Extremal sections for a Trudinger-Moser functional on vector bundle over a closed Riemann surface (Q6143102)

Let \((E,h,D)\) be a vector bundle of finite rank and over a closed Riemann surface \((\Sigma,g)\), \(\Gamma(E)\) be a set of all smooth sections and \(\mathcal{D}\) be a \(T^*\Sigma\otimes\Gamma(E)\)-valued connection on \(\Gamma(E)\) such that \(T^*\Sigma\) represents the cotangent bundle over \((\Sigma,g)\). Let \(W^{1,2}(\Sigma,E)\) be the completion of \(\Gamma(E)\) w.r.t. the norm \[ \|\sigma\|_{W^{1,2}(\Sigma,E)}=\left(\int_\Sigma(|\mathcal{D}\sigma|^2+|\sigma|^2)dv_g\right)^{1/2}. \] Let \(\mathcal{H}_0=\mathcal{D}^{-1}(\{0\})\) and \(\mathcal{H}_1=\left\{\sigma\in W^{1,2}(\Sigma,E):\displaystyle\int_{\Sigma}\langle\sigma,v\rangle dv_g=0,\text{ for all }v\in\mathcal{H}_0\right\}\). Define the bundle Laplace-Beltrami operator \(\Delta_E\) as \[ \int_\Sigma\langle \Delta_E\sigma,\psi\rangle dv_g=\int_\Sigma\langle\mathcal{D}\sigma,\mathcal{D}\phi\rangle dv_g,\text{ for all }(\sigma,\phi)\in \Gamma(E)\times W^{1,2}(\Sigma,E). \] The first eigenvalue associated to \(\Delta_E\) is define as \(\displaystyle\lambda_1(E)=\inf_{\sigma\in \mathcal{H}_1,||\sigma||^2_2=1}\left[\int_\Sigma|\mathcal{D}\sigma|^2dv_g\right].\) For \(\alpha<\lambda_1(E)\), we define a new norm on \(\|\sigma\|_{1,\alpha}=\left(\displaystyle\int_\Sigma\left(|\mathcal{D}\sigma|^2-\alpha|\sigma|^2\right)dv_g\right)^{1/2}\) for \(\sigma\in \Gamma(E)\). Then the authors state that for any \(\alpha<\lambda_1(E)\) and for \(\beta\le 4\pi\), the supremum of \(\displaystyle \int_\Sigma \exp(\beta|\sigma|^2)dv_g\) on the set of sections \(\sigma\in\mathcal{H}_1\) such that \(\|\sigma\|_{1,\alpha}\le 1\) can be reached by a section \(\sigma_0\in\mathcal{H}_1\cap C^1(\Sigma,E)\) such that \(\|\sigma_0\|_{1,\alpha}=1\) (Theorem 1). Also, they state that the supremum of \(\displaystyle \int_\Sigma \exp(\beta(1+\alpha\|\sigma\|_2^2)|\sigma|^2)dv_g\) is finite on the set \(A=\{\sigma\in\mathcal{H}_1\) such that \(\|\mathcal{D}\sigma\|_2=1\}\) and it is infinite either when \((\alpha<\lambda_1(E)) \text{ and } \beta>4\pi\)) or when \((\alpha\ge \lambda_1(E) \text{ and } \beta>0)\). Besides they state that there exists a constant \(\alpha_0>0\) such that if \(\alpha\in [0,\alpha_0]\) and \(\beta\le 4\pi\), then the supremum of \(\displaystyle \int_\Sigma \exp(\beta(1+\alpha\|\sigma\|_2^2)|\sigma|^2)dv_g\) on \(A\) is attained by a section \(\sigma_0\in\mathcal{H}_1\cap C^1(\Sigma,E)\) such that \(\|\mathcal{D}\sigma_0\|_2=1\) (Theorem 2).