Skip to main content
1 of 3
Andreas Thom
  • 25.5k
  • 4
  • 82
  • 142

Matrix identities in two variables

The famous Amitsur-Levitzki-Theorem states that the algebra $M_n(\mathbb C)$ satisfies no polynomial identity of degree less than $2n$ and it satisfies $$p(x_1,\dots,x_{2n}) = \sum_{\sigma \in \Sigma(2n)} {\rm sign} \cdot (\sigma) x_{\sigma(1)} \cdots x_{\sigma(2n)},$$ a polynomial in $2n$ non-commuting variables.

Question: What is the smallest degree of a non-trivial polynomial identity of $M_n(\mathbb C)$ in two variables?

By inserting non-commutative monomials in two variables into the Amitsur-Levitzki polynomial one gets an upper abound of the degree of $$2n \lceil \log_2(2n) \rceil.$$ Is $O(n \log(n))$ optimal? The question might be approachable for low $n$, but I am more interested in asymptotic properties.

Andreas Thom
  • 25.5k
  • 4
  • 82
  • 142