A sharp form of trace Moser-Trudinger inequality on compact Riemannian surface with boundary (Q970205)

The well-known Moser-Trudinger inequalities are important in the study of two-dimensional problems such as the conformal geometry problems [see \textit{J. Moser}, Indiana Univ. Math. J. 20, 1077--1092 (1971; Zbl 0213.13001)]. There has been many extensions on such inequalities. In this paper, a sharp form of trace Moser-Trudinger inequality on the boundary of compact Riemannian surface is established and the existence of extremal function is proved via the method of blowing up analysis. Let \((\Sigma, g)\) be a compact Riemannian surface with smooth boundary \(\partial \Sigma\) and let \(H^{1,2} (\Sigma)\) be the completion of \(C^{\infty} (\Sigma)\) in the norm \[ \|u\| = \left( \int_{\Sigma} (|u|^2 + | \nabla u|^2 ) d v_g \right)^{\frac{1}{2}} < +\infty , \] where \(dv_g\) denotes the area element on \(\Sigma\). Let \[ \mathcal{H} : = \left\{ u \in H^{1,2}(\Sigma) : \|\nabla u\|_2 : = \left(\int_{\Sigma} |\nabla u|^2 dv_g \right)^{\frac{1}{2}} =1, \quad \int_{\partial \Sigma} u ds_g =0 \right\} , \] where \(ds_g\) denotes the induced length element on \(\partial \Sigma\). Define \[ \begin{aligned} \lambda (\partial \Sigma ) : &= \inf_{u\in H^{1,2} (\Sigma), \int_{\partial \Sigma} u ds_g , u \neq 0} \frac{\| \nabla u\|_2^2}{\int_{\partial \Sigma} u^2 ds_g}, \\ C_{\alpha} (\partial \Sigma ) : &= \sup_{u\in \mathcal{H}} \int_{\partial \Sigma} e^{\pi u^2 (1+\alpha\|u\|^2_{L^2(\partial \Sigma)})} ds_g, \\ J_{\beta}^{\alpha} (u) : &= \int_{\partial \Sigma} e^{\beta u^2 (1+\alpha\|u\|^2_{L^2(\partial \Sigma)})} ds_g, \end{aligned} \] where \(\|u\|^2_{L^2(\partial \Sigma)} = \int_{\partial \Sigma} u^2 ds_g\). It is well-known that \(\lambda (\partial \Sigma) > 0\). It follows from the main result of [\textit{Y. Li} and \textit{P. Liu}, Math. Z. 250, No. 2, 363--386 (2005; Zbl 1070.26017)] that \(C_{\alpha} (\partial \Sigma ) < +\infty\) for \(\alpha \leq 0\). The author proves that (1) \(C_{\alpha} (\partial \Sigma) <+\infty \) for any \(0\leq \alpha < \lambda(\partial \Sigma )\), (2) \(C_{\alpha} (\partial \Sigma) = +\infty \) for any \( \alpha \geq \lambda(\partial \Sigma )\), (3) there exists a \(u_{\alpha} \in \mathcal{H} \cap C^{\infty} (\overline{\Sigma})\) such that \(J^{\alpha}_{\pi} (u_{\alpha}) = C_{\alpha} (\partial \Sigma)\) for sufficiently small \(\alpha>0\). This result is a trace version of the main result in [\textit{Adimurthi} and \textit{O. Druet}, Commun. Partial Differ. Equations 29, No. 1--2, 295--322 (2004; Zbl 1076.46022)] and gives more information than the main result in [Li and Liu, loc. cit.].