6

5
1


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







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


FORCING FOR hL AND hd
ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
Abstract. The present paper addresses the problem of attainment of the
supremums in various equivalent denitions of hereditary density hd and hered-
itary Lindelof degree hL of Boolean algebras. We partially answer two prob-
lems of J. Donald Monk, [13, Problems 50, 54], showing consistency of dierent
attainment behaviour and proving that (for the considered variants) this is the
best result we can expect.
0. Introduction
We deal with the attainment problem in various denitions of two cardinal func-
tions on Boolean algebras: the hereditary density hd and the hereditary Lindel of
degree hL. These two cardinal functions are closely related, as it is transparent
when we pick the right variants of (equivalent) denitions. Also they both are
somewhat related to the spread s of Boolean algebras. So, for a Boolean algebra
B, we dene
s(B) = sup : there is an idealindependent sequence of length ,
hd(B) = sup : there is a leftseparated sequence of length ,
hL(B) = sup : there is a rightseparated sequence of length .
Let us recall that a sequence a

: < ) of elements of a Boolean algebra is

idealindependent if a

_
w
a

for each < and a nite set w ,

leftseparated if a

_
w
a

for each < and a nite set w ( +1),

rightseparated if a

_
w
a

for each < and a nite set w .

The above denitions of the three cardinal functions are of special use, see e.g.
[15, 1]. However, neither these denitions explain the names of the functions, nor
they are good enough justications for the interest in them. But all three functions
originate in the cardinal functions of the topological space Ult(B) (of ultralters on
B). And thus, for a Boolean algebra B, we may dene (or prove that the following
equalities hold true):
s(B) = sup[X[ : X Ult(B) is discrete in the relative topology ,
1991 Mathematics Subject Classication. Primary 03E35, 03G05, 54A25; Secondary 03E05,
06Exx.
Key words and phrases. Boolean algebras, spread, hereditary density, hereditary Lindelof de-
gree, attainment.
The rst author thanks the KBN (Polish Committee of Scientic Research) for partial support
through grant 2 P03 A 01109.
The research of the second author was partially supported by the Israel Science Foundation.
Publication 651.
1
6
5
1


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







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


2 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
hd(B) = supd(X) : X Ult(B), where
d(X) = min[Y [ : Y X is dense in X ,
hd(B) = supL(X) : X Ult(B), where
L(X) = min : every open cover of X has a subcover of size .
The respective pairs of cardinal numbers are dened using sup, so even if we know
that they are equal we still may expect dierent attainment properties: one of the
families of cardinals may have the largest member while the other not. Also we
may ask if the sup has to be attained. Situation may seem even more complicated if
one notices that there are more than just two equivalent denitions of the cardinal
functions s, hd, hL: Monk [13] lists six equivalent denitions for spread (see [13,
Theorem 13.1]), nine denitions for hd, and nine for hL (see [13, Theorems 16.1,
15.1]). Fortunately, there is a number of dependencies here.
First, all of the equivalents of spread have the same attainment properties. More-
over, spread is always attained for singular strong limit cardinals and for singular
cardinals of countable conality (for these and related results see Hajnal and Juh asz
[3], [4], [5], Juh asz [8], [9], Roitman [14], Kunen and Roitman [11], Juh asz and She-
lah [10]). Then Shelah [20] proved that 2
cf(s(B))
< s(B) implies that the spread is
attained (see 1.3 here). Finally, it is shown in Shelah [18, 4] that, e.g., if is a
singular strong limit cardinal such that < cf() < 2

, then there is a Boolean

algebra B such that [B[ = s(B) = and the spread is not obtained. Thus, to some
extend, the problem of attainment for spread is settled.
Many of the results mentioned above can be carried out for (some) variants of
hd and hL. However, the dierence between these two cases and the case of the
spread is that the various equivalent denitions of the respective cardinal function
might have dierent attainment properties.
Let us introduce some of the equivalents of hL, hd. They will be called hL
(n)
,
hd
(n)
, with the integer n referring to the respective cardinal
n
as used in the
proofs of [13, 15.1 and 16.1], respectively. Also, we will have hd
+
(n)
and hL
+
(n)
to
have proper language to deal with the attainment questions. Let us start with
the hereditary Lindel of degree hL. First, for a topological space X we dene the
Lindel of degree L(X) of the space X as
L(X) = min : every open cover of X has a subcover of size .
Denition 0.1. Let B be an innite Boolean algebra. For an ideal I in a Boolean
algebra B we let
cof(I) = min[A[ : A I and (b I)(a A)(b a).
Now we dene
hL
(+)
(0)
(B) = supL(X)
(+)
: X is a subspace of Ult(B) ,
hL
(+)
(1)
(B) = supcof(I)
(+)
: I is an ideal of B ,
hL
(+)
(7)
(B) = sup
(+)
: there is a rightseparated sequence a

: < ) in B .
The superscript (+) in the above denitions means that each of the formulas has
two versions: one with + and one without it.
The cardinals mentioned in 0.1 are among those listed in [13, Theorem 15.1], and
so hL
(0)
(B) = hL
(1)
(B) = hL
(8)
(B). The attainment properties can be described
using the versions with +: hL
+
(i)
(B) = hL
(i)
(B) means that the supremum is not
6
5
1


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







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


FORCING FOR hL AND hd 3
obtained; hL
+
(i)
(B) = hL
+
(j)
(B) means that the respective two denitions of hL have
the same attainment for B. It is not dicult to note that
hL
+
(7)
(B) = hL
(7)
(B) hL
+
(1)
(B) = hL
(1)
(B)
and
hL
(0)
(B) = hL
+
(0)
(B) is a regular cardinal hL
+
(7)
(B) = hL
(7)
(B)
(and the attainment of hL in senses not listed in 0.1 can be reduced to those three;
see [13, p. 190, 191] for details). Also, if hL(B) is a strong limit cardinal or if it
has countable conality, then hL
(7)
(B) < hL
+
(7)
(B) (see Juh asz [9, 4.2, 4.3]).
In 1.4 we will show that if hL(B) is a singular cardinal such that 2
cf(hL(B))
<
hL(B), then hL
+
(0)
(B) = hL
+
(1)
(B) = hL
+
(7)
(B) = (hL(B))
+
. Thus, e.g., under GCH,
the sups in all equivalent denitions of hL are attained at singular cardinals. Next,
in section 3, we use forcing to show that, consistently, there is a Boolean algebra B
such that
hL
(7)
(B) < hL
+
(7)
(B) and hL
+
(1)
(B) = hL
(1)
(B)
(see 3.7). This still leaves some aspects of [13, Problem 50] open: are there any
implications between attainment in hL
(0)
and hL
(1)
sense? Between hL
(0)
and hL
(7)
sense?
We also carry out the parallel work for the hereditary density. Let us introduce
the respective denitions. The density d(X) of a topological space X is dened as
the minimal size of a dense subset of X. The topological density d(B) of a Boolean
algebra B is the density of the space Ult(B) of ultralters on B. The algebraic
density (sometimes also called the weight ) of a Boolean algebra B is
(B) = min[A[ : A B 0 and (b B 0)(a A)(a b) .
Denition 0.2. For an innite Boolean algebra B we let:
hd
(+)
(0)
(B) = supd(X)
(+)
: X is a subspace of Ult(B) ,
hd
(+)
(5)
(B) = sup
(+)
: there is a leftseparated sequence of length ,
hd
(+)
(7)
(B) = sup(B

)
(+)
: B

is a homomorphic image of B ,
hd
(+)
(8)
(B) = supd(B

)
(+)
: B

is a homomorphic image of B .
(Again, the superscripts (+) mean that we have two variants for each cardinal:
with and without +.)
Like before, the cardinals mentioned in 0.2 correspond to those listed in [13,
Theorem 16.1], and the variants with + reect the attainment properties. The
known dependencies here are
hd
+
(5)
(B) = hd
(5)
(B) hd
+
(7)
(B) = hd
(7)
(B)
hd
+
(0)
(B) = hd
(0)
(B) hd
+
(8)
(B) = hd
(8)
(B)
and
hd
(0)
(B) = hd
+
(0)
(B) is a regular cardinal hd
(5)
= hd
+
(5)
(B)
(and Monk [13, Problem 54] asked for a complete description of dependencies).
Like for hL, if hd(B) is a strong limit cardinal or if it has countable conality, then
hd
(5)
(B) < hd
+
(5)
(B) (see Juh asz [9, 4.2, 4.3]).
6
5
1


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







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


4 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
In 1.5 we note that if hd(B) is a singular cardinal such that 2
cf(hd(B))
< hd(B),
then hd
+
(8)
(B) = hd
+
(7)
(B) = hd
+
(5)
(B) = hd
+
(0)
(B) = (hd(B))
+
. Consequently, GCH
implies that the sups in all equivalent denitions of hd are attained at singular
cardinals. Then, in section 4, we show that, consistently, there is a Boolean algebra
B such that
hd
(5)
(B) < hd
+
(5)
(B) and hd
+
(7)
(B) = hd
(7)
(B)
(see 4.5). This still leaves several aspects of [13, Problem 54] open.
Finally, in the last section of the paper we show that (if we start with the right
cardinals , , cf() < ) adding a Cohen real produces a Boolean algebra B
such that hL
+
(7)
(B) = hd
+
(5)
(B) = s
+
(B) = (put 5.4, 5.6 together). This result is
of interest as it shows how easily we may have algebras in which the three cardinal
functions do not attain their supremums. (But of course there is the semi-ZFC
result of [18, Theorem 4.2].)
Notation: Our notation is standard and compatible with that of classical text-
books on set theory (like Jech [7]) and Boolean algebras (like Monk [12], [13]).
However in forcing considerations we keep the older tradition that
the stronger condition is the greater one.
Let us list some of our notation and conventions.
(1) A name for an object in a forcing extension is denoted with a dot above
(like

X) with one exception: the canonical name for a generic lter in a
forcing notion P will be called
P
. For a Pname

X and a Pgeneric lter
G over V, the interpretation of the name

X by G is denoted by

X
G
.
(2) i, j, , , , , . . . will denote ordinals and , , , will stand for (always
innite) cardinals.
(3) For a set X and a cardinal , [X]
<
stands for the family of all subsets of
X of size less than . If X is a set of ordinals then its order type is denoted
by otp(X).
(4) Sequences of ordinals will be typically called , , , ; the length of a se-
quence is lh(); means that the sequence in an initial segment of
. The set of all sequences of length with values in will be denoted by

. The lexicographic order on sequences of ordinals will be called <

lex
.
(5) In Boolean algebras we use (and
_
), (and
_
) and for the Boolean
operations. If B is a Boolean algebra, x B then x
0
= x, x
1
= x. The
Stone space of the algebra B (the space of ultralters) is called Ult(B).
When working in the Stone space, we identify the algebra B with the eld
of clopen subsets of Ult(B).
(6) For a subset Y of an algebra B, the subalgebra of B generated by Y is
denoted by Y )
B
and the ideal generated by Y is called id
B
(Y ).
Acknowledgements: We would like to thank the referee for valuable com-
ments and suggestions.
1. Golden Oldies: the use of [Sh:233]
In this section we recall how [20] applies to the attainment problems. The
proofs of 1.2 and 1.3 were presented in [20], but we recall them here, as we have an
impression that those beautiful results went somehow unnoticed. Also, as the results
6
5
1


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







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


FORCING FOR hL AND hd 5
of sections 3 and 4 complement the consequences of [20, Lemma 5.1] presented here,
it may be convenient for the reader to have all the proofs presented as well.
Hypothesis 1.1. Let , be cardinals, and =
i
: i < cf()) be an increasing
sequence of regular cardinals such that
cf() < =
_
2
cf()
_
+
< = sup
i<cf()

i
and <
0
.
Theorem 1.2 (See [20, Lemma 5.1]). Let X be a topological space with a basis B
consisting of clopen sets. Suppose that is a function assigning cardinal numbers
to subsets of X such that (X) and
(i) (A) (A B) (A) + (B) +
0
for A, B X,
(ii) for each closed set Y X such that (Y ) and for i < cf(), there are
u

