4

4
6


r
e
v
i
s
i
o
n
:
1
9
9
3
-
0
8
-
2
9







m
o
d
i
f
i
e
d
:
1
9
9
3
-
0
8
-
2
9


Baire Property and Axiom of Choice
Haim Judah
Bar-Ilan University
Abraham Fraenkel Center
for
Mathematical Logic
Department of Mathematics and Computer Science
52-900 Ramat-Gan, Israel
Saharon Shelah
Hebrew University
Institute of Mathematics
Jerusalem, Israel
October 6, 2003
1 Introduction
In 1979 Shelah proved that in order to obtain a model in which every set of
reals has Baire property, a large cardinal assumption is not necessary. The
model he constructed satised
L
1
=
1
. Therefore Woodin asked if we can
get a model for ZF + DC(
1
) + each set of reals has Baire property.
Recall here that DC(
1
) is the following sentence:
if 1 is a relation such that (X)(Y )(1(X, Y )) then there is a
sequence < Z

: <
1
> such that
( <
1
)(1(< Z

: < > , Z

)).
Note that DC(
1
) implies the following version of choice:
1
4
4
6


r
e
v
i
s
i
o
n
:
1
9
9
3
-
0
8
-
2
9







m
o
d
i
f
i
e
d
:
1
9
9
3
-
0
8
-
2
9


if 1
1
R
then there exists a choice function f :
1
R such that
1(, f()) for each <
1
.
In [JS1] we studied the consistency strength of ZFC + variants of MA +
suitable sets of reals have Baire property. We showed that Baire property
for
1
3
-sets of reals plus MA(-centered) implied that
1
is a Mahlo cardinal
in L.
The natural question that arises at this point is:
Do we need large cardinals to construct a model in which all
projective sets of reals have Baire property and the union of any

1
meager sets is meager?
Note that if unions of
1
many null sets are null then every
1
2
-set of reals is
Lebesgue measurable. Consequently if each projective sets of reals has Baire
property and any union of
1
null sets is null then
1
is inaccessible in L.
The aim of the present paper is to prove the following two theorems:
Theorem 1.1 If ZF is consistent then the following theory is consistent:
ZF + DC(
1
) + Every set of reals has Baire property
Theorem 1.2 If ZF is consistent then the following theory is consistent:
ZFC + Every projective set of reals has Baire property + Any
union of
1
meager sets is meager
Our notation is standard and derived from [Jec]. There is one exception,
however. We write p q to say that q is a stronger condition then p.
denotes the smallest element of a forcing notion.
2 Basic denitions and facts
In this section we recall some denitions and results from [She]. They will
be applied in the next section.
The basic tool in the construction of models in which denable sets have
Baire property is the amalgamation. To dene this operation we need the
following denition.
2
4
4
6


r
e
v
i
s
i
o
n
:
1
9
9
3
-
0
8
-
2
9







m
o
d
i
f
i
e
d
:
1
9
9
3
-
0
8
-
2
9


Recall that P < P

means P P

and each maximal antichain in P is a

maximal antichain in P

. For a forcing notion P let

P
be a P-name for the
generic subset of P.
Denition 2.1 Suppose that P < BA(Q). Then (Q: P) is the P-name of
a forcing notion which is a subset of Q,
(Q: P) = q Q : q is compatible with every p
P
.
Thus p q (Q: P) if and only if every p

P, p

p is compatible with
q. Recall that if P < BA(Q) then forcing notions Q and P (Q : P) are
equivalent.
Denition 2.2 Let P
0
, P
1
and P
2
be forcing notions. Suppose that f
1
:
P
0
11
BA(P
1
), f
2
: P
0
11
BA(P
2
) are complete embeddings (i.e. they
preserve order and f
i
[P
0
] < BA(P
i
) ). We dene the amalgamation of P
1
and P
2
over f
1
, f
2
by P
1

