Groups and Fields

Lecture Notes

Lior Silberman
 These are rough notes for the fall 2009 course. Problem sets and solutions were posted on an
internal website.
 Contents

Chapter 1. Introduction 5
1.1. About the course 5
1.2. Problems to be discussed 5
1.3. Definitions 6
Math 422/501: Problem set 1 (due 16/9/09) 11

Chapter 2. Group theory 12

2.1. Group actions 12
Math 422/501: Problem set 2 (due 23/9/09) 14
2.2. p-groups 17
2.3. Sylow Subgroups 17
Math 422/501: Problem set 3 (due 30/9/09) 19
2.4. Solvable groups 21
2.5. Continuation (30/9/09) 22
2.6. Sn and An (30/9/09) 22
2.7. Omitted 23
Math 422/501: Problem set 4 (due 7/10/09) 24

Chapter 3. Fields and Field extensions 25

3.1. Rings of Polynomials (5/10/09) 25
Math 422/501: Problem set 5 (due 14/10/09) 28
3.2. Field extensions 31
Math 422/501: Problem set 6 (due 21/10/09) 33
3.3. Straightedge and compass 35
Math 422/501: Problem set 7 (due 28/10/09) [extended till 30/10/09] 36

Chapter 4. Monomorphisms, Automorphisms and Galois Theory 39

4.1. Splitting fields and normal extensions 39
4.2. Separability 40
4.3. Automorphism Groups 41
Math 422/501: Problem set 8 (due 4/11/09) 43
4.4. The group action 45
4.5. Galois groups and the Galois correspondence (2/11/09) 45
4.6. Examples and applications 46
Math 422/501: Problem set 9 (due 13/11/09) 48
4.7. Solubility by radicals 50
Math 422/501: Problem set 10 (due 18/11/09) 53
4.8. The group action and Grothendieck’s Galois Correspondence 55
3
Chapter 5. Algebraic Number Theory 56
5.1. Intro (16-18/11/09) 56
Math 422/501: Problem set 11 (due 25/11/09) 57
5.2. Integrality and Integral basis (23-25/11/09) 59
5.3. Unique factorization (30/11/09) 60
5.4. Splitting of primes (2/12/09) 61
Bibliography 63
Bibliography 63

4
 CHAPTER 1

Introduction

Lior Silberman, [email protected], http://www.math.ubc.ca/~lior

Office: Math Building 229B
Phone: 604-827-3031

1.1. About the course

Course plan.

1.2. Problems to be discussed

Classification of groups.
Duplicating the cube, trisecting the angle, squaring the circle.
Insolubility of the quintic.
Cyclotomic extensions.
5
 1.3. Definitions
1.3.1. Set Theory.
N OTATION 1. We write 0/ for the empty set, [n] = {0, . . . , n − 1} for the standard set of size n.
N OTATION 2. For a set A write:
[ def \ def
A = {x | ∃y ∈ A : x ∈ y} , A = {x | ∀y ∈ A : x ∈ y} ,

P(A) = {a | a ⊆ A} .
For two sets A, B we write A ∪ B, A ∩ B for {A, B} and {A, B} respectively. Also write
S T

def def
A \ B = {x ∈ A | x ∈
/ B} A∆B = (A \ B) ∪ (B \ A) ,
def
A × B = {x | ∃a ∈ A, b ∈ B : x = (a, b)} .
D EFINITION 3. A relation on a set S is any subset R ⊂ S × S. We write xRy for (x, y) ∈ R, and
for A ⊂ S also R[A] = {y | ∃x ∈ A : (x, y) ∈ R}. We call a relation:
(1) Reflexive if ∀x ∈ S : xRx;
(2) Symmetric if ∀x, y ∈ S : xRy ↔ yRx;
(3) Transitive if ∀x, y, z ∈ S : (xRy ∧ yRz) → xRz;
If S0 ⊂ S we write R S0 for the induced relation R ∩ S0 × S0 .
D EFINITION 4. A relation which is reflexive and transitive is called a partial order. A partial
order is called a linear order if for every distinct x, y ∈ S exactly one of xRy and yRx holds. A
subset of a A partially ordered set S is called a chain if R A is a linear order on A.
If (S, ≤) is a partial order and A ⊂ S we say m ∈ S is an upper bound for A if for any a ∈ A we
have a ≤ m. We say m ∈ S is maximal if for any m0 ∈ S such that m ≤ m0 we have m = m0 . Note
that maximal elements are not necessarily upper bounds for S.
A XIOM 5. (Zorn’s Lemma) Let (S, ≤) be a partial order such that every chain in S has an
upper bound. Then S has maximal elements.
D EFINITION 6. A function is a set f of ordered pairs such that ∀x, y, y0 ((x, y) ∈ f ∧ (x, y0 ) ∈ f ) →
y = y0 . For a function f write Dom( f ) = {x | ∃y : (x, y) ∈ f }, Ran( f ) = Im( f ) = {y | ∃x : (x, y) ∈ f }
for its domain and range (image), respectively, and if x ∈ Dom( f ) write f (x) for the unique y such
that (x, y) ∈ f . Say that f is a function from X to Y if Dom( f ) = X and Ran( f ) ⊂ Y , in which case
we write f : X → Y . Write Y X for the set of functions from X to Y .
Given a function f and A ⊂ Dom( f ) write f [A] for the image { f (x) | x ∈ A} and f A for the
restriction {(x, y) ∈ f | x ∈ A}. This is a function with domain A and range f [A].
Say that a function f is injective if ∀x, x0 : ( f (x) = f (x0 )) → (x = x0 ); say that f : X → Y is
surjective if f [X] = Y , bijective if it is injective and surjective.
A XIOM 7. (Axiom of Choice) Let X be a set. Then there exists a function c with domain X such
that for all x ∈ X such that x is non-empty, c(x) ∈ x.
FACT 8. Under the usual axioms of set theory, AC is equivalent to Zorn’s Lemma.
6
 N OTATION 9. Let A be a function with domain I. We write:
[ def [ \ def \
A(i) = Ran(A), A(i) = Ran(A)
i∈I i∈I
and
def
×i∈I A(i) = { f | f is a function with domain I and ∀i ∈ I : f (i) ∈ A(i)} .
Note that the axiom of choice is the following assumption: let A be a function on I such that
for all i ∈ I, A(i) is non-empty. Then ×i∈I Ai is non-empty.
D EFINITION 10. For two sets A, B write |A| ≤ |B| if there exists an injective function f : A → B,
|A| = |B| if these exists a bijection between A and B. Both relations are clearly transitive and
reflexive. The second is clearly symmetric.
T HEOREM 11. (Cantor-Schroeder-Bernstein) |A| ≤ |B| and |B| ≤ |A| together imply |A| = |B|.
(Using Zorn’s Lemma) Given A, B at least one of|A| ≤ |B| and |B| ≤ |A| holds.
N OTATION 12. For sets A, B write BA = { f ⊂ A × B | f is a function with Dom( f ) = A}. For
a set A and a cardinal κ We set κA = {x ∈ P(A) | |x| = κ} (read “A choose κ”).

1.3.2. Groups.
D EFINITION 13. A group is a quadruplet (G, e, ι, ·) where e ∈ G, ι : G → G, · : G × G → G
and:
(1) ∀g, h, k ∈ G : (g · h) · k = g · (h · k) [associative law].
(2) ∀g ∈ G : e · g = g [identity].
(3) ∀g ∈ G : ι(g) · g = e [inverse].
Call the group G Abelian (or commutative) if for all x, y ∈ G, g · h = h · g.
E XAMPLE 14. The symmetric group on X is the set SX of all bijections X → X, with the
composition operation.
N OTATION 15. We then write Sn for S[n] (“the symmetric group on n letters”).
L EMMA 16. Let G be a group, g, h ∈ G. Then g · e = g, g · ι(g) = e, and the equations gx = h
and xg = h have unique solutions. In particular the identity elements and inverses are unique, and
we usually write g−1 for ι(g).
D EFINITION 17. A non-empty subset H ⊂ G is a subgroup if e ∈ H and if ι(H), H · H ⊂ H. In
that case we write H < G, and (H, e, ι H , · H×H ) is a group. The subgroup H is normal (denoted
H C G) if for all g ∈ G, g H = gHg−1 = H.
When H < G write G/H = {gH | g ∈ G}, H\G = {Hg | g ∈ G}, and [G : H] for the cardinality
of either of these sets (the index) of H in G. ι induces a bijection between G/H and H\G.
T HEOREM 18. (Lagrange) Let H < G. Then there is a set-theoretic bijection between H ×G/H
and G. In particular, if |G| is finite then |H| | |G|.
def
L EMMA 19. If N is normal in G we have G/N = N\G and setting aN · bN = = abN defines a
group structure on G/N. We write qN for the map g 7→ gN.
7
 D EFINITION 20. Let H, G be groups. A map f : H → G is a group homomorphism if f (ab) =
f (a) f (b) for all a, b ∈ H. This implies that f (eH ) = eG and that f ◦ ιH = ιG ◦ f . The set of
homomorphisms will be denoted Hom(H, G). The kernel of f ∈ Hom(H, G) is the set Ker( f ) =
{h ∈ H | f (h) = eG }. The image of f is the set Im( f ) = Ran( f ). When N C G the map qH is a
group homomorphism called the quotient map.
FACT 21. Let f ∈ Hom(H, G). Then f is a monomorphism iff it is injective, an epimorphism
iff it is surjective, and an isomorphism iff it is bijective.
L EMMA 22. Let f ∈ HomR (H, G). Then Ker( f ) C H, Im( f ) < G. Moreover there is a unique
isomorphism f¯ : H/ Ker( f ) → Im( f ) such that f¯ ◦ qKer( f ) = f .
Let H, N < G with N normal. Then HN < G, N is normal in HN, H ∩ N is normal in H, and
HN/N ' H/H ∩ N.
Finally, let N C G. Then qN induces an order- and normality-preserving bijection between
subgroups of G containing N and subgroups of G/N. If N < H < G and H C G as well then
G/H ' (G/N)/(N/H).
def T
D EFINITION 23. If X ⊂ G write hXi = {H < G | X ⊂ H} for subgroup generated by X.
E XAMPLE 24. The infinite cyclic groups is the additive group of Z; the finite cyclic groups are
its quotients Cn ' (Z/nZ, +), n ∈ Z≥1 .
L EMMA 25. If x ∈ G then hxi is isomorphic to a cyclic group. The order of hxi is called the
order of x and is equal to the smallest n ≥ 1 such that xn = e.
N OTATION 26. We write Cn for the cyclic group of order n, D2n = C2 n Cn for the dihedral
group of order 2n ({±1} ∈ (Z/nZ)× acting on (Z/nZ, +) by multiplication).
1.3.3. Rings. All rings in this course are commutative unless noted otherwise.
D EFINITION 27. A (commutative) ring is a quintuple (R, 1, 0, +, ·) consisting of a set R, two
elements 0, 1 ∈ R and two binary operations +, · : R × R → R, such that:
(1) (R, 0, +) is an Abelian group;
(2) ∀x, y, z ∈ R : (x · y) · z = x · (y · z) [associative law];
(3) ∀x ∈ R : 1 · x = x · 1 = x [multiplicative identity];
(4) ∀x, y ∈ R : x · y = y · x [commutative law];
(5) ∀x, y, z ∈ R : x · (y + z) = x · y + x · z ∧ (y + z) · x = y · x + z · x [distributive law];
(6) 0 6= 1 [non-degeneracy].
L EMMA 28. Let R be a ring.
(1) The neutral elements are unique.
(2) For any r ∈ R we have 0 · r = r · 0 = 0.
E XAMPLE 29. (Rings)
(1) Z, Q, R, C, Z/mZ.
(2) For a ring R and a set X, the space of functions RX with point-wise operations.
D EFINITION 30. Let R, S be rings. The map f : R → S is a ring homomorphism if:
(1) f (0R ) = 0S .
(2) f (1R ) = 1S .
8
 (3) For all x, y ∈ R, f (x +R y) = f (x) +S f (y).
(4) For all x, y ∈ S, f (x ·R y) = f (x) ·S f (y).
The set of homomorphisms from R to S will be denoted Hom(R, S).
L EMMA 31. Let f ∈ Hom(R, S). Then f is a monomorphism iff it is injective, an epimorphism
iff it is surjective, and an isomorphism iff it is bijective.
D EFINITION 32. Let R be a ring, and let r ∈ R.
(1) Say that r is invertible (or that it is a unit) if these exists r̄ ∈ R such that r · r̄ = r̄ · r = 1R .
Write R× for the set of units.
(2) Say that r is reducible if r = ab for some non-units a, b ∈ R, irreducible otherwise.
(3) Say that r is a zero-divisor if these exists a non-zero s ∈ R such that rs = 0 or sr = 0.
L EMMA 33. Let r ∈ R be invertible. Then it has a unique multiplicative inverse, to be denoted
r−1 from now on. Writing R× for the set of invertible elements,(R× , 1, ·) is a group.
N OTATION 34. Call R× the multiplicative group of R.
D EFINITION 35. An additive subgroup I ⊂ R is an ideal if for all r ∈ R and a ∈ I, ra ∈ I. We
write I C R.
L EMMA 36. There is a unique ring structure on the additive group R/I such that the quotient
map qI : R → R/I is a ring homomorphism.
The kernel of any ring homomorphism is an ideal; if f ∈ Hom(R, S) then there exists a unique
isomorphism f¯ : R/ Ker( f ) → Im( f ) so that f = f¯ ◦ qKer( f ) .
1.3.4. Fields.
D EFINITION 37. We say that a ring R is:
(1) An integral domain if its only zero-divisor is 0R , that is if (0) is a prime ideal.
(2) A field if its only non-unit is 0R , that is if (0) is a maximal ideal..
E XAMPLE 38. Z is an integral domain; Q, R, C are fields. Z/mZ is an integral domain iff m
is prime, in which case it is a field.
L EMMA 39. Let F be a field, R a ring, and f ∈ Hom(F, R). Then f is injective.
1.3.5. Modules (not a pre-requisite). Let R be a ring.
D EFINITION 40. An R-module is a quadruplet (V, 0, +, ·) where (V, 0, +) is an abelian group ,
and · : R ×V → V is such that:
(1) For all v ∈ V , we have 1R · v = v.
(2) For all α, β ∈ R and v ∈ V , α · (β · v) = (αβ ) · v (αβ denotes the product in R).
(3) For all α, β ∈ R and u, v ∈ V , (α + β ) (u + v) = α · u + β · u + α · v + β · v (note that the
RHS is meaningful since addition is associative and commutative).
If V,W are R-modules we call a map f : V → W a homomorphism of R-modules if it is a homomor-
phism of abelian groups such that for all α ∈ R and v ∈ V , f (α · v) = α · f (v). Write HomR (V,W )
for the set of R-module homomorphisms from V to W (the R may be omitted when clear from
context). The kernel and image of a homomorphism are its kernel and image as a map of abelian
groups.
9
 L EMMA 41. Let f ∈ HomR (V,W ). Then f is a monomorphism iff it is injective, an epimorphism
iff it is surjective, and an isomorphism iff it is bijective.
E XAMPLE 42. Let X be a set, R a ring. Then RX has the structure of an R-module under the
diagonal action of R. We usually write Rn for R[n] .
Complex conjugation is an element of HomR (C, C) but not of HomC (C, C).
L EMMA 43. Let V be an R-module. Then for every v ∈ V we have 0R · v = 0.
D EFINITION 44. Let V be an R-module. An subgroup W ⊂ V is an R-submodule if for all
α ∈ R and w ∈ W , αw ∈ W .
L EMMA 45. Let f ∈ HomR (V,W ). Then Ker( f ) ⊂ V and Im( f ) ⊂ W are R-submodules.
1.3.6. Vector spaces. Let F be a field. F-modules are called vector spaces over F; homomor-
phism of F-modules are called F-linear maps.
D EFINITION 46. Let V be a vector space over F, and let S ⊂ V . Say that S is linearly dependent
if there exist r ≥ 1 and finite sequences {vi }ri=1 ⊂ S, {αi }ri=1 ⊂ F with vi pairwise distinct and the
the αi not all zero such that
r
∑ αivi = 0 .
i=1
The empty sum (the case r = 0) is by definition equal to 0. Call S linearly independent if it is not
dependent. Finally, say that v ∈ V depends on S if v is a linear combination of vectors from S (in
particular, 0 depends on every set). Say that S spans a set W ⊂ V if every v ∈ W depends on S and
write Span(S) for the set of vectors depending on S. Finally say that B ⊂ V is a basis if it is linearly
independent and spans V .
L EMMA 47. Span(S) = {W | S ⊂ W ⊂ V and W is a subspace}. In particular the span of S
T

is the minimal subspace containing S.