: < ) B and y

: < ) Y such that

(a) y

Y ,
(b) (v B)(y

v (v Y )
i
),
(c) (g : 2
cf()
)(, < )(g() = g() & y

/ u

),
(iii) if A

: < ) is a sequence of subsets of X such that (A

)
i
(for
< ) then (

<
A

)
i
.
Then there is a sequence v
i
: i < cf()) B such that
(i < cf())((v
i

_
j=i
v
j
)
i
).
Proof. First, by induction on i < cf(), we choose families K
i
of clopen subsets
of X, and sets D
i
X such that [K
i
[ = [D
i
[ = . So suppose that K
j
, D
j
have
been dened for j < i. For each | [

j<i
K
j
]
<cf()
such that (X

|)
pick y
U

: < ) X | and u
U

: < ) B as guaranteed by (ii) (for i and

Y = X

|). Let D
i
consist of all y
U

(for | as above and < ); note that

[D
i
[ = . Let K
i
be a family of clopen sets such that [K
i
[ = and for each | as
above:
u
U

K
i
for all < ,
if y
U

u
U

u
U

, , < , then there is u K

i
B such that y
U

u u
U

u
U

,
if u K
i
then X u K
i
.
Let K =

i<cf()
K
i
(clearly [K[ = ) and let
Z
i
= x X : if v

: < cf() K and x

<cf()
v

then (

<cf()
v

)
i
.
Claim 1.2.1. If Y X is a closed set such that (Y )
i
, then Z
i
Y ,= .
Proof of the claim. Suppose that for each x Y we have a sequence v
x

: <
cf()) K such that x

<cf()
v
x

and (

<cf()
v
x

) <
i
. There are at most
possibilities for such sequences, so we get a set W [Y ]

such that
Y
_
xW

<cf()
v
x

.
6
5
1


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







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


6 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
Use the assumption (iii) to conclude that (

xW

<cf()
v
x

)
i
, and next use (i)
to get a contradiction with (Y ) .
For each i < cf() x z
i
Z
i
.
Now, by induction on i < cf(), choose v
i
K
i
and x
i
Z
i
such that
() x
i
v
i


j<i
v
j
, v
i
B,
() x
j
/ v
i
for j < i,
() z

/ v
i
for i < < cf().
Suppose that x
j
, v
j
have been dened for j < i. Let | = v
j
: j < i and
Y = X

| (so it is a closed subset of X). By (), for > i we have z

Y
and thus (Y )

(just look at the denition of Z

; remember X v
j
K),
and hence (Y ) . Consequently, we have sequences y
U

: < ) D
i
and
u
U

: < ) K
i
as chosen before (so they are as in (ii)). Consider a function g
dened on such that
g() = u
U

(z

: < cf() x
j
: j < i).
So by (ii)(c) we nd distinct , < such that g() = g() and y
U

/ u
U

. Then,
by the denition of K
i
, we nd v
i
K
i
B such that y
U

v
i
u
U

u
U

. It
follows from (ii)(b) that (v
i
Y ) = (v
i

j<i
v
j
)
i
. By claim 1.2.1 we may
pick x
i
Z
i
v
i
Y = Z
i
v
i


j<i
v
j
. Since, by our choices, v
i
is disjoint from
z

: < cf() x
j
: j < i, the inductive step is complete.
After the inductive construction is carried out, look at the sequence v
i
: i <
cf()). Since x
i
Z
i
v
i


j=i
v
j
we easily conclude that (v
i


j=i
v
j
)
i
.
Corollary 1.3 (See [20, 3.3., 5.4]). If B is a Boolean algebra satisfying s(B) =
then s
+
(B) =
+
.
Proof. Suppose s(B) = . Then for each i < cf() we may pick a discrete set
A
i
Ult(B) of size
i
. Let X =

i<cf()
A
i
(and the topology of X is the one
inherited from Ult(B)) and let B = b X : b B. Finally let (A) = [A[ for
A X. Note that X, B, clearly satisfy clauses 1.2(i,iii). Suppose that the demand
in 1.2(ii) fails for i < cf() and a closed set Y X (so [Y [ = ). Let
Y

i
= y Y : (v B)(y v [v Y [
i
).
Case 1: [Y

i
[ < .
Then [Y Y

i
[ = . For each y Y Y

i
pick v
y
B such that y v
y
and
[v
y
Y [ <
i
. Consider the function
F : Y Y

i
T(Y Y

i
) : y v
y
Y Y

i
.
By the Hajnal Free Set Theorem (see Hajnal [2]) there is an Ffree set S Y Y

i
of size . Then y / F(y

) for distinct y, y

S, and thus v
y
S = y for y S.
Consequently S is discrete and s
+
(B) > .
Case 2: [Y

i
[ .
For some j < cf() we have [Y

i
A
j
[ , so we may choose distinct y

Y

i
A
j
for < . The set y

: < is discrete (as A

j
is so), so we may pick u

B
6
5
1


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







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


FORCING FOR hL AND hd 7
such that (, < )(y

= ). Then y

, u

: < ) is as required in
1.2(ii), contradicting our assumption that this clause fails.
So we may assume that the assumptions of 1.2 are satised, and therefore we
may nd v
i
: i < cf()) B such that [v
i


j=i
v
j
[
i
for each i < cf(). Then,
for every i < cf(), there is (i) < cf() such that
[A
(i)
v
i

_
j=i
v
j
[
i
.
Let
A =
_
i<cf()
(A
(i)
v
i

_
j=i
v
j
).
Clearly [A[ = and easily A is discrete.
Theorem 1.4. If B is a Boolean algebra satisfying hL(B) = then
hL
+
(0)
(B) = hL
+
(1)
(B) = hL
+
(7)
(B) =
+
.
Proof. So assume hL(B) = .
If s
+
(B) > , that is if B has an ideal independent sequence of length , then
easily all sups in the equivalent denitions of hL are obtained. So we may assume
() s
+
(B) and thus, by 1.3, s
+
(B) < . We may also assume that s
+
(B) <

0
.
Let X = Ult(B), B = B, and for Y X let
(Y ) = sup : there is a right separated sequence in Y of length .
(Recall that in a topological space Y , a sequence y

: < ) is right separated

whenever all initial segments of the sequence are open in the relative topology.) We
are going to apply 1.2 to X, B, , and for that we need to check the assumptions
there. Clauses (i) and (iii) are obvious, and let us verify 1.2(ii).
Let i < cf() and let Y Ult(B) be a closed set such that (Y ) = . Let
x

: <
+
i
) Y be a right separated sequence, and let b

B be such that
x

and x

/ b

for < <

+
i
. Let
Z = <
+
i
: cf() =
i
& (a B)(x

a & (a Y ) <
i
).
Claim 1.4.1. Z is not stationary in
+
i
.
Proof of the claim. Assume Z is stationary. For Z pick a

B such that
x

and (a

Y ) <
i
. Note that then for some () < we have
( < )(x

< ()).
By the Fodor lemma, for some

the set Z

= Z : () =

is stationary.
Now look at the set Y

= x

: Z

& >

: we have
( Z

+ 1))((a

) Y

= x

).
Consequently Y

is a discrete set of size
+
i
, contradicting ().
Thus we may pick an increasing sequence () : < ) of ordinals below
+
i
such that cf(()) =
i
and () / Z (for < ). Let y

= x
()
and u

= b
()
.
Then y

, u

: < ) is as required in 1.2(ii) (for Y, i).

6
5
1


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







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


8 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
Consequently we may apply 1.2 to choose a sequence v
i
: i < cf()) B such
that
(i < cf())((v
i

_
j=i
v
j
)
i
).
For i < cf() choose a right separated sequence y
i

: <
i
) v
i+1


j=i+1
v
j
.
Let I consist of those b B that for some nite set W cf() and a sequence
(i) : i W)

iW

i
we have
(i < cf())( <
i
)(y
i

b i W & < (i)).

Claim 1.4.2. I is an ideal in B and cof(I) = . Consequently hL
+
(1)
(B) =
+
and
hence hL
+
(7)
(B) =
+
.
Proof of the claim. Plainly, I is an ideal in B. Suppose that A I is of size less
than , and for b A let W
b
[cf()]
<
,
b
(i) : i W
b
)

iW
b

i
witness b I.
Let i < cf() be such that
i
> [A[ and let sup
b
(i) : (b A)(i W
b
) < <
i
.
Take b B such that y
i

b and ( <
i
)( < y
i

/ b). Then
y
j

b v
i+1
j = i & ,
so v
i+1
b I, but it is not included in any member of Z.
Let Y = y
i

: i < cf() & <

i
.
Claim 1.4.3. L(Y ) = , and consequently hL
+
(0)
(B) =
+
.
Proof of the claim. For i < cf() and <
i
, let U
i,
be an open subset of v
i+1
such that
( <
i
)(y
i

U
i,
).
Put |
i
= U
i,
: <
i
, | =

i<cf()
|
i
. It should be clear that if |

|
i
is of size
less than
i
then Y

,= Y

|
i
. Also y
i

/

|
j
v
j
for i ,= j, so we may
conclude that no subfamily of | of size less than covers Y , showing the claim.

Theorem 1.5. If hd(B) = then hd

+
(8)
(B) =
+
(and thus also hd
+
(0)
(B) =
hd
+
(7)
(B) = hd
+
(5)
(B) =
+
).
Proof. We may follow like in 1.4 and use 1.2 to get our conclusion. However, an
alternative way is to use a result of

Sapirovski that for every compact space X,
hd(X) s(X)
+
(see

Sapirovski [17] or Hodel [6, 7.17]). Consequently, in our
situation, hd(B) = s(B) and by 1.3 we conclude that s
+
(B) =
+
. But this implies
that there is a homomorphic image B

of B with the cellularity c(B

) = (see [13,
Theorem 3.25 and p. 175]). Clearly d(B

) c(B

), so we get our conclusion.

6
5
1


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







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


FORCING FOR hL AND hd 9
2. Some combinatorics
Arguments based on the lemma are very important in forcing considerations.
The result quoted below is a variant of the lemma and in various forms was
presented, proved and developed in [21, 6], [19, 6] and [23, 7].
Lemma 2.1 (see [19, 6.1]). Assume that:
(i) , are regular cardinals and is a cardinal,
(ii) ( < )([[

< ),
(iii) T is a complete lter on containing all co-bounded subsets of ,
(iv)

: < ) is a sequence of ordinals (for < ),

(v) X is such that X ,= mod T.
Then there are a sequence

: < ) and a set w such that:

(a) ( w)( cf(

) ),
(b) the set
B
def
= X : if w then

,
if w then sup

: < ,

<

<

<

is not modulo the lter T,

(c) if

<

(for w) then
B : ( w)(

<

) ,= mod T.
The above version of the lemma will have multiple use in our proofs in the
next two sections. Among others, it will be applied to lters given by 2.2, 2.3 below.
Lemma 2.2. Suppose that B is a Boolean algebra generated by x

: < ). Let
I B be an ideal with cof(I) = and let
0
< < . Then there are a regular
cardinal [, ], a (< )complete lter T on and a sequence a

: < ) I
such that
(
1
) all co-bounded subsets of are in the lter T, and for every b I:
< : a

b = mod T,
(
2
) for each < , a

/ id
B
(a

: < ),
(
3
) every a

(for < ) is of the form

a

<n
x
t(,)
(,)
(where n < , (, ) < , t(, ) < 2).
Proof. It is basically like [22, 2.2, 2.3], but for readers convenience we present the
proof fully.
Claim 2.2.1. Assume
0
< . Then there are a regular cardinal [
0
, ] and a
set Y [I]

, such that
(Z [I]
<
)(b Y )(a Z)(b , a).
Proof of the claim. Assume not. By induction on [Y [ we show that then
() if Y [I]

then there is Y

I such that [Y

[ =
0
and
(b Y )(a Y

)(b a).
6
5
1


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







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


10 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
If [Y [
0
, then there is nothing to do.
Suppose now that Y I and [Y [ >
0
is a regular cardinal. Then, using the
assumption that the claim fails, we may nd a set Z I such that [Z[ < [Y [ and
(b Y )(a Z)(b a). Now apply the induction hypothesis to Z and get a set
Z

I of size
0
such that (a Z)(c Z

)(a c) clearly the set Z

works
for Y too.
So suppose now that Y I and [Y [ is a singular cardinal >
0
. Let Y =

<cf(|Y |)
Y

,
where [Y

[ < [Y [ (for < cf([Y [)). For each apply the inductive hypothesis to get
Y

I such that [Y

[ =
0
and (b Y

)(a Y

)(b a). Put Y

+
=

<cf(|Y |)
Y

and note that [Y

+
[ cf([Y [)
0
< [Y [. Again, apply the inductive hypothesis (),
this time to Y
+
, to get the respective Y

and note that it works for Y too.
To nish the proof of the claim note that the statement in () contradicts the
assumption that
0
< = cof(I).
If a set Y I is given by 2.2.1 for I,
0
, then we say that it is temporarily
(I,
0
, )good.
Claim 2.2.2. Suppose that Y I is temporarily (I, , )good, < [Y [. Assume
Y =

<
Y

. Then for some < the set Y

is temporarily (I, , )good.

