Solutions to Exercises in Eisenbud’s Commutative

Algebra
Sam P. Fisher
April 21, 2021

1 Roots of Commutative Algebra

Exercise 1.1. Prove that the following conditions on a module M over a commutative ring
R are equivalent (the fourth is Hilbert’s original formulation; the first and third are the ones
most often used). The case M = R is the case of ideals.
1. M is Noetherian (that is, every submodule of M is finitely generated).
2. Every ascending chain of submodules of M terminates (“ascending chain condition”).
3. Every set of submodules of M contains elements maximal under inclusion.
4. Given any sequence of elements f1 , f2 , ·P · · ∈ M , there is a number m such that for
m
each n > m there is an expression fn = i=1 ai fi with ai ∈ R.

(1. ⇒ 2.) Let N1 S⊆ N2 ⊆ · · · be an ascending chain of submodules of M . Consider

the submodule N = i∈N Ni . Since M is Noetherian, N is finitely generated, say by the
elements f1 , . . . , fn . There is some k such that each fi is contained in Nk , and therefore the
ascending chain of submodules terminates at k.
(2. ⇒ 3.) Let {Nα }α∈A be an arbitrary collection of submodules of M . Assume that
the collection does not have a maximal element under inclusion. Let Nα1 be an arbitrary
submodule in the collection. By assumption, it is not maximal, so there is some Nα2 such
that Nα1 ⊂ Nα2 . Inductively assume that we have found Nα1 ⊂ · · · ⊂ Nαi . Since Nαi is
not maximal under inclusion, we have that Nαi ⊂ Nαi+1 . Like this we construct a chain of
modules violating the ascending condition.
(3. ⇒ 4.) Consider the chain of submodules (f1 ) ⊆ (f1 , f2 ) ⊆ · · · . Since there is a
maximal element among these, there is some m such that (f1 , . . . , fn ) = (f1 , . . . , fm ) for all
n > m. This is clearly equivalent to what was to be shown.
(4. ⇒ 1.) Let N be an arbitrary submodule of M . Suppose that N is not finitely
generated. Then it is clear that we can choose f1 , f2 , . . . such that (f1 ) ⊂ (f1 , f2 ) ⊂ · · · .
This violates (4).

Exercise 1.2. Prove that if R is Noetherian, and I ⊂ R is an ideal, then among the primes
of R containing I there are only finitely many that are minimal with respect to inclusion
(these are usually called the minimal primes of I, or the primes minimal over I
as follows: Assuming that the proposition fails, the Noetherian hypothesis guarantees the
existence of an ideal I maximal among ideals in R for which it fails. Show that I cannot be
prime, so we can find elements f and g in R, not in I, such that f g ∈ I. Now show that
every prime minimal over I is minimal over one of the larger ideals (I, f ) and (I, g).

1
 Let {Iα }α∈A be the collection of ideals for which the result of the exercise fails (we assume
the result of the exercise fails, so this collection is nonempty). The Noetherian condition
guarantees that there is a maximal ideal I under inclusion in the collection. Note that I
cannot be prime, because otherwise there is only one prime containing I that is minimal
under inclusion, namely I itself. Since I is not prime, there are elements f, g ∈ R − I such
that f g ∈ I. Note that any prime ideal containing I must then contain f or g, which implies
that it must contain (I, f ) or (I, g). Since there infinitely many prime ideals containing I
that are minimal under inclusion, the same is true of either (I, f ) or (I, g). But both of
these ideals are strictly larger than I, which violates maximality of I.
Exercise 1.3. Let M 0 be a submodule of M . Show that M is Noetherian iff both M 0 and
M/M 0 are Noetherian.
Let M be Noetherian. Let N be a submodule of M 0 ; then N is also a submodule of
M , and is therefore finitely generated since M is Noetherian. Let N be a submodule of
M/M 0 and let π : M → M/M 0 be the natural projection. The preimage π −1 (N ) is finitely
generated, and the image of any finite generating set under π becomes a finite generating
set for N . Therefore M/M 0 is Noetherian.
Conversely, suppose that M 0 and M/M 0 are Noetherian, and let N be a submodule of
M . Then M 0 ∩ N and π(N ) are submodules of M 0 and M/M 0 and are therefore finitely
generated. Let {a1 , . . . , am } and {b1 , . . . , bn } be generating sets for M 0 ∩ N and π(N ). Let
ci ∈ N be any element such that π(ci ) = bi . We will show that {a P1 , . . . , am , c1 , . . . , cn }
generates N . P To this end, let α ∈ N . Then we can write π(α) = i ri bi . Consider the
element β = i ri ci for some P ri ∈ R. Then α − β ∈ ker π, so we P have α P − β ∈ N ∩ M 0.
Hence, we can write α−β = j r̃j aj for some r̃j ∈ R. Thus, α = j r̃j aj + i ri ci , proving
that {a1 , . . . , am , c1 , . . . , cn } generates N , and therefore that M is Noetherian.
Exercise 1.4. We have seen from Corollary 1.3 that any finitely generated algebra over a
field is Noetherian. The converse is quite false, and we shall see many important examples
of rings that are Noetherian but not finitely generated (for instance localizations and com-
pletions). But the converse is true for graded rings R where R0 is a field, as the following
result shows.
Let R = R0 ⊕ R1 ⊕ · · · be a graded ring. Prove that the following are equivalent:
1. R is Noetherian.
2. R0 is Noetherian and the irrelevant ideal R1 ⊕ R2 ⊕ · · · is finitely generated.
3. R0 is Noetherian and R is a finitely generated R0 -algebra.

(1. ⇒ 2.) Since R0 is a summand of R, it is also Noetherian as the next exercise will
show. The irrelevant ideal is finitely generated by the definition of a Noetherian ring.
(2. ⇒ 3.) R0 is Noetherian by assumption, so it remains only to show that R is a finitely
generated R0 -algebra. Suppose {x1 , . . . , xn } is a finite generating set for the irrelevant ideal.
We will show that this set also generates R as an R0 -algebra. To this end, let α ∈ R; we
will induct on the degree of α. For the base case, suppose that α has degree 0. Then α ∈ R0
and there is nothing to show. Now suppose that α ∈ Rk for some k > 0. Then we can
write α = β + γ, where β ∈ R0 ⊕ · · · ⊕ Rk−1 and γ ∈ PRk . By induction, β ∈ hx1 , . . . , xn i.
Since γ is in the irrelevant ideal, we have that γ = i ri xi for some ri ∈ R. Since every
xi is in the irrelevant ideal, we have deg xi ≥ 1 for every i. Hence, the ri can be chosen
such that deg ri ≤ k − 1 for every i. By induction, ri ∈ hx1 , . . . , xn i for every i. Therefore,
γ ∈ hx1 , . . . , xn i, which proves that {x1 , . . . , xn } generates R as an R0 -algebra.

2
 (3. ⇒ 1.) For some n ∈ N, there is a surjection R0 [x1 , . . . , xn ] → R. Since R0 is Noethe-
rian, the basis theorem implies that R0 [x1 , . . . , xn ] is as well. Therefore, R is Noetherian.
Exercise 1.5. Although the Noetherian property does not usually pass from a ring to a
subring, it does when the subring is a summand:
Let R ⊆ S be rings, and assume that R is a summand of S as an R-module, that is,
there is a homomorphism φ : S → R of R-modules every element of R. Prove that if S is
Noetherian, then R is Noetherian.
Every homomorphic image of a Noetherian module is Noetherian (the proof is the same
as the proof in the case of rings). Therefore, R is Noetherian as an R-module. But this
is the same as being Noetherian as a ring, since the ideals of R correspond precisely to its
R-submodules.
Exercise 1.6. The following proof of the assertions of Example 1.1 is from Van der Waerden
[1971]. We shall systematically develop this method in chapter 15. Let Σ and f1 , . . . , fn be
as defined in Example 1.1.
Order the monomials of the polynomial ring S = k[x1 , . . . , xr ] according to the degree-
lexicographic order, defined as follows: Let A = xm mr
1 · · · xr
1
and B = xn1 1 · · · xnr r be two
monomials. We say that A > B if either deg A > deg B, or else deg A = deg B and the
sequence of exponents (m! , . . . , mr ) is greater than the sequence (n1 , . . . , nr ) in the lexico-
graphic order; that is, the difference mi − ni > 0 for the first index i for which it is not
zero.
Given any polynomial p(x1 , . . . , xr ), we define the initial term of p to be the term
involving the greatest monomial in the order >.
a. Show that for each monomial A there are only finitely many monomials B such that
A > B.
b. Show that if p is invariant under Σ, then the initial term of p is an element of k times
a monomial xm 1 mr
1 · · · xr with m1 ≥ m2 ≥ · · · ≥ mr . Pr
c. Show that the initial term of the product f1µ1 · · · frµr is xm mr
1 · · · xr where mi =
1
j≥i µj .
d. Show that the function Zr → Zr defined by
r
X
(µ1 , . . . , µr ) 7→ (m1 , . . . , mr ) with mi = µj
j≥i

is a monomorphism. Conclude that a monomial xm mr

1 · · · xr
1
with m1 ≥ m2 ≥ · · · ≥ mr
is the initial monomial of a unique product of fi .
e. Now show that any element of S Σ can be written uniquely as a polynomial in the fi .
a. If A > B, then deg A ≥ deg B, so it suffices to show that there are finitely many
monomials in r variables of a given degree d. Indeed, if xm 1 · · · xr
1 mr
has degree d, then
d = m1 + · · · + mr . This implies that 0 ≤ mi ≤ d for every i. Since there are only finitely
many choices for each mi , we conclude that there are only finitely many monomials of degree
d.
b. Suppose not. Then the initial term is A = xm mr
1 · · · xr
1
and there is some i such that
mi+1 mi
mi < mi+1 . Since p is invariant under Σ, we know that B = xm mr
1 · · · xi+1 xi · · · xr
1
is
also a monomial in p with a nonzero coefficient. But deg B > deg A, a contradiction.
c. We first prove the following claim: if f, g ∈ S have initial terms A, B, then the initial
term of f g is AB. The only way AB might not be the initial term of f g is if there are

3
monomials A0 , B 0 in f, g such that A0 B 0 ≥ AB. By the definition of the lexicographical
order, we must have that deg A = deg A0 and deg B = deg B 0 . Let
0 m0 m0r
A = xm mr
1 · · · xr , A = x1 · · · xr
1 1

n0 n0
B = xn1 1 · · · xnr r , B 0 = x1 1 · · · xr r .
Since A > A0 and B > B 0 , there are integers i, j such that mi − m0i > 0 and nj − n0j > 0.
We choose i, j to be minimal in the sense that mk = m0k and nl = n0l for k < i and l < j.
Without loss of generality, we assume that i ≤ j. But then we easily see that AB > A0 B 0 ,
since mi + ni > m0i + n0i and i is the first index where the exponents differ. This proves the
claim.
We now prove the assertion by induction on k = µ1 + · · · + µr . For k = 0, the assertion is
immediate. Suppose that the initial term of f1µ1 · · · frµr is as above, and consider the product
fi · f1µ1 · · · frµr = f1µ1 · · · fiµi +1 · · · frµr . The initial term of fi is x1 · · · xi , so the initial term
i +1 mi+1
of fi · f1µ1 · · · frµr is xm1
1 +1
· · · xm
i xi+1 · · · xm
r , which is as required.
r

d. It is clear that the function defined is a homomorphism. Now suppose that (µ1 , . . . , µr ) 7→
(0, . . . , 0). By the definition of the map, we immediately Pr have that µr = 0. Inductively as-
sume that µk = · · · = µr = 0. Then µk−1 = j=k−1 µr = 0. Hence, (µ1 , . . . , µr ) =
(0, . . . , 0), the map is a monomorphism.
A monomial xm 1 · · · xr
1 mr
with m1 ≥ · · · ≥ mr is always the initial term of
m −mr
f1m1 −m2 · · · fr−1
r−1
frmr
by the result of part c. Since the function defined in part d. is injective, there can only be
one product of fi ’s with initial term xm mr
1 · · · xr .
1

e. Let p ∈ S . By part c. the initial term of p is αxm

Σ mr
1 · · · xr
1
with mi such m1 ≥ · · · ≥
mr and α ∈ k − {0}. By part d. there is a unique product f = kf1µ1 · · · frµr with the same
initial term as p. Then p − f ∈ S Σ and its initial term is strictly small than that of p with
respect to <. Part a. guarantees that after repeating this procedure a finite number of times
we obtain that p is a polynomial in the fi .
Now suppose that p can be expressed in two different ways as a polynomial in the fi ,
i.e. let p(x1 , . . . , xr ) = p1 (f1 , . . . , fr ) = p2 (f1 , . . . , fr ). Let A be the initial term of p. By
part d. there is a unique product of fi ’s with initial term A. We may remove this product
from p1 and p2 . Continuing this process inductively shows that p1 = p2 .
Exercise 1.7. a. (The simplest group action whose ring of invariants is not a polynomial
ring) Suppose that k is a field of characteristic 6= 2. Let the generator g of the group
G := Z2 act on the polynomial ring k[x, y] in two variables by sending x to −x and
y to −y. Show that the ring of invariants is k[x2 , xy, y 2 ]. Prove that k[x2 , xy, y 2 ] ∼
=
k[u, v, w]/(uw − v 2 ). Show that this is not isomorphic to any polynomial ring over a
field.
b. More generally, let G be any finite Abelian group, acting linearly on the space of linear
forms of the ring S = k[x1 , . . . , xr ]. Assume that G acts by characters; that is assume
that there are homomorphisms αi : G → k × , and g(xi ) = αi (g)xi for all g ∈ G, where
k × is the multiplicative group of the field k. (As long as the characteristic of k does
not divide the order of G, this could be achieved, by a suitable choice of variables, for
any
Q aaction of G.) Show that the invariants of G are generated by those monomials
xi i whose exponent vectors (a1 , . . . , ar ) are in the kernel of a map from Zr to a
certain finite Abelian group. Conclude that the quotient field S G is isomorphic to a
field of rational functions in r variables.

