150 (1996)
On automorphisms of Boolean algebras embedded in P (ω)/fin
by
Magdalena G r z e c h (Warszawa)
Abstract. We prove that, under CH, for each Boolean algebra A of cardinality at most
the continuum there is an embedding of A into P (ω)/fin such that each automorphism of A can be extended to an automorphism of P (ω)/fin. We also describe a model of ZFC + MA(σ-linked) in which the continuum is arbitrarily large and the above assertion holds true.
It is well known that, under CH (the continuum hypothesis), each Boolean algebra of cardinality at most 2
ωcan be embedded in P (ω)/fin (see e.g. [5]). This implication cannot be reversed: there is a model of set theory in which 2
ω> ω
1and the above conclusion still holds ([1]). It is also known that CH is equivalent to the following condition: each Paroviˇcenko algebra (i.e. algebra of cardinality 2
ω, atomless and having neither countable limits nor countable unfilled gaps) is isomorphic to P (ω)/fin. We begin by proving the following.
Proposition 1. If CH holds, then for every Boolean algebra A of car- dinality at most the continuum there is an embedding i : A → P (ω)/fin such that each automorphism of i(A) can be extended to an automorphism of P (ω)/fin.
P r o o f. Assume CH. Let A be a Boolean algebra of cardinality at most the continuum. We will construct an extension A
?of A such that:
1. A
?is a Paroviˇcenko algebra;
2. S
α<ω1
A
α= A
?, where (A
α: α < ω
1) is an increasing sequence of algebras satisfying the following two conditions:
(?) card A
α≤ 2
ω, A
0= A, A
λ= S
α<λ
A
αfor every limit λ < ω
1,
1991 Mathematics Subject Classification: Primary 03E50; Secondary 03E05, 06E05.
This paper is the author’s PhD thesis presented to the Institute of Mathematics, Polish Academy of Sciences in November 1995.
[127]
(??) for every automorphism φ
αof A
αthere exists an automorphism φ
α+1of A
α+1such that φ
α⊆ φ
α+1.
It is clear that if an algebra A
?satisfies the above conditions then each automorphism of A can be extended to an automorphism of A
?. Thus to prove our theorem it suffices to construct A
?.
Fix a pairing k : ω
1×ω
1→ ω
1(one-to-one and onto) such that k(ζ, ξ) ≥ ζ for all ζ, ξ < ω
1. It remains to describe the successor step from α to α + 1.
Suppose that we have defined a sequence (A
γ: γ ≤ α) satisfying (inductive) conditions (?) and (??).
Let E
γ(φ, a, c) abbreviate the statement: φ is an automorphism of A
γsuch that φ(a) = c.
Assume that at each stage γ ≤ α we chose an enumeration (x
γξ: ξ < ω
1) of the collection of the following families:
{(c
i, d
j: i, j < ω) : ∃φ ∀i, j < ω [E
γ(φ, a
i, c
i) ∧ E
γ(φ, b
i, d
i)]},
where (a
i, b
j: i, j < ω) is a countable ordered gap a
0< a
1< . . . < b
1< b
0of elements of A
γ, {(b
i: i < ω) : ∃φ ∀i < ω E
γ(φ, a
i, b
i)}, where (a
i: i < ω) is a decreasing chain of elements of A
γ, and the set of all atoms of the algebra A
γ. (Since we assumed CH, we have at most ω
1objects to enumerate.)
We identify A
αwith the field B(X
α) of open-closed subsets of the associa- ted Stone space X
α. The ordinal α determines a certain object, namely x
ζξ, where ξ and ζ are ordinals such that k(ζ, ξ) = α (ζ < α). If x
ζξis a family of chains or gaps, we take A
α+1to be the subfield of P (X
α) generated by A
α= B(X
α) and by
n b = \
i<ω
b
i: (a
j, b
i: i, j < ω) ∈ x
ζξo
when x
ζξis a collection of gaps, or by n
b = \
i<ω
b
i: (b
i: i < ω) ∈ x
ζξo
when x
ζξconsists of countable chains. Using the Sikorski theorem (on exten- ding homomorphisms, see e.g. [7], [5]) we extend each automorphism of A
αto an automorphism of A
α+1and therefore (?), (??) hold.
Now, suppose that x
ζξis a set of nonzero elements of A
α. Each element
of the family is an atom of A
ζbut it need not remain an atom in A
α. If
there are at least countably many elements e
i< a, a ∈ x
ζξ, then we put
A
α+1= A
α. (Note that the property (??) implies that if some element of
x
ζξis an atom then all elements of the set are atoms.) Suppose that each
a ∈ x
ζξis a finite sum of atoms a = e
1+ . . . + e
n. (n is the same for all
elements of x
ζξby (??).) Atoms of A
αcorrespond to isolated points of X
α. We delete the isolated points e
i< a for all a ∈ x
ζξ, and put into their places copies of a one-point compactification of the discrete ω. Let X
α+1denote the topological space thus obtained. We put A
α+1= B(X
α+1). Since X
αis a continuous image of X
α+1, we have A
α⊆ A
α+1. Obviously, each automorphism of A
αcan be extended to an automorphism of A
α+1. This finishes the proof.
