Arxiv:1812.00960V1 [Math.LO] 3 Dec 2018 N Ch4,H Rvdta 2 That Proved He [Coh64], and Ucinblw the Below

arXiv:1812.00960v1 [math.LO] 3 Dec 2018 n Ch4,h rvdta 2 that proved he [Coh64], and ucinblw The below. Function n iglrcardinal singular any olw ht2 that follows with asoffi Hu4 wt ale esosi Hu7 n [H and [Hau07] in versions 2 earlier that (with [Hau14] in Hausdorff The L isl,Kr ¨dlpoe n[Gö40] that in Gödel proved Kurt Firstly, au f2 of value fteCniumFnto.The Function. Continuum the of nsadcmie hmi a htwshneot nw sth as known henceforth was that way a in them combines and ings ohsswsaogtefis ttmnsta eesont ei be to shown were that statements first the among was pothesis rein true o iglrcardinals singular For fwihi titybetween strictly is which of For ig fstter.I 88 er atravne the advanced Cantor Georg (CH) 1878, In theory. set of nings Introduction. otecriaiyo t oe set power its of cardinality the to nteohrhn,we alChnivne h ehdo for of method the invented Cohen Paul when hand, other the On . nEso-ieTermfrZreoFane Set Zermelo-Fraenkel for Theorem Easton-like An regular eeaie otnu yohss(GCH) Hypothesis Continuum Generalized FC ZF swell-ordered”( is powersets adnl.Ti opeet u eut rm[K6,weew prove we where [FK16], from ZF results our complements This cardinals. nwr usino hlhi Se6,weeh mhszsthat emphasizes he where [She16], in Shelah of question a answers eso hti h theory the in that show We possible. AX κ [a7],wihsae ht2 that states which ([Can78]), hoywt h xo fDpnetChoice Dependent of Axiom the with Theory L = 4 (without se[¨4].I 90 ila atnpoe h following the proved Easton William 1970, [Gö40]). In (see κ κ sa“esnbeter,weemc fstter n combinato and theory set of much where “reasonable”theory, a is + o singular for [a7].I i ocn osrcin etks“ay Coh “many” takes he construction, forcing his In ([Eas70]). cardinals o l cardinals all for κ = ´( κ The κ DC + ) . neFregladPtrKoepke Peter and Fernengel Anne a aeams rirr auso n e funcountable of set any on values arbitrary almost take can AX ,aypsil eairo the of behavior possible any ), κ iglrCriasHptei (SCH) Hypothesis Cardinals Singular κ otnu Function Continuum κ n esnbebhvo fte2 the of behavior reasonable any , ihtepoet ht2 that property the with κ oee,testaini o oeivle,snethe since involved, more lot a is situation the however, , 4 ,the ), ssrnl nune ytebhvo fteContinuum the of behavior the by influenced strongly is ℵ 0 ℵ κ 0 ti a is it , n h adnlt of cardinality the and θ ZF ol eaycardinal any be could fnto esrn h ujciesz fthe of size surjective the measuring -function ´( GCH κ Abstract + ) ℵ a enivsiae ic h al begin- early the since investigated been has , DC 0 global 1 = + scnitn with consistent is CH ℵ freeycardinal every “for 1 ..teei osttecardinality the set no is there i.e. , ttmn bu osbebehaviors possible about statement κ od ntecntutbeuniverse constructible the in holds λ θ ↦ fnto a eraie;and realized; be can -function < 2 κ κ od o all for holds hc asaycardinal any maps which , ´(ℵ κ otnu Hypothesis Continuum fucutbycofinality. uncountably of a omltdb Felix by formulated was 0 κ ) fnto sconsistent is -function h otnu Hy- Continuum The . FC ZF λ dpnetof ndependent h set the , u8) Asserting au08]). e ZF λ mle htfor that implies atnproduct Easton ic tholds it since , igi [Coh63] in cing < lblresult: global + κ htin that isis rics talready it , DC [ λ ] nforc- en ℵ + 0 ZF κ . : It turned out that the negation of the SCH is tightly linked with the existence of large cardinals. Among the first results in this direction was a theorem by Menachem Magidor ([Mag77b] and [Mag77a]), who proved that, assuming a huge cardinal, it is possible that GCH first fails at a singular strong limit cardinal. On the other hand, Ronald Jensen and Keith Devlin proved in [DJ75] that the negation of 0♯ implies SCH. Motik Gitik determined in [Git89] and [Git91] the consistency strength of SCH being the existence of a measurable cardinal λ of ¬ Mitchell order σ λ = λ++. ( ) There are many more results about possible behaviors of the Continuum Function starting from large cardinals. For instance, by a theorem of Carmi Merimovich, the theory ZFC κ 2κ = κ+n is consistent for each n < ω ([Mer07]). + ∀ ( )

On the other hand, Silver’s Theorem ([Sil75]) states that for any singular cardinal κ of uncountable cofinality such that 2λ = λ+ holds for all λ < κ, it already follows that 2κ = κ+. Hence, the SCH holds if it holds for all singular cardinals of countable cofinality. This result was extended by Fred Galvin and András Hajnal shortly after ([GH75]). Another prominent example concerning upper bounds on the Continuum Function of singular cardinals is the following theorem by Saharon Shelah ([She94]):

If 2ℵn < for all n < ω, then 2ℵω < . ℵω ℵω4

This makes clear that there are signiﬁcant constraints on possible behaviors of the Continuum Function in ZFC. In particular, a result like Easton’s Theorem can not exist for singular cardinals.

All of the aforementioned results essentially involve the Axiom of Choice. Without AC, however, there is a lot more possible. In [AK10], Arthur Apter and Peter Koepke examine the consistency strength of the negation of SCH in ZF AC. + ¬ In this context, one has to distinguish between injective and surjective failures. An injective failure of SCH at κ is a model of ZF AC with a singular cardinal +¬ κ such that GCH holds below κ, but there is an injective function ι λ κ ∶ ↪ ´( ) for some λ ≥ κ++. A surjective failure of SCH at κ is a model of ZF AC + ¬ with a singular cardinal κ such that GCH holds below κ, but there is a surjective function f κ → λ for some cardinal λ ≥ κ++. ∶ ´( ) On the one hand, Arthur Apter and Peter Koepke construct injective failures of the SCH at , and that would contradict the theorems by Shelah and Silver ℵω ℵω1 ℵω2 in the ZFC-context, but have fairly mild consistency strengths in ZF AC. On +¬ the other hand, regarding a surjective failure of the SCH, they prove that for every α ≥ 2, ZFC together with the existence of a measurable cardinal is equiconsistent with the theory ZF AC “GCH holds below ” + ¬ + ℵω +

2 “there exists a surjective function f ω → ”. + ∶ [ℵω] ℵω+2 It follows that also without the Axiom of Choice, injective failures of the SCH are inevitably linked to large cardinals. Regarding surjective failures however, it is not possible to replace in their argument the surjective function f ω → ∶ [ℵω] ℵω+2 by a surjection f → , so the following question remained: Is it pos- ∶ ´(ℵω) ℵω+2 sible, for λ ≥ , to construct a model of ZF AC where GCH holds below ℵω+2 + ¬ and there is a surjection f → λ without any large cardinal assumptions? ℵω ∶ ´(ℵω) This question was positively answered by Motik Gitik and Peter Koepke in [GK12], where a ground model V ⊧ ZFC GCH with a cardinal λ ≥ is extended + ℵω+2 via symmetric forcing, in a way such that the extension N = V G preserves all ( ) V -cardinals, the GCH holds in N below , and there is a surjective function ℵω f → λ. ∶ ´(ℵω) More generally, in the absence of the Axiom of Choice, where the power set κ ´( ) of a cardinal κ is generally not well-ordered, the “size” of κ can be measured ´( ) surjectively by the θ-function

θ κ = sup α ∈ Ord f κ → α surjective function , ( ) ∶ { S ∃ ∶ ´( ) } which generalizes the value θ = θ ω prominent in descriptive set theory. In the ∶ ( ) AC-context, the θ-function provides a surjective substitute for the Continuum ¬ Function κ ↦ 2κ.

One can show that in the model constructed in [GK12], it follows that indeed, θ = λ. This gives rise to the question whether the behavior of the θ-function (ℵω) might be essentially undetermined in ZF .

In [FK16], we could prove that indeed, the only constraints on the θ-function in ZF are the obvious ones: weak monotonicity and θ κ ≥ κ++ for all κ. In other ( ) words: In ZF , there is an analogue of Easton’s Theorem for regular and singular cardinals.

Theorem. ([FK16]) Let V be a ground model of ZFC GCH with a function F + on the class of inﬁnite cardinals such that the following properties hold:

• κ F κ ≥ κ++ ∀ ( ) • κ, λ κ ≤ λ → F κ ≤ F λ . ∀ ( ) ( ) Then there is a cardinal-preserving extension N ⊇ V with N ⊧ ZF such that θN κ = F κ holds for all κ. ( ) ( ) In our construction, we introduce a forcing notion P whose elements p are functions on trees t, ≤ with ﬁnitely many maximal points. The trees’ levels are indexed ( t) by cardinals, and on any level κ, there are ﬁnitely many vertices κ, i with i < ( )

3 F κ . For successor cardinals κ+, the value p κ+, i is a partial 0-1-function on ( ) ( ) the interval κ, κ+ . Thus, for any condition p and κ, i ∈ dom p, it follows that [ ) ( ) p ν+, j ν+, j ≤ κ, i is a partial function on κ with values in 0, 1 . Since ⋃{ ( ) S ( ) t ( )} { } we do not allow splitting at limits for the trees, it follows that this forcing indeed adds F κ -many new κ-subsets for every cardinal κ. ( ) We discussed in [FK16, Chapter 6] whether it might be possible to modify our construction and use trees with countably many maximal points, which could result in a countably closed forcing, giving rise to a symmetric extension N with N ⊧ ZF DC. However, this modiﬁcation would have a drastic impact on the + forcing notion, destroying a crucial homogeneity property. Seemingly, our construction relies on certain ﬁniteness properties, and hence does not opt for a symmetric extension N with N ⊧ DC.

In this paper, we treat the question whether the θ-function is still essentially undetermined, if we consider a model N ⊧ ZF DC. Starting from a ground model + V ⊧ ZFC GCH, we construct a cardinal-preserving symmetric extension N ⊇ V + with N ⊧ ZF DC, and this time, we generalize the forcing in [GK] to obtain a + countably closed forcing notion P.

The Axiom of Dependent Choice (DC), introduced by Paul Bernays in 1942 ([Ber42]), states that for every nonempty set X with a binary relation R such that for all x ∈ X there is y ∈ X with yRx, it follows that there is a sequence x n < ω ( n S ) in X such that xn+1Rxn for all n < ω.

When dealing with real numbers, surprisingly often DC is sufficient (instead of the full Axiom of Choice), and the theory ZF DC provides an interesting framework + for real analysis. Concerning combinatorial set theory however, investigations under ZF DC + seemed rather hopeless in the first place. A crucial step in the other direction was a paper by Saharon Shelah ([She97]) with the main result in ZF DC that + whenever µ is a singular cardinal of uncountable cofinality such that H µ = µ, S ( )S then µ+ is regular and non-measurable. In the case that the power sets α are ´( ) well-orderable for all α < ω1 with ⋃α<ℵ α = ω1 , it essentially follows that ℵ S ω1 ´( )S ℵ also is well-orderable. ´(ℵω1 ) Subsequently (see [She10] and [She16]), Shelah showed that much of pcf-theory is possible in ZF DC, if an additional axiom is adopted: +

AX For every cardinal λ, the set λ ℵ0 can be well-ordered. ( 4) [ ]

Starting from a ground model V ⊧ ZFC, any symmetric extension by countably closed forcing yields a model of ZF DC AX (see [She10, p.3 and p.15]). + + 4 In [She16, 0.1], Shelah concludes that ZF ZF AX is “a reasonable theory, + + 4 for which much of combinatorial set theory can be generalized”. For example, he

4 proves a rather strong version of the pcf-theorem, gives a representation of λκ for λ >> κ, and proves that certain covering numbers exist. Concerning applications to cardinal arithmetic, Shelah emphasizes that we “cannot say much” on possible cardinalities of κ , and suggests to investigate possible cardinalities of ´( ) κℵ0 κ ∈ Card rather than κ κ ∈ Card ([She10, p. 2]). In [She16, 0.2] he S ´( ) S asks, referring to [GK12], if there are any bounds on θ κ for singular cardinals κ ( ) in ZF DC AX . + + 4 In this paper, we give a negative answer to this question. We prove that in ZF DC AX again, the only restrictions on the θ-function on a set of un- + + 4 countable cardinals are the obvious ones:

Given a ground model V ⊧ ZFC GCH with “reasonable” sequences of uncount- + able cardinals κ η < γ and α η < γ for some ordinal γ, we construct a ( η S ) ( η S ) cardinal-preserving symmetric extension N ⊇ V with N ⊧ ZF DC AX , such + + 4 that θN κ = α holds for all η < γ. ( η) η More precisely, we prove:

Theorem. Let V be a ground model of ZFC GCH with γ ∈ Ord and sequences + of uncountable cardinals κ η < γ and α η < γ , such that κ η < γ is ( η S ) ( η S ) ( η S ) strictly increasing and closed, and the following properties hold:

• η < η′ < γ α ≤ α ′ , i.e. the sequence α η < γ is increasing, ∀ η η ( η S ) • η < γ α ≥ κ+ +, ∀ η η • η < γ cf α > ω, ∀ η • η < γ α = α+ → cf α > ω . ∀ ( η ) Then there is a cardinal- and cofinality-preserving extension N ⊇ V with N ⊧ ZF DC AX such that that θN κ = α holds for all η < γ. + + 4 ( η) η Our paper is structured as follows: In Chapter 1, we briefly review some basic definitions and fact about forcing and symmetric extensions. In Chapter 2, we state our theorem and then argue why it is not possible to drop any of our requirements on the sequences κ η < γ and α η < γ . ( η S ) ( η S ) In Chapter 3, we introduce our forcing notion which blows up the power sets κ η < γ according to α η < γ ; and in Chapter 4 construct a group ´( η) S ( η S ) A of P-automorphisms and a normal filter F on A, giving rise to our symmetric extension N = V G . In Chapter 5, we prove that sets of ordinals located in N can ∶ ( ) be captured in fairly “mild” V -generic extensions, which implies that all cardinals and cofinalities are N-V -absolute. It remains to show that indeed, θN κ = α ( η) η holds for all η < γ. The first part, θN κ ≥ α , follows by our construction ( η) η (see Chapter 6.1), while for the second part, θN κ ≤ α , we assume that there ( η) η

5 was a surjective function f κ → α in N, and obtain a contradiction by ∶ ´( η) η capturing a restricted version f β in an intermediate generic extension V Gβ which [ ] is suﬃciently cardinal-preserving (see Chapter 6.2 and 6.3). We treat the values θ λ for cardinals λ in the “gaps” λ ∈ κ , κ and λ ≥ κ = sup κ η < γ in ( ) ( η η+1) γ ∶ { η S } Chapter 6.4 and 6.5, respectively, and show that they are the smallest possible. We conclude with several remarks in Chapter 7.

1 Basics.

In this chapter, we brieﬂy establish some basic notations and terminology about forcing and symmetric extensions. We write Ord and Card for the class of all ordinals and cardinals, respectively. For our construction, we work with a countable transitive model V of ZFC, our ground model, with a notion of forcing P, ≤, 1 ∈ V . The class of P-names, ( ) Name P , is deﬁned recursively as follows: Name P = , Name P = Name P ( ) 0( ) ∶ ∅ α+1( ) ∶ ´ α( ) P for α ∈ Ord, and Name P = Name P whenever λ is a limit ordinal. × λ( ) ∶ ⋃α<λ α( ) Then

Name P = Nameα P . ( ) ∶ α∈Ord ( ) As usual, we denote P-names asx ˙. Forx ˙ ∈ Name P withx ˙ ∈ Name P ( ) α+1( )∖ Name P , we write rk x˙ = α for the rank ofx ˙. α( ) ( ) ∶ Let G be a V -generic ﬁlter on P. The V -generic extension by G is V G = x˙ G x˙ ∈ [ ] ∶ { S Name P , where the interpretation function ⋅ G is deﬁned recursively on the ( )} ( ) Name P -hierarchy as usual. Then V G is a transitive model of ZFC with α( ) [ ] V ⊆ V G . [ ] For an element a of the ground model, its canonical name is denoted bya ˇ. When- ever x, y ∈ V G with x = x˙ G, y = y˙G, there is a canonical P-name for the pair [ ] x,y , which will be abbreviated by ORP x,˙ y˙ . ( ) ( ) Regarding the construction of symmetric extensions, we follow the presentation in [Dim11], where the standard method for forcing with Boolean values as described in [Jec06] and [Jec73] is translated to partial orders.

For the rest of this chapter, fix a partial order P. Let Aut P denote the automor- ( ) phism group of P. Any π ∈ Aut P can be extended to an automorphism π of the ( ) ̃ name space Name P by the following recursive definition: ( ) π x˙ ∶= π y˙ ,πp y,p˙ ∈ x˙ . ̃( ) { (̃( ) ) S ( ) } We confuse any π ∈ Aut P with its extension π (which does not lead to ambigu- ( ) ̃ ity). For any canonical namea ˇ and π ∈ Aut P , it follows recursively that π aˇ = aˇ. ( ) ( ) The forcing relation ⊩ can be defined in an outer model as follows:

6 If ϕ v , . . . , v is a formula of set theory andx ˙ ,..., x˙ ∈ Name P , then ( 0 n−1) 0 n−1 ( ) p ⊩ ϕ x˙ ,..., x˙ if for every V -generic ﬁlter G on P, it follows that V G ⊧ ( 0 n−1) [ ] ϕ x˙ G,..., x˙ G . ( 0 n−1) The forcing relation ⊩ can also be deﬁned in the ground model V , and the forcing theorem holds:

Theorem (Forcing Theorem). If ϕ v , . . . , v is a formula of set theory and G a ( 0 n−1) V -generic ﬁlter on P, then for every x˙ ,..., x˙ ∈ Name P , it follows that V G ⊧ 0 n−1 ( ) [ ] ϕ x˙ G,..., x˙ G if and only if there is a condition p ∈ P with p ⊩ ϕ x˙ ,..., x˙ . ( 0 n−1) ( 0 n−1)

Let now A ⊆ Aut P denote a group of P-automorphisms. A normal ﬁlter on A is ( ) a collection F of subgroups B ⊆ A such that F ≠ ∅, F is closed under supersets and ﬁnite intersections, and for any B ∈ F and π ∈ A, it follows that the conjugate π−1Bπ is contained in F, as well.

An important property of P-automorphisms is the symmetry lemma:

Lemma (Symmetry Lemma). For a formula of set theory ϕ v , . . . , v , an auto- ( 0 n−1) morphism π ∈ Aut P and P-names x˙ ,...,x , it follows that p ⊩ ϕ x˙ ,..., x˙ ( ) 0 n−1 ( 0 n−1) if and only if πp ⊩ ϕ πx˙ ,...,πx˙ . ( 0 n−1)

The proof is by induction over the complexity of ϕ.

Fix a normal ﬁlter F on A. A P-namex ˙ is symmetric if the stabilizer group π ∈ A πx˙ = x˙ is contained in F. Recursively, a namex ˙ is hereditarily symmet- { S } ric,x ˙ ∈ HS, ifx ˙ is symmetric andy ˙ ∈ HS for ally ˙ ∈ domx ˙.

For a V -generic ﬁlter G on P, the symmetric extension by F and G is deﬁned as follows: N ∶= V G ∶= x˙ G x˙ ∈ HS . ( ) { S } Then N is a transitive class with V ⊆ N ⊆ V G and N ⊧ ZF . [ ]

The symmetric forcing relation ⊩s can be deﬁned informally as follows: If ϕ v , . . . , v is a formula of set theory, andx ˙ ,..., x˙ ∈ HS, then p ⊩ ( 0 n−1) 0 n−1 s ϕ x˙ ,..., x˙ if for every V -generic ﬁlter G on P, it follows that V G ⊧ ϕ x˙ G,..., ( 0 n−1) ( ) ( 0 x˙ G . n−1)

Note that the symmetric forcing relation ⊩s can be deﬁned in the ground model similar to the ordinary forcing relation ⊩, but with the quantiﬁers and variables ranging over HS. It has most of the basic properties as ⊩. In particular, the forcing theorem holds for ⊩s, and the symmetry lemma is true, as well.

We will consider the following weak versions of the Axiom of Choice:

7 The Axiom of Dependent Choice (DC) states that whenever X is a nonempty set with a binary relation R, such that for all x ∈ R there exists y ∈ R with yRx, there is a sequence x n < ω in X with x Rx for all n < ω. ( n S ) n+1 n The Axiom of Countable Choice (ACω) states that every countable family of nonempty sets has a choice function.

The Axiom of Choice implies DC, and DC implies ACω.

The following lemma follows from [Kar14, Lemma 1]:

Lemma ([Kar14, Lemma 1]). If P is countably closed with a group of automorphisms A ⊆ Aut P and a normal ﬁlter F on A, such that F is countably closed ( ) as well, then DC holds in the corresponding symmetric extension N = V G . ( )

The theory ZF + DC is suﬃcient to develop most of real analysis; while combinatorial set theory in ZF + DC seemed rather hopeless in the ﬁrst place. In [She10] and [She16], Shelah suggests that when working with ZF + DC, another axiom should be adopted to set the framework for a “reasonable” set theory, where now, surprisingly much of combinatorial set theory can be realized:

AX For every cardinal λ, the set λ ℵ0 can be well-ordered. ( 4) [ ]

Note that AX holds true in any symmetric extension by countably closed forcing ( 4) (see [She10, p.3 and p.15]).

2 The Theorem

We start from a ground model V ⊧ ZFC + GCH and a reasonable behavior of the θ-function: There will be sequences of uncountable cardinals κ 0 < η < γ and ( η S ) α 0 < η < γ in V , where γ ∈ Ord, for which we aim to construct a symmetric ( η S ) extension N ⊇ V with N ⊧ ZF + DC + AX4, such that V and N have the same cardinals and coﬁnalities and θN κ = α holds for all η. ( η) η (Later on, we will set κ0 ∶= ℵ0, α0 ∶= ℵ2 for technical reasons – therefore, we talk about sequences κ 0 < η < γ , α 0 < η < γ here, excluding κ and α . ) ( η S ) ( η S ) 0 0 First, we want to discuss what properties the sequences κ 0 < η < γ and ( η S ) α 0 < η < γ must have to allow for such construction. ( η S ) W.l.o.g. we can assume that (κ 0 < η < γ is strictly increasing and closed. η S ) The following conditions must be satisﬁed:

′ • For η < η , it follows from κη < κη′ that αη ≤ αη′ must hold, i.e. the sequence α 0 < η < γ must be increasing. ( η S )

8 • For any cardinal κ, it is possible to construct a surjection s ∶ ´ κ → κ+ ( ) ++ without making use of the Axiom of Choice. Hence, αη ≥ κη must hold for all η.

• Since N ⊧ ACω, it follows that cf αη > ω for all η: Assume towards a contradiction, there were cardinals κ, α with θN κ = α, but cf N α = ω. ( ) ( ) Let α = α . By deﬁnition of θN κ , it follows that for every i < ω, there ⋃i<ω i ( ) exists in N a surjection from ´ κ onto α . Now, AC allows us to pick in ( ) i ω N a sequence s i < ω such that each s ∶´ κ → α is a surjection. This ( i S ) i ( ) i yields a surjective function s ∶´ κ × ω → α, where s X, i ∶= s X for each ( ) ( ) i( ) X, i ∈ ´ κ × ω; which can be easily turned into a surjection s ∶´ κ → α. ( ) ( ) ( ) Contradiction, since θN κ = α. Hence, it follows that cf α > ω for all η. ( ) η + • Finally, for every αη a successor cardinal with αη = α , we will need that cf α > ω.

In this setting, it is not possible to drop this requirement: We start from a ground model V ⊧ ZFC + GCH with sequences κ 0 < η < γ , α 0 < ( η S ) ( η S η < γ , and aim to construct N ⊇ V with N ⊧ ZF + DC such that V and N ) have the same cardinals and coﬁnalities, and θN κ = α holds for all η. If ( η) η there was some η with θN κ = α+, where cf α = ω, one could construct in ( η) N a surjective function s ∶´ κ → α+ as follows: ( η) Take a surjection s ∶ ´ κ → α in N. Firstly, the canonical bijection ( η) κ ↔ κ × ω gives a surjection s ∶ 2κ → 2κ ω. Secondly, the surjection 0 ( ) s ∶´ κ → α yields in N a surjection s ∶ 2κ ω → αω, by setting s X i < ( η) 1 ( ) 1( i S ω ∶= s X i < ω . Then s is surjective, since for α i < ω ∈ αω ) ( ( i) S ) 1 ( i S ) given, one can use AC to obtain a sequence Y i < ω with Y ∈ s−1 α ω ( i S ) i [{ i}] for all i < ω. Then s Y i < ω = α i < ω . Thirdly, it follows from 1( i S ) ( i S ) ω + cf α = ω that there is a surjection s2 ∶ α → α in V . Then s2 ∈ N, and since αω N ⊇ αω V and α+ N = α+ V , we obtain a surjection s ∶ αω → α+ in ( ) ( ) ( ) ( ) 2 N. κ + Thus, it follows that s2 ○ s1 ○ s0 ∶ 2 → α is a surjective function in N; contradicting that θN κ = α+. ( η)

Hence, in our setting, where we want to extend a ground model V ⊧ ZFC + + GCH cardinal-preservingly, it is not possible to have αη = α with cf α = ω.

The following question arises: More generally, without referring to a ground

model V , could there be N ⊧ ZF + DC + AX4 with cardinals κ, α, such that cf N α = ω and θN κ = α+? The answer is no: Let s ∶ 2κ → α ( ) ( ) denote a surjective function in N. Then with DC, it follows as before that there is also a surjective function s ∶ 2κ ω → αω in N; and we also have 1 ( ) a surjective function s ∶ 2κ → 2κ ω. Since αω is well-ordered by AX , 0 ( ) ( 4)

9 a diagonalization argument as in K¨onig’s Lemma shows that there is also a ω + α + surjection s2 ∶ α → α . Hence, s2 ○ s1 ○ s0 ∶ 2 → α is a surjective function in N as desired.

We conclude that all the requirements on the sequences κ 0 < η < γ and ( η S ) α 0 < η < γ listed above, are necessary for a model N ⊧ ZF + DC + AX . ( η S ) 4

In addition, one could ask if there exists a model N ⊧ ZF + DC (without AX4) with cardinals κ, α, such that θN κ = α+ and cf N α = ω. It is not diﬃcult to ( ) ( ) see that this is not possible under ¬0♯ (cf. Chapter 7). Hence, if one wishes to + avoid large cardinal assumptions, then for every αη = α a successor cardinal, one has to require cf α > ω.

Our main theorem states that these are the only restrictions on the θ-function for set-many uncountable cardinals in ZF + DC + AX4:

Theorem. Let V be a ground model of ZFC + GCH with γ ∈ Ord and sequences of uncountable cardinals κ 0 < η < γ and α 0 < η < γ such that κ 0 < η < γ ( η S ) ( η S ) ( η S ) is strictly increasing and closed, and the following properties hold:

• ∀ 0 < η < η′ < γ α ≤ α ′ , i.e. the sequence α 0 < η < γ is increasing, η η ( η S ) • + + ∀ 0 < η < γ αη ≥ κη ,

• ∀ 0 < η < γ cf αη > ω,

• ∀ 0 < η < γ α = α+ → cf α > ω . ( η )

Then there is a cardinal- and coﬁnality-preserving extension N ⊇ V with N ⊧ ZF + DC + AX such that that θN κ = α holds for all 0 < η < γ. 4 ( η) η In our construction, we will make sure that for any cardinal λ in a “gap” κ , κ , ( η η+1) the value θN λ is the smallest possible, i.e. θN λ = max α , λ++ . Also, if we ( ) ( ) { η } set κ ∶= κ 0 < η < γ , α ∶= α 0 < η < γ , then for every λ ≥ κ , we γ ⋃{ η S } γ ⋃{ η S } γ will again make sure that θN λ takes the smallest possible value: We will have ( ) θN λ = max α++, λ++ in the case that cf α = ω, θN λ = max α+, λ++ in the ( ) { γ } γ ( ) { γ } case that α = α+ for some cardinal α with cf α = ω, and θN λ = max α , λ++ , γ ( ) { γ } else.

This allows us to assume w.l.o.g. that the sequence α 0 < η < γ is strictly ( η S ) increasing: If not, one can start with the original sequences κ 0 < η < γ ( η S ) and α 0 < η < γ , and successively strike out all κ for which the value α ( η S ) η η is not larger than the values αη before. This procedure results in sequences κ 0 < η < γ ∶= κ 0 < η < γ and α 0 < η < γ ∶= α 0 < (̃η S ̃) ( s(η) S ̃) (̃η S ̃) ( s(η) S η < γ for some γ ≤ γ and a strictly increasing function s ∶ γ → γ, such that ̃) ̃ ̃ α ∶= α 0 < η < γ = α 0 < η < γ = α 0 < η < γ = α , ̃̃γ ⋃{̃η S ̃} ⋃{ s(η) S ̃} ⋃{ η S } γ

10 and α 0 < η < γ = α 0 < η < γ is strictly increasing. If we then use (̃η S ̃) ( s(η) S ̃) the sequences κ 0 < η < γ , α 0 < η < γ for our construction and make (̃η S ̃) (̃η S ̃) sure that not only θN κ = α holds for all 0 < η < γ, but additionally, θN λ (̃η) ̃η ̃ ( ) takes the smallest possible value for all cardinals λ within the “gaps” κ , κ , (̃η ̃η+1) and also make sure that θN λ takes the smallest possible value for all cardinals ( ) λ ≥ κ ∶= κ 0 < η < γ , then it follows, that for all κ in the original sequence ̃̃γ ⋃{̃η S ̃} η κ 0 < η < γ , the values θN κ = α are as desired. ( η S ) ( η) η Hence, from now on, we assume w.l.o.g. that the sequence α 0 < η < γ is ( η S ) strictly increasing.

3 The Forcing

In this chapter, we will deﬁne our forcing notion P.

We start from a ground model V ⊧ ZFC + GCH with sequences κ 0 < η < γ , ( η S ) α 0 < η < γ that have all the properties mentioned in Chapter 2. ( η S ) We will have to treat limit cardinals and successor cardinals separately. Let Lim ∶= 0 < η < γ κ is a limit cardinal , and Succ ∶= 0 < η < γ κ is a successor cardinal . { S η } { S η } + For η ∈ Succ, we denote by κη the cardinal predecessor of κη; i.e. κη = κη . Our forcing will be a product P = P0 × P1, where P0 deals with the limit cardinals κη, and P1 is in charge of the successor cardinals.

The forcing P0 is a generalized version of the forcing notion in [GK12].

Roughly speaking, for every η ∈ Lim we add αη-many κη-subsets, which will be linked in a certain fashion, in order to make sure that not too many κ-subsets for cardinals κ < κη make their way into the eventual model N.

For technical reasons, let κ0 ∶= ℵ0, α0 ∶= ℵ2. For all η with η + 1 ∈ Lim, we take a sequence of cardinals κ j < cf κ coﬁnal in κ , such that κ = κ , the ( η,j S η+1) η+1 η,0 η sequence κ j < cf κ is strictly increasing and closed, and any κ is a ( η,j S η+1) η,j+1 ++ successor cardinal with κη,j+1 ≥ κη,j for all j < cf κη+1. These “gaps” between the cardinals κη,j and κη,j+1 will be necessary for further factoring arguments.

For all 0 < η < γ for which η + 1 ∈ Succ, i.e. κη+1 is a successor cardinal, we set

κη,0 ∶= κη, and cf κη+1 ∶= 1 for reasons of homogeneity.

Now, in the case that η ∈ Lim, the forcing P η will be deﬁned like an Easton-support product of Cohen forcings for the intervals κ , κ ⊆ κ : [ ν,j ν,j+1) η Deﬁnition 1. For η ∈ Lim, we let the forcing notion P η, ⊇, ∅ consist of all func- ( ) tions p ∶ dom p → 2 such that dom p is of the following form:

11 There is a sequence δ ν < η , j < cf κ with δ ∈ κ , κ for all ν < η, ( ν,j S ν+1) ν,j [ ν,j ν,j+1) j < cf κν+1 with dom p = κν,j, δν,j , ν<η [ ) j

For a set S ⊆ κ , we let P η ↾ S ∶= p ∈ P η dom p ⊆ S = p ↾ S p ∈ P η . Then for η { S } { S } any κ < κ , the forcing P η is isomorphic to the product P η ↾κ × P η ↾ κ , κ , ν,j η ν,j [ ν,j η) + where the ﬁrst factor has cardinality ≤ κν,j, and the second factor is ≤ κν,j-closed.

This helps to establish:

Lemma 2. For all η ∈ Lim, the forcing P η preserves cardinals and the GCH.

Proof. Let Gη denote a V -generic ﬁlter on P η. It suﬃces to show that for all cardinals α in V , η 2α V [G ] ≤ α+ V , ( ) ( ) which implies that cardinals are V -V Gη -absolute: If not, there would be a V - [ ] cardinal α with a surjection s ∶ β → α in V Gη for some V Gη -cardinal β < α. [ ] [ ] Then there is also a surjection s ∶ β → β+ V in V Gη , which gives a surjection V [Gη] ( ) [ ] s ∶ β → 2β . Contradiction. η • In the case that α ≥ κ+, it follows that 2α V [G ] ≤ ´ α ⋅ P η V ≤ 2α V = η ( ) S ( S S)S ( ) α+ V by the GCH in V . ( ) • Now, assume α ∈ κ , κ for some κ < κ . Then the forcing P η can ( ν,j ν,j+1) ν,j η be factored as P η ↾ κ × P η ↾ κ , κ , where P η ↾ κ has cardinality ν,j [ ν,j η) ν,j ≤ κ+ ≤ α, and P η ↾ κ , κ is ≤ α-closed. Hence, ν,j [ ν,j η ) V [Gη ] V [Gη↾κ ] V V 2α ≤ 2α ν,j ≤ ´ α ⋅ P η ↾ κ ≤ 2α = α+ V . T S ν,jS T ( )

• If α = κ for some regular κ < κ , then P η ↾κ = κ and P η ↾ κ , κ ν,j ν,j η S ν,jS ν,j [ ν,j η) is ≤ κν,j-closed; so the same argument applies. η If α = κ is regular, then 2α V [G ] ≤ α+ V follows from P η ≤ κ . η ( ) ( ) S S η η κν,j V [G ] + V It remains to show that 2 = κν,j for all singular κν,j < κη, and η ( ) ( ) 2κη V [G ] ≤ κ+ V in the case that κ itself is singular. ( ) ( η ) η

We only prove the ﬁrst part (the argument for κη is similar).

η • Assume the contrary and let κ least with λ ∶= cf κ < κ and 2κν,j V [G ] > ν,j ν,j ν,j ( ) κ+ V . Take α i < λ coﬁnal in κ . By assumption and by what we ( ν,j) ( i S ) ν,j α V [Gη] + V have shown before, it follows that 2 = α for all α < κν,j. Hence, ( ) ( ) η CardV ∩ κ + 1 = CardV [G] ∩ κ + 1 , and 2αi V [G ] = α+ V for all ( ν,j ) ( ν,j ) ( ) ( i ) i < λ. Thus, κν,j αi λ κν,j κν,j 2 ≤ M 2 ≤ κν,j ≤ κν,j = 2 i<λ

12 holds true in V and V Gη . Let λ ∈ κ , κ for some κ < κ . If [ ] [ µ,m µ,m+1) µ,m ν,j λ > κ , then P η ↾κ ≤ κ + ≤ λ, and P η ↾ κ , κ is ≤ λ-closed. In µ,m S µ,mS ( µ,m) [ µ,m η) the case that λ = κ , it follows by regularity of λ that P η ↾κ ≤ κ = λ, µ,m S µ,mS µ,m as well. In either case,

V [Gη] V [Gη] V [Gη↾κ ] V [Gη↾κ ] 2κν,j = κλ ≤ κλ µ,m ≤ 2κν,j µ,m ≤ ν,j ν,j V V ≤ ´ κ ⋅ P η ↾κ ≤ ´ κ ⋅ κ+ = κ+ V , T ν,j S µ,mS T T ( ν,j µ,m) T ( ν,j) which gives the desired contradiction.

Corollary 3. For every η ∈ Lim, the forcing P η preserves coﬁnalites.

