MATH 123 NOTES

AMY FENG & EMMA CARDWELL & ZIYONG CUI

Please let us know if you catch any errors, thank you!

Contents

1. Monday, January 22: Introduction to Rings 2

2. Friday, January 26: Prime and Maximal Ideals 3

3. Monday, January 29 7

4. Friday, February 2: Ideals in K[x] and Beyond 8

5. Monday, February 5 10

6. Friday, February 9 12

7. Monday, February 12 13

8. Friday, February 16: More Tensor Products (ft many commutative diagrams) 14

9. Friday, February 23: Wedges and the Determinant 17

10. Monday, February 26: 19

11. Friday, March 1: Review 20

12. Friday, March 8 21

13. Monday, March 18 24

14. Friday, March 22: WOW more Galois! + symmetric functions 25

15. Monday, 03/25 29

16. Friday, March 29: Splitting Fields 30

17. Monday, April 1 32

18. Friday, April 5 33

Date: Spring 2024.
1
2 AMY FENG & EMMA CARDWELL & ZIYONG CUI

19. Monday, April 8 34

20. Friday, April 12 35

21. Monday, April 15 37

22. 04/19/2024 38

23. 04/22/2024 40

1. Monday, January 22: Introduction to Rings

There will be semi-frequent homeworks, which are due a week after they are assigned, and
two assessments:

(1) Midterm: TBD, 3/4

(2) Final Exam: 2pm, 5/10

Professor Kisin’s OH are 2-3pm on Wednesday in SC234. The current plan is to go over
Rings & Modules, Quadratic Fields (maybe), and Galois Theory. We will use Artin’s Algebra
as the primary textbook for this course.

Recall: a group is a set with an operation (G, +).

(1) associativity: (a+b) + c = a+(b+c)

(2) identity: ∃0 ∈ G, s.t. 0 + a = a + 0 = a
(3) inverse: ∀a ∈ G, ∃ − a ∈ G s.t. a + (−a) = 0
(4) For abelian groups and a, b ∈ G, a + b = b + a

Examples include (R, +), Dn = Z/nZ ⋊ Z/2Z, SO(3)(R), Sn .

Rings: informally, a ring is a structure (set) that has addition

√ and multiplication.
√ Examples
include Z, Z[x] = {a0 +a1 x+a2 x2 +· · ·+an xn : ai ∈ Z}, Z[ −1] = {a+b −1 : a, b ∈ Z} ⊆ C.
Formally we have: a ring is a set R with operations (+, ·) s.t.

(1) (R, +) is an abelian group =⇒ 0 ∈ R

(2) (a · b) · c = a · (b · c)
(3) a · (b + c) = a · b + a · c

For rings with units: (4) ∃1 ∈ R s.t. a · 1 = 1 · a = a. For commutative rings: (5) a · b = b · a.
Examples and constructions:

(1) If R is a ring, Mn (R) is a ring

(2) If R is a ring, R[x] is a ring
 MATH 123 NOTES 3
P
(3) Let G be a finite group (which may be non-abelian), then Z[G] = { g∈G ag · g : ag ∈
Z} = tupes of integers indexed by elements of G is a ring
(4) If R is a ring, R[G] is a ring where G is a group.
(5) If X is a set, let CC (X) =functions : f : X → C where addition and multiplication
are defined pointwise. This is a ring where f (x) = 1∀x ∈ X is the multiplicative
identity
(6) If X is a topological space, then C CT S = continuous functions on X is a ring.
(7) C ∞ = infinitely differentiable functions is a ring.

Consider Yoga: any ring has the above form, you just need to define X appropriately (at
least when R is commutative).

Examples of rings without units:

(1) The even numbers

(2) If X = R2 (non-compact space), C(X) ⊃ C C (X) = {f : R2 → C} where f = 0 outside
of a big disc

Now consider Quotient Rings. In groups, if G is a group and H is a subgroup, G/H is a

set G/H = {g + H : g ∈ G}. For G/H to be a group, H needs to be normal: ∀g ∈ G,
g + H − g = H (gHg −1 = H). Note that if G is abelian, then any subgroup H ⊆ G is
normal. If R is a ring and I ⊆ R is an additive subgroup, when is (R/I, +) a ring? Assume
that R is commutative. If r1 , r2 ∈ R, try to define (r1 + I) · (r2 + I) = r1 r + 2 + I. For this to
be well-defined, this should only depend on the cosets r1 + I and r2 + I. If r1′ ∈ r1 + I and
r2′ ∈ r2 + I, then r1′ = r1 + j1 and r2′ = r2 + j2 for j1 , j2 ∈ I. Thus, we need r1 r2 + I = r1′ r2′ + I
where r1′ r2′ = (r1 + j1 )(r2 + j2 ) = r1 r2 + r1 j2 + j1 r2 + j1 j2 for all j1 , j2 ∈ I. Thus, we need
three terms to be in I: ∀r1 , r2 ∈ R and j1 , j2 ∈ I, we need R · I ⊂ I and I · R ⊂ I. These
are the necessary conditions.This allows us to arrive at the following definition: an additive
subgroup (I, +) ⊂ (R, +) is called an ideal if the following holds: I · R and R · I ⊂ I. Also,
prop: If R is a ring and I ⊂ R is an ideal, then R/I has a ring structure that is inherited
from R. Examples of Ideals include:

(1) Z, n ∈ Z, n · Z ⊂ Z is an ideal
(2) Q[x]
(3) If R is a commutative ring,, r1 , · · · , rn , {a1 r1 + · · · + an rn : a1 , · · · , an ∈ R} is an
ideal.

2. Friday, January 26: Prime and Maximal Ideals

2.1. Announcements.

• References: Artin Algebra, currently following Chapter 11. We will move to Chap-
ters 12 and 13 in the next couple weeks.
4 AMY FENG & EMMA CARDWELL & ZIYONG CUI

2.2. Ideals. Last time:

Proposition 2.1. Let R be a commutative ring with I an ideal. Then R/I is a ring.

Proof. (proved last time). □

Examples of ideals:

• Working in the ring Z, then for any n ∈ Z, n · Z = {. . . , −2n, 0, n, 2n, . . .} is an ideal.

• For the ring Q[x] and any element f ∈ Q[x], f · Q[x] = {f · h : h ∈ Q[x]} is an ideal.
Definition 2.2. For any commutative ring R and any finite collection of elements r1 , . . . , rk ∈
R, we can define the ideal generated by r1 , . . . , rn as
I = {a1 r1 + . . . + ak rk : ai ∈ R}.

We can also define the ideal generated by an infinite set by requiring all but a finite number
of the coefficients ai to be 0.
Definition 2.3. If R1 , R2 are rings with units, a map of rings f : R1 → R2 is a map of
additive abelian groups that respects multiplication, i.e. f (a) · f (b) = f (ab) for all a, b ∈ R1 .
Proposition 2.4. If f : R1 → R2 is a map of commutative rings, then
I = ker(f ) = {r ∈ R1 : f (r) = 0}
is an ideal, and f induces an injective map R1 /I → R2 , i.e. we have f (a) = f (b) if and only
if a = b.

Proof. We first check that I is an ideal. Consider r ∈ R1 such that f (r) = 0, so r ∈ I. It

suffices to show that r · s ∈ I for any s ∈ R1 . Since the map preserves the multiplicative
structure, we have
f (r · s) = f (r) · f (s) = 0 · f (s) = 0.
This shows that r · s ∈ I.

The second statement follows from the analogous result for abelian groups. □
Example 2.5. Consider the map
f : Q[x] → C,
map is defined by sending x 7→ a for some a ∈ C. For example, if we define the
where the √
map x 7→ 5, then a polynomial such as p(x) = x2 + 3x + 4 gets mapped via
√ √
f (p(x)) = ( 5)2 + 3( 5) + 4 ∈ C.
We then observe
√
ker(f ) = {g ∈ Q[x] : g( 5) = 0}.
This implies (x2 − 5) ⊆ ker(f ), and in this case, we actually have equality! (We will prove
this later).
 MATH 123 NOTES 5

Note that sometimes, the kernel is trivial. For example, the map given by f : x 7→ e, Euler’s
constant, as e is not the root of any polynomial with rational coefficients. This all connects
to a more general statement:

NOTE: From now on, we will assume ALL rings are commutative with unit unless explicitly
stated otherwise!
Lemma 1. Let h : R → S be a map of rings and β ∈ S. Then there exists a unique map
h̃ : R[x] → S such that h̃ = h on R and h̃(x) = β:
h̃
R[x] S

⊆ h

Proof. We define h̃ by
h̃(an xn + . . . + a1 x + a0 ) = h(an )β n + h(an−1 )β n−1 + . . . + h(a1 )β + h(a0 ) ∈ S.
It remains to check that this definition is compatible with addition and multiplication to
show that it is indeed a ring map. This is left as an exercise. □

2.3. Prime Ideals. We’ve seen that for the integers, n · Z is an ideal for any n ∈ Z. We also
have the notion of prime numbers - formally, nonzero n ∈ Z is prime if n | a · b implies n | a
or n | b for all a, b ∈ Z. It is a natural question to ask whether we can define an analogous
notion for ideals.
Definition 2.6. An ideal I ⊊ R is called a prime ideal if for any a, b ∈ R such that a·b ∈ I,
then either a ∈ I or b ∈ I.

Returning to the integers, this formulation is the same as that of prime numbers - if ab ∈ pZ,
then a ∈ pZ or b ∈ pZ. For any prime p, the ideal pZ is a prime ideal. In fact, there is only
one other prime ideal in Z - the zero ideal, defined as 0 · Z = {0}.

We might want to know if anything interesting happens when we take the quotient of a ring
by a prime ideal. But first:
Definition 2.7. A ring R is an (integral) domain if for a, b ∈ R such that a · b = 0, we
have a = 0 or b = 0.

This leads to a very useful statement:

Proposition 2.8. Let I ⊆ R be an ideal. Then I is a prime if and only if R/I is a domain.

Proof. Consider the quotient map f : R → R/I. First consider I prime. Then if we have
a, b ∈ R such that a · b ∈ I, then a ∈ I or b ∈ I. Then we know that if f (a · b) = 0, then
6 AMY FENG & EMMA CARDWELL & ZIYONG CUI

f (a) = 0 or f (b) = 0. This shows that f (R) is a domain. But by definition, f is surjective,
so we can conclude R/I is a domain. The reverse direction is analogous. □

As a simple example, we can see that Z is a domain, and Z/{0} = Z, so the zero ideal is
prime in Z.

2.3.1. Structure of Prime Ideals in a Ring - Yoga? Let S be a set, consider C(S) = {f :
S → C}. Yoga: any commutative ring comes from such a construction (see lecture 1). We
might wonder if, given S, we can recover the set of functions C(S). For any s ∈ S, we have
hs : C(S) → C given by the evaluation map f 7→ f (s). The image of hs is a domain, so
using Prop. 2.8, we see that ker hs is a prime ideal of C(S). In fact, ker hs turns out to be a
maximal ideal (more on this later). Anyways, we get a map
S → {prime ideals of C(S)}
s 7→ ker hs .
This turns out to be injective, but not necessarily bijective. If S is finite, then this is bijective.
Corollary 2.9. Consider h : R1 → R2 a map of rings, and R2 is a domain. Then ker(h) is
a prime ideal.

Proof. Any subring of a domain is a domain, so ker(h) is prime by Prop. 2.8. □

2.4. Maximal Ideals.

Definition 2.10. An ideal I ⊊ R is called maximal if, for any ideal J ⊊ R such that I ⊂ J,
we have I = J.
Proposition 2.11. Every ring has (at least) one maximal ideal.

Proof. Start with any ideal I1 ⊂ R. Then either I1 is maximal, or there exists I2 ⊂ R such
that IS1 ⊊ I2 . Either I2 is maximal, or there exists some I3 , and so on. Then, consider
Iω = j Ij . Either Iω is maximal, or we need to find some even bigger Ij ’s. To resolve this
we use the Axiom of Choice, which tells us that we can construct such a union. □
Definition 2.12. A ring R is a field if for any nonzero a ∈ R, there exists b ∈ R such that
ab = 1, i.e. b = a−1 .

Remarks:

• Some examples of fields: Q, R, C.