Now we consider the case of ¬CH. It is known that there exists a model of ZFC+MA+¬CH in which the algebra P (ω
1) is not embeddable in P (ω)/fin ([2]). On the other hand, it is consistent with ZFC and MA(σ-linked) that the cardinality of the continuum is arbitrarily large and each Boolean algebra of cardinality ≤ 2
ωcan be embedded in P (ω)/fin ([1]). Thus the existence of such embeddings does not imply CH. The assertion of Proposition 1 is stronger and we may ask if the converse holds. The answer is negative. We prove that:
Theorem 1. It is consistent with ZFC + MA(σ-linked) that the cardina- lity of the continuum is arbitrarily large and for each Boolean algebra B of cardinality ≤ 2
ω, there is an embedding i : B → P (ω)/fin such that each automorphism of i(B) can be extended to an automorphism of P (ω)/fin.
P r o o f. Let V be a ground model satisfying the generalized continuum hypothesis (GCH). Thus there exists a regular cardinal κ in V such that κ > ω
1and if κ = λ
+, then cf(λ) > ω, moreover ♦
κ(the diamond principle) holds in the form:
There is a sequence (T
α: α < κ, cf(α) = ω
1) such that for every set X ⊆ H(κ) the set {α < κ : cf(α) = ω
1, X ∩ H
α= T
α} is stationary in κ.
H(κ) denotes as usual the family of all sets of hereditary power < κ, H(κ) = S
α<κ
H
α, and (H
α: α < κ) is a continuously increasing sequence of sets of cardinality < κ.
We will define a finite support iteration (P
α: α < κ) having the c.c.c.
(countable chain condition) such that, in the corresponding generic exten-
sion V[G], the conclusion of Theorem 1 will be satisfied. The model V will
be extended in such a way that given a system of generators for a certain
Boolean algebra B (card B < 2
ω) there will be an embedding sending the
generators to generic sets added at some steps α < κ. The embedding will
be defined by induction: If a certain monomorphism embeds the subalgebra
B
0of B generated by the initial α (α < 2
ω) generators and if the image
(under the monomorphism) of each of them is a generic subset of P (ω)/fin
then the next generator determines in B
0two sets (which form a gap): one
consists of elements less than the generator (called the “lower class”), and
the elements of the other (called the “upper class”) are disjoint from the generator. An image of the gap is a gap in the subalgebra of P (ω)/fin which is generated by some generic sets. The monomorphism can be extended if there is an element of P (ω)/fin which fills this gap. Thus, to ensure embed- dability of Boolean algebras, we will add generic sets X
α⊆ ω filling gaps in subalgebras of P (ω)/fin generated by some previously added X
β, β < α.
Simultaneously, in a similar way, we will extend automorphisms of the sub- algebras. In constructing embeddings of Boolean algebras and extensions of their automorphisms, we have to avoid the following problem: It is well known that there are gaps in P (ω)/fin which are unfillable by c.c.c. forcing.
It could happen, unless steps are taken to prevent it, that an image (under an extension of an automorphism of one of the embedded algebras) of some gap filled in a later step is an unfillable gap.
To ensure that every automorphism of an embedded Boolean algebra can be extended we use the ♦ principle. It guarantees that each such automor- phism is “approximated” by an increasing sequence of automorphisms which belong to models V[G|α]. To be more precise: if F is a canonical P
κ-name for some automorphism f of a given algebra B (from the model V[G]) then there is a subset A of κ such that
card A = κ and [
α∈A
(F ∩ H
α) = [
α∈A
T
αand T
αis a P
α-name for an automorphism from V[G|α]. We will extend automorphisms using those of the T
α’s which are their names. To obtain MA(σ-linked) we will enumerate at some stages (with repetition) all σ-linked forcings R with card R < κ (cf. [1], [4]).
Assume the following notation:
Let X be a set and let T
ξdenote a homomorphism. Then for ε ∈ {−1, 1}, εX denotes X, if ε = 1, or \X, if ε = −1. Moreover, T
ξεis T
ξ, if ε = 1, or T
ξ−1, if ε = −1. (We abbreviate (T
ξε)
nto T
ξεn.)
For ϕ : α → {0, 1} let B(ϕ) be the subalgebra generated by {X
β: ϕ(β) = 1}, where X
βis a generic subset of ω added at stage β. If s is a finite sequence with dom(s) ⊆ {β : ϕ(β) = 1} and rg (s) ⊆ {−1, 1} then
X(s) = \
s(ξ)=1
X
ξ∩ \
s(ζ)=−1
(ω \ X
ζ).
Thus B(ϕ) consists of finite unions of sets of the form X(s). A gap in B(ϕ) is a system of the form
L = ({X(s) : s ∈ S}, {X(t) : t ∈ T }),
where X(s) ∩ X(t) =
?∅ for all s ∈ S and t ∈ T .
An increasingly ordered gap L of type (λ, γ) is a gap as above such that there are enumerations S = {s
α: α < λ} and T = {t
β: β < γ} satisfying
α
1< α
2< λ ⇒ X(s
α1) ⊆
?X(s
α2), β
1< β
2< γ ⇒ X(t
β1) ⊆
?X(t
β2).
We assume that each gap except the increasingly ordered ones satisfies the condition: if s
1, . . . , s
n∈ S and X(s) ⊆
?X(s
1) ∪ . . . ∪ X(s
n) then s ∈ S (and similarly for T ).