L EMMA 48. Every vector space has a basis. In fact, every linearly independent set can be
extended to a basis and every spanning set contains a basis. Moreover, any two bases of a vector
space V have the same cardinality, called the dimension of V and denoted dimF V .
T HEOREM 49. Let U,V be vector spaces over F. Let B,C be bases for U,V respectively. Then
HomF (U,V ) is in bijection with {A ∈ MB×C (F) | ∀b ∈ B : # {c ∈ C | Abc 6= 0} < ∞}.
D EFINITION 50. Let V be an n-dimensional vector space over F. Let A : V n → F be a non-zero
alternating form. For ϕ ∈ EndF (V ) we have A ◦ ϕ = cA for some c ∈ F. Set det(ϕ) = c and call it
the determinant of F.
D EFINITION 51. Let x be a formal variable, and let Pϕ (x) = det(xI − ϕ) ∈ F[x] be the char-
acteristic polynomial of F. Then Pϕ (0) = (−1)n det(ϕ). Let t be the coefficient of xn−1 in Pϕ (A).
We set Tr(A) = −t.
T HEOREM 52. Fix a basis {vi }ni=1 for V , and let A ∈ Mn (F) be the matrix associated to ϕ.
Then det(A) = ∑π∈Sn sgn(π) ∏ni=1 Ai,σ (i) and Tr(A) = ∑ni=1 Aii .
T HEOREM 53. (Cayley-Hamilton) Pϕ (ϕ) = 0.

10
 Math 422/501: Problem set 1 (due 16/9/09)
Some group theory
1. (Cyclic groups)
(a) Show that the infinite cyclic group Z is the unique group which has non-trivial proper
subgroups and is isomorphic to all of them.
(b) [optional] which groups have no non-trivial proper subgroups?
2. (Groups with many involutions) Let G be a finite group, and let I = g ∈ G | g2 = e \ {e} be

its subset of involutions (e is the identity element of G).
(a) Show that G is abelian if it has exponent 2, that is if G = I ∪ {e}.
(b) Show that G is abelian if |I| ≥ 43 |G|.

Some polynomial algebra

3. Show that (x −y) divides (xn −yn ) in Z[x, y]. Conclude that for any ring R, polynomial P ∈ R[x]
and element a ∈ R such that P(a) = 0 one has (x − a)|P in R[x].
4. Let R be an integral domain, P ∈ R[x], {ai }ki=1 ⊂ R distinct zeroes of P. Show that ∏i (x − ai )|P
in R[x]. Give a counterexample when R has zero-divisors.
j−1
5. Let Vn (x1 , . . . , xn ) ∈ Mn (Z[x1 , . . . , xn ]) be the Vandermonde matrix (Vn )i j = xi . Let Vn (x) =
det (Vn (x)) ∈ Z[x]. Show that there exists cn ∈ Z so that Vn (x) = cn ∏i> j (xi − x j ).
Hint: Consider Vn as an element of (Z[x1 , . . . , xn−1 ]) [xn ].
6. Setting xn = 0 show that cn = cn−1 , hence that cn = 1 for all n.
Some abstract nonsense
D EFINITION . Let G, H be groups, and let f : G → H be a homomorphism. Say that f is a
monomorphism if for every group K and every two distinct homomorphisms g1 , g2 : K → G, the
compositions f ◦ g1 , f ◦ g2 : K → H are distinct. Say that f is an epimorphism if for every group K
and every two distinct homomorphisms g1 , g2 : H → K the compositions gi ◦ f : G → K are distinct.

7. Show that a homomorphism of groups is a monomorphism iff it is injective, an epimorphism

iff it is surjective.
8. (Variants)
(a) Same as 7, but replace “group” with “vector space over the field F” and “homomorphism”
with “F-linear map”.
(b) Consider now the case of rings and ring homomorphisms. Show that monomorphisms are
injective, but show that there exist non-surjective epimorphisms.
*9. Replacing “groups” with “Hausdorff topological spaces” and “homomorphism” with “contin-
uous map” show that:
(a) A continuous map is a monomorphism iff it is injective.
(b) A continuous map is an epimorphism iff its image is dense.

11
 CHAPTER 2

Group theory

2.1. Group actions

D EFINITION 54. Let G be a group, X a set. A left (right) action of G on X is a map· : G×X → X
(respectively · : X × G → X) so that:
(1) ∀x ∈ G : e · x = x (respectively x · e = x).
(2) ∀g, h ∈ G, x ∈ X : g · (h · x) = x (respectively, (x · g) · h = x.
Let G act on the left on the sets X,Y . A map f : X → Y is said to be G-equivariant if f (g · x) =
g · f (x) for all g ∈ G, x ∈ X.
L EMMA 55. (Group actions) Let G be a group, X a set.
(1) Let · be a left action of G on X. Then (x, g) 7→ g−1 x is a right action. Conversely, let · be
a right action. Then (g, x) 7→ x · g−1 is a left action.
(2) For each g ∈ G the map σg (x) = gx is a permutation of X, and the map g 7→ σg is a
homomorphism G 7→ SX .
(3) Conversely, if σ : G → SX is a homomorphism then (g, x) 7→ (σ (g)) (x) is a left action of
G on X.
D EFINITION 56. Let G be a group. A G-set is a pair (X, ·) where X is a set, · a left action of G
on X. For G-sets X,Y we have HomG (X,Y ) for the set of G-equivariant maps between them.
R EMARK 57. When we talk about a group acting on a structure, we usually assume that the
action preserves the structure.
E XAMPLE 58. (group actions)
(1) For any group G and set X we have the trivial action g · x = x for all g, x.
(2) For any set X, the symmetric group SX acts on X.
(3) For a ring R, the group R× acts on R by multiplication.
(4) Let F be a field, V a vector space, GL(V ) = End(V )× . Then GL(V ) acts on V ,
(5) G = (V, E) a finite graph. AutG = { f ∈ SV | ∀e ∈ E : f (e) ∈ E}.
(6) Let (X, d) be a metric space. Then Isom(X, d) acts on X.
(7) π1 (M) acts on M̃.
E XAMPLE 59. Let G be a group. Then we have the left-regular action (g, h) 7→ gh and the
conjugation action (g, h) 7→ ghg−1 = g h. In both cases G acts on itself.
T HEOREM 60. (Cayley [1]) The left-regular action G 7→ SG is an injective map. In other words,
a group of order n is isomorphic to a subgroup of Sn .
P ROPOSITION 61. The following are group actions:
(1) G  X. Then G acts on P(X) by gA = {ga | a ∈ A}. In fact, for a cardinal α G acts on
X

α .
12
 (2) G  X, Y ⊂ X is G-invariant (gY = Y for all g). Then G  Y .
(3) G a group acting on X,Y . Then Y X is a G-set where for f : X → Y we set (g f ) (x) =
g ·Y f (g−1 ·X x) .


(4) {Xi }i∈I are G-sets. Then their disjoint union i∈I (X × {i}) is a G-set with the action
S

g · (x, i) = (gx, i) for any g ∈ G, i ∈ I, x ∈ Xi . It is a product in the category of G-sets.

(5) {Xi }i∈I are G-sets. Then their Cartesian product ∏i∈I Xi is a G-set with the action (g f ) (i) =
g · ( f (i)) for any g ∈ G, f ∈ ∏i Xi , i ∈ I.
(6) G a group acting on X, H < G. Then (h, x) 7→ hx is an H-action of H on X denoted
ResG H X.
(7) H a group acting on X, H < G. Let IndG −1 f (g) . This is a

H X = f : G → X | f (gh) = h
G-set where (g f ) (g0 ) = f (g−1 g0 ).
D EFINITION 62. Say that a G-set X is transitive or that it is an orbit if for all x, y ∈ X there
exists g ∈ G so that gx = y.
E XAMPLE 63. The G action on G/H is transitive.
Fix a group G and a G-set X.
L EMMA 64. Then the relation x ∼ y ⇐⇒ ∃g ∈ G : gx = y is an equivalence relation. The
equivalence classes are called the orbits of the action. The set of orbits is denoted G\X, and the
orbit of x is denoted O(x) or G · x.
O∈G\X O.
F
C OROLLARY 65. X =
D EFINITION 66. For x ∈ X write Gx = StabG (x) = ZG (x) for the stabilizer subgroup {g ∈ G | gx = x}.
L EMMA 67. Ggx = gGx g−1 = g Gx .
P ROPOSITION 68. Let X be a G-set, x ∈ X. Then f : G · x → G/Gx given by f (gx) = gGx and
h : G/Gx → X given by h(gGx ) = gx are inverse G-morphisms.
T HEOREM 69. Isomorphism classes of transitive G-sets are in bijection with conjugacy classes
of subgroups of G.
C OROLLARY 70. Let X be a G-set, and Flet R ⊂ X be a set of representatives for the orbits.
Then we have an isomorphism of G-sets X ' x∈R G/Gx . In particular, we have the class formula
(2.1.1) |X| = ∑ [G : Gx ] .
x∈R

13
 Math 422/501: Problem set 2 (due 23/9/09)
Direct and semidirect products

1. Let G be a group, and let A, B be subgroups of G so that B is normal and A ∩ B = {e}.

(a) Show that A n B = {a · b | a ∈ A, b ∈ B} is a subgroup of G; and that every element of it
can be uniquely written as a product a · b. We call this subgroup the internal semidirect
product of A, B.
(b) Assuming that A is normal as well show that ab = ba for all a ∈ A, b ∈ B. In that case we
say that the subgroup AB is the internal direct product of A, B.
2. Let G, H be groups. Let G × H = {(g, h) | g ∈ G, h ∈ H} and give it the group structure (g, h) ·
(g0 , h0 ) = (gg0 , hh0 ). Show that this makes G × H into a group (called the direct product of
G, H) and find normal subgroups Ḡ, H̄ < G × H isomorphic to G, H respectively so that G × H
is the internal direct product of Ḡ and H̄.
3. Let G, H be groups and let G act on H by automorphisms (in other words, for each g ∈ G you
are given a group isomorphism αg : H → H such
 that αgh = αg ◦ αh ). Give the set G × H the
group structure (g0 , h0 ) · (g, h) = g0 g, αg−1 (h0 )h . Show that this gives a group structure called
the semidirect product G n H. Show that the semidirect product contains subgroups Ḡ, H̄ with
H̄ normal such that G n H is the internal semidirect product of G, H.

p-Groups

4. Let G be a non-abelian group of order p3 , p a prime. Show that Z(G) has order p and that
G/Z(G) ' C p ×C p .

Cyclic group actions and cycle decompositions

5. Let G be a group acting on a set X, and let g ∈ G. Show that a subset Y ⊂ X is invariant under
the action of the subgroup hgi of G iff gY = Y . When Y is finite show that assuming gY ⊂ Y is
enough.
6. For α ∈ Sn write supp(α) for the set {i ∈ [n] | α(i) 6= i}.
(a) Show that supp(α) is invariant under the action of hαi.
(b) Show that if supp(α) ∩ supp(β ) = 0/ then αβ = β α.

14
7. (Cycle decomposition) Call σ ∈ Sn a cycle if its support is a single orbit of hσ i, in which case
we call the size of the support the length of the cycle.
(a) Let α ∈ Sn , and let O ⊂ [n] be an orbit of hαi of length at least 2. Show that there exists
a unique cycle β ∈ Sn supported on O so that α O = β O (that is, the restrictions of the
functions α, β to the set O are equal).
(b) Let α ∈ Sn and let {βO | O an orbit of hαi} be the set of cycles obtained in part (a). Show
that they all commute and that their product is α.
(c) Show that every element of Sn can be written uniquely as a product of cycles of disjoint
support.
(d) Consider the action of [4]35 = 4 + 35Z ∈ Z/35Z by multiplication on Z/35Z. Decompose
this permutation into a product of cycles.
8. (The conjugacy classes of Sn )
(a) Let α, β ∈ Sn with α a cycle. Show that β αβ −1 is a cycle as well.
(b) Show that α, β ∈ Sn are conjugate iff for each 2 ≤ l ≤ n the number of cycles of length l
in their cycle decomposition is the same.
Hint: Constructs a bijection from [n] to [n] that converts one partition into orbits into the
other.

Affine algebra

D EFINITION 71. Let F be a field, V /F a vector space. An affine combination is a formal sum
∑ni=1 ti vi
where ti ∈ F, vi ∈ V and ∑ni=1 ti = 1. If V,W are vector spaces then a map f : V → W is
called an affine map if for every affine combination in V we have
!
n n
f ∑ tivi = ∑ ti f (vi ) .
i=1 i=1

9. (The affine group) Let U,V,W be vector spaces over F, f : U → V , g : V → W affine maps.
(a) Show that g ◦ f : U → W is affine.
(b) Assume that f is bijective. Show that its set-theoretic inverse f −1 : V → U is an affine
map as well.
(c) Let Aff(V ) denote the set of invertible affine maps from V to V . Show that Aff(V ) is a
group, and that it has a natural action on V .
(d) Assume that f (0U ) = 0V . Show that f is a linear map.
10. (Elements of the affine group)
(a) Given a ∈ V show that Ta x = x + a (“translation by a”) is an affine map.
(b) Show that the map a 7→ Ta is a group homomorphism from the additive group of V to
Aff(V ). Write T(V ) for the image.
(c) Show that T(V) acts transitively on V . Show that the action is simple: for any x ∈ V ,
StabT(V ) (x) = T0 .
(d) Fixing a basepoint 0 ∈ V , show that every A ∈ Aff(V ) can be uniquely written in the form
A = Ta B where a ∈ V and B ∈ GL(V ). Conclude that Aff(V ) = T(V ) · GL(V ) setwise.
(e) Show that T(V ) ∩ GL(V ) = {1} and that T(V ) is a normal subgroup of Aff(V ). Show that
Aff(V ) is isomorphic to the semidirect product GL(V ) n (V, +).

15
 Additional (not for credit)

V /F a vector space of dimension n. Find a formula for

A. Let F be a finite field with q elements,
the Gaussian binomial coefficient nk q , the number of k-dimensional subspaces of V . Show
that this is a polynomial in q and that its limit as q → 1 is the usual binomial coefficient nk .

B. Let F be a field, V a finite-dimensional F-vector space. A flag in V is a nested sequence

{0} = W0 ( W1 ( · · · ( Wk ( Wk+1 = V of subspaces of V . An ordered basis {vi }ni=1 ⊂ V is
said to be adapted to W if for all j, W j = Span {vi }di=1
k
where d j = dimW j .
(a) Show that G = GL(V ) acts on the space of flags.
(b) Find the orbits of the action and find which are isomorphic G-sets. Point stabilizers are
called parabolic subgroups.
(c) Let F be finite (say with q elements). Find the size of each orbit.
Hint: The set of subspaces of V containing W is in bijection with the set of subspaces of
the quotient vector space V /W .
(d) Let B < G be the stabilizer of a maximal flag (“Borel subgroup”). Find the order of B,
hence the order of G.

16
 2.2. p-groups
T HEOREM 72. (Cauchy) Let G be a finite group of order divisible by the prime p. Then G has
an element of order p.
P ROOF. By induction on the order of G we may assume that every proper subgroup of G has
order prime to p. Now if G has a proper normal subgroup N then from |G| = |N| · |G/N|it follows
that G/N has order divisible by p. Let gN have order p in G/N. Let r be the order of g in G. Then
r
the order of gN in G/N divides r, so g p has order p.
Finally, assume G is simple and has no proper subgroups of order divisible by p, and consider
the action of G on itself by conjugation. Every non-central conjugacy class has size divisible by
p (the index of the centralizer is divisible by p). It follows that p | |Z(G)|, and since the center is
normal that by G = Z(G). It follows that G is cyclic group of prime order, and we are done. 
C OROLLARY 73. Let G be a finite group, p a prime. Then every element of G has order a
power of p iff the order of G is a power of p.
D EFINITION 74. Call G a p-group if every element of G has order a power of p.
Observe that if G is a finite p-group then the index of every subgroup is a power of p. It
follows that every orbit of a G-action has either size 1 or size divisible by p. By the class equation
we conclude that if G is a finite p-group and X is a finite G-set, we have:
(2.2.1) |X| ≡ |{x ∈ X | StabG (x) = G}| mod p .
T HEOREM 75. Let G be a finite p-group. Then Z(G) 6= 1.
P ROOF. Let G act on itself by conjugation. The number of conjugacy classes of size 1 must be
divisible by p. 

It follows that all groups of order p2 are abelian, and in fact isomorphic to one of C p and
C p ×C p .

