6

9
1


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







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


Weak reection at the successor of a singular
cardinal
Mirna Dzamonja
School of Mathematics
University of East Anglia
Norwich, NR47TJ, UK
M.Dzamonja@uea.ac.uk
http://www.mth.uea.ac.uk/people/md.html
Saharon Shelah
Mathematics Department
Hebrew University of Jerusalem
91904 Givat Ram, Israel
and
Rutgers University
New Brunswick, New Jersey
USA
shelah@sunset.huji.ac.il
http://www.math.rutgers.edu/ shelarch
October 6, 2003
Abstract
The notion of stationary reection is one of the most important
notions of combinatorial set theory. We investigate weak reection,
which is, as its name suggests, a weak version of stationary reection.
Our main result is that modulo a large cardinal assumption close to 2-
hugeness, there can be a regular cardinal such that the rst cardinal
1
6
9
1


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







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


weakly reecting at is the successor of a singular cardinal. This
answers a question of Cummings, Dzamonja and Shelah.
1
0 Introduction and the statement of the re-
sults.
Stationary reection is a compactness phenomenon in the context of station-
ary sets. To motivate its investigation, let us consider rst the situation of
a regular uncountable cardinal and a closed unbounded subset C of .
For every limit point of C we have that C is closed unbounded in
. Now let us ask the same question, but starting with a set S which is
stationary, not necessarily club, in . Is there necessarily < such that
S is stationary in -or as this situation is known in set theory, S reects
at ? The answer to this question turns out to be very intricate, and in
fact the notion of stationary reection is one of the most studied notions of
combinatorial set theory. This is the case not only because of the historical
signicance stationary reection achieved through by now classical work of
R. Jensen [Je] and later work of J.E. Baumgartner [Ba], L. Harrington and S.
Shelah [HaSh 99], M. Magidor [Ma] and many later papers, but also because
of the large number of applications it has in set theory and allied areas. In
set theory, stationary reection is known to have deep connections with var-
ious guessing and coherence principles, the simplest one of which is Jensens
([Je]), and the notions from pcf theory, such as good scales (for a long
list of results in this area, as well as an excellent list of references, we refer
the reader to [CuFoMa]), and some connections with saturation of normal
lters ([DjSh 545]). In set-theoretic topology, various kinds of spaces have
been constructed from the assumption of the existence of a non-reecting
stationary set (for references see [KuVa]), and in model theory versions of
1
This paper is numbered 691 (10/98) in Saharon Shelahs list of publications. Both
authors thank NSF for partial support by their grant number NSF-DMS-97-04477, as well
as the United States-Israel Binational Science Foundation for a partial support through a
BSF grant.
Keywords: stationary reection, weak reection, successor of singular, 2-huge cardinal.
AMS 2000 Classication: 03E05, 03E35, 03E55.
2
6
9
1


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







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


stationary reection have been shown to have a connection with decidability
of monadic second-order logic ([Sh 80]).
We investigate the notion of weak reection, which, as the name suggests,
is a weakening of the stationary reection. For a regular cardinal , we say
that > weakly reects at i for every function f : , there is
< of conality (we say S

) such that f e is not strictly increasing

for any e a club of . The negation of this principle is a strong form of
non-reection, called strong non-reection. The notions were introduced by
Dzamonja and Shelah in [DjSh 545] in connection with saturation of normal
lters, as well as the guessing principle

(
+
), which is a relative of another
popular guessing principle, . It is proved in [DjSh 545] that, in the case
when =
+
and
0
< = cf() < , if weak reection of at holds
relativized to every stationary subset of S

, then

(S

) holds. The exact

statement of the principle is of no consequence to us here, so we omit the
denition. We simply note that this statement is stronger than just

(),
which holds just from the given cardinal assumptions.
Weak reection was further investigated by Cummings, Dzamonja and
Shelah in [CuDjSh 571], more about which will be mentioned in a moment.
A very interesting application of weak reection was given by Cummings
and Shelah in [CuSh 596], where they used it as a tool to build models where
stationary reection holds for some conalities but badly fails for others.
It was proved in [DjSh 545] that if there is which weakly reects at ,
the rst such is a regular cardinal. It is also not dicult to see that the rst
such cannot be weakly compact. On the other hand, in [CuDjSh 571] Cum-
mings, Dzamonja and Shelah proved that, modulo the existence of certain
large cardinals, it is consistent to have a cardinal which weakly reects
at unboundedly many regular below it, and strongly non-reects at un-
boundedly many others. The forcing notion used in this can be used to get
models where there is such that the rst cardinal weakly reecting at is
the successor of a regular cardinal.
A question we attempted in these investigations but did not succeed in
resolving was if it is consistent to have for which the rst which weakly re-
ects at is the successor of a singular cardinal. In this paper we answer this
3
6
9
1


