Professional Documents
Culture Documents
0
8
r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
6
-
2
2
m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
6
-
2
2
Simple forcing notions and Forcing Axioms
Andrzej Roslanowski
Institute of Mathematics
The Hebrew University of Jerusalem
91904 Jerusalem, Israel
and
Mathematical Institute of Wroclaw University
50384 Wroclaw, Poland
Saharon Shelah
Institute of Mathematics
The Hebrew University of Jerusalem
91904 Jerusalem, Israel
and
Department of Mathematics
Rutgers University
New Brunswick, NJ 08854, USA
October 6, 2003
The research partially supported by Basic Research Foundation of the Israel Academy
of Sciences and Humanities. Publication 508
0
5
0
8
r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
6
-
2
2
m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
6
-
2
2
[RoSh:508] October 6, 2003 1
0 Introduction
In the present paper we are interested in simple forcing notions and Forcing
Axioms. A starting point for our investigations was the article [JR1] in which
several problems were posed. We answer some of those problems here.
In the rst section we deal with the problem of adding Cohen reals by simple
forcing notions. Here we interpret simple as of small size. We try to establish
as weak as possible versions of Martin Axiom sucient to conclude that some
forcing notions of size less than the continuum add a Cohen real. For example
we show that MA(-centered) is enough to cause that every small -linked
forcing notion adds a Cohen real (see 1.2) and MA(Cohen) implies that every
small forcing notion adding an unbounded real adds a Cohen real (see 1.6). A
new almost
) implies that every small forcing notion adding a new real adds an
unbounded real (see 1.13).
In the second section we are interested in AntiMartin Axioms for simple
forcing notions. Here we interpret simple as nicely denable. Our aim is to show
the consistency of AMA for as large as possible class of ccc forcing notions
with large continuum. It has been known that AMA(ccc) implies CH, but it
has been (rightly) expected that restrictions to regular (simple) forcing notions
might help. This is known under large cardinals assumptions and here we try
to eliminate them. We show that it is consistent that the continuum is large
(with no real restrictions) and AMA(projective ccc) holds true (see 2.5).
Lastly, in the third section we study the inuence of MA on
1
3
absoluteness
for some forcing notions. We show that MA
1
(P) implies
1
3
(P)absoluteness
(see 3.2).
Notation: Our notation is rather standard and essentially compatible with that
of [Je] and [BaJu]. However, in forcing considerations we keep the convention
that a stronger condition is the greater one.
For a forcing notion P and a cardinal let MA
,= for
all < .
For a class / of forcing notions the sentence MA
(P);
MA
is the sentence MA
(ccc).
For a forcing notion P, the canonical Pname for the generic lter on P will be
called
P
. The incompatibility relation on P is denoted by
P
(so ,
P
means
compatible).
c stands for the cardinality of the continuum. For a tree T 2
<
, [T] is the
set of all branches through T.
The family of all sets hereditarily of cardinality < (for a regular cardinal )
is denoted by H().
5
0
8
r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
6
-
2
2
m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
6
-
2
2
[RoSh:508] October 6, 2003 2
1 Adding a Cohen real
In this section we obtain several results of the form a (weak) version of MA
implies that small forcing notions (of some type) add Cohen reals. As a con-
sequence we answer Problem 5.3 of [JR1] (see 1.9, 1.10 below).
Proposition 1.1 Suppose P is a forcing notion and
h is a function such that
1. dom(
h) P, rng(
h) 2
<
,
2. if p
1
, p
2
dom(
h), p
1
,
P
p
2
then either
h(p
1
)
h(p
2
) or
h(p
2
)
h(p
1
),
3. if q P then there is
0
2
<
such that
( 2
<
,
0
)(p
dom(
h))(p
,
P
q &
h(p
)).
Then P adds a Cohen real.
Proof Though this is immediate, we present the proof fully for readers
convenience. Let
h : dom(
h) 2
<
be the function given by the assumptions.
Dene a P-name c by
P
c =
h(p) : p dom(
h)
P
.
First note that, by the properties of
h, for every lter G P the set
h(p) : p
dom(
h) G is a chain in (2
<
, ). Hence
P
c 2
.
But really c is a name for a member of 2
0
be such that lh() > m. We nd p
dom(
h) such that p
,
P
q and
h(p
). Thus p
P
h(p
) c, a contradiction.
To show that
P
c is a Cohen real over V
suppose that we have a closed nowhere dense set A 2
and a condition q P
such that
q
P
c A.
