Andrzej Roslanowski and Saharon Shelah - Norms On Possibilities I: Forcing With Trees and Creatures

4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Norms on possibilities I: forcing with trees and
creatures
Andrzej Ros lanowski
Saharon Shelah
Institute of Mathematics, The Hebrew University of Jerusalem,
91904 Jerusalem, Israel, and Mathematical Institute of Wroclaw Uni-
versity, 50384 Wroclaw, Poland
Current address: Department of Mathematics and Computer Science, Boise
State University, Boise ID 83725, USA
E-mail address: roslanow@math.idbsu.edu
URL: http://math.idbsu.edu/roslanow
Institute of Mathematics, The Hebrew University of Jerusalem,
91904 Jerusalem, Israel, and Department of Mathematics, Rutgers
University, New Brunswick, NJ 08854, USA
E-mail address: shelah@math.huji.ac.il
URL: http://www.math.rutgers.edu/shelah
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

1991 Mathematics Subject Classication. Primary 03E35;
Secondary 03E40, 03E05
The rst author thanks the Hebrew University of Jerusalem and the Lady Davis
Foundation for the Golda Meir Postdoctoral Fellowship, and the KBN (Polish
Committee of Scientic Research) for partial support through grant 2P03A01109.
The research of the second author was partially supported by Basic Research
Foundation of the Israel Academy of Sciences and Humanities. Publication 470.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Contents
Annotated Content v
Chapter 0. Introduction 1
0.1. The content of the paper 2
0.2. Notation 3
Chapter 1. Basic denitions 5
1.0. Prologue 5
1.1. Weak creatures and related forcing notions 7
1.2. Creatures 12
1.3. Tree creatures and treelike forcing notions 15
1.4. Non proper examples 20
Chapter 2. Properness and the reading of names 27
2.1. Forcing notions Q
s
(K, ), Q
w
(K, ) 27
2.2. Forcing notion Q
f
(K, ): bigness and halving 31
2.3. Treecreating (K, ) 39
2.4. Examples 43
Chapter 3. More properties 53
3.1. Old reals are dominating 53
3.2. Preserving non-meager sets 54
3.3. Preserving non-null sets 59
3.4. (No) Sacks Property 63
3.5. Examples 63
Chapter 4. Omittory with Halving 67
4.1. What omittory may easily do 67
4.2. More operations on weak creatures 69
4.3. Old reals are unbounded 71
4.4. Examples 80
Chapter 5. Around not adding Cohen reals 94
5.1. (f, g)-bounding 94
5.2. (
t,

T)bounding 98
5.3. Quasi-generic and preserving them 104
5.4. Examples 114
Chapter 6. Playing with ultralters 123
6.1. Generating an ultralter 123
6.2. Between Ramsey and p-points 126
iii
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

iv CONTENTS
6.3. Preserving ultralters 129
6.4. Examples 140
Chapter 7. Friends and relatives of PP 147
7.1. BalcerzakPlewik number 147
7.2. An iterable friend of the strong PPproperty 149
7.3. Bounded relatives of PP 153
7.4. Weakly non-reducible p-lters in iterations 156
7.5. Examples 157
List of denitions 165
Bibliography 167
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Annotated Content
Chapter 0: Introduction
Chapter 1: Basic definitions We introduce a general method of building
forcing notions with use of norms on possibilities and we specify the two cases we
are interested in.
1.0 Prologue
1.1 Weak creatures and related forcing notions [We dene weak crea-
tures, weak creating pairs and forcing notions determined by them.]
1.2 Creatures [We introduce the rst specic case of the general schema:
creating pairs and forcing notions of the type Q
C(nor)
(K, ).]
1.3 Tree creatures and treelike forcing notions [The second case of the
general method: forcing notions Q
tree
e
(K, ) in which conditions are trees
with norms; tree creatures and treecreating pairs.]
1.4 Non proper examples [We show several examples justifying our work
in the next section: the method may result in forcing notions collapsing
1
, so special care is needed to ensure properness.]
Chapter 2: Properness and the reading of names We dene properties
of weak creating pairs which guarantee that the forcing notions determined by
them are proper. Typically we get a stronger property than properness: names for
ordinals can be read continuously.
2.1 Forcing notions Q
s
(K, ), Q
w
(K, ) [We show that the respective
forcing notions are proper if (K, ) is nitary and either growing or cap-
tures singletons.]
2.2 Forcing notion Q
f
(K, ): bigness and halving [We introduce an
important property of creatures: bigness. We note that it is useful for
deciding bounded names without changing the nite part of a condition
in forcing notions discussed in 2.1. Next we get properness of Q
f
(K, )
when the creating pair (K, ) is big and has the Halving Property.]
2.3 Treecreating (K, ) [We show that properness is natural for forcing
notions Q
tree
e
(K, ) determined by treecreating pairs (with our norm
conditions). With more assumptions on (K, ) we can decide names on
fronts.]
2.4 Examples [We recall some old examples of forcing notions with norms
putting them in our setting and we build more of them.]
Chapter 3: More properties We formulate conditions on weak creating
pairs which imply that the corresponding forcing notions: do not add unbounded
reals, preserve non-null sets or preserve non-meager sets.
v
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

vi ANNOTATED CONTENT
3.1 Old reals are dominating [From the results of section 2 we conclude
that various forcing notions are
bounding.]
3.2 Preserving non-meager sets [We deal with preservation of being a non-
meager set. We show that if a treecreating pair (K, ) is T-omittory then
the forcing notion Q
tree
1
(K, ) preserves non-meager sets. We formulate
a weaker property (being of the NMPtype) which in the nitary case
implies that forcing notions Q
tree
1
(K, ), Q
f
(K, ) preserve non-meager
sets. We get a similar conclusion for Q
w
(K, ) when (K, ) is a nitary
creating pair which captures singletons.]
3.3 Preserving non-null sets [We formulate a property of treecreating
pairs which implies that the forcing notion Q
tree
e
(K, ) preserves non-null
sets.]
3.4 (No) Sacks Property [An easy condition ensuring no Sacks property
for forcing notions of our type.]
3.5 Examples [We build a treecreating pair (K, ) such that the forcing
notion Q
tree
1
(K, ) is proper,
-bounding, preserves the outer measure,

preserves non-meager sets but does not have the Sacks property.]
Chapter 4: Omittory with Halving We explain how omittory creating
pairs with the weak Halving Property produce almost
-bounding forcing no-

tions.
4.1 What omittory may easily do [We show why natural examples of
forcing notions Q
s
(K, ) (for an omittory creating pair (K, )) add a
Cohen real and make ground model reals meager.]
4.2 More operations on weak creatures [Just what the title says: we
present more ways to put weak creatures together.]
4.3 Old reals are unbounded [We say when a creating pair (K, ) is of the
ABtype and we show that this property may be concluded from easier
tocheck properties. We show that Q
s
(K, ) is almost
-bounding if
(K, ) is growing condensed and of the ABtype. For omittory creating
pairs we do not have to assume condensed but then we require a stronger
variant of the AB, AB
+
.]
4.4 Examples [We generalize the forcing notions from [Sh 207], [RoSh 501]
building examples for properties investigated before.]
Chapter 5: Around not adding Cohen reals We try to ensure that the
forcing notions built according to our schema do not add Cohen reals even if it-
erated. We generalize (f, g)bounding and further we arrive to a more general
iterable condition implying no Cohen reals.
5.1 (f, g)bounding [We present easy ways to make sure that our method
results in (f, g)bounding forcing notions.]
5.2 (
t,

T)bounding [We introduce a natural (in our context) generalization
of (f, g)bounding property. For the sake of completeness we show that
the new property is preserved in CS iterations.]
5.3 Quasigeneric and preserving them [We formulate a reasonably
weak but still iterable condition for not adding Cohen reals. We dene
tsystems, we say when is quasi-W-generic and when a forcing notion is

genericity preserving. These notions will be crucial in the next section
too.]
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

ANNOTATED CONTENT vii
5.4 Examples [We construct a sequence W
k
,h
: h T
, k, ) such that W
k
,h
are

tsystems and various forcing notions (including the random alge-
bra) are (, W
k
,h
)genericity preserving (for quasi-generic ). We build a
forcing notion Q
w
(K, ) which is proper,
-bounding, (f, g)-bounding,

makes ground model reals null and we use the technology of genericity
to conclude that its CS iterations with Millers forcing, Lavers forcing and
random algebra do not add Cohen reals.]
Chapter 6: Playing with ultrafilters Our aim here is to build a model
in which there is a p-point generated by
1
elements which is not a q-point and
m
1
=
2
.
6.1 Generating an ultralter [We say when and how quasi-W-generic
determines an ultralter on .]
6.2 Between Ramsey and p-points [We dene semiRamsey and almost
Ramsey ultralters and we have a short look at them.]
6.3 Preserving ultralters [We give conditions on a treecreating pair (K, )
which imply that the forcing notion Q
tree
1
(K, ) preserves T is an ultra-
lter for T which is Ramsey, almost Ramsey. We say when the lter
generated by T in the extension is almost Ramsey.]
6.4 Examples [We construct

tsystems W
n
L
such that if a quasi-W
N
L
-generic
generates a semiRamsey ultralter then it generates an almostRamsey
ultralter, and we build a suitable quasi-generic . For a function
we give a treecreating pair (K, ) such that the forcing notion Q

tree
1
(K, )
preserves T is an almostRamsey ultralter and it adds a function

W
(with

W(m) [(m)]
m + 1
) such that for each partial function h
mdom(h)
(m) innitely often h(m)

W(m). Next we apply it to get an
answer to Matets problem.]
Chapter 7: Friends and relatives of PP We deal with BalcerzakPlewik
number and various properties resembling PPproperty.
7.1 BalcerzakPlewik number [We recall the denition of
BP
and we show
that it is bounded by the dominating number of the relation determined
by the strong PPproperty.]
7.2 An iterable friend of strong PPproperty [We introduce a property
slightly stronger than the strong PPproperty but which can be easily
handled in CS iterations. We show that this property is natural for forc-
ings Q
tree
0
(K, ), Q
w
(K, ) (in nitary cases).]
7.3 Bounded relatives of PP [We dene various PP-like properties for
localizing functions below a given function. We say how one gets them for
our forcing notions and how we may handle them in iterations.]
7.4 Weakly non-reducible p-lters in iterations [We show that a prop-
erty of lters, crucial for getting PP-like properties for our forcing notions,
is easy to preserve in CS iterations.]
7.5 Examples [For a perfect set P 2
we build a creating pair (K, )

such that the forcing notion Q
f
(K, ) is proper,
-bounding and adds

a perfect subset Q of P whith property that (K []
)(Q`K ,= 2
K
).
We use this forcing notion to get consistency of d <
BP
. We show how
forcing notions from other parts of the paper may be used to distinguish
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

viii ANNOTATED CONTENT
PP-like properties (and the corresponding cardinal invariants). We build
an example of a forcing notion which is
bounding and preserves non-

meager sets but which does not have the strong PP-property.]
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

CHAPTER 0
Introduction
Set Theory
1
began with Georg Cantors work when he was studying some spe-
cial sets of reals in connection with the theory of trigonometric series. This study
led Cantor to the following fundamental question: does there exist a bijection be-
tween the natural numbers and the set of real numbers? He answered this question
negatively by showing that there is no such function. Cantors work did not stop
here and with his sharp intuition he discovered new concepts like the alephs scale:
0, 1, . . . ,
0
,
1
, . . . ,
,
+1
, . . . .
Thus Cantors theorem says that
0
< 2
0
and Cantors question was: is 2
0
equal
to
1
?
A real advance on Cantors question was given by Kurt G odel when he showed
that it is (relatively) consistent that 2
0
=
1
. In 1963 Paul Cohen showed that if
the ZFaxioms for Set Theory are consistent then there is a model for Set Theory
where the continuum is bigger than
1
. Cohens work is the end of classical set
theory and is beginning of a new era.
When the cardinality of the continuum is
1
(i.e. CH holds) most of the com-
binatorial problems are solved. When the continuum is at least
3
then most of
the known technology fails and we meet very strong limitations and barriers.
When the continuum is
2
there are many independence results; moreover there
are reasonably well developed techniques for getting them by sewing the countable
support iterations of proper forcing notions together with theorems on preservation
of various properties.
The aim of this paper is to present some tools applicable in the last case.
We present here a technique of constructing of (proper) forcing notions that was
introduced by Shelah for solving problems related to cardinal invariants like the un-
bounded number or the splitting number as well as questions of existence of special
kinds of P-points (see Blass Shelah [BsSh 242] and Shelah [Sh 207], [Sh 326]).
That method was successfully applied in Fremlin Shelah [FrSh 406], Ros lanowski
Shelah [RoSh 501], Ciesielski Shelah [CiSh 653] and other papers. The rst at-
tempt to present a systematic study of the technique was done in the late eighties
when the second author started work on preparation of a new edition of [Sh:b].
For a long time the new book, [Sh:f], was supposed to contain 19 chapters. The
last chapter, Norms on possibilities, contained a series of general denitions and
statements of some basic results. However, there was no new application (or: a
good question to solve) and the author of the book decided to put this chapter
1
A large part of the beginning of this introduction is based on notes of Haim Judah. I really
think that they t to the present paper, though Saharon Shelah is not convinced Andrzej
Ros lanowski.
1
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

2 0. INTRODUCTION
aside. Several years later, when the rst author started his cooperation with She-
lah some new applications of Norms on possibilities appeared. But the real shape
was given to the work due to questions of Tomek Bartoszy nski and Pierre Matet.
The answers were very stimulating for the development of the general method.
This paper is meant as the rst one in a series of works presenting applicability
of the method of norms on possibilities. In [RoSh 670] we will present more
applications of this technique for example we develop the ideas of Ciesielski Shelah
[CiSh 653] to build models without magic sets and their relatives. Though one
can get an impression here that our method results in non-ccc forcing notions, we
managed to generalize it slightly and get a tool for constructing ccc forcing notions.
That was successfully applied in [RoSh 628] to answer a problem of Kunen by
constructing a ccc Borel ideal on 2
which is translationinvariant indexinvariant

and is distinct from the null ideal, the meager ideal and their intersection. It
should be pointed out here, that already in Judah Ros lanowski Shelah [JRSh 373]
an example of a ccc forcing notion built with the use of norms on possibilities
was given (the forcing notion there can be presented as some Q
tree
1
(K, ) in the
terminology here). Investigations of ccc forcing notions constructed according to our
schema are continued in [RoSh 672]. There are serious hopes that the technique
presented here might be used to deal with problems of large continuum due to
special products. This would continue Goldstern Shelah [GoSh 448]. Another
direction is study of -ideals related to forcing notions built according to the schema.
Let us note that most of the forcing notions constructed here fall into the
category of snepforcing notions of [Sh 630]. Consequently, the general machinery
of denable forcing notions is applicable here. We may use it to improve some of
our results, and also to get more tools for handling iterations (see [Sh 630] and
[Sh 669] for more details).
We want to emphasize that though the aim of the paper is strongly related
to independence proofs it should have some value for those rmly committed to
unembellished ZFC, too. This is nicely expressed by the following:
Thesis 0.0.1. We cannot discover the (candidates for) Theorems of ZFC with-
out having good forcing techniques to show they are hard nuts.
0.1. The content of the paper
Most of the results of the paper originated in answering a particular question
by constructing an example of a forcing notion. However, the general idea of the
paper is to extract those properties of the example which are responsible for the
fact that it works, with the hope that it may help in further applications of the
method. That led us to separation of the general theory from its applications, and
caused us to introduce a large number of denitions specifying various properties
of weak creating pairs. Each chapter ends with a section presenting examples and
applications of the tools developed in previous sections. Moreover, at the end of
the paper we present the list of all denitions which appeared in it. This is not a
real index, but should be helpful. The rst chapter introduces basic denitions and
the general scheme. In the next chapter we deal with the fundamental question of
when our forcing notions are proper. The rst two chapters are a basis for the rest
of the paper. After reading them one can jump to any of the following chapters.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

0.2. NOTATION 3
In the third chapter we show how we may control some basic properties of
forcing notions built according to our scheme. The properties we deal with here
are related to measure and category and they lead us to example 3.5.1 of a proper
forcing notion which is
bounding, preserves non-meager sets and the outer

measure, but does not have the Sacks Property. This answers Bartoszy nskis request
[Ba94, Problem 5].
The fourth part continues [RoSh 501], dealing with localizations of subsets of
. Though the example constructed here is a minor modication of the one built
there, it is presented according to our general setting. We show explicitly how the
weak Halving Property works in this type of examples.
A serious problem in getting models of ZFC with given properties of measure
and category is that of not adding Cohen reals. What is disturbing here is that we
do not have any good (meaning: suciently weak but iterable) conditions for this.
In the fth chapter we show how one can ensure that our general scheme results in
forcing notions not adding Cohen reals. A new iterable condition for this appears
here and quite general tools are developed (see 5.3.6, 5.4.2). Finally, in 5.4.3, 5.4.4,
we fully answer another request of Bartoszy nski formulated in [Ba94, Problem 4].
We build a proper
bounding forcing notion which preserves non-meager sets,

makes ground model reals null, is (f, g)bounding and such that countable support
iterations of this forcing with Laver forcing, Miller forcing and random real forcing
do not add Cohen reals.
The next chapter leads to answering a question of Matet and Pawlikowski.
In 6.4.6 we show that it is consistent that there exists a p-point generated by
1
elements which is not a q-point and that for every
and a family T of
1
partial innite functions h : dom(h) such that h(n) < (n) for n dom(h)
there is W

n
[(n)]
n + 1
with (h T)(
n dom(h))(h(n) W(n)).
Several general results on preserving special properties of ultralters are presented
on the way to this solution.
A starting point for chapter 7 was a problem of Balcerzak and Plewik. We
show that the BalcerzakPlewik number
BP
(see 7.1.1) is bounded by a cardinal
invariant related to the strong PPproperty (in 7.1.3). Next we show the consistency
of both d <
BP
(in 7.5.2) and
BP
< c (in 7.5.3). We treat our solution as
a good opportunity to look at various properties of forcing notions related to the
PPproperty (and corresponding cardinal invariants).
0.2. Notation
Most of our notation is standard and compatible with that of classical textbooks
on Set Theory (like Bartoszy nski Judah [BaJu95] or Jech [J]). However in forcing
we keep the convention that a stronger condition is the larger one.
Notation 0.2.1. (1) R
0
stands for the set of non-negative reals. The
integer part of a real r R
0
is denoted by r|.
(2) For two sequences , we write < whenever is a proper initial
segment of , and _ when either < or = . The length of a
sequence is denoted by g().
(3) A tree is a family of nite sequences closed under initial segments. (In
1.3.1 we will dene more general objects.) For a tree T the family of all
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

4 0. INTRODUCTION
branches through T is denoted by [T]. We may use the notation lim(T)
for this object too (see 1.3.1, note that a tree is a quasi tree).
(4) The quantiers (
n) and (
n) are abbreviations for

(m )(n > m) and (m )(n > m),
respectively.
(5) For a function h : X X and an integer k we dene h
(k)
as the k
th
iteration of h: h
(1)
= h, h
(k+1)
= h h
(k)
.
(6) For a set X, [X]
, [X]
<
and T(X) will stand for families of countable,
nite and all, respectively, subsets of the set X. The family of k-element
subsets of X will be denoted by [X]
k
. The set of all nite sequences with
values in X is called X
<
(so domains of elements of X
<
are integers).
The collection of all nite partial functions from to X is X
.
(7) The Cantor space 2
and the Baire space
are the spaces of all functions

from to 2, , respectively, equipped with natural (Polish) topology.
(8) For a forcing notion P,
P
stands for the canonical Pname for the generic
lter in P. With this one exception, all Pnames for objects in the exten-
sion via P will be denoted with a dot above (e.g. ,

X).
(9) c stands for the cardinality of the continuum. The dominating number
(the minimal size of a dominating family in
in the ordering of eventual

dominance) is denoted by d and the unbounded number (the minimal size
of an unbounded family in that order) is called b. /, ^ stand for the
ideals of meager and null sets on the real line, respectively.
General Definitions 0.2.2. (1) For an ideal of subsets of a space X
we dene its cardinal characteristics (called additivity, covering number,
uniformity and conality, respectively):
add() = min[/[ : / &

/ / ,
cov() = min[/[ : / &

/ = X,
non() = min[Y [ : Y X & Y / ,
cof () = min[/[ : / & (A )(B /)(A B).
(2) Assume that X, Y are Polish spaces and R X Y is a Borel relation.
Suppose that V V
are models of ZFC and that all parameters we

need are in V. We say that the extension (V, V
) has the R-localization

property if
(x X V
)(y Y V)((x, y) R).

If x X V
, y Y V and (x, y) R then we say that y R-localizes x.

We say that a forcing notion P has the Rlocalization property if every
generic extension of V via P has this property.
(3) For a relation R XY we dene two cardinal numbers (the unbounded
and the dominating number for R):
b(R) = min[B[ : (y Y )(x B)((x, y) / R)
d(R) = min[D[ : (x X)(y D)((x, y) R).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

CHAPTER 1
Basic denitions
In this chapter we introduce our heroes: forcing notions built of weak creatures.
The Prologue is intended to give the reader some intuitions needed to get through
a long list of denitions. A general scheme is presented in the second part, where
we dene weak creatures, subcomposition operations, weak creating pairs (K, )
and corresponding forcing notions Q
C(nor)
(K, ). However, in practice (at least
in this paper) we will be interested in two special cases of the scheme. The rst
main family of weak creating pairs (and related forcing notions) are creating pairs
determined by composition operations on creatures. The second family consists of
treecreating pairs coming from tree compositions on treecreatures. These two
options are introduced in the following two parts of the chapter. It should be
underlined here that the rest of the paper will deal with these two (essentially
disjoint and parallel) cases of the general scheme. In the last section we give some
justications for our work in the next chapter, showing that without extra care our
schema may result in forcing notions collapsing
1
.
Note: Our terminology (weak creatures, creatures, treecreatures etc) might be
slightly confusing, but it was developed during a long period of time (see introduc-
tion) and large parts of it are established in literature already.
Basic Notation: In this paper H will stand for a function with domain such
that (m )([H(m)[ 2). We usually assume that 0 H(m) (for all m
); if it is not the case then we x an element of H(m) and we use it whenever
appropriate notions refer to 0. Moreover we x a suciently large uncountable
regular cardinal and we assume that at least H H() (the family of sets of
cardinality hereditarily less than ) or, what is more natural, even H H(
1
).
1.0. Prologue
If one looks at forcing notions appearing naturally in the Set Theory of Reals
(i.e. the forcing notions adding a real with certain properties and preserving various
properties of the ground model reals) then one realizes that they often have a com-
mon pattern. A condition in such a forcing notion determines an initial segment of
the real we want to add and it puts some restrictions on possible further extensions
of the initial segment. When we pass to a stronger condition we extend the deter-
mined part of the generic real and we put more restrictions on possible extensions.
But we usually demand that the amount of freedom which is left by the restrictions
still goes to innity in a sense. The basic part of the denition of such a forcing
notion is to describe the way a condition puts a restriction on possible initial seg-
ments of the generic real. Typically the restriction can be described locally by use
of atoms or black boxes, in our terminology called weak creatures. So a weak
creature t has a domain (contained in nite sequences) and for each sequence w
5
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

6 1. BASIC DEFINITIONS
from the domain it gives a family of extensions of w (this is described by a relation
val[t]: if u, v) val[t] then v is an allowed extension of u). Moreover, such a t
has a norm nor[t] which measures the amount of freedom it leaves. Further, we
are told what we are allowed to do with weak creatures: typically we may shrink
them, glue together or just forget about them (i.e. omit them). The results of
permitted operations on a family o of weak creatures are elements of (o) in our
notation (where is a sub-composition operation on the considered family K of
weak creatures), see 1.1.4. Now a condition in our forcing notion can be viewed as
(w, o), where w is a nite sequence (the determined part of the generic real) and
o is a countable family of weak creatures satisfying some demands on its structure
and requirements on nor[t] for t o. When we want to build a stronger condition
then we take t o such that w is in the domain of t and we pick up one of the
possible extensions of w allowed by t. We may repeat this procedure nitely many
times and we get a sequence w
extending w. Next we choose a family o
of weak
creatures such that each s o
is obtained by permitted procedures from some

o
s
o (i.e. s (o
s
)). The pair (w
, o
) is an extension of (w, o) provided o
satises the structure demands and norm requirements.

However, this general schema breaks to two cases, which, though very similar,
are of dierent avors. In the rst case we demand that the family o in a con-
dition (w, o) has a linear structure. Then we usually represent the condition as
(w, t
0
, t
1
, t
2
, . . .), where for some sequence 0 m
0
< m
1
< m
2
< . . . <
w is a sequence of length m
0
and for each i < :
t
i
is a weak creature saying in which way sequences of length m
i
may be extended to sequences of length m
i+1
.
So it is natural in this context to consider only weak creatures t such that for
some integers m
t
dn
< m
t
up
(dn stands for down), the domain of t is contained
in sequences of length m
t
dn
and every extension of a sequence from the domain
allowed by t is of length m
t
up
. In other words we require that if u, v) val[t] then
g(u) = m
t
dn
and g(v) = m
t
up
. In applications the domain of the relation val[t]
consist of all legal sequences of length m
t
dn
. Let us describe a simple example of
this kind. Consider the Silver forcing Q below 2
n
: a condition in Q is a function
p : dom(p) such that
dom(p) & [ dom(p)[ = & (n dom(p))(p(n) < 2
n
).
Let us look at this forcing in a dierent way.
Let K consist of all triples t = (nor[t], val[t], dis[t]) such that
dis[t] = (m
t
, k
t
) where m
t
< and k
t
2
m
t
,
nor[t] = m
t
if k
t
= and nor[t] = 0 otherwise,
val[t] = u, v)

i<m
t
2
i

im
t
2
i
: u < v & (k
t
,= v(m
t
) = k
t
).
For o K we let (o) = if [o[ ,= 1 and
if t K, k
t
,= then (t) = t,
if t K, k
t
= then (t) consists of all s K with m
s
= m
t
.
Now, a condition p in Q may be represented as a sequence (w
p
, t
p
0
, t
p
1
, . . .), where
(a) each t
p
i
is from K,
(b) w
p
is a nite sequence of length m
t
p
0
such that w
p
(n) < 2
n
for n < m
t
p
0
,
(c) m
t
p
i
= m
t
p
0
+i,
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

1.1. WEAK CREATURES AND RELATED FORCING NOTIONS 7
(d) lim sup
i
nor[t
p
i
] = .
The order of Q is dened by (w
p
, t
p
0
, t
p
1
, t
p
2
, . . .) (w
q
, t
q
0
, t
q
1
, t
q
2
, . . .) if and only if
for some N < :
w
p
_ w
q
, g(w
q
) = g(w
p
) +N,
w
q
`g(w
p
) +i, w
q
`(w
p
) +i + 1) val[t
p
i
] for each i < N, and
t
q
j
(t
p
N+j
) for every j < .
The pair (K, ) is an example of a creating pair and the forcing notion Q (repre-
sented as above) is Q
w
(K, ) (see 1.2.2 and 1.2.4).
On the other pole of possible weak creatures we have those which provide
possible extensions for only one sequence (i.e. those t for which [dom(val[t])[ = 1).
Weak creatures of this type are called treecreatures and they say to us simply:
I know what the restrictions on extensions of a single sequence
are, and I do not look at other sequences at all.
Treecreatures are fundamental for building forcing notions in which conditions are
trees of a special kind. In these forcing notions a condition p = (w, o) is such that
w is a root (stem) of a tree T
p
and each t o is a part of the tree T
p
; usually such
a t describes how one passes from an element T
p
to its extensions in T
p
(not
necessarily immediate successors). It is natural to put some requirements on sub
composition operations when the weak creatures we consider are treecreatures,
and this leads to the denition of tree composition and treecreating pair, see 1.3.3.
Moreover, it turns out to be very practical to consider special demands on the
norms nor[t] to take an advantage of the treeform of a condition, see 1.3.5. (Note
that in further denitions we do not require that T
p
is a tree but we demand that
it is a quasi tree only. This will simplify the notation a little bit.) Let us illustrate
this by a suitable representation of the Laver forcing L. Recall that a condition in
L is a tree T
<
such that if T, root(T) _ then [succ
T
()[ = .
Let K consist of all triples t = (nor[t], val[t], dis[t]) such that
dis[t] = (
t
, A
t
), where A
t
[]
and
t

<
,
nor[t] = g(
t
),
val[t] =
t
, ) :
t
< & g() = g(
t
) + 1 & (g(
t
)) A
t
.
For t K we let (t) = s K :
s
=
t
& A
s
A
t
and for o K
with [o[ ,= 1 we declare (o) = . Now, a condition p in L can be represented
as (
p
, t
p
: T
p
)), where T
p

<
is a tree such that root(T
p
) =
p
and
for each T
p
, root(T
p
) _ we have succ
T
p() = : , ) val[t
p
] (so
o is t
p
: T
p
here). Moreover, we demand that for each innite branch
through T
p
the norms nor[t
p
i
] go to innity. The order of L is given by
(
p
, t
p
: T
p
)) (
q
, t
q
: T
q
)) if and only if
q
T
p
and
( T
q
)( T
p
& t
q
(t
p
)).
The pair (K, ) dened above is a tree creating pair and the forcing notion L is
Q
tree
1
(K, ).
1.1. Weak creatures and related forcing notions
Definition 1.1.1. (1) A triple t = (nor, val, dis) is a weak creature for
H if:
(a) nor R
0
,
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

(b) val is a non-empty subset of
x, y)
_
m0<m1<
[
i<m0
H(i)
i<m1
H(i)] : x < y,
(c) dis H().
The family of all weak creatures for H is denoted by WCR[H].
(2) In the above denition we write nor = nor[t], val = val[t] and dis =
dis[t].
[val is for value, nor is for norm, dis is for distinguish.]
Remark 1.1.2. The dis[t] in a weak creature t plays the role of an additional
parameter which allows as to have distinct creatures with the same values of val
and nor. This may be sometimes important in dening sub-composition operations
on K (see 1.1.4 below): we will be able to have distinct values of (t
0
), (t
1
)
though val[t
0
] = val[t
1
] and nor[t
0
] = nor[t
1
]. One may think that this additional
parameter describes the way the weak creature t was constructed (while val[t],
nor[t] give the nal eect of the construction). We may sometimes forget to
mention dis[t] explicitly in most of the results and applications dis[t] might be
arbitrary. In the examples we construct, if we do not mention dis[t] we mean that
either it is 0 or its form is clear.
Definition 1.1.3. (1) If we omit H we mean for some H or the H is
clear from the context, etc.
(2) We say that His nitary (or of a countable character, respectively) if H(n)
is nite (countable, resp.) for each n . We say that K WCR[H] is
nitary if H is nitary and val[t] is nite for each t K.
Definition 1.1.4. Let K WCR[H].
(1) A function : [K]
T(K) is a sub-composition operation on K if:

(a) (transitivity) if o [K]
and for each s o we have s (o

s
)
then (o) (
sS
o
s
),
(b) r (r) for each r K and () = .
[Note that (o) may be empty for non-empty o; in future dening we
will describe it only for the cases it provides a non-empty result, in all
other cases we will assume that (o) = .]
(2) In the situation described above the pair (K, ) is called a weak creating
pair for H.
(3) Suppose that (K, ) is a weak creating pair, t
0
, t
1
K. We say that
t
0
, t
1
are equivalent (and we write then t
0

t
1
) if nor[t
0
] = nor[t
1
],
val[t
0
] = val[t
1
] and for each o [Kt
0
, t
1
]

we have (o t
0
) =
(o t
1
).
Remark 1.1.5. Note that the relation
as dened in 1.1.4(3) does not have

to be transitive in a general case (so, perhaps, we should not use the name
equivalent). However, if (K, ) is either a creating pair (see 1.2.2) or a treecreating
pair (see 1.3.3) then
is an equivalence relation. Then, if additionally (o) is non-

empty for nite o only, the value of (o) depends on the
equivalence classes
of elements of o only. Therefore we will tend to think in these situations that we
identify all
equivalent elements of K (or just consider a selector K
K of
K/
). If (o) may be non-empty for an innite o K (which may happen

4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

for treecreating pairs), then we have to be more careful before we consider this
identication: we should check that the values of depend on
equivalence
classes only.
Definition 1.1.6. Let (K, ) be a weak creating pair for H.
(1) For a weak creature t K we dene its basis (with respect to (K, )) as
basis(t) = w
_
m<
i<m
H(i) : (s (t))(u)(w, u) val[s]).
(2) For w

m<
i<m
H(i) and o [K]
we dene the set pos(w, o) of

possible extensions of w from the point of view of o (with respect to
(K, )) as:
pos
(w, o) = u : (s (o))(w, u) val[s]),

pos(w, o) = u : there are disjoint sets o
i
(for i < m < ) with

i<m
o
i
= o
and a sequence 0 <
0
< . . . <
m1
< g(u) such that
u`
0
pos
(w, o
0
) and
u`
1
pos
(u`
0
, o
1
) & . . . & u pos
(u`
m1
, o
m1
).
(3) Whenever we use basis or pos we assume that the weak creating pair
(K, ) with respect to which these notions are dened is understood.
Definition 1.1.7. Suppose (K, ) is a weak creating pair for H and ((nor) is
a property of -sequences of weak creatures from K (i.e. ((nor) can be thought of
as a subset of K
). We dene the forcing notion Q

C(nor)
(K, ) by
conditions are pairs (w, T) such that for some k
0
< :
(a) w
i<k0
H(i)
(b) T = t
i
: i < ) where:
(i) t
i
K for each i,
(ii) w basis(t
i
) for some i < and for each u pos(w, t
i
: i I
0
),
I
0
there is i I
0
such that u basis(t
i
),
(c) the sequence t
i
: i < ) satises the condition ((nor);
the order is given by: (w
1
, T
1
) (w
2
, T
2
) if and only if
for some disjoint sets o
0
, o
1
, o
2
, . . . we have:
w
2
pos(w
1
, t
1
: o
0
) and t
2
i
(t
1
: o
i+1
) for each i <
(where T
= t
i
: i < )).
If p = (w, T) we let w
p
= w, T
p
= T and if T
p
= t
i
: i < ) then we let
t
p
i
= t
i
. We may write (w, t
0
, t
1
, . . .) instead of (w, T) (when T = t
i
: i < )).
Proposition 1.1.8. If (K, ) is a weak creating pair and ((nor) is a property
of sequences of elements of K then Q
C(nor)
(K, ) is a forcing notion.
Remark 1.1.9. The reason for our notation ((nor) for the property relevant
for (c) of 1.1.7 is that in the applications this conditions will say that the norms
nor[t
i
] go to the innity in some sense. Some of the possibilities here are listed in
1.1.10 below.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Definition 1.1.10. For a weak creature t let us denote
m
dn
(t) = ming(u) : u dom(val[t]).
We introduce the following (basic) properties of sequences of weak creatures which
may serve as ((nor) in 1.1.7:
(s) A sequence t
i
: i < ) satises (
s
(nor) if and only if
(i < )(nor[t
i
] > maxi, m
dn
(t
i
)).
() A sequence t
i
: i < ) satises (

lim
i
nor[t
i
] = .
(w) A sequence t
i
: i < ) satises (
w
lim sup
i
nor[t
i
] = .
Let f : . We dene the property introduced by f by
(f) A sequence t
i
: i < ) satises (
f
(k < )(
i)(nor[t
i
] > f(k, m
dn
(t
i
))).
For notational convenience we will sometimes use the empty norm condition:
() Each sequence t
i
: i < ) satises (
(nor).
The forcing notions corresponding to the above properties (for a weak creating
pair (K, )) will be denoted by Q
s
(K, ), Q
(K, ), Q
w
(K, ), Q
f
(K, ) and
Q
(K, ), respectively.
Remark 1.1.11. 1) Note that the second component of a pair (w, T)
Q
C(nor)
(K, ) is a sequence of weak creatures, and in the most general case the order
of its members may be important. For example the property (
s
(nor) introduced
in 1.1.10 is not permutation invariant and some changes of the order in the sequence
t
i
: i < ) may produce a pair (w, T
) which is not a legal condition. This is not

what we would like to have here, so in applications in which this kind of problems
appears we will restrict ourselves to suborders Q
C(nor)
(K, ) of Q
C(nor)
(K, ) in
which we put additional structure demands on the sequences t
i
: i < ) (see 1.2.6).
Moreover, to get properness for forcing notions Q
s
(K, ) we will have to put some
demands on (K, ) (see 2.1.6). These demands will cause that various variants of
the norm condition (
s
(nor) result in equivalent forcing notions (see 2.1.3). So,
from the point of view of applications, the main reason for introducing (
s
(nor)
is a notational convenience.
2) Note that
Q
s
(K, ) Q
(K, ) Q
w
(K, ) Q
(K, )
where the inclusions mean suborder (but often not complete suborder). If
we put some conditions on f (e.g. f is fast, see 1.1.12) then we may easily have
Q
f
(K, ) Q
(K, ).
3) In our applications we will consider the forcing notions Q
f
(K, ) only for
functions f : which are growing fast enough (see 1.1.12 below).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Definition 1.1.12. A function f : is fast if
(k )
_
)(f(k, ) f(k, + 1) & 2 f(k, ) < f(k + 1, )
_
.
The function f is H-fast if additionally (H is nitary and) for each k, :
2
H()
(f(k, ) +
H
() + 2) < f(k + 1, ),
where
H
() = [
i<
H(i)[.
Definition 1.1.13. Suppose that (K, ) is a weak creating pair and ((nor)
is a property of sequences of elements of K. Let

W be a Q
C(nor)
(K, )-name such
that
'
Q
C(nor)
(K,)

W =
_
w
p
: p
Q
C(nor)
(K,)
.
Proposition 1.1.14. Suppose that (K, ) is a weak creating pair and ((nor)
is a property such that the forcing notion Q
C(nor)
(K, ) is non-empty. Then:
(1) '
Q
C(nor)
(K,)

W is a member of

i<
H(i).
(2) If (i )(H(i) = 2) then '
Q
C(nor)
(K,)

W is a real.
(3) If for every t K, u basis(t) the set pos(u, t) has at least two elements
then '
Q
C(nor)
(K,)

W / V.
Remark 1.1.15. 1) We will always assume that the considered weak creating
pairs (K, ) (and norm conditions ((nor)) are such that Q
C(nor)
(K, ) ,= . Usu-
ally, it will be enough that K contains enough creatures with large norms and in
each particular example this requirement will be easy to verify.
2) In general, the

W dened in 1.1.13 does not have to encode the generic l-
ter. We may formulate a condition ensuring this. Let (K, ) be a weak creating
pair and ((nor) be a norm condition such that Q
C(nor)
(K, ) is not empty. For
p Q
C(nor)
(K, ) dene
S(p)
def
= w
_
n
i<n
H(i) : (q p)(w _ w
q
)
Clearly S(p) is a subtree of

n
i<n
H(i). Moreover, for each w
i<n
H(i) and
p, q Q
C(nor)
(K, ):
p ,'
Q
C(nor)
(K,)
w

W if and only if w S(p) and
p '
Q
C(nor)
(K,)

W [S(q)] if and only if S(p) S(q).
Now we may dene a Q
C(nor)
(K, )name

H by
'
Q
C(nor)(K,)
H = p Q
C(nor)
(K, ) :

W [S(p)]
and we may want to claim that '

H =
Q
C(nor)
(K,)
. But for this we need to
know that any two conditions in

H are compatible. A sucient and necessary
requirement for this is:
() if p, q Q
C(nor)
(K, ) and S(p) S(q)
then p ' q
Q
C(nor)
(K,)
(or in other words p q modulo the
equivalence of conditions).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

In most of our examples and applications, the condition () will be easy to check.
We will not mention it in future as we will not use its consequences.
Note however that it is very easy to build examples of weak creating pairs (K, )
(even creating pairs or tree creating pairs) for which () fails. Some of these
examples might appear naturally.
1.2. Creatures
Now we will deal with the rst specic case of the general scheme: creating
pairs and forcing notions Q
C(nor)
(K, ). Notation and denitions introduced here
are applicable to this case only and should not be confused with that for tree
creating pairs.
Definition 1.2.1. Let t be a weak creature for H.
(1) If there is m < such that (u, v) val[t])(g(u) = m)
then this unique m is called m
t
dn
.
(2) If there is m < such that (u, v) val[t])(g(v) = m)
then this unique m is called m
t
up
.
(3) If both m
t
dn
and m
t
up
are dened then t is called an (m
t
dn
, m
t
up
)creature
(or just a creature).
(4) CR
m
dn
,mup
[H] = t WCR[H] : m
t
dn
= m
dn
and m
t
up
= m
up
,
CR[H] =

m
dn
<mup<
CR
m
dn
,mup
[H].
Definition 1.2.2. Suppose that K CR[H] and is a sub-composition op-
eration on K. We say that is a composition on K (and we say that (K, ) is a
creating pair for H) if:
(1) if o [K]

and (o) ,= then o is nite and for some enumeration
o = t
0
, . . . , t
m1
we have m
ti
up
= m
ti+1
dn
for all i < m1, and
(2) for each s (t
0
, . . . , t
m1
) we have m
s
dn
= m
t0
dn
and m
s
up
= m
tm1
up
.
In this paper we will always assume that the creating pair under considerations is
additionally nice and smooth (see 1.2.5 below) and we will not repeat this demand
later.
Remark 1.2.3. Sets of creatures with pairwise distinct m
t
dn
s might be natu-
rally ordered according to this value and therefore in similar situations we identify
sets of creatures with the corresponding sequences of creatures.
Definition 1.2.4. (1) For K CR[H] and a composition operation
on K we dene nite candidates (FC) and pure nite candidates (PFC)
with respect to (K, ):
FC(K, ) = (w, t
0
, . . . , t
n
) : w basis(t
0
) and for each i n
t
i
K, m
ti
up
= m
ti+1
dn
and pos(w, t
0
, . . . , t
i
) basis(t
i+1
),
PFC(K, ) = (t
0
, . . . , t
n
) : (w basis(t
0
))((w, t
0
, . . . , t
n
) FC(K)).
(2) We have a natural partial order on FC(K, ) (like in 1.1.7). The partial
order on PFC is dened by
(t
0
, . . . , t
n1
) (s
0
, . . . , s
m1
) if and only if m
tn1
up
= m
sm1
up
, and
(w basis(t
0
))
_
w basis(s
0
) and (w, t
0
, . . . , t
n1
) (w, s
0
, . . . , s
m1
)
_
(so (t
0
) (s
0
) means that s
0
(t
0
) and basis(t
0
) basis(s
0
)).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

1.2. CREATURES 13
(3) A sequence t
0
, t
1
, t
2
, . . .) of creatures from K is a pure candidate with
respect to a creating pair (K, ) if
(i < )(m
ti
up
= m
ti+1
dn
) and
(w basis(t
0
))(i < )(pos(w, t
0
, . . . , t
i
) basis(t
i+1
)).
The set of pure candidates with respect to (K, ) is denoted by PC(K, ).
The partial order on PC(K, ) is dened naturally.
(4) For a norm condition ((nor) the family of ((nor)-normed pure candidates
is
PC
C(nor)
(K, )
def
= t
0
, t
1
, . . .) PC(K, ) : t
0
, t
1
, . . . , ) satises ((nor).
Definition 1.2.5. Let (K, ) be a creating pair for H. We say that
(1) (K, ) is nice if for all t
0
, . . . , t
n1
K and s (t
0
, . . . , t
n1
) we have
basis(t
0
) basis(s).
(2) (K, ) is smooth provided that:
if (w, t
0
, . . . , t
n1
) FC(K, ), m < n and u pos(w, t
0
, . . . , t
n1
)
then u ` m
tm
dn
pos(w, t
0
, . . . , t
m1
) and u pos(u ` m
tm
dn
, t
m
, . . . , t
n1
).
(3) K is forgetful if for every creature t K we have:
[w, u) val[t] & w
n<m
t
dn
H(n)] w
, w
u ` [m
t
dn
, m
t
up
)) val[t].
(4) K is full if dom(val[t]) =

n<m
t
dn
H(n) for every t K.
As we said in 1.2.2, we will always demand that a creating pair is nice and
smooth (but these properties occur naturally in applications). The main reason for
the rst assumption is to have the eect presented in 1.2.8(2) below and the second
demand is to get the conclusion of 1.2.10. Before we state these observations let us
modify a little bit the forcing notions we are interested in.
Definition 1.2.6. Let (K, ) be a creating pair and ((nor) be a property
of -sequences of creatures. The forcing notion Q
C(nor)
(K, ) is a suborder of
Q
C(nor)
(K, ) consisting of these conditions (w, t
0
, t
1
, . . . ) for which additionally
(
1.2.6
) (i < )(m
ti
up
= m
ti+1
dn
).
Remark 1.2.7. 1) The forcing notions introduced in 1.2.6 t better to the
idea of creatures and compositions on them. Moreover in most of the applications
the forcing notions Q
C(nor)
(K, ) and Q
C(nor)
(K, ) will be equivalent. Even in the
most general case they are not so far from each other; note that if p Q
C(nor)
(K, )
and q Q
C(nor)
(K, ), p q (in Q
C(nor)
(K, )) then q Q
C(nor)
(K, ). Of course
it may happen that Q
C(nor)
(K, ) is trivial this usually suggests that the tree
approach is more suitable (see 1.3.3).
2) Several notions simplify for the forcing notions Q
C(nor)
(K, ). For example if
t
0
, . . . , t
n1
K are such that m = m
t0
dn
, m
ti
up
= m
ti+1
dn
and w
i<m
H(i) then
pos(w, t
0
, . . . , t
n1
) = u : for some 0k
1
<. . . <k
<n 1 we have
u ` m
t
k
1
up
pos
(w, t
0
, . . . , t
k1
) &
u ` m
t
k
2
up
pos
(u ` m
t
k
1
up
, t
k1+1
, . . . , t
k2
) & . . . &
u pos
(u ` m
t
k
up
, t
k
+1
, . . . , t
n1
).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

3) The norm condition (s) (see 1.1.10) can be presented slightly simpler for
Q
s
(K, ). For (w, t
0
, t
1
, . . .) Q
s
(K, ) it says just that
(i < )(nor[t
i
] > m
ti
dn
).
Proposition 1.2.8. Suppose (K, ) is a creating pair for H.
(1) Assume that K is full. Then (K, ) is nice and if ((nor) is one of the
conditions (s), (), or (f) (where f is any fast function), then the
forcing notion Q
C(nor)
(K, ) is a dense subset of Q
C(nor)
(K, ).
(2) If (K, ) is nice, (w, t
0
, t
1
, t
2
, . . .) is a condition in Q
(K, ) and s
n
:
n ) is such that for some 0 = k
0
< k
1
< . . . < , s
n
(t
i
: k
n
i <
k
n+1
) (for all n ) then (w, s
0
, s
1
, s
2
, . . .) is a condition in Q
(K, )
(stronger than (w, t
0
, t
1
, t
2
, . . .)).
(3) If (K, ) is forgetful then it is full.
Definition 1.2.9. Let (K, ) be a creating pair, ((nor) be a norm condition,
p Q
C(nor)
(K, ) and be a Q
C(nor)
(K, )-name for an ordinal. We say that
(1) p essentially decides the name if
(m )(u pos(w
p
, t
p
0
, . . . , t
p
m1
))((u, t
p
m
, t
p
m+1
, . . .) decides the value of ),
(2) p approximates at n (or at t
p
n
) whenever:
for each w
1
pos(w
p
, t
p
0
, . . . , t
p
n1
), if there is a condition r Q
C(nor)
(K, )
stronger than p and such that w
r
= w
1
and r decides the value of then
the condition (w
1
, t
p
n
, t
p
n+1
, . . .) decides the value of .
Lemma 1.2.10. Suppose that (K, ) is a smooth creating pair, ((nor) is a norm
condition and is a Q
C(nor)
(K, )name for an ordinal. Assume that a condition
p Q
C(nor)
(K, ) essentially decides (approximates at each n, respectively).
Then each q p essentially decides (approximates at each n, respectively).
Proof. Immediate by smoothness.
Definition 1.2.11. Let (K, ) be a creating pair for H.
(1) For a property ((nor) of -sequences of creatures from K and conditions
p, q Q
C(nor)
(K, ) we dene
p
apr
q (in Q
C(nor)
(K, )) if and only if
p q and for some k we have (i < )(t
p
i+k
= t
q
i
)
(so then w
q
pos(w
p
, t
p
0
, . . . , t
p
k1
) too).
(2) We dene relations
s
n
(for n < ) on Q
s
(K, ) by:
() p
s
0
q (in Q
s
(K, )) if p q and w
p
= w
q
,
() p
s
n+1
q (in Q
s
(K, )) if p
s
0
q and t
p
i
= t
q
i
for i < n + 1.
(3) Relations
n
on Q
(K, ) (for n < ) are dened by:

() p
0
q (in Q
(K, )) if p q and w
p
= w
q
,
(
) p
n+1
q (in Q
(K, )) if p
0
q and
t
p
j
= t
q
j
for all j mini < : nor[t
p
i
] > n + 1 and
t
q
i
: i < & nor[t
q
i
] n + 1 t
p
i
: i < .
(4) Relations
w
n
on Q
w
(K, ) are dened by
() p
w
0
q (in Q
w
(K, )) if p q and w
p
= w
q
,
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

1.3. TREE CREATURES AND TREELIKE FORCING NOTIONS 15
(
+
) p
w
n+1
q (in Q
w
(K, )) if p
w
0
q and
t
p
j
= t
q
j
p
i
] > n + 1.
(5) Let f : be a fast function. Relations
f
n
on Q
f
(K, ) are
dened by:
() p
f
0
q (in Q
(K, )) if p q and w
p
= w
q
,
(
f
) p
f
n+1
q (in Q
f
(K, )) if p
f
0
q and
t
p
j
= t
q
j
p
i
] > f(n + 1, m
t
p
i
dn
) and
t
q
i
: i < & nor[t
q
i
] f(n + 1, m
t
q
i
dn
) t
p
i
: i < .
(6) We may omit superscripts in
s
n
,
n
,
w
n
and
f
n
if it is clear from
the context in which forcing notion we are working (i.e. what is the norm
condition we deal with).
Remark 1.2.12. The dierence between e.g. (3) and (4) is in the last condition
of (3), of course.
Proposition 1.2.13. Suppose (K, ) is a creating pair for H. Let ((nor)
be one of the following properties of -sequences: (
s
(nor), (
(nor), (
w
(nor)
or (
f
(nor) for some fast function f (see 1.1.10) and let
n
be the corresponding
relations (dened in 1.2.11). Then
(1)
apr
is a partial order (stronger than ) on Q
C(nor)
(K, ).
(2)
n
(with superscripts) are partial orders (stronger than ) on the respec-
tive Q
C(nor)
(K, ) and p
n+1
q implies p
n
q.
(3) Suppose that p
n
Q
C(nor)
(K, ) (for n ) are such that
(n )(p
n

n+1
p
n+1
).
Then the naturally dened limit condition p = lim
n
p
n
satises:
p Q
C(nor)
(K, ) and (n < )(p
n

n+1
p).
Remark 1.2.14. A natural property one could ask for in the context of creating
pairs is some kind of monotonicity:
basis(t) = dom(val[t]) and pos(u, t) = v : u, v) val[t],
for t K and u basis(t). However, there is no real need for it, as all our demands
and assumptions on creating pairs will refer to pos (and not val). But for tree
creating pairs we will postulate the respective demand, mainly to simplify notation
(and have explicit treerepresentations of conditions), see 1.3.3(3).
1.3. Tree creatures and treelike forcing notions
Here we introduce the second option for our general scheme: forcing notions
in which conditions are trees with norms. This case, though parallel to the one
of creating pairs, is of dierent character and therefore we reformulate all general
denitions for this particular context.
Definition 1.3.1. (1) A quasi tree is a set T of nite sequences with the
<-smallest element denoted by root(T).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

(2) A quasi tree T is a tree if it is closed under initial segments. If T is a quasi
tree then dcl(T) is the smallest tree containing T (the downward closure
of T).
(3) For a quasi tree T and T we dene the successors of in T, the
restriction of T to , the splitting points of T and maximal points of T by:
succ
T
() = T : < & ( T)( < < ),
T
[]
= T : _ ,
split(T) = T : [succ
T
()[ 2,
max(T) = T : there is no T such that < .
We put

T = T max(T).
(4) The set of all limit innite branches through a quasi tree T is
lim(T)
def
= : is an sequence and (
n)(`n T).
The quasi tree T is well founded if lim(T) = .
(5) A subset F of a quasi tree T is a front in T if no two distinct members of
F are <-comparable and
( lim(T) max(T))(n )( ` n F).
Remark 1.3.2. Note the dierence between lim(T) and lim(dcl(T)) for a quasi
tree T. In particular, it is possible that a quasi tree T is well-founded but there is
an innite branch through dcl(T). Moreover, a front in T does not have to be a
front in dcl(T).
Definition 1.3.3. (1) A weak creature t WCR[H] is a treecreature if
dom(val[t]) is a singleton and no two distinct elements of rng(val[t])
are <comparable;
TCR[H] is the family of all treecreatures for H.
(2) TCR
[H] = t TCR[H] : dom(val[t]) = .

(3) A sub-composition operation on K TCR[H] is a tree composition
(and then (K, ) is called a treecreating pair (for H)) if:
(a) if o [K]
, (o) ,= then o = s
:

T for some well founded
quasi tree T

n<
i<n
H(i) and a system s
:

T) K such that
for each nite sequence

T
s
TCR
[H] and rng(val[s
]) = succ
T
(),
and
(b) if t (s
:

T) then t TCR
root(T)
[H] and rng(val[t])
max(T).
If

T = root(T), t = t
root(T)
TCR
root(T)
[H] and rng(val[t]) = max(T)
then we will write (t) instead of (t
:

T).
(4) A tree-composition on K is bounded if for each t (s
:

T) we
have
nor[t] maxnor[s
] : ( rng(val[t]))( < ).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Remark 1.3.4. 1) Note that sets of tree creatures relevant for tree composi-
tions have a natural structure: we identify here o with s
(s)
: s o where (s)
is such that s TCR
(s)
and s
(s)
= s.
2) To check consistency of our notation for tree creatures with that of 1.1.7 note
that in 1.3.3(3), if s
(s
,
:

T
) for each

T, T is a well founded quasi
tree as in (3)(a) of 1.3.3 then T
def
=

T
T
is a well founded quasi tree,

T
and s
,
:

T,

T
) is a system for which may be non-empty, i.e. it satises

the requirements of 1.3.3(3)(a).
3) Note that if (K, ) is a treecreating pair for H, t TCR
[H] then basis(t) =

and pos(, t) = rng(val[t]) (see 1.1.6). For this reason we will write pos(t) for
pos(, t) and rng(val[t]) in the context of treecreating pairs.
4) Treecreating pairs have the properties corresponding to the niceness and
smoothness of creating pairs (see 1.2.5, compare with 1.3.9).
When dealing with treecreating pairs it seems to be more natural to consider
both very special norm conditions and some restrictions on conditions of the forcing
notions we consider. The second is not very serious: the forcing notions Q
tree
e
(K, )
(for e < 5) introduced in 1.3.5 below are dense subsets of the general forcing
notions Q
C(nor)
(K, ) (for suitable conditions ((nor)). We write the denition of
Q
tree
e
(K, ) fully, not referring the reader to 1.1.7, to show explicitly the way tree
creating pairs work.
Definition 1.3.5. Let (K, ) be a treecreating pair for H.
(1) We dene the forcing notion Q
tree
1
(K, ) by letting:
conditions be sequences p = t
: T) such that
(a) T

n
i<n
H(i) is a non-empty quasi tree with max(T) = ,
(b) t
TCR
[H] K and pos(t
) = succ
T
() (see 1.3.4(3)),
(c)
1
for every lim(T) we have:
the sequence nor[t
k
] : k < , `k T) diverges to innity;
the order be given by:
t
1
: T
1
) t
2
: T
2
) if and only if
T
2
T
1
and for each T
2
there is a well founded quasi tree T
0,

(T
1
)
[]
such that t
2
(t
1
:

T
0,
).
If p = t
: T) then we write root(p) = root(T), T

p
= T, t
p
= t
etc.
(2) Similarly we dene forcing notions Q
tree
e
(K, ) for e = 0, 2, 3, 4 replacing
the condition (c)
1
by (c)
e
respectively, where:
(c)
0
for every lim(T):
limsupnor[t
k
] : k < , `k T) = ,
(c)
2
for every T and n < there is such that < T and
nor[t
] n,
(c)
3
for every T and n < there is such that < T and
( T)( < nor[t
] n),
(c)
4
for every n < , the set
T : ( T)( < nor[t
] n)
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

contains a front of the quasi tree T.
(3) If p Q
tree
e
(K, ) then we let p
[]
= t
p
: (T
p
)
[]
) for T
p
.
(4) For the sake of notational convenience we dene partial order Q
tree
(K, )
in the same manner as Q
tree
e
(K, ) above but we omit the requirement
(c)
e
(like in 1.1.10; so this is essentially Q
(K, )).
Remark 1.3.6. 1) In the denition above we do not follow exactly the notation
of 1.1.7: we omit the rst part w
p
of a condition p as it can be clearly read from the
rest of the condition. Of course the missing item is root(p). In this new notation
the name

W of 1.1.13 may be dened by
'
Q
tree
e
(K,)

W =
_
root(p) : p
Q
tree
e
(K,)
.
2) Note that
Q
tree
4
(K, ) Q
tree
1
(K, ) Q
tree
0
(K, ) Q
tree
2
(K, ) and
Q
tree
1
(K, ) Q
tree
3
(K, ) Q
tree
2
(K, )
but in general these inclusions do not mean complete suborders. If the tree
creating pair is t-omittory (see 2.3.4) then Q
tree
4
(K, ) is dense in Q
tree
2
(K, ) and
thus all these forcing notions are equivalent. If (K, ) is

2big (see 2.3.2) then
Q
tree
4
(K, ) is dense in Q
tree
1
(K, ) (see 2.3.12).
Let us give two simple examples of treecreating pairs.
Let H(i) = (for i ).
Let K
0
TCR[H] consists of these treecreatures s that if s TCR
[H] then
rng(val[s])
k) : k and
nor[s] =
_
g() if val[s] is innite,
0 otherwise.
The operation
0
gives non-empty values for singletons only; for s K
0
we let
0
(s) = t K
0
: val[t] val[s] (an operation dened in this manner will be
further called trivial ). Clearly (K
0
,
0
) is a treecreating pair. Note that:
(a) the forcing notions Q
tree
2
(K
0
,
0
) and Q
tree
0
(K
0
,
0
) are equivalent to Millers
Rational Perfect Set Forcing;
(b) the forcing notions Q
tree
1
(K
0
,
0
), Q
tree
3
(K
0
,
0
), Q
tree
4
(K
0
,
0
) are equiv-
alent to the Laver forcing
(thus Q
tree
1
(K
0
,
0
) is not a complete suborder of Q
tree
0
(K
0
,
0
), and Q
tree
3
(K
0
,
0
)
is not a complete suborder of Q
tree
2
(K
0
,
0
)).
Let us modify the norms on the treecreatures a little. For this we dene a
function f :
<
by
f()) = 0, f(
k)) =
_
f() + 1 if k = 0
f() otherwise.
Now, let K
1
consist of tree creatures s TCR[H] such that rng(val[s])
k) :
k (where s TCR
[H]) and
nor[s] =
_
f() if val[s] is innite,
0 otherwise.
Let
1
be the trivial tree-composition on K
1
, so it is nonempty for singletons only
and then
1
(s) = t K
1
: val[t] val[s] (clearly (K
1
,
1
) is a tree creating
pair). Then
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

(a) the forcing notion Q
tree
0
(K
1
,
1
) is a dense suborder of Q
tree
2
(K
1
,
1
),
(b) the partial orders Q
tree
1
(K
1
,
1
) and Q
tree
4
(K
1
,
1
) are empty, but
(c) Q
tree
3
(K
1
,
1
) is nottrivial (it adds a new real) and it is not a complete
suborder of Q
tree
2
(K
1
,
1
) (e.g. incompatibility is not preserved) and it is
disjoint from Q
tree
0
(K
1
,
1
).
Definition 1.3.7. Let (K, ) be a tree creating pair, e < 5, p Q
tree
e
(K, ).
A set A T
p
is called an e-thick antichain (or just a thick antichain) if it is an
antichain in (T
p
, <) and for every condition q Q
tree
e
(K, ) stronger than p the
intersection A dcl(T
q
) is non-empty.
Proposition 1.3.8. Suppose that (K, ) is a treecreating pair for H, e < 5,
p Q
tree
e
(K, ) and T
p
. Then:
(1) Q
tree
e
(K, ) is a partial order.
(2) Each e-thick antichain in T
p
is a maximal antichain. Every front of T
p
is an e-thick antichain in T
p
.
(3) If e 1, 3, 4, n < then the set
B
n
(p)
def
= T
p
: (i) ( T
p
)( _ nor[t
] > n) but
(ii) no
< ,
T
p
satises (i)
is a maximal <-antichain in T
p
. If e = 4 then B
n
(p) is a front of T
p
.
(4) For every m, n < the set
F
m
n
(p)
def
= T
p
: (i) nor[t
] > n and
(ii) [
T
p
:
< & nor[t
] > n[ = m
is a maximal <-antichain of T
p
. If e 0, 1, 4 then F
m
n
(p) is a front of
T
p
.
(5) If K is nitary (so [val[t][ < for t K, see 1.1.3) then every front of
T
p
is a front of dcl(T
p
) and hence it is nite.
(6) If is bounded then each F
m
n
(p) is a thick antichain of T
p
.
(7) p p
[]
Q
tree
e
(K, ) and root(p
[]
) = .
Remark 1.3.9. One of the useful properties of treecreating pairs (K, ) and
forcing notions Q
tree
e
(K, ) is the following:
()
1.3.9
Suppose that p, q Q
tree
e
(K, ), p q (so in particular T
q
T
p
), T
q
and < , T
p
.
Then p
[]
q
[]
.
Definition 1.3.10. Let p, q Q
tree
e
(K, ), e < 3 (and (K, ) a treecreating
pair). We dene relations
e
n
for n by:
(1) If e 0, 2 then:
p
e
0
q (in Q
tree
e
(K, )) if p q and root(p) = root(q),
p
e
n+1
q (in Q
tree
e
(K, )) if p
e
0
q and
if F
0
n
(p) (see 1.3.8(4)) and T
p
, _ then T
q
and t
q
= t
p
.
(2) The relations
1
n
(on Q
tree
1
(K, )) are dened by:
p
1
0
q (in Q
tree
1
(K, )) if and only if p q and root(p) = root(q),
p
1
n+1
q (in Q
tree
1
(K, )) if p
1
0
q and
if F
0
n
(p) (see 1.3.8(4)) and T
p
, _ then T
q
, t
p
= t
q
,
and
t
q
: T
q
& nor[t
q
] n t
p
: T
p
.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