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







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


question positively, modulo the existence of a certain large cardinal whose
strength is in the neighbourhood of (and seemingly less than) being 2-huge.
In our model both and are successors of singulars. Cummings has mean-
while obtained in [Cu] an interesting result indicating that it would be very
dicult to obtain a result similar to ours with e.g. =
2
and =
+1
,
as there is an interplay with a closely related compactness phenomenon. To
state our results more precisely, let us now give the exact denition of weak
reection and the statement of our main theorem.
Denition 0.1. Given
0
< = cf() and > . We say that weakly
reects at i for every function f : , there is S

such that f e
is not strictly increasing for any e a club of .
Theorem 0.2. (1) Let V be a universe in which, for simplicity, GCH holds
and let
0
be a cardinal such that there is an elementary embedding j : V M
with the following properties:
(i) crit(j) =
0
,
(ii) For some

which is the successor of a singular cardinal

and for some

, we have

0
<

<
1
def
= j(
0
) <

def
= j(

) < cf() = <

2
def
= j(
1
),
(iii)

M M.
Then there is a generic extension of V in which cardinals and conalities

0
are preserved, GCH holds above
0
, and the rst weakly reecting
at

is

(hence, the successor of a singular).

(2) In (1), we can replace the requirement that

is the successor of a singular

by (

) holds for any of the following meanings of (x):

(a) x is inaccessible,
(b) x is strongly inaccessible,
(c) x is Mahlo,
4
6
9
1


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







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


(d) x is strongly Mahlo,
(e) x is -(strongly) inaccessible for < x,
(f) x is -(strongly) Mahlo for < x,
and have the same conclusion (hence in place of

is the successor of a
singular, V
P
will satisfy (

)).
(3) With the same assumptions as in (1), there is a generic extension of V
in which

=
53
and

, the successor of a singular, is the rst cardinal

weakly reecting at

.
Remark 0.3. (1) Our assumptions follow if
0
is 2-huge and GCH holds.
The integer 53 in the statement of part (3) above is to a large extent arbitrary.
(2) Notice that

-cc forcing notions preserve that

is a regular uncountable
cardinal and that

is the rst cardinal weakly reecting at

, as well as
the fact that

is the successor of a singular cardinal, but not necessarily the

fact that

is the successor of a singular cardinal. It is natural to consider

the possibility of

=
+1
and

=
++1
but we have not considered
this for the moment.
The proof of (1) of the Theorem uses as a building block a forcing notion
used by Cummings, Dzamonja and Shelah in [CuDjSh 571], which introduces
a function witnessing strong non-reection of a given cardinal to a cardinal
. An important feature of this forcing is that it has a reasonable degree
of (strategic) closure, provided that strong non-reection of at already
holds for (, ), and hence it can be iterated. This forcing is a rather
homogeneous forcing, so the term forcing associated with it has strong deci-
sion properties. The forcing that we actually use is a term forcing associated
with a certain product of the strong non-reection forcing and a Laver-like
preparation. Using this, we force the strong non-reection of at

for all
<

(say >

, as the alternative situation is trivial), and the point is

to prove that in the extension

weakly reects on

. If we are given a
condition and a name forced to be a strongly non-reecting function, we can
use the large cardinal assumptions to pick a certain model N, for which we
5
6
9
1


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







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


are able to build a generic condition, whose existence contradicts the choice
of the name. To build the generic condition we use the preparation and the
fact that we are dealing with a term forcing. Proofs of (2) and (3) are easy
modications of the proof of (1).
We recall some facts and denitions.
Notation 0.4. (1) Reg stands for the class of regular cardinals.
(2) If p, q are elements of a forcing notion, then p q means that q is an
extension of p. For a forcing notion Q we assume that
Q
is the minimal
member of Q.
(3) For p a condition in the limit of an iteration P

, Q

, <

),
we let
Dom(p)
def
= <

: (p p() =
Q

).
(4) The statement that weakly reects at is denoted by WR(, ). Its
negation (including the situation ) is denoted by SNR(, ).
Remark 0.5. It is easily seen that weakly reects at i [[ does, so
we can without loss of generality, when discussing weak reection of to
assume that is a cardinal.
Denition 0.6. (1) For a forcing notion and a limit ordinal , we dene the
game (P, ) as follows. The game is played between players I and II, and it
lasts steps, unless a player is forced to stop before that time. For < , we
denote the -th move of I by p

, and that of II by q

. The requirements are

that I commences by
P
and that for all we have p

, while for <

we have q

.
I wins a play of (P, ) i lasts steps.
(2) For P and as above, we say that P is -strategically closed i I has a
winning strategy in (P, ). We say that P is (< )-strategically closed i it
is -strategically closed for all < .
1 Proofs.
We give the proof of Theorem 0.2. The main point is the proof of part (1) of
the Theorem. With minimal changes, this proof can be adapted to prove the
6
6
9
1


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







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


other parts of the theorem. The necessary changes are described at the end of
the section. Let V , j and the cardinals mentioned in the assumptions of the
Theorem be xed and satisfy the assumptions. Note that the elementarity
of j guarantees that

is the successor of a singular cardinal. We shall build

a generic extension in which

remains the successor of a singular cardinal

and is made to be the rst cardinal which weakly reects at

. Let us
commence by describing the forcing for obtaining strong non-reection at a
given (favourably prepared) pair of cardinals.
Denition 1.1. Suppose that we are given cardinals and satisfying

