Anda di halaman 1dari 16

(

7
8
4
)


r
e
v
i
s
i
o
n
:
2
0
0
4
-
0
3
-
0
1







m
o
d
i
f
i
e
d
:
2
0
0
4
-
0
3
-
0
2


FORCING AXIOM
FAILURE FOR ANY >
1
SH784
Saharon Shelah
Institute of Mathematics
The Hebrew University
Jerusalem, Israel
Rutgers University
Mathematics Department
New Brunswick, NJ USA
Annotated Content
1 A forcing axiom for >
1
fails
[The forcing axiom is: if P is a forcing notion preserving stationary subsets
of any regular uncountable and I
i
is dense open subset of P for i <
then some directed G P meets every I
i
.
We prove (in ZFC) that it fails for every regular >
1
. In our coun-
terexample the forcing notion P adds no new sequence of ordinals of length
< ).
2 There are
1
-semi-proper forcing notions
I would like to thank Alice Leonhardt for the beautiful typing.
Publ. 784 - Revised with proofreading for the Journal
Latest version - 04/Mar/2
Typeset by A
M
S-T
E
X
1
(
7
8
4
)


r
e
v
i
s
i
o
n
:
2
0
0
4
-
0
3
-
0
1







m
o
d
i
f
i
e
d
:
2
0
0
4
-
0
3
-
0
2


2 SAHARON SHELAH
1 A forcing axiom for >
0
fail
David Aspero asks on the possibility of, see Denition below, the forcing axiom
FA(K,
2
) for the case K = the class of forcing notions preserving stationarily of
subsets of
1
and of
2
. We answer negatively for any regular >
1
(even
demanding adding no new sequence of ordinals of length < ), see 1.16 below.
1.1 Denition. 1) Let FA(K, ), the -forcing axiom for K mean that K is a family
of forcing notions and for any P K and dense open sets J
i
P for i < there is
a directed G P meeting every J
i
.
2) If K = P we may write P instead of K.
1.2 Denition. Let be regular uncountable. We dene a forcing notion P = P
2

as follows:
(A) if p P i p = (,

S,

W) = (
p
,

S
p
,

C
p
) satisfying
(a) <
(b)

S
p
= S

: ) = S
p

: )
(c)

C
p
= C

: ) = C
p

: )
such that
(d) S

is a stationary subset of consisting of limit ordinals


(e) C

is a closed subset of
(f) if is a limit ordinal then C

is a closed unbounded subset of


(g) if C

then C

= C

(h) C

=
(i) for every and C

we have S

= S

(B) order: natural


p q i
p

q
,

S
p
=

S
q
(
p
+ 1) and

C
p
=

C
q
(
p
+ 1).
1.3 Observation: 1) P
2

is a (non empty) forcing notion of cardinality 2

.
2) J
i
= p P
2

:
p
i is dense open for any i < .
Proof. 1) Obvious.
2) Given p P
2

if
p
i we are done. So assume
p
< i and for (
p
, i]
let S
q

be S

for any stationary subset S

of < : > i a limit ordinal


which does not belong to S
p

:
p
and let C
q

= j :
p
< j < and
(
7
8
4
)


r
e
v
i
s
i
o
n
:
2
0
0
4
-
0
3
-
0
1







m
o
d
i
f
i
e
d
:
2
0
0
4
-
0
3
-
0
2


FORCING AXIOM FAILURE FOR ANY >
1
SH784 3
q = (i,

S
p
S
q

: (
p
, i]),

C
p
C
q

: (
p
, i])).
It is easy to check that p q P
2

and q J
i
.
1.3
1.4 Claim. Let = cf() be regular uncountable and P = P
2

. For any stationary


S and P
2

-name f

of a function from

to the ordinals or just to V and


p P there are q, such that:
(i) p q P
(ii)
q
= + 1
(iii) S if

=
(iv) q forces a value to f

)
(v) if <

and
P
Rang(f

) then q
P
f

() < .
Proof. Without loss of generality S is a set of limit ordinals. We prove this by induc-
tion on

, so without loss of generality

= [

[ and without loss of generality

<

= cf(

), but if

< the set S is immaterial so without loss of generality


< & S cf()

