Saharon Shelah - A More General Iterable Condition Ensuring 1 Is Not Collapsed

(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

A MORE GENERAL ITERABLE CONDITION
ENSURING
1
IS NOT COLLAPSED
SH311
Saharon Shelah
Institute of Mathematics
The Hebrew University
91 904 Jerusalem, Israel
Rutgers University
Mathematics Department
New Brunswick, NJ USA
Abstract. In a self-contained way, we deal with revised countable support iterated
forcing for the reals; improve theorems on preservation of a property UP, weaker than
semi proper, and hopefully improve the presentation. We continue [Sh:b, Ch.X,XI]
(or see [Sh:f, Ch.X,XI]), and Gitik Shelah [GiSh 191] and [Sh:f, Ch.XIII,XIV] and
particularly Ch.XV; concerning no new reals see lately Larson Shelah [\LarSh:746
]. In particular, we fulll some promises from [Sh:f] and give a more streamlined
version.
F144, based on Chapter XV of PIF
I would like to thank Alice Leonhardt for the beautiful typing.
Previous Version - 01/Feb/27
Latest Revision - 06/July/25
Typeset by A
M
S-T
E
X
1
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

2 SAHARON SHELAH
Annotated Content
1 Preliminaries, p.5
[We agree that forcing notion P has actually also pure (
pr
) quasi-order
and very pure (<
vpr
) quasi-order. For a -increasing sequence

Q of forcing
notions we dene what is a

Q-named and a

Q-named [j, )-ordinal. Then
we dene Sp
e
(W)-iterations (revised support of size < , including
the case inaccessible) with nite apure support, countable pure support
(the revised version) and Easton or W-Easton very pure support, similar
to [Sh:f, XIV] and prove its basic properties (this is done by simultaneous
induction).]
2 Trees of Models, p.31
[We quote the basic denitions and theorems concerning trees with levels
tagged by ideal and partition theorems.]
3 Ideals and Partial Orders, p.36
[We can replace the families I of ideals by corresponding partial orders or
quasi orders (we ignore the distinction). This is essentially equivalent
(for some -complete I with for some -complete L) but the Ls have
better pullback from forcing extensions, so we can replace L
in a forcing
extension of V by L
in V preserving L
RK
L
and preserving the

amount of completeness we have, so a similar situation holds for a set of
ideals; in the cases we have in mind here increasing those sets I or L do
not matter.]
4 UP Reintroduced, p.42
[We dene when

N is an I-tagged tree of models, when it is I-suitable, or
(I, W)-suitable, and when it is strictly or -strictly, etc., where I is a family
of ideals. Similarly we dene N is -strictly (I, S, W)-suitable; i.e. can serve
as N
and prove some basic properties. Such models will fulll here the
role that any countable N (H (), ) fullls in theorems on semi-proper
iteration. Lastly, we dene when a forcing notion P satises UP
(I, W) for
= 0, 1, 2, and here UP
1
, UP
2
replace W-proper, W-semi-proper, where
W is a stationary subset of
1
. All those properties imply that forcing
by P does not collapse
1
, preserve the stationarity of W and even of
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

GENERAL ITERABLE CONDITION 3
any stationary subset of it. They are all relatives of semi-properness for
strictly (I, W)-suitable models, they speak on I-tagged trees of countable
models.]
5 An Iteration Theorem for UP
1
, p.53
[We prove that satisfying UP
1
(W), i.e. satisfying it for some family I
of ideals, complete enough, is preserved by
1
Sp
6
(W)-iterations,

Q =
P
i
, Q
i
: i < ); that is if each Q
i
is like that then P
=
1
Sp
6
(W)-
Lim(
Q) is like that, provided some mild condition holds (say Q

i
is
UP
1
(I
i
, W), I
i
is
i
-complete, P
i
satises the
i
-c.c.; we can even make I
i
,
i
to be just P
i+1
-names, see there). The proof is more similar to the proofs
of preservation of properness and semi-properness than with the proofs in
[Sh:b, XI], (=[Sh:f, XI]), [GiSh 191], [Sh:f, XV], and hopefully more trans-
parent. The proof will be non-trivially shorter if we use just the particular
case of the revised countable support (i.e.,
vpr
is equality and
pr
is ).
We give a sucient condition for not being collapsed by P
e.g. is
strongly inaccessible, < [P
[ < and: W stationary in (so is

Mahlo) or
vpr
is equality and the iteration is suitable enough. Lastly, if
e.g. is the rst strongly inaccessible, i < [P
i
[ < we give a sucient
condition for not being collapsed.]
6 Preservation of UP
0
, p.72
[Here we make the condition more similar to semi-proper iteration, that is
the demand is that for suitable models N (one on which the right trees
grow) above each p Q N there is an (N, Q)-semi-generic q. There is
some price though.
[?] However, if Q satises UP
0
and the -c.c., then Q Levy
(
1
, < ) is
appropriate in our iteration.]
7 No New Reals - replacements for completeness, p.91
[Here we deal with the parallel of Q add no new real because it is
W-complete for some stationary W
1
.]
8 Examples, p.98
[We show that various forcing notions fall under our context. In particular
(variants of) Namba forcing, shooting a club through a stationary S <
: cf() =
0
where = cf() >
0
, and prove that the older condition
from [Sh:f] implies the present one.]
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

4 SAHARON SHELAH
9 Reection in [
2
]
0
, p.104
[We answer a question of Jech, on the consistency of 2
0
=
2
+D
1
is
2
-
saturated + every stationary subset of [
2
]
0
reects and there is a special
projectively stationary subset of [
2
]
0
.]
10 Mixing nitary norms and ideals, p.110
[We consider a common generalization of creature forcing (see [RoSh 470])
and relatives of Namba forcing.]
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

1 Preliminaries
1.1 Denition/Notation. 1) A forcing notion here, P, is a nonempty set (abusing
notation, it too is denoted by P) and three partial orders
0
,
1
,
2
(more exactly
quasi-orders and
=
P
) and a minimal element

P
P (so
P

p for p P) and
for = 0, 1 we have [p
q p
+1
q]. We call p P very pure if
P

0
p and
we call q a very pure extension of p if p
0
q. We call p P pure if
P

1
p and we
call q a pure extension of p if p
1
q. Let be
2
, let
pr
be
1
and
vpr
be
0
.
We call P -vp-complete if: for any <
vpr
-increasing sequence p
: < ), <
with p
0
=
P
there is a
vpr
-upper bound p. We dene vp--complete similarly
waiving p
0
=
P
. We dene
-complete and
-complete similarly.
The forcing relation, of course, refers to the partial order . We denote forcing
notions by P, Q, R. Let P
1
P
2
mean p P
1
p P
2
,
P
1
=
P
2
P
1
for = 0, 1, 2
and let P
1
ic
P
2
means P
1
P
2
and for 2, if p, q P
1
are
-incompatible in
P
1
, then they are
-incompatible in P
2
. Let P
1
P
2
means P
1
ic
P
2
& (P
1
,
) (P
2
, ).
2)

P denotes a -increasing sequence of forcing notions.

Q denotes a sequence of
the form P
i
, Q
i
: i < ) such that P
i
: i < ) is a -increasing sequence. Usually
Q
i
is a P
i
-name,
P
i
P
i+1
/P
i

= Q
i
.
3) Convention: If

Q = P
i
, Q
i
: i < ), P
i
is -increasing, we may write

Q instead
of P
i
: i < ).
4) For a forcing notion P (as in part (1)) we dene

P:
(a) the set of elements of

P is
_
A :A P, and for some p A (called a witness) we have
(i) (q A)(r)[q r A & p
vpr
r]
(ii) there is an upper bound p
P of A such that p
vpr
p
moreover (p
A)(p
vpr
p
vpr
p
)
_
(we call p
an outer witness for A or for A

P if clause (ii) hold), and
(b)

P is ordered by: A B i: A = B or A = or for some q B, (p
A)(p q) and we call q a witness to A B
(c) we dene the order
on

P by: A
B i A B and A ,= B implies that

for every witnesses p for A and every witness q for B we have p A & q
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

6 SAHARON SHELAH
B & (p
)(p
vpr
p
A p
q); we call such a pair (p, q) a witness

for A
B. [See 1.10(5)]
(d) we stipulate sometime
A for every A

P or =
P
[Saharon].
5) CC(P) is the minimal regular uncountable cardinal such that P satises the
-c.c. We may add
(iii) if p
A satises clause (i) + (ii) then there is p
A such that p
vpr
p
& p
vpr
p
.
1.2 Observation.: 1) For any forcing notion P, (as in 1.1(1), of course), also

P is a
forcing notion (in particular
is a quasi order for 2) and P

ic

P and P

P
and P is
vpr
-dense in

P when we identify p and p.
2) If A
i

B for i < i
then B
vpr
B
+
where B
+
=
_
i<i
A
i
B.
3) If 0, 1, 2 and (P,
) is -complete (i.e., an increasing sequence of length

< has an upper bound) then so is (
P,
).
Proof. 1) Check.
2) Easy.
3) If < , A
i
: i < ) is
-increasing let (p
i
, q
i
) witness A
i

P. If p
i
: i < )
is eventually constant then A
i
: i < ) is eventually constant and A
j
for j large
enough can serve. If not, without loss of generality (i < )p
i
,= p
i+1
and let (p
i
, q
i
)
witness A
i

A
i+1
. Clearly q
i
: i < ) has a
-upper bound in P, call it q. Now

q is as required.
1.2
1.3 Denition. Let MAC(P) be the set of maximal antichains of the forcing
notion P.
1.4 Remark. 1) Note: [MAC(P)[ 2
|P|
, P satises the [P[
+
-c.c. and if P satises
the -c.c. then [MAC(P)[ [P[
<
.
2) Note:
() if Q is a forcing notion, =
<
> [Q[ +
0
,
Q
( < )
0
< and
Q
= Q Levy(
1
, < ) then [MAC(Q
)[ = [Q
[ = .
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

1.5 Notation. Car is the class of cardinals.
IRCar is the class of (innite) regular cardinals.
RCar = IRCar 1.
URCar is the class of uncountable regular cardinals.
D
cb
is the lter of co-bounded subsets of .

D
is the club lter on for regular uncountable.
= (g() 1) for a nite sequence of length > 0.

1.6 Notation. 1) H () is the family of sets with transitive closure of power < ;
let <
be a well ordering of H ().

2) Let W be a function from the set of strongly inaccessible cardinals to 0, 1,
1
2
;
if / Dom(W) we understand W() = 0 and let W means W() = 1.
1.7 Denition. 1) Assume

P is a -increasing sequence of forcing notions.
Let
Gen
r
(
P) =:
_
G : for some (set) forcing notion P
we have
i<
P
i
P
and for some G
generic over V we have

G = G
_
i<
P
i
_
.
2) If

Q = P
i
: i < ) or

Q = P
i
, Q
i
: i < ) where P
i
is a -increasing we
dene a

Q-name
almost as we dene (
_
i<
P
i
)-names, but we do not use maximal
antichains of
_
i<
P
i
, that is:
()
is a function, Dom(
)
_
i<
P
i
and for every G Gen
r
(
Q),
[G] is dened
i Dom(
) G ,= and then
[G] V[G] [from where every G. . . is

taken? E.g., V is countable, G any set from the true universe] and
[G]
is denable with parameters from V and the parameter
_
i<
P
i
G (so
is really a rst-order formula with the variable

_
i<
P
i
G and parameters
from V).
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

8 SAHARON SHELAH
Now
Q
has a natural meaning.
3) For p

Q (i.e. p
_
i<
P
i
) and

Q-names
0
, . . . ,
n1
we let
0
, . . . ,
n1
be
the name for the set that contains exactly those
i
[
Q] that are dened. We let

p
= x if for every G such that p G Gen

r
(
Q) we have
[G] = x. If <
and G
, we let
[G
] = x means that for some p G
we have p
Q

= x,
so possibly no p
_
i<
P
i
forces a value to
and no such p forces
is not denable.
4) We say a

Q-name x
is full if x
[G] is well dened for every G Gen

r
(
Q).
5) A simple

Q-named
1
[j, )-ordinal
is a

Q-name
such that: if G Gen

r
(
Q)
and
[G] = then j < and (p G P
)p
Q

= (where =
g(
Q)); however, we allowed
[G] to be undened. If we omit [j, ) we mean

[0, g(
Q)) = [0, ). If we omit simple, we mean replacing (p G P
) by
(p G P
(+1)
) (this is used in [Sh:f, Ch.X,1], we shall only remark on it
here).
6) A simple

Q-named
2
[j, )-ordinal
is a simple

Q-named
2
[j, )-ordinal of depth
for some ordinal , where this is dened below by induction on . In all cases
it is a

Q-name of an ordinal from the interval [j, ) so may be undened, i.e., we
allow non full such names.
Case 1: = 0.
This is an ordinal [j, ), or is undened (in the full case this is forbidden).
Case 2: > 0.
For some < g(
Q) and maximal antichain I = p
: <
of P
, there is
a sequence
: <
) such that
is a simple

Q-named
2
[Maxj, , )-ordinal
of depth
< and:
[G
] = i and for some we have p
and
[G
] = (including the case: not dened). If we omit [j, ) we mean

[0, g(
Q)) = [0, ).
7) If we omit simple in (6) we mean that in case 2,
i
is a not necessarily simple
Q-name
2
and I P
+1
.
8) We say

P is W-continuous or (
P, W) is continuous when for every W g(
P)
if (i < ) [density (P
i
) < , or just P
i
satises the cf()-c.c.], then P
=
_
i<
P
i
.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

We say

P is W-smooth or (
P, W) is smooth if W P
=
_
i<
P
i
. We say
is a simple
Q, W)-named [j, )-ordinal if it is a simple

Q-named ordinal and
W (g(
Q) + 1) ( < )(
/ [, )).
1.8 Claim. 1) Assume that

Q is W-continuous. If
is a simple

Q-named
2
[0, )-
ordinal, W is regular and < implies density(P
) < or just (P
satises
the cf())-c.c.), then for some < ,
is a simple

Q -named
2
[0, )-ordinal.
2) If

Q is W-continuous and W regular and
is a simple

Q-named
2
[0, )-
ordinal then
is a simple (
Q, W)-named
2
[0, )-ordinal.
3) If
is a simple

Q-named
2
[0, )-ordinal then there is a full simple

Q-named
2
[0, )-
ordinal
such that
Q
if
is well dened then it is equal to
.
Proof. By induction on the depth of
.
1.8
1.9 Remark. 1) We can restrict in the denition of Gen
r
(
Q) to P
in some class K,
and get a K-variant of our notions.
2) Note: even if in 1.7(1) we ask Dom(
) to be a maximal antichain of
_
i<
P
i
it will
not be meaningful as in the appropriate P
, we have
i<
P
i
P
but not necessarily

_
i<
P
i
P
hence it will not in general be a maximal antichain.

3) Note that in the simple case we wrote P
not P
(+1)
. Compare this with
the remark [Sh:f, Ch.XIV,1.1B]. Here in the main case we use full simple

Q-named
2
ordinals, though we shall remark on the aect of the non-simple case; as a result
we will not have a general associativity law, but the denition of Sp
3
Lim
Q)
will be somewhat simplied. As said earlier, we can interchange decisions on this
matter. Of course, also [Sh:f, Ch.XV] can be represented with this iteration.
4) The name
1
is necessary for the >
1
case, but name
2
is preferable for
=
1
, so we could have concentrated on name
1
for >
1
, name
2
for =
1
,
but actually we concentrated on simple, name
2
for =
1
; see 1.15(B) below.
1.10 Fact. 1) For

P = P
i
: i < g(
P)), a -increasing sequence of forcing notions,

1, 2 and simple

P-named
[j, )-ordinal
and p
_
i<
P
i
there are , q and q
1
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

10 SAHARON SHELAH
such that p q
_
i<g(
P)
P
i
and: either q
P
q
1
G
, q
1
P
, < , [p P

q = q
1
] and q
1

P

= or q
P

is not dened (and even p

P

is not
dened).
2) For

P and 1, 2 as above, and simple

P-named
[j, )-ordinals
, also
max
and min
are simple

Q-named
[j, )-ordinals (naturally dened,

so max
[G] is dened i a
[G],
[G] are dened, and min
[G] is dened
i
[G] is dened or
[G] is dened). If
are full then so are max
and
min
.
3) For

P and as above, n < and simple

P-named
ordinals
1
, . . . ,
n
and
p
_
i<g(
P)
P
i
there are < and q P
such that, rst: p q or at least

q
P
p P
i
/G
for some i < g(
P) and second: for some 1, . . . , n we

have q
P
=
= max
1
, . . . ,
n
or q
P
max
1
, . . . ,
n
not dened.
Similarly for min.
4) The same holds for simple (
Q, W)-names
and we can omit simple.

5) If
i
is a simple

Q-named
1
[
i
,
i
)-ordinal for i < i
then
1
sup
i
: i < i
is
simple

Q-named
1
[Min
i
i
, sup
i
i
)-ordinal.
6) Similarly for
i
: i and when we omit simple.
7A) A simple

P-name
[j, )-ordinal
(a) is a

P-named
[j, )-ordinal;
(b) if j
2
j
1
<
1

2
then any [simple]

P-named [j
1
,
1
)-ordinal is a [simple]
P-named
[j
2
,
2
)-ordinal;
(c) a [simple]

P-named
2
[j, ) ordinal is a [simple]

P-named
1
[j, )-ordinal
(d) if
g(
P) then any [simple]

P-named
[j, )-ordinal is a (
)-
named
[j, )-ordinal.
Proof. Straight.
1.11 Discussion. We have in dening our iteration several possible variants, some
of our particular choices are not important: we can make it like revised countable
1
this seems lacking for name
2
.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

support as in [Sh:f, Ch.X,1] or like
1
-RS in [Sh:f, Ch.XIV,1], or as in [Sh:f,
Ch.XIV,2] (as here); for most uses =
1
and we could restrict ourselves to
pr
=
vpr
as equality; but in [\GoSh:511 ] we need the three partial orders.
So below we have nite support for non-pure, countable for pure and Easton for
very pure.
1.12 Denition. 1) For a forcing notion P (as in 1.1) let P
[pr]
be dened like P
except that we make
P

pr
p for every p P.
2) For a forcing notion P (as in 1.1) let P
[vp]
be dened like P except that we make
P

vp
p (and
pr
P, of course) for every p P.
1.13 Fact. 1) For a forcing notion P and x pr,vp, P
[x]
is also a forcing notion,
and they are equivalent as forcing notions.
2) For x pr,vp
r
the operations P

P and P P
[x]
commute.
3) If (x
1
, x
2
, x
3
) (pr,pr,pr),(vpr,pr,vpr),(vpr,vpr,pr)),(vpr,vpr,vpr) then P
[x
3
]
=
P
[x
1
][x
2
]
.
4) -completeness is preserved in the natural cases.
1.14 Discussion. 1) Why do we bother with P
[pr]
, P
[vpr]
? If in the iteration dened
below we use only Q
[pr]
i
, Q
[vp]
i
, we get a variant of the denition without the need
to repeat it. We may want that: if g(
Q) = inaccessible and i < [P

i
[ <
then
_
i<
P
i
= P
(here as done in [Sh:f, Ch.XIV,2] we can just impose it).

Some other restrictions are for simplicity only.
2) Below the case e = 6 is the main one. [Saharon]
1.15 Denition/Claim. We dene and prove the following by induction on .
Below =
1
, e = so we can omit them (they are meaningful in 11)
(A) [Denition]

Q = P
i
, Q
i
: i < ) is a Sp
e
-iteration for W or
Sp
e
(W)-iteration (if W is absent we mean : strongly inaccessible);
is called the length of

Q, g(
Q).
(B) [Denition] A simple (
Q, W)-named
e
ordinal
and
[, ).
(C) [Denition] A simple (
Q, W)-named
e
atomic condition q
(or atomic
[j, )-condition where j ); also we dene q
, q
, q
[, ) for
a simple

Q-named
e
atomic condition q
and ordinals < , (or

(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

12 SAHARON SHELAH
simple

Q-named
e
ordinals
instead , ). We may add pure/very pure as

adjectives to the condition.
(D) [Claim] Assume
is a simple (
Q, W)-named
e
[j, )-ordinal. Then for any
,
is a simple (
Q, W)-named [j, min, )-ordinal and

Q
if
<
then
; if
or
is undened, then
is undened, also

is a simple (
Q , W)-named ordinal.
Similarly
[
1
,
2
). If
2
are simple (
Q, W)-named [
1
,
2
)-
ordinals (the
2
means
Q

2
), then
1
,
2
) is a simple
(
Q, W)-named [
1
,
2
)-ordinal and
Q
if
1
,
2
), then
2
,
2
)
otherwise
[
1
,
2
) is undened. If in addition = Min,
2
, g(
Q)
and ,
1

1
then
1
,
2
) is a simple (
Q , W)-named [
1
, )-
ordinal. Also if n < , for 1, . . . , n,
is a simple

Q-named [
1
,
2
)-
ordinal then Max
1
, . . . ,
n
is a simple

Q-named [
1
,
2
)-ordinal. Simi-
larly for Min.
(E) [Claim] If q is a simple (
Q, W)-named atomic [j, )-condition, < ,

then q
is a simple (
Q , W)-named atomic [j, min, )-condition and

q
is a P
-name of a member of Q
or undened (and then it may be

assigned the value
Q
, the minimal member of Q
). If q is a simple (
Q, W)-
named atomic condition,
are simple (
Q, W)-named ordinals then

q
) is a simple (
Q, W)-named ordinal. Also q
= q
[, +1), and
if q
is a simple (
Q, W)-named [, )-ordinal,
<
and
Q

),
then it is a simple (
Q, W)-named [
)-ordinal. Also pure and very

pure are preserved by restriction.
(F) [Denition] The Sp
e
(W)-limit of

Q, Sp
e
(W)-Lim
Q), denoted by
P
for

Q as in clause (A) in particular of length , and p and Dom(p)
for p Sp
e
(W)-Lim
Q), an ordinal ; (similarly for a simple (
Q, W)-
named [0, g(
Q)) ordinal
, etc. We also dene P
for
a (
Q, W)-named
ordinal.
(G) [Theorem] Sp
e
(W)-Lim
Q) is a forcing notion (in the sense of 1.1(1)).

(H) [Theorem] Assume < = g(
Q) or more generally,
is a full simple
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

(
Q, W)-named ordinal (see end of clause (F) above). Then P

ic
Sp
e
(W)-
Lim
Q) (so a submodel with the three partial orders, even compatibilities

are preserved) and [p P
p = p] and [P
[= p
q P
[=
p
q ] (for = 0, 1, 2, of course) and P
[= p
p. Also
q P
, p Sp
e
(W)-Lim
Q) are compatible i q, p are compatible in

P
. In fact, if q P
, P
[= p q then q (p [, )) belongs to P
and is a least upper bound of p, q and if P
[= p
q even a
-least
upper bound of q. Hence P
(Sp
e
(W)-Lim
( q)).
(I) [Claim] The set of p P
such that for every < we have

P
p
is a singleton or empty, is a dense subset of P
. Also we can replace Q
by
and the set of old p P
is a dense subset of the new (but actually

do not use this).
(J) [Claim] If is strongly inaccessible, < [P
[ < or just <

CC(P
) and W, then P
=
_
<
P
.
Proof and Denition:
(A)

Q = P
i
, Q
i
: i < ) is a Sp
e
(W)-iteration if

Q is a Sp
e
(W)-
iteration for every < , and if = +1, then P
= Sp
e
(W)-Lim
Q )
and Q
is a P
-name of a forcing notion as in 1.1(1) here.

(B) We say
is a simple (
Q, W)-named
e
[j, )-ordinal if
is a simple (
Q, W)-
named
2
[j, )-ordinal.
(C) We say q
is a simple (
Q, W)-named atomic [j, )-condition when: q
is a

Q-
name, and for some
, a simple (
Q, W)-named [j, )-ordinal, we have
has a value i q
has, and if they have then j
< min, g(
Q)
and q
. If we omit [j, ) we mean [0, ). Now q
will have
a value i
has a value < and then its value is the value of q
. Lastly,
q
will have a value i
has value and then its value is the value of

q
. Similarly for q
[, ) and q
, q
, q
). We say q
is pure if
Q
for < , if
q
= and q
then Q
[=
Q

pr
q
. We say q
is very
pure if
Q
for < , if q
, then Q
[=
Q
i

vpr
q
.
(D), (E) Left to the reader.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

14 SAHARON SHELAH
(F) We are dening Sp
e
(W)-Lim
Q) (where

Q = P
, Q
: < ), of course).
It is a quadruple P
= (P
, ,
pr
,
vpr
) where
(a) P
is the set of p = q
i
: i < i
satisfying for some witness

:
(i) each q
i
is a simple (
Q, W)-named atomic condition,

and for every < , we have
p =: q
i
: i < i
0
Q
(ii) if W is strongly inaccessible > CC(P

i
) +
for every i < , then i
<
(iii)

: < j) where j < and

each
is a simple (
Q, W)-named [0, )-ordinal,

[the reader should think of
: < j
as the non-very-pure support of p]
(iv) for every < we have (we may replace
Q
by
P
as we use simple names)
Q
if ( < j)(
[G
] is ,= ) (for example is
not well dened) then
Q

vpr
p in

Q
(v) if < then p =: q
i
: i < i
belongs to P
(vi) if W is strongly inaccessible > CC(P

i
) +
for every i < then for some <
every
i
is a simple (
Q, W)-named [0, )-ordinal;

needed, e.g., in 6.12 (note: this demand follows by 1.8)
(b) for p Sp
e
(W)-Lim
Q) and < g(
Q) we let:
p =: r : r p
p =: r : r p
we dene similarly p [, ), p
, p [
).
(c) P
[= p
1
vpr
p
2
i for every < we have (letting
p
= q
i
: i < i
() for = 1, 2):
p
2

P
[= p
1

vpr
p
2

(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

(d) P
[= p
1
pr
p
2
i
(i) for every < g(
Q), we have
2
p
2

P
then

Q
[= p
1

pr
p
2

(ii) for some ordinal j < and simple (
Q, W)-named
[0, )-ordinals
for < j, for every < g(
Q) we have:
p
2

P
if for no < j do we have
[G
] = , then
p
1

vpr
p
2
in

Q
; we call
: < j) a witness.
(e) P
[= p
1
p
2
i
(i) for every < g(
Q) we have:
p
2
[= p
1
p
2

(ii) as in the denition of
pr
; clause (ii)
(iii) for some n < and simple (
Q, W)-named ordinals
1
, . . . ,
n
we have:
for each < g(
Q) we have
p
2

P
if ,=
[G
] for
= 1, . . . , n and
(
Q

vp
p
1
)]
then:

Q
[= p
1

pr
p
2

note that the truth value of =
is a P
-name so this is
well dened.
We then (i.e. if (i) + (ii) + (iii)) say: p
1
p
2
over
1
, . . . ,
n
.
(f) Lastly, for p P
we let Dom(p) = Dom

vp
(p) =
: q
p and
Dom
pr
(p) =
: < j where

=
: < j) is as in clause (F)(a)

above (we can make it part of p).
2
recall 1.1(4)(d) we can omit p
1
{xi} =
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

16 SAHARON SHELAH
(g) We still have to dene P
for
a full simple (
Q, W)-named ordinal,
it is p : p = q
i
: i < i
and
Q

q
i
<
that is if < then

P
if
i
[G
P
] = and
[G
P
] is well dened then it is > .