0
< = cf() < .
P(, ) is the forcing notion whose elements are functions p with dom(p)
an ordinal < , the range rge(p) , and the property
[ S

& dom(p)] = (c a club of ) [p c is strictly increasing],

ordered by extension.
Fact 1.2 (Cummings, Dzamonja and Shelah, [CuDjSh 571]). Let and
be such that P(, ) is dened, then
(1) [P(, )[ [
>
[ =
<
.
(2) Suppose that for all (, ) we have SNR(, ). Then P(, ) is
(< )-strategically closed.
(3) If GCH holds then P(, ) preserves cardinals and conalities.
We recall the following
Denition 1.3. A set A of ordinals is an Easton set i
Reg (sup(A) + 1) = sup(A ) < .
The following forcing notion enforces strong non-reection through an
iteration of forcing of the form P(, ).
7
6
9
1


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







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


Denition 1.4. Given
0
< cf() = < .
Q
(,)
is the result of a reverse Easton support iteration of P(, ) for
= cf() (, ). More precisely, let

Q = Q

, R

: , < ),
where
(1)
Q
R

= unless Reg (, ), in which case

Q
R

= P

(, ).
(2) For we dene by induction on that
p Q

i p is a function with domain such that for all < we

have p Q

and
Q
p() R

and letting
Dom(p) = < : (
Q
p() =
R

)
we have that Dom(p) is an Easton subset of .
(3) p q i for all < we have q
Q

q() p().
Fact 1.5 (Cummings, Dzamonja, Shelah, [CuDjSh 571]). Let

Q, and
be as in Denition 1.4. For all :
(1) Whenever is regular, [Q

[
<
,
(2)
Q
[R

[
<||
,
(3) If
0
, then Q
+1
has ([[
<||
)
+
-cc. In addition, if is strongly
Mahlo, then Q

has -cc.
(4)
Q
R

is (< )-strategically closed.
(5) For all < , we have that Q

/Q

is (< )-strategically closed.

8
6
9
1


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







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


(6) Q

preserves all cardinals and conalities ([[

<||
)
+
, and all strongly
inaccessible cardinals and conalities [[
2
, as well as all cardinals
and conalities ()
+
. If is strongly Mahlo, then Q

preserves
cardinalities and conalities .
(7)
Q
SNR(, ) for all < .
The forcing notion we shall use will be a term forcing associated with
iterations of the above type. We rst need some general notation for such
forcing.
Notation 1.6. (1) For a forcing notion Q of the form Q = P
1
P

2
, we
denote by Q

the term forcing associated with Q, dened by

Q

def
= (
P
1
, q

) : q

is a canonical P
1
-name for a condition in P
2
,
(in particular Q

Q), with the order inherited from Q. Following

the usual practice, we may write (, q) in place of (
P
1
, q

) when the
meaning is clear from the context.
(2) For a triple (R, , ) with cf() = < , and R a forcing notion
preserving the fact that > is a cardinal and = cf() >
0
, we
dene Q

(R,,)
to be [R Q

(,)
]

.
(3) In the situation when the notation Q

(R,,

)
makes sense, we abbreviate
it as Q

(R,)
.
Observation 1.7. Q

(R,,)
, when dened, is (<
+
)-strategically closed.
The forcing notion we shall use will have a preparatory component, P

described below, which will be followed by a component made up of term

forcings described above. We give a precise denition in the following.
2
Note that this implies that if GCH holds, all cardinalities and conalities are
preserved.
9
6
9
1


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







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


Denition 1.8. We dene P

to be the forcing whose elements are functions

h, with dom(h) an Easton subset of
0
consisting of cardinals, with the
property that
< dom(h)
0
= h() H()
3
.
The order on P

is the extension.
In order to be able to dene the next component of the forcing we need
to ascertain a preservation property of P

. Recall that we are assuming that

V [= GCH.
Claim 1.9. Forcing with P

preserves cardinals and conalities

0
, and
GCH above and at
0
. If p P

and dom(p) is (strongly) inaccessible,

then p forces that the conality of any
0
whose V -conality is > ,
remains > , while 2

remains
+
.
Proof of the Claim. First notice that [P

[ =
0
, so P

has
+
0
-cc and
preserves cardinals and conalities
+
0
, as well as GCH above and at
0
.
Now suppose that p P

and p forces that for some

0
and < cf()
with dom(p) inaccessible, the conality of in V [G] is . Let
P
<
def
= q : q P

& q p
and
P

def
= q [,
0
) : q P

& q p,
both ordered by the extension. The mapping q (q [,
0
), q ) shows
that P

/p
def
= q P

: q p is isomorphic to P

P
<
(we are using the
fact that dom(p)). We have that P

is (<
+
)-closed, so P
<
adds a co-
nal function from to . However, [P
<
[ (as is strongly inaccessible),
and so it preserves cardinals and conalities
+
, a contradiction.
We can similarly observe that
p
P
2

=
+
.