Take
0
2
<
given by condition (3) (for q). Since A is nowhere dense we may
choose 2
<
such that
0
and [] A = . By the choice of
0
, there
is a condition p
dom(
h) such that p
,
P
q and
h(p
) (so p
P
c / A), a
contradiction.
5
0
8
r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
6
-
2
2
m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
6
-
2
2
[RoSh:508] October 6, 2003 3
Theorem 1.2 Assume MA
. We are
going to show that (under our assumptions) there exists a function
h as in the
assumptions of proposition 1.1. Since P is -linked there are sets T
n
P such
that
n
T
n
= P and any two members of T
n
are compatible in P (i.e. each
T
n
is linked). Let N be a countable elementary submodel of (H((
7
)
+
), , <
)
such that P, T
n
: n ), . . . N.
We dene a forcing notion R = R(P):
conditions are pairs r = h, w) = h
r
, w
r
) such that
(a) h is a nite function, dom(h) P N, rng(h) 2
<
,
(b) if p
1
, p
2
dom(h) then either p
1
P
p
2
or p
2
P
p
1
or p
1
P
p
2
,
(c) if p
1
, p
2
dom(h), p
1
P
p
2
then h(p
1
) h(p
2
),
(d) w [P]
<
,
the order is such that r
1
R
r
2
if and only if
() h
r1
h
r2
,
() w
r1
w
r2
,
() if q w
r1
, p dom(h
r1
), p, q are compatible in P and no p
dom(h
r1
)
satises p
P
p
, p ,= p
, p
,
P
q
then the set
J
r1,r2
p,q
def
= h
r2
(p
1
) : p p
1
dom(h
r2
) & p
1
,
P
q &
(p
2
)(p
1
< p
2
dom(h
r2
) p
2
P
q)
contains a front of 2
<
above h
r1
(p) (i.e. for every 2
such that
h
r1
(p) there is k with k J
r1,r2
p,q
).
Claim 1.2.1 (R,
R
) is a partial order.
Proof of the claim: The relation
R
is reexive as J
r,r
p,q
= h
r
(p) for all
relevant p, q. For the transitivity suppose that r
1
R
r
2
and r
2
R
r
3
. Clearly
the conditions (), () for the pair r
1
, r
3
are satised. To get condition () note
that if
0
, . . . ,
k1
is a front in 2
<
above and
0
0
, . . . ,
l1
0
is a front in
2
<
above
0
then
0
0
, . . . ,
l1
,
1
, . . . ,
k1
is a front above .
5
0
8
r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
6
-
2
2
m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
6
-
2
2
[RoSh:508] October 6, 2003 4
Claim 1.2.2 R is -centered.
Proof of the claim: Note that if r
1
, r
2
R, h = h
r1
= h
r2
then h, w
r1
w
r2
)
is a common upper bound of r
1
, r
2
.
Claim 1.2.3 Suppose p PN, q P, r
0
R and m . Then the following
sets are dense in R:
1. I
0
p
def
= r R : (q w
r
)[p,
P
q (p
dom(h
r
))(p
P
p
& p
,
P
q)],
2. I
1
q
def
= r R : q w
r
,
3. I
2
r0,m
def
= r R : r
R
r
0
or for every q w
r0
and p dom(h
r0
) such
that p,
P
q & (p
dom(h
r0
))([p
P
p
& p ,= p
] p
P
q)
and for every 2
m
such that h
r0
(p) there is
p
dom(h
r
) with p
P
p
, p
,
P
q and h
r
(p
).
Proof of the claim: 1) Assume p P N, r
0
R. Let q
l
: l < l
) be an
enumeration of q w
r0
: q,
P
p. Choose conditions p
l
(for l < l
) such that
1. p
l
P N
2. for each p
dom(h
r0
) either p
P
p
l
or p
P
p
l
,
3. p
P
p
l
,
4. p
l
: l < l
P
p
+
l
(for l < l
). Next we choose p
++
l
P, p
+
l
P
p
++
l
such that the clauses (2)(4) are satised (remember that P is atomless and
dom(h
r0
) is nite). Let n
l
be such that p
++
l
T
n
l
. As N is an elementary
submodel of (H(
+
7
), , <
) we nd p
l
: l < l
) N such that p
l
T
n
l
and the
clauses (2)(4) are satised. But now we have (1) too. Moreover this sequence
satises (5) since p
l
, p
++
l
T
n
l
and the second condition is stronger than q
l
(remember that the sets T
n
l
are linked).
Dene h
r
by
dom(h
r
) = dom(h
r0
) p
l
: l < l
, h
r0
h
r
and
h
r
(p
l
) =
h
r0
(p
) : p
P
p
l
& p
dom(h
r0
).
