Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
8
7
r
e
v
i
s
i
o
n
:
1
9
9
3
-
0
8
-
2
8
m
o
d
i
f
i
e
d
:
1
9
9
3
-
0
8
-
2
8
FULL REFLECTION OF STATIONARY SETS BELOW
: S
3
1
) be such that each C
1
. Clearly, at most
1
of the sets S
i
can meet each C
S
i
= for all i i().
There is i <
2
such that i() = i for a stationary set of s. Let A S
3
1
be
this stationary set and let S = S
i
. Then S C
3
Proposition 1.2. Let k < m < n 1. There exist stationary sets S S
n
k
and
A S
n
m
such that S does not reect at any A.
Consequently, if n > 2 then full reection in S
n
m
is possible only if m = n1. This
motivates our Main Theorem.
1.3 Main Theorem. Let
2
<
3
< <
n
< be a sequence of supercompact
cardinals. There is a generic extension V [G] in which
n
=
n
for all n 2, and
such that
(a) every stationary subset of S
2
0
reects fully in S
2
1
, and
(b) for every n 3, every stationary subset of S
n
k
for all k = 0, , n3, reects
fully in S
n
n1
.
We will show that the result of the Main Theorem is best possible. But rst we
prove a corollary:
1.4 Corollary. In the model of the Main Theorem we have for all n 2 and all
m, 0 < m < n:
(a) Any
m
stationary subsets of S
n
0
reect simultaneously at some S
n
m
.
(b) For every k m2,any
m
stationary subsets of S
n
k
reect simultaneously
at some S
n
m
.
Proof. Let us prove (a) as (b) is similar. Let m < n and let S
, <
m
, be
stationary subsets of S
n
0
. First, each S
reects fully in S
n
n1
and so there exist
club sets C
, <
m
, such that each S
reects at all C
S
n
n1
. As the club
lter is
n
- complete, there exists an S
n
n1
such that S
is stationary, for
all <
m
. Next we apply full reection of subsets of S
n1
0
in S
n1
n2
(to the ordinal
of conality
n1
rather than to
n1
itself) and the
n1
- completeness of the
club lter on
n1
, to nd S
n
n2
such that S
is stationary.
Note that the amount of simultaneous reection in 1.4 is best possible:
1.5 Proposition. If cf =
m
and if S
, <
m+1
, are disjoint stationary sets then
some S
.
3
8
7
r
e
v
i
s
i
o
n
:
1
9
9
3
-
0
8
-
2
8
m
o
d
i
f
i
e
d
:
1
9
9
3
-
0
8
-
2
8
4 THOMAS JECH AND SAHARON SHELAH
By Corollary 1.4, the model of the Main Theorem has the property that whenever
2 m < n, every stationary subset of S
n
k
reects quite strongly in S
n
m
, provided
k m 2. This cannot be improved to include the case of k = m 1, as the
following proposition shows:
1.6 Proposition. Let m 2. Either
(a) for all k < m1 there exists a stationary set S S
m
k
that does not reect
fully in S
m
m1
,
or
(b) for all n > m there exists a stationary set A S
n
m1
that does not reect at
any S
n
m
.
We shall give a proof of 1.6 in Section 3. In our model we have, for every m 2,
full reection of subsets of S
m
0
in S
m
m1
(and of subsets of S
m
k
for k m 3) and
therefore 1.6 (a) fails in the model. Thus the model necessarily satises 1.6 (b),
which shows that the consistency result is best possible.
2. Proof of Main Theorem
Let
2
<
3
< <
n
< be a sequence of cardinals with the property that
for each n 2,
n
is a <
n+1
- supercompact cardinal, i.e. for every <
n+1
there exists an elementary embedding j : V M with critical point
n
such that
j(
n
) > and M
M.
1
We construct the generic extension by iterated forcing,
an iteration of length with full support. The rst stage of the iteration P
1
makes
2
=
2
, and for each n, the n
th
stage P
n
(a forcing notion in V (P
1
P
n1
))
makes
n+1
=
n+1
. In the iteration, we repeatedly use three standard notions of
forcing: Col (, ), C() and CU(, T).
Denition. Let be a regular uncountable cardinal.