2.3. Sylow Subgroups

R EMARK 76. Pronounce “Silof” not “sigh-low”.
D EFINITION 77. Let G be a group, p a prime number. A Sylow p-subgroup of G is a maximal
(under inclusion) p-subgroup of G.
It is immediate from Zorn’s lemma that Sylow p-subgroups exist, in fact that every p-subgroup
is contained in a maximal one. If G is finite then Cauchy’s Theorem 72 implies that for p| |G| the
Sylow p-subgroups are non-trivial.
T HEOREM 78. (Sylow) Let G be a finite group of order n = pe m where p is a prime, e ≥ 1, and
(p, m) = 1. Them:
(1) All Sylow p-subgroups are conjugate.
(2) Their number is a divisor of m congruent to 1 modulu p.
(3) They all have order pe .
17
 P ROOF. Let F denote the set of Sylow p-subgroups. G acts on this set by conjugation since
the definition of Sylow subgroups is a group automorphism invariant. For P 6= Q ∈ F we have
P ∩ NG (Q) = 0/ (otherwise (P ∩ NG (Q)) n Q would be a p-subgroup of G properly containing Q.
Letting P ∈ F act on F be conjugation it follows that P has exactly one fixed point, P itself. By
equation (2.2.1) it follows that the number of subgroups is congruent to 1 modulu p. In fact, the
same applies to the orbit O = {g P}g∈G . If there was Q ∈ F \ O then Q would act on O without
fixed points, implying that |O| ≡ 0 (p) which is impossible. Claim (1) thus follows, and from
[G : NG (P)] = |F| ≡ 1 (p) we conclude that the number of subgroups is a divisor of m (it is a
divisor of G prime to p) and that pe k |NG (P)|. Regarding claim (3), if NG (P) is a proper subgroup
of G then we are done by induction so it remains to consider the case of P normal in G, in which
case it suffices to show that [G : P] is prime to p. Indeed, were p to divide the order of G/P then
Cauchy’s Theorem would give a subgroup of order p in this group. Its preimage in G would be a
p-subgroup properly containing P, which is not possible. 
E XAMPLE 79. Classification of groups. Let p < q be odd primes.
(1) Groups of order 2p are either cyclic or dihedral: there must be a unique (hence, normal)
subgroup P of order p. Next, Aut(P) contains a unique element of order 2.
(2) D E or (if q ≡ 1 (p)) isomorphic to the semidirect product
Groups of order pq are either cyclic
q−1
x, y | x p = yq = 1, xyx−1 = y p . Indeed a Sylow q-subgroup must be unique, and it
has an automorphism of order p iff p | q − 1.
(3) Groups of order 30 are never simple. Indeed, if the Sylow 5-subgroups are not normal
there are 6 of them, contributing 24 elements of order 5. If the Sylow 3-subgroups are
not normal there are 10 of them, contributing 20 elements of order 3. Since C3 has no
automorphism of order 5 and C5 has no automorphism of order 5, it follows that in either
case the 3-subgroup and 5-subgroup commute element-wise. It follows that the number of
such subgroups is a divisor of 2, hence that G has a normal cyclic subgroup of order 15, on
which an involution acts by an automorphism. There are 4 choices for this automorphism,
giving four isomorphism classes of groups of order 30.
(4) Groups of order 36 are never simple: If there is not one Sylow 3-subgroup there are 4 of
them, say Q1 , . . . , Q4 . The conjugation action on this set gives a homomorphism G → S4 .
It is not injective and the kernel furnishes the desired subgroup.

18
 Math 422/501: Problem set 3 (due 30/9/09)
Groups of small order

1. Let m be a positive integer. Let Cm be the cyclic group of order m. Show that Aut(Cm ) '
(Z/mZ)× .
Hint: Fix a generator g of Cm , and given ϕ ∈ Aut(Cm ) consider ϕ(g).
2. (Quals September 2008) Show that every group of order 765 is Abelian.
Hint: To start with, let G act by conjugation on a normal Sylow p-subgroup.
3. Let G be a group of order 36 and assume that it does not have a normal Sylow 3-subgroup.
Obtain a non-trivial homomorphism G → S4 and conclude that G is not simple.

Index calculations

4. Let G be a group, H < G a subgroup of finite index. Show that there exists a normal subgroup
N C G of finite index such that N ⊂ H.
Hint: You can get inspiration from problem 3.
5. (Normal p-subgroups)
(a) Let G be a finite group, N C G a normal subgroup which is a p-group. Use the conjugacy
of Sylow subgroups to show that N is contained in every Sylow p-subgroup of G.
(b) Now let G be any group, N C G a normal subgroup which is a p-group. Let P < G be
another p-subgroup. Show that PN is a p-subgroup of G and conclude that N is contained
in every Sylow p-subgroup of G.

Commutators

Let G be a group. For x, y ∈ G write [x, y] = xyx−1 y−1 for the commutator of x, y. Write G0 for
the subgroup of G generated by all commutators and call it the derived subgroup of G.
6. (The Abelianization)
(a) Show that x, y ∈ G commute iff [x, y] = e.
(b) Show that G0 is a normal subgroup of G.
Hint: Show that it is enough to show that the set of commutators is invariant under conju-
gation. Then show that g[x, y]g−1 is a commutator.
(c) Show that Gab = G/G0 is abelian.
(d) Let A be an Abelian group, and let f ∈ Hom(G, A). Show that G0 ⊂ Ker f . Conclude that
f can be written uniquely as the composition of the quotient map G  Gab and a map
f ab : Gab → A.
OPTIONAL Let G, H be groups and let f ∈ Hom(G, H). Does f extend to a map Gab → H ab ?

19
7. (Groups of Nilpotence degree 2) Let G be group, Z = Z(G) its center.
(a) Show that the commutator [x, y] only depends on the classes of x, y in G/Z(G).
From now on assume that G is non-Abelian but that A = G/Z is.
(b) Show that G0 < Z(G).
Hint: 6(d).
(c) Show that the commutator map of G descends to an anti-symmetric bi-linear pairing
[·, ·] : A × A → Z(G).

8. Let G be a non-abelian group of order p3 .

(a) Show that Z(G) < G0 .
Hint: 6(b) and general properties of p-groups.
(b) Show that Z(G) = G0 .
Hint: Show that G/Z(G) is abelian and use 7(b).

Optional: Example of a Sylow subgroup

A. Let k be field, V a vector space over k of dimension n. A maximal flag F in V is a sequence

{0} = F0 ( F1 ⊆ · · · ( Fn = V of subspaces. Let F(V ) denote the space of maximal flags in V .
(a) Show that the group GL(V ) of all invertible k-linear maps V → V acts transitively on
F(V ). 
(b) Let F ∈ F(V ) and let B < GL(V ) be its stabilizer. Let N = b ∈ B | ∀ j ≥ 1∀v ∈ Fj : bv − v ∈ Fj−1 .
Show that N is a normal subgroup of B.
(c) Show that B/N ' (k× )n .

B. Assume |k| = q = pr for a prime p. Let V = kn , Let G = GL(V ) = GLn (F), and let B ⊂ G
 k
be the point stabilizer of the standard flag Vk = Span e j j=1 where e j is the jth vector of the
standard basis.
(a) What is |F(F n )|?
Hint: For each one-dimensional subspace W ⊂ V show that the set flags containing W is
in bijection with the set flags F(V /W ).
(b) Show that q is relatively prime to |F(V )|. Conclude that B contains a Sylow p-subgroup
of G.
(c) Show that N is a Sylow p-subgroup of G.

Optional: Infinite Sylow Theory

C. Let G be any group, P < G a p-subgroup of finite index. We will show that Sylow’s Theorems
apply in this setting.
(a) Show that G has a normal p-subgroup N of finite index.
(b) Show that every Sylow p-subgroup contains N.
(c) Deduce a version of Sylow’s Theorem for G from Sylow’s Theorems for G/N.

20
 2.4. Solvable groups
• Composition Series
• Statement of Jordan-Hölder
D EFINITION 80. Call a group G solvable if there exist subgroups G0 = G ⊃ G1 ⊃ G2 ⊃ · · · ⊃
Gk = {e} so that Gi+1 C Gi and Gi /Gi+1 are normal for all 0 ≤ i ≤ k − 1.
R EMARK 81. A finite group is solvable iff its composition factors are cyclic p-groups.
E XAMPLE 82. Every finite p-group is solvable.
P ROOF. The composition factors are simple p-groups. Every such group is a cyclic p-group.

E XAMPLE 83. S3 is solvable.
P ROOF. The subgroup of elements of order 3 is Abelian and of index 2. 
L EMMA 84. Every group of order 12 is solvable, hence S4 is solvable.
P ROOF. Let G have order 12, and let P be its set of 2-Sylow subgroups. |P| ∈ {1, 3} since it
must be an odd divisor of 12. If P has a unique member then G has a normal subgroup of order
4. Otherwise the conjugation action of G on P gives a homomorphismTG → S3 . It is not injective
since |G| = 12 > 6 = |S3 |, and therefore has a non-trivial kernel N = P (the point stabilizer of
each Sylow subgroup is itself since each is a maximal subgroup in our case). N is abelian (it has
order 2 or 4 = 22 ) And G/N is solvable (it either has order 6 = 2 · 3 or 3). 
P ROPOSITION 85. Every group of order p2 q is solvable.

P ROOF. Assume the Sylow p-subgroups are not normal. Then these are P1 , . . . , Pq and q ≡
1 (p). It follows that q − 1 ≥ p. Next, if the Sylow q-Subgroups are not normal then there are p2
of them, and p2 ≡ 1 (q). But then q divides one of p − 1 and p + 1 so q ≤ p + 1. We conclude
q = p + 1, which is only possible if q = 3, p = 2 and |G| = 12. 
FACT 86. (later this week) Sn , n ≥ 5 is not solvable.
P ROPOSITION 87. Let G be a group, H a subgroup, N a normal subgroup.
(1) If G is solvable then so are H and G/N.
(2) If N and G/N are solvable then so is G.
P ROOF. Let {Gi }ki=0 be a series as in the definition. Set Hi = H ∩ Gi , and let h ∈ Hi and
g ∈ Hi+1 . Then hgh−1 ∈ H and ghg−1 ∈ Gi+1 so hgh−1 ∈ Hi+1 . Composing the inclusion Hi ,→ Gi
with the quotient map Gi → Gi /Gi+1 gives a map Hi → Gi /Gi+1 with kernel Hi ∩ Gi+1 = Hi+1 .
It follows that Hi /Hi+1 embeds in Gi /Gi+1 and in particular that it is commutative. Next,  let
q : G → G/N be the quotient map and set N̄i = q(Gi ) = Gi N/N. Then N̄0 = G/N, N̄k = eG/N
and since Gi normalizes Gi+1 and N it normalizes Gi+1 N, so its image N̄i normalizes N̄i+1 . Finally,
the map Gi → N̄i /N̄i+1 is surjective and its kernel contains Gi+1 . It follows that N̄i /N̄i+1 is a
quotient of the abelian group Gi /Gi+1 hence abelian.
Conversely, let N = N0 ⊃ N1 ⊃ · · · ⊃ Nk = {e} and let Ḡ0 = G/N ⊃ Ḡ1 ⊃ · · · ⊃ Ḡl = {eN} be
normal series with abelian quotients in N and G/N, respectively. For 0 ≤ i ≤ l let Gi be the inverse
image of Ḡi , and for i ≤ l ≤ l + k let Gi = Ni−l . This is a normal series and the quotients come
from the two series combined. 
21
 E XAMPLE 88. Every finite p-group is solvable.
P ROOF. Let G be a finite p-group. Then Z(G) is non-trivial and solvable, and G/Z(G) is
solvable by induction. 

2.5. Continuation (30/9/09)

• Hall π-subgroups and π 0 -subgroups; Hall’s Theorem
• Feit-Thompson.
• Classification of FSG.
• Nilpotent groups and commutators.
• Finite groups are nilpotent iff they are direct products of p-groups.

2.6. Sn and An (30/9/09)

L EMMA 89. Let k, l ≥ 0 and let a, b, c1 , · · · , ck , d1 , · · · , dl ∈ [n] be distinct. Then
(a b)(a c1 · · · ck b d1 · · · dk ) = (a c1 · · · ck )(b d1 · · · dk )
(a, b)(a c1 · · · ck )(b d1 · · · dk ) = (a c1 · · · ck b d1 · · · dk ) .
C OROLLARY 90. Sn is generated by the transpositions.
P ROOF. The Lemma shows that every cycle is a product of a transposition and shorter cycles.
Moreover, every element of Sn is a product of cycles. 
D EFINITION 91. For σ ∈ Sn write sgn(σ ) = (−1)n−t where t is the number of cycles in the
decomposition of σ (including fixed points!).
E XAMPLE 92. If β is a cycle of length l then sgn(β ) = (−1)l−1 . Indeed, β has n − l fixed
points.
L EMMA 93. Let a 6= b. Then sgn((a b)σ ) = − sgn(σ ).
P ROOF. Let σ = ∏tj=1 β j be the cycle decomposition of σ , where we may assume a ∈ supp(β1 )
without loss of generality. If b ∈ supp(β1 ) as well then (a b)σ is the product of t + 1 disjoint cycles
by the first part of Lemma 89. Otherwise, we may assume b ∈ supp(β2 ). Then (a b)σ is the product
of t − 1 cycles by the second part of the Lemma. 
P ROPOSITION 94. ∀α, β ∈ Sn : sgn(αβ ) = sgn(α) sgn(β ).
P ROOF. Let T = {α ∈ Sn | ∀β ∈ Sn : sgn(αβ ) = sgn(α) sgn(β )}. By Lemma 93 T contains
all transpositions. T is also closed under multiplication: if α, α 0 ∈ T and β ∈ Sn then:
sgn (αα 0 )β = sgn α(α 0 β )



(associativity)
= sgn(α) sgn(α 0 β ) α ∈T
= sgn(α) sgn(α 0 ) sgn(β ) α0 ∈ T
= sgn(αα 0 ) sgn(β ) α ∈T.
That T = Sn now follows from Corollary 90. 
D EFINITION 95. An = Ker(sgn). It is a normal subgroup of Sn of index 2.
L EMMA 96. (Generation and conjugacy in An ) Let n ≥ 5.
22
 (1) All cycles of length 3 are conjugate in An ..
(2) All elements which are a product of two disjoint transpositions are conjugate in An .
T HEOREM 97. An is simple if n ≥ 5.
P ROOF. Let N C An be normal and non-trivial and let σ ∈ N \ {id} have a maximal num-
ber of fixed points. Assume σ moves at least 1, 2, 3, 4, 5 with σ (1) = 2, σ (2) = 3 and let γ =
(3 4 5)σ (3 4 5)−1 σ −1 ∈ N. Then γ fixes every point that σ does, and also γ(2) = 2 but γ(3) = 4, a
contradiction. Thus either γ is a product of disjoint transpositions or γ = (1 2 3). In the first case
another conjugation trick also gives (1 2 3) ∈ N. We are now done by the Lemma. 
2.7. Omitted
Characteristic subgroups, minimal normal subgroups and the socle, characteristically simple
groups.
Every solvable group has a normal abelian subgroup.

23
 Math 422/501: Problem set 4 (due 7/10/09)
Solvable groups

1. Show the following are equivalent:

(a) Every finite group of odd order is solvable.
(b) Every non-abelian finite simple group is of even order.
Aside: That (a) holds is a famous Theorem of Feit and Thompson (1963).

2. Let F be a field. Let G = GLn (F), let B < G be the subgroup of upper-triangular matrices,
N < B the subgroup of matrices with 1s on the diagonal. Next, for 0 ≤ j ≤ n − 1 write N j for
the matrices
 with 1s on the main
diagonal and 0s on the next When n = 4 we
j diagonals. have:

 1 ∗ ∗ ∗   
 1 0 ∗ ∗   
 1 0 0 ∗  
1 ∗ ∗ 1 0 ∗ 1 0 0 
     
N = N0 =    , N1 =   , N2 =   etc

 1 ∗   

 1 0   

 1 0  
 1   1   1 
.
(a) Show that N C B and that B/N ' (F × )n (direct product of n copies).
(b) For each 0 ≤ j < n − 1, N j+1 C N j and N j /N j+1 ' F n− j−1 (direct products of copies of
the additive group of F).
(c) Conclude that B is solvable.

D EFINITION . Let G bea group.

 The derived series of G is the sequence of subgroups defined
0
by G(0) = G and G(i+1) = G(i) (commutator subgroups).

3. Let G be a group and assume G(k) = {e}. Show that G is solvable.

4. Let G be a solvable group, say with normal series G = G0 B G1 B G2 B · · · B Gk = {e}. Show
that G(i) < Gi for all i. Conclude that G is solvable iff the derived series terminates.
OPTIONAL Let S∞ ⊂ SNSdenote the set of permutations of finite support.
S∞ = n Sn with respect to the natural inclusion of Sn = S[n] in S∞ .
(a) Show that S
(b) Let A∞ = n An with respect to the same inclusion. Show that A∞ is a subgroup of S∞ of
index 2.
(c) Show that A∞ is simple.