First note that all conditions p
dom(h
r0
) satisfying p
P
p
l
are compatible
in P and hence (by (b) for r
0
) they are pairwise comparable and thus (by (c)
for r
0
) the set
h
r0
(p
) : p
P
p & p
dom(h
r0
)
5
0
8
r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
6
-
2
2
m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
6
-
2
2
[RoSh:508] October 6, 2003 5
is a (nite) chain in (2
<
, ). Hence h
r
(p
l
) 2
<
(actually h
r
(p
l
) = h
r0
(p
l
)
for the
P
-maximal p
l
dom(h
r0
) such that p
l
P
p
l
; if there is no such p
l
then h
r
(p
l
) = )). Consequently h
r
satises (a). One can easily check that h
r
satises conditions (b), (c) too and thus r = h
r
, w
r0
) I
0
p
. Conditions (), ()
for the pair r
0
, r are clear. To check the clause () suppose that p
dom(h
r0
),
q w
r0
are relevant for it. If for each l < l
either p
l
P
q or p
P
p
l
(which
implies p
P
p
l
) then J
r0,r
p
,q
= h
r0
(p
) as the property of p
there is preserved.
Otherwise J
r0,r
p
,q
= h
r
(p
l
) : l < l
, p
l
,
P
q, p
P
p
l
. But due to condition (b)
for r
0
we have that each condition from dom(h
r0
) weaker than any p
l
such that
p
l
,
P
q, p
P
p
l
is weaker than p
. Consequently h
r
(p
l
) = h
r0
(p
dom(h
r
) with p <
P
p
, q,
P
p
.
(It is possible that l
l
such that
1. p
l
P N,
2. for each p dom(h
r
) either p
P
p
l
or p
P
p
l
,
3. p
l
P
p
l
,
4. p
l
: l < l
l
,
P
q
l
.
For this we follow exactly the lines of the respective part of the proof of 1) (so
this is another place we use the assumptions on P).
Next dene h
r1
= h
r
(p
l
,
l
) : l < l
, w
r1
= w
r
, r
1
= h
r1
, w
r1
).
Similarly as in 1) one checks that r
1
R.
The condition r
1
is stronger than r: clauses (), () are clear. For ()
suppose that q w
r
, p dom(h
r
) are relevant for this clause. If m lh(h
r
(p))
then each 2
m
extending h
r
(p) appears as
l
= h
r1
(p
l
) for some l < l
such
that q
l
= q, p
l
= p. Hence J
r,r1
p,q
contains a front above h
r
(p). If m < lh(h
r
(p))
then the pair (p, q) does not appear as (p
l
, q
l
). Note that for each l < l
, if q
l
,
P
p
then p
l
P
p (as p
l
cannot be stronger than p since lh(h
r1
(p
l
)) m) and hence
p
P
p. If q
l
P
p then we get the same conclusion (though p
l
might be weaker
than p, the demands (5), (2) of the choice of p
l
imply that p
P
p). Consequently
the maximality property of p is preserved in dom(h
r1
) and J
r,r1
p,q
= h
r
(p).
5
0
8
r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
6
-
2
2
m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
6
-
2
2
[RoSh:508] October 6, 2003 6
To prove that r
1
I
2
r0,m
suppose that q w
r0
and p dom(h
r0
) is maximal
(in dom(h
r0
)) compatible with q. Let 2
m
extend h
r0
(p). Since r
0
R
r
we nd p
dom(h
r
) stronger than p, maximal (in dom(h
r
)) compatible with
q and such that , h
r
(p
) then
we are done. So suppose h
r
(p
we have q = q
l
,
p
= p
l
and =
l
. By the choice of p
l
and the denition of h
r1
(p
l
) we get
p
P
p
P
p
l
& p
l
,
P
q
l
= q & =
l
= h
r1
(p
l
).
The claim is proved.
Since we have assumed MA
dom(h
r0
) such that
p
P
p
& p
,
P
q & (p
dom(h
0
))([p
P
p
& p
,= p
] p
P
q).
Assume that h
r0
(p
) 2
m
.
By (
2
) we nd r H I
2
r0,m
. As r
0
, r H, H is a lter, we cannot have
r
R
r
0
. Consequently the second part of the denition of I
2
r0,m
applies to r.
Looking at this denition (with p
dom(h
r
)
with
p
P
p
& p
,
P
q & h
r
(p
).
So
0
=
h(p
(R(P)) implies
P adds a Cohen real, provided P is atomless linked of size .
