a
r
X
i
v
:
m
a
t
h
/
9
9
1
1
2
2
5
v
1
[
m
a
t
h
.
L
O
]
2
8
N
o
v
1
9
9
9
ON CARDINALITIES IN QUOTIENTS OF INVERSE
LIMITS OF GROUPS
SAHARON SHELAH AND RAMI GROSSBERG
Abstract. Let be
0
or a strong limit of conality
0
. Suppose
that G
m
,
m,n
: m n < ) and H
m
,
t
m,n
: m n < ) are
projective systems of groups of cardinality less than and suppose
that for every n < there is a homorphism : H
n
G
n
such
that all the diagrams commute.
If for every < there exists f
i
G
: i < ) such that
i ,= j = f
i
f
1
j
,
(H
) then there exists f
i
G
: i < 2
)
such that i ,= j = f
i
f
1
j
,
(H
).
1. Introduction
The main result of this paper was motivated by our interest
in the structure of the group Ext
p
(G, Z) for G abelian torsion free.
For basic results about the structure of Ext(G, Z) the reader is refered
to sections 47 and 52 of Laszlo Fuchs book [Fu], however all we need
is Denition 1.21 below. Since Shelahs proof of the independence of
Whiteheads problem of ZFC (see [Sh 44]) much was done since that
paper, for a summary see the introduction to [GrSh] and Chapter XII
of Eklof & Meklers book is dedicated ([EK]) to the structure of Ext.
In [GrSh] we have dealt with the cardinality of Ext
p
(G, Z). The
main Theorem of [GrSh] states that for a strong limit of conality
0
for every torsion free G of cardinality either
[Ext
p
(G, Z)[ < or [Ext
p
(G, Z)[ = 2
.
In section 2 of [GrSh] we indicated that the proof of the main
theorem can be adapted to give a result concerning cardinalities of
inverse systems of abelian groups subject to certain conditions (See
Theorem 1.1 below). We did not include a proof there. Recently we
Date: February 8, 2008.
Partially supported by the United States-Israel Binational Science Foundation
and by the National Science Fundation, NSF-DMS97-04477.
1
2 SAHARON SHELAH AND RAMI GROSSBERG
were asked to supply a complete proof to that theorem. Charles Meg-
ibben in a widely circulated preprint [Me] (which to our knowledge
did not appear yet in print) even claimed that he proved a result that
contradicts Theorem 1.1.
The aim of this paper is to present a complete proof of Theorem
1.1 below.
Notice that we do not make any assumptions on the groups, in
particular the groups need not be commutative and can be even locally
nite. See more on the subject in [Sh 664].
Theorem 1.1. [The Main Theorem] Suppose is
0
or it is strong
limit cardinal of conality
0
.
(1) Let G
m
,
m,n
: m n < ) be an inverse system of groups
of cardinality less than whose inverse limit is G
with
n,
such that [G
n
[ < . (
m,n
is a homomorphism from G
m
to
G
n
,
,
,
=
,
and
,
is the identity).
(2) Let I be a nite index set. Suppose that for every t I, H
t
m
,
t
m,n
:
m n < ) is an inverse system of groups of cardinality less
than and H
t
with
t
n,
be the corresponding inverse limit.
(3) Let for every t I,
t
n
: H
t
n
G
n
be a homomorphism such that
all diagrams commute (i.e.
m,n
t
n
=
t
m
t
m,n
for m n < ),
and let
t
be the induced homomorphism from H
t
into G
.
Assume that for every < there is a sequence f
i
G
: i <
) such that for i ,= j and t I f
i
f
1
j
/ Rang(
t
). Then there is
f
i
G
: i < 2
) such that i ,= j and t I f
i
f
1
j
/ Rang(
t
).
Notation 1.2. Since has conality
0
we can x
n
< for n <
such that =
n<
n
, for all n < ,
n
is regular and 2
n
<
n+1
<
and [G
n
[ +
tI
[H
t
n
[
n
.
Denote by e
G
, e
H
t
the unit elements. Without loss of generality
the groups are pairwise disjoint.
Denition 1.3. (1) For let H
tI
H
t
and H
<
=
<
H
, H
.
(2) For g H
let lev( g) = , for g H
t
let lev(g) = (without
loss of generality this is well dened).