1.9
3
Note that this implies that each h() is bounded in
0
.
10
6
9
1


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







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


Denition 1.10. (1) For <
0
strongly inaccessible let R

and

be the
following P

-names:
for a condition p P

, if dom(p) and
p() = (, R) with < cf() = <
0
,
and R H(
0
) a forcing notion which preserves the fact that is a
regular uncountable cardinal (and it by necessity preserves that

is a
cardinal larger than ), then p forces

to be and R

to be Q

(R,

)
.
If Dom(p) but or p() do not satisfy the conditions above, or p
has no extension q with Dom(q), then p forces

= 0 and R

to
be the trivial forcing, which will for notational purposes be thought of
as (, ). In these circumstances we think of R

= .
It follows from Claim 1.9 that the above denition is correct and that over
a dense subset of P

each R

is a P

-name of a forcing notion from V ,

is
a P

-name of an ordinal <

0
. In the following item (2), clearly

<
0
R

is
a P

-name of a product of forcing notions from V , but R

below is forced not

to be from V .
(2) For a P

-name f

<
0
R

and
0
, let
A

,
def
= <
0
: f

() = (, q

) with (
R

/ Dom(q

)).
(3) Let R

be a P

-name for:

<
0
R

: (
0
) [A

,
is an Easton set ]

,
ordered by the order inherited from

<
0
R

.
We proceed to discuss the preservation properties of the forcing notions
we dened.
Notation 1.11. If we write (p, r

) P

, we mean that p P

and

P
r

= (
R
, r

()) : <
0
).
11
6
9
1


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







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


Denition 1.12. (1) Given (p, r

) P

and = cf() <

0
. For
(q, s

) P

, we dene
(i) (q, s

)
pr,
(p, r

) i
() (q, s

) (p, r

),
() q ( + 1) = p ( + 1),
() for <
0
with (q R

is trivial), we have
(q,
R

) if

< , then s

() (

, ] = r

() (

, ].
(ii) (q, s

)
apr,
(p, r

) i
() (q, s

) (p, r

),
() q ( + 1,
0
) = p ( + 1,
0
),
() for <
0
with (q R

is trivial), we have
(q,
R

) s

() (,

) = r

() (,

).
(2) For (p, r

) P

and = cf()
0
, we let
Q

(p,r

),
def
= (q, s

) : (q, s

)
apr,
(p

, r

) for some (p

, r

)
pr,
(p, r

),
ordered as a suborder of P

.
Claim 1.13. Given (p, r

) (q, s

) in P

, and a regular <

0
.
Then there is a unique (t, z

) such that
(p, r

)
pr,
(t, z

)
apr,
(q, s

).
Proof of the Claim. Let t
def
= p ( + 1) q ( + 1,
0
). Hence t P

and p t q (note that + 1 / dom(p)). We dene a P

-name z

by
letting for <
0
z

()
def
=

() (

, ] q

() (,

] if dened
r

() otherwise.

1.13
12
6
9
1


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







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


Notation 1.14. (t, z

) as in Claim 1.13 is denoted by intr((p, r

), (q, s

)).
Claim 1.15. For = cf() <
0
, the forcing notion (P

,
pr,
) is
(< + 2)-strategically closed.
Proof of the Claim. For every <
0
we have by Fact 1.5(5) applied to

+
(or +2, recalling that for coordinates (,
+
) in the iteration Q

)
the factor R

is trivial)

P
R

non-trivial = Q

)
/Q

,]
is (< + 2)-strategically closed.
Hence we can nd names St

of the winning strategies exemplifying the

corresponding instances of the above statement.
Suppose that +1 and p

= p
0

, p

) : < ), q

= q
0

, q

) : < )
are sequences of elements of P

such that
(1) For all < we have p

pr,
q

,
(2) For all < and < we have q

pr,
p

and for <

0
with
(p
0

is trivial), we have (p
0

, (
R

, p

() (

, ]))
P

,]
p

() (,

) = St

(p

() (,

) : < ), q

() (,

) : < ))
We dene p

as follows. First let p

0

def
=

q
0

: < . Notice that p

0

and p
0

( + 1) = p
0
0
( + 1).
For <
0
with (p

is trivial), we let p

() be the name given

by
p

() (

, ]
def
= p

1
0
() (

, ]
and
p

() (,

)
def
= St

(p

() (,

) : < ), q

() (,

) : < )).
The conclusion follows because we have just described a winning strategy for
I in ((P

