3

9
7


r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
3
-
1
0







m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
3
-
1
0


FACTOR = QUOTIENT, UNCOUNTABLE BOOLEAN ALGEBRAS
FACTOR = QUOTIENT,
UNCOUNTABLE BOOLEAN ALGEBRAS,
NUMBER OF ENDOMORPHISM AND WIDTH
Saharon Shelah
June 22, 1991
Abstract. We prove that assuming suitable cardinal arithmetic, if B is a Boolean
algebra every homomorphic image of which is isomorphic to a factor, then B has
locally small density. We also prove that for an (innite) Boolean algebra B, the
number of subalgebras is not smaller than the number of endomorphisms, and other
related inequalities. Lastly we deal with the obtainment of the supremum of the
cardinalities of sets of pairwise incomparable elements of a Boolean algebra.
We show in the rst section:
0.1 Conclusion. It is consistent, that for every Q = F Boolean algebra B

,
for some n < , x : B

[x has a density
n
is dense (so B

has no independent
subset of power
n
).
Where:
0.2 Denition. A B.A. is Q = F (quotient equal factor) if: every homomor-
phic image of B is isomorphic to some factor of B i.e B[a for some a B.
The consistent is really a derivation of the conclusion from a mild hyphothesis
on cardinal arithmetic (1.2). The background of this paper is a problem of Bonnet
whether every Q = F Boolean algebra is superatomic.
Noting that : B[x has density
n
is a weakening of x is as atom of B

,
we see that 0.1 is relevant.
The existence of non trivial example is proved in R. Bonnet, S. Shelah [2].
M. Bekkali, R. Bonnet and M. Rubin [1] characterized all interval Boolean alge-
bras with this property.
In the second section we give a more abstract version. In a paper in preparation
, Bekkali and the author use theorem 2.1 to show that every Q = F tree Boolean
algebras are superatomic.
In the third section we deal with the number of endomorphism (e.g. aut(B)
0

end(B)) and in the fourth with the width of a Boolean algebra.

We thank D. Monk for detecting an inaccuracy in a previous version.
Notation
B denote a Boolean algebra.
Partially supported by Basic Research Fund of the Israeli Academy of Sciences. Publ. No. 397
Typeset by A
M
S-T
E
X
1
3
9
7


r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
3
-
1
0







m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
3
-
1
0


FACTOR = QUOTIENT, UNCOUNTABLE BOOLEAN ALGEBRAS
B
+
is the set of non zero members of B.
B[x where x B
+
is B[y : y x.
compB is the completion of B, so it is an extension of B.
Remember : if B
1
is a subalgebra of B
2
, B is complete, ha homormorphism
from B
1
to B, then h can be extended to a homomorphism from B
2
to B.
id
A
is the identity function on A.

= : is a sequence of lenghts of ordinals < .

>
=

<

.
1 Maybe every quotient equal factor BA has locally small density .
1.1 Hypothesis.
(1) B

is a Q = F Boolean algebra.
(2) for each n < , B

has a factor B
n
s.t.: 0 < x B
n
density(B
n
[x)
n
.
1.2 Hypothesis. For we have 2
||
>

.
1.3 Desired Conclusion. Contradiction.
We shall use 1.1 all the times, but 1.2 only in 1.17.
1.4 Denition. K

is the class of Boolean algebra such that:

(x B
+
)[density(B[x) = ].
1.5 Claim. If B is atomless, x B
+
then for some y, 0 < y x,and innite
cardinal we have B[y K

density(B|x)
.
1.6 Claim. If B K

, B B

comp(B) then,
B

.
1.7 Claim. If B K

,
0
< , regular then some subalgebra B

of B is
in K

. If in addition B is a Q = F algebra, then some homomorphic image B

of
B is in K

.
Proof. Choose by induction on i < , B
i
B, [B
i
[ , [i < j B
i
B
j
]
such that:
if x B
+
i
then there is y(x, i) B
(i+1)
satisfying:
(1) 0 < y(x, i) x,
(2) for no z B
i
, 0 < z y(x, i).
(possible as for each i and x B
+
i
density(B
i
[x) = ). Let now B

=
i<
B
i
; it
is a subalgebra of B. Now x (B

)
+
density(B

[x) = as on the one hand

[B

i<
[B
i
[ = implies density(B

[x) for every x B

and on
the other hand if x (B

)
+
, A B

[x, [A[ < then for some i < , A B

i
, hence
y(x, i) (B

)
+
wittness A is not dense in B

[x, Now, if B is a Q = F Boolean

Algebra, then id
B
can be extended to a homomorphism h

from B into comp(B

),
so h

(B) is as required.
3
9
7


r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
3
-
1
0







m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
3
-
1
0


FACTOR = QUOTIENT, UNCOUNTABLE BOOLEAN ALGEBRAS
1.8 Conclusion. : regular and B

has a factor in K

is an initial segment
of

regular.
1.9 Denition.
(1) () is minimal such that for no >
()
does B

has a factor in K

. Let

= () =: [()[.
(2) Let for <

, b

be such that B

[b

+1
.
(3) Let J

be the ideal b B

: B

[b K

+1
.
1.10 Denition. B[
>
] is the Boolean algebra generated freely by x

:
>
except x

x
|m
(m lg (),
>
).
1.11 Claim.
(1) For < <

, J

= 0 and J

is an ideal.
(2) For no B

and proper ideal I of B

, [x B

I density(B

[x/I) >

()
].
Proof.
(1) Trivial.
(2) By 1.5 for some > () and proper ideal J of B

(of the form x

B

: x b I) we have B

/I K

so there is a homomorphism h from B

into
comp(B

/I) extending x x/I (x B

). So B

has a factor isomorphic to Rang h,

but this Boolean algebra is in K

>
()
. So by 1.7 we get contradiction to
the choice of ().
1.12 Denition. 1) I

=: x B

:

