Saharon Shelah - The Generalized Continuum Hypothesis Revisited

(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

THE GENERALIZED CONTINUUM
HYPOTHESIS REVISITED
SH460
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
110 Frelinghuysen Road
Piscataway, NJ 08854-8019 USA
Partially supported by the Israeli Basic Research Fund and the BSF and I would like to thank
Alice Leonhardt for the beautiful typing.
Done - 8/91(1), 9/91(2,3),
Latest Revision - 06/Oct/26
Publ. No.460
Revised according to the proofreading for the Journal
Typeset by A
M
S-T
E
X
1
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

2 SAHARON SHELAH
Annotated Content
0 Introduction
[We explain why we consider the main theorem here a reasonable revision
of GCH (but provable in ZFC).]
1 The generic ultrapower proof
[We prove that for strong limit >
0
for every > for some < ,
there is P []
<
of cardinality such that every A []
<
is the union
of < members of P. We do it using generic ultrapowers. We draw some
immediate conclusions.]
2 The main theorem revisited
[We give a somewhat stronger theorem, with a proof using pcf calculus. We
then show that if is a limit cardinal satisfying the conclusion of the main
theorem, then a Reg & [a[ < [pcf(a)[ . Then we prove a
converse: if (a Reg)([a[ < [pcf(a)[ < ) (or somewhat less) then
the conclusion on the main theorem holds.]
3 Application
[We draw a conclusion on diamonds (and (D)), hence on the omitting types
theorem for L(Q) in the
+
-interpretation and on relatives of the Cantor
discontinuum partition problem: can we divide a topological space such
that no part contains a compactness.]
Appendix: Existence of tiny models
[We show the close connection of the problems to the existence of families
of large sets with pairwise nite intersections to the caes of totally transcen-
dental theories.]
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

THE GENERALIZED CONTINUUM HYPOTHESIS REVISITED SH460 3
0 Introduction
I had a dream, quite a natural one for a mathematician in the twentieth century:
to solve a Hilbert problem, preferably positively. This is quite hard for (at least)
three reasons:
(a) those problems are almost always hard
(b) almost all have been solved
(c) my (lack of) knowledge excludes almost all.
Now (c) points out the rst Hilbert problem as it is in set theory; also being the
rst it occupy a place of honor.
The problem asks is the continuum hypothesis true?, i.e.,
(1) is 2
0
=
1
?
More generally, is the generalized continuum hypothesis true? Which means:
(2) is 2
=
+1
for all ordinals ?
I think the meaning of the question is what are the laws of cardinal arithmetic;
it was known that addition and multiplication of innite cardinals is trivial, i.e.
previous generations have not left us anything to solve:
+ = = max, .
This would have certainly made elementary school pupils happier than the usual
laws, but we have been left with exponentiation only. As there were two operations
on innite cardinals increasing them 2
and
+
it was most natural to assume
that those two operations are the same; in fact, in this case also exponentiation
becomes very simple; usually
= max,
+
, the exception being that when
cf() < we have
=
+
where
cf() =: min : there are
i
< for i < such that =
i<
i
. Non-set
theorists may be reminded that =
+
if =
and =
+1
, and then is
called the successor of and we know cf(
+1
) =
+1
; we call a cardinal regular
if cf() = and singular otherwise. So successor cardinals are regular and also
0
,
but it is hard to come by other regular cardinals so we may ignore them. Note
n<
n
is the rst singular cardinal, and for a limit ordinal > [[ we have
singular, but there are limit =
for which
is singular.
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

4 SAHARON SHELAH
Probably the interpretation of Hilberts rst problem as nd all laws of cardi-
nal arithmetic is too broad
1
, still is cardinal arithmetic simple is a reasonable
interpretation.
Unfortunately, there are some diculties. On the one hand, Godel had proved
that GCH may be true (specically it holds in the universe of constructible sets,
called L). On the other hand, Cohen had proved that CH may be false (by increas-
ing the universe of sets by forcing), in fact, 2
0
can be anything reasonable, i.e.,
cf(2
0
) >
0
.
Continuing Cohen, Solovay proved that 2
n
for n < can be anything reason-
able: it should be non-decreasing and cf(2
) > . Continuing this, Easton proved

that the function 2
for regular cardinals is arbitrary (except for the laws

above). Well, we can still hope to salvage something by proving that (2) holds for
most cardinals; unfortunately, Magidor had proved the consistency of 2
>
+
for all in any pregiven initial segment of the cardinals and then Foreman and
Woodin [FW] for all .
Such diculties should not deter the truly dedicated ones; rst note that we
should not identify exponentiation with the specic case of exponentiation 2
, in
fact Eastons results indicate that on this (for regular) we cannot say anything
more, but they do not rule out saying something on
when < , and we can

rephrase the GCH as
(3) for every regular < we have
= .
Ahah, now that we have two parameters we can look again at for most pairs
of cardinals (3) holds. However, this is a bad division, because, say, a failure for
=
1
implies a failure for =
0
.
To rectify this we suggest another division, we dene to the revised power of ,
for regular < as
[]
= Min
_
[P[ :P a family of subsets of each of cardinality
such that any subset of of cardinality
is contained in the union of < members of P
_
.
This answers the criticism above and is a better slicing because:
(A) for every > we have:
= i 2
and for every regular ,
[]
= .
1
On this see [Sh:g] or [Sh 400a], note that under this interpretation of the problem there is
much to say.
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

(B) By Gitik, Shelah [GiSh 344], the values of, e.g.,
[
0
]
, . . . ,
[
n
]
are essentially
independent.
Now we rephrase the generalized continuum hypothesis as:
(4) for most pairs (, ),
[]
=
Is such reformulation legitimate? As an argument, I can cite, from the book [Br]
on Hilberts problems, Lorentzs article on the thirteenth problem. The problem
was
() Prove that the equation of the seventh degree x
7
+ax
3
+bx
2
+cx+1 = 0 is
not solvable with the help of any continuous functions of only two variables.
Lorentz does not even discuss the change from 7 to n and he shortly changes it
to (see [Br, Ch.II,p.419])
()
Prove that there are continuous functions of three variables not represented
by continuous functions of two variables.
Then, he discusses Kolmogorovs solution and improvements. He opens the second
section with ([Br, p.421,16-22]): that having disproved the conjecture is not solving
it, we should reformulate the problem in the light of the counterexamples and prove
it, which in his case: (due to Vituvskin) the fundamental theorem of the Dierential
Calculus: there are r-times continuously dierential functions of n variables not
represented by superpositions of r times continuously times dierential functions
of less than n variables.
Concerning the fth problem, Gleason (who makes a major contribution to its
solution) says (in [AAC90]): Of course, many mathematicians are not aware that
the problem as stated by Hilbert is not the problem that has been ultimately called
the Fifth Problem. It was shown very, very early that what he was asking people to
consider was actually false. He asked to show that the action of a locally-euclidean
group on a manifold was always analytic, and thats false. Its only the group itself
thats analytic, the action on a manifold need not be. So you had to change things
considerably before you could make the statement he was concerned with true.
Thats sort of interesting, I think. Its also part of the way a mathematical theory
develops. People have ideas about what ought to be so and they propose this as a
good question to work on, and then it turns out that part of it isnt so.
In our case, I feel that while the discovery of L (the constructible universe)
by Godel and the discovery of forcing by Cohen are fundamental discoveries in
set theory, things which are and will continue to be in its center, forming a basis
for ourishing research, and they provide for the rst Hilbert problem a negative
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

6 SAHARON SHELAH
solution which justies our reinterpretation of it. Of course, it is very reasonable
to include independence results in a reinterpretation.
Back to rmer grounds, how will we interpret for most? The simplest ways are
to say for each for most or for each for most . The second interpretation
holds in a non-interesting way: for each for many s,
= hence
[]
=
(e.g.
when 2). So the best we can hope for is: for every for most small
s (remember we have restricted ourselves to regular quite smaller than ). To
x the dierence we restrict ourselves to >
> . Now what is a reasonable

interpretation of for most <
? The reader may well stop and reect. As

all is forbidden (by [GiSh 344] even nitely many exceptions are possible), the
simplest oer I think is for all but boundedly many.
So the best we can hope for is (
is for deniteness):
(5) if >
, for every large enough regular <
,
[]
=
(and similarly replacing
by any strong limit cardinal).

If the reader has agreed so far, he is trapped into admitting that here we solved
Hilberts rst problem positively (see 0.1 below). Now we turn from fun to business.
A consequence is
()
6
for every
for some n and

2
P []
<
of cardinality , every
a []
<
is the union of <

n
members of P.
The history above was written just to lead to (5), for a fuller history see [Sh:g].
More fully our main result is
0.1 The revised GCH theorem. Assume we x an uncountable strong limit
cardinal (i.e., >
0
, ( < )(2
< ), e.g. =
n
where
0
=
0
,
n+1
= 2
n
).
Then for every for some < we have:
(a) < & is regular
[]
=
(b) there is a family P of subsets of each of cardinality < such that every
subset of of cardinality is equal to the union of < members of P.
Proof. It is enough to prove it for singular .
Clause (a) follows by clause (b) (just use P
= a P : [a[ ) and clause

(b) holds by 1.2(4).
2
where []
<
= {a : |a| < }
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

In 1 we prove the theorem using a generic embedding based on [Sh:g, Ch.VI,1]
(hence using simple forcing) and give some applications, mainly, they are reformu-
lations. For example, for
for every regular <
large enough, there is

no tree with nodes and > -branches. Also we explain that this is sucient for
proving that e.g. a topology (not necessarily even T
0
!) with a base of cardinality

and > open sets has at least

+1
open sets relying on [Sh 454a].
In 2.1 we give another proof (so not relying on 1), more inside pcf theory and
saying somewhat more. In 2.10 we show that a property of =
which suces
is: is a limit cardinal such that [a[ < [pcf(a)[ < giving a third proof.
This is almost a converse to 2.9. Now 3 deals with applications: we show that for

, 2
=
+
is equivalent to
+ (moreover =
<
is equivalent to (D)
, a
weak version of diamond). We also deal with a general topology problem: can every
space be divided to two pieces, no one containing a compactum (say a topological
copy of

2), showing its connection to pcf theory, and proving a generalization
when the cardinal parameter is >
. Lastly in an appendix we prove there are no

tiny models for theories with a non-trivial type (see [LaPiRo]) of cardinality
,
partially solving a problem from Laskowski, Pillay and Rothmaler [LaPiRo].
For other applications see [Sh 575, 8]. This work is continued in [Sh 513], for
further discussion see [Sh 666]. For more on the Cantor discontinuum partition
problem see [Sh 668].
We thank Todd Eisworth for many corrections and improving presentation.
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

8 SAHARON SHELAH
1 The generic ultrapower proof
1.1 Theorem. Assume is strong limit singular and > . Then there are only
boundedly many < such that for some (, ) we have pp
(
+
,)
() (so
cf() < < ).
We list some conclusions, which are immediate by older works.
1.2 Conclusion. For every strong limit such that cf() = < < , for some
< we have:
(1) for every a Reg (, ) of cardinality we have
sup pcf
complete
(a) ,
(2) there is no family P of > subsets of such that for some regular (, )
we have: A ,= B P [A B[ < & [A[
(3) cov(,
+
,
+
, ) (equivalently cov(, , , ) as without loss of
generality cf() > ).
Hence
(4) there is P []
<
such that [P[ = and every A []
is equal to the
union of < members of P
(5) there is no tree with nodes and > -branches when (, ) is regular.
Proof. By [Sh:g]; in detail (we repeat rather than quote immediate proofs).
1) Let be as in 1.1. Without loss of generality cf() / [, ).
Note that sup(pcf
-complete
(a)) suppp
(|a|
+
,)
(
) :
= sup(a
) and cf(
)
so cf(
) [a[ < , and easily the latter is by 1.1.

