You are on page 1of 2

Proof of Proposition 3F.

12
We begin with two preliminary observations which are consequences of Proposition 3F.11:
(a) An inclusion B A induces a surjection Ext(A, Z) Ext(B, Z). Hence Ext(A, Z) is
uncountable if Ext(B, Z) is.
(b) If A A/B is a quotient map with B nitely generated, then the rst term in the
exact sequence Hom(B, Z) Ext(A/B, Z) Ext(A, Z) is countable, so Ext(A, Z) is
uncountable if Ext(A/B, Z) is.
There are two explicit calculations that will be used in the proof:
(c) If A =

Z
p
, a direct sum of innitely many nite cyclic groups Z
p
for various p, then
Ext(A, Z) is uncountable since it is isomorphic to

Ext(Z
p
, Z)

Z
p
.
(d) If p is prime then from Example 3F.10 we have Ext(Z
p
, Z)

Z
p
which is uncountable.
Consider now the map A A given by a pa for a xed prime p. Denote the kernel,
image, and cokernel of this map by
p
A, pA, and A
p
, respectively. The functor A A
p
is the same as A A Z
p
. We call the dimension of A
p
as a vector space over Z
p
the
p-rank of A.
Suppose the p-rank of A is innite. Then Ext(A
p
, Z) is uncountable by (c). From the exact
sequence 0 pA A A
p
0 we have Hom(pA, Z) Ext(A
p
, Z) Ext(A, Z) exact,
so either Hom(pA, Z) or Ext(A, Z) is uncountable. Also, the exact sequence 0
p
A
A pA 0 gives an exact sequence 0 Hom(pA, Z) Hom(A, Z) Hom(
p
A, Z)
whose last term is 0 since
p
A is a torsion group. Hence Hom(pA, Z) Hom(A, Z). Thus
we have shown that either Hom(A, Z) or Ext(A, Z) is uncountable if A has innite p-rank
for some p.
In the remainder of the proof we will show that Ext(A, Z) is uncountable if A has nite
p-rank for all p (and is not nitely generated).
Let C be a nontrivial cyclic subgroup of A, either nite or innite. If there is no maxi-
mal cyclic subgroup of A containing C then there is an innite ascending chain of cyclic
subgroups C = C
1
C
2
. If the indices [C
i
: C
i1
] involve innitely many distinct
prime factors p then A/C contains an innite sum

Z
p
for these p so Ext(A/C, Z) is
uncountable by (a) and (c) and hence also Ext(A, Z) by (b). If only nitely many primes
are factors of the indices [C
i
: C
i1
] then A/C contains a subgroup Z
p
so Ext(A/C, Z)
is uncountable in this case as well by (a), (b), and (d). Thus we may assume that each
nonzero element of A lies in a maximal cyclic subgroup.
If A has positive nite p-rank we can choose a cyclic subgroup mapping nontrivially to
A
p
and then a maximal cyclic subgroup C containing this one will also map nontrivially
to A
p
. The quotient A/C has smaller p-rank since C A A/C 0 exact implies
C
p
A
p
(A/C)
p
0 exact, as tensoring with Z
p
preserves exactness to this extent.
1
By (b) and induction on p-rank this gives a reduction to the case A
p
= 0, so A = pA.
If A is torsionfree, the maximality of the cyclic subgroup C in the preceding paragraph
implies that A/C is also torsionfree, so by induction on p-rank we reduce to the case that
A is torsionfree and A = pA. But in this case A has no maximal cyclic subgroups so this
case has already been covered.
If A has torsion, its torsion subgroup T is the direct sum of the p-torsion subgroups T(p)
for all primes p. Only nitely many of these T(p)s can be nonzero, otherwise A contains
nite cyclic subgroups not contained in maximal cyclic subgroups. So we may assume only
nitely many T(p)s are nonzero. If some T(p) is not nitely generated then by (a) we can
assume A = T(p). In this case the reduction from nite p-rank to p-rank 0 earlier in the
proof stays within the realm of p-torsion groups. But if A = pA we again have no maximal
cyclic subgroups, so we are done in the case that T is not nitely generated. Finally, when
T is nitely generated then we can use (b) to reduce to the torsionfree case by passing
from A to A/T.
2

You might also like