(<())
J

is dense below x.
2) For A () : I

A
=: x B

: A is dense below x.
1.13 Claim.
(1) I

, I

A
are ideals of B.
(2) A B I

A
I

B
I

.
1.14 Claim. For every A (), there are c
A
, h
A
such that:
(1) c
A
B

and I

A
is dense below c
A
,
(2) h
A
is a homomorphism from B

onto B

[c
A
,
(3) h
A
[I

A
is one to one,
(4) If B[x I

A
= 0 then h
A
(x) = 0,
(5) B

[c
A
is a subalgebra of the completion of the subalgebra h
A
(x) : x I

A
.
Proof. Let ba(A) be
I

A
1 x : x I

A
,
this is a subalgebra of B

.
Let h
1
be a homomorphism from B

to comp

b a(A)

extending id
ba(A)
and
h
1
(x) = 0 if B

[x I

A
= 0.
Now h
1
(B

) is a quotient of B

hence there is an isomorphism h

2
from h
1
(B)
onto some B

[c
A
.
3
9
7


r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
3
-
1
0







m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
3
-
1
0


FACTOR = QUOTIENT, UNCOUNTABLE BOOLEAN ALGEBRAS
Let
h
A
= h
2
h
1
so (1), (2), (3), (4), (5) holds.
1.15 Claim. Let A ()
(1) If x

A
J

then h
A
(x)

A
J

.
(2) If x I

A
J

then h
A
(x) ,

A
J

.
(3)

A
h
A
(J

) is dense and downwared closed in B

[c
A
.
(4) c
A
I

A
.
(5) If x J

and A then h
A
(x) J

.
Proof.
(1) By (5) of 1.14.
(2) is easy too.
(3) is easy too.
(4) is easy too.
(5) is easy too.
1.16 Claim. We can nd x

for
>
where = 2
()
such that:
(1) m < lg() 0 < x

x
(|m)
.
(2) If

k
e=1

e
,
>
,

e


then
B

,
k

e=1
x
e
.
Proof. Now we can choose A
0

:
>
) which is a family of subsets of ()
such that any non trivial Boolean combination of then has cardinality ().
Let for
>
A

=
def

elg()
A
|e
. Let
(a) x
<>
= c
A<>
= h
A<>
(1
B
).
(b) x
<i>
= h
A<>
(c
A<i>
) = h
A<>
h
A<i>
(1
B
)
and generally,
x
<i0,i1,...,in1>
= h
A<>
h
A<i
0
>
h
A<i
0
,i
1
>
. . . h
A<i
0
,i
1
,...,i
n1
>
(1
B
).
We prove (a) by induction of lg.
The reader may check
1.17 Final Contradiction. x

:
>
from Claim 1.16 contradict by
1.11(2) and the choice of
()
, because = 2
()
= 2
|()|
>
()
.
[of course
()
[B

[]
Actually, we have prove more.
3
9
7


r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
3
-
1
0







m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
3
-
1
0


FACTOR = QUOTIENT, UNCOUNTABLE BOOLEAN ALGEBRAS
1.18 Remark. (1) So we have in 1.17 prove that if set theory is as in Hypoth-
esis 1.2, then there is no Boolean algebra as in 1.1, hence proving Conclusion
0.1
(2) Note: if 1.2, any Q = F Boolean algebra has no factor T().
2 Q = F Boolean algebras: a general theorem .
2.1 Theorem. Suppose:
(1) B

is a (Q = F) Boolean algebra.
(2) N is a family of (non zero) members of B

(the nice elements).

(3) a cardinal (
0
), K

: < ) a sequence.
(4) K

is family of Boolean algebras closed under isomorphism and for ,=

we have K

= .
(5) for every some factor of B

is in K

.
(6) if x (B

)
+
, (B

[x) K

then for some y x, B

[y K

, y N.
(7) if x
1
, x
2
N, B

[x
1
K
1
, B

[x
2
K
2
,
1
,=
2
then x
1
x
2
= 0.
(8) if x B

, N

N then
() for some y
1
, , y
n
N

,for every z N

we have B

x z
y
1
y
n
or
() for some y N

: y x.
Actually we use (8) only for N

of the formy N : ()[ A&B

[y
K

.
(9) if x < y B

, B

[x K

and B

[y K

, then = .
Then the Boolean algebra B[
>
(2

)] can be embedded into B

, remember
>
(2

)
is
the tree : a nite sequence of ordinals < 2

and Def 1.10.

Proof of theorem 2.1.
Let for < , Y

=: y N : B

[y K

.
For A let us dene
I
A
=the ideal generated by Y
A
=

A
Y

= y N : B

[y K

for some
A and
J
A
=: z B

: for every y Y
A
we have z y = 0.
Clearly J
A
is an ideal.
Now for each A B

/J
A
is a quotient of B

. Hence by condition (1) there

are y

A
B

and an isomorphism h
A
: B

/J
A
B

[y

A
onto.
Let g
A
: B

/J
A
be canonical, so h
A
g
A
(1
B
) = y

A
. Let f
A
= h
A
g
A
.
Dene for y B

the following: cont(y) =: : (y

y)[B

[y

].
(i.e. the content of y). We next prove
()
1
cont (y

A
) A.
Proof. By conditon (5) for each A, there is x

, such that: B

[x

.
By condition (6) wlog x

N, hence x

I
A
, hence g
A
[(B

[x

), is one to one,
hence B

[x

= B

[f
A
(x

), hence by condition (4) B

[f
A
(x

) K

; nowRangf
A
=
B

[y

A
, so cont(y

A
). So we have prove ()
1
.
()
2
cont(y

A
) A
Proof. Suppose cont(y

A
), so there is z y

A
, such that B

[z K

. As f
A
is
a homomorphism from B

onto B

[y

A
, there is x B

such that f
A
(x) = z.
3
9
7


r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
3
-
1
0







m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
3
-
1
0


FACTOR = QUOTIENT, UNCOUNTABLE BOOLEAN ALGEBRAS
Now the kernel of f
A
is J
A
, and B

[0 , K

so x , J
A
; and clearly (B

[x)/J
A
is
in K

.
Hence by ()
3
below A.
()
3
if x B

J
A
and (B

/J
A
)[f
A
(x) K

then A
Proof. We apply condition (8) to x and N

= Y
A
. So one of the following two
cases ocurrs:
Case : There are n < , y
1
, , y
n
N

such that:
(z N

) x z y
1
. . . y
n
.
So x (y
1
y
n
) J
A
(by denition of J
A
).
Let x
1
= x (y
1
y
n
) hence B

