Documente Academic
Documente Profesional
Documente Cultură
6
5
5
)
r
e
v
i
s
i
o
n
:
2
0
0
1
-
1
1
-
1
2
m
o
d
i
f
i
e
d
:
2
0
0
1
-
1
1
-
1
2
ITERATION OF -COMPLETE FORCING NOTIONS NOT
COLLAPSING
+
.
ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
Abstract. We look for a parallel to the notion of proper forcing
among -complete forcing notions not collapsing
+
. We suggest such
a denition and prove that it is preserved by suitable iterations.
0. Introduction
This work follows [Sh 587] and [Sh 667] (and see history there), but we
do not rely on those papers. Our goal in this and the previous papers is to
develop a theory parallel to properness in CS iterations for iterations with
larger supports. In [Sh 587], [Sh 667] we have presented parallels to [Sh 64]
and [Sh:98], whereas here we try to have parallels to [Sh 100], [Sh:b, Ch.III],
[Sh:b, Ch.V,5-7] and hopefully [Sh:f, Ch.VI], [Sh:f, Ch.XVIII].
It seems too much to hope for a notion fully parallel to proper among
-complete forcing notions as even for
+
-c.c. -complete there are prob-
lems. We should also remember about ZFC limitations for possible iteration
theorems. For example, if in the denition of the forcing notion Q
in Sec-
tion 3 we demand h
p
e
= o
, and (A
, h
: o
: A
2
and we ask for success of the uniformization on an end segment of A
(for
all such A
: o
P
. In iterations, if
Q = P
, Q
: <
) and p lim(
for
Dom(p).
(4) Ordinal numbers will be denoted be the lower case initial letters of
the Greek alphabet (, , . . .) and also by i, j (with possible sub-
and superscripts).
(5) A bar above a letter denotes that the object considered is a sequence;
usually
X will be X
i
: i < )), where denotes the length of
X.
Often our sequences will be indexed by a set of ordinals, say o ,
and then
X will typically be X
:
o) is a Ddiamond sequence if o D
+
, F
for o, and
(f
)( o : F
f D
+
).
We may also call such
F a (D, o)diamond sequence.
(2) We say that (D, o) has diamonds if there is a (D, o)diamond. We
say that D has diamonds if D is a normal lter on and for every
o D
+
there is a (D, o)diamond.
Denition 0.3. A forcing notion P is complete if every
P
increasing
chain of length less than has an upper bound in P. It is lubcomplete if
every
P
increasing chain of length less than has a least upper bound in
P.
Proposition 0.4. (1) If D is a lter on , then the family of all diago-
nal intersections of members of D constitutes a normal lter (but in
general not necessarily proper). We call this family the normal lter
generated by D.
(2) If P is a complete forcing notion and D is a normal lter on
, then in V
P
the lter D generates a proper normal lter on .
(
6
5
5
)
r
e
v
i
s
i
o
n
:
2
0
0
1
-
1
1
-
1
2
m
o
d
i
f
i
e
d
:
2
0
0
1
-
1
1
-
1
2
ITERATION OF -COMPLETE FORCING NOTIONS 3
[Abusing notation, we will denote this lter also by D or, if we want
to stress that we work in the forcing extension, by D
V[G
P
]
.]
Moreover, by the completeness of P, if X D
+
V, then
P
X
D
+
, and if X V, p
P
X D
V
P
then X D.
(3) If P is a complete forcing notion and
F = F
: o) is a
Ddiamond sequence, then
P
F is a Ddiamond sequence .
Denition 0.5 and Proposition 0.6 below are not central for us, but they
may be used to get somewhat stronger results, see [Sh:F509].
Denition 0.5. Let pr be a denable pairing function on , for example
pr(, ) =
+
+, and let
F = F
: o) be a Ddiamond sequence.
For an ordinal < we let
F
[]
= F
[]
is a
function with domain and such that
F
[]
() =
F
[]
is also a Ddiamond sequence.
Throughout the paper we will assume the following:
Context 0.7. (a) is an uncountable cardinal, =
<
, and
(b) D is a normal lter on (usually D is the club lter T
on ),
(c) o D
+
contains all successor ordinals below , 0 / o, and o
= o
is unbounded in ,
(d) there is a (D, o)diamond sequence.
1. The definitions
In this section we dene a special genericity game, properness over (D, o)
semi diamonds and the class of forcing notions we are interested in.
Denition 1.1. Let P be a forcing notion and let N (H(), , <
) be
such that |N| = , N
<
N and , P, D, o N. Let h : N be
such that the range Rang(h) of the function h includes P N.
(1) We say that
F = F
for o and
() for every
P
increasing sequence p = p
: < ) P N we
have
o : ( < )(h F
() = p
) D
+
.
(2) Let
F be a (D, o)semi diamond. A sequence q = q
: o) N
P is called an (N, h, P)candidate over
F (or: (N, h, P,
F)candidate)
whenever
(
6
5
5
)
r
e
v
i
s
i
o
n
:
2
0
0
1
-
1
1
-
1
2
m
o
d
i
f
i
e
d
:
2
0
0
1
-
1
1
-
1
2
4 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
() for every open dense subset 1 N of P
o : q
1 = o mod D,
and
() if o is a limit ordinal and h F
() : < ) is a
P
i
, r
i
P and a set C
i
D are chosen such that
r
i
N, r
i
r
i
, r r
i
,
(j < i)(r
j
r
i
& r
j
r
i
), and
the generic player chooses r
i
, r
i
, C
i
if i o, and the anti-generic
player chooses r
i
, r
i
, C
i
if i o
.
