HOMEWORK 2 SOLUTIONS

MATH 121

R is an integral domain with quotient eld Q and let N

U = R be the set of nonzero elements in R and dene U 1 N
to be the set of equivalence classes of ordered pairs of elements (u, n) with u U and
n N under the equivalence relation (u, n) (u0 , n0 ) if and only if u0 n = un0 in N .
1
(a) Prove that U
N is an abelian group under the operation dened by (u1 , n1 )+
(u2 , n2 ) = (u1 u2 , u2 n1 + u1 n2 ). Prove that r(u, n) = (u, rn) denes an action
1
of R of U
N making it into an R-module. (This is an example of localization
Problem (10.4.8). Suppose

be any

R-module.

Let

considered in general in Section 4 of Chapter 15, cf. also Section 5 in Chapter

7.)

QN

U 1 N

dened by sending (a/b, n) to (b, an)

a R, b U , n N , is an R-balanced map, so induces a homomorphism
f from Q R N to U 1 N . Show that the map g from U 1 N to Q R N dened
by g((u, n)) = (1/u) n is well-dened and is an inverse homomorphism to
f. Conclude that Q R N
= U 1 N as R-modules.
Conclude from (b) that (1/d) n is 0 in Q R N if and only if rn = 0 for
some nonzero r R.
If A is an abelian group, show that Q Z A = 0 if and only if A is a torsion
abelian group (i.e., every element of A has nite order).

(b) Show that the map from

to

for

(c)
(d)

Solution.

(b)

(a) It is straightforward to check that letting the identity element be

(1, 0) and the inverse of (u, n) be (u, n) gives an abelian group structure
1
on U
N , and that the suggested R-action in fact does make U 1 N into an
R-module.
The only interesting thing to check in regard to the R-balancedness is that, if
we call the map , then (a/b, n) + (c/d, n) = (a/b + c/d, n). We have
(a/b, n) + (c/d, n) = (b, an) + (d, cn)
= (bd, adn + bcn)
= ((ad + bc)/(bd), n)
= (a/b + c/d, n),
as desired.

Hence,

induces a map

straightforward.

Date

: 18 January, 2011.

Q R N U 1 N .

The rest is again

MATH 121
(c) Suppose

rn = 0

for some nonzero

r R.

We have

(1/d) n = g((d.n)) = g((rd, rn)) = (1/rd) (rn) =

as desired. Now, if

(1/d) n = 0,

1
0 = 0,
rd

then we have

0 = (1/d) n = 1 dn = g((1, dn)),

so

dn = 0.

(d) We know from part (c) that the torsion elements of

become zero in

Q Z A.

Hence

Q Z A = 0

are exactly those that

if and only if

is torsion.

R is an integral domain with quotient eld Q and let N be

any R-module. Let Q R N be the module obtained from N by extension of scalars
from R to Q. Prove that the kernel of the R-module homomorphism : N Q R N
is the torsion submodule of N (cf. Exercise 8 in Section 1). [Use the previous exercise.]
Problem (10.4.9). Suppose

Solution. The proof is very similar to that of 10.4.8(c).

Problem (10.4.13). Prove that the usual dot product of vectors dened by letting

(a1 , . . . , an ) (b1 , . . . , bn )

be

a1 b 1 + + an b n

Solution. We have to show that

is a bilinear map from

Rn Rn

(r1 a1 + r2 a2 ) b = r1 (a1 b) + r2 (a2 b).

to

R.

This is

either easily checked by hand or else recalled from elementary linear algebra.

R is commutative and let I and J be ideals of R, so R/I

R-modules.
Prove that every element of R/I R R/J can be written as a simple tensor of
the form (1 mod I) (r mod J).
Prove that there is an R-module isomorphism R/I R R/J
= R/(I + J)
0
0
mapping (r mod I) (r mod J) to rr mod (I + J).

Problem (10.4.16). Suppose

and

R/J

(a)
(b)

are naturally

Solution.

(a) We'll rst check that a simple tensor can be rewritten in the desired

form. A typical simple tensor is of the form

(r

mod I) (s

(r mod I) (s mod J),

mod J) = r(1 mod I) (s

mod J)

= (1 mod I) (rs

mod J).

but

Now, suppose instead that we have a sum of simple tensors, which looks like

(ri

mod I) (si

mod J).

By the above, this is equal to

(1 mod I) (ri si

mod J) = (1 mod I)

which is a simple tensor of the desired form.