[f
A
(x) = (B

[x)/J
A

= (B

[x
1
)/J
A

=
B

[x
1
,(last isomorphism as
e
y
e
Y
A
hence y
1
y
n
I
A
hence f
A
[(B

[x
1
)
is one to one). So B

[x
1
K

, hence by condition (6) for some x

2
x
1
,we have
x
2
N&B

[x
2
K

.
Let B

[y
e
K
e
where
e
A(by denition of Y
A
). Clearly x
2
x
1
y
1

y
n
, so for some e, y
e
x
2
,= 0 hence =
e
A (by condition (7)) so we get the
trivial desired conclusion.
Case : There is y N

, y x.
As y N

= Y
A
=
A
Y

for some A we have y K

, also y N

N so
J
A
(B

[y) = 0 so (B

/J
A
)[g
A
(y) K

. Remembering (B

/J
A
)[g
A
(x) K

, as
B

/J
A

= B

[y

A
we get by condition (9) that = . So ()
3
hence ()
2
is proved.
Next we prove
()
4
if B A , and cont(y) = B, then cont[f
A
(y)] = B.
Proof.
inclusion
First let B, then for some x y,B

[x K

, and by condition (6) wlog x

N, hence x I
A
(as B A) hence f
A
[(B

[x) is one to one and onto B

[f
A
(x)
so f
A
(x) f
A
(y), B

[f
A
(x) K

, so cont(f
A
(y)).
inclusion
Second let us assume cont[f
A
(y)]. So (as f
A
is onto B

[y

A
, and if f
A
(x)
f
A
(y) then f
A
(x y) = f
A
(y), x y y) there is x y such that B

[f
A
(x) K

.
Now apply condition (8) to x and Y
A
. So case () or case () below holds.
Case : There are n < , y
1
, , y
n
Y
A
such that for every z Y
A
we have
x z y
1
y
n
.
Hence x (y
1
y
n
) J
A
let x
1
= x (y
1
... y
n
) so f
A
(x) = f
A
(x
1
) so
B

[f
A
(x
1
) K

and of course x
1
y
1
y
n
, (and x
1
y) so f
A
[(B

[x
1
) is one
to one.
Now f
A
is one to one on B

[x
1
hence B

[x
1

= B

[f
A
(x
1
) K

. Now x
1
x y,
so x
1
wittness cont(y), which is B.
Case : There is t x, t Y
A
.
Now t x y, B

[t K

for some A as t Y
A
. Now f
A
is one to one on
B

[f
A
(t) K

.
Also f
A
(t) f
A
(x) hence by assumption (9) we have = . Also t x y so
t wittness cont(y), so = cont(y) = B as required.
So we have proved ()
4
3
9
7


r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
3
-
1
0







m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
3
-
1
0


FACTOR = QUOTIENT, UNCOUNTABLE BOOLEAN ALGEBRAS
end of proof of theorem 2.1: Let = 2

.
Let |

:
>
) be a family of subsets of , any nite Boolean combination
of them has power (or just ,= ).
Let |

=
elg
|
|e
. Now dene for every
>
and e lg() an element y
e

of B

:
y
e

=
def
f
U

|e
f

U
|(e+1)
f
U

|(n1)
f
U

(1
B
) and y

=
def
y

Now: (a) prove for each

n
by downward induction on e 0, 1, ..., n that
cont(y
e

) = |

; for e = n this is ()
1
+ ()
2
as y
n

= y

U
;
for e < n (assuming for e + 1) this is by ()
4
.
Next note: (b) if = < > then y

[prove by domnward induction for e 0, 1, ..., lg we have : y

e+1
v
y
e

; remem-
ber f
U
is order preserving].
Lastly note (c) if
>
, n < , and
e

>
is not initial segment of for
e = 1, . . . , n then y

n
e=1
y

e
,=; this follows by (a) and the denition of cont(y

).
Now by (a), (b), (c) there is an embedding g from the subalgebra of B

which
y

:
>
generates mapping y

to x

.
3 The Number of Subalgebras .
3.1 Denition. For a BA A
(1) Sub(A) is the set of subalgebras of A.
(2) Id(A) is the set of ideals of A.
(3) End(A) is the set of endomorphisms of A.
(4) Pend(A) is the set of partial endomorphisms ( i.e. homomorphisms from a
subalgebra of A into A).
(5) Psub(A) is the family of subsets of A closed under union ,intersection and
substruction but 1 may be not in it though 0 is [so not neccessarily closed
under complementation].
(6) We let sub(A), id(A), end(A), aut(A), pend(A), psub(A) be the cardinality
of Sub(A), Id(A), End(A), Aut(A), Pend(A) and Psub(A) respectively
In D. Monk [4] list of open problems appear:
PROBLEM 63. Is there a BA A such that aut(A) > sub(A) ?
See [4] page 125 for backgraound.
3.2 Theorem. For a BAA we have: aut(A) is not bigger than sub(A).
3.3 Conclusion.
(1) For a BAA we have end(A) is not bigger than sub(A).
(2) For a BAA we have pend(A) is exactly sub(A).
(3) For a BAAand a in A, 0 < a < 1 we have sub(A) = Maxsub(A[a), sub(A[
a).
We shall prove it in 3.5
Remark. Of course - A is innite- we many times forget to say so.
3.4 Proof of the theorem. Let be sub(A).
3.4A Observation: Psub(A) has cardinality Sub(A) [why? for the less trivial
inequality, , for every X in Psub(A) which is not a subset of 0, 1 choose a
member a ,= 0, 1 in X and let Y
a
[X] be the subalgebra generated by x : x
3
9
7


r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
3
-
1
0







m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
3
-
1
0


FACTOR = QUOTIENT, UNCOUNTABLE BOOLEAN ALGEBRAS
X, x a let Z
a
[X] be the subalgebra generated by x : x X, x a = 0; now
X can be reconstructed from a, Y, Z) as x A : x a Y and x a Z.
So [Psub(A)[ [A[ [Sub(A)[
2
+ 4 = [Sub(A)[ +
0
= [Sub(A)[. (remember
that [A[
0
, and that [A[ [Sub(A)[, as the number of non-atoms of A is
[Z
a
[A] : a A[).
For any automorphism f of A we shall choose a nite sequence of members of
Psub(A) ( in particular of ideals of A ), and this mapping is one to one, thus we
shall nish.
Let J =
def
x : x A, and for every y A below x, we have f(y) = y; let I

be the ideal of A generated by I, the set of elements x for which f(x) x = 0.

