No label defined (Q4802423)

A subalgebra \(R\) of \({\mathbb C}[x,y]\) is called a retract if there is an idempotent endomorphism \(\pi\) of \({\mathbb C}[x,y]\) (called a retraction or a projection) such that \(\pi({\mathbb C}[x,y])=R\). In other words, \(R\subseteq {\mathbb C}[x,y]\) is a retract if and only if \(\pi({\mathbb C}[x,y])=R\) for some endomorphism \(\pi\) of \({\mathbb C}[x,y]\) which acts identically on \(R\). There are several equivalent descriptions of the retracts of \({\mathbb C}[x,y]\): (i) \({\mathbb C}[x,y]=R\oplus I\) for some ideal \(I\) of \({\mathbb C}[x,y]\); (ii) \({\mathbb C}[x,y]\) is a projective extension of \(R\) in the category of \({\mathbb C}\)-algebras; (iii) by a theorem of \textit{L. J. Costa} [J. Algebra 44, 492-502 (1977; Zbl 0352.13008)], every proper retract of \({\mathbb C}[x,y]\) (i.e. different from \({\mathbb C}[x,y]\) and \(\mathbb C\)) is of the form \({\mathbb C}[p]\) for some \(p(x,y)\in{\mathbb C}[x,y]\). By \textit{V. Shpilrain} and \textit{Jie-Tai Yu} [Trans. Am. Math. Soc. 352, 477-484 (2000; Zbl 0944.13011)] there exists an automorphism of \({\mathbb C}[x,y]\) which brings \(p(x,y)\) to the form \(x+yq(x,y)\), \(q(x,y)\in {\mathbb C}[x,y]\), and every polynomial of the form \(x+yq(x,y)\) generates a proper retract of \({\mathbb C}[x,y]\). (iv) By a theorem in the same paper by Shpilrain and Yu, \(p(x,y)\) generates a retract of \({\mathbb C}[x,y]\) if and only if there is an endomorphism of \({\mathbb C}[x,y]\) which takes \(p(x,y)\) to \(x\). The purpose of this note is to give one more characterization of retracts of \({\mathbb C}[x,y]\) and to remove the very essential requirement for idempotency in the definition of retracts. The main result is the following theorem: The polynomial \(p(x,y)\in {\mathbb C}[x,y]\) belongs to a proper retract of \({\mathbb C}[x,y]\) if and only if \(p(x,y)\) is fixed by some endomorphism of \({\mathbb C}[x,y]\) with non-trivial kernel. The polynomial \(p(x,y)\) is called a test polynomial if every endomorphism of \({\mathbb C}[x,y]\) fixing it is an automorphism. The test elements do not belong to any proper retract and there is a conjecture that any element which is not in a proper retract is a test element. Let us generalize the notion of test polynomials in the following obvious way. If \({\mathcal A}\subseteq {\mathcal B}\subseteq\text{ End}({\mathbb C}[x,y])\), then we say that \(p(x,y)\in{\mathbb C}[x,y]\) is a test polynomial which recognizes the endomorphisms from \({\mathcal A}\) in the class \({\mathcal B}\), if every endomorphism \(\phi\in{\mathcal B}\) fixing \(p(x,y)\) belongs to \(\mathcal A\). Then our theorem has the following straightforward corollary: A polynomial of positive degree \(p(x,y)\in{\mathbb C}[x,y]\) does not belong to a proper retract if and only if it recognizes monomorphisms in the class of all endomorphisms.