4
 a. Let f ∈ k[x, y] be a polynomial invariant under the action of G. It is straightforward
to see that every monomial in the support of f be of the form xm y n where m and n are
either both even or both odd (this uses the fact the characteristic of k is not 2). Therefore,
f ∈ k[x2 , xy, y 2 ]. Conversely, x2 , y 2 , and xy are all invariant under the G-action. Therefore,
the ring of invariants is k[x2 , xy, y 2 ]. To prove the isomorphism, consider the map φ :
k[u, v, w] → k[x2 , xy, y 2 ] sending u, v, w to x2 , xy, y 2 , respectively. Then φ is clearly onto, so
we just need to determine its kernel. It is clear that (uw − v 2 ) ⊆ ker φ. On the other hand,
let p ∈ k[u, v, w] be a polynomial such that p(x2 , xy, y 2 ) = 0. Using the division theorem
over the ring k[u, w], we obtain that
p(u, v, w) = q(u, v, w)(v 2 − uw) + r(u, v, w),
for some polynomials q and r, and where the v-degree of r is at most 1. If degv r = 0,
we can write r(u, v, w) = r̃(u, w). After plugging in u = x2 , v = xy, w = y 2 , we find that
r̃(x2 , y 2 ) = 0, which means that r̃ is the zero polynomial. Thus, r = 0 and we have that
v 2 − uw divides p as desired. Now consider the case where degv r = 1. Then there are
polynomials a, b ∈ k[u, w] such that
r(u, v, w) = a(u, w)v + b(u, w).
Making the same substitution as earlier, we have that a(x2 , y 2 )xy + b(x2 , y 2 ) is the zero
polynomial. Since every term in a(x2 , y 2 )xy has odd x- and y-degrees and every term in
b(x2 , y 2 ) has even x- and y-degrees, we must have that a and b are actually zero polynomials.
Hence, r is once again zero and we have the desired result. This proves the isomorphism.
To prove that k[x2 , xy, y 2 ] is not isomorphic to any polynomial ring, we prove that
it is not a UFD. First, note that x2 , xy, and y 2 are all irreducible in k[x2 , xy, y 2 ]. To
prove this for x2 , assume that x2 = f (x2 , xy, y 2 )g(x2 , xy, y 2 ) for some polynomials f and
g. We must then have that f and g are independent of y, and therefore of their second
and third variables. But then by degree considerations we conclude that either f or g is
constant. The proof that y 2 is irreducible is similar. To see that xy is irreducible, again
write xy = f (x2 , xy, y 2 )g(x2 , xy, y 2 ). By looking at the x- and y-degrees of both sides, we
see that f and g can only depend on their second variable. Then either f or g must be a
constant polynomial. However, x2 y 2 = (x2 )(y 2 ) = (xy)(xy), so k[x2 , xy, y 2 ] is not a unique
factorization domain.
b. A monomial xa1 1 . . . xar r is G-invariant if and only if α1 (g)a1 · · · αr (g)ar = 1 for every
g ∈ G. Moreover, f ∈ S G ig and only if it is the sum of invariant monomials (this is because
the G-action does not permute the variables). Hence, S G is generated by monomials whose
exponent vectors are in the kernel of the following map
φ : Zr → (k × )|G| , (a1 , . . . , ar ) 7→ (α1 (g a1 ) · · · αr (g ar ))g∈G
The image of each αi is finite, so the image of φ is finite. Hence, S G is generated by
monomials in the kernel of a map from Zr to a finite Abelian subgroup. Therefore ker φ
must have finite index in Zr , which implies that ker φ must be a free Abelian group of rank
r. Thus, the fraction field of S G is generated by r monomials over k (hint: we can use
division and multiplication to add and subtract vectors on the the level of monomials). Let
Ai = xa1 i1 · · · xar ir , 1 ≤ i ≤ r, denote the generators of the fraction field of S G . To show that
k(A1 , . . . , Ar ) is isomorphic to a fraction field of polynomials in r variables is equivalent to
showing that the Ai are algebraically independent. We do this by computing the Jacobian
of the Ai : when evaluated at x1 = · · · xr = 1, it is equal to det((aij )) 6= 0. Therefore, the
Ai are algebraically independent as desired.

5
Exercise 1.8. Let X and J be partially ordered sets, and suppose that I : X → J and
Z : J → X are functions such that
i) I and Z reverse the order in the sense that x ≤ y ∈ X implies I(x) ≥ I(y), and
i ≤ j ∈ J implies Z(i) ≥ Z(j).
ii) ZI and IZ are increasing functions, in the sense that x ∈ X implies ZI(x) ≥ x, and
i ∈ J implies IZ(i) ≥ i.
a. Show that I and Z establish a one-to-one correspondence between the subsets
I(X) ⊆ J and Z(J) ⊆ X.
b. Let k be a field. Call an ideal I ⊆ k[x1 , . . . , xn ] formally radical if it is of
the form I(X) for some X ⊆ k n . Use part a. to prove that there is a one-to-
one correspondence between formally radical ideals and algebraic subsets of k n .
(Hilbert’s Nullstellensatz identifies the formally radical ideals with the ordinary
radical ideals when k is algebraically closed.)
a. Let x = Z(J). Then x = Z(i) for some i ∈ J. This implies I(x) = IZ(i) ≥ i, which
in turn implies x ≤ ZI(x) ≤ Z(i) = x. Therefore ZI(x) = x. The proof that IZ(i) = i for
i ∈ J is the same. Thus, I and Z are inverses of each other when restricted to I(X) and
Z(J).
b. In this setting, X is the set of all subsets of k n and J is the set of all ideals of
k[x1 , . . . , xn ]. Then I : X → J is the function that assigns to X the ideal of polynomials
vanishing on X, and Z : J → X is the function that assigns to I the set where all polynomials
in I vanish. Both X and J are ordered by inclusion. For an ideal of polynomials J, IZ(J)
is the ideal of polynomials that vanish on the zero-set of J, so clearly J ⊆ IZ(J). For a set
X ⊆ k n , ZI(X) is the set where the polynomials in I(X) vanish, so we have X ⊆ ZI(X).
The result of part b. is then immediate from that of part a.
Exercise 1.9. Let S = k[x1 , . . . , xr ], with k an algebraically closed field. Show that under
the correspondence of radical ideals in S and algebraic subsets of Ar , the prime ideals cor-
respond to the algebraic sets that cannot be written as a proper union of smaller algebraic
sets.
Let J / S be a prime ideal, and suppose that Z(J) is the proper union of two algebraic
subsets. Then there are radical ideals J1 and J2 such that Z(J) = Z(J1 )∪Z(J2 ) = Z(J1 J2 ).
Since the union is proper, neither algebraic subset is contained in the other. Thus, the
Nullstellensatz implies that there are exist elements j1 ∈ J1 −J2 and j2 ∈ J2 −J1 . Moreover,
since
√ Z(J1 ), Z(J2 ) ⊆ Z(J), we have that J ⊆ J1 ∩ J2 . Finally, note that j1 j2 ∈ J1 J2 ⊆
J1 J2 = J by the Nullstellensatz. However, j1 , j2 ∈
/ J. This contradicts that J is prime.
Conversely, suppose that X ⊆ An is an algebraic set that is not the proper union of two
smaller algebraic subsets. Let X = Z(J), and assume that J is not prime. Then there are
elements j1 , j2 such that the product j1 j2 ∈ J but j1 , j2 ∈
/ J. Let J1 = (j1 ) and J2 = (j2 ).
Then J1 J2 ⊆ J, Z(J) ⊆ Z(J1 ) ∪ Z(J2 ). Since neither J1 nor J2 is contained in J, Z(J) is
not contained in either Z(J1 ) nor Z(J2 ). Therefore Z(J) = (Z(J) ∩ Z(J1 )) ∪ (Z(J) ∩ Z(J2 ))
expresses Z(J) as a proper union of two algebraic subsets, a contradiction.
Exercise 1.10. Find rings to represent the following figures (see textbook). The first rep-
resents the union of a circle and a parabola in the plane, and the second shows the union
of two skew lines in 3-space. (You may use the Nullstellensatz to prove that your answer is
right.)
For the first figure, we have the ring k[x, y]/(x2 + (y − 1)2 − 1, x2 − 2y). For the second
figure, note that a line in 3-space is parametrized by (x, y, z) = (at + x0 , bt + y0 , ct + z0 ). At

6
least one of the constants a, b, c is nonzero, let’s assume that it is a. Then our parametrization
is equivalent to the equations

0 = b(x − x0 ) − a(y − y0 ), z = c(x − x0 ) − a(z − z0 ).

Let L1 (x, y) and L2 (x, z) be the right-hand sides of the above equations. For the other skew
lines, we obtain similar equations L01 (x, y) and L02 (x, z). Note that it might be the case
that it is a variable other than x that appears in both linear polynomials. Then the ring is
k[x, y]/(L1 , L2 , L01 , L02 ).
Exercise 1.11. When we draw pictures representing algebraic sets, we often draw the same
picture for for all ground fields k, although by rights it generally represents best the case
k = R. In fact, we are usually interested in the case of an algebraically closed field k, such
as k = C, where the Nullstellensatz applies. The main way in which the pictures can be
misleading is illustrated by the following examples.
a. If k = R then the ring k[x, y]/(x2 + y 2 − 1) ∩ (y − 2 − x2 ) corresponds to the union of
a circle and a parabola. If k = C, show that there are four points in the intersection
of these two components. Show that there is a bijection given by polynomial maps
between the parabola and the line x = 0. Show further that “projection from the north
pole” gives a bijection (given by rational functions) between the circle minus one point
and the line minus two points. (Can you find such a bijection between the circle and
the line minus one point?)
b. Show that the polynomial f (x, y) = y 2 − (x − 1)x(x + 1) is irreducible over any field
k. Thus X = Z(f ) ⊆ A2 is an irreducible algebraic set. This is not so obvious from
the real picture, which approximately resembles the following image (see textbook).
a. To find intersections between the components, we want to find complex solutions to
the system of equations x2 + y 2 = 1 and y = 2 + x2 . Using the second equation to replace y
in the first, we obtain a quadratic equation in x2 . Using the quadratic formula easily yields
four complex values for x. A bijection between the parabola and the line is easily found via
(x, y) 7→ (x, 0) (this gives a bijection to the y = 0 axis, but we can easily obtain a bijection
to the x = 0 axis by swapping coordinates). For the next part, the usual stereographic
projection formulas work, i.e.

z2 − 1
 
x 2z
(x, y) 7→ with inverse z 7→ , .
1−y z2 + 1 z2 + 1

Interestingly, since we are working over C, we get that z = ±i are missing from the image
of the stereographic projection (unlike in the real case where all of R is hit). To improve
things, we can make the substitution z = w1 + i. What happens here after this substitution
is that the point z = i gets sent to infinity, reducing the number of missing points on the
complex plane. It is an easy computation to show that our new map only misses the point
w = 0.
b. Suppose that f (x, y) = g(x, y)h(x, y) is a factorisation of f . If degy g = 2, then we
have g(x, y) = (a(x)y 2 +b(x)y+c(x)) and h(x, y) = d(x) for some polynomial a, b, c, d ∈ k[x].
This implies that a(x)d(x) = 1, meaning that h(x, y) is a unit. On the other hand, suppose
that degy g = degy h = 1. Then g(x, y) = a(x)y + b(x) and h(x, y) = c(x)y + d(x). This
implies that a(x)c(x) = 1 so we may assume that g(x, y) = y + b(x) and h(x, y) = y + d(x).
Multiplying g and h together and comparing with f , we have c(x) = −d(x) and c(x)2 =
(x − 1)x(x + 1), which is impossible for degree reasons. We conclude that f is irreducible

7
(over any field since we never specified k). Since irreducible elements generate prime ideals,
we see that the disjoint union drawn in the textbook is actually irreducible.
Exercise 1.12. Find equations for a parabola meeting a circle just once in the complex
plane, represented by the following picture (see textbook).
The circle is given by x2 + y 2 = 1 and the parabola by y = 12 x2 − 1. To see that they
meet just once, observe that the norm squared of the points on the parabola is 14 x4 + 1,
which equals 1 only at the point (x, y) = (0, −1).
√
Exercise 1.13. Suppose that I is an ideal in √ a commutative ring. Show that if I is finitely
generated, then for some integer N we have ( I)N ⊆ I. Conclude that in a Noetherian ring
the ideals I and J have the same radical if and only if there is some integer N such that
I N ⊆ J and J N ⊆ I. Use the Nullstellensatz to deduce that if I, J ⊆ S = k[x1 , . . . , xr ] are
ideals and k is algebraically closed, then Z(I) = Z(J) if and only if I N ⊆ J and J N ⊆ I for
some N .
√
Let I = (f1 , . . . , fm ). We will prove the first claim by induction on m. Let Jk =
(f1 , . . . , fk ). We will prove by induction on k that there is some Nk such that JkNk ⊆ I. If
k = 1, then since there is some N1 such that f1N ∈ I, then J1N1 = (f1 )N1 ⊆ I. Now assume
Nk−1 √
k > 1. By induction there is an integer Nk−1 such that Jk−1 ⊆ I. By definition of I
there is some l such that fkl ∈ I. Let α ∈ Jk be arbitrary; then we may write α = β + rfk ,
where β ∈ Jk−1 and r is an element of the ring containing I. By the binomial √ theorem,
αNk−1 +l = (β + rfk )Nk−1 +l ⊆ I, so we can let Nk = Nk−1 + l. Since Jm = I, this proves
the claim. √ √
Now, let R be a Noetherian √ N ring and let √ I, N
J / R be such that I = J. Then there
is an integer N such that ( I) ⊆ I and ( J) ⊆ J (the first part guarantees a different √
integer for I and J, then √ we can √just take N to be the maximum of the two). Since J ⊆ J,
we have that J n ⊆ ( J)N = ( I)N ⊆ I. Similarly, we have that I N ⊆ √J. Conversely,
suppose that there is an integer N such that J N ⊆ I and I N ⊆ J. Let √ α ∈√ I. Then there
is some M such that αM ∈ I. But √ then
√ α MN
∈ J, implying that I ⊆ J. The reverse
inclusion is proved similarly, so I = J.
√Finally, √suppose the I, J / S and Z(I) = Z(J). By the Nullstellensatz, this is equivalent
to I = J, which is equivalent to I N ⊆ J and J N ⊆ I for some integer N as we just
showed.
Exercise 1.14. Not all interesting graded rings are generated by forms of degree 1, as are
the homogeneous coordinate rings of projective varieties. For example, k[x, y]/(y 2 − x3 ), the
ring corresponding to the cusp becomes a graded ring if we give x degre 2 and y degree 3.
Prove that the map k[x, y] → k[t] sending x to t2 and y to t3 induces an isomorphism
k[x, y]/(y 2 − x3 ) ∼
= k[t2 , t3 ] ⊆ k[t].
The map clearly has image k[t2 , t3 ] and it is clear that (y 2 −x3 ) is contained in its kernel,
so it suffices to show that every element of the kernel is a multiple of y 2 − x3 . To this end,
let f (x, y) ∈ k[x, y] be a polynomial such that f (t2 , t3 ) = 0. Viewing f as a polynomial in
y over the ring k[x], we have that
f (x, y) = q(x, y)(y 2 − x3 ) + r(x, y),
where degy r = 0 or 1. If degu r = 0, then r depends only on x. Plugging in x = t2 and
y = t3 , we have that r(t2 , t3 ) = 0. But since r only depends on x, we have that r is zero as

8
a polynomial in t2 , which means that r is just the zero polynomial. If degy r = 1, then we
have that r(x, y) = a(x)y + b(x) for some polynomials a and b. Then 0 = a(t2 )t3 + b(t2 ). If
a is nonzero, then a(t2 )t3 contains only odd-degree terms, while b(t2 ) has only even degree
terms. Thus, a must be zero, implying that b = 0. We conclude that y 2 − x3 divides f , as
desired.