Observe : x I implies f(x) I [why? as y = f(x), x I implies f(x) x = 0
hence f(y) y = f(f(x)) f(x) = f(f(x) x) = f(0) = 0].
Observe: J I is a dense subset of A [if x B
+
and there is no y J, 0 < y x
then wlog x ,= f(x). If x , f(x) then z =: x f(x) satises 0 < z x, f(z) z
f(x) z = f(x) (x f(x)) = 0; so z I. If x f(x) then x < f(x); as f is
an automorphism of B

, for some z in B

we have f(z) = x, so x < f(x) means

f(z) < f(f(z)) hence z < f(z), and z

= x z = f(z) z is in I, is > 0 but < x,

as required.]
Next let x
i
: i < be a maximal sequence of distinct members of I satisfying :
for any i, j < we have x
i
f(x
j
) = 0, let I
1
be the ideal generated by x
i
: i <
and let I
2
be the ideal generated by f(x
i
) : i < .
Clearly I
1
I
2
= 0 let I
0
= y : y I

, and for every x I

1
, y x =
0 and f(y) I
1
and let I
3
be the ideal of A generated by f(x) : x I
2

Observe : each I
t
(t = 0, 1, 2, 3) is an ideal of A, contained in I

[why? for
t = 1 by choice each x
i
is in I, t = 0 by its denition , I
2
, I
3
by their denition
and an observation above, i.e., x I f(x) I]
Observe: for t = 0, 1, 2 we have: I
t
I
t+1
= 0 [why ? for t = 1, by the choice
of x
i
: i < , for t = 0 by denition of I
0
, lastly for t = 2 applies its denition +
f being an automorphism.]
Observe: I
0
I
1
I
2
is a dense subset of I

[why? assume x in I

but below
it there is no non zero member of this union , so we can replace it by any non zero
element below it; as x I

, there is below it a non zero element y with y f(y) = 0

so wlog x f(x) = 0; why have we not choose x

= x? there are two case:

Case 1: For some j < , x f(x
j
) is not zero
so there is a non zero element below x in I
2
.
Case 2: for some j < , f(x) x
j
is not zero
so there is a non zero y in I
1
below f(x) hence (as f is an automorphism) there
is x

below x such that f(x

) = y so wolg f(x) is in I
1
, but then by its denition,
either below x there is a member of I
1
or x is in I
0
so we have nished proving the
observation.]
Observe: for t = 0, 1, 2, x I
t
f(x) I
t+1
[check].
Now we dene C
t
= C
f
t
, a member of Psub(A) for t = 0, 1, 2 as follows: C
t
is
the set x f(x) : x I
t
. The closure under the relevant operations follows as I
t
is closed under them and f is an isomorphism and for x, y in I
t
, x f(y) = 0, this
is needed for substraction.
Also for every automorphismf of Awe assign the sequence J, I, I
0
, I
1
, I
2
, I
3
, C
0
, C
1
,
C
2
) (some reddundancy). Suppose for f
1
, f
2
Aut(A) we get the same tuple; it
is enough to show that their restriction to J and to I are equal -as the union is
3
9
7


r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
3
-
1
0







m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
3
-
1
0


FACTOR = QUOTIENT, UNCOUNTABLE BOOLEAN ALGEBRAS
dense. Concerning J this is trivial - they are the identity on it so we discuss I, by
an observation above it is enough to choeck it for I
t
, t = 0, 1, 2 but for each t, from
C
t
and I
t
, I
t+1
we can [or see below] reconstruct f[I
t
.
So we have nished the proof.
3.4B Remark. We can phrased this argument as a claim: let I, J, be ideals of
A with intersection 0; for every f, a one to one homomorphism from I to J let
X
f
be the set x f(x) : x I then f mapped to X
f
is a one to one mapping
from HOM(I, J) to Psub(A) (the former include g[I : g Aut(A), g maps I onto
J , for which we use this). For subalgebra relative to , only f needs not be one
to one.
3.5 Proof of the conclusion from the theorem.
(1) For a given Boolean Algebra B assume =: sub(A) is < end(A). For any
endomorphism f of A we attach the pair (Kernel(f), Range(f)) . The
number of possible such pairs is at most id(A) sub(A) , which is at most
(we are dealing with innite BAs and id(A) psub(A) sub(A)) so
as we assume < end(A), there are distinct f
i
, endomorphisms of A for
i <
+
, an ideal I and a subalgebra R of B such that for every i we have
Kernel(f
i
) = I and Range(f
i
) = R.
We now dene a homomorphism g
i
from B/I to R by : g
i
(x/I) = f
i
(x)
. Easily the denition does not depend on the representative, so g
i
is as
required and it is one to one and onto. So g
i
(g
0
)
1
: i <
+
is a set of

+
distinct automorphisms of R.
So
()
1
< aut(R)
but, by the theorem
()
2
aut(R) sub(R)
obviously
()
3
sub(R) sub(A)
remember
()
4
= sub(A)
together a contradiction
(2) As R ( identity map on R) is one to one from sub(A) into Pend(A),
obviously sub(A) pend(A), so we are left with proving the other inequal-
ity. Same proof, only the domain is a subalgebra too and it has an ideal.
So for every such partial endomorphism h of A we attach two subalgebras
D
h
= Domain(h) and R
h
= Range(h) an ideal I
h
of D
h
x : x D
h
and
3
9
7