2) By part (4) it is easy (let P
4
[]
<
be as in part (4) and , P
2
be a
counterexample to part (2), so for every A P
2
we can nd P
A
P
4
such
that [P
A
[ < and A = B : B P
A
hence there is B
A
P
A
such that
[B
A
[ = . So A B
A
is a function from P
2
into P
4
and B
A
[A]
and
A
1
,= A
2
P
2
[A
1
A
2
[ < & [A
1
[ & [A
2
[ so the function is
one-to-one so [P
2
[ [P
4
[ , contradiction).
3) By [Sh:g, Ch.II,5.4].
4) Let P
0
[]
<
be such that [P
0
[ and every A []
is included in the
union of < members of P
0
(exists by part (3)). Dene P = B : for some A
P
0
, B A so P []
<
and [P[ [P
0
[ sup2
|A|
: A P
0
= .
Now for every A []
we can nd < and B

i
P
0
for i < such that
A
_
i<
B
i
. Let B
i
= A B
i
for i < so B
i
P and A =
_
i<
B
i
as required.
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

5) Follows by part (2): if the tree is T, without loss of generality its set of nodes is
and the set of -branches cannot serve as a counterexample.
1.2
1.3 Remark. We can let be regular (strong limit >
0
) if we restrict ourselves
in 1.2(1) to [a[ < , and in 1.2(3),(4) to A []
<
as if for
< :
strong limit singular, (
, ) is as in 1.2, then by Fodors lemma for some =

() the set S
< : (
, ) = is stationary: this can serve.

The stimulation for proving this was in [Sh 454a] where we actually use:
1.4 Conclusion. Assume is strong limit, . Then for some < and family
P, [P[ we have: for every n < and (, ) and f : [
n
()
+
]
n+1
, for
some A
n
()
+
of cardinality
+
we have f A P.
Proof. Let be as in 1.2 (or 1.3), and P as in 1.2(4), and let
P
1
= f : f a function from some bounded subset A of into some B P
(hence [B[ < ). As is strong limit and [P[ , clearly [P
1
[ . Now
for any given f : [
n
()
+
]
n+1
, we can nd < and B
i
P for i <
such that Rang(f)
_
i<
B
i
. Dene g : [
n
()
+
]
n+1
by: g(w) = Mini <
: f(w) B
i
, so by the Erdos-Rado theorem for some A
n
()
+
, we have:
[A[ =
+
and g A is constantly i(). Now f A P
1
so we have nished.
1.4
1.5 Conclusion. If =
0
or strong limit of conality
0
, (, T ) is a topology
(i.e. the set of points, T the family of open sets; the topology is not necessarily
Hausdor or even T
0
), B T a basis (i.e. every member of T is the union of
some subfamily of B), and [T [ > [B[ + then [T [ 2
.
Proof. By [Sh 454a] - the only missing point is that for >
0
, we need: for
arbitrarily large < there is (
2
()
+
, ) such that cov([B[,
+
,
+
, ) [B[,
which holds by 1.1 (really in the proof there we use 1.4).
1.5
1.6 Proof of 1.1. Assume this fails. By Fodors Lemma (as in 1.3) without loss of
generality cf() =
0
.
Without loss of generality for our given , is the minimal counterexample.
Let =
n<
n
,
n
= cf(
n
) < ; so for each n there is
n
(, ) such that
pp
(
+
,
n
)
(
n
) ; hence for some a
n
Reg (,
n
) of cardinality and
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

10 SAHARON SHELAH
n
-complete ideal J
n
J
bd
a
n
we have
n
= sup(a
n
) and a
n
/J
n
has true conality
which is . Let
n
= cf(
n
), so
n

n
[a
n
[.
Without loss of generality
n
>
0
hence without loss of generality [a
n
[ < hence
without loss of generality [a
n
[ <
n+1
(and really even [pcf(a
n
)[ <
n+1
), hence
the
n
s are distinct hence the
n
s are distinct, and without loss of generality for
n < we have
n
<
n+1
and
n
<
n+1
< , hence necessarily (by s minimality)
=
n<
n
, hence without loss of generality (see [Sh:E12, 5.2]) tcf(a
n
,
J
n
) =
+
.
It is clear that forcing by a forcing notion Q of cardinality < changes nothing,
i.e., we have the same minimal , etc. (only omit some
n
s). So without loss
of generality
0
=
0
= [a
0
[ = [pcf(a
0
)[ =
1
, and for some increasing sequence
i
: i <
1
) of regular cardinals <
0
()
0
=
i<
1
i
and
i<
1
i
/D
1
has true conality
+
(D
1
is the club lter on
1
).
(Of course, we can alternatively use the generalization of normal lters as in [Sh
410, 5] hence avoid forcing). (How do we force? First by Levy(
0
, <
0
) then
Levy(
0
, [pcf(a
0
)[); there is no change in the pcf structure for a set of cardinals
> [ pcf(a
0
)[, so now [a
0
[ =
1
, sup pcf
1
-complete
(a
0
) > and pcf(a
0
) has cardinality
1
, let a
0
=
: <
1
, pcf(a
0
) =
: <
1
, choose by induction on
<
1
, an ordinal () <
1
such that
,
/ b
[a
0
] : < and
< ,
so
<
1
()
/J
bd
1
is
+
-directed hence without loss of generality
()
: <
1
) is
(strictly) increasing, so we get () and the statement before it). Without loss of
generality
()
1
<
n
[[
1
+
3
(
1
) <
n
for n 1.
Now by [Sh:g, Ch.VI,1] there is a forcing notion Q of cardinality
3
(
1
) (< !)
and a name D
of an ultralter on the Boolean Algebra P(

1
)
V
(i.e. not on subsets
of
1
which forcing by Q adds) which is normal (for pressing down functions from
V ), extends D
1
and, the main point, the ultrapower M =: V
1
/D
(computed in
V
Q
but the functions are from V ) satises:
()
2
for every >
3
(
1
) regular or at least cf() >
3
(
1
), for some
g

1
Ord from V (but depending on the generic subset G of Q), the set
g/
D
: g (
1
Ord)
V
, g<
D
is -like (i.e. of cardinality but every

proper initial segment has cardinality < ), the order being <
D
of course.
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

We shall say in short g
/D
is -like, note that for each there is at most

one such member in M (as the ordinals of M are linearly ordered).
However, we should remember V
1
/D
is, in general, not well-founded; still there is

a canonical elementary embedding j of V into M = V
1
/D
(of course it depends

on G). Note that j maps the natural numbers onto x M : M [= x j(),
but this fails for
1
; without loss of generality j (+1) is the identity. If M [= x
an ordinal let card
M
(x) be the cardinality in V
Q
of y : M [= y < x. Note: also
j() is -like and j(
n
) : n < is unbounded in j().
Without loss of generality for every n 1,
n
> [Q[, and Min(a
n+1
) >
n
. For
every regular (
1
,
+
] there is x
= g
/D
which is -like. Note: g
V (not
V
Q
V ), but we need the generic subset of Q to know which member of V it is. Let
g
,i
: i j() and g
+/D
hence without loss of generality g
+ =
: <
1
)
(why? see (), by [Sh:g, Ch.V] for some normal lter D on
1
and
we have
<
1
/D is
+
-like, and force as above; by renaming we have the above).
Now also without loss of generality for regular (,
+
] and i but <
0
of cardinality
1
(as without
loss of generality g
,i
() <
for <
1
and recall
<
0
). For n 1 denote
c
n
=: Rang(g
,i
) : a
n
, i j() but
< j(
0
), of cardinality j([a
n
[) < j(
n+1
) < j(). Also for every a
n
we have
M [= x
d
n
as x
= g
,i
/D
for some i < i
and Rang(g
,i
) c
n
.
We can apply the theorem on the structure of pcf ([Sh:g, Ch.VIII,2.6]) in M
(as M is elementarily equivalent to V ) and get b
y
[d
n
] : y pcf(d
n
)) M and
f
d
n
,y
t
: t < y) : y pcf(d
n
)
_
M (this is not a real sequence, only M thinks
so).
For y M such that M [= y a limit ordinal (e.g. a cardinal) let
y
be the
conality (in V
Q
) of (x : M [= x an ordinal < y, <
M
). So
()
3
=
(x
)
for Reg, > [Q[
()
4
assume [a : a
M
j(
m
)[ <
n
, then assuming for simplicity 1 < m < n
M [= sup pcf
j(
m
)-complete
(d
n
g
+/D
) g
+/D
.
[Why? Assume not, so M [= sup pcf
j(
m
)-complete
(d
n
g
+/D
) < g
+/D
hence
M [= for every g (d
n
g
+/D
) for some (y
, a
) : < j(
m
)), y
pcf(d
n

