General principles in factoring

A general principle in factoring polynomials is to determine whether or not the polynomial has multiple factors is the following one. First, for any polynomial $ f(x)=a_0+a_1x+a_2x^2+...+a_nx^n\in F[x]$, let

$\displaystyle f'(x)=a_1+2a_2x+...+na_nx^{n-1}
$

denote the derivative polynomial.

Example 2.4.5   If the field is not characteristic zero then funny things can happen when you differentiate! For example, if $ f(x)=x^{25}+x^7\in {\mathbb{F}}_5[x]$ then $ f'(x)=25x^{24}+7x^6=2x^6$, so $ deg(f(x))=25$ while $ deg(f'(x))=6$. Even stranger things than this can happen: if $ f(x)=x^5+1\in {\mathbb{F}}_5[x]$ then $ f'(x)=5x^4=0$. In other words, some non-constant polynomials have derivative equal to 0!

Lemma 2.4.6   If $ F$ is a ring, $ f(x)$ is a polynomial in $ F[x]$, and $ f'(x)\not=0$ (i.e., $ f'(x)$ is not the zero polynomial) and then $ f(x)$ has a multiple factor if and only if $ gcd(f(x),f'(x))\not= 1$.

proof: The proof uses the fact that if $ f(x)=g(x)^kh(x)$ then we have the ``product rule'' $ f'(x)=kg(x)^{k-1}g'(x)h(x)+g(x)^kh'(x)$. This implies $ g(x)^{k-1}\vert gcd(f(x),f'(x))$, which proves the lemma. $ \Box$

Example 2.4.7   Let

$\displaystyle f(x)=(x+2)^3(x^2+1)^5,
$

so

\begin{displaymath}
\begin{array}{c}
f'(x)=
3\,\left (x+2\right )^{2}\left ({x}^...
...left (x+2\right )^{2}\left ({x}^{2}
+1\right )^{4},
\end{array}\end{displaymath}

so $ gcd(f(x),f'(x))=(x+2)^2(x^2+1)^4$.

Lemma 2.4.8   If $ F$ is a field of characteristic zero and $ f(x)$ is an irreducible polynomial in $ F[x]$, $ f(x)$ has cannot have a multiple root in any field extension $ E/F$.

proof: Since $ f(x)$ is an irreducible polynomial in $ F[x]$, by the previous lemma and the Euclidean algorithm, there are polynomials $ p(x)$ and $ q(x)$ such that $ f(x)p(x)+f'(x)q(x)=1$. This also holds over any field extension.

Suppose, to get a contradition, $ f(x)$ had a multiple root in $ E$. By the previous lemma, it would have a root in common with $ f'(x)$, say $ c\in E$. Plugging this into the equation $ f(x)p(x)+f'(x)q(x)=1$ would yield $ 0=1$, which is impossible. $ \Box$



David Joyner 2007-09-03