewhere, i.e., there exists a bounded domain \(\Lambda \subset \mathbb{R}^3\) such that \(\min_\Lambda V < \min_{\partial\Lambda} V\). Suppose also that \(f\) satisfies the same growth conditions as in [\textit{J. Zhang} et al., J. Lond. Math. Soc., II. Ser. 90, No. 3, 827--844 (2014; Zbl 1317.35247)], namely {\parindent=0.7cm \begin{itemize}\item[{\((f_1)\)}] \(\lim_{t\to 0^+} f(t)/t=0\), \item[{\((f_2)\)}] \(\lim_{t \to +\infty} f(t)/t^5=1\), and \item[{\((f_3)\)}] there exist \(\lambda>0\) and \(2<q<6\) such that \(f(t)\geq \lambda t^{q-1}+t^5\) for \(t\geq 0\). \end{itemize}} The main result is that under these assumptions for all \(\varepsilon>0\) sufficiently small there exists a positive variational solution \(u_\varepsilon \in C^{2,\alpha}_{\text{loc}} (\mathbb{R}^3)\) provided either (i) \(4<q<6\) or (ii) \(2<q\leq 4\) and \(\lambda>0\) is sufficiently large. Moreover, this solution concentrates near the set \(\{x \in \Lambda: V(x)=\min_{y\in\Lambda} V(y) \}\) as \(\varepsilon \to 0\), in the sense that there is a point \(x_\varepsilon \in \mathbb{R}^3\) where \(u_\varepsilon\) attains its maximum, whose distance to this set tends to zero, and \(u_\varepsilon\) decays as \(C_1\exp (-C_2|x-x_\varepsilon|/\varepsilon)\) away from \(x_\varepsilon\). The principle difficulty (and novelty) lies in the critical growth assumption on \(f\), which is amplified by the fact that the domain \(\mathbb{R}^3\) is unbounded, meaning the usual methods do not work. Instead, a penalization method is used, [\textit{J. Byeon} and \textit{L. Jeanjean}, Arch. Ration. Mech. Anal. 185, No. 2, 185--200 (2007; Zbl 1132.35078)] or [\textit{J. Byeon} and \textit{Z.-Q. Wang}, Calc. Var. Partial Differ. Equ. 18, No. 2, 207--219 (2003; Zbl 1073.35199)], together with a rather subtle and careful analysis of the resulting solutions. This complements previous work of \textit{G. M. Figueiredo} et al. [Arch. Ration. Mech. Anal. 213, No. 3, 931--979 (2014; Zbl 1302.35356)], where the nonlinearity was assumed to have subcritical growth.