We will use two notions of forcing: Kunen’s forcing filling a gap, and the other, which adds an uncountable antichain to Kunen’s forcing of type (ω
1, ω
1).
Now we describe the two forcings:
Let L = ({X(s) : s ∈ S}, {X(t) : t ∈ T }) be a gap. Kunen’s forcing Q(L) consists of elements of the form (u
q, x
q, w
q), where u
qand w
qare finite subsets of S and T (respectively) and x
qis a finite zero-one sequence.
Moreover,
[
s∈uq
X(s) ∩ [
t∈wq
X(t) ⊆ dom(x
q).
Let p = (u
p, x
p, w
p) and q = (u
q, x
q, w
q); then p is an extension of q (written p ≤ q) iff u
q⊆ u
p, w
q⊆ w
p, x
q⊆ x
pand for each i with dom(x
q) ≤ i <
dom(x
p), if i ∈ [
s∈uq
X(s) then x
p(i) = 1 and if i ∈ [
t∈wq
X(t) then x
p(i) = 0.
It is known that if L is separated, then Q(L) has the c.c.c.
Now let L = ({X(s
α) : α < ω
1}, {X(t
β) : β < ω
1}) be an increasingly ordered gap. A condition of forcing E(L) is a finite set e consisting of se- quences of the type (α, s
α, t
α) such that if (α, s
α, t
α), (β, s
β, t
β) ∈ e and α 6= β then either X(s
α) ∩ X(t
β) 6= ∅ or X(s
β) ∩ X(t
α) 6= ∅. E(L) is ordered by inverse inclusion. It is well known that if L is an unfilled gap then E(L) has the c.c.c. and
E(L) ° “Q(L) has an uncountable antichain”.
The definition of the iteration is inductive and uses a “bookkeeping”
technique. At each inductive step α < κ we enumerate some objects in V
(Pα), and at higher stages we add some generic sets to them. The objects occur in an order determined by a function Nb. To be more precise, we divide κ into five unbounded sets:
A = {α < κ : cf(α) = ω
1}, M = {α ∈ κ \ A : α is odd},
E = k(A), Q
1= k(M ), Q
2= k(κ \ (A ∪ M )),
where k : κ → κ \ (A ∪ M ) is an increasing bijection.
Let {ν
α: α < κ} be an increasing enumeration of the set {β < κ : β ≥ λ}, if κ = λ
+, or of the set {β < κ : β is a cardinal and cf(β) > ω}, if κ is a limit cardinal.
Let n : κ × κ → κ be a pairing function satisfying:
ξ, ζ < n(ξ, ζ) for all ξ, ζ < κ, n(α, β) ∈ M for all α ∈ M, β < κ, n(α, β) ∈ Q
1for all α ∈ Q
1, β < κ, n(α, β
1) < n(α, β
2) for all α ∈ A, β
1< β
2< κ, n(α, β) ∈ Q
2for all α ∈ A, β < ν
α, n(α, β) ∈ E for all α ∈ A, β > ν
α,
n(α
1, β
1) < n(α
2, β) for β
1< ν
α1, α
1< α
2, β < κ.
Using this function we will define (by induction) a function Nb. At stages ξ ∈ M , we will add generic filters to σ-linked forcings. In steps ξ ∈ Q
1, we add (by Kunen’s forcing) the generic set X
ξwhich fills a gap consisting of some sets previously added (in Q
1steps). In the model V[G], each Boolean algebra will be embedded in a certain algebra generated by sets obtained in these steps. At stage ξ ∈ Q
2we also add (by the same forcing) the generic set X
ξwhich separates a gap, but this gap is generated by sets previously added both in Q
1and Q
2steps. In the model V[G] each of these sets X
ξ, ξ ∈ Q
2, will be an image (under one of the extended automorphisms) of some element of P (ω)/fin which appeared in some model V[G|δ], δ < ξ. In steps ξ ∈ E we will add uncountable antichains to Kunen’s forcing to keep gaps in the ranges (of the extended automorphisms) unfilled.
The sequence in which new elements of P (ω)/fin appear is important in our construction. It will be described by the function Ind from P (ω)/fin into κ, defined inductively simultaneously with iteration. We begin with the condition: if x ∈ P (ω)/fin ∩ V then Ind(x) = 0. At each higher stage we extend the function Ind according to the rule:
If P
α+1° “x 6∈ dom(Ind) and x ∈ P (ω)/fin” then Ind(x) = α + 1.
If ξ < κ, cf(ξ) = ω
1and P
ξ° “T
ξis an automorphism of B(ϕ)” (T
ξis an element of the ♦-sequence), then we begin to define (inductively) families of monomorphisms according to the following conditions:
(a) T
ξξ= T
ξ. (b) For γ ≥ ξ,
P
γ° “T
ξγis a monomorphism from a subalgebra of P (ω)/fin into P (ω)/fin”.
(c) If γ
1< γ
2then T
ξγ2is an extension of T
ξγ1.