Exercise 1.15. Here is an example of the simplification brought to geometry by the idea of
projective space. As the reader probably remembers from high school, a conic in the affine
real plane R2 (that is, the locus defined by a quadratic equation in two variables x and y
with real coefficients) belongs to one of the following eight types:
a. The empty set (as with x2 + y 2 + 1 = 0)
b. A single point (as with x2 + y 2 = 0)
c. A line (x2 = 0)
d. The union of two coincident lines (xy = 0)
e. The union of two parallel lines (x(x − 1) = 0)
f. A parabola (y − x2 = 0)
g. A hyperbola (xy − 1 = 0)
h. An ellipse (x2 + 2y 2 − 1 = 0).
Any two examples of one of these types differ only by an invertible linear transformation of
the coordinates.
a. Show that in the complex affine plane C2 there are only five types of loci defined by
equations of degree 2: Types a and b disappear, and types g and h coincide.
b. Show that in the complex projective plane P2 (C) there are only three types of loci
represented by quadratic equations; they are represented by types c, d, and h on the above
list. More generally, there are exactly n types of nonzero
P quadratic forms in n variables,
classified by rank (where the rank of a quadratic form i<j aij xi xj is defined to be the rank
of the symmetric matrix (aij )).
c. Show that the different types in part a correspond to the relative placement of the conic
and the line at infinity, in the sense that a parabola is a rank-3 conic tangent to the line
at infinity, while an ellipse/hyperbola is a rank-3 conic meeting the line at infinity at two
distinct points. The classification over the real numbers may be recovered from the position
of these points: A real rank-3 conic meeting the line at infinity in two points is a hyperbola if
the points are real, and an ellipse if the points are nonreal (they are then conjugate complex
points). If the affine plane is represented by points (x, y, z) ∈ P2 with z = 1, the ellipse is
a circle if and only if it meets the line at infinity in the points (1, i, 0) and (1, −i, 0), the
“circular points at infinity.”
a. For every fixed value of y ∈ C, any quadratic f (x, y) ∈ C[x, y] has at least one root
(in x), so the locus of f in the affine plane C2 is necessarily infinite. This shows that there
are no conics of type a or b in the affine plane. I will not show that types c, d, e, and f are
distinct in the complex case (one would have to show that the different families of equations
are not related by any linear transformation of the variables). To see that√g and h are in
√ of the same type over C, note that we can replace x with x + i 2y and y with
fact the
x − i 2y to go from the equation xy − 1 = 0 to the equation x2 + 2y 2 − 1 = 0.
b. A conic in P2 (C) is the zero-set of a quadratic form in 3 variables, so classifying
the conics in P2 (C) is the same as classifying the homogeneous forms in 3 variables. More
generally, we classify all quadratic forms in n variables as follows. To every quadratic form
A : Cn → C, there is a unique symmetric matrix A such that A(x) = xt Ax (similarly,
to each symmetric matrix we can associate a quadratic). We will assume that A has real

9
coefficients so that A is orthogonally diagonalizable; hence, there is a real orthogonal matrix
Q such that Qt AQ is diagonal. Moreover, there is a diagonal complex matrix D such that
Dt Qt AQD is a diagonal matrix whose only nonzero entries are 1’s. Thus, quadratic forms
in n variables are completely classified by the rank of their associated symmetric matrices
(if we are allowed to re-scale coordinates in C, otherwise they are classified by the number of
positive and negative eigenvalues their matrices have). There are then three types of conics
in P2 (C), and they take the forms x2 = 0 (type c), x2 + y 2 = 0 (type d, transform to xy = 0
using x, y 7→ x + iy, x − iy), and x2 + y 2 + z 2 (type h, by setting z = 1, we obtain a type h
conic in the affine plane, sitting inside P2 (C)).
c. A parabola is given by a homogeneous equation of the form yz = x2 (the general
case is not much different). Since the plane at infinity is given by points with z = 0, we
get that x = 0, therefore the parabola meets the line at infinity only at the point [0 : 1 : 0].
Intuitively, the parabola lies on one side of the line at infinity, and is therefore tangent to
it. The hyperbola is given by the homogeneous equation xy − x2 = 0, so it meets the line at
infinity at the point with xy = 0. In homogeneous coordinates, these are the points [1 : 0 : 0]
and [0 : 1 : 0].

Exercise 1.16. a. Let I be a homogeneous ideal in S = k[x0 , . . . , xr ], and suppose that

the projective algebraic set corresponding to I is nonempty. Let Y ⊆ k r+1 be the affine
algebraic set associated to I. Show that Y is a union of one-dimensional subspaces of
k r+1 , and that these one-dimensional subspaces are precisely the points of the projective
algebraic set associated to I.
b. Show also that if k is an infinite field and X ⊆ Ar (k) is a union of lines through the
origin, then I(X) is a homogeneous ideal.
a. Let x ∈ Y . Then f (x) = 0 for every f ∈ I. Moreover, f (λx) = 0 for every λ ∈ R
and every homogeneous f ∈ I. Since I is homogeneous, it is generated by homogeneous
polynomials. Therefore, every polynomial in I is the sum of homogeneous polynomials,
which implies that f (λx) for every λ ∈ R. We conclude that Y is the union of lines
through the origin. Let L be a one-dimensional subspace of k r+1 contained in Y . Then
every polynomial vanishes on this line, and in particular the homogeneous ones vanish on L.
Thus, L corresponds to a point in the projective algebraic set associated to I. Conversely,
let p be a point of the projective algebraic set associated to I. Then every homogeneous
polynomial vanishes on the line corresponding to p. Since I is homogeneous, we have that
this line is in Y .
b. We begin by proving that if a polynomial vanishes on a line through the origin, then
each of its homogeneous components must vanish on that line. Let f ∈ k[x1 , . . . , xr ] be
a polynomial vanishing on the 1-dimensional subspace spanned by the point (c1 , . . . , cr ).
Write
f (x1 , . . . , xr ) = a0 (x1 , . . . , xr ) + · · · + ak (x1 , . . . , xr ),
where ai is the degree i homogeneous component of f . Letting αi = ai (c1 , . . . , cr ), we have
that the polynomial f˜(x) = α0 + α1 x + · · · + αk xk ∈ k[x] vanishes for all x ∈ k. Since k is
an infinite field, we must have that f˜ is the zero polynomial. Therefore αi = 0. Since the ai
are homogeneous, we conclude that they vanish on the entire line spanned by (c1 , . . . , cr ).
Now, let f ∈ I(X). By the above claim, every homogeneous component of f vanishes
on X and is therefore contained in I(X). Thus, I(X) is generated by the homogeneous
components of its polynomials and is therefore a homogeneous ideal.

10
Exercise 1.17. Let I ⊆ k[x1 , x2 , x3 ] be the ideal (x21 + x2 , x21 + x3 ), and let X ⊆ A3 be
the affine algebraic set Z(I). Let X ⊆ P3 be the projective closure of X. Show that the
homogeneous ideal I(X) is not generated by the homogenizations of x21 + x2 and x21 + x3 .
I(X) is generated by the homogenization of all the elements of I(X). We first check
that I is a radical ideal, so that I(X) = I by the Nullstellensatz. To see this, note that

k[x1 , x2 , x3 ] k[x1 , x2 , x3 ] ∼ k[x1 , x2 ] ∼

= 2 = 2 = k[x],
(x21 2
+ x2 , x1 + x3 ) (x1 + x2 , x2 − x3 ) (x1 + x2 )

where the last isomorphism is induced by the map k[x1 , x2 ] → k[x] sending x1 7→ x and
x2 7→ −x2 . Since k[x] is an integral domain, we have that I is a prime ideal, and in particular
a radical ideal. Therefore, I(X) = I.
Now, I(X) is generated by the homogenizations of all the elements in I. Therefore,
x2 − x3 ∈ I(X). Now consider the ideal J = (x21 + x0 x2 , x21 + x0 x3 ) generated by the
homogenizations of the generators of I. No polynomial in J will contain a term of degree
1, so J 6= I(X).
Exercise 1.18. Let k be a field. Compute the Hilbert function and polynomial for the ring

k[x, y, z, w]/(x, y) ∩ (z, w)

corresponding to the disjoint union of two lines in projective 3-space. Compare these to the
filbert function and polynomial of the ring corresponding to one projective line, k[x, y].

We begin by showing that (x, y) ∩ (z, w) = (x, y)(z, w). Let f (x, y, z, w) ∈ (x, y) ∩
(z, w). Then we can qrite f = ax + by = cz + dw for some a, b, c, d ∈ k[x, y, z, w]. Then
f (x, y, 0, 0) = a(x, y, 0, 0)x + b(x, y, 0, 0)y = 0. Let ã(x, y, z, w) = a(x, y, z, w) − a(x, y, 0, 0)
and b̃(x, y, z, w) = b(x, y, z, w) − b(x, y, 0, 0). Note that ã and b̃ do not have any terms of
the form xn y m (in other words, every term involves z or w). Moreover, f = ãx + b̃y, so
f ∈ (x, y)(z, w) and therefore (x, y) ∩ (z, w) ⊆ (x, y)(z, w). The reverse inclusion is obvious.
The map φ : k[x, y, z, w] → k[x, y] ⊕ k[z, w], f 7→ (f (x, y, 0, 0), f (0, 0, y, z)) is surjective
and has kernel (x, y)(z, w), so the above shows that k[x, y, z, w]/(x, y) ∩ (z, w) ∼ = k[x, y] ⊕
k[z, w]. It is also easy to see that this is actually an isomorphism of k[x, y, z, w]-modules,
where the actions are

f · g = f g and f · (g(x, y), h(z, w)) = (f (x, y, 0, 0)g(x, y), f (0, 0, z, w)h(z, w)).

Now, k[x, y]⊕k[z, w] is clearly graded by degree, and {xi y j , z i wj : i+j = n} is a basis for the
degree-n polynomials, with n ≥ 1. Then the Hilbert function for M = k[x, y, z, w]/(x, y) ∩
(z, w) is 
2n + 2 if n ≥ 1

HM (n) 1 if n = 0 .

0 otherwise


We immediately get that the Hilbert polynomial is PM (n) = 2n + 2.

Things are simpler for the k[x, y]-module k[x, y]. Everything is graded by degree. The
{xi y j : i + j = n} is a basis for the degree-n polynomials, with n ≥ 0. The Hilbert function
is then n + 1 for n ≥ 0, and equal to zero otherwise. The Hilbert polynomial in this case is
n + 1.

11
Exercise 1.19. Let k be a field. Let I ⊆ k[x, y, z, w] be the ideal generated by the 2 × 2
minors of the matrix  
x y z
,
y z w
that is, I = (yw − z 2 , xw − yz, xz − y 2 ).
Show that R = k[x, y, z, w]/I is a finitely generated free module over S = k[x, w]. Exhibit
a basis for R as an S-module. Show that there is a ring homomorphism R → k[s, t] such
that x 7→ s3 , y 7→ s2 t, z 7→ st2 , w 7→ t3 . Use the basis you constructed to show that it is a
monomorphism. Conclude that I is prime. From the rank of R as a free S-module, and
the degrees of the generators, deduce the Hilbert function of R. Show that R is not finitely
generated as a module over k[x, y].

It is not hard to show that the images of 1, y, and z in the quotient span R over S.
I’m not sure how to prove that these elements are linearly independent however. There is
certainly a map k[x, y, z, t] → k[s, t] satisfying the requirements of the question. It suffices
to show that I is in the kernel of this map, and this is a straightforward verification. To
check that the map is a monomorphism, let f1 (x, w) + f2 (x, w)y + f3 (x, w)z be an element
of the kernel. Then f1 (s3 , t3 ) + f2 (s3 , t3 )s2 t + f3 (s3 , t3 )st2 = 0. Considering the degrees of
s and t in this expression modulo 3 shows that f1 = f2 = f3 = 0. We conclude that the
map is a monomorphism. Since R is then isomorphic to a subring of an integral domain, it
is itself an integral domain. Hence, I is prime.
Since the generators all have degree 3 in k[s, t], we can just give them degree 1 to simplify
things, and grade R by degree over k[x, y, z, w]. From our basis, every element of R can be
written in the form f1 (x, w) + f2 (x, w)y + f3 (x, w)z for unique polynomials f1 , f2 , f3 . There
are n + 1 monomials in two variables of degree n. Thus, HR (n) = (n + 1) + n + n = 3n + 1.
Assume that R is finitely generated as a k[x, y]-module. Then it has a basis p1 , . . . , pk
of polynomials, and assume that maxi (degw pi ) = n. Since the basis elements span R over
k[x, y], we can write wn+1 as a linear combination of the basis elements, which will be a
polynomial p(x, y, z, w) with w-degree at most n. Then wn+1 − p ∈ I, which is clearly
impossible.