(h) We call p P
full if it has a witness
: < j) with each
full.
(G) Let us check Denition 1.1(1) for P
=: Sp
e
(W)-Lim
Q):
Proof of
P
is a partial order.
Suppose p
0
p
1
p
2
. Let n
1
, . . . ,
and j
(< )
(for < j
)
appear in the denition of p
p
+1
. Let n = n
0
+n
1
, and
i
=
_
_
_
0
i
if 1 i n
1
1
in
if n
1
< i n
1
+n
2
.
Let j = j
0
+j
1
and
=
_
_
_
if < j
0
1
j
0
if [j
0
, j
0
+j
1
).
Let us check the three clauses of (e) of part (D).
Clause (i):
Let < g(
Q) so for = 0, 1
p
+1

P
p
+1
in

Q
.
As P
[= p
1
p
2
(by the induction hypothesis, clause (H))
clearly p
2
forces both assertions. As

Q
is a partial order (under ) the

conclusion follows.
Clause (ii):
Let < g(
Q), so similarly p
2

P
if ,=
m
for m = 1, . . . , n
and
= 0, 1 (i.e.,
m
[G
] is ,= or is not well dened), then p

0

pr
p
1

in

Q
and p
1

pr
p
2
in

Q
from which the result follows.

(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

Clause (iii):
Lastly, for < we have p
2

P
if /
[G
P
] :
[G
P
] well dened,
< j
and 0, 1 then p
0

vpr
p
1

vpr
p
2
hence
p
0

vpr
p
2
.
So we have proved the three conditions needed for p
0
p
2
by the denition
above so really p
0
p
2
holds, so is a partial order.
Proof of
pr
is a partial order.
Similar proof.
Proof of
vpr
is a partial order.
Similar proof just easier.
Proof of p
pr
q p q:
By the denition; easy.
Proof of p
vpr
q p
pr
q:
By the denition, check.
So in 1.1(1) all the requirements on P
holds.
(H), (I), (J) We leave the checking to the reader (actually we prove (I) in the proof of
1.20 below).
1.15
1.16 Fact. 1) If

Q is a Sp
e
-iteration and for each i < g(
Q) we have it is forced
(i.e.,
P
i
) that
Q
i
pr
=
Q
i
and
Q
i
vpr
is equality, then

Q is a variant of -RS iteration
(as in [Sh:f, Ch.XIV,1]), i.e. they are the same if we use there only simple

Q-named
ordinals (or allow here non-simple ones so the version here is exactly as in [Sh:f,
Ch.XIV,2.6]).
Proof. Straightforward.
1.17 Claim. 1) In 1.15 in Denition (F), clause (a)(iii) we can demand that each
is full (simple (
Q, W)-named [0, )-ordinal) and similarly in (d)(ii), (=(e)(ii))

and (e)(iii).
2) Suppose

Q = P
i
, Q
i
: i < ) is a Sp
e
(W)-iteration (so P
= Sp
e
(W)-
Lim
Q)). If p q in P
, then there are r, n < and

1
< . . . <
n
< such
that:
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

18 SAHARON SHELAH
(a) r P
(b) q r
(c) p r above
1
, . . . ,
n
.
3) If p
1
, p
2
P
and 1.15(F)(e) holds but we allow n
to be a full simple (
Q, W)-
named ordinal, then p
2
p
1
G
.
Remark. In fact, in 1.17(1) we can have r [
n
, ) = q [
n
, ).
Proof. 1) As increasing those sets (
: < ,
1
, . . . ,
n
, respectively) cause no
harm.
2) We prove this by induction on .
Case 1: = 0.
Trivial.
Case 2: = + 1.
Apply the induction hypothesis to

Q , p , q (clearly

Q is an
Sp
e
(W)-iteration, p P
, q P
and P
[= p q, by 1.15, clause
(H)). So we can nd r
, m < and
1
, . . . ,
m
such that:
(a)
(b)
[= q r
(c)
p r
(in P
) above
1
, . . . ,
m
(d)
1
< . . . <
m
.
Let n =: m+ 1 and
=
_
if 1, . . . , m
if = n
and lastly r = r
(q ).
Case 3: is a limit ordinal.
Let p q (in P
) above
1
, . . . ,
n
. We choose by induction on n, the objects
r
such that:
() r
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

() r
r
+1
() q
()

+1
<
()
0
= 0, r
0
=
P
0
() for 1, . . . , n we have: either r
and
or
=
1
& r
= r
1
and r
(q [, ))
P
is not dened or is
.
Carrying the denition is straight: for i = 0 use clause (). For +1 n when the
second possibility of clause () fails there is r
, such that r
(q [, )) r
,
and r
+1
is dened and is < , so there are r
+1
< such that r
and r
+1
=
+1
so as
+1
is a simple

Q-named ordinal we know
that
+1
< and r
+1

P
+1
+1
=
+1
. Let
+1
=: max
+1
, and
r
+1
=: r

+1
. So we have carried the induction.
We now apply the induction hypothesis to

Q
n
, p
n
, r
n
; it is applicable as
n
< , and P
n
[= p
n
q
n
r
n
. So there are m < ,
1
< . . . <
m
<
n
and r
such that P
n
[= r
n
r
and p r
(in P
n
) above
1
, . . . ,
m
. Now
let r =: r
(q [
n
, )), clearly q r and p r above
1
, . . . ,
m
,
n
.
3) So p
1
, p
2
,
: = 1, . . . , n are given. Assume that p

2
q
1
P
. We can nd
q
2
such that q
1

2
P
and q
2
forces a value to n
say n(). Next we can nd q

3
such that q
2
q
3
P
and q
3
forces values to
( = 1, . . . , n(), say
0
1
, . . . ,
0
n()
.
Now repeating the proof of
P
is a partial order (in clause (H) of 1.15) with

p
1
, p
2
, q
3
here standing for p
0
, p
1
, p
2
there we let P
[= p
1
q
3
. As we have assumed
just p
2
q
1
P
and q
1
q
3
we are done.
1.17
1.18 Claim. Let

Q be a Sp
e
(W)-iteration of length .
0) If
is a simple (
Q, W)-named ordinal then for some ordinal we have:
is a
simple (
Q, W)-named [0, )-ordinal.

1)
(i) If < and
is a P
-name of a [full] simple (
Q, W)-named [, )-ordinal
then for some [full] simple (
Q, W)-named [, )-ordinal
we have
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

20 SAHARON SHELAH
(ii) if ,
1

2
and
is a P
-name of a [full] simple (
Q, W)-named
[
1
,
2
)-ordinal then for some [full] simple (
Q, W)-named [
1
,
2
)-ordinal
we have
Q

.
2) The same holds if we replace ordinal by atomic condition (so in (ii) we
should demand
2
).
3) If and
is a full simple (
Q, W)-named [0, )-ordinal, and for each

< ,
is a [full] simple (
Q, W)-named [, )-ordinal then for some [full] simple

(
Q, W)-named [0, )-ordinal
we have
Q
if
[G
] = (so < ) then
[G
] =
[G
].
4) If
is a full simple (
Q, W)-named [0, )-ordinal and for each < , p
is a
(
Q, W)-named [, )-atomic condition then for some (
Q, W)-named atomic condi-

tion p we have
Q
if
[G
] = then p
[G
] = p
[G
].
Proof. Easy.
1.18
> scite1.16 ambiguous
1.19 Claim. 1) Suppose F is a function, then for every ordinal there is
Sp
e
(W)-iteration

Q = P
i
, Q
i
: i <
), such that:
(a) for every i we have Q
i
= F(
Q i),
(b)

(c) either
= or F(
Q) is not an Sp
e
(W) Lim
Q)-name of a forcing
notion.
2) Suppose < , G
is generic over V, then in V[G
],

Q/G
= P
i
/G
, Q
i
:
i < ) is an Sp
e
-iteration and Sp
e
(W) Lim(
Q) = P
(Lim(
Q)/G

)
(essentially; more exactly up to equivalence) where, of course, P
i
/G
= p P
i
:
p G
.
3) If

Q is an Sp
e
(W)-iteration, p Sp
e
(W)Lim
Q), P
i
= q P
i
: q p i,
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

Q
i
= p Q
i
: p p i and
Q
i
= p i,
Q
=
Q
i
for = 0, 1, 2 then

Q =
P
i
, Q
i
: i < g(
Q)) is (essentially) a Sp
e
(W)-iteration (and Sp
e
(W)Lim
)
is P
g(
Q)
).
Proof. Should be clear.
1.20 Claim. Suppose
(a)

Q = P
i
, Q
i
: i < ) is a Sp
e
(W)-iteration (and P
= Sp
e
(W)
Lim
Q))
(b) () 0, 1
(c)
P
i
(Q
i
,
()
) is a -complete for each i < .
Then:
1) (P
,
()
) is -complete, i.e. if < , p
i
: i < ) is
()
-increasing then it has
an
()
-upper bound provided that:
or () = 0 & Min : W is strongly inaccessible and ( <
)([P
[ < ).
2) Moreover for < we have (P
/P
,
()
) is -complete.
3) In fact, we can get
()
-lub (provided that there are such lubs for each Q
i
.
Remark. We deal with -complete rather than strategically -complete (here and
later) just for simplicity presentation, as it does not matter much by [Sh:f, CH.XIV,2.4].
Proof. Straightforward but we elaborate.
1) So assume < and p
i
P
for i < and [i < j < p

i

()
p
j
]. Now it is
enough to nd p P
such that
i < p
i

()
p.
Let p
i
= q
: <
i
and for each <
i
, q
is a simple (
Q, W)-named atomic
condition, say
Q
q
, where
is a simple (
Q, W)-named ordinal (which

is
). Now for each < let <
be a P
-name of a well ordering of Q
. For each
i() < , () <
i
let r
i()
()
be the following simple (
Q, W)-named atomic condition:

(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

22 SAHARON SHELAH
Let < , G
generic over V and
i()
()
[G
] = , now work in V[G
],
let w
= i < : for some <

i
we have
[G
] = . We let u
i
= <
i
:
[G
] = for each i w
. (As p
i
is
()
-increasing, w
is an end segment of
and i() w
, () u
i()
). For i w
let q
i,
= (p
i
)[G
]. Now dene
r
i()
()
[G
] as follows.
Case 1: For some j < the sequence q
i,
: i w
j) is constant.
Let r
i()
()
[G
] = q
Min(w
\j),i
.
Case 2: Not Case 1, but for some < 2, j < the sequence q
i,
: i w
j)
is
-increasing, without loss of generality minimal (on all possible j) and then
r
i()
()
[G
] Q
[G
] is the <
-rst
-upper bound of q
i,
: i w
j where
<
= <
[G
]. It exists by 1.2(2).
Case 3: Neither Case 1 nor Case 2 r
i()
()
[G
] is
Q
.
Let p = r
i()
()
: i() < and <
i()
.
If () = 0 (i.e., p
i
is
vpr
-increasing) and
j
: j < j
) witness p
0
P
, then it
witnesses p P
and easily i < p

i

0
p.
So assume () = 1, that is p
i
is
pr
-increasing. For i
0
< i
2
< let
i
0
,i
1
j
: j <
j
0
,
1
be a witness to p
i
0

pr
p
i
1
and let
i
j
: j < j
i
witness p
i
P. Letting
be
the maximal cardinal < , clearly
i
0
,i
1
j
: i
0
< i
1
< and j < j
i
0
,i
1
has cardinality
, so we can order it as
i
0
(),i
1
()
j()
: <
and some
i()
j
()
: <
list
i
j
: i < , j < j
i
. Now
i()
j
()
: <
witness p P
and
i
0
(),i
1
()
j()
: <
witness p
i

1
p
for every i < .

2), 3) Similar proof.
1.20
1.21 Denition. Let

Q = P
i
, Q
i
: i < ) be an Sp
e
(W)-iteration.
1) We say y
is a (
Q, W,
)-name if: y
is a P
-name,
is a simple

Q-named [0, )-
ordinal, and: if < , G
P
is generic over V and for some r G

P
we have r
Q

= , then y
[G
P
] V[G
P
] is well dened and depends only on

G
P
so we write y
[G
P
]; and if G
P
is generic over V and
[G
P
]
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

not well dened then y
[G
P
] is not well dened (do not arise if
is full).
2) If p P
and G
P
is generic over V, (or just in Gen

r
(
Q)), then p[G

P
]
is the following function, Dom(p[G
]) =
[G
P
] : q
p and (p[G
P
])() =
q
[G
P
] : q
p and
[G
P
] = .
1.22 Claim. Suppose
(a)

Q = P
i
, Q
i
: i < ) is a Sp
e
(W)-iteration.
(b) p P
and
is a simple (

(c) r
is a (
Q, W,
)-named member of P
/P
.
Then:
1) There is q P
satisfying p q such that:

() if < , G
generic over V, then

()
[G
] =
implies (p )[G
] = (q )[G
] and
()
[G
] = implies
(q [, ))[G
] = r
[G
].
2) If in addition (for any < 3) clause (c)
+
below, then we can in () add p
q
(c)
+
r
is a (
Q, W, r
)-named member of P
/P
which is
-above p [
, ).
Central here is pure decidability.
1.23 Denition. 1) A forcing notion Q has pure (
1
,
2
)-decidability if: for every
p Q and Q-name
<
1
, there are a
1
, [a[ <
2
(but [a[ > 0) and r Q such
that p
pr
r and r
Q

a (for
1
= 2, alternatively,
is a truth value). If we
write
2
we mean [a[ [
2
[.
2) A forcing notion Q has pure -decidability where is an ordinal if: for every
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

24 SAHARON SHELAH
p Q and Q
-name
< there are < and r Q such that p

pr
r and r
Q
if
< then
= and if is a limit ordinal then
< .
1.24 Observation. 1) If
0
>
2
> 2 then pure (
2
, 2)-decidability is equivalent to
pure (2, 2)-decidability.
2) If Q is purely semi-proper (see [Sh:f, X] or here xxx) or just Q satises UP
0
(I, W)
(see 5) then Q has pure (
1
,
1
)-decidability.
3) If Q is purely proper, then Q has (,
1
)-decidability for every .
4) If Q has the c.c.c. (and we let
pr
be equality if not dened), then Q is purely
proper.
5) If
pr
=, then Q has pure (, 2)-decidability for every .
Proof. Think of the denitions.
1.25 Denition. 1) A forcing notion P is purely proper for large enough (e.g.,
P(P) H () is enough) and N is an elementary submodel of (H(), ) to which
P belongs and p N P then there is (N, P)-semi-generic satisfying p
pr
q P,
see below.
2) q is (N, P)-generic if q
P
if
which belongs to N is a P-name of an ordinal

then
[G
P
] N Ord.
3) A forcing notion P is purely semi-proper if in part 4) we replace (N, P)-generic
by (N, P)-semi-generic.
4) q is (N, P)-semi-generic if q
P
if
which belongs to N, is a P-name of a

countable ordinal then
[G
P
] N
1
.
1.26 Claim. Let

Q be a Sp
e
(W)-iteration.
1) The property Q has pure
-decidability and pure (2, 2)-decidability is pre-

served by
1
Sp
e
(W)-iterations if
is a limit ordinal.
2) The property Q has pure (2, 2)-decidablity is preserved by
1
Sp
e
(W)-iterations.
1.27 Remark. 1) This is like [Sh:f, Ch.XIV,2.13] and is reasonable for iterations not
adding reals. For getting rid of pure (2, 2)-decidability at the expense of others,
natural demands, see 5.
2) Is this not suitable for name
1
ordinals only? By UP help.
3) See proof of 6.8 for use of
_
n
q
n
p
and more cases phrase as a subclaim?

(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

Proof. Let = g(
Q) and let
be a well ordering of H (), let
,P
be a P
-
name of a well ordering of H ()
V
P
. Let p P
and
be a P
-name of an ordinal
< , 2,
and let

0
=
: < j) be a witness for p (see 1.15(F), clause

(a)) without loss of generality each
is full. For each < j and < below we

shall r
0
,
,
0
,
and t
,
such that r
0
,
is a P
+1
-name of a condition with domain
( + 1, ),
0
,
is a P
+1
-name of an ordinal < and t
,
is a P
+1
-name of a
truth value, satisfying the following. Let < , G
P
+1
P
+1
be generic over V
and
[G
+1
] = :
()
1
if there are r P
/G
P
+1
and < satisfying (a) + (b) below then =
0
,
[G
P
+1
], r = r
0
,
[G
P
+1
] are such objects, rst by the xed well ordering
<
and t
,
[G
P
+1
] is truth, and does not depend on ; if there are no
such , r we let r
0
,
[G
P
+1
] be the empty condition,
0
,
[G
P
+1
] = 0 and
t
,
[G
P
+1
] is false.
(a) p
pr
r P
/G
P
+1
, p ( + 1) = r ( + 1) and we have r
P
/G
P
+1
if = 2 then
= and if
0
then
<
(b) if
1
< then no r
satises (a) with

1
, G
P
+1
P
1
+1
.
So by 1.18(4), there is in P
a condition r
which is r
0
,
if
[G] = , t
,
[G] =
truth, is well dened (the
there is
here!, hence it is full). So easily p

1
= pr
:
< j belongs to P
and is a pure extension of p (using <
, noting that for each

+ 1 and G
P
+1
P
+1
generic over V, if
0
+1
[G
P
+1
] = =
1
[G
P
+1
] then
r
1
[G
P
+1
] = r
2
[G
P
+1
]).
We now dene p
2
= p
1
r
: < j where r
is an atomic

Q-named condition
with
r
dened as follows
() if < , G
P
generic over V and
[G
P
] = then in V[G
P
] we have
r Q
[G] is
,P
[G
P
]-minimal such that

(i)

Q
[G
P
] [= p
2

pr
r

Q
[G
P
]
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

26 SAHARON SHELAH
(ii) for some r
1
P
+1
and < we have: r
1
G
P
and r
1
= r
and: r
1

P
+1
= 2 &
0
,
= or
0
&
0
,
< and t
,
=
truth or r
1
forces (
P
+1
) that t
,
= false.
Let us choose now and r
1
with minimal such that
r
1
P
and there are q P
and < such that p

2

pr
q and q r
1
and r
1
(q [, )) = 2,
= or ,= 2,
< .
There is such as = (= g(
)) is O.K.
Case 1: = 0.
We are done.
Case 2: is limit.
Without loss of generality, by 1.17 for some n < and
1
< . . . <
n
<
we have: p
2
r
1
above
1
, . . . ,
n
. If n = 0 we are done (as = 0)
so assume n > 0. Let
=
n
+ 1, r
= r
1
(
n
+ 1) and there is q
dened by
q
(
n
+1) = q (
n
+1), q
(
n
+1, ) is r
1
(
n
+1, ) if r
1
(
n
+1) G
n
+1
and
is r
1
(
n
+1, ) otherwise and lastly q
[, ) = q [, ). Now
, r
, q
satises:
r
n
+1
= P
, p
2

pr
q
, q
and r
q [
, )
P
= 2,
= or
,= 2,
< and
< . So we get a contradiction to the choice of .

Case 3: r
1

0
/
: < j where =
0
+ 1.
The proof is similar to the one of case 2 using
=
0
.
Case 4: None of the above.
So by neither case 1 nor case 2 we have =
0
+1, and as we can increase r
1
without loss of generality r
1
forces
0

: < j, so without loss of generality r

1

0
=
where < j.
Let r
1
G
with G
generic over V; let G
= G
for
. So
[G
0
] =
0
.
We rst ask: is there
1
< j such that
1
[G
0
] is well dened so call it , (so
necessarily
0
) and t
1
,
[G
] is truth?
If yes, then we get a contradiction to the minimality of as can serve by the
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

choice of
2
, so assume not. Now considering r
0
,
,
0
,
, clause (b) holds and r
1
, q
exemplies t
,
[G
p
] = truth.
1.26
1.28 Remark. 1) You may ask why we do not use the
dened by
[G
+1
] = +1
if t
,
[G
+1
] = truth for some < j? The reason is that (as for e = 6, =
1
) this
seems not to be a simple

Q-named ordinal.
2) By the proof, if

Q is a -Sp
e
(W)-iteration, g(
Q), p P
and for each

< , t
is a P
-name of a truth value, p
a P
-name of some p P
/G
such
that
P
p
pr
p
, p = p
then we can nd q such that p

pr
q P
and:
if G
is generic over V, < , t
[G
] = truth and < t
[G
] =
false then q
P
/G
P
.
We now consider some variants of the -c.c.
1.29 Denition. 1) We say P satises the local
-c.c. if
is a P-name and
p P : P q : p q P satises the
-c.c. and p
P

for some
is
dense in P.
2) We say P satises the local
-c.c. purely if the set above is dense in (P,

pr
).
3) We say P satises lc.pr.
-c.c. if:
(a)
is a (P,
pr
)-name, usually of a regular cardinal of V
(could use just a partial function from P to cardinals such that
(p) =
p
pr
q
(q) = , but abusing notation we write q
= if
(q) = )
(b) for every p P for some q, we have p
pr
q, q
(P,
pr
)

= and P
q
satises the -c.c.
(we could use: if
(p) = then P
p
i.e. (q : p q P,
P
) satises the
-c.c.)
4) If P satises the lc.pr.
-c.c. and q P let

mcc
(q, P) = means q
(P,
pr
)
= and P
q
satises the -c.c.
5) Let
mcc
(P) be minimal such that P satises the lc.pr.
-c.c.; that is
(q, P) =
Min : P
q
satises that -c.c. and
(q, P) = if (r
)(q
pr
r
(r, P) = )
(see below) and let
mcc
(q, P) = mean
mcc
(P)
(q, P) = where
mcc
(P).
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

28 SAHARON SHELAH
1.30 Claim. 1) For a forcing notion P (as in 1.1) the (P,
pr
)-name
mcc
(P) is
well dened, so
2) If P satises the lc.pr.
-c.c. and p P then for some q we have p

pr
q and
mcc
(q, P) is well dened.
Proof. Straight.
1.31 Denition. 1) We say Q has strong pr. (
1
,
2
)-decidability when
2
are
(Q,
pr
)-names of regular cardinals of V and if p Q, p
(Q,
pr
)

1
=
1
and
2
=
2
and
is a Q-name of an ordinal <

1
for <
<
2
then for some
a
1
of cardinality <
2
and q such that p
pr
q Q we have q
Q

a for
<
.
2) We say Q has strong pr.
-decidability if for any it has pr. (,
)-decidability
(i.e. each
is a Q-name of an ordinal < ).

3) We use weak instead of strong in parts (1), (2) if above we restrict ourselves
to the case
= 1.
4) We let
(P) is the minimal (P,

pr
)-name
of a regular cardinal from V such

that P has weak pr.
-decidability. Similarly
St
(P) for strong pr.
-decidability.
1.32 Note/Observation. If

Q is an Sp
e
(W)-iteration, then
(a) (P
i
,
P
i
pr
) : i g(
Q)) is a -increasing sequence

(b) if i < j g(
Q), q P
j
, q i
pr
p P
i
then p, q has a
pr
-lub, pq [i, j).
Proof. Check.
1.33 Claim. 1) If P satises the lc.pr.
-c.c. and =
mcc
(p, P)
P
is
regular then P has a strong pr
-decidability.
2) Let

Q be a
1
Sp
e
(W)-iteration e 4. If g(
Q) is a limit ordinal, u an
unbounded subset of , for i u we have P
i
has the strong pr.
i
-decidability then
letting
= Min : a regular cardinal in V and
i
for i u we have
(i)
is a (P
,
pr
)-name of a regular cardinal of V
(ii) P
has weak pr.
-decidability.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

3) Similarly for P
/G
P
when < and even P
/P
where
is a simple

Q-named
[0, )-ordinal.
4) If P satises the strong pr
-decidability and p P, =
(p, P) then p is
a regular cardinal.
Proof. 1) Trivial.
2),3) Similar to the proof of 1.26 [Saharon!]
4) Trivial.
1.33
1.34 Convention: Let

Q = P
j
, Q
i
: j , i < ) be a -Sp
e
(W)-iteration.
1.35 Claim. Assume (
Q, W) is smooth, (see Denition 1.7(8)).

