Download as pdf or txt
Download as pdf or txt
You are on page 1of 42

5

9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


More on cardinal invariants of Boolean algebras
Andrzej Roslanowski
Institute of Mathematics
The Hebrew University
91 904 Jerusalem, Israel
and
Mathematical Institute
Wroc law University
50 384 Wroc law, Poland
Saharon Shelah

Institute of Mathematics
The Hebrew University
91 904 Jerusalem, Israel
and
Department of Mathematics
Rutgers University
New Brunswick, NJ 08903, USA
August 17, 2011
Abstract
We address several questions of Donald Monk related to irredun-
dance and spread of Boolean algebras, gaining both some ZFC knowl-
edge and consistency results. We show in ZFC that irr(B
0
B
1
) =
maxirr(B
0
), irr(B
1
). We prove consistency of the statement there
is a Boolean algebra B such that irr(B) < s(B B) and we force
a superatomic Boolean algebra B

such that s(B

) = inc(B

) = ,
irr(B

) = Id(B

) =
+
and Sub(B

) = 2

+
. Next we force a su-
peratomic algebra B
0
such that irr(B
0
) < inc(B
0
) and a superatomic
algebra B
1
such that t(B
1
) > Aut(B
1
). Finally we show that con-
sistently there is a Boolean algebra B of size such that there is no
free sequence in B of length , there is an ultralter of tightness (so
t(B) = ) and / Depth
Hs
(B).

Partially supported by Basic Research Foundation of the Israel Academy of Sci-


ences and Humanities. Publication 599.
0
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 1
0 Introduction
In the present paper we answer (sometimes partially only) several questions
of Donald Monk concerning cardinal invariants of Boolean algebras. Most
of our results are consistency statements, but we get some ZFC knowledge
too.
For a systematic study and presentation of current research on cardinal
invariants of Boolean algebras (as well as for a long list of open problems)
we refer the reader to Monk [M2]. Some of the relevant denitions are listed
at the end of this section.
Content of the paper: In the rst section we show that the dierence
between s
n
(B) and s
N
(B) (for n < N) can be reasonably large, with the
only restriction coming from the inequality s
n
(B) 2
s
N
(B) (a consistency
result; for the denitions of the invariants see below). It is relevant for the
description of the behaviour of spread in ultraproducts: we may conclude
that it is consistent that s(

n
B
n
/D) is much larger than

n
s(B
n
)/D. In
the following section we answer [M2, Problem 24] showing that irr(B
0
B
1
) =
maxirr(B
0
), irr(B
1
) (a ZFC result). A partial answer to [M2, Problem 27]
is given in the third section, where we show that, consistently, there is a
Boolean algebra B such that irr(B) < s(BB). In particular, this shows that
the parallel statement to the result of section 2 for free product may fail.
Note that proving the result of section 3 in ZFC is a really dicult task, as so
far we even do not know if (in ZFC) there are Boolean algebras B satisfying
irr(B) < [B[. In section 4 we force a superatomic Boolean algebra B such
that s(B) = inc(B) = , irr(B) = Id(B) =
+
and Sub(B) = 2

+
. This gives
answers to [M2, Problems 73, 77, 78] as stated (though the problems in ZFC
remain open). Next we present some modications of this forcing notion and
in the fth section we answer [M2, Problems 79, 81] forcing superatomic
Boolean algebras B
0
, B
1
such that irr(B
0
) < inc(B
0
) and Aut(B
1
) < t(B
1
).
Finally in the last section we show that (consistently) there is a Boolean
algebra B of size such that there is no free sequence in B of length , there
is an ultralter in Ult(B) of tightness (so t(B) = ) and / Depth
Hs
(B).
This gives answers to [M2, Problems 13, 41]. Lastly we use one of the results
of [Sh 233] to show that 2
cf(t(B))
< t(B) implies t(B) Depth
Hs
(B).
Notation: Our notation is rather standard and compatible with that of
classical textbooks on set theory (like Jech [J]) and Boolean algebras (like
Monk [M1], [M2]). However in forcing considerations we keep the older
tradition that
the stronger condition is the greater one
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 2
Let us list some of our notation and conventions.
Notation 0.1 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
.
2. , , , , . . . 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. 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 is called Ult(B).
5. For a subset Y of an algebra B, the subalgebra of B generated by Y is
denoted by Y )
B
.
6. The sign stands for the operation of the free product of Boolean
algebras and the product is denoted by .
The invariants: Below we recall some denitions and formalism from
[RoSh 534] (see [M2] too).
Denition 0.2 For a (not necessary rst order) theory T in the language
of Boolean algebras plus one distinguished unary predicate P
0
plus, possibly,
some others P
1
, P
2
, . . . we dene cardinal invariants inv
T
, inv
+
T
of Boolean
algebras by (for a Boolean algebra B):
inv
T
(B)
def
= sup[P
0
[ : (B, P
n
)
n
is a model of T,
inv
+
T
(B)
def
= sup[P
0
[
+
: (B, P
n
)
n
is a model of T.
We think of the spread s(B) of a Boolean algebra B as
(
s
) s(B) = sup[X[ : X B is ideal-independent
(it is one of the equivalent denitions, see [M2, Thm 13.1]). Thus we can
write s(B) = s

(B), where
Denition 0.3 1.
s
n
is the formula saying that no member of P
0
can
be covered by union of n + 1 other elements of P
0
.
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 3
2. For 0 < n let T
n
s
=
s
k
: k < n.
3. For a Boolean algebra B and 0 < n : s
(+)
n
(B) = inv
(+)
T
n
s
(B).
The hereditary density hd(B) and the hereditary Lindelof degree hL(B) of a
Boolean algebra B are treated in a similar manner. We use [M2, Thm 16.1]
and [M2, Thm 15.1] to dene them as
(
hd
) hd(B) = sup[[ : there is a strictly decreasing sequence of ideals (in
B) of the length ,
(
hL
) hL(B) = sup[[ : there is a strictly increasing sequence of ideals (in
B) of the length .
This leads us directly to the following denition.
Denition 0.4 1. Let the formula say that P
1
is a well ordering of P
0
(denoted by <
1
).
2. For n < let
hd
n
,
hL
n
be the following formulas:

hd
n
& (x
0
, . . . , x
n+1
P
0
)(x
0
<
1
. . . <
1
x
n+1
x
0
, x
1
. . .x
n+1
)

hL
n
& (x
0
, . . . , x
n+1
P
0
)(x
n+1
<
1
. . . <
1
x
0
x
0
, x
1
. . .
x
n+1
).
3. For 0 < n we let T
n
hd
=
hd
k
: k < n, T
n
hL
=
hL
k
: k < n.
4. For a Boolean algebra B and 0 < n :
hd
(+)
n
(B) = inv
(+)
T
n
hd
(B), hL
(+)
n
(B) = inv
(+)
T
n
hL
(B).
We use the following characterization of tightness (see [M2, 12]):
t(B) = sup[[ : there exists a free sequence of the length in B.
Denition 0.5 1. Let be the sentence saying that P
1
is a well ordering
of P
0
(we denote the respective order by <
1
). For k, l < let
t
k,l
be
the sentence asserting that
for each x
0
, . . . , x
k
, y
0
, . . . , y
l
P
0
if x
0
<
1
. . . <
1
x
k
<
1
y
0
<
1
. . . <
1
y
l
then
_
ik
x
i
,
_
il
y
i
,
and let the sentence
ut
k,l
say that
for each distinct x
0
, . . . , x
k
, y
0
, . . . , y
l
P
0
we have

ik
x
i
,

il
y
i
.
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 4
2. For n, m let T
n,m
t
=
t
k,l
: k < n, l < m and T
n,m
ut
=

ut
k,l
: k < n, l < m and for a Boolean algebra B:
t
n,m
(B) = inv
T
n,m
t
(B) & ut
n,m
(B) = inv
T
n,m
ut
(B).
The irredundance irr(B) of a Boolean algebra B is the supremum of
cardinalities of sets X B such that (x X)(x / X x)
B
).
Denition 0.6 (compare [M2, p. 144]) Let n and let T
n
irr
be the
theory of the language of Boolean algebras plus a predicate P
0
, which says
that for each m < n and a Boolean term (y
0
, . . . , y
m
) we have
(x P
0
)(x
0
, . . . , x
m
P
0
x)(x ,= (x
0
, . . . , x
m
)).
For Boolean algebra B we dene irr
(+)
n
(B) = inv
(+)
T
n
irr
(B) (so irr

(B) = irr(B)).
The incomparability number inc(B) is the supremum of cardinalities of sets
of pairwise incomparable elements. The number of ideals in B is denoted
by Id(B), Aut(B) stands for the number of automorphisms of the algebra B,
and the number of subalgebras of B is denoted by Sub(B).
1 Forcing for spread
The aim of this section is to show that for N much larger than n, the
inequalities 2
s
N
(B) s
n
(B) s
N
(B) (see [M2, Thm 13.6]) seem to be the
only restriction on the jumps between s
N
and s
n
. The forcing notion dened
in 1.1(2) below is a modication of the one from [Sh 479, 2] and a relative
of the forcing notion from [Sh 620, 15].
Denition 1.1 1) 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
_
u
1
x

_
u
0
(x

) = 0.
2) Let be cardinals, 0 < n < . We dene forcing notion Q
,
n
:
a condition is a pair p = (w
p
, F
p
) such that w
p
[]
<
, F
p
2
w
p
,
[F
p
[ < and for every u [w
p
]
n
there is f

: w
p
u 2 such that
if h : u 2 then f

h F
p
;
the order is given by p q if and only if w
p
w
q
and
(f F
q
)(f w
p
cl(F
p
)) and (f F
p
)(g F
q
)(f g).
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 5
Proposition 1.2 (see [Sh 479, 2.6]) 1. If p Q
,
n
, f F
p
then f
extends to a homomorphism from B
p
to 0, 1 (i.e. it preserves the
equalities from the denition of B
p
).
2. If p Q
,
n
, (y
0
, . . . , y
k
) is a Boolean term and
0
, . . . ,
k
w
p
are
distinct then
B
p
[= (x

0
, . . . , x

k
) ,= 0 if and only if
(f F
p
)(0, 1 [= (f(
0
), . . . , f(
k
)) = 1).
3. If p, q Q
,
n
, p q then B
p
is a subalgebra of B
q
.
Proposition 1.3 Assume
<
= , 0 < n < . Then
1. Q
,
n
is a complete forcing notion of size
<
,
2. Q
,
n
satises
+
-cc.
Proof This is almost exactly like [Sh 479, 2.7]. For (1) no changes are
required; for (2) one has to check that the condition dened as there is
really in Q
,
n
. So suppose that p

: <
+
) Q
,
n
. Applying standard
cleaning procedure nd
0
<
1
<
+
such that
otp(w
p
0
) = otp(w
p
1
),
if H : w
p
0
w
p
1
is the order preserving mapping then H (w
p
0

w
p
1
) is the identity on w
p
0
w
p
1
and F
p
0
= f H : f F
p
1

(remember
<
= ; use lemma). Let w
q
= w
p
0
w
p
1
and
F
q
= f g : f F
p
0
& g F
p
1
& f (w
p
0
w
p
1
) = g (w
p
0
w
p
1
).
To check that q = (w
q
, F
q
) is in Q
,
n
suppose that u [w
q
]
n
and let
u

= H
1
[u w
p
1
] (u w
p
0
) [w
p
0
]
n
. Let f

0
: w
p
0
u

2 be
such that if h : u

2 then f

0
h F
p
0
. Next, let f

: w
p
0
u 2 be
such that f

0
f

and if u

u then f

() = 0, and let g

: w
p
1
u 2
be such that f

0
H
1
g

and if H[u

] u then g

() = 0. Now it
should be clear that
if h : u 2 then (f

) h F
q
.
Verifying that both p

0
q and p

1
q is even easier.
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 6
Let

B be the Q
,
n
name for

B
p
: p
Q
,
n
. It follows from 1.2 that

Q
,
n


B is a Boolean algebra generated by x

: <
and, for a condition p Q
,
n
,
p
Q
,
n
x

: w
p
)

