9

5
7


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







m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
3
-
2
3


PARTITION THEOREMS FROM CREATURES AND
IDEMPOTENT ULTRAFILTERS
ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
Abstract. We show a general scheme of Ramsey-type results for partitions
of countable sets of nite functions, where one piece is big is interpreted in
the language originating in creature forcing. The heart of our proofs follows
Glazers proof of the Hindman Theorem, so we prove the existence of idempo-
tent ultralters with respect to suitable operation. Then we deduce partition
theorems related to creature forcings.
0. Introduction
A typical partition theorem asserts that if a set with some structure is divided
into some number of nice pieces, then one of the pieces is large from the point
of the structure under considerations. Sometimes, the underlying structure is com-
plicated and it is not immediately visible that the arguments in hands involve a
partition theorem. Such is the case with many forcing arguments. For instance,
the proofs of propernes of some forcing notions built according to the scheme of
norms on possibilities have in their hearts partition theorems stating that at some
situations a homogeneous tree and/or a sequence of creatures determining a con-
dition can be found (see, e.g., Roslanowski and Shelah [6, 7], Roslanowski, Shelah
and Spinas [8], Kellner and Shelah [5, 4]). A more explicit connection of partition
theorems with forcing arguments is given in Shelah and Zapletal [9].
The present paper is a contribution to the Ramsey theory in the context of ni-
tary creature forcing. We are motivated by earlier papers and notions concerning
norms on possibilities, but we do not look at possible forcing consequences. The
common form of our results here is as follows. If a certain family of partial nite
functions is divided into nitely many pieces, then one of the pieces contains all
partial functions determined by an object (a pure candidate) that can be inter-
preted as a forcing condition if we look at the setting from the point of view of the
creature forcing. Sets of partial functions determined by a pure candidate might
be considered as large sets.
Our main proofs are following the celebrated Glazers proof of the Hindman
Theorem, which reduced the problem to the existence of a relevant ultralter on
in ZFC. Those arguments were presented by Comfort in [2, Theorem 10.3, p.451]
with [2, Lemma 10.1, p.449] as a crucial step (stated here in 2.7). The arguments
of the second section of our paper really resemble Glazers proof. In that section
we deal with the easier case of omittorylike creatures (loose FFCC pairs of 1.2(2))
Date: January, 2011.
1991 Mathematics Subject Classication. Primary 03E05; Secondary: 03E02, 05D10, 54D80.
This research was supported by the United States-Israel Binational Science Foundation. Pub-
lication 957.
1
9
5
7


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







m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
3
-
2
3


2 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
and in the proof of the main conclusion (2.10) we use an ultralter idempotent with
respect to operation (dened in 2.4). The third section deals with the case of tight
FFCC pairs of 1.2(4). Here, we consider partitions of some sets of partial functions
all of which have domains being essentially intervals of integers starting with some
xed n < . While the general scheme of the arguments follows the pattern of
the second section, they are slightly more complicated as they involve sequences of
ultralters and operations on them. As an application of this method, in 3.9 we give
a new proof of a partition theorem by Carlson and Simpson [1, Theorem 6.3]. The
next section presents a variation of the third section: under weaker assumptions
on the involved FFCC pairs we get weaker, yet still interesting partition theorem.
Possible applications of this weaker version include a special case of the partition
theorem by Goldstern and Shelah [3] (see 4.9). These results motivate the fourth
section, where we develop the parallel of the very weak bigness for candidates with
limsup demand on the norms.
Our paper is self-contained and all needed creature terminology is introduced
in the rst section. We also give there several examples of creating pairs to which
our results may be applied.
Notation: We use standard set-theoretic notation.
An integer n is the set 0, 1, . . . , n 1 of all integers smaller than n, and the
set of all integers is called . For integers n < m, the interval [n, m) denotes the
set of all integers smaller than m and greater than or equal to n.
All sequences will be indexed by natural numbers and a sequence of objects is
typically denoted by a bar above a letter with the convention that x = x
i
: i < y),
y .
For a set X the family of all subsets of X is denoted by T(X). The domain of
a function f is called dom(f).
An ideal J on is a family of subsets of such that
(i) all nite subsets of belong to J but / J, and
(ii) if A B J, then A J and if A, B J then A B J.
For an ideal J, the family of all subsets of that do not belong to J is denoted by
J
+
, and the lter dual to J is called J
c
.
1. Partial creatures
We use the context and notation of Roslanowski and Shelah [6], but below we
recall all the required denitions and concepts.
Since we are interested in Ramsey-type theorems and ultralters on a countable
set of partial functions, we will use pure candidates rather than forcing notions
generated by creating pairs. Also, our considerations will be restricted to creating
pairs which are forgetful, smooth ([6, 1.2.5]), monotonic ([6, 5.2.3]), strongly nitary
([6, 1.1.3, 3.3.4]) and in some cases omittorylike ([6, 2.1.1]). Therefore we will
reformulate our denitions for this restricted context (in particular, val[t] is a set
of partial functions), thus we slightly depart from the setting of [6].
Context 1.1. In this paper H is a xed function dened on and such that H(i)
is a nite non-empty set for each i < . The set of all nite non-empty functions f
such that dom(f) and f(i) H(i) (for all i dom(f)) will be denoted by T
H
.
9
5
7


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







m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
3
-
2
3


PARTITION THEOREMS FROM CREATURES AND IDEMPOTENT ULTRAFILTERS 3
Denition 1.2. (1) An FP creature
1
for H is a tuple
t = (nor, val, dis, m
dn
, m
up
) = (nor[t], val[t], dis[t], m
t
dn
, m
t
up
)
such that
nor is a non-negative real number, dis is an arbitrary object and
m
t
dn
< m
t
up
< and
val is a non-empty nite subset of T
H
such that dom(f) [m
t
dn
, m
t
up
)
for all f val.
(2) An FFCC pair
2
for H is a pair (K, ) such that
(a) K is a countable family of FP creatures for H,
(b) for each m < the set K
m
:= t K : m
t
up
m is nite and the
set K
m
:= t K : m
t
dn
m & nor[t] m is innite,
(c) is a function with the domain dom() included in the set
(t
0
, . . . , t
n
) : n < , t

K and m
t

up
m
t
+1
dn
for < n
and the range included in T(K) ,
(d) if t (t
0
, . . . , t
n
) then (t K and) m
t0
dn
= m
t
dn
< m
t
up
= m
tn
up
,
(e) t (t) (for each t K) and
(f) if t (t
0
, . . . , t
n
) and f val[t], then
dom(f)
_
[m
t

dn
, m
t

up
) : n
and f[m
t

dn
, m
t

up
) val[t

] for n, and
(g) if

t
0
, . . . ,

t
n
dom() and

t =

t
0

. . .

t
n
dom(), then
_
(s
0
, . . . , s
n
) : s

) for n (

t).
(3) An FFCC pair (K, ) is loose if
(c
loose
) the domain of is
dom() = (t
0
, . . . , t
n
) : n < , t

K and m
t

up
m
t
+1
dn
for < n.
(4) An FFCC pair (K, ) is tight if
(c
tight
) the domain of is
dom() = (t
0
, . . . , t
n
) : n < , t

K and m
t

up
= m
t
+1
dn
for < n,
(f
tight
) if t (t
0
, . . . , t
n
) and f val[t], then f[m
t

dn
, m
t

up
) val[t

] for all
n, and
(h
tight
) if s
0
, s
1
K, m
s0
up
= m
s1
dn
, f
0
val[s
0
], f
1
val[s
1
] and f = f
0
f
1
,
then there is s (s
0
, s
1
) such that f val[s].
Denition 1.3 (Cf. [6, Denition 1.2.4]). Let (K, ) be an FFCC pair for H.
(1) A pure candidate for (K, ) is a sequence

t = t
n
: n < ) such that t
n
K,
m
tn
up
m
tn+1
dn
(for n < ) and lim
n
nor[t
n
] = .
A pure candidate

t is tight if m
tn
up
= m
tn+1
dn
(for n < ).
The set of all pure candidates for (K, ) is denoted by PC

(K, ) and the

family of all tight pure candidates is called PC
tt

(K, ).
1
FP stands for Forgetful Partial creature
2
FFCC stands for smooth Forgetful monotonic strongly Finitary Creature Creating pair
9
5
7


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







m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
3
-
2
3


4 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
(2) For pure candidates

t, s PC

(K, ) we write

t s whenever there is a
sequence u
n
: n < ) of non-empty nite subsets of satisfying
max(u
n
) < min(u
n+1
) and s
n
(

tu
n
) for all n < .
(3) For a pure candidate

t = t
i
: i < ) PC

(K, ) we dene
(a) o(

t) = (t
i0
, . . . , t
in
) : i
0
< . . . < i
n
< for some n < , and
(b)

t) =

( s) : s o(

t) and
tt
(

t) =

(t
0
, . . . , t
n
) : n < ,
(c) pos(

t) =

val[s] : s

t) and pos
tt
(

t) =

val[s] : s
tt
(

t),
(d)

t n = t
n+k
: k < ).
(Above, if s / dom() then we stipulate ( s) = .)
Remark 1.4. Loose FFCC and tight FFCC are the two cases of FFCC pairs treated
in this article. The corresponding partition theorems will be slightly dierent in
the two cases, though there is a parallel. In the loose case we will deal with

t),
pos(

t) and ultralters on the latter set. In the tight case we will use
tt
(

t), pos
tt
(

t)
and sequences of ultralters on pos
tt
(t n) (for n < ).
We will require two additional properties from (K, ): weak bigness and weak
additivity (see 1.5, 1.6). Because of the dierences in the treatment of the two
cases, there are slight dierences in the formulation of these properties, so we have
two variants for each: lvariant and tvariant (where l stands for loose and
t stands for tight, of course).
Plainly, PC
tt

(K, ) PC

(K, ),
tt
(

t)

t) and pos
tt
(

t) pos(

t). Also,
if

t PC
tt

(K, ), then

t n PC
tt

(K, ) for all n < .

