18.

785 Number theory I Fall 2021

Lecture #4 09/20/2021

4 Étale algebras, norm and trace

4.1 Separability
In this section we briefly review some standard facts about separable and inseparable field
extensions that we will use repeatedly throughout the course. Those familiar with this
material should feel free to skim it. In this section K denotes any field, K is an algebraic
closure that we will typically P P L/K
choose to contain any extensions under consideration,
and for any polynomial f = ai xi ∈ K[x] we use f 0 := iai xi−1 to denote the formal
derivative of f (this definition also applies when K is an arbitrary ring).

Definition 4.1. A polynomial f in K[x] is separable if (f, f 0 ) = (1), that is, gcd(f, f 0 ) is a
unit in K[x]. Otherwise f is inseparable.

If f is separable then it splits into distinct linear factors over over K, where it has deg f
distinct roots; this is sometimes used as an alternative definition. Note that the property of
separability is intrinsic to the polynomial f , it does not depend on the field we are working in;
in particular, if L/K is any field extension the separability of a polynomial f ∈ K[x] ⊆ L[x]
does not depend on whether we view f as an element of K[x] or L[x].

Warning 4.2. Older texts (such as Bourbaki) define a polynomial in K[x] to be separable if
all of its irreducible factors are separable (under our definition); so (x−1)2 is separable under
this older definition, but not under ours. This discrepancy does not change the definition of
separable elements or field extensions.

Definition 4.3. Let L/K be an algebraic field extension. An element α ∈ L is separable over
K if it is the root of a separable polynomial in K[x] (in which case its minimal polynomial is
necessarily separable). The extension L/K is separable if every α ∈ L is separable over K;
otherwise it is inseparable.

Lemma 4.4. An irreducible polynomial f ∈ K[x] is inseparable if and only if f 0 = 0.

Proof. Let f ∈ K[x] be irreducible; then f is nonzero and not a unit, so deg f > 0. If f 0 = 0
then gcd(f, f 0 ) = f 6∈ K × and f is inseparable. If f is inseparable then g := gcd(f, f 0 ) is a
nontrivial divisor of f and f 0 . This implies deg g = deg f , since f is irreducible, but then
deg f 0 < deg f = deg g, so g cannot divide f 0 unless f 0 = 0.

Corollary 4.5. Let f ∈ K[x] be irreducible and let p ≥ 0 be the characteristic of K. We

n
have f (x) = g(xp ) for some irreducible separable g ∈ K[x] and integer n ≥ 0 that are
uniquely determined by f .

Proof. If f is separable the theorem holds with g = f and n = 0; for uniqueness, note that
n
if p = 0 then pn 6= 0 if and only if n = 0, and if p > 0 and g(xp ) is inseparable unlessPn = 0
n n n
because g(xp )0 = g 0 (xp )pnPxp −1 = 0 (by the previous lemma). Otherwise f (x) := fr xr
0
is inseparable and f (x) = rfr x r−1 = 0 (by the lemma), and this can occur only if p > 0
and fr = 0 for all r ≥ 0 not divisible by p. So f = g(xp ) for some (necessarily irreducible)
g ∈ K[x]. If g is separable we are done; otherwise we proceed by induction. As above, the
n
uniqueness of g and n is guaranteed by the fact that g(xp )0 = 0 for all n > 0.

Corollary 4.6. If char K = 0 then every algebraic extension of K is separable.

Lecture by Andrew V. Sutherland

Lemma 4.7. Let L = K(α) be an algebraic field extension contained in an algebraic closure
K of K and let f ∈ K[x] be the minimal polynomial of α over K. Then

# HomK (L, K) = #{β ∈ K : f (β) = 0} ≤ [L : K],

with equality if and only if α is separable over K.

Proof. Each element of HomK (L, K) is uniquely determined by the image of α, which must
be a root β of f (x) in K. The number of these roots is equal to [L : K] = deg f precisely
when f , and therefore α, is separable over K.

Definition 4.8. Let L/K be a finite extension of fields. The separable degree of L/K is

[L : K]s := # HomK (L, K).

The inseparable degree of f is

[L : K]i := [L : K]/[L : K]s

We will see shortly that [L : K]s always divides [L : K], so [L : K]i is an integer (in fact a
power of the characteristic of K), but it follows immediately from our definition that

[L : K] = [L : K]s [L : K]i .

holds regardless.
Theorem 4.9. Let L/K be an algebraic field extension. and let φK : K → Ω be a homomor-
phism to an algebraically closed field Ω. Then φK extends to a homomorphism φL : L → Ω.
Proof. We use Zorn’s lemma. Define a partial ordering on the set F of pairs (F, φF ) for
which F/K is a subextension of L/K and φF : F → Ω extends φK by defining

(F1 , φF1 ) ≤ (F2 , φF2 )

whenever F2 contains
S F1 and φF2 extends φF1 . Given any totally ordered subset C of F, let
E be the field {F : (F, φF ) ∈ C} and define φE : E → Ω by φE (x) = φF (x) for x ∈ F ⊆ E
(this does not depend on the choice of F because C is totally ordered). Then (E, φE ) is a
maximal element of C, and by Zorn’s lemma, F contains a maximal element (M, φM ).
We claim that M = L. If not, then pick α ∈ L−M and consider the field F = M (α) ⊆ L
properly containing M , and extend φM to ϕF : F → Ω be letting φF (α) be any root of αM (f )
in Ω, where f ∈ M [x] is the minimal polynomial of α over M and αM (f ) is the image of
f in Ω[x] obtained by applying φM to each coefficient. Then (M, φM ) is strictly dominated
by (F, φF ), contradicting its maximality.