Proof of the claim. Suppose that Y =

<
Y

, < [Y [ and no Y

is temporarily
(I, , )good. For < choose Z

I such that [Z

[ < [Y [ = and
(b Y

)(a Z

)(b a),
and put Z =

<
Z

. Then Z contradicts Y is temporarily (I, , )good. The

claim is shown.
Now, let Y I be a temporarily (I, , )good set, = [Y [, and let Y = b

:
< be an enumeration. Each b

can be represented as
b

j<j

<n
x
t(,j,)
(,j,)
.
By 2.2.2 we nd n

, j

and A []

such that ( A)(j

= j

& n

= n

) and
the set Y

= b

: A is temporarily (I, , )good. For j < j

and A let
b
j

=
_
<n

x
t(,j,)
(,j,)
and let Y
j
= b
j

: A. We claim that for some j < j

the
set Y
j
is temporarily (I, , )good. If not, then we nd Z
j
I (for j < j

) such
that [Z
j
[ < and ( A)(a Z
j
)(b
j

a). Put
Z = a
0
. . . a
j

1
: a
0
Z
0
, . . . , a
j

1
Z
j

and note that this set contradicts Y

is temporarily (I, , )good.
So let j
0
< j

be such that the set Y

def
= b
j0

: A is temporarily (I, , )
good and let Y

= a

: < be an enumeration.
For b I let F
b
= < : a

, b and let T
0
be the (< )complete lter of
subsets of generated by F
b
: b I.
First note that if < and b

: < ) I then (by the choice of Y

) we may
nd < such that ( < )(a

, b

). Consequently

<
F
b

,= and we may
6
5
1


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







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


FORCING FOR hL AND hd 11
conclude that T
0
is a proper lter on . Since / F
a
, we get that T
0
extends the
lter of co-bounded subsets of .
Claim 2.2.3. The set A
+
def
= < : a

id
B
(a

: < ) does not belong to

the lter T
0
.
Proof of the claim. Assume toward contradiction that A
+
T
0
. Thus we have
a sequence b

: < ) I, < , such that

<
F
b

A
+
. It follows from
the choice of Y

that Y

, id
B
(b

: < ). So let < be the rst such

that a

/ id
B
(b

: < ). This implies that a

<
F
b

A
+
, and thus
a

id
B
(a

: < ). By the minimality of we have id

B
(a

: < )
id
B
(b

: < ), and we get a contradiction.

Take the set A
+
from 2.2.3 and let T = X A
+
: X T
0
. Then the lter
T and a

: A
+
) satisfy the demands (
1
)(
3
) (after taking the increasing
enumeration of A
+
).
Lemma 2.3 (see [22, 2.2, 2.3]). Suppose cf() < , < . Assume that B is
a Boolean algebra generated by x

: < ) and I B is an ideal such that

(B/I) = . Then there are a regular cardinal [, ], a (< )complete lter T
on and a sequence a

: < ) B I such that

(
1
) the lter T contains all co-bounded subsets of and for every b B I:
< : b a

mod I = mod T,
(
2
) if < < then a

(a

) / I,
(
3
) every a

(for < ) is of the form

a

<n
x
t(,)
(,)
(where n < , (, ) < , t(, ) < 2).
Proof. It is an easy modication of [22, 2.2, 2.3] (and the proof is fully parallel to
that of Lemma 2.2 here).
One of the ways of describing Boolean algebras is giving a dense set of ultralters
(equivalently: homomorphisms from the algebra into 2). This is useful when we
want to force a Boolean algebra by smaller approximations (see the forcing notions
used in [22], [16]).
Denition 2.4. For a set w and a family F 2
w
we dene
cl(F) = g 2
w
: (u [w]
<
)(f F)(f u = g u),
B
(w,F)
is the Boolean algebra generated freely by x

: w except that
if u
0
, u
1
[w]
<
and there is no f F such that f u
0
0, f u
1
1
then
_
u1
x

_
u0
(x

) = 0.
Proposition 2.5 (see [22, 2.6]). Let F 2
w
. Then:
(1) each f F extends (uniquely) to a homomorphism from B
(w,F)
to 0, 1
(i.e. it preserves the equalities from the denition of B
(w,F)
),
(2) if (y
0
, . . . , y

) is a Boolean term and

0
, . . . ,

w are distinct then

B
(w,F)
[= (x
0
, . . . , x

) ,= 0 if and only if
(f F)(0, 1 [= (f(
0
), . . . , f(
k
)) = 1),
6
5
1


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







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


12 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
(3) if w w

, F

2
w

and
(f F)(g F

)(f g) and (g F

)(g w cl(F))
then B
(w,F)
is a subalgebra of B
(w

,F

)
.
Remark 2.6. Let F 2
w
. We will use the same notation for a member f of F and
the homomorphism from B
(w,F)
determined by it. Hence, for a Boolean term , a
nite set v w and f F, we may write f((x

: v)) etc.
Proposition 2.7. Let B be a Boolean algebra.
(1) A sequence a = a

: < ) of elements of B is:

ideal independent if and only if for each < there is a homomor-
phism f

: B 0, 1 such that
f

(a

) = 1 and ( < )( ,= f

(a

) = 0);
leftseparated if and only if for each < there is a homomorphism
f

: B 0, 1 such that
f

(a

) = 1 and ( < )( < f

(a

) = 0);
rightseparated if and only if for each < there is a homomorphism
f

: B 0, 1 such that
f

(a

) = 1 and ( < )(f

(a

) = 0).
(2) If the algebra B is generated by a sequence x

: < ), and there is an

ideal independent (leftseparated, rightseparated, respectively) sequence of
elements of B of length , then there is such a sequence with terms of the
form
a

k<k
x
t(,k)
(,k)
and where (, k) < , t(, k) 0, 1 and (, k) ,= (, k

) whenever
k < k

< k

.
3. Forcing for hL
In this section we show that consistently there is a Boolean algebra B of size
in which there is a strictly increasing sequence of ideals but every ideal in B is
generated by less than elements. This answers [12, Problem 43] (and thus a part
of [13, Problem 50]). The problem if the respective example can be constructed
just from cardinal arithmetic assumptions remains open.
Denition 3.1. (1) A good parameter is a tuple S = (, , ) such that ,
are cardinals satisfying
=
<
< cf() < and ( < cf())( < )(

< cf()),
and =
i
: i < cf()) is a strictly increasing sequence of regular cardinals
such that cf() <
i
< , (i < cf())(
<
i
=
i
) and = sup
i<cf()

i
.
(2) A good parameter S = (, , ) is a convenient parameter if additionally
cf() =
+
.
6
5
1


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







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


FORCING FOR hL AND hd 13
Denition 3.2. Let S = (, , ) be a convenient parameter and let the set
A
S
def
= (i, ) : i < cf() & <
i

be equipped with the lexicographic order

S
(i.e., (i, )
S
(i

) if and only if
either i < i

, or i = i

and <

).
(1) We dene a forcing notion Q
S
as follows.
A condition is a tuple p = w
p
, u
p
, f
p
i,
: (i, ) u
p
)) such that
(a) w
p
[cf()]
<
, u
p
[A
S
]
<
,
(b) (i w
p
)((i, 0) u
p
) and if (i, ) u
p
then i w
p
,
(c) for (i, ) u
p
, f
p
i,
: u
p
2 is a function such that
(j, ) u
p
& (j, )
S
(i, ) f
p
i,
(j, ) = 0,
and f
p
i,
(i, ) = 1,
the order is given by: p q if and only if
() w
p
w
q
, u
p
u
q
, and
() ((i, ) u
p
)(f
p
i,
f
q
i,
), and
() for each (i, ) u
q
one of the following occurs:
either f
q
i,
u
p
= 0
u
p,
or i w
p
and for some ,
i
we have (i, ) u
p
and f
q
i,
u
p
=
(f
p
i,
)

, where (f
p
i,
)

: u
p
2 is dened by
(f
p
i,
)

(j, ) =
_
0 if j = i, < ,
f
p
i,
(j, ) otherwise,
or i / w
p
and f
q
i,
u
p
= (f
p
j,
)

for some (j, ) u

p
and ,
j
,
where (f
p
j,
)

is dened as above.
(2) We say that conditions p, q Q
S
are isomorphic if the linear orders
(u
p
,
S
u
p
) and (u
q
,
S
u
q
)
are isomorphic, and if H : u
p
u
q
is the
S
isomorphism then:
() H(i, ) = (j, 0) if and only if = 0,
() f
p
i,
= f
q
H(i,)
H (for (i, ) u
p
).
In this situation we may call H an isomorphism from p to q.
Remark 3.3. (1) Of course,
S
is a well ordering of A
S
in the order type .
(2) The forcing notion Q
S
is a relative of the one used in [16, 7].
(3) There are isomorphism types of conditions in Q
S
(remember
<
= ).
A condition p Q
S
is determined by its isomorphism type and the set u
p
.
Proposition 3.4. Let S = (, , ) be a convenient parameter. Then Q
S
is a
(< )complete
+
cc forcing notion.
Proof. First we should check that Q
S
is really a partial order and for this we have
to verify the transitivity of . So suppose that p q and q r and let us justify
that p r. The only perhaps unclear demand is clause 3.2(1). Assume that
(i, ) u
r
and f
r
i,
u
p
,= 0
u
p and consider two cases.
Case 1: i w
p
.
Then i w
q
and, by the denition of (clause ()), we may pick
i
such
that (i, ) u
q
and f
r
i,
u
q
= (f
q
i,
)

. Again by clause (), for some

we have
6
5
1


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







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


14 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
(i,

) u
p
and f
q
i,
u
p
= (f
p
i,

. Now look at the denition of the operation ()

it should be clear that f

r
i,
u
p
= (f
p
i,
)

for some

.
Case 2: i / w
p
.
If i w
q
then for some , we have f
r
i,
u
q
= (f
q
i,
)

and f
q
i,
u
p
= (f
p
j,

for
some j,

. Now, since i / w
p
we may write f
r
i,
u
p
= (f
q
i,
)

u
p
= (f
p
j,
)

and we are done. Suppose now that i / w

q
. Then f
r
i,
u
q
= (f
q
j,
)

(for some
j, , ) and we ask if j w
p
. If so, then for some

we have f
q
j,
u
p
= (f
p
j,

and hence f
r
i,
u
p
= (f
p
j,
)

(for some

). If not (i.e., if j / w
p
) then like before
we easily conclude that f
r
i,
u
p
= (f
q
j,
)

u
p
= f
q
j,
u
p
= (f
p
j

(for some
j

).
Thus Q
S
is a forcing notion. To check that it is (< )complete suppose that
< and p

