D-MATH Algebra II FS19

Prof. Rahul Pandharipande

Solution 25
Finite fields

1. Let k be a field.

(a) Show that k is an extension of a field k0 , called prime field, given by k0 = Q

if char (k) = 0 k0 = Fp and if char (k) = p > 0.
(b) Show that any field homomorphism restricts to the identity on the prime
field.

Solution:

(a) The characteristic of the field k is precisely the non-negative generator of the
kernel of the unique ring homomorphism ϕ : Z → k.
If char (k) = 0, then ϕ is an injective map. Since Q is the field of fractions
of Z, the inclusion ϕ extends to an inclusion of Q inside k.
If char (k) > 0, then it is a prime number p and by the first homomorphism
theorem ϕ induces an injection ϕ : Fp := Z/pZ → k, and k0 coincides with
the additive subgroup of k generated by 1k .
(b) If θ : k → ` is a field homomorphism, then the composition of ring homo-
ϕk θ
morphisms Z −→ k −→ ` must coincide with the unique homomorphism
ϕ` : Z → `. Moreover θ is necessarily injective (as is every field homomor-
phism, because the image of x ∈ k × = k r {0} has inverse θ(x−1 ), hence it
cannot be zero). Thus

ker(ϕ` ) = {m ∈ Z : ϕk (m) ∈ ker(θ)} = {m ∈ Z : ϕk (m) = 0} = ker(ϕk )

so that k and ` have the same characteristic.

If the two fields have characteristic p > 0, then they contain the prime field
Fp as images of ϕk and ϕ` and those prime fields are mapped ”identically”
because ϕ` = θ ◦ ϕk .
If the two fields have characteristic 0, then θ maps each integer m·1k to m·1` .
The inclusion ϕk : Z → k extends to an inclusion ϕk : Q → k by sending m/n
to ϕk (m)ϕk (n)−1 for m, n ∈ Z with n 6= 0. Similarly, ϕ` extends to ϕ` : Q → `.
In order to conclude, it is enough to prove that ϕ` = θ ◦ ϕk , so that θ restricts

1
 to the identity on the prime fields Q seen as images of ϕk and ϕ` . This is
again done by using the fact that ϕ` = θ ◦ ϕk : for all m, n ∈ Z with n 6= 0,

(θ ◦ ϕk )(m/n) = θ(ϕ` (m/n)) = θ(ϕ` (m)ϕ` (n)−1 )

= (θ ◦ ϕ` )(m) · (θ ◦ ϕ` )(n)−1 = ϕ` (m)ϕ` (n)−1 = ϕ` (m/n).

2. We say that a field k is perfect if every algebraic field extension of k is separable.

(a) Prove that k is perfect if and only if every irreducible polynomial in k[X] is
separable, i.e. has no multiple roots.
(b) Let f ∈ k[X] be an irreducible polynomial. Show that f is separable if and
only if its derivative is nonzero.
(c) For f as in Part (b), show that the derivative of f is zero if and only if
char (k) = p > 0 and f (X) = g(X p ) for some irreducible g ∈ k[X].
(d) Suppose that char (k) = p > 0. Prove that k is perfect if and only if the
Frobenius homomorphism ϕ : k → k, x 7→ xp is surjective.
(e) Deduce that fields of characteristic zero and finite fields are perfect.
Solution:
(a) Suppose k is a perfect field and let f ∈ k[X] be an irreducible polynomial
with x a root of f in an algebraic closure k̄ of k. Then k(x) is a field extension
of k and it is separable because k is perfect. Hence x is a separable element,
meaning that its minimal polynomial f is separable.
Conversely, assume that every irreducible polynomial in k[X] is separable
and let `/k be an algebraic extension. Every α ∈ ` has a minimal polynomial
over k because `/k is algebraic; it is a separable polynomial by assumption,
meaning that α is separable. Hence `/k is a separable field extension.
(b) Let a1 , . . . , ar ∈ k̄ be the distinct roots of f with respective multiplicities
n1 , . . . , nr > 1. Over k̄ we thus have the factorization
r
Y
f= (X − ai )ni ,
i=1

with derivative
r
X Y
f0 = ni (X − ai )ni −1 · (X − aj )nj .
i=1 j6=i

From this we see that f 0 (ai ) = ni (ai − ai )ni −1 · j6=i (ai − aj )nj is nonzero if
Q
ni = 1, proving “⇒”.
Conversely, suppose f has a multiple root a in its splitting field E. Then
from the above we see that a is a root of both f and f 0 , so X − a divides