Lemma 4.10. Let L/F/K be a tower of finite extensions of fields and K be an algebraic
closure of K that contains L. Then

# HomK (L, K) = # HomK (F, K)# HomF (L, K).

Proof. The result is immediate when F = K or L = F (the RHS is 1 times the LHS), so
we assume K ( F ( L and decompose the extensions L/F and F/K into finite towers of
non-trivial simple extensions

K = K0 ( K1 ( · · · ⊆ Km = F = Km ( Km+1 ( · · · ( Kn = L,

18.785 Fall 2021, Lecture #4, Page 2

where Ki = Ki−1 (αi ) for 1 ≤ i ≤ n. To prove the lemma it suffices to show that
n
Y
# HomK0 (Kn , K) = # HomKi−1 (Ki , K)
i=1

for any tower of proper simple extensions K0 ( K1 ( · · · ( Kn . We now consider the map
n
Φ : HomK (Kn , K) → K
ϕ 7→ (ϕ(α1 ), · · · , ϕ(αn )).

This map Φ is injective, since ϕ : Kn = K0 (α1 , . . . , αn ) → K is uniquely determinedQby the

images ϕ(α1 ), . . . ϕ(αn ), so it suffices to show that the image of Φ has cardinality ni=1 hi ,
where hi := # HomKi−1 (Ki , K).
For any polynomial f ∈ K[x] let r(f ) := {β ∈ K : f (β) = 0} be the set of its K-roots.
Each σ ∈ AutK (K) induces a bijection r(f ) → r(σ(f )) with #r(f ) = #r(σ(f )).
Let fi be the minimal polynomial of αi over Ki−1 . Lemma 4.7 implies hi = #r(fi ), and
for any ϕ ∈ HomK (Kn , K) we have hi = #r(ϕ(fi )), since Theorem 4.9 allows us to extend ϕ
to an element of HomK (K, K) = AutK (K).
Let ϕ0 : K → K be the inclusion map. For each ϕ ∈ HomK (Kn , K) we may define
a compatible sequence of pairs (ϕi , βi ) with ϕi ∈ HomK (Ki , K) satisfying ϕi |K = ϕi−1
i−1
and βi ∈ r(ϕi−1 (fi )) by putting ϕi = ϕ|K and βi = ϕ(αi ); note that (β1 , . . . , βn ) uniquely
i
determine ϕ and the sequence of pairs (αi , βi ), so the set of such compatible sequences of pairs
is in bijection with the image of Φ. We can construct such a compatible sequence by initially
choosing β1 ∈ r(f1 ) and letting ϕ1 (α1 ) = β1 uniquely determine ϕ1 ∈ HomK (K1 , K); there
are exactly h1 choices for the pair (α1 , β1 ). To extend this sequence we choose β2 ∈ r(ϕ1 (f2 ))
and let ϕ2 (α2 ) = β2 and ϕ2 |K = ϕ1 uniquely determine ϕ2 ∈ HomK (K2 , K). There are
1
exactly
Qn h 2 choices for the pair (α 2 , β2 ). Continuing in this fashion, we find there are exactly
i=1 hi sequences of pairs (αi , βQ i ) corresponding to some ϕ ∈ HomK (Kn , K). The image of
the map Φ thus has cardinality ni=1 hi as desired.

Corollary 4.11. Let L/F/K be a tower of finite extensions of fields. Then

[L : K]s = [L : F ]s [F : K]s
[L : K]i = [L : F ]i [F : K]i

Proof. The first equality follows from the lemma and the second follows from the identities
[L : K] = [L : F ][F : K] and [L : K] = [L : K]s [L : K]i .

Theorem 4.12. Let L/K be a finite extension of fields. The following are equivalent:
(a) L/K is separable;
(b) [L : K]s = [L : K];
(c) L = K(α) for some α ∈ L separable over K;
(d) L ' K[x]/(f ) for some monic irreducible separable polynomial f ∈ K[x].

Proof. The equivalence of (c) and (d) is immediate (let f be the minimal polynomial of α
and let α be the image of x in K[x]/(f )), and the equivalence of (b) and (c) is given by
Lemma 4.7. That (a) implies (c) is the Primitive Element Theorem, see [2, §15.8] or
[3, §V.7.4] for a proof. It remains only to show that (c) implies (a).