(d) If γ
1≤ γ
2, 0 < ξ
i≤ γ
i, P
ξi° “T
ξiis an automorphism of B(ϕ
i)”
(i = 1, 2) and
P
max(ξ1,ξ2)° “For some ordinal %, ϕ
1¹ % = ϕ
2¹ % and
T
ξ1and T
ξ2agree on B(ϕ
1¹ %)”, then
T
ξγ11¹ {X ∈ P (ω)/fin : Ind(X) < %} ∩ dom(T
ξγ11
)
= T
ξγ22
¹ {X ∈ P (ω)/fin : Ind(X) < %} ∩ dom(T
ξγ22
).
(e) If λ ≤ α is a limit ordinal then T
ξλ= S
γ<λ
T
ξγ.
At each stage we will compute card P (ω)/fin using the following two (well known) theorems (see e.g. [5], [6]):
Theorem 2. Assume that P has the c.c.c. in V and let ν be a cardinal in V such that V ° “card P ≤ ν, ν
ω= ν”. Let Q be such that P ° “card Q ≤ ν”.
Then card P ? Q ≤ ν in V.
Theorem 3. Assume that P has the c.c.c. in V and λ, ν ≥ ω are cardi- nals in V such that V °“card P ≤ ν and λ = ν
ω”. Let G be P-generic over V. Then 2
ω≤ λ in V[G].
Thus we have to show that for each α, P
αhas the c.c.c. We will do that in the second part of the proof; now we assume that it is true.
We describe the inductive step α ⇒ α + 1. Assume that the forcing P
αand families of monomorphisms T
ξγ(ξ ≤ γ ≤ α) satisfying the above conditions (a)–(e) are already defined. Assume also that card P
α≤ ν
αand P
α° “2
ω≤ ν
α”. Since the cardinality of each of the forcings occurring in Cases 1 to 5 below is ≤ ν
α, by Theorems 1 and 2 we have card P
α+1≤ ν
α+1and P
α° “2
ω≤ ν
α+1”.
We distinguish five cases. In Cases 1, 2 and 5 we set T
ξα+1= T
ξα. C a s e 1: α ∈ M . We enumerate all P
α-names of σ-linked forcings of cardinality < κ so that each forcing occurs κ times in the enumeration and the following holds:
If R is ξth element of the enumeration then P
α° “card R ≤ ξ”.
We extend the function Nb: if
P
α° “R is σ-linked and card R < β ”
and R is the βth element of the above enumeration then Nb(R) = n(α, β).
If there are γ < α and β < κ such that α = n(γ, β), then we put P
α+1= P
α? R,
where P
γ° “R is σ-linked and card R < κ” and Nb(R) = α.
C a s e 2: α ∈ Q
1. In this case we enumerate all pairs (L, ϕ) of P
α-names such that P
αforces the following properties:
(a) ϕ ∈ D
α, where D
αconsists of all ψ ∈ V
Pαwith dom(ψ) ≤ α, rg(ψ) ⊆ {0, 1} and Γ = {γ : ψ(γ) = 1} ⊆ Q
1, and such that if {γ
ξ: ξ < δ}
is an increasing enumeration of Γ , then for each ξ < δ there is a gap L
ξin B(ψ ¹ γ
ξ+ 1) satisfying γ
ξ+1= Nb(L
ξ, ψ ¹ γ
ξ+ 1).
(b) L is a gap in B(ϕ).
Each of these pairs occurs κ times in the enumeration.
If there are γ < α and β < κ such that n(γ, β) = α then we put P
α+1= P
α? Q(L),
where P
γ° “L is a gap in B(ϕ)” for some ϕ ∈ D
γ. C a s e 3: cf(α) = ω
1. If
P
α° “For some γ < α and ϕ ∈ D
γ, T
αis an automorphism of B(ϕ)”
then two cases are possible:
(?) P
α° “There is no ξ < α such that T
ξis an automorphism of B(ψ) and T
αand T
ξagree on B(ψ
ξ), where ψ
ξ= ϕ ¹ %
ξ= ψ ¹ %
ξfor some ordinal
%
ξ≤ ξ”, and
(??) P
α° “There are ordinals ξ, %
ξand a function ψ ∈ D
ξsuch that
%
ξ≤ ξ < α, cf(ξ) = ω
1, ψ
ξ= ϕ ¹ %
ξ= ψ ¹ %
ξ, T
ξis an automorphism of B(ψ) and T
αis an extension of T
ξ¹ B(ψ
ξ)”.
Let Υ denote the set of all pairs (%
ξ, ξ) such that P
αforces that T
ξis an automorphism of B(ψ), ψ
ξ= ϕ ¹ %
ξ= ψ ¹ %
ξand T
αis an extension of T
ξ¹ B(ψ
ξ). Let ζ = sup{%
ξ: (%
ξ, ξ) ∈ Υ }. We enumerate all triples (X, T
α, εn) of P
α-names such that
P
α° “X ∈ P (ω)/fin”,
and Ind(X) < dom(ϕ) (case (?)), or ζ ≤ Ind(X) < dom(ϕ) (case (??)), ε ∈ {−1, 1}, n ∈ ω. We fix a function j from the set of these triples into κ with the following properties:
(a) If X ∈ B
dom(ϕ)= {X
γ: γ ∈ Q
1, γ < dom(ϕ)} and Y 6∈ B
dom(ϕ)then
j((X, T
α, εn)) < j((Y, T
α, εm)) for all n, m < ω.