2
 their gcd g (over E), i.e. g has degree at least 1. Moreover, if f 0 6= 0, then g
has degree strictly less than that of f . But the gcd over E is the same as the
gcd over k (see Solution 16, Exercise 1(a)), so f is divisible by g over k and
hence not irreducible – contrary to the assumption.
(c) If char (k) = p > 0 and f (X) = g(X p ), then f 0 (X) = pX p−1 · g 0 (X) = 0.
Pn i
For the converse, write f = i=0 ai X ∈ k[X] with an 6= 0. Then we have
f 0 = ni=1 i·ai X i−1 = 0 if and only if i·ai = 0 for all 1 6 i 6 n. In particular,
P
nan = 0; hence n = 0 in k, which implies that k has positive characteristic p.
Moreover, for any index i not divisible by p, the equation i · ai = 0 yields
Pn/p
ai = 0. Thus, we can write f (X) = j=0 ajp X jp =: g(X p ) ∈ k[X p ]. Note
that any factorization of g yields one of f . Thus g is irreducible because f is.
(d) Suppose that k is a perfect field. We want to show that each y ∈ k has a
p-th root in k. Since k is perfect, the polynomial f = X p − y ∈ k[X] must
be either separable, or reducible by Part (a). Let x ∈ k̄ be a root of f , i.e.
xp = y. Since k has characteristic p, we can compute

(X − x)p = X p − xp = X p − y = f.

Hence x is the only root of f in k̄ and so f is not separable; in fact, a factor

of f in k[X] has no multiple roots in k̄ if and only if it is a linear factor.
As each irreducible factor of f in k[X] must separable, the only possibility is
that f splits completely in k[X]. In particular, x ∈ k.
Conversely, suppose that the Frobenius map ϕ : k → k is surjective. By (a)
it suffices to prove that every irreducible polynomial f in k[X] is separable.
Suppose f ∈ k[X] is irreducible and has multiple roots. Then by Part (c)
we have f ∈ k[X p ]. Moreover, every coefficient of f is a p-th power of an
element in k, since ϕ is surjective by assumption. So we can write
n n
!p
X X
f= bpi X pi = bi X i ,
i=0 i=0

which is a proper factorization of f in k[X], contradicting the assumption

that f is irreducible. Hence f has no multiple roots.
(e) If k is a field of characteristic zero, then by Part (c), the derivative of any
irreducible polynomial over k is nonzero. By Part (b), this implies that every
such polynomial is separable, which by Part (a) is equivalent to k being
perfect.
Let k be a finite field of characteristic p. The Frobenius homomorphism ϕ
from Part (d) is a generator of Gal(k/Fp ) (see Assignment 17, Exercise 3).
In particular, it is an automorphism, hence surjective. By Part (d) k is thus
perfect.

3
3. Let k be a finite field and consider a finite field extension k(α, β)/k such that
k(α) ∩ k(β) = k (inside an algebraic closure of k). Prove that k(α, β) = k(α + β).
Hint. Study the cardinality of the involved fields.
Solution: Clearly, k(α + β) ⊂ k(α, β) since α + β ∈ k(α, β).
For the reverse inclusion, let q = |k| be a power of a prime p. We write k = Fq and
we know that char (k) = p. Fix an algebraic closure k̄. Then, as seen in Algebra I,
for each power q t of q there exists a unique subfield of k̄ containing q t elements: it
t
consists of those elements α ∈ k̄ such that αq = α. The proof of Assignment 13,
Exercise 1(b) generalizes to q and tells us that Fqs ⊂ Fqt if and only if s divides t.
Let n, m ∈ N be such that k(α) = Fqn and k(β) = Fqm . Here n is the minimal
h
positive integer h such that αq = α, because otherwise k(α) would be contained
in a strictly smaller subfield of Fqn . Since k = k(α) ∩ k(β) is the largest subfield
of k̄ contained in both Fqn and Fqm , we deduce that gcd(m, n) = 1. In particular,
p is not a common divisor of m and n. Without loss of generality, assume that p
does not divide n. Also, note that k(α, β) is the smallest subfield of k̄ containing
both Fqn and Fqm , so k(α, β) = Fqmn .
We write k(α + β) = Fqt . This means that
t t t
αq + β q = (α + β)q = α + β,

implying that
t t
αq − α = −(β q − β) ∈ k(α) ∩ k(β) = k.
t
Write αq = α + λ for λ ∈ Fq . Repeatedly raising to the q t -th power, we deduce
inductively that
tp
αq = α + pλ = α.

This means that n | tp and since p - n we obtain n | t. Thus, by uniqueness

of subfields mentioned above, k(α + β) = Fqt contains k(α) and, in particular,
α ∈ k(α + β). This implies that β = (α + β) − α ∈ k(α + β), as well. Hence
k(α, β) ⊂ k(α + β) and we conclude that k(α, β) = k(α + β).

4. Give a detailed proof of Wedderburn’s theorem: Every finite skew-field is a field.

Solution: See N. Jacobson, Basic Algebra I, 2nd Edition, Section 7.7 or
R. Lidl, H. Niederreiter, Finite Fields, Ch. 2, Section 6, Theorem 2.55, first proof.