f
1
,f
2
P
2
=
(p
1
, p
2
)P
1
P
2
: (pP
0
)(p p
1
(P
1
: f
1
[P
0
]) &p
2
(P
2
: f
2
[P
0
]))
P
1

f
1
,f
2
P
2
is ordered in the natural way: (p
1
, p
2
) (p

1
, p

2
) if and only if
p
1
p

1
, p
2
p

2
.
Note that P
1
, P
2
can be completely embedded into the amalgamation
P
1

f
1
,f
2
P
2
by p
1
P
1
(p
1
, ) and p
2
P
2
(, p
2
). Thus we think of
P
1

f
1
,f
2
P
2
as an forcing notion extending both P
1
and P
2
.
The amalgamation is applied in constructing of Boolean algebras admit-
ting a lot of automorphisms. The mapping
f
1
2
f
1
1
: f
1
[P
0
] P
2
can be naturally extended to an embedding
: P
1
P
1

f
1
,f
2
P
2
.
Now. suppose that B is a complete Boolean algebra such that for suciently
many pairs (P
1
, P
2
) of complete suborders of B and for complete embeddings
f
i
: P
0
P
i
, (i = 1, 2) the algebra B contains the amalgamation P
1

f
1
,f
2
P
2
. Then B is strongly Cohen-homogeneous:
Suppose is a B-name for an
1
-sequence of ordinals. Then there exists a
complete subalgebra B

of the algebra B such that

3
4
4
6


r
e
v
i
s
i
o
n
:
1
9
9
3
-
0
8
-
2
9







m
o
d
i
f
i
e
d
:
1
9
9
3
-
0
8
-
2
9


is a B

-name,
if B

< B

< B, B

(B

: B

) is the Cohen algebra and f : B

B
is a complete embedding such that f[B

= id
B

then there exists an automorphism : B

onto
B extending f.
For more details on extending homomorphisms see [JuR].
Solovay showed the connection between the strong homogeneity of the
algebra B and the fact that in generic extensions via B all projective sets of
reals have Baire property. Let S
1
be the class of all
1
-sequences of ordinal
numbers.
Theorem 2.3 (Solovay) Let B be a strongly Cohen homogeneous complete
Boolean algebra satisfying ccc. Suppose that for any B-name for an
1
-
sequence of ordinals
B the union of all meager Borel sets coded in V[] is meager.
Then B any set of reals denable over S
1
has Baire property.
Proof See theorem 2.3 of [JuR].
The class HOD(S
1
) consists of all sets hereditarily ordinal denable over S
1
.
Theorem 2.4 (Solovay) Assume that every set of reals ordinal denable
over S
1
has Baire property. Then
HOD(S
1
)[= ZF + DC(
1
) + every set of reals has Baire property.
Proof See [Sol].
In the next section we will built a model in which there exists an algebra
B satisfying the assumptions of theorem 2.3 and such that
B the union of
1
meager sets is meager.
To be sure that the algebra B satises ccc we will use the following notion.
Denition 2.5 A triple (P, T, E
n

n
) is a model of sweetness if
1. P is a notion of forcing and T is a dense subset of P,
2. E
n
are equivalence relations on T such that
4
4
4
6


r
e
v
i
s
i
o
n
:
1
9
9
3
-
0
8
-
2
9







m
o
d
i
f
i
e
d
:
1
9
9
3
-
0
8
-
2
9


each E
n
has countably many equivalence classes (the equivalence
class of the element p T in the relation E
n
will be denoted by
[p]
n
),
equivalence classes of all relations E
n
are upward directed,
if p
i
: i T, p
i
[p

]
i
for all i then for every n < there
exists q [p

]
n
which is stronger than all p
i
for i n,
if p, q T, p q and n then there exists k such that
(p

