Descriptive set theory on κκ

Yurii Khomskii
April 13, 2014

1 Topology on κκ
1.1 Preliminaries
We consider a generalization of descriptive set theory on ω ω and 2ω to the generalized Baire
space κκ and generalized Cantor space 2κ , for uncountable regular cardinals κ satisfying
|κ<κ | = κ. For simplicity, we’ll present most things for κκ , but for 2κ everything is the
same.
Notation 1.1. Since we are considering generalizations from ω to κ, it will be convenient
to redefine the following piece of notation:

• “∀∞ α < λ” means “∃α0 < λ such that for all α with α0 < α < λ”, and

• “∃∞ α < λ” means “∀α0 < λ ∃α with α0 < α < λ”.

Usually λ will just be κ (for a fixed κ), in which case we will just write “∀∞ α” and
∞
“∃ α”.
Definition 1.2. For σ ∈ κ<κ , let [σ] := {x ∈ κκ | σ ⊆ x}, as usual. The standard topology
on κκ is the topology generated by the basic open sets {[σ] | σ ∈ κ<κ }.
Definition 1.3. A κ-sequence hxα | α < κi converges to x iff

∀σ ⊆ x ∀∞ α (σ ⊆ xα ).

A function f : κκ → κκ is continuous at x iff

∀τ ⊆ f (x) ∃σ ⊆ x ∀y (σ ⊆ y → τ ⊆ f (y)).

Some standard properties apply:

Lemma 1.4.

1. The basic open sets [σ] are clopen.

2. For fixed α, β, Aαβ := {x | x(α) = β} is clopen.
3. The intersection of less than κ basic open sets is either empty of basic open.

4. The intersection of less than κ open sets is open.

1
Proof.

1. Write κκ \ [σ] = {[τ ] | |τ | = |σ| and τ 6= σ}.

S

2. Write Aαβ = {[σ] | |σ| > α and σ(α) = β} and κκ \ Aαβ = {[σ] | |σ| > α and
S S
σ(α) 6= β}
S
3. Let λ < κ and consider {[sα ] | α < κ}. If sα ⊥sβ holds for two α, β < λ, then
this
S intersection is empty. Otherwise, all sα ’s are mutually compatible, so<κlet sλ :=
Sα<λ sα . But as κ is regular, sλ cannot have cofinal length, i.e., sλ ∈ κ . Hence
{[sα ] | α < κ} = [sλ ] is basic open.
4. A λ-intersection of a union of basic open sets is a union of λ-intersections of basic
open sets, each of which is open by the previous point. Hence, the union is open.

1.2 κ-trees
Trees are similar to the standard situation, but somewhat more exotic because of the
existence of various anomalies (Kurepa trees, Aronszajn tress etc.) and also because we
should distinguish between infinite branches and cofinal branches.

Definition 1.5.

1. A tree on κκ is a subset of κ<κ closed under initial segments.

2. For a node t ∈ T , we denote SuccT (t) := {s ∈ T | s = t_ hαi for some α}.
3. A node t is called
• terminal if SuccT (t) = ∅,
• non-splitting if |SuccT (t)| = 1,
• splitting if |SuccT (t)| > 1,
• λ-splitting if |SuccT (t)| = λ,
• club-splitting if {α | t_ hαi ∈ SuccT (t)} contains a club, and
• totally splitting if {α | t_ hαi ∈ SuccT (t)} = κ.

Moreover, t is called a successor node if |t| is a successor ordinal and a limit node if
|t| is a limit ordinal.
4. A tree T is called pruned if it has no terminal nodes. It is called cofinally-pruned, or
<κ-closed, if, in addition,
S for every increasing sequence {sα | α < λ} of nodes from
T , λ < κ, the limit α<λ sα is also in T .
5. A tree T is splitting-closed if it is cofinally-pruned,
S and moreover, for every sequence
{sα | α < λ} of splitting nodes the limit α<λ sα is also a splitting node.
6. A (cofinal ) branch through T is a real x ∈ κκ such that ∀α (xα ∈ T ). [T ] denotes
the set of all (cofinal) branches through T .