: < ) Q
S
is an increasing sequence of conditions. Put
w
p
=

<
w
p
, u
p
=

<
u
p
and for (i, ) u
p
let
f
p
i,
=
_
f
p
i,
: (i, ) u
p
, < .
Plainly, w
p
, u
p
, f
p
i,
: (i, ) u
p
)) Q
S
is an upper bound to p

: < ).
Now assume that / Q
S
is of size
+
. Since
<
= and cf() =
+
we may
use lemma and standard cleaning and nd conditions p, q / such that
(i) p, q are isomorphic (and let H : u
p
u
q
be the isomorphism),
(ii) H (u
p
u
q
) is the identity on u
p
u
q
,
(iii) sup(w
p
w
q
) < min(w
p
w
q
) sup(w
p
w
q
) < min(w
q
w
p
).
Now we are going to dene an upper bound r to p, q. To this end we put w
r
=
w
p
w
q
, u
r
= u
p
u
q
and for (i, ) u
r
we dene f
r
i,
: u
r
2 as follows.
If (i, ) u
p
, i w
p
w
q
then f
r
i,
= f
p
i,
(f
q
H(i,)
)

,
if (i, ) u
q
, i w
p
w
q
then f
r
i,
= (f
p
H
1
(i,)
)

f
q
i,
,
if (i, ) u
p
, i w
p
w
q
then f
r
i,
= f
p
i,
f
q
H(i,)
,
if (i, ) u
q
, i w
q
w
p
then f
r
i,
= 0
u
p f
q
i,
.
It should be clear that in all cases the functions f
r
i,
are well dened and that they
satisfy the demand 3.2(1c). Hence r = w
r
, u
r
, f
r
i,
: (i, ) u
r
)) Q
S
and one
easily checks that it is a condition stronger than both p and q. So we may conclude
that Q
S
satises the
+
chain condition.
For a condition p Q
S
let F
p
= 0
u
p (f
p
i,
)

: ,
i
, (i, ) u
p
, where
(f
p
i,
)

: u
p
2 is dened like in 3.2(1):
(f
p
i,
)

(j, ) =
_
0 if j = i, < ,
f
p
i,
(j, ) otherwise.
Further, let B
p
be the Boolean algebra B
(u
p
,F
p
)
(as dened in 2.4). Note that p q
implies that B
p
is a subalgebra of B
q
(remember 2.5). Let

B
0
S
be a Q
S
name such
that
QS

B
0
S
=

B
p
: p
QS
and for (i, ) A
S
let

f
i,
be a Q
S
name such
that

QS


f
i,
=
_
f
p
i,
: (i, ) u
p
, p
QS
.
Proposition 3.5. Assume that S = (, , ) is a convenient parameter. Then in
V
QS
:
6
5
1


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







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


FORCING FOR hL AND hd 15
(1)

f
i,
: A
S
2 (for (i, ) A
S
) is such that

f
i,
(i, ) = 1 and
((j, ) A
S
)((j, )
S
(i, )

f
i,
(j, ) = 0).
(2)

B
0
S
is the Boolean algebra B
(XS,

F)
(see 2.4), where

F = (

f
i,
)

: (i, ) A
S
,
i

and (

f
i,
)

: A
S
2 is such that
(

f
i,
)

(j, ) =
_
0 if j = i, < ,

f
i,
(j, ) otherwise,
(for (j, ) A
S
).
(3) The sequence x
i,
: (i, ) A
S
) is rightseparated in

B
0
S
(when we consider
A
S
with the well ordering
S
).
Proof. Should be clear (for the third clause remember that each

f
i,
extends to a
homomorphism from

B
0
S
to 0, 1, see 2.5; remember 2.7).
Theorem 3.6. Assume S = (, , ) is a convenient parameter. Then

QS
there is no ideal I

B
0
S
such that cof(I) = .
Proof. Let

I be a Q
S
name for an ideal in

B
0
S
, p Q
S
, and suppose that p
QS
cof(

I) = .
Fix i < cf() for a moment.
It follows 2.2 that we may choose p
i
,
i
, n
i
,

T
i
, e
i
and

t
i
such that
() p
i
Q
S
is a condition stronger than p,
i
is a regular cardinal,
+
i
<
i
<
and n
i
,
()

T
i
is a Q
S
name for a (<
i
)complete lter on
i
extending the lter of
co-bounded subsets of
i
,
()
QS
e
i
:
i
n
i
A
S
and

t
i
:
i
n
i
2 ;
for <
i
let a
i

be a Q
S
name for an element of

B
0
S
such that

QS
a
i

=

<ni
x

ti(,)
ei(,)
,
() p
i

QS
a
i

I for each <
i
,
() p
i

QS
if b

I then <
i
: a
i

b = mod

T
i
and
a
i

/ id

B
0
S
( a
i

: < ) for each <

i
.
For each <
i
choose an antichain p
i
,
: < of conditions stronger than p
i
,
maximal above p
i
, and such that each p
i
,
decides the values of e
i
(, ),

t
i
(, ).
Let
p
i
,

QS
e
i
(, ) = e

i
(, ) &

t
i
(, ) = t

i
(, ) (for < n
i
).
Plainly, we may demand that i w
p
i
,
and e

i
(, ) u
p
i
,
(for <
i
, < ,
< n
i
).
Suppose now that G Q
S
is a generic lter (over V) such that p
i
G and work
in V[G] for a while. Since the lter

T
G
i
is (<
i
)complete we nd ordinals
G
i
<
i
6
5
1


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







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


16 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
and

G
i
< such that the set

X
G
i
def
=
_
<
i
:
G
i
and p
i

G
i
,

G
i
, p
i
,

G
i
G and w
p
i

G
i
,

G
i
= w
p
i
,

G
i
,
the conditions p
i

G
i
,

G
i
, p
i
,

G
i
are isomorphic, and
if H : u
p
i

G
i
,

G
i
u
p
i
,

G
i
is the isomorphism then
( < n
i
)(H(e

G
i
i
(
G
i
, )) = e

G
i
i
(, ) & t

G
i
i
(
G
i
, ) = t

G
i
i
(, ))
and if j i, (j, ) A
S
then
(j, ) u
p
i

G
i
,

G
i
(j, ) u
p
i
,

G
i
_
is not modulo

T
G
i
(remember that in V[G] we still have cf()
<
= cf() and

<
i
=
i
). Let

G
i
= otp(u
p
i

G
i
,

G
i
,
S
) and for

X
G
i
let s
,i

: <

G
i
) be the

S
increasing enumeration of u
p
i
,

G
i
. Apply Lemma 2.1 to
+
,
i
,

G
i
,

T
G
i
and
s
,i

: <

G
i
) here standing for , , , T and

: < ) (respectively) there.

(Remember
S
is a well ordering of A
S
in the order type .) So we nd a sequence
s
,i

: <

G
i
) A
S
and a set v
G
i

G
i
such that
(i) (

G
i
v
G
i
)(
+
cf(s A
S
: s
S
s
,i

,
S
)
i
),
(ii) the set

B
G
i
def
=
_


X
G
i
: if v
G
i
then s
,i

= s
,i

, and
if

G
i
v
G
i
then
sup
S
s
,i

: <

G
i
, s
,i

S
s
,i

S
s
,i

S
s
,i

_
is not modulo the lter

T
G
i
,
(iii) if s

S
s
,i

for

G
i
v
G
i
then
_


B
G
i
: (

G
i
v
G
i
)(s

S
s
,i

)
_
,= mod

T
G
i
.
As there was no special role assigned to
G
i
(other than determining the order type
of a condition) we may assume that
G
i


B
G
i
.
Now we go back to V and we choose a condition q
i
Q
S
, ordinals
i
,
i
,
i
,
a set v
i
and a sequence s
,i

: <
i
) A
S
such that q
i
p
i
i,i
and q
i
forces
that these objects have the properties listed in (i)(iii) above. Note that if some
condition stronger than q
i
forces that

B
i
, then any condition stronger than
both q
i
and p
i
,i
does so. Then the conditions p
i
,i
and p
i
i,i
are isomorphic and
the isomorphism is the identity on u
p
i
,
i
u
p
i

i
,
i
, and it preserves e
i
i
, t
i
i
. Also then
w
p
i
,
i
= w
p
i

i
,
i
and u
p
i
,
i
(j
j
) = u
p
i

i
,
i
(j
j
) for j i. In this situation
we will use s
,i

: <
i
) to denote the
S
increasing enumeration of u
p
i
,
i
(and
so s
,i

= s
,i

for v
i
, and sup
S
s
,i

: <
i
, s
,i

S
s
,i

S
s
,i

S
s
,i

for
v
i
).
Claim 3.6.1. If j i < cf(), < n
i
and e
i
i
(
i
, ) = (j, ) (for some ) then
t
i
i
(
i
, ) = 1.
Proof of the claim. Suppose that the claim fails for some j
0
i,
0
<
j0
and
0
<
n
i
(i.e., t
i
i
(
i
,
0
) = 0 and e
i
i
(
i
,
0
) = (j
0
,
0
)). Choose such that
i
< <
i
and letting r
1
= p
i
i,i
, r
2
= p
i
,i
we have:
6
5
1


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







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


FORCING FOR hL AND hd 17
the conditions r
1
, r
2
are isomorphic and if H is the isomorphism from r
1
to
r
2
then H(e
i
i
(
i
, )) = e
i
i
(, ) and t
i
i
(
i
, ) = t
i
i
(, ) (for < n
i
),
w
r1
= w
r2
and the isomorphism H is the identity on u
r1
u
r2
,
(j, )
S
H(j, ) for (j, ) u
r1
u
r2
, and
if j i, (j, ) A
S
then (j, ) u
r1
(j, ) u
r2
.
Why is the choice possible? Let G Q
S
be generic over V such that q
i
G.
It follows from clauses (ii), (iii) that we may nd

B
G
i
(
i
+ 1) such that
(
i
v
i
)(s
i,i

S
s
,i

). Then the two ordinals

i
, have the required properties
in V[G], and hence clearly in V too.
Next we let w
r
= w
r1
= w
r2
, u
r
= u
r1
u
r2
and for (j, ) u
r
we dene
f
r
j,
: u
r
2 as follows.
If (j, ) u
r1
u
r2
then f
r
j,
= f
r1
j,
f
r2
j,
,
if (j, ) u
r1
u
r2
then f
r
j,
= f
r1
j,
f
r2
H(j,)
,
if (j, ) u
r2
u
r1
then f
r
j,
= (f
r1
H
1
(j,)
)

f
r2
j,
.
Check that the functions f
r
j,
are well dened and that
r = w
r
, u
r
, f
r
j,
: (j, ) u
r
)) Q
S
is a condition stronger than r
1
, r
2
. Let
1
=
_
<ni
x
t

i
i
(i,)
e

i
i
(i,)
and
2
=
_
<ni
x
t

i
i
(,)
e

i
i
(,)
.
Suppose that (j, ) u
r
and <
j
. If j i then (j
j
) u
r1
=
(j
j
) u
r2
and therefore (f
r
j,
)

(
1
) = (f
r
j,
)

(
2
). If j > i then necessarily
(f
r
j,
)

(j
0
,
0
) = 0, so (f
r
j,
)

(
1
) = (f
r
j,
)

(
2
) = 0. Consequently B
r
[=
1
=
2
and hence r a
i
i
= a
i

, contradicting clause () (and so nishing the proof of the

claim).
Take n < , < , v and an unbounded set Y cf() such that for i, j Y :
n
i
= n,
i
= , v
i
= v, and
the conditions p
i
i,i
, p
j
j,j
are isomorphic, and the isomorphism maps e
i
i
(
i
, )
and t
i
i
(
i
, ) onto e
j
j
(
j
, ), t
j
j
(
j
, ), respectively.
Now apply Lemma 2.1 to nd a sequence s
,
: < ) A
S
(cf(), 0) and a
set v

such that
(a) ( v

)(cf(s A
S
: s
S
s
,
,
S
) =
+
),
(b) the set
C
def
=
_
i Y : if v

then s
,i

= s
,
, and
if v

then
sup
S
s
,
: < , s
,

S
s
,

S
s
,i

S
s
,
_
is unbounded in cf(),
(c) if s

S
s
,
for v

, then the set

i C : ( v

)(s

S
s
,i

)
is unbounded in cf().
[So , , , T and

: < ) : < ) in 2.1 correspond to cf() =

+
,

and
the lter of co-bounded subsets of cf() and s
,i

: < ) : i < cf()) here.]