[p]
k
)(q

[q]
n
)(p

).
Note that if (P, T, E
n

n
) is a model of sweetness then P is -centered.
Denition 2.6 We say that a model of sweetness (P
2
, T
2
, E
2
n

n
) ex-
tends a model (P
1
, T
1
, E
1
n

n
) (we write
(P
1
, T
1
, E
1
n

n
) < (P
2
, T
2
, E
2
n

n
)) whenever
1. P
1
< P
2
, T
1
T
2
and E
1
n
= E
2
n
[T
1
for each n ,
2. if p T
1
, n then [p]
2
n
T
1
,
3. if p q, p T
2
, q T
1
then p T
1
.
Lemma 2.7 a) The relation < is transitive on models of sweetness.
b) Suppose that (P
i
, T
i
, E
i
n

n
) are models of sweetness such that
(P
i
, T
i
, E
i
n

n
) < (P
j
, T
j
, E
j
n

n
)
for all i < j < ( <
1
). Then
lim
i<
(P
i
, T
i
, E
i
n

n
) = (

i<
P
i
,

i<
T
i
,

i<
E
i
n

n
)
is a model of sweetness extending all models (P
i
, T
i
, E
i
n

n
).
The sweetness may be preserved by the amalgamation.
Lemma 2.8 Suppose that (P
i
, T
i
, E
i
n

n
) for i = 1, 2 are models of
sweetness and f
i
: P
0
BA(P
i
) are complete embeddings. Then there ex-
ists a model of sweetness (P
1

f
1
,f
2
P
2
, T

, E

n
) based on the amalgama-
tion P
1

f
1
,f
2
P
2
and extending both (P
1
, T
1
, E
1
n

n
) and (P
2
, T
2
, E
2
n

n
).
5
4
4
6


r
e
v
i
s
i
o
n
:
1
9
9
3
-
0
8
-
2
9







m
o
d
i
f
i
e
d
:
1
9
9
3
-
0
8
-
2
9


Proof see lemmas 7.5, 7.12 of [She].
To ensure that our algebra satises
B the union of
1
meager sets is meager
we will use the Hechler order D. Recall that D consists of all pairs (n, f)
such that n , f

. It is ordered by
(n, f) (n

, f

) if and only if
n n

, f[n = f

[n and (k)(f(k) f

(k)).
The forcing with D adds both a dominating real and a Cohen real. Conse-
quently
D

D the union of all Borel meager sets coded in the ground
model is meager.
The iteration with D preserves sweetness.
Lemma 2.9 Let (P, T, E
n

n
) be a model of sweetness and let

D be
a P-name for the Hechler forcing. Then there exists a model of sweetness
(P

D, T

, E

n
) based on P

D and extending the model (P, T, E
n

n
).
Proof Similar to the proof of lemmas 7.6, 7.11 of [She].
3 The proof of the main result
In this section we present proofs of theorems 1.2 and 1.1.
Denition 3.1 Let / be the class consisting of all sequences

P = < (P
i
, M
i
) : i <
1
> such that
1. M
i
is a model of sweetness based on P
i
,
2. if i < j <
1
then P
i
< P
j
.
If

P / is as above then we put P

1
=

i<
1
P
i
.
6
4
4
6


r
e
v
i
s
i
o
n
:
1
9
9
3
-
0
8
-
2
9







m
o
d
i
f
i
e
d
:
1
9
9
3
-
0
8
-
2
9


Note that if

P / then each P
i
is -centered. Consequently P

1
satises
ccc.
We dene the relation on /.
Denition 3.2 Let

P
1
,

P
2
/. We say

P
1


P
2
if P

1
1
< P

1
2
and there
exists a closed unbounded subset C of
1
such that
(!) if i C then M
i
1
< M
i
2
(!!) if i C, q P

1
1
, p P
i
1
and p
P
i
1
q (P

1
1
: P
i
1
)
then p
P
i
2
q (P

1
2
: P
i
2
).
Clearly the relation is transitive and reexive.
Lemma 3.3 Suppose that

P
m
/ for m < are such that m
1
< m
2
<
implies

P
m
1


P
m
2
(and let C
m
1
,m
2
witness it). Let C =

m
1
<m
2
<
C
m
1
,m
2
.
Put
P
i

m<
P
(C\i)
m
, M
i

= lim
m<
M
(C\i)
m
.
Then

P

= < (P
i

, M
i

) : i <
1
> / and

P
m


P

for each m < .