(a) Col (, ) is the forcing that collapses with conditions of size < :
A condition is a function p from a subset of of size < into ; a condition q is
stronger than p if q p.
(b) C() is the forcing that adds a Cohen subset of : A condition is an 0-1-
function p on a subset of of size < ; a condition q is stronger than p if q p.
1
We note in passing that the condition about the
n
is equivalent to every
n
is <
-
supercompact where
= sup
m<
m
.
3
8
7
r
e
v
i
s
i
o
n
:
1
9
9
3
-
0
8
-
2
8
m
o
d
i
f
i
e
d
:
1
9
9
3
-
0
8
-
2
8
FULL REFLECTION OF STATIONARY SETS BELOW
5
(c) CU(, T) is the forcing that shoots a club through a stationary set T :
A condition is a closed bounded subset of T; a condition q is stronger than p if
q end-extends p.
The rst stage P
1
of the iteration P = P
n
: n = 1, 2, ) is a forcing of size
2
that is - closed
2
, satises the
2
- chain condition and collapses each cardinal
between
1
and
2
(it is essentially the Levy forcing with countable conditions.)
For each n 2, we construct (in V (P[n)) the n
th
stage P
n
such that
(2.1) (a) [P
n
[ =
n+1
(b) P
n
is
n2
closed
(c) P
n
satises the
n+1
- chain condition
(d) P
n
collapses each cardinal between
n
(=
n
) and
n+1
(e) P
n
does not add any
n1
- sequences of ordinals
and such that P
n
guarantees the reection of stationary subsets of
n
stated in the
theorem.
It follows, by induction, that each
n
becomes
n
: Assuming that
n
=
n
in V (P[n), the n
th
stage P
n
preserves
n
by (e), and the rest of the iteration
P
n+1
, P
n+2
, ) also preserves
n
because it is
n1
- closed by (b); P
n
makes
n+1
the successor of
n
by (c) and (d).
We rst dene the forcing P
1
:
P
1
is an iteration, with countable support, Q
: <
2
) where for each ,
Q
= Col (
1
,
1
+) C(
1
).
It follows easily from well known facts that P
1
is an -closed forcing of size
2
,
satises the
2
- chain condition and makes
2
=
2
.
Next we dene the forcing P
2
. (It is a modication of Magidors forcing from
[2], but the added collapsing of cardinals requires a stronger assumption on
2
than weak compactness. The iteration is padded up by the addition of Cohen
forcing which will make the main argument of the proof work more smoothly). The
denition of P
2
is inside the model V (P
1
), and so
2
=
2
:
P
2
is an iteration, with
1
- support, Q
: <
3
) where for each ,
Q
= Col (
2
,
2
+) C(
2
) CU(T
)
2
A forcing notion is - closed if every descending sequence of length has a lower bound.
3
8
7
r
e
v
i
s
i
o
n
:
1
9
9
3
-
0
8
-
2
8
m
o
d
i
f
i
e
d
:
1
9
9
3
-
0
8
-
2
8
6 THOMAS JECH AND SAHARON SHELAH
where T
is, in V (P
1
P
2
[), some stationary subset of
2
. We choose the T
s
so that each T
is -closed.
The crucial property of the forcing P
2
will be the following:
Lemma 2.2. P
2
does not add new
1
- sequences of ordinals.
One consequence of Lemma 2.2 is that the conditions (p, q, s) Q
can be taken
to be sets in V (P
1
) (rather than in V (P
1
P
2
[)). Once we have Lemma 2.2, the
properties (2.1) (a) - (e) follow easily.
It remains to specify the choice of the T
3
- chain condition, we can enumerate all potential subsets of
2
by a sequence
S
: <
3
) in such a way that each S
is already in V (P
1
P
2
[). At the stage
of the iteration, we let T
=
2
, unless S
is, in V (P
1
P
2
[), a stationary set
of ordinals of conality . If that is the case, we let
T
= (Tr(S
) S
2
1
) S
2
0
Assuming that Lemma 2.2 holds, we now show that in V (P
1
P
2
), every station-
ary S S
2
0
reects fully in S
2
1
:
The set S appears as S
P
2
), it is stationary in the smaller model V (P
1
P
2
[). The forcing Q
creates a
closed unbounded set C such that C S
2
1
Tr(S) (note that because P
2
does not
add
1
- sequences, the meaning of Tr(S) or of S
2
1
does not change).
