Sunteți pe pagina 1din 24

3

6
8


r
e
v
i
s
i
o
n
:
1
9
9
9
-
0
5
-
1
9







m
o
d
i
f
i
e
d
:
2
0
0
2
-
0
8
-
2
6


THE CICHO

N DIAGRAM
TOMEK BARTOSZY

NSKI, HAIM JUDAH, AND SAHARON SHELAH


Abstract. We conclude the discussion of additivity, Baire number, unifor-
mity and covering for measure and category by constructing the remaining 5
models. Thus we complete the analysis of Cicho ns diagram.
1. Introduction
The goal of this paper is to describe the relationship between basic properties of
measure and category.
Denition 1.1. Let ^ and / denote the ideals of null subsets of the real line
and meager subsets of the real line respectively.
Dene the following ten sentences:
A(m) unions of fewer than 2
0
null sets is null,
B(m) ' is not the union of fewer than 2
0
null sets,
U(m) every subset of ' of size less than 2
0
is null,
C(m) ideal of null sets does not have a basis of size less than 2

0
.
Sentences A(c), B(c), U(c) and C(c) are dened analogously by replacing word
null by the word meager in the denitions above.
In addition dene
wD F [

]
<2

0
g

f F

n f(n) < g(n)


and
D F [

]
<2

0
g

f F

n f(n) < g(n).


The relationship between these sentences is described in the following diagram
which is called Cicho ns diagram:
B(m) U(c) C(c) C(m)

D wD


A(m) A(c) B(c) U(m)
In addition
A(c) B(c) & D
and
C(c) U(c) wD.
1991 Mathematics Subject Classication. 03E35, 03E15.
Key words and phrases. cardinal invariants, measure and category.
The second author would like to thank NSF under grant DMS-8505550 and MSRI for partial
support.
The third author would like to thank US-Israel Binational Science Foundation and MSRI for
partial support. Publication number 368.
1
3
6
8


r
e
v
i
s
i
o
n
:
1
9
9
9
-
0
5
-
1
9







m
o
d
i
f
i
e
d
:
2
0
0
2
-
0
8
-
2
6


2 TOMEK BARTOSZY

NSKI, HAIM JUDAH, AND SAHARON SHELAH


The proofs of these inequalities can be found in [1], [4] and [7].
In context of this diagram a natural question arises:
Are those the only implications between these sentences that are provable in
ZFC?
It turns out that the answer to this question is positive. Every combination
of those sentences which does not contradict the implications in the diagram is
consistent with ZFC. This is proved in step-by-step fashion and this paper contains
constructions of the last 5 models.
The tables below contain all known results on the subject. They are not sym-
metric but still one can recognize some patterns here. Let L be the set of sentences
obtained from sentences A, B, U, C, D and wD using logical connectives. Dene

: L L as

=
_

if =

2
if =
1

2
C if = A
U if = B
B if = U
A if = C
wD if = D
D if = wD
for L.
It turns out that if is consistent with ZFC then

is consistent with ZFC.


Moreover, in most cases one can nd a notion of forcing P such that
2
-iteration
of P over a model for CH gives a model for while
1
-iteration of P over a model
for MA & CH gives a model for

.
The rst table known as, the Kunen-Miller chart, gives consistency results con-
cerning sentences A, B, U, C only. It was completed by H. Judah and S. Shelah
in [5]. The remaining three tables give corresponding information including all 3
consistent combinations of D and wD.
3
6
8


r
e
v
i
s
i
o
n
:
1
9
9
9
-
0
5
-
1
9







m
o
d
i
f
i
e
d
:
2
0
0
2
-
0
8
-
2
6


THE CICHO

N DIAGRAM 3
Add T F F F F F
Category Baire T T T F F F
Measure Unif T T F T F F
Add Baire Unif Cov T T T T T F
T T T T A
F T T T B C D
F T F T E=E

F F T T F G H=H

I =I

F F F T D

F F F F A

A
2
-iteration with nite (countable) support of amoeba reals over a model
for CH or any model for CH or MA works.
A

2
-iteration with nite (countable) support of amoeba reals over a model
for CH or
2
-iteration of Sacks or Silver reals over a model for CH.
B
2
-iteration of random and dominating reals over a model for CH. [7]
B

1
-iteration of random and dominating reals over a model for CH & B(c).
C
2
-iteration with nite support of random reals over a model for CH. [7]
C

1
-iteration with nite support of random reals over a model for CH& D.
[7]
D Countable support
2
-iteration of innitely equal reals (see section 3) and
random reals over a model for CH. [7]
D

2
-iteration of Laver reals ([5]). We do not know if there exists a notion
of forcing P such that
2
-iteration of P over a model for CH gives D and

1
-iteration of P over a model for MA & CH gives D

.
E=E

2
random reals over a model for CH. This model is self-dual.
F
2
-iteration with nite support of any -centered notion of forcing adding
dominating reals over a model for CH . [7]
3
6
8


r
e
v
i
s
i
o
n
:
1
9
9
9
-
0
5
-
1
9







m
o
d
i
f
i
e
d
:
2
0
0
2
-
0
8
-
2
6


4 TOMEK BARTOSZY

NSKI, HAIM JUDAH, AND SAHARON SHELAH


F

1
-iteration with nite support of any -centered notion of forcing adding
dominating reals over a model for MA & CH . We can also get a model
for this case by an
2
-iteration of innitely equal reals over a model for CH.
G
2
-iteration with nite support of eventually dierent reals (see [7]) over a
model for CH.
G

1
-iteration with nite support of eventually dierent reals over a model
for CH & B(c).
H=H

2
Cohen reals over a model for CH. This model is self-dual.
I =I

2
-iteration of Mathias forcing over a model for CH [7]. This model is self
dual.
wD & D Add T F F F F F
Category Baire T T T F F F
Measure Unif T T F T F F
Add Baire Unif Cov T T T T T F
T T T T
F T T T A B
F T F T C
F F T T D E=E

F =F

F F F T B

F F F F
A
2
-iteration with nite support of random reals over a model for CH.
A

1
-iteration with nite support of random reals over a model for CH &D.
B
2
-iteration with countable support of forcing from [10] and random reals
over a model for CH (see section 5).
B

2
-iteration with countable support of rational perfect set forcing and forc-
ing Q
f,g
from [11] over a model for CH (see section 5).
C
2
Cohen and then
2
random reals over a model for CH. This model is
self-dual.
3
6
8


r
e
v
i
s
i
o
n
:
1
9
9
9
-
0
5
-
1
9







m
o
d
i
f
i
e
d
:
2
0
0
2
-
0
8
-
2
6


THE CICHO

N DIAGRAM 5
D
2
-iteration of eventually dierent reals over a model for CH. [7]
D

