9

7
2


r
e
v
i
s
i
o
n
:
2
0
1
0
-
0
7
-
3
0







m
o
d
i
f
i
e
d
:
2
0
1
0
-
0
7
-
3
0


MONOTONE HULLS FOR ^ /
ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
Abstract. Using the method of decisive creatures (see Kellner and Shelah [6])
we show the consistency of there is no increasing
2
chain of Borel sets and
non(N) = non(M) =
2
= 2

. Hence, consistently, there are no monotone

Borel hulls for the ideal M N. This answers Balcerzak and Filipczak [1,
Questions 23, 24]. Next we use FS iteration with partial memory to show that
there may be monotone Borel hulls for the ideals M, N even if they are not
generated by towers.
0. Introduction
Brendle and Fuchino [3, Section 3] considered the following spectrum of cardinal
numbers
DO =
_
cf(otp(X, RX))) : R

2

2 is a projective binary relation,

X

2 and R X
2
is a well ordering of X
_
and they introduced a cardinal invariant do = sup DO. The invariant do satises
minnon(1), cov(1) do for every ideal 1 on R with Borel basis (see [3, Lemma
3.6]). The proof of Kunen [7, Theorem 12.7] essentially shows that adding any
number of Cohen (or random) reals to a model of CH results in a model in which
do =
1
. Thus both
non(^) = cov(/) =
2
+ non(/) = cov(^) = do =
1
, and
non(/) = cov(^) =
2
+ non(^) = cov(/) = do =
1
are consistent (where /, ^ stand for the ideals of meager and null sets, respec-
tively). This naturally leads to the question if
() non(/) = non(^) =
2
+ do =
1
(= cov(^) = cov(/))
is consistent. In this note we show the consistency of () using the method of
decisive creatures developed in Kellner and Shelah [6], and this method is in turn a
special case of the method of norms on possibilities of Roslanowski and Shelah [9].
Note that if there is a increasing chain of Borel subsets of

2, then cf()
DO. (Just consider a relation R on

2

2

2 given by: (x, y) R (x

, y

) if
and only if y, y

are Borel codes and x belongs to the set coded by y

; cf. Elekes
and Kunen [4, Lemma 2.4].) Thus if we set
d
B
= sup
_
cf() : there is a increasing chain of Borel subset of R of length
_
then d
B
do. If d
B
is smaller than the conality of the uniformity number non(1)
of a Borel ideal 1, then there is no monotone Borel hull operation on 1 (see Elekes
and M athe [5, Theorem 2.1], Balcerzak and Filipczak [1, Theorem 5]). Thus
Date: July 29, 2010.
1991 Mathematics Subject Classication. Primary 03E17; Secondary: 03E35, 03E15.
Both authors acknowledge support from the United States-Israel Binational Science Foundation
(Grant no. 2002323). This is publication 972 of the second author.
1
9
7
2


r
e
v
i
s
i
o
n
:
2
0
1
0
-
0
7
-
3
0







m
o
d
i
f
i
e
d
:
2
0
1
0
-
0
7
-
3
0


2 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
() if 1 is an ideal with Borel basis on R, d
B
< non(1) and non(1) is a regular
cardinal, then there is no monotone mapping : 1 Borel(R) 1.
Therefore in our model for () we will have (Corollary 3.2)
there are no monotone Borel hull operations on the ideals /, ^ and /^.
This answers Balcerzak and Filipczak [1, Question 23].
We also obtain a positive result providing a new situation in which monotone
hulls exist. Consistently, the ideals /, ^ do not possess towerbasis but they do
admit monotone Borel hulls (Corollary 3.8).
Notation Most of our notation is standard and compatible with that of classi-
cal textbooks (like Bartoszy nski Judah [2]). However in forcing we keep the older
convention that a stronger condition is the larger one.
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(). A tree is a family T of nite sequences closed under initial segments. For
a tree T, the family of all branches through T is denoted by [T].
The Cantor space

2 is the space of all functions from to 2, equipped with
the product topology generated by sets of the form [] =

2 : for

>
2. This space is also equipped with the standard product measure .
For a forcing notion P, all Pnames for objects in the extension via P will be
denoted with a tilde below (e.g. A

). The canonical name for a Pgeneric lter

over V is denoted G

P
. Our notation and terminology concerning creatures and
forcing with creatures will be compatible with that in [6] (except of the reversed
orders). While this is a slight departure from the original terminology established
for creature forcing in [9], the reader may nd it more convenient when verifying
the results on decisive creatures that are quoted in the next section.
1. Background on decisive creatures
As declared in the introduction, we will follow the notation and the context of
[6] (which slightly diers from that of [9]). For readers convenience we will recall
here all relevant denitions and results from that paper.
Let H : H(
0
) (where H(
0
) is the family of all hereditarily nite sets).
A creating pair for H is a pair (K, ), where
K =

n<
K(n), where each K(n) is a nite set; elements of K are called
creatures, each creature c K(n) has some norm nor(c) (a non-negative
real number) and a non-empty set of possible values val(c) H(n),
if c K(n), nor(c) > 0, then [val(c)[ > 1
: K T(K) is such that if c K(n) and c

(c), then c

K(n),
c (c) and c

(c) implies (c

) (c),
if c

(c), then nor(c

) nor(c) and val(c

) val(c).
If c K and x H(n), then we write x (c) if and only if x val(c). For
x H(n) we also set (x) = val(x) = x.
Denition 1.1 (See [6, Denitions 3.1, 4.1]). Let 0 < r 1, B, K, m be positive
integers and (K, ) be a creating pair for H.
(1) A creature c is rhalving if there is a half(c) (c) such that
nor(half(c)) nor(c) r, and
9
7
2


r
e
v
i
s
i
o
n
:
2
0
1
0
-
0
7
-
3
0







m
o
d
i
f
i
e
d
:
2
0
1
0
-
0
7
-
3
0


MONOTONE HULLS FOR N M 3
if d (half(c)) and nor(d) > 0, then there is a d

(c) such that

nor(d

) nor(c) r and val(d

) val(d).
K(n) is rhalving, if all c K(n) with nor(c) > 1 are rhalving.
(2) A creature c is (B, r)big if for every function F : val(c) B there is
a d (c) such that nor(d) nor(c) r and the restriction Fval(d) is
constant. We say that c is hereditarily (B, r)-big, if every d (c) with
nor(d) > 1 is (B, r)-big. Also, K(n) is (B, r)big if every c K(n) with
nor(c) > 1 is (B, r)big.
(3) We say that c is (K, m, r)decisive, if for some d