(b) If X
1, X
2∈ B
dom(ϕ)[resp. Y
1, Y
26∈ B
dom(ϕ)] and Ind(X
1) < Ind(X
2) [resp. Ind(Y
1) < Ind(Y
2)] then
j((X
1, T
α, εn)) < j((X
2, T
α, εm)) [resp. j((Y
1, T
α, εn)) < j((Y
2, T
α, εm))].
(c) For all X with Ind(X) < dom(ϕ) and all n ∈ ω,
j((X, T
α, −n)) = j((X, T
α, n)) + 1,
j((X, T
α, n + 1)) = j((X, T
α, −n)) + 1.
Since P
α° “card P (ω)/fin ≤ ν
α”, the domain of the sequence of the triples is ≤ ν
α.
Using ordinals > ν
αwe also enumerate all triples (L, T
α, εn) of P
α-names such that P
αforces: L = ({X(s
γ) : γ < ω
1}, {X(t
β) : β < ω
1}) is an increasingly ordered gap of the type (ω
1, ω
1), Ind(X(s
γ)) ≤ dom(ϕ) and Ind(Y (t
β)) ≤ dom(ϕ) for all γ < ω
1and β < ω
1, and Q(L) does not have the c.c.c. We can assume that (L, T
α, −n) follows (L, T
α, n) and precedes (L, T
α, n + 1).
We extend the function Nb to the set of objects described above in the following way:
Nb((X, T
α, εn)) = n(α, j(X, T
α, εn))
and if (L, T
α, n) is the βth element of the (second) enumeration then Nb((L, T
α, n)) = n(α, β).
We set
P
α+1= P
α.
We also define T
αα= T
αα+1= T
αin case (?) and T
αα= T
αα+1= monomor- phism generated by T
αand S
(%ξ,ξ)∈Υ
T
ξα¹ {X ∈ P (ω)/fin : Ind(X) < %
ξ} in case (??). The families T
γαdefined at earlier stages are not changed:
T
γα+1= T
γα.
It is easy to check by using Sikorski’s theorem and the following lemma that the above definitions are correct.
Lemma 1. Let X be an element of P (ω)/fin in V, let p = (u
p, x
p, w
p) ∈ Q(L) and let X
γstand for a generic subset added by Q. Then we have:
if p ° “X ⊆
?X
γ” then X ⊆
?[
s∈up
X(s),
if p ° “X ∩ X
γ=
?∅” then X ⊆
?[
t∈wp
X(t).
C a s e 4: α ∈ Q
2. Suppose that α = Nb((X, T
γ, εn)), where P
γ° “T
γis an automorphism of B(ϕ), X 6∈ B(ϕ)
00.
If ε = 1 and X 6∈ dom(T
ξα) or ε = −1 and X 6∈ rg(T
ξα) then we extend the monomorphism T
ξα.
Suppose that ε = 1. Let L be a gap in rg(T
ξα) defined by X:
L = ({(T
ξα)
εn(Z) : Z ⊆
?X}, {(T
γα)
εn(Y ) : X ⊆
?Y })
(P
αforces all the properties). All elements of the gap have been defined at the previous stages, because of the definition of j (Case 3). We set
P
α+1= P
α? Q(L).
We extend T
ξαsetting
T
γα+1= homomorphism generated by T
ξα∪ {((T
ξα)
εn(X), X
α+1)}, where
X
α+1= {i ∈ ω : ∃p ∈ G [x
p(i) = 1]}, and G ⊆ Q(L) is a generic filter. If T
ξis a P
α-name such that P
α° “T
ξis an automorphism of B(ψ) and
for some ordinal %, ϕ ¹ % = ψ ¹ % and T
ξand T
γagree on B(ϕ ¹ %)”, and Ind(X) < dom(B(ψ)), then
T
ξα+1= homomorphism generated by T
ξα∪ {((T
ξα)
εn(X), X
α+1)}, and T
ζα+1= T
ζαin the remaining cases. If ε = −1 we proceed with the construction in a similar way: we add a generic set to the domain of T
ξαand to the domains of each of the T
γα’s which agree with T
ξαon an “initial segment” of their domains.
(It is easy to prove, by using Lemma 1 and Sikorski’s theorem, that the definitions of the monomorphism T
γαare correct.)
C a s e 5: α ∈ E. Assume that α = Nb((L, T
γ, εn)), where P
γ° “L is an increasingly ordered gap in P (ω)/fin and Q(L) does not have the c.c.c.”
Suppose that L = ({X(s
ζ) : ζ < ω
1}, {X(t
β) : β < ω
1}) and let L
?denote the gap
({(T
γα)
εn(X(s
ζ)) : ζ < ω
1}, {(T
γα)
εn(X(t
β)) : β < ω
1}).
We set
P
α+1= P
α? E(L
?) and T
γα+1= T
γα.
For limit ordinals λ < κ we define P
λas a direct limit of {P
α: α < λ}.
We also assume P
α+1= P
αin all cases not mentioned above. This completes the definition of the iteration.
We conclude this part of the proof by checking that the above construc- tion is correct, i.e. that there is no gap L of the type (ω
1, ω
1), consisting of generic subsets of ω, which is an image (under one of the extending monomorphisms) of some gap L
0such that Q(L
0) does not have the c.c.c.
and which is filled by the generic set X
γat some stage γ < κ.
