MATH 221 NOTES: PART 1

YI SUN

Contents
Introduction 1
1. The Tensor Product 1
1.1. Existence and Uniqueness for Modules 1
1.2. Flat Modules 5
1.3. Projective Modules 5
1.4. The Commutative Case 6
1.5. For Algebras 7
2. Finite Generation 8
2.1. Finitely Generated Modules 8
2.2. Finitely Generated Algebras 11
3. The Nullstellensatz 11

Introduction
These are notes from the course Math 221: Commutative Algebra taught by Professor Dennis
Gaitsgory at Harvard University in Fall 2008.

1. The Tensor Product

1.1. Existence and Uniqueness for Modules. We wish to construct the tensor product for
modules over an arbitrary ring. In this section, let R be a (non necessarily commutative) ring, M
a right R-module, and N a left R-module.
Theorem 1.1. There exists an abelian group M ⊗ N with an associated map φuniv : M × N →
R
M ⊗ N such that for any abelian group A and any map φ : M × N → A of abelian groups, there
R
exists a unique map ψ : M ⊗ N → A such that φ = ψ ◦ φuniv .
R

The group M ⊗ N is referred to as the tensor product of M and N , and the property is referred
R
to as the universal property of tensor products. We call elements in the image of φuniv pure tensors
and we denote m ⊗ n := φuniv (m, n). The tensor product also satisfies the following naturality
property.
Proposition 1.2. Let T : M1 → M2 be a map of right R-modules and N a left R-module. Assume
that M1 ⊗ M2 and M2 ⊗ N exist. Then, there exists a unique map T ⊗ N : M1 ⊗ N → M2 ⊗ N
R R R R R
such that φ2,univ = (T ⊗ N ) ◦ φ1,univ .
R

Proof. Apply the universal property of M1 ⊗ N to the map M1 × N → M2 ⊗ N given by (m, n) 7→

R R
φ2,univ (T (m), n). 
1
2 YI SUN

Example. Suppose that M = R. Then, for any N , take R ⊗ N := N and φuniv : R × N → N

R
given by (r, n) 7→ r · n.
Lemma 1.3. The pair (N, φuniv ) as defined above satisfies the universal property of the tensor
product, meaning that N = R ⊗ N .
R

Proof. Given φ : R × N → A, we can take ψ(n) = φ(1, n). 

Before we prove Theorem 1.1, we recall some preliminaries about modules.

Lfor each i ∈ I, let Mi be a right R-module. Then, for any right

Definition. Let I be a set, and
R-module M , their direct sum i∈I Mi satisfies the universal property
!
M Y
HomR Mi , M = HomR (Mi , M ).
i∈I i∈I

Q each i ∈ I, let Mi be a right R-module. Then, for any right

Definition. Let I be a set, and for
R-module M , their direct product i∈I Mi satisfies the universal property
!
Y Y
HomR M, Mi = HomR (M, Mi ).
i∈I i∈I

Recall that the direct sum is realized as the set of assignments i 7→ mi ∈ Mi with mi = 0 for all
but finitely many i, and the direct product is realized as the set of all assignments i 7→ mi ∈ Mi .
Recall also that the direct sum and direct product are isomorphic for finite sets I.
We will now show that the tensor product is nicely behaved under some common operations.
L

Lemma 1.4. If Mi ⊗ N exists for all i ∈ I, then i∈I Mi ⊗ N also exists.
R R

Proof. Let φiuniv be the universal map for each Mi ⊗ N and let M =
L
i∈I Mi . We will claim that
R
!
M M
Mi ⊗ N := (Mi ⊗ N )
R R
i∈I i∈I
L
satisfies the universal property with the universal map φuniv : M × N → i∈I (Mi ⊗ N ) given by
R

φuniv (⊕ mi , n) 7→ ⊕ φiuniv (mi , n).

We can easily verify that the universal property holds for this map. 
From Lemmas 1.3 and 1.4, then, we see that
!
M M
R ⊗N = N.
R
I I
We now construct the tensor product for cokernels via the following lemma:
T
Lemma 1.5. Let M1 → M2 → M2 → M3 → 0 be short exact, and suppose that M1 ⊗ N and
R
M2 ⊗ N exist. Then, M3 ⊗ N also exist and is isomorphic to coker(M1 ⊗ N → M2 ⊗ N ).
R R R R

Proof. Take φ3,univ : M3 × N → coker(M1 ⊗ N → M2 ⊗ N ) as follows: Given (m3 , n) ∈ M3 × N , let

R R
m2 be a preimage of m3 in M2 . Consider φ2,univ (m2 , n) ∈ M2 ⊗ N and project it to coker(M1 ⊗ N →
R R
M2 ⊗ N ).
R
 MATH 221 NOTES: PART 1 3

We first claim that φ3,univ is well-defined. It suffices to check that φ2,univ (T (m1 ), n) ∈ Im(T ⊗ N ).
R
But indeed φ2,univ (T (m1 ), n) = (T ⊗ N )(φ1,univ (m1 , n)).
R
We now check the universal property. Take any bilinear map φ : M3 × N → A. Observe that
a map ψ : coker(M1 ⊗ N → M2 ⊗ N ) → A is equivalent to a map ψ2 : M2 ⊗ N → A such that
R R R
ψ2 ◦ (T ⊗ N ) = 0. Now, consider the following diagram:
R

T × id π × id φ
M1 × N > M2 × N > M3 × N >
>
A
∩

φ1,univ φ2,univ
ψ2
∨ T ⊗N ∨
R
M1 ⊗ N > M2 ⊗ N
R R

Observe that φ ◦ (π × id) is a bilinear map M2 × N → A, so it lifts to a map ψ2 : M2 ⊗ N → A.