, d
+
(c) we have:
d
+
is hereditarily (2
K
m
, r)big, and [val(d

)[ K and nor(d

), nor(d
+
)
nor(c) r. c is (m, r)decisive if c is (K

, m, r)decisive for some K

.
(4) K(n) is (m, r)decisive if every c K(n) with nor(c) > 1 is (m, r)decisive.
Lemma 1.2 (See [6, Lemma 4.3]). Assume that (K, ) is a creating pair for H,
k, m, t 1, 0 < r 1. Suppose that K(n) is (k, r)decisive and c
0
, . . . , c
k1
K(n)
are hereditarily (2
m
t
, r)big with nor(c
i
) > 1 + r (k 1) (for each i < k). Let
F :

i<k
val(c
i
) 2
m
t
. Then there are d
0
, . . . , d
k1
K(n) such that:
d
i
(c
i
), nor(d
i
) nor(c
i
) r k, and F

ik
val(d
i
) is constant.
A creating pair (K, ) determines the forcing notion Q

(K, ) and its special

product P
I
(K, ) as described by the following denition. (The forcing notion
P
I
(K, ) is a relative of the CS product of Q

(K, ) indexed by the set I.)

Denition 1.3 (See [6, Denitions 2.1, 5.2, 5.3]). (1) A condition in the forc-
ing Q

(K, ) is an sequence p = p(i) : i < ) such that for some n <

(called the trunk-length of p) we have p(i) H(i) if i < n, p(i) K(i) and
nor(p(i)) > 0 if i n, and lim
t
(nor(p(i))) = .
The order on Q

(K, ) is dened by q p if and only if (both belong

to Q

(K, ) and) q(i) (p(i)) for all i.

1
(2) Let I be a non-empty (index) set. A condition p in P
I
(K, ) consists of a
countable subset dom(p) of I, of objects p(, n) for dom(p), n , and
of a function trunklg(p, ) : dom(p) satisfying the following demands
for all dom(p):
() If n < trunklg(p, ), then p(, n) H(n).
() If n trunklg(p, ), then p(, n) K(n) and nor(p(, n)) > 0.
() Setting supp(p, n) = dom(p) : trunklg(p, ) n, we have
[supp(p, n)[ < n for all n > 0 and lim
n
([supp(p, n)[/n) = 0.
() lim
n
(min(nor(p(, n)) : supp(p, n))) = .
The order on P
I
(K, ) is dened by q p if and only if (both belong to
P
I
(K, ) and) dom(q) dom(p) and
() if dom(p) and n , then q(, n) (p(, n)),
() the set dom(p) : trunklg(q, ) ,= trunklg(p, ) is nite.
Note that for dom(p) the sequence p(, n) : n ) is in Q

(K, ).
However, P
I
(K, ) is not a subforcing of the CS product of I copies of Q

(K, )
because of a slight dierence in the denition of the order relation.
1
Remember our convention that for x H(i) and c K(i) we write x (c) i x val(c)
9
7
2


r
e
v
i
s
i
o
n
:
2
0
1
0
-
0
7
-
3
0







m
o
d
i
f
i
e
d
:
2
0
1
0
-
0
7
-
3
0


4 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
Proposition 1.4 (See [6, Lemmas 5.4, 5.5]). (1) If J I, then P
J
(K, ) =
p P
I
(K, ) : dom(p) J is a complete subforcing of P
I
(K, ).
(2) Assume CH. Then P
I
(K, ) satises the
2
chain condition.
Denition 1.5 (See [6, Denition 5.6]). (1) For a condition p P
I
(K, ) we
dene
2
val

(p, <n) =

dom(p)

m<n
val(p(, m)).
(2) If w dom(p) and t

m<n
H(m), then p t is dened by
trunklg(p t, ) =
_
max(trunklg(p, ), n) if w,
trunklg(p, ) otherwise
and
(p t)(, m) =
_
t(, m) if m < n and w,
p(, m) otherwise.
(3) If

is a name of an ordinal, then we say that p <ndecides

, if for every
t val

(p, <n) the condition p t forces a value to

. The condition p
essentially decides

, if p <n-decides

for some n.
Proposition 1.6. (1) p t P, and if t val

(p, <n), then p t p.

(2) val

(p, <n)

m<n
[H(m)[
m
.
(3) p t : t val

(p, <n) is predense above p

Theorem 1.7 (See [6, Theorems 5.8, 5.9]). Let (<n) =

m<n
[H(m)[
m
and r(n) =
1/(n
2
(<n)). Assume that each K(n) is (n, r(n))-decisive and r(n)halving (for
n ).
(1) The forcing notion P
I
(K, ) is proper and

-bounding. If [I[ 2 and
= [I[
0
, then P
I
(K, ) forces [I[ 2
0
.
(2) Moreover, if

(n) is a P
I
(K, )name for an ordinal (for n < ) and
p P
I
(K, ), then there is a condition q p which essentially decides all
the names

(n).
(3) Assume, additionally, that each K(n) is (g(n), r(n)) big, where g

is strictly increasing. Suppose that

(n) is a P
I
(K, )name and p
P
I
(K, ) forces that

(n) < 2
g(n)
for all n < . Then there is a q p
which <ndecides

(n) for all n.