24
 CHAPTER 3

Fields and Field extensions

3.1. Rings of Polynomials (5/10/09)

D EFINITION 98. Let R be a ring. A formal power series over R in the variable x is a formal
sum ∞
f (x) = ∑ ai xi
i=0
with ai ∈ R, that is a function a : Z≥0 → R. We write R[[x]] for the set of these formal power series.
For f , g ∈ R[[x]] for the form f = ∑i≥0 ai xi , g = ∑ j≥0 b j x j and r ∈ R. We make the following
definitions:
def
f +g = ∑ (ai + bi)xi ;
i≥0
!
def
f ·g = ∑ ∑ ai b j x k ;
k≥0 i+ j=k
def
α·f = ∑ (αai) xi .
i≥0
P ROPOSITION 99. These definitions give R[[x]] the structure of a commutative R-algebra. R[[x]]
is an integral domain iff R is. Note that the additive group of R[[x]] is isomorphic to the countable
direct product of copies of the additive group of R.
L EMMA 100. The subset R[x] ⊂ R[[x]] of formal power series with finitely many non-zero co-
efficients is a subalgebra. The subset of polynomials of the form rx0 , r ∈ R, is a further subalgebra
isomorphic to R and we identify the two.
D EFINITION 101. R[x] is called the ring of polynomials over R in the variable x. For a non-zero
f ∈ R[x] set deg( f ) = max {i | ai 6= 0} and call it the degree of f , call adeg( f ) the leading coefficient,
and call f monic if adeg( f ) = 1 (we also set deg(0) = −∞).
From now on assume that the ring R is a field F.
L EMMA 102. (Degree valuation) Let f , g ∈ F[x]. Then deg( f g) = deg( f )+deg(g) and deg( f +
g) ≤ max {deg f , deg g}, with equality if deg f 6= deg g.
C OROLLARY 103. Let f , g ∈ F[x]. Then
(1) (zero-divisors) f g = 0 only if one of f , g is zero.
(2) (units) f g = 1 only if deg f = deg g = 0 and f g = 1 in R.
T HEOREM 104. (Division with remainder) Let f , g ∈ F[x] with f 6= 0. Then there exists unique
q, r ∈ F[x] with deg r < deg f so that
g = qf +r.
25
 C OROLLARY 105. F[x] is a principal ideal domain. An ideal I C F[x] is prime iff it is maximal,
iff I = ( f ) with f irreducible.
3.1.1. Divisors, GCD, LCM and unique factorization.
D EFINITION 106. f , g, h ∈ F[x].
• Say that f divides g, or that g is a multiple of f is there exists h such that f h = g.
• Say that f is irreducible if whenever f = gh one of g, h is a unit, reducible if f = gh for
some g, h both of degree at least 1.
• Say that f is prime if whenever f |gh we have either f |g or f |h (or both).
• If f , g ∈ F[x] and f = αg for α ∈ F × we say that f , g are associate. This is an equivalence
relation, and every equivalence class has a unique monic member.
D EFINITION 107. Let f , g ∈ F[x]. A greatest common divisor of f , g is the monic polynomial
h of maximal degree which divides both of them.
T HEOREM 108. Let f , g be polynomials. Then the Euclidean algorithm will compute a GCD,
which can be written in the form h f + kg for some h, k ∈ F[x].
P ROPOSITION 109. Every polynomial can be written as a product of irreducibles. A polyno-
mial is irreducible iff it is prime. Every polynomial has a unique factorization into primes (up to
associates).
3.1.2. Irreducibility in Q[x].
T HEOREM 110. (Gauss) Let f ∈ Z[x]. Be irreducible. Then f is irreducible in Q[x] as well.
P ROOF. Assume that f is reducible in Q[x], and let a ∈ Z≥1 be minimal such that
a f = gh
For g, h ∈ Z[x] for degree at least 1 (that a exists follows from clearing denominators). If a = 1 we
are done, so let p be a prime divisor of a. Letting bar denotes reductions mod p we have:
0̄ = ḡh̄ in (Z/pZ) [x] .
Since Z/pZ is a field, we have without loss of generality that ḡ = 0̄, in other words that every
coefficient of g is divisible by p. It then follows that
a g
f = h in Z[x] ,
p p
a contradiction to the minimality of a. 
T HEOREM 111. (Eisenstein) Let f = ∑ni=0 ai xi ∈ Z[x] and let p be a prime such that: p - an ,
p|ai for 0 ≤ i ≤ n − 1 but p2 - a0 . Then f is irreducible in Q[x].
P ROOF. Assume that f = gh. Say that deg(g) = r, deg(h) = s with leading coefficients br
and cs , respectively. Then r + s = n and br cs = an . In particularly, both br and cs are prime to
p. Reducing modulu p we find ān xn = ḡh̄ in (Z/pZ) [x]. It follows that ḡ = b̄r xr and h̄ = c̄s xs .
Assuming deg(g), deg(h) ≥ 1 this means that the constant coefficients of g, h are both divisible by
p, which would make the constant coefficient of f divisible by p2 . Otherwise one of g, h is an
integer, so f , g are associates in Q[x]. 
p
−1
E XAMPLE 112. The cyclotomic polynomial Φ p (x) = xx−1 = ∑ p−1 j
j=0 x is irreducible.
26
 P ROOF. The map x 7→ y + 1 and y 7→ x + 1 are isomorphisms of Z[x] and Z[y]. It follows that
p
it is enough to consider the irreducibility of Φ p (y + 1) = (y+1)y −1 = y p−1 + ∑ p−1
j=1
p
j−1
j y . Since
p
p
2
p| j for 1 ≤ j ≤ p − 1 and 1 = p is not divisible by p we are done. 

27
 Math 422/501: Problem set 5 (due 14/10/09)
Ideals in Rings

Let R be a ring. Recall that an ideal I C R is an additive subgroup I ⊂ R so that rI ⊂ I for all
r ∈ R.
1. (Working with ideals)
(a) Let I be a set of ideals in R. Show that I is an ideal.
T
def T
(b) Given a non-empty S ⊂ R show that (S) = {I | S ⊂ I C R} is the smallest ideal of R
containing S.
(c) Show that (S) = {∑ni=1 ri si | n ≥ 0, ri ∈ R, si ∈ S}.
(d) Let a ∈ R× . Show that a is not contained in any proper ideal.
Hint: Show that a ∈ I implies 1 ∈ I.

2. (Prime and maximal ideals) Call I C R prime if whenever a, b ∈ R satisfy ab ∈ I, we have a ∈ I

or b ∈ I. Call I maximal if it is not contained in any proper ideal of R.
(a) Show that R is an integral domain iff (0) = {0} C R is prime.
(b) Show that I C R is prime iff R/I is an integral domain.
(c) Show that R is a field iff (0) is its unique ideal (equivalently, a maximal ideal).
(d) Use the correspondence theorem to show that I is maximal iff R/I is a field.
(e) Show that every maximal ideal is prime.
Hint: Every field is an integral domain.

3. (Polynomials and maps of fields) Let ϕ : F → K be a map of fields, and let α ∈ K.

(a) Show that ϕ is injective.
Hint: consider the kernel of the map.
(b) Show that there is a unique homomorphism of rings (“evaluation at α”) ϕ̃α : F[x] → K
compatible with the embeddings F ,→ F[x] and F → K so that ϕ̃α (x) = α. For f ∈ F[x]
we usually write f (α) for ϕ̃α ( f ).
Hint: Write ϕ̃α as the composition of a map F[x] → K[x] and a map K[x] → K.
(c) Show that ϕ̃α is not injective iff there exists f ∈ F[x] (or f ∈ K[x] with coefficients in the
image of ϕ) so that f (α) = 0.
OPTIONAL If ϕ̃ is not injective, show that its kernel is of the form ( f ) for an irreducible
polynomial f ∈ F[x] and its image is a subfield of K.

4. (The field of rational functions) Let F be a field.

(a) For f , g, h, k ∈ F[x] with g, k 6= 0 say gf ∼ hk if f k = gh. Show that ∼ is an equivalence
relation. n o
f
(b) Show that F(x) = g | f , g ∈ F[x], g 6= 0 / ∼ is a field, and that the natural map ι : F[x] →
F(x) given by f 7→ 1f is an embedding.
(c) ϕ : F → K be an embedding of fields, and let α ∈ K. Assume that f (α) 6= 0 for all non-zero
f ∈ F[x], and show that ϕ extends uniquely to a map ϕ̃α : F(x) → K so that ϕ̃α (x) = α.

D EFINITION . Call α ∈ K algebraic over F if the situation of 3(c) holds, transcendental over
F if the situation of 4(c) holds.
28
 Irreducible polynomials and zeroes

5. Let f ∈ Z[x] be non-zero and let ab ∈ Q be a zero of f with (a, b) = 1. Show that constant
coefficient of f is divisible by a and that the leading coefficient is divisible by b. Conclude
that if f is monic then any rational zero of f is in fact an integer.
6. Decide while the following polynomials are irreducible:
(a) t 4 + 1 over R.
(b) t 4 + 1 over Q.
(c) t 3 − 7t 2 + 3t + 3 over Q.

7. Show that t 4 + 15t 3 + 7 is reducible in Z/3Z but irreducible in Z/5Z. Conclude that it is
irreducible in Q[x].
8. Let R be the field of real numbers. Let C = {a + bi | a, b ∈ R} where i is a formal symbol, and
def def
define (a + bi) + (c + di) = (a + b) + (c + d)i, (a + bi)(c + di) = (ac − bd) + (ad + bc)i.
(a) Show that the definition makes C into a ring.
(b) Show that {a + 0i | a ∈ R} is a subfield of C isomorphic to R.
(c) Show that the complex conjugation map τ(a+bi) = a−bi is a ring isomorphism τ : C → C
which restricts to the identity map on the image of R from part (b).
(d) Show that for z ∈ C the condition z ∈ R and τz = z are equivalent. Conclude that Nz =
def
NRC z = z · τz is a multiplicative map C → R.
(e) Show that C is a field.
Hint: Show first that if z ∈ C is non-zero then Nz is non-zero.
9. (Quadratic equations in C).
(a) Let d ∈ C. Show that there exist z ∈ C such that z2 = d.
(b) Let a, b, c ∈ C with a 6= 0, and let d = b2 − 4ac. Show that the equation az2 + bz + c = 0
has two solutions in C when d 6= 0, and one solution when d = 0.

Optional - A split R-algebra

8’. Let R be the field of real numbers. Let A = {a + bi | a, b ∈ R} where i is a formal symbol, and
def def
define (a + bi) + (c + di) = (a + b) + (c + d)i, (a + bi)(c + di) = (ac + 2bd) + (ad + bc)i.
(a) Show that the definition makes A into a ring.
(b) Show that {a + 0i | a ∈ R} is a subfield of A isomorphic to R.
(c) Show that the complex conjugation map τ(a +bi) = a −bi is a ring isomorphism τ : A → A
which restricts to the identity map on the image of R from part (b).
(d) Show that for z ∈ A the condition z ∈ R and τz = z are equivalent. Conclude that Nz =
def
NRA z = z · τz is a multiplicative map A → R.
(e) Show that A ' R ⊕ R, and in particular that it is not a field.
def
(f) Assume that multiplication is defined by (a + bi)(c + di) = (ac + tbd) + (ad + bc)i for
some fixed t ∈ R. For which t is the algebra a field? Find the isomorphism class of the
algebra, depending on t.

29
 Optional - The field of Laurent series

D EFINITION . Let R be a ring. A formal Laurent series over R is a formal sum f (x) = ∑i≥i0 ai xi ,
in other words a function a : Z → R for which there exists i0 ∈ Z so that ai = 0 for all i ≤ i0 . We
define addition and multiplication in the obvious way and write R((x)) for the set of Laurent series.
For non-zero f ∈ R((x)) let v( f ) = min {i | ai 6= 0} (“order of vanishing at 0”; also set v(0) = ∞).
Then set | f | = q−v( f ) (|0| = 0) where q > 1 is a fixed real number.
A. Show that R((x)) is a ring, and that R[[x]] is a subring.
B. (Invertibility)
(a) Show that 1 − x is invertible in R[[x]].
1
Hint: Find a candidate series for 1−x and calculate the product.
× ×
(b) Show that R[[x]] = {a + x f | a ∈ R , f ∈ R[[x]]}.
(c) Show that f ∈ R((x)) is invertible iff it is non-zero and av( f ) ∈ R× .
(d) Show that F((x)) is a field for any field F.
C. (Locality) Let F be a field.
(a) Let I C F[[x]] be a non-zero ideal. Show that I = xn F[[x]] for some n ≥ 1.
Hint: Show that f ∈ F[[x]] can be written in the form xv( f ) g(x) where g ∈ F[[x]]× .
(b) Show that the natural map F[x]/xn F[x] → F[[x]]/xn F[[x]] is an isomorphism.
D. (Completeness)
(a) Show that v( f g) = v( f ) + v(g), equivalently that | f g| = | f | |g| for all f , g ∈ R((x)).
(b) Prove the ultrametric inequality v( f + g) ≥ min {v( f ), v(g)} ⇐⇒ | f + g| ≤ max {| f | , |g|}
and conclude that d( f , g) = | f − g| defines a metric on f .
(c) Show that { fn }∞ n=1 ⊂ R((x)) is a Cauchy sequence iff there exists i0 such that v( f n ) ≥ i0
for all n, and if for each i there exists N = N(i) and r ∈ R so that for n ≥ N the coefficient
of xi in fn is r.
(d) Show that (R((x)), d) is complete metric space.
(e) Show that R[[x]] is closed in R((x)).
(f) Show that R[[x]] is compact iff R is finite.
E. (Ultrametric Analysis) Let {an }∞ ∞
n=1 ⊂ R((x)). Show that ∑n=1 an converges in R((x)) iff
limn→∞ an = 0.
Hint: Assume first that an ∈ R[[x]] for all n, and for each k consider the projection of ∑N
n=1 an
to R[[x]]/xk R[[x]].
F. (The degree valuation) Let F be a field.
(a) For f ∈ F[x] set v∞ ( f ) = − deg( f ) (and set v∞ (0) = ∞). Show that v∞ ( f g) = v∞ ( f ) +
v∞ (g). Show that v∞ ( f + g) ≥ min {v∞ ( f ), v∞ (g)}.
(b) Extend v∞ to the field F(x) of rational functions and show that it retains the properties
above. For a rational function f you can think of v∞ ( f ) as “the order of f at ∞”, just like
v( f ) measures the order of f at zero.
(c) Show that the completion of F(x) w.r.t. the metric coming from v∞ is exactly R(( 1x )).

30
 3.2. Field extensions
D EFINITION 113. A field extension is a homomorphism of fields.If ι : K → L is an inclusion
one may identify K with ι(K). In that case write L : K.
If K is a subfield of L and S ⊂ L we write K(S) for the intersection of all subfields of L con-
taining K and S.
Say ι : K → L and ι 0 : K 0 → L0 are isomorphic if there exist isomorphisms λ : K → K 0 and
η : L → L0 intertwining the two.
√ √ √
E XAMPLE 114. Q(i). Q( 3 2), Q(i, −i, 5, − 5).
D EFINITION 115. Let K ,→ L be an extension. Call α ∈ L algebraic over K if there exists
p ∈ K[x] such that p(α) = 0, transcendental otherwise. Call K ,→ L algebraic if every α ∈ L is
algebraic over K.
f (t) f (α)
L EMMA 116. Let α ∈ L be transcendental over K. Then K(α) ' K(t) via the map g(t) 7→ g(α) .

C OROLLARY 117. If α is transcendental over K then dimK K(α) = ∞.

Consider the map ϕ : K[x] → L given by evaluation at α. It is a ring homomorphism, and its
image is an integral domain contained in K(α). Its kernel is a prime ideal I of K[x], consisting of
all polynomials in K[x] which vanish at α. Since K[x] is a PID it follows that I = (m) for some
irreducible m, and that the image of the map is a field which must then be equal to K(α). . We
have obtained:
L EMMA 118. Let α ∈ L be algebraic over K. Then
(1) Every element of K(α) is of the form p(α) for some p ∈ f
(2) There is a unique monic irreducible polynomial m ∈ K[x] such that m(α) = 0, called the
minimal polynomial of α over K.
(3) If p ∈ K[x] satisfies p(α) = 0 then m|p.
D EFINITION 119. Call L : K simple if L = K(α) for some α.
P ROPOSITION 120. Let m ∈ K[x] be irreducible. Then there exists a simple extension L = K(α)
with m(α) = 0, and this extension unique up to isomorphism, which can be taken to map the images
of α.
P ROOF. K ,→ K[x]/(m) is such an extension, which we have already seen to be isomorphic to
any such K(α). 
C OROLLARY 121. dimK K(α) = deg m.
P ROOF. The polynomials of degree less than m are mapped injectively into K[x]/(m) (the
difference of two of them cannot be divisible by m unless zero). They are mapped surjectively by
division with remainder. 
T HEOREM 122. α is algebraic over K iff dimK K(α) < ∞.
C OROLLARY 123. Let α be algebraic over K. Then K(α) is algebraic over K.
P ROOF. Let β ∈ K(α). Then K(β ) ⊂ K(α) so dimK K(β ) ≤ dimK K(α) < ∞. 
D EFINITION 124. Let K ,→ L be an extension of fields. Call dimK L the degree of the extension
and denote it [L : K].
31
 P ROPOSITION 125. (Multiplicativity) Let K ,→ L ,→ M. Then [M : K] = [M : L] · [L : M].

P ROOF. Let {λi }i∈I be a basis for L over K. Let µ j j∈J be a basis for M over L. We will see

that λi µ j (i, j)∈I×J is a basis for M over K. First, assume that ∑i, j ai j λi µ j = 0 with a : I × J → K


finitely supported. Then ∑ j ∑i ai j λi µ j = 0. Since the µ j are independent over L, ∑i ai j λi = 0
for each j. Now get ai j = 0 for all i, j. Next, let m ∈ M. Then there exists b : J → L of finite
support such that ∑ j b j µ j = m. Next, for each j there exists a j : I → K of finite support such that
∑i ai j λi = b j . It follows that m = ∑i, j αi j λi µ j . 
C OROLLARY 126. Let α, β ∈ L be algebraic over K. Then so are α + β , −α, αβ , and α −1 .
P ROOF. β is algebraic over K, hence over K(α), and [K(α, β ) : K] = [K(α, β ) : K(α)][K(α) :
K] < ∞. 
D EFINITION 127. Let K ,→ L. The algebraic closure of K in L is the set {α ∈ L | [K(α) : K] < ∞}.
It is a subfield of L containing every algebraic extension of K contained in L
The algebraic closure of Q in C is called the field of algebraic numbers.

