On the variety of automorphisms of the affine plane (Q1372668)

The main subject of this study is \(GA_{2,n}\), the variety of automorphisms of the affine plane of degree bounded by a positive integer \(n\). After detailing some definitions and notations in section 1, we give in section 2 an algorithm to decide whether an endomorphism of the affine plane over an integral domain is a tame automorphism. Then, by applying this algorithm to the Nagata automorphism, we recover easily its known results. If \(n\) is a positive integer, let \(GA_2\) be the group of automorphisms of \(\mathbb{A}^2\) and \(E_{2,n}\) be the vector space of \(\mathbb{C}\)-endomorphisms \(f\) of \(\mathbb{A}^2_\mathbb{C}\) satisfying \(\deg(f)\leq n\). Let us set \(GA_{2,n}=GA_2\cap E_{2, n}\). We set \(J_{2, n}= \{f\in E_{2,n}\), such that Jac\((f)=1\}\) and \(G_{2,n}= GA_{2,n}\cap J_{2,n}\). \(GA_{2,n}\) is an algebraic variety. We have \(GA_{2,n}\simeq \mathbb{C}^* \times G_{2, n}\), via \(f=(f_1,f_2) \mapsto(\text{Jac}(f),(f_1/ \text{Jac} (f), f_2))\), where Jac means the Jacobian variety, such that the irreducible components of \(GA_{2,n}\) are in one to one correspondence with the irreducible components of \(G_{2,n}\). Our interest in the irreducible components of \(G_{2,n}\) is motivated by the Jacobian conjecture in dimension 2 and degree \(n\) asserting that an element of \(E_{2,n}\) is invertible if and only if its Jacobian is a nonzero constant. This conjecture is equivalent to the equality \(J_{2,n}= G_{2,n}\). \textit{H. Bass, E. H. Connell} and \textit{D. Wright} note [Bull. Am. Math. Soc., New Ser. 7, 287-330 (1982; Zbl 0539.13012); remark 1.7] that to show this equality, it would suffice to show, if possible, that \(\dim J_{2,n}= \dim G_{2,n}\) and that the algebraic variety \(J_{2,n}\) is irreducible (because we have of course \(J_{2,n} \subset G_{2,n})\). We answer here negatively to this hope. Indeed, on one hand \(J_{2,n}=G_{2,n}\) when \(n\leq 100\), by \textit{T. T. Moh} [J. Reine Angew. Math. 340, 140-212 (1983; Zbl 0525.13011)] and on the other hand we show that \(G_{2,n}\) is reducible when \(n\geq 4\) (proposition 11). We see thus that the variety \(J_{2,n}\) is reducible when \(4\leq n\leq 100\). The results of section 3 lead us to believe that we can get some insight into the decomposition in irreducible components of \(GA_{2,n}\) via purely combinatorial methods. Definition: \(U_d=\{f\in GA_2\) such that \(d(f)\leq d\}\), \(V_d=\{f\in GA_2\) such that \(d(f)=d\}\), where \(d(f)\) means the multidegree. We note that by the definition of \(U_d\) and \(V_d\), their Zariski closures in \(GA_{2,\deg(d)}\) are equal. We set \(W_d=\overline U_d=\overline V_d\) (Zariski closure in \(GA_{2, \deg (d)})\). Conjecture. If \(n\) is a positive integer, then \(GA_{2,n}=\bigcup_{d\in C_n}W_d\) is the decomposition in irreducible components of \(GA_{2,n}\). We will prove this conjecture for \(n\leq 9\) by using the next lemma: If \(n\) is an integer greater than or equal to two and if \(d\in C_n\) with \(d\neq (n)\), then we have \(W_d\not\subset W_{(n)}\) and \(W_{(n)}\not\subset W_d\). The proofs are based on a precise decomposition theorem for automorphisms given in section 3 and a characterization of length one automorphisms given in section 5. Finally, in section 7, we give some details on the case \(n=4\).