2
Lemma 1.6. If T is a tree then [T ] is closed. If C is a closed set then C = [T ] for some
tree T .
Proof. Let {xα | α < κ} be a sequence of reals from [T ], converging to x. We must show
that x ∈ [T ]. Let s ⊆ x be an initial segment. By definition, ∀∞ α (s ⊆ xα ), implying, in
particular, that s ∈ T . So indeed x ∈ [T ].
For the second claim, let T := {s | s ⊆ x for some x ∈ C}. If x ∈ C then by definition
x ∈ [T ], so it remains to prove the converse. But if x ∈ [T ] then, for every α, we know by
definition that there exists some xα ∈ C such that xα ⊆ xα . But that means that the
sequence {xα | α < κ} converges to x, implying that x ∈ C since C is closed.

1.3 Baire category and Fubini

The two classical properties of the ordinary Baire space—the Baire category theorem and
the Fubini theorem for category—remain valid in the generalized setting.
Lemma 1.7. Basic open sets are not meager (where meager is defined as κ-unions of
nowhere dense sets).
Proof. Let [s] be basic open and let {Xα | α < κ} be a κ-sequence of nowhere dense sets.
S sα+1 ⊆ sα , such thatS|sα+1 | > |sα |
Inductively, let s0 := s, and at successor stages find
and [sα+1 ] ∩ Xα = ∅. At limit stages λ, let sλ := α<λ sα . The result x := α sα is a real
which is in [s] but avoids all Xα .

Lemma 1.8. Fubini.

Proof. Later.

3
2 Descriptive stuff
2.1 Borel and projective sets
As usual we define the Borel sets to be the sets generated by the basic open sets and the
operations of κ-union, κ-intersection and complement. The Σ11 sets are then defined as
projections of Borel (or closed) sets, Π11 are the complements of Σ11 , etc. as usual.
Lemma 2.1. Borel 6= ∆11 .
Proof. Later.
Remark 2.2. Notice that every sequence {xα | α < λ} for λ < κ can be coded by one real
x (using the bijection κ×λ ∼ = κ. In particular, any quantifier ranging over < κ-sequences of
ordinals is still a first-order-quantifier, and only those ranging over reals are second-order
quantifiers. For example, for E ⊆ κ × κ, the statement “E is well-founded” is equivalent
to
@0 hαn | n < ωi ∀0 n (αn+1 Eαn )
which is just a closed statement.

2.2 Wellorder in L and absoluteness

In the ω ω case, the well-order of the reals of L is Σ12 -good. This is witnessed by saying
that x <L y iff ∃Lδ containing x and y and Lδ |= x <Lδ y, which is like saying ∃1 M (M
is countable and transitive, x, y ∈ M and M |= ZF C ∗ + V = L and M |= x <L y). Here,
“M is countable and transitive” is actually “E is wellfounded”, which is a Π11 -statement.
Hence, the entire statment is Σ12 . By a similar argument we know that actually the well-
order is Σ12 -good. However, by Remark 2.2, being wellfounded is closed in the κκ -setting.
As a consequence we get:
Fact 2.3. In L there is a Σ11 -good well-order of the generalized reals κκ .
Next, we discuss absoluteness. In the ω ω -setting we have
1. Σ11 -absoluteness between any model (including countable M ) and V , and
2. Σ12 -absoluteness (Shoenfield absoluteness) between any model M such that ω1 ⊆ M
and V .
In the generalized setting, we don’t have either of the above results.
Fact 2.4. Σ11 -absoluteness fails between XX and XX.
Proof. Something about destroying a club....
However, we have Σ11 -absoluteness between V and V P if V P is a forcing extension of V
by a <κ-closed forcing.
Definition 2.5. A forcing notion P is <κ-closed if for every increasing sequence hpα | α < λi
of conditions, with λ < κ, there exists a stronger condition q such that q ≤ pα for all α.
Theorem 2.6 (Who?). Σ11 -absoluteness holds between V and V P if P is a <κ-closed
forcing.
Proof. Something about extending one branch and witnessing stuff along this branch......
I still don’t get this AT ALL!

4
3 Forcing on κκ
Some stuff about forcing on κκ , particularly proper forcing, iterated proper forcing, master
conditions and what this means.

