2
$\begingroup$

I asked this question in MSE few days ago but there was no response.

Suppose $\mathbb{F}=\mathbb{F}_{q^2}$, where $q$ is a prime power. The conjugate of elements in $\mathbb{F}$ is defined by $\overline{x}=x^q$. I need to find the number of $n\times n$ unitary circulant matrices over $\mathbb{F}$.

The number of invertible circulant matrices over a finite field can be seen elsewhere, such as when $n,q$ coprime and my question when $n=\operatorname{char} q$.

Is there any better method to calculate this number other than considering each entry?

This is equivalent to the order of the centraliser of the permutation matrix of $(1,2,\dots,n)$ in $\operatorname{GU}_n(q)$.


Added on 30 May 2020 MSE:

Let $C$ be the subgroup of $\operatorname{GL}_n(q^2)$ of all circulant matrices. Is $C\operatorname{GU}_n(q)$ a subgroup of $\operatorname{GL}_n(q^2)$? That is, is $C\operatorname{GU}_n(q)=\operatorname{GU}_n(q)C$? If that is correct then $C\operatorname{GU}_n(q)=\operatorname{GL}_n(q^2)$ and so $|C\cap\operatorname{GU}_n(q)|$ follows. Here we denote by $\operatorname{GU}_n(q)$ the general unitary group over $\mathbb{F}_{q^2}$.

$\endgroup$
0
4
$\begingroup$

Let $\tau$ denote the permutation matrix corresponding to $(1,2,\ldots,n)$. Consider it first as an element of $\mathrm{M}_n(q^2)$.

This matrix has minimal polynomial equal to $X^n-1$, which is equal to its characteristic polynomial. It is therefore cyclic, and its centralizer is isomorphic to $\mathbb{F}_{q^2}$-algebra $\mathbb{F}_{q^2}[X]/(X^n-1)$. For simplicity, I'll assume $\mathbb{F}_{q^2}$ has no $N$-th root of unity, so that this algebra is isomorphic to $\mathbb{F}_{q^{2n}}$. Mapping $X\to \bar{X}^t$ defines a field automorphism of order $2$ of $\mathbb{F}_{q^{2n}}$ which, by Hilbert 90 (which is an overkill, but does the job), restricts to a surjective map $\mathbb{F}_{q^{2n}}^\times\to \mathbb{F}_{q^n}^\times$. The centralizer you're seeking is precisely the kernel of this map, and is of carinality $$ \frac{q^{2n}-1}{q^n-1}=q^n+1.$$

Does this make sense?

Now, if $X^N-1$ splits in $\mathbb{F}_{q^2}$, then $\mathbb{F}_{q^2}[X]/(X^N-1)$ is a product of finite fields, and the same type of argument works per coordinate, which should result in a formula of the form $\prod_{i=1}^r q^{d_i}+1$ for suitable $d_i$'s such that $\sum d_i=n$.

$\endgroup$
1
  • $\begingroup$ Thank you for your answer. But what if $(n,q)\ne 1$ or in particular $n=\operatorname{char}q$? $\endgroup$ – Hongyi Huang Jun 2 '20 at 6:54

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.