Normalized ground states to a cooperative system of Schrödinger equations with generic \(L^2\)-subcritical or \(L^2\)-critical nonlinearity (Q2135308)

For prescribed numbers \(a=(a_1,\dots,a_M)\in(0,\infty)^M\) and a large class of non-linearities \(F\in C^1(\mathbb{R}^M)\), the author studies existence and properties of solutions \((\lambda,u)=(\lambda_1,\dots,\lambda_M,u_1,\dots,u_M)\in\mathbb{R}^M\times \mathcal{S}(a)\) to the system of equations \begin{align*} \begin{cases} -\Delta u_j - \frac{\partial F(u)}{\partial u_j} = -\lambda_j u_j \\ \int_{\mathbb{R}^N} u_j^2 = a_j^2 \\ (\lambda_j,u_j)\in\mathbb{R}\times H^1(\mathbb{R}^N) \end{cases} \qquad \text{for all}\ j\in\{1,\dots,M\}\,.\tag{1} \end{align*} Here \[ \mathcal{S}(a) := \{u\in H^1(\mathbb{R}^N)^M :\, \int_{\mathbb{R}^N}u_j^2 = a_j^2\ \text{for all}\ j\in\{1,\dots,M\}\}\,. \] Particular emphasis is put on ground states, i.e., minimizers of the corresponding energy functional \[ J: u\in H^1(\mathbb{R}^N)^M \mapsto J(u) := \int_{\mathbb{R}^N}\left(\frac12 \sum_{j=1}^M|\nabla u_j|^2 - F(u)\right)\,dx \in\mathbb{R} \] over the set \[ \mathcal{D}(a) := \{u\in H^1(\mathbb{R}^N)^M :\, \int_{\mathbb{R}^N}u_j^2 \leq a_j^2\ \text{for all}\ j\in\{1,\dots,M\}\}\,. \] The work mainly deals with non-linearities that are either \(L^2\)-subcritical (in which case \(J(u)\) is bounded from below for all \(a\in(0,\infty)^M\)) or \(L^2\)-critical (in which case (un)boundedness of \(J(u)\) from below depends on \(a\)). To state the results of the paper more precisely, let \(2_\#:=2+4/N\) and consider the assumptions \begin{itemize} \item[(F1)] \(\eta_\infty:=\limsup_{|u|\to\infty}\frac{F(u)}{|u|^{2_\#}} < \infty\). \item[(F2)] \(\lim_{u\to0}\frac{F(u)}{|u|^2} = 0\). \item[(F3)] \(\eta_0:=\liminf_{u\to0}\frac{F(u)}{|u|^{2_\#}} > 0\). \item[(a1)] \(2\eta_\infty C_{N,2_\#}^{2_\#} |a|^{4/N}<1\). \item[(a2)] \(2\eta_0 C_{N,2_\#}^{2_\#} M^{2/N}\min_{1\leq j\leq M}a_j^{4/N}>1\). \end{itemize} Here \(C_{N,p}>0\) denotes the best constant in the Gagliardo-Nirenberg inequality \(\|u\|_{L^p(\mathbb{R}^N)}\leq C_{N,p}\|u\|_2^{1-\delta_p}\|\nabla u\|_2^{\delta_p}\) with \(2<p<2N/(N-2)\) and \(\delta_p=N(1/2-1/p)\). Examples of non-linearities \(F\) for which the main results of the paper (Theorems 1.1 and 1.5, Propositions 1.2 and 1.4) hold are given at the end of the introduction. For \(M=1\), the author shows under the assumptions (F1)--(F3), (a1), and (a2) the existence of a solution \((\lambda,u)\in(0,\infty)\times \mathcal{S}(a)\) to the system (1) such that \(0>J(u)=\inf_{\mathcal{D}(a)}J=\inf_{\mathcal{S}(a)}J\). In particular, for even \(F\), the minimizer has constant sign and is, up to translations, radial and radially monotone (Theorem~1.1). Under slightly stronger assumptions on \(F\), the minimizer \(u\) is shown to be strictly radially monotone and to obey \(|u|>0\) (Proposition~1.2). Proposition~1.4 provides an existence result for non-radial minimizers of \(J(u)\). For \(M\geq2\), the author considers even, non-negative, and