• Every field is an integral domain - for a field R, we observe that if a, b ∈ R and
a, b ̸= 0, then (ab)a−1 b−1 = 1, so ab ̸= 0.
Proposition 2.13. A ring R is a field iff the only proper ideal is the zero ideal, I = {0}.
 MATH 123 NOTES 7

Proof. First, assume R is a field with I ⊊ R a proper ideal. Consider 0 ̸= a ∈ I, then

1 = a · a−1 ∈ a · R ⊆ I · R ⊆ I. Then, since I contains 1, we have I = R, so I is not proper.

Next, assume that the only proper ideal is I = {0}. For any 0 ̸= a ∈ R, we then have
a · R ̸= 0. The ideal a · R is not proper, so a · R = R. This implies there exists some b ∈ R
such that a · b = 1. □
Proposition 2.14. For a ring R with ideal I ⊆ R, then I is maximal iff R/I is a field.

Proof. We will prove this next time! □

3. Monday, January 29

Theorem 3.1. If I ⊂ R is an ideal then I is maximal if and only if R/I is a field.

Proof.
Lemma 2. There is a bijection between the set of ideals J in R containing I and ideals J
in R/I.

We map J to J/I = J under the quotient map. As an exercise, prove that the quotient map
actually induces a bijection between these sets of ideals.

Now recall that a ring S is a field if and only if the only proper ideal of S is the 0 ideal. So
if R/I is a field, there must be two ideals in R containing I. These can only be I and R, so
I is maximal. If I is maximal, then R/I has two ideals (R/I and the zero ideal), so R/I is
a field. □
Definition 3.2. An ideal I ⊂ R is principal if I is generated by a single element.

Note: to show that an ideal is not principal, it doesn’t suffice to show that the ideal can
have multiple generators. For example, in Z, the ideal generated by 10 and 15 is generated
by neither 10 nor 15, but is generated by 5.

In Z, all ideals are principal - any ideal is generated by the GCD of its elements.
Theorem 3.3. If R is a field, then every ideal in R[x] is principal.

Proof. We basically proceed by the Euclidean algorithm. If f, g, ∈ R[x] with g ̸= 0, we show

that we can find r with deg(r) < deg(g) such that f = q · g + r.

We can write r = f − q · g. Take r to be the element of {f − qg : q ∈ R[x]} with minimal

degree. Now we want to show that deg r < deg g. Suppose deg r ≥ deg g. Write
r = bm xm + bm−1 xm−1 + · · · + b0 , bm ̸= 0, bi ∈ R
g = ct xt + ct−1 xt−1 + · · · + c0 , cj ∈ R, ct ̸= 0
8 AMY FENG & EMMA CARDWELL & ZIYONG CUI

where m ≥ t. Now we write

r′ = r − bm (c−1
t )x
m−t
+ lower order terms
which has lower degree than r. To complete the proof, take an element g ∈ I with lowest
degree, and now observe that for any other f ∈ I, f must have remainder 0 when divided by
g. Otherwise, we could write the remainder r as f − qg, and r is in I and has lower degree
than g. This is a contradiction.
Definition 3.4. A principal idea domain is a domain R where every ideal is principal.
Definition 3.5. Let R be a domain. 0 ̸= f ∈ R is a irreducible if f is not a unit and for
a, b ∈ R, if f | ab then f | a or f | b.
Corollary 3.6. An ideal I ⊂ R[x] is prime if and only if I = (f ) for irreducible f .
Remark 3.7. In any ring, a maximal ideal is prime.

Taking R/I where I is maximal ideal yields a field, which must be a domain. Then R/I is
a domain only if I is prime. □

In a polynomial ring, nonzero prime ideals are maximal.

4. Friday, February 2: Ideals in K[x] and Beyond

4.1. Prime and Maximal Ideals in K[x]. A note about notation: we usually use K to
represent fields and R to represent rings.
Proposition 4.1. Let K be a field, 0 ̸= I ⊆ K[x] an ideal. Then:

(1) I is a prime iff I = (f ) for some irreducible f ∈ K[x].

(2) If I is prime, then I is maximal.

Proof. We proved (1) last time.

To prove (2), first consider a nonzero prime ideal I. Suppose that I is not maximal. Then,
there exists some maximal ideal J such that I ⊆ J ⊊ K[x].

Note that J = (g) for some g ∈ K[x], and I = (f ) for f irreducible. Since f ∈ I ⊆ J, we
have f = g · h for some h ∈ K[x]. It follows from the irreducibility of f that either g is a
unit or h is a unit. If g is a unit, then J = (g) = K[x], contradicting the assumption that J
is maximal. If h is a unit, then the ideal generated by g will be equal to the ideal generated
by f , so I = J. As such, it follows that I is maximal.

□
Remark 4.2. The proof here relies on the statement that f ∈ K[x] is irreducible iff
whenever we have f = gh, then either g or h is a unit.
 MATH 123 NOTES 9

Definition 4.3. A field K is algebraically closed if for any nonconstant polynomial f ∈

K[x], there exists a ∈ K such that f (a) = 0.

More informally, being algebraically closed means that the field contains all of the roots of
all polynomials with coefficients in the field.
Proposition 4.4. If K is algebraically closed, then every maximal ideal I ⊆ K[x] has the
form (x − a) for some a ∈ K.

Proof. If a ∈ K, then x − a is irreducible, so by the previous proposition, (x − a) is maximal.

For the reverse direction, suppose we have I ⊆ K[x] with I = (f ). Since K is algebraically
closed, there exists a ∈ K such that f (a) = 0, so (x − a) | f (explanation: 1). Every maximal
ideal is prime, so f is irreducible. As such, we have f = k(x − a) for some k ∈ K, so f, x − a
generate the same ideal I. □
Remark 4.5. For non-algebraically closed fields K, we can have maximal ideals in K[x]
generated by polynomials of degree greater than 1. For example, (x2 + 1) is maximal in R[x].
Example 4.6. Consider K = C, which is an algebraically closed field. By by Prop. 4.4, we
have a bijection between
{maximal ideals of C[x]} ←→ {a ∈ C} = points of the complex plane
given by the map
C[x] −→ function : C → C
f 7−→ {f˜ : a 7→ f (a)}.

This is another example of Yoga - we can get the complex numbers from the ring, and we
can recover the ring from the complex numbers.

4.2. Nullstellensatz. Recall for any ring R, we can consider R[x]. We can repeatedly do
this:
R[x1 , . . . , xn ] = R[x1 , . . . , xn−1 ][xn ].

If you happen to be curious about the ideals of K[x1 , . . . , xn ] in particular, good news! Here’s
a cool theorem that you will talk about more if you take Math 137...
Theorem 4.7 (Hilbert’s Nullstellensatz). There is a bijection between
{maximal ideals of K[x1 , . . . , xn ]} ←→ {(a1 , . . . , an ) ∈ K n }
Remark 4.8. The map is clear in one direction: if we take (a1 , . . . , an ) ∈ K n and consider
the map h : K[x1 , . . . , xn ] → K given by xi 7→ ai , then ker(h) is a maximal ideal given by
(x1 −a1 , x2 −a2 , . . . , xn −an ) = {r1 (x1 −a1 )+r2 (x2 −a2 )+. . .+rn (xn −an ) | ri ∈ K[x1 , . . . , xn ]}.
1To see (x − a) | f : note that by the Euclidean algorithm, we have f = (x − a) · q + r, with deg r <
deg(x − a) = 1, so r = 0 or 0 ̸= r ∈ K. Evaluating at x = a, we get r(a) = 0, so we must have r = 0)
10 AMY FENG & EMMA CARDWELL & ZIYONG CUI

Note: it is clear that (x1 − a1 , x2 − a2 , . . . , xn − an ) is contained in ker(h) - it is left as an

exercise to show that we actually have equality!
Corollary 4.9. For any nonconstant f ∈ K[x1 , . . . , xn ], there exists (a1 , a2 , . . . , an ) such
that f (a1 , a2 , . . . , an ) = 0.

Before we prove the Nullstellensatz, we will review some other constructions:

4.3. Fields of Fractions. Let R be a domain. We want to construct the smallest field
containing R. Consider formal fractions ab for a, b ∈ R, b ̸= 0. We say ab ∼ dc (is
equivalent) if ad = bc in R. We define addition and multiplication in the usual way:
a c ac
· =
b d bd
a c ad + bc
+ = .
b d bd

Then, consider
F = {fractions}/ ∼ .

Proposition 4.10. (1) F is a field (this is the field of fractions, or fraction field of R)
(2) We can define an injective map h : R ,→ F given by a 7→ a1 .
(3) For any injective map R ,→ F ′ for F ′ a field, then there exists a unique extension of
h such that the map R ,→ F ′ factors through F :
h
R F′

∃!

Proof. We will prove these statements next time. □

5. Monday, February 5

We prove Prop. 4.10 from last lecture.

a
Proof. If b
∈ F is nonzero, then a is nonzero. We construct h̃, an extension of h. We define
a h(a)
h̃ = = h(b)−1 h(a)
b h(b)
Since h is injective, this makes sense in F ′ .
 MATH 123 NOTES 11

As for uniqueness, take any h̃ that extends h. Then we have

h̃(a/b) = h̃(1/b)(h(a)) = h̃(b)−1 h̃(a) = h(b)−1 h(a)
h̃(1/b)b̃ = h̃(1) = 1
h̃(1/b) = h(b)−1
□

Ly K be a field.
Definition 5.1. A vector space over K is a n abelian group (V, +) together with an operation
K × V → V , (r, v) → r · v such that

(1) (r1 · r2 ) · v = r1 · r2
(2) (r1 + r2 )v = r1 v + r2 v
(3) r1 (v1 + v2 ) = r1 v1 + r1 v2

Examples: If we have any ring R and K ⊂ R a field, then R is a vector space over R. K[x]
is a vector space over K.

Another point of view: Suppose (V, +) is an additive abelian group. Take the set of endomor-
phisms of V . To make this into a ring, define (f + g)v = f (v) + g(v) and (f · g)(v) = f (g(v)).
Then we can check that this is a ring.

For example, if V is Zn , the endomorphisms of V form the space of n × n matrices of Z.

A vector space over K is an abelian group (V, +), together with a ring map ϕ : K → End(V )
where (r, v) = ϕ(r)(v) = r · v.
Definition 5.2. A R − module is an abelian group (M, +) together with a ring map R →
End(M ).

Suppose we have M an R-module with M ⊂ R. Then M is an ideal.

Definition 5.3. Let {vi }i∈I ⊂ V . Then {vi }i∈I is independent if any finite linear combina-
tions of elements of {vi }i∈I is nonzero (aside from the trivial linear combination).
Definition 5.4. {vi }i∈I spans V if any v ∈ V can be written as a finite linear combinations
of elements of {vi }i∈I .
Definition 5.5. V is finite-dimensional if it has a finite spanning set.
Proposition 5.6. Let V be finite-dimensional over K, then any minimal spanning set of
independent, and any two minimal spanning sets have the same cardinality.

Then the cardinality of a minimal spanning set is the dimension of V .

Proof. If {v1 , v2 , . . . , vn } is spanning and minimal, take the equation a1 v1 + · · · + an vn = 0.

If a1 ̸= 0, multiply by a11 , then v1 + a2 a−1 −1
1 v2 + · · · + an a1 vn = 0. So {v2 , . . . , vn } is a smaller
spanning set. Thus a minimal spanning set is necessarily independent.
12 AMY FENG & EMMA CARDWELL & ZIYONG CUI

The second part is proved by taking a minimal spanning set {v1 , v2 , . . . , vn } and constructing
ϕ : K n → V where (a1 , a2 , . . . , an ) 7→ a1 v1 + a2 v2 + . . . , an vn . Then ϕ is a bijection because
{v1 , v2 , . . . , vn } is both spanning and independent.

Now we prove the second part of the Nullstellensatz (stated in previous lecture).