B
= B
p
.
Theorem 1.4 Assume
<
= and 0 < N, n < are such that
2
n/2
+n N. Then

Q
,
n
ind
+
n
(

B) =
+
and t
+
1,N
(

B) = t
+
N,1
(

B) = ind
+
(

B) =
+
.
Proof It follows immediately from the denition of Q
,
n
(by density
arguments, remembering 1.2) that

Q
,
n
the sequence x

: < ) is nindependent .
Suppose now that a

: <
+
) is a Q
,
n
name for a
+
sequence of
elements of

B, p Q
,
n
. For each <
+
choose a condition p

p, a
Boolean term

and ordinals (, 0) < . . . < (,

) < such that


p


Q
,
n
a

(x
(,0)
, . . . , x
(,

)
).
By system arguments, passing to a subsequence and increasing p

s, we
may assume that
(i)

= ,

= and (, 0), . . . , (, ) w
p

,
(ii) otp(w
p

0
) = otp(w
p

1
) and otp(w
p

0
(
0
, j)) = otp(w
p

1
(
1
, j))
for j ,
0
,
1
<
+
,
(iii) w
p

: <
+
forms a system of sets with heart w

,
(iv) if H

0
,
1
: w
p

0
w
p

1
is the order preserving mapping then H

0
,
1

w

is the identity on w

and F
p

0
= f H

0
,
1
: f F
p

1
.
After this cleaning procedure look at the conditions p
0
, . . . , p
N
. We want
to show that they have a common upper bound q Q
,
n
such that q
Q
,
n
a
0

_
j<N
( a
1+j
) = 0 . To this end dene:
w
q
= w
p
0
. . . w
p
N
and
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 7
F
q
= f
0
. . . f
N
: f
0
F
p
0
, . . . , f
N
F
p
N
, f
0
w

= . . . = f
N
w

,
and if 0, 1 [= (f
0
( (0, 0)), . . . , f
0
( (0, ))) = 1
then for some j [1, N]
0, 1 [= (f
j
( (j, 0)), . . . , f
j
( (j, ))) = 1 .
Let us check that q = (w
q
, F
q
) is in Q
,
n
. Clearly each f F
q
is a function
from w
q
to 2 and [F
q
[ < . Suppose now that u [w
q
]
n
. Let u

= uw

and u
+
=

iN
H
i,0
[u w
p
i
] [w
p
0
]
n
. One of the sets u

, u
+
u

has
size at most n/2, and rst we deal with the case [u

[ n/2. Choose f

:
w
p
0
u
+
2 such that (h : u
+
2)(f

h F
p
0
). For each v u

choose h
v
: u
+
2 such that h
v
v 1, h
v
(u

v) 0 and
if there is h : u
+
2 satisfying the above demands and such
that 0, 1 [= ((f

h)( (0, 0)), . . . , (f

h)( (0, ))) = 1


then h
v
has this property.
Since 2
|u

|
+n N we may choose distinct i
v
[1, N] for v u

such that
w
p
iv
u = u

. Now we dene functions f

i
: w
p
i
u 2 (for i N) as
follows:
if i = i
v
, v u

then f

i
= (f

h
v
) H
i,0
,
if i / i
v
: v u

then f

i
f

H
i,0
is such that f

i
() = 0 for all
H
i,0
[u

] u.
Suppose that h : u 2 and let f
i
= f

i
(h (uw
p
i
)). It should be clear
that for each i N we have f
i
F
p
i
and f
i
w

= f
0
w

(remember the
choice of f

). Assume that 0, 1 [= (f
0
( (0, 0)), . . . , f
0
( (0, ))) = 1. Look
at v = h
1
[1] u

and the corresponding i


v
. By the above assumption
and the choice of h
v
, f

iv
we have
0, 1 [= (f
iv
( (i
v
, 0)), . . . , f
iv
( (i
v
, ))) = 1.
This shows that

iN
f
i
F
q
and hence we conclude q Q
,
n
. If [u
+
u

[
n/2 then we proceed similarly: for v u
+
u

we choose distinct i
v
[1, N]
such that w
p
iv
u = u

. We pick f

as in the previous case and we dene


f

i
: w
p
i
u 2 (for i N) as follows
if i = i
v
, v u
+
u

then f

i
= f

H
i,0
and ( u
+
u

)(f

i
(H
0,i
()) =
1 v),
if i / i
v
: v u
+
u

then f

i
f

H
i,0
is such that f

i
() = 0 for
all H
i,0
[u
+
] u.
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 8
Next we argue like before to show that q Q
,
n
.
Checking that q is a common upper bound of p
0
, . . . , p
N
is straightfor-
ward. Finally, by the denition of F
q
and by 1.2(2) we see that
q
Q
,
n
a
0

j=1
( a
j
) = 0 .
Thus we have proved that
Q
,
n
t
+
1,N
(

B)
+
. The same arguments
show that
Q
,
n
t
+
N,1
(

B)
+
(just considering a

instead of a

and
0, . . . , N 1, N as the two groups of indexes there).
To show that the equalities hold one can prove even more: in V
Q
,
n
,
there is an independent subset of

B of size . The construction of the set
is easy once you note that if p Q
,
n
, w
p
and w
q
= w
p
,
F
q
= f 2
w
q
: f w
p
F
p
then q = (w
q
, F
q
) is a condition in Q
,
n
stronger than p.
Conclusion 1.5 Assume that
<
= < . Then there is a forcing
notion P which does not change cardinalities and conalities and such that in
V
P
: 2

and there are Boolean algebras B


0
, B
1
, B
2
, . . . of size satisfying
ind
+
n+1
(B
n
) =
+
and hd
+
(B
n
) = hL
+
(B
n
) = ind
+
(B
n
) =
+
.
Consequently, in V
P
, for every non-principal ultralter T on we have
inv(

n<
B
n
/T) =

and

n
inv(B
n
)/T =

,
where inv ind, t, hd, hL, s.
Proof Let P
0
be the forcing notion adding many Cohen subsets of
(with conditions of size < ) and for n > 0 let P
n
be Q
,
n
. Let P be the
<support product of the P
n
s (so if = then P is the nite support
product of the P
n
s and otherwise it is the full product).
Claim 1.5.1 P is a closed
+
cc forcing notion of size
<
.
Proof of the claim: Modify the proof of 1.3.
Let

B
n
be the P
n+1
name (and so Pname) for the Boolean algebra added
by forcing with P
n
.
Claim 1.5.2 For n , inv ind, t, hd, hL, s we have

P
ind
+
n+1
(

B
n
) =
+
and inv
+
(

B
n
) =
+
.
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 9
Proof of the claim: Repeat the proof of 1.4 with suitable changes to show
that in V
P
, for each n, we have
ind
+
n+1
(

B
n
) =
+
and t
+
1,2
n
+n
(

B
n
) = t
2
n
+n,1
(

B
n
) = ind
+
(

B
n
) =
+
.
Now note that for a Boolean algebra B
t
(+)
1,N
= hd
(+)
N
(B) and t
(+)
N,1
(B) = hL
(+)
N
(B)
(and remember that ind
(+)
(B) s
(+)
(B) hd
(+)
(B), hL
(+)
(B)).
The consequently part of the conclusion should be clear (or see [RoSh 534,
Section 1]).
Remark 1.6 Note that the examples when the spread of ultraproduct is
larger than the ultraproduct of the spreads which were known before pro-
vided a successor dierence only. Conclusion 1.5 shows that the jump
can be larger, but we do not know if one can get it in ZFC (i.e. assuming
suitable cardinal arithmetic only).
Problem 1.7 Can one improve 1.4 getting it for N = n + 1?
2 Irredundance of products
In theorem 2.1 below we answer [M2, Problem 24]. A parallel question for
free products of Boolean algebras will be addressed in the next section. It
should be noted here that the proof of the ZFC result was written as a result
of an analysis why a forcing proof of consistency of an inequality (similar to
the one from the next section) failed.
Theorem 2.1 For Boolean algebras B
0
, B
1
:
irr(B
0
B
1
) = maxirr(B
0
), irr(B
1
).
Proof Clearly irr(B
0
B
1
) maxirr(B
0
), irr(B
1
), so we have to deal
with the converse inequality only. Assume that a sequence x = (x
0

, x
1

) :
< ) B
0
B
1
is irredundant. Thus, for each < , we have homomor-
phisms f
0

, f
1

: B
0
B
1
0, 1 such that f
0

(x
0

, x
1

) = 0, f
1

(x
0

, x
1

) = 1
and
( )(f
0

(x
0

, x
1

) = f
1

(x
0

, x
1

)).
By shrinking the sequence x if necessary, we may assume that one of the
following occurs:
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 10
(i) ( < )(f
0

(1, 0) = f
1

(1, 0) = 0),
(ii) ( < )(f
0

(1, 0) = f
1

(1, 0) = 1),
(iii) ( < )(f
0

(1, 0) = 0 & f
1

(1, 0) = 1),
(iv) ( < )(f
0

(1, 0) = 1 & f
1

(1, 0) = 0).
If the rst clause occurs then we may dene (for < ) homomorphisms
h
0

, h
1

: B
1
0, 1 by h

(x) = f

(1, x) (remember that in this case we


have f

(0, 1) = 1). Clearly these homomorphisms witness that the sequence


x
1

: < ) B
1
is irredundant (and thus irr
+
(B
1
) > ). Similarly, if (ii)
holds then the sequence x
0

: < ) B
0
is irredundant and irr
+
(B
0
) > .
Since f

(1, 0) = 0 f

(0, 1) = 1 and the algebras B


0
, B
1
are in
symmetric positions, we may assume that clause (iv) holds, so f

(0, 1) =
(for < 2, < ).
For < and < 2 let g

: 2 be given by g

() = f

(x
0

, x
1

) for
< . Note that ,= implies g
0

() = g
1

() (remember the choice of the


f

s). Next, for < 2 let F

= g

: < and let B

be the algebra B
(,F

)
(see 1.1(1)).
Claim 2.1.1 Assume that A and < 2 are such that
(

A
) the mappings x

: A 0, 1 : x

g
k

() (for k = 0, 1 and
A) extend to homomorphisms from x

: A)
B

onto 0, 1.
Then the sequence x

: A) B

is irredundant.
Proof of the claim: First note that the assumption (

A
) implies that the
sequence x

: A) B

is irredundant. Now, the mapping x

extends to a homomorphism from the algebra x

: < )
B

onto B

. [Why?
Note that, since f
0

(1, 0) = 1 = f
1

(0, 1), the mappings x

(x
0

, x
1

) =
g

() extend to homomorphisms from B

onto 0, 1. Now look at the


denition of the algebra B

; remember 1.2(2).] Consequently we get that


the sequence x

: A) B

is irredundant.
It follows from claim 2.1.1 that if there are A []

and < 2 such that


(

A
) holds true then the algebra B

has an irredundant sequence of length


(i.e. irr
+
(B

) > ). So the proof of the theorem will be concluded when


we show the following claim.
Claim 2.1.2 Let < 2. Assume that there is no A []

such that (

A
)
holds. Then s
+
(B
1
) > (so irr
+
(B
1
) > too).
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 11
Proof of the claim: By induction on < we build a sequence (u

, v

) :
< ) such that for each < :
(a) u

, v

[]
<
are disjoint,
(b) (u

)

<
(u

) = ,
(c) B

[=
_
u


_
v

(x

) = 0,
(d) B

1
[=
_
u


_
v

(x

) ,= 0.
Suppose we have dened u

, v

for < . The set A =



<
(u

) is
of size , so (by our assumptions) (

A
) fails. This means that one of the
mappings
x

: A 0, 1 : x

g
k

(), (k = 0, 1, A)
does not extend to a homomorphism from x

: A)
B

. But, by the
denition of B

, the mappings x

() do extend (see 1.2(1)). So we


nd nite disjoint sets u

, v

A such that B

[=
_
u


_
v

(x

) = 0,
but for some < , g
1

1 and g
1

0. The latter implies that


B

1
[=
_
u


_
v

(x

) ,= 0. This nishes the construction.