Denition 1.5. Let (K, ) be an FFCC pair for H and

t = t
i
: i < )
PC

(K, ).
(1) We say that the pair (K, ) has weak ladditivity for the candidate

t if for
some increasing f : , for every m < we have:
if s
0
, s
1

t), nor[s
0
] f (m), m
s0
dn
f (m), nor[s
1
] f (m
s0
up
) and
m
s1
dn
> f (m
s0
up
), then we can nd s

t) such that
m
s
dn
m, nor[s] m, and val[s] f g : f val[s
0
], g val[s
1
].
(2) The pair (K, ) has weak tadditivity for the candidate

t if for some in-
creasing f : , for every n, m < we have:
if s
0
(t
n
, . . . , t
k
), k n, nor[s
0
] f (n + m), s
1
(t
k+1
, . . . , t

),
nor[s
1
] f (k +m) and > k, then we can nd s (t
n
, . . . , t

) such that
nor[s] m and val[s] f g : f val[s
0
], g val[s
1
].
(3) The pair (K, ) has ladditivity if for all s
0
, s
1
K with nor[s
0
], nor[s
1
] >
1 and m
s0
up
m
s1
up
there is s (s
0
, s
1
) such that
nor[s] minnor[s
0
], nor[s
1
] 1 and val[s] f g : f val[s
0
], g val[s
1
].
The pair (K, ) has tadditivity if for all s
0
, s
1
K with nor[s
0
], nor[s
1
] >
1 and m
s0
up
= m
s1
up
there is s (s
0
, s
1
) such that
nor[s] minnor[s
0
], nor[s
1
] 1.
We say that (K, ) has tmultiadditivity if for all s
0
, . . . , s
n
K with
m
s

up
= m
s
+1
dn
(for < n) there is s (s
0
, . . . , s
n
) such that nor[s]
maxnor[s

] : n 1.
Denition 1.6. Let (K, ) be an FFCC pair for H and

t = t
i
: i < )
PC

(K, ).
9
5
7


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







m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
3
-
2
3


PARTITION THEOREMS FROM CREATURES AND IDEMPOTENT ULTRAFILTERS 5
(1) We say that the pair (K, ) has weak lbigness for the candidate

t whenever
the following property is satised:
()

t
l
if n
1
, n
2
, n
3
< and pos(

t) =

: < n
1
, then for some s

t)
and < n
1
we have
nor[s] n
2
, m
s
dn
n
3
, and val[s] T

.
(2) We say that the pair (K, ) has weak tbigness for the candidate

t whenever
the following property is satised:
()

t
t
if n, n
1
, n
2
< and pos
tt
(

t n) =

: < n
1
, then for some
s
tt
(

t n) and < n
1
we have
nor[s] n
2
and val[s] T

.
(3) We say that the pair (K, ) has bigness if for every creature t K with
nor[t] > 1 and a partition val[t] = F
1
F
2
, there are 1, 2 and s (t)
such that nor[s] nor[t] 1 and val[s] F

.
Denition 1.7. Let (K, ) be an FFCC pair for H.
(1) (K, ) is simple except omitting if for every (t
0
, . . . , t
n
) dom() and
t (t
0
, . . . , t
n
) for some n we have val[t] val[t

].
(2) (K, ) is gluing on a candidate

t = t
i
: i < ) PC

(K, ) if for every

n, m < there are k n and s (t
n
, . . . , t
k
) such that nor[s] m.
The following two observations summarize the basic dependencies between the
notions introduced in 1.5, 1.6 separately for the two contexts (see 1.4).
Observation 1.8. Assume (K, ) is a loose FFCC pair,

t PC

(K, ).
(1) If (K, ) has bigness (ladditivity, respectively), then it has weak lbigness
(weak ladditivity, respectively) for the candidate

t.
(2) If (K, ) has the weak lbigness for

t, k < and pos(

t) =

<k
T

, then for
some s PC

(K, ) and < k we have

t s and (n < )(val[s

n
] T

).
(3) Assume that (K, ) has the weak lbigness property for

t PC

(K, )
and it is simple except omitting. Let k < and pos(

t) =

<k
T

. Then for
some s

t and < k we have pos( s) T

.
Observation 1.9. Assume (K, ) is a tight FFCC pair,

t PC
tt

(K, ).
(1) If (K, ) has bigness and is gluing on

t, then it has the weak tbigness for
the candidate

t.
(2) If (K, ) has tadditivity, then it has the weak tadditivity for

t.
(3) If (K, ) has the tmultiadditivity, then it has the tadditivity and it is
gluing on

t.
In the following two sections we will present partition theorems for the loose and
then for the tight case. First, let us oer some easy examples to which the theory
developed later can be applied.
Example 1.10. Let H
1
(n) = n+1 for n < and let K
1
consist of all FP creatures
t for H
1
such that
dis[t] = (u, i, A) = (u
t
, i
t
, A
t
) where u [m
t
dn
, m
t
up
), i u, , = A H
1
(i),
nor[t] = log
2
([A[),
9
5
7


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







m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
3
-
2
3


6 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
val[t]

ju
H
1
(j) is such that f(i) : f val[t] = A.
For t
0
, . . . , t
n
K
1
with m
t

up
m
t
+1
dn
let
1
(t
0
, . . . , t
n
) consist of all creatures
t K
1
such that
m
t
dn
= m
t0
dn
, m
t
up
= m
tn
up
, u
t
=
_
n
u
t

, i
t
= i
t

, A
t
A
t

for some

n,
and val[t]
_
f
0
. . . f
n
: (f
0
, . . . , f
n
) val[t
0
] . . . val[t
n
]
_
.
Also, let

1
be
1
restricted to the set of those tuples (t
0
, . . . , t
n
) for which m
t

up
=
m
t
+1
dn
(for < n). Then
(K
1
,
1
) is a loose FFCC pair for H
1
with bigness and ladditivity,
(K
1
,

1
) is a tight FFCC pair for H
1
with bigness and tmultiadditivity,
and it is gluing on every

t PC
tt

(K
1
,

1
).
Example 1.11. Let H
2
(n) = 2 for n < and let K
2
consist of all FP creatures t
for H
2
such that
,= dis[t] [m
t
dn
, m
t
up
),
,= val[t]
dis[t]
2,
nor[t] = log
2
([val[t][).
For t
0
, . . . , t
n
K
2
with m
t

up
m
t
+1
dn
let
2
(t
0
, . . . , t
n
) consist of all creatures
t K
2
such that
m
t
dn
= m
t0
dn
, m
t
up
= m
tn
up
, dis[t] = dis[t

], and val[t] val[t

] for some

n.
Then (K
2
,
2
) is a loose FFCC pair for H
1
which is simple except omitting and
has bigness.
Example 1.12. Let H be as in 1.1 and let K
3
consist of all FP creatures t for H
such that
,= dis[t] [m
t
dn
, m
t
up
),
val[t] f T
H
: dis[t] dom(f) [m
t
dn
, m
t
up
) satises
(g

idis[t]
H(i))(f val[t])(g f),
nor[t] = log
957
([dis[t][).
For t
0
, . . . , t
n
K
2
with m
t

up
m
t
+1
dn
let
3
(t
0
, . . . , t
n
) consist of all creatures
t K
3
such that
m
t
dn
= m
t0
dn
, m
t
up
= m
tn
up
, dis[t]

n
dis[t

], and
if f val[t], then dom(f)

[m
t

dn
, m
t

up
) : n and f[m
t

dn
, m
t

up
)
val[t

] for all n.
Also, for t
0
, . . . , t
n
K
2
with m
t

up
= m
t
+1
dn
let

3
(t
0
, . . . , t
n
) consist of all creatures
t K
3
such that
m
t
dn
= m
t0
dn
, m
t
up
= m
tn
up
, dis[t]

n
dis[t

], and
if f val[t], then dom(f)

[m
t

dn
, m
t

up
) : n and f[m
t

dn
, m
t

up
)
val[t

] for all n.
Then
(K
3
,
3
) is a loose FFCC pair for H with bigness and ladditivity,
9
5
7