Next we use clauses (c), (a) and (iii), (i) to choose inductively a set C
+
C of
size cf() and ordinals
i
<
i
(for i C
+
) such that for every i C
+
:
6
5
1


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







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


18 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
(d) if v

then for all j C

+
i and < we have
s
,j

S
s
,
s
,j

S
s
,i

and s
j,j

S
s
,
s
j,j

S
s
,i

,
(e) some condition stronger than q
i
forces that
i


B
i
(see clause (ii) earlier),
(f) if v then for all j C
+
i and < we have
s
,j

S
s
,i

s
,j

S
s
i,i

and s
j,j

S
s
,i

s
j,j

S
s
i,i

,
(g) if v

, s
,
= (j, ) then j < min(C
+
).
Note that then
i, j C
+
& , < & s
j,j

= s
i,i

= v v

.
So s
i,i

: < ) : i C
+
) is a system of sequences with the heart s
,
:
v v

). Let u

= s
,
: v v

and w

= j < cf() : (j, 0) u

.
Pick i

C
+
such that [C
+
i

[ = .
Claim 3.6.2.
q
i

QS
(

B
i
)(j
1
, j
2
C
+
)( a
i

a
j1
j
1
a
j2
j
2
& p
j1
, p
j2

QS
) .
Proof of the claim. We are going to show that for every condition q q
i
and an
ordinal <
i
such that q

B
i
, there are a condition r q and ordinals
j
1
, j
2
C
+
such that
r a
i

a
j1
j
1
a
j2
j
2
& p
j1
, p
j2

QS
.
So suppose q q
i
and q

B
i
. We may assume that p
i

,
i

q (see the
denition of

X
i
,

B
i
). Choose j
1
C
+
i

and j
2
C
+
(i

+ 1) such that
u
q
u
p
j
1

j
1
,
j
1
= u
q
u
p
j
2

j
2
,
j
2
= u

and sup(w
q
) < min(w
p
j
2

j
2
,
j
2
w

).
(Remember that u
p
j

j
,
j
: j C
+
forms a system with heart u

and hence
w
p
j

j
,
j
: j C
+
forms a system with heart w

.)
To make the notation somewhat simpler let q
0
= p
i

,
i

, q
1
= p
j1
j
1
,j
1
and
q
2
= p
j2
j
2
,j
2
. Note that the conditions q
0
, q
1
, q
2
are pairwise isomorphic, and
the isomorphisms are the identity on the u

(which is the common part of any two

u
q
k
s). Put

0
=

<n
x
t

(,)
e

(,)
and
k
=

<n
x
t

j
k
j
k
(j
k
,)
e

j
k
j
k
(j
k
,)
(for k = 1, 2).
Thus
k
is an element of the algebra B
q
k . Clearly, for k, k

< 3, the isomorphism

H
k,k

from q
k
to q
k

carries
k
to
k
.
Now we are going to dene a condition r Q
S
stronger than q, q
1
and q
2
. For
this we put w
r
= w
q
w
q
1
w
q
2
, u
r
= u
q
u
q
1
u
q
2
and we dene functions
f
r
i,
: u
r
2 considering several cases.
(1) If (i, ) u
q
1
and i w

then we put f
r
i,
= f
q
H
1,0
(i,)
f
q
1
i,
f
q
2
H
1,2
(i,)
(note
that this includes the case (i, ) u

).
(2) If (i, ) u
q
1
, i / w

then we put f
r
i,
= 0
u
q f
q
1
i,
0
u
q
2 .
6
5
1


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







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


FORCING FOR hL AND hd 19
(3) If (i, ) u
q
u

then we look at f
q
i,
u
q
0
. If it is 0
u
q
0 then we let
f
r
i,
= f
q
i,
0
u
q
1 0
u
q
2 . Otherwise we nd (j, ) u
q
0
and
j
such
that f
q
i,
u
q
0
= (f
q
0
j,
)

and if i w
q
0
then i = j, and we dene:
() if j w

, j < i sup(w

) then f
r
i,
= f
q
i,
(f
q
1
H
0,1
(j,)
)
j
(f
q
2
H
0,2
(j,)
)
j
,
() if i = j w

then f
r
i,
= f
q
i,
(f
q
1
H
0,1
(j,)
)

(f
q
2
H
0,2
(j,)
)

, where

= max, ,
() if j w

, i < j then f
r
i,
= f
q
i,
(f
q
1
H
0,1
(j,)
)

(f
q
2
H
0,2
(j,)
)

,
() if either i > sup(w

) or j / w

then we rst choose j

w
q
2
and

j
such that (j

) u
q
2
and (f
q
2
j

(j

) = 0 whenever
(j

) u
q
2
, j

, and (f
q
2
j

(
2
) = 1 if possible (under our
conditions); next we let f
r
i,
= f
q
i,
0
u
q
1 (f
q
2
j

,
)

.
(4) If (i, ) u
q
2
u

, i w

then we let f
r
i,
= (f
q
H
2,0
(i,)
)

(f
q
1
H
2,1
(i,)
)

f
q
2
i,
.
(5) If (i, ) u
q
2
, i / w

then we put f
r
i,
= 0
u
q 0
u
q
1 f
q
2
i,
.
It should be a routine to check that in all cases the function f
r
i,
is well dened
and that r = w
r
, u
r
, f
r
i,
: (i, ) u
r
)) Q
S
is a condition stronger than q, q
1
, q
2
(and thus stronger than p
j1
, p
j2
). [Remember that w

min(C
+
), so for j w

we have (j, ) u
q
0
(j, ) u
p
i

i
,
i

and hence, when checking clause 3.2(1c)

in Case 1, we may use clauses (d), (f) of the choice of the set C
+
. They imply
that if (i, ) u
q
1
, i w

then (i, ) _
S
H
1,0
(i, ) _
S
H
1,2
(i, ). Considering Case
3() with j / w

, use the fact that min(w

q
0
w

) sup(w

) (it follows from our

choices). Similarly in Case 2 remember min(w
q
1
w

) sup(w

).]
We claim that B
r
[=
0

1

2
and for this we have to show that there is no
function f F
r
with f(
0
) = 1 and f(
1
) = f(
2
) = 0 (see 2.5). So suppose toward
contradiction that f F
r
is such a function. Note that f cannot be 0
u
r as then the
values given to all the terms would be the same (remember they are isomorphic).
So for some (i, ) u
r
and
i
we have f = (f
r
i,
)

. Let us look at all the

cases appearing in the denition of the functions f
r
j,
s (we keep labeling as there
so we do not repeat the descriptions of the cases).
Case 1: Clearly f
r
i,
(
0
) = f
r
i,
(
1
). It follows from the demands (d), (f) of the
choice of C
+
that if i w

, (i, ) u
q
0
, (i

) = H
0,1
(i, ), then i

= i and

.
Consequently, we may use 3.6.1 to conclude that (f
r
i,
)

(
0
) (f
r
i,
)

(
1
), what
contradicts the choice of f.
Case 2: Plainly (f
r
i,
)

(
0
) = (f
r
i,
)

(
2
).
Case 3: Note that f
r
i,
(
0
) = f
r
i,
(
1
) and, as j < i sup(w

), necessarily
i / w
q
0
w
q
1
. Hence easily (f
r
i,
)

(
0
) = (f
r
i,
)

(
1
).
Cases 3, , 4: Like in cases 1, 3 we conclude (f
r
i,
)

(
0
) (f
r
i,
)

(
1
).
Case 3: It follows from the choice of

, j

there that f
r
i,
(
0
) f
r
i,
(
2
). If
i / w
q
0
then (as also i / w
q2
) we have f(
0
) = f
r
i,
(
0
) and f(
2
) = f
r
i,
(
2
), so we
are done. If i w
q
0
then i = j and we easily nish by the choice of

, j

.
Case 5: Clearly (f
r
i,
)

(
0
) = (f
r
i,
)

(
1
), a contradiction.
6
5
1


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







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


20 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
Thus we may conclude that r a
i

a
j1
j
1
a
j2
j
2
, nishing the proof of the
claim.
Now we may easily nish the theorem: take a generic lter G Q
S
over V
such that q
i
G and work in V[G]. Since the lter

T
G
i
is (<
i
)complete and
cf() <
i
, we nd j
1
, j
2
C
+
such that p
j1
, p
j2
G and


B
G
i
: ( a
i

)
G
( a
j1
j
1
)
G
( a
j2
j
2
)
G
,= mod

T
G
i

(remember

B
G
i
,= mod

T
G
i
by (ii)). But then also ( a
j1
j
1
)
G
( a
j2
j
2
)
G

I
G
, so we
get a contradiction to clause ().
Conclusion 3.7. It is consistent that there is a Boolean algebra B of size such that
there is a rightseparated sequence of length in B, (so hL
+
(7)
(B) =
+
), but there is
no ideal I B with the generating number (and thus hL
+
(1)
(B) = hL
(1)
(B) = ).
Problem 3.1. Does there exist a Boolean algebra B as in 3.7 in semi-ZFC? I.e.,
can one construct such an algebra for from cardinal arithmetic assumptions?
4. Forcing for hd
Here we deal with a problem parallel to the one from the previous section and
related to the attainment question for hd. We introduce a forcing notion P
S
comple-
mentary to Q
S
and we use it to show that, consistently, there is a Boolean algebra
B of size in which there is a strictly decreasing sequence of ideals but every
homomorphic image of B has algebraic density less than . This gives a partial
answer to [13, Problem 54]. Again, we do not know if an example like that can be
constructed from cardinal arithmetic assumptions.
Denition 4.1. Let S = (, , ) be a good parameter (see 3.1) and let A
S
,
S
be as dened in 3.2.
(1) We dene a forcing notion P
S
as follows.
A condition is a tuple p = w
p
, u
p
, f
p
i,
: (i, ) u
p
)) such that
(a) w
p
[cf()]
<
, u
p
[A
S
]
<
,
(b) (i w
p
)((i, 0) u
p
) and if (i, ) u
p
then i w
p
,
(c) for (i, ) u
p
, f
p
i,
: u
p
2 is a function such that
(j, ) u
p
& (i, )
S
(j, ) f
p
i,
(j, ) = 0,
and f
p
i,
(i, ) = 1,
the order is given by: p q if and only if
() w
p
w
q
, u
p
u
q
, and
() ((i, ) u
p
)(f
p
i,
f
q
i,
), and
() for each (i, ) u
q
one of the following occurs:
either f
q
i,
u
p
= 0
u
p,
or i w
p
and for some , <
i
we have (i, ) u
p
and f
q
i,
u
p
=
(f
p
i,
)

, where (f
p
i,
)

: u
p
2 is dened by
(f
p
i,
)

(j, ) =
_
0 if j = i, <
i
,
f
p
i,
(j, ) otherwise,
6
5
1


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







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


FORCING FOR hL AND hd 21
or i / w
p
and either f
q
i,
u
p
= (f
p
j,
)