Proof First note that C is a closed unbounded subset of
1
. Since
C

m<
C
m,m+1
we may apply lemma 2.7 b) to conclude that each M
i

is
a model of sweetness based on P
i

.
Claim: If i < j <
1
then P
i

< P
j

.
Indeed, let i < j. We may assume that i, j C (recall that P
i

= P
(C\i)

).
Note that P
i
m
< P
i

and P
i
m
< P
j
m
for each m . Let / P
i

be a maximal
antichain. Clearly it is an antichain in P
j

but we have to prove that it is

maximal. Let q P
j

. Then q P
j
m
for some m < . Let
Z = r P
i
m
: (p
r
/)(r
P
i
m
p
r
(P
i

: P
i
m
))
Clearly Z is dense in P
i
m
. Hence we nd r Z such that r
P
i
m
q (P
j
m
: P
i
m
).
Let p
r
/ witness r Z. Take k such that p
r
P
i
k
, m < k < . Consider

P
m
and

P
k
. Since i, j C C
m,k
we may apply condition (!!) to conclude
that
r
P
i
k
q (P
j
k
: P
i
k
).
7
4
4
6


r
e
v
i
s
i
o
n
:
1
9
9
3
-
0
8
-
2
9







m
o
d
i
f
i
e
d
:
1
9
9
3
-
0
8
-
2
9


By the choice of p
r
we have
r
P
i
m
p
r
(P
i
k
: P
i
m
).
Thus p
r
and r are compatible and any p

P
i
k
, p

r, p
r
is compatible with
q. Consequently q and p
r
are compatible. The claim is proved.
It follows from the above claim that

P

/.
Claim: The club C witness that

P
m


P

for each m < .

Indeed, rst note that
P

i<
1
P
i

i<
1

m<
P
(C\i)
m
=

m<
P

1
m
.
Since P

1
m
1
< P

1
m
2
for each m
1
< m
2
we see that P

1
m
< P

. It follows from
the denition of M
i

and lemma 2.7 that if i C then M

i
m
< M
i

. Thus
we have to check condition (!!) only. Suppose i C, q P

1
m
, p P
i
m
and
p
P
i
m
q (P

1
m
: P
i
m
). Assume p ,
P
i

q (P

: P
i

). Then we nd r P
i

such that r p and r is incompatible with q. Let k > m be such that r P

i
k
.
Since i C
m.k
we have p
P
i
k
q (P

1
k
: P
i
k
) (by condition (!!) for

P
m
,

P
k
).
But r
P
i
k
q , (P

1
k
: P
i
k
) - a contradiction.
Lemma 3.4 Assume that

/ for <
1
,
if < <
1
then

P

is witnessed by the club C

,

1
,
if <
1
is a limit ordinal and i

<<
C
,
then M
i

= lim
<
M
i

.
Let
C = <
1
: is limit & ( < <)( C
,
)
and let C(i) = (Ci) for i <
1
. Put P
i

1
= P
C(i)
C(i)
, M
i

1
= M
C(i)
C(i)
.
Then

P

1
/ and ( <
1
)(

1
).
Proof First note that the set <
1
: ( < <)( C
,
) is the
diagonal intersection of clubs

<
C
,
(for <
1
). Hence C is closed and
unbounded and

P

1
is well dened.
Claim: If i < j <
1
then P
i

1
< P
j

1
.
8
4
4
6


r
e
v
i
s
i
o
n
:
1
9
9
3
-
0
8
-
2
9







m
o
d
i
f
i
e
d
:
1
9
9
3
-
0
8
-
2
9


Indeed, suppose i < j <
1
. Then P
i

1
= P
C(i)
C(i)
, P
j

1
= P
C(j)
C(j)
and we
may assume that C(i) < C(j). By 3.1 2) we have that P
C(i)
C(i)
< P
C(j)
C(i)
.
Since C consists of limit ordinals only and C(j)