The next theorem is a consequence of (the proof of) [3, Corollaries 4.8(e), 3.9(b)].
However, the results in [3] are stated for products, while P
I
(K, ) is not exactly a
product (though it does have all the required features). Therefore we will present
the relatively simple proof of this result fully.
Theorem 1.8. Assume CH. Let r, , K and be as in the assumptions of Theorem
1.7. Then
PI(K,)
do = d
B
=
1
.
Proof. Without loss of generality, [I[
2
.
Every bijection : I
onto
I determines an automorphism of the forcing
P
I
(K, ) in a natural way. Then, for J I, P
J
(K, ) is an isomorphism
2
Remember our convention that, for x H(i), val(x) = {x}
9
7
2


r
e
v
i
s
i
o
n
:
2
0
1
0
-
0
7
-
3
0







m
o
d
i
f
i
e
d
:
2
0
1
0
-
0
7
-
3
0


MONOTONE HULLS FOR N M 5
from P
J
(K, ) onto P
[J]
(K, ). Also, gives rise to a natural bijection from
val

(p, < n) onto val

( (p), < n); we will denote this mapping by as well.

Suppose that (x, y,

) is a projective denition of a binary relation on

2,
where

is a P
I
(K, )name for a real parameter. Assume towards contradiction
that there are P
I
(K, )names

(for <
2
) and a condition p P
I
(K, ) such
that
(i) p
PI(K,)

_
, <
2
__
(

) <
_
.
For each <
2
choose a condition p

p which essentially decides all

(n) (for
n < ). Then we may also pick an increasing sequence

N

= N

n
: n < )
and a mapping f

n<
val

(p

, <N

n
) such that for each t val

(p

, <N

n
)
we have (p

t)

(n) = f

(t).
Since P
I
(K, ) satises the
2
cc we may choose a set J I of size
1
such
that dom(p) J and

is a P
J
(K, )name (see 1.4).
Now, by CH and a standard system argument, we may nd a set X [
2
]

2
and bijections
,
: dom(p

)
onto
dom(p

) for , X such that

(ii) dom(p

) J = dom(p

) J and
,
(dom(p

) J) is the identity,
(iii)
,
(p

) = p

,

N

=

N

, and f

= f

,
.
Pick < from X. Let be a bijection from I onto I such that
,
,
(
,
)
1
and J is the identity. Then
(iv) (p

) = p

, (p

) = p

and (

) =

.
The conditions p

, p

are compatible, so let q be a condition stronger than both

of them. Note that p

does not have to be a condition in P

I
(K, ) as the
demand 1.3(2)() may fail. But extending nitely many trunks well easily resolve
this problem. We may even do this in such a manner that the resulting condition
q will also satisfy (q) = q. Since q p

, p

and by (iii) we have

(v) q (

) =

& (

) =

.
Since q p and < we have q (

). Applying the automorphism and

remembering (v) we conclude that then also (q) = q (

), contradicting
(i).
2. Consistency of do < non(/ ^)
Denition 2.1. Let n < .
(1) A basic nblock is a nite set B of functions from some non-empty v
[]
<
to 2 (i.e., B
v
2) such that [B[/2
|v|
< 2
n
. If
>
2

2 and
B
v
2 is a basic block, then we write B whenever v B. For an
nblock B
v
2 we set v(B) = v.
(2) Let H
n
be the family of all pairs (b, B) such that b is a positive integer and
B is a non-empty nite set of basic nblocks.
(3) We dene a function pnor : H
n
by declaring inductively when
pnor(b, B) k. We set pnor(b, B) 0 always, and then
pnor(b, B) 1 if and only if (F [

2]
b
)(B B)( F)( B),
pnor(b, B) k+1 if and only if there are positive integers b
0
, . . . , b
M1
and disjoint sets B
0
, . . . , B
M1
B such that
() M > b
k+1
, b
0
b and
() pnor(b
i
, B
i
) k and (b
i
)
2
2
|Bi|
n
< b
i+1
for all i < M.
9
7
2


r
e
v
i
s
i
o
n
:
2
0
1
0
-
0
7
-
3
0







m
o
d
i
f
i
e
d
:
2
0
1
0
-
0
7
-
3
0


6 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
Proposition 2.2. Let n < , (b, B), (b

, B

) H
n
.
(1) pnor(b, B) is well dened and 2
pnor(b,B)
[B[.
(2) If B B

and b

b, then pnor(b, B) pnor(b

, B

).
(3) For each N there is (b

, B

) H
n
such that
b

N and pnor(b

, B

) N and min(v(B)) > N for all B B

.
(4) If pnor(b, B) k + 1 2 and c : B 0, . . . , b 1, then for some < b
we have pnor(b, c
1
[]) k.
Proof. (1,2) Easy induction on pnor(b, B).
(3) Note that if w []
<
, 2
n
N < 2
|w|
and B
w
consists of all basic nblocks B
with v(B) = w, then pnor(N, B
w
) 1. Now proceed inductively.
(4) Induction on k 1. Assume pnor(b, B) 2 and c : B b. We claim that
for some < b we have pnor(b, c
1
[]) 1. If not, then for each < b we may
choose F

2]
b
such that
_
B B
__
F

__
c(B) = B
_
.
Set F =

<b
F

. Let b
0
, . . . , b
M1
, B
0
, . . . , B
M1
witness pnor(b, B) 2, in particu-
lar, b
1
> b
2
and pnor(b
1
, B
1
) 1. Since [F[ b
2
we conclude that there is B B
1
such that ( F)( B). Then B contradicts the choice of F
c(B)
.
Now, for the inductive step, assume our statement holds for k. Let pnor(b, B) k+2
and c : B 0, . . . , b 1. Let (b
i
, B
i
) : i < M witness pnor(b, B) (k +1) +1,
so M > b
k+2
and pnor(b
i
, B
i
) k + 1 and b
i
b. For each i < M apply the
inductive hypothesis to choose
i
< b such that pnor(b
i
, B
i
c
1
[
i
]) k. Choose

< b such that [i < M :

=
i
[ b
k+1
. Then (b
i
, B
i
c
1
[
i
]) :
i
=

witnesses that pnor(b, c

1
[

]) k + 1.
Now, by induction on n < we dene the following objects
()
n

H
(<n), r
H
(n), a(n), N
n
, g(n), H

(n), K

(n),

(n),
H
(= n).
We start with stipulating N
0
= 0,
H
(< 0) = 1.
Assume we have dened objects listed in ()
k
for k < n, and that we also have
dened integers N
n
,
H
(<n). We set
(i) g(n) = 2
Nn
+
H
(<n), r
H
(n) =
1
(n+1)
2