1
-iteration of eventually dierent reals over a model for CH & B(c).
E=E

2
Cohen reals over a model for CH. This model is self-dual.
F =F

2
-iteration with countable support of forcing Q from [2] over a model for
CH. This model is self-dual.
Models in the following two tables are dual to each other.
D Add T F F F F F
Category Baire T T T F F F
Measure Unif T T F T F F
Add Baire Unif Cov T T T T T F
T T T T A
F T T T B C
F T F T F
F F T T E D
F F F T G
F F F F
3
6
8


r
e
v
i
s
i
o
n
:
1
9
9
9
-
0
5
-
1
9







m
o
d
i
f
i
e
d
:
2
0
0
2
-
0
8
-
2
6


6 TOMEK BARTOSZY

NSKI, HAIM JUDAH, AND SAHARON SHELAH


wD Add T F F F F F
Category Baire T T T F F F
Measure Unif T T F T F F
Add Baire Unif Cov T T T T T F
T T T T
F T T T G

F T F T F

F F T T D

F F F T C

F F F F A

A
2
-iteration of amoeba reals over a model for CH or any model for MA.
A

2
-iteration of amoeba reals over a model for CH.
B
2
-iteration of dominating and random reals over a model for CH. [7]
B

2
-iteration of dominating and random reals over a model for CH &B(c).
C
2
-iteration with countable support of Mathias and random reals (see sec-
tion 5).
C

2
-iteration with countable support of forcing Q
f,g
from [11] (see section 2
and 3).
D
2
-iteration with countable support of Mathias reals over a model for CH.
D

2
-iteration with countable support of Q
f,g
and innitely equal reals over
a model for CH. (section 2)
E
2
-iteration of dominating reals over a model for CH. [7]
E

2
-iteration of dominating reals over a model for CH &MAor
2
-iteration
with countable support of eventually equal reals.
F
2
random reals over a model for MA & 2
0
=
2
.
F

2
random reals over a model for CH.
G
2
-iteration with countable support of Laver reals over a model for CH.
3
6
8


r
e
v
i
s
i
o
n
:
1
9
9
9
-
0
5
-
1
9







m
o
d
i
f
i
e
d
:
2
0
0
2
-
0
8
-
2
6


THE CICHO

N DIAGRAM 7
G

2
-iteration with countable support of innitely equal and random reals
over a model for CH. [5]
2. Not adding unbounded reals
Our rst goal is to construct a model for ZFC & wD & U(c) & B(m) & U(m).
We start with the denition of the forcing which will be used in this construction.
This family of forcing notions was dened in [11].
Denition 2.1. Let f

and g

be two functions such that


(1) f(n) >

j<n
f(j) for n ,
(2) g(n, j + 1) > f(n)
2
g(n, j) for n, j ,
(3) minj : g(n, j) > f(n + 1)
n
.
Let
Seq
f
=
_
n

j<n
f(j).
For a tree T dene T
[s]
= t T : s t or t s, succ
T
(s) = t T : t
s, lh(t) = lh(s) + 1. If T = T
[s]
for some s T then s is called a stem of T.
Let Q
f,g
be the following notion of forcing: T Q
f,g
i
(1) T is a perfect subtree of Seq
f
,
(2) there exists a function h

diverging to innity such that


n m n s T
m
[succ
T
(s)[ g(m, h(m)).
Elements of Q
f,g
are ordered by .
Let Q

f,g
Q
f,g
be the set dened as follows: T Q

f,g
i there exists s
0
Seq
f
such that T = T
[s0]
and there exists an increasing function h

such that
m lh(s
0
) s T
m1
[succ
T
(s)[ g(m, h(m)).
Clearly Q

f,g
is dense in Q
f,g
and therefore from now on we will work with
conditions in this form. Notice that
Lemma 2.2. V
Q
f,g
[= V

is meager in

.
Proof Notice that if r is a Q
f,g
-generic real then by an easy density argument
we show that
h V

n h(n) ,= r(n).
Therefore V

n h(n) ,= r(n) which is a meager set.


Denition 2.3. We say that notion of forcing P is

-bounding if
V
P

| r V

n (n) r(n).
The following theorem was proved in [11], we prove it here for completeness;
Theorem 2.4. Q
f,g
is

-bounding.
Proof We will need the following
Denition 2.5. For T, T

Q
f,g
and

k dene T

k
T

if
(1) T T

,
(2) s T succ
T
(s) ,= succ
T
(s) [succ
T
(s)[ g(lh(s),

k).
3
6
8


r
e
v
i
s
i
o
n
:
1
9
9
9
-
0
5
-
1
9







m
o
d
i
f
i
e
d
:
2
0
0
2
-
0
8
-
2
6


8 TOMEK BARTOSZY

NSKI, HAIM JUDAH, AND SAHARON SHELAH


Claim 2.6. Suppose that T
n
: n is a sequence of elements of Q
f,g
such that
T
n+1

kn
T
n
for n where k
n
: n is an increasing sequence of natural
numbers. Then there exists T Q
f,g
such that T
kn
T
n
for n .
Proof For n dene
u
n
= minj : k j s T
n

k
[succ
Tn
(s)[ g(k, k
n
).
Let T =

n
T
n
u
n
. Function h(m) = k
n1
for m [u
n1
, u
n
) witnesses that
T Q
f,g
.
Lemma 2.7. Let T Q
f,g
and be such that T | . Suppose that

k .
Then there exists

T

k
T and n such that
s

T
n
a
s


T
[s]
| = a
s
.
Proof Let S T be the set of all t T such that T
[t]
satises the lemma.
In other words
S = t T : n
t

k
T
[t]
s

T
nt
a
s


T
[s]
| = a
s
.
We want to show that stem of T belongs to S. Notice that if s , S then
[succ
T
(s) S[ g(lh(s),

k).
Suppose that stem of T does not belong to S and by induction on levels build a
tree

S

k
T such that for s

S,
succ

S
(s) =
_
succ
T
(s) if [succ
T
(s) S[ g(lh(s),

k)
succ
T
(s) succ
S
(s) otherwise
.
Clearly

S Q
f,g
since g(lh(s), m)g(lh(s),

k) g(lh(s), m

k) for all s and m >

k.
Find

S
1


S and n such that

S
1
| = n. Now get t T and

S
2


S
1
such
that

S
2

k
T
[t]
. But that contradicts the denition of the condition

S.
We nish the proof of the theorem. Suppose that T |

. Build by
induction sequences T
n
: n and k
n
: n such that for n ,
(1) T
n+1

kn
T
n
,
(2) s T
n+1

kn
a
s
T
[s]
n+1
| (n) = a
s
.
Let T = lim
n
T
n
and let r(n) = maxa
s
: s T
kn
for n . Then
T | n (n) r(n)
which nishes the proof.
Notice that in fact we proved that
Lemma 2.8. If T |

then there exists a sequence k


n
: n and a tree

T T such that
s

T
kn
a
s


T
[s]
| (n) = a
s
.
Our next goal is to show that forcing with Q
f,g
does not add random reals. We
will need the following
3
6
8


r
e
v
i
s
i
o
n
:
1
9
9
9
-
0
5
-
1
9







m
o
d
i
f
i
e
d
:
2
0
0
2
-
0
8
-
2
6


THE CICHO

N DIAGRAM 9
Denition 2.9. Let f

and let X
f
=

n=0
f(n). Dene S
f
as follows: T S
f
if T is a perfect subtree of Seq
f
and
lim
n
[T
n
[

n1
m=1
f(m)
= 0.
Notion of forcing Q is called f-bounding if
X
f
V
Q
T S
f
V n n T.
Theorem 2.10. Let P be a notion of forcing. We have the following
(1) If P is an f-bounding notion of forcing then P does not add random reals.
(2) If P is