Proof. We show that every regular V -cardinal λ is still regular in V Gη . If not, [ ] there would be in V Gη a regular cardinal λ < λ with a cofinal function f ∶ λ → λ. [ ] Let λ ∈ κ , κ . The forcing P η is isomorphic to the product P η ↾ κ × P η ↾ [ ν,j ν,j+1) ν,j κ , κ , where the second factor is ≤ λ-closed. If λ > κ , then the first factor [ ν,j η) ν,j + has cardinality ≤ κν,j ≤ λ. In the case that λ = κν,j, the first factor has cardinality ≤ κ = λ by regularity of λ. Hence, f ∈ V G ↾ κ . However, since P η ↾ κ < λ, ν,j [ ν,j] S ν,jS it follows that λ is still a regular cardinal in the generic extension V Gη ↾ κ . [ ν,j] Contradiction. Thus, it follows that P η preserves cofinalities as desired.

σ Our eventual forcing notion P0 will contain ασ-many copies of P for every σ ∈ Lim. σ σ They will be labelled Pi , where i < ασ. All the Pi for σ ∈ Lim, i < ασ, will be γ linked with a forcing notion P∗, which is a two-dimensional version of P , adding κ -many Cohen subsets to every interval κ , κ : ν,j+1 [ ν,j ν,j+1) Deﬁnition 4. We denote by P , ⊇ ∅ the forcing notion consisting of all func- ( ∗ ) tions p∗ ∶ dom p∗ → 2 such that dom p∗ is of the following form:

There is a sequence δ ν < γ , j < cf κ with δ ∈ κ , κ for all ν < γ, ( ν,j S ν+1) ν,j [ ν,j ν,j+1) j < cf κν+1 with 2 dom p∗ = κν,j, δν,j , ν<γ [ ) j

As in Lemma 2, it follows that P∗ preserves cardinals and the GCH.

13 Now, we are ready to deﬁne our forcing notion P0. Every p0 ∈ P0 is of the form

p = p , pσ, aσ 0 ( ∗ ( i i )σ∈Lim ,i<ασ ) with p ∈ P and pσ ∈ P σ for all σ, i . ∗ ∗ i ( ) σ σ The linking ordinals ai will determine how the i-th generic κσ-subset Gi , given σ by the projection of the generic ﬁlter G onto Pi , will be eventually linked with the P∗-generic ﬁlter G∗.

Deﬁnition 5. Let P be the collection of all p = p , pσ, aσ such that: 0 0 ( ∗ ( i i )σ∈Lim ,i<ασ ) • The support of p , supp p , is countable with pσ = aσ = ∅ whenever σ, i ∉ 0 0 i i ( ) supp p0.

• We have p ∈ P , and pσ ∈ P σ for all σ, i ∈ supp p . ∗ ∗ i ( ) 0 • σ The domains of the pi are coherent in the following sense: If dom p = κ , δ 2, then for every σ, i ∈ supp p , it follows ∗ ⋃ν<γ,j

Concerning the partial ordering ≤ , any linking ordinal ξ = aσ ∩ κ , κ set- 0 { } i [ ν,j ν,j+1) tles that whenever q ≤ p , the extension qσ ⊇ pσ within in the interval κ , κ 0 0 i i [ ν,j ν,j+1) is determined by q ξ : ∗( ) For p = p , pσ, aσ , q = q , qσ, bσ ∈ P , let q ≤ p if the follow- 0 ( ∗ ( i i )σ,i) 0 ( ∗ ( i i )σ,i) 0 0 0 0 ing holds: q ⊇ p ; qσ ⊇ pσ , bσ ⊇ aσ for all σ, i ∈ supp p , and whenever ∗ ∗ i i i i ( ) 0 ζ ∈ dom qσ ∖ dom pσ ∩ κ , κ with aσ ∩ κ , κ = ξ , then ξ ∈ dom q ( i i ) [ ν,j ν,j+1) i [ ν,j ν,j+1) { } 0 with qσ ζ = q ξ,ζ (we call this the linking property ). i ( ) ∗( ) The maximal element of P is 1 ∶= ∅, ∅, ∅ . 0 0 ( ( )σ<γ,i<ασ ) Let G denote a V -generic filter on P , and gσ ∶= aσ p = p , pσ, aσ ∈ G . 0 0 i ⋃{ i S ( ∗ ( i i )σ,i) 0} Note that by our strong independence property, every interval κ , κ will be [ ν,j ν,j+1) blown up to size sup α σ ∈ Lim in a P -generic extension. { σ S } 0 Hence, since we want our eventual symmetric submodel N preserve all V -cardinals, we will have so make sure that N “does not know” the sequence of linking ordinals gσ σ ∈ Lim , i < α . ( i S σ) A major difference between our forcing and the basic construction in [GK12] is the following: The forcing conditions in [GK12, Definition 2] have finite linking

14 σ σ ordinals ai ; so the according generics gi are not contained in the ground model V . With our deﬁnition however, it follows for any p ∈ G with σ, i ∈ supp p 0 ( ) σ σ that gi = ai ∈ V . By countable support, also countable sequences of linking σ ordinals g j j < ω are contained in V ; but for σ ∈ Lim not the whole sequence ( ij S ) gσ i < α . ( i S σ) σ This modiﬁcation helps to establish that any generic Gi can be described using only G∗ and sets from the ground model V (see below).

Next, we deﬁne our forcing notion P1, which will be in charge of the successor + cardinals. For every σ ∈ Succ with κσ = κσ , it follows that σ =∶ σ + 1 must be a successor ordinal, since we have assumed in the beginning that the sequence κ 0 < σ < γ is closed. ( σ S ) We denote by P σ the Cohen forcing

P σ ∶= p ∶ dom p → 2 dom p ⊆ κ , κ , dom p < κ , { T [ σ σ) S S σ} and let

Cσ ∶= p ∶ dom p → 2 dom p = dom p × dom p ⊆ α × κ , κ , dom p < κ . { S x y σ [ σ σ) S S σ}

σ σ <κσ κσ Then both P and C are < κσ-closed, and if 2 = κσ, i.e. 2 = κσ, then they + satisfy the κσ-chain condition and hence, preserve cardinals. In particular, any forcing P σ or Cσ preserves cardinals if we are working in our ground model V with V ⊧ GCH, or any V -generic extension by ≤ κσ-closed forcing.

Deﬁnition 6. The forcing notion P , ≤ , ∅ consists of all p = pσ with ( 1 1 1) 1 ( )σ∈Succ countable support supp p ∶= σ ∈ Succ pσ ≠ ∅ , and pσ ∈ Cσ for all σ ∈ Succ. For 1 { S } p = pσ , q = qσ ∈ P , we let q ≤ p if qσ ⊇ pσ for all σ ∈ Succ; and 1 ( )σ∈Succ 1 ( )σ∈Succ 1 1 1 1 1 ∶= ∅ is the maximal element. 1 ( )σ∈Succ For σ ∈ Succ and i < α , we set pσ = ζ,pσ i, ζ i, ζ ∈ dom pσ . σ i {( ( )) S ( ) }

Our main forcing will be the product P ∶= P0 × P1 with maximal element 1 ∶= 1 , 1 and order relation ≤. In order to simplify notation, we write conditions ( 0 1) p ∈ P in the form p = p , pσ, aσ , pσ . ( ∗ ( i i )σ∈Lim,i<ασ ( )σ∈Succ) It is not diﬃcult to verify:

Proposition 7. P is countably closed.

This is important to make sure that DC holds in our eventual symmetric extension N.

η For 0 < η ≤ γ (with η ∈ Lim or η ∈ Succ or η = γ), we deﬁne a forcing P like P η is deﬁned in the case that η ∈ Lim:

15 η Let P consist of all functions p ∶ dom p → 2 such that there is a sequence δ ν < ( ν,j S η, j < cf κ with δ ∈ κ , κ for all κ < κ , and ν+1) ν,j [ ν,j ν,j+1) ν,j η

dom p = κν,j, δν,j , ν,j [ ) such that p↾κ < κ whenever κ is a regular cardinal, and p < κ in the case S ν,S ν, ν, S S η that κη itself is regular.

η λ For any 0 < η < λ with κλ a limit cardinal, it follows that P = P ↾κη.

Let now G be a V -generic ﬁlter on P. It induces

G ∶= q ∈ P ∃ p = p , pσ, aσ , pσ ∈ G ∶ q ⊆ p , ∗ { ∗ ∗ S ( ∗ ( i i )σ,i ( )σ) ∗ ∗} and for λ ∈ Lim, k < αλ:

Gλ ∶= qλ ∈ P λ ∃ p = p , pσ, aσ , pσ ∈ G ∶ qλ ⊆ pλ . k { k S ( ∗ ( i i )σ,i ( )σ) k k}

λ λ As usually, these ﬁlters G∗, Gk are identiﬁed with their union ⋃ G∗, ⋃ Gk . Then λ any Gk can be regarded a subset of κλ. Moreover, let gλ ∶= aλ p = p , pσ, aσ , pσ ∈ G . k { k S ( ∗ ( i i )σ,i ( )σ) } Then gλ = aλ for any p ∈ G with λ, k ∈ supp p ; and gλ hits any interval k k ( ) 0 k κ , κ ⊆ κ in exactly one point. By the independence property, it follows [ ν,j ν,j+1) λ λ0 λ1 that g ∩ g = ∅ whenever λ0, k0 ≠ λ1, k1 . k0 k1 ( ) ( ) For λ ∈ Succ, set

Gλ ∶= pλ p = p , pσ, aσ , pσ ∈ G , { S ( ∗ ( i i )η,i ( )σ) } and Gλ ∶= pλ p = p , pσ, aσ , pσ ∈ G k { k S ( ∗ ( i i )σ,i ( )σ) } for any k < αλ.

λ λ λ λ Again, we confuse these ﬁlters G , Gk with their union ⋃ G , ⋃ Gk . Let now ξ ∈ κ , κ . We denote by [ ν,j ν,j+1) G ξ ∶= q ∶ κ , δ → 2 δ ∈ κ , κ , ∃ p = p , pσ, aσ , pσ ∈ G ∶ ∗( ) [ ν,j ν,j) U ν,j [ ν,j ν,j+1) ( ∗ ( i i )σ,i ( )σ)

∀ ζ ∈ dom q q ζ = p ξ,ζ ( ) ∗( ) the ξ-th section of G∗.

If a ⊆ κγ is a set that hits any interval κν,j, κν,j+1 ⊆ κγ in at most one point, we γ [ ) denote by G a the set of all q ∈ P such that there is p ∈ G with q ⊆ p a . ∗( ) ∗( )

16 As before, we identify any G ξ and G a with their union G ξ and G a , ∗( ) ∗( ) ⋃ ∗( ) ⋃ ∗( ) respectively. Then any G ξ with ξ ∈ κ , κ can be regarded as a function ∗( ) [ νj ν,j+1) G ξ ∶ κ , κ → 2, and any G a becomes a function G a ∶ dom G a → ∗( ) [ ν,j ν,j+1) ∗( ) ∗( ) ∗( ) 2, where dom G a ⊆ κ is the union of those intervals κ , κ with a ∩ ∗( ) γ [ ν,j ν,j+1) κ , κ ≠ ∅. [ ν,j ν,j+1)

Now, the linking property implies that any Gλ ↾ κ , κ with λ ∈ Lim, k < α , k [ ν,j ν,j+1) λ is eventually equal to G ξ , where ξ ∶= aλ ∩ κ , κ . ∗( ) { } k [ ν,j ν,j+1) Indeed, the symmetric diﬀerence Gλ ⊕ G gλ is always an element of the ground k ∗( k ) model V : Take a condition p ∈ G with λ, k ∈ supp p , such that for any interval ( ) 0 κ , κ ⊆ κ with dom p ∩ κ , κ ≠ ∅ and ξ ∶= aλ ∩ κ , κ , it [ ν,j ν,j+1) λ 0 [ ν,j ν,j+1) { } k [ ν,j ν,j+1) follows that ξ ∈ dom p . This does not interfere with the condition that dom p 0 0 has to be bounded below all regular κν,, since we do not bother the intervals κ , κ with dom p ∩ κ , κ = ∅. [ ν,j ν,j+1) 0 [ ν,j ν,j+1)

• Firstly, Gλ ζ ⊕ G gλ ζ = 0 whenever ζ ∉ dom p : Let ζ ∈ κ , κ , k( ) ∗( k )( ) 0 [ ν,j ν,j+1) ζ ∉ dom p with ξ ∶= gλ ∩ κ , κ = aλ ∩ κ , κ . Take q ∈ G, q ≤ p { } k [ ν,j ν,j+1) k [ ν,j ν,j+1) with ζ ∈ dom q0. Then by the linking property, it follows that ξ ∈ dom q0 with qλ ζ = q ξ,ζ . Hence, Gλ ζ = qλ ζ = q ξ,ζ = G gλ ζ , and k ( ) ∗( ) k ( ) k ( ) ∗( ) ∗( k )( ) Gλ ζ ⊕ G gλ ζ = 0. k ( ) ∗( k )( ) • If ζ ∈ dom p then Gλ ζ ⊕ G gλ ζ = pλ ζ ⊕ p ξ,ζ , where again, ζ ∈ k ( ) ∗( k )( ) k( ) ∗( ) κ , κ and ξ ∶= gλ ∩ κ , κ = aλ ∩ κ , κ . Here we use [ ν,j ν,j+1) { } k [ ν,j ν,j+1) k [ ν,j ν,j+1) that for any interval κ , κ with dom p ∩ κ , κ ≠ ∅, it follows [ ν,j ν,j+1) [ ν,j ν,j+1) that aλ ∩ κ , κ ⊆ dom p . k [ ν,j ν,j+1) 0 Hence, Gλ ⊕ G gλ can be calculated in V . k ∗( k ) This will be employed to keep control over the surjective size of ´ κ in the even- ( λ) tual symmetric extension N.

σ σm m Now, we consider countable products ∏m<ω P and ∏m<ω P :

Definition 8. Let σ , i m < ω be a sequence of pairwise distinct pairs (( m m) S ) σm with 0 < σm < γ, i < ασm for all m < ω. We denote by ∏m<ω P the set of all p m m < ω with p m ∈ P σm for all m < ω (with full support), and similarly, ( ( ) S σ ) ( ) σ P m ∶= p m m < ω ∀ m < ω p m ∈ P m . ∏m<ω ( ( ) S ) T ( ) For any interval κ , κ ⊆ κ , it follows that P σm ↾ κ has cardinality [ ν,j ν,j+1) γ ∏m<ω ν,j + ≤ κν,j in the case that κν,j is regular, and cardinality ≤ κν,j, else. Moreover, P σm ↾ κ , κ is ≤ κ -closed. Hence, as in Lemma 2 and Corollary 3, ∏m<ω [ ν,j σm ) ν,j σm one can show that the product ∏m<ω P preserves cardinals, cofinalities and the GCH. σm Similarly, ∏m<ω P preserves cardinals, cofinalities and the GCH.

17 The next lemma implies that countable products G gσm are V -generic ∏m<ω ∗ im σm ( ) over ∏m<ω P :

Lemma 9. Consider a sequence a m < ω of pairwise disjoint sets such that ( m S ) for all m < ω, the following holds: am is a subset of κσm for some 0 < σm < γ, and for all κ < κ , it follows that a ∩ κ , κ = 1, i.e. a hits every interval ν,j σm S m [ ν,j ν,j+1)S m κν,j, κν,j+1 ⊆ κσm in exactly one point. Then ∏m<ω G∗ am ∶= p m m < [ ) ( )σ ( ( ) S ω ∀ m < ω p m ∈ G a is a V -generic ﬁlter on P m . ) T ( ) ∗( m) ∏m<ω

σm Proof. Let D ⊆ ∏m<ω P be an open dense set in V . We show that D ∩ G a ≠ ∅. Let ∏m<ω ∗( m) D ∶= p = p , pσ, aσ , pσ ∈ P p a m < ω ∈ D . ( ∗ ( i i )σ,i ( )σ) T ( ∗( m) S ) It suﬃces to prove that D is dense in P. Assume p = p , pσ, aσ , pσ ∈ P ( ∗ ( i i )σ,i ( )σ) given, and denote by q m < ω an extension of p a m < ω in D. We ( m S ) ( ∗( m) S ) have to construct p ≤ p such that p a ⊇ q for all m < ω. ∗( m) m Consider an interval κ , κ ⊆ κ . In the case that dom p ∪ dom q ∩ [ ν,j ν,j+1) γ ⋃m<ω m κ , κ = ∅, let δ ∶= κ . Otherwise, we pick δ ∈ κ , κ such that [ ν,j ν,j+1) ν,j ν,j ν,j [ ν,j ν,j+1) ﬁrstly,

dom p ∪ dom qm ∩ κν,j, κν,j+1 ⊆ κν,j, δν,j ; m<ω [ ) [ ) secondly, for all m < ω, it follows that a ∩ κ , κ ⊆ κ , δ ; and thirdly, m [ ν,j ν,j+1) [ ν,j ν,j ) aσ ∩ κ , κ ⊆ κ , δ for all σ, i ∈ supp p. This is possible, since the sets i [ ν,j ν,j+1) [ ν,j ν,j) ( ) a ∩ κ , κ and aσ ∩ κ , κ are singletons or empty, all the domains m [ ν,j ν,j+1) i [ ν,j ν,j+1) dom p ∩ κ , κ and dom q ∩ κ , κ for m < ω are bounded below [ ν,j ν,j+1) m [ ν,j ν,j+1) κν,j+1, and κν,j+1 is always a successor cardinal. Let

dom p0 ∶= κν,j, δν,j . ν,j [ )

Then dom p0 is bounded below all regular κν,, since this holds true for dom p and dom q . We define p on κ , δ 2 as follows: Consider an in- ⋃m<ω m ∗ ⋃ν,j[ ν,j ν,j) terval κ , δ ≠ ∅ and ξ, ζ ∈ κ , δ . For ξ,ζ ∈ dom p × dom p, let [ ν,j ν,j ) [ ν,j ν,j) ( ) p ξ,ζ ∶= p ξ,ζ . If ξ = a ∩ κ , κ for some m < ω and ζ ∈ dom q , ∗( ) ∗( ) { } m [ ν,j ν,j+1) m we set p ξ,ζ ∶= q ζ . This is not a contradiction towards p ↾ κ , δ 2 ⊇ ∗( ) m( ) ∗ [ ν,j ν,j) p ↾ κ , δ 2, since q ⊇ p a for all m < ω. Also, the a are pairwise dis- ∗ [ ν,j ν,j) m ∗( m) m joint, so for any ξ ∈ κ , δ , there is at most one m with ξ ∈ a . For all the [ ν,j ν,j) m remaining ξ,ζ ∈ dom p , we can set p ξ,ζ ∈ 0, 1 arbitrarily. This defines p ( ) ∗ ∗( ) { } ∗ on κ , δ 2. ⋃ν,j[ ν,j ν,j) For all σ, i ∈ supp p ∶= supp p , we set aσ ∶= aσ, and define pσ ⊇ pσ on the corre- ( ) 0 0 i i i i sponding domain ⋃ κν,j, δν,j according to the linking property: Whenever κν,j <κσ [ ) ζ ∈ dom p ∖ dom p ∩ κ , κ and aσ ∩ κ , κ = ξ , then ξ ∈ dom p ( 0 ) [ ν,j ν,j+1) i [ ν,j ν,j+1) { } 0

18 follows, so we can set pσ ζ ∶= p ξ,ζ . For the ζ ∈ dom pσ ∖ dom pσ remaining, we i ( ) ∗( ) i i can deﬁne pσ ζ arbitrarily. This completes the construction of p . i ( ) 0 P Let p1 ∶= p1. It is not diﬃcult to check that p ≤ p indeed is a condition in with p a ⊇ q for all m < ω. Hence, p a m < ω ∈ D, and p ∈ D as desired. ∗( m) m ( ∗( m) S )

In particular, for σ , i m < ω a sequence of pairwise distinct pairs as before (( m m) S ) with σ ∈ Lim, i < α for all m < ω, it follows that G gσm is a V -generic m m σm ∏m<ω ∗ im σm ( ) ﬁlter over ∏m<ω P .

Similarly, one can show:

Lemma 10. Let σ , i m < ω denote a sequence of pairwise distinct pairs (( m m) S ) σm with 0 < σm < γ, im < ασ for all m < ω. Then ∏ G ∶= p m m < m m<ω im ( ( ) S ω ∀ m < ω p m ∈ Gσm is a V -generic ﬁlter on ∏ P σm . ) T ( ) im m<ω

4 Symmetric names.

4.1 Outline.

For constructing our symmetric submodel N, we will define a group A of P- automorphisms and a normal filter F on A. More accurately: In our setting, an automorphism π ∈ A will not be defined on P itself, but only on a dense subset

Dπ ⊆ P. We call such π ∶ Dπ → Dπ a partial automorphism. Hence, the set A is not quite a group, but has a very similar structure: For any π, σ ∈ A with π ∶ D → D , σ ∶ D → D and p ∈ D ∩ D , the image σ p π π σ σ π σ ( ) will be an element of Dπ ∩ Dσ as well; and A will contain a map ν ∶ Dν → Dν such that Dν = Dπ ∩ Dσ and ν = π ○ σ on Dν. (We will call ν the concatenation π ○ σ.)

Moreover, for any π ∈ A, there will be a map ν in A with Dν = Dπ such that −1 π ○ ν = ν ○ π = idDν = idDπ is the identity on Dν = Dπ. (We call ν the inverse π .) There will also be an identity element id ∈ A, which is the identity map on its domain Did, where Did ⊇ Dπ for all π ∈ A.

This does not quite give a group structure: For instance, for any π ∈ A, the con- −1 −1 catenation π ○ π = π ○ π = idDπ is not really the identity element id, which usually has a larger domain Did.

In this setting, the standard approach would be using Boolean-valued models for the construction of the symmetric submodel N: Any automorphism π ∶ Dπ → Dπ can be uniquely extended to an automorphism of the complete Boolean algebra P B P , and thereby induces an automorphism of the Boolean valued model V B( ). ( ) Then one can consider the group consisting of these extended automorphisms, deﬁne a normal ﬁlter and construct the corresponding symmetric submodel as described in [Jec73, Chapter 5].

19 We try to avoid Boolean valued models here, and work with partial orders and automorphisms π ∶ Dπ → Dπ on dense subsets Dπ ⊆ P instead.

We will have a collection D of dense subsets D ⊆ P with certain properties, and a collection A of partial automorphisms π ∶ Dπ → Dπ with Dπ ∈ D for any π ∈ A. Whenever D ∈ D is ﬁxed, the automorphisms π ∈ A D = D will form a group { S π } ′ ′ that we denote by AD. Moreover, for any D, D ∈ D with D ⊆ D and π ∈ AD′ , we will have π D = D, and the restriction π ↾ D is an element of A . Hence, there [ ] D ′ are canonical homomorphisms φD′D ∶ AD′ → AD, π ↦ π ↾ D for any D, D ∈ D with D ⊆ D′. This gives a directed system, and we can take the direct limit

A ∶= lim AD = + AD ∼ Ð→ ′ with the following equivalence relation “∼”: Whenever π ∈ AD and π ∈ AD′ , then π ∼ π′ iﬀ there exists D′′ ∈ A, D′′ ⊆ D ∩ D′, such that π and π′ agree on D′′. Since for any D, D′ ∈ A, the intersection D ∩ D′ will be contained in D as well, and P is separative, this will be the case if and only if π and π′ agree on the intersection D ∩ D′.

For π ∈ A, we denote by π its equivalence class: [ ] π ∶= σ ∈ A σ ∼ π = σ ∈ A π ↾ D ∩ D = σ ↾ D ∩ D . [ ] { S } { S ( π σ) ( π σ)} Then A = π π ∈ A becomes a group as follows: For π, σ ∈ A, let π ○ σ ∶= ν , [ ] S [ ] [ ] [ ] where ν ∈ A with D = D ∩ D and ν p = π σ p for all p ∈ D ∩ D . Such ν ν π σ ( ) ( ( )) π σ will always exists by our construction of A, and ν is well-deﬁned: If π = π′ , [ ] [ ] [ ] σ = σ′ and ν, ν′ as above, then for all p ∈ D ∩ D ∩ D ′ ∩ D ′ , it follows [ ] [ ] ( π σ) ( π σ ) that ν p = π σ p = π′ σ′ p = ν′ p . Hence, ν = ν′ . ( ) ( ( )) ( ( )) ( ) [ ] [ ]

The identity element id is the identity map on its domain Did ∈ D, with Did ⊇ Dπ for all π ∈ A (then π ○ id = id ○ π = π for all π ∈ A follows). [ ] [ ] [ ] [ ] [ ] Finally, for π ∈ A, let π −1 ∶= ν , where ν ∈ A with D = D and ν = π−1 on D , [ ] [ ] ν π π i.e. ν π p = π ν p = p for all p ∈ D . Again, such ν will always exists by our ( ( )) ( ( )) π construction of A, and ν is well-defined: Whenever π = π′ and ν, ν′ as above, [ ] [ ] [ ] it follows that ν p = ν′ p must hold for all p ∈ D ∩ D ′ = D ∩ D ′ ; hence, ( ) ( ) π π ν ν ν = ν′ . Moreover, π ○ ν = ν ○ π = id = id . [ ] [ ] [ ] [ ] [ ] [ ] [ Dπ ] [ ] Hence, A is a group. Later on, we will define a collection of A-subgroups generating a normal filter F on A, giving rise to our notion of symmetry.

However, we ﬁrst have to extend our partial automorphisms π ∈ A to the name space Name P . ( ) D For any D ∈ D, we deﬁne a hierarchy Name P recursively: α( ) D • Name P ∶= ∅ 0( )

20 D D • Name P ∶= x˙ ∈ Name P x˙ ⊆ Name P × D , and α+1( ) { ( ) S α( ) } D D • Name P ∶= Name P for λ a limit ordinal. λ( ) ⋃α<λ α( ) Let D D Name P ∶= Nameα P . ( ) α∈Ord ( )

D In other words: Name P is the collection of all P-namesx ˙ where only conditions ( ) p ∈ D occur.

Whenever π ∈ A, π ∶ Dπ → Dπ, then the image πx˙ can be deﬁned es as usual as Dπ Dπ long asx ˙ ∈ Name P . In the case thatx ˙ is a P-name withx ˙ ∉ Name P , ( ) ( ) however, it is not clear how to apply π, so the namex ˙ has to be modiﬁed.

Given D ∈ D, we define recursively forx ˙ ∈ Name P : ( ) xD ∶= yD,p y˙ ∈ domx,p ˙ ∈ D,p ⊩ y˙ ∈ x˙ . {( ) S } D Then xD ∈ Name P withx ˙ G = xD G for any G a V -generic filter on P. ( ) ( ) We will call a P-namex ˙ symmetric if the collection of all π with πxDπ = xDπ is [ ] contained in our normal filter F. Hereby, we have to make sure that this definition does not depend on which rep- resentative of π we choose: In Lemma 14 and 15, we prove that whenever π and [ ] π′ belong to the same equivalence class π = π′ , then πxDπ = xDπ holds if and [ ] [ ] only if π′ xDπ′ = xDπ′ .

Then we set N ∶= V G ∶= x˙ G x˙ ∈ HS , ( ) { S } where HS denotes the class of all hereditarily symmetric P-names, deﬁned recursively as usual: For everyx ˙ ∈ Name P , we havex ˙ ∈ HS ifx ˙ is symmetric and ( ) dom x˙ ⊆ HS.

We have to verify that also with this modiﬁed notion of symmetry, N = V G is a ( ) model of ZF . This will be done in Chapter 5.

4.2 Constructing A and A.

We start with constructing A, our collection of partial P-automorphisms with the properties mentioned in Chapter 4.1.

We will have A = A0 × A1, where A0 is a collection of partial P0-automorphisms, and A1 is a collection of partial P1-automorphisms.

Every π0 ∈ A0 will be an order-preserving bijection π0 ∶ Dπ0 → Dπ0 , where Dπ0 is contained in our collection D0, deﬁned as follows:

21 Let D0 denote the collection of all sets D ⊆ P0 given by

• a countable support supp D ⊆ σ, i σ ∈ Lim, i < α , and {( ) S σ} • a domain dom D ∶= κ , δ such that δ ∈ κ , κ for ⋃ν<γ,j

bounded below κν,, such that D is the set of all p = p , pσ, aσ ∈ P with ( ∗ ( i i )σ,i) 0 • supp p ⊇ supp D , dom p ⊇ dom D , and

• for all intervals κ , κ with dom p ∩ κ , κ ≠ ∅, it follows that [ ν,j ν,j+1) [ ν,j ν,j+1) aσ ∩ κ , κ ⊆ dom p. i [ ν,j ν,j+1) (σ,i)∈supp p

In other words, D is the collection of all p ∈ P0, the domain and support of which covers a certain domain and support given by D; with the additional property that all the linking ordinals ξ = aσ ∩ κ , κ contained in any interval { } i [ ν,j ν,j+1) κ , κ hit by dom p, are already contained in dom p. [ ν,j ν,j+1)

It is not diﬃcult to see that any D ∈ D0 is dense in P0. The sets D ∈ D0 are not open dense; but whenever p, q ∈ P0 with p ∈ D and q ≤ p such that supp q = supp p, then by the linking property, it follows that also q ∈ D.

′ ′ Whenever D, D ∈ D0, then the intersection D ∩ D is contained in D0, as well, with supp D ∩ D′ = supp D ∪ supp D′, dom D ∩ D′ = dom D ∪ dom D′. ( ) ( )

We now describe the two types of partial P0-automorphisms that will generate A0:

Our ﬁrst goal is that for any two conditions p, q ∈ P with the same “shape”, i.e. σ σ dom p = dom q, supp p = supp q and ⋃ ai = ⋃ bi , there is an automorphism π0 ∈ A0 with π0p = q. This homogeneity property will be achieved by giving the maps

π0 ∈ A0 a lot of freedom regarding what can happen on supp π0 and dom π0.

For κν,j < κγ , let

supp π ν, j ∶= σ, i ∈ supp π κ < κ . 0( ) {( ) 0 S ν,j σ} Regarding the linking ordinals, we want that for any p ∈ D , p = p , pσ, aσ π0 ( ∗ ( i i )σ,i) with πp = p′ = p′ , p′ σ, a′ σ , the sets of linking ordinals for p and p′ are (( )∗ (( )i ( )i )σ,i) the same, i.e. aσ = a′ σ. In other words, for any interval κ , κ , the ⋃ i ⋃( )i [ ν,j ν,j+1) linking ordinals ξ ∈ κ , κ will be exchanged between the coordinates σ, i ∈ [ ν,j ν,j+1) ( ) supp π ν, j , which is described by an isomorphism F ν, j ∶ supp π ν, j → 0( ) π0 ( ) 0( ) supp π ν, j . 0( )

22 Regarding the p′ σ for σ, i ∈ supp π , there will be for every ζ ∈ κ , κ ∩ ( )i ( ) 0 [ ν,j ν,j+1) dom π a bijection π ζ ∶ 2supp π0(ν,j) → 2supp π0(ν,j) with 0 0( )

p′ σ ζ σ, i ∈ supp π ν, j ∶= π ζ pσ ζ σ, i ∈ supp π ν, j . ( )i ( ) S ( ) 0( ) 0( ) i ( ) S ( ) 0( )

Concerning p′ , we will have a similar construction for the p′ ξ,ζ in the case ∗ ∗( ) that ζ ∈ dom π and ξ is a linking ordinal contained in aσ. For all ξ,ζ ∈ 0 ⋃ i ( ) dom π ∩ κ , κ 2, we will have a bijection π ξ,ζ ∶ 2 → 2, with p′ ξ,ζ = 0 [ ν,j ν,j+1) ∗( ) ∗( ) π ξ,ζ p ξ,ζ whenever ξ, ζ ∈ dom π and ξ ∉ aσ. ∗( )( ∗( )) 0 ⋃ i Our second goal is that for any interval κ , κ and σ, i , λ, k ∈ supp π ν, j , [ ν,j ν,j+1) ( ) ( ) 0( ) there is an isomorphism π ∈ A such that π G λ ∩ κ , κ = Gσ ∩ κ , κ . 0 0 ( 0 )k [ ν,j ν,j+1) i [ ν,j ν,j+1) Thus, every π ∈ A will be equipped with bijections G ν, j ∶ supp π ν, j → 0 0 π0 ( ) 0( ) suppπ ν, j for every κ , such that the following holds: For every p ∈ D , 0( ) ν,j π0 p′ ∶= πp and ζ ∈ dom p ∖ dom π , σ, i ∈ supp π ν, j , we have p′ σ ζ = pλ ζ 0 ( ) 0( ) ( )i ( ) k( ) with λ, k ∶= G ν, j σ, i . ( ) π0 ( )( ) Whenever ζ ∈ dom π and σ, i ∈ supp π ν, j , the values p′ σ ζ are described 0 ( ) ( ) ( )i ( ) by the maps π ζ mentioned in context with “our ﬁrst goal” above, which al- 0( ) lows for setting p′ σ ζ ∶= pλ ζ for any pair σ, i , λ, k ∈ supp π ν, j with ( )i ( ) k( ) ( ) ( ) 0( ) λ, k = G ν, j σ, i . ( ) π0 ( )( )

Roughly speaking, A0 will be generated by these two types of isomorphism. Re- ′ garding the construction of p∗, some extra care is needed concerning the values p′ ξ,ζ for ζ ∉ dom π and ξ ∈ aσ a linking ordinal, since we have to make sure ∗( ) 0 ⋃ i that the maps π0 ∈ A0 are order-preserving: Whenever p,q ∈ Dπ0 with q ≤ p, then also q′ ≤ p′ must hold; in particular, whenever ξσ ∶= aσ ∩ κ , κ is a linking { i } i [ ν,j ν,j+1) ordinal and ζ ∈ dom q ∖ dom p (hence, ζ ∉ dom π ), then ξσ = a′ λ ∩ κ , κ 0 { i } ( )k [ ν,j ν,j+1) with σ, i = F ν, j λ, k , and q′ ξσ,ζ = q′ λ ζ by the linking property for ( ) π0 ( )( ) ∗( i ) ( )k ( ) q′ ≤ p′. Moreover, q′ λ ζ = qµ ζ with µ, l = G ν, j λ, k , and qµ ζ = ( )k( ) l ( ) ( ) π0 ( )( ) l ( ) q ξµ,ζ with ξµ = aµ ∩ κ , κ by the linking property for q ≤ p. Hence, ∗( l ) l l [ ν,j ν,j+1) q′ ξσ,ζ = q ξµ,ζ must hold, where µ, l = G ν, j ○ F ν, j −1 σ, i . ∗( i ) ∗( l ) ( ) π0 ( ) ( π0 ( )) ( ) This gives rise to the following deﬁnition:

Deﬁnition 11. Let A0 consist of all automorphisms π0 ∶ Dπ0 → Dπ0 such that there are

• a countable set supp π ⊆ σ, i σ ∈ Lim, i < α 0 {( ) S σ} (for κ < κ , we set supp π ν, j ∶= σ, i ∈ supp π κ < κ ), ν,j γ 0( ) {( ) 0 S ν,j γ }

• a domain dom π = κ , δ such that δ ∈ κ , κ for 0 ⋃ν<γ,j

bounded below κν, (for κ < κ , we set dom π ν, j ∶= dom π ∩ κ , κ ), ν,j γ 0( ) 0 [ ν,j ν,j+1)

23 such that

D = p = p , pσ, aσ ∈ P suppp ⊇ suppπ , dom p ⊇ dom π , and π0 ( ∗ ( i i )σ,i) 0 T 0 0

∀ κ < κ ∶ dom p ∩ κ , κ ≠ ∅ ⇒ aσ ∩ κ , κ ⊆ dom p ; ν,j γ [ ν,j ν,j+1) i [ ν,j ν,j+1) (σ,i)∈supp p moreover, there are

• for all ν < γ, j < cf κν+1, a bijection

F ν, j ∶ supp π ν, j → supp π ν, j π0 ( ) 0( ) 0( )

(which will be in charge of permuting the linking ordinals as mentioned above),

and a bijection

G ν, j ∶ supp π ν, j → supp π ν, j π0 ( ) 0( ) 0( )

σ (which will be in charge of permuting the verticals pi outside dom π0 on the interval κ , κ ), [ ν,j ν,j+1) • for all ν < γ, j < cf κ and ζ ∈ κ , κ ∩ dom π , a bijection ν+1 [ ν,j ν,j+1) 0

π ζ ∶ 2supp π0(ν,j) → 2supp π0(ν,j) 0( )

(this map will be in charge of setting the values πp σ ζ for σ, i ∈ suppπ ν, j , ( )i ( ) ( ) 0( ) ζ ∈ dom π0),

• for all ν < γ, j < cf κ , ζ ∈ κ , κ ∩ dom π , and ν+1 [ ν,j ν,j+1) 0

ξσ σ, i ∈ supp π ν, j ∈ dom π ν, j supp π0(ν,j) i S ( ) 0( ) ( 0( ))

a sequence of pairwise distinct ordinals, a bijection

π ζ ξσ σ, i ∈ suppπ ν, j ∶ 2supp π0(ν,j) → 2supp π0(ν,j) ( 0)∗( ) i S ( ) 0( )

(which will be in charge of setting the values πp ξσ,ζ for ξσ = aσ ∩ ( )∗( i ) { i } i κ , κ a linking ordinal and ζ ∈ κ , κ ∩ dom π ), [ ν,j ν,j+1) [ ν,j ν,j+1) 0 • for all ν < γ, j < cf κ and ξ,ζ ∈ κ , κ 2, a bijection ν+1 ( ) [ ν,j ν,j+1)

π ξ,ζ ∶ 2 → 2 ( 0)∗( )

such that π ξ,ζ is the identity whenever ξ,ζ ∉ dom π 2 ∗( ) ( ) ( 0) (which will be in charge of the values πp ξ,ζ in the case that ξ ∉ aσ ( )∗( ) ⋃σ,i i is not a linking ordinal);

24 which deﬁnes for p ∈ D , p = p , pσ, aσ , the image πp =∶ p′ = p′ , p′ σ, a′ σ π0 ( ∗ ( i i )σ,i) ( ∗ (( )i ( )i )σ,i) as follows:

We will have supp p′ = supp p, dom p′ = dom p. Moreover:

• Concerning the linking ordinals, for all σ, i ∈ supp p′ = supp p and κ < κ : ( ) ν,j σ – a′ σ ∩ κ , κ = aσ ∩ κ , κ for σ, i ∉ supp π ν, j , ( )i [ ν,j ν,j+1) i [ ν,j ν,j+1) ( ) 0( ) – a′ σ ∩ κ , κ = aλ ∩ κ , κ with λ, k = F ν, j σ, i in ( )i [ ν,j ν,j+1) k [ ν,j ν,j+1) ( ) π0 ( )( ) the case that σ, i ∈ supp π ν, j . ( ) 0( ) • Concerning the p′ σ with σ, i ∈ supp π , ( )i ( ) 0

– for ζ ∈ dom π0,

p′ σ ζ σ, i ∈ supp π ν, j = π ζ pσ ζ σ, i ∈ supp π ν, j , ( )i ( ) S ( ) 0( ) 0( ) i ( ) S ( ) 0( )

– and in the case that ζ ∉ dom π0, we will have

p′ σ ζ = pλ ζ with λ, k = G ν, j σ, i . ( )i ( ) k( ) ( ) π0 ( )( ) • Whenever σ, i ∉ supp π , then p′ σ = pσ. ( ) 0 ( )i i • We now turn to p′ . Consider an interval κ , κ . For any σ, i ∈ ∗ [ ν,j ν,j+1) ( ) supp π ν, j , let ξσ ∶= aσ ∩ κ , κ . For ζ ∈ κ , κ ∩ dom π , 0( ) { i } i [ ν,j ν,j+1) [ ν,j ν,j+1) 0 we will have p′ ξσ,ζ σ, i ∈ supp π ν, j = ∗( i ) S ( ) 0( ) π ζ ξσ σ, i ∈ supp π ν, j p ξσ,ζ σ, i ∈ supp π ν, j . ( 0)∗( ) i S ( ) 0( ) ∗( i ) S ( ) 0( ) In the case that ζ ∈ κ , κ ∩ dom p ∖ dom π , we will have for σ, i ∈ [ ν,j ν,j+1) ( 0) ( ) supp π ν, j : 0( ) p′ ξσ,ζ ∶= p ξλ,ζ , ∗( i ) ∗( k ) where λ, k = G ν, j ○ F ν, j −1 σ, i . ( ) π0 ( ) ( π0 ( )) ( ) Finally, if ξ,ζ ∈ dom p 2 with ξ,ζ ∈ κ , κ such that ξ ∉ aσ, then ( ) ( ) [ ν,j ν,j+1) ⋃σ,i i p′ ξ,ζ = π ξ,ζ p ξ,ζ . ∗( ) ( 0)∗( ) ∗( )

For any π ∈ A0, we have Dπ0 ∈ D0 with supp Dπ0 ∶= supp π0 and dom Dπ0 ∶= dom π0. ′ P Moreover, whenever p is a condition in Dπ0 , then p ∶= π0p ∈ 0 is well-deﬁned with p′ ∈ D , since supp p′ = supp p, dom p′ = dom p, and aσ = a′ σ by π0 ⋃σ,i i ⋃σ,i( )i construction. P Here we use that π0 is only deﬁned on Dπ0 and not on the entire forcing 0, since we have to make sure that in our construction of the p′ ξσ,ζ for ζ ∉ dom π , we ∗( i ) 0 do not run out of dom p.