If at some moment during the play there is no legal move for one of
the players, then the anti-generic player wins. If the play lasted
moves, then the generic player wins the play whenever
() if o
i<
C
i
is a limit ordinal, and h F
() : < ) =
r
: < ), then q
.
(4) Let q be an (N, h, P,
F)candidate,
F a (D, o)semi diamond. A
condition r P is (N, h, P)generic for q over
F if the generic player
has a winning strategy in the game (r, N, h, P,
F, q).
Observation 1.2. (1) In the game (r, N, h, P,
F, q), for each of the play-
ers, if it increases conditions r
i
, r
i
, its choice can only improve its
situation. Making sets C
i
(for i o) smaller can only help the
generic player.
(2) If forcing with P does not add new subsets to , then the game in
Denition 1.1(5) degenerates as without loss of generality r forces
a value to G
P
N; the condition does not degenerate, in fact this
condition (which implies adding no new sequences) is preserved
by (<
+
)support iterations (see [Sh 587]).
(3) Also if o
1
o mod D, o
1
D
+
, then in Denition 1.1 we can
replace o by o
1
. (Again, the generic player can guarantee C
i
o
1
o.)
(4) If P is complete and r is (N, P)generic (in the usual sense, i.e.,
r
P
N[G
P
] V = N ), then both players have always legal moves
in the game (r, N, h, P,
F, q).
Also if the forcing notion P is lubcomplete, then both players
have always legal moves in the game (r, N, h, P,
F, q) (for any r).
Denition 1.3. (1) Let o D
+
. We say that a forcing notion P
is proper over (D, o)semi diamonds whenever (there is a (D, o)
diamond and):
(
6
5
5
)
r
e
v
i
s
i
o
n
:
2
0
0
1
-
1
1
-
1
2
m
o
d
i
f
i
e
d
:
2
0
0
1
-
1
1
-
1
2
ITERATION OF -COMPLETE FORCING NOTIONS 5
(a) P is complete, and
(b) if is large enough, p P and N (H(), , <
), |N| = ,
N
<
N and , p, P, D, o, . . . N, and h : N satises
PN Rang(h), and
F is a (D, o)semi diamond for (N, h, P),
and q = q
: o) is an (N, h, P,
F)candidate,
then there is r P stronger than p and such that r is (N, h, P)
generic for q over
F.
(2) P is said to be proper over Dsemi diamonds if it is proper over
(D, o)semi diamonds for every o D
+
(so D has diamonds). The
family of forcing notions proper over Dsemi diamonds is denoted
K
1
D
.
(3) A forcing notion P is proper over if it is proper over Dsemi dia-
monds for every normal lter D on which has diamonds.
Remark 1.4. Does D matter? Yes, as we may use some large D and be
interested in preserving its largeness properties.
Proposition 1.5. If P is a
+
complete forcing notion, then P is proper
over .
Proof. Straightforward.
Proposition 1.6. (1) If N, P, h are as in 1.1, P is -complete, and
F
is a (D, o)semi diamond, then there is an (N, h, P,
F)candidate.
In fact we can even demand:
(+) if 1 N is an open dense subset of P, then q
1 for every
large enough .
(2) Let r be (N, h, P)generic over
F for some (N, h, P,
F)candidate q.
Then
(a) if r
i
, r
i
, C
i
: i < ) is a result of a play of (r, N, h, P,
F, q) in
which the generic player uses its winning strategy, then
G
= p P N : (i < )(p r
i
)
is a subset of P N generic over N, and
(b) r is (N, P)generic (in the usual sense).
(3) If P is proper over (D, o)semi diamonds, , Y []
, Y V,
then:
(a) forcing with P does not collapse
+
,
(b) forcing with P preserves the following two properties:
(i) Y is a conal subset of []
(under inclusion),
(ii) for every large enough and x H(), there is N
(H(), ) such that |N| = , N
+
+
, N
<
N,
N Y (i.e., the stationarity of Y under the relevant
lter).
Proof. 1) Immediate (by the completeness of P).
2) Clause (a) should be clear (remember 1.1(2)()). For clause (b) note
that 0 o
=
0
for some ordinal
0
/ N, what
guarantees it to win the play.
3) Straightforward from 2).
Very often checking properness over Dsemi diamonds (for particular ex-
amples of forcing notions) we get somewhat stronger properties, which mo-
tivate the following denition.
Denition 1.7. We say that a condition r P is Ngeneric for Dsemi
diamonds if it is (N, h, P)generic for q over
F whenever h, q,
F are as in
1.1. Omitting D we mean for every normal lter D with diamonds.
The following notion is not of main interest in this paper, but surely it is
interesting from the point of view of general theory.
Denition 1.8. Let 0 < <
+
.