Proof. Let (a1 , . . . , an ) ∈ K n . Suppose m is a maximal ideal of K[x1 , . . . , xn ] for K uncount-

able and algebraically closed. Then K[x1 , . . . , xn ]/m is either K or is not algebraic over F .
1
In the first case, m is just (xi − ai ) for some ai . In the second case, { r−k : k ∈ K} is linearly
independent for transcendental r. But this is impossible because then K[x1 , . . . , xn ]/m has
uncountable dimension. □

6. Friday, February 9

Theorem 6.1. (Nullstellensatz) Suppose K is algebraically closed and uncountable. There is

a correspondence between the maximal ideals in K[x1 , · · · , xn ] and tuples (a1 , · · · , an ) ∈ K n
where there is a bijection from (a1 , · · · , an ) → (x1 − a1 , · · · , xn − an ).

Proof. Last time, we showed that (x1 − a1 , · · · , xn − an ) is a maximal ideal. Now, we will
show that every maximal ideal amounts to this form. Suppose m ⊂ K[x1 , · · · , xn ] is a
maximal ideal. This means that F = K[x1 , · · · , xn ]/m is a field. Consider the mapping
ϕ1 : K[x1 ] → K[x1 , · · · , xn ] → F . If m = (x1 − a1 , · · · , xn − an ), then F is isomorphic to K,
and the Ker(ϕ1 ) = (x1 −a1 ). Without knowing m has this form, if Ker(ϕ1 ) ̸= 0, then Ker(ϕ1 )
is a nonzero prime ideal, which is then also maximal because we are working in a polynomial
ring with one variable. Thus, it has the form Ker(ϕ1 ) = (x1 −a1 ). We then claim that ϕ1 is not
injective, which would mean, from above, that Ker(ϕ1 ) = (x1 − a1 ). Then, we can apply this
to Ker(ϕ2 ), · · · , Ker(ϕn ) where ϕ1 : K[xi ] → F to see that Ker(ϕi ) = (xi − ai ), which then
tells us that (x1 −a1 , · · · , xn −an ) ⊂ m. Since we already know that this is maximal, we arrive
at the desired result. Thus, it suffices to show that ϕ1 is not injective. Suppose FSOC that ϕ1
is injective and consider the fraction field K(x1 ). By injectivity, K(x1 ) could be thought of as
being embedded in the field F . Consider the following three claims: (1) the K-vector space
F = K[x1 , · · · , xn ]/m has a countable spanning set, which we will denote as {vi }i∈I , (2) if V
is a vector space with a countable spanning set, then every independent set in V is countable,
and (3) in the K vector space K(x1 ) the vectors { x11−a |a ∈ K} are independent. By (1), F
has a countable spanning set and (2) tells us that every independent set must be countable.
But (3) gives us an uncountable independent set in K(x1 ) ⊂ F, the desired contradiction.
Now, let us prove the three claims. For (1), the monomials in K[x1 , · · · , xn ] form a countable
spanning set in K[x1 , · · · , xn ], and so the images in F are also a countable spanning set. For
(2), suppose S = {v1 , v2 , · · · } ⊂ V is a countable spanning set and denote Sn = {v1 , · · · , vn }.
Let Vn = {a1 v1 + · · · + an vn : a1 , · · · , an ∈ K}, which is the span of Sn . Let L ⊂ V be an
independent set S and let Ln = L ∩ VnS , which is certainty independent as well in V (and Vn ).
Note that V = n Vn , and so L = n Ln , implying that L is countable. Finally, for (3),
 MATH 123 NOTES 13

b1 b2 bn
take distinct a1 , a2 , · · · , an ∈ K. We would like to show f = x−a 1
+ x−a 2
+ · · · + x−a n
=
0 implies that bi = 0 for all i (where bi ∈ K). Eliminating the denominators, we get
b1 (x − a2 ) · · · (x − an ) + b2 (x − a1 )(x − a3 ) · · · (x − an ) + · · · + bn (x − a1 ) · · · (x − an−1 ). Now,
take x = a1 , we get b1 (a1 − a2 ) · · · (a1 − an ) = 0 But since all the ai are distinct, then b1 = 0.
Reapplying this logic for all of the bi , we finish the proof for (3).

□
Corollary 6.2. Let R be a ring which is a quotient K[x1 , · · · , xn ]/I where I = (f1 , · · · , fn ) ⊂
K[x1 , · · · , xn ] is an ideal. Then there is a bijection between (a1 , · · · , an ) ∈ K where all
fi (a1 , · · · , an ) = 0 and maximal ideals m ∈ R.

Proof. Maximal ideals in R correspond to maximal ideals m ∈ K[x1 , · · · , xn ] such that

I ⊂ m. By the Nullstellensatz, any maximal ideal m ⊂ K[x1 , · · · , xn ] has the form (x1 −
a1 , · · · , xn −an ) for a1 , · · · , an ∈ K. The condition I ⊂ m just means that all ai are contained
in the kernel, which is equivalent to fi (a1 , · · · , an ) = 0 for all fi ∈ I. □

Consider the following example: f = y 2 − x(x − 1)(x − 2) ∈ C[x, y]. By above, we know that
the complex solutions of f = 0 correspond to the maximal ideals in C[x, y]/f .

7. Monday, February 12

7.1. R-modules.
Definition 7.1. Let R be a commutative ring with 1. An R-module is an abelian group M
on which R acts. That is, we have a definition of multiplication of elements of M by elements
of R where r(m1 + m + 2) = rm1 + rm2 , (r1 + r2 )m = r1 m + r2 m, and r1 (r2 m) = (r1 r2 )m.
Example 7.2. Let M = R. Then multiplication is just usual multiplication in R.
Example 7.3. Let R = Z. Then any M can be defined as a Z-module. For n ∈ Z, m ∈ M ,
we let nm = m + · · · + m (n times).
Example 7.4. Let M = Rn . Then multiplication is defined pointwise.

If M1 and M2 are R-modules, a map of R-modules is a homomorphism f : M1 → M2 of

abelian groups that respects multiplication. That is, r · f (m) = f (r · m).

Like with vector spaces, we can take the direct sum and tensor product of modules.

If M and N are R-modules, homR (M, N ) also has the structure of an R-module. We define
(r · f )(m) = rf (m) = f (r · m).

homR (M, R) is a special case of R-module which refer to as the dual space of M . If I is an
ideal of R, I is an R-module (why doesn’t this construction work when I is just a subring)?

If M is an R-module with N a submodule, then M/N is a module, where we can take the
quotient as the usual quotient of abelian groups.
Example 7.5. Let R = C[x], f ∈ R any polynomial.T Then C[x]/f C[x] is a C[x]-module.
14 AMY FENG & EMMA CARDWELL & ZIYONG CUI

7.2. Tensor Products. Let E1 , E2 , F be R-modules. A map h : E1 × E2 → F is called

bilinear if it is linear in each component. That is, if we fix e2 ∈ E2 , then h induces a linear
map E1 × e2 → F . h also induces a linear map e1 × E2 → F for any e1 ∈ E1 .
Definition 7.6. A tensor product E1 ⊗ E2 is an R-module equipped with a bilinear map
ϕ : E1 × E2 → E1 ⊗ E2 such that for any bilinear map ψ : E1 × E2 → F , there is a unique
ϕ′ : E1 ⊗ E2 → F such that ϕ′ circϕ = ψ.

Any two tensor products are canonically isomorphic, which is why we speak of ‘the’ tensor
product.

8. Friday, February 16: More Tensor Products (ft many commutative

diagrams)

8.1. The Tensor Product. Let R be a commutative ring with unit. Let E1 , E2 , F be
R-modules.
Definition 8.1. A tensor product E1 ⊗ E2 is an R-module with a bilinear map µ :
E1 × E2 → E1 ⊗ E2 such that for any bilinear map φ : E1 × E2 → F , there exists a unique
linear map φ̃ : E1 ⊗ E2 → F :

E1 × E2 φ F

µ
∃!φ̃

E1 ⊗ E2

Remark 8.2. Note that we sometimes notate this as E1 ⊗R E2 to specify we are tensoring
with respect to R.
Proposition 8.3. E1 ⊗ E2 exists.

Proof. Consider RE1 ×E2 , the free module over R generated from the set E1 × E2 . Consider
a map E1 × E2 → RE1 ×E2 given by
(e1 , e2 ) 7→ basis under ve1 ×e2 .

Then, consider another module F with bilinear map φ : E1 × E2 → F . Then there exists a
unique R-linear map φ̃ such that

E1 × E2 φ F

µ
∃!φ̃

RE1 ×E2
 MATH 123 NOTES 15

The only problem preventing RE1 ×E2 from satisfying the definition of a tensor product is
the map µ not necessarily being bilinear. To fix this, we can find a quotient RE1 ×E2 /N such
that µ′ is bilinear:
E1 × E2
µ′
µ

RE1 ×E2 RE1 ×E2 /N

To do this, we let N ⊆ RE1 ×E2 be the R-submodule generated in the following way:

We can define the generating set of N by considering all e1 , e′1 ∈ E1 , e2 , e′2 ∈ E2 , r ∈ R, and
for v(e1 ,e2 ) ∈ RE1 ×E2 , consider

• v(e1 ,e2 +e′2 ) − v(e1 ,e2 ) − v(e1 ,e′2 )

• v(e1 +e′1 ,e2 ) − v(e1 ,e2 ) − v(e′1 ,e2 )
• r · v(e1 ,e2 ) − v(re1 ,e2 )
• r · v(e1 ,e2 ) − v(e1 ,re2 ) .

Ok great, almost there. Now we have a bilinear map - to finish it off, we need to verify that
φ : E1 × E2 → F will induce a linear map φ̃ : M/N → F . To do this, we show that the
maps factor through the following diagram:

φ
E1 × E2 F

˜
φ̃

RE1 ×E2 φ̃

RE1 ×E2 /N

˜ ) = 0.
To show that this factors through φ̃, we want to show that φ̃(N

Then, since φ is bilinear, the R-linear map φ̃˜ factors through a map φ̃. □
Example 8.4. Let K be a field, V1 , V2 vector spaces of dimension m, n ∈ N+ . Let e1 e2 , . . . , en ∈
V and f1 , f2 , . . . , fm ∈ V2 be a set of basis elements. Then, V1 ⊗K V2 has a basis given by
ei ⊗ fj and dim V1 ⊗ V2 = mn.
Example 8.5. If I ⊆ R is an ideal and M is an R-module, then M ⊗R (R/I) ≃ M/(I · M ),
where I · M is a submodule of M generated by {i · M : i ∈ I, m ∈ M }. To see this, observe
16 AMY FENG & EMMA CARDWELL & ZIYONG CUI

that if j ∈ I, then for m ∈ M and r ∈ R/I, we have

j(m ⊗ r) = m ⊗ r · j = 0.
Example 8.6. If m, n are coprime integers, then (Z/mZ) ⊗Z (Z/nZ) = 0. This follows from
8.5, as we get
(Z/mZ) ⊗Z (Z/nZ) ≃ (Z/mZ)/nZ = Z/(m, n)Z = 0.
Remark 8.7. We can also do this for multilinear maps. Consider R-modules E1 , E2 , . . . , Es , F
with a multilinear map φ : E1 × E2 × . . . × Es → F such that for each i, if we fix ej ∈ Ej
for all j ̸= i, the map φ : e1 × . . . × Ei × . . . × es → F is linear. Then, we define the tensor
product analogously, with φuniv : E1 × E2 × . . . × Es → E1 ⊗ E2 ⊗ . . . ⊗ Es a multilinear map:

φ
E1 × E2 × . . . × Es F

φuniv ∃!

E1 ⊗ E2 ⊗ . . . ⊗ Es

It turns out that we can use our bilinear tensor product construction to interpret this mul-
tilinear tensor product:
Proposition 8.8. (E1 ⊗ E2 ) ⊗ E3 ≃ E1 ⊗ E2 ⊗ E3 ≃ E1 ⊗ (E2 ⊗ E3 ).