H
(<n)
and a(n) = 2
1/r
H
(n)
.
Choose (b

, B

) H
n
such that
(ii) b

> g(n), min(v(B)) > N

n
for all B B

and pnor(b

, B

) > a(n)
n+972
(possible by 2.2(3)). Set
(iii) N
n+1
= max
_
v(B) : B B

_
+ 1.
We let H

(n) be the set of all basic nblocks B such that v(B) [N

n
, N
n+1
), and
K

(n) consist of all triples c = (k

c
, b
c
, B
c
) such that
(b
c
, B
c
) H
n
, B
c
H

(n), b
c
> g(n), and k
c
, k
c
< pnor(b
c
, B
c
) 1.
For c K

(n) we set
(iv) nor(c) = log
a(n)
_
pnor(b
c
, B
c
) k
c
_
, val(c) = B
c
and

(c) = d K

(n) : k
c
k
d
, b
c
b
d
, B
d
B
c
.
Finally, we put
H
(=n) = [H

(n)[
n
and
H
(<n+1) =
H
(<n)
H
(=n). This
completes our inductive denition.
9
7
2


r
e
v
i
s
i
o
n
:
2
0
1
0
-
0
7
-
3
0







m
o
d
i
f
i
e
d
:
2
0
1
0
-
0
7
-
3
0


MONOTONE HULLS FOR N M 7
Proposition 2.3. (K

) is a creating pair for H

such that, for each n < ,

K

(n) is (n, r
H
(n))decisive, r
H
(n)halving and (g(n), r
H
(n))big.
Proof. To verify halving, for each c K

(n) with nor(c) > 1 set

half(c) = (k
c
+
1
2
(pnor(b
c
, B
c
) k
c
), b
c
, B
c
).
Note that nor(c) > 1 implies pnor(b
c
, B
c
) k
c
> 2 and hence
k
c
+
1
2
(pnor(b
c
, B
c
) k
c
) < pnor(b
c
, B
c
) 1.
Therefore, half(c)

(c) and nor(half(c)) nor(c) r

H
(n). Now suppose d

(half(c)), so k
c
+
1
2
(pnor(b
c
, B
c
) k
c
) k
d
, b
c
b
d
and B
d
B
c
. Also,
k
d
pnor(b
d
, B
d
) 1, so pnor(b
d
, B
d
) k
c
+
1
2
(pnor(b
c
, B
c
) k
c
) + 1. Consider
d

= (k
c
, b
d
, B
d
). Plainly d

(c), val(d

) val(d) and
nor(d

) log
a(n)
_

1
2
(pnor(b
c
, B
c
) k
c
) + 1
_
log
a(n)
_
1
2
(pnor(b
c
, B
c
) k
c
)
_
= nor(c) r
H
(n).
It follows from 2.2(4) that
() if c K

(n), nor(c) > r

H
(n), then c is (b
c
, r
H
(n))big.
Hence K

(n) is (g(n), r
H
(n))big (remember the denition of K

(n)).
Now suppose c K

(n), nor(c) > 1. Then pnor(b

c
, B
c
) k
c
> 2, so by the
denition of pnor (see 2.1(3)) we may nd b
c
b
0
< b
1
< . . . < b
M1
and disjoint
B
0
, . . . , B
M1
B
c
such that pnor(b
i
, B
i
) pnor(b
c
, B
c
)1 and (b
i
)
2
2
|Bi|
n
< b
i+1
.
Set
d

= (k
c
, b
0
, B
0
), d
+
= (k
c
, b
1
, B
1
), and K = [B
0
[.
Plainly, d

, d
+
(c), minnor(d

), nor(d
+
) nor(c) r
H
(n) and [val(d

)[ =
K. Also d
+
is hereditarily (2
K
n
, r
H
(n))big (remember b
1
> 2
K
n
, use ()). Thus
d

, d
+
witness that c is (K, n, r
H
(n))decisive.
Denition 2.4. (1) For a cardinal we consider the forcing notion P

(K

)
determined by the creating pair (K

) as in 1.3(2). Let < . A

P

(K

)name

is dened by

(K

=
_
_
p(, n) : dom(p) & n < trunklg(p, ) & p G

(K

)
_
.
(2) For

n<
H

(n) we set F() =

_


2 : (

n < )( (n)
_
.
Plainly, for each < ,
P

(K

n<
H

(n). Also, for

n<
H

(n),
the set F() is a meager and null
0
2
subset of

2.
Theorem 2.5. Assume CH. Let be an uncountable cardinal, =
0
.
(1) Forcing with P

(K

) preserves cardinalities and conalities and

(K

)
2
0
= .
(2) If < and

is a P
\{}
(K

)name for a member of

2, then

(K

F(

) .
(3) Consequently,
P

(K

)
non(^) = non(/) = .
Proof. (1) It follows from 2.3+1.4(2)+1.7.
(2) The proof is parallel to that of [6, Lemma 9.1]. Assume p P

(K

).
Remembering 1.4(1) we may use 1.7(3) to nd a condition q p such that
9
7
2


r
e
v
i
s
i
o
n
:
2
0
1
0
-
0
7
-
3
0







m
o
d
i
f
i
e
d
:
2
0
1
0
-
0
7
-
3
0


8 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
()
1
the condition q( ) <ndecides the value of

N
n
(for each n), and
()
2
trunklg(q, ) > 972 for all dom(q) and nor(q(, m)) > 972 whenever
supp(q, m), and
()
3
dom(q) and if supp(q, m) ,= , then [supp(q, m)[ > 972.
Thus, for each n, we have a mapping E
n
: val

(q( ), <n)
Nn
2 such that
_
q( )
_
t
P
\{}
(K

N
n
= E
n
(t) .
We will further strengthen q to a condition q

such that dom(q

) = dom(q) and
()
goal
for all n trunklg(q

, ) and t val

_
q

( ), <(n + 1)
_
we have
_
B q

(, n)
__
E
n+1
(t) B
_
.
Then clearly we will have q

(K

F(

) and the proof of 2.5(2) will

follow by the standard density argument.
To construct the condition q

we set dom(q

) = dom(q), trunklg(q

, ) =
trunklg(q, ), and we dene q

(, m) by induction on m so that:
q

(, m) = q(, m) whenever / supp(q, m) or / supp(q, m), and

q

(, m)

(q(, m)), nor(q

(, m)) nor(q(, m)) 2 for supp(q, m).