(1) Let o D
+
. We say that a forcing notion P is proper over (D, o)
semi diamonds whenever
(a) P is complete, and
(b) if is large enough, p P and
N = N
| = , N
<
,
, p, P,
N N
, and
= F
: o), F
(for < ),
h
: N
, PN
Rang(h
) and h
,
F
: < )
N
, and
, h
, P),
and
q
= q
: o) is an (N
, h
, P)candidate over
F
,
and q
: < ) N
,
then there is r P above p which is (N
, h
, P)generic for q
over
F
D
) and P
is proper over in a way parallel to 1.3(2,3).
Remark 1.9. Note that for = 1 (in Denition 1.8) we get the same notions
as in Denition 1.3.
2. The preservation theorem
In 2.7 below we prove a preservation theorem for our forcing notions. It
immediately gives the consistency of the suitable Forcing Axiom, see 4.1.
Also the proof actually species which semi-diamond sequences
F are used.
First, recall that
(
6
5
5
)
r
e
v
i
s
i
o
n
:
2
0
0
1
-
1
1
-
1
2
m
o
d
i
f
i
e
d
:
2
0
0
1
-
1
1
-
1
2
ITERATION OF -COMPLETE FORCING NOTIONS 7
Proposition 2.1. Suppose that P
, Q
: <
) is a (<
+
)support
iteration such that for each <
P
Q
is complete.
Then the forcing P
is complete.
Before we engage in the proof of the preservation theorem, let us prove
some facts of more general nature than the one of our main context. If, e.g.,
all iterands are lubcomplete, then Proposition 2.3 below is obvious.
Temporary Context 2.2. Let
Q = P
, Q
: <
) be a (<
+
)support
iteration of complete forcing notions. We also suppose that N is a model
as in 1.1,
Q, . . . N.
Proposition 2.3. Suppose that (
is such that
( N)
r is (N, P
)generic.
N P
r).
Then there are conditions s N P
and r
+
P
such that s r
+
, r r
+
and ( < )(s
s).
Proof. Let i
, r
such that
() r
P
i
N is above (in P
i
) of all s
for < ,
() r
P
i
, r
P
i
r
, and r i
,
() if < < cf() then r
and r
.
(The choice is clearly possible as r i
is (N, P
i
)generic.)
Let r
+
P
be an upper bound of r
N.
We let Dom(s) =
Dom(r
+1
)[i
, i
+1
) : < cf(), and for Dom(s),
i
< i
+1
, we let s() be a P
]
(for a generic lter G
P
over V):
(i) If r
+1
()[G
P
] is an upper bound of s
()[G
P
] : < in Q
[G
P
],
then s()[G
P
] = r
+1
()[G
P
].
(ii) If not (i), but s
()[G
P
[G
P
],
then s()[G
P
] is the <
] = s
0
()[G
P
].
It should be clear that s P
N. Now,
s r
+
.
Why? By induction on N we show that s r
+
. Steps = 0
and limit are clear, so suppose that we have proved s r
+
,
i
< i
+1
(and we are interested in the restrictions to + 1). Assume
that G
P
. Since
(
6
5
5
)
r
e
v
i
s
i
o
n
:
2
0
0
1
-
1
1
-
1
2
m
o
d
i
f
i
e
d
:
2
0
0
1
-
1
1
-
1
2
8 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
s
i
+1
r
+1
r
+1
r
+
, we also have s
: < G
P
and
r
+1
G
P
. Hence r
+1
()[G
P
] is an upper bound of s
()[G
P
] : <
. Therefore, s()[G
P
] = r
+1
()[G
P
] r
+1
()[G
P
] r
+
()[G
P
] (see
(i) above) and we are done.
The proof of the proposition will be nished once we show
( < )(s
s).
Why does this hold? By induction on N we show that s
s
for all < . Steps = 0 and limit are as usual clear, so suppose
that we have proved s
s (for < ), i
< i
+1
(and we are
interested in the restrictions to + 1). Assume that G
P
is a generic
lter over V such that s G
P
: < G
P
and therefore s
()[G
P
] : < ) is an increasing
sequence of conditions from the (complete) forcing Q
[G
P
]. Thus this
sequence has an upper bound, and s()[G
P
+1)]
<
is a closed set, 0,
, and
()
1
for every two successive members
<
of the set w, p [
)
is a P
name of an element of P
).
The family of all RSconditions in P
is denoted by P
RS
.
(2) If (p, w) P
RS
and G
P
G
P
whenever
()
2
for every two successive members
<
of the set w,
(p [
))[G
P
] G
P
.
(3) If (p
1
, w
1
), (p
2
, w
2
) P
RS
, then we write (p
1
, w
1
)
(p
2
, w
2
) when-
ever
()
3
for every generic G
P
over V, if (p
2
, w
2
)
G
P
then
(p
1
, w
1
)
G
P
<
of
the set w
1
w
2
we have
(p
1
[
))[G
P
]
P
(p
2
[
))[G
P
].
Remark 2.5. If (p, w) P
RS
<
name for a P
name of a member of Q
there is q P
such that
(p, w)
(q, 0,
).
(2) If (p, w) P
and q P
, then there is q
stronger than q
and such that for each successive members
<
of w the condition
q
decides p [
) (i.e., q
p [
) = p
,
for
some p
,
P
).
(3) Let (p
i
, w
i
) P
RS
N, r P
be
such that
s r and (j < i < )((p
j
, w
j
)
(p
i
, w
i
)
(r, 0,
).