Corollary 1.4 Assume MA
over V
(b) there exists a sequence /
n
: n ) of maximal antichains of P such that
1. /
0
= 0,
2. (n )(p /
n+1
)(q /
n
)(q
P
p),
3. (n )(p /
n
)(|q /
n+1
: p
P
q| = ),
4. (q P)(n )(|p /
n
: p,
P
q| = ).
Proof Easy, left for the reader.
Theorem 1.6 Suppose that P is a ccc forcing notion such that |P| < cov(/)
(i.e. unions of |P| many meager sets are meager) and
P
there is an unbounded real over V.
Then
P
there is a Cohen real over V.
Proof We are going to apply proposition 1.1 and for this we will construct
a function
h satisfying (1)(3) of 1.1.
Let /
n
: n ) be a sequence of maximal antichains of P given by (b) of
proposition 1.5. Take a countable elementary submodel N of (H(
+
7
), , <
)
such that P, /
n
: n ), . . . N. Consider the following partial order:
conditions are nite functions h such that
a. dom(h)
n
/
n
, rng(h) 2
<
,
b. if p
1
, p
2
dom(h), p
1
P
p
2
then h(p
1
) h(p
2
),
the order is the inclusion; h
1
C
h
2
i h
1
h
2
.
Clearly C is (isomorphic to) the Cohen forcing notion.
Claim 1.6.1 Let p
n
/
n
, h
0
C, q P, m . Then the following sets
are dense in C:
1. J
0
p
def
= h C : p dom(h),
2. J
1
q,m,h0
def
= h C : h
C
h
0
or for every p dom(h
0
) such that for some
n , p /
n
and the set p
/
n+1
: p
P
p
& p
,
P
q is
innite we have: for every 2
m
extending h
0
(p)
there is p
dom(h) with p
P
p
, p
,
P
q and h(p
) =
5
0
8
r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
6
-
2
2
m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
6
-
2
2
[RoSh:508] October 6, 2003 8
Proof of the claim: 1) Assume p
n
/
n
, h C. Extend h to h
by putting
h
(p) =
h(p
) : p
dom(h) & p
P
p.
Easily this h
satises h
J
0
p
, h
C
h
.
2) Suppose that q P, m , h
0
C, h C. We may assume that h
0
C
h.
Let (p
l
, n
l
,
l
) : l < l
/
n+1
: p
P
p
& p
,
P
q is innite
() h
0
(p) .
Next (using () above) choose p
l
/
n
l
+1
such that
1. p
l
P
p
l
2. p
l
: l < l
either p
P
p
l
or p
P
p
l
.
Now put dom(h
) = dom(h) p
l
: l < l
, h
(p
l
) =
l
and h
dom(h) = h.
Easily h
C, h
C
h
and h
J
1
q,m,h0
. This nishes the claim.
Since |P| < cov(/) we nd a lter H C such that H J
0
p
,= and
H J
1
q,m,h0
,= for all q P, m , h
0
C and p
n
/
n
. Put
h =
H.
Then clearly
h :
n
/
n
2
<
is a function satisfying the requirements (1),
(2) of 1.1. To check the third condition there suppose q P. Take n and
p
/
n
such that the set p
/
n+1
: p
P
p
& p
,
P
q is innite (possible
by the choice of the /
k
s). Since H J
0
p
,= we nd a condition h
0
H
such that p
dom(h
0
). Suppose that 2
<
,
h(p
dom(h
0
),
h(p
) 2
m
we nd p
dom(h
1
) such that p
, p
,
P
q and h
1
(p
) = .
Since h
1
(p
) =
h(p
) we conclude that
0
=
h(p
-bounding if
for each P-name
f for an element of
V:
(p
P
p)(p
P
(
n X)(
f(n) < g(n))).
5
0
8
r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
6
-
2
2
m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
6
-
2
2
[RoSh:508] October 6, 2003 9
Lemma 1.8 1. Suppose that P is a ccc forcing notion such that for every
integer n the product forcing notion P
n
does not add unbounded real and
satises the ccc. Then the -product P
-
bounding and satises the ccc.
2. Finite support iteration of ccc almost
is almost
-bounding. Let
f be a P
and g
n
k
: /
n
k
such that
(q /
n
k
)(
n
k
(q)n = q &
n
k
(q)
P
f(k) = g
n
k
(q))
(possible as P
n
< P
dened by
(k )(q /
n
k
)(q
P
n g
n
(k) = g
n
k
(q)).
Since P
n
does not add unbounded reals and satises the ccc we nd a function
g
n
such that
P
n (m )(k m)( g
n
(k) < g
n
(k)).
Take g
P
(X []
V)(
k X)(
f(k) < g(k)).