Exercise 1.20. Given a number s0 , find an example of a graded k[x1 , . . . , xr ]-module M

generated by elements of degree 0 for which the function HM (s) is not equal to the Hilbert
polynomial PM (s) for any s < s0 . If you find this too easy, can you find torsion-free
k[x1 , . . . , xr ]-modules of this sort?

Consider the ideal M = (xm y n , xn y m ), with n − m very large, viewed as k[x, y]-module
graded by degree. We shift the grading by n + m so that the generators have degree 0.
Every element of M is written in the form f (x, y)xm y n + g(x, y)xn y m . Thus, for degrees
d less than n − m, the dimension over k is 2d + 2. However, in larger degrees, there are
relations between the two generators, so the dimension will be less than 2d + 2. Therefore,
the Hilbert function of M does not agree with the Hilbert polynomial in dimensions less
than n − m.
Exercise 1.21. Consider the subring T ⊆ Q[n] of rational polynomials that take integral
values at sufficiently large integers. T is of interest to us because it contains all the Hilbert
polynomials discussed in
this chapter. The ring T obviously contains Z[n], but it is larger:
It contains things like n2 = (n2 − n)/2.

12
 a. Let F (n) be a function defined for sufficiently large integers n, and set G(n) := F (n +
1) − F (n). Show that F (n) ∈ Q[n] if and only if G(n) ∈ Q[n], and that if these
conditions are satisfied then deg F = 1 + deg G.
b. Show by induction on the degree that T is a free Abelian group with basis given by the
functions  
n
Fk = = n(n − 1) · · · (n − k + 1)/k! 0 ≤ k < ∞,
k
where Fk is a polynomial function in n of degree k.
c. Although Q ⊗Z T = Q[n], the ring T itself is not finitely generated as an algebra over
Z; show that T is not even Noetherian. We shall meet T again as a free divided power
algebra in Appendix A2.
a. If F (n) ∈ Q[n], then clearly G(n) ∈ Q[n]. Conversely, suppose that G(n) ∈ Q[n].
Let n0 be the least integer for which F is defined. Since F (n) is a polynomial if and
only if F (n + c) is a polynomial for some constant c, we may assume that n0 = 1. Then
F (n) = F (0) = (F (n) − F (n − 1)) + · · · + (F (2) − F (1)) = G(n − 1) + · · · + G(1). To show
that this is a polynomial, it suffices to show that expressions of the form 1k + · · · + nk are
polynomials in n. Here is a proof of this using induction on k: note that
n
X n
X
1+ (i + 1)k+1 = (n + 1)k+1 + ik+1 .
i=1 i=1

ik+1 on each side, we have

P P k
Foiling on the left-hand side and cancelling i as a polyno-
mial in n by induction on k. This proof also gives the desired result about degrees.
b. Linear independence of the Fi is obvious, so we only need to show that they span
T over Z. It is clear that F0 = 1 is forms a basis for the degree 0 polynomials in T . Now
suppose that {F0 , . . . , Fk } is a basis for the polynomials of degree k in T . It is clear that
we can write any degree k + 1 polynomial f as a linear combination
   
n n
f (n) = a0 + · · · + ak+1 .
0 k+1
Note that    
n n
f (n + 1) − f (n) = a1 + · · · + ak+1
0 k
is polynomial in T written in terms of the basis, so we must have that a1 , . . . , ak+1 ∈ Z by
induction on the degree. It then follows that a0 must be an integer for f to be in T .
c. We will show that the chain
(F1 ) ⊆ (F1 , F2 ) ⊆ (F1 , f2 , F3 ) ⊆ · · ·
never stabilizes. Suppose that it did. Then for some large enough prime p, we would have
(F1 , . . . , Fp−1 ) = (F1 , . . . , Fp ). Thus, we can write
     
n n n
= a1 (n) + · · · + ap−1 (n) ,
p 1 p−1
where the ai (n) are Z-linear combinations of the basis elements Fj found in part b. Can-
celling an n from each side and multiplying by p!, we have
(n − 1) · · · (n − (p − 1)) = p!a1 (n) + · · · + pap−1 (n)(n − 1) · · · (n − (p − 2)).

13
Since ai (0) is always an integer, evaluating both sides at n = 0 yields that (p − 1)! is a
multiple of p, which is a contradiction. We conclude that T is not Noetherian.
Exercise 1.22. Let R = k[x]. Use the structure theorem for finitely generated modules over
a principal ideal domain to show that every finitely generated R-module has a finite free
resolution.
Let M be a finitely generated R-module. The structure theorem says that

M∼
= Rf ⊕ R/(d1 ) ⊕ · · · ⊕ R/(dn )

for some natural numbers n and f , where (d1 ) ⊇ · · · ⊇ (dn ) is a decreasing sequence of
proper ideals in R. It is an easy consequence of this theorem that submodules of a free
module over a principal ideal domain are free. Suppose M is generated by the elements
f1 , . . . , fk , and let π : Rk → M be the map that sends a generator of the ith copy of R to
fi . Then ker π is free, being a submodule of the free R-module Rk , so we have the finite
free resolution
π
0 → ker π ,→ Rk −
→ M → 0.
Exercise 1.23. Let R = k[x]/(xn ). Compute a free resolution of the R-module R/(xm ), for
any m ≤ n. Show that the only R-modules with finite free resolutions are the free modules.
R/(xm ) has the infinite free resolution

xn−m xm xn−m xm
· · · −−−−→ R −−→ R −−−−→ R −−→ R → R/(xm ) → 0,

where the labels on the arrows indicate that the map is multiplication by that element.
To prove the second part of the problem, we first show that any linearly independent
L
set in a free R-module M can be extended to a basis for M . To this end, M = I R and
let X ⊆ M be a linearly independent, where L I is an arbitrary index set. We claim that the
set Y := {(fα (0))α∈I : (fα )α∈I ∈ X} ⊆ I k is linearly independent: let λ1 , . . . , λm ∈ k be
such that
λ1 fα1 (0) + · · · + λm fαm (0) = 0
for some fα1 , . . . , fαm ∈ X. Then

λ1 xn fα1 + · · · + λm xn fαm = 0.

By linear independence of X we have thatL λi xn = 0, which implies that λi = 0, for every

i. Thus, Y is linearly independent. Since I k is a vector space, Y can be extended to a
basis B.
It is then not too hard to see that X ∪ (B − Y ) is a basis for M , where we view
elements of B − Y is vectors of constant polynomials. To show that it is spanning, note that
xn (X ∪ (B − Y )) is a basis for the elements of M whose components are all polynomials of
the form xn f (x) (that is, they have a single term of degree n or are just zero). Assuming
we can write all polynomials of the form xj+1 v ∈ M as linear combinations of elements in
j
X ∪ (B − Y ), we show that we can L do the same for elements of the form x v as follows:
use the fact that B is a basis for I k to write the homogeneous component of degree j of
xj v as a linear combination of vectors in X ∪ (B − Y ). Once we have done this, add the
appropriate vector of the form xj+1 w to this vector to obtain xj v. Continuing inductively
shows that X ∪ (B − Y ) spans M .

14
 We now show that X ∪ (B − Y ) is linearly independent. Let f1 v1 + · · · + fm vm = 0
for some fi ∈ R and vi ∈ X ∪ (B − Y ). Multiplying both sides by xn , we see that the
constant term of each fi is zero. Then multiplying the equation by xn−1 , we get that the
linear term of each fi is zero. Continuing in this way, we see that each fi = 0. This proves
that X ∪ (B − Y ) is linearly independent, and proves our claim.
Now, let
0 → Fm → · · · → F1 → F0 → M → 0
be a finite free resolution of a (not necessarily free) R-module M . Denote the maps ∂i : Fi →
Fi−1 and ∂0 : F0 → M . We prove by reverse induction on i that im ∂i is a free submodule
of Fi . Since ∂m is an injection, the base case is true. Now suppose that im ∂i+1 ⊆ Fi
is a free submodule. Since a basis for im ∂i+1 can be extended to a one for Fi , we have
that Fi / im ∂i+1 is free. But Fi / im ∂i + 1 = Fi / ker ∂i ∼
= im ∂i , which proves our claim by
induction. Thus M is free (note that the induction still works at i = 0 without assuming
that M is free, all that is required is that F0 be free).
Exercise 1.24. In the text we defined the Zariski topology on an algebraic set over any
algebraically closed field k. We may identify X with the set max-Spec A(X). The subset
Z(I) is identified with the set of maximal ideals containing I. This suggests a way of
defining a topology on the set of maximal ideals of any ring. The corresponding idea can
also be applied to the set of all prime ideals, and it turns out to be even more useful there.
These ideas were first pursued by Oscar Zariski, and the resulting topology bears his name.
Definition. Let R be any ring. The subsets of Spec R of the form

Z(I) := {p a prime ideal of R : p ⊇ I},

for ideals I of R are called Zariski-closed subsets. When there is no danger of confusion,
we shall simply call them closed subsets.
a. Prove that finite unions and arbitrary intersections of closed subsets are closed, and
therefore the closed subsets define a topology, called the Zariski topology, on Spec R.
The induced topology on the subset max-Spec R is also called the Zariski topology.
If k = R or C (or some other topological field), then we have two topologies on k n :
the topology induced from the topology of k, called the classical topology, and the
Zariski topology. The Zariski topology has many fewer closed sets than the classical
topology.
b. Suppose for simplicity that k is an algebraically closed field, in the Zariski topology on
A1 (k) (that is, on the maximal ideals of k[x]) show that the open sets are exactly the
complements of finite sets. In particular this topology is not Hausdorff. Show that the
Zariski topology on An (k) = k n is not the product topology, even for n = 2.
c. We define a distinguished open set of Spec R to be an open set of the form U (f ) :=
{p a prime ideal of R : f ∈ / p} for some f ∈ R. Show that the distinguished open
sets form a basis for the Zariski topology, in the sense
S that every open set is a union
of distinguished open sets. Show that Spec R = i U (fi ) for some collection fi of
elements of R if and only if the ideal generated by all the fi is the unit ideal (1).
d. Show that if R is any ring then Spec R is compact in the Zariski topology (that is,
every open covering has a finite refinement).
a. To show that finite unions of closed sets are closed, it is enough to show that a
union of two closed sets is closed. To this end, let I, J / R be ideals. Then we claim that

15
Z(I) ∪ Z(J) = Z(IJ). If p ∈ Z(I) ∪ Z(J), then assume p ∈ Z(I). Then p ⊇ I ⊇ IJ, which
shows that Z(I) ∪ Z(J) ⊆ Z(IJ). Conversely, suppose that p ∈ Z(IJ), so p ⊇ IJ. If p
contains neither I nor J, then there are elements i ∈ I − p and j ∈ J − J − p. But then
ij ∈ p, contradicting the fact that p is prime. Therefore p contains either I or J, proving
that Z(I) ∪ Z(J) = Z(IJ). T
Let {Iα }α∈A be an arbitrary collection of ideals of R. Then p ∈ α∈A Z(Iα ) if and
only
T if p contains each Iα if and only if p contains the ideal generated by the Iα . Hence,
α∈A Z(Iα ) = Z((Iα )α∈A ), so arbitrary intersections of closed sets are closed. Note that
this didn’t use the fact that the elements of Spec R are prime.
b. Let I / k[x] be an ideal. Since k[x] is a principal ideal domain, I = (f ) for some
polynomial f . Moreover, f factors as (x − a1 ) · · · (x − an ) where n = deg f and ai ∈ k, since
k is algebraically closed. Thus, if m / k[x] is maximal, then m ⊇ I if and only if m = (x − ai )
for some i. Hence, Z(I) = {(x − a1 ), . . . , (x − an )}, which proves that the closed sets are
all finite sets of points, or all of A1 (k). Equivalently, all open sets are complements of finite
sets. Let p, q ∈ A1 (k) be any two points and let U, V be any open sets with p ∈ U and
q ∈ V . Since k is algebraically closed, it is infinite, and since U and V are complements of
finite sets, they must then meet in infinitely many points. Hence, the Zariski topology on
A1 (k) is not Hausdorff.
We will show that the topology on An (k) = k n is not the product topology by showing
that the diagonal {(x, . . . , x) : x ∈ k} ⊆ k n is closed in the Zariski topology but not in
the product topology. Consider the ideal I = (x1 − x2 , . . . , x1 − xn ) ⊆ k[x1 , . . . , xn ]. Let
m = (x1 − a1 , . . . , xn − an ) be an arbitrary maximal ideal of k[x1 , . . . , xn ] containing I.
This occurs if and only if a1 = · · · = an . This proves that the diagonal is closed in the
Zariski topology. Now suppose that the diagonal is closed in the product topology. Then
the complement of the diagonal would be open, meaning there is a product neighbourhood
that is disjoint from the diagonal. But since every product open set contains all but finitely
many elements of k, every product set must contain an element of the diagonal. Hence,
An (k) is not the n-fold topological product of A1 (k).
c. Let U be an arbitrary open set, and let p ∈ U ; we will show that p is contained in
some U (f ) ⊆ U , which implies that U is the union of distinguished open sets. Since U is
the complement of Z(I) for some ideal I, we have that there is an element f ∈ I − p. Then
clearly p ∈ U (f ). Moreover, if q ∈ U (f ), then q does not contain I, which is equivalent to
q ∈ U . Thus, p ∈ U (f ) ⊆ U , as desired. Hence, the distinguished open sets form a basis for
the Zariski topology. S
Suppose that Spec R = i U (fi ) for some collection of fi . Consider the ideal I generated
by the fi . If I 6= R, then it is contained in some maximal ideal p, which S is also prime. But
the fi ∈ p for every i, contradicting the assumption that Spec R = i U (fi ). Conversely,
suppose that the fi generate all of R. Let p be any prime ideal. If p contained every fi , then
it would be the entire ring R, contradicting S the definition of a prime ideal. Hence fi ∈ / p for
some i, which implies that Spec R = i U (fi ).
d. Since the distinguished open sets form a basis for the Zariski topology, it is enough to
consider open covers of Spec R by distinguished open sets. Let {U (fα )}α∈A be an open cov-
ering of Spec R. By the part c, the fi generate the unit ideal, which means that we can write
1 = a1 fα1 + · · · + an fαn . Thus, fα1 , . . . , fαn generate the unit idea, so {U (fα1 ), . . . , U (fαn )}
is a finite subcover by part c. Thus, Spec R is compact.
Exercise 1.25. Let X be a compact Hausdorff space and R = C(X) be the ring of con-
tinuous real-valued functions on X. Let µ : X → max-Spec R be the map taking a point

16
x ∈ X to the maximal ideal mx of all continuous functions vanishing at x. Prove that µ is
a homeomorphism, so that X can be reconstructed algebraically from C(X), as follows:
a. µ is surjective: Let m be a maximal ideal of C(X). We wish to prove that m = mx
for some x. Let V = V (m) be the set of common zeros of the functions in m. If V
is empty, then for each x ∈ X there exists fx ∈ m such that fx (x) 6= 0. Since fx is
continuous, there is an open neighbourhood U (x) of x such that fx does not vanish on
U (x). By compactness, X is the union of finitely many of these neighbourhoods, say

X = U (x1 ) ∪ · · · ∪ U (xn ).