Assume that either r is (N, P
)generic, or
is a limit ordinal of
conality cf(
the condition r is
(N, P
)generic.
Then there are conditions s
N P
and r
such that
s s
, r r
(s
, 0,
)).
Proof. (1), (2) Straightforward (use the completeness of P
).
(3) If r is (N, P
) < ). Let
i
: < cf(
i
= p
i
i
, w
i
= (w
i
i
) i
(clearly
(p
i
, w
i
) P
RS
i
). Since r i
is (N, P
i
)generic, we may inductively pick
conditions s
, r
)) such that
s i
P
i
N, r r
,
(i < )((p
i
, w
i
)
(s
, 0, i
)), s
,
if < < cf(
) then s i
and r
.
Let r
, Q
: <
)
is a (<
+
)support iteration such that for each <
P
Q
= lim(
),
|N| = , N
<
N and , p,
Q, P
, D, o, . . . N, and h : N
satises P
: o) is
a (D, o)semi diamond and q = q
: o) is an (N, h, P
,
F)candidate.
We may assume that for each o
() if h F
() : < ) is not a
P
N,
then h F
for q over
F, let us introduce
notation used later and give two important facts.
Let i N (
+ 1) and let G
P
i
P
i
be generic over V. We dene:
h
i
[G
P
i
] : N[G
P
i
] is such that
if h() is a function, i Dom(h()) and (h())(i) is a P
i
name, then
(h
i
[G
P
i
])() = (h())(i)[G
P
i
], otherwise it is ;
h
[i]
: N is dened by
h
[i]
() = (h()) i provided h() is a function, and otherwise;
o
i
[G
P
i
] = o : if is limit, then q
i G
P
i
;
q
i
[G
P
i
] is q
(i)[G
P
i
] : o
i
[G
P
i
]);
q
[i]
= q
i : o);
F
i
[G
P
i
] is F
: o
i
[G
P
i
]).
Observe that h
[i]
: N is such that P
i
N Rang(h
[i]
) and h
i
[G
P
i
] is
such that N[G
P
i
] Q
i
[G
P
i
] Rang(h
i
[G
P
i
]).
Plainly, by (),
Claim 2.7.1. Assume i N (
+1). Then
F is a (D, o)semi diamond
sequence for (N, h
[i]
, P
i
) and q
[i]
is an (N, h
[i]
, P
i
,
F)candidate.
Claim 2.7.2. Assume that i N(
+1) and r P
i
is (N, h
[i]
, P
i
)generic
for q
[i]
over
F. Let G
P
i
P
i
be a generic lter over V, r G
P
i
. Then in
V[G
P
i
]:
(1) o
i
[G
P
i
] D
+
,
(2)
F
i
[G
P
i
] is a (D, o
i
[G
P
i
])semi diamond for
(N[G
P
i
], h
i
[G
P
i
], Q
i
[G
P
i
]), and
(3) q
i
[G
P
i
] is an (N[G
P
i
], h
i
[G
P
i
], Q
i
[G
P
i
],
F
i
[G
P
i
])candidate.
Proof of the Claim. (1) Will follow from (2).
(2) Assume that this fails. Then we can nd a condition r
P
i
, a P
i
name q
= q
i
, and P
i
names A
for members of D V such that r
P
i
r
G
P
i
and
r
P
i
( o
i
<
A
)(h
i
F
() : < ) , = q
) .
Consider a play r
j
, r
j
, C
j
: i < ) P
i
of the game (r, N, h
[i]
, P
i
,
F, q
[i]
)
in which the generic player uses its winning strategy and the anti-generic
player plays as follows. In addition to keeping the rules of the game, it
makes sure that at stage j o
:
r
j
r
(so r
0
r
for < j.
(
6
5
5
)
r
e
v
i
s
i
o
n
:
2
0
0
1
-
1
1
-
1
2
m
o
d
i
f
i
e
d
:
2
0
0
1
-
1
1
-
1
2
ITERATION OF -COMPLETE FORCING NOTIONS 11
Let A
D V be such that r
j
A
= A
.
Note that the sequence r
j
) : j < ) is
P
i+1
increasing. So, as
D is normal and A
, C
j
D and
F is a semi diamond for (N, h
[i+1]
, P
i+1
)
(by 2.7.1), we may nd a limit ordinal o
<
A
j<
C
j
such that
h
[i+1]
F
(j) : j < ) = r
j
) : j < ). Then also h
[i]
F
(j) : j < ) =
r
j
: j < ), and since the play is won by the generic player, we conclude
q
i r
we have
r
j
o
i
<
A
& h
i
F
() : < ) = q
,
a contradiction.
(3) Should be clear.
Fix a bijection :
. Also let (
i
,
i
) : i < ) list all
pairs (
, ) N such that
, cf() and
is a P
name for an
ordinal.