(3) We may omit the superscript e in
e
n
if it is clear in which of the forcing
notions Q
tree
e
(K, ) we are working.
Proposition 1.3.11. Let (K, ) be a treecreating pair for H, e < 3.
(1) The relations
e
n
are partial orders on Q
tree
e
(K, ) stronger than . The
partial order
e
n+1
is stronger than
e
n
.
(2) Suppose that conditions p
n
Q
tree
e
(K, ) are such that p
n

e
n+1
p
n+1
.
Then lim
n
p
n
Q
tree
e
(K, ) and (n )(p
n

e
n+1
lim
n
p
n
) (where the
limit condition p = lim
n
p
n
is dened naturally; T
p
=

n
T
pn
).
1.4. Non proper examples
In the next chapter we will see that if one combines a norm condition with suit-
able properties of a weak creating pair then the resulting forcing notion is proper.
In particular we will see that (with the norm conditions dened in 1.3.5) getting
properness in the case of treecreating pairs is relatively easy. Here, however,
we show that one cannot expect a general theorem like Q
tree
C(nor)
(K, ) is always
proper and that we should be always careful a little bit. The forcing notions re-
sulting from our general schema may collapse
1
! For example, looking at the norm
conditions introduced in 1.3.5(2) one could try to consider the following condition
(c)
5
(k )(
n)( T
p
)(g() n nor[t
] k).
If a creating pair (K, ) is nitary then, clearly, the forcing notions Q
tree
5
(K, )
and Q
tree
4
(K, ) are the same.
The forcing notion Q
tree
5
(K, ) might be even not proper. The following exam-
ple shows this bad phenomenon which may be made quite general.
Example 1.4.1. Let H(i) = for i .
There is a tree creating pair (K
1.4.1
,
1.4.1
) for H which is simple (see 2.1.7) and
tree
5
(K
1.4.1
,
1.4.1
) is not proper (and collapses c onto ).
Construction. To dene the family K
1.4.1
of tree creatures for H choose
families ,= S
[]
and functions h
: R
(for
<
, 0 <
g()) such that for every , we have:
() R
, h
() = g() + 1,
() if F S
then h
(F) = , each S
is innite,
() if F
0
, F
1
S
, F
0
,= F
1
then F
0
F
1
= ,
() if A R
, h
(A) + 1 then for each F S
A F R
and h
(A F) = ,
() if A
0
, A
1
R
, A
0
A
1
then h
(A
0
) h
(A
1
).
There are several possibilities to construct S
, R
, h
as above. One can do it for

example in the following way. Fix
n
. Take a system K
:
n
of
innite subsets of such that
a) K
= ,
b)
0
<
1

n
K
1
K
0
,
c)
0
,
1

& n &
0
,=
1
K
0
K
1
= ,
d)
<n
K
=

m
K
m
.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

1.4. NON PROPER EXAMPLES 21
Now put R
I
K
: ,= I
n
. For A R
we declare that
e) h
(A) 1, h
() = h
(K
) = n + 1,
f) if for some
n
, n the set A contains the set K
then h
(A)
+ 1,
g) if A
0
A
1
, A
0
, A
1
R
then h
(A
0
) h
(A
1
).
Next we put
F
,m
m
:
n
, for 0 < n, m and
S
= F
,m
: m .
It should be easy to check that S
, R
, h
(for
<
, 0 < g()) dened in
this way satisfy the requirements ()().
A tree creature t TCR
[H] is in K
1.4.1
if
<
, g() > 0 and
dis[t] R
, val[t] = ,
m)) : m dis[t] and nor[t] = h
(dis[t]).
If A R
then the unique treecreature t TCR
[H] K
1.4.1
such that dis[t] = A
will be denoted by t
,A
.
The operation
1.4.1
is trivial: it gives a non-empty result for singletons only and
then
1.4.1
(t) = s K
1.4.1
: val[s] val[t].
Claim 1.4.1.1. The forcing notion Q
tree
5
(K
1.4.1
,
1.4.1
) collapses c onto .
Proof of the claim. Fix
<
) for a moment.
Elements of S
are pairwise disjoint so we may naturally order them according to

the smallest element. Say S
= F
,m
: m < . Let f : [g(), ) . We

dene a condition p
f,
Q
tree
5
(K
1.4.1
,
1.4.1
) putting (we keep the notation as for
the forcing notions Q
tree
e
(K, )):
root(p
f,
) = ;
let k
0
= g(), k
+1
= f(k
) +k
+ 1 (for < );
if T
p
f,
, k
1
g() < k
then t
p
f,
= t
,F
,f(g())
and
succ
T
p
f, () =
m) : m F
,f(g())
.
Clearly this denes p
f,
Q
tree
5
(K
1.4.1
,
1.4.1
). Note that
if f, g : [g(), ) are distinct
then the conditions p
f,
, p
g,
are incompatible in Q
tree
5
(K
1.4.1
,
1.4.1
)
(by the requirement ()). Let be a Q
tree
5
(K
1.4.1
,
1.4.1
)name for a function
dened on
<
such that ' () V & () : [g(), ) for
<
and
for f : [g(), ) we have
p
f,
'
Q
tree
5
(K1.4.1,1.4.1)
() = f.
This denition is correct as p
f,
: f : [g(), ) is an antichain (of course it
is not necessarily maximal in Q
tree
5
(K
1.4.1
,
1.4.1
)). The claim will be shown if we
prove that
'
Q
tree
5
(K1.4.1,1.4.1)
(g
V)(
<
)(n g())( ()(2n) = g(n)).
For this suppose that g
, p Q
tree
5
(K
1.4.1
,
1.4.1
). Choose an increasing
sequence g(root(p)) < k
0
< k
1
< . . . of odd integers such that for each <
( T
p
)(k
g() nor[t
p
] + 2).
Let f : [k
0
, ) be a function such that:
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

(1) f(k
) = k
+1
k
1
(2) if n k
0
then f(2n) = g(n).
(Note that these clauses are compatible as the k
s are odd. Of course there is still

much freedom left in dening f.)
Choose T
p

k
0
and look at the condition p
f,
. Due to the requirement
() this condition is compatible with p: dene r Q
tree
5
(K
1.4.1
,
1.4.1
) by
root(r) = , T
r
T
p
and
if T
r
, k
1
g() < k
then
t
r
= t
,A
and succ
T
r () =
m) : m A
where A
= dis[t
p
] F
,f(g())
.
Note that by the choice of k
and the requirement () we have

t
,A

1.4.1
(t
p
)
1.4.1
(t
p
f,
) and nor[t
,A
] = .
Consequently the denition of r is correct. Clearly r is stronger than both p and
p
f,
. Thus
r '
Q
tree
5
(K1.4.1,1.4.1)
(n k
0
)( ()(n) = f(n))
which together with
(n k
0
)(f(2n) = g(n)) and k
0
= g()
nishes the proof of the claim.
Our next example shows that the assumption that (K, ) is nitary in 2.1.6 is
crucial.
Example 1.4.2. Let H(i) = for i .
There is a creating pair (K
1.4.2
,
1.4.2
) for H which is forgetful and growing (see
2.1.1(3)) but the forcing notion Q
(K
1.4.2
,
1.4.2
) is not proper (and collapses c
onto ).
(By 2.1.3 we may replace Q
by either Q
w
or Q
s
or Q
f
(for a fast function f).)
Construction. This is similar to 1.4.1: for 0 < i < choose ,= S
i

R
i
[]
and functions h
i
: R
i
satisfying the requirements ()() of the
construction of 1.4.1 (with i instead of and g()) and
()

S
i
= for each 0 < i <
(this additional condition is satised by the example constructed there). Fix an
enumeration S
i
= F
,m
i
: m .
A creature t K
1.4.2
may be described in the following way. For each i
[m
t
dn
, m
t
up
) we have a set A
i
R
i
. Now:
dis[t] = A
i
: m
t
dn
i < m
t
up
)
val[t] = u, v)

i<m
t
dn
H(i)

i<m
t
up
H(i): u < v & (i [m
t
dn
, m
t
up
))(v(i) A
i
),
nor[t] = maxh
i
(A
i
) : i [m
t
dn
, m
t
up
).
If creatures t
0
, . . . , t
n1
K
1.4.2
are determined by sets A
j
i
R
i
(for j < n,
i [m
tj
dn
, m
tj
up
)) in the way described above and m
tj+1
dn
= m
tj
up
(for j < n 1) then
1.4.2
(t
0
, . . . , t
n1
) consists of all creatures t K
1.4.2
which are determined (in
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

the way described above) by some sets A
i
R
i
(for i [m
t0
dn
, m
tn1
up
)) such that
A
i
A
j
i
whenever m
tj
dn
i < m
tj
up
, j < n.
It is easy to check that
1.4.2
is a composition operation on K
1.4.2
. The creating
pair (K
1.4.2
,
1.4.2
) is forgetful and growing.
(K
1.4.2
,
1.4.2
) collapses c onto .
Proof of the claim. We proceed like in 1.4.1.1 (with small modications how-
ever). Let : : n (
0
(n),
1
(n)) be a bijection. Let
n
:
[n, )
(for n ) be mappings dened in the following manner. Let f : [n, ) ;

inductively dene n
0
= n, n
+1
= n
+
0
(f(n
)) +
1
(f(n
)) + 2
(for < ) and then m
= n
+
0
(f(n
)) +1 (so n
< m
< n
+1
);
now put
n
(f)() = f(m
) (for ).
For 0 < n < , u

i<n
H(i) and a function f : [n, ) dene a condition
p
u,f
Q
(K
1.4.2
,
1.4.2
):
k
0
= n, k
2+1
= k
2
+
0
(f(k
2
))+1, k
2+2
= k
2+1
+
1
(f(k
2
))+1,
w
p
u,f
= u,
if k
2
i+n < k
2+2
then t
p
u,f
i
K
1.4.2
is such that m
t
p
u,f
i
dn
= n+i,
m
t
p
u,f
i
up
= n +i + 1, dis[t
p
u,f
i
] = F
+1,f(i+n)
i+n
).
As in 1.4.1.1, if f, g : [n, ) are distinct, u

i<n
H(i) then the conditions
p
u,f
, p
u,g
are incompatible. Consequently we may choose a Q
(K
1.4.2
,
1.4.2
)name
for a function on such that ' (n )( (n) : [n, ) ) and p
u,f
' (n) = f.
To nish it is enough to show that
'
Q
(K1.4.2,1.4.2)
(g
V)(n )(
n
( (n)) = g).
Suppose that p Q
(K
1.4.2
,
1.4.2
), g
. Choose 2 < i
0
< i
1
< . . . < such
that nor[t
p
i
] + 2 and next choose k

0
< k
1
< k
2
< . . . < such that for each
:
m
t
p
i
dn
k
< m
t
p
i
up
and for some set A
R
k
we have
h
k
(A
) + 2 and (n A
)(u, v) val[t
p
i
])(v(k
) = n)
(possible by the way we dened (K
1.4.2
,
1.4.2
)). Choose any v pos(w
p
, t
p
0
, . . . , t
p
i0
)
and let u = v`k
0
. Next choose f : [k
0
, ) such that for each :
0
(f(k
2
)) = k
2+1
k
2
1,
1
(f(k
2
)) = k
2+2
k
2+1
1, f(k
2+1
) = g(),
and if k (k
2
, k
2+2
) k
2+1
, m
t
p
i
dn
k < m
t
p
i
up
( < , i < ) then
(u, v) val[t
p
i
])(v(k) F
+1,f(k)
k
).
One easily checks that the choice of f is possible (remember the additional require-
ment ()) and that the conditions p
u,f
and p are compatible in Q
(K
1.4.2
,
1.4.2
).
As
p
u,f
'
Q
(K1.4.2,1.4.2)

n
( (n)) =
n
(f) = g,
we nish the proof of the claim.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

One could expect that the main reason for collapsing c in the two examples
constructed above is that the (K, )s there are not nitary. But this is not the
case. Using similar ideas we may build a nitary creating pair (K, ) for which the
forcing notion Q
(K, ) collapses c onto as well. This is the reason why we have

to use forcing notions Q
f
(K, ) with (K, ) satisfying extra demands (including
Halving and bigness, see 2.2.12) and why Q
(K, ) is used only for growing (K, )

(so then Q
s
(K, ) is dense in Q
(K, ), see 2.1). This bad eect can be made

quite general and we will present it in this way, trying to show the heart of the
matter. One could try to cover the previous examples by our negative theory
too, but this would involve much more complications.
Definition 1.4.3. We say that a weak creating pair (K, ) is local if for every
t K, w basis(t) and u pos(w, t) we have g(u) = g(w) + 1.
Definition 1.4.4. Let (K, ) be a (nice and smooth) creating pair for H
which is local (so t K m
t
up
= m
t
dn
+ 1) and simple (which means that
(o) ,= [o[ = 1; see 2.1.7(1)).
We say that (K, ) is denitely bad if there are a perfect tree T
<
and
mappings F
0
, F
1
such that
(1) T
m
is nite for each m , dom(F
0
) = dom(F
1
) =

m<
i<m
H(i),
(2) if v

im
H(i), m < then F
1
(v) : T
m
2 and F
0
(v) : T
m
T is such that ( T
m
)(F
0
(v)() succ
T
()),
(3) if t K, nor[t] 2, m = m
t
dn
> 0, i < 2 and F
: T
m
T is such
that F
() succ
T
() for T
m
then there is s (t) such that
nor[s] nor[t] 1 and for each T
m1
there is succ
T
()
with
(u basis(s))(v pos(u, s))(F
0
(v)() = F
() & F
1
(v)() = i).
Proposition 1.4.5. Suppose that (K, ) is a local, simple and denitely bad
creating pair for H such that (t) is nite for each t K. Then
(a) the forcing notion Q
(K, ) collapses c onto ,

(b) if f : is a fast function then the forcing notion Q
f
(K, )
collapses c onto .
Proof. In both cases the proof is exactly the same, so let us deal with (a)
only. So suppose that a nitely branching perfect tree T
<
and functions
F
0
, F
1
witness that (K, ) is denitely bad. Let G
0
: T

i<
H(i) [T] and
G
1
: [T]
i<
H(i) 2
be dened by
G
0
(, W)`g() = , and
G
0
(, W)`(m + 1) = F
0
(W`(m+ 1))(G
0
(, W)`m) for m g(),
G
1
(, W)(n) = F
1
(W`(n + 1))(`n) for n .
We are going to show that
'
Q
(K,)
(r 2
V)( T)(
n )(r(n) = G
1
(G
0
(,

W),

W)(n)),
where

W is the Q
(K, )name dened in 1.1.13. To this end suppose that r 2
and p Q
(K, ). We may assume that g(w

p
) > 0 and (i )(nor[t
p
i
] 3).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Fix i
0
for a moment. By downward induction on i i
0
we choose s
i0
i

(t
p
i
) and F
i0,i
: T
g(w
p
) +i 1
T such that
() nor[s
i0,i
] nor[t
p
i
] 1,
() F
i0,i
() succ
T
() for T, g() = g(w
p
) +i 1,
() for all sequences u basis(s
i0,i
) and v pos(u, s
i0,i
) and every T of
length g() = g(w
p
) +i 1 we have
F
0
(v)(F
i0,i
()) = F
i0,i+1
() and F
1
(v)(F
i0,i
()) = r(g(w
p
) +i)
[for i = i
0
we omit the rst part of the above demand].
It should be clear that the choice of the s
i0,i
s and F
i0,i
s as above is possible by
1.4.4(3). All levels of the tree T are nite, so for each i there are nitely many
mappings F
: T
g(w
p
) +i 1
T
g(w
p
) +i
. Moreover, each (t
p
i
) is
nite (for i ). Hence, by K onigs lemma, we nd a sequence t
q
i
: i ) and a
mapping F
: T T such that
(i) ( T)(F
() succ
T
()) and
(ii) for each i there is j(i) > i such that for every j i
s
j(i)
j
= t
q
j
and ( T
g(w
p
) +j 1
)(F
() = F
j(i),j
()).
By () we have that q = (w
p
, t
q
0
, t
q
1
, . . .) Q
(K, ) and it is stronger than p.

Take any T with g() = g(w
p
) 1 and let
0
= F
(),
i+1
= F
(
i
), and
+
= lim(
i
) [T]. It follows from () and (ii) (e.g. inductively using smoothness)
that for each n and v pos(w
p
, t
q
0
, . . . , t
q
n1
) we have
(v, t
q
n
, t
q
n+1
, . . .) ' G
0
(
0
,

W)`g(v) =
+
`g(v) and
G
1
(G
0
(
0
,

W),

W)`[g(
0
), g(v)) = r`[g(
0
), g(v)).
Hence we conclude
q '
Q
(K,)
(n g(
0
))(G
1
(G
0
(
0
,

W),

W)(n) = r(n)),
nishing the proof.
Example 1.4.6. Let f : be a fast function (for example f(k, ) =
2
2k
( + 1)). There are a nitary function H and a creating pair (K
1.4.6
,
1.4.6
) for
H such that
(a) (K
1.4.6
,
1.4.6
) is local, simple, forgetful and denitely bad (and smooth),
(b)
1.4.6
(t) is nite for each t K
1.4.6
,
(c) the forcing notions Q
(K
1.4.6
,
1.4.6
) and Q
f
(K
1.4.6
,
1.4.6
) are not trivial
and thus collapse c onto .
Construction. Let f : be fast. For n let k
n
= 2
f(n+1,n+1)
.
Next, for n , let H(n) consist of all pairs z
0
, z
1
) such that
z
0
:
i<n
k
i
k
n
and z
1
:
i<n
k
i
2.
Immediately by the denition, one sees that H is nitary. Now we dene the
creating pair (K
1.4.6
,
1.4.6
) for H. A creature t CR[H] with m
t
dn
> 0 is in K
1.4.6
if:
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

dis[t] = m
t
dn
, A
t
:

i<m
t
dn
1
k
i
), F
t
0
, F
t
1
), where
A
t
k
m
t
dn
1
for

i<m
t
dn
1
k
i
,
F
t
0
:
i<m
t
dn
k
i
: (m
t
dn
1) A
t
(m
t
dn
1)
k
m
t
dn
,
F
t
1
:

i<m
t
dn
k
i
: (m
t
dn
1) A
t
(m
t
dn
1)
2,
val[t] consists of all pairs w, u)

i<m
t
dn
H(i)

im
t
dn
H(i) such that w < u
and if u(m
t
dn
) = z
0
, z
1
) then z
0
F
t
0
and z
1
F
t
1
.
nor[t] = k
m
t
dn
max[A
t
[ :
i<m
t
dn
1
k
i
.
If t CR[H], m
t
dn
= 0 then we take t to K
1.4.6
if:
nor[t] = 0, dis[t] = 0, x
t
, i
t
), where x
t
k
0
, i
t
< 2, val[t] = ), u ), where
u

i0
H(i), u(0) = x
t
, i
t
) (both x
t
and i
t
are treated here as functions from
)).
The composition operation
1.4.6
is the trivial one (so
1.4.6
(o) is non-empty for
singletons only and
1.4.6
(t) = s K
1.4.6
: val[s] val[t]). Easily (K
1.4.6
,
1.4.6
)
is a local, simple and forgetful creating pair. Note that if n > 0 and t K
1.4.6
is such that m
t
dn
= n and A
t
= for each
i<n1
k
i
then nor[t] = k
n1
>
f(n, n), so the forcing notions Q
(K
1.4.6
,
1.4.6
), Q
f
(K
1.4.6
,
1.4.6
) are non-trivial.
Finally let T =

n
i<n
k
i
and for v

im
H(i) let F
0
(v), F
1
(v) be such that
v(m) = F
0
(v), F
1
(v)). It should be clear that T, F
0
, F
1
witness that (K
1.4.6
,
1.4.6
)
is denitely bad.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

CHAPTER 2
Properness and the reading of names
This chapter is devoted to getting the basic property: properness. The rst
two sections deal with forcing notions determined by creating pairs. We dene
properties of creating pairs implying that appropriate forcing notions are proper.
Some of these properties may look articial, but in applications they appear natu-
rally. The third part deals with forcing notions Q
tree
e
(K, ) (determined by a tree
creating pair). Here, properness is an almost immediate consequence of our choice
of norm conditions. In most cases, proving properness of a forcing notion we get
much stronger property: continuous reading of names for ordinals. This property
will be intensively used in the rest of the paper. Finally, in the last part of the
chapter we give several examples for properties introduced and studied before.
2.1. Forcing notions Q
s
(K, ), Q
w
(K, )
Definition 2.1.1. Let (K, ) be a creating pair for H.
(1) For t K, m
0
m
t
dn
, m
t
up
m
1
we dene the creature s
def
= t [m
0
, m
1
)
by:
nor[s] = nor[t],
dis[s] = 4, m
0
, m
1
)
dis[t]),
val[s] = w, u)
i<m0
H(i)
i<m1
H(i) : u`m
t
dn
, u`m
t
up
) val[t] &
w < u & (i [m
0
, m
t
dn
) [m
t
up
, m
1
))(u(i) = 0).
[Note that t [m
0
, m
1
) is well dened only if val[s] ,= above and then
m
s
dn
= m
0
, m
s
up
= m
1
.]
(2) The creating pair (K, ) is omittory if:
(
0
) if t K and u basis(t) then u
0
[m
t
dn
,m
t
up
)
pos(u, t) but there is
v pos(u, t) such that v`[m
t
dn
, m
t
up
) ,= 0
[m
t
dn
,m
t
up
)
,
(
1
) for every (t
0
, . . . , t
n1
) PFC(K, ) and i < n:
t
i
[m
t0
dn
, m
tn1
up
) (t
0
, . . . , t
n1
),
(
2
) if t, t [m
0
, m
1
) K then for every u basis(t [m
0
, m
1
)) and
v pos(u, t [m
0
, m
1
)) we have
v(n) ,= 0 & n [g(u), g(v)) n [m
t
dn
, m
t
up
).
[Note that (
0
) implies that in the cases relevant for (
1
), t
i
[m
t0
dn
, m
tn1
up
)
is well dened.]
(3) (K, ) is growing if for any (t
0
, . . . , t
n1
) PFC(K, ) there is a creature
t (t
0
, . . . , t
n1
) such that nor[t] max
i<n
nor[t
i
].
Proposition 2.1.2. If (K, ) is omittory then it is growing.
27
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

28 2. PROPERNESS AND THE READING OF NAMES
Proposition 2.1.3. Suppose that a creating pair (K, ) is growing.
(1) Then Q
s
(K, ) is a dense subset of both Q
(K, ) and Q
w
(K, ) and
Q
f
(K, ) (for every fast (see 1.1.12) function f : ). Conse-
quently whenever we work with growing creating pairs we may interchange
the respective forcing notions as they are equivalent.
(2) Moreover, if g : , p Q
w
(K, ) then there is q Q
s
(K, )
such that p
w
0
q and
(n )(nor[t
q
n
] > g(n, m
t
q
n
dn
)).
Proof. Suppose that g : and (w, t
0
, t
1
, . . .) Q
w
(K, ). Choose
an increasing sequence k
0
< k
1
< k
2
< . . . such that
nor[t
k0
] > g(0, m
t0
dn
) and nor[t
kn+1
] > g(n + 1, m
t
kn
up
)
(exists by 1.1.10(w)) and choose s
0
(t
0
, . . . , t
k0
), s
n+1
(t
kn+1
, . . . , t
kn+1
)
such that nor[s
n
] nor[t
kn
] (exist by 2.1.1(3)). Hence (by 1.2.8(2); remember
that we assume (K, ) is nice)
q
def
= (w, s
0
, s
1
, s
2
, . . .) Q
s
(K, )
and clearly (w, t
0
, t
1
, t
2
, . . .)
0
q.
Theorem 2.1.4. Assume (K, ) is a nitary creating pair. Further assume
that p Q
s
(K, ) and for n < we have a Q
s
(K, )-name
n
such that
'
Q
s
(K,)

n
is an ordinal and < . Then there is q = (w
p
, s
0
, s
1
, s
2
, . . .)
such that:
(a) p
s
q Q
s
(K, ) and
(b) if n < , m m
sn1
up
then the condition q approximates
m
at s
n
(see
1.2.9(2)).
Proof. Let p = w
p
, t
p
0
, t
p
1
, t
p
2
, . . .). Let s
i
= t
p
i
for i < . Now, by induction
on n we dene q
n
, s
n
, t
n
n+1
, t
n
n+2
, . . . such that:
(i) q
= p,
(ii) q
n+1
= (w
p
, s
0
, . . . , s
n
, t
n
n+1
, t
n
n+2
, . . .) Q
s
(K, )
(iii) q
n

s
n
q
n+1
(iv) if w
1
pos(w
p
, s
0
, . . . , s
n1
), m m
sn1
up
and there is a condition r
Q
s
(K, ),
s
0
stronger than (w
1
, s
n
, t
n
n+1
, t
n
n+2
, . . .) which decides the
value of
m
then the condition (w
1
, s
n
, t
n
n+1
, t
n
n+2
, . . .) does it.
Arriving at the stage n + 1 > we have dened
q
n
= (w
p
, s
0
, . . . , s
n1
, t
n1
n
, t
n1
n+1
, . . .).
Fix an enumeration (w
n
i
, m
n
i
) : i < k
n
) of
pos(w
p
, s
0
, . . . , s
n1
) (m
sn1
up
+ 1)
(since each H(m) is nite, k
n
is nite). Next choose by induction on k k
n
conditions q
n,k
Q
s
(K, ) such that:
() q
n,0
= q
n
() q
n,k
is of the form (w
p
, s
0
, . . . , s
n1
, t
n,k
n
, t
n,k
n+1
, t
n,k
n+1
, . . .)
() q
n,k

n
q
n,k+1
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

2.1. FORCING NOTIONS Q
s
(K, ), Q
w
(K, ) 29
() if, in Q
s
(K, ), there is a condition r
0
(w
n
k
, t
n,k
n
, t
n,k
n+1
, t
n,k
n+2
, . . .) which
decides (in Q
s
(K, )) the value of
m
n
k
, then
(w
n
k
, t
n,k+1
n
, t
n,k+1
n+1
, t
n,k+1
n+2
, . . .) Q
s
(K, )
is a condition which forces a value to
m
n
k
.
(Note: (w
n
k
, t
n,k
n
, t
n,k
n+1
, t
n,k
n+2
, . . .) Q
s
(K, ).)
For this part of the construction we need our standard assumption that (K, ) is
nice. Note that choosing (w
n
k
, t
n,k+1
n
, t
n,k+1
n+1
, t
n,k+1
n+2
, . . .) we want to be sure that
(w
p
, s
0
, . . . , s
n1
, t
n,k+1
n
, t
n,k+1
n+1
, t
n,k+1
n+2
, . . .) Q
s
(K, )
(remember that 1.1.7(b)(ii) might fail). But by 1.2.8(2) it is not a problem. Next,
the condition q
n+1
def
= q
n,kn
Q
s
(K, ) satises (iv): the keys are the clause ()
and the fact that
(w
n
k
, t
n,k+1
n
, t
n,k+1
n+1
, t
n,k+1
n+2
, . . .)
s
0
(w
n
k
, t
n,kn
n
, t
n,kn
n+1
, t
n,kn
n+2
, . . .) Q
s
(K, ).
Thus s
n
def
= t
n,kn
n
, q
n+1
and t
n
n+k
def
= t
n,kn
n+k
are as required.
Now, by 1.2.13:
q
def
= (w
p
, s
0
, s
1
, . . . , s
l
, s
l+1
, . . .) = lim
n
q
n
Q
s
(K, ).
Easily it satises the assertions of the theorem.
A small modication of the proof of 2.1.4 shows the corresponding result for
w
(K, ):
Theorem 2.1.5. Assume (K, ) is a nitary creating pair and p Q
w
(K, ).
Let
n
be Q
w
(K, )-names for ordinals (for n < ), < . Then there is a
condition q = (w
p
, s
0
, s
1
, . . .) Q
w
(K, ) such that
(a) p
q and
(b) there is an increasing sequence = k
0
< k
1
< k
2
< . . . < such that if
n < and m m
s
kn1
up
then the condition q approximates
m
at s
kn
.
As an immediate corollary to theorem 2.1.4 we get the following.
Corollary 2.1.6. Assume that (K, ) is a nitary creating pair.
(a) Suppose that
n
are Q
s
(K, )-names for ordinals and q Q
s
(K, )
is a condition satisfying (b) of 2.1.4 (for = 0). Further assume that
q r Q
s
(K, ), n < g(w
r
) and r '
n
= (for some ordinal ).
Then for some q
s
(K, ), q
apr
q

0
r, we have q
'
n
= .
(Note: q
s
(K, ): q
apr
q
is countable provided

i<
H(i) is countable.)
(b) The forcing notion Q
s
(K, ) is proper (and -proper for <
1
).
It should be underlined here that 2.1.6 applies to forcing notions Q
(K, ) for
nitary growing creating pairs (remember 2.1.3). To get the respective conclusion
for Q
w
(K, ) we need to assume more.
Definition 2.1.7. We say that:
(1) A weak creating pair (K, ) is simple if (o) is non-empty for singletons
only.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

(2) A creating pair (K, ) is gluing if it is full and for every k < there is
n
0
< such that for every n
0
n < , (t
0
, . . . , t
n
) PFC(K, ), for
some s (t
0
, . . . , t
n
) we have
nor[s] mink, nor[t
0
], . . . , nor[t
n
].
In this situation the integer n
0
is called the gluing witness for k.
The two properties dened above are, in a sense, two extremal situations under
which we may say something on (K, ). The demand either simple or gluing
(like in 2.2.11) should not be surprising if one realizes that then we know what may
happen when is applied, at least in terms of m
t
dn
, m
t
up
.
Corollary 2.1.8. Assume that (K, ) is a nitary and simple creating pair.
(a) Suppose that
n
are Q
w
(K, )names for ordinals, k
0
< k
1
< . . . <
and q Q
w
(K, ) are as in 2.1.5(b). Suppose that q r Q
w
(K, ),
and r decides the value of one of the names
n
, say r '
n
= .
Then for some q
w
(K, ) we have
q
apr
q
, q
'
n
= and q
, r are compatible.
(b) The forcing notion Q
w
(K, ) is proper (and even more).
Remark 2.1.9. Note the presence of simple in the assumptions of 2.1.8. In
practical applications of forcing notions of the type Q
w
we can get more, see
2.1.12 below.
Definition 2.1.10. Let (K, ) be a creating pair for H. We say that (K, )
captures singletons if (K, ) is forgetful and for every (t
0
, . . . , t
n
) PFC(K, )
and for each u basis(t
0
)(=

m<m
t
0
dn
H(m)) and v pos(u, t
0
, . . . , t
n
) there is
(s
0
, . . . , s
k
) PFC(K, ) such that (t
0
, . . . , t
n
) (s
0
, . . . , s
k
) (see 1.2.4) and
pos(u, s
0
, . . . , s
k
) = v, m
t0
dn
= m
s0
dn
, m
s
k
up
= m
tn
up
.
[Note that we put no demands on the norms of the s
i
s.]
Proposition 2.1.11. Suppose that (K, ) is a creating pair which captures
singletons (so in particular it is forgetful), p Q
w
(K, ) and is a Q
w
(K, )
name for an ordinal. Then there is q Q
w
(K, ) such that
p
w
0
q and q decides .
Proof. Take r Q
w
(K, ) such that p r and r ' = (for some ).
Look at w
r
: for some n we have w
r
pos(w
p
, t
p
0
, . . . , t
p
n1
). By 2.1.10 we nd
s
0
, . . . , s
k
such that pos(w
p
, s
0
, . . . , s
k
) = w
r
and
(w
p
, t
p
0
, t
p
1
, . . .) (w
p
, s
0
, . . . , s
k
, t
r
0
, t
r
1
, . . .)
def
= q Q
w
(K, ).
Clearly p
w
0
q. To show that q ' = we use our standard assumption that
(K, ) is smooth. Suppose that q
q is such that q
' ,= . We may assume that

g(w
q
) > m
s
k
up
. By the smoothness we have w
q
`m
s
k
up
pos(w
p
, s
0
, . . . , s
k
), and so
w
q
`m
s
k
up
= w
r
, and w
q
pos(w
r
, t
r
0
, . . . , t
r
) (for a suitable < ). Consequently

q
r and this is a contradiction.

Corollary 2.1.12. Assume (K, ) is a nitary creating pair which captures
singletons.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

2.2. FORCING NOTION Q
f
(K, ): BIGNESS AND HALVING 31
(1) Let p Q
w
(K, ),
n
be Q
w
(K, )-names for ordinals (for n < )
and < . Then there is a condition q Q
w
(K, ) such that
(a) p
q and
(b) the condition q essentially decides (see 1.2.9(1)) each name
n
.
(2) The forcing notion Q
w
(K, ) is proper.
Proof. 1) Follows from 2.1.5 and 2.1.11.
2) Follows from 1).
2.2. Forcing notion Q
f
(K, ): bigness and halving
Definition 2.2.1. Let r = r
m
: m < ) be a non-decreasing sequence of
integers 2. For a creating pair (K, ) for H we say that:
(1) (K, ) is big if for every k < there is m < such that:
if t K, nor[t] m, u basis(t) and c : pos(u, t) 0, 1
then there is s (t) such that nor[s] k, and c`pos(u, s) is constant.
In this situation we call m a bigness witness for k.
(2) (K, ) is rbig if for each t K such that nor[t] > 1 and u basis(t)
and c : pos(u, t) r
m
t
dn
there is s (t) such that nor[s] nor[t] 1
and c ` pos(u, s) is constant.
Remark 2.2.2. Clearly, for a creating pair (K, ), rbig implies implies big.
To show how the notions introduced in denition 2.2.1 work we start with
proving an application of 2.1.4 to the case when the creating pair is additionally
big and growing.
Proposition 2.2.3. Assume (K, ) is a nitary, growing and big creating pair.
If p Q
s
(K, ), p ' < m, m < then there is q, p
0
q Q
s
(K, ) such
that q ' = m
0
for some m
0
.
Proof. Let h
be such that h(k) is a bigness witness for k (remember

that (K, ) is big, see 2.2.1(1)). Note that 2.1.4 + 2.1.3(2) give us a condition
p
= (w
p
, s
0
, s
1
, s
2
, . . .) Q
s
(K, ) such that p
s
0
p
and
() nor[s
] h
(m|pos(w
p
,s0,...,s
1
)|)
(m
s
dn
+ 1) for all < and
() p
approximates the name at each n < .

Using iteratively the choice of h(k) we will have then
() for every < and each function
d : u, v) : u pos(w
p
, s
0
, . . . , s
1
) & v pos(u, s
) m+ 1
there is a creature s (s
) such that nor[s] > m

s
dn
and
d ` u, v) dom(d) : v pos(u, s)
depends on the rst coordinate only.
(Since, as usual, we assume that (K, ) is nice, we have in () above that basis(s)
pos(w
p
, s
0
, . . . , s
1
).)
Now apply () to nd s
(s
) (for < ) such that nor[s
] > m
s
dn
= m
s
dn
and for every u pos(w
p
, s
0
, . . . , s
1
) we have
() for each v
0
, v
1
pos(u, s
), i < m
(v
0
, s
+1
, s
+2
, . . .) ' = i i (v
1
, s
+1
, s
+2
, . . .) ' = i.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Look at q
def
= (w
p
, s
0
, s
1
, s
2
, . . .). First it is a condition in Q
s
(K, ) as (K, ) is
nice and s
(s
), nor[s
] > m
s
dn
. To show that
(m
0
< m)(q ' = m
0
)
take a condition r = (w
1
, t
0
, t
1
, . . .) q such that r decides the value of ,
w
1
pos(w
p
, s
0
, . . . , s
1
) and is the smallest possible. By () we know that
the condition (w
1
, s
, s
+1
, . . .) forces a value to , say m
0
. We claim that = 0,
i.e. w
1
= w
p
(which is enough as then q decides the value of ). Why? Suppose that
> 0 and look at the requirement () for 1, u = w
1
` m
s
1
dn
. By the smoothness
u pos(w
p
, s
0
, . . . , s
2
) and consequently, by (), for each v pos(u, s
1
) we
have
(v, s
, s
+1
, . . .) ' = m
0
.
Applying smoothness once again we note that for each w pos(u, s
1
, s
, . . . , s
k
)
w ` m
s
dn
pos(u, s
1
) and w pos(w ` m
s
dn
, s
, . . . , s
k
).
Hence for each such w we have
(w ` m
s
dn
, s
, s
+1
, . . . , ) ' = m
0
and
(w, s
k+1
, s
k+2
, . . .) (w ` m
s
dn
, s
, s
+1
, . . .)
and so
(w, s
k+1
, s
k+2
, . . . , ) ' = m
0
.
Hence we may conclude that (u, s
1
, s
, . . . ) ' = m
0
which contradicts the choice
of .
Remark 2.2.4. One may notice that the assumptions of 2.2.3 are dicult to
satisfy in most natural cases. First examples of growing creating pairs one has in
mind are omittory creating pairs. However, if we demand that an omittory creating
pair (K, ) is smooth then we get to
t K & u basis(t) u
0
[m
t
dn
,m
t
up
)
pos(u, t).
This excludes bigness as dened in 2.2.1. Thus it is desirable to consider in this
case a weaker condition, which more ts to specic properties of omittory creating
pairs.
Definition 2.2.5. An omittory creating pair (K, ) is omittorybig if for every
k < there is m < such that:
if t K, nor[t] m, u basis(t) and c : pos(u, t) 0, 1 then there is s (t)
such that nor[s] k and c`(pos(u, s) 0
[m
t
dn
,m
t
up
)
) is constant.
We may call m an omittory-bigness witness for k.
Proposition 2.2.6. Assume (K, ) is a nitary, omittory and omittorybig
creating pair. Suppose that p Q
s
(K, ), p ' < m, m < . Then there is a
condition q Q
s
(K, ) such that p
s
0
q and q decides the value of .
Proof. We start as in the proof of 2.2.3, but in () there we say that
d ` u, v) dom(d) : v pos(u, s) & (k [g(u), g(v)))(v(h) ,= 0)
depends on the rst coordinate only, and therefore we get s
(s
) as there but
with () replaced by
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

f
()
for each v
0
, v
1
pos(u, s
) u
0
[m
s
dn
,m
s
up
)
, i < m
(v
0
, s
+1
, s
+2
, . . .) ' = i i (v
1
, s
+1
, s
+2
, . . .) ' = i.
Now comes the main modication of the proof of 2.2.3. We choose an innite set
I = i
0
, i
1
, i
2
, . . . such that for every i < m we have
if: k <
< , w pos(w
p
, s
0
, . . . , s
i
k
), v
0
pos(w, s
[m
s
i
k
up
, m
s
up
)),
v
1
pos(w, s
[m
s
i
k
up
, m
s
up
)), and
(m [m
s
dn
, m
s
up
))(v
0
(m) ,= 0) and (m [m
s
dn
, m
s
up
))(v
1
(m) ,= 0)
then: (v
0
, s
+1
, s
+2
, . . .) ' = i i (v
1
, s
+1
, s
+2
, . . .) ' = i
and a similar condition for the case of w = w
p
, 0
< . The construction

of the set I is rather standard (by induction) and it goes like in the proof of the
suitable property for the Mathias forcing (see e.g. [BaJu95, 7.4.6]). Next we look
at
q
def
= (w
p
, s
i0
[m
s
0
dn
, m
s
i
0
up
), s
i1
[m
s
i
0
up
, m
s
i
1
up
), s
i2
[m
s
i
1
up
, m
s
i
2
up
), . . .).
It should be clear that it is a condition in Q
s
(K, ) which is
s
0
stronger than
p. Note that, as (K, ) is omittory (remember the demand (
0
) of 2.1.1(2)), by
the choice of the set I we have
if w
1
pos(w
q
, t
q
0
, . . . , t
q
k
), k < , w
1
`[m
t
q
k
dn
, m
t
q
k
up
) = 0
[m
t
q
k
dn
,m
t
q
k
up
)
and
the condition (w
1
, t
q
k+1
, t
q
k+2
, . . .) decides the value of
then (w
1
`m
t
q
k
dn
, t
q
k
, t
q
k+1
, t
q
k+2
, . . .) does so.
Now, like in 2.2.3 we show that the condition q decides the value of , using the
remark above and the choice of the set I.
Definition 2.2.7. Let (K, ) be a creating pair.
(1) We say that (K, ) has the Halving Property if there is a mapping
half : K K
such that
(a) for each t K, half(t) (t) and nor[half(t)]
1
2
nor[t],
(b) if t
0
, . . . , t
n
K, minnor[t
i
] : i n 2 and a creature t
(half(t
0
), . . . , half(t
n
)) is such that nor[t] > 0 then there is s
(t
0
, . . . , t
n
) such that
nor[s] min
1
2
nor[t
i
] : i n and (ubasis(t
0
))(pos(u, s) pos(u, t)).
(2) We say that (K, ) has the weak Halving Property if there is a mapping
half : K K which satises (a) above and
(b)
if t
0
K, nor[t
0
] 2 and t (half(t
0
)) is such that nor[t] > 0
then there is a creature s (t
0
) such that
nor[s]
1
2
nor[t
0
] and (u basis(t
0
))(pos(u, s) pos(u, t)).
(3) Whenever we say that (K, ) has the (weak) Halving Property we assume
that the function half : K K witnessing this is xed.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Remark 2.2.8. Remember that we standardly assume that creating pairs are
nice, so in 2.2.7(1b), 2.2.7(2b
) we have basis(t
0
) basis(t) and basis(t
0
)
basis(s). Of course, the Halving Property implies the weak Halving Property.
Moreover, the two notions agree for simple creating pairs.
Our next lemma shows how we are going apply the Halving Property.
Lemma 2.2.9. Assume that (K, ) is a creating pair with the Halving Property
(witnessed by a mapping half : K K). Suppose that f : is a fast
function, is a Q
f
(K, )-name for an ordinal, n < , 0 < 1 and p Q
f
(K, )
is a condition such that
(i )(nor[t
p
i
] f(n, m
t
p
i
dn
)).
Further assume that there is a condition r Q
f
(K, ) such that
(i )(nor[t
r
i
] > 0) and (w
p
, half(t
p
0
), half(t
p
1
), . . .)
f
0
r and
r essentially decides (see 1.2.9). Then there is a condition q Q
f
(K, ) such
that
(i )(nor[t
q
i
]

2
f(n, m
t
q
i
dn
)), p
f
0
q and q essentially decides .
Proof. First note that the niceness implies that (w
p
, half(t
p
0
), half(t
p
1
), . . .) is
a condition in Q
f
(K, ) (by 2.2.7(1)(a) and 1.2.8(2); remember that f is fast). Now
suppose that r Q
f
(K, ) is as in the assumptions of the lemma. Take m < so
large that
(u pos(w
p
, t
r
0
, . . . , t
r
m1
))((u, t
r
m
, t
r
m+1
, . . .) decides the value of ) and
(i m)(nor[t
r
i
]

2
f(n, m
t
r
i
dn
))
(for the rst requirement remember that (K, ) is smooth; the second is possible
since 1). Next choose integers 0 = i
0
< i
1
< . . . < i
m1
< i
m
such that
( < m)(t
r
(half(t
p
i
), . . . , half(t
p
i
+1
1
))).
Applying the Halving Property (see 2.2.7(1b); remember that we have assumed
nor[t
r
] > 0 for each ) we nd s
(t
p
i
, . . . , t
p
i
+1
1
) (for < m) such that
nor[s
] min
1
2
nor[t
p
i
] : i
i < i
+1
and
(ubasis(t
p
i
))(pos(u, s
) pos(u, t
r
)).
Then easily
q
def
= (w
p
, s
0
, . . . , s
m1
, t
r
m
, t
r
m+1
, . . .) Q
f
(K, ), p
f
0
q and
(i )(nor[t
q
i
]

2
f(n, m
t
q
i
dn
))
(for the last statement remember that f is fast so f(n, ) is non-decreasing). More-
over q essentially decides the value of as
pos(w
p
, s
0
, . . . , s
m1
) pos(w
p
, t
r
0
, . . . , t
r
m1
).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

f
Remark 2.2.10. One could ask why we cannot in the conclusion of 2.2.9 require
that simply nor[t
q
i
]
1
2
nor[t
p
j
] (for j chosen somehow suitably, e.g. such that
m
t
q
i
dn
= m
t
p
j
dn
). The reason is that the upgrading procedure given by 2.2.7(1b)
takes care of possibilities only: no other relation between s, t there is required.
In particular we do not know if (t) (s) ,= . Consequently, if we apply this
procedure to all m (replacing each t
r
m
by suitable s
m
) then we may get a condition
incompatible with r.
Theorem 2.2.11. Assume that a creating pair (K, ) for H is nitary,

2-big
and has the Halving Property. Further suppose that a function f : is
H-fast and that (K, ) is either simple or gluing (see 2.1.7). Let
m
be Q
f
(K, )-
names for ordinals (for m ), p Q
f
(K, ) and n < .
Then there is a condition q Q
f
(K, ) such that p
f
n
q and q essentially decides
(see 1.2.9) all the names
m
(for m ).
Proof. Let (K, ) and f be as in the assumptions of the theorem.
Claim 2.2.11.1. Suppose that is a Q
f
(K, )-name for an ordinal, n < , a
condition p Q
f
(K, ) and a real are such that
2
|pos(w
p
,t
p
0
)|
2
H(m
t
p
1
dn
)
1
(where
H
() = [
i<
H(i)[, see 1.1.12), and
nor[t
p
0
] > 1 and (i > 0)(nor[t
p
i
] f(n + 1, m
t
p
i
dn
)).
f
(K, ) such that p
f
0
q, t
q
0
(t
p
0
), nor[t
q
0
]
nor[t
p
0
] 1, q essentially decides the value of and
(i > 0)(nor[t
q
i
]

2
|pos(w
p
,t
p
0
)|
f(n + 1, m
t
q
i
dn
)).
Proof of the claim: First note that our assumptions on (and the fact that f is
H-fast) imply that if t K is such that m
t
dn
m
t
p
1
dn
and
nor[t]

2
|pos(w
p
,t
p
0
)|
f(n + 1, m
t
dn
)
then
nor[t] > f(n, m
t
dn
) +
H
(m
t
dn
) + 2.
Let w
0
m
: m < m
0
enumerate pos(w
p
, t
p
0
). We inductively choose conditions p
0
m
for m m
0
such that
() p
0
0
= (w
0
0
, t
p
1
, t
p
2
, . . .),
() if there is a condition r Q
f
(K, ) such that p
0
m

0
r and
(i )(nor[t
r
i
]

2
m+1
f(n + 1, m
t
r
i
dn
))
and r essentially decides then we choose such an r and we put
p
0
m+1
= (w
0
m+1
, t
r
0
, t
r
1
, . . .),
() if we cannot apply the clause () (i.e. there is no r as above) then
p
0
m+1
= (w
0
m+1
, half(t
p
0
m
0
), half(t
p
0
m
1
), . . .).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

[For the sake of the uniformity of the inductive denition we let w
0
m0
to be e.g. w
0
0
.]
Note that by the niceness there are no problems in the above construction (i.e.
p
0
m
Q
f
(K, )). Let c : pos(w
p
, t
p
0
) 2 be such that
c(w
0
m
) =
_
0 if the clause () was applied to choose p
0
m+1
,
1 otherwise.
Due to the bigness of (K, ) we nd a creature s
0
(t
p
0
) such that c is constant
on pos(w
p
, s
0
) and nor[s
0
] nor[t
p
0
] 1. Note that
q
def
= (w
p
, s
0
, t
p
0
m
0
0
, t
p
0
m
0
1
, t
p
0
m
0
2
, . . .) Q
f
(K, )
(the niceness applies here once again, see 1.2.8; the norm condition should be
obvious as p
0
m0
Q
f
(K, )). Moreover
nor[t
p
0
m
0
i
]

2
|pos(w
p
,t
p
0
)|
f(n + 1, m
t
p
0
m
0
i
dn
)
and hence, in particular,
nor[t
p
0
m
0
i
] > f(n, m
t
p
0
m
0
i
dn
) +
H
(m
t
p
0
m
0
i
dn
) + 2.
If the constant value of c ` pos(w
p
, s
0
) is 0 then easily the condition q satises the
requirements of the claim (use 1.2.10).
So we want to exclude the possibility that the constant value is 1. For this
assume that it is the case. First note that then, due to the way we constructed
p
0
m0
, we may apply lemma 2.2.9 and conclude that there are no u pos(w
p
, s
0
) and
r Q
f
(K, ) such that nor[t
r
i
] > 0 for all i and
(u, t
p
0
m
0
0
, t
p
0
m
0
1
, . . .)
0
r and r essentially decides the value of .
Now we inductively choose an increasing sequence
0
<
1
< . . . of integers, crea-
tures s
1
, s
2
, . . . K and conditions p
0
, p
1
, . . . Q
f
(K, ) such that
(1) p
0
= q (dened above),
0
= 0,
(2)
k+1
is such that
k+1
>
k
and
(i
k+1
)(nor[t
p
k
i
] f(n +k + 2, m
t
p
k
i
dn
)),
(3) s
i
= t
p
k
i
for i
k
and nor[t
p
k
i
] > f(n + k, m
t
p
k
i
dn
) +
H
(m
t
p
k
i
dn
) + 2 for
i >
k
,
(4) there are no u pos(w
p
, s
0
, . . . , s
k
) and r Q
f
(K, ) such that
(i )(nor[t
r
i
] > 0)
and (u, t
p
k
k
+1
, t
p
k
k
+2
, . . .)
0
r and r essentially decides ,
(5) for i (
k
,
k+1
], s
i
(t
p
k
i
) and nor[s
i
] > f(n +k, m
si
dn
),
(6) nor[s
k+1
] > f(n +k + 1, m
s
k+1
dn
) and p
k

f
k
p
k+1
.
Suppose we have dened
k
, p
k
and s
i
for i
k
. Choose
k+1
according to the
requirement (2) above. Fix an enumeration
w
k
m
: m < m
k
= pos(w
p
, s
0
, . . . , s
k
, t
p
k
k
+1
, . . . , t
p
k
k+1
).
We inductively choose conditions p
k
m
Q
f
(K, ) (in the way analogous to the
construction of p
0
m
s):
()
k
p
k
0
= (w
k
0
, t
p
k
k+1
+1
, t
p
k
k+1
+2
, . . .),
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

f
()
k
if there is a condition r Q
f
(K, ) such that p
k
m

0
r and
(i )(nor[t
r
i
]
1
2
m+1
f(n +k + 2, m
t
r
i
dn
))
and r essentially decides then we choose such an r and we put
p
k
m+1
= (w
k
m+1
, t
r
0
, t
r
1
, . . .),
()
k
if we cannot apply the clause ()
k
then
p
k
m+1
= (w
k
m+1
, half(t
p
k
m
0
), half(t
p
k
m
1
), . . .).
As previously, due to the assumptions about (K, ), we can carry out the con-
struction. Let c
k
: pos(w
p
, s
0
, . . . , s
k
, t
p
k
k
+1
, . . . , t
p
k
k+1
) 2 be a function such
that
c
k
(w
k
m
) =
_
0 if the clause ()
k
was applied to choose p
k
m+1
1 otherwise.
Applying successively

2-bigness (to each t
p
k
i
for i =
k+1
,
k+1
1, . . . ,
k
+ 1) we
nd creatures s
k
+1
, . . . , s
k+1
such that for each i [
k
+ 1,
k+1
]
s
i
(t
p
k
i
) and nor[s
i
] nor[t
p
k
i
] [pos(w
p
, s
0
, . . . , s
k
, t
p
k
k
+1
, . . . , t
p
k
i1
)[
and for each u pos(w
p
, s
0
, . . . , s
k
, s
k
+1
, . . . , s
k+1
) the value of c
k
(u) depends
on u ` m
s
k
up
only. If the constant value of c
k
` pos(v, s
k
+1
, . . . , s
k+1
) (for some
sequence v pos(w
p
, s
0
, . . . , s
k
)) is 0 then easily
(v, s
k
+1
, . . . , s
k+1
, t
p
k
m
k
0
, t
p
k
m
k
1
, . . .) Q
f
(K, )
(remember that (K, ) is nice), it is
0
-stronger than (v, t
p
k
k
+1
, t
p
k
k
+2
, . . .) and it
essentially decides (by 1.2.10). This contradicts the inductive assumption (4). So
the constant value is always 1 and consequently c
k
` pos(w
p
, s
0
, . . . , s
k
, . . . , s
k+1
)
is constantly 1. Put
p
k+1
def
= (w
p
, s
0
, . . . , s
k+1
, t
p
k
m
k
0
, t
p
k
m
k
1
, . . .) Q
f
(K, ).
Note that for all i
nor[t
p
k
m
k
i
] > f(n +k + 1, m
t
p
k
m
k
i
dn
) +
H
(m
t
p
k
m
k
i
dn
) + 2
and thus p
k+1
satises (3). By the construction (and 2.2.9) the condition p
k+1
satises the inductive requirement (4) (for k + 1). Since f is H-fast we have that
for i [
k
+ 1,
k+1
] (by the inductive assumption (3))
nor[t
p
k+1
i
] nor[t
p
k
i
] [pos(w
p
, s
0
, . . . , s
k
, t
p
k
k
+1
, . . . t
p
k
i1
)[ > f(n +k, m
t
p
k+1
i
dn
),
and for i >
k+1
nor[t
p
k+1
i
] > f(n +k + 1, m
t
p
k+1
i
dn
)
and (if k > 0)
nor[t
p
k
k
] = nor[s
k
] > f(n +k, m
t
p
k
k
dn
).
Hence p
k

f
k
p
k+1
. Moreover nor[s
k+1
] > f(n + k + 1, m
s
k+1
dn
). Thus the require-
ments (5) and (6) are satised.
Finally look at the limit condition
p
= (w
p
, s
0
, s
1
, . . .) = lim
k
p
k
Q
f
(K, ).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Now we have to distinguish the two cases: (K, ) is simple and (K, ) is gluing.
If our creating pair is simple then we take a condition r p
which decides the

value of and such that nor[t
r
i
] > 0 for all i . We may assume that for some
k we have w
r
pos(w
p
, s
0
, . . . , s
k
). Then (w
r
, t
p
k
k
+1
, t
p
k
k
+2
, . . .)
0
r and this
contradicts the assumption (4) about p
k
.
Now suppose that (K, ) is gluing. Note that choosing
k+1
we may take it
arbitrarily large. So we may assume that (additionally)
k+1
k
is larger then the
gluing witness for f(k, m
s
k
up
). Then we nd s
k+1
(s
k
+1
, . . . , s
k+1
) such that
nor[s
k+1
] minf(k, m
s
k
up
), nor[s
k
+1
], . . . , nor[s
k+1
] = f(k, m
s
k
up
).
Put s
0
= s
0
and consider the condition
p
= (w
p
, s
0
, s
1
, s
2
, . . .) p
.
Now we nish choosing the r as earlier above p
. The claim is proved.

Claim 2.2.11.2. Suppose that is a Q
f
(K, )-name for an ordinal, n < and
p Q
f
(K, ). Then there is q Q
f
(K, ) such that p
f
n
q and q essentially
decides .
Proof of the claim: Take i
0
< so large that
(i i
0
)(nor[t
p
i
] > f(n + 1, m
t
p
i
dn
)).
Let w
m
: m < m
enumerate pos(w
p
, t
p
0
, . . . , t
p
i0
). Choose inductively q
m
,
m
(for
m < m
) such that
q
0
is given by 2.2.11.1 for (w
0
, t
p
i0+1
, t
p
i0+2
, . . .),
0
= 1
1
=
1
2
|pos(w
0
,t
p
i
0
+1
)|
,
m+1
=
m
2
|pos(wm,t
q
m1
0
)|
(for m > 0),
q
m+1
is given by claim 2.2.11.1 for (w
m+1
, t
qm
0
, t
qm
1
, . . .),
m+1
.
Note that arriving at the stage m+ 1 < m
of this construction we have

[pos(w
0
, t
p
i0+1
)[ +
m
[pos(w
+1
, t
q
0
)[
H
(m
t
qm
1
dn
),
so
m+1
satises the respective demand. Moreover,
nor[t
qm
i
]
m+1
f(n + 1, m
t
qm
i
dn
) > f(n, m
t
qm
i
dn
) for i > 0, m < m
,
and for each m < m
nor[t
qm
0
] nor[t
p
i0+1
] m > f(n, m
t
qm
0
dn
) + 2.
Consequently we may carry out the construction and nally letting
q
def
= (w
p
, t
p
0
, . . . , t
p
i0
, t
q
m
1
0
, t
q
m
1
1
, . . .)
we will clearly have a condition as required in the claim.
Applying inductively claim 2.2.11.2 to
m
and n +m we nish the theorem (using
1.2.13 and 1.2.10).
Corollary 2.2.12. Suppose that a creating pair (K, ) is nitary,

2-big and
has the Halving Property. Further assume that it is either simple or gluing. Let
f : be an H-fast function. Then the forcing notion Q
f
(K, ) is proper.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

2.3. TREECREATING (K, ) 39
2.3. Treecreating (K, )
Proposition 2.3.1. Let e < 3, n < , p Q
tree
e
(K, ) and let A T
p
be an
antichain in T
p
such that ( A)( F
0
n
(p))( < ). Assume that for each
A we have a condition q
Q
tree
e
(K, ) such that p
[]
0
q
and
if e = 1 then t
q
: T
q
& nor[t
q
] n t
p
: T
p
.
Then there exists q Q
tree
e
(K, ) such that p
e
n+1
q, A T
q
, q
[]
= q
for A
and if T
p
is such that there is no A with _ then T
q
and t
p
= t
q
.
Definition 2.3.2. Let (K, ) be a tree-creating pair for H, k < .
(1) A tree creature t K is called k-big if nor[t] > 1 and for every function
h : pos(t) k there is s (t) such that h`pos(s) is constant and
nor[s] nor[t] 1.
(2) We say that (K, ) is k-big if every t K with nor[t] > 1 is k-big.
Remark 2.3.3. The dierence with 2.2.1 is not serious - we could have inter-
fering.
Definition 2.3.4. A tree creating pair (K, ) is t-omittory if for each system
s
:

T) K such that

T is a well founded quasi tree, root(s
) = , pos(s
) =
succ
T
() (for

T) and for every
0

T such that pos(s
0
) max(T) there is
s (s
:

T) such that
nor[s] nor[s
0
] 1 and pos(s) pos(s
0
).
Remark 2.3.5. The name t-omittory comes from treeomittory: it is a
natural notion corresponding to omittory creating pairs for the case of treecreating
pairs. The main point of being t-omittory is that if p, q Q
tree
e
(K, ), p q then
we have a condition r Q
tree
e
(K, ) such that p
e
0
r and dcl(T
r
) dcl(T
q
) and
t
r
= t
q
for each T
r
, root(r) < . [Why? Let = root(q) and let T
T
p
be a
well founded quasi tree such that
(

T
)(succ
T
() = pos(t
p
)), and root(T
) = , and t
q
(t
p
:

T
).
Let T
= root(p)
pos(t
p
) : < & T
p
pos(t
q
). Clearly T
is a well
founded quasi tree and we may apply 2.3.4 to t
q
, t
p
: < & T
p
) and . Thus
we get t
r
root(p)
(t
q
, t
p
: < & T
p
) such that pos(t
r
root(p)
) pos(t
q
). Note
that, by transitivity of , t
r
root(p)
(t
p
:

T
). For pos(t
r
root(p)
) let
t
r
= t
q
and so on. Easily, this denes a condition r as required.]

Moreover this property implies that Q
tree
4
(K, ) is dense in Q
tree
2
(K, ).
Lemma 2.3.6. Suppose (K, ) is a treecreating pair, e < 3, p Q
tree
e
(K, ),
n < and is a Q
tree
e
(K, )-name for an ordinal. Further assume that if e = 2
then (K, ) is bounded (see 1.3.3(4)). Then:
(1) There exist a condition q Q
tree
e
(K, ) and a maximal antichain A T
q
of T
q
such that:
() p
e
n
q,
() for every A the condition q
[]
decides the value of ,
() A is an e-thick antichain of T
p
(see denition 1.3.7).
(2) Assume additionally that either e = 0 and (K, ) is t-omittory or e = 1
and (K, ) is 2-big. Then there are q Q
tree
e
(K, ) and a front F of T
q
satisfying clauses () and () of (1) above.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Proof. 1) Put
A
0
= T
p
: ( F
0
n
(p))( < ) and there is q Q
tree
e
(K, ) such that
p
[]
e
0
q, q decides the value of and
if e = 1 then ( T
q
)(nor[t
q
] > n).
Claim 2.3.6.1. (r Q
tree
e
(K, ))(p r A
0
T
r
,= ).
Proof of the claim: Suppose r Q
tree
e
(K, ) is such that p r. We may
assume that r decides the value of and if e = 1 then additionally we have (
T
r
)(nor[t
r
] > n) (by 1.3.8(3)). Now, as F

0
n
(p) is an e-thick antichain of T
p
(see
1.3.8(4), (6); remember that if e = 2 then we assume that (K, ) is bounded), we
nd T
r
such that ( F
0
n
(p))( < ). Look at the condition r
[]
: it witnesses
that A
0
.
Thus
A
def
= A
0
: there is no
< which is in A
0
is an e-thick antichain of T
p
and for each A we may take a witness q
for
A
0
. Now apply 2.3.1 to nd q Q
tree
e
(K, ) such that p
e
n
q and q
[]
= q
for
A.
2) If e = 0 and (K, ) is t-omittory then for each F
1
n
(p) we may choose a
condition q

0
p
[]
deciding the value of (see remark 2.3.5). Now apply 2.3.1.
Assume now that (K, ) is 2-big, e = 1. Let
A
1
def
= T
p
: ( F
1
n
(p))( _ ) and there is q
0
p
[]
such that
t
q
: T
q
& nor[t
q
] n t
p
: T
p
and
there is a front F of T
q
with ( F)(q
[]
decides ).
Our aim is to show that F
1
n
(p) A
1
which will nish the proof (applying 2.3.1
remember that F
1
n
(p) is a front of T
p
above F
0
n
(p)).
Fix
0
F
1
n
(p). For each
0
_ T
p
such that nor[t
p
] > 1 the creature t

p
is
2-big so there is s
(t
p
) such that nor[s
] nor[t
p
] 1 and
either pos(s
) A
1
= or pos(s
) A
1
.
Claim 2.3.6.2. If
0
_ T
p
, nor[t
p
] n + 2 and pos(s
) A
1
,= then
A
1
.
Proof of the claim: By the the choice of s
we know that then pos(s
) A
1
so for pos(s
) we may choose a condition q
and a front F
T
q
witnessing
A
1
. Look at the quasi tree
T
r
def
= pos(s
)
_
pos(s)
T
q
.
It determines a condition r Q
tree
1
(K, ). It follows from the assumption nor[t
p
]
n + 2 that nor[s
] > n and therefore

t
r
: T
r
& nor[t
r
] n t
p
: T
p
[]
.
Hence the condition r together with F
=

pos(s)
F
(which is clearly a front of

T
r
) witness that A
1
, nishing the proof of the claim.
Now we construct inductively a condition q:
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

2.3. TREECREATING (K, ) 41
root(T
q
) =
0
, T
q
T
p
,
if T
q
and nor[t
p
] < n + 2 then t
q
= t
p
, succ
T
q () = pos(t
p
),
if T
q
and nor[t
p
] n + 2 then t
q
= s
, and succ
T
q () = pos(s
).
It should be clear that q Q
tree
1
(K, ) and
t
q
: T
q
& nor[t
q
] n t
p
: T
p
[
0
]
.
Claim 2.3.6.3. B
def
= T
q
: ( T
q
)( _ nor[t
q
] n +2) A
1
.
Proof of the claim: Suppose that B. Take a condition r Q
tree
1
(K, ),
q
[]
r which decides . We may assume that ( T
r
)(nor[t
r
] n + 2) (by
1.3.8(3)). Consequently = t
r
: T
r
& nor[t
r
] n t
p
: T
p
and
root(r) A
1
. But now we note that for each T
q
, if _ < root(r) then
nor[t
p
] n + 2 (as B) and t
q
= s
. Thus we may apply 2.3.6.2 inductively to

conclude that all these , including , are in A
1
, nishing the proof of the claim.
Claim 2.3.6.4. If T
q
, / A
1
then there is pos(t
q
) such that / A
1
.
Proof of the claim: Should be clear.
Claim 2.3.6.5.
0
A
1
.
Proof of the claim: Assume not. Then we inductively choose a sequence
0
<
1
<
2
< . . . T
q
such that
(i )(
i
/ A
1
& nor[t
q
i+1
] < n + 2).
For this suppose that we have dened
i
/ A
1
. Take
pos(t
q
i
) A
1
(possible by
2.3.6.4). By claim 2.3.6.3 we know that
/ B so there is (T
q
)
[
]
such that
nor[t
q
] < n + 2. Let
i+1
be the shortest such , i.e.
i+1
is such that
[ T
q
&
_ <
i+1
] nor[t
q
] n + 2 and nor[t
q
i+1
] < n + 2.
By repeating applications of 2.3.6.2 we conclude that
i+1
/ A
1
, as otherwise
A
1
.
Now look at the branch through T
q
determined by
i
: i < ) it contradicts
q Q
tree
1
(K, ). This nishes the proof of the claim and the lemma.
Theorem 2.3.7. Suppose (K, ) is a treecreating pair, e < 3, p Q
tree
e
(K, ).
Further suppose that if e = 2 then is bounded. Let n < and let
k
be Q
tree
e
(K, )-
names for ordinals (for k ). Then:
(1) There exist a condition q Q
tree
e
(K, ) and e-thick antichains A
k
T
q
of T
q
such that for each k :
() p
e
n
q,
() for every A
k
the condition q
[]
decides the value of
k
,
() ( A
k+1
)( A
k
)( < ).
(2) If either e = 0 and (K, ) is t-omittory or e = 1 and (K, ) is 2-big then
there are q Q
tree
e
(K, ) and fronts F
k
of T
q
such that for each k
the conditions ()() of (1) above are satised.
Proof. This is an inductive application of 2.3.6 and 2.3.1 (and 1.3.8 + 1.3.11).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Lemma 2.3.8. Let (K, ) be a treecreating pair, be a Q
tree
3
(K, )name for
an ordinal, p Q
tree
3
(K, ), n < . Then there are a condition q Q
tree
3
(K, )
and a 3thick antichain A T
p
of T
p
such that q p and
() ( A)( T
p
)( < T
q
& t
q
= t
p
),
() for every A the condition q
[]
decides the value of ,
() ( A)( T
q
)( _ nor[t
q
] > n).
Proof. Look at the set
B
def
= T
p
: there is a condition q p
[]
such that root(q) = ,
( T
q
)(nor[t
q
] > n) and q decides the name .