32
 Math 422/501: Problem set 6 (due 21/10/09)
√
Q( 3 2)

1. Let K = Q(α) where α 3 = 2. By Eisenstein’s criterion (p = 2), x3 − 2 ∈ Q[x] is irreducible.

Without using the tools from abstract algebra:
(a) Show by hand that 1, α, α 2 ⊂ K is linearly independent over Q.
Hint: You may use the irreducibility of x3 − 2.
(b) Show by hand that 1, α, α 2 isa basis for K.
Hint: It is enough to show that a + bα + cα 2 ⊂ K is closed under addition and multi-
plication, and that each element has an inverse.
– Conclude that [K : Q] = 3.

2. (The hard way) Let β ∈ K satisfy β 3 = 2.

(a) Write β = a + bα + cα 2 , and convert the equation β 3 = 2 = 2 + 0α + 0α 2 to a system of
three non-linear equations in the three
 variables a, b, c.
2
Hint: You need to use the fact that 1, α, α is a basis at some point.
(b) Taking a clever linear combination of two of the equations, show that a = 0.
(c) Now show that b = 1, c = 0, that is that β = α.

3. (The easy way) Let β ∈ K satisfy β 3 = 2 and assume that β 6= α.

(a) Let γ = β /α and show that γ 3 = 1.
(b) Let m(x) ∈ Q[x] be the minimal polynomial of γ over Q. Show that deg m = 2.
Hint: Start by showing that m is an irreducible factor of x3 − 1.
(c) Consider the field Q(γ) ⊂ K. Show that [Q(γ) : Q] = 2 and obtain a contradiction.
Hint: [K : Q] = [K : Q(γ)] · [Q(γ) : Q].

Prime fields and the characteristic

4. Let K be a field.
(a) Show that there is a unique ring homomorphism ϕ : Z → K.
(b) Let p ≥ 0 be such that Ker(ϕ) = (p). Show that either p = 0 or p is prime.
D EFINITION . We call p the characteristic of K.
(c) Let K be a field of characteristic p > 0. Show that the image of ϕ is the intersection of all
subfields of K, and that it isomorphic to the field F p = Z/pZ.
(d) Let K be a field of characteristic zero. Show that there is a unique homomorphism Q ,→ K
and conclude that the minimal subfield of K is isomorphic to Q.

5. (Finite fields)
(a) Let K be a finite field. Show that there exists a prime p and a natural number n so |K| = pn .
(b) Show that there exists a field of order 4.
Hint: Construct an irreducible quadratic polynomial in F2 [x].
(c) Show that there is a unique field of order 4.
R EMARK . We will see that for each prime power there is a field of that order, unique up to
isomorphism.
33
 Quadratic fields

Let K be a field of characteristic not equal to 2. Write K × for the multiplicative group of K,
(K × )2 for its subgroup of squares.
6. (Reduction to squares) Let L : K be an extension of degree 2.
(a) Show that there exists α ∈ L such that K(α) = L. What is the degree of the minimal
polynomial of α? √
(b) Show that there exist d ∈ K × such that L : K is isomorphic to K( d) : K.
Hint: Complete the square.
7. (Classifying the extensions)
√ n √ o
(a) Assume that d ∈ K × is not a square. Using the representation K( d) = a + b d | a, b ∈ K
√
show that e ∈ K is a square in K( d) iff e = d f 2 for some f ∈ K. Where did you use the
assumption about the characteristic?
√ √ 2
(b) Show that the extensions K( d) and K( e) are isomorphic iff de ∈ (K × ) (in general, the
√ √
isomorphism will not send d to e). √
√
Hint: Construct a K-homomorphism K( e) → K( d). Why is it surjective? Injective?
(c) Show that quadratic extensions of K are in bijection with non-trivial elements of the group
K × /(K × )2 .
8. (Applications)
(a) Show that R has
√ a unique
√ quadratic
√ √extension.
(b) Show that Q( 2√ + 3) =√Q( √2, 3). √ √ √
Hint: Show that 6 ∈ Q( 2, 3) but that 2 + 3 6= a + b 6 for any a, b ∈ Q.

Simple extensions

9. Let K(α) : K be a simple extension.

(a) If α is algebraic, show that there are finitely many subfields L of K(α) containing K.
Hint: consider the minimal polynomial of α over L.
(b) If α is transcendental, show that there are infinitely many intermediate fields L.
10. Let L : K be an extension of fields with finitely many intermediate subfields.
(a) Show that the extension is algebraic.
(b) Show that the extension is finitely generated: there exists a finite subset S ⊂ L so that
L = K(S).
(c) Show that L : K is finite.
11. Let L : K be an extension of infinite fields with finitely many intermediate fields.
(a) Given α, β ∈ L find γ ∈ L so that K(α, β ) = K(γ).
Hint: Consider elements of the form γ = α + kβ where k ∈ K.
(b) Show that L : K is a simple algebraic extension.

Algebraicity
12. Let M : L and L : K be algebraic extensions of fields. Show that M : K is algebraic.

34
3.3. Straightedge and compass

35
 Math 422/501: Problem set 7 (due 28/10/09) [extended till 30/10/09]
Splitting fields and normal closures

1. Construct subfields of C which are splitting fields over Q for the following polynomials:
(a) t 3 − 1;
(b) t 4 + 5t 2 + 6;
(c) t 4 + 7t 2 + 6;
(d) t 6 − 8.
Find the degrees of the splitting fields as extensions of Q.

2. Construct a splitting field for the following polynomials over F3 :

(a) t 3 + 2t + 1;
(b) t 3 + t 2 + t + 2.
(c) Are the two fields isomorphic?

3. Let f ∈ K[x] and let Σ : K be a splitting field for f over K. Let K ⊂ M ⊂ Σ be an intermediate
field. Show that Σ is a splitting field for f over M.

4. Let f ∈ K[x] have degree n and let Σ : K be a splitting field for f over K. Show that [Σ : K] ≤ n!.

Algebraic closures

D EFINITION . A field extension K ,→ K̄ is called an algebraic closure if it is algebraic, and if

polynomial in K[x] splits in K̄[x]. We also say informally that K̄ is an algebraic closure of K.
5. Let K ,→ L be an algebraic extension.
(a) If K is finite, show that |L| ≤ ℵ0 .
(b) If K is infinite, show that |L| = |K|.

6. Let K ,→ K̄ be an algebraic closure. Show that every algebraic extension of K̄ is isomorphic to

K̄.

7. (Existence of algebraic closures) Let K be a field, X an infinite set containing K with |X| > |K|.
Let 0, 1 denote these elements of K ⊂ X. Let
F = {(L, +, ·) | K ⊂ L ⊂ X, (L, 0, 1, +, ·) is a field with K ⊂ L an algebraic extension} .
Note that we are assuming that restricting +, · to K gives the field operations of K.

(OPT) Show that F is a set. Note that {(ϕ, L) | L is a field and ϕ : K → L is an algebraic extension}
is not a set.
(a) Show that every algebraic extension of K is isomorphic to an element of F.
(b) Given (L, +, ·) and (L0 , +0 , ·0 ) ∈ F say that (L, +, ·) ≤ (L0 , +0 , ·0 ) if L ⊆ L0 , + ⊆ +0 , · ⊆ ·0 .
Show that this is a transitive relation.
(c) Let K̄ ∈ F be maximal with respect to this order. Show that K̄ is an algebraic closure of
K.
(d) Show that K has algebraic closures.

36
8. (Uniqueness of algebraic closures) Let K ,→ K̄ and K ,→ L be two algebraic closures of K.
Show that the two extensions are isomorphic.
Hint: Let G be the set of K-embeddings intermediate subfields K ⊂ M ⊂ L into K̄, ordered by
inclusion.

Symmetric polynomials

Let R be a ring. Then Sn acts on the polynomial ring R[x1 , . . . , xn ] by permuting the variables,
and we write R[x]Sn for the set of fixed points.
9. (Basic structure)
(a) Show that R[x]Sn is a subring of R[x], the ring of symmetric polynomials.
(b) For α ⊂ [n] write xα for the monomial ∏i∈α xi . For 1 ≤ r ≤ n let
sr (x) = ∑ xα ∈ R[x] .
α∈([n]
r)

Show that sr (x) ∈ R[x]Sn . These are called the elementary symmetric polynomials.
10. (Generation) Define the height of a monomial ∏ni=1 xiαi to be ∑ni=1 iαi . Define the height of
p ∈ R[x] to be the maximal height of a monomial appearing in p.
β
(a) Given p ∈ R[x]Sn find β ∈ Zn≥0 and r ∈ R so that the highest term of q = r ∏nr=1 sr r occurs
in p.
(b) Show that p − q either fewer highest terms than p or smaller height than p.
(c) Show that every symmetric polynomial can be written as a polynomial of equal or smaller
degree in the elementary symmetric polynomials.

Derivatives, derivations and separability

11. The formal derivative of the Laurent series f = ∑i≥i0 ai xi ∈ R((x)) over the ring R is the Laurent
def
series D f = ∑i≥i0 iai xi−1 .
(a) Show that D : R((x)) → R((x)) is R-linear: that D(α f + β g) = αD f + β Dg for α, β ∈ R
and f , g ∈ R((x)).
(b) Show that D is a derivation: that D( f g) = D f · g + f · Dg (this is called the Leibniz rule).
(c) Show that D( f k ) = k · f k−1 · D f for all k ≥ 0.
(d) Show that if f is a polynomial then D f is a polynomial as well, that is that D restricts to a
map R[x] → R[x].
12. (Derivative criterion for separability) Let K be a field.
(a) Let α ∈ K be a zero of f ∈ K[x]. Show that (x − α)2 | f iff D f (α) = 0 iff (x − α)|D f .
(b) Let ϕ : K → L be an extension of fields, and let f , g ∈ K[x]. Let ( f , g) = (h) as ideals of
K[x], (ϕ( f ), ϕ(g)) = (h0 ) as ideals of L[x]. Taking h, h0 monic show that h0 = ϕ(h).
(c) Show that f ∈ K[x] is has no repeated roots in any extension (is separable) iff ( f , D f ) = 1.
(d) Show that an irreducible f ∈ K[x] is separable iff D f 6= 0.

37
 Optional problems

A. Construct an embedding K(x) ,→ K((x)) and show that D restricts to a map K(x) → K(x).

For the rest fix a ring R.

B. Let A be an R-algebra, and consider the map A × A → A given by the commutator bracket
[a, b] = ab − ba, .
(a) Show (A, [·, ·]) is a Lie algebra, that is that the commutator is R-bi-linear and anti-symmetric,
and satisfies the Jacobi identity [a, [b, c]] + [b, [c, a]] + [c, [a, b]] = 0.
(c) Show that for a fixed a ∈ A the map b 7→ [a, b] is an element ad(a) ∈ EndR (A).
(d) Show that ad(a) is a derivation: (ad(a)) (bc) = [(ad(a)) (b)] c + b [(ad(a)) (c)].
C. Let A be an R-algebra. Let DerR (A) = {D ∈ EndR (A) | D is a derivation}.
(a) Show that DerR (A) ⊂ EndR (A) is an R-submodule.
(b) Give an example showing that DerR (A) need not be an R-subalgebra (that is, closed under
multiplication=composition).
(c) Show that DerR (A) is closed under the commutator bracket of EndR (A).
D. Let A an R-algebra. Show that the map ad : A → DerR (A) is a map of Lie algebras, that is a
map of R-modules respecting the brackets.

38
 CHAPTER 4

Monomorphisms, Automorphisms and Galois Theory

4.1. Splitting fields and normal extensions

D EFINITION 128. Let L : K be an extension of fields. Say f ∈ K[x] splits in L if its image in
L[x] is a product of linear factors there. Say that L is a splitting field for f over K if f splits in L be
not in any intermediate field K ⊂ M ( L.
T HEOREM 129. (Splitting fields)
(1) For every field K and f ∈ K[x] there exists a splitting field L/K, in fact one with[L : K] ≤
(deg( f ))!.
(2) Splitting fields are unique up to isomorphism of extensions: if κ : K → K 0 is an isomor-
phism of fields, f ∈ K[x], and ι : K → L, ι 0 : K 0 → L0 are splitting fields for f and κ( f )
respectively, then there exists an isomorphism λ : L → L0 so that (κ, λ ) is an isomorphism
of the extensions ι and ι 0 .
P ROOF. First, if f ∈ K[x] splits in L, say f = c ∏i (x − αi ), then M = K({αi }) is a splitting field:
f splits there, and any sub-extension of M where f splits contains the {αi } hence is equal to M. It
is thus enough to construct an extension where f splits (with the given bound of the degree). We
prove this by induction on the degree of f . If deg( f ) ≤ 1 there’s nothing to prove. Otherwise let
f
g be an irreducible factor of f and let M = K(α) where α is a root of g. By induction x−α ∈ M[x]
has a splitting field. It is clear that f splits there as well. The degree bound is an exercise.
We prove the second part by a similar induction. Let g ∈ K[x] be an irreducible factor of f
and let α ∈ L be a root of g, α 0 ∈ L0 a root of κ(g) which is also irreducible. Then K(α) : K and
K 0 (α 0 ) : K 0 are isomorphic

0 0 0 0
extensions, say by (κ, κ ). Then L : K(α) and K : K (α ) are splitting
f f κ( f )
fields for x−α and κ 0 x−α = x−α 0 0
0 respectively so by induction there is λ : L → L so that (κ , λ )

is an isomorphism of the extensions. It follows that (κ, λ ) is an isomorphism of extensions. 

D EFINITION 130. Call L : K normal if every irreducible f ∈ K[x] which has a root in L splits
in L.
P ROPOSITION 131. If L : K is normal and M is an intermediate field then L : M is normal.
P ROOF. (Exercise). 
T HEOREM 132. L/K is normal and finite iff it is a splitting field.
P ROOF. If L/K is finite it is finitely generated, say L = K(α1 , . . . , αr ). Let gi ∈ K[x] be the
minimal polynomial of αi . Then f = ∏i gi splits in L (each gi does by normality), while every
subfield of L where f splits contains all the αi and hence is L. For the converse let L : K be the
splitting field of f ∈ K[x] and let α ∈ L have minimal polynomial g. In the splitting field M of
f g (which contains a unique copy of L) let α 0 be another root of g. Then K(α) : K and K(α 0 ) : K
39
are isomorphic extensions, hence of the same degree. Next, L(α) : K(α) and L(α 0 ) : K(α 0 ) are
splitting fields for the same polynomial f (the isomorphism of K(α) and K 0 (α) fixes K). Thus they
are isomorphic extensions and also have the same degree. It follows that [L(α) : K] = [L(α 0 ) : K].
Dividing by [L : K] shows [L(α 0 ) : L] = [L(α) : L] = 1 so α 0 ∈ L as well so M = L and g splits in
L. 
D EFINITION 133. A normal closure of an extension of fields L : K is an extension N : L so that
N : K is normal while every proper intermediate extension of N : L is not.
P ROPOSITION 134. Normal closures exist and are unique up to isomorphism of fields.
P ROOF. Assume [L : K] < ∞, say L = K(α1 , . . . , αr ). Let gi ∈ K[x] be a minimal polynomial
of αi . Then N is a normal closure iff it is a splitting field for ∏i gi . The infinite case is left as an
exercise (one way is to construct it in an algebraic closure). 
E XAMPLE 135. Every quadratic extension is normal.

4.2. Separability
D EFINITION 136. Let L : K be an extension. Call f ∈ K[x] separable if every irreducible factor
of f has distinct roots in the splitting field. Call α ∈ L separable over K if its minimal polynomial
in K[x] is separable. Call L : K separable if every α ∈ L is separable over K, purely inseparable if
every α ∈ L separable over K belongs to K.
P ROPOSITION 137. If L : K is separable and M is an intermediate field then L : M and M : K
are separable.
P ROOF. (Exercise). 
P ROPOSITION 138. (Construction of monomorphisms) Let L/K be finite. Then there are at
most [L : K] K-monomorphisms of L into a normal closure N/K. If L is generated over K by
separable elements then the number of monomorphisms is precisely [L : K], and conversely if the
number if [L : K] then the extension is separable.
P ROOF. Induction on the degree. Assuming that n = [L : K] > 1 let α ∈ L \ K have minimal
polynomial f ∈ K[x] and let {αi }ei=1 be the roots of f in N (including α1 = α), and note that
e ≤ d = deg( f ). Then K(α) has precisely e embeddings into N. By induction L has at most
n
d = [L : K(α)] K(α)-embeddings into N with α mapping to αi . Since every embedding maps α
to one of the αi it follows that the total number of embedding is at most e · dn ≤ d · dn = n. If we can
choose α ∈ L which is not separable then we’d have e < d and so the number of embedding would
be strictly less than n. If L/K is generated by separable elements then we take α to be one of them
so e = d; since L/K(α) is also generated by separable elements it has precisely dn embeddings and
we are done. 
We obtain several corollaries:
T HEOREM 139. (Separability) A finite extension L/K is separable iff it is generated by sepa-
rable elements. Thus:
(1) An extension generated by separable elements is separable.
(2) Let K ,→ L ,→ M with L/K separable and let α ∈ M be separable over L. Then α is
separable over K. In particular, M/K is separable iff M/L and L/K are.
40
 (3) If M/K is an extension then the subset L ⊂ M of elements which are separably algebraic
over K is a subfield, the separable closure of K in M. If M/K is algebraic the extension
M/L is purely inseparable.
P ROOF. The initial claim is immediate.
(1) let L = K(S) with S ⊂ L separable algebraic over K. For each α ∈ L there is a finite subset
T ⊂ S so that α ∈ K(T ) and we may apply the Proposition to the extension K(T ) : K.
(2) Let f ∈ L[x] be the minimal polynomial of α and let R = K( f ) be the subfield generated
by its coefficients. Let N/K be a normal closure of M/K. Then R has [R : K] embeddings
into N and each can be extended in [R(α) : R] = deg( f ) ways to embeddings of R(α). It
follows that R(α) has [R(α) : K] K-embeddings into N so R(α) is separable and so is α.
(3) The field extension generated by the separable elements is separable, hence is equal to
that set. Any element of the extension separable over the separable closure is separable
over the base field.