Next, by induction, we choose a sequence (p
i
, w
i
) : i < ) P
RS
N
such that
(i) (p, 0,
(p
i
, w
i
)
(p
j
, w
j
) for i < j < ,
(ii) if i < j < and () i, then Dom(p
i
) and p
i
() = p
j
(),
(iii) if i < is a limit ordinal, then w
i
is the closure of
j<i
w
j
, and
if, additionally, Dom(q
i
) is such that () i (and i o, of
course), then Dom(p
i
) and p
i
() is such that
() for every generic G
P
G
P
,
and two successive members
of the set w
i
such that
<
we have:
if p
j
()[G
P
][G
P
] : j < i q
i
()[G
P
] has
an upper bound in Q
[G
P
],
then p
i
()[G
P
][G
P
name
N we have:
sup
<
i
: () i (w
i
i
)
< , w
i+1
= w
i
, p
i+1
=
p
i
and
if G
P
G
P
,
then
i
[G
P
i
] =
[G
P
]
(It should be clear that there are no problems in the induction and it is
possible to pick (p
i
, w
i
) as above.) From now on we will treat each p
i
() as
a P
.
Now we are going to dene an (N, h, P
N and for
each i
N
r i r(i) p
(i)
(i) is (N[G
P
i
], h
i
, Q
i
)generic for q
i
over
F
i
.
(
6
5
5
)
r
e
v
i
s
i
o
n
:
2
0
0
1
-
1
1
-
1
2
m
o
d
i
f
i
e
d
:
2
0
0
1
-
1
1
-
1
2
12 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
Main Claim 2.7.3. For every (
,
F, q
[]
).
Proof of the Claim. We will prove the claim by induction on (
+1)N.
For
is (N, h
[
]
, P
)
generic for q
[
]
over
F for all
,
F, q
[]
). The
inductive hypothesis is not used in the full strength in the denition of
the strategy, but we need it in several places, e.g., to know that r is well
dened as well as that we have the st
i
s below. Also note that it implies
that (p
i
, w
i
)
i
= p
i
and w
i
=
(w
i
). Moreover, during the play, both players will always have legal
moves. Why? By the inductive hypothesis we know that r
is (N, P
)
generic for all
)generic
(remember clause (iv) of the choice of the p
i
s!), and thus 1.2(4) applies. If
is a limit ordinal of conality cf() < , then we may use 2.3.
Let st
i
be a P
i
name for the winning strategy of the generic player in
(r(i), N[G
P
i
], h
i
, Q
i
,
F
i
, q
i
), and let
E
0
def
= < : is a limit of points from o
.
Plainly, E
0
is a club of .
Let the generic player play as follows. Aside it will construct sequences
r
(), r
() : j
< , N) and C
() : j
, < , N) so that
r
() is a P
N[G
P
], r
() is a P
name
for a member of Q
, C
() is a P
() : < ), r
(), and
r
() are dened.
So suppose that j
o and r
j
, r
j
, C
j
: j < j
< min(o
(j
0
+ 1)) = j
1
, j
0
o
.
First the generic player picks conditions s
, s P
, s
N such that
r
j
0
s
s, r
j
0
s and for each N we have
s
(i < j
0
)(p
i
() s
()) .
(
6
5
5
)
r
e
v
i
s
i
o
n
:
2
0
0
1
-
1
1
-
1
2
m
o
d
i
f
i
e
d
:
2
0
0
1
-
1
1
-
1
2
ITERATION OF -COMPLETE FORCING NOTIONS 13
[Why possible? By 2.6(3).]
Now the generic player looks at
) = < j
1
. It picks
P
names r
), r
), C
) so that s
forces that
r
), r
),
<
C
) : j
< j
1
)
is a play according to st
in which the moves of the anti-generic player are
determined as follows. First, it keeps the convention that if j
o o
,
then (r
), r
),
<
C
j
0
(
) = s
), r
j
0
(
) = s(
), C
j
0
(
) =
jj
0
C
j
, and next we continue
up to j
1
keeping our convention. If j
0
< j
1
, then the generic player lets
r
0
(
) = s
), r
0
(
) = s(
), C
0
(
) =
jj
0
C
j
and then it plays the
game according to st
up to j
1
keeping our convention for all j
/ o
.
Next, the generic player picks a condition r
and P
names
j
(
)
N (for < j
1
,
< , j
0
< j
< j
1
) such that
r
< j
1
, and
r
P
r
) r
) & r
) =
j
(
) ,
for every j, < j
1
and < j
1
with
), and
r
) ( + 1) = C
) ,
where C
) D V.
Then it lets r
N P
(for j
(j
0
, j
1
)) be conditions such that
Dom(r
) = Dom(s
: < j
1
&
< ,
and for Dom(r
)
r
if () < j
1
and
j
() : j
0
< j < j
1
) is an increasing
sequence of conditions stronger than s
(),
then r
() =
j
(), otherwise r
() = s
() .
Finally, for j
(j
0
, j
1
) it plays r
, r
) : j
, , < j
1
,
< .
Case 2: supi o
: i < j
= j
0
j
< min(o
j
0
) = j
1
, j
0
o.
The generic player proceeds as above, the dierence is that now j
0
belongs
to the generic player, and that it is a limit of moves of the anti-generic
player. Again, we look at
) = < j
1
.
(
6
5
5
)
r
e
v
i
s
i
o
n
:
2
0
0
1
-
1
1
-
1
2
m
o
d
i
f
i
e
d
:
2
0
0
1
-
1
1
-
1
2
14 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
If < j
0
, then every condition in P
(for j < j
0
)
forces that
r
), r
),
<
C
) : j
< j
0
)
is a legal play in which the generic player uses st
. The generic player
determines r
), r
), and C
) for j
[j
0
, j
1
) playing the game
as earlier (with the same convention that if j
o o
, then the j
th
move of the anti-generic player is stipulated as the <
for j < j
0
forces that) r
j
() : j < j
0
), r
j
() : j < j
0
) are increasing, and r
j
(
)
r
j
(
) and r(
) is (N[G
P
], Q
0
(
), r
0
(
)) be the <
j
() r
0
(
) N[G
P
], r
j
(
) r
0
(
). It also lets C
0
(
) =
j<j
0
C
j
.