Use these ideas to construct a function f ∈ m that does not vanish anywhere on X.
Derive a contradiction.
It now follows that m ⊆ mx for some x. Since m is maximal, the ideas are equal.
b. µ is injective: Use Urysohn’s lemma to show that if x 6= y, then there is a continuous
function vanishing at x but not at y.
c. µ is a homeomorphism: A subbasis for the topology of X is given by the sets

Uf = {x ∈ X : f (x) 6= 0}, f ∈ C(X)

while a subbasis for the topology of max-Spec R is given by the sets

Vf = {m ∈ max-Spec R : f ∈
/ m}, f ∈ C(X).

Show that µ(Uf ) = Vf .

a. For each i = 1, . . . , n, let gi ∈ C(X) be defined by
(
fxi (x) if fxi (x) ≥ 0
gi (x) = .
0 otherwise

Then gi is clearly well defined, and it is continuous by the gluing lemma. Then gi fxi ≥ 0
on all of X, and is strictly positive on U (xi ). It follows that f := g1 fx1 + · · · + gn fxn is
positive on all of X. Note also that f ∈ m and 1/f ∈ C(X). Thus, m contains a unit,
implying that m = C(X), contradicting the definition of maximality. We conclude that the
functions in m must have a common zero, say x. Then m ⊆ mx . Since mx is not equal to all
of C(X) (it doesn’t contain nonzero constant functions for instance), we have that m = mx
by maximality.
b. Urysohn’s lemma states that for distinct points x, y ∈ X, there is a continuous
function f : X → R such that f (x) = 0 and f (y) = 1 (the lemma also states that the image
of f is contained in [0, 1], but that is not important here). Hence, f ∈ mx − my , implying
that mx 6= my . Hence, µ, is injective.
c. The collections of sets Uf and Vf are actually bases for the topologies on X and
max-Spec R. The collection of sets Vf is simply the restriction of the collection of distin-
guished open sets to max-Spec R, and is therefore a basis for the Zariski topology by part
c of the preceding problem. To show that the Uf form a basis, we will use a more general
form of Urysohn’s lemma, which states that in a normal space, disjoint closed sets can be
separated by continuous functions, in the sense that if A and B are disjoint closed sets,
then there is a continuous function f : X → [0, 1] such that f (a) = 0 for every a ∈ A and
f (b) = 1 for every b ∈ B. Now, X is compact and Hausdorff, so it is normal and we can
apply Urysohn’s lemma. Let U ⊆ X be an arbitrary open set, and let x ∈ U . Then {x} and

17
X − U are disjoint closed sets ({x} is closed because X is Hausdorff), so there is a function
f ∈ C(X) such that f (x) = 1 and f (y) = 0 for all y ∈ X − U . Thus, Uf ⊆ U . Since every
point of U has a neighbourhood of the form Uf , we conclude that U can be written as a
union of sets Uf , and therefore that the Uf form a basis for the topology on X.
Now,

mx ∈ µ(Uf ) ⇔ x ∈ Uf
⇔ f (x) 6= 0
⇔f ∈
/ mx
⇔ mx ∈ Vf .

Thus, µ(Uf ) = Vf . Since the Uf and Vf form bases, any open set can be written as a union
of such sets. Since preimages commute with unions, we conclude that µ and its inverse are
continuous, that is, that µ is a homeomorphism.

2 Localization
Exercise 2.1. Check that the definitions really do make R[U −1 ] into a ring and M [U −1 ]
into an R[U −1 ]-module (and thus also an R-module). Check that the map R → R[U −1 ]
sending r → r/1 is a ring homomorphism, and the map M → M [U −1 ] sending m to m/1
is a homomorphism of R-modules.

We will just check that addition and multiplication in R[U −1 ] are well-defined. Checking
the rest of the ring axioms is then trivial. Let r1 /u1 = r10 /u01 and r2 /u2 = r20 /u02 . Then
there are elements v and w of U such that

v(r1 u01 − r10 u1 ) = 0 = w(r2 u02 − r20 u2 ).

To show addition is well-defined, first note that

r1 r2 r1 u2 − r2 u1 r0 r0 r0 u0 − r0 u0
+ = and 10 + 20 = 1 2 0 0 2 1 .
u1 u2 u1 u2 u1 u2 u1 u2

Then, one can check that

vw [u01 u02 (r1 u2 − r2 u1 ) − u1 u2 (r10 u02 − r20 u01 )] = 0,

which proves that addition is well-defined. For multiplication, we have that

vw(r1 r2 u01 u02 − r10 r20 u1 u2 ) = vw(r1 r2 u01 u02 − r10 r2 u1 u02 + r10 r2 u1 u02 − r10 r20 u1 u2 )
= vw[r2 u02 (r1 u01 − r10 u1 ) + r10 u1 (r2 u02 − r20 u2 )] = 0,

proving that it is well-defined.

Exercise 2.2. Let R be a ring and let U be any subset of R. Show that R[U −1 ] is the result
of adjoining inverses of elements of U to R in the freest possible way, in the sense that

R[U −1 ] ∼
= R[{xu }u∈U ]/({uxu − 1}u∈U ).

18
 Define φ : R → R[{xu }]/({uxu − 1}), r 7→ r. Since every element of U gets mapped to
some u with inverse xu , φ(U ) consists of units. By the universal property of localization,
there is a map φ0 : R[U −1 ] → R[{xu }]/({uxu − 1}). Explicitly, φ0 (r/u) = rxu . We show
that φ0 is an isomorphism by finding its inverse.
There is a map
 
1 1
ψ : R({xu }) → R[U −1 ], f (xu1 , . . . , xun ) 7→ f ,..., .
u1 un

Clearly, ψ induces a well-defined map ψ 0 : R({xu })/({uxu − 1}) → R[U −1 ]. It is also clear
that ψ 0 and φ0 are inverses.
Exercise 2.3. Here is a collection of results that allow one to do many things in a lo-
calization without having to admit that there is any such thing. Suppose that U ⊆ R is
a multiplicatively closed subset of a ring. Show that there is a one-to-one correspondence,
preserving sums and intersections, between the ideals in R[U −1 ] and the ideals I in R such
that (I : f ) = I for all f ∈ U . Show that this correspondence respectsPthe property of being
prime. Show that for any ideal J ⊆ R we have R ∩ JR[U −1 ] = ∞
f ∈U (J : f ) where
∞
S∞ n
(J : f ) = n=1 (J : f ). Show that the ideals I ⊆ R such that (I : f ) = I are exactly
the image of the map J 7→ R ∩ JR[U −1 ]. Historically, constructions like (I : f ) were used
before localizations were defined, to accomplish the same ends.
We will show that the correspondence of ideals is given by I 7→ φ−1 (I), where φ : R →
R[U −1 ] is the natural map r 7→ r/1. By Proposition 2.2, we know that it the preserves
intersections and the property of being prime. Before proving that the correspondence, we
prove that it preserves sums: let I1 , I2 /R[U −1 ] be ideals. Let f ∈ φ−1 (I1 +I2 ). Then φ(f ) ∈
I1 + I2 . Then φ(f ) = f1 /u1 + f2 /u2 for some fi /ui ∈ Ii and ui ∈ U . Since fi /1 ∈ Ii , we have
that f ∈ φ−1 (I1 ) + φ−1 (I2 ). Conversely, it is clear that φ−1 (I1 ) + φ−1 (I2 ) ⊆ φ−1 (I1 + I2 ),
so we conclude that the correspondence also preserves sums.
Now we prove the correspondence, which boils down to proving that the ideals of the
form φ−1 (I) are exactly those satisfying the condition (φ−1 (I) : f ) = φ−1 (I) for every
f ∈ U . First, consider the ideal φ−1 (I) of R. Let f ∈ U and let r ∈ (φ−1 (I) : f ). Then
rf ∈ φ−1 (I), so (rf )/1 ∈ I. But then (1/f )((rf )/1)) = r/1 ∈ I, so r ∈ φ−1 (I). The other
inclusion is trivial, so we conclude that (φ−1 (I) : f ) = φ−1 (I). Now suppose J /R is an ideal
such that (J : f ) = J for every f ∈ U . This means that every f ∈ U is a non zero-divisor
mod J, since rf ∈∈ J implies that r ∈ (J : f ) = J. Proposition 2.2 then implies that
J = φ−1 (I) for some I.
We now tackle the next claim. Let r ∈ R ∩ JR[U −1 ]. P Then r = j(s/f ) for some
∞
j ∈ J, s ∈ R, f ∈ U , so f r ∈ J, implying that r ∈ (J : f ) ⊆ f ∈U (J : f ). Conversely, let
r ∈ f ∈U (J : f ∞ ). Then r = r1 + · · · + rn for some ri ∈ (J : fi∞ ), where fi ∈ U . Then
P
there are integers ni such that ri ∈ (J : fini ). If m = max1≤i≤n {ni }, then (f1 · · · fn )m r ∈ J,
implying that r ∈ R ∩ JR[U −1 ] as desired.
For the last claim, let J /R be an ideal such that (J : f ) = J. We showed that J = φ−1 (I)
for some ideal I / R. By Proposition 2.2, J = φ−1 (JR[U −1 ]) = R ∩ JR[U −1 ].
Exercise 2.4. Let k be a field, and let Z denote (as usual) the ring of integers. Let m, n
be integers. Describe as explicitly as possible:
a. HomZ (Zn , Zm ) and Homk[x] (k[x]/(xn ), k[x]/(xm )).
b. Zn ⊗Z Zm and k[x]/(xn ) ⊗k[x] k[x]/(xm ).
c. k[x] ⊗k k[x] (describe this as an algebra).

19
 a. We claim that HomZ (Zn , Zm ) ∼
= Zd , where d = gcd(m, n). Let φ ∈ HomZ (Zn , Zm ) be
determined by φ(1) = m/d. We claim that

Zd → HomZ (Zn , Zm ), k 7→ kφ

is an isomorphism. To prove surjectivity, we need to show that every map in HomZ (Zn , Zm )
is a multiple of φ. Any homomorphism ψ ∈ HomZ (Zn , Zm ) is determined by ψ(1) = k, so it
suffices to determine the allowed values of k. We require that 0 = ψ(0) = ψ(n) = nψ(1) = nk
modulo m, which implies that m divides nk, which in turn implies that m/d divides k.
Hence, ψ is indeed a multiple of φ. is an isomorphism. To prove injectivity, suppose that
kφ is the zero map. Then km/d = 0 modulo m, which implies that k/d is an integer, which
is equivalent to k = 0 modulo d. Thus, Ψ is injective and therefore an isomorphism.
To describe Homk[x] (k[x]/(xn ), k[x]/(xm )), there are two cases to consider. First assume
that m ≥ n. Again, every homomorphism φ ∈ Homk[x] (k[x]/(xn ), k[x]/(xm )) is completely
determined by φ(1). One can easily check that the map

k[x] → Homk[x] (k[x]/(xn ), k[x]/(xm )), g(x) 7→ φg ,

where φg (1) = xm−n g(x), is a well-defined homomorphism of k[x]-modules with kernel (xn ).
Therefore, Homk[x] (k[x]/(xn ), k[x]/(xm )) ∼
= k[x]/(xn ). In the case n > m, the map

k[x] → Homk[x] (k[x]/(xn ), k[x]/(xm )), g(x) 7→ ψg ,

where ψg (1) = g(x), is easily seen to be a well-defined homomorphism with kernel (xm ).
Therefore, Homk[x] (k[x]/(xn ), k[x]/(xm )) ∼= k[x]/(xm ).
b. We claim that the map Z → Zn ⊗Z Zm , a 7→ a(1 ⊗ 1) is a surjection with kernel
(gcd(n, m)). It is clear that it is a surjection since 1 ⊗ 1 generates Zn ⊗Z Zm . Now suppose
that a is in the kernel of this map. Then a⊗1 = 0. Note that the bilinear map φ : Zn ×Zm →
Zd , (a, b) 7→ ab, where d = gcd(n, m), is well-defined, so there is a commutative diagram
⊗
Zn × Zm Zn ⊗Z Zm
.
φ
Zd

Hence, a ⊗ 1 = 0 implies that a = 0 modulo d. Conversely, if a = 0 modulo d, then a is a

multiple of d, which means that we can write a = sn + tm for some integers s and t. But
then it immediately follows that a ⊗ 1 = 0. We conclude that Zn ⊗Z Zm ∼ = Zd .
Now consider k[x]/(xn ) ⊗k[x] k[x]/(xm ). For definiteness, suppose that n ≤ m. We
claim that k[x] → k[x]/(xn ) ⊗k[x] k[x]/(xm ), f (x) 7→ f (x)(1 ⊗ 1) has kernel (xn ). Suppose
that f (x) ⊗ 1 = 0. There is a well defined bilinear map ψ : k[x]/(xn ) × k[x]/(xm ) →
k[x]/(xn ), (f (x), g(x)) 7→ f (x)g(x). Thus, we have a commutative diagram
⊗
k[x]/(xn ) × k[x]/(xm ) k[x]/(xn ) ⊗k[x] k[x]/(xm )
.
ψ
k[x]/(xn )

Hence, if f (x)⊗1 = 0, then f (x) = 0 modulo (xn ) as desired. conversely, if f (x) = 0 modulo
(xn ), then clearly f (x) ⊗ 1 = 0. We conclude that k[x]/(xn ) ⊗k[x] k[x]/(xm ) ∼= k[x]/(xn ).