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







m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
3
-
2
3


PARTITION THEOREMS FROM CREATURES AND IDEMPOTENT ULTRAFILTERS 7
(K
3
,

3
) is a tight FFCC pair for H with bigness and tmultiadditivity
and it is gluing on every

t PC
tt

(K
3
,

3
).
Example 1.13. Let N > 0 and H
N
(n) = N. Let K
N
consist of all FP creatures
t for H
N
such that
dis[t] = (X
t
,
t
), where X
t
[m
t
dn
, m
t
up
), and
t
: X
t
N,
nor[t] = m
t
up
,
val[t] =
_
f
[m
t
dn
,m
t
up
)
N :
t
f and f is constant on [m
t
dn
, m
t
up
) X
t
_
.
For t
0
, . . . , t
n
K
N
with m
t

up
= m
t
+1
dn
(for < n) we let
N
(t
0
, . . . , t
n
) consist of
all creatures t K
N
such that
m
t
dn
= m
t0
dn
, m
t
up
= m
t0
up
, X
t0
. . . X
tn
X
t
,
for each n,
either X
t
[m
t

dn
, m
t

up
) = X
t

and
t
[m
t

dn
, m
t

up
) =
t

,
or [m
t

dn
, m
t

up
) X
t
and
t
[m
t

dn
, m
t

up
) val[t

].
Then
(i) (K
N
,
N
) is a tight FFCC pair for H
N
,
(ii) it has the tmultiadditivity and
(iii) it has the weak tbigness and is gluing for every candidate

t PC
tt

(K, ).
Proof. (i) All demands in 1.2(2,4) are easy to verify. For instance, to check
1.2(4)(h
tight
) note that:
if s
0
, s
1
K
N
, m
s0
up
= m
s1
dn
, f

val[s

] (for = 0, 1) and s K
N
is such that
m
s
dn
= m
s0
dn
, m
s
up
= m
s1
up
, X
s
= X
s0
[m
s1
dn
, m
s1
up
),
s
X
s0
=
s0
,
s
[m
s1
dn
, m
s1
up
) = f
1
,
then s
N
(s
0
, s
1
) and f
0
f
1
val[s].
(ii) The s constructed as in (i) above for s
0
, s
1
will witness the tadditivity as
well. In an analogous way we show also the multiadditivity.
(iii) Let

t = t
i
: i < ) PC
tt

(K
N
,
N
). Suppose that n, n
1
, n
2
< and
pos
tt
(

t n) =

: < n
1
. By the HalesJewett theorem (see [?]) there is
k > n
2
such that for any partition of
k
N into n
1
parts there is a combinatorial line
included in one of the parts. Then we easily nd s
N
(t
0
, . . . , t
k1
) such that
val[s] T

for some < n

1
. Necessarily, nor[s] k 1 n
2
. This proves the
weak tbigness for

t. Similarly to (ii) we may argue that (K
N
,
N
) is gluing on

t.
2. Ultrafilters on loose possibilities
Here we introduce ultralters on the (countable) set T
H
(see 1.1) which contain
sets large from the point of view of pure candidates for a loose FFCC pair. Then
we use them to derive a partition theorem for this case.
Denition 2.1. Let (K, ) be a loose FFCC pair for H.
(1) For a pure candidate

t PC

(K, ), we dene
/
0

t
= pos(

t n) : n < ,
/
1

t
is the collection of all sets A T
H
such that for some N < we
have
_
s

t)
__
nor[s] N & m
s
dn
N val[s] A ,=
_
,
9
5
7


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







m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
3
-
2
3


8 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
/
2

t
is the collection of all sets A T
H
such that for some N < we
have
(

t
1


t)(

t
2


t
1
)(s

t
2
))(nor[s] N val[s] A ,=
_
.
(2) For < 3 we let uf

t
(K, ) be the family of all ultralters D on T
H
such
that /

t
D. We also set (for < 3)
uf

(K, )
def
=
_
uf

t
(K, ) :

t PC

(K, ).
Proposition 2.2. Let (K, ) be a loose FFCC pair for H,

t PC

(K, ).
(1) /
0

t
/
1

t
/
2

t
and hence also uf
2

t
(K, ) uf
1

t
(K, ) uf
0

t
(K, ).
(2) uf
0

t
(K, ) ,= .
(3) If (K, ) has the weak lbigness for each

t

t, then uf
2

t
(K, ) is not
empty.
(4) If (K, ) has the weak lbigness for

t, then uf
1

t
(K, ) ,= .
(5) Assume CH. Suppose that (K, ) is simple except omitting (see 1.7(1)) and
has the weak lbigness on every candidate

t PC

(K, ). Then there is

D uf
2

t
(K, ) such that
_
A D
__

t PC

(K, )
__
pos(

t) D & pos(

t) A
_
.
Proof. (2) Note that /
0

t
has the nite intersection property (p).
(3) It is enough to show that, assuming (K, ) has the weak lbigness for all

t, /
2

t
has p. So suppose that for < k we are given a set A

/
2

t
and let
N

< be such that

()

t
1


t)(

t
2


t
1
)(s

t
2
))(nor[s] N

val[s] A

,=
_
.
Let N = maxN

: < k. Then we may choose

t

t such that
() (s

))(nor[s] N ( < k)(val[s] A

,= )
_
.
[Why? Just use repeatedly ()

for = 0, 1, . . . , k 1; remember

t

implies

).]
For
k
2 set
T

= f pos(

) : ( < k)(() = 1 f A

).
Then pos(

) =

:
k
2 and (K, ) has the weak lbigness for

t

, so we
may use Observation 1.8(2) to pick
0

k
2 and s

t

such that val[s

n
] T
0
for all n < . Consider n < such that nor[s
n
] > N. It follows from () that
val[s
n
] A

,= for all < k. Hence, by the choice of s,

0
() = 1 for all < k and
therefore ,= val[s
n
]

<k
A

.
(4) Similarly to (3) above one shows that /
1

t
has p.
(5) Assuming CH and using Observation 1.8(3) we may construct a sequence

: <
1
) PC

(K, ) such that

if < <
1
then (n < )(

n)),
if A T
H
then for some <
1
we have that either pos(

) A or
pos(

) A = .
(Compare to the proof of [6, 5.3.4].) Then the family
pos(

n) : <
1
& n <
generates the desired ultralter.
9
5
7


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







m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
3
-
2
3


PARTITION THEOREMS FROM CREATURES AND IDEMPOTENT ULTRAFILTERS 9
Observation 2.3. The sets uf

t
(K, ) (for < 3) are closed subsets of the (Haus-
dor compact topological space)

(T
H
) of non-principal ultralters on T
H
. Hence
each uf

t
(K, ) itself is a compact Hausdor space.
Denition 2.4. (1) For f T
H
and A T
H
we dene
f A
def
= g T
H
: max(dom(f)) < min(dom(g)) and f g A.
(2) For D
1
, D
2
uf
0
(K, ) we let
D
1
D
2
def
=
_
A T
H
: f T
H
: (f A) D
1
D
2
_
.
Proposition 2.5. (1) If A
1
, A
2
T
H
and f T
H
, then
f (A
1
A
2
) = (f A
1
) (f A
2
) and
g T
H
: max(dom(f)) < min(dom(g)) (f A
1
) = f (T
H
A
1
).
(2) If D
1
, D
2
, D
3
uf
0
(K, ), then D
1
D
2
is a non-principal ultralter on
T
H
and D
1
(D
2
D
3
) = (D
1
D
2
) D
3
.
(3) The mapping : uf
0
(K, ) uf
0
(K, )

(T
H
) is right continuous
(i.e., for each D
1
uf
0
(K, ) the function uf
0
(K, ) D
2
D
1
D
2

(T
H
) is continuous).
Proof. Straightforward, compare with 3.3.
Proposition 2.6. Assume that a loose FFCC pair (K, ) has the weak ladditivity
(see 1.5(1)) for a candidate

t PC

(K, ). If D
1
, D
2
uf
1

t
(K, ), then D
1
D
2

uf
1

t
(K, ).
Proof. Let f : witness the weak ladditivity of (K, ) for

t, and let D =
D
1
D
2
, D
1
, D
2
uf
1

t
(K, ). We already know that D is an ultralter on T
H
(by
2.5(2)), so we only need to show that it includes /
1

t
.
Suppose that A /
1

t
and let N < be such that
()
1
_
s

t)
__
nor[s] N & m
s
dn
N val[s] A ,=
_
.
Claim 2.6.1. For every s

t), if nor[s] f (N) and m

s
dn
f (N), then val[s]
f T
H
: f A D
1
,= .
Proof of the Claim. Suppose s
0

t), nor[s
0
] f (N), m
s0
dn
f (N). Set
B =
_
f A : f val[s
0
].
We are going to argue that
()
2
B /
1

t
.
So let M = f (m
s0
up
) + m
s0
up
+ 957 and suppose s
1
(

t) is such that nor[s

1
] M
and m
s1
dn
M. Apply the weak additivity and the choice of M to nd s

t)
such that
m
s
dn
N, nor[s] N and val[s] f g : f val[s
0
] & g val[s
1
].
Then, by ()
1
, val[s] A ,= so for some f val[s
0
] and g val[s
1
] we have f g
A (and max(dom(f)) < min(dom(g))). Thus g (f A) val[s
1
] B val[s
1
]
and ()
2
follows.
Since D
1
uf
1

t
(K, ) we conclude from ()
2
that B D
1
and hence (as val[s
0
]
is nite) f A D
1
for some f val[s
0
], as desired.
9
5
7