Easily, B T
r
,= for every condition r p. Hence the set
A
def
= B : ( B)( < )
is a 3thick antichain of T
p
. Now we nish in a standard way.
Theorem 2.3.9. Let (K, ) be a treecreating pair,
k
be Q
tree
3
(K, )names
for ordinals (for k < ) and p Q
tree
3
(K, ). Then there are a condition q
Q
tree
3
(K, ) and 3thick antichains A
k
of T
p
such that q p and for every k :
() A
k
T
q
,
() if A
k
then q
[]
decides
k
,
() ( A
k+1
)( A
k
)( < ).
Proof. Build the condition q by induction using 2.3.8.
Corollary 2.3.10. If e < 4, (K, ) is a tree creating pair which is bounded if
e = 2, and

i
H(i) is countable then the forcing notion Q
tree
e
(K, ) is proper (and
even more).
Proof. By 2.3.7, 2.3.9 (or rather the proofs of them) and the denition of
thick antichains (remember 1.3.9).
Lemma 2.3.11. Assume that (K, ) is a 2big treecreating pair, n < , and
p Q
tree
1
(K, ). Then there are q Q
tree
1
(K, ) and a front F of T
q
such that
p
1
n
q and
( F)( T
q
)( _ nor[t
q
] > n + 1).
Proof. It is like 2.3.6(2). We consider the set
A
1
def
= T
p
: ( F
1
n
(p))( _ ) and there is q
0
p
[]
such that
t
q
: T
q
& nor[t
q
] n t
p
: T
p
and
there is a front F of T
q
with
( F)( T
q
)( _ nor[t
q
] > n + 1).
We proceed exactly as in 2.3.6(2) to show that F
1
n
(p) A
1
.
Corollary 2.3.12. Suppose that (K, ) is a 2-big tree creating pair, n < ,
p Q
tree
1
(K, ). Then there are q Q
tree
1
(K, ) and fronts F
m
of T
q
(for m )
such that
p
1
n
q and ( T
q
)( F
m
)( _ nor[t
q
] m).
Hence, in particular, Q
tree
4
(K, ) is dense in Q
tree
1
(K, ).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

2.4. EXAMPLES 43
2.4. Examples
In this part we will give several examples of weak creating pairs, putting some
of the known forcing notions into our setting. It seems that the main ingredient of
any application of our technique is, next to an appropriate choice of the function
H, the denition of the norm we use to measure possibilities. Often such a norm
is an application of a particular type of prenorms.
Definition 2.4.1. A function H : T(A) R
0
is a pre-norm on the set A
(or rather T(A)) if
(a) B C A implies H(B) H(C)
(b) H(A) > 0 and if a A then H(a) 1.
A pre-norm H on A is nice if additionally
(c) if B C A, H(C) > 1 then either H(B) H(C) 1 or H(C B)
H(C) 1.
Definition 2.4.2. (1) For a non-empty nite set A we let dp
0
(A) = [A[.
(2) For a nite family A []
<
such that (a A)([a[ > 1) we dene
dp
1
(A) by the following induction
dp
1
(A) 0 always,
dp
1
(A) 1 if A ,= ,
dp
1
(A) n + 2 if for every set X one of the following conditions holds:
dp
1
(a A : a X) n + 1 or
dp
1
(a A : a X) n + 1.
(3) For a non-empty nite family A of non-empty subsets of we let
dp
2
(A) = min[I[ : (a A)(a I ,= ).
(4) For n , i < 3 and A in the domain of dp
i
we let
dp
i
n
= log
2+n
(dp
i
(A)).
Proposition 2.4.3. (1) Let i < 3, n . Suppose that A is a nite set
in the domain of dp
i
such that dp
i
n
(A) > 0. Then dp
i
n
`T(A) is a nice
pre-norm on T(A).
(2) If H is a nice pre-norm on T(A), r < H(A) is a positive real number and
H
r
: T(A) R
0
is dened by
H
r
(B) = max0, H(B) r for B A,
then H
r
is a nice pre-norm on T(A).
Proof. 1) Note that (in all cases), if B, C A then
dp
i
(B C) dp
i
(B) + dp
i
(C).
The only unclear instance here might be i = 1, but note that if

A [m
0
, m
1
),
B A then
dp
1
(B) k + 2 if and only if
for every partition I
: <
0
) of [m
0
, m
1
),
0
2
k+1
, there are
b B and <
0
such that b I
.
2) Check.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Remark 2.4.4. 2.4.3(2) will be of special importance when dening creating
pairs with the Halving Property. Then we will use H
r
for r =
1
2
H(A)| (see 2.4.6,
4.4.2 and 7.5.1).
Our rst example of a creating pair recalls BlassShelah forcing notion applied
in [BsSh 242] to show the consistency of the following statement:
if T
1
, T
2
are non-principal ultralters on
then there is a nite-to-one function f
such that f(T

1
) =
f(D
2
).
(The suitable model was obtained there by a countable support iteration of forcing
notions close to Q
s
(K
2.4.5
,
2.4.5
) over a model of CH.)
Example 2.4.5. Let H(m) = 2 for m . We build a full, omittory and
omittorybig (and smooth) creating pair (K
2.4.5
,
2.4.5
) for H.
Construction. A creature t CR[H] is in K
2.4.5
if m
t
dn
+2 < m
t
up
and there
is a sequence A
t
u
: u

i<m
t
dn
H(i)) such that for every u

i<m
t
dn
H(i):
() A
t
u
is a non-empty family of subsets of [m
t
dn
, m
t
up
), each member of A
t
u
has at least 2 elements,
() u, v) val[t] if and only if
u < v

i<m
t
up
H(i) and i [m
t
dn
, m
t
up
) : v(i) = 1 A
t
u
,
() nor[t] = mindp
1
0
(A
t
u
) : u
i<m
t
dn
H(i).
[Note that we do not specify here what are the dis[t] for t K
2.4.5
. We have a
total freedom in this, we may allow all possible values of dis[t] to appear.]
2.4.5
on K
2.4.5
is dened as follows. Suppose that
t
0
, . . . , t
n
K
2.4.5
are such that m
ti
up
= m
ti+1
dn
for i < n. Then
s
2.4.5
(t
0
, . . . , t
n
) if and only if
s K
2.4.5
, m
s
dn
= m
t0
dn
, m
s
up
= m
tn
up
and for every u, v) val[s]
for each i n we have v`m
ti
dn
, v`
ti
up
) val[t
i
].
It is an easy exercise to check that (K
2.4.5
,
2.4.5
) is a full, omittory, smooth and
omittorybig creating pair (for the last property use 2.4.3). Note that the forcing
notion Q
s
(K
2.4.5
,
2.4.5
) is non-trivial as for each m
0
< m
0
+ 2
n+1
< m
1
there is
t K
2.4.5
CR
m0,m1
[H] such that nor[t] = log
2
(n + 1).
One may consider a modication of (K
2.4.5
,
2.4.5
) making it forgetful. For
this we let K
2.4.5
= t K
2.4.5
: (u
0
, u
1

i<m
t
dn
H(i))(A
t
u0
= A
t
u1
) and
2.4.5
(t
0
, . . . , t
n
) =
2.4.5
(t
0
, . . . , t
n
)K
2.4.5
(for suitable t
0
, . . . , t
1
K
2.4.5
). Check
that (K
2.4.5
,
2.4.5
) is a forgetful, omittory and omittorybig creating pair.
Example 2.4.6. We dene functions H : and f : and a
creating pair (K
2.4.6
,
2.4.6
) for H such that:
f is Hfast,
(K
2.4.6
,
2.4.6
) is

2big, forgetful, simple and has the Halving Property,
f
(K
2.4.6
,
2.4.6
) is nontrivial.
Construction. Let F
be an increasing function. Dene inductively

functions H = H
F
and f = f
F
such that for each n, k, :
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

2.4. EXAMPLES 45
(i) H(n) = F(
H
(n)) 2
f(n,n)
(remember
H
(n) =

i<n
[H(i)[,
H
(0) = 1),
(ii) f(0, ) = + 1, f(k + 1, ) = 2
H()
(f(k, ) +F(
H
())
H
() + 2)
(note that (i)+(ii) uniquely determine H and f and f is Hfast).
A creature t CR[H] belongs to K
2.4.6
if m
t
up
= m
t
dn
+ 1 and
dis[t] = m
t
dn
, A
t
, H
t
), where A
t
is a subset of H(m
t
dn
) and H
t
: T(A
t
)
is a nice pre-norm,
val[t] = u, v)
i<m
t
dn
H(i)
im
t
dn
H(i) : u < v & v(m
t
dn
) A
t
,
nor[t] = H
t
(A
t
).
For t K
2.4.6
let
2.4.6
(t) consist of all s K
2.4.6
such that m
s
dn
= m
t
dn
, A
s

A
t
and (B A
s
)(H
s
(B) H
t
(B)). (And if o K
2.4.6
, [o[ ,= 1 then we let
2.4.6
(o) = .) Next, for t K
2.4.6
we dene half(t) K
2.4.6
as follows:
if nor[t] < 2 then half(t) = t,
if nor[t] 2 then half(t) K
2.4.6
is (the unique creature) such that
m
half(t)
dn
= m
t
dn
, A
half(t)
= A
t
, and H
half(t)
= (H
t
)
r
(see 2.4.3(2)),
where r =
1
2
nor[t]|.
It should be clear that (K
2.4.6
,
2.4.6
) is a forgetful, simple and

2big creating pair
(for the last remember the denition of nice pre-norms). Moreover, the function half
witnesses that (K
2.4.6
,
2.4.6
) has the weak Halving Property (and so the Halving
Property). [Why? Note that if s
2.4.6
(half(t)), nor[s] > 0, nor[t] 2 then
A
s
A
t
and
1 H
s
(A
s
) H
half(t)
(A
s
) = H
t
(A
s
)
1
2
nor[t]|.
Thus H
t
(A
s
)
1
2
nor[t]| + 1 >
1
2
nor[t] 1. Now look at a creature t
K
2.4.6
such that dis[t
] = m
t
dn
, A
s
, H
t
`T(A
s
)).] Finally note that if m < and t K
2.4.6
is such that dis[t] = m, H(m), dp
0
0
`T(H(m))) then nor[t] > f(m, m).
Note that the creating pair (K, ) described in the Prologue to represent the
Silver forcing Q below 2
n
is an example of a nitary creating pair which captures
singletons.
The rst serious application of treecreating pairs appeared in [Sh 326]. The
forcing notion LT
f
d
constructed there was later modied in various ways and several
variants of it found their applications (see e.g. [BJSh 368], [FrSh 406] and 2.4.8
below). This forcing notion is essentially the Q
tree
1
(K
0
2.4.7
,
0
2.4.7
). (One should note
similarities with the forcing notion Q
s
(K
2.4.5
,
2.4.5
).)
Example 2.4.7. Let f
be a strictly increasing function, H(m) = f(m)+1

for m . We construct nitary treecreating pairs (K
2.4.7
,
2.4.7
), < 4, for H
such that
(1) K
0
2.4.7
= K
1
2.4.7
, K
2
2.4.7
= K
3
2.4.7
,
(2) (K
0
2.4.7
,
0
2.4.7
), (K
1
2.4.7
,
1
2.4.7
) are 2big local treecreating pairs,
(3) (K
2
2.4.7
,
2
2.4.7
), (K
3
2.4.7
,
3
2.4.7
) are 2big t-omittory treecreating pairs.
Construction. First we dene (K
0
2.4.7
,
0
2.4.7
). A tree creature t TCR
[H]
(where

n<
i<n
H(i)) is in K
0
2.4.7
if
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

dis[t] = , A
t
, H
t
), where A
t
H(g()) and H
t
is a nice pre-norm on
T(A
t
),
val[t] = , ) : < & g() = g() + 1 & (g()) A
t
,
nor[t] = H
t
(A
t
).
The operation
0
2.4.7
is the trivial one and for t TCR
[H] K
0
2.4.7
:
0
2.4.7
(t) = s TCR
[H] K
0
2.4.7
: A
s
A
t
& H
s
= H
t
`T(A
s
).
Easily, (K
0
2.4.7
,
0
2.4.7
) is a nitary 2big simple treecreating pair.
To dene
1
2.4.7
on K
1
2.4.7
= K
0
2.4.7
we let for t TCR
[H]:
1
2.4.7
(t) = t TCR
[H] K
1
2.4.7
: A
s
A
t
& (B A
s
)(H
s
(B) H
t
(B)).
Plainly, (K
1
2.4.7
,
1
2.4.7
) is a nitary 2big simple treecreating pair too.
To have t-omittory variants of the tree-creating pairs dened above we declare
that a treecreature t TCR
[H] is in K
2
2.4.7
= K
3
2.4.7
if
dis[t] = ,
t
, A
t
, H
t
), where _
t

n<
i<n
H(i), A
t
H(g(
t
)) and
H
t
is a nice pre-norm on T(A
t
),
val[t] = , ) :
t
< & g() = g(
t
) + 1 & (g(
t
)) A
t
,
nor[t] = H
t
(A
t
).
The operations
2
2.4.7
,
3
2.4.7
are such that if T is a well founded quasi tree, s
:

T) is a system of treecreatures from K
2
2.4.7
such that
(

T)(s
TCR
[H] & rng(val[s
]) = succ
T
())
then
2
2.4.7
(s
:

T) consists of all tree-creatures s K
2
2.4.7
TCR
root(T)
[H] such
that for some
0

T we have
rng(val[s]) rng(val[s
0
]) max(T) and H
s
= H
s
0
`T(A
s
),
and
3
2.4.7
(s
:

T) is dened in a similar manner but we replace the last demand
(on H
s
) by (B A
s
)(H
s
(B) H
s
0
(B)). Now check that
2
2.4.7
,
3
2.4.7
are t-
omittory 2big tree compositions on K
2
2.4.7
= K
3
2.4.7
.
Let us nish our overview of classical examples recalling FremlinShelah forc-
ing notion. This forcing notion is essentially Q
tree
1
(K
2.4.8
,
2.4.8
), and it is a relative
of Q
tree
1
(K
1
2.4.7
,
1
2.4.7
). It was applied in [FrSh 406] to construct a model in which
there is a countable relatively pointwise compact set of Lebesgue measurable func-
tions which is not stable.
Example 2.4.8. We build a function H and a nitary, local 2big treecreating
pair (K
2.4.8
,
2.4.8
).
Construction. Choose inductively increasing sequences n
k
: k < ) and
m
k
: k < ) such that n
0
= m
0
= 4, m
k+1
> m
k
2
n
k
and
(n
k+1
)
(m
k+1
)
(k+6)
(m
k+1
)
(k+6)
> log
2
(n
k+1
), n
k+1
> 2
(m
k+1
)
k+6
(m
k+1
)
k+6
.
For i let H(i) = a n
i
:
|a|
ni
1
1
2
i+2
.
A treecreature t TCR
[H] is taken to K
2.4.8
if
dis[t] = , A
t
), where A
t
H(g()) is non-empty,
val[t] = , ) : < & g() = g() + 1 & (g()) A
t
,
nor[t] =
g()+1
log
2
(n
g()
)(g()+2)
dp
2
0
(A
t
).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

2.4. EXAMPLES 47
[Note that dp
2
0
(H(i)) log
2
(n
i
) (i + 2).] The operation
2.4.8
is trivial and for
t TCR
[H] K
2.4.8
2.4.8
(t) = s K
2.4.8
TCR
[H] : A
s
A
t
.
Check that (K
2.4.8
,
2.4.8
) is a local 2big treecreating pair.
Remark 2.4.9. (1) Note that if

W is the Q
tree
1
(K
2.4.8
,
2.4.8
)name for
the generic real (see 1.1.13), then
'
Q
tree
1
(K2.4.8,2.4.8)
(i )(

W(i) n
i
&
|

W(i)|
ni
1
1
2
i+2
) and
( V

i<
n
i
)(
i )((i) /

W(i)) .
Thus, after forcing with Q
tree
1
(K
2.4.8
,
2.4.8
), the ground model reals are
of measure zero.
(2) One can dene a t-omittory variant of (K
2.4.8
,
2.4.8
) (similarly to the
denition of the pair (K
2
2.4.7
,
2
2.4.7
); in forcing this would correspond to
considering Q
tree
0
(K
2.4.8
,
2.4.8
)).
(3) In practical applications, forcing notions of the type Q
tree
0
(K, ) can be
represented in an equivalent form as Q
tree
1
(K
) for some t-omittory

pair (K
).
In the next two examples we want to show that the choice of the type of forcing
notion or the norm condition may be very crucial. Even if we use the same or very
similar weak creating pairs, dierent approaches may result in forcing notions with
extremely dierent properties.
Example 2.4.10. There exists a nitary, local and 2big treecreating pair
(K
2.4.10
,
2.4.10
) such that the forcing notion Q
tree
1
(K
2.4.10
,
2.4.10
) is
bounding
but the forcing notion Q
tree
3
(K
2.4.10
,
2.4.10
) adds an unbounded real.
Construction. Let H(i) = 2
i+1
for i . A treecreature t TCR
[H] is
taken to K
2.4.10
if
dis[t] = , A
t
) for some A
t
H(g()),
val[t] = , ) : < & g() = g() + 1 & (g()) A
t
,
nor[t] = dp
0
0
(A
t
).
The operation
2.4.10
is trivial and
2.4.10
(t) = s K
2.4.10
: val[s] val[t].
Plainly, (K
2.4.10
,
2.4.10
) is a nitary, local and 2big tree creating pair. By
2.3.7(2) we conclude that Q
tree
1
(K
2.4.10
,
2.4.10
) is
bounding (compare 3.1.1).

Note that, by 2.3.12, Q
tree
4
(K
2.4.10
,
2.4.10
) is dense in Q
tree
1
(K
2.4.10
,
2.4.10
).
Suppose now that p Q
tree
3
(K
2.4.10
,
2.4.10
). By induction on i < we build
an increasing sequence n
i
: i < ) , a condition q Q
tree
3
(K
2.4.10
,
2.4.10
) and
a function f : T
q
: (i < )(g() = n
i
) such that
() q p and root(q) = root(p),
() n
0
= g(root(q)), f(root(q)) = 0,
() if , T
q
, g() = n
i
, g() = n
i+1
and < then f() f(), f() +
1,
() for each T
q
such that g() = n
i
there is exactly one T
q
such that
< , g() = n
i+1
and f() = f() + 1,
() if T
q
, n
i
g() < n
i+1
then nor[t
q
] = f(`n
i
).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

The construction is quite straightforward. Suppose we have dened n
i
, T
q
n
i
T
p

n
i
and f`T
q

n
i
in such a way that
( T
q

n
i
)( T
p
)( _ nor[t
p
] f()).
By the denition of Q
tree
3
(K
2.4.10
,
2.4.10
) we nd n
i+1
> n
i
such that for each
T
q

n
i
there is
T
p

n
i+1
such that <
and
( T
p
)(
_ nor[t
p
] f() + 1)
(remember T
q

n
i
is nite). Now, for each T
q

n
i
we continue building
the condition q above in such a manner that each t
q
(for n
i
g() < n
i+1
) has
norm f() and the
is taken to T
q
. Declare f(
) = f() + 1 and f(
) = f()
for all
T
q

n
i+1
extending but dierent from
.
It is easy to check that q built in this manner is a condition in Q
tree
3
(K
2.4.10
,
2.4.10
)
stronger than p.
Note that for each m the set T
q
: dom(f) & f() = m+1 is the
B
m
(q) (and thus it is a 3thick antichain of T
q
). We will be done if we show the
following claim.
Claim 2.4.10.1. If q r Q
tree
3
(K
2.4.10
,
2.4.10
) then for some m the set
T
r
: dom(f) & f() = m is innite.
Proof of the claim: Choose T
r
such that
( T
r
)( _ nor[t
r
] 2) and g() = n
i
(for some i ).
Let m = f() + 1. Note that if T
r
, g() = n
j
n
i
, _ and f() = f()
then:
(1) [
T
r
: <
& g(
) = n
j+1
[ 4,
(2) there is at most one
T
r
such that
g(
) = n
j+1
, <
, and f(
) = m,
(3) there are j
> j and
T
r
such that
<
, g(
) = n
j
and f(
) = m.
Hence the set T
r
: _ & dom(f) & f() = m is innite.
Remark 2.4.11. Note that the proof that Q
tree
3
(K
2.4.10
,
2.4.10
) adds an un-
bounded real does not use the specic form of (K
2.4.10
,
2.4.10
). With not much
changes we may repeat it for any local tree creating pair (K, ) such that
(1) if pos(t), t K and m nor[t], m
then there is s (t) such that nor[s] = m and pos(s), and
(2) if nor[t] > 2 then [pos(t)[ > 2.
In the last example of this section we try to show the dierence between the
use of tree creating pairs and that of creating pairs.
Example 2.4.12. Let n
i
: i < ) , f : and H : []
<
be such that
(1) f(0, ) = + 1, f(k + 1, ) = 2
H()+1
(f(k, ) +
H
() + 2),
(2) n
0
= 0, n
i+1
(n
i
+ 1) 2
2
2f(i,i)
H
(i)
2
+2
+n
i
,
(3) H(i) = [[n
i
, n
i+1
)]
n
i
+ 1
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

2.4. EXAMPLES 49
(so f is Hfast).
We construct weak creating pairs (K
0
2.4.12
,
0
2.4.12
) and (K
1
2.4.12
,
1
2.4.12
) for H such
that
(a) (K
0
2.4.12
,
0
2.4.12
) is a 2big, local and nitary tree creating pair,
(b) (K
1
2.4.12
,
1
2.4.12
) is a simple,

2big, nitary and forgetful creating pair
with the Halving Property,
(c) if P is either Q
tree
1
(K
0
2.4.12
,
0
2.4.12
) or Q
f
(K
1
2.4.12
,
1
2.4.12
) and

W is the
corresponding name for the generic real (see 1.1.13) interpreted as an
innite subset of , then
'
P
(i )([

W [n
i
, n
i+1
)[ = n
i
+ 1) and
(X[]
V)(
i )(

W [n
i
, n
i+1
) X or

W [n
i
, n
i+1
) X = ).
Construction. We try to dene minimal forcing notions adding a set

W
[]
with the property stated in the clause (c). The most natural way is to use
weak creatures giving approximations to

W with norms related to dp
1
.
Dening (K
0
2.4.12
,
0
2.4.12
) we may follow the simplest possible pattern presented
already in 2.4.8 and 2.4.10. So a treecreature t TCR
[H] is in K
0
2.4.12
if
dis[t] = , A
t
), where A
t
H(g()),
val[t] = , ) : < & g() = g() + 1 & (g()) A
t
,
nor[t] =
dp
1
0
(At)
2H(g())
2
.
The operation
0
2.4.12
is trivial (and so s
0
2.4.12
(t) if and only if val[s] val[t]).
One easily checks that (K
0
2.4.12
,
0
2.4.12
) is a local 2big and nitary tree creating
pair. Note that (see the proof of 2.4.3)
dp
1
(H(i)) 2
2H(i)
2
f(i,i)
+ 1 and thus dp
1
0
(H(i)) > f(i, i) 2
H
(i)
2
.
Consequently, Q
tree
1
(K
0
2.4.12
,
0
2.4.12
) is a non-trivial forcing notion. Checking that
it satises the demand (c) is easy if you remember the denition of dp
1
.
Now we want to dene a creating pair (K
1
2.4.12
,
1
2.4.12
) in a similar way as
(K
0
2.4.12
,
0
2.4.12
). However, we cannot just copy the previous case (making suitable
adjustments) as we have to get a new quality: the Halving Property. But we use
2.4.3(2) for this. Thus a creature t CR[H] is taken to K
1
2.4.12
if m
t
up
= m
t
dn
+ 1
and
dis[t] = m
t
dn
, B
t
, r
t
), where B
t
H(m
t
dn
) and r
t
is a non-negative real,
val[t] = (u, v)

i<m
t
dn
H(i)

im
t
dn
: u < v & v(m
t
dn
) B
t
,
nor[t] = max0,
dp
1
0
(Bt)
2H(m
t
dn
)
2
r
t
.
The operation
1
2.4.12
is dened by:
1
2.4.12
(t) = s K
1
2.4.12
: m
t
dn
= m
s
dn
& B
s
B
t
& r
s
r
t
.
It is not dicult to verify that (K
1
2.4.12
,
1
2.4.12
) is a nitary, forgetful, simple and
2big creating pair (remember 2.4.3(2)). Let half : K

1
2.4.12
K
1
2.4.12
be such that
s = half(t) if and only if m
s
dn
= m
t
dn
, B
s
= B
t
and r
s
= r
t
+
1
2
nor[t].
We claim that the function half witnesses the Halving Property for (K
1
2.4.12
,
1
2.4.12
).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Clearly half(t) (t) and
nor[half(t)] = max0,
dp
1
0
(Bt)
2H(m
t
dn
)
2
r
t

1
2
nor[t] =
max0,
dp
1
0
(Bt)
2H(m
t
dn
)
2
r
t

1
2

dp
1
0
(Bt)
2H(m
t
dn
)
2
+
1
2
r
t
=
1
2
max0,
dp
1
0
(Bt)
2H(m
t
dn
)
2
r
t
=
1
2
nor[t].
Suppose now that t
0
K
1
2.4.12
, nor[t
0
] 2 and t
1
2.4.12
(half(t
0
)) is such that
nor[t] > 0. Then m
t
dn
= m
t0
dn
, B
t
B
t0
and r
t
r
t0
+
1
2
nor[t
0
]. Let s
1
2.4.12
(t
0
) be such that B
s
= B
t
and r
s
= r
t0
. Clearly val[s] = val[t] and nor[s] =
max0,
dp
1
0
(Bs)
2H(m
s
dn
)
2
r
t0
. But we know that
0 < nor[t] = max0,
dp
1
0
(Bt)
2H(m
t
dn
)
2
r
t
=
dp
1
0
(Bs)
2H(m
s
dn
)
2
r
t

dp
1
0
(Bs)
2H(m
s
dn
)
2
r
t0

1
2
nor[t
0
]
and hence
1
2
nor[t
0
] nor[s].
Moreover, by standard arguments, the forcing notion Q
f
(K
1
2.4.12
,
1
2.4.12
) is not
trivial and satises the demand (c).
Let us try to show what may distinguish the two forcing notions. We do not
have a clear property of the extensions, but we will present a technical hint that
they may work dierently. Let us start with noting the following property of the
pre-norm dp
1
.
Claim 2.4.12.1. Suppose that A
0
, . . . , A
k1
[]
<
are nite families of sets
with at least 2 elements, m >
k(k+3)
2
and dp
1
(A
i
) m for each i < k. Then there
is a set X such that for each i < k
both dp
1
(a A
i
: a X) m
k(k+3)
2
and dp
1
(a A
i
: a X = ) m
k(k+3)
2
.
Proof of the claim: We prove the claim by induction on k.
Step k = 1.
We have A
0
T([m
0
, m
1
)) such that dp
1
(A
0
) m > 2. Take a set X [m
0
, m
1
)
of the smallest possible size such that dp
1
(a A
0
: a X) m 1. Pick any
point n X and let Y = X n. Then
dp
1
(a A
0
: a Y ) < m1 and dp
1
(a A
0
: a n) = 0.
Hence we get dp
1
(a A
0
: a X = ) m2 (remember the characterization
of dp
1
from the proof of 2.4.3(1)). Consequently, the set X is as required.
Step k + 1.
Suppose A
0
, . . . , A
k1
, A
k
T([m
0
, m
1
)) are such that dp
1
(A
i
) m >
(k+1)(k+4)
2
.
Let X
0
[m
0
, m
1
) be such that for each i < k
dp
1
(a A
i
: a X
0
) m
k(k+3)
2
and
dp
1
(a A
i
: a X
0
= ) m
k(k+3)
2
(exists by the inductive hypothesis). Since dp
1
(A
k
) m, one of the following holds:
dp
1
(a A
k
: a X
0
) m1 or dp
1
(a A
k
: a X
0
= ) m1.
We may assume that the rst takes place. Take X
1
X
0
such that both
dp
1
(a A
k
: a X
1
) m3 and dp
1
(a A
k
: a X
0
X
1
) m3.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

2.4. EXAMPLES 51
Let I
0
= i < k : dp
1
(a A
i
: a X
0
X
1
) m
k(k+3)
2
1. If I
0
= k then
we nish this procedure, otherwise we x i
0
k I
0
and we choose a set X
2
X
1
such that
dp
1
(a A
k
: a X
2
) m5,
dp
1
(a A
k
: a X
1
X
2
) m5, and
dp
1
(a A
i0
: a X
1
X
2
) m
k(k+3)
2
2.
We let I
1
= i < k : i I
0
or dp
1
(a A
i
: a X
1
X
2
) m
k(k+3)
2
2. Note
that I
0
I
1
. We continue in this fashion till we get I
= k. Note that this has to

happen for some k. Look at the set X
+1
constructed at this stage. It has the
property that
dp
1
(a A
k
: a X
+1
) m(2k + 1),
dp
1
(a A
k
: a X
0
X
+1
) m(2k + 1), and for i < k
dp
1
(a A
i
: a X
0
X
+1
) m
k(k+3)
2
(k + 1) m
(k+1)(k+4)
2
.
So let X = X
0
X
+1
and check that it is as required for A
0
, . . . , A
k
(and k + 1).
Now we may show an extra property of

W which we may get in the case of
Q
tree
1
(K
0
2.4.12
,
0
2.4.12
).
Claim 2.4.12.2. The following holds in V
Q
tree
1
(K
0
2.4.12
,
0
2.4.12
)
:
there are sequences i
k
: k < ), X
i
: i < ) from V such that
(1) i
0
< i
1
< . . . < , X
i
[n
i
, n
i+1
),
(2) for each k , for exactly one i [i
k
, i
k+1
) we have

W [n
i
, n
i+1
) X
i
,
(3) the set i :

W [n
i
, n
i+1
) X
i
is not in V.
[The last demand is to avoid a triviality like X
i
, [n
i
, n
i+1
).]
Proof of the claim: Let p Q
tree
1
(K
0
2.4.12
,
0
2.4.12
). We may assume that (
T
p
)(nor[t
p
] > 4). Let i g(root(p)) and let T

p
, g() = i. Then
dp
1
(A
t
p
) > 2
8H(i)
2
>

H
(i)(
H
(i) + 3)
2
.
Since [ T
p
: g() = i[
H
(i), we may use 2.4.12.1 to nd a set X
i

[n
i
, n
i+1
) such that for every T
p
with g() = i we have both
dp
1
(a A
t
p
: a X
i
) dp
1
(A
t
p
)
H(i)(H(i)+3)
2
and
dp
1
(a A
t
p
: a X
i
= ) dp
1
(A
t
p
)
H(i)(H(i)+3)
2
.
Note that
log
2
(dp
1
(A
t
p
)
H(i)(H(i)+3)
2
)
2
H
(i)
2
nor[t
p
] 1.
Therefore we may inductively build a condition q Q
tree
1
(K
0
2.4.12
,
0
2.4.12
) and a
sequence i
k
: k < ) such that
(1) p q, root(q) = root(p), g(root(q)) = i
0
< i
1
< i
2
< . . . < ,
(2) for each T
q
, if g() = i then either
A
t
q
= a A
t
p
: a X
i
= or A
t
q
= a A
t
p
: a X
i
(and so nor[t
q
] nor[t
p
] 1),
(3) for each T
q
with g() = i
k
there is exactly one i = i() [i
k
, i
k+1
)
such that
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

if < T
q
, g() = i
k+1
and
0
= ì
then A
t
q
0
= a A
t
p
0
: a X
i
,
and for distinct as above the values of i() are distinct.
Now check that the condition q forces that X
i
: i < ) and i
k
: k < ) are as
required.
Finally look at 2.4.12.2 in the context of the forgetful creating pair (K
1
2.4.12
,
1
2.4.12
).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

CHAPTER 3
More properties
While the properness is the rst property we usually ask for when building a
forcing notion, the next request is preserving some properties of ground model reals.
In this chapter we start investigations in this direction dealing with three properties
of this kind. We formulate conditions on weak creating pairs which imply that the
corresponding forcing notions: do not add unbounded reals, preserve nonnull sets
or preserve nonmeager sets. Applying the methods developed here we answer
Bartoszy nskis request (see [Ba94, Problem 5]), building a proper forcing notion P
which
(1) preserves nonmeager sets, and
(2) preserves nonnull sets, and
(3) is
-bounding, and
(4) does not have the Sacks property.
A forcing notion with these properties is associated with the conality of the null
ideal (see [BaJu95]). The construction is done in 3.5.1, 3.5.2 and it fullls promise
of [BaJu95, 7.3A].
3.1. Old reals are dominating
Recall that a forcing notion P is
bounding if it does not add unbounded

reals, i.e.
'
P
(x
)(y
V)(
n)(x(n) < y(n)).

Any countable support iteration of proper
bounding forcing notions is
bounding (see [Sh:f, Ch VI, 2.8AC, 2.3]).

Conclusion 3.1.1. Suppose that (K, ) is a nitary tree-creating pair.
(1) If (K, ) is 2-big then the forcing notion Q
tree
1
(K, ) is
-bounding.
(2) If (K, ) is t-omittory then the forcing notion Q
tree
0
(K, ) is
-bounding.
Proof. By 2.3.7(2) and 1.3.8(5).
Conclusion 3.1.2. Let (K, ) be a nitary creating pair for H, and let f :
be an H-fast function. Suppose that (K, ) is

2-big, has the Halving
Property and is either simple or gluing. Then the forcing notion Q
f
(K, ) is
-
bounding.
Proof. By 2.2.11.
Conclusion 3.1.3. Let (K, ) be a nitary creating pair which captures sin-
gletons. Then the forcing notion Q
w
(K, ) is
-bounding.
Proof. By 2.1.12.
53
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

54 3. MORE PROPERTIES
3.2. Preserving non-meager sets
An important question concerning forcing notions is if large sets of reals from
the ground model remain large after the forcing. Here we interpret large as
non-meager. Preserving this property in countable support iteration is relatively
easy. Any countable support iteration of proper
bounding forcing notions which

preserve non-meager sets is of the same type (see [BaJu95, 6.3.21, 6.3.22]). If
we omit
bounding then we may consider a condition slightly stronger than

preserving non-meager sets:
Definition 3.2.1. Let P be a proper forcing notion. We say that P is Cohen
preserving if
meager
P
for every countable elementary submodel N of (H(), , <
), a condition
p P and a real x 2
such that p, P, . . . N and x is a Cohen real over

N, there is an (N, P)generic condition q P stronger than p such that
q '
P
x is a Cohen real over N[
P
].
In practice, forcing notions preserving non-meagerness of sets from the ground
model are Cohenpreserving. Now, to deal with iterations we may use [Sh:f, Ch
XVIII, 3.10] (considering (

R, S, g) as there with g
a
being a Cohen real over a).
Theorem 3.2.2. Suppose that

i
H(i) is countable and (K, ) is a t-omittory
treecreating pair. Then the forcing notion Q
tree
0
(K, ) is Cohenpreserving.
Proof. Suppose that N (H(), , <
) is countable, H, K, , p, . . . N,
p Q
tree
0
(K, ) and x2
is a Cohen real over N. Let

n
: n < ),
k
n
i
: i < ):
n < ),
n
i
: i < ) : n < ) list all Q
tree
0
(K, )names from N for ordinals and
sequences of integers and sequences of nite functions, respectively, such that for
each n < :
(a) '
Q
tree
0
(K,)
the sequence
k
n
i
: i < ) is strictly increasing,
(b) '
Q
tree
0
(K,)
(i < )(
n
i
: [
k
n
i
,

k
n
i+1
) 2).
Thus each
k
n
i
,
n
i
: i < ) is essentially a name for a canonical co-meager set
y 2
: (
i)(y`[
k
n
i
,

k
n
i+1
) =
n
i
).
Of course, the enumerations are not in N, but their initial segments are there.
Claim 3.2.2.1. Suppose that q Q
tree
0
(K, ) N, T
q
, n . Then there
exists a condition q
n
Q
tree
0
(K, ) N such that
(1) q
[]
0
0
q
n
,
(2) nor[t
q
n
] > n,
(3) q
n
decides the values of

m
for m n,
(4) for every m n:
q
n
' (i
0
< . . . < i
n
< )
_
x`[
k
m
i0
,

k
m
i0+1
) =
m
i0
& . . . & x`[
k
m
in
,

k
m
in+1
) =
m
in
_
.
Proof of the claim: First, using 2.3.7(2), choose a condition q
Q
tree
0
(K, ) N
such that q
[]
q
, q

m
for m n and for some fronts F
j
of T
q
(for j < ) we have

() (j )(
0
F
j+1
)(
1
F
j
)(
1
<
0
),
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

3.2. PRESERVING NON-MEAGER SETS 55
() for each m n, j , F
j
and i j, the condition (q
)
[]
decides the
values of

k
m
i
,
m
i
,
() F
j
: j < ) N
(remember that N is an elementary submodel of (H(), , <
)). Now take an

innite branch N lim(T
q
) through the quasi tree T

q
. Take an increasing
sequence
j
: j < ) N of integers such that `
j
F
j
. Then, by (), we have
two sequences k
m
j
: m n, j < ) and
m
j
: m n, j < ), both in N, such that
for m n, j < :
(q
)
[j]
'
Q
tree
0
(K,)

k
m
j
= k
m
j
&
m
j
=
m
j
.
Look at the set
y 2
: (m n)(
j)(y`[k
m
j
, k
m
j+1
) =
m
j
).
It is a dense
0
2
set (coded) in N and therefore x belongs to it. Hence we nd
j
< such that

(m n)([j < j
: x`[k
m
j
, k
m
j+1
) =
m
j
[ > n).
Now take
T
(q
)
[
j
]
such that nor[t
q
] > n + 1. Since (K, ) is t-omittory

we nd a condition q
n
Q
tree
0
(K, ) N such that
q
[]
0
0
q
n
, pos(t
q
n
) pos(t
q
), nor[t
q
n
] > n, and
t
q
n
= t
q
for all T
q
n
, < (compare 2.3.5). Now one easily checks that q

n
is
as required in the claim.
We inductively build a sequence q
n
: n < ), a condition q Q
tree
0
(K, ) and
an enumeration
n
: n < ) of T
q
such that for all m, n :
(1)
n
<
m
n < m,
(2)
n
: n < ) T
p
,
(3) q
n
Q
tree
0
(K, ) N, p
[n]
0
0
q
n
and if
m
<
n
(so m < n) and
n
pos(t
qm
m
) then (q
m
)
[n]
0
0
q
n
,
(4) nor[t
qn
n
] > n,
(5) the condition q
n
m
for m n,
(6) for every m n:
q
n
' (i
0
< . . . < i
n
< )
_
x`[
k
m
i0
,

k
m
i0+1
) =
m
i0
& . . . & x`[
k
m
in
,

k
m
in+1
) =
m
in
_
,
(7) t
q
n
= t
qn
n
.
The construction is actually described by the conditions above: with a suitable
bookkeeping we build sequences
n
: n ) and A
n
: n < ) T(). Arriving at
the stage n of the construction we know
m
for m n and q
m
for m < n. Applying
3.2.2.1 we nd q
n
Q
tree
0
(K, ) N such that the requirements (3)(6) above are
satised. Next we choose A
n

m<n
A
m
of size [pos(t
qn
n
)[ and we assign numbers
from A
n
to elements of pos(t
qn
n
) in such a way that pos(t
qn
n
) =
k
: k A
n
, the
set

mn
A
m
is innite and min(

m<n
A
m
) A
n
.
The q constructed above is a condition in Q
tree
0
(K, ) due to (4), and it is stronger
than p by (3) and (7). Clearly q is not in N, but as every nite step of the
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

construction takes place in N, the condition q is (N, Q
tree
0
(K, ))generic (by (5)).
Moreover, by (6),
q '
Q
tree
0
(K,)
(m < )(
i)(x`[
k
m
i
,

k
m
i+1
) =
m
i
),
what implies that
q '
Q
tree
0
(K,)
x is a Cohen real over N[
Q
tree
0
(K,)
].
This nishes the proof.
The denition 3.2.3 below was inspired by 3.2.2 and its proof. We distinguish
here the two cases: (K, ) is a creating pair and (K, ) is a treecreating pair,
but in both of them the avor of being of the NMPtype is the same: it generalizes
somehow the notion of t-omittory treecreating pairs. (Note that if (K, ) is a t-
omittory treecreating pair then it is of the NMPtype.) One could formulate
a uniform condition here, but that would result in unnecessary complications in
formulation.
Definition 3.2.3. (1) A nitary creating pair (K, ) is of the NMP
type if the following condition is satised:
()
NMP
Suppose that (w, t
0
, t
1
, . . .) Q
(K, ) is such that

(k )(nor[t
k
] >
H
(m
t
k
dn
))
and let n
0
< n
1
< n
2
< . . . < . Further, assume that
g :
_
i
pos(w, t
0
, . . . , t
ni1
)
_
i
pos(w, t
0
, . . . , t
ni+11
)
is such that g(v) pos(v, t
ni
, . . . , t
ni+1
) for v pos(w, t
0
, . . . , t
ni1
)
(so v < g(v)). Then there are 0 < i < and a creature s
(t
n0
, . . . , t
ni1
) such that
() nor[s] minnor[t
m
]
H
(m
tm
dn
) : n
0
m < n
i
,
() for each v pos(w, t
0
, . . . , t
n01
, s) there is j < i such that
v`m
tn
j+1
1
up
= g(v`m
tn
j
1
up
).
(2) A treecreating pair (K, ) is of the NMPtype if the following condition
is satised:
()
tree
NMP
Suppose that t
: T) Q
tree
(K, ) (see 1.3.5(4)) is such that

( T)(nor[t
] > 1)
and let F
0
, F
1
, F
2
, . . . be fronts of the quasi tree T such that for i < :
( F
i+1
)(
F
i
)(
< ).
Assume that a function g :

i
F
i

i
F
i+1
is such that < g()
F
i+1
provided F
i
. Then there are 0 < i < and a treecreature
s (t
: ( F
i
)( < )) such that
()
tree
nor[s] infnor[t
] 1 : T,
()
tree
for each pos(s) there are j < i and k < g() such that
`k F
j
and g(`k) _ .
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

3.2. PRESERVING NON-MEAGER SETS 57
Theorem 3.2.4. Assume (K, ) is a nitary, gluing and

2big creating pair.
Suppose that (K, ) is of the NMPtype and has the Halving Property. Let f :
be an H-fast function. Then the forcing notion Q
f
(K, ) is Cohen
preserving.
Proof. By 3.1.2 we know that the forcing notion Q
f
(K, ) is
-bounding.
Consequently it is enough to show that if A 2
is a non-meager set then

'
Q
f
(K,)
A is not meager
(see [BaJu95, 6.3.21]). So suppose that A 2
is not meager but some condition

in Q
f
(K, ) forces that this set is meager. Thus we nd a condition p
0
Q
f
(K, )
and Q
f
(K, )names

k
n
,
n
such that
p
0
'
Q
f
(K,)

k
0
<

k
1
< . . . < and
n
: [
k
n
,

k
n+1
) 2 (for n < ) and
(x A)(
n)(x`[
k
n
,

k
n+1
) ,=
n
).
As Q
f
(K, ) is
-bounding, we nd a condition p
1
p
0
, a sequence 0 = k
0
<
k
1
< k
2
< . . . < and names
n
such that
p
1
'
Q
f
(K,)

n
: [k
n
, k
n+1
) 2 (for n < ) and
(x A)(
n)(x`[k
n
, k
n+1
) ,=
n
).
Further, applying 2.2.11 we nd p
2
p
1
such that p
2
essentially decides all the
names
n
. Clearly we may assume that nor[t
p2
n
] > f(2, m
t
p
2
n
dn
) for all n < (and thus
nor[t
p2
n
] >
H
(m
t
p
2
n
dn
)). Choose 0 = n
0
< n
1
< . . . < and
0
<
1
<
2
< . . . <
such that
(m )(
H
(m
t
p
2
nm
dn
) <
m+1

m
)
and for each m < and every sequence w pos(w
p2
, t
p2
0
, . . . , t
p2
nm+11
) the condition
(w, t
p2
nm+1
, t
p2
nm+1+1
, . . .) decides all the names
j
for j [
m
,
m+1
) (remember the
choice of p
2
). Let
g :
_
m
pos(w
p2
, t
p2
0
, . . . , t
p2
nm1
)
_
m
pos(w
p2
, t
p2
0
, . . . , t
p2
nm+11
)
be such that v < g(v) pos(w
p2
, t
p2
0
, . . . , t
p2
nm+11
) for v pos(w
p2
, t
p2
0
, . . . , t
p2
nm1
).
Next, for each v pos(w
p2
, t
p2
0
, . . . , t
p2
nm1
), m < x (v) [
m
,
m+1
) and (v)
such that
there are no repetitions in (v): v pos(w
p2
, t
p2
0
, . . . , t
p2
nm1
)),
(v) : [k
(v)
, k
(v)+1
) 2 is such that
(g(v), t
p2
nm+1
, t
p2
nm+1+1
, . . .) '
Q
f
(K,)

(v)
= (v) .
Now we apply successively 3.2.3(1) to the condition (w
p2
, t
p2
0
, t
p2
1
, . . .), the sequence
n
i
: i < ) and the mapping g. As a result we construct an increasing sequence
0 = i
0
< i
1
< i
2
< . . . < of integers and creatures s
j
(t
p2
ni
j
, . . . , t
p2
ni
j+1
1
) such
that for all j < :
(1) nor[s
j
] minnor[t
p2
ni
j
]
H
(m
t
p
2
n
i
j
dn
), . . . , nor[t
p2
ni
j+1
1
]
H
(m
t
p
2
n
i
j+1
1
dn
),
(2) for each v pos(w
p2
, t
p2
0
, . . . , t
p2
ni
j
1
, s
j
) there is i
[i
j
, i
j+1
) such that
v`m
t
p
2
n
i
+1
dn
= g(v`m
t
p
2
n
i
dn
).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

It should be clear that (w
p2
, s
0
, s
1
, . . .) Q
f
(K, ) (note that if nor[t
p2
n
] > f(k +
1, m
t
p
2
n
dn
) for all n [n
ij
, n
ij+1
) then nor[s
j
] > f(k, m
sj
dn
)). Look at the set
A
def
= x 2
: for innitely many j , for every i
[i
j
, i
j+1
)
(v pos(w
p2
, t
p2
0
, . . . , t
p2
n
i
1
))(x`[k
(v)
, k
(v)+1
) = (v)).
It is a dense
0
2
subset of 2
and hence A
A ,= . Take x A
A. Note
that the choice of the s
j
s (see clause 2. above) implies that for each j < and
v pos(w
p2
, s
0
, . . . , s
j
) there is i
[i
j
, i
j+1
) such that letting v
= v`m
t
p
2
n
i
dn
we
have
(v, s
j+1
, s
j+2
, . . .) '
Q
f
(K,)

(v
)
= (v
) .
Hence (w
p2
, s
0
, s
1
, . . .) '
Q
f
(K,)
(
n )(x`[k
n
, k
n+1
) =
n
) , a contradiction.
Theorem 3.2.5. Assume that (K, ) is a nitary 2-big treecreating pair of

the NMPtype. Then the forcing notion Q
tree
1
Proof. Like 3.2.4 but using 3.1.1, 2.3.7(2), 2.3.12, and 3.2.3(2): we choose p
0
,
k
n
,
n
, p
1
, k
n
and
n
as there. Further we inductively build an increasing sequence
m
: m ) of integers, a condition p
2
p
1
and fronts F
m
of T
p2
(for m < )
such that [F
m
[ <
m+1
m
and for every F
m+1
_ T
p2
nor[t
p2
] m+ 1 and p
[]
2
decides all
j
for j [
m
,
m+1
),
_
and the front F
m+1
is above F
m
. For F
m
we dene g() F
m+1
, ()
[
m
,
m+1
) and () : [k
()
, k
()+1
) 2 in a manner parallel to that in the proof
of 3.2.4. Next we build a condition q p
2
and an increasing sequence m
i
: i )
such that each F
mi
T
q
is a front of T
q
and
if F
mi
T
q
, i then pos(t
q
)

F
m
: m
i
< m m
i+1
and for every pos(t

q
) there are j and k < g() such that

`k F
j
and g(`k) _ .
Finally we let
A
def
= x 2
: (
i )(m [m
i
, m
i+1
))( F
m
)(x`[k
()
, k
()+
) = ())
and we nish as in 3.2.4.
Theorem 3.2.6. Suppose that (K, ) is a nitary creating pair which captures
singletons. Then the forcing notion Q
w
Proof. Like 3.2.4, but using 3.1.3, 2.1.12 and 2.1.10. Note that the last implies
that the pair (K, ) has the following property:
() If (t
0
, . . . , t
N
) PFC(K, ),
H
(m
t0
dn
) k, 0 = n
0
< n
1
< . . . <
n
k
< N and u pos(u`m
t0
dn
, t
0
, . . . , t
N
) then there are s
0
, . . . , s
K
such that pos(u`m
s0
dn
, s
0
, . . . , s
) = u, m
s0
dn
= m
t0
dn
, m
s
up
= m
tN
up
and
t
0
, . . . , t
N
) s
0
, . . . , s
). Consequently, choosing an enumeration

w
j
: j <
H
(m
t0
dn
) of basis(t
0
) and letting u
j
= w
j
u`[m
t0
dn
, m
tn
j+1
1
up
)
we will have (remember (K, ) is forgetful)
() u
j
pos(w
j
, t
0
, . . . , t
nj+11
),
() (t
0
, . . . , t
N
) (s
0
, . . . , s
),
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

3.3. PRESERVING NON-NULL SETS 59
() for each w basis(s
0
) and v pos(w, s
0
, . . . , s
) there is j <
H
(m
t0
dn
)
such that v`m
tn
j+1
1
up
= u
j
.
Thus we may repeat the proof of 3.2.4 with not many changes.
Let us note that it is not an accident that we have the results on preserving of
non-meager sets only for forcing notions which are very much like the ones coming
from t-omittory treecreating pairs. If we look at the opposite pole: local weak
creating pairs, then we notice that they easily produce forcing notions making the
ground reals meager.
Definition 3.2.7. Let H be of countable character. We say that a weak
creating pair (K, ) for H is trivially meagering if for every t K with nor[t] > 1
and each u basis(t) and v pos(u, t) there is s (t) such that nor[s]
nor[t] 1 and v / pos(u, s).
Proposition 3.2.8. (1) If (K, ) is a local trivially meagering treecreating
pair for H, H is of countable character and e = 1, 3 then
'
Q
tree
e
(K,)

V is meager .
(2) If (K, ) is a simple nitary and trivially meagering creating pair for H
and f : is Hfast then
'
Q
f
(K,)

V is meager .
Proof. In the rst case remember that if p Q
tree
e
(K, ), e = 1, 3 then for
some T
p
we have ( T
p
)( _ nor[t
p
] 2). Hence, as (K, ) is local

and trivially meagering we easily get
'
Q
tree
e
(K,)
(x
m
H(m) V)(
m )(

W(m) ,= x(m)) .
For the second case suppose that p Q
f
(K, ). Since f(n+1, ) >
H
() +f(n, )
using the assumptions that (K, ) is simple and trivially meagering we immediately
see that
p '
Q
f
(K,)
(x
m
H(m) V)(
n )(

W`[m
t
p
n
dn
, m
t
p
n
up
) ,= x`[m
t
p
n
dn
, m
t
p
n
up
)) ,
nishing the proof.
Later, in 4.1.3, we will see that the forcing notions Q
s
(K, ) may make the
ground model reals meager too. This suggests that if one wants to build a forcing
notion preserving non-meagerness then the most natural approach is Q
tree
e
for e =
0, 2 or Q
w
.
3.3. Preserving non-null sets
In this section we introduce a property of treecreating pairs which implies that
forcing notions Q
tree
e
(K, ) preserve non-null sets. Though preserving non-null sets
alone is not enough to use the preservation theorem of [Sh:f, Ch. XVIII, 3], one
may apply the methods of [Sh 630], [Sh 669] when dealing with countable support
iterations of forcing notions of the type presented here.
Definition 3.3.1. We say that a weak creating pair (K, ) is of the NNPtype
if the following condition is satised:
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

()
NNP
there are increasing sequences a = a
n
: n < ) (0, 1) and
k = k
n
: n ) such that lim
n
a
n
< 1 and:
if t K, nor[t] > 2, v basis(t), g(v) k
n
, N < and a function
g : pos(v, t) T(N)
is such that (u pos(v, t))(a
n+1

|g(u)|
N
) then the set
n < N : there is s (t) such that nor[s] nor[t] 1 and
basis(t) basis(s) and (u pos(v, s))(n g(u))
has not less than N a
n
elements.
(The sequences a,
k from ()
NNP
are said to witness NNP.)
Definition 3.3.2. We say that a treecreating pair (K, ) for H is gluing
(respectively: weakly gluing) if for each well founded quasi tree T

n<
i<n
H(i)
and a system s
:

T) K such that
(

T)(s
TCR
[H] & pos(s
) = succ
T
())
there is s (s
:

T) such that
nor[s] supnor[s
] 1 :

T
(nor[s] infnor[s
] 1 :

T, respectively).
Remark 3.3.3. The above denition, though dierent from 2.1.7(2) (for creat-
ing pairs), has actually the same meaning: we may glue together creatures without
loosing too much on norms.
Definition 3.3.4. We say that a weak creating pair (K, ) is strongly nitary
if K is nitary (see 1.1.3(2)) and (o) is nite for each o K. If
(see 1.1.4(3))
is an equivalence relation on K and depends on
equivalence classes only, then

what we actually require is that (o)/
is nite.
Theorem 3.3.5. Suppose (K, ) is a strongly nitary treecreating pair of the
NNPtype. Further suppose that:
either (K, ) is t-omittory and e = 0
or (K, ) is 2-big and weakly gluing and e = 1.
If A 2
is a non-null set then '

Q
tree
e
(K,)
A is not null.
Proof. In both cases (i.e. e = 0 and e = 1) the proof is actually the same
so let us deal with the case e = 1 only (and thus we assume that (K, ) is 2-big
and weakly gluing). Suppose that A 2
is a set which is not null but for some

p
0
Q
tree
1
(K, )
p
0
'
Q
tree
1
(K,)
A is null.
We may assume that A is of outer measure 1 just consider the set
x 2
: (y A)(
n)(x(n) = y(n)).
Let a = a
n
: n < ),

k = k
n
: n < ) witness that (K, ) is of the NNPtype.
Let

T be a Q
tree
1
(K, )-name for a subtree of 2
<
such that
p
0
'
Q
tree
1
(K,)
([

T]) > lim
n
a
n
& [

T] A = .
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

3.3. PRESERVING NON-NULL SETS 61
By 2.3.7(2) we nd a condition p
1
p
0
and fronts F
n
of T
p1
such that for each
n < :
( F
n
)(the condition p
[]
1
decides

T 2
n
).
Clearly we may assume that
F
0
= root(p
1
) and (n)( F
n+1
)(
F
n
)(
< & k
n+1
g())
and (n )( F
n
)( T
p1
)( _ nor[t
p1
] > 4 +n) (remember 2.3.12).

As (K, ) is weakly gluing and nitary we nd a condition p
2
p
1
such that
T
p2
=
_
n
F
n
and ( T
p2
)(nor[t
p2
] > 3).
For F
n
, n < let g
: pos(t
p2
) T(2
n + 1
) be a function such that
( pos(t
p2
))(p
[]
2
'
Q
tree
1
(K,)

T 2
n + 1
= g
()).
Clearly, if F
n
, n < and pos(t
p2
) then a
n+1

|g()|
2
n+1
. Let (for F
n
,
n < )
X
n
def
= 2
n + 1
: there is s (t
p2
) such that
nor[s] nor[t
p2
] 1, and ( pos(s))( g
()).
Due to ()
NNP
we know that 2
n+1
a
n
[X
n
[ (for all F
n
, n ). Fix n < .
By downward induction on m < n we dene sets X
n
for F
m
:
if F
m
, m < n then
X
n
= 2
n+1
: there is s (t
p2
) such that
nor[s] nor[t
p2
] 1 and ( pos(t
p2
))( X
n
)).
Now we may apply the choice of a,
k (remembering that g() k

m
for each
F
m
) to conclude (by the downward induction on m n) that
(m n)( F
m
)([X
n
[ 2
n+1
a
m
).
Hence, in particular, 2
n+1
a
0
[X
n
root(p2)
[ for each n < . So look at the set
F
def
= x 2
: (
n)(x`(n + 1) X
n
root(p2)
).
Necessarily (F) a
0
> 0 and thus we may take x F A. For each n < such
that x`(n+1) X
n
root(p2)
we may choose a well founded quasi tree S
n
and a system
s
n
:

S
n
) of creatures from K such that:
S
n

_
mn
F
m
, max(S
n
) F
n+1
, root(S
n
) = root(p
2
) and for all

S
n
we have
pos(s
n
) = succ
Sn
(), s
n
(t
p2
), nor[s
n
] nor[t
p2
] 1 and x`(n + 1) X
n
,
and if S
n
F
n
,
pos(s
n
) then x`(n+1) g
). Note that if one constructs

a condition q
n
such that
S
n
T
qn
m<
F
m
,
(m n)( T
qn
F
m
)(

S
n
& t
qn
= s
n
) and
(m > n)( T
qn
F
m
)(t
qn
= t
p2
)
then q
n
'
Q
tree
1
(K,)
x`(n + 1)

T. Hence, in particular,
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

() if n is such that x`(n + 1) X
n
root(p2)
, m n + 1 and S
n
F
m
then p
[]
2
' x`m

T.
Applying the K onig Lemma (remember that (K, ) is strongly nitary!) we nd a
quasi tree S

n
F
n
and a system s
: S) such that
(1) max(S) = ,
(2) for all S:
succ
S
() = pos(s
) and nor[s
] nor[t
p2
] 1 and s
(t
p2
),
(3) for some increasing sequence 0 < n
0
< n
1
< n
2
< . . . < we have:
(i )(x`(n
i
+ 1) X
ni
root(p2)
) and
(i )( S F
i
)(j i)( S
nj
& s
= s
nj
).
The quasi tree S determines a condition q Q
tree
1
(K, ) stronger than p
2
. We
claim that q '
Q
tree
1
(K,)
(i < )(xì

T). Why? Let i and S F
i
. Then
s
= s
ni
, and S
ni
F
i
, i n
i
and x`(n
i
+ 1) X
ni
root(p2)
.
But now we may use () to conclude that for each such
p
[]
2
'
Q
tree
1
(K,)
xì

T,
and hence q ' xì

T. Consequently q ' x [

T], contradicting q p
0
.
We may get a variant of 3.3.5 for tree creating pairs which are not gluing (e.g.
for local (K, ), see 1.4.3). Then, however, we have to require more from the
witnesses for the NNPtype.
Theorem 3.3.6. Suppose that (K, ) is a strongly nitary 2big tree creating
pair of the NNPtype with witnesses a,
k such that k
n
= n. Then
'
Q
tree
1
(K,)
A is non-null
whenever A 2
is a set of positive outer measure.

Proof. It is similar to 3.3.5. We start exactly like there choosing A,

T, p
1
,
fronts F
n
of T
p1
and functions g
n
: F
n
T(2
n
) such that p
[]
1
'

T 2
n
= g
n
()
for each n and F
n
. Next we dene g
n
() for T
p1
below F
n
by downward
induction, in such a way that:
if T
p1
and there is
0
F
n
such that <
0
and g
n
() has been
dened already for all pos(t
p1
) and ( pos(t
p1
))(a
g()

|gn()|
2
n
) then
g
n
() = 2
n
: there is s (t
p1
) such that
nor[s] nor[t
p1
] 1, and ( pos(s))( g
n
()).
We continue as in 3.3.5 getting suitable S
n
for n and applying the K onig lemma
we get S T
p1
with the corresponding properties.
Note that the main dierence is that, in the above construction, we may keep the
demand
|gn()|
2
n
a
g()
(and this is the replacement for weakly gluing).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

3.5. EXAMPLES 63
3.4. (No) Sacks Property
Recall that a forcing notion P has the Sacks property is for every Pname x for
a real in
we have
'
P
(F
n
[]
n + 1
V)(n )( x(n) F(n)).
The Sacks property is equivalent to preserving the basis of the null ideal: every
Lebesgue null set in the extension may be covered by a null set (coded) in the
ground model. Here we are interested in refusing this property, i.e. getting forcing
notions which do not preserve the basis of the null ideal.
Definition 3.4.1. We say that a weakcreating pair (K, ) strongly violates
the Sacks property if
()
3.4.1
for some nondecreasing unbounded function f
we have
if t K, nor[t] > 1
then for each u basis(t) there is n g(u) such that
f(n) < [w(n) : w pos(u, t) & n < g(w)[.
Theorem 3.4.2. Let H be of countable character and let (K, ) be a weak
creating pair for H which strongly violates the Sacks property. Assume that
either (K, ) is a creating pair and P is one of Q
s
(K, ), Q
(K, ),
Q
w
(K, ), Q
f
(K, )
or (K, ) is a treecreating pair and then P is one of Q
tree
e
(K, ) (e < 5).
Then the forcing notion P fails the Sacks property.
Proof. For simplicity we may assume that H(i) = (for all i ). Take an
increasing sequence n
k
: k < ) of positive integers such that k + 1 < f(n
k
) for
all k . Let

W be the Pname for the generic real (see 1.1.13) and let x be the
P-name for an element of
such that
'
P
(k )( x(k) = (

W`n
k
)).
where :
<
is the canonical bijection. Now we claim that
'
P
(F
n
[]
n + 1
V)(
n)( x(n) / F(n)).