To this end suppose that X []
, p P
P
n
stronger
than p and an integer m
0
such that p
P
n (k m
0
)( g
n
(k) < g
n
(k)). By the
choice of g we nd m
1
such that (k m
1
)(g
n
(k) < g(k)). Let k X be
such that k > m
0
+ m
1
+ N. Since /
n
k
is a maximal antichain of P
n
we may
take a condition q /
n
k
compatible with p
. Let p
and
n
k
(q) in P
. Then (p
f(k) = g
n
k
(q) = g
n
(k) < g
n
(k) < g(k)
(remember k is above m
0
, m
1
). Since k X is greater than N we nish by
standard density arguments.
2) See [Sh:f, Ch VI, 3.6+3.17] or [BaJu, 6.5.3].
Theorem 1.9 Assume MA
(ccc &
(ccc &
P
n there is an unbounded real over V.
Clearly n > 1. By proposition 1.5 we nd maximal antichains /
k
P
n
(for
k < ) satisfying conditions (1)(4) of clause (b) there.
We may think that P is an ordering on . Let N be an elementary submodel
of (H(
+
7
), , <
) such that
P,
P
, /
k
: k ), . . . N, + 1 N and |N| = .
Let : N M be the Mostowski collapse of N, M a transitive set. Note
that (P) = P, (/
k
) = /
k
etc. Since P
n1
is ccc and adds no unbounded real
we may apply our restricted version of MA
/
H
k
: p,
P
p
| < )
and thus
(k )(|( p
, p
) /
k
: p
H & p,
P
p
| < ).
Since P
n1
adds no unbounded real (and this is true in M too) we nd nite
sets A
k
/
k
(for k ) and a condition p P
n1
such that for each k
M [= p
P
n1 ( p
, p
) /
k
: p
P
n1 & p,
P
p
A
k
.
5
0
8
r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
6
-
2
2
m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
6
-
2
2
[RoSh:508] October 6, 2003 11
This means that if ( p
, p
) /
k
A
k
then either p
P
n1 p
or p
P
p
. Hence the
condition ( p, p) P
n
is a counterexample to the fourth property of /
k
: k ).
The claim is proved.
It follows from 1.9.1 and 1.8 that (under our assumptions) P is -centered.
So now we may use the following claim.
Claim 1.9.2 Every -centered atomless forcing notion adds an unbounded real.
Proof of the claim: Folklore; see e.g. 5.2 of [JR1].
Corollary 1.10 It is consistent that c >
1
, every atomless ccc forcing notion
of the size < c adds a Cohen real but MA
1
(ccc) fails.
As we saw in 1.6, if we assume a small part of MA
:
conditions are pairs a, w) such that w [2
]
<
and a [2
<
]
<
,
the order is dened by: a
0
, w
0
)
Q
a
1
, w
1
) if and only if
a
0
a
1
, w
0
w
1
and ( w
0
)(l )( l a
1
l a
0
).
Lemma 1.12 1. Q
is an almost
Q
A =
a : (w)(a, w)
Q
).
Then
()
Q
( 2
V)(
n )(n /
A)
()
Q
if T 2
<
is a perfect tree from the ground model
then (
n )(T 2
n
A ,= ).
Proof 1) Clearly if a
0
= a
1
, a
0
, w
0
), a
1
, w
1
) Q
then a
0
, w
0
w
1
) Q
is -
centered. Next note that
5
0
8
r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
6
-
2
2
m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
6
-
2
2
[RoSh:508] October 6, 2003 12
a
0
, w
0
)
Q
a
1
, w
1
) if and only if
either there are w
0
, l such that l a
1
a
0
or the symmetrical condition holds (interchanging 0 and 1).
Hence if a
0
, a
1
2
l
0
, l
0
: w
1
= l
0
: w
2
then
a
0
, w
0
)
Q
a
1
, w
1
) i a
0
, w
0
)
Q
a
1
, w
2
).
Since the product space (2
)
n
is compact we may conclude that
if / Q
is a maximal antichain, n , a [2
<
]
<
then there is a nite set A = A
a,n
/ such that for every w 2
,
|w| = n there is r A with a, w),
Q
r.
The above property easily implies that Q
is almost
h is a Q
-
bounding) suppose that X []
. Assume that
r
0
Q
(
n X)(g(n)
h(n)),
so we have r
1
and k such that
r
1
Q
(n > k)(n X g(n)
h(n)).
Now take k
X such that k
> k and if r
1
= a, w) then a 2
k
, |w| =
n k
. By the denition of A
a,n,k
we nd r A
a,n,k
compatible with r
1
.