Proof. Consider

φ
(e1 , e2 , e3 ) E1 × E2 × E3 F

φuniv ×E3
∃! bilinear

((e1 ⊗ e2 ), e3 ) (E1 ⊗ E2 ) × E3 ∃! linear

(e1 ⊗ e2 ) ⊗ e3 (E1 ⊗ E2 ) ⊗ E3

Checking that E1 × E2 × E3 → (E1 ⊗ E2 ) ⊗ E3 is trilinear is left as an exercise! □

Proposition 8.9. For a commutative ring R, ⊗ is associative and commutative.
 MATH 123 NOTES 17

Proof. Associativity follows from Prop 8.8. To check commutativity, we have E1 × E2 ≃

E2 ×E1 . Since E2 ×E1 → E2 ⊗E1 is a bilinear map, we get a bilinear map E1 ⊗E2 → E2 ⊗E1 .
Analogously, we get maps from E2 × E1 ≃ E1 × E2 → E1 ⊗ E2 . We can also construct inverse
maps in the same way, so we get an isomorphism:

E1 × E2 ≃ E2 × E1 E2 ⊗ E1

∃!

E1 ⊗ E2

9. Friday, February 23: Wedges and the Determinant

Intuition for problem set 2, question 9: we have the duality:

Rings ←→ Spaces
ρ(X) ←
− X
ρ(X1 ) ⊗ ρ(X2 ) ←
− X1 × X2

Now back to class. . .

We have R a ring with unit and E1 , E2 are R-modules. Some properties of tensor products:

• Associativity: for E1 , E2 , E3 all R-modules, we have

(E1 ⊗ E2 ) ⊗ E3 ≃ E1 ⊗ (E2 ⊗ E3 ).

• E1 ⊗ E2 ≃ E2 ⊗ E1
• E1 ⊗R E2 is an R-module.

Definition 9.1. A bilinear map ϕ : M ⊗ M → N is called alternating if ϕ(m, m) = 0.

Remark 9.2. If m1 , m2 ∈ M , then

0 = ϕ(m1 + m2 , m1 + m2 )
= ϕ(m1 , m1 ) + ϕ(m2 , m1 ) + ϕ(m1 , m2 ) + ϕ(m2 , m2 )
=⇒ ϕ(m1 , m2 ) = −ϕ(m2 , m1 ).

Remark 9.3. If 2 ∈ R× , then the identity ϕ(m1 , m2 ) = −ϕ(m2 , m1 ) implies ϕ is alternating

bilinear.
18 AMY FENG & EMMA CARDWELL & ZIYONG CUI

Theorem 9.4. There exists ∧2 M such that for any alternating bilinear map ϕ : M ×M → N ,
we have
ϕ
M ×M N
∃! linear

M ⊗R M ∃! linear

∧2 M

Proof. We proceed by constructing ∧2 M . Consider ∧2 M = M ⊗R M/S, where S is the

R-submodule generated by m ⊗ m for m ∈ M :
n o
S = r1 (m1 ⊗ m1 ) + r2 (m2 ⊗ m2 ) + . . . + rn (mn ⊗ mn ) .

Since ϕ is alternating bilinear, we see that ker(ϕ) = S. □

Example 9.5. Let K be a field, V a 2-dimensional vector space over K with basis v1 , v2 .
What is ∧2 V ? Note that V ⊗K V is spanned by v1 ⊗ v2 , v1 ⊗ v1 , v2 ⊗ v1 , v2 ⊗ v2 . We then
have
∧2 V = V ⊗K V /span{v ⊗ v ∀v ∈ V }.
Equivalently, observe that
v ⊗ v = (av1 + bv2 )(av1 + bv2 )
= a2 v1 ⊗ v1 + abv1 ⊗ v2 + abv2 ⊗ v1 + b2 v2 ⊗ v2 .
As such, we can also view ∧2 V as
∧2 V = V ⊗ V /span{v1 ⊗ v1 , v2 ⊗ v2 , v1 ⊗ v2 + v2 ⊗ v1 }.
Furthermore, we can see that ∧2 V has dimension 1 and is spanned by v1 ⊗ v2 = v1 ∧ v2
(notation when we are working with wedge product).
Definition 9.6. Let M be an R-module. Then
M ⊗n = M ⊗ M ⊗ . . . M (n times)
∧n M = M ⊗n /S,
where S is the submodule generated by elements of the form
m1 ⊗ m2 ⊗ . . . ⊗ mn
such that mi = mj for some i ̸= j.

We can formulate a definition for a universal property from this.

 MATH 123 NOTES 19

Definition 9.7. A multilinear map

ϕ : M ⊗ M ⊗ ... ⊗ M → N
is called alternating if
0 = ϕ(m1 , m2 , . . . , mn )
if mi = mj for some i = j. (i.e. if two entries are equal, the expression is equal to 0).
Remark 9.8. Note: the condition above works in all cases. If 2 ∈ R× , this condition is
equivalent to the statement
ϕ(m1 , . . . , mi , . . . , mj , . . . , mn ) = −ϕ(m1 , . . . , mj , . . . , mi , . . . , mn ).
(i.e., if we swap two mi , mj with i ̸= j, the resulting expression is equal to −1 times the
original).
Remark 9.9. We represent elements in multilinear wedge products as m1 ∧ . . . ∧ mn .
Theorem 9.10. If ϕ is a multilinear alternating map, then
ϕ
M × M × ... × M N

M ⊗n ∃! linear

∧n M

9.1. The Determinant. Suppose M ≃ Rn for some n with basis v1 , . . . , vn .

Example 9.11. For example, let n = 2 and let R = K be a field. Then ∧2 M is a 1-
dimensional space spanned by v1 ∧ v2 . Suppose h : M → M is a linear map. Then we get
an induced map ∧2 h : ∧2 M → ∧2 M .

10. Monday, February 26:

Proposition 10.1. If M ∼ Rn then ∧∗ M is free of rank 1. That is, ∧∗ M ∼ R.

Proof. Consider permutations; that is, bijections

σ : {1, 2, . . . , n} → {1, 2, . . . , n}
We have M ∼ Rn , so let M have basis {r1 , r2 , . . . , rn }. Then M ⊗ · · · ⊗ M is generated
by w1 ⊗ w1 ⊗ wn where each wj = σj over all permutations σ. Then we take the map
M ⊗ · · · ⊗ M → ∧n M where w1 ⊗ · · · ⊗ wn 7→ w1 ∧ · · · ∧ wn .

̸ wj for some i ̸= j then wi = vσ(i) for some permutation σ. So we can get to any
If wi =
basis tensor of M ⊗ · · · ⊗ M to any other basis tensor by applying some permutation to its
20 AMY FENG & EMMA CARDWELL & ZIYONG CUI

components. In the wedge product, this just corresponds to multiplying the tensor by ±1
based on sign. So we can pick w1 ∧ · · · ∧ wn to generate the whole tensor. □

Now assume R = K is a field. We have dim ∧n M = 0 or 1. To show that ∧n M = 1, we have

to construct some nonzero alternating map M × · · · × M → R.

Define a map M × · · · × M → R where

(
0 wj = wj for some i ̸= j
(m1 , m2 . . . , mn ) 7→
ε(r) w1 = wj
Lemma 3. Let Sn be the symmetric group. Then there is a homomorphism ε : Sn → {±1}
where if σ is the product of j transpositions then ϵ(σ) = (−1)j .
Q
Proof. Let P (x1 , . . . , xn ) = i<j (xi − xj ). For σ ∈ Sn , let
P (xσ(1) , . . . , xσ(n) )
ε(σ) = = ±1
P (x1 , . . . , xn )

Then for σ, τ ∈ Sn , we have

P (xστ (1) , . . . , xστ (n ) P (xτ (1) , . . . , xτ (n )
ε(στ ) = · = ε(σ)ε(τ )
P (x1 , . . . , xn ) P (xτ (1) , . . . , xτ (n) )
Then it is easy to check that for some transposition π, ε(π) = −1. □

11. Friday, March 1: Review

We let R be a commutative ring with unit.

Definition 11.1. An ideal m ⊊ R is maximal if for any ideal n ⊊ R such that m ⊆ n, we
have m = n.
Definition 11.2. An ideal p ⊆ R is prime if whenever ab ∈ p for a, b ∈ R, then either a ∈ p
or b ∈ p.
Proposition 11.3. Every prime ideal is maximal.
Proposition 11.4. An ideal I is maximal if and only if R/I is a field. I is prime if and
only if R/I is a domain (integral domain).
Proposition 11.5 (Correspondence Theorem from Artin). There is a bijection between the
set of ideals of R/I and the set of ideals in R that contain I. Equivalently:
A = { Ideal n : I ⊆ n ⊊ R} ←→ { Ideal n ⊊ R/I} = B.
Corollary 11.6. An ideal I is maximal in R if and only if A = {I}, which is true if and
only if B = {0}.
Definition 11.7. Let S be a set. The free module RS is defined as
RS = {(rj )j∈S : rj ∈ R}.
 MATH 123 NOTES 21

Proposition 11.8. If M is an R-module,

Hom(RS , M ) ≃ Homsets (S, M ).
Definition 11.9. If M1 , M2 , N are R-modules, there exists M1 ⊗ M2 such that for any
bilinear map M1 × M2 → N , the map factors through a unique linear map M1 ⊗R M2 → N .

bilinear
M1 × M2 N

(canonical map)
∃! linear

M1 ⊗ M2

For full proof of existence of tensor product, see Friday, Feb. 16 lecture notes. Essentially,
we define M1 × M2 → RM1 ×M2 → M1 ⊗ M2 .

We can also define the wedge product, where we consider M × M → M ⊗ M → ∧2 M . See

Friday, Feb. 23 lecture notes.
Theorem 11.10 (Nullstellensatz). If K is an algebraically closed field, any maximal ideal
in K[x1 , . . . , xn ] has the form (x1 − a, . . . , xn − an ) for a1 , . . . , an ∈ K.
Remark 11.11. If a1 , . . . , an ∈ K, then we get the map K[x1 , . . . , xn ] → K by taking
xi 7→ ai . This shows that (x1 − a, . . . , xn − an ) is maximal. The proof that every maximal
ideal has this form is harder. For full proof, see Friday, Feb. 9.

12. Friday, March 8

12.1. Modules over a PID. Let R be a domain.

Definition 12.1. R is a principal ideal domain (PID) if every ideal I ⊊ R is principal,
I = (f ).
Example 12.2. R = Z, R = K[x] for any field K are PIDs.

Recall that f ∈ R is called irreducible if the following are all satisfied:

• f ̸= 0
• f is not a unit (recall that a unit refers to any element u ∈ R such that there exists
some element v that satisfies uv = 1, ie it has a multiplicative inverse).
• If f = ab, a, b ∈ R, then a or b is a unit.
Proposition 12.3 (Unique Factorization in PIDs). If R is a PID, every nonzero element
h ∈ R that is not a unit can be written as a product h = f1 · · · fn for irreducible elements
fi . Furthermore, this factorization is unique up to multiplication by units.

We have an analogous concept when working with modules. First, some preparation:
22 AMY FENG & EMMA CARDWELL & ZIYONG CUI

• For a ring R, if f ∈ R, R/f · R is an R-module.

• If M1 , M2 are R-modules, M1 ⊕ M2 is an R-module
• An R-module M is called finitely-generated if there exists m1 , . . . , ms ∈ M such
that

M = {a1 m1 + . . . + as ms | ai ∈ R}.

Now, onto the main statement:

Proposition 12.4 (Structure Theorem for Finitely Generated Modules over a PID). Let M
be a finitely generated module over a PID R. Then M is a finite direct sum

M = M0 ⊕ M1 ⊕ . . . ⊕ Ms , s ∈ N+ ,

where M0 = Rt for t ∈ N+ and for i ≥ i, we have Mi = R/fini for fi irreducible.

2
This is a generalization of the structure theorem for finitely generated abelian groups.

12.2. Jordan Canonical Form. We will now look at a cool application: the Jordan canon-
ical form.

Proposition 12.5. Let K be an algebraically closed field. Let V be a finite dimensional

K-vector space. Consider φ : V → V a K-linear map. Then there exists a decomposition

V = V1 ⊕ V2 ⊕ . . . ⊕ Vm

such that φ : φ1 ⊕ φ2 ⊕ . . . ⊕ φm such that φi : Vi → Vi and ϕi is given by a Jordan block,

which is a matrix of the following form:
 
λi 1 0
..
 . 
ϕi ∼  λi ∈ K,
 
 . .. 1 


0 λi

i.e. V has a basis e1 , . . . , emi where mi := dim Vi .