ON CARDINALITIES IN QUOTIENTS OF INVERSE LIMITS OF GROUPS 3
(3) For , g H
t
let g H
t
=
t
,
(g) and we say g H
t
is below g and g is above g H
t
or extend g H
t
.
(4) For , f G
let f G
=
,
(f).
We will now introduce the rank function used in the proof of
Theorem 1.1, it is a measure for the possibility to extend functions in
Lemma 1.7 we show that it is an ultrametric valuation.
Denition 1.4. (1) For g H
t
n
, f G
we say that (g, f) is a nice
t-pair if
t
n
(g) = f G
n
.
(2) Dene a ranking function rk
t
(g, f) for any nice t-pair. First by
induction on the ordinal (we can x f G
), we dene when
rk
t
(g, f) simultaneously for all n < and every g H
t
n
(a) rk
t
(g, f) 0 i (g, f) is a nice t-pair
(b) rk
t
(g, f) for a limit ordinal i for every < we have
rk
t
(g, f)
(c) rk
t
(g, f) + 1 i (g, f) is a nice t-pair, and letting n =
lev(g) there exists g
H
t
n+1
extending g such that rk
t
(g
, f)
(d) rk
t
(g, f) 1.
(3) For an ordinal or 1 (stipulating 1 < < for any ordinal
) we have rk
t
(g, f) = i rk
t
(g, f) and it is false that
rk
t
(g, f) + 1.
(4) rk
t
(g, f) = i for every ordinal we have rk
t
(g, f) .
The following two claims give the principal properties of rk
t
(g, f).
Claim 1.5. Let (g, f) be a nice t-pair.
(1) The following statements are equivalent:
(a) rk
t
(g, f) =
(b) there exists g
H
t
extending g such that
t
(g
) = f.
(2) If rk
t
(g, f) < , then rk
t
(g, f) <
+
.
(3) If g
is a proper extension of g and (g
, f) is also a nice t-pair
then
(a) rk
t
(g
, f) rk
t
(g, f) and
(b) if 0 rk
t
(g, f) < then the inequality is strict.
Proof. (1) (a) (b): Let n be such that g H
t
n
. It is enough to
dene g
k
H
t
k
for k < , k n such that
(a) g
n
= g
(b) g
k
is below g
k+1
that is
t
k,k+1
(g
k+1
) = g
k
and
(c) rk
t
(g
k+1
, f) = :
4 SAHARON SHELAH AND RAMI GROSSBERG
Let g
:= lim
g
k
it is as required. The denition is by induction
on k n. For k = n let g
0
= g. For k n, suppose g
k
is dened.
By (iii) we have rk
t
(g
k
, f) = , hence there exists g
H
t
k+1
extending g
k
such that rk
t
(g
, f) = , and let g
k+1
:= g
.
(b) (a): Since g is below g
, it is enough to prove by induc-
tion on that for every k n when g
k
:= g
H
t
k
we have that
rk
t
(g, f) .
For = 0, since
t
(g
) = f G
n
clearly for every k we have
t
k
(g
k
) = f G
k
so (g
k
, f) is a nice t-pair.
For limit , by the induction hypothesis for every < and
every k we have rk
t
(g
k
, f) , hence by Denition 1.4(2)(b),
rk
t
(g
k
, f) .
For = + 1, by the induction hypothesis for every k n
we have rk
t
(g
k
, f) . Let k
0
n be given. Since g
k
0
is
below g
k
0
+1
and rk
t
(g
k
0
+1
, f) , Denition 1.4(2)(c) implies
that rk
t
(g
k
0
, f) +1; i.e. for every k n we have rk
t
(g
k
, f)
. So we are done.
(2) Let g H
t
n
and f G
be given. It is enough to prove that if
rk
t
(g, f)
+
then rk
t
(g, f) = . Using part (1) it is enough
to nd g
H
t
such that g is below g
and f =
t
(g
).
We dene by induction on k < , g
k
H
t
n+k
such that g
k
is
below g
k+1
and rk
t
(g
k
, f)
+
. For k = 0 let g
k
= g. For
k + 1, for every <
+
, as rk
t
(g
k
, f) > by 1.4(2)(c) there
is g
k,
G
n+k+1
extending g
k
such that rk
t
(g
k,
, f) . But
the number of possible g
k,
is [H
t
n+k+1
[ 2
n+k+1
<
+
hence
there are a function g and a set S
+
of cardinality
+
such
that S g
k,
= g. Now take g
k+1
= g.