20
 c. We claim that the map

k[x] ⊗k k[x] → k[x] → k[x, y], f (x) ⊗ g(x) 7→ f (x)g(y)

is an isomorphism of k-algebras. It is well-defined by the universal property of tensor

products, and it is not too hard to check that it is bijective.

Exercise 2.5. Suppose k is an infinite field, and Tlet U be the set of nonzero elements of the
polynomial ring k[x] in one variable. Show that ( a∈k (x − a))[U −1 ] 6= a∈k ((x − a)[U −1 ]).
T
Thus, Corollary 2.6 would be false for infinite intersections.
T
Note that a∈k (x − a) = 0; indeed, if any polynomial f is T in this ideal then it vanishes
that every value of k, but since k is infinite, f = 0. Hence, ( a∈k (x − a))[U −1 ] = 0. On
−1
the other hand, for any a ∈ k we have that (x − a)[U ] is the field of all rational functions,
which is nonzero and independent of a. Thus, a∈k ((x − a)[U −1 ]) 6= 0. This provides a
T
counter-example to Corollary 2.6.
Exercise 2.6. Let R be a ring,T and Q let Q1 , . . . , Qn be ideals of R such that Qi + Qj = R
for all i 6= j. Show that R/( i Qi ) ∼= i R/Q Qi as follows:
a. Consider the map of rings φ :TR → i R/Qi obtained from the n projection maps
R → R/Qi . Show that ker φ = i Qi .
b. Let m be a maximal ideal of R. Show that the hypothesis that Qi + Qj = R for all
i 6= j means that at most one of the Qi is contained in m. Now use Corollary 2.9 to
show that φ is surjective.
T
a. r ∈ ker φ if and only if r vanishes in each quotient R/Qi if and only if r ∈ i Qi .
b. If m contains Qi and Qj for i 6= j, then it contains Qi + Qj = R. But m is maximal
and therefore not equal to all of R. By Corollary 2.9, we can check that surjectivity of φ by
checking that the induced map on localizations at maximal ideals is surjective. First note
that if Qi is not contained in m, then (Qi )m = Rm . Furthermore, localization commutes
with finite direct products, so
!
Y Y Y
R/Qi = (R/Qi )m = Rm /(Qi )m
i m i i

has at most one non-zero factor,Tand therefore φm is surjective. We conclude that φ is

surjective, and therefore that R/( i Qi ) ∼
Q
= i R/Qi .
Exercise 2.7. Show that the universal property of the localization given in the text
characterizes R → R[U −1 ] up to unique isomorphism in the sense that if another map
R → S has the same property, then there is a unique isomorphism R[U −1 ] → S making the
diagram commute.

R[U −1 ]

21
 Let φ : R → S be such that φ(U ) ⊆ S × . We begin by proving existence of a homomor-
phism R[U −1 ] → S. Define

ψ : R[U −1 ] → S, r/u 7→ φ(r)φ(u)−1 .

To show that ψ is well defined, let r/u = r0 /u0 . Then there is some v ∈ U such that
v(ru0 − r0 u) = 0. Since φ(v) ∈ S × , we have that φ(r)φ(u0 ) − φ(r0 )φ(u) = 0, which in turn
implies that φ(r)φ(u)−1 = φ(r0 )φ(u0 )−1 . Thus, ψ is well-defined. It is also clearly a ring
homomorphism.
Now suppose that ψ 0 : R[U −1 ] → S is another map making the diagram commute. Then
for any r ∈ R, ψ 0 (r/1) = φ(r). Therefore, for any u ∈ U ,
r u r  u  r  r
φ(r) = ψ 0 = ψ0 · = ψ0 ψ0 = φ(u)ψ 0 ,
1 1 u 1 u u
which implies ψ 0 (r/u) = φ(r)φ(u)−1 . Hence, ψ = ψ 0 , proving uniqueness.

Exercise 2.8. Show that the following universal property similarly characterizes M →
M [U −1 ]: Given a map φ from M to and R-module N on which the elements of U act
by multiplication as automorphisms, there is a unique extension φ0 : M [U −1 ] → N . In
particular, if M and N are R[U −1 ]-modules, then the maps of R-modules from M to N are
the same as the maps of R[U −1 ]-modules.

Similarly as in the previous exercise, we first prove existence by defining

φ0 : M [U −1 ] → N, m/u 7→ u−1 φ(m),

where u−1 is the automorphism of N inverse to u. To show that ψ is well-defined, let

m/u = m0 /u0 . Then there is some v ∈ U such that v(u0 m − um0 ) = 0. Since φ is an
R-module homomorphism, we have that v(u0 φ(m) − uφ(m0 )) = 0. Since v acts by an
automorphism of N , we furthermore have that u0 φ(m) = uφ(m0 ), from which we obtain
u−1 φ(m) = (u0 )−1 φ(m0 ). It is easily check that φ0 is a homomorphism of R-modules.
To prove uniqueness, suppose that we have another extension ψ : M [U −1 ] → N ; by
extension we mean that ψ(m/1) = φ(m). Then for any u ∈ U we have
m u m m
φ(m) = ψ =ψ · = uψ ,
1 1 u u
implying ψ(m/u) = u−1 φ(m), thus proving uniqueness.
For the last statement, let φ : M → N be a map of R modules. Then
u r  r 
rφ(m) = φ(rm) = φ · m = uφ m ,
1 u u
implying that (r/u) = φ((r/u)m). Therefore, φ is also a map of R[U −1 ]-modules. Con-
versely, every map of R[U −1 ]-modules is also clearly a map of R-modules.

Exercise 2.9. One way of describing the ring R[U −1 ] is to say what its modules are: Show
that an R[U −1 ]-module is the same thing as an R-module on which the elements of U act
as automorphisms. In particular, the map M → M [U −1 ] is an isomorphism if and only if
the elements of U act as automorphisms on M .

22
 If M is an R[U −1 ]-module, then the action of u ∈ U on M must be an automorphism,
since the action of 1/u is inverse to it. Conversely, if the elements of U act as automorphisms
on M , then for every u ∈ U we define u1 ·m = u−1 ·m, where u−1 is the inverse automorphism
to u.
Suppose M → M [U −1 ] is an isomorphism (of R-modules). Then clearly M can be
given an R[U −1 ]-module structure by pulling back the R[U −1 ]-module structure on M [U −1 ].
Conversely, suppose the elements of U act as automorphisms on M . We will show that
M → M [U −1 ], m 7→ m/1 is an isomorphism of R-modules. If m/1 = 0 in M [U −1 ], then
there is some u ∈ U such that um = 0. But u acts as an automorphism, so m = 0. It is easy
to see that m/u is the image ofu−1 · m, so we conclude that m 7→ m/1 is an isomorphism
as desired.
Exercise 2.10. Show that every finitely generated module over R[U −1 ] is the localization
of a finitely generated module over R. Here is a truly trivial statement that sounds deeper:
The same is true without the condition finitely generated.
Let M be a finitely generated R[U −1 ]-module, generated by the elements {m1 , . . . , mn }.
Let M 0 be the R-submodule of M (viewed as an R-module) generated by the same elements.
It is then easy to show that M ∼ = M 0 [U −1 ].
The second claim has the same proof as the first one, but I don’t see why it’s a truly
trivial statement (or at least why it is more trivial than the first one).
Exercise 2.11. Let N 0 ⊆ M [U −1 ] be an R[U −1 ]-submodule, and let N ⊆ M be the preimage
of N 0 . Show that N 0 = N [U −1 ].
Define
n 1 n n
φ : N [U −1 ] → N 0 , 7→ · = .
u u 1 u
Let φ(n/u) = 0 in N 0 . Then vn = 0 for some v ∈ U , which implies that n/u = 0 in N [U −1 ]
and therefore that φ is injective. Let m/u ∈ N 0 . By definition, there is some n ∈ N such
that n/1 = m/u. Thus, φ(n/1) = m/u, proving that φ is surjective. Hence, N0 ∼ = N [U −1 ].
Exercise 2.12. Show that Proposition 2.10 is sharp in the following sense: Consider the
ring R = Z of integers.
L∞Let U be the set of powers of 2. Consider the statement of the
−1
proposition with M = i=1 R, N = R, S =
∞R[U ].
a. If N = R, show that the element 21i i=1 is not in the image of the map
∞ ∞
!
Y Y
α: R [U −1 ] → (R[U −1 ])
i=1 i=1

of the proposition. L
∞
b. Now suppose N = i=1 R/(2i ), with M and S as before. Show that the map sending
the generator of the ith factor of M to the generator of the ith factor of N is nonzero
in HomR (M, N )[U −1 ], but goes to 0 under α.
a. Chasing through the definitions and identifications, we find that
(ni )∞
   
i=1 ni ∞
α k
= k .
2 2 i=1

∞
If 21i i=1 was in the image of α, then there would be a fixed k ∈ N and a sequence (ni )∞
i=1
of integers such that ni = 2k−i for all i ∈ N. But this clearly cannot occur.

23
 b. To prove the first statement, we need to show that the map φ described in the problem
is not annihilated by any element of U . Suppose that 2k φ = 0 for some k ∈ N. If ei is
the element of M with a 1 in the ith component and zeros everywhere else, we have that
2k φ(ei ) = 0 implies that 2k ≡ 0 (mod 2i ). But this is true for every i ∈ N, which is clearly
impossible.
On the other hand, we want to show that the map

id ⊗Z φ : Z[U −1 ] ⊗Z M → Z[U −1 ] ⊗Z N

is zero. Indeed, every element of Z[U −1 ] ⊗Z M can be written as a linear combination of

a ⊗ ei , where a ∈ Z[U ∈] and ei is define as before. But a ⊗ ei = 2ai ⊗ 2i ei goes to zero in
Z[U −1 ] ⊗Z N , which show that φ goes to 0 under α.

Exercise 2.13. Suppose that

i j
0→A→
− B−
→C→0 (∗)
is a short exact sequence of R-modules.
a. Show that (∗) is split if and only if the map HomR (C, B) → HomR (C, C) induced by
the right-hand map of (∗) is an epimorphism.
b. Suppose that (∗) is locally split in the sense that for each maximal ideal P ⊆ R the
localized sequence 0 → AP → BP → CP → 0 is split. If C is finitely presented, show
that (∗) is split by using part a and Proposition 2.10.
a. Suppose that (∗) is split and let φ : C → C be an arbitrary homomorphism. Let s be
a splitting, that is let s be such that j ◦ s = idC . Then φ = j ◦ s ◦ φ = j∗ (s ◦ φ). Hence,
j∗ : HomR (C, B) → HomR (C, C) is surjective.
Conversely, let j∗ : HomR (C, B) → HomR (C, C) be an epimorphism. Then, in particu-
lar, there is some s ∈ HomR (C, B) such that j∗ (s) = j ◦ s = idC . But then s is a splitting
of (∗).
b. Suppose that (∗) is locally split and that C is finitely presented. By Proposition 2.10,
for every maximal ideal P / R, the diagram
∼
=
HomRP (CP , BP ) HomR (C, B)P
(jP )∗ (j∗ )P
∼
=
HomRP (CP , CP ) HomR (C, C)P

commutes, where the maps are homomorphisms of RP -modules. Since (∗) is locally split, we
have that that (jP )∗ is surjective by part a. Then (j∗ )P must also be surjective. Since this
is true for every maximal ideal P of R, Corollary 2.9 tells us that j∗ must be a surjection.
We conclude that (∗) is split by part a.
Exercise 2.14. Show that an ideal I of a Z-graded ring R is homogeneous if and only if
for every element f ∈ I, all the homogeneous components of f are in I.
We do the easy direction first. If all the homogeneous components of an element of I
are also in I, then I is generated by the homogeneous components of its elements, and is
therefore a homogeneous ideal.
On the other hand, suppose that I is homogeneous. Let f ∈ I, and let f = f1 + · · · + fn
be a homogeneous decomposition for f , where deg fi < deg fi+1 for each i and each fi is

24
nonzero. We will show that each fi ∈ I by induction on n. If n = 1, this is trivial. Now
assume that the claim holds when f has up to n−1 terms in its homogeneous decomposition.
Since I is homogeneous, we can write f = g1 + · · · + gk , where gj ∈ I is homogeneous for
every j = 1, . . . , k. Moreover, we can write

f = (g1 + · · · + gk1 ) + (gk1 +1 + · · · + gk2 ) + · · · + (gkn−1 +1 + · · · + gkn )

where each of the bracketed terms are homogeneous, and we may assume that they are
nonzero as well. It is clear from here that we must have g1 + · · · + gk1 = f1 . Hence, f1 ∈ I
and f2 + · · · + fn = f − (g1 + · · · + gk1 ) ∈ I. By induction, f2 , · · · , fn ∈ I, which proves the
claim.