2For
R = Z, the proposition is the classification of finitely generated abelian groups: If A is a finitely
generated abelian group, we have

A ≃ Zt ⊕ (Z/pm ms
1 Z) . . . ⊕ (Z/ps Z)
1

for primes p1 , . . . , ps ∈ Z.
 MATH 123 NOTES 23

Then, φi (ej ) = λi ej + ej+1 , and φ can be represented as

 
λ1 1
.. ..
. .
 
 
 . 

 λ1 . . 

 .. .. 

 . . 
.
 ... ... 
 
 . 

 λm . . 

 .. 
 . 1
λm

Proof. We make V into a K[X] module by letting X act as φ. I.e. if v ∈ V , we let

X · v = φ(v), X 2 · v = φ(φ(v)) = φ2 (v), and so on.

Note: if a ∈ K, then a · X = X · a ∈ K[X], so we want a · φ(v) = φ(a · v). This is alright as

φ is linear. Now, V is a K[X] module iff we get a map K[X] → EndV given by X 7→ ϕ.

By the Proposition on Modules over a PID (see 12.4), there exists a decomposition
V = V0 ⊕ V1 ⊕ . . . ⊕ Vm ,
where V0 ≃ K[X]t , Vi ≃ K[X]/fimi for fi irreducible, mi ∈ N+ .

As V is a finite dimensional vector space over K, we have V0 = 0. As K is algebraically

closed, we have fi = X −λi for λi ∈ K. As such, each Vi is of the form Vi = K[X]/(X −λi )mi .

Let φi : Vi → Vi given by v 7→ X · v.

Now, we can construct a basis for Vi . Let

e1 = 1
e2 = X − λi
..
.
..
.
emi = (X − λi )mi −1

Note: why is this a basis? Let Y = X − λi . We have K[X] ≃ K[Y ]. Then, observe that
K[X]/(X − λi )mi ≃ K[Y ]/Y mi , so the e′i s correspond to the basis 1, Y, Y 2 , . . . , Y mi −1 .

Then, we have
φ(ej ) = X · ej = (X − λi )ej + λi ej
= λi ej + ej+1 .
As such, we have a representation of φ in the desired form. □
24 AMY FENG & EMMA CARDWELL & ZIYONG CUI

12.3. Characteristic Polynomials. If M is a finitely generated free R module, M = Rn

for n ∈ N+ .

If φ : M → M is a linear map, we have

det φ : ∧n M → ∧n M.

Then consider the map

T − φ : M ⊗R R[T ] → M ⊗R R[T ].
If φ ∼ A is an n × n matrix, then T − φ ∼ T · Id − A.
Definition 12.6. The characteristic polynomial is
Pφ (T ) = Char(φ|M ) = det(T − φ|M ⊗R R[T ]) = det(T · Id − A).
Corollary 12.7 (Cayley-Hamilton Theorem). Pφ (φ) = 0.
Remark 12.8 (Proof sketch of C-H for algebraically-closed fields). If R = K an algebraically
closed field, then φ can be put in Jordan Canonical form. Then to check that φ satisfies its
own characteristic polynomial, it suffices to check that each φi satisfies its own characteristic
polynomial by observing (φ − λi )mi acts as (X − λi )mi ≡ 0 on V .

13. Monday, March 18

Let K be a field. We say that L/K is a field extension of K if K ⊂ L. L/K is a finite

extension if the dimension of L over K is finite.

Note: the slash notation in a field extension is simply to indicate which field is being extended
and does not correspond to any notion of “quotient.”
Definition 13.1. Gal(L/K) = Aut(L/K) = {σ ∈ Aut(L) | σ|K = id}. We call this the
Galois group of L/K.
√ √
Example
√ 13.2. If L/Q is
√ quadratic, dimQ (L) = 2. Q( 2) = {a + b 2 | a, b√
∈ Q}. That is,
{1, 2} is√a basis for Q( 2) over Q. We can define an automorphism of Q( 2) that
√ √ sends
2 to − 2. Since the automorphism must fix everything in Q, knowing where 2 goes
determines the entire automorphism.
√ √
Observe also that Q( 2) is isomorphic
√ to Q[x]/(x2 − 2), since we can send 2 to x or −x.
We can do the same for any Q( d) where d is a square-free integer.
Example 13.3. Let ζn be an nth root of unity. For ζp where p is prime, Gal(Q(ζp )) is
isomorphic to Z/Zp× . If we instead take Q(ζn ) where n is not necessarily prime, Gal(Q(ζp ))
has cardinality φ(n), where φ is the totient function. It is isomorphic to Z/Zn× .
Lemma 4. Suppose K ⊂ L is a field. Let α be an element such that f (α) = 0 for some
f ∈ K[x]. Then K(α) is a subfield of L.

Proof. Since f (α) = 0, we have αn = −cn−1 αn−1 − · · · − c0 . So K(α) is spanned by

1, α, · · · , αn−1 , as any higher power of α can also be written in terms of 1, α, · · · , αn−1 . □
 MATH 123 NOTES 25
√ √
What is the Galois group of Q( 3 2)? We know√that in C, the cube
√ roots of 2 are 3 2 and two
complex numbers. So an automorphism of Q( 3 2) must send 3 2 to itself, as it only contains
real numbers.
√ √
What happens when we adjoin ζ3 ? Remember that 3 2ζ3 is a cube root of 2, and 3 2ζ32√is
3
the remaining cube root of 2. Suppose we take a automorphism √ of this
√ field extension.
√ 2
may go to a real or complex cube root of 2. That is, it can go to 3 2, 3 2ζ3 , or 3 2ζ32 . But ζ3
must go to another complex root of unity, so it can go to itself or ζ32 . Thus the Galois group
has order 6 and is isomorphic to Z/3/Z · Z/2Z.

Remark 13.4. If L/K is a finite extension, |Gal(L/K)| ≤ [L : K], where [L : K] denotes

the dimension of L over K.

When these two quantities are equal, we say that L/K is a Galois extension.

Theorem 13.5. If L/K is Galois, there is a 1-1 correspondence between subgroups of

Gal(L/K) and subfields of L containing K.

14. Friday, March 22: WOW more Galois! + symmetric functions

Recall that we are working with finite field extensions L/K, and

• Gal(L/K) = Aut(L/K) = {σ : L → L | σ|K = Id}

• L/K is called Galois if |Gal(L/K)| = [L : K] = dimLK .
√
3
Example 14.1. As an example, consider the field extension Q(ζ3 , 2)/Q. We have
 √ 
3
Gal ζ3 , 2/Q ≃ Z/3Z ⋊ Z/2Z.

Also, the degrees of the extensions are:

√
• [Q(ζ3 , 3 2) : Q(ζ3 )] = 3
• [Q(ζ3 ) : Q] = 2.

Suppose we have some intermediate extension, K ⊆ K ′ ⊆ L. We define

GK ′ = {σ ∈ Gal(L/K) | σ|K ′ = Id}.

Additionally, suppose we have a subgroup of the Galois group G ⊆ Gal(L/K). We define

KG′ = {k ∈ L | σ(k) = k ∀σ ∈ G}.

This allows us to state a very important theorem:

26 AMY FENG & EMMA CARDWELL & ZIYONG CUI

Theorem 14.2 (Galois Correspondence). If L/K is Galois, we have a 1 to 1 correspondence

between
{ subgroups G ⊆ Gal(L/K)} ←→ {K ⊆ K ′ ⊆ L}
G −→ K/G
GK ′ →−7 K ′ .

Proof. Proved last time. □

We will now look at an application of Galois theory:

14.1. Symmetric functions. Let u1 , u2 , . . . , un be variables and consider the action of the
symmetric group Sn (group of permutations on n elements) on {u1 , . . . , un }. A question
we may be interested in is, what functions in these variables are invariant under this group
action?

For example, a polynomial of the form

(x − u1 ) · · · (x − un )
is invariant under the action of Sn , as Sn just permutes the roots!

On the other hand, a polynomial of the form

(x − u1 )2 (x − u2 ) · · · (x − un )
would not be invariant under the action of Sn , as swapping u1 with any other uj will yield
a different polynomial!

Before we try to classify all of the polynomials invariant under Sn , consider

s0 = 1
s1 = u1 + u2 + . . . + un
!
X
s2 = u1 u2 + u1 u3 + . . . ui uj
i̸=j
..
.
sn = u1 u2 · · · un .

Punchline: if g is a polynomial in the ui and σ(g) = g for all σ ∈ Sn , then g is a polynomial

in the si ’s!

We can state this more formally: let R be a ring and consider the action of Sn on R[u1 , . . . , un ].
We define
R[u1 , . . . , un ]Sn = {g ∈ R[u1 , . . . , un ] | σ(g) = g ∀σ ∈ Sn }.
This is essentially defining the elements of R[u1 , . . . , un ] that are fixed under the action of
Sn .
 MATH 123 NOTES 27

Theorem. If g ∈ R[u1 , . . . , un ]Sn , then there exists some unique G ∈ R[z1 , . . . , zn ] such that
g = G(s1 , . . . , sn ).

Before we prove this, we want to make our statement even more precise!
Definition 14.3. If g ∈ R[u1 , . . . , un ] such that
X
g= aI u I
where I = {i1 , i2 , . . . , in } and uI = ui11 ui22 · · · uinn , then we define the degree of a monomial
uI as
X
deg uI = ij ,
and we define the degree of g as
deg g = max deg uI .
Remark 14.4. We have deg si = i.
Definition 14.5. For a monomial z I = z1i1 z2i2 · · · znin , we define the weighted degree as
degw z I = i1 + 2i2 + . . . + nin .
ai z I as
P
We then define the weighted degree for an arbitrary polynomial G =
degw G = max degw z I

With these different notions of degree, we get a more precise theorem statement:
Theorem 14.6. If g ∈ R[u1 , . . . , un ]Sn , then there exists some unique G ∈ R[z1 , . . . , zn ] such
that g = G(s1 , . . . , sn ) and for this G, we have degw G = deg g.

But what does this have to do with Galois theory? Observe that
X
(x − u1 ) · · · (x − un ) = (−1)i si xn−i .
We can take any field (in this case we just pick Q) and add to it the set of elementary
symmetric functions. Then, observe that
   
Q(s1 , . . . , sn ) = Frac Q[s1 , . . . , sn ] ⊆ Frac Q[u1 , . . . , un ] = Q(u1 , . . . , un ).
Somehow, thinking more about the relationships between all of these fields (using Galois
theory!) will give us insight into our symmetric polynomials question.

Now for some Galois theory. . .

We have a field extension Q(u1 , . . . , un )/Q(s1 , . . . , sn ). What is the Galois group of this
extension? Without rigorously proving anything, we can already see that the Galois group
contains the symmetric group Sn , as all elements of Q(s1 , . . . , sn ) are invariant under Sn . It
turns out that the Galois group is exactly equal to Sn :)

Then we can see a special example of the Galois correspondence - if we take G to be the
full symmetric group, the Galois correspondence tells us that the only elements of the upper
field Q(u1 , . . . , un ) fixed under G are the elements of Q(s1 , . . . , sn ).
28 AMY FENG & EMMA CARDWELL & ZIYONG CUI

Now we will more on to the proof:

Proof of Theorem 15.1. One direction is more straightforward (exercise: do it!)

For the harder direction, it suffices to find a polynomial G such that degw G ≤ deg g. We
induct on n.

For n = 1, there’s no nontrivial group action, so we are done!

Let g 0 (u1 , u2 , . . . , un−1 ) = g(u1 , u2 , . . . , un−1 , 0). We can see that g 0 (u1 , . . . , un−1 ) ∈ R[u1 , . . . , un−1 ].
By induction, there exists some unique Q ∈ R[z1 , . . . , zn−1 such that
g 0 (u1 , . . . , un−1 ) = Q s01 , . . . , s0n−1 ,


(1)
where
s01 = u1 + . . . + un−1
..
.
sn−1 = u1 · · · un−1 .

We also have degw Q = deg g 0 by our induction hypothesis.

Now, let
p(u1 , . . . , un ) = g − Q(s1 , s2 , . . . , sn−1 ).