(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

12 SAHARON SHELAH
g
+/D
), a
an ordinal < y
we have g < sup

<j(
m
)
f
y
. In V
Q
we have a
n
/J
n
is
+
-directed hence
a
n
(t : t <
M
x
, <
M
)/J
n
is
+
-directed (by ()
3
) hence there
is a function g
such that
(a) Dom(g
) = a
n
(b) g
() <
M
x
= g
/D
(c) if M [= y pcf(d
n
g
+/D
) and a < y then

a
n
: M [= f
d
n
,y
a
(x
) <
M
g
() = a
n
mod J
n
.
By 1.7(1) below we can nd Y V such that [Y [ < [Q[
+
+ = and
a
n
M [= g
() j(Y ). There is g
M such that M [= g
d
n
and
g
() = (sup(j(Y )) ) + 1 < for d

n
(as M [= Min(d
n
) > j()).
By the choice of Y clearly a
n
g
() <
M
g
().
By the choice of
f
d
n
,y
t
: t < y) : y pcf(d
n
)
_
(in Ms sense) and the assumption
toward contradiction we have:
M [= there is a subset of pcf(d
n
g
+/D
) of cardinality < j(
m
) and a
:
) such that ( d
n
)(
() < f
d
n
,
a
()).
Choose such a sequence a
: ) in M and let
i
: i < i()) list the
M
,
so i() <
n
by the hypothesis of ()
4
. Let a
n,i
= a
n
: letting = x
M we
have g
() < f
d
n
,
i
a
i
() V
Q
. Now as g
() < g
(x
), clearly a
n
=
_
i<i()
a
n,i
. So
for some i < i() we have a
n,i
J
+
n
, and we get a contradiction to the choice of g
hence at last we have proved ()

4
.]
Clearly j(c
n
: n < )) is a sequence of length j() = hence j(c
n
: n < )) =
d
n
: n < ), i.e. with n-th element d
n
. Let z M be such that M [= p =
(k
n
, t
n
, s
n
) : n < ) dened by: k
n
< is the maximal k such that g
+/D
sup pcf
j(
k
)-complete
(d
n
g
+/D
), and t
n
is the minimal cardinal such that sup
pcf
j(
n
)-complete
(d
n
) is g
+/D
and cf() = s
n
so s
n
j(
n
). As j() is
-like clearly (m < )(n < )(m < n & [x M : x
M
(j(
m
))[ <
n
) hence
by ()
4
above necessarily (m < )(n < ) [[[s
n
][
m
], but j() is the limit of
j(
n
) : n < ) M, hence M [= j() = lims
n
: n < ). Now
()
5
M [= j(), g
+/D
form a counterexample to the Theorem 1.1.

(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

But as j is an elementary embedding of V to M, the choice of (minimal) implies
M [= there is no
< j() such that j(),
form a counterexample to the theorem.

But as Rang [g
+/D
] < j(
0
) < j(), clearly we have M [= g
/D
< j().
By the last two sentences we get a contradiction to ()
5
.
1.1
1.7 Observation. Let Q, D
, G Q, V
Q
, M, j be as in the proof 1.6. Let for z M,
[z] = t : M [= t y. So
(1) If Y V
Q
, Y M, = Max
_
[Y [
V
Q
, [Q[
V
_
then for some y V, [y[
V

and x[x Y M [= x j(y)].
(2) Assume M [= d is a set of regular cardinals > [d[, > j
_
[Q[
V
_
and
y
(when M [= y limit ordinal) is as in 1.6 (its conality in V
Q
).
(a) If M [= y pcf(d), J is (in V
Q
) the ideal on [d] generated by
[b
[d]] : M [= pcf(d) & < y [d b

y
[d]]
then (in V
Q
)
x[d]
x
/J has true conality
y
(b) cf(
y
: y [d]) = max
y
: y [pcf d].
Proof. Straightforward (and we use only part (1)). For (2)(b) remember
M [= y is nite [y] nite.
1.8 Remark. Of course, the proof of 1.1 gives somewhat more than stated (say after
xing
0
=
1
). E.g.,
the cardinal satises the conclusion of 1.1 for
if
> cf() =
0
(as before this suces) and = sup < : is regular
uncountable and there is a forcing notion Q satisfying the -c.c. of cardinality

0
< such that
Q
for every
1
-complete lter D on from V
containing the co-countable sets there is an ultralter D
on P()
V
extend-
ing D as in [Sh:g, Ch.VI,1] for regular cardinal >
+
which is complete for
partitions of from V to countably many parts.
Alternatively, we can phrase the theorem after xing D.
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

14 SAHARON SHELAH
2 The Main Theorem Revisited
We give another proof and get more rened information. Note that in 2.1 if is
strong limit, we can choose R
such that: if < are in R
then 2
< and then
0
R
,,
1
is immediate.
2.1 Theorem. Suppose is a limit singular cardinal satisfying:
for any R Reg unbounded, for some R, > cf() and

1
, cf() <
1
,
1

1
< and R
R
+
unbounded in we have:
0
R
,,
1
if < are in R
, f
: for < , I
a -complete ideal on
extending J
bd
and J is a -complete ideal on , then for some A I

+
and
B
for A satisfying B
= mod J we have
< [f
() : A and B
[ <
1
.
Then
for every > we have:
1
,
for some < we have:
1
,,
for every a (, ) Reg of cardinality < , pcf
complete
(a) .
Before we prove it, note:
2.2 Observation. Assume:
(a) w
n
i
: i <
) is a sequence of pairwise disjoint sets, w

n
=
_
i<
w
n
i
(possibly w
n
i
= for some n and i)
(b)
_
sup
n,i
[w
n
i
[
+
_
< and is uncountable
(c) J
n
is a -complete ideal on w
n
such that w
n
/ J
n
(d) h
n
i
is a function from w
n+1
i
to w
n
i
and h
n
=
_
i<
h
n
i
(e) for every A J
n+1
the set x w
n
: (y w
n+1
)[h
n
(y) = x y A]
belongs to J
n
.
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

Then for some i there are x
n
w
n
i
such that
n
h
n
(x
n+1
) = x
n
.
2.3 Remark. Hence for the J
m
-majority of y w
m
there is x
n
: n < ) as above
such that y = x
m
.
Proof. Without loss of generality w
n
i
: n < , i <
) are pairwise disjoint. Now

we dene by induction on the ordinal for each i <
a set u
i
w
i
=:
_
n<
w
n
i
by:
u
i
=
_
x w
i
:x
_

)[u
i
= u
i
], hence there is for each i <
a unique [i] <

1
+ [w
i
[
+
such that
u
i
= u
[i]
i
[i].
If for some i we have u
[i]
i
,= w
i
, we can easily prove the conclusion so assume
u
[i]
i
= w
i
for every i. Let = sup
i
([w
i
[
+
+
1
), so except when
1
(hence
=
1
= ) we know < . Now we can use clause (e) to prove by induction on
for all n that
u
i
w
n
i
: i <
J
n
(we use J
n
is -complete, > ). But as i = u
i
w
n
i
= w
n
i
we get w
n
J
n
, a contradiction. We are left with the case =
1
so each w
n
i
is nite and
i <
[i] < but then for each m we have u

m
i
w
0
i
: i <
J
0
, so as
J
0
is -complete there is x w
0
such that for each m < and i <
we have
x / u
m
i
w
0
i
. For some i(), x w
0
i()
, so as x / u
n
i()
for some x
n
w
n
i()
we have
h
n1
h
n2
h
0
(x
n
) = x. By Konigs Lemma (as all w
n
i()
are nite) we nish.
2.2
Before we continue we mention some things which are essentially from [Sh:g], and
more explicitly, [Sh 430, 6.7A].
We forgot there to mention the most obvious demand
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

16 SAHARON SHELAH
2.4 Subclaim. In [Sh 430, 6.7A] we can add:
(j) max pcf(b
[a]) = (when dened).

Also in [Sh 430, 6.7] we can add
() max pcf(b
) = .
Proof. This is proved during the proof of [Sh 430, 6.7] (see ()
4
in that proof,
p.103). Actually we have to state it earlier in ()
2
there, i.e. add
() max pcf(b
i,j
) .
We then quote [Sh:g, Ch.VIII,1.3,p.316], but there this is stated.
Lastly, concerning [Sh 430, 6.7A] the addition is inherited from [Sh 430, 6.7].
2.4
2.5 Subclaim. In [Sh 430, 6.7A] we can deduce:
() if a
_
<
a
, [a
[ < , a
, then for some () < and nite

c =
0
,
1
, . . . ,
n
a
()
we have
(i)
>
+1
(ii) a
_
n
b
()
[a]
(iii) ((), ) b
[a] a
= b
()
[a] a
(iv)
= max pcf(a
_
k<
b
()
k
[a])
() moreover,
: n) is denable from a
, () and b
()
[a] : a
)
uniformly;
() if a
: < ) N
, < , [a
[ < then we can have one () for all a
and so
,
: n()) : < ) N
()
.
Proof. Clause (). We choose
by induction on . For = 0 clearly for

some
0
< we have a
0
so a
0
, hence
0
= max pcf(a
) belongs to
N
0
hence to a
for [
0
, ) so by clause (i) of [Sh 430, 6.7A], b
0
[a] :
[
0
, )) is increasing hence b
0
[a] a
: [
0
, )) is eventually constant, say for
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

[
0
, ),
0
(
0
, ). For + 1 apply the case = 0 to a
_
k
b
k
[a] and get
+1
,
+1
.
Clauses (), (). Easier.
2.5
2.6 Claim. 1) Assume
0
is regular, a cardinal, J the -complete ideal
generated by J
<
[a] for a set a of regular cardinals > [a[, a / J, a
i
J for i < <
[a[
+
, a =
_
i<
a
i
and max pcf(a
i
) < .
Then
3
we can nd b, b
i
(i < ) and I such that:
(a) b
i
pcf(a
i
) is nite
(b) b =
_
i<
b
i
(c) I is an ideal on b
(d) for w we have
_
iw
a
i
J
_
iw
b
i
I
(e) I is the -complete ideal generated by J
<
[b]
(f) we have b
i
=
i,
: < n
i
and if I
1
is an
1
-complete ideal on b extending
I (so I
1
= I is O.K. if >
0
), then for any d I
+
1
there are B and
< such that:

()
i,
: i B d,
()
i,
: i B I
+
1
() for every B
B we have
_
iB
b
i
I
1

i,
: i B
I
1
.
2) Assume in addition pcf
i
complete
(a
i
)
i
and
i
then we can nd b,
b
i
(i < ) and I such that:
(a)
b
i
pcf
i
complete
(a
i
)
i
has cardinality <
i
and (b) (e) hold.
3) Assume
(i) I an ideal on
(ii) J an ideal on
(iii)
i
: i < ) a sequence of regular cardinals with tcf(
i<
i
/I) =
3
note that without loss of generality i < a
i
= so necessarily || |a|.
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

18 SAHARON SHELAH
(iv) for i < ,
i
j
: j < ) is a sequence of regular cardinals with
tcf(
j<
i
j
/J) =
i
(v)
j
: j < ) is a sequence of regular cardinals
(vi) [[ +[[ +
j<
j
< min
i
j
: i < , j < .
Then there are for each j < an ordinal
j
<
j
and sets b
j
: <
j
) such that
(a)
_
<
j
b
j

i
j
: i < and if max pcf
i
j
: i < , j < = then equality
holds
(b)
j
=: max pcf(b
j
) is in pcf
j
-complete
(b
j
),
(c) letting J
be the ideal with domain

_
j<
j
j
dened by A J
i
max pcf
j
: (j, ) A < , we have = max pcf

j
: j < , <
j
(d) if w J
, then i < : j < : <

x
[
i
j
b
j
(j, ) w / J I.
(Note that J
is a proper ideal and
(j,) Dom(J
/J
is -directed by basic pcf

theory.)
Proof. By the proof of [Sh:g, Ch.VIII,1.5] or by [Sh 430, 6.7,6.7A,6.7B] (for (1)(f),
shrink A to make n
i
constantly n
, then prove by induction on n
). In more detail:
1) Without loss of generality Min(a) > [a[
+3
. To be able to use [Sh 430] freely
in its notation rename a
i
as e
i
. We apply [Sh 430, 6.7A,p.104] with a, , there
standing for a, [a[
++
, [a[
+
here and without loss of generality e
i
: i < ) N
0
,
N
0
. By the subclaim 2.5 above for each i < there are (i) < [a[
+
and nite
b
i
pcf(e
i
) a
(i)
such that [(i), [a[
+
) e
i

_
b
i
b
+1
[a]. Moreover
(b
i
, (i)) : i < ) N
()
for () = (sup
i<
(i)) + < [a[
+
and let b =
_
i<
b
i
and
I = c b : we can nd < and c
: < ) such that c =

_
<
c
and max
pcf(c
) < . Let us check all the clauses of the desired conclusion.