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







m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
3
-
2
3


10 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
It follows from 2.6.1 that f T
H
: f A D
1
/
1

t
and hence (as D
2

uf
1

t
(K, )) f T
H
: f A D
1
D
2
. Consequently, A D
1
D
2
.
Lemma 2.7 (See [2, Lemma 10.1, p.449]). If X is a non-empty compact Hausdor
space, an associative binary operation which is continuous from the right (i.e.
for each p X the function q pq is continuous), then there is a idempotent
point p X (i.e. p p = p).
Corollary 2.8. Assume that a loose FFCC pair (K, ) has weak ladditivity and
the weak lbigness for a candidate

t PC

(K, ). Then
(1) uf
1

t
(K, ) is a non-empty compact Hausdor space and is an associative
right continuous operation on it.
(2) There is D uf
1

t
(K, ) such that D = D D.
Proof. (1) By 2.2(3), 2.3, 2.5(2,3) and 2.6.
(2) It follows from (1) above that all the assumptions of Lemma 2.7 are satised
for and uf
1

t
(K, ), hence its conclusion holds.
Theorem 2.9. Assume that (K, ) is a loose FFCC pair,

t PC

(K, ). Let an
ultralter D uf
1

t
(K, ) be such that DD = D. Then
_
A D
__
s

t
__
pos( s) A
_
.
Proof. The main ingredient of our argument is given by the following claim.
Claim 2.9.1. Let (K, ),

t and D be as in the assumptions of 2.9. Assume A D
and n < . Then there is s

t) such that
()
1
val[s] A, nor[s] n, m
s
dn
n, and
()
2
(f val[s])(f A D).
Proof of the Claim. Let A

:= f T
H
: f A D and A

:= A A

. Since
A D = D D we know that A

D and thus A

D. Hence T
H
A

/ /
1

t
(remember D uf
1

t
(K, )). Therefore, there is s

t) such that
nor[s] n, m
s
dn
n, and val[s] (T
H
A

) = .
Then val[s] A and for each f val[s] we have f A D, as desired.
Now suppose A D. By induction on n we choose s
n
, A
n
so that
(a) A
0
= A, A
n
D and A
n+1
A
n
,
(b) s
n

t), nor(s
n
) n and m
sn
up
m
sn+1
dn
,
(c) val[s
n
] A
n
,
(d) if f A
n+1
, then m
sn
up
min(dom(f)),
(e) if f val[s
n
], then A
n+1
f A
n
.
Suppose we have constructed s
0
, . . . , s
n1
and A
n
so that demands (a)(e) are
satised. Set N = m
sn1
up
+ n + 1 (if n = 0 stipulate m
sn1
up
= 0) and use Claim
2.9.1 to nd s
n

t) such that
()
n
1
val[s
n
] A
n
, nor[s
n
] N, m
sn
dn
N, and
()
n
2
(f val[s
n
])(f A
n
D).
Put
A
n+1
:= A
n
g T
H
: m
sn
up
< min(dom(g))

fval[sn]
f A
n
.
9
5
7


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







m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
3
-
2
3


PARTITION THEOREMS FROM CREATURES AND IDEMPOTENT ULTRAFILTERS 11
Since /
0

t
D we know that g T
H
: m
sn
up
< min(dom(g)) D and since val[s
n
]
is nite (and by ()
n
2
) also

fval[sn]
f A
n
D. Thus A
n+1
D. Plainly the other
requirements hold too.
After the above construction is carried out we set s = s
n
: n < ). Clearly
s PC

(K, ) and s

t (remember clause (b)).
Claim 2.9.2. If n
0
< . . . < n
k
< and f

val[s
n

] for k, then

k
f

A
n0
.
Proof of the Claim. Induction on k. If k = 0 then clause (c) of the choice of s
n0
gives the conclusion. For the inductive step suppose the claim holds true for k
and let n
0
< n
1
< . . . < n
k
< n
k+1
, f

val[s
n

] (for k + 1). Letting

g = f
1
. . . f
k+1
we may use the inductive hypothesis to conclude that g A
n1
.
By (a)+(e) we know that A
n1
A
n0+1
f
0
A
n0
, so g f
0
A
n0
. Hence
f
0
g = f
0
f
1
. . . f
k+1
A
n0
.
It follows from 2.9.2 that pos( s) A (remember (a) above and 1.2(2)(f)).
Conclusion 2.10. Suppose that (K, ) is a loose FFCC pair with weak lbigness
and weak ladditivity over

t PC

(K, ). Assume also that pos(

t) is the nite
union T
0
. . . T
n
. Then for some i n and s PC

(K, ) we have
pos( s) T
i
and

t s.
Proof. By 2.8 there is D uf
1

t
(K, ) such that D = D D. Clearly for some
i n we have T
i
D. By 2.9 there is s PC

(K, ) such that

t s and
pos( s) T
i
.
3. Ultrafilters on tight possibilities
In this section we carry out for tight FFCC pairs considerations parallel to that
from the case of loose FFCC pairs. The main dierence now is that we use sequences
of ultralters, but many arguments do not change much.
Denition 3.1. Let (K, ) be a tight FFCC pair for H,

t = t
n
: n < )
PC
tt

(K, ).
(1) For f pos
tt
(

t n) let x
f
= x

t
f
be the unique m > n such that f val[s]
for some s (t
n
, . . . , t
m1
). (Note 1.2(4)(f
tight
).)
(2) If f pos
tt
(

t n), n < , A T
H
, then we set
f A = f
t
A =
_
g pos
tt
(

t x
f
) : f g A
_
.
(3) We let suf
t
(K, ) be the set of all sequences

D = D
n
: n < ) such that
each D
n
is a non-principal ultralter on pos
tt
(

t n).
(4) The space suf
t
(K, ) is equipped with the (Tichonov) product topology of

n<

_
pos
tt
(

t n)
_
. For a sequence

A = A
0
, . . . , A
n
) such that A

pos
tt
(

t ) (for n) we set
Nb
A
=
_

D suf
t
(K, ) : ( n)(A

)
_
.
(5) For

D = D
n
: n < ) suf
t
(K, ), n < and A pos
tt
(

t n) we let
set
n

t
(A,

D) =
_
f pos
tt
(

t n) : f A D
x
f
_
.
9
5
7


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







m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
3
-
2
3


12 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
(6) For

D
1
,

D
2
suf
t
(K, ) we dene

D
1

D
2
to be a sequence D
n
: n < )
such that for each n
D
n
=
_
A pos
tt
(

t n) : set
n

t
(A,

D
1
) D
2
n
_
.
Observation 3.2. Let (K, ) be a tight FFCC pair for H and

t PC
tt

(K, ).
Suppose f pos
tt
(

t n), g pos
tt
(

t x
f
). Then
(1) f g pos
tt
(

t n) (note 1.2(4)(h
tight
)) and
(2) (f g) A = g (f A) for all A T
H
.
(3) suf
t
(K, ) is a compact Hausdor topological space. The sets Nb
A
for

A = A
0
, . . . , A
n
), A

pos
tt
(

t ), n < , form a basis of the

topology of suf
t
(K, ).
Proposition 3.3. Let (K, ) be a tight FFCC pair for H and

t PC
tt

(K, ).
(1) If

D
1
,

D
2
suf
t
(K, ), then

D
1

D
2
suf
t
(K, ).
(2) The mapping : suf
t
(K, )suf
t
(K, ) suf
t
(K, ) is right continuous.
(3) The operation is associative.
Proof. (1) Let

D
1
,

D
2
suf
t
(K, ), n < , and
D
n
=
_
A pos
tt
(

t n) : set
n

t
(A,

D
1
)

D
2
n
_
.
Let A, B pos
tt
(

t n).
(a) If f pos
tt
(

t n) and A is nite, then f A is nite as well, so it does not

belong to D
1
x
f
. Consequently, if A is nite then set
n

t
(A,

D
1
) = and A / D
n
.
(b) set
n

t
(pos
tt
(

t n),

D
1
) = pos
tt
(

t n) D
2
n
(note 3.2(1)). Thus pos
tt
(

t n)
D
n
.
(c) If A B then set
n

t
(A,

D
1
) set
n

t
(B,

D
1
) and hence
A B & A D
n
B D
n
.
(d) set
n

t
(A B,

D
1
) = set
n

t
(A,

D
1
) set
n

t
(B,

D
1
) and hence
A, B D
n
A B D
n
.
(e) set
n

t
(pos
tt
(

t n) A,

D
1
) = pos
tt
(

t n) set
n

t
(A,

D
1
), and hence
A / D
n
pos
tt
(

t n) A D
n
.
It follows from (a)(e) that D
n
is a non-principal ultralter on pos
tt
(

t n) and
hence clearly

D
1

D
2
suf
t
(K, ).
(2) Fix

D
1
suf
t
(K, ) and let

A = A

: n), A

pos
tt
(

t ).
For n put B

= set

t
(A

,

D
1
) and let

B = B

: n). Then for each

D
2
suf
t
(K, ) we have

D
1

D
2
Nb
A
if and only if

D
2
Nb
B
.
(3) Let

D
1
,

D
2
,

D
3
suf
t
(K, ). Suppose n < , A pos(

t n). Then
(i) A
_
(

D
1

D
2
)