5
4 Regularity properties
4.1 Arboreal forcing on κκ
Generalizing the theory of regularity properties related to tree-like forcing notions in the
ω ω -context, we try to do a similar thing in the κκ -context, following the abstract setting
of Ikegami [Ike10].
Definition 4.1. A forcing notion P is called κ-arboreal if the conditions of P are trees in
κκ , with the additional property that for all T ∈ P and all s ∈ T the tree T ↑ s is also in
P. The ordering is given by q ≤ p iff q ⊆ p.
Given a κ-arboreal forcing P we can define:
Definition 4.2. Let A be a subset of κκ (or 2κ ). Then
1. A is P-null if ∀T ∈ P ∃S ≤ T such that [S] ∩ A = ∅. We denote the ideal of P-null
sets by NP
2. A is P-meager if it is a κ-union of P-null sets. We denote the κ-ideal of P-meager
sets by IP .
3. A is P-measurable if ∀T ∈ P ∃S ≤ T such that [S] ⊆∗ A or [S] ∩ A =∗ ∅, where ⊆∗
and =∗ refers to “modulo IP .
As in the classical setting, the property of being P-measurable usually reduces to one
of two simpler properties. Precisely:
Definition 4.3.

1. A κ-arboreal forcing notion P is topological if {[T ] | T ∈ P} is a topology base (i.e.,

∀S, T ∈ P, [S] ∩ [T ] is either empty or contains [R] for some R ∈ P).
2. A κ-arboreal forcing notion P satisfies Axiom A∗ iff there are orderings {≤α | α < κ},
with ≤0 =≤, satisfying:
(a) T ≤β S implies T ≤α S, for all α ≤ β.
(b) For hTα | α < λi, for λ ≤ κ (in particular λ = κ) satisfying

Tα ≥α Tα+1

for all α, then there exists T ∈ P such that T ≤ Tα for all α.

(c) For allST ∈ P, D dense below T , and α < κ, there exists S ≤α T such that
[S] ⊆ D.

S proves a stronger property, namely one finds an E ⊆ D such

In practice, one usually
that |E| = κ and [S] = E, which implies both point (c) above end that E is predense
below S, which is needed to prove the standard Axiom A. Notice that, in general, Axiom A
does not imply Axiom A∗ , since e.g. Cohen forcing satisfies the former but not the latter.
Lemma 4.4.

1. If P is topological then a set A is P-measurable iff it satisfies the property of Baire in

the topology generated by P.

6
 2. If P satisfies Axiom A∗ then NP = IP , and consequently a set A is P-measurable iff
∀T ∈ P ∃S ≤ T ([S] ⊆ A or [S] ∩ A = ∅) (i.e., we can forget about “modulo IP ).
Moreover, the collection of P-measurable sets forms a “κ+ -algebra” (i.e., it is closed
under κ-unions, κ-intersections and complements).

Proof. The proofs are essentially analogous to the classical situation, but let us present
them anyway since they are not so well-known.

1. First of all, notice that if P is topological then NP is exactly the collection of nowhere
dense sets in the P-topology and IP is exactly the ideal of meager sets in the P-
topology.
First assume A satisfies the P-Baire property, then let O be open such that A4O is
P-meager. Given any T ∈ P, we have two cases: if [T ] ∩ O = ∅ then we are done
since [T ] ∩ A =∗ ∅. If [T ] ∩ O is not empty then there exists a S ≤ T such that
[S] ⊆ [T ] ∩ O. Then [S] ⊆∗ A holds, so again we are done.
The converse direction is somewhat more involved (cf. [Kec95, Theorem 8.29]). As-
sume A is P-measurable. Let
• D1 be a maximal mutually disjoint subfamily of {T ∈ P | [T ] ⊆∗ A},
• D2 be a maximal mutually disjoint subfamily of {T ∈ P | [T ] ∩ A =∗ ∅}, and
• D := D1 ∪ D2 .
S
We will show that A4 D1 is P-meager.
S
Claim 1. D is P-open dense.

Proof of Claim. Start with any T . By assumption there exists S ≤ T such that
[S] ⊆∗ A or [S] ∩ A =∗ ∅. In the former case, note that by maximality, there must
be some S 0 ∈ SD1 such that [S] ∩ [S 0 ] 6= ∅. Then find S 00 such that [S 00 ] ⊆ [S] ∩ [S 0 ].
Then [SS ] ⊆ D1 . Likewise, in the case [S] ∩ A =∗ ∅ we find a stronger S 00 with
00

[S 00 ] ⊆ D2 .  (Claim 1).

S S
Claim 2. A ∩ D2 and D1 \ A are P-meager.