Then the generic player chooses r
), r
), and C
) for 0 < j
< j
1
playing the game with the strategy st
(and keeping the old convention
for j
/ o
).
Next the generic player picks a condition r
names
j
(
) N and sets C
) D V (for j
, , < j
1
)
as in the previous case. Then it chooses conditions s
NP
and r
+
P
such that r
r
+
and (j < j
0
)(r
j
s
r
+
). [Why possible? If is limit
of conality cf() < , use 2.3; otherwise we know that r is (N, P
)generic.]
Next it denes conditions r
N P
(for j
0
j
< j
1
) so that
Dom(r
) = Dom(s
: < j
1
&
< ,
and for Dom(r
)
r
if () < j
1
and
j
() : j
0
j < j
1
) is an increasing
sequence of conditions above all r
j
() for j < j
0
,
then r
() =
j
(), otherwise r
() = s
() .
Finally, for j
0
j
< j
1
it plays r
, r
+
,
) : j
, , < j
1
,
< .
Why does the strategy described above work? Suppose that r
j
, C
j
: j <
) is a play of the game (r , N, h
[]
, P
,
F, q
[]
) in which the generic player
used this strategy and let r
() : j
< , N) and C
() : j
, <
, N) be the sequences it constructed aside. (As we said earlier,
the game surely lasted steps and thus the sequences described above have
length .)
Let us argue that condition 1.1(3)() holds.
Assume that a limit ordinal o
j<
C
j
(so in particular E
0
) is
such that
()
h
[]
F
() : < ) = r
: < ).
(
6
5
5
)
r
e
v
i
s
i
o
n
:
2
0
0
1
-
1
1
-
1
2
m
o
d
i
f
i
e
d
:
2
0
0
1
-
1
1
-
1
2
ITERATION OF -COMPLETE FORCING NOTIONS 15
We are going to show that q
(i < )(j
< )(p
i
() r
()) .
Now look at the clause (iii) of the choice of the p
) implies that
r
() is an upper bound to q
() p
i
() : i < .
thus, r
() p
() r
forces that r
(), r
(),
<
C
j
() : j < j
1
) is a play
according to st
and
j,<
C
j
() o
).
is dened as follows:
a condition in P
is the inclusion .
Clearly,
Proposition 3.1. P
is lubcomplete and [P
[ =
+
.
But also,
Proposition 3.2. P
is proper over .
Proof. Assume N (H(), , <
) is as in 1.1, and p P
N.
Put j
= N
+
and r = p j
, j
).
Claim 3.2.1. (1) If r
P, r r
, then r
NP
and r
.
(
6
5
5
)
r
e
v
i
s
i
o
n
:
2
0
0
1
-
1
1
-
1
2
m
o
d
i
f
i
e
d
:
2
0
0
1
-
1
1
-
1
2
16 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
(2) If r
P, r r
, and r
N P
, then r
is
stronger than both r
and r
.
(3) If p = p
: <
) P
is increasing and
, and q j
: <
).
Proof of the Claim. (1) By the denition of P
,
Dom(r
) j
() < j
() N.
(2), (3) Should be clear.
Claim 3.2.2. r is Ngeneric for semidiamonds (see 1.7).
Proof of the Claim. Suppose that D is a normal lter on , o D
+
. Let
h : N be such that N P
Rang(h),
F = F
: o) be a (D, o)
semi diamond, and let q = q
: o) be an (N, h, P
,
F)candidate.
We have to show that the condition r is (N, h, P
,
F, q). Note that the set
E
0
def
= < : is a limit of members of o
is a club of (so E
0
D). Now, the strategy that works for the generic
player is the following one:
At stage o of the play, when a sequence r
i
, r
i
, C
i
: i < ) has been
already constructed, the generic player lets C
= E
0
( + 1) and it asks:
() Is there a common upper bound to r
i
: i < q
?
If the answer to () is yes, then the generic player puts Y = r
i
: i <
q
; otherwise it lets Y = r
i
: i < . Now it chooses r
to be the
<
rst element of P
= r
N.
Why the strategy described above is the winning one? Let r
i
, r
i
, C
i
:
i < ) be a play according to this strategy. Suppose that o
i<
C
i
is a
limit ordinal such that h F
() : < ) = r
: < ). So, q
is stronger
than all r
= r
.
Therefore, q
: < q
has an upper
bound. Now look at the choice of r
.
The proposition follows immediately from 3.2.2.
Proposition 3.3. (1) P
K
,s
D
[
F] if and only if < .
Proof. (1) Follows from (2).
(2) Assume < .
(
6
5
5
)
r
e
v
i
s
i
o
n
:
2
0
0
1
-
1
1
-
1
2
m
o
d
i
f
i
e
d
:
2
0
0
1
-
1
1
-
1
2
ITERATION OF -COMPLETE FORCING NOTIONS 17
Let
N = N
: < ), h
: N
and q
be as in 1.8(1b), p P
N
0
.