D
3
_
n
i set
n

t
(A,

D
1

D
2
) D
3
n
i
_
f pos
tt
(

t n) : f A (

D
1

D
2
)
x
f
_
D
3
n
, and
(ii) A
_

D
1
(

D
2

D
3
)
_
n
i set
n

t
(A,

D
1
) (

D
2

D
3
)
n
i
set
n

t
_
set
n

t
(A,

D
1
),

D
2
_


D
3
n
i
_
f pos
tt
(

t n) : f set
n

t
(A,

D
1
)
D
2
x
f
_
D
3
n
.
9
5
7


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







m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
3
-
2
3


PARTITION THEOREMS FROM CREATURES AND IDEMPOTENT ULTRAFILTERS 13
Let us x f pos
tt
(

t n) for a moment. Then

f A
_

D
1

D
2
_
x
f
i set
x
f

t
(f A,

D
1
) D
2
x
f
i
_
g pos
tt
(

t x
f
) : g (f A) D
1
xg
_
D
2
x
f
i
_
g pos
tt
(

t x
f
) : (f g) A D
1
xg
_
D
2
x
f
i
_
g pos
tt
(

t x
f
) : (f g) set
n

t
(A,

D
1
)
_
D
2
x
f
i f set
n

t
(A,

D
1
) D
2
x
f
.
Consequently,
_
f pos
tt
(

tn) : f A (

D
1

D
2
)
x
f
_
=
_
f pos
tt
(

tn) : f set
n

t
(A,

D
1
) D
2
x
f
_
and (by (i)+(ii)) A
_

D
1
(

D
2

D
3
)
_
n
if and only if A
_
(

D
1

D
2
)

D
3
_
n
.
Denition 3.4. Let (K, ) be a tight FFCC pair for H and

t PC
tt

(K, ).
(1) For n < , B
n

t
is the family of all sets B pos
tt
(

t n) such that for some

M we have:
if s
tt
(

t n) and nor[s] M, then val[s] B ,= .

(2) suf

t
(K, ) is the family of all

D = D
n
: n < ) suf
t
(K, ) such that
B
n

t
D
n
for all n < .
Proposition 3.5. Let (K, ) be a tight FFCC pair for H and

t PC
tt

(K, ).
(1) suf

t
(K, ) is a closed subset of suf
t
(K, ).
(2) If (K, ) has the weak tbigness for

t, then suf

t
(K, ) ,= .
(3) If (K, ) has the weak tadditivity for

t, then suf

t
(K, ) is closed under
.
Proof. (1) Suppose

D suf
t
(K, ) suf

t
(K, ). Let n < and B B
n

t
be
such that B / D
n
. Set A
n
= pos
tt
(

t n) B and A

= pos
tt
(

t ) for < n, and

let

A = A
0
, . . . , A
n
). Then

D Nb
A
suf
t
(K, ) suf

t
(K, ).
(2) It is enough to show that, assuming the weak tbigness, each family B
n

t
has
p. To this end suppose that B
0
, . . . , B
m1
B
n

t
. Pick M
0
such that
()
_
s
tt
(

t n)
__
< m
__
nor[s] M
0
B

val[s] ,=
_
.
For
m
2 set C

= f pos
tt
(

t n) : ( < m)(f B

() = 1). By the
weak tbigness we may choose and s
tt
(

t n) such that nor[s] > M

0
and
val[s] C

. Then (by ()) we also have () = 1 and val[s] B

for all < m.

Hence ,= val[s]

<m
B

.
(3) Let f : witness the weat tadditivity of (K, ) for

t. Suppose
that

D
1
,

D
2
suf

t
(K, ),

D =

D
1

D
2
. We have to show that for each n < ,
B
n

t
D
n
(remember 3.3(1)). To this end assume that B B
n

t
and let M be such
that
(s
tt
(

t n))(nor[s] M val[s] B ,= ).
Claim 3.5.1. If s
tt
(

t n) is such that nor[s] f (n +M),

then val[s] set
n

t
(B,

D
1
) ,= .
Proof of the Claim. Fix s
0
(t
n
, . . . , t
m1
) such that nor[s
0
] f (n + M). Let
A =

f B : f val[s
0
]. We claim that
() A B
m

t
.
[Why? Set N = f (m+M) + 957. Suppose s
1

tt
(

t m) has norm nor[s

1
] N.
By the weak tadditivity and the choice of N we can nd s
tt
(

t n) such that
nor[s] M and val[s] f g : f val[s
0
], g val[s
1
]. By the choice of M we
9
5
7


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







m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
3
-
2
3


14 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
have B val[s] ,= , so for some f val[s
0
] and g val[s
1
] we have g f B.
Thus val[s
1
] A ,= and we easily conclude that A B
m

t
.]
But

D
1
suf

t
(K, ), so B
m

t
D
1
m
and hence, for some f val[s
0
], we get
f B D
1
x
f
. Then f val[s
0
] set
n

t
(B,

D
1
).
It follows from 3.5.1 that set
n

t
(B,

D
1
) B
n

t
D
2
n
, so B D
n
as required.
Corollary 3.6. Assume that (K, ) is a tight FFCC pair with the weak tadditivity
and the weak tbigness for

t PC
tt

(K, ). Then there is

D suf

t
(K, ) such
that

D

D =

D.
Proof. By 2.7+3.2(3)+3.3+3.5.
Theorem 3.7. Assume that (K, ) is a tight FFCC pair,

t = t
n
: n < )
PC
tt

(K, ). Suppose also that

(a)

D = D
n
: n < ) suf

t
(K, ) is such that

D

D =

D, and
(b)

A = A
n
: n < ) is such that A
n
D
n
for all n < .
Then there is s = s
i
: i < ) PC
tt

(K, ) such that s

t, m
s0
dn
= m
t0
dn
and if
i < , s
i

tt
(

t k), then pos

tt
( s i) A
k
.
Proof. Let (K, ),

t,

D and

A be as in the assumptions. Then, in particular, B
k

t

D
k
for all k < .
Claim 3.7.1. If M, k < and B D
k
, then there is s
tt
(

t k) such that
val[s] B, nor[s] M and (f val[s])(f B D
x
f
).
Proof of the Claim. Since

D =

D

D and B D
k
, we know that set
k

t
(B,

D) D
k
and thus B set
k

t
(B,

D) D
k
. Hence pos
tt
(

t k) (B set
k

t
(B,

D)) / B
k

t
and we
may nd s
tt
(

t k) such that nor[s] M and val[s] B set

k

t
(B,

D). This s
is as required in the assertion of the claim.
Now we choose inductively s
i
, B
i
, k
i
(for i < ) such that
(i) B
0
= A
0
, k
0
= 0,
(ii) B
i
D
ki
, B
i
A
ki
, k
i
< k
i+1
< , s
i
(t
ki
, . . . , t
ki+11
),
(iii) val[s
i
] B
i
, nor[s
i
] i + 1,
(iv) if f val[s
i
], then B
i+1
f B
i
D
ki+1
.
Clause (i) determines B
0
and k
0
. Suppose we have already chosen k
i
and B
i
D
ki
.
By 3.7.1 we may nd k
i+1
> k
i
and s
i
(t
ki
, . . . , t
ki+11
) such that
nor[s
i
] i + 1, val[s
i
] B
i
and (f val[s
i
])(f B
i
D
ki+1
).
We let B
i+1
= A
ki+1

f B
i
: f val[s
i
] D
ki+1
. One easily veries the
relevant demands in (ii)(iv) for s
i
, B
i+1
, k
i+1
.
After the above construction is carried out, we set s = s
i
: i < ). Plainly,
s PC
tt

(K, ), s

t and m
s0
dn
= m
t0
dn
.
Claim 3.7.2. For each i, k < and s (s
i
, . . . , s
i+k
) we have val[s] B
i
.
Proof of the Claim. Induction on k < . If k = 0 then the assertion of the claim
follows from clause (iii) of the choice of s
i
. Assume we have shown the claim for
k. Suppose that s (s
i
, . . . , s
i+k
, s
i+k+1
), i < , and f val[s]. Let f
0
=
f[m
si
dn
, m
si
up
) val[s
i
] and f
1
= f[m
si+1
dn
, m
s
i+k+1
up
) pos
tt
( s (i + 1)) (remember
1.2(4)(f
tight
) and 3.2(1)). By the inductive hypothesis we know that f
1
B
i+1
, so
by clause (iv) of the choice of s
i
we get f
1
f
0
B
i
and thus f = f
0
f
1
B
i
.
9
5
7


