Some families of polynomial automorphisms (Q1886784)

Let \(G\) be the group of polynomial automorphisms of the complex plane \(\mathbb{C}^{2}.\) By the Jung-van der Kulk theorem each \(\sigma\in G\) can be represented as a composition \[ \sigma=\alpha_{1}\circ\beta_{1}\circ\ldots\alpha_{k}\circ\beta_{k}\circ \alpha_{k+1}, \] where \(\alpha_{i}\) are affine automorphisms and \(\beta_{j}\) are triangular automorphisms. If this composition is reduced i.e. \(\alpha_{i}\) (\(2\leq i\leq k\)) are not triangular and \(\beta_{j}\) are not affine automorphisms then the number \(k\) is unique and it is called the length of \(\sigma\) and denoted \(l(\sigma).\) The multidegree \(d(\sigma)\) of \(\sigma\) is defined as \[ d(\sigma)=(\deg\beta_{1},\ldots,\deg\beta_{k}). \] It is a finite sequence of integers \(\geq2.\) Let us denote by \(G_{d}\) the set of automorphisms whose multidegree is \(d=(d_{1},\ldots,d_{k}),\) \(d_{i}\geq2.\) In the Zariski topology of \(G\) (it is an infinite dimensional algebraic variety) the length \(l(\sigma)\) is a lower semicontinuous function on. It gave rise to the question whether for any \(G_{d}\) there exists a subset \(E(d)\) of multidegrees such that \[ \overline{G_{d}}= {\displaystyle\bigcup\limits_{e\in E(d)}} G_{e}. \] The authors prove that this conjecture is not true. In particulary they show that \[ G_{(19)}\cap\overline{G_{(11,3,3)}}\neq\emptyset\quad \text{and}\quad G_{(19)}\not \subset \overline{G_{(11,3,3)}}. \]