Let j
= N
+
(for < ) and put r = p (j
, j
) : < . Clearly
r P
and r j
). By the
proof of 3.2, the condition r j
+1
is (N
, h
, P
)generic for q
over
F
To show that P
/ K
,s
D
[
F] for , it is enough to do this for = .
So, pick any
N = N
: < ), h
: N
and q
<
N
.
Let
be a P
P
. We claim that
()
P
( < )(i Dom(
) N
)(
(i) / N
) .
Why? Let p P
. Take
0
< such that Dom(p) N
0
(remember
[p[ < ). If for some i Dom(p) N
0
we have p(i) / N
0
, then
p (i Dom(
) N
0
)(
(i) / N
0
) .
Otherwise, we let
= N
0
+
and
= N
0
+1
+
, and we put q =
p (
). Then clearly q P
) N
0
+1
)(
(i) / N
0
+1
).
It should be clear that () implies that there is no condition r P
which
is (N
, h
, P
)generic for q
are as in 0.7,
(b) o
def
= <
+
: cf() = ,
(c) A
, h
: o
:
(d) A
, otp(A
) = and A
is a club of , and
(e) h
: A
.
The forcing notion Q
is dened as follows:
a condition in Q
is a tuple p = (u
p
, v
p
, e
p
, h
p
) such that
(a) u
p
[
+
]
<
, v
p
[o
]
<
,
(b) e
p
= e
p
: v
p
), where each e
p
,
and e
p
u
p
, and
(c) if
1
<
2
are from v
p
, then
sup(e
2
) >
1
and sup(e
1
) > sup(A
2
1
),
(d) h
p
: u
p
is such that for each v
p
we have
h
p
e
: otp( e
) o
;
the order of Q
is an end-extension of e
p
.
A tuple p = (u
p
, v
p
, e
p
, h
p
) satisfying clauses (a), (b) and (d) above will
be called a pre-condition. Note that every pre-condition can be extended to
a condition in Q
.
(
6
5
5
)
r
e
v
i
s
i
o
n
:
2
0
0
1
-
1
1
-
1
2
m
o
d
i
f
i
e
d
:
2
0
0
1
-
1
1
-
1
2
18 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
Plainly:
Proposition 3.5. The forcing notion Q
is lubcomplete. Also Q
sat-
ises the
+
chain condition.
Proposition 3.6. Q
is proper over .
Proof. Assume N (H(), , <
) is as in 1.1, A
, h
: o
) N and
p Q
, let r N be such
that u
rN
= u N, v
rN
= v N, e
rN
= e
r
N, h
rN
= h
r
N. Note that
r N N. Let us describe the winning strategy of the generic player in the
game (p, N, h, P
,
F, q). For this we rst x a list j
i
: i < of N o
,
and we let E
0
= < : is a limit of members of o .
Suppose that we arrive to a stage o and r
i
, r
i
, C
i
: i < ) is the
sequence played so far. The generic player rst picks a condition r
stronger
than all r
i
s played so far and, if possible, stronger than q
. Then it plays a
condition r
above r
such that
if v
r
, then otp(e
r
) > , and
j
i
: i < v
r
,
and r
= r
N. The set C
N j
i
: i < and otp(A
(max(e
r
)+1)) <
for all v
r
.
Why is this a winning strategy? Let r
i
, r
i
, C
i
: i < ) be a play according
to this strategy, and suppose that o
i<
C
i
is a limit ordinal such that
h F
() : < ) = r
: < ).
Note that then
(i) if
i<
v
r
i
then
i<
e
r
i
is an unbounded subset of A
: otp(
A
) < , and
(ii)
i<
v
r
i
N = j
i
: i < .
We want to show that there there is a common upper bound to r
i
: i <
q
i<
v
r
i
,
if v
q
, then we let e
r
= e
q
, if
i<
v
r
i
v
q
, then
e
r
e
r
i
: i < , v
r
i
the
th
member of A
,
u
r
= u
q
i<
u
r
i
the
th
member of A
: v
r
v
q
,
h
r
h
q
i<
h
r
i
.
(
6
5
5
)
r
e
v
i
s
i
o
n
:
2
0
0
1
-
1
1
-
1
2
m
o
d
i
f
i
e
d
:
2
0
0
1
-
1
1
-
1
2
ITERATION OF -COMPLETE FORCING NOTIONS 19
Why is the choice possible? As / o
and adding
to e
r
(and extending u
r
, h
r
suitably).
Our last example is a natural generalization of the forcing notion D
from
Newelski and Roslanowski [NeRo93]. Let us work in the context of 0.7.
Denition 3.7. (1) A set T
<
is a complete tree if
(a) ( T)( T)( ), and T has the smallest element
called root(T),
(b) (,
<
)(root(T) T T),
(c) if
i
: i < ) T is a increasing chain, < , then there is
T such that
i
for all i < .
Let T
<
be a complete tree.
(2) For T we let succ
T
() = < :
) T.
(3) We let split(T) = T : [succ
T
()[ > 1.
(4) A sequence
is a branch through T if
( < )(lh(root(T)) T).
The set of all branches through T is called lim
(T).
(5) A subset F of the tree T is a front in T if no two distinct members
of F are comparable and
( lim
:
A condition in D
i<
i
split(T).
