Why is x^p - a irreducible over a field of characteristic p?

  • Level: Graduate 
  • Thread starter Thread starter imurme8
  • Start date Start date
  • Tags Tags
    Characteristic Field
Join the discussion
Registration is free. Start your own thread to ask a follow-up.
4 replies · 11K views
imurme8
Messages
45
Reaction score
0
If [itex]K[/itex] is a field of characteristic [itex]p[/itex], and there exists an element [itex]a \in K[/itex]
which is not a [itex]p[/itex]th power (i.e. the Frobenius endomorphism is not
surjective), then I am told we can show [itex]x^p - a[/itex] is an irreducible polynomial
(and since it is not separable our field is imperfect). I see that
[itex]x^p - a[/itex] has no roots in [itex]K[/itex], but how do we know that there does not exist
any factorization of [itex]x^p -a[/itex] into factors of lesser degree?
 
Physics news on Phys.org
Hint: Let [itex]f(x) = x^p - a[/itex], let F be a splitting field for f and let [itex]\alpha[/itex] be a root of f(x) in F. Can you take it from there?
 
Last edited:
OK, how about this? Over [itex]F[/itex] we have [itex]\alpha[/itex] as the only root of [itex]f(x)[/itex] (with multiplicity [itex]p[/itex]). Let [itex]f(x)=p_1(x)\dotsb p_n(x)[/itex] be a factorization of [itex]f(x)[/itex] in [itex]K[x][/itex] into monic irreducibles. Then each of these must be the minimal polynomial of [itex]\alpha[/itex] over [itex]K[/itex]. So they all must have the same degree. So the degree of [itex]p_1(x)=\dotsb=p_n(x)[/itex] divides [itex]p[/itex], so it is either 1 or [itex]p[/itex]. But it cannot be 1, so it must be [itex]p[/itex], so [itex]f(x)[/itex] itself is irreducible.

Is that what you would suggest? I don't see an easier way to show [itex](x-\alpha)^m\notin K[x][/itex] for all positive integers [itex]m<p[/itex].
 
Your argument looks OK. Here's what I had in mind: We have that [itex]f(x) = (x - \alpha)^p = x^p - \alpha^p[/itex] and so [itex]a = \alpha^p[/itex]. Suppose that f(x) = g(x)h(x) is a proper, non-trivial factorization of f(x) over K. Then, by unique factorization, [itex]g(x) = (x - \alpha)^s[/itex], for some 0 < s < p. The constant term of g(x) is [itex]\alpha^s[/itex], so [itex]\alpha^s \in K[/itex]. Since gcd(s, p) = 1, there exist integers m and n such that sm + pn = 1. Therefore, [itex]\alpha = \alpha^{sm+pn} = (\alpha^s)^m (\alpha^p)^n[/itex] is an element of K. But [itex]a = \alpha^p[/itex], contradicting the assumption that a is not a pth power in K. Therefore, no such factorization of f(x) exists.
 
  • Like
Likes   Reactions: Warren Leywon