r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
3
-
1
0







m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
3
-
1
0


FACTOR = QUOTIENT, UNCOUNTABLE BOOLEAN ALGEBRAS
f(x) = 0 . They are all in Psub(A), so their number is at most sub(A) and
if we xed them the amount of freedom we have left is : an automorphism
of R (and aut(R) sub(R) sub(A)).
(3) Let B be a subalgebra of A. We shall attached to it several ideals and
subalgebras of B[a, B[(a) such that B can be reconstructed from them;
this clearly suce. Let C be the subalgebra B, a), C
o
= x C : x
a, C
1
= x C : x 1 a. The number of possible C is clearly the
number of pairs C
o
, C
1
) which is clearly sub(A[a) +sub(A[ a) ; x C. Let
I =: x : x B, x a, it is an ideal of C[a ; so the number of such I is at
most
id(C[a) pend(C[A) sub(C[a) sub(A[a).
So we can x it . Similarly we can x J = y : y B, y (1 a),
now I and J are subsets of B , now check : the amount of freedom we have
left is an isomorphism g from (C[a)/I onto (C[ a)/J such that B = the
subalgebra of C generated by I J x y : x (C[a), y (C[ a) and
g(x/I) = (y/J).
So we can nish easily.
We originally want to prove Aut(A)
0
sub A and even
[f End(A) : f is onto [
0
subA. But we get more: intermediate invariants
with reasonable connections.
3.6 Denition: For a Boolean algebra A
(0) A partial function f from A to A is everywhere onto if:
x Dom(f)&y Rang(f)&y f(x) (z)[z x&f(z) = y].
(1) End
0
(A) = End(A).
End
1
(A) = f End(A) : Rang(f) include a dense ideal.
End
2
(A) is the set of endomorphism f of A which are onto.
End
3
(A) = f : for some dense ideal I, f is an onto endomorphism of I.
End
4
(A) = f : for some ideal I of A, f is an onto endomorphism of I.
End
5
(A) = f : for some dense ideals I, J, f is an homorphism from I
onto J.
End
6
(A) = f : for some ideals I, J of A, f is an endomorphism from I
onto J.
Note I = A is allowed. All kinds of endomorphism, commute with , ,
preserve 0 but not necessarily .
(2) For l = 1 , 6 we let Aut
l
(A) = f End
l
(A) : f(x) = 0 x = 0 for
x Domf.
(3) For function f, g whose domain is A let: f g if Kerf = Kerg and x
: f(x) = g(x) or both are dened not include a dense ideal of B/Kerf.
(4) Let end
e
(A) = [End
e
(A)[, aut
e
(A) = [Aut
e
(A)[.
Let end

e
(A) = [f/ : f End
e
(A)[ and aut

e
(A) = [f/ : f
Aut
e
(A)[.
(5) We allow to replace A by an ideal I with the natural changes.
(6) We dene Endv
e
(A), endv
e
(A) similarly replacing onto by everywhere
onto and dene Endl
e
(A), endl
e
(A) similarly omitting onto. We dene
Endu
e
(A) as the set of f End
e
(A) such that for every x Domf, f(x) ,=
0 and ideal I of A[x which is dense, we have f(x) = sup
A
f(y) : y I.
We dened naturally Autv
e
(A), Autl
e
(A), Autv
e
(A) etc.
3
9
7


r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
3
-
1
0







m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
3
-
1
0


FACTOR = QUOTIENT, UNCOUNTABLE BOOLEAN ALGEBRAS
Note: In Endv
1
(A) we mean: for some dense ideal I of A, y f(x)
I (z x)(f(z) = y) and Endl
1
(A) = End(A).
3.7 Claim.
(1) End
6
(A) End
5
(A) End
4
(A) End
5
(A) End
4
(A) End
3
(A)
End
2
(A) and End
2
(A) End
1
(A) End
0
(A) = EndA.
(2) end
6
(A) maxend
5
(A), end
4
(A) minend
5
(A), end
4
(A) end
3
(A)
end
2
(A), and end
2
(A) end
1
(A) end(A).
(3) In (1) we can replace End by Aut or Endv or aEndl or Autv or Autl, or
Endu or Autu and in (2) end by aut etc. We can in (2) replace end
l
(or
aut
l
) by end

l
(or aut

l
) etc.
(4) Aut
l
(A) End
l
(A) hence aut
l
(A) end
l
(A); and Endv
e
(A) End
e
(A)
Endl
e
(A) hence endv
e
(A) end
e
(A) endl
e
(A); Endv
e
(A) = End
e
(A) if
e = 0. Also Aut(A) = Aut

2
(A) hence aut(A) = aut

2
(A).
(5) If f, g, Endv
6
(A), f g then f, g, are compatible functions,i.e. x
Domf Domg f(x) = g(x).
(6) sub(A) aut
6
(A) and aut(A) aut

3
(A).
(7) Aut
e
(A) = Autv
e
(A).
(8) end
e
(A) id(A) +end

e
(A) etc.
Proof. E.g.
(5) Supposse x Domf Domg, f(x) ,= g(x), so wlog f(x) , g(x) so let
z =: f(x)g(x) ,= 0. As z f(x) for some t Domf we have 0 < f(t) z,
and wlog f(t) = g(t), and we get contradiction.
(6) As in proof of 3.2 or see proof of 3.12 (noting [I : I a dense ideal of
A[ sub(A) by 3.4A).
3.8 Denition. For a Boolean Algebra A
(1) Idc(A) = I : I an ideal of A, and I
c
= I where I
c
= x A : I is dense
below x.
(2) idc(A) = [Idc(A)[.
(3) Did(A) = I : I a dense ideal of A.
(4) For ideals I, J I + J is the minimal ideal I of A which include I J i.e.
x y : x I and y J. Similarly

<0
I

.
(5) If we replace A by an ideal I (in 3.8 (1),(2),(3), 3.1 (2)) means we restrict
ourselves to subideals of it.
3.9 Claim. For a Boolean algebra A:
(1) id(A) idc(A) = [comp(A)[.
(2) [A[ idc(A) = idc(A)
0
(when A is innite, of course).
(3) If f Endu
5
(A), then f has a unique extension to an endomorphism f
+
of
comp(A) where f
+
(x) = sup
A
f(y) : y x, y Domf. If f is everywhere
onto it is the unique extension of f in End(compA).
(4) For g End(A), (f End
5
(A))f
+
g i g Endu(A).
(5) For f, g End
5
(A) we have f g f
+
= g
+
.
(6) For f End u
6
(A), letting a = sup
comp(A)
x : B[x Domf comp(A)
and b = sup
comp(A)
x : A[x Rangf comp(A), we dene f
+

HOM(comp(A[a), comp(A[b)) extending f by f(x) = supf(y) : y x, y

Domf, also f
+
is onto.
3
9
7


r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
3
-
1
0







m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
3
-
1
0


FACTOR = QUOTIENT, UNCOUNTABLE BOOLEAN ALGEBRAS
3.10 Claim.
(1) id(A) endl
3
(A).
(2) idc(A) endl