The demand (d) means that (by 1.2) for each < we nd

< such
that g
1

1 and g
1

0. On the other hand, by (c), there is


no < such that g

1 and g

0. But now, if / u

then g
1

(u

) = g

(u

), so necessarily

. Let
y

=
_
u

x
1


_
v

(x
1

) B
1
and h

: x
1

: < )
B
1
0, 1
be a homomorphism dened by h

(x
1

) = f
1

(x
0

, x
1

) = g
1

(). It
follows from the above discussion that (h

is well dened and)


h

(y

) = 1 if and only if = ,
showing that the sequence y

: < ) is ideal independent (and irredun-


dant). This nishes the proof of the claim and that of the theorem.
3 Forcing for spread and irredundance
In this section we show that, consistently, there is a Boolean algebra B such
that irr(B) < s(B B). This gives a partial answer to [M2, Problem 27].
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 12
Moreover, it shows that a statement parallel to 2.1 for the free product
(instead of product) is not provable in ZFC. Note that before trying to
answer [M2, Problem 27] in ZFC one should rst construct a ZFC example
of a Boolean algebra B such that irr(B) < [B[ so far no such example is
known.
Denition 3.1 1. We dene a forcing notion Q

by:
a condition is a tuple p = u
p
, f
p
0,
, f
p
1,
, f
p
2,
: u
p
)) such that
(a) u
p

1
is nite,
(b) f
p
,
: u
p
2 0, 1 for < 3, u
p
,
(c) f
p
0,
(u
p
) 2 = f
p
1,
(u
p
) 2 = f
p
2,
(u
p
) 2 for
u
p
,
(d) f
p
0,
(, 0) = 1, f
p
0,
(, 1) = 0 (for u
p
),
(e) f
p
1,
(, 0) = 0, f
p
1,
(, 1) = 1 (for u
p
),
(f ) f
p
0,
(, 0) = 0 or f
p
1,
(, 1) = 0 (for distinct , u
p
),
(g) f
p
0,
(, 0) = 0 or f
p
2,
(, 1) = 0 (for , u
p
),
(h) f
p
1,
(, 1) = 0 or f
p
2,
(, 0) = 0 (for , u
p
),
(i) f
p
2,
(, 0) = 0 or f
p
2,
(, 1) = 0 (for , u
p
);
the order is dened by: p q if and only if u
p
u
q
, and f
q
,

(u
p
2) = f
p
,
for u
p
, < 3 and for each u
q
, < 3:
f
q
,
(u
p
2) f
p
k,
: u
p
, k < 3.
2. For a condition p Q

let B

p
be the algebra B
(w,F)
, where w = u
p
2
and F = f
p
,
: u
p
, < 3 (see 1.1(1)).
3. Let

B

,

f
,
(for < 3, <
1
) be Q

-names such that

Q


B

=
_
B

p
: p
Q
and

f
,
=
_
f
p
,
: p
Q
, u
p
.
Proposition 3.2 1. Q

is a ccc forcing notion.


2. If p, q Q

, p q then B
p
is a subalgebra of B
q
.
3. In V
Q

,

f
,
:
1
2 2 (for <
1
and < 3) and

B

is the Boolean
algebra B
(w,F)
, where w =
1
2 and F =

f
,
: <
1
, < 3.
Proof 1) Suppose that / Q

is uncountable. Applying system


arguments nd p, q / such that letting u

= u
p
u
q
we have:
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 13
(i) max(u

) < min(u
p
u

) max(u
p
u

) < min(u
q
u

),
(ii) [u
p
[ = [u
q
[ and if H : u
p
u
q
is the order isomorphism, u
p
and
< 3 then f
p
,
= f
q
,H()
(H id).
Now let u
r
= u
p
u
q
and for < 3 and u
r
let:
f
r
,
=
_

_
f
p
,
f
q
,
if u
p
u
q
,
f
p
,
f
q
2,H()
(u
q
u
p
) if u
p
u
q
,
f
q
,
f
p
2,H
1
()
(u
p
u
q
) if u
q
u
p
.
It is a routine to check that this denes a condition in Q

stronger than both


p and q.
2) Should be clear.
3) Note that if p Q

,
0
u
p
and
1
u
p
then letting u
q
= u
p

and
f
q
,
=
_

_
f
p
,
((, 0), 0), ((, 1), 0) if u
p
, < 3,
f
p
2,
0
((, 0), 1 ), ((, 1), ) if = , < 2,
f
p
2,
0
((, 0), 0), ((, 1), 0) if = , = 2,
we get a condition q Q

stronger than p and such that w


q
. Now, the
rest should be clear.
Proposition 3.3
Q
s(

) =
1
.
Proof To avoid confusion between the two copies of

B

in

B

, let us
denote an element ab

B

such that a is from the rst copy of



B

and
b is from the second one, by a, b). With this convention, for each <
1
let y

= x
,0
, x
,1
) and let

f

:

B

0, 1 be a homomorphism such
that (for <
1
, i < 2)

(x
,i
, 1)) =

f
0,
(, i) and

f

(1, x
,i
)) =

f
1,
(, i).
Note that, by 3.1(d,e), for each <
1

( y

) =

f
0,
(, 0)

f
1,
(, 1) = 1,
and if
1
then (by 3.1(f))

( y

) =

f
0,
(, 0)

f
1,
(, 1) = 0.
Hence we conclude that

Q
y

: <
1
) is idealindependent ,
nishing the proof.
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 14
Theorem 3.4
Q
irr
+
5
(

) =
0
.
Proof Let a

: <
1
) be a Q

name for an
1
sequence of elements
of

B

, p Q

. For <
1
choose a condition p

p, a Boolean term

,
ordinals (, 0) . . . (,

) <
1
and (, 0), . . . , (,

) 0, 1 such
that
p


Q
a

(x
(,0),(,0)
, . . . , x
(,

),(,

)
).
Applying standard cleaning procedure we may assume that for ,
0
,
1
<

1
:
(i)

= ,

= ,
(ii) ( (, j), (, j)) : j = u
p

2 is an enumeration which does not


depend on if we treat it modulo otp (so 2 [u
p

[ = +1 and we may
write (x
,i
: u
p

, i < 2)),
(iii) u
p

: <
1
forms a system of sets with the heart u

, and if

0
<
1
<
1
then
max(u

) < min(u
p

0
u

) max(u
p

0
u

) < min(u
p

1
u

),
(iv) [u
p

0
[ = [u
p

1
[ and if H

0
,
1
: u
p

0
u
p

1
is the order preserving
mapping then f
p

0
k,
= f
p

1
k,H()
(H

0
,
1
id) (for u
p

0
, k < 3).
Now we are going to dene a condition q stronger than p
0
, . . . , p
5
. We put
u
q
=

i<6
u
p
i
and we dene functions f
q
,
: u
q
2 2 (for u
q
and
< 3) as follows.
() If u

, < 3 then f
q
,
=

i<6
f
p
i
,
.
(
0
) If u
p
0
u

then
f
q
0,
= f
p
0
0,
f
p
1
0,H
0,1
()
f
p
2
0,H
0,2
()
f
p
3
2,H
0,3
()
f
p
4
2,H
0,4
()
f
p
5
2,H
0,5
()
,
f
q
1,
= f
p
0
1,
f
p
1
2,H
0,1
()
f
p
2
2,H
0,2
()
f
p
3
1,H
0,3
()
f
p
4
1,H
0,4
()
f
p
5
2,H
0,5
()
,
f
q
2,
=

i<6
f
p
i
2,H
0,i
()
.
(
i
) If u
p
i
u

, 0 < i < 6 and < 3 then f


q
,
= f
p
i
,


j=i
f
p
j
2,H
i,j
()
.
It follows from (iv) and 3.1(c) that the functions f
q
,
are well dened.
Claim 3.4.1 The tuple q = u
q
, f
q
,
: < 3, u
q
)) is a condition in Q

stronger than p
0
, . . . , p
5
.
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 15
Proof of the claim: To show that q Q

one has to check the demands


(a)(i) of 3.1. The only possible problems could be caused by clauses (f)(i).
If functions f
q
,
were dened in clauses (), (
i
) then easily these demands
are met. To deal with instances of (
0
) (i.e. when u
p
0
u

) note that
in the denition of f
q
,
( < 2, u
p
0
u

) a part of the form f


p
j
,H
i,j
()
meets f
p
j
2,H
i,j
()
on the side of f
q
1,
. Therefore, by (g), (h) of 3.1, we have
no problems with checking demand (f). Clause 3.1(i) is immediate and (g),
(h) should be clear too.
Claim 3.4.2
(x
,i
: u
p
0
, i < 2) (x
,i
: u
p
j
, i < 2) : 0 < j < 6)
B

q
.
Consequently q
Q
a
0
a
j
: 0 < j < 6)

B
.
Proof of the claim: Suppose that
(x
,i
: u
p
0
, i < 2) / (x
,i
: u
p
j
, i < 2) : 0 < j < 6)
B

q
.
Then we nd two homomorphisms h
0
, h
1
: B

q
0, 1 such that
h
0
((x
,i
: u
p
0
, i < 2)) ,= h
1
((x
,i
: u
p
0
, i < 2)) but
h
0
((x
,i
: u
p
j
, i < 2)) = h
1
((x
,i
: u
p
j
, i < 2)) for 0 < j < 6.
By the denition of the algebra B

q
each its homomorphism into 0, 1 is
generated by one of the functions f
q
,
(for < 3, u
q
). So we nd

0
,
1
< 3 and
0
,
1
u
q
such that h
k
f
q

k
,
k
. Now we have to consider
several cases corresponding to the way the f
q

k
,
k
were dened.
Case A:
k
u

,
1k
u
p
i
, i < 6.
Then look at the denition () of f
q

k
,
k
it copies f
p
0

k
,
k
everywhere (re-
member (iv)). On the other hand, whatever clause was used to dene
f
q

1k
,
1k
, there is j (0, 6) such that f
q

1k
,
1k
(u
p
j
2) is a copy of
f
q

1k
,
1k
(u
p
0
2). Hence we may conclude that (for this j)
h
1k
((x
,i
: u
p
j
, i < 2)) ,= h
k
((x
,i
: u
p
j
, i < 2)),
a contradiction.
Case B:
k
u
p
0
u

,
1k
u
p
i
u

, 0 < i < 6.
Then we repeat the argument of the previous Case, choosing j in such a
way that j ,= i and: if
k
= 0 then j 1, 2, if
k
= 1 then j 3, 4.
Case C:
k
u
p
i

,
1k
u
p
i

, 0 < i

, i

< 6.
Like above, but now take j 1, . . . , 5 i

, i

.
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 16
Case D:
0
,
1
u
p
0
u

.
This is the most complicated case. We may repeat the previous argument
in some cases letting:
j =
_
1 if (
0
,
1
) (0, 0), (0, 2), (2, 0), (2, 2)
3 if (
0
,
1
) (1, 1), (1, 2), (2, 1)
This leaves us with two symmetrical cases: (
0
,
1
) = (0, 1) or (
0
,
1
) =
(1, 0). So suppose that
0
= 0,
1
= 1 and let
x
def
= h
0
((x
,i
: u
p
5
, i < 2)) = h
1
((x
,i
: u
p
5
, i < 2)).
Since f
q
0,
0
(u
p
4
2) is a copy of f
q
0,
0
(u
p
5
2) we conclude that
x = h
0
((x
,i
: u
p
4
, i < 2)) = h
1
((x
,i
: u
p
4
, i < 2)),
and, since f
q
1,
1
(u
p
4
2) is a copy of f
q
1,
1
(u
p
0
2) we get
() x = h
1
((x
,i
: u
p
0
, i < 2)).
Next, f
q
1,
1
(u
p
2
2) is a copy of f
q
1,
1
(u
p
5
2) and therefore
x = h
1
((x
,i
: u
p
2
, i < 2)) = h
0
((x
,i
: u
p
2
, i < 2)).
But f
q
0,
0
(u
p
2
2) is a copy of f
q
0,
0
(u
p
0
2), so we conclude that
() x = h
0
((x
,i
: u
p
0
, i < 2)).
But now () +() contradict the choice of h
0
, h
1
. The other case is similar.
This nishes the proof of the claim and of the theorem.
Conclusion 3.5 It is consistent that there exists a Boolean algebra B such
that