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







m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
3
-
2
3


PARTITION THEOREMS FROM CREATURES AND IDEMPOTENT ULTRAFILTERS 15
It follows from 3.7.2 that for each i < we have pos
tt
( s i) B
i
A
ki
, as
required.
Conclusion 3.8. Suppose that (K, ) is a tight FFCC pair with weak tbigness and
weak tadditivity for

t PC
tt

(K, ).
(a) Assume that, for each n < , k
n
< and d
n
: pos
tt
(

t n) k
n
. Then
there is s = s
i
: i < ) PC
tt

(K, ) such that s

t, m
s0
dn
= m
t0
dn
and for
each n < ,
if n is such that s
i

tt
(

t n), then d
n
pos
tt
( s i) is constant.
(b) Suppose also that (K, ) has tmultiadditivity. Let d
n
: pos
tt
(

t n) k
(for n < ), k < . Then there are s = s
i
: i < ) PC
tt

(K, ) and
< k such that s

t and for each i < , if n is such that s
i

tt
(

t n)
and f pos
tt
( s i), then d
n
(f) = .
Now we will use 3.8 to give a new proof of CarlsonSimpson Theorem. This
theorem was used as a crucial lemma in the (inductive) proof of the Dual Ramsey
Theorem [1, Theorem 2.2].
Theorem 3.9. [Carlson and Simpson [1, Theorem 6.3]] Suppose that 0 < N < ,
X =

n<
n
N and X = C
0
. . . C
k
, k < . Then there exist a partition Y Y
i
:
i < of and a function f : Y N such that
(a) each Y
i
is a nite non-empty set,
(b) if i < j < then max(Y
i
) < min(Y
j
),
(c) for some k:
if i < , g : min(Y
i
) N, f min(Y
i
) g and gY
j
is constant for j < i,
then g C

.
Proof. For f X let d
0
(f) = min k : f C

. Consider the tight FFCC

pair (K
N
,
N
) dened in Example 1.13. It satises the assumptions of 3.8. Fix
any

t PC
tt

(K
N
,
N
) with m
t0
dn
= 0 and use 3.8(a) to choose s PC
tt

(K
N
,
N
)
such that s

t, m
s0
dn
= m
t0
dn
= 0 and d
0
pos
tt
( s) is constant. Set Y =

i<
X
si
,
f =

i<

si
and Y
i
= [m
si
dn
, m
si
up
) X
si
for i < .
4. Very weak bigness
The assumptions of Conclusion 3.8 (weak tbigness and weak tadditivity) are
somewhat strong. We will weaken them substantially here, getting weaker but still
interesting conclusion.
Denition 4.1. Let (K, ) be a tight FFCC pair for H,

t PC
tt

(K, ).
(1) For n < m < we dene
pos(

t[n, m)) =
_
val[s] : s (t
n
, . . . , t
m1
)
and we also keep the convention that pos(

t[n, n)) = .
[Note that pos(

t[n, m)) = f
n
. . . f
m1
: f

val[t

] for < m
(remember 1.2(4)(f
tight
) and 3.2(1)).]
(2) We say that (K, ) has the very weak tbigness for

t if
9
5
7


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







m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
3
-
2
3


16 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
()
vw

t
for every n, L, M < and a partition T
0
. . . T
L
= pos
tt
(

t n),
there are i
0
= n i
1
< i
2
i
3
, L and g
0
pos(

t[i
0
, i
1
)),
g
2
pos(

t[i
2
, i
3
)) and s (t
i1
, . . . , t
i21
) such that
nor[s] M and (g
1
val[s])(g
0
g
1
g
2
T

).
Observation 4.2. If a tight FFCC pair (K, ) has the weak tbigness for

t, then
it has the very weak tbigness for

t.
Denition 4.3. Let (K, ) be a tight FFCC pair for H and

t PC
tt

(K, ).
(1) For n < , (
n

t
is the family of all sets B pos
tt
(

t n) such that for some

M we have:
if i
0
= n i
1
< i
2
i
3
, g
0
pos(

t[i
0
, i
1
)), g
2
pos(

t[i
2
, i
3
)) and
s (t
i1
, . . . , t
i21
), nor[s] M, then B g
0
g
1
g
2
: g
1
val[s] , = .
(2) suf

t
(K, ) is the family of all

D = D
n
: n < ) suf
t
(K, ) such that
(
n

t
D
n
for all n < .
Proposition 4.4. Let (K, ) be a tight FFCC pair for H and

t PC
tt

(K, ).
(1) suf

t
(K, ) is a closed subset of suf
t
(K, ), suf

t
(K, ) suf

t
(K, ).
(2) If (K, ) has the very weak tbigness for

t, then suf

t
(K, ) ,= .
(3) If

D suf

t
(K, ), n < and B (
n

t
, then set
n

t
(B,

D) = pos
tt
(

t n).
(4) suf

t
(K, ) is closed under the operation (dened in 3.1(6)).
Proof. (1) Since in 4.3(1) we allow i
1
= i
0
and i
3
= i
2
(so g
0
= g
2
= ), we easily
see that (
n

t
B
n

t
. Hence suf

t
(K, ) suf

t
(K, ). The proof that suf

t
(K, ) is
closed is the same as for 3.5(1).
(2) Like 3.5(2).
(3) Let M be such that B g
0
g
1
g
2
: g
1
val[s] ,= whenever g
0

pos(

t[n, i
1
)), g
2
pos(

t[i
2
, i
3
)), s (

t[i
1
, i
2
)), nor[s] M, n i
1
< i
2
i
3
.
We will show that this M witnesses f B (
x
f

t
for all f pos
tt
(t n).
So suppose that f pos
tt
(

t n) and x
f
i
1
< i
2
i
3
, g
0
pos(

t[x
f
, i
1
)),
s (

t[i
1
, i
2
)), nor[s] M and g
2
pos(

t[i
2
, i
3
)). Then f g
0
pos(

t[n, i
1
))
(remember 3.2(1)) and consequently (by the choice of M) B
_
(f g
0
) g
1
g
2
:
g
1
val[s]
_
,= . Let g

1
val[s] be such that f g
0
g

1
g
2
B. Then
g
0
g

1
g
2
f B witnessing that (f B)
_
g
0
g
1
g
2
: g
1
val[s]
_
,= .
Since (
x
f

t
D
x
f
we conclude now that f B D
x
f
so f set
n

t
(B,

D).
(4) Suppose

D
1
,

D
2
suf

t
(K, ),

D =

D
1


D
2
. Let B (
n

t
, n < . By (3)
we know that set
n

t
(B,

D
1
) = pos
tt
(

t n) D
2
n
and thus B D
n
. Consequently,
(
n

t
D
n
for all n < , so

D suf

t
(K, ).
Corollary 4.5. Assume that (K, ) is a tight FFCC pair with the very weak t
bigness for

t PC
tt

(K, ). Then there is

D suf

t
(K, ) such that

D

D =

D.
Theorem 4.6. Assume that (K, ) is a tight FFCC pair for H,

t = t
n
: n <
) PC
tt

(K, ). Let

D suf

t
(K, ) be such that

D

D =

D and suppose that
A
n
D
n
for n < . Then there are sequences n
i
: i < ), g
3i
, g
3i+2
: i < ) and
s
3i+1
: i < ) such that for every i < :
() 0 = n
0
n
3i
n
3i+1
< n
3i+2
n
3i+3
< ,
() if j = 3i or j = 3i + 2, then g
j
pos(

t[n
j
, . . . , n
j+1
)),
() if j = 3i + 1, then s
j
(t
nj
, . . . , t
nj+11
) and nor[s
j
] j,
9
5
7


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







m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
3
-
2
3


PARTITION THEOREMS FROM CREATURES AND IDEMPOTENT ULTRAFILTERS 17
() if g
3+1
val[s
3+1
] for [i, k), i < k, then
3k1

j=3i
g
j
A
n3i
.
Proof. Parallel to 3.7, just instead of val[s
i
] use g
i1
g g
i+1
: g val[s
i
].
Conclusion 4.7. Assume that (K, ) is a tight FFCC pair with the very weak t
bigness for

t PC
tt

(K, ). Suppose that for each n < we are given k

n
<
and a mapping d
n
: pos
tt
(

t n) k
n
. Then there are sequences n
i
: i < ),
g
3i
, g
3i+2
: i < ), s
3i+1
: i < ) and c
i
: i < ) such that for each i < :
() 0 = n
0
n
3i
n
3i+1
< n
3i+2
n
3i+3
< , c
i
k
n3i
,
() if j = 3i or j = 3i + 2, then g
j
pos(

t[n
j
, . . . , n
j+1
)),
() if j = 3i + 1, then s
j
(t
nj
, . . . , t
nj+11
) and nor[s
j
] j,
() if i < k and f pos(

t[n
3i
, n
3k
)) are such that
g
3
g
3+2
f and f[m
s
3+1
dn
, m
s
3+1
up
) val[s
3+1
] for all [i, k),
then d
n3i
(f) = c
i
.
Example 4.8. Let (G, ) be a nite group. For a function f : S G and a G
we dene a f : S G by (a f)(x) = a f(x) for x S. Let H
G
(m) = G (for
m < ) and let K
G
consist of all FP creatures t for H
G
such that
nor[t] = m
t
up
, dis[t] = ,
val[t]
[m
t
dn
,m
t
up
)
G is such that (f val[t])(a G)(a f val[t]).
For t
0
, . . . , t
n
K
G
with m
t
+1
dn
= m
t

up
(for < n) we let
G
(t
0
, . . . , t
n
) consist of
all creatures t K
G
such that
m
t
dn
= m
t0
dn
, m
t
up
= m
tn
up
,
val[t] f
[m
t
dn
,m
t
up
)
G : ( n)(f[m
t

dn
, m
t

up
) val[t

]).
Then
(1) (K
G
,
G
) is a tight FFCC pair for H
G
.
(2) If [G[ = 2, then (K
G
,
G
) has the very weak tbigness for every candidate

t PC
tt

(K
G
,
G
).
Proof. (1) Straightforward.
(2) Let G = (1, 1, ). Suppose that

t PC
tt

(K
G
,
G
) and pos
tt
(

t n) =
T
0
. . . T
L
, n, L, M < . For future use we will show slightly more than needed
for the very weak bigness.
We say that N n +M is good (for L) if
()

there are j
2
j
1
> N, g
0
pos(

t[n, N)), g
2
pos(

t[j
1
, j
2
)) and s

G
(

t[N, j
1
)) such that g
0
g
1
g
2
: g
1
val[s] T

.
(Note that if s is as in ()