Proof of Claim. Since the proof of both statements is analogous, we only do the first.
Enumerate D2 := {Tα | α < |κκ |}. For each α, S let {X
α
i | i < κ} be a collection
α
of P-nowhere dense sets, such that [Tα ] ∩ A = i<κ Xi . Now, for every i < κ, let
Yi := α<|κκ | Xiα . We will show that each Yi is P-nowhere dense. So fix i and pick
S

any T ∈ P: if [T ] is disjoint from all [Tα ]’s then clearly also [T ] ∩ Yi = ∅. Else, let Tα
be such that [T ] ∩ [Tα ] 6= ∅. Then there exists S ≤ T such that [S] ⊆ [T ] ∩ [Tα ]. By
assumption, [Tα ] is disjoint from all [Tβ ]’s, and hence from all Xiβ ’s, for all β 6= α.
Next, since Xiα is P-nowhere dense, we can find S 0 ≤ S such that [S 0 ] ∩ Xiα = ∅. But
then [S 0 ] ∩ Yi = ∅, proving that Yi is indeed P-nowhere dense.
S
Now clearly D1 ∩ A is completely covered by the collection {Yi | i < κ}, therefore
it is meager.  (Claim 2).

7
 S S S S
Now it follows from both claims that A4 D1 = ( D1 \ A) ∪ (A ∩ D2 ) ∪ (A \ D)
is a union of three meager sets, hence it is meager.
2. Assume P satisfies strong-Axiom A, S and let {Ai | i < κ} be a collection of P-null
sets. We want to show that A := i<κ Ai is also P-null. For each i let Di := {T |
[T ] ∩ Ai = ∅}. By assumption, each Di is dense. Now let T0 ∈ P be given. Using
strong-Axiom A find, inductively, a sequence {Ti | i < κ} such that
• Ti+1 ≤i Ti for all i and
S
• [Ti ] ⊆ Di for all i.
This can always be done by condition (c) of strong-Axiom A. Then, by condition (b)
there is T such that T ≤ Ti for all i, and
T hence, [T ] ⊆ Di for all i. In particular,
[T ] ∩ Ai = ∅ for all i < κ, proving that Ai is P-null.
A similar proof shows that the collection of P-measurable sets is closed under κ-unions
and -intersections.
Corollary 4.5. If P is either topological or satisfies Axiom A∗ then the collection of P-
measurable sets forms a κ+ -algebra and all Borel sets are P-measurable.
Proof. Basic open sets [s] are clearly P-measurable: given any T ∈ P, by definition T ↑s ∈ P.
So if s ∈ T then [T ↑s] ⊆ [s] and if not then [T ↑s] ∩ [s] = ∅. The rest follows from the
above theorem.

4.2 Regularity of ∆11 and Σ11 sets

Let us abbreviate “all sets of complexity Γ are P-measurable” by “Γ(P)”. In the ω ω case,
ZFC proves Σ11 (P) (and by symmetry Π11 (P)), but Σ12 (P) and ∆12 (P) are independent of
ZFC. In the generalized case, however, things turn out differently. Halko and Shelah [?]
first proved that Σ11 (C) is false in ZFC, i.e., the Baire property for analytic sets fails.
We can generalize this proof somewhat, and for that, we need to introduce two specific
arboreal forcing notions which can be seen as the “weakest” forcing notions on 2κ and κκ ,
respectively.
Definition 4.6. A tree T on 2κ is called a κ-Sacks tree if it is splitting-closed and perfect,
i.e.,
S
• for every sequence hsα | α < λi of nodes in T , s := α<λ sα is in T ,
S
• for every sequence hsα | α < λi of splitting nodes in T , s := α<λ sα is a splitting
node in T , and
• every node t ∈ T has a splitting extension in T .
κ-Sacks forcing, denoted by Sκ , is the partial order of κ-Sacks trees ordered by inclusion.
Definition 4.7. A tree T on κκ is called a κ-Miller tree if it is:
S
• cofinally-pruned: for every sequence hsα | α < λi of nodes in T , s := α<λ sα is in
T,
• club-perfect: every node t ∈ T has an extension s ⊇ t in T which is club-splitting,