R
Now, note that

ψ2 ◦ (T ⊗ N ) ◦ φ1,univ = ψ2 ◦ φ2,univ ◦ (T × id) = φ ◦ (π ◦ id) ◦ (T × id) = 0,

R

which means that ψ2 ◦ (T ⊗ N ) is induced by the zero map M1 × N → A, so it must itself be zero.
R
Observe now that the ψ corresponding to this ψ2 satisfies the desired property. 

Finally, we wish to establish that all modules can be written as cokernels via the following lemma:

Lemma 1.6. Every module is the quotient of some free module.

L
Proof. For a module M over a ring R, consider the free module N = M R and the map N  M
that sends the mth coordinate rm to m. This is surjective, giving the desired representation. 

Corollary 1.7. Every module can be written as the cokernel of a map of free modules.
0
L
Proof. Given a module
L M , by Lemma 1.6, choose I giving a surjection I R  M . Let M be the
0
L
kernel of the map I R  M , and choose J giving a surjection J R  M . Then, we have the
exact sequence
M M
R  M 0 ,→ R  M 0 → 0,
J I

M0
L L
meaning that is the cokernel of the map J R→ I R. 

We are now ready to put everything together to prove the existence of the tensor product.

Proof of Theorem 1.1. First, by Lemmas 1.3 and 1.4, we know that the tensor product M ⊗ N
L L R
exists for M a free module. Now, if M is not
L a free module, write
L M = coker( J R → I R) for
some I and J by Corollary 1.7. Because ( J R) ⊗ N and ( I R) ⊗ N exist, we have by Lemma
R R
1.5 that M ⊗ N exists, establishing the theorem. 
R

Corollary 1.8. If M1  M2 is a surjection, then M1 ⊗ N → M2 ⊗ N is a surjection.

R R
4 YI SUN

Proof. Recall that a map T : A1 → A2 is a surjection in the categorical sense, if for every object A
and map S : A2 → A, then S ◦ T = 0 implies that S = 0. Now, consider the following diagram:
T ×N
M1 × N > M2 × N

φ2
∨ T ⊗N ∨

>
R
M1 ⊗ N > M2 ⊗ N > A
R R ψ2
Suppose that T : M1  M2 is a surjection, so T × N is a surjection. Now, take any A and ψ2 :
M2 ⊗ N → A. Then, we see that ψ2 ◦(T ⊗ N ) = 0 ⇐⇒ φ2 ◦(T ×N ) = 0 ⇐⇒ φ2 = 0 ⇐⇒ ψ2 = 0,
R R
as needed. 
Having established the existence of the tensor product, we now wish to consider its uniqueness,
given by the following lemma.
Lemma 1.9. If (M ⊗ N )0 and M ⊗ N satisfy the universal property of M ⊗ N , then they are
R R R
canonically isomorphic.
Proof. Consider the following diagram:
φ0univ
M ×N > (M ⊗ N )0
R
∧

φuniv
> ∨
M ⊗N
R

By the universal property, we obtain maps ψ0 : (M ⊗ N )0 → M ⊗ N and ψ : M ⊗ N → (M ⊗ N )0

R R R R
such that ψ 0 ◦ φ0univ = φuniv and ψ ◦ φuniv = φ0univ . This means that (ψ 0 ◦ ψ) ◦ φuniv = φuniv . Observe
that ψ 0 ◦ ψ ◦ φuniv : M × N → M ⊗ N is a bilinear map which factors through the identity map
R
M ⊗ N → M ⊗ N . By the universal property, this factorization is unique, so ψ 0 ◦ ψ = id. Similarly,
R R
ψ ◦ ψ 0 = id, as desired. 
Note Lthat to Lcompute the tensor product
L in practice,
L one should find I, J small so that M =
coker( J R → I R). Now, note that if J R → I R → M → 0 is exact, then tensoring with
the right exact functor − ⊗ N gives the exact sequence
R
M M
N→ N → M ⊗ N → 0.
R
J I

We now wish to discuss the relation between the tensor product and various homomorphisms of
rings. Given a ring homomorphism φ : A → B, if M is a B-module, we can view it as an A-module
with action induced by φ. Note that we can view B ⊗ N as a B-module via the natural action of
A
B on itself.
Lemma 1.10. If N is an A-module, B ⊗ N has the structure of B-module such that for every
A
B-module M , we have a natural isomorphism HomA (N, M ) ' HomB (B ⊗ N, M ).
A
 MATH 221 NOTES: PART 1 5

Proof. For any map B ⊗ N → M , take the corresponding φ : B ×N → M and map it to φ0 : N → M

A
given by φ0 (n) = φ(1, n). Map ψ : N → M to ψ 0 : B ⊗ N → M given by ψ 0 (b ⊗ n) = b · ψ(n). It’s
A
easy to check that these maps define the desired isomorphism. 
Lemma 1.11. Let R1 , R2 be rings, L a right R1 -module, M a (R1 , R2 )-bimodule, and N a left
R2 -module. Then, M ⊗ N is a left R1 -module.
R2

