Graded monomial identities and almost non-degenerate gradings on matrices (Q2676746)

Let \(F\) be a field of characteristic zero, \(G\) be a group and \(M_n(F)\) be the algebra of matrices of size \(n\) with entries from \(F\) with a \(G\)-grading. \textit{Y. Bahturin} and \textit{V. Drensky} [Linear Algebra Appl. 357, No. 1--3, 15--34 (2002; Zbl 1019.16011)] proved that if the \(G\) grading on \(M_n(F)\) is elementary and the neutral component of \(M_n(F)\) is commutative, then the graded identities of \(M_n(F)\) follow from three basic types of identities and monomial identities of length \(\ge 2\) bounded by a function \(f(n)\) of \(n\). In this paper, the authors prove that the best upper bound is \(f(n) = n\). More generally, they prove that all the graded monomial identities of an elementary \(G\)-grading on \(M_n(F)\) follow from those of degree at most \(n\). The authors also study gradings which satisfy no graded multilinear monomial identities but the trivial ones, which they call \textit{almost non-degenerate} gradings. The description of nondegenerate elementary gradings on matrix algebras is reduced to the description of non-degenerate elementary gradings on matrix algebras that have commutative neutral component. The authors provide necessary conditions so that the grading on \(M_n(F)\) is almost non-degenerate and apply the results on monomial identities to describe all almost non-degenerate \(\mathbb Z\)- gradings on \(M_n(F)\) for \(n\le 5\). The paper looks special, but in the paper [\textit{A. R. Kemer}, Izv. Vyssh. Uchebn. Zaved., Mat. 1989, No. 6(325), 71--76 (1989; Zbl 0678.16013)] the graded technique is used, in fact the problems reduce somehow to the elementary graded case. I personally feel that results of such kind can be used for the classical non-graded case.