,
pr,
), ).
1.15
13
6
9
1


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







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


Claim 1.16. Suppose (p, r

: Ord, where is regular <

0
.
Then there is (q, s

)
pr,
(p, r

) and a Q

(q,s

),
-name

such that
(q, s

.
Proof of the Claim. We dene a play of ((P

,
pr,
), +1) as follows.
Player I starts by playing (p, r

)
def
= p
0
. At the stage , player II
chooses q

such that q

forces a value to

(), and we let q

def
= intr(p

, q

).
At the stage 0 < < , we let I play according to the winning strategy for
((P

,
pr,
), +1) applied to (p

: < ), q

: < )). At the end, we

let (q, s

) = p

. This process denes

by letting

() be the Q

(q,s

),
-name
such that for (q

t) Q

(q,s

),
we have
(q

t)

() =

().
Note that q

(q,s

),
and

is a Q

(q,s

),
-name.
1.16
Claim 1.17. If (p, r

) P

, and <
0
is inaccessible with dom(p),
then Q

(p, r

),
has
0
-cc.
Proof of the Claim. Given q = q
i
= q
0
i
, q

1
i
) : i <
0
), with q
i
Q

(p, r

),
.
Suppose for contradiction that the range of this sequence is an antichain.
We have that for all i <
0
dom(q
0
i
) ( + 1,
0
) dom(p) ( + 1,
0
).
As dom(p) is an Easton set, without loss of generality we have that all
dom(q
0
i
) ( +1,
0
) are the same. If this set has the largest element, let us
denote its successor by . Otherwise, let
def
= [ sup[dom(p) ( + 1,
0
)][
+
.
In either case, we have that for all i and dom(q
0
i
(+1,
0
)) the relation
q
0
i
() H() holds. Hence for all i <
0
q
0
i
() : dom(q
0
i
) ( + 1,
0
)) H(),
as dom(q
0
i
) is an Easton set, so without loss of generality all q
0
i
are the same.
As dom(p), for each i we have that dom(q
0
i
) and hence q
0
i
H()
14
6
9
1


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







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


and we can assume that all q
0
i
are the same, and hence all q
0
i
are the same
condition in P

, which we shall call q

. Let G

be P

-generic over V with

q

. Hence in V [G

] the sequence q
1
i
def
= (, q

i
) : i <
0
) is an antichain
in

<
0
[R

(,

)
]

, and by the choice of the initial sequence, we have that

(, q

i
() (+1)) : <
0
) : i <
0
) is an antichain in

<
0
[R

(,)
]

.
For every i <
0
,
A
i
def
= <
0
: q

( + 1) ,=
is the union of Easton sets. Hence by Claim 1.9 (i.e. as in V [G

] the
cardinal
0
is still strongly Mahlo), without loss of generality, A
i
s form a
-system with root A

. Note that A

is a bounded subset of
0
. We can
conclude that
, q

( + 1)) : sup(A

)) : i <
0
)
forms an antichain in

sup(A

)
[R

(,)
]

, in contradiction with Fact 1.5

(in fact the number of possible values is <
0
.
1.17
Claim 1.18. Forcing with P

preserves cardinals and conalities

0
.
Proof of the Claim. Suppose conalities
0
are not preserved and let
be the rst conality
0
destroyed. Hence is regular, and for some

, condition (p, r

) and some regular < , we have (p, r

: is
conal.
Case 1. <
0
. Without loss of generality, is inaccessible, and by
increasing and p if necessary, we can assume that dom(p). By Claim
1.16, there is (q, s

) (p, r

) and a Q

(q, s

),
-name

such that (q, s

.
Hence (q, s

: is conal, contradicting the fact that Q

(p, r

),
satises
0
-cc, which follows from Claim 1.17.
Case 2.
0
.
As for every <
0
with R

non-trivial we have

)
/Q

,]
is (< ( + 1))-strategically closed,
15
6
9
1


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







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


there is (q, s

) (p, r

) and a P

<
0
[R

,]
]

-name

such that
(q, s

: is conal. But this forcing has

+
-cc, a contradiction.

1.18
Combining Claim 1.18 with Fact 1.5, we obtain
Corollary 1.19. Forcing with P

)
preserves cardinalities and
conalities
0
, it preserves GCH from
0
on, and forces SNR(,

) for
(

).
By our assumptions it follows that the predecessor of

is singular, so

remains the successor of a singular after we have forced by P

)
,
and clearly

remains regular.
Now we have arranged the situation so that we are left with the main
point of the argument, which is that after forcing with P

)
we
shall have weak reection of

at

. The basic forcing for enforcing strong

non-reection and the actual forcing we have used posses a convenient ho-
mogeneity property that will become relevant in the main argument, so we
formulate these in the following. As the proof below suggests, the homogene-
ity these forcing notions have is actually stronger than the mild homogeneity
dened in the following denition, but as all we need is mild homogeneity,
we formulate our claims using that notion.
Denition 1.20. We shall call a forcing notion P mildly homogeneous i for
every formula (x
0
, . . . , x
n1
) of the forcing language of P and a
0
, . . . , a
n1
(canonical names of) objects in V , we have
P
| (a
0
, . . . , a
n1
).
Claim 1.21. (1) P(, ) is mildly homogeneous, for
0
< cf() = < ,
(2) Q
(,)
is mildly homogeneous, for
0
< cf() = < .
(3) If R is mildly homogeneous and Q