18.785 Fall 2021, Lecture #4, Page 3

 So let L = K(α) with α separable over K. For any β ∈ L we can write L = K(β)(α),
and we note that α is separable over K(β), since its minimal polynomial over K(β) divides
it minimal polynomial over K, which is separable. Lemma 4.7 implies [L : K]s = [L : K]
and [L : K(β)]s = [L : K(β)] (since L = K(α) = K(β)(α)), and the equalities

[L : K] = [L : K(β)][K(β) : K]
[L : K]s = [L : K(β)]s [K(β) : K]s

then imply [K(β) : K]s = [K(β) : K]. So β is separable over K (by Lemma 4.7). This
applies to every β ∈ L, so L/K is separable and (a) holds.

Corollary 4.13. Let L/K be a finite extension of fields. Then [L : K]s ≤ [L : K] with
equality if and only if L/K is separable.

Proof. We have already established this for simple extensions, and otherwise we my decom-
pose L/K into a finite tower of simple extensions and proceed by induction on the number
of extensions, using the previous two corollaries at each step.

Corollary 4.14. Let L/F/K is a tower of finite extensions of fields. Then L/K is separable
if and only if both L/F and F/K separable.

Proof. The forward implication is immediate and the reverse implication follows from Corol-
laries 4.11 and 4.13.

Corollary 4.15. Let L/F/K be a tower of algebraic field extensions. Then L/K is separable
if and only if both L/F and F/K are separable.

Proof. As in the previous corollary the forward implication is immediate. To prove the
reverse implication, we assume L/F and F/K are separable and show that every β ∈ L is
separable over K. If β ∈ F we are done, and if not we at least know that β is separable
over F . Let M/K be the subextension of F/K generated by the coefficients of the minimal
polynomial f ∈ F [x] of β over F . This is a finite separable extension of K, and M (β) is also
a finite separable extension of M , since the minimal polynomial of β over M (β) is f , which
is separable. By the previous corollary, M (β), and therefore β, is separable over K.

Corollary 4.16. Let L/K be an algebraic field extension, and let

F = {α ∈ L : α is separable over K}.

Then F is a separable field extension of K.

Proof. This is clearly a field, since if α and β are both separable over K then K(α) and
K(α, β) are separable extensions of K (by the previous corollary), thus every element of
K(α, β), including αβ and α + β, is separable over K and lies in F . The field F is then
separable by construction.

Definition 4.17. Let L/K be an algebraic field extension. The field F in Corollary 4.16 is
the separable closure of K in L. When L is an algebraic closure of K it is simply called a
separable closure of K and denoted K sep .

When K has characteristic zero the notions of separable closure and algebraic closure
necessarily coincide. This holds more generally whenever K is a perfect field.

18.785 Fall 2021, Lecture #4, Page 4

Definition 4.18. A field K is perfect if every algebraic extension of K is separable.

All fields of characteristic zero are perfect. Perfect fields of positive characteristic are
characterized by the following property.

Theorem 4.19. A field K of characteristic p > 0 is perfect if and only if K = K p , that is,
every element of K is a pth power, equivalently, the map x 7→ xp is an automorphism.

Proof. If K 6= K p then for any α ∈ K − K p the polynomial xp − α is irreducible and

the extension K[x]/(xp − α) is inseparable, implying that K is not perfect. Now suppose
n
K = K p and let f ∈ K[x] be irreducible. By Corollary 4.5, we havef (x) = g(xp ) for some
separable g ∈ K[x] and n ≥ 0. If n > 0 then
n n−1
f (x) = g(xp ) = g̃(xp )p ,

where g̃ is the polynomial obtained from g by replacing each coefficient with its pth root
(thus g̃(x)p = g(xp ), since we are in characteristic p). But this contradicts the irreducibility
of f . So n = 0 and f = g is separable. The fact that every irreducible polynomial in K[x]
is separable implies that every algebraic extension of K is separable, so K is perfect.

Corollary 4.20. Every finite field is a perfect field.

n n−1
Proof. If a field K has cardinality pn then #K × = pn − 1, thus α = αp = (αp )p for all
α ∈ K and every element of K is a pth power.

Definition 4.21. A field K is separably closed if K has no nontrivial finite separable ex-
tensions. Equivalently, K is equal to its separable closure in any algebraic closure of K.

Definition 4.22. An algebraic extension L/K is purely inseparable if [L : K]s = 1.

Remark 4.23. The trivial extension K/K is both separable and purely inseparable (but
not inseparable!); conversely, an extension that is separable and purely inseparable is trivial.

Example 4.24. If K = Fp (t) and L = K[x]/(xp − t) = Fp (t1/p ), then L/K is a purely

inseparable extension of degree p.

Proposition 4.25. Let K be a field of characteristic p > 0. If L/K is purely inseparable of

degree p then L = K(a1/p ) ' K[x]/(xp − a) for some a ∈ K − K p .

Proof. Every α ∈ L − K is inseparable over K, and by Corollary 4.5 its minimal polynomial
over K is of the form f (x) = g(xp ) with f monic. We have 1 < deg f ≤ [L : K] = p, so
g(x) must be a monic polynomial of degree 1, which we can write as g(x) = x − a. Then
f (x) = xp − a, and we must have a 6∈ K p since f is irreducible (a difference of pth powers
can be factored). We have [L : K(α)] = 1, so L = K(α) ' K[x]/(xp − a) as claimed.

