Anda di halaman 1dari 25

7

9
9


r
e
v
i
s
i
o
n
:
2
0
0
5
-
0
2
-
2
8







m
o
d
i
f
i
e
d
:
2
0
0
5
-
0
2
-
2
8


COFINALITY OF THE NONSTATIONARY IDEAL
PIERRE MATET, ANDRZEJ ROS LANOWSKI, AND SAHARON SHELAH
Abstract. We show that the reduced conality of the nonstationary ideal
NS on a regular uncountable cardinal may be less than its conality, where
the reduced conality of NS is the least cardinality of any family F of nonsta-
tionary subsets of such that every nonstationary subset of can be covered
by less than many members of F. For this we investigate connections of
the various conalities of NS with other cardinal characteristics of

and
we also give a property of forcing notions (called manageability) which is pre-
served in <support iterations and which implies that the forcing notion
preserves non-meagerness of subsets of

(and does not collapse cardinals
nor changes conalities).
0. Introduction
Let be a regular uncountable cardinal. For C and , we say that is a
limit point of C if

(C ) = > 0. C is closed unbounded if C is a conal subset


of containing all its limit points less than . A set A is nonstationary if A
is disjoint from some closed unbounded subset C of . The nonstationary subsets
of form an ideal on denoted by ^S

. The conality of this ideal, cof(^S

),
is the least cardinality of a family T of nonstationary subsets of such that every
nonstationary subset of is contained in a member of T. The reduced conality
of ^S

, cof(^S

), is the least cardinality of a family T ^S

such that every


nonstationary subset of can be covered by less than many members of T. This
paper addresses the question whether cof(^S

) = cof(^S

). Note that

+
cof(^S

) cof(^S

) 2

,
so under GCH we have cof(^S

) = cof(^S

).
Let

2 be endowed with the box product topology, 2 itself considered discrete.
We say that a set W

2 is meager if there is a sequence U

: < ) of dense
open subsets of

2 such that W

<
U

= . The covering number for the category


of the space

2, denoted cov(M
,
), is the least cardinality of any collection A of
meager subsets of

2 such that

A =

2. It is not hard to verify that
cov(M
,
) cof(^S

)
_
cof(^S

)
_
<
.
1991 Mathematics Subject Classication. 03E05. 03E35, 03E55.
Key words and phrases. Nonstationary ideal, conality.
The second author thanks University Committee On Research of the University of Nebraska
at Omaha for partial support. He also thanks his wife, Ma lgorzata JankowiakRos lanowska for
supporting him when he was preparing the nal version of this paper.
The research of the third author was partially supported by the Israel Science Foundation.
Publication 799.
1
7
9
9


r
e
v
i
s
i
o
n
:
2
0
0
5
-
0
2
-
2
8







m
o
d
i
f
i
e
d
:
2
0
0
5
-
0
2
-
2
8


2 PIERRE MATET, ANDRZEJ ROS LANOWSKI, AND SAHARON SHELAH
It follows that if cof(^S

) < cov(M
,
) and the Singular Cardinals Hypothesis
holds true, then cf(cof(^S

)) < and cof(^S

) = (cof(^S

))
+
. We prove:
Theorem 0.1. Assume GCH. Then there is a complete,
+
cc forcing notion
P such that

P
cof(^S

) =
+
and cof(^S

) =
+(+1)
.
What about the consistency of cof(^S

) is regular and cof(^S

) < cof(^S

)?
We establish:
Theorem 0.2. It is consistent, relative to the existence of a cardinal such that
o() =
++
, that cof(^S
1
) =
+1
and cov(M
1,1
) =
+2
.
The structure of the paper is as follows. In Section 1, for each innite cardinal
we introduce the <conality cof
<
(^S

) and the <dominating number


d
<

, and we show that these two numbers are equal. Section 2 is concerned with
a variant of d
<

denoted by d
cl,<

(where cl stands for club). We establish that


d
<

= d
cl,<

if > .
^S

is the smallest normal ideal on . Section 3 deals with ^S

,
, the small-
est normal ideal on T

(). We compute cof


<
(^S

,
) and give examples of
situations when cof
<
(^S

,
) < cof(^S

,
).
In the following section we present some basic facts regarding the ideal of
meager subsets of

2 and its covering number.
The nal three sections of the paper present the consistency results mentioned in
Theorems 0.1, 0.2 above. First, in Section 5 we introduce manageability, a property
of <complete
+
cc forcing notions which implies preservation of non-meagerness
of subsets of

and which can be iterated. Next, in Section 6, we dene one-step
forcing and verify that it has all required properties. The nal section gives the
applications obtained by iterating this forcing notion.
Notation 0.3. Our notation is rather standard and compatible with that of classical
textbooks (like Jech [J]). In forcing we keep the older (Cohens) convention that a
stronger condition is the larger one. Some of our conventions are listed below.
(1) For a forcing notion P,
P
stands for the canonical Pname for the generic
lter in P. With this one exception, all Pnames for objects in the extension
via P will be denoted with a dot above (e.g. ,

X). The weakest element
of P will be denoted by
P
(and we will always assume that there is one,
and that there is no other condition equivalent to it). In iterations, if

Q = P

,

Q

: <

) and p lim(

Q), then we keep convention that


p() =

Q
for

Dom(p).
(2) Ordinal numbers will be denoted by , , , , , , and also by i, j (with
possible sub- and superscripts).
Innite cardinal numbers will be called , , , , (with possible sub- and
superscripts); is our xed regular uncountable cardinal, will
denote a xed cardinal > (in Section 3).
(3) By we will denote a suciently large regular cardinal and by H() the
family of all sets hereditarily of size less than . Moreover, we x a well
ordering <

of H().
(4) A bar above a letter denotes that the object considered is a sequence;
usually

X will be X
i
: i < ), where denotes the length of

X. For a set
7
9
9


r
e
v
i
s
i
o
n
:
2
0
0
5
-
0
2
-
2
8







m
o
d
i
f
i
e
d
:
2
0
0
5
-
0
2
-
2
8


COFINALITY OF THE NONSTATIONARY IDEAL 3
A and a cardinal , the set of all sequences of members of A of length
(length < , respectively), will be denoted by

A (
<
A, respectively).
1. cof
<
(^S

)
Denition 1.1. (1) For a set A and a cardinal , T

(A) = a A : [a[ < .


(2) Given two innite cardinals , u(, ) is the least cardinality of a
collection A T

() such that T

() =

aA
T(a).
Denition 1.2. Let S be an innite set and be an ideal on S (containing all
singletons).
(1) cof() is the least cardinality of any A such that for every A ,
there is B A with A B.
(2) add() is the least cardinality of any A such that

A / .
(3) For an innite cardinal add(), cof
<
() is the least cardinality of a
family A such that for every A , there is } T

(A) such that


A

}.
(4) We let cof() = cof
<add(J)
().
The following proposition collects some trivialities.
Proposition 1.3. Let S be an innite set and be an ideal on S that contains
all singletons. Then:
(i) cof
<
() = cof().
(ii) If , are two innite cardinals with add(),
then cof
<
() cof
<
().
(iii) cof() u(, cof
<
()) for every innite cardinal add().
(iv) add() cof().
The following is well-known (see, e.g., Matet, Pean and Shelah [MPSh 713]):
Lemma 1.4. Let be a regular innite cardinal. Then u(,
+n
) =
+n
for every
n < .
Proposition 1.5. Let S be an innite set and be an ideal on S such that
_
add()
_
+
cof(). Then
_
add()
_
+
cof().
Proof. Use Lemma 1.4.
With these preliminaries out of the way, we can concentrate on ideals on . If
there is a family of size
+
of pairwise almost disjoint conal subsets of , then
there is a complete ideal on such that cof() < cof() (see Matet and
Pawlikowski [MPxx]).
Proposition 1.6. Suppose is a normal ideal on and is a limit cardinal.
Then cof() = cof
<
() for some innite cardinal < .
Proof. Assume that the conclusion fails. Fix A such that [A[ = cof() and
=
_
T(

X) : X T

(A).
Set } = A : A A & . Note that [}[ = cof(). For each innite
cardinal < we may select a set B

so that B

Y for any Y T

(}).
Now let B be the set of all < such that B

for some innite cardinal < .