0
= irr(B) and s(B B) = irr(B B) =
1
.
Remark 3.6 We may use any cardinal =
<
instead of and
+
instead
of
1
in 3.1 and then 3.2, 3.3. But we do not know if the dierence between
the respective cardinal invariants can be larger.
Problem 3.7 Is it consistent that there is a Boolean algebra B such that
(irr(B))
+
< [B[ ? (irr(B))
+
< s(B B) ?
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 17
4 Forcing a superatomic Boolean algebra
In this section we give partial answers to [M2, Problems 73, 77, 78] showing
that, consistently, there is a superatomic Boolean algebra B such that s(B) =
inc(B) < irr(B) = Id(B) < Sub(B). The forcing notion we use is a variant of
the one of Martinez [Ma92], which in turn was a modication of the forcing
notion used in Baumgartner Shelah [BaSh 254]. For more information on
superatomic Boolean algebras we refer the reader to Koppelberg [Ko89],
Roitman [Rt89] and Monk [M2].
Denition 4.1 Let be a cardinal. For a pair s = (, )
+
we will
write (s) = and (s) = . We dene a forcing notion P

as follows:
a condition is a tuple
p =
_
w
p
, u
p
, a
p
, f
p
s
: s u
p
), y
p
s
0
,s
1
: s
0
, s
1
u
p
, s
0
,= s
1
, (s
0
) (s
1
))
_
such that
(a) a
p
w
p
[
+
]
<
, u
p
[w
p
]
<
, and w
p
(, 0), (, 1) u
p
,
(b) for s u
p
, f
p
s
: u
p
0, 1 is such that f
p
s
(s) = 1 and
(t u
p
)((t) (s) & t ,= s f
p
s
(t) = 0),
(c) if < , , a
p
then f
p
,0
(, 0) = f
p
,1
(, 0),
(d) if s
0
, s
1
u
p
are distinct, (s
0
) (s
1
) then
y
p
s
0
,s
1
[u
p
((s
0
) )]
<
and for every t u
p
f
p
t
(s
0
) = 1 & f
p
t
(s
1
) ,= f
p
s
0
(s
1
) (s y
p
s
0
,s
1
)(f
p
t
(s) = 1);
the order is given by p q if and only if w
p
w
q
, u
p
u
q
, a
p
= a
q
w
p
,
y
q
s
0
,s
1
= y
p
s
0
,s
1
(for distinct s
0
, s
1
u
p
such that (s
0
) (s
1
)), f
p
s
f
q
s
(for s u
p
) and
(s u
q
)(t u
p
)(f
q
s
u
p
= f
p
t
or f
q
s
u
p
= 0
u
p).
Denition 4.2 We say that conditions p, q P

are isomorphic if there is


a bijection H : u
p
u
q
(called the isomorphism from p to q) such that
(s u
p
)(otp((s) w
p
) = otp((H(s)) w
q
) & (s) = (H(s))),
( w
p
)((H(, 0)) a
q
a
p
),
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 18
(s u
p
)(f
p
s
= f
q
H(s)
H),
(s
0
, s
1
u
p
)((s
0
) (s
1
) y
s
0
,s
1
= s u
p
: H(s) y
q
H(s
0
),H(s
1
)
).
Proposition 4.3 Assume
<
= . Then P

is a complete
+
cc forcing
notion of size
+
.
Proof It should be clear that P

is complete and [P

[ =
+
. Moreover,
there is many isomorphism types of conditions in P

(and a condition in
P

is determined by its isomorphism type and the set w


p
). Now, to show the
chain condition assume that / P

is of size
+
. Applying lemma choose
pairwise isomorphic conditions p
0
, p
1
, p
2
/such that w
p
0
, w
p
1
, w
p
2
forms
a system with heart w

and such that for i < j < 3


sup(w

) < min(w
p
i
w

) sup(w
p
i
) < min(w
p
j
w

)
(remember
<
= ). For i, j < 3 let H
i,j
: u
p
i
u
p
j
be the isomorphism
from p
i
to p
j
. We are going to dene a condition q P

which will be an
upper bound to p
1
, p
2
(note: not p
0
!). To this end we rst let
w
q
= w
p
0
w
p
1
w
p
2
, u
q
= u
p
0
u
p
1
u
p
2
, a
q
= a
p
1
a
p
2
.
To dene functions f
q
s
we use the approach which can be described as put
zero whenever possible. Thus we let:
if s u
p
1
u
p
0
then f
q
s
= 0
u
p
0 f
p
1
s
0
u
p
2 ,
if s u
p
2
u
p
0
then f
q
s
= 0
u
p
0
0
u
p
1
f
p
2
s
,
if s u
p
0
then f
q
s
= f
p
0
s
f
p
1
H
0,1
(s)
f
p
2
H
0,2
(s)
.
It should be clear that the functions f
q
s
are well dened. Now we are going
to dene the sets y
q
s
0
,s
1
for distinct s
0
, s
1
u
q
such that (s
0
) (s
1
). It
is done by cases considering all possible congurations. Thus we put:
if s
0
, s
1
u
p
i
, i < 3 then y
q
s
0
,s
1
= y
p
i
s
0
,s
1
,
if s
0
u
p
1
u
p
0
, s
1
u
p
2
u
p
0
then y
q
s
0
,s
1
= H
2,0
(s
1
),
if s
0
u
p
0
, s
1
u
p
i
, i 1, 2 then
y
q
s
0
,s
1
=
_

_
if H
i,0
(s
1
) = s
0
,
H
i,0
(s
1
) if (H
i,0
(s
1
)) < (s
0
),
y
p
0
s
0
,H
i,0
(s
1
)
if (s
0
) (H
i,0
(s
1
)), s
0
,= H
i,0
(s
1
).
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 19
We claim that
q =
_
w
q
, u
q
, a
q
, f
q
s
: s u
q
), y
q
s
0
,s
1
: s
0
, s
1
u
q
, s
0
,= s
1
, (s
0
) (s
1
))
_
is a condition in P

and for this we have to check the demands of 4.1. Clauses


(a) and (b) should be obvious. To check 4.1(c) note that a
q
w
p
0
= a
q
w

and therefore there are no problems when a


q
w
p
0
. If a
q
(w
p
1
w
p
0
)
and < a
q
(w
p
2
w
p
0
) then f
q
,0
(, 0) = f
q
,1
(, 0) = 0. In all other
instances we use the clause (c) of 4.1 for p
1
, p
2
.
Now we have to verify the demand 4.1(d). Suppose that s
0
, s
1
are distinct
members of u
q
and (s
0
) (s
1
). If s
0
, s
1
u
p
i
for some i < 3 then easily
the set y
q
s
0
,s
1
has the required property. So suppose now that s
0
u
p
1
u
p
0
,
s
1
u
p
2
u
p
0
(so then f
q
s
0
(s
1
) = 0) and let t u
q
be such that f
q
t
(s
1
) =
1 = f
q
t
(s
0
). Then necessarily t u
p
0
and f
q
t
(H
2,0
(s
1
)) = f
q
t
(s
1
) = 1, so we
are done in this case. Finally, let us assume that s
0
u
p
0
and s
1
u
p
i
,
0 < i < 3. Note that if f
q
t
(s
0
) = 1 then t u
p
0
. Now, if H
i,0
(s
1
) = s
0
then
f
q
t
(s
0
) = f
q
t
(s
1
) for every t u
p
0
and there are no problems (i.e. no f
q
t
has to
be taken care of). If (H
i,0
(s
1
)) < (s
0
) then the set y
q
s
0
,s
1
= H
i,0
(s
1
) will
work as for every t u
p
0
we have f
q
t
(H
i,0
(s
1
)) = f
q
t
(s
1
) (and f
q
s
0
(s
1
) = 0).
For the same reason the set y
q
s
0
,s
1
has the required property in the remaining
case too.
Checking that the condition q is stronger than both p
1
and p
2
is straight-
forward (note: we do not claim that q is stronger than p
0
).
Lemma 4.4 If p P

, t
+
then there is q P

such that p q and


t u
q
.
Proof Suppose t = (, ) (
+
) u
p
. Put w
q
= w
p
, a
q
= a
p
and u
q
= u
p
(, 0), (, 1), (, ). For s u
p
let f
q
s
= f
p
s
0
u
q
\u
p and for
s u
q
u
p
let f
q
s
be such that f
q
s
(s) = 1 and f
q
s
u
q
s 0. Finally, for
distinct s
0
, s
1
u
q
such that (s
0
) (s
1
) let
y
q
s
0
,s
1
=
_
y
p
s
0
,s
1
if s
0
, s
1
u
p
,
otherwise.
Check that q = w
q
, u
q
, a
q
, f
q
s
: s u
q
), y
q
s
0
,s
1
: s
0
, s
1
u
q
)) P

is as
required.
For p P

let B
p
be the algebra B
(u
p
,F
p
)
(see 1.1(1)), where F
p
= f
p
s
:
s u
p
0
u
p, and let

B

be a P

name such that

P


B

=
_
B
p
: p
P
.
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 20
Furthermore, for s
+
let

f
s
be a P

name such that

P


f
s
=
_
f
p
s
: p
P
& s u
p
.
Proposition 4.5 Assume
<
= . Then in V
P
:
1.

B

is the algebra B
(W,

F)
, where W =
+
and

F =

f
s
: s

+
0

,
2. the algebra

B

is superatomic,
3. if s
+
and b

B

then there are nite v


0
v
1
(s) such
that either x
s
b or x
s
(b) equals to

x
t

v
1
(t

)<(t)
(x
t
) : t v
0
,
4. the height of

B

is
+
and x
,
: are representatives of atoms
of rank + 1,
5. irr(

) =
+
.
Proof 1) First note that if p q then B
p
is a subalgebra of B
q
. Next,
remembering 4.4, conclude that

P


B

is a Boolean algebra generated by x


s
: s
+
) .
Clearly, by 4.4,

f
s
:
+
0, 1 and p

f
s
u
p
= f
p
s

(for s u
p
, p P

). So it should be clear that


P

= B
(W,

F)
, where
W =
+
and

F =

f
s
: s
+
0

.
2) It follows from 4.1(b) that for each s
+

P


f
s
(s) = 1 and (t
+
)((t) (s) & t ,= s

f
s
(t) = 0) .
Now work in V
P
. Let

J

be the ideal in

B

generated by x
,
: < ,
(for
+
; if = 0 then

J

= 0). It follows from the previous


remark that x
,
/

J

(for all ; remember 1.2).


