Login
finite field
A finite field (also called a Galois field) is a field that has finitely many elements. The number of elements in a finite field is sometimes called the order of the field. We will present some basic facts about finite fields.
Size of a finite field
Existence of finite fields
Now that we know every finite field has $p^n$ elements, it is natural to ask which of these actually arise as cardinalities of finite fields. It turns out that for each prime $p$ and each natural number $n$ , there is essentially exactly one finite field of size $p^n$ .
In general, the polynomial $f(X) := X^{p^n} - X \in \F_p[X]$ has derivative $-1$ and thus is separable over $\F_p$ . We claim that the splitting field $F$ of this polynomial is a finite field of size $p^n$ . The field $F$ certainly contains the set $S$ of roots of $f(X)$ . However, the set $S$ is closed under the field operations, so $S$ is itself a field. Since splitting fields are minimal by definition, the containment $S \subset F$ means that $S = F$ . Finally, $S$ has $p^n$ elements since $f(X)$ is separable, so $F$ is a field of size $p^n$ .
For the uniqueness part, any other field $F'$ of size $p^n$ contains a subfield isomorphic to $\F_p$ . Moreover, $F'$ equals the splitting field of the polynomial $X^{p^n} - X$ over $\F_p$ , since by Lemma 2.1 every element of $F'$ is a root of this polynomial, and all $p^n$ possible roots of the polynomial are accounted for in this way. By the uniqueness of splitting fields up to isomorphism, the two fields $F$ and $F'$ are isomorphic. ![]()
Note: The proof of Theorem 2.2 given here, while standard because of its efficiency, relies on more abstract algebra than is strictly necessary. The reader may find a more concrete presentation of this and many other results about finite fields in [1, Ch. 7].
Units in a finite field
Henceforth, in light of Theorem 2.2, we will write $\F_q$ for the unique (up to isomorphism) finite field of cardinality $q = p^n$ . A fundamental step in the investigation of finite fields is the observation that their multiplicative groups are cyclic:
where $\phi$ denotes the Euler totient function. It is proved as follows. For every divisor $d$ of $k$ , the cyclic group $C_k$ of size $k$ has exactly one cyclic subgroup $C_d$ of size $d$ . Let $G_d$ be the subset of $C_d$ consisting of elements of $C_d$ which have the maximum possible order of $d$ . Since every element of $C_k$ has maximal order in the subgroup of $C_k$ that it generates, we see that the sets $G_d$ partition the set $C_k$ , so that $$ \sum_{d \div k} |G_d| = |C_k| = k. $$ The identity (1) then follows from the observation that the cyclic subgroup $C_d$ has exactly $\phi(d)$ elements of maximal order $d$ .
We now prove the theorem. Let $k = q-1$ , and for each divisor $d$ of $k$ , let $\psi(d)$ be the number of elements of $\F_q^*$ of order $d$ . We claim that $\psi(d)$ is either zero or $\phi(d)$ . Indeed, if it is nonzero, then let $x \in \F_q^*$ be an element of order $d$ , and let $G_x$ be the subgroup of $\F_q^*$ generated by $x$ . Then $G_x$ has size $d$ and every element of $G_x$ is a root of the polynomial $x^d - 1$ . But this polynomial cannot have more than $d$ roots in a field, so every root of $x^d - 1$ must be an element of $G_x$ . In particular, every element of order $d$ must be in $G_x$ already, and we see that $G_x$ only has $\phi(d)$ elements of order $d$ .
We have proved that $\psi(d) \leq \phi(d)$ for all $d \div q-1$ . If $\psi(q-1)$ were 0, then we would have $$ \sum_{d \div q-1} \psi(d) < \sum_{d \div q-1} \phi(d) = q-1, $$ which is impossible since the first sum must equal $q-1$ (because every element of $\F_q^*$ has order equal to some divisor $d$ of $q-1$ ). ![]()
A more constructive proof of Theorem 3.1, which actually exhibits a generator for the cyclic group, may be found in [2, Ch. 16].
Automorphisms of a finite field
Observe that, since a splitting field for $X^{q^m} - X$ over $\F_p$ contains all the roots of $X^q - X$ , it follows that the field $\F_{q^m}$ contains a subfield isomorphic to $\F_q$ . We will show later (Theorem 4.2) that this is the only way that extensions of finite fields can arise. For now we will construct the Galois group of the field extension $\F_{q^m}/\F_q$ , which is normal by Corollary 2.3.
It is enough to show that none of $(\Frob_q)^k$ , for $k = 1, 2, \ldots, m-1$ , is the identity map on $\F_{q^m}$ , for then we will have shown that $\Frob_q$ is of order exactly equal to $m$ . But, if any such $(\Frob_q)^k$ were the identity map, then the polynomial $X^{q^k} - X$ would have $q^m$ distinct roots in $\F_{q^m}$ , which is impossible in a field since $q^k < q^m$ . ![]()
We can now use the Galois correspondence between subgroups of the Galois group and intermediate fields of a field extension to immediately classify all the intermediate fields in the extension $\F_{q^m}/\F_q$ .
The subfields of $\F_{p^n}$ can be obtained by applying the above considerations to the extension $\F_{p^n}/\F_p$ . ![]()
Bibliography
- 1
- Kenneth Ireland & Michael Rosen, A Classical Introduction to Modern Number Theory, Second Edition, Springer-Verlag, 1990 (GTM 84).
- 2
- Ian Stewart, Galois Theory, Second Edition, Chapman & Hall, 1989.