7
9
9


r
e
v
i
s
i
o
n
:
2
0
0
5
-
0
2
-
2
8







m
o
d
i
f
i
e
d
:
2
0
0
5
-
0
2
-
2
8


4 PIERRE MATET, ANDRZEJ ROS LANOWSKI, AND SAHARON SHELAH
Since B (by normality of ), there must be X T

(A) such that B

X.
Let be any innite cardinal such that [X[ < < . Then B

AX
(A( +1)),
which is a contradiction.
Arguing as in Proposition 1.6, we get:
Proposition 1.7. Suppose is a complete ideal on and is an uncountable
limit cardinal < . Then there is an innite cardinal < such that cof
<
() =
cof
<
(). Moreover, the least such is either , or a successor cardinal.
The remainder of this section is concerned with cof
<
(^S

). Let us recall the


denition of the bounding number b

:
Denition 1.8. The bounding number b

is the least cardinality of any T


with the property that for every g



, there is f T such that
[ < : g() f()[ = .
The following is proved in Matet and Pawlikowski [MPxx]:
Proposition 1.9. (i) cof(^S

) b

.
(ii) If cof(^S

) = b

, then cof(^S

) = cof(^S

).
Proposition 1.10. Let be an innite cardinal . Then
either cf(cof
<
(^S

)) < , or cf(cof
<
(^S

)) b

.
Proof. Suppose to the contrary that cf(cof
<
(^S

)) = < b

. For <
select A

^S

so that
(i) [A

[ < cof
<
(^S

),
(ii) A

for < ,
(iii) ^S

T(

X) : X T

<
A

).
For < , set }

= A : A A

& and pick B

^S

so that
B



Y for any Y T

(}

). By a result of Balcar and Simon (see [BS89,


Theorem 5.25]), there is B ^S

such that [B

B[ < for every < . Select


X T

<
A

) so that B

X. There is < such that X A

. Then
B

AX
(A ) for some , which is a contradiction.
Denition 1.11. Let . A family T

is called
a dominating family if
(h

)(f T)(j < )(h(j) < f(j)),
a <dominating family if
(h

)(F T

(T))(j < )(h(j) < supf(j) : f F).


We dene dominating numbers d

, d
<

by:
d

= min[T[ : T

is a dominating family ,
d
<

= min[T[ : T

is a <dominating family .
We let

d

= d
<

and for an innite cardinal < we put d

= d
<
+

.
Note that d
<

= d

. Landver [La90] established that cof(^S

) = d

. His result
can be generalized as follows:
7
9
9


r
e
v
i
s
i
o
n
:
2
0
0
5
-
0
2
-
2
8







m
o
d
i
f
i
e
d
:
2
0
0
5
-
0
2
-
2
8


COFINALITY OF THE NONSTATIONARY IDEAL 5
Theorem 1.12. Let be an innite cardinal . Then cof
<
(^S

) = d
<

.
Proof. Set = cof
<
(^S

). First we will argue that d


<

. Select a family
( of size of closed unbounded subsets of so that for every closed unbounded
subset D of , there is X T

(() with

X D. For U T

(() dene
f
U


by f
U
() = min
_
U ( + 1)
_
. Note that f
V
() f
U
() whenever
V T(U) . Now given g

, let D be the set of all limit ordinals < such
that g() < for every < . Pick X T

(() so that

X D. Dene
h

by
h() = sup
_
f
U
() : U T

(X)
_
.
We are going to show that g < h. Let < and C X. First, suppose that there
is W T

(X) such that h() = f


W
(). Then h() = f
W{C}
() and hence
h() C. Next suppose that f
U
() < h() for all U T

(X) . Then h()


is a limit ordinal. Set = cf(h()) and pick an increasing sequence

: < )
conal in h(). For < , select T

(X) with

< f
T

(), and set

= f
T

{C}
(). Note that

C. Obviously, the sequence

: < ) is conal
in h(), and consequently h() C. Thus for each < , h() belongs to

X
and therefore to D. Since clearly h() > , it follows that h() > g().
It remains to show that d
<

. Let F be the set of all strictly increasing


functions from to . Select T F so that
(a) [T[ = d
<

, and
(b) given g

, there is F
g
T

(T) such that


( < )(g() < supf() : f F
g
).
For f F, let C
f
be the set of all limit ordinals < such that f() < for every
< . Easily
^S

= A : (g F)(A C
g
= )
(see, e.g., [MPxx]) and (as

fFg
C
f
C
g
for every g F) it follows that [T[.
It follows from Propositions 1.6 and 1.7 and Theorem 1.12 that to determine the
value of cof
<
(^S

) for every innite cardinal , it suces to compute d

and
d

for every innite cardinal < .


2. d
cl,<

It is straightforward to check that d


<

is the least cardinality of a family T


such that
(g

)(F T

(T))([
_
: g() supf() : f F
_
[ < ).
In this section we discuss the variant that arises if we replace has cardinality <
by is nonstationary.
Denition 2.1. (1) d
cl

is the least cardinality of a family T



with the
property that for every g

, there is f T such that
: g() f() ^S

.
(2) For an innite cardinal , d
cl,<

is the least cardinality of a family


T

with the property that for every g

, there is F T

(T) such
that
_
: g() supf() : f F
_
^S

.
7
9
9


r
e
v
i
s
i
o
n
:
2
0
0
5
-
0
2
-
2
8







m
o
d
i
f
i
e
d
:
2
0
0
5
-
0
2
-
2
8


6 PIERRE MATET, ANDRZEJ ROS LANOWSKI, AND SAHARON SHELAH
Note that d
cl,<

= d
cl

. It is simple to check that cf(d


cl

) b

.
Theorem 2.2. For every uncountable cardinal ,
d
cl,<

= d
<

.
Theorem 2.2 easily follows from the next two lemmas.
Lemma 2.3. Let be an uncountable limit cardinal . Then d
cl,<

= d
cl,<

for
some innite cardinal < .
Proof. The proof is similar to that of Proposition 1.7. Suppose that the conclusion
fails. Fix a family T

such that [T[ = d
cl,<

and
(g

)(F T

(T))(
_
: g() supf() : f F
_
^S

).
For each innite cardinal < we may select g



so that for every F T

(T)
we have
_
: g

() supf() : f F
_
/ ^S

.
Dene g

so that g() g

() for every innite cardinal < such that


< . Now pick F T

(T) such that


_
: g() supf() : f F
_
^S

.
Let be any innite cardinal with [F[ < < . Obviously, F T

(T) and
_
: g

() supf() : f F
_
^S

,
a contradiction.
To establish the following lemma, we adapt the proof of Theorem 5 in Cummings
and Shelah [CuSh 541].
Lemma 2.4. Let be a regular uncountable cardinal . Then d
<

= d
cl,<

.
Proof. Select a family T

such that
(a) every member of T is increasing,
(b) [T[ = d
cl,<

, and
(c) for each g

, there is F T

(T) such that


_
: g() supf() : f F
_
^S

.
We claim that the family
T

def
=
_
f

:
_
, <
__
g T
__
f & f[, ) = g[, )
__
is <dominating. So let g

. Stipulate that g
1
= g. By induction on n
choose a closed unbounded subset C
n
of , g
n
, h
n


and F
n
T

(T) so that
(i) C
n+1
C
n
,
(ii) g
n1
() < supf() : f F
n
for all C
n
,
(iii) h
n
() = min(C
n
( + 1)),
(iv) g
n
() = sup
_
Rng(g
n1
(h
n
() + 1))
_
.
Note that, by (iii) and (iv), g() g
0
() g
1
() . . . for all . Set
F =

n
F
n
and = supmin(C
n
) : n . We are going to show that g() <
supf() : f F whenever < < . To this end suppose that < < . By
(i), there are m and such that = sup( C
n
) whenever m n < .
7
9
9


r
e
v
i
s
i
o
n
:
2
0
0
5
-
0
2
-
2
8







m
o
d
i
f
i
e
d
:
2
0
0
5
-
0
2
-
2
8


COFINALITY OF THE NONSTATIONARY IDEAL 7
By (iii), h
m
() and so (by (iv)) g() g
m1
() g
m
(). Since > we also
have C
m+1
,= . Hence C
m+1
and consequently, by (ii),
g() g
m
() < supf(): f F
m+1
supf(): f F
m+1
supf(): f F.

Theorem 2.2 implies that d

d
cl

. We mention that it was shown in


Cummings and Shelah [CuSh 541] that d
cl

= d

if >

.
3. cof
<
(^S

,
)
Throughout this section denotes a xed cardinal > . Our object of study will
be the ideal ^S

,
, a T

() version of ^S

.
Denition 3.1. For a regular uncountable cardinal and a cardinal ,
,
is
the set of all A T

() such that for some a T

() we have b A : a b = .
It is straightforward to check that
,
is a complete ideal on T

().
Denition 3.2. (1) An ideal of T

() is normal if given A
+
and
f : A such that f(a) a for all a A, there is B
+
T(A)
such that f is constant on B.
(2) The smallest normal ideal on T

() containing
,
is denoted by ^S

,
.
(3) For f

(T

()) we let
C
f
def
= a T

() : a ,= and
_
a
f() a.
The following lemma is due to Abe.
Lemma 3.3 (Abe [Ab97]). Let A T

(). Then
A ^S

,
if and only if (f

(T

()))(A C
f
= ).
Our purpose in this section is to compute the value of cof
<
(^S

,
). We will
need an analogue of d
<

dened in 3.4(1) below.