Theorem 4.26. Let L/K be an algebraic extension and let F be the separable closure of K
in L. Then L/F is purely inseparable.

Proof. If L/K is separable then L = F the theorem holds, so we assume otherwise, in

which case the characteristic p of K must be nonzero. Fix an algebraic closure K of K that
contains L. Let α ∈ L − F have minimal polynomial f over F . Use Corollary 4.5 to write
n
f (x) = g(xp ) with g ∈ F [x] irreducible and separable, and n ≥ 0. We must have deg g = 1,
since otherwise the roots of g would be separable over F , and therefore over K, but not lie

18.785 Fall 2021, Lecture #4, Page 5

 n
in in the separable closure F of K in L. Thus f (x) = xp − a for some a ∈ F (since f is
monic and deg g = 1). Since we are in characteristic p > 0, we can factor f in F (α)[x] as
n n n
f (x) = xp − αp = (x − α)p .

There is thus only one F -homomorphism from F (α) to K. The same statement applies to
any extension of F obtained by adjoining any set of elements of L (even an infinite set).
Therefore # HomF (L, K) = 1, so [L : F ]s = 1 and L/F is purely inseparable.

Corollary 4.27. Every algebraic extension L/K can be uniquely decomposed into a tower
of algebraic extensions L/F/K with F/K separable and L/F purely inseparable.

Proof. By Theorem 4.26, we can take F to be the separable closure of K in L, and this is
the only possible choice, since we must have [L : F ]s = 1.

Corollary 4.28. The inseparable degree of any finite extension of fields is a power of the
characteristic.

Proof. This follows from the proof of Theorem 4.26.

4.2 Étale algebras

We now want to generalize the notion of a separable field extension. By Theorem 4.12,
every finite separable extension L/K can be explicitly represented as L = K[x]/(f ) for
some separable irreducible f ∈ K[x]. If f is not irreducible then we no longer have a field,
but we do have a ring K[x]/(f ) that is also a K vector space, in which the ring multiplication
is compatible with scalar multiplication. In other words, L is a (unital) commutative K-
algebra whose elements are all separable over K. The notion of separability extends to
elements of a K-algebra (even non-commutative ones): an element is separable over K if
and only it is the root of some separable polynomial in K[x] (in which case its minimal
polynomial must be separable). Recall that the minimal polynomial of an element α of a
K-algebra A is the monic generator of the kernel of the K-algebra homomorphism K[x] → A
defined by x 7→ α; note that if A is not a field, minimal polynomials need not be irreducible.
It follows from the Chinese remainder theorem that if f is separable then the K-algebra
K[x]/(f ) is isomorphic to a direct product of finite separable extensions of K. Indeed, if
f = f1 · · · fn is the factorization of f into irreducibles in K[x] then

K[x] K[x] K[x] K[x]

= ' × ··· × ,
(f ) (f1 · · · fn ) (f1 ) (fn )

where the isomorphism is both a ring isomorphism and a K-algebra isomorphism. The
separability of f implies that the fi are separable and the ideals (fi ) are pairwise coprime
(this justifies our application of the Chinese remainder theorem). We thus obtain a K-
algebra that is isomorphic to finite product of separable field extensions K[x]/(fi ) of K.
Algebras of this form are called étale algebras (or separable algebras).

Definition 4.29. Let K be a field. An étale K-algebra is a K-algebra L that is isomorphic

to a finite product of separable field extensions of K. The dimension of an étale K-algebra
is its dimension as a K-vector space. When this dimension is finite we say that L is a finite
étale K-algebra. A homomorphism of étale K-algebras is a homomorphism of K-algebras
(which means a ring homomorphism that commutes with scalar multiplication).

18.785 Fall 2021, Lecture #4, Page 6

Remark 4.30. One can define the notion of an étale A-algebra for any noetherian domain A
(we will consider this in a later lecture).

Example 4.31. If K is a separably closed field then every étale K-algebra A is isomorphic
to K n = K × · · · × K for some positive integer n (and therefore a finite étale K-algebra).

Étale algebras are semisimple algebras. Recall that a (not necessarily commutative) ring
R is simple if it is nonzero and has no nonzero proper (two-sided)Qideals, and R is semisimple
if it is isomorphic to a nonempty finite product of simple rings Ri .1 A commutative ring
is simple if and only if it is a field, and semisimple if and only if it is isomorphic to a finite
product of fields; this applies in particular to commutative semisimple K-algebras. Every
étale K-algebra is thus semisimple (but the converse does Q not hold).
The ideals of a semisimple commutative ring R = ni=1 Ri are easy to describe; each
corresponds to a subproduct. To see this, note that the projection maps R → Ri are
surjective homomorphisms onto a simple ring, thus for any R-ideal I, its image in Ri is
either the zero ideal or the whole ring (note that the image of an ideal under a surjective
ring homomorphism is an ideal). In particular, for each index i, either every (r1 , . . . , rn ) ∈ I
has ri = 0 or some (r1 , . . . , rn ) ∈ I has ri = 1; it follows that I is isomorphic to the product
of the Ri for which I projects onto Ri .
Q
Proposition 4.32. Let A = Ki be a K-algebra written that is a product of field exten-
sions Ki /K. Every surjective homomorphism ϕ : A → B of K-algebras corresponds to the
projection of A on to a subproduct of its factors.
Q
Proof. The ideal ker ϕ is a subproduct of Ki , thus A ' ker ϕ × im ϕ and B = im ϕ is
isomorphic to the complementary subproduct.