8
 S
• for every sequence hsα | α < λi of club-splitting nodes in T , s := α<λ sα is a club-
splitting node of T . Moreover, continuous club-splitting is required,
S i.e., if Cα is
the club witnessing the fact that sα is club-splitting, then C := α<λ witnesses the
club-splitting of s.
κ-Miller forcing, denoted by Mκ , is the partial order of κ-Miller trees ordered by inclu-
sion.
The generalized κ-Sacks forcing was introduced and studied by Kanamori in [?], and
the κ-Miller forcing is its natural variant, studied e.g. by Friedman and Zdomskyy in [?].
The property of being “continuous club-splitting” might seem somewhat artificial, but it
is necessary e.g. in the proof of [?].
The conditions on κ-Sacks seem to be the minimal ones required of an arboreal forcing
notion on 2κ , assuming we want it to be <κ-closed. So it is natural to assume that any
arboreal forcing on 2κ is a refinement of Sκ , i.e., its conditions are a particular kind of
κ-Sacks trees. A similar case could be made for κ-Miller and arboreal forcing notions of
κκ —although here the analogy does not quite apply, since we may also consider trees on
κκ that are only λ-splitting for some λ < κ, or even finitely splitting etc. Still, a large class
of arboreal forcings on κκ do form a refinement of Mκ . In our next result, we generalize
the Halko-Shelah theorem by proving that for any arboreal forcing P refining Sκ or Mκ ,
Σ11 (P) fails.
Theorem 4.8. Let P be an arboreal forcing notion on 2κ such that its conditions are κ-
Sacks trees, or an arboreal forcing notion on κκ such that its conditions are κ-Miller trees.
Then Σ11 (P) fails.
Proof. Let’s start with the first case. Define
A := {x ∈ 2κ | ∃C club such that ∀α ∈ C (x(α) = 1)}.
As “being a club set” is Borel, the set A is Σ11 .
If A were P-measurable then, in particular, we would have a T ∈ P such that [T ] ⊆∗ A or
[T ] ∩ A =∗S∅. First deal with the former case: let {Xα | α < κ} be P-null sets such that
[T ] \ A = α<κ Xα . Inductively, construct an increasing sequence of splitting nodes in T
in such a way that:
• s−1 := stem(T ),
• given sα , first extend to s0α ∈ T such that [T ↑s0α ] ∩ Xα+1 = ∅, then extend further
to a splitting node sα+1 ∈ T .
• at limit stages λ, note that s0λ := α<λ sα is a splitting node by assumption. Let
S
sλ := s0λ _ h0i.
S
Now let x := α<κ sα . Then x is a branch through T , x ∈ / Xα for all α, and moreover, there
exists a club set C 0 such that x(α) = 0 for all α ∈ C 0 . In particular, x ∈/ A—contradiction.
To deal with the second case that [T ] ∩ A =∗ ∅, proceed analogously except that at limit
stages, pick sλ := s0λ _ h1i. Then it will follow that x ∈ A.

Next, suppose P is an arboreal forcing such that its conditions are κ-Miller trees. Fix a
partition of κ into two stationary sets S0 and S1 , and define
A := {x ∈ 2κ | ∃C club such that ∀α ∈ C (x(α) ∈ S0 )}.

9
Again this set is Σ11 . As before, assume there is T ∈ P such that [T ] ⊆∗ A or [T ] ∩ ∗
SA = ∅.
Dealing with the former case let {Xα | α < κ} be P-null sets such that [T ] \ A = α<κ Xα .
Inductively, construct an increasing sequence of club-splitting nodes in T in such a way
that:

• s−1 := stem(T ),
• given sα , first extend to s0α ∈ T such that [T ↑s0α ] ∩ Xα+1 = ∅, then extend further
to a club-splitting node sα+1 ∈ T .
• at limit stages λ, note that s0λ := α<λ sα is club-splitting. Therefore, there exists
S
α ∈ S1 such that sλ := s0λ _ hαi ∈ T .
S
Now let x := α<κ sα . Then x is a branch through T , x ∈ / Xα for all α, and moreover, there
exists a club set C 0 such that x(α) ∈ S1 for all α ∈ C 0 . In particular, x ∈
/ A—contradiction.
To deal with the second case that [T ] ∩ A =∗ ∅, proceed analogously except that at limit
stages, pick sλ := s0λ _ hβi ∈ T such that β ∈ S0 . Then it will follow that x ∈ A.