Proof. We need to construct a map R1 → EndZ (M ⊗ N ). We have a map R1 → EndR2 (M ) because

R2
M is a left R1 -module. The desired action then follows from the functoriality of − ⊗ N . 
R2

Note further that L ⊗ (M ⊗ N ) ' (L ⊗ M ) ⊗ N . Observe also that we can obtain another
R1 R2 R1 R2
proof of Lemma 1.10 by viewing A → B as a map of right A-modules and using the functoriality
of − ⊗ N .
A

1.2. Flat Modules. We now investigate the exactness of the tensor product more closely. Let
M1 ,→ M2 be an injection of right R-modules.
Claim. M1 ⊗ N → M2 ⊗ N need not always be injective.
R R
2
Proof. Take R = M1 = M2 = Z with the map M1 ,→ M2 given by multiplication. Now let
N = Z/2Z. In this case, we see that M1 ⊗ N = M2 ⊗ N = Z/2Z, where the map M1 ⊗ N → M2 ⊗ N
R R R R
is given by multiplication by 2. Hence, it is the zero map, which is not injective. 
Definition. A left R-module N is called flat if the functor − ⊗ N is left exact (i.e., it transforms
R
injections to injections).
Example. All free modules are flat because tensor products distribute over direct sums.
1.3. Projective Modules. Now, consider the following situation. Let φ : N1  N2 be a surjection
of modules and ψ : N → N2 a map. We wish to determine when there exists a lift ψ̃ : N → N1 so
that φ ◦ ψ̃ = ψ.
φ
N1 >> N2
.<...

>
....
....
....
....

ψ̃ ψ
....
....
....

N
Observe that such a lift always exists when we are working over a field (i.e., N, N1 , N2 are vector
spaces). However, for R = Z, N1 = Z, and N2 = N = Z/2Z, it does not.
Definition. A module N is called projective if, for any surjection N1  N2 , any map N → N2 can
be lifted to a map N → N1 .
Example. (1) N = R is projective L
(2) If {Ni } are projective, then so is i∈I Ni .
Lemma 1.12. Let P be projective, and let P1 be a direct summand of P (i.e. we can write
P = P1 ⊕ P2 ). Then, P1 is projective.
Proof. Take a surjection N1  N2 and suppose we have a map φ : P1 → N2 . Then, φ extends to a
map φ̃ : P → N2 by sending P2 to zero. We can then finish the proof by using the lifting property
of P . 
6 YI SUN

Proposition 1.13. Any projective module is a direct summand of a free module.

L
Proof. Recall that
L we can take a surjection φ : I R  P for some I. Then, lift the identity map
id : P → P to I R. Thus,
L φ is a surjection with a left inverse, meaning exactly that its image P
is a direct summand of I R. 
Corollary 1.14. Any projective module is flat.
Corollary 1.15. Any vector space is free, hence projective and flat.
1.4. The Commutative Case. From this point onwards, we take our rings to be commutative.
Let R be a commutative ring and recall that every R-module M is equipped with an R-action
act : R × M → M such that act(r1 , act(r2 , m)) = act(r1 · r2 , m). If R is commutative, then every
left or right module is automatically a bi-module under the identification r · m = m · r. From now
on, we will simply refer to such modules as R-modules.
Remark. It is not true that every R-bimodule comes from such a construction. Observe that
R ⊗ R is a R-bimodule, but evidently the actions on the left and right are distinct.
Z

Now, let M, N be R-modules; then, M ⊗ N acquires a priori two R-module structures, as we

R
can either view it as a left R-module or a right R-module.
Lemma 1.16. These two structures coincide.
Proof. It suffices to show that the left and right actions of R on M ⊗ N coincide. This is a
R
consequence of the following equalities: r · (m ⊗ n) = (r · m) ⊗ n = (m · r) ⊗ n = m ⊗ (r · n) =
m ⊗ (n · r) = (m ⊗ n) · r. 
We now construct the tensor product over a commutative ring in a functorial manner. Given a
map φ̃ : M × N → L for R-modules M, N, L, we say that φ̃ is bilinear if it satisfies the conditions:
φ̃(m1 + m2 , n) = φ̃(m1 , n) + φ̃(m2 , n)
φ̃(m, n1 + n2 ) = φ̃(m, n1 ) + φ̃(m, n2 )
φ̃(rm, n) = rφ̃(m, n) = φ̃(m, rn).
We can now define the tensor product of commutative rings.

Definition. Let R be a commutative ring. For R-modules M and N , the R-module M

^ ⊗ N is a
R
R-module equipped with a bilinear map φ̃univ : M ×N → M
^ ⊗ N satisfying the following universal
R
property: Any bilinear map of R-modules φ̃ : M × N → L factors through a map of R-modules
ψ̃ : M
^ ⊗ N → L such that the following diagram commutes:
R

φ̃
M ×N > L
∧

ψ̃
φ̃univ
>

M
^ ⊗N
R

Theorem 1.17. For any R-modules M and N , the tensor product M

^ ⊗ N exists.
R
 MATH 221 NOTES: PART 1 7

Proof. View M as a right R-module and N as a left R-module. Take M

^ ⊗ N := M ⊗ N with
R R
φ̃univ = φuniv : M × N → M ⊗ N . Now, consider any bilinear map φ̃ : M × N → L. By the
R
universal property of the M ⊗ N , we have a map of a abelian groups ψ : M ⊗ N → L such that
R R
φ̃ = ψ ◦ φ̃univ . It remains only to check that ψ is actually a map of R-modules, which follows from
the fact that ψ(r · (m ⊗ n)) = r · ψ(m ⊗ n). 