X

ri si


mod J ,

HOMEWORK 2 SOLUTIONS

: R/I R R/J R/(I + J) be the map in the problem, and let

: R/(I +J) R/I R R/J be given by (r mod (I +J)) = (1 mod I)(r
mod J). These are both homomorphisms; let's show that they are inverses.

(b) Let

We have

((r

mod I) (r

mod J)) = (rr0

= (1

by part (a), this is equal to

mod (I + J))

mod I) (rr0

(r mod I) (r0 mod J),

mod J);

so this composition is

the identity. For the other direction, we have

(r

mod (I + J)) = ((1 mod I) (r

=r

mod J))

mod (I + J),

so this composition is also the identity. Hence

is an isomorphism.

Problem (10.4.24). Prove that the extension of scalars from

Z[i]

of the ring

is isomorphic to

as a ring:

Z to the Gaussian integers

Z[i] Z R = C as rings.

Solution. We'll check that

C satises the universal property for the tensor product.

This means that, if we let : Z[i] R C be the tensor product map (a + bi, r) =
ar+bri and : ZZ[i]R A is a bilinear map, then there's a unique map : C A
so that = . We dene ( + i) = (1, a) + (i, b). Hence, for a + bi Z[i]
and r R, we have
(a + bi, r) = (ar + bri)
= (1, ar) + (i, br)
= a(1, r) + b(i, r)
= (a + bi, r),
as desired. In fact, this only shows that

is an isomorphism of modules, but since

all the maps in sight can be taken to be ring maps, the isomorphism is in fact an
isomorphism of rings.
Well, perhaps there's a bit of a story to be told in that last remark, actually.
Whenever we deal with universal properties, we're talking about categorical notions.
These are generally not too heavily dependent in which category we're working in. In
this case, we begin by talking about the tensor product of two modules, but now we're
suddenly discussing the tensor product of two algebras (or rings). The reason we can
move so easily between these two seemingly dierent notions is that their denitions
are essentially identical: when all our maps are only required to be module maps,
we nd that certain things are isomorphic as modules. When we require them to be
ring maps, our arguments generally go through just as before to show that they're
isomorphic as rings.

MATH 121

Problem (10.4.27).

(a) Write down a formula for the multiplication of two ele-

a 1 + b e2 + c e3 + d e4 and a0 1 + b0 e2 + c0 e3 + d0 e4 in the example

A = C R C following Proposition 21 (where 1 = 1 1 is the identity of A).
1
1
Let 1 = (11+ii) and 2 = (11ii). Show that 1 2 = 0, 1 +2 = 1,
2
2
2
and j = j for j = 1, 2 (1 and 2 are called orthogonal idempotents in A.)
Deduce that A is isomorphic as a ring to the direct product of two principal
ideals: A
= A1 A2 (cf. Exercise 1, Section 7.6).
Prove that the map : C C C C by (z1 , z2 ) = (z1 z2 , z1 z 2 ), where z 2
denotes the complex conjugate of z2 , is an R-bilinear map.
Let be the R-module homomorphism from A to C C obtained from in
(c). Show that (1 ) = (0, 1) and (2 ) = (1, 0). Show also that is C-linear,
where the action of C is on the left tensor factor in A and on both factors in
C C. Deduce that is surjective. Show that is a C-algebra isomorphism.
ments

(b)

(c)
(d)

(a) We have

Solution.

(a 1 + b e2 + c e3 + d e4 )
(a0 1 + b0 e2 + c0 e3 + d0 e4 ) = (aa0 bb0 cc0 + dd0 ) 1
+ (ab0 + ba0 cd0 dc0 ) e2
+ (ac0 + ca0 bd0 db0 ) e3
+ (ad0 + da0 bc0 cb0 ) e4 .
(b) The rst part follows directly from the formula above.
decompose in this way is a general property: the map
by

The fact that rings

A A1 A2

is given

a 7 (a1 , a2 ).

(c) Just check it.

(d) We have

(1 1) = (1, 1) = (1, 1),

(i i) = (i, i) = (1, 1),
(1 ) = 12 ((1, 1) + (1, 1)) = (0, 1) and (2 ) = 12 ((1, 1) (1, 1)) = (1, 0).
To show that is C-linear, we compute

so

(z(z1 z2 )) = (zz1 zz2 ) = (zz1 z2 , zz1 z 2 ) = z(z1 z2 , z1 z 2 ),

as desired. Now,

is surjective because it's

C-linear,

and its image contains

a basis of the codomain. Finally, it's an isomorphism because it's a surjective

map of

C-vector

Problem (1). Let

spaces of the same (nite) dimension.