Claim 1. Let B
i(i = 1, 2) denote one of the following subalgebras of P (ω)/fin : B(ϕ
i) (where ϕ
i∈ D
ηi); the domain of T
γα; the range of T
ξα. Assume that S
ni=1
X(s
i) ∈ B
1, S
mj=1
X(t
j) ∈ B
2and S
ni=1
X(s
i) ⊆
?S
mj=1
X(t
j). Then there are finite functions r
1, . . . , r
ksuch that rg(r
l) ⊆ {−1, 1} (l = 1, . . . , k) and for each ξ ∈ S
kl=1
dom(r
l) we have X
ξ∈ B
1∩ B
2and
[
n i=1X(s
i) ⊆
?[
k l=1X(r
l) ⊆
?[
m j=1X(t
j).
This is proved by using Lemma 1.
Lemma 2. Let α < κ. Assume that P
αhas the c.c.c. Suppose that P
αforces the following:
(1) For each β < α and each ϕ ∈ D
β, if L
ϕis a gap in B(ϕ), then Q(L
ϕ) has the c.c.c.
(2) If L
ζ= (S
ζ, U
ζ) is a gap in the domain or range of (T
ζα)
εksuch that P
α° “ Q(L) has the c.c.c.” and for all X(s) ∈ S
ζand X(t) ∈ U
ζthere are S
ni=1
X(s
i) ∈ S
ζand S
mj=1
X(t
j) ∈ U
ζsuch that X(s) =
[
n i=1X(s
i) ∧ X(t) = [
m j=1X(t
j)
∧ Ind
[
ni=1
X(s
i)
< Ind
(T
ζα)
εk[
ni=1
X(s
i)
∧ Ind
[
mj=1
X(t
j)
< Ind
(T
ζα)
εk[
mj=1
X(t
j)
then P
α° “Q((T
ζα)
εk(L)) has the c.c.c.”
(3) L
ξ= ({(T
ξα)
ε1m(X(s
η)) : X(s
η) ∈ S}, {(T
ξα)
ε1m(X(t
η)) : X(t
η) ∈ U}) is an increasingly ordered gap such that P
ξ° “T
ξis an automorphism of B(ψ), L = (S, U) is an increasingly ordered gap of the type (ω
1, ω
1) in P (ω)/fin and Q(L) does not have the c.c.c.”
Under the above assumptions, there is an α
0< ω
1such that for each γ > α
0and any finite subsets {X(s
1), . . . , X(s
n)}, {X(t
1), . . . , X(t
m)} of the lower and upper classes (respectively) of the gap L
ϕor L
ζthe following holds:
X(s
γ) 6⊆
?[
n i=1X(s
i) or X(t
γ) 6⊆
?[
m j=1X(t
j)
where X(s
α0), X(s
γ) and X(t
α0), X(t
γ) are elements of the lower and upper classes of L
ξ(respectively).
P r o o f. Assume to the contrary that for every γ ∈ ω
1there are s
γ1, . . . . . . , s
γn, t
γ1, . . . , t
γmsuch that
h
X(s
γ) ⊆
?[
n i=1X(s
γi) i
∧ h
X(t
γ) ⊆
?[
m j=1X(t
γj) i
.
Applying Claim 1 to X(s
γ) ⊆
?S
ni=1
X(s
γi) and X(t
γ) ⊆
?S
mj=1
X(t
γj) (for each γ < ω
1) we obtain elements S
nγi=1
X(r
iγ), S
mγj=1
X(p
γj) such that X(s
γ) ⊆
?S
nγi=1
X(r
iγ) ⊆
?S
ni=1
X(s
γi), X(t
γ) ⊆
?S
mγj=1
X(p
γj) ⊆
?S
mj=1
X(t
γj) and X
η∈ B ∩ rg((T
ξα)
ε1m) for each η ∈ S
nγi=1
dom(r
i) ∪ S
mγj=1
dom(p
j). (Here B denotes B(ϕ) or the domain or range of T
ζα.) Thus
L
0=
n [
nγi=1
X(r
iγ) : γ < ω
1o
, n
m[
γj=1
X(p
γj) : γ < ω
1o
is a gap in B ∩ rg(T
ξα).
If B = B(ϕ) then L
0is a gap in B(ϕ ∩ ψ). Thus L
0ξ, the image of L
0under (T
ξα)
−ε1m, is a gap in B(ψ) and by (1), Q(L
0ξ) has the c.c.c., but by (3) it does not have the c.c.c., a contradiction.
If B = dom(T
γα) then (T
ξα)
−ε1m(L
0) = L
00is a gap in dom((T
ξα)
ε1m). By (2), Q(L
00) has the c.c.c. but by (3) it does not have the c.c.c., a contradic- tion.
We show that the assertion of Theorem 1 holds in the extension V[G], where G ⊆ P
κis a generic filter. It is clear that V[G] ° “2
ω= κ” and (by Theorems 2 and 3), V[G|α] ° “2
ω< κ” for each α < κ. Let B be a Boolean algebra in V[G] with card B = κ. There are elements b
γ∈ B for γ < κ such that B = S
α<κ
B
α, where B
αis the subalgebra generated by b
γ, γ ≤ α.