(3) Immediate.
Lemma 1.6. (1) Let (g, f) be a nice t-pair. Then we have rk(g, f)
rk(g
1
, f
1
).
(2) For every nice t-pair (g, f) we have rk(g, f) = rk(g
1
, f
1
).
Proof. (1) By induction on prove that rk(g, f) rk(g
1
, f
1
)
(see more details in Lemma 1.7).
(2) Apply part (1) twice.
In the following lemma we show that the rank is indeed ultra-
metric (ordinal valued).
ON CARDINALITIES IN QUOTIENTS OF INVERSE LIMITS OF GROUPS 5
Lemma 1.7. Let n < be xed, and let (g
1
, f
1
), (g
2
, f
2
) be nice t-pairs
with g
H
t
n
( = 1, 2).
(1) If (g
1
, f
1
) and (g
2
, f
2
) are t-nice pairs, then (g
1
g
2
, f
1
f
2
) is a nice
pair and rk
t
(g
1
g
2
, f
1
f
2
) Minrk
t
(g
, f
) : = 1, 2.
(2) Let n, (f
1
, g
1
) and (f
2
, g
2
) be as above. If rk
t
(g
1
, f
1
) ,= rk
t
(g
2
, f
2
),
then rk
t
(g
1
g
2
, f
1
f
2
) = Minrk
t
(g
, f
) : = 1, 2.
Proof. (1) It is easy to show that the pair is t-nice. We show by
induction on simultaneously for all n < and every g
1
, g
2
H
t
n
that Minrk(g
, f
) : = 1, 2 implies that rk(g
1
g
2
, f
1
f
2
)
.
When = 0 or is a limit ordinal this is easy. Suppose
= + 1 and that rk(g
, f
) + 1; by the denition of rank
for = 1, 2 there exists g
H
t
n+1
extending g
such that (g
, f
)
is a nice pair and rk
t
(g
, f
) . By the induction assumption
rk
t
(g
1
g
2
, f
1
f
2
) . Hence g
1
g
2
is as required in the denition of
rk
t
(g
1
g
2
, f
1
f
2
) + 1.
(2) Suppose without loss of generality that rk(g
1
, f
1
) < rk(g
2
, f
2
),
let
1
= rk(g
1
, f
1
) and let
2
= rk
t
(g
2
, f
2
). By part (1),
rk
t
(g
1
g
2
, f
1
f
2
)
1
, by Proposition 1.6, rk
t
(g
1
2
, f
1
2
) =
2
>
1
.
So we have
1
= rk
t
(g
1
, f
1
) = rk
t
(g
1
g
2
g
1
2
, f
1
f
2
f
1
2
)
Minrk
t
(g
1
g
2
, f
1
f
2
), rk
t
(g
1
2
, f
1
2
)
= rk
t
(g
1
g
2
, f
1
f
2
)
1
.
Hence the conclusion follows.
Denition 1.8. (1) Let < and let =
t
: t I) where
t
is an ordinal less or equal to
+
. We say that
f = f
i
: i < )
-exemplies
n
(or
f is a -witness for
n
) i
(a) f
i
G
and f
i
G
n
= e
Gn
(b) for i ,= j and t I we have rk
t
(e
H
t
n
, f
i
f
1
j
) <
t
(possibly is
1).
(2) Let
n
=
: =
t
: t I),
t
an ordinal
+
,
and for every < there is a sequence f
i
: i < )
which -exemplies
n
.
6 SAHARON SHELAH AND RAMI GROSSBERG
(3)
n
=
n
: for no
we have
n
,
(i.e.
tJn
t
) and
,= .
Claim 1.9. (1)
n
is not empty.
(2)
n
is not empty in fact (
n
)(
n
)(
).
Proof. (1) Let
t
= suprk
t
(g, f)+1 : g H
t
n
, f G
and rk
t
(g, f) <
, by 1.5(2), this is a supremum on a set of ordinals <
+
(as
1 + 1 = 0) hence is an ordinal
+
. So
t
: t I) is as
required.
(2) If not, then choose by induction on < a sequence
n
such that
0
= ,
+1
+1
,=
. So for each t I, the
sequence
t
: < ) is a non-increasing sequence of ordinals
hence is eventually constant, say for some
t
< we have
[
t
, )
t
=
t
t
, so as I is nite, () = max
t
: t I < ,
so
()
=
()+1
, a contradiction.