Z[X] denote the ring of polynomials with coecients in the integers,

2
2
denote the subring of all polynomials in X . Let Z denote the Z[X ]2
2
module whose underlying group is Z and where X acts by X n = 0 for all n.

and let

Z[X 2 ]

HOMEWORK 2 SOLUTIONS
Describe the abelian group structure of

Z[X] Z[X 2 ] Z,

5
i.e. provide a decomposition

of abelian groups

for some choice of

Z[X] Z[X 2 ] Z
= Zr Z/n1 Z/nk
r and integers ni .
Z[X] Z[X 2 ] Z
= Z2 .

2
To do this, we'll show that Z satises
2
2
the universal property for the tensor product. Let : Z[X] Z Z be the Z[X ]P
i
bilinear map dened by (
ai x , n) = (a1 n, a0 n). Now, let : Z[X] Z A be any
2
Z[X ]-bilinear map. We need to show that there's a unique map : Z2 A so that
= . The only possibility is to set (a, b) = (ax + b, 1). Hence, Z2 satises
Solution. We claim that

the desired universal property.

There was a fair amount of magic in that solution: we had to guess, somehow,
2
that Z t there, and then we had to dene the map carefully. We can, in fact,
do this problem a bit more systematically, using some properties of tensor products
of polynomial rings, as follows: Z[X]
= Z[X 2 , Y ]/(Y 2 X 2 ), and Z
= Z[X 2 ]/(X 2 ).
Hence we have

Z[X] Z[X 2 ] Z
= Z[X 2 , Y ]/(Y 2 X 2 ) Z[X 2 ] Z[X 2 ]/(X 2 )

= Z[X 2 , Y ]/(Y 2 X 2 , X 2 )

= Z[X 2 , Y ]/(Y 2 , X 2 )

= Z[Y ]/(Y 2 )

= Z2 .
Problem (2). Let

Z[ p1 ]

denote the abelian group of rational numbers whose denomi-

nators are powers of the prime

p.

Determine whether the group

 
 
1
1
Z
Z Z
/Z
p
p
is zero or nonzero. Prove your assertion.
Solution. The group is zero. It suce to check that any simple tensor is zero. Let

(a/pm ) (b/pn )

be a simple tensor. (The second term should really be

but we'll be sloppy with our notation.) We have

(a/pm ) (b/pn ) = (apn /pm+n ) (b/pn )

= pn (a/pm+n ) (b/pn )
= (a/pm+n ) (bpn /pn )
= (a/pm+n ) b
= (a/pm+n ) 0
= 0.

b/pn mod Z,

MATH 121

Hence, any simple tensor is zero, so the entire group is also zero.
This problem should be considered in analogy with Problem 10.4.8 above. In that
problem, we saw that when we tensor with

Q,

we annihilate all the torsion. Here,

we saw that when we tensor with Z[1/p], then we annihilate p-power torsion: every
n
element of Z[1/p]/Z is p torsion for some n. More generally, if we tensor with a ring
in which a prime
of

is invertible, we annihilate all

q -power

torsion. Hence, in the case

in which every prime is invertible, we annihilate torsion at all primes.

Problem (3). Prove the isomorphism

k[x] k k[y]
= k[x, y]

for any eld

k.

k -bilinear map : k[x] k[y] k[x, y] given by (p(x), q(y)) 7

p(x)q(y). Let : k[x] k[y] A be any k -bilinear map. We must show that
there is a unique k -module (or vector space) homomorphism : k[x, y] A so that
= . It suces to dene on monomials of the form xi y j , since we can then
i j
i j
extend it by linearity. We dene (x y ) = (x , y ); this clearly makes the diagram

Solution. We have a

commute. Furthermore, this is the only option to make the desired relation hold,
i j
i j
since (x , y ) = x y . Hence, k[x, y] satises the universal property for the tensor
product, so we have

k[x] k k[y]
= k[x, y],

Problem (4). Describe the

Q-vector

space

as desired.

Q[x]/(x) Q[x] Q[x]/(x2 + 1),

its dimension.
Solution. By Problem 10.4.16 above, we have

Q[x]
Q[x]
Q[x]
Q[x]

Q[x] 2
=
=
= 0.
2
(x)
(x + 1)
(x, x + 1)
Q[x]

i.e. compute