On the images of generalized polynomials evaluated on matrices over an algebraically closed skew field (Q2228139)

The notion of an algebraically closed skew field \(K\) was introduced by \textit{L. Makar-Limanov} [J. Algebra 93, 117--135 (1985; Zbl 0558.16007)]. Let \(C\) be the centre of \(K\). Let \(X=\{x_1,x_2,\ldots\}\) be an infinite countable set of (non-commuting) variables, and let \(C\{X\}\) be the corresponding free associative algebra over the field \(C\). Denote by \(K\{X\}=K*_C C\{X\}\) the free product over \(C\). The elements of the algebra \(K\{X\}\) are the generalized polynomials. Let \(F(x_1,\ldots,x_n)\) be a non-constant multilinear generalized polynomial. The authors prove, in Theorem 2.2, that for every matrix algebra \(M_m(K)\) over \(K\), its image under \(F\) is the whole algebra \(M_m(K)\). They also address the case when one drops the condition on multilinearity. Indeed, it turns out that without requiring \(F\) multilinear, their Theorem 2.2 is no longer valid as stated above. The situation is not hopeless though: the authors prove in their Theorem 3.3 that even when \(F\) is not assumed to be multilinear, the diagonal matrices are always in the image of \(F\). Moreover they prove that the image of \(F\) always contains upper triangular (and lower triangular) non-diagonal matrices. As a consequence they get that in fact \(M_m(K)\) is the span of the values of \(F\). Moreover every element of \(M_m(K)\) can be obtained as a sum of at most three elements in the image of \(F\). At the end of the paper the authors pose an interesting problem: Is it possible to obtain every element of \(M_m(K)\) as a sum of at most two images of \(F\)?