Assume inductively that we have an embedding i : B
α→ P (ω)/fin such that i(b
ξ) = X
βξwith β
ξ∈ Q
1for each ξ < α. We define a sequence ϕ
α: sup{β
ξ: ξ < α} → {0, 1} putting ϕ
α(β
ξ) = 1 for each ξ < α, and ϕ
α(ζ) = 0 otherwise. Thus B(ϕ
α)/fin is an isomorphic image of the algebra B
α. Let
b(s) = Y
s(ζ)=1
b
ζ·
Y
s(η)=−1
−b
η,
where s is a finite function on α with rg(s) ⊆ {−1, 1}. The next generator b
αdetermines a gap
L
Bα= ({b(s) : b(s) ≤ b
α}, {b(t) : b(t) · b
α= 0})
in the algebra B
α. Let L be the image of L
Bαunder i. So L is a gap in B(ϕ
α) and
L = ({X(s
i)}, {X(t
i)}),
where s
iis defined on {β
ξ: ξ ∈ dom(s)} by the equality s
i(β
ξ) = s(ξ) (t
iis defined similarly).
Let γ > sup(ϕ
α), γ ∈ Q
1and γ = Nb(L, ϕ
α). We define i(b
α) = X
γand
ϕ
α+1= ϕ
α∪ {(β, 0) : dom(ϕ
α) < β < γ} ∪ {(γ, 1)}. This extends i to an
embedding from B
α+1onto B(ϕ
α+1)/fin (we check this using Lemma 1).
Let Φ = S
α<κ
ϕ
α. It is clear that B is isomorphic to B(Φ).
Let f be an automorphism of B. Then i ◦ f ◦ i
−1is an automorphism of B(Φ) and there is a canonical name F for it, F ⊆ H(κ), consisting of some pairs ((x, y)
(Pκ), p), where x, y are canonical names for the elements of B(Φ) and the set F (x, y) = {p ∈ P
κ: ((x, y), p) ∈ F } is an antichain. Since P
κhas the c.c.c., the set
N
1= {α < κ : ∀x, y [x, y ∈ V
(Pα)→ F (x, y) ⊆ P
α]}
is ω
1-club (closed and unbounded) in κ. For any α ∈ N
1the restriction F
α= F ∩ (V
(Pα)× P
α)
is a P
α-name and F
α[G|α] = i ◦ f ◦ i
−1∩ V[G|α]. So, for all α ∈ N
1, the monomorphism F
α[G|α] belongs to V[G|α]. On the other hand, the sets
N
2= {α < κ : β < α, cf(α) = ω
1, F ∩ H
α= F
α}
are ω
1-club for all β < κ. From the diamond principle it follows that there is an increasing sequence {γ
β∈ N
1∩ N
2: β < κ} such that F
γβ= T
γβ. Let A(F ) = S
β<κ
T
γγββ+1and f = A(F )[G]. Then f is an automorphism of P (ω)/fin and i ◦ f ◦ i
−1⊆ f .
It remains to show that P
αhas the c.c.c. for each α ≤ κ. Let P
0αconsist of all p ∈ P
αsatisfying the following conditions:
1. For each γ ∈ supp(p) ∩ (Q
1∪ Q
2) there are u
γ(p), x
γ(p), w
γ(p) such that
p ¹ γ ° “p(γ) = (u
γ(p), x
γ(p), w
γ(p))”
and dom(s) ⊆ supp(p) for each s ∈ u
γ(p) ∪ w
γ(p). Moreover, for each γ ∈ supp(p) ∩ (Q
1∪ Q
2), the number dom(x
γ(p)) is constant (independent of γ).
We write l(p) for this value.
2. For each γ ∈ supp(p) ∩ E there are (α
1, s
α1, t
α1), . . . , (α
n, s
αn, t
αn) such that
p ¹ γ ° “p(γ) = {(α
1, s
α1, t
α1), . . . , (α
n, s
αn, t
αn)}”
and dom(s
αi) ∪ dom(t
αi) ⊆ supp(p) for i ≤ n.
Let P
?α⊆ P
0αbe the set of all p ∈ P
0αwith the property:
3. If γ ∈ M then there is an n ∈ ω such that p ¹ γ ° “h
γ(p(γ)) = n”, where h
γis a P
γ-name of a function such that
P
γ° “h
γ: R
γ→ ω and ∀n ∈ ω [h
−1γ(n) is linked]”.
(We can choose the h
γsince P
γ° “R
γis σ-linked”.)
Lemma 3. For each p ∈ P
αand m ∈ ω, there is a q ∈ P
?αsuch that p ≥ q
and l(q) ≥ m.
P r o o f. The proof (except for the case β ∈ E) is similar to the proof of Lemma 4.4 in Chapter 9 of [5].
Assume (inductively) that the lemma holds for α, β = α+1, β ∈ supp(p) and β ∈ E. There is a p
1≤ p¹ β such that
p
1° “p(β) = {(α
1, s
α1, t
α1), . . . , (α
n, s
αn, t
αn)}”
for some P
β-names (α
1, s
α1, t
α1), . . . , (α
n, s
αn, t
αn). We may assume that dom(s
αi) ∪ dom(t
αi) ⊆ supp(p) for i ≤ n. By the inductive assumption there is a p
2≤ p
1such that p
2∈ P
?βand l(p
2) ≥ m. Thus, the element p
2? p(β) has all the required properties.
