2

I'm trying to understand the proof of the following problem:

Let $L / K$ be a field extension, and assume that $K$ has $q=p^{r}$ elements. Let $P(X)$ be a irreducible polynomial in $K[X]$. Let $M$ be a subfield of $K(\alpha)$ for $\alpha$ a root of $P(X)$, and suppose that the degree of $K(\alpha)$ over $M$ is $r .$ Show that $P(X)$ factors, in $M[X],$ as a product of $\frac{d}{r}$ irreducible polynomials of degree $r$.

Proof: We have the previous results: (a) $P\left(\alpha^{q}\right)=P(\alpha)^{q}$ for all $\alpha \in L$, (b) If $P(X)$ is irreducible, and has a root $\alpha$ in $L$, then $P(X)$ factors in $L$ as $(X-\alpha)\left(X-\alpha^{q}\right)\left(X-\alpha^{q^{2}}\right) \ldots\left(X-\alpha^{q^{d-1}}\right),$ where $d$ is the degree of $P(X)$. Now, let $Q(X)$ be an irreducible element of $M[X]$ that divides $P(X) .$ Then, since $P(X)$ factors into linear factors in $K(\alpha), Q(X)$ has a root in $K(\alpha) ;$ by (b) this root is of the form $\alpha^{q^{s}}$ for some $s$. Applying (b) with $M$ in place of $K$ and $Q(X)$ in place of $P(X)$ (and noting that $M$ has $q^{\frac{d}{r}}$ elements) we see that the other roots are $\left(\alpha^{q^{s}}\right)^{q^{k} \frac{d}{r}},$ for $0 \leq k \leq r-1,$ so $Q(X)$ has degree $r$ as claimed.

This proof feels a bit too quick for me. Could anyone explain the line "Applying (b) with $M$ in place of $K$ and $Q(X)$ in place of $P(X)$ (and noting that $M$ has $q^{\frac{d}{r}}$ elements) we see that the other roots are $\left(\alpha^{q^{s}}\right)^{q^{k} \frac{d}{r}},$ for $0 \leq k \leq r-1,$ so $Q(X)$ has degree $r$ as claimed?" More precisely,

(1) Why is $0 \le k \le r-1$, and why are the $\left(\alpha^{q^{s}}\right)^{q^{k} \frac{d}{r}}$ distinct for distinct $k$?

(2) How do we conclude $Q(X)$ has degree $r$?

Partial answers (answering only some of the questions) are very welcome! (I will upvote them too)

By the way, this is a problem from a basic algebra course. I do not know Galois Theory.

Albert
  • 1,609
  • We have already covered this question from many angles on our site. Not sure whether your exact question is covered, because many of the earlier answers did rely on basic Galois theory of finite fields. I tried to do without. If we locate a good duplicate I will delete/CW my answer according how close a match is found. – Jyrki Lahtonen Jan 09 '21 at 11:35
  • @JyrkiLahtonen Thanks for your comment. I was trying to understand the exact sequence of logic that the above proof follows. (For example, if using (b) like the above proof does, then we get that $0 \le k \le deg(Q)-1$, not $0 \le k \le r-1$. Putting $r$ there seemed to me that the degree of $Q$ was being assumed to be $r$, which is circular logic.) – Albert Jan 09 '21 at 11:39
  • 1
    Ah! I see what the concern is (+1). I will revise my answer. – Jyrki Lahtonen Jan 09 '21 at 11:42
  • The revision may address this concern. I assume that (a) and (b) also contain the fact that the elements $\alpha^{q^i}$, $i=0,1,\ldots,d-1$, are distinct. – Jyrki Lahtonen Jan 09 '21 at 12:13
  • The closest hits I could find include 1 and 2. Both are phrased in a more general setting of Galois theory. I guess it is ok (at least for now), to leave an answer specific to finite fields in place. – Jyrki Lahtonen Jan 09 '21 at 12:36

1 Answers1

3

From the factorization of $P(X)$ we know that the elements $\alpha^{q^i}$, $i=0,1,\ldots,d-1$, are all distinct. It may be better to restate this in the form $$ \alpha^{q^i}=\alpha^{q^j} $$ if and only if $i\equiv j\pmod d$

If $M$ is an intermediate field, $K\subset M\subset K(\alpha)$, by the tower law of field extensions we know that $\ell:=[M:K]$ is a factor of $d=[K(\alpha):K]$. Also, $M$ has $q^\ell$ elements.

As noted, the polynomial $Q(X)$ has a root, say, $\beta=\alpha^{q^s}$. By (b) the other roots of $Q(X)$ are $$ \beta_j:=\beta^{q^{\ell j}} $$ for $j=0,1,\ldots, t-1$, $t=\deg Q(X)$. For all $i$ we have $\beta_i=\alpha^{q^{s+\ell i}}.$

Observe that by using (a) and induction on $i$, we see that the element $\beta_i$ is a zero of $Q(X)$ for all natural numbers $i$. That is, not only when $0\le i\le t-1$.

The earlier observation says that $$\beta_j=\beta_i$$ if and only if $$s+\ell i\equiv s+\ell j\pmod d$$ if and only if $$ i\equiv j\pmod {d/\ell}. $$ So there are $d/\ell$ distinct elements $\beta_i$. We know that the zeros of $Q(X)$ are distinct, so we can conclude that $t=d/\ell$.

This answers both parts as we see that $\ell=d/r$.

Jyrki Lahtonen
  • 140,891
  • In a sense the key is that we get the minimal polynomial of any element $\gamma$ over a field of $q^m$ elements by repeatedly raising the previous roots to power $q^m$, and stop at the exact point we start getting repetitions. All those powers must be included iin the set of zeros, but no repetitions. – Jyrki Lahtonen Jan 09 '21 at 12:16