These demands guarantee that q

is a condition in P

(K

) stronger than q.
Fix an n trunklg(q, ). Put A = supp(q, n) and note that that A, A has
at least 972 elements (remember ()
3
), and [A[ < n (by 1.3(2)()).
Set c
0

= q(, n) for A.
We choose inductively an enumeration
0
, . . . ,
|A|1
of A and creatures c

k
(for k) and d

k
from

(c
0

k
). So assume that for some 0 we already
have dened a list
k
: k < of distinct elements of A and creatures c

for
A
0
, . . . ,
1
. Each c

is (K

, n, r
H
(n))decisive for some K

. Put
K

= min(K

: A
0
, . . . ,
1
), and choose

such that K

= K

.
Let d

(c

) be such that [val(d

)[ K

and nor(d

) nor(c

) r
H
(n).
For A
0
, . . . ,

, let c
+1

(c

) be hereditarily (2
(K

)
n
, r
H
(n))big and
such that nor(c
+1

) nor(c

) r
H
(n). Iterate this procedure [A[ 1 times. At
the end, there remains one that has not been listed as an

, so we set
|A|1
=
and d

|A|1
= c
|A|1

.
Since c
+1

+1
is hereditarily (2
(K

)
n
, r
H
(n))big, we see that 2
(K

)
n
< K
+1
. Let
m be such that =
m
, and put
K = K
m
, S =

: < m, L =

: > m.
It is possible that (at most) one of the sets S, L is empty. By our choices,
()
4
(a) d

(q(, n)), nor(d

) nor(q(, n)) (n1) r

H
(n) > 900, and
(b) if S ,= then d

is (2
(Km1)
n
, r
H
(n))big and hence in particular
(K
m1
)
n2
< K; if S = then K = K
0
,
(c)

S
[val(d

)[ (K
m1
)
n2
< K and [val(d

)[ K,
(d)
H
(<n) < K
0
K (remember that K(n) is (g(n), r
H
(n))big and
g(n) >
H
(<n)),
(e) if L, then d

is (2
K
n
, r
H
(n))big.
Let Z = t val

(q( ), <(n+1)) : t(, n) val(d

) for A and
for s

L
val(d

) let Z
s
= t Z : t(, n) = s() for L. Next, for t Z put
(
t
= B B
d

: E
n+1
(t) B.
9
7
2


r
e
v
i
s
i
o
n
:
2
0
1
0
-
0
7
-
3
0







m
o
d
i
f
i
e
d
:
2
0
1
0
-
0
7
-
3
0


MONOTONE HULLS FOR N M 9
If S = , then in what follows ignore

S
val(d

) and set K
m1
= 1. Assume L is
non-empty (otherwise move to ()
6
). For each s

L
val(d

) consider a function
c(s) : val

(q( ), <n)

S
val(d

) T(val(d

))
such that c(s)(t
0
, t
1
) = (
t0

t1

s
, where t
0

t
1

s Z
s
is obtained by natural con-
catenation. This determines a coloring c on

L
val(d

) with the range of size at

most
_
2
K
_

H
(<n)(Km1)
n2

_
2
K
_
KK
= 2
K
3
< 2
K
n
.
Since K

(n) is (n, r
H
(n))decisive, and each d

is hereditarily (2
K
n
, r
H
(n))big
(for L), nor(d

) > 900 and [L[ n 2, therefore we may use Lemma 1.2 to

nd q

(, n)

(d

) for L such that

()
5
(a) nor(q

(, n)) nor(d

) r
H
(n) n nor(q(, n)) 2, and
(b) c

L
val(q

(, n)) is constant.
If L = then the procedure described above is not needed. In any case, letting
X = val

(q( ), <n)

S
val(d

),
we have a mapping d : X T(val(d

)) and q

(, n) for L such that

()
6
if t Z and t(, n) val(q

(, n)) for L, then (

t
= d(t
0
, t
1
), where
t
0
= t
_
(dom(q) ) n
_
val

(q( ), <n) and t

1
= t(S n)

S
val(d

).
For each (t
0
, t
1
) X x one t = t[t
0
, t
1
] Z such that t(, n) val(q

(, n)) for
L, t
0
= t
_
(dom(q) ) n
_
and t
1
= t(S n). Now, for B val(d

)
we (try to) x (t
B
0
, t
B
1
) X such that B (
t[t
B
0
,t
B
1
]
, if possible. Consider a coloring
e : val(d

)
Nn+1
2 dened by
e(B) =
_
E
n+1
(t[t
B
0
, t
B
1
]) if (t
B
0
, t
B
1
) X is dened,
otherwise.
Since [X[
H
(<n) (K
m1
)
n2
max(K
m1
)
n1
,
H
(<n), we know that
the range of the coloring e has at most max(K
m1
)
n1
,
H
(<n) + 1 members.
Thus d

is ([rng(e)[, r
H
(n))big and we may choose q

(, n)

(d

) such that
nor(q

(, n)) nor(d

) r
H
(n) nor(q(, n)) 2 > 900 and eval(q

(, n)) is
constant. If the constant value were
Nn+1
2, then we would have B for all
B val(q

(, n)), contradicting nor(q

(, n)) > 1. Therefore,

()
7
(t
B
0
, t
B
1
) is dened for no B val(q

(, n)) and hence

val(q

(, n))
_
(
t[t0,t1]
: (t
0
, t
1
) X = .
For S we set q

(, n) = d

. Now note that

()
8
if t Z is such that t(, n) q

(, n) for S L and B val(q

(, n)),
then E
n+1
(t) B.
Why? Assume towards contradiction that E
n+1
(t) B, i.e., B (
t
. Represent
t as t = t
0

t
1

s where (t
0
, t
1
) X. Then (
t
= (
t[t0,t1]
(by ()
6
) and therefore
B (
t[t0,t1]
, contradicting ()
7
.
9
7
2


r
e
v
i
s
i
o
n
:
2
0
1
0
-
0
7
-
3
0







m
o
d
i
f
i
e
d
:
2
0
1
0
-
0
7
-
3
0


10 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
This completes the denition of q