We precede the next two lemmas with the following note: Fix α < κ and suppose that P
αhas the c.c.c. and the assumptions of Lemma 2 are satisfied.
Let P
αforce that L is the image under T
γαof an increasingly ordered gap L
0such that
P
γ° “Q(L
0) does not have the c.c.c.”
Suppose that {p
ξ: ξ < ω
1} ⊆ P
?αis a set of pairwise compatible conditions and that e
ξare P
α-names of conditions of the forcing E(L) such that
∀ξ < ω
1[p
ξ° “e
ξ= {(α
ξ1, s
αξ 1, t
αξ1
), . . . , (α
ξnξ, s
αξ nξ, t
αξnξ
)}”].
Let z
i, i = 1, . . . , n, be finite functions with dom(z
i) ⊆ T
ξ<ω1
supp(p
ξ) ∩ (Q
1∪Q
2). From Lemma 2 it follows that there are (at most) two possibilities:
1. There is an uncountable set B ⊆ ω
1such that
∀ξ
1, ξ
2∈ B [r
ξ1,ξ2° “e
ξ1= e
ξ2”], where r
ξ1,ξ2≤ p
ξ1, p
ξ2.
2. Any set A ⊆ ω
1satisfying the following condition:
If ξ
1, ξ
2∈ A then for some i
0∈ {1, . . . , n
ξ1} and j
0∈ {1, . . . , n
ξ2} we have
p
ξ1° “X(s
αξ1 i0) ⊆
?[
n i=1X(z
i)” and p
ξ2° “X(t
αξ2 j0) ⊆
?ω \ [
n i=1X(z
i)”
and for all r
ξ1,ξ2≤ p
ξ1, p
ξ2, r
ξ1,ξ2° “∀k ∈ {1, . . . , n
ξ2} [α
ξi10
6= α
ξk2] and ∀l ∈ {1, . . . , n
ξ1} [α
ξj20
6= α
ξl1]”
is at most countable.
Lemma 4 ([5]). Let p ° “X(s) ∈ fin” and γ = max dom(s). If p ∈ P
0αthen there is an r ∈ P
0αwith r ≤ p and l(p) = l(r) such that if r ¹ γ ° “r(γ) =
(u
rγ, x
rγ, w
γr)”, then r ° “s ¹ γ ∈ u
rγ” (if s(γ) = −1) or r ° “s ¹ γ ∈ w
rγ” (if
s(γ) = 1).
Lemma 5. Assume that p, q ∈ P
?α+1satisfy the following conditions:
1. p¹ α and q ¹ α are compatible.
2. If ξ ∈ supp(p) ∩ supp(q) ∩ M and
p ¹ ξ ° “p(ξ) ∈ h
−1ξ(n)” and q ¹ ξ ° “q(ξ) ∈ h
ξ(m)”
then n = m.
3. If ξ ∈ supp(p) ∩ supp(q) ∩ (Q
1∪ Q
2) and
p ¹ ξ ° “p(ξ) = (u
pξ, x
pξ, w
ξp)” and q ¹ ξ ° “q(ξ) = (u
qξ, x
qξ, w
qξ)”
then x
pξ= x
qξ.
4. Let ξ ∈ supp(p) ∩ supp(q) ∩ E and p ¹ ξ ° “p(ξ) = {(α
ξ1, s
αξ1
, t
αξ1
), . . . , (α
nξξ, s
αξ nξ, t
αξnξ
)}”, q ¹ ξ ° “q(ξ) = {(β
1ξ, s
βξ1
, t
βξ1
), . . . , (β
mξξ, s
βξ mξ, t
βξmξ
)}”.
Define A
ξ= {i : (α
ξi, s
αξ i, t
αξi
) ∈ p(ξ) and α
ξi6= β
jξfor all j such that (β
ξj, s
βξj
, t
βξj
) ∈ q(ξ)} (B
ξis defined in a similar way). Assume that for any i ∈ A
ξand j ∈ B
ξthere is no s
lwith dom(s
l) ⊆ supp(p) ∩ supp(q) such that
p ° “X(s
αξi
) ⊆
?[
X(s
l)” and q ° “X(t
βξj
) ⊆
?ω \ [
X(s
l)”.
Then p and q are compatible.
P r o o f. Denote by ∆ the set supp(p) ∩ supp(q) ∩ E. The required condi- tion will be constructed in the following way: First we define extensions of the conditions p and q by extending zero-one sequences x
pξand x
qξsuch that
p ¹ ξ ° “p(ξ) = (u
pξ, x
pξ, w
ξp)” and q ¹ ξ ° “q(ξ) = (u
qξ, x
qξ, w
qξ)”.
This will be done in such a way that if % ∈ ∆ and if some extension r of the conditions p and q forces
X(s
αi) = ε
i1X
γi1
∩ . . . ∩ ε
iniX
γini
, i ∈ A
%, and
X(t
βi) = ε
j1X
ξj1
∩ . . . ∩ ε
jmjX
ξjmj
, j ∈ B
%, then for some n ≥ l(p),
x
ξjk
(n) = 0, ε
ξjk
= −1, 1, ε
ξjk
= 1, x
γil
(n) =
0, ε
γi l= −1, 1, ε
γil
= 1, x
ξjk
⊂ x
ξjk
, x
γil