′ ′ It is not diﬃcult to see that for any p, q ∈ Dπ0 with q ≤ p, also q ≤ p holds. The linking property follows readily from our deﬁnition of the p′ ξσ,ζ for ζ ∉ dom π . ∗( i ) 0

25 Whenever π0 ∈ A0 and D ∈ D0 with D ⊆ Dπ0 , it follows that the map π0 ∶= π0 ↾ D is contained in A0, as well. Here we have to use that the maps π0 do not disturb the conditions’ domain or support, and merely permute the linking ordinals. In particular, whenever p ∈ D, it follows that the image π0p is contained in D, as well.

It remains to verify that A0 can be endowed with a group structure. More precisely: We will show that for any D ∈ D , the collection A ∶= π ∈ A D = D gives 0 ( 0)D { 0 S π0 } a group; and then take the direct limit A0 ∶= lim A0 D. Ð→ ( )

Firstly, for any π0 ∈ A0, it is not diﬃcult to write down a map ν0 ∈ A0 with

Dν0 = Dπ0 such that ν0 is the inverse of π0:

Let supp ν ∶= supp π and dom ν ∶= dom π . For any κ < κ , we set F ν, j ∶= 0 0 0 0 ν,j γ ν0 ( ) F ν, j −1, G ν, j ∶= G ν, j −1; and for ζ ∈ κ , κ , we let ν ζ ∶= ( π0 ( )) ν0 ( ) ( π0 ( )) [ ν,j ν,j+1) 0( ) π ζ −1. Regarding ν we use the following notation: ( 0( )) ( 0)∗ For sets I, J with a bijection b ∶ I → J and a sequence x j ∈ J , we denote by ( j S ) b x j ∈ J the induced sequence parametrized by I: ( j S ) b x j ∈ J ∶= y i ∈ I with y ∶= x for all i ∈ I. ( j S ) ( i S ) i b(i) Whenever ζ ∈ κ , κ ∩ dom π , and [ ν,j ν,j+1) 0 ξσ σ, i ∈ suppπ ν, j ∈ dom π ν, j supp π0(ν,j) i S ( ) 0( ) ( 0( )) is a sequence of pairwise distinct ordinals, we set ν ξσ σ, i ∈ supp π ν, j ∶= ( 0)∗ i S ( ) 0( )

−1 −1 −1 F ν, j ○ π F ν, j ξσ σ, i ∈ suppπ ν, j ○ F ν, j , π0 ( ) ( 0)∗ π0 ( ) i S ( ) 0( ) π0 ( ) which is a bijection on 2supp π0(ν,j).

−1 For ξ,ζ ∈ κ , κ 2, let ν ξ,ζ ∶= π ξ,ζ . ( ) [ ν,j ν,j+1) ( 0)∗( ) ( 0)∗( ) It is not diﬃcult to verify that indeed, π ν p = ν π p = p holds for all 0( 0( )) 0( 0( )) p ∈ Dπ0 = Dν0 .

Secondly, one has to make sure that for any π0 ∶ Dπ0 → Dπ0 , σ0 ∶ Dσ0 → Dσ0 in A0, there is a map τ ∈ A with D = D ∩ D such that τ p = π σ p holds for 0 0 τ0 π0 σ0 0( ) 0( 0( )) all p ∈ Dτ0 . Setting dom τ0 ∶= dom π0 ∪ dom σ0, supp τ0 ∶= supp π0 ∪ supp σ0, one can write down τ explicitly (similarly as for the inverse map), and verify that τ p = π σ p 0 0( ) 0( 0( )) holds for all p ∈ Dτ0 .

Finally, the identity element id0, which is the identity map on its domain Did0 ∶=

p ∈ P ∀ κ < κ ∶ dom p ∩ κ , κ ≠ ∅ ⇒ aσ ∩ κ , κ ⊆ dom p , 0 T ν,j γ [ ν,j ν,j+1) i [ ν,j ν,j+1) (σ,i)∈supp p

26 is contained in A0 as well with Did0 ⊇ Dπ0 for all π0 ∈ A0.

It follows that for any D ∈ D , the collection A ∶= π ∈ A D = D gives 0 ( 0)D { 0 0 S π0 } ′ a group with the identity element idD ∶= id0 ↾ D; and whenever D, D ∈ D with D ⊆ D′, then there is a canonical homomorphism φ ′ ∶ A ′ → A that D D ( 0)D ( 0)D maps any π ∈ A ′ to its restriction π ↾ D. We take the direct limit 0 ( 0)D 0

A0 ∶= lim A0 D = + A0 D ∼ , Ð→ ( ) ( )

′ ′ with π0 ∼ π iﬀ π0 p = π p holds for all p ∈ Dπ ∩ Dπ′ . Then A0 becomes a 0 ( ) 0( ) 0 0 group with concatenation and the inverse elements as described in Chapter 4.1, and the identity element id ∶= id as deﬁned above. 0 [ 0]

Now, we turn to P1 and deﬁne A1, our collection of partial P1-isomorphisms. Ev- ery π1 ∈ A1 will be a bijection π1 ∶ Dπ1 → Dπ1 with a dense set Dπ1 ∈ D1, where D1 is deﬁned as follows:

Let D1 denote the collection of all D ⊆ P1 given by:

• a countable support supp D ⊆ Succ, and

• for every σ ∈ supp D, κ = κ +, a domain dom D σ = dom D σ × dom D σ ⊆ σ σ ( ) x ( ) y ( ) α × κ , κ with dom π σ < κ , σ [ σ σ) S 1( )S σ such that

D = p ∈ P suppp ⊇ suppD ∧ ∀ σ ∈ supp D dom pσ ⊇ dom D σ . { 1 S ( )} ′ Then every set D ∈ D1 is open dense; and whenever D, D ∈ D1, then the intersection D ∩ D′ is contained in D as well, with supp D ∩ D′ = supp D ∪ supp D′, and 1 ( ) dom D ∩ D′ σ = dom D σ ∪ dom D′ σ , dom D ∩ D′ σ = dom D σ ∪ x( )( ) x ( ) x ( ) y( )( ) y ( ) dom D′ σ for all σ ∈ supp D ∩ D′ . y ( ) ( )

We now describe the two types of partial P1-isomorphisms that will generate A1:

Like for A0, our ﬁrst goal is that for any p,q ∈ P1 which have “the same shape”, i.e. supp p = supp q and dom pσ = dom qσ for every σ ∈ supp p, there is an isomorphism

π1 ∈ A1 with π1p = q. These isomorphisms will be of the following form: For every σ ∈ supp π , we will have a collection of π σ i, ζ ∶ 2 → 2 for i, ζ ∈ dom π σ , 1 ( )( ) ( ) 1( ) such that for any p ∈ D , the map π changes the value of pσ i, ζ if and only if π1 1 ( ) π σ i, ζ ≠ id. In other words, π p σ i, ζ = π σ i, ζ pσ i, ζ . 1( )( ) ( 1 ) ( ) 1( )( ) ( ) This allows for constructing an isomorphism π1 with π1p = q for any pair of conditions p, q that have the same supports and domains: One can simply set π σ i, ζ = id if pσ i, ζ = qσ i, ζ , and π σ i, ζ ≠ id in the case that pσ i, ζ ≠ 1( )( ) ( ) ( ) 1( )( ) ( ) qσ i, ζ . ( )

σ σ ′ Secondly, for every pair of generic κσ-subsets Gi and Gi′ for σ ∈ Succ and i, i < ασ, σ σ we want an isomorphism π ∈ A1 such that πGi = Gi′ . Therefore, we include into A1

27 all isomorphisms π = π σ σ ∈ supp π such that for every σ ∈ supp π , there is 1 ( 1( ) S 1) 1 a bijection f σ on a countable set suppπ σ ⊆ α ; and π is deﬁned as follows: π1 ( ) 1( ) σ 1 Whenever p ∈ D , then π p σ i, ζ = pσ f σ i , ζ for all i, ζ ∈ dom pσ. π1 ( 1 ) ( ) π1 ( )( ) ( ) Then πGσ = Gσ . i fπ1 (σ)(i)

Roughly speaking, A1 will be generated by these two types of isomorphisms. In order to retain a group structure, the values π p σ i, ζ for i, ζ ∈ dom π σ ( 1 ) ( ) ( ) 1( ) and i ∈ supp π σ have to be treated separately: For every ζ ∈ dom π σ , there 1( ) y 1( ) will be a bijection π ζ ∶ 2supp π1(σ) → 2supp π1(σ) such that π p σ i, ζ i ∈ 1( ) ( 1 ) ( ) S supp π σ = π ζ pσ i, ζ i ∈ supp π σ . 1( ) 1( ) ( ) S 1( ) This gives the following deﬁnition:

Deﬁnition 12. A consists of all isomorphisms π ∶ D → D , π = π σ σ ∈ 1 1 π1 π1 1 ( 1( ) S supp π with countable support supp π ⊆ Succ, such that for all σ ∈ supp π , 1) 1 1 + κσ = κσ , there are

• a countable set supp π σ ⊆ α with a bijection f σ ∶ supp π σ → 1( ) σ π1 ( ) 1( ) supp π σ , 1( ) • a domain dom π σ = dom π σ × dom π σ ⊆ α × κ , κ with dom π σ < 1( ) x 1( ) y 1( ) σ [ σ σ) S 1( )S κ , such that suppπ σ ⊆ dom π σ , σ 1( ) x 1( ) • for every i, ζ ∈ α × κ , κ a bijection π σ i, ζ ∶ 2 → 2, with π σ i, ζ = ( ) σ [ σ σ) 1( )( ) 1( )( ) id whenever i, ζ ∉ dom π σ , and ( ) 1( ) • for every ζ ∈ dom π σ a bijection π ζ ∶ 2supp π1(σ) → 2supp π1(σ) y 1( ) 1( ) with

D = p ∈ P supp p ⊇ supp π ∧ ∀ σ ∈ supp π dom pσ ⊇ dom π σ , π1 { 1 S 1 1 1( )} and for every p ∈ Dπ1 , the image π1p is deﬁned as follows:

We will have supp π p = supp p with π p σ = pσ whenever σ ∉ supp π . Moreover, ( 1 ) ( 1 ) 1 for σ ∈ suppπ1,

• for every i, ζ ∈ dom pσ with i ∉ supp π σ , we have π p σ i, ζ = π σ i, ζ pσ i, ζ , ( ) 1( ) ( 1 ) ( ) 1( )( ) ( ) • for every i ∈ supp π σ and ζ ∈ dom pσ ∖ dom π σ , 1( ) y y 1( )

π p σ i, ζ = pσ f σ i ,ζ , and ( 1 ) ( ) π1 ( )( )

• for all ζ ∈ dom π σ , y 1( )

π p σ i, ζ i ∈ supp π σ = π ζ pσ i, ζ i ∈ suppπ σ . ( 1 ) ( ) S 1( ) 1( ) ( ) S 1( )

28 In other words: Outside the domain dom π σ , we have a swap of the horizontal 1( ) lines pσ i, ⋅ for i ∈ suppπ σ , according to f σ . If ζ ∈ dom π σ , then the ( ) 1( ) π1 ( ) 1( ) values π p σ i, ζ for i ∈ supp π σ are given by the map π ζ . Any remaining ( 1 ) ( ) 1( ) 1( ) value π p σ i, ζ with i ∉ supp π σ is equal to pσ ζ, i or not, depending on ( 1 ) ( ) 1( ) ( ) whether π σ i, ζ ∶ 2 → 2 is the identity or not. 1( )( ) We need the dense sets D in order to make sure that dom π p σ = dom pσ. In π1 ( 1 ) particular, we do not want to run out of dom pσ when permuting the pσ i, ⋅ for x ( ) i ∈ supp π σ . 1( )

It is not diﬃcult to see that any map π1 ∶ Dπ1 → Dπ1 as in Deﬁnition 12 is order- preserving.

Whenever π1 ∈ A1 and D ∈ D1 with D ⊆ Dπ1 , then the map π1 ∶= π1 ↾ D is contained in A1, as well. Here we have to use that the maps π1 do not disturb the conditions’ support or domain. In particular, whenever p ∈ D, it follows that π1p is contained in D, as well.

It remains to verify that A1 can be endowed with a group structure, which happens similarly as for A0:

Firstly, for any π1 ∈ A1, π1 ∶ Dπ1 → Dπ1 , one can write down a map ν1 ∈ A1 with

Dν1 = Dπ1 such that ν1 is the inverse of π1.

Secondly, whenever π1,σ1 ∈ A1, π1 ∶ Dπ1 → Dπ1 , σ1 ∶ Dσ1 → Dσ1 , one can explicitly write down a map τ ∈ A with D = D ∩ D such that τ p = π σ p holds 1 1 τ1 π1 σ1 1( ) 1( 1( )) for all p ∈ Dτ1 . Thirdly, A1 contains the identity element id1, which is the identity on its domain P Did1 = 1.

As before, we set A ∶= π ∈ A D = D for D ∈ D , and take the direct ( 1)D { 1 1 S π1 } 1 limit

A1 ∶= lim A1 D = + A1 D ∼ , Ð→ ( ) ( ) ′ ′ with π1 ∼ π iﬀ π1 p = π p holds for all p ∈ Dπ ∩ Dπ′ . Then A1 becomes a 1 ( ) 1( ) 1 1 group with the identity element id ∶= id . 1 [ 1] Now, we can deﬁne our group A:

Definition 13. Let D ∶= D0 × D1 and A ∶= A0 × A1; i.e. A is the collection of all π = π ,π with π ∈ A and π ∈ A as defined above, and domain D = ( 0 1) 0 0 1 1 π Dπ0 × Dπ1 ∈ D. Let A ∶= A0 × A1. This is a group with the identity element id = id , id . ( 0 1) D For D ∈ D, we define Name P as in Chapter 4.1. For D ∈ D andx ˙ ∈ Name P , ( ) ( ) we define recursively:

xD ∶= yD,p y˙ ∈ domx,p ˙ ∈ D,p ⊩ y˙ ∈ x˙ . {( ) S }

29 D Then xD ∈ Name P ; and for any G a V -generic ﬁlter on P, it follows that ( ) xD G = x˙ G. ( ) It is not diﬃcult to check that whenever D, D′ ∈ D andx ˙ ∈ Name P , then ′ ( ) D ′ xD = xD .

The next lemma is important to establish a notion of symmetry that is coherent with the equivalence relation ∼:

Lemma 14. Let π, π′ ∈ A with π ∼ π′, i.e. π ↾ D ∩ D ′ = π′ ↾ D ∩ D ′ . ( π π ) ( π π ) Then for x˙ ∈ Name P , it follows that πxDπ = xDπ if and only if π′xDπ′ = xDπ′ . ( ) We prove the following more general statement by induction over α:

Lemma 15. Let π, π′ ∈ A with π ∼ π′, i.e. π ↾ D ∩ D ′ = π′ ↾ D ∩ D ′ , ( π π ) ( π π ) and α ∈ Ord. Then for any y˙, z˙ ∈ Name P with rky ˙ = rkz ˙ = α, it follows that ( ) πyDπ = zDπ if and only if π′yDπ′ = zDπ′ .

Proof. W.l.o.g. we can assume that D ′ ⊆ D , since the map π ∶= π ↾ D ∩ D ′ = π π ̃ ( π π ) π′ ↾ D ∩ D ′ is contained in A as well, with D = D ∩ D ′ . ( π π ) ̃π π π Consider α ∈ Ord, and assume that the statement is true for all β < α. Lety ˙, z˙ ∈ Name P with rky ˙ = rkz ˙ = α. ( ) “⇒”: First, assume that πyDπ = zDπ . We only prove zDπ′ ⊆ π′yDπ′ ; the other inclusion is similar.

D ′ D ′ Let x π , p ∈ z π , i.e.x ˙ ∈ domz ˙, p ∈ D ′ , and p ⊩ x˙ ∈ z˙. Then also ( ) π p ∈ D holds. Hence, xDπ , p ∈ zDπ , and zDπ = πyDπ by assumption; so π ( ) there must beu ˙ ∈ domy ˙ with xDπ = πuDπ . Setting q ∶= π−1p, it follows that q ⊩ uDπ ∈ yDπ and q ⊩ u˙ ∈ y˙.

Since rku ˙ = rkx ˙ < α, our inductive assumption implies that xDπ′ = π′uDπ′ holds. Hence, xDπ′ , p = πuDπ′ ,πq , which is contained in π′yDπ′ , since ( ) ( ) −1 u˙ ∈ domy ˙, q ∈ Dπ′ (since p ∈ Dπ′ and q = π p), and q ⊩ u˙ ∈ y˙.

“⇐”: Now, we assume that π′yDπ′ = zDπ′ . As before, we only prove the inclusion zDπ ⊆ πyDπ . Consider xDπ , p ∈ zDπ , i.e.x ˙ ∈ domz ˙, p ∈ D and p ⊩ x˙ ∈ z˙. Let p ≤ p with ( ) π ̃ D ′ D ′ D ′ p ∈ D ′ . Then x π , p ∈ z π = π′y π , so there must beu ˙ ∈ domy ˙ with ̃ π ( ̃) xDπ′ = π′uDπ′ . By the inductive assumption, it follows that xDπ = πuDπ , since rku ˙ = rkx ˙ < α. Let q ∶= π−1p. We have to show that πuDπ ,πq ∈ πyDπ . ( ) Sinceu ˙ ∈ domy ˙ and q ∈ Dπ, it suﬃces to verify that q ⊩ u˙ ∈ y˙. We prove

that whenever r ≤ q, r ∈ Dπ′ , then r ⊩ u˙ ∈ y˙. Consider r ≤ q with r ∈ Dπ′ . D ′ D ′ Then πr ∈ D ′ , and πr ≤ p implies that πr ⊩ x˙ ∈ z˙. Hence, x π , πr ∈ z π , π ( ) and zDπ′ = π′yDπ′ by assumption. Now, π′uDπ′ ,π′r = xDπ′ , πr ∈ π′yDπ′ ( ) ( ) implies that r ⊩ uDπ′ ∈ yDπ′ ; hence, r ⊩ u˙ ∈ y˙ as desired.

30 4.3 Constructing F. Now, we deﬁne our collection of A-subgroups that will generate a normal ﬁlter F on A, establishing our notion of symmetry.

We will introduce two diﬀerent types of subgroups.

Firstly, for any 0 < η < γ, i < αη (with η ∈ Lim or η ∈ Succ), let

F ix η, i ∶= π ∈ A ∀p ∈ D πp η = pη . ( ) [ ] S π ( )i i Whenever π ∼ π′, it follows that πp η = pη for all p ∈ D if and only if π′p η = pη ( )i i π ( )i i for all p ∈ D ′ . Hence, F ix η, i is well-deﬁned, and clearly, any F ix η, i is a π ( ) ( ) subgroup of A.

By including F ix η, i into our ﬁlter F, we make sure that any canonical name ( ) ˙ η η ˙ η ˙ η Gi for the i-th generic κη-subset Gi is hereditarily symmetric, since πGi = Gi for all π ∈ F ix η, i . Hence, our eventual model N will contain any generic κη-subset η ( ) Gi .

Now, we turn to the second type of A-subgroup. For any 0 < λ < γ and k < αλ (with λ ∈ Lim or λ ∈ Succ), we need in N a surjection s ∶ ´ κ → k in order to make ( λ) sure that θN κ ≥ α . However, the sequence Gλ i < α must not be included ( λ) λ ( i S λ) into N, since θN κ ≤ α , so N must not contain a surjection s ∶´ κ → α . ( λ) λ ( λ) λ The idea (which appears in [GK12] in a slightly different setting, and in [FK16] similar as here) is that for any 0 < λ < γ and k < αλ, we define a “cloud” around each Gλ for i ≤ k, denoted by Gλ (k), and make sure that the “sequence of clouds” i (Èi ) Gλ (k) i < k makes its way into N. ( (Èi ) S ) When defining these subgroups, we have to treat limit cardinals and successor cardinals separately.

For λ ∈ Lim, k < αλ, let

Hλ ∶= π ∈ A ∃ κ < κ ∀ κ ∈ κ , κ ∀ i ≤ k ∶ k [ ] U ν, λ ν,j [ ν, λ)

λ, i ∉ supp π ν, j ∨ G ν, j λ, i = λ, i . ( ) 0( ) π0 ( )( ) ( ) λ It is not difficult to verify that any Hk is well-defined and indeed a subgroup of A. Roughly speaking, Hλ contains all π ∈ A such that above some κ < κ , there k [ ] ν, λ is no permutation of the vertical lines P λ ↾ κ , κ for i ≤ k. i [ ν, λ) This implies that for any i, j < k with i ≠ j and π ∈ Hλ, it is not possible that [ ] k λ λ λ πGi = Gj . Hence, for any i < k, we can define a “cloud” around Gi as follows:

˙ Dπ GÈλ (k) ∶= πGλ , 1 π ∈ Hλ . ( i ) i S [ ] k

31 ˙ G With Gλ (k) ∶= Gλ (k) , it follows that Gλ (k) is the orbit of Gλ under Hλ; (Èi ) (Èi ) (Èi ) i k so two distinct orbits Gλ (k) and Gλ (k) for i ≠ j are disjoint. The sequence (Èi ) (Èj ) Gλ (k) i < k , which has a canonical symmetric name stabilized by all π with (Èi ) S π ∈ Hλ, gives a surjection s ∶´ κ → k in N (see Chapter 6.1). [ ] k ( λ)

Now, we consider the case that λ ∈ Succ. For k < αλ, let

Hλ ∶= π ∈ A ∀ i ≤ k i ∉ supp π λ ∨ f λ i = i . k [ ] S 1( ) π1 ( )( )

λ Again, one can easily check that Hk is well-deﬁned and indeed an A-subgroup.

Whenever π is contained in Hλ, then π does not interchange any Gλ and Gλ for [ ] k i j i, j < k in the case that i ≠ j. Thus, as for λ ∈ Lim, we can define “clouds” Gλ (k) (Èi ) for i ≤ k and obtain a surjection s ∶´ κ → k in N (see Chapter 6.1). ( η) We are now ready to define our normal filter F on A. Note that the F ix η, i ( ) and Hλ are not normal A-subgroups: For instance, if π ∈ F ix η, i for some k [ ] ( ) η ∈ Lim, i < α , and σ ∈ A with G ν, j η, i = η, i′ for all κ , < κ such that η σ0 ( )( ) ( ) ν,j η π ∉ F ix η, i′ , then in general, σ −1 π σ is not contained in F ix η, i . [ ] ( ) [ ] [ ][ ] ( ) However, it is not difficult to verify:

Lemma 16. • For all σ ∈ A, and η ∈ Lim, i < αη,

σ F ix η, i σ −1 ⊇ F ix η, i ∩ F ix η , i m < ω, η , i ∈ supp σ . [ ] ( )[ ] ( ) { ( m m) S ( m m) 0}

In the case that σ ∈ A, and η ∈ Succ, i < αη,

σ F ix η, i σ −1 ⊇ F ix η, i ∩ F ix η, i m < ω, i ∈ supp σ η . [ ] ( )[ ] ( ) { ( m) S m 1( )}

• For σ ∈ A and λ ∈ Lim, k < αλ,

σ Hλ σ −1 ⊇ Hλ ∩ F ix η , i m < ω, η , i ∈ supp σ . [ ] k [ ] k { ( m m) S ( m m) 0}

In the case that λ ∈ Succ, k < αλ,

σ Hλ σ −1 ⊇ Hλ ∩ F ix λ, i m < ω, i ∈ supp σ λ . [ ] k [ ] k { ( m) S m 1( )}

Hence, it follows that countable intersections of the A-subgroups F ix η, i and Hλ ( ) k generate a normal ﬁlter on A.

Definition 17. Let F denote the filter on A defined as follows:

A subgroup B ⊆ A is contained in F if there are η , i m < ω , λ , k m < (( m m) S ) (( m m) S ω with ) B ⊇ F ix η , i ∩ Hλm . m m km m<ω ( ) m<ω

32 Then by Lemma 16, it follows that F is a normal ﬁlter on A; which is countably closed.

Now, we can use F to establish our notion of symmetry.

Deﬁnition 18. A P-name x˙ is symmetric if

π ∈ A πxDπ = xDπ ∈ F. [ ] S Recursively, a name x˙ is hereditarily symmetric, x ∈ HS, if x˙ is symmetric, and y˙ is hereditarily symmetric for all y˙ ∈ domx ˙.

By Lemma 14, this is well-deﬁned, since for π ∼ π′ andx ˙ ∈ Name P , it follows ( ) that πxDπ = xDπ if and only if π′xDπ′ = xDπ′ .

We will use the following properties: Ifx ˙ ∈ HS and π ∈ A, then ﬁrstly, it is not diﬃcult to verify that also xDπ ∈ HS holds, and secondly, πxDπ ∈ HS. For the second claim, one can check that whenever σ ∈ A with σxDσ = xDσ , then

Dπσπ−1 Dπσπ−1 πσπ−1 πxDπ = πxDπ , ( ) and then use the normality of F.

5 The symmetric submodel.

Let G be a V -generic ﬁlter on P. Our symmetric extension is

N ∶= V G = x˙ G x˙ ∈ HS . ( ) { S } Since we do not use the standard method for constructing symmetric extensions, we ﬁrst have to make sure that N ⊧ ZF .

The symmetric forcing relation “ ⊩s” can be deﬁned in the ground model as usually, and the forcing theorem holds.

Wheneverx ˙,y ˙ ∈ HS and p ∈ P, then p ⊩s y˙ ∈ x˙ if and only if p ⊩ y˙ ∈ x˙ (with the ordinary forcing relation ⊩) and p ⊩s x˙ = y˙ if and only if p ⊩ x˙ = y˙. In particular, for anyx ˙ ∈ HS and D ∈ D, we have

xD = yD,p y˙ ∈ domx,p ˙ ∈ D,p ⊩ y˙ ∈ x˙ . {( ) S s }

Since V ⊆ V G ⊆ V G and V G is transitive, it follows that V G satisﬁes the ( ) [ ] ( ) ( ) axioms of Emptyset, F oundation, Extensionality and Infinity. The proofs of the axioms of P airing, Union and Separation are similar to the proofs for the standard construction, with some extra care needed to make sure that all the involved names are indeed symmetric.

We give a proof of P ower Set and Replacement.

33 Lemma 19. V G ⊧ P ower Set. ( ) Proof. Consider X ∈ N, X = X˙ G with X˙ ∈ HS. We have to show that Ń X ∈ N. ( ) Let B˙ ∶= Y,p˙ Y˙ ∈ HS , Y˙ ⊆ dom X˙ × P ,p ⊩ Y˙ ⊆ X˙ . {( ) S s } Then B˙ G = Ń X , since for any Y ∈ N with Y ⊆ X, there exists a name Y˙ ∈ HS, ( ) Y˙ G = Y , such that Y˙ ⊆ dom X˙ × P. It remains to make sure that the name B˙ is symmetric. Consider π ∈ A with D πX π = X. Then D D B π = Y π , p Y˙ ∈ HS , Y˙ ⊆ dom X˙ × P , p ⊩ Y˙ ∈ B˙ , p ∈ D . ( ) S s π It is not difficult to check that D D B π = Y π , p Y˙ ∈ HS , Y˙ ⊆ dom X˙ × P , p ⊩ Y˙ ⊆ X˙ , p ∈ D , ( ) S s π

since for any p ∈ Dπ and Y˙ ∈ HS, Y˙ ⊆ dom X˙ × P, it follows that p ⊩s Y˙ ∈ B˙ if and

only if p ⊩s Y˙ ⊆ X˙ . Hence, D D D D πB π = πY π ,πp Y˙ ∈ HS , Y˙ ⊆ dom X˙ × P ,πp ⊩ πY π ⊆ πX π ,πp ∈ D . ( ) S s π D D It remains to show that B π = πB π ; then D D D D π ∈ A πB π = B π ⊇ π ∈ A πX π = X π ∈ F [ ] S [ ] S as desired.

D D D D For the inclusion B π ⊆ πB π , consider Y π , p ∈ B π as above. It suﬃces to ( ) ˙ ˙ ˙ P Dπ Dπ −1 construct Y0 ∈ HS, Y0 ⊆ dom X × with πY0 = Y . Then setting p0 ∶= π p, it Dπ Dπ Dπ Dπ Dπ Dπ follows that Y , p = πY0 ,πp0 ∈ πB , since p ⊩s Y ⊆ X and πX = Dπ ( ) Dπ( Dπ ) X gives πp0 ⊩s πY0 ⊆ πX . Let D D Y˙ ∶= z,p˙ z˙ ∈ dom X,p˙ ∈ D ,πzDπ ∈ dom Y π , πp ⊩ πzDπ ∈ Y π . 0 ( ) S π s Then D D πY π = πzDπ ,πp z˙ ∈ dom X,p˙ ∈ D ,πzDπ ∈ dom Y π ,p ⊩ z˙ ∈ Y˙ . 0 ( )S π s 0 D Dπ π We ﬁrst show that wheneverz ˙ ∈ dom X˙ , p ∈ Dπ and πz ∈ dom Y as above, D Dπ π then p ⊩s z˙ ∈ Y˙0 if and only if πp ⊩s πz ∈ Y . D “⇒”: If πp ⊩ πzDπ ∈ Y π , it follows that z,p˙ ∈ Y˙ ; hence, p ⊩ z˙ ∈ Y˙ as desired. s ( ) 0 s 0

“⇐”: Now, assume that p ⊩s z˙ ∈ Y˙0. Let H be a V -generic ﬁlter on P with D πp ∈ H. We have to show that πzDπ H ∈ Y π H . Let H′ ∶= π−1H. Then ′ ( ) ′ ( )′ Dπ H H ′ H ˙ H πz = z˙ , and p ∈ H . Hence,z ˙ ∈ Y0 implies that there must ( ) ′ ′ D H H ′ Dπ π be u,˙ r ∈ Y˙0 withu ˙ = z˙ and r ∈ H . Then πr ⊩s πu ∈ Y by ( ) D Dπ H π H construction of Y˙0. Since πr ∈ H, it follows that πu ∈ Y , with ′ ′ ( ) ( ) πuDπ H = u˙ H = z˙H = πzDπ H as desired. ( ) ( )

34 Hence,

D D D πY π = πzDπ ,πp z˙ ∈ dom X,p˙ ∈ D ,πzDπ ∈ dom Y π , πp ⊩ πzDπ ∈ Y π . 0 ( )S π s

Dπ Dπ Dπ Dπ We have to make sure that πY0 = Y . The inclusion πY0 ⊆ Y is clear. Re- D D D π π Dπ π garding Y ⊆ πY0 , consider u ,q ∈ Y withu ˙ ∈ dom Y˙ ⊆ dom X˙ and q ∈ Dπ ( ) D D Dπ π π such that q ⊩s u˙ ∈ Y˙ . From u ∈ dom X = dom πX , it follows that there must D Dπ Dπ −1 Dπ Dπ π bev ˙ ∈ dom X˙ with u = πv . Let r ∶= π q. Then u ,q = πv , πr ∈ πY0 , D ( ) ( ) Dπ π since q ⊩s u˙ ∈ Y˙ implies πr ⊩s πv ∈ Y as desired.

Dπ Dπ Thus, we have constructed Y˙0 ⊆ dom X˙ × P with πY0 = Y . It remains to make sure that Y˙0 ∈ HS. Firstly, Y˙0 ⊆ dom X˙ × P, so dom Y˙0 ⊆ HS. Secondly, for any D D σ ∈ A with σY σ = Y σ , it follows that

D −1 D −1 D − D π σπ D π σπ π−1σπ Y π 1σπ = π−1σπ Y π = π−1σπ π−1Y π , ( ) 0 ( ) 0 ( ) D D and since σY σ = Y σ , one can easily check that

D − D − D π 1σπ D π 1σπ π−1σπ π−1Y π = π−1Y π , ( ) and

D −1 D −1 D π σπ D π σπ D − −1 π π π 1σπ π Y = Y0 = Y0 .

Since the name Y˙ is symmetric, it follows by normality of F that Y˙0 is symmetric, Dπ Dπ as well. Hence, Y˙0 has all the desired properties; and it follows that B ⊆ πB . D D The inclusion πB π ⊆ B π is similar.