(dened above) for some

(j, ) u
p
, <
j
or f
q
i,
u
p
= (f
p
j,
)
j
for some (j, ) u
p
and
j

j, where (f
p
j,
)
j
: u
p
2 is dened by
(f
p
j,
)
j
(j

) =
_
0 if j

,
f
p
j,
(j

) otherwise.
(2) Conditions p, q P
S
are said to be isomorphic if the well orderings
(u
p
,
S
u
p
) and (u
q
,
S
u
q
)
are isomorphic, and if H : u
p
u
q
is the
S
isomorphism then:
() H(i, ) = (j, 0) if and only if = 0,
() f
p
i,
= f
q
H(i,)
H (for (i, ) u
p
).
Proposition 4.2. Let S = (, , ) be a good parameter. Then P
S
is a (< )
complete
+
cc forcing notion.
Proof. Plainly P
S
is a (< )complete forcing notion (compare the proof of 3.4).
To verify the chain condition suppose that / P
S
, [/[ =
+
. Apply the
lemma and standard cleaning to choose isomorphic conditions p, q / such that
if H : u
p
u
q
is the isomorphism from p to q then H (u
p
u
q
) is the identity
on u
p
u
q
. Put w
r
= w
p
w
q
, u
r
= u
p
u
q
and for (i, ) u
r
dene a function
f
r
i,
: u
r
2 as follows.
If (i, ) u
p
, i w
p
w
q
then f
r
i,
= f
p
i,
(f
q
H(i,)
)
+1
,
if (i, ) u
q
, i w
p
w
q
then f
r
i,
= (f
p
H
1
(i,)
)
+1
f
q
i,
,
if (i, ) u
p
, i w
p
w
q
then f
r
i,
= f
p
i,
(f
q
H(i,)
)
i
,
if (i, ) u
q
, i w
q
w
p
then f
r
i,
= (f
p
H
1
(i,)
)
i
f
q
i,
.
It is a routine to check that the functions f
r
i,
are well dened and that they satisfy
the demand 4.1(1c). Hence r = w
r
, u
r
, f
r
i,
: (i, ) u
r
)) P
S
and one easily
checks that it is an upper bound to both p and q.
For a condition p P
S
let
F
p
= (f
p
i,
)

, (f
p
i,
)
j
: (i, ) u
p
, <
i
, j i,
where (f
p
i,
)

, (f
p
i,
)
j
: u
p
2 are dened like in 4.1(1):
(f
p
i,
)

(i

) =
_
0 if i = i

,
f
p
i,
(i

) otherwise,
(f
p
i,
)
j
(i

) =
_
0 if j i

,
f
p
i,
(i

) otherwise.
Like in the previous section, B
p
is the Boolean algebra B
(u
p
,F
p
)
(see 2.4) (note that
p q implies that B
p
is a subalgebra of B
q
). Let

B
1
S
be a P
S
name such that

PS


B
1
S
=
_
B
p
: p
PS
,
and for s A
S
let

f
s
be a P
S
name such that

PS


f
s
=
_
f
p
s
: s u
p
, p
PS
.
Proposition 4.3. Assume that S = (, , ) is a good parameter. Then in V
PS
:
6
5
1


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







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


22 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
(1) For s A
S
,

f
s
: A
S
2 is such that

f
s
(s) = 1 and
(s

A
S
)(s
S
s

f
s
(s

) = 0).
(2)

B
1
S
is the Boolean algebra B
(XS,

F)
(see 2.4), where

F = (

f
i,
)

, (

f
i,
)
j
: (i, ) A
S
, <
i
, j i,
and (

f
i,
)

, (

f
i,
)
j
: A
S
2 are such that
(

f
i,
)

(i

) =
_
0 if i = i

f
i,
(i

) otherwise,
(

f
i,
)
j
(i

) =
_
0 if j i

f
i,
(i

) otherwise.
(3) The sequence x
s
: s A
S
) is leftseparated in

B
1
S
(when we consider A
S
with the well ordering
S
).
Theorem 4.4. Assume S = (, , ) is a good parameter. Then

PS
there is no ideal I

B
1
S
such that (

B
1
S
/I) = .
Proof. Not surprisingly, the proof is similar to the one of 3.6. Let

I be a P
S
name
for an ideal in

B
1
S
, p P
S
, and suppose that p
PS
(

B
1
S
/

I) = .
Fix i < cf(). Use 2.3 to choose p
i
,
i
, n
i
,

T
i
, e
i
and

t
i
such that
() p
i
P
S
is a condition stronger than p,
i
is a regular cardinal,
+
i
<
i
<
and n
i
,
()

T
i
is a P
S
name for a (<
i
)complete lter on
i
extending the lter of
co-bounded subsets of
i
,
()
PS
e
i
:
i
n
i
A
S
and

t
i
:
i
n
i
2 ;
for <
i
let a
i

be a P
S
name for an element of

B
1
S
such that

PS
a
i

<ni
x

ti(,)
ei(,)
,
() p
i

PS
a
i

B
1
S


I for each <
i
,
() p
i

PS
if b

B
1
S


I then <
i
: b a
i

mod

I = mod

T
i
and
( <
i
)( < )( a
i

( a
i

) /

I) .
For each <
i
choose a maximal above p
i
antichain p
i
,
: < such that each
p
i
,
p
i
decides the values of e
i
(, ),

t
i
(, ). Let
p
i
,

PS
e
i
(, ) = e

i
(, ) &

t
i
(, ) = t

i
(, ) (for < n
i
),
and we may assume that (i, 0), e

i
(, ) u
p
i
,
for <
i
, < n
i
and < . Take a
generic lter G P
S
such that p
i
G and work in V[G]. Choose ordinals
G
i
<
i
6
5
1


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







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


FORCING FOR hL AND hd 23
and

G
i
< such that the set

X
G
i
def
=
_
<
i
:
G
i
and p
i

G
i
,

G
i
, p
i
,

G
i
G and w
p
i

G
i
,

G
i
= w
p
i
,

G
i
,
the conditions p
i

G
i
,

G
i
, p
i
,

G
i
are isomorphic, and
if H : u
p
i

G
i
,

G
i u
p
i
,

G
i is the isomorphism then
( < n
i
)(H(e

G
i
i
(
G
i
, )) = e

G
i
i
(, ) & t

G
i
i
(
G
i
, ) = t

G
i
i
(, ))
and if j i, (j, ) A
S
then
(j, ) u
p
i

G
i
,

G
i (j, ) u
p
i
,

G
i
_
is not modulo

T
G
i
. Let

G
i
= otp(u
p
i

G
i
,

G
i
,
S
) and for

X
G
i
let s
,i

: <

G
i
)
be the
S
increasing enumeration of u
p
i
,

G
i
. Apply Lemma 2.1 to nd a sequence
s
,i

: <

G
i
) A
S
and a set v
G
i

G
i
such that
(i) (

G
i
v
G
i
)(
+
i
cf(s A
S
: s
S
s
,i

,
S
)
i
),
(ii) the set

B
G
i
def
=
_


X
G
i
: if v
G
i
then s
,i

= s
,i

, and
if

G
i
v
G
i
then
sup
S
s
,i

: <

G
i
, s
,i

S
s
,i

S
s
,i

S
s
,i

_
is not modulo the lter

T
G
i
,
(iii) if s

S
s
,i

for

G
i
v
G
i
then


B
G
i
: (

G
i
v
G
i
)(s

S
s
,i

) ,= mod

T
G
i
.
We may assume that
G
i


B
G
i
.
Now, in V, we choose a condition q
i
P
S
, ordinals
i
,
i
,
i
, a set v
i
and a
sequence s
,i

: <
i
) A
S
such that q
i
p
i
i,i
, and q
i
forces that these objects
are as described above. If some condition stronger than q
i
forces that

B
i
, then
we will use s
,i

: <
i
) to denote the
S
increasing enumeration of u
p
i
,
i
.
Next, like in the proof of 3.6, we pick an unbounded set Y cf() and n < ,
< , v such that for i, j Y :
n
i
= n,
i
= , v
i
= v, and
the conditions p
i
i,i
, p
j
j,j
are isomorphic, and the isomorphism maps e
i
i
(
i
, )
and t
i
i
(
i
, ) onto e
j
j
(
j
, ), t
j
j
(
j
, ), respectively.
Now use Lemma 2.1 to nd a sequence s
,
: < ) A
S
(cf(), 0) and a set
v

such that
(a) ( v

)(cf(s A
S
: s
S
s
,
,
S
) = cf()),
(b) the set
C
def
=
_
i Y : if v

then s
,i

= s
,
, and
if v

then
sup
S
s
,
: < , s
,

S
s
,

S
s
,i

S
s
,
_
is unbounded in cf(),
6
5
1


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







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


24 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
(c) if s

S
s
,
for v

, then the set

i C : ( v

)(s

S
s
,i

)
is unbounded in cf().
Next choose a set C
+
[C]
cf()
and ordinals
i
<
i
<
i
(for i C
+
) such that
for every i C
+
:
(d) if v

then for all j C

+
i and < we have
s
,j

S
s
,
s
,j

S
s
,i

, and
s
j,j

S
s
,
s
j,j

S
s
,i

, and
s
j,j

S
s
,
s
j,j

S
s
,i

,
(e) some condition stronger than q
i
forces that
i
,
i


B
i
,
(f) if v and x
i
,
i
, then for all j C
+
i and < we have
s
,j

S
s
,i

s
,j

S
s
x,i

and s
j,j

S
s
,i

s
j,j

S
s
x,i

, and
s
j,j

S
s
,i

s
j,j

S
s
x,i

and s
i,i

S
s
,i

s
i,i

S
s
i,i

,
(g) if v

, s
,
= (j, ) then j < min(C
+
).
Then s
i,i

, s
i,i

: < ) : i C
+
) forms a system of sequences with heart
s
,
: v v

); but note that s

i,i

= s
i,i

for v. Let u

= s
,
: v v

and w

= j < cf() : (j, 0) u

.
Claim 4.4.1. For each i
0
C
+
:
q
i0

PS
(

B
i0
)(i

C
+
)( a
i

( a
i

) a
i0

& p
i

PS
)
(where

B
i0
was dened in (ii)).
Proof of the claim. Let i
0
C
+
. We will show that for every condition q q
i0
and
an ordinal <
i0
such that q

B
i0
, there are i

C
+
and a condition r
stronger than both q and p
i
, and such that r a
i

( a
i

) a
i0

.
So suppose q q
i
, q

B
i0
. We may assume that p
i0
,i
0
q. Choose
i

C
+
(i
0
+ 1) such that
u
q
u
p
i

i
,
i

= u
q
u
p
i

i
,
i

= u

and w
q
i

,
Let p
i0
,i
0
= q
0
, p
i

i
,
i

= q
1
, p
i

i
,
i

= q
2
, and

0
=

<n
x
t

i
0
i
0
(,)
e

i
0
i
0
(,)
,
1
=

<n
x
t

(
i
,)
e

(
i
,)
and
2
=

<n
x
t

(
i
,)
e

(
i
,)
(so q
0
q and
0
B
q
0 B
q
,
1
B
q
1 ,
2
B
q
2 ). Note that the conditions
q
0
, q
1
, q
2
are pairwise isomorphic and the isomorphism H
k,k

from q
k
to q
k

carries

k
to
k
. Moreover, H
k,k

is the identity on u
q
k
u
q
k

. Also note that w

q
1
=
w
p
i

i
,
i

= w
q
2
and, as w
q
i

, our choices imply H

k,0
(i, ) _
S
(i, ) for k = 1, 2,
(i, ) u
q
k
.
Now we dene a condition r stronger than q, q
1
, q
2
. We put w
r
= w
q
w
q
1
,
u
r
= u
q
u
q
1
u
q
2
and we dene functions f
r
i,
: u
r
2 as follows.
(1) If (i, ) u
q
1
u
q
2
, i w
q
then we let f
r
i,
= f
q
H
1,0
(i,)
f
q
1
i,
f
q
2
i,
.
[Note that by (d)+(ii) we have (i, 0) _
S
H
1,0
(i, ) _
S
(i, ).]
6
5
1


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







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


FORCING FOR hL AND hd 25
(2) If (i, ) u
q
1
u
q
2
, i / w
q
then we rst choose

such that, if possible,

(f
q
0
H
1,0
(i,)
)

(
0
) = 1, and then we let f
r
i,
= (f
q
H
1,0
(i,)
)

f
q
1
i,
f
q
2
i,
.
[Note that H
1,0
(i, )
S
(i, ), and thus if H
1,0
(i, ) = (j, ) then j < i,
j / w
q
1
.]
(3) If (i, ) u
q
2
u
q
1
(so i > i

sup(w
q
)) then we rst choose

such
that, if possible, (f
q
0
H
2,0
(i,)
)