-bounding and P does not random reals then P is f-bounding for


every f

.
Proof Dene a measure on X
f
as a product of equally distributed, nor-
malized measures on f(n).
(1) Every element of S
f
corresponds to a closed, measure zero subset of X
f
. This
nishes the proof as X
f
is isomorphic to the Cantor space with standard measure.
(2) Suppose that | X
f
. Since we assume that P does not add random reals
we can nd a null G

subset H V of X
f
such that | H.
Claim 2.11. Suppose that H X
f
. Then (H) = 0 i there exists a sequence
J
n
Seq
f

n
: n such that
(1) H x X
f
:

n xn J
n
,
(2)

n=0
(x X
f
: xn J
n
) < .
Proof () This implication is an immediate consequence of Borel-Cantelli
lemma.
() Since (H) = 0 there are open sets G
n
: n covering H such that
(G
n
) <
1
2
n
for n . Write each G
n
as a union of disjoint basic sets i.e.
G
n
=
_
m
[s
n
m
] for n .
Let J
n
= s Seq
f

n
: s = s
l
k
for some k, l for n . Verication of (1)
and (2) is straightforward.
Let J
n
: n be a sequence obtained by applying the above to the set H.
In particular n : n J
n
is innite. Using the fact that forcing P is

-
bounding nd a function h

such that n m [h(n), h(n + 1)) m J


m
.
Let
C =

n
h(n+1)
_
m=h(n)
_
sJm
[s].
It is easy to see that C is a closed set and that | C. As C is a closed set C is
a set of branches of some tree T. This tree has required properties.
The following theorem was proved in [11], we prove it here for completness.
Theorem 2.12. Forcing Q
f,g
is f-bounding.
Proof We start with the following
3
6
8


r
e
v
i
s
i
o
n
:
1
9
9
9
-
0
5
-
1
9







m
o
d
i
f
i
e
d
:
2
0
0
2
-
0
8
-
2
6


10 TOMEK BARTOSZY

NSKI, HAIM JUDAH, AND SAHARON SHELAH


Lemma 2.13. If

T | n (n) f(n) then there exists tree

T

T such that
s

T
n
a
s
f(n)

T
[s]
| (n) = a
s
.
Proof By applying 2.8 we get a tree T

T and a sequence k
n
: n
such that
s T
kn
a
s
T
[s]
| (n) = a
s
.
Without loss of generality we can assume that k
n
n for all n . Suppose
that function h

witnesses that T Q
f,g
. In other words [succ
T
(s)[
g(lh(s), h(lh(s))) for s T.
Build by induction a family of trees T
n,l
: n , n l k
n
such that
(1) T
n,l
T
n,l
for l l

, n ,
(2) T
n,l
l = T
n,l
l for l l

, n ,
(3) T
n,l
T
m,l
for n < m and all l, l

,
(4) T
n,l
n = T
m,l
n for n < m and all l, l

,
(5) n s T
n,l

l
a
s
f(n) T
[s]
n,l
| (n) = a
s
,
(6) n s T
n,n

n
[succ
T
n,l
(s)[ g(lh(s), h(lh(s)) 1).
It is clear that

T = lim
n
T
n,n
has the required properties and the function h

(n) = h(n) 1 witnesses that



T
Q
f,g
.
Suppose that the tree T
n,n
is given for some n . Trees T
n+1,kn
T
n+1,kn1

. . . T
n+1,n+1
are constucted by induction as follows:
Let T
n+1,kn
= T
n,n
and suppose that T
n+1,l
is given. Tree T
n,l1
will be dened
in the following way: T
n,l1
l 1 = T
n,l
l 1 and for each t T
n,l

l1
we will
specify which of the immediate successors of t belong to T
n,l1
.
Take t T
n+1,l

l1
and let s succ
T
n+1,l
(t). By (5) there exists a
s
f(n)
such that T
[s]
n+1,l
| (n) = a
s
. That denes a partition of the set succ
T
n+1,l
(t)
into f(n) many pieces. Let the set of immediate successors of t in T
n+1,l1
be the
largest piece in this partition.
Notice that for t T
n
the set succ
T
(t) will be altered at most n times and
each time its size will decrease by a factor f(i) for i n. Therefore
[succ
Tn,n
(t)[ >
g(n, h(n))

in
f(i)
g(n, h(n) 1).
This veries (6) and nishes the proof of the lemma.
Now we can prove the theorem. Let be a Q
f,g
-name such that

T | n (n)
f(n) for some

T Q
f,g
.
Let

T

T be the condition as in the lemma above. The tree T

we are looking
for will be dened as follows:
s T

i t

T

T
[t]
| lh(s) = s.
By trimming

T some more we can see that
[T


n
[

n
m=1
f(m)

[

T
n
[

n
m=1
f(m)
n
0 .
3
6
8


r
e
v
i
s
i
o
n
:
1
9
9
9
-
0
5
-
1
9







m
o
d
i
f
i
e
d
:
2
0
0
2
-
0
8
-
2
6


THE CICHO

N DIAGRAM 11
To conclude this section we need some preservation theorems. We have to show
that a countable support iteration of

-bounding forcings is

-bounding. This
has been proved for proper forcings (see [9]). Here we present a much easier proof
that works for a more limited class of partial orderings. Similarly we need to know
that the iterations we use do not add random reals. Unfortunately f-boundedness
is not preserved by a countable support iteration. We will prove it only for certain
partial orderings. For a general preservation theorem of a slightly stronger property
called (f, g)-boundedness see [12].
Denition 2.14. Let P be a notion of forcing satisfying axiom A (see [3]). We
say that P has property () if for every p P, n and a P-name for a natural
number there exists N and q
n
p such that q | < N.
It is easy to see that partial orderings having property () are

-bounding.
Theorem 2.15. Let P

,

Q

: < be a countable support iteration of forcings


that have the property (). Then P

= lim
<
P

is

-bounding.
Proof For p, q P

, F []
<
and n write p
F, n
q if
(1) p q,
(2) F p | p()
n
q().
The proof of the theorem is based on the following general fact:
Lemma 2.16. Suppose that p P

, F []
<
and n are given. Let be
a P-name for a natural number. Then there exists q
F, n
p and N such that
q | < N.
Proof It will be proved by induction on ([F[, min F) over all possible models.
Suppose that [F[ = n+1 and min F =
0
< . By induction hypothesis in V
P
0
+1
the lemma is true for F