3
(A).
(3) endl
e
(A) = id(A) +endl

e
(A) for e = 3, 4, 5, 6.
(4) If f End
5
(A) then f
+
End
1
(compA).
(5) aut
e
(A) id(A) +aut

m
(A) when e, m 3, 5 or e, m 4, 6.
Proof.
(1) For I Id(A) let J
I
= x B : (B[x) I = 0, so J
I
is an ideal,
J
I
I = 0, J
I
I is dense. Let F
I
be the following map: Domf
I
=
I +J
I
, f
I
[I = id
I
, f
I
[J
I
= 0
JI
. Now I f
I
is a one to one mapping from
Id(A) to Endl
3
(A).
(2) The mapping above works.
(3) Note that id(A) end
3
(A) by part(1), and endl

3
(A) endl
3
(A) trivially.
A is innite hence all those cardinals are innite so =: id(A)+endl

3
(A)
endl
3
(A). The inverse inequality is easy too.
(4), (5) Left to the reader.
3.11 Claim.
(1) For x A:

f Aut(A)

[x ,= f(x)] i

f Aut
6
(A)

x Domf&x ,= f(x)

i (y, z)[0 < y x&0 < z

1 x&A[y

= A[z].
(2) If I

: < ) is a sequence of ideals of A ,[ ,= I

= 0] and
I =

i<
I
i
then:
id(I)
<
id(I

).
idc(I)
<
idc(I

).
aut(I)
<
aut(I

).
end(I)
<
end(J

).
Similarly for end
l
, aut
l
, end

l
, aut

l
etc.
There are many more easy relations, but for our aim the main point is
3.12 Main Lemma. For an innite Boolean Algebra A:
(1) aut

e
(A) for e = 3, 5 are equal or both nite (and we can restrict ourselves
to automorphisms of order 2).
(2) If for some e 3, 4, 5, 6 we have aut

e
(A) > idc(A) then aut

e
(A) for
e = 3, 4, 5, 6 are all equall.
(3) aut

3
(A) = aut

3
(A)
0
or aut
3
(A) is nite.
(4) autv

e
(A) = autv
3
(A) +idc(A) for e = 4, 6.
Proof. Let J =: x A : for every f Aut(A), f[(A[x) = id
A|x
.
The function F
1,
, F
5
satisfying y x F
e
(y) F
e
(x) are functions from A to
ord dened below; and we let:
K =: x A :
(i) for some y, x y = 0 and A[x

= A[y.
3
9
7


r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
3
-
1
0







m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
3
-
1
0


FACTOR = QUOTIENT, UNCOUNTABLE BOOLEAN ALGEBRAS
(ii) for e = 1, , 5 and 0 < y < x we have F
e
(y) = F
e
(y).
Where:
F
1
(x) = the cardinality of A[x.
F
2
(x) = idc(A[x).
F
3
(x) = aut(A[x).
F
4
(x) = aut
3
(A[x).
F
5
(x) = aut

3
(A[x).
Now
()
1
J is an ideal of B.
()
2
K is downward closed.
()
3
K J is dense.
Choose x
i
: i < maximal such that:
(a) x
i
K, x
i
> 0,
(b) if i ,= j, 0 < y

x
i
, 0 < y

x
j
then A[y

= A[y

.
Let K
i
= y : for some y

x
i
, A[y

= A[y

, and K
+
i
: the ideal K
i
generate.
Now
()
4

i<
K
i
is dense in K [hence J

i<
K
i
is dense in A].
()
5
For i ,= j we have K
i
K
j
= 0, K
i
J = 0.
Clearly,
()
6
for f Aut
6
(A) we have:
(a) f is the identity on J Domf.
(b) for x Domf we have x K
i
f(x) K
i
.
So
()
7
aut

e
(A) =
i<
aut

e
(K
+
i
) for e = 3, 5 and aut

e
(A) = idc(J)
i<
aut

e
(K
+
i
)
for e = 4, 6
We shall prove:
()
8
For each i, one of the following ocurrs:
(a) aut

e
(K
+
i
) for e = 3, 4, 5, 6 are all nite > 1,
(b) For some innite we have aut

e
(K
+
i
) = autv

e
(K
+
i
)

idc(K
i
) for e =
3, 4, 5, 6
(really we can use F
6
(i) = sup
+
: B[x has pairwise disjoint non zero
members and any such is OK for (b)).
Case 1. x
i
is an atom.
This is clear: let
i
=: [K
i
[, if it is innite, aut

e
(K
+
i
) = 2
i
for e = 3, 4, 5, 6 so
case (b) in ()
8
ocurrs.
If
i
is nite, 1 < aut

e
(K
+
i
) <
0
(we can compute exactly), so case (a) in ()
8
.
In fact in all cases we can use just automorphism of order 2.
So we can assume
Case 2 . not Case 1, so B[x
i
is atomless, hence idc(B[x
i
) 2
0
Let J
i,
, J

i,
: <
i
) be a sequence such that:
() J
i,
, J

i,
are ideals K
+
i
and ,= 0,
() 0 ,= J
i,
A[x
i
,
() J
i,

= J

i,
, an h
i,
an isomorphism from J
i,
onto J

i,
,
()
<<i
J

i,
J

i,
= 0,
3
9
7


r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
3
-
1
0







m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
3
-
1
0


FACTOR = QUOTIENT, UNCOUNTABLE BOOLEAN ALGEBRAS
() if y K
+

is disjoint to all members

i,
of then for some y

y and
z J
i,
, B[z

= B[y

hence
() if < <
i
then J
i,
J
i,
is a dense subset of J
i,
Now
()
9

<i
J

i,
is a dense subset of K
+
i
.
()
10

i
2 (as there is y, x
i
y = 0 and A[y

= A[x
i
as x
i
K).
()
11
idc(K
+
i
) =
<i
idc(J

i,
).
By the denition of K (see choice of F
2
), we have 0 < y x
i
idc(A[y) = (A[x)
hence idc(J

i,
) = idc(A[x
i
) so
()
12
idc(K
+
i
) = [idc(A[x
i
)]
|i|
hence [idc(K
+
i
)]
|i|
= idc(K
+
i
).
Easily
()
13
aut

