Least energy solutions for nonlinear Schrödinger equation involving the fractional Laplacian and critical growth (Q524530)

In this paper, the authors study the existence and behaviour of least-energy solutions for the one-parameter family of nonlinear fractional Schrödinger problems \[ (1.1)\; \begin{cases} (-\Delta)^s u +(\lambda V(x) -\mu)u = |u|^{2^\sharp_s-2}u\quad\text{ in }\mathbb{R}^N,\\ u\geq 0, \quad u\in H^s(\mathbb{R}^N), \end{cases} \] where \(0 < s < 1\), \(N > 5+2s\), and \(2^\sharp_s=\frac{2N}{N-2s}\) denotes the critical fractional Sobolev exponent. For the classical Laplacian (\(s=1\)), this kind of problems where first investigated by \textit{T. Bartsch and Z.-Q. Wang} [Comm. PDE 20, No. 9--10, 1725--1741 (1995; Zbl 0837.35043)] for a subcritical nonlinearity, and it has attracted a lot of interest since then. For the fractional Laplacian (\(0 < s < 1\)) the subcritical problem has been studied by the second author [Comm. Pure Appl. Anal. 13, No. 1, 237--248 (2014; Zbl 1291.35366)]. In the present paper, the following assumptions are made on the potential \(V\) and the parameter \(\mu\): \textbf{(V1)} \(V\geq 0\) is continuous and its zero set has nonempty interior \(\Omega=\text{int }V^{-1}(0)\) and smooth boundary \(\partial\Omega\); \textbf{(V2)} \(\liminf\limits_{|x|\to\infty}V(x)>0\), and \textbf{(V3)} \(0<\mu<\mu_1\), where \(\mu_1\) is the first eigenvalue of the \(s\)-Laplacian on \(\Omega\) with Dirichlet boundary condition. In a first result (Theorem 1.2), the authors show that (1.1) has a least-energy solution when \(\lambda>0\) is large. Moreover, it is shown that the equation \[ (1.10)\; \begin{cases} (-\Delta)^s u -\mu u = |u|^{2^\sharp_s-2}u\quad\text{ in }\Omega,\\ u\geq 0\quad\text{ in }\Omega, \quad u=0\quad\text{ on } \mathbb{R}^N\backslash\Omega. \end{cases} \] plays the role of a limit problem, in the following sense: If \((\lambda_n)_n\) is a sequence of positive numbers such that \(\lambda_n\to\infty\), and \((u_n)_n\) is a sequence of least-energy solutions of (1.1) with \(\lambda=\lambda_n\). Then, up to a subsequence, \((u_n)_n\) converges strongly in \(H^s(\mathbb{R}^N)\) to a least-energy solution \(u_0\) of (1.10). In other words, as \(n\to\infty\), the sequence \((u_n)_n\) localizes at the bottom \(\Omega=\text{int}V^{-1}(0)\) of the potential well \(\lambda V(x)-\mu\). As the authors point out, in the case where \(\Omega\) consists of several connected components, the limit \(u_0\) will in general be positive on more than one component, due to the nonlocality of the fractional Laplacian. In a second result (Theorem 1.3), the authors prove that the same localization phenomenon on \(\Omega\) occurs for every sequence of solutions \((u_n)_n\) of (1.1) with \(\lambda=\lambda_n\to\infty\), provided the energy of the sequence is bounded from above. The proofs rely on the \(s\)-harmonic extension of the problems (1.1) and (1.10) to \(\mathbb{R}^{N+1}_+\), in the spirit of \textit{L. Caffarelli and L. Silvestre} [Comm. PDE 32, No. 8, 1245--1260 (2007; Zbl 1143.26002)]. Variational arguments are applied to the energy functionals of the extended problems. The condition on the space dimension \(N>5+2s\) is used in estimating the least-energy level of the Caffarelli-Silvestre extension of (1.10) and guarantees the existence of a least-energy solution to this limit problem.