Exercise 2.15. Many basic operations on ideals, when applied to homogeneous ideals in
Z-graded rings, lead to homogeneous ideals. For example, let I be a homogeneous ideal in a
Z-graded ring R. Show that:
a. The radical of I is homogeneous; that is, the radical of I is generated by all the homo-
geneous elements f such that f n ∈ I for some n.
b. If I and J are homogeneous ideals of R, then

(I : J) := {f ∈ R : f J ⊆ I}

is a homogeneous ideal
c. Suppose that for all f, g homogeneous elements of R such that f g ∈ I, one of f and g
is in I. Show that I is prime
√
a. Let f ∈ I, and let f = f1 + · · · + fn be a homogeneous decomposition of f , where fi
is nonzero and deg fi < deg fi+1 for every√ i. We will show by induction on n that each of f ’s
homogeneous components must lie in I. If n = 1, this is trivial. Assume the claim holds
when f has fewer than n > 1 homogeneous components. Suppose that f m ∈ I. Since I is
m m m
homogeneous and the lowest degree √ homogeneous component of f is f√1 , we have f1 ∈ I
by exercise 2.14. But then f1 ∈ I, which implies that f2 + · · · + fn ∈ I. The claim then
follows from the inductive hypothesis.
b. Let f ∈ (I : J) and let f = f1 + · · · + fn be a homogeneous decomposition of f as
in part a. Let j ∈ J and let j = j1 + · · · + jm be a homogeneous decomposition of j with
each jk ∈ J. Then f jk = f1 jk + · · · + fn jk ∈ I is a homogeneous decomposition of f jk ,
and since I is homogeneous, we must have that fi jk ∈ J for every i, k. Hence, fi j ∈ J for
every i. Since j was arbitrary, we conclude that each fi ∈ (I : J), proving that (I : J) is
homogeneous.
c. Let f, g ∈ R be such that f g ∈ I. Moreover, let f = f1 + · · · + fm and g = g1 + · · · + gn
be homogeneous decompositions of f and g such that f1 and g1 are of minimal degree among
the the fi and gi , respectively. We will prove that either f ∈ I or g ∈ I by induction on m+n.
If m+n = 2, then f and g are homogeneous, and the claim follows by hypothesis. Otherwise,
note that f1 g1 is a homogeneous component of I and therefore f1 ∈ I by homogeneity of I,
without loss of generality. Then f1 g ∈ I, implying that f g − f1 g = (f2 + · · · + fm )g ∈ I. By
induction, either f2 + · · · + fm ∈ I or g ∈ I. In the latter case we are done. In the former,
simply note that f1 ∈ I and f2 + · · · + fm ∈ I, so f = f1 + (f2 + · · · + fm ) ∈ I.
Exercise 2.16. Let R be a Z-graded ring and let M be a graded R-module. Show that if
x ∈ R is a homogeneous element of nonzero degree, then u := 1 − x is a nonzerodivisor on
M . The element u is a unit if and only if x is nilpotent.

25
 Let m ∈ M and let m = m1 + · · · + mn be a homogeneous decomposition of m, with
m1 of minimal degree among the mi . If deg x > 0, then m1 the minimal degree term of
(1 − x)m. If deg u < 0, then mn is the maximal degree term of (1 − x)m. In either case,
this shows that um = (1 − x)m is a nonzerodivisor on M .
Suppose that u is a unit, and let u−1 = r1 + · · · + rk be a homogeneous decomposition
of u−1 , where deg ri < deg ri+1 . Assume that deg x > 0. The deg x < 0 case is similar. We
have
1 = (1 − x)(r1 + · · · + rk ).
The minimal degree term on the right-hand side above is r1 , so we must have that r1 = 1.
Rearranging, we obtain
x = (1 − x)(r2 + · · · + rk ).
The minimal degree term on the right-hand side above is r2 , so r2 = x. Continuing like
this, we obtain that ri = xi−1 for each i. Then

1 = (1 − x)(1 + x + · · · + xk−1 ) = 1 − xk ,

implying that xk = 0. Conversely, if x is nilpotent then xk = 0 for some k, and the previous
line shows that u−1 = 1 + x + · · · + xk−1 .
Exercise 2.17. Suppose R is a Z-graded ring and 0 6= f ∈ R1 . Show that R[f −1 ] is again
a Z-graded ring. Let S = R[f −1 ]0 .
a. Show that R[f −1 ] ∼= S[x, x−1 ], where x is a new variable. (The ring S[x, x−1 ] is the
ring of Laurent polynomials over S. We make the convention that if S is the zero
ring, then S[x, x−1 ] is also the zero ring.)
b. Show that S = R[f −1 ]0 ∼ = R/(f − 1).
c. Let U ⊆ R be a multiplicatively closed set of homogeneous elements containing at least
one nonzero element of R1 . Show that R[U −1 ] ∼ = (R[U −1 ])0 [x, x−1 ], where x is an
indeterminate of degree 1.
d. Now let P be a homogeneous prime ideal of R, and let U be the multiplicative set of
homogeneous elements not in P . Note that R[U −1 ] is naturally a Z-graded ring. We
define R(P ) to be the degree-0 component of R[U −1 ]0 of R[U −1 ]. If P does not contain
R1 , then by part c, R[U −1 ] ∼
= R(P ) [x, x−1 ].
We claim that R[f −1 ] = ⊕n∈Z Sn , where
 
r
Sn = : r is homogeneous, k ≥ 0, and deg(r) − k = n .
fk

It is clear that every element of R[f −1 ] can be expressed as the sum of elements in the Sn .
To show that such a sum is unique, let
r1 rn
k
+ · · · + kn = 0,
f 1 f

with deg(ri ) − ki 6= deg(rj ) − kj for i 6= j. Let N ≥ max k1 , . . . , kn . Then

f N −k1 r1 + · · · + f N −kn rn = 0

in R[f −1 ]. After multiplying the above equation by a suitable power of f , we have that the
above equation holds in R, so we may assume that the equation already holds in R. Since

26
R is Z-graded, we have that f N −ki ri = 0 in R, which implies that ri /f ki = 0 in R[f −1 ] for
each i. This proves our original claim. It is clear that Si Sj ⊆ Si+j .
a. Define a map
 
−1 −1 r r
ϕ : R[f ] → S[x, x ], 7→ xdeg(r)−k .
fk f deg(r)

on homogeneous elements of R[f −1 ], and extend by linearity to all of R[f −1 ]. We now

show that ϕ is well-defined. If r/f k = 0, then r/f deg(r) = 0. If r/f k = r0 /f k0 6= 0, then
deg(r) − k = deg(r0 ) − k0 by the Z-grading of R[f −1 ]. Hence,
r r0 r0 r r0
= = ⇒ = ,
f deg(r)−k0 f deg(r)−k f deg(r 0 )−k0 f deg(r) f deg(r 0)

proving that ϕ is well-defined.

We now prove that ϕ is a ring homomorphism. To check that ϕ preserves addition, let
r/f k and r0 /f k0 be such that deg(r) − k = deg(r0 ) − k0 . Without loss of generality, assume
that k ≥ k0 . Then,

r + f k−k0 r0 r + f k−k0 r0
 
r r0
+ = →
7 xdeg(r)−k
fk f0k fk f deg(r)
   
r deg(r)−k r0
= x + xdeg(r)−k
f deg(r) f k0 +deg(r)−k
   
r deg(r)−k r0
= x + xdeg(r0 )−k0 .
f deg(r) f deg(r0 )

It then follows by definition of ϕ that it preserves addition on all of R[f −1 ]. To show that
ϕ preserves multiplication, first note that this is obvious for homogeneous elements. It
then follows that ϕ preserves multiplication on all of R[f −1 ] since we already know that it
preserves addition.
To see that ϕ is surjective, simply note that ϕ(r/f deg(r)−m ) = (r/f deg(r) )xm for any
m ∈ Z. Surjectivity then follows by linearity. For injectivity, first note that ϕ preserves
degrees, where S[x, x−1 ] is graded in the obvious way (by declaring deg(x) = 1). Thus, it
suffices to show that ϕ does not map any nonzero homogeneous element to zero. This is
easy: if ϕ(r/f k ) = 0, then r/f deg(r) = 0, implying that r/f k = 0. Thus, R[f −1 ] = S[x, x−1 ].
b. Define ψ : R → R[f −1 ]0 , r 7→ r/f deg(r) on homogeneous elements, and extend by
linearity. It is easy to see that ψ is a ring homomorphism. It is also clear that (f −1) ⊆ ker ψ,
so it suffices to show that ker ψ ⊆ (f −1). Let r1 +· · ·+rn ∈ ker ψ, with each ri homogeneous
and deg(ri ) < deg(ri+1 ). Then there is some N ∈ N large enough so that

f N −deg(r1 ) r1 + · · · + f N −deg(rn ) rn = 0

in R. But then
 
r1 + · · · + rn = r1 + · · · + rn − f N −deg(r1 ) r1 + · · · + f N −deg(rn ) rn
   
= 1 − f N −deg(r1 ) r1 + · · · + 1 − f N −deg(rn ) rn ∈ (f − 1).

This proves the claim.

27
 c. This is a generalization of a that follows very similarly. R[U −1 ] has a grading by
declaring deg(r/u) = deg(r) − deg(u) when r ∈ R is homogeneous and u ∈ U (this is similar
to the case of R[f −1 ]). It is not hard that this definition does indeed turn R[U −1 ] into a
Z-graded ring.
Let f ∈ U be a nonzero element of degree 1. Then we define
 
−1 −1 −1 r r
ϕ : R[U ] → (R[U ])0 [x, x ], 7→ xdeg(r)−deg(u) ,
u uf deg(r)−deg(u)
and notice that this reduces to the formula from part a in that context. It follows similarly
that φ is an isomorphism.
d. This is immediate from part c.
Exercise 2.18. Show that if R is a graded ring with no nonzero homogeneous prime ideals,
then R0 is a field and either R = R0 or R = R0 [x, x−1 ].
A Zorn’s lemma argument shows that every proper homogeneous ideal is contained in
a maximal homogeneous ideal. Let I be a maximal homogeneous ideal. To see that it is
prime, it suffices to check that for f g ∈ I implies that either f ∈ I or g ∈ I whenever f and
g are homogeneous by Exercise 2.15c. Suppose f ∈ / I. Then (I, f ) = R since I is maximal
homogeneous and (I, f ) is homogeneous. Hence, 1 = ar + bf for some a, b ∈ R and r ∈ I.
But then g = agr + bf g ∈ I, proving that I is prime.
Let r ∈ R0 . If r is a non-unit, then it is contained in some maximal homogeneous ideal I.
But, as we just showed, I is a homogeneous prime ideal, so we conclude that r = 0. Hence,
R0 is a field. Note that this proof actually shows that all nonzero homogeneous elements
are units. R0 is the only homogeneous component that is a field since it is the only one that
is multiplicatively closed.
If R does not contain any nonzero elements of nonzero degree, then R = R0 trivially.
Otherwise, we may assume that R has a nonzero element of degree 1. To see this, take the
let f be the nonzero homogeneous element of least positive degree n. Then deg(f −1 ) = −n.
Let g be a nonzero homogeneous element of degree m. Writing n = qm + r with remainder
0 6 r < n, we see that 0 ≤ deg(f g −qm ) = r < n (recall that g is a unit). If r 6= 0,
this contradicts the definition of f . Hence, R is concentrated in degrees that are integer
multiples of n. By rescaling, we may just as as well assume that deg(f ) = 1. Thus, by the
previous exercise, R = (R[f −1 ])0 [x, x−1 ] = R0 [x, x−1 ], since f is a unit.
Exercise 2.19. Let R be a ring and let M be an R-module. Suppose that {fi } is a set of
elements of R that generate the unit ideal. Prove:
a. If m ∈ M goes to 0 in each M [fi−1 ], then m = 0.
b. If mi ∈ M [fi−1 ] are elements such that mi and mj go to the same element of
M [fi−1 fj−1 ], then there is an element m ∈ M such that m goes to mi in M [fi−1 ]
for each i. Note that by part a, the element m is unique.
Throughout the problem, we assume that {fi } is finite with m elements. This is fine
since 1 is a finite linear combination of the fi .
n
P a. We first show that {fi P } generates R for all n ∈ N. There are some ri ∈ R such that
mn
i ri fi = 1. We then have ( i ri fi ) = 1, and each term in the left-hand side must be
a multiple of fin for some i, proving our claim. The assumption tells us that there exists
ni
nPi ∈ N such that fi m P = 0 in R. We may choose n > maxi {fi } and si ∈ R such that
( i si fin ). Hence, m = ( i si fin ) m = 0.

28
 b. Let mi = mi /fini . The map M [fi−1 ] → M [fi−1 fj−1 ] sends mi to fjni mi /(fi fj )ni . By
assumption, we have
fjni mi n
fi j mj
= in M [fi−1 fj−1 ]
(fi fj )ni (fi fj )nj
for every pair (i, j). Thus, there is some Nij ∈ N such that
n
(fi fj )Nij (fi fj )nj fjni mi = (fi fj )Nij (fi fj )ni fi j mj in M.

Therefore,
Nij +ni +nj Nij +ni
fj mi = fj mj in M [fi−1 ].
P
Let N > max(i,j) {Nij } + i ni . Then we have
N −nj
fjN mi = fj mj in M [fi−1 ]

for every pair (i, j).