, then also nor[s] = m

tj
1
dn
j
1
> N M.) We are
going to argue that
() almost every N n +M is good for some L.
So suppose that () fails and we have an increasing sequence n + M < N(0) <
N(1) < N(2) < . . . such that N(k) is not good for any L (for all k < ). Let
m = L+957 and for each i [n, N(m)] x f
i
val[t
i
] (note that then f
i
val[t
i
]
as well). Next, for j < m dene
h
j
=
N(j)1
_
i=n
f
i

N(m)
_
i=N(j)
f
i
9
5
7


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







m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
3
-
2
3


18 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
and note that h
j
pos
tt
(

t n). For some L and j < k < m we have

h
j
, h
k
T

. Set
g
0
=
N(j)1
_
i=n
f
i
= h
j
m
t
N(j)
dn
= h
k
m
t
N(j)
dn
,
g
2
=
N(m)
_
i=N(k)
f
i
= h
j
[m
t
N(k)
dn
, m
t
N(m)
up
) = h
k
[m
t
N(k)
dn
, m
t
N(m)
up
)
and let s
G
(

t[N(j), N(k))) be such that

val[s] = h
j
[m
t
N(j)
dn
, m
t
N(k)
dn
), h
k
[m
t
N(j)
dn
, m
t
N(k)
dn
).
Then g
0
g
1
g
2
: g
1
val[s] = h
j
, h
k
T

, so g
0
, g
2
and s witness ()

for
N(j), a contradiction.
The following conclusion is a special case of the partition theorem used in Gold-
stern and Shelah [3] to show that a certain forcing notion preserves a Ramsey
ultralter (see [3, 3.9, 4.1 and Section 5]).
Corollary 4.9. Let Y =

n<
n
1, 1. Suppose that Y = C
0
. . . C
L
, L < .
Then there are a sequence n
i
: i < ), a function f : 1, 1 and < L
such that
(a) 0 = n
0
n
3i
n
3i+1
< n
3i+2
n
3i+3
< ,
(b) if g : n
3i
1, 1 for each j < i satises
g[n
3j
, n
3j+1
) g[n
3j+2
, n
3j+3
) f and
g[n
3j+1
, n
3j+2
) f[n
3j+1
, n
3j+2
), f[n
3j+1
, n
3j+2
)
then g C

.
Proof. By 4.7+4.8.
5. Limsup candidates
Denition 5.1. Let (K, ) be a tight FFCC pair for H and J be an ideal on .
(1) A limsup
J
candidate for (K, ) is a sequence

t = t
n
: n < ) such that
t
n
K, m
tn
up
= m
tn+1
dn
(for all n) and for each M
m
tn
dn
: n < & nor[t
n
] > M J
+
.
The family of all limsup
J
candidates for (K, ) is denoted by PC
J
w
(K, ).
(2) A nite candidate for (K, ) is a nite sequence s = s
n
: n < N), N < ,
such that s
n
K and m
sn
up
= m
sn+1
dn
(for n < N). The family of all nite
candidates is called FC(K, ).
(3) For s = s
n
: n < N) FC(K, ) and M < we set
base
M
( s) = m
sn
dn
: n < N & nor[s
n
] M.
(4) Let

t,

PC
J
w
(K, ), s FC(K, ). Then we dene

t n,
tt
(

t),
pos
tt
(

t),

t

t

, pos(

t[n, m)) and pos( s) as in the case of tight pure candi-

dates (cf. 1.3, 4.1).
9
5
7


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







m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
3
-
2
3


PARTITION THEOREMS FROM CREATURES AND IDEMPOTENT ULTRAFILTERS 19
(5) Let

t PC
J
w
(K, ). The family of all nite candidates s = s
n
: n <
N) FC(K, ) satisfying
(n < N)(k, )(s
n
(

t[k, ))) and m

s0
dn
= m
t0
dn
is denoted by
seq
(

t).
Denition 5.2. Let (K, ) be a tight FFCC pair, J be an ideal on and

t
PC
J
w
(K, ).
(1) We say that (K, ) has the Jbigness for

t if
()
J

t
for every n, L, M < and a partition T
0
. . . T
L
= pos
tt
(

t n),
there are L and a set Z J
+
such that
(z Z)( s
seq
(

t n))(z base
M
( s) & pos( s) T

).
(2) The pair (K, ) captures singletons (cf. [6, 2.1.10]) if
(t K)(f val[t])(s (t))(val[s] = f).
(3) We dene suf
t
(K, ) as in 3.1(3), set
n

t
(A,

D) (for A pos
tt
(

t n) and

D suf
t
(K, )) as in 3.1(5) and the operation on suf
t
(K, ) as in 3.1(6).
(4) For n < , T
n,J

t
is the family of all sets B pos
tt
(

t n) such that for

some M < and Y J
c
we have:
if s
seq
(

t n) and base
M
( s) Y ,= , then B pos( s) ,= .
(5) suf
J

t
(K, ) is the family of all

D = D
n
: n < ) suf
t
(K, ) such that
T
n,J

t
D
n
for all n < .
Remark 5.3. Note that no norms were used in the proofs of 3.2, 3.3, so those
statements are valid for the case of

t PC
J
w
(K, ) too.
Observation 5.4. (1) Assume that (K, ) is a tight FFCC pair with bigness
(see 1.6(3)). If (K, ) captures singletons or it has the tmultiadditivity
(see 1.5(3)), then (K, ) has the Jbigness for any

t PC
J
w
(K, ).
(2) The tight FFCC pairs (K
1
,

1
), (K
3
,

3
) and (K
N
,
N
) dened in 1.10,
1.12 and 1.13, respectively, have Jbigness on every

t PC
J
w
(K, ).
Every tight FFCC pair can be extended to a pair capturing singletons while
preserving pos
tt
(

t).
Denition 5.5. Let (K, ) be a tight FFCC pair for H. Dene K
sin
as the family
of all FP creatures t for H such that
dis[t] = K, nor[t] = 0 and [val[t][ = 1.
Then we let K
s
= K K
sin
and for t
0
, . . . , t
n
K
s
with m
t

up
= m
t
+1
dn
(for < n)
we set

sin
(t
0
, . . . , t
n
) consists of all creatures t K
sin
such that m
t
dn
= m
t0
dn
,
m
t
up
= m
tn
up
and
if val[t] = f then f[m
t

dn
, m
t

up
) val[t

] for all n;
if t
0
, . . . , t
n
K, then
s
(t
0
, . . . , t
n
) = (t
0
, . . . , t
n
)
sin
(t
0
, . . . , t
n
);
if t

K
sin
for some n, then
s
(t
0
, . . . , t
n
) =
sin
(t
0
, . . . , t
n
).
Observation 5.6. Let (K, ) be a tight FFCC pair for H.
(1) (K
s
,
s
) is a tight FFCC pair for H and it captures singletons.
(2) If (K, ) has bigness then so does (K
s
,
s
) and consequently then (K
s
,
s
)
has the Jbigness on any

t PC
J
w
(K
s
,
s
).
9
5
7


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







m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
3
-
2
3


20 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
(3) If

t PC
J
w
(K, ), then

t PC
J
w
(K
s
,
s
) and pos
tt
(

t) with respect to
(K, ) is the same as pos
tt
(

t) with respect to (K
s
,
s
).
Observation 5.7. Let G = (1, 1, ) and (K
G
,
G
) be the tight FFCC pair
dened in 4.8. Suppose that

t PC
J
w
(K
G
,
G
). Then (K
s
G
,
s
G
) (sic!) has the
Jbigness for

t.
Proof. Note that PC
J
w
(K
G
,
G
) PC
tt

(K
G
,
G
) and remember () from the
proof of 4.8(2).
Proposition 5.8. Assume that (K, ) is a tight FFCC pair for H, J is an ideal
on and

t PC
J
w
(K, ).
(1) suf
J

t
(K, ) is a closed subset of the compact Hausdor topological space
suf
t
(K, ).
(2) If (K, ) has the Jbigness for

