Minimizers of mass critical Hartree energy functionals in bounded domains (Q1669806)

The paper is devoted to studying the mass critical Hartree energy functional in a bounded domain \(\Omega\subset\mathbb{R}^4\): \[ e(a):= \inf\limits_{u\in H_0^1(\Omega),\;\|u\|_{L_2(\Omega)}=1} E_a(u), \] where the energy functional \(E_a(u)\) is defined by \[ E_a(u):=\int\limits_{\Omega} \|\nabla u(x)\|^2 + V(x)\|u(x)\|^2\,dx - \frac{a}{2}\int\limits_{\Omega\times\Omega} \frac{|u(x)|^2|u(y)|^2}{|x-y|^2}\, dx\,dy \] and the trapping potential \(V(x)\geqslant 0\) satisfies \[ \min_{x\in\overline{\Omega}}V (x) = 0\quad\text{and}\quad V\in C^\alpha(\overline{\Omega})\quad\text{for some \(0<\alpha<1\)}. \] The boundary \(\partial\Omega\) is assumed to be \(C^1\)-smooth and to satisfy the interior ball condition. It is known that the following nonlinear Hartree equation \[ -\Delta u+ u - \left(\int\limits_{\mathbb{R}^4} \frac{|u(y)|^2}{|x-y|^2}\,dy\right) u=0\quad\text{in}\quad\mathbb{R}^4,\qquad u\in H^1(\mathbb{R}^4) \] possesses a unique positive solution \(Q=Q(|x|)\). The first result of the paper is as follows. Theorem. If \(0\leqslant a<a^*:=\|Q\|_{L_2(\Omega)}^2\), then there exists at least one minimizer for \(e(a)\). If \(a\geqslant a^*\), then there is no minimizer for \(e(a)\). It is also shown that each minimizer is sign-definite, and this is why only positive minimizers are studied below. Assume that the potential \(V(x)\) has \(n\geqslant 1\) distinct zero points \(x_i\in\overline{\Omega}\) satisfying \(V(x_i) = 0\) and \(V(x)> 0\) for any \(x\neq x_i\), \( i = 1,\ldots,n\). Moreover, we suppose that there exist \(p_i > 0\) and \(C >0\) such that \( V (x) = h(x) \prod_{i=1}^n|x-x_i|^{p_i}\quad\text{and}\quad C<h(x)<1/C\quad\text{in}\quad\overline{\Omega}, \) where \(\lim\limits_{x\to x_i} h(x)\) exists for \(1\leqslant i \leqslant n\). Set \[ \begin{aligned} &p:=\max\{ p_1,\ldots,p_n\}, \qquad \kappa_i:=\lim\limits_{x\to x_i} \frac{V(x)}{|x - x_i|^p}, \\ &\lambda_i:= \left(\frac{p \kappa_i}{2}\int\limits_{\mathbb{R}^4} |x|^p |Q(x)|^2\,dx\right)^{\frac{1}{p+2}}, \\ &\kappa:=\min\{\kappa_1,\ldots,\kappa_n\},\qquad \lambda:=\min\{\lambda_i:\, x_i\in\Omega,\, 1 \leqslant i \leqslant n\}, \\ &Z_1:=\{x_i\in\Omega:\, p_i=p\},\qquad Z_0:=\{x_i\in\partial\Omega:\, p_i=p\}. \end{aligned} \] The next result of the paper is the following theorem. Theorem. Suppose \(V(x)\) satisfies the above assumptions and \(Z_1\not=\emptyset\). Let \(u_a\) be a positive minimizer of \(e(a)\) for \(a<a^*\). Then for any sequence \(\{a_k\}\) satisfying \(a_k\nearrow a^*\) as \(k\to\infty\), there exist a subsequence, still denoted by \(\{a_k\}\), and \(x_0\in Z_1\) such that each \(u_{a_k}\) has a unique maximum point \(z_{a_k}\) satisfying \(\lim_{k\to\infty} z_{a_k} = x_0\), \[ \lim\limits_{k\to\infty} (a^* - a_k)^{\frac{2}{2+p}} u_{a_k} \left( (a^* - a_k)^\frac{1}{2+p}x+ z_{a_k} \right) =\frac{\lambda^2 Q(\lambda x)}{\|Q\|_{L_2(\Omega)}^2}\qquad\text{strongly in}\quad H^1(\mathbb{R}^4). \] The main result of the paper reads as follows. Theorem. Suppose that \(V(x)\) satisfies the above assumptions and \(Z_1 =\emptyset\). Let \(u_a\) be a positive minimizer of \(e(a)\) for \(a<a^*\). Then for any sequence \(\{a_k\}\) satisfying \(a_k\nearrow a^*\) as \(k\to\infty\), there exist a subsequence, still denoted by \(\{a_k\}\), \(x_0\in Z_1\) such that \(u_{a_k}\) satisfies \[ \lim\limits_{k\to\infty} \epsilon_{a_k}^2 u_{a_k} ( \epsilon_{a_k} x + z_{a_k}) =\frac{ Q(x)}{\|Q\|_{L_2(\Omega)}^2}\qquad\text{strongly in}\quad H^1(\mathbb{R}^4), \] where \(z_{a_k}\) is the unique maximum point of \(u_{a_k}\) satisfying \[ |z_{a_k}-x_0|\approx \frac{p+2}{2} \epsilon_{a_k} \ln\frac{1}{\epsilon_{a_k}}\qquad\text{as}\quad k\to\infty, \] and \(\epsilon_{a_k}>0\) satisfies \[ \epsilon_{a_k}\approx\left(\frac{2^{1+p}}{p a^* k}\right)^\frac{1}{2+p} (a^*-a_k)^\frac{1}{2+p} \left(\ln\frac{1}{a_*-a_k}\right)^{-\frac{p}{2+p}}\qquad\text{as}\quad k\to\infty. \]