Suppose that s
0
, s
1
are distinct, (s
0
) = (s
1
) = <
+
and suppose
that t
+
is such that

f
t
(s
0
) =

f
t
(s
1
) = 1. Let p
P
be such
that t, s
0
, s
1
u
p
. It follows from 4.1(d) that there is s y
p
s
0
,s
1
such that
f
p
t
(s) = 1. Hence (applying 1.2) we may conclude that

[= x
s
0
x
s
1

x
s
: s y
p
s
0
,s
1
,
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 21
and therefore x
s
0
x
s
1


J

.
Now suppose that s
0
, s
1

+
are such that (s
0
) < (s
1
) and let p

P
be such that s
0
, s
1
u
p
. If f
p
s
0
(s
1
) = 0 then, by similar considerations as
above, we have x
s
0
x
s
1


J

. Similarly, if f
p
s
0
(s
1
) = 1 then x
s
0
(x
s
1
)

. Hence we conclude that x


s
0
/

J

is an atom in

B

/

J

.
Finally, note that the ideal

J

+ is maximal (as x
s
: s
+
are
generators of the algebra

B

) and hence the algebra



B is superatomic.
3) For
+
let

J

be the ideal of

B

dened as above. Note that if


a

J

0 then there is a nite set v such that


a

tv
x
t
and (t v)(x
t
a /

J
(t)
).
A set v with these properties will be called a good cover for a.
We know already that x
s
/

J
(s)
is an atom in

B

/

J
(s)
and therefore
either x
s
b

J
(s)
or x
s
(b)

J
(s)
. We may assume that the rst takes
place. Applying repeatedly the previous remark nd a nite set v
1
(s)
such that for every t v
1
s:
if x
t
(x
s
b)

J
(t)
0 then there is a good (t)cover v v
1
for
x
t
(x
s
b),
if x
t
(x
s
b)

J
(t)
0 then there is a good (t)cover v v
1
for
x
t
(x
s
b).
Now let v
0
= t v
1
: x
t
(x
s
b) /

J
(t)
and check that
x
s
b =

x
t

v
1
(t

)<(t)
(x
t
) : t v
0
,
as required.
4) Almost everything what we need for this conclusion was done in clause 2)
above except that we have to check that, for each <
+
, x
,
/

J

: <
lists all atoms of the algebra

B

/

J

. So suppose that b/

J

is an atom in

/

J

. We may assume that b =


_
tw
x
t

_
tu
(x
t
) and that (t) > for
t w (otherwise either b

J

or b/

J

= x
s
/

J

for some s with (s) = ).


Suppose that w = . Let p P

. We may nd a condition q p such


that u u
q
and then take t ( ) u
q
. Exactly as in the proof of
4.4 we dene a condition r P

stronger than q and such that t u


r
. Note
that for this condition we have r x
t
b and we easily nish.
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 22
Let s w (so (s) > ) and b

=
_
tw\{s}
x
t

_
tu
(x
t
) (so b = b

x
s
).
It follows from the third clause that we nd nite sets v
0
v
1
(s)
such that
c
def
=

x
t

v
1
(t

)<(t)
(x
t
) : t v
0
x
s
b

, x
s
(b

).
If c = x
s
(b

) then we repeat arguments similar to those from the previous


paragraph but with a modied version of 4.4: dening the condition r with
the property that t u
r
, we use the function f
q
s
(t, 1) as f
r
t
(check that no
changes are needed in the denition of y
r
s
0
,s
1
). Then easily r x
t
x
s
(c).
Finally, if c = x
s
b

then we take s

v
0
such that (s

) is maximal possible.
If (s

) > then similarly as in the previous case we nd a condition r which


forces that x
t
x
s
b

= b, if (s

) it is even easier. In all cases we


easily nish nding an element x
,
which is

J

smaller than b.
5) Look at the demand 4.1(c): it means that if , a
def
=

a
p
: p
P

are distinct then



f
,0
(, 0) =

f
,1
(, 0). As

f
,0
(, 0) = 1,

f
,1
(, 0) = 0
we conclude that

f
,0
,

f
,1
determine homomorphisms from

B

to 0, 1
witnessing x
,0
/ x
,0
: a )

B
. Since clearly [ a[ =
+
the proof
is nished.
Proposition 4.6 Assume
<
= . Then

P
inc(

) = s(

) = .
Proof Suppose that

: <
+
) is a P

name for a
+
sequence of
elements of

B

and
p
P

: <
+
) are pairwise incomparable .
Applying lemma and standard cleaning choose pairwise isomorphic
conditions p
0
, p
1
, p
2
stronger than p, sets v
1
, v
2
, a Boolean term and
1
<

2
<
+
such that
w
p
0
, w
p
1
, w
p
2
forms a system with heart w

,
sup(w

) < min(w
p
i
w

) sup(w
p
i
) < min(w
p
j
w

) for i < j < 3,


v
i
[u
p
i
]
<
for i = 1, 2,
if H
i,j
is the isomorphism from p
i
to p
j
then v
2
= H
2,1
[v
1
],
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 23
p
i


b

i
= (x
s
: s v
i
) for i = 1, 2.
Considering two cases, we are going to dene a condition r stronger than
p
1
, p
2
. The condition r will be dened in a similar manner as the condition
q in the proof of 4.3.
Case A: 0, 1 [= (0 : t v
1
) = 0.
First choose s

u
p
2
u
p
0
such that
if there is s u
p
2
u
p
0
with the property that
0, 1 [= (f
p
2
s
(t) : t v
2
) = 1
then s

is like that.
Now we proceed as in 4.3 using f
p
2
s
instead of 0
u
p
2
. So we let
w
r
= w
p
0
w
p
1
w
p
2
, u
r
= u
p
0
u
p
1
u
p
2
, a
r
= a
p
1
a
p
2
,
and we dene functions f
r
s
as follows:
if s u
p
0
then f
r
s
= f
p
0
s
f
p
1
H
0,1
(s)
f
p
2
H
0,2
(s)
,
if s u
p
1
u
p
0
then f
r
s
= 0
u
p
0 f
p
1
s
f
p
2
s
,
if s u
p
2
u
p
0
then f
r
s
= 0
u
p
0 0
u
p
1 f
p
2
s
(check that the functions f
r
s
are well dened). Next, for distinct s
0
, s
1
u
r
such that (s
0
) (s
1
), we dene the sets y
r
s
0
,s
1
:
if s
0
, s
1
u
p
i
, i < 3 then y
r
s
0
,s
1
= y
p
i
s
0
,s
1
,
if s
0
u
p
1
u
p
0
, s
1
u
p
2
u
p
0
then y
r
s
0
,s
1
= H
1,0
(s
0
),
if s
0
u
p
0
, s
1
u
p
i
, i 1, 2 then
y
r
s
0
,s
1
=
_

_
if H
i,0
(s
1
) = s
0
,
H
i,0
(s
1
) if (H
i,0
(s
1
)) < (s
0
),
y
p
0
s
0
,H
i,0
(s
1
)
if (s
0
) (H
i,0
(s
1
)), s
0
,= H
i,0
(s
1
).
Exactly as in 4.3 one checks that
r =
_
w
r
, u
r
, a
r
, f
r
s
: s u
r
), y
r
s
0
,s
1
: s
0
, s
1
u
r
, s
0
,= s
1
, (s
0
) (s
1
))
_
is a condition in P

stronger than both p


1
and p
2
. Moreover, it follows from
the denition of f
r
s
s that
B
r
[= (x
t
: t v
1
) (x
t
: t v
2
)
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 24
(see 1.2). Consequently r

b

1


b

2
, a contradiction.
Case B: 0, 1 [= (0 : t v
1
) = 1.
Dene r almost exactly like in Case A, except that when choosing s

u
p
2
u
p
0
ask if there is s u
p
2
u
p
0
such that
0, 1 [= (f
p
2
s
(t) : t v
2
) = 0
(and if so then s

has this property). Continue like before getting a condition


r stronger than p
1
, p
2
and such that
B
r
[= (x
t
: t v
1
) (x
t
: t v
2
)
and therefore r

b

1


b

2
, a contradiction nishing the proof.
Theorem 4.7 Assume
<
= . Then

P
Id(

B) = 2

= (2

)
V
.
Proof Let / be the a family of all pairs (p, ) such that p P

and
= (x
s
: s v) is a Boolean term, v u
p
. For each ordinal <
+
we
dene a relation E

on / as follows:
(p
0
,
0
) E

(p
1
,
1
) if and only if
(i) the conditions p
0
, p
1
are isomorphic,
(ii) w
p
0
= w
p
1
,
(iii) if H : u
p
0
u
p
1
is the isomorphism from p
0
to p
1
then
1
= H(
0
)
(i.e.
0
= (x
s
: s v),
1
= (x
H(s)
: s v)).
A relation E

on / is dened by
(p
0
,
0
) E

(p
1
,
1
) if and only if (p
0
,
0
) E

(p
1
,
1
) and
(iv) if w
p
0
then
sup(w
p
0
) = H() sup(w
p
1
H()) mod and
sup(w
p
0
) + if and only if H() sup(w
p
1
H()) +.
Claim 4.7.1 For each <
+
, E

, E

are equivalence relations on / with


many equivalence classes.
Claim 4.7.2 Suppose that <
+
, (p
0
,
0
) E

(p
1
,
1
) and p
0
q
0
. Then
there is q
1
P

such that p
1
q
1
and (q
0
,
0
) E

(q
1
,
1
).
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 25
Claim 4.7.3 Suppose that

I is a P

name for an ideal in the algebra



B

and let /(

I) = (p, ) / : p

I. Then there is = (

I) <
+
such
that
/(

I) =
_
(p, )/E

: (p, ) /(

I).
Proof of the claim: Assume not. Then for each <
+
we nd (p

0
,

0
)
/(

I) and (p

1
,

1
) / /(

I) such that (p

0
,

0
) E

(p

1
,

1
). Take q

1
p

1
such
that q

1
/

I and use 4.7.2 to nd q

0
p

0
such that (q

0
,

0
) E

(q

1
,

1
).
Now use lemma and clause (i) of the denition of E

to nd
0
<
1
<

2
<
3
<
+
such that letting q
2
= q

2
1
,
2
=

2
1
and q
i
= q

i
0
,
i
=

i
0
for i ,= 2 we have
the conditions q
0
, . . . , q
3
are pairwise isomorphic (and for i, j < 4 let
H
i,j
: u
q
i
u
q
j
be the isomorphism from q
i
to q
j
),
w
q
0
, w
q
1
, w
q
2
, w
q
3
forms a system with heart w

,
sup(w

) < min(w
q
i
w

) sup(w
q
i
w

) < min(w
q
j
w

) when
i < j < 4,

i
= H
i,j
(
j
) (i.e. we have the same term).
Now we dene a condition q P

in a similar manner as in 4.3, 4.6. First


we x s

u
q
3
u
q
0
such that
if there is s u
q
3
u
q
0
with the property that f
q
3
s
(
3
) = 1
then s

is like that.
We put
w
q
= w
q
0
. . . w
q
3
, u
q
= u
q
0
. . . u
q
3
, a
q
= a
q
1
a
q
2
a
q
3
,
and we dene f
q
s
as follows:
f
q
s
=
_

i<4
f
q
i
H
0,i
(s)
if s u
q
0
,
0
u
q
0 f
q
1
s
f
q
2
H
3,2
(s

)
f
q
3
s
if s u
q
1
u
q
0
,
0
u
q
0 0
u
q
1 f
q
2
s
f
q
3
s
if s u
q
2
u
q
0
,
0
u
q
0 0
u
q
1 0
u
q
2 f
q
3
s
if s u
q
3
u
q
0
.
Finally, for distinct s
0
, s
1
u
q
such that (s
0
) (s
1
), we dene
y
q
s
0
,s
1
=
_

_
y
q
i
s
0
,s
1
if s
0
, s
1
u
q
i
, i < 4,
H
i,0
(s
0
) if s
0
u
q
i
u
q
0
, s
1
u
q
j
u
q
0
, 0 < i < j < 4,
if s
0
u
q
0
, s
1
u
q
i
, 0 < i < 4, H
i,0
(s
1
) = s
0
,
H
i,0
(s
1
) if s
0
u
q
0
, s
1
u
q
i
, 0 < i < 4, (H
i,0
(s
1
)) < (s
0
),
y
q
0
s
0
,H
i,0
(s
1
)
otherwise.
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 26
It should be a routine to check that this denes a condition q P

stronger
than q
1
, q
2
, q
3
and that (by 1.2) B
q
[=
2

1

3
(remember that the terms
are isomorphic). But this means that
q
P

2
1

1
0

3
0
&

2
1
/

I &

1
0
,

3
0


I ,
a contradiction nishing the proof of the claim.
Now, using 4.7.3, we may easily nish: if

I
0
,

I
1
are P

names for ideals


in

B

such that /(

I
0
) = /(

I
1
) then

I
0
=

I
1
. But 4.7.3 says that /(

I) is
determined by (

I) and a family of equivalence classes of E
(

I)
. So we have
at most
+
2

= 2

possibilities for /(

I). Finally note that [P

[ =
+
and
P

satises the
+
cc, so
P
2

= (2

)
V
.
Conclusion 4.8 It is consistent that there is a superatomic Boolean algebra
B such that
s(B) = inc(B) = , irr(B) = Id(B) =
+
and Sub(B) = 2

+
.
5 Modications of P

In this section we modify the forcing notion P

of 4.1 and we get two new


models. The rst model shows the consistency of there is a superatomic
Boolean algebra B such that irr(B) < inc(B) answering [M2, Problem 79].
Next we solve [M2, Problem 81] showing that possibly there is a superatomic
Boolean algebra B with Aut(B) < t(B).
Denition 5.1 Let be a cardinal. A forcing notion P
0

is dened like P

of 4.1 but the demand 4.1(c) is replaced by:


(c
0
) if < , , a
p
then (s u
p
)(f
p
s
(, 0) = 1 & f
p
s
(, 0) = 0).
Naturally we have a variant of denition 4.2 of isomorphic conditions for
the forcing notion P
0

(with no changes) and similarly as for the case of P

we dene algebras B
p
(for p P
0

) and P
0

names

B
0

,

f
0
s
(for s
+
).
Proposition 5.2 Assume
<
= . Then P

is a complete
+
cc forcing
notion of size
+
.
Proof Repeat the proof of 4.3 (with no changes).
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 27
Proposition 5.3 Assume
<
= . Then in V
P
:
1.