Now, let M1 , M2 , . . . , Mn be a finite collection of R-modules We want to construct the R-module

M1 ⊗ M2 ⊗ · · · ⊗ M n .
R R R

Given a map φ : M1 × M2 × · · · ⊗ Mn → L, we say that φ is multilinear if it is

• Additive in each argument
• φ(rm1 , m2 , . . . , mn ) = φ(m1 , rm2 , . . . , mn ) = · · · = φ(m1 , . . . , mn−1 , rmn ) = r·φ(m1 , . . . , mn )
Definition. For R-modules M1 , . . . , Mn , the tensor product M1 ⊗ M2 ⊗ · · · ⊗ Mn is a R-module
R R R
equipped with a multilinear map φuniv : M1 × · · · × Mn → M1 ⊗ · · · ⊗ Mn satisfying the following
R R
universal property: For any R-module L and any multilinear map φ : M1 × · · · × Mn → L, there
exists a map ψ : M1 ⊗ · · · ⊗ Mn → L of R-modules such that ψ ◦ φuniv = φ.
R R

We wish to show that this tensor product also exists for all n. Here we will just demonstrate the
proof in the case where n = 3.
Proposition 1.18. The tensor product M1 ⊗ M2 ⊗ M3 is given by M1 ⊗ (M2 ⊗ M3 ) with the
R R R R
universal map φuniv : M1 × M2 × M3 → M1 ⊗ (M2 ⊗ M3 ) given by (m1 , m2 , m3 ) 7→ m1 ⊗ (m2 ⊗ m3 ).
R R

Proof. It suffices to show that for any trilinear φ : M1 × M2 × M3 → N in the following diagram,
we have a factorization ψ.
φ
M1 × M2 × M3 >N
>

φuniv
ψ
∨
M1 ⊗ (M2 ⊗ M3 )
R R

Taking ψ = m1 ⊗ (m2 ⊗ m3 ) 7→ φ(m1 , m2 , m3 ) finishes the proof. 

Note finally that if R is commutative, we have an isomorphism M1 ⊗ M2 ' M2 ⊗ M1 given by

R R
exchanging the order of factors.

1.5. For Algebras. Let A be a commutative ring.

Definition. An A-module is a ring R with an additional structure of an A-module such that
(r1 a) · r2 = r1 (a · r2 ).
Claim. For an A-module R, a structure of an A-algebra on R is equivalent to a map of R-modules
m
R⊗RR
A
8 YI SUN

such that the following diagram commutes

idR ⊗m
R⊗R⊗R > R⊗R
A A A

m ⊗ idR m
∨ m ∨
R⊗R > R.
A

Example. Let φ : A → R be a map of rings such that φ(A) is central. Then, R is naturally an
A-algebra under a · r = φ(a) · r.
We now wish to define the notion of the tensor product of A-algebras. This will be the coproduct
in the category of A-algebras. Let R1 , R2 be A-algebras.
Definition. In the unital case, the coproduct of R1 and R2 is another A-algebra Coprod(R1 , R2 )
equipped with two maps φ1univ : R1 → Coprod(R1 , R2 ) and φ2univ : R2 → Coprod(R1 , R2 ) such that,
for any A-algebra R equipped with maps p1 : R1 → R and p2 : R2 → R whose images commute,
we obtain a unique map φ : Coprod(R1 , R2 ) → R such that φ ◦ φ1univ = p1 and φ ◦ φ2univ = p2 .
In other words, for maps p1 , p2 with commuting images in the following diagram, there is a
unique factoring map φ:
φ1univ φ2
R1 > Coprod(R1 , R2 ) < univ R2

φ
p1 p2
> ∨
<
R
Definition. In the non-unital case, we can give R1 ⊗ R2 the structure of an A-algebra via
A
(R1 ⊗ R2 ) ⊗ (R1 ⊗ R2 ) → (R1 ⊗ R1 ) ⊗ (R2 ⊗ R2 ) → R1 ⊗ R2 .
A A A A A A A
In other words, the multiplication is given by (r10 ⊗ r20 ) · (r100 ⊗ r200 ) = (r10 · 00
r1 ) ⊗ (r20 · r200 ).

2. Finite Generation
2.1. Finitely Generated Modules. Let M be a left R-module.
Definition. We say that M is finitely generated if there exists a surjection R⊕n  M .
Lemma 2.1. Let 0 → M1 → M2 → M3 → 0 be a short exact sequence. If M1 , M3 are finitely
generated, then so is M2 .
Proof. Take surjections Rn1  M1 and Rn3  M3 . Because Rn3 is free and hence projective, we can
lift to a map Rn3 → M2 . Now, this map, together with the composition of maps Rn1 → M1 → M2 ,
determines a map Rn1 ⊕ Rn3 → M2 . We wish to show that this is surjective. Observe that its
image contains the image of M1 in M2 ; further, it projects to all of M3 , so it must be all of M2 ,
giving the desired surjectivity. 
Definition. We say that R is left Noetherian if every left ideal I ⊂ R is finitely generated. We say
that M is left Noetherian if every submodule of M is finitely generated.
Theorem 2.2. The following conditions are equivalent:
(a) R is left Noetherian.
 MATH 221 NOTES: PART 1 9

(b) Any finitely generated R-module is Noetherian.