Then, p is symmetric (invariant under Sn ), and from our construction of Q (see Equation
1), we see that p0 (u1 , . . . , un−1 ) = 0. This implies un | p0 . By symmetry, since p is invariant
under the action of Sn , we can see that ui | p for each i (key observation!). Then, since each
ui divides p, their product must also divide p, so we have sn = u1 · · · un | p. As such, we
have
p(u1 , . . . , un ) = sn · h, h ∈ R[u1 , . . . , un ].

We can also calculate the degree of h:

n o
deg h = deg p − n ≤ max deg g, deg Q(s1 , s2 , . . . , sn ) − n
= deg g − n.

Note that
deg Q(s1 , . . . , sn−1 ) ≤ degw Q(z1 , . . . , zn−1 ) = deg g 0 ≤ deg g.

Take g = Q(s1 , . . . , sn−1 ) + sn h. By induction on degree of g, there exists H ∈ R[z1 , . . . , zn ]

such that h = H(s1 , . . . , sn ) with degw H = deg h. Then, take
G = Q(z1 , . . . , zn−1 ) + zn · H(z1 , . . . , zn ).
Then G(s1 , . . . , sn ) = g, and degw G = degw (Q + zn H) ≤ max(degw Q, degw zn H).
 MATH 123 NOTES 29

As such, we see that

degw G ≤ max(deg g, deg h + n)
= deg g.
This shows existence. Next time, we will prove uniqueness. □

15. Monday, 03/25

We finish the proof from last time:

Theorem 15.1. If g ∈ R[u1 , . . . , un ]Sn , then there exists some unique G ∈ R[z1 , . . . , zn ] such
that g = G(s1 , . . . , sn ) and for this G, we have degw G = deg g.

Proof. We induct on n and deg g. Our base case is n = 1, deg g = 0. We assume the
statement holds for all polynomials g ′ in n′ variables where either n′ < n or n′ = n and
deg′ < deg g.
Lemma 5. If G ∈ RR[z1 , . . . , zn ] and G(s1 , . . . , sn ) = 0, then G ≡ 0.

Proof. We induct on n. Assume G ̸= 0. Let znt be the biggest power of zn dividing G. Then
(G/znt )(s1 , . . . , sn ) = 0, 0 ≡ G(s1 , . . . , sn ) = (G/znt )(s1 , . . . , sn )stn . Now set un = 0. Recall
that g(u, . . . , un−1 , 0) = g 0 (u1 , . . . , un−1 ).

Now
0 ≡ (G/znt )(s01 , . . . , s0n−1 , 0)
=⇒ (G/znt )(z1 , . . . , zn−1 , 0) = 0
=⇒ zn | G/znt .
which contradicts the minimality of t. □

If G1 , G2 ∈ R[z2 , . . . , zn ] with G1 (s1 , . . . , sn ) = G2 (s1 , . . . , sn ) = g, then (G1 −G2 )(s1 , . . . , sn ) =

0. Applying the lemma, we get G1 − G2 ≡ 0, so G1 = G2 . Now we are done. □
Definition 15.2. Let L/K be a field extension of finite degree. Take f ∈ K[x]. We say
that f splits completely in L if f (x) = (x − α1 ) · · · (x − αn ) with αi ∈ L. We say that L is a
splitting field for f if f splits completely in L and L = K(α1 , . . . , αn ).
√ √ √ √
Example 15.3. Q( 3 2, ζ3 ), f (x) = x3 − 2. Then f (x) = (x − 3 2)(x − 3 2ζ3 )(x − 3 2ζ32 .
Corollary 15.4. Let L/K be a finite degree extension. Let f ∈ K[x] be monic such that
f splits completely in L. So f (x) = (x − α1 ) · · · (x − αn ). If g(u1 , . . . , un ) ∈ K[u1 , . . . , un ]Sn
then g(α1 , . . . , αn ) ∈ K.

Proof. Let g = G(s1 , . . . , sn ) with GQ∈ K[z1 , . . . , zn ]. It’s enough to show that si (α1 , . . . , αn ) ∈
K. f (x) = (x − α1 ) · · · (x − αn ) = ni=0 (−1)i si (α1 , . . . , αn )xn−i .
30 AMY FENG & EMMA CARDWELL & ZIYONG CUI

Lemma 6. (a) If F ⊂ K ⊂ L are finite extensions of fields, then L/F splitting implies that
L/K is splitting. (b) Every f ∈ K[x], where f is monic, has a splitting field. (c) Every finite
L′ /K is contained in a splitting field.

Proof. (a) is clear. If L = F (α1 , . . . , αn ) for some f ∈ F [x], f (x) = (x − α) · · · (x − αn ).

Then L = K(α1 , . . . , αn .

(b) We take L to be K adjoined with the roots of f . We want to show that there is some
L̄/K, a finite extension, such that f splits in L̄. We induct on deg f . If deg f = 1, we are
done. Otherwise, let f1 be an irreducible factor of f . Write L1 = K[y]f1 (y). Write f (x) =
f1 (x)g1 (x). We know that (x − α1 ) divides f (x) in L1 [x]. Apply the induction hypothesis to
f (x)/(x−α1 ). Then there is some L̄ extending L1 where f (x)/(x−α1 ) = (x−α2 ) · · · (x−αn )
for ai ∈ L̄.

(c) Let K ′ /K be a finite extension. □

16. Friday, March 29: Splitting Fields

Definition 16.1. A finite extension of fields L/K is called a splitting field if L = K(α1 , . . . , αn )
with αi ∈ L such that f (x) = (x − α1 ) · · · (x − αn ) ∈ K[x].
√
Example 16.2. Q( 3 2) is not a splitting field over Q, see Splitting Theorem (Theorem 16.3)!
Lemma 7. (1) If we have a tower of finite field extensions F ⊆ K ⊆ L with L/F
splitting, then L/K is also splitting.
(2) Every polynomial f ∈ K[x] has a splitting field.
(3) Every finite extension L′ /K is contained in a splitting field.

Proof. For (3), let L′ = K(γ1 , . . . , γn ) for γi ∈ L′ (we can do this because L′ is a finite
extension). Each γi is the root of some polynomial gi ∈ K[x], so gi (γi ) = 0. We now want
to construct some larger extension L ⊇ L′ ⊇ K such that each gi factors into monic terms
in this larger extension. We do this by letting L be a splitting field for the polynomial
g = g1 · · · gn ∈ K[x] ⊆ L′ [x]. More explicitly, we have g = (x − α1 ) · · · (x − αs ) and let
L = L′ (α1 , . . . , αs ). Then, L is a splitting field for the polynomial g ∈ K[x], so L/K is
splitting. □

Note that being a splitting field means that there exists a polynomial whose roots will
generate the extension. But, given a field extension, how do we know whether it is splitting?
We could spend a possibly infinite amount of time by testing every single possible polynomial,
or we can use the splitting theorem!
Theorem 16.3 (Splitting Theorem). Let L/K be a splitting field for f ∈ K[x]. If an
irreducible polynomial g(x) ∈ K[x] has a root in α ∈ L with g(α) = 0, then g splits
completely in L.
 MATH 123 NOTES 31

Remark 16.4. This gives us a way to determine if a field is NOT a splitting field - if we
can find an irreducible polynomial with one root in the extension L but the polynomial does
NOT split completely, then the field extension is NOT splitting.

Proof of Splitting Theorem. Let f (x) = (x − α1 ) · · · (x − αn ) for αi ∈ L. We have L =

K(α1 , . . . αn ). We’ve seen that K(αi ) = K[αi ] ≃ K[x]/(f (x)) if f is irreducible, so we have

L = K(α1 , . . . , αn ) = K[α1 , . . . , αn ].

If β1 ∈ L is a root of g, write β1 = p1 (α1 , . . . , αn ) for p1 ∈ K[u1 , . . . , un ]. Then consider

σ ∈ Sn (set of permutations on n elements). We have

{p1 , . . . , pk! } = {σ(p1 ) = p1 (uσ(1) , . . . , uσ(n) | σ ∈ Sn }.

Then, let βj = pj (α1 , . . . , αn ).

We claim that

h(x) := (x − β1 ) · · · (x − βk ) ∈ K[x].

Assuming this claim, then we have g(x) | h(x). Since g is irreducible, this implies g must be
the minimal polynomial of β1 . (this was a lemma from last class). This then shows that g
splits completely in L.

Now to prove the claim: let w1 , . . . , wk and let s′1 (w), . . . , s′k (w) be elementary symmetric
functions in the w’s. The coefficients of h are (±1)

s′1 (β1 , . . . , βk ), s′2 (β1 , . . . , βk ), . . . , s′k (β1 , . . . βk ).

We want to show that these all lie in K.

If σ ∈ Sn , we have
   
σ s′i (p1 (u1 , . . . , un ), . . . , pk (u1 , . . . , un ) = s′i p1 (u1 , . . . , un ), . . . , pk (u1 , . . . , un ) .

This implies
 
s′i p1 (u1 , . . . , un ), . . . , pk (u1 , . . . , un ) ∈ K[u1 , . . . , un ]

is symmetric in u1 , . . . , un . As such, it is in K. □

Theorem 16.5. Let L/K be a finite extension of fields of characteristic 0. Then, there
exists α ∈ L such that L = K(α).

Proof. (Next time). □

32 AMY FENG & EMMA CARDWELL & ZIYONG CUI

17. Monday, April 1

Theorem 17.1. Let L/K be a Galois extension. Then there is a bijective correspondence
between K ′ such that {K ⊂ K ′ ⊂ L} and {H ⊂ G} where G is the Galois group of L/K.
K ′ maps to the subgroup of G that fixes K ′ , and a subgroup H of G maps to its fixed field.
Lemma 8. If f ∈ K[x] is irreducible, α and α′ are roots of f in L and L′ , respectively,
for extensions L/K and L′ /K. Then there is some unique isomorphism between K(α) and
K(α′ ).

Proof. K(α) ∼
= K[x]/f (x) ∼
= K(α′ ). The first map sends α to x, and the second map sends
′
x to α . □
Proposition 17.2. (a) Let f ∈ K[x], an extension L/K contains at most the splitting field
for f .

(b) Any two splitting fields for f are isomorphic.

Proof. (a) If L contains a splitting field for f then

f (x) = (x − α1 ) · · · (x − αn ) ∈ L[x]
and any splitting field = K(α1 , · · · , αn .

(b) If K1 , K2 are splitting fields for f , then K1 = K(α) for some α ∈ K1 . Let g ∈ K[x] be
a minimal polynomial for α.

Now let α′ be a root of g in some extension L/K. Then by the lemma, we have K1 ∼
= K(α) ∼
=
′
K(α ) ⊂ L ⊂ K2 . By (a), we have K1 = K2 in L. So K1 and K2 are isomorphic. □

Assume that K has characteristic zero.

Theorem 17.3. If L/K is finite, L = K(α) for some α ∈ L. This is called the primitive
element theorem.

Proof. Since L/K is finite, we can write L = K(α1 , . . . , αn ) with αi ∈ L. We induct on n.

We assume that K(α1 , . . . , αn−1 ) = K(β) for some β ∈ K(α1 , . . . , αn−1 . Then we can write
L = K(α, β).
Lemma 9. For all but finitely many c ∈ K, γ = β + cα. Then L = K(γ). Observe that this
implies the theorem.

Let f (x) ∈ K[x] be irreducible. We prove that the roots of f in any extension L/K are
distinct. That is, if z ∈ L and (x − z) | f (x), then (x − z)2 does not divide f (x).

Replace L by K adjoined with the roots of f in L, so L/K is finite. Let L′ /L be a splitting

field for f . Then f (x) = (x − α1 ) · · · (x − αn ) in L′ [x]. We want to show that the αi are
distinct. Since f (x) is irreducible, it is the minimal polynomial of each αi over K. Consider
 MATH 123 NOTES 33

f ′ (x) = dx
d
f (x). Then f ′ (α1 ) ̸= 0, and its degree is less than that of f . Note that this step
requires the fact that K has characteristic zero.

We have f ′ (α1 ) = (α1 − α2 ) · · · (α1 − αn ) ̸= 0. Then α1 ̸= αn . Applying this argument over

all αi , we see that they are distinct.

