2

Problem 7.5 in Chapter 15 of Artin's Algebra asks to factorize $x^{27}-x$ over $\mathbb F_3$.

Here is what I have done.

$x^{27}-x=x(x^{26}-1)= x(x^{13}-1)(x^{13}+1)$.

In am having trouble factorizing $x^{13}-1$ and $x^{13}+1$ in $\mathbb F_3$.

$x^{13}-1=(x-1)(x^{12}+x^{11}+\cdots+x+1)$.

I know that $x^{12}+x^{11}+\cdots+x+1$ is reducible over $\mathbb F_3$ because the degree of this polynomial, which is $12$, does not divide $3$. ($3$ is the exponent of $3$ which gives $27$).

But I can't see how to factorize this polynomial.

Similarly for $x^{13}+1$.

Can somebody help?

Henno Brandsma
  • 242,131

3 Answers3

6

The irreducible factors of $x^{3^3}-x$ are precisely the monic irreducible polynomials of degree $1$ and $3$ over $\mathbb F_3$. So there are the $3$ linear factors and $8$ irreducible polynomials of degree $3$. Giving them all is only a matter of patience.

You can do it like this:

Make a list of all polynomials of the form $x^3+ax^2+bx+c$ with $c \neq 0$. Cancel all those polynomials with $f(1) = 0$ or $f(-1)=0$. What is left are precisely the $8$ polynomials you are looking for.

MooS
  • 31,390
3

Shaking (not stirring) my bag of tricks for this problem. The following dropped out.

Start with $$ p_1(x)=x^3-x+1,\qquad p_2(x)=x^3-x-1. $$ The irreducibility of these has been explained many times on our site (search for $x^p-x+a$ over $\Bbb{F}_p$).

The first trick is that if a polynomial $p(x)$ of degree $k>1$ is irreducible, then so is its reciprocal polynomial $$ \tilde{p}(x):=x^kp(\frac1x). $$ For you can easily check that $p(x)=a(x)b(x)\Leftrightarrow \tilde{p}(x)=\tilde{a}(x)\tilde{b}(x)$. (the reciprocal of the reciprocal is the original polynomial).

This gives us two more irreducibles: $$ p_3(x)=\tilde{p_1}(x)=x^3-x^2+1,\qquad p_4(x)=\tilde{p_2}(x)=x^3+x^2-1. $$

The next trick is that if $p(x)$ is and irreducible cubic, then so are $p(x\pm1)$. These won't give us anything new on $p_1$ or $p_2$, because those polynomials are invariant under this linear substitution. However, we do get $$ p_5(x)=p_3(x+1)=(x^3+1)-(x^2+2x+1)+1=x^3-x^2+x+1 $$ and $$ p_6(x)=p_4(x+1)=(x^3+1)+(x^2+2x+1)-1=x^3+x^2-x+1. $$

Leaving it to you to figure out which of the tricks you should reapply to get the missing two irreducible cubics.


I was planning on using the following trick as well, but looks like we don't need it. Assume that $\alpha$ is a zero of $p_1(x)$ in $\Bbb{F}_{27}$. Let's find the minimal polynomial of $\alpha^2$.

To that end consider the polynomial $$ p_1(x)p_1(-x)=(x^3-x+1)(-x^3+x+1)=-x^6+2x^4-x^2+1=-q(x^2), $$ where $q(x)=x^3+x^2+x-1$. We see that $q(\alpha^2)=-p_1(\alpha)p_2(\alpha)=0$. It is easy to see that $\alpha^2$ cannot be an element of the prime field, so its minimal polynomial is also a cubic ...

Jyrki Lahtonen
  • 133,153
  • 1
    Do you know something about this: Starting with the Artin-Schreier-Polynomials $x^p-x+a \in \mathbb F_p[x]$, will we get all the irreducible polynomials of degree $p$ with those tricks? – MooS Feb 22 '15 at 13:45
  • A good question, @MooS. I don't know. If we think in terms of the zeros of those polynomials, the tricks amount 1) adding an element of the prime field, 2) taking the inverse, 3) squaring, 4) multiplying by an element of the prime field (not used here). The question is what is the minimal subset of the field closed under the above operation that contains the zeros of the Artin-Schreier polynomials. I don't see it. At least not right away. – Jyrki Lahtonen Feb 22 '15 at 14:46
1

Claim: $x^{27}-x$ factors over $\Bbb{F}_3$ as the product of all irreducible polynomial of degree $1$ and $3$.

Proof:

First notice that $x^3-x \mid x^{27}-x$ since $3\mid 27$, so $\Bbb{F}_3\subset \Bbb{F}_{27}$ and so every element of $\Bbb{F}_3$ divides $x^{27}-x$. I.e.: $x-a\mid x^{27}-x \, \forall a \in \Bbb{F}_3$. So there are $3$ irreducible polynomial of degree $1$.

Let $g(x) \in \Bbb{F}_3[x]$ be a monic irreducible polynomial of degree $3$ and let $\beta$ be one of its roots in $\Bbb{F}_{27}$. Then $[\Bbb{F}_3(\beta):\Bbb{F}_3] = 3$, so, since all the finite fields of a fixed order are isomorphic to each other, $\Bbb{F}_3(\beta) \approx \Bbb{F}_{27}$. Therefore $x-\beta \mid x^{27}-x$. Since you can do this reasoning for all the roots of $g$ over $\Bbb{F}_{27}$ we have $g(x) \mid x^{27}-x$. Since there are $8$ irreducible polynomial of degree $3$ we are done, since the degree of the multiplication of all the monic irreducible polynomial of degree $1$ and $3$ is $3+8\cdot 3 = 27$.

This might also help you.

Bman72
  • 2,854
  • 1
  • 16
  • 30
  • Using the number of irreducible polynomials of a certain given degree to show the decomposition of $x^{p^n}-x$ ends up in circular reasoning.

    Of course it works with $p=3$ and degree $3$, as we can just compute all irreducibles by hand due to the above method. But in general you need another argument. Of course you can find the proof in any algebra textbook.

    – MooS Feb 22 '15 at 13:35