Proof. (b) =⇒ (a). This follows because an ideal I ⊂ R is a submodule of R.
(a) =⇒ (b). We first claim that it is enough to consider M = Rn , since every finitely generated
module is a quotient of Rn for some n.
We induct on n. The base case n = 1 is provided by (2). Now suppose the claim is shown for
m < n and consider some submodule M ⊂ Rn . Then, consider the following diagram
0 > Rn−1 > Rn > R > 0
∧ ∧ ∧

∪ ∪ ∪
n−1
0 M2 = R
> ∩M >M > M1 >0

where M1 = M/(M ∩ Rn−1 ). Note that M2 is finitely generated by the inductive hypothesis, and
M1 is an ideal of R, hence finitely generated. Now by Lemma 2.1, M is finitely generated, finishing
the induction. 
Theorem 2.3. The following conditions are equivalent:
(a) M is Noetherian.
(b) For any sequence M1 ⊂ M2 ⊂ · · · of submodules of M , there exists an N such that
Mi = MN for all i > N .
S
Proof. (a) =⇒ (b). Notice that L = Si Mi is also a submodule of M . By (a), we may choose a
finite set of generators {li } for L. Since i Mi = L, we must have that li ∈ Mji for some ji . Letting
j = max{ji }, we see that Mj = L, so we may take N = j to get the conclusion.
(b) =⇒ (a). Suppose for contradiction that we had some submodule L of M that was not finitely
generated. Now, construct a sequence of elements {li } of L as follows. Choose l1 arbitrarily and
take li+1 ∈ L − span{l1 , . . . , li }; note that the last set is non-empty by our assumption that L was
not finitely generated. Now, take Li = span{l1 , . . . , li } and notice that we have L1 ⊂ L2 ⊂ · · · .
By condition (b), we may find some N such that Li = LN for i > N , which is a contradiction,
because lN +1 ∈ / LN by our choice of {li }. Hence L must be finitely generated, so M is Noetherian
as desired. 
Example. Any field k is Noetherian.
Lemma 2.4. If we have a surjection φ : A  B, then A Noetherian implies B Noetherian.
Proof. For any ideal I ⊂ B, note that φ−1 (I) ⊂ A is finitely generated, and the projections of the
generators of φ−1 (I) generate I. 
Theorem 2.5 (Hilbert Basis Theorem). If A is Noetherian, then A[x] is Noetherian.
Proof. Take any ideal I ⊂ A[x]. First, let J ⊂ A be the ideal generated by the highest degree
coefficients of elements of I. Let a1 , . . . , an be generators for J (these exist because A is Noetherian).
Let f1 , . . . , fn be polynomials in A[x] such that ai is the highest degree coefficient in fi . Let
d = max{deg(f1 ), . . . , deg(fn )}. Consider I ∩ (A[x])≤d , which is finitely generated as a submodule
of Ad+1 ; let g1 , . . . , gm be a set of generators.
Now, we claim that f1 , . . . , fn , g1 , . . . , gm generate I. Suppose otherwise for contradiction.
Let f ∈ I be an element of minimal degree not belonging to the A[x]-submodule generated by
f1 , . . . , fn , g1 , . . . , gm . First, note that
P deg(f ) > d. Observe now that the leading coefficient a of f
belongs to J, so we can write a = i ci ai for some coefficients ci . Then, consider
X
f− ci fi xdeg(f )−deg(fi ) ,
i
10 YI SUN

which is an element of degree less than deg(f ) in I which does not belong to the A[x]-submodule
generated by f1 , . . . , fn , g1 , . . . , gm . This is a contradiction, completing the proof. 
For the rest of this section, all rings will be Noetherian and commutative. Consider a map
A → B of rings.
Definition. We say that B is A-finite or finite as an A-module if there exists a surjection An  B
of A-modules.
Definition. We say that B is finite over A or finite as an A-algebra if there exists a surjection
A[x1 , . . . , xn ]  B of A-algebras.
Lemma 2.6. For maps A → B → C, if C is finite over B and B is finite over A, then C is finite
over A.
Proof. Let the maps be φ : A → B and ψ : B → C. Take a (finite) set of generators {ci } for C over
B, and a set of generators {bj } for B over A. Then, for any c ∈ C, we can write
 
Xn Xn Xm n X
X m
c= ψ(bi ) · ci = ψ φ(aij ) · bj  · ci = ψ(φ(aij ))ψ(bj ) · ci ,
i=1 i=1 j=1 i=1 j=1