But each member of A
a,n,k
forces that
h(k
) < g(k
), a contradiction.
2) Straightforward.
Theorem 1.13 Assume MA
(Q
V ,= 2
such that
P
r / V. For a condition q P let
T
q
def
= 2
<
: q ,
P
r.
By our assumptions on r we know that each T
q
is a perfect tree in 2
<
. Next
x
q
[T
q
] (for q P). Since we have assumed MA
(Q
) we may apply
lemma 1.12 to nd a set A 2
<
such that for each q P:
5
0
8
r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
6
-
2
2
m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
6
-
2
2
[RoSh:508] October 6, 2003 13
() (
n )(
q
n / A) and
() (
n )(T
q
2
n
A ,= ).
Now dene a P-name
K for a subset of by:
P
K = n : rn A.
First note that
K is a P-name for an innite subset of : Why? Suppose that
q P and N . By the property () of A we nd A T
q
such that
lh() > N. Then we have a condition p
P
lh()
K.
Suppose now that q P, g
P
the increasing enumeration of
K is an unbounded real over V
nishing the proof.
Remark 1.14 The forcing notion Q
is a
Borel ccc forcing notion). Hence we may put together 1.6 and 1.13 and we get
the following corollary.
Corollary 1.15 Assume MA
(Q
(P) be the
following sentence:
there exists a sequence G
i
: i < ) of lters on P such that for every
maximal antichain / P for some i
0
< we have
(i i
0
)(G
i
/ ,= ).
For a class / of forcing notions the axiom AMA
P1
(H(
1
)
V[
P
1
P0]
, )
n
(H(
1
), ).
Instead of < we may write <
0
.
Denition 2.3 Let be a cardinal number.
1. (
of (
(for n ):
(
0
= (
,
(
n+1
(
n
P < P
P <
n+1
P
,
(
=
n<
(
n
.
Lemma 2.4 Let be a cardinal such that
= , n .
1. If P
0
, P
1
(
n
, P
0
< P
1
then P
0
<
n
P
1
.
5
0
8
r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
6
-
2
2
m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
6
-
2
2
[RoSh:508] October 6, 2003 15
2. Suppose <
+
, cf()
1
and P
i
(
n
def
=
i<
P
i
(
n
and if P (
, P
i
<
n
P
for every i < then P
<
n
P.
3. If P (
then there is P
(
n
.
4. If P (
, P
i
<
n+1
P for each i < . By the inductive hypothesis
we know that P
<
n
P, P
(
n
) we have
P
i
<
n+1
P
]
, ) and
() (H(
1
)
V[GPi]
, )
n+1
(H(
1
)
V[G]
, ).
Let (x, y) be a
n
-formula and y
0
H(
1
)
V[GP
]
. Take i < such that
y
0
H(
1
)
V[GPi]
(remember cf() > ). If (H(
V[G]
1
), ) [= x(x, y
0
) then
(H(
V[GPi]
1
), ) [= x(x, y
0
) (by ()) and H(
V[GP
]
1
), ) [= x(x, y
0
) (by
()). This shows P
<
n+1
P. To prove that P
(
n+1
suppose that P (
n
,
P
and consequently
P
i
<
n+1
P. By the previous part we get P
<
n+1
P nishing 2.
3) Let P (
, P < P
0
, P
i
< P
j
, P
=
i<
P
i
,
2. p
i
P
i
,
3.
i
is a
n
-formula,
i
is a P
i
-name for a nite sequence of elements of
H(
1
),
5
0
8
r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
6
-
2
2
m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
6
-
2
2
[RoSh:508] October 6, 2003 16
4. (p
i
,
i
,
i
) : i < ) lists all triples (p, , ) such that = (x, y) is a
n
-formula, is a (canonical) P
,
5. if i is limit, cf(i) = then P
i
(
n
is such that
<i
P
< P
i
,
6. if i is limit, cf(i) > then P
i
=
<i
P
(
n
,
7. if there is P
(
n
such that P
i
< P
we have
p
,
P
p
i
and
p
P
(H(
1
), ) [= x
i
(x,
i
)
then P
i+1
is an example of such P
.
The construction is fully described by the above conditions (and easy to carry
out; remember about the inductive hypothesis and the assumption that
= ).
Clearly P
(
n
. We have to
show that actually P
(
n+1
(
n
such that
P
< P
but P
,<
n+1
P
and a
n
-formula and a P
P
(H(
1
), ) [= x(x, ) but (H(
1
)
V[
P
P]
, ) [= x(x, ).
Take p P
such that p
,
P
p and there is no condition p
such that
p
P
p
and p
P
p
Pi+1
(H(
1
), ) [= x(x, ).