.
Let be large enough (e.g. = (
3
())
+
), <

is a well ordering of H () and


choose

N = N
i
: i < ) such that
(a) N
i
(H (), , <

) is increasing continuous
(b) , p, f

, S belongs to N
i
hence P N
i
(c) |N
i
| <
(d) N
i

(e) N
j
: j i) belong to N
i+1
; hence i N
i
so N
i
: i < .
Let
i
= N
i
, and let i() = Mini : i < is a limit ordinal and
i
S, it is
well dened as
i
: i < ) is strictly increasing continuous hence
i
: i < is a
club of ; so by we know that

< cf(i()) = cf(


i()
)

. Let

i
be

i
for i i() a limit ordinal and be
i
+ 1 for i < i() a non limit ordinal. Now
by induction on i i() choose p

i
and if i < i() also p
i
and prove on them the
following:
()(i) p
i
, p

i
P N
i+1
(ii) p
i
is increasing
(
7
8
4
)


r
e
v
i
s
i
o
n
:
2
0
0
4
-
0
3
-
0
1







m
o
d
i
f
i
e
d
:
2
0
0
4
-
0
3
-
0
2


4 SAHARON SHELAH
(iii)
p
i
>

i
(and
i+1
>
p
i
follows from p P N
i+1
)
(iv) S
p
i

i
= S and C
p
i

i
=

j
: j < i
(v) p

i
is the <

-rst q satisfying:
q P
j < i p
j
q

q
>
i
S
q

i
= S and
C
q

i
=

j
: j < i
(vii) p
i
is the <

-rst q such that:


q P
p

i
q
q forces a value to f

(i) if

<
q forces a value to f


i
if

= .
There is no problem to carry the denition, recalling the inductive hypothesis on

and noting that (p

j
, p
j
) : j < i) N
i+1
by the <

-rst being used to make our


choices as N
j
: j i) N
i+1
hence
j
: j i) N
i+1
and also

j
: j i) N
i+1
(and p, f

N
0
N
i+1
).
Now p

i()
is as required.
1.4
1.5 Conclusion: Let = cf() >
0
. Forcing with P
2

add no bounded subset of


and preserve stationarity of subsets of (and add no new sequences of ordinals of
length < ).
Proof. Obvious from 1.4.
1.6 Claim. Let = cf() >
0
. If FA(P
2

), (the forcing axiom for the forcing


notion P
2

, dense sets) holds, then there is a witness (

S,

C) to where
1.7 Denition. 1) For regular uncountable, we say that (

S,

C) is a witness to
or (

S,

C) is a -witness if:
(a)

S = S

: < )
(b)

C = C

: < )
(c) for every < , (,

S ( + 1),

C ( + 1)) P
2

.
(
7
8
4
)


r
e
v
i
s
i
o
n
:
2
0
0
4
-
0
3
-
0
1







m
o
d
i
f
i
e
d
:
2
0
0
4
-
0
3
-
0
2


FORCING AXIOM FAILURE FOR ANY >
1
SH784 5
2) For (

S,

C) a witness for , let F = F
(

S,

C)
be the function F : dened by
F() = Min : S

= S

.
3) For < let W

S,

C)
= < : F
(

S,

C)
() = .
Proof of 1.6. Let J
i
= p P
2

:
p
i, by 1.3(2) this is a dense open subset of P
2

,
hence by the assumption there is a directed G P
2

such that i < J


i
G ,= .
Dene
S

= S
p

, C

= C
p

for every p G such that


p
.
Now check.
1.6
1.8 Observation: Let (

S,

C) be a witness for and F = F
(

S,

C)
.
1) If < then F() .
2) If < is limit then F() < .
3) If < then W
F()
(

S,

C)
.
4) If < and i = F() and C

then / S

= S
i
.
Proof. Easy (for part (4) remember that each S

is a set of limit ordinals < and


that for limit
p
, p P
2

we have = sup(C

) and S

= ).
1.9 Claim. Assume (

S,

C) is a -witness and S

satises S

cf()
>
0
and F
(

S,

C)
S

is constant and S

is stationary. Then there is a club E

of such that: (

S,

C, S

, E

) is a strong (, )-witness, where


1.10 Denition. 1) We say that p = (

S,

C, S

, E

) is a strong -witness if
(a) (

S,

C) is a -witness
(b) S

is a set of limit ordinals and is a stationary subset of


(c) E

is a club of
(d) for every club E of , for stationarily many S

we have
= sup C

: < Suc
1
C

(, E

) E
where
()(i) Suc
0
C

() = Min(C

( + 1)),
(ii) Suc
1
C

(, E

) = sup(E

Suc
0
C

()).
(
7
8
4
)


