Existence of Solutions of a Non-Linear Eigenvalue Problem with a Variable Weight

We study the non-linear minimization problem on $H^1_0(\Omega)\subset L^q$ with $q=\frac{2n}{n-2}$, $\alpha>0$ and $n\geq4$~: $$\inf_{\substack{u\in H^1_0(\Omega) \|u\|_{L^q}=1}}\int_\Omega a(x,u)|\nabla u|^2 - \lambda \int_{\Omega} |u|^2.$$ where $a(x,s)$ presents a global minimum $\alpha$ at $(x_0,0)$ with $x_0\in\Omega$. In order to describe the concentration of $u(x)$ around $x_0$, one needs to calibrate the behaviour of $a(x,s)$ with respect to $s$. The model case is $$\inf_{\substack{u\in H^1_0(\Omega) \|u\|_{L^q}=1}}\int_\Omega (\alpha+|x|^\beta |u|^k)|\nabla u|^2 - \lambda \int_{\Omega} |u|^2.$$ In a previous paper dedicated to the same problem with $\lambda=0$, we showed that minimizers exist only in the range $\beta<kn/q$, which corresponds to a dominant non-linear term. On the contrary, the linear influence for $\beta\geq kn/q$ prevented their existence. The goal of this present paper is to show that for $0<\lambda\leq \alpha\lambda_1(\Omega)$, $0\leq k\leq q-2$ and $\beta > kn/q + 2$, minimizers do exist.