(R,,)
is well dened, then Q

(R,,)
is
mildly homogeneous.
(4) It is forced by P

that R

is mildly homogeneous.
16
6
9
1


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







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


(5) It is forced by P

that R

)
is mildly homogeneous.
Proof of the Claim. (1) Suppose not, and let p, q P
def
= P(, ) force
contradictory statements about (a
0
, . . . , a
n1
) for some a
0
, . . . , a
n1
V .
Let = dom(q) and consider the function F = F
(,)
: P P such that
F(f) = g i q g and for i dom(f) we have g( +i) = f(i).
This function is an isomorphism between P and P/q
def
= g P : g q,
so it induces an isomorphism between the canonical P-names for objects in V
and the canonical P/q-names for the same objects. In particular, F(p) forces
in P/q the same statements about a
0
, . . . , a
n1
that p does in P. If G is
P-generic over V such that F(p) G (then also q G ). As q G and F(p)
force contradictory statements about a
0
, . . . , a
n1
, we obtain a contradiction.
(2) Suppose that p, q Q = Q
(,)
force contradictory statements about
(a
0
, . . . , a
n1
) for some a
0
, . . . , a
n1
V . We dene a function F
+
: Q Q
so that such that F(f) = g i Dom(g) = Dom(q)Dom(f) and for Dom(g)
we have that q g() = F
(,)
(p()), where F
(,)
is dened as in (1)
above. One can now check that F
+
is an isomorphism between Q and Q/q,
and then the conclusion follows as in (1).
(3)-(5) Similar proofs.
1.21
Remark 1.22. The homogeneity properties discussed in Claim 1.21 are not
enjoyed by the forcing notion P

.
Main Claim 1.23. After forcing with P
def
= P

)
, the weak reec-
tion of

holds at

.
Proof of the Main Claim. Suppose otherwise, and let p

= (p, q

, r

) force

to be a function exemplifying the strong non-reection of

at

. As
R

)
is forced to be mildly homogeneous by Claim 1.21, without loss of
generality p

= (p, , ). We proceed through a series of lemmas that taken

together suce to prove the Claim.
Lemma 1.24. There are cardinals , and

and a model N H() such

that
17
6
9
1


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







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


(i) N
0
is an inaccessible cardinal <
0
,
(ii) otp(N

) =

,
(iii) otp(N
1
) =
0
,
(iv)
>
N N,
(v) (N, ) is isomorphic to H(

) for some regular

< ,
(vi) [N

[ is a regular cardinal <

0
, in fact otp(N

) = ,
(vii)

,
0
,
1
,

, P, p

N.
Proof of the Lemma. We use the notation of the main Theorem 0.2, in
particular M described in the assumptions of the Theorem. Since we assume
that

M M and
<
= we conclude that j(H()) M. Now we can
use elementarity to obtain N V whose relationship to H() mirrors that
of the relationship between j(H()) and j(H()) in M. In M we have that
j(H()) j(
0
) =
0
since the critical point of j is
0
, so in M this is an
inaccessible cardinal < j(
0
), and so we can obtain that N satises (i) above.
Requirement (ii) follows because in M we have
j(

) j(H()) = j(

)
since

< , and hence this set has order type

= j(

). We argue similarly
for requirement (iii), using that
1
< and j(
0
) =
1
. For (iv) we choose
the denition of as the mirror image of
0
and the fact that j(H())
is closed under sequences of length <
0
, as
0
is the critical point of the
embedding. For (v) we use the fact that j(H()) is isomorphic to H(),
which can be said in M because M is closed under sequences, and (vi) and
(vii) are handled similarly to the above.
1.24
We x N as guaranteed by Lemma 1.24 and let
def
= sup(N

). Some
simple consequences of the choice of N are given in the following
Lemma 1.25. The following hold
18
6
9
1


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







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


(i) S

,
(ii) N is a stationary subset of , and it remains such after forcing with
P.
Proof of the Lemma. The rst claim holds because otp(N

) is

. As
for (ii), notice that the set E dened as the closure of N is a club of .
Letting S
def
= S

0
N we have that [ E & cf() =
0
] = S (the
analogue of this is true even withcf() < in place of cf() =
0
). As
P is an (<
1
)-closed forcing notion, S remains stationary after forcing with
P and hence so does N .
1.25
The next step of our argument is to extend p to a P

-generic condition
over N, which is done as in the following. We use the notation Most
N
to
denote the function of the Mostowski collapse of the structure (N, ).
Note that one of the properties of N is that it is isomorphic to H(

) for
some regular cardinal

< , in particular the Mostowski collapse of N is

H(

), and it follows from the other properties of N that

H(

) and

= Most
N
(

).
As p

N and N
0
= , we have that dom(p) and since P

N
also for all dom(p) we have p() H(). Hence we can extend p to the
condition
p
+
def
= p , (, Most
N
((P