B
0

is the algebra B
(W,

F)
, where W =
+
and

F =

f
0
s
: s

+
0

,
2. the algebra

B
0

is superatomic (of height


+
) and x
,
: are
representatives of atoms of rank + 1,
3. inc(

B
0

) =
+
.
Proof The proofs of the rst two clauses are repetitions of that of 4.5(1
4) (so we have the respective version of 4.5(3) too).
To show the third clause let a
def
=

a
p
: p
P
0

. It should be clear
that [ a[ =
+
. Note that if , a
p
, < then, by 5.1(c
0
), B
p
[= x
,0
,
x
,0
and by the respective variant of 4.1(b) we have B
p
[= x
,0
, x
,0
.
Consequently the sequence x
,0
: a) witnesses inc(

B
0

) =
+
.
Proposition 5.4 Assume
<
= . Then
P
0

irr
+
3
(

B
0

) =
+
.
Proof Let

: <
+
) be a P
0

name for a
+
sequence of elements of

B
0

and let p P
0

. Find pairwise isomorphic conditions p


i
, sets v
i
, ordinals

i
(for i < 7) and a Boolean term such that
p p
0
, . . . , p
7
,
0
<
1
< . . . <
6
<
+
, v
i
[u
p
i
]
<
for i < 7,
w
p
0
, . . . , w
p
6
forms a system with heart w

,
sup(w

) < min(w
p
i
w

) sup(w
p
i
) < min(w
p
j
w

) for i < j < 7,


if H
i,j
is the isomorphism from p
i
to p
j
then v
j
= H
i,j
[v
i
] (for i, j < 7),
p
i


b

i
= (x
s
: s v
i
) for i < 7.
Now we are going to dene an upper bound q to the conditions p
3
, . . . , p
6
.
For this we let
w
q
=
_
i<7
w
p
i
, u
q
=
_
i<7
u
p
i
, a
q
=
_
2<i<7
a
p
i
and for s w
q
we dene
f
q
s
=
_

j<7
f
p
j
H
0,j
(s)
if s u
p
0
,
0
u
p
0u
p
2u
p
4 f
p
1
s
f
p
3
H
1,3
(s)
f
p
5
H
1,5
(s)
f
p
6
H
1,6
(s)
if s u
p
1
u
p
0
,
0
u
p
0u
p
1u
p
5 f
p
2
s
f
p
3
H
2,3
(s)
f
p
4
H
2,4
(s)
f
p
6
H
2,6
(s)
if s u
p
2
u
p
0
,
0
u
p
0u
p
1u
p
2u
p
6 f
p
3
s
f
p
4
H
3,4
(s)
f
p
5
H
3,5
(s)
if s u
p
3
u
p
0
,
0
u
q
\u
p
i
f
p
i
s
if s u
p
i
u
p
0
, 3 < i.
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 28
Next, for distinct s
0
, s
1
u
q
such that (s
0
) (s
1
), we dene y
q
s
0
,s
1
considering all possible congurations separately. Thus we put:
if s
0
, s
1
u
p
i
, i < 7 then y
q
s
0
,s
1
= y
p
i
s
0
,s
1
,
if s
0
u
p
0
u
p
1
, s
1
u
p
i
u
p
0
, 0 < i < 7 then
y
q
s
0
,s
1
=
_

_
if H
i,0
(s
1
) = s
0
,
H
i,0
(s
1
) if (H
i,0
(s
1
)) < (s
0
),
y
p
0
s
0
,H
i,0
(s
1
)
otherwise,
if s
0
u
p
i
u
p
0
, s
1
u
p
j
u
p
0
, 0 < i < j < 7 then
y
q
s
0
,s
1
=
_

_
H
i,k
(s
0
) : k < i if H
j,i
(s
1
) = s
0
,
H
i,k
(s
0
) : k < i H
j,i
(s
1
) if (H
j,i
(s
1
)) < (s
0
),
H
i,k
(s
0
) : k < i y
p
i
s
0
,H
j,i
(s
1
)
otherwise.
It is not dicult to check that the above formulas dene a condition q P
0

stronger than p
3
, p
4
, p
5
, p
6
(just check all possible cases). Moreover, applying
1.2, one sees that
B
q
[= (x
s
: s v
3
) = ((x
s
: s v
4
) (x
s
: s v
5
))
((x
s
: s v
4
) (x
s
: s v
6
))
((x
s
: s v
5
) (x
s
: s v
6
)).
Hence
q
P
0

4
,

b

5
,

b

6
)

B
0

,
nishing the proof.
Conclusion 5.5 It is consistent that there is a superatomic Boolean algebra
B such that inc(B) =
+
and irr(B) = .
For the next model we need a more serious modication of P

involving
a change in the denition of the order.
Denition 5.6 For an uncountable cardinal we dene a forcing notion
P
1

like P

of 4.1 except that the clause 4.1(c) is replaced by


(c
1
) if < , , a
p
then f
p
,0
(, 0) = 1
and we add the following requirement
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 29
(e) if (1, ) u
p
then the set < : f
p
0,
(1, ) = 1 is innite.
Moreover, we change the denition of the order demanding additionally that,
if p q,
() if (1, ) u
p
, (0, ) u
q
u
p
then f
q
0,
(1, ) = 0, and
() if (1, ) u
q
u
p
then the set (0, ) u
p
: f
p
0,
(1, ) = 1 is nite.
Like before we have the respective variants of 4.34.5 for P
1

which we for-
mulate below. The P
1

names

B
1

and

f
1
s
are dened like

B

and

f
s
.
Proposition 5.7 Assume
0
< =
<
. Then P
1

is a complete
+
cc
forcing notion.
Proof Repeat the arguments of 4.3 with the following small adjust-
ments. First note that we may assume [w

[ > 2. Next, if a
p
2
w

,=
then we let = min(a
p
2
w

) and dening f
q
s
for s u
p
1
u
p
0
we put
f
q
s
= 0
u
p
0 f
p
1
s
f
p
2
,0
. (No other changes needed.)
Proposition 5.8 Assume
0
< =
<
. Then in V
P
1

:
1.

B
1

is the algebra B
(W,

F)
, where W =
+
and

F =

f
1
s
: s

+
0

,
2. the algebra

B
1

is superatomic,
3. if s
+
and b

B
1

then there are nite v


0
v
1
(s) such
that either x
s
b or x
s
(b) equals to

x
t

v
1
(t

)<(t)
(x
t
) : t v
0
,
4. the height of

B
1

is
+
and x
,
: are representatives of atoms
of rank + 1,
5. t(

B
1

) =
+
.
Proof (1)(3) Repeat the arguments of 4.5(13) with no changes.
(4) Like 4.5(4), but the cases = 0 and = 1 are considered separately
(for > 1 no changes are required).
(5) Let a
def
=

a
p
: p
P
and look at the sequence x
,0
: a). It
easily follows from 5.6(c
1
) that it is a free sequence (so it witnesses t(

B
1

) =

+
).
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 30
Theorem 5.9 Assume
0
< =
<
. Then
P
1

Aut(

B
1

) = .
Proof It follows from 5.7 that, in V
P
1

, =
<
. By 5.8(2,4) we have
that each automorphism of

B
1

is determined by its values on atoms of



B
1

and x
0,
: < is the list of the atoms of

B
1

. Therefore it is enough to
show that in V
P
1

:
if

h :

B
1



B
1

is an automorphism then [ < :



h(x
0,
) ,= x
0,
[ < .
So assume that

h is a P
1

name for an automorphism of the algebra



B
1

and
p P
1

is such that 0, 1 w
p
. Now we consider three cases.
Case A: for each q p there are r P
1

and distinct , < such that


q r, (0, ), (0, ) u
r
u
q
, f
r
0,
u
q
0 and r
P
1


h(x
0,
) = x
0,
.
Construct inductively a sequence q
n
,
n
,
n
: n < ) such that
q
n
P
1

,
n
,
n
< , p = q
0
q
1
q
2
. . .,
(0,
n
), (0,
n
) u
q
n+1
u
qn
and f
q
n+1
0,n
u
qn
0,
q
n+1


h(x
0,n
) = x
0,n
.
Choose < such that (1, ) /

n<
u
qn
. Now we are dening a condition
r P
1

. First we put
w
r
=
_
n<
w
qn
, u
r
= (1, )
_
n<
u
qn
and a
r
=
_
n<
a
qn
.
Next for s u
q
we put
f
r
s
=
_

_
(1, ), 1)

m>n
f
qm
s
if s = (0,
n
), n ,
(1, ), 0)

m>n
f
qm
s
if s u
qn
(0,

) : n, n ,
0
u
r
\{s}
s, 1) if s = (1, ).
Furthermore, if s
0
, s
1
u
r
are distinct and such that (s
0
) (s
1
) then we
dene y
r
s
0
,s
1
as follows:
if (1, ) / s
0
, s
1
then y
r
s
0
,s
1
= y
qn
s
0
,s
1
, where n < is such that s
0
, s
1
u
qn
,
if s
0
= (1, ), s
1
u
qn
, n < then y
r
s
0
,s
1
= (0,
m
) : m n,
if s
1
= (1, ), s
0
u
qn
, (s
0
) = 1, n < then y
r
s
0
,s
1
= (0,
m
) : m n.
It is not dicult to check that the above formulas dene a condition r P
1

stronger than all q


n
(verifying 4.1(d) remember that f
q
n+1
0,n
u
qn
0). Note
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 31
that r (n < )(x
0,n
x
1,
) and hence r (n < )(x
0,n


h(x
1,
)).
Take a condition r

stronger than r and such that for some < we


have (1, ) u
r

and r


h(x
1,
)/

J
1
= x
1,
/

J
1
, where

J
1
is the ideal
of

B
1

generated by atoms (remember 5.8(4)). Then for some N we have


r

(n N)(x
0,n
x
1,
). Now look at the denition of the order in P
1

:
by 5.6() we have (1, ) u
r
. If (1, ) u
qn
for some n < then we get
immediate contradiction with 5.6(), so the only possibility is that = .
But then look at the denition of the functions f
r
0,m
they all take value 0
at (1, ) so r x
0,n
, x
1,
, a contradiction. Thus necessarily Case A does
not hold.
Case B: there are p

p and t u
p

such that
for each q p

there are r P
1

and distinct , < with:


q r, (0, ), (0, ) u
r
u
q
, r
P
1


h(x
0,
) = x
0,
, f
r
0,
(t) = 1 and
(s u
q
)((s) < (t) f
r
0,
(s) = 0).
First note that (by 5.6()) necessarily (t) > 1. Now apply the procedure
of Case A with the following modications. Choosing q
n
,
n
,
n
we demand
that q
0
= p