r
e
v
i
s
i
o
n
:
2
0
0
4
-
0
3
-
0
1







m
o
d
i
f
i
e
d
:
2
0
0
4
-
0
3
-
0
2


6 SAHARON SHELAH
2) We say (

S,

C, S

, E

) is a strong (, )-witness if in addition


(e) S

cf() .
3) For (

S,

C, S

, E

) a strong -witness we let



C

= C

: S

acc(E

)), C

=
C

Suc
1
C

(, E

) : C

; if p = (

S,

C, S

, E

) we write

C

=

C

p
and

S
p
=

S,

C
p
=

C, S

p
= S

, E

p
= E

. We call (

S,

C, S

, E

,

C

) an expanded strong
-witness (or (, )-witness).
1.11 Observation. In Denition 1.10(3) for S

acc(E

) we have:
C

is a club of , Min(C

) sup(E

Min(C

)) and if
1
<
2
are successive
members of C

then C

(
1
,
2
) has at most one member (which necessarily
is sup(E


2
)) hence acc(C

) = acc(C

) and C

< Suc
1
C

()
/ C

and acc(C

) = acc(C

).
Proof of 1.9. As in [Sh:g, III], but let us elaborate, so assume toward contradiction
that for no club E

of is (

S,

C, S

, E

) a strong (, )-witness. We choose by


induction on n sets E

n
, E
n
, A
n
such that:
(a) E

n
, E
n
are clubs of
(b) E

0
=
(c) E
n
is a club of such that the following set is not stationary (in )
A
n
= S

: acc(E

n
) and
= sup C

: < Suc
1
C

(, E

n
) E
n

(d) E

n+1
is a club of included in acc(E

n
E
n
) and disjoint to A
n
.
For n = 0, E

n
is dened by clause (b).
If E

n
is dened, choose E
n
as in clause (c), possible by our assumption toward
contradiction, also A
n
S

is dened and not stationary. So obviously E

n+1
as
required in clause (d) exists.
So E

=: E

n
: n < is a club of and let () be the constant value of
F
(

S,

C)
S

, exists by an assumption of the claim. Recall that S


()
is a stationary
subset of , so clearly E

=: E

: = sup( E

S
()
) is a club of . As
S

is a stationary subset of , we can choose

. For each n < we


(
7
8
4
)


r
e
v
i
s
i
o
n
:
2
0
0
4
-
0
3
-
0
1







m
o
d
i
f
i
e
d
:
2
0
0
4
-
0
3
-
0
2


FORCING AXIOM FAILURE FOR ANY >
1
SH784 7
have

n+1
hence

/ A
n
hence

n
= sup C

:
< Suc
1
C

(, E

n
) E
n
is <

but C

. But

so cf(

) >
0
,
hence

= sup

n
, Min(C

) : n < is <

but Min(C

) and it belongs
to C

. As

, we know that

= sup(

S
()
) hence there is

S
()
( Suc
0
C

),

). But

S
()
recalling by the choice of
() above F
(

S,

C)
(

) = () hence by Claim 1.8(4), i.e., Denition 1.2(1), clause


(A)(h) and Denition 1.7(1) we have C

S
()
= hence

/ C

. But

>

>

Min(C

) and C

is a closed subset of

hence

= max(C

)
is well dened and so, recalling

we have
(n < )(

< Suc
0
C

).
Let

= Suc
0
C

) so clearly

). Now for every n we have sup(

n
) [

] as

S
()
E

n
.
So recalling

<

clearly

< sup(

n
); if also sup(

n
) E
n
then
recalling

= Suc
0
C

), Suc
1
C

, E

n
) sup(

n
) we have

n
(see
its choice and see the choice of

n
above), but this contradicts

Suc
0
C

) >

n
and the denition of A
n
(see clause (c) of ()), contradiction. So necessarily
sup(

n
) does not belong to E
n
hence does not belong to E

n+1
, hence sup(

n
) > sup(

n+1
).
So sup(

n
) : n < ) is a strictly decreasing sequence of ordinals, contra-
diction.
1.9
1.12 Denition. Assume
()
1
(