Clause (a): b
i
pcf(e
i
) is nite.
Holds by the choice of b
i
.
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

Clause (b): b =
_
i<
b
i
.
Holds by the choice of b.
Clause (c): I an ideal on b.
By [Sh:g, Ch.I] and the denition of I.
Clause (d): For w we have
_
iw
e
i
J
_
iw
b
i
I.
Why? By the denition of I and J it suces to prove for each subset w of that
max pcf(
_
iw
e
i
) < max pcf(
_
iw
b
i
) < .
First assume max pcf(
_
iw
e
i
) < . Now j w b
j

_
iw
pcf(e
i
) hence (by [Sh:g,
Ch.I,1.11]) pcf(
_
jw
b
j
) pcf(
_
iw
e
i
) so max pcf(
_
iw
b
i
) max pcf(
_
iw
e
i
) < , as
required.
If the other implication fails, then there is w which exemplies it in N
()
(as all the relevant parameters are in it), so we need only consider w N
()
.
Assuming w N
()
and max pcf(
_
iw
b
i
) < let b
=:
_
iw
b
i
, so b
N
()
a
()
and max pcf(b
) < , and by [Sh 430, 6.7A,](h) for some nite c pcf(b
) N
()
we have
_
c
b
()
[a] includes b
, recalling that () is a limit ordinal.

By [Sh 430, 6.7A](f), i.e., smoothness
b
b
()
[a]
_
c
b
()
[a]
hence
_
iw
e
i

iw
e
i
iw
b
()
[a] : b
i

iw
_
c
b
()
[a]

_
c
b
()
[a].
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

20 SAHARON SHELAH
So
_
iw
e
i

_
c
b
()
[a] hence
max pcf(
_
iw
e
i
) max pcf(
_
c
b
()
[a])
max
c
(max pcf b
()
[a]) = max(c) <

(we use subclaim 2.5 above).
Clause (e): I is the -complete ideal generated by J
<
[b].
By the choice of I.
Clause (f): As I
1
is
1
-complete for some n
the set d b
i
: [b
i
[ = n
belongs
to I
+
1
. Now we try to choose by induction on n
+ 1 a set B
decreasing
with such that:
()
i,k
d : i B
and k < I
+
1
() for each k < the set
i,k
: i B
belongs to I
1
.
For = 0, the set B
0
= i < : [b
i
[ = n
is O.K.: in clause () we ask

_
i<
b
i
d I
+
1
, by which we mean d I
+
1
which is assumed and Clause () is
empty (no k < !) lastly by the choice of n
we are done.
For +1, if , B
are not as required, then there is B
such that the statements
_
iB
b
i
I
1
and
i,
: i B
I
1
have dierent truth values.
By obvious monotonicity this means
_
iB
b
i
/ I
1
,
i,
: i B
I
1
so let B
+1
=
B
.
If B
n
+1
is well dened we have by clause () that
i,k
: i B
n
+1
and k
n
+ 1 I
+
1
but as B
n
+1
B
0
this set is empty, easy contradiction.
2) Same proof except that, for dening b
i
, instead of quoting 2.5 we use
[Sh 430, 6.7A](h)
+
. We could have used it in the proof of part (1) here.
3) We apply [Sh 430, 6.7A] to a =:
i
j
: i < , j <
i
: i < and
without loss of generality
i
: i < ), I, J,
j
: j < ) and
i
j
: j < ) : i <
_
belong to N
0
. Let a
[a] be such that J
[a] = J
<
[a] +a
and let e
j
=
i
j
:
i < a
but if possible a
= a. Again by [Sh 430, 6.7A,](h)

+
for each j there is
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

c
j
pcf
j
-complete
(e
j
) such that e
j

_
c
j
b
+1
[a]. Let c
j
=
j
: <
j
with no
repetitions and let b
j
= b
+1
[a] e
j
.
Now clause (a) holds by the choices of c
j
and b
j
. As for clause (b) note that

max pcf(b
j
) =
j
by 2.4, i.e. clause (j) of [Sh 430, 6.7A] and clearly

j

pcf
j
-complete
(e
j
) but
j
/ pcf(e
j
b
j
) by clause (e) of [Sh 430, 6.7A] so neces-

sarily
j
pcf
j
-complete
(b
j
), that is clause (b) holds.

Let J
be the ideal with domain

_
j<
j
j
dened by J
= A Dom(J
) :
max pcf
j
: (j, ) A < . By transitivity of pcf, pcf(

i
j
: i < , j < )
hence by the choice of a
, e
j
clearly = max pcf(
_
j<
e
j
).
As in the proof of clause (d) of part (1) we have
() for w we have
max pcf(
_
iw
e
i
) < max pcf(
_
iw
c
i
) < .
We conclude that = max pcf(
_
<
c
i
) hence J
satises clause (c) (well maybe

c
i
1
c
i
2
,= ? Remember [Sh:g, Ch.I,1]).
Lastly, we prove clause (d) so assume w J
, so by the denition of J
, we have
max pcf(d) < where d =
j
: (j, ) w. So by transitivity of pcf ([Sh:g,

Ch.I,1.11]) as = tcf(
i<
i
/I) necessarily B =: i < :
i
pcf(d) I. Now
for each i B we have
i
/ pcf(d) hence
i
/ pcf(d e
i
) but
j
0
, pcf
complete
(a) then for some , = cf()
[a[, and
i
: i < ) we have:
i
regular,
i
pcf(a) and for some -complete
ideal I J
bd
we have = tcf(
i<
i
/I).
Proof. Without loss of generality = max pcf(a), otherwise replace it by b
[a];
let J be the -complete lter on a which J
<
[a] generates. Let be minimal such
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

22 SAHARON SHELAH
that J is not
+
-complete so necessarily = cf() [a[; as we can replace a
by any a
a, a
/ J
<
[a] without loss of generality a is the union of members of
J, so for some a
i
J (for i < ) we have a =
_
i<
a
i
, as J is -complete without loss
of generality a
i
J
<
[a]. By 2.6(1), we have b
i
: i < ), b and I as there. As J is
-complete
_
iw
b
i
: [w[ < I, so by applying clause (f), we can nish.
2.7
Proof of 2.1. We shall prove
1
,
by induction on . Arriving to , assume it
is a counterexample so necessarily > , cf() = cf(). For each < there is
a (, ) such that [a[ < and pcf
-complete
(a) , so by [Sh:g, ChIX,4.1] without
loss of generality for some -complete ideal J on a,
+
= tcf(a/J). So (by 2.7)
the following subset of (cf(), ) Reg is unbounded in (by 2.7):
R =:
_
:cf() < = cf() < and there is
,
: < ),
a sequence of regular cardinals (, )
and a complete ideal I
on extending J
bd
such that
<
,
/I
has true conality

+
_
.
Let ,
1
, R
be witnesses for
0
(i.e.
0
R
,,
1
holds), without loss of generality
otp(R
) = cf() and remember cf() <

1
,
+
< Min(R
), R. Let
= , we
now dene by induction on n the following: J
n
, w
n
, w
n
i
: i < ),
x
: x w
n
), h
n
as in observation 2.2 such that x w
n
:
x

+
J
n
and h
n
i
(y) = x &
x
>
+

y
<
x
, so we shall get a contradiction (as the domain of h
n
is
the set x w
n+1
:
x
>
+
). We also demand
xw
n
x
/J
n
is
+
-directed and
[x w
n

+
<
x
< ] hence
+
x
< . We let w
0
i
= i,
i
=
,i
, and J
0
= I
.
Suppose all have been dened for n. Now by the induction hypothesis on (as
= sup(R
)) for every x w
n
, if
x
>
+
then for some = [
x
] R
we have
a (,
x
) & [a[ < pcf
-complete
(a)
x
.
Remember J
n
is [R
[
+
-complete (as > cf()), so it is enough to deal separately
with each u
n,
= u(n, ) =: x w
n
: [
x
] = and
x
>
+
where R
. If
u
n,
J
n
we have nothing to do. Otherwise choose
> , and I
,
: <
) witnessing
R. By [Sh:g, Ch.IX,4.1] applied to
,
<
+
=
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

tcf
xu(n,)
x
/J
n
, for each <
we can nd a sequence
n,,
x
: x u
n,
),
n,,
x
regular <
x
but
+
and
xu(n,)
n,,
x
/J
n
has true conality
,
.
Now apply 2.6(3) with , , I, J, ,
i
: i < ),
i
j
: j < ),
j
: j < )
there standing for
, u(n, ), I
, J
n
u(n, ),
+
,
,
: <
),
n,,
x
: x
u(n, )), : x u(n, )). This gives us objects b
n,,
x
: x u(n, ), <
x
) and
J
n,
as there. We could have changed some values of
n,,
x
to
+
to guarantee that
+
= max pcf
n,,
x
: x u(n, ), <
, so without loss of generality

n,,
x
:
<
=
_
<
x
b
n,,
x
. By 2.6(3),clause () we have
()
1
if w Dom(J
n,
) and
<
: x u(n, ) : ( <
x
)[
n,,
x
b
n,,
x
& (x, ) w] / J
n
/
I
, then w / J
n,
.
Let I
n,
be the ideal on Dom(J
n,
) dened by
w I
n,
_
<
: x u(n, ) :( <
x
)[
n,,
x
b
n,,
x
&
(x, ) w] / J
n
_
I
.
Now ()
1
tells us that J
n,
I
n,
. Note that since I
and J
n
are -complete
proper ideals we assumed u(n, ) / J
n
we have that I
n,
is a -complete proper
ideal on Dom(J
n,
). This means that if we want to verify that a set is not in the
-complete ideal generated by J
n,
, it suces to see it is not in I
n,
.
By 2.6(3), clause (b) we have
n,,
x
=: max pcf(b
n,,
x
) is in pcf
-complete
(b
n,,
x
).
Since b
n,,
x

x
, our choice of [
x
] = guarantees
()
2

n,,
x
= max pcf(b
n,,
x
) <
x
.
For <
, let f
n,
: u(n, ) be dened by f
(x) = Min <

x
:
n,,
x

b
n,,
x
. Now we can apply the choice of
1
, (i.e. for them
0
R
,,
1
holds) only
instead of J a -complete ideal on we have here J
n
is a -complete ideal on
a set of cardinality and actually use J
n
u
n,
. So we get A
n,
I
+
and
B
n,
= u(n, ) mod J
n
for A
n,
such that:
()
3
x u
n,