Proposition 4.32 can be viewed as a generalization of the fact that every surjective
homomorphism of fields is an isomorphism.

Corollary 4.33. The decomposition of an étale algebra into field extensions is unique up to
permutation and isomorphisms of factors.

Proof. Let A be an étale K-algebra and suppose A is isomorphic (as a K-algebra) to two
products of field extensions of K, say
m
Y n
Y
Ki ' A ' Lj .
i=1 j=1
Q
ComposingQ with isomorphisms yields surjective K-algebra homomorphisms πi : Lj → K i
and πj : Ki → Lj . Proposition 4.32 then implies that each Ki must be isomorphic to one
of the Lj and each Lj must be isomorphic to one of the Ki (and m = n).

Our main interest in étale algebras is that they naturally arise from (and are stable
under) base change, a notion we now recall.

Definition 4.34. Let ϕ : A → B be a homomorphism of rings (so B is an A-module),

and let M be any A-module. The tensor product of A-modules M ⊗A B is a B-module
(with multiplication defined by b(m ⊗ b0 ) := m ⊗ bb0 ) called the base change (or extension of
scalars) of M from A to B. If M is an A-algebra then its base change to B is a B-algebra.
1
There are many equivalent (and a few inequivalent) definitions, but this is the simplest.

18.785 Fall 2021, Lecture #4, Page 7

 We have already seen one example of base change: if M is an A-module and p is a prime
ideal of A then Mp = M ⊗A Ap (this is another way to define the localization of a module).

Remark 4.35. Each ϕ : A → B determines a functor from the category of A-modules to

the category of B-modules via base change. It has an adjoint functor called restriction of
scalars that converts a B-module M into an A-module by the rule am = ϕ(a)m (if ϕ is
inclusion this amounts to restricting the scalar multiplication by B to the subring A).

The ring homomorphism ϕ : A → B will often be an inclusion, in which case we have a

ring extension B/A (we may also take this view whenever ϕ is injective, which is necessarily
the case if A is a field). We are specifically interested in the case where B/A is a field
extension and M is a finite étale A-algebra.

Proposition 4.36. Suppose L is a finite étale K-algebra and K 0 /K is any field extension.
Then L ⊗K K 0 is a finite étale K 0 -algebra of the same dimension as L.

Proof. Without loss of generality we assume that L is actually a field; if not L is a product
of fields and we can apply the following argument to each of its factors.
By Theorem 4.12, L ' K[x]/(f ) for some separable f ∈ K[x], and if f = f1 f2 · · · fm is
the factorization of f in K 0 [x], we have isomorphisms of K 0 -algebras
Y
L ⊗K K 0 ' K 0 [x]/(f ) ' K 0 [x]/(fi ),
i

in which each factor K 0 [x]/(fi ) is a finite separable extension of K 0 (as discussed above, this
follows from the CRT because f is separable). Thus L ⊗K K 0 is a finite étale K 0 -algebra,
and dimK L = deg f = dimK 0 K 0 [x]/(f ), so the dimension is preserved.

Example 4.37. Any finite dimensional real vector space V is a finite étale R-algebra (with
coordinate-wise multiplication with respect to some basis); the complex vector space V ⊗R C
is then a finite étale C-algebra of the same dimension.

Note that even when an étale K-algebra L is a field, the base change L ⊗K K 0 will often
not be a field. For example, if K = Q and L 6= Q is a number field, then L ⊗K C will never
be a field, it will be isomorphic to a C-vector space of dimension [L : K] > 1.

Remark 4.38. In the proof of Proposition 4.36 we made essential use of the fact that the
elements of an étale K-algebra are separable. Indeed, the proposition does not hold if L is
a finite semisimple commutative K-algebra that contains an inseparable element.

Corollary 4.39. Let L ' K[x]/(f ) be a finite separable extension of a field K defined by
an irreducible separable polynomial f ∈ K[x]. Let K 0 /K be any field extension, and let
f = f1 · · · fm be the factorization of f into distinct irreducible polynomials fi ∈ K 0 [x]. We
have an isomorphism of finite étale K 0 -algebras
Y
L ⊗K K 0 ' K 0 [x]/(fi )
i

where each K 0 [x]/(fi ) is a finite separable field extension of K 0 .

Proof. This follows directly from the proof of Proposition 4.36.

18.785 Fall 2021, Lecture #4, Page 8

 The following proposition gives several equivalent characterizations of finite étale alge-
bras, including a converse to Corollary 4.39 (provided the field K is not too small). Recall
that an element α of a ring is nilpotent if αn = 0 for some n, and a ring is reduced if it
contains no nonzero nilpotents.