, f
q
n+1
0,n
(t) = 1 and (s u
qn
)((s) < (t) f
q
n+1
0,n
(t) = 1).
Next, dening the condition r we declare that f
r
1,
=

n<
f
qn
t
(1, ), 1)
and in the denition of y
r
s
0
,s
1
we let
if s
0
= (1, ) and either s
1
= t or (s
1
) < (t) then y
r
s
0
,s
1
= ,
if s
0
= (1, ) and (s
1
) (t), s
1
,= t then y
r
s
0
,s
1
= y
qn
t,s
1
, where n < is
such that s
1
u
qn
.
Continuing as in the Case A we get a contradiction.
Case C: neither Case A nor Case B hold.
Let q
0
p witness that Case A fails. So for each r q
0
and distinct , <
such that (0, ), (0, ) u
r
u
q
0
if r

h(x
0,
) = x
0,
then (t u
q
0
)(f
r
0,
(t) = 1).
Now, since Case B fails and P
1

is complete (and > ) we may build a


condition q
1
q
0
such that
if t u
q
1
, r q
1
, (0, ), (0, ) u
r
u
q
1
, r

h(x
0,
) = x
0,
,
f
r
0,
(t) = 1 and (s u
q
1
)((s) < (t) f
r
0,
(s) = 0)
then = .
But then clearly
q
1

P
1

( < )(

h(x
0,
) ,= x
0,
(0, ) u
q
1
) ,
nishing the proof.
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 32
Conclusion 5.10 It is consistent that there is a superatomic Boolean alge-
bra B such that t(B) =
+
and Aut(B) = .
6 When tightness is singular
In this section we will show that, consistently, there is a Boolean algebra
with tightness and such that there is an ultralter with this tightness
but there is no free sequence of length and no homomorphic image of the
algebra has depth . This gives partial answers to [M2, Problems 13, 41].
Next we show some bounds on possible consistency results here showing
that sometimes we may nd quotients with depth equal to the tightness of
the original algebra.
Let us recall that a sequence b

: < ) of elements of a Boolean


algebra B is (algebraically) free if for each nite sets F, G such that
max(F) < min(G) we have
B [=

F
b

G
(b

) ,= 0.
Existence of algebraically free sequences of length is equivalent to the
existence of free sequences of length in the space ultralters Ult(B).
Denition 6.1 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() <
0
, (i < cf())(
<
i
=
i
) and = sup
i<cf()

i
.
2) Let S = (, , ) be a good parameter. Put A
S
= (i, ) : i < cf() & 0

+
i
and dene a forcing notion Q
S
as follows.
A condition is a tuple p =
p
, w
p
, u
p
, f
p
i,,
: (i, ) u
p
, <
p
)) such
that
(a)
p
< , w
p
[cf()]
<
, u
p
[A
S
]
<
,
(b) (i w
p
)((i, 0), (i,
+
i
) u
p
) and if (i, ) u
p
then i w
p
,
(c) for (i, ) u
p
and <
p
, f
p
i,,
: u
p
2 is a function such that
if < then f
p
i,,
(i, ) = 0, if
+
i
then f
p
i,,
(i, ) = 1, and
f
p
i,,
(u
p
i
+
i
) = f
p
i,0,
(u
p
i
+
i
);
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 33
the order is given by p q if and only if
p

q
, w
p
w
q
, u
p
u
q
,
f
p
i,,
f
q
i,,
(for (i, ) u
p
, <
p
) and
((i, , ) u
q

q
)(f
q
i,,
u
p
f
p
j,,
: (j, , ) u
p

p
0
u
p).
3) We say that conditions p, q Q
S
are isomorphic if
p
=
q
, otp(w
p
) =
otp(w
q
) and there is a bijection H : u
p
u
q
(called the isomorphism from
p to q) such that if H
0
: w
p
w
q
is the order preserving mapping then:
() H(i, ) = (H
0
(i), ) for some ,
() for each i w
p
, the mapping
H
i
:
+
i
: (i, ) u
p

+
H
0
(i)
: (H
0
(i), ) u
q

given by H(i, ) = (H
0
(i), H
i
()) is the order preserving isomorphism,
() ( <
p
)((i, ) u
p
)(f
p
i,,
= f
q
H(i,),
H).
Remark: Variants of the forcing notion Q
S
are used in [RoSh 651] to deal
with attainment problems for equivalent denitions of hd, hL.
Proposition 6.2 Let S = (, , ) be a good parameter. Then Q
S
is a
complete
+
cc forcing notion.
Proof Easily Q
S
is closed. To show the chain condition suppose
that / Q
S
is of size
+
. Since
<
= we may apply standard cleaning
procedure and nd isomorphic conditions p, q / such that if H : u
p
u
q
is the isomorphism from p to q and H
0
: w
p
w
q
is the order preserving
mapping then
H
0
w
p
w
q
is the identity on w
p
w
q
, and
H u
p
u
q
is the identity on u
p
u
q
.
Next put
r
=
p
=
q
, w
r
= w
p
w
q
, u
r
= u
p
u
q
. For (i, ) u
r
and
<
r
we dene f
r
i,,
as follows:
if (i, ) u
p
, i w
p
w
q
then f
r
i,,
= f
p
i,,
f
q
H
0
(i),0,
,
if (i, ) u
q
, i w
q
w
p
then f
r
i,,
= f
p
H
0
1
(i),0,
f
q
i,,
,
if i w
p
w
q
then
f
r
i,,
= (f
p
i,0,
f
q
i,0,
) (u
r
i
+
i
)0
({i}[0,))u
r 1
({i}[,
+
i
])u
r
.
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 34
Checking that r
def
=
r
, w
r
, u
r
, f
r
i,
: (i, ) u
r
)) Q
S
is a condition
stronger than both p and q is straightforward.
For a condition p Q
S
let B
p
be the Boolean algebra B
(u
p
,F
p
)
for F
p
=
f
p
i,,
: (i, ) u
p
, <
p
0
u
p (see 1.1). Naturally we dene Q
S
names

S
and

f
i,,
(for i < cf(), <
+
i
, < ) by:

Q
S


B

S
=
_
B
p
: p
S
,

f
i,,
=
_
f
p
i,,
: (i, , ) u
p

p
, p
Q
S
.
Further, let

B
S
be the Q
S
name for the subalgebra x
i,
: i < cf(), <

+
i
)

S
of

B

S
.
Proposition 6.3 Assume S = (, , ) is a good parameter. Then in V
Q
S
:
1.

f
i,,
: A
S
2 (for < i < cf() and
+
i
),
2.

B

S
is the Boolean algebra B
(X
S
,

F)
, where

F =

f
i,,
: (i, ) A
S
, <
,
3. for each i < cf(), the sequence x
i,
: <
+
i
) is (algebraically) free
in the algebra

B
S
,
4. 0
X
S
cl(

F), so it determines a homomorphism from

B

S
to 2 (so
ultralter). Its restriction 0
X
S


B
S
has tightness .
Proof 1)3) Should be clear.
4) First note that if p Q
S
and i < cf() then there is a condition q Q
S
stronger than p and such that
( <
q
)(f
q
i,0,
(u
p
i (
+
i
+ 1)) 0).
Hence we immediately conclude that 0
X
S
cl(

F). Now we look at the
restriction 0
X
S


B
S
. First x i < cf() and let

Y
i
=

f
i,,


B
S
: <

+
i
, < (so

Y
i
is a family of homomorphisms from

B
S
to 2 and it can
be viewed as a family of ultralters on

B
S
). It follows from the previous
remark (and 6.1(2c)) that 0
X
S


B
S
cl(

Y
i
). We claim that 0
X
S


B
S
is not
in the closure of any subset of

Y
i
of size less than
+
i
. So assume that

X is
a Q
S
name for a subset of

Y
i
such that [

X[
i
(and we will think that


X
+
i
). Since Q
S
satises the
+
cc we nd <
+
i
such that

X
. Now note that 6.1(2c) implies that ((, )

X)(

f
i,,
(i, ) = 1),
so 0
X
S


B
S
/ cl(

X). Hence the tightness of the ultralter 0
X
S


B
S
is .
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 35
Theorem 6.4 Assume that S = (, , ) is a good parameter. Then in
V
Q
S
:
1. there is no algebraically free sequence of length in

B
S
,
2. if

I is an ideal in

B
S
then Depth(

B
S
/

I) < .
Proof 1) Assume that

: < ) is a Q
S
name for a sequence
of elements of

B
S
and p Q
S
. For each i < cf() and <
+
i
choose a
condition p
i,
Q
S
stronger than p, a nite set v
i,
u
p
i,
and a Boolean
term
i,
such that
p
i,

Q
S

i
+
=
i,
(x
j,
: (j, ) v
i,
).
Let us x i < cf() for a moment. Applying lemma arguments and
standard cleaning (and using the assumption that
<
i
=
i
= cf(
i
)) we
may nd a set Z
i
[
+
i
]

+
i
such that
()
i
all conditions p
i,
for Z
i
are isomorphic,
()
i
u
p
i,
: Z
i
forms a system with heart u
i
,
()
i
if
0
, Z
i
and H : u
p
i,
0
u
p
i,
1
is the isomorphism from p
i,
0
to
p
i,
1
then H[v
i,
0
] = v
i,
1
and H u
i
is the identity on u
i
,
()
i

i,
=
i
(for each Z
i
),
()
i
u
p
i,
0
(j, ) : j < i & <
+
j
= u
p
i,
1
(j, ) : j < i & <
+
j

whenever
0
,
1
Z
i
.
Apply the cleaning procedure and lemma again to get a set J [cf()]
cf()
such that
()

if i
0
, i
1
J,
0
Z
i
0
,
1
Z
i
1
then the conditions p
i
0
,
0
, p
i
1
,
1
are
isomorphic,
()

u
i
: i J forms a system with heart u

,
()

if i
0
, i
1
J,
0
Z
i
0
,
1
Z
i
1
and H : u
p
i
0
,
0
u
p
i
1
,
1
is the
isomorphism from p
i
0
,
0
to p
i
1
,
1
then H[v
i
0
,
0
] = v
i
1
,
1
, H[u
i
0
] = u
i
1
and H u

is the identity on u

,
()

i
= (for i J)
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 36
(remember the assumptions on cf() in 6.1(1)). Now choose i
0
J such
that supi < cf() : (i, 0) u

) < i
0
and pick
0
0
,
0
1
Z
i
0
,
0
0
<
0
1
. Next
take i
1
J such that
i
1
> i
0
+ supi < cf() : (i, 0) u
p
i
0
,
0
0
u
p
i
0
,
0
1

and u
i
1
(u
p
i
0
,
0
0
u
p
i
0
,
0
1
) = u

. Finally pick
1
0
,
1
1
Z
i
1
such that
1
0
<
1
1
and, for < 2,
u
p
i
1
,
1

(u
p
i
0
,
0
0
u
p
i
0
,
0
1
) = u

.
To make our notation somewhat simpler let p
k

= p
i
k
,
k

,
k

= (x
j,
: (j, )
v
i
k
,
k

) (for k, < 2) and let H


k
0
,
0
k
1
,
1
: u
p
k
0

0
u
p
k
1

1
be the isomorphism from
p
k
0

0
to p
k
1

1
(for k
0
, k
1
,
0
,
1
< 2).
It follows from the choice of i
k
,
k

that:
(i) if (i, 0) u

, k < 2, <
+
i
then (i, ) u
p
k
0
(i, ) u
p
k
1
,
(ii) if i (w
p
0
0
w
p
0
1
) (w
p
1
0
w
p
1
1
) then (i, 0) u

