3

I am stuck trying to prove the following corollary:

Let $f = X^n -1 \in \mathbb{F}_q[X]$ with $gcd(q,n)=1$. Let k = order of q mod n. Then the degree of every irreducible factor of $f$ divides k.

It is supposed to be a relatively straightforward consequence of the following fact:

Let $q$ be a prime power and $n$ such that $gcd(n,q)=1$. Let $e$ be the order of the congruence class of $q$ in $(\mathbb{Z}/ n \mathbb{Z})^*$. Then the splitting field K of $X^n-1 \in \mathbb{F}_q[X]$ is isomorphic to $\mathbb{F}_{q^e}$.

I unfortunately fail to see how I can apply the second statement to show the first.

  • 1
    This is a little outside my comfort zone (but what isn't? ), but there's something called the tower law, for field extensions. – suckling pig Jun 20 '24 at 21:55
  • By the Galois correspondence (if it applies), this is also about subgroups, and Lagrange. – suckling pig Jun 20 '24 at 21:57
  • I have walked around this theme on our beloved site more often than I care to admit. 1, 2, 3, 4. and then some. Different guises appear, but the technique is always to play with the simple Galois theory of finite fields (Hail, Frobenius!) and appropriate roots of unity. – Jyrki Lahtonen Jun 22 '24 at 03:45
  • (cont'd) I have no less than three applicable dupehammers, but I am reluctant to vote to close this as a duplicate, because I hate doing that when I have answered the target question(s) myself. Also, I think it is harsh to expect relative newbies (including every one who has posted in this thread, +1 to all) to be aware of past appearances. After all, this isn't a run-of-the-mill calculus or congruence question. I do expect more from old-timers :-) – Jyrki Lahtonen Jun 22 '24 at 03:48

2 Answers2

2

If $f$ is an irreducible factor of $t^n-1$ over $\mathsf k=\mathbb F_q$, then if $K$ is the splitting field of $g(t)=t^n-1$, $f$ splits in $K$, and hence if we take the subfield generated by one of its roots, we get an intermediate field $E\cong \mathsf k[t]/\langle f\rangle$ and so

$$ e=[K:\mathsf k] = [K:E][E:\mathsf k] = [K:E]\deg(f), $$ so that $\deg(f)\mid e$.

It is also not too hard to derive the result from scratch:

If $K$ is the splitting field of $t^n-1$ then $K$ is a finite extension of $\mathsf k$ and so is $\mathbb F_{q^m}$ for some integer $m$. But now $\mathbb F_{q^m}^{\times}$ is a (finite subgroup of) the multiplicative group of a field, it is cyclic. But then since $t^n-1$ splits in $K$ it forms a subgroup of $K^\times$ of order $n$ and hence we must have $n\mid q^m-1$ so that $q^m\cong 1 \mod n$ so that $e\mid m$ where $e$ is the order of $q$ in $(\mathbb Z/n\mathbb Z)^\times$. But the same argument shows that if $e$ is the order of $q$ in $(\mathbb Z/n\mathbb Z)^\times$ then $t^n-1$ must split in $\mathbb F_{q^e}$ so that $K=\mathbb F_{q^e}$.

Proof that any finite subgroup of the multiplicative group of a field is cyclic:

Let $G_n$ be a subgroup of the multiplicative group of a field $K$ with $|G_n|=n$. Any subgroup $H$ of $G_n$ has, by Lagrange's theorem, order a divisor, $d$ say, of $n$, and its elements all have order dividing $d$, so that, since a polynomial has at most its degree many roots over a field, $H = \{x \in K: x^d=1\}=G_d$. It follows that $G_n$ has at most one subgroup $G_d$ of order $d$ for each divisor $d$ of $n$.

But if $\varphi$ denotes the Euler $\varphi$-function, then it is easy to see that if $G_n$ contains an element $\zeta$ of order $d$, then $\zeta\in G_d$ and $G_d$ is therefore cyclic, so that it contains exactly $\varphi(d)$ elements of order $d$. Thus if we set $c_d =1$ if $G_n$ contains an element of order $d$ and $c_d=0$ otherwise, it follows that $$ n=|G_n| = \sum_{d\mid n} c_d\varphi(d)\leq \sum_{d\mid n} \varphi(d) $$ But by, for example, considering the case $K=\mathbb C$ where $c_d =1$ for all $d \mid n$ we see that $\sum_{d\mid n} \varphi(d) =n$, and hence we must have $c_d=1$ for all $d \mid n$. Thus $G_n$ is a cyclic group as required.

krm2233
  • 7,230
1

I came up with a proof that uses less terminology but that may be different from what you brought up in your post [my eyes still glaze over at the words 'congruence class', I am sorry!]