Denition 3.4. Let be an innite cardinal.
(1) d
,<
,
is the least cardinality of a family A of functions from to T

()
with the property that
_
g

(T

())
__
X T

(A)
__

__
g()
_
fX
f()
_
.
(2) cov(,
+
,
+
, ) is the least cardinality of a family A T

+() such that


_
B T

+()
__
X T

(A)
__
B
_
X
_
.
Theorem 3.5. Let be an innite cardinal . Then
cof
<
(^S

,
) = d
,<
,
= maxd
<

, cov(,
+
,
+
, ).
Theorem 3.5 is an immediate consequence of Lemmas 3.63.9 below.
Lemma 3.6. Let be an innite cardinal . Then
cov(,
+
,
+
, ) cof
<
(^S

,
).
7
9
9


r
e
v
i
s
i
o
n
:
2
0
0
5
-
0
2
-
2
8







m
o
d
i
f
i
e
d
:
2
0
0
5
-
0
2
-
2
8


8 PIERRE MATET, ANDRZEJ ROS LANOWSKI, AND SAHARON SHELAH
Proof. By 3.3 we may pick a family A

(T

()) with the property that [A[ =


cof
<
(^S

,
) and for every function g : T

() there is X T

(A) such that

fX
C
f
C
g
. For f A, let B
f
=

<
f() T

+().
Suppose now that B T

+(). Pick a function g : T

() such that
B

<
g(). There is X T

(A) such that

fX
C
f
C
g
. We are going to show
that B

fX
B
f
. To this end suppose < and let us argue that g()

fX
B
f
.
For n < let a
n
T

fX
B
f
) be dened by
a
0
= , and a
n+1
= a
n

_
fX
_
an
f(),
and let a =

n<
a
n
. Then a

fX
C
f
C
g
and consequently g() a

fX
B
f
.
Lemma 3.7. Let be an innite cardinal . Then d
<

cof
<
(^S

,
).
Proof. By Theorem 1.12, it suces to establish that cof
<
(^S

) cof
<
(^S

,
).
Let a family A

(T

()) be such that [A[ = cof


<
(^S

,
) and
_
B ^S

,
__
X T

(A)
__
B

fX
C
f
=
_
.
For f A, let Z
f
be the set of all limit ordinals < such that
( < )(f() ).
Plainly, Z
f
is a closed unbounded subset of . Now given a closed unbounded
subset T of , set B
T
= a T

() : a / T. A simple argument (see, e.g.,


[MPSh 713]) shows that B
T
^S

,
. Hence there is X
T
T

(A) such that


B
T


fXT
C
f
= . We will show that

fXT
Z
f
T. Thus let

fXT
Z
f
. Setting
a =

fXT

<
f(), it is easy to see that a = and a

fXT
C
f
. It follows
that = a T.
Lemma 3.8. Let be an innite cardinal . Then cof
<
(^S

,
) d
,<
,
.
Proof. The inequality easily follows from the following observation.
Suppose h : T

() and X T

(T

())
_
are such that
_
<
__
h()
_
fX
f()
_
.
Then

fX
C
f
C
h
.
Lemma 3.9. Let be an innite cardinal . Then
d
,<
,
maxd
<

, cov(,
+
,
+
, ).
Proof. Fix T

so that [T[ = d
<

and
_
h

__
F T

(T)
__
<
__
h() supf() : f F
_
.
7
9
9


r
e
v
i
s
i
o
n
:
2
0
0
5
-
0
2
-
2
8







m
o
d
i
f
i
e
d
:
2
0
0
5
-
0
2
-
2
8


COFINALITY OF THE NONSTATIONARY IDEAL 9
Also, x A T

+() such that [A[ = cov(,


+
,
+
, ) and
T

+() =
_
T(

X) : X T

(A).
For each a A, select a mapping
a
:
onto
a. Now for f T and a A dene
g
f,a
: T

() by
g
f,a
() =
a
() : < f().
Suppose now that g : T

(). By the choice of A, there is X T

(A) such
that

<
g()

X. Choose h

such that
_
<
__
g()
_
aX

a
() : < h()
_
.
Next pick F T

(T) such that ( < )(h() supf() : f F). Then


_
<
__
g()
_
fF
_
aX
g
f,a
()
_
.

Another formula worth noting is:


cof
<
(^S

,
) = maxcof
<
(^S

), cof
<
(

+
,
).
This identity follows from Theorems 1.12 and 3.5 and the next proposition.
Proposition 3.10. Let be an innite cardinal . Then cov(,
+
,
+
, ) =
cof
<
(

+
,
).
Proof. The result easily follows from the following observation.
Suppose that A T

+() and X T

(A) . Then

aX
c T

+() : a c = c T

+() :

X c,
and therefore for each b T

+()
b

X if and only if

aX
c T

+() : a c c T

+() : b c.

We next consider special cases when cof


<
(^S

,
) < cof(^S

,
).
Lemma 3.11. Let be an innite cardinal . Then cov(,
+
,
+
, ) .
Proof. It is shown in Matet, Pean and Shelah [MPSh 813] that cof(

+
,
) .
Now observe that (by Proposition 3.10) cov(,
+
,
+
, ) cof(

+
,
).
Lemma 3.12. Suppose is singular and is a cardinal such that cf() < .
Then cov(,
+
,
+
, ) supu(
+
, ) : < < .
Proof. Let

: < cf()) be an increasing sequence of cardinals conal in .


Then, for every a T

+(), a =

<cf()
a

. The desired inequality follows.


Proposition 3.13. Let be an uncountable cardinal . Then
cof
<
_
^S

,
+
_
= maxd
<

,
+
.
7
9
9


r
e
v
i
s
i
o
n
:
2
0
0
5
-
0
2
-
2
8







m
o
d
i
f
i
e
d
:
2
0
0
5
-
0
2
-
2
8


10 PIERRE MATET, ANDRZEJ ROS LANOWSKI, AND SAHARON SHELAH
Proof. By Lemmas 1.4, 3.11 and 3.12 we have
cov(
+
,
+
,
+
, ) =
+
,
so the result follows from Theorem 3.5.
Thus, if d
<1


+
, then
cof
<1
(^S

,
+) < cof(^S

,
+).
Lemma 3.14. Assume the Singular Cardinals Hypothesis. If 2

, then
u(
+
, ) =
_

+
if cf() ,
otherwise.
Proof. Plainly, u(
+
, )

. It follows immediately that u(


+
, ) = if
cf() > . For the other case, use the wellknown fact (see, e.g., [MPSh 713]) that
cf(u(
+
, ))
+
.
Proposition 3.15. Assume the Singular Cardinals Hypothesis. If 2

and

0
, then
cof
<
(^S

,
) =
_

+
if cf() ,
otherwise.
Proof. By Lemma 3.11, cov(,
+
,
+
, ) d

d
<

, so by Theorem 3.5
cof
<
(^S

,
) = cov(,
+
,
+
, ).
Case: cf() > .
By Lemmas 3.11 and 3.14 we have cov(,
+
,
+
, ) u(
+
, ) , and hence
cov(,
+
,
+
, ) = .
Case: cf() .
By Lemma 3.14 we know that
cov(,
+
,
+
, ) u(
+
, )
+
and
+
u(
+
, )
_
cov(,
+
,
+
, )
_
<
.
Since
<
= , it follows that cov(,
+
,
+
, ) =
+
.
Case: cf() < .
By Lemmas 3.11, 3.12 and 3.14 we have
cov(,
+
,
+
, ) supu(
+
, ) : < < ,
and consequently cov(,
+
,
+
, ) = .
Thus, if the Singular Cardinals Hypothesis holds, and 2

, then
cof
<
(^S

,
) < cof(^S

,
) if and only if cf() < .
4. cov(M
,
)
Let us recall some basic facts and denitions related to the combinatorics of the
meager ideal M
,
on

2.
Denition 4.1. (1) The Baire number n(X) of a topological space X (also
called the Novak number of X) is the least number of nowhere dense subsets
of X needed to cover X.
7
9
9


r
e
v
i
s
i
o
n
:
2
0
0
5
-
0
2
-
2
8







m
o
d
i
f
i
e
d
:
2
0
0
5
-
0
2
-
2
8


COFINALITY OF THE NONSTATIONARY IDEAL 11
(2) For a topological space X and a cardinal , the (<)complete ideal of sub-
sets of X generated by nowhere dense subsets of X is denoted by M
<
(X);
M
<
+(X) will be also denoted by M

(X). The ideal M

(X) is the ideal


of meager subsets of X.
(3) The space

(respectively

2) is endowed with the topology obtained by
taking as basic open sets and O
s
for s
<
(respectively s
<
2),
where O
s
= f

: s f (respectively O
s
= f

2 : s f).
(4) The ideals of meager subsets of

,

2 are denoted by M

,
and M
,
,
respectively.
Remark 4.2. (1) Clearly, for a topological space X, n(X) is the least number
of open dense subsets of X with empty intersection. If n(X), then
M
<
(X) is a proper ideal (i.e., X / M
<
(X)).
(2) Following the tradition of the set theory of the reals, we may consider the
covering number cov(M
<
(X)) of the ideal M
<
(X):
cov(M
<
(X)) = min[/[ : / M
<
(X) &
_
/ = X.
By the denition, n(X) = cov(M
<0
(X)). But also for every < n(X) we
have cov(M
<
(X)) = n(X); also cov(M
<n(X)
(X)) = cf(n(X)).
(3) Plainly, n(

) > and n(

2) > (remember, is assumed to be regular).