Why? Suppose that p P, F

n
[]
n
, N . By 3.4.1, in all relevant cases,
we nd k > N and n [n
k
, n
k+1
) such that p allows more than f(n) values for
W(n). But k + 2 f(n

k
) f(n) and thus the condition p allows more than
k + 1 values for x(k + 1).
3.5. Examples
Example 3.5.1. We build a treecreating pair (K
3.5.1
,
3.5.1
) which is: strongly
nitary, 2-big, t-omittory, gluing, of the NNPtype, and which strongly violates
the Sacks property.
Construction. Before we dene (K
3.5.1
,
3.5.1
) let us note some basic prop-
erties of the nice pre-norm dp
0
k
dened in 2.4.2(1),(4).
Claim 3.5.1.1. Let M < , k < .
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

(1) If A M then either dp
0
k
(A) dp
0
k
(M) log
2+k
(2)
or dp
0
k
(M A) dp
0
k
(M) log
2+k
(2).
(2) Suppose that a (0, 1), N < and a function g : M T(N) is such
that
(m < M)(N
a(k + 1) + 1
k + 2
[g(m)[).
Then a N [n < N : dp
0
k
(m < M : n g(m)) dp
0
k
(M) 1[.
Proof of the claim: 2) Let u(n) = m < M : n g(m) and
X = n < N : dp
0
k
(u(n)) dp
0
k
(M) 1.
Look at the set Y
def
= (m, n) M N : n g(m) and note that
[Y [ =
m<M
[g(m)[ N
a(k + 1) + 1
k + 2
M.
On the other hand, noticing that n X [u(n)[
M
k+2
, we have:
[Y [ =

n<N
[u(n)[
nX
[u(n)[ +

n/ X
M
k + 2
[X[ M + (N [X[)
M
k + 2
.
Consequently
a(k+1)+1
k+2
N [X[ +
N|X|
k+2
and hence a N [X[, proving the claim.
Let H(n) = (n + 2)
n
.
Let K
3.5.1
be the collection of all treecreatures t for H such that
(1) dis[t] is a pair (d
0
(t), d
1
(t)) such that d
0
(t)
m<n
H(m) is a nite
tree, [d
0
(t)[ > 2 and d
1
(t) _ root(d
0
(t)),
(2) nor[t] = mindp
0
k
(succ
d0(t)
()) : split(d
0
(t)) & g() = k,
(3) val[t] = d
1
(t), ) : max(d
0
(t)).
For a well founded quasi tree T and a system s
:

T) of treecreatures from
K
3.5.1
such that the requirement (a) of 1.3.3 is satised we let
3.5.1
(s
:

T) = s K
3.5.1
: d
1
(s) = root(T) & rng(val[s]) max(T).
It should be clear that
3.5.1
is a tree-composition on K
3.5.1
. Now, the tree
creating pair (K
3.5.1
,
3.5.1
) is strongly nitary, t-omittory and gluing. For the last
two properties we apply the procedure similar to the one below.
Note that if t K
3.5.1
and d
0
(t) is a splitting point of d
0
(t) then choosing
max(d
0
(t)) (for succ
d0(t)
()) such that _
we may build a tree

creature s
3.5.1
(t) such that pos(s) =
: succ
d0(t)
(). Then we will
have nor[s] nor[t]. If additionally d
0
(t) is a splitting point such that
nor[t] = dp
0
g()
(succ
d0(t)
()) then nor[s] = nor[t] (see the denition of K
3.5.1
). In
this case, let us call the respective treecreature s(t) (here we just x one such).
Considering suitable s(t)s and using 3.5.1.1(1) one can easily show that the
creating pair (K
3.5.1
,
3.5.1
) is 2-big.
Let k
n
= 2
n+4
2, a
0
=
1
2
, a
n+1
=
an(kn+1)+1
kn+2
= a
n

an
2
n+4
+
1
2
n+4
. We are
going to show that the sequences

k = k
n
: n < ), a = a
n
: n < ) witness that
(K
3.5.1
,
3.5.1
) is of the NNPtype.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

3.5. EXAMPLES 65
First note that a,
k are strictly increasing, a (0, 1) and lim

n
a
n

5
8
. Now
suppose that t K
3.5.1
, k
n
m
t
dn
, N < and g : pos(t) T(N) is such that
( pos(t))(a
n+1

[g()[
N
).
Take s(t) (dened above) and look at h = g`pos(s(t)). We may think that actually
h : succ
d0(s(t))
() T(N), where is the unique splitting point of s(t) (note that
g() m
s(t)
dn
k
n
). Applying claim 3.5.1.1(2) we get
a
n
N [n < N : dp
0
kn
(m : n h(m)) dp
0
kn
(succ
d0(s(t))
()) 1[.
For each n < N from the set on the right hand side of the inequality above choose
s
n

3.5.1
(s(t)) such that pos(s
n
) = pos(s(t)) : n h((g())). By the
denition of dp
0
k
, dp
0
k
(w
0
) dp
0
k
(w
1
) 1 implies that dp
0
k+1
(w
0
) dp
0
k+1
(w
1
)
1, and therefore dp
0
kn
(m : n h(m)) dp
0
kn
(succ
d0(s(t))
()) 1 implies that
nor[s
n
] nor[s(t)] 1. Now we may conclude that the set
n < N : there is s
3.5.1
(t) such that nor[s] nor[t] 1 and
( pos(s))(n g()).
has not less than a
n
N elements.
Finally note that (K
3.5.1
,
3.5.1
) satises the condition (
3.4.1
) for f(n) = n
(so it strongly violates the Sacks property).
Conclusion 3.5.2. The forcing notions Q
tree
e
(K
3.5.1
,
3.5.1
) for e<5 are equiv-
alent. They are proper, preserve the outer measure, preserve non-meager sets, are
-bounding, but do not have the Sacks property.

Example 3.5.3. We construct a nitary, 2big and local treecreating pair
(K
3.5.3
,
3.5.3
) which is trivially meagering and of the NNPtype with the sequence
k = n : n < ) witnessing it.

Construction. This is similar to 3.5.1. We dene

k
,

k and a letting k
n
=
2
n+4
2, k
n
= n, a
0
=
1
2
, a
n+1
=
an(k
n
+1)+1
k
n
+2
. Let H(n) = 2
(n+4)
2
.
The family K
3.5.3
consists of these t TCR[H] that:
dis[t] = , A
t
), where is such that t TCR
[H] and A
t
H(g()),
val[t] = , ) : < & g() = g() + 1 & (g()) A
t
,
nor[t] = dp
0
k
g()
(A
t
).
The operation
3.5.3
is trivial and s
3.5.3
(t) if and only if val[s] val[t].
Plainly, (K
3.5.3
,
3.5.3
) is a 2big local and trivially meagering treecreating
pair. Checking that it is of the NNPtype (with witnesses a and

k) is exactly like
in 3.5.1 (just apply 3.5.1.1).
Conclusion 3.5.4. Q
tree
1
(K
3.5.3
,
3.5.3
) is a proper
bounding forcing no-

tion which preserves outer measure but makes the ground model reals meager.
Example 3.5.5. We dene functions H, f and a creating pair (K
3.5.5
,
3.5.5
)
such that
(1) H is nitary, f : is Hfast,
(2) (K
3.5.5
,
3.5.5
) is gluing, forgetful,

2big, has the Halving Property and is
of the NMPtype,
(3) the forcing notion Q
f
(K
3.5.5
,
3.5.5
) is not trivial.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Construction. Dene inductively H and f such that for all n, k, :
(i) H(0) = 8, H(n) = 2
H(n)+f(n,n)
,
(ii) f(0, ) = + 1, f(k + 1, ) = 2
H()+1
(f(k, ) +
H
() + 2)
(compare with 2.4.6; note that the above conditions uniquely determine H and f).
By their denition H is nitary and f is Hfast.
A creature t CR
m0,m1
[H] belongs to K
3.5.5
if:
dis[t] = m
0
, m
1
, X
t
, H
t
, A
k
t
, H
k
t
, u
k
t
: k X
t
)), where X
t
[m
0
, m
1
),
H
t
is a nice pre-norm on T(X
t
), and for each k X
t
:
A
k
t
H(k), H
k
t
is a nice pre-norm on T(A
k
t
) and u
k
t

i[m0,m1)\{k}
H(i),
val[t] = u, v)

i<m0
H(i)

i<m1
H(i) : u < v & (k X
t
)(u
k
t

v & v(k) A
k
t
),
nor[t] = minH
t
(X
t
), H
k
t
(A
k
t
) : k X
t
.
Now we describe the operation
3.5.5
. Suppose that t
0
, . . . , t
n
K
3.5.5
are such
that m
ti
up
= m
ti+1
dn
for i < n. Let
3.5.5
(t
0
, . . . , t
n
) consist of all creatures s K
3.5.5
such that m
s
dn
= m
t0
dn
, m
s
up
= m
tn
up
and for some i n
() X
s
X
ti
, (B X
s
)(H
s
(B) H
ti
(B)) and for every k X
s
:
() A
k
s
A
k
ti
and (A A
k
s
)(H
k
s
(A) H
k
ti
(A)), and
() u
k
ti
u
k
s
and for every j (n + 1) i for some X
tj
we have
u
tj
u
k
s
and u
k
s
() A
tj
.
3.5.5
,
3.5.5
) is a gluing and forgetful creating pair for H.
It is

2big as for each t K
3.5.5
both H
t
and H
k
t
(for k X
t
) are nice pre-norms.
We may use similar arguments as in 3.2.6 to show that (K
3.5.5
,
3.5.5
) is of the
NMPtype. Now dene function half : K
3.5.5
K
3.5.5
by
half(t) = s if and only if
m
s
dn
= m
t
dn
, m
s
up
= m
t
up
, X
s
= X
t
, A
k
s
= A
k
t
for k X
s
and
H
s
= (H
t
)
1
2
nor[t]
, H
k
s
= (H
k
t
)
1
2
nor[t]
for k X
s
(see 2.4.3(2)).
It is not dicult to check that the function half witnesses that (K
3.5.5
,
3.5.5
) has
the Halving Property.
Note that (K
3.5.5
,
3.5.5
) resembles an omittory creating pair.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

CHAPTER 4
Omittory with Halving
In 2.2.11, 3.1.2 we saw how the Halving Property and the bigness apply to
forcing notions Q
f
(K, ). In this chapter we will look at another combination:
omittory creating pairs with the weak Halving Property. Since an omittory creating
pair cannot be big, it is natural that we consider in this context the (s) norm
condition. The rst example of a forcing notion of the type Q
s
(K, ) for an
omittory creating pair (K, ) with the weak Halving Property appeared in [Sh 207]
(but in the real application there a dierent norm condition was used). A direct
application of a forcing notion of this type was presented in [RoSh 501]. In the
last part of this chapter we will develop the example from that paper. Before, in
the rst section, we show that the forcing notions Q
s
(K, ) with (K, ) omittory
tend to add Cohen reals and make ground reals meager. Next we introduce some
general operations on creating pairs and, in the third section, we explain how the
weak Halving Property may prevent them from adding dominating reals.
4.1. What omittory may easily do
Natural examples of omittory creating pairs with the weak Halving Property
are meagering and anti-big (see 4.1.2 below). We will show how these properties
cause that forcing notions Q
s
(K, ) do some harm to the old reals. Examples
and applications are presented in the last part of this chapter.
First note the following easy observation.
Proposition 4.1.1. If (K, ) is an omittory creating pair such that for each
t K, u basis(t)
nor[t] > 0 [pos(u, t)[ > 2
then '
Q
s
(K,)
there is an unbounded real over V.
(1) We say that (K, ) is meagering if for every (t
0
, . . . , t
n1
) PFC(K, ),
t (t
0
, . . . , t
n1
) and k
i
: i < n) such that for each i < n:
nor[t
i
] > 2 and m
ti
dn
k
i
< m
ti
up
and nor[t] > 2
there is s (t) satisfying
nor[s] nor[t] 1 and
(u basis(t
0
))(v pos(u, s))(k [g(u), g(v)))(v(k) ,= 0) and
(u basis(t
0
))(v pos(u, s))(i < n)(v(k
i
) = 0).
(2) The creating pair (K, ) is called anti-big if there are colourings
c
t
:
_
ubasis(t)
pos(u, t) 3 for t K
67
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

68 4. OMITTORY WITH HALVING
such that: if (t
0
, . . . , t
n1
) PFC(K, ), nor[t
i
] > 1 (for i < n) and
t (t
0
, . . . , t
n1
), nor[t] > 1 then for each u basis(t
0
) there are
v
0
, v
1
pos(u, t) and < n satisfying
v
0
`m
t
dn
= v
1
`m
t
dn
, c
t
(v
0
`m
t
up
) = 0, c
t
(v
1
`m
t
up
) = 1, and
(i n )(c
ti
(v
0
`m
ti
up
) = c
ti
(v
1
`m
ti
up
) = 2).
Theorem 4.1.3. Let (K, ) be a growing creating pair.
(1) If (K, ) is meagering then
'
Q
s
(K,)

V is meager.
(2) If (K, ) is anti-big then
'
Q
s
(K,)
there is a Cohen real over V.
Proof. 1) Let p = (w
p
, t
p
0
, t
p
1
, . . .) Q
s
(K, ) and for n let f(n) =
T([m
t
p
n
dn
, m
t
p
n
up
)). The space

n
f(n) equipped with the product topology (of the
discrete f(n)s) is a perfect Polish space. Thus it is enough to show that
p '
Q
s
(K,)

n
f(n) V is a meager subset of
n
f(n) .
Note that if X
n
f(n) is such that (
n )(X(n) ,= ) then the set

Y
n
f(n) : (
n )(Y (n) = or Y (n) ,= X(n))

is meager in

n
f(n). Let

X be a Q
s
(K, )name for an element of

n
f(n)
such that
p '
Q
s
(K,)
(n )(

X(n) = k [m
t
p
n
dn
, m
t
p
n
up
) :

W(k) ,= 0),
where

W is the Q
s
(K, )name for the generic real (see 1.1.13). It follows from
the remarks above that it is enough to show that
() p '
Q
s
(K,)
(
n )(

X(n) ,= ) and
() p '
Q
s
(K,)
(Y

n
f(n) V)(
n )(Y (n) = or Y (n) ,=

X(n)).
To this end suppose that p q = (w
q
, t
q
0
, t
q
1
, . . .) Q
s
(K, ) is such that g(w
q
) >
1 and nor[t
q
i
] > m
t
q
i
dn
+ 2 for i and let Y

n
f(n). For each n choose
k
n
[m
t
p
n
dn
, m
t
p
n
up
) such that Y (n) ,= k
n
Y (n). Let 0 n
0
< n
1
< n
2
< . . . <
be such that w
q
pos(w
p
, t
p
0
, . . . , t
p
n01
) and t
q
i
(t
p
ni
, . . . , t
p
ni+11
) for i .
Note that necessarily m
t
p
n
0
dn
2 and thus nor[t
p
n
] > 2 for each n n
0
. Applying
4.1.2(1) we nd s
i
(t
q
i
) such that for each i :
()
1
i
nor[s
i
] nor[t
q
i
] 1 > m
t
q
i
dn
, and
()
2
i
(u pos(w
p
, t
p
0
, . . . , t
p
ni1
))(v pos(u, s
i
))(k [g(u), g(v)))(v(k) ,=
0),
()
3
i
(u pos(w
p
, t
p
0
, . . . , t
p
ni1
))(v pos(u, s
i
))(n [n
i
, n
i+1
))(v(k
n
) = 0).
Look at r
def
= (w
q
, s
0
, s
1
, s
2
, . . .). Clearly r Q
s
(K, ) is a condition stronger
than q. Moreover, by the choice of the s
i
s we have
r '
Q
s
(K,)
(n n
0
)(

W(k
n
) = 0)
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

4.2. MORE OPERATIONS ON WEAK CREATURES 69
and therefore, by the choice of the k
n
s, we get
r '
Q
s
(K,)
(n n
0
)(Y (n) ,= Y (n) ,=

X(n)).
Further note that if v pos(w
p
, s
0
) is given by ()
2
0
above for w
q
then
(v, s
1
, s
2
, . . .) ' (n [n
0
, n
1
))(

X(n) ,= ).
We nish by density arguments.
2) Let p = (w
p
, t
p
0
, t
p
1
, . . .) Q
s
(K, ) and let

Z be a Q
s
(K, )name for a
subset of such that
p '
Q
s
(K,)

Z = n : c
t
p
n
(

W`m
t
p
n
up
) < 2.
Note that p '

Z is innite. Why? Suppose p q Q
s
(K, ), g(w
q
) > 1.
Let n
0
< n
1
< be such that w
q
pos(w
p
, t
p
0
, . . . , t
p
n01
), t
q
0
(t
p
n0
, . . . , t
p
n11
).
Necessarily nor[t
p
i
] > 1 for i [n
0
, n
1
) and nor[t
q
0
] > 1. So we nd [n
0
, n
1
) and
v
0
, v
1
pos(w
q
, t
q
0
) as in 4.1.2(2). Now look at the condition r = (v
0
, t
q
1
, t
q
2
, . . .)
Q
s
(K, ). It is stronger than q and forces that

Z.
Now let c be a Q
s
(K, )name for a real in 2
such that
p '
Q
s
(K,)
if k and n is the k
th
member of

Z then c(k) = c
t
p
n
(

W`m
t
p
n
up
) .
We claim that p ' c is a Cohen real over V. So suppose that p q Q
s
(K, ),
g(w
q
) > 1 and | 2
is an open dense set. Let 0 n

0
< n
1
< . . . < be such
that
w
q
pos(w
p
, t
p
0
, . . . , t
p
n01
) and t
q
i
(t
p
ni
, . . . , t
p
ni+11
) for i .
Let m = [n < n
0
: c
t
p
n
(w
q
`m
t
p
n
up
) < 2[ and let 2
m
be such that (k) =
c
t
p
n
(w
q
`m
t
p
n
up
) if k < m and n < n
0
is the k
th
member of the set n < n
0
:
c
t
p
n
(w
q
`m
t
p
n
up
) < 2. Choose 2
<
such that < and
(x 2
)( < x x |).
Let j = g() m. Use 4.1.2(2) to dene inductively u pos(w
q
, t
q
0
, . . . , t
q
j1
) such
that for each i < j, for some [n
i
, n
i+1
) we have
c
t
p
(u`m
t
p
up
) = () and (k [n
i
, n
i+1
) )(c
t
p
k
(u`m
t
p
k
up
) = 2).
Look at the condition r = (u, t
q
j
, t
q
j+1
, . . .) Q
s
(K, ): it is stronger than q and
it forces that < c. We nish by density argument.
4.2. More operations on weak creatures
Below we dene some operations on creatures and treecreatures which provide
for (some) systems of weak creatures a new weak creature (of the same type). These
operations may be used to dene subcomposition operations.
Definition 4.2.1. Suppose 0 < m < and for i < m we have t
i
CR[H]
such that m
ti
up
m
ti+1
dn
. Then we dene the sum of the creatures t
i
as a creature
t =
sum
(t
i
: i < m) such that (if well dened then):
(a) m
t
dn
= m
t0
dn
, m
t
up
= m
tm1
up
,
(b) val[t] is the set of all pairs h
1
, h
2
) such that:
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

g(h
1
) = m
t
dn
, g(h
2
) = m
t
up
, h
1
< h
2
,
and h
2
`m
ti
dn
, h
2
`m
ti
up
) val[t
i
] for i < m,
and h
2
`[m
ti
up
, m
ti+1
dn
) is identically zero for i < m1,
(c) nor[t] = minnor[t
i
] : i < m,
(d) dis[t] = 0)
dis[t
i
] : i < m).
If for all i < m1 we have m
ti
up
= m
ti+1
dn
then we call the sum tight.
Remark 4.2.2. Note that the sum
sum
(t
i
: i < m) is dened only for these
sequences t
i
: i < m) CR[H] for which m
ti
up
m
ti+1
dn
and part (b) of the denition
gives a nonempty value of val[t].
Definition 4.2.3. If m < , u m, d H() is a function such that dom(d)
(R
0
)
u
and rng(d) R
0
, and for i < m creatures t
i
CR[H] are such that
m
ti
up
m
ti+1
dn
then we dene the (d, u)-sum t =
sum
d,u
(t
i
: i < m) of the t
i
s by:
(a) m
t
dn
= m
t0
dn
, m
t
up
= m
tm1
up
,
(b) val[t] is the set of pairs h
1
, h
2
) such that:
g(h
1
) = m
t
dn
, g(h
2
) = m
t
up
, h
1
< h
2
and
h
2
`m
ti
dn
, h
2
`m
ti
up
) val[t
i
] for i u,
h
2
`[m
ti
dn
, m
ti
up
) is identically zero for i / u and
h
2
`[m
ti
up
, m
ti+1
dn
) is identically zero for i < m1.
(c) nor[t] = d(nor[t
i
] : i u)),
(d) dis[t] = 1, d, u)
dis[t
i
] : i < m).
[Note: the (d, u)-sum is dened only if clause (b) gives a nonempty value for val[t].]
Definition 4.2.4. (1) For a pre-norm H on (see 2.4.1) let D
H
be the
family of all functions d such that for some nite set u
d
, H(u
d
) > 0
and
d : (R
0
)
u
d
R
0
: r
i
: i u
d
) minH(u
d
), r
i
: i u
d
.
(2) We say that a creating pair (K, ) is saturated with respect to a pre-norm
H on if for each d D
H
and (t
i
: m
0
i < m
1
) PFC(K, ) such
that u
d
[m
0
, m
1
) and nor[t
i
] > 0 for i u
d
:
sum
d,u
d
(t
i
: m
0
i < m
1
) is well dened and belongs to (t
i
: m
0
i < m
1
),
and if t (
sum
d,u
d
(t
i
: m
0
i < m
1
)), nor[t] > 0 then for some d
D
H
and s
i
(t
i
) (for m
0
i < m
1
) we have
u
d
u
d
, val[
sum
d
,u
d
(s
i
: m
0
i < m
1
)] val[t], and nor[s
i
] > 0 for i u
d
.
We say that (K, ) is saturated with respect to (nice) prenorms if for each
(nice) pre-norm H on , (K, ) is saturated with respect to H. Similarly
for other classes of prenorms.
Remark 4.2.5. Note that in practical realizations of 4.2.4(2) the additional
parameter dis may play a crucial role. Looking at a creature t we may immediately
recognize if it comes from the operation
sum
d,u
d
and we do not have to worry that the
last demand gives a contradiction. It may happen that for distinct ds from D
H
we
get (as a result of
sum
d,u
d
) creatures with the same values of val, nor, however they
are distinguished by dis. Moreover, the same eect appears for distinct prenorms
H: we can read from dis the function d and consequently the function H restricted
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

4.3. OLD REALS ARE UNBOUNDED 71
to subsets of u
d
. In applications we may redene dis[
sum
d,u
d
(t
i
: i < m)], but we
should keep this coding property.
Definition 4.2.6. Let T

n<
i<n
H(i) be a well founded quasi tree and let
s
:

T) TCR[H] be a system of tree creatures such that for each

T:
dom(val[s
]) = and pos(s
) = succ
T
().
(1) The treesum t =
tsum
(s
:

T) of tree creatures s
(for

T) is
dened by:
() val[t] = root(T), ) : max(T),
() nor[t] = infnor[s
] :

T & nor[s
] 1, if nor[s
] < 1 for all

T then we let nor[t] = 0,
() dis[t] = 2)
:

T).
(2) For a function g
, the special treesum t =

tsum
g
(s
:

T) of tree
creatures s
(for

T) with respect to g is dened in a similar manner
as
tsum
but the conditions (), () introducing the norm and dis are
replaced by
()
g
nor[t] = maxk < : ( max(T))([ < g() : `

T and
nor[s
] k[ g(k)),
()
g
dis[t] = 3, g)
dis[s
] :

T).
4.3. Old reals are unbounded
Recall that a forcing notion P is almost
bounding if for every Pname

f for
an element of
and any p P we have

(g
)(A []
)(q p)(q '

P
(
n A)(

f(n) < g(n))).
Almost
bounding forcing notions do not add dominating reals (i.e. they force
that (x
)(y
V)(
n)(x(n) < y(n))). If Q is a forcing notion not

adding dominating reals and
'
Q

P is almost
bounding
then the composition Q

P does not add a dominating real (see [Sh:f, Ch VI, 3.6]).
Thus the notion of being almost
bounding is very useful from the point of

view of iterations: in a countable support iteration of proper forcing notions no
dominating reals are added at limit stages (see [Sh:f, Ch VI, 3.17]). (Note that
not adding dominating reals is not preserved by compositions.)
In the denition 4.3.1(1) below one can think about the following situation
(explaining the name decision function). Suppose that (K, ) is a creating pair,
is a Q
s
(K, )name for an ordinal and p Q
s
(K, ), N
0
< are such that p
approximates at each n N
0
(see 1.2.9, remember 2.1.4). Let us dene a function
z : pos(w
p
, t
p
0
, . . . , t
p
N01
) PC(K, )
_
kN0
pos(w
p
, t
p
0
, . . . , t
p
k1
)
such that for every v pos(w
p
, t
p
0
, . . . , t
p
N01
) and t
0
, t
1
, . . .) PC(K, ) satisfying
t
p
N0
, t
p
N0+1
, . . .) t
0
, t
1
, . . .):
(1) If (v, t
0
, t
1
, . . .) Q
s
(K, ) then z(v, t
0
, t
1
, . . .)) is the rst (in a xed or-
dering of

n<
m<n
H(m)) of the shortest v
such that v
pos(v, t
0
, . . . , t
k1
)
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

(for some k < ) and (v
, t
p
m
, t
p
m+1
, . . .) decides the value of (m is just
suitable: m
t
p
m
dn
= m
t
k1
up
).
(2) If (v, t
0
, t
1
, . . .) / Q
s
(K, ) then we take the rst k such that nor[t
k
]
m
t
k
dn
and we ask if there is t
0
, t
1
, . . .) PC(K, ) such that
(v, t
p
N0
, t
p
N0+1
, . . .) (v, t
0
, t
1
, . . .) Q
s
(K, ), and for some k
z(v, t
0
, t
1
, . . .)) pos(v, t
0
, . . . , t
1
) and t
0
, . . . , t
1
) = t
0
, . . . , t
1
).
If the answer is yes then we choose such a sequence t
0
, t
1
, . . . ) and we
let
z(v, t
0
, t
1
, . . . )) = z(v, t
0
, t
1
, . . . ))
(note that this does not depend on the choice of the particular t
0
, t
1
, . . . );
see the previous case). If the answer is no then z(v, t
0
, t
1
, . . .)) is the
rst element of pos(v, t
0
, . . . , t
k
).
This z is a canonical example of a decision function for p, N
0
, (K, ); we will call
it z(p, N
0
, ) (assuming that (K, ) is understood).
(1) Let p Q
(K, ), N
0
. We say that a function
z : pos(w
p
, t
p
0
, . . . , t
p
N01
) PC(K, )
_
kN0
pos(w
p
, t
p
0
, . . . , t
p
k1
)
is a decision function for p, N
0
, (K, ) if:
()
4.3.1
for every v pos(w
p
, t
p
0
, . . . , t
p
N01
) and t
0
, t
1
, . . .) PC(K, ) such
that t
p
N0
, t
p
N0+1
, . . .) t
0
, t
1
, . . .), there is k such that:
z(v, t
0
, t
1
, . . .)) pos(v, t
0
, . . . , t
k1
)
and if t
p
N0
, t
p
N0+1
, . . .) t
0
, t
1
, . . .) PC(K, ) is such that t
i
for all i < k, then
z(v, t
0
, t
1
, . . .)) = z(v, t
0
, t
1
, . . .)).
(2) We say that (K, ) is of the AB-type whenever the following two condi-
tions are satised:
()
0
AB
if (t
0
, . . . , t
n1
) PFC(K, ), k < n then there is t (t
0
, . . . , t
n1
)
such that
nor[t] minnor[t
] : < n
and if (w, t
0
, . . . , t
n1
) FC(K, ), t
(t), nor[t
] > 0, then there

is t
(t
k
) such that nor[t
] > 0 and
(u
pos(w, t
0
, . . . , t
k1
))(u
pos(u, t
))(v pos(w, t
))(u
_ v);
()
1
AB
if p Q
(K, ), N
0
, nor[t
p
i
] > 2 for i N
0
and
z : pos(w
p
, t
p
0
, . . . , t
p
N01
) PC(K, )
_
kN0
pos(w
p
, t
p
0
, . . . , t
p
k1
)
0
, (K, )
then there are N
1
> N
0
and t
(t
p
N0
, . . . , t
p
N11
) such that
nor[t
]
1
2
minnor[t
p
N0
], . . . , nor[t
p
N11
] and
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

for each v pos(w
p
, t
p
0
, . . . , t
p
N01
) and t (t
) with nor[t] > 0

there is t
0
, t
1
, . . .) PC(K, ) such that t
p
N0
, t
p
N0+1
, . . .) t
0
, t
1
, . . .)
and
() if t
i
(t
p
N0+k
, . . . , t
p
N0+k+
) (i, k, < )
then nor[t
i
]
1
2
minnor[t
p
N0+k
], . . . , nor[t
p
N0+k+
], and
() (w pos(v, t))(z(v, t
0
, t
1
, t
2
, . . .)) _ w).
(3) We say that (K, ) is condensed if for every (w, t
0
, . . . , t
n1
) FC(K, )
with nor[t
i
] > 0 for i < n, and t (t
0
, . . . , t
n1
), nor[t] > 0, there exist
k < n, a creature s (t
k
) and v pos(w, t
0
, . . . , t
k1
) such that
nor[s] > 0 and (u pos(v, s))(u
pos(w, t))(u _ u
).
Remark 4.3.2. Note that the condition ()
0
AB
is easy to satisfy: e.g. if (K, )
is omittory and has the property that for every t K:
if m
0
m
t
dn
< m
t
up
m
1
then (t [m
0
, m
1
)) = s [m
0
, m
1
) : s (t)
then it satises this requirement (the ()
0
AB
for (t
0
, . . . , t
n1
), k < n is witnessed
by t
k
[m
t0
dn
, m
tn1
up
) ).
Theorem 4.3.3. Suppose that (K, ) is a nitary, growing and condensed cre-
ating pair of the ABtype. Then the forcing notion Q
s
(K, ) is almost
-
bounding.
Proof. Let us start with the following claim which will be used later too.
Claim 4.3.3.1. Let (K, ) be as in the assumptions of 4.3.3. Suppose that is
a Q
s
(K, )name for a function in
, q Q
s
(K, ) and n . Then there
are a condition p = (w
p
, t
p
0
, t
p
1
, . . .) Q
s
(K, ) and a strictly increasing function
g
such that q
s
n
p and for every
(
) for each v pos(w

p
, t
p
0
, . . . , t
p
n1
, . . . , t
p
n+1
) and t (t
p
n+
) with nor[t] >
0 there is w pos(v, t) such that the condition (w, t
p
n++1
, t
p
n++2
, . . .) de-
cides the value of (g()) and the decision is smaller than g( + 1).
Proof of the claim: We dene inductively conditions p
s
(K, ) and the values
g() for such that g(0) = 0, q = p
0

s
n
p
1

s
n+1
. . .
s
n+
p
n+

s
n++1
. . .
and p
+1
, g(), g( + 1) have the property stated in (
). It should be clear that

then the limit condition p = lim
(see 1.2.13(3)) is as required in the claim.

Suppose we have dened p
, g(). Using 2.1.3 and 2.1.4 we nd a condition

p
s
(K, ) such that
p

n+
p
, nor[t
p
k
] > 2 m
t
p
k
+ 2 for all k n + and
p
approximates (g()) at each t

p
k
for k n +.
Let z
= z(p
, n + , (g())) be the canonical decision function as dened before

4.3.1 (remember the choice of p
). Thus
z
: pos(w
q
, t
p
0
, . . . , t
p
n+1
) PC(K, )
_
mn+
pos(w
q
, t
p
0
, . . . , t
p
m1
)
is a decision function such that
if w pos(w
q
, t
p
0
, . . . , t
p
n+1
), (w, t
0
, t
1
, . . .) Q
s
(K, ),
(w, t
p
n+
, t
p
n++1
, . . .) (w, t
0
, t
1
, . . .)
then z
(w, t
0
, t
1
, . . .)) pos(w, t
0
, t
1
, . . . , t
m1
) (for some m )
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

is such that the condition (z
(w, t
0
, t
1
, . . .)), t
p
M
, t
p
M+1
, . . .) gives a
value to (g()), where M is such that m
t
m1
up
= m
t
p
M1
up
.
Now apply ()
1
AB
to p
, n+ and z
to nd N > n+ and t
(t
p
n+
, . . . , t
p
N1
)
such that
nor[t
]
1
2
minnor[t
p
n+
], . . . , nor[t
p
N1
] > m
t
p
n+
dn
+ 1
and for each v pos(w
q
, t
p
0
, . . . , t
p
n+1
) and t (t
) with nor[t] > 0 there is

w pos(v, t) for which (w, t
p
N
, t
p
N+1
, . . .) decides (g()). For this note that if
t
0
, t
1
, . . .) PC(K, ) is given by ()
1
AB
for z
, t, v then, as for some k

i

i
<
t
i
(t
p
n++ki
, . . . , t
p
n++i
), and
nor[t
i
]
1
2
minnor[t
p
n++ki
], . . . , nor[t
p
n++i
] > m
t
i
dn
,
the condition (v, t
0
, t
1
, . . .) Q
s
(K, ) is stronger than (v, t
p
n+
, t
p
n++1
, . . .). Thus
our requirements on z
k
apply. Finally we dene
p
+1
= (w
q
, t
p
0
, . . . , t
p
n+1
, t
, t
p
N
, t
p
N+1
, . . .) and
g( + 1) = 1 +g()+
+maxi < : (vpos(w
q
, t
p
0
, . . . t
p
n+1
, t
))((v, t
p
N
, t
p
N+1
, . . .) ' (g()) = i).
Clearly they are as required. This nishes the inductive construction and the proof
of the claim.
Now we are going to show that Q
s
(K, ) is almost
bounding. For this

suppose that is a name for a strictly increasing function in
and q Q
s
(K, ).
Applying claim 4.3.3.1 to , q and n = 0 we get a condition p q and an increasing
function g
as there (so they satisfy (
) for ). Note that, as is (forced

to be) increasing, for every we have
if v pos(w
p
, t
p
0
, . . . , t
p
1
) and t (t
p
) is such that nor[t] > 0

then (w, t
p
+1
, t
p
+2
, . . .) ' () < g( + 1), for some w pos(v, t).
We will be done when we show the following claim.
Claim 4.3.3.2. For each A []
there is p
p such that
p
'
Q
s
(K,)
(
k A)( (k) < g(k + 1)).

Proof of the claim: Let A []
. Choose 0 = n
0
< n
1
< . . . < and creatures
t
i
(t
p
ni
, . . . , t
p
ni+11
) and k
i
A such that
(1) n
i
k
i
< n
i+1
, nor[t
i
] minnor[t
p
k
] : n
i
k < n
i+1
,
(2) if w pos(w
p
, t
p
0
, . . . , t
p
ni1
), t
(t
i
), nor[t
] > 0 then there is t

(t
p
ki
) such that nor[t
] > 0 and
(u
pos(w, t
p
ni
, . . . , t
p
ki1
))(u
pos(u
, t
))(v pos(w, t
))(u
_ v)
(possible by ()
0
AB
). Now let p
= (w
p
, t
0
, t
1
, . . .). Plainly, p
s
(K, ), p p
.
We want to show that
p
'
Q
s
(K,)
(
k A)( (k) < g(k + 1)).

So assume not. Thus we nd a condition p
+
p
and k
such that
p
+
'
Q
s
(K,)
(k A)(k
k (k) g(k + 1)).

4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Let i be such that w
p
+
pos(w
p
, t
0
, t
1
, . . . , t
i1
). As we may pass to an
extension of p
+
we may assume that k
< k
i
and g(w
p
+
) > 1. Let j > i be such
that t
p
+
0
(t
i
, . . . , t
j1
). Since (K, ) is condensed we nd [i, j), s (t
)
and v pos(w
p
+
, t
i
, . . . , t
1
) such that
nor[s] > 0 and (u pos(v, s))(u
pos(w
p
+
, t
p
+
0
))(u _ u
).
Now look at our choice of t
: since v pos(w
p
, t
p
0
, . . . , t
p
n
1
) and s (t
), nor[s] >
0, therefore we nd t
(t
p
k
) and u
pos(v, t
p
n
, . . . , t
p
k
1
) such that nor[t
] > 0
and
(u
pos(u
, t
))(u
+
pos(v, s))(u
_ u
+
).
But now look at the choice of p: since u
pos(w
p
, t
p
0
, . . . , t
p
k
1
), t
(t
p
k
) and
nor[t
] > 0, we nd w pos(u
, t
) such that the condition (w, t

p
k
+1
, t
p
k
+2
, . . .)
forces that (k
) < g(k
+ 1). But now, going back, we know that there is u

+
pos(v, s) such that w _ u

+
. Further we nd u
pos(w
p
+
, t
p
+
0
) such that u
+
_ u
.
So look at the condition (u
, t
p
+
1
, t
p
+
2
, . . .). It is stronger than p
+
and it forces that
(k
) < g(k
+ 1), contradicting the choice of p

+
and k
< k
i
k
. This nishes
the proof of the claim and the theorem.
Lemma 4.3.4. Suppose that (K, ) is a strongly nitary and omittory creating
pair with the weak Halving Property which is saturated with respect to nice pre
norms with values in (see 4.2.4(2)). Further suppose that for each t K
(
1
) (s (t))(val[s] val[t]) and
(
2
) if m
0
m
t
dn
< m
t
up
m
1
, s (half(t))
then s [m
0
, m
1
) (half(t [m
0
, m
1
))).
Assume that p Q
(K, ), N
0
, nor[t
p
i
] > 2 for i N
0
, m 1 and
z : pos(w
p
, t
p
0
, . . . , t
p
N01
) PC(K, )
_
nN
pos(w
p
, t
p
0
, . . . , t
p
n1
)
0
, (K, ).
Then there are a nice prenorm H : []
<
(so (K, ) is saturated with
respect to H) and d D
H
(see 4.2.4) such that u
d
= [N
0
, N
1
), H([N
0
, N
1
)) m
and
if t
_
sum
d,u
d
(half(t
p
N0
), . . . , half(t
p
N11
))
_
, v pos(w
p
, t
p
0
, . . . , t
p
N01
), and
nor[t] > 0
then there is t
0
, t
1
, . . .) PC(K, ) such that for each i , for some k
<
t
i
(t
p
N0+k
, . . . , t
p
N0+
) & nor[t
i
]
1
2
minnor[t
p
N0+k
], . . . , nor[t
p
N0+
]
and (w pos(v, t))(z(v, t
0
, t
1
, . . .)) _ w).
Proof. This is essentially [Sh 207, 2.14].
First note that if i
0
< . . . < i
k
, j j
0
< . . . < j
< , i
0
, . . . , i
k
j
0
, . . . , j
(and k < ), w pos(w

p
, t
p
0
, . . . , t
p
j1
) and s
jn
(t
p
jn
) (for n ) then
w, v) val[
sum
(s
i0
[m
t
p
j
dn
, m
si
0
up
), s
i1
, . . . , s
i
k
)]
implies
w, v
0
[m
s
i
k
up
,m
s
i
up
)
) val[
sum
(s
j0
[m
t
p
j
dn
, m
sj
0
up
), s
j1
, . . . , s
j
)].
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Why? Suppose that w, v) val[
sum
(s
i0
[m
t
p
j
dn
, m
si
0
up
), s
i1
, . . . , s
i
k
)]. If j
0
= i
0
this immediately implies w, v`m
sj
0
up
) val[s
j0
[m
t
p
j
dn
, m
sj
0
up
)]. Otherwise necessar-
ily j
0
< i
0
and v`[m
t
p
j
dn
, m
sj
0
up
) is constantly zero. Now, w pos(w
p
, t
p
0
, . . . , t
p
j1
),
so using the assumptions that (K, ) is omittory and (
1
) we get w, v`m
sj
0
up
)
val[s
j0
[m
t
p
j
dn
, m
sj
0
up
)]. Proceeding in this fashion further we get the desired con-
clusion.
Note that above we use val[
sum
(. . .)] and not pos(w,
sum
(. . .)) as we do
not claim that
sum
(. . .) is in K.
Let us dene the function H : []
<
by:
H(u) 0: always,
H(u) 1: if [u N
0
[ > 1, u N
0
= i
0
, . . . , i
k1
(the increasing enumera-
tion), and
if s
(half(t
p
i
)), nor[s
] > 0 (for < k) and v pos(w

p
, t
p
0
, . . . , t
p
N01
)
then there exists t
0
, t
1
, . . .) PC(K, ) such that for each i , for some
k <
t
i
(t
p
N0+k
, . . . , t
p
N0+
) & nor[t
i
]
1
2
minnor[t
p
N0+k
], . . . , nor[t
p
N0+
]
and for some w
v, w) val[
sum
(s
0
[m
t
p
N
0
dn
, m
t
p
i
0
up
), s
1
, . . . , s
k1
)] & z(v, t
0
, t
1
, t
2
, . . .)) _ w,
H(u) n + 1: if for every u
u either H(u
) n or H(u u
) n (for
n > 0).
Note that this denes correctly a nice prenorm on []
<
; for monotonicity use
the remark we started with. Thus (K, ) is saturated with respect to H.
Now it is enough to nd N
1
> N
0
such that H([N
0
, N
1
)) m and then
take d D
H
with u
d
= [N
0
, N
1
). Why? Suppose that we have such an N
1
(and the respective d) and let t (
sum
d,u
d
(half(t
p
N0
), . . . , half(t
p
N11
))), nor[t] >
0. By 4.2.4(2) (the second demand) we have that there are d
D
H
and s
i

(half(t
p
i
)) (for i [N
0
, N
1
)) such that u
d
u
d
and nor[s
i
] > 0 for i u
d
, and
val[
sum
d
,u
d
(s
i
: N
0
i < N
1
)] val[t]. But now look at the denition of H (and
remember the denitions of
sum
,
sum
d
,u
d
; note H(u
d
) > 0).
As H is monotonic, it is enough to nd a set u [ N
0
]
<
with H(u) m.
We will do this by induction on m for all p, N
0
, z.
Case 1: m = 1
For t (half(t
p
N0+i
)) with nor[t] > 0, i x s(t) (t
p
N0+i
) such that
nor[s(t)]
1
2
nor[t] and (w basis(t))(pos(w, s(t)) pos(w, t))
(possible by 2.2.7(2)(b
)). We may additionally require that if t
then s(t) =
s(t
). Let
A
def
= t
0
, t
1
, . . .) PC(K, ) : (i )(t
i
(half(t
p
N0+i
)) & nor[t
i
] > 0).
As (K, ) is strongly nitary each (half(t
p
N0+i
)) is nite (up to
equivalence,
but we may consider representatives only) and thus the space A equipped with the
product topology (of discrete (half(t
p
N0+i
))s) is compact.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

For w pos(w
p
, t
p
0
, . . . , t
p
N01
) and t
0
, t
1
, . . .) A let M(w, t
0
, t
1
, . . .)) be the
unique M < such that
z(w, s(t
0
), s(t
1
), . . .)) pos(w, s(t
0
), . . . , s(t
M1
)).
Note that the function M : pos(w
p
, t
p
0
, . . . , t
p
N01
) A is continuous. Why?
Look at the denition of decision functions: if t
0
, t
1
, . . .) A is such that t
i
for all i < M(w, t
0
, t
1
, . . .)) then
z(w, s(t
0
), s(t
1
), . . .)) = z(w, s(t
0
), s(t
1
), . . .)).
Hence, by compactness of A, the function M is bounded. Let N
1
> 1 be such that
rng(M) N
1
and let N
1
= N
0
+N
1
. We want to show H([N
0
, N
1
)) 1. So suppose
that s
(half(t
p
N0+
)), nor[s
] > 0 (for < N
1
) and v pos(w
p
, t
p
0
, . . . , t
p
N01
).
Look at s
0
, . . . , s
N
1
1
, half(t
p
N1
), half(t
p
N1+1
), . . .) A, and let
s = s(s
0
), . . . , s(s
N
1
1
), s(half(t
p
N1
)), s(half(t
p
N1+1
)), . . .) PC(K, ).
By the choice of N
1
we know that for some k < N
1
:
z(v, s) pos(v, s(s
0
), . . . , s(s
k
)) pos(v, s
0
, . . . , s
k
)
(by the choice of s(t)s). Take w
pos(v, s
0
, . . . , s
N
1
1
) such that z(v, s) _ w
.
Applying (
1
) we get v, w
) val[
sum
(s
0
, . . . , s
N
1
1
)]. To nish this case note
that nor[s(s
)]
1
2
nor[t
p
N0+
] for < N
1
and nor[s(half(t
p
k
))]
1
2
nor[t
p
k
] for
k N
1
.
Case 2: m
= m+ 1 2
Now suppose that we always can nd a nite subset of of the prenorm H at
least m. Thus we nd an increasing sequence N
0
=
0
<
1
< . . . < such that
H([
i
,
i+1
)) m for each i. Consider the space of all increasing functions
such that `N
0
is the identity and (i )((N
0
+i) [
i
,
i+1
)) - it is a compact
space. For each from the space we may consider a condition
(w
p
, t
p
0
, . . . , t
p
N01
, t
p
(N0)
[m
t
p
0
dn
, m
t
p
(N
0
)
up
), t
p
(N0+1)
[m
t
p
(N
0
)
up
, m
t
p
(N
0
+1)
up
), . . .)
(and call it p
) and the respective prenorm H
dened like H but for p
, N
0
, z
(note that p p
(K, ), so z may be interpreted as a decision function for

p
).
Claim 4.3.4.1. For each nite set u :
H
(u) H((k) : k u).

Proof of the claim: Suppose H
(u) 1. We may assume that u N

0
. Let
s
k
(half(t
p
(k)
)), nor[s
k
] > 0 for k u and v pos(w
p
, t
p
0
, . . . , t
p
N01
). By the
assumption (
2
) we know that
s
k
def
= s
k
[m
t
p
k
dn
, m
t
p
k
up
)
_
half(t
p
(k)
[m
t
p
k
dn
, m
t
p
k
up
))
_
= (half(t
p
k
)).
So, as H
(u) 1, we nd t
0
, t
1
, . . .) PC(K, ) such that
(1) t
i
(t
p
ki
, . . . , t
p
ni
) (for some N
0
k
i
n
i
< ),
(2) nor[t
i
]
1
2
minnor[t
p
ki
], . . . , nor[t
p
ni
], and
(3) for some w
v, w) val
_
sum
(s
k
: k u) [m
t
p
N
0
dn
, m
t
p
max(u)
up
)
& z(v, t
0
, t
1
, . . .)) _ w.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

But then also v, w) val
_
sum
(s
k
: k u) [m
t
p
N
0
dn
, m
t
p
(max(u))
up
)
. As
minnor[t
p
ki
], . . . , nor[t
p
ni
] minnor[t
p
(ki1)+1
], . . . , nor[t
p
(ni)
]
and t
i
(t
p
(ki1)+1
, . . . , t
p
(ni)
), we may conclude that 1 H((k) : k u).
Next we easily proceed by induction, nishing the proof of the claim.
By the induction hypothesis for each suitable function we nd N
> N
0
such
that H
([N
0
, N
)) m. By the compactness of the space of all these functions

we nd one n such that H
([N
0
, N
0
+ n)) m (for each ). Look at the interval
[N
0
,
n
) we claim that its H-norm is greater or equal than m + 1. Why? By
the choice of
i
s we have that H([N
0
,
n
)) m. Suppose that u [N
0
,
n
). If
u [
k
,
k+1
) ,= for each k < n then we may take a function from our space
such that [[N
0
, N
0
+n)] u. But H
([N
0
, N
0
+n)) m and by 4.3.4.1
m H((k) : N
0
k < N
0
+n) H(u).
If u [
k
,
k+1
) = for some k < n then necessarily
m H([
k
,
k+1
)) H([N
0
,
n
) u).
This nishes the induction and the proof of the lemma.
Theorem 4.3.5. Assume that (K, ) is a strongly nitary and omittory creat-
ing pair with the weak Halving Property. Further suppose that (K, ) is saturated
with respect to nice prenorms with values in and for each t K:
(
1
) (s (t))(val[s] val[t]),
(
2
) if m
0
m
t
dn
< m
t
up
m
1
, s (half(t))
then s [m
0
, m
1
) (half(t [m
0
, m
1
))),
(
3
) (t [m
0
, m
1
)) = s [m
0
, m
1
) : s (t).
Then the creating pair (K, ) is of the ABtype. Consequently if (K, ) is addi-
tionally condensed then the forcing notion Q
s
(K, ) is almost
-bounding.
Proof. By 4.3.2, 4.3.4 and 4.3.3.
Remark 4.3.6. The assumptions of 4.3.5 may look very complicated, but in
the real examples they are relatively easy to check and appear naturally. Sometimes
it is easy to check directly that a creating pair is of the ABtype, but then it may
happen that it is not condensed (this happens e.g. for (K
2.4.5
,
2.4.5
); see 4.4.1). To
get that Q
s
(K, ) is almost
bounding we do not have to require that (K, )

is condensed, but then we should strengthen the demands of 4.3.1(2) a little bit.
Definition 4.3.7. We say that a creating pair (K, ) is of the AB
+
type if it
satises the demand ()
0
AB
of 4.3.1 and the following strengthening of ()
1
AB
:
(
+
)
1
AB
if p Q
s
(K, ), N
0
, and
z : pos(w
p
, t
p
0
, . . . , t
p
N01
) PC(K, )
_
kN0
pos(w
p
, t
p
0
, . . . , t
p
k1
)
0
, (K, ) then there are N
1
> N
0
and t

(t
p
N0
, . . . , t
p
N11
) such that
nor[t
]
1
2
minnor[t
p
N0
], . . . , nor[t
p
N11
]
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

and for each v pos(w
p
, t
p
0
, . . . , t
p
N01
, t
) such that
(k [m
t
dn
, m
t
up
))(v(k) ,= 0)
there is t
0
, t
1
, . . .) PC
s
(K, ) such that t
p
N0
, t
p
N0+1
, . . .) t
0
, t
1
, . . .)
and z(v`m
t
dn
, t
0
, t
1
, t
2
, . . .)) _ v.
Theorem 4.3.8. Suppose that (K, ) is a nitary and omittory creating pair
of the AB
+
type such that for each t K
(
0
) if nor[t] > 1 and u basis(t) then [pos(u, t)[ > 2 and
(
3
) (t [m
0
, m
1
)) = s [m
0
, m
1
) : s (t).
Then the forcing notion Q
s
(K, ) is almost
bounding.
Proof. It is fully parallel to 4.3.3. First one proves that
Claim 4.3.8.1. If is a Q
s
, q Q
s
(K, )
and n , then there are a condition p Q
s
(K, ) and an increasing g
such that q
s
n
p and for every
(
+
) if w pos(w
p
, t
p
0
, . . . , t
p
n1
, . . . , t
p
n+
) is such that w`[m
t
p
n+
dn
, m
t
p
n+
up
) ,= 0
then the condition (w, t
p
n++1
, t
p
n++2
, . . .) decides the value of (g()) and
the decision is smaller than g( + 1).
Proof of the claim: Repeat the proof of 4.3.3.1.
Next, assuming that is a name for a strictly increasing function in
, n = 0,
and q Q
s
(K, ), we take the condition p
0
q and the function g
given
by 4.3.8.1. They have the property that for each
if v pos(w
p
, t
p
0
, . . . , t
p
) is such that (k [m
t
p
dn
, m
t
p
up
))(v(k) ,= 0)
then (v, t
p
+1
, t
p
+2
, . . .) ' () < g( + 1).
To show that for every A []
there is p
p such that
p
'
Q
s
(K,)
(
k A)( (k) < g(k + 1))

we slightly modify the proof of 4.3.3.2. So suppose A []
. Choose 0 = n
0
< n
1
<
. . . < and k
i
A such that n
i
k
i
< n
i+1
and let t
p
i
= t
p
ki
[m
t
p
n
i
dn
, m
t
p
n
i+1
dn
). Look
at the condition p
= (w
p
, t
p
0
, t
p
1
, . . .). Assume that p
+
p
, k
, g(w
p
+
) > 1
and k
i
> k
, where i is such that w

p
+
pos(w
p
, t
p
0
, . . . , t
p
i1
). Take j > i such
that t
p
+
0
(t
p
i
, . . . , t
p
j1
). Choose v pos(w
p
+
, t
p
+
0
) such that v() ,= 0 for some
[m
t
p
+
0
dn
, m
t
p
+
0
up
) (exists by (
0
)). Let i
[i, j) be such that [m

t
p
dn
, m
t
p
up
). By
smoothness we know that v`m
t
p
up
pos(v`m
t
p
dn
, t
p
i
), and therefore, by (
3
) and the
choice of t
p
i
we get
v`m
t
p
k
i
dn
pos(w
p
, t
p
0
, . . . , t
p
k
i
) and m
t
p
k
i
dn
< m
t
p
k
i
up
.
Hence (v`m
t
p
k
i
up
, t
p
k
i
+1
, t
p
k
i
+2
, . . .) ' (k
i
) < g(k
i
+ 1), so we are done.
Let us nish this section by proving a parallel of 4.3.5 for the treelike forcing
notions.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Theorem 4.3.9. Suppose that

i
H(i) is countable and (K, ) is a t-omittory
tree creating pair for H. Then the forcing notion Q
tree
0
(K, ) is almost
-bounding.
Proof. Let p Q
tree
0
(K, ) and let

f be a Q
tree
0
(K, )name for a function
in
. For simplicity we may assume that for every t T

p
we have [pos(t)[ =
or at least that above each T
p
we nd may an innite front of T
p
[]
(compare
3.1.1(2)).
Like in 2.3.6(2) we may construct a condition q Q
tree
0
(K, ) stronger than p
and fronts F
n
of T
q
such that for all n :
(1) F
n
=
s
: s
n + 1
(just a xed enumeration), and for each s
n+1
:
(2) if m then
s
<
s
m
,
(3)
s
m
: m is a front of T
q
[
s
]
and nor[t
q
s] n + 1,
(4) the condition q
[
s
]

f`
_
n + 1 +

kn
s(k)
_
.
For m let g(m) be
1 + max < : (s
m+1
)
_

k<g(s)
s(k) 4m & q
[
s
]
'

f(m) =
_
.
Let A []
. For s
<
choose m(s) A such that g(s) +

k<g(s)
s(k) < m(s)
and let c(s) = s
m(s)). Note that q

[
c(s)
]
'

f(m(s)) < g(m(s)). Now build
inductively a condition p
Q
tree
0
(K, ) such that q
0
0
p
, F
0
T
p
and for each

n :
if s
2n + 1
,
s
F
2n
T
p
then
t
p
s
_
t
p
: ( F
2n+1
)(
s
_ _ )
_
, pos(t
p
s ) pos(t
q
c(s)
) and
nor[t
p
s] 2n + 1
(possible as (K, ) is t-omittory and by the third requirement on F
n
s). We claim
that
p
'
Q
tree
0
(K,)
(
m A)(

f(m) < g(m)).
To see this suppose that p
, N . Choose s
2N + 1
such that
s
F
2N
dcl(T
p
). Then necessarily pos(t

p
s )dcl(T
p
) ,= so we may choose T
p
such that some initial segment of is in pos(t

p
s) pos(t
q
c(s)
) (see the construction
of p
). But now we conclude

(p
)
[]
'
Q
tree
0
(K,)

f(m(s)) < g(m(s))
what nishes the proof as N < m(s) A.
4.4. Examples
Let us start with noting that the BlassShelah forcing notion Q
s
(K
2.4.5
,
2.4.5
)
is a good application of the notions introduced in this section.
Proposition 4.4.1. The creating pair (K
2.4.5
,
2.4.5
) (see the end of the con-
struction for 2.4.5) is meagering, of the AB
+
type and satises the demands (
0
),
(
3
) of 4.3.8. Consequently, the forcing notion Q
s
(K
2.4.5
,
2.4.5
):
() makes ground model reals meager,
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

4.4. EXAMPLES 81
() adds an unbounded real,
() is almost
bounding,
() does not add Cohen reals.
Proof. To show that (K
2.4.5
,
2.4.5
) is meagering assume that (t
0
, . . . , t
n1
)
PFC(K
2.4.5
,
2.4.5
), t
2.4.5
(t
0
, . . . , t
n1
) and k
i
: i < n) are such that nor[t
i
]
2, m
ti
dn
k
i
< m
ti
up
(for i < n) and nor[t] 2. Fix u

i<m
t
0
dn
H(i). By the
denition of K
2.4.5
(see clause () there) we know that 2 nor[t] dp
1
0
(A
t
u
).
Moreover, by the denition of
2.4.5
we know that no element of A
t
u
is included in
k
i
: i < n (by clause () of 2.4.5; remember a [m
ti
dn
, m
ti
up
) A
ti
u(am
t
i
dn
)

for all a A
t
u
). Consequently, if A
s
u
= a A
t
u
: a k
i
: i < n = , then
dp
1
(A
s
u
) dp
1
(A
t
u
) 1 and dp
1
0
(A
s
u
) nor[t] 1. This determines a condition
s
2.4.5
(t) which is as required in 4.1.2(1).
2.4.5
,
2.4.5
) satises the conditions (
0
), (
3
) of
4.3.8 (actually, (
3
) is satised if interpreted modulo
2.4.5
, but this makes no
problems). The proof that (K
2.4.5
,
2.4.5
) is of the AB
+
type follows exactly the
lines of [BsSh 242, 2.6] (see [BaJu95, 7.4.20] too) and is left to the reader.
Consequently, the assertion () follows from 4.1.3(1), clause () is a conse-
quence of 4.1.1 and () follows from 4.3.8. To show () one uses 2.2.6, or see
6.3.8.
Note, that if

W is interpreted as a name for an innite subset of , then
'
Q
s
(K
2.4.5
,
2.4.5
)
(X []
V)([

W X[ < or [

W X[ < ).
Thus forcing with Q
s
(K
2.4.5
,
2.4.5
) makes ground model reals null too.
Now we will present an application of forcing notions determined by omittory
creating pairs with the weak Halving Properties to questions coming from localizing
subsets of . These problems were studied in [RoSh 501] and our example is
built in a manner similar to that of the forcing notion constructed in [RoSh 501,
2.4]. Moreover, all these examples are relatives of the forcing notion presented in
[Sh 207]. The creating pair constructed there can be build like (K
4.4.2
,
4.4.2
) for
1.
Example 4.4.2. Let
be a non-decreasing function, (0) > 0. We

construct a creating pair (K
4.4.2
,
4.4.2
) which:
() is strongly nitary, forgetful and omittory,
() has the weak Halving Property,
() is saturated with respect to nice pre-norms with values in ,
() is condensed and satises the demands (
1
), (
2
) and (
3
) of 4.3.5
[thus, by 4.3.5, (K
4.4.2
,
4.4.2
) is of the ABtype],
() is anti-big and meagering.
Construction. Let H(m) = 2 for m . First we describe which creatures
t CR[H] are taken to be in K
4.4.2
. So, t = (nor[t], val[t], dis[t]) K
4.4.2
if:
dis[t] = (T[t], L[t], R[t], D[t], NOR[t], m
t
dn
, m
t
up
), where, letting T = T[t],
L = L[t], R = R[t], D = D[t] and NOR = NOR[t]:
(a) T is a nite tree, D T : succ
T
() ,= and
(i) ( T D)(succ
T
() = or [succ
T
()[ = (L())),
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

(ii) if T D and succ
T
() then either D or succ
T
() =
,
(b) L, R : T [m
t
dn
, m
t
up
) are such that for each T
(i) L() R(),
(ii) if succ
T
() then L() L() R() R(),
(iii) if succ
T
() = then L() = R(),
(iv) if
0
,
1
succ
T
(),
0
,=
1
then
[L(
0
), R(
0
)] [L(
1
), R(
1
)] = ,
(c) NOR is a function on D such that for each D, NOR() is a nice
pre-norm on succ
T
() with values in ,
if ) / D and ( succ
T
()))(succ
T
() = ), or D = then nor[t] = 0,
otherwise nor[t] = minNOR()(succ
T
()) : D,
val[t] = u, v) 2
m
t
dn
2
m
t
up
: u < v & i [m
t
dn
, m
t
up
) : v(i) = 1
L() : T & succ
T
() = .
For t K
4.4.2
we dene
nor
0
[t] =
_
minNOR[t]()(succ
T[t]
()) : D[t] if D[t] ,= ,
0 otherwise.
Note that nor[t] nor
0
[t] and in most cases they agree. One could use nor
0
[t] as
the norm of t and get the same forcing notion. We take nor[t] for technical reasons
only. Now we are going to describe a composition operation
4.4.2
on K
4.4.2
by
giving basic operations which may be applied to creatures from K
4.4.2
.
(1) For a creature t K
4.4.2
let half(t) K
4.4.2
be such that
if nor[t] < 2 then half(t) = t,
if nor[t] 2 then val[half(t)] = val[t], T[half(t)] = T[t], L[half(t)] = L[t],
R[half(t)] = R[t], D[half(t)] = D[t] and
if D[half(t)], A succ
T[half(t)]
() then
NOR[half(t)]()(A) = max0, NOR[t]()(A)
nor[t]
2
|.
[Thus m
half(t)
dn
= m
t
dn
, m
half(t)
up
= m
t
up
and nor[half(t)] = nor[t]
nor[t]
2
| when
nor[t] 2.]
(2) For t K
4.4.2
, m
0
m
t
dn
, m
1
m
t
up
let s = S
m0,m1
(t) K
4.4.2
be a creature
such that dis[s] = (T[t], L[t], R[t], D[t], NOR[t], m
0
, m
1
), nor[s] = nor[t] and
val[s] =
_
u, v) 2
m
0
2
m
1
: u < v & v`[m
0
, m
t
dn
) = 0 & v`[m
t
up
, m
1
) = 0 &
& v`m
t
dn
, v`m
t
up
) val[t].
Thus, essentially, S
m0,m1
(t) = t [m
0
, m
1
), the small dierence in the denition of
dis is immaterial.
(3) For t K
4.4.2
let
4.4.2
(t) consist of all s K
4.4.2
such that
(i) m
s
dn
= m
t
dn
, m
s
up
= m
t
up
, T[s] T[t], D[s] = D[t] T[s],
(ii) ( T[s])(succ
T[s]
() = succ
T[t]
() = ) (thus if T[s] D[s]
then succ
T[s]
() = succ
T[t]
()),
(iii) L[s] = L[t]`T[s] and R[s] = R[t]`T[s],
(iv) if D[s], A succ
T[s]
() then NOR[s]()(A) = NOR[t]()(A).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

4.4. EXAMPLES 83
(4) Suppose that t
0
, . . . , t
k1
K
4.4.2
are such that k = (L[t
0
]())), m
ti
up
m
ti+1
dn
(for i < k 1) and
(i < k)() D[t
i
] or succ
T[ti]
()) = ).
Let s = S
(t
0
, . . . , t
k1
) K
4.4.2
be a creature such that
(i) m
s
dn
= m
t0
dn
, m
s
up
= m
t
k1
up
and T[s] is a tree such that [succ
T[s]
())[ = k
and for every succ
T[s]
()) there is a unique i = i() < k such that
T[s] : _ =
T[t
i
] & L[s]() = L[t
i
]()) & R[s]() = R[t
i
]()),
(ii) D[s] =
: succ
T[s]
()) &
D[t
i()
],
(iii) L[s]()) = L[t
0
]()), R[s]()) = R[t
k1
]()) and for all succ
T[s]
())
(
T[t
i()
])(L[s](
) = L[t
i()
](
) & R[s](
) = R[t
i()
](
)),
(iv) if succ
T[s]
()),
D[t
i()
] and A succ
T[t
i()
]
(
) then
NOR[s](
)(
A) = NOR[t
i()
](
)(A).
(5) Suppose that H : T(m) is a nice pre-norm and t
0
, . . . , t
m1
K
4.4.2
are such that m
ti
up
m
ti+1
dn
for i < m1. Let s = S
H
(t
0
, . . . , t
m1
) K
4.4.2
be a
creature such that
(i) m
s
dn
= m
t0
dn
, m
s
up
= m
tm1
up
and T[s] is a tree such that [succ
T[s]
())[ = m
and for every succ
T[s]
()) there is a unique j = j() < m such that
T[s] : _ =
T[t
j
] & L[s]() = L[t
j
]()) & R[s]() = R[t
j
]()),
(ii) D[s] = )
: succ
T[s]
()) &
D[t
j()
],
(iii) L[s]()) = L[t
0
]()), R[s]()) = R[t
m1
]()) and for every succ
T[s]
())
(
T[t
j()
])(L[s](
) = L[t
j()
](
) & R[s](
) = R[t
j()
](
)),
(iv) if A succ
T[s]
()) then NOR[s]())(A) = H(j() : A),
(v) if succ
T[s]
()),
D[t
j()
] and A succ
T[t
j()
]
(
) then
NOR[s](
)(
A) = NOR[t
j()
](
)(A).
Note that, under the respective assumptions, the procedures described in (1)(5)
above determine creatures in K
4.4.2
, though (in cases (4) and (5)) not uniquely:
there is some freedom in dening succ
T[s]
()). However, this freedom becomes
irrelevant when we identify creatures that look the same. The last operation (
4.4.2
below) is a way to describe which creatures are identied.
(6) For t K
4.4.2
, let
4.4.2
(t) consist of all creatures s K
4.4.2
such that
m
s
dn
= m
t
dn
, m
s
up
= m
t
up
and there is an (order) isomorphism : T[s] T[t]
which preserves L, R, D and NOR.
Finally, if t
0
, . . . , t
m1
K
4.4.2
are such that m
ti
up
= m
ti+1
dn
(for i < m 1)
then
4.4.2
(t
0
, . . . , t
m1
) consists of all creatures s K
4.4.2
such that m
s
dn
= m
t0
dn
,
m
s
up
= m
tm1
up
and s may be obtained from t
0
, . . . , t
m1
by use of the operations
half, S
m0,m1
, S
, S
H
,
4.4.2
and
4.4.2
(with suitable parameters).
Let us check that (K
4.4.2
,
4.4.2
) has the required properties. It should be
clear that (K
4.4.2
,
4.4.2
) is a nitary, forgetful and omittory creating pair. The
relation
4.4.2
(see 1.1.4(3)) is an equivalence relation on K
4.4.2
and
4.4.2
depends
on
4.4.2
equivalence classes only (remember the denition of
4.4.2
; note that
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

4.4.2
(t
0
) =
4.4.2
(t
1
) implies that t
0
, t
1
are the same up to the isomorphism of
the trees T[t
0
], T[t
1
]). Thus the value of
4.4.2
(t
0
, . . . , t
m1
) does not depend
on the particular representation of the trees T[t
i
] (for i < m). Hence, if t
4.4.2
(t
0
, . . . , t
m1
) then we may think that T[t] is a tree built of T[t
0
], . . . , T[t
m1
]
in the following sense. There are s
i

