The \(\overline\partial\)-cohomology groups, holomorphic Morse inequalities, and finite type conditions (Q612271)

The authors study the spectral behavior of the complex Laplacian on forms with values in the \(k\)-th tensor power of a holomorphic line bundle over a smoothly bounded domain with degenerated boundary (i.e., with degenerated Levi form) in a complex manifold. More precisely, let \(\Omega\subset\subset X\) be a domain with smooth boundary in a complex manifold of dimension \(n\), and let \(E\) be a holomorphic line bundle over \(\Omega\) that extends smoothly to \(b\Omega\). Let \(h^q(\Omega,E)\) be the dimension of the Dolbeault cohomology group on \(\Omega\) for \((0,q)\)-forms with values in \(E\), and let \(\tilde{h^q}(\Omega,E)\) be the dimension of the corresponding \(L^2\)-cohomology group for the \(\overline{\partial}\)-operator. It was proved by Hörmander that when \(b\Omega\) satisfies the conditions \(a_q\) and \(a_{q+1}\) for the Levi form, then these two cohomology groups are isomorphic. Furthermore, there exists a defining function \(r\) of \(\Omega\) and a constant \(c_0>0\), independent of \(E\), such that these cohomology groups are in turn isomorphic to their counterparts on \[ \Omega_c=\{z\in\Omega: r(z)<-c\} \] for all \(c\in (0,c_0)\). Combining Hörmander's theorem with a result of Diederich-Fornæss, the authors observe: Theorem 1.1. Let \(\Omega\subset\subset X\) be a smooth pseudoconvex domain. Assume that there exists a neighborhood \(U\) of \(b\Omega\) and a bounded continuous function whose complex Hessian is bounded from below by a positive constant on \(U\cap\Omega\). Then \(h^q(\Omega,E)=\tilde{h^q}(\Omega,E)\) for all \(1\leq q\leq n\). Furthermore, there exists a defining function \(r\) of \(\Omega\) and a constant \(c_0>0\), independent of \(E\), such that \(b\Omega_c\) is strictly pseudoconvex and \(\tilde{h^q}(\Omega,E)=\tilde{h^q}(\Omega_c,E)\) for all \(c\in (0,c_0)\). It follows from the work of Catlin that smooth pseudoconvex domains of finite type satisfy the assumptions above. Together with Berman's local version of holomorphic Morse inequalities, the authors deduce that, for such domains, \(h^q(\Omega,E^k) \leq C\cdot k^n\) for some constant \(C>0\). For pseudoconcave domains, the authors show: Theorem 1.2. Assume that \(\Omega\subset\subset X\) is pseudoconcave and \(b\Omega\) does not contain the germ of any complex hypersurface. Then \(h^0(\Omega,E^k)\leq C\cdot k^n\) for some constant \(C>0\). The above result shows that \(h^q(\Omega,E^k)\) is insensitive to the order of degeneracy of the Levi form on the boundary. This is related to the fact that the dimension of the cohomology groups (i.e., the multiplicity of the zero eigenvalues of the \(\overline{\partial}\)-Neumann Laplacian) alone, even though they can determine pseudoconvexity, are not sufficient to detect other geometric features, such as the finite type conditions of the boundary. For this, one needs to consider higher eigenvalues as well. Let \(N_k(\lambda)\) be the number of eigenvalues of the \(\overline{\partial}\)-Neumann Laplacian on \(\Omega\) for \((0,1)\)-forms with values in \(E^k\) that are less or equal to \(\lambda\). The following is the main theorem of the present paper: Theorem 1.3. Let \(\Omega\subset\subset X\) be a pseudoconvex domain with smooth boundary in a complex surface \(X\). Let \(E\) be a holomorphic line bundle over \(\Omega\) that extends smoothly to \(b\Omega\). Then \(b\Omega\) is of finite type if and only if \(N_k(C\cdot k)\) has at most polynomial growth as \(k\rightarrow\infty\) for any \(C>0\). The proof of Theorem 1.3 is a modification of an earlier work of the first author. One (only) needs to establish here that effects of the curvature of the basic metric on the complex surface \(X\) and the fiber metric of the line bundle \(E^k\) are negligible.