)
N
))).
Note that this is a well dened element of P

because <
0
(see clause
(vi)) and Most
N
((P

)
N
) is a well dened element of H(
0
) and a forcing
notion that preserves the fact that is a regular uncountable cardinal. In
fact we have
Claim 1.26. The condition p
+
a well dened extension of p which is P

-
generic over N.
Proof of the Claim. We have already discussed the fact that p
+
is a
well dened extension of p. By the choice of its last coordinate it is actu-
ally P

-generic over N, since any condition q p

+
will satisfy that for all
N dom(q) we have q() H().
1.26
19
6
9
1


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







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


Claim 1.27. The condition p
+
forces that

= and R

= Most
N
((P

)
N
),
hence that R

is [Most
N
((P

)
N
) Q

(,

)
]

.
Proof of the Claim. This follows by the denition of

and R

.
1.27
Claim 1.28. The condition (p
+
, ) is P

-generic over N.
Proof of the Claim. This follows by the previous claims and the denition
of term forcing.
1.28
Using the denition of the Mostowski collapse, one can establish a con-
nection between Q
N
(

)
and Q
(,

)
. Namely, let F be the inverse of the
Mostowski collapse of N

, so an order preserving function from

onto
N

. We also let F(

) =

. If N = cf() (

), then F
1
()
is an ordinal in (,

) and if for such we have N < is an ordinal,

then F
1
() is an ordinal < F
1
(). For such , if N r is a function from
to

, then
(, F
1
()) : < F
1
() & N r(F()) =
is a function from F
1
() into F
1
(

). Continuing in a similar fashion,

we can see that F
1
induces a mapping from P(

, )
N
into P(, F
1
()),
which we again denote by F
1
. Suppose that N

and A is an
Easton set, then F
1
()

and F
1
(A) F
1
() is an Easton set, for
if Reg sup(F
1
(A)) + 1 then N F() Reg sup(A) + 1 and so
sup(F
1
(A) ) < . This shows that F
1
induces a mapping from Q
N
(

)
into Q
(,

)
, which we again denote by F
1
.
This process can be pushed one step further. As p
+
is P

-generic over
N, it forces that N[G

] V = N, so using that P

N and an analysis
similar to the one carried so far, one shows that over p
+
it is forced that
[Most
N
((P

)
N
) Q

(,

)
]

is Most
N
(([(P

) Q

)
]

)
N
) and hence
that there is a function F
1
that induces a mapping between the P

-names
20
6
9
1


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







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


for conditions in the latter forcing and those in [(P

) Q

)
]

. This
now shows that p
+
forces that R

is Most
N
(([(P

) Q

)
]

)
N
).
Let H

be the canonical P

-name for a subset of Most

N
([(P

)Q

)
]

)
N
)
such that p
+
forces H

to be Most
N
([(P

) Q

)
]

)
N
)-generic. The
Mostowski collapse induces a mapping that maps H

into a P

-name for a
subset H

of ([(P

) Q

)
]

)
N
.
We proceed to dene q as follows: q
def
= (p
+
,
R
, r

), where r

is a P

-name
over (p
+
,
R

) of a condition in Q

)
dened by letting
Dom(r

)
def
=

Dom(h

) : p
+

P
(
R

, h

) H

,
and for with (p
+
,
R

) Dom(r

), we let
r

()
def
=

() : (p
+
,
R

) Dom(h

) & h

.
We have to verify that q is a condition in P and we also claim that q p

.
Let us check the relevant items by proving a series of short Lemmas:
Lemma 1.29. If (p
+
,
R

) is strongly inaccessible (

), then
(p
+
,
R

) [Dom(r

) [ < .
Proof of the Lemma. We have that (p
+
,
R

) forces that
Dom(r

Dom(f

) : f

and [H

< (as [N

[ =

), so Dom(r

) is forced to be bounded
in .
1.29
Lemma 1.30. If (p
+
,
R

) Dom(r

), then (p
+
,
R

) r

() is a func-
tion whose domain is an ordinal < and range a subset of

.
Proof of the Lemma. As (p
+
,
R

) H

is directed, we have that

(p
+
,
R

) r

() is a function. If
(p
+
,
R

) Dom(h

) & F(h

) H

,
21
6
9
1


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







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


then (p
+
,
R

) forces
( Dom(F(h

))) [F(h

)() is a function with domain ],

so by the denition of F and the fact that [H

< = cf() we have

(p
+
,
R

) dom(h

()) is an element of
and (p
+
,
R

) dom(r

()) is an element of .
1.30
Lemma 1.31. (p
+
,
R

) r

)
.
Proof of the Lemma. We know by Lemma 1.30 that (p
+
,
R

) Dom(r

is an Easton set. It is forced by (p

+
,
R

) that for relevant the set dom(r

())
is the union of a subset of H(
++
)N which has cardinality [
++
N[ <

,
and clearly h() : h H

has no last element, by density and genericity.