non-decreasing functions \(F_j,\tilde F_{i,j}\in C^1(\mathbb{R})\setminus\{0\}\) to construct non-linearities of the type \[ F(u) = \sum_{j=1}^M F_j(u_j) + \sum_{\ell=1}^L \prod_{j=1}^M \tilde F_{\ell,j}(u_j) \] for some \(L\in\mathbb{N}\) such that (F1)--(F3), (a1), and (a2) are satisfied or \[ F(u) = \sum_{j=1}^M F_j(u_j) + \sum_{i<j}^M \tilde F_{i,j}(u_i)\tilde F_{j,i}(u_j) \] such that \(\eta_0=\infty\) and (F1)--(F3), and (a1) are satisfied. Under further suitable conditions on the functions \(F_j\) and \(\tilde F_{i,j}\), the author shows existence of solutions \((\lambda,u)\in(0,\infty)^M\times\mathcal{S}(a)\) to (1) such that \(0>J(u)=\inf_{\mathcal{D}(a)}J=\inf_{\mathcal{S}(a)}J\) and that each component of \(u\) is radial, positive, and radially non-increasing (Theorem 1.5). Besides the main results, Propositions~1.7 and 4.1 state orbital stability of ground state solutions to (1) and provide a more detailed analysis of the ground state energy map \((0,\infty)^M\ni a\mapsto\inf\{J(u):\,u\in\mathcal{D}(a)\}\). In regard of finding reasonable assumptions for \(F\) to prove existence of solutions to (1) in the \(L^2\)-(sub)critical case, the present work joins, among others, the works [\textit{T. Bartsch} and \textit{L. Jeanjean}, Proc. R. Soc. Edinb., Sect. A, Math. 148, No. 2, 225--242 (2018; Zbl 1393.35035); \textit{T. Cazenave} and \textit{P.-L. Lions}, Commun. Math. Phys. 85, 549--561 (1982; Zbl 0513.35007); \textit{T. Gou} and \textit{L. Jeanjean}, Nonlinear Anal., Theory Methods Appl., Ser. A, Theory Methods 144, 10--22 (2016; Zbl 1457.35068); \textit{N. Ikoma}, Adv. Nonlinear Stud. 14, No. 1, 115--136 (2014; Zbl 1297.35218); \textit{L. Jeanjean} and \textit{S.-S. Lu}, ``On global minimizers for a mass constrained problem'', Preprint, \url{arXiv:2108.04142}; \textit{M. Shibata}, Manuscr. Math. 143, No. 1--2, 221--237 (2014; Zbl 1290.35252); \textit{M. Shibata}, Math. Z. 287, No. 1--2, 341--359 (2017; Zbl 1382.35012); \textit{C. A. Stuart}, Proc. Lond. Math. Soc. (3) 45, 169--192 (1982; Zbl 0505.35010)]. The main difficulty that has to be circumvented in this and the above-mentioned works is to show that minimizing sequences do not lose \(L^2\)-norm. Since the limit of a weakly convergent sequence in \(\mathcal{D}(a)\) belongs to \(\mathcal{D}(a)\) again, the author chooses to work with \(\mathcal{D}(a)\) instead of \(\mathcal{S}(a)\) directly. Proving that a limit point in \(\mathcal{D}(a)\) actually belongs to \(\mathcal{S}(a)\) is more difficult for \(M\geq2\) since \(\mathcal{S}(a)\subsetneq\partial\mathcal{D}(a)\) (unlike in the case \(M=1\), where \(\mathcal{S}(a)=\partial\mathcal{D}(a)\)). The idea of introducing the set \(\mathcal{D}(a)\) appeared, e.g., in [\textit{B. Bieganowski} and \textit{J. Mederski}, J. Funct. Anal. 280, No. 11, Article ID 108989, 26 p. (2021; Zbl 1465.35151)] for \(M=1\) and in [\textit{J. Mederski} and \textit{J. Schino}, Calc. Var. Partial Differ. Equ. 61, No. 1, Paper No. 10, 31 p. (2022; Zbl 1490.35367)] for \(M\geq2\).