= F
0
. Therefore there exists a Q
0
name V
P
0
such that
V
P
0
[= p(
0
) | q
F

, n
p(
0
, ) q | < .
Since Q
0
has property () in V
P
0
we can nd q


n
p(
0
) and N such that
V
P
0
[= q

| < N.
The last statement is forced by a condition q
0
P
0
. Let q = q

0
q

. It is the
condition we were looking for.
Let p
0
be any element of P

. Suppose that p
0
|

. Using 2.16 dene by


induction sequences p
n
: n , F
n
: n and a function r

such that
(1) p
n+1

Fn,n
p
n
for n ,
(2) supp(p
n
) j F
j
,
(3) F
n
F
n+1
for n ,
(4) p
n+1
| (n) < r(n).
Let q be the limit of p
n
: n . Then q | n (n) < r(n).
Finally we can prove:
Theorem 2.17. Con(ZFC) Con(ZFC & wD & U(c) & B(m) & U(m)).
Proof The following notion of forcing was introduced in [7]: let f

.
Dene
p Q
f
i
3
6
8


r
e
v
i
s
i
o
n
:
1
9
9
9
-
0
5
-
1
9







m
o
d
i
f
i
e
d
:
2
0
0
2
-
0
8
-
2
6


12 TOMEK BARTOSZY

NSKI, HAIM JUDAH, AND SAHARON SHELAH


(1) p : dom(p) ,
(2) dom(p) and dom(p) is innite,
(3) n p(n) f(n).
For p, q Q
f
p q if p q and for n p
n
q i p q and the rst n elements
of dom(p) and dom(q) are the same.
The following fact is well known:
Lemma 2.18. Let P be a notion of forcing. If P has the Laver property then P is
f-bounding for all functions f

.
Lemma 2.19. Let f

be a strictly increasing function such that f(n) > 2


n
for
n . Then
(1) V 2

has measure zero in V


Q
f
,
(2) Q
f
is f-bounding.
Proof (1) It is enough to show that X
f
V has measure zero in V
Q
f
. Notice
that for h X
f
the set
H
h
= x X
f
:

n x(n) = h(n)
has measure zero. It is easy to see that
X
f
V H
hG
where h
G
is a generic real.
(2) Let p
0
be any element of Q
f
. Suppose that p
0
| X
f
. Dene by
induction sequences p
n
: n , k
n
: n and J
n
: n such that
(1) J
n
Seq
f

kn
for n ,
(2) p
n+1

n
p
n
for n ,
(3) p
n+1
| k
n
J
n
for n ,
(4)
[J
n
[

kn
m=1
f(m)

1
n
for n .
Let q p
0
be the limit of p
n
: n and T =

n
J
n
. By removing all nodes
whose ancestors are missing we can make sure that T is a tree. Then q forces that
is a branch through T and by (4) T has measure zero.
Let P

,

Q

: <
2
be a countable support iteration such that
|


= Q
f,g
if is even
|


= Q
f
if is odd.
Let P = P
2
. Then V
P
[= wD since P is

-bounding, and V
P
[=
U(c) & U(m) by the properties of forcings Q
f,g
and Q
f
(note that Q
f,g
has prop-
erty ()). To nish the proof we need
Lemma 2.20. P is f-bounding.
Proof For p, q P, F [
2
]
<
and n denote p
F, n
q if
(1) p q,
(2) F p | p()
n
q().
Let p
0
be any element of P. Suppose that p
0
| X
f
. Using the fact that
both Q
f,g
and Q
f
are f-bounding and arguing as in the proofs of 2.13 and 2.19,
dene by induction sequences p
n
: n , F
n
: n , k
n
: n and
J
n
: n such that
3
6
8


r
e
v
i
s
i
o
n
:
1
9
9
9
-
0
5
-
1
9







m
o
d
i
f
i
e
d
:
2
0
0
2
-
0
8
-
2
6


THE CICHO

N DIAGRAM 13
(1) J
n
Seq
f

kn
for n ,
(2) p
n+1

Fn,n
p
n
for n ,
(3) supp(p
n
) j F
j
,
(4) F
n
F
n+1
for n ,
(5) p
n+1
| k
n
J
n
for n ,
(6)
|Jn|

kn
m=1
f(m)

1
n
for n .
Let q p
0
be the limit of p
n
: n and T =

n
J
n
. As before, by removing
non-splitting nodes we can assume that T is a tree. Then q forces that is a branch
through T and by (6) T has measure zero.
Notice that 2.20 can be proved in the same way for many other forcings including
perfect set forcing from section 5.
3. Preserving old reals have outer measure 1
In this section we construct a model for ZFC & wD& U(c) & U(m) & B(m).
It is obtained by
2
-iteration with countable support of Q
f,g
.
The main problem is to verify that U(m) holds in that model.
We will use the following technique from [5].
Denition 3.1. Let P be a notion of forcing. Dene

1
[P] i for every suciently large cardinal , and for every countable elementary
submodel N H(, ), if P N and

I
n
: n N is a P-name for a sequence
of rational intervals and p
n
: n N is a sequence of elements of P such that
p
0
|

n=1
(

I
n
) < and p
n
|

I
n
= I
n
for n then for every random real x
over N, if x ,

n
I
n
then there exists q p
0
such that
(1) q is (N, P)-generic,
(2) q | x is random over N[G] for every P-generic lter over N containing p
0
,
(3) q | x ,

I
n
.

2
[P] i for every P-name

A for a subset of 2

and every p P, if p | (

A)
then

(x 2

: q p q | x ,

A) 1 .

3
[P] i for every A V 2

of positive measure V
P
[=

(A) > 0.

4
[P] i for every suciently large cardinal , and for every countable elementary
submodel N H(, ), if P N and p
n
: n N is a sequence of P and


A
n
: n N is a sequence of elements of P-names such that for n
p
n
|

A
n
is a Borel set of measure
n
, and lim
n

n
= 0 then for every random
real x over N there exists a condition q P such that
(1) q is (N, P)-generic,
(2) q | x is random over N[G] for every P-generic lter over N containing p
0
,
(3) there exists n such that q p
n
and q | x ,

A
n
.
In [5] it is proved that
Lemma 3.2. For every notion of forcing P,
(1) If P is weakly homogenous then
2
[P]
3
[P],
(2)
1
[P]
4
[P].
3
6
8


r
e
v
i
s
i
o
n
:
1
9
9
9
-
0
5
-
1
9







m
o
d
i
f
i
e
d
:
2
0
0
2
-
0
8
-
2
6


14 TOMEK BARTOSZY

NSKI, HAIM JUDAH, AND SAHARON SHELAH


Lemma 3.3. Suppose that P has property
1
. Then V
P
[=V2

is not measur-
able.
Proof It is enough to show that V 2

has positive outer measure. Let

I
n
: n be a P-name for a sequence of rational intervals such that p
0
|

n
(

I
n
) < 1. Find sequences p
n
: n , j
n
: n , and I
n
: n
such that for n
(1) p
n+1
p
n
,
(2) p
n+1
|