E XAMPLE 140. Let char(K0 ) = p and let K = K0 (t) be the function field in one variable over
K0 . Then x p − t ∈ K[x] is irreducible and inseparable. Indeed if L/K is a field and s ∈ L is a root
then (x − s) p = x p − s p = x p − t so s is the unique root of x p − t in L. Also, any monic divisor of
x p − t in L[x] has the form (x − s)r for some 0 ≤ r ≤ p. If 1 ≤ r < p then the constant coefficient
of this divisor is sr/p ∈
/ K (this elements generates K(s) as well) so the divisor is not in K[x]. One
p
can also see that x − t is irreducible using Eisenstein’s criterion in K0 [t][x].

4.3. Automorphism Groups

D EFINITION 141. Let L be a field. Aut(L) will be the group of automorphisms of L. If L : K is
an extension of fields we write AutK (L) for the group of automorphisms fixing K element-wise.
√
E XAMPLE 142. Quadratic extensions in characteristic different from 2, Q( 3 2) and its normal
closure, the inseparable extension.
L EMMA 143. (Dedekind) Let K, L be fields. Then Hom(K, L) is linearly independent over L
(as a subset of LK ).
P ROOF. Let 0 = ∑ri=1 ai fi be a minimal linear combination. Then the fi distinct and all the ai
are non-zero. We have r ≥ 2 since 0 ∈ / Hom(K, L). Let y ∈ K be such that f1 (y) 6= fr (y) (then y 6= 0
as are f1 (y), fr (y)), and see that for all x ∈ K we have:
r−1 r r
∑ (ai ( fi(y) − ai fn(y))) fi(x) = ∑ ai fi(y) fi(x) − fn(y) ∑ ai fi(x)
i=1 i=1 i=1
! !
r r
∑ ai fi (yx) − fn (y) ∑ ai f i (x)
i=1 i=1
= 0.

R EMARK 144. In fact, we have shown that if H is a group then Hom(H, L× ) is linearly inde-
pendent (take H = K × ).
41
 C OROLLARY 145. Let [L : K] = n. Then # AutK (L) ≤ n2 .
P ROOF. AutK (L) is a linearly independent subset of the space of linear maps on an n-dimensional
K-vector space. 
P ROPOSITION 146. Let [L : K] = n. Then # AutK (L) ≤ n.
P ROOF. Let {ωi }ni=1 be a basis for L over K. Each σ ∈ AutK (L) is determined by the vec-
tor (σ (ωi ))ni=1 ∈ Ln , and these vectors are linearly independent over L: if aσ ∈ L are such that
∑σ aσ σ (ωi ) = 0 for each i, then ∑σ aσ σ is a K-linear map L → L which vanishes on a basis, hence
vanishes identically, which forces all the aσ to vanish by the Lemma. Since dimL Ln = n we are
done. 
D EFINITION 147. For σ ∈ Aut(L) write Fix(σ ) = {x ∈ L | σ (x) = x}, a subfield of L. For
S ⊂ Aut(L) write Fix(S) = ∩σ ∈S Fix(σ ). Note that Fix(S) = Fix(hSi).
P ROPOSITION 148. Let L be a field, G ⊂ Aut(L) a finite subgroup of order n, and let K =
Fix(G). Then [L : K] = n.
P ROOF. To each ω ∈ L associate the vector ω G = (σ (ω))σ ∈G ∈ LG . Let Ω ⊂ L be a basis over
K and let ∑ri=1 ai ωiG = 0 be a minimal linear dependence in LG over L. Then for each σ ∈ G we
have ∑i ai σ (ωi ) = 0 with ai ∈ L× . For τ ∈ G note that we have ∑i τ(ai )(τσ )(ωi ) = 0 for all σ ,
so ∑i τ(ai )ωiG = 0 as well. Since minimal combinations are unique up to scalar, there is b ∈ L× so
that τ(ai ) = bai for all i. Then τ a−1 −1


1 ai = a1 ai for all i. Since τ was arbitrary it follows that
there are ci ∈ K × so that ai = a1 ci . Dividing by a1 it follows that ∑ri=1 ci ωiG = 0. In particular the
co-ordinate of the identity gives ∑ri=1 ci ωi = 0, which is impossible. It follows that ω G ω∈Ω ⊂ LG
are linearly independent over L, and hence that |Ω| ≤ |G|, that is [L : K] ≤ |G|. In particular ,[L : K]
is finite, and we then have |G| ≤ [L : K] as well. 

42
 Math 422/501: Problem set 8 (due 4/11/09)
Monomorphisms of fields

1. (From class)
(a) Let L/K be a finite extension and σ ∈ HomK (L, L). Show that σ is an automorphism.
(b) Let L/K be an algebraic extension and σ ∈ HomK (L, L). Show that σ is an automorphism.
(c) Give an example showing there exist extensions with endomorphisms which are not auto-
morphisms.

2. Let ϕ : R → R be a homomorphism of rings. Show that ϕ is the identity map.

3. (The Frobenius map) Let K be a field of characteristic p > 0
(a) Show that the map x 7→ x p defines a monomorphism K → K.
r
(b) Conclude by induction that the same holds for the map x 7→ x p for any r ≥ 1.
(c) When K is finite show that the Frobenius map is an automorphism.
OPT When K = F̄ p show that the Frobenius map is again an automorphism, and obtain a
group homomorphism Z 7→ Gal(F̄ p : F p ).
FACT The image of this homomorphism is dense.
√ √
4. (The Galois correspondence for K = Q( 2, 3)).
(a) Find all Q-automorphisms of K and give their group structure.
(b) Find all subfields of K.
(c) Show that the map H 7→ Fix(H) gives a bijection between subgroups of the automorphism
group and subfields of K.

Finite fields

5. (Multiplicative groups)
(a) Let G be a finite p-group such that for every d, g ∈ G | gd = e ≤ d. Show that G is

cyclic.
(b) Let G be a finite group such that for every d, g ∈ G | gd = e ≤ d. Show that G is

cyclic.
(c) Let F be a field, G ⊂ F × a finite group. Show that G is cyclic.

6. (Uniqueness of finite fields) Fix a prime p and let q = pr for some r ≥ 1.

(a) Show that the polynomial xq − x ∈ F p [x] is separable.
(b) Let F be a finite field with q elements. Show that F is a splitting field for xq − x over F p .
(c) Conclude that for each q there is at most one isomorphism class of fields of order q. If
such a field exists it is denoted Fq .

43
7. (Existence of finite fields) Fix a prime p and let q = pr for some r ≥ 1.
(a) Let F/F p be a splitting field for xq − x, and let σ : F → F be the map σ (x) = xq . Show
that the fixed field of σ is also a splitting field.
(b) Conclude that the field F has order q.
8. Let F be a finite field, K/F a finite extension.
(a) Show that the extension K/F is normal and separable.
Hint: 7(a).
(b) Show that there exists α ∈ K so that K = F(α).
Hint: Consider the group K × .

Optional – Finite fields

A. (Galois correspondence for finite fields) Fix a prime p.

(a) Assume that F pr embeds in F pk . Show that r|k.
(b) Let r|k. Show that F pk has a unique subfield isomorphic to F pr , consisting of the fixed
r
points of the map x 7→ x p .
r
(c) Show that Gal(F pk : F pr ) is cyclic and generated by the Frobenius map x 7→ x p .
(d) Obtain the Galois correspondence for extensions of finite fields.

44
 4.4. The group action
If L/K is an extension of fields, then AutK (L) acts on L. We now investigate the orbits of this
action. For this note that if L, M are extensions of K and σ ∈ HomK (L, M), and if f ∈ K[x], α ∈ L
then σ ( f (α)) = f (σ (α)). In particular, α is root of f iff σ (α) is.
C OROLLARY 149. Let L/K be an algebraic extension, let N/K be a normal extension and
let M/N be a further extension. Assume we have a K-monomorphism σ : L → N. Then every
K-monomorphism τ ∈ HomK (L, M) has its image in N.
P ROOF. For every α ∈ L, τ(α) ∈ M is a root of the minimal polynomial of α, and this polyno-
mial already has the root σ (α) ∈ N. 
and σ ∈ AutK (L) then σ ( f (α)) = f (σ (α)). It follows that if α is algebraic over K then its
orbit is contained in the set of roots of its minimal polynomial.
L EMMA 150. Let f ∈ K[x] be irreducible and let N/K be a finite normal extension. If f splits
in N then AutK (N) acts transitively on the roots of f .
P ROOF. Let α, β be roots of f in N. By Theorem 132 there exist g ∈ K[x] be such that N is the
splitting field of g over K, hence also over K(α) and K(β ). By Theorem 129 the K-isomorphism
of K(α) and K(β ) carrying α to β extends to an isomorphism of N to itself. 
We can generalize this:
P ROPOSITION 151. (Construction of monomorphisms) Let L/K be a finite algebraic extension,
let N/K be a finite normal extension and let σ , τ ∈ HomK (L, N). Then there exists ρ ∈ AutK (N)
so that τ = ρσ .
P ROOF. Again let g ∈ K[x] be such that N is the splitting field of g. Then σ , τ : L → N are both
splitting fields for g, and are therefore isomorphic. 
We can rephrase this by saying that AutK (N) acts transitively on HomK (L, N).
T HEOREM 152. Let L/K be an algebraic extension, let N/K be a normal algebraic extension,
and let σ , τ ∈ HomK (L, N). Then there exists ρ ∈ AutK (N) so that τ = ρσ .
P ROOF. Identifying L with σ (L) we may assume σ = id. Consider the set of functions µ
whose domain is a subfield of N containing L, whose range is contained in N, and which are field
monomorphisms extending τ, ordered by inclusion. Let ρ be a maximal element of the set (this
exists by Zorn’s Lemma). If the domain of ρ is a proper subfield M of N let α ∈ N \ M. 

4.5. Galois groups and the Galois correspondence (2/11/09)

D EFINITION 153. If L/K is normal and separable we say that it is a Galois extension, call
AutK (L) the Galois group, and denote it Gal(L : K).
T HEOREM 154. Let [L : K] = n. Then the following are equivalent:
(1) L/K is a Galois extension.
(2) AutK (L) has order n.
(3) The fixed field of AutK (L) is precisely K.
45
 P ROOF. By Proposition 138 if L/K is normal and separable there are n = [L : K] K-automorphisms
of L. By Proposition 148, K is the fixed field of AutK (L) iff # AutK (L) = [L : K]. If the fixed is
indeed K let N be a normal closure of L/K. Then # AutK (L) consists of [L : K] K-monomorphisms
from L to N. By Proposition 138 we see that L/K is separable. Also, every K-monomorphism of
L into N has its image in L. Let f ∈ K[x] be irreducible and have a zero α ∈ L. Then f splits in N,
and for every other root β ∈ N there is a K-monomorphism of L into N mapping α to β (otherwise
there would not be n = [L : K] monomorphisms in total). It follows that β ∈ L, that is that L is
normal. 
T HEOREM 155. (Galois Correspondence) Let L : K be a finite Galois extension. Then the
inclusion-reversing maps H 7→ Fix(H), M 7→ Gal(L : M) between subgroups H < Gal(L : K) and
intermediate fields K ⊂ M ⊂ L are inverse to each other. Further:
(1) M : K is normal iff Gal(L : M) is normal in Gal(L : K).
(2) If M : K is normal then Gal(M : K) ' Gal(L : K)/ Gal(L : M).
P ROOF. Clearly if M ⊂ M 0 ⊂ L then AutM 0 (L) ⊂ AutM (L): every M 0 -automorphism of L is an
M-automorphism. Similarly, if H ⊂ H 0 then every α ∈ L fixed by H 0 is fixed by H. Also, for any
intermediate field M, L : M is normal and separable hence Galois. Now for H < Gal(L : K) we
have H ⊂ Gal(L : Fix(H)). By Proposition 148 [L : Fix(H)] = #H and by the previous Theorem
[L : Fix(H)] = # Gal(L : Fix(H)). It follows that H = Gal(L : Fix(H)). Similarly for an intermediate
field M, the index of Fix(Gal(L : M)) in L is the same as the index of M. Since the two are contained
in each other they are equal.
Finally, let σ ∈ Gal(L : K) and let H < Gal(L : K). Then the fixed field of σ Hσ −1 is exactly
σ Fix(H). If Fix(H) is normal than any K-automorphism of L must leave Fix(H) invariant since
it maps roots of polynomials to roots of polynomials, so σ Hσ −1 has the same fixed field as H
and hence is equal. Conversely, if H is normal then Fix(H) is an invariant set for the action of
Gal(L : K); since the orbits of the action are precisely the sets of roots of irreducible polynomials,
it follows that M = Fix(H) is normal over K. Restricting the action of the Galois group to M we
obtain a map Gal(L : K) → Gal(M : K). By definition, the kernel of this map is Gal(L : M). It is
surjective since by Proposition 151 every K-automorphism of M extends to a K-automorphism of
L. 
P ROPOSITION 156. Let L/K Galois extension, and let α ∈ L. Let O ⊂ L be the orbit of α
under Gal(L : K). Then O is finite and f = ∏β ∈O (x − β ) is the minimal polynomial of α over K.
P ROOF. Since O is contained in the roots of the minimal polynomial, it is finite, and we may
then take L finite. From now on we only assume that G = AutK (L) has order n = [K : L], this
giving an alternative proof of the converse part of Theorem 154. First, for σ ∈ G we have σ ( f ) =
∏β ∈O (x − σ (β )) = f so f belongs to the fixed field of G, that is K. Note that f has distinct roots
by construction, so α is separable. f is also irreducible, since a product of the form ∏β ∈S (x − β )
is G-fixed if an only if S is G-invariant set, and it follows that every irreducible in K[x] which has
a root in L splits in L, so L is normal. 

4.6. Examples and applications

4.6.1. The primitive element Theorem.
T HEOREM 157. Let L/K be a finite, separable extension. Then L = K(θ ) for some θ ∈ K.
46
 P ROOF. Assume first that K is infinite, and let N/K be a normal closure of L/K. Then N/K is
finite by Proposition 134 and separable since the Proposition shows it is generated by separable el-
ements. Since Gal(N/K) is finite it has finitely many subgroups, and by the Galois correspondence
it follows that there are finitely many intermediate fields between N and K, hence also between L
and K. The claim now follows from problem 11 of problem set 6. When K is finite L is finite as
well and the claim is problem 8 of problem set 8. 
4.6.2. Symmetric combination and Galois’s outlook. Let f ∈ K[x] split in L[x] with roots
{α1 , . . . , αr }. Let s ∈ K[y]Sr . Then s(α) ∈ K.
4.6.3. Cyclotomic fields. See Problem Set 9.
4.6.4. The polynomial t 4 − 2. See [2, Ch. 12]

47
 Math 422/501: Problem set 9 (due 13/11/09)
Galois theory

1. Let L/K be a finite Galois extension. Let K ⊂ M1 , M2 ⊂ L be two intermediate fields. Show
that the following are equivalent:
(1) M1 /K and M2 /K are isomorphic extensions.
(2) There exists σ ∈ Gal(L : K) such that σ (M1 ) = M2 .
(3) Gal(L : Mi ) are conjugate subgroups of Gal(L : K).
2. (V -extensions) Let K have characteristic different from 2.
(a) Suppose L/K is normal, separable, with Galois group C2 × C2 . Show that L = K(α, β )
with α 2 , β 2 ∈ K.
√ √
(b) Suppose a, b ∈ K are such that none of a, b, ab is a square in K. Show that Gal(K( a, b) :
K) ' C2 ×C2 .

The fundamental theorem of algebra

3. (Preliminaries)
(a) Show that every simple extension of R has even order.
(b) Show that every quadratic extension of R is isomorphic to C.

4. (Punch-line)
(a) Let F : R be a finite extension. Show that [F : R] is a power of 2.
Hint: Consider the 2-Sylow subgroup of the Galois group of the normal closure.
(b) Show that every proper algebraic extension of R contains C.
(c) Show that every proper extension of C contains a quadratic extension of C.
(d) Show that C : R is an algebraic closure.