<<C(j)
C
,
we get
P
C(j)
C(j)
=

<C(j)
P
C(j)

(and it is a direct limit). Since C(i) < C(j) we conclude

P
C(j)
C(i)
< P
C(j)
C(j)
and consequently P
C(i)
C(i)
< P
C(j)
C(j)
. The claim is proved.
Since each M
i

1
is a model of sweetness based on P
i

1
we have proved that

1
/. Let <
1
.
Claim: P

< P

1
First note that
P

1
=

i<
1
P
i

1
=

i<
1
P
C(i)
C(i)
=

,i<
1
P
i

<
1
P

.
Since
1
<
2
<
1
implies

P

1


P

2
we have P

1
< P

2
for
1
<
2
<
1
.
Consequently P

< P

1
.
Claim: If i C( + 1) then M
i

< M
i

1
.
If i C( + 1) then C(i) = i > . Moreover it follows from our
assumptions that M
i
i
= lim
<i
M
i

. By lemma 2.7 we get M

i

< M
i
i
= M
C(i)
C(i)
=
M
i

1
.
Claim: Suppose i C( +1), q P

, p P
i

and p
P
i

q (P

: P
i

).
Then p
P
i

1
q (P

1
: P
i

1
).
Assume not. Then we have r P
i

1
= P
i
i
, r p such that r and q are
incompatible. There is (, i) such that r P
i

. Thus p ,
P
i

q (P

: P
i

).
Since i C
,
we get a contradiction with condition (!!) for

P

.
We have proved that the club C( + 1) witness

P

1
.
Suppose

P = < (P
i
, M
i
) : i <
1
> /. Let
P
i
D
= (p, ) P

D : p P
i
& is a P
i
-name .
Note that P
i
D
is isomorphic to P
i

D. Let M
i
D
be the canonical model of
sweetness based on P
i
D
and extending the model M
i
(see lemma 2.9). Let

P
D
= < (P
i
D
, M
i
D
) : i <
1
> .
Lemma 3.5

P
D
/,

P

P
D
and P

1
D
= P

D.
9
4
4
6


r
e
v
i
s
i
o
n
:
1
9
9
3
-
0
8
-
2
9







m
o
d
i
f
i
e
d
:
1
9
9
3
-
0
8
-
2
9


Proof The last assertion is a consequence of the fact that P

1
is a ccc
notion of forcing. It follows from properties of Souslin forcing (cf [JS2]) that
P
i
D
< P
j
D
provided i < j. Consequently

P
D
/. To show

P

P
D
note that
M
i
< M
i
D
for all i <
1
and P

1
< P

1
D
. Suppose now that i <
1
, p P
i
,
q P

1
and p
P
i q (P

1
: P
i
). Assume that p ,
P
i
D
q (P

1
D
: P
i
D
). Then
we nd a condition r = (r
0
, ) P
i
D
above p which is inconsistent with q.
Note that q may be a member of P
i

D < P
j
(for some j > i) but we consider
it as an element of P

1
, while r is an element of P

D. Consequently
incompatibility of q and r means that q and r
0
are not compatible. But
r
0
P
i
lies above p - a contradiction.
Lemma 3.6 Suppose that B, (, T, (
0
are complete Boolean algebras such
that
(1) B < T < (, (
0
< (
Let B
0
= B (
0
, T
0
= T (
0
(note that B
0
< T
0
< (
0
). We assume that
(2) B (T: B) is a subset of ((
0
: B)
(3) if b B, b
0
B
0
and b
0

B
0
b (B: B
0
) then b
0

C
0
b (( : (
0
).
Then
(3

) if d T, d
0
T
0
and d
0

D
0
d (T: T
0
) then d
0

C
0
d (( : (
0
).
Proof
Claim: Suppose c (
0
, d
0
T
0
and d
0

D
0
c ((
0
: T
0
). Then
d
0

D
c (( : T).
We have to prove that each d d
0
, d T is compatible with c. Let
d d
0
, d T. By (2) we nd b B and d
1
T
0
such that
b
B
d (T: B) & d
(D:B)
d
1

