8
$\begingroup$

Suppose that $I, X_1, \ldots, X_{d-1}$ are $n \times n$ matrices with integer entries whose $\mathbb{Z}$-span is a subalgebra of $\mathrm{Mat}_n(\mathbb{Z})$. Suppose that, thought of as a subalgebra of $\mathrm{Mat}_n(\mathbb{C})$, this algebra is semisimple and non-commutative. Thus, by Wedderburn's Theorem, it is isomorphic to a direct product of complete matrix algebras $\mathrm{Mat}_r(\mathbb{C})$, with $r \ge 2$ for at least one factor.

It is possible that there exists a prime $p$ and a field $K$ of characteristic $p$ such that, regarding $I, X_1, \ldots, X_{d-1}$ as elements of $\mathrm{Mat}_n(K)$ by reduction modulo $p$, the subalgebra of $\mathrm{Mat}_n(K)$ spanned by $I, X_1, \ldots, X_{d-1}$ is commutative, and still of dimension $d$?

As a follow-up, in my specific setup, the matrices are the orbital matrices for the action of a finite group $G$ on a set $\Omega$: the orbital matrix with $1$ in position $(\alpha,\beta)$ has $1$s exactly in the positions $(\alpha g, \beta g)$, for $g \in G$. It is known that these matrices span the centralizer algebra of the permutation module (over $\mathbb{Z}$ or any field). Moreover, the algebra is commutative in characteristic zero if and only if the associated permutation character $\pi(g) = |\mathrm{Fix}_\Omega(g)|$ is multiplicity-free.

Is it possible that in this situation the centralizer algebra is non commutative in characteristic zero, but commutative after reduction modulo a prime?

$\endgroup$
0

2 Answers 2

4
$\begingroup$

Example 5.10 of

Towers, Matthew, Endomorphism algebras of transitive permutation modules for $p$-groups., Arch. Math. 92, No. 3, 215-227 (2009)

(whose author you might know) gives a positive answer to the second question. The group action involved is of $$G=\langle x,y,x\mid x^4, y^4, z^4, [x,y]=z\in Z(G)\rangle$$ (which has order $64$) acting on the cosets of the Klein four subgroup generated by $x^2$ and $y^2$,

$\endgroup$
1
  • 2
    $\begingroup$ Thank you very much Jeremy. Indeed, Matt and I overlapped as Ph.D. students of Karin Erdmann. What a nice example! I now don't feel so stupid for not being able to prove it yesterday. Somewhat embarrassingly I see I am thanked in the acknowledgements, so I really should have remembered this paper. $\endgroup$ Commented May 10, 2021 at 11:09
6
$\begingroup$

Yes. For instance, a quaternion algebra is commutative modulo 2.

$\endgroup$
5
  • $\begingroup$ The regular action of quaternions doesn't identify $\pm 1$ as scalar multiples of each other which is needed for this example $\endgroup$ Commented May 9, 2021 at 19:03
  • 1
    $\begingroup$ Benjamin Steinberg's comment points out that this example doesn't (as I wrongly claimed in a previous comment) answer the second question. E.g. in the regular action of quaternions we have distinct matrices with zero/one entries representing $1$ and $-1$ and so $ij = k$ and $ji = -k$ are also distinct matrices. $\endgroup$ Commented May 9, 2021 at 19:17
  • $\begingroup$ I must have misunderstood the question. There is a $\mathbb{Z}$-algebra spanned by $1,i,j,k$, for example a subalgebra of $\mathrm{Mat}_4(\mathbb{Z})$, which I believe answers the first question. Actually, I think I didn't even see the second part of the question. $\endgroup$ Commented May 10, 2021 at 13:06
  • $\begingroup$ @Theo Johnson-Freyd: your answer is perfect for the first question, which was probably all you could see when you answered. $\endgroup$ Commented May 10, 2021 at 13:29
  • $\begingroup$ @TheoJohnson-Freyd Benjamin's comment was replying to a previous comment that has now been deleted, not to your answer. $\endgroup$ Commented May 10, 2021 at 14:16

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.