6
(K
+
i
) aut

3
(K
+
i
) idc(A[x
i
).
(and even by automorphism of order 2).
Last inequality: for each z Idc(A[x
i
) there is z

, z

x
i
= 0, and A[z

= A[z

and
let g be such isomorphism, let I
z
be the ideal of A generated by x K
+
i
: x z
or x z

or x z = x z

= 0 g
z
Aut(I
z
) Aut
3
(K
+
3
), g
z
(y) =: (y z z

)
g(y z) g
1
(y z

).
Case 2A.
i
<
0
(and not Case 1)
Let for n < x
i,n
x
i
, x
i,n
,= 0, and [n ,= m x
i,n
x
i,m
= 0], and
I
i,
= y K
+
i
:
n
y x
i,n
= 0 and I
i,n
= z : z x
i,n
: So aut

3
(K
+
i
)
aut

3
(

I
i,
)
n<
aut

3
(B[x
i,n
) = aut

3
(B[x
i
)
0
.
Similarly to the proof of Case 2B below (but easier) we can show that Case (b)
of ()
8
ocurrs. From now on we assume
Case 2B. Not Case 2A, so
i

0
For every f Aut
6
(K
+
i
) let, for , <
i
:
L
f
,
= x J

i,
: f(x) J

i,
.
M
f
,
= f(x) : x J

i,
, f(x) J

i,
.
So the number of possible

L
f
= supL
f
,
: ,= <
i
) and

M
f
= sup(M
f
,
) :
,= < ) is [idc(A[x
i
)]
|i|
and for xed

L,

M the number of f aut

6
(K
+
i
)
for which

L
f
=

L,

M
f
=

M is
,
[f/ : f an isomorphism from a dense
subset of L
,
onto a dense subset of M
,
[
,
aut

3
(L
,
)
,
aut

3
(A[x
i
) =
[aut

3
(A[x
i
)]
|i|
).
(In the last equality we use F
5
in the denition of K: for the last , note we
can replace L
,
by isomosrphic ideal L

,
of A[x
i
, and letting L

,
= y : : y
A[x
i
, (z L

,
)[y z = 0] we can extend every f Aut
3
(L

(, ) to f

Aut
3
(A[x) by letting f

(x y) = f(x) y for x Dom(f), y L

,
)
So
()
14
aut

6
(K
+
i
) [aut

3
(A[x
i
) +idc(A[x
i
)]
|i|
.
Now for each such that 2 +1 <
i
we can choose y
i,2
J
i,2
, y
i,2+1
J
i,2+1
such that A[y
i,2

= A[y
i,2+1
and g
i,,
: < aut

3
(A[x
i
)) such that g
i,,
is an
isomorphism from a dense subset of A[y
i,,
onto a dense subset of A[y
i,2+1
, g
+
i,,
:
< aut

3
(A[x
i
)) pairwise distinct. Let

i
be minimal such that 2

i

i
.
Now for every sequence =

: <

i
),

< aut

3
(A[x
i
), we dene g
i,

Aut
3
(K
+
i
) of order two (see condition ()):
g
i,
[(y
i,2
y
i,2+1
) = h
2+1
g
i,,
h
1
2
h
2
g
1
i,
h
1
2+1
3
9
7


r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
3
-
1
0







m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
3
-
1
0


FACTOR = QUOTIENT, UNCOUNTABLE BOOLEAN ALGEBRAS
and if y K
+
i
0 and
<

i
(y y
i,
= 0) then g
i,
(y) = 0.
Lastly if y is the disjoint union of y
0
, , y
n
and each g

(y
e
) was dened then
we dene g

(y
0
y
n
) = g

(y
0
) g

(y
n
). The reader may check that
g

Aut
3
(K
+
i
).
The mapping g
i,
show:
()
15
aut

3
(K
+
i
) aut

3
(A[x
i
)
|i|
.
Easily, choosing y
i
J
i,1
(possible as
i
2)
()
16
aut

3
(K
+
i
) aut

3
(A[(x
i
y
i
)) idc(A[y) = idc(A[x
i
)
0
.
Together by ()
13
, ()
14
, ()
15
, ()
16
for e = 3, 6:
aut

e
(K
+
i
) = aut

3
(A[x
i
)
|i|
.
By 3.7(1) this holds for e = 4, 5, hence ()
8
has been proved.
Now by ()
7
+ ()
8
the four parts of 3.12 follows.
3.13 Conclusion : aut(A)
0
sub(A), also aut
e
(A)
0
sub(A) for e =
3, 4, 5, 6.
Note: even if aut(A) is nite, A innite, still 2
0
sub(A) (for A innite).
Proof. By 3.7(3) + 3.7(6) aut(A) = aut

0
(A) id(A) + aut

3
(A) id(A) +
Pend(A) sub(A) but by S. Shelah [6] id(A)
0
= id(A) and by 3.12(3) aut

3
(A)
0
=
aut

3
(A), together we can nish the rst inequality, the second is similar using
3.10(5).
3.14 Claim : For an (ininite) Boolean algebra A we have:
(1) end
e
(A) = id(A) +aut
3
(A) for e = 3, 4, 5, 6.
(2) end
6
(A)
0
sub(A).
(3) end
e
(A)
0
sub(A) for e = 2, 3, 4, 5, 6.
Proof.
(1) Clearly end
3
(A) aut
3
(a) and end
3
(A) id(A) (as the mapping in the
proof of 3.10(1) examplify).
On the other hand we can attach to every f End
6
(A) three ideals
I
1
(f) = Domf, I
2
(f) = Rang(f) and I
3
(f) = Ker(f). Now the number of
triples