S,

C, S

, E

,

C

) is an expanded strong -witness so



C

= C

:
S

), C

= C

Suc
1
C

(, E

) : C

or just
()
2
S

is a stationary set of limit ordinals,



C

= C

: S

), C

is a club
of and E

is a club of .
We dene a forcing notion P = P
C

(A) c P i
(a) c is a closed bounded subset of
(b) if S

c then
C

: Suc
0
C

() c is bounded in
Let
c
= sup(c).
(B) order: c
1
c
2
i c
1
is an initial segment of c
2
.
(
7
8
4
)


r
e
v
i
s
i
o
n
:
2
0
0
4
-
0
3
-
0
1







m
o
d
i
f
i
e
d
:
2
0
0
4
-
0
3
-
0
2


8 SAHARON SHELAH
1.13 Claim. Let P = P
C
be as in Denition 1.12.
1) P is a (non empty) forcing notion.
2) For i < the set J
i
= c P : i < sup(c) is dense open.
Proof. 1) Trivial.
2) If c P, i < and c / J
i
then let c
2
= c i +1, clearly (c
2
c) S

= as S

is a set of limit ordinals hence c


2
P and obviously c c
2
J
i
.
1.13
1.14 Claim. Assume p = (

S,

C, S

, E

,

C

) is an expanded strong -witness.


Forcing with P = P
C
add no new bounded subsets of , no new sequence of
ordinals of length < and preserve stationarity of subsets of .
Proof. Assume p P,

and f

is a P-name of a function from

to the
ordinals or just to V and S is stationary and we shall prove that there are q,
satisfying (the parallel of) of 1.4, i.e.,
(i) p q P
(ii)
q
= + 1
(iii) S if

=
(iv) q forces a value to f

)
(v) if <

and p

then q
P
f

() < .
This is clearly enough for all the desired consequences. We prove this by induction
on

, so without loss of generality

= [

[ and without loss of generality

<

= cf(

), but if

< then S is immaterial so without loss of generality

<
& S cf()

. Also we can shrink S as long as it is a stationary subset


of and recall that F
(

S,

C)
is regressive on limit ordinals (see Observation 1.8(2))
so without loss of generality F
(

S,

C)
S is constantly say ().
Let be large enough and choose

N = N
i
: i < ) such that
(a) N
i
(H (), , <

) is increasing continuous
(b) , p, f

, S belongs to N
i
hence P N
i
(c) |N
i
| <
(d) N
i

(e) N
j
: j i) belong to N
i+1
(hence i N
i
, so N
i
: i < )
(f) N
i+1
S
()
and N
0
S
()
.
(
7
8
4
)


r
e
v
i
s
i
o
n
:
2
0
0
4
-
0
3
-
0
1







m
o
d
i
f
i
e
d
:
2
0
0
4
-
0
3
-
0
2


FORCING AXIOM FAILURE FOR ANY >
1
SH784 9
Let
i
= N
i
and i() = Mini : i < is a limit ordinal and
i
S, it is well
dened as
i
: i < ) is (strictly increasing continuous) hence
i
: i < is a club
of , hence

< cf(i()) = cf(


i()
)

.
Let
1
W =: i i(): i > 0 and if i < i() and j < i then C

i()

i+1

j+1
.
Clearly W i() is a closed subset of i() and as
i()
= sup(C

i()
), also W i()
is unbounded in i(). Also as by 1.11 we have ( acc C

i()
) C

i()

C

= C

i()
clearly
() if i W then N
j
: j W (i + 1)) N
i+1
.
Also note that
() if i < i() is nonlimit, then
i
> sup(C

i()

i
) hence
i
> sup(C

i()

i
).
[Why? By 1.8(4) as
i()
S S
()
recalling the choice of () clearly
C

i()
S
()
= but by clause (f) we have
i
S
()
so
i
/ C

. But
C

i()
is a closed subset of
i()
hence
i
> sup(C

i()

i
), and C

i()

i
sup(C

i()

i
) has at most two members (see 1.11) so C

i()

i
is a
bounded subset of
i
so we are done.]
Now by induction on i W we choose p
i
, p

i
and prove on them the following:
()(i) p
i
, p

i
P N
i+1
(ii) p
i
is increasing (in P)
(iii) max(p
i
) >
i
(of course
i+1
> max(p
i
) as p
i
P N
i+1
)
(iv) p

i
= p sup(
i
(C

i()

i+1
)) + 1 if i = Min(W)
(v) if 0 < i = sup(W i) and
i
= max(C

i()

