Asymptotic behavior of the Bergman kernel and associated invariants in certain inænite type pseudoconvex domains (Q3150737)

The authors study the asymptotic behavior of the Bergman kernel and metric and of the holomorphic sectional curvature near an isolated infinite type point in the boundary of a pseudoconvex domain \(G\) in \(\mathbb C^2\) (here it is known that this behavior depends only on the local situation), of the following special type : for some neighborhood \(U\) of \((0,0)\), \(G\cap U = \{ (z_1,z_2) \in U : \Re z_1 + \varphi (|z_2|^2)<0 \}\), where \(\varphi(x)\) behaves like \(\exp(-|x|^{-m})\). To give a sample of the results, we need some notation. Let \(q(t):=(q_1(t),q_2(t))\) tending to \((0,0)\) as \(t \to 0\), \(d(t)\) the shortest distance from \(q(t)\) to \(\partial G\), \(d^*(t)\) the distance from \(q(t)\) to \(\partial G\) along the unique complex-tangential direction to \(\partial G\) at the projection of \(q(t)\) on \(\partial G\), and \(\lambda(t)\) the distance between the point \(q(t)\) and the \(\Re u_1\)-axis (this definition is lacking in the paper, I thank the authors for providing it). Define \(\tau_0 (t) = d^*(t)/|\lambda(t)|\). Let \(K_G (z,w)\) stand for the Bergman kernel of the domain \(G\). Theorem. 1. If \(\lim_{t\to 0} \tau_0 (t)\) exists and lies in \((0, \infty]\) (note that this condition is much wider than the nontangential approach), then there is a constant \(C\) depending on this limit so that \[ \lim_{t\to 0} d(t)^2 d^*(t)^2 K_G(q(t),q(t)) = C. \] 2. If \(\lim_{t\to 0} \tau_0 (t)=0\) and: (a) \(\lim_{t\to 0} \tau_0 (t)|\lambda(t)|^{-2m}=\infty\), or (b) \(\lim_{t\to 0} \tau_0 (t)|\lambda(t)|^{-2m} \in (0,\infty)\), or (c) \(\lim_{t\to 0} \tau_0 (t)|\lambda(t)|^{-2m} =0\), then the same result holds for the Bergman kernel (with different constants, given in terms of the hypotheses). There hold, however, differentiated results for the metric and curvature, which we skip because they would require the definition of a basis of vector fields, and thus be longer to state. The proofs rely on the Bergman-Fuchs formulae and sequences of well-chosen renormalizations (scaling, centering, further localization, and final scaling), so that the eventual result depends on the (unbounded) limit domains that arise out of those renormalizations. In case 1, this domain is an arbitrarily small perturbation (as \(t\to 0\)) of \(\mathbb H \times D\), where \(\mathbb H\) is the right half plane and \(D\) a disc, in case 2a an arbitrarily small perturbation of \(\mathbb H \times \mathbb H \), in case 2b an arbitrarily small perturbation of a domain given by a defining function of the type \(\rho(u_1,u_2)= \psi_1 (\text{Re} u_1,\text{Re} u_2) + \exp \psi_2(\text{Re} u_2)\), where \(\psi_1\) and \(\psi_2\) are affine and real-valued, in case 2c an arbitrarily small perturbation of an unbounded domain biholomorphic to the unit ball. The computations are intricate; beware of a few unfortunate typos. Note that in fact \(\beta=p_2\), \(b=q_2\) on pages 786 ff.