(
0
) = 1, and then we let f
r
i,
= (f
q
H
2,0
(i,)
)

f
q
1
H
2,1
(i,)
f
q
2
i,
.
[Note that H
2,0
(i, )
S
(i, 0)
S
H
2,1
(i, )
S
(i, ); remember w
q
1
= w
q
2
.
Also, if H
2,0
(i, ) = (j, ), then j / w
q
1
.]
(4) If (i, ) u
q
1
u
q
2
then, like above, we choose

such that if possible

then (f
q
0
H
1,0
(i,)
)

(
0
) = 1, and next we put f
r
i,
= (f
q
H
1,0
(i,)
)

f
q
1
i,

(f
q
2
H
1,2
(i,)
)
+1
.
(5) If (i, ) u
q
u
q
1
then we look at f
q
i,
u
q
0
. If it is 0
u
q
0 then we let
f
r
i,
= f
q
i,
0
u
q
1 0
u
q
2 . Otherwise, we consider the following three cases.
() Suppose i w
q
0
. Then for some <
i
, + 1 we have
f
q
i,
u
q
0
= (f
q
0
i,
)

and we let:
if i w
q
1
then f
r
i,
= f
q
i,
(f
q
1
H
0,1
(i,)
)

(f
q
2
H
0,2
(i,)
)

,
if i / w
q
1
then f
r
i,
= f
q
i,
(f
q
1
H
0,1
(i,)
)
i
(f
q
2
H
0,2
(i,)
)
i
.
[Note that if i w
q
1
then (i, ) _
S
H
0,1
(i, ) = H
0,2
(i, )
S
(i + 1, 0),
and if i / w
q
1
then (j, 0) _
S
H
0,1
(i, ) _
S
H
0,2
(i, )
S
(j + 1, 0) for some
j > i.]
() Suppose i / w
q
0
(so i / w
q
1
) and f
q
i,
u
q
0
= (f
q
0
i

, (i

) u
q
0
,

<
i
.
If i

w
q
1
and i

< i, then put f

r
i,
= f
q
i,
(f
q
1
H
0,1
(i

)
)

(f
q
2
H
0,2
(i

)
)

.
If i

w
q
1
and i < i

, then we put f
r
i,
= f
q
i,
(f
q
1
H
0,1
(i

)
)
i

(f
q
2
H
0,2
(i

)
)
i
.
If i

/ w
q
1
, then let f
r
i,
= f
q
i,
(f
q
1
H
0,1
(i

)
)
i
(f
q
2
H
0,2
(i

)
)
i
.
() Suppose i / w
q
0
and f
q
i,
u
q
0
= (f
q
0
i

)
j
, j

mini, i

, (i

) u
q
0
.
Let f
r
i,
= f
q
i,
(f
q
1
H
0,1
(i

)
)
j
(f
q
2
H
0,2
(i

)
)
j
.
Verifying that the functions f
r
i,
are well dened and that r = w
r
, u
r
, f
r
i,
: (i, )
u
r
)) P
S
is a condition stronger than q, q
1
, q
2
is left to the reader. Let us argue
that B
r
[=
1
(
2
)
0
. If not then we have a function f F
r
such that
f(
0
) = f(
2
) = 0 and f(
1
) = 1. Clearly f cannot be 0
u
r , so it is either (f
r
i,
)

or (f
r
i,
)
j
. Let us look at the denition of the functions f
r
i,
and consider each case
there separately.
Cases 1, 5, , : Plainly f
r
i,
(
1
) = f
r
i,
(
2
) and also (f
r
i,
)
j
(
1
) = (f
r
i,
)
j
(
2
)
(remember w
q
1
= w
q
2
). As far as the operation ()

is concerned, note that (i

6
5
1


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







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


26 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH

i
) u
q
1
= (i
i
) u
q
2
, so (in these cases) we easily get (f
r
i,
)

(
1
) = (f
r
i,
)

(
2
),
a contradiction.
Case 2: Again, f
r
i,
(
1
) = f
r
i,
(
2
) and (f
r
i,
)
j
(
1
) = (f
r
i,
)
j
(
2
) (for each j). So
suppose that f = (f
r
i,
)

for some , and look at the choice of

in the current
case. Since 1 = (f
r
i,
)

(
1
) = (f
q
1
i,
)

(
1
), we conclude that 1 = (f
q
0
H
1,0
(i,)
)

(
0
) =
f
r
i,
(
0
) = (f
r
i,
)

(
0
), a contradiction.
Case 3: Note that f
r
i,
(
1
) = f
r
i,
(
2
) (and also (f
r
i,
)
j
(
1
) = (f
r
i,
)
j
(
2
)). Now,
if for some we have (f
r
i,
)

(
1
) = 1, then look at the choice of

necessarily
(f
r
i,
)

(
0
) = f
r
i,
(
0
) = 1 (remember (i, 0)
S
H
2,1
(i, )
S
(i, )).
Case 4: Like above: if for some we have (f
r
i,
)

(
1
) = 1, then necessarily
f
r
i,
(
0
) = (f
r
i,
)

(
0
) = 1. Moreover, (f
r
i,
)
j
(
1
) = (f
r
i,
)
j
(
2
) for all j i.
In all cases we get a contradiction showing that B
r
[=
1
(
2
)
0
, and hence
r a
i

( a
i

) a
i0

, nishing the proof of the claim.

Finally we note that 4.4.1 and clauses (), () give an immediate contradiction,
showing the theorem.
Conclusion 4.5. It is consistent that there is a Boolean algebra B of size such
that there is a leftseparated sequence of length in B (and thus hd
+
(5)
(B) =
+
),
but there is no ideal I B with (B/I) = (so hd
+
(7)
(B) = hd
(7)
(B) = ).
Problem 4.1. Can one construct a Boolean algebra B as in 4.5 for from any
cardinal arithmetic assumptions?
5. More on the attainment problem
In this section we will assume the following:
Hypothesis 5.1. S = (, , ) is such that , are cardinals satisfying
=
<
< cf() < 2

,
and =
i
: i < cf()) is a strictly increasing continuous sequence of cardinals
such that

0
= 0, cf() <
1
, cf(
i+1
) =
i+1
, and sup
i<cf()

i
= .
For < let j() < cf() be such that
j()
<
j()+1
.
Denition 5.2. (1) A pair ( , A) is a base for S = (, , ) if
(a) A
<
, =

: < )

,
(b) if < < , j() = j() then

/ A, and
(c) if Y []

then there are distinct , Y such that

A.
(2) ( , A) is called a base
+
for S if it satises demands (a), (b) (stated above)
and
(c
+
) if Y []

and t 0, 1, then there are , Y such that

< ,

A, and

<
lex

i t = 0.
6
5
1


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







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


FORCING FOR hL AND hd 27
For a topological space X, a (
0
,
1
)Lusin set in X is a set L X such that
[L[ =
0
and for every meager subset Z of X the intersection Z L is of size less
than
1
. (See, e.g., Cicho n [1] for a discussion of sets of this type.) Below, the
space

is equipped with the topology generated by sets of the form

[] =

:
for
<
.
Proposition 5.3. Assume that for some i

< cf() there is a (,

i
)Lusin set L
in

. Then there is a base

+
for S.
Proof. Choose sequences
i
: i < cf())

and

: < ) L, both with no

repetitions. For < let

be dened by

(2 ) =
j()
() and

(2 + 1) =

()
(for < ), and let A =

<

2
. We claim that (

: < ), A) is a base
+
for
S. The conditions 5.2(1)(a,b) should be clear. Let us verify 5.2(2)(c
+
). So suppose
that Y []

and t 0, 1. Choose sequences Y

i
: i < cf()) and j
i
: i < cf())
such that
Y
i
Y , ( Y
i
)(j() = j
i
), and [Y
i
[ =
i
(so

: Y
i
is not
meager),
the sequence j
i
: i < cf()) is strictly increasing.
For each i < cf() pick
i

<
such that
(
<
)(
i
[]

: Y
i
,= ).
We may pick i
0
< i
1
< cf() such that

i0
=
i1
=

and
ji
0
<
lex

ji
1
i t = 0.
(Remember that, under the assumptions of 5.1, (

, <
lex
) contains no monotonic
sequences of length cf().) Let = lh(
ji
0

ji
1
) and take

<
such that

and < lh(

). Now pick
0
Y
i0
and
1
Y
i1
such that

0

1
(there are such
0
,
1
by the choice of
i0
=
i1
=

). Note that then necessarily

0
<
1
, lh(
0

1
) = 2 (so
0

1
A) and
0
<
lex

1
i t = 0.
Proposition 5.4. Let P = (2
<
, ) be the Cohen forcing notion. Then

P
there is a base
+
for S (and S is still as in 5.1) .
Proof. Pick sequences
i
: i < cf()) and

: < ) of pairwise distinct elements

of

. Let

A

be a Pname for the generic subset of (added by P) and let

A be
a Pname such that

P


A =
<
: lh()

A

.
For < , let

be a Pname for a function in

such that

P
(

A

)(

() =
j()
(otp(

A

)) & (

A

)(

() =

(otp(

A

))).
We claim that

P
(

: < ), A) is a base
+
for S .
Clauses 5.2(1)(a,b) should be clear, so let us prove 5.2(2)(c
+
) only. Let

: < )
be a Pname for an increasing sequence of elements of , and let t 0, 1, p P.
For each < pick a condition p

p and an ordinal

such that p

.
6
5
1


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







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


28 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
Necessarily, there are X []

and p

P such that p

= p

for X. Then also

0
<
1
for
0
<
1
from X. Shrinking X a little we may also demand that for
some sequences
j

lh(p

) + 2
(for j < cf()) we have
X & j(

) = j
j

.
Now pick
0
<
1
from X such that letting j
0
= j(
0
) and j
1
= j(
1
) we have
j
0
< j
1
and
j0
=
j1
and
j0
<
lex

j1
i t = 0.
Let a condition q p

be such that lh(q) = lh(p

) + lh(
j0

j1
) + 2 and q() = 1
for all lh(q) lh(p

). It should be clear that

0
<
1
and
q

1


A and

0
<
lex

1
i t = 0 .

Denition 5.5. Let b = ( , A) be a base for S, =

: < ). We dene the

Boolean algebra B
b
determined by b. First, functions f
b

: 2 (for < ) are

such that
f
b

() =
_
1 if = or ,= &

A &

<
lex

,
0 otherwise.
Next, we let F
b
= f
b

: < and B
b
= B
(,F
b
)
(see 2.4).
Theorem 5.6. If b is a base for S = (, , ), then
hL(B
b
) = hd(B
b
) = s
+
(B
b
) = .
If additionally b is a base
+
for S then also
hL
+
(7)
(B
b
) = hd
+
(5)
(B
b
) = .
Proof. Let b = ( , A), =

: < ). Clearly [B
b
[ = .
Claim 5.6.1. hL(B
b
) = hd(B
b
) = s(B
b
) = .
Proof of the claim. By 5.2(1)(b), f
b

() = 0 whenever ,= and j() = j().