Lemma 4.3. Suppose that X is a topological space, < n(X), and Y

are open
subsets of X (for < ). Assume also that Y =

<
Y

is dense in X. Then, if Y
is equipped with the subspace topology, n(Y ) = n(X).
Proof. Let U

(for < n(X)) be open dense subsets of X such that

<n(X)
U

= .
Then U

Y are open dense subsets of Y (remember Y is dense) and

<n(X)
(U


Y ) = . This shows that n(Y ) n(X).
Now, let V

Y (for < n(Y )) be open dense subsets of Y such that

<n(Y )
V

=
. Take open subsets U

of X such that V

= U

Y clearly U

s are dense in
X (as Y is so). Then =

<n(Y )
(U

Y ) =

<n(Y )
U

<
Y

, and hence
n(X) n(Y ) + and therefore n(X) n(Y ).
Proposition 4.4. cov(M

,
) = n(

) = n(

2) = cov(M
,
).
Proof. For s
<
2 and < let F(s, )
<
2 be such that lh(F(s, )) =
lh(s) + + 1 and
F(s, )lh(s) = s, F(s, )[lh(s), lh(s) +) 1, and F(s, )(lh(s) +) = 0.
Now, let :
<

<
2 be such that
()) = ), (s

)) = F((s), ) for s
<
, and
if s

: < )
<
is increasing, < , s =

<
s

,
then (s) =

<
(s

).
7
9
9


r
e
v
i
s
i
o
n
:
2
0
0
5
-
0
2
-
2
8







m
o
d
i
f
i
e
d
:
2
0
0
5
-
0
2
-
2
8


12 PIERRE MATET, ANDRZEJ ROS LANOWSKI, AND SAHARON SHELAH
Then induces a one-to-one mapping

:



2 :

<
(). The range
of

is
Rng(

) =

2 : ( < )( < )( < & () = 0).
Plainly, Rng(

) is the intersection of many open dense subsets of



2. Moreover,

is a homeomorphism from

onto Rng(

). Therefore, using Lemma 4.3, we


get n(

) = n(Rng(

)) = n(

2). The rest should be clear (remember Remark


4.2(2,3)).
Proposition 4.5. cov(M
,
) d

.
Denition 4.6. C
,
is the forcing notion for adding Cohen functions in

with
<support. Thus a condition in C
,
is a function q such that
Dom(q) , Rng(q) and [q[ < .
The order of C
,
is the inclusion.
Proposition 4.7. Assume 2
<
= < . Then
C,
cov(M
,
) .
5. Manageable forcing notions
In this section we introduce a property of forcing notions which is crucial for
the consistency results presented later: (, , )manageability. This property has
three ingredients: an iterable variant of
+
cc (see Denition 5.1), completeness
and a special property implying preservation of non-meagerness of subsets of

(see Proposition 5.9). Since later we will work with <support iterations, we also
prove a suitable preservation theorem (see Theorem 5.11).
Throughout the section we will assume that our xed (uncountable)
regular cardinal satises 2
<
= (so also
<
= ).
Denition 5.1 (See Shelah [Sh 288, Denition 1.1] and [Sh 546, Denition 7]).
Let P be a forcing notion, and < be a limit ordinal.
(1) We dene a game
cc
,
(P) of two players, Player I and Player II. A play
lasts steps, and at each stage < of the play q

, p

are chosen so
that:
q
0
=
P
: i <
+
),
0
:
+

+
: i 0;
If > 0, then Player I picks q

such that
(i) q

= q

i
: i <
+
) P satises
( < )(i <
+
)(p

i
q

i
),
(ii)

:
+

+
is regressive, i.e., (i <
+
)(

(i) < 1 +i);


Player II answers choosing a sequence p

= p

i
: i <
+
) P such
that (i <
+
)(q

i
p

i
).
If at some stage of the game Player I does not have any legal move, then he
loses. If the game lasted steps, Player I wins a play q

, p

: < )
if there is a club C of
+
such that for each distinct members i, j of C
satisfying cf(i) = cf(j) = and ( < )(

(i) =

(j)), the set


p

i
: < p

j
: <
has an upper bound in P.
(2) The forcing notion P satises condition ()

if Player I has a winning strat-


egy in the game
cc
,
(P).
7
9
9


r
e
v
i
s
i
o
n
:
2
0
0
5
-
0
2
-
2
8







m
o
d
i
f
i
e
d
:
2
0
0
5
-
0
2
-
2
8


COFINALITY OF THE NONSTATIONARY IDEAL 13
Remark 5.2. Condition ()

is a strong version of
+
cc (easily, if < is limit,

= , and P satises ()

, then P satises
+
cc). This condition was used
in a number of papers, e.g., to obtain a series of consistency results on partition
relations; see Shelah and Stanley [ShSt 154], [ShSt 154a], Shelah [Sh 80], [Sh 288],
[Sh 546]. Its primary use comes from the fact that it is preserved in <support
iterations.
Proposition 5.3 (See Shelah [Sh 288, Iteration Lemma 1.3] and [Sh 546, Theorem
35]). Let < be a limit ordinal, =
<
. Suppose that

Q = P

,

Q

: < ) is a
<support iteration such that for each <

satises ()

.
Then P

satises ()

.
Denition 5.4. A forcing notion P is <complete if every
P
increasing chain
of length less than has an upper bound in P. It is <lubcomplete if every

P
increasing chain of length less than has a least upper bound in P.
Denition 5.5. Let and be cardinals such that < and
<
= . Let P be
a <
+
lubcomplete forcing notion.
(1) A model N (H(), , <

) is (P, , )relevant if P, N, N,
[N[ = and
<
N N.
(2) For a (P, , )relevant model N we dene a game
m
(N, , P) of two play-
ers, He and She, as follows. A play lasts moves, and in the i
th
move
conditions p
i
, q
i
P are chosen so that:
q
i
N P, q
i
p
i
,
(j < i)(q
j
q
i
& p
j
p
i
),
She chooses p
i
, q
i
if i is odd, He picks p
i
, q
i
if i is even.
She wins the play q
i
, p
i
: i < ) whenever
if p

is a least upper bound of p


i
: i < ), and q

is a least upper
bound of q
i
: i < ),
then (q N P)(q

q q, p

are compatible ).
(3) The forcing notion P is weakly (, , )manageable if (it is <
+
lubcomplete
and) there is an x H() (called a witness) such that for every (P, , )
relevant model N H() with x N, She has a winning strategy in the
game
m
(N, , P).
(4) The forcing notion P is (, , )manageable if it is <complete, weakly
(, , )manageable, and satises the condition ()

.
Remark 5.6. Suppose that P is <
+
lubcomplete and N is (P, , )relevant. Then
both players have always legal moves in the game
m
(N, , P). Moreover, if q
i
, p
i
:
i < ) is a (legal) play of
m
(N, , P), then there are least upper bounds q

N P
of q
i
: i < ), and p

P of p
i
: i < ) (and q

).
Denition 5.7. Let N be a (P, , )relevant model, and let q N P, p P be
such that q p. We say that a pair (q

, p

) is an Ncover for (q, p), if


q q

N P, p p

P, q

, and
every condition q

N P stronger than q

is compatible with p

.
Lemma 5.8. Suppose that P is a <
+
lubcomplete forcing notion, N is a (P, , )
relevant model, and She has a winning strategy in the game
m
(N, , P). Then:
7
9
9


r
e
v
i
s
i
o
n
:
2
0
0
5
-
0
2
-
2
8







m
o
d
i
f
i
e
d
:
2
0
0
5
-
0
2
-
2
8


14 PIERRE MATET, ANDRZEJ ROS LANOWSKI, AND SAHARON SHELAH
(1) For all conditions q NP and p P such that q p, there is an Ncover
(q

, p

) for (q, p).