Hence the sup of this union has conality <

(as having conality

is
preserved by the forcing P

), so r

() is forced to be in P

(,

), and hence
r

is forced to be an element of Q

)
.
1.31
Lemma 1.32. (p
+
,
R

) forces that if Dom(r) then for all S

, there
is a club e of on which r

() is strictly increasing, for dom(r

).
Proof of the Lemma. Again modulo what is forced by (p
+
,
R

), the set
h() : h H

, Dom(h) is linearly ordered, as the domains are ordinals,

so the conclusion follows because the analogue holds for each h H

.
1.32
Conclusion 1.33. The condition q = (p
+
, , r

) is an element of P and is
above any condition of the form (p
+
, , h) such that (p
+
, ) forces that h H

.
Consequently, q is P-generic over N.
We shall now see that q forces

to be constant on a stationary subset of

, a contradiction, as S

, and remains there after forcing with P. We

need to consider what q forces about

() for N

. Such

() is a
22
6
9
1


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







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


P-name of an ordinal <

, by the choice of

, see the beginning of the proof

of the Main Claim. Let
1

def
= (, , t

) P : (, , t

) forces

() to be equal to a P

-name.
Hence 1

N. As P

forces that Q

)
is (

+ 1)-strategically closed
(this is by Fact 1.5(5) applied to any (

+2, (

)
+
), and as the cardinality
of P

is

, we have that 1

is dense in [(P

) Q

)
]

. By the
denition of H

, there is (, , h

) 1

N such that (, , h

) q. Let

exemplify this, so

N.
Hence q forces

() to be in the set of all

N[G

], where

is a
P

-name of an ordinal <

. The cardinality of the set

T =

N :

is a P R

-name for an ordinal <

is forced to be [T(

) N[, which is <

0
. Since N was arbitrary, q
forces the range of

(N ) to be a set of size <

0
<

, hence

will
be constant on a stationary subset of N (as N is stationary). More
elaborately, one checks the stationarity in V
P

)
. Forcing with Q
(

)
adds no subsets to

, hence is irrelevant. Forcing with P

preserves the
uncountability of conalities, so as N contains e : cf() =
0
for
some club e of , clearly N is stationary in V
P

)
, and

is still
a cardinal, hence we are done.
1.23
Part (2) of Theorem 0.2 Same proof.
Part (3) of Theorem 0.2 In short, follow the forcing from (1) by a Levy
collapse. We are making use of the following
Claim 1.34. Suppose weakly reects at and P is a -cc forcing.
Then weakly reects at in V
P
.
Proof of the Claim. Suppose that
p
P
f

: & E

is a club .
We dene f

: by letting f()
def
= sup < : (p f

() ,= ).
As P is -cc, the range of f

is indeed contained in . Let S

be such
23
6
9
1


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







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


that f

S is constant on a stationary set S (the existence of such

follows as WR(, ) holds). Hence p f

S is bounded, so f

does not
witness SNR(, ) in V
P
, as S remains stationary.
1.34
So, for example to get

=
n
, we could in V
P
from (1) rst make GCH
hold below
0
by collapsing various cardinals below
0
, and then collapse

to
n
.

0.2
24
6
9
1


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







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


References
[Ba] J.E. Baumgartner, A new class of order types, Annals of Mathe-
matical Logic 9 (1976), 187-222.
[Cu] J. Cummings, On strong non-reection, private communication.
[CuFoMa] J. Cummings, M. Foreman and M. Magidor, Squares, scales
and stationary reection, Journal of Mathematical Logic 1 No.
1 (2001) 35-98.
[CuSh 596] J. Cummings and S. Shelah, Some independence results on re-
ection, Journal of the London Mathematical Society 59 (1999)
37-49.
[CuDjSh 571] J.Cummings, M. Dzamonja and S. Shelah, A consistency result
on weak reection, Fundamenta Mathematicae 148 (1995), 91-100.
[DjSh 545] M. Dzamonja and S. Shelah, Saturated lters at successors of sin-
gulars, weak reection and yet another weak club principle, Annals
of Pure And Applied Logic 79 (1996) 289-316.
[HaSh 99] L. Harrington and S. Shelah, Some exact equiconsistency results
in set theory, Notre Dame Journal of Formal Logic, 1985.
[Je] R.B. Jensen, The ne structure of the constructible hierarchy, An-
nals of Mathematical Logic 4 (1972), 229-308.
[KuVa] K. Kunen and J. Vaughan (eds.), Handbook of Set-theoretic Topol-
ogy, North-Holland 1983.
[La] R. Laver, Making the supercompactness of indestructible under
-directed closed forcing, Israel Journal of Math, 29 (1978), 385-
388.
[Ma] M. Magidor, Reecting Stationary Sets, Journal of Symbolic Logic
47 No. 4 (1982) 755-771.
25
6
9
1


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







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


[Sh 80] S. Shelah, A weak generalization of MA to higher cardinals, Israel
Journal of Mathematics, 30 (1978), 297-360.
[Sh 667] S. Shelah, Not collapsing cardinals in (< )-support iterations
II, in preparation.
26