(the last means that bd = bd

1
). Thus bd
1
d
0
= bdd
0
= bd
0
,= 0. We nd
b
0
B
0
such that b
0

B
0
b (B: B
0
) and b
0
d
1
d
0
,= 0 (it is enough to take
b
0
such that b
0

B
0
bd
1
d
0
(T: B
0
)). Note that then b
0

C
0
b (( : (
0
) (by
(3)). Since b
0
d
1
d
0
T
0
and it is stronger than d
0
we get b
0
d
1
d
0
c ,= 0. The
last condition is stronger than b
0
and belongs to (
0
. Hence bb
0
d
1
d
0
c ,= 0.
10
4
4
6


r
e
v
i
s
i
o
n
:
1
9
9
3
-
0
8
-
2
9







m
o
d
i
f
i
e
d
:
1
9
9
3
-
0
8
-
2
9


Finally note that bb
0
d
1
d
0
c bd
1
= bd d so d and c are compatible.
The claim is proved.
Now suppose that d T, d
0
T
0
and d
0

D
0
d (T: T
0
). Let c (
0
, c d
0
.
Take d

T
0
such that d

d
0
and d

D
0
c ((
0
: T
0
). By the claim we
have d

D
c (( : T). Since d

d
0
we have d

d ,= 0, d

d T and
consequently d

dc ,= 0. Hence d and c are compatible and we are done.

Suppose that

P
0
,

P
1
,

P
2
,

P
3
/ and the club C
1
witness that both

P
0


P
1
and

P
2


P
3
. Assume that Q
0
, Q
2
are complete Boolean algebras
such that for some i
0
<
1
BA(P

1
0
) < Q
0
< BA(P

1
1
), BA(P

1
2
) < Q
2
< BA(P

1
3
)
BA(P

1
0
) (Q
0
: BA(P

1
0
)) (BA(P
i
0
1
): BA(P

1
0
))
BA(P

1
2
) (Q
2
: BA(P

1
2
)) (BA(P
i
0
3
): BA(P

1
2
))
Let f : Q
0
Q
2
be an isomorphism such that f[Q
0
BA(P
i
1
)] = Q
2

BA(P
i
3
) for all i Ci
0
. For i Ci
0
put
P
i
= (p
1
, p
2
) P

1
1

id,f
P

1
3
: p
1
P
i
1
& p
2
P
i
3
,
where id stands for the identity on Q
0
. It follows from lemma 3.6 that P
i
is
isomorphic to P
i
1

f
1
,f
3
P
i
3
, where f
3
= f[Q
0
BA(P
i
1
) and f
1
is the identity
on Q
0
BA(P
i
1
). Therefore we have the canonical model of sweetness M
i
based on P
i
and extending both models M
i
1
and M
i
2
(compare lemma 2.8).
Let

P
1

f

P
3
= < (P
i
, M
i
) : i <
1
> .
Note that

i<
1
P
i
= P

1
1

id,f
P

1
3
.
Lemma 3.7

P
1

f

P
3
/ and

P
1
,

P
3


P
1

f

P
3
.
Proof To prove

P
1

f

P
3
/ we have to show the following
Claim: P
i
< P
j
for each i < j <
1
, i, j Ci
0
.
Let / P
i
be a maximal antichain and let (p
1
, p
2
) P
j
. Let q Q
0
be
such that
q p
1
(P