Thus in V (P
1
P
2
) we have full reection of subsets of S
2
0
in S
2
1
. The later stages
of the iteration do not add new subsets of
2
and so this full reection remains true
in V (P).
We postpone the proof of Lemma 2.2 until after the denition of the rest of the
iteration.
We now dene P
n
for n 3. We work in V (P
1
P
n1
). By the induction
hypothesis we have
n
=
n
.
P
n
is an iteration with
n1
- support, Q
: <
n+1
), where for each ,
Q
= Col(
n
,
n
+) C(
n
) CU(T
)
where T
is a P
n
[ - name for a subset of
n
. To specify the T
s, let S
: <
n+1
)
be an enumeration of all potential subsets of
n
such that each S
is a P
n
[ - name.
3
8
7
r
e
v
i
s
i
o
n
:
1
9
9
3
-
0
8
-
2
8
m
o
d
i
f
i
e
d
:
1
9
9
3
-
0
8
-
2
8
FULL REFLECTION OF STATIONARY SETS BELOW
7
At the stage , let T
=
n
unless S
S
n
k
for
some k = 0, , n 3, in which case let
T
= (Tr(S
) S
n
n1
) (S
n
0
S
n
n2
)
= <
n
: cf
n2
or S
is stationary
Due to the selection of the T
s, Q
is
n2
- closed, and so is P
n
. The crucial
property of the forcing is the analog of Lemma 2.2:
Lemma 2.3. P
n
does not add new
n1
- sequences of ordinals.
Given this lemma, properties (2.1) (a) - (e) follow easily. The same argument as
given above for P
2
shows that in V (P
1
P
n
), and therefore in V (P) as well,
every stationary subset of S
n
k
, k = 0, , n 3, reects fully in S
n
n1
.
It remains to prove Lemmas 2.2 and 2.3. We prove Lemma 2.3, as 2.2 is an easy
modication.
Proof of Lemma 2.3.
Let n 3, and let us give the argument for a specic n, say n = 4. We want to
show that P
4
does not add
3
-sequences of ordinals.
We will work in V (P
1
P
2
) (and so consider the forcing P
3
P
4
). As P
1
P
2
has size
3
,
4
is a <
5
- supercompact cardinal in V (P
1
P
2
), and
3
=
3
. The
forcing P
3
is an iteration of length
4
that makes
4
=
4
and is
1
- closed; then
P
4
is an iteration of length
5
. By induction on <
5
we show
(2.4) P
4
[ does not add
3
- sequences of ordinals.
As P
4
has the
5
- chain condition, (2.4) is certainly enough for Lemma 2.3. Let
<
5
.
Let j be an elementary embedding j : V M (as we work in V (P
1
P
2
), V
means V (P
1
P
2
)) such that j(
4
) > and M
= M[G(j(P
3
)[)].
Since P
4
[ is
2
- closed, it is isomorphic in M
= (j(P
3
))() = Col(
3
,
3
+)C(
3
)CU(T
), so G[Q
= G
Col
G
C
G
CU
,
and using G
C
and the isomorphism between P
4
[ and C(
3
) we obtain H. Since
the quotient forcing j(P
3
)/(P
3
C(
3
)) is an iteration of
1
- closed forcings, it is
1
- closed.
Lemma 2.7. In M[G] there is a condition p j(P
4
[) that extends p, and extends
every member of j
H.
Lemma 2.7 will complete the proof of (2.5): since every value of
F is decided by
some condition in H, every value of j(
F) is decided by some condition in j
H, and
therefore by p.
Proof of Lemma 2.7. Working in M[G], we construct p j(P
4
[), a sequence
p
: < j()) of length j(), by induction. When is not in the range of j, we let
p
be the trivial condition; that guarantees that the support of p has size [[ which
is
3
(because < j(
4
) =
4
in M[G]). So let < be such that p[j() has been
dened, and construct p
j()
.