t, then suf
J

t
(K, ) ,= .
(3) If

D suf
J

t
(K, ), n < and B T
n,J

t
, then set
n

t
(B,

D) T
n,J

t
.
(4) suf
J

t
(K, ) is closed under the operation .
Proof. (1) Same as 3.5(1).
(2) Similar to 3.5(2).
(3) Let B T
n,J

t
be witnessed by M < and Z J
c
. We are going to show that
then for each s
seq
(

t n) with base
M
( s)Z ,= we have pos( s)set
n

t
(B,

D) ,= .
So let s = s
0
, . . . , s
k
)
seq
(

t n), base
M
( s) Z ,= and let x be such that
m
s
k
up
= m
tx
dn
. Set A =

f B : f pos( s). Suppose that r

seq
(

t x).
Then s

r
seq
(

t n) and base
M
( s

r) base
M
( s), so pos( s

r) B ,= . Let
g pos( s

r) B and f
0
= gm
s
k
up
, f
1
= g[m
s
k
up
, ). Necessarily f
0
pos( s),
f
1
pos( r) and (as g = f
0
f
1
B) f
1
f
0
B. Consequently A pos( r) ,= .
Now we easily conclude that A T
x,J

t
D
x
. Hence for some f pos( s) we have
f B D
x
, so f set
n

t
(B,

D).
(4) Follows from (3).
Corollary 5.9. Assume that (K, ) is a tight FFCC pair, J is an ideal on and

t PC
J
w
(K, ). If (K, ) has the Jbigness for

t, then there is

D suf
J

t
(K, )
such that

D

D =

D.
Denition 5.10. Let J be an ideal on .
(1) A game
J
between two players, One and Two, is dened as follows. A play
of
J
lasts steps in which the players construct a sequence Z
i
, k
i
: i < ).
At a stage i of the play, rst One chooses a set Z
i
J
+
and then Two
answers with k
i
Z
i
. At the end, Two wins the play Z
i
, k
i
: i < ) if and
only if k
i
: i < J
+
.
(2) We say that J is an Rideal if player One has no winning strategy in
J
.
Remark 5.11. If J is a maximal ideal on , then it is an Rideal if and only if the
dual lter J
c
is a Ramsey ultralter. Also, the ideal []
<
of all nite subsets of
is an Rideal.
Theorem 5.12. Assume that (K, ) is a tight FFCC pair, J is an Rideal on
and

t PC
J
w
(K, ). Suppose that

D suf
J

t
(K, ) satises

D

D =

D and let
A
n
D
n
for n < . Then there are s PC
J
w
(K, ) and 0 = k(0) < k(1) <
k(2) < k(3) < . . . < such that

t s, m
s0
dn
= m
t0
dn
and
if i < j, < , s
k(i)

tt
(

t ), then pos(s
k(i)
, s
k(i)+1
, . . . , s
k(j)1
) A

.
9
5
7


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







m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
3
-
2
3


PARTITION THEOREMS FROM CREATURES AND IDEMPOTENT ULTRAFILTERS 21
Proof. The proof follows the pattern of 3.7 with the only addition that we need to
make sure that at the end s PC
J
w
(K, ), so we play a round of
J
. First,
Claim 5.12.1. Assume M, < and B D

. Then for some set Z J

+
, for
every x Z, there is s
seq
(

t ) such that
x base
M
( s), pos( s) B and (f pos( s))(f B D
x
f
).
Proof of the Claim. Similar to 3.7.1. Since

D

D =

D and B D

, we know that
pos
tt
(

t ) (B set

t
(B,

D)) / T
,J

t
. Therefore, for each Y J
c
there are x Y
and s
seq
(

t ) such that x base

M
( s) and pos( s) B set

t
(B,

D). So the
set Z of x as above belongs to J
+
.
Consider the following strategy for player One in the game
J
. During the course
of a play, in addition to his innings Z
i
, One chooses aside
i
< , B
i
D
i
and
s
i
FC(K, ). So suppose that the players have arrived to a stage i of the play
and a sequence Z
j
, k
j
, s
j
,
j
, B
j
: j < i) has been constructed. Stipulating
1
= 0
and B
1
= A
0
, One uses 5.12.1 to pick a set Z
i
m
t

i1
dn
such that Z
i
J
+
and for all x Z
i
there exists s
seq
(

t
i1
) with
x base
i+1
( s) & pos( s) B
i1
& (f pos( s))(f B
i1
D
x
f
).
The set Z
i
is Ones inning in
J
after which Two picks k
i
Z
i
. Now, One chooses
s
i

seq
(

t
i1
) such that
()
i
k
i
base
i+1
( s
i
),
()
i
pos( s
i
) B
i1
, and
()
i
(f pos( s
i
))(f B
i1
D
x
f
).
He also sets
()
i

i
= x
f
for all (equivalently: some) f pos( s
i
), and
()
i
B
i
= A
i

f B
i1
: f pos( s
i
) D
i
.
The strategy described above cannot be winning for One, so there is a play
Z
i
, k
i
: i < ) in which One follows the strategy, but k
i
: i < J
+
. In the
course of this play One constructed aside a sequence
i
, B
i
, s
i
: i < ) such that
s
i

seq
(

t
i1
) and conditions ()
i
()
i
hold (where we stipulate
1
= 0,
B
1
= A
0
). Note that s
i
s
i+1
. . .

s
i+k

seq
(

t
i1
) for each i, k < . Also
s
def
= s
0
s
1
s
2
. . . PC
J
w
(K, ) and s

t.
Claim 5.12.2. For each i, k < , pos( s
i
s
i+1
. . .

s
i+k
) B
i1
A
i1
.
Proof of the Claim. Induction on k; fully parallel to 3.7.2.
Now the theorem readily follows.
Conclusion 5.13. Assume that (K, ) is a tight FFCC pair, J is an Rideal on
and

t PC
J
w
(K, ). Suppose also that (K, ) has Jbigness for

t. For n <
let k
n
< and let d
n
: pos
tt
(

t n) k
n
. Then there are s PC
J
w
(K, ) and
0 = k(0) < k(1) < k(2) < . . . < and c
i
: i < ) such that

t s, m
s0
dn
= m
t0
dn
, and
for each i, n < ,
if s
k(i)

tt
(

t n), i < j < and f pos(s

k(i)
, s
k(i)+1
, . . . , s
k(j)1
),
then d
n
(f) = c
i
.
9
5
7


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







m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
3
-
2
3


22 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
Corollary 5.14. Let H

: 0 be increasing, Z

n<

i<n
H

(i) and let

J be an Rideal on . Suppose that Z

= C
0
. . . C
L
, L < . Then there are
sequences k
i
, n
i
: i < ) and E
i
: i < ) and L such that
(a) 0 = n
0
k
0
< n
1
. . . < n
i
k
i
n
i+1
. . . < , k
i
: i < J
+
,
and for each i < :
(b) ,= E
i
H

(i), [E
ki
[ = i + 1, and

j<ni
E
j
C

.
References
[1] Timothy J. Carlson and Stephen G. Simpson. A dual form of ramseys theorem. Adv. in Math.,
53:265290, 1984.
[2] W. Wistar Comfort. Ultralters: some old and some new results. Bulletin of the American
Mathematical Society, 83:417455, 1977.
[3] Martin Goldstern and Saharon Shelah. Ramsey ultralters and the reaping numberCon(r <
u). Annals of Pure and Applied Logic, 49:121142, 1990.
[4] Jakob Kellner and Saharon Shelah. Creature forcing and large continuum. The joy of halving.
Archive for Mathematical Logic, accepted. 1003.3425.
[5] Jakob Kellner and Saharon Shelah. Decisive creatures and large continuum. Journal of Sym-
bolic Logic, 74:73104, 2009. math.LO/0601083.
[6] Andrzej Roslanowski and Saharon Shelah. Norms on possibilities I: forcing with trees
and creatures. Memoirs of the American Mathematical Society, 141(671):xii + 167, 1999.
math.LO/9807172.
[7] Andrzej Roslanowski and Saharon Shelah. Measured creatures. Israel Journal of Mathematics,
151:61110, 2006. math.LO/0010070.
[8] Andrzej Roslanowski, Saharon Shelah, and Otmar Spinas. Nonproper Products. Bulletin of
the London Mathematical Society, submitted. 0905.0526.
[9] Saharon Shelah and Jindrich Zapletal. Ramsey theorems for product of nite sets with sub-
measures. Combinatorica, accepted.
Department of Mathematics, University of Nebraska at Omaha, Omaha, NE 68182-
0243, USA
E-mail address: roslanow@member.ams.org
URL: http://www.unomaha.edu/logic
Einstein Institute of Mathematics, Edmond J. Safra Campus, Givat Ram, The He-
brew University of Jerusalem, Jerusalem, 91904, Israel, and, Department of Mathe-
matics, Hill Center - Busch Campus, Rutgers, The State University of New Jersey, 110
Frelinghuysen Road, Piscataway, NJ 08854-8019 USA
E-mail address: shelah@math.huji.ac.il
URL: http://shelah.logic.at