Example: Cyclotomic fields

µn ⊂ C× will denote the group of nth roots of unity, Sn ⊂ µn the primitive nth roots of unity.
p −1
5. (prime order) Let p be an odd prime, and recall the proof from class that Φ p (x) = xx−1 is
irreducible in Q[x].
(a) Let ζ p be a root of Φ p . Show that Q(ζ p ) is a splitting field for Φ p . What is its degree?
(b) Show that G = Gal(Q(ζ p ) : Q) is cyclic.
(c) Show that Q(ζ p ) has a unique subfield K so that [K : Q] = 2.
(d) Show that there is a unique non-trivial homomorphism χ : G → {±1}.
(e) Let g = ∑σ ∈G χ(σ )σ (ζ ) (“Gauss sum”). Show that g ∈ K and that g2 ∈ Q.
p−1
OPT Show that g2 = (−1) 2 p, hence that K = Q(g).

48
6. Let ζn ∈ C be a primitive nth root of unity.
(a) Show that Q(ζn ) is normal over Q.
Hint: Show that every embedding of Q(ζn ) in C is an automorphism.
j(σ )
(b) Let G = Gal(Q(ζn ) : Q). Show for every σ ∈ G there is j ∈ (Z/nZ)× so that σ (ζn ) = ζn
and that j : G → (Z/nZ)× is an injective homomorphism.
(c) Let Φn (x) = ∏ζ ∈Sn (x − ζ ). Show that Φn (x) ∈ Q[x] (in fact, Φn (x) ∈ Z[x]). Show that the
degree of Φn is exactly φ (n) = #(Z/nZ)× .
(d) Show that the definitions of Φ p (x) in problems 5 and 6(c) agree.
7. (prime power order) Let p be prime, r ≥ 1 and let n = pr .
pr
(a) Show that Φn (x) = xpr−1−1 .
x −1
(b) Show that Φn is irreducible.
Hint: Change variables to Φn (1 + y) and reduce mod p.
(c) Conclude that Gal(Φ pr ) ' (Z/pr Z)× .
ri
8. (general order) Let n = ∏si=1
 pi with pi  distinct primes. Let G, j be as in 6(b).
(a) Show that Q(ζn ) = Q ζ pr1 , . . . , ζ prss .
1

×
(b) For each i let πi : (Z/nZ)× → Z/pri i Z be the natural quotient map. Show that the maps

×
πi ◦ j : G → Z/pri i Z are surjective.
(c) [deferred]

Example: Cubic extensions

9. Let K be a field, f ∈ K[x] of degree n, and let {αi }ni=1 ⊂ Σ be the roots of f in a splitting field
Σ, counted with multiplicity.
(a) Let {sr }nr=1 be the elementary symmetric polynomials in n variables, thought of as ele-
ments of K[y1 , . . . , yn ]. Show that sr (α1 , . . . , αn ) ∈ K.
Hint: Consider the factorization of f in Σ.
(b) Let t ∈ K[y]Sn be any symmetric polynomial. Show that t(α1 , . . . , αn ) ∈ K.

10. Let K be a field of characteristic zero, and let f ∈ K[x] be an irreducible cubic. Let Σ be a
splitting field for f , and let {αi }3i=1 be the roots.
(a) Show that [Σ : K] ∈ {3, 6} and that Gal(Σ : K) is isomorphic to C3 or S3 .
Hint: The Galois group acts transitively on the roots.
(b) Let δ = (α1 − α2 ) (α2 − α3 ) (α3 − α1 ), and let ∆ = δ 2 . Show that ∆ ∈ K × .
(c) Let M = K(δ ). Show that [Σ : M] = 3 and hence that [Σ : K] = 3 iff δ ∈ K. Conclude that
f is still irreducible in M[x].
(d) Assume that K ⊂ R and that Σ ⊂ C. Show that Σ ⊂ R iff M ⊂ R iff δ ∈ R iff ∆ > 0.
— We now adjoin ω so that ω 3 = 1.
(e) Show that [Σ(ω) : M(ω)] ∈ {1, 3}, and in the first case that Σ is contained in a radical
extension.
(f) Assuming [Σ(ω) : M(ω)] = 3 show that this extension is still normal.
(g) Let y = α1 + ωα2 + ω 2 α3 ∈ Σ(ω). Show that for any σ ∈ Gal(Σ(ω) : M(ω)) there is j so
that σ y = ω j y. Conclude that y3 ∈ M(ω).

49
 4.7. Solubility by radicals
In this section all fields have characteristic zero.
def
D EFINITION 158. f ∈ K[x] separable. Then Gal( f ) = Gal(Σ( f ) : K) where Σ( f ) is the splitting
field.
Call L/K radical if L = K(α1 , . . . , αs ) and for each i there is ri so that αiri ∈ K(α1 , . . . , αi−1 ).
Call f ∈ K[x] soluble by radicals if there exists a radical extension containing Σ( f ). If f is irre-
ducible enough to show K[x]/( f ) is contained in a radical extension.
T HEOREM 159. f ∈ K[x] is soluble by radicals iff Gal( f ) is a solvable group.
4.7.1. Radical extensions are solvable (4-9/11/09).
L EMMA 160. L/K contained in a radical then normal closure N/K contained in a radical.
P ROOF. Enough to show that the normal closure of a radical extension is radical. Indeed, let
L = K(α1 , . . . , αs ) be radical, let N be the normal closure, G = Gal(N : K). Then N = K ({σ (αi ) | σ ∈ G, 1 ≤ i ≤ s}
Ordering this lexicographically with i most significant than σ exhibits this as a radical extension.
In the alternative let r = lcm {r1 , . . . , rs } and let N = L(µr ). Then N is normal (contains all
conjugates of the αi ) and radical. 
L EMMA 161. Gal(Σ(t p − 1) : K) is Abelian.
P ROOF. Automorphisms raise generator to a power. 
L EMMA 162. If µn ⊂ K then Σ(t n − a) : K is abelian.
P ROOF. Galois group maps the root α to the root ζ α where ζ is a root of unity. 
P ROPOSITION 163. L/K normal and radical implies Gal(L : K) solvable.
P ROOF. Induction on number of roots. Can assume ri are all prime. Say α p ∈ K but α ∈/ K.
p
Let M ⊂ L be splitting field for t − 1. Then M : K, M(α) : M are normal and abelian. L : M(α)
solvable by induction. 
T HEOREM 164. L/K contained in radical extension. Then AutK (L) is solvable.
P ROOF. K ⊂ L ⊂ R ⊂ N where R/K is radical, N/K its normal closure. Then N/K is radical so
Gal(N/K) is solvable. Let H = {σ ∈ Gal(N/K) | σ (L) ⊂ L}; restriction gives a map H → AutK (L)
with kernel Gal(N : L). This map is surjective since every K-automorphism of L is extends to an
automorphism of N since N is a splitting field of some f ∈ K[x]. Now H is solvable as a subgroup
of a solvable group, and AutK (L) is solvable as a quotient of a solvable group. 
4.7.2. Insoluble polynomials (9/11/09).
P ROPOSITION 165. Let p be prime, f ∈ Q[x] irreducible of degree p with precisely two complex
roots. Then Gal( f ) ' S p .
P ROOF. Let A ⊂ C be the roots of f , Σ = Q(A) the splitting field, G = Gal(Σ : Q). Then
G acts transitively on a set of size p, giving an embedding G ,→ S p . If α ∈ A is any root then
[Q(α) : Q)] = p so p| [Σ : Q] = #G so the image of the map contains an element of order p, which
is hence a p-cycle σ ∈ S p . Let τ ∈ G be the restriction of complex conjugation to Σ. Then τ is a
2-cycle, say τ = (1 2). Any non-identity power of σ is also a p-cycle, and by transitivity there is
one of the form (1 2 . . . p). These two together generated S p . 
50
 E XAMPLE 166. t 5 − 6t + 3 ∈ Q[x] is irreducible by Eisenstein. Its derivative is 5t 4 − 6 which

1/4
1/4
is positive if |t| > 56 = u and negative in |t| < 65 . Since f (−u) = − 65 u + 6u + 3 > 0
and f (u) = 65 u − 6u + 3 = 3 − 4.8u < 0 since u > 1, it follows that f has three real roots (one in
(−∞, −u), one in (−u, u) and one in (u, ∞)).
4.7.3. Solvable extensions are radical (9/11/09).
D EFINITION 167. Let L/K be a finite extension and let α ∈ L. If L/K is Galois set TrLK (α) =
∑σ ∈Gal(L/K) σ α, NKL (α) = ∏σ ∈Gal(L/K) σ α. In general let TrLK (α) and NKL (α) be, respectively, the
trace and determinant of multiplication by α, thought of as a K-linear map L → L.
E XERCISE 168. (PS10 problem 3) The two definitions coincide when they intersect.
(PS10 problem 2) Let 1 < [L : K] < ∞ be prime to char(K). Then L = L0 ⊕ K as K-vector
spaces. In particular, there exist α ∈ L \ K with trace zero.
The key step in our induction will be the following:
P ROPOSITION 169. Let L/K be a Galois extension of prime index p, and assume µ p ⊂ K. Then
L is radical over K.
P ROOF. Let σ generate G = Gal(L/K) (a group of order p hence cyclic). For α ∈ L and ζ ∈ µ p
consider the Lagrange Resolvent
Θ(α, ζ ) = ∑ ζ bσ b(α) .
b(p)

Then:
σ (Θ(α, ζ )) = ζ −1 Θ(α, ζ ) .
If Θ 6= 0 and ζ 6= 1 this would show Θ(α, ζ) ∈/ K but (Θ(α, ζ )) p ∈ K, finishing the proof. For
α fixed let Θ(α) be the vector Θ(α, ζ pa ) a∈Z/pZ = Z · α G where Z ∈ Mn (K) is the Vandermonde

matrix Zab = ζ pab and α G = σ b (α) b . Note that (Zα)0 = Tr α and choose α ∈ L \ K so that

Tr(α) = 0. Then α G 6= 0 so Zα G 6= 0 and it follows that there is a 6= 0 so that (Zα)a 6= 0. 

P ROPOSITION 170. (”Base change”) Let T : K be an extension of fields, and let L, M ⊂ T be
intermediate extensions with L/K is a finite Galois extension. Let LM ⊂ T be the field generated
by L, M. Then LM : M is a finite Galois extension, and restriction to L is an injective map Gal(LM :
M) → Gal(L : K) (in particular, [LM : M] ≤ [L : K]). Moreover, if L/K is cyclic, abelian or solvable
then so is LM/M.
P ROOF. Assume that L is the splitting field of the separable polynomial f ∈ K[x]. Then LM
is the splitting field of f over M. It follows that LM : M is a finite Galois extension. Since L is
normal every σ ∈ AutK (LM) maps L to L, so restriction to L gives a map AutM (LM) → AutK (L).
If σ belongs to the kernel of this map then σ ∈ Aut(LM) fixes M (assumption on the domain) and
L (assumption on the image). It follows that σ is trivial. 
T HEOREM 171. Let L/K be a finite solvable Galois extension. Then there exists a radical
extension M of K containing L.
51
 P ROOF. Let [L : K] = n. We will show that L(µn! ) : K is radical. It is clearly enough to show
that L(µn! ) : K(µn! ) is radical, and by the base change proposition this is a solvable extension of
degree at most n. We now prove by induction on N that if L/K is solvable, and K contains µn!
then L/K is radical. For this let G = Gal(L/K), and let H < G be normal of prime index p. Let
M = Fix(H).Then M : K is Galois, with Galois group G/H ' C p . Since p ≤ n, we may apply the
first Proposition to see that M/K is radical. Also, L : M is solvable and [L : M]! | [L : K]! so M
contains all the requisite roots of unity to apply the induction hypothesis. 

52
 Math 422/501: Problem set 10 (due 18/11/09)
The trace

When L/K is a finite Galois extension and α ∈ L we used in class the combination (“trace”)
TrLK (α) = ∑σ ∈Gal(L/K) σ α, which we needed to be non-zero. We will study this construction when
L/K is a finite separable extension, fixed for the purpose of the problems 1-3.

1. Let N/K be a normal extension containing L.

(a) For α ∈ L we provisionally set TrLK (α) = ∑µ : L→N µα (“trace of α”), NKL (α) = ∏µ : L→N µα
(“norm of α”). Show that the definition is independent of the choice of N.
(b) Making a judicious choice of N show that the trace and norm defined in part (a) are ele-
ments of K.
(c) Show that when L/K is a Galois extension the definition from part (a) reduces to the
combination used in class.

2. (Elements of zerotrace) In class we had the occasion to need elements α ∈ L with trace zero.
For this, let L0 = α ∈ L | TrLK (α) = 0 .
(a) Show that TrLK is a K-linear functional on L, so that L0 is a K-subspace of L.
(b) When char(K) = 0, show that L = K ⊕ L0 as vector spaces over K (direct sum of vector
spaces; the analogue of direct product of groups). Conclude that when [L : K] ≥ 2 the set
L0 \ K is non-empty.
(c) Show that TrLK is a non-zero linear functional in all characteristics.
(d) Show that L0 is not contained in K unless [L : K] = char(K) = 2, in which case L0 = K, or
[L : K] = 1 in which case L0 = {0}.
3. (Yet another definition) We continue with the separable extension L/K of degree n.
(a) Let f ∈ K[x] be the (monic) minimal polynomial of α ∈ L, say that f = ∑di=0 ai xi with
K(α) K(α)
ad = 1. Show that TrK (α) = −ad−1 and that NK (α) = (−1)d a0 .
n/d
(b) Show that TrLK (α) = − dn ad−1 and that NKL (α) = (−1)n a0 .
Hint: Recall the proof that [L : K] has n embeddings into a normal closure.
(c) Show that TrLK (α) and NKL (α) are, respectively, the trace and determinant of multiplication
by α, thought of as a K-linear map L → L.
Hint: Show that we have L ' (K(α))n/d as K(α)-vector spaces,.

R EMARK . From now on we define the trace and norm of α as in 3(c). Note that this definition
makes sense even if L/K is not separable.

4. (Transitivity) Let K ⊂ L ⊂ M be a tower of finite extensions. Show that

(a) TrM L M
K = TrK ◦ TrL .
(b) NKM = NKL ◦ NLM .

53
 Purely inseparable extensions

5. Let L/K be an purely inseparable algebraic extension of fields of characteristic p.

r
(a) For every α ∈ L show that there exists r ≥ 0 so that α p ∈ K. In fact, show that the minimal
r r
polynomial of α is of the form x p − α p .
Hint: Consider the minimal polynomials of α and α p
(b) Conclude that when [L : K] is finite it is a power of p.
(c) When [L : K] is finite show that TrLK is identically zero.

Some examples

6. Solve the equation t 6 + 2t 5 − 5t 4 + 9t 3 − 5t 2 + 2t + 1 = 0 by radicals.

Hint: Try u = t + 1t .

7. Let K have characteristic zero and consider the system of equations over the field K(t):

2
x = y + t

y2 = z + t .
z2 = x + t


(a) Let (x, y, z) be a solution in a field extension of K(t). Show that x satisfies either x2 = x + t
or a certain sextic equation over K(t).
OPT Use a computer algebra system to verify that the sextic is relatively prime both to x2 −
x − t and to its own formal derivative.
(b) Show that the Galois group of the splitting field of the sextic preserves an equivalence
relation among its six roots.
Hint: Find an permutation of order 3 acting on the roots. This is visible in the original
system.
(c) Let {α, β , γ} be an equivalence class of roots, and let s(a, b, c) be a symmetric polynomial
in three variables. Show that s(α, β , γ) belongs to an extension of K(t) of degree 2 at most.
Hint: If s(α, β , γ) is a root of a quadratic, what should the other root be? Show that the
coefficients of the putative quadratic are indeed invariant by the Galois group.
(d) Show that the system of equations can be solved by radicals.
Hint: For each equivalence class construct a cubic whose roots are the equivalence class
and whose coefficients lie in a radical extension.
OPT Show that knowing [K(t, x + y + z) : K(t)] = 2 where x, y, z are roots of the original sys-
tem would have been enough.

OPTIONAL Let L = C(x) (the field of rational functions in variable) and for f ∈ L let (σ ( f )) (x) =
f ( 1x ), (τ( f )) (x) = f (1 − x).
(a) Show that σ , τ ∈ Aut(L) and that σ 2 = τ 2 = 1.
(b) Show that G = hσ , τi is a subgroup of order 6 of Aut(L) and find its isomorphism class.
(c) Let K = Fix(G). Find this field explicitly.

54
 4.8. The group action and Grothendieck’s Galois Correspondence
[to be added later]

55
 CHAPTER 5

Algebraic Number Theory

5.1. Intro (16-18/11/09)

P ROPOSITION 172. TFAE for α ∈ C:
(1) There exists a monic f ∈ Q[x] with f (α) = 0.
(2) Q(α) is finite-dimensional.
(3) There exists a finite-dimensional Q-subspace W ⊂ C so that αW ⊂ W .
P ROOF. We have seen (1) ⇒ (2), (2) ⇒ (3) and clear and (3) ⇒ (1) is Cayley-Hamilton. 
D EFINITION 173. α ∈ C is an algebraic number if it satisfies any of these properties.
P ROPOSITION 174. TFAE for α ∈ C:
(1) There exists a monic f ∈ Z[x] with f (α) = 0.
(2) Z[α] is finite-dimensional.
(3) There exists a finite-dimensional Z-submodule W ⊂ C so that αW ⊂ W .
P ROOF. Same (see xxx below) 
D EFINITION 175. α ∈ C is an algebraic number if it satisfies any of these properties.