1
1
: Q
0
) & p
2
(P

1
3
: f[Q
0
]).
Take r
1
P
i
1
such that r
1

P
i
1
p
1
, q (P

1
1
: P
i
1
) (note that q and p
1
are
compatible). Next nd q

Q
0
such that q

q and q

r
1
(P

1
1
: Q
0
)
11
4
4
6


r
e
v
i
s
i
o
n
:
1
9
9
3
-
0
8
-
2
9







m
o
d
i
f
i
e
d
:
1
9
9
3
-
0
8
-
2
9


(recall that r
1
and q are compatible). Since p
2
and f(q

) are compatible we
nd r
2
P
i
3
such that r
2

P
i
3
p
2
, f(q

) (P

1
3
: P
i
3
). Consider the pair
(r
1
, r
2
). There is q

Q
0
, q

such that q

r
2
(P

1
3
: f[Q
0
]). Then
q

r
1
(P

1
1
: Q
0
) & r
2
(P

1
3
: f[Q
0
])
and consequently (r
1
, r
2
) P
i
. Since (r
1
, r
2
) has to be compatible with some
element of / we are done.
Claim: Suppose q P

1
1
, i Ci
0
, p P
i
1
are such that p
P
i
1
q
(P

1
1
: P
i
1
). Then p
P
i q (P

1
: P
i
).
Suppose r P
i
is stronger than p. Let r = (r
1
, r
2
) and let r
0
Q
0
witness r P

1
1

id,f
P

1
3
. We may get r
0
Q
0
BA(P
i
1
). Remember that
really we have p (p, ), q (q, ). Since r
0
, r
1
BA(P
i
1
) are compatible
and r
1
p we nd r

1
P

1
1
above r
0
, r
1
and q. Then (r

1
, r
2
) P

1
and it is
a condition stronger than both (r
1
, r
2
) and (q, ). The claim is proved.
Since M
i
1
< M
i
for each i Ci
0
it follows from the above claim that

P
1


P
1

f

P
3
(and Ci
0
is a witness for it). Similarly one can prove

P
3


P
1

f

P
3
.
Lemma 3.8 Suppose

P
0
,

P
1
/,

P
0


P
1
. Let Q
0
, Q
1
be complete
Boolean algebras such that (for k = 0, 1):
BA(P

1
0
) < Q
k
< BA(P

1
1
)
BA(P

1
0
) (Q
k
: BA(P

1
0
)) is the Cohen algebra
Let f : Q
0
Q
1
be an isomorphism such that f[BA(P

1
0
) = id.
Then there exist p / and an automorphism : P

1
onto
P

1
such that

P
1


P and f .
Proof We may apply lemma 3.7 to get that

P
2
=

P
1

f

P
1
/. The
amalgamation over f produces an extension of f there is f
1
: P

1
1
P

1
2
such that f f
1
(we identify p P

1
1
with (, p) P

1
2
). Moreover

P
1
,

P
2
, f
1
satisfy assumptions of lemma 3.7 and thus

P
3
=

P
2

f
1

P
2
/. If we
identify p P

1
2
with (p, )

P
3
we get a partial isomorphism f
2
such that
f
1
f
2
and rng(f
2
) = P

1
2
. And so on, we build

P
m
/ and partial
isomorphisms f
m
such that

P
m


P
m+1
, f
m
f
m+1
and either P

1
m

12
4
4
6


r
e
v
i
s
i
o
n
:
1
9
9
3
-
0
8
-
2
9







m
o
d
i
f
i
e
d
:
1
9
9
3
-
0
8
-
2
9


dom(f
m
) or P

1
m
rng(f
m
). Next we apply lemma 3.3 to conclude that

/ and f

=

m
f
m
: P

onto
P

is the desired automorphism.

Denition 3.9 We dene the following notion of forcing
R =

P / : p H(
2
)

R
is the relation of 3.2.
A notion of forcing P is (
1
+ 1)-strategically closed if the second player
has a winning strategy in the following game of the length
1
+ 1.
For i = 0 Player I gives p
0
P;
Player I gives in the i-th move a dense subset D
i
of P;
Player II gives p
i+1
p
i
, p
i+1
D
i
, for a limit i Player II gives
p
i
above all p
j
(for j < i).
Player II looses if he is not able to give the respective element of P for some
i
1
.
Note that (
1
+1)-strategically closed notions of forcings do not add new