Lemma 20. V G ⊧ Replacement. ( ) Proof. Consider a ∈ N such that N ⊧ ∀x ∈ a ∃y ϕ x,y . We have to show that ( ) there is b ∈ N with N ⊧ ∀x ∈ a ∃ y ∈ b ϕ x,y . ( ) Let a = a˙ G witha ˙ ∈ HS. We proceed like in the proof of Replacement in ordinary forcing extensions. Forx ˙ ∈ doma ˙ and p ∈ P, let

α x,p˙ ∶= min α ∃ w˙ ∈ Name P ∩ HS ∶ p ⊩ ϕ x,˙ w˙ ∧ x˙ ∈ a˙ ( ) { S α( ) s ( ( ) )} if such α exists, and α x,p˙ ∶= 0, else. ( ) By Replacement in V , take β ∈ Ord with β ≥ sup α x,p˙ x˙ ∈ doma,p ˙ ∈ P . Let { ( ) S }

b˙ ∶= y,˙ 1 y˙ ∈ Name P ∩ HS , {( ) S β( ) }

35 and b ∶= b˙G. Then for all x ∈ a, it follows that there exists y ∈ b with N ⊧ ϕ x,y . ( ) It remains to show that the name b˙ is symmetric. Let π ∈ A. Then

D b π = yDπ ,q y˙ ∈ Name P ∩ HS, q ∈ D , {( ) S β( ) π} and D πb π = πyDπ ,πq y˙ ∈ Name P ∩ HS, q ∈ D . {( ) S β( ) π} D D We will show that πb π = b π .

Dπ Since it is not possible to apply π to arbitrary P-namesy ˙ withy ˙ ∉ Name P , we ( ) construct an alternative π which is suﬃcient for our purposes here. Recursively, ̃ we deﬁne for P-namesy ˙:

π y˙ ∶= π z˙ ,πq ∃ z,q˙ ∈ y˙ , q ≤ q , q ∈ D . ̃( ) (̃( ) ) S ( ) π Then for ally ˙ ∈ Name P , it follows that π y˙ ∈ Name P , as well. β( ) ̃( ) β( ) Whenever H is a V -generic ﬁlter on P, H′ ∶= π−1H andy ˙ ∈ Name P , then ′ ( ) π y˙ H = y˙H , and one can easily check that (̃( ))

Dπ πyDπ = π y˙ . ̃( )

Moreover, whenever σ ∈ A with σyDσ = yDσ , then setting τ ∶= πσπ−1, it follows recursively that Dτ Dτ τ π y˙ = π y˙ . ̃( ) ̃( ) Hence,

Dτ Dτ τ ∈ A τ π y˙ = π y˙ ⊇ π σ π −1 σ ∈ A, σyDσ = yDσ . [ ] T ̃( ) ̃( ) [ ][ ][ ] T [ ]

In the case thaty ˙ is symmetric, i.e. σ ∈ A σyDσ = yDσ ∈ F, it follows [ ] T Dτ Dτ by normality that also τ ∈ A τ π y˙ = π y˙ ∈ F. Hence, π y˙ ∈ HS [ ] T ̃( ) ̃( ) ̃( ) whenevery ˙ ∈ HS.

D D D Now, πb π = b π follows: For the inclusion “⊆”, consider πyDπ ,πq ∈ πb π with ( ) Dπ y˙ ∈ Name P ∩ HS, q ∈ D . Then also πq ∈ D , and πyDπ = π y˙ , where β( ) π π ̃( ) Dπ D π y˙ ∈ Name P ∩ HS; so πyDπ ,πq = π y˙ ,πq ∈ b π follows. The inclusion ̃( ) β( ) ( ) (̃( ) ) “⊇” is similar.

Hence, b˙ ∈ HS as desired.

Thus, our symmetric extension N is a model of ZF . Since our forcing P is countably closed (Proposition 7), and our normal ﬁlter F generating N is countably closed, it follows that N ⊧ DC (see for example [Kar14, Lemma 1]). Moreover, N ⊧ AX (see [She10, p.3 and p.15]): For any cardinal λ, we have λ ℵ0 N = 4 ([ ] )

36 λ ℵ0 V ; so the set λ ℵ0 can be well-ordered in N, using the according well- ([ ] ) [ ] ordering of λ ℵ0 in V . [ ] Next, we want to show that N preserves all V -cardinals; which will follow from the fact that any set of ordinals X ⊆ α, X ∈ N, can be captured in a “mild” V -generic extension by a forcing notion as in Lemma 9 and Lemma 10. This Approximation Lemma demonstrates how our symmetric extension N can be approximated from within by fairly nice V -generic extensions. Later on, this will be a crucial step in keeping control over the values θN κ . ( η) Lemma 21 (Approximation Lemma). Consider X ∈ N, X ⊆ α with X = X˙ G such D D that πX π = X π holds for π ∈ A with π contained in the intersection [ ]

λm λm F ix ηm, im ∩ F ix η , im ∩ H ∩ H , m km k m<ω ( ) m<ω ( ) m<ω m<ω m where η , i m < ω , η , i m < ω , λ , k m < ω and λ , k m < (( m m) S ) (( m m) S ) (( m m) S ) (( m m) S ω denote sequences with η ∈ Lim, i < α ; η ∈ Succ, i < α for all m < ω; ) m m ηm m m ηm and λ ∈ Lim, k < α ; λ ∈ Succ, k < α for all m < ω. m m λm m m λm Then X ∈ V Gηm × Gηm . M im M i m<ω m<ω m Proof. Let

X′ ∶= β < α ∃ p = p , pσ, aσ , pσ ∶ p ⊩ β ∈ X˙ , ∀m ∶ η , i ∈ suppp , T ( ∗ ( i i )σ,i ( )σ) s ( m m) 0

ηm ηm ηm ηm ηi ηm ∀m ∶ a = g , p m ω ∈ G , p m ω ∈ G . im im im < M im i < M i ( ) m<ω ( m ) m<ω m Then X′ ∈ V Gηm × Gηm , M im M i m<ω m<ω m ηm ′ since the sequence g m<ω is contained in V . It remains to show that X = X . ( im ) The inclusion X ⊆ X′ follows from the forcing theorem. Concerning “⊇”, assume towards a contradiction there was β ∈ X′ ∖ X. Take p as above with η , i ∈ ( m m) suppp0 for all m < ω, and

ηm ηm ηm ηm ηi ηm p ⊩s β ∈ X˙ , ∀m ∶ a = g , p m ω ∈ G , p m ω ∈ G . im im im < M im i < M i ( ) m<ω ( m ) m<ω m

Since β ∉ X, we can take p′ ∈ G, p′ = p′ , p′ σ, a′ σ , p′ σ with p′ ⊩ β ∉ ∗ ( )i ( )i σ,i ( ) σ s X˙ , such that η , i ∈ supp p′ for all m < ω. ( m m) 0 First, we want to extend p and p′ and obtain conditions p ≤ p, p′ ≤ p′, p = p , pσ, aσ , pσ , p′ = p′ , p′ σ, a′ σ , p′ σ such that the following ( ∗ ( i i )σ,i ( )σ) ∗ ( )i ( )i σ,i ( ) σ holds:

• ∀ m < ω pηm = p′ ηm , aηm = a′ ηm im ( )im im ( )im

37 • ∀ m < ω pηm = p′ ηm im ( )im • ′ dom p0 = dom p0

• ′ supp p0 = supp p0

• σ ′ ′ σ ⋃ σ,i supp p a = ⋃ supp a ( )∈ 0 i (σ,i)∈ p0 ( )i • ∀ ν, j ∶ dom p ∩ κ , κ ≠ ∅ → aσ ∩ κ , κ ⊆ dom p , ( ) 0 [ ν,j ν,j+1) ⋃σ,i i [ ν,j ν,j+1) 0 ∀ ν, j ∶ dom p′ ∩ κ , κ ≠ ∅ → a′ σ ∩ κ , κ ⊆ dom p′ ( ) 0 [ ν,j ν,j+1) ⋃σ,i ( )i [ ν,j ν,j+1) 0 • ′ supp p1 = supp p1

• ∀ σ ∈ supp p = supp p′ ∶ dom p σ = dom p′ σ . 1 1 1( ) 1( ) ′ We will now describe how p0 and p0 can be constructed. First, we need a set ′ supp0 ∶= supp p0 = supp p0. Consider

s ∶= sup κ σ ∈ Lim , ∃ i < α ∶ σ, i ∈ supp p ∪ supp p′ . { σ S σ ( ) 0 0} Then by closure of the sequence κ 0 < σ < γ , it follows that s = κ for ( σ S ) γ some γ ≤ γ. If γ = γ, then cf κ = ω and we can take σ , l k < ω with γ (( k k) S ) σ ∈ Lim, l < α for all k < ω such that κ k < ω is coﬁnal in κ , and k k σk ( σk S ) γ σ , l ∉ supp p ∪ supp p′ for all k < ω. Let ( k k) 0 0 supp ∶= supp p ∶= suppp′ ∶= supp p ∪ suppp′ ∪ σ , l k < ω . 0 0 0 0 0 {( k k) S } If γ < γ, we can set σ ∶= γ ∈ Lim for all k < ω and take l k < ω such that k ( k S ) l < α with σ , l = γ, l ∉ supp p ∪ supp p′ for all k < ω. Let k σk ( k k) ( k) 0 0 supp ∶= supp p ∶= supp p′ ∶= supp p ∪ supp p′ ∪ σ , l k < ω 0 0 0 0 0 {( k k) S } as before.

The next step is to deﬁne the linking ordinals. Take a set X ⊆ κγ such that for all intervals κ , κ ⊆ κ , it follows that X ∩ κ , κ = ℵ ; and [ ν,j ν,j+1) γ S [ ν,j ν,j+1)S 0 σ ′ σ X ∩ ⋃(σ,i)∈supp p a ∪ ⋃(σ,i)∈supp p′ a = ∅. Let 0 i 0 ( )i σ ′ σ X ∶= X ∪ ai ∪ a i . σ,i σ,i( )

Our aim is to construct p and p′ such that aσ = a′ σ = X. ⋃σ,i i ⋃σ,i ( )i Consider an interval κ , κ ⊆ κ . For every σ, i ∈ supp p with κ > κ , [ ν,j ν,j+1) γ ( ) 0 σ ν,j we let aσ ∩ κ , κ ∶= aσ ∩ κ , κ . i [ ν,j ν,j+1) i [ ν,j ν,j+1) Deﬁne σ ξk ν, j k < ω ∶= X ∩ κν,j, κν,j+1 ∖ ai . { ( ) S } [ ) σ,i

This set has cardinality ℵ0 by construction of X.

38 Moreover, let

σ ν, j , l ν, j k < ω =∶ σ, i ∈ supp p ∖ supp p κ > κ . k( ) k( ) S {( ) 0 0 S σ ν,j}

This set also has cardinality ℵ0 by construction of supp p0 = supp0.

Now, for any k < ω, let

σk(ν,j) a ∩ κν,j, κν,j+1 ∶= ξk ν, j . lk(ν,j) [ ) { ( )} Together with same construction for p′, we obtain the linking ordinals aσ, a′ σ i ( )i for σ, i ∈ supp = suppp = supp p′ such that the independence property holds, ( ) 0 0 ( )0 and aσ = a′ σ = X. ⋃σ,i i ⋃σ,i( )i Next, we construct dom ∶= dom p = dom p′ ∶= κ , δ as follows: Con- 0 0 ( )0 ⋃ν,j[ ν,j ν,j ) sider an interval κ , κ ⊆ κ . In the case that dom p ∩ κ , κ = [ ν,j ν,j+1) γ 0 [ ν,j ν,j+1) dom p′ ∩ κ , κ = ∅, let δ ∶= κ . Otherwise, take δ ∈ κ , κ ( )0 [ ν,j ν,j+1) ν,j ν,j ν,j [ ν,j ν,j+1) such that dom p ∪ dom p′ ∪ X ∩ κ , κ ⊆ κ , δ . (This is possible ( 0 0 ) [ ν,j ν,j+1) [ ν,j ν,j) since the set X ∩ κ , κ is countable, and any κ is a successor cardinal.) [ ν,j ν,j+1) ν,j+1 Let ′ dom0 ∶= dom p0 ∶= dom p0 ∶= κν,j, δν,j . ν,j[ ) ′ This set is bounded below all regular κν, by construction, since dom p0 and dom p0 are bounded below all regular κν,.

Now, for σ, i ∈ supp p , let pσ ∶ dom pσ → 2 on the corresponding domain with ( ) 0 i i dom pσ = dom p ∩ κ , such that pσ ⊇ pσ for all σ, i ∈ suppp , and in the case i 0 σ i i ( ) 0 ηm ′ ηm that σ, i = ηm, im for some m < ω, we additionally require that p ⊇ p . ( ) ( ) im ( )im This is possible, since p′ ∈ G and pηm ∈ Gηm , so pηm and p′ ηm are compatible. im im im ( )im We deﬁne p on the according domain κ , δ 2 such that p ⊇ p , and the ∗ ⋃ν,j[ ν,j ν,j) ∗ ∗ linking property holds for p ≤ p : Consider an interval κ , κ with δ > κ . 0 0 [ ν,j ν,j+1) ν,j ν,j For ζ ∈ dom p ∖ dom p ∩ κ , κ and ξ ∶= aσ ∩ κ , κ for some ( 0 0) [ ν,j ν,j+1) { } i [ ν,j ν,j+1) σ, i ∈ supp p , it follows by construction that ξ ∈ dom p . Let p ξ,ζ ∶= pσ ζ . ( ) 0 0 ∗( ) i ( ) For all ξ, ζ ∈ dom p ∩ κ , κ , we set p ξ,ζ ∶= p ξ,ζ ; and p ξ,ζ ∈ 0, 1 0 [ ν,j ν,j+1) ∗( ) ∗( ) ∗( ) { } arbitrary for the ξ, ζ ∈ dom p0 remaining.

Concerning p′, we set p′ ηm = pηm for all m < ω. Then p′ ηm ⊇ p′ ηm by ( )im im ( )im ( )im construction. For the σ, i ∈ supp p′ = suppp remaining, we can set p′ σ ( ) ( )0 0 ( )i arbitrarily on the given domain such that p′ σ ⊇ p′ σ. ( )i ( )i Finally, we let p′ ⊇ p′ according to the linking property for p′ ≤ p′ (same ( )∗ ( )∗ 0 0 construction as for p∗).

′ ′ ′ It follows that p0 ≤ p0 and p0 ≤ p0, and p0 and p0 have all the required properties.

′ ′ The construction of p1 ≤ p1 and p1 ≤ p1 is similar.

Our aim is to write down an isomorphism π ∈ A with the following properties:

39 ′ • p ∈ Dπ with πp = p ,

• λm λm π ∈ ⋂m<ω F ix ηm, im ∩ ⋂m<ω F ix ηm, im ∩ ⋂m<ω H ∩ ⋂m<ω H ( ) ( ) km km D D (then πX π = X π follows).

Dπ ′ Dπ From p ⊩s β ∈ X˙ , we will then obtain πp ⊩s β ∈ πX ; hence, p ⊩s β ∈ X . This ′ will be a contradiction towards p ⊩s β ∉ X˙ .

′ ′ We start with π0. Let dom π0 ∶= dom p0 = dom p0, and supp π0 ∶= suppp0 = supp p0.

• Consider an interval κ , κ . We define F ν, j ∶ supp π ν, j → [ ν,j ν,j+1) π0 ( ) 0( ) supp π ν, j as follows: Let F ν, j σ, i ∶= λ, k in the case that a′ σ ∩ 0( ) π0 ( )( ) ( ) ( )i κ , κ = aλ ∩ κ , κ . This is well-defined by the independence [ ν,j ν,j+1) k [ ν,j ν,j+1) property, and since we have arranged aσ = a′ σ. ⋃σ,i i ⋃σ,i( )i • For every interval κ , κ , let G ν, j σ, i = σ, i for all σ, i ∈ [ ν,j ν,j+1) π0 ( )( ) ( ) ( ) supp π ν, j . 0( ) (These maps G ν, j will be the only parameters of π which are not de- π0 ( ) 0 ′ termined by the requirement that π0p0 = p0. However, in order to make sure λm that π ∈ ⋂m<ω F ix ηm, im ∩ ⋂m<ω H , we firstly need Gπ ν, j ηm, im = ( ) km 0 ( )( ) η , i for all m < ω; and secondly, whenever m < ω and i ≤ k , we need ( m m) m that G ν, j λ , i = λ , i for all κ above a certain κ .) π0 ( )( m ) ( m ) ν,j ν, • For ζ ∈ κ , κ ∩ dom π , we define π ζ ∶ 2supp π0(ν,j) → 2supp π0(ν,j) as [ ν,j ν,j+1) 0 0( ) follows: For ǫ σ, i ∈ suppπ ν, j ∈ 2supp π0(ν,j) given, let π ζ ǫ (σ,i) S ( ) 0( ) 0( ) (σ,i) S σ, i ∈ supp π ν, j ∶= ǫ σ, i ∈ supp π ν, j such that ǫ = ǫ ( ) ( ) ̃(σ,i) S ( ) 0( ) ̃(σ,i) (σ,i) whenever pσ ζ = p′ σ ζ , and ǫ ≠ ǫ in the case that pσ ζ ≠ p′ σ ζ . i ( ) ( )i ( ) ̃(σ,i) (σ,i) i ( ) ( )i ( ) • Let now ζ ∈ dom π ∩ κ , κ , and ξσ ν, j σ, i ∈ supp π ν, j ∈ 0 [ ν,j ν,j+1) i ( ) S ( ) 0( ) dom π ν, j supp π0(ν,j). The map π ζ ξσ ν, j σ, i ∈ suppπ ν, j ∶ 0( ) ∗( ) i ( ) S ( ) 0( ) 2supp π0(ν,j) → 2supp π0(ν,j) is defined as follows: A sequence ǫ σ, i ∈ (σ,i) S ( ) supp π ν, j is mapped to ǫ σ, i ∈ supp π ν, j with ǫ = ǫ if 0( ) ̃(σ,i) S ( ) 0( ) ̃(σ,i) (σ,i) p ξσ ν, j ,ζ = p′ ξσ ν, j ,ζ , and ǫ ≠ ǫ in the case that p ξσ ν, j ,ζ ∗( i ( ) ) ∗( i ( ) ) ̃(σ,i) (σ,i) ∗( i ( ) ) ≠ p′ ξσ ν, j ,ζ . ∗( i ( ) ) • For ξ,ζ ∈ κ , κ 2, the map π ξ,ζ ∶ 2 → 2 is defined as follows: We ( ) [ ν,j ν,j+1) ∗( ) let π ξ,ζ = id in the case that ξ,ζ ∉ dom π ν, j 2. If ξ,ζ ∈ dom π ν, j , ∗( ) ( ) ( 0( )) 0( ) let π ξ,ζ = id if p ξ,ζ = p′ ξ,ζ , and π ξ,ζ ≠ id in the case that ∗( ) ∗( ) ∗( ) ∗( ) p ξ,ζ ≠ p′ ξ,ζ . ∗( ) ∗( ) ′ This defines π0. Directly by construction, it follows that π0p0 = p0: Let π0p0 =∶ πp , πp σ, πa σ , πpσ . Then for any σ, i ∈ supp π p = supp p (( )∗ (( )i ( )i )σ,i ( )σ) ( ) ( 0 0) 0 and κ < κ , we have πa σ ∩ κ , κ = aλ ∩ κ , κ , where λ, k = ν,j σ ( )i [ ν,j ν,j+1) k [ ν,j ν,j+1) ( ) F ν, j σ, i ; hence, aλ ∩ κ , κ = a′ σ ∩ κ , κ as desired. π0 ( )( ) k [ ν,j ν,j+1) ( )i [ ν,j ν,j+1)

40 For any ζ ∈ dom p , it follows by deﬁnition of π ζ that πp σ ζ σ, i ∈ 0 0( ) ( )i ( ) S ( ) suppπ ν, j = p′ σ ζ σ, i ∈ supp π ν, j , and similarly, πp ξ,ζ = 0( ) ( )i ( ) S ( ) 0( ) ( )∗( ) p′ ξ,ζ for all ξ,ζ ∈ dom πp = dom p . ∗( ) ( ) ( )∗ ∗ ′ Hence, π0p0 = p0.

λm It remains to verify that π ∈ ⋂m F ix ηm, im ∩ ⋂m H . Consider a condition r ∈ ( ) km D and let r′ ∶= π r. Take an interval κ , κ ⊆ κ . Then for any m < ω with π0 0 [ ν,j ν,j+1) γ ′ ηm ηm ηm, im ∈ supp π0 ν, j and ζ ∈ dom π0 ν, j , it follows that r ζ = r ζ ( ) ( ) ( ) ( )im ( ) im ( ) ηm ′ ηm by construction of the map π0 ζ , since we have arranged p ζ = p ζ . ( ) im ( ) ( )im ( ) In the case that ζ ∈ κ , κ with ζ ∈ dom r ∖ dom π , it follows for m < ω [ ν,j ν,j+1) 0 0 ′ ηm λ that r ζ = r ζ , where λ, k = Gπ ν, j ηm, im = ηm, im as desired. ( )im ( ) k ( ) ( ) 0 ( )( ) ( ) ′ ηm ηm Hence, r = r for all m < ω. Since r ∈ Dπ was arbitrary, it follows that ( )im im 0 π ∈ F ix η , i . 0 ⋂m<ω ( m m)

λm Similarly, π0 ∈ ⋂m H follows from the fact that Gπ ν, j = id for all intervals km 0 ( ) κ , κ ⊆ κ . [ ν,j ν,j+1) γ

Now, we turn to the map π1.

Let supp π ∶= supp p = supp p′ , and dom π σ ∶= dom p σ = dom p′ σ for 1 1 1 1( ) 1( ) 1( ) σ ∈ suppπ . We set supp π σ ∶= ∅ for all σ ∈ supp π . Then we only have to de- 1 1( ) 1 ﬁne maps π σ i, ζ ∶ 2 → 2 for σ ∈ supp π, i, ζ ∈ dom π σ : Let π σ i, ζ = id 1( )( ) ( ) 1( ) 1( )( ) if p σ i, ζ = p′ σ i, ζ , and π σ ≠ id in the case that p σ i, ζ ≠ p′ σ i, ζ . ( )( ) ( )( ) 1( ) ( )( ) ( )( ) Clearly, π p = p′ . Moreover, π ∈ F ix η , i : Let m < ω and r ∈ D 1 1 1 ⋂m ( m m) π1 with η ∈ supp r and i ∈ dom r η . In the case that η ∈ supp π , it follows m m x ( m) m 1 for any ζ ∈ dom r η that πr η i ,ζ = π η i ,ζ r η i ,ζ = y ( m) ( )( m)( m ) 1( m)( m ) ( m)( m ) r η i ,ζ by construction of π , since we have arranged that p′ η i ,ζ = ( m)( m ) 1 ( m)( m ) p η i ,ζ whenever i ,ζ ∈ dom p η = dom p′ η = dom π η . If η ∉ ( m)( m ) ( m ) ( m) ( m) 1( m) m suppπ , then πr η = r η by construction. 1 ( )( m) ( m) λm Finally, π ∈ ⋂m H follows from the fact that supp π1 λ = 0 for all λ ∈ supp π1. km ( ) Hence, the map π has all the desired properties.

This ﬁnishes the proof of X = X′, and

X = X′ ∈ V Gηm × Gηm M im M i m<ω m<ω m follows.

It is not diﬃcult to see that with the exception of the maps G ν, j , all the π0 ( ) parameters describing π are given by the requirement that πp = p′. We call an isomorphism π ∈ A of this form a standard isomorphism for πp = p′.

With the same proofs as for Lemma 2 and 3, one can show:

41 Lemma 22. Let σ , i m < ω , σ , i m < ω with σ ∈ Lim, i < α , (( m m) S ) (( m m) S ) m m σm σm σm and σm ∈ Succ, im < ασm for all m < ω. Then ∏m<ω P × ∏m<ω P preserves cardinals, coﬁnalities and the GCH.

Hence, the Approximation Lemma 21 implies:

Corollary 23. Cardinals and coﬁnalities are V -N-absolute.

We will now take a closer look at the intermediate generic extensions introduced in the Approximation Lemma 21. Firstly, we replace the generic ﬁlters Gσm by im σm G∗ g , and secondly, we factor at κη (or κη+1). ( im ) Deﬁnition 24. For 0 < η < γ, we say that a , σ , i is an η-good ( m)m<ω ( m m)m<ω pair if the following hold:

• a m < ω is a sequence of pairwise disjoint κ -subsets, such that for all ( m S ) η m < ω and κ < κ , it follows that a ∩ κ , κ = 1, ν, η S m [ ν, ν,+1)S • for all m < ω, we have σm ∈ Succ with σm ≤ η, im < ασm ,

• if m ≠ m′, then σ , i ≠ σ ′ , i ′ . ( m m) ( m m ) As in Lemma 9 and Lemma 10, it follows that for any η-good pair a , σ , i , ( m)m<ω ( m m)m<ω G a × Gσm M ∗ m M i m<ω ( ) m<ω m η is a V -generic ﬁlter on P ω × P σm . ∏m<ω( ) ∏m<ω + Proposition 25. Let 0 < η < γ and X ∈ N with X ⊆ κη. If κη+1 > κη (or κη = κγ with γ = γ +1), it follows that there is an η-good pair a , σ , i with ( m)m<ω ( m m)m<ω X ∈ V G a × Gσm . M ∗ m M i m<ω ( ) m<ω m Proof. By the Approximation Lemma 21, there are sequences σ , i m < ω , (( m m) S ) σ , i m < ω of pairwise distinct pairs with σ ∈ Lim, i < α ; σ ∈ Succ, (( m m) S ) m m σm m im < ασm for all m < ω, such that

X ∈ V Gσm × Gσm . M im M i m<ω m<ω m The sequence of linking ordinals gσm m < ω is contained in V , and by the ( im S ) σm σm linking property, it follows that V ∏ G = V ∏ G∗ g . [ m<ω im ] [ m<ω ( im )] Hence, X ∈ V G gσm × Gσm . M ∗ im M i m<ω ( ) m<ω m

σm σm The forcing ∏m<ω P × ∏m<ω P can be factored as

P σm ↾κ × P σm × P σm ↾ κ , κ × P σm , M η M M [ η σm ) M m<ω σm≤η m<ω σm>η

42 + where the “lower part” has cardinality ≤ κη by the GCH in V , and the “upper + part” is ≤ κη -closed: If κη+1 is a limit cardinal, this follows from the fact that ++ ++ κη,j+1 ≥ κη,j for all j < cf κη+1 by construction (in particular, κη,1 ≥ κη ); and if + κη+1 is a successor cardinal, we use our assumption that κη+1 > κη . Hence,

σm σm X ∈ V M G∗ gi ∩ κη × M G . ( m ) im m<ω σm≤η Setting a ∶= gσm ∩ κ for m < ω, it follows by the independence property that m im η a , σ , i is an η-good pair with ( m)m<ω ( m m)m<ω, σm≤η

σm X ∈ V M G∗ am × M G . [ ( ) im m<ω σm≤η

+ In the case that κη+1 = κη , we use our notion of an η-almost good pair, which is deﬁned like an η-good pair, with the exception that for an η-almost good pair a , σ , i , we have a ⊆ κ for all m < ω. ( m)m<ω ( m m)m<ω m η+1 Deﬁnition 26. For 0 < η < γ with κ = κ+, we say that a , σ , i η+1 η ( m)m<ω ( m m)m<ω is an η-almost good pair if the following hold:

• a m < ω is a sequence of pairwise disjoint κ -subsets, such that for all ( m S ) η+1 m < ω and κ < κ , it follows that a ∩ κ , κ = 1, ν, η+1 S m [ ν, ν,+1)S • for all m, we have σm ∈ Succ with σm ≤ η, and im < ασm ,

• if m ≠ m′, then σ , i ≠ σ ′ , i ′ . ( m m) ( m m ) The counterpart of Proposition 25 states:

+ Proposition 27. Let 0 < η < γ and X ∈ N with X ⊆ κη. In the case that κη+1 = κη , there is an η-almost good pair a , σ , i with ( m)m<ω ( m m)m<ω X ∈ V G a × Gσm × Gη+1 . M ∗ m M i m<ω ( ) m<ω m Proof. We follow the proof of Proposition 25 with a slightly diﬀerent factorization: Let X ∈ V G gσm × Gσm M ∗ im M i m<ω ( ) m<ω m as before with σm ∈ Lim, im < ασm ; σm ∈ Succ, im < ασm for all m < ω. The forcing σm σm ∏m<ω P × ∏m<ω P can be factored as im im P σm ↾κ × P σm × P σm ↾ κ , κ × P σm , M η+1 M M [ η+1 σm ) M m<ω σm≤η+1 m<ω σm>η+1 + where the “lower part” has cardinality ≤ κη+1 by the GCH in V (since κη+1 = κη ), and the “upper part” is ≤ κη+1-closed. Hence,

σm σm σm σm η+1 X ∈ V M G∗ gi ∩ κη+1 × M G ⊆ V M G∗ gi ∩ κη+1 × M G × G . ( m ) im ( m ) im m<ω σm≤η+1 m<ω σm≤η

43 σm With am ∶= g ∩ κη+1 for m < ω, it follows that am m<ω, σm, im m<ω, σ ≤η is im ( ) ( ) m an η-almost good pair with

σm η+1 X ∈ V M G∗ am × M G × G [ ( ) im m<ω σm≤η as desired.

N 6 ∀η θ (κη)= αη.

It remains to make sure that in our ZF -model N, the values θN κ are as desired. ( η) Firstly, in Chapter 6.1, 6.2 and 6.3, we will show that θN κ = α holds for all ( η) η 0 < η < γ. After that, in Chapter 6.4 and 6.5, we will see that for any cardinal λ ∈ κ , κ in a “gap”, or λ ≥ κ = sup κ 0 < η < γ , the value θN λ is the ( η η+1) γ { η S } ( ) smallest possible. By our remarks from Chapter 2, this justiﬁes our assumption from the beginning that the sequence α 0 < η < γ is strictly increasing. ( η S )

N 6.1 ∀η θ (κη) ≥ αη. η Using the subgroups Hk , it is not diﬃcult to see that for all k < αη, there exists in N a surjection s ∶´ κ → k. ( η) Proposition 28. Let 0 < η < γ. Then θN κ ≥ α . ( η) η Proof. Let k < α . We construct in N a surjection s ∶ ´ κ → k. As already η ( η) η outlined in Chapter 4.3, we deﬁne around each Gi with i < k a “cloud” as follows:

(k) ˙ (k) G Gη ∶= Gη , Èi Èi where

˙ (k) Dπ Gη ∶= πGη , 1 π ∈ Hη ; Èi ( i ) S [ ] k and we take the following canonical name for the i-th generic κη-subset:

η η G˙ ∶= a, p p ∈ P , ∃ ζ < κ ∃ ǫ ∈ 0, 1 ∶ a = ORP ζ,ˇ ǫˇ ∧ p ζ = ǫ . i ( ) T η { } ( ) i ( ) (k) Roughly speaking, Gη is the orbit of Gη under the A-subgroup Hη; hence, its Èi i k ˙ (k) canonical name Gη is ﬁxed by all automorphisms in Hη. Èi k More precisely:

Let σ ∈ A with σ ∈ Hη. Then [ ] k

Dσ Dσ ˙ Dπ Gη (k) = πGη ,p π ∈ Hη , p ∈ D . (Èi ) i S [ ] k σ

44 Moreover, for all π,

Dσ ηDπ D ηDπ πG = a σ , p p ∈ D , p ⊩ a ∈ πG , ∃ ζ < κ ∃ ǫ ∈ 0, 1 ∶ a = ORP ζ,ˇ ǫˇ , i S σ s i η { } ( ) D D D σ D σ D since for any a = ORP ζ,ˇ ǫˇ as above, it follows that πa π = a π = a σ . ( ) ηDπ Now, it is not diﬃcult to see that p ∈ Dσ with p ⊩s a ∈ πGi if and only if p ∈ Dσ and for all q ≤ p with q ∈ D ∩ D and ζ ∈ dom q , it follows that π−1q η ζ = ǫ. π σ 0 ( )i ( ) Also, σaDσ = aDσ holds for all σ.

Hence,

Dσ ηDπ D σ πG = σa σ , σp p ∈ D , ∃ ζ < κ ∃ ǫ ∈ 0, 1 a = ORP ζ,ˇ ǫˇ , i S σ η { } ( )

∀ q ∈ D ∩ D q ≤ p ∧ ζ ∈ dom q ⇒ π−1q η ζ = ǫ π σ ( 0) ( )i ( )

D = a σ , p p ∈ D , ∃ ζ < κ ∃ ǫ ∈ 0, 1 a = ORP ζ,ˇ ǫˇ , S σ η { } ( ) ∀ q ∈ D ∩ D q ≤ p ∧ ζ ∈ dom q ⇒ π−1σ−1q η ζ = ǫ . π σ ( 0) ( )i ( ) Setting τ ∶= σπ, it follows that

Dσ Dσ ηDπ ηDτ σπGi = τGi .

D ˙ σ Now, any element of σ Gη (k) is of the form (Èi )

Dσ Dπ σπGη ,σp i with π ∈ Hη and p ∈ D . Since [ ] k σ

Dσ Dσ Dπ Dτ σπGη ,σp = τGη , p , i i where τ ∶= σπ and p ∶= σp satisfy τ ∈ Hη and p ∈ D , it follows that [ ] k σ

Dσ Dσ Dπ ˙ σπGη ,σp ∈ Gη (k) . i (Èi )

Hence, D D ˙ σ ˙ σ σ Gη (k) ⊆ Gη (k) . (Èi ) (Èi ) The inclusion “⊇” is similar.

Thus, ˙η (k) ˙η (k) G i < k ∶= ORP ˇi, G , 1 i < k , Èi T Èi U

45 (k) is a name for the sequence Gη i < k that is stabilized by all σ with σ ∈ Hη. Èi S [ ] k (k) Hence, Gη i < k ∈ N. Èi S

Now, we can deﬁne in N a surjection s ∶´ κη → k as follows: For X ∈ N, X ⊆ κη, (k) ( ) let s X ∶= i in the case that X ∈ Gη if such i exists, and s x ∶= 0, else. ( ) Èi ( ) The surjectivity of s is clear, since Gη ∈ N for all i < k with s Gη = i. It remains i i to show that s is well-deﬁned; i.e. for any i, i′ < k with i ≠ i′, it follows that η (k) η (k) G ∩ G ′ = ∅. Èi Èi First, let η ∈ Lim, and take i, i′ < k with i ≠ i′.

η The point is that the automorphisms in Hk do not permute the vertical lines η η η Pi ↾ κν,, κη and Pi′ ↾ κν,, κη above some κν, < κη. Thus, the orbits of Gi [η ) η [ ) and Gi′ under Hk must be disjoint: η (k) η (k) Assume towards a contradiction there was X ∈ G ∩ G ′ . Then we have Èi Èi

ηDπ G η Dτ G πG = τG ′ i i η η −1 η −1 η for some π,τ with π ∈ H and τ ∈ H . Hence, π G = τ G ′ . Take [ ] k [ ] k ( )i ( )i κ < κ such that for all κ ∈ κ , κ and l < k, it follows that G ν, j η, l = ν, η ν,j [ ν, η) π0 ( )( ) η, l whenever η, l ∈ supp π ν, j , and G ν, j η, l = η, l whenever η, l ∈ ( ) ( ) 0( ) τ0 ( )( ) ( ) ( ) supp τ ν, j . 0( ) By genericity, take q ∈ G with q ∈ D ∩ D such that there is ζ ∈ dom q ∖ dom π ∩ π τ ( 0 η η dom τ , ζ ∈ κ , κ with q ζ ≠ q ′ ζ . 0) [ ν, η) i ( ) i ( ) η η −1 ′ −1 W.l.o.g., let q ζ = 1, q ′ ζ = 0. With r ∶= π q, r ∶= τ q, it follows by construc- i ( ) i ( ) η η ′ η η tion of the isomorphism that r ζ = q η = 1 and r ′ ζ = q ′ ζ = 0, which i ( ) i ( ) ( )i ( ) i ( ) −1 η −1 η would contradict π G = τ G ′ . ( )i ( )i Hence, s ∶´ κ → k is a well-deﬁned surjection in N. ( η) The case η ∈ Succ is similar.

+ → N 6.2 ∀η κη+1 > κη Ð θ (κη) ≤ αη . Let 0 < η < γ. Throughout this Chapter 6.2, we assume that

+ κη+1 > κη.

Then Proposition 25 can be applied.

+ In Chapter 6.3, we discuss the case that κη+1 = κη , where the proof can be structured the very same way; except that the intermediate generic extensions where the κη-subsets in N are located are given by Proposition 27. Thus, we will have to take care of an extra factor Gη+1 in our products describing these intermediate generic extensions, which will lead to a couple of modiﬁcations. In Chapter 6.3,

46 we take a brief look at each step in the proof presented here, and go through the major changes.