(2) N P < P.
Proof. 1) Consider a play q
i
, p
i
: i < ) of
m
(N, , P) in which He starts with
q
0
= q, p
0
= p, and then he always plays the <

rst legal moves, and She uses


her winning strategy. Let q

N P, p

P be least upper bounds of q


i
: i < ),
p
i
: i < ), respectively. Plainly, as She won the play, the pair (q

, p

) is an
Ncover for (q, p).
2) Suppose that / N P is a maximal antichain in N P, but p P is
incompatible with all members of /. Let (q

, p

) be an Ncover for (
P
, p). The
condition q

is compatible with some q /, so let q


+
NP be such that q
+
q

,
q
+
q /. By the choice of (q

, p

) we know that the conditions q


+
and p

are
compatible, and hence q and p are compatible, a contradiction.
The rest follows from the elementarity of N.
Proposition 5.9. Assume < =
<
< . Suppose that a set Y

cannot be covered by the union of less than nowhere dense subsets of



, and P
is a weakly (, , )manageable forcing notion not collapsing cardinals. Then

P
Y is not the union of < nowhere dense subsets of

.
Proof. Let P be weakly (, , )manageable with a witness x H(). Suppose
toward contradiction that a condition q P is such that
q Y is the union of < nowhere dense subsets of

.
Passing to a stronger condition if needed, we may assume that for some < and
Pnames

A

(for < ) we have:


q

A


<
& (s
<
)(t

A

)(s t) and
q (y Y )( < )(t

A

)(t y) .
For each < pick a (P, , )relevant model N

(H(), , <

) such that q,

A

:
< ), x, N

. Then [

<
N

[ = < , so we may pick a y Y such that


y O for all open dense subsets O of

from

<
N

. By our assumptions, there


are < and p q such that
p (t

A

)(t y) .
Let (q

, p

) be an N

cover for (q, p) (there is one by Lemma 5.8(1)). Put


A = s
<
: (q

)(q

s

A

).
Clearly A N

, A N

, and O =

sA
O
s
N

is an open dense subset of



.
Hence s y for some s A. Let q

P be a condition stronger than q

and
such that q

s

A

. The condition q

is compatible with p

, and so with p. Take


a condition q
+
stronger than both q

and p. Then
q
+
s

A

& s y and q
+
(t

A

)(t y) ,
a contradiction.
7
9
9


r
e
v
i
s
i
o
n
:
2
0
0
5
-
0
2
-
2
8







m
o
d
i
f
i
e
d
:
2
0
0
5
-
0
2
-
2
8


COFINALITY OF THE NONSTATIONARY IDEAL 15
Corollary 5.10. Suppose that < =
<
and cov(M
,
) > . Let P be a
(, , )manageable forcing notion. Then

P

_
cov(M
,
)
_
V
cov(M
,
) .
Proof. Remembering Proposition 4.4, apply Proposition 5.9 to = cov(M
,
) =
cov(M

,
) and Y =

to get

P
(

)
V
is not the union of < nowhere dense sets .
But this clearly implies
P
cov(M
,
) = cov(M

,
) .
Theorem 5.11. Assume that < =
<
. Let

Q = P

,

Q

: < ) be
<support iteration such that for each <

is (, , )manageable .
Then P

is (, , )manageable.
Proof. Let , , and

Q be as in the assumptions of the theorem.
First note that the limits of <support iterations of <complete forcing no-
tions satisfying the condition ()

are <complete
+
cc (as
<
= ; remember
Proposition 5.3). Therefore no such iteration collapses cardinals nor changes co-
nalities nor adds sequences of ordinals of length < . Hence the assumed properties
of , and hold in all intermediate extensions V
P

and our assumption on



Q

s
is meaningful.
Plainly, P

is <complete, <
+
lubcomplete and satises condition ()

. We
have to show that P

is weakly (, , )manageable.
For < let x

be a P

name for a witness for



Q

being weakly manageable


and let x = x

: < ). Suppose that N (H(), , <

) is a (P

, , )relevant
model such that ( x,

Q) N.
Since each P

is <complete and satises the


+
cc (and +1 N) we know
that if N and G

is generic over V, then in V[G

] we have:
N[G

] V = N and N[G

] (H(), , <

)
V[G

]
and
<
N[G

] N[G

].
Since clearly

Q
G

N[G

], we conclude that N[G

] is (

Q
G

, , )relevant, and
x
G

N[G

]. Therefore, She has a winning strategy in the game


m
(N[G

], ,

Q
G

).
Let

st

be a P

name for such a strategy. We may assume that the strategy



st

is
such that
() if i < is even and q
i
= p
i
=

,
then

st

instructs Her to play q


i+1
= p
i+1
=

.
We dene a strategy st for Her in the game
m
(N, , P

) as follows. At an odd stage


i < of the game, the strategy st rst instructs Her to choose (side) conditions
q

i
, p

i
P

and only then pick conditions q


i
N P

and p
i
P

which are to
be played. These conditions will be chosen so that if q
j
, p
j
: j < i) is a legal play
of
m
(N, , P

) in which She uses st, and q

j
, p

j
are the side conditions picked by
her (for odd j i), then
()
i
Dom(q
i
) = Dom(q

i
) = Dom(p
i1
) N, Dom(p

i
) = Dom(p
i1
),
()
i
p
i1
p

i
p
i
, q
i1
q

i
p

i
, q
i1
q
i
p
i
,
7
9
9


r
e
v
i
s
i
o
n
:
2
0
0
5
-
0
2
-
2
8







m
o
d
i
f
i
e
d
:
2
0
0
5
-
0
2
-
2
8


16 PIERRE MATET, ANDRZEJ ROS LANOWSKI, AND SAHARON SHELAH
()
i
letting (q

j
, p

j
) be (q

j
, p

j
) if j i is odd and (q
j
, p
j
) if j < i is even, for
every Dom(q
i
) we have
p
i

P

q
i
() = q

i
() and
the sequence q

j
(), p

j
() : j i) is a legal play of

m
(N[
P

], ,

Q

) in which She uses the strategy



st

.
So suppose that i < is odd, q
j
, p
j
: j < i) is a partial play of
m
(N, , P

) in
which She uses st (and the side conditions for odd j < i are q

j
, p

j
), and the clauses
()
j
()
j
hold for all odd j < i. Let (q

j
, p

j
) be (q

j
, p

j
) if j < i is odd and (q
j
, p
j
)
if j < i is even.
We rst declare that Dom(q

i
) = Dom(p
i1
) N, Dom(p

i
) = Dom(p
i1
) and
p

i
() = p
i1
() for all Dom(p

i
) N. Next, by induction on Dom(q

i
)
we dene q

i
(), p

i
(). So suppose that Dom(q

i
) and q

i
, p

i
have been
dened so that q
i1
q

i
p

i
and p
i1
p

i
. Then, by clauses ()
j
,
p

i

P

the sequence q

j
(), p

j
() : j < i) is a legal play of

m
(N[
P

], ,

Q

) in which She uses the strategy



st

.
(Remember our assumption () on

st

and our convention regarding


P
stated in
Notation 0.3(1).) Let q

i
() and p

i
() be P

names for members of



Q

such that
q

i

P

i
() N[
P

] & q
i1
() q

i
() ,
and
p

i

P

(q

i
(), p

i
()) is what

st

tells Her to play


as the answer to q

j
(), p

j
() : j < i) .
(So q

i
() is a name for a member of N[
P

], but it does not have to be from


N.) This completes the denition of q

i
, p

i
P

. Now we use the fact that P

is <complete and [Dom(q

i
)[ < to pick a condition p
i
P

stronger than p

i
and names

N (for Dom(q

i
)) such that p
i

P

i
() =

. Since
<
N N, the sequence

: Dom(q

i
)) is in N. Hence we may nd a condition
q
i
N P

such that
Dom(q
i
) = Dom(q

i
), and
for each Dom(q
i
),

if

q
i1
(), then q
i
() =

, otherwise q
i
() = q
i1
() .
(For deniteness we pick the <

rst p
i
, q
i
as above.) It should be clear that
q

i
, q
i
, p

i
, p
i
satisfy conditions ()
i
()
i
. This nishes the description of the strat-
egy st. Let us argue that it is a winning strategy for Her.
To this end suppose that q
i
, p
i
: i < ) is the result of a play of
m
(N, , P