1) If p
P
j
then P
j
, Q
i
: j , i < ) is a
1
-Sp
e
(W)-iteration where
P
j
= p P
j
: p
j p,
Q
j
= p Q
j
:
P
j
p
j p in

Q
j
.
2) If < and G
j
P
is generic over V then P

+j
/G
, Q
+i
: j and
i < ) is a
1
Sp
e
(W)-iteration.
Proof. Straight.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

30 SAHARON SHELAH
2 Tree of Models
We present here needed information on trees and tagged trees. On partition of
tagged trees, see Rubin, Shelah [RuSh 117] and, [Sh:b], [Sh:f, XI,3.5,3.5A,3.7,XV,2.6,2.6A,2.6B]
and [Sh 136, 2.4,2.5,p.111-113]; or the representation in [Sh:e, AP,1]; on history
see [RuSh 117] and [Sh:f].
2.1 Denition. A tagged tree is a pair (T, I) such that:
(1) T is a -tree, which here means a nonempty set of nite sequences of ordinals
such that if T then any initial segment of belongs to T. T is ordered
by initial segments; i.e., i is a proper initial segment of .
(2) I is a partial function such that for every T Dom(I) : I() is an ideal
of subsets of some set called the domain of I
, Dom(I
), and
3
Suc
T
() =: : is an immediate successor of in T Dom(I
),
Usually I
is
2
-complete.
(3) For every T we have Suc
T
() ,= .
2.2 Convention. For any tagged tree (T, I) we can dene I
by:
Dom(I
) = : T Dom(I) and Suc

T
() dom(I
) and Suc
T
() / I
and
I
=
_
: ) A : A I
_
;
we sometimes, in an abuse of notation, do not distinguish between I and I
; e.g. if
I
is constantly I
, we write I
instead of I. Also, if I = I
x
, we may write I
x
for
I
x
().
3
in this section it is not unreasonable to demand equality but this is very problematic in
Denition 4.2(1), clause (d)
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

2.3 Denition. 1) Let be called a splitting point of (T, I) if I
is well dened and

Suc
T
() / I
(normally this follows but we may forget to decrease the domain of

I). Let split(T, I) be the set of splitting points of (T, I). We usually consider trees
where each -branch meets split(T, I) innitely often (see Denition 2.5(6), i.e. are
normal).
2) For T let T
[]
=: T : = or or .
3) For a tree T, let lim(T) be the set of branches of T; i.e. all -sequences of
ordinals, such that every nite initial segment of them is a member of T : lim(T) =
s

Ord : (n)s n T. We call them also -branches.
4) A subset Z of a tree T is a front if: ,= Z implies none of them is an
initial segment of the other, and every lim(T) has an initial segment which is
a member of Z.
2.4 Denition. We now dene orders between tagged trees:
(a) (T
1
, I
1
) (T
2
, I
2
) if T
2
T
1
, and split(T
2
, I
2
) split(T
1
, I
1
), and
for every split(T
2
, I
2
) we have I
2
() Suc
T
2
() = I
1
() Suc
T
2
()
(where I A = B : B A and B I).
(So every splitting point of T
2
is a splitting point of T
1
, and if we suppose
T Suc
T
(h) = Dom(I
) then I
2
is completely determined by I
1
and
split(T
2
, I
2
) and T
2
.)
(b) (T
1
, I
1
)
(T
2
, I
2
) if (T
1
, I
1
) (T
2
, I
2
) and
split(T
2
, I
2
) = split(T
1
, I
1
) T
2
.
(c) For any set A, (T
1
, I
1
)
A
(T
2
, I
2
) if T
1
T
2
and AT
2
Suc
T
2
() =
Suc
T
1
() and T
2
split(T
1
, I
1
)A Suc
T
2
() ,= mod I
1
.
(d) In (c) we may omit the subscript A when A = T
2
split(T
1
, I
1
).
(e) (T
1
, I
1
)
(T
2
, I
2
) if T
2
T
1
is a subtree and if lim(T
2
), I
is
-complete, then for some k g(), I
1

RK
I
2
k
, I
2
k
is -complete and
k split(T
2
, I
2
). We can replace and -complete by and satisfying
, e.g. I.
(f) (T
1
, I
1
)
(T
2
, I
2
) when T
2
T
1
, and if T
2
split(T
1
, I
1
) and I
1
is
-complete then split(T
2
, I
2
) and I
2
= I
1
and every split(T

2
, I
2
)
is like that.
2.5 Denition. 1) For a set I of ideals, a tagged tree (T, I) is an I-tree if for every
splitting point T we have I
I (up to an isomorphism).
2) For a tagged tree (T, I) and set I of ideals let I I = I Dom(I) and
I
I.
3) Let in (2), (T, I) I = (T, I I).
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

32 SAHARON SHELAH
4) A tagged tree (T, I) is standard if for every non-splitting point T, [Suc
T
()[ =
1.
5) A tagged tree (T, I) is full if every T is a splitting point.
6) A tagged tree (T, I) is normal if for every lim(T) for innitely many k <
we have k split(T, I).
2.6 Remark. 1) Of course, the set lim(T) is not absolute; i.e., if V
1
V
2
are
two universes of set theory then in general (lim(T))
V
1
will be a proper subset of
(lim(T))
V
2
.
2) However, the notion of being a front is absolute:
(a) Z a point i Z contains a front and ,= Z ( )
(b) V
1
[= Z contains a front of T, i Z T and T : k g()(
k Z has no -branch i Z T there is a depth function : T Ord
satisfying & k g()[ k / Z] () > (). (Levy absoluteness
theorem) This function will also witness in V
2
that Z is a front.
3) Z T contains a front i Z meets every branch of T. So if Z T contains
a front of T and T
T is a subtree, then Z T
contains a front of T
. This is
absolute, too.
2.7 Denition. 1) An ideal I is -complete if any union of less than members
of I is still a member of I.
2) A tagged tree (T, I) is -complete if for each T Dom(I) or just
split(T, I), the ideal I
is -complete.
3) A family I of ideals is -complete if each I I is -complete. We will only
consider
2
-complete families I.
4) A family I is called restriction-closed if: I I, A Dom(I), A / I implies
I A = B I : B A belongs to I.
5) The restriction closure of I, res-c(I) is I A : I I, A Dom(I), A / I.
6) I is -indecomposable if for every A Dom(I), A / I and h : A there is
Y , [Y [ < such that h
1
(Y ) / I. We say I (or we say I) is -indecomposable
if each I
where split(T, I) (or I I) is -indecomposable.

7) I is strongly -indecomposable if for any A
i
I (i < ) and A Dom(I), A / I
we can nd B A of cardinality < such that for no i < does A
i
include B.
8) Let I
[]
= I I : I is -complete.
2.8 Fact. If is a regular cardinal and I is a strongly -indecomposable, then I is
-indecomposable.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

Proof. Given A, h as in 2.7(6), let A
i
= h
1
(j : j < i) and A
i
= h
1
(i);
if for some i < , A
i
/ I we are done, otherwise by Denition 2.7(7) there is
B A, [B[ < such that: i < B A
i
. But as is regular, B A, [B[ <
and A
i
: i < ) is a -increasing sequence of sets with union A, clearly for some
j < , B A
j
, contradiction.
2.8
2.9 Denition. For a subset A of (an -tree) T we dene by induction on the
length of a sequence , res
T
(, A) for each T. Let res
T
(), A) = ). Assume
res
T
(, A) is already dened and we dene res
T
(), A) for all members )
of Suc
T
(). If A then res
T
(), A) = res
T
(, A)), and if / A then
res
T
(), A) = res
T
(, A)0). If lim(T), we let
res(, A) =
_
k
res( k, A).
2.10 Explanation. Thus res(T, A) =: res
T
(, A) : T is a tree obtained by
projecting T; i.e., glueing together all members of Suc
T
() whenever / A.
We state now (Lemma 2.11 is [Sh:f, Ch.XI,5.3;p.559] and Lemma 2.12 is
[Sh:f, XV,2.6;p.738] and Lemma 2.13 is [Sh:f, XI,3.5](2); p.546-7.
2.11 Lemma. Let , be uncountable cardinals satisfying
<
= and let (T, I)
be a tagged tree in which for each T either [Suc
T
()[ < or I() is
+
-complete.
Then for every function H : T there exists a subtree T
of T satisfying (T, I)
(T
, I) such that for

1
,
2
T
satisfying res
T
(
1
, A) = res
T
(
2
, A) we have:
(i) H(
1
) = H(
2
)
(ii)
1
A
2
A,
(iii) if T
A, then Suc
T
() = Suc
T
().
Proof. See [Sh:f, XV,2.6;p.738].
2.12 Lemma. Let be an uncountable regular cardinal (the main case here is
=
1
). Assume
() I be a family of
+
-complete ideals,
() (T
0
, I) a tagged tree,
() A =: T : [Suc
T
0
()[ ,
() [ T
0
A I
I & Suc
T
0
() / I
]
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

34 SAHARON SHELAH
( [ A Suc
T
0
() i) : i < ]
() H : T
0
and
() e = e
: A, [Suc
T
0
()[ = ) is such that e
is a club of .
Then there is a club C of such that: for each C there is T
T
0
satisfying:
(a) T
is a tree
(b) if T
, [Suc
T
0
()[ < , then Suc
T
() = Suc
T
0
() and if Suc() = , then
Suc
T
() = i) : i < and e
(c) T
A implies Suc
T
() / I
(d) for every T
we have H() < .

Proof. See [Sh:f, XV,2.6].
2.13 Lemma. Let (T, I) be a
+
-complete I-tree, and assume = cf(). If
lim(T) =
_
i<
B
i
where B
i
is a Borel subset of lim(T), then for some i < and
(T
, I
) we have (T, I)
(T
, I
) and lim(T
) B
i
.
Proof. By [Sh:f, XI,3.5](2);p.546-7.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

3 Ideals and Partial Orders
3.1 Denition. 1) We call an ideal J non-atomic if x J for every
x Dom(J).
2) We call the ideal with domain 0, which is , the trivial ideal.
3.2 Denition. 1) For ideals J
1
, J
2
we say
J
1

RK
J
2
if there is a function h : Dom(J
2
) Dom(J
1
) such that
for every A Dom(J
2
) we have : A ,= mod J
2
h
(A) ,= mod J
1
or equivalently,
J
2
h
1
(A) : A J
1
.
2) For families I
1
, I
2
of ideals we say I
1

RK
I
2
if there is a function H witnessing
it; i.e.,
(i) H is a function from I
1
into I
2
(ii) for every J I
1
we have J
RK
H(J).
3) For families I
1
, I
2
of ideals, I
1

RK
I
2
if I
1

RK
I
2
& I
2

RK
I
1
.
3.3 Claim. 1) If an ideal J is not non-atomic then J
RK
the trivial ideal.
2)
RK
is a partial quasi-order (among ideals and also among families of ideals).
Proof. Easy.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

36 SAHARON SHELAH
3.4 Denition. 1) For an (upward) directed quasi order
4
L = (B, <) we dene
an ideal id
L
:
id
L
=
_
A B : for some y L we have A x L : y x
_
.
Equivalently, the dual lter l
L
is generated by the cones, where the cone of L
dened by y L is
L
y
=: x L : y x.
We call such an ideal a partial order ideal. We let Dom(L) = Dom(id
L
)(= B), but
we may use L instead of Dom(L) (like x L).
2) For a partial order L let dens(L) = min[X[ : X Dom(L) is dense
5
; i.e.
(a Dom(L))(b X)[a b]) (this applies also to ideals considered as (I, )).
3) For a family L of directed quasi orders let id
L
= id
L
: L L.
3.5 Fact. 1) id
L
is -complete i L is -directed.
2) dens(L) = dens(id
(L,<)
, ).
3) If h : L
1
L
2
preserves order (i.e. x, y L, (x y h(x) h(y))) and has
conal (= dense) range (i.e. (x L
2
)(y L
1
)[x h(y)]) then id
L
2

RK
id
L
1
.
4) Let h : L
1
L
2
. Now h exemplies id
L
2

RK
id
L
1
i for every x
2
L
2
there
is x
1
L
1
such that: (y)[y L
1
& x
1

L
1
y x
2

L
2
h(y)]; (equivalently for
every y L
1
: (x
2

L
2
h(y)) (x
1

L
1
y); note that h is not necessarily order
preserving).
5) If L
1
L
2
and L
1
is a dense in L
2
, then id
L
1

RK
id
L
2
.
6) If = density(L) then id
L
is -based; i.e. X Dom(id
L
), X / id
L
(Y
X)[[Y [ & Y / id
L
].
Proof. Straight. E.g.
3) If A L
1
and A / id
L
1
, then (x L
1
)(y)(x
L
1
y A) hence (x
L
2
)(y, z L
1
)(x
L
2
h(y) & y
L
1
z A) hence (x L
2
)(z)(x
L
2
z
h
(A)) hence h
(A) / id
L
2
(and trivially h
(A) L
1
).
By Dention 3.2(1) this shows id
L
2

RK
id
L
1
.
4) Note: h exemplies id
L
2

RK
id
L
1
i
4
no real dierence if we ask partial order or just quasi orders; i.e., partial orders satisfy x
y & y x x = y, quasi order not necessarily; note that in the case there is x
L such that
(y L)(y x
) gives an ideal which is not non-atomic

5
also called conal in this context
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

(A L
1
)(A ,= mod id
L
1
h
(A) ,= mod id
L
2
)
i
(A L
1
)(x
2
L
2
)[A ,= mod id
L
1
h
(A) y L
2
: x
2

L
2
y , = ]
i
(x
2
L
2
)(A L
1
)(A ,= mod id
L
1
h
(A) y L
2
: x
2

L
2
y , = )
i
(x
2
L
2
)(y L
1
: (x
2

L
2
h(y)) = mod id
L
1
]
i
(x
2
L
2
)(x
1
L
1
)(y L
1
)[(x
2

L
2
h(y)) x
1

L
1
y]
i
(x
2
L
2
)(x
1
L
1
)(y L
1
)[x
1

L
1
y x
2

L
2
h(y)].
5) Letting h
1
be the identity map on L
1
by part (3) we get id
L
2

RK
id
L
1
; choose
h
2
: L
2
L
1
which extends h
1
, by part (4) we get id
L
1

RK
id
L
2
, together we are
done.
6) Easily.
3.5
3.6 Fact. 1) For any ideal J (such that (Dom(J)) / J), letting J
1
= id
(J,)
, we
have
(i) J
1
is a partial order ideal
(ii) [Dom(J
1
)[ = [J[ 2
|Dom(J)|
(iii) J
RK
J
1
(iv) if J is -complete, then (J, ) is -directed hence J
1
is -complete
(v) dens(J, ) = dens(J
1
, )
(vi) dens(J, ) [J[ 2
|Dom(J)|
.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

38 SAHARON SHELAH
2) For every ideal J and dense X J we can use id
(X,)
and get the same
conclusions.
3) For every ideal J there is a directed order L such that:
J
RK
id
L
, dens(J) = dens(L)
and: for every if J is -complete, then so is id
L
.
Proof. Least trivial is (1)(iii), let h : J Dom(J) be such that h(A) (Dom(J))A
(exists as (Dom(J)) / J). Let J
1
= id
(J,)
, so h is a function from Dom(J
1
) into
Dom(h) and we shall prove that h exemplies the desired conclusion J
RK
J
1
by
Denition 3.2(1)
Assume toward contradicion that X Dom(J
1
) = J, X / J
1
and A =: h
(X)
belongs to J. So Y =: B J : (A B) id
(J,)
= J
1
(by the denition of
J
1
= id
(J,)
) hence (as X / J
1
) for some B X we have B / Y hence by denition
A B, so h(B) h
(X) = A contradicting the choice of h(B) (as A B).

3.6
3.7 Remark. So we can replace a family of ideals by a family of directed quasi orders
without changing much the relevant invariants such as completeness or density as
long as we do not mind adding larger ones in the appropriate sense.
3.8 Conclusion. For any family of ideals I there is a family of L of quasi order
such that:
(i) I
RK
id
(L,<)
: (L, <) L
(ii) [L[ [I[
(iii) sup[L[ : (L, <) L = sup[J[ : J I sup2
Dom(J)
: J I
(iv) supdens(L, <) : (L, <) L = supdens(J, ) : J I
(v) I is -complete i every (L, <) L is -directed.
Proof. Easy.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

3.9 Denition. For a forcing notion Q, satisfying the -c.c., a Q-name L
of a
quasi order with Y V,
Q
Y = Dom(L
) V for notational simplicity given (i.e.,

is not just a Q-name); let L
= ap
(L
) = ap
(L
, Q) be the following quasi order

Dom(L
) = a : a Y ) and [a[ < V

a
b i
Q
(y a)(x b)[L
[= y < x]
(this is a quasi-order only, i.e. maybe a
a but a ,= b).
3.10 Claim. 1) For a forcing notion Q satisfying the -c.c. and a Q-name L
of a
-directed quasi order (with Dom(L
) V given, not just a Q-name, for simplicity)

such that we have:
(i) ap
(L
) is -directed quasi order (in V and also in V

Q
)
(ii) [ap
(L
)[ [Dom(L
)[
<
(iii)
Q
id
L
[G]

RK
id
ap
(L
2) For a forcing notion Q satisfying the -c.c., the local
-c.c. and a Q-name L
of
a -directed quasi order such that we can nd L V such that
(i) L is a family of < -directed quasi orders (in V and also in V
Q
)
(ii) for each L L for some we have [L[
<
and [Dom(L
)[ , = ,
=
(iii)
Q
id
L
[G]

RK
id
L
for some L L.
3) In (2), of course, we can replace I by a -complete
RK
-upper bound.
Proof. 1) We leave (i), (ii) to the reader. We check (iii). Let G Q be generic over
V, and in V[G] we dene a function h from ap
(L
) to Dom(L
[G]) as follows: h(a)

will be an element of Dom(L
[G]) such that

(x a)[L
[G] [= x h(a)].
It exists by the -directed, assumptions. We can now easily verify the
condition in 3.5(4).
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

40 SAHARON SHELAH
2) Let p
i
: i < i
be a maximal antichain of Q such that: p

i

=
i
and
Q
p
i
satises the
i
-c.c and without loss of generality p
i
Dom(L
) =
i
. Let
Q
i
= Q
p
i
and L
i
be L
restricted to Q
i
and lastly let L = ap
i
(L
i
, Q
i
) : i < i
,
by part (1) we have p
i

Q
id
L
[G
Q
]

RK
ap
i
(L
i
, Q
i
) by p
i
: i < i
is a maximal
antichain of Q hence
Q
id
L
[G
Q
]

RK
ap
i
(L
i
, Q
i
) : i < i
.
A base [ap
i
(L
i
, Q
i
)[ [Dom(L
i
)[
<
i
=
<
i
i
and ap
i
(L
i
, Q
i
) is -directed. To-
gether we are done as necessarily i
< .
3) Easy by 3.13(3),(4) below.
3.10
3.11 Conclusion. 1) Suppose Q is a forcing notion satisfying the local -c.c., I
1
a
Q-name of a family of -complete lters and . Then there is, (in V), a family
I
2
of -complete lters such that:
(i)
Q
I
1

RK
I
2
(ii) if Q satises the -c.c. then [I

2
[ = [I
1
[ + i.e.
Q
[I
2
[ = [I
1
[
V
+
(iii) sup[Dom(J)[ : J I
2
= sup(2
)
<
: some q Q
forces that some J I
1
has domain of power .
2) If in V
Q
the set I
1
has the form id
(L,<)
: (L, <) L
1
; i.e., is a family of quasi
order ideals, then in (iii) we can have
(iii)
sup[Dom(J)[ : J I
2
= sup
<
: some q Q
force some (L, <) L has power
(iv) I
2
is a family of quasi order ideals.
Proof. Easy.
3.12 Remark. The aim of 3.10, 3.11 is the following: we will consider iterations
P
i
, Q
i
: i < ) where
P
i
Q
i
satises UP(I
i
), but I
i
may not be a subset of the
ground model V. Now 3.11 gives us a good
RK
-bound I
i
of I
i
in V, and we can
prove (under suitable assumptions) that P
will satisfy the UP(

_
i<
I
i
).
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

3.13 Denition. 1) For a family L of directed quasi order let the (< )-closure
c
(L) be
L
_
i<
L
i
: L
i
L for i < and <
_
(the partial order on
i<
L
i
is natural).
2) L is (< )-closed if for any < and L
i
L for i < there is L L such
that i < L
i

RK
L.
3.14 Claim. 1) If L is -complete, then c
(L) is -complete.
2) L
j

RK
i<
L
i
for j < .
3) If is regular, then c
(L) is -directed under

RK
and is (< )-closed.
Proof. Use 3.5(4).
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

42 SAHARON SHELAH
4 UP reintroduced
The reader may concentrate on the case S =
1
.
4.1 Convention. I will be a set of quasi order ideals, i.e. I = I
L
.
4.2 Denition. Fix I, S, W assuming
()(a) I a family of
2
-complete quasi-order ideals
(b) S is a set of regular cardinals to which
1
belongs
(c) W
1
is stationary.
1) We say

N is an I-tagged tree of models (for or for (, x)) if there is an I-tagged
tree (T, I) such that

N = N
: (T, I)) satises the following:

(a) for T we have N
(H (), , <
) is a countable model
(b) I N
and x N
(if x is present, we can use I as a predicate)

(c) T implies N
(d) for T we have N
and I
I.
Whenever we have such an I-tagged tree

N of models, we write N
=
_
k<
N
k
for
each lim(T).
1A)

N = N
: (T, I)) is a tagged tree of models if this occurs for some

2
-
complete I. If x is not mentioned, we assume it contains all necessary information,
in particular I, S, W.
1B) In part (1) we say weak I-tagged tree of models if we replace clause (d) by
(d)
(i) for T Dom(I) we have I
I
(ii) if Split(T, I) then for some one-to-one function f N
with domain
Suc
T
() and range Dom(I
) we have Suc
T
() f() N
.
We say weaker if we omit clause (d)
(ii). In all the denitions below we can use

this version (i.e., adding weak/weaker and replacing (d) by (d)
/(d)
(i)
2) We say

N is truely I-suitable (tagged tree of models) if clauses (a) (d) and:
(e) if T and I I N
then the set

T
[]
: split(T) and I
= I N
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

[]
. So

N is truly I-suitable tree (of models) does not imply

N is an I-tagged tree of models as possibly I / N
<>
.
3) We say

N is I-suitable (a tagged tree of models) if clauses (a) (d) and:
(e)
if T and I I N, then the set

T
[]
: split(T) and I
RK
I
N
[]
.
4) We say

N is -strictly (I, W)-suitable if

N is I-suitable and in addition
(b)
+
I N
, W N and x N
(f) one of the following cases holds:

(i) =
2
and for some W, for all T we have: N

1
=
(ii) = cf() >
1
, N
<>

1
W and T N
<
(i.e.,
N
is an initial segment of N

(note: if =
+
this means N
= N
so no contradiction with
the case =
2
)
(iii) =
1
and we demand nothing.
If we omit , we mean =
2
. So = cf() >
1
implies T
N

1
= N
<>

1
W. If we omit strictly we demand only (b)
+
.
5) We say

N is S-strictly (I, W)-suitable, if in addition to clauses (a) - (d),(e)
we
have:
(b)
+
I N
, W N
, S N
and x N
(g) for all T and S N
there is
, < such that

( T)[ T sup(N
) =
].
6) We say

N is
+
-uniformly (I, W)-suitable if it is (I, W)-suitable and
(b)
+
I N
, W N
, N
and x N
(f)
for some a []
0
for every lim(T) we have
N
= a
and
if
+
=
2
then a W.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

44 SAHARON SHELAH
6A) We say

N is S-uniformly (I, W)-suitable if it is (I, W)-suitable and
(b)
+
I N
, W N
, S N
and x N
(g)
for every T, S N
for some a []
0
for every satisfying
lim(T) we have
N
= a.
7) In 4), 5) if we add truely if (e)
is replaced by (e).
8) If S
is a P-name then in the clauses above we mean S
= :
P
/ S
.
9) I
[]
= I I : I is -complete.
4.3 Denition. 1) We say N (H(), , <
) is S
-strictly or -strictly (I, W)-

suitable model if there is an S
-strictly or a -strictly (I, W)-suitable

N such that
N = N
, (see 4.2(5),(9), and see 4.2(8), it applies). We can add truely.

2) We say N is (I, W)-suitable if it is strictly (I, W)-suitable, that is
2
-strictly
(I, W)-suitable.
4.4 Denition. In Denitions 4.2, 4.3 we may omit W when it is
1
, and may
omit S when S =
2
, for 4.2(5) and 4.2(6), 4.2(6A). We may replace S by if
S = UReg. Let (T, I) means T and we write T when I is clear.
4.5 Claim. Assume
(i) S
, P, I, W, x H () and S
is as in 4.2(8)
(i.e., S
= :
1
= cf() [P[ and
P
S

1
,
e.g., S
=
1
), and
(ii) I is a
2
-complete or for each I I, S, I is -indecomposable.
Then there is an S-strictly, truely (I, W)-suitable tree

N with x N
.
Proof. We will construct this tree in three steps: rst we nd a suitable tree, then
we thin it out to be a S
-uniformly suitable tree, then we blow up the models to

make it S
-strict. For notational simplicity let S
=
1
(the reader can check the
others).
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

First Step: An easy bookkeeping argument (to ensure 4.2(1)(e)) yields a truly
(I J
bd
1
)-suitable tree N
: (T, I)) satisfying sup(N

1
) <
sup(N

1
) such that N
(H (), , <
) and x N
<>
; so for lim(T) we
let N
=
_
<
N
.
Moreover we can get that for all lim(T), for each I (I N
) J
bd
1
, there
are innitely many k such that k split(T, I) and I
k
= I and Suc
T
( k) =
x) : x Dom(I).
Second Step: Dene H : T
1
by H() = sup(N

1
) <
1
. Apply 2.12 to get
a subtree T
and a limit ordinal W

1
such that clauses (a) - (d) of 2.12 hold
for . By clause (d) of 2.12, for all T
, N
1
. Let
0
<
1
< . . .
_
n
n
= ,
and let
T
=
_
T
: for each k < g(), if Suc

T
( k) = k) : <
1
so Suc
T
( k) = k) : < )
then (k) =
k
_
.
Clearly T
will be
1
-uniformly suitable; i.e.
lim(T) N
,

1
= .
Third Step: For T
2
, let N
= the Skolem Hull of N
in (H (), , <
). So
N

1
. Conversely, let lim(T
2
), , then N
, so N
hence N
. So N

1
= , i.e. N
: T) is an
1
-strict, (I, S
, W)-tree
of models (see Denition 4.2(4)).
We claim that this tree is still truly suitable. Indeed, assume T
2
,
lim(T
2
), and I I N
. Then for some < , I is in the Skolem hull

of N
. Let k < be such that N

k

1
and k g(). Then since
N
: T
2
) was suitable, there is k such that I
= I. So N
: T
2
) is
also (I, S
, W)-suitable.
4.5
4.6 Fact. Assume I
RK
I, where I, I
are families of ideals.

