ALGEBRA HOMEWORK SET IV SOLUTIONS

JAMES CUMMINGS

(1) (a) (internal sum) Prove that if M1 and M2 are submodules of an R-

module M , then the submodule generated by M1 ∪ M2 is
M1 + M2 = {m1 + m2 : mi ∈ Mi }.
Easy! Just check its a submodule, and note that any submodule which
contains M1 ∪ M2 is closed under sums so contains it.
(b) (external direct sum) Prove that if M1 and M2 are R-modules and we
define operations on
M1 ⊕ M2 = {(m1 , m2 ) : mi ∈ MI }
by r(m1 , m2 ) = (rm1 , rm2 ), (m1 , m2 ) + (n1 , n2 ) = (m1 + n1 , m2 + n2 )
then the resulting structure is an R-module.
Routine.
(c) (internal direct sum) Prove that if M1 , M2 ≤ M and M1 ∩ M2 = 0
then M1 + M2 ' M1 ⊕ M2 . What can you say about the structure of
M1 + M2 in general?
Define φ : (m1 , m2 ) 7→ m1 + m2 . Easily it’s a surjective HM (an
epimorphism). The kernel is the set of pairs with m1 + m2 = 0, but
then m1 = −m2 ∈ M1 ∩ M2 = 0, so the kernel is trivial. In general
the same argument shows that M1 + M2 is isomorphic to the quotient
of M1 ⊕ M2 by a submodule isomorphic to M1 ∩ M2 .
(2) Let M be an R-module and I an ideal of R. Prove thatP
(a) If we define IM to be the set of all finite sums i ri mi with ri ∈
I, mi ∈ M then IM ≤ M .
Routine!
(b) M/IM has the structure of an R/I module if we define (r + I)(m +
IM ) = rm + IM .
Routine, the main point is that this is well defined (in fact IM is
the smallest submodule that makes the quotient have an R/I-module
structure).
(3) Consider a sequence of R-modules Mi where the index i runs through some
interval of integers, together with R-module HMs αi : Mi → Mi+1 . The
sequence is said to be exact at Mi if im(αi−1 ) = ker(αi ).
(a) Show that 0 → M1 → M2 is exact at M1 iff α1 is injective.
α1 is injective iff the kernel of α1 is 0 iff the kernel of α1 is the image
of α0 .
(b) Show that M1 → M2 → 0 is exact at M2 iff α1 is surjective.
α1 is surjective iff the image of α1 is M2 iff the image of α1 is the
kernel of α2 .
(c) When is the sequence 0 → M1 → M2 → 0 exact at both of M1 , M2 ?
By the above, this happens iff α1 : M1 ' M2 .
1
2 JAMES CUMMINGS

(d) Suppose that 0 → M1 → M2 → M3 → 0 is exact at all of M1 , M2 , M3 .

What does this tell you about the relation between the modules Mi ?
Let N ≤ M2 be the image of α1 . Then N is the kernel of α2 , and
is isomorphic to M1 . Also M3 ' M2 /N . So this sequence is a slick
description of the quotient construction in diagrammatic form.
(4) Recall that when R is an ID we defined the field of fractions by considering
the set X = {(r, s) : r ∈ R, s ∈ R \ {0}} and introducing an equivalence
relation (r1 , s1 ) ' (r2 , s2 ) iff r1 s2 ' r2 s1 .
(a) Show by example that if R is not an ID then the binary relation defined
in this way may not be an equivalence relation.
Easy (look at the proof that it is an ER when R is an ID if you are
stuck)
(b) Let R be an arbitrary ring. A subset S ⊆ R is called multiplicatively
closed iff 1 ∈ S and S is closed under multiplication.
Prove that if we define a relation on R × S by (r1 , s1 ) '∗ (r2 , s2 ) iff
there is s3 ∈ S such that s3 (r1 s2 − r2 s1 ) = 0 then '∗ is an equivalence
relation.
Easy!
What is this in the case when R is an ID and S = R \ {0}?
It reduces to the old ER in this case.
(c) With the same assumptions as the last part, we write r/s for the '∗ -
class of the pair (r, s) and RS −1 for the set of such classes. Prove that
if we attempt to define
r1 /s1 × r2 /s2 = (r1 r2 )/(s1 s2 ), r1 /s1 + r2 /s2 = (r1 s2 + r2 s1 )/(s1 s2 ),
then we get well-defined operation which make RS −1 into a ring.
Routine and very very dull.
(d) Prove that the map r 7→ r/1 is a HM from R to RS −1 , and that every
element of S is mapped to a unit in RS −1 . When is the map injective?
When is it surjective?
The defn of the ring structure ensures that it’s a HM. We have explic-
itly added an inverse 1/s for each element s/1 in the image of S.
Now we analyse the kernel. r is in the kernel iff r/1 = 0/1 iff there is
s ∈ S such that sr = 0. So the map is injective iff S does not contain
zero divisors.
As for surjectivity, that amounts to the assertion that for all r ∈ R
and s ∈ S there is r1 ∈ R such that r1 /1 = r/s, that is to say that for
some s1 ∈ S we have s1 (r − sr1 ) = 0.
(5) Let R be a ring and let Spec(R) be the set of prime ideals of R. For each
ring element a, let Oa = {P ∈ Spec(R) : a ∈ / P }. Say that a set X of prime
ideals is open if for every P ∈ X there exists a such that P ∈ Oa ⊆ X.
(a) Prove that Oa ∩ Ob = Oab , O0 = ∅, O1 = Spec(R).
For a prime ideal P , ab ∈ P iff a ∈ P or b ∈ P . So ab ∈ / P iff a ∈
/P
and b ∈/ P . Every prime ideal contains 0 and no prime ideal contains
1.
(b) Prove that the collection of open sets form a topology for Spec(R),
and describe it when R = Z.
The properties of the Oa listed above make this easy, in fact they form
a basis for this Zariski topology. Now the prime ideals of Z are (0) and
 ALGEBRA HOMEWORK SET IV SOLUTIONS 3