Theorem 4.40. Let L be a commutative K-algebra of finite dimension and assume that the
dimension of L is less than the cardinality of K. The following are equivalent:

(a) L is a finite étale K-algebra.

(b) Every element of L is separable over K.
(c) L ⊗K K 0 is reduced for every extension K 0 /K.
(d) L ⊗K K 0 is semisimple for every extension K 0 /K.
(e) L = K[x]/(f ) for some separable f ∈ K[x].

The implications (a) ⇔ (b) ⇔ (c) ⇔ (d) ⇐ (e) hold regardless of the dimension of L.
Qn
Proof. To show (a) ⇒ Qn(b), let L = i=1 Ki with each Ki /K separable, and consider α =
(α1 , . . . , αn ) ∈ L = i=1 Ki . Each αi ∈ Ki is separable over K with separable minimal
polynomial fi ∈ K[x], and α is a root of f := lcm{f1 , . . . , fn }, which is separable (the LCM
of a finite set of separable polynomials is separable), thus α is separable.
To show (b) ⇒ (c), note that if α ∈ L is nonzero and separable over K it cannot be
nilpotent (the minimal polynomial of a nonzero nilpotent is xn for some n > 1 and is
therefore not separable), and separability is preserved under base change.
The equivalence (c) ⇔ (d) follows from Lemma 4.42 below.
To show (d) ⇒ (a), we first note we can assume L is semisimple (take K 0 = K), and it
suffices to treat the case where L is a field. By base-changing to the separable closure of K
in L, we can further reduce to the case that L/K is a purely inseparable field extension. If
L = K we are done. Otherwise we may pick an inseparable α ∈ L, and, as in the proof of
n
Theorem 4.26, the minimal polynomial of α has the form f (x) = xp − a for some a ∈ K
and n ≥ 1. Now consider
γ := α ⊗ 1 − 1 ⊗ α ∈ L ⊗K L
n n n
We have γ 6= 0, since γ ∈ / K, but γ p = αp ⊗ 1 − 1 ⊗ αp = a ⊗ 1 − 1 ⊗ a = 0, so γ is a
nonzero nilpotent and L ⊗K L is not reduced, contradicting (c) ⇔ (d).
We have (e) ⇒ (a) form Corollary 4.39. For the converse, suppose L = ni=1 Li with
Q
each Li /K a finite separable extension of K. Pick a monic irreducible separable polynomial
f1 (x) so that L1 ' K[x]/(f1 (x)), and then do the same for i = 2, . . . , n ensuring that each
polynomial fj we pick is not equal to fi for any i < j. This can be achieved by replacing
fj (x) with fj (x + a) for some a ∈ K × if necessary. Here we use the fact that there are at
least n distinct choices for a, under our assumption that the dimension of L is less than the
cardinality of K (note that if f (x) is irreducible then the polynomials f (x+a) are irreducible
and pairwise coprime as a ranges over K). The polynomials f1 , . . . fn are then coprime and
separable, so their product f is separable and L = K[x]/(f ), as desired.

Remark 4.41. K-algebras of the form L = K[x]/(f (x)) are monogenic (generated by one
element). Theorem 4.40 implies that finite étale K-algebras are monogenic whenever the
base field K is big enough. This always holds if K is infinite, but if K is a finite field then
not every finite étale K-algebra is monogenic. The recent preprint [5] gives exact bounds
on the maximal number of generators needed for a finite étale K-algebra over a finite field.

18.785 Fall 2021, Lecture #4, Page 9

 The following lemma is a standard exercise in commutative algebra that we include for
the sake of completeness.
Lemma 4.42. Let K be a field. A commutative K-algebra of finite dimension is semisimple
if and only if it is reduced.
Proof. If A is semisimple it is clearly reduced (otherwise we could project a nonzero nilpotent
of A to a nonzero nilpotent in a field); we only need to prove the converse. Every ideal of a
commutative K-algebra A is also a K-vector space; this implies that when dimK A is finite
A satisfies both the ascending and descending chain conditions and is therefore noetherian
and artinian. This implies that A has finitely many maximal ideals M1 , . . . Mn and that the
intersection of these ideals (the radical of A) is equal to the set of nilpotent elements of A
(the nilradical of A); see Exercises 19.12 and 19.13 in [1], for example.
Taking the productL of the projection maps A  A/Mi yields a surjective ring ho-
n
momorphism ϕ : A  i=1 A/Mi from A to a product of fields. If A is reduced then
ker ϕ = ∩Mi = {0} and ϕ is an isomorphism, implying that A is semismiple.

Proposition 4.43. Suppose L is a finite étale K-algebra and Ω is a separably closed field
extension of K. There is an isomorphism of finite étale Ω-algebras
∼
Y
L ⊗K Ω −→ Ω
σ∈HomK (L,Ω)

that sends β ⊗ 1 to the vector (σ(β))σ for each β ∈ L.

