Precise subelliptic estimates for a class of special domains (Q464290): Difference between revisions

It is by now classical that subelliptic estimates for the \(\overline{\partial}\)-Neumann problem hold on a domain if and only if the boundary is of finite (D'Angelo) type (work of Catlin and D'Angelo in the eighties). However, it is still not understood how to obtain the ``optimal'' subelliptic gain (say near a point) in terms of boundary data. The paper under review represents a contribution to this important question. The domains under consideration are the so-called regular coordinate domains, defined by \(2\mathrm{Re}\, z_{n+1} + \sum_{j=1}^{j=n}|f_{j}(z)|^{2} < 0\;,\;(z,z_{n+1}) \in U\times\mathbb{C}\), where \(U\) is a neighborhood of \(0 \in \mathbb{C}^{n}\) and the \(f_{j}\)s are holomorphic functions on \(U\) satisfying \(f_{j}(z) = f_{j}(z_{1}, \dots, z_{j})\), \(f_{j}(0)=0\), but \(f_{j}(0,\dots,0,z_{j})\) not identically vanishing. Denote by \(m_{j}\) the smallest integer such that \(\partial_{z_{j}}^{m_{j}}(0)\neq 0\), and set \(m:= m_{1} \cdots m_{n}\). If \(D\) denotes the (D'Angelo) type of the boundary point \(0\), then by Catlin's results, the subelliptic gain \(\varepsilon\) at \(0\) satisfies \(\varepsilon \leq 1/D\). D'Angelo conjectured that \(\varepsilon\geq 1/2m\) (note that \(D\leq 2m\)). The authors give a positive answer to this conjecture in their Theorem 2.1. This result was also obtained independently by \textit{D. W. Catlin} and \textit{J.-S. Cho} [``Sharp estimates for the \(\bar{\partial}\)-Neumann problem on regular coordinate domains'', \url{arXiv:0811.0830}]. Both proofs construct (families of) plurisubharmonic functions whose complex Hessians blow up like \((1/\delta(z))^{m}\) near the boundary, where \(\delta(z)\) denotes the boundary distance (roughly speaking). The construction of these families of plurisubharmonic functions appears to be simpler in the present paper. Moreover, the authors improve on this result. They define a number \(\gamma\), taking into account the vanishing of the \(f_{j}\)s in different directions, such that always \(\gamma \geq 1/m\) and in many cases \(\gamma > 1/m\). They then prove that there is a subelliptic estimate at \(0\) with \(\varepsilon = \gamma/2\) (Theorem 2.4). There are families of examples where \(\gamma/2 = 1/D\); in these examples the subelliptic gain is thus optimal.