) in
which She uses st. Let q

, p

be least upper bounds of q


i
: i < ), p
i
: i < ),
respectively. Then for every Dom(p

) we have
p

() is a least upper bound of p


i
() : i < ) .
We may also assume that Dom(q

) =

i<
Dom(q
i
) = Dom(p

) N.
Let q N P

be a condition stronger than q

(and thus stronger than all q


i
for
i < ). We dene a condition p P

as follows. First, we declare that Dom(p) =


Dom(q)Dom(p

), and p() = q() for Dom(q)Dom(p

), and p() = p

() for
Dom(p

) Dom(q) = Dom(p

) N. Now suppose that Dom(q) Dom(p

)
7
9
9


r
e
v
i
s
i
o
n
:
2
0
0
5
-
0
2
-
2
8







m
o
d
i
f
i
e
d
:
2
0
0
5
-
0
2
-
2
8


COFINALITY OF THE NONSTATIONARY IDEAL 17
and we have already dened p so that q p and p

p. Then, by our
choices,
p
P

the sequence q
j
(), p

j
() : j < ) is a legal play of

m
(N[
P

], ,

Q

) in which She uses the strategy



st

, and
q() N[
P

] is stronger than all q


j
() for j < .
(Above, p

j
are as in the denition of st: either p
j
or p

j
, depending on the parity
of j.) Consequently,
p q() and p

() are compatible ,
so we may pick a P

name p() for a condition in



Q

such that
p q() p() and p

() p() .
This completes the choice of p P

. Plainly, p is an upper bound of q and p

showing that they are compatible.


Remark 5.12. Note that
if P is weakly (, , )manageable,
then it satises the
+
cc.
Hence we may use a slight modication of the proof of 5.11 to show (by induction
on ) that
if < =
<
,

Q = P

,

Q

: < ) is a (<)support iteration


of <complete weakly (, , )manageable forcing notions,
then P

is weakly (, , )manageable and complete (and thus


also
+
cc).
6. The one-step forcing
In this section we introduce a forcing notion Q for adding a small family of
functions in

which dominates

V. Iterating this type of forcing notions
we will get models with d

small. Our forcing is (of course) manageable for suitable


parameters, and thus it preserves non-meagerness of subsets of . Throughout this
section we assume the following.
Context 6.1. (i) = cf() < cf() = = 2
<
,
(ii) =

: < ) is an increasing sequence of regular cardinals,


0
,
(iii) [

<

[ = 2

and :

<



is a bijection.
We will write

for (). Also for a set u

<

we let
T(u)
def
= : < & u
Denition 6.2. (1) We dene a forcing notion Q = Q(, , ) as follows.
A condition in Q is a tuple p = (i, u,

f, g) = (i
p
, u
p
,

f
p
, g
p
) such that
(a) i < , u T

_
<

_
,
(b)

f = f

: T(u)) and f

: i for T(u),
(c) g : u i + 1, and if u, g() j < i, then

(j) < supf

(j) : < .
7
9
9


r
e
v
i
s
i
o
n
:
2
0
0
5
-
0
2
-
2
8







m
o
d
i
f
i
e
d
:
2
0
0
5
-
0
2
-
2
8


18 PIERRE MATET, ANDRZEJ ROS LANOWSKI, AND SAHARON SHELAH
The order of Q is such that for p, q Q we have
p q if and only if
i
p
i
q
, u
p
u
q
, g
p
g
q
and f
p

f
q

for T(u
p
).
(2) For a set U

<

we let QU = p Q : u
p
U, and for a condition
q Q we put
qU =
_
i
q
, u
q
U,

f
q
T(u
q
U), g
q
(u
q
U)
_
.
Proposition 6.3. (1) Q is a <lubcomplete forcing notion of size 2

.
(2) Let U

<

be of size . Then [QU[ .


Proof. 1) Plainly, (Q, ) is a partial order of size 2

. To prove the completeness


suppose that p

: <

) is an increasing sequence of members of Q and

< .
Put
i
q
= sup
<

i
p

, u
q
=
_
<

u
p

, g
q
=
_
<

g
p

and f
q

f
p

: <

& T(u
p

) for T(u
q
). Clearly q = (i
q
, u
q
,

f
q
, g
q
)
Q is the least upper bound of p

: <

).
2) Should be clear.
Proposition 6.4. The forcing notion Q satises the condition ()

(see 5.1(2))
for any limit ordinal < .
Proof. Let < be a limit ordinal. To give the winning strategy for Player I in
the game
cc
,
(Q) we need two technical observations.
Claim 6.4.1. If p, q Q are such that i
p
= i
q
and g
p
(u
p
u
q
) = g
q
(u
p
u
q
) and
f
p

= f
q

for T(u
p
) T(u
q
), then the conditions p, q have a least upper bound.
Proof of the Claim. Let i
r
= i
p
= i
q
, u
r
= u
p
u
q
, g
r
= g
p
g
q
and
f
r

=
_
f
p

if T(u
p
)
f
q

if T(u
q
).
Then r = (i
r
, u
r
,

f
r
, g
r
) Q is the least upper bound of p, q.
Claim 6.4.2. Suppose q = q
j
: j <
+
) Q. Then there is a regressive function

q
:
+

+
such that
if j < j

<
+
, cf(j) = cf(j

) = and
q
(j) =
q
(j

),
then i
qj
= i
q
j

, and g
qj
(u
qj
u
q
j

) = g
q
j

(u
qj
u
q
j

), and f
qj

= f
q
j

for
T(u
qj
) T(u
q
j

).
Proof of the Claim. Take a sequence

: <
+
)

<

such that for each j <

+
of conality and an < j we have u
q

: < j. Let U =

: <
+

and U
j
=

: < j for j <


+
. By 6.3(2) we know that [QU
j
[ (for j <
+
)
and [QU[
+
, and hence we may pick a mapping
0
:
+
QU such that
_
j <
+
__
cf(j) = Rng(
0
j) = QU
j
_
.
Also, for j <
+
, let F(U
j
) be the set
_

f = f

: Dom(

f)) : Dom(

f) T

_
T(U
j
)
_
& ( Dom(

f))(f


<
)
_
,
7
9
9


r
e
v
i
s
i
o
n
:
2
0
0
5
-
0
2
-
2
8







m
o
d
i
f
i
e
d
:
2
0
0
5
-
0
2
-
2
8


COFINALITY OF THE NONSTATIONARY IDEAL 19
and F(U) =

j<
+
F(U
j
). Note that [F(U
j
)[ and [F(U)[
+
. Choose a
function
1
:
+
F(U) such that
_
j <
+
__
cf(j) = Rng(
1
j) = F(U
j
)
_
.
Finally, let c :
+

+

+
be a bijection such that
_
j <
+
__
cf(j) = Rng(c(j j)) = j
_
.
Now let
q
:
+

+
be a regressive function such that for j <
+
of conality
we have

q
(j) = c(min <
+
:
0
() = q
j
U
j
, min <
+
:
1
() =

f
qj
T(U
j
)).
Easily,
q
is as required.
Now we may complete the proof of Proposition 6.4. Consider the following
strategy st for Player I in the game
cc
,
(Q). Suppose that the players arrived at
stage > 0 of the play and they have already constructed a sequence q

, p

:
< ). Then, for each j <
+
, the sequence p

j
: < ) is increasing, so Player
I can take its least upper bound q

j
. This determines q

played by Player I; the


function

played at this stage is the


q
given by Claim 6.4.2.
One easily veries that the strategy st is a winning one (remember Claim 6.4.1).

Theorem 6.5. Suppose and are cardinals such that = cf() < =
<
.
Then the forcing notion Q is (, , )manageable.
Proof. For each

<

<

x a sequence



<

such that

. Let
=

:

<

<

).
Suppose that N is a (Q, , )relevant model such that ( , , ) N.
For a condition p Q we dene conditions cl
+
N
(p) = q and cl

N
(p) = r by
i
r
= i
q
= i
p
,
u
r
= (u
p
N)

: T(u
p
) N, u
q
= u
p

: T(u
p
) N,
f
r

= f
p

for T(u
p
) N, f
q

= f
p

for T(u
p
), and f
q

(j) = f
r

(j) = i
p
for T(u
q
) T(u
p
), j < i
p
,
g
r
() = g
p
() for u
p
N and g
r
() = i
p
for u
r
u
p
;
g
q
() = g
p
() for u
p
and g
q
() = i
p
for u
q
u
p
.
Plainly, cl
+
N
(p), cl

N
(p) are conditions in Q and cl

N
(p) belongs to N (remember
<
N N). If p N then also cl
+
N
(p) N.
Claim 6.5.1. Suppose that p Q, q N Q are such that q p. Then
(1) cl
+
N
(q) = cl

