Degree estimate for commutators. (Q1048945)

Let \(K[X]\) and \(K(X)\) stand for the polynomial algebra, respectively, the free associative algebra over a field \(K\) of characteristic 0, \(X=\{x_1,\dots,x_d\}\). If \(f\), \(g\) are two polynomials (in either one of these two algebras), one considers the subalgebra generated by \(f\) and \(g\). A natural question one may ask is to determine the least degree of a nonconstant element of this subalgebra. The answer to such a question is well known whenever \(f\) and \(g\) are algebraically dependent. In the case of the polynomial algebra the above subalgebra is of the form \(K[h]\); in the case of \(K(X)\) \(f\) and \(g\) must commute, and they have the same centralizer which is a subalgebra \(K[h]\) of \(K(X)\). It is not difficult to see that one cannot obtain a good lower bound of the least degree of the elements of our subalgebra. The authors of the paper under review formulate, as a main problem, the following. Suppose \(f\) and \(g\) are algebraically independent but their homogeneous components of maximal degrees are algebraically dependent. Suppose further \(\deg f\) does not divide \(\deg g\) nor \(\deg g\) divides \(\deg f\). Under these assumptions, find an estimate for the least degree of nonconstant elements in the subalgebra generated by \(f\) and \(g\). In the paper under review the authors construct examples showing that two conjectures, one due to \textit{J.-T. Yu} [in Degree estimate for subalgebras and automorphisms of free algebras, Proc. Fourth Int. Congress Chinese Mathematicians., Hangzhou, China 2007, Higher Educ. Press, Beijing, 2008, 359-366; Internat. Press, Boston, MA, (2009)]; and the other to \textit{L. Makar-Limanov} and \textit{J.-T. Yu} [J. Eur. Math. Soc. (JEMS) 10, No. 2, 533-541 (2008; Zbl 1137.16029)], fail. These examples are based on the following approach. Let \(u\) be a monomial in \(K(X)\) which is not a proper power. Consider \(K(X)\) as a bimodule over \(K[u]\): it turns out to split into three types of bimodules. These are first \(K[u]\), then bimodules generated by a single monomial, and finally two-generated bimodules with a nontrivial relation (the latter bimodules do appear). Afterwards the authors solve the equation \([u^m,r]=[u^n,s]\) where \(r\) and \(s\) are the unknown elements of \(K(X)\). The paper is well written and quite ``reader-friendly''. The authors give all necessary background together with an extensive survey on the existing results in the area. The paper may be considered as another proof that the combinatorial and computational aspects of free algebras is a vast area of research. Moreover the authors state a couple of interesting open problems.