N −nj
By part a, there are rj ∈ R such that 1 = j rj fjN . Let m = j rj fj
P P
mj ∈ M .
Then m goes to
 
N −nj
X X X
rj fj mj = rj fjN mi =  rj fjN  mi = mi in M [fi−1 ],
j j j

as desired.
If m0 goes to mi in M [fi−1 ] for every i, then m − m0 goes to zero in each localization.
Thus, m = m0 by part a, so the element m constructed above is unique.
Exercise 2.20. There are other collections of localizations that have the property of the set
of localizations at all maximal ideals described in Corollary 2.9. Perhaps the most important
type is the following, which generalizes the covering of an affine set by open affine subsets:
Show that for a collection of elements f1 , . . . , fm ∈ R, the following properties are equivalent:
a. The ideal generated by f1 , . . . , fm is R.
b. An R-module M is zero if and only if each of the modules M [fi−1 ] is zero.
Let f1 , . . . , fm generate R. If M is zero, then it is clear that M [f −1 ] is zero as well.
Conversely, suppose that each M [fi−1 ] = 0. The m ∈ M goes to zero in each M [fi−1 ], which
implies that m = 0 by part a of the previous exercise. Since m ∈ M is arbitrary, this implies
that M = 0
Conversely, suppose b holds. Consider that R-module M = R/(f1 , . . . , fm ). Then
M [fi−1 ] = 0 for each i, since each fi annihilates M . But then M = 0, which implies that
R = (f1 , . . . , fm ).
Exercise 2.21. a. Show that P1 is obtained by gluing two copies of A1 as follows: Con-
sider the two open subsets U0 = {(a0 , a1 ) ∈ P1 : a0 6= 0} and U1 = {(a0 , a1 ) ∈ P1 :
a1 6= 0}, as described in Chapter 1. Each of these is identified with an affine space:
With notation as above we may write the identifications as U0 ∼
= X by (a0 , a1 ) 7→ a1 /a0
and U1 ∼= Y by (a0 , a1 ) 7→ a0 /a1 . Show that U0 ∩ U1 is taken by these two identifi-
cations to X − X 0 and Y − Y 0 , respectively. Show that the composite identification
X − X 0 → U0 ∩ U1 → Y − Y 0 is the map s 7→ t−1 .
b. Formulate a corresponding “gluing” description of Pn .

29
 Parts a and b are done in most textbooks on differential geometry: the sets Ui are the
standard sets in a chart for projective space. See, for example, Lee’s Introduction to Smooth
Manifolds.
Exercise 2.22. An ideal maximal with respect to not being finitely generated is prime; thus
a ring whose primes are finitely generated is Noetherian.

Let I be maximal with respect to not being finitely generated, and let a, b ∈ R be such
that ab ∈ I but a, b ∈ / I. Then (I, a) is finitely generated, P since it strictly contains I. Let
(I, a) = (f1 , . . . , fm , a) for some fi ∈ R. Write fi = ri a + j rij sj , where ri , rij ∈ R and
sj ∈ I for all i, j. Since fi − ri a ∈ I, we may assume that the fi are already in I.
Note that (I : a) is finitely generated: it is clear that (I : a) ⊇ I, and moreover, b ∈ I
since ab ∈ I. Therefore, (I : a) strictly contains I, meaning it is finitely generated. Let
(I : a) = (g1 , . . . , gn ). We claim that I = (f1 , . . . , fm , ag1 , . . . , agn ). All of the fi and agj
lie in I, so we just need to show that anyP r ∈ I is a linear combination of these elements.
Since I ⊆ (I : a), we can write r = qa + i pi fi for some q, pi ∈ R. But then qa ∈ I, so
q can be written as a linear combination of the gj . But then r ∈ (f1 , . . . , fm , ag1 , . . . , agn ),
proving the claim. This contradicts the assumption that I is finitely generated; therefore, I
is prime.
Exercise 2.23. An ideal maximal among those that are not principal is prime.
Let I be maximal among non-principal ideals, and let a, b ∈ I be such that ab ∈ I.
Suppose that a ∈ / I. Then I ⊂ (I, a), so (I, a) = (a0 ) for some a0 ∈ R by maximality
of I. Then I = a0 (I : a0 ) essentially by definition. If (I : a0 ) is principal, then I is
principal, contradicting the definition of I. Therefore (I : a0 ) cannot be principal, and since
I ⊆ (I : a0 ), we must have I = (I : a0 ) by maximality of I. Now, there is some f ∈ I and
r ∈ R such that a0 = f + ra. Then ba0 = bf + rab ∈ I, so b ∈ (I : a0 ) = I, proving that I is
prime.
Exercise 2.24. Let R be a Noetherian ring, and let n be a natural number. Show that there
are only finitely many primes P of R such that the cardinality of R/P is 6 n as follows:
a. Suppose that R has infinitely many such primes. let I ⊆ R be an ideal maximal
among those for which R/I has infinitely many such primes. Show that I must be
prime. Replacing R by R/I, we may assume from the outset that R is a domain and
that every proper homomorphic image of R satisfies the desired statement.
b. Note that R must be infinite (otherwise R has only
Q finitely many ideals!). Let a0 , . . . ,
an+1 be distinct elements of R, and let p = i<j (ai − aj ) be the product of their
differences. Because R is a domain, p 6= 0. If P ⊆ R is a prime ideal, and p ∈ / P,
show that the cardinality of R/P is greater that n. Using the hypothesis at the end
of step a, show that there are only finitely many primes P containing p for which the
cardinality of R/P is n or less than n.

a. Let ab ∈ I, and suppose that neither a ∈ / I and b ∈ / I. Then R/(I, a) and R/(I, b)
each have finitely many primes. By the correspondence theorem, the primes of R/I are in
one-to-one correspondence with the primes of R containing I. Any such prime also contains
either (I, a) or (I, b), since ab ∈ I. But there are only finitely many primes containing either
(I, a) and (I, b), contradicting the assumption that R/I has infinitely many primes.
b. If p ∈
/ P , then each of the factors ai − aj ∈/ P , since P is prime. We claim that the
elements a0 − ai with i ≥ 1 descend to distinct elements of R/P . Indeed, if a0 − ai = a0 − aj

30
modulo P , then ai − aj = 0 modulo P . This implies that ai − aj ∈ P , which in turn implies
that i = j. Hence, R/P contains at least n + 1 elements.
Since R/(p) is a proper homomorphic image of R (we use the fact that p 6= 0 here), we
have that there are only finitely many primes Q of R/(p) such that (R/(p))/Q has cardinality
at most n. Every prime P such that R/P has cardinality at most n must contain (p) by
the previous paragraph, and therefore corresponds to a prime Q such that (R/(p))/Q has
cardinality at most n (by the isomorphism theorems). Hence, there are finitely many primes
P such that R/P has cardinality at most n.
Remark. The proof feels a bit funny to me: we assume for a contradiction that R has
infinitely many primes such that R/P has cardinality at most n. Then the contradiction we
derive is that there are finitely many primes such that R/P has cardinality at most n, so
we conclude that R must have finitely many primes such that R/P has cardinality at most
n.
Exercise 2.25. If R is a ring, then as in Exercise 1.25 we write Spec R for the topological
space whose points are the prime ideals of R and whose closed sets are the sets of prime
ideals containing a given ideal of R. Show that Spec R is disconnected —that is, Spec R
is the disjoint union of two nonempty closed sets, say X1 , X2 , if and only if R contains a
nontrivial idempotent —that is, an element e 6= 0, 1 such that e2 = e, as follows.
First, if e is a nontrivial idempotent, show that 1 − e is also a nontrivial idempotent, and
e(e − 1) = 0. Take X1 and X2 to be the sets of primes containing e and 1 − e, respectively.
Show that Spec R is the disjoint union of X1 and X2 and that these sets are nonempty.
For the converse, suppose that Spec R is the disjoint union of nonempty closed sets X1
and X2 .
a. Since Xj is closed, there is an ideal Ij such that P ∈ Xj if and only if P ⊇ Ij . Show
that I1 + I2 = R and every element of I1 I2 is nilpotent.
b. Write 1 = a1 + a2 with aj ∈ Ij . By part a, a1 a2 is nilpotent, say (a1 a2 )n = 0. By
splitting up the right-hand side of the expression 1 = (a1 + a2 )2n = a2n 2n
1 + · · · + a2
suitably, show that a1 and a2 may be replaced with elements e1 and e2 such that
e1 + e2 = 1 and e1 e2 = 0.
c. Show that e1 and e2 are nontrivial idempotents.
If e is an idempotent, then (1 − e)2 = 1 − 2e + e2 = 1 − 2e + e = 1 − e, so 1 − e is also
an idempotent, and it is nontrivial since e 6= 0, 1. Moreover, e(e − 1) = e2 − e = e − e = 0.
Since every prime ideal contains 0, every prime ideal must contain either (e) or (1 − e).
This proves that Spec R = X1 ∪ X2 . If an ideal contains both (e) and (1 − e), then it must
contain 1. But prime ideals never contain units, so no prime ideal can contain both (e) and
(1 − e). This proves that Spec R = X1 t X2 , as desired.
a. The first statement is just the definition of closed sets in the Zariski topology on
Spec R. If I1 + I2 6= R, then I1 + I2 is contained in some prime ideal P . But then P contains
both ideals Ij , which is impossible because then P ∈ X1 ∩ X2 = ∅. Thus, I1 + I2 = R.
Next, I1 I2 ⊆ I1 ∩ I2 is contained in every prime ideal of R. Therefore, I1 I2 is contained in
the nilradical of R, which is precisely the set of nilpotent elements.
b. Every term in the expansion either has a factor of an1 or an2 . Thus, we may write
1 = r1 an1 + r2 an2 for some elements rj ∈ R. Let ej = rj anj . Then we obviously have
e1 + e2 = 1 and e1 e2 = 0.
c. We have e21 − e1 = e1 (e1 − 1) = e1 e2 = 0, so e1 is an idempotent. Similarly, e2 is an
idempotent. Note that e1 cannot be a unit, because e1 ∈ I1 , so if e1 is a trivial idempotent,

31
then e1 = 0. But then e2 = 1, which is impossible since e2 ∈ I2 . Thus, the idempotents e1
and e2 are nontrivial.
Exercise 2.26. Show that R can be written as a direct product of two or more (nonzero)
rings if and only if R contains a non-trivial idempotent. Show that if e is an idempotent,
then R = Re × R(1 − e), and that Re may be realized as a localization, Re = R[e−1 ].
If R = R1 × R2 is a nontrivial decomposition of R as a direct product, then (1, 0) and
(0, 1) are nontrivial idempotents. Conversely, let e be a nontrivial idempotent, and consider
the map ϕ : R → Re × R(1 − e), r 7→ (re, r(1 − e)). Let ϕ(r) = (re, r(1 − e)) = (0, 0).
Then r = re + r(1 − e) = 0, so ϕ is injective. Let (r1 e, r2 (e − 1)) ∈ Re × R(1 − e). Then
ϕ(r1 e + r2 (1 − e)) = (r1 e2 + r2 e(1 − e), r1 e(1 − e), r2 (1 − e)2 ) = (r1 e, r2 (1 − e)), so ϕ is
surjective. Hence, R ∼ = Re × R(1 − e), which proves the converse.
For the second statement, let e be an idempotent. The proof above does not use the
fact that e is a nontrivial idempotent to show that R ∼ = Re × R(1 − e). The only reason we
needed e to be nontrivial is so that neither Re nor R(1 − e) is the 0 ring.
Consider the map ψ : Re → R[e−1 ], re 7→ r/e. Suppose that ψ(re) = 0. Then r/e = 0,
which means that en r = 0 for some n ∈ N. But en = e for every n ∈ N since e is an
idempotent, so er = 0, proving that ψ is injective. Again since en = e for all n ∈ N, every
element of R[e−1 ] is of the form r/e, so ψ is clearly surjective. Thus, Re ∼ = R[e−1 ]. ‘
Q
Exercise 2.27. If Rγ , γ ∈ Γ, are rings, then the direct product of rings γ∈Γ Rγ is the
direct product of the sets Rγ , with componentwise ring operations. If R is a Noetherian ring
all of whose primesQPγ are maximal, then by the Chinese remainder theorem, Proposition
2.14, we have R = γ Rγ , and the product is finite. In this case Rγ = RP is a localization
of R. We have seen in Proposition 2.17 that the modules over R are made by taking a direct
product of modules, one Q over each RP . These phenomena are somewhat more general:
a. Suppose that R = γ∈Γ Rγ is a finite direct product of rings. Write eγ ∈ R for the
element whose γth component is 1 and whose δth component is 0 for δQ6= γ. Show that
every R-module M is uniquely expressible as a direct product M = γ∈Γ Mγ , where
Mγ is an Rγ -module and the action is componentwise; show in fact that Mγ = eγ M =
M [e−1
γ ]. Show that any homomorphism
Y Y
ϕ: M = Mγ → N = Nγ
γ∈Γ γ∈Γ

of R-modules is a Q direct product of homomorphisms ϕγ : Mγ → Nγ of Rγ -modules.

b. Show that if R = γ∈Γ Rγ is an infinite direct product, then not every module is the
product of Rγ -modules.
Q
P a. Consider the map ψ : M → γ∈Γ Mγ , m 7→ (eγ m)γ∈Γ . If ψ(m) = 0, then m =
γ∈Γ eγ m = 0, which proves
P
that ψ is injective (we are using finiteness of Γ here). Moreover,

ψ is surjective, since ψ eγ mγ = (eγ mγ )γ∈Γ . Hence, M = ∼Q Mγ =
Q
eγ M .
γ∈Γ γ∈Γ γ∈Γ
The second statement follows from the observation that ϕ(Mγ ) = ϕ(eγ M ) ⊆ eγ ϕ(M ) ⊆
e γ N = Nγ .
Q
b. Inspired by the result of Exercise 2.12a, we claim that the i∈N Z-module
!
Y
M= Z [2−1 ]
i∈N

32
is not a direct product of Z-modules. For a contradiction, suppose that
!
Y Y
ϕ: Z [2−1 ] → Mi ,
i∈N i∈N

is an isomorphism, where Mi is a Z-module for every i ∈ N. It is not hard to see that ϕ

carries ei M into Mi , and therefore that ϕ restricts to an isomorphism from ei M ∼
= Z[2−1 ]
to Mi . Hence, there is an isomorphism
!
Y Y
α: ϕ: Z [2−1 ] → Z[2−1 ],
i∈N i∈N
Q
which is the same α as in Exercise 2.12, except now we view it as a map of i∈N Z-
modules. However, as we showed in Exercise 2.12, α is not surjective, and therefore is
not an isomorphism, a contradiction.

33