1) If N
: (T, I)) is a I-suitable tree and I
, then N
: (T, I)) is also

I
-suitable.
2) If N
: (T, I)) is I-suitable and I
, then there is a tree (T
, I
) satisfying
the following for some T
, f:
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

46 SAHARON SHELAH
(a) T
T and f is an isomorphism from T
onto T
, (i.e., is one to one onto

preserving length and and its negation) and T

RK
I
f()
, I
,=
I
(b) N
: (T
, I
)) is truely I
-suitable when we let I
= I
f()
(c) split(T
, I
) = T
split(T, I) if I
RK
I.
3) We can weaken the hypothesis to I
RK
I the trivial ideal. The same holds
in similar situations.
4) In Denition 4.4, if S = :
1
= cf() , = cf(), then clause (f)
of 4.2(4) (i.e.,
+
-strictly) and clause (g) of 4.2(5) (i.e. the demand concerning S,
i.e., S-strictly) are equivalent. Similarly 4.2(6), 4.2(6A) are equivalent.
5) In part (2), N
: (T
, I)) is a weak I
-tagged tree, truely I
-suitable; moreover
it is enough to assume N
: (T, I)) is a weak I-suitable tree (see 5.2).

6) If N
: (T, I)) is a weak I-tagged tree then for some tree T
and tree
isomorphic f from T
onto T letting I
= I
f()
: T
) we have N
f()
:
(T
, I
)) is a I-tagged tree. All relevant properties are preserved. [Check, see 5.2.]
Proof. 1) Should be clear, as
RK
is transitive (as a relation among ideals and also
among families of ideals).
2) For every Y =: T : (I
)(I
) and split(T, I) pick an

ideal I
I N
, I

RK
I
such that: for every T, for every I
I N
the
set T
[]
: I
= I
and and Y contains a front of T

[]
. This can be
done using a bookkeeping argument.
Now dene T
as follows. We choose by induction on n, a function f

n
with
domain T
n
Ord, such that Dom(f
) f() N

n
Ord. Let f
0
be the
identity on <>. Assume f
n
has been dened and Dom(f
n
), and we shall
dene f
n+1
Suc
T
(). If TY , then Dom(f
n+1
) Suc
T
() Suc
T
() is a
singleton
and let f
n+1
(
) = f
n
()(
(n)). If T Y , then I
is already
dened and it belongs to N
. Let g
be a witness for I

RK
I
and stipulate
Dom(I
) x : < x > Suc

T
(). Now g
introduces an equivalence relation

on Dom(I
). Let A
be a selector set for this equivalence relation; i.e. g
is
1-1 and has the same range as g
. Note that we can choose g
in N
as I
, I
(whereas Suc
T
() does not necessarily belong to N
) and then choose A
and let
A
=: x A
: (y Suc
T
())[g
(x) = g
(y)] (so possibly A
/ N
). Now
for x A
n
so < x > Suc
T
() we let f
n+1
( < x >) = f
n
()g
n
(x)) so
Dom(f
n+1
) Suc
T
() = < x >: x A
. Lastly, let T
= Rang(f
n
) : n <
, T
= Dom(f
n
) : n < and for Dom(f
n
), n < let I
f
n
()
= I
and
f = f
n
: n < . Now check.
3), 4) Left to the reader.
4.6
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

4.7 Denition. 1) Let >
0
, I H () a set of ideals and S H () a set
of regular cardinals (or just limit ordinals) such that
1
S. For N a count-
able elementary submodel of B = (H (), ) (or B an expansion of (H (), )
with countable vocabulary) such that I, S N we dene Dp(N) = Dp
I
(N) =
Dp
I
(N, B) = Dp
I
(N, S, B) Ord , by dening when Dp(N) for an
ordinal , by induction on :
Dp(N) i N is as above and for every I I N
and for every < and X I there is M satisfying :
(i) Dp(M) (hence M B is countable)
(ii) N M
(iii) sup(M
1
) = sup(N
1
) moreover
S N sup(N ) = sup(M )
(iv) M Dom(I)X ,= .
2) We dene Dp
(N) = Dp
I
(N) by dening: Dp
(N) i N is as above and for

every J I N and < for some I N I we have J
RK
I and for every
X I there is M satisfying (i)-(iv) above.
4.8 Claim. 1) In Denition 4.7:
(a) Dp(N) Ord if well dened
(b) if Dp(N) = , I IN then we can nd Y I
+
(i.e., Y Dom(I), Y / I)
and

N = N
t
: t Y ) such that:
(i) Dp(N
t
) =
(ii) N N
t
(iii) sup(N
1
) = sup(N
t

1
) moreover S N sup(N ) =
sup(N
t
).
2) If I
1

RK
I
2
and I
1
, I
2
N and Dp
I
2
(N) then Dp
I
1
(N) .
3) Dp
I
(N) = Dp
I
(N).
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

48 SAHARON SHELAH
4.9 Claim. 1) Let N (H (), ) be countable, I N, N
1
W . Then
(a) N is strictly (I, W )-suitable i Dp
I
(N) =
(b) N is (I, W )-suitable i Dp
I
(N) = .
2) Similarly with S.
Proof. Easy.
4.10 Denition. 1) The forcing notion P satises UP
(I, S
, W) (note if ,= 2 we
may omit
) (adopting the conventions of 4.2(8);
is a purely decidable P-name

of a V-cardinal) when: 0, 1, 2 and if is large enough and

N is S
-strictly
(I, W)-suitable and p N
P and P N
, of course, then there is q P such

that p
pr
q P and:
(a) if = 0 then q N
[G
P
]
1
= N

1
and sup(N
<>
[G
P
] ) =
sup(N
<>
) if S
(b) if = 1 then q for some lim(T) we have N
[G
P
]
1
= N

1
and for every S
we have sup(N
[G
P
] ) = sup(N
) where N
=
N
: < and is not necessarily from V

(c) if = 2 then q N
[G
P
] is (S
)-strictly (I
[
]
, W)-suitable and
sup(N
<>
[G
P
] ) = sup(N
<>
) for every S
(in particular
V
1
) where
=
V[G
P
]
2
.
2) If we omit we mean = 0.
If W =
1
we may omit it. We write instead of S
if
S
= : V[G
P
] [= UReg
V[G
P
]
. If we omit S
we mean
V
1
.
3) The forcing notion P satises UP
4
(I, S
, W) when
are (P,
pr
)-names of
regular V-cardinals and for some x we have: if is large enough and

N = N
:
(T, I)) is an S
-strictly (I, W)-suitable tree of models for (, x) and p P N
and
, P, I) N
, then for some q, T
we have:
(a) p
pr
q P
(b) q is (

N T
, P)-semi
4
generic (see below).
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

3A) If
= we can replace T
by
such that T
n : n < (see below) so

lim(T) and then in clause (b) write N
instead of

N T
. We then may
omit . We may
if
(P), see Denition 1.29(5).

3B) We say that q is (

N T
, P)-semi
4
generic where

N, T
is as above if:
q
P
(i) T
is a subtree of T (so T
T is closed under
initial segments <> T
, T
Suc
T
() ,= )
(ii) N
[G
P
]
1
= N

1
for T
(iii)

N[G
P
] T
has (
)-covering which means: if
is an -branch of T
and y N
[G
P
]
is a set of <
[G
P
] ordinals
(if
[G
P
] is not a cardinal, this means [
[G
P
][)
then for some A V
_
<
N
we have
[A[
V
<
[G
P
] and y A
(iv) N
[G
P
] : (T
, I)) is a strictly
I
[
[G
P
]]
-suitable tree of models.
4) We dene UP
3
(I, S
, W) similarly replacing clause (b) by the weaker

(b)
q is (N
, P)-semi
3
-generic, (see below).
4A) If
= we can replace T
by
such that T
n : n < so
P

lim(T) and replace

N T
by N
. We may omit
if it is (but see 4.12(0)). We

then may omit
if it is , too.
4B) q is (N,
, P)-semi
3
-generic is dened as in (4) only replacing clause (iii) in
of (3B) by
(iii)

N[G
P
] T
has (
, 1)-covering which means for every y V N
[G
P
] for
some A V N
we have [A[
V
<
[G
P
] and y A, recalling N
[G
P
] =
N
[G
P
] : < .
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

50 SAHARON SHELAH
5) We allow to use I
, a P-name of an element of V as above if:

(a) it is decidable purely
(b) if q P decides I
= I then P
q
satises UP
(I, S
, W).
6) We say that P satises the UP
5
(I, S
, W) i
(a)
and
are P-names of V-cardinals such that P satises the
-c.c. purely
locally
(b) I
is a P-name of a set which belongs to V, it is a set of ideals and I
is
decidable purely
(c) is -complete if
P
(
= & I I
),
(d) if p P forces I
= I, I I
= ,
= and I
I is a set of -complete
ideals then for some x
if

N = N
: (T, I)) is strictly I
-suitable, x N
<>
, then for some
q, T
we have: p
pr
q P and q is (

N T
, P)-semi
5
-generic, (see
below).
7) We dene P satises UP
5
(I
, S
, W) as in part (6) but restrict ourselves to I
= I.
[others?]
8) Assume

N = N
: (T, I)) and P N

<>
satises UP
(I
, W) and P, I
, W, . . . )
N
<>
. We say q is (

N,
, P)-semi
5
-generic (for

N when not understood from the
context) if:
q for some T
we have (T, I
(T
, I), see 2.4, clause (f) and

T
[G
P
]
1
= N
<>

1
and
T
& S
sup(N
[G
P
] ) = sup(N
).
9) We write UP
3
(I
, S, W) for UP
3
(I
, S, W) where
is
(P) see 1.29(5), 1.30(1).

4.11 Denition. [?] We call I to be a name if it is a name of an old family of
ideals purely decidable.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

4.12 Claim. 0) In Denition 4.10(3)-(6), if
(P), then the demand concerning
(i.e., clause (iii) of 4.10(3B) holds trivially (as increasing p purely, p
=
and P
p
satises the -c.c).
1) If Q satises UP
4
(I, S
, W), then it satises UP

3
(I, S
, W). If Q satises
UP
3
(I, S
, W), then Q satises UP

1
(I, S
, W) and UP
2
(I, W). If Q satises UP

2
(I, S
, W)
or UP
1
(I, S, W) then it satises UP
0
(I, S, W). If 3, 4 and Q satises UP
(I, I
, W)
and
, [ = 4
1
=
], then Q satises UP
1
,
1
(I, S
, W).
1A) If Q satises UP
(I, S
, W), then it satises UP

0
(I, S
, W) which implies Q
has pure
1
-decidability, see Denition 1.23(2).
2) The forcing notion Q satises UP
(I, S
, W) i its completion (i.e., Q

1
, or equiv-
alently its completion to a complete Boolean algebra) satises it assuming
pr
=.
3) If Q satises UP(I, , W), (i.e., see 4.10(2)) and I is
+
-complete (e.g., I = ),
then any new countable set of ordinals < is included in an old countable set
of ordinals; i.e., one from V.
4) Q satises UP(, ) i Q is purely proper (see Denition 1.25(1)).
5) Q satises UP(,
1
) i Q is purely semiproper (see Denition 1.25(2))
6
.
6) If Q satises UP(I, S
, W) and I I
1
, S
1
S
and W
1
W then Q satises
UP
(I
1
, S
1
, W
1
).
7) In Denition 4.2, if P satises the -c.c. (e.g. = [P[
+
) then:
(a) we can replace S
by any set S
of uncountable regular cardinals of V, such

that
P
S
= S
.
8) In Denition 4.10 (in all the variants), if we demand for large enough, for
some x H (), for every

N such that x N
and . . . we get an equivalent

denition.
9) In Denition 4.10 we can use weak I-tagged trees, i.e. we get with this an
equivalent denition.
Proof. 1), 2) Trivial.
3) Straightforward.
4) Use 4.5 below.
5),6) If I = , then every N (H (), , <
) is a I-model.
6
that is: if Q N (H (), ), N is countable, p Q N, then for some q we have p
pr
q
and q
Q

N for every Q-name
N of a countable ordinal
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

52 SAHARON SHELAH
7) Easy.
8) Check the Denition.
9) As in [Sh:f].
10) The weak version allows more trees of models so apparently is a stronger
condition, but by 4.8(4) it is equivalent.
4.12
4.13 Conclusion. If P satises UP
(I, S
, W) and S
is as in 4.2()(b) (or S
=
1
)
(recall that this notation implies I is
2
-complete,
1
S
and W
1
stationary)
then
P
W is stationary. Moreover, if W
W is stationary then also

P
W
is a stationary subset of
1
.
Proof. The moreover fact is by 4.12(7) (i.e., monotonicity in W).
Assume that p C
is a club of
1
and C
W = . By 4.5 we can nd an
1
-strictly (I, S
, W)-suitable tree of models N
: (T, I)) with C
, p N
. Let
= N
<>
1
, so W. By UP
(I, S
, W) we can nd a condition q as in Denition

4.2 in particular p
pr
q. Clearly q N
[G]
1
= and, trivially p
P
C
is
unbounded in N
[G]
1
hence p N
[G]
1
C
. So q
Q
C
W,
contradiction.
4.13
4.14 Remark. Usually we assume I, S satises 4.2(*)(a) + (c), S
=
1
is the main
case.
4.15 Remark. 1) From the proof of 4.5 we can conclude that in 4.2; we can replace
S-strictly (I, W)-suitable, N

1
= W by (I, S, W)-suitable, and then
the condition q will be N
-semi-generic.
2) As at present S
=
1
seem to suce, we shall use only it for notational
simplicity.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

5 An iteration theorem for UP
5.1 Claim. 1) If

N = N
: (T, I)) is a tagged tree of models for (, x, P)), P

a forcing notion and P N
, then
P
N
[G
P
] : (T, I)) is a tagged tree of
models for (, x, P, G)).
2) If in addition P satises the -c.c. and I N
is -closed (see Denition

3.13(2)) and

N is I-suitable, then
()
P
N
[G
P
] : (T, I)) is I-suitable.
3) If in part (2) assume in addition that
, I
to be P-names of objects from V

such that I
is purely decidable and P satises the local
-c.c. purely, then for every

p N
<>
P there is q, p
pr
q N
<>
P forcing () above. If I
is purely decidable
and P is locally
-c.c. purely then we can nd q satisfying p

pr
q N
<>
P and
forcing ().
Proof. 1) Straight.
2) So P satises the -c.c. and let G P be generic over V. Now from Denition
4.2(1), clearly N
[G] : (T, I)) satises clauses (a) - (d), so it is enough to check

clause (e)
of Denition 4.2(3). So let I I N
[G] where T. Hence there is

I
such that I
is a P-name and I
[G] = I. Let I
= J I : for some p P
we have p
P
I
= J. So I
belongs to V and is a subset of I of cardinality

< and I
hence there is I
I such that (J)(J I
J
RK
I
), so
without loss of generality I
, hence as N is I-suitable clearly :

T and I
RK
I
contains a front of T[]. Hence in V[G], the set T : I

RK
I
[]
as required.
3) Left to the reader.
5.1
The point of the following claim is to get more from some UP
than seems on the

surface; our aim is to help iterating.
5.2 Claim. 1) Assume 3, 4 and the forcing notion Q satises UP
(I
, W)
and
, I
, I
+
are Q-names with pure decidability and I
+
I
is
-complete. Then
the forcing notion Q satises the UP
(I
+
, W).
2) Suppose
(a) I
0
, I
1
, I
2
, I
3
are sets of quasi-order ideals, I
1
I
0
I
2
, I
3
= I
1
(I
2
I
0
), I
2
I
0
=
I
3
I
1
is -closed, and I
2
I
0
is -complete
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

54 SAHARON SHELAH
(b)

N = N
: (T
, I
)) is a strict truely I
2
-suitable tree of models (for
and x = Q, I
0
, I
1
, I
2
, , )
(c) p N
<>
, 3, 4 and Q
p
satises the -c.c.
(d) () is a property with

N, G
Q
as parameters (and possibly others)
(e) for any T
a subtree of T
such that N
: (T
, I)) is a truely I
0
-suitable
tree of models there are q = q
T
, T
= T
(T
) such that
(i) p
pr
q Q
(ii) (T
, I
) (T
, I
)
(iii) q
Q
N
[G
Q
] : (T
, I)) is a truely I
1
-suitable tree of models and
for every T
we have
() N
[G
Q
]
1
= N
<>

1
and
() []
() if = 3, y N
[G
Q
] is a member of V then
: T
, and for some A N
, A a set of cardinality <

we have y A contains a front of T
[]
() if = 4 and y N
[G
Q
] is a set of < members of V then
: T
and for some A N
a set of cardinality <

we have y A contains a front of T
[]
.
Then there are q, T
such that:
(i) p
pr
q Q
(ii) (T
, I
) (T
, I
)
(iii) q
Q
N
[G
Q
] : (T
, I)) is a truely I
3
-suitable tree of models and for
every T
we have
() N
[G
Q
]
1
= N
<>

() [= ()
(), () as in clause (iii) of (e) above.
3) In part (2), if Q satises the -c.c., then we can omit (), () in their two
appearances as they follow.
4) In part (2), we can replace truely I
-suitable by weakly I
-suitable.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

5.3 Remark. In part (2) clause (e) we can restrict T
to those needed.
Proof. 1) As in the denition of UP
(I
+
, W) let

N = N
: (T, I)) be a strict

I
+
-suitable tree of models for and x Q, I
, I
+
, W,
) and p N
<>
. We can
nd p
N
<>
, p
pr
p
Q which forces
, I
, I
+
to be say , , I, I
+
respectively.
Now we can apply part (2) of the claim with

N, I, I
[]
, I
+
, I
[]
(I
+
I), x = x here
standing for

N, I
0
, I
1
, I
2
, I
3
, there.
2) Let T = T : T is a subtree of T
such that

N T is a truely I
0
-suitable tree
of models or just T = T : T a subtree of T
such that (<> T, T , =

Suc
T
() Suc
T
(), T closed under initial segments and): if split(T
, I
) &
I
I
0
then Suc
T
() I
+
and if / split(T
, I
) I
/ I
0
then [Suc
T
()[ = 1.
For any T T by assumption (e) there are q
T
, T
Q
[T] as required there.
We shall show that some such q
T
is as required. We dene a Q-name T
as follows:
for G
Q
Q generic over V we let
T
= T
[G
Q
] = T
:N
[G]
1
= N
<>

1
and g() ( ).
Clearly
()
1
q
T

Q
T
Q
[T] T
for every T T .
Working in V[G
Q
] we dene a depth function Dp function from T
to Ord
by dening for any ordinal when Dp() as follows:
Dp() i the following conditions hold:
() T
() for every < there is Suc

T
() such that Dp()
() if < and g() and I N
[G] I
3
then for some with
Dp() we have T
& I = I
I
3
and
Suc
T
() : Dp() , = mod I
() if < and y
is a Q-name of a member of V when = 3 and

is a set of cardinality < when = 4 then for some
with
Dp(
) and A N
of cardinality < we have [ = 3 y
[G
Q
]
A] and [ = 4 y
[G
Q
] A].
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

56 SAHARON SHELAH
Clearly it is enough to show in V that for some T T we have q
T
Dp(<>) =
. Note
()
2
if T
then Dp() Dp().

Clearly in V
Q
we have
()
3
if T
, Dp() < and I
I
3
then Suc
T
() : Dp() <
Dp() = mod I
()
4
if T
, Dp() < and I
/ I
3
, Suc
T
() then Dp() Dp().
For each T
such that I
I
2
I
0
dene the set A
as follows:
First, A
is the minimal family of sets satisfying

(i) if = 3 and y
is a Q-name of a member of V then the set A

3
,y
=:
Suc
T
() : for some set A N
V of cardinality < , y
A belongs to
A
(ii) if = 4, parallely (using A

4
,y
), i.e., if y N
is a Q-name of a family
of < members of V then the set A
4
,y
=: Suc
T
(): for some set
A N
V of cardinality , we have y
A belongs to A
(iii) if g(), = Dp( ) then the set A
,
= Suc
T
() : Dp()
belongs to A
.
Let A
= A A
: A = mod I
, note that A
is a countable family of members

of I
(more exactly the ideal I

V
Q
, which I
generates in V
Q
), and so actually we
have dened a Q-name A

= A : A A
.
Now we can dene in V a sequence B
: T
) such that
()
5
B
= mod I
for T
()
6
(i) if Suc
T
() = mod I
or I
/ I
2
I
0
then B
=
(ii) if Suc
T
() ,= mod I
and I
I
2
I
0
then
p
Q
A
if T
.
This is possible as Q
p
satises the -c.c. and each I I
2
I
0
is -complete.
Now we dene
T
0
= T
: for every < g() we have ( + 1) / B
.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

Clearly we can nd T
1
T such that T
1
T
0
(in particular by ()
6
(i)). So if
q
T
1

Q
Dp(<>) = then we are done so toward contradiction we assume that
this fails. Hence, there is a subset G
Q
of Q generic over V such that q
T
1
G
Q
and
V[G
Q
] [= Dp(<>) < . As q
T
1
G
Q
clearly T
(T)[G] T
[G
Q
].
By the choice of G
Q
we have Dp(<>) < hence by ()
2
we have T
(T
1
)[G
Q
]
T
[G
Q
] Dp() < .
Now we shall prove by induction on Ord that T
(T
1
)[G
Q
] Dp() .
For = 0, limit we have no problem, so let = +1, assume toward contradiction
Dp() = for some T
(T
1
)[G
Q
], hence by ()
2
and the induction hypothesis we
have
2
T
(T
1
)[G
Q
] Dp() =
3
if I (I
2
I
0
) N
[G
Q
] then for some (T
(T
1
)[G
Q
], (in fact many
s) and J I
3
N
we have I = I
[why? by clause (e)(iii) of the assumption, i.e. choice of q

T
1
, T
(T
1
).]
4
if I I
3
N
[G
Q
] then there is satisfying T
(T
1
)[G
Q
] such that
I = I
[why? if I I
2
I
1
by
3
, if I I
0
use the choice of T
(T
1
).]
Now
5
if I N
I
3
then for some satisfying T
(T
1
)[G
Q
] we have I
= I
and
Suc
T
() : Dp() , = mod I
.
[Why true? We can choose such that I = I
and T
(T)[G
Q
] and
Suc
T
(T
1
)[G
Q
]
() ,= mod I
and choose
Suc
T
(T
1
)[G
Q
]
() T
[G
Q
]. First
assume I
I
3
I
0
.
Now easily
A
,g()
[G
Q
] = mod I
,g()
[G
Q
] B
/ A

[G
Q
]
by the denition of A
,g()
[G
Q
] and we get
Suc
T
[G
Q
]
() : Dp() Dp()(= )
easy contradiction.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

58 SAHARON SHELAH
Next assume I I
1
. Now Suc
T
(T
1
)[G
Q
]
() is a set witnessing the requirement.]
Now for T
(G
1
)[G
Q
] we check the denition of Dp() + 1: clause () holds
as T
(T
1
)[G
Q
] T
[G
Q
], clause () holds by the induction hypothesis, clause ()
holds by
3
+
4
+
5
and clause () by the choice of q
T
1
, T
(T
1
). So Dp() +1
contradiction.
3),4) Similar to the proof of part (2).
5.2
We now can deduce more implications between the UP
-s.
5.4 Conclusion. Assume that
are purely decided Q-names and I

[
]
is (<
)-
closed, (which just means: if r
= ,
= and I
= I where r P then I
[]
is
-closed) and Q satises the local
-c.c. purely.
1) If Qsatises UP
1
(I, W) then Qsatises UP

2
(I
, W) and UP
4
(I
, W), UP
3
(I
, W).
2) UP
1
(I
, W) implies UP
5
(I
, W) (for Q) if
is a Q-name decidably pure, Q

satises the local
-c.c.
3) 4-5 ll!!!
Proof. Why? By Denition 4.10 and Lemma 5.2.
5.4
5.5 Denition. 1) We say

Q = P
i
, Q
i
, I
i
,
i
, S
i
: i < ) is UP
4,e
(W, W)-suitable
iteration if:
(a) P
i
, Q
j
: i < , j < ) is an
1
Sp
e
(W)-iteration
7
(b) I
i
is a P
i
-name of a set of quasi order ideals with domain a cardinal in
V
P
i
for notational simplicity or even just a P
i+1
-name of such objects (i.e.,
P
i+1
I
V
P
i
) such that in V
P
i
, I
i
/G
P
i
which is a Q
i
-name, is purely
decidable
(c) W
1
is stationary
(d) for each i < , we have:
i
is a P
i
-name of a regular uncountable cardinal
of V
P
i
, purely decidable
(e)
P
i
Q
i
satises UP
4
i
,
i+1
(I
i
, S
i
, W) and I
i
is
i
-complete set of partial
order ideals (from V) and (Q
i
,
vpr
) is
1
-complete (see 1.1)
7
the reader can x W as the class of strongly inaccessible cardinals
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

(f) or i < j we have
P
j

(g) P
i+1
satises the
i
-c.c., or just for every p P
i+1
there are
, q such that
() p
pr
q P
i+1
() q
(P
i+1
,
pr
)

i+1
=
()
UReg and
is a P
i
-name of an ordinal for <
<
and
p
, q i p
P
i
then there is q
, p
pr
q
P
i
and set a of <
ordinals such that q
a for <
(h) [?] for any i < for some n < , if i < j < and p P
j+1
and
j
(p, P
j+1
) = and
is a (P
i
)
pi
-name of an ordinal for <
<
then for some q and a we have: p i
pr
q P
i
and q
P
i

a for
<
where a is a set of < ordinals.