Proof. We may reduce to the case that L = K[x]/(f ) is a separable field extension, and
we may then factor f (x) = (x − α1 ) · · · (x − αn ) over Ω, with the αi are distinct. We have
a bijection between HomK (K[x]/(f ), Ω) and the set {αi }: each σ ∈ HomK (K[x]/(f ), Ω)
is determined by σ(x) ∈ {αi }, and for each αi , the map x 7→ αi determines a K-algebra
homomorphism σi ∈ HomK (K[x]/(f ), Ω). As in the proof of Proposition 4.36 we have
Ω-algebra isomorphisms
n n
K[x] ∼ Ω[x] ∼
Y Ω[x] ∼ Y
⊗K Ω → → → Ω.
(f ) (f ) (x − αi )
i=1 i=1

which map
x ⊗ 1 7→ x 7→ (α1 , . . . , αn ) 7→ (σ1 (x), . . . , σn (x)).
The element x ⊗ 1 generates L ⊗K Ω as an Ω-algebra, and it follows that β ⊗ 1 7→ (σ(β))σ
for every β ∈ L.

Remark 4.44. The proof of Proposition 4.43 does not require Ω to be separably closed. If
L ' K[x]/(f ) as in Theorem 4.40 (with f not necessarily irreducible), we can take Ω to be
any extension of K that contains the splitting field of f .
Example 4.45. Let L/K = Q(i)/Q and Ω = C. We have Q(i) ' Q[x]/(x2 + 1) and
Q[x] C[x] C[x] C[x]
Q(i) ⊗Q C ' 2
⊗Q C ' 2 ' × ' C × C.
(x + 1) (x + 1) (x − i) (x + i)
As C-algebra isomorphisms, the corresponding maps are determined by

i ⊗ 1 7→ x ⊗ 1 7→ x 7→ (x, x) ≡ (i, −i) 7→ (i, −i).

18.785 Fall 2021, Lecture #4, Page 10

Taking the base change of Q(i) to C lets us see the two distinct embeddings of Q(i) in C,
which are determined by the image of i. Note that Q(i) is canonically embedded in its base
change Q(i) ⊗Q C to C via α 7→ α ⊗ 1. We have
−1 = i2 = (i ⊗ 1)2 = i2 ⊗ 12 = −1 ⊗ 1 = −(1 ⊗ 1)
Thus as an
 isomorphism  of C-algebras, the basis (1 ⊗ 1, i ⊗ 1) for Q(i) ⊗Q C is mapped to
the basis (1, 1), (i, −i) for C × C. For any (α, β) ∈ C × C, the inverse image of

α+β α−β
(α, β) = (1, 1) + (i, −i)
2 2i
in Q(i) ⊗ C under this isomorphism is
α+β α−β α+β α−β
(1 ⊗ 1) + (i ⊗ 1) = 1 ⊗ +i⊗ .
2 2i 2 2i
Now R/Q is an extension of rings, so we can also consider the base change of the Q-algebra
Q(i) to R. But note that R is not separably closed and in particular, it does not contain a
subfield isomorphic to Q(i), thus Proposition 4.43 does not apply. Indeed, as an R-module,
we have Q(i) ⊗Q R ' R2 , but as an R-algebra, Q(i) ⊗Q R ' C 6' R2 .

4.3 Norms and traces

We now introduce the norm and trace map associated to a finite free ring extension B/A.
These are often defined only for field extensions, but in fact the same definition works without
modification whenever B is a free A-module of finite rank. One can generalize further to
projective modules (with some restrictions), but we will not need this.
Definition 4.46. Let B/A be a (commutative) ring extension in which B is a free A-module
of finite rank. The (relative) norm NB/A (b) and trace TB/A (b) of b (down to A) are the
determinant and trace of the A-linear multiplication-by-b map B → B defined by x 7→ bx.
As a special case, note that if A is a field and B is a finite A-algebra (a field extension,
for example) then B is an A-vector space of finite dimension, hence a free A-module of finite
rank. In practice one computes the norm and trace by picking a basis for B as an A-module
and computing the matrix of the multiplication-by-b map with respect to this basis; this is
an n × n matrix with entries in A whose determinant and trace are basis independent.
It follows immediately from the definition that NB/A is multiplicative, TB/A is additive,
we have group homomorphisms
NB/A : B × → A× and TB/A : B → A,
and if B1 /A and B2 /A are two ring extensions that are free A-modules of finite rank then
NB1 ×B2 /A (x) = NB1 /A (x1 )NB2 /A (x2 ) and TB1 ×B2 /A = TB1 /A (x1 ) + TB2 /A (x2 )
for all x = (x1 , x2 ) ∈ B1 × B2 .
Example 4.47. Consider A = R and B = C, which has the A-module basis (1, i). For
×b
b = 2 + 3i the matrix of B → B with respect to this basis can be written as 23 −3


