My problem is that I don't have idea how to complete one side of the proof.

For any field $k$ an irreducible polynomial $p \in k[x] $ is called separable if for all field extensions of $k$ all roots of $p$ are simple or equivalently $p$ has no multiple roots.


An irreducible polynomial $p$ is separable iff $p' \neq 0$ as $k \to k $ function.


(assume for simplicity that $p$ is monic)

($\Rightarrow$) If $p$ is separable then in splitting field $K$ $$p'(x) = \sum^d_{i =1}\frac{1}{x- a_i}\prod^d_{j=1}(x - a_j)$$ where $a$ are roots of $p$ in $K$ (all distinct) and $d = \deg p$. It's clear that $p'(a_1) = \prod_{j = 2}^d(a_i - a_j)$ is non zero by definition of the field. And hence $p'$ is non zero as function over $K$ and as consequence over $k$.

($\Leftarrow$) And here I am stacked.

Here we just assume that $p' \neq 0$ as a function. It's clear that $\deg p' = d -1$ and that if $a$ is a multiple root of $p$ then $p'(a) = 0$ . Moreover, multiplicity of such roots is reduced by one. I don't understand how to deduce contradiction from this facts, for example that $p | p'$ and $\deg p' > \deg p$ or that $p' |p$ and $p$ is not irreducible, so I need Your help here.

Thank You in advance.


Suppose $p$ has a multiple root $r$ in some field extension $K$ of $k$. Then $r$ is a root of both $p$ and $p'$, so $x - r$ divides both $p$ and $p'$, i.e. the gcd $g$ of $p$ and $p'$ (over $K$) has degree $\ge 1$. Moreover if $p' \ne 0$, $g$ has degree less than that of $p$. But the gcd over $K$ is the same as the gcd over $k$ (it is computed by the same extended Euclidean algorithm), so $p$ is divisible by $g$ as a polynomial over $k$ and is not irreducible.

  • 2
    $\begingroup$ Thank you, It turned out hat my problem was that I forgot that gcd is field invariant, so I was imagening it as something exclusively belonging to $K[X]$. $\endgroup$ – Nik Pronko Oct 6 '16 at 22:12
  • 1
    $\begingroup$ Why do you need to consider $K[x]$ at all? Isn't $p'\in k[x]$? $\endgroup$ – Akababa Jan 31 '18 at 17:27

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.