i+1
) so
i

i
<
i+1
then
p

i
= p
j
: j W i
i
,
i
+ 1
(vi) if j < i are in W then p
j
p

i
p
i
(vii) i W, i < i() and j < i satises j = Max(W i) and
i
= max(
i

(C


i+1
)) so
i

i
<
i+1
then p

i
= p
j

i
+ 1
(viii) p
i
is the <

-rst q P satisfying
() p

i
q P
() if

< then q forces a value to f

(otp(j < i : j W and otp(j W)


is a successor ordinal)
() if

= then q forces a value to f


i
1
if cf(
i()
) >
0
then W = {i < i() :
i
C

i()
} {
i()
} is O.K.
(
7
8
4
)


r
e
v
i
s
i
o
n
:
2
0
0
4
-
0
3
-
0
1







m
o
d
i
f
i
e
d
:
2
0
0
4
-
0
3
-
0
2


10 SAHARON SHELAH
(ix) p

j<i
p
j
and p
i
p

i
are disjoint to C

i()
acc(C

i()
), which include the
set Suc
1
C

i()
(, E

) : C

i()
and < Suc
1
C

i()
().
Note that clause (ix) follows from the rest; we now carry the induction.
Case 1: i = Min(W).
Choose p

i
just to fulll clauses (iv), note that
i

i
<
i+1
as i W i()
and then choose p
i
to fulll clause (viii).
Case 2: i = Min(W(j + 1)) and j W.
Choose p

i
by clauses (vii) and then p
i
by clause (viii).
Case 3: 0 < i = sup(W i).
A major point is p
j
: j < i) N
i+1
, this holds as p

j
, p
j
, j i W) is denable
from

N
i
, f

, p, C

i()
N
i+1
all of which belong to N
i+1
and N
i+1
(H (),
, <

).
Let p

i
be dened by clause (v), note that
i

i
<
i+1
as i W and p

i
P
as:
() (j < i)[p
j
P] and
()
i
= sup(
j
: j < i and j W).
[Why? As
i
< max(p
j
) <
i+1
by clause (iii)] and
() p

i
S

sup(p

i
C

acc(C

)) < .
[Why? If <
i
then for some j i W we have <
j
so p
j
is an initial
segment of p

i
hence sup(p

i
C

) = sup(p
j
C

) < . If =
i
we can
assume S

but clearly =
i
C

i()
by the denition of W and the
assumption of case 3; so by (

S,

C) being a -witness, C

i
= C

i()

i
so by
clause (ix) the demand (in ()) hold.]
So easily p

i
is as required. If i < i() we can choose p
i
by clause (viii) using the
induction hypothesis if

= . So we have carried the denition and p

i()
is as
required.
1.14
1.15 Conclusion: 1) If p = (

S,

C, S

, E

) is a strong -witness and



C

=

C

p
and
P = P
C
, then FA(P, ) fails.
2) In part (1), P
C
is a forcing of cardinality 2
<
, add no new sequence of ordinals
of length < and preserve stationarity of subsets of any = cf() [
1
, ].
(
7
8
4
)


r
e
v
i
s
i
o
n
:
2
0
0
4
-
0
3
-
0
1







m
o
d
i
f
i
e
d
:
2
0
0
4
-
0
3
-
0
2


FORCING AXIOM FAILURE FOR ANY >
1
SH784 11
Proof. 1) Recall that by Claim 1.13(2), J
i
is a dense open subset of P. Now if
G P
C
is directed not disjoint to J
i
for i < , let E = p : p G. By
the denition of P
C
and J
i
clearly E is an unbounded subset of and by the
denition of P
C
and G being directed, p G E (max(p) + 1) = p and (p is
closed) hence E is a closed unbounded subset of . So E contradicts the denition
of (

S,

C,

S

,

E

,

C

) being a strong -witness.