for some aij ∈ A; hence, {ψ(bj ) · ci } is a finite set of generators for C over A. 
Definition. For b ∈ B, take a homomorphism φ : A[x] → B given by φ(x) = b. We say that b is
integral over A if Im(φ) is finite as an A-module.
Lemma 2.7. The following are equivalent:
(1) b is integral over A.
(2) B satisfies a monic polynomial equation in A. i.e., there exist a0 , . . . , an−1 ∈ A such that
bn + an−1 bn−1 + · · · + a0 = 0.
Proof. (2) =⇒ (1): Note that Im(φ) is generated by {1, b, . . . , bn−1 }.
(1) =⇒ (2): Take Mi = span{1, . . . , bi }. We have that i Mi = Im(φ); further, Mi will eventually
S
contain all generators of Im(φ), so we can find some n with Im(φ) = Mn ; in particular, this means
that bn+1 is generated by 1, b, . . . , bn , finishing the proof. 
Lemma 2.8. Suppose that b1 , . . . , bn are integral over A and let φn : A[x1 , . . . , xn ] → B the
corresponding map of algebras. Then, Im(φn ) is A-finite.
Proof. Consider the map φi = φn |A[x1 ,...,xi ] : A[x1 , . . . , xi ] → B. Then, consider the sequence of
maps
A → Im(φ1 ) → Im(φ2 ) → · · · → Im(φn−1 ) → Im(φn ).
Since b1 is integral over A, we see that Im(φ1 ) is finite over A. We now claim that Im(φi ) is finite
over Im(φi−1 ) for all i. Since bi is integral over A, we can find some finite set of generators {bi,t } for
the image of the map A[x] → B given by x 7→ bi (as an A-module); we claim that {bi,t } generate
Im(φi ) as a Im(φi−1 )-module. Take any φi (p(x1 , . . . , xi )) ∈ Im(φi ) and write
qj (x1 , . . . , xi−1 )xji
X
p(x1 , . . . , xi ) =
j

for some qj ∈ k[x1 , . . . , xi−1 ]. This means that

φi−1 (qj (x1 , . . . , xi−1 ))bji =
X X X
φi (p(x1 , . . . , xi )) = φi−1 (qj (x1 , . . . , xi−1 )) ci,j,t bi,t .
j j t

But φi−1 (qj (x1 , . . . , xi−1 )), ci,j,t ∈ Im(φi−1 ) (since ci,j,t ∈ φi−1 (A)), so we’ve just shown that bi,t
generate Im(φi ) as a Im(φi−1 )-module. This means that Im(φi ) is finite over Im(φi−1 ), as desired.
Now, apply Lemma 2.6 repeatedly to see that Im(φn ) is finite over A. 
 MATH 221 NOTES: PART 1 11

Proposition 2.9. Let b be integral over A and φ : A[x] → B the corresponding map. Then, every
element b0 ∈ Im(φ) is integral over A.
Proof. Let φ0 : A[x0 ] → B be given by x0 7→ b0 . Because x0 ∈ Im(φ), we see that Im(φ0 ) ⊂ Im(φ) is
finite over A because A is Noetherian. 
Definition. B is integral over A if all elements b ∈ B are integral over A.
2.2. Finitely Generated Algebras. Let A → B be a map of A-algebras.
Definition. We say that B is finitely generated as an A-algebra if there exist finitely many b1 , . . . , bn
such that
φn : A[x1 , . . . , xn ] → B
is surjective.
Lemma 2.10. If A is Noetherian and B a finitely generated A-algebra, then B is Noetherian.
Proof. Take a surjection A[x1 , . . . , xn ]  B. Note that A[x1 , . . . , xn ] is Noetherian by the Hilbert
Basis Theorem. We’ve just represented B as a quotient of A[x1 , . . . , xn ], so B is also Noetherian. 
Lemma 2.11. An A-algebra B is finite as an A-module if and only if it is finitely generated by
elements integral over A.
Proof. First suppose that B is finite over A. Then, it’s clearly finitely generated as an A-algebra
(since it is already finitely generated as an A-module). Now, for any b ∈ B, let φb : A[x] → B be
the evaluation map at b. Then, Im(φb ) ⊂ B is a sub-module of the finitely generated A-module B;
because A is Noetherian, it is finitely generated as well. Thus, b is integral and hence B is integral
over A.
Now, suppose that B is integral and finitely generated as an A-algebra. Choose generators
b1 , . . . , bk for B as an A-algebra, and consider the map φk : A[x1 , . . . , xk ] → B given by evaluation
at (b1 , . . . , bk ). Since B is integral, we see that b1 , . . . , bk are integral, so, by Lemma 2.8, we find
that B = Im(φk ) is finite over A, as desired. 
Corollary 2.12. An A-algebra B is finite as an A-module if and only if it is finitely generated as
an A-algebra and integral over A.

3. The Nullstellensatz
Before introducing the Nullstellensatz, we provide some motivation from algebraic geometry.
First, what is algebraic geometry?
• Take a field k and f ∈ k[x1 , . . . , xn ]. If c = (c1 , . . . , cn ) ∈ k n , we denote the evaluation of f
at c as a polynomial by f (c) ∈ k.
• For α in some index set A, take fα ∈ k[x1 , . . . , xn ]. Then, we take the algebraic subset of
k n defined by {fα } to be V ({fα }) = {c ∈ k n , fα (c) = 0}.
Note that V ({fα }) depends only on the ideal I generated by fα and that I is finitely generated
because k[x1 , . . . , xn ] is Noetherian. We can therefore denote this algebraic subset by V (I). The
question we seek to answer in algebraic geometry is:
If we know V (I), can we reconstruct I?
Remark. Not every X ⊂ k n arises as V (I) for some I. For instance, take n = 1, and X an
arbitrary infinite subset of k. Then, V (I) 6= X for any I, because any polynomial has a finite
number of roots.
Remark. For I not equal to k[x1 , . . . , xn ], V (I) can be empty. Take k = R and I = (x2 + 1). So
we would like k to be algebraically closed.
12 YI SUN