Assume towards a contradiction that there was a surjective function f ∶´ κη → αη D D ( ) in N. Let f = f˙G with f˙ ∈ HS, such that πf π = f π holds for all π ∈ A with π [ ] contained in the intersection

F ix η , i ∩ Hλm A . m m km f˙ m<ω ( ) m<ω ( )

By Proposition 25, it follows that any X ∈ dom f is of the form

σm ∏m<ω G∗(am) × ∏m<ω G X = X˙ im , where a , σ , i is an η-good pair. ( m)m<ω ( m m)m<ω

Our proof will be structured as follows: We pick some β < αη large enough for the intersection A (we give a deﬁnition of this term on the next page) and consider ( f˙) a map f β, which will be obtained from f by restricting its domain to those X that are contained in a generic extension

V G a × Gσm M ∗ m M i m<ω ( ) m<ω m for an η-good pair a , σ , i such that i < β for all m < ω. ( m)m<ω ( m m)m<ω m β We wonder if this restricted function f could still be surjective onto αη.

The main steps of our proof (similar as in [FK16, Chapter 5]) can be outlined as follows:

β β First, we assume that also f ∶ dom f → αη was surjective onto αη.

A) We define a forcing notion Pβ ↾ η + 1 , which will be obtained from P by ( ) essentially “cutting off ” at height η + 1 and width β. We show that there is a projection of forcing posets ρβ ∶ P → Pβ ↾ η + 1 . Then the V -generic filter ( ) G on P induces a V -generic filter Gβ ↾ η + 1 on Pβ ↾ η + 1 . ( ) ( ) B) We show that f β is contained in an intermediate generic extension similar to V Gβ ↾ η + 1 . [ ( )] C) We prove that the forcing Pβ ↾ η + 1 preserves cardinals ≥ α . ( ) η D) We construct in V Gβ ↾ η + 1 a set ´ κ ⊇ dom f β with an injection [ ( )] ̃( η) ι ∶ ´ κ ↪ β. ̃( η)

Then D) together with B) and C) gives the desired contradiction.

β β Hence, f ∶ dom f → αη must not be surjective.

47 β E) We consider α < αη with α ∈ rgf ∖ rgf , and use an isomorphism argument to obtain a contradiction, again.

We see that either case, whether f β was surjective or not, leads into a contradiction. Thus, our initial assumption must be wrong, and we can ﬁnally conclude:

There is no surjective function f ∶´ κ → α . ( η) η Before we start with Chapter 6.2 A), we first define our term large enough for the intersection A : ( f˙) Definition 29. A limit ordinal β < α is large enough for the intersection A ̃ η ( f˙) if the following hold:

• + β̃ > κη

• β > sup i η ≤ η ̃ { m S m } • β > sup k λ ≤ η ̃ { m S m }

++ (We use that αη ≥ κη , and cf αη > ω.)

Fix a limit ordinal β < α large enough for the intersection A , and let β ∶= β +κ+ ̃ η ( f˙) ̃ η (addition of ordinals).

The restriction f β is deﬁned as follows:

Deﬁnition 30.

f β ∶= X, α ∈ f ∃ a , σ , i η-good pair ∶ ( ) U ( m)m<ω ( m m)m<ω

σ ∗ m η ω σm ∏m<ω G (am) × ∏m<ω G ∀m im < β ∧ ∃ X˙ ∈ Name P × M P X = X˙ im . ( ) ( ) m<ω

β β First, we assume towards a contradiction that f ∶ dom f → αη is surjective.

A) Constructing Pβ ↾ (η + 1).

Our aim is to construct a forcing notion Pβ ↾ η + 1 that is obtained from P by ( ) essentially “cutting oﬀ” at height η and width β; i.e. only the cardinals κσ for

σ ≤ η should be considered, and for any such κσ, we add at most β-many new σ κσ-subsets Gi . Regarding our V -generic ﬁlter G on P, we need that the restriction Gβ ↾ η + 1 ∶= ( ) G↾ Pβ ↾ η + 1 is a V -generic ﬁlter on Pβ ↾ η + 1 , which will be guaranteed by ( ) ( ) making sure that the canonical map ρβ ∶ P → Pβ ↾ η + 1 , p ↦ pβ ↾ η + 1 is a ( ) ( ) projection of forcing posets.

A ﬁrst attempt to deﬁne Pβ ↾ η + 1 could be the following: ( )

48 For p ∈ P, let

pβ ↾ η + 1 = p ↾κ2, pσ, aσ , pσ ↾ min α , β × dom pσ ( ) ∗ η ( i i )σ≤η,i

Pβ ↾ η + 1 ∶= pβ ↾ η + 1 p ∈ P . ( ) { ( ) S }

But then, Gβ ↾ η + 1 ∶= pβ ↾ η + 1 p ∈ G would not be a V -generic ﬁlter on ( ) { ( ) S } Pβ ↾ η + 1 : Consider a linking ordinal ξ ∈ gσ for some σ, i , such that η < σ < γ, ( ) i ( ) i < α holds; or σ ≤ η, β ≤ i < α . The set D ∶= p ∈ Pβ ↾ η + 1 ξ ∈ aσ σ σ { ( ) S ⋃σ≤η,i<β i } is dense in Pβ ↾ η + 1 ; but D ∩ Gβ ↾ η + 1 = ∅ by the independence property. ( ) ( ) Hence, Gβ ↾ η + 1 can not be a V -generic ﬁlter on Pβ ↾ η + 1 . ( ) ( ) This shows that the conditions in Pβ ↾ η + 1 should contain some information ( ) σ about which linking ordinals are “forbidden” for ⋃σ≤η,i<β ai , being already occu- pied by some index σ, i with σ > η or i ≥ β. ( ) Thus, for p ∈ P, we add to pβ ↾ η + 1 a new coordinate X , which is essentially ( ) p σ the union of all ai ∩ κη for σ > η or i ≥ β. Then Xp is a subset of κη that hits any interval κ , κ in at most countably many points. [ ν,j ν,j+1) Let η ∶= sup σ < η σ ∈ Lim . By closure of the sequence κ 0 < σ < γ , it follows ̃ { S } ( σ S ) that η ∈ Lim with η = max σ ≤ η η ∈ Lim , and κ = sup κ σ ∈ Lim, σ < η . ̃ ̃ { S } η̃ { σ S ̃} W.l.o.g. we restrict to the case that

β < αη̃ or Lim ∩ (η,γ)≠∅ ; which is the same as requiring that there exist coordinates σ, i with σ ∈ Lim, ( ) and σ > η or i ≥ β. (Otherwise, the forcing Pβ ↾ η + 1 already contains all ( ) coordinates σ, i with σ ∈ Lim, and there are no “forbidden” linking ordinals. ( ) In that case, we can indeed set Pβ ↾ η + 1 ∶= p ↾ κ2, pσ, aσ , pσ ↾ ( ) ∗ η ( i i )σ≤η,i<β ( β × dom pσ p ∈ P , and obtain that Gβ ↾ η + 1 is a V -generic ﬁlter on ( y ))σ≤η S ( ) Pβ ↾ η + 1 . ) ( ) For a condition p ∈ P, let

X ∶= aσ ∩ κ σ ∈ Lim with σ > η or i ≥ β , p i η̃ S ( ) and

pβ ↾ η + 1 ∶= p ↾ κ2, pσ, aσ , pσ ↾ β × dom pσ , X . ( ) ∗ η ( i i )σ≤̃η,i<β ( ( y ))σ≤η p

For reasons of homogeneity, we include into Pβ ↾ η + 1 only those conditions ( ) pβ ↾ η + 1 for which the set X hits every interval κ , κ ⊆ κ in countably ( ) p [ ν,j ν,j+1) η̃ many points, which is the same as requiring σ, i ∈ suppp σ > η or i ≥ β = ℵ . S{( ) 0 S ̃ }S 0

49 Deﬁnition 31. Pβ ↾ η + 1 ∶= ( )

pβ ↾ η + 1 p ∈ P , σ, i ∈ supp p σ > η or i ≥ β = ℵ ( ) T S{( ) 0 S }S 0

∪ 1β , { η+1} 1β with η+1 as the maximal element. For conditions pβ ↾ η +1 , qβ ↾ η +1 in Pβ ↾ η +1 ∖ 1β , let qβ ↾ η +1 ≤β ( ) ( ) ( ) { η+1} ( ) η+1 pβ ↾ η + 1 if X ⊇ X , and q ↾ κ2, qσ, bσ , qσ ↾ β × dom qσ ≤ ( ) q p ∗ η ( i i )σ≤η,i<β ( ( y )σ≤η p ↾ κ2, pσ, aσ , pσ ↾ β × dom pσ regarded as conditions in P. ∗ η ( i i )σ≤η,i<β ( ( y )σ≤η In other words: Pβ ↾ η + 1 is the collection of all p , pσ, aσ , pσ , X ( ) ( ∗ ( i i )σ≤η,i<β ( )σ≤η p) such that

• p ∶= p , pσ, aσ , pσ is a condition in P with dom p ⊆ κ , supp p ⊆ ( ∗ ( i i )σ≤η,i<β ( )σ≤η) 0 η 0 σ, i σ ≤ η, i < β , and supp p ⊆ η + 1 with ∀ σ ∈ supp p ∶ dom pσ ⊆ β, {( ) S } 1 1 x • X ⊆ κ with ∀ κ , κ ⊆ κ X ∩ κ , κ = ℵ , and X ∩ p η̃ [ ν,j ν,j+1) η̃ S p [ ν,j ν,j+1)S 0 p σ ⋃σ≤η,i<β ai = ∅.

For p, q ∈ P with q ≤ p and σ, i ∈ supp p σ > η or i ≥ β = ℵ , it follows that S{( ) 0 S }S 0 qβ ↾ η + 1 ≤ pβ ↾ η + 1 . ( ) ( ) Deﬁnition 32.

Gβ ↾ η+1 ∶= p ∈ Pβ ↾ η+1 ∃ p ∈ G ∶ σ, i ∈ supp p σ > η or i ≥ β = ℵ , ( ) ( ) T S{( ) 0 S }S 0

pβ ↾ η + 1 ≤β p . ( ) η+1

We will now show that Gβ ↾ η + 1 is a V -generic ﬁlter on Pβ ↾ η + 1 . ( ) ( ) Let P ⊆ P denote the collection of all p ∈ P with the property that σ, i ∈ S {( ) supp p σ > η ∨ i ≥ β = ℵ , together with the maximal element 1. Then P is a 0 S } S 0 dense subforcing of P.

Proposition 33. The map ρβ ∶ P → Pβ ↾ η + 1 with p ↦ pβ ↾ η + 1 in the case ( ) ( ) that σ, i ∈ supp p σ > η ∨ i ≥ β = ℵ , and 1 ↦ 1β , is a projection of forcing S {( ) 0 S } S 0 η+1 posets:

• ρβ 1 = 1β , ( ) η+1 • if p, q ∈ P with q ≤ p, it follows that ρβ q ≤β ρβ p , ( ) η+1 ( ) • for any p ∈ P and q ∈ Pβ ↾ η + 1 with q ≤β ρβ p , there exists q ∈ P such ( ) η+1 ( ) that q ≤ p and ρβ q ≤ q. ( ) Hence, Gβ ↾ η + 1 is a V -generic ﬁlter on Pβ ↾ η + 1 . ( ) ( )

50 Proof. Clearly, the map ρβ as deﬁned above is order-preserving with ρβ 1 = 1β . ( ) η+1 Consider p = p , pσ, aσ , pσ ∈ P and q = q ↾ κ2, qσ, bσ , qσ , X ∈ ( ∗ ( i i )σ,i ( )σ) ( ∗ η ( i i )σ≤η,i<β ( )σ≤η q) Pβ ↾ η + 1 with q ≤β ρβ p = pβ ↾ η + 1 . Then ( ) η+1 ( ) ( ) q ↾ κ2, qσ, aσ ≤ p ↾ κ2, pσ, aσ in P , ( ∗ η ( i i )σ≤η,i<β) 0 ( ∗ η ( i i )σ≤η,i<β) 0 qσ ≤ pσ ↾ β × dom pσ in P , and ( )σ≤η 1 ( ( y ))σ≤η 1 X ⊇ aσ ∩ κ σ > η ∨ i ≥ β . q { i η̃ S } σ We have to construct q ∈ P, q = q , qσ, b , qσ , with q ≤ p and ρβ q = qβ ↾ ( ∗ ( i i )σ,i ( )σ) ( ) η + 1 ≤β q. ( ) η+1

We start with q0: • In order to achieve Xq ⊇ Xq, we will enlarge supp p0 ∪ supp q0 by countably many η,mˆ k k < ω , whereη ˆ > η or mk ≥ β for all k < ω, and arrange that (( ) S ) ηˆ ∈ ∖ any ξ Xq Xp occurs as a linking ordinal in some b mk . More precisely: Let supp q0 ∶= supp p0 ∪ supp q0 ∪ supp∗, where supp∗ ∶= η,mˆ k < ω such that η,mˆ ∉ suppp ∪ supp q for all k < ω, and {( k) S } ( k) 0 0 since we are working in the case that β < α or η, γ ∩ Lim ≠ ∅, we can η̃ ( ) take eitherη ˆ ∶= η and m ∈ β, α for all k < ω; orη ˆ ∈ η, γ ∩ Lim. Then ̃ k ( η̃) ( ) for all η,mˆ , it follows thatη ˆ > η or m ≥ β. ( k) k σ • Next, we deﬁne the linking ordinals b for σ, i ∈ supp q such that X ⊇ X . i ( ) 0 q q σ For σ, i ∈ supp q , we let b ∶= bσ ⊇ aσ; and in the case that σ, i ∈ supp p ∖ ( ) 0 i i i ( ) 0 σ σ ηˆ supp q0, we set b ∶= a . Finally, we deﬁne b k < ω with the following i i ( mk S ) properties:

ηˆ – as usual, every b is a subset of κˆ that hits every interval κν,j, κν,j 1 ⊆ mk η [ + ) κηˆ in exactly one point , ηˆ – ⋃ b ∩ κη̃ k < ω ⊇ Xq ∖ Xp , mk S ηˆ σ – b ∩ b = ∅ for all k < ω and σ, i ∈ supp q0 , mk i ( ) ηˆ σ – b ∩ a = ∅ for all k < ω and σ, i ∈ supp p ∖ supp q0 mk i ( ) 0 (since q ≤β pβ ↾ η + 1 , it follows that in this case, σ > η or i ≥ β) , η+1 ( ) ηˆ ηˆ – b ∩ b = ∅ whenever k ≠ k′. mk mk′

This is possible, since X ∩ bσ = ∅ for any σ, i ∈ suppq by construction q i ( ) of Pβ ↾ η + 1 ; and whenever σ, i ∈ supp p ∖ supp q , then σ > η or i ≥ β ( ) ( ) 0 0 implies aσ ⊆ X ; thus X ∖ X ∩ aσ = ∅. i p ( q p) i • We now deﬁne dom q . For any interval κ , κ ⊆ κ , take δ ∈ 0 [ ν,j ν,j+1) η ν,j κ , κ as follows: In the case that dom q ∩ κ , κ = ∅, let [ ν,j ν,j+1) 0 [ ν,j ν,j+1) δν,j ∶= κν,j. If dom q0 ∩ κν,j, κν,j+1 ≠ ∅, we take δν,j ∈ κν,j, κν,j+1 such that σ [ ) ( ) b σ, i ∈ supp q ∩ κ , κ ⊆ κ , δ and dom q ∩ κ , κ ⊆ ⋃{ i S ( ) 0} [ ν,j ν,j+1) [ ν,j ν,j ) 0 [ ν,j ν,j+1)

51 κ , δ . Since dom q is bounded below all regular cardinals κ , this is [ ν,j ν,j ) 0 ν, also true for κ , δ κ < κ . Let ⋃{ [ ν,j ν,j) S ν,j η }

dom q ∩ κ ∶= κ , δ κ < κ , 0 η { [ ν,j ν,j) S ν,j η }

and dom q ∩ κ , κ ∶= dom p ∩ κ , κ . 0 [ η γ ) 0 [ η γ ) • We take q ↾ κ2 ⊇ q ↾ κ2 arbitrary on the given domain; and q ↾ κ , κ 2 ∶= ∗ η ∗ η ∗ [ η γ ) p ↾ κ , κ 2. ∗ [ η γ ) • It remains to define qσ for σ, i ∈ supp q . i ( ) 0 σ σ For σ, i ∈ suppq0, we define q ⊇ q on the given domain ⋃ κν,j, δν,j ( ) i i κν,j <κσ [ ) according to the linking property: Consider an interval κ , κ with [ ν,j ν,j+1) σ δν,j > κν,j. For any ζ ∈ dom q0 ∖dom q0 ∩ κν,j, κν,j+1 , set qi ζ ∶= q∗ ξ,ζ , ( σ ) [ ) ( ) ( ) where ξ ∶= bσ ∩ κ , κ = b ∩ κ , κ . (Note that ξ ∈ dom q { } i [ ν,j ν,j+1) i [ ν,j ν,j+1) 0 by construction). For σ, i ∈ supp p ∖ supp q , we set qσ ↾ κ , κ ∶= pσ ↾ ( ) 0 0 i [ η γ ) i κ , κ , and define qσ ↾ κ ⊇ pσ ↾ κ on the given domain according to the [ η γ ) i η i η linking property as before. ηˆ Finally, q mk for k < ω can be arbitrary on the given domain.

σ σ P Then q0 = q∗, qi , bi σ,i is a condition in 0. In particular, the independence ( ( ) ) σ ηˆ < property holds for the linking ordinals bi : Firstly, by construction of b mk k ω , ηˆ σ ( S ) it follows that b ∩ b = ∅ for any σ, i ∈ supp q0 ∪ supp p . Secondly, whenever mk i ( ) 0 σ0, i0 ∈ supp q0 and σ1, i1 ∈ supp p0 ∖ supp q0, then σ1 > η or i1 ≥ β; hence, (σ ) ( σ) σ σ b 1 = aσ1 ⊆ X ⊆ X . Since b 0 ∩ X = bσ0 ∩ X = ∅, this implies b 0 ∩ b 1 = ∅ as i1 i1 p q i0 q i0 q i0 i1 desired. Thus, the independence property holds for q0.

σ Moreover, ρβ q = q ↾ κ2, qσ, b , X ≤ q by construction; in particu- ( ( ))0 ( ∗ η ( i i )σ≤η,i<β q) 0 lar, X ⊆ X : Consider ξ ∈ X . In the case that ξ ∈ X , it follows that ξ ∈ aσ for q q q p i some σ, i ∈ supp p0 with σ > η, or i ≥ β. Then σ, i ∈ supp p0 ∖ supp q0; hence, σ ( ) σ ( ) b = aσ, and it follows that ξ ∈ b ⊆ X as desired. In the case that ξ ∈ X ∖ X , we i i i q q p ηˆ ∈ < ∈ have ξ b mk for some k ω; so again, ξ Xq as desired.

Finally, q0 ≤ p0 by construction; and it follows that q0 has all the desired properties.

The construction of q is similar. Thus, the map ρβ ∶ P → Pβ ↾ η + 1 as deﬁned 1 ( ) above, is indeed a projection of forcing posets.

It follows that Gβ ↾ η + 1 is a V -generic ﬁlter on Pβ ↾ η + 1 : For genericity, ( ) ( ) consider an open dense set D ⊆ Pβ ↾ η + 1 . It suﬃces to show that the set ( ) D ∶= p ∈ P pβ ↾ η + 1 ∈ D is dense in P. Take a condition p ∈ P, and let { S ( ) } p ≤ p with p ∈ P. Since D ⊆ Pβ ↾ η + 1 is dense, there exists q ∈ Pβ ↾ η + 1 ( ) ( ) with q ≤β pβ ↾ η + 1 . By what we have just shown, we there exists q ≤ p with η+1 ( ) qβ ↾ η + 1 ≤ q. Then q is an extension of p in D as desired. ( )

52 B) Capturing f β.

In this section, we will show that the map f β is contained in a generic extension similar to V Gβ ↾ η + 1 . [ ( )] + Recall that we are working in the case that κη+1 > κη, and β < αη̃ or (η,γ̃ ) ∩ Lim ≠∅, where η ∶= max σ < η σ ∈ Lim . ̃ { S } Recall that any X ∈ dom f is of the form

σm ∏m<ω G∗(am) × ∏m<ω G X = X˙ im ,

η ˙ ω σm where X ∈ Name P × ∏m<ω P and am m<ω, σm, im m<ω is an η-good (( ) ) im ) ( ) ( ) pair. Moreover,

f β ∶= X, α ∈ f ∃ a , σ , i η-good pair ∶ ( ) T ( m)m<ω ( m m)m<ω

σ ∗ m η ω σm ∏m<ω G (am) × ∏m<ω G ∀m im < β ∧ ∃ X˙ ∈ Name P × M P X = X˙ im . ( ) ( ) m<ω η ω Fix an η-good pair ̺ = am m<ω, σm, im m<ω . We use recursion over the Name P ( ) ( ) η (( ) σm ω σm P × ∏m<ω P -hierarchy to deﬁne a map τ̺ ∶ Name P × ∏m<ω P → Name ) η (( ) ) ( ) that maps any name Y˙ ∈ Name P ω × P σm to a name τ Y˙ ∈ Name P (( ) ∏m<ω ) ̺( ) ( ) such that σm ∏m<ω G∗(am) × ∏m<ω G G Y˙ im = τ Y˙ . ( ̺( ))

Deﬁnition 34. For an η-good pair ̺ = am m<ω, σm, im m<ω , we deﬁne recur- η ( ) ( ) sively for Y˙ ∈ Name P ω × P σm : (( ) ∏m<ω )

σm τ̺ Y˙ ∶= τ̺ Z˙ ,q q ∈ P , ∃ Z˙ , p∗ am m<ω , p m<ω ∈ Y˙ ∶ ( ) ( ( ) ) T ( ( )) ( im )

σm σm ∀m q∗ am ⊇ p∗ am , q ⊇ p . ( ) ( ) im im

σm ∏m<ω G∗(am) × ∏m<ω G G It is not diﬃcult to check that indeed, Y˙ im = τ̺ Y˙ holds η ( ( )) for all Y˙ ∈ Name P ω × P σm . (( ) ∏m<ω ) Now, we deﬁne a map f β ′ ⊇ f β, which is contained in an intermediate generic ( ) extension similar to V Gβ ↾ η + 1 . We will then use an isomorphism argument [ ( )] to show that actually, f β ′ = f β. ( ) D D Recall that f = f˙G, where πf π = f π whenever π contained in the intersection [ ] λm ⋂m<ω F ix ηm, im ∩ ⋂m<ω H denoted by A ˙ . ( ) km ( f )

β ηm The idea is that we include into P ↾ η+1 the verticals P for ηm ∈ Lim, ηm > η. ( ) im Below κη, the linking property will be important, so we also have to include the linking ordinals aηm ∩ κ . im η

53 For a condition p ∈ P, we set

X ∶= aσ ∩ κ σ ∈ Lim , σ, i ≠ η , i for all m < ω , σ > η or i ≥ β . ̃p { i η̃ S ( ) ( m m) ( )} Then X is similar to X , but excludes the linking ordinals aηm for η ∈ Lim. ̃p p im m For reasons of notational convenience and better clarity, we introduce the following ad-hoc notation:

Let

β (ηm ,im)m<ω 2 σ σ ηm ηm p ↾ η+1 ∶= p∗ ↾ κ , p , a σ η,i β , p ↾ κη, a ∩ κη m<ω , η >η, ( ( )) η ( i i ) ≤ < ( im im ) m pσ ↾ β × dom pσ , X . ( ( y ))σ≤η ̃p Then pβ ↾ η+1 (ηm ,im)m<ω can be obtained from pβ ↾ η+1 by using X instead ( ( )) ( ) ̃p ηm ηm of Xp, and including p ↾ κη, a ∩ κη for ηm ∈ Lim with ηm > η. (Note that ( im im ) ηm ηm for ηm ≤ η, it follows that im < β, so p , a is already part of the condition ( im im ) pβ ↾ η + 1 .) ( ) We are now ready to define our forcing notion Pβ ↾ η + 1 (ηm,im)m<ω . The order ( ( )) relation is defined similarly as for the forcing notion Pβ ↾ η + 1 ; but additionally, ( ) we require for qβ ↾ η + 1 (ηm ,im)m<ω ≤ pβ ↾ η + 1 (ηm ,im)m<ω that the linking ( ( )) ( ( )) property below κ holds for all η , i with η ∈ Lim, η > η. η ( m m) m m Definition 35. Let Pβ ↾ η + 1 (ηm ,im)m<ω denote the collection of all pβ ↾ ( ( )) ( η + 1 (ηm ,im)m<ω such that p ∈ P (i.e. p ∈ P with σ, i ∈ supp p σ > η or i ≥ ( )) S{( ) 0 S β = ℵ ); together with 1β (ηm,im)m<ω as the maximal element. }S 0 ( η+1) For conditions p, q ∈ P, let qβ ↾ η + 1 (ηm,im)m<ω ≤ pβ ↾ η + 1 (ηm,im)m<ω if ( ( )) ( ( )) • X̃q ⊇ X̃p,

• q ↾ κ2, qσ, bσ , qσ ↾ β × dom qσ ≤ p ↾ κ2, pσ, aσ , pσ ↾ ∗ η ( i i )σ≤η,i<β ( ( y ))σ≤η ∗ η ( i i )σ≤η,i<β ( β × dom pσ regarded as conditions in P, ( y ))σ≤η • ∀ η > η ∶ qηm ↾ κ ⊇ pηm ↾ κ , m im η im η

ηm ηm • ∀ ηm > η , ηm, im ∈ supp p ∶ b = a , ( ) im im

ηm • for all intervals κν,j, κν,j+1 ⊆ κη and ηm > η with a ∩ κν,j, κν,j+1 = ξ , [ ) im [ ) { } ηm it follows that q ζ = q∗ ξ,ζ whenever ζ ∈ dom q ∖ dom p ∩ κν,j, κν,j+1 . im ( ) ( ) ( ) [ ) Finally, for constructing our intermediate generic extension for capturing f β, we also have to include the verticals P ηm ↾ κ , κ for η > η. [ η ηm ) m This gives a product

Pβ (ηm ,im)m<ω ηm ↾ η + 1 × M P ↾ κη, κηm , ( ( )) m<ω [ )

54 which is the set of all

β (ηm ,im)m<ω ηm p ↾ η + 1 , p ↾ κη, κη m<ω ( ( )) ( im [ m )) such that p ∈ P (i.e. p ∈ P with σ, i ∈ supp p0 σ > η or i ≥ β = ℵ0); together β S{( ) S }S with a maximal element 1 (ηm,im)m<ω . ( η+1) Then Gβ ↾ η + 1 (ηm ,im)m<ω × Gηm ↾ κ , κ M im η ηm ( ( )) m<ω [ )

β (ηm ,im)m<ω ηm is the set of all p ↾ η + 1 , p ↾ κη, κη m<ω such that there ( ( )) ( im [ m )) exists q ∈ G ∩ P with qβ ↾ η + 1 (ηm,im)m<ω ≤ pβ ↾ η + 1 (ηm ,im)m<ω and ( ( )) ( ( )) ηm ηm q ↾ κη, κη ⊇ p ↾ κη, κη for all m < ω; together with the maximal element im [ m ) im [ m ) β 1 (ηm,im)m<ω . ( η+1)

β (ηm,im)m<ω ηm In order to show that G ↾ η + 1 × ∏ G ↾ κη, κη is a V - ( ( )) m<ω im [ m ) generic ﬁlter on Pβ ↾ η + 1 (ηm,im)m<ω × P ηm ↾ κ , κ , we proceed ( ( )) ∏m<ω [ η ηm ) similarly as in Proposition 33:

Proposition 36. The map ρβ (ηm,im)m<ω ∶ P → Pβ ↾ η + 1 (ηm ,im)m<ω × ( ) ( ( )) P ηm ↾ κ , κ , ∏m<ω [ η ηm )

β (ηm ,im)m<ω ηm p ↦ p ↾ η + 1 , p ↾ κη, κη m<ω ( ( )) ( im [ m )) β in the case that σ, i ∈ supp p σ > η ∨ i ≥ β = ℵ , and 1 ↦ 1 (ηm,im)m<ω , S{( ) 0 S }S 0 ( η+1) is a projection of forcing posets.

Proof. We closely follow the proof of Proposition 33. Consider p ∈ P with σ, i ∈ S{( ) suppp σ > η ∨ i ≥ β = ℵ , and a condition 0 S }S 0

2 σ σ ηm ηm q = q∗ ↾ κ , q , b σ η,i β , q ↾ κη, b ∩ κη m<ω,η >η , η ( i i ) ≤ < ( im im ) m

σ ηm q σ≤η , Xq , q ↾ κη, κη m<ω ( ) ̃ ( im [ m )) in Pβ (ηm ,im)m<ω ηm ↾ η + 1 × M P ↾ κη, κηm ( ( )) m<ω [ ) σ with q ≤ ρβ (ηm,im)m<ω p . We have to construct q ≤ p, q = q , qσ, b , qσ , ( ) ( ) ( ∗ ( i i )σ,i ( )σ) such that ρβ (ηm,im)m<ω q ≤ q. ( ) ( )

We start with q0.

• Similarly as in Proposition 33, we construct supp = η,mˇ k < ω such ∗ {( k) S } thatη ˇ > η or m ≥ β, and η,mˇ ∉ supp p ∪ supp q for all k < ω; with k ( k) 0 0 the additional property that for all k < ω, we have η,mˇ ∉ η , i m < ( k) {( m m) S ω , η ∈ Lim . We set supp q = supp p ∪ supp q ∪ supp ∪ η , i m < m } 0 0 0 ∗ {( m m) S ω , η ∈ Lim . m }

55 σ • Next, we deﬁne the linking ordinals b for σ, i ∈ supp q , such that X ⊇ X i ( ) 0 ̃q ̃q holds:

First, we consider the case that σ, i ∉ ηm, im m < ω , ηm ∈ Lim . For σ ( ) {( σ ) S } σ, i ∈ supp q , we let b ∶= bσ ⊇ aσ, and b ∶= aσ in the case that σ, i ∈ ( ) 0 i i i i i ( ) supp p0 ∖ supp q0. ηˇ We construct b k < ω as in Proposition 33. ( mk S ) η After that, we deﬁne the linking ordinals b m m < ω , η ∈ Lim with the ( im S m ) following properties:

η – As usual, every b m is a subset of κ that hits any interval κ , κ ⊆ im ηm [ ν,j ν,j+1) κηm in exactly one point. ηm ηm ∩ σ = ∅ < – The bim are pairwise disjoint, and bim bi for every m ω and σ, i ∈ supp q with σ, i ≠ η , i . ( ) 0 ( ) ( m m) η – For every η , i ∈ supp p , we set b m ∶= aηm ; for every η , i ∈ m m 0 im im m m ( ) η ( ) supp q ∖suppp with η ≤ η, we set b m ∶= bηm ; and whenever η , i ∈ 0 0 m im im m m η ( ) supp q ∖ supp p with η > η, we let b m ⊇ bηm ∩ κ . 0 0 m im im η σ This concludes our construction of the linking ordinals bi .

• We deﬁne dom q = κ , δ as follows: Let dom ∶= dom p ∪ dom q ∪ 0 ⋃ν,j[ ν,j ν,j ) 0 0 ηm ⋃ dom q ↾ κη, κη . For every interval κν,j, κν,j+1 with dom ∩ ηm∈Lim im [ m ) [ ) κ , κ = ∅, we set δ ∶= κ ; and whenever dom ∩ κ , κ ≠ ∅, [ ν,j ν,j+1) ν,j ν,j [ ν,j ν,j+1) we pick δν,j ∈ κν,j, κν,j+1 with the property that dom ∩ κν,j, κν,j+1 ⊆ (σ ) [ ) κ , δ , and b ∩ κ , κ ⊆ κ , δ for all σ, i ∈ supp q . [ ν,j ν,j ) i [ ν,j ν,j+1) [ ν,j ν,j) ( ) 0 ηm Since dom p , dom q0 and the domains dom q ↾ κη, κη are bounded 0 im [ m ) below all regular cardinals, this is also true for dom and dom q0.

• 2 2 We take q∗ ↾ κη ⊇ q∗ ↾ κη arbitrary on the given domain. The verticals qσ ↾ κ for σ, i ∈ supp q ∪ supp p ∖ η , i m < ω , η ∈ i η ( ) ( 0 0) {( m m) S m Lim can be defined according to the linking property as in Proposition 33. } The verticals q ηˆ ↾ κ with η,mˆ ∈ supp can be set arbitrarily on the mk η ( k) ∗ given domain. Now, consider η , i with η ∈ Lim. In the case that η , i ∈ supp q ( m m) m ( m m) 0 with η ≤ η, we can proceed as before, and define qηm ⊇ qηm according to the m im im linking property as in Proposition 33. ηm Concerning the verticals q ↾ κη for ηm, im ∈ supp q0 with ηm > η, we im ( ) ηm ηm define q ↾ κν,j, κν,j+1 ⊇ q ↾ κν,j, κν,j+1 on intervals κν,j, κν,j+1 ⊆ κη im [ ) im [ ) [ ) according to the linking property, and use that we have incorporated the ηm β (ηm ,im)m<ω linking ordinals b ∩ κη into our forcing notion P ↾ η + 1 : im ( ( )) ηm For ζ ∈ dom q ∖ dom q0 ∩ κν,j, κν,j+1 , we set q ζ ∶= q ξ,ζ , where ( 0 ) [ ) im ( ) ∗( )

56 ηm ηm ξ = b ∩ κν,j, κν,j+1 = b ∩ κν,j, κν,j+1 . (Note that ξ ∈ dom q by { } im [ ) im [ ) 0 construction.) In the case that η , i ∉ supp q , it follows that also η , i ∉ supp p , ( m m) 0 ( m m) 0 and we can set qηm ↾ κ arbitrarily on the given domain. im η

• Next, consider an interval κ , κ ⊆ κ , κ . We ﬁrst set the verti- [ ν,j ν,j+1) [ η γ ) cals qσ ↾ κ , κ for σ, i ∈ supp q , σ > η, on the given domain, with i [ ν,j ν,j+1) ( ) 0 ηm ηm the property that q ↾ κν,j, κν,j+1 ⊇ q ↾ κν,j, κν,j+1 for all m < ω with im [ ) im [ ) η , i ∈ supp q , and qσ ↾ κ , κ ⊇ pσ ↾ κ , κ whenever σ, i ∈ ( m m) 0 i [ ν,j ν,j+1) i [ ν,j ν,j+1) ( ) supp p . After that, we deﬁne q ↾ κ , κ 2 ⊇ p ↾ κ , κ 2 accord- 0 ∗ [ ν,j ν,j+1) ∗ [ ν,j ν,j+1) ing to the linking property: Whenever ζ ∈ dom q0 ∖ dom p0 ∩ κν,j, κν,j+1 σ ( ) [ ) and ξ = aσ ∩ κ , κ = b ∩ κ , κ for some σ, i ∈ supp p , then { } i [ ν,j ν,j+1) i [ ν,j ν,j+1) ( ) 0 q ξ,ζ ∶= qσ ζ . Otherwise, q ξ,ζ can be set arbitrarily. ∗( ) i ( ) ∗( )

This deﬁnes q0. The construction of q1 is similar; and it is not diﬃcult to see that q ≤ p with ρβ (ηm,im)m<ω q ≤ q. ( ) ( ) Hence, ρβ (ηm,im)m<ω is a projection of forcing posets. ( )

β (ηm,im)m<ω ηm Thus, it follows that G ↾ η+1 × ∏ G ↾ κη, κη is a V -generic ( ( )) m<ω im [ m ) ﬁlter on the forcing notion Pβ ↾ η + 1 (ηm,im)m<ω × P ηm ↾ κ , κ . ( ( )) ∏m<ω [ η ηm ) The aim of Chapter 6.2 B) is to show that f β is contained in the intermediate β (ηm ,im)m<ω ηm V -generic extension V G ↾ η + 1 × ∏ G ↾ κη, κη . [ ( ( )) m<ω im [ m ) ] Deﬁnition 37. Let f β ′ denote the set of all X, α for which there exists an ( ) ( ) η-good pair ̺ = a , σ , i with i < β for all m < ω such that ( m)m<ω ( m m)m<ω m

σm ∏m G∗(am) × ∏m G X = X˙ im , and there is a condition p ∈ P with

• σ, i ∈ suppp σ > η or i ≥ β = ℵ , S{( ) 0 S }S 0 • p ⊩ τ X˙ , α ∈ f˙ s ̺( )

β (ηm ,im)m<ω ηm β (ηm ,im)m<ω • p ↾ η+1 , p ↾ κη, κη m<ω ∈ G ↾ η+1 × ( ( )) ( im [ m )) ( ( )) ηm ∏ G ↾ κη, κη , m<ω im [ m )

ηm ηm • ∀ ηm ∈ Lim ∶ ηm, im ∈ supp p0 with a = g . ( ) im im

β ′ β (ηm ,im)m<ω ηm Then f ∈ V G ↾ η + 1 × ∏ G ↾ κη, κη , since the se- ( ) [ ( ( )) m<ω im [ m ) ] quence gηm m < ω is contained in the ground model V . ( im S ) We will now use an isomorphism argument and show that f β = f β ′. ( ) Proposition 38. f β = f β ′. ( )

57 Proof. By the forcing theorem, it follows that f β ′ ⊇ f β. Assume towards a ( ) contradiction, there was X, α ∈ f β ′ ∖ f β. Let ( ) ( )

σm ∏m<ω G∗(am) × ∏m<ω G X = X˙ im for an η-good pair ̺ = a , σ , i with i < β for all m < ω. Take ( m)m<ω ( m m)m<ω m p ∈ P as in Definition 37 with p ⊩ τ , α ∈ f˙; and since X, α ∉ f β, we can s ( ̺(X˙ ) ) ( ) take p′ ∈ G with p′ ⊩ τ X˙ , α ∉ f˙ and η , i ∈ supp p′ for all η ∈ Lim. s ( ̺( ) ) ( m m) 0 m Our first step will be to extend the conditions p and p′ and obtain p ≤ p, p′ ≤ p′ such that p and p′ have “the same shape” similarly as in the Approximation Lemma 21; but additionally, pβ ↾ η + 1 = p′ β ↾ η + 1 holds, and pηm = p′ ηm for all ( ) ( ) ( ) im ( )im ηm ′ ηm m < ω, and a = a for all m < ω with ηm ∈ Lim. im ( )im After that, we construct an isomorphism π such that firstly, πp = p′; secondly, π Dπ Dπ should not disturb the forcing Pβ ↾ η+1 (which will imply π τ X˙ = τ X˙ ); ( ) ̺( ) ̺( ) λm and thirdly π should be contained in the intersection ⋂m F ix ηm, im ∩ ⋂m H [ ] ( ) km D D (which implies πf π = f π ). Dπ D Then from p ⊩ τ X˙ , α ∈ f˙ it follows πp ⊩ π τ X˙ , α ∈ πf π . Together s ( ̺( ) ) s ( ̺( ) ) Dπ D with p′ ⊩ τ X˙ , α ∉ f π , this gives our desired contradiction. s ( ̺( ) ) In order to make such an isomorphism π possible, the extensions p ≤ p and p′ ≤ p′ will satisfy the following properties:

• ′ supp0 ∶= suppp0 = supp p0

• ′ dom0 ∶= dom p0 = dom p0

• ⋃ a ∶= ⋃ aσ = ⋃ a′ σ (σ,i)∈supp0 i (σ,i)∈supp0 ( )i • ∀ ν , j ∶ dom ∩ κ , κ ≠ ∅ ⇒ a ∩ κ , κ ⊆ dom 0 [ ν,j ν,j+1) ⋃ [ ν,j ν,j+1) 0 • ′ supp1 ∶= suppp1 = supp p1

• ∀ σ ∈ supp ∶ dom σ ∶= dom pσ = dom p′ σ. 1 1( ) ( )

Additionally, we want:

• ∀ m < ω ∶ pηm = p′ ηm im ( )im

ηm ′ ηm • ∀ m < ω , ηm ∈ Lim ∶ a = a im ( )im • pβ ↾ η + 1 = p′ β ↾ η + 1 , i.e. ( ) ( ) ( ) 2 ′ 2 – p∗ ↾ κη = p∗ ↾ κη – ∀ σ ∈ Lim, σ ≤ η, i < min α , β ∶ pσ = p′ σ, aσ = a′ σ { σ } i ( )i i ( )i – ∀ σ ∈ Succ σ ≤ η ∶ pσ ↾ β × dom pσ = p′ σ ↾ β × dom p′ σ . ( y ) ( ) ( y( ) )

58 Then it follows that X̃p = X̃p′ .

ηm ′ ηm ηm ′ ηm ηm Note that a = a for ηm ∈ Lim follows automatically, since a = a = g im ( )im im ( )im im by assumption.