4.4.2
(t
i
) (for i < m), a front F of T[t] and
a one-to-one mapping : F m such that for F
T[t] : _ =
T[s
()
]
and the L[t], R[t], D[t] above in T[t], F, look like L[s
()
], R[s
()
], D[s
()
]
(but the norms given by NOR may be substantially dierent, still their values
may be only smaller). Now it should be clear that (K
4.4.2
,
4.4.2
) is strongly
nitary. The pair (K
4.4.2
,
4.4.2
) has the weak Halving Property as witnessed
by the function half dened in (1) above. [Why? Note that if t
0
K
4.4.2
,
nor[t
0
] 2, t
4.4.2
(half(t
0
)), nor[t] > 0 (so nor[t] 1) then t is obtained
from t
0
by alternate applications of half and shrinking (i.e.
4.4.2
). Look at the
tree T[t] with L[t], R[t], D[t]: necessarily the last three objects are the restrictions
of L[t
0
], R[t
0
], D[t
0
] to T[t]. Let s
4.4.2
(t
0
) be such that T[s] = T[t] (it should be
clear clear that there is one; actually the s is uniquely determined by the tree T[s]).
Since in the process of building t the norms were decreased only, and we started with
half(t
0
), we may conclude that nor[s] nor[t] +
1
2
nor[t
0
]|
1
2
nor[t
0
]. Clearly
val[s] = val[t].]
Suppose that H
0
: T() is a nice pre-norm, d D
H0
(see 4.2.4). Note
that
sum
d,u
d
(t
i
: n
0
i < n
1
) (for t
i
: n
0
i < n
1
) PFC(K
4.4.2
,
4.4.2
),
u
d
[n
0
, n
1
)) corresponds to a creature t obtained from t
n0
, . . . , t
n11
by suit-
able applications of S
m0,m1
and S
H
(the last for the pre-norm H = H
0
`T(u
d
),
the rst is to omit creatures t
i
for i [n
0
, n
1
) u
d
). The dierence is in dis, but
this causes no real problem as we may read H
0
`T(u
d
) from the resulting creature
(and it is essentially the NOR[t]())). We could have changed the denition of
(K
4.4.2
,
4.4.2
) to make this correspondence more literal, but that would result in
unnecessary complications in the denition. Now note that
if t
4.4.2
(
sum
d,u
d
(t
0
, . . . , t
m1
)), nor[t] > 0 and d
D
H0
is such
that u
d
succ
T[t]
()) and s
i

4.4.2
(t
i
) for i < m are such that
succ
T[t]
())
T[s
j()
] = T[t] : _ ,
then val[t] = val[
sum
d
,u
d
(s
0
, . . . , s
m1
)]. Moreover, if nor[t
i
] > 0
for i < m then nor[s
i
] > 0 for i u
d
.
This shows that (K
4.4.2
,
4.4.2
) is saturated with respect to nice pre-norms with
values in .
Suppose that t
4.4.2
(t
0
, . . . , t
n1
), nor[t] > 0, nor[t
i
] > 0 (for i < n). Then
for some i < n and T[t] we nd s
4.4.2
(t
i
) such that
T[t] : _ =
T[s]
and L[t], R[t] and D[t] above are like L[s], R[s], D[s], but remember that NOR[t]
may have nothing in common with NOR[s]: the operation half may be involved.
However, by the denition of nor[t], we know that if
D[s] then
NOR[t](
)(succ
T[t]
(
)) > 0,
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

4.4. EXAMPLES 85
and this is enough to conclude that NOR[s](
)(succ
T[s]
(
)) > 0. Moreover D[s] ,=

as nor[t
i
] > 0. Consequently nor[s] > 0. Since
v`[m
t0
dn
, m
s
dn
) = 0 & v, u) val[s] (u
)(v, u
) val[t] & u _ u
)
we conclude that (K
4.4.2
,
4.4.2
) is condensed.
One easily checks that the demands (
1
), (
2
) and (
3
) of 4.3.5 are satised
(remember that t [m
0
, m
1
) is, basically, S
m0,m1
(t)).
Finally, let us check that (K
4.4.2
,
4.4.2
) is meagering and anti-big. For the rst
note that if t K
4.4.2
, nor[t] > 2 and T[t] is a maximal node of T[t] then
there is s
4.4.2
(t) such that / T[s] and nor[s] nor[t] 1. [Why? Take the
shortest < such that D[t] there must be one as nor[t] > 0 and choose
s
4.4.2
so that T[s] = T[t] : < (
_ ), where
succ
T[t]
()
is such that
_ . Since NOR[t]() is a nice pre-norm we have

NOR[t]()(succ
T[s]
()) NOR[t]()(succ
T[t]
()) 1
and hence s is as required.] Now suppose that (t
0
, . . . , t
n1
) PFC(K
4.4.2
,
4.4.2
),
nor[t
i
] > 2, k
i
[m
ti
dn
, m
ti
up
) and t
4.4.2
(t
0
, . . . , t
n1
), nor[t] > 2. Then there is
a front F of T[t] such that for every F, for a unique i = i() < n and some
s
4.4.2
(t
i
):
T[t] : _ =
T[s
i
]
and D[t], L[t], R[t] above look like D[s
i
], L[s
i
], R[s
i
] (but NOR[t] might be dierent
than that of s
i
: the values may be smaller). Now apply the previous remark to
choose s
4.4.2
(s
i
) such that nor[s
i
] nor[s
i
] 1 and k
i
/ L[s
i
]() :
T[s
i
] & succ
T[s
i
]
() = . Finally let s
4.4.2
(t) be such that F T[s] and for
each F
T[s] : _ =
T[s
i()
].
It is easy to check that this determines a creature s as required in 4.1.2(1) for
k
i
, t
i
: i < n), t.
To verify that (K
4.4.2
,
4.4.2
) is anti-big dene colourings
c
t
:
_
ubasis(t)
pos(u, t) 3
for t K
4.4.2
by:
c
t
(v) =
_
_
_
0 if [k [m
t
dn
, m
t
up
) : v(k) = 1[ is even > 0,
1 if [k [m
t
dn
, m
t
up
) : v(k) = 1[ is odd ,
2 if v`[m
t
dn
, m
t
up
) = 0
[m
t
dn
,m
t
up
)
.
Suppose (t
0
, . . . , t
n1
) PFC(K
4.4.2
,
4.4.2
), t
4.4.2
(t
0
, . . . , t
n1
), nor[t
i
] > 1
and nor[t] > 1. Clearly for some i < n
[L[t]() : L[t
i
]()) L[t]() R[t
i
]()) and succ
T[t]
() = [ 2.
Let u basis(t
0
) =

i<m
t
dn
2. Take v
0
, v
1

i<m
t
up
2 such that for = 0, 1
u < v, [k [m
t
dn
, m
t
up
) : v
(k) = 1[ = + 1, and
k [m
t
dn
, m
t
up
) : v
(k) = 1 is contained in
L[t]() : L[t
i
]()) L[t]() R[t
i
]()) & T[t] & succ
T[t]
() = .
Now check that the v
0
, v
1
are as required in 4.1.2(2) for t
i
: i < n), t.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Corollary 4.4.3. Q
s
(K
4.4.2
,
4.4.2
) is a proper almost
bounding forcing
notion which makes ground reals meager and adds a Cohen real.
Proof. By 4.4.2, 4.3.5 and 4.1.3.
Definition 4.4.4. (1) For an innite set X []
let
X
: X be
the increasing enumeration of X.
(2) Let
be non-decreasing. We dene relations S
dn
, S
up
[]
[]
by
(X, Y ) S
dn
if and only if
(
n )(i < (
Y
(n)))([[
Y
(n +i),
Y
(n +i + 1)) X[ 2),
(X, Y ) S
up
if and only if
(
n )(i < (
Y
(n + 1)))([[
Y
(n +i),
Y
(n +i + 1)) X[ 2).
If is a constant function, say k, then S
may be called S
k
.
Remark 4.4.5. We will consider the notions of S
dn
and S
up
localizations as
given by 0.2.2 for these relations as well as the corresponding dominating numbers
d(S
dn
) and d(S
up
). Note that for a non-decreasing
, S
up
localization implies
S
dn
localization (and d(S
dn
) d(S
up
)). If (
n )((n) (n)) then S

localization implies S
localization and for eventually constant , S
up
localization
is the same as S
dn
localization.
Proposition 4.4.6. Suppose ,
are non-decreasing functions such that

(
n )(1 (n) < (n)). Then the forcing notion Q
s
(K
4.4.2
,
4.4.2
) does
not have the S
up
localization property.
Proof. This is parallel to [RoSh 501, 2.4.3]. The main step is done by the
following claim, which is essentially a repetition of [RoSh 501, 2.4.2].
Claim 4.4.6.1. Suppose that t K
4.4.2
is such that nor
0
[t] 11. Let Y []
.
Then there is s
4.4.2
(t) such that nor
0
[s] nor
0
[t] 10 and for every n
there is i (
Y
(n + 1)) such that
L(s)() : T[s] & succ
T[s]
() = [
Y
(n +i),
Y
(n +i + 1)) = .
Proof of the claim: The proof is by induction on the height of the tree T[t].
One could try just to apply the inductive hypothesis to creatures determined by
T[t] : _ for each succ
T[t]
()). However, this would not be enough.
What we need to do is to shrink t to separate the sets
L[t]() : _ & T[t] & succ
T[t]
() =
for distinct succ
T[t]
()) by intervals [
Y
(m),
Y
(m+1)). This will prevent bad
events occurring above distinct succ
T[t]
()) from accumulating. The shrinking
procedure depends on the character of ) in T[t], so the arguments brake into two
cases.
Case 1: ) D[t].
Let
A
0
= succ
T[t]
()) : (m )(
Y
(2m) L[t]() R[t]() <
Y
(2m + 1)),
A
1
= succ
T[t]
()) : (m )(
Y
(2m1) L[t]() R[t]() <
Y
(2m)),
A
2
= succ
T[t]
()) : (L[t](), R[t]()] Y ,= .
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

4.4. EXAMPLES 87
Note that A
0
A
1
A
2
= succ
T[t]
()) and NOR[t]())(succ
T[t]
())) nor
0
[t]. As
NOR[t]()) is a nice pre-norm, at least one of the following holds:
(0) NOR[t]())(A
0
) nor
0
[t] 2,
(1) NOR[t]())(A
1
) nor
0
[t] 2,
(2) NOR[t]())(A
2
) nor
0
[t] 2.
Suppose that (0) holds. Let s
4.4.2
(t) be a creature such that
T[s] = T[t] : ( A
0
)( _ or < ).
[It should be clear that this uniquely denes the creature s and nor
0
[s] nor
0
[t]
2.] Note that by the denition of the set A
0
, if
1
,
2
A
0
are distinct and
R[t](
1
) L[t](
2
) then there are m
1
< m
2
< such that
Y
(2m
1
) L[s](
1
) R[s](
1
) <
Y
(2m
1
+ 1) <
Y
(2m
2
)
L[s](
2
) R[s](
2
) <
Y
(2m
2
+ 1).
Hence we may conclude that for every n there is i < 2 such that
L[s]() : T[s] & succ
T[s]
() = [
Y
(n +i),
Y
(n +i + 1)) = ,
and thus s is as required in the assertion of the claim.
If (1) holds then we may proceed in the same manner (considering the set A
1
).
So suppose that (2) holds true. Divide the set A
2
into three sets A
0
2
, A
1
2
, A
2
2
such
that for each i < 3 and
1
,
2
A
i
2
with R[t](
1
) < L[t](
2
), there is m such
that
R[t](
1
) <
Y
(m) <
Y
(m + 1) < L[t](
2
)
[e.g. each A
i
2
contains every third element of A
2
counting according to the values
of L[t]()]. For some i < 3 we have
NOR[t]())(A
i
2
) NOR[t]())(A
2
) 2 nor
0
[t] 4.
Fix A
i
2
for a moment and let T
= T[t] : < or _ . If T
D[t] = )
then necessarily T
does not contain sequences of length > 2 (remember clause

(a)(ii) of the denition of K
4.4.2
) and
[ T
: succ
T[t]
() = [ 1, (L[t]()).
Let n
be maximal such that

Y
(n
) L[t]() (if there is no such n
the we
let n
= 1). Then, for every n n
, (
Y
(n + 1)) (
Y
(n
+ 1)) (L[t]())
(as is non-decreasing). Hence, letting T
= T
we have that for each n there

is i (
Y
(n + 1)) such that
L[t]() : T
& succ
T[t]
() = [
Y
(n +i),
Y
(n +i + 1)) = .
If T
D[t] ,= ) then we may look at a creature t
4.4.2
such that T[t
] =
T[t] and L[t
], R[t
], D[t
], NOR[t
] are copied in a suitable manner

from t (so they are restrictions of the corresponding objects to T
) and m
t
dn
= m
t
dn
,
m
t
up
= m
t
up
. Clearly nor
0
[t
] nor
0
[t] and the height of T[t
] is smaller than that

of T[t]. Thus we may apply the inductive hypothesis and we nd s
4.4.2
(t
)
such that nor
0
[s
] nor
0
[t
] 10 nor
0
[t] 10 and for all n there is
i (
Y
(n + 1)) such that
L[s
]() : T[s
] & succ
T[s
]
() = [
Y
(n +i),
Y
(n +i + 1)) = .
Let T
T[s
].
Look at the tree T
=

A
i
2
T
. It determines a creature s
4.4.2
(t) (i.e. s is such
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

that T[s] = T
, D[s] = D[t] T
etc). Clearly nor

0
[s] nor
0
[t] 10 and it is easy
to check that s is as required (remember the choice of the A
i
2
and the T
s).
Case 2: ) / D[t].
Since T[t] ,= ) (as nor
0
[t] > 0) we have [succ
T[t]
())[ = (L[t]())). Moreover,
for each succ
T[t]
())
either D[t] or succ
T[t]
() =
(remember the demand (a)(ii) of the denition of K
4.4.2
). Note that necessarily
D[t] succ
T[t]
()) ,= (as nor
0
[t] > 0). Fix D[t] succ
T[t]
()).
Choose a set A
succ
T[t]
() such that NOR[t]()(A
) nor
0
[t] 5 and one of
the following holds:
(m )( A
)(
Y
(m) L[t]() R[t]() <
Y
(m+ 1)),
there are m
0
< m
1
< m
2
< m
3
< such that
( A
)(
Y
(m
1
) L[t]() R[t]() <
Y
(m
2
)) and
( succ
T[t]
())(L[t]()
Y
(m
0
)) and
( succ
T[t]
())(R[t]()
Y
(m
3
)).
Why is the choice of A
possible? For m let

B
m
0
= succ
T[t]
() :
Y
(m) L[t]() R[t]() <
Y
(m+ 1).
If there is m such that NOR[t]())(B
m
0
) nor
0
[t] 5 then we may take the respec-
tive B
m
0
as A
. So suppose that
(m )(NOR[t]())(B
m
0
) < nor
0
[t] 5).
Let B
0
=

m
B
m
0
and suppose that NOR[t]()(B
0
) nor
0
[t] 1. Let k
0
, k
1
be the
two smallest elements of m : B
m
0
,= and let k
2
, k
3
be the two largest elements
of this set (note that [m : B
m
0
,= [ 6; remember that NOR[t]() is a nice
pre-norm). We let
A
= succ
T[t]
() :
Y
(k
1
+ 1) L[t]() R[t]() <
Y
(k
2
),
and m
0
= k
0
+ 1, m
1
= k
1
+ 1, m
2
= k
2
, m
3
= k
3
. Easily NOR[t]()(A
)
NOR[t]()(B
0
) 4 nor
0
[t] 5 and since B
k0
0
,= , B
k3
0
,= we see that A
is as
required. So we are left with the possibility that NOR[t]()(B
0
) < nor
0
[t] 1. In
this case we have
NOR[t]()( succ
T[t]
() : (L[t](), R[t]()] Y ,= ) nor
0
[t] 1.
Let
0
,
1
,
2
,
3
succ
T[t]
() B
0
be such that L[t](
0
) < L[t](
1
) are the rst two
members of L[t]() : succ
T[t]
() B
0
and L[t](
2
) < L[t](
3
) are the last two
members of this set. Let m
i
be such that
Y
(m
i
) (L[t](
i
), R[t](
i
)] (for i < 4)
and let
A
= succ
T[t]
() B
0
:
Y
(m
1
) L[t]() R[t]() <
Y
(m
2
).
Since
NOR[t]()(succ
T[t]
() B
0

0
,
1
,
2
,
3
) nor
0
[t] 5,
one easily checks that A
is as required.
Let T
T[t] & ( A
)( <
or
_ ). We would
like to apply the inductive hypothesis to the creature determined by T
(with
L, R, D and NOR copied in a suitable way from t). However, this creature may
have too small norm: it may happen that NOR[t]()(A
) < 11. But we may repeat

4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

4.4. EXAMPLES 89
the procedure of Case 1, noticing that the inductive hypothesis was applied there
above some elements of succ()). Here, this corresponds to applying the inductive
hypothesis to creatures determined by
T[t] for some A
, and
these creatures have norms not smaller than nor
0
[t]. Consequently we will get a
tree
T
: <
& ( A
)( _
)
corresponding to a creature s
with nor
0
[s
] nor
0
[t] 10 and such that for each
n , for some i (
Y
(n + 1)) we have
L[t](
) :
& succ
T[t]
(
) = [
Y
(n +i),
Y
(n +i + 1)) = .
[Note that the procedure of Case 1 may involve further shrinking of A
and drop-
ping the norm by 4. Still, 5 +4 < 10 so the norm of s
is above nor
0
[t] 10.] Next
let
T
=
_
T
: succ
T[t]
()) D[t] succ
T[t]
()) ),
and let s be the restriction of the creature t to T
. Check that s
4.4.2
(t) is as
required. This nishes the proof of the claim.
Now we may prove the proposition. We are going to show that, in the model
V
Q
s
(K
4.4.2
,
4.4.2
)
, the set m :

W(m) = 1 witnesses that the S
up
localization
fails. So suppose that p = (w, t
0
, t
1
, t
2
. . .) Q
s
(K
4.4.2
,
4.4.2
) and Y []
.
Since (K
4.4.2
,
4.4.2
) is omittory we may assume that (k) < (k) for k g(w)
and for each i
() nor[t
i
] 11 +i +m
ti
dn
, ) D[t
i
] and
() [(R[t
i
]()), L[t
i+1
]())) Y [ > 2, [(g(w), L[t
0
]) Y [ > 2.
Apply 4.4.6.1 to get s
i

4.4.2
(t
i
) such that nor
0
[s
i
] nor
0
[t
i
] 10 and for every
n there is j (
Y
(n + 1)) such that
L[s
i
]() : T[s
i
] & succ
T[si]
() = [
Y
(n +j),
Y
(n +j + 1)) = .
Note that by () and the denition of
4.4.2
and nor
0
we have
nor
0
[t
i
] = nor[t
i
], nor
0
[s
i
] = nor[s
i
].
Hence, letting q = (w, s
0
, s
1
, s
2
, . . .) we will have p q Q
s
(K
4.4.2
,
4.4.2
) and
for every n > g(w)
q ' (j < (
Y
(n + 1)))([
Y
(n +j),
Y
(n +j + 1)) m :

W(m) = 1 = );
remember that q forces that m :

W(m) = 1 is a subset of
m < m
s0
dn
: w(m) = 1 L[s
i
]() : i < & T[s
i
] & succ
T[si]
() = .
This nishes the proof.
Proposition 4.4.7. Let ,
be non-decreasing, (0), (0) > 0 and

lim
n
(n) lim
n
(n). Suppose that N (H(), , <
) is countable, p N
Q
s
(K
4.4.2
,
4.4.2
), , N and Y []
is such that
(X []
N)(
n )(i < (
Y
(n)))([[
Y
(n +i),
Y
(n +i + 1)) X[ 2).
Then there is an (N, Q
s
(K
4.4.2
,
4.4.2
))generic condition q stronger than p and
such that
q ' for every X []
N[
Q
s
(K
4.4.2
,
4.4.2
)
],
(
n )(i < (
Y
(n)))([[
Y
(n +i),
Y
(n +i + 1)) X[ 2).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Consequently, if is unbounded then the forcing notion Q
s
(K
4.4.2
,
4.4.2
) has
the S
dn
localization property for every non-decreasing . If is bounded, say
lim
n
(n) = k, then Q
s
(K
4.4.2
,
4.4.2
) has the S
k
-localization property.
Proof. This is like [RoSh 501, 2.4.5]. We will deal with the case lim
n
(n) =
(if is bounded then the arguments are similar).
Suppose that , , N, p, Y are as in the assumptions. Let
n
: n < ) enu-
merate all Q
s
(K
4.4.2
,
4.4.2
)names from N for ordinals and let

X
n
: n < ) list
all names for innite subsets of . Further, for n , let
n
N be a name for a
function in
such that
'
Q
s
(K
4.4.2
,
4.4.2
)
(m )(i n)([[m,
n
(m))

X
i
[ > 2).
We inductively construct sequences p
n
, p
n
: n < ), g
n
: n < ) and i
n
: n < )
such that for each n :
(a) p
n
, p
n
Q
s
(K
4.4.2
,
4.4.2
)N, g
n

N is strictly increasing, i
n
,
(b) p
s
0
p
n

s
n
p
n

s
n+1
p
n+1

s
n+1
p
n+1
, i
n
< i
n+1
,
(c) p
n
approximates
n
at each t
p
n+
for ,
(d) for each and v pos(w
p
, t
pn
0
, . . . , t
pn
n+1
),
if t
4.4.2
(t
pn
n+
), nor[t] > 0 then there is w pos(v, t) such that
(w, t
pn
n++1
, t
pn
n++2
, . . .) '
n
(g
n
()) < g
n
( + 1),
(e) (g
n
()) < (L[t
pn
n+
]())) for every ,
(f) ) D[t
pn
m
] for all m ,
(g) i
n
is such that (i < (
Y
(i
n
)))([[
Y
(i
n
+i),
Y
(i
n
+i+1))rng(g
n
)[ 2),
(h) t
p
n
m
= t
pn
m
for m < n,
(i) for i < (
Y
(i
n
)) let j
n
(i) = minj : g
n
(j) [
Y
(i
n
+i),
Y
(i
n
+i +
1)) and for i [(
Y
(i
n
)), (L[t
pn
n+j
n
(0)
]()))) let j
n
(i) = j
n
((
Y
(i
n
))
1) +i;
then t
p
n
n
= S
m0,m1
(S
(t
pn
n+j
n
(i)
: i < (L[t
pn
n+j
n
(0)
]())))) (see clauses (2),
(4) of the denition of (K
4.4.2
,
4.4.2
) in 4.4.2, remember (f ) above), where
m
0
= m
t
pn
n1
up
, m
1
= m
t
pn
k
up
, k = n +j
n
((L[t
pn
n+j
n
(0)
]())) 1),
(j) t
p
n
n+1+m
= t
pn
k+1+m
for every m (where k is as in (i) above).
The construction is quite straightforward and essentially described by the require-
ments (a)(j) above. Having dened p
n
we rst choose a condition p
n+1
N
such that it approximates
n+1
at each m n + 1 and p
n

s
n+1
p
n+1
(by 2.1.4).
Then we use 4.3.3.1 inside N to nd a condition p
n+1

n+1
p
n+1
, p
n+1
N and
a function g
n+1
N such that the demands (a), (d)(f ) are satised. Note here,
that the creatures t
p
constructed in the proof of 4.3.3.1 came from the application

of
1
AB
of 4.3.1. This condition, in turn, is exemplied in our case by the use of
the operation S
H
(see item (5) of the denition of (K
4.4.2
,
4.4.2
), see 4.3.4, 4.3.9).
Consequently we may ensure that (f ) holds. As far as (e) is concerned, note that
in the inductive construction of the condition p in 4.3.3.1, when choosing p
, we
may require additionally that (g()) < (L[t
p
]())) (remember (K
4.4.2
,
4.4.2
) is
omittory, is unbounded). Thus we have dened p
n+1
, g
n+1
satisfying (a)(f ).
By the assumptions on Y we know
(
m )(i < (
Y
(m)))([[
Y
(m +i),
Y
(m +i + 1)) rng(g
n+1
)[ 2)
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

4.4. EXAMPLES 91
(remember rng(g
n+1
) []
N). Hence we may choose i

n+1
> i
n
as required in
(g). Clauses (h)(j), (b) fully describe the condition
p
n+1
= (w
p
, t
pn+1
0
, . . . , t
pn+1
n
, t
p
n+1
n+1
, t
p
n+1
n+2
, . . .)
n+1
p
n+1
.
Note here that
(
Y
(i
n+1
)) (g
n+1
(j
n+1
(0))) < (L[t
pn+1
n+1+j
n+1
(0)
]()))
(by (e); remember is non-decreasing), so there are no problems with the denition
of t
p
n+1
n+1
in clause (i). Moreover, t
p
n+1
n+1

4.4.2
(t
pn+1
n+1
, . . . , t
pn+1
k
), where k = n +1 +
j
n+1
((L[t
pn+1
n+1+j
n+1
(0)
]())) 1, and nor[t
p
n+1
n+1
] > m
t
p
n+1
n+1
dn
. Clearly p
n+1
N.
Now let q = lim
n
p
n
= lim
n
p
n
Q
s
(K
4.4.2
,
4.4.2
) (see 1.2.13). We claim that
q is as required in the assertion of the proposition. Clearly it is stronger than p (by
(b)) and is (N, Q
s
(K
4.4.2
,
4.4.2
))generic (by (c) and 1.2.10). The proof will be
nished if we show that for each k
q ' (
m )(i < (
Y
(m)))([[
Y
(m +i),
Y
(m +i + 1))

X
k
[ 2).
To this end suppose that k, , q
s
(K
4.4.2
,
4.4.2
), q q
. Passing to an ex-
tension of q
we may assume that for some n > k we have w

q
pos(w
q
, t
q
0
, . . . , t
q
n1
)
and < i
n
(remember (b)). Let m n be such that t
q
4.4.2
(t
q
n
, . . . , t
q
m
). Since
nor[t
q
0
] m
t
q
0
dn
> 0 we nd n
0
[n, m] and s
4.4.2
(t
p
n0
) such that nor[s] > 0
and
L[s](): T[s] & succ
T[s]
() = L[t
q
0
](): T[t
q
0
] & succ
T[t
q
0
]
() =
(compare the proof that (K
4.4.2
,
4.4.2
) is condensed). Look at the creature t
q
n0
: it
is t
p
n
0
n0
. So look at the clause (i): the creature t
p
n
0
n0
was obtained from t
pn
0
n0+j
n
0 (i)
(for i < (L[t
pn
0
n0+j
n
0 (0)
]()))) by applying the operation S
and S
m0,m1
. Since
s
4.4.2
(t
q
n0
) we have
[succ
T[s]
())[ = (L[t
pn
0
n0+j
n
0 (0)
]())) = (L[s]()))
and for each succ
T[s]
()) for unique i = i() < (L[t
pn
0
n0+j
n
0 (0)
]())), the tree
T[s] above and L[s], R[s] and D[s] look exactly like T[s
i
], L[s
i
], R[s
i
] and D[s
i
]
for some s
i

4.4.2
(t
pn
0
n0+j
n
0 (i)
) with nor[s
i
] > 0. Now, applying successively clause
(d) to each t
pn
0
n0+j
n
0 (i)
, s
i
(for i < (L[s]()))) we nd w pos(w
q
, t
q
0
) such that
for every succ
T[s]
())
(w, t
pn
0
j
, t
pn
0
j+1
, . . .) '
Q
s
(K
4.4.2
,
4.4.2
)

n0
(g
n0
(j
n0
(i()))) < g
n0
(j
n0
(i()) + 1) ,
where j is such that m
t
pn
0
j
dn
= m
t
q
0
up
. By the denition of the name
n0
and the fact
that k < n n
0
we get that for each succ
T[s]
())
(w, t
pn
0
j
, t
pn
0
j+1
, . . .) ' [[g
n0
(j
n0
(i())), g
n0
(j
n0
(i()) + 1)

X
k
[ 2 .
By the choice of i
n0
, j
n0
(i) we know that
Y
(i
n0
+i) g
n0
(j
n0
(i)) < g
n0
(j
n0
(i) + 1) <
Y
(i
n0
+i + 1)
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

for each i < (
Y
(i
n0
)) (g
n0
(j
n0
(0))) (L[s])()). Therefore
(w, t
pn
0
j
, t
pn
0
j+1
, . . .) ' (i <(
Y
(i
n0
)))([[
Y
(i
n0
+i),
Y
(i
n0
+i + 1))

X
k
[ 2).
We nish noticing that (w, t
pn
0
j
, t
pn
0
j+1
, . . .) (w, t
q
1
, t
q
2
, . . .) and < i
n
i
n0
.
Corollary 4.4.8. The following is consistent with ZFC:
d = cov(/) = non(/) =
2
+
for every non-decreasing unbounded
, d(S
up
) =
2
+
for every non-decreasing
, d(S
dn
) =
1
.
Proof. Start with a model for CH and build a countable support iteration
P
,

Q
: <
2
) and a sequence

: <
2
) such that
(a)

is a P
name for a non-decreasing unbounded function in
,
(b) '
P

Q
= Q
s
(K

4.4.2
,

4.4.2
) ,
(c) for each P
2
name

for a non-decreasing unbounded function in
, for
2
many <
2
, '
P

=

.
By 4.4.3 we have
'
P
2
cov(/) = non(/) =
2
= c & b =
1

and by 4.4.6 we get
'
P
2
d =
2
+ for each non-decreasing unbounded
, d(S
up
) =
2
.
To show that
'
P
2
d(S
dn
) =
1
for every non-decreasing
we use 4.4.7 and [Sh:f, Ch XVIII, 3.6] and we show that the property described in
4.4.7 is preserved in countable support iterations.
So suppose that
is non-decreasing and dene a context (R
, S
, g
)
(see [Sh:f, Ch XVIII, 3.1]) as follows. First, for
let X
[]
be such that
X
(n) =

in
((i) + 1). Next we let (in the ground model V):
S
is the collection of all N H(

1
) for N a countable elementary sub-
model of (H(), , <
),
for each a S
, d[a] = c[a] = = d
[a] = c
[a],

= 1,
R
= R
0
is the relation determined by the S
dn
localization:
R
g if and only if (, g
and)
(
n )(i < (
Xg
(n)))([[
Xg
(n +i),
Xg
(n +i + 1)) X
[ 2)
(remember that
= 1 so we have R
0
only),
g
= g
a
: a S
is such that for every a S
, for each a
we have R
g
a
(exists as each a is countable, e.g. one may take as g
a
any real dominating a).
By the choice of (R
, S
, g
) we know that it covers in V (see [Sh:f, Ch XVIII,

3.2]).
Claim 4.4.8.1. Let P be a proper forcing notion such that
'
P
(R
, S
, g
) covers.
Then:
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

4.4. EXAMPLES 93
(1) '
P
(R
, S
, g
) strongly covers by Possibility B (see [Sh:f, Ch XVIII,

3.3]),
(2) for every P-name

for a non-decreasing unbounded function in
,
'
P
s
(K

4.4.2
,

4.4.2
) is (R
, S
, g
)preserving.
Proof of the claim: 1) As
= 1 it is enough to show that the second player

has an absolute (for extensions by proper forcing notions) winning strategy in the
following game G
a
(for each a S
).
The play lasts moves.
Player I, in his n
th
move chooses functions f
n
0
, . . . , f
n
n

such
that
f
n
`b
n1
= f
n1
`b
n1
for < n and f
n
g
a
for each n.
Player II answers choosing a nite set b
n
, b
n1
b
n
.
At the end the second player wins if and only if for each
_
n
f
n
`b
n
R
g
a
.
But it should be clear that Player II has a (nice) winning strategy in this game. In
his n
th
move he chooses as b
n
a suciently long initial segment of to provide new
witnesses for the quantier (
n ) in the denition of R
(for all f
n
0
, . . . , f
n
n
).
2) Since
= 1 what we have to prove is exactly the statement of 4.4.7 (see [Sh:f,

Ch XVIII, 3.4A]), so we are done with the claim.
Now we may use [Sh:f, Ch XVIII, 3.5, 3.6] (and 4.4.8.1) to conclude that '
P
2
(R
, S
, g
) covers and hence immediately

'
P
2
d(S
dn
) =
1
.
As every function from
V
P
2
appears in
V
P
for some <
2
we nish
the proof.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

CHAPTER 5
Around not adding Cohen reals
The starting point for this chapter was the following request of Bartoszy nski
(see [Ba94, Problem 4]): construct a proper forcing notion P such that:
(1) P is
bounding,
(2) P preserves nonmeager sets,
(3) P makes ground reals to have measure zero,
(4) P has the Laver property,
(5) countable support iterations of P with Laver forcing, random real forcing
and Millers rational perfect set forcing do not add Cohen reals.
A forcing notion with these properties would correspond to the invariant non(^)
(the minimal size of a non-null set; see [BaJu95, 7.3C]). Forcing notions with prop-
erties (1)(4) were known. The fourth property is a kind of technical assumption
and might be replaced by
4
. P is (f, g)bounding for some f, g
(with g(n) f(n), of course).

At least we believe that that was the intension (see an example presented in
[BaJu95, 7.3C], see 5.4.3 here too). However it was not clear how one should
take care of the last required property. The problem comes from the fact that we
do not have any good (meaning: iterable and suciently weak) condition for not
adding Cohen reals. The diculty starts already at the level of compositions of
forcing notions: adding rst a dominating real and then innitely often equal real
below it one produces a Cohen real. Various iterable properties implying no Co-
hen reals are in use, but the point is to nd one capturing as many of them as
possible. The rst section deals with (f, g)bounding property. We generalize this
property in the following section (a special case of the methods developed there is
presented in [BaJu95, 7.2E], however not fully). The (
t,

T)bounding property
seems to be still not weak enough to capture the measure algebra. So we weaken
this further and we present a good candidate for a property responsible for not
adding Cohen reals in the third part of this chapter (see 5.4.2 too). The tools
developed in this section are very general and will be used later too.
5.1. (f, g)-bounding
Let us recall that a proper forcing notion P is (f, g)bounding (for some in-
creasing f, g
) if
'
P
(x
i
f(i))(S V
i
[]
<
)(i )([S(i)[ g(i) & x(i) S(i)).
It is almost obvious that (f, g)bounding forcing notions add neither Cohen reals
nor random reals (see e.g. [BaJu95, 7.2.15]). For the treatment of this property in
countable support iterations see [Sh 326, A2.5] or [Sh:f, Ch VI, 2.11A-C].
94
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

5.1. (f, g)-BOUNDING 95
Definition 5.1.1. Let f
. We say that a weak creating pair (K, ) for

H is essentially f-big if
(
f
5.1.1
) for every weak creature t K and u basis(t) such that 0 < g(u) and
nor[t] > f(0) and each function h : pos(u, t) f(g(u)) there is s (t)
such that u basis(s), h`pos(u, s) is constant and nor[s] nor[t]
f(0)
g(u)
.
Remark 5.1.2. Denition 5.1.1 may be thought of as a kind of strengthening
of 2.2.1 and 2.3.2.
Lemma 5.1.3. Let f
be increasing. Suppose that (K, ) is a nitary

essentially f-big tree-creating pair, p Q
tree
1
(K, ) and is a Q
tree
1
(K, )-name
such that
p '
Q
tree
1
(K,)

is such that (n )( (n) < f(n)).

tree
1
(K, ) stronger than p and such that for every
T
q
the condition q
[]
forces a value to `(g() + 1).
Proof. First note that the essential fbigness of (K, ) implies that if t K,
nor[t] > f(0) and g(root(t)) > f(0) + 1 then the tree creature t is 2-big. This is
more than enough to carry out the proofs of 2.3.6(2) and 2.3.7(2) and thus we nd
a condition q
0
p, g(root(q
0
)) > f(0) + 1, and fronts F
0
, F
1
, F
2
, . . . of T
q0
such
that
(n )( F
n
)(g() > n and q
[]
0
decides (n))
and ( T
q0
)(nor[t
q0
] > 2 f(0) + 1). For each n we have a function

h
n
: F
n
f(n) such that
( F
n
)(q
[]
0
'
Q
tree
1
(K,)
(n) = h
n
()).
Suppose that T
q0
is such that pos(t
q0
) F
n
and g() n (note that there are
T
q0
such that pos(t
q0
) F
n
as F
n
is nite). Then we may apply (
f
5.1.1
) and
we nd s (t
q0
) such that h
n
`pos(s) is constant and nor[s] nor[t
q0
]
f(0)
g()
.
Repeating this process downward and for all n we nd a quasi tree T
T
q0
and s
(t
q0
) for T
such that
() root(T
) = root(q
0
),
() nor[s
] nor[t
q0
] (g() + 1)
f(0)
g()
nor[t
q0
] (f(0) + 1),
() pos(s
) = succ
T
(),
() if ,
0
,
1
T
, n g(), <
0
, <
1
,
0
,
1
F
n
then h
n
(
0
) =
h
n
(
1
).
This denes a condition q
Q
tree
1
(K, ). Clearly it is stronger than q
0
and, by ()
above, it has the required property.
Remark 5.1.4. If (K, ) is a local tree-creating pair (see 1.4.3), p Q
tree
e
(K, )
then
( dcl(T
p
))(root(p) _ T
p
).
Conclusion 5.1.5. Suppose that f
is increasing, (K, ) is essentially

f-big nitary and local tree creating pair. Then the condition q Q
tree
1
(K, )
provided by the assertion of 5.1.3, gives at most [T
q

m<n
H(m)[ possible values
to `(n + 1) (for each n). Hence, if g(n) =

m<n
[H(m)[ then Q
tree
1
(K, ) is (f, g)-
bounding.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

96 5. AROUND NOT ADDING COHEN REALS
Definition 5.1.6. A weak creating pair is reducible if for each t K with
nor[t] > 3 there is s (t) such that
nor[t]
2
nor[s] nor[t] 1.
Definition 5.1.7. Let h : . We say that a weak creating pair
(K, ) is hlimited whenever
if t K, u basis(t), g(u) m
0
and nor[t] m
1
then [pos(u, t)[ h(m
0
, m
1
).
If the function h does not depend on the rst coordinate (i.e. h(m
0
, m
1
) = h
0
(m
1
))
then we say that (K, ) is hnorm-limited. We may say then that (K, ) is h
0
norm-limited or just h
0
limited.
Theorem 5.1.8. Suppose that f, g
are increasing, h : and

(
n )(
m<n
h(m, m) < g(n) < f(n)).
Assume that (K, ) is a reducible nitary tree creating pair which is h-limited and
essentially f-big. Then the forcing notion Q
tree
1
(K, ) is (f, g)-bounding.
Proof. Let N be such that (n N)(

m<n
h(m, m) < g(n) < f(n)). Suppose
that p '
Q
tree
1
(K,)

n<
f(n). By 5.1.3 we nd q p such that for every T
q
the condition q
[]
decides `(g() +1). As (K, ) is reducible we may assume that
( T
q
)(nor[t
q
] g()) and g(root(q)) > N. For n let

F
n
def
= T
q
: g() n and ( T
q
)( < g() < n).
Clearly each F
n
is a front of T
q
and if F
n
then q
[]
decides the value of (n).
Now note that [F
n
[ = 1 for n g(root(q)) and [F
n
[

m<n
h(m, m) < g(n) for all
other n. This allows us to nish the proof.
Theorem 5.1.9. Assume that (K, ) is a nitary and reducible creating pair
which is hlimited for some function h. Further suppose that (K, ) is either grow-
ing and big or omittory and omittorybig. Then the forcing notion Q
s
(K, ) is
(f, g)bounding for any strictly increasing functions f, g
.
Proof. Suppose that is a Q
s

n
f(n) and
p Q
s
(K, ). Applying repeatedly 2.2.3 (or 2.2.6 in the second case) we may
construct inductively an increasing sequence n
0
< n
1
< . . . < and a condition
q = (w
q
, t
q
0
, t
q
1
, . . .) Q
s
(K, ) such that p
s
0
q (so w
q
= w
p
) and for all k :
(
0
) g(n
k
) >
ik
h(m
t
q
i
dn
, 2 m
t
q
i
dn
+ 1),
(
1
) nor[t
q
k
] 2 m
t
q
k
dn
+ 1, and
(
2
) if w pos(w
q
, t
q
0
, . . . , t
q
k1
) then the condition (w, t
q
k
, t
q
k+1
, . . .) decides the
value of `(n
k
+ 1).
This is straightforward; to get (
1
) we use the assumption that (K, ) is reducible.
Now we note that for each k:
[pos(w
q
, t
q
0
, . . . , t
q
k
)[
ik
h(m
t
q
i
dn
, 2 m
t
q
i
dn
+ 1)
and so the condition q allows less than g(n
k
) candidates for values of on the
interval (n
k
, n
k+1
].
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

5.1. (f, g)-BOUNDING 97
Theorem 5.1.10. Let (K, ) be a reducible and nitary creating pair. Suppose
that increasing functions f, g
and a function h : are such that

(1) (K, ) is hlimited,
(2) (K, ) is essentially fbig,
(3) (
n)(

m<n
h(m, m) < g(n) < f(n)).
Lastly assume that (K, ) captures singletons. Then the forcing notion Q
w
(K, )
is (f, g)bounding.
Proof. Take N such that

m<n
h(m, m) < g(n) < f(n) for all n N.
Let be a Q
w

n
f(n), and let p Q
w
(K, ).
First note that, as (K, ) captures singletons, for each t K we may nd s (t)
such that for some u basis(s) (equivalently: for each u, remember (K, ) is
forgetful) we have [pos(u, s)[ = 1. Using this remark and 2.1.12 we nd a condition
(w
p
, s
0
, s
1
, s
2
, . . .) Q
w
(K, ) and a sequence 0
0
<
1
<
2
< . . . < such
that:
() nor[s
0
] 2 f(0)+2, nor[s
i+1
] 2 f(0) [pos(w
p
, s
0
, . . . , s
i
)[ +2(i +1),
() if n
0
,
1
,
2
, . . . then for some u basis(s
n
) we have [pos(u, s
n
)[ =
1,
() N + 4 < m
s
0
dn
, p (w
p
, s
0
, s
1
, s
2
, . . .),
() for each i , u pos(w
p
, s
0
, . . . , s
i
) the condition (u, s
i+1
, s
i+2
, . . .)
decides `(m
s
i
up
+ 1).
Next we slightly correct creatures s
i
to ensure that the value of `(m
s
i
dn
+ 1) is
decided by any u pos(w
p
, s
0
, . . . , s
i1
). For this we use the procedure similar to
that in the proof of 5.1.1 (and based on the assumption that (K, ) is essentially
fbig). Thus we get creatures t
i
(s
i
) such that (for i ):
(u pos(w
p
, s
0
, . . . , s
i1
))
_
(u, t
i
, s
i+1
, s
i+2
, . . .) decides `(m
s
i
dn
+ 1)
_
and
nor[t
i
] nor[s
i
] (m
s
i
dn
+ 1)
f(0)
m
s
i
dn
[pos(w
p
, s
0
, . . . , s
i1
)[.
Note that [pos(w
p
, s
0
, . . . , s
01
)[ = 1 and hence nor[t
0
] 1. Moreover, for
each i , [pos(w
p
, s
0
, . . . , s
i
)[ = [pos(w
p
, s
0
, . . . , s
i
, . . . , s
i+11
)[ and therefore
nor[t
i+1
] 2(i + 1). Finally, as (K, ) is reducible, we may choose t
i
(t
i
)
(for i ) such that
i+2
2
nor[t
i
] m
t
i
dn
. Now we let
w
q
= w
p
, t
p
m
=
_
s
m
if m
0
,
1
, . . .,
t
i
if m =
i
, i .
This denes a condition q Q
w
(K, ) stronger than p and such that for each
i :
(a) (u pos(w
q
, t
q
0
, . . . , t
q
i1
))
_
(u, t
q
i
, t
q
i+1
, . . .) decides `(m
t
q
i
dn
+ 1)
_
,
(b) [pos(w
q
, t
q
0
, . . . , t
q
i+11
)[

ji
h(m
t
q
j
dn
, m
t
q
j
dn
) < g(m
t
q
i
dn
+ 1),
(c) [pos(w
q
, t
q
0
, . . . , t
q
01
)[ = 1.
Now we easily nish.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Definition 5.1.11. Let H be nitary, F
be increasing. A function
f : is called (H, F)fast if it is Hfast (see 1.1.12) and additionally
(n, )(f(n + 1, ) > f(n, ) +F()
H
() ).
Theorem 5.1.12. Suppose that (K, ) is a nitary local and

2big creating
pair for H which has the (weak) Halving Property. Let F
be increasing and
f : be (H, F)fast. Then the forcing notion Q
f
(K, ) is (F,
H
)
bounding.
Proof. Suppose that is a Q
f

m<
F(m) and
p Q
f
(K, ). By 2.2.11 we nd a condition q p which essentially decides all
the values (m) (for m ). We may assume that (i )(nor[t
q
i
] > f(2, m
t
q
i
dn
)).
Applying, in a standard by now way, the bigness (like in 2.2.3 or 5.1.1) we build
a condition r
0
q such that t
r
i
(t
q
i
), nor[t
r
i
] nor[t
q
i
] F(m
t
r
i
dn
)
H
(m
t
r
i
dn
)
m
t
r
i
dn
and for each i and u pos(w
p
, t
r
0
, . . . , t
r
i1
) the condition (u, t
r
i
, t
r
i+1
, . . .)
decides the value of `(g(u) + 1). Now we easily nish (remembering that (K, )
is local).
5.2. (
t,

T)bounding
Here we introduce and deal with a property which, in our context, is a natural
generalization of the notion of (f, g)bounding forcing notions. This is a rst step
toward handling not adding Cohen reals and, in some sense, it will be developed
in the next parts of this chapter. After we formulate and prove some basic results
we show how one may treat this property in countable support iterations.
A particular case of this machinery was presented in [BaJu95, 7.2E].
Definition 5.2.1. Let (K, ) be a creating pair,

t = t
n
: n ) PC(K, ).
(1) For a function h
we dene U
h
(
t) as the set
s = s
n
: n )PC(K, ):

t s & (n)(nor[s
n
] h(m
sn
dn
)).
For n and h
we let
V
n
h
(
t) = s (t
n
) : nor[s] h(m
s
dn
).
(2) Let h
1
, h
2

. We say that a forcing notion P is (
t, h
1
, h
2
)bounding if
'
P
( s U
h1
(
t))( s
U
h2
(
t) V)( s
s).
Remark 5.2.2. (1) We will be interested in the notions introduced in
5.2.1 only for

t PC
(K, ) (i.e. lim

n
nor[t
n
] = ) and (
n)(h
1
(n) <
h
2
(n)), lim
n
h
1
(n) = lim
n
h
2
(n) = .
(2) Note that if (K, ) is nice and simple (see 2.1.7) then
U
h
(
t) =
n
V
n
h
(
t).
Definition 5.2.3. For a creating pair (K, ) on H we say that:
(1) (K, ) is monotonic if for each t K, s (t) we have val[s] val[t].
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

5.2. (
t,

F)BOUNDING 99
(2) (K, ) is strictly monotonic if it is monotonic and for all n , t
0
, . . . , t
n

K and s (t
0
, . . . , t
n
) such that nor[s] maxnor[t
] 1 : n we
have:
(u basis(t
0
))
_
pos(u, s) pos(u, t
0
, . . . , t
n
)
_
.
(3) (K, ) is spread if for each t K, u basis(t) and v pos(u, t) there is
s (t) such that
nor[s]
1
2
nor[t] and v pos(u, s).
Proposition 5.2.4. Let (K, ) H(
1
) be a strictly monotonic and spread
creating pair for H. Suppose that

t = t
n
: n ) PC
(K, ) and h
1
, h
2

are such that

(n)(0 < h
1
(m
tn
dn
) h
2
(m
tn
dn
) nor[t
n
]) and (
n)(h
2
(m
tn
dn
) nor[t
n
] 1).
Then every (
t, h
1
, h
2
)bounding forcing notion does not add Cohen reals.
Proof. Let w basis(t
0
) be such that
(n )
_
pos(w, t
0
, . . . , t
n1
) basis(t
n
)
_
.
Look at the space
A = x
m
H(m) : (n )(x`m
tn
up
pos(w, t
0
, . . . , t
n
))
equipped with the natural (product) topology. It is a perfect Polish space (note that
as (K, ) is strictly monotonic and spread, by lim
n
nor[t
n
] = , for suciently
large n, for each u basis(t
n
) we nd two distinct v
0
, v
1
pos(u, t
n
)). Thus, if
a forcing notion P adds a Cohen real then it adds a Cohen real c A. In V[c],
choose (e.g. inductively) a sequence s = s
n
: n ) U
h1
(
t) such that
(n )
_
s
n
(t
n
) & c`m
sn
dn
pos(w, s
0
, . . . , s
n1
) basis(s
n
)
_
(possible by 5.2.3(3), remember that (K, ) is nice). We claim that there is no
s
U
h2
(
t) V with s
s. Why? Suppose that s
U
h2
(
t) V. Working in V,
consider the set
O
def
= x A : (n )(x`m
s
n
dn
/ pos(w, s
0
, . . . , s
n1
)).
This set is open dense in A (for the density use strict monotonicity of (K, );
remember that for suciently large n , nor[s
n
] h
2
(m
s
n
dn
) nor[t
m
] 1,
where m is such that m
tm
dn
= m
s
n
dn
). Consequently, in V[c], c O and s
, s.
Definition 5.2.5. Let (K, ) be a weak creating pair.
(1) We say that a weak creature t K is (n, m)additive if for all t
0
, . . . , t
n1

(t) such that nor[t
i
] m (for i < n) there is s (t) such that
t
0
, . . . , t
n1
(s) and nor[s] maxnor[t
] : < n + 1.
(2) madditivity of a weak creature t K is dened as
add
m
(t) = supk < : t is (k, m)additive.
[Note that each t is at least (1, m)additive.]
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

(3) We say that (K, ) is (g, h)additive (for g, h
) if add
h(m
dn
(t))
(t)
g(m
dn
(t)) for all t K.
Similarly, if (K, ) is a creating pair and

t PC(K, ) then we say that
t is (g, h)additive if (n )(add

h(m
tn
dn
)
(t
n
) g(m
tn
dn
)).
(4) If the function g is constant, say g n, then instead of (g, h)additive
we may say (n, h)additive etc.
[Note that for creatures we have m
dn
(t) = m
t
dn
.]
Remark 5.2.6. The notion of additivity of a weak creature is very close to that
of bigness: in most applications they coincide. One can easily formulate conditions
under which (k, m)additivity is equivalent to kbigness.
Let us recall that a forcing notion P has the Laver property if it is (f, g
)
bounding for every increasing function f
, where g
(n) = 2
n
(g
may be
replaced by any other xed increasing function in
).
Proposition 5.2.7. Assume that (K, ) is a strongly nitary (see 3.3.4) and
simple (see 2.1.7) creating pair,

t = t
n
: n < ) PC(K, ) and h
1
, h
2

are
such that
(n )(h
1
(m
tn
dn
) h
2
(m
tn
dn
)) and (
n)(h
1
(m
tn
dn
) + 1 h
2
(m
tn
dn
)).
(1) If f
is such that (n )([V

n
h1
(
t)[ f(m
tn
dn
)), g
is strictly in-
creasing and

t is (g, h
1
)additive then every (f, g)bounding forcing notion
is (
t, h
1
, h
2
)bounding.
(2) If

t is (g
, h
1
)additive (where g
(n) = 2
n
) then every forcing notion with
Laver property is (
t, h
1
, h
2
)bounding.
Proof. 1) Suppose that s
n
: n ) is a P-name for an element of U
h1
(
t),
p P. Since (K, ) is simple we know that p ' s
n
(t
n
). Consequently, we may
apply the assumption that P is (f, g)bounding (remember the property of f) and
we get a condition p
0
p and a sequence s
+
n,
: < g
+
(m
tn
dn
), n < ) such that
(n < )( < g(m
tn
dn
))
_
s
+
n,
V
n
h1
(
t)
_
and
p
0
'
P
(n )( s
n
s
+
n,
: < g
+
(m
tn
dn
)),
where g
+

is such that for each n :

g
+
(m
tn
dn
) =
_
g(m
tn
dn
) if h
1
(m
tn
dn
) + 1 h
2
(m
tn
dn
),
1 otherwise.
Since

t is (g, h
1
)-additive we nd s
n
: n ) U
h2
(
t) such that for each n

( < g
+
(m
tn
dn
))(s
+
n,
(s
n
)).
Clearly, p
0
' s
n
: n < ) s
n
: n < ).
2) Similarly.
Definition 5.2.8. Let (K, ) be a creating pair and

t = t
n
: n )
PC
(K, ).
(1) We say that a partial ordering

T = (T, <
F
) on T
is

tgood if:
(a)

T is a dense partial order with no maximal and minimal elements,
(b) for each h T
(n )(1 < h(m
tn
dn
) nor[t
n
]) and lim
n
[nor[t
n
] h(m
tn
dn
)] = ,
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

5.2. (
t,

F)BOUNDING 101
(c) if h
1
, h
2
T, h
1
<
F
h
2
then
(n )(h
1
(m
tn
dn
) h
2
(m
tn
dn
)) and lim
n
[h
2
(m
tn
dn
) h
1
(m
tn
dn
)] = .
(2) Let

T be a

tgood partial order (on T
). We say that a forcing

notion P is (
t,

T)bounding if P is (
t, h
1
, h
2
)bounding for all h
1
, h
2
T
such that h
1
<
F
h
2
.
It should be clear that if

T is

tgood,

t PC
(K, ) then the composition of

(
t,

T)bounding forcing notions is (
t,

T)bounding. To deal with the limit stages
(in countable support iterations) we have to apply the technique of [Sh:f, Ch VI,
1]. Theorem 5.2.9 below fullls the promise of [BaJu95, 7.2.29].
Theorem 5.2.9. Let (K, ) H(
1
) be a simple and reducible creating pair
and let

t = t
n
: n < ) PC
(K, ), nor[t
n
] 2. Suppose that

T = (T, <
F
)
is a

tgood partial order such that

t is (2, h)additive for all h T. Assume that
P
,

Q
: < ) is a countable support iteration of proper forcing notions such that

for each < :
'
P

Q
is (
t,

T)bounding.
Then P
is (
t,

T)bounding.
Proof. We are going to use [Sh:f, Ch VI, 1.13A] and therefore we will closely
follow the notation and terminology of [Sh:f, Ch VI, 1], checking all necessary
assumptions. First we dene a ne covering model (D

F,
t
, R
F,
t
, <
F,
t
) (see [Sh:f,
Ch VI, 1.2]).
For h T and n we x a mapping
h
n
:
onto
V
n
h
(
t). Next, for h T

and
we let
h
() =
h
n
((n)) : n < ). Note that, as (K, ) is nice,
h
() U
h
(
t) (for all
). Further, for h T and for s = s

n
: n < )

t we
let
T
s,h
=
<
: (n < g())(
h
n
((n)) (s
n
)).
Clearly each T
s,h
is a subtree of
<
and any node in T
s,h
has a proper extension
in T
s,h
(remember that (K, ) is reducible). We dene:
D

F,
t
is H(
1
) = H(
1
)
V
(we want to underline here that in the iteration
D

F,
t
is xed and consists of elements of the ground universe),
for x, T D

F,
t
we say that x R
F,
t
T if and only if
x = h
, h) and T = T
s,h
for some h
, h T, h
<
F
h, s U
h
(
t)
D

F,
t
,
for h
, h), h
, h
) dom(R
F,
t
) we say that h
, h) <
F,
t
h
, h
) if and
only if h
= h
<
F
h <
F
h
.
Claim 5.2.9.1. (D

F,
t
, R
F,
t
) is a weak covering model in V (see [Sh:f, Ch VI,
1.1]).
Proof of the claim: The demand (a) of the denition [Sh:f, Ch VI, 1.1] of weak
covering models holds by the way we dened R

F,
t
. The clause (b) there is satised
as for each
and x = h
, h) such that h
, h T, h
<
F
h, we have
() U
h
(
t) U
h
(
t), x R
F,
t
T
(),h
and lim(T
(),h
).
Claim 5.2.9.2. A forcing notion P is (
t,

T)bounding if and only if
it is (D

F,
t
, R
F,
t
)preserving (see [Sh:f, Ch VI, 1.5]).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Proof of the claim: Suppose P is (
t,

T)bounding. Let be a Pname for an
element of
and x = h
, h) dom(R

F,
t
) (so h
, h T and h
<
F
h). Then
'
P

h
( ) U
h
(
t) and, as by 5.2.8(2) P is (
t, h
, h)bounding,
'
P
( s U
h
(
t) V)( s
h
( ))
and hence
'
P
(T D

F,
t
)(x R
F,
t
T & lim(T))
(so (D

F,
t
, R
F,
t
) covers in V
P
).
On the other hand suppose that P is (D

F,
t
, R
F,
t
)preserving. Take h
, h T
such that h
<
F
h and let

s be a Pname for an element of U
h
(
t). Let be a
Pname for an element of
such that '

P

h
( ) =

s. As, in V
P
, (D

F,
t
, R
F,
t
)
still covers (and x = h
, h) dom(R

F,
t
)) we have
'
P
( s D

F,
t
)( s U
h
(
t) & lim(T
s,h
)).
Hence we may conclude that P is (
t, h
, h)bounding.
Claim 5.2.9.3. (D

F,
t
, R
F,
t
, <
F,
t
) is a ne covering model (see [Sh:f, Ch VI,
1.2]).
Proof of the claim: We have to check the requirements of [Sh:f, Ch VI, 1.2(1)].
We will comment on each of them, referring to the enumeration there.
() (D

F,
t
, R
F,
t
) is a weak covering model (by 5.2.9.1).
() <
F,
t
is a partial order on dom(R

F,
t
) = h
, h) T T : h
<
F
h such
that:
(i) there is no minimal element in <
F,
t
,
(ii) <
F,
t
is dense as <
F
is such,
(iii) if x
1
<
F,
t
x
2
(so x
1
= h
, h
1
), x
2
= h
, h
2
) and h
<
F
h
1
<
F
h
2
)
and x
1
R
F,
t
T then there is T
D

F,
t
such that T T
and
x
2
R
F,
t
T namely T itself may serve as T
,
(iv) if x
1
, x
2
dom(R
F,
t
), x
1
<
F,
t
x
2
and x
1
R
F,
t
T
1
, x
1
R
F,
t
T
2
then there is T D

F,
t
such that
x
2
R
F,
t
T, T
1
T, and (n )( T
2
)(`n T
1
T).
For ()(iv) we use the assumption that

t is (2, h)additive for h T. Let x
1
=
h
, h
1
), x
2
= h
, h
2
) (so h
<
F
h
1
<
F
h
2
) and let s
= s
,m
: m < ) U
h1
(
t)
be such that T
= T
s
,h
(for = 1, 2). By 5.2.8(1c) we nd n < such that
(m n)(h
1
(m
tn
dn
) + 1 < h
2
(m
tn
dn
)).
For each m n we choose s
m
(t
m
) such that
s
1,m
, s
2,m
(s
m
) and nor[s
m
] maxnor[s
1,m
], nor[s
2,m
] + 1 < h
2
(m
tn
dn
)
(remember that t
m
is (2, h
1
(m
tm
dn
))additive). For m < n we let s
m
= s
1,m

V
m
h1
(
t) V
m
h2
(
t). Finally let s = s

m
: m < ). As (K, ) is nice, and by the
choice of the s
m
s we have s U
h2
(
t). Look at T
def
= T
s,h
D

F,
t
. By denitions,
x
2
R
F,
t
T, T
1
T (remember that (s
1,m
) (s
m
) for all m ) and if T
2
,
`n T
1
then T (as (s
2,m
) (s
m
) for m n and s
m
= s
1,m
for m < n).
Now comes the main part: conditions () and () of [Sh:f, Ch VI, 1.2(1)]. As
the second one is stronger, we will verify it only. Let us state what we have to
show.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

5.2. (
t,

F)BOUNDING 103
() If V
is a generic extension of V and, in V
, (D

F,
t
, R
F,
t
) is a weak
covering model (i.e. it still covers) then the following two requirements
are satised (in V
).
(a) If x, x
+
, x
n
dom(R
F,
t
) and T
n
D

F,
t
are such that for each n :
x
n
<
F,
t
x
n+1
<
F,
t
x
+
<
F,
t
x and x
n
R
F,
t
T
n
then there are T
D

F,
t
and W []
such that x R
F,
t
T
and

: (i W)(` min(W (i+1))

_
j<i,
jW
T
j
T
0
) lim(T
).
(b) If ,
n

are such that `n =

n
`n for every n and x
dom(R

F,
t
) then
(T D

F,
t
)(
n)
_
x R
F,
t
T &
n
lim(T)
_
.
So suppose that V V
is a generic extension and V
[=(D

F,
t
, R
F,
t
) covers.
We work in V
.
()(a) Let h
, h
n
, h
+
, h T be such that x
n
= h
, h
n
), x = h
, h), x
+
=
h
, h
+
) and h
<
F
h
n
<
F
h
n+1
<
F
h
+
<
F
h. Choose inductively an increasing
sequence 0 < n
0
< n
1
< . . . < such that
(m n
0
)(h
0
(m
tm
dn
) + 1 < h
1
(m
tm
dn
) h
+
(m
tm
dn
)) and
(i )(m n
i+1
)(h
ni
(m
tm
dn
) +i + 2 < h
ni+1
(m
tm
dn
) h
+
(m
tm
dn
))
(possible by 5.2.8(1c)). Let s
n
= s
n,m
: m < ) U
hn
(
t) D

F,
t
be such that
T
n
= T
sn,h
. (Note: each s
n
, h
n
is in V but the sequences s
n
: n < ), h
n
: n < )
do not have to be there.) Using (2, h
+
)additivity of

t we choose s
+
m
(t
m
) such
that for all m we have nor[s
+
m
] h
+
(m
sm
dn
) and
if m < n
1
then s
+
m
= s
0,m
(so s
0,m
(s
+
m
)),
if n
i+1
m < n
i+2
, i < then s
0,m
, s
n0,m
, . . . , s
ni,m
(s
+
m
)
(remember the choice of the n
i
s). As (K, ) is nice we have s
+
= s
+
m
: m < )
U
h
+(
t). Since h
+
<
F
h and V
[=(D

F,
t
, R
F,
t
) covers we nd s U
h
(
t) D

F,
t
such that s s
+
(compare the proof of 5.2.9.2). Look at T
def
= T
s,h
D

F,
t
. By
the denitions, x R
F,
t
T
. Let W
def
= n
i
: i . Suppose that
is such
that
(i )(`n
i+1

_
j<i
T
nj
T
0
).
This means that for each m :
if m < n
1
then `(m+ 1) T
0
= T
s0,h
and so
m
((m)) (s
0,m
) (s
+
m
) (s
m
),
if n
i+1
m < n
i+2
, i then `(m + 1) T
0
T
n0
. . . T
ni
and so
m
((m)) (s
0,m
) (s
n0,m
) . . . (s
ni,m
) (s
+
m
) (s
m
).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Hence (m )(
h
m
((m)) (s
m
)) what implies that lim(T
s,h
) = lim(T
),
nishing the proof of ()(a).
()(b) This is somewhat similar to ()(a). Let h
, h T be such that x = h
, h)
(so h
<
F
h). As

T is a dense partial order we may take h
+
T such that
h
<
F
h
+
<
F
h. Choose 0 < n
0
< n
1
< . . . < such that
(i )(m n
i
)(h
(m
tm
dn
) +i + 2 < h
+
(m
tm
dn
)).
Now take s
+
m
(t
m
) such that for m we have nor[s
+
m
] h
+
(m
sm
dn
) and
if m < n
0
then s
+
m
=
h
m
(
n0
(m)) (so
h
m
(
n0
(m)) (s
+
m
)), and
if n
i
m < n
i+1
, i < then
h
m
(
n0
(m)), . . . ,
h
m
(
ni+1
(m))
(s
+
m
)
(possible as

t is (2, h
+
)additive; remember the choice of the n
i
s). Note that, since
`n =
n
`n for all n , we have that
(i )(m )(
h
m
(
ni
(m)) (s
+
m
)).
Let s
+
= s
+
m
: m < ). Thus s
+
U
h
+(
t) and, as (D