I of ideal of A has cardinality id(A) and for each such

I:
f End
6
(A) : I
e
(f) = I
e
for e = 1, 2, 3 has cardinality [Aut(I
2
)[
which is aut
3
(A). By 3.7(2) we can nish.
(2) Remember also id(A)
0
= id(A) by [6] and part (1) and 3.13.
(3) By part (2) and 3.7.
4 The width of the Boolean algebra .
4.1 Denition. For a Boolean algebra B let: (1) A B is an antichain if
x A&y A&x ,= y x , y (i.e. A is a set of pairwise incomparable elements).
(2) Width of B, w(B) is sup[A[ : A B antichain, w
+
(B) = [A[
+
: A B
antichain
E. C. Milner and M. Poizat [3], answering a question of E. K. van Dowen, D.
Monk and M. Rubin [7] proved cf(w
+
(B)) ,=
0
.
In S. Shelah [5] we claim: if > cf >
0
, for some generic extension of
the universe preserving cardinalities and conalities, for some B, w
+
(B) = . We
retract this and replace it by the theorem 4.2 below.
3
9
7


r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
3
-
1
0







m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
3
-
1
0


FACTOR = QUOTIENT, UNCOUNTABLE BOOLEAN ALGEBRAS
For weakly inaccessibles we still have the consistency. Moreover, if is a limit
uncountable regular cardinal, S stationary not reecting and
S
it then we
have such an example for .
4.2 Theorem. For an innite Boolean algebra B, w
+
(B) is an uncountable
regular cardinal.
Proof. As B is innite it has an antichain A, [A[ =
0
, [if B has nitely many
atoms clear, if not it has a subalgebra which is atomless, without loss of generality
countable and check]. So =: w
+
(B) >
0
. Assume =: cf < ; let =

i<

i
, cf+

j<i

j
<
i
< and let A
i
B be an antichain of cardinality
+
i
( exist by the choice of ). Let A =

i<
A
i
, so [A[ = . Choose such A
i
: i < )
such that, if possible
(*)i < j < , x A
i
, y A
j
y , x.
For x B let A[>, x] = y A : y > x, A[<, x] = y A : y < x,
A[>, ] = x A : [A[>, x][ < , A[<, ] = x A : [A[<, x][ < .
Case 1. For some < , A[>, ] has cardinality .
By Hajnal free subset theorem, there is a set E A[>, ] of cardinality such
that:
x ,= y &x E &y E x , A[>, y] &y , A[>, x]. So E witness w
+
(B) > .
Case 2. For some < , A[<, ] has cardinality .
Same proof.
Case 3. For every i < there is x A such that
i
< [A[>, x][ < .
Let for i < , x
i
A be such that
i
< [A[>, x
i
][ < . Let u be such
that: = supu and for i u,
i
>

jui
[A[>, x
i
][ (choose the members of u
inductively). By renaming without loss of generality u = . Clearly A[>, x
i
]
j<i
A[>, x
j
] has cardinality >
i
.
As
i
> (by its choice) and A =
j<
A
j
, clearly for each i there is (i) <
such that (A[>, x
i
]
j<i
A[>, x
j
])A
(i)
has cardinality >
i
; necessarily (i) i.
For some unbounded u we have [i u&j u&i < j (i) < j]; without
loss of generality u = ,
i
= i. Let A

i
be a subset of (A[>, x
i
]
j<i
A[<, x
j
])A
(i)
of cardinality
+
i
. Now A

i
: i < ) satises: A

i
B is an antichain of cardinality

+
i
and
()

i < j, x A

i
, y A

j
x , y (otherwise x
i
x y , A[>, x
i
],
contradiction).
So A

i
=: 1
B
x : x A

i
: i < ) satises A

i
B is an antichain of B of
cardinality
+
i
and also (*) above (check). So by the choice of A
i
: i < ), it
satises (*). By () + ()

, A

=
i<
A

i
is an antichain of B of cardinality , so
w
+
(B) > .
Case 4. For every i < x there is x A such that
i
< [A[<, x][ < .
Similar to Case 3.
Case 5. None of the previous cases.
By not Case 3 for some i() < , for no x A is
i()
< [A[>, x][ < . By not
Case 2, A[<,
+
i()
] has cardinality < . By not Case 1 A[>,
+
i()
] has cardinality
< .
Choose x

AA[<,
+
i()
]A[>,
+
i()
] so A[>, x

] has cardinality
+
i()
>
i()
,
hence by the choice of i() we have A[>, x

] has cardinality .
3
9
7


r
e
v
i
s
i
o
n
:
1
9
9
6
-
0
3
-
1
0







m
o
d
i
f
i
e
d
:
1
9
9
6
-
0
3
-
1
0


FACTOR = QUOTIENT, UNCOUNTABLE BOOLEAN ALGEBRAS
As
i()
> , for some j(), A[<, x

] A
j()
has cardinality >
i()
, so choose
distinct y
i
A[<, x

] A
j()
for i < . Now y
i
< x

(as y
i
A[<, x

]), and
[i ,= j y
i
, y
j
] (as they are distinct and in A
j()
).
Let A

i
= A
i
A[>, x

], so A

i
A
i
hence is an antichain of B, and

i
A

i
= (A
i
) A[>, x

] = A A[>, x

] = A[>, x

]
So each A

i
is an antichain, its member are > x

and [
i
A

i
[ is as [A[>, x

][ is.
Now
A

=
i<
y
i
(x x

) : x A

is an antichain of B of cardinality , so w
+
(B) > , as required.
References
1. M. Bekkali, R. Bonnet and M. Rubin, Spaces for which every closed subspace is homeomorphic
to a closen subspace, to appear Order.
2. R. Bonnet and S. Shelah, On HCO spaces. An uncountable compact T
2
space dierent from

1
+ 1 which is homeomorphic to every of its uncountable closed subspaces, to appear Israel
J. Math.
3. E. C. Milner and M. Pouzet, On the width of ordered sets and Boolean Algebras, Algebra
Universatis 23 (1986), 242253.
4. J. D. Monk, Cardinal Functions on Boolean Algebra, Lectures in Mathematics, ETH Zurich,
Birkhause Verlag, Basel, Boston , Berlin, 1990.
5. S. Shelah, Construction of many complicated uncountable structures and Boolean Algebras,
Israel J. Math. 23 (1983), 100146, [Sh 136].
6. S. Shelah, Remarks on the number of ideals of Boolean algebras and open sets of a topology,
Springer-
Verlag Lecture Notes 1182 (1982), 151187, [Sh 233].
7. E. K. van Dowen, D. Monk and M. Rubin, Some questions about Boolean Algebras, Algebra
Universatis 11 (1980), 220243.
The Hebrew University of Jerusalem, Israel, Rutgers University, New Brunswick,
NJ USA, MSRI, Berkeley, Calif, U.S.A.