1
> [f
n,
(x) : A
n,
, x B
n,
[.
Let us dene
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

24 SAHARON SHELAH
w
n+1
i,
= (x, , ) : ( A
n,
)[x B
n,
& = f
n,
(x) & x w
n
i
]
h
n
i,
: w
n+1
i,
w
n
i
is h
n
i,
((x, , )) = x when
x
>
+
x u
n,

(x,,)
=
n,,
x
.
Recall we are assuming u
n,
J
+
n
, if i u
n,
J
n
we let w
n+1
i,
= . Now we
switch integrating on all R
:
w
n+1
i
=
_
R
w
n+1
i,
We let
w
n+1
=
_
R
_
i<
w
n+1
i,
, h
n
=
_
R
_
i<
h
n
i,
.
J
n+1
=
_
u w
n+1
: for some i < and u
j
u for j max pcf
(x,,)
: (x, , ) u
j
_
.
Most of the verication that w
n+1
, h
n
and J
n+1
are as required is routine; we
concentrate on a few important points
0
[w
n+1
i
[ <
1
[Why? By ()
3
, as cf() <
1
< so the does not cause a problem.]
1
if x w
n
,
x
>
+
and h
n
(y) = x, then
y
<
x
[Why? Choose such that x u(n, ). If u(n, ) J
n
then
y
=
+
<
x
.
If u(n, ) / J
n
then we are done by ()
2
.]
2
w
n+1
/ J
n+1
[Why? Choose R
with u(n, ) / J
n
, and let v(n, ) = (x, ) :
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

(x, , ) w
n+1
.
For A
n,
,
B
n,
x u(n, ) : ( <
x
)[
n,,
x
b
n,,
x
(x, ) v(n, )],
and so v(n, ) / I
n,
. Thus v(n, ) is not in the -complete ideal generated
by J
n,
, and the denitions of J
n,
and J
n+1
imply w
n+1
/ J
n+1
.]
3
For every A J
n+1
the set B =: x w
n
: (y w
n+1
)[h
n
(y) = x y
A] belongs to J
n
.
[Why? Suppose toward contradiction that B J
+
n
, and choose R
such that B u(n, ) J

+
n
. Let A = (x, , ) w
n+1
: x B, and let
A
= (x, ) : (x, , ) A. For A

n,
as B
n,
= u(n, ) mod J
n
clearly
B B
n,
J
+
n
; also
B B
n,
x u(n, ) : ( <
x
)[
n,,
x
b
n,,
x
(x, ) A
],
and since BB
n,
J
+
n
by the denition of I
n,
, we know A
/ I
n,
hence
A
1
/ J
n+1
but by the denition of B, A clearly A
1
A hence A / J
n+1
,
contradiction.]
Thus we have carried out the induction and hence get by 2.2 the contradiction and
nish the proof.
2.1
2.8 Remark. 1) We can be more specic phrasing 2.1: let R
be unbounded,
: R
),
a set of ideals on ; the desired conclusion is: for every

> for some
< we have: if R
,
i
(, ) Reg for i < , J,
J
then pcf
i >
1
, an uncountable limit cardinal and we have:
1.5
,
for every (,
], we have
1
,
(from the conclusion of 2.1).
Then
,
() a (,
), a Reg, [a[ < [
pcf(a)[
() if is regular then (for a Reg):
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

26 SAHARON SHELAH
a (,
) & [a[ < [
pcf(a)[ < .
Proof. Let () =
_
+
if is singular
if is regular .
So assume a (,
) Reg, [a[ < , and
pcf(a) has cardinality ().

Let
0
= Min(a) and
i+1
: i ()) list the rst (() + 1)-members of
(pcf(a))
0
(remember pcf(a) has a last member), and for limit (), let
=
_
i
_
<
i(), and
i()
> sup pcf
-complete
(
i()
: < ).
Why is this possible? We know pcf
-complete
(
i()
: < ) cannot have a member

()
, (hence >
()
being regular), by the choice of . Also
pcf
-complete
(
i()
: < ) cannot be unbounded in
()
(because cf(
()
) =
() (remember () is regular) as then it will have a member >
()
, see
[Sh:g, Ch.I,1.11]). So it is bounded below
()
hence i() exists.
Now we get contradiction to [Sh 410, 3.5], version (b) of (iv) there
(use e.g.
i()
: < ( +[a[)
+4
)). (Alternatively to [Sh 430, 6.7F](5)).
2.9
2.10 Theorem. Let be a limit uncountable singular cardinal, < and
[[a[ < & a Reg (, ) [ pcf(a)[ < ]
or at least:
,
for every large enough Reg , we have:
,
if a Reg (, ), [a[ < then [ pcf
-complete
(a)[ <
Then for every large enough < we have
1
,
of 2.1, hence cov(, , , ) = .
Remark. This proof relies on [Sh 420, 5].
Proof. Without loss of generality cf() =
0
(e.g. force by Levy(
0
,cf()) as nothing
relevant is changed, or argue as in 1.3, as
,
implies that for each [cf(), ),
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

the cardinal sup pcf
-complete
(a) : a Reg (, ) and [a[ , however, we
can just repeat the proof).
Assume this fails. Without loss of generality is minimal, so cf() =
0
. Failure
means (by 2.7) that = sup(R) when
R =
_
: Reg and for some
Reg (, ) for < ,

and -complete ideal I on
we have
+
= tcf(
<
/I)
_
.
For simplicity assume that for < and A (2
)
+
, in K[A] there are Ramsey
cardinals > . This makes a minor restriction say for one we may get
+
instead of <
+
(which is equivalent to < ).
So by [Sh 420, 5], for some uncountable regular < from R cf()
+
,
,
from
the assumption of the theorem holds and for some family E of ideals on normal
by a function : and J E and
i
= cf(
i
) (, ),
+
= tcf
_
i<
i
/J
_
and
i
: i < ), J minimal in a suitable sense, that is () = rk
3
J
(
i
: i < ), E)
is minimal so without loss of generality rk
3
J
(
i
: i < ), E) = rk
2
J
(
i
: i < ), E).
Hence we do not have A , A / J and
i
(, ) Reg such that
i
:
i < ) <
J+A

i
: i < ) and
+
= tcf(
i<
i
/J). As cf() =
0
, we can nd
n
: n < ), <
n
R and =
_
n<
n
. As is minimal there is a partition
u(n) : n < ) of , such that:
() i u(n), n < , [a[ < , a Reg (,
i
) pcf
n
-complete
(a)
i
.
So for some n we have u(n) J
+
. Without loss of generality (i < )(
i
>
+
)
and (as >
0
) for some n = n() we have u(n) = (i.e. the minimality of ()
is preserved). Choose R large enough such that (a)
_
a Reg (, ) &
[a[
n()
+ [ pcf
-complete
(a)[ <
_
. (Why is this possible? As
,
which
holds by the choice of ). As R we can choose a sequence
: < ) and
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

28 SAHARON SHELAH
I J
bd
a -complete ideal on such that
(, ) and tcf
_
_
<
/I
_
_
=
+
.
By [Sh:g, Ch.IX,4.1] we can nd
i
= cf(
i
) (,
i
),
i
<
i
such that
=
tcf
_
i<
i
/J
_
.
Now a =: pcf
-complete
i
: i < , < has cardinality < (by the choice
of ) and has a smooth closed representation b
(a) : a) (see [Sh 430, 6.7]).

For i < there is c
i
pcf
n()
-complete
i
: < such that [c
i
[ <
n()
and
<
i
b
(a) : c
i
(by the choice of n() and by [Sh 430, 6.7], note that
< <
n()
by their choices hence pcf
n()
-complete
i
: < a hence all is
O.K.). Also c
i

i
because we are assuming u
n()
= .
Let
d =:
_
tcf(
iA
i
/(J +A)) :
i
: i < )
i<
c
i
and A J
+
and
tcf(
iA
i
/(J +A)) is well dened
_
Let c =
_
i<
c
i
. So [c[ +
n()
hence pcf
-complete
(c) has cardinality < , and
d by the choice of () and d pcf
-complete
(c) hence [d[ < (by the choice of
).
Now if
+
pcf(c) then
B
= < : i < :
i
b
[a] / J I.
[Why? Otherwise B
pcf(b
[a]) hence pcf(b
[a]) includes
pcf
: B
, but as B
/ I the cardinal
+
belongs to the latter; but max
pcf(b
[a]) = < contradiction].