The condition p
j()
has three parts u, v, s where u Col(j(
4
), j(
4
) +j()), v
C((
4
)) and s CU(T
j()
). It is easy to construct the u - part and the v - part, as
follows: The lter H[P
4
() has three parts; a collapsing function f of
4
onto
4
+,
a 0-1-function g on
4
, and a club subset C of T
9
the s - part, let s = jC
4
. In order that this set be a condition in CU(T
j()
),
we have to verify that
4
T
j()
.
This is a nontrivial requirement if S
j()
is in M(j(P
3
) (j(P
4
)[j())) a stationary
subset of j(
4
) and is a subset of either S
4
0
or of S
4
1
(of S
n
k
for n = 4 and k n3).
Then
4
has to be reecting point of S
j()
, i.e. we have to show that S
j()
4
is
stationary, in M(j(P
3
) (j(P
4
)[j()).
By the assumption and by elementarity of j, S
is a stationary subset of
4
in
V (P
3
P
4
[), and S
S
4
0
or S
S
4
1
, i.e. consists of ordinals of conality
1
.
Since S
j()
4
= j(S
)
4
= S
is stationary in M(P
3
P
4
[).
Secondly, j(P
4
) is
1
- closed, and by Lemma 2.6, M(j(P
3
)) is an
1
- closed
forcing extension of M(P
3
P
4
[), and so the proof is completed by application of
the following lemma (taking =
0
or
1
, =
4
).
Lemma 2.8 Let < be regular cardinals and assume that for all < and all
< ,
: < ) be an
elementary chain of submodels of M such that each N
is an ordinal, and N
+1
contains all < - sequences in N
. Since S
is stationary, there exists a S such that N
= . As cf = , N = N
is
closed under < - sequences.
Let
N greater than
, q
+1
|
C. At successor
stages, q
+1
exists because in N, q
forces that
C is unbounded. At limit stages
< , the - sequence q
S
3
1
, <
3
, there exists a sta-
tionary set S S
3
0
such that S ,< A
for all .
Proof. Let A
, <
3
, be stationary subsets of S
3
1
. By [3], there exist
4
almost
disjoint stationary subsets of S
3
0
; let S
i
, i <
4
, be such sets. Assuming that each
S
i
reects fully in some A
(i)
, we can nd
4
of them that reect fully in the same
A
. Take any
2
of them and reduce each by a nonstationary set to get
2
pairwise
disjoint stationary subsets T
: <
2
of S
3
0
, such that each of them reects fully
in A
, <
2
, such that Tr(T
) A
for every .
Let
<
2
C
. Then every T
: <
2
. This is a contradiction because has a closed
unbounded subset of size cf =
1
.
The next result uses the fact that under GCH there exists a - sequence for S
3
1
.
Proposition 3.2. (GCH) There exists a stationary set A S
3
1
that is not the trace
of any S S
3
0
; precisely: for every S S
3
0
the set A(Tr(S) S
3
1
) is stationary.
Proof. Let S
: S
3
1
) be a - sequence for S
3
1
; it has the property that for
every set S
3
, the set D(S) = S
3
1
: S = S
is stationary. Let
A = S
3
1
: S
is nonstationary.
The set A is stationary because A D(). If S is any stationary subset of S
3
0
,
then for every in the stationary set D(S), A i / Tr(S), and so D(S)
ATr(S).
The remaining negative results use the following theorem of Shelah which proves
the existence of sets with the square property.
3
8
7
r
e
v
i
s
i
o
n
:
1
9
9
3
-
0
8
-
2
8
m
o
d
i
f
i
e
d
:
1
9
9
3
-
0
8
-
2
8
FULL REFLECTION OF STATIONARY SETS BELOW
11
Theorem ([4], Lemma 4.2). Let 1 k n 2. The set S
n
k
is the union of
n1
stationary sets A, each having the following property. There exists a collection
C
is a club
subset of of order type
k
, consisting of limit ordinals of conality <
k
, and
such that for all
1
,
2
A and all , if C
1
C
2
then C
1
= C
2
.
Square sequences can be used to construct a number of counterexamples. For
instance, if S
n
, n < , are
0
stationary subsets of S
3
0
, then Tr(
n=0
S
n
) =
n=0
S
n
.
