2

I have to prove the following theorem :

Let $p$ be a prime number and let $n \ge 1$,be any integer, then there exists a field of order $p^n$.

My attempt

I started off by considering the polynomial $f(x)$=$x^{p^n}-x \in \Bbb Z_p[x]$.

I took $F$ to be the splitting field of $f(x)$ over $\Bbb Z_p$.

Since $F$ is the splitting field of $f(x)$ over $\Bbb Z_p$ therefore $f(x)$ has exactly $p^n$ zeros in $F$ counting multiplicity.

Also since $f'(x)$ = $p^nx^{p^n-1}-1$ , therefore $f(x)$ will not have any multiple zeros and hence it will have $p^n$ distinct zeros in $F$.

Then I took $K =\{k \in F | k^{p^n} = k\}$.

I was trying to prove that $K$ is a sub-field of $F$.

$1 \in K$ thus $K$ is non-empty.

For $a,b \in K$ $a^{p^n}=a$ and $b^{p^n}=b$

So $(a+b)^{p^n}$ (should be) = $a^{p^n}$+$b^{p^n}$, this is where I am currently stuck!

I know that this is true when $p$ is the characteristic of a field, but why would this expansion be true here?

Adam Hughes
  • 36,777
johny
  • 1,609

2 Answers2

3

In $\mathbb Z_p$ and hence its extension $F$ we have $$ p a = \underbrace{a + \cdots + a}_{p \text{ times}} = (\underbrace{1 + \cdots + 1}_{p \text{ times}})a = 0 \cdot a = 0. $$

By the binomial theorem $$ (a + b)^{p^n} = \sum_{k=0}^{p^n}\binom{p^n}{k} a^k b^{p^n - k}. $$

All terms in the latter expansion are multiples of $p$ except for $a^{p^n}$ and $b^{p^n}$. It follows that in $\mathbb Z_p$ and $F$ we have $$(a + b)^{p^n} = a^{p^n} + b^{p^n}.$$

Now show that the roots of $f(x)$ are closed under addition, multiplication and taking additive and multiplicative inverses. This shows that $F$ consists entirely of the roots of $f(x)$. We conclude that the order of $F$ is $p^n$ as desired.

Ayman Hourieh
  • 39,603
1

We note that it is sufficient to show there exists an irreducible polynomial, $Q(x)$ of degree $n$ over $\Bbb F_p$, the field with $p$ elements (which you denote by $\Bbb Z_p$). This is so because in this case you have that $\Bbb F_p[x]/(Q(x))$ is a field of degree $n$ over $\Bbb F_p$, hence has $p^n$ elements.

Consider the polynomial $x^{p^n}-x$, which has no repeated roots, since its derivative is $-1$. Then I claim the splitting field of this polynomial exists and has degree $n$ over $\Bbb F_p$. For each divisor $d | n$ with $d < n$, consider the polynomial $\Phi_d(x)$ which we define to be the product of all irreducible degree-$d$ polynomials over $\Bbb F_p$. Then define

$$P(x) = \left(x^{p^n-1}-1\right) \bigg/ \prod_{d|n,\; d<n} \Phi_d(x)$$

Since all the roots of $x^{p^d}-x$ are in one of the factors we've factored out, the representation of $P(x)$ as a product of irreducibles satisfies the fact that every root of every irreducible factor has multiplicative order exactly $p^n$. Choose one such root, $\omega$. Then we have that $\deg \omega\ge n$ because $x^{p^d}\ne 1$ for $d<n$. However we also see that the $\deg \omega \le n$ because if it generated an extension of degree $k<n$, we can use the fact that the multiplicative group of a field is a group, and we'd have $x^{p^k}=1$, a contradiction. So $\deg\omega =n$, and if $m_\omega(x)$ is it's minimal polynomial we have that $m_\omega(x)=Q(x)$ is the desired polynomial.

Adam Hughes
  • 36,777
  • Why is $P\left(x\right) \neq 1$? – darij grinberg May 28 '19 at 21:13
  • It's degree is clearly positive, so it can't be a constant – Adam Hughes May 28 '19 at 21:29
  • How do you prove that its degree is positive, short of counting aperiodic necklaces? – darij grinberg May 28 '19 at 21:41
  • The definition of the degree of a polynomial (in one variable) is the largest power of the variable term which has a non-zero coefficient. In our case the highest degree term has a coefficient of 1 which is not 0 in a field by the field axioms – Adam Hughes May 28 '19 at 22:08
  • How does this prevent $P\left(x\right)$ from being $1$? – darij grinberg May 28 '19 at 22:13
  • I'm not sure I understand your question, two elements of a ring, $a,b\in R$ are equal iff $a-b=0$. The zero polynomial is by definition the one for which all coefficients are 0, and P(x)-1 clearly fails to satisfy that definition as its leading coefficient is 1 – Adam Hughes May 28 '19 at 22:35
  • The leading coefficient of the constant polynomial $1$ is $1$, too. – darij grinberg May 28 '19 at 22:56
  • Yes but as we've away discussed the leading term here has positive degree – Adam Hughes May 28 '19 at 22:59
  • Not sure where this positivity come from. Why can't the $\Phi_d$ take out all the irreducible factors of $x^{p^n-1}-1$? – darij grinberg May 28 '19 at 23:34
  • Basic group theory tells you there are more roots to the numerator than can possibly be in the denominators (the standard argument using LaGrange's theorem is taught in any good algebra class). It might help if you explain your background, it seems you're not terribly familiar with these results, and I may be able to recommend a book to help you brush up on the basics-- it will probably help you more than going back and forth in these comments – Adam Hughes May 28 '19 at 23:45
  • 1
    I am fairly familiar with these results -- I'm currently looking for the most elementary proof that I can find, but I'm well aware of the proofs using necklace counting and using splitting fields. What I don't understand is the specifics of your argument -- it appears to use the exact thing it is proving. – darij grinberg May 29 '19 at 00:40
  • Ah, you certainly won't find the most elementary proof from me: I usually choose the most elegant, which often uses more advanced machinery. I truly Elementary proof won't appeal to any group theory and will probably count the roots of the denominator by hand using Mobius inversion. I don't know of any books offhand that take this approach, so I'm sorry to say I probably can't point you further in the right direction – Adam Hughes May 29 '19 at 00:44