But we know that [d[ < , and I is -complete and d pcf(c), so
X =
_
< : for some d we have
i < :
i
b
[a] / J
_
_
d
B
I.
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

So there is some
X, and for i < choose

i
c
i
such that
i
b
i
[a]
(well dened by the choice of c
i
). So by smoothness of the representation
d i < :
i
b
[a] i < :
i
b
[a] J.
Now by the pcf theorem for some A J
+
we have
iA
i
/J has true conality
which we call , so necessarily pcf
-complete
(
i
: i A) d (see the
denition of d) but this contradicts the previous sentence (recall d by the
minimality of ()).
2.10
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

30 SAHARON SHELAH
3 Applications
Of course
3.1 Claim. If is as in 2.1, then the conclusions of 1.2 and 1.1 hold.
3.2 Claim. If
then:
(a) 2
=
+

+
(b) =
<
i (D)
.
Where we remember
3.3 Denition. 1) (D)
means that:
is regular uncountable and there is

P = P
: < ) such that P
is a
family of < subsets of satisfying:
() for every A , < : A P
is a stationary subset of .
2) (D)
S
(S stationary) means regular and there is

P as above such that:
() for every A we have S : A x / P
is not stationary.
3) (D)
+
S
where S is stationary, regular uncountable means that: for some
P as above:
() for every A for some club C of we have:
S C A P
& C P
.
4) Let be regular uncountable, S stationary. Now
S
means that there is
A
: S) such that A
and for every A the set S : A = A
is a stationary subset of .
5) For regular uncountable and S stationary (D)
S
means that for some
P
: S) as above for every A the set S : A P
is stationary.
3.4 Remark. 1) If is a successor cardinal, (D)
is equivalent to
(by Kunen),
so (a) is a particular case of (b) in 3.2.
2) By [Sh 82], [HLSh 162], if (D)
then the omitting types theorem for L(Q) for

-compact models in the
+
-interpretation holds (and more). Now =
<
is the
standard assumption to the completeness theorem of L(Q) in the
+
-interpretation;
and is necessary and sucient when we restrict ourselves to -compact models. So
the question arises, how strong is this extra assumption? If G.C.H. holds (D)

(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

=
<
for every ,=
1
(by [Sh 82], continuing Gregory [Gre]); and more there.
Here we improve those theorems. Now 3.2 says that above
, the two conditions

are equivalent.
3) We may consider the function h : Car, demanding [P
[ < h().
4) Remember that for >
0
regular and stationary S = S
1
S
2
we have
(D)
+
S
(D)
S
(D)
S
and (D)
S
1
(D)
S
2
but (D)
S
(D)
S
1
, (D)
+
S
2
(D)
+
S
1
.
3.5 Proof of 3.2. By 3.4(1) it suces to prove clause (b). Trivially (D)
<
, so assume =
<
, and let
A
i
: i < list the bounded subsets of , each appearing times.
For each < let
R
= <
: cov([[,
+
,
+
, ) < and is regular.
We know (by 1.2(3)) that for each (
, ), R
contains a co-bounded subset

of Reg
, say Reg
. So for some n
<
S
= < : >
, n
< n
is unbounded in ; hence trivially S
= (
, ). So R =: < : is regular, 2
<
, and for every < we have cov([[,
+
,
+
, ) < contains Reg (
n
,
).
As = cf() >
, for each < , R there is P
, a family of < subsets of

of cardinality such that if A , [A[ = then A is included in the union of
< members of P
.
Let P
= B : for some R(+1) and A P
we have B A so P
is a
family of < subsets of . For each A we dene h
A
: by dening h
A
()
by induction on : for non-limit h
A
() is the rst ordinal i >
_
<
h
A
()+1 such
that A = A
i
and for limit h
A
() =
_
<
h
A
(). So h
A
() is strictly increasing
continuous, hence h
A
() and h() = [( limit) & ( < )(h
A
() < )].
Let
P
0
=:
_
_
B
A
: B P
_
P
=: P
0

_
< : h
A
() = : A P
0
_
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

32 SAHARON SHELAH
(remember A
: < ) lists the bounded subsets of each appearing unboundedly

often).
Now for any A we have E =: E
A
=:
_
< : limit and
<
h
A
() <
_
is a
club of , and
()
1
cf() < E & cf() R A P
0
[Why? Let =: cf(), and let

j
: j < ) be an increasing sequence of successor
ordinals with limit , hence h
A
(
j
) : j < ) is (strictly) increasing with limit ;
so for some < = cf() and B
i
P
cf()
for i < we have h

A
(
j
) : j <
_
i<
B
i
, so for some i, h
A
(
j
) : j < , h
A
(
j
) B
i
is unbounded in , and clearly
B
=: h
A
(
j
) : j , strong limit then for some
< we
have
+
{<:
<cf()<}
.
2) Similarly, for =
<
inaccessible, strong limit < for some
< ,
(D)
+
{<:
< cf()<}
holds.
3) If =
<
, and
S = < : cf() < , 2
cf()
< , and [ > cov([[, cf()
+
, cf()
+
, cf()] then
(D)
+
S
; so if is a successor cardinal we have
+
S
.
4) Assume
4
=
<
> = cf() > = cf(),
< ,
+
< , S , S :
cf() = is stationary,

C = C
: S), for S, C
is a closed subset of
, [ C
cf() = S & C
= C
] and S cf() =
otp(C
) = . Assume further that for no < is there P a : [a[ =

satisfying [P[ = , such that [a P & b P & a ,= b [a b[ < ], and
[a RegMin(a) > & [a[ (b [a]
)( > max pcf(a)] (e.g.

successor has such

C).
Then (D)
S
holds where S
= S : cf() = .
Proof. Easy, e.g.: 4) By [Sh:g, Ch.III,2] without loss of generality for every club
E of for some E S, C
E if
+
< and otp(C
E) < otherwise. Let

4
if =
+
, = cf() > = cf() > = cf() then there are S,

C as in 3.6(4)(see [Sh 351, 4]
or [Sh:g, Ch.III,2.14] + [Sh:E12]). Of course, we get not just guessing on a stationary set but on
a positive set modulo a larger ideal.
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

=
3
()
+
, let M
i
: i < ) be such that: M
i
(H (), , <
), |M
i
| < ,
M
i
, M
i
an ordinal, M
j
: j i) M
i+1
. Let for S
, P
= M
+1
P().
It is enough to show that

P = P
: S
) exemplies (D)
S
. So let X
: <
) M
0
list the bounded subsets of each appearing times. Let X , E
0
be
a club of ; we dene by induction on , h
X
() < as the rst < such that
>
_
<
h
X
() and X = X
. Let M
i
: i < ) be chosen as above but also
h
X
M
0
, M
i
: i < ) M
0
, E
0
M
0
. Let E =: E
0
: M
= = M
,
clearly it is a club of . Let S E, cf() = be such that C
E. Now we
imitate the proof [Sh 410, 6] or directly as in [Sh 420, 1] for h
X
(C
E).
3.6
3.7 Claim. Above instead demanding on = cf() & 2
< & [ <

cov([[,
+
,
+
, ) < ] it suces to demand = cf() < and if T is a tree
with -levels and < nodes then T has < -branches.
See [Sh 589, 2] for a pcf-characterization of this property.

3.8 Lemma. 1) Suppose c is an operation on X, i.e., c is a function from P(X)
to P(X). Assume further
< =
and we let
P
=
_
A X :[A[ = and for every B A satisfying [B[ =
there is
B
B, [B
[ = such that c(B
) A,
and [c(B
)[ =
_
.
If
<
() then there is function h : X such that: if A P
then
h A is onto .
2) Actually instead of
() we just need a conclusion of it:

()
1
= ()
1
,
( )()[ Reg &
& cov(,
+
,
+
, ) = ],
or even just a conclusion of that:
()
2
= ()
2
,
for every for some < ,
we have:
=
,
: there is no family P of > subsets of each of

cardinality with the intersection of any two having cardinality <
.
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

34 SAHARON SHELAH
3.9 Remark. 1) The holding of ()
2
is characterized in [Sh 410, 6].
2) On earlier results concerning such problems and earlier history see Hajnal,
Juhasz, Shelah [HJSh 249]. In particular, the following is quite a well known
problem:
the Cantor discontinuum partition problem: Can every topological space

be divided into two pieces, such that no part contains a closed homeomor-
phic copy of

2 (or any topological space Y such that every scattered set
is countable, and the closure of a non-scattered set has cardinality contin-
uum)?
3) Note that the condition in ()
2
holds if = 2
0
>
, =
0
,
=
1
and
1
(from 2.1) (which holds e.g. if V = V

P
0
, P a c.c.c. forcing making the continuum
>
V
0
). So in this case the answer to is positive.

4) Also if = 2
0
>
1
, and ()[ 2
0

,
1
] then the answer to in

(2) is yes; now on
,
1
see [Sh 410, 6].

Proof. We prove by induction on [, [X[] that:
()
if Z, Y are disjoint subsets of X, [Y [ , then there are h, Y

+
such that
(a) Y Y
+
XZ
(b) [Y
+
[
(c) h is a function from h : Y
+
to
(d) if A P
< ,
, [A Y
+
[ and [A Z[ < then
h (A Y
+
) is onto .
Case 1. = , so [Y [ .
Without loss of generality [B Y & [B[ & [c(B)[ = c(B)Z Y ].
Now just note that P
Y
=: c(B) Y : B Y, [B[ , [c(B) Y [ = has
cardinality =
, and by the denition of P
(using the demand [A Z[ <

in ()
), it suces that h satises: [A P

Y
h Z is onto ], which is easily
accomplished.
Case 2. > .
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

Let =
_
2
_
+
, N
i
: i ) an increasing continuous sequence of elementary
submodels of (H (), , <
), X, c, Y, Z, ) N
0
, +1 N
0
, N
i
: i j) N
j+1
(when i < ) and |N
i
| = +[i[.
We dene by induction on i < , a set Y
+
i
and a function h
i
as follows:
(Y
+
i
, h
i
) is the <
-rst pair (Y
, h
) such that:
(a) Y
X (Z
_
j<i
Y
+
j
)
(b) Y N
i
_
j<i
Y
+
j
Z X N
i
_
j<i
Y
+
j
Z Y
(c) [Y
[ = +[i[
(d) h
: Y

(e) if A P
+|i|
,
< , [A Y
[ ,
[A (Z
_
j<i
Y
+
j
)[ < then h
(A Y
) is onto .
Note: (Y
+
i
, h
i
) exists by the induction hypothesis applied to the cardinal +[i[
and the sets Z
_
j<i
Y
+
j
, X N
i
_
j<i
Y
+
j
. Also it is easy to check that (Y
+
j
, h
j
) :
j i) N
i+1
(as we always choose the <
-rst, hence Y
+
i
N
i+1
).
Let Y
+
=
_
i<
Y
+
i
, h =
_
i<
h
i
. Clearly Y
_
i<
N
i
, hence by requirement (b)
clearly Y Y
+
, (and even XN
Z Y
+
); by requirements (c) (and (a)) clearly
[Y
+
[ = , by requirement (a) clearly Y
+
XZ and even Y
+
= X N
Z.
By requirements (a) + (d), h is a function from Y
+
to . Now suppose A P
< , [A Y
+
[ , [A Z[ < ; we should prove h (A Y
+
) is onto
. So [A N
[ . Choose (
) a pair such that:

(i)

(ii)
+|
|
,
<
(iii) [A N
[ or
=
(iv) under (i) + (ii) + (iii),
is minimal.
This pair is well dened as (, ) satises requirement (i) + (ii) + (iii).
Subcase 1.
is zero.
So [Y
+
0
A[
hence by the choice of h