I
j
= I
j
for j j
n
,
(3) p
n+1
|

j=jn
(

I
j
)
1
n
.
It is easy to see that

n
(I
n
) .
Choose a countable, elementary submodel N of H() containing Pand p
n
, j
n
,

I
n
, I
n
:
n . Since N is countable there exists x V 2

such that x is a random real


over N and x ,

n
I
n
. Using
1
[P] we get q p such that q | x ,

I
n
.
Since

I
n
: n was arbitrary it shows that
V
P
[=

(V 2

) = 1
which nishes the proof.
The lemma above would be even easier to prove if we assume
3
[P]. The reason
for using property
1
[P] is in the following:
Theorem 3.4 ([5]). Suppose that P

,

Q

: < is a countable support iteration


such that |

has property
1
for < . Let P = P

. Then P has property

1
.
To construct the model satisfying ZFC & wD & U(c) & U(m) & B(m)
we show that forcing Q
f,g
has property
1
. At the rst step we show that it has
property
3
i.e.
Theorem 3.5. Let A 2

be such that (A) =


0
> 0. Then V
Q
f,g
[=

(A) > 0.
Proof Suppose that this theorem is not true. Then there exists a set A 2

such that

(A) =
0
> 0, a condition T Q
f,g
and a sequence

I
n
: n of
Q
f,g
-names for rational intervals such that
(1) T |

n=1
(

I
n
) = 1,
(2) T | A

n>m

I
n
.
Let s
0
be the stem of T. By 2.8 without losing generality we can assume that there
exists an increasing sequence of natural numbers k
n
: n such that
(1) For every s T
kn
T
[s]
forces a value to

I
j
: j n,
(2) T |

nlh(s0)
(

I
n
) <
1
2

0
,
(3)

n=lh(s0)
(1
1
f(n)
) >
1
2
.
For s T and j dene
I
s
j
=
_
I if T
[s]
|

I
j
= I
otherwise
.
Suppose that a function h

witnesses that T Q
f,g
and consider a function
h

such that h

(n) h(n) for n .


Claim 3.6. For x 2

the following condition are equivalent:


3
6
8


r
e
v
i
s
i
o
n
:
1
9
9
9
-
0
5
-
1
9







m
o
d
i
f
i
e
d
:
2
0
0
2
-
0
8
-
2
6


THE CICHO

N DIAGRAM 15
(1) There exists T

T such that h

witnesses that T

Q
f,g
and T

| x ,

I
n
,
(2) For every k lh(s
0
) there exists a nite tree t of height k such that
(a) t T
k
,
(b) [succ
t
(s)[ g(lh(s), h

(lh(s))) for s t
lh(s0)
,
(c) If s t
k
then x ,

j
I
s
j
.
Proof (1) (2) If T

satises (1) then T

k satises (2)
(2) (1) Build a sequence t
k
: k satisfying (2) and apply the compactness
theorem to construct T

.
Dene a set D 2

as follows:
y D i there exists T

Q
f,g
such that
(1) T

T has the same stem as T (=s


0
),
(2) T

| y ,

nlh(s0)

I
n
,
(3) n lh(s
0
) s T


n
[succ
T
(s)[ g(n, h(n) 1).
Notice that the set D is dened in V and since T | A

nlh(s0)

I
n
we have
(2

D) >
0
.
For k lh(s
0
) dene sets D
k
as follows:
y D
k
i there exists a nite tree t such that
(1) t T
k
,
(2) n lh(s
0
) s t
n
[succ
t
(s)[ g(n, h(n) 1),
(3) s t
k
y ,

nlh(s0)
I
s
n
.
By the above claim D =

k
D
k
. Since sets D
k
form a decreasing family we can
nd k such that (2

D
k
) >
0
.
For every s T such that lh(s
0
) lh(s) k dene set D
k,s
as follows:
y D
k,s
i there exists a nite tree t such that
(1) t T
k
and t = t
[s]
,
(2) n lh(s) s

t
n
[succ
t
(s

)[ g(n, h(n) 1),


(3) s

t
k
y ,

nlh(s0)
I
s

n
.
Notice that D
k
= D
k,s0
. Observe also that for s T
k
(2

D
k,s
)

nlh(s0)
(I
s
n
) <

0
2
.
Claim 3.7. Suppose that for some m [lh(s
0
), k 1] and s T
m
,
(2

D
k,t
) a for t succ
T
(s).
Then
(2

D
k,s
)
a
1
g(m, h(m) 1)
g(m, h(m))
.
Proof Notice that y , D
k,s
i [t succ
T
(s) : y , D
k,t
[ > g(m, h(m))
g(m, h(m) 1).
Claim 3.8. Let N
1
> N
2
be two natural numbers. Suppose that A
j
: j N
1
is
a family of subsets of 2

of measure a. Let U = x 2

: x belongs to at least
N
2
sets A
j
. Then
(U) a
N
1
N
2
.
3
6
8


r
e
v
i
s
i
o
n
:
1
9
9
9
-
0
5
-
1
9







m
o
d
i
f
i
e
d
:
2
0
0
2
-
0
8
-
2
6


16 TOMEK BARTOSZY

NSKI, HAIM JUDAH, AND SAHARON SHELAH


Proof Let
Ai
be the characteristic function of the set A
i
for i N
1
. It
follows that
_
iN1

Ai
N
1
a and therefore

_
_
_
_
_
x 2

iN1

Ai
(x) N
2
_
_
_
_
_

N
1
N
2
a.
By applying the claim above we get
(2

D
k,s
) a
g(m, h(m))
g(m, h(m)) g(m, h(m) 1)
=
a
1
g(m, h(m) 1)
g(m, h(m))
.
Finally by induction we have
(2

D
k
) = (2

D
k,s0
)

0
2

1
M
where
M =
m=k

lh(s0)
_
1
g(m, h(m) 1)
g(m, h(m))
_

m=k

lh(s0)
_
1
1
f(m)
_
>
1
2
.
Therefore (2

D
k
) <
0
which gives a contradiction.
Now we can prove
Theorem 3.9. Q
f,g
has property
1
.
Proof We will need several denitions:
Denition 3.10. Let

I
n
: n be a Q
f,g
-name for a sequence of rational
intervals. We say that T Q
f,g
interprets

I
n
: n if there exists an increasing
sequence k
n
: n such that for every j n and s T
kn
T
[s]
decides
a value of

I
j
i.e. T
[s]
|

I
j
= I
s
j
for some rational interval I
s
j
.
By 2.8 we know that
T Q
f,g
: T interprets

I
n
: n
is dense in Q
f,g
. Suppose that T Q
f,g
. Subset S T is called front if for every
branch b through T there exists n such that bn S.
Suppose that D Q
f,g
is an open set. Dene
cl(D) = T Q
f,g
: s T : T
[s]
D is a front in T.
Let