The order of D
is proper over .
Proof. First let us argue that D
<
T
is a condition in D
<
root(T
). By
clause (c) (for T
N. Choose
inductively complete trees T
N and fronts F
(of T
) such
that
(i) root(T
) = root(T),
(
6
5
5
)
r
e
v
i
s
i
o
n
:
2
0
0
1
-
1
1
-
1
2
m
o
d
i
f
i
e
d
:
2
0
0
1
-
1
1
-
1
2
20 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
(ii) if < , then T
T and F
split(T
), and
(iii) if F
)) = ,
(iv) if is limit, then T
<
T
,
(v) if o is limit and h F
N and
<
hF
() T
, and =
<
root(h
F
()) F
we have
F
+1
and q
(T
+1
)
[]
.
Now we let r =
<
T
.
Claim 3.8.1. r is (N, h, D
,
F, q). Let E
0
be the club of limits of mem-
bers of o
j
, r
j
, C
j
: j < i)
has been already constructed. If i / E
0
then the generic player chooses
r
i
, r
i
D
such that
(A) r
i
j<i
r
j
, r
j<i
r
j<i
T
j
, and r
i
N, r
i
r
i
,
(B) root(r
i
) = root(r
i
) F
(i)
for some (i) > i,
and lets C
i
= E
0
((i) + 1). If i E
0
then the generic player picks r
i
, r
i
satisfying (A) + (B) and such that
(C) if possible, then q
i
and it takes C
i
as earlier.
Why is this a winning strategy? First, as D
() : < ) = r
: < ).
Let =
<
root(r
<
r
is included in T
, for some T
we have F
+1
and q
(T
+1
)
[]
. But
this immediately implies that it was possible to choose r
i
stronger than q
in (C) (remember r =
<
T
).
4. Discussion
4.1. The Axiom. We can derive Forcing Axiom as usual, see [Sh:f, Ch.
VII, VIII]. E.g., if is a supercompact cardinal larger than , then we can
nd a cc complete, proper over Dsemi diamonds forcing notion P of
cardinality such that
(
6
5
5
)
r
e
v
i
s
i
o
n
:
2
0
0
1
-
1
1
-
1
2
m
o
d
i
f
i
e
d
:
2
0
0
1
-
1
1
-
1
2
ITERATION OF -COMPLETE FORCING NOTIONS 21
P
2
=
P collapses every (
+
, ), no other cardinal is collapsed,
in V
P
:
if Q is a forcing notion proper over Dsemi diamonds, 1
are open
dense subsets of Q for <
+
,
then there is a directed set G Q intersecting every 1
(for <
+
).
If we restrict ourselves to [Q[ = , it is enough that is indiscrebable enough.
In ZFC, we have to be more careful concerning Q.
4.2. Future applications. Real applications of the technology developed
here will be given in a forthcoming paper Roslanowski and Shelah [RoSh 777],
where we will present more examples of proper for forcing notions (concen-
trating on the case of inaccessible ). We start there developing a theory par-
allel to that of Roslanowski and Shelah [RoSh 470], [RoSh 628], [RoSh 672]
aiming at generalizing many of the cardinal characteristics of the continuum
to larger cardinals.
4.3. Why our denitions? The main reason why our denitions are (per-
haps) somewhat complicated is that, in addition to ZFC limitations, we
wanted to cover some examples with large creatures (to be presented in
[RoSh 777]). We also wanted to have a real preservation theorem: the (limit
of the) iteration is of the same type as the iterands (though for many appli-
cations the existence of (N, P
is stationary and o
o is also stationary.
One of possible modication of the present denitions for the case of
inaccessible , can be sketched as follows. We have
: o),
= (
)
||
;
q = q
: o) is replaced by q = q
,t
: o, t Par
,
) (where
= N
+
), and
Par = Par
,
: o) V is constant for the iteration
(like D).
In the forcing P: for p = p
j
: j < ), o, t Par
,
, there is an upper
bound q[ p, t] of p (this is a part of P).
(
6
5
5
)
r
e
v
i
s
i
o
n
:
2
0
0
1
-
1
1
-
1
2
m
o
d
i
f
i
e
d
:
2
0
0
1
-
1
1
-
1
2
22 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
For each , each Par
,N
iN
Par
,
has cardinality
= (
)
||
(N
is
of cardinality [[; is the length of the iteration). Having p
j
: j < ) N
we can nd q
t
: t Par
,N
) as in [Sh 587].
Several (more complex) variants of properness over semidiamonds will
be presented in [Sh:F509] and Roslanowski and Shelah [RoSh 777].
References
[AbSh 146] Uri Abraham and Saharon Shelah. Forcing closed unbounded sets. The Jour-
nal of Symbolic Logic, 48:643657, 1983.
[J] Thomas Jech. Set theory. Springer Monographs in Mathematics. Springer-
Verlag, Berlin, 2003. The third millennium edition, revised and expanded.
[NeRo93] Ludomir Newelski and Andrzej Ros lanowski. The ideal determined by the un-
symmetric game. Proceedings of the American Mathematical Society, 117:823
831, 1993.
[RoSh 628] Andrzej Ros lanowski and Saharon Shelah. Norms on possibilities II: More ccc
ideals on 2