N
(q), q cl

N
(p) cl
+
N
(p), and p cl
+
N
(p),
(2) if q

N Q is stronger than cl

N
(p), then q

and p are compatible,


(3) if p

P is stronger than cl
+
N
(p), then cl

N
(p) cl

N
(p

).
Proof of the Claim. 1) Just check.
2) Suppose cl

N
(p) q

N Q. Put
i
r
= i
q

, u
r
= u
q

u
p
, g
r
() =
_
g
q

() if u
q

,
g
p
() if u
p
u
q
and:
7
9
9


r
e
v
i
s
i
o
n
:
2
0
0
5
-
0
2
-
2
8







m
o
d
i
f
i
e
d
:
2
0
0
5
-
0
2
-
2
8


20 PIERRE MATET, ANDRZEJ ROS LANOWSKI, AND SAHARON SHELAH
if T(u
q

), then f
r

= f
q

, and if T(u
p
) T(u
q

), then
f
p

f
r

and f
r

(j) = sup

(j) : u
p
+ 1 for i
p
j < i
q

.
Note that if u
p
u
q

, then for some < we have / N (so / T(u


q

)).
Hence we may easily verify that r = (i
r
, u
r
,

f
r
, g
r
) Q and clearly r is stronger
than q

. To check that it is also stronger than p it is enough to note that:


if u
p
u
q

, then ( u
cl

N
(p)
and hence) g
q

() = g
cl

N
(p)
() = g
p
(), and
if T(u
p
) T(u
q

), then ( T(u
cl

N
(p)
) and hence) f
p

= f
cl

N
(p)

f
q

.
3) Note that if cl
+
N
(p) p

, then
(u
p
N)

: T(u
p
) N u
p

N,
so checking the conditions for cl

N
(p) cl

N
(p

) is pretty straightforward.
Claim 6.5.2. Suppose that a sequence p

: <

) Q is increasing,

< is
a limit ordinal, and cl
+
N
(p

) = p
+1
for all even <

. Let p

be the least upper


bound of p

: <

). Then cl

N
(p

) is the least upper bound of cl

N
(p

) : <

).
Proof of the Claim. It follows from Claim 6.5.1(3) that cl

N
(p

) cl

N
(p

) (for <

). To show that cl

N
(p

) is actually the least upper bound it is enough to note


that
i
cl

N
(p

)
= i
p

= supi
p

: <

= supi
cl

N
(p

)
: <

,
and
u
p

N =

u
p
+1
N : <

& even =

u
cl

N
(p

)
: <

& even,

: T(u
p

) N =

: T(u
p

) N & <

u
cl

N
(p

)
: <

,
so u
cl

N
(p

)
= u
p

N =

u
cl

N
(p

)
: <

.
Now we may describe a strategy st for Her in the game
m
(N, , Q). Suppose
that i < is even and (q
i
, p
i
) is His move at this stage of the play (so q
i
N P,
q
i
p
i
P). Then st instructs Her to play q
i+1
= cl

N
(p
i
) and p
i+1
= cl
+
N
(p
i
). It
follows from Claim 6.5.1(1) that (q
i+1
, p
i+1
) is a legal move. It follows from Claims
6.5.2 and 6.5.1(2) that the strategy st is a winning one.
Thus we have shown that Q is weakly (, , )manageable (remember 6.3(1)).
The rest follows from Propositions 6.3 and 6.4.
Denition 6.6. We dene Qnames

f

(for

<

<

) and g by

Q


f

f
p

: p
Q
& T(u
p
) ,

Q
g =

g
p
: p
Q
.
Proposition 6.7. (1)
Q
g :

<

.
(2) For each

<

<

we have
Q


f

: .
(3) For each

<

Q

_
j <
__
g() j

(j) < sup



f

(j) : <
_
.
7
9
9


r
e
v
i
s
i
o
n
:
2
0
0
5
-
0
2
-
2
8







m
o
d
i
f
i
e
d
:
2
0
0
5
-
0
2
-
2
8


COFINALITY OF THE NONSTATIONARY IDEAL 21
Proof. For

<

and i < let


1

= p Q : u
p
and 1
i
= p Q : i < i
p
.
We claim that these are open dense subsets of Q. First, suppose / u
p
, p Q
and let i
r
= i
p
, u
r
= u
p
, g
p
g
r
, g
r
() = i
r
, f
r

= f
p

for T(u
p
) and
f
r

(j) = 1 if / T(u
p
), < . Then r 1

is stronger than p. (Thus the sets


1

are dense.)
Now suppose that p P is such that i
p
i < . Put i
r
= i +1, u
r
= u
p
, g
r
= g
p
and for T(u
r
) let f
r

f
p

be such that (Dom(f


r

) = i
r
and) for j i
r
i
p
we
have f
r

(j) = sup

(j) : u
r
+ 1. This way we have dened a condition r Q
stronger than p and such that r 1
i
. (Thus the sets 1
i
are dense.)
Using the above observation and the denition of the order of Q one easily
justies (1) and (2). (Note also that, as Q is <complete,
Q

<

=
_
<

_
V
.) Then (3) follows immediately once you note that
p g
p
g & f
p

,
(for T(u
p
), p Q); remember Denition 6.2(1)(c).
7. The models
Theorem 7.1. Assume that
(a) = 2
<
,
(b) is a cardinal such that cf() < < <
cf()
= 2

,
(c) there is an increasing sequence =

: < cf()) of regular cardinals


such that
( < cf())(

)
cf()

+1
) and = sup

: < cf().
Then there is a forcing notion P such that:
(i) P has a dense subset of size 2

,
(ii) P is (, , )manageable for all cardinals , satisfying cf() = < =

<
,
(iii)
P
d
cf()

,
(iv) if cov(M
,
) > , then
P
d
cf()

= <
_
cov(M
,
)
_
V
cov(M
,
) .
Proof. Assume , , =

: < cf()) are as in the assumptions (a)(c). Note


that then also

<cf()

= 2

(by Tarskis theorem).


The forcing notion P is built as the limit of a <support iteration

Q = P

,

Q

:
<
+
). The names

Q

are dened by induction on <


+
so that
() P

has a dense subset of size 2

,
and for all cardinals , satisfying cf() = < =
<
,
() the forcing notion P

is (, , )manageable and
()
P

is (, , )manageable .
So suppose that P

is already dened (and clauses (), () hold). Then P

is <
complete
+
cc, and hence the properties of , and stated in (a)(c) hold in
7
9
9


r
e
v
i
s
i
o
n
:
2
0
0
5
-
0
2
-
2
8







m
o
d
i
f
i
e
d
:
2
0
0
5
-
0
2
-
2
8


22 PIERRE MATET, ANDRZEJ ROS LANOWSKI, AND SAHARON SHELAH
V
P

. Take a P

name

such that

<cf()



is a bijection ,
and let

Q

be a P

name for the forcing notion Q(

, , ). Then clause () holds


(remember Theorem 6.5).
It follows from Proposition 6.3(1) that the demand () is preserved at successor
stages and it is preserved at limits
+
by the support we use. By Theorem 5.11,
the clause () holds for each P

(
+
). So our P = P

+ satises (i)+(ii).
For <
+
let

f

(for

<cf()

<

) and g

be P
+1
names for functions
added by Q(

, , ) (see Denition 6.6). Then the family


T =

f

: <
+
&
_
<cf()

<

is of size . Since P is
+
cc, for each Pname

h for a member of

there are
<
+
and a P

name

h

such that
P


h =

h

. Thus using Proposition 6.7(3)


we get

P
(j

< )(F T
cf()
+(T))(j < )(j

j

h(j) < supf(j) : f F) .
Now we easily conclude that demand (iii) holds.
To show (iv) let us assume cov(M
,
) > . The forcing notion P is (
0
, , )
manageable, so by Corollary 5.10 and Proposition 4.5 we have

P
<
_
cov(M
,
)
_
V
cov(M
,
) d

.
By (iii) we know
P
d
cf()

, but as for each < cf():

P
(

)
cf()

+1
< < d

,
we immediately get that
P
d
cf()

= (remember d

(d
cf()

)
cf()
).
Corollary 7.2. Assume GCH. Then there a complete
+
cc forcing notion P

such that

P


d

= d
0

=
+
and cov(M
,
) = 2

=
+(+1)
.
Proof. Let P
0
= C

+(+1)
,
be the forcing adding
+(+1)
many Cohen functions in

(with <support). Note that

P0
, =
+
and =
+n
: n < ) are as in Theorem 7.1(a)(c) .
Therefore we have a P
0
name

P for a forcing notion satisfying 7.1(i)(iv). (Note
that
P0
cov(M
,
) =
+(+1)
, so the assumption of 7.1(iv) holds.) Let P

=
P
0


P.
Note that Theorem 0.1 follows from Corollary 7.2, Theorem 1.12, and the fact
that cov(M
,
) cof(^S

).
Theorem 7.3. Assume that (a)(c) of Theorem 7.1 hold and
(d) is a regular cardinal such that < < 2

.
Then there is a forcing notion P

satisfying (i)+(ii) of Theorem 7.1 and


(iii)
+


P
d

= for every cardinal satisfying cf() < ,


(iv)

P

_
cov(M
,
)
_
V
cov(M
,
) .
7
9
9


r
e
v
i
s
i
o
n
:
2
0
0
5
-
0
2
-
2
8







m
o
d
i
f
i
e
d
:
2
0
0
5
-
0
2
-
2
8


COFINALITY OF THE NONSTATIONARY IDEAL 23
Proof. The forcing notion P

is the limit of <support iteration



Q = P

,

Q

: <
), where

Q

are dened as in the proof of Theorem 7.1 (so the only dierence is the
length of the iteration). As there, P

satises (i)+(ii) and


P
d
cf()

. Since
cov(M
,
) > and P

is (
0
, , )manageable, we get (iv)

(by Corollary 5.10).