I
n
: n be a Q
f,g
-name for a sequence of rational intervals such that
for some T
0
Q
f,g
T
0
|

n=1
(

I
n
) < < 1 and T
0
interprets

I
n
: n .
Let N H() be a countable model containing Q
f,g
, T
0
,

I
n
: n .
Dene a set Y 2

as follows:
x Y i there exists

T Q
f,g
such that
(1)

T T
0
,
(2) If D N is an open, dense subset of Q
f,g
then there exists T

cl(D) N
such that

T T

,
(3)

T | x ,

I
n
,
3
6
8


r
e
v
i
s
i
o
n
:
1
9
9
9
-
0
5
-
1
9







m
o
d
i
f
i
e
d
:
2
0
0
2
-
0
8
-
2
6


THE CICHO

N DIAGRAM 17
(4) Suppose that J =

I
n
: n N is a Q
f,g
-name for a sequence of
rational intervals such that |

n=1
(

I
n
) < and let D
J
= T Q
f,g
:
T interprets

I
n
: n (with sequence k
T
n
: n ). Then there exists
T D
J
N and k such that
m k s

T
k
T
m
x , I
T,s
m
.
Notice that (2) guarantees that

T is (N, Q
f,g
)-generic while (4) guarantees that x
is random over N[G].
Lemma 3.11.
(1) Y is a
1
1
set of reals (in V),
(2) (Y ) 1 .
Proof (1) It is easy to see that conditions (1)-(4) in the denition of Y are
Borel provided that we have an enumeration (we can code as a real number) of the
objects appearing in (2) and (4).
(2) easy computation using the fact that Q
f,g
has property
3
and
2
.
Work in N. Let G Coll(
0
, 2
0
) be generic over N and let x be a random real
over N[G]. Let B denotes the measure algebra. Since parameters of the denition
of Y are in N[G] we can ask whether N[G][x] [= x Y .
Since in N[G], Y is a measurable set we can nd two disjoint, Borel sets A and B
such that (AB) = 1 and A |
B
x Y and B |
B
x , Y . Morover (A) 1.
In other words A Y a.e. and B 2

Y a.e.
Since x is a random real over N as well we have
Coll(
0
, 2

0
) B

= Q
x


R

= B

R
where Q
x
is the smallest subalgebra which adds x.
Find a Borel set of positive measure A

such that
N [= A

|
B
p

R p | x A

and
N [= 2

|
B
| x B

.
It is clear that A

A has measure zero and therefore (A

) 1 .
Notice that the denitions above do not depend on the choice of random real x
as long as x A

. Thus if x is any random real over N such that x A

then we
can nd an N-generic lter G Coll(
0
, 2
0
) such that (G, x) is Coll(
0
, 2
0
) B-
generic over N and N[G][x] [= x Y . Since Y is a
1
1
set it means that V [= x Y .
In other words there exists a Borel set A

of measure 1 such that if x VA

is a random real over N then x Y .


Now we nish the proof of the theorem. Let N , p
n
: n ,

I
n
: n and
x be such that
(1) p
n+1
p
n
for n ,
(2) p
n
|

I
n
= I
n
for n ,
(3) x ,

n
I
n
,
(4)

n=1
(I
n
) = .
Dene for n , Y
n
= set Y dened for model N, condition p
n
and set

I
m+n
:
m .
By the above remarks we can nd Borel sets A

n
: n N such that for
n (A

n
) 1 (

jn
(I
j
)) and for every x V A

n
if x is random over
3
6
8


r
e
v
i
s
i
o
n
:
1
9
9
9
-
0
5
-
1
9







m
o
d
i
f
i
e
d
:
2
0
0
2
-
0
8
-
2
6


18 TOMEK BARTOSZY

NSKI, HAIM JUDAH, AND SAHARON SHELAH


N then x Y
n
. Since (

n
A

n
) = 1 if x is random over N then x A

n
for some
n . Therefore x Y
n
and this nishes the proof as Y
n
Y
0
for all n . From
the fact that x Y
0
follows the existence of the condition witnessing
1
.
Theorem 3.12. Con(ZFC) Con(ZFC & wD & U(c) & U(m) & B(m)).
Proof Let P

,

Q

: <
2
be a countable support iteration such that
|

= Q
f,g
for <
2
. Let P = P

2
. Then V
P
[= U(m) because P has
property
1
and V
P
[= B(m) and wD since P is f-bounding and

-bounding
by 2.20 and 2.15. Finally V
P
[= U(c) by 2.2.
4. Rational perfect set forcing
Our next goal is to construct a model for
ZFC & wD & D & U(c) & U(m) & B(m).
We will do it in the next section. H This model is obtained as a
2
-iteration with
countable support of Q
f,g
and rational perfect set forcing. In this section we will
prove several facts about rational perfect set forcing which we will need later.
Recall that rational perfect set forcing is dened as follows:
T R i T is a perfect subtree of
<
and for every s T there exists s t T
such that succ
T
(t) is innite.
Elements of R are ordered by .
Without loss of generality we can assume that for every T R and s T the
set succ
T
(s) is either innite or contains exactly one element since elements of this
form are dense in R.
For T R dene
split(T) = s T : succ
T
(s) is innite .
For T, T

R let
T
0
T

if T T

and T and T

have the same stem.


T


n
T if T

T and for every s split(T) if exactly n proper segments of s


belong to split(T) then s split(T

).
First we have to show that forcing R preserves outer measure.
Denition 4.1. Let

I
n
: n be an R-name for sequence of rational intervals
such that |

n=1
(

I
n
) =
1
2
.
We say that T R interprets

I
n
: n if for every s split(T) there exist
rational intervals I
s
1
, . . . , I
s
ns
such that
(1) T
[s]
| j n
s

I
j
= I
s
j
,
(2) for every > 0 and every branch y through T there exists m such that
for k m

_
_
_
jn
yk
I
yk
j
_
_

1
2
.
Lemma 4.2. Suppose that

I
n
: n is an R-name for sequence of rational
intervals. Assume that T |

n=1
(

I
n
) =
1
2
. Then there exists

T T such that

T interprets

I
n
: n .
3
6
8


r
e
v
i
s
i
o
n
:
1
9
9
9
-
0
5
-
1
9







m
o
d
i
f
i
e
d
:
2
0
0
2
-
0
8
-
2
6


THE CICHO

N DIAGRAM 19
Proof Construct a sequence T
n
: n R such that T
n+1

n
T
n
for
n as follows:
T
0
= T and suppose that T
n
is already constructed.
For every s split(T
n
) such that exactly n proper segments of s belong to
split(T
n
) and every m such that s

m succ
Tn
(s) extend T
[s

{m}]
to decide
a suciently long part of

I
n
: n . Paste all extensions together to get T
n+1
.
Clearly

T =

n
T
n
has required property.
Now we are ready to show:
Theorem 4.3. If A 2

and (A) = 1 then |


R

(A) > 0.
Proof Suppose not. Then there exists a measure one set A 2

, a R-name
for sequence of rational intervals

I
n
: n and a condition T R such that
(1) T |

n=1
(

I
n
) =
1
2
.
(2) T | A

mn

I
m
.
By the above lemma we can assume that T interprets

I
n
: n .
For s split(T) and > 0 dene
h

(s) = min
_
_
_
j :

ij
(I
s
i
)
1
2

_
_
_
and
A

s
=
_
ih

(s)
I
s
i
.
Note that h

(s) may be undened for some and s.