Since P
i+1
<
n
P
P
(H(
1
), ) [= x(x, ).
The choice of p implies p
+
,
P
p
P
(H(
1
)
V[P]
, ) [= x(x, ).
4) Let F
0
k
:
i<
P P (for k ) be functions such that if Q P, Q is
closed under all F
0
k
then Q <
n
P (they are given by the inductive hypothesis 4.).
Let A
i,j,k
0 be disjoint innite sets (for i, j, k ). For a
n
-formula
(x, y
0
, . . . , y
1
) and m we choose a function F
,m
:
i<
P P satisfying
the condition described below.
5
0
8
r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
6
-
2
2
m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
6
-
2
2
[RoSh:508] October 6, 2003 17
Let p
m
: m < ) P. For k < we try to dene a P-name
k
for a real in
by
(m A
i,j,k
)(p
m
P
k
(i) = j).
If this denition is correct then we ask if these reals encode (in the canonical
way) elements of H(
1
) (which we identify with the names
k
themselves). If
yes then we ask if
p
0
P
(H(
1
), ) [= x(x,
0
, . . . ,
1
).
If the answer is positive then we x a P-name for (a real encoding) a member
of H(
1
) such that
p
0
P
(H(
1
), ) [= ( ,
0
, . . . ,
1
).
This name can be represented similarly as names
k
(for k < ) so we have a
sequence q
m
: m < ) P encoding it. Finally we want F
,m
to be such that
if the above procedure for p
m
: m < ) works then F
,m
(p
m
: m < ) = q
m
.
Now take all the functions F
0
k
, F
,m
; it is easy to check that they work.
Lastly note that the case n = follows immediately from the lemma for
n < . (For 3. construct an increasing sequence P
i
: i <
1
) such that P < P
0
and if <
1
is limit, k < then P
+k
(
k
.)
Theorem 2.5 Suppose that , are cardinals such that
1
= cf()
=
P
c = & AMA
(projective ccc).
Proof The forcing notion P which we are going to construct will be es-
sentially a nite support iteration of length of ccc forcing notions. One
could try to force with all possible ccc orders in the iteration. However some
care is necessary to make sure that several notions (including being a max-
imal antichain) are suciently absolute for intermediate stages. Therefore
we use forcing notions from the class (
, P
=
i<
P
i
(
,
2. P
i
< P
j
,
3. (
i
,
i
,
i
) : i < ) lists with conal repetitions all triples (, , ) such
that is a formula with n+1 variables, is a formula with n+2 variables
and is a P
Pi
i
(x,
i
),
i
(x
0
, x
1
,
i
)) denes in (H(
1
), ) a ccc partial order
Q
i
then P
i
Q
i
< P
i+1
.
It is easy to carry the construction (use a book-keeping argument, remembering
Q
= (x
0
, x
1
) H(
1
) H(
1
) : (H(
1
), ) [= (x
0
, x
1
, r).
Let be a P-name for r. We may assume that
P
(x, ), (x
0
, x
1
, )) denes (in (H(
1
), )) a ccc partial order.
There is an increasing conal in sequence i
j
: j < ) such that is a
P
i0
-name and (
ij
,
ij
,
ij
) = (, , ). Since P, P
ij
(
we have that
P
(H(
1
)
V[
P
Pi
j
]
, ) (H(
1
), )
and hence the formulas (x, r), (x
0
, x
1
, r)) dene (in (H(
1
)
V[GPi
j
]
, )) the
partial order QH(
1
)
V[GPi
j
]
. The incompatibility relation in this partial or-
der is expressible in (H(
1
), ) and thus it is the restriction of
Q
. Consequently
QH(
1
)
V[GPi
j
]
is ccc in V[G] and hence in V[GP
ij
]. Hence in V[GP
ij+1
]
we have a lter G
j
QH(
1
)
V[GPi
j
]
generic over V[GP
ij
] (here we apply
condition 4 of the construction). Look at the sequence G
j
: j < ). Let / Q
be a maximal antichain. It is countable and hence for suciently large j <
we have / V[GP
ij
]. Moreover the antichain can be coded as a one real and
the fact that it is a maximal antichain in the partial order dened by , ) is
expressible in (H(
1
), ). Applying P
ij
(
we get that
V[G P
ij
] [= / is a maximal antichain in Q H(
1
)
V[GPi
j
]
.
Consequently for suciently large j < we have
G
j
/ ,= .
This nishes the proof.