(p) for p prime. When n is nonzero and composite we see that On is

the cofinite set of ideals (p) for p not dividing n together with the ideal
(0). So the open sets are ∅ and all cofinite sets of prime ideals which
contain (0).
(c) (Trickier) Prove that this topology is compact.
We need to show that any open cover has a finite subcover. Enough
to show that any cover by sets of form Oa has a finite subcover. So let
Oa for a ∈ X be a cover. This means that every prime ideal is in Oa
for some a ∈ X, that is no prime ideal contains X, that is no prime
ideal contains the ideal generated by X. But any proper ideal extends
to a maximal (hence prime) ideal so X generates R. Hence a finite
subset of X generates R, and from this we read off an open subcover.
(6) Let p be prime. Let Rn = Z/pn Z, and let πn : Rn+1 7→ Rn be the surjective
HM which maps a + pn+1 Z to a + pn Z. Define a ring Zp as follows: the
elements are infinite sequences (r0 , r1 , . . .) such that ri ∈ Ri and πi (ri+1 ) =
ri for all i. Addition are multiplication are defined coordinatewise
Prove that
(a) Zp is uncountable.
Diagonalise a la Cantor.
(b) Zp is an ID which contains an isomorphic copy of Z.
Routine.
(c) The sequence (2, 2, 2, . . .) has a square root in Z7 .
Build a square root by induction, starting either with r0 = 3 or r0 = 4.
At stage n suppose we have rn2 = 2 mod pn . Then we must choose rn+1 =
rn + xpn so that rn+12
= 2 mod pn+1 . This is an easy problem in mod p
arithmetic.
Cultural note: this is the ring of p-adic integers and we just did an easy
example of Hensel’s Lemma.
(7) The R-module M is said to be Artinian iff there is no infinite strictly
decreasing sequence of submodules. R is Artinian iff it is Artinian as an
R-module, that is there is no infinite strictly decreasing sequence of ideals.
(a) Give an example of an infinite Artinian ring.
Any field.
(b) Prove that if R is a field, the classes of Artinian and Noetherian mod-
ules coincide.
Using basic facts about dimension, both classes coincide with the class
of finite dimensional R-modules.
(c) Give an example of a Noetherian module which is not Artinian.
Z is Noetherian but not Artinian.
(d) Give an example of a Artinian module which is not Noetherian.
We work with Z-modules, that is abelian groups. Consider the sub-
group of Q/Z consisting of elements of form a/2n for integers a, n
where n ≥ 0. Easily the subgroups generated by 1/2n form an infinite
strict increasing chain.
Let H be any subgroup, let h ∈ H and write h = a/2n for a odd. So
gcd(a, 2n ) = 1, and by elementary number theory there are integers
C, D with Ca + D2n = 1, that is Ca/2n + D = 1/2n . So easily
H contains the subgroup generated by 1/2n . Therefore either H is
4 JAMES CUMMINGS

generated by 1/2n for some n or it’s the whole group. It follows easily
that the group is Artinian as a Z-module.
Note: all Artinian rings are Noetherian, but it is not completely easy to
see this.