Let f and g be the minimal polynomials of α and β, so f, g ∈ K[x]. Let α1 , . . . , αn and

β1 , . . . , βm be their roots.
β ′ =β
Now let γij = βj + cαi . Let c = αji −αij′ . Choose c such that γij ̸= γi′ j ′ . Let γ = β1 + cα1 ∈ L.
Finally, let h(x) = g(γ − cx) ∈ K(γ)[x]. We have h(α1 ) = g(β1 ) = 0. If i > 1, h(ai =
g(γ − cαi ). If this equals 0, then γ − cαi = βj for some j, but this cannot hold. So h(αi ) ̸= 0.

The gcd of f and h is x − α1 . We have x − α1 ∈ K(γ)[x]. So α1 ∈ K(γ). Then β = γ − cα ∈

K9γ), so K(α, β) = K(γ) and we are done. □
Theorem 17.4. Let L be a field and H ⊂ Aut(L) be a subgroup. Let K = LH . Then
[L : K] = |H|.

18. Friday, April 5

Lemma 10. Let H be a finite group of automorphisms of a field L. If β1 ∈ L such that

{h(β1 ) : h ∈ H} = {β1 , . . . , βr }, then
g(x) = (x − β1 ) · · · (x − βr ) ∈ K[x], K = LH
is the minimal polynomial for β1 .

(Proved last time).

Corollary 18.1. β1 is algebraic over K and deg g = r.
Theorem 18.2 (Fixed Field Theorem). If K = LH , then [L : K] = |H|.

Proof. If β ∈ L, then β is algebraic (i.e. is the root of a polynomial of a polynomial in K[x])

over K and its degree divides |H|. Then,
deg β = deg min polynomial of β over K = [K(β) : K].
If g is the minimal polynomial of β over K, then K(β) ≃ K[x]/g(x), and we have [K(β) :
K] = dim K(β) = deg g. By the primitive element theorem, there exists γ ∈ L such that
L = K(γ). This implies [L : K] | |H|, but this only works if we assume [L : K] is finite. To
show that the extension is finite, we use the following lemma and corollary:
Lemma 11. If K̃/K is an algebraic extension of fields and K̃/K is note finite, then {K(β) :
K] : β ∈ K̃} contains arbitrarily large integers.

Proof of Lemma. Choose α1 ∈ K̃, α1 ̸∈ K. Then K ⊊ K(α1 ). We have K(α1 )/K is finite,
but K̃/K is not finite. Then K̃ ̸= K(α1 ), so there exists α2 ∈ K such that
K ⊊ K(α1 ) ⊆ K(α1 , α2 ) ⊊ . . . ⊊ K(α1 , . . . , αn ).
34 AMY FENG & EMMA CARDWELL & ZIYONG CUI

So, [K(α1 , . . . , αr ) : K] ≥ 2r . 3 Then, by the primitive element theorem, there exists γ ∈ K̃

such that K(α1 , . . . , αr ) = K(γr ), and [K(γr ) : K] = 2r → ∞. □
Corollary 18.3. [L : K] is finite.

Proof of Corollary. If β ∈ L, we saw [K(β) : K] | |H|, so [L : K] is finite. □

Then, let Hγ = {h ∈ H : h(γ) = γ}. The Hγ acts trivially on L. But H ⊆ Aut(L), so

Hγ = {1}. So, |H · γ| = |H/Hγ | = |H|.

So, [L : K] = deg γ = |H · γ| = |H|. □

Remark 18.4. Let L be a field with H ⊆ Aut(L) a finite group. Let K = LH . We know
that

• L/K is called a Galois extension if |Gal(L/K) = [L : K].

• By the fixed field theorem, we have |H| = [L : K], with H ⊆ Gal(L/K).
• In fact, always true that |Gal(L/K)| ≤ [L : K] (this is something we will prove).
Lemma 12. Let L/K be a finite extension, γ ∈ L a primitive element with minimal polyno-
mial f ∈ K[x]. If γ1 , . . . , γr are the roots of f in L, then |Gal(L/K)| = r, and for i = 1, . . . , r,
then there exists unique σi ∈ Gal(L/K) such that σi (r1 ) = ri .

Proof. Let L = K(γ) ≃ K[x]/f (x). If σ ∈ Gal(L/K), then

∼
σ : γ ∈ L = K[x]/f (x) K(x)/f (x) ≃ L

σ̃ (σ̃i (x)=γi )

K[x]
Here, σ̃i (f (x)) = f (σ̃i (x)) = f (γi ) = 0. This implies σ̃i factors as K[x]/(f (x)) → K[x]/f (x).
By definition, σ̃i is surjective. Then, Imσ̃i = K(γi ), so [K(γi : K] = deg γi = deg f . As such,
we observe [K(γ) : K] = [L : K]. □

19. Monday, April 8

Theorem 19.1. Let L be a field, H ⊂ Aut(L) a finite subgroup, K = LH . Then |H| = [L :

LH ]
Lemma 13. Let L/K be a finite extension, γ ∈ L a primitive element, with f a minimal
polynomial for γ. Let {γ1 , . . . , γn } be the roots of f in L. There is some unique σi in
Gal(L/K) where σi (γ1 ) = γi .
Lemma 14. (a) If L/K is finite, Gal(L/K) is finite and |Gal(L/K)| divides [L : K].

(b) If H is a finite subgroup of Aut(L), then L/LH is finite, Galois, Gal(L/LH ) = H.

3Generalfact: if K2 ⊇ K1 ⊇ K is a finite extension of fields, then [K2 : K] = [K2 : K1 ] · [K1 : K], this is
a general fact for vector spaces.
 MATH 123 NOTES 35

Proof. (a) G = Gal(L/K) is finite by the first lemma. G embeds into the group of permuta-
tions on {γ1 , . . . , γn }. By the fixed field theorem, we have |G| = [L : LG ], K ⊂ LG ⊂ L. We
have [L : K] = [L : LG ][LG : K]. Then |G| = [L : LG ] divides [L : K].

(b) We have |H| = [L : LH ] is finite. H ⊂ Gal(L/LH ). By (a) we have |Gal(L/LH )| divides

[L : LH ] = |H|. Then H = Gal(L/LH ) and L/LH is Galois. □
Theorem 19.2. Let L/K be finite, G be the galois group of L/K. The following are
equivalent:

(a) L/K is Galois.

(b) LG = K

(c) L/K is a splitting extension.

Proof. We prove that (a) implies (b), and vice versa. By the Fixed Field Theorem, |G| =
[L : LG ]. So L/K is Galois if and only if |G| = [L : K], it and only if LG = K.

We prove that (a) if and only if (c). Let L = K(γ), where (γ) is a primitive element. Let f
be the minimal polynomial of γ and {γ1 , . . . , γr } be the roots of f in L.

Recall that there is a unique σ ∈ Gal(L/K) such that σi (γ1 ) = γi . Then |G| = r.

We have (a) if and only if |G| = [L : K] only if r = [L : K] only if r is the degree of f , only
if L contains all the roots of f . So then L must be the splitting fields of f . Thus we have
(a) implies (c).

Now we prove that (c) implies (a). Observe that L/K is splitting implies that L contains
the roots of f by the splitting theorem. □
Corollary 19.3. (a) Every finite L/K is contained in a Galois extension.

(b) If K ⊂ K ′ ⊂ L, then L/K is Galois implies that L/K ′ is Galois.x

20. Friday, April 12

Theorem 20.1. Let L/K be a Galois extension. There are inverse bijections
{K ⊆ K ′ ⊆ L} ←→ {subgroups H ⊆ Gal(L/K)}
K ′ −→ Gal(L/K ′ )
LH ←− H.

This was proved last time.

Theorem 20.2. H is normal iff K ′ /K is Galois, in which case Gal(K ′ /K) ≃ Gal(L/K)/H.
Lemma 15. Let L/K be a Galois extension with Gal(L/K) = G. Let g ∈ K[x] be a
polynomial that splits completely in L with roots {β1 , . . . , βn }. Then:
36 AMY FENG & EMMA CARDWELL & ZIYONG CUI

(1) G acts on the set of roots by permutation G −→ Sn .

(2) If L is a splitting field for g, then G ,→ Sn .
(3) If g is irreducible, the action of G on the roots {β1 , . . . , βn } is transitive.
(4) If g is irreducible and L is a splitting field for g, then the action of G on the roots
{β1 , . . . , βn } is transitive and faithful.

Proof of Lemma. We’ve seen the proof of (1) - any element of G is a field automorphism, so
it must send roots to roots. As such, it can be viewed as a permutation on the set of roots.

For (2), then L = K(β1 , . . . , βn ). If σ ∈ ker(G → Sn ), then σ fixes all of the βi ’s, so σ fixes
L. As such, the kernel of G −→ Sn is injective.

For (3), if g is irreducible, the orbit {σ ∈ Gal(L/K) : σ(βi )} has size deg(βi ) = deg g.

Finally, (4) follows from (2) and (3). □

Proof of Theorem 20.2. Let ϵ1 ∈ LH be a primitive element. Let g be the minimal polyno-
mial of ϵ1 in K[x]. The roots of g are ϵ1 , . . . , ϵn ∈ L. We know that LH /K is Galois iff LH /K
is splitting. By the splitting theorem, we also know that LH /K is a splitting extension iff
ϵ1 , . . . , ϵn ∈ LH .

Then, we note that LH = K[ϵ1 ), so [LH : K] = [K(ϵ1 ) : K] = deg g. Similarly, since g is

irreducible, we have [K(ϵi ) : K] = deg g for each i, so LH = K(ϵi ). This implies
Hi = {σ ∈ Gal(L/K) : σ(ϵi ) = ϵi } = H.

For any i, Hi = τi Hτi−1 , where τi ∈ Gal(L/K) such that τi (ϵi ) = ϵi . Then, we observe the
following chain of iff statements:
σ ∈ Hi ⇐⇒ σ(ϵi ) = ϵi ⇐⇒ στi (ϵ1 ) = τi (ϵ1 ) ⇐⇒ τi−1 στi ∈ H ⇐⇒ τi−1 στi (ϵ1 ) = ϵ1 .
Then if ϵ1 . . . , ϵn ∈ LH , then H = τi Hτi−1 for all τi such that τi (ϵ1 ) = ϵi . This shows that H
is a normal subgroup.

If H is normal, then we have Hi = τi Hτi−1 = H, so H fixes ϵi for all i. This implies ϵi ∈ LH

for all i, so LH /K is Galois. □
Corollary 20.3. If LH /K is Galois, then Gal(LH /K) ≃ Gal(L/K)/H.

Proof. For σ ∈ Gal(L/K), σ|LH is an automorphism of LH as σ(ϵi ) = ϵj , and LH = K(ϵi )

for any i. This shows that σ(LH ) = LH . This implies Gal(L/K) → Gal(LH /K) is given by
σ 7→ σ|LH .

Then, we observe
σ|LH = 1 ⇐⇒ σ fixes LH ⇐⇒ σ ∈ H.

This implies H is the kernel of the map Gal(L/K) → Gal(LH /K), so we get an injective
[L:k]
mapping Gal(L/K)/H ,→ Gal(LH /K). This then also implies [L:LH ] = [L
H
: K]. □
 MATH 123 NOTES 37

For an irreducible polynomial f (x) ∈ K[x] with deg f = N , let L/K be its splitting field.
Then, applying the lemma, we get an injective map Gal(L/K) ,→ Sn = {α1 , . . . , αN }.
Question: can we describe the image?

• If N = 2, then Gal(L/K) = S2 .
• If N = 3, then |S3 | = 6, but |Gal(L/K)| ≥ 3. We also know that [L : K] | 6. As
such, there are two possibilities: either Gal(L/K) ≃ Z/3Z ⊆ S6 = Z/3Z ⋊ Z/2Z, or
Gal(L/K) ≃ S3 .

Consider a polynomial
F (x) = (x − u1 ) · · · (x − un ) = xn − s1 xn−1 + s2 xn−2 . . . + (−1)n Sn .

Define D(u) = i<j (ui − uj )2 = (−1)i i̸=j (ui − uj ). For any field K and irreducible
Q Q

polynomial f ∈ K[x],
f (x) = xn − a1 xn−1 + a2 xn−2 + . . . + (−1)n an ∈ K[x].

Then, define D(f ) = D(a1 , . . . , an ). For example:

• n = 2, then D(u) = (u1 − u2 )2 = s21 − 4s2 .