Let N be a countable, elementary submodel of H() for suciently big .
Let x A be a random real over N . The following holds in N[x].
Lemma 4.4. For every > 0 there exists a tree T

T such that
(1) T

has no innite branches.


(2) for every s T

either x A

s
or n : s

n succ
T
(s) succ
T
(s)
is nite.
Proof Fix > 0. For s split(T) dene an ordinal r

(s) as follows:
r

(s) = 0 i x A

s
,
r

(s) = limsupr

(t) + 1 : t succ(s) and r

(t) is dened.
In other words r

(s) i for all < there exists innitely many t succ


T
(s)
such that r

(t) .
Claim 4.5. For every s split(T) ordinal r

(s) is well dened.


Proof If not we inductively build a condition T

T
[s]
such that r

(t) is not
dened for all t split(T

). But then T

| x ,

nh

(s)

I
n
. Contradiction.
Let s
0
be the stem of T. Dene
T

= s T : s
0
s or for all k < l if r

(sk) and r

(sl) are dened then


r

(sk) > r

(sl))
It is easy to see that T

has no branches since for every branch y through T there


exists m such that for k m r

(yk) = 0.
On the other hand if x , A

s
then by the denition of rank the set t succ(s) :
r

(t) r

(s) is at most nite which veries (2).


3
6
8


r
e
v
i
s
i
o
n
:
1
9
9
9
-
0
5
-
1
9







m
o
d
i
f
i
e
d
:
2
0
0
2
-
0
8
-
2
6


20 TOMEK BARTOSZY

NSKI, HAIM JUDAH, AND SAHARON SHELAH


By the above lemma for every > 0 there exists a tree T

together with a
function r

: split(T

)
1
such that
s, t split(T

) s t r

(s) > r

(t).
Since N[x] is a generic extension of N there exists Borel set B 2

of positive
measure such that
N [= B |
B
> 0 there exist r

and T

as in 4.4 .
Fix
0
= (B)/2 and let r and

T be B-names for r
0
and T
0
.
We can nd Borel set B

B such that (B

) >
1
2
(B) and for s split(T)
(1) n : B

| s

T & s

n ,

T & (B

A
0
s
) = 0 is
nite,
(2)
1
: B

| r(s) = is nite.
To show this we use the fact that the measure algebra B is

-bounding and

T is
forced to satisfy 4.4(2).
Now dene in N

T = s T : B

|
B
s

T
and
r(s) = max( <
1
: B

|
B
s

T & r(s) = ).
Notice that these denitions do not depend on the initial choice of random real
x as long as x B

.
Lemma 4.6.
(1)

T is a subtree of T,
(2) If s

T and x B

is any random real over N such that x , A


0
s
and
s

T[x] then n : s

n

T[x]

T is nite,
(3) If t s

T then r(t) > r(s).
Proof (1) and (2) follow immediately from the denition of

T and the choice
of the set B

.
(3) Suppose that r(s) = . It means that there exists a set B

such that
B

| r(s) = .
Thus
B

| r(t) is well dened and >


so < r(t).
In particular it follows from (3) that the tree

T is well-founded, i.e. has no
innite branches, and that r :

T
1
is a rank function such that
s t

T r(s) > r(t).
By induction on rank dene sets X
s
2

for s split(

T) as follows:
If r(s) = 0 then X
s
= A
0
s
. If r(s) > 0 then X
s
= z 2

: z belongs to all but


nitely many sets X
t
where t is an immediate successor of t is split(

T).
It is easy to check that (X
s
)
0
for s split(

T).
Choose x A (B

X
s0
) which is random over N. Since x , X
s0
we can nd
innitely many immediate successors s of s
0
in split(

T) such that x , X
s
. Choose
one of them, say s
1
s
0
3
6
8


r
e
v
i
s
i
o
n
:
1
9
9
9
-
0
5
-
1
9







m
o
d
i
f
i
e
d
:
2
0
0
2
-
0
8
-
2
6


THE CICHO

N DIAGRAM 21
such that x , X
s1
and s
1


T[x]. By repeating this argument with s
1
instead
of s
0
and so on we construct a branch through

T[x]. Contradiction since the tree

T[x] is well-founded. .
By repeating the proof of 3.9 we get
Theorem 4.7. R has property
1
.
5. Not adding dominating and Cohen reals
In this section we construct models for
(1) ZFC & D & B(m) & B(c) & U(m),
(2) ZFC & wD & D & B(c) & B(m) & U(m),
(3) ZFC & wD & D & U(c) & U(m) & B(m)).
We need the following denitions.
Denition 5.1. Let P be a notion of forcing. We say that P is almost

-bounding
if for every P-name such that p |

there exists a function f V

such that for every subset A V []

there exists q p such that


q |

n A (n) f(n).
We say that P is weakly

-bounding if for every P-name such that p |

there exists a function f V

such that there exists q p such that


q |

n (n) f(n).
We will use the following two preservation theorems.
Theorem 5.2 ([10]). Let P

,

Q

: < be a countable support iteration such


that for <
|

is almost

-bounding.
Then P

= lim
<
P

is weakly

-bounding.
Denition 5.3. Let P be a notion of forcing satisfying axiom A. We say that P
has Laver property if there exists a function f
P

such that for every nite set


A V, P-name a , p P and n if p | a A then there is q
n
p and a set
B A of size f
P
(n) such that q | a B.
Notice that this denition is actually stronger than standard denition of Laver
property.
Theorem 5.4 ([6]). Let S and suppose that P

,

Q

: < is a countable
support iteration such that
|

is a random real forcing if S


|

has Laver property if , S.


Let P = P

. Then no real in V
P
is Cohen over V.
Now we can prove that:
Theorem 5.5.
(1) Con(ZFC) Con(ZFC & D & B(m) & B(c) & U(m)),
(2) Con(ZFC) Con(ZFC & wD & D & B(c) & B(m) & U(m)),
(3) Con(ZFC) Con(ZFC & wD & D & U(c) & U(m) & B(m)).
3
6
8


r
e
v
i
s
i
o
n
:
1
9
9
9
-
0
5
-
1
9







m
o
d
i
f
i
e
d
:
2
0
0
2
-
0
8
-
2
6


22 TOMEK BARTOSZY

NSKI, HAIM JUDAH, AND SAHARON SHELAH


Proof (1) Let P

,

Q

: <
2
be a countable support iteration such that
|

is a random real forcing if is even


|

is Mathias forcing if is odd.