Using a square sequence we get:
Proposition 3.3. There is a stationary set A S
3
1
and stationary subsets S
i
, i <
1
,
of S
3
0
such that Tr(S
i
) A = for each i but Tr(
i<
1
S
i
) A.
Proof. Let A be a stationary subset of S
3
1
with a square sequence C
: A,
and let S =
A
C
. Clearly, S S
3
0
is stationary, and Tr(S) A. For each <
1
,
let
S
= S : order type (C
) =
(this is independent of the choice of A). For every S and every <
1
,
the set S
is a well known
trick]
The argument used in the above proof establishes the following:
Proposition 3.4. If a stationary set A S
n
m
has a square sequence and if k < m
then there exists a stationary S S
n
k
that does not reect at any A.
Proof of Proposition 1.6. Let 2 m < n and let us assume that (b) fails, i.e.
that every stationary set A S
n
m1
reects at some of conality
m
. We shall
prove that (a) holds. For each k < m 1 we want a stationary set S S
m
k
that
does not reect fully in S
m
m1
. Let k < m1.
Let A be a stationary subset of S
n
m1
that have a square sequence C
: A.
The set A reects at some of conality
m
. Let C be a club subset of of order
type
m
. Using the isomorphism between C and
m
, the sequence C
C : A
becomes a square sequence for a stationary subset B of S
m
m1
. It follows that there
is a stationary subset of S
m
k
that does not reect at any B.
3
8
7
r
e
v
i
s
i
o
n
:
1
9
9
3
-
0
8
-
2
8
m
o
d
i
f
i
e
d
:
1
9
9
3
-
0
8
-
2
8
12 THOMAS JECH AND SAHARON SHELAH
The last counterexample also uses a square sequence.
Proposition 3.5. (GCH) There is a stationary set A S
3
1
and
4
stationary sets
S
i
S
3
0
such that the sets Tr(S
i
) A : i <
4
are stationary and pairwise almost
disjoint.
Proof. Let A be a stationary subset of S
3
1
with a square sequence C
: A),
and let S =
A
C
. Let f
i
: i <
4
be regressive functions on S
3
0
S
3
1
with the
property that for any two f
i
, f
j
, the set : f
i
() = f
j
() is nonstationary (such
a family exists by [3]). For each i and each A, the function f
i
is regressive
on C
: f
i
() < is
stationary. Let T
i,
1
be the stationary set o.t.(C
) : f
i
() < and let
H
i,
be the function on T
i,
(with values < ) dened by H() = f
i
(
th
element of
C
) if = o.t.(C
) T then f
i
() = H()
S
i
= S : o.t.(C
) T and( , C
) if = o.t.(C
) T then f
i
() = H()
By the denition of T and H, each A
i
is a stationary set, and each S
i
reects at
every point of A
i
. We claim that if A and S
i
is stationary then A
i
. So
let A be such that S
i
is stationary. Let T and let be the
th
element
of C
of all limit
points of C is disjoint from A
i
A
j
. If C
. Since T is stationary in
1
, there is a T such that the
th
element
of C
is in C, and therefore f
i
() ,= f
j
(); it follows that cannot be both in A
i
and in A
j
.
3
8
7
r
e
v
i
s
i
o
n
:
1
9
9
3
-
0
8
-
2
8
m
o
d
i
f
i
e
d
:
1
9
9
3
-
0
8
-
2
8
FULL REFLECTION OF STATIONARY SETS BELOW
13
[1] T. Jech, Stationary subsets of inaccessible cardinals, in Axiomatic Set Theory
(J. Baumgartner, ed.), Contemporary Math. 31, Amer. Math. Soc. 1984,
115-142.
[2] M. Magidor, Reecting stationary sets, J. Symbolic Logic 47 (1982), 755-771.
[3] S. Shelah [Sh 247], More on stationary coding, in Around Classication Theory
of Models, Springer-Verlag Lecture Notes 1182 (1986), pp. 224-246.
[4] S. Shelah [Sh 351], Reecting stationary sets and successors of singular cardinals.
The Pennsylvania State University The Hebrew University