. It follows from ()
8
(for n trunklg(q

, ))
that ()
goal
is satised.
(3) Follows from (2) and the fact that F() ^ / for

n<
H(m).
Corollary 2.6. It is consistent that
non(^) = non(/) = non(^ /) =
2
= 2
0
and do =
1
.
Proof. Start with a model of CH and force with P
2
(K

). It follows from 2.5

and 1.8 that the resulting model is as required.
In models for the statement in Corollary 2.6 necessarily cov(^) = cov(/) =
1
.
However, it is not clear if we could not get a parallel result for d
B
and cov.
Problem 2.7. Is it consistent that
cov(^) = cov(/) =
2
= 2
0
and d
B
=
1
?
In particular, is it consistent that do > d
B
?
Directly from 2.6 we also obtain
Corollary 2.8. It is consistent that non(^/) =
2
and there is no increasing
chain of Borel subset of

2 of length
2
.
3. Monotone hulls
The interest in Corollary 2.8 came from the questions concerning Borel hulls.
Denition 3.1. Let Borel(

2) be the family of all Borel subsets of

2, 1 be a
ideal on

2 with Borel basis and o
I
be the algebra of subsets of

2 generated
by Borel(

2) 1. Let T o
I
. A monotone Borel hull on T with respect to 1 is a
mapping : T Borel(

2) such that
A (A) and (A) A 1 for all A T, and
if A
1
A
2
are from T, then (A
1
) (A
2
).
If the range of consists of sets of some Borel class /, then we say that is a
monotone / hull operation.
As discussed in Balcerzak and Filipczak [1, Question 24], 2.8 implies the following.
Corollary 3.2. It is consistent that
there are no monotone Borel hulls on / with respect to /, and
there are no monotone Borel hulls on ^ with respect to ^, and
there are no monotone Borel hulls on / ^ with respect to / ^.
The non-existence of monotone Borel hulls on 1 implies non-existence of such
hulls on o
I
. While some partial results were presented in [5] and [1], not much is
known about the converse implication.
Problem 3.3 (Cf. Balcerzak and Filipczak [1, Question 26]). Let 1 /, ^. Is
it consistent that there exists a monotone Borel hull on 1 (with respect to 1) but
there is no such hull on o
I
? In particular, is it consistent that add(1) = cof(1)
but there is no monotone Borel hull on o
I
?
9
7
2


r
e
v
i
s
i
o
n
:
2
0
1
0
-
0
7
-
3
0







m
o
d
i
f
i
e
d
:
2
0
1
0
-
0
7
-
3
0


MONOTONE HULLS FOR N M 11
It was noted in [1, Proposition 7] (see also Elekes and M athe [5, Theorem 2.4])
that add(1) = cof(1) implies that there exists a monotone Borel hull on 1 (with
respect to 1). It appears that was the only situation in which the positive result of
this kind was known. Using FS iteration with partial memory we will show in this
section that, consistently, we may have add(1) < cof(1) (for 1 ^, /) and yet
there are monotone hulls for 1.
Denition 3.4. Let 1 be an ideal of subsets of

2.
(1) We say that a family B Borel(

2) 1 is an mhgbase for 1 if
3
(a) B is a basis for 1, i.e., (A 1)(B B)(A B), and
(b) if B
i
: i <
1
) is a sequence of elements of B, then for some i < j <
1
we have B
i
B
j
.
(2) Let

be limit ordinals. An

base for 1 is a sequence B

,
:
<

& <

) of Borel sets from 1 such that it forms a basis for 1 (i.e.,

(a) above holds) and
(c) for each
0
,
1
<

,
0
,
1
<

we have
B
0,0
B
1,1

0

1
&
0

1
.
Proposition 3.5. Assume that B
,
: <

& <

) is an

base for
1. Then:
(i) B
,
,= B

,
whenever (, ) ,= (

), ,

<

, ,

<

.
(ii) B
,
: <

& <

is an mhgbase for 1.
(iii) add(1) = mincf(

), cf(

) and cof(1) = maxcf(

), cf(

).
Proof. Straightforward.
Proposition 3.6. Suppose that an ideal 1 on

2 has an mhgbase B Borel(

2)
1. Then there exists a monotone hull operation : 1 Borel(

2) 1 on 1. If,
additionally, B
0

, <
1
, then can be taken to have values in
0

.
Proof. For a set A 1 let o
A
be the family of all sequences

B = B
i
: i < ) B
satisfying
()
1
(i < )(A B
i
) and
()
2
(i < j < )(B
i
B
j
).
Note that for each

B o
A
we have g(

B) <
1
(by 3.4(1)(b) and ()
2
). Clearly,
every increasing chain of elements of o
A
has a upper bound in o
A
, so we
may choose

B
A
= B
A
i
: i <
A
) o
A
which has no proper extension in o
A
. Put
(A) =

i<A
B
A
i
. Plainly, A (A) 1 and (A) is a Borel set, and if B
0

then also (A)

0

.
Claim 3.6.1. (A) =

B B : A B
Proof of the Claim. By ()
1
we see that (A)

B B : A B. To show
the converse inclusion suppose B B, A B. By the choice of

B
A
we know that

B
A

B) / o
A
and hence B
A
i
B for some i <
A
. Consequently (A) B.
It follows from the above claim that A
1
A
2
1 implies (A
1
) (A
2
).
3
mhg stands for monotone hull generating
9
7
2


r
e
v
i
s
i
o
n
:
2
0
1
0
-
0
7
-
3
0







m
o
d
i
f
i
e
d
:
2
0
1
0
-
0
7
-
3
0


12 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
Theorem 3.7. Let , be cardinals of uncountable conality, . There is a
ccc forcing notion Q
,
of size
0
such that