F,
t
, R
F,
t
) covers in V
and
h
+
<
F
h, we nd s = s
m
: m < ) U
h
(
t) D

F,
t
such that s s
+
. Now we have
(i )(m )(
h
m
(
ni
(m)) (s
m
)),
and therefore (i )(
ni
lim(T
s,h
)). As x R
F,
t
T
s,h
, we nish the proof of
the claim.
Now, to nish the proof of the theorem we put together 5.2.9.2, 5.2.9.3 and
[Sh:f, Ch VI, 1.13A].
5.3. Quasi-generic and preserving them
Here we will develop the technique announced in [Sh:f, Ch XVIII, 3.14, 3.15],
putting it in a slightly more general setting, more suitable for our context. We will
get a reasonably weak, but still easily iterable, condition for not adding Cohen reals
this will be used in 5.4.3, 5.4.4. But the general schema presented here will be
applied in the next chapter too (to preserve some ultralters on ).
Definition 5.3.1. Suppose that (K, ) is a creating pair,

t = t
k
: k < )
PC(K, ).
(1) A function W :
T(K) is called a

tsystem for (K, ) if:
(a) if k < and : [m
t
k
dn
, m
t
up
) then W(m
t
k
dn
, m
t
up
, )
(t
k
, . . . , t
), in all other instances W(m

0
, m
1
, ) is empty,
(b) if s (t
k
, . . . , t
), k < then there is n = n

W
(s) [k, ] such
that for each
0
,
1
: [m
t
k
dn
, m
t
up
)
if
0
`[m
tn
dn
, m
tn
up
) =
1
`[m
tn
dn
, m
tn
up
)
then s W(m
s
dn
, m
s
up
,
0
) s W(m
s
dn
, m
s
up
,
1
),
(c) if k
0
< k
1
< . . . < k
i
, s
j
(t
kj
, . . . , t
kj+11
) for j < i, s
(s
0
, . . . , s
i1
) and j
0
< i is such that n
W
(s) [k
j0
, k
j0+1
) (see
(b) above) and : [m
s
dn
, m
s
up
) then
s
j0
W(m
sj
0
dn
, m
sj
0
up
, `[m
sj
0
dn
, m
sj
0
up
)) s W(m
s
dn
, m
s
up
, ),
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

5.3. QUASI-GENERIC AND PRESERVING THEM 105
(d) for some unbounded non-decreasing function G : (1, ) R
0
(called sometimes the weight of W), for every s (t
k
, . . . , t
), k
< , nor[s] > 1, and each : [m
t
k
dn
, m
t
dn
) there is t
W(m
t
k
dn
, m
t
up
, ) such that
t (s) and nor[t] G(nor[s]).
If the function G might be G(x) = x 1 then we call the

tsystem
W regular.
(2) For a norm condition ((nor) we let
P
C(nor)
(
t, (K, ))
def
= s PC
C(nor)
(K, ) :

t s.
It is equipped with the partial order inherited from PC
C(nor)
(K, ).
We introduce another relation _
C(nor)
on P
C(nor)
(
t, (K, )) letting
s
0
_
C(nor)
s
1
if and only if
there is s
2
P
C(nor)
(
t, (K, )) such that s

0
s
2
and the
sequence s
2
is eventually equal to s
1
.
If the norm condition ((nor) is clear we may omit the index to _.
(3) For a

tsystem W (for (K, )) and P
C(nor)
(
t, (K, )) we say that

is quasi-W-generic in P
C(nor)
(
t, (K, )) if
(a) (, _) is directed (i.e. ( s
0
, s
1
)( s )( s
0
_ s & s
1
_ s)) and
countably closed (i.e. if s
n
: n < ) is _increasing then there
is s such that (n )( s
n
_ s)),
(b) for every function
there is s = s
m
: m < ) such that
(
m)(s
m
W(m
sm
dn
, m
sm
up
, `[m
sm
dn
, m
sm
up
))).
Remark 5.3.2. The demand 5.3.1(1d) is to ensure the existence of quasi-W-
generic sets (see 5.3.4(2) below). Conditions 5.3.1(1b) and 5.3.1(1c) are to pre-
serve quasi-W-genericity in countable support iterations. As formulated, they will
be crucial in the proof of 5.3.12.2(2).
Natural applications of the notions introduced in 5.3.1 will be when (K, ) is
simple or simple plus at most omitting. In both cases it will be easy to check
demands 5.3.1(1b,c). In the rst case they are trivial, see 5.3.3 below.
Proposition 5.3.3. Suppose that (K, ) is a creating pair and

t PC(K, ).
(1) If (K, ) is simple then the condition 5.3.1(1b) is empty (so may be omit-
ted) and the condition 5.3.1(1c) is equivalent to
(c)
if t W(m
tn
dn
, m
tn
up
, ), : [m
tn
dn
, m
tn
up
) and s (t)
then s W(m
tn
dn
, m
tn
up
, ).
(2) If W is a

tsystem for (K, ),
and s
= s
,m
: m < )
P
t, (K, )) (for < 2) are such that

s
0
_ s
1
and (
m)(s
0,m
W(m
s0,m
dn
, m
s0,m
up
, `[m
s0,m
dn
, m
s0,m
up
)))
then
(
m)(s
1,m
W(m
s1,m
dn
, m
s1,m
up
, `[m
s1,m
dn
, m
s1,m
up
))).
Proposition 5.3.4. Suppose that (K, ) is a creating pair, ((nor) is one of
the norm conditions introduced in 1.1.10 (i.e. it is one of (s), (), (w) or (f)
for some fast function f) and

t PC
C(nor)
(K, ).
(1) (P
C(nor)
(
t, (K, )), _
C(nor)
) is a countably closed partial ordering.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

(2) Assume CH. Further suppose that if ((nor) (s), (w) then (K, )
is growing. Let W :
T(K) be a

tsystem which is regular
if ((nor) = (f) (for some fast function f). Then there exists a quasi-W-
generic in P
C(nor)
(
t, (K, )).
Proof. We will show this for ((nor) = (). In other instances the proof is
similar and requires very small changes only.
1) It should be clear that _ is a partial order on P
t, (K, )). To show that it is

countably closed suppose that s
n
= s
n,m
: m < ) P
t, (K, )) are such that

s
n
_ s
n+1
for all n . Choose an increasing sequence m
0
< m
1
< . . . < such
that (i )(m )(m
i
= m
si,m
dn
) and for each n i < :
if m < , m
i
m
sn,m
dn
then nor[s
n,m
] i and
if m < , m
i
m
si,m
dn
then (m
< m
< )(s
i,m
(s
n,m
, . . . , s
n,m
1
)).
Now choose s = s
m
: m < ) PC(K, ) such that
if m , m
s0,m
dn
< m
1
then s
m
= s
0,m
,
if m , m
s
k+1
,m
dn
[m
k+1
, m
k+2
) then s
k+1,m
= s
m
for some m
.
Clearly the choice is possible (and uniquely determined) and, by the niceness, we
are sure that s PC(K, ). Moreover, by the choice of m
i
s, we have that s
PC
(K, ) and so s P
t, (K, )). Plainly, s

n
_ s for all n .
2) First note that if s P
t, (K, )),
then there is s
= s
m
: m < )
P
t, (K, )) such that s s
and (
m)(s
m
W(m
s
m
dn
, m
s
m
up
, `[m
s
m
dn
, m
s
m
up
))) (by
5.3.1(1d), remember that the weight of W is unbounded and non-decreasing).
Using this remark, (1) above, and the assumption of CH we may build a _-
increasing sequence s
: <
1
) P
t, (K, )) such that

(
)( <
1
)(
n)(s
,n
W(m
s,n
dn
, m
s,n
up
, `[m
s,n
dn
, m
s,n
up
))).
This sequence gives a quasi-W-generic = s
: <
1
.
Note that proving (2) for ((nor) (s), (w) we have to assume something
about the creating pair (K, ). The assumption that it is growing is the most
natural one (in our context). It allows us to obtain the respective version of the
rst sentence of the proof of (2) for (). Similarly, if ((nor) is (f) then we need
too assume something about the weight of the system W. The assumption that W
is regular is much more than really needed.
Remark 5.3.5. If W
i
are

tsystems (for i
1
) then we may construct in a
similar way (under CH) which is quasi-W
i
-generic for all i <
1
.

t PC
C(nor)
(K, ) (where
((nor) is a norm condition). Suppose that W :
T(K) is a

tsystem
for (K, ), and P
C(nor)
(
t, (K, )) is quasi-W-generic. We say that a proper

forcing notion P is (, W)genericity preserving (or genericity preserving if W is
clear) if '
P
is quasi-W-generic.
Remark 5.3.7. (1) Note that if P is a proper forcing notion and
P
C(nor)
(
t, (K, )) is quasi-W-generic then

'
P
(, _) is directed and countably closed.
Which may fail after the forcing is condition 5.3.1(3b), so the real meaning
of 5.3.6 is that this condition is preserved.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

(2) The composition of genericity preserving forcing notions is clearly
genericity preserving. To handle the limit stages in countable support
iterations we use the main result of [Sh:f, Ch XVIII, 3], see 5.3.12 below.

t PC(K, ). We say that a
tsystem W :
T(K) is
(1) Cohensensitive if for all suciently large m
<
(s (t
m
, . . . , t
m
))( : [m
s
dn
, m
s
up
) )(s / W(m
s
dn
, m
s
up
, )),
(2) directed if for every m
< ,
0
, . . . ,
m
: [m
t
m
dn
, m
t
m
up
)
(m < ) there is : [m
t
m
dn
, m
t
m
up
) such that
W(m
t
m
dn
, m
t
m
up
, )
m
W(m
t
m
dn
, m
t
m
up
,
).
Proposition 5.3.9. Suppose that (K, ) is a creating pair,

t PC
C(nor)
(K, ),
W :
T(K) is a

tsystem and P
C(nor)
(
t, (K, )) is quasi-W-
generic. Let P be a proper forcing notion.
(1) If W is Cohensensitive and P is genericity preserving then P does not
add Cohen reals.
(2) If W is directed, (K, ) is simple and P is
bounding then P is
genericity preserving.
Proof. 1) Note that if
is a Cohen real over V, W is Cohensensitive

and s = s
m
: m < ) PC(K, ) V then
(
m)(s
m
/ W(m
sm
dn
, m
sm
up
, `[m
sm
dn
, m
sm
up
))).
2) Suppose that is a Pname for a real in
, p P. As P is
bounding we
nd a function
and a condition q p such that q '

P
(n < )( (n) < (n)).
Since W is directed we can nd
such that for each m <

W(m
tm
dn
, m
tm
up
,
`[m
tm
dn
, m
tm
up
))
W(m
tm
dn
, m
tm
up
, ) :
m
tm
dn
k<m
tm
up
(k).
Next, as is quasi-W-generic, we nd s = s
m
: m < ) such that
(
m)(s
m
W(m
sm
dn
, m
sm
up
,
`[m
sm
dn
, m
sm
up
))).
We nish noting that (m < )(s
m
(t
m
)), as (K, ) is simple.
Definition 5.3.10. Suppose that (K
) are simple creating pairs and

t
=
t
,m
: m < ) PC(K
) (for < 2) are such that (m < )(m

t0,m
dn
= m
t1,m
dn
).
(1) Let h
0
, h
1

. We say that

t
1
systems W
0
, W
1
:
T(K
1
)
(for (K
1
,
1
)) are (
t
0
, h
0
, h
1
)coherent if there are functions
0
,
1
(called
(
t
0
, h
0
, h
1
)coherence witnesses) such that
(a)
0
,
1
:
0
(t
0,m
) : m < ,
(b) if : [m
t1,m
dn
, m
t1,m
up
) , m < then
()
0
(t
0,m
) and
nor[
()] h
(m
t0,m
dn
) (for = 0, 1),
(c) for suciently large m < , for every
0
: [m
t1,m
dn
, m
t1,m
up
) , there
is
0
: [m
t1,m
dn
, m
t1,m
up
) such that
W
1
(m
t1,m
dn
, m
t1,m
up
,
1
) W
0
(m
t1,m
dn
, m
t1,m
up
,
0
)
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

whenever
1
: [m
t1,m
dn
, m
t1,m
up
) is such that
0
(
0
)
0
(
1
(
1
));
the sequence
0
, as well as
0
(
0
), will be called a
0
cover for
0
,
(d) for each s
0
(t
0,m
), m < , if nor[s] h
1
(m
t0,m
dn
) then there is
: [m
t1,m
dn
, m
t1,m
up
) such that
1
() = s.
(2) Suppose that

T = (T, <
F
) is a

t
0
good partial order (see 5.2.8). We say
that a family W
k
h
: h T, k ) of

t
1
systems for (K
1
,
1
) is (
t
0
,

T)
coherent if for every k and h
0
T there is h
1
T such that h
0
<
F
h
1
and the systems W
k
h0
, W
k+1
h1
are (
t
0
, h
0
, h
1
)coherent.
Theorem 5.3.11. Let (K
0
,
0
), (K
1
,
1
) be simple creating pairs and let
t
0
PC
(K
0
,
0
),

t
1
PC
C(nor)
(K
1
,
1
)
be such that (m < )(m
t0,m
dn
= m
t1,m
dn
). Assume that

T = (T, <
F
) is a

t
0
good
partial order and W
k
h
: h T, k ) is a (
t
0
,

T)coherent family of

t
1
systems for
(K
1
,
1
). Further suppose that P
C(nor)
(
t
1
, (K
1
,
1
)) is quasi-W
k
h
-generic for
all h T, k . Then every (
t
0
,

T)bounding proper forcing notion is (, W
k
h
)
genericity preserving for all h T, k .
Proof. Let P be a proper (
t
0
,

T)bounding forcing notion. We have to show
that for all h T, k
'
P
is quasi-W
k
h
-generic.
For this suppose that is a Pname for a real in
, p P, k and h
0
T.
Take h
1
T such that h
0
<
F
h
1
and the systems W
k
h0
, W
k+1
h1
are (
t
0
, h
0
, h
1
)
coherent and let functions
0
,
1
:
0
(t
0,m
) : m witness this fact.
Let

s = s
m
: m < ) be a Pname for an element of U
h0
(
t
0
) such that for some
N
p '
P
(m N)( s
m
is a
0
cover for `[m
t1,m
dn
, m
t1,m
up
)) (see 5.3.10(1c))
(remember clause 5.3.10(1b)). Now, as P is (
t
0
,

T)bounding and h
0
<
F
h
1
, we
nd a condition q p and s
= s
m
: m < ) U
h1
(
t
0
) such that q '
P
s

s.
Since (K
0
,
0
) is simple this means that q '
P
(m )( s
m

0
(s
m
)). Let

be such that for each m we have

1
(`[m
t1,m
dn
, m
t1,m
up
)) = s
m
(remember
nor[s
m
] h
1
(m
s
m
dn
); see clause 5.3.10(1d)). We know that is quasi-W
k+1
h1
-generic,
so there is s = s
m
: m < ) such that
(
m)(s
m
W
k+1
h1
(m
sm
dn
, m
sm
up
, `[m
sm
dn
, m
sm
up
))).
But (K
1
,
1
) is simple too, so m
sm
dn
= m
s
m
dn
, m
sm
up
= m
s
m
up
. Thus we may apply
5.3.10(1c) and conclude that for suciently large m
q '
P
s
m
W
k+1
h1
(m
sm
dn
, m
sm
up
, `[m
sm
dn
, m
sm
up
)) W
k
h0
(m
sm
dn
, m
sm
up
, `[m
sm
dn
, m
sm
up
)),
so we are done.
Theorem 5.3.12. Suppose that (K, ) H(
1
) is a creating pair and ((nor)
is a norm condition. Assume that
t = t
k
: k < ) PC
C(nor)
(K, ),
W :
T(K) is a

tsystem and
P
C(nor)
(
t, (K, )) is a quasi-W-generic for (K, ).

4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Let P
,

Q
: < ) be a countable support iteration of proper forcing notions such

that for each < :
'
P

Q
is genericity preserving .
Then P
is genericity preserving.
Proof. We will use the preservation theorem [Sh:f, Ch XVIII, 3.6] and there-
fore we will follow the notation and terminology of [Sh:f, Ch XVIII, 3], checking
all necessary details. First we have to dene our context (

R
,W
, S
,W
, g
,W
) (see
[Sh:f, Ch XVIII, 3.1]).
For each m m
t
k
dn
: k we x a mapping
m
:
onto
s K : (k < )(s (t
k
, . . . , t
) & m
t
k
dn
= m).
Next, for
we let () =
m0
((0)),
m1
((1)), . . .), where m
t0
dn
= m
0
<
m
1
< m
2
< . . . < are chosen in such a way that m
m
k
((k))
up
= m
k+1
. Note that
() P
t, (K, )), though it does not have to be in P
C(nor)
(
t, (K, )) (we do not

control the norms).
Now we choose (

R
,W
, S
,W
, g
,W
) such that:
S
,W
is, in the ground model V, the collection of all intersections N
H(
1
), where N is a countable elementary submodel of (H(), , <
); so
S
,W
([H(
1
)
V
]

0
)
V
is stationary (we could replace S
,W
by any
stationary subset),
for each a S
,W
we let d[a] = c[a] = (so d
[a] = c
[a] = ),

= ,
for n <
and , g
we let
R
n
g if and only if
if (g) = s
g
m
: m < ) and m is such that m
s
g
m
up
n then
s
g
m
W(m
s
g
m
dn
, m
s
g
m
up
, `[m
s
g
m
dn
, m
s
g
m
up
)),

R
,W
is a three place relation such that (, n, g)

R
,W
if and only if
, g
, n and R
n
g
(note: this is a denition of a relation, not a xed object from V),
g
,W
= g
a
: a S
,W
)
is such that for every a, a
S
,W
:
() (g
a
) = g
a,m
: m < ) ,
() ( a
)(
m)(g
a,m
W(m
g
a,m
dn
, m
g
a,m
up
, `[m
g
a,m
dn
, m
g
a,m
up
))),
() if a
a S
,W
, s a then (g
a
) _ (g
a
) and s _ (g
a
).
Note that we may choose g
a
by induction for a S
,W
considering all a

a S
,W
, s a and a
. So, before we choose g

a
, we rst take
s
= s
,m
: m < ) for a
such that
(
) (
m)(s
,m
W(m
s,m
dn
, m
s,m
up
, `[m
s,m
dn
, m
s,m
up
)))
(possible by 5.3.1(3b)). Next, as (, _) is directed and countably closed (by
5.3.1(3a)) we may nd s
0
such that
( a
)(a
a S
,W
)( s a )((g
a
) _ s
0
& s _ s
0
& s
_ s
0
).
Let g
a

be such that (g
a
) = s
0
. It is easy to check that g
a
is as required (in
()() above; for () we use 5.3.3(2)).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Claim 5.3.12.1. (1) (

R
,W
, S
,W
, g
,W
) covers in V (see [Sh:f, Ch XVIII,
3.2]), i.e.:
if x V then there is a countable elementary submodel N of
(H(), , <
) such that a
def
= NH(
1
)
V
S
,W
, (

R
,W
, S
,W
, g
,W
),
x N and ( N
)(n < )( R
n
g
a
).
(2) Let P be a proper forcing notion. Then the following conditions are equiv-
alent:
()
1
'
P
(

R
,W
, S
,W
, g
,W
) covers,
()
2
P is -genericity preserving,
()
3
if p P and N is a countable elementary submodel of (H(), , <
)
such that p, P, S
,W
, g
,W
N, a
def
= N H(
1
) S
,W
then there
is an (N, P)generic condition q P stronger than p and such that
q '
P
(
N[
P
])(n < )( R
n
g
a
).
Proof of the claim: 1) By the choice of (

R
,W
, S
,W
, g
,W
) (see condition () of
the choice of g
,W
).
2) Assume ()
1
. Let be a P-name for a real in
, p P. By the assumption
we nd q p, a S
,W
and a P-name

N for an elementary submodel such that
q '
P

N H(
1
)
V
= a &

N & (

N)(n < )( R
n
g
a
).
But, as (g
a
) , this is enough to conclude ()
2
(see the denitions of R
n
,
R
,W
).
Now, suppose that ()
2
holds true. Let N, p be as in the assumptions of ()
3
(so
a
def
= N H(
1
) S
,W
). Let q P be any (N, P)generic condition stronger than
p. Then, by ()
2
, the condition q forces in P that
(
N[
P
])(s
m
: m<)N)(
m)(s
m
W(m
sm
dn
, m
sm
up
, `[m
sm
dn
, m
sm
up
)))
(note that rng(g
,W
) N is a conal subset of ). But now, using 5.3.3(2) and
clause () of the choice of g
,W
, we conclude
q '
P
(
N[
P
])(n < )( R
n
g
a
).
The implication ()
3
()
1
is straightforward.
Claim 5.3.12.2. Suppose that P is a proper forcing notion such that
'
P
(

R
,W
, S
,W
, g
,W
) covers.
Then:
(1) If

Q is a P-name for a proper -genericity preserving forcing notion then
'
P

Q is (

R
,W
, S
,W
, g
,W
)preserving (for Possibility A
)
(see [Sh:f, Ch XVIII, 3.4]).
(2) '
P
(

R
,W
, S
,W
, g
,W
) strongly covers in the sense of Possibility A
(see
[Sh:f, Ch XVIII, 3.3]).
Proof of the claim: 1) We have to show that the following condition holds true
in V
P
:
() Assume
(i)
1
is large enough, > 2
1
,
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

(ii) N is a countable elementary submodel of (H(), , <
), a
def
= N
H(
1
)
V
S
,W
, and

Q, S
,W
, g
,W
,
1
, . . . N,
(iii) (
N)(n < )( R
n
g
a
),
(iv)
0
N is a

Q-name for a real in
,
(v)
,
(vi) p, p
n

Q N are such that p
Q
p
n

Q
p
n+1
for all n ,
(vii)
0
, p
n
: n < ) N,
(viii) (x )(
n)(p
n
'
Q

0
(x) =
0
(x)),
(ix) n
0
< is such that
0
R
n0
g
a
,
(xi) there is a countable elementary submodel N
1
of (H(
1
), , <
1
) such
that

Q, S
,W
, g
,W
N
1
N, and for every open dense subset 1 of
Q, 1 N
1
for some n we have p
n
1 N
1
(i.e. p
n
: n ) is a
generic sequence over N
1
).
Then there is an (N,

Q)generic condition q

Q stronger than p and such
that
(a) q '
Q

0
R
n0
g
a
and
(b) q '
Q
(
N[
Q
])(n < )( R
n
g
a
).
So suppose that
1
, , N, N
1
, a,
0
,
0
, n
0
, p and p
n
(for n ) are as in the assump-
tions of (). Passing to a subsequence (in N) we may assume that p
n
'
Q

0
`n =
0
`n. Remember that we work in V
P
.
So, as (

R
,W
, S
,W
, g
,W
) covers and N (H(), , <
), we nd a countable
elementary submodel N
2
of (H(
1
), , <
1
) such that
Q, S
,W
, g
,W
,
0
,
0
, p
n
: n < ), N
1
, . . . N
2
N, a
2
def
= N
2
H(
1
)
V
S
,W
and (
N
2
)(n < )( R
n
g
a2
).
By the choice of g
,W
we know that (g
a2
) _ (g
a
) (as a
2
a) and hence we nd
m
[n
0
, ) such that
if m < , m
g
a,m
up
m
then for some m
< we have
g
a,m
(g
a2,m
, . . . , g
a2,m
).
Now, working in N, we inductively choose sequences n
: < ), k
: < ),
m
: < ), q
: < ) and
: < ), all from N.

Step = 0.
The n
0
is given already. Let m
0
be the rst such that m
m
g
a
2
,m
0
dn
and let
k
0
= n
0
, q
0
= p
n0
,
0
=
0
`m
g
a
2
,m
0
up
.
Step + 1.
Suppose we have dened n
, k
, m
, q
. We let n
+1
= m
g
a
2
,m
up
and we choose
an (N
2
,

Q)generic condition q
+1
p
n
+1
, integers k
+1
[n
+1
, ) and m
+1

(m
, ), and a nite function

+1
: [n
+1
, m
g
a
2
,m
+1
up
) such that:
() q
+1
'
Q
(
N
2
[
Q
])(n < )( R
n
g
a2
),
() q
+1
'
Q

0
R
k
+1
g
a2
,
() m
+1
is the rst such that k
+1
m
g
a
2
,m
+1
dn
,
() q
+1
'
Q

0
`[n
+1
, m
g
a
2
,m
+1
up
) =
+1
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

(possible as, in V
P
,

Q is a proper -genericity preserving forcing notion, remember
5.3.12.1(2)). Note that all parameters needed for the construction are in N. After
it we have
n
0
= k
0
m
g
a
2
,m
0
dn
< m
g
a
2
,m
0
up
= n
1
k
1
m
g
a
2
,m
1
dn
< m
g
a
2
,m
1
up
= n
2
k
2
. . .
and
0
: [0, n
1
) ,
+1
: [n
+1
, n
+2
) . Let
def
=
0
. . . N
.
By (iii), we nd > 0 such that R
n
g
a
. We claim that q
'
Q

0
R
n0
g
a
. If
not, then we nd a condition q q
and m < such that m

g
a,m
up
n
0
and
q '
Q
g
a,m
/ W(m
g
a,m
dn
, m
g
a,m
up
,
0
`[m
g
a,m
dn
, m
g
a,m
up
)).
Let n = n
W
(g
a,m
) (see 5.3.1(1b)) and let m
< be such that m

g
a
2
,m
dn
m
tn
dn
<
m
tn
up
m
g
a
2
,m
up
. Consider the following three possibilities.
Case 1: m
g
a
2
,m
dn
k
.
By the choice of m
0
and m
(remember > 0, so m
g
a,m
up
> m
g
a
2
,m
dn
k
1
> m
) we
know that for some m
<
g
a,m
(g
a2,m
, . . . , g
a2,m
)
and by the choice of k
we know that
q
'
Q
g
a2,m
W(m
g
a
2
,m
dn
, m
g
a
2
,m
up
,
0
`[m
g
a
2
,m
dn
, m
g
a
2
,m
up
)).
By 5.3.1(1c) we conclude that
q
'
Q
g
a,m
W(m
g
a,m
dn
, m
g
a,m
up
,
0
`[m
g
a,m
dn
, m
g
a,m
up
))
(remember the choice of m
, note m
), a contradiction.
Case 2: m
g
a
2
,m
dn
< n
.
Then, by the choice of n
(remember > 0), we have m

g
a
2
,m
up
n
(and m
).
As q
'
Q

0
`n
0
`n
and
0
R
n0
g
a
we immediately get a contradiction
(remember 5.3.1(1b) and the choice of m
). So we are left with the following

possibility.
Case 3: n
m
g
a
2
,m
dn
< k
(so m
< m
).
Now the choice of , n
and clause () of the choice of q
work: we know that

m
g
a
2
,m
up
m
g
a
2
,m
dn
< m
g
a
2
,m
up
= n
+1
,
q
'
Q

0
`[n
, n
+1
) = `[n
, n
+1
) and R
n
g
a
.
Consequently we get a contradiction like in the previous cases.
Now, choosing an (N,

Q)generic condition q q
such that
q '
Q
(
N[
Q
])(n < )( R
n
g
a
)
(possible by 5.3.12.1(2)) we nish.
2) Work in V
P
. We know that (

R
,W
, S
,W
, g
,W
) covers. Clearly each R
n
(for
n < ) is (a denition of) a closed relation on
. So what is left are the following

two requirements:
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

if a
1
, a
2
S
,W
, a
1
a
2
then for every
we have
(n < )( R
n
g
a1
) (n < )( R
n
g
a2
);
1
if

Q,
0
, p, N, N
1
, G
1
, n
0
are such that (in V
P
):
(a)

Q is a proper forcing notion,
(b) N (H(), , <
) is countable, a
def
= N H(
1
)
V
S
,W
,
(
N)(n < )( R
n
g
a
),
Q, S
,W
, g
,W
,
1
, . . . N, p

Q N,
(c)
0
N is a

Qname for a function in
,
(d)
1
< (
1
large enough), N
1
N, N
1
(H(
1
), , <
1
) is count-
able,

Q, p, S
,W
, g
,W
,
0
, . . . N
1
, and G
1
N is a

Qgeneric lter
over N
1
, p G
1
,
(e)
0
[G
1
] R
n0
g
a
,
then for every y N H(
1
) there are N
2
, G
2
satisfying the parallel of
clause (d) and such that y N
2
and
0
[G
2
] R
n0
g
a
.
Now, concerning , look at our choice of g
a
s: by () we know that
a
1
a
2
(g
a1
) _ (g
a2
).
Thus we may use 5.3.3(2) and the denition of

R
,W
to get .
To show
1
we proceed similarly as in the proof of (1) above. So suppose that

Q,

0
, p, N, N
1
, G
1
, n
0
are as in the assumptions of
1
and y N H(
1
). We work
in the universe V
P
. Choose a
Q
increasing sequence p
n
: n ) N,

Qgeneric
over N
1
, such that p
n
: n G
1
and p
n
0
`n. Let N
2
N
be a countable elementary submodel of (H(
1
), , <
1
) such that N
1
, y, . . . N
2
and then choose a countable N
+
2
(H(
1
), , <
1
) such that N
2
N
+
2
N,
a
2
def
= N
+
2
H(
1
) S
,W
and
( N
+
2

)(n )( R
n
g
a2
)
(remember that (

R
,W
, S
,W
, g
,W
) covers in V
P
). Let m
n
0
be such that
(m )
_
m
g
a,m
up
m
(m
< )(g
a,m
(g
a2,m
, . . . , g
a2,m
))
_
.
Next, working in N, construct inductively sequences n
: < ),
: < ),
G
: < ) (all from N) such that

() G
is

Qgeneric over N
2
, p
n
N
+
2
(so
0
[G
] N
+
2
),
()
0
[G
] R
n
+1
g
a2
, n
+1
> n
+m
and n
+1
= m
g
a
2
,k
dn
for some k ,
()
=
0
[G
]`[n
, n
+1
).
Finally let =
0
[G
1
]`n
0
2
. . .
N and let > 0 be such that

R
n
g
a
. As in (1), one shows now that
0
[G
] R
n0
g
a
.
Claim 5.3.12.3. The forcing notion P
is (

R
,W
, S
,W
, g
,W
)preserving and
hence
'
P
(

R
,W
, S
,W
, g
,W
) covers.
Proof of the claim: Due to 5.3.12.2, we may apply [Sh:f, Ch XVIII, 3.6(1)] to get
the conclusion.
Putting together 5.3.12.3 and 5.3.12.1(2) we nish the proof of the theorem.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

5.4. Examples
Example 5.4.1. Let F
be strictly increasing.
There are increasing functions f
F
= f, g
F
= g
, and (K
F
5.4.1
,
F
5.4.1
) =
(K
5.4.1
,
5.4.1
), (K
,F
5.4.1
,
,F
5.4.1
) = (K
5.4.1
,
5.4.1
),

t
F
=

t,

t
F
=

t
,

T
F
=

T
=
(T
, <
) (for < ) and W

k
,h
: h T
, k, < ) such that for every < :

(1) (K
5.4.1
,
5.4.1
), (K
5.4.1
,
5.4.1
) are simple, strongly nitary and forgetful
creating pairs for H
5.4.1
, H
5.4.1
, respectively;
(2)

t PC
(K
5.4.1
,
5.4.1
),

t
PC
(K
5.4.1
,
5.4.1
);
(3)

T
is a countable

t
good partial order;

(4)

t
is (2
g
, h)additive for each h T
;
(5) W
k
,h
: h T
, k ) is a (
,

T
)coherent sequence of regular

tsystems;
(6) each W
k
,h
(for h T
, k ) is Cohen sensitive;
(7) (h T
)(
m)([V
m
h
(
)[ < f(m));
(8) (m )(g(m+ 1) = F(f(m))).
Moreover, the sequence W
k
,h
: h T
, k, < ) has the following property:

()
5.4.1
if P
(K
5.4.1
,
5.4.1
) is quasi-W
k
,h
-generic for every k, < , h T
and B is a measure algebra (i.e. adding a number of random reals) then

B is (, W
k
,h
)genericity preserving for all k, < , h T
.
Construction. Let F
be a strictly increasing function. We inductively

dene f, g
, n
i
,
i
, k
i
: i < ) and a function : such that:
() g(0) = F(1), n
0
> 2
2g(0)
satises 2
2g(0)+1
<
(n
0
)
n
0
(n
0
)!
,
()
i
is such that n
i
<
i
and
(
i
n
i
)
n
i
n
i
n
i
)!
(
i
n
i
)!
2,
() k
i
= n
i
(
i
)
n
i
+n
i
+ 1,
(
0
) (i, 0) = 2
g(i)(i+1)
(n
i
)
k
i
,
(
1
) (i, + 1) = [2
2g(i)(i+1)+(i,)
((i, )!)]
g(i)+1
(n
i
)
k
i
,
() f(i) = 2
(i,i)
, g(i + 1) = F(f(i)),
() n
i+1
is such that 2
2g(i+1)
k
i
< n
i+1
and
2
2g(i+1)(i+1)
2
+1
<
(n
i+1
)
n
i+1
(n
i+1
)!
.
Why is the choice possible? For clauses (), () remember that lim
n
n
n
n!
= . For
clause () note that
N
n
(N n
i
)!
N!
=
N
n
i
N (N 1) . . . (N (n
i
1))
N
1.
Now, we dene H
5.4.1
, H
5.4.1
. For i we let
H
5.4.1
(i) =
e T([(n
i
, k
i
)]
n
i
) :
e = (n
i
, k
i
) & (u, u
e)(u ,= u
u u
= ),
H
5.4.1
(i) = (j, x) : j < (i, ) & x : (n
i
, k
i
) n
i
.
A creature t CR[H
5.4.1
] is in K
5.4.1
if for some m , B T([(n
m
, k
m
)]
n
m
):
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

5.4. EXAMPLES 115
val[t] = u, w)

i<m
H
5.4.1
(i)

im
H
5.4.1
(i) : u < w & B w(m),
nor[t] =
log
2
1+
k
m
n
m
1
n
m
|B|
log
2
(
m
)
=
log
2
1+(
m
)
n
m|B|
log
2
(
m
)
,
dis[t] = B.
5.4.1
on K
5.4.1
is given by
5.4.1
(t) = s K
5.4.1
: m
s
dn
= m
t
dn
& dis[t] dis[s] for t K
5.4.1
.
Now we dene (K
5.4.1
,
5.4.1
) for < . The family K
5.4.1
consists of all creatures
t CR[H
5.4.1
] such that for some m and a nonempty set C H
5.4.1
(m) such
that ((j, f), (j
) C)(j = j
f = f
) we have:
val[t] = u, w)
i<m
H
5.4.1
(i)
im
H
5.4.1
(i) : u < w & w(m) C,
nor[t] =
log
2
(|C|)
g(m)
,
dis[t] = C.
5.4.1
on K
5.4.1
is such that for t K
5.4.1
:
5.4.1
(t) = s K
5.4.1
: m
s
dn
= m
t
dn
& dis[s] dis[t],
where dis[s] dis[t] means that there is an embedding
i : j < (m, ) : (x)((j, x) dis[s])
11
j < (m, ) : (x)((j, x) dis[t])
such that (j dom(i))(x)((j, x) dis[s] (i(j), x) dis[t]). Later we
may identify elements s
0
, s
1

5.4.1
(t) such that dis[s
0
] dis[s
1
] and dis[s
1
]
dis[s
0
]. Therefore we may think that we have the following inequality:
[
5.4.1
(t)[ 2
dis[t]
.
5.4.1
,
5.4.1
), (K
5.4.1
,
5.4.1
) are strongly nitary, simple
and forgetful creating pairs. Now we have to dene

t,

t
. The rst is the minimal

member of PC(K
5.4.1
,
5.4.1
):
t = t
m
: m < ) is such that (m )(m
tm
dn
= m & dis[t
m
] = ).
Next, for each m (and ), we choose t
,m
(K
5.4.1
,
5.4.1
) such that
m
t
,m
dn
= m and
(x : (n
m
, k
m
) n
m
)
_
[j < (m, ) : (j, x) dis[t
,m
][ =
(m, )
(n
m
)
k
m
n
m
1
_
.
Then we let

t
= t
,m
: m < ). Note that
nor[t
m
] =
log
2
_
1 + (
m
)
n
m
_
log
2
(
m
)
and nor[t
,m
] =
log
2
((m, ))
g(m)

(when m goes to , is xed). Moreover, if n is such that
2
g(m)(n+1)
<
(m, )
(n
m
)
k
m
n
m
1
(where m, < )
then the creature t
,m
is (2
g(m)
, n)additive. Why? Note that if nor[s
i
] n, s
i

5.4.1
(t
,m
) then

i<2
g(m)
[dis[s
i
][ 2
g(m)(n+1)
and it is smaller than
(m,)
(n
m
)
k
m
n
m
1
,
which is the number of repetitions of each function from (n
m
)
(n
m
, k
m
)
in dis[t
,m
].
For each we choose a countable

t
good partial order

T
= (T
, <
) such
that for every h T
:
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

(i) g(m)(h(m) + 1) < log
2
_
(m,)
(n
m
)
k
m
n
m
1
_
for all m ,
(ii) there is h
such that
h <
and (m )(h
(m) h(m) +m),

(iii) there is a function h
T
+1
such that
h
(m) g(m)(m + 1) +(m, ) + log

2
((m, )!) (for all m ).
There should be no problems in carrying the construction of the T
. Note that we
may do this inductively, building a linear order (and so it will be isomorphic to
rationals). The clause (iii) is not an obstacle (in the presence of (i)) as (m, ) is
increasing fast enough:
(n
m
+ 1) log
2
(n
m
) +g(m)(g(m)(m + 1) +(m, ) + log
2
((m, )!) + 2) <
(n
m
+ 1) log
2
(n
m
) + (g(m) + 1)(2g(m)(m+ 1) +(m, ) + log
2
((m, )!)) =
log
2
_
(m,+1)
(n
m
)
k
m
n
m
1
_
.
Note that the clause (i) and the previous remark imply that

t
is (2
g
, h)additive
for each h T
. Moreover, by the choice of the function f we have that for every

m < and h T
[V
m
h
(
)[ < [(t
,m
)[ 2
|dis[t
,m
]|
= 2
(m,)
2
(m,m)
= f(m).
Finally, we are going to dene

tsystems W
k
,h
for k, and h T
. First, for
each , h T
we x a function
h
:
m
V
m
h
(
) such that for m :
h
`
[m, m+ 1)
:
[m, m+ 1)
onto
V
m
h
(
).
Next, for m and : [m, m + 1) (and , k , h T
) we let:
if m k then W
k
,h
(m, m+ 1, ) =
5.4.1
(t
m
),
if m > k then
W
k
,h
(m, m + 1, ) =
_
s
5.4.1
(t
m
) : for some u dis[s] we have
[(j, x) dis[
h
()] : x[u] = n
m
[ <
|dis[
h
()]|
2
g(m)(m+1)(k+1)
_
(in all other instances we let W
k
,h
(m
, m
, ) = ).
Claim 5.4.1.1. For each k, and h T
, the function W
k
,h
is a Cohen
sensitive regular

tsystem.
Proof of the claim: First we have to check that W
k
,h
is a

tsystem. Immediately
by its denition we have that 5.3.1(1a-c) are satised (remember (K
5.4.1
,
5.4.1
) is
simple; see 5.3.3(1)). What might be problematic is 5.3.1(1d). So suppose that
k, , m , m > k (otherwise trivial), h T
, : [m, m+1) , s
5.4.1
(t
m
),
nor[s] > 1. The last means that
k
m
n
m
1
n
m
[dis[s][ >
m
1.
Let N =
m
n
m
. Choose a set X (n
m
, k
m
) such that [X[ = N and X
dis[s] =
. Note that for each (j, x) dis[
h
()] we have
[u [X]
n
m
: x[u] = n
m
[
[[X]
n
m
[
=
[x
1
[0] X[ . . . [x
1
[n
m
1] X[
_
N
n
m
_
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

5.4. EXAMPLES 117
_
_
i<n
m
[x
1
[i] X[
n
m
_
_
n
(n
m
)! (N n
m
)!
N!
= N
n
(N n
m
)!
N!

(n
m
)!
(n
m
)
n
m
.
Now look at clause () of the choice of
m
. It implies that N
n
(Nn
m
)!
N!
2 and
hence
[u [X]
n
m
: x[u] = n
m
[
[[X]
n
m
[
2
(n
m
)!
(n
m
)
n
m
<
1
2
2g(m)m
2

1
2
g(m)(m+1)(k+1)
(the second inequality follows from clause () of the choice of n
m
, remember m > k).
Consequently, applying the Fubini theorem, we nd u
0
[X]
n
m
such that
[(j, x) dis[
h
()] : x[u
0
] = n
m
[
[dis[
h
()][
<
1
2
g(m)(m+1)(k+1)
.
Thus, choosing s

5.4.1
(s) such that dis[s
] = dis[s] u
0
we will have
nor[s
] nor[s] 1 and s
W
k
,h
(m, m + 1, ). This shows that each W
k
,h
is a regular

tsystem.
Finally we show that W
k
,h
is Cohensensitive. Suppose that s
5.4.1
(t
m
),
m > k. Choose a function x
: (n
m
, k
m
) n
m
such that (u dis[s])(x
[u] =
n
m
). Next take : [m, m + 1) such that dis[
h
()] = (j
0
, f
) for some
j
0
< (m, ). It should be clear that s / W
k
,h
(m, m+ 1, ).
Claim 5.4.1.2. For each , the sequence W
k
,h
: h T
, k ) is (
,

T
)
coherent.
Proof of the claim: Let h
0
T
, k . By the demand (ii) of the choice of T
we
nd h
1
T
such that
h
0
<
h
1
and (m )(h
1
(m) h
0
(m) +m).
We want to show that the systems W
k
,h0
and W
k+1
,h1
are (
, h
0
, h
1
)coherent and
that this is witnessed by the functions
h0
,
h1
. Clearly these functions satisfy the
demands (1a), (1b) and (1d) of 5.3.10, so what we have to check is 5.3.10(1c) only.
Suppose that m > k+1,
0
: [m, m+1) . Look at the creature s =
h0
(
0
).
We know that nor[s] h
0
(m) and hence [dis[s][ 2
g(m)h0(m)
. Since 2
g(m)h0(m)
<
(m,)
(n
m
)
k
m
n
m
1
(remember clause (i) of the choice of T
) we nd a creature s
5.4.1
(t
,m
) such that dis[s] dis[s
], nor[s
] = h
0
(m) (i.e. [dis[s
][ = 2
g(m)h0(m)
)
and for each function x
: (n
m
, k
m
) n
m
we have
[(j, x) dis[s] : x = x
[
[dis[s][
2
[(j, x) dis[s
] : x = x
[
[dis[s
][
.
How? We just repeat each (j, x) from dis[s] successively, till we get the required
size. We have enough space for this as the number of the required repetitions for
each function from (n
m
, k
m
) to n
m
is less than 2
g(m)h0(m)
.
Take
0
: [m, m + 1) such that
h0
(
0
) = s
. We want to show that this
0
is a
h0
cover for
0
. So suppose that
1
: [m, m + 1) is such that
h0
(
0
)
5.4.1
(
h1
(
1
)). Let t W
k+1
,h1
(m, m + 1,
1
). This means that we can
nd u dis[t] such that
[(j, x) dis[
h1
(
1
)] : x[u] = n
m
[ <
[dis[
h1
(
1
)][
2
g(m)(m+1)(k+2)

2
g(m)h1(m)
2
g(m)(m+1)(k+2)
.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Hence, remembering that dis[s
] dis[
h1
(
1
)] and the choice of h
1
, we get
[(j, x) dis[s
] : x[u] = n
m
[ <
<
2
g(m)h0(m)
2
g(m)(m+1)(k+1)

2
mg(m)
2
g(m)(m+1)
=
1
2
g(m)

[dis[s
][
2
g(m)(m+1)(k+1)
.
But we are interested in s. By the choice of s
we have
[(j, x) dis[s] : x[u] =n
m
[
[dis[s][
2
[(j, x) dis[s
] : x[u] =n
m
[
[dis[s
][
<
1
2
g(m)(m+1)(k+1)
and therefore t W
k
,h0
(m, m+ 1,
0
). Thus we have proved
W
k+1
,h1
(m, m+ 1,
1
) W
k
,h0
(m, m + 1,
0
)
whenever
h0
(
0
)
5.4.1
(
h1
(
1
)). This nishes the claim.
Claim 5.4.1.3. Suppose that P
(K
5.4.1
,
5.4.1
) is quasi-W
k
,h
-generic for
all k, , h T
. Let B be a measure algebra. Then

'
B
is quasi-W
k
,h
-generic for all k, , h T
.
Proof of the claim: Let be a -additive measure on the complete Boolean algebra
B. Let k, , h T
. Suppose that is a Bname for a real in
, b B
+
(i.e. (b) > 0). To simplify notation let us dene, for m ,
K
m
= [V
m
h
(
)[, M
m
= 2
g(m)(m+1)
K
m
, N
m
= (m, )!
and let h
T
+1
be the function given by the clause (iii) of the choice of

T
+1
.
Fix m for a moment.
For s V
m
h
(
) choose a creature t(s)

+1
5.4.1
(t
+1,m
) such that
[dis[t(s)][ = N
m
,
for each x
: (n
m
, k
m
) n
m
[(j, x) dis[s] : x = x
[
[dis[s][
=
[(j, x) dis[t(s)] : x = x
[
[dis[t(s)][
(possible be the choice of N
m
and the fact that each x
is repeated more than N

m
times in dis[t
+1,m
]). Further, we choose integers g(m, s) < M
m
for s V
m
h
(
)
such that
sV
m
h
(
)
g(m, s) = M
m
and [
_
[[
h
( `[m, m+ 1)) = s]]
B
b
_
(b)

g(m, s)
M
m
[
1
M
m
(where [[]]
B
stands for the Boolean value). Take a creature t
m

+1
5.4.1
(t
+1,m
) such
that for some sequence A
s
: s V
m
h
(
)) of disjoint subsets of (m, +1) we have

[A
s
[ = g(m, s) N
m
,

A
s
: s V
m
h
(
) = j < (m, + 1) : (x)((j, x) dis[t
m
]),
for some bijection
s
: g(m, s) dis[t(s)] A
s
we have
(k < g(m, s))((j, x) dis[t(s)])
_
(
s
(k, (j, x)), x) dis[t
m
]
_
.
Why is the choice of the t
m
possible? Note that our requirements imply that
[dis[t
m
][ = M
m
N
m
= 2
g(m)(m+1)
[V
m
h
(
)[ (m, )! <
(m, + 1)
(n
m
)
k
m
n
m
1
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

5.4. EXAMPLES 119
and the last number says how often each function is repeated in dis[t
+1,m
]. More-
over
nor[t
m
] =
g(m)(m+1)+log
2
(|V
m
h
(
)|)+log
2
((m,)!)
g(m)
<
g(m)(m+1)+(m,)+log
2
((m,)!)
g(m)

h
(m)
g(m)
< h
(m).
Thus t
m
V
m
h
t
+1
).
Let
be such that
+1
h
(`[m, m + 1)) = t
m
. Since is quasi-W
k+1
+1,h
-
generic, we nd s = s
m
: m < ) such that for some m
> k + 4
(m m
)(s
m
W
k+1
+1,h
(m, m+ 1, `[m, m+ 1)).

Fix m m
for a moment. We know that for some u dis[s

m
] we have
[(j, x) dis[t
m
] : x[u] = n
m
[ <
[dis[t
m
][
2
g(m)(m+1)(k+2)
.
For s V
m
h
(
) let
Y
m
(s)
def
=
[(j, x) dis[s] : x[u] = n
m
[
[dis[s][
=
[(j, x) dis[t(s)] : x[u] = n
m
[
[dis[t(s)][
and note that
s
Y
m
(s) g(m, s) N
m
= [(j, x) dis[t
m
] : x[u] = n
m
[. Let
A
m
= s V
m
h
(
) : Y
m
(s)
1
2
g(m)(m+1)(k+1)
(so
h
() / A
m
implies s
m
W
k
,h
(m, m + 1, )). Note that
sXm
g(m, s) N
m
2
g(m)(m+1)(k+1)

sXm
Y
m
(s) g(m, s) N
m

M
m
N
m
2
g(m)(m+1)(k+2)
and therefore

sXm
g(m,s)
Mm
1
2
g(m)(m+1)
. Hence
sXm
_
[[
h
( `[m, m+ 1)) = s]]
B
b
_
(b)

sXm
_
g(m, s)
M
m
+
1
M
m
_
1
2
g(m)m+g(m)1
.
Let b
m
=

B
sXm
[[
h
( `[m, m + 1)) = s]]
B
b. By the above estimations we have
(b
m
)
(b)
2
m+4
(remember m m
> k + 4 4). Look at the condition b
=
b (
B
mm
b
m
). Clearly (b
) > 0 and
b
'
B
(m m
)(
h
( `[m, m+ 1) / A
m
))
and therefore
b
'
B
(m m
)(s
m
W
k
,h
(m, m + 1, `[m, m+ 1))).
This nishes the proof of the claim and thus checking that the construction is as
required.
Conclusion 5.4.2. Let F
be strictly increasing and f, g, (K

5.4.1
,
5.4.1
),
(K
5.4.1
,
5.4.1
),

t,

t
,

T
= (T
, <
) (for < ) and W

k
,h
: h T
, k, < ) be
given by 5.4.1 (for F). Suppose that P
(K
5.4.1
,
5.4.1
) is quasi-W
k
,h
-generic
for every k, < , h T
(exists e.g. under CH, see 5.3.4(2), 5.3.5).

4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

(1) Countable support iterations of proper forcing notions which are (, W
k
,h
)
genericity preserving for all k, , h T
is (, W
k
,h
)genericity pre-
serving (for k, , h T) and hence does not add Cohen reals.
(2) Every (f, g)-bounding proper forcing notion (this includes proper forcing
notions with the Laver property) and random real forcing are (, W
k
,h
)
genericity preserving (for k, , h T
),
(3) Assume CH. Then any countable support iteration of proper forcing no-
tions of one of the following types:
(f, g)bounding, Laver property, random forcing
does not add Cohen reals.
Proof. By 5.4.1, 5.3.9(1), 5.3.11, 5.2.7 and 5.3.12.
Example 5.4.3. We dene a strictly increasing function F
and a creating
pair (K
5.4.3
,
5.4.3
) which: captures singletons, is strongly nitary, reducible, for-
getful, simple, essentially f
F
big and hlimited for some function h :
such that (
n)(

m<n
h(m, m) < g
F
(n) < f
F
(n)), where f
F
, g
F
are given by
5.4.1 for F.
Construction. For N < we dene a nice pre-norm H
N
on T(T(N) N)
by:
H
N
(A) = log
2
(1 + maxk < : (x [N]
k
)(a A)(x a))
(for A T(N) N). Note that if B C T(N) N, H
N
(C) > 1 then:
(1) maxk < : (x [N]
k
)(a C)(x a) 2,
(2) H
N
(B) H
N
(C),
(3) maxH
N
(B), H
N
(C B) H
N
(C) 1.
For 3) above note that if it fails then we nd k
0
such that
1 + maxH
N
(B), H
N
(C B) log
2
(2k
0
) < H
N
(C).
By the rst inequality we nd x
0
, x
1
[N]
k
0
such that
(a B)(x
0
, a) and (a C B)(x
1
, a).
But the second inequality implies that there is a C such that x
0
x
1
a, what
gives a contradiction.
As clearly H
N
(a) = 0 for a T(N) N, H
N
is really a nice prenorm.
Let F
be dened by F(m) = 2
(m+1)
3
2
2
m
(for m ) and let f
F
, g
F
be from
5.4.1 (for F). To simplify notation let M
n
= 2
f
F
(n)
, N
n
= (n+1)
2
2
Mn
(for n ).
Let H : T([]
<
) be given by H(n) = [N
n
]
2
Mn
.
Let K
5.4.3
consist of creatures t CR[H] such that, letting m = m
t
dn
,
(a) dis[t] is a subset of H(m),
(b) val[t] = w, u)
i<m
H(i)
im
H(i) : w < u & u(m) dis[t],
(c) nor[t] =
HNm
(dis[t])
(m+1)f
F
(m)
.
5.4.3
is the trivial one: it gives a nonempty result for
singletons only and then
5.4.3
(t) = s K
5.4.3
: m
s
dn
= m
t
dn
& dis[s] dis[t].
Now, we have to check that (K
5.4.3
,
5.4.3
) has the required properties. Clearly
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

5.4. EXAMPLES 121
it is a strongly nitary, simple and forgetful creating pair. Plainly, it captures
singletons. Note that if t
m
K
5.4.3
(for m ) is such that dis[t
m
] = H(m) then
nor[t
m
] =
H
Nm
(H(m))
(m+ 1) f
F
(m)
=
log
2
(1 + 2
Mm
)
(m + 1) f
F
(m)
>
2
f
F
(m)
(m + 1) f
F
(m)
m
.
Consequently the forcing notions Q
(K
5.4.3
,
5.4.3
), Q
w
(K
5.4.3
,
5.4.3
) etc will be
non-trivial.
To verify that (K
5.4.3
,
5.4.3
) is reducible use the fact that H
N
is a nice pre-norm
on T(T(N) N): if a C T(T(N) N), H
N
(C) > 1 then H
N
(C) 1
H
N
(C a). For similar reasons (K
5.4.3
,
5.4.3
) is essentially f
F
big, remember
that we divide the respective value of H
Nm
by (m + 1) f
F
(m) (where m = m
t
dn
).
Finally, let h(m, k) = 2
Nm
for m, k . Then [H(m)[ =
_
Nm
2
Mm
_
< 2
Nm
for all
m , so (K
5.4.3
,
5.4.3
) is hlimited (actually much more). Moreover, for every
m :
g
F
(m+1) = F(f
F
(m)) > 2
(m+1)
3
2
2
f
F
(m)
= 2
(m+1)
3
2
Mm
>
km
2
N
k
=
km
h(k, k).
Thus (K
5.4.3
,
5.4.3
) is as required.
Conclusion 5.4.4. The forcing notion Q
w
(K
5.4.3
,
5.4.3
):
(1) is proper and
bounding,
(2) preserves non-meager sets,
(3) is (f
F
, g
F
)bounding,
(4) makes the ground model reals have measure zero.
Moreover, assuming CH, countable support iterations of Q
w
(K
5.4.3
,
5.4.3
) with
Lavers forcing notion, Millers forcing notion and random forcing do not add Cohen
reals.
Proof. The rst required property is a consequence of 2.1.12 and 3.1.3. The
second follows from 3.2.6 and the third property is a consequence of 5.1.10. The
moreover part holds true by 5.4.2.
Let A =

m
N
m
be equipped with the product measure. Of course, this space
is measure isomorphic to the reals, so what we have to show is the following claim.
Claim 5.4.4.1. '
Q
w
(K5.4.3,5.4.3)
V A is a null set.
Proof of the claim: Let

W be the Q
w
(K
5.4.3
,
5.4.3
)name for the generic real
(see 1.1.13) and let

A be a Q
w
(K
5.4.3
,
5.4.3
)name for a subset of A such that
'
Q
w
(K5.4.3,5.4.3)

A = x A : (
n)(x(n)

W(n)).
Note that ' (m )(

W(m) [N
m
]
2
Mm
) and
2
Mm
Nm
=
1
(m+1)
2
. Consequently,
'

A is a null subset of A. By the denition of (K
5.4.3
,
5.4.3
) (remember the
denition of H
N
) one easily shows that ' VA

A, what nishes the proof of
the claim and the conclusion.
One may consider tree versions of 5.4.3.
Example 5.4.5. Let F
, h : and H : T([]
<
) be as
dened in 5.4.3. There are nitary treecreating pairs (K
5.4.5
,
5.4.5
) (for < 3)
for H such that
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

(a) (K
0
5.4.5
,
0
5.4.5
) is local, reducible, hlimited and essentially f
F
big (where
f
F
is given by 5.4.1),
(b) (K
1
5.4.5
,
1
5.4.5
) is t-omittory, reducible and essentially f
F
big,
(c) (K
2
5.4.5
,
2
5.4.5
) is reducible, hlimited and essentially f
F
big,
(d) the forcing notions Q
tree
0
(K
0
5.4.5
,
0
5.4.5
), Q
tree
1
(K
1
5.4.5
,
1
5.4.5
) and
Q
tree
1
(K
2
5.4.5
,
2
5.4.5
) are equivalent,
(e) the forcing notion Q
tree
1
(K
0
5.4.5
,
0
5.4.5
) is proper,
-bounding, makes
ground model reals meager and null (even more), is (f
F
, g
+
)bounding,
where g
+
(n) =

i<n
[H(i)[,
(f) the forcing notions Q
tree
1
(K
1
5.4.5
,
1
5.4.5
) and Q
tree
1
(K
2
5.4.5
,
2
5.4.5
) are proper,
bounding, preserve non-meager sets, make ground model reals null

and are (f
F
, g
F
)bounding.
Construction. Let F, N
m
, M
m
, H, h be as in 5.4.3.
A tree creature t TCR
[H] is in K
0
5.4.5
if
dis[t] H(g()),
val[t] = , ) : < & g() = g() + 1 & (g()) dis[t],
nor[t] =
HN
g()
(dis[t])
(g()+1)f
F
(g())
.
The tree composition
0
5.4.5
on K
0
5.4.5
is trivial:
0
5.4.5
(t) = s K
0
5.4.5
: val[s]
val[t].
The family K
1
5.4.5
consists of these treecreatures t TCR[H] that for some _

n<
i<n
H(i) we have
dis[t] H(g(
)),
val[t] = , ) :
< & g() = g(
) + 1 & (g(
)) dis[t],
nor[t] =
HN
g(
)
(dis[t])
(g(
)+1)f
F
(g(
))
.
1
5.4.5
is such that
1
5.4.5
(t
:

T) = t K
1
5.4.5
: dom(val[t])=root(T) & rng(val[t]) max(T).
In a similar manner we dene (K
2
5.4.5
,
2
5.4.5
). A tree creature t TCR
[H] is in
K
2
5.4.5
if
dis[t] = (A
t
,
t
x
: x A
t
)), where A
t
H(g()) and
t
x

n<
i<n
H(i)
(for x A
t
) are such that <
x) _
t
x
,
val[t] = ,
t
x
) : x A
t
,
nor[t] =
HN
g()
(At)
(g()+1)f
F
(g())
,
and the tree composition
2
5.4.5
is dened like
1
5.4.5
.
Checking that (K
5.4.5
,
5.4.5
) have the desired properties is straightforward and
similar to 5.4.3 (remember 5.1.5, 5.1.8, 3.2.2, 3.2.8).
Corollary 5.4.6. Let F
be an increasing function, f, H be as dened in

2.4.6 (for F) and let F
(n) = F(
H
(n)). Then the forcing notion Q
f
(K
2.4.6
,
2.4.6
)
(dened as in 2.4.6 for F) is proper,
bounding, (F
,
H
)bounding and makes
ground model reals meager.
Proof. By 3.1.2, 5.1.12 and 3.2.8(2).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

CHAPTER 6
Playing with ultralters
This chapter originated in the following question of Matet and Pawlikowski.
Are the cardinals m
1
, m
2
equal, where
m
1
is the least cardinality of a set Z

E
E
such that
(i) (F
)(f Z)(n dom(f))(f(n) < F(n)),

(ii) the family dom(f) : f Z has the nite intersection property, and
(iii) (x
n
[]
n
)(f Z)(
n dom(f))(f(n) / x(n));
m
2
is dened in a similar manner, but (iii) is replaced by
(iii)
(g
)(f Z)(
n dom(f))(f(n) ,= g(n))?
It was known that m
2
m
1
, where is the least size of a basis of an ideal on
which is not a weak q-point (see Matet Pawlikowski [MaPa9x]). We answer the
Matet Pawlikowski question in 6.4.6, showing that it is consistent that =
1
(and so m
2
=
1
) and m
1
=
2
. On the way to this result we have to deal with
preserving some special ultralters on . The technology developed in the previous
section is very useful for this (both to describe the required properties and to
preserve them at limit stages of countable support iterations).
In the rst part of the chapter we present the framework: ultralters gener-
ated by quasi-W-generic . Then we introduce several properties of ultralters and
discuss relations between them. The third section shows how forcing notions con-
structed according to our schema may preserve some special ultralters. Finally, in
the last part, we apply all these tools to answer the Matet Pawlikowski question.
6.1. Generating an ultralter
Definition 6.1.1. We say that a creating pair (K, ) generates an ultralter
if
(
6.1.1
) for every k < there is k
< such that

if t K, nor[t] k
and c : [m
t
dn
, m
t
up
) 2,
then for some s (t) and i
< 2 we have nor[s] k and

[u basis(s) & v pos(u, s) & m
t
dn
m < m
t
up
& v(m) ,= 0] c(m) = i
and if nor[t] > 0, u basis(t) then there is v pos(u, t) such that for some
m [m
t
dn
, m
t
up
) we have v(m) ,= 0.
Proposition 6.1.2. Assume that (K, ) is an omittory and monotonic (see
5.2.3) creating pair which generates an ultralter. Then
'
Q
s
(K,)
m :

W(m) ,= 0 induces an ultralter on the algebra T()
V
/Fin.
123
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

124 6. PLAYING WITH ULTRAFILTERS
Proof. For k , let g(k) be the k
given by (
6.1.1
) (for k). Let

Z be a
Q
s
(K, )-name such that
'
Q
s
(K,)

Z = m :

W(m) ,= 0.
Clearly '
Q
s
(K,)

Z []
(by the second requirement of 6.1.1). We have to show

that for each A []
'
Q
s
(K,)
either

Z A or

Z A is nite.
So suppose that p Q
s
(K, ), A []
are such that

p '
Q
s
(K,)
both

Z A and

Z A are innite.
Since (K, ) is omittory we may assume that p = (w
p
, t
0
, t
1
, . . .) where nor[t
]
g( + m
t
dn
) (see 2.1.3.(2)). Let c
: [m
t
dn
, m
t
up
) 2 be the characteristic function
of A [m
t
dn
, m
t
up
). Applying the condition (
6.1.1
) and the choice of g for each
< we nd s
(t
) such that nor[s
] +m
s
dn
and for some i
< 2 we have
(
) if u basis(s
), v pos(u, s
), m
s
dn
m < m
s
up
and v(m) ,= 0
then c
(m) = i
.
Now choose i
< 2 and an increasing sequence

k
: k < ) such that i
k
= i
(for k < ) and take s

0
= s
0
[m
s
0
dn
, m
s
0
up
), s
k+1
= s
k+1
[m
s
k
up
, m
s
k+1
up
). Once
again, since (K, ) is omittory we get q
def
= (w
p
, s
0
, s
1
, s
2
, . . .) Q
s
(K, ) and
it is stronger than p. Suppose i
= 0. We claim that in this case q '

Q
s
(K,)
Z A m
s0
dn
(contradicting the choice of p, A). If not then we nd k < ,
w pos(w
p
, s
0
, . . . , s
k
) and n [m
s0
dn
, m
s
k
up
)A such that w(n) ,= 0. By smoothness
we may additionally demand that, if k > 0 then n [m
s
k
dn
, m
s
k
up
) = [m
s
k1
up
, m
s
k
up
).
By the smoothness and monotonicity of (K, ) we have
w pos(w`m
s
k
dn
, s
k
) = v : w`m
s
k
dn
, v) val[s
k
].
Now, by the choice of s
k
we may conclude that
w`[m
s
k
dn
, m
s
k
dn
) = 0
[m
s
k
dn
,m
s
k
dn
)
and thus n [m
s
k
dn
, m
s
k
up
).
Moreover w`m
s
k
dn
, w) val[s
k
] so we may apply (
) (to
k
) and conclude that
c
k
(n) = i
= 0, a contradiction. Similarly one shows q '

ZA m
s0
dn
if i
= 1.
Definition 6.1.3. Suppose that (K, ) is a creating pair,

t PC
C(nor)
(K, )
and W :
T(K) is a

tsystem (see 5.3.1). Let P
C(nor)
(K, ) be
quasi-W-generic.
(1) We dene T() as the family of all sets A such that: for some
s = s
n
: n < ) and N < , for every w basis(s
0
) and u
pos(w, s
0
, . . . , s
N
) we have
(m > N)(v pos(u, s
N+1
, . . . , s
m
))(k [g(u), g(v)): v(k) ,= 0 A).
(2) We say that generates a lter (an ultralter, respectively) on if T()
is a lter (an ultralter, resp.).
Remark 6.1.4. Note the close relation of 6.1.3 and 6.1.2. Below it becomes
even closer.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

6.1. GENERATING AN ULTRAFILTER 125
Definition 6.1.5. A creating pair (K, ) is interesting if for each creature
t K such that nor[t] > 0 and every u basis(t) we have
[m [m
t
dn
, m
t
up
) : (v pos(u, t))(v(m) ,= 0)[ > 1.
Proposition 6.1.6. Let (K, ) be a forgetful creating pair,

t PC
(K, )
and W :
T(K) be a

tsystem.
(1) If P
t, (K, )) is quasi-W-generic then T() is a lter on contain-

ing all co-nite sets. If, additionally, (K, ) is interesting and condensed
(see 4.3.1(3)) and A is such that
(
n)([A [m
tn
dn
, m
tn
up
)[ 1)
then A / T().
(2) If is quasi-W-generic in P
t, (K, )) then T() is a p-point (see 6.2.1).