Now, we construct the conditions p and p′.

We start with the linking ordinals aσ and a′ σ, with our aim that aσ = i ( )i ⋃σ,i i a′ σ =∶ a. We closely follows our construction from the Approximation ⋃σ,i( )i ⋃ Lemma 21; but now, some extra care is needed, since we additionally have to make sure that aσ = a′ σ holds for all σ ≤ η, i < β. i ( )i Similarly as in the Approximation Lemma 21, let

s ∶= κ ∶= sup κ σ ∈ Lim , ∃ i < α σ, i ∈ supp p ∪ supp p′ . δ { σ S σ ( ) 0 0}

Recall that we are assuming β < α or Lim ∩ (η,γ̃ )≠∅, where η ∶= max σ ≤ η̃ ̃ { η σ ∈ Lim . S } In the case that κ = κ , we set γ ∶= δ and take σ , l k < ω such that δ γ (( k k) S ) sup κ k < ω = κ = κ , and σ , l ∉ suppp ∪ supp p′ for all k < ω, with the { σk S } γ γ ( k k) 0 0 additional property that σ > η or l ≥ β for all k < ω. k ̃ k If κ < κ and Lim ∩ η, γ ≠ ∅, let γ ∈ Lim ∩ η, γ with γ ≥ δ, and take δ γ (̃ ) (̃ ) σ , l k < ω such that σ , l = γ, l ∉ supp p ∪ supp p′ for all k < ω. (( k k) S ) ( k k) ( k) 0 0 Finally, if κ < κ and Lim ∩ η, γ = ∅, then β < α follows. In this case, let δ γ (̃ ) η̃ γ ∶= η ≥ δ, and take σ , l k < ω with σ , l = γ, l ∉ supp p ∪ supp p′ for ̃ (( k k) S ) ( k k) ( k) 0 0 all k < ω; with the additional property that lk ≥ β for all k < ω.

Let

supp ∶= supp p ∶= supp p′ ∶= suppp ∪ supp p′ ∪ σ , l k < ω . 0 0 0 0 0 {( k k) S }

We now construct the linking ordinals aσ. For any σ, i ∈ supp p , we set aσ ∶= aσ; i ( ) 0 i i and whenever σ, i ∈ supp p′ ∖ supp p with σ ≤ η, i < β, then aσ ∶= a′ σ. ( ) 0 0 i ( )i Now, take a set Z ⊆ κ such that for all intervals κ , κ ⊆ κ , we have γ [ ν,j ν,j+1) γ σ ′ σ Z ∩ κν,j, κν,j+1 = ℵ0, and Z ∩ ⋃(σ,i)∈supp p a ∪ ⋃(σ,i)∈supp p′ a = ∅. Let S [ )S 0 i 0 ( )i

σ ′ σ Z ∶= Z ∪ ai ∪ a i . σ,i σ,i( )

Our aim is to construct p and p′ with aσ = a′ σ = a ∶= Z. ⋃σ,i i ⋃σ,i( )i ⋃ Fix an interval κ , κ ⊆ κ . Let [ ν,j ν,j+1) γ

Z ∶= aσ σ, i ∈ suppp ∪ a′ σ σ, i ∈ suppp′ ,σ ≤ η, i < β ∩ ν,j { i S ( ) 0} {( )i S ( ) 0 }

∩ κ , κ [ ν,j ν,j+1)

59 and ξ ν, j k < ω ∶= Z ∩ κ , κ ∖ Z . { k( ) S } [ ν,j ν,j+1) ν,j

This set has cardinality ℵ0 by construction of Z.

Now, let

σ , l k < ω =∶ σ, i ∈ supp p ∖ supp p κ < κ and σ > η or i ≥ β . {( k k) S } {( ) 0 0 S ν,j σ ( )}

This set also has cardinality ℵ0 by construction of supp p0. Now, for any k < ω, we let

σk a ∩ κν,j, κν,j+1 ∶= ξk ν, j . lk [ ) { ( )} We apply the same construction to the linking ordinals a′ σ for σ, i ∈ supp p′ = ( )i ( ) 0 supp . It is not diﬃcult to see that aσ = a′ σ = a = Z, the independence 0 ⋃σ,i i ⋃σ,i( )i ⋃ property holds, and aσ = a′ σ whenever σ ≤ η, i < β. i ( )i Next, take dom ∶= dom p = dom p′ = κ , δ with the property that ﬁrstly, 0 0 0 ⋃ν,j[ ν,j ν,j ) dom p ∪ dom p′ ⊆ dom , and secondly, for every interval κ , κ ⊆ κ with 0 0 0 [ ν,j ν,j+1) γ dom ∩ κ , κ ≠ ∅, it follows that Z ∩ κ , κ ⊆ dom . 0 [ ν,j ν,j+1) [ ν,j ν,j+1) 0 It remains to construct p , p′ , and pσ, p′ σ for σ, i ∈ supp . ∗ ∗ i ( )i ( ) 0 First, we consider an interval κ , κ ⊆ κ . [ ν,j ν,j+1) η We start with the construction of p ↾ κ , κ 2 = p′ ↾ κ , κ 2. ∗ [ ν,j ν,j+1) ∗ [ ν,j ν,j+1) Let ξ, ζ ∈ κ , κ ∩ dom . [ ν,j ν,j+1) 0

• In the case that ξ,ζ ∈ dom p × dom p , we set p′ ξ,ζ ∶= p ξ,ζ ∶= p ξ,ζ . ( ) 0 0 ∗( ) ∗( ) ∗( ) • If ξ,ζ ∈ dom p′ × dom p′ , then p′ ξ,ζ ∶= p ξ,ζ ∶= p′ ξ,ζ . ( ) 0 0 ∗( ) ∗( ) ∗( )

For ξ,ζ ∈ dom p × dom p ∩ dom p′ × dom p′ , this is not a contradiction, ( ) ( 0 0) ( 0 0) 2 ′ 2 since p∗ ↾ κη and p∗ ↾ κη are compatible.

• ′ If ζ ∈ dom p0 ∖ dom p0 and ξ ∉ dom p0, we proceed as follows: In the case that ξ = aσ ∩ κ , κ for some σ, i ∈ supp p with σ ≤ η, i < β or { } i [ ν,j ν,j+1) ( ) 0 σ, i ∈ η , i m < ω , we set p′ ξ,ζ ∶= p ξ,ζ ∶= pσ ζ . Otherwise, we ( ) {( m m) S } ∗( ) ∗( ) i ( ) set p′ ξ,ζ = p ξ,ζ arbitrarily. ∗( ) ∗( ) • ′ ′ In the case that ζ ∈ dom p0 ∖ dom p0 and ξ ∉ dom p0, we proceed as before: If ξ = a′ σ ∩ κ , κ for some σ, i ∈ supp p′ with σ ≤ η, i < β or { } ( )i [ ν,j ν,j+1) ( ) 0 σ, i ∈ η , i m < ω , then p′ ξ,ζ ∶= p ξ,ζ ∶= p′ σ ζ . Otherwise, ( ) {( m m) S } ∗( ) ∗( ) ( )i ( ) we set p′ ξ,ζ = p ξ,ζ arbitrarily. ∗( ) ∗( ) • In all other cases, p′ ξ,ζ = p ξ,ζ can be set arbitrarily. ∗( ) ∗( )

60 This deﬁnes p ↾ κ , κ 2 = p′ ↾ κ , κ 2. ∗ [ ν,j ν,j+1) ∗ [ ν,j ν,j+1) Now, consider σ, i ∈ supp . We deﬁne pσ and p′ σ on the interval κ , κ ⊆ ( ) 0 i ( )i [ ν,j ν,j+1) κη as follows:

• For σ, i ∈ supp p , we deﬁne pσ ↾ κ , κ ⊇ pσ ↾ κ , κ ac- ( ) 0 i [ ν,j ν,j+1) i [ ν,j ν,j+1) cording to the linking property: Let ξ ∶= aσ ∩ κ , κ and con- { } i [ ν,j ν,j+1) sider ζ ∈ κ , κ ∩ dom . If ζ ∈ dom p , we set pσ ζ ∶= pσ ζ ; and [ ν,j ν,j+1) 0 0 i ( ) i ( ) pσ ζ ∶= p ξ,ζ in the case that ζ ∈ dom ∖ dom p . (Note that ξ ∈ dom i ( ) ∗( ) 0 0 0 follows by construction.)

• In the case that σ, i ∈ supp p′ , we deﬁne p′ σ ↾ κ , κ ⊇ p′ σ ↾ ( ) 0 ( )i [ ν,j ν,j+1) ( )i κ , κ according to the linking property as before: Let ξ ∶= a′ σ ∩ [ ν,j ν,j+1) { } ( )i κ , κ , and consider ζ ∈ κ , κ ∩ dom . If ζ ∈ dom p′ , we set [ ν,j ν,j+1) [ ν,j ν,j+1) 0 0 p′ σ ζ ∶= p′ σ ζ ; and p′ σ ζ ∶= p′ ξ,ζ in the case that ζ ∈ dom ∖ ( )i ( ) ( )i ( ) ( )i ( ) ( ∗)( ) 0 ′ dom p0. (Again, ξ ∈ dom0 by construction.)

• For σ, i ∈ supp p ∖ suppp′ , let p′ σ ↾ κ , κ ∶= pσ ↾ κ , κ . ( ) 0 0 ( )i [ ν,j ν,j+1) i [ ν,j ν,j+1) • For σ, i ∈ supp p′ ∖ suppp , let pσ ↾ κ , κ ∶= p′ σ ↾ κ , κ . ( ) 0 0 i [ ν,j ν,j+1) ( )i [ ν,j ν,j+1) • If σ, i ∈ supp ∖ supp p ∪ supp p′ , then pσ ↾ κ , κ = p′ σ ↾ ( ) 0 ( 0 0) i [ ν,j ν,j+1) ( )i κ , κ can be set arbitrarily on the given domain. [ ν,j ν,j+1)

This deﬁnes all pσ and p′ σ for σ, i ∈ supp on intervals κ , κ ⊆ κ . i ( )i ( ) 0 [ ν,j ν,j+1) η We now have to verify that pσ = p′ σ for any σ, i ∈ supp with σ ≤ η, i < β. We i ( )i ( ) 0 only have to treat the case that σ, i ∈ supp p ∩ supp p′ . ( ) 0 0 Consider an interval κ , κ ⊆ κ ⊆ κ . Then p′ ∈ G and [ ν,j ν,j+1) σ η

pβ ↾ η + 1 (ηm ,im)m<ω ∈ Gβ ↾ η + 1 (ηm,im)m<ω ( ( )) ( ( )) implies that pσ and p′ σ are compatible, and aσ ∩ κ , κ = a′ σ ∩ κ , κ =∶ i ( )i i [ ν,j ν,j+1) ( )i [ ν,j ν,j+1) ξ . { } Let ζ ∈ κ , κ ∩ dom . [ ν,j ν,j+1) 0

• If ζ ∈ dom p ∩ dom p′ , then pσ ζ = pσ ζ = p′ σ ζ = p′ σ ζ . 0 0 i ( ) i ( ) ( )i ( ) ( )i ( ) • For ζ ∈ dom ∖ dom p ∪ dom p′ , it follows that pσ ζ = p ξ,ζ = p′ ξ,ζ = 0 ( 0 0) i ( ) ∗( ) ∗( ) p′ σ ζ by construction, since we have arranged p ↾ κ2 = p′ ↾ κ2. ( )i ( ) ∗ η ∗ η • Let now ζ ∈ dom p ∖ dom p′ , ξ ∉ dom p . Then pσ ζ = pσ ζ , and p′ σ ζ = 0 0 0 i ( ) i ( ) ( )i ( ) p′ ξ,ζ . Since p′ ξ,ζ = pσ ζ by construction of p′ , this gives pσ ζ = ∗( ) ∗( ) i ( ) ∗ i ( ) p′ σ ζ as desired. ( )i ( ) ′ ′ The case that ζ ∈ dom p0 ∖ dom p0, ξ ∉ dom p0, can be treated similarly.

61 • If ζ ∈ dom p ∖ dom p′ and ξ ∈ dom p , it follows that pσ ζ = pσ ζ and 0 0 0 i ( ) i ( ) p′ σ ζ = p′ ξ,ζ as before; but in this case, we have set p′ ξ,ζ ∶= p ξ,ζ , ( )i ( ) ∗( ) ∗( ) ∗( ) so it remains to verify that pσ ζ = p ξ,ζ . i ( ) ∗( ) Since p′ ∈ G, pβ ↾ η + 1 (ηm ,im)m<ω ∈ Gβ ↾ η + 1 (ηm,im)m<ω , we can ( ( )) ( ( )) take q ∈ G with qβ ↾ η + 1 (ηm ,im)m<ω ≤ pβ ↾ η + 1 (ηm,im)m<ω , and ( ( )) ( ( )) assume w.l.o.g. that q ≤ p′. Then qσ ζ = q ξ,ζ by the linking property i ( ) ∗( ) for q ≤ p′, since a′ σ ∩ κ , κ = ξ . Moreover, pσ ζ = qσ ζ and ( )i [ ν,j ν,j+1) { } i ( ) i ( ) p ξ,ζ = q ξ,ζ , and we are done. ∗( ) ∗( ) • ′ ′ The remaining case is that ζ ∈ dom p0 ∖ dom p0 and ξ ∈ dom p0. Then pσ ζ = p ξ,ζ = p′ ξ,ζ and p′ σ ζ = p′ σ ζ , and it remains to ver- i ( ) ∗( ) ∗( ) ( )i ( ) ( )i ( ) ify that p′ σ ζ = p′ ξ,ζ . As before, take q ∈ G with q ≤ p′ and qβ ↾ ( )i ( ) ∗( ) ( η+1 (ηm ,im)m<ω ≤ pβ ↾ η+1 (ηm ,im)m<ω . The latter gives qσ ζ = q ξ,ζ ( )) ( ( )) i ( ) ∗( ) by the linking property, since σ ≤ η, i < β, aσ ∩ κ , κ = ξ and i [ ν,j ν,j+1) { } ζ ∈ dom q ∖ dom p . Moreover, from q ≤ p′ it follows that p′ σ ζ = qσ ζ 0 0 ( )i ( ) i ( ) and p′ ξ,ζ = q ξ,ζ ; hence, p′ σ ζ = p′ ξ,ζ as desired. ∗( ) ∗( ) ( )i ( ) ∗( ) Thus, it follows that pσ = p′ σ holds for all σ, i ∈ supp with σ ≤ η, i < β. i ( )i ( ) 0

ηm ′ ηm If m < ω with ηm ≤ η, then im < β follows by construction of β. Hence, p = p . im ( )im It remains to make sure that whenever m < ω with η > η, then pηm ↾ κ = m im η ′ ηm σ ′ σ p ↾ κη holds; which can be shown similarly as p = p in the case that ( )im i ( )i σ ≤ η, i < β: We use that aηm = a′ ηm and pηm ζ = p′ ηm ζ for all m < ω and im ( )im im ( ) ( )im ( ) ′ ζ ∈ dom p0 ∩ dom p0; and now, it is important that we are using the forcing notion Pβ ↾ η + 1 (ηm,im)m<ω instead of Pβ ↾ η + 1 ; since we need the linking property ( ( )) ( ) below κ for the η , i with η > η. η ( m m) m It remains to construct p ↾ κ , κ 2, p′ ↾ κ , κ 2, and pσ ↾ κ , κ , p′ σ ↾ ∗ [ η γ ) ∗ [ η γ ) i [ η γ ) ( )i κ , κ for all σ, i ∈ supp with σ > η. [ η γ ) ( ) 0

ηm ηm • For ηm, im with ηm > η, we take p ↾ κη, κη ⊇ p ↾ κη, κη , ( ) im [ m ) im [ m ) ′ ηm ′ ηm p ↾ κη, κη ⊇ p ↾ κη, κη on the given domain, such that ( )im [ m ) ( )im [ m ) ηm ′ ηm ′ p ↾ κη, κη = p ↾ κη, κη . This is possible, since p ∈ G and im [ m ) ( )im [ m ) pβ ↾ η + 1 (ηm ,im)m<ω ∈ Gβ ↾ η + 1 (ηm,im)m<ω ; so pηm and p′ ηm are ( ( )) ( ( )) im ( )im compatible for all m < ω.

• For the σ, i ∈ supp remaining, we set pσ ↾ κ , κ ⊇ pσ ↾ κ , κ and ( ) 0 i [ η γ ) i [ η γ ) p′ σ ↾ κ , κ ⊇ p′ σ ↾ κ , κ arbitrarily on the given domain. ( )i [ η γ ) ( )i [ η γ ) • Consider an interval κ , κ ⊆ κ , κ . We deﬁne p ↾ κ , κ 2 ⊇ [ ν,j ν,j+1) [ η γ ) ∗ [ ν,j ν,j+1) p ↾ κ , κ 2 according to the linking property: Whenever ζ ∈ dom ∖ ∗ [ ν,j ν,j+1) 0 dom p and ξ = aσ ∩ κ , κ for some σ, i ∈ supp p , then p ξ,ζ ∶= 0 { } i [ ν,j ν,j+1) ( ) 0 ∗( ) pσ ζ . i ( ) The construction of p′ ↾ κ , κ 2 ⊇ p′ ↾ κ , κ 2 is similar. ∗ [ ν,j ν,j+1) ∗ [ ν,j ν,j+1)

62 ′ ′ This completes our construction of p0 ≤ p0 and p0 ≤ p0 with all the desired properties.

′ ′ ′ Similarly, one can construct p1 ≤ p1, p1 ≤ p1 such that supp1 ∶= supp p1 = supp p1, dom σ ∶= dom p σ = dom p′ σ for all σ ∈ supp ; and pσ = p′ σ for all σ ≤ η, 1( ) 1( ) 1( ) 1 i ( )i ηm ′ ηm i < β with σ ∈ Succ, and p = p for all m < ω with ηm ∈ Succ. im ( )im We now proceed similarly as in the Approximation Lemma 21 and construct an isomorphism π such that π a standard isomorphism for πp = p′. This determines all parameters of π except the maps G ν, j ∶ supp π ν, j → supp π ν, j , which 0( ) 0( ) 0( ) will be deﬁned as follows: Consider an interval κ , κ . Recall that we have [ ν,j ν,j+1) the map F ν, j ∶ suppπ ν, j → supp π ν, j , which is in charge of permuting π0 ( ) 0( ) 0( ) the linking ordinals: We set F ν, j σ, i ∶= λ, k for a′ σ ∩ κ , κ = aλ ∩ π0 ( )( ) ( ) ( )i [ ν,j ν,j+1) k κ , κ . We deﬁne G ν, j ∶= F ν, j for all κ < κ , and G ν, j ∶= id [ ν,j ν,j+1) π0 ( ) π0 ( ) ν,j η π0 ( ) whenever κν,j ≥ κη.

By construction, it follows that πp = p′. We will now check that π is contained [ ] λm in the intersection ⋂m F ix ηm, im ∩ ⋂m H . ( ) km

• Consider m < ω with η ∈ Lim and r ∈ D , r′ ∶= πr, with η , i ∈ supp r . m π ( m m) 0 For an interval κ , κ ⊆ κ and ζ ∈ dom π ∩ κ , κ , it follows [ ν,j ν,j+1) ηm 0 [ ν,j ν,j+1) ′ ηm ηm ηm by construction of the map π0 ζ that r ζ = r ζ holds; since p = ( ) ( )im ( ) im ( ) im p′ ηm . ( )im ′ ηm In the case that ζ ∈ κν,j, κν,j+1 ∩ dom r0 ∖ dom π0 , it follows that r ζ = [ ) ( ) ( )im ( ) rλ ζ with λ, k = G ν, j η , i . If κ < κ , then λ, k = G ν, j η , i k ( ) ( ) π0 ( )( m m) ν,j η ( ) π0 ( )( m m) ηm ′ ηm = Fπ ν, j ηm, im = ηm, im , since a = a . In the case that κν,j ≥ κη, 0 ( )( ) ( ) im ( )im we have G ν, j = id; so again, λ, k = η , i . π0 ( ) ( ) ( m m) ηm ′ ηm Hence, r ζ = r ζ holds for all ζ ∈ dom r0 ∩ κη . im ( ) ( )im ( ) m This proves π ∈ F ix η , i in the case that η ∈ Lim. For η ∈ Succ, we [ ] ( m m) m m obtain π ∈ F ix η , i as in the Approximation Lemma. [ ] ( m m) • Consider m < ω with λ ∈ Lim. In the case that λ > η, we have G ν, j λ , i m m π0 ( )( m ) λm = λm, i for all κν,j ∈ κη, κλ , and π ∈ H follows. If λm ≤ η, it follows ( ) [ m ) [ ] km that km < β by construction of β. Hence, whenever κν,j < κλm and i ≤ km, we have G ν, j λ , i = F ν, j λ , i = λ , i ; since aλm = a′ λm follows π0 ( )( m ) π0 ( )( m ) ( m ) i ( )i from λm ≤ η, i < β.

λm In the case that λm ∈ Succ, we obtain π ∈ H as in the Approximation [ ] km Lemma.

λm Dπ Thus, we have shown that π ∈ ⋂m F ix ηm, im ∩ ⋂m H ; which implies πf = [ ] ( ) km D Dπ Dπ f π . It remains to make sure that πτ X˙ = τ X˙ . ̺( ) ̺( )

63 Recall that we have an η-good pair ̺ = am m<ω, σm, im m<ω with im < β for η (( ) ( ) ) all m < ω, and X˙ ∈ Name P ω × P σm with ( ) ∏m<ω

σm τ̺ X˙ = τ̺ Y˙ ,q q ∈ P , ∃ Y˙ , p∗ am m<ω , p m<ω ∈ X˙ ∶ ( ) ( ) T ( ( )) ( im )

σm σm ∀ m q∗ am ⊇ p∗ am ,q ⊇ p . ( ) ( ) im im Then

Dπ Dπ τ X˙ = τ Y˙ ,q q ∈ D , Y˙ ∈ dom X,q˙ ⊩ τ Y˙ ∈ τ X˙ , ̺( ) ̺( ) T π s ̺( ) ̺( ) and

Dπ Dπ πτ X˙ = πτ Y˙ ,πq πq ∈ D , Y˙ ∈ dom X,q˙ ⊩ τ Y˙ ∈ τ X˙ . ̺( ) ̺( ) S π s ̺( ) ̺( ) We will now check that π is the identity on Pβ ↾ η + 1 . More precisely: Let ( ) q ∈ D , q = q , qσ, bσ , qσ with πq = q′ = q′ , q′ σ, b′ σ , q′ σ . We π ( ∗ ( i i )σ,i ( )σ) ∗ (( )i ( )i )σ,i (( ) )σ prove that q′ ↾ κ2 = q ↾ κ2; moreover, q′ σ = qσ, b′ σ = bσ for all σ ≤ η, i < β ∗ η ∗ η ( )i i ( )i i with σ ∈ Lim, and q′ σ = qσ for all σ ≤ η, i < β with σ ∈ Succ. ( )i i • ′ ′ 2 2 Since π is a standard isomorphism for πp = p , it follows that q∗ ↾ κη = q∗ ↾ κη for all q ∈ D ; since firstly, p ↾ κ2 = p′ ↾ κ2, and secondly, G ν, j = π ∗ η ∗ η π0 ( ) F ν, j for all κ < κ . The latter makes sure that q′ ξσ ν, j ,ζ = π0 ( ) ν,j η ∗( i ( ) ) q ξσ ν, j ,ζ whenever ζ ∈ dom q ∖dom π , and ξσ ν, j ∶= bσ ∩ κ , κ ∗( i ( ) ) 0 0 { i ( )} i [ ν,j ν,j+1) for some σ, i ∈ supp π ν, j : We have q′ ξσ ν, j ,ζ = q ξλ ν, j ,ζ with ( ) 0( ) ∗( i ( ) ) ∗( k ( ) ) λ, k = G ν, j ○ F ν, j −1 σ, i ; so from G ν, j = F ν, j it follows ( ) π0 ( ) ( π0 ( )) ( ) π0 ( ) π0 ( ) that q′ ξσ ν, j ,ζ = q ξσ ν, j ,ζ as desired. ∗( i ( ) ) ∗( i ( ) ) • Let now σ, i ∈ supp π = supp p with σ ≤ η, i < β and σ ∈ Lim. Then ( ) 0 0 aσ = a′ σ; hence, F ν, j σ, i = σ, i for all κ < κ . This gives b′ σ = bσ i ( )i π0 ( )( ) ( ) ν,j σ ( )i i as desired. For ζ ∈ dom π = dom p , it follows from pσ = p′ σ by construction 0 0 i ( )i of π that q′ σ ζ = qσ ζ holds. Finally, if ζ ∈ dom q ∖ dom π , and ζ 0 ( )i ( ) i ( ) ( 0 0) is contained in an interval κ , κ ⊆ κ , then q′ σ ζ = qλ ζ with [ ν,j ν,j+1) σ ( )i ( ) k ( ) λ, k = G ν, j σ, i = F ν, j σ, i = σ, i as desired. Hence, it follows ( ) π0 ( )( ) π0 ( )( ) ( ) that q′ σ = qσ for all σ ≤ η, i < β. ( )i i • In the case that σ ≤ η, i < β with σ ∈ Succ, we obtain q′ σ = qσ from ( )i i pσ = p′ σ as in the Approximation Lemma 21. i ( )i Hence, π is the identity on Pβ ↾ η + 1 . ( ) η Now, it is not difficult to prove recursively that for every Z˙ ∈ Name P ω × (( ) σm P ˙ πH ∏m<ω P the following holds: If H is a V -generic filter on , then τ̺ Z = ) −1 ( ( )) τ Z˙ H = τ Z˙ π H . ( ̺( )) ( ̺( )) Dπ Dπ This implies τ X˙ = πτ X˙ , since for every q ∈ D and Y˙ ∈ dom X˙ , we have ̺( ) ̺( ) π q ⊩ τ Y˙ ∈ τ X˙ if and only if πq ⊩ τ Y˙ ∈ τ X˙ holds. s ̺( ) ̺( ) s ̺( ) ̺( )

64 Summing up, this gives our desired contradiction: Since p ⊩ τ X˙ , α ∈ f˙, it s ̺( ) Dπ D Dπ D follows that πp ⊩ πτ X˙ , α ∈ πf π ; hence, p′ ⊩ τ X˙ , α ∈ f π . But s ̺( ) s ̺( ) this contradicts p′ ⊩ τ X˙ , α ∉ f˙. s ( ̺( ) ) Thus, our assumption that X, α ∈ f β ′ ∖ f β was wrong, and it follows that ( ) ( ) f β ′ = f β as desired. ( )

β β (ηm ,im)m<ω ηm Hence, f ∈ V G ↾ η + 1 × ∏ G ↾ κη, κη . [ ( ( )) m<ω im [ m ) ]

β (ηm,im)m<ω ηm C) P ↾ (η + 1)) × P ↾ [κ ,κ m ) preserves cardinals ≥ α . ( ∏m<ω η η η The next step is to show that cardinals ≥ α are absolute between V and V Gβ ↾ η [ (ηm ,im)m<ω ηm η + 1 × ∏ G ↾ κη, κη . ( )) m<ω im [ m )] Recall that we are assuming GCH in our ground model V , which will be used im- plicitly throughout this Chapter 6.2 C): When we claim that a particular forcing notion preserves cardinals, then we mean it preserves cardinals under the assumption that GCH holds, if not stated diﬀerently.

First, we have a look at the cardinality of Pβ ↾ η + 1 (ηm,im)m<ω . Recall that β ( ( )) was an ordinal large enough for the intersection A with κ+ < β < α . ( f˙) η η

Lemma 39. Pβ ↾ η + 1 (ηm ,im)m<ω ≤ β +. S( ( )) S S S

Proof. The forcing notion Pβ ↾ η + 1 (ηm ,im)m<ω is the set of all ( ( ))

σ σ σ σ ηm ηm p∗ ↾ κη , p , a σ≤η,i<β, p ↾ β × domy p σ≤η , p ↾ κη, a ∩ κη m<ω , η >η , Xp ( i i ) ( ( )) ( im im ) m ̃ for p ∈ P with σ, i ∈ supp p σ > η ∨ i ≥ β = ℵ , together with the max- S{( ) 0 S }S 0 imal element 1β (ηm,im)m<ω . Since X ⊆ κ , there are only κ+ ≤ β -many ( η+1) ̃p η η S S possibilities for X ; and there are only ≤ κ+ ≤ β -many possibilities for p ↾ κ2 ̃p η S S ∗ η ηm ηm σ σ ℵ0 + and p ↾ κη, a ∩ κη m<ω. Concerning p , a σ η,i β, there are β ≤ β - ( im im ) ( i i ) ≤ < S S S S many possibilities for the countable support; and with the support ﬁxed, we have 2κη ℵ0 ≤ κ+ ≤ β -many possibilities for countably many pσ, aσ with σ ≤ η, i < β. ( ) η S S ( i i ) Finally, for pσ ↾ β × dom pσ , there are only η ω ≤ κ+ ≤ β -many pos- ( ( y ))σ≤η S S η S S sibilities for the countable support; and with the countable support ﬁxed, there are ≤ 2SβS ⋅ κη ℵ0 = β +-many possibilities for countably many pσ with dom pσ ⊆ ( ) S S β × κ ⊆ β × κ . Hence, it follows that the forcing notion Pβ ↾ η + 1 (ηm ,im)m<ω σ η ( ( )) has cardinality ≤ β +. S S

Corollary 40. If β + < α , then Pβ ↾ η + 1 (ηm,im)m<ω × P ηm ↾ κ , κ S S η ( ( )) ∏m<ω [ η ηm ) preserves cardinals ≥ αη.

65 Proof. With the same arguments as in Lemma 2, one can show that the forcing P ηm ↾ κ , κ preserves all cardinals. By Lemma 39 above, the ∏m<ω [ η ηm ) forcing Pβ ↾ η + 1 (ηm ,im)m<ω has cardinality ≤ β + (in V ; and hence, also ( ( )) S S in any P ηm ↾ κ , κ -generic extension). It follows that the product ∏m<ω [ η ηm ) Pβ ↾ η +1 (ηm ,im)m<ω × P ηm ↾ κ , κ preserves all cardinals ≥ β ++. ( ( )) ∏m<ω [ η ηm ) S S It remains to consider the case that β + = α . Then by our assumptions on the S S η sequence α 0 < η < γ (cf. Chapter 2), it follows that cf β > ω. Hence, GCH ( η S ) S S gives β ℵ0 = β < α ; and by our proof of Lemma 39, it follows that all compo- S S S S η nents of Pβ ↾ η + 1 (ηm ,im)m<ω have cardinality ≤ β < α ; with the exception ( ( )) S S η of pσ ↾ β × dom pσ , where there might be 2SβS ⋅ κη ℵ0 = β + = α -many ( ( y ))σ≤η ( ) S S η possibilities.

We now have to distinguish several cases depending on whether η is a limit ordinal or not, and depending on whether κη is a limit cardinal or a successor cardinal (i.e. η ∈ Lim or η ∈ Succ). We will have to separate one or two components P σ ↾ β × κ , κ , where ( [ σ σ)) + β (ηm,im)m<ω σ ∈ Succ, σ ≤ η, κσ = κσ , from the forcing notion P ↾ η + 1 ; and ′ ( ( )) obtain a forcing Pβ ↾ η + 1 (ηm,im)m<ω which has cardinality < α , while the ( ( )) η product of the remaining P σ ↾ β × κ , κ and P ηm ↾ κ , κ preserves ( [ σ σ)) ∏m<ω [ η ηm ) cardinals.

Proposition 41. The forcing notion Pβ ↾ η + 1 (ηm ,im)m<ω × P ηm ↾ ( ( )) ∏m<ω κ , κ preserves all cardinals ≥ α . [ η ηm ) η Proof. By Corollary 40, we only have to treat the case that α = β +. Then η S S cf β > ω and β ℵ0 = β . S S S S S S First, we assume that η is a limit ordinal. Then by closure of the sequence κ 0 < σ < γ , it follows that η ∈ Lim, i.e. κ = sup κ 0 < σ < η is a limit ( σ S ) η { σ S } cardinal.

Since the sequence α 0 < σ < γ is strictly increasing (cf. Chapter 2), it follows ( σ S ) that α < β for all σ < η. Hence, for any σ ∈ Succ with σ < η, the forcing σ S S notion P σ ↾ β × κ , κ = P σ ↾ α × κ , κ has cardinality ≤ α+ ≤ β ; ( [ σ σ)) ( σ [ σ σ)) σ S S and we conclude that there are only ≤ η ℵ0 ⋅ β ℵ0 = β -many possibilities for S S S S S S pσ ↾ β × dom pσ . ( ( y ))σ≤η Hence, by the proof of Lemma 39, it follows that Pβ ↾ η + 1 (ηm ,im)m<ω has ( ( )) cardinality ≤ β < α . Like in Corollary 40, this implies that the product Pβ ↾ S S η ( η + 1 (ηm ,im)m<ω × P ηm ↾ κ , κ preserves all cardinals ≥ β + = α as ( )) ∏m<ω [ η ηm ) S S η desired.