Q
, the meager ideal / has a basis consisting of
0
2
sets, and
the null ideal ^ has a basis consisting of
0
2
sets .
Proof. The forcing notion Q
,
will be obtained by means of nite support iteration
of ccc forcing notions. The iterands will be products of the Amoeba for Category B
and Amoeba for Measure A but considered over partial sub-universes only. Thus it
is yet another application of FS iterations with partial memories used in Shelah
[10, 11, 12], Mildenberger and Shelah [8] and Shelah and Thomas [13]. We will use
the notation and some basic facts stated in the third section of the latter paper.
Let us recall the forcings A and B used as iterands.
A condition in A is a tree T
>
2 such that ([T]) >
1
2
and ([t][T]) > 0
for all t T. The order
A
of A is the reverse inclusion.
A condition in B is a pair (n, T) such that n , T
>
2 is a tree with
no maximal nodes and [T] is a nowhere dense subset of

2. The order
B
of B is given by:
(n, T)
B
(n

, T

) if and only if n n

, T T

and T
n
2 = T

n
2.
Both A and B are (nice denitions of) ccc forcing notions, B is centered and if
V

are universes of set theory then A

V

is still ccc in V

. We will use the

following immediate properties of these forcing notions.
()
1
If G A is generic over V, F =

[T] : T G, then F is a closed subset

of

2, (F) =
1
2
and F is disjoint from every Borel null set coded in V.
Hence the set F

= x

2 : (y F)(

n)(x(n) ,= y(n)) is a null

0
2
set and it includes all Borel null sets coded in V.
Let F

A
, F

A
be Anames for the sets F, F

, respectively.
()
2
If G B is generic over V, F =

[T] : (n)((n, T) G), then F is

a closed nowhere dense subset of

2. Letting F

= x

2 : (y
F)(

n)(x(n) = y(n)) we get a meager

0
2
set including all Borel meager
sets coded in V.
Let F

B
, F

B
be Bnames for the sets F, F

, respectively.
()
a
3
If T A, t T, then there is T

A
T such that T

A
[t] F

A
,= .
()
b
3
If T A, n , then there is N > n such that for each
[n,N)
2 there is
T

A
T with T

A
(y F

A
)(y[n, N) ,= ).
()
a
4
If (n, T) B, t T, m
1
> m
0
n and
[m0,m1)
2, then there are
(n

, T

)
B
(n, T) and s T

such that t s and s[m

0
, m
1
) = (and
(n

, T

)
B
s F

B
).
()
b
4
If (n, T) B, m
0
< , then there are m
1
> m
0
and
[m0,m1)
2 and
(n

, T

)
B
(n, T) such that (n

, T

)
B
(y F

B
)(y[m
0
, m
1
) ,= ).
Fix an ordinal and a bijection :
onto
such that

0

1
< &
0

1
< (
0
,
1
) (
1
,
1
).
For i = (
1
,
1
) let a
i
= (
0
,
0
) :
0

1
&
0

1
i. We say that a set
b is closed if a
i
b for all i b. It follows from our choice of that for each
i < we have
()
5
a
i
i and the sets a
i
, i, a
i
i are closed.
9
7
2


r
e
v
i
s
i
o
n
:
2
0
1
0
-
0
7
-
3
0







m
o
d
i
f
i
e
d
:
2
0
1
0
-
0
7
-
3
0


MONOTONE HULLS FOR N M 13
Now, by induction we dene P
i
, Q

i
, F

0
i
, F

1
i
, F

A
i
, F

B
i
: i < ) and P

b
for closed b
simultaneously proving the correctness of the denition and the desired properties
listed below.
4
()
6
P
j
, Q

i
: j , i < ) is a nite support iteration of ccc forcing notions.
()
7
P

b
=
_
p P

: supp(p) b & p(i) is a P

ai
name (for a member of Q

i
) for
each i supp(p)
_
.
()
8
P

b
is a complete suborder of P

, P

ai{i}
is isomorphic with the composition
P

ai
Q

i
.
()
9
Q

i
is a P

ai
name for the product
5
A B.
()
10
F

0
i
, F

1
i
, F

A
i
, F

B
i
are P

ai{i}
names for the sets F

A
, F

B
, F

A
, F

B
added by the
forcings at the last coordinate of P

ai{i}
P

ai
(A B).
()
11
(a) P

i
is a dense subset of P
i
(for i ).
(b) If q P

, then qb P

b
.
(c) If p, q P

, p q and i supp(q) then pa

i

P

a
i
qa
i
and qa
i

P

a
i
p(i) q(i).
(d) If q P

, p P

b
and p q, then p
P

b
qb.
(e) If q P

b
, p P

, pb
P

b
q and r is dened by
r() =
_
q() if b,
p() otherwise
for <
then r P

and r q, r p.
Also,
()
12
if

is a canonical
6
P

name for a member of

2, then

is a P

ai
name for
some i < .
[Why? Note that if (
n
,
n
) , n < , then there is (

) such
that
n

,
n

for all n < .]

The main technical point of our argument is given in the following observation.
()
13
Suppose i, j < , i / a
j
, j / a
i
, i ,= j, 0, 1. Assume that p P

,

n
2, n < and p
P

[]F

i
,= . Then there are
[n,N)
2, n < N <
and q
P

p such that
q
P

] F

i
,= and
_
y F

j
__
y[n, N) ,=
_
.
[Why? Let us provide detailed arguments for = 0. By ()
b
3
+ ()
9
+ ()
11
we
may nd N > n and a condition p

0
P

aj
such that p

0
pa
j
and
p

0

P

a
j
for each
[n,N)
2 there is p
j

Q

j
p(j) such that
p
j

Q

j
(y F

A
)(y[n, N) ,= ) .
Let p
0
P

be such that p
0
() = p

0
() for a
j
and p
0
() = p() otherwise (see
()
11
(e); so p
0
is a common extension of p

0
and p). Note that p
0
(j) = p(j). Use
()
a
3
to choose
[n,N)
2 and a condition p

