A sharp Trudinger-Moser type inequality involving \(L^n\) norm in the entire space $\mathbb{R}^n$ (Q2420508)

As well-known, if $\Omega\subseteq\mathbb{R}^n$ is a bounded open set and if $q\in[1,n)$, the Sobolev embedding theorem states that the following continuous embedding holds: \[ W_0^{1,q}(\Omega)\hookrightarrow L^s(\Omega) \quad\text{for all }1\leq s \leq q^* := \frac{qn}{n-q}. \] On the other hand, in the particular case $q = n$, it is known that \[ W_0^{1,n}(\Omega) \hookrightarrow L^s(\Omega)\quad\text{for any }n\leq s < \infty,\eqno{(1)} \] but \textit{it is not true that} $W_0^{1,n}(\Omega)\subseteq L^\infty(\Omega)$. The `right' substitute for (1) when $s = n$ is the celebrated Moser-Trudinger inequality, which reads as follows: \[ \sup_{\begin{subarray}{c} u\in W_0^{1,n}(\Omega) \\ \|\nabla u\|_{L^n(\Omega)}\leq 1 \end{subarray}} \int_\Omega\exp\bigg(\alpha|u|^{\frac{n}{n-1}}\bigg)\,\mathrm d x < \infty\,\,\Longleftrightarrow\,\, \alpha\leq \alpha_n := n\omega_{n-1}^{\frac{1}{n-1}},\eqno{(2)} \] where $\omega_{n-1}$ is the $(n-1)$-dimensional Hausdorff measure of $\delta B(0,1)\subseteq\mathbb{R}^n$. In this paper, the authors establish a sharp generalization of the Moser-Trudinger inequality (2), by using suitable $L^n$-norms on the entire space $\mathbb{R}^n$. As it results from several papers, in the case $\Omega = \mathbb{R}^n$ the exponential function appearing in (2) must be replaced with the more `natural' one, which is \[ \Phi(t) := e^t - \sum_{j = 1}^{n-2}\frac{t^j}{j!}. \] Thus, one of the main result of the paper is the following. Theorem 1. For every $\alpha\in[0,1)$, one has \[ S:=\sup_{\begin{subarray}{c} u\in W^{1,n}(\mathbb{R}^n) \\ \|u\|_{W^{1,n}(\mathbb{R}^n)}=1 \end{subarray}} \int_{\mathbb{R}^n}\Phi\bigg(\alpha_n\,|u|^{\frac{n}{n-1}} \Big(1+\alpha\|u\|_{L^n(\mathbb{R}^n)}^n\Big)^{\frac{1}{n-1}}\bigg)\mathrm d x < \infty, \eqno{(3)} \] where $\alpha_n$ is as in (2). Furthermore, $S = \infty$ for every $\alpha \geq 1$. Furthermore, the authors prove the `optimality' of Theorem 1 for $\alpha\sim 0$. Theorem 2. For every $\alpha$ sufficiently small, there exists $u_\alpha\in W^{1,n}(\mathbb{R}^n)$ such that \[ \int_{\mathbb{R}^n}\Phi\bigg(\alpha_n\,|u_\alpha|^{\frac{n}{n-1}} \Big(1+\alpha\|u_\alpha\|_{L^n(\mathbb{R}^n)}^n\Big)^{\frac{1}{n-1}}\bigg)\mathrm d x = S. \] As described by the authors at the end of the introduction, the proof of the first part of Theorem 1 relies on a suitable blow-up analysis on the (non-linear) Euler-Lagrange equation associated to the Moser-Trudinger functional; the proof of the second part (i.e., $S = \infty$ for all $\alpha\geq 1$), instead, is based on a test-function argument which needs to take care of the fact that \[ \inf_{u\in W^{1,n}(\Omega)\setminus\{0\}} \frac{\|\nabla u\|^n_{L^n(\Omega)}}{\|u\|^n_{L^n(\Omega)}} \] is not attained when $\Omega = \mathbb{R}^n$. Finally, the proof of Theorem 2 adapts some ideas which have been proved to be fruitful in a similar setting: it is first derived an upper bound for (3) when the blow-up arises, and then a suitable sequence of functions is constructed to show that this upper bound can be surpassed.