(3) Assume CH. Suppose that, additionally, (K, ) is nitary, omittory, mono-
tonic and generates an ultralter. Then there exists a quasi-W-generic
P
(K, ) such that T() is an ultralter on .

Proof. 1), 2) Should be obvious.
3) Modify the proof of 5.3.4(2), noting that if s PC
(K, ), A then there is

s
= s
n
: n < ) PC
(K, ) such that s

s
and either
(ubasis(s
0
))(n<)(vpos(u, s
0
, . . . , s
n
))(k[g(u), g(v)): v(k),=0 A)
or a similar requirement with A instead of A holds. (Compare the proof of 6.1.2,
remember (K, ) is forgetful.)
Conclusion 6.1.7. Suppose that (K, ) is a forgetful and monotonic creating
pair,

t PC
(K, ) and W is a

tsystem. Assume that P
t, (K, )) is quasi-
W-generic and T() is an ultralter. Let be a limit ordinal and P
,

Q
: < )
be a countable support iteration of proper forcing notions such that for each < :
(a) '
P
is quasi-W-generic,
(b) '
P
T() is an ultralter.
Then
(1) '
P
is quasi-W-generic,
(2) '
P
T() is an ultralter.
Proof. 1) It follows from 5.3.12.
2) Since, by 6.1.6(2), T() is a p-point we may use [Sh:f, Ch VI, 5.2] (another
presentation of this result might be found in [BaJu95, 6.2]).
Remark 6.1.8. (1) If

t = t
n
: n < ) PC
(K, ), W is a

tsystem
and is quasi-W-generic generating a lter then we make think of T()
as a lter on

i
i (m
ti
up
m
ti
dn
) (just putting the intervals [m
ti
dn
, m
ti
up
)
vertically). This will be our approach in the further part, where we will
consider ultralters on

i
i (i + 1).
(2) We may treat T() as a canonical lter on with a property described by
(or, more accurately, by the

tsystem W). This is the way we are going
to use 6.1.7 later: it will allow us to claim that the additional property of
an ultralter is preserved at limit stages of an iteration.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

6.2. Between Ramsey and p-points
Here we recall some denitions of special properties of ultralters on and we
introduce more of them. Then we comment on relations between these notions.
Definition 6.2.1. Let T be a lter on . We say that:
(1) T is Ramsey if for each colouring F : []
2
2 there is a set A T
homogeneous for F.
(2) T is a p-point if for every partition A
n
: n ) of into sets from the
dual ideal (i.e. A
n
T) we nd a set A T with
(n )([A
n
A[ < ).
(3) T is a q-point if for every partition A
n
: n ) of into nite sets there
is a set A T with
(n )([A
n
A[ 1).
(4) T is a weak q-point if for each set B such that B / T and a
partition A
n
: n ) of B into nite sets there is a set A B such that
A / T and (n )([A
n
A[ 1).
Remark 6.2.2. Clearly, if T is an ultralter on which is a weak q-point then
T is a q-point. (So the two notions coincide for ultralters).
Definition 6.2.3. (1) For a lter T on let G
R
(T) be the game of two
players, I and II, in which Player I in his n
th
move plays a set A
n
T
and Player II answers choosing a point k
n
A
n
. Thus a result of a play
is a pair of sequences A
n
: n ), k
n
: n )) such that k
n
A
n
T.
Player I wins the play of the game G
R
(T) if and only if
the result A
n
: n ), k
n
: n )) satises: k
n
: n / T.
(2) Similarly we dene the game G
p
(T) allowing the second player to play
nite sets a
n
A
n
(instead of points k
n
A
n
).
Player I wins if

n
a
n
/ T
Remark 6.2.4. Let us recall that if T is an ultralter on then the following
conditions are equivalent (see [Sh:f, Ch VI, 5.6] or [BaJu95, 4.5]):
(a) T is Ramsey,
(b) T is both a p-point and a q-point,
(c) Player I does not have a winning strategy in the game G
R
(T).
Similarly, an ultralter T is a p-point if and only if Player I does not have a winning
strategy in G
p
(T).
As we are interested in ultralters which are not q-points (see the discussion of
the Matet Pawlikowski problem at the beginning of this chapter) it is natural
to x a partition of which witnesses this. Thus, after renaming, we may consider
ultralters on

i
i(i +1) instead (compare with the last remark of the previous
section).
Definition 6.2.5. Let T be a lter on

i<
i (i + 1).
(1) We say that the lter T is interesting if for each function h

i
(i + 1)
the set (i, h(i)) : i is not in T.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

6.2. BETWEEN RAMSEY AND P-POINTS 127
(2) Let G
sR
(T) be the game of two players in which Player I in his n
th
move
plays a set A
n
T and the second player answers choosing an integer i
n
and a set a
n
[A
n
i
n
(i
n
+ 1)]
n
.
Finally, Player I wins the play if

n
a
n
/ T.
(3) The game G
aR
(T) is a modication of G
sR
(T) such that now, the rst
Player (in his n
th
move) chooses a set A
n
T, L
n
< and a function
f
n
:
_
i<
[i (i + 1)]
n
L
n
.
Player II answers playing i
n
and a set a
n
[A
n
i
n
(i
n
+ 1)]
n
homogeneous for f
n
(i.e. such that f
n
`[a
n
]
k
is constant for k n).
Player I wins the play if

n
a
n
/ T.
(4) We say that the lter T is semiRamsey if the rst player has no winning
strategy in the game G
sR
(T).
(5) The lter T is almost Ramsey if it is semiRamsey and for every colouring
f :
_
i<
[i (i + 1)]
n
L, (n, L < )
there is a set A T which is almost homogeneous for f in the following
sense:
(i )(f`[A (i (i + 1))]
k
is constant for each k n).
Proposition 6.2.6. Suppose T is a non-principal ultralter on

i<
i(i+1).
(1) If T is interesting then it is not a q-point.
(2) If T is Ramsey then T is almost Ramsey.
(3) If T is semiRamsey then it is a p-point.
(4) If T is semiRamsey then
T
def
= A :
_
iA
i (i + 1) T
is a Ramsey ultralter.
Proof. Compare the games and denitions.
Theorem 6.2.7. Assume CH. There exists an ultralter T on

i<
i (i +1)
which is semiRamsey but not almost Ramsey.
Proof. For i let k
i
be the integer part of the square root of i +1. Choose
partitions e
m
i
: m < k
i
) of i (i + 1) such that (m < k
i
)([e
m
i
[ k
i
). Let
f :

i
[i (i + 1)]
2
2 be such that
f((i,
0
), (i,
1
)) = 1 if and only if (m < k
i
)((i,
0
) e
m
i
(i,
1
) e
m
i
).
Assuming CH, we will construct a semiRamsey ultralter containing no almost
homogeneous set for f. To this end we choose an enumeration
: <
1
of all
functions from to [
i
i (i + 1)]
. By induction on <
1
dene sequences
i
n
: n < ) and a
n
: n < ) such that for < <
1
:
(a) i
0
< i
1
< i
2
< . . . < ,
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

(b) a
n
[i
n
(i
n
+ 1)]
n
(for n ),
(c) either (m )(n )(a
(m) = ) or (n )(a
n+1

(i
n
)),
(d) lim
n
[m < k
i
n
: e
m
i
n
a
n
,= [ = ,
lim
n
min[a
n
e
m
i
n
[ : m < k
i
n
& a
n
e
m
i
n
,= = ,
(e) [

n
a

n
a
n
[ < .
There should be no problems with carrying out the construction. Let A
=

n
a
n
.
Clearly the sequence A
: <
1
) generates a non-principal ultralter T on
i
i(i +1) (remember that the constant functions are among the
s). By the
demand (d), no set A
is almost homogeneous for f, so T is not almost Ramsey.

To show that T is semiRamsey suppose that is a winning strategy for the
rst player in the game G
sR
(T). Then is a function dened on nite sequences
x = (i
0
, a
0
), . . . , (i
n1
, a
n1
)) such that i
0
< . . . < i
n1
and
( < n)(a
[i
(i
+ 1)]

)
and with values in T. For j < put
(j) =
_
((i
0
, a
0
), . . . , (i
n1
, a
n1
)) : i
0
< . . . < i
n1
j and
a
[i
(i
+ 1)]

_
.
Thus : T, so for some <
1
we have =
. But now look at the

sequence a = (i
0
, a
0
), (i
1
, a
1
), (i
2
, a
2
), . . .). Since A
T and
(m) T for all

m we necessarily have
(n )(a
n+1

(i
n
) ((i
0
, a
0
), . . . , (i
n
, a
n
))).
This means that the sequence a is a result of a legal play of the second player
against the strategy . Hence A
=

n
a
n
/ T, a contradiction.
Theorem 6.2.8. Suppose that T is a semiRamsey ultralter on

i<
i(i+1).
Then the following conditions are equivalent.
(a) T is almost Ramsey,
(b) the rst player has no winning strategy in the game G
aR
(T),
(c) for each m, L and a colouring f :

i<
[i (i + 1)]
m
L, the
rst player has no winning strategy in the following modication G
sR
f
(T)
of the game G
sR
(T): rules are like in G
sR
(T) but the sets a
n
chosen by
the second player have to be homogeneous for f.
Proof. The implications (b) (c) (a) are immediate by the denitions.
The implication (a) (b) is easy too: suppose that is a strategy for the rst
player in the game G
aR
(T). Let
be a strategy for Player I in G

sR
(T) such that
if ((i
0
, a
0
), . . . , (i
n1
, a
n1
)) = (f
n
, A
n
) then
((i
0
, a
0
), . . . , (i
n1
, a
n1
)) T is
an almost f
n
homogeneous subset of A
n
(exists by the assumption (a)). Now,
cannot be the winning strategy for Player I as T is semiRamsey. But then the
play witnessing this shows that is not winning in G
aR
(T).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

6.3. PRESERVING ULTRAFILTERS 129
Theorem 6.2.9. Assume that T is a semiRamsey ultralter on

i<
i(i+1).
Suppose that P is a proper
bounding forcing notion such that

'
P
T generates an ultralter .
Then
'
P
T generates a semiRamsey ultralter .
Proof. This is very similar to [Sh:f, Ch VI, 5.1]. We know that, in V
P
,
T generates an ultralter. What we have to show is that Player I has no winning
strategy in the game G
sR
(T) (in V
P
). So suppose that V
P
is a winning strategy
of the rst player in G
sR
(T). We may assume that the values of are elements
of T (so from the ground model). But now, as P is proper and
bounding, we
nd a function
+
V such that dom(
+
) = dom(), rng(
+
) [T]
<
and
( a)
+
( a) for all a dom(). Letting
( a) =
+
( a) for a dom() we will
get, in V, a winning strategy for Player I in G
sR
(T), a contradiction.
Remark 6.2.10. One may note that we did not mention anything about the
existence of almost Ramsey ultralters. Of course it is done like 6.2.7, under CH.
However we want to have an explicit representation of the ultralter as T() for
some quasi-generic . This will give us the preservation of the colouring part of
the denition of almost Ramsey ultralters at limit stages. As the representation
is very specic we postpone it for a moment and we will present this in Examples
(see 6.4.1, 6.4.2).
6.3. Preserving ultralters
In this section we show when forcing notions of the type Q
tree
1
preserve ultra-
lters introduced in the previous part. The key property of a treecreating pair
needed for this is formulated in the following denition.
Definition 6.3.1. Let T be a lter on . We say that a tree creating pair
(K, ) is of the UP(T)
tree
type if the following condition is satised:
()
tree
UP(D)
Assume that 1 m < , p Q
tree
(K, ), nor[t
p
] > m + 1 for each

T
p
, and F
0
, F
1
, . . . are fronts of T
p
such that
(n )( F
n+1
)( F
n
)( < ).
Further suppose that u
n
F
n
(for n ) are sets such that there is no
system s
:

T) K with:
(a) T T
p
: ( F
n
)( _ ) is a (well founded) quasi tree with
max(T) u
n
and root(T) = root(T
p
),
(b) for each

T:
root(s
) = , pos(s
) = succ
T
(), nor[s
] m and
s
(t
p
:

S
) for some (well founded) quasi tree S
T
p
.
tree
(K, ) such that

() p q,
() the set
Z
def
= n : u
n
dcl(T
q
) =
is not in the ideal T
c
dual to T,
() ( T
q
)(nor[t
q
] minnor[t
p
] m: _ T
p
).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

If we may additionally demand that the condition q Q
tree
(K, ) above satises

() root(q) = root(p) and (n Z)(F
n
T
q
is a front of T
q
)
then we say that (K, ) is of the sUP(T)
tree
type.
If T is the lter of all co-nite subsets of then we say that (K, ) is of the UP
tree
type (sUP
tree
type, respectively) instead of UP(T)
tree
type (sUP(T)
tree
type,
resp.).
Theorem 6.3.2. Let T be a Ramsey ultralter on . Suppose that (K, ) is a
nitary 2big treecreating pair of the UP(T)
tree
type. Then:
'
Q
tree
1
(K,)
T generates an ultralter on .
Consequently, '
Q
tree
1
(K,)
T generates a Ramsey ultralter on .
Proof. Let

X be a Q
tree
1
(K, )name for a subset of , p
0
Q
tree
1
(K, ).
Consider the following strategy for Player I in the game G
R
(T):
in the n
th
move he chooses a condition p
n+1
Q
tree
1
(K, ) such
that p
n

1
n
p
n+1
and p
n+1
' k
n

X (where k
n
is the last point
played so far by Player II). Then he plays the set
B(p
n+1
, n + 1)
def
= k : (q Q
tree
1
(K, ))(p
n+1

1
n+1
q & q ' k

X).
As T is Ramsey, this strategy cannot be the winning one for Player I and therefore
there is a play (determined by k
0
, k
1
, k
2
, . . .) according to this strategy in which
Player I looses. This means that one of the following two possibilities holds:
Case A In the course of the play all sets B(p
n+1
, n + 1) are in the ultralter T
and k
0
, k
1
, . . . T.
In this situation we look at the sequence p
n
: n ). By 1.3.11 it has the limit
p
= lim
n
p
n
. Clearly p
' k
0
, k
1
, . . .

X and we are done.
Case B In the course of the play it occurs that for some n the set B(p
n+1
, n+
1) is not in the ultralter.
Take q Q
tree
1
(K, ), p
n+1

1
n+1
q and fronts F
, F
k
of T
q
such that the condition
q
[]
decides the truth value of k

X for each F
k
and k and
( F
)( T
q
)( _ nor[t
q
] > n + 3),
(k )( F
k
)( T
q
)( _ nor[t
q
] > 2
k
+n + 3)
(possible by 2.3.7(2) and 2.3.11). Of course we may assume that
(k )( F
k+1
)( F
k
)( < )
and that the fronts F
k
are above F
and F
is above F
1
n+1
(q). Further let
u
k
= F
k
: q
[]
'
Q
tree
1
(K,)
k

X.
Look at the set C = B(p
n+1
, n + 1) T. If k C then necessarily for some
F
there is no s
:

T) K with:
(1) T T
q
[]
: ( F
k
)( _ ) is a (well founded) quasi tree with
max(T) u
k
and root(T) =
(2) for each

T:
root(s
) = , pos(s
) = succ
T
(), nor[s
] n + 2 and
s
(t
q
: S
) for some (well founded) quasi tree S
T
q
.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

(Otherwise we could build a condition
1
n+1
stronger than q (and thus than p
n+1
)
and forcing that k

X, contradicting k / B(p
n+1
, n +1).) As F
is nite we have
one F
for which the above holds for all k C

0
for some C
0
T, C
0
C.
But now we may apply the fact that (K, ) is of the UP(T)
tree
type (see
6.3.1, remember the choice of F
): we get q
Q
tree
(K, ) such that q

[]
q
,
Z
def
= k C
0
: u
k
dcl(T
q
) = T and
nor[t
q
] minnor[t
q
] (n + 2) : _ T
q
.
Now we easily see that in fact q
Q
tree
1
(K, ) (by the norm requirement and the
choice of q and F
k
s) and q
'
Q
tree
1
(K,)
Z

X = .
For the consequently part, note that, by 3.1.1, the forcing notion Q
tree
1
(K, )
is proper and
bounding. Therefore we may apply [Sh:f, Ch VI, 5.1]. This

nishes the proof of the theorem.
Definition 6.3.3. We say that a tree creating pair (K, ) for H is rich if:
for every system s
:

T) K, n and u such that
(1) T
m<k
H(m) is a well founded quasi tree, u max(T),
(2) root(s
) = , pos(s
) = succ
T
(), nor[s
] > n + 3,
(3) there is no s
:

T
) K such that
T
T, max(T
) u, root(T
) = root(T), root(s
) = ,
pos(s
) = succ
T
(), nor[s
] > n + 1, and
s
(s
:

T
) for some T
T
there is s
+
:

T
+
) K such that
T
+
T, max(T
+
) max(T) u, root(T
+
) = root(T), root(s
+
) = ,
pos(s
+
) = succ
T
+(), nor[s
+
] minnor[s
] : T (n + 2),
and s
+
(s
:

T
) for some T
T.
Theorem 6.3.4. Assume (K, ) is a nitary 2-big rich treecreating pair of the
sUP
tree
type. Let T be an almost Ramsey interesting ultralter on

i
i(i +1).
Then
'
Q
tree
1
(K,)
T generates an interesting ultralter on

i
i (i + 1) .
Proof. First note that if we show that, in V
Q
tree
1
(K,)
, T generates an ultra-
lter then the ultralter has to be interesting (remember that T is interesting).
Let

X be a Q
tree
1
(K, )name for a subset of

i
i (i + 1).
We say that a condition p Q
tree
1
(K, ) is (

X, n)special if:
there is a set C T such that:
for every i and a [C i (i +1)]
n
there are a condition
p
p and a front F of T
p
such that
root(p) = root(p
), ( T
p
)(nor[t
p
] > n + 1),
( T
p
)(F)( _ t
p
= t
p
) and p
'
Q
tree
1
(K,)
a

X.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

The condition p is nspecial if it is either (

X, n)special or (

X, n)special.
Note that the part of the denition of special conditions concerning the exis-
tence of a front F (of T
p
) is purely technical and usually easy to get (once we have

the rest):
If the condition p is such that ( T
p
)(nor[t
p
] > n+1) and the

values of

X i (i + 1) are decided on some fronts F
i
of T
p
then if we have a condition p

0
p such that
( T
p
)(nor[t
p
] > n + 1)
and p
' a

X, then we may nd one (weaker than p
) which has
this property and a front F as there. Moreover, in this situation,
if p
0
p
1
and p
1
is (

X, n)special then p is (

X, n)special.
Note that if n m and p is (

X, n)special then it is (

X, m) special.
Claim 6.3.4.1. Let n < . Suppose that p Q
tree
1
(K, ) is such that
( T
p
)(nor[t
p
] > (2
2n
+ 1)(n + 3))
and there are fronts F
i
of T
p
(for i ) with
(i )( F
i
)(p
[]
decides

X i (i + 1)).
Then p is n-special.
Proof of the claim: Let f
+
, f
:

i
[i (i + 1)]
n
2 be such that
f
+
(v) = 1 if and only if there are q
0
p and a front F of T
q
such that q '
Q
tree
1
(K,)
v

X and
( T
q
)(F)( _ t
q
= t
p
) & ( T
q
)(nor[t
q
] > n + 1),
and f
is dened similarly replacing

X by

X.
As T is almost Ramsey and interesting we nd j
+
, j
< 2 and a set C T

such that for each i :
if C i (i + 1) ,=
then [C i (i + 1)[ 2n, f
+
`[C i (i + 1)]
n
= j
+
and
f
`[C i (i + 1)]
n
= j
.
If either j
+
= 1 or j
= 1 then plainly p is n-special. So suppose that j

+
= j
= 0
(and we want to get a contradiction).
Take i such that [C i (i + 1)[ 2n (remember the choice of C) and
x v [C i (i + 1)]
2n
. For each v
1
v let
u
i
v1,v
= F
i
: p
[]
'
Q
tree
1
(K,)
v

X = v
1
.
Since (K, ) is rich we nd v
1
and s
+
:

T
+
) K such that
T
+
T
p
, max(T
+
) u
i
v1,v
, root(T
+
) = root(p), pos(s
+
) = succ
T
+(),
root(s
+
) = , nor[s
+
] > n + 1 and s
+
(s
:

T
) for some T
T
p
.
[How? We try successively each v
1
v. If we fail with one, we use 6.3.3 to pass to
a subtree with the minimum of norms dropping down by at most n +2 and we try
next candidate. For some v
1
v we have to succeed.]
Now look at this v
1
(and suitable s
+
:

T
+
)). Since j
+
= 0 we necessarily have
[v
1
[ < n: if not then we may take v
2
[v
1
]
n
and then f
+
(v
2
) = 1 as witnessed
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

by the condition q starting with s
+
:

T
+
). Similarly, by j
= 0, we have
[v v
1
[ < n. Together contradiction to [v[ = 2n. Thus the claim is proved.
Now, let p Q
tree
1
(K, ). By 2.3.12 and 2.3.7(2) we nd p
1
p and fronts F
n
of T
p1
such that for n :
(1) ( F
n+1
)( F
n
)( < ),
(2) ( F
n
)( T
p1
)( _ nor[t
p1
] > (2
2n+1
+ 5)(n + 3)),
(3) ( F
n
)(p
[]
1
decides

X n (n + 1)).
Then, by 6.3.4.1, for each F
n
the condition p
[]
1
is nspecial. Let
u
n
= F
n
: p
[]
1
is (

X, n)special.
Now we consider two cases.
Case A: There are n , F
n
such that for each m > n there is no system
s
:

T) K with:
T T
p1
, max(T) u
m
, root(T) = , root(s
) = , pos(s
) = succ
T
(),
nor[s
] (2
2n
+ 2)(n + 3) and s
(t
p1
:

T
) for some T
T
p1
.
Since (K, ) is of the sUP
tree
type and
( T
p1
)( _ nor[t
p1
] > (2
2n+1
+ 5)(n + 3)),
we nd a condition q Q
tree
1
(K, ) such that
() p
[]
1

1
0
q,
() Z
def
= m > n : u
m
dcl(T
q
) = []
,
() ( T
q
)(nor[t
q
] minnor[t
p1
] : _ T
p1
(2
2n
+ 2)(n + 3)),
() (m Z)(F
m
T
q
is a front of T
q
).
Let m Z (so then F
m
T
q
is a front of T
q
and u
m
T
q
= ) and let F
m
T
q
.
By () and (2) above we know that
( T
q
)( _ nor[t
q
] > (2
2m
+ 1)(m+ 3)).
Consequently we may use 6.3.4.1 to conclude that q
[]
is mspecial. It cannot be
(

X, m)special as then the condition p
[]
1
would be (

X, m)special (compare the
remark after the denition of special conditions) contradicting / u
m
. Thus q
[]
is (

X, m)special.
For m Z and F
m
T
q
x a set C
m
T witnessing the fact that q

[]
is
(

X, m)special . Let B
m
=

FmT
q
C
m
T (for m Z).
Consider the following strategy for Player I in the game G
sR
(T):
at position number 0:
Player I writes down to the side: m
0
= min Z, q
0
= q and he plays: B
m0
.
[Note that he is (trivially) sure that if (i
0
, a
0
) is an answer of Player II then he may
nd q
1

0
q
0
such that q
1
' a
0

X and for some front F of q
1
, ( F)(
T
q1
)( _ t
q1
= t
q
).]
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

at position number k + 1:
Player I looks at the last move (i
k
, a
k
) of his opponent. He chooses a condition
q
k+1

k
q
k
and a front F of T
q
k+1
such that
( F)( T
q
k+1
)( _ t
q
k+1
= t
q
) and q
k+1
' a
k

X.
Now he takes m
k+1
Z so large that m
k+1
> m
k
and the front F
m
k+1
is above
both F and F
1
k+2
(q
k+1
). Finally:
Player I writes down to the side: m
k+1
, q
k+1
and he plays: B
m
k+1
.
[After this move he is sure that if (i
k+1
, a
k+1
) is a legal answer of the second player
then he may nd a condition q
k+2

k+1
q
k+1
such that q
k+2
' a
k+1

X and
for some front F of T
q
k+2
, _ t
q
k+2
= t
q
whenever F, T
q
k+2
. Why?
Remember the choice of Z, m
k+1
, B
m
k+1
and F
m
k+1
; see 2.3.1, clearly m
k+1
k+1.]
The strategy described above cannot be the winning one. Consequently there is a
sequence (i
0
, a
0
), (i
1
, a
1
), . . .) such that (i
n
, a
n
) are legitimative moves of Player II
against the strategy and

k
a
k
T. But in this play, Player I constructs (on a side)
a sequence p q = q
0

0
q
1

1
q
2

2
. . . of conditions such that q
k
' a
k

X.
Take the limit condition q
= lim
k
q
k
; it forces that

k
a
k

X, nishing the
proof of the theorem in Case A.
Case B: Not Case A.
Thus for every F
n
, n we nd m > n and s
:

T) K such that:
T T
p1
, max(T) u
m
, root(T) = , root(s
) = , pos(s
) = succ
T
(),
nor[s
] (2
2n
+ 2)(n + 3) and s
(t
p1
:

T
) for some T
T
p1
.
If s
:

T), m > n are as above then we will say that s
:

T) is (m, n)good
for . For each n and u
n
x a set C
n
T witnessing the fact that the

condition p
[]
1
is (

X, n)special. Now consider the following strategy for Player I
in the game G
sR
(T):
For each F
0
, Player I chooses n() > 0 and s
:

T
) K which is
(n(), 0)good. He builds a condition q
0
which starts like these sequences. Thus q
0
is such that:
(a) root(q
0
) = root(p
1
),
(b) if T
p1
is below the front F
0
then T
q0
, t
q0
= t
p1
,
(c) if

T
for some F
0
then T
q0
, t
q0
= s
,
(d) if T
p1
and there are F
0
and max(T
) such that _ then

T
q0
, t
q0
= t
p1
.
Note that F
0
def
=

F0
max(T
) is a front of T
q0
above F
0
2
(q
0
).
Now, Player I writes down to the side: q
0
, F
0
and he plays:
A
0
=
C
n()
: F
0
, max(T
).
[Thus Player I knows that F
0
is a front of T
q0
above F
0
2
(q
0
), and for each F
0
the condition p
[]
1
= q
[]
0
is (

X, 1)special and the set A
0
witnesses this fact.]
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

at the position number k + 1:
Player I looks at the last move (i
k
, a
k
) of his opponent. He chooses a condition
q
k+1

k
q
k
and a front F
k+1
of T
q
k+1
such that q
k+1
' a
k

X, the front F
k+1
is
above F
0
k+3
(q
k+1
) and for each F
k+1
the condition q
[]
k+1
= p
[]
1
is (

X, k + 2)
special. How does he nd q
k+1
and F
k+1
? He has q
k
and F
k
and he knows that
F
k
is a front of T
q
k
above F
0
k+2
(q
k
) and for each F
k
, the condition q
[]
k
= p
[]
1
is (

X, k + 1)-special and the set A
k
(played by Player I before) witnesses this fact
(this is our inductive hypothesis). Now, as a
k
A
k
, for each F
k
the rst
player may choose a condition q
+

0
q
[]
k
= p
[]
1
which forces that a
k

X and such
that ( T
q
+
)(nor[t
q
+
] > k + 1) and for some front F

+
of q
+
, if _ T
p1
,
F
+
then T
q
+
, t
q
+
= t
p1
. We may assume that the fronts F

+
are such
that F
+
F
m
+
()
for some m
+
() > k + 2 and F
+
is above F
0
k+3
(q
+
). For each
F
+
, F
k
we may choose m() > m
+
() and s
:

T
) K such that
s
:

T
) is (m(), m
+
())good. Let
F
k+1
=
_
max(T
) : ( F
k
)( < F
+
).
The condition q
k+1
is such that:
(a) below F
k
it agrees with q
k
,
(b) if _ < F
+
, F
k
, T
q
+
then T
q
k+1
, t
q
k+1
= t
q
+
,
(c) if < _

T
, F
k
, F
+
then T
q
k+1
, t
q
k+1
= s
,
(d) if < <
0
_
1
, F
k
, F
+
,
0
max(T
),
1
T
p1
then
1
T
q
k+1
and t
q
k+1
1
= t
p1
1
.
Thus q
k+1

k
q
k
, F
k+1
is a front of T
q
k+1
, and if < < F
k+1
, F
k
, F
+
then q
[]
k+1
= p
[]
1
is (

X, m())special, m() > m
+
() > k + 2 (so m() > k + 3).
Now the rst player writes down to the side q
k+1
, F
k+1
and he plays:
A
k+1
=
C
m()
: ( F
k
)( F
+
& < < u

m()
).
[Note that for each F
k+1
, the set A
k+1
witnesses that the condition q
[]
k+1
= p
[]
1
is (

X, k + 2)-special; the front F
k+1
is above F
0
k+3
(q
k+1
).]
The strategy described above cannot be the winning one. Consequently, there
is a play according to this strategy in which Player I loses. Thus we have moves
(i
0
, a
0
), (i
1
, a
1
), . . . of Player II (legal in this play) for which

k
a
k
T. But in the
course of the play the rst player constructs conditions p
1
q
0

0
q
1

1
q
2

2
. . .
such that q
k
' a
k

X. Then the limit condition q
= lim
k
q
k
forces that

k
a
k

X. This nishes the proof of the theorem.

Definition 6.3.5. For n, k, m let R
n
(k, m) be the smallest integer such
that for every colouring f : [R
n
(k, m)]
n
k there is a [R
n
(k, m)]
m
homoge-
neous for f (so this is the respective Ramsey number).
Theorem 6.3.6. Let T be an almost Ramsey ultralter on

i
i (i + 1).
Suppose that (K, ) is a nitary 2-big rich treecreating pair such that
'
Q
tree
1
(K,)
T generates an ultralter on

i
i (i + 1) .
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Then '
Q
tree
1
(K,)
the ultralter

T generated by T is almost Ramsey .
Proof. Due to 3.1.1(1) we may apply 6.2.9 and get that
'
Q
tree
1
(K,)
T generates a semiRamsey ultralter on

i
i (i + 1) .
So, what we have to do is to show that, in V
Q
tree
1
(K,)
, the ultralter

T generated by
T has the colouring property of 6.2.5(5). Suppose that a condition p Q
tree
1
(K, ),
n, L < and a Q
tree
1
(K, )name are such that p ' :

i
[i(i+1)]
n
L.
Note that if T is not interesting and it is witnessed by h

i
(i + 1) then the set
(i, h(i)) : i is almost homogeneous for and it is in T. Consequently, we
may assume that T is interesting.
We say that a condition q Q
tree
1
(K, ) is m-beautiful (for m ) if:
there is a set C T such that:
for every i and a [C i (i + 1)]
m
there are a condition
q
q and a front F of T
q
such that
root(q) = root(q
), ( T
q
)(nor[t
q
] > m+ 1),
q
' a is homogeneous and ( T

q
)(F)( _ t
q
= t
q
).
Clearly, if q is m-beautiful and k m then q is k-beautiful.
Let R
(m) = R
n
(L, m) (see 6.3.5).
Claim 6.3.6.1. Let m < , q Q
tree
1
(K, ). Assume that
( T
q
)(nor[t
q
] > (
_
R
(m)
m
_
+ 1)(m + 3))
and there are fronts F
i
of T
q
(for i ) such that conditions q
[]
(for F
i
,
i ) decide the value of `[i (i + 1)]
n
. Then the condition q is mbeautiful.
Proof of the claim: Look at the following colouring f :

i
[i (i + 1)]
m
2:
f(v) = 1 if and only if there are q

0
q and a front F of T
q
such that
q
'
Q
tree
1
(K,)
v is homogeneous for and
( T
q
)( F)( _ t
q
= t
q
) & ( T
q
)(nor[t
q
] > m+ 1).
Since T is almost Ramsey and interesting we nd j < 2 and a set C T such that
for each i :
if C i (i + 1) ,=
then [C i (i + 1)[ R
(m), f`[C i (i + 1)]

m
= j.
If j = 1 then easily the condition q is m-beautiful. Thus we have to exclude the
other possibility. So assume j = 0. Take i such that C i (i + 1) ,= and
choose v [C i (i + 1)]
R
(m)
. For v
1
[v]
m
put
u
i
v1,v
= F
i
: q
[]
'
Q
tree
1
(K,)
v
1
is homogeneous.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Note that, by the denition of R
(m), for each F

i
we nd v
1
[v]
m
such that
u
i
v1,v
. Consequently we may apply the assumption that (K, ) is rich and we
nd v
1
[v]
m
and s
+
:

T
+
) K such that
T
+
T
q
, max(T
+
) u
i
v1,v
, root(T
+
) = root(q), pos(s
+
) = succ
T
+(),
root(s
+
) = , nor[s
+
] > m+ 1 and s
+
(s
:

T
) for some T
T
q
.
[Exactly like in 6.3.4.1.] But with this in hands we easily conclude that f(v
1
) = 1,
contradicting v
1
[C i (i + 1)]
m
and the choice of j, C.
Choose a condition q p such that for some fronts F
m
of T
q
(for m ) we
have
(1) ( F
m+1
)( F
m
)( < ),
(2) ( F
m
)( T
q
)( _ nor[t
q
] > (
_
R
(m)
m
_
+ 1)(m + 3)),
(3) ( F
m
)(q
[]
decides `[m (m+ 1)]
n
)
(possible by 2.3.12 and 2.3.7(2)). By 6.3.6.1 we know that for each F
m
, m
the condition q
[]
is mbeautiful. So for every m < and F
m
we may x a set
C
m
T witnessing q
[]
is mbeautiful.
Consider the following strategy of the rst player in the game G
sR
(T):
Player I writes down to the side q
0
= q, F
0
= F
0
. He plays
C
0
: F
0
T.
arriving at position k + 1:
Player I has a condition q
k
and a front F
k
of T
q
k
such that q
k
above F
k
agrees with
q. Moreover, the set played by him before witnesses that each q
[]
is k-beautiful
(for F
k
). He looks at the last move (i
k
, a
k
) of his opponent. For each F
k
,
Player I can nd a condition q

0
q
[]
k
= q
[]
such that
(a) ( T
q
)(nor[t
q
] > k + 1),
(b) for some fronts F
of T
q
, q
above F
agrees with q and

(c) q
'
Q
tree
1
(K,)
a
k
is homogeneous.
We may think that for some m > i
k
the fronts F
are contained in F
m
(for F
k
).
Now let q
k+1
be such that
() below F
k
it agrees with q
k
,
() if _ < F
, F
k
, T
q
then T
q
k+1
, t
q
k+1
= t
q
,
() if < _ T
q
, F
k
, F
then T
q
k+1
, t
q
k+1
= t
q
= t
q
.
Let F
k+1
=

F
: F
k
. Clearly F
k+1
is a front of T
q
k+1
contained in F
m
,
q
k+1

k
q
k
and q
k+1
forces that a
k
is -homogeneous. Now:
Player I writes down to the side q
k+1
, F
k+1
and he plays the set
C
m
: F
k+1
T.
[Note that for every F
k+1
the condition q
[]
k+1
= q
[]
is k + 1-beautiful and the
set played by Player I witnesses it.]
This strategy cannot be winning for the rst player. Consequently he loses some
play according to it. Let (i
0
, a
0
), (i
1
, a
1
), . . .) be the sequence of the respective
moves of the second player (so

k
a
k
T) and let q
0
, q
1
, q
2
, . . . be the sequence of
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

conditions written down by the rst player during the play. Let q
be the limit
condition lim
k
q
k
. Then we have:
q
'
Q
tree
1
(K,)
(k )(a
k
is homogeneous ).
This nishes the proof of the theorem.
Let us nish this section with a theorem showing how several types of forcing
notions built according to our scheme may preserve genericity in the context of
ultralters. The proof of theorem 6.3.8 below resembles the proof that BlassShelah
forcing notion preserves ppoints (see [BsSh 242, 3.3]).
Definition 6.3.7. (1) A creating pair (K, ) is simple except omitting if
it is omittory, [pos(u, t)[ > 1 whenever t K, nor[t] > 0 and u basis(t),
and for every (t
0
, . . . , t
n1
) PFC(K, ) and s (t
0
, . . . , t
n1
) there is
k < n such that s (t
k
[m
s
dn
, m
s
up
)).
(2) Suppose that (K, ) is an omittory creating pair,

t = t
k
: k < )
PC
(K, ) and W :
K is a

tsystem. We say that W is
omittory compatible if
() k < n < and s (t
[m
t
k
dn
, m
tn
dn
)) imply n
W
(s) = (where
n
W
(s) is as in 5.3.1(1b,c)), and
() k < , : [m
t
k
dn
, m
t
up
) imply
W(m
t
k
dn
, m
t
up
, ) =
_
s[m
t
k
dn
, m
t
up
): s W(m
tn
dn
, m
tn
up
, `[m
tn
dn
, m
tn
up
)), kn.
Theorem 6.3.8. Suppose that a creating pair (K
0
,
0
) is simple except omit-
ting, forgetful and monotonic. Let

t PC
(K, ) and W be an omittorycompatible
tsystem. Assume that PC
(K
0
,
0
) is quasi-W-generic and generates an ul-
tralter.
(1) If (K
0
,
0
) is strongly nitary, (K, ) is a nitary, monotonic, omittory
and omittorybig creating pair then the forcing notion Q
s
(K, ) is
genericity preserving.
(2) If (K, ) is a nitary creating pair which captures singletons then the
forcing notion Q
w
(K, ) is genericity preserving.
(3) If (K, ) is a nitary tomittory tree creating pair then the forcing notion
Q
tree
1
(K, ) is genericity preserving.
Proof. 1) We have to show that, in V
Q
s
(K,)
, the demand 5.3.1(3b) is
satised (as we know that Q
s
(K, ) is proper). So suppose that is a Q
s
(K, )
name for an element of
and p Q
s
(K, ). Since (K
0
,
0
) is strongly nitary
we may repeatedly use 2.2.6 and we get a condition q p such that for each n <
and v pos(w
q
, t
q
0
, . . . , t
q
n1
) the condition (v, t
q
n
, t
q
n+1
, . . .) decides the value of
W(m
tn
dn
, m
tn
up
, `[m
tn
dn
, m
tn
up
)), where

t = t
0
, t
1
, t
2
, . . .). Fix v pos(w
q
, t
q
0
, . . ., t
q
n1
),
n < for a moment. For k n let v
k
= v
0
[g(v),m
t
q
k
up
)
pos(v, t
q
n
, . . . , t
q
k
)
(remember that (K, ) is omittory). Note that, for each k n, the condition
(v
k
, t
q
k+1
, t
q
k+2
, . . .) decides the value of W(m
t
k+1
dn
, m
t
k+1
up
, `[m
t
k+1
dn
, m
t
k+1
up
)). Thus
we nd a function (v)
such that, for each n,

(v, t
q
n
, t
q
n+1
, . . .) ' W(m
t
dn
, m
t
up
, `[m
t
dn
, m
t
up
)) = W(m
t
dn
, m
t
up
, (v)`[m
t
dn
, m
t
up
))
and for each k n the condition (v
k
, t
q
k+1
, t
q
k+2
, . . .) forces that
W(m
t
k+1
dn
, m
t
k+1
up
, `[m
t
k+1
dn
, m
t
k+1
up
)) = W(m
t
k+1
dn
, m
t
k+1
up
, (v)`[m
t
k+1
dn
, m
t
k+1
up
)) .
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Since is quasi-W-generic we nd s = s
m
: m < ) such that for every n <
and v pos(w
q
, t
q
0
, . . . , t
q
n1
) we have
(
m )(s
m
W(m
sm
dn
, m
sm
up
, (v)`[m
sm
dn
, m
sm
up
)))
(remember (, _) is directed countably closed). Next we choose inductively an
increasing sequence (0) < (1) < (2) < . . . < such that
if v pos(w
q
, t
q
0
, . . . , t
q
n1
), m
tn
up
m
s
(i)
up
, n, i
then (m (i + 1))(s
m
W(m
sm
dn
, m
sm
up
, (v)`[m
sm
dn
, m
sm
up
))).
For j < 4 let Y
j
=

i
[m
s
(4i+j)
dn
, m
s
(4i+j+1)
dn
). Since generates an ultralter,
exactly one of these sets is in T(). Without loss of generality we may assume that
Y
2
T() (otherwise start the sequence of the (i)s from (1), or (2) or (3)).
This means that we nd s
= s
m
: m < ) such that s s
and for suciently

large m (say m m
), there are i = i(m) and k = k(m) such that

m
s
(4i+2)
dn
m
t
k
dn
< m
t
k
up
m
s
(4i+3)
dn
and s
m
(t
k
[m
s
m
dn
, m
s
m
up
))
(remember (K
0
,
0
) is simpleexceptomitting and monotonic). Let x(i)
be such that m
t
x(i)
dn
= m
s
(4i+3)
dn
(for i ). Now we dene a condition q
=
(w
q
, t
q
0
, t
q
1
, . . .):
w
q
= w
q
, t
q
0
= t
q
x(0)
[g(w
q
), m
t
q
x(0)
up
), and t
q
i+1
= t
q
x(i+1)
[m
t
q
x(i)
up
, m
t
q
x(i+1)
up
).
Plainly, this denes a condition in Q
s
(K, ) stronger than q. We claim that
q
' (m > m
)(s
m
W(m
s
m
dn
, m
s
m
up
, `[m
s
m
dn
, m
s
m
up
))).
Assume not. Then we nd q
and m > m
such that
q
' s
m
/ W(m
s
m
dn
, m
s
m
up
, `[m
s
m
dn
, m
s
m
up
))).
Of course we may assume that g(w
q
) m
t
q
i(m)
up
. Let k
0
, k
1
, k
2
be such that m
t
k
dn
=
m
s
(4i(m)+)
dn
and let v = w
q
`m
t
q
k
0
dn
. Clearly
m
t
q
i(m)
dn
< m
t
q
k
0
up
< m
t
q
k
1
up
< m
t
q
k
2
up
m
t
q
k(m)
up
< m
t
q
x(i(m))
up
= m
t
q
i(m)
up
,
v pos(w
q
, t
q
0
, . . . , t
q
k
0
1
) (by smoothness) and w
q
`[m
t
q
k
0
dn
, m
t
q
x(i(m))
dn
) 0 (by the
denition of t
q
i(m)
, remember (K, ) is monotonic). Consequently for each k
[k
0
, x(i(m))) we have v
k
_ w
q
. Since k
0
< k(m) < x(i(m)) we conclude that the
condition q
forces
W(m
t
k(m)
dn
, m
t
k(m)
up
, `[m
t
k(m)
dn
, m
t
k(m)
up
)) = W(m
t
k(m)
dn
, m
t
k(m)
up
, (v)`[m
t
k(m)
dn
, m
t
k(m)
up
)) .
Since m
t
k
0
up
< m
s
(4i(m)+1)
up
, for each i [(4i(m) + 2), (4i(m) + 3)) we have s
i

W(m
si
dn
, m
si
up
, (v)`[m
si
dn
, m
si
up
)) (remember that choice of the sequence of the (i)s).
Now look at the condition 5.3.1(1c). Since W is omittorycompatible we have
n
W
(s
m
)=k(m)[k
2
, x(i(m))) and therefore s
m
W(m
s
m
dn
, m
s
m
up
, (v)`[m
s
m
dn
, m
s
m
up
)).
Hence, by 6.3.7(2), we easily get
q
' s
m
W(m
s
m
dn
, m
s
m
up
, `[m
s
m
dn
, m
s
m
up
)) ,
a contradiction.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

2) Repeat the proof of 1) noting that we do not have to assume that (K
0
,
0
)
is strongly nitary as we use 2.1.12 instead of 2.2.6. (Dening v
k
we use xed
sequences u
n
(for n ) such that u
n
pos(u, t
q
n
) for each u basis(t
q
n
).)
3) Similarly (remember that (K, ) is nitary).
Remark 6.3.9. In 6.3.8(3) we need the assumption that the tree creating pair
(K, ) is nitary. The forcing notion D
of [NeRo93], in which conditions are trees

<
such that each node has an extension which has all possible successors in
the tree, adds a Cohen real (see [NeRo93, 2.1]). This forcing may be represented
as Q
tree
1
(K
) for some tomittory (not nitary) tree creating pair (K
).
6.4. Examples
Example 6.4.1. Let H(m) = 2 for m < . We construct a creating pair
(K
6.4.1
,
6.4.1
) for H,

t PC
(K
6.4.1
,
6.4.1
) and

t-systems W
n
L
(for n, L < )
such that
(1) (K
6.4.1
,
6.4.1
) is simple except omitting, nitary, forgetful, monotonic,
interesting, condensed and generates an ultralter, the systems W
n
L
are
omittorycompatible,
(2) if P
t, (K
6.4.1
,
6.4.1
)) is quasi-W
n
L
-generic then T() is a lter on
i
i (i +1) such that for every colouring f :

i
[i (i +1)]
n
L
there is a set A T() almost homogeneous for f,
(3) if P
t, (K
6.4.1
,
6.4.1
)) is _-directed and countably closed and T()
is a lter such that for every colouring f :

i
[i (i + 1)]
n
L
(n, L < ) there is a set A T() almost homogeneous for f then is
quasi-W
n
L
-generic for all n, L.
Construction. A creature t CR[H] is in K
6.4.1
if for some non-empty
subset a
t
of [m
t
dn
, m
t
up
) we have nor[t] = log
2
([a
t
[) and
val[t] = u, v) 2
m
t
dn
2
m
t
up
: (n[m
t
dn
, m
t
up
))(v(n) = 1 n a
t
).
We dene
6.4.1
by:
if t
0
, . . . , t
n
K
6.4.1
are such that m
t
up
= m
t
+1
dn
(for < n) then
6.4.1
(t
0
, . . . , t
n
) = s K
6.4.1
: m
t0
dn
= m
s
dn
& m
tn
up
= m
s
up
& ( n)(a
s
a
t
).
6.4.1
,
6.4.1
) is a nitary, forgetful, monotonic and simple
except omitting creating pair. It is interesting as nor[t] > 0 implies [a
t
[ > 1
(for t K
6.4.1
). By the denition of
6.4.1
, one easily shows that (K
6.4.1
,
6.4.1
) is
condensed and generates an ultralter. Moreover, (K
6.4.1
,
6.4.1
) is strongly nitary
modulo
6.4.1
. Now, let

t = t
i
: i < ) PC
(K
6.4.1
,
6.4.1
) be such that for
i :
m
ti
dn
=
i(i + 1)
2
, m
ti
up
=
(i + 1)(i + 2)
2
, a
ti
= [m
ti
dn
, m
ti
up
)
(so nor[t
i
] = log
2
(i +1)). We will identify the interval [m
ti
dn
, m
ti
up
) with i(i +1)
(for i ). Thus, if P
t, (K
6.4.1
,
6.4.1
)) is quasi-W-generic for some

t
system W, then we may think of T() as a lter on

i
i (i + 1). The lter
T() is interesting by 6.1.6(1).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

6.4. EXAMPLES 141
Fix n, L < . For each i choose a mapping
i
:
[m
ti
dn
, m
ti
up
) onto
f : f is a function from [[m
ti
dn
, m
ti
up
)]
n
to L.
Next, for i j < and : [m
ti
dn
, m
tj
up
) dene
W
n
L
(m
ti
dn
, m
tj
up
, ) = t
6.4.1
(t
i
, . . . , t
j
) : if i k j, a
t
a
t
k
then a
t
is homogeneous for
k
(`[m
t
k
dn
, m
t
k
up
))
(in all other instances we let W
n
L
(m
, m
, ) = ).
Claim 6.4.1.1. W
n
L
is an omittorycompatible

tsystem.
Proof of the claim: The requirement 5.3.1(1a) is immediate by the denition
of W
n
L
. For 5.3.1(1b,c) remember the way we dened the composition operation
6.4.1
: if s
6.4.1
(s
0
, . . . , s
k
) then a
s
a
s
for some 0 k. Finally note

that if s
6.4.1
(t
k
, . . . , t
), k < , nor[s] > log

2
(R
n
(L, m)) (see 6.3.5) and
: [m
s
dn
, m
s
up
) then there is t W
n
L
(m
s
dn
, m
s
up
, ) such that t
6.4.1
(s)
and nor[t] log
2
(m) (by the denition of R
n
(L, m)). This gives the suitable for
5.3.1(1d) function G. Thus we have veried that W
n
L
is a

tsystem. It should be
clear that W is omittorycompatible.
Note that if P
t, (K
6.4.1
,
6.4.1
)) is _directed then T() is the lter
generated by all sets A
N
s
def
=

Nn
a
sn
, for s = s
n
: n < ) and N < .
Therefore we easily check that the systems W
n
L
(for n, L < ) are as required in 2,
3 of 6.4.1, noting that each colouring f :

i
[i (i + 1)]
n
L corresponds via
i
: i < ) to some function
.
Proposition 6.4.2. Assume CH. Then there exist P
t, (K
6.4.1
,
6.4.1
))
which is quasi-W
n
L
-generic for all n, L < and such that T() is a semiRamsey
ultralter on

i
i (i +1). Consequently, T() is an interesting almost Ramsey
ultralter on

i
i (i + 1).
Proof. This is somewhat similar to 6.1.6(3) (and so to 5.3.4(2)), but we have
to be more careful to ensure that T() is semiRamsey. For this, as a basic step
of the inductive construction of , we use the following observation.
Claim 6.4.2.1. Suppose that s P
t, (K
6.4.1
,
6.4.1
)), n, L < ,
and : [
i
i (i + 1)]
. Then there exists s
= s
m
: m < )
P
t, (K
6.4.1
,
6.4.1
)) such that s s
and
(1) (
m)(s
m
W
n
L
(m
s
m
dn
, m
s
m
up
, `[m
s
m
dn
, m
s
m
up
))),
(2) (m )([a
s
m
[ m + 1),
(3) if i
m
< (for m ) are such that a
s
m
i
m
(i
m
+1)
= [m
t
i
m
dn
, m
t
i
m
up
)
then
either (k )(m )(a
s
m
(k) = ) or (m )(a
s
m+1
(i
m
)).
Proof of the claim: Let s = s
m
: m < ) and let i
m
be such that a
sm

i
m
(i
m
+ 1) (remember

t s, see the denition of (K
6.4.1
,
6.4.1
)). We know
that lim
m
[a
sm
[ = , so we may choose m
0
< m
1
< m
2
< . . . < such that
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

(k )([a
sm
k
[ R
n
(L, k)) (see 6.3.5). Choose b
k
[a
sm
k
]
k
homogeneous for the
colouring of [i
m
k
(i
m
k
+ 1)]
n
(with values in L) coded by `[m
tim
k
dn
, m
tim
k
up
).
Next choose k(0) < k(1) < . . . < and c
[b
k()
]
+ 1
(for < ) such that
either (k )( )(c
(k) = ) or (m )(c
+1
(i
m
k()
)).
Let s
0
K
6.4.1
be such that m
s
0
dn
= m
s0
dn
, m
s
0
up
= m
sm
k(0)
up
, a
s
0
= c
0
and let
s
+1
K
6.4.1
(for ) be such that m
s
+1
dn
= m
sm
k()
up
, m
s
+1
up
= m
sm
k(+1)
up
and
a
s
+1
= c
+1
. Easily, the sequence s
= s
n
: n < ) is as required.
Assume CH. Using 6.4.2.1, we may construct a sequence s
: <
1
)
P
t, (K
6.4.1
,
6.4.1
)) such that
() < <
1
s
_ s
,
() for each n, L < we have
(
)( <
1
)(
m)(s
,m
W
n
L
(m
s,m
dn
, m
s,m
up
, `[m
s,m
dn
, m
s,m
up
))),
() for each function : [
i
i (i + 1)]
there is <
1
such that
if a
s,m
i
,m
(i
,m
+ 1) (for m ) then
either (k )(m )(a
s,m
(k) = ) or (m )(a
s,m+1
(i
,m
)).
Like in 6.2.7 and 5.3.4(2) we check that
def
= s
: <
1
is quasi-W
n
L
-generic for
all n, L < and T() is a semiRamsey ultralter on

i
i (i + 1).
Conclusion 6.4.3. Assume CH. Let (K
6.4.1
,
6.4.1
),

t, W
n
L
be given by 6.4.1,
and let P
t, (K
6.4.1
,
6.4.1
)) be quasi-W
n
L
-generic for all n, L < such that
T() is a semiRamsey ultralter on

i
i (i + 1) (see 6.4.2). Suppose that
is a limit ordinal and P
,

Q
: < ) is a countable support iteration of proper
bounding forcing notions such that for each < :

'
P
generates an interesting almost Ramsey ultralter.
Then '
P
generates an interesting almost Ramsey ultralter.

Proof. By 6.4.1(3) we have
'
P
is quasi-W
n
L
-generic for all n, L <
(for each < ). Hence, by 6.1.7, we get
'
P
is quasi-W
n
L
-generic for each n, L < and generates an ultralter.
As P
is
bounding (by [Sh:f, Ch VI, 2.3, 2.8]) we may apply 6.2.9 to conclude
that
'
P
T() is a semiRamsey ultralter.

Consequently, by 6.4.1(2), we have
'
P
T() is an interesting almost Ramsey ultralter.
Example 6.4.4. Let
be such that (n )((n) > (n + 1)

2
).
We build a tree creating pair (K
6.4.4
,
6.4.4
) which is: nitary, 2-big, rich (see 6.3.3)
and of the sUP(T)
tree
-type (see 6.3.1) for every Ramsey ultralter T on .
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

6.4. EXAMPLES 143
Construction. Let H(n) = [(n)]
n + 1
.
For

m<n
H(m) (n ) and A

m<
m (m) we will write A if
(m < n)(k < (m))((m, k) A k (m)).
Now we dene (K
6.4.4
,
6.4.4
). A treelike creature t TCR
[H] is in K
6.4.4
if:
(1) val[t] is nite and
(2) nor[t] = log
2
(min[A[ : A

mg()
m (m) & ( pos(t))(A ,
)).
By the denition, K
6.4.4
is nitary. The tree composition
6.4.4
is generated simi-
larly to
tsum
of 4.2.6 but with norms as above. Thus, if t
:

T) K
6.4.4
is a
system of tree-creatures such that T is a well founded quasi tree, root(t
) = , and
rng(val[t
]) = succ
T
() (for

T) then we dene S
(t
:

T) as the unique
creature t
in K
6.4.4
with rng(val[t
]) = max(T), dom(val[t
]) = root(T) and
dis[t
] = dis[t
] :

T). Now we let
6.4.4
(t
:

T) = t K
6.4.4
: val[t] val[S
(t
:

T)].
Clearly,
6.4.4
is a tree-composition on K
6.4.4
. Note that if t K
6.4.4
then we may
identify elements of
6.4.4
(t) with subsets of pos(t): for each non-empty u pos(t),
t
u
is the unique creature in K
6.4.4
with pos(t
u
) = u and root(t
u
) = root(t). More
general, if p Q
tree
(K
6.4.4
,
6.4.4
), F is a front of T
p
, u F then there is a unique
creature t(p, u) K
6.4.4
such that
pos(t(p, u))=u, root(t(p, u))=root(p) and t(p, u)
6.4.4
(t
p
: (F)( < )).

Claim 6.4.4.1. nor[S
(t
:

T)] minnor[t
] :

T.
Proof of the claim: Suppose that m < 2
min{nor[t]:
T}
, but there is a set A
kn0
k (k) (where n
0
= g(root(T))) such that [A[ = m and A , for each
max(T). Now we build inductively a bad max(T): since m < 2
nor[t
root(T)
]
we nd
0
pos(t
root(T)
) such that A
0
. Next we look at A
1
= A

kg(0)
k
(k). Since m < 2
nor[t
0
]
we nd
1
pos(t
0
) such that A
1

1
. Continuing in
this fashion, after nitely many steps, we get
k
max(T) such that A
k
, a
contradiction.
Claim 6.4.4.2. (K
6.4.4
,
6.4.4
) is 2-big.
Proof of the claim: Let t K
6.4.4
, nor[t] > 0 and let pos(t) = u
0
u
1
. Take sets
A
0
, A
1

mn0
m (m) (where n
0
= g(root(t))) such that for i = 0, 1:
( u
i
)(A
i
, ) and log
2
([A
i
[) = nor[t
ui
].
Look at A = A
0
A
1
. Clearly ( pos(t))(A , ). Hence, for some i < 2:
nor[t] log
2
([A[) log
2
([A
i
[) + 1 = nor[t
ui
] + 1.
Claim 6.4.4.3. (K
6.4.4
,
6.4.4
) is rich.
Proof of the claim: Suppose that s
:

T) K
6.4.4
, n and u are such that
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

(1) T

k
m<k
H(m) is a well founded quasi tree, u max(T),
(2) root(s
) = , pos(s
) = succ
T
(), nor[s
] > n + 3,
(3) there is no system s
:

T
) K
6.4.4
such that
T
T, max(T
) u, root(T
) = root(T), pos(s
) = succ
T
()
root(s
) = , nor[s
] > n + 1, and
s
6.4.4
(s
:

T
) for some T
T.
Let t = S
(s
:

T). By 6.4.4.1 we have nor[t] > n + 3 (remember 2. above).
But now, considering t
u
(dened as before) we note that necessarily nor[t
u
]
n + 1 < nor[t] 1. Let v = pos(t) u. By the bigness (see 6.4.4.3) we have
nor[t
v
] nor[t] 1, nishing the claim (remember 6.4.4.1).
Claim 6.4.4.4. Let T be a Ramsey ultralter on . Then (K
6.4.4
,
6.4.4
) is of
the sUP(T)
tree
type.
Proof of the claim: Assume that 1 m < , p Q
(K
6.4.4
,
6.4.4
), nor[t
p
] > m+1
for each T
p
and F
0
, F
1
, . . . are fronts of T
p
such that
(n )( F
n+1
)( F
n
)( < ).
Further suppose that u
n
F
n
are such that there is no system s
:

T) with
pos(S
(s
:

T)) u
n
, root(T) = root(T
p
), and nor[s
] m.
In particular, this means that nor[t(p, u
n
)] < m for each n (t(p, u
n
) is as
dened earlier). Thus, for each n , we nd a set
A
n

_
kg(root(p))
k (k)
such that
[A
n
[ = 2
m
and ( u
n
)(A
n
, ).
Now we use the assumption that T is Ramsey: we nd sets Z
0
T and A
i
i (i) such that
(n
0
, n
1
Z
0
)(n
0
< n
1
A
n0
A
n1
= A
).
[How? Just consider the colouring f of []
2
such that f(n
0
, n
1
) codes the trace
of A
n0
on A
n1
(for n
0
< n
1
, in the canonical enumerations of A
n
s) and take an
f-homogeneous set.] Now choose Z Z
0
, Z T such that if n
0
< n
1
, n
0
, n
1
Z
then
mini : (A
n1
A
) i (i) ,= > maxg() : F

n0
.
[How? Consider the following strategy for Player I in the game G
R
(T):
at stage k +1 of the game he looks at the last move i
k
of the second player and he
chooses N such that if n N, n Z
0
then
(A
n
A
)
_
_
i (i) : i maxg() : F
i
k
_
= .
Now he plays Z
0
(N, ). This strategy cannot be the winning one.] Using the set
Z we build the suitable condition q Q
tree
(K
6.4.4
,
6.4.4
). It will be constructed
in such a way that p
1
0
q, T
q
root(q)

Z
F
and each F
T
q
will be a
front of T
q
(for Z). Let
0
= min Z and v
0
= F
0
: A
0
. By
the choice of A
0
we know that v
0
u
0
= . Let t
q
root(q)
= t(p, v
0
). Note that
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

6.4. EXAMPLES 145
2
nor[t(p,v
0
)]
+2
m
2
nor[t(p,F
0
)]
, and hence, as nor[t(p, F
0
)] > m+1 (see 6.4.4.1),
we have
nor[t(p, v
0
)] nor[t(p, F
0
)] m minnor[t
p
] : root(p) _ T
p
m.
We put t
q
root(q)
= t(p, v
0
). Suppose that we have dened T
q
up to the level of F
,
Z (thus we know T
q
F
already). Let T
q
F
and let
= min(Z( +1)).
By the rst step of the construction we know that A
. By the choice of Z we
have that
(A
)
_
i<g()
i (i) = .
We take v
,
= F
: < & A
. As before, v
,
u
= and
nor[t(p
[]
, v
,
)] nor[t(p
[]
, F
0
T
p
[]
)] m minnor[t
p
] : _ T
p
m
(remember 6.4.4.1).
Now we easily check that the condition q constructed above is as required by
6.3.1 to show that (K
6.4.4
,
6.4.4
) is of the sUP(T)
tree
type.
One easily checks that the forcing notion Q
tree
1
(K
6.4.4
,
6.4.4
) is non-trivial
(i.e. K
6.4.4
contains enough tree like creatures with arbitrarily large norms). Let us
show another property of the tree creating pair (K
6.4.4
,
6.4.4
).
Proposition 6.4.5. Let (K
6.4.4
,
6.4.4
) be the tree creating pair of 6.4.4. Then:
(1) '
Q
tree
1
(K
6.4.4
,
6.4.4
)
(m )(

W(m) [(m)]
m + 1
) and
(2) if h is a partial function, dom(h) []
and (m dom(h))(h(m) <

(m)) then
'
Q
tree
1
(K
6.4.4
,
6.4.4
)
(
m dom(h))(h(m)

W(m))
(where

W is the generic real added by Q
tree
1
(K
6.4.4
,
6.4.4
), see 1.1.13).
Proof. 1) Should be clear.
2) Let p Q
tree
1
(K
6.4.4
,
6.4.4
) and let h be as in the assumptions. We may
assume that ( T
p
)(nor[t
p
] > 5). Let m

0
= min
_
dom(h) g(root(p))
_
. Take
a front F of T
p
such that ( F)(g() > m
0
) and look at the set u = F :
h(m
0
) (m
0
). Plainly, nor[t(p, u)] nor[t(p, F)] 1 4. Let q be a condition
in Q
tree
1
(K
6.4.4
,
6.4.4
) such that root(q) = root(p), T
q
T
p
, t
q
root(q)
= t(p, u) and
if u, _ T
p
then T
q
, t
q
= t
p
. Clearly q p and q ' h(m

0
)

W(m
0
).
Now we may easily nish.
Conclusion 6.4.6. The following is consistent with ZFC:
(1) there is an almost Ramsey interesting ultralter on

i
i (i +1) which
is generated by
1
elements (so m
2
= =
1
, see the introduction to this
chapter) and
(2) m
1
=
2
, and even more: for each function
and a family T of
1
partial innite functions h : dom(h) such that
(m dom(h))(h(m) < (m))
there is W

m
[(m)]
m + 1
such that
(h T)(
m dom(h))(h(m) W(m)).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Proof. Start with V [= CH. By 6.4.2 we have an interesting almost Ram-
sey ultralter T = T() on

i
i (i + 1) generated by a quasi generic
P
t, (K
6.4.1
,
6.4.1
)) as there. Build a countable support iteration P
,

Q
: <
2
) and a list

: <
2
) such that for each <
2
:
(1)

is a P
name for a function in
such that (n )(

(n) > (n +
1)
2
),
(2)

Q
is the P
name for the forcing notion Q

tree
1
(K

6.4.4
,

6.4.4
),
(3)

: <
2
) lists with
2
repetitions all (canonical) P
2
names for
functions
such that (n )((n) > (n + 1)

2
).
We claim that if G P
2
is a generic lter over V, then, in V[G], the two sentences
of the conclusion hold true. Why? One can inductively show that for each
2
:
'
P
generates an almost Ramsey interesting ultralter on
_
i
i (i + 1)
(at successor stages use 6.3.4, 6.3.6 and 6.4.4; at limit stages use 6.4.3). Hence, in
V[G], the ultralter T() witnesses the rst property. For the second assertion use
6.4.5.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

CHAPTER 7
Friends and relatives of PP
In this chapter we answer a question of Balcerzak and Plewik, showing that the
cardinal number
BP
(see 7.1.1) may be smaller than the continuum (7.5.3) and
that it may be larger than the dominating number (7.5.2). As this cardinal turns
out to be bounded by a cardinal number related to the strong PPproperty, we
take this opportunity to have a look at several properties close to the PPproperty.
7.1. BalcerzakPlewik number
For an ideal of subsets of 2
it is natural to ask if it has the following

property (P):
(P)
J
every perfect subset of 2
contains a perfect set from .