.
Now we are dening a condition q stronger than all p
k

. So we put

q
=
p
0
0
, w
q
= w
p
0
0
w
p
0
1
w
p
1
0
w
p
1
1
, u
q
= u
p
0
0
u
p
0
1
u
p
1
0
u
p
1
1
, and for
(j, ) u
q
and <
q
we dene f
q
j,,
: u
q
2 in the following manner.
We declare that
f
q
j,,
(j [0, )) u
q
0 and f
q
j,,
(j [,
+
j
]) u
q
1,
and now we dene f
q
j,,
on u
q
(j [0,
+
j
]) letting:
if (j, 0) u

then
f
q
j,,

_
,k<2
f
p
k

j,0,
(u
p
k

j
+
j
),
[note that in this case we have: f
q
j,,
(
k
0
) = f
q
j,,
(
k
1
) for k = 0, 1]
if (j, 0) u
i
0
u

then
f
q
j,,

_
<2
f
p
0

j,0,
(u
p
0

j
+
j
)
_
<2
f
p
1

H
0,0
1,0
(j,0),
,
[note that then f
q
j,,
(
1
0
) = f
q
j,,
(
1
1
)]
if (j, 0) u
p
k

u
p
k

: (k

) ,= (k, ), k

< 2 then
f
q
j,,
=
_
k

<2
f
p
k

H
k,
k

(j,),
,
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 37
[again, f
q
j,,
(
1
0
) = f
q
j,,
(
1
1
)]
if (j, 0) u
i
1
u

and, say, (j, ) u


p
1
0
then let j

w
p
0
0
be the isomorphic
image of j (in the isomorphism from p
1
0
to p
0
0
). Choose

<
+
j
such that,
if possible then, f
p
0
0
j

,
(
0
0
) = 0 (if there is no such

take

= 0). Let

= min : (j

, ) u
p
0
1
&

and
f
q
j,,
f
p
0
0
j

,
f
p
1
0
j

,

_
<2
f
p
1

j,0,
(u
p
1
1
j
+
j
)
[note that f
q
j,,
(
0
0
) f
q
j,,
(
1
1
)].
It is a routine to check that q =
q
, w
q
, u
q
, f
q
j,,
: (j, ) u
q
, <
q
))
Q
S
is a condition stronger than all p
k

. It follows from the remarks on


f
q
j,,
(
1
1
) we made when we dened f
q
j,,
that, by 1.2, B
q
[=
0
0

0
1

1
0

1
1
.
Hence we conclude that q forces that the sequence

: < ) is not free


as witnessed by
i
0
+
0
0
,
i
0
+
0
1
,
i
1
+
1
0
and
i
1
+
1
1
.
2) Suppose that

I is a Q
S
name for an ideal in

B
S
and p Q
S
is such that
p
Q
S
Depth(

B
S
/

I) = . Then for each i < cf() we nd a Q
S
name

b
i,
: <
+
i
) for a sequence of elements of

B
S
such that
q
Q
S
( < <
+
i
)(0/

I <

b
i,
/

I <

b
i,
/

I) .
Repeat the procedure applied in the previous clause, now with

b
i,
instead
of

b

i
+
there, and get i
0
, i
1
,
0
0
,
0
1
,
1
0
,
1
1
as there (and we use the same
notation p
k

,
k

, H
k
0
,
0
k
1
,
1
as before). Now we dene a condition q stronger
than all the p
k

. Naturally we let
q
=
p
0
0
, w
q
= w
p
0
0
w
p
0
1
w
p
1
0
w
p
1
1
,
u
q
= u
p
0
0
u
p
0
1
u
p
1
0
u
p
1
1
. Suppose (j, ) u
q
and <
q
. We dene
f
q
j,,
: u
q
2 declaring that
f
q
j,,
(j [0, )) u
q
0 and f
q
j,,
(j [,
+
j
]) u
q
1,
and:
if (j, 0) u

then f
q
j,,


,k<2
f
p
k

j,0,
(u
p
k

j
+
j
),
if (j, 0) u
p
k

but (j, 0) / u
p
k

for (k

) ,= (k, ) then
f
q
j,,
=
_
k

<2
f
p
k

H
k,
k

(j,),
,
if (j, 0) u
i
1
u

then
f
q
j,,

_
<2
f
p
1

j,0,
(u
p
1

j
+
j
)
_
<2
f
p
0

H
1,0
0,0
(j,0),
,
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 38
if (j, 0) u
i
0
u

then rst take

= min
+
j
: (j, ) u
p
0

&
(for < 2) and next put
f
q
j,,
= f
p
0
0
j,
0
,
f
p
0
1
j,
1
,
f
p
1
0
H
0,1
1,0
(j,
1
),
f
p
1
1
H
0,0
1,1
(j,
0
),
[remember that H
0,1
1,0
[u
i
0
] = H
0,0
1,1
[u
i
0
] = u
i
1
and both isomorphisms are the
identity on u

].
It should be a routine to verify that q =
q
, w
q
, u
q
, f
q
j,,
: (j, ) u
q
, <

q
)) Q
S
is a condition stronger than all p
k

. Note that the only case when


f
q
j,,
(
0
0
) ,= f
q
j,,
(
0
1
) is (j, 0) u
i
0
u

. But then f
q
j,,
(
1
0
) = f
q
j,,
(
0
1
) and
f
q
j,,
(
1
1
) = f
q
j,,
(
0
0
). Hence (by 1.2) B
q
[=
0
1
(
0
0
)
1
0
(
1
1
) and
therefore q

b
i
0
,
0
1
(

b
i
0
,
0
0
)

b
i
1
,
1
0
(

b
i
1
,
1
1
) . Now, q

b
i
1
,
1
0
/

b
i
1
,
1
1
/

I so we conclude q

b
i
0
,
0
1
(

b
i
0
,
0
0
)

I . But the last statement
contradicts q

b
i
0
,
0
0
/

I <

b
i
0
,
0
1
/

I , nishing the proof.


Conclusion 6.5 It is consistent that there is a Boolean algebra B of size
such that there is an ultralter x Ult(B) of tightness , there is no free
sequence in B and t(B) = / Depth
Hs
(B) (i.e. no homomorphic image of
B has depth ).
Let us note that in the universe V
Q
S
we have 2
cf ()
. This is a real
limitation we can prove that 2
cf ()
cannot be small in this context. In the
proof we will use the following theorem cited here from [Sh 233].
Theorem 6.6 (see [Sh 233, Lemma 5.1(3)]) Assume that = sup
i<cf()

i
,
cf() <
i
< , = (2
cf ()
)
+
. Let X be a T
3
1
2
topological space with a basis
B. Suppose that is a function assigning cardinal numbers to subsets of X
such that:
(i) (A) (A B) (A) +(B) +
0
for A, B X,
(ii) for each i < cf() there is a sequence u

: < ) B such that


(g : 2
cf()
)( ,= )(g() = g() & (u

cl
X
(u

))
i
),
(iii) for suciently large < , if A

: < ) is a sequence of subsets of


X such that (A

) then (

<
A

) .
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 39
Then there is a sequence u
i
: i < cf()) B such that
(i < cf())((u
i

_
j=i
u
j
)
i
).
Theorem 6.7 Suppose that B is a Boolean algebra satisfying 2
cf (t(B))
<
t(B). Then for some ideal I on B we have Depth(B/I) = t(B).
Proof Let = t(B) and let
i
: i < cf()) be an increasing conal in
sequence of successor cardinals,
0
> cf() and let = (2
cf ()
)
+
. Further,
let X be the Stone space Ult(B) and thus we may think that B = B is a
basis of the topology of X. Now dene a function on subsets of X by
(Y ) = sup: there are sequences y

: < ) Y and u

: < ) B
such that (, < )(y

< ).
We are going to apply 6.6 to these objects and for this we should check
the assumptions there. The only not immediate demands might be (ii) and
(iii). So suppose i < cf(). Since
i
< = t(B) we can nd a free sequence
u

: <
+
i
) B. Next, for each <
+
i
we may choose an ultralter
y

X such that ( <


+
i
)(y

< ). Now, for < , let


u

= u

. Suppose g : 2
cf()
and take any < < such that
g() = g(). Note that
u

cl
X
(u

) = u

:
i
< <
i
( + 1)
and easily (y

:
i
< <
i
(+1)) =
i
. Thus (u

cl
X
(u

))
i
and the demand 6.6(ii) is veried. Assume now that < < and A

X
(for < ) are such that (

<
A

) > . Let sequences y

: <
+
)

<
A

and u

: <
+
) B witness this. Then for some C [
+
]

+
and
< we have y

: C) A

and therefore y

, u

: C) witness
(A

)
+
. This nishes checking the demand 6.6(iii).
So we may use 6.6 and we get a sequence u
i
: i < cf()) B such that
(i < cf())((u
i

_
j=i
u
j
)
i
).
Then for each i < cf() we may choose sequences y
i

: <
i
) u
i


j=i
u
j
and w
i

: <
i
) B such that
y
i

w
i

< ,
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 40
and we may additionally demand that w
i

u
i
(for each <
i
). Now let
I
def
= b B : (i < cf())( <
i
)(y
i

/ b).
It should be clear that I is an ideal in the Boolean algebra B (identied
with the algebra of clopen subsets of X). Fix i < cf() and suppose that
< <
i
. By the choices of the w
i

s we have y
i

w
i

w
i

and no y
i

belongs to w
i

w
i

. As w
i

u
i
we conclude B/I [= w
i

/I < w
i

/I. Thus
the sequence w
i

/I : <
i
) (for i < cf()) is strictly decreasing in B/I
and consequently Depth(B/I) . Since there is many y
i

s only, we may
easily check that there are no decreasing
+
sequences in B/I (remember
the denition of I), nishing the proof.
References
[BaSh 254] James E. Baumgartner and Saharon Shelah. Remarks on su-
peratomic Boolean algebras. Annals of Pure and Applied Logic,
33:109129, 1987.
[J] Thomas Jech. Set theory. Springer Monographs in Mathematics.
Springer-Verlag, Berlin, 2003. The third millennium edition,
revised and expanded.
[Ko89] Sabine Koppelberg. Handbook of Boolean Algebras, volume 1.
NorthHolland, 1989. Monk, D. and Bonnet, R. eds.
[Ma92] Juan C. Martinez. A consistency result on thintall superatomic
Boolean algebras. Proceedings of the American Mathematical
Society, 115:473477, 1992.
[M1] J. Donald Monk. Cardinal Invariants of Boolean Algebras. Lec-
tures in Mathematics. ETH Zurich, Birkhauser Verlag, Basel
Boston Berlin, 1990.
[M2] J. Donald Monk. Cardinal Invariants of Boolean Algebras, vol-
ume 142 of Progress in Mathematics. Birkhauser Verlag, Basel
BostonBerlin, 1996.
[Rt89] Judy Roitman. Superatomic Boolean algebras. In Handbook
of Boolean Algebras, volume 3, pages 719740. NorthHolland,
1989. Monk, D. and Bonnet, R. eds.
5
9
9


r
e
v
i
s
i
o
n
:
1
9
9
7
-
0
4
-
1
8







m
o
d
i
f
i
e
d
:
1
9
9
7
-
0
4
-
1
8


[RoSh 599] August 17, 2011 41
[RoSh 534] Andrzej Roslanowski and Saharon Shelah. Cardinal invariants of
ultrapoducts of Boolean algebras. Fundamenta Mathematicae,
155:101151, 1998. math.LO/9703218.
[RoSh 651] Andrzej Roslanowski and Saharon Shelah. Forcing for
hL and hd. Colloquium Mathematicum, 88:273310, 2001.
math.LO/9808104.
[Sh 233] 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 Mathematics,
pages 151187. Springer, Berlin, 1986.
[Sh 479] Saharon Shelah. On Monks questions. Fundamenta Mathemat-
icae, 151:119, 1996. math.LO/9601218.
[Sh 620] Saharon Shelah. Special Subsets of
cf()
, Boolean Algebras
and Maharam measure Algebras. Topology and its Applica-
tions, 99:135235, 1999. 8th Prague Topological Symposium
on General Topology and its Relations to Modern Analysis and
Algebra, Part II (1996). math.LO/9804156.

You might also like