0
we are done.
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

36 SAHARON SHELAH
Subcase 2.
= i + 1.
So [A N
i
[ < , hence [A
_
j<i
Y
+
j
[ < , hence [A (Z
_
j<i
Y
+
j
)[ < . Clearly
+|i|
holds (as + [i[ = + [
[), so if [A Y
+
i
[
we are done by the choice

of h
i
; if not [A (Z
_
j<i+1
Y
+
j
)[ < and A Y
+
i+1
A N
i+1
= A N
has
cardinality
(and
|Y
+
i+1
|
holds) so we are done by the choice of h
i+1
.
Subcase 3.
limit < .
So for some i <
, [A N
i
[
[why? as
< < ]. Now in N

i+1
there is
a maximal family Q [X N
i
]
satisfying [B
1
,= B
2
Q [B
1
B
2
[ <
]
hence [Q[ + [
[ and without loss of generality Q N

i+1
, hence Q N
so there is B Q, B N
, [B A[
, but [B[ =
< =
hence
[B
[B A]
B A N
]. As A P
there is B
[BA]
with c(B
) A,
[c(B
)[ = . Clearly c(B
) N
hence for some j (i,
), c(B
) N
j
hence
c(B
) X N
j
. So [AN
j
[ . By assumption for some
, ),
+|j|
, so
(j,
) contradicts the choice of (
).
Subcase 4.
limit = .
As N
0
, there is a maximal family Q []
satisfying [B
1
,= B
2
Q
[B
1
B
2
[ <
] which belongs to N
0
. By the assumption ()
2
, we know [Q[ .
We dene by induction on j , a one-to-one function g
j
from N
j
XZ onto an
initial segment of increasing continuous in j, g
j
the <
-rst such function. So

clearly g
j
N
j+1
and let Q
= g
(B) : B Q (i.e. g
(x) : x B : B Q
note: g
is necessarily a one to one function from N
XZ onto ). So for some

B Q
, [B
A[
, so as in subcase 3, for some B
, B
B A,
[B
[ = , c(B
) A, [c(B
)[ = ; so for some i < , c(B
) N
i
. But [A Z[ <
so [A Y
+
i
[ = and by assumption ()
2
, for some ,
< we have
+|i|
,
contradicting the choice of (
) (i.e. minimality of
).
3.8
3.10 Discussion. 1) So if we return to the topological problem (see of 3.9(2)), by
3.8 + 3.9(4), if 2
0
>
1
we can try =
2
,
=
0
, =
1
. So a negative
answer to (i.e. the consistency of a negative answer) is hard to come by: it
implies that for some ,
,
1
, a statement which, when >

1
at present we do
not know is consistent (but clearly it requires large cardinals).
2) If we want = 2
0
=
2
, =
1
=
we should consider a changed framework.

We have a family I of ideals on cardinals < which are -based (i.e. if A I
+
,
I I (similar to [HJSh 249]) then B [A]
(B I
+
)) and in 3.8 replace P
by
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

P
= P
I
=:
_
A X :[A[ = and for every pairwise distinct
x
A for < we have

u : [cx
: u[ <
is included in some I I
_
.
and replace ()
2
by
()
3
For every assume
F (, I, f) : I I, = Dom(I), f : is one to one
and if (
, I
, f
) F for = 1, 2 are distinct then <

2
: f
2
() Rang f
1
I
2
.
Then [F[ .
Note that the present P
ts for dealing with of 3.9(2) and repeating the proof

of 3.8.
3.11 Discussion of Consistency of no: There are some restrictions on such theo-
rems. Suppose
() GCH and there is a stationary S <
+1
: cf() =
1
and
A
: S) such that: A
= sup A
, otp(A
) =
1
and
1
,=
2
[A
1
A
2
[ <
0
.
(This statement is consistent by [HJSh 249, 4.6,p.384] which continues [Sh 108].)
Now on
1
we dene a closure operation:
c(u) ( S)[ A
& (u A
)
0
].
This certainly falls under the statement of 3.8(2) with =
=
0
, =
1
except
the pcf assumptions ()
1
and ()
2
fail. However, this is not a case of our theorem.

(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

38 SAHARON SHELAH
Appendix: Existence of tiny models
We deal now with a model theoretic problem, the existence of tiny models; we
continue Laskowski, Pillay, and Rothmaler [LaPiRo]; our main result is in 3.17.
3.12 Context. Assume T is a complete rst order theory. Let [T[ be the number of
rst order formulas ( x), x = x
: < n), n < , up to equivalence modulo T.

Assume T is categorical in all cardinals > =: [T[ and call a model M of T tiny
if |M| < (= [T[). It is known that a T with a tiny model satises exactly one of
the following:
(a) T is totally transcendental, trivial (i.e. any regular type is trivial)
(b) T is not totally transcendental.
3.13 Question. For which < are there T, [T[ = (which is categorical in
+
and) with a tiny model of cardinality ?
3.14 Discussion. By [LaPiRo] we can deal with just the following two cases
(see [LaPiRo], 0.3,p.386 and 387
121
and 1.7,p.390).
Case A. x = x is a minimal formula and its prime model consists of individual
constants.
Case B. T is superstable not totally transcendental and is uni-dimensional, the
formula x = x is weakly minimal, regular types are trivial and its prime model
consists of individual constants.
They proved: ()[
0

+
in case A, =
0
], (see [LaPiRo,2.1,p.341]).
Actually more is true by continuing their argument.
3.15 Lemma. If , , T are as above, in Case A, then:
(i) <
,
(ii) we can nd
n
: n < ) such that:
0
= ,
n

n+1
,
=
n<
n
and ()
,
n
,
n+1
(hence in particular ()
,,
+), where
()
,,
there is a family of subsets of each of cardinality , with the
intersection of any two being nite, or equivalently functions from
to such that for any two such two distinct functions f
, f
we have
i < : f
(i) = f
(i) is nite.
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

Proof. By 1.2(2),(ii) (i), so let us prove (ii). Let M be a tiny model of T,
|M| = .
For n 0, let B
n
be the family of denable (with parameters) subsets of
n+1
M.
Clearly [T[
n<
[B
n
[, also = |M| [B
n
[, [B
n
[ [B
n+1
[. Also [B
0
[ = |M| as
M is minimal which means
0
= ; let
n
=: [B
n
[, so
n

n+1
; =
n<
n
and
it is enough to prove ()
,
n
,
n+1
when
n
<
n+1
. For each R B
n+1
we dene a
function f
R
from M to B
n
, f
R
(a) =
b
n
M :

b < a > R. So f
R
: R B
n+1
is a family of
n+1
functions from M to B
n
, hence it is enough to show:
dene R
1
R
2
a M
0
: f
R
1
(a) = f
R
2
(a) is co-nite
then
() is an equivalence relation on B
n+1
() each -equivalence class has cardinality
n
() if [R
1
R
2
], R
1
B
n+1
, R
2
B
n+1
then
a M : f
R
1
(a) = f
R
2
(a) is nite.
Now clause () is straight, for clause () just compute, for clause () remember
x = x is a minimal formula. Together, a set of representations for B
n+1
/ will
have cardinality
n+1
(as [B
n+1
[ =
n+1
>
n
= [B
n
[ by clauses (), ()) and
f
R
: R is a set of functions as required.
3.15
3.16 Lemma. Suppose ()
,,
, < . Then
(a) there is a group G of permutations of such that [G[ = and
f ,= g G < : f() = g() is nite
(b) there is a theory T as in 1.1, [T[ = , with a tiny model of cardinality of
Case A.
Proof. As (a) (b) is proved in [LaPiRo], p.392
2331
we concentrate on (a). Let
pr(, ) be a pairing function on i.e. pr is one-to-one from onto . So
let A
: < []
be such that ,=
0
> [A
[. Clearly
0

hence there is a list =
: < ) of distinct members of

. By renaming
we can have the family A
,n
: < , n < , such that (A
,n
[]
, [(, n) ,=
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

40 SAHARON SHELAH
(, m) [A
,n
A
,m
[ <
0
] and)
_
<
A
,n

_
<
A
,m
= for n ,= m, and
,= (n)(m)[n m < A
,n
A
,n
= ]; use . Let g
0
,n

be g
0
,n
()
= the th member of A
,n
and g
1
,n
() = pr(, g
0
,n
()) so also g
1
,n
is a function
from to .
We dene the set A = (
>
1, +1), clearly [A[ = ; , let x, y vary on
1, +1. Now for < we dene a permutation f
of A, by dening f
+1

( ) = f
( ), f
1
( ) for
n
1, +1 by induction on n
(so in the end, f
1
is the inverse of f
= f
+1
).
For n = 0, =<> and let for x 1, +1, f
x
(, <>) = (g
1
,0
(), < x >).
For n + 1, = < y >
n+1
1, +1 we let
() f
x
(, ) = (, ) when
x = y, f
y
(, ) = (, ) (by the previous stage)

() f
x
(, ) = (g
1
,n+1
(), < x >) when () does not apply.
Easily f
is a well-dened permutation of A.
Now f
: < generates a group G of permutations of A. We shall prove it

generates G freely, moreover:
if n < , t = ((), x()) : n) is such that () < , x() 1, 1, and
for no < n do we have () = ( + 1) & x() = x( + 1)
(i.e.
n
f
x()
()
is a non-trivial group term) then A
t
= a A : (
n
f
x()
()
)(a) =
a is nite.
As [A[ = , this clearly suces.
As this property of
n
f
x()
()
is preserved by conjugation without loss of generality
()
0
n () ,= ( +1) x() ,= x( +1) where n+1 is interpreted as zero.
For any a A
t
let
()
1
b
t
m
[a] = (
n
=m
f
x()
()
)(a) for m n + 1
(so b
t
n+1
[a] = a = b
t
0
[a] and for m = 0, . . . , n we have b
t
m
[a] = f
x(m)
(m)
(b
t
m+1
[a]))
()
2
b
t
m
[a] = (
t
m
[a],
t
m
[a])
Choose m
< large enough such that:

()
3
if m m
and 0
1
<
2
n and (
1
) ,= (
2
) then
A
(
1
),m
A
(
2
),m
= .
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

For a A
t
let m = m[a] n + 1 be such that lg(
t
m
[a]) is maximal and call the
length k = k[a]. As f
(, )) = , ) implies lg() lg() 1, lg() +1, clearly

