Limits of tame automorphisms of \(k[x_ 1,\dots ,x_ N]\) (Q790891): Difference between revisions

Let k be a field of characteristic 0 and let \(S=k[x_ 1,...,x_ N]\). End S denotes the ring of all endomorphisms of S which fix k pointwise. Aut S is the group of units of End S. End S and \(S^ N\) are identified via \(f\leftrightarrow(f(x_ 1),...,f(x_ N)).\) It is well-known that Aut S is contained in \(J=\{f\in End S:\quad \det [\partial f(x_ i)/\partial x_ j]\) is a non-zero field elemen\(t\}\). The group G of tame automorphisms is generated by all automorphisms f for which each \(f(x_ i)\) is linear and all automorphisms f with \(f(x_ 1)=x_ 1+g(x_ 2,...,x_ N),\) \(g\in S\), \(f(x_ i)=x_ i\) for \(i=2,...,N\). Two major questions in this context are whether \(Aut S=J\) (Jacobian problem) and whether \(G=Aut S\) (tame automorphism problem). The author looks at \(End S=S^ N\) as a ring with componentwise addition and multiplication and introduces the ''formal power series topology'' on End S in which two elements \(f,g\in End S\) are ''close'' if, for all i, \(f(x_ i)\) and \(g(x_ i)\) agree in enough lower degree homogeneous components. Aut S, J, G, and other subsets of End S are furnished with the induced topology. The author proves that J is closed in End S and that G is dense in J. This theorem transforms the Jacobian problem into the question whether G is closed in Aut S, and the tame automorphism problem into the question whether limits of invertible endomorphisms are invertible. The author actually proves that a subgroup of G (proper at least for \(N=2)\) is already dense in J, and furthermore obtains results on the structure of G too technical to reproduce here.