2) We may write UP
4,e
instead UP
4,e
(
1
, the class of strongly inaccessibles). If we
omit S
i
we mean
1
.
If we omit I
i
, we mean some I
as required (note that the requirements on I
i
are
actually on each member so the family of candidates to being I
is closed under
union). If we omit e we mean e = 6. We may omit W if it is the class of strongly
inaccessible cardinals.
If we omit
i
we mean some such P
i+1
-name. (Can we eliminate names? Well if
we use the iteration as in [Sh:f, Ch.X,1], (RCS) no, but if we waive associativity
as done here, we can).
3) We dened UP
3,e
(W, W)-suitable iterations similarly but replace clause (e) by:
(e)
a
as above replacing UP
4
i
,
i+1
by UP
3
i
,
i+1
.
4) We dene a UP
(I, W, W)-iterations as above but with I
i
= I
i
]
[Saharon like
6, straight in successor, in limit work it out, Question:
i
pure name?]]
We can also deal with strong preservation
5.6 Denition. We say that

Q = P
i
, Q
i
,
j
: i < , j ) is a weak UP
4
(W, W)-
iteration if
(a) P
i
, Q
i
: i < ) is a
1
Sp
6
(W)-iteration,
i
is (P
i
,
P
i
pr
)-name such that
j < i
i
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

60 SAHARON SHELAH
(b) if i < j and i is non-limit, p P
i
and p
(P
i
,
pr
)

i
=
i
then
p
P
i
P
j
/G
P
i
satises UP
4
i
,
j
(I
, W) for some P
i
-name I
of
i
-complete
ideals.
5.7 Lemma. Assume that W
1
is stationary and

Q = P
i
, Q
i
, I
i
,
i
: i < ) is
a UP
4
(W, W)-suitable iteration, and P
= Sp(W) Lim(
Q) be the limit.
For j we dene
j
, a (P
j
,
pr
)-name of a member of UReg
V
:
j
=
Min UReg
V
: i < j
i
and
2
and j.
For simplicity we can assume
there are
i
: i < ), P
i
satises the
i
-c.c. and for each i, i + for
some
(P
i
,
pr
)

i
,
(P
i+1
,
pr
)

i+1
(and
= P
=
_
i<
P
i
(i.e.
is strongly inaccessible > [P

i
[ for i < or change support?)
1) For each , P
satises UP
5
(I
, W) for some
-complete I
V.
1A) If , p
P
, p G
P
then in V[G
P
] the forcing notions

P
/G
P
satises UP
5
(I
,
, W) for some
j
-complete I
,
V
P
(so this justi-

es the weak in 5.6).
2) In fact each I I
has domain of cardinality (sup

<
:
P
(I
_
i<j
I
i
)( = [Dom(I)[) and
i
> for some i and [I
<
(
0
+ [P
[ +
sup
:
P
([
_
i
I
i
[ and
i
> for some i ). Similarly for I
,
.
3) Similarly for weak UP
5
(W, W)-iterations.
5.8 Remark. 1) We can also get the preservation version of this Lemma.
2) The reader can concentrate on the case that
s are objects and not names.

Proof. 1) We prove this by induction on , so without loss of generality = . For
each < let J
=: q P
+1
: q forces a value to
, called
,q
and q forces
I
to be equal to a P
-name called I
,q
and q forces that [I
[ is
,q
but no
q
such that q
pr
q
forces a smaller bound. Let
= sup
qJ
,q
.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

Let q
P
= I
,
: <
,q

,q
for q J
and let J
,
= q J
,q
> and q Dom(I
,
) is
,q,
and no q
such that q
pr
q
forces a smaller bound and let I
,
be id
L
,q,
, so L
,q,
is a P
-name of a
,q
-
directed quasi order on some

,q,
(but
P
if [I
[ <
then let
L
,
be trivial). We can assume L
,q,
is a quasi order on
,q,
(putting every

,q,
Dom(I
,q,
) at the bottom.
For q J
let L
,q,
be ap
,q,q
(L
,
) for the forcing notion
P
[q]
= p P
: q
P
pr
p from Denition 3.9, so it is dened in V. So by Claim
3.10
(i) L
,q,
is
,q
-directed partial order on [
,q,
]
<
,q
(ii) [L
,q,
[ (
,q,
)
<
,q
(iii) q
P
,
= id
L
,q,

RK
id
L
,q,
.
Let
= sup
,q
: < and q J
.
Let I
be the (<
)-closure of id
L
,q,
: < , q J
, <
,q
(see Denition
3.13(1)).
Let

N = N
: (T
, I)) be a strict truely (I
, W)-suitable tree of models for

(, x), x coding enough information (so

Q, I
, W, W N
); why truely? see ?.

> scite4.X undened
Let T
N
be the set of quadruples (, q,
, T
) such that:
1
, q P
, T
is a P
-name of a subtree of T
,
q
(P
,
pr
)

and
q
P
[G
] : (T
, I T
))
is strictly ((I
)
[
]
, W)-suitable tree,
N
[G
]
1
= N

1
and
and , N
[G
]
and

N[G
] has ()-covering.
Now T
N
is dened similarly as the set of quadruples (
, q,
, T
) such that:
is a
simple (

Q, W)-named [0, )-ordinal, q P
a P
-name and
[G
P
]. (I.e. if
< , G
P
is generic over V and =
[G
P
] then r q
r
[G
P
] < , i.e.
is well dened < or is forced (
P
/G
P
) to be not well dened), and q

P
lim(T).
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

62 SAHARON SHELAH
We consider the statements, for <
,
for any (, q,
, T
) T
N
and
a P
-name such that

q
P
and N
[G

]
and p
a P
-name such that q

P
[G
P
] N
[G
] P
/G
P
and
(p
[G
P
])
pr
q and p
[G
P
] forces (
(P
,
pr
)
) a value
to
(usually
redundant) there is (, q
, T
) T
N
such that p
pr
q
(i.e., p
P
[G
]
pr
q) and q
= q and q
.
For simple (
Q, W)-names [0, )-ordinals
we dene
similarly ( <
) (
,
) and
.
Observation: If <
, (
,
) and
are simple

Q-named [0, )-ordinals

<
then
(dened naturally).
Proof. By induction on the depth of
(see 6.8, fact A).

We prove by induction on that
(a) P
has pure
-covering; i.e. if
is a P
-name of an ordinal <
and
p P
then for some q and a we have: p

pr
q P
, a V is a set of
ordinals and q [a[ <
&
a (even over P
)
(b) P
has pure (
1
,
1
)-decidability
(c) for every we have
,
(but for 5.7(3) we have to restrict ourselves
to non-limit ).
Note that for = the statement in clause (c) is trivial hence we shall consider
only < .
Case 1: = 0.
Trivial.
Case 2: a successor ordinal.
Clauses (a), (b) follows easily from clause (c) so let us concentrate on clause (c).
As trivially
0
,
1
&
1
,
2

0
,
2
, clearly without loss of generality
= + 1.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

Let G
P
be such that q G
P
and G
P
generic over V.
Let T
= : T
[G
P
],

N
= N
: (T
, I
)) where N
= N
[G
P
], I
=
I
.
By 5.2 applied to

N
we can nd p
, T
as required.
Case 3: a limit ordinal.
[Saharon note: it would be if
is a (P
,
pr
+)-name, < , G
generic we
should dene
/G
anyhow we can use real names but then
is just a (P
,
pr
)-
name. But if
are real cardinals no problem. But see clause (g) of denition of

UP.]
Proof of Clause (a). If we use the c.c. version: easier, hardest case is
= , so
strongly inaccessible.
Note that we have to prove the weak version. If the property fails for , p,
(so
p
P
an ordinal, etc.), then by the induction hypothesis is minimal so den-

able in (H (), ) from

Qhence necessarily N
<>
and without loss of generality p,
N
<>
. Also without loss of generality p
(P
,
pr
)

for some
UReg
V.
We shall now choose p
1
as in the proof of 1.26. Let
: < j) be a witness for p

(see 1.15 clause (F), in particular (a)(iii) of (F)), so without loss of generality j .
For each < j and < , let a
,
be a P
+1
-name of a set of <
ordinals and let

r
,
be a P
+1
-name of a member of P
/G
+1
with domain [ +1, ) such that if
G
+1
P
+1
[G
+1
P
] = , then r = r
[G
+1
] satises:
(a) if possible p
pr
(p ( + 1)) r P
and (p ( + 1)) r
P
a
where a = a
,
[G
+1
] V[G
+1
] is a set of <
ordinals and t
,
[G
+1
] =
true
(b) if not possible r = r
,
[G
+1
] is the empty function and a
,
[G
+1
] = and
t
,
[G
+1
] = false.
Also we can demand that:
(c)
1
[G
] = =
2
[G
] then r
1
,
[G
+1
P
] = r
2
,
[G
].
Let r
[G
] = r i for some we have
[G
] = and r = r
,
[G
], similarly
we dene a
. Let p
1
= p
: < j. Clearly p
pr
p
1
P
.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

64 SAHARON SHELAH
Next dene p
2
as in 1.26: (recall (g) + (h) of Denition 5.4). I.e. for each <
j, < , G
generic over V, p G
, r
[G
] = there are r
,
, a
,

V[G
] such that

Q
[G
] [= p
1

pr
r
1
,
and r
,
a
,
. So
really without loss of generality we have P
-names r
,
, a
,
such that
1
[G
] =
=
2
[G
] implies r
1
,
[G
] = r
2
,
[G
] and a
1
,
[G
] = a
1
,
[G
]. We dene r
, a
by: r
[G
] = r i for some < we have
[G
] = and r = r
,
[G
]
and similarly a
. Now let p
2
= p
1
r
: < j so p
1

pr
p
2
. We can nish as in
the proof of 1.26 [ll]!!!
Clause (b):
As in the proof of clause (a) without loss of generality we have , p,
N
<>
.
We dene also p
1
as in the proof of clause (a) trying to force a countable bound
for
. Let
: < ) be a witness for p P
(see Denition 1.15, clause (F)

in particular (a)(iii) of (F) and without loss of generality
: n < ) belong to
N
<>
). We now choose by induction on n a quadruple (
n
, q
n
,
n
, t
n
) such that
(i)
n
is a simple (
Q , W)-named [0, )-ordinal

(ii)
0
= 0,
n
<
n+1
(iii)
n
+ 1
n+1
(iv) (
n
, q
n
,
n
, T
n
) T
N
(v) q
n
= q
n+1

n
(vi) p
n

pr
q
n
.
No problem to carry it by the observation above.
Let p
2
=
_
n<
q
n
p, clearly p
pr
p
2
, q
n
p
2
, and so it is enough to prove
p
2

< N
<>

1
. So let p
2
G
with G
a subset of P
generic over V.
So there is p
+
2
G
satisfying p p
2
p
+
2
such that p
+
2

<
1
and
so p p
+
2
so without loss of generality p p
+
2
above
0
< . . . <
m1
, using 1.17.
There is n such that [
n
[G
],
_
<
[G
]) is disjoint to
0
, . . . ,
m1
, hence for
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

some < , letting =
[G
] dening r
0
,
[G
P
+1
] we get t
0
,
= truth (see?
1.26) and <
n
[G
].
Now consider N
= N
<>
[G
P
+1
] we know N

1
= N
<>

2
(by clause
(iv) above), and in it we have p ( +1) r
0
,
[G
p
+1
] forces a bound to
, but
the condition is
pr
p
1

pr
p
2
, and it belongs to N
, so the value is < N

<>

1
and we are done.
Clause (c):
By 5.2 it suces to prove
2
there are r,
such that:
is a P
-name, r P
, r = q , p
pr
q and r
P
lim(T
) and
N
[G
]
1
= N

1
and for every y V
_
<
N
[G
P
] for some
A V
_
<
N
we have [A[
V
<
[G
P
] and y A,
We shall choose by induction on n < ,
n
, q
n
,
n
, T
n
, k
n
, p
n
such that:
(a) (
n
, q
n
,
n
, T
n
) T
N
(so
n
is a simple

Q-named ordinal)
(b) k
n
is a P
n
-name of a natural number
(c)
n
is a P
n
-name
(d) q
n

P
n
T
n

k
n
Ord
(e)
0
= and
Q

n
<
n+1
< and
n+1
non-limt
i.e., if < and G
P
n
[G
P
] then
r q
n

n
[G
] < (i.e., is well dened < or is forced to be not well

dened),
(f) q
n+1

n
= q
n
(g) q
n+1

P
n+1
n+1
, so k
n
< k
n+1
and T
n+1
T
(h) p
n
is a P
n
-name, p
0
= p, p
n

pr
q
n
and
q
n

P
n
p
n
N
n
[G
P
n
] P
and p
n

n
G
P
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

66 SAHARON SHELAH
(i) q
n

P
n
p
n

P
pr
p
n+1
N
n+1
[G
P
n
] P
(j) letting
,
: < ) list the P
-names of ordinals from N
: for m, n we
have:
q
n

P
n
p
n+1
force that: a) if
n
m,
is a countable ordinal, m k
n
then it is smaller than some
n
m,
N
n+1
[G
P
n+1
],
a P
n
-name of a countable ordinal
b) for some A V N
n+1
k
n+1
,
[A[
V
<
and
n
m,
A.
The induction is straight (later we shall show that
_
n<
q
n
and
=
_
n<
n
are as
required in
2
) by clauses (a) + (b) proved above.

Because we need and have ()
1
or ()
2
+ ()
3
below:
()
1
Assume
pr
,
vpr
are equal to
(i.e.,
P
vpr
is
Q
for each < ), if p P
, < ,
a P
-name of
an ordinal then there are p
such that:
(i)
is a P
-name of an ordinal
(ii) p
pr
p
and p = p

(iii) p
.
[why? straight by 1.18].
[Saharon: maybe below we are stuck with
[, ), but this suces - need

to change?]
()
2
old proof of clause (b): if p P
, < ,
is a P
-name of a countable
ordinal, then there are p
such that
(i)
is a P

(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

(ii) p
pr
p
and p = p

(iii) p
,
[why ()
2
? let
be the following simple

Q-named [, )-ordinal:
for G
is generic over V for [, ) we let
[G
] = if
(a) p / G
or: for some p
we have p
= p and
P [= p
pr
p
and p
/G
<
for some countable ordinal
(b) for no [, ) does clause (a) hold for , G
.
Now if for some [, ) we have p
P
= we are done. Also
[G
] is well dened as if p G
and G
is generic over
V, then for some q G
and countable ordinal
we have q
.
By the denition of
1
Sp
e
(W)-iteration for some [, ) we have
[, ) [p
Q
pr
q or e = 4 & p not dened[?]].
Dene p
by: p
= p
, and for [
, ) we let p
be q if:
p
Q
pr
q or e = 4 & p not dened. We shall show that
p
is as required, hence really

P
[, ) is well dened. So there is

a P
-name of p
as appearing in the denition of
and it is, essentially, a

member of P
. Now as we have nite apure support, the proof of
[G
]
is well dened gives
P
is not a limit ordinal > . Lastly

P
is
not a successor ordinal > is proved by the property of each Q
.]
()
3
old proof of clause (a): if p P
, < ,
a P
-name of a set of <

ordinals, then there are p
such that:
(i)
is a P
-name of a set of < ordinals

(ii) p
pr
p
(iii) p
.
[Why? Similar to the proof of ()
2
; note that it is automatic if the
i
s
increase fast enough].
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

68 SAHARON SHELAH
Finishing the induction we let
=
_
n<
n
and we dene q
n
= q
n
,
q
[
_
n<
n
, ) is dened as
vpr
-upper bound of p
m
[
_
n<
n
, ) : m < ).
More formally, let
, and G
be such that: G
is generic over V,
=
_
n<
n
,
n
=
n
[G
], let p
n
= p
n
[G
n
], let p
n
[
, ) = r
: <
n
where
r
is a simple [
, )-named atomic condition.

Now we dene s
, a simple [
, )-named atomic condition as follows:

(a)
s
=
r
(b) if [
, ), G
, G
generic over V,
r
[G
] = then s
[G
] is
the <
V [G
]
-rst elements of Q
[G
] which satises the following:

()() (p
n
)
vpr
s
() if
Q
[G
]
)(p
m
)[G
] : m (n, )) has a
vpr
-upper bound
then s
[G
] is such upper bound.

Now actually such
vpr
-upper bound actually exists, and q
is as required.
5.7
5.9 Claim. Suppose W
1
is stationary and

Q = P
i
, Q
i
, I
i
,
i
: i < ) is a
UP
(W, W)-suitable iteration (where 4, 5), and P
= Sp
6
(W) Lim(
Q).
Each of the following is a sucient condition for
_
<
P
is a dense subset of
P
:
(A)
W is
8
strongly inaccessible and
<
density (P
) <
(B)
for every i,
Q
i
vpr
is equality and V [= cf() =
1
or at least for some
< we have
P
cf() =
1
.
Proof.
Case 1: (A)
Straight by the denition of Sp

6
(W)-iteration (see 1.15).
8
remember W is a parameter in the denition of Sp
e
(W)- iteration
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

Case 2: (B)
Follows by 5.7.
5.9
5.10 Conclusion. Assume W
1
is stationary and

Q = P
i
, Q
i
, I
i
,
i
: i < ) is a
UP
4
(W, W)-iteration with Sp
e
(W)-limit P
.
If < :

Q satises (A)
(B)
from 5.9 is a stationary subset of

and < P
satises the cf()-c.c. (e.g. has cardinality or at least density

< cf()), then P
satises the cf()-c.c.

Proof. Straight.
We may like to iterate up to e.g. the rst inaccessible (we may below weaken
[P
[ < to P
satises the -c.c. if P
=
_
<
P
).
5.11 Claim. [See 6.12]?? Assume
(a) W
1
is stationary and

Q = P
i
, Q
i
, I
i
,
i
: i < ) is a UP
4
(W, W)-
iteration (y a, b) with Sp
6
(W)-limit P
(b)
P
if in Q
,
Q
,= p
vpr
q, then p = q and
+
Q
is ,=
Q
but
,= p Q

+
Q
pr
p
) [?]
(c) S < : cf() =
1
is stationary; for a club E of , ES &
implies
P
(r Q
:
Q

vpr
r,
vpr
) is
+
-directed (question directed
above p, <
vpr
p
(d) / W is strongly inaccessible and: < P
< .
Then:
() forcing with P
does not collapse

() any function from any () < to ordinals in V
P
belongs to some V
P
.
Proof. Clearly clause () follows from clause (), so we shall prove just clause ().
If W is stationary, then by 5.10 we are done, so assume not and let E be a
club of disjoint to W, without loss of generality < E density(P
) < .
Suppose p P
and p
P
: () Ord is not in any V

P
for < where

(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

70 SAHARON SHELAH
() < .
We choose by induction on < the tuple (p
, q
) such that:
(a)
< is increasing continuous in and

+1
> Min(EB
)
(b) [?] p
is such that () <
, =
Q
or
Q
(c) for < , p
= p
(d) p
(e) q
) =
but there is no q
compatible with p
which
forces this and
< (), of course

(f) if cf() =
1
, then for some
<
,
[
) q

vpr
= q

(g) q
+1
and for every [
, ) we have
p
+1

P
if
Q
<
vpr
q
in

Q
then
p
+1
= q
;
if not [?] then r Q
such that
Q
<
vpr
r and
Q
if there is none
(see clause (b) in the assumptions of 5.11).
Having carried the denition, for some stationary W
S < : cf() =
1
and
< and
we have: W
&
< .
As [P
[ < = cf(), without loss of generality W
= q
. Now
choose < in W
then q
, q
are compatible and an easy contradiction to clause

(c) (with q
here playing the role of q
there).
5.11
Now we can rene 1.19 to the iteration theorem of this section.
5.12 Claim. 1) Suppose W
1
be stationary, F is a function, then for every
ordinal there is UP
4
(W)-iteration

Q = P
i
, Q
i
,
i
: i <
), such that:
(a) for every i we have Q
i
= F(
Q i) and
i
=
cc
(P
i
Q
i
)
(b)

(c) either
= or the following fails:

() F(
Q) is an (Sp
e
(W)Lim(
Q))-name of a forcing notion forced to sat-

isfy UP
4
(I, W) for some I -complete set of ideals, where is minimal
such that Sp
e
(W) Lim(
Q) satises the -c.c

(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

(d) Sp
e
(W) Lim(
Q) does not collapse

1
and preserve stationary subsets of
W (in fact it satises UP
1
(W).
Proof. Straight.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

72 SAHARON SHELAH
6 Preservation of UP
0
Here we present alternatives to 5, i.e. to UP
4
(W)-iterations. In UP
6
-iteration
P
j
, Q
i
,
j
: k , i < ) the demands are weak but the
i+j
may be large and
it is quite similar to semi-properness (but for fewer models). In UP
6
-iteration we
carry with us trees.
Recall [?]
6.1 Denition. 1) We say q is (N, , Q)-semi-generic if q is (N, Q)-semi-generic
and q if y N[G
] V then for some A N, [A[

V
< and y A.
2) Similarly with
instead of .
Remark. In 6.1(1) without loss of generality y N[G
] Ord.
6.2 Lemma. Suppose
(A) Q is a forcing notion
(B) I N is a family of ideals, I is -closed,
2
(and it is natural but not
needed to assume that I is -complete)
(C) N is -strictly (I, W)-suitable for (, x) as witnessed by
N = N
: (T, I)), Q N (so N (H (), , <
) is countable, N =
N
<>
and, of course, x codes Q, I, W)), see Denition 4.2)
(D) q is (N, , Q)-semi-generic
(E) at least one of the following holds:
() [MAC(Q)[ <
() q is (

N, , Q)-semi-generic, i.e. (N
, , Q)-semi-generic for every
T
() q T
T is a subtree and every T
for some ,
split(T
, I), I

RK
I
and T
[G
Q
]
1
= N

1
.
Then q [= N[G
Q
] is -strictly (I, W)-suitable for (, x, G
Q
)) (in fact, if
N
: (T, I)) was a witness for N then (N
[G
] : (T, I)) is a witness for

N[G
] being (I, W)-suitable for (, x)).

(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

Proof. Let G Q be generic over V such that q G.
In V
Q
, i.e., in V[G], clearly N
[G] (H()
V
Q
, , <
) (see [Sh:f, III,2.11,p.104],

N
[G] is countable (trivially) and N

k
[G] N
[G] for k g().

As q is (N, , Q)-semi-generic and N = N
, clearly q is (N
, Q)-semi-generic.
First assume [MAC(Q)[ < : now Q N hence MAC(Q) N hence [MAC(Q)[
N, and as N
cap <
clearly N
[MAC(Q)[ = N
[MAC(Q)[ hence also

N
MAC(Q) = N
MAC(Q) hence every Q-name
of an ordinal belongs
to N
(essentially). Hence q is (N
, , Q)-semi-generic, so q
Q
N
[G]
1
=
N

1
= N
1
, (even N
[G] <
[G]). So q is (N
, , Q)-semi-generic for
any T and even (lim T)
V
or (lim T)
V
Q
.
This almost shows that N
[G] : (T, I)) is a witness to N[G] being -strictly

(I, W)-suitable for (, x).
The missing point is clause (e)
of Denition 4.2, that is, that there may be

T, and J I N
[G] such that J / N
. So there is a Q-name
satisfying
J =
[G]; choose a
like that with min[Y [ : Y N
and
[G] Y minimal, and

let the set Y be J
i
: i < (without loss of generality Y I), so without loss of
generality J
i
: i < ) belongs to N
: by the minimality of [Y [ = [[ and q being

(N
, , Q)-semi-generic we have < .

