Estimates for invariant metrics near non-semipositive boundary points (Q1948725)

Let \(D\subset \mathbb C^n\) denote a domain, and \(S_D,A_D\) and \(K_D\) the differential metrics of Sibony, Azukawa, and Kobayashi, respectively. For a point \(z \in D\) the indicatrix of a differential metric \(F_D\) on \(D\) is defined by \[ I_zF_D:=\Big\{v \in T_z^{\mathbb C}D\,\,\Big|\,\, F_D(z,v)<1\Big\}. \] The Kobayashi-Buseman metric \(\widehat K_D\) is the invariant metric whose indicatrices are the convex hulls of the indicatrices of \(K_D\). The authors introduce also the new differential metric \(\widetilde K_D\) as the largest invariant metric all of whose indicatrices are pseudoconvex. For each point \(z \in D\) the indicatrix \(I_z\widetilde K_D\) is the envelope of holomorphy for \(I_z K_D\). Assume now that \(a \in \partial D\) and \(\partial D\) is \(C^2\)-smooth near \(a\). The point \(a\) is called semipositive if the Leviform of \( \partial D\) has only nonnegative eigenvalues on \(T_a^{\mathbb C}\partial D\). Otherwise it is called non-semipositive. By \(\nu_a\) we denote the outward unit normal vector to \(\partial D\) at \(a\). By results of \textit{S. G. Krantz} [Rocky Mt. J. Math. 22, No. 1, 227--233 (1992; Zbl 0760.32010)] and of \textit{J. E. Fornaess} and \textit{L. Lee} [Complex Var. Elliptic Equ. 54, No. 3--4, 293--301 (2009; Zbl 1166.32010)] it is known that for a non-semipositive boundary point \(a\) of \(D\) for small \(t>0\) one has \( K_D( a-t \nu_a; \nu_a ) \asymp t^{-3/4}\) and \(S_D( a-t \nu_a; \nu_a ) \asymp t^{-1/2} \). In the present paper the authors obtain the following generalization of this: Proposition 1. For a non-semipositive boundary point \(a\) of \(D\) and sufficiently small \(t>0\) one has \[ F_D(a-t \nu_a; X) \asymp \frac{|\langle\nabla d (a-t \nu_a),X\rangle|}{\sqrt{t}} +|X| \,\,, \] where \(F_D\) denotes one of the metrics \(S_D,A_D, \widehat K_D\), and \(\widetilde K_D\). Here by \(d\) the boundary distance function is denoted. Proposition 2. (i) In the case that the boundary of \(D\) is only \(C^{1,\varepsilon}\) - smooth near \(a\), then \[ \displaystyle{ S_D(z;X) \gtrsim \frac{|\langle \nu_{a'} , X\rangle |}{d(z)^{1-\frac{1}{1+\varepsilon} } } +|X| } \] for any \(X\in \mathbb C^n \) and any point \(z \in D\) near \(a\), where \(a'\) is the orthogonal projection of \(z\) to \(\partial D\). \smallskip (ii) In the case that the boundary of \(D\) is \(C^{2,\varepsilon}\) - smooth near \(a\) and \(a\) is semipositive, then \[ \displaystyle{S_D(a-t \nu_a ;X) \gtrsim \frac{|\langle \nabla d(a-t \nu_a) , X\rangle |}{t^{1-\frac{1}{2+\varepsilon} } } +|X| } \] for any \(X\in \mathbb C^n \) and sufficiently small \(t>0\). The article concludes with a section where some properties of \(\widetilde K_D\) are discussed.