To show that (iii)
+

holds, suppose that



T is a P

name for a family of functions


in

of size < . Then

T is essentially a P

name for some < . Since P


+
adds a subset of which is Cohen over V
P

, (

)
V
P

is not a dominating family in


(

)
V
P
, and hence for any <

P


T is not dominating
(remember that P

is <complete).
Now, Theorem 0.2 follows from Theorem 1.12 and the following corollary.
Corollary 7.4. It is consistent, relative to the existence of a cardinal such that
o() =
++
, that

d
1
=
+1
and cov(M
1,1
) =
+2
.
Proof. Gitik and Woodin (see Gitik [Gi89] also Gitik and Merimovich [GiMe97])
constructed a model of 2
n
<

for every n < , 2


0
=
1
and 2

=
+2

from o() =
++
. Add
+2
Cohen subsets of
1
(with countable support) to that
model and then apply Theorem 7.3 (with =
1
, =

and =
+1
).
Theorem 7.5. Assume that
(a) = cf() = 2
<
, n <
(b)
0
,
1
, . . . ,
n
are cardinals such that

0
>
1
> . . . >
n
> and cf(
0
) < cf(
1
) < . . . < cf(
n
) < ,
(c) (

)
cf(

)
= 2

for = 0, . . . , n,
(d) for = 0, . . . , n, there is an increasing sequence

: < cf(

)) of
regular cardinals such that
( < cf(

))(

= (

)
cf(

)
) and

= sup

: < cf(

).
(e) cov(M
,
) = 2

.
Then there is a forcing notion P such that
(i) P has a dense subset of size 2

,
(ii) P is (, , )manageable for all cardinals , satisfying cf() = < =

<
,
(iii)
P
d
cf(

for = 0, . . . , n and cov(M


,
) = 2

.
Proof. Let A
0
, . . . , A
n
be a partition of
+
into sets of size
+
. The forcing notion
P is the limit of a <support iteration

Q = P

,

Q

: <
+
) dened like in the
proof of Theorem 7.1, but
if A

, then

is a P

name for a bijection from

<

onto

, and

is Q(

, ).
We argue that P has the required properties similarly as in Theorem 7.1.
Remark 7.6. Of course the assumption (e) in Theorem 7.5 is not very important:
we may start with adding 2

many Cohen subsets of .


7
9
9


r
e
v
i
s
i
o
n
:
2
0
0
5
-
0
2
-
2
8







m
o
d
i
f
i
e
d
:
2
0
0
5
-
0
2
-
2
8


24 PIERRE MATET, ANDRZEJ ROS LANOWSKI, AND SAHARON SHELAH
Corollary 7.7. It is consistent, relative to the existence of a strong cardinal, that
d
1
2
=
1
, d
0
2
=
1+
and cov(M
2,2
) =
1++1
.
Proof. As was pointed out to us by Moti Gitik, by work of Magidor [Mg], Meri-
movich [Me03], and Segal [Se96], it is consistent relative to a strong cardinal that
(a) 2
1
=
2
,
0
1
=
1
, and
(b) for every <
1
,

1
++1
=
++1
and 2
++1
=
++2
,
and
(c)
1
1
=
0
1+
=
1++1
.
After adding
1++1
Cohen subsets of
2
(with <
2
support) to a model of (a)
(c) we will get a model in which the assumptions of Theorem 7.5 are satised for
=
2
, n = 2,
0
=
1+
,
1
=
1
.
[References of the form math.XX/ refer to arXiv.org ]
References
[Ab97] Yoshihiro Abe. A hierarchy of lters smaller than CF

. Archive for Mathematical


Logic, 36:385397, 1997.
[BS89] Bohuslav Balcar and Petr Simon. Disjoint renement. In Handbook of Boolean alge-
bras, volume 2, pages 333388. North-Holland, Amsterdam, 1989.
[CuSh 541] James Cummings and Saharon Shelah. Cardinal invariants above the continuum.
Annals of Pure and Applied Logic, 75:251268, 1995. math.LO/9509228.
[Gi89] Moti Gitik. The negation of the singular cardinal hypothesis fromo() =
++
. Annals
of Pure and Applied Logic, 43:209234, 1989.
[GiMe97] Moti Gitik and Carmi Merimovich. Possible values for 2
n
and 2

. Annals of Pure
and Applied Logic, 90:193241, 1997.
[J] Thomas Jech. Set theory. Springer Monographs in Mathematics. Springer-Verlag,
Berlin, 2003. The third millennium edition, revised and expanded.
[La90] Avner Landver. Singular Baire numbers and related topics. PhD thesis, University
of Wisconsin, Madison, 1990.
[Mg] Menachem Magidor. On the singular cardinals problem I. Israel J. Math., 28:131,
1977.
[MPxx] Pierre Matet and Janusz Pawlikowski. Qpointness, Ppointness and feebleness of
ideals. Journal of Symbolic Logic, 68:235261, 2003.
[MPSh 713] Pierre Matet, Cedric Pean, and Saharon Shelah. Conality of normal ideals on P(),
I. Archive for Mathematical Logic. math.LO/0404318.
[MPSh 813] Pierre Matet, Cedric Pean, and Saharon Shelah. Conality of normal ideals on P(),
II. Israel Journal of Mathematics, 150:253283, 2005.
[Me03] Carmi Merimovich. Extender-based Radin forcing. Transactions of the American
Mathematical Society, 355:17291772, 2003.
[Se96] Miri Segal. Master Thesis. The Hebrew University of Jerusalem, 1996. Menachem
Magidor, adviser.
[Sh 80] Saharon Shelah. A weak generalization of MA to higher cardinals. Israel Journal of
Mathematics, 30:297306, 1978.
[Sh 288] Saharon Shelah. Strong Partition Relations Below the Power Set: Consistency, Was
Sierpi nski Right, II? In Proceedings of the Conference on Set Theory and its Appli-
cations in honor of A.Hajnal and V.T.Sos, Budapest, 1/91, volume 60 of Colloquia
Mathematica Societatis Janos Bolyai. Sets, Graphs, and Numbers, pages 637668.
1991. math.LO/9201244.
[Sh 546] Saharon Shelah. Was Sierpi nski right? IV. Journal of Symbolic Logic, 65:10311054,
2000. math.LO/9712282.
7
9
9


r
e
v
i
s
i
o
n
:
2
0
0
5
-
0
2
-
2
8







m
o
d
i
f
i
e
d
:
2
0
0
5
-
0
2
-
2
8


COFINALITY OF THE NONSTATIONARY IDEAL 25
[ShSt 154] Saharon Shelah and Lee Stanley. Generalized Martins axiom and Souslins hypothesis
for higher cardinals. Israel Journal of Mathematics, 43:225236, 1982. Corrections
in [Sh:154a].
[ShSt 154a] Saharon Shelah and Lee Stanley. Corrigendum to: Generalized Martins axiom and
Souslins hypothesis for higher cardinals [Israel Journal of Mathematics 43 (1982),
no. 3, 225236; MR 84h:03120]. Israel Journal of Mathematics, 53:304314, 1986.
Mathematiques, Universite de Caen CNRS, BP 5186, 14032 Caen Cedex, France
E-mail address: matet@math.unicaen.fr
Department of Mathematics, University of Nebraska at Omaha, Omaha, NE 68182-
0243, USA
E-mail address: roslanow@member.ams.org
URL: http://www.unomaha.edu/logic
Institute of Mathematics, The Hebrew University of Jerusalem, 91904 Jerusalem,
Israel, and Department of Mathematics, Rutgers University, New Brunswick, NJ 08854,
USA
E-mail address: shelah@math.huji.ac.il
URL: http://shelah.logic.at