So, with Σ11 (P) already being false in ZFC in many reasonable cases, we are left with
the ∆11 -level. Recall that this is strictly between the Borel and the Σ11 levels.
Lemma 4.9. If V = L then ∆11 (P) fails for all arboreal P.
Proof. Use the Σ11 -good wellorder of the reals of L, and proceed as in the ω-case, obtaining
a ∆11 -counterexample as opposed to a ∆12 -one.

On the other hand, we now show that, assuming P satisfies κ-properness (the natural
generalization of properness), a κ+ -iteration of P yields ∆11 (P). First we need an analogy
with the classical case showing that master conditions can be strengthened in a particular
way. The idea comes from Zapletal, cf. [Zap08, Proposition 2.2.2].
Definition 4.10. For P, M ≺ Hθ an elementary submodel of size κ:
1. T ∈ P is a (M, P)-master condition if T “Ġ is generic over M ” (the usual definition),
and

2. T ∈ P is a strong-(M, P)-master condition if [T ] ⊆∗ {x | x is an (M, P)-generic real},

where ⊆∗ means “modulo IP ”.
The key observation in the above definition is that the set

Bgen/M := {x | x is an (M, P)-generic real}

is Borel, since x is (M, P)-generic iff

∀0 D ∈ M (M |= “D is dense” → ∃0 T ∈ (D ∩ M ) (x ∈ [T ])).

Lemma 4.11. For P arboreal and M ≺ Hθ elementary, |M | = κ:

1. if T is a strong-(M, P)-master condition then it is an (M, P)-master condition.

10
 2. if T is an (M, P)-master condition then there is S ≤ T such that S is a strong
(M, P)-master condition.
Proof.

1. Let Bgen/M := {x | x is a (M, P)-generic real}, and suppose [T ] ⊆∗ B. Since [T ] ∩

Bgen/M is a Borel set in IP , and one can show that for such Borel sets, the generic real
is forced out of it by every condition [this might require additional assumptions—
check!] it follows that T ẋgen ∈ Bgen/M , so T “ẋgen is (M, P)-generic” and hence
also T “Ġ is (M, P)-generic”.

2. Suppose T is (M, P)-generic. By Lemma 4.4, Bgen/M is P-measurable. Hence, for

every T 0 ≤ T there exists T 00 ≤ T 0 such that [T 00 ] ⊆∗ Bgen/M or [T 00 ] ∩ Bgen/M =∗ ∅.
If the first case occurs for at least one T 00 , then we are done. But if not, then the set
D := {T 00 | [T 00 ] ∩ Bgen/M =∗ ∅} is dense below T , implying that T ẋgen ∈ / Bgen/M ,
contradicting the assumption.

Theorem 4.12. If P is arboreal and κ-proper, then a κ+ -iteration of P (κ-support??)

yields ∆11 (P).
Proof. In V [Gκ+ ], let A be ∆11 , defined by Σ11 -formulas φ and ψ. Without loss of generality
(i.e., κ-properness) the parameters of φ and ψ can be assumed to be in V . Then, let x0
be the first P-generic real over V . Without loss of generality assume V [Gκ+ ] |= φ(x0 ). By
Theorem 2.6, Σ11 -absoluteness holds between V [x0 ] and V [Gκ+ ], since (the iteration of) P is
a <κ-closed forcing. Therefore V [x0 ] |= φ(x0 ). In V , let T be such that T φ(ẋgen ). Now,
again in V , take an elementary submodel M of size κ, containing T , so that by elementarity
M |= “T φ(ẋgen )”. Going back to V [Gκ+ ], by κ-properness, let S ≤ T be an (M, P)-
master condition, and using Lemma 4.11 (2) find S 0 ≤ S such that [S 0 ] ⊆∗ Bgen/M = {x | x
is (M, P)-generic over M }. Then [S 0 ] ⊆∗ {x | φ(x)} = A, as we had to show.

References
[Ike10] Daisuke Ikegami. Forcing absoluteness and regularity properties. Ann. Pure Appl.
Logic, 161(7):879–894, 2010.
[Kec95] Alexander S. Kechris. Classical descriptive set theory, volume 156 of Graduate
Texts in Mathematics. Springer-Verlag, New York, 1995.
[Zap08] Jindřich Zapletal. Forcing idealized, volume 174 of Cambridge Tracts in Mathe-
matics. Cambridge University Press, Cambridge, 2008.

11