Remark. If I 6= J, we can still have V (I) = V (J). Take k algebraically closed, I = (x), and
J = (x2 ).
The last remark is a special case of the following theorem, which is the goal of this section.
Theorem 3.1 (Nullstellensatz). Let k be algebraically closed, I an ideal in k[x1 , . . . , xn ]. Suppose
that f |V (I) = 0. Then, there exists n such that f n ∈ I.
There are several related and weaker versions of this theorem which we will consider before
proving it. Recall first the following criterion from ring theory for an ideal to be maximal.
Lemma 3.2. An ideal m ⊂ R is maximal if and only if R/m is a field.
Lemma 3.3. The set k n is in bijection with maps of k-algebras k[x1 , . . . , xn ] → k.
Proof. For c ∈ k n , let φc : k[x1 , . . . , xn ] → k be defined by φc (xi ) = ci . This gives the desired
bijection. Note that we obtain in this way a short exact sequence
0 → mc → k[x1 , . . . , xn ] → k → 0,
where ker(φc ) = mc is a maximal ideal. 
The following theorem gives a converse to the construction of Lemma 3.3. We defer the proof.
Theorem 3.4 (Weak Nullstellensatz). Let k be algebraically closed. Every maximal ideal in
k[x1 , . . . , xn ] is of the form mc for some c ∈ k n .
Corollary 3.5. Let A be a finitely generated k-algebra (where k is algebraically closed). Then,
every maximal ideal in A is of the form ker φ for φ : A → k a map of k-algebras.
Proof. Take a surjection ψ : k[x1 , . . . , xn ]  A. For any maximal ideal ma ⊂ A, consider the
composition φ : k[x1 , . . . , xn ] → A → A/ma = K. Now, take ker(φ) = ψ −1 (ma ) = m, which is
maximal in k[x1 , . . . , xn ] because A is a finitely generated k-algebra. This means that Im(φ) =
k[x1 , . . . , xn ]/m = k and hence that ma = ker(A → k), as needed. 
Now, let us consider the short exact sequence
0 → I → k[x1 , . . . , xn ] → A → 0.
We then have the following correspondence between maps A → k and points in the variety V (I).
Lemma 3.6. The restriction of the bijection {k[x1 , . . . , xn ] → k} ' k n gives a bijection {A →
k} ' V (I) = {c ∈ k n | f (c) = 0 for all f ∈ I}.
Proof. Take some φ : k[x1 , . . . , xn ] → k that factors through A. Recall that under the identification
{k[x1 , . . . , xn ] → k} ' k n , φ maps to c = (φ(x1 ), . . . , φ(xn )). Now, for f ∈ I, notice that f (c) =
φ(f ) = 0, giving one direction. The other direction is similar; any c ∈ k n with f (c) = 0 for all
f ∈ I corresponds to the map φ : k[x1 , . . . , xn ] → k such that φ(xi ) = ci . For any f ∈ I, we see
that φ(f ) = f (c) = 0, so φ|I = 0, completing the proof. 
Given A, write Spech(A) = {A → k} for the set of k-algebra maps A → k and Specm(A) = {m |
m maximal in A}. Note that Spech(k[x1 , . . . , xn ]) ' k n . Observe that there is a map Spech(A) →
Specm(A) given by φ 7→ ker(φ).
Theorem 3.7 (Maximalidealensatz (MIS)). When k is algebraically closed and A is finitely gen-
erated over k, the map Spech(A) → Specm(A) is an isomorphism.
We defer the proof for now in favor of some remarks. Notice that this fails if A = k 0 ⊃ k is an
infinite field extension. In particular, for A = k(x), Spech(A) = k ∗ , while Specm(A) = (0). For
non-algebraically closed fields, we have the following refinement.
 MATH 221 NOTES: PART 1 13

Theorem 3.8 (Maximalidealensatz (non-algebraically closed k)). There is a bijection between

Specm(A) and maps of k algebras A  k 0 , where k 0 is an algebraically closed extension of k and
we take two maps φ1 : A → k10 and φ2 : A → k20 to be equivalent if there exists an isomorphism
completing the following diagram:
φ1
A > k10
w
w
w
w
∼
w
w
φ2 w
w
w

