Endpoint resolvent estimates for compact Riemannian manifolds (Q517170)

The \(L_p(\mathbb R^n)\to L_q(\mathbb R^n)\) mapping properties of the resolvent \((-\Delta-z)^{-1}\) such that \(q\) is the Hölder conjugate of \(p\in \left[\frac{2n}{n+2},\frac{2(n+1)}{n+3}\right)\) has been already conducted. The purpose of the present article is to treat the case of \(p=\frac{2(n+1)}{n+3}\) on an \(n\)-dimensional compact manifold \(M^n\). Precisely, for \(\delta>0\), the authors state that \[ ||(-\Delta-z)^{-1}f||_{L_{\frac{2(n+1)}{n-1}}(M^n)}\leq C|z|^{-\frac{1}{n+1}}||f||_{L_{\frac{2(n+1)}{n+3}}(M^n)},\leqno{(*)} \] whenever \(z\in\mathbb C\) such that \(\Im\left(\sqrt{z}\right)\geq\delta\) (Theorem 1). The proof of \((*)\) is essentially based on the following apparatus: construction of the Hadamard parametrix on \(M^n\), which is given in term of an integral operator, covering \(M^n\) by geodesics balls of identical radius [the injectivity radius], using a suitable partition of unity subordinated to these balls, and using some adequate Lorentz spaces. The authors observe that the region \(\left\{z\in \mathbb C\text{ such that }\Im\left(\sqrt{z}\right)\geq \delta\right\}\) cannot be protracted for some specific manifolds (Proposition 2).