So as J
i
: i < N
I and I is -closed there is J N
I such that
i<
J
i

RK
J hence the set : split(T) and J
RK
I
is a front of
T
[]
. So N
[G] : (T, I)) is a witness for N[G] is -strictly I-suitable for

(, , x, G)) (see Denition 4.2) so by 4.6 we know that N[G] is I-suitable.
Now if the second phrase of (E) holds, the proof is similar and if the third holds,
we can prove that without loss of generality the second holds.
6.2
The proof suggests some denitions, but we rst consider:
6.3 Denition. 1) For a forcing notion Q, family I of ideals and cardinal and Q-
names
of a cardinal and stationary W

1
, we say Q satises UP
6
(I, ,
, W)
or UP
,
(I, W) if:
() for every regular large enough, p Q and N a strictly (I
[]
, W)-suitable
model for satisfying p, Q, ,
N, there is q satisfying p
pr
q Q
such that q is (N, Q)-semi-generic and q
Q
N[G
Q
] is strictly I
[
]
-suitable
model for , .
1A) We say q is (N, I
, Q)-semi
6
generic if q is (N, Q)-semi genric and q N[G
Q
]
is strictly I
-suitable.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

74 SAHARON SHELAH
2) In part (1), if we omit , we mean = Max : I is -complete, so is regular
>
1
.
3) For Q, I, W, ,
as in part (1) and
a Q-name of a cardinal; we say that Q

satises UP
6
,
(I, W) if:
() if I
+
is a set of partial orders ideal extending I, large enough, p Q N
and N a strictly (I, W)-suitable model for , p, Q, ,
N, then there
is q satisfying p
pr
q Q such that q is (N, I
[
]
(I
+
I)
[
]
, Q)-semi
6
-generic
(see (1A) above).
4) We say Q satises UP
5
,
(I, W) if: for any (I

[]
, W)-suitable tree N
: (T, I))
of models and p N
<>
there are q, T
such that p
pr
q Q, q T
T is a subtree
and N
[G
P
] : T
) is a (I
[]
, W)-suitable tree of models and N
[G
Q
]
1
=
N
<>

1
. [Saharon but ?!]
> scite4.X undened
Some variants of this Denition are equivalent by the following claim.
6.4 Claim. 1) If is regular uncountable and Q satises the -c.c. then: q Q
is (N, , Q)-semi-generic i q Q is (N, Q)-semi-generic.
2) If N is (I, W)-suitable for (, x) (see Denition 4.3), I is -complete, is regular
and ( < )([[
0
< ), then there is a -strictly (I, W)-suitable N
for (, x)
such that N N
and N

1
= N
1
.
3) If q is (N, , Q)-semi generic, [MAC(Q)[ < , I N is -closed -complete and
N is I-suitable then q is (N, I
[]
, Q)-semi
6
generic.
4) If [MAC(Q)[ < and Q is UP
0
(I
[]
, W) then Q is UP
6
,
(I, W). Similarly for
Q-name such that for a dense set of p we have p
= and MAC(Q
p
) < .
Proof. Straight.
1) Reect.
2) Use the partition theorem 2.11.
3) By 6.2 using possibility (A).
4) By 6.2, too.
6.4
6.5 Claim. 1) If Q satises UP
6
,
(I, W) and W
1
W,
1
and
Q

1
,
then Q satises UP
6
1
,
1
(I, W
1
).
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

2) If Q
0
is a forcing notion satisfying UP
6
0
,
1
(I, W) and Q
1
is a Q-name of a
forcing notion satisfying UP
6
1
,
2
(I, W
1
), then Q
0
Q
1
is a forcing notion satisfying
UP
6
0
,
2
(I, W
1
).
3) If Q satises UP
6
,
(I, W
1
) and
Q

and I I
+
an I
+
I is -complete
then Q satises UP
6
,
(I
, W
1
).
6.6 Denition. 1) We say that

Q = P
j
, Q
i
,
j
: j and i < ) is a UP
6,e
(I, W, W)-
suitable iteration (with e = 6 if not mentioned explicitly) if:
(a) P
j
, Q
i
: j < , i < ) is an
1
Sp
e
(W)-iteration
(b) I is a set of partial order ideals such that I
[]
is -closed for any regular
,
2
[P
[
+
(c) W
1
is stationary
(d) for each i < we have:
i
is a P
i
-name of a regular cardinal >
1
in V
[purity?],
(e) (i) for i < j we have
P
j

(ii) if is a limit ordinal, then P
satises the local
-c.c. purely
[and (?) if
,= then some -complete ideal on

belongs to I]
(f)
P
i
Q
i
satises UP
6
i
,
i+1
(I
i
, W) and (Q
i
,
vpr
) is
1
-complete.
2) We say P
j
, Q
i
,
j
: j , i < ) is a UP
4
(I, W)-iteration if:
(a) (e) as above
(f) Q
i
satises UP
5
i
,
i+1
(I, W).
6.7 Denition. 1) We say that P
j
, Q
i
,
j
: j and i < ) is a weak UP
6,e
(I, W)-
iteration if (when e = 6 we may omit it):
(a) P
i
, Q
i
: i < ) is an
1
-SP
e
(W)-iteration (and, of course, P
= Sp
e
(W)-
Lim
Q))
(b) I is a set of partial order ideals
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

76 SAHARON SHELAH
(c) W
1
stationary
(d)
j
is a P
j
-name of a member of RCar
V
2
, increasing with j
(e) for i < j , i nonlimit we have
P
i
P
j
/P
i
satises the UP
i
,
j
(I, W)
(f)
P
i
(Q
i
,
vpr
) is
1
-complete.
2) We say

Q = P
j
, Q
i
,
j
: j , i < ) is a UP
5
(I, W)-iteration if:
(a) (d), (f) as above
(g) if i < j , i non limit we have
P
i
P
j
/G
P
i
satises UP
5
i
,
j
(I, W).
6.8 Claim. 1) If P
j
, Q
i
,
j
: j and i < ) is a UP
6
(I, W, W)-iteration, then
it is a weak UP
6
(I, W, W)-iteration, moreover, in clause (e) also limit i is O.K.
2) Assume

Q = P
j
, Q
i
,
j
: j , i < ) is an
1
Sp
e
(W)-iteration
(a) if is a limit ordinal and <

Q = P
j
, Q
i
,
j
: j , i < ) is
a weak UP
6
(I, W)-iteration and
= sup
j
: j < , then

Q is a weak
UP
6
(I, W)-iteration
(b) if = + 1,

Q is a weak UP
6
(I, W)-iteration and in V
P
, Q
satises
UP
6
+1
(I, W) then

Q is a weak UP
6
(I, W)-iteration.
[Saharon - compare with 6.9 and 1.27(2)]
Proof. Let and we need
,
Assume G
is generic over V,
i
=
i
[G
i
] and let N V[G
], N is
strictly I
[]
-suitable, N
1
W (so N (H(), ) is countable) and
p P
/G
and p P
N and

Q, , N. Then we can nd q satisfying
p
pr
q P
/G
and q
P
/G
N[G

]
1
= N
1
and N[G

] is I
[
]
-
suitable.
Without loss of generality [??] p forces a value to
moreover
(p, P
)].
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

Naturally, we prove this by induction on (for all ). The case = holds trivially
so assume < . If = 0, we have nothing to prove. If is a successor ordinal say
1
+1 so
1
, now we use rst
,
1
and then the demand on Q
1
in denition
6.7, in clause (f).
So from now on we shall assume that is a limit ordinal. As in the proof of 5.7 we
can note
Fact A: If
2
are simple

Q-named [0, )-ordinals then
1
,
2
holds.
Proof of the fact: Here we use e = 6 rather than e = 4. On P
see Denition
1.15(F)(g). We prove it by induction on the depth of
2
, see Denition 1.7(5).
So we are given G
2
P
1
generic over V and in particular let
1
=
2
[G
1
], (so
it is simpler to say that G
1
P
1
is generic over V,
1
[G
1
] =
1
). Let
1
=
1
[G
1
] and we are also given N which is strictly I
[]
-suitable, p P
2
/G
1
, p
N,
Q,
1
, G
1
N,
2
N. We have to nd q P
2
/G
1
such that p
pr
q, q is
(N, P
2
/P
1
)-generic and q N[G
P
2
/G
1
] is strictly I
[
2
]
-suitable.
If the depth of
2
is 0, then
2
=
2
and we can use
1
,
2
. So assume the
depth of
2
is > 0, and so for some
and a sequence
2,
: <
) of simple
Q-named [Max
, , )-ordinals and P
-name
we have
Q

2
=
2,
. So
without loss of generality
2,
: <
),
N. Let
1
= Max,
1
,
Q
1

2
.
Now clearly
1
is a simple

Q-named [0, )-ordinal,
Q

2
and
1
,
1
&
1
,
1
,
2
and
1
,
2
easily holds (by the cases proved above) so it
is enough to prove
1
,
2
. This just means that without loss of generality
1

and even
1
=
. Now
[G
1
] N so we use the induction hypothesis to get the
desired q.
fact
Fact B: If
is a simple

Q-named [0, )-ordinal, p P
and
is a P
-name of a
countable ordinal, then there are
and q such that:

()(i) P
[= p
pr
q
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

78 SAHARON SHELAH
(ii) q
= p
(iii)
is a P

(iv) q
P
<
.
Proof. Let
n
: n < ) be a witness for p, so each
n
is a simple

Q-named
[0, )-ordinal. For each n we dene a P
n
-name of t
n
of a truth value and r
n
of
a member of P
, r
n
= r
n
[
n
, ), as follows: if G
n
P
n
(
n
[G
n
]
[G
n
]), and there are q P
/G
n
and <
1
, P
p
pr
q and
q
P
/G
n
< , then t
n
[G
n
] = truth and then r
n
[G
n
] is q [
n
, ) for some such
q otherwise t
n
[G
n
] = false, r
[G
n
] = . Let r
n
be the following P
n
-name:
for G
n
P
n
if:
(a) t
n
[G
n
] = truth and
(b) for no m < , for some r G
n
forces a value to
m
, say
m
,
m
<
n
[G
n
]
(
m
=
n
[G
n
] & m < n)
then
n
[G
n
] is r
n
, otherwise
n
[G
n
] = . Let p
= p
_
n
r
n
, easily q P
and
p
pr
p
.
Let
n
= Min
+ 1,
0
+ 1, . . . ,
n1
+ 1,
n
is a simple

Q-named [0, )-ordinal,
0
=
n
<
n+1
,
n+1
. Let N be a strictly (I, W)-suitable model for such
that
Q, p,
n
,
n
: n < ), r
n
, r
n
: n < belongs to N, it exists by 4.5. Now we
choose q
n
by induction on n < such that
q
n
P
n
is (N, I
[
n
]
, P
n
) semi
6
generic
p

n
,
pr
q
n
q
n+1

n
= q
n
.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

This is possible by Fact A. Now (see 1.26) q
=:
_
n<
q
n
p
belongs to P
, p
pr
q
.
It is enough to show that q
N
1
, assume not so there is r, q
r P
such that r forces a value

1
(N
1
) to
.
By 1.17(1) without loss of generality q
r above
1
, . . . ,
m
for some m <
and
< . There are r
, k,
k
such that: r r
,
k
< , r
[
k
, ) = r [
k
, ), r
forces
k
is
k
and for each = 1, 2, . . . , m we have
<
k
or
(r
)( < )(k
)[r r
& r
k
= &
< ].
Clearly r
forces that (n)[
n

k
t
n
is truth] and we easily nish.
Fact B
.
Now we do not just have to nd q satisfying P
/G
[= p
pr
q and q is
(N, P
/G
)-semi-generic, but we need more in the (N, I

[
]
P
/G
)-semi
6
generic.
Now for G
generic over V, in V[G
] for every countable elementary sub-

model M of (H ()
V
, G
, <
),
Q, , , I) M, we have (in 4.8) dened Dp(M),

an ordinal or and I
M
such that
(A) Dp(M) = i M[G
] =
[G
] :
M a P
-name includes M, has

the same countable ordinals, is (H ()
V[G
]
, ) and M[G
] is strictly
I
[
[G
]]
-suitable
(B) if Dp(M) = < then I
M
is a member of IM which is
[G
]-complete,
and
Y
M
= t Dom(I) :there is N as above, M N, M
1
= N
1
and t N and Dp(N) = mod I.
So we have P
-names Dp
, I
M
.
Consider
K =
_
(, G
, N) : < , nonlimit, G
generic over V, in V[G
], N[G
] is
I
[
[G
]]
-suitable
_
.
We now dene by induction (
n
, q
n
, p
n
, N
n
) such that
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

80 SAHARON SHELAH
(a)
n
is a simple

Q-named [, )-ordinal, (as e = 6, it is full)
(b) N
n
is a P
n
-name, q
n
P
n
, p
n
is a P
m
-name of a member of
N
n
[G
n
] (P
/G

n
)
(c)
0
= , N
0
= N
(d) if G
n
P
n
is generic over V, q
n
G
n
, G
G
n
,
n
[G
n
] =
n
(so essentially
G
n
is just a generic subset of G
n
over V such that
n
[G
n
] =
n
), then
N
n
[G
n
] is a countable elementary submodel of (H ()
V[G
n
]
, ) to which
Q, , , I belongs and is strictly I

[
n
[G
n
]]
-suitable
(e) N
n
N
n+1
, N
1
= N
n+1
1
, q
n+1

n
= q
n
, p
n

n
q
n
, p
n

pr
p
n+1
(f) for G
n
,
n
, N
n
as in (d) and I IN
n
there is k > n such that: if G
k
,
k
, N
k
are as in (d), G
n
G
k
, then there is t Dom(I) N
k
Y
(N
n
[G
n
])[G
]
(i.e.
forced to be there)
(recall p forces a value to
) sn
(g) for G
n
n
,
n
, N
n
as in clause (d) above and
N
n
a P
-name of a countable
ordinal there is k > n such that: if G
k
,
k
, N
k
are as in clause (d), G
n
G
k
then p
k
forces (
P
/G
k) to be < N
1
.
No problem to carry the denition. Now
()
a
q =
_
n<
q
n
P
Here we use the q

n+1

n
= q
n
below
_
n<
n
and (Q
i
,
vpr
) is
1
-
complete above
_
n<
n
()
b
for G
n
, N
n
,
n
as in clause (d), q is (N
n
, P
/G
n
)-semi generic and above
p
n
[why? note clause (g)]
()
c
if q G
, G
is generic over V extending G
then in V[G
], Dp(N
n
[G
])
is well dened
()
d
Dp((N
0
[G
])) is
(use clause (B) in the demands Dp and clause (f) above). We use I
[]
is
-closed for the relevant s.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

So we are done.
6.8
6.9 Claim. 1) Assume that

Q = P
j
, Q
i
,
j
: j , i < ) satises:
() is a limit ordinal
() if < then P
j
, Q
i
,
j
: j , i < ) is a weak UP
6
(I, W, W)-iteration
() P
is the
1
Sp
6
(W)-limit of P
i
, Q
i
: i < )
()
, a P
-name is sup
i
: i < .
Then

Q is a weak UP
6
(I, W, W)-iteration.
2) Assume that
() P
j
, Q
i
,
i
: j , i < ) is a weak UP
6
(I, W, W)-iteration
() in V
P
, Q is a forcing notion satisfying UP
6
I,
, W), where
is the
interpretation of
) and let Q
be P
-names of those objects.

Then there is a UP
6
(I, W, W)-iteration P
i
, Q
j
,
i
: i + 1, j < ) with Q
=
Q
+1
=
.
Proof. 1) By the proof of 6.8.
2) Straightforward.
6.10 Claim. Assume
(a)

Q = P
j
, Q
i
,
j
: j
, i < ) is a weak UP
6
(I, W, W)-iteration
(b) <
(c) G
is generic over V
(d) N is a strictly (I, W)-suitable model N for (,
Q, , )) in V[G
]
(e) p N (P
/G
).
Then there is q such that:
() p
pr
q P
/G
() p = q
() q is (N, P
/G
)-semi generic
() q has a witness listing
N :
a simple

Q-named [, )-ordinal.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

82 SAHARON SHELAH
Proof. Like 6.8, forgetting to take care of the I IN
n
so we can use N
= N[G
n
].
6.10
6.11 Claim. Suppose F
f
, F
c
are functions (possibly classes), W
1
is station-
ary, I is a class of (
2
-complete) quasi order ideals, W a class of strongly inacces-
sible cardinals.
Then for every ordinal there is a unique

Q = P
j
, Q
i
,
j
: j
, i < ) such
that:
(a)

Q is a UP
6
(I, W, W (
+ 1))-iteration
(b) for every i <
we have Q
i
= F
f
(
Q i),
i
= F
c
(
Q i)
(c)
c
(d) for limit
we have
= sup
: <
(e) if
< then the following is impossible

() F
f
(
Q i) is a P
-name of a forcing notion

() F
c
(
Q i) is a P
-name of a F
f
(
Q i)-name of a V-cardinal
1
()
P
F
f
(
Q i) is UP
6
(I, W).
Proof. Straight.
Next we give sucient conditions for
_
i<
P
i
being a dense subset of P
j
6.12 Claim. Assume that P
j
, Q
i
,
j
: j
, i <
) is a UP
6
(I, W, W)-
iteration.
1) Assume
()
1
(i <
)[
P
i

Q
i
vpr
is equality].
If cf() =
1
&
then
_
i<
P
i
is a dense subset of P
even under
pr
.
2) Assume
()
2
(i <
)(
P
i
(r Q
i
:
Q
i

vpr
r,
Q
i
vpr
) is directed).
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

If
, cf() =
1
then
(a)
_
i<
D
,i
is a dense subset of P
even under
pr
where D
,i
= p P
: P
[=
p i
vpr
p
(b) if i < , p
0
, p
1
D
,i
and p
0
i = p
j
i then p
0
, p
1
has an upper bound p
even in D
,i
, p i = p
0
i = p
1
i.
3) If W &
& (i < ) [density(P

i
) < ] then P
=
_
i<
P
i
.
Proof. 1), 2) Let p P
, choose large enough. There is a strictly (I, W)-suitable

countable model N (H (), ) to which
Q, , p belongs. Applying 6.8 for =

0, = (i.e.
,
from the proof) we can nd q P
, p
pr
q which is (N, I
[
]
, P
)-
semi
6
generic and q has a witness
N :
a simple

Q-named [0, )-ordinal.
As cf() =
1
there is an increasing continuous

=
: <
1
) with limit ,
without loss of generality

N so q N[G
] =
: N[G
]
1
but q
: N[G
]
1
=
: N
1
=
N
1
, so clearly
P
[= q
N
1

vpr
. For (1) it follows that q P
N
and we are done.
For (2) just reect.
3) Straight.
6.13 Conclusion. Let

Q = P
j
, Q
i
,
j
: j
, i <
) be a UP
6
(I, W, W)-iteration
and = cf()
and (i < ) (density(P

i
) < ).
1) If < : = cf() W is stationary. Then P
satises the -c.c. (in fact a

strong version and even under
P
vpr
.
2) If ()
1
from 6.12(1), i.e.
P
i

Q
i
pr
is equality for i < then P
satises the
-c.c. (in fact a strong version) even for
P
vpr
)(
vpr
) is (2
||
)
+
-complete.
3) Assume / W and S < : cf() =
1
is stationary and i &
S
P
i
(r :
Q
i

vpr
r. Then forcing with P
does not add a function in

Ord
not in
_
<
V
P
, even any function in

()
Ord
_
<
V
P
, () < .
Proof. 1) Let S = < : = cf() W and let P
: S) be a sequence of
members of P
. So for each S for some i() < we have p
P
i()
and we can
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

84 SAHARON SHELAH
nd a pressing down function h on S such that h(
1
) = h(
2
) p
1

1
= p
2

2
.
Clearly there is a club E or such that
i

2
s &
2
S S E p
1
P
2
.
Lastly, if
1
,
2
E S and f(
1
) = f(
2
) then p
1
p
2
is a common upper
bound of p
1
, p
2
(even a
vpr
one).
2) Similar using S = < : cf() =
1
and < density(P
) < .
3) If W is stationary in use part (2), so let E be a club of disjoint to W.
Assume toward contradiction that p
P
the function
:
1

Ord is not in
_
<
V
P
, let S be as in part (2). We choose by induction on j < , (p

j
,
j
) and if
j
Salso(q
j
,
j
,
j
,
j
) such that:
(i)
j
E is increasing continuous
(ii) p
j
P
, p
0
= p
(iii) i < j p
j

i
= p
i
(iv) i < j & [
i
,
j
)
P

vpr
p
j
in

Q
(v) if
i
S, j <
i
then for every p
i
q D
i
,j
there is
i
(q) <
i
such that:
if there are , r such that p
i+1

vpr
r, r
i
= q, <
1
, r forcing a value
to
() but for no r, q r
i
, does P
i+1
r
forces a value to
() then
=
i
(q) satises this
(vi)
i
S then p
i+1
q
i
, q
i

(
i
) =
i
,
i
< (),
i
<
i
, q
i

i

vpr
q
i

i
(vii) there is no q, q
i

i
q P
i
such that p
i+1
(q
i

i
) forces a value to
(
i
).
For any j we choose (p
j
,
j
).
For j = 0 let p
0
= p and by 6.9 for some
0
< , p
0
P
0
. For j = i + 1, rst
choose p
j
to satisfy clause (v) and then
j
such that p
j
, q P
j
. Lastly for j limit
let p
j
=
_
i<j
p
i
,
j
=
_
i<j
i
and check. The contradiction is easy.
Let G
i
P
i
be generic over V such that p
i
G
i
. Clearly for some
< () for no q P
i
do we have q p
i+1
forces a value to
() as other-
wise p
i+1

P
/G
V[G
i
]. Choose
i
< () as above, choose q
i
G
i
which forces this choose q
i
P
above q
i
p
i+1
which forces a value to
(
i
) and
without loss of generality there is
i
<
i
such that q
i

i

vpr
q
i

i
. Lastly let
i+1
be such that [
i+1
, )
P

vpr
p
i+1
.
[Saharon: the role of W].
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

6B On UP
2
-iteration
6.1(?)Lemma. Assume that W
1
is stationary and

Q = P
i
, Q
i
, I
i
,
i
: i < )
is a UP
2,e
(W, W)-suitable iteration, and P
= Sp
e
(W) Lim
Q) be the limit and
() = sup
: < (this is a P
-name of a V -cardinal) and
() = Sup :
P
,=
() for some < .

1) For each , P
satises UP
0,e
()
(I
, W) for some
2
-complete I
V.
2) In fact, I
is
-complete where
= min : for some < we have
P
+1

,= , and each I I
has domain of cardinality

(sup
<
<
+1
:
P
(I I
)( = [Dom(I)[)) and
[I
<
(
0
+[P
[ + min :
P
[I
[ )
<
.
3) Similarly for UP
0,e
a
and weak UP
0,e
y
(W)-iterations.
6.2 Remark. We can also get the preservation version of this Lemma.
Proof. For each < let J
=: q P
+1
, called
,q
and q forces I
to be equal to a P
-name I
,q
and q forces a value to [I
[ says
,q
is this is purely decidable, if not, just an upper bound to it; let J
a
maximal antichain. Let
= sup
qJ
,q
.
Let q
P
= I
,
: <
,q
for q J
and let J
,
= q J
:
,q
>
and q Dom(I
,
) is
,q,
and let I
,
be id
L
,
, so L
,q,
is a P
-name of a
,q
-directed partial order on
,q,
(but
P
if [I
[ <
then let L
,
be
trivial).
For q J
let L
,q,
be ap
,q
(L
,
) for the forcing notions
P
[q]
= p P
: q
P
pr
p from Denition 3.9. So by Claim 3.10
(i) L
,q,
is
,q
,q,
]
<
,q
(ii) [L
,q,
[ (
,q,
)
<
,q
(iii) q
P
,
= id
L
,

RK
id
L
,q,
.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

86 SAHARON SHELAH
Let
= sup
,q
: < , q J
.
Let I
be the (<
)-closure of id
L
,q,
: < , q J
, <
,q
(see Denition
3.13(1)).
Let N be (I
, W)-suitable model for (, ), x code enough informtion so for some
N, N = N
and

N = N
: (T, I)) be a strict (I
, W)-suitable tree of models

for (, x), x coding enough information (so

Q, I
, S
, W N
).
Let T
N
be the set of pairs (, p) such that:
, p P
, and for some ,

p
P
N[G
]
1
= N

1
and N[G
P
].
T
N
is dened similarly as the set of pairs (
, p) such that:
is a simple (
Q, W)-
named ordinal, p P
. (I.e. if < , G
P
n
[G
P
]
then r q
n

n
[G
] < , i.e. is well dened < or is forced (

P
/G
) to be not
well dened, and p
P
lim T).
We consider the statements, for < (or restrict ourselves to non-limit)
,
for any (, p) T
N
and such that
p
P
and p
a P
-name such that p

P
[G
P
] N[G
] P
/G
P
and
(p
[G
P
])
pr
p there is (, q) T such that p
pr
q
(i.e. p
P
[G
]
pr
q) and q = p.
We prove by induction on that ( )
,
(but for 6.1(3), we use (
non-limit )
,
), note that for = the statement is trivial hence we shall
consider only < .
Case 1: = 0.
Trivial.
Case 2: a successor ordinal (for part (3), successor of non-limit ordinal).
As trivially
0
,
1
&
1
,
2

0
,
2
= + 1.
Let G
P
be such that p G
P
and G
P
generic over V .
Let

N
= N
[G
P
] : (T
, I)).
In V[G
P
] we apply ? for =
2
to

N
and nd T
T such that V[G

P
] [=
> scite6.2 undened
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

N
[G
P
] : (T
, I
)) is strict (I
)
[()]
, W)-suitable. So we can apply clause
(f) of ?.
> scite6.4A undened
Discussion: This is a question whether there is an I-tree of model N
: (T, I))
such that:
if I
is (I
)-complete, (I
) regular, < (I
) [[
0
< (I
), then
Suc
T
() N
<
.
This would make 6.2? more eective.
We shall choose by induction on n < ,
n
, q
n
, p
n
such that:
(a) (
n
, q
n
) T
N
(so
n
is a simple (
Q, W)-named ordinal)
(b)
0
= and
Q

n
<
n+1
<
i.e. if < and G
P
n
[G
P
] then
r q
n

n
[G
] < (i.e. is well dened < or is forced to be not well

dened),
(c) q
n+1

n
= q
n
(d) p
n
is a P
n
-name, p
0
= p, p
n

pr
q
n
and
q
n

P
n
p
n
N
n
[G
P
n
] P
and p
n

n
G
P
(e) q
n

P
n
p
n

P
pr
p
n+1
N[G
P
n
] P
(f) letting
: < ) list the P
-names of ordinals from N: for m, n we

have:
q
n

P
n
p
n+1
force (
P
n+1
) that: if
is a countable ordinal, then it is smaller

than some
N[G
P
n+1
],
a P
n
-name of a countable ordinal.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

88 SAHARON SHELAH
The induction is straight and
_
n<
q
n
are as required noting we need and have ()
1
or ()
2
below:
()
1
Assume
pr
,
vpr
are equal to
(i.e.,
P
vpr
is
Q
for each < ), if p P
, < ,
a P
-name of
an ordinal then there are p
such that:
(i)
is a P
-name of an ordinal
(ii) p
pr
p
and p = p

(iii) p
.
[why? straight by 1.18].
()
2
if p P
, < ,
is a P
-name of a countable ordinal, then there are p
such that
(i)
is a P

(ii) p
pr
p
and p
= p
(iii) p
[why ()
2
? let
be the following simple (
Q, W)-named [, )-ordinal:
G
is generic over V for [, ) we let
[G
] = if
(a) p / G
or:
for some p
we have p
= p and P
[= p
pr
p
and p
/G
<
for some countable ordinal
(b) for no [, ) does clause (a) hold for , G
.
Now if p
P
= we are done. Also

P
[G
] is well dened
as if p G
and G
is generic over V, then for some q G
and
countable ordinal
we have q
. By the denition of
1
Sp
e
(W)-
iteration for some [, ) we have [, ) [p
Q
pr
q or
e = 4 & p not dened].
Dene p
by: p
= p
, and for [, ) we let p
be q if:
p
Q
pr
q or e = 4 & p not dened. Now p
is as required.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

So there is a P
-name of p
as appearing in the denition of
and it is,
essentially, a member of P
. Now as we have nite apure support, the proof

of
[G
] is well dened gives

P
is not a limit ordinal > . Lastly
is not a successor ordinal > is proved by the property of each

Q
.]
Finishing the induction we dene q
n
= q
n
, q
[
_
n<
n
, ) is dened as
vpr
-
upper bound of p
m
[
_
n<
n
, ) : m < ).
More formally, let
, and G
be such that: G
is generic over V,
=
_
n<
n
,
n
=
n
[G
], let p
n
= p
n
[G
n
], let p
n
[
, ) = r
: <
n
where
r
is a simple [
, )-named atomic condition.