Proof #1: Let $k$ be the smallest positive integer s.t. $q^k \equiv 1 \pmod n$, or equivalently, $n$ divides $q^k-1$. Let $f$ be an irreducible polynomial that divides $X^n-1$ in $\mathbb{F}_q[X]$ an let $\alpha$ be a root of $f$. Then the equation $\alpha^n = 1$ holds, and as $n$ divides $q^k-1$ it follows that $\alpha^{q^k-1} = 1$ and $\alpha^{q^k} = \alpha$. Now, let $d$ be the smallest positive integer such that the equation $\alpha^{q^d}=\alpha$ holds.

Then check that $\alpha = \alpha^{q^d} = (\alpha^{q^d})^2=(\alpha^{q^d})^3 = \cdots = (\alpha^{q^d})^{\ell} = \cdots$, and thus in particular $\alpha^{q^{\ell d}} = \alpha$ for all positive integers $\ell$. And so writing $k = \ell d +r$ [with $r \in \{0,\ldots, d-1\}$] plugging in the equation $\alpha^{q^k}=\alpha$ yields $$\alpha = \alpha^{q^k}$$ $$= \alpha^{q^{\ell d+ r}} = \alpha^{q^{\ell d}q^{r}}$$ $$= (\alpha^{q^{\ell d}})^{q^{r}} = \alpha^{q^r}.$$ Given that $d$ is the smallest positive integer such that $\alpha^{q^d}=\alpha$ and $\alpha^{q^{r}} = \alpha$ with $r \in \{0,1,\ldots, d-1\}$, it follows that $r$ must be $0$, and thus on the one hand, $d$ must divide $k$.

On the other hand, as $\alpha^{q^d}-\alpha = 0$ and $f$ is a irreducible polynomial in $\mathbb{F}_q[x]$, it follows that $f$ must divide $X^{q^d}-X$ [because they both share a common root namely $\alpha$]. However, from an elementary result in finite fields, we know the following: Let $f$ be an irreducible polynomial in $\mathbb{F}_q[x]$ that divides $x^{q^d}-x$. Then letting $c$ be the degree of $f$, the relation $c$ divides $d$ holds. So in particular, the degree $c$ of the irreducible polynomial $f \in \mathbb{F}_q[x]$ that has $\alpha$ as a root, must divide $d$.

So letting $\alpha$ be a root of $X^n-1$ in $\mathbb{F}_q[X]$, and letting $d$ be the smallest positive integer s.t. $\alpha^{q^d}-\alpha = 0$, we have established the following: On the one hand, $d$ must divide $k$. On the other hand, the degree $c$ of the irreducible polynomial $f \in \mathbb{F}_q[x]$ that has $\alpha$ as a root, divides $d$. So $c$ must divide $k$, and the result follows.


ETA: Here is another proof Proof #2, that is along the lines of what was outlined in your OP: Let $\mathbb{F}_{q^k}$ denote the field on $q^k$ elements. Note that $(\mathbb{F}_{q^k})^{\times}$ is a cyclic group on $q^k-1$ elements. As $n|q^k-1$ then it follows that there are exactly $n$ elements $\alpha \in (\mathbb{F}_{q^k})^{\times}$ s.t. $\alpha^n=1$, or equivalently, $X^n-1$ factors completely in $(\mathbb{F}_{q^k})^{\times}$. [Use the fact that a cyclic group $G$ on $N$ elements has exactly $n$ elements $g \in G$ satisfying $g^n = e$ iff $n|N$.] So now then let $f$ be an irreducible polynomial in $\mathbb{F}_q[X]$ that divides $X^n-1$. Then, as $X^n-1$ factors completely in $\mathbb{F}_{q^k}$ there is a element $\alpha \in (\mathbb{F}_{q^k})^{\times}$ s.t. $\alpha$ that satisfies $f(\alpha)=0$. Furthermore, as $f$ is irreducible in $\mathbb{F}_q[X]$ and $\alpha$ is in $\mathbb{F}_{q^k}$, it turns out that $\mathbb{F}_q(\alpha)$ is a subfield of $\mathbb{F}_{q^k}$ of cardinality $q^c$, where $c$ is the degree of $f$. We finish this proof that noting that a field of $q^k$ elements has a subfield of $q^c$ elements iff $c|k$, and so the result follows.

Mike
  • 23,484