5
0
8
r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
6
-
2
2
m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
6
-
2
2
[RoSh:508] October 6, 2003 19
Remark 2.6 In 2.4 and 2.5 we used H(
1
) as we were mainly interested in
AMA
1
and projective ccc forcing notions. But we may replace it by H() for
any uncountable regular cardinal such that
<
||
= . Then in 2.4(2) we
consider <
+
such that cf() and in 2.5 we additionally assume that
.
3 Absoluteness and embeddings
In this section we answer positively Problem 4.4 of [JR1] (see 3.2) and we give
a negative answer to Problem 3.3 of [JR1] (see 3.5).
Denition 3.1 Let P be a forcing notion. We say that
1
n
(P)-absoluteness
holds if for every
1
n
formula (with parameters in V) and a generic lter
G P over V
V[G] [= if and only if V [= .
Obviously
1
2
(P)-absoluteness holds for any forcing notion P.
Theorem 3.2 Assume MA
1
(P). Then
1
3
(P)absoluteness holds.
Proof Suppose that is a
1
3
sentence (with a parameter a
). Using
the tree representation of
1
2
sets we nd a tree T (constructible from a) over
1
such that
(x
)(f
1
)(n )(x n, f n) / T)
(x
P
(x
such that
P
the tree T( r) is well founded.
Consequently we have a P-name for a function such that
P
: T( r) Ord is a rank function.
For n ,
1
n
put
J
0
n
= p P : (m )(p
P
r(n) = m),
5
0
8
r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
6
-
2
2
m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
6
-
2
2
[RoSh:508] October 6, 2003 20
J
1
= p P : either p
P
/ T( r)
or ( Ord)(p
P
T( r) & ( r n, ) = ).
Clearly these are dense subsets of P. By MA
1
(P) we nd a lter G on P such
that G J
0
n
,= for n and G J
1
,= for
1
<
. Using this lter we
may interpret the name r to get r = r
G
J
1
G then p
,
P
/ T( r) and hence for some ordinal we have
p
P
T( r) & ( r n, ) = . Moreover if
0
,
1
are such that for some
p
0
, p
1
G we have
p
i
P
T( r) & ( r n, ) =
i
then (as p
0
,
P
p
1
) we cannot have
0
,=
1
.]
We claim that is a rank function on T(r). Suppose that n
0
< n
1
,
0
1
n
0
,
1
1
n
1
,
0
1
and r n
0
,
0
), r n
1
,
1
) T. Take a condition
p G
m<n1
J
0
m
. Then
p
P
r n
1
= r n
1
&
0
,
1
T( r).
Next choose conditions p
0
, p
1
G such that
p
i
P
i
T( r) & ( r n
i
,
i
) = (
i
).
Take p
G stronger than p
0
, p
1
, p. Since is (forced to be) a rank function on
T( r) we have
p
P
(
0
) = ( r n
0
,
0
) > ( r n
1
,
1
) = (
1
).
Hence (
0
) > (
1
) and we may conclude our theorem: the tree T(r) is well
founded so V [= .
Proposition 3.3 Suppose that Q is a ccc Souslin forcing notion (i.e. Q,
Q
and
Q
are
1
1
sets), r is a Qname for a function from 2
<
to 2. Let
A[ r]
def
= 2
: (p Q)(
m )(p
Q
r( m) = 1).
Then A[ r] is an analytic set.
5
0
8
r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
6
-
2
2
m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
6
-
2
2
[RoSh:508] October 6, 2003 21
Proof For each 2
<
choose a maximal antichain p
,l
: l ) in Q
and a set I
p
,l
Q
r() = 0 and l / I
p
,l
Q
r() = 1.
Now note that for each 2
we have
A[ r] (p Q)(
n )(l I
n
)(p
n,l
Q
p).
Proposition 3.4 For every A 2
Q
A r =
r : (w)(r, w)
Q
A).
It should be clear that Q
A
is -centered, |Q
A
| = |A| +
0
and
Q
A r : 2
<
2.
Moreover for each 2
and r, w, ) Q
A
:
(
m)(r, w)
Q
A r( m) = 1) i w.
Consequently A = A[ r].
Corollary 3.5 If A 2
1
, Pacic Journal of Mathematics, vol.115(1984): 399-407.
[Sh:480] Shelah, Saharon, How special are Cohen and Random forcings, Israel
Journal of Mathematics, vol.88(1994): 153174.
[Sh:f] Shelah, Saharon, Proper and improper forcing, Perspectives in
Mathematical Logic, Springer, accepted.
[To] Todorcevic, Stevo, Remarks on Martins Axiom and the Continuum
Hypothesis, Canadian Journal of Mathematics, vol.43 (1991): 832
851.