The remaining case is that η is a successor ordinal. Let η = η + 1. We now have to distinguish four cases, depending on whether κη and κη are successor cardinals or limit cardinals.

66 If η ∈ Lim and η ∈ Lim, it follows for any P σ ↾ β × κ , κ with σ ≤ η, σ ∈ ( [ σ σ)) Succ that σ < η must hold; hence, α < α < α = β +, which implies α < β . Thus, σ η η S S σ S S the corresponding forcing notion P σ ↾ β × κ , κ = P σ ↾ α × κ , κ has car- ( [ σ σ)) ( σ [ σ σ)) dinality ≤ α+ ≤ β ; and as before, it follows that the forcing Pβ ↾ η+1 (ηm ,im)m<ω σ S S ( ( )) has cardinality ≤ β ℵ0 = β . Like in Corollary 40, this implies that the product S S S S Pβ ↾ η + 1 (ηm ,im)m<ω × P ηm ↾ κ , κ preserves all cardinals ≥ β + = α ( ( )) ∏m<ω [ η ηm ) S S η as desired.

′ If η ∈ Lim and η ∈ Succ, we consider the forcing notion Pβ ↾ η+1 (ηm ,im)m<ω , ( ( )) which is obtained from Pβ ↾ η + 1 (ηm ,im)m<ω by excluding P η ↾ β × κ , κ ; ( ( )) ( [ η η)) i.e. we consider

pσ ↾ β × dom pσ = pσ ↾ β × dom pσ ( ( y ))σ<η ( ( y ))σ<η ′ instead of pσ ↾ β × dom pσ . Then Pβ ↾ η+1 (ηm ,im)m<ω has cardinality ( ( y ))σ≤η ( ( )) ≤ β as before; and it suﬃces to check that the remaining product S S

η ηm P ↾ β × κη, κη × M P ↾ κη, κηm ( [ )) m<ω [ ) preserves all cardinals. The forcing notion P ηm ↾ κ , κ preserves cardinals. Moreover, P ηm ↾ ∏m<ω [ η ηm ) ∏m<ω κ , κ is ≤ κ -closed. Hence, in any V -generic extension by P ηm ↾ [ η ηm ) η ∏m<ω κ , κ the following holds: Firstly, P η ↾ β × κ , κ is the same forcing no- [ η ηm ) ( [ η η)) tion as in V ; and secondly, P η ↾ β × κ , κ preserves cardinals, since 2<κη = κ . ( [ η η)) η Thus, it follows that the product P η ↾ β × κ , κ × P ηm ↾ κ , κ ( [ η η)) ∏m<ω [ η ηm ) preserves all cardinals as desired.

If η ∈ Succ and η ∈ Lim, we proceed similarly, but exclude P η ↾ β × κ , κ ( [ η η)) instead of P η ↾ β × κ , κ . ( [ η η)) If η ∈ Succ and η ∈ Succ, then both P η ↾ β × κ , κ and P η ↾ β × κ , κ ( [ η η)) ( [ η η)) have to be parted from Pβ ↾ η+1 (ηm,im)m<ω . As before, it follows that ﬁrstly, the ( ( )) ′′ remaining forcing notion, denoted by Pβ ↾ η + 1 (ηm ,im)m<ω , has cardinality ( ( )) ≤ β ; and secondly, the remaining product S S

η η ηm P ↾ β × κη, κη × P ↾ β × κη, κη × M P ↾ κη, κηm ( [ )) ( [ )) m<ω [ ) preserves all cardinals.

It follows that Pβ ↾ η + 1 (ηm ,im)m<ω × P ηm ↾ κ , κ preserves all car- ( ( )) ∏m<ω [ η ηm ) dinals ≥ αη.

This concludes our proof by cases.

67 ̃ β ̃ ℵ0 D) A set ´(κη) ⊇ dom f with an injection ι ∶ ´(κη) ↪ SβS .

β (ηm ,im)m<ω ηm In this section, we construct in V G ↾ η+1 × ∏ G ↾ κη, κη [( ( )) m<ω im [ m )] a set ´ κ with ´ κ ⊇ dom f β, together with an injective function ι ∶ ´ κ ↪ ̃( η) ̃( η) ̃( η) ℵ0 V β β (ηm ,im)m<ω ηm β < αη. Since f is contained in V G ↾ η + 1 × ∏ G ↾ (S S ) [( ( )) m<ω im κ , κ by Deﬁnition 37 and Proposition 38, and Pβ ↾ η + 1 (ηm ,im)m<ω × [ η ηm )] ( ( )) P ηm ↾ κ , κ preserves cardinals ≥ α by Proposition 41, this will contra- ∏m<ω [ η ηm ) η β β dict our initial assumption that f ∶ dom f → αη was surjective.

σm Fix an η-good pair ̺ = am m<ω, σm, im m<ω . Then ∏m G∗ am × ∏m G is a im (η ) ( ) ( ) σm V -generic ﬁlter on ∏m P × ∏m P ; and as in Lemma 2, it follows that this forcing preserves cardinals and the GCH. Hence, there is an injection χ ∶´ κ ↪ κ+ V ( η) ( η ) σm in V ∏m G∗ am × ∏m G . [ ( ) im ] Let M be the set of all η-good pairs a , σ , i in V with the prop- β ( m)m<ω ( m m)m<ω erty that i < β for all m < ω. Then M has cardinality ≤ 2κη ℵ0 ⋅ η ℵ0 ⋅ β ℵ0 ≤ m β ( ) S S S S β ℵ0 . S S First, we consider the case that SβS+ = α . Then cf β > ω; hence, GCH gives η S S β ℵ0 = β and there is an injection ψ ∶ M ↪ β in V . S S S S β S S

β By construction of f (cf. Deﬁnition 30), it follows that any X ⊆ κη with X ∈ β σm dom f is contained in a model V ∏m G∗ am × ∏m G for some η-good pair [ ( ) im ] a , σ , i ∈ M . Hence, dom f β is a subset of ( m)m<ω ( m m)m<ω β

´ κ ∶= ´ κ ∩ V G a × Gσm a , σ , i ∈ M . ̃ η η M ∗ m M i m m<ω m m m<ω β ( ) ( ) [ m ( ) m m ] U ( ) ( )

β (ηm ,im)m<ω ηm The set ´ κη can be deﬁned in V G ↾ η + 1 × G ↾ κη, κη , ̃( ) [( ( )) im [ m )] since for any a , σ , i ∈ M , we have a ⊆ κ , and σ ≤ η, i < β ( m)m<ω ( m m)m<ω β m η m m for all m < ω.

η For the rest of this section, we work in V Gβ ↾ η + 1 (ηm ,im)m<ω × ∏ G m ↾ [( ( )) m<ω im κ , κ , and construct there an injective function ι ∶ ´ κ ↪ β V . [ η ηm )] ̃( η) S S For a set X ∈ ´ κ , let ̃( η)

ι X ∶= a , σ , i ̃( ) (( m)m<ω ( m m)m<ω)

σm if am m<ω, σm, im m<ω ∈ Mβ with X ∈ ´ κη ∩ V ∏m G∗ am × ∏m G , and (( ) ( ) ) ( ) [ ( ) im ] ψ a , σ , i is least with this property. (( m)m<ω ( m m)m<ω) η Now, we use the Axiom of Choice in V Gβ ↾ η + 1 (ηm,im)m<ω × ∏ G m ↾ [( ( )) m<ω im κ , κ , and choose for all a , σ , i ∈ M an injection [ η ηm )] (( m)m<ω ( m m)m<ω) β

σm + V χ ∶ ´ κη ∩ V G∗ am × G ↪ κ . ((am)m<ω,(σm,im)m<ω) M M i η ( ) [ m ( ) m m ] ( )

68 Now, we can deﬁne ι ∶ ´ κ ↪ κ+ V ⋅ β V as follows: For X ∈ ´ κ , let ̃( η) ( η ) S S ̃( η)

ι X ∶= χ X , ψ ι X . ( ) ̃ι(X)( ) (̃( ))

Since ψ and the maps χ for am m<ω, σm, im m<ω ∈ Mβ are ((am)m<ω ,(σm,im)m<ω) (( ) ( ) ) injective, it follows that also ι is injective; which ﬁnishes our construction in the case that β + V = α . (S S ) η If SβS+ < α in V , we can take an injection ψ ∶ M ↪ β + V , and construct an injec- η β (S S ) + V + V β (ηm ,im)m<ω ηm tive function ι ∶ ´ κη ↪ κ ⋅ β in V G ↾ η+1 × ∏ G ↾ ̃( ) ( η ) (S S ) [( ( )) m<ω im κ , κ similarly as before. [ η ηm )] This gives the following proposition:

Proposition 42. If β + V = α , then there is in V Gβ ↾ η + 1 (ηm ,im)m<ω × (S S ) η [( ( )) ηm V ∏ G ↾ κη, κη an injection ι ∶ ´ κη ↪ β , where m<ω im [ m )] ̃( ) S S

´ κ ∶= ´ κ ∩ V G a × Gσm a , σ , i ∈ M . ̃ η η M ∗ m M i m m<ω m m m<ω β ( ) ( ) [ m ( ) m m ] U (( ) ( ) )

+ V β (ηm,im)m<ω ηm If β < αη, there is in V G ↾ η + 1 × ∏ G ↾ κη, κη an (S S ) [( ( )) m<ω im [ m )] injection ι ∶ ´ κ ↪ β + V . ̃( η) (S S )

β β This leads to our desired contradiction: We assumed that f ∶ dom f → αη was surjective. By Chapter 6.2 B), Deﬁnition 37 and Proposition 38, it follows β β (ηm ,im)m<ω ηm β that f ∈ V G ↾ η + 1 × ∏ G ↾ κη, κη ; where G ↾ η + [( ( )) m im [ m )] ( ( (ηm,im)m<ω ηm 1 × ∏ G ↾ κη, κη is a V -generic ﬁlter on the forcing notion )) m im [ m ) Pβ ↾ η + 1 (ηm,im)m<ω × P ηm ↾ κ , κ , which preserves cardinals ≥ α by (( ( )) ∏m [ η ηm ) η Chapter 6.2 C), Proposition 41. However, since dom f β ⊆ ´ κ and β V < α , it follows that f β together with ̃( η) S S η the map ι from Proposition 42 above, collapses the cardinal α in V Gβ ↾ η [( (ηm ,im)m<ω ηm η + 1 × ∏ G ↾ κη, κη . Contradiction. ( )) m im [ m )]

β β Thus, we have shown that our initial assumption that f ∶ dom f → αη was surjective, was wrong.

β Hence, there must be α < αη with α ∉ rg f .

E) We use an isomorphism argument and obtain a contradiction.

β We ﬁx an ordinal α < αη with α ∉ rg f . By surjectivity of f, there must be X ⊆ κη, X ∈ N, with f X = α. Hence, there is an η-good pair ̺ = a , σ , i ( ) (( m)m<ω ( m m)m<ω) σm β with X ∈ V ∏m G∗ am × ∏m G ; but since X ∉ dom f , there must be at least [ ( ) im ] one index m < ω with i ≥ β. Let S denote the set of all σ , i with i < β, m 0 ( m m) m and let S be the set of all σ , i with i ≥ β. Then S ≥ 1. 1 ( m m) m S 1S

69 For better clarity, we now switch to a slightly diﬀerent notation, and write λ , k ∶= ( m m) σ , i in the case that m ∈ S . We denote our η-good pair ̺ by ( m m) 1

̺ = a , σ , i , λ , k . ( m)m<ω ( m m)m∈S0 ( m m)m∈S1

Then σm λm ∏m G∗(am) × ∏m∈S G × ∏m∈S G X = X˙ 0 im 1 km

η for some X˙ ∈ Name P ω × ∏ P σm × ∏ P λm , such that the following (( ) m∈S0 m∈S1 ) holds:

• a m < ω is a sequence of pairwise disjoint κ -subsets, such that for all ( m S ) η m < ω and κ < κ , it follows that a ∩ κ , κ = 1, ν, η S m [ ν, ν,+1)S

• S ⊆ ω, and for all m ∈ S , we have σ ∈ Succ with σ ≤ η, i < min α , β , 0 0 m m m { σm }

• if m, m′ ∈ S with m ≠ m′, then σ , i ≠ σ ′ , i ′ , 0 ( m m) ( m m )

• ∅ ≠ S1 ⊆ ω, and for all m ∈ S1, we have λm ∈ Succ with λm ≤ η, km ∈ β, α , [ λm )

• if m, m′ ∈ S with m ≠ m′, then λ , k ≠ λ ′ , k ′ . 1 ( m m) ( m m )

Since X, α ∈ f, take p ∈ G with ( )

p ⊩ τ X˙ , α ∈ f.˙ s ( ̺( ) )

Since we are using countable support, we can asssume w.l.o.g. that σm ∈ supp p1, i ∈ dom p σ for all m ∈ S ; and λ ∈ suppp , k ∈ dom p λ for all m x 1( m) 0 m 1 m x 1( m) m ∈ S1.

+ The idea can roughly be explained as follows: Recall that we have β = β̃ + κη (addition of ordinals), where the ordinal β is large enough for A . In particular, ̃ ( f˙) + κη < β̃ < β < αη. We will now extend p and obtain a condition q ∈ G, q ≤ p, such that there is a sequence l m ∈ S with l ∈ β, β for all m ∈ S , such that ( m S 1) m (̃ ) 1 λm λm q = q for all m ∈ S1. Then we construct an isomorphism π ∈ A that swaps any km lm λ , k -coordinate with the according λ , l -coordinate. ( m m) ( m m) λm Then πq = q; and we will see that π ∈ ⋂m F ix ηm, im ∩ ⋂m H , since β is large ( ) km ̃ D D enough for A . Hence, πf π = f π ; so from q ⊩ τ X˙ , α ∈ f˙, we obtain that ( f˙) s ( ̺( ) ) Dπ D q ⊩ πτ X˙ , α ∈ f π . Setting s ( ̺( ) )

Dπ G Y ∶= πτ X˙ , ̺( ) it follows that Y, α ∈ f. ( )

70 σ λ G∗ a × G m × G m ∏m ( m) ∏m∈S0 ∏m∈S1 However, we will see that Y = X˙ im lm ; where im < β ′ for all m ∈ S0, but also lm < β for all m ∈ S1. But then, the η-good pair ̺ = a , σ , i , λ , l is an element of M , and it follows that (( m)m<ω (( m m)m∈S0 ( m m)m∈S1 )) β

σm λm ∏m G∗(am) × ∏m∈S G × ∏m∈S G β Y, α = X˙ 0 im 1 lm , α ∈ f . ( ) But this would be a contradiction towards α ∉ rg f β.

We start our proof with the following lemma:

Lemma 43. Let D be the set of all q ∈ P for which there exists a sequence of pairwise distinct ordinals l m ∈ S with l ∈ β, β ∖ i m ∈ S for all ( m S 1) m (̃ ) { m S 0} λm λm m ∈ S1, such that q = q holds for all m ∈ S1. Then D is dense below p. km lm Proof. Consider q ∈ P with q ≤ p. We have to construct q ≤ q with q ∈ D. The idea is that for every m ∈ S , we enlarge dom q λ by some suitable k , and set 1 x ( m) m q λ k ,ζ ∶= q λ l ,ζ = q λ l ,ζ for all ζ ∈ dom q λ = dom q λ . ( m)( m ) ( m)( m ) ( m)( m ) y ( m) y ( m)

Note that for every m ∈ S1, we have λm ∈ suppq1 with domx q1 λm < κ ≤ κη, S ( )S λm since λ ≤ η. Hence, it follows that dom q λ ≤ κ < κ+; and similarly, m S ⋃m∈S1 x ( m)S η η dom q σ ≤ κ < κ+. Thus, the set S ⋃m∈S0 x ( m)S η η ∆ ∶= β, β ∖ dom q λ ∪ dom q σ (̃ ) x ( m) x ( m) m∈S1 m∈S0

+ has cardinality κη .

Recall that for every m ∈ S , we have assumed that i ∈ dom p σ ⊆ dom q σ ; 0 m x ( m) x ( m) hence im ∉ ∆.

For m ∈ S1, we have km ∈ β, α ; hence, β < α and ∆ ⊆ β, β ⊆ α follows. [ λm ) λm (̃ ) λm We take a sequence of pairwise distinct ordinals l m ∈ S in ∆ (then l m ∈ ( m S 1) { m S S ⊆ β, β ∖ i m ∈ S ), and deﬁne the extension q ≤ q as follows: 1} (̃ ) { m S 0}

Set q0 ∶= q0, and supp q1 = supp q1. (From q ≤ p it follows that λm ∈ suppq1 for all m ∈ S .) For σ ∈ suppq with σ ∉ λ m ∈ S , we set q σ ∶= q σ . For 1 1 { m S 1} ( ) ( ) σ ∈ λ m ∈ S , we proceed as follows: Let S σ ∶= m ∈ S σ = λ . We set { m S 1} 1( ) { 1 S m} dom q σ ∶= dom q σ , and dom q σ ∶= dom q σ ∪ l m ∈ S σ . Note y ( ) y ( ) x ( ) x ( ) { m S 1( )} that by construction of ∆ this union is disjoint, since l ∉ dom q σ = dom q λ m x ( ) x ( m) for all m ∈ S σ . 1( ) Note that for every m ∈ S σ , we have k ∈ dom p σ ⊆ dom q σ ⊆ dom q σ . 1( ) m x ( ) x ( ) x ( ) We let q σ i, ζ ∶= q σ i, ζ whenever i, ζ ∈ dom q σ × dom q σ . If i, ζ ∈ ( )( ) ( )( ) ( ) x ( ) y ( ) ( ) dom q σ ∖ dom q σ , then ζ ∈ dom q σ and i = l for some n ∈ S σ , i.e. ( ) ( ) y ( ) n 1( ) n ∈ S with σ = λ . In this case, we set q σ i, ζ = q λ l ,ζ ∶= q λ k ,ζ = 1 n ( )( ) ( n)( n ) ( n)( n ) q σ k ,ζ . ( )( n )

71 λm λm This deﬁnes q ≤ q with the property that q = q holds for all m ∈ S1. km lm Thus, it follows that D is dense below p.

Since p ∈ G, we can now take q ∈ G, q ≤ p with q ∈ D. Take l m ∈ S as in ( m S 1) λm λm the deﬁnition of D, with lm ∈ β, β ∖ im m ∈ S0 and q = q for all m ∈ S1. (̃ ) { S } km lm Then the sets λ , l m ∈ S and σ , i m ∈ S are disjoint. {( m m) S 1} {( m m) S 0} Since q ≤ p, we have q ⊩ τ X˙ , α ∈ f.˙ s ( ̺( ) ) The next step is to construct an isomorphism π that swaps every λ , k - ( m m) coordinate with the according λ , l -coordinate for m ∈ S , and does nothing ( m m) 1 else.

Deﬁnition 44. We deﬁne an isomorphism π ∈ A as follows:

• P The map π0 is the identity on Dπ0 = 0.

• We set supp π ∶= supp q , and for every σ ∈ supp q , we let dom π σ ∶= 1 1 1 1( ) dom q σ . 1( ) Then for all m ∈ S1, it follows that λm ∈ supp p1 ⊆ suppq1 = supp π1; and k ∈ dom p λ ⊆ dom q λ = dom π λ , l ∈ dom q λ = m x 1( m) x 1( m) x 1( m) m x 1( m) dom π λ . x 1( m) • Consider σ ∈ supp π with κ = κ +. In the case that σ ∉ λ m ∈ S , we 1 σ σ { m S 1} set supp π σ ∶= ∅, and let π σ i, ζ ∶ 2 → 2 be the identity map for all 1( ) 1( )( ) i, ζ ∈ α × κ , κ . ( ) σ [ σ σ) • For σ ∈ λ m ∈ S , consider the set S σ ∶= m ∈ S σ = λ , and let { m S 1} 1( ) { 1 S m} supp π σ ∶= k m ∈ S σ ∪ l m ∈ S σ . Then supp π σ is a 1( ) { m S 1( )} { m S 1( )} 1( ) subset of dom π σ . x 1( )

The map f σ ∶ supp π σ → supp π σ is deﬁned as follows: Let f σ k = π1 ( ) 1( ) 1( ) π1 ( )( m) l , and f σ l = k for all m ∈ S σ . m π1 ( )( m) m 1( ) Then f σ is well-deﬁned and bijective, since k ≥ β for all m ∈ S , and π1 ( ) m 1 lm < β for all m ∈ S1.

It remains to deﬁne the maps π ζ ∶ 2supp π1(σ) → 2supp π1(σ) for ζ ∈ dom π σ : 1( ) y 1( ) Let π1 ζ ǫi i ∈ supp π1 σ ∶= ǫi i ∈ supp π1 σ , where ǫ ∶= ǫ , ( )( S ( )) (̃ S ( )) ̃km lm ǫ ∶= ǫ for all m ∈ S1 σ . ̃lm km ( )

Finally, for every i, ζ ∈ α × κ , κ , we let π σ i, ζ ∶ 2 → 2 be the ( ) σ [ σ σ) 1( )( ) identity.

72 This deﬁnes our automorphism π ∈ A.

Dπ G Lemma 45. For Y ∶= πτ X˙ , it follows that Y, α ∈ f. ̺( ) ( )

Proof. By construction of π it follows that whenever r is a condition in Dπ with ′ ′ λm λm r ∶= πr, then the following holds: Firstly, for all m ∈ S1, we have r = r ( )km lm ′ λm λm and r = r . Secondly, whenever σ ∈ supp r1, i ∈ domx r σ with σ, i ∉ ( )lm km ( ) ( ) λ , k m ∈ S ∪ λ , l m ∈ S , then r′ σ = rσ. {( m m) S 1} {( m m) S 1} ( )i i ′ σm′ σm′ ′ In particular, r = r holds for all m ∈ S0: ( )im′ im′ On the one hand, we have σ ′ , i ′ ∉ λ , k m ∈ S for all m′ ∈ S , since ( m m ) {( m m) S 1} 0 i ′ < β; but k ≥ β for all m ∈ S . On the other hand, σ ′ , i ′ ∉ λ , l m ∈ m m 1 ( m m ) {( m m) S S for all m′ ∈ S follows by construction of the set D. 1} 0 In other words: The map π swaps for all m ∈ S the λ , k -coordinate with the 1 ( m m) according λ , l -coordinate, and does nothing else. ( m m) λm λm Hence, it follows that πq = q; since q = q for all m ∈ S1. km lm

λm Dπ Dπ Next, we want to show that π ∈ ⋂m F ix ηm, im ∩ ⋂m H . Then πf = f ( ) km follows. Regarding π ∈ F ix η , i , it suﬃces to make sure that for all m < ω, ⋂m ( m m) we have η , i ∉ λ ′ , k ′ m′ ∈ S ∪ λ ′ , l ′ m′ ∈ S . But this follows ( m m) {( m m ) S 1} {( m m ) S 1} ′ from the fact that λm′ ≤ η and km′ ≥ β > β̃, lm′ > β̃ for all m ∈ S1; but β̃ is large enough for A , so for any η with η ≤ η, it follows that i < β. This implies ( f˙) m m m ̃ η , i ∉ λ ′ , k ′ m′ ∈ S ∪ λ ′ , l ′ m′ ∈ S for all m < ω as desired. ( m m) {( m m ) S 1} {( m m ) S 1} Hence, π ∈ F ix η , i . ⋂m ( m m) λ Regarding π ∈ H m , we have to make sure that whenever λ = λ ′ for some ⋂m km m m m < ω and m′ ∈ S , then supp π λ = supp π λ ′ ⊆ k , α holds; i.e. 1 1( m) 1( m ) ( m λm ) km′ > km and lm′ > km. Again, this follows from the fact that λm′ ≤ η and k ′ ≥ β > β, l ′ > β for all m′ ∈ S ; and β is large enough for A , so whenever m ̃ m ̃ 1 ̃ ( f˙) λ ≤ η, then k < β follows. Hence, π ∈ Hλm . m m ̃ ⋂m km D D Thus, it follows that πf π = f π .

Dπ D Now, from q ⊩ τ X˙ , α ∈ f˙, we obtain πq ⊩ πτ X˙ , α ∈ πf π ; hence, s ( ̺( ) ) s ( ̺( ) ) Dπ D q ⊩ πτ X˙ , α ∈ f π . With s ( ̺( ) )

Dπ G Y ∶= πτ X˙ , ̺( ) it follows from q ∈ G that Y, α ∈ f as desired. ( )

We will now show that Y, α ∈ f implies that also Y, α ∈ f β must hold. This ( ) ( ) ﬁnally gives our desired contradiction, since α ∉ rg f β.

73 Indeed, we will prove that

σm λm ∏m G∗(am) × ∏m∈S G × ∏m∈S G Y = X˙ 0 im 1 lm .

Since im < β for all m ∈ S0 and lm < β for all m ∈ S1, it follows that the η-good pair ̺′ ∶= a , σ , i , λ , l ( m)m<ω (( m m)m∈S0 ( m m)m∈S1 ) is an element of M . Hence, Y, α ∈ f would then imply that also Y, α ∈ f β β ( ) ( ) must hold, and we are done.

Recall that ̺ ∶= a , σ , i , λ , k , ( m)m<ω (( m m)m∈S0 ( m m)m∈S1 ) η ω σm λm X˙ ∈ Name P × ∏ P × ∏ P , and τ̺ X˙ is the canonical exten- (( ) m∈S0 m∈S1 ) ( ) sion of X˙ to a name for P (see Deﬁnition 34).

We will show recursively:

η Lemma 46. For every Y˙ ∈ Name P ω × ∏ P σm × ∏ P λm , it follows ( ) m∈S0 m∈S1 that Dπ Dπ π τ Y˙ = τ ′ Y˙ . ̺( ) ̺ ( ) η ˙ ∈ ω × σm × λm Proof. Consider Y Nameα+1 P ∏m∈S0 P ∏m∈S1 P , and assume (( ) η ) ω σm recursively that the claim was true for all Z˙ ∈ Nameα P × ∏ P × (( ) m∈S0 ∏ P λm . m∈S1 ) First,

Dπ Dπ τ Y˙ = τ Z˙ , r r ∈ D , Z˙ ∈ dom Y˙ , r ⊩ τ Z˙ ∈ τ Y˙ , ̺( ) ̺( ) T π s ̺( ) ̺( ) and

Dπ Dπ πτ Y˙ = πτ Z˙ , πr r ∈ D , Z˙ ∈ dom Y˙ , r ⊩ τ Z˙ ∈ τ Y˙ . ̺( ) ̺( ) T π s ̺( ) ̺( ) η ω σm Now, for any H a V -generic ﬁlter on P and Z˙0 ∈ Name P × ∏ P × (( ) m∈S0 ∏ P λm , it follows by construction of the map π that m∈S1 )

σm λm H ∏m<ω H∗(am) × ∏m∈S H × ∏m∈S H τ Z˙ = Z˙ 0 im 1 km ̺( 0) 0 σm λm ∏m<ω(πH)∗(am) × ∏m∈S (πH) × ∏m∈S (πH) ˙ 0 im 1 lm = Z0 πH = τ ′ Z˙ , ̺ ( 0) since π swaps any λ , k -coordinate with the according λ , l )-coordinate, ( m m) ( m m and does nothing else.

74 Hence, whenever r ∈ D , then r ⊩ τ Z˙ ∈ τ Y˙ if and only if πr ⊩ τ ′ Z˙ ∈ π s ̺( ) ̺( ) s ̺ ( ) τ ′ Y˙ . Thus, by our recursive assumption, ̺ ( )

Dπ Dπ πτ Y˙ = πτ Z˙ , πr πr ∈ D , Z˙ ∈ dom Y˙ , πr ⊩ τ ′ Z˙ ∈ τ ′ Y˙ ̺( ) ̺( ) T π s ̺ ( ) ̺ ( ) Dπ = τ ′ Z˙ , r r ∈ D , Z˙ ∈ dom Y˙ , r ⊩ τ ′ Z˙ ∈ τ ′ Y˙ ̺ ( ) T π s ̺ ( ) ̺ ( ) Dπ = τ ′ Y˙ . ̺ ( )

Hence, Dπ G Dπ G G Y = πτ X˙ = τ ′ X˙ = τ ′ X˙ = ̺( ) ̺ ( ) ̺ ( ) σm λm ∏m G∗(am) × ∏m∈S G × ∏m∈S G = X˙ 0 im 1 lm .

Hence, by Lemma 45 above, it follows that

σm λm ∏m G∗(am) × ∏m∈S G × ∏m∈S G X˙ 0 im 1 lm , α ∈ f. ( )

But im < β for all m ∈ S0 and lm < β for all m ∈ S1; hence,

σm λm ∏m G∗(am) × ∏m∈S G × ∏m∈S G β X˙ 0 im 1 lm , α ∈ f . ( ) But this contradicts our choice of α ∉ rg f β.

Thus, in either case our assumption of a surjective function f ∶´N κ → α in N ( η) η has lead to a contradiction, and it follows that indeed, θN κ ≤ α . ( η) η

+ Recall that we have assumed throughout our proof that κη+1 > κη . In the next + Chapter 6.3, we will treat the case that κη+1 = κη , and discuss where the arguments from Chapter 6.2 have to be modiﬁed.

+ Ð→ N 6.3 ∀ η κη+1 = κη θ (κη) ≤ αη .

+ If κη+1 = κη , we need the notion of an η-almost good pair (cf. Deﬁnition 26 and Proposition 27): For any X ∈ N, X ⊆ κη, there exists an η-almost good pair σm η+1 am m<ω, σm, im m<ω such that X ∈ V ∏m G∗ am × ∏m G × G . (( ) ( ) ) [ ( ) im ] Throughout this Chapter 6.3, we assume that

+ κη+1 = κη.

As before in Chapter 6.2, we assume towards a contradiction that there was a D D surjective function f ∶´N κ → α in N with πf π = f π for all π ∈ A with π ( η) η [ ] contained in the intersection

F ix η , i ∩ Hλm A . m m km f˙ m<ω ( ) m<ω ( )

75 We take β large enough for A as in Chapter 6.2, Deﬁnition 29, and set β ∶= β+κ+ ̃ ( f˙) ̃ η (addition of ordinals).

Now, we can adapt our deﬁnition of f β to η-almost good pairs, and obtain: f β ∶= X, α ∈ f ∃ a , σ , i η-almost good pair ∶ ∀m i < β ∧ ( ) T (( m)m<ω ( m m)m<ω) ( m )

σm η+1 η ω σ η+1 ∏m G∗(am) × ∏m G × G ∃ X˙ ∈ Name P × M P m × P X = X˙ im . ( ) m β β First, we assume towards a contradiction that f ∶ dom f → αη is surjective.

A) Constructing P̃β ↾ (η + 1).

As before, we only treat the case that

β < αη̃ or Lim ∩ (η,γ) ≠ ∅, where η ∶= sup σ < η σ ∈ Lim , i.e. we presume that there exist σ, i with ̃ { S } ( ) σ ∈ Lim and i ≥ β or σ > η.

This times, we construct a forcing notion Pβ ↾ η +1 instead of Pβ ↾ η +1 ; which ̃ ( ) ( ) should be like Pβ ↾ η + 1 , except that ﬁrstly, we use restrictions p ↾ κ2 instead ( ) ∗ η+1 2 η+1 of p∗ ↾ κη, and secondly, we include P .

Deﬁnition 47. For p ∈ P, let

pβ ↾ η + 1 ∶= p ↾ κ2 , pσ, aσ , pσ ↾ β × dom pσ ,pη+1, X , ̃ ( ) ∗ η+1 ( i i )σ≤η,i<β ( ( y ))σ≤η p and denote by Pβ ↾ η + 1 the collection of all pβ ↾ η + 1 such that p ∈ P (i.e. ̃ ( ) ̃ ( ) p ∈ P with σ, i ∈ supp p σ > η ∨ i ≥ β = ℵ ); together with the maximal S{( ) 0 S }S 0 1β ≤β element ̃η+1. The order relation ̃η+1 is defined as in Definition 31. Like in Chapter 6.2 A), one can write down a projection of forcing posets ρβ ∶ ̃η+1 P → Pβ ↾ η + 1 and conclude that ̃ ( ) Gβ ↾ η + 1 ∶= p ∈ Pβ ↾ η + 1 ∃ q ∈ G ∩ P qβ ↾ η + 1 ≤β p ̃ ( ) ̃ ( ) T ̃ ( ) ̃η+1 is a V -generic filter on Pβ ↾ η + 1 . ̃ ( )

B) Capturing f β.

We deﬁne a forcing notion Pβ ↾ η + 1 (ηm ,im)m<ω , which will be obtained from (̃ ( )) Pβ ↾ η + 1 by using X instead of X (cf. Chapter 6.2 B) ), and including for ̃ ( ) ̃p p η ∈ Lim, η > η the verticals pηm ↾ κ , and also aηm ∩ κ , the according m m im η+1 im η+1 linking ordinals up to κη+1.

The restriction pβ ↾ η + 1 (ηm ,im)m<ω for p ∈ P is deﬁned as follows: (̃ ( ))

β (ηm ,im)m<ω 2 σ σ ηm ηm p ↾ η+1 ∶= p∗ ↾ κ , p , a σ η,i β , p ↾ κη+1, a ∩ κη+1 m<ω , η >η, (̃ ( )) η+1 ( i i ) ≤ < ( im im ) m

76 pσ ↾ β × dom pσ , X , pη+1 . ( ( y ))σ≤η ̃p Roughly speaking, the diﬀerence with the restrictions pβ ↾ η+1 (ηm ,im)m<ω intro- ( ( )) + duced in Chapter 6.2 B) is, that we are now reaching up to κη+1 = κη instead of κη.

We denote by Pβ ↾ η + 1 (ηm,im)m<ω the collection of all pβ ↾ η + 1 (ηm ,im)m<ω (̃ ( )) (̃ ( )) for p ∈ P together with the maximal element 1β (ηm,im)m<ω . The order relation (̃η+1) “≤” is deﬁned like in Deﬁnition 35.

Finally, we include the verticals P ηm ↾ κ , κ for η > η + 1, which gives the [ η+1 ηm ) m product Pβ + (ηm ,im)m<ω × ηm ̃ ↾ η 1 M P ↾ κη+1, κηm . ( ( )) m<ω [ ) Let Gβ ↾ η + 1 (ηm ,im)m<ω × Gηm ↾ κ , κ ̃ M im η+1 ηm ( ( )) m<ω [ ) denote the collection of all

β (ηm ,im)m<ω ηm p ↾ η + 1 , p ↾ κη+1, κη m<ω (̃ ( )) ( im [ m ) ) such that there exists q ∈ G ∩ P with qβ ↾ η + 1 ≤ pβ ↾ η + (̃ ( ))(ηm ,im)m<ω (̃ ( (ηm,im)m<ω ηm ηm 1 and q ↾ κη+1, κη ⊇ p ↾ κη+1, κη for all m < ω. )) im [ m ) im [ m ) As in Proposition 36, one can construct a projection of forcing posets

β (ηm,im)m<ω ∶ P Pβ + (ηm,im)m<ω × ηm ρ → ̃ ↾ η 1 M P ↾ κη+1, κηm , (̃ ) ( ( )) m<ω [ )

β (ηm ,im)m<ω ηm and it follows that G ↾ η+1 × ∏ G ↾ κη+1, κη is a V -generic ( ̃ ( )) m<ω im [ m ) ﬁlter on Pβ ↾ η + 1 (ηm,im)m<ω × P ηm ↾ κ , κ . (̃ ( )) ∏m<ω [ η+1 ηm ) Like in Chapter 6.2 B), we want to deﬁne a map f β ′ contained in V Gβ ↾ ( ) [ ̃ (ηm ,im)m<ω ηm η + 1 × ∏ G ↾ κη+1, κη , and then use an isomorphism ar- ( )) m<ω im [ m )] gument to show that f β = f β ′. ( )

Before that, we have to modify our transformations of names τ̺ (where ̺ is an η-good pair), and deﬁne transformations τ (where ̺ is an η-almost good pair) with ̺̃

η+1 ω σm η+1 τ̺ ∶ Name P × M P × P → Name P ̃ (( ) m<ω ) ( ) as follows (cf. Deﬁnition 34):

Deﬁnition 48. For an η-almost good pair ̺ = am m<ω, σm, im m<ω , deﬁne η+1 ( ) ( ) recursively for Y˙ ∈ Name P ω × P σm × P η+1 : (( ) ∏m<ω )

σm η+1 τ̺ Y˙ ∶= τ̺ Z˙ ,q q ∈ P , ∃ Z˙ , p∗ am m<ω , p m<ω ,p ∈ Y˙ ∶ ̃ ( ) (̃ ( ) ) T ( ( )) ( im )

σm σm η+1 η+1 ∀m q∗ am ⊇ p∗ am , q ⊇ p , q ⊇ p . ( ) ( ) im im

77 σm η+1 ∏m<ω G∗(am) × ∏m<ω G × G G η+1 ω Then Y˙ im = τ Y˙ holds for all Y˙ ∈ Name P × (̺̃( )) (( ) P σm × P η+1 . ∏m<ω )

Deﬁnition 49. Let f β ′ denote the set of all X, α for which there exists an ( ) ( ) η-almost good pair ̺ = a , σ , i with i < β for all m < ω, such (( m)m<ω ( m m)m<ω) m that σm η+1 ∏m G∗(am) × ∏m G × G X = X˙ im , and there is a condition p ∈ P with the following properties:

• σ, i ∈ supp p σ > η or i ≥ β = ℵ , S{( ) 0 S }S 0 • p ⊩ τ X˙ , α ∈ f˙, s ̺̃( )

β (ηm,im)m<ω ηm β (ηm,im)m<ω • p ↾ η+1 , p ↾ κη+1, κη m<ω ∈ G ↾ η+1 × (̃ ( )) ( im [ m )) ( ̃ ( )) ηm ∏ G ↾ κη+1, κη , m<ω im [ m )

ηm ηm • ∀ ηm ∈ Lim ∶ ηm, im ∈ supp p0 with a = g . ( ) im im

β ′ β (ηm,im)m<ω ηm Then f ∈ V G ↾ η + 1 × ∏ G ↾ κη+1, κη . ( ) [( ̃ ( )) m<ω im [ m )] Proposition 50. f β = f β ′. ( ) Proof. We briefly outline, where the isomorphism argument form Proposition 38 σm η+1 β ′ β ∏m G∗(am) × ∏m G × G has to be modified. We start with X, α ∈ f ∖f , X = X˙ im , ( ) ( ) for an η-almost good pair ̺ = a , σ , i . Take p as in the defini- (( m)m<ω (( m m)m<ω)) tion of f β ′ with p ⊩ τ X˙ , α ∈ f˙, and p′ ∈ G with p′ ⊩ τ X˙ , α ∉ f˙. ( ) s (̺̃( ) ) s (̺̃( ) ) The first step is the construction of extensions p ≤ p, p′ ≤ p′, such that p and p′ have “the same shape”, agree on Pβ ↾ η +1 ; and pηm = p′ ηm holds for all m < ω, ̃ ( ) im ( )im ηm ′ ηm and a = a holds for all m < ω with ηm ∈ Lim. im ( )im We proceed as in Proposition 38, with the following modifications:

• ′ The construction of p∗, p∗ that we used in the Proposition 38 for intervals κ , κ ⊆ κ , has to be applied to all intervals κ , κ ⊆ κ now, [ ν,j ν,j+1) η [ ν,j ν,j+1) η+1 ′ 2 since we need p∗ and p∗ agree on κη+1.