Claim 1.10. (1) If
and f
i
: i <
),
-exemplify
n
and h :
is one to one, then f
h(i)
: i < ), -exemplies
n
.
(2) If f
i
: i < ), -exemplify
n
and f
i
G
n+1
= f for i < ,
then f
i
f
1
0
: i < ), -exemplify
n+1
.
(3) If
n
, then
n+1
.
(4) If
n
, then some
belongs to
n+1
.
(5) For some n < there is
mn
n
.
(6) In clause (b) of Denition 1.8(1) it suces to deal with i < j.
Proof. (1) Trivial.
(2) Clearly.
Clause (a):
(f
i
f
1
0
) G
n+1
=
n+1
(f
i
f
1
0
) = (
n+1
(f
i
))(
n+1
(f
0
))
1
=
ff
1
= e
G
n+1
.
Clause (b):
For i ,= j and t I, note that
(f
i
f
1
0
)(f
j
f
1
0
) = f
i
f
1
0
f
0
f
1
j
= f
i
f
1
j
so we can use the assumption.
(3) So let < and we should nd a -witness for
n+1
. We can
choose
such that [G
n+1
[ <
< . As
n
, clearly there
is a
-witness f
i
: i <
) for it. Now the number of possible f
i
ON CARDINALITIES IN QUOTIENTS OF INVERSE LIMITS OF GROUPS 7
G
n+1
is [G
n+1
[ (really) even [Rang(
n+1,
) Ker(
n,n+1
)[)
hence for some f G
n+1
and Y
we have: [Y [ and
i Y f
i
G
n+1
= f. By renaming i : i < Y , now
f
i
f
1
0
: i < ) is a -witness by part (1).
(4) Follows by 1.10(2) and 1.9(2).
(5) By 1.10(3) by the well foundedness of the ordinals (as in the
proof of 1.9(2),(8).
(6) Because for i < j, (f
j
f
1
i
)
1
= (f
i
f
1
j
) and 1.6(2).
Convention 1.11. By renaming and 1.10(4), without loss of general-
ity
n
for every n.
Claim 1.12. Each
t
(t I) is a non-successor ordinal (i.e. limit or
zero).
Proof. Fix n < .
Assume s I is a counterexample. So
s
=
+ 1,
0.
Let
=
t
: t I) be dened as follows:
t
is
t
if t ,= s and is
if t = s. We shall prove that
n+1
thus getting a contradiction.
So let < and we shall nd a -witness for
n+1
. Let
be
such that [G
n+1
[ <
< . As
n
(see 1.11) there is a
-
witness f
i
: i <
) for
n
, as earlier without loss of generality
i < f
i
G
n+1
= f for some f. We shall prove that f
i
f
1
0
: i < )
is a -witness for
n+1
. Let f
i
= f
i
f
1
0
for i < .
Clause (a):
f
G
n+1
= (f
0
f
1
0
) G
n+1
= e
G
n+1
because f
i
G
n+1
= f
0
G
n+1
.
Clause (b):
Let i ,= j < . If t Is then
rk
t
(e
G
n+1
, f
i
(f
j
)
1
) = rk
t
(e
G
n+1
, f
i
f
1
j
) rk
t
(e
Gn
, f
i
f
1
j
)
t
=
t
.
(Why? By group theory, by 1.5(3)(), by choice of
f, by choice of
t
,
respectively).
If t = s, then rk
t
(e
Gn
, f
i
f
1
j
) < rk
t
(e
G
n+1
, f
i
f
1
j
) by 1.5(3)(),
and proceed as above.
8 SAHARON SHELAH AND RAMI GROSSBERG
Notation 1.13. For let T
:=
k<
k
, T :=
n<
T
n
(note:
treeness used).
Claim 1.14. There are for n < , a sequence f
n,i
: i <
n
) and an
ordinal
t
n
<
t
(
t
is the ordinal from 1.11) such that
(1) f
n,i
G
, f
n,i
G
n+1
= e
G
n+1
for all i <
n
;
(2) for each t I for every h H
t
n
and i < j <
n
we have:
rk
t
(h, f
n,i
f
1
n,j
)
t
n
;
(3) rk
t
(e
H
t
n
, f
n,i
f
1
n,j
)
t
n1
for i < j <
n
and
t
n1
0 rk
t
(e
H
t
n
, f
n,i
f
1
n,j
) >
t
n1
(4)
t
n1
<
t
n
if
t
> 0 and
t
n
= 1 if
t
= 0.
We delay the rest of proof for a while.
Convention 1.15. Let
t
n
, g
n,i
(n < , i <
n
) be as in 1.14.
Denition 1.16. We set f
= g
n1,(n1)
g
n2,(n2)
. . . g
0,(0)
for
T
n
. Then dene f
for T
as follows: f
is the element of G
satisfying f
G
n
= f
n
. It is well dened by:
Fact 1.17. (1) For T
and m n < we have
f
n
G
n+1
= f
m
G
n+1
.
(2) For T
we have f
is well dened (as the inverse limit
of f
n
G
n
: n < ), so n < f
G
n
= f
n
.
Proof. (1) As
n,
is a homomorphism it is enough to prove
(f
n
(f
m
)
1
) G
n+1
= e
G
n+1
, hence it is enough to prove
n k < (f
k
f
1
(k+1)
) G
n+1
= e
G
n+1
which follows from
k < f
k
f
1
(k+1)
G
k+1
= e
G
k+1
, which means
f
k,(k)
G
k+1
= e
G
k+1
which holds by clause (a) of 1.11.
(2) Follows by part (1) and G
being an inverse limit.
Proposition 1.18. Let , T
. If ,= and t I, then f
f
1
(H
t
).
ON CARDINALITIES IN QUOTIENTS OF INVERSE LIMITS OF GROUPS 9
Proof. Suppose for the sake of contradiction that for some g H
t
we
have
t
(g) = f
f
1
.
Let k be minimal such that k = k, (k) ,= (k),
without loss of generality (k) < (k). For k let
be rk
t
(g
H
t
, f
(+1)
f
1
(+1)
). We will reach a contradiction by showing that
k 0
t
k
and > k
+1
<
.
Note
()
1
if , then rk
t
(g H
t
, f
f
1
) 0 as
t
(g H
t
) =
t
(g)
G
t
= (f
f
1
) F
t
and 1.17.
For = k, we show that
k
t
k
. Let i = [k], j = [k].
By the choice of k, i ,= j. In this case f
(+1)
f
1
(+1)
= f
k,(k)
f
1
k,(k)
by the minimality of k and, of course, f
k,(k)
f
1
k,(k)
= f
k,i
f
1
k,j
, hence
k
= rk
t
(g H
h
k
, f
k,i
f
1
k,j
)
k
by clause (b) of 1.14. Note: if
t
=
0, then
t
m
= 1 for m < hence
k
= 1, but (f
f
1
) G
k
=
(f
(k+1)
f
1
(k+1)
) G
k
immediate contradiction. So assume
t
0
hence 0
t
m
<
t
m+1
.
Now we proceed inductively. We assume that
k
and show
that
+1
<
. Let i = [ + 1], j = [ + 1], and let
= rk
t
(g H
t
+1
, f
(+1)
f
1
(+1)
). Observe:
()
2
< rk
t
(g H
t
, f
(+1)
f
1
(+1)
) =
[why? by 1.5(3) and ()
1
above.]
So
()
3
+1
= rk
t
(g H
t
+1
, f
(+2)
f
1
(+2)
)
= rk
t
(g H
t
+1
, f
+1,(+1)
(f
(+1)
f
1
(+1)
)f
+1,(+1)
)
= rk
t
(e
H
t
+1
(g H
t
+1
)e
H
t
+1
, f
+1,(+1)
(f
(+1)
f
1
(+1)
)f
+1,(+1)
).
Now:
()
4
rk
t
(e
H
t
+1
, f
+1,(+1)
) >
t
(why? by clause (c) of 1.14)
()
5
rk
t
(g H
t
+1
, f
(+1)
f
1
(+1)
) =
k
t
k
t
(why? the equality by the denition of
, the rst inequality by the
induction hypothesis and the second inequality was proved above (for
= k), the last inequality by 1.14 clause (d)
10 SAHARON SHELAH AND RAMI GROSSBERG
()
6
rk
t
(e
H
t
+1
, g
+1,(+1)
) >
t
(why? by clause (c) of 1.14).
Hence by 1.5(3)
()
7
rk
t
(e
H
t
+1
(g H
t
+1
)e
H
t
+1
, f
+1,()
(f
(+1)
f
1
(+1)
)f
+1,(+1)
)
= rk(g G
+1
, f
(+1)
f
1
(+1)
).
Together we get the induction demand for + 1.
Before proving 1.14 and nishing we prove
Claim 1.19. Assume 1
t
<
t
for t I and n < and < .
Then we can nd f
i
: i < ) such that
(1) f
i
G
and f
i
G
n+1
= e
G
n+1
(2) t I and i ,= j rk
t
(e
H
t
n
, f
i
f
1
j
) [
t
,
t
)
(3) t I and i < = rk
t
(e
H
t
n
, f
i
) [
t
,
t
).
Proof. For each s I we dene
s
=
s
t
: t I) by:
s
t
=
t
if t ,= s
t
if t = s
So
s
,=
, so as
n<
m
necessarily
s
/
n
, hence for
some
s
< there is no
s
-witness for
s
and n (check the denition
of
n
).
Let
1
< be > + max
s
: s I.
Let < be large enough (so that it will be possible to use the
nite Ramsey theorem when =
0
and when >
0
the Erdos Rado
theorem we require that (
1
)
2
where = 2
t
|H
t
n
|
).
Let f
i
: i < ) be a -witness
n
and even
n+1
. For
each t I, h H
t
n
dene the two place function F
t,h
from []
2
to 0, 1
for i < j < let
F
t,h
i, j :=
0 if rk
t
(h, f
i
f
1
j
) <
t
1 Otherwise.
Dene the two-place function F from []
2
: For i < j < let
Fi, j = F
t,h
(i, j) : t I, h H
t
n
).
Clearly [Rang(F)[ 2
t
|H
t
n
|
.
ON CARDINALITIES IN QUOTIENTS OF INVERSE LIMITS OF GROUPS 11
Hence an application of one of the above partition theorems
provides us with a set Y , [Y [ =
1
such that F [Y ]
2
is constant.
Without loss of generality Y =
1
.
For each s I, clearly f
i
f
1
0
: i <
s
) is not a
s
-witness for
s
, but the only thing that may go wrong is the inequality, i < j <
s
rk
s
(e
H
s
n
, f
i
f
1
j
) <
s
, so for some i < j <
s
we have that
rk
s
(e
H
s
n
, f
i
f
1
j
)
s
holds, hence
() s I and i < j <
1
rk
s
(e
H
s
n
, f
i
f
1
j
)
s
.
This means clause (b) holds and clause (a) by denition of f
i
:
i < ) is a -witness for
n
. Clause (c) follows. So f
i
: i < ) is
as required.
Proof. of 1.14
Stipulate
t
1
: if
t
> 0 it is 0, otherwise is it 1. Assume n <
and
t
n1
: t I) is well dened,
t
n1
<
t
. Let
t,
n
be:
t
n1
+1 if
t
is a limit ordinal and
t
n1
= 1 otherwise (i.e.
t
= 0, see 1.12). Note
that to construct the family f
n,i
: i <
n
we will combine Claim 1.19
with a second application of the Erdos Rado Theorem.
Let = (2
|H
t
n
||H
t
n
|
) [I[ and < be such that (
n
+2)
3
(exists by Ramsey theorem if =
0
and by Erdos Rado theorem if
>
0
). Apply Claim 1.19 to get a family f
i
: i < satisfying:
(1) f
i
G
n+1
= e
G
n+1
,
(2) for i ,= j and t I, we have
t,
n1
rk
t
(e
H
t
n
, f
i
f
1
j
) <
t
.
For t I, g = g
1
, g
2
), g
1
, g
2
H
t
n
such that
t
n
(g) = e
Gn
dene
a coloring F
t, g
of [I]
3
by two colors according to the following scheme:
for < < < , let
F
t,g
, , :=
red if rk
t
(g
1
, f
i
f
1
) rk
t
(g
2
, f
f
1
);
green if rk
t
(g
1
, f
i
f
1
) > rk
t
(g
2
, f
f
1
)
.
By the Ramsey theorem (if =
0
) or Erdos Rado Theorem if >
0
there is a set J , otp(J) =
n
+2 such that each coloring is constant
on [J]
3
. Let the value of F
t, g
on [J]
3
be denoted c
t, g
. Observe that c
t, g
is
never green as this would produce a descending -sequence of ordinals
as if
J,
<
+1
for < , then rk
t
(g, f
f
1
+1
) > rk
t
(g, f
+1
f
1
+2
),
so rk
t
(g, f
2
f
1
2+1
) : < ) is strictly decreasing.
12 SAHARON SHELAH AND RAMI GROSSBERG
Let () = Min(J) and J
0
= J : otp( J) < and
is the
n
-th member of J, the (
n
+ 1)-th member of J and let
t
n
= rk
t
(e
H
t
n
, f
f
1
), by clause (b) above
t,
n
t
n
<
t
so
t
= 0
t
n
= 1 and
t
> 0
t
n1
<
t
n
.
We claim that f
i
f
1
()
: i J
0
(remember J
0
J, [J
0
[ =
n
)
provides a set that can play the role of f
n,i
: i <
n
. We note
()
1
rk
t
(g, f
f
1
)
n
t
for < in J
0
[why? clearly < < < are
in J hence by the choice of J we have rk
t
(g, f
f
1
) rk
t
(g, f
f
1
)
rk
t
(g, f
f
1
) =
t
n
].
Now clauses (1), (4) of 1.14 holds by clause (1) above, clause
(3) of 1.14 holds by ()
1
and clause (4) of 1.14 holds by the choice of
the
t
. We are left with clause (2). Let h H
t
n
, as above clearly for
< < < in J we have rk
t
(h, f
f
1
) rk
t
(h, f
f
1
). Hence for
< < in J
0
we have
n
t
rk
t
(e
H
t
n
, f
f
1
)
= rk
t
(h
1
, (f
f
1
)(f
f
1
)
1
)
Minrk
t
(h, f
f
1
), rk
t
(h
1
, (f
f
1
)
1
= Minrk
t
(h, f
f
1
), rk
t
(h, f
f
1
)
Minrk
t
(h, f
f
1
), rk
t
(h, f
f
1
)
= rk
t
(h, f
f
1
).
So giving also clause (2) of 1.14.
1.1
Remark 1.20. The result about the cardinality of Ext
p
(G, Z) can be
derived from Theorem 1.1 using the following denition (which con-
structs an isomorphic group ot Ext
p
(G, Z)).
Denition 1.21. Given an abelian group G, let G
:= Hom(G, Z)
and for a prime p denote by G
p
the group Hom(G, Z/pZ). For g G
let g g/p be the natural homomorphism from G
into G
p
. By G
/p
denote the subgroup of G
p
which is the image ofG
/p under g g/p.
Finally
Ext
p
(G, Z) := G
p
/(G
/p).
ON CARDINALITIES IN QUOTIENTS OF INVERSE LIMITS OF GROUPS 13
Recall that when is
0
or strong limit of conality
0
then
0
= 2
.
The group H
corresponde to the subgroup G
/p and the s
are inclusions.
We have learned from Paul Eklof that Christian U. Jensen in
his book [Jen] have a proof of Theorem 1.0 of [GrSh] for the case that
=
0
.
References
[EK] Paul C. Eklof and Alan H. Mekler. Almost Free Modules, set-theoretic
methods. North-Holland Publishing .Co., Amsterdam, 1990.
[Fu] Lazslo Fuchs, Innite Abelian Groups I, Academic Press 1970
[Gr] Rami Grossberg. A course in model theory. Book in preparation.
[GrSh] Rami Grossberg and Saharon Shelah. On the structure of Ext
p
(G, Z). Jour-
nal of Algebra, 121:117128, 1989.
[Je] Thomas Jech. Set Theory. Academic Press, New York, 1978.
[Jen] Christian U. Jensen. Les Foncteurs Derives de lim
et leurs Applications
en Theorie des Modules. Springer-Verlag, Lecture Notes in Mathematics
volume 254, 1972.
[Me] Charles Megibben. ..... A preprint.
[Sh 44] Saharon Shelah. Innite abelian groups, Whitehead problem and some con-
structions. Israel Journal of Mathematics, 18:243256, 1974.
[Sh 664] Saharon Shelah. On strong dichotomy of cardinality, in preparation.
E-mail address, Saharon Shelah: [email protected]
Institute of Mathematics, The Hebrew University, Jerusalem,
Israel, & Rutgers University, Mathematics Department, New Brunswick,
NJ 08903 USA
E-mail address, Rami Grossberg: [email protected]
Department of Mathematical Sciences, Carnegie Mellon Uni-
versity, Pittsburgh, PA 15213