2 , thus
 
2 −3
NC/R (2 + 3i) = det = 13,
3 2
 
2 −3
TC/R (2 + 3i) = tr = 4.
3 2

18.785 Fall 2021, Lecture #4, Page 11

Warning 4.48. In order to write down the matrix of an A-linear transformation B → B
with respect to basis for B as a free A-module of rank n, we not only need to pick a basis,
we need to decide whether to represent elements of B ' An as row vectors with linear
transformations acting via matrix multiplication on the right, or as column vectors with
linear transformations acting via matrix multiplication on the left. The latter convention is
often implicitly assumed in the literature (as in the example above), but the former is often
used in computer algebra systems (such as Magma).
We now verify that the norm and trace are well behaved under base change.
Lemma 4.49. Let B/A be ring extension with B free of rank n over A, and let ϕ : A → A0
be a ring homomorphism. The base change B 0 = B ⊗A A0 of B to A0 is a free A0 -module of
rank n, and for every b ∈ B we have
ϕ(NB/A (b)) = NB 0 /A0 (b ⊗ 1) and ϕ(TB/A (b)) = TB 0 /A0 (b ⊗ 1).
Proof. Let b ∈ B, let (b1 , . . . , bn ) be a basis for B as an A-module, and let M = (mij ) ∈ An×n
×b
be the matrix of B → B with respect to this basis. Then (b1 ⊗ 1, . . . , bn ⊗ 1) is a basis for
B 0 as an A0 -module (thus B 0 is free of rank n over A0 ) and M 0 = (ϕ(mij )) ∈ A0n×n is the
×b⊗1
matrix of B 0 → B 0 , and we have
ϕ(NB/A (b)) = ϕ(det M ) = det M 0 = NB 0 /A0 (b ⊗ 1)
ϕ(TB/A (b)) = ϕ(tr M ) = tr M 0 = NB 0 /A0 (b ⊗ 1)
Theorem 4.50. Let K be a field with separable closure Ω and let L be a finite étale K-
algebra. For all α ∈ L we have
Y X
NL/K (α) = σ(α) and TL/K (α) = σ(α).
σ∈HomK (L,Ω) σ∈HomK (L,Ω)

Proof. Let n be the rank of L as a K-module. By the previous lemma and Proposition 4.43,
n
Y
NL/K (α) = N(L⊗K Ω)/Ω (α ⊗ 1) = NΩn /Ω (σ1 (α), . . . , σn (α)) = σi (α).
i=1

The isomorphism L ⊗K Ω → σ Ω = Ωn of Prop. 4.43 sends α ⊗ 1 to (σ1 (α), . . . , σn (α)).

Q
Using the standard basis for Ωn , the matrix of multiplication-by-(σ1 (α), . . . , σn (α)) is just
the diagonal matrix with σi (α) in the ith diagonal entry. Similarly,
n
X
TL/K (α) = T(L⊗K Ω)/Ω (α ⊗ 1) = TΩn /Ω (σ1 (α), . . . , σn (α)) = σi (α).
i=1

The proof above demonstrates a useful trick: when working over a field that is not
algebraically/separably closed, base change to an algebraic/separable closure. This often
turns separable field extensions into étale algebras that are no longer fields.
Proposition 4.51. Let L/K be a (not necessarily separable) finite extension, let K be an
×
Qdcontaining L. Let α ∈ L have minimal polynomial f ∈ K[x] with
algebraic closure of K
factorization f (x) = i=1 (x − αi ) in K[x], and let e = [L : K(α)]. We have
d
Y d
X
NL/K (α) = αie and TL/K (α) = e αi .
i=1 i=1
Pd i
In particular, if f (x) = i=0 ai x , then NL/K (α) = (−1)de ae0 and TL/K (α) = −ead−1 .

18.785 Fall 2021, Lecture #4, Page 12

Proof. See Problem Set 2.

Corollary 4.52. Let A be an integrally closed domain with fraction field K and let L/K be
a finite extension. if α ∈ L is integral over A then NL/K (α) ∈ A and TL/K (α) ∈ A.

Proof. This follows immediately from Propositions 1.28 and 4.51.

Theorem 4.53 (Transitivity of Norm and Trace). Let A ⊆ B ⊆ C be rings with C

free of finite rank over B and B free of finite rank over A. Then C is free of finite rank over
A and
NC/A = NB/A ◦ NC/B and TC/A = TB/A ◦ TC/B .

Proof. See [3, §III.9.4].

References
[1] Allen Altman and Steven Kleiman, A term of commutative algebra, Worldwide Center
of Mathematics, 2013.

[2] Michael Artin, Algebra, 2nd edition, Pearson, 2010.

[3] Nicolas Bourbaki, Algebra I: Chapters 1–3 , Springer, 1989.

[4] Nicolas Bourbaki, Algebra II: Chapters 4–7 , Springer, 1989.

[5] Uriya First, Zinovy Reichstein, Santiago Salazar, On the number of generators of a
separable algebra over a finite field , arXiv:1709.06982, 2017.

[6] Anthony W. Knapp, Advanced Algebra, Digital Second Edition, 2016.

[7] Joseph J. Rotman, Advanced Modern Algebra, 2nd edition, Graduate Studies in Mathe-
matics 114, AMS, 2010.

18.785 Fall 2021, Lecture #4, Page 13