• Analogously, the construction of pσ, p′ σ for σ ∈ Lim, i < α for intervals i ( )i σ κ , κ ⊆ κ , has to be applied to all intervals κ , κ ⊆ κ now, [ ν,j ν,j+1) η [ ν,j ν,j+1) η+1 in the case that σ > η + 1.

• Additionally, we have to make sure that pη+1 = p′ η+1. ( ) D The next step is the construction of an isomorphism π such that πp = p′, πf π = D Dπ Dπ f π , and π τ X˙ = τ X˙ . Again, we take for π a standard isomorphism for ̺̃( ) ̺̃( ) πp = p′; but this time, we set G ν, j ∶= F ν, j for all intervals κ , κ ⊆ π0 ( ) π0 ( ) [ ν,j ν,j+1)

78 κ (instead of only intervals κ , κ ⊆ κ ), and G ν, j = id for all κ ≥ η+1 [ ν,j ν,j+1) η π0 ( ) ν,j κ (instead of all κ ≥ κ ). Then as before, it follows that π ∈ F ix η , i ∩ η+1 ν,j η ⋂m ( m m) Hλm . ⋂m km Dπ Dπ For verifying π τ X˙ = τ X˙ , we now additionally have to make sure that ̺̃( ) ̺̃( ) π is the identity on P η+1. But since we have arranged pη+1 = p′ η+1, this is clear ( ) by construction of π1.

Dπ D Now, it follows from p ⊩ τ X˙ , α ∈ f˙ that πp ⊩ π τ X˙ , α ∈ πf π . Hence, s ̺̃( ) s ̺̃( ) Dπ D p′ ⊩ τ X˙ , α ∈ f π , which is a contradiction towards p′ ⊩ τ X˙ , α ∉ s ̺̃( ) s ̺̃( ) f˙.

β β ′ β (ηm,im)m<ω ηm Thus, f = f ∈ V G ↾ η + 1 × ∏ G ↾ κη+1, κη as desired. ( ) [( ̃ ( )) m im [ m )]

P̃β (ηm,im)m<ω ηm C) ( ↾ (η + 1)) × ∏m P ↾ [κη+1,κηm ) preserves cardinals ≥ αη.

Now, we will show that cardinals ≥ α are absolute between V and V Gβ ↾ η [( ̃ (ηm ,im)m<ω ηm η + 1 × ∏ G ↾ κη+1, κη . ( )) m<ω im [ m )] As in Chapter 6.2 C), we are using that GCH holds in our ground model V , and when we argue that a particular forcing notion preserves cardinals, we mean that it preserves cardinals under GCH, if not stated diﬀerently.

Lemma 51. If β + < α , then Pβ ↾ η + 1 (ηm,im)m<ω × P ηm ↾ κ , κ S S η (̃ ( )) ∏m [ η+1 ηm ) preserves cardinals ≥ αη.

Proof. We closely follow the proof of Lemma 39 and Corollary 40.

The forcing notion Pβ ↾ η + 1 (ηm ,im)m<ω is the set of all (̃ ( ))

2 σ σ ηm ηm p∗ ↾ κ , p , a σ η,i β , p ↾ κη+1, a ∩ κη+1 m<ω,η >η, η+1 ( i i ) ≤ < ( im im ) m

pσ ↾ β × dom pσ , X , pη+1 , ( ( y ))σ≤η ̃p where p ∈ P with σ, i ∈ suppp σ > η ∨ i ≥ β = ℵ . S{( ) 0 S }S 0 + 2 ηm ηm Since κη+1 = κ , it follows that the p∗ ↾ κ , as well as p ↾ κη+1, a ∩ κη+1 for η η+1 ( im im ) m < ω are bounded below κ ; which gives only ≤ κ ⋅ 2κη ω = κ = κ+ ≤ β - η+1 ( η+1 ) η+1 η S S many possibilities. Since X ⊆ κ , there are only ≤ κ+ ≤ β -many possibilities for X , as well. Regard- ̃p η η S S ̃p ing pσ, aσ and pσ ↾ dom pσ × β , it follows as in Lemma 39 that ( i i )σ≤η,i<β ( ( x ))σ≤η there are only ≤ β + ⋅ κ+ = β +-many possibilities. S S η S S ′ We denote by Pβ ↾ η + 1 (ηm ,im)m<ω the forcing notion that is obtained from (̃ ( )) Pβ (ηm,im)m<ω η+1 Pβ (ηm ,im)m<ω ̃ ↾ η + 1 by excluding P . Then ̃ ↾ η + 1 is iso- ( ( )) ′ ( ( )) morphic to the product Pβ ↾ η + 1 (ηm,im)m<ω × P η+1. (̃ ( ))

79 Pβ By what we have just argued, it follows that the forcing notion ̃ ↾ η + ′ ( ( 1 (ηm ,im)m<ω has cardinality ≤ β +; and the remaining product P η+1 × P ηm ↾ )) S S ∏m<ω κη+1, κηm preserves all cardinals by similar arguments as in Proposition 41. [ ) ′ Hence, it follows that Pβ ↾ η +1 (ηm ,im)m<ω × P η+1 × P ηm ↾ κ , κ (̃ ( )) ∏m<ω [ η+1 ηm ) preserves all cardinals ≥ β ++. S S

Proposition 52. The forcing Pβ ↾ η + 1 (ηm ,im)m<ω × P ηm ↾ κ , κ (̃ ( )) ∏m [ η+1 ηm ) preserves cardinals ≥ αη.

Proof. We only have to treat the case that α = β +. Then cf β > ω, and GCH η S S gives β ℵ0 = β . The proof is similar as for Proposition 41: We distinguish several S S S S ′′ cases, and construct Pβ ↾ η + 1 (ηm,im)m<ω from Pβ ↾ η + 1 (ηm ,im)m<ω by (̃ ( )) (̃ ( )) splitting up P η+1, and also one or two factors P σ ↾ β × κ , κ for σ ∈ Succ ( [ σ σ)) with σ = η, or σ = η in the case that η is a successor ordinal with η = η + 1. Then ′′ as in the proof of Proposition 41, it follows that Pβ ↾ η + 1 (ηm ,im)m<ω has (̃ ( )) cardinality ≤ β < α , and the product of the remaining P σ ↾ β × κ , κ , P η+1 S S η ( [ σ σ)) and P ηm ↾ κ , κ preserves all cardinals. ∏m [ η+1 ηm )

̃ β ̃ ℵ0 D) A set ´(κη) ⊇ dom f with an injection ι ∶ ´(κη) ↪ SβS .

For an η-almost good pair ̺ = a , σ , i , it follows that G a × (( m)m<ω ( m m)m<ω) ∏m ∗( m) η η ∏ G m × Gη+1 is a V -generic ﬁlter on ∏ P ω × ∏ P σm × P η+1, and m im m( ) m

σm η+1 α ∏m G∗(am) × ∏m G × G + V 2 im = α ( )

η+1 holds for all α ≤ κη by the same proof as for Lemma 2, since P ist ≤ κη-closed. + V σm η+1 Thus, there is an injection χ ∶´ κη ↪ κη in V ∏m G∗ am × ∏m G × G . ( ) ( ) [ ( ) im ] Let M denote the set of all η-almost good pairs ̺ = a , σ , i Èβ (( m)m<ω ( m m)m<ω) in V with the property that im < β for all m < ω. Then MÈβ has cardinality ≤ κ ⋅ η ℵ0 ⋅ β ℵ0 = β ℵ0 . η+1 S S S S S S Moreover, dom f β is a subset of ´ κ ∶= ̃( η)

´ κ ∩ V G a × Gσm × Gη+1 a , σ , i ∈ M . η M ∗ m M i m m<ω m m m<ω Èβ ( ) [ m ( ) m m ] U ( ) ( )

Now, we can proceed as in Chapter 6.2 D) and construct in V Gβ ↾ η+1 (ηm ,im)m<ω × [( ̃ ( )) ηm V + V ∏ G ↾ κη+1, κη an injection ι ∶ ´ κη ↪ β in the case that αη = β , m im [ m )] ̃( ) S S (S S ) and an injection ι ∶ ´ κ ↪ β + V in the case that α > β + V . Together with ̃( η) (S S ) η (S S ) Chapter 6.3 B) and 6.3 C), this gives the desired contradiction.

β β Thus, we have shown that the map f ∶ dom f → αη must not be surjective.

80 E) We use an isomorphism argument and obtain a contradiction.

+ The arguments for this part are the very same as in the case that κη+1 > κη , except that we are now working with η-almost good pairs ̺ = a , σ , i in- (( m)m<ω ( m m)m<ω) stead of η-good pairs.

Thus, also in the case that κ = κ+, it follows that θN κ = α . η+1 η ( η) η

6.4 The remaining cardinals in the “gaps” λ ∈ (κη,κη+1).

So far, we have shown that θN κ = α holds for all 0 < η < γ. Recall that in ( η) η the very beginning (see Chapter 2), we started by “thinning out” our sequence κ 0 < η < γ and assuming w.l.o.g. that α 0 < η < γ is strictly increasing. ( η S ) ( η S ) Thus, it remains make sure that for all cardinals λ ∈ κ , κ in the “gaps” , ( η η+1) θN λ gets the smallest possible value, i.e. θN λ = max α , λ++ . This will be ( ) ( ) { η } our aim for this Chapter 6.4.

After that, in Chapter 6.5, we make sure that also for all cardinals λ ≥ κγ , the value θN λ will be the smallest possible. ( )

We consider a cardinal λ in a “gap” λ ∈ κ , κ (then κ > κ+), and set ( η η+1) η+1 η α λ ∶= max λ++, α . Then θN λ ≥ α λ is clear, and it remains to make sure ( ) { η} ( ) ( ) that there is no surjective function f ∶Ń λ → α λ in N. ( ) ( ) First, we want to describe the intermediate generic extensions where the λ-subsets X ∈ Ń λ are located. ( ) Let λ ∈ κ , κ , where  < cf κ in the case that η + 1 ∈ Lim, and  = 0 with [ η, η,+1) η+1 λ ∈ κ , κ = κ , κ in the case that η + 1 ∈ Succ. ( η,0 η,1) ( η η+1) We will modify our definition of an η-good pair and obtain the notion of an η-good pair for λ, which will be used to describe the intermediate generic extensions where the sets X ∈ Ń λ are located: ( ) Definition 53. We say that a , σ , i is an η-good pair for λ, if ( m)m<ω ( m m)m<ω the following hold:

• a m < ω is a sequence of pairwise disjoint subsets of κ , such that for ( m S ) η, all κ < κ , it follows that a ∩ κ , κ = 1, ν,l η, S m [ ν,l ν,l+1)S • for all m < ω, we have σm ∈ Succ with σm ≤ η, and im < ασm ,

• if m ≠ m′, then σ , i ≠ σ ′ , i ′ . ( m m) ( m m ) Similarly as in Proposition 25, we obtain:

Proposition 54. For every X ∈ N, X ⊆ λ, there is an η-good pair for λ, denoted σm by ̺ = am m<ω, σm, im m<ω , such that X ∈ V ∏m<ω G∗ am × ∏m<ω G . ( ) ( ) [ ( ) im ]

81 Proof. As in Proposition 25, it follows by the Approximation Lemma 21 that any X ∈ N, X ⊆ λ is contained in a generic extension

σm σm X ∈ V G∗ g × G , M im M i [m<ω ( ) m<ω m ] where σ , i m < ω and σ , i m < ω are sequences of pairwise distinct (( m m) S ) (( m m) S ) pairs with σm ∈ Lim, im < ασm , and σm ∈ Succ, im < ασm for all m < ω.

σm σm The forcing ∏m<ω P × ∏m<ω P can be factored as

P σm ↾ κ × P σm × P σm ↾ κ , κ × P σm . M η, M M [ η, ηm ) M m<ω σm≤η m<ω σm>η In the case that λ ∈ κ , κ , it follows that the “lower part” has cardinality ( η, η,+1) + ≤ κη, ≤ λ, and the “upper part” is ≤ λ-closed. + + If λ = κη,, then ﬁrstly, the “lower part” has cardinality ≤ κη, = λ , and secondly, ++ it follows that  > 0 and κη+1 ∈ Lim, so κη,+1 ≥ κη, by construction. Hence, the “upper part” is ≤ λ+-closed. In either case, we obtain

σm σm X ∈ V M G∗ gi ∩ κη, × M G . [ ( m ) im ] m<ω σm≤η With a ∶= gσm ∩ κ for all m < ω, it follows by the independence property that m im η, a , σ , i is an η-good pair for λ with ( m)m<ω ( m m)m<ω

σm X ∈ V M G∗ am × M G . [ ( ) im ] m<ω σm≤η

η+1 × σm (Note that ∏m G∗ am ∏σm≤η G is a V -generic ﬁlter on the forcing P ↾ ( ) im ( ω σm κη, × ∏ P ). ) σm≤η

As before, we assume towards a contradiction that there was a surjective function D D f ∶´N λ → α λ in N, where πf π = f π holds for all π ∈ A with π contained ( ) ( ) [ ] in the intersection

F ix η , i ∩ Hλm A . m m km f˙ m<ω ( ) m<ω ( ) We take β large enough for the intersection A as in Chapter 6.2, Deﬁnition 29, ̃ ( f˙) + and set β ∶= β̃+ κη (addition of ordinals).

Let f β ∶= X, α ∈ f ∃ a , σ , i η-good pair for λ ∶ ∀m i < β ∧ ( ) T ( m)m<ω ( m m)m<ω ( m )

σm η+1 ω σ ∏m G∗(am) × ∏m G ∃ X˙ ∈ Name P ↾ κ × P m X = X˙ im . ( η,) M σm≤η First, we assume towards a contradiction that f β ∶ dom f β → α(λ) is surjective.

82 A) Constructing P̃β ↾ (η + 1) P 2 We proceed as in Chapter 6.2 A) and 6.3 A), except that we have to use ∗ ↾ κη, P 2 η+1 instead of ∗ ↾ κη, and do not include P : For p ∈ P, we set pβ ↾ η + 1 ∶= p ↾ κ2 , pσ, aσ , pσ ↾ β × dom pσ , X , ̃ ( ) ∗ η, ( i i )σ≤η,i<β ( ( y ))σ≤η p and denote by Pβ ↾ η + 1 the collection of all pβ ↾ η + 1 , where p ∈ P (i.e. p ∈ P ̃ ( ) ̃ ( ) with σ, i ∈ suppp0 σ > η ∨ i ≥ β = ℵ0 ); together with the maximal element S{( ) S β }S 1β ≤ ̃η+1, and the order relation ̃ η+1 defined similarly as in Definition 31. β Pβ We denote by G̃ ↾ η + 1 the set of all p ∈ ̃ ↾ η + 1 such that there exists ( ) β ( ) q ∈ G ∩ P with q β ↾ η + 1 ≤ p. Then as in Chapter 6.2 A), Proposition 33, it ̃ ( ) ̃ η+1 follows that Gβ ↾ η + 1 is a V -generic filter on Pβ ↾ η + 1 . ̃ ( ) ̃ ( )

B) Capturing f β

For p ∈ P, the restriction pβ ↾ η + 1 (ηm ,im)m<ω is defined as follows: (̃ ( )) β (ηm ,im)m<ω 2 σ σ ηm ηm p ↾ η+1 ∶= p∗ ↾ κ , p , a σ η,i β , p ↾ κη,, a ∩ κη, m<ω , η >η , (̃ ( )) η, ( i i ) ≤ < ( im im ) m pσ ↾ β × dom pσ , X . ( ( y ))σ≤η ̃p We define Pβ ↾ η + 1 (ηm ,im)m<ω and Gβ ↾ η + 1 (ηm,im)m<ω as in Chapter 6.2 (̃ ( )) ( ̃ ( )) B) and 6.3 B). Then Gβ ↾ η + 1 (ηm ,im)m<ω × Gηm ↾ κ , κ ̃ M im η, ηm ( ( )) m<ω [ ) is a V -generic filter on

Pβ + (ηm ,im)m<ω × ηm ̃ ↾ η 1 M P ↾ κη,, κηm . ( ( )) m<ω [ ) The construction of f β ′ as well as the proof of f β = f β ′ are as in Chapter 6.2 ( ) ( ) B) and 6.3 B); except that this time, the isomorphism π from the proof of Propo- P 2 P 2 sition 38 has to be the identity on ∗ ↾ κη, (not only on ∗ ↾ κη). This can be ′ achieved by the following modiﬁcations: Firstly, we demand that p∗ and p∗ cohere P 2 P 2 2 ′ 2 on ∗ ↾ κη, (not only ∗ ↾ κη); secondly, we arrange p∗ ↾ κη, = p∗ ↾ κη, (instead of 2 ′ 2 just p∗ ↾ κη = p∗ ↾ κη); and thirdly, when constructing the isomorphism π, we set G ν, j ∶= F ν, j for all κ < κ now, and G ν, j = id whenever κ ≥ κ . π0 ( ) π0 ( ) ν,j η, π0 ( ) ν,j η,

β β ′ β (ηm,im)m<ω ηm It follows that f = f ∈ V G̃ ↾ η + 1 × ∏ G ↾ κη,, κη . ( ) [( ̃ ( )) m im [ m )]

P̃β (ηm,im)m<ω ηm C) ( ↾ (η + 1)) × ∏m<ω P ↾ [κη,,κηm ) preserves cardinals ++ ≥ α(λ) = max{λ , αη}.

The arguments here are similar as in Chapter 6.2 C) and 6.3 C), since there are κ ℵ0 + + 2 ηm only ≤ 2 η, = κ ≤ λ < α λ -many possibilities for p∗ ↾ κ and p ↾ ( ) η, ( ) η, ( im ηm κη,, a ∩ κη, m<ω. im )

83 D) A set ´̃(λ) ⊇ dom f β with an injection ι ∶ ´̃(λ) ↪ λ+ ⋅ SβSℵ0 .

We proceed as in Chapter 6.2 D) and 6.3 D). Whenever a , σ , i (( m)m<ω ( m m)m<ω) σm is an η-good pair for λ, it follows that ∏m G∗ am × ∏m G is a V -generic ﬁlter ( ) im η+1 on P ↾ κ ω × P σm ; and ( η,) ∏m

σm α V [∏m G∗(am) × ∏m G + V 2 im = α ] ( ) holds for all cardinals α by the same proof as in Lemma 2. Hence, it follows that σm + V in V ∏m G∗ am × ∏m G , there is an injection χ ∶´ λ ↪ λ . [ ( ) im ] ( ) ( ) Let M be the set of all ̺ = a , σ , i in V such that ̺ is an η-good Èβ (( m)m<ω ( m m)m<ω) pair for λ with the property that im < β for all m < ω. Then MÈβ has cardinality ≤ κ+ ℵ0 ⋅ β ℵ0 ≤ λ+ ⋅ β ℵ0 . ( η,) S S S S By construction, it follows that dom f β is a subset of

´ λ ∶= ´ λ ∩ V G a × Gσm a , σ , i ∈ M . ̃ M ∗ m M i m m<ω m m m<ω Èβ ( ) ( ) [ m ( ) m m ] T (( ) ( ) )

η As in Chapter 6.2 D), we can now work in V G̃β ↾ η + 1 (ηm,im)m<ω × ∏ G m ↾ [( ̃ ( )) m im κ , κ and construct there an injection ι ∶ ´ λ ↪ λ+ V ⋅ β V in the case [ η, ηm )] ̃( ) ( ) S S that α = β + V , and an injection ι ∶ ´ λ ↪ λ+ V ⋅ β + V in the case that η (S S ) ̃( ) ( ) (S S ) α > β + V . η (S S ) Together with Chapter 6.4 B) and 6.4 C), this gives the desired contradiction.

Hence, it follows that there must be α < α λ with α ∉ rg f β. ( )

E) We use an isomorphism argument and obtain a contradiction.

The arguments for this part are the same as in Chapter 6.2 E); except that we are working with η-good pairs for λ now (instead of η-good pairs).

Thus, we have shown that for all cardinals λ ∈ κ , κ in a “gap” , the value ( η η+1) θN λ is the smallest possible: θN λ = α λ = max λ++, α . ( ) ( ) ( ) { η}

It remains to consider the cardinals λ ≥ κ ∶= sup κ 0 < η < γ . We prove that γ { η S } for all λ ≥ κ , again, θN λ takes the smallest possible value. γ ( ) This will be the aim of the next Chapter 6.5.

6.5 The cardinals λ ≥ κγ ∶= sup{κη S 0 < η < γ}. Let α ∶= sup α 0 < η < γ , and consider a cardinal λ ≥ κ . We want to show γ { η S } γ that θN λ takes the smallest possible value α λ , deﬁned as follows: ( ) ( )

• In the case that cf α = ω, we set α λ = max α++, λ++ . γ ( ) { γ }

84 • In the case that α = α+ for some α with cf α = ω, we set α λ = max α+, λ++ . γ ( ) { γ }

• In other cases, we set α λ ∶= max α , λ++ . ( ) { γ }

Then by our remarks from Chapter 2, it follows that indeed, θN λ ≥ α λ holds ( ) ( ) for all λ ≥ κγ .

First, we assume that

α(λ) > αγ.

It remains to prove that there is no surjective function f ∶´N λ → α λ in N. ( ) ( ) We start with the following observation (again, we use that V ⊧ GCH):

Lemma 55. Let λ ≥ κ with α λ > α . Then P preserves cardinals ≥ α λ . γ ( ) γ ( )

Proof. For every p ∈ P, p = p , pσ, aσ , pσ , there are ( ∗ ( i i )σ<γ,i<ασ ( )σ<γ )

• + ≤ κγ -many possibilities for p∗,

• ≤ αℵ0 -many possibilities for the countable support of pσ, aσ , γ ( i i )σ<γ,i<ασ

• ≤ κ+-many possibilities for pσ, aσ when the support is given. γ ( i i )σ<γ,i<ασ

In the case that γ is a limit ordinal, it follows by the strict monotonicity of the sequence α 0 < σ < γ that α < α holds for all 0 < σ < γ. Hence, for ( σ S ) σ γ σ + any σ ∈ Succ, the forcing notion P has cardinality ≤ ασ ≤ αγ ; and it follows by countable support that we have ≤ γ ℵ0 ⋅ αℵ0 = αℵ0 -many possibilities for pσ . S S γ γ ( )σ<γ P + ℵ0 + ℵ0 Hence, the forcing has cardinality ≤ κγ ⋅ αγ ≤ λ ⋅ αγ . If cf αγ > ω, GCH gives P ≤ λ+ ⋅ α , and α λ = max λ++, α+ . Hence, P preserves cardinals ≥ α λ as S S γ ( ) { γ } ( ) desired. If cf α = ω, then α λ = max λ++, α++ ≥ P +; and again, it follows that γ ( ) { γ } S S P preserves cardinals ≥ α λ . ( ) It remains to consider the case that γ = γ + 1 is a successor ordinal. Then our sequences κ 0 < σ < γ = κ 0 < σ ≤ γ and α 0 < σ < γ = α 0 < σ ≤ γ ( σ S ) ( σ S ) ( σ S ) ( σ S ) have a maximal element, and κγ = κγ , αγ = αγ.

If γ ∈ Lim, i.e. κγ is a limit cardinal, it follows that for any σ ∈ Succ, we have σ < γ; hence, α+ ≤ α = α . This gives P ≤ κ+ ⋅ αℵ0 ≤ λ+ ⋅ αℵ0 as before, and σ γ γ S S γ γ γ α λ ≥ P +. ( ) S S γ If γ ∈ Succ, i.e. κγ is a successor cardinal, then P has to be treated separately. We factor P ≅ P′ × P γ with P′ ∶= p , pσ, aσ , pσ p ∈ P . Then P γ { ( ∗ ( i i )σ≤γ,i<ασ ( )σ<γ ) S } preserves cardinals, and P′ has cardinality ≤ λ+ V ⋅ αℵ0 V as before (in V , and ( ) ( γ ) hence, also in any P γ-generic extension); where α λ ≥ P′ +. Hence, the forcing ( ) S S P ≅ P′ × P γ preserves cardinals ≥ α λ as desired. ( )

85 Now, we assume towards a contradiction that there was a surjective function f ∶Ń λ → α λ in N. ( ) ( ) By the Approximation Lemma 21, it follows that any X ∈ N, X ⊆ λ, is contained in σm an intermediate generic extension V ∏ G , with a sequence σm, im m < [ m<ω im ] (( ) S ω of pairwise distinct pairs in V such that 0 < σ < γ, i < α for all m < ω. ) m m σm Denote by M the collection of these σ , i m < ω . Then M ≤ αℵ0 in V ; and (( m m) S ) S S γ αℵ0 < α λ as argued before. γ ( ) σm The product ∏m P preserves cardinals and the GCH. Hence, it follows that in any generic extension V ∏ Gσm , there is an injection χ ∶ ´ λ ↪ λ+ V . [ m im ] ( ) ( ) Now, one can argue as in Chapter 6.2 D), and define in V G a set ´ λ ⊇ Ń λ [ ] ̃( ) ( ) with an injection ι ∶ ´ λ ↪ λ+ V ⋅ α , or ι ∶ ´ λ ↪ λ+ V ⋅ α+ V in the case ̃( ) ( ) γ ̃( ) ( ) ( γ ) that α λ ≥ α++ V . Together with Lemma 55, this gives the desired contradiction. ( ) ( γ ) Thus, we have shown that in the case that α λ > α , there can not be a surjective ( ) γ function f ∶´ λ → α λ in N. ( ) ( ) It remains to consider the case that

α(λ) = αγ.

+ + Then λ < αγ , cf αγ > ω; and if αγ = α for some α, then cf α > ω.

Assume towards a contradiction that there was a surjective function f ∶´N λ → D D ( ) α λ in N, f = f˙G with πf π = f π for all π ∈ A with π contained in the ( ) [ ] intersection

F ix η , i ∩ Hλm A . m m km f˙ m<ω ( ) m<ω ( ) Similary as before, we take β < α λ large enough for the intersection A , i.e. ̃ ( ) ( f˙) β > λ+ with β > sup i m < ω ∪ sup k m < ω (this is possible, since ̃ ̃ { m S } { m S } cf α λ > ω . Let β ∶= β + κ+ (addition of ordinals). Then κ+ ≤ λ+ < α λ gives ( ) ) ̃ γ γ ( ) λ+ < β < α λ . ( ) By the Approximation Lemma 21, it follows as in Proposition 25 that any X ∈ N, σm X ⊆ λ, is contained in an intermediate generic extension V ∏m G∗ am × ∏m G , [ ( ) im ] where a , σ , i is a good pair for κ , i.e. (( m)m<ω ( m m)m<ω) γ • a m < ω is a sequence of pairwise disjoint subsets of κ , such that for ( m S ) γ all κ < κ and m < ω, it follows that a ∩ κ , κ = 1, ν, γ S m [ ν, ν,+1)S • for all m < ω, we have σm ∈ Succ, 0 < σ < γ, and im < ασm ,

• if m ≠ m′, then σ , i ≠ σ ′ , i ′ . ( m m) ( m m ) As before, let f β ∶= X, α ∈ f ∃ a , σ , i good pair for κ ∶ ∀m i < β ∧ ( ) T (( m)m<ω ( m m)m<ω) γ ( m )

86 σm γ ω σ ∏m G∗(am) × ∏m G ∧ ∃ X˙ ∈ Name P × M P m X = X˙ im . ( ) m First, we assume towards a contradiction that f β ∶ dom f β → α(λ) is surjective.

A) + B) Constructing Pβ and capturing f β.

For a condition p ∈ P, let

pβ ∶= p , pσ, aσ , pσ ↾ β × dom pσ , X , ( ∗ ( i i )σ∈Lim ,i<β ( ( x ))σ∈Succ p ) where X ∶= aσ σ ∈ Lim , i ≥ β . p { i S } We deﬁne Pβ and Gβ as before.

The construction of f β ′ ∈ V Gβ and the isomorphism argument for f β = f β ′ ( ) [ ] ( ) are as in Chapter 6.2 and 6.3; except that when constructing the isomorphism π, we now have to set G ν, j ∶= F ν, j for all κ < κ . π0 ( ) π0 ( ) ν,j γ

β C) P preserves cardinals ≥ α(λ) = αγ = sup{αη S 0 < η < γ} .

The arguments here are similar as in Chapter 6.2 C): If α > β +, it follows as γ S S in Lemma 39 that Pβ ≤ κ+ ⋅ β ℵ0 ≤ λ+ ⋅ β + < α . In the case that α = β +, it S S γ S S S S γ γ S S follows that cf β > ω, and as before, we distinguish several cases, whether γ is S S a limit ordinal or γ = γ + 1, and in the latter case, whether γ ∈ Lim, or γ ∈ Succ with γ = γ + 1 etc. We separate P γ (or P γ , or both), and obtain that P γ (or P γ, or the product P γ × P γ) preserves cardinals, while the remaining forcing is now suﬃciently small.

D) A set ´̃(λ) ⊇ dom f β with an injection ι ∶ ´̃(λ) ↪ λ+ ⋅ SβSℵ0 .

As in Chapter 6.2 D) and 6.4 D), we construct in V Gβ a set ´ λ ⊇ dom f β with [ ] ̃( ) an injection ι ∶ ´ λ ↪ λ+ V ⋅ β + V in the case that β + V < α λ , and an ̃( ) ( ) (S S ) (S S ) ( ) injection ι ∶ ´ λ ↪ λ+ V ⋅ β V in the case that β + V = α λ . ̃( ) ( ) S S (S S ) ( ) Together with Chapter 6.5 B) and 6.5 C), this gives the desired contradiction.

Hence, it follows that there must be some α < α λ with α ∉ rg f β. ( )

E) We use an isomorphism argument and obtain a contradiction.

With the same isomorphism argument as in Chapter 6.2 E), it follows that θN λ = ( ) α λ as desired. ( ) Thus, we have shown that also for all cardinals λ ≥ κ , θN λ takes the smallest γ ( ) possible value.

This was the last step in the proof of our main theorem.

87 7 Discussion and Remarks

Our result conﬁrms Shelah’s thesis from [She10, p.2] that in ZF + DC + AX4 it is “better” to look at κ ℵ0 κ a cardinal rather than ´ κ κ a cardinal , in [ ] S ( ) S the sense that by what we have shown, the only restrictions that can be imposed on the θ-function on a set of cardinals in ZF + DC + AX4, are the obvious ones.

From Theorem 1 in [AK10], it follows that increasing the surjective size of ℵ ℵ0 [ ω] together with preserving GCH below ℵω requires a measurable cardinal, which again underlines how diﬀerently ´ ℵ and ℵ ℵ0 behave without the Axiom of ( ω) [ ω] Choice. In further investigation, one might look at the cardinal arithmetic in our constructed model, such as possible (surjective) sizes of κλ κ a cardinal for ( S ) λ << κ.

Another question one might ask is, under what circumstances certain ¬AC-large cardinal properties are preserved in our symmetric extension N. As an example, we will now brieﬂy look at the question whether an inaccessible cardinal κ in the ground model could remain inaccessible in N.

The notion of inaccessibility in ZFC reads as follows: A cardinal κ is inaccessible (or strongly inaccessible) if κ is regular and 2λ < κ holds for all cardinals λ < κ. Hence, it can not be transferred directly to the ¬AC-context, since the power sets ´ λ for λ < κ are ususally not well-ordered. In [BDL07, Chapter 2], one can ﬁnd ( ) several characterizations how inaccessibility can be deﬁned in ZF :

Deﬁnition ([BDL07]). • A regular uncountable cardinal κ is i-inaccessible if for all λ < κ, there is an ordinal α < κ with an injection ι ∶´ λ ↪ α. ( ) • A regular uncountable cardinal κ is v-inaccessible if for all λ < κ, there is

no surjection s ∶ Vλ → κ.

• A regular uncountable cardinal κ is s-inaccessible if for all λ < κ, there is no surjection s ∶´ λ → κ. ( ) Note that i-inaccessibility implies v-inaccessibility, and v-inaccessibility implies s- inaccessibility. It is not diﬃcult to see that a cardinal κ is v-inaccessible if and only if Vκ is a model of second-order ZF (see [BDL07, Chapter 2]).

Let now κ be an inaccessible cardinal in the setting of our theorem: V ⊧ ZFC + GCH with sequences κ 0 < η < γ , α 0 < η < γ as before, with the additional ( η S ) ( η S ) property that for all κη < κ, it follows that also αη < κ. Then by construction, it follows that κ is s-inaccessible in N, while i-inaccessibility of κ is out of reach, since we do not have our power set well-ordered. The question remains whether κ is v-inaccessible in N. By induction over λ, we could prove (using several isomorphism and factoring arguments, similar to those

88 in Chapter 6.2):

Proposition. Let V be a ground model of ZFC + GCH with γ ∈ Ord, and sequences of uncountable cardinals κ 0 < η < γ and α 0 < η < γ with the ( η S ) ( η S ) properties listed in Chapter 2. Moreover, let N ⊇ V denote the symmetric extension constructed in Chapter 3, 4 and 5.

If κ is an inaccessible cardinal in V with the property that for all κη < κ it follows that αη < κ, then κ is v-inaccessible in N: For any λ < κ, there is no surjective N function s ∶ Vλ → κ in N.

In our inductive proof, we show that for any cardinal λ < κ, there exists some κ λ < κ and a cardinal β < κ with an injection ι ∶ V N ↪ β in V G ↾ κ λ . ν,( ) λ λ λ [ ν,( )]

Our next remark is about the following requirement that we put on the sequences κ 0 < η < γ , α 0 < η < γ : ( η S ) ( η S ) ∀ η α = α+ → cf α > ω . ( η ) We mentioned in Chapter 2 that this condition is necessary if we want ∀ηθN κ = ( η) αη under AX4. Moreover, we proved in Chapter 2 that whenever we start from a ground model V ⊧ ZFC + GCH, and construct a symmetric extension N ⊇ V with N ⊧ ZF + DC such that V and N have the same cardinals and coﬁnalities, the following holds:

If κ, α ∈ Card with θN κ = α+, then cfN α > ω. ( ) ( ) One could ask what happens if we drop the requirement that N should extend a ground model V ⊧ ZFC + GCH cardinal-preservingly. Can there be any inner model N ⊧ ZF + DC with cardinals κ, α such that θN κ = α+ and cf N α = ω? ( ) ( ) Let s ∶ 2κ → α denote a surjective function in N. Then with DC, it follows that there is also a surjection s ