(

7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


ON NICE EQUIVALENCE
RELATIONS ON

2
SH724
Saharon Shelah
The Hebrew University of Jerusalem
Einstein Institute of Mathematics
Edmond J. Safra Campus, Givat Ram Jerusalem 91904, Israel
Department of Mathematics Hill Center-Busch Campus
Rutgers, The State University of New Jersey
Piscataway, NJ 08854-8019 USA
Abstract. Let E be an equivalence relation on the powerset of an uncountable
set, which is reasonably denable. We assume that any two subsets with symmetric
dierence of size exactly 1 are not equivalent. We investigate whether for E there
are many pairwise non-equivalent sets.
2000 Mathematics Subject Classication. (2000); FILL!! 03E47,03E35; 20K94.
Key words and phrases. set theory; denable equivalence relations, generalizing descriptive set
theory to uncountable cardinals; perfect sets of pairwise non-equivalent; abelian groups, Ent.
I would like to thank Alice Leonhardt for the beautiful typing
Written 99/11
First Typed - 2000/Feb/12; (revised with proofreading for the journal and little more)
Latest Revision - 03/Jan/29
Typeset by A
M
S-T
E
X
1
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


2 SAHARON SHELAH
Annotated content
0 Introduction
1 Dichotomical results on nice equivalence relations
[Assume E is a
1
1
[]-equivalence relation on

2 such that , are not E-
equivalent whenever they dier in exactly one place. Assume further that
this holds even after adding a -Cohen subset of . If =
<

(alternatively, E is more nicely dened or other requirement on ) then E

has a perfect set (so 2

elements) of pairwise non E-equivalent members of

2. There are related results.]

2 Singular of uncountable conality
[Assume =
<
> cf() = >
0
. We nd on

quite nice equivalence
relations for which the parallel of the results of 1 fail badly. If is strong
limit, we can use

2.]
3 Countable conality: positive results
[Assume that > cf() =
0
and is the limit of measurables, or just
a related property (which consistently holds for

) is satised. We
prove the parallel of the result in 1 on

.]
4 The countable conality case: negative results
[We show that if our universe is far enough from large cardinals (and close
to L) then we can build counterexamples as in 2.]
5 On r
p
(Ext(G, Z))
[We return to the p-rank of the abelian group Ext(G, Z) where G is tor-
sion free abelian group (
1
-free, without loss of generality see [Fu](xxx)).
We show that if is compact, strong limit (singular) cardinal > and
r
p
(Ext(G, Z)) then r
p
(Ext(G, Z)) 2

. This is preserved by adding

Cohens, super-compact. If GCH holds above we have a complete
characterization of Ext(G, Z) : G.]
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


ON NICE EQUIVALENCE RELATIONS ON

2 SH724 3
0 Introduction
The main topic here is the possible generalizations of the following theorem from
[Sh 273] on simple equivalence relation on

2 to higher cardinals.
0.1 Theorem. 1) Assume that
(a) E is a Borel 2-place relation on

2
(b) E is an equivalence relation
(c) if ,

2 and (!n)((n) ,= (n)), then , are not E-equivalent.
Then there is a perfect subset of

2 of pairwise non E-equivalent members.
2) Instead of E is Borel, E is analytic (or even a Borel combination of analytic
relations) is enough.
3) If E is a
1
2
relation which is an equivalence relation satisfying clauses (b) +
(c) also in V
Cohen
, then the conclusion of (1) holds.
In [Sh 273], Theorem 0.1 was used to prove a result on the homotopy group: if X
is a Hausdor metric topological space which is compact, separable, arc-connected,
and locally arc-connected, and the homotopy group is not nitely generated then
it has the cardinality of the continuum; the proof of 0.1 used forcing in [Sh 273],
see [PaSr98] without the forcing.
We may restrict E to be like the natural equivalence relation in presenting
r
p
(Ext(G, Z)) or just closer to group theory as in Grossberg Shelah [GrSh 302],
[GrSh 302a], Mekler-Roslanowski-Shelah [MRSh 314], [Sh 664]. In 5 we say some-
what more. We here continue [Sh 664] but do not rely on it.
Turning to

2 the problem split according to the character of and the simplic-
ity of E. If E is
1
1
and =
<
and

(or just (Dl)

holds), a generalization
holds. If E is
1
1
and =
<
, the generalization in general fails; all this in 1. Now
if is singular, strong limit for simplicity, it is natural to consider
cf()
instead of

2. If has uncountable conality we get strong negative results in 2. If has

countable conality, and is the limit of somewhat large cardinals, e.g. measurable
cardinals, (but =

may be O.K., i.e., consistently) the generalization holds (in

3), but if the universe is close to L (e.g. in L there is no weakly compact) then
we get negative results (see 4). Note that theorems of the form if E has many
equivalence classes it has continuum many equivalence classes do not generalize
well, see [ShVs 719] even for weakly compact.
We thank Alex Usuyatsov for many helpful comments and corrections.
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


4 SAHARON SHELAH
0.2 Denition. For a cardinal let B

be

2 (or

or
cf()
); we write B for
such set.
1) For a logic L we say that E is a L-nice, (say 2-place for simplicity), relation
on B if there is a model M with universe and nite vocabulary , and unary
function symbols F
1
, F
2
/ (denoting possibly partial unary functions), such that
letting
+
= F
1
, F
2
, for some sentence = (F
1
, F
2
) in L(
+
) we have
for any
1
,
2
B letting M

1
,
2
= (M,
1
,
2
) be the
+
-model expanding
M with F
M

1
,
2

for = 1, 2 we have

1
E
2
(M,
1
,
2
) [= .
We may write M [= [
1
,
2
] and [
1
,
2
, M] or (x, y, M) or write a
coding M instead of M.
2) E is a
1
1
-relation on Bmeans that above we allow to be of the form (X )
where is rst order or even in inductive logic (i.e., we have variables on sets and
are allowed to form the rst x point for formula (x, X) such that (x, X
1
) &
X
1
X
2
(x, X
2
)); if we allow just rst order we say strictly if we allow
formulas from L we say L-strictly. Similarly
1
1
,
1
2
, projective; writing nice
means L is L(induction) i.e. rst order + denition by induction. We may write
E nice(B

),
1
1
[B] etc, and may replace B by if this holds for every B =

2.
We write very nice for L-nice when L is L rst order logic.

0.3 Notation:
(

i < ) means for every large enough i < .

J
bd

is the ideal of bounded subsets of .

L denotes a logic, L() denotes the language (i.e, a set of formulas, for the logic
L in the vocabulary ), L denotes rst order logic, L
,
denotes the extension of
L by allowing

<()

(when () < ) and (x

0
, . . . , x
i
)
i<()
for () < .
We note the obvious (by now) relation
0.4 Fact. 1) If =
<
and R is a [strict] L

+
,
-nice relation on B

then R is a
[strict]
1
1
-relation and also a [strict]
1
1
-relation (with parameter a relation of ,
of course). If >
0
, L

+
,
= L

+
,
(induction).
2) If R is a L

+
,
(induction)-nice relation on B

and >
0
, then R is a L

+
,
-
[strict]
1
1
-relation on B

and
1
1
-relation on B.
3) If cf() >
0
then if R is L(induction)-nice relation on B

then R is strictly

1
1
-nice (hence being
1
1
is equivalent to being strictly
1
1
).
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


ON NICE EQUIVALENCE RELATIONS ON

2 SH724 5
Proof. 1) The quantication on X can code the satisfaction relation for any
subformula.
2) Easy.
3) It is well known that a linear order <

on such is a well ordering i for every

< , <

: < is isomorphic to (, <) for some < (e.g. [Na85]).

0.4
0.5 Denition. Let (D)

means that is regular, uncountable and there is a

sequence

P = P

: < ) such that P

is a family of < subsets of and for

every X the set < : X P

is stationary; hence =
<
. (By [Sh
460], =
<

(D)

and (by Kunen) =

+
(D)

).
0.6 Denition. Q

2 is called perfect or -perfect if:
(a) Q ,=
(b) if Q then g( ) : Q is an unbounded subset of
(c) the set : Q and is closed under the union of -increasing
sequences.
Equivalently, Q =

:

2 such that
(a)

2
(b)

1
,=
2


2

2
,=

2
(c)

if
0
,
1
,
2


2 are distinct and (
1

2
) (
1

0
) (so
1

2
,=
1

0
)
then (

2
) (

0
) and

1
(g(

2
)) =
1
(g(
1

2
)).
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


6 SAHARON SHELAH
1 Dichotomical results on nice equivalence relations on

2
We here continue [Sh 664, 2], the theorem and most proofs can be read without
it. The claims below generalize [Sh 273].
1.1 Claim. Assume

1
(a) =
<
and

or just (Dl)

(see 0.4)
(b) E is a nice 2-place relation on

2
(c)() E is an equivalence relation on

2
() if ,

2 and (! < )(() ,= ()) then (E).
Then E has 2

equivalence classes, moreover a perfect set of pairwise non E-

equivalent members of

2.
Proof. Note that
If P is a -complete forcing (or just -strategically complete) then
P
clauses (c), (), () are still true.
So we can apply 1.2 below.
1.1
A relative is
1.2 Claim. Assume

2
(a), (c) as in
1
(b) E is a
1
1
[] 2-place relation on

2, say dened by (Z)(x, y, Z, a)
see Denition 0.2
(c)
+
= (c)
+
Cohen
if P = (
>
2, ), i.e. -Cohen, then in V
P
clauses (c) from 1.1 still hold.
Then the conclusion of 1.1 holds.
Proof. Stage A: Let (
0
,
1
)

2

2 be generic over V for the forcing Q =
(
>
2) (
>
2) ordered naturally, i.e., (
0
,
1
) (
0
,
1
) i
0

0
&
1

1
. Now
do we have V[
0
,
1
] [=
0
E
1
? If so, then for some (p
0
, p
1
) (
>
2) (
>
2) we
have (p
0
, p
1
)
Q

0
E

1
, let < be > g(p
0
), g(p
1
) and by clause (c)
+
() in
V[
0
,
1
] we can nd

1


2 such that

1
=
1
, and for some (, ),

1

[, ) =
1
[, ), (here = +1 is O.K. but not so in some generalizations) and
V[
0
,
1
] [= (

1
E
1
).
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


ON NICE EQUIVALENCE RELATIONS ON

2 SH724 7
So V[
0
,
1
] [= (
0
E

1
) (again as in V[
0
,
1
], E is an equivalence relation
by clause (c)
+
and we are assuming for the time being that V[
0
,
1
] [=
0
E
1
).
But also (
0
,

1
) is generic over V for (
>
2) (
>
2) with (p
0
, p
1
) in the generic set
and V[
0
,
1
] = V[
0
,

1
] so we get a contradiction to (p
0
, p
1
) (

0
E

1
). Hence

1

(
>
2)(
>
2)
(

0
E

1
).
Stage B:
Let be large enough and let N (H(), ) be such that |N| = , N
<
N
and the denition of E belongs to N. Note that

2
if (
0
,
1
) (

2) (

2) (and is in V) and N[
0
,
1
] [= (
0
E
1
), then
(
0
E
1
).
[Why? As E is
1
1
, in N[
0
,
1
], there is a witness

2 for failure, and it
also witnesses in V that (
0
E
1
).]
Clearly to nish proving 1.1, it suces to prove
1.3 Subclaim. 1) Assume =
<
and (Dl)

.
If H () N, N
<
N, |N| = and N [= ZFC

, then there is a perfect

Q

2 such that for any
0
,=
1
from Q the pair (
0
,
1
) is generic over N for
[(
>
2) (
>
2)]
N
.
2) Assume that is regular and
(a) T is a tree with levels each of cardinality < and 2

-branches (or just

) and
(b)

N = N

: < ) is -increasing,

N ( + 1) N
+1
, T N
0
and
N

, |N

| < and N =

<
N

and T

N
+1
(if is regular it is
enough that

N ( + 1) N, T

N)
(c) <

is a well ordering of N such that <

N
+1
.
Then for some X

2, [X[ = 2

(or just [X[ = ) and

0
,=
1
X the pair
(
0
,
1
) is generic over N for (
>
2) (
>
2).
3) Like part (2) but we weaken clause (a) to
(a)

T is a tree with levels each of cardinality and

Y = Y

: ), Y

is a set of < nodes of T of level if < and a set of -branches of T

if = and [Y

[ and ,= Y

( < )( , Y

).
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


8 SAHARON SHELAH
1.4 Remark. A tree T as in clause (a) of 1.3(2) is called a -Kurepa tree and much
is known on its existence (and non existence). E.g. if is strong limit then such
T exists.
Proof. 1) Let P

: < ) be such that P

P(), [P

[ < , and for every

X the set : X P

is stationary. So by coding we can nd P

(
0
,
1
) :
0
,
1


2 of cardinality < such that for every
0
,
1


2 the set
< : (
0
,
1
) P

is stationary. Lastly, let I

: < ) list the dense

open subsets of (
>
2) (
>
2) which belong to N. Now we dene by induction on
< ,

:

2) such that:
(a)

>
2
(b) < g()

(c)

(d) if is a limit ordinal and (

0
,
1
) P

,
0
< 2,
1
< 2 and
0

0
) , =
1

1
)
then (

.
There is no problem to carry the denition (using [P

[ < = cf()) and

<

:


2 is a perfect set as required.
2) Similar. We choose by induction on ,

: T

) such that (a),(b),(c) above

hold and
(d)

if
0
,=
1
are in T
+1
then (

0
,

1
) I : I is a dense open subset of
P and belongs to N

(e) if

: T

) is the <

-sequence satisfying (a)-(d).

So

: T

) can be dened from N

: < ).
The proof in part (2) is easier as we can assume that such a tree belongs to N.
3) Left to the reader.
So we have nished proving claim 1.2 hence claim 1.1.
1.3
,
1.2
1.5 Claim. 1) In claims 1.1, 1.2 we can weaken clause () (in (c), (c)
+
, call it
(c)

, (c)

respectively) to:
()

if

2 and < then for some (, ) and
[,)
2 the sequences
, (( ) [, )) are not E-equivalent .
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


ON NICE EQUIVALENCE RELATIONS ON

2 SH724 9
2) In claims 1.1, 1.2 and in 1.5(1), for any

we can replace E by E

: <

),
each E

satisfying clauses (b) and (c), (c)

+
, (c)

, (c)

there respectively and we

strengthen the conclusion:
() there is a -perfect set Q such that
() Q =

:

2 and
() if
1
,=
2
are from

2 then

1
,=

2
and <

1
E

2
)
() for

2 the set g(

) :

2 is a closed unbounded
subset of .
3) In 1.2, 1.5(1),(2) we can weaken (c)
+
or (c)

to
() for a stationary set of N [H (
+
)]

there is (in V)

2 which is
Cohen over N such that
1
1
[] sentences are absolute from N[] to V (for

1
1
[]-sentences this is necessarily true) and clause (c) (or (c)

) holds.
Proof. 1), 2) The same as the proof of 1.1.
3) The only place it makes a dierence is in Stage A of the proof of Claim 1.1. We
choose N, as in () of 1.5(3), and let

= (2 +) : < ) in N[] = N[
0
,
1
]
instead of working with V[

0
,

1
].
1.5
Now we would like not to restrict ourselves to
1
1
[]-equivalence relations.
1.6 Claim. 1) Assume
(a) =
<
, 2

(b) E is a
1
2
[] 2-place relation on

2, say denable by (Z
1
)(Z
2
)(x, y, Z
1
, Z
2
, a)
(c)() E is an equivalence relation on

2
() if ,

2 and (! < )(()) ,= ()) then (E)
(c)
+
if

2 is generic over V for (
>
2, ), i.e. is a Cohen sequence over V
then in V[], clause (c) still holds
(note that for
1
,
2
(

2)
V
anyhow V [=
1
E
2
V[] [=
1
E
2
)
(d) for every A and > 2

there are N,

: < ) such that

(i) N (H (), ), N
<
N, |N| = , A N
(ii)

2 and [ <

,=

]
(iii) for ,= the pair (

) is generic over N for the forcing notion

(
>
2
>
2)
(iv)
1
1
[] formulas are preserved from N[

] to V for < < .

(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


10 SAHARON SHELAH
Then E has equivalence classes.
2) We can replace by perfect in the conclusion if in (d),

: <

2
is perfect [see 0.6].
3) We can replace
>
2 by a subtree T
>
2 such that forcing with T adds no
bounded subset to .
Proof. By [Sh 664, 2.2t].
1.7 Denition. Clause (d) of 1.6 is called is [, )-weakly Cohen-absolute:
[, )-w.c.a., in short (as in [Sh 664, 2.1t]s notation).
1.8 Claim. We can strengthen 1.6 just as 1.5 strenghthens 1.1.
We may wonder when does clause (d) of 1.6 hold.
1.9 Claim. 1) Assume
(i) =
<
in V
(ii) P is a forcing notion
(iii)

: < ) is a sequence of P-names,

(iv)
P

,=

2 for < <
(v) if A , p P, large enough then there are N (H (), ), |N| =
, N
<
N, A, p N and q such that p q P, q is (N, P)-generic,
q (
>
2)
V
P
N[G

P
] and P

P such that q
P
for some u []

, for
every ,= from u, the pair (

) is generic over N[G

P
] for (
>
2
>
2)
V
P
and the forcing P/(P

) is -complete (or at least -strategically

complete).
Then is (, )-w.c.a. (see 1.7) in the universe V
P
.
Proof. Straightforward.
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


ON NICE EQUIVALENCE RELATIONS ON

2 SH724 11
2 Singulars of uncountable cofinality
In this section we show that the natural generalization of 0.1 usually provably
fails badly for
cf()
, singular of uncountable conality.
2.1 Claim. Assume
(a) > = cf() >
0
(b) 2

+
<
= .
Then there is E such that
() E is an equivalence relation on

() E is very nice
1
(see Denition 0.2)
() if
1
,
2


and (

i < )(
1
(i) =
2
(i)) then
1
E
2

1
=
2
() E has exactly equivalence classes.
2.2 Observation. In 2.1, and in the rest of this section: (of course, we have to
translate the results; we leave it as an exercise to the reader).
1) We can restrict ourselves to

i<

i
where i <
i
< =

j<

j
, see the proof;
similarly in 2.4.
2) We can consider

as a subset of

2, in fact a very nice one:
we identify

with

2 when

(i) = 1 i pr(, ()) : < for

any choice of a pairing function pr, that is, any one to one function pr from
onto is O.K.
3) If is strong limit we can identify

i<

i
with

2 as follows: without loss of generality
i
=
2

i
with
i
increasing, let g
i

: <

i
2) list the functions from [

j<i

j
,
i
) to 0, 1
and we identify

i<

i
with

i<
g
i
(i)


2.
4) We can translate our results to any

i<

i
when
i
=

i<

i
= lim sup
i
:
i < ).
5) Even without assuming 2

+
<
= , the union of Tichonov closed subsets
1
In fact we have a closed division of

to

2 sets such that E renes this division and on each
part E is closed, see 2.2(5)
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


12 SAHARON SHELAH
of (

) (

) is very nice where A

is closed when: if (, )

and
for every nite u for some (

) A we have u =

u & u =

u
then (, ) A.
6) If =
<
(5) holds even for tree closed subsets (the topology we normally use).
Proof of 2.2. (1),(2),(3) left to the reader.
4) Dene the function F from

to

i<

i
by dening F() by induction on
g() as follows:
(a) F(<>) =<>
(b) F()) is F()
,
when:
,
= Min : <
g(F())+
,
,
=
0

,
1 +)
(c) for of limit length, F() =

<g()
F( ).
Clearly g() g(F()) and , are -incomparable implies F(), F() are -
incomparable, so F is one to one. Also F maps

into

i<

i
continuously so
Range(F) is a closed set.
Also, when cf() >
0
for any ,

we have (

)(() = ()) (

)((F())() =
F())()).
This is enough to translate 2.1 to

i<

i
instead of

.
Alternatively, we can repeat the proof.
5) Why is it very nice? Assume E = E
i
: i < i(), i() and each E
i
is a
closed subset of (

) (

). Let

: < list
>
with no repetitions, and we
dene a model M:
its universe is
F
0
is unary, F
0
() = g(

)
F
1
is binary, F
1
(, ) = i

(min, g(

))
R is a three-place relation, R
M
(, , i) i for some (
0
,
1
) E
i
we have

0
,

1
P is unary predicate P
M
= i()
< is binary relation, the order on , an individual constant.
Now for f, g : we have
fEg i (M, f, g) [= (i)[P(i) & ( < )(, )(F
0
() = & F
0
() =
& R(, , i) & ( < )[f() = F
1
(, ) & g() = F
1
(, )]).
Normally we do not elaborate such things.
2.2
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


ON NICE EQUIVALENCE RELATIONS ON

2 SH724 13
Proof of 2.1. We choose

=
i
: i < ), nondecreasing, i.e. i < j
i

j
with limit , (e.g.
i
= which is the case stated in the claim) let
j
=

i<j

i
so

j
and let

f
i
= f
i

: <
i
) list

j<i

j
or be just a set of representatives of

j<i

j
/J
bd
i
.
For every

i<

i
let
(a) for limit i < let
i
() = Min : i = f
i

mod J
bd
i

(b) for < let B

() = i : i < is a limit ordinal, < i and f

i

i
()
() = ()
and lastly
(c) A() = < : B

() is not stationary.
Now we dene two binary relations E
0
, E
1
on

i<

i
:
(d)
1
E
0

2
i for every < we have B

(
1
) = B

(
2
)
(e)
1
E
1

2
i
1
E
0

2
&
1
A(
1
) =
2
A(
2
).
Clearly
() E
0
is an equivalence relation on

i<

i
with 2

classes
() E
1
is an equivalence relation on

, rening E
0
() E
0
, E
1
are very nice; in details:
(a) E
0
is a closed subset of (

i<

i
) (

i<

i
) (under the initial seg-
ment topology, that is, for (
0
,
1
) (

i<

i
) (

i<

i
) the family
u

(
0
,
1
)
: < where u

= (
0
,
1
) (

i<

i
) (

i<

i
) : (
0

,
1
) = is a neighborhood basis of (
0
,
1
))
(b) E
1
is the union of 2

closed subsets of (

) (

) under the initial

segment topology
[Why?
(a) as if (
0
,
1
)

i<

i<

i
E
0
, then for some <
and i < , we have (i B

(
0
)) (i / B

(
1
)) so < i <
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


14 SAHARON SHELAH
and so u = u
i
(
0
,
1
)
is a neighbor of (
0
,
1
) and by the
denition of B

() we have u E
0
= hence u E
1
=
(b) for

B = B

: < ), B

let
B
=

i<

i
: B

() = B

for every < .

Now
B
:

B

P()) list the E
0
-equivalence classes
(and ) and each E
1

B
is closed.]
() if
1
,
2


and
1
E
0

2
then A(
1
) = A(
2
)
[Why? Check the denitions]
() for

, A() is a bounded subset of
[why? otherwise let C = < : = sup(A() ), it is a club of , and
for each i C there is j
i
< i such that [j
i
, i) = f
i

i
()
[j
i
, i), clearly j
i
exists by the denition of
i
(). By Fodor lemma, for some j() < the
set S
j()
= i C : j
i
= j() is stationary, now choose A()j(), so
clearly B

() includes S
j()
hence is a stationary subset of hence by the
denition of A() clearly does not belong to A(), contradiction.]
So clearly
() E
1
has (

i<

i
/E
0
) +

j<i

j
: i < equivalence classes.
Now
() if
1
,
2


and
1
=
2
mod J
bd

then for every limit i < large enough

we have
i
(
1
) =
i
(
2
)
[why? let i

= supj +1 :
1
(j) ,=
2
(j) so by the assumption, if i is a limit
ordinal and i (i

, ) then
1
i =
1
i mod J
bd
i
hence
i
(
1
) =
i
(
2
)
by the denition of
i
(), which is the desired conclusion of clause ().]
() if
1
,
2


and
1
=
2
mod J
bd

then
1
E
1

2

1
=
2
[why? if
1
=
2
clearly
1
E
1

2
; so assume
1
E
1

2
and we shall show that

1
=
2
, i.e. <
1
() =
2
(). By the denition of E
1
we have

1
E
0

2
hence by clause () we have A(
1
) = A(
2
), call it A. If A,
by the denition of E
1
we have
1
A =
2
A hence
1
() =
2
().
So assume A, rst we can nd j

< such that for every limit

i (j

, ) we have
i
(
1
) =
i
(
2
), it exists by clause (). Second, the sets
B

(
1
), B

(
2
) are stationary (as / A(

)) and equal (as

1
E
0

2
); so we
can nd i B

(
1
) B

(
2
) which satisfy i > j

. Now
1
() = f
i

i
(
1
)
()
by the denition of B

(
1
) as i B

(
1
) and
i
(
1
) =
i
(
2
) as i > j

and f
i

i
(
2
)
() =
2
() by the denition of B

(
2
) as i B

(
2
); together

1
() =
2
(). So we have completed the proof that <
1
() =
2
()
thus proving
1
=
2
as required.]
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


ON NICE EQUIVALENCE RELATIONS ON

2 SH724 15
() E
1
has
j
equivalence classes for any j <
[why? let

i<

i
and for <
j
let

be dened by

() is if
= j and is

() otherwise. By clause () we have < <

j

E
1

,
hence [

i<

i
/E
1
[
i
.]
() E
1
has exactly equivalence classes
[why? by clause (), E
1
has sup
i
: i < = equivalence classes and
by clause (), E has equivalence classes.]
We could have dened E
0
as
()
1
E
0

2
i for every < we have B

(
1
) = B

(
2
) mod D

where D

is
the club lter on .
This causes no change except that E
0
is not a closed subset of (

) (

), but a
union of 2

ones.
2.1
2.3 Claim. Assume
(a) > = cf() >
0
(b) 2

+
<
=
(c)

.
Then there is E such that
() E is an equivalence relation on

() E is very nice
2
() if
1
,
2


and
1
=
2
mod J
bd

then
1
E
2

1
=
2
() E has exactly equivalence classes.
Proof. Let

be as in the proof of 2.1 except that we add i <

j<i

j

i
, (this
holds if e.g. if i <
i
= ). We can nd a tree T
>
with nodes and
exactly -branches ([Sh 262]); we can easily manage that ,= lim

(T )
(

i < )((i)) ,= (i)). We proceed as in the proof of 2.1, but in the denition of
E
1
we add
2
in fact, again union of 2

closed sets of pairs

(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


16 SAHARON SHELAH

1
lim

(T)
2
lim

(T ) & (
1
lim

(T )
1
=
2
).

2.3
2.4 Claim. In Claim 2.1 we can replace clauses (), () by
()
1
E is very nice, moreover is the union of closed sets minus the union of
closed sets
()
1
for every

, the set

: =

modJ
bd

is a set of representa-
tives for the family of E-equivalence classes.
Proof. Let

be as there but <
+

i
. Let K
i
be a group, with universe

i
and unit 0
K
i
. Let <

be a well ordering of

(P()). For every

i<

i
let

= B

() : < ) :

i<

i
and = mod J
bd

.
So

is a non-empty subset of

(P()) and let

B

= B

,
: < ) be its <

-rst
member. Note that
for
1
,
2

i<

i
if
1
=
2
mod J
bd

then

B

1
=

B

2
and

1
=

2
.
Let

i<

i
: B

() = B

,
for every < and = mod J
bd

.
Now note
()
0

,= .
[Why? By the denition of

,

B

and

.]
()
1
if

then for every limit i < large enough we have

i
() =
i
().
[Why? As = mod J
bd

.]
()
2
if
1
,
2

and < , then for every limit i large enough we have:

i
(
1
) =
i
(
2
) hence f
i

i
(
1
)
() = f
i

i
(
2
)
().
Now for

i<

i
we dene

i<

i
by

() is :f
i

i
()
() for every i B

,
large enough
if B

,
is stationary
0
K
i
if B

,
is not stationary.
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


ON NICE EQUIVALENCE RELATIONS ON

2 SH724 17
It is easy to see that
()
3
if

i<

i
then

() = () for every < large enough.

[Why? We can nd

i<

i
such that = mod J
bd

and B

() : <
) =

B

. Now apply () inside the proof of 2.1.]

hence
()
4

= mod J
bd

()
5
if
1
,
2

i<

i
and
1
=
2
mod J
bd

then

1
=

2
.
Lastly, we dene the equivalence relation E:
for
1
,
2

i<

i
we dene:

1
E
2
i (for every i < we have K
i
[=
1
(i)(

1
(i))
1
=
2
(i)(

2
(i))
1
).
Now clearly
()
6
if
1
,
2

i<

i
and
1
=
2
mod J
bd

then
1
E
2

1
=
2
.
[Why? By ()
5
we have

1
=

2
, call it ; we are done by and the
properties of groups (i.e. x
1
y
1
= x
2
y
1
x
1
= x
2
.]
()
7
if

i<

i
then

i<

i
and

= mod J
bd

is a set of represen-
tatives of the E-equivalence classes.
[Why? Let ,

i<

i
and we shall dene

i<

i
such that

/E
and

= mod J
bd

. For i < we choose

(i) K
i
, i.e. <
i
such that
K
i
[=

(i)(

(i))
1
= (i)(

(i))
1
.
[Why this is solvable? As K
i
is a group and

(i), (i),

(i) are well de-

ned members of K
i
.] Also we know that =

mod J
bd

by ()
4
hence
for some i
1
< we have i [i
1
, ) (i) =

(i); this implies that

i [i
1
, )

(i) =

(i), so

mod J
bd

; however

= mod J
bd

hence

= mod J
bd

, as required. Hence

so by the denition of

we have K
i
[=

(i)(

(i))
1
= (i)(

(i))
1
which means that

E,
so we have proved ()
7
.]
Lastly, how complicated is E? Dene a two-place relation E

on

i<

i
:
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


18 SAHARON SHELAH

2

1
E

2
i
(a)

B

1
=

B

2
.
Clearly
()
8
E

is an equivalence relation on

and is the union of closed minus
the union of closed subsets of (

i<

i
) (

i<

i
) with 2

equivalence
classes
()
9
on each E

-equivalence class the function

is continuous (even under

the Tichonov topology, even more)
()
10
if Y
1
, Y
2
are E

-equivalence classes, then E(Y

1
Y
2
) is closed (even under
the Tichonov topology).
Now check.
2.4
We may like to weaken the cardinal arithmetic assumptions.
2.5 Remark. Assume that =
+
and instead the ideal J
bd

we use the ideal []

<
.
Then we can dene
j
() for

i<

i
and j < if cf(j) = cf(). Let
j
() be
Min : f
j

= j mod J
j
where J
j
= A j: for some i < j we have [Ai[ <
so J
j
replaces J
bd
j
in the earlier proof.
So = mod []
<
implies that
j
() =
j
() for all suitable j. There are no
marked changes.
Now
() if
1
E

2
then B

(
1
) = B

(
2
),

1
=

2
and

B

1
=

B

0
E

can serve as well and it is an equivalence relation with 2

equiv-
alence classes, each closed even under the Tichonov topology.
We can use > , J = []
<
but in general the number of ideals necessary is

. Most interesting is the case =

0
dealt with in the next claim.
2.6 Claim. 1) Assume
(a) > = cf() >
0
(b)

0
< =

0
.
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


ON NICE EQUIVALENCE RELATIONS ON

2 SH724 19
Then the results 2.1, 2.4 and 2.3 holds if we replace the ideal J
bd

by the ideal []
<
0
.
2) This applies also to 2.3 if
(c)

and there is a tree T with nodes and -branches.

3) The natural topology for (1) + (2) is the
1
-box product.
Proof. Without loss of generality
i
>

0
,
i
: i < ) as in the proof of 2.1. Let
D
i
: i <

0
) list the subsets of of order type and let

f
i
= f
i

: <

jD
i

j
)
list

jD
i

j
(or just a set of representatives modulo J
bd
D
i
). For

<

let
(a)

i
() = Min : D
i
= f
i

mod J
bd
D
i
for i <

0
(b)

for < let B

() = i <

0
: D
i
and () = f

i
()
()
(c)

A() = < : B

() is nite
(d)

() = i B

() : i B

() is nite.
With those choices the proofs are similar.
2.6
2.7 Claim. 1) If 2

0
< =
<
,
0
< = cf() < , then we can nd E as in
2.1(), (), () (but not necessarily ()) and
()

if

and i < then X
,i
=

: (j)(j < & j ,= i (j) =
(j) is a set of representatives for E.
2) If 2

0
=

0
,
0
< = cf() < , 1 and ( < )[( + )
<
),
then we can nd E as in 2.1(), () and
()

if

and i < then X
,i
contains a set of representatives
()

E has equivalence classes.

Proof. 1) First the proof in short.
We choose
i
= for i < . We let K be a group with universe and let D
j
: j <

0
) be as in the proof of 2.6 and dene E by: E i K [=

iA()
((i)(

(i))
1
) =

iA()
((i)(

(i))
1
). We give a more detailed proof below.
2) First, the proof in short. We choose
i
but
i
; without loss of generality each
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


20 SAHARON SHELAH

i
is a subgroup of K but we use equality in cosets of xK
1
= yK
1
, K
1
a subgroup of
K such that [K : K
1
] = and a, c K abcK
1
: b : <
0
= bK
1
; b <
.
Now in detail (for (2) so including a proof of (1)).
We repeat the proof of 2.4 + 2.6, so for

we let

= B

() : < ) :

and = mod[]
<
0
where B

() = j <

0
: f

j
()
() = () and let

B

be
the <

-rst member of

and let

=

: B

() = B

,
for every <
and = mod []
<
0
and for

let

be dened by
(a)

() = f

j
()
() if ()(

& () = f

j
()
()) & j B

,
(b)

() = 0 if there are no j, as in (a).

Easily

is well dened and

= mod []
<
0
.
Lastly, let a

= < : () ,=

() and we dene the two-place relation E on

i<

i
by
1
E
2
i a

1
= a

2
& (

ia

1
()

(i)
1
K
1
= (

2
()
2
()
1
)K
1
.
Is this well dened? The product

() is a nite product in the group K, so

in general we have to choose an order of

() : a

), i.e., of a

i
. We use the
most natural choice: the order on (if K is abelian clearer). Obviously E is an
equivalence relation on

<

and it has [xK

1
: K[ = [K : K
1
] equivalence
classes. Now suppose that

and < and we shall prove that X
,
is
the set of representatives for E, recall X
,
is dened in ()

of 2.7(1). Let a

=
a

, a
+
= a

( +1), let g

ia

((i)(

(i))
1
) and g
+
=

ia
+
((i)(

(i))
1
),
so:
() g

, g
+
K again well dened as a

, a
+
are nite
() if X
,
then a

and

ia

((i)(

(i))
1
) = g

()g
+
K,
the product in K, of course.
Now for part (1), g

the sequence () : X
,
) lists K without repetition (as
the universe of K is ) hence (by basic group theory), g
1
()g
+
: X
,
)
lists K without repetitions hence

ia

((i)(

(i))
1
) : X
,
) lists K without
repetitions, so if we use the trivial K
1
, X
,i
is a set of representatives of E, as
required.
For (2) the sequence g

K
2
: X
,
) lists xK
1
: x K possibly with repetition.

3.5
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


ON NICE EQUIVALENCE RELATIONS ON

2 SH724 21
2.8 Concluding Remark. 1) Instead of J
bd
D
i
: i <

0
) we can use (D
i
, J
i
) : i <
i

), D
i
, J
i
an ideal on D
i
such that [

D
i

/J
i
[ , I = D : for every
i < i

we have D D
i
J
i
is included in J
bd

. The author has not pursued this.

2) Assume K is a group of cardinality , K
1
a subgroup and [K : K
1
] = .
Then we can nd B K, [B[ = such that if K

is a subgroup of K including
B such that

K,K
1
,K
if a, c K

then acK
1
: b K

= bK
1
: b K.
[Why? Let b
i
: i < be such that b
i
K
1
: i < = bK
1
: b K and
let B = b
i
: i < . If B K

K and ac K

and i < there is b

such that ab

c = b
i
so ab

cK
1
= b
i
K
1
.]
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


22 SAHARON SHELAH
3 Countable cofinality: positive results
We rst phrase sucient conditions which relate to large cardinals. Then we
prove that they suce. The proof of 3.1 is presented later in this section.
3.1 Lemma. Assume
(a) is strong limit of conality
0
(b) is a limit of measurables, or just
(b)

for every < for some , satisfying < , there is a (, , )-

witness (see Denition 3.2 below)
(c) E is a nice equivalence relation on

(or has enough absoluteness, as proved
in 3.12), i.e., fact 3.13, so being
1
1
() over zc is enough
(d) if ,

and (!n)((n) ,= (n)) then (E).
Then E has 2

equivalence classes, moreover if

n
<
n+1
< =
n<

n
then
there is a subtree of
>
isomorphic to

n<m

n
, whose -branches are pairwise
non E-equivalent (even somewhat more, see 3.17).
Remark. For the simplest example of witness dened below see 3.4(2) so a witness
is a weak form of being measurable.
3.2 Denition. 1) We say (Q, s
1
, s
2
) is a (, , )-witness if ( and):
(a) Q is a -complete forcing notion
(b) s
1
is a function from Q to P()
(c) s
2
is a function from Q to A : A (, ) : < <
(d) if Q [= p q then s

(q) s

(p) for = 1, 2
(e) (, ) s
2
(p) , s
1
(p) for p Q
(f) for every p Q there is q such that p q Q and
()(, )[ s
1
(q) (, ) s
2
(p) & (, ) s
2
(p)]
(g) if p Q and A , then for some q we have p q Q and (s
2
(q)
A) (s
2
(q) A = )
(h) if p Q then for some Y []

for every < from Y we have (, )

s
2
(p) (hence Y s
1
(p)).
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


ON NICE EQUIVALENCE RELATIONS ON

2 SH724 23
2) We say (Q, s
1
, s
2
) is a (, , , )-witness if is a cardinal and we can
strengthen clause (g) to
3
(g)
+

if f :
2
and p Q then for some q we have p q Q and f s
2
(q) is
constant.
3) We call (Q, s
1
, s
2
) a uniform (, , )-witness if = s
1
(p) : p Q and for
every p Q and < for some q we have p q Q and s
1
(q) = .
Similarly a uniform (, , , )-witness.
4) We replace by < if we demand only (g)
+
<
which means that Rang(f) is a
subset of of cardinality < . We write < instead of if in clause (h) of
Denition 3.2(1) we demand just that for each < there is Y of order type
and as there (so can be an ordinal).
3.3 Denition. 1) We say that (Q, s) is a (, , , ; n)-witness if ,
and s = s
m
: m = 1, . . . , n) and
(a) Q is a -complete forcing
(b) s
m
is a function from Q to P( : =

: < m)
m
and

<
+1
<
for < m1)
(c) if Q [= p q and m 1, . . . , n then s
m
(q) s
m
(p)
(d) if

: < m+ 1) s
m+1
(p) and k < m+ 1 then

: < k)

: = k + 1, . . . , m) s
m
(p)
(e) for every m 1, . . . , n1, k < m and p Q there is q satisfying p q Q
and ( s
m
(q))(

s
m+1
(p))[ = (

k)(

[k + 1, m))]
(f)
+
if m 1, . . . , n and f :
m
and p Q then for some q we have
p q Q and f s
m
(q) is constant
(g) if p Q then for some Y []

every increasing
n
Y belongs to s
n
(p).
2) (Q, s) is a (, , , ; )-witness is dened similarly (i.e., s = s
m
: m [1, )))
and in clause (g) the same Y works for all n.
2A) (Q, s) is a (, , , ; )-witness is dened similarly, except that in clause (h),
for each n < there is Y []

such that every increasing

n
Y belongs to
s
n
(p).
3) If = 2 we may omit it, as in Denition 3.2. Also uniform and < and
< mean as in Denition 3.2.
We rst give some basic facts on witnesses, including cases of existence.
3
note that (g)
+

is equal to (g) if = 2
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


24 SAHARON SHELAH
3.4 Claim. 1) If (Q, s) is a (, , ; n)-witness and < , n < , then (Q, s) is a
(, , , 2

; n)-witness.
2) If D is a normal ultralter on so is a measurable cardinal and we choose,
Q = (D, ), s
1
(A) = A, s
2
(A) = (, ) : < are from A, then (Q, s
1
, s
2
) is a
uniform (, , , < )-witness.
3) If in (2), s
m
(A) = : =

: < m) is increasing,

A, s = s
m+1
:
1 + m n) and n then (Q, s) is a (, , , < ; n)-witness.
4) If there is a (, , , ; n)-witness and 2
<
, then there is such (Q, s) with
[Q[ 2

.
5) Denition 3.2(1) is the case n = 2 of Denition 3.3(1) that is, (Q, s
1
, s
2
) is a
(, , , )-witness i (Q, (s
1
, s
2
)) is a (, , , ; 2)-witness.
6) If (Q, s) is a (, , )-witness and p Q, then we can nd q such that p q Q
and for every s
1
(q) there are
1
<
2
< such that (
1
, ), (
2
, ) s
2
(p)
(this strengthens clause (f ) of 3.2).
Proof. Easy.
1) Checking Denition 3.3 the least easy clause is (f)
+
, so assume m 1, . . . , n
and p Q and f is a function from
m
to 2

and we should nd q satisfying

p q Q and f s
m
(q) is constant. Let h be a one to one function from 2

into

2 and dene f

:
m
0, 1 for < by f

( s) = (h(f( s)))(). Now we choose

p

Q, increasing (by
Q
) by induction on such that p
0
= p, f

s
m
(p
+1
) is
constant, say is

. For = 0 this is trivial, for successor use (Q, s) is (, , ; n)-

witness, i.e. clause (f)
+
in Denition 3.3. For a limit ordinal we use Q is
-complete, i.e., clause (a) in Denition 3.3 for (Q, s) is a (, , ; n)-witness, recall
< .
Lastly, let q = p

so we are done.
2), 3) Note that Q is -complete as D is -complete as D is a -complete ultralter
(being normal) and clause (f)
+
holds because if f
n
: []
n
and < then for
some A
n
D we have f [A]
n
is constant (see, e.g., [J]) and as D is closed under
intersection of < (hence of
0
) we are done (if p Q, let q = p

n<
A
n
).
4) Let (Q, s) be a (, , , ; n)-witness and let be large enough. Choose an
elementary submodel N of (H (), ) to which (Q, s) satisfying |N| = 2

, [N]

N so 2

N.
Lastly, choose Q

= Q N and s

m
= s
m
Q

. Now check that (Q

, s

m+1
: m <
n)) is a (, , , ; n)-witness recalling , , .
5) Read the denitions.
6) For 0, 1, 2 let A

= s
1
(p): the number [

< : (

, ) s
2
(p)[ is
equal to or = 2 and the number is .
So A
0
, A
1
, A
2
) is a partition of s
1
(p).
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


ON NICE EQUIVALENCE RELATIONS ON

2 SH724 25
Dene a function f from A
1
to : for A
1
, f() is the unique

< s
2
(p).
It is known (and easy) that we can nd a partition B
1
, B
2
, B
3
) of A
1
such that
1, 2, 3 & B

f() / B

. Let B
0
= A
0
, B
4
= A
2
, so B
0
, . . . , B
4
)
is a partition of
2

=0
A

that is of s
1
(p). By clause (g) of Denition 3.2 (applied
three times, see 3.4(1)) we can nd () < 5 and q Q such that p q Q and
s
1
(q) B
()
. s s
2
(q) ,= necessarily () = 4 and so we are done.
3.4
Something of the largeness remains if we collapse a large cardinal, see, e.g.,
[JMMP]. We shall need
3.5 Claim. 1) Assume
(a) 2 n < and =
n1
()
+
(b) is a compact cardinal or just a -compact cardinal
(c) =
<
<
(d) P = Levy(, < ).
Then in V
P
(and of course in V), there is a (, , ; n)-witness (Q, s) which is even
a (, , , < ; n)-witness.
2) If there are
n
for n < ,
n
<
n+1
and
n
is 2
(2

n
)
+
-compact and =

n
: n < , then for some set forcing P, in V
P
the cardinal =

is
dichotomically good (see Denition 3.8 below).
Proof. By [Sh 124].
3.6 Remark. 1) In fact we can weaken the consistency strength. Assume that
(G.C.H. holds for simplicity) and:
(a)
n
: n < ) is strictly increasing sequence of cardinals
(b)
n

n
<
n+1
(c) D
n+1
is a
n+1
-complete ultralter on I
n+1
= a [
n+1
] <
n+1
:
min(a) <
n+1

(d) let
n+1
: I
n+1

n+1
is
n+1
(a) = min(
n+1
a) and if A D
n+1
, f :
A
n+1
is regressive, i.e., f(a) <
n+1
(a) then f is constant on some
B D
n+1
, B A
(e) if g : [
n+1
]
n

n
then a I
n+1
: g [a
n+1
(a)]
n
is constant D
n+1
(f) Q
0
= Levy(
0
,
0
), Q
n+1
= Levy(
++
n
, <
n+1
), Q =

n<
Q
n
.
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


26 SAHARON SHELAH
Then V
Q
is as required in 3.5.
2) If
n
is
+(n+2)
n
-hyper-measurable and we let
n
=
+(n+1)
n
and
n
<
n+1
, then
there is j
n
: V M
n
,
n
is the critical cardinal of j
n
, M

n
n
M
n
, j
n
(
n
) >
+(n+2)
n
and H (
n
) M
n
. So in V we can nd b [j
n
(
n
)
n
]

n1
such that
() if f : [
n
]
n

n1
then j
n
(f) [b]
n
is constant.
Let a =
n
b so a M
n
and D
n
= A [
n
]

n1
: a j
n
(A).
Those D
n
are as required for
n
=
n
.
Toward proving Lemma 3.1 assume (from 3.10 till the end of this section) that
3.7 Hypothesis. m =
n
,
n
,
n
, P
n
, s
n,1
, s
n,2
)
n<
=
m
n
,
m
n
,
m
n
, P
m
n
, s
m
n,1
, s
m
n,2
)
n<
satises =
n
: n < and
1
+ 2

: < n <
n

n
and (P
n
, s
n,1
, s
n,2
)
is a (
n
, <
+
n
,
n
)-witness and for simplicity
n
<
n+1
and =

n
: n < .
3.8 Denition. We call dichotomically good if there is m, i.e., there are

n
,
n
,
n
, P
n
, s
n,1
, s
n,2
as in 3.7.
The hypothesis 3.7 is justied because
3.9 Observation. 1) If satises (a) + (b) or at least (a) + (b)

of Lemma 3.1
then is dichotomically good.
2) It is consistent that G.C.H. and

is dichotomically good (if CON(ZFC + there

is a supercompact cardinal).
3) For proving 3.1 without loss of generality E is a nice equivalence relation on

n<

n
satisfying clause (d) of 3.1.
Proof. 1) By 3.4(2) we know (b) (b)

in 3.1, now read the denitions.

2) By 3.5.
3) ??
3.10 Denition. 1) We dene the forcing notion Q
1
(really Q
1
= Q[m]) as follows:
(a) Q
1
=

p : p = (,

A) = (
p
,

A
p
) such that letting n
p
= n(p) = g()
we have n
p
< ,
p

<n[p]

and

A
p
= A
p

: [n(p), )) and A
p

(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


ON NICE EQUIVALENCE RELATIONS ON

2 SH724 27
(b) p
Q
1
q i
p

q
(so n(p) n(q) and [ [n(q), ) P

[= A
p

A
q

]
and [n(p) < n(q)
q
() s
1
(A
p

)]
(c) We dene the Q
1
-name

by:

[G] =
p
: p G

Q
1

(d) We dene
() p
Q
1
pr
q i p
Q
1
q & n(p) = n(q)
() p
Q
1
apr
q i p
Q
1
q &

n(q)
(A
q

= A
p

)
() p
Q
1
pr,n
q i p
Q
1
pr
q and n > n(p)

A
p
[n(p), n) =

A
q
[n(p), n).
2) We dene the forcing notion Q
2
(really Q
2
[m]) by:
(a) Q
2
=

p : p = (
0
,
1
,

A) = (
p
0
,
p
1
,

A
p
) where for some n(p) < we have:

p
0
,
p
1

<n(p)

and

A
p
= A
p

: [n(p), )) and A
p

(b) p
Q
2
q i
(i) n(p) n(q)
(ii)
p

for = 0, 1
(iii) A
q

A
p

for [n(q), )
(iv) the pair (
q
0
(),
q
1
()) is from s
2
(A
p

) for [n(p), n(q))

(c) we dene the Q
2
-name

(for = 0, 1) by

[G] =
p

: p G

Q
2

(d) we dene
() p
Q
2
pr
q i p
Q
1
q & n(p) = n(q) and
() p
Q
2
apr
q i p
Q
2
q &

n(q)
A
q

= A
p

and
() p
Q
2
pr,n
q i p
Q
2
pr
q and n > n(p)

A
p
[n(p), n) =

A
q
[n(p), n).
3) If for a xed k < , we have (P
n
, s
n
) is a (
n
,
n
,
n
; k)-witness for n < then
we can dene Q
k
naturally.
4) If (P
n
, s
n
) is a (
n
,
n
,
n
; n)-witness for n < then we can dene
Q = (,

A) : n < , ()

(

) and

A = A

: [n, ), A

with the
natural order.
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


28 SAHARON SHELAH
Remark. 1) We shall not pursue here parts (3) and (4) of Denition 3.10 because we
deal with equivalence relations which are binary. We can prove parallel theorems
for relations with higher arity using 3.10(3),(4).
2) In the denition of the set of elements p of Q
2
, why dont we ask ( <
n
p
)(
p
0
() <
p
1
())? To be able to construct the perfect set, but, of course,
p
Q
2

0
() <

1
() for [n(p), ).
3) Those forcing notions are in the (large) family of relatives of Prikry forcing.
3.11 Fact. Let 1, 2.
0) For p, q Q

we have:
(i) p
Q

pr
q p
Q

q
(ii) p
Q

apr
q p q
(iii) p
Q

pr,n+1
q p
Q

pr,n
q p
Q

pr
q.
1) If p
Q

r then for some q we have p

Q

pr,n(q)
q
Q

apr
r.
2) If p = p
i
: i < ) is
Q

pr
-increasing and <
n(p
0
)
(=
m
n(p
0
)
), then p has a

pr
-upper bound; similarly for
Q

pr,n
and <
n
.
3) If

is a Q

-name of an ordinal and p Q

, then for some q and n we have:

(a) p
pr
q
(b) if q
apr
r and n(r) n, then r forces a value to

.
4) In (3), if
Q

< or just <

<
n(p)
then without loss of generality n =
n(p).
Proof. Easy.
3.12 Claim. Recall that by 3.9(3) without loss of generality E is a nice denition
of a two-place relation on

n<

n
. Then forcing by Q
2
preserves E is an equiva-
lence relation on

n<

n
satisfying clause (d) of 3.1 or more exactly the denition
E denes in V
Q
2
an equivalence relation on

n<

n
satisfying clause (d) of 3.1
(and, of course, E
(V
Q
2
)
(

n<

n
)
V
= E
V
).
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


ON NICE EQUIVALENCE RELATIONS ON

2 SH724 29
Proof. Assume toward contradiction that p

Q
2

0
,

1
,

<

form a coun-
terexample, that is:

0
E

1
E

0
E

2
or

0
E

0
or

0
E

1
E

0
or

0
E

1
(!n)(

0
(n) ,=

1
(n)).
Choose large enough and

N = N
n
: n < ), N such that:

N
(i) N
n

L

+
n
,
+
n
(H (), ) and |N
n
| = 2

n
and p

, E,

0
,

1
,

2
, N
0
, . . . , N
n1

belong to N
n
(ii) N
n
N
n+1
hence N
n

L

+
n
,
+
n
N
n+1
and N =

n<
N
n
so N (H(), ).
Now we choose p
n
by induction on n < such that:
()(i) p
0
= p

,
(ii) p
n
N
n
Q
2
and n(p
n
) = maxn, n(p

)
(iii) p
n
p
n+1
(iv) if

N
n
is a Q
2
-name of an ordinal then for some k
n
(

) > n + 1 we have:
if p
n+1
q and n(q) k
n
(

) then q forces a value to

.
This is possible by 3.11(2),(3). Now let G = q : q N Q
2
and q p
n
or
just p
n
q G

for some n; it is a subset of Q

N
2
generic over N. (Why?
If N [= I Q
2
is dense then I Q
2
is dense and there is I

I, a
maximal antichain of Q
2
which belongs to N hence to some N
n
; there is g N
n
,
a one to one function from I

onto [I

[, so it denes a Q
2
-name

by

[G] =
(q)(q I

G f(q) = ) (q)(q I

G & f(q) = ), so
k
n
(

) < is well dened (see clause (iv) above) and so p

k
n
(

)
forces a value to

hence forces q G

for some q I

I, hence q G so G I ,= as required).
Now by straightforward absoluteness argument,

0
[G],

1
[G],

2
[G]

<

give
contradiction to an assumption.
In details let

[G]. Let M be the Mostowski collapse of N, so there is an

isomorphism g from N onto M. Clearly
n
N
n
hence N hence + 1 N
so g(x) = x if x + 1 or x + 1 or x H (). Clearly G

= g

(G) is a
generic subset of Q

2
= (g(Q
2
))
M
and M

= M[G] is a generic extension of M (for

g(Q
2
)
M
) and so

1
M

is a transitive model of enough set theory (i.e. of ZC if is strong

limit) which includes H () , H ()
n
: n < ).
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


30 SAHARON SHELAH
Also easily in M

[G

] =

, so as g(p

) G

, clearly (E stands for the formula

dening it, its parameter is a subset of so it is mapped by g to itself):
M

[=
0
,
1
,
2

n
: n < and

0
E
1
&
1
E
2
&
0
E
1
or

0
E
0
or
0
E
1
&
1
E
0
or

0
E
1
& (!n)(
0
(n) ,=
1
(n)).
So it is enough to prove (see Lemma 3.1, clause (c)).
3.12
3.13 Fact. Assume M

satises
1
above, E is a nice two-place relation on
n
:
n < so a denition with parameter which is a subset of (equivalently: a model
on ) as in Denition 0.2(1).
Then

2
if M

satises
1
E
2
&
3
E
4
and
0
,
1
,
2
,
3

n
: n < then
so does V.
Proof. Immediate.
In fact
3.14 Observation. Assume
(a)(i)

is strong limit of conality

0
,
(ii)

n<

n
(iii)

n
<

n+1
for n < , for simplicity 2

n
<

n+1
(b)(i) Q is a forcing notion
(ii)
pr
is included in
Q
(iii) n : Q is a function satisfying for each n the set I
n
= p Q : n(p)
n is a dense subset of Q
(iv) for p Q, q Q : p
pr
q is

n(p)
-complete
(v) Q has pure decidability for Q-names of truth values
(vi) if p Q and

is a Q-name of an ordinal, then there are m < and q

satisfying: p
pr
q and (q r & m n(r)) (r forces a value to

)
(c) N, N
n
: n < ) as in the proof of 3.12 for

n
: n < ),
Q, ,
pr
N
0
.
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


ON NICE EQUIVALENCE RELATIONS ON

2 SH724 31
Then there is G Q
N
generic over N hence H ()
N[G]
= H () = H ()
N
.
Proof. Should be clear.
3.15 Claim. Assume that F is a permutation of (

<n()

) (

<n()

) and let
Q
n()
2
= p Q
2
: n(p) n(). We let

F be the following function from Q
n()
2
to Q
n()
2

F(p) = q i n(q) = n(p)

(
q
0
n(),
q
1
n()) = F((
p
0
n(),
p
1
n()))

q
0
[n(), n(p)) =
p
0
[n(), n(p))

q
1
[n(), n(p)) =
p
1
[n(), n(p))

A
q
=

A
p
.
Then the following holds:
1) For p Q
n()
2
,

F(p) is well dened Q
n()
2
.
2)

F is a permutation of Q
n()
2
preserving ,
pr
,
pr,n
,
apr
and their negations,
and F

F is a group homomorphism (hence embedding).
3) If G Q
2
is generic over V then
(a)

F(G) =: r Q
2
: for some q G Q
n()
2
we have r

F(q) is a subset
of Q
2
generic over V
(b) G = p Q
2
: there is q Q
n()
2
such that p
Q
2
q and

F(q)

F(G)
(c) and V[

F(G)] = V[G] and even N[

F(G)] = N[G] if, e.g.,
N (H (), ), Q
2
N, F N, N.
Proof. Easy.
3.16 Claim.

Q
2

0
E

1
.
Proof. If not, let p Q
2
be such that p
Q
2

0
E

1
. Now by clause (f) of
Denition 3.2(1), we can nd p
1
such that:
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


32 SAHARON SHELAH
(i) Q
2
[= p
pr
p
1
(ii) if n(p) n < and s
1
(A
p
1
n
) then for some , we have (, ), (, )
s
2
(A
p
n
).
Let G
1
Q
2
be generic over V such that p
1
G
1
and let

[G
1
] for = 1, 2
so V[G
1
] [=
0
E
1
. By 3.12 in V[G
1
], E is still an equivalence relation satisfying
clause (d) of 3.1 and trivially n [n(p), )
1
(n) s
1
(A
p
1
n
). Let n

=: n(p), by
3.4(6)we can nd

<
n
such that

<
1
(n

),

,=
0
(n

) and (

,
1
(n

))
s
2
(A
p
n
). Let us dene

n<

n
by

0
(n) is

if n = n

and
0
(n) otherwise;
as

<
1
(n

) <
0
(n

) necessarily
0
,=

0
.
Now the pairs (
0
(n()+1),
1
(n()+1)) and (

0
(n())+1),
1
(n()+1)) are
from (

nn()

n
) (

nn()

n
), so there is a permutation F of this set interchanging
those two pairs and is the identity otherwise. Let

F be the automorphism of
Q
(n

+1)
2
from Claim 3.15. Let G
2
=

F(G
1
). Now by 3.15:
()
1
G
2
is a generic subset of Q
2
over V
()
2
V[G
2
] = V[G
1
]
()
3

0
[G
2
] =

0
,

1
[G
2
] =
1
.
By 3.12 (and the choice of

0
) we have
()
4
V[G
1
] [=
0
E

0
.
As p p
1
G
1
, by the choice of p clearly
()
5
V[G
1
] [=
0
E
1
.
By the choice of p
1
and (,
1
(n

)) clearly p (

0
[n()+1)),

1
(n()+1),

A
[n() + 1, ) G
2
so (using ()
1
)
()
6
V[G
2
] [= (

0
[G
2
])E(

1
[G
2
])
hence by ()
2
+ ()
3
we have
()
7
V[G
2
] [=

0
E
1
.
Now ()
4
+ ()
5
+ ()
7
contradict 3.12.
3.16
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


ON NICE EQUIVALENCE RELATIONS ON

2 SH724 33
3.17 Claim. 1) Fix > large enough and choose N
n

L

n
,
n
(H (), ) such
that |N
n
| = 2

n
, E, m N

: < n belongs to N
n
(hence Q
2
N
n
), and let
N =

n<
N
n
; (certainly can be done). Then we can nd

<n

n
and
n < ) and
4
()

<g()

()
1

2

2
() if
1
,
2

<n

and m k < n,
1
m =
2
m and
1
(m) <
2
(m) then

1
(k) <

2
(m)
() if

<

then

=:

n<

n
is generic for (N, Q
1
)
() if
0
,
1

<

and
0
<
lex

1
then (

0
,

1
) is generic for (N, Q
2
) hence
() if
0
,=
1

<

then (

0
E

1
).
2) Also, for some p Q
2
, n(p) = 0 and non-principal ultralter D on we have
() if ,

n<
s
1
(A
p
n
) and /D ,= /D then (E).
3) Moreover, there is a lter J on to which all co-nite subsets of belong and
for ,

n<
s
1
(A
p
n
) we have E = mod J.
Proof. Let M
0

L

1
,
1
N
0
be such that |M
0
| = 2

0
and E, m M
0
.
As above we choose p
n
by induction on n such that:

1
(i) p
n
Q
2
(ii) p
n
N
n
(iii) n(p
0
) = 0
4
why not {

: < n}? First we like

(n) : {

: n} to be increasing with
(the s are linearly ordered by lexicographic order) so the order type is the ordinal product

n

n1
. . . has cardinality but order type > . Second and more seriously we intend
to use clause (h) of Denition 3.2 which gives us Y of cardinality ; note if we use 3.4(1) we get

n
=
n
but not if we use 3.5(1).
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


34 SAHARON SHELAH
(iv) p
n

pr
p
n+1
(hence p
0

pr
p
n
so < n(p

) = 0)
(v) for every Q
2
-name of an ordinal

N
n
, for some k
n
(

) [n, ) we have:
if Q
2
[= p
n+1
q and n(q) k
n
(

) then q forces a value to

(vi) if

M
0
is a Q
2
-name of a natural number then p
0
forces a value to it.
Moreover,
(vii) if n < and
0
,
1

<n

then p
1
n+1
= (
0
,
1
, A
p
n+1

: [n, ]) Q
2
satises, too, clause (v).
We can nd p

Q
2
such that n < p
n

pr
p

and we can nd p

such that
p

and (n)()(, )[ s
1
(A
p

n
) (, ), (, ) s
2
(A
p

n
)] and we shall
show that p

is as required in parts (2) and (3), for p. Now clearly

2
if
0
,
1

n<

n
and ( < 2)(n < )(

(n) s
1
(A
p

n
)) and for every
n < large enough (
0
(n),
1
(n)) s
2
(A
p

n
) then
(a) for some subset G of Q
N
2
generic over N to which p
0
belongs we have

0
[G] =
0
,

1
[G] =
1
(b)
0
E
1
.
[Why? Let k

< be such that k

k <
0
(k

) ,=
1
(k

),
it exists by the denition of order on Q
2
. For every k k

we dene
a condition q = q
k

0
,
1
Q
2
by: n(q) = k,
q
0
=
0
k,
q
1
=
1
k
and A
q
n
= A
p

n
for n [k, ) and let G

0
,
1
=: r : r Q
2
, r N
and r
Q
2
q
k

0
,
1
for some k < . By
1
and the proof of 3.12 easily
G

0
,
1
is a subset of Q
2
N generic over N, so clause (a) holds. By 3.16
clearly N[G

0
,
0
] [= (
0
E
1
). By using absoluteness (as in 3.12(1)),
also clause (b) holds.]
This suces for part (1), in detail: by clause (h) of Denition 3.2(1) recalling the

+
n
is Hypothesis 3.7, we can nd Y
n

n
of order type
n

n1
. . .
0
from
N
n
such that for any < from Y
n
the pair (, ) belong to s
2
(A
p

n
). Now we
can choose by induction on n,

<n

) as required in (), (), () of 3.17(1)

such that () Y

, they are as required.

We are left with proving part (2). For B let

B
be the following Q
2
-name:
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


ON NICE EQUIVALENCE RELATIONS ON

2 SH724 35

B
(n) is

1
(n) if n B and is

0
(n) if n B.
Clearly

B
is a Q
2
-name of a member of

and

B
M
0
(recall that |M
0
| = 2

0
)
hence for B
1
, B
2
the following Q
2
-name of a truth value, the truth value of
(

B
1
E

B
2
), is decided by p
0
, say it is t(B
1
, B
2
).
Dene a two place relation E

on P() : B
1
E

B
2
i t(B
1
, B
2
) = truth.
Let J = B : t(, B) = truth, that is, J = B : E

B.
Clearly
()
0
E

is an equivalence relation on P().

[Why? By E being (forced to be) an equivalence relation.]
()
1
/ J, moreover [n, ) / J.
[Why? By
2
.]
()
2
if B
1
, B
2
J then B
1
E

B
2
.
[Why? As E

is an equivalence relation.]
Let
0
n
<
1
n
<
2
n
<
3
n
<
4
n
<
5
n
be from Y
n
for n < and for h

0, 1, 2, 3, 4
let
h

n<

n
be
h
(n) =
h(n)
n
. If g
1
, g
2


1, 2, 3, 4 and B we let
h
g
1
,g
2
,B


1, 2, 3, 4 be h
g
1
,g
2
,B
(n) =

g
1
(n) if n / B
g
2
(n) if n B.
Easily
()
3
if g
1
, g
2


1, 2, 3, 4 and (n < )((n B
1
B
2
) (n B
2
B
1
) g
1
(n) <
g
2
(n)) and B
1
, B
2
then B
1
E

B
2
i
h
g
1
,g
2
,B
1
E
h
g
1
,g
2
,B
2
[Why? That is, let h

= h
g
1
,g
2
,B

for = 1, 2 and note that n (B

1
B
1
)
(B
1
B
2
) h
1
(n) = h
2
(n).
We dene

0
,

n
y
n
as follows:
(a) if n (B
1
B
2
) (B
2
B
1
) then

0
(n) =
n
g
1
(n)
,

1
(n) =
n
g
2
(n)
(b) if n B
1
B
2
then

0
(n) =
n
g
2
(n)1
,

1
(n) =
n
g
2
(n)
(c) if n B
1
B
2
then

0
(n) =
n
g
1
(n)
,

1
(n) =
n
g
1
(n)+1
.
Now choose G as in clause (a) of
2
with (

0
,

1
) here standing for (
0
,
1
)
there and note that
(d)
h
1
=

B
1
[G],
h
2
=

B
2
[G].
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


36 SAHARON SHELAH
[Why? Because as
n B
1
B
2
(

B
1
(n),

B
2
)[G] = (g
2
(n), g
1
(n)) = (
h
1
(n),
h
2
(n)) and
n B
2
B
1
(

B
1
(n),

B
2
(n))[G] = (g
1
(n), g
2
(n)) = (
h
1
(n),
h
2
(n))
and also for the other ns.]
Now
h
g
1
,g
2
,B
1
E
h
g
1
,g
2
,B
2
mean (by the choice of h
1
, h
2
) that
h
1
E
h
2
which
by clause (d) means

B
1
[G]E

B
2
[G] which by 3.13 is equivalent to N[G] [=

B
1
[G]E

B
2
[G] which means that B
1
E

B
2
, so we are done.]
()
4
if B
1
, B
2
, B = B
1
B
2
then B
1
E

B
2
B
1
E

B & B
2
E

B.
[Why? The implication holds as E is an equivalence relation so let us
proof . By the symmetry it is enough to show that B
1
E

B. We choose
h

1, 2, 3, 4 for = 1, 2, 3 by: if n (B
1
B
2
) then (h
1
(n), h
2
(n), h
3
(n)) =
(2, 3, 1), if n B
2
B
1
then (h
1
(n), h
2
(n), h
3
(n)) = (1, 2, 3), if n (B
1

B
2
) or n B
1
B
2
= B then (h
1
(n), h
2
(n), h
3
(n)) = (1, 1, 1).
Now we choose functions g
a
1
, g
a
2
, g
b
1
, g
b
2
, g
c
1
, g
c
2


1, 2, 3, 4 as follows: for
n < , the six-tuple (g
a
1
(n), g
a
2
(n), g
b
2
(n), g
b
2
(n), g
c
1
(n), g
c
2
(n) is:
(i) (1, 2; 1, 3; 2, 3) if n B
1
B
2
(ii) (1, 3; 2, 3; 1, 2) if n B
2
B
1
(iii) (1, 1; 1, 1; 1, 1) if n (B
1
B
2
) ((B
1
B
2
))
So
h
1
E
h
3
as we are asuming B
1
E

B
2
, using ()
3
for (g
a
1
, g
a
2
) the only if
part because (h
1
, h
3
) = (h
g
a
1
,g
a
2
,B
1
, h
g
a
1
,g
a
2
,B
2
. Also
h
2
E
h
3
similarly using
(g
b
1
, g
b
2
).
Together it follows that
h
1
E
h
2
as E is an equivalence relation. Using
()
3
again for (g
c
1
, g
c
2
) this time, by the if part it follows that B
1
E

B as
required.]
Similarly
()
5
if B
1
B
2
then B
1
E

B
2
(B
2
B
1
) J.
[Why? This follows by ()
3
.]
()
6
if B
1
B
2
B
3
and B
1
E

B
3
then B
1
E

B
2
& B
2
E

B
3
[Why? We dene h
1
, h
2
, h
3


1, 2, 3, 4 by:
(a) if n B
3
B
2
then (h
1
(n), h
2
(n), h
3
(n)) = (2, 2, 4)
(b) if n B
2
B
1
then (h
1
(n), h
2
(n), h
3
(n)) = (2, 3, 4)
(c) if n (B
3
B
1
) then (h
1
(n), h
2
(n), h
3
(n)) = (1, 1, 1). Now
h
1
E
h
3
as we are assuming B
1
E

B
3
using ()
3
with (g
a
1
, g
a
2
) the only if part.
Similarly
h
2
E
h
3
using ()
3
with (g
b
1
, g
b
2
).
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


ON NICE EQUIVALENCE RELATIONS ON

2 SH724 37
As E is an equivalence relation we deduce
h
1
E
h
2
hence B
1
E

B
2
by ()
3
using
(g
c
1
, g
c
2
) the if part.
By E

being an equivalence relation we can deduce B

2
E

B
3
so we are done.]
()
7
J is an ideal
[Why? If B
1
B
2
& B
2
J we have B
1
B
2
& E

B
2
so by
()
6
we have E

B
1
as required. If B
1
, B
2
J are disjoint members of J,
then E

B
1
& E

B
2
by the denition J, so by E

being an equivalence
relation B
1
E

B
2
. Now B
1
J and so by ()
5
applied with B
1
, B
1
B
2
here
standing for B
1
, B
2
there we get B
1
E

(B
1
B
2
) so by transitivity of E

we
have E

(B
1
B
2
) which means B
1
B
2
J.]
()
8
0, . . . , n J
[Why? By
2
.]
()
9
B
1
E

B
2
i B
1
B
2
J
[Why? Let B = B
1
B
2
; if B
1
B
2
J then we have B
1
B, B
2
B J
so by ()
5
we have B
1
E

B & BE

B
2
hence B
1
E

B
2
. If B
1
E

B
2
then
B
1
E

B & B
2
EB by ()
4
, hence B
1
B, B
2
B J by ()
5
so by ()
7
B
1
B
2
J.]
So by ()
7
+ ()
2
there is an ultralter D on disjoint to J, and by ()
8
it is
non-principal, and by ()
9
it has the desired property so we have proved also part
(2). Part (3) has been proved by ()
7
+ ()
9
.
3.17

3.1
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


38 SAHARON SHELAH
4 The countable cofinality case: negative results
In the previous section we have gotten positive results, however, the assump-
tions are such that they may fail in ZFC (for every ). Can we eliminate those
assumptions? We below show that we cannot eliminate them: for reasonable the
conclusion fails strongly (as in 2), if fails the free subset property (a well known
property, see, e.g., [J]). So e.g. if 0
#
, the results of 3 fail.
4.1 Claim. Assume
(a) > cf() =
0
(b) ( < )[[[

0
< ]
(c) there is an algebra B with universe , with < functions and with no
innite free subset, see Denition below.
Then there is E such that
() E is an equivalence relation on

() E is very nice (see Denition 0.2)

() if ,

and =

(i.e. (
<
0
n)((n) ,= (n)) then E =
() E has equivalence classes.
4.2 Denition. A subset Y of an algebra B is free if: a Y a / c
B
(Y a)
where c
B
(Y

) means the subalgebra of B generated by Y

.
4.3 Remark. 1) We can replace

by the set of increasing -sequences or by

n<

n
when
n
<
n+1
< =

m<

m
or by A : (n)(!)( A &

<n

<

n
).
2) We can omit clause (b) if we weaken clause (). We can imitate 2.4 and 2.7, see
4.4 below.
Proof. Without loss of generality B has
0
function and the individual con-
stants : <
0
and there are no other individual constants. Let

n
=
(x
0
, . . . , x
n1
) : (x
0
, . . . , x
n1
) a
B
-term and <

n
a well ordering of

n
where

0
< , of course.
We dene a two place E
0
on

by
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


ON NICE EQUIVALENCE RELATIONS ON

2 SH724 39
E
0
i : if n < and k, k
1
, . . . , k
n
< then
(a) there is (x
0
, . . . , x
n1
)

n
such that
(k) = ((k
1
), . . . , (k
n
)) i there is
(x
0
, . . . , x
n1
)

n
such that (k) = (((k
1
), . . . , (k
n
))
(b) if in (a) they hold then the <

n
-rst term (x
0
, . . . , x
n1
)

n
such that (k) = ((k
1
), . . . , (k
n
)
is the <

n
-rst term (x
0
, . . . , x
n1
)

n
such that
(k) = ((k
1
), . . . , (k
n
)).
So E
0
is an equivalence relation with

0
0
< equivalence classes. For

let
A() = k : for some k

< there are no n < , k

1
, . . . , k
n
[k

, ) and B-term
(x
1
, . . . , x
n
) such that (k) = ((k
1
), . . . , (k
n
)).
Lastly, we dene E
1
by
E
1
i E
0
& A() = A().
The rest is as in 2.
4.1
4.4 Claim. 1) In 4.1 we can demand
() for each

, /J
bd

is a set of representatives of E.
2) We can weaken in 4.1 assumption (b) to
(b)

(
0
+[(B)[)

0
< .
3) If in 4.1 we change clause () in the conclusion to ()

below, we can omit clause

(b) of the assumption
()

for every

the set
,n
: < ) is a set of representatives of E with
no repetition where
,n


is:
,n
() = if = n and
,n
() = ()
otherwise.
Proof. 1) We imitate 2.4 only letting

k, k
1
, . . . , k
n
, ) : (k) = (k
1
, . . . , k
n
) :


, /J
bd

= /J
bd

.
2) The same proof.
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


40 SAHARON SHELAH
3) For

let n() < be the minimal n [n(), ) c
B
() : [n, ) =
c
B
() : [n(), ). Let K be an additive group with universe , K
1
a sub-
group, [K
1
[ = , [K : K
1
] = and E i

n<n()
(n) =

n<n()
() mod K
1
.

4.4
Remark. We can imitate in 2 the proof of 4.1: use a function F :

such
that there is no innite independent set for the algebra (, F) see [\EH71 ].
4.5 Question: 1) What about having (, 2

) equivalence classes?
2) Assume, e.g., is strong limit singular and 2

>
+
, does have the free subset
property? (See in [Sh 513]).
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


ON NICE EQUIVALENCE RELATIONS ON

2 SH724 41
5 On r
p
(Ext(G, Z)
5.1 Denition. For an abelian group G and prime p let r
p
(G) be the rank of
G/pG as a vector space over Z/pZ. Let r
0
(G) be the rank of G/Tor(G).
There has been much interest in Ext(G, Z) for G torsion free abelian group see
[EM], and later [MRSh 314]. This group is divisible so the ranks r
p
(G) above and
r
0
(Ext(G, Z)) determine it up to isomorphism.
Instead using a denition of the abelian group Ext(G, Z), we quote (see [Fu]) a
result which gives a characterization of the cardinal r
p
(Ext(G, Z)) directly from G.
5.2 Claim. For a torsion free abelian group G and prime p, r
p
(Ext(G, Z)) is the
rank of Hom(G, Z/pZ)/(Hom(G, Z)/pZ) where
(a) Hom(G, Z/pZ) is the abelian group of homomorphisms from G to Z/pZ,
(b) Hom(G, Z)/pZ is the abelian group of homomorphism h from G to Z/pZ
such that for some homomorphism g from G to Z we have x G
g(x)/pZ = h(x).
More generally (see [Sh 664, 3] except separating g

), the point is that asking what

can r
p
(Ext(G, Z)) be when G is an abelian group of cardinality , we can translate
the situation to a -system:
5.3 Denition. 1) We say Y = (

A,

K,

G,

D) is a -system if
(A)

A = A
i
: i ) is an increasing sequence of sets, A = A

= A
i
: i <
(B)

K = K
t
: t A) is a sequence of nite groups
(C)

G = G
i
: i ) is a sequence of groups, G
i

tA
i
K
t
, each G
i
is closed
(under the Tichonov topology) and i < j G
i
= g A
i
: g G
j

and
G

= g

tA

K
t
: (i < )(g A
i
G
i
), that is, G is the inverse limit
of G
i
: i < ) under the restriction maps
(D)

D = D

: (a limit ordinal)), D

an ultralter on such that <

[, ) D

; the D

s are used to choose limits canonically.

Of course, formally we should write A
Y
i
, K
Y
t
, G
Y

, D
Y

, g
Y
i
, etc., but if clear from
the context we shall not write this.
2) Let Y

be the same omitting D

and we call it a lean -system.

3) We say g

is a Y -candidate if
(E) g

= g

i
: i < ), g

i
G

and g

i
A
i
= e
G
i
= e
K
t
: t A
i
).
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


42 SAHARON SHELAH
We can deduce the result of Sageev Shelah [SgSh 148] (if [G[ = is weakly
compact (>
0
) and p is prime, then r
p
(Ext(G, Z)) r
p
(Ext(G, Z)) = 2

).
(We later get more.) For this note
5.4 Claim. 1) Assume
(a) Y is a -system
(b)

H = H
i
: i < ) is a sequence of groups, =
i,j
: i < j < ),

i,j
Hom(H
j
, H
i
), commuting
(c)

h = h
i
: i < ), h
i
Hom(H
i
, G
Y
i
), and i < j < & x H
j
(h
j
(x))
A
i
= h
i
(
i,j
(x))
(d) H

,
i,
(i < ) form the inverse limit of H
i
,
i,j
: i < j < ), and h = h

the inverse limit of h

i
: i < )
(e) E
h
is the following 2-place relation on G

: f
1
E
h
f
2
f
1
f
1
2
Rang(h)
similarly E
h

for < .
Then
() h Hom(H

, G

)
() if (i < )([A
i
[ & [H
i
[ & [G
i
[ ), then E
h
is a
1
1
[]-
equivalence relation on G

() if (i < )([A
i
[ < & [H
i
[ < ) and is weakly compact uncountable
cardinal, then
(a) the 2-place relation E = E
h
on G

(from clause (e)) is a very nice

equivalence relation
(b) if f
1
, f
2
G

and f
1
f
1
2
/ Rang(h) then for every < large enough
(f
1
A
i
)(f
2
A
i
)
1
/ Rang(h
i
) that is (f
1
E

f
2
) (

< )(f
1

A
i
)
() under ()s assumptions, if [G : Rang(h)] then [G : Rang(h)] = 2

.
2) If for < () we have H

i
: i < ),

i,j
: i < j < ), h

i
: i )
are as in (a) - (e) above and below (which follows for weakly compact) and
i < [H
i
[ + [A
i
[ < , and for every < there are f

i
G

(for i < ) such

that (f

i
E
h

j
) for i < j < & < (), then there are f
i
G for i < 2

such
that i < j < 2

& <

(f
i
E
h

f
j
)
is strong limit and for any f, g G

and < () such that fg

1
/
Rang(h

) for some < we have (fg

1
) A

/ Rang(h

).
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


ON NICE EQUIVALENCE RELATIONS ON

2 SH724 43
Proof. Straightforward.
1) Clause (): Easy.
Clause (): By (b) of clause () proved below it is enough (in Denition 0.2) to
code E
i
for every i < and as is strong limit this is easy.
Clause (): The point is that if f G

Rang(h

) then for some i < we have

i,
(f) G
i
Rang(h
i
) by the tree property of , (one of the equivalent forms of
being weakly compact).
Clause (): By part (2).
2) We shall show the proof such that it works for any strong limit except one point
where we use weak compactness. For each i < , as is strong limit, let

be
(

tA

[K
t
[)
+
if regular,

tA

[K
t
[ + cf() if singular. By the assumption we can
nd f

i
: i < (2

)
+
) such that f

i
G and < & i < j <

(f

i
E
h

j
).
By the choice of without loss of generality i < (2

)
+
f

i
A

= f

0
A

. By
the weak compactness (i.e., see clause () of part (1)) for any i < j <

there is

(i, j) < such that < (f

i
(f

j
)
1
) A

(i,j)
/ Rang(h

(i,j)
).
If = cf() let

= sup

(i, j) : i < j < (

)
+
. Note if is regular then triv-
ially

< and if > cf() by Erdos-Rado theorem without loss of generality

=
sup

(i, j) : i < j <

+

< . So for some club E of we have E

<
Min(E(+1)). Now for any

we dene g

= g
,
: < ) as follows:
g
,
G

is f

()
if E and is e
G

if / E and let f

= f
g

be dened as in
[Sh 664, 3]. Easily (see there)
f

and if
1
=
2
, E,
1
() ,=
2
() and = Min(E(+
1)) then < () (f

1
f
1

2
) A

/ Rang(h

).
Easily we can nd B

[
+

for E such that:

()
1
,
2

and
1
,=
2
B

and < () and = Min(E(+1))

then
(f
g

1
,
:<f

,
1

)(f
g

2
,
:<f

)
1
A

/ Rang(h

).
So restricting ourselves to f

E
B

) we are done, that is, if < () and

1
,=
2

E
B

then we can nd such that

1
=
2
,
1
() =
2
(), so
letting = Min(E(+1)),

( +1) for = 1, 2 we have f

= f

for = 1, 2 so (f

1
f
1

2
) A

/ Rang(h

) hence (f

1
f
1

2
) A

/ Rang(h

)
hence f

1
f
1

2
/ Rang(h

) as required.
5.4
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


44 SAHARON SHELAH
5.5 Remark. We can phrase 5.4(2) forgetting h

, etc., using only E

( < ) and
E

i
= (f, g) G

: (fg
1
) A
i
Rang(h

i
).
5.6 Claim. Assume
(A)(a) is a strong limit cardinal and is a compact cardinal <
(b) K
i
is a group for i <
(c) I is a directed partial order, t I A(t) and

tI
A(t) =
(d) for t I, G
t
is a subgroup of K
i
: i A(t)
(e) for s t from I we have A(s) A(t) and f G
t
f A(s) G
s
(f) G

is the inverse limit of the G

t
s, i.e, f

i<
K
i
: f A
t
G
t
for every
t I
(B)(a) ()
(b) for < (), H

u
,

u,w
: u w from I) is an inversely directed system of
groups
(c) (i) h

u
Hom(H

u
, G
u
) for u I, < ()
(ii) if u w in I and H

w
then (h

w
(x)) A(u) = h

u
(

u,w
(x))
(d) H

, h

, h

,u
are the limit of the inverse system
(e) E

is the equivalence relation on G

: fE

g fg
1
Rang(h

)
(C) for every < we can nd f

: < ) from G

such that
< () & < (f

)
(D) is > sup
i<
[K
i
[ + sup
tI
[A(t)[ and also sup
tI,<()
[H

t
[.
Then there are f

G for < 2

such that < () & < < 2

(f

).
Proof. Let = cf(),
i
: i < ) be increasing with limit . We can choose by
induction on i < , I
i
, A
i
such that
() A
i
, [A
i
[ +[i[ and j < i A
j
A
i
,
i
A
i
() I
i
I is directed, [I
i
[ +[i[ and j < i I
j
I
i
and t I
i
A(t) A
i
() if we restrict ourselves to A
i
, I
i
, there is a sequence f
i

: <
i
), such that
f
i

G
I
i

= Lim
I
i
G
u
, f
u,w
: u w from I
i
) and < () & <
i

(f
i

E
I
i

f
i

) and

j<i
A
j
B
i
, f
i

() = e
K

.
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


ON NICE EQUIVALENCE RELATIONS ON

2 SH724 45
This is straightforward (see the proof of 5.9, rst case). We can extend f
i

to

f
i

such that i B
i


f
i

(i) = e
K
i
. Now we can apply the proof of 5.4.

5.6
5.7 Claim. 1) Assume
(a) > cf() = , and is a measurable cardinal, say D a normal ultralter
on
(b) G is a torsion free abelian group
(c) [G[ =
(d) p is a prime number.
If r
p
(Ext(G, Z)) and =
<
then r
p
(Ext(G, Z))

.
2) Assume
(a) of part (1)
(b) G
i
: i ) is an increasing continuous sequence of torsion free abelian
group
(c)
i
= r
p
(Ext(G
i
, Z)) for i .
Then
() if f Hom(G, Z/pZ) but f / Hom(G, Z)/pZ then for some i < , f
G
i
Hom(G
i
, Z/pZ), f / Hom(G, Z)/pZ
()

i<

i
.
Proof. 1) Choose
i
: i < ) an increasing continuous sequence of cardinal with
limit . Let G
i
: i < ) be an increasing sequence of pure subgroups of G with
union G satisfying i < [G
i
[ =
i
. Now
() if g Hom(G, Z/pZ) and i < g G
i
Hom(G
i
, Z)/pZ then g
Hom(G, Z)/pZ.
[Why? Let g G
i
= h
i
/pZ where h
i
Hom(G, Z) and let h a function
from G to Z be dened as h(x) = n i < : h
i
(x) = n D. Clearly
h Hom(G, Z) and g = h/pZ, as required.]
The result follows by 5.4(2).
2) Similar.
5.7
A complimentary claim is
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


46 SAHARON SHELAH
5.8 Claim. Assume that G
i
: i ) is a purely increasing sequence of torsion
free abelian groups, = cf() for notational simplicity.
1) If r
p
(Ext(G
i
, Z)) : i < ) is not eventually constant then for some closed un-
bounded set C(Ext(G
i
, Z)) we have
(a) r
p
(Ext(G
i
, Z)) : i C) is strictly increasing
(b) there are f
i

: i C, < r
p
(Ext(G
i
, Z))) such that
() f

Hom(G

, Z/pZ)
() f
i

G
i
is constantly zero (of the abelian group Z/pZ)
() if i C, j = Min(C(i + 1)) and < < (G
i
) then (f
i

f
i

)
G
j
/ (Hom(G, Z)/pZ); moreover, f
i

G
i
) + (Hom(G
j
Z)/pZ : <
r
p
(Ext(G
i
, Z))) is independent.
2) If C = sup(C) and the sequence f
i

: i C, <
i
) is as above then
r
p
(Ext(G
i
, Z))

i
.
Proof. Straight.
5.9 Conclusion. If
(a) is a strong limit cardinal and such that () () where
() is above some compact cardinal
() cf() is a measurable cardinal
(b) G is a torsion free abelian group and p is a prime.
Then r
p
(Ext(G, Z)) r
p
(Ext(G, Z)) = 2

.
Proof.
First Case: Let < be a compact cardinal.
For any < we can nd a sequence f
i
: i < ) of members of Hom(G, Z/pZ)
such that i < j f
j
f
i
h/pZ : h Hom(G, Z). As is compact for
i < j < we can nd a pure subgroup G
i,j
of G of cardinality < such that
f
j
G
i,j
= f
i
G
i,j
/ h/pZ : h Hom(G
i,j
, Z).
Let G

be a pure subgroup of G of cardinality + which includes G

i,j
:
i < j < . So r
p
(Ext(G

, Z)) . By 5.4(2) we are done.

(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


ON NICE EQUIVALENCE RELATIONS ON

2 SH724 47
Second Case: Should be clear by the two previous claims.
5.9
5.10 Remark. 1) So for strong limit singular the problem of the existence of G
such that [G[ = , r
p
(Ext(G, Z)) = is not similar to the problem of the existence
of M such that |M| = , nu(M) = where nu(M) = N/

=: N is a model of
cardinality |M|, L
,
-equivalent to M.
That is, we know (in ZFC) that for strong limit singular of uncountable co-
nality, for some model M of cardinality we have nu(M) = (see Shelah and
Vaisanen [ShVs 644] and history there). Now 5.9 is a strong negation of the parallel
of this result for r
p
(Ext(G, Z)).
2) There has been much eort to characterize the class Ext(G, Z) : G a torsion free
abelian group of abelian groups under the assumption V = L (see [MRSh 314] and
references there). We note another possible characterization (in a dierent model
of ZFC).
5.11 Claim. Assume is supercompact, ()( 2

< 2

+
) and Q is the
forcing of adding Cohen reals. Then in V
Q
we have

1
if G is a torsion free abelian group, p a prime and r
p
(Ext(G, Z)) > 0 then for
some (pure) subgroup G

of G of cardinality < 2

0
we have r
p
(Ext(G

, Z)) >
0

2
if G is a torsion free abelian group, then r
p
(Ext(G, Z)), if not nite, has
the form 2

3
in (2) r
p
(Ext(G, Z)) = 2
fr-rk
[p]
(G)
, see below.
5.12 Denition. For a prime p.
1) Let K
p
= G : G is a torsion free abelian group such that even if we add [G[
+
Cohen reals still r
p
(Ext(G, Z)) = 0.
2) For a torsion free abelian group G let
fr-rk
[p]
(G) = Minrk(G

) :G

is a pure subgroup of
G and G/G

K
p
.
Proof. Essentially by [MkSh 418].
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


48 SAHARON SHELAH
REFERENCES.
[EM] Paul C. Eklof and Alan Mekler. Almost free modules: Set theoretic
methods, volume 46 of NorthHolland Mathematical Library. North
Holland Publishing Co., Amsterdam, 1990.
[Fu] Laszlo Fuchs. Innite Abelian Groups, volume I, II. Academic Press,
New York, 1970, 1973.
[GrSh 302] Rami Grossberg and Saharon Shelah. On the structure of Ext
p
(G, Z).
Journal of Algebra, 121:117128, 1989. See also [GrSh:302a] below.
[GrSh 302a] Rami Grossberg and Saharon Shelah. On cardinalities in quotients of
inverse limits of groups. Mathematica Japonica, 47(2):189197, 1998.
math.LO/9911225.
[J] Thomas Jech. Set theory. Springer Monographs in Mathematics.
Springer-Verlag, Berlin, 2003. The third millennium edition, revised
and expanded.
[JMMP] Thomas Jech, Menachem Magidor, William Mitchell, and Karel
Prikry. On precipitous ideals. J. of Symb. Logic, 45:18, 1980.
[MRSh 314] Alan H. Mekler, Andrzej Roslanowski, and Saharon Shelah. On the
p-rank of Ext. Israel Journal of Mathematics, 112:327356, 1999.
math.LO/9806165.
[MkSh 418] Alan H. Mekler and Saharon Shelah. Every coseparable group
may be free. Israel Journal of Mathematics, 81:161178, 1993.
math.LO/9305205.
[Na85] Mark Nadel. L

and admissible fragments. In J. Barwise and S. Fe-

ferman, editors, Model Theoretic Logics, Perspectives in Mathematical
Logic, chapter VIII, pages 271316. Springer-Verlag, New York Berlin
Heidelberg Tokyo, 1985.
[PaSr98] N. Pandey and S. M. Srivastava. A Powerless Proof of a Result of
Shelah on Fundamental Groups. Preprint.
[SgSh 148] Gershon Sageev and Saharon Shelah. Weak compactness and the
structure of Ext(A, Z). In Abelian group theory (Oberwolfach, 1981),
volume 874 of Lecture Notes in Mathematics, pages 8792. Springer,
Berlin-New York, 1981. ed. Goebel, R. and Walker, A.E.
[Sh 124] Saharon Shelah.

may have a strong partition relation. Israel Jour-

nal of Mathematics, 38:283288, 1981.
(
7
2
4
)


r
e
v
i
s
i
o
n
:
2
0
0
3
-
0
7
-
0
2







m
o
d
i
f
i
e
d
:
2
0
0
3
-
1
1
-
1
8


ON NICE EQUIVALENCE RELATIONS ON

2 SH724 49
[Sh 273] Saharon Shelah. Can the fundamental (homotopy) group of a space
be the rationals? Proceedings of the American Mathematical Society,
103:627632, 1988.
[Sh 262] Saharon Shelah. The number of pairwise non-elementarily-embeddable
models. The Journal of Symbolic Logic, 54:14311455, 1989.
[Sh 460] Saharon Shelah. The Generalized Continuum Hypothesis revisited.
Israel Journal of Mathematics, 116:285321, 2000. math.LO/9809200.
[Sh 664] Saharon Shelah. Strong dichotomy of cardinality. Results in Mathe-
matics, 39:131154, 2001. math.LO/9807183.
[Sh 513] Saharon Shelah. PCF and innite free subsets in an algebra. Archive
for Mathematical Logic, 41:321359, 2002. math.LO/9807177.
[ShVs 644] Saharon Shelah and Pauli Vaisanen. On inverse -systems and the
number of L
,
-equivalent, non-isomorphic models for singular.
Journal of Symbolic Logic, 65:272284, 2000. math.LO/9807181.
[ShVs 719] Saharon Shelah and Pauli Vaisanen. On equivalence relations sec-
ond order denable over H(). Fundamenta Mathematicae, 174:121,
2002. math.LO/9911231.