Now we dene s
, a simple (
Q, W)-named [
, )-ordinal atomic condition as fol-

lows:
(a)
s
=
r
(b) if [
, ), G
, G
generic over V,
r
[G
] = then s
[G
]
is the <
V[G
]
-rst elements of Q
[G
] which satises the following:

()(a) (p
n
)
(b) if
Q
[G
]
)(p
m
)[G
] : m (n, )) has a
vpr
-upper bound
then s
[G
] is such upper bound.

Now actually such
vpr
-upper bound actually exists, and q
is as required.
6.1
Now we can rene 1.19 to our iteration theorem.
Saharon revise.
6.3 Claim. 1) Suppose F is a function, W
1
stationary, I a class of quasi
order ideals, then for every ordinal there is UP
6
(I, W)-iteration

Q = P
j
, Q
i
,
j
:
j
, i <
), such that:
(a) for every i <
we have Q
i
= F(

Q i),
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

90 SAHARON SHELAH
(b)

(c) for
is as in clause (d) of Denition ?

> scite6.3 undened
(d) either
= or F(
Q) is not a pair (Q
) such that:
is a P
-name
of a cardinal from V and
P
satises UP
6
(I, W).
2) Suppose < , G
is generic over V, then in V[G
],

Q/G
= P
i
/G
:
Q
i
,
i
: i < ) is an UP
6
(I
[
[G
]]
, W)-iteration.
3) If

Q is an UP
6
(I, W)-iteration, p Sp
e
(W) Lim(
Q), P
i
= q P
i
: q p
i, Q
i
= p Q
i
: p p i, then

Q = P
i
, Q
i
: i < g(
Q)) is (essentially) an
UP
6
(I, W)-iteration.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

7 No New Reals, Replacements for Completeness
Now we turn to No New Reals, there are versions corresponding to [Sh:f, Ch.V,1-
3] (W-complete), [Sh:f, Ch.V,5-7] (
<
1
-proper +D-completeness) and better
[Sh:f, Ch.VIII, 4] (making the previous preserved) and in dierent directions [Sh:f,
Ch.XVIII,2] and [Sh 656].
We deal here with the rst (here we are interested in the cases
pr
=)
7.1 Denition. For p Q let Q
pr
p
= q Q : p
pr
q. A point which may
confuse is that the pure extension notion used in Denition 7.2, is not necessarily
the one used seriously in the iteration. This is the reason for the case e = 5 in 1.
[Saharon check: main question: do we really need the purity in the iteration for
Nm
. For Nm it is not needed (as in [Sh:f, Ch.XI].]

7.2 Denition. 1) UP
4
com
(I, W) is satised by the forcing notion Q, i: for any
N
: (T, I)) a strict (I, W)-suitable tree of models for (, x), x coding enough
information, we have ()
1
()
2
where:
()
1
for every , T, of the same length we have (N
, Q)

= (N
, Q) and
letting h
,
be the isomorphism from N
onto N
we have h
,
(x) = x and
< g() h
,
h
,
; (for , lim(T) let h
,
=
_
<
h
,
)
()
2
if
lim(T), p N
G and G
is a
pr
-directed subset of N
Q
pr
p
=
_
<
N
Q
pr
, not disjoint to any dense subset of Q
pr
p

_
m<
N
m
denable
in (
_
m<
N
m
, N
m
, I
m
)
m<
, then there is q Q such that p
pr
q
and q
Q
there is lim(T) (in V
Q
) such that
_
<
h
,
(G N
) is
a subset of G
Q
.
2) UP
4
stc
(I, W) is satised by the forcing notion Q i for any

N = N
: (T, I))
a strict (I, W)-suitable tree of models for (, x), x coding enough information we
have ()
1
()
2
where
()
1
for every , T, of the same length we have (N
, Q)

= (N
, Q) and
letting h
,
be the isomorphism from N
onto N
we have h
,
(x) = x and
< g() h
,
h
,
; (for , lim(T) let h
,
=
_
<
h
,
)
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

92 SAHARON SHELAH
()
2
if
lim(T); N
<>
G then in the following game =
Q,p,

N,
; player
complete has a winning strategy. The plays last -moves, in the n-th move
(we can incorporate the last into the game) where t
n
= T x(Suc(
n)
I
+
) a condition p
n
P N
is chosen such that p p

0
, [n > 0 p
n1

p
n
]. In the n-th move, the anti-completeness player chooses q
n
Q N
such that n = 0 p = q
n
and n > 0 p
n1

pr
q
n
and the completeness
player chooses p
n
Q N such that q
n
p
n
Q N
. In the end the

completeness player wins i one of the following occurs:
() for some Q-name
of a countable ordinal, no p
n
forces a value
() not () but there is q such that: p
pr
q Q and q
Q
there is
lim(T) such that n < h
n,
n
(p
n
) G
Q
.
3) We dene Q satises UP
1
com,
(I, W) is dened as in (1) but we replace the

conclusion of ()
2
by: there is q Q such that p
pr
q and
q
Q
there is (T
, I) satisfying (T, I)
(T
, I)
(see 2.4(d)) such that for every
lim(T
) we have h
,
(G
) G
Q
.
4) Similarly UP
1
stc,
(I, W).
5) We say UP
4
stc
(I, W) if letting p = (
n
, I
n
,
(n) : I
n
, (t
n
), n < ) be such
that
n
: n < list the Q-names of ordinals in N
, I
n
: n < lists IN
, the
winning strategy on each stage depends just on , N
<>
1
and in p continuously.
7.3 Remark. 1) This property relates to the UP(I, W) just as E-complete relates
to E-proper (see [Sh:f, Ch.V,1]).
2) Who satises this condition? See section 8, so W-complete forcing notions,
Nm
(D),Nm(D)(D is
2
-complete) Nm
()
(T, I) (when I is
2
-complete), and shoot-
ing a club through a stationary subset of some = cf() >
1
consisting of ordinals
of conality
0
(and generally those satisfying the I-condition from [Sh:f, Ch.XI]).
7.4 Claim. If Q satises UP
1
com
(I, W) or UP
1
stc
(I, W) and I is (2
0
)
+
-complete,
and Q has (
1
, 2)-pure decidability, then forcing by Q add no new real.
Proof. Immediate.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

7.5 Claim. Suppose:
(a) Q is a forcing notion satisfying the UP
1
com
(I, W) and the local
-c.c. where
is a purely decidable Q-name

(b)

N = N
: (T, I)) is a strict (I, W)-suitable tree of models (for and

x = Q,
, I, W))) satisfying ()
1
of Denition 7.2
(c) the family I
=: I I : I is
-complete is (<
)-closed.
Then Q satises UP
1
stc,
(I, W).
Proof. Let (T
, I
) and

N = N
: (T
, I
)), h
,
: , T
n
Ord for some n)
be as in Denition 7.2.
Let
lim(T
), p QN
<>
be given and we choose as our strategy for proving
UP
1
stc
(I, W) the same strategy that exists as UP
1
stc
(I, W) and let p
n
: n < ) be a
play as in Denition 7.2 in which the completeness player uses his winning strategy.
Let T = T : (T
, I
(T, I
T). As we can replace p by p
if p
pr
p
QN
<>
, without loss of generality p forces a value to . So for every T T there
are q and
such that
p
pr
q Q and q
Q

lim T
and h =
_
h
n,
n
satises n < h() G
Q
(hence N
n

1
= N
<>

1
).
Remember that we can replace
by any
lim(T). Let
T
[G
Q
] = T
:G
Q
N
h
,
g()
(r),
r Q N
(g())
and r p
n
for some n
clearly it is a subtree. We continue as in the proof of 5.2.
7.5
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

94 SAHARON SHELAH
7.6 Claim. Suppose:
(a)

Q = P
i
, Q
i
, I
i
,
i
: i < ) is a UP
1,e
(W, W)-iteration with
Sp
e
(W)-limit P
(b)
P
i
Q
i
satises UP
1
stc
(I
i
, W)
(c)
() = Min
: < ,
+
() = Sup : for some < ,
P
,= ,
() = Min : for some < we have

P
,= .
Then 1) for each , P
satises UP
1
com
(I
, W) for some
()-complete set
I
of (partial order) ideals.

2) In fact, I
is
-complete where
= min : for some < we have

P
+1

,= , and each I I
has domain of cardinality (sup

<
<
+1
:
P
(I I
)( = [Dom(I)[))
and [I
<
(
0
+[P
[ + min( :
P
[I
[ )
<
.
7.7 Remark. We can use this for iteration as in 5.11, the version with clauses (b),
(d) or (d)
, W = . To prove P
does not add reals, it is enough to prove that

for each < , forcing with P
does not add reals. By p P
: ( < )
P
< p is
vpr
-dense. This should be useful in [\GoSh:511 ].
SAHARON: 1) Use less
.
2) What requirements will resurrect
vpr
?
Proof. Similar to the one of 5.7.
For each < let J
=: q P
+1
, called
,q
and q
forces I
to be equal to a P
-name I
,q
and q forces a value to [I
[ says
,q
;
let J
be a maximal antichain. Let
= sup
qJ
,q
.
Let q
P
= I
,
: <
,q
for q J
and let J
,
= q J
:
,q
>
and q Dom(I
,
) is
,q,
if this is purely decidable and let I
,
be id
L
,
,
so L
,q,
is a P
-name of a
,q
-directed partial order on
,q,
(but
P
if [I
[
<
then let L
,
be trivial).
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

For q J
let L
,q,
be ap
,q
(L
,
) for the forcing notions
P
[q]
= p P
: q
P
pr
p from Denition 3.9. So by Claim 3.10
(i) L
,q,
is
,q
,q,
]
<
,q
(ii) [L
,q,
[ (
,q,
)
<
,q
(iii) q
P
,
= id
L
,

RK
id
L
,q,
.
Note:
() = Min
,q
: < , q J
.
Let I
be the (<
())-closure of id
L
,q,
: < , q J
, <
,q
(see
Denition 3.13(1)).
Let

N = N
: (T
, I)) be a strict (I
, W)-suitable tree of models for (, x), x

coding enough information (so

Q, I
, W, W N
). For any < and G
generic over V, T T
and I
[
()]
-tree and T and
lim(T) and p
N
[G
] (P
/G
) then let

N
,T
[G] = N
[G] : T) then we can nd a win-

ning strategy St for the completeness player in the game =
N
,T
[G], p,
, P
/G
of 7.2(2). Without loss of generality if
1
,
2
lim(T) the isomorphism from
N
1
[G] onto N
2
[G] commutes with the winning strategies; so the choice of
is
not important. Of course, we have a name St
= St
,T
.
Now x
lim(T
) and we dene a strategy St for the game. For each simple

(
Q, W)-name of an [0, )-ordinal
, let (
n
, I
n
,
(n), I
n
, t
m
) : n < )
be as in 7.2(2) for Q
.
We dene St such that if p = p
n
: n < ) is a play in which the completeness
player uses his winning strategy then this holds for p
n
(
) : n < ) for each , i.e.,

() if < , G
is generic over V and p G and
[G
] = and
T V[G
] is a subtree of T
, I
[
()]
-large and
lim(T)
V [G
]
and
p
= (h
,
(p
n
))()[G
] Q
[G
], then in p
n
: n < ) the completeness
player uses his winning strategy from above.
So x such p
= p
n
: n < ), we would like to nd q as in Denition 7.2(2).
Let T
N
be the set of quadruples (, q,
, T
) such that:
1
, q P
, P
[= p
pr
q and
q
P
()
, where
, T
are P
-names
() N
[G
]
1
= N

1
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

96 SAHARON SHELAH
()
_
<
N
[G
P
]
() N
,
[G
] : T
is a strictly (I
)
[
()]
-suitable tree,
() for every
lim(T
) we have
h
,
(p
n
) : n < is a subset of G
.
Now T
N
is dened similarly as the set of quadruples (
, q,
, T
) such that: as in
1
but we have
is a simple (
Q, W)-named ordinal, q P
and in clause ()

[G
P
]. (I.e., if < , G
P
n
[G
P
] then
r q
n

n
[G
] < , i.e., is well dened < or is forced (

P
/G
) to be not well
dened, and p
P
lim(T)).
We consider the statements, for <
,
for any (, p,
, T
) T
N
and
such that
p
P
and p
a P
-name such that

p
P
[G
] N
[G
] P
/G
and
(p
[G
P
]) p there is (, q,
, T
) T such that
p
q (i.e., p
P
[G
] q) and q = p and T
T .
We prove by induction on that ( )
,
(or, for strong preservation),
that ( non-limit )
,
), note that for = the statement is trivial hence
we shall consider only < .
Case 1: = 0.
Trivial.
Case 2: a successor ordinal.
As trivially
0
,
1
&
1
,
2

0
,
2
= + 1.
Let G
P
be such that p G
P
and G
P
generic over V .
Let T
= : T
[G
P
],

N
= N
: (T
, I
)) where N
= N
[G
P
],
I
= I
.
By 7.5 applied to

N
we can nd p
, T
as required.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

By 7.5 it suces to prove
2
there are q and
such that:
is a P
-name,
q P
, q = p , p q and q
lim(T
) and
_
<
N
[G
]
1
= N
1
and h
,
(r) : r G
.
We should choose by induction on n < ,
n
, q
n
,
n
,
n
, k
n
such that:
(a) (
n
, q
n
,
n
) T
N
(so
n
is a

Q-named ordinal)
(b) k
n
is a P
n
-name of a natural number
(c)
n
is a P
n
-name (of a member of T)
(d) q
n

P
n
k
n
=
(e)
0
= and
Q

n
<
n+1
< and
n+1
non-limit
i.e., if < and G
P
n
[G
P
] then
r q
n

n
[G
] < (i.e. is well dened < or is forced to be not well

dened),
(f) q
n+1

n
= q
n
(g) q
n+1

P
n+1
n+1
, so k
n
< k
n+1
.
Finishing the induction we let
=
_
n<
n
and we dene q
n
= q
n
and
q
P
_
n<
n
.
We shall check that
2
holds which is straight.
7.8 Discussion. 1) As in 6 (not 5)?
2) The other NNR.
Like V and like XVIII.
A. Like XVIII,2 - seem straight but check.
B. Like V,6 - think.
3) Explain the specic choice for 7.2.
4) Think Ch.VI,1, =
pr
.
3 not necessarily.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

98 SAHARON SHELAH
8 Examples
Namba [Nm] denes Nm(J
bd
) (and also with ideals) as examples of forcing notion

preserving
1
but changing the conality of some = cf() to
0
.
More [RuSh 117], [Sh:f, X,XI,XV,XIV,5].
8.1 Denition. 1) For an ideal I on a cardinal , let the forcing notion Nm(I) be
Nm(I) =
_
T :T
>
is non-empty, closed under initial segments and
( T)()[ T & (
I
+
< )() T)]
_
where (
I
+
< )Pr() means < : Pr() I
+
and I
+
= A : A / I
ordered by inverse inclusion and let <
pr
= and
vpr
be the equality p Nm(I)
is normal if p [Suc()[ = 1 Suc
T
() ,= mod I.
2) For an ideal I and a cardinal , let the forcing notions Nm
(I) be
Nm
(I) =
_
T : T
>
is non-empty, closed under initial segments and for some
n = n(T) < we have :
(i) n [T
[ = 1
(ii) T & g() n (
I
+
< )[) T]
_
we call the T
n(T)
the trunk of T and denote it by tr(T))
ordered by inverse inclusion and let
pr
=
(see 2) and
vpr
be the equality.
3) Writing a lter D means the dual ideal.
8.2 Claim. Let I be a -complete ideal on ,
2
, I I, I is (restriction
closed and) -complete.
1) Nm(I) and Nm
(I) satises UP
1
(I) and UP
4
+(I) and UP
6
+(I) so does not col-

lapse
1
.
) If I is uniform, then
Nm(I)
cf() =
0
and
Nm
(I)
cf() =
0
, in fact if
[A I
+
I A is
-decomposable] and
is regular, then the same holds for
.
3) [Nm(I)[, [Nm
(I)[ 2
.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

4) If in addition 2
0
< , then forcing with Nm(I) does not add reals, moreover it
satises the condition from 7.2, UP
4
com
(I).
5) If in addition 2
0
< then forcing with Nm(I), Nm
(I) does not add reals;

moreover, they satisfy the condition UP
4,+
stc
(I).
Proof. 1) We will use the following fact about Q = Nm(I) and Q = Nm
(I):
() If p Q,
is a Q-name of an ordinal,
then there is q, p
pr
q such that the set
q : for some we have q
[]

= contains a front.
This fact follows easily from 2.13 (let H : p 0, 1 (i.e., Dom(H) = T
p
) be
dened by H() = 1 i (q)[p
[]
pr
q & q decides
], dene H() =
lim
n<
(H( n)) for lim(p), and nd q such that H is constant on lim(q)).
Let Y = T
q
: H() = 1 & ()( H() = 0], so Y is a front of q. For
Y let q
be such that p
[]
pr
q
and q
forces a value to
let r be such that

T
r
=
_
Y
T
q
. So clearly r is as required, Y such a front.

Now let N
: (T, I)) be a strictly I-suitable tree of models for , x satisfying

p, I, I N
where p Q N
<>
is a condition. We can now nd a condition
q, p
pr
q, a family p
: p) of conditions and a function f : q T satisfying

the following:
(1) If in q, then f() f().
(2) For all in q, Suc
T
(f()) ,= 0 mod I and I
= I.
(3) For all in q, Suc
q
() Suc
T
(f()).
(4) For all in q, p
N
f()
, tr(p
) = , p
[]
pr
p
.
(5) For all in q, p

pr
q
[]
.
(6) For all in q, all names
in N
f()
, the set
q : p
decides contains a front of p

[]
.
We can do this as follows: by induction on p we dene f(), p
and Suc
q
().
We can nd f() satisfying (2) + (3) because T is I-suitable and I I and I is
restriction closed. We choose p
using a bookkeeping argument to take care of a

case of (6), using (). Then we choose Suc
q
() such that (3) are satised.
Lastly, let q = : for some , p
is well dened and . Clearly p

pr
q Q.
Now let G be Q-generic, q G. Now G denes a generic branch through q.
This induces a branch through T : =
_
n<
f( n). Let
N
k
, then there is
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

100 SAHARON SHELAH
such that p
= and N
f()
N
.
2) It is enough to prove the second version for any condition p, let I
p
be I mapped
to Suc
p
().
For any condition p, for each p such that Suc
p
() ,= mod I let h
: Suc
p
()
be such that ( <
)[ Suc
p
() : h
() < I
p
]. Now letting
Ord,
=
n
Ord r for any r G
Q
large enough, we let A
= h
( +1)) :
< .
So easily
Q
A
is unbounded.
3) Trivial.
4) Without loss of generality I is -complete (as we can decrease it). So by 7.4
it suces to prove UP
1
com
(I, W). So assume N
: (T, I)), h
,
(for ,
T lim(T), g() = g()) are as in Denition 7.2, ()
1
and
lim(T), G
as in the assumption of ()
2
there. Now choose inductively on n < , p
n
and k
n
such that: p
0
= p, k
0
= 0, p
n
G
, p
n
N
k
n+1
, and p
n
< p
n+1
, k
n
< k
n+1
,
n
is the trunk of p
n
,
n

n+1
, Suc
p
n
(
n
) ,= mod I (as in proof of part (1)) and
Suc
T
( k
n+1
) = Suc
p
n
(
n
) and p
[
n
(k
n+1
)]
n

pr
p
n+1
and if
is a
Nm(I)-name of a countable ordinal then for some n, p
n
decides its value.
5) The winning strategy of the completeness player is, given q
n
, let = tr(T)
and let n be minimal such that q
n
N
n
and I Suc
q
n
() = I
n
and let
p
n
= (q
n
)
[
(n)]
.
8.2
8.3 Denition. 1) We can consider an I-suitable tree of models
N = N
: (T
, I
)), and let

a) Q
N
=
_
T T
: T non-empty, closed under initial segments

such that N
: (T, I T)) is an
I-suitable tree of models
_
ordered by inverse inclusion.
2) We can consider for any tagged tree (T
, I
)
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

Q
0
(T
,I
)
=
_
T T
: T non-empty, closed under initial segments

such that for some n = n(T),
(i) n [T
n
Ord[ = 1
(ii) if T & g() n & Suc
T
() ,= mod I
then Suc
T
() ,= mod I
_
is ordered by inverse inclusion.
Q
1
(T,I
)
=
_
(T, I) :(T
, I
) (T, I), and for every lim(T)

we have (k)(
n)[ n is a splitting point of (T, I)

and I
k

RK
I
n
_
ordered by inverse inclusion. [Saharon [Q
0
,= Q
2
?]
Q
2
(T,I
)
=
_
(T, I) : (T
, I
)
[]
(T, I), for some T
_
ordered by inverse inclusion.
8.4 Claim. For the forcing notions dened in Denition 8.3 for I being
2
-complete,
of course, we have: if P Q
N
, Q
0
(T
,I
)
, Q
1
(T,I
)
, then
(a) P satises UP
1
(I)
(b) if I I [Dom(I)[ < = cf(), then [P[ 2
<
and even 2
for some
<
(c) if for regular
(I I)(A (I)
+
[I A is not -indecomposable]
then
P
cf() =
0
(d) if P = Q
0
(T
,I
)
then ( lim(T
))
mn
A (I
n
)
+
[I
A is not
-indecomposable] then
P
cf() =
0
(e) if I is (2
0
)
+
-complete then forcing with P add no new reals, moreover it
satises UP
4,+
stc
(I) and if p Q
N
, Q
0
(T
,I
)
then it satises UP
4
com
(I).
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

102 SAHARON SHELAH
Proof. Left to the reader.
8.4
8.5 Denition. Let = cf() >
1
, S < : cf() =
0
be stationary and
club
S
(S) =
_
h : for some non-limit <
1
,
h is an increasing function from to S
_
ordered by inverse inclusion,
pr
=,
vpr
is equality.
8.6 Claim. For , S as in Denition 8.5 we have (for any I, I is an
2
-complete
ideal on extending J
bd
).
1) Club(S) satises UP
1
(I) of cardinality
0
.
2) If I is (2
0
)
+
-complete and I I, then Club(S) satises UP
4
com
(I), hence
UP
4,+
stc
(I).
Proof. Left to the reader or follows from [Sh:f, XI,4.6] by 8.9 below.
8.6
8.7 Lemma. Let

W = W
i
: i <
1
) be a sequence of stationary subsets of <
: cf() = where = cf() >
0
and let the forcing notion P[

W] be dened by
P[

W] =:
_
f : f is an increasing and continuous function from
+ 1 into W
0
for some <
1
,
such that for every i we have f(i) W
i
_
(ordered by inclusion). If I J
bd
be
2
-complete, then P[

W] satises UP
4,+
(I) for
any I such that I I and if I is (2
0
)
+
-complete it also satises UP
4
com
(I) hence
UP
4,+
stc
(I).
Proof. Left to the reader or follows from [Sh:f, Ch.XI,4.6A] by 8.9 above.
8.7
Concerning W-completeness (see [Sh:f, Ch.V]):
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

8.8 Claim. Assume W
1
is stationary and Q is W-complete forcing notion
(i.e., if is large enough, Q N (H (), , <
), N countable, p
n
Q N is
Q
-increasing and (I N)(I Q is dense
n
p
n
I) then p
n
: n <
has an upper bound in Q).
Then Q satises UP
4,+
com
(I, W) (i.e., for any I).
Proof. Trivial (see Denition ?), for the -branch
of T there is q, p q, q is an
> scite7.1 undened
upper bound of Q
hence is (
_
<
N
, Q)-generic.
8.8
Comparing to [Sh:f] we have
8.9 Claim. 1) If Q satises the I-condition of [Sh:f, Ch.XI,Def.2.6,2.7] I is (2
0
)
+
-
complete, then Q satises UP
4
com
(I).
2) If Q satises condition UP(I, W) of [Sh:f, Ch.XV, Denition 2.7A], then it
satises UP
4
(I, W) here.
3) If Q is a proper or just semi-proper forcing notion, then Q satises UP
6
(I) and
all the UP
(I).
Proof. Part (2) holds by [Sh:f, XV,2.11].
Part (1) follows by part (2) and [Sh:f, Ch.XV,2.11].
Part (3) is immediate by the denition (can use any x branch).
8.9
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

104 SAHARON SHELAH
9 Reflection in [
2
]
0
As an exercise we answer a question (and variants) of Jech
9
.
9.1 Theorem. Assume
(A) is large enough (supercompact or just Woodin).
Then
1) there is a -c.c. semi-proper forcing notion P of cardinality such that in V
P
()
2
=
V
1
,
2
= and cardinal are the same as in V and 2
0
=
2
() every stationary subset of [
2
]
0
reect (in a set of cardinality
1
)
() D
1
is
2
-saturated
() there is a projectively stationary S [
2
]
0
, see below such that there is
no sequence a
i
: i <
1
) increasing continuous, a
i
S, a
i
,= a
i+1
.
2) Assume in addition
: measurable is not in the weakly compact ideal of .
We can add to (1) the statement (on V
P
)
() every stationary S S
2
0
= <
2
: cf() =
2
contains a closed copy of
1
.
3) We may strengthen clause () of (1) to S is S
2
1
-projectively stationary.
9.2 Denition. 1) We call S [
2
]
0
projectively stationary if:
for every club E of [
2
]
0
and stationary co-stationary W
1
we can nd a
sequence a
i
: i <
1
) increasing continuous, a
i
[
2
]
0
, a
i
E and i W : a
i
/
S is not a stationary subset of
1
.
2) We say S [
2
]
0
is S
2
1
-projectively stationary for W
1
stationary co-
stationary, for stationarily many S
2
1
if we let =
_
i<
1
a
i
, a
i
countable increasing
continuous we have i W : a
i
/ S non-stationary.
Proof. Like [Sh:f, Ch.XVI,2.4]s proof.
1), 2) We dene a RCS iteration P
i
, Q
j
: i , j < ) such that:
9
Done 10/97
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

[Q
j
[ (2
2
)
V
P
j
in V
P
j
, Q
j
is the disjoint union of the following (so the choice by which of them we
force is generic), but if j is non-limit only (a) is allowed
(a) Q
0
j
= Levy(
1
, 2
1
) Cohen
(b) Q
1
j
= sealing all semi-proper maximal antichanges of D
1
provided that
strong Chang conjecture holds in V
P
j
(true if j is measurable >
0
)
(c) if we like to have () and in V
P
j
, strong Chang conjecture holds then allow:
Q
2
j,S
where S S
2
0
stationary not containing a closed copy of
1
and Q
2
j,S
semi-proper where Q
2
j,S
shoot an
1
-increasingly continuous chain i.e.
Q
2
j,S
= (S, f) :S S
2
0
stationary, Dom(f) is a successor
countable ordinal, f is increasingly continuous into S
(S, f
1
) (S, f
2
) S
1
= S
2
f
1
f
2
.
So
P
j
Q
j
is semi-proper of cardinality (2
1
)
V
P
j
.
So by [Sh:f, Ch.XVI,2,2.4,2.5]
1
for i < j , P
j
/P
i
is semi-proper, so
1
is not collapsed
2
P
collapses every (
1
, ), satises the -c.c. and has cardinality
3

P

1
is
2
-saturated.
By preliminary forcing, without loss of generality there is S
0
= < : strong limit, cf() =
0
, stationary in , reecting only in inaccessibles. Let S
1
= < : is measurable
so we know S
1
is stationary. If we are proving the version with (), note that
S
1
in V
P
, the strong Chang conjecture holds ([Sh:f, Ch.XIII,1.9]) hence

Q
2
,S
is semi-proper for every stationary S < : V
P
[= cf() =
0
. Also if
holds then
< : in V
P
, cf() =
0
is stationary
S
1
= S
1
: S
is a P
-name of a stationary
subset of < : cf() =
0
in V
P
is stationary.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

106 SAHARON SHELAH
So
4
if we are proving (), then in V
P
every stationary S < : cf() =

0
contains a close copy of

1
.
Now we have to deal with the projectively stationary. We can nd function h,
Dom(h) = S
0
, h() is a P
-name of a stationary co-stationary subset of

1
(even
P
-name for some < ) such that: every such name appears stationarily often.
Let a
i
: i <
1
) be a P
-name such that
i
is countable unbounded in increasingly continuous in i
and =
_
i<
1
a
i
.
Let
W

= a
i
: i h()
W
<
= W
: S
0
= W
<
.
So
5
W
is a P
-name of a subset of []
0
and S
0
is stationary in V
P
(as P
[= -
c.c.)
6

P
is stationary.
[why? If
P
is a model with countable vocabulary and universe then E
=
< : M
is a P
-name and is an elementary submodel of M is a P
-name of
a club of hence contains a club E
of from V. So for a club i <

1
, M a
i
is
an elementary submodel of M. But for stationarily many i <
1
, a
i
W
W , so
really W is stationary. If W
is a P
-name of a stationary co-stationary subset of
1
then for some, even for stationarily many E
S
0
we have h() = W
and
so easily
7

P
is projectively stationary.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

Lastly, why would W contain no increasing
1
-chains? Assume p
i
: i <
1
)
is increasing continously and a
i
W. So without loss of generality for some
either
() p
sup(a
i
) is strictly increasing with limit
or
() p
sup a
i
is constantly
for i i
, i
<
1
so
without loss of generality i
= 0.
Case A: The possibility () holds.
Necessarily
S
0
, p
P
a
i
: i <
1
W
and as
P
h(
) is co-
stationary subset of
1
and P
/P
is semi-proper hence preserve stationarity of

subsets of
1
we are done.
Case B: Possibility () holds and
not strongly inaccessible. So S

0

is not
stationary in
hence W
0
=
_
S
0
W
is not even stationary.