1
-sequences of elements of the ground model.
Proposition 3.10 The forcing notion R is
1
-closed and (
1
+1)-strategically
closed. Consequently forcing with R does not collapse
1
and
2
.
Proof For the rst assertion use lemma 3.3. The second follows from
3.3 and 3.4.
Note that [R[ = 2

1
. Thus if we assume that 2

1
=
2
then forcing with
R does not collapse cardinals.
Suppose V [=GCH.
Let G R be a generic over V. Let P =

P

1
:

P G.
Proposition 3.11 1. P is a ccc notion of forcing.
2. If is a P-name for an
1
-sequence of ordinals then
P the union of all Borel meager sets coded in V[] is meager.
13
4
4
6


r
e
v
i
s
i
o
n
:
1
9
9
3
-
0
8
-
2
9







m
o
d
i
f
i
e
d
:
1
9
9
3
-
0
8
-
2
9


3. The Boolean algebra BA(P) is strongly Cohen-homogeneous.
4. P any union of
1
meager sets is meager.
Proof 1. Work in V. Suppose that

/ is a R-name for an
1
-sequence
of pairwise incompatible elements of P. Let

P R. By proposition 3.10
there is

P
1


P which decides all values of

/. We may assume that all these
elements belong to P

1
1
. A contradiction.
2. Let be a P-name for an
1
-sequence of ordinals. Then is actually
an
1
-sequence of (countable) antichains in P. Therefore V and it is a
P

1
0
-name for some

P
0
G. By density arguments we have that (

P
D
)
D
G
for some

P

P
0
(compare lemma 2.9). Hence
P the union of all Borel meager sets coded in V[G][] is
meager
3. Work in V[G]. Let be a P-name for an
1
-sequence of ordinals. As
in 2. we nd

P
0
G such that is a P

1
0
-name. Suppose now that
BA(P

1
0
) < B < BA(P),
BA(P

1
0
) (B: BA(P

1
0
)) is the Cohen algebra,
f : B BA(P) is a complete embedding such that f[BA(P

1
0
) = id.
Note that B and f are determined by countably many elements. Each element
of BA(P) is a countable union of elements of P. Consequently B, f V and
there is

P
1
G such that B, rng(f) BA(P

1
1
),

P
0


P
1
. By density
argument and lemma 3.8 we nd

P
2
G and f
2
such that

P
1


P
2
and f
2
is an automorphism of BA(P

1
2
) extending f. Similarly, if

P
4
G,

P
3


P
4
and f
3
is an automorphism of BA(P

1
3
) then there are

P
5
G, f
5
such that
f
5
is an automorphism of BA(P

1
5
) extending f
3
.
It follows from the above that, in V [G], we can extend f to an automor-
phism of BA(P).
4. Similar arguments as in 1. and 2.
Theorems 1.2 and 1.1 follow directly from the above proposition and
theorems 2.3 and 2.4.
14
4
4
6


r
e
v
i
s
i
o
n
:
1
9
9
3
-
0
8
-
2
9







m
o
d
i
f
i
e
d
:
1
9
9
3
-
0
8
-
2
9


References
[Jec] T.Jech, Set Theory, Academic Press, New York 1978.
[JS1] H.Judah, S.Shelah, Martins axioms, measurability and equiconsis-
tency results, Journal of Symbolic Logic, 54(1989).
[JS2] H.Judah, S.Shelah, Souslin forcing, Journal of Symbolic Logic,
53(1988).
[JuR] H.Judah, A.Roslanowski, On Shelahs amalgamation, submitted to
Set Theory of Reals, Proceedings of Bar-Ilan Conference in honour of
Prof. Abraham Fraenkel, Bar-Ilan University 1991.
[She] S.Shelah, Can you take Solovays inaccessible away?, Israel Journal
of Mathematics, 48(1984), pp 1:47.
[Sol] R.Solovay, A model of set theory in which every set of reals is
Lebesgue measurable, Annals of Mathematics 92(1970), pp 1:56.
15