Therefore, by 2.7(1), the sequence x

:
i
<
i+1
) is ideal independent (for
each i < cf()).
The main part is to show that s
+
(B
b
) = (and/or under the additional as-
sumption, that hL
+
(7)
(B
b
) = hd
+
(5)
(B
b
) = ), and for this we will need the following
technical claim.
Claim 5.6.2. Suppose that k

< ,
k
,
,k
< (for k < k

, <

) and

0
, . . . ,
k

1

<
are such that
()
0
, . . . ,
k

1
are pairwise incomparable,
()
k

k
,
k

,k
(for <

, k < k

),
() for each k < k

one of the following occurs:

(i)
k
=
,k
for some <

, or
(ii) there are
1
,
2
,
3
<

such that

1
,k

2
,k

3
,k
, and

1
,k
,

2
,k
A, and

1
,k
<
lex

k
<
lex

2
,k
.
6
5
1


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







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


FORCING FOR hL AND hd 29
Let t(k) 0, 1 for k < k

. Then
B
b
[=

k<k

x
t(k)

<

k<k

x
t(k)

,k
.
Proof of the claim. We are going to show that, under our assumptions, for each
f F
b
there is <

such that (k < k

)(f(
k
) = f(
,k
)). So let us x < ,
and we consider f
b

. First note that

(
k
) if
k
is not an initial segment of

, then f
b

(
k
) = f
b

(
,k
) for all <

.
[Why? Suppose
k

. Then clearly
k
,= ,=
,k
(for <

) and

,k

and

k
<
lex

,k
<
lex

.
Now look at the denition of f
b

.]
If no
k
is an initial segment of

, then (by (
k
)) we conclude f
b

(
k
) = f
b

(
,k
)
for all <

, k < k

. So suppose that
m

, m < k

. Then for all k < k

,
k ,= m, we have
k

and thus f
b

(
k
) = f
b

(
,k
) (for all <

). Thus it
is enough to nd <

such that f
b

(
m
) = f
b

(
,m
). If
m
=
,m
for some
<

, then this works. So suppose

m
,=
,m
for all <

. Then clause ()(ii)

holds true for m, and let
1
,
2
,
3
be as there. If
m

3
,m
, then
easily f
b

(
m
) = f
b

(
3,m
). Otherwise
m

3
,m

m

, and f
b

(
1,m
) ,=
f
b

(
2,m
), so either
1
or
2
works.
Claim 5.6.3. s
+
(B
b
) = .
Proof of the claim. Suppose that a

: < ) is an ideal independent sequence

in B
b
. We may assume that a

=
_
k<k

x
t(,k)
(,k)
and (, k) ,= (, k

) whenever
k < k

< k

(remember 2.7(2)). Also we may assume that k

= k

for all < (as

cf() > ).
Fix i < cf() for a moment.
After possibly re-enumerating the sequences (, k) : k < k

), we may nd a
set S
i
[
i
,
i+1
), an ordinal

i
< , a sequence
i
k
: k < k

) of pairwise distinct
elements of

i
, and t
i
k
0, 1 and j
i
k
< cf() (for k < k

) such that
(i) S
i
is unbounded in
i+1
,
(ii) t(, k) = t
i
k
and j((, k)) = j
i
k
for all S
i
and k < k

,
(iii)
i
k

(,k)
for k < k

and S
i
,
(iv) (, k) : k < k

) : S
i
) is a system of sequences with heart
i
k
:
k < k(i)),
(v) the sequence (, k) : S
i
) is strictly increasing for k(i) k < k

,
(vi) j
i
k
i for k(i) k < k

(it follows from (ii)+(iv)).

Next pick a set S [cf()]
cf ()
such that (possibly after some re-enumerations)
(vii) k(i) = k
+
, t
i
k
= t
k
,

i
=

and
i
k
=

k
for k < k

, i S,
(viii)
i
k
: k < k
+
) : i S) is a system of sequences with heart
k
: k <
k

),
(ix) j
i
k
: k < k

) : i S) is a system of sequences with heart j

k
: k w),
w k

.
Note that then k

w k
+
. Also, possibly further shrinking S and the S
i
s (for
i S), we may demand that
6
5
1


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







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


30 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
(x) if i
1
< i
2
, i
1
, i
2
S, then j
i1
k
< i
2
(for k < k

),
(xi) if i
1
, i
2
S are distinct,
1
S
i1
and
2
S
i2
, then
(
1
, k) : k < k

(
2
, k) : k < k

=
k
: k < k

.
Let S

=

iS
S
i
. For < and k
+
k < k

let
S
L
,k
=
_
S

: ( S

)
_
> lh(
(,k)

(,k)
) or
(,k)

(,k)
/ A or

(,k)

lex

(,k)
__
,
S
R
,k
=
_
S

: ( S

)
_
> lh(
(,k)

(,k)
) or
(,k)

(,k)
/ A or

(,k)

lex

(,k)
__
.
We claim that both [S
L
,k
[ < and [S
R
,k
[ < . Why? Assume, e.g., [S
L
,k
[ = .
Note that, by (v)+(vi)+(x), (, k) < (, k) for < from S

. Pick

and
a set X [S
L
,k
]

such that ( X)(

(,k)
). By 5.2(1)(c), there are distinct
, X such that
(,k)

(,k)
A. Clearly lh(
(,k)

(,k)
) and we
easily get a contradiction with , S
L
,k
. Similarly for S
R
,k
.
For k
+
k < k

let
S

k
=
_
S

: for all < there exists S

such that
(,k)
<
lex

(,k)
,
and lh(
(,k)

(,k)
) and
(,k)

(,k)
A,
and
for all < there exists S

such that
(,k)
<
lex

(,k)
,
and lh(
(,k)

(,k)
) and
(,k)

(,k)
A
_
.
Note that S

k
=

<
(S
L
,k
S
R
,k
), and hence [S

k
[ < for each k [k
+
, k

).
Fix distinct

m1

k=k
+
S

k
such that j(

) = j(

). For each k [k
+
, k

) pick

k
1
,
k
2
,
k
3
S

such that

k

(

,k)

(
k
1
,k)

(

,k)

(
k
2
,k)

(

,k)

(
k
3
,k)
,

,k)

(
k
1
,k)
,
(

,k)

(
k
2
,k)
A,

(
k
1
,k)
<
lex

(

,k)
<
lex

(
k
2
,k)
.
Now look: letting
k
= (

, k),
,k
: <

be the suitable enumeration of

_
(
k

n
, k) : k
+
k

< k

& n 1, 2, 3
_
(

, k), and
k
=

k
, we get that
the clauses ()() of 5.6.2 are satised. Hence
a

k<k

x
t
k
(

,k)

k<k

x
t
k
(,k)

3

n=1
k

=k
+

k<k

x
t
k
(
k

n
,k)
= a

n=1
k

=k
+
a

n
.
Since clearly

n
: k
+
k

< k

, n = 1, 2, 3, we get a contradiction.
Claim 5.6.4. If b is a base
+
then also hL
+
(7)
(B
b
) = hd
+
(5)
(B
b
) = .
Proof of the claim. It is similar to 5.6.3. Suppose that a

: < ) is a right
separated sequence in B
b
. Like before we may assume that a

=
_
k<k

x
t(,k)
(,k)
and
(, k) ,= (, k

) whenever k < k

< k

. Next we apply the same cleaning

6
5
1


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







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


FORCING FOR hL AND hd 31
procedure as in 5.6.3 getting S, S
i
,

k
, t
k
, j
i
k
etc such that clauses (i)(xi) are
satised. We let S

=

iS
S
i
and for < and k
+
k < k

we dene
S
+
,k
=
_
S

: ( S

)
_
>lh(
(,k)

(,k)
) or
(,k)

(,k)
/ A
or
(,k)

lex

(,k)
__
,
S

,k
=
_
S

: ( S

)
_
>lh(
(,k)

(,k)
) or
(,k)

(,k)
/ A
or
(,k)

lex

(,k)
__
.
Then both [S
+
,k
[ < and [S

,k
[ < . [It is like before: assume, e.g., [S
+
,k
[ = .
Pick

and a set X [S
+
,k
]

such that ( X)(

(,k)
). Note that
(, k) < (, k) for < from S

. Use 5.2(2)(c
+
) to nd < , both from X,
such that
(,k)

(,k)
A and
(,k)
<
lex

(,k)
. A clear contradiction.]
Next for k
+
k < k

we let S

k
= S

<
(S
+
,k
S

,k
). Choose

<

from
m1

k=k
+
S

k
such that j(

) = j(

). And next for each k [k

+
, k

) pick
k
1
,
k
2
,
k
3

S

like those in the proof of 5.6.3. Finish in the same way.

References
[1] Jacek Cicho n. On two-cardinal properties of ideals. Transactions of the American Mathemat-
ical Society, 314:693708, 1989.
[2] Andras Hajnal. Proof of a conjecture of S.Ruziewicz. Fundamenta Mathematicae, 50:123128,
1961/1962.
[3] Andras Hajnal and Istvan Juhasz. Discrete subspaces of topological spaces. Nederl. Akad.
Wetensch. Proc. Ser. A = Indag. Math., 29:343356, 1967.
[4] Andras Hajnal and Istvan Juhasz. Discrete subspaces of topological spaces. II. Nederl. Akad.
Wetensch. Proc. Ser. A = Indag. Math., 31:1830, 1969.
[5] Andras Hajnal and Istvan Juhasz. Some remarks on a property of topological cardinal func-
tions. Acta Mathematica Academiae Scientiarum Hungaricae, 20:2537, 1969.
[6] Richard E. Hodel. Cardinal functions. i. In K. Kunen and J. E. Vaughan, editors, Handbook
of set-theoretic topology, pages 161. North-Holland, 1984.
[7] Thomas Jech. Set theory. Academic Press, New York, 1978.
[8] Istvan Juhasz. Cardinal functions in topology, volume 34 of Mathematical Centre Tracts.
Mathematisch Centrum, Amsterdam, 1971.
[9] Istvan Juhasz. Cardinal functions in topology ten years later, volume 123 of Mathematical
Centre Tracts. Mathematisch Centrum, Amsterdam, 1980.
[10] Istvan Juhasz and Saharon Shelah. How large can a hereditarily separable or hereditarily
Lindelof space be? Israel Journal of Mathematics, 53:355364, 1986.
[11] Kenneth Kunen and Judith Roitman. Attaining the spread at cardinals of conality . Pacic
Journal of Mathematics, 70:199205, 1977.
[12] Donald Monk. Cardinal Invariants of Boolean Algebras. Lectures in Mathematics. ETH
Zurich, Birkhauser Verlag, Basel Boston Berlin, 1990.
[13] Donald Monk. Cardinal Invariants of Boolean Algebras, volume 142 of Progress in Mathe-
matics. Birkhauser Verlag, BaselBostonBerlin, 1996.
[14] Judith Roitman. Attaining the spread at cardinals which are not strong limits. Pacic Journal
of Mathematics, 57:545551, 1975.
[15] Andrzej Roslanowski and Saharon Shelah. Cardinal invariants of ultrapoducts of Boolean
algebras. Fundamenta Mathematicae, 155:101151, 1998, math.LO/9703218
1
.
[16] Andrzej Roslanowski and Saharon Shelah. More on cardinal functions on Boolean algebras.
Annals of Pure and Applied Logic, 103:137, 2000, math.LO/9808056.
1
References of the form math.XX/ refer to the arXiv.org/archive/math archive
6
5
1


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







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


32 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
[17] Boris

Sapirovski. Canonical sets and character. Density and weight in bicompacta. (Russian).
Dokl. Akad. Nauk SSSR, 218:5861, 1974.
[18] Saharon Shelah. Constructing Boolean algebras for cardinal invariants. Algebra Universalis,
accepted, math.LO/9712286.
[19] Saharon Shelah. PCF and innite free subsets. Archive for Mathematical Logic, accepted,
math.LO/9807177.
[20] Saharon Shelah. Remarks on the numbers of ideals of Boolean algebra and open sets of a
topology. In Around classication theory of models, volume 1182 of Lecture Notes in Math-
ematics, pages 151187. Springer, Berlin, 1986.
[21] Saharon Shelah. Further cardinal arithmetic. Israel Journal of Mathematics, 95:61114, 1996,
math.LO/9610226.
[22] Saharon Shelah. On Monks questions. Fundamenta Mathematicae, 151:119, 1996,
math.LO/9601218.
[23] Saharon Shelah. Special Subsets of
cf()
, Boolean Algebras and Maharam measure Algebras.
Topology and its Applications, 99:135235, 1999, math.LO/9804156. 8th Prague Topological
Symposium on General Topology and its Relations to Modern Analysis and Algebra, Part II
(1996).
Department of Mathematics, University of Nebraska at Omaha, Omaha, NE 68182-
0243, USA, and Mathematical Institute of Wroclaw University, 50384 Wroclaw, Poland
E-mail address: roslanowski@unomaha.edu
URL: http://www.unomaha.edu/aroslano
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