The property (P) has numerous consequences and applications (see e.g. Balcerzak
[Ba91], some related results and references may be found in Balcerzak Ros lanowski
[BaRo95]) and it is usually easy to decide if (P)
J
holds. However, that was not
clear for some of Mycielskis ideals /
2,K
.
Suppose that / []
is a non-empty family such that

() (X /)(X
0
, X
1
/)(X
0
, X
1
X & X
0
X
1
= ).
Let /
2,K
consist of these sets A 2
that
(X /)(f : X 2)(g A)(f g).
It is easy to check that /
2,K
is a ideal of subsets of 2
. These ideals are relatives

of the ideals from Mycielski [My69] and were studied e.g. in Cicho n Ros lanowski
Stepr ans W eglorz [CRSW93] and [Ro94]. If the family / is countable than easily
the ideal /
2,K
determined by it has the property (P). D ebski, Kleszcz and Plewik
[DKP92] showed that the ideal /
2,[]
does not satisfy (P). Then Balcerzak and

Plewik dened the following cardinal number
BP
(see Balcerzak Plewik [BaPl96]).
Definition 7.1.1. The BalcerzakPlewik number
BP
is the minimal size of a
family / []
such that
for some perfect set Q 2
, for every perfect subset P of Q there

is X / such that
P`X
def
= f 2
X
: (g P)(f g) = 2
X
.
[Note that
BP
is the minimal size of / []
satisfying () for which the ideal

/
2,K
does not have the property (P).]
They proved that d
BP
and asked if
it is consistent that
BP
< c,
it is consistent that d <
BP
.
147
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

148 7. FRIENDS AND RELATIVES OF PP
A full answer to these questions will be given in the nal part of this chapter. Now
we want to give an upper bound to
BP
.
Definition 7.1.2. Let A be the space of all sequences w = w
i
: i B) such
that B []
and (i B)(w
i
[]
i
). We dene a relation R
sPP

A by
(, w) R
sPP
if and only if
, w A, and (i dom( w))((i) w

i
).
Note that the space A carries a natural Polish topology (inherited from the
product space of all w = w
i
: i < ) such that for each i , either w
i
= or
[w
i
[ = i). The relation R
sPP
describes the strong PPproperty of [Sh:f, Ch VI,
2.12E]: a proper forcing notion P has the strong PPproperty if and only if it has
the R
sPP
localization property (see 0.2.2).
Theorem 7.1.3.
BP
d(R
sPP
).
Proof. Construct inductively a perfect tree T 2
<
and an increasing se-
quence 0 = k
0
< k
1
< k
2
< . . . < such that for every i :
() ( T 2
k
i
)(
0
,
1
T 2
k
i+1
)( <
0
& <
1
&
0
,=
1
),
() for each colouring f : T 2
k
i
2 there is n [k
i
, k
i+1
) such that
( T 2
k
i+1
)((n) = f(`k
i
)).
The construction is straightforward. It is not dicult to check that if P [T] is a
perfect set then there is X []
such that P`X = 2

X
(or see [DKP92]).
Let D A be such that [D[ = d(R
sPP
) and
(
)( w D)((, w) R
sPP
).
Let N be an elementary submodel of H() such that D, T, k
i
: i < ) N, D N
and [N[ = [D[. We are going to show that
for each perfect set P [T] there is X N []
such that P`X = 2

X
(what will nish the proof of the theorem). To this end suppose that T
T is
a perfect tree. Since N
is a dominating family (as the strong PPproperty

implies
bounding) we may choose an increasing sequence n

i
: i < ) N
such that
(i )( T
2
k
ni
)([ T
2
k
ni+1
: < [ > 2(i + 1)).
As we may encode (in a canonical way) subsets of 2
n
as integers, we may use the
choice of D and N and nd a sequence w
i
: i B) N such that for each i B:
(i) 0 < [w
i
[ i,
(ii) a w
i
a T 2
k
ni
,
(iii) T
2
k
ni
w
i
.
By shrinking each w
i
if necessary, we may additionally demand that [a[ 2min(B)
for each a w
min(B)
and if i < j both are from B, a w
j
then
(iv) for each a the set a : `k
ni
= `k
ni
has at least 2j elements
(remember the choice of the n
i
s). Now, working in N, we inductively build a perfect
tree T
+
T. First for each a w
min(B)
we choose
a,0
,
a,1
a such that there
are no repetitions in
a,
: a w
min(B)
, < 2) (possible as [w
min(B)
[ min(B)
and [a[ 2 min(B) for a w
min(B)
). We declare that
T
+
2
k
n
min(B)
=
a,
: a w
min(B)
, < 2.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

7.2. AN ITERABLE FRIEND OF THE STRONG PPPROPERTY 149
Suppose that we have dened T
+
2
k
ni
, i B and j = min(B (i +1)). For each
T
+
2
k
ni
and a w
j
we choose
a,0
,
a,1
T 2
k
nj
such that
(1) there are no repetitions in
a,0
,
a,1
: a w
j
),
(2) if a w
j
is such that ( a)( < ) then
a,0
,
a,1
a.
Again, the choice is possible by (iv). We declare that
T
+
2
k
nj
=
a,
: T
+
2
k
ni
, a w
j
, < 2.
This fully describes the construction (in N) of the tree T
+
T. Next, working
still in N, we choose integers m
j
[k
nj
, k
nj+1
) such that for each j B:
(+) if T
+
2
k
ni
, i B is such that j = min(B (i + 1)) (or = ) and
j = min(B)) and a w
j
and
a,
< T 2
k
nj+1
then (m
j
) =
(possible by () of the choice of the tree T and the rst demand of the choice of
the
a,
s). Let X = m
j
: j B []
N. Suppose f : X 2. By clause (iii)

we know that for each j B, b
j
def
= T
2
k
nj
w
j
. Consequently we may build
inductively an innite branch [T
] such that `k
n
min(B)
=
b
min(B)
,f(m
min(B)
)
and
for each i B, if j = min(B (i +1)),
i
= `k
ni
then `k
nj
=
i
bj ,f(mj)
. It follows
from (+) that f .
Remark 7.1.4. Note that a sequence w
i
: i B), where w
i
[]
i
, may
be interpreted as a sequence v
i
: i B), where [v
i
[ = i and members of v
i
are functions from the interval [
i(i+1)
2
,
(i+1)(i+2)
2
) to . Consequently, considering
suitable diagonals, we may use Bartoszy nskiMillers characterization of non(/)
(see [BaJu95, 2.4.7]) and show that non(/) d(R
sPP
). As clearly d d(R
sPP
)
we conclude that cof (/) d(R
sPP
) (by [BaJu95, 2.2.11]). On the other hand,
it follows from Bartoszy nskis characterization of cof (^) (see [BaJu95, 2.3.9])
that d(R
sPP
) cof (^) (just note that the Sacks property implies the strong PP
property).
7.2. An iterable friend of the strong PPproperty
The PPproperty is preserved in countably support iterations of proper forcing
notions (see [Sh:f, Ch VI, 2.12]). However, it is not clear if the strong PPproperty
is preserved. Here, we introduce a property stronger then the strong PP which is
preserved in countable support iterations.
Definition 7.2.1. Let T be a lter on , x
be a non-decreasing function
and let

T = (T, <
F
) be a partial order on T
.
(1) The lter T is weakly non-reducible if it is non-principal and for every
partition X
n
: n < ) of into nite sets there exists a set Y []
such that

n\Y
X
n
T.
If above we allow partitions into sets from the dual ideal T
c
then we
say that T is non-reducible.
(2) The partial order

T is PPok if it is dense, has no maximal and mini-
mal elements, each member of T is non-decreasing, the identity function
belongs to T and
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

() if h
0
, h
1
T, h
0
<
F
h
1
then (n )(1 h
0
(n) h
1
(n)) and
lim
n
h1(n)
h0(n)
= and there is h T such that
(
n )(h(n)
h
1
(n)
h
0
(n)
).
(3) We say that a proper forcing notion P has the (T, x)strong PPproperty
if
'
P
for every
there are B T V and w

i
: i B) V such that
(i B)([w
i
[ x(i) & (i) w
i
) .
(4) A proper forcing notion P has the (T,

T)strong PPproperty if it has the
(T, x)strong PPproperty for every x T.
Remark 7.2.2. (1) Each non-principal ultralter on is non-reducible.
Clearly non-reducible lters are weakly non-reducible.
(2) One can easily construct a countable partial order

T which is PPok and
such that x T for any pregiven non-decreasing unbounded function
x
.
Theorem 7.2.3. Suppose that T is a non-principal p-lter on (see 6.2.1(2))
and

T = (T, <
F
) is a PPok partial order. Then:
(1) every proper forcing notion which has the (T,

T)strong PPproperty has
the strong PPproperty,
(2) the (T,

T)strong PPproperty is preserved in countable support iterations
of proper forcing notions.
Proof. 1) Should be clear.
2) We will apply [Sh:f, Ch VI, 1.13A], so we will follow the terminology of [Sh:f,
Ch VI, 1]. However, we will not quote the conditions which we have to check, as
that was done in the proof of 5.2.9 (and the proof here is parallel to the one there).
We will present the proof in a slightly more complicated way than needed, but later
we will be able to refer to it in a bounded context (in 7.3.6). Moreover, in this way
the analogy to 5.2.9 will be more clear.
For each m 1 we x a function
m
:
onto
[]
m
and for h T we dene
h
:

n
[]
<
by
h
()(n) =
h(n)
((n)). Further, for h
, h T, B T
and w = w
i
: i B) such that h
<
F
h and (i B)(w
i
[]
h(i)
) we put
T
w,h
=
<
: (i B g())(
h
(i)
((i)) w
i
).
Each T
w,h
is a perfect subtree of
<
. Now we dene:
D
D,

F
is H(
1
)
V
,
for x, T D
D,

F
we say that x R
D,

F
T if and only if
x = h
, h) and T = T
w,h
for some h
, h T and w = w
i
: i B)
D
D,

F
such that h
<
F
h, B T, and (i B)(w
i
[]
h(i)
),
for h
, h), h
, h
) dom(R
D,

F
) we say that h
, h) <
D,

F
h
, h
) if and
only if h
= h
<
F
h <
F
h
.
Claim 7.2.3.1. (1) (D
D,

F
, R
D,

F
) is a weak covering model in V.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

7.2. AN ITERABLE FRIEND OF THE STRONG PPPROPERTY 151
(2) In any generic extension V
of V in which (D
D,

F
, R
D,

F
) covers, a forcing
notion P is (D
D,

F
, R
D,

F
)preserving if and only if it has the (T,

T)strong
PPproperty.
[Compare 5.2.9.1, 5.2.9.2.]
Proof of the claim: 1) Check.
2) Suppose that (D
D,

F
, R
D,

F
) covers in V
and P V
is a forcing notion with

the (T,

T)strong PPproperty. Let h
, h) dom(R
D,

F
) and
(V
)
P
.
Choose h
0
, h
1
T such that h
0
<
F
h
1
and (
n )(h
1
(n)
h(n)
h
(n)
) (possible by
7.2.1(2)). Take w
= w
i
: i B
) V
such that B
T V and
(i B
)(w
i
[]
h
0
(i)
& (i) w
i
).
Let
be such that
h0(i)
(
(i)) = w
i
(for i B
). Since (D
D,

F
, R
D,

F
)
covers in V
and h
0
, h
1
) dom(D
D,

F
), we nd w = w
i
: i B) D
D,

F
such
that
lim(T
w,h0
) and (i B)(w
i
[]
h
1
(i)
). Let B
+
= B B
T V,
w
+
i
=

kwi
(i)
(k) for i B
+
. Note that for suciently large i B
+
[w
+
i
[ [w
i
[ h
(i) h
1
(i) h
(i) h(i)
and we may assume that this holds for all i B
+
. Letting w
+
= w
+
i
: i B
+
) we
will have h
, h) R
D,

F
T
w
+
,h
and lim(T
w
+
,h
), and hence (D
D,

F
, R
D,

F
) covers
in (V
)
P
.
The converse implication is even simpler.
Claim 7.2.3.2. (D
D,

F
, R
D,

F
, <
D,

F
) is a ne covering model.
[Compare 5.2.9.3.]
Proof of the claim: Immediately by the denition of (D
D,

F
, R
D,

F
, <
D,

F
) one sees
that the demands (), ()(i)(iii) of [Sh:f, Ch VI, 1.2(1)] are satised. To verify
the condition ()(iv) of [Sh:f, Ch VI, 1.2(1)] suppose that h
, h) <
D,

F
h
, h
) and
h
, h) R
D,

F
T
w
,h
, w
= w
i
: i B
), B
T (for = 1, 2). Take n such that

(m n)(2 h(m) < h
(m)) (possible by 7.2.1(2)) and let B = B

1
B
2
[n, ) T.
Put w
i
= w
1
i
w
2
i
for i B and look at the tree T
w,h
. Clearly h
, h
) R
D,

F
T
w,h
and T
w
1
,h
T
w
2
,h
T
w,h
(so more than needed).
Checking clauses () and () of [Sh:f, Ch VI, 1.2(1)] we restrict ourselves to
the stronger condition (). So suppose that V
is a generic extension (via a proper

forcing notion) of V such that V
[= (D
D,

F
, R
D,

F
) covers.
(a) Assume that x, x
+
, x
n
dom(R
D,

F
), T
n
D
D,

F
are such that for n
x
n
<
D,

F
x
n+1
<
D,

F
x
+
<
D,

F
x and x
n
R
D,

F
T
n
.
Let x = h
, h), x
n
= h
, h
n
), x
+
= h
, h
+
) (so h
<
F
h
n
<
F
h
n+1
<
F
h
+
<
F
h)
and let w
n
= w
n
i
: i B
n
) D
D,

F
be such that T
n
= T
w
n
,h
(so B
n
T V
and [w
n
i
[ h
n
(i)). Look at the sequence B
n
: n ) T V. It does not
have to belong to V, but it may be covered by a countable set from V (as V
is a
proper forcing extension of V). Hence, as T is a p-lter, we nd a set B T such
that (n )([B B
n
[ < ). Take h
0
, h
1
T such that h
0
<
F
h
1
and (
n
)(h
1
(n)
h(n)
h
+
(n)
) (remember () of 7.2.1(2)) and choose an increasing sequence
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

m
0
< m
1
< m
2
< . . . < such that (i m
0
)(h
1
(i)
h(i)
h
+
(i)
), B B
0
m
0
and
for n :
B B
mn
m
n+1
and (i m
n+1
)((n + 2) h
mn
(i) < h
+
(i)).
Let
be such that if i B[m

n
, m
n+1
) then
h
+
(i)
((i)) = w
0
i

k<n
w
m
k
i
,
and let
be such that
h
0
(i)
(
(i)) = (i) for each i . Since

(D
D,

F
, R
D,

F
) covers in V
we nd T
w
,h
0
D
D,

F
such that
lim(T
w
,h
0
) and
h
0
, h
1
) R
D,

F
T
w
,h
0
. Let B
def
= B dom( w
) m
0
T. By the choice of ,
T
w
,h
0
and h
1
we nd w
= w
i
: i B
) D
D,

F
such that [w
i
[ h
1
(i) h
+
(i)
h(i) and w
0
i

k<n
w
m
k
i
w
i
whenever i B
[m
n
, m
n+1
), n . Clearly
T
w
,h
D
D,

F
and h
, h) R
D,

F
T
w
,h
. Put W
def
= m
0
, m
1
, m
2
, . . . and suppose
that
is such that for every n , `m

n+1

k<n
T
m
k
T
0
. If i B
,
m
n
i < m
n+1
then, by the assumptions on ,
h
(i)
((i)) w
0
i

k<n
w
m
k
i
w
i
.
Hence lim(T
w
,h
) (remember B
[m
0
, )).
(b) Assume that x = h
, h) dom(R
D,

F
),
n
,
are such that `n =

n
`n
for n . Take h
T such that h
<
F
h
<
F
h and choose an increasing sequence
0 = m
0
< m
1
< m
2
< . . . < such that for each n
(m m
n+1
)((n + 2) h
(m) < h
(m)).
Let
(in V
) be such that
if m [m
n
, m
n+1
), n , 0 < k < n + 1
then
h
(m)
(
m
k
(m))
h
(m)
(
(m))
(remember the choice of the m
n
s). Since (D
D,

F
, R
D,

F
) covers in V
we nd
w = w
i
: i B) such that h
, h) R
D,

F
T
w,h
(so in particular B T and
[w
i
[ h(i)) and
lim(T
w,h
). But now look at the tree T
w,h
. Clearly it satises
h
, h) R
D,

F
T
w,h
. Moreover, one can inductively show that
mn+1
lim(T
w,h
)
for each n . [Why? Plainly for each m B we have
h
(m)
(
m1
(m))
(m)
(
(m)) w
m
, so
m1
lim(T
w,h
). Looking at
mn+1
, n > 0, note
that
mn+1
`m
n
=
mn
`m
n
and for each m m
n
we have
h
(m)
(
mn+1
(m))
(m)
(
(m)).] Thus T
w,h
is as required, nishing the proof of the claim.
Finally, due to 7.2.3.1, 7.2.3.2 we may apply [Sh:f, Ch VI, 1.13A] to conclude
the theorem.
Theorem 7.2.4. Let T be a weakly non-reducible lter on , x
be an
unbounded non-decreasing function.
(1) If (K, ) is a nitary t-omittory treecreating pair then the forcing notion
Q
tree
1
(K, ) has the (T, x)strong PPproperty.
(2) If (K, ) is a nitary creating pair which captures singletons then the
forcing notion Q
w
(K, ) has the (T, x)strong PPproperty.
Proof. 1) Suppose that is a Q
tree
1
,
p Q
tree
1
(K, ). Choose a condition q p and fronts F
n
of T
q
such that for each
n
(1) if F
n
then the condition q
[]
decides the value of (n),
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

7.3. BOUNDED RELATIVES OF PP 153
(2) ( F
n
)(nor[t
q
] > n + 1),
(3) ( F
n+1
)(
F
n
)(
< )
(possible by 2.3.7(2), 2.3.5). Next choose an increasing sequence 0 = n
0
< n
1
<
n
2
< . . . < such that for each k
[
_
pos(t
q
) : F
n
k
[ < x(n
k+1
).
Since T is weakly non-reducible we nd Y []
such that B
def
=

k\Y
[n
k
, n
k+1
)
T. Now construct inductively a condition q
q such that root(q
) = root(q) and
(a) T
q
root(q)
pos(t
q
) : F
n
k
& k Y ,
(b) if T
q
then pos(t
q
) pos(t
q
) and nor[t
q
] nor[t
q
] 1 for some

kY
F
n
k
.
It should be clear that one can build such q
(remember (K, ) is t-omittory). Note

that if k
0
, k
1
Y , k
0
< k
1
and (k
0
, k
1
) Y then for each n (n
k0
, n
k1
] we have
[dcl(T
q
) F
n
[ = [dcl(T
q
)F
n
k
0
+1
[ [
_
pos(t
q
): F
n
k
0
[ < x(n
k0+1
) x(n).
For n B let w
n
= m : ( F
n
dcl(T
q
))(q
[]
' (n) = m). By the above
remark we have [w
n
[ x(n) (for n B) and clearly q
' (n B)( (n) w

n
).
2) Similar.
7.3. Bounded relatives of PP
In the following denition we introduce relations which determine localization
properties (see 0.2.2(2)) close to the PPproperty when restricted to functions from
n
f(n). Not surprisingly they include (the relation responsible for) the (f, g)
bounding property too.
Definition 7.3.1. Let f, g
be non-decreasing functions such that (n

)(0 < g(n) < f(n)). Dene:
(1) S
f,g
=

n
[f(n)]
g(n)
, S
f,g
= S
f,g
[]
,
(2) R
f,g
, R
f,g

n
f(n) S
f,g
are given by
R
f,g
A if and only if

n
f(n),

A = A
n
: n ) S
f,g
and (
n )((n) A
n
),
R
f,g
A if and only if

n
f(n),

A = A
n
: n ) S
f,g
and (
n )((n) A
n
),
(3) a relation R
n
f(n)
n
f(n) is dened by
0
R
f

1
if and only if
0
,
1

n
f(n) and (
n )(
0
(n) =
1
(n)),
(4) a relation R
f,g

n
f(n) S
f,g
is such that
R
f,g
(

A, K) if and only if

n
f(n),

A = A
n
: n ) S
f,g
, K = k
0
, k
1
, k
2
, . . . []
(the
increasing enumeration) and (m )(n [k
m
, k
m+1
))((n) A
n
).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Remark 7.3.2. (1) The spaces S
f,g
, S
f,g
and

n
f(n) carry natural (prod-
uct) Polish topologies.
(2) The relation R
f,g
corresponds to the (f, g)bounding property, of course.
The cardinal number d(R
f,g
) is the c(f, g) of [GoSh 448] (see there for
various ZFC dependencies between the cardinals determined by dierent
functions as well as for consistency results).
(3) Note that the relation R
f,g
(or actually the corresponding localization
property) is really very close to the PPproperty. The cardinal numbers
d(R
f
) and d(R
f,g
) appear naturally in [BRSh 616].
(4) There are other natural variants of relations introduced in 7.3.1. We
will deal with them (and the corresponding cardinal invariants) in the
continuation of this paper.
Below we list some obvious relations between the localization properties intro-
duced in 7.3.1 and the corresponding cardinal numbers.
Proposition 7.3.3. Let f, g, h
be non-decreasing functions such that

0 < g(n) < f(n) for each n . Then:
(1) The R
f,g
localization implies the R
f,g
localization and d(R
f,g
) d(R
f,g
).
(2) Suppose that for some increasing sequence m
0
< m
1
< m
2
< . . . < we
have
(n )(g(n) m
n+1
m
n
& f(n)
k[mn,mn+1)
h(k)).
Then the R
f,g
h
h
) d(R
f,g
).
(3) The R
f
f,g
f,g
) d(R
f
).
(4) The R
f,g
localization plus
bounding imply the R
f,g
localization. The
R
f,g
f,g
localization. Hence d(R
f,g
) d(R
f,g
)
maxd, d(R
f,g
).
(5) If g is unbounded then the strong PPproperty implies the R
f,g
localization,
and d(R
f,g
) d(R
sPP
).
(6) Assume ground model reals are not meager. Then the extension has the
R
f
localization property and thus d(R
f
) non(/).
(7) The R
f,g
localization implies that there is no Cohen real over the ground
model, and thus cov(/) d(R
f,g
).
For getting the R
f,g
localization (i.e. (f, g)bounding property) for forcing no-
tions built according to our schema see 5.1. Let us note that the other properties
appear naturally too.
Proposition 7.3.4. Let f, g
. Suppose that P is a forcing notion one of

the following type
Q
s
(K, ) for a nitary creating pair (K, ) which is either growing and
big or omittory and omittorybig,
Q
w
(K, ) for a nitary creating pair which captures singletons,
Q
tree
0
(K, ) for a nitary t-omittory treecreating pair (K, ).
Then P has the R
f,g
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

7.3. BOUNDED RELATIVES OF PP 155
Proof. It should be clear, so we will sketch the proof for the rst case only. Let
be a Q
s

n
f(n) and let p Q
s
(K, ). Using
2.2.3 or 2.2.6 construct a condition q Q
s
(K, ), an enumeration u
k
: k ) of
n
pos(w
q
, t
q
0
, . . . , t
q
n1
) and a sequence m
k
: k < ) such that
(1) p
0
q, m
0
< m
1
< . . . < ,
(2) if u
k
pos(w
q
, t
q
0
, . . . , t
q
n1
) then the condition (u
k
, t
q
n
, t
q
n+1
, . . .) decides
the value of (m
k
).
Plainly the construction is possible and easily it nishes the proof.
It may be not clear how one can preserve (in countable support iterations)
the localization properties introduced in 7.3.1. To deal with the R
f,g
localization
property we may adopt the approach of 7.2.3. It slightly changes the meaning of
this notion but the change is not serious and makes dealing with compositions much
easier.
Definition 7.3.5. Let h
0
, h
1
, f
be non-decreasing unbounded functions,

T be a lter on and

T = (T, <
F
) be a partial order on T

n
f(n). We say
that a proper forcing notion P:
(1) has the T-R
f,h0,h1
localization property if
'
P
for every

n
[f(n)]
h
0
(n)
there are B T V and w
i
: i B) V
such that (i B)([w
i
[ h
1
(i) & (i) w
i
) ,
(2) has the (T,

T)R
f
localization property if it has the T-R
f,h0,h1
localization
property for every h
0
, h
1
T such that h
0
<
F
h
1
.
Proposition 7.3.6. Suppose that T is a non-principal p-lter on , f
is non-decreasing unbounded and

T = (T, <
F
) is a PPok partial order on T
n
f(n) (see 7.2.1(2), except that it does not have to contain the identity function).
Let P
,

Q
: < ) be a countable support iteration such that for each <

'
P

Q
is a proper forcing notion which has

the (T,

T)R
f
Then the forcing notion P
has the (T,

T)R
f
Proof. Repeat the proof of 7.2.3 making suitable adjustments to the fact that
we are below the function f. No real changes are required.
Remark 7.3.7. Note that with no serious changes we may formulate and prove
a variant of 7.3.6 which would be an exact reformulation of 5.2.9 for the current
context.
Proposition 7.3.8. Let f, h
0
, h
1

be non-decreasing unbounded functions

such that (n )(1 h
0
(n) h
1
(n) f(n)) and lim
n
h1(n)
h0(n)
= . Assume T
is a non-reducible p-lter on . Suppose that (K, ) is a nitary, omittory and
omittorybig creating pair. Then the forcing notion Q
s
(K, ) has the T-R
f,h0,h1
Proof. Like 7.2.4 plus 7.3.4.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Remark 7.3.9. Note that if x(n)
h1(n)
h0(n)
for n then the (T, x)strong
PPproperty implies the T-R
f,h0,h1
localization property. Consequently we may
use 7.2.4 to get the conclusion of 7.3.8 for the two types of forcing notions specied
in 7.2.4.
7.4. Weakly non-reducible p-lters in iterations
One could get an impression that 7.2.3, 7.3.6 together with 7.2.4 and 7.3.8 are
everything we need: the properties involved are iterable and we may get them for
various forcing notions. However, to be able to make a real use of 7.2.4 or 7.3.8
we have to know that if we start with a weakly non-reducible p-lter and then we
iterate suitable forcing notions, the lter remains weakly non-reducible. One could
start with a p-point and consider forcing notions which are p-point preserving only.
However this is much too restrictive: we may iterate forcing notions mentioned in
7.2.4 and 7.3.8 and the iterations will preserve the fact that the lter is weakly
non-reducible. The rst step in proving this is the following observation.
Proposition 7.4.1. Suppose that T is a weakly non-reducible lter on . Let
P be an almost
-bounding forcing notion. Then

'
P
(The ler generated by) T is weakly non-reducible .
[Note that this covers
-bounding forcing notions.]

Proof. Suppose that

X
n
: n ) is a Pname for a partition of into nite
sets. Let

f be a Pname for a function in
such that
'
P
(n )(m )(

X
m
[n,

f(n))) .
Suppose p P. Since P is almost
-bounding we nd an increasing function

g
such that
(A []
)(q p)(q '

P
(
n A)(

f(n) < g(n)) ).
Let 0 = n
0
< n
1
< n
2
< . . . < be such that g(n
k
) < n
k+1
. As the lter T is
weakly non-reducible, we nd Y []
such that Z
def
=

k\Y
[n
k
, n
k+1
) T. Let
A = n
k
: k Y . By the choice of the function g, there is a condition q p such
that
q '
P
(
n A)(

f(n) < g(n)) .
Now look at the choice of

f necessarily
q '
P
(
m )(

X
m
Z = ) ,
which is enough to conclude the proposition.
Note that 7.4.1 captures almost all forcing notions mentioned in 7.2.4, 7.3.8.
So what is needed more is that T is weakly non-reducible is preserved at limit
stages of countable support iterations of proper forcing notions. This is done like
preserving unbounded families (i.e. by [Sh:f, Ch VI, 3]).
Theorem 7.4.2. Let T be a weakly non-reducible p-lter on . Suppose that
P
,

Q
: < ) is a countable support iteration of proper forcing notions such that

is limit and for each <
'
P
(The lter generated by) T is weakly non-reducible .
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

7.5. EXAMPLES 157
Then '
P
(The lter generated by) T is weakly non-reducible .

Proof. We will use [Sh:f, Ch VI, 3.13] and thus we will follow the notation
there. Let F
be the family of all increasing enumerations of elements of

T (i.e. F =
X
: X T, see 4.4.4). Let R be (a denition of) the following
twoplace relation on
:
g R f if and only if (g, f
and)
(
k)([n
g
k
, n
g
k+1
) rng(f) = ), where n
g
0
= 0, n
g
k+1
= n
g
k
+g(k) +1 for k .
As T is weakly non-reducible, the family F is Rbounding (i.e. (g
)(f
F)(g R f)).
Claim 7.4.2.1. (F, R) is Snice (see [Sh:f, Ch VI, 3.2]; here S [F]
is
arbitrary).
Proof of the claim: We have to show that for each N S there is g F such that
for each m
0
(the n
0
of [Sh:f, Ch VI, 3.2.3()] is irrelevant here) the second
player has an absolute winning strategy in the following game.
At the stage k of the game, Player I chooses f
k

and g
k

F N such that f
k
`m
+1
= f
`m
+1
for all 0 < k and f
k
R g
k
.
Then Player II answers playing an integer m
k+1
> m
k
.
Player II wins the game if (

k
f
k
`m
k
) R g.
But this is easy: let g F be such that rng(g)
rng(f) for all f F N

(remember T is a p-lter). Then
f
k
R g
k
implies (
)([n
f
k
, n
f
k
+1
) rng(g) = ).
Thus, at stage k of the game, the second player may choose m
k+1
> m
k
such that
[n
f
k
, n
f
k
+1
) rng(g) = for some (m
k
, m
k+1
).
As we iterate proper forcing notions, countable subsets of F from V
P
can be
covered by countable subsets of F from V. By our assumptions, F is Rbounding
in each V
P
(for < ) and it is nice there (like in the claim above). Consequently
we may apply [Sh:f, Ch VI, 3.13(3)] and we conclude that
'
P
F is Rbounding .
But this is exactly what we need.
7.5. Examples
Example 7.5.1. Let P 2
be a perfect set. We construct a nitary function

H
P
, an H
P
fast function f
P
: and a

2big trivially meagering simple
creating pair (K
P
7.5.1
,
P
7.5.1
) for H
P
with the (weak) Halving Property such that
'
Q
f
P
(K
P
7.5.1
,
P
7.5.1
)
there is a perfect set Q P such that
(K []
V)(Q`K ,= 2
K
) .
Construction. The creating pair (K
P
7.5.1
,
P
7.5.1
) will be constructed in a way
slightly similar to (K
1
2.4.12
,
1
2.4.12
). For positive integers i, m let R
i
2
(2, m) be the
minimal integer k such that for every function :

<i
[k]
2
2 there are sets
a
0
, . . . , a
i1
[k]
m
such that `[a
0
]
2
. . . [a
i1
]
2
is constant. [Thus this is the
Ramsey number for polarized partition relations; R
1
2
(2, m) is essentially R
2
(2, m)
of 6.3.5.]
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Dene inductively Z
i
2
(k) for k by Z
i
2
(0)=R
i
2
(2, 4), Z
i
2
(k+1)=R
i
2
(2, 2Z
i
2
(k)),
and for a nite set X let
H
i
(X)
def
= mink : [X[ Z
i
2
(k).
Let T 2
<
be a perfect tree such that P = [T]. Now construct inductively
functions H
P
= H and f
P
= f and an increasing sequence n = n
i
: i ) such
that
(i) f(0, ) = + 1, f(k + 1, ) = 2
H()+1
(f(k, ) +
H
() + 2) (compare
1.1.12),
(ii) n
0
= 0, n
i+1
is the rst such that for every T 2
n
i
H
2
i ( T 2
n
i+1
: < ) 2
f(i,i)
,
(iii) H(i) is the family of all non-empty subsets of T 2
n
i+1
.
It should be clear that the clauses (i)(iii) uniquely determine H, f and n.
Call a sequence u

i<m
H(i) acceptable if for each i
0
< i
1
< m
[u(0)[ = 2, u(i
0
) = `n
i0+1
: u(i
1
), and
( u(i
0
))([ u(i
0
+ 1) : < [ = 2).
Note that if W

m
H(m) is such that each W`m is acceptable then the sequence
W determines a perfect tree
T
W
def
= T : (m )( W(m))( _ ) T
with the property that [T
W
2
n
i
[ = 2
i
and each node from T
W
2
n
i
has a
ramication below n
i+1
.
A creature t CR[H] is in K
P
7.5.1
if m
t
up
= m
t
dn
+ 1 = i + 1 and
dis[t] = m
t
dn
, B
t
: T 2
n
i
), r
t
), where r
t
is a non-negative real and
B
t
T 2
n
i+1
: < (for all T 2
n
i
; remember i = m
t
dn
),
val[t] = u, v)

k<m
t
dn
H(k)

km
t
dn
H(k) : u < v both are acceptable
and
if v(m
t
dn
) then B
t
ni
,
nor[t] = max0, minH
2
i (B
t
) : T 2
n
i
r
t
.
The operation
P
7.5.1
is dened by
P
7.5.1
(t) = s K
P
7.5.1
: m
t
dn
= m
s
dn
& ( T 2
n
i
)(B
s
B
t
) & r
s
r
t
.
P
7.5.1
,
P
7.5.1
) is a simple nitary creating pair and the
forcing notion Q
f
(K
P
7.5.1
,
P
7.5.1
) is not trivial.
To check that (K
P
7.5.1
,
P
7.5.1
) is

2big suppose that t K
P
7.5.1
, nor[t] > 1, u
basis(t) and c : pos(u, t) 2 (note that basis(t) = dom(val[t]) and pos(u, t) =
v rng(val[t]) : u < v). Then u is acceptable and, if g(u) > 0, [u(g(u) 1)[ =
2
g(u)
. Let i = g(u) = m
t
dn
. Let k = minH
2
i (B
t
) : u(i 1). By the

denition of the norm of t we know that k nor[t] + r
t
> 1, so necessarily k 2.
Note that under natural interpretation

u(i1)
[B
t
]
2
pos(u, t), so we may restrict
our colouring c to this set and use the denition of H
2
i (and the choice of k). Thus
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

7.5. EXAMPLES 159
we nd sets B
B
t
(for u(i 1)) such that

( u(i 1))([B
[ = 2 Z
2
i
2
(k 2)) and c`(
u(i1)
B
) is constant.
Note that H
2
i (B
) k 1 nor[t] 1 + r
t
> r
t
. Let s K
P
7.5.1
be a creature
determined by
m
s
dn
= m
t
dn
, r
s
= r
t
, B
s
= B
if u(i 1), and B

s
= B
t
otherwise.
Clearly s
P
7.5.1
(t), nor[s] nor[t] 1 and c`pos(u, s) is constant.
Plainly (K
P
7.5.1
,
P
7.5.1
) is trivially meagering, as if x, y X then
H
i
(X x, y) H
i
(X) 1.
Let half : K
P
7.5.1
K
P
7.5.1
be such that if nor[t] 2 then
dis(half(t)) = m
t
dn
, B
t
: T, g() = n
m
t
dn
), r
t
+
1
2
nor[t]),
and half(t) = t otherwise. Exactly like in 2.4.12 one checks that the function half
witnesses the fact that (K
P
7.5.1
,
P
7.5.1
) has the (weak) Halving Property.
To show the last assertion of 7.5.1 we prove that
'
Q
f
(K
P
7.5.1
,
P
7.5.1
)
(K []
V)([T

W
]`K ,= 2
K
) ,
where

W is the name for the generic real (see 1.1.13) and T

W
is the tree dened
before. To this end suppose that p Q
f
(K
P
7.5.1
,
P
7.5.1
) and K []
. We may
assume that g(w
p
) = j
0
> 0 and (i )(nor[t
p
i
] > f
P
(0, m
t
p
i
dn
) 2). Choose
j
1
such that [[n
j0
, n
j1
) K[ 2
j0
and x one-to-one mapping
k : w
p
(j
0
1) [n
j0
, n
j1
) K : k()
(remember that w
p
is acceptable, so [w
p
(j
0
1)[ = 2
j0
and w
p
(j
0
1) T 2
n
j0
).
Fix w
p
(j
0
1) for a moment. Let i() = i < j
1
j
0
be such that
k() [n
j0+i
, n
j0+i+1
). For each T 2
n
j0+i
such that _ choose c
j0+i
() 2
such that the set B
def
= B
t
p
i
: (k()) = c
j0+i
() has at least
1
2
[B
t
p
i
[ elements
(so then H
2
j
0
+i (B
) H
2
j
0
+i (B
t
p
i
) 1). If = then we nish the procedure.

Otherwise, for each T 2
n
j0+i1
such that _ we choose c
j0+i1
() 2 such
that the set B
def
= B
t
p
i1
: c
j0+i1
() = c
j0+i
() has at least
1
2
[B
t
p
i1
[ elements
(and so H
2
j
0
+i1(B
) H
2
j
0
+i1(B
t
p
i1
) 1). Continuing this procedure downward

till we arrive to we determine sets B
: _ T 2
n
j
, j
0
j j
0
+i()) and
c
j0
() 2 such that
()
if _ T 2
n
j
, j
0
j j
0
+i()
then B
B
t
p
jj
0
and H
2
j (B
) H
2
j (B
t
p
jj
0
) 1,
()
if _ T 2
n
j1
is such that (j [j
0
, j
0
+i()))(`n
j+1
B
nj
)
then (k()) = c
j0
().
For each i < j
1
j
0
choose a creature s
i

P
7.5.1
(t
p
i
) such that r
si
= r
t
p
i
and
for every T 2
n
j0+i
if i(`n
j0
) i then B
si
= B
nj
0
, otherwise B
si
= B
t
p
i
.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Clearly nor[s
i
] nor[t
p
i
] 1 and thus q = (w
p
, s
0
, . . . , s
j1j01
, t
p
j1j0
, . . .) is a
condition in Q
f
P
(K
P
7.5.1
,
P
7.5.1
) stronger than p. Let : K [n
j0
, n
j1
) 2 be
such that (k()) = 1 c
j0
() for each w
p
(j
0
1). Note that
u pos(w
p
, s
0
, . . . , s
j1j01
) ( u(j
1
1))(`(K [n
j0
, n
j1
) ,= )),
what nishes the proof.
Conclusion 7.5.2. It is consistent that
BP
= non(/) = d(R
sPP
) = c =
2
and d =
1
.
Proof. Start with a model for CH and build inductively (with a suitable
bookkeeping) a countable support iteration P
,

Q
: <
2
) and a sequence
: <
2
) such that
()

P
: <
2
) lists with
2
repetitions all P
2
names for perfect subsets
of 2
; each

P
is a P
name,
()

Q
is a P
name for the forcing notion Q
f

P
(K

P
7.5.1
,

P
7.5.1
).
By 2.2.12 and 3.1.2 we know that each

Q
is a (name for) proper
bounding
forcing notion and hence '
P
2
d =
1
. By 3.2.8(2) we easily conclude that '
P
2
non(/) =
2
and nally we note that by the last property of Q
f
P
(K
P
7.5.1
,
P
7.5.1
)
stated in 7.5.1, and by the choice of

P
: <
2
), we have '
P
2

BP
=
2
. To
nish remember 7.1.3.
Conclusion 7.5.3. It is consistent that
BP
= d(R
sPP
) =
1
and non(^) =
c =
2
.
Proof. Force over a model of CH with countable support iteration,
2
in
length, of forcing notions Q
w
(K
5.4.3
,
5.4.3
).
By 7.2.4(2) we know that the forcing notion Q
w
(K
5.4.3
,
5.4.3
) has the (T, x)-
strong PPproperty for any weakly non-reducible lter T on and an unbounded
non-decreasing x
. Consequently, if (in V) we take a p-point T and a PP

ok partial order

T then the iteration will have the (T,

T)-strong PPproperty, so
in particular the strong PPproperty (by 7.2.3; remember that by 7.4.1+7.4.2 the
lter generated by T in the intermediate universes is weakly non-reducible). Hence,
in the resulting model we have d(R
sPP
) =
1
and thus
BP
=
1
(by 7.1.3). Finally,
it follows from 5.4.4 that in this model non(^) = c =
2
.
Note that one can use the forcing notion Q
tree
1
(K
1
5.4.5
,
1
5.4.5
) of 5.4.5 as well.
Let us recall the following notions from [Sh 326].
Definition 7.5.4. Let T
and g
.
(1) We say that the family T is gclosed if
(f T)(f
+
, f
T)(
n )(f(n)
g(n)
f
(n) &
m<n
(f(m) + 1) f
+
(n)).
(2) We say that a proper forcing notion P has the (T, g)bounding property
if it has the (f, g
)bounding property for each > 0 and f T.

These notions are important when we want to iterate (f, g)bounding forcing
notions: if T is a gclosed family then each countable support iteration of proper
(T, g)bounding forcing notions is (T, g)bounding (see [Sh 326, A2.5], compare
to 5.2.9).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

7.5. EXAMPLES 161
Proposition 7.5.5. Suppose that T
is a gclosed family and
is
an increasing function. Let g
= g and let T
= f : f T. Then T
is
g
closed.
Proof. Check.
Conclusion 7.5.6. Let g(n) = n
n
for n and let T
be a countable
gclosed family. Suppose that F
is an increasing function which dominates

all elements of T (i.e. (f T)(
n )(f(n) < F(n))) and let H = H

F
, f = f
F
(and
H
) be as dened in 2.4.6 for F. Next, let f
0
T and m
k
: k )
and h
be such that m
0
= 0, m
k+1
= m
k
+
H
(k)
H(k)
, h is non-decreasing
and h(m
k+1
) f
0
(
H
(k)) (for k ). Assume that P
2
is the countable support
iteration of the forcing notions Q
f
(K
2.4.6
,
2.4.6
). Then
'
P
2
d(R
H
) =
2
& d = d(R
h
) = d(R
f0H,gH
) =
1
.
Proof. We know that T
H
= f

H
: f
T is g
H
closed, where
g
H
= g
H
. By 5.4.6 and the choice of g, F we have that the forcing no-
tion Q
f
(K
2.4.6
,
2.4.6
) is proper,
bounding, (T
H
, g
H
)bounding and so is the
iteration. Hence, for each f
1
T, '
P
2
d(R
f1H,gH
) = d =
1
. Next note
that for the function h dened in the assumptions and for suciently large k we
have
n[m
k
,m
k+1
)
h(n) h(m
k+1
)
H(k)
((f
0

H
)(k))
H(k)
(f
0

H
)(k),
where f
0
T is such that (
n )(f
0
(n)
g(n)
f
0
(n)). Use 7.3.3(2) to conclude
that '
P
2
d(R
h
) =
1
. Finally, note that if

W is the name for the generic real
then
'
Q
f
(K2.4.6,2.4.6)
(x
n
H(n) V)(
n )(

W(n) ,= x(n))
and therefore '
P
2
d(R
H
) =
2
.
Conclusion 7.5.7. It is consistent that non(/) = d =
2
and cov(/) =
b =
1
= d(R
f,g
) for every non-decreasing unbounded g
and any f
such that lim

n
f(n)
g(n)
= .
Proof. Start with a model of CH and iterate
2
times with countable sup-
port the Blass-Shelah forcing notion Q
s
(K
2.4.5
,
2.4.5
). By 4.4.1 we immediately
conclude that the iteration forces non(/) = d =
2
& b =
1
. As each
function in
appears in an intermediate model we may restrict our attention to

f, g
V. By 7.3.8 and 7.3.6 we conclude that the iteration has the R
f,g
localization property (just build a suitable PPok partial order

T and take any
p-point T V; by 7.4.2, 7.4.1 and 4.4.1 we know that T generates a non-reducible
p-lter in the intermediate universes). Hence we get that in the resulting model
cov(/) = d(R
f,g
) =
1
.
Example 7.5.8. We construct a nitary 2big treecreating pair (K
7.5.8
,
7.5.8
)
of the NMPtype (see 3.2.3(2)) such that the forcing notion Q
tree
1
(K
7.5.8
,
7.5.8
)
does not have the strong PPproperty.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Construction. This example is similar to that of 6.4.4 (what is not surprising
if you notice some kind of duality between m
1
and d(R
sPP
)).
Let H(n) = n
n
. Let / be the family of all pairs (n, x) such that x [H(n)]
n
.
For

k<m
H(k), m
0
< m and A / we will write
m0
A if
((n, x) A)(m
0
n < m & (n) x).
Now we dene (K
7.5.8
,
7.5.8
). A treelike creature t TCR
[H] is taken to be in
K
7.5.8
if:
val[t] is nite, and
nor[t] = log
2
(min[A[ : A / & ( rng(val[t]))(
g()
A)).
7.5.8
is dened like in 6.4.4: if t
:

T) K
7.5.8
is
a system such that T is a well founded quasi tree, root(t
) = , and rng(val[t
]) =
succ
T
() (for

T) then we dene S
(t
:

T) as the unique creature t
in K
7.5.8
with rng(val[t
]) = max(T), dom(val[t
]) = root(T) and dis[t
] = dis[t
] :
T). Next we put
7.5.8
(t
:

T) = t K
7.5.8
: val[t] val[S
(t
:

T)].
7.5.8
,
7.5.8
) is a nitary treecreating pair and the forcing
notion Q
tree
1
(K
7.5.8
,
7.5.8
) is non-trivial.
Claim 7.5.8.1. (K
7.5.8
,
7.5.8
) is 2big.
Proof of the claim: Let t K
7.5.8
, nor[t] > 0 and suppose that pos(t) = u
0
u
1
.
Let s

7.5.8
(t) be such that pos(s
) = u
(for = 0, 1). Take A
/ such that
[A
[ = 2
nor[s
]
and ( pos(s
))(
m0
A
),
where m
0
= g(root(s
)) = g(root(t)). Clearly ( pos(t))(
m0
A
0
A
1
) and
thus
nor[t] log
2
([A
0
[ +[A
1
[) 1 + maxnor[s
0
], nor[s
1
].
Claim 7.5.8.2. (K
7.5.8
,
7.5.8
) is of the NMPtype (see 3.2.3(2)).
Proof of the claim: Suppose that t
: T) Q
tree
(K
7.5.8
,
7.5.8
) is such that
( T)(nor[t
] > 1) and F
0
, F
1
, F
2
, . . . are fronts of T such that
( F
i+1
)(
F
i
)(
< ).
Clearly these fronts are nite (as (K
7.5.8
,
7.5.8
) is nitary). Further suppose that
g :

i
F
i

i
F
i+1
is such that < g() F
i+1
for F
i
.
Let r = min2
nor[t]
: T. Choose increasing sequences n
k
: k r
r
+ r),
i
k
, j
k
: k < r
r
+r) such that for each k < r
r
+r:
(i) i
k
< j
k
< i
k+1
and if k [r, r +r
r
) then j
k
= i
k
+ 1,
(ii) if k < r, F
i
k
then [ F
j
k
: < [ r,
(iii) n
k
< ming() : F
i
k
< maxg() : F
j
k
< n
k+1
.
Choose a mapping : F
jr1
r
r
such that
()
0
if <
0
F
jr1
, <
1
F
jr1
, F
j
k
, k < r
then (
0
)(k) = (
1
)(k),
()
1
for each F
i
k
, k < r and < r there is F
jr1
such that < and
()(k) = .
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

7.5. EXAMPLES 163
(It is easy to dene such a mapping if you remember clause (ii) above.) Let
:
r
r
r
r
be the isomorphism of r
r
equipped with the lexicographical order and
r
r
with the natural order of integers. Take a treecreature s
7.5.8
(t
: (
F
j
r
r
+r1
)( < )) such that
rng(val[s]) = F
j
r
r
+r1
: if `m
0
F
jr1
, m
0
and k =
((`m
0
))
and `m
1
F
i
r+k
, m
1
then g(`m
1
) < .
(Note that we may nd a suitable s by the denition of
7.5.8
.) By the choice, this
s satises the demand ()
tree
of 3.2.3(2). But why does it have large enough norm?
Suppose that A / is such that [A[ < r. Let k
0
< r be such that
(n, x) A n / [n
k0
, n
k0+1
).
Since [A[ < 2
nor[t]
for each T we may inductively build a sequence
0
F
i
k
0
such that
((n, x) A)(g(root(T)) n < g(
0
)
0
(n) / x).
Let
0
: k
0
r be such that
0
(k) is the value of ()(k) for each F
jr1
,
0
< . Take < r such that
if r
r
,
0
) _
then there is no (n, x) A with n
()+r
n < n
()+r+1
(remember the choice of
and that [A[ < r). Now take

1
F
j
k
0
such that
0
<
1
and
( F
jr1
)(
1
_ ()(k
0
) = )
(possible by ()
0
+ ()
1
). By the choice of k
0
we know that
((n, x) A)(g(root(T)) n < g(
1
)
1
(n) / x)
(look at (iii)). Next continue like at the beginning to get F
jr1
such that
1
_
and (
g(root(T))
A). We are sure that
0
) _ () and therefore there is no

(n, x) A with n
(())+r
n < n
(())+r+1
. Consequently we may continue
the procedure applied to build and we construct
F
j
r
r
+r1
such that
<
, if
`m F
i
r+
(())
then g(
`m) <
, and (
g(root(T))
A).
Since, by its construction, the sequence
is in rng(val[s]), it exemplies that A

cannot witness the minimum in the denition of nor[s]. Consequently, nor[s]
log
2
(r) and thus the treecreature s satises the demand ()
tree
of 3.2.3(2).
tree
1
(K
7.5.8
,
7.5.8
) does not have the strong
PPproperty.
Proof of the claim: We will show that the generic real

W shows that the strong
PPproperty fails for Q
tree
1
(K
7.5.8
,
7.5.8
). So suppose that w
i
: i B) is such that
B []
and (i B)(w
i
[]
i
) and let p Q
tree
1
(K
7.5.8
,
7.5.8
). We may assume
that nor[t
p
] > 1 for each T

p
. Take i
0
B g(root(p)) and build inductively
a condition q p such that for each T
q
and pos(t
q
)
t
q

7.5.8
(t
p
) and nor[t
q
] nor[t
p
] 1 and (i
0
< g() (i
0
) / w
i0
)
(remember the denition of the norm of elements of K
7.5.8
). Now clearly q '
W(i
0
) / w
i0
, nishing the proof of the claim and the construction.
Conclusion 7.5.9. It is consistent that cof (/) < d(R
sPP
).
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Proof. Start with a model of CH and force with countable support iteration of
length
2
of forcing notions Q
tree
1
(K
7.5.8
,
7.5.8
). We know that Q
tree
1
(K
7.5.8
,
7.5.8
)
is proper,
bounding and Cohenpreserving (by 3.2.5 + 3.1.1). Consequently

the iteration is of the same type (see [BaJu95, 6.3.21, 6.3.22]) and, by standard
arguments, in the nal model we have non(/) = d =
1
. But this implies that
cof (/) =
1
too (see [BaJu95, 2.2.11]). Finally, as Q
tree
1
(K
7.5.8
,
7.5.8
) does
not have the strong PPproperty we easily conclude that the iteration forces that
d(R
sPP
) =
2
.
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

List of denitions
1.1.1 weak creatures, WCR[H];
1.1.3 nitary H, nitary K;
1.1.4 sub-composition operation, weak creating pair, the relation
;
1.1.6 basis basis(t), possibilities pos(w, o);
1.1.7 forcing notion Q
C(nor)
(K, ) (for a weak creating pair (K, ) and a
norm condition ((nor));
1.1.10 m
dn
(t), norm conditions and corresponding forcing notions Q
s
(K, ),
Q
(K, ), Q
w
(K, ), Q
f
(K, ), Q
(K, );
1.1.12 fast function, H-fast function f : ;
1.1.13 name for the generic real

W;
1.2.1 m
t
dn
, m
t
up
, creatures, CR[H];
1.2.2 composition operation on K, creating pairs (K, );
1.2.4 nite candidates FC(K, ), pure nite candidates PFC(K, ), pure can-
didates PC(K, ), ((nor)normed pure candidates PC
C(nor)
(K, ) and
partial orders on them;
1.2.5 creating pairs which are: nice, smooth, forgetful, full;
1.2.6 forcing notions Q
C(nor)
(K, ) for creating pairs (K, );
1.2.9 when a condition p essentially decides a name , approximates ;
1.2.11 partial orders
apr
,
s
n
,
n
,
w
n
,
f
n
;
1.3.1 quasi trees, well founded quasi trees, downward closure dcl(T), succes-
sors succ
T
() of in T, T
[]
, split(T), max(T),

T, lim(T), fronts of a quasi
tree T;
1.3.3 treecreatures, TCR[H], treecomposition, bounded treecomposition;
1.3.5 forcing notions Q
tree
e
(K, ) for e < 5, Q
tree
(K, ), condition p
[]
for
p Q
tree
e
(K, ), T
p
;
1.3.7 e-thick antichains in T
p
for p Q
tree
e
(K, );
1.3.10 partial orders
e
n
(for e < 3);
1.4.3 local weak creating pairs;
2.1.1 creature t [m
0
, m
1
), for a creating pair (K, ) we say when it is
omittory, growing;
2.1.7 creating pairs which are: gluing, simple;
2.1.10 creating pairs which capture singletons;
2.2.1 big creating pairs;
2.2.5 omittorybig creating pairs;
2.2.7 Halving Property and weak Halving Property;
2.3.2 big treecreating pairs;
2.3.4 tomittory tree creating pairs;
2.4.1 prenorm on T(A), nice pre-norm;
2.4.2 prenorms dp
i
, dp
i
n
(for i < 3, n );
165
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

166 LIST OF DEFINITIONS
3.2.1 Cohenpreserving proper forcing notions;
3.2.3 creating pairs of the NMPtype, tree creating pairs of the NMP
tree
type;
3.2.7 trivially meagering weak creating pairs;
3.3.1 weak creating pairs of the NNPtype;
3.3.2 gluing and weakly gluing tree creating pairs;
3.3.4 strongly nitary creating pairs;
3.4.1 when a weak creating pair (K, ) strongly refuses Sacks property;
4.1.2 creating pairs which are: meagering, antibig;
4.2.1
sum
;
4.2.3 (d, u)sum
sum
d,u
;
4.2.4 when a creating pair is saturated with respect to a family of prenorms;
4.2.6
tsum
;
4.3.1 decision functions, creating pairs of the AB-type, condensed creating
pairs;
4.3.7 creating pairs of the AB
+
type;
5.1.1 essentially f-big weak creating pairs;
5.1.6 reducible weak creating pairs;
5.1.7 hlimited weak creating pairs;
5.1.11 (H, F)fast function;
5.2.1 U
h
(
t), V
n
h
(
t), (
t, h
1
, h
2
)bounding forcing notions;
5.2.3 creating pairs which are monotonic, strictly monotonic, spread;
5.2.5 madditivity add
m
(t) of a weak creature t, (g, h)-additive weak creating
pairs;
5.2.8

tgood families of functions, (
t, T)bounding forcing notions;

5.3.1

tsystems, regular

tsystems, P
C(nor)
(
t, (K, )) and the partial order _

on it, quasi-W-generic ;
5.3.6 (, W)genericity preserving forcing notions;
5.3.8 Cohen sensitive

tsystems, directed

tsystems;
5.3.10 (
t
0
, h
1
, h
2
)coherent

tsystems, (
t
0
,

T)coherent sequences of

tsystems;
6.1.1 creating pair which generates an ultralter;
6.1.3 when generates a lter (ultralter), T();
6.1.5 interesting creating pair;
6.2.1 Ramsey lter, ppoint, qpoint, weak qpoint;
6.2.5 interesting ultralters, games G
sR
(T), G
aR
(T), semiRamsey ultral-
ters, almost Ramsey ultralters;
6.3.1 tree creating pairs of the UP(T)
tree
, sUP(T)
tree
types;
6.3.3 rich tree creating pairs;
6.3.5 R
n
(k, m);
6.3.7 simple except omitting creating pairs, omittorycompatible

tsystems;
7.1.1
BP
;
7.1.2 R
sPP
;
7.2.1 non-reducible lters, PPok partial orders, forcing notions with (T, x)-
strong PPproperty, (T,

T)-strong PPproperty;
7.3.1 R
f,g
, R
f,g
, R
f
, R
f,g
;
7.3.5 (T,

T)-R
f
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

Bibliography
[Ba91] Marek Balcerzak. Some properties of ideals of sets in Polish spaces. Habilitation
Thesis. Lodz University Press, 1991.
[BaPl96] Marek Balcerzak and Szymon Plewik. Property (P) and game ideals. Tatra Moun-
tains Mathematical Publications, 8:153158.
[BaRo95] Marek Balcerzak and Andrzej Ros lanowski. Coinitial families of perfect sets. Journal
of Applied Analysis, 1:181204, 1995.
[Ba94] Tomek Bartoszy nski. Some old and new problems. Circulated notes of November
1994.
[BaJu95] Tomek Bartoszy nski and Haim Judah. Set Theory: On the Structure of the Real Line.
A K Peters, Wellesley, Massachusetts, 1995.
[BJSh 368] Tomek Bartoszy nski, Haim Judah, and Saharon Shelah. The Cicho n diagram. Journal
of Symbolic Logic, 58:401423, 1993.
[BRSh 616] Tomek Bartoszy nski, Andrzej Ros lanowski, and Saharon Shelah. After all, there are
some inequalities which are provable in ZFC. Journal of Symbolic Logic, submitted.
[BsSh 242] Andreas Blass and Saharon Shelah. There may be simple P
1
- and P
2
-points and
the Rudin-Keisler ordering may be downward directed. Annals of Pure and Applied
Logic, 33:213243, 1987.
[CRSW93] Jacek Cicho n, Andrzej Ros lanowski, Juris Steprans, and Bogdan Weglorz. Combina-
torial properties of the ideal P
2
. Journal of Symbolic Logic, 58:4254, 1993.
[CiSh 653] Krzysztof Ciesielski and Saharon Shelah. A model with no magic sets. Preprint, 1997.
[DKP92] Wojciech Debski, Jan Kleszcz, and Szymon Plewik. Perfect sets of independent func-
tions. Acta Universitatis Carolinae Mathematica et Physica, 33:3133, 1992.
[FrSh 406] David H. Fremlin and Saharon Shelah. Pointwise compact and stable sets of measur-
able functions. Journal of Symbolic Logic, 58:435455, 1993.
[GoSh 448] Martin Goldstern and Saharon Shelah. Many simple cardinal invariants. Archive for
Mathematical Logic, 32:203221, 1993.
[J] Thomas Jech. Set theory. Academic Press, New York, 1978.
[JRSh 373] Haim Judah, Andrzej Ros lanowski, and Saharon Shelah. Examples for Souslin Forc-
ing. Fundamenta Mathematicae, 144:2342, 1994.
[MaPa9x] Pierre Matet and Janusz Pawlikowski. Ideals over and cardinal invariants of the
continuum. Preprint, 1997.
[My69] Jan Mycielski. Some new ideals of sets on the real line. Colloquium Mathematicum,
XX:7176, 1969.
[NeRo93] Ludomir Newelski and Andrzej Ros lanowski. The ideal determined by the unsymmet-
ric game. Proceedings of the American Mathematical Society, 117:823831, 1993.
[Ro94] Andrzej Ros lanowski. Mycielski ideals generated by uncountable systems. Colloquium
Mathematicum, LXVI:187200, 1994.
[RoSh 670] Andrzej Ros lanowski and Saharon Shelah. Norms on possibilities III: strange subsets
of the real line. in preparation.
[RoSh 672] Andrzej Ros lanowski and Saharon Shelah. Norms on possibilities IV: ccc forcing no-
tions. in preparation.
[RoSh 501] Andrzej Ros lanowski and Saharon Shelah. Localizations of innite subsets of .
Archive for Mathematical Logic, 35:315339, 1996. A special volume dedicated to
Prof. Azriel Levy.
[RoSh 628] Andrzej Ros lanowski and Saharon Shelah. Norms on possibilities II: more ccc ideals
on 2
. Journal of Applied Analysis, 3:103127, 1997.

167
4
7
0

r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
6
-
2
1

m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
6
-
2
1

168 BIBLIOGRAPHY
[Sh 630] Saharon Shelah. Non-elementary proper forcing notions. Journal of Applied Analysis,
submitted.
[Sh 669] Saharon Shelah. Non-elementary proper forcing notions II: iterations. in preparation.
[Sh:f] Saharon Shelah. Proper and improper forcing. Perspectives in Mathematical Logic.
Springer-Verlag, 1998.
[Sh:b] Saharon Shelah. Proper forcing, volume 940 of Lecture Notes in Mathematics.
Springer-Verlag, Berlin-New York, xxix+496 pp, 1982.
[Sh 207] Saharon Shelah. On cardinal invariants of the continuum. In Axiomatic set theory
(Boulder, Colo., 1983), volume 31 of Contemp. Mathematics, pages 183207. Amer.
Math. Soc., Providence, RI, 1984. Proceedings of the Conference in Set Theory, Boul-
der, June 1983; ed. Baumgartner J., Martin, D. and Shelah, S.
[Sh 326] Saharon Shelah. Viva la dierence I: Nonisomorphism of ultrapowers of countable
models. In Set Theory of the Continuum, volume 26 of Mathematical Sciences Re-
search Institute Publications, pages 357405. Springer Verlag, 1992.

Andrzej Roslanowski and Saharon Shelah - Norms On Possibilities I: Forcing With Trees and Creatures

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

Andrzej Roslanowski and Saharon Shelah - Norms On Possibilities I: Forcing With Trees and Creatures

Uploaded by

Copyright:

Available Formats

4

-bounding, preserves the outer measure,

-bounding forcing no-

tsystems, we say when is quasi-W-generic and when a forcing notion is

-bounding, (f, g)-bounding,

we give a treecreating pair (K, ) such that the forcing notion Q

we build a creating pair (K, )

-bounding and adds

bounding and preserves non-

which is translationinvariant indexinvariant

bounding, preserves non-meager sets and the outer

bounding forcing notion which preserves non-meager sets,

n) are abbreviations for

and the Baire space

are the spaces of all functions

in the ordering of eventual

are models of ZFC and that all parameters we

) has the R-localization

)(y Y V)((x, y) R).

, y Y V and (x, y) R then we say that y R-localizes x.

extending w. Next we choose a family o

is obtained by permitted procedures from some

) is an extension of (w, o) provided o

satises the structure demands and norm requirements.

T(K) is a sub-composition operation on K if:

and for each s o we have s (o

as dened in 1.1.4(3) does not have

is an equivalence relation. Then, if additionally (o) is non-

equivalent elements of K (or just consider a selector K

). If (o) may be non-empty for an innite o K (which may happen

we dene the set pos(w, o) of

(w, o) = u : (s (o))(w, u) val[s]),

). We dene the forcing notion Q

(nor) if and only if

) which is not a legal condition. This is not

(K, ) (for n < ) are dened by:

[H] = t TCR[H] : dom(val[t]) = .

[H] and rng(val[s

is a well founded quasi tree,

) is a system for which may be non-empty, i.e. it satises

[H] then basis(t) =

[H] K and pos(t

: T) then we write root(p) = root(T), T

< & nor[t

(A) + 1 then for each F S

as above. One can do it for

m)) : m dis[t] and nor[t] = h

then the unique treecreature t TCR

are pairwise disjoint so we may naturally order them according to

: m < . Let f : [g(), ) . We

s are odd. Of course there is still

and the requirement () we have

(for n ) be mappings dened in the following manner. Let f : [n, ) ;

] + 2 and next choose k

(K, ) collapses c onto as well. This is the reason why we have

(K, ) is used only for growing (K, )

(K, ), see 2.1). This bad eect can be made

(K, ) collapses c onto ,

(K, )name dened in 1.1.13. To this end suppose that r 2

(K, ). We may assume that g(w

(K, ) and it is stronger than p.

' ,= . We may assume that