56
 Math 422/501: Problem set 11 (due 25/11/09)
The discriminant

Let L/K be a separable extension, and let N/K be its normal closure. Let n = [L : K] =
n
# HomK (L, N), with an enumeration HomK (L, N) = {µi }ni=1 . Given ω j j=1 ⊂ L let Ω ∈ Mn (L)
be the matrix with Ωi, j = µi (ω j ) and set:
dL/K (ω1 , . . . , ωn ) = (det Ω)2 .
In particular, write dL/K (α) = dL/K 1, α, α 2 , . . . , α n−1 .

 n
1. Let ω j j=1 ⊂ L.
(a) Show that dL/K (ω1 , . . . , ωn ) ∈ K.
 n
(b) Show that dL/K (ω1 , . . . , ωn ) 6= 0 iff ω j j=1 is a basis for L over K.
(c) Show that dL/K (α) 6= 0 iff L = K(α).
(d) Show that if dL/K (α) 6= 0 then it is the discriminant of the minimal polynomial of α.
n on
n
2. (The case K = Q) Let L be a number field of degree n over Q. Let i i=1 , ω 0j
{ω } ⊂ L be
j=1
Q-bases of L so that the abelian groups M = Zω1 ⊕ · · · ⊕ Zωn and N = Zω10 ⊕ · · · ⊕ Zωn0 satisfy
N ⊂ M.
(a) Show that the sum ⊕ni=1 (Zωi ) is indeed direct.
(b) Show that dL/Q (ω10 , . . . , ωn0 ) = DdL/Q (ω1 , . . . , ωn ) for some positive integer D.
Hint: Relate the matrices Ω and Ω0 .
(c) Show that when M = N we have dL/Q (ω1 , . . . , ωn ) = dL/Q (ω10 , . . . , ωn0 ), in other words that
the discriminant of a basis is really a function of the Z-module generated by that basis.
(d) Say ω 0j = a j ω j for some a j ∈ Z. Show that D = [M : N]2 .
REMARK (c),(d) are special cases of the general identity dL/Q (N) = [M : N]2 dL/Q (M).

Rings of integers

FACT. (Integral basis Theorem) Let K be a number field of degree n (that is, [K : Q] = n), and
let OK ⊂ K be the set of algebraic integers in K. Then there exists a basis {αi }ni=1 of K over Q so
def
that OK = ⊕ni=1 Zαi . Moreover, dK = dK/Q (α1 , . . . , αn ) is an integer.
3. Let D be√ a square-free integer (this means a product of distinct primes up to sign) and let
K = Q( D).
(a) Let α ∈ K. Show that α is an algebraic integer iff Tr α, Nα ∈ Z (trace and norm from K
to Q). √
(b) Show that 1+2 D is an algebraic integer iff D ≡ 1√(4).
√ √
(c) Show that Z[ D] = Z ⊕ Z D ⊂ OK ⊂ Z 21 ⊕ Z 2D .
√
Hint: write α ∈ K in the form a + b D for a, b ∈ Q.
(d) By considering
√ the equation
 √ x2 − y2 D ≡ 0 (4) in Z/4Z, show that if D ≡ 2, 3 (4) then
OK = Z[ D] = a + b D | a, b ∈ Z .
57
 h √ i n √ o
1+ D a+b D
(e) Show that when D ≡ 1 (4) OK = Z 2 = 2 | a, b ∈ Z, a ≡ b (2) .
— What about D ≡ 0 (4)?

4. (Dedekind) Let K = Q(θ ) where θ is a root of f (x) = x3 − x2 − 2x − 8.

(a) Show that f is irreducible over Q and find its Galois group.
(b) Show that 1, θ , θ 2 are all algebraic integers.
2
(c) Let η = θ 2+θ . Show that η 3 − 3η 2 − 10η − 8 = 0 and conclude that is an algebraic integer
a well.
(d) Show that 1, θ , η are linearly independent over Q.
(e) Let M = Z ⊕ Zθ ⊕ Zη and let N = Z[θ ] = Z ⊕ Zθ ⊕ Zθ 2 . Show that N ⊂ M.
(f) Show that dK/Q (θ ) = ∆( f ) = −4 · 503.
(g) Find dK/Q (1, θ , η).
Hint: You can be confident in your answer by consulting 2(a).
(h) Show that {1, θ , η} is an integral basis.
dK/Q (1,θ ,η)
Hint: Let {α, β , γ} be an integral basis and consider dK/Q (α,β ,γ) .
(i) Let δ = A + Bθ + Cη with A, B,C ∈ Z. Show that 2|dK/Q (δ ). Conclude that the set of
algebraic integers of K is not of the form Z[δ ].

58
 5.2. Integrality and Integral basis (23-25/11/09)
Let K be an algebraic extension of Q, OK ⊂ K be the set of algebraic integers.
P ROPOSITION 176. a ∈ OK iff Z[α] is a finitely generated Z-module iff there is a non-zero
finitely generated Z-module M ⊂ K so that αM = M.
P ROOF. Say α is integral of degree d. Then Z[α] = ⊕d−1 i i
i=0 Zα and αα ∈ Z[α] for all i. The
third part implies the first by Cayley-Hamilton. 
C OROLLARY 177. OK is a subring of K. If α ∈ OK then:
(1) Every conjugate of α is integral over Q;
(2) The minimal polynomial of α over Q is monic and belongs to Z[x];
(3) If K is finite over Q then TrK K
Q (α), NQ (α) ∈ Z and,
(4) α ∈ OK× iff NQK α ∈ {±1}.
P ROOF. Given α, β ∈ OK say αM ⊂ M and αN ⊂ N then MN is finitely generated and
Z[α, β ]MN ⊂ MN. Also, every conjugate of α satisfies the same polynomials that α does. The
minimal polynomial of α is ∏µ∈HomQ (Q(α),Q̄) (x − µα) ∈ OQ̄ [x] ∩ Q[x] since OQ̄ is a ring. Since
OQ̄ ∩ Q = Z we have claim (2). Claim (3) now follows by taking specific coefficients. For (4) note
that for α ∈ OK, Nα
α = ∏µ6=1 µα ∈ OQ̄ ∩ K = OK. Thus α|Nα in OK, so if Nα is invertible so is
α. Conversely, if αβ = 1 then NαNβ = 1. 
L EMMA 178. Let α ∈ K. Then there is m ∈ Z so that mα ∈ OK .
P ROOF. Let f = ∑di=0 ai xi ∈ Q[x] be the (monic) minimal polynomial of α. Then ∑di=0 md−i ai (mα)i =
0. If m is large enough then md−i ai ∈ Z for all 0 ≤ i < d. 
C OROLLARY 179. There exists a basis of K consisting of elements of OK .
From now on assume that K is a number field, that is a finite extension of Q. We write n =
[K : Q].
L EMMA 180. The quadratic form (x, y) 7→ Tr(xy) is non-degenerate.
P ROOF. Tr(x · x−1 ) = n. 
P ROPOSITION 181. There exist a basis {ωi∗ }ni=1 ⊂ K so that OK ⊂ ⊕i Zωi∗ .
P ROOF. Take {ωi∗ } be the basis dual to a basis contained in OK w.r.t. the trace form. 
C ONCLUSION 182. The Z-module OK embeds in Zn and contains a copy of Zn .
The structure theorem of finitely generated abelian groups now immediately shows that OK ='
Zn as Z-modules. We give a self-contained proof of this fact.
L EMMA 183. Let f : Zr → Zs be an isomorphism of Z-modules. Then r = s.
P ROOF. In short: f ⊗Z 1 : Zr ⊗Z Q → Zs ⊗Z Q is an isomorphism of vector spaces. In detail,
define f˜ : Qr → Qs by setting f˜(v) = m1 f (mv) for any v ∈ Qr and m ∈ Z \ {0} so that mv ∈ Zr .
This is easily seen to be a well-defined isomorphism of vector spaces. 
D EFINITION 184. A Z-module M is called free if M ' Zr for some r. In that case we call r the
rank of M.
59
 P ROPOSITION 185. Let A be a free Z-module of finite rank, B < A a Z-submodule. Then B is
free and rank(B) ≤ rank(A).
P ROOF. By induction on r = rank(A). If r = 1 then A ' Z and this is problem 1(a) of Problem
Set 1. Assume then that A = ⊕r+1 0 r 0 0 0
i=1 Zxi and let A = ⊕i=1 Zxi , B = B ∩ A . By induction, B is free
of rank at most r. Also, B/B0 embeds in A/A0 ' Z so by the case of rank 1, B/B0 is free of rank 0
or 1. In the first case B ' B0 and we are done. Otherwise, let y ∈ B project to a generator of B/B0 .
Then the sum B0 + Zy is direct and equal to B, finishing the proof. 
T HEOREM 186. (Integral Basis Theorem) Let [K : Q] = n. Then is OK a free Z-module of rank
n.
P ROOF. Applying Proposition 185 to the inclusion of Proposition 181 we see that OK is free
of rank at most n. Applying Proposition 185 to the inclusion of Corollary 179 show that OK has
rank at least n. 

5.3. Unique factorization (30/11/09)

Fix a number field K of degree n.
P ROPOSITION 187. (Ideals of OK ) Fix a non-zero proper ideal a C OK .
(1) a ∩ Z is a non-zero proper ideal of Z.
(2) [OK : a] < ∞.
(3) a is finitely generated. In fact, rankZ a = n.
(4) If a is prime then it is maximal, and a ∩ Z = (p) for a prime number p.
P ROOF. a ∩ Z is certainly an ideal. It does not contain 1 since a doesn’t, and is not zero since
it contains Nα for all α ∈ a. Setting a ∩ Z = (m) we have mOK ⊂ a ⊂ OK . This shows that a is a
free Z-module of rank n, and that [OK : a] ≤ [OK : mOK ] = mn .
If a is prime then OK /a is a finite integral domain, hence a field, so a is maximal. Z/(Z ∩ a) is
also an integral domain since it embeds in OK /a. 
D EFINITION 188. An OK -submodule a ⊂ K is a fractional ideal if there is α ∈ K × so that
αa ⊂ OK .
Given fractional ideals a, b let ab be the module generated by all products xy, x ∈ a, y ∈ b.
Multiplication of fractional ideals is commutative and associative, and has the unit (1) = OK . We
call a fractional ideal invertible if it is invertible in this commutative semigroup.
L EMMA 189. Every proper ideal of OK contains a product of primes.
P ROOF. Let a be a maximal counterexample. It is not prime so there are x, y ∈ OK \ a with
xy ∈ a. Then (a, x) (a, y) = a, a contradiction. 
def
P ROPOSITION 190. Let p C OK be prime. Then p−1 = {x ∈ K | xp ⊂ OK } is a fractional ideal
properly containing OK . In particular, pp−1 = OK .
P ROOF. Let p ∈ Z be the prime lying below p. Let x, y ∈ p−1 and α ∈ OK . First, (αx + y) p ⊂
xp + yp ⊂ OK + OK = OK . Second, (px)p =⊂ pOK ⊂ p. Since p is a finitely generated Z-module,
it follows that px is integral over Z, and hence that p · p−1 ⊂ OK . To see that p−1 ) OK fix a
non-zero x ∈ p. Then xOK ⊂ p and hence contains a product of prime ideals. Let p1 · · · pr be a
60
minimal such product. Since p contains this product, it contains one factor, and hence equal to it
(all primes are maximal). Let b ∈ p2 · · · pr \ xOK (this exists by minimality of r). Then bp ⊂ xOK
so bx p ⊂ OK but bx ∈/ OK . Finally, p ⊂ pp−1 ⊂ OK . Since pp−1 is an OK -submodule of OK and
p is a maximal ideal one side must be an equality. If pp−1 = p held then every y ∈ p−1 would be
integral, a contradiction. 
T HEOREM 191. All ideals of OK are invertible; every ideal can be uniquely written in the form
e
∏ri=1 pi i with pi prime and ei ∈ Z>0 . a|b in the monoid of ideals iff b ⊂ a.
P ROOF. First, let a C OK be a proper ideal and let a ⊂ p C OK be a maximal ideal. Then
p−1 a ⊂ p−1 p = O −1
K and p a 6= a since p
−1 6⊂ O .
K
Now let a C OK be maximal among the non-invertible non-zero ideals. Then p−1 a is invertible,
and hence so is a. Similarly, let a be maximal ideal without representation as a product of primes.
Then p−1 a can be written as such a product, and hence so can a. Finally, let ∏ri=1 pi = ∏tj=1 q j .
Then pr contains the product on the left, hence the product on the right, hence equal to one of the
factors. Multiplying by p−1
r the claim follows by induction on r.
If ac = b then certainly b ⊂ a. Conversely, if b ⊂ a then a−1 b ⊂ a−1 a = OK . 
C OROLLARY 192. Every fractional ideal is invertible, so that the fractional ideals form a
group. Every element of this group has a unique representation in the form ∏ri=1 pei i with ei ∈
Z \ {0}.
D EFINITION 193. Call a fractional ideal principal if it is of the form αOK for some α ∈
K ×. Say that two fractional ideals a, b are in the same class if ab−1 is principal (note that every
fractional ideal is in the same class as an ideal by definition). The principal fractional ideals form a
subgroup of the group of fractional ideals. The class group of K is the quotient Cl(K) of the group
of ideals by the group of principal ideals. It measures the failure of unique factorization.
T HEOREM 194. (Dedekind) Cl(K) is a finite group.
P ROPOSITION 195. (Kummer) Let α, β ∈ OK be relatively prime and let their product be a
pth power, where p - # Cl(K). Then α, β are pth powers up to units.
P ROOF. Say that αβ = γ p . Then (α)(β ) = (γ) p and so (α), (β ) are pth powers in the monoid
of ideals, say (α) = a p , (β ) = b p . But then the classes of a, b are trivial in the class group. 
D EFINITION 196. (Kummer) Call the odd prime p regular if p - Cl(Q(ζ p )).
T HEOREM 197. (Kummer) Let p be a regular prime. Then x p + y p + z p = 0 has only the trivial
solution x = y = z = 0.
j
P ROOF. We consider the equation x p − y p = z p in Z[ζ p ]. We have there ∏ p−1
j=0 (x − ζ p y) =
z p. 
5.4. Splitting of primes (2/12/09)
D EFINITION 198. If p C OK is a non-zero prime ideal and pZ = p ∩ Z we say that p divides p
or lies above it, and write p | p.
We have seen that every prime of OK lies above a unique prime of Z. There is a partial converse.
L EMMA 199. Let pZ be a prime of Z. Then there exist primes p of OK lying above (p).
61
 P ROOF. let p be a maximal ideal of OK containing the ideal pOK . Then p ∩ Z is a proper ideal
containing the maximal ideal (p) hence equal to it. 
Next, let pOK = ∏gi=1 pei i for g = g(p). Then kp = OK /pi is a finite extension of Z/pZ, say of
degree fi . By the CRT, we have
g
OK /pOK ' ∏ OK /pei i .
i=1
Now dimkp pr /pr+1 = 1 so dimk p OK /pe = e and dimF p OK /pei i = ei fi . It follows that
g
n = ∑ ei fi .
i=1
D EFINITION 200. Say p ramifies in K if ei ≥ 2 for some i.
T HEOREM 201. p ramifies iff p|dK . In particular, there are only finitely many ramified primes.
When K/Q is Galois, we have more structure. Note that if σ ∈ G = Gal(K : Q) and p is a
prime ideal then so is σ (p) and both have the same intersection with Z.
T HEOREM 202. Gal(K/Q) acts transitively on the set of primes lying above p.
In particular, fi , ei are constant and we have n = e(p) f (p)g(p).
D EFINITION 203. Dp = {σ ∈ G | σ (p) = p}. This group acts on OK fixing p setwise hence on
kp fixing Z/pZ element-wise, giving a homomorphism Dp → Gal(kp : F p ). Write Ip for the kernel
of this map.
T HEOREM 204. Dp surjects onto the Galois group of the residue field.
C OROLLARY 205. [G : Dp ] = g(p) by transitivity, [Dp : Ip ] = f (p) by surjectivity. It then fol-
lows that #Ip = e(p) since e f g = n.
If p is unramified then e = 1 so for each p lying above p there is a unique element mapping to
the Frobenius element of kp . Write Frob p for the resulting conjugacy class.
T HEOREM 206. (Chebotarev density theorem) Let C ⊂ G be a conjugacy class. Then

# p ≤ x | p unramified and Frob p = C #C
−−−→ .
# {p ≤ x} x→∞ #G

62
 Bibliography

[1] Arthur Cayley. Desiderate and suggestions: No. 1. the theory of groups. Amer. J. Math., 1(1):50–52, 1878.
[2] Ian Stewart. Galois Theory. Chapman & Hall/CRC Mathematics. Chapman & Hall/CRC, Boca Raton, FL, third
edition, 2004.

63