Case C: Possibility () holds and not case B, in V
P
strong Chang conjecture

fails.
Then Q
is Levy(
1
, 2
1
) Cohen (as in clauses (b) and (c) in V
P
strong Chang
conjecture holds), so as clearly in V
P
, 2
0
=
2
(by the Cohen in (a), i.e. Q
0
j
),
then in V
P
for every club E
of [
0
, we can nd some <
and 2
0
members
of E
[]
0
W
<
. So in V
P
, [
0
W
<
is stationary and Q
is proper so
this holds in V
P
+1
. But P
/P
+1
preserves stationarity of subsets of
1
hence in
V
P
0
< W < is stationary, so we are done.
Case D: Possibility () holds, not case B and in V
P
strong Chang conjecture

holds.
Just note: in V
P
, let p Q
, let large enough N (H (), ) is countable

to which

Q,
, G
P
, p, W
: S
0

) belong, then we can nd (see [Sh:f,

Ch.XIII]) T
>
2
closed under initial segments T N = , satisfying (
T)(
2
)() T) and N
: T) such that
(i) N
<>
= N
(ii) N
(H (), ) is countable
(iii) N
, N

1
= N
1
(iv) N
(v) if I = A
: <
is a maximal antichain of D
1
which is semi-proper and
I N
then for some k < , T & g() k N

1

_
N
.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

108 SAHARON SHELAH
Let
E = <
2
: if
>
then N

2
is a bounded subset of .
Now if p Q
0
we do as in Case C. If p Q
1
, choose E, cf() =
0
, and such
that for every T
>
, = otp < : ) T and ) T & <
sup(N
2
) < . Now we can by cardinality considerations (2
0
>
1
) nd
lim(T)
such that letting M =

_
<
N
, M
2
= M / W
<
. So there
is q Q
which is (M, Q
)-generic, p q (by the denition of Q

1
). Now q forces
a
M
1
= a
M
1
to be M
2
which is not in W
<
.
Lastly if q Q
2
j,S
(in V
P
) as S does not reect we can nd E as above, S,

cf() =
0
and choose , M as above.
3) We may like to adapt the proof above.
We omit the choice of a
i
: i <
1
), but in Q
j
if j S
0
we also choose a P
j
-name of a
countable unbounded subset of , a
and let W

= a
so Q
j
is replaced by Q
j
a
:
a
a name as above. Now h

0
has domain S
= : strongly inaccessible, in
V
P
, strong Chang conjecture holds, h

0
() a P
-name of a stationary co-stationary

subset of
1
and we add to clauses (a), (b), (c) above also
(d) dene in V
P
:
Q
3
=
_
M
i
: i j) :the ordinal j is countable and
M
i
(H ((2
)
+
), ) is countable increasing
continuous, and: if M
i

1
h
0
() then
M
i
W
<
and if M
i

1
/ h() then M
i
/ W
<
_
.
Now again we use N
: T) and choosing M it is enough to show that
1
for some limT,
_
i<
N

2
W
<
2
some limT,
_
<
N

2
/ W
<
.
Now
2
is as before,
1
O.K. by the way W
<
is dened.
9.1
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

10 Mixing finitary norms and ideals
We may consider replacing families of ideals by families of creatures see [RoSh 470]
on creatures:
We hope it will gain something
10.1 Denition. : 1) A -creature c consists of (D
c
, , val
c
, nor
c
,
c
), where:
c
=
D
c
the domain,
a partial order on D
c
,
val
c
is a function from D
c
to P()
nor
c
: D
c
or to??
2) It is called simple if nor
c
is always > 0 (without loss of generality constant, e.g.
Rang(val
c
) = I
+
, I an ideal on ). A creature is a -creature for some .
I will be a set of creatures.
10.2 Denition. 1) An I-tree is (T, I, d) such that:
for some ordinal , T
>
closed under initial segments, ,=
I is a partial function Dom(I) T, I
I,
d has domain Dom(I), d() D
I
val
I
(d()) = : ) T.
2) Let (T
, I
, d
) be an I-tree, such that

() ( lim T
)[lim sup n < nor

I
n
(d
) = ]
and Dom(I
) = T
.
We dene a forcing notion Q = Q
(T
,I
,d
)
:
Q =
_
(T, I, d) :T T
, I = I
T,
(T, I, d) an I-tree,
( T)(d
)
and (( lim T
) lim sup nor

I
n
(d
) = )
_
.
Order: natural.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

110 SAHARON SHELAH
3) Let (T
, I
, d
) be an I-tree such that

() ( lim T
)(n)(
)(nor
I
n
(d
) n)
(i.e. lim inf = ).
and dene Q
= Q
(T
,I
,d
)
parallelly.
4) For p Q (or p Q
) we write p = (T
p
, I
p
, d
p
). In this case for T
p
we dene
q = p
[]
by:
T
[q]
= T
p
: or , I
q
= I
p
T
q
, d
q
= d
p
T
p
. Clearly p Q p
q Q and p Q
p q Q
.
10.3 Claim. Let (T, I
, d
) and Q, Q
be as in 10.2. A sucient condition for

1
not collapsed is:
(a) for Q: () below
(b) for Q
: () + () below where
() I has
1
-bigness:
(c I)(x D
c
)
_
nor
c
(x) > 0 (h
(
c
)
1
)(y)
[x
c
y & nor
c
(y) nor
c
(x) 1 &
(h val
c
(y) is constant]
_
() I is (
1
,
1
)-indecomposable where I is (, )-indecomposable means:
I,,
if c I and x D
c
satises nor
c
(x) > 2 and A

c
for < are such that (y)(x y D
c
val
c
(y) A

nor
c
(y) + 2 nor
c
(x)), then we can nd u
c
of cardinality
< such that for every large enough < we have u A
.
Proof for Q. Lets use given p = (T, I, d) Q and Q-name
such that
:
1
.
Now we choose by induction on n, p
n
, A
n
such that:
(a) p
n
p
n+1
(b) A
0
, . . . , A
n
are fronts of p
n
which means ( lim
p
n
)(!n)( n A
)
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

(c) A
below A
+1
which means
( A
+1
)( ) A
(so A
n
T
p
n+1
, ( T
p
n
T
p
n+1
)( )( A
n
))
(d) ( A
n
)( Suc
T
p
n ())(p
[]
n
forces a value to
(n))
(e) A
n
nor
I
(d
p
n
) n & d
p
n
= d
p
n+1
, it follows that < n &

A
d
p
= d
p
n
& Suc
T
p
n
() = Suc
T
p
().
So p
is dened by T
p
n<
T
p
n
, I
p
= I T
p
, d
p
=
_
n
d
p
n
: A
n
and g() : n < belong to Q and is an upper bound of p
n
: n < .
We dene h : T
p

1
as follows: if T
p
, A
1
,
(if = 0 omit
, so just
), then p
[]
forces value to
call it ( )
p
[]
and let
h() = Sup Rang(
)
p
[]
.
For notational simplicity A
n
= T
p
n
Ord.
We now dene a game =
T
p
for each <
1
:
A play of the game last moves, in the (n 1)-th move a member
n
of A
n
is
chosen such that m < n
m

n
, and xing some
1
A
n
.
In the n-the move:
(a) the anti-decidability player chooses a set A
n
Suc
T
p
(
n1
) such that
B
n
= (nor
I
(A
n
) n 2, n 2)
1
B
n
,= or n 3 and for no d, satisfying d
p
() d nor
I
2
(d) n2
do we have ) A
n
val
I
(d)
(b) the decidability player chooses
n
A
n
such that ( B
n
)(
n
) and
n 1 (
n
(g(
n1
) 1)) .
Without loss of generality A
0
= <>.
If for some and decidability player has a winning strategy, we can produce a
condition as required.
If not, for every <
1
the antidecidability player has a winning strategy St
.
For each T
p
and <
1
, we consider the play of the game in which the an-
tidecidability player has winning strategy St
and in some move n the decidability

(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

112 SAHARON SHELAH
player chooses
n
= . Reecting there is no freedom left so there is at most one
such play and n and let the antidecidability player choose set B
,
as there (if no
such game let B
,
= ).
Case 1: For some n and A
n
, we have: there is no countable u Suc
T
p
()
such that for every large enough <
1
, u B
,
so by the assumption () we
get a contradiction.
Case 2: Not Case 1.
We can choose by induction on n, a countable subset u
n
A
n
such that: u
0
=
<>
if A
n
then for some
<
1
for every [
,
1
),
if in the play in which the antidecidability player uses St
and they arrive to , there is
u
n
+ 1
which is a legal response of the decidability player.
mn Now let
=suph() + 1 : for some

_
n
u
n
, , Dom(h)+
sup
:
_
n
u
n
and we can nd a play of
as above where the decidability player chooses s

from
_
n<
u
n
. We get a contradiction.
Proof for Q
.? We should make changes: in p

n+1
we shrink p
[]
n
for each
T
p
n
n
Ord, to q
, p
[]
n

pr
[] and for each n, if possible, q
forces a bound to
() and, of course, p
[]
n+1
= q
for each such and T

p
n+1
Ord = T
p
n
n
Ord
and d
p
n+1
n
Ord = d
p
n
n
Ord. So let p
n
p
n
be naturally dened, and
we use 2-bigness to prove enough times q
forces a bound.
Now we give details.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

Proof for Q. Given p = (T, I, d), for notational simplicity tr(p) =<>and nor
d
p
(Suc
T
p()) >
2 and P-name
such that
:
1
we choose by induction on n, p
n
such that:
(a) p
n
p
n+1
, and p
n
has trunk n
(b) A
0
, . . . , A
n
are fronts of p
n
(c) A
below A
+1
which means
( A
+1
)( ) A
(so A
n
T
p
n+1
, ( T
p
n
T
p
n+1
)( )( A
n
))
(d) A
0
= <>[ A
n
T
p
n
nor
I
(d
p
n
) > n + 2]
(e) when A
n
let =
n be maximal such that there are

m
<
1
, m <
and q satisfying trq = , p
[]
n
q P, q
m<
(m) <
n
and
T
q
nor
d
(d
q
) n and we demand: p
[]
n+1
satises the demand on q for
some
m
: m <
), note possible
= then we are left with demand on

norm.
So p
, T
p
n<
T
p
n
is an upper bound of p
n
: n < .
Clearly p
P and n < p
n
p
. Let for T
p
, let n() = Maxn :

there is , A
n
and
be in A
n
and
<
1
be minimal such that p
[
]
n+1
forces (0), . . . , (
1) <
. Using games as in the proof for Q
there is p
+
such
that:
(a) p
P
(b) T
p
+
nor
I
(d
p
+
) nor
I
(d
p
) 1
(c)
= sup
: T
p
+
<
1
.
We continue as in [Sh:f, Ch.XIV,5].

Discussion: We can continue to do iteration.
But more urgent: can Q, Q
like this do anything not already covered by composi-

tion?
A natural thought is splitting or reaping numbers. We can think of the tree
splitting in T
as a list of the reals. BUT, what is the norm?

Not nished...check the betters theorem proof?
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

114 SAHARON SHELAH
Assignment: [Sh:f, XIII,XVI] and here put together, so does the reection Pr
a
(, f)
works for ???
97/2/2 - Discussion:
Saying a creature is -complete meana that for pure extensions, increasing chains
of length < have pure upper bounds? Probably pure means not changing the
norm; maybe the
1
-indecomposable is enough.
So the I-th condition has a new meaning.
Question: Does the theorem here hold?
Question: Does this new context have real applications?
The rst result to be discussed is moving from I
to one in the ground model.

The second are 5.2, ? preservation of N being suitable.
> scite6.2 undened
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

11 Variants of the iteration
As mentioned in 1, we can consider -RS iteration and variants of the Sp
e
iteration.
11.1 Denition/Claim. Let be a successor cardinal or an innite ordinal not
a cardinal but an ordinal of power [[, x
10
. We dene and prove the following
by induction on (here e = 3, 4, 5, 6). If =
1
, we may omit it and this is the
main case.
We repeat 1.15 with the following changes in the proof and denition:
(B) We say
is a simple

Q-named
e
[j, )-ordinal if
()
1

is a simple

Q-named
1
[j, )-ordinal and may restrict ourselves to
=
1
e 3, 4
()
2
if e 5, 6 and =
1
, then
is a simple

Q-named
2
[j, )-ordinal.
(F)(a) in (v) replace the remark in the end by:
if e 5, 6, w then this demand follows by 1.8
and add:
(vii) if e = 3, 5 then for some n < and simple

Q-named [0, g(
Q))-ordinals
1
, . . . ,
n
we have, for every < g(
Q)
P
if for = 1, . . . , n we have
[G
] ,= (for example
[G
P
] not well dened) then

Q

pr
p
in

Q
(F)(e)(iii) inside change p

2

P
. . . by
p
2

P
if ,=
[G
] for
= 1, . . . , n and
[e = 4 e = 6 (
Q

vp
p
1
)]
then:

Q
[= p
1

pr
p
2
.
10
For inaccessible, see ?.
> scite{1.22} undened
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

116 SAHARON SHELAH
11.2 Claim. 1) As in 1.16 adding ,=
1
e 3, 4.
2) If

Q is an Sp
e
(W)-iteration, and for each i the quasi-order
Q
i
pr
is equality
hence
Q
i
vpr
is equality, then

Q is essentially a nite support iteration.
[Saharon: maybe restrict yourself above the constantly function
Q
, so we
have to use > g(
Q).]
11.3 Claim. 1) Add
(d) if e = 3, 5 then r is pure outside
1
, . . . ,
n
.
2) In the proof on
1
?? we say, i.e., simple
1
if e 3, 4 and simple
2
if
e 5, 6.
11.4 Claim. 5) If e 3, 4 and
11
for each < g(
Q), t
is a P
-name of a truth
value, then there is a simple (
such that
[G
] = i
t
[G
] = truth and < t
[G
] = false for any subset G
of P
generic over
V.
We can deal with the parallel of hereditarily countable names. This is not used
in later sections.
11.5 Denition. We dene for an Sp
e
(W)-iteration

Q, and cardinal (
regular), when is a (
Q, W)-name hereditarily < , and in particular when a (
Q, W)-
named [j, )-ordinal is hereditarily < and a (
Q, W)-named [j, )-atomic condition

hereditarily < , and which conditions of Sp
e
(W)-Lim
Q) are hereditarily < .

For simplicity we are assuming that the set of members of Q
i
is in V. This is done
by induction on = g(
Q).
First Case: = 0.
Trivial.
Second Case: > 0.
(A) A

Q-named [j, )-ordinal
hereditarily < is a (
Q, W)-named [j, )-ordinal

which can be represented as follows: there is (p
i
,
i
) : i < i
), i
< , each
i
an ordinal in [j, ), p
i
P
i
is a member of P
i
hereditarily < and for
any G Gen
r
(
Q),
[G] is i for some i we have:

(a) p
i
G,
i
=
(b) if p
j
G then
i
<
j
(
i
=
j
& i < j)
11
for the parallel for e {5, 6}, =
1
we need pure decidability and restrict ourselves to
above p for purely dense sets of p s
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

(B) A (
Q, W)-named [j, )-atomic condition q
hereditarily < , is a (
Q, W)-
named [j, )-atomic condition which can be represented as follows: there
is (p
i
,
i
, q
i
) : i < i
), i
< ,
i
[j, ), p
i
P
i
, q
i
V, and for any
G Gen
r
(
Q), q
[G] is q i for some i we have:

(a) p
i
G, q = q
i
, and p
i

P
i
q Q
(b) if p
j
G then
i
<
j
(
i
=
j
& i < j)
(C) A member p of P
= Sp
e
(W)-Lim
Q) is hereditarily < if each member

of r is a (
Q, W)-named atomic condition hereditarily <

(D) A (
Q, W)-name of a member of V hereditarily < is dened as in clause

(B), similarly for member x V
P
such that y transitive closure of

x [y[ < .
11.6 Concluding Remarks. 1) We have not really dealt with the case is inacces-
sible. The point is that in this case, we do not know a priori the length of the list
of the members of a condition (which are atomic conditions). It is natural to work
on it together with decidability on bound on
< by pure extensions, see 1.23

below.
2) We can think of putting together [Sh:f, Ch.XIV] and [Sh 587].
3) We can ask: Does Souslin forcing notions help?
11.7 Claim. e 4, 5 is O.K.
11.8 Claim. In the proof of 1.26, in case 3 add:
(the point is that e 4, 6). Instead e 4, 6 it is enough to assume:
for every q
, q
0
we have
vpr
q
pr
q
.
11.9 Remark. 1) Add:
but for e = 4 we could use appropriate p
1
= p r
1
, r
an atomic (
Q, W)-named
condition,
, see 1.7(5).
2) Holds for e 4, 6.
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

118 SAHARON SHELAH
11.10 Denition. 0) Let

Q be a
1
-Sp
e
(W)-iteration of length . Let
denote a
simple

Q-named [0, )-ordinal or a simple

Q-named
2
[0, )-ordinals and a count-
able set of such objects.
1) For an atomic simple

Q-named condition r, r
is dened by r
[G] = r
if
[G]
r
[G], r[G] = r
and
p
otherwise.
2) For q P
, q
= r : r q and q =
_
q .
3) P
= p P
: p
= p, i.e., for every G P
generic over V, p
[G] = p[G]
P
= p P
: p = p
both with the order inherited from P
.
11.11 Claim. Let

Q be an
1
Sp
e
(W)-iteration.
1) If
1
is a simple

Q-named [
1
,
2
)-ordinal,
2
is a simple

Q-named
1
[
1
,
2
)-
ordinal, then there is a simple

Q-named [
1
,
2
)-ordinal
such that for G P
is
generic over V:
(a) if
1
[G] =
1
[GP
] = and Min: some p GP
decided to be or be
undened > then
[G] =
[G P
] =
(b) otherwise undened.
2) Let
be a simple

Q-named
1
ordinal. For r an atomic

Q-named condition r
is an atomic

Q-named condition.
3) For q P
we have q
.
4) For q
1
, q
2
P
, q
1
q
2
q
1

q
2

.
5) If q P
, p P
, p q then p q P
is a lub of p and q.
6) P
.
7) If G P
is generic over V, =
[G] then G P
, G P
are essentially the

same.
8) The parallel statements with instead of
.
Remark. In fact by part (1), part (6) follows from the parts (2)-(5).
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

11.12 Claim. Let e = 4(2). Assume
n
is a simple

Q-named for n < ,
n
<
n+1
and for every G P
generic over V, for some n, (n, GP
n
[G]
). Then for some
simple

Q-name ordinal
, we have
for some n,
[G
] =
n
[G
] and (n, G
n
[G
P
]
).
(
3
1
1
)

r
e
v
i
s
i
o
n
:
2
0
0
6
-
0
7
-
2
6

m
o
d
i
f
i
e
d
:
2
0
0
6
-
0
7
-
2
7

120 SAHARON SHELAH
REFERENCES.
[GiSh 191] Moti Gitik and Saharon Shelah. On the I-condition. Israel Journal of
Mathematics, 48:148158, 1984.
[RoSh 470] Andrzej Roslanowski and Saharon Shelah. Norms on possibilities I:
forcing with trees and creatures. Memoirs of the American Mathemat-
ical Society, 141(671):xii + 167, 1999. math.LO/9807172.
[RuSh 117] Matatyahu Rubin and Saharon Shelah. Combinatorial problems on
trees: partitions, -systems and large free subtrees. Annals of Pure
and Applied Logic, 33:4381, 1987.
[Sh 656] Saharon Shelah. NNR Revisited. Preprint. math.LO/0003115.
[Sh:e] Saharon Shelah. Nonstructure theory, accepted. Oxford University
Press.
[Sh:b] Saharon Shelah. Proper forcing, volume 940 of Lecture Notes in Math-
ematics. Springer-Verlag, Berlin-New York, xxix+496 pp, 1982.
[Sh 136] Saharon Shelah. Constructions of many complicated uncountable struc-
tures and Boolean algebras. Israel Journal of Mathematics, 45:100146,
1983.
[Sh:f] Saharon Shelah. Proper and improper forcing. Perspectives in Mathe-
matical Logic. Springer, 1998.
[Sh 587] Saharon Shelah. Not collapsing cardinals in (< )
support iterations. Israel Journal of Mathematics, 136:29115, 2003.
math.LO/9707225.

Saharon Shelah - A More General Iterable Condition Ensuring 1 Is Not Collapsed

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Saharon Shelah - A More General Iterable Condition Ensuring 1 Is Not Collapsed

Uploaded by

Copyright:

Available Formats

(

and preserving the

Q) is like that, provided some mild condition holds (say Q

[ < and: W stationary in (so is

) and a minimal element

an outer witness for A or for A

B i A B and A ,= B implies that

q); we call such a pair (p, q) a witness

A satises clause (i) + (ii) then there is p

is a quasi order for 2) and P

) is -complete (i.e., an increasing sequence of length

-upper bound in P, call it q. Now

is the lter of co-bounded subsets of .

is the club lter on for regular uncountable.

= (g() 1) for a nite sequence of length > 0.

be a well ordering of H ().

and for some G

generic over V we have

[G] V[G] [from where every G. . . is

is really a rst-order formula with the variable

Q] that are dened. We let

= x if for every G such that p G Gen

] = x means that for some p G

and no such p forces

[G] is well dened for every G Gen

such that: if G Gen

[G] = then j < and (p G P

Q)); however, we allowed

[G] to be undened. If we omit [j, ) we mean

Q)) = [0, ). If we omit simple, we mean replacing (p G P

Q) and maximal antichain I = p

] = i and for some we have p

] = (including the case: not dened). If we omit [j, ) we mean

P, W) is continuous when for every W g(

Q, W)-named [j, )-ordinal if it is a simple

is well dened then it is equal to

but not necessarily

hence it will not in general be a maximal antichain.

P)), a -increasing sequence of forcing notions,

is not dened (and even p

[j, )-ordinals (naturally dened,

[G] are dened, and min

are full then so are max

such that, rst: p q or at least

for some i < g(

P) and second: for some 1, . . . , n we

and we can omit simple.

P) then any [simple]

Q) = inaccessible and i < [P

(here as done in [Sh:f, Ch.XIV,2] we can just impose it).

and ordinals < , (or

instead , ). We may add pure/very pure as

Q, W)-named [j, min, )-ordinal and

Q, W)-named atomic [j, )-condition, < ,

Q , W)-named atomic [j, min, )-condition and

or undened (and then it may be

, the minimal member of Q

Q, W)-named ordinals then

Q, W)-named ordinal. Also q

)-ordinal. Also pure and very

Q), an ordinal ; (similarly for a simple (

, etc. We also dene P