2) Follows from 1.14 and direct checking.
1.15
1.16 Conclusion: Let be regular >
1
.
Then there is a forcing notion P such that:
() P of cardinality 2

() forcing with P add no new sequences of ordinals of length <


() forcing with P preserve stationarity of subsets of (and by clause () also
of any = cf() [
1
, ))
() FA(P, ) fail.
Proof. We try P
2

, it satises clause (), (), () (see 1.3(1), 1.5, 1.6). If it satises


also clause () we are done otherwise by Claim 1.6 there is a -witness (

S,

C). Let
S

< : cf() >


0
be stationary, so by 1.9 for some club E

of , the
quadruple p = (

C,

S, S

, E

) is a strong -witness (see Denition 1.10), and let

=

C

p
.
Now the forcing notion P = P
C
(see Denition 1.12) satises clauses (), (), ()
by claims 1.15(2) and also clause () by claim 1.15(1). So we are done.
1.16
(
7
8
4
)


r
e
v
i
s
i
o
n
:
2
0
0
4
-
0
3
-
0
1







m
o
d
i
f
i
e
d
:
2
0
0
4
-
0
3
-
0
2


12 SAHARON SHELAH
2 There are
1
-semi-proper not proper forcing notion
2
By [Sh:f, XII,2], it was shown that if remnant of large cardinal properties
holds (e.g. 0
#
) then every quite semi-proper forcing is proper, more fully UReg-
semi-properness implies properness. This leaves the problem
() is the statement

for every forcing notion P, P is proper follows from P is


semi-proper, i.e.,
1
-semi proper

consistent or is the negation provable


in ZFC.
David Asparo raises the question and we answer armatively: there are such forcing
notions. So the iteration theorem for semi proper forcing notions in [Sh:f, X] is not
covered by the one on proper forcing notions even if 0
#
does not exist.
2.1 Claim. There is a forcing notion P of cardinality 2

2
which is not proper but
is
1
-semi proper. This follows from 2.2 using =
2
.
2.2 Claim. Assume = cf() >
1
, = 2

. Then there is P such that


(a) P is a forcing notion of cardinality 2

(b) if > , p P N (H (), ), N countable, then there is q P above p


such that
q NN[G

P
] ( means initial segment); this gives P is
1
-semi
proper and more
(c) there is a stationary S []

0
such that
P
S is not stationary
(d) P is not proper.
Proof. We give many details.
Stage A: Preliminaries.
Let M

= (, F
n,m
)
n,m<
, with F
n,m
an (n + 1)-place function be such that
for every n < and n-place function f from to there is m < such that
(i
1
, . . . , i
n
< )( < )[f(i
1
, . . . , i
n
) = F
n,m
(, i
1
, . . . , i
n
)].
Let S
1
, S
2
be disjoint stationary subsets of of conality
0
(i.e. S
1
S
2

cf() =
0
).
Let
2
done 2001/8/8
(
7
8
4
)


r
e
v
i
s
i
o
n
:
2
0
0
4
-
0
3
-
0
1







m
o
d
i
f
i
e
d
:
2
0
0
4
-
0
3
-
0
2


FORCING AXIOM FAILURE FOR ANY >
1
SH784 13
S =

a []

0
: for some b []

0
we have
() a b are closed under F
n,m
for n, m < ,
() sup(a ) S
1
, sup(b ) S
2
() (a ) (b ) ( is being an initial segment)

P = P
S
=

a : a = a
i
: i ) is an increasing continuous sequence
of members of []

0
S of length <
1

Clearly clause (a) of 2.2 holds.


Stage B: S is a stationary subset of []

0
.
Why? Let N

be a model with universe and countable vocabulary, it is enough


to nd a S such that N

a N. Without loss of generality N

has Skolem
functions and N

expands M

. Choose for < , N

, |N

| < , <
N

, N

, N

increasing continuous.
So C =: < : a limit ordinal and N

= is a club of . Choose
1
<
2
from C such that
1
S
1
,
2
S
2
. Choose a countable c
1

1
unbounded in
1
,
and a countable c
2

2
unbounded in
2
.
Choose a countable M N

2
such that M N

1
N

1
and c
1
c
2
. Let
a = M N

1
, b = M N

2
. As N

expands M

, clearly a, b are closed under the


functions of M

. Also c
1
M
1
= M (N

1
) = a N

1
=
1
hence

1
= sup(c
1
) sup(a )
1
so sup(a ) =
1
. Similarly sup(b ) =
2
.
Lastly, obviously a b so b witnesses a S, as required.
Stage C:
P
S is not stationary.
Why? Dene a

= a

: a G

P
, g( a) > . Clearly
()
0
P ,= .
[Why? Trivial.]
()
1
for <
1
, I
1

= a P : g( a) > is a dense open subset of P.