()
4
lg(
t
m1
[a]) = lg(
t
m+1
[a]) = lg(
t
m
[a]) 1 (where m 1, m + 1 means
mod n + 1).
Clearly
()
5
(a) b
t
m
[a] = f
x(m)
(m)
(b
t
m+1
[a])
(b) b
t
m1
[a] = f
x(m1)
(m1)
(b
t
m
[a]) hence (as (f
x(m1)
(m1)
)
1
= f
x(m1)
(m1)
) we have
(b)
b
t
m
[a] = f
x(m1)
(m1)
(b
t
m1
[a]).
Looking at the denition of f
x(m1)
(m1)
(b
t
m1
[a]), as m = m[a] by ()
4
clause () in
the denition of f applies so
()
6
(a) f
x(m1)
(m1)
(b
t
m1
[a]) = (g
1
(m1),k[a]
(
t
m1
[a]), (
t
m1
[a])x(m1))).
Similarly looking at the denition f
x(m)
(m)
(b
t
m+1
[a]), by ()
4
clause () applies so
()
6
(b) f
x(m)
(m)
(b
t
m+1
[a]) = (g
1
(m),k[a]
(
t
m+1
[a]), (
t
m+1
[a])x(m))).
By ()
5
(b)
+ ()
6
(a) we have
()
7
(a) b
t
m
[a] = (g
1
(m1),k[a]
(
t
m1
[a]), (
t
m1
[a])x(m1))).
By ()
5
(a) + ()
6
(b) we have
()
7
(b) b
t
m
[a] = (g
1
(m),k[a]
(
t
m+1
[a]), (
t
m+1
[a])x(m)))).
We can conclude by ()
7
(a) + ()
7
(b) that
()
8
x(m) = x(m1) hence x(m) ,= x(m1).
So by ()
0
applied to m1 we get
()
9
(m) ,= (m1).
Clearly by ()
7
(a) + ()
7
(b) we have
()
10
g
1
(m),k[a]
(
t
m+1
[a]) = g
1
(m1),k[a]
(
t
m1
[a]).
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

42 SAHARON SHELAH
Now by the choice of the g
1
s (and the pairing function) and ()

10
()
11

t
m+1
[a] =
t
m1
[a] and
g
0
(m),k[a]
(
t
m+1
[a]) = g
0
(m1),k[a]
(
t
m1
[a]).
So by ()
11
and the choice of the g
0
s
()
12
g
0
(m),k[a]
(
t
m+1
[a]) = g
0
(m1),k[a]
(
t
m1
) A
(m),k[a]
A
(m1),k[a]
.
If k[a] > m
we get a contradiction (by ()

3
), so remembering m = m[a] necessarily
lg(
t
m[a]
[a]) m
+ 1, hence by the choice of m[a] we have
lg(
t
[a]) m
.
So
t
[a] : < n + 1) : a A
t
is nite, hence it suces to prove for each

n+1
1, 1 the niteness of
A
t,
= a A
t
:
t
[a] : < n + 1) = .
Let us x .
As for a A
t,
we have g(
t
m
[a]) m
for n + 1, it is enough to prove that

for each

k = k
: n) the following set is nite:

A
t, ,
k
=: a A
t,
: g(
t
[a]) = k
for < n + 1.
Let K(
k) = n + 1 : k
is k
1
, k
+1
(i.e. a local maximum).
For each m K(
k), the arguments in ()

3
()
12
apply, so by ()
11
, if a
A
t, ,
k
then the value g(
t
m
[a]) is determined and g
0
(m),k
m
(
t
m+1
[a]) A
(m),k
m

A
(m1),k
m
, but the latter is nite so we can x g
0
(m),k
m
(
t
m+1
[a]) =
m
but
g
1
(m),k
m
(
t
m+1
[a]) can be computed from = g
0
(m),k
m
(
t
m+1
[a]) and ((m), k
m
)
i.e. as pr(otp(A
(m),k
m
),
m
).
But by ()
7
(b) the latter is
t
m
[a] and as
t
m
[a] =
m
the value of b
t
m
[a] is uniquely
determined. Similarly by induction we can compute the other b
t
m
[a] for every m
,
in particular b
t
0
[a] = a, so we are done.
3.17
3.17 Conclusion. For a cardinal , the following are equivalent:
(a) there is a T as in 3.16(b) (i.e. T categorical in [T[
+
, [T[ > ), with a tiny
model M, |M| = as in Case A above
(b) ()
,,
+
(c) there is a group G of permutations of , [G[ =
+
such that for g G,
< : g() = is nite or is .
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

REFERENCES.
[AAC90] D.L. Alben, G.L. Alexanderson, and C. Reid (editors). More Mathe-
matical People. Harcourt Brace Jovanovich, 1990.
[Br] F.E. Browder (editor). Mathematical developments arising from
Hilberts Problems. Proc. of Symposium in Pure Math, 28:421, 1974.
[FW] Matthew Foreman and Hugh Woodin. The generalized continuum hy-
pothesis can fail everywhere. Annals Math., 133:136, 1991.
[GiSh 344] Moti Gitik and Saharon Shelah. On certain indestructibility of strong
cardinals and a question of Hajnal. Archive for Mathematical Logic,
28:3542, 1989.
[Gre] John Gregory. Higher Souslin trees and the generalized continuum
hypothesis. Journal of Symbolic Logic, 41(3):663671, 1976.
[HJSh 249] Andras Hajnal, Istvan Juhasz, and Saharon Shelah. Splitting strongly
almost disjoint families. Transactions of the American Mathematical
Society, 295:369387, 1986.
[HLSh 162] Bradd Hart, Claude Laamme, and Saharon Shelah. Models with sec-
ond order properties, V: A General principle. Annals of Pure and
Applied Logic, 64:169194, 1993. math.LO/9311211.
[LaPiRo] Michael C. Laskowski, Anand Pillay, and Philipp Rothmaler. Tiny
models of categorical theories. Archive for Mathematical Logic, 31:385
396, 1992.
[Sh:E12] Saharon Shelah. Analytical Guide and Corrections to [Sh:g].
math.LO/9906022.
[Sh 108] Saharon Shelah. On successors of singular cardinals. In Logic Collo-
quium 78 (Mons, 1978), volume 97 of Stud. Logic Foundations Math,
pages 357380. North-Holland, Amsterdam-New York, 1979.
[Sh 82] Saharon Shelah. Models with second order properties. III. Omitting
types for L(Q). Archiv fur Mathematische Logik und Grundlagen-
forschung, 21:111, 1981.
[Sh 351] Saharon Shelah. Reecting stationary sets and successors of singular
cardinals. Archive for Mathematical Logic, 31:2553, 1991.
[Sh 400a] Saharon Shelah. Cardinal arithmetic for skeptics. Ameri-
can Mathematical Society. Bulletin. New Series, 26:197210, 1992.
math.LO/9201251.
(
4
6
0
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
1
0
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
1
0
-
2
9

44 SAHARON SHELAH
[Sh 420] Saharon Shelah. Advances in Cardinal Arithmetic. In Finite and In-
nite Combinatorics in Sets and Logic, pages 355383. Kluwer Academic
Publishers, 1993. N.W. Sauer et al (eds.). 0708.1979.
[Sh 410] Saharon Shelah. More on Cardinal Arithmetic. Archive for Mathemat-
ical Logic, 32:399428, 1993. math.LO/0406550.
[Sh:g] Saharon Shelah. Cardinal Arithmetic, volume 29 of Oxford Logic
Guides. Oxford University Press, 1994.
[Sh 454a] Saharon Shelah. Cardinalities of topologies with small base. Annals of
Pure and Applied Logic, 68:95113, 1994. math.LO/9403219.
[Sh 430] Saharon Shelah. Further cardinal arithmetic. Israel Journal of Math-
ematics, 95:61114, 1996. math.LO/9610226.
[Sh 589] Saharon Shelah. Applications of PCF theory. Journal of Symbolic
Logic, 65:16241674, 2000.
[Sh 575] Saharon Shelah. Cellularity of free products of Boolean alge-
bras (or topologies). Fundamenta Mathematica, 166:153208, 2000.
math.LO/9508221.
[Sh 666] Saharon Shelah. On what I do not understand (and have some-
thing to say:) Part I. Fundamenta Mathematicae, 166:182, 2000.
math.LO/9906113.
[Sh 513] Saharon Shelah. PCF and innite free subsets in an algebra. Archive
for Mathematical Logic, 41:321359, 2002. math.LO/9807177.
[Sh 668] Saharon Shelah. Antihomogeneous Partitions of a Topological Space.
Scientiae Mathematicae Japonicae, 59, No. 2; (special issue:e9,
449501):203255, 2004. math.LO/9906025.

Saharon Shelah - The Generalized Continuum Hypothesis Revisited

Diunggah oleh

Informasi Dokumen

Deskripsi Asli:

Judul Asli

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

Saharon Shelah - The Generalized Continuum Hypothesis Revisited

Diunggah oleh

Hak Cipta:

Format Tersedia

(

singular, but there are limit =

) > . Continuing this, Easton proved

for regular cardinals is arbitrary (except for the laws

when < , and we can

and for every regular ,

> . Now what is a reasonable

? The reader may well stop and reect. As

, for every large enough regular <

by any strong limit cardinal).

for some n and

is the union of <

= a P : [a[ ) and clause

for every regular <

large enough, there is

and > open sets has at least

. Lastly in an appendix we prove there are no

) [a[ < , and easily the latter is by 1.1.

we can nd < and B

strong limit singular, (

, ) is as in 1.2, then by Fodors lemma for some =

, ) = is stationary: this can serve.

of an ultralter on the Boolean Algebra P(

is -like (i.e. of cardinality but every

is -like, note that for each there is at most

is, in general, not well-founded; still there is

(of course it depends

which is -like. Note: g

V be a set such that

hence without loss of generality g

for some i < i

we have g < sup

) and a < y then

() = (sup(j(Y )) ) + 1 < for d

hence at last we have proved ()

form a counterexample to the Theorem 1.1.

< j() such that j(),

form a counterexample to the theorem.

[d]] : M [= pcf(d) & < y [d b

such that: if < are in R

< and then

for any R Reg unbounded, for some R, > cf() and

and J is a -complete ideal on , then for some A I

for every > we have:

) is a sequence of pairwise disjoint sets, w

) are pairwise disjoint. Now

a unique [i] <

[i] < but then for each m we have u

[a]) = (when dened).

, then for some () < and nite

[ < then we can have one () for all a

by induction on . For = 0 clearly for

< such that:

be the ideal with domain

: (j, ) A < , we have = max pcf

, then i < : j < : <

is a proper ideal and

is -directed by basic pcf

, then prove by induction on n

: < ) such that c =

) < . Let us check all the clauses of the desired conclusion.

) < , and by [Sh 430, 6.7A,](h) for some nite c pcf(b