A remark on the Kim's theorem (Q991554)

The author has detected an error in a paper by \textit{K.-T. Kim} [Trans. Am. Math. Soc. 319, No.~1, 139--153 (1990; Zbl 0705.32008)] and provides a correct proof for the main result there along the lines of the original one but with somewhat more sophisticated arguments. Actually Kim's proof is based on the following proposition concerning scaling sequences in domains satisfying the so-called (L) condition. Let \(p\) be a boundary point of a domain \(\Omega\subset\mathbb C^n\) such that \(\partial\Omega\) is real-analytic near \(p\) and is of finite type \(2k\) at \(p\) and there is an open neighborhood \(U\) of \(p\) along with a biholomorphic map \(F:U\to\mathbb C^n\) with convex image for \(U\cap\Omega\). Suppose \(p_0\in\Omega\) and \(\{ g_j\} \subset \text{Aut}(\Omega)\) is a sequence with \(\lim_{j} g_j(p_0) =p\). Then the maps \(\sigma_j(z):= \big[ dg_j(z) \big]^{-1} \big( g_j(z)-p\big)\) admit a subsequence converging to a biholomorphic mapping \(\Omega\to\mathbb C^n\). The author's counterexample to this statement is the unit disc \(\Omega:=\Delta=(| z|<1)\) with \(p:=1\) and \(g_j(z):= (z+a_j)/(1+\overline{a_j}z)\) where \(a_j:=1+i/j+1/j^2\). The corresponding scaling sequence in Kim's proposition is \(\sigma_j(z)=\big(a_j-1+z(1-\overline{a_j}) \big) \big/\big( (1-| a_j|^2 (1-\overline{a_j}z)\big)\) which is easily seen to be not normal.