1
P

ai{i}
such that p

1
p
0
(a
i
i)
and p

1

P

a
i
{i}
[

] F

0
i
,= . Let p
1
P

be such that p
1
() = p

1
() if
4
See [13, 3.13.7] for the order in which these should be shown.
5
Since B
V
P

a
i
is centered we know that the product is ccc.
6
i.e., determined in a standard way by a sequence of maximal antichains
9
7
2


r
e
v
i
s
i
o
n
:
2
0
1
0
-
0
7
-
3
0







m
o
d
i
f
i
e
d
:
2
0
1
0
-
0
7
-
3
0


14 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
a
i
i and p
1
() = p
0
() otherwise. Then p
1
is stronger than both p

1
and p
0
,
and p
1
(j) = p
0
(j) = p(j). Hence
p
1
a
j

P

a
j
there is p
j

Q

j
p
1
(j) such that p
j

Q

j
(y F

A
)(y[n, N) ,= ) .
Let q(j) be a P

aj
name for a p
j
as above and let q() = p
1
() for ,= j. Clearly
q P

and q(a
j
j)
P

a
j
{j}
(y F

0
j
)(y[n, N) ,= ), and (as q(a
i
i) =
p
1
(a
i
i) = p

1
) q(a
i
i)
P

a
i
{i}
[

] F

0
i
,= . Using ()
8
+()
10
+()
11
we get that the condition q is as required. If = 1 then the arguments are similar,
but instead of ()
a
3
, ()
b
3
we use ()
a
4
, ()
b
4
.]
For < , < let B

A
,
= F

A
(,)
and B

B
,
= F

B
(,)
. Immediately from
()
12
+ ()
1
+ ()
2
we conclude that
()
14

P

A
,
: < & < is a basis for ^ and
B

B
,
: < & < is a basis for /
and
()
15
if
0

1
< ,
0

1
< , (
0
,
0
) ,= (
1
,
1
), then

A
0,0
B

A
1,1
& B

B
0,0
B

B
1,1
.
Also
()
16
if
0
,
1
< ,
0
,
1
< and
_

0

1
&
0

1
_
then

A
0,0
B

A
1,1
& B

B
0,0
B

B
1,1
.
[Why? If
1

0
and
1

0
, then ()
15
applies, so we may assume additionally

1

0
&
1

0
_
. Then our assumptions on
0
,
1
,
0
,
1
mean that, letting
j = (
0
,
0
) and i = (
1
,
1
), we have i / a
j
, j / a
i
, i ,= j. So using ()
13
for
= 0 we easily build a P

name

for a member of

2 such that

[F

0
i
]

2 F

A
i
=

2 B

A
1,1
&

A
j
= B

A
0,0
.
Similarly, using ()
13
for = 1 and interchanging the role of i and j we may
construct a P

name

such that
P

B
0,0
B

B
1,1
. ]
Finally we note that P

has a dense subset of size

0
, so we may choose it as
our desired forcing Q
,
.
Corollary 3.8. It is consistent that
add(^) = add(/) < cof(^) = cof(/) = 2

, and
there is a monotone
0
3
hull operation on / with respect to /, and
there is a monotone
0
2
hull operation on ^ with respect to ^, and
there is a monotone
0
3
hull operation on / ^ with respect to / ^.
Proof. Start with a universe satisfying CH and use the forcing given by Theorem
3.7 for =
1
and =
2
. Propositions 3.6 and 3.5 imply that the resulting model
is as required.
Remark 3.9. In Theorem 3.7 we obtained a universe of set theory in which both
^ and / have bases that are (with respect to the inclusion) order isomorphic to
. We may consider any partial order (S, ) such that
(a) [S[ = and (S, ) is well founded, and
(b) every countable subset of S has a common upper bound.
9
7
2


r
e
v
i
s
i
o
n
:
2
0
1
0
-
0
7
-
3
0







m
o
d
i
f
i
e
d
:
2
0
1
0
-
0
7
-
3
0


MONOTONE HULLS FOR N M 15
Then by a very similar construction we get a forcing extension in which both ^
and / have bases order isomorphic to (S, ). If additionally
(c) for every sequence s
i
: i <
1
) S there are i < j <
1
such that s
i
s
j
,
then those bases will be mhg. (Note that forcings with the Knaster property pre-
serve the demand described in (c).)
References
[1] Marek Balcerzak and Tomasz Filipczak. On monotone hull operations. Mathematical Logic
Quarterly, accepted (2010).
[2] Tomek Bartoszy nski and Haim Judah. Set Theory: On the Structure of the Real Line. A K
Peters, Wellesley, Massachusetts, 1995.
[3] J org Brendle and Sakae Fuchino. Coloring ordinals by reals. Fundamenta Mathematicae,
196:151195, 2007.
[4] M arton Elekes and Kenneth Kunen. Transnite sequences of continuous and Baire class 1
functions. Proc. Amer. Math. Soc., 131:24532457, 2003.
[5] M arton Elekes and Andr as M athe. Can we assign the Borel hulls in a monotone way? Fun-
damenta Mathematicae, 205:105115, 2009.
[6] Jakob Kellner and Saharon Shelah. Decisive creatures and large continuum. Journal of Sym-
bolic Logic, 74:73104, 2009.
[7] Kenneth Kunen. Inaccessibility properties of cardinals. PhD thesis, Stanford University, 1968.
[8] Heike Mildenberger and Saharon Shelah. Changing cardinal characteristics without changing
sequences or conalities. Annals of Pure and Applied Logic, 106:207261, 2000.
[9] 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.
[10] Saharon Shelah. Covering of the null ideal may have countable conality. Fundamenta Math-
ematicae, 166:109136, 2000.
[11] Saharon Shelah. Was Sierpi nski right? IV. Journal of Symbolic Logic, 65:10311054, 2000.
[12] Saharon Shelah. The null ideal restricted to some non-null set may be
1
-saturated. Funda-
menta Mathematicae, 179:97129, 2003.
[13] Saharon Shelah and Simon Thomas. The Conality Spectrum of The Innite Symmetric
Group. Journal of Symbolic Logic, 62:902916, 1997.
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
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