Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

I have the following definition of what it means for a polynomial to be separable:

Let $E$ be a field and $P \in E[x]$ be irreducible. Then we say $P$ is separable if it has no repeated root in any field extension of $E$. If $P\in E[x]$ is reducible, we say $P$ is separable if all of its irreducible factors are separable.

In the irreducible case, we have shown $P$ is separable if it has no repeated root in a splitting field extension.

Can anyone give me an example of a polynomial which is not separable? I don't see why this definition isn't trivial. For example, if I had a double root like $(x-2)^2$ then the polynomial is reducible so we consider $(x-2)$ and $(x-2)$ separately, each of which does not have a double root. So when could we ever have a polynomial which is not separable?

share|improve this question

1 Answer 1

up vote 9 down vote accepted

The typical example of a polynomial that is not separable: let $F=\mathbb{F}_p(x)$ be the field of all rational functions with coefficients in the field with $p$ elements. Consider the polynomial $y^p-x$ in $F[y]$. If $u$ is a root of $y^p-x$ (in some extension of $F$), we have $u^p-x=0$, so $u^p=x$. Because we are in charactersitic $p$, the binomial theorem becomes $$(a\pm b)^p = a^p \pm b^p,$$ so therefore we have $$(y-u)^p = y^p-u^p = y^p-x,$$ so $u$ is the only root of $y^p-x$ (by unique factorization), which is a perfect $p$ power in its splitting field.

Note also that $y^p-x$ is irreducible over $F$ (so that $u$ actually "lives" in some extension strictly larger than $F$). To prove this, consider e.g. by Eisenstein's criterion applied to it as an element of $(\mathbb{F}_p[x])[y]$: $\mathbb{F}_p[x]$ is a UFD where $x$ is prime; every coefficient of $y^p-x$ except the leading one is a multiple of $x$, and the constant term is not divisible by $x^2$; so by Eisenstein's Criterion $y^p-x$ is irreducible in $(\mathbb{F}_p[x])[y]$. Now Gauss's Lemma lets us conclude that the polynomial is also irreducible over $F$.

Over a field of characteristic $0$, all irreducible polynomials are separable. This follows because a polynomial has a multiple root if and only if the gcd with its formal derivative is not equal to $1$. But if $p(x)$ is irreducible, then $p'(x)$ is a polynomial of strictly smaller degree than $p(x)$, and therefore no nonconstant factor of $p(x)$ can be a factor of $p'(x)$. Thus, $\gcd(p(x),p'(x))=1$, so $p(x)$ is separable.

So inseparability is a purely positive characteristic phenomenon. In fact, one can show that an irreducible polynomial over a field of characteristic $p$ is not separable if and only if it can be written as a polynomial in $x^p$ (that is, every exponent that occurs is a multiple of $p$).

share|improve this answer
    
I don't quite follow. How is $y^p-x$ irreducible if you just factored it as $(y-u)^p$? Do you have a more concrete example? –  nullUser May 19 '12 at 21:01
1  
@nullUser: The polynomial is irreducible *over $F$*. The factorization $(y-u)^p$ involves $u$ which is not in $F$. –  Ted May 19 '12 at 21:06
    
@Ted: Oops; didn't read that comment carefully enough. Thanks. –  Arturo Magidin May 19 '12 at 21:09
1  
@nullUser: Sorry, I wasn't very clear. If $u$ is a root of the polynomial, possibly in some extension of $F$*, then we see that $y^p-x$ is equal to $(y-u)^p$. *Then we can prove (using Eisenstein's Criterion and Gauss's Lemma) that the polynomial is in fact irreducible over $F$, so that its roots do indeed "live" in some extension of $F$. So the polynomial is irreducible when considered as a polynomial in $F[y]$, but it has multiple roots when considered as a polynomial in $K[y]$, where $K$ is any field that contains roots for it. –  Arturo Magidin May 19 '12 at 21:11

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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