[Why? If a
i
: i j) P, j < <
1
we let a
i
=: a
j
for i (j, ] and then
a
i
: i j)
P
a
i
: i ).]
Also
(
7
8
4
)


r
e
v
i
s
i
o
n
:
2
0
0
4
-
0
3
-
0
1







m
o
d
i
f
i
e
d
:
2
0
0
4
-
0
3
-
0
2


14 SAHARON SHELAH
()
2
for < , J
2

= a P : a

for some < g( a) is a dense open


subset of P.
[Why? Given a = a
i
: i j). Choose S
2
such that > sup( (a
j

) let c be countable unbounded in and let a
j+1
= a
j
c; so
trivially sup(a
j+1
) = S
2
hence a
j+1
/ S. Now let a
+
= a
i
: i
j + 1). Now check.]
So
()
3

P

i
: i <
1
) is an increasing continuous sequence of members of
([]

0
)
V
S whose union is
hence
()
4

P
a

i
: i <
1
) witness S is not stationary (subset) of []

0
.
So we have nished Stage C.
Stage D: Clauses (c),(d) of 2.2 holds.
Why? By Stage B and Stage C.
Stage E: Clause (b) of 2.2 holds.
So let > , N a countable elementary submodel of (H (), , <

) to which P
and p P belong hence M

, , , S N (they are denable from P or demand it).


In the next stage we prove
there is a countable M (H (), <

) such that N M, (N) (M)


and M / S.
Let J
n
: n < ) list the dense open subsets of P which belong to M. Choose by
induction on n, p
n
NP : p
0
= p, p
n

P
p
n+1
J
n
. So let p
n
= a
i
: i
n
), by
()
1
of Stage C the sequence
n
: n < ) is not eventually constant. Dene q by:
q = a
i
: i ) where =
n
: n < and a

= M . Trivially a
i
M
and by ()
2
of Stage C clearly a

= a
i
: i < hence a
i
: i ) is increasing
continuous and i a
i
[]

0
and i < a
i
[]

0
S. So the only non
trivial point is a

/ S which holds by .
Clearly p q and q is (M, P)-generic hence q N[G] M[G] and N (N[G]
) M[G] = M so as (N)(M) necessarily (N[G] ) (N[G] )
as required.
Stage F: Proving .
If N / S let M = N and we are done so assume M S. Let
a = N []

0
and let b []

0
witness a = N S [the rest should by
(
7
8
4
)


r
e
v
i
s
i
o
n
:
2
0
0
4
-
0
3
-
0
1







m
o
d
i
f
i
e
d
:
2
0
0
4
-
0
3
-
0
2


FORCING AXIOM FAILURE FOR ANY >
1
SH784 15
now be clear but we elaborate]. Let M be the Skolem Hull in (H(), , <

) of
N(b) (exists as <

is a well ordering of H () so (H (), , <

) has (denable)
Skolem functions).
If M then we can nd a denable function f of (H (), , <

) and x N
(recall that in N we can use m-tuple for every m) and
1
. . .
n
b such that
= f(x,
1
, . . . ,
n
). Fixing x, f the mapping (
1
, . . . ,
n
) f(x,
1
, . . . ,
n
) is
an n-place function from to denable in N hence belong to N and M

N
hence for some N and m < we have (
1
, . . . ,
n
< )[f(x,
1
, . . . , a
n
) =
F
n,m
(,
1
, . . . ,
n
)].
But
1
, . . . ,
n
b b and N b = b and as b being in S is
closed under F
n,m
clearly = f(x,
1
, . . . ,
n
) = F
n,m
(,
1
, . . . ,
n
) b but
so b . So M b but of course b M so b = M . So
a = (N ) = N ; but a b by the choice of b so N =
a b = M .
Lastly, sup(M ) = sup(b ) S
2
hence M / S. So M is as required in
and we are done.
2.2
(
7
8
4
)


r
e
v
i
s
i
o
n
:
2
0
0
4
-
0
3
-
0
1







m
o
d
i
f
i
e
d
:
2
0
0
4
-
0
3
-
0
2


16 SAHARON SHELAH
REFERENCES.
[Sh:g] Saharon Shelah. Cardinal Arithmetic, volume 29 of Oxford Logic Guides.
Oxford University Press, 1994.
[Sh:f] Saharon Shelah. Proper and improper forcing. Perspectives in Mathematical
Logic. Springer, 1998.

Anda mungkin juga menyukai