Matrix algebras with degenerate traces and trace identities (Q2057234): Difference between revisions

A \textit{trace identity} for an algebra \(A\) over a field \(F\) is an element of the free algebra with trace that vanishes under substitutions from elements of \(A\). For example, \(2\times 2\) matrices satisfy the following identity (equivalent to the Cayley-Hamilton theorem): \[x^2-\operatorname{tr}(x)x+\frac12(\operatorname{tr}(x)^2-\operatorname{tr}(x^2)).\] A trace identity is called \emph{pure} if variables occur only inside of traces. The paper under review considers trace identities of algebras with degenerate traces, and the results are various. They first consider \(A\) equal to \(n\times n\) diagonal matrices or, equivalently, \(A=F^n,\) with trace \[(a_1,\ldots,a_n)\mapsto t_1a_1+\cdots+t_na_n\] for some fixed \(t_1,\ldots,t_n\in F\), denoting this algebra with trace as \(D^{t_1,\ldots,t_n}\). Such a trace will be degenerate if one or more of the \(t_i\) is 0. The authors' main theorem about \(D^{t_1,\ldots,t_n}\) is that all trace identities are consequences of commutativity and pure trace identities. Along the way they show how the identities of \(D^{t_1,\ldots,t_n}\) in the case of \(t_n=0\) and the remaining \(t_i\ne 0\) can be derived from the pure trace identities of \(D^{t_1,\ldots,t_{n-1}}\). They then determine generating sets of identities and codimension sequences in the cases of \(n\) equal to 2 or 3. The next algebras considered are truncated polynomial algebras, \(F\langle x\rangle/(x^{n+1})\). This algebra can be embedded into upper triangular \(n\times n\) matrices, which is the setting in which they are considered here. The most general trace function would be \[\sum_{k=0}^n a_kx^k\mapsto t_1a_1+\cdots+t_na_n\] for some fixed \(t_1,\ldots,t_k\in F\). The two cases considered are \(n=2\) and \(a_2=\cdots=a_n=0\). in each of these cases they again determine generating sets of identities and codimension sequences. In the last section the authors study algebras whose trace codimension sequences are polynomially bounded.