>
k20
Theorem 3.9 (Endlich-Ergzeute-Korper-Satz (EEKS)). Let k ⊂ F be a field extension such that
F is finitely generated as a k-algebra. Then, F is a finite field extension of k.
Proof. We will first prove the following lemma.
Lemma 3.10. Let A be Noetherian and A ⊂ B ⊂ C algebras such that
• C is finitely generated as an A-algebra.
• C is finitely generated as a B-module.
Then, B is finitely generated as an A-algebra.
Proof. Let x1 , . . . , xn generate C Pas an A-algebra and y1 , . . . , ym generate C as a B-module. This
means that we can write xi = j bij yj for some bij ∈ B. Further, there are some bijk such that
P
yi yj = k bijk yk . Now, let B0 ⊂ B be the A-subalgebra in B generated by {bij , bijk }. This is a
finitely generated algebra over A and is in particular Noetherian. P
We now claim that y1 , . . . , ym generate C as a B0 -module. First, note that xi = j bij yj .
Further, the xi generate C as an A-algebra. Thus each monomial of the xi ’s is in the linear span
of y1 , . . . , ym with coefficients bijk by our definition, giving the claim.
Now, note that B0 is Noetherian and B ⊂ C, so B is finitely generated as a B0 -module. In
particular, it is finitely generated as B0 -algebra, meaning that it is finitely generated as an A-
algebra via the chain A → B0 → B of finitely generated algebras. 
We can now proceed to the proof of the theorem. Let x1 , . . . , xn ∈ F be generators of F as a field
over k. Let x1 , . . . , xm be a maximal subset of algebraically independent elements; i.e., we have
a tower of field extensions k ,→ k(x1 , . . . , xm ) ⊂ F , where the last extension k(x1 , . . . , xm ) ⊂ F
is algebraic and hence finite. By Lemma 3.10, we see that k(x1 , . . . , xm ) is finitely generated as a
k-algebra and in particular as an algebra over k(x1 , . . . , xm−1 ). This is a contradiction, since K(x)
is never a finitely generated K-algebra for any field K (here we take K = k(x1 , . . . , xm−1 )). 
Theorem 3.9 is the key technical assertion that allows us to prove the various formulations of
the Nullstellensatz.
Proof of Theorem 3.4. Let m be a maximal ideal in k[x1 , . . . , xn ]. Then, we have the projection
k[x1 , . . . , xn ] → k[x1 , . . . , xn ]/m = K for K a field extension of k. Observe that K is finitely
generated as a k-algebra (since the images of x1 , . . . , xn are generators), so by Theorem 3.9, it
is finitely generated as a k-module. But k was algebraically closed, so it has no non-trivial finite
extensions, meaning that K = k. Thus we have the exact sequence 0 → m → k[x1 , . . . , xn ] → k → 0,
as desired. 
Proof of Theorem 3.7. We first claim that it is enough to prove the theorem for A a polynomial
algebra. Suppose that the theorem held for polynomial algebras. Then, for a finitely generated A,
we’d have a short exact sequence
0 → I → k[x1 , . . . , xn ] → A → 0.
14 YI SUN

Take m ⊂ A and consider m0 = φ−1 (m). Note that I ⊂ m0 , so there exists φ0 : k[x1 , . . . , xn ] → k
such that ker(φ0 ) = m0 . This means that I ⊂ ker(φ0 ) and hence φ0 factors through A.
Take some maximal ideal m ⊂ A, and consider A/m = F . We want to show that k ,→ F is an
isomorphism. Note that F is a finitely generated k algebra, so F is a finite field extension of k.
Since k is algebraically closed, this means that k = F . 
Definition. For I ⊂ k[x1 , . . . , xn ], define its radical to be rad(I) = {a ∈ A | an ∈ I for some n}.
We therefore have the following restatement of the Nullstellensatz (Theorem 3.1).
Theorem 3.11. Let k be algebraically closed. Then, if f ∈ k[x1 , . . . , xn ] vanishes on V (I), then
f ∈ rad(I).
Corollary 3.12. If V (I1 ) = V (I2 ), then rad(I1 ) = rad(I2 ) (for k algebraically closed).
Theorem 3.13. Let A be a finitely generated algebra over an algebraically closed field k. Then,
if f ∈ A is such that f vanishes under every element of Spech(A), then f is nilpotent.
We will first prove that Theorems 3.11 and 3.13 are equivalent and then prove Theorem 3.13 to
establish the full Nullstellensatz.
Proof of equivalence of Theorems 3.11 and 3.13. Take a surjection k[x1 , . . . , xn ]  A, giving the
short exact sequence
0 → I → k[x1 , . . . , xn ] → A → 0.
˜
Now, for f ∈ A, take f ∈ k[x1 , . . . , xn ] mapping to f . By Lemma 3.6 and Theorem 3.7, f vanishes
under each element of Spech(A) if and only if f˜ vanishes on V (I). The desired equivalence then
follows from the fact that f˜ is in rad I if and only if f is nilpotent. 
Proof of Theorem 3.13. Suppose for contradiction that f were not nilpotent. Since f vanishes
under all elements of Spech(A), by Theorem 3.7, we see that f ∈ m for all maximal ideals m ∈ A.
We first claim that (tf − 1) is a proper ideal in A[t]. Suppose otherwise for contradiction; then,
we’d have that 1 = p(t) · (tf − 1) for some p(t) = a0 + a1 t + · · · + an tn ∈ A[t], meaning that
p(t) · (tf − 1) = f an tn+1 + (f an−1 − an )tn + · · · + (f a0 − a1 )t − a0 = 1.
This implies that a0 = −1, f an = 0, and ai+1 = f ai for 1 ≤ i ≤ n, so we may conclude that
f n+1 a0 = −f n+1 = 0, which is a contradiction because f is not nilpotent.
Now, consider the algebra B = A[t]/(tf − 1); since A[t] is a finitely generated k-algebra, so is
B. Now, consider the map φ : A ,→ A[t] → B. Take any maximal ideal m0 ⊂ B. Notice that φ(f )
/ m0 . Thus, we see that f ∈
is invertible in B, so φ(f ) ∈ / φ−1 (m0 ), which is maximal because A is a
finitely generated k-algebra. This is a contradiction, so we are done. 
Remark. In the proof of Theorem 3.13, the construction of B = A[t]/(tf − 1) may at first seem
like an unmotivated trick. However, it is in fact the localization of A at f , forming an instance of
a much more general construction.