Let P = P

2
. Then
V
P
[= D & B(m) & U(m) because Mathias and random reals are added
conally in the iteration and
V
P
[= B(c) by 5.4.
(2) Let P

,

Q

: <
2
be a countable support iteration such that
|

is a random real forcing if is even


|

is Shelah forcing from [2] if is odd.


Let P = P

2
. Then
V
P
[= wD & B(m) & U(m) because of properties of Shelah forcing and random
forcing. To show that V
P
[= B(c) we use 5.4 and the fact that Shelah forcing has
the Laver property.
(3) Let P

,

Q

: <
2
be a countable support iteration such that
|


= Q
f,g
if is even
|


= R if is odd.
Let P = P

2
. Since R is has Laver property ([8]) exactly as in 2.20 we show
that P is f-bounding. Therefore V
P
[= B(m). V
P
[= U(m) since Q
f,g
and R
have property
1
. Also V [= wD & U(c) since R adds unbounded reals and by 2.2.
To nish the proof of (2) and (3) we have to check that forcings used there do
not add dominating reals. By 5.2 it is enough to verify that both Shelah forcing
and rational perfect set forcing are almost

-bounding and this will be proved in


the next theorem.
Theorem 5.6. (1) Rational perfect set forcing R is almost

-bounding,
(2) The Shelah forcing is almost

-bounding.
Proof Let be an R-name such that T |

for some T R. As in
4.2 we can assume that for every s split(T) and t succ
T
(s), T
[t]
decides the value
of lh(s). Notice that in this case every branch through T gives an interpretation
to . Let N be a countable, elementary submodel of H() such that R, T and
belong to N. Let g V

be a function which dominates all elements of


N

. Fix a set A V []

. Since forcing R has absolute denition it is


enough to show that for every m and every condition T

N R, T T

there exists a condition T

N R, T

and n A [0, m] such that


N [= T

| (n) g(n). Choose T

T and let b N be a branch through T

.
Let
b
N

be the interpretation of obtained using b. By the assumption


there exists n A, n m such that
b
(n) g(n). Choose T

= T

[t]
where
t = bn.
(2) The proof presented here uses notation from [2]. Since the denition of
Shelahs forcing and all the necessary lemmas can be found in [2] we give here only
a skeleton of the proof.
Let p = (w, T) S and let be an S-name for an element of

. Let q be a
pure extension of p satisfying 2.4 of [2]. Suppose that q = (w, t
0
, t
1
, . . . ). We dene
by induction a sequence q
l
: l satisfying the following conditions:
(1) q
0
= q,
(2) q
l+1
= (w, t
l+1
0
, t
l+1
1
, . . . ) is an l-extension of q
l
,
3
6
8


r
e
v
i
s
i
o
n
:
1
9
9
9
-
0
5
-
1
9







m
o
d
i
f
i
e
d
:
2
0
0
2
-
0
8
-
2
6


THE CICHO

N DIAGRAM 23
(3) if k l + 1 and (w, w

) t
l+1
0
. . . t
l+1
k
and w

[n(t
l+1
k
), m(t
l+1
k
)) ,= when
t
l+1
k
K
n(t
l+1
k
),m(t
l+1
k
)
then (w

, t
l+1
k+1
, t
l+1
k+1
, . . . ) forces value for k,
(4) Dp(t
l+1
l+1
) > l.
Before we construct this sequence let us see that this is enough to nish the proof.
Let q

= (w, t
1
1
, t
2
2
, . . . ). By (4), q

S.
Let g(n) = maxk : w

(w, w

) t
1
1
. . . t
n
n
and (w

, t
n+1
n+1
, t
n+2
n+2
, . . . ) | (n) = k
for n .
Clearly g

. Suppose that A . Dene


p
A
= (w, (t
i
i
: i A)).
It is easy to see that
p
A
|

n A (n) g(n)
which nishes the proof.
We build the sequence q
l
: l by induction on l. Suppose that q
l
is already
given. By the denition of S it is enough to build the condition for some xed
w

= w m(t
l
0
, . . . t
l
l
).
Dene a function C :
<
2 as follows:
C(v) = 1 i k (w

, v) t
l
l+1
, . . . , t
l
k
and (v, t
l
k+1
, t
l
k+2
, . . . ) forces value for (l).
Using lemma 2.6 from [2] we get a condition where the function C is constantly 0
or 1. The rst is impossible since the set of conditions forcing a value for (l) is
dense. Therefore we get a condition q = q
l+1
on which C is constantly 1. Moreover
we can assume that q
l+1
is an l-extension of q
l
.
This nishes the induction and the proof.
References
[1] T.Bartoszynski Additivity of measure implies additivity of category, Transactions of
AMS 1984.
[2] A.Blass, S.Shelah There may be simple P

1
- and P

2
-points and the Rudin-Keisler order
may be downward directed, Annals of Pure and Applied Logic, vol. 33, 1987
[3] J.Baumgartner Iterated forcing in Surveys in set theory, London Mathematical Society
Lecture Note Series, No. 8, Cambridge University Press, Cambridge, 1983.
[4] D.Fremlin On Cicho ns diagram, Initiation a lAnalyse, Universite Pierre et Marie
Curie, Paris 1985
[5] H. Judah, S. Shelah The Kunen-Miller chart, Journal of Symbolic Logic, vol.55 (1990)
[6] J.Judah, S.Shelah
1
3
sets to appear in Journal of Symb. Logic
[7] A.Miller Some properties of measure and category, Trans. AMS,1983.
[8] A. Miller Rational perfect set forcing, Contemporary Mathematics vol.31 1983
[9] S. Shelah Proper Forcing, Springer Lecture Notes in Mathematics, 1982
[10] S. Shelah On cardinal invariants of the continuum, Contemporary Mathematics vol.
31, 1983
[11] S. Shelah Vive la dierence, in Set theory of the continuum, Springer Verlag 1992
[12] S. Shelah Proper and improper forcing, to appear
3
6
8


r
e
v
i
s
i
o
n
:
1
9
9
9
-
0
5
-
1
9







m
o
d
i
f
i
e
d
:
2
0
0
2
-
0
8
-
2
6


24 TOMEK BARTOSZY

NSKI, HAIM JUDAH, AND SAHARON SHELAH


Department of Mathematics University of California Berkeley, CA 94720
Current address: Department of Mathematics Boise State University, Boise, Idaho 83725
E-mail address: tomek@math.idbsu.edu
Department of Mathematics University of California, Berkeley and Mathematical
Sciences Research Institute, 1000 Centennial Drive, Berkeley, California
Current address: Department of Mathematics Bar Ilan University Ramat Gan, Israel 52900
E-mail address: judah@bimacs.cs.biu.ac.il
Department of Mathematics Rutgers University, New Brunswick, New Jersey and
Mathematical Sciences Research Institute, 1000 Centennial Drive, Berkeley, Califor-
nia and Department of Mathematics Hebrew University Jerusalem
Current address: Department of Mathematics, Hebrew University, Jerusalem
E-mail address: shelah@sunrise.huji.ac.il

S-ar putea să vă placă și