• n = 3, then D(u) = −4s31 s3 + s21 s2 + 18s1 s2 − 4s32 − 27s23 . If a1 = 0, we get D(f ) =
−4a32 − 27a23 .
Theorem 20.4. Let K be a field of characteristic 0, and let f ∈ K[x] be an irreducible, cubic
polynomial with splitting field L. Then, D(f ) ∈ (K × )2 = {a2 : a ∈ K × } ⇐⇒ Gal(L/K) ≃
Z/3Z.

For example, if f (x) = x3 − 2, then D(f ) = −27 · 22 ̸∈ (Q× )2 .

21. Monday, April 15

Theorem 21.1. Let k be a field of char 0, f ∈ K[x] irreducible where L is the splitting field
of f . Then D ∈ (K × )2 ⇐⇒ Gal(L/K) = Z/3Z. D ∈ / (K × )2 ⇐⇒ Gal(L/K) = S 3 .
Theorem 21.2. Gal(Q(ζn )/Q) ∼ (Z/nZ)× .

Proof. We prove this when N = p is prime. We want to show that the minimal polynomial of
p −1
ζp has degree p−1. That is, xx−1 is irreducible. Recall Eisenstein’s criterion: Let f (x) ∈ Z[x],
where f is monic. If there is a prime p such that p | ai for 0 ≤ i ≤ n − 1 but p2 ∤ a0 , then f
is irreducible.

Now let x = y + 1, so
xp − 1 (y + 1)p − 1 y p + py p−1 + · · · + py + 1 − y
= = = y p−1 + · · · + p
x−1 y+1−1 y
which satisfies Eisenstein’s criterion.
38 AMY FENG & EMMA CARDWELL & ZIYONG CUI

Now we prove Eisenstein’s criterion. By Gauss’ lemma if f is not irreducible, f (x) = g(x)h(x)
where g, h ∈ Z[x]. Reducing this mod p, we get f¯(x) = ḡ(x)h̄(x) ∈ Z/pZ[x]. But by
assumption, all coefficients of f are divisible by p except for the leading coefficient, so
f¯(x) = xn = xi xj for some i and j. Now we have ḡ(0) = f¯(0) = 0. But then p | g(0) and
p | h(0) so p2 ∤ f (0), which is a contradiction. □

Question. When is a regular n-gon constructible using ruler and compass?

Note that the numbers 1, ζn , ζn2 , . . . , ζnn−1 form a regular n-gon. So we also want to know
when n is constructible.

If this is possible, there is a sequence of fields Q ⊂ K1 ⊂ K2 ⊂ · · · ⊂ Km = Q(ζn ) where

[Ki+1 : K] = 2.

When n = p is prime, then 2m = p − 1, p = m2 + n1 so m = 2n .

22. 04/19/2024

Question: when does a polynomial of degree n have a solution in radicals?

Definition 22.1. A finite group G is solvable if there exists a sequence {0} ⊆ H1 ◁H2 ◁. . .◁G
such that Hi+1 /Hi is an abelian group.
Theorem 22.2. Let K be a field, f ∈ K[x] an irreducible polynomial, and L(f ) the splitting
field of f . Then Gal(L(f )/K) is solvable iff f has a solution in radicals.
Definition 22.3. A polynomial f ∈ K[x] has a solution in the radicals if there exists a
√
tower K ⊂ K1 ⊂ . . . ⊂ Kr ⊂ L ⊂ L(f ) such that Ki+1 = Ki ( n αi ) for αi ∈ Ki , and L
contains a root of f .
Lemma 16. Suppose ζp ∈ K where ζp is a√primitive p-th root of unity. If√β ∈ K and β p ∈ K,
then xp − β ∈ K[x] is irreducible, and K(√p β)/K is Galois with Gal(K( p β)/K) ≃ Z/pZ. If
L/K is Galois of degree p, then L = K( p β) for some β ∈ K.
√ √ √ √ √
Proof. The roots of X p − β are p β, ζp p β, ζp2 p β, . . . , ζpp−1 p β. This implies K( p β) is a
√
splitting field for X p − β = 0, so K(
√
p
β)/K is Galois. Any element in the Galois√group that
gets sent to the identity must fix p β, so we have an injective mapping Gal(K( p β)/K) ,→
Z/pZ.

For the second part, by the Galois correspondence, if L/K is Galois, then Gal(L/K) ≃ Z/pZ.
Consider an automorphism of L as a K vector space. Since the extension has finite order,
there is some K-basis of L, and we can represent any generator σ ∈ Gal(L/K) as
 n1 
ζp
 ... 
 
nj
σ∼ ζp
 


 . ..


ns
ζp
 MATH 123 NOTES 39

This implies there exists an eigenvector α ∈ L, i.e. (σ − ζpn )(α) = 0 for some ζpn . So, σ(x) =
ζpn − α, and we have σ(αp ) = αp , so αp = β ∈ K. Then since 1 ̸= [K(α) : K] | [L : K] = p,
this shows [K(α) : K] = [L : K], so we have L = K(α). □

Notation: suppose K ⊆ K̃ with K̃ algebraically closed. If K1 , K2 ⊆ K̃ such that K1 /K, K2 /K

are finite extensions, we have K1 · K2 ⊆ K̃ for the subfield generated by K1 , K2 .
Lemma 17. Let L/K be Galois, K ′ /K any finite extension.

L · K′

K′ L

Then L · K ′ /K ′ is Galois and Gal(L · K ′ /K ′ ) ⊆ Gal(L/K).

If K ′ /K is Galois, then L · K ′ /K is also Galois and we get an injective map ϕ : Gal(L ·

K ′ /K) ,→ Gal(L/K) × Gal(K ′ /K).

Proof. If L is a splitting field for f ∈ K[x], then L · K ′ is a splitting field for f ∈ K ′ [x], so
L · K ′ /K ′ is Galois. We want to show that σ ∈ Gal(L · K ′ /K ′ ) preserves L. It suffices to
show that σ permutes the roots of f , but this already follows from the fact that L · K ′ /K ′
is a splitting field for f . Then, if ϕ(σ) = 1, then σ|L = id, but σ|K ′ = id, so σ fixes K ′ and
L, so σ fixes L · K ′ .

If K ′ /K is Galois, then K ′ is the splitting field of g for some g ∈ K[x]. The composite
L · K ′ /K is splitting field for f · g. □
Theorem (Theorem 22.2). Let f ∈ K[x] be irreducible, with splitting field L(f ). Then
Gal(L(f )/K) is solvable iff f is solvable in radicals.

Proof. Consider sequences of finite extensions

(2) K = K0 ⊂ K 1 ⊂ . . . ⊂ Kr = L
with L containing a root of f , Ki+1 /Ki finite. Then
√
(1) There exists (2) with Ki = Ki−1 ( ni βi ) for βi ∈ Ki−1
(2) There exists (2) with Ki /Ki−1 Galois and of prime order
(3) There exists (2) with Ki /Ki−1 Galois abelian
(4) There exists (2) with Ki /Ki−1 Galois abelian and L/K Galois.

The theorem says (1) ⇐⇒ (4), but we can prove all of these statements.

We’ve discussed (2) =⇒ (3) before.

40 AMY FENG & EMMA CARDWELL & ZIYONG CUI

For (3) =⇒ (2), we note that if (2) as in (3), then K = K0 ⊂ K1 ⊂ . . . ⊂ Kr = L,

and
Q Kin/K i−1 is Galois abelian, Gal(Ki /Ki−1 ) = Gi is a finite abelian group. Then Gi =
j
(Z/pj Z) for some prime pj and nj ∈ Z>0 . Then, we can construct a filtration by subgroups
{0} ⊂ G1i ⊂ G2i ⊂ . . . ⊂ Gki = Gi such that Gti /Gt−1
i ≃ Z/pj Z for some j. By the Galois
correspondence, this gives us a sequence Ki−1 ⊂ Kit ⊂ . . . ⊂ Ki2 ⊂ Ki1 ⊂ Ki . This shows
(3) and (2) are equivalent. Next time: prove (1) and (2) are equivalent and (3) and (4) are
equivalent. □

23. 04/22/2024

We finish the proof of 23.3 from last time.

Proof. We prove that (2) =⇒ (1). Let m = n1 , . . . , nr , ni = [Ki : Ki−1 ].

√
If ζm ∈ K0 then Ki = Ki−1 ( ni βi )√for some βi ∈ Ki−1 , by a lemma from last time. Let
′
Ki′ = Ki (ζm ). Then Ki′ = Ki−1 ( ni βi ). We have K0 ⊂ K0′ ⊂ K1′ ⊂ · · · ⊂ Kr′ , and this
sequence satisfies (1).

We prove that (1) =⇒ (3). We may assume that all ni are prime. We know that Ki /Ki−1
is Galois abelian if ζni ∈ Ki−1 . Let Ki′ = Ki (ζm ) with m defined as above. Then we have
′
K0 ⊂ K0′ is Galois abelian. We saw last time that Gal(Ki′ /Ki−1 ) ⊂ Gal(Ki /Ki−1 ) for i ≥ 1,
′ ′ ′ ′ ′
so Ki /Ki−1 is Galois abelian, and so K0 ⊂ K0 ⊂ K1 ⊂ · · · ⊂ Kr , satisfies (2)

We have shows (1), (2) and (3) are equivalent, and obviously (4) =⇒ (3). We prove that
(2) =⇒ (4). Let K0 ⊂ K1 ⊂ · · · ⊂ Kr−1 ⊂ Kr = L where nr is prime. We’ll show by
induction on r, that there is some K = K0′ ⊂ K1′ ⊂ · · · ⊂ Kr′ such that Kr′ /K is solvable
′
and Galois, and Ki′ /Ki−1 is abelian. When r = 0, there is nothing to show.

Let γ ∈ Kr−1 be a primitive element, so that Kr−1 = K(γ). By induction we may assume
there exists K0 = K0′ ⊂ K1′ ⊂ · · · ⊂ Kr−1
′ ′
, with γ ∈ Kr−1 ′
, Kr−1 /K Galois, solvable, and
′ ′ ′
Ki /Ki−1 is abelian. We may also assume ζm ∈ K1 , as above.

Let Kr′′ = Kr · Kr−1 ′

. By a lemma from last time, Kr′′ /Kr−1 ′
is Galois, and Gal(Kr′′ /Kr−1
′
)⊂
Gal(Kr /Kr−1 ). By our assumptions, the latter group has prime degree nr , and so either √
[Kr′′ : Kr−1
′
] = 1 or nr . In the former case we are done. In the latter case Kr′′ = Kr−1 ′
( nr β)
′ ′
for some β in Kr−1 . Let g ∈ K[X] be the minimal polynomial of β. As Kr−1 /K is Galois,
′
Kr−1 contains all the roots of g, by the splitting theorem. Call these β1 , . . . , βs . Now let
Kr′ = Kr−1
′ ′
p p
(ζni r nr βj )i,j = Kr−1 ( nr βj )j ,
so that Kr′ is the splitting field of h(X)g(X p ) ∈ K[X], where h ∈ K[X] is the minimal
polynomial of the primitive element γ above. (Alternatively, Kr′ is the composite of Kr−1 ′
,
p
and the splitting field of g(X ) hence is Galois).

Then Kr′ /K is Galois, and Kr′ /Kr−1′ ′ ′

p
is abelian, as Gal(Kr−1 ( nr βj )/Kr−1 ) is abelian, and
′ ′
the product of these groups over j contains Gal(Kr /Kr−1 ). Thus the tower
K0 = K0′ ⊂ K1′ ⊂ · · · ⊂ Kr′ ,
 MATH 123 NOTES 41

satisfies (4).

Now to prove the theorem, note that (1) is just a rewording of the condition that f is solvable
in radicals, so it is enough to show that (4) holds if and only if Gal(L(f )/K) is solvable.
If Gal(L(f )/K) is solvable, then a tower as in (4) exists with L = L(f ). The fields Ki
can be taken to be the fixed fields of a sequence of subgroups witnessing the solvability of
Gal(L(f )/K). If (4) holds, then L contains all the roots of f, as L/K is Galois, and hence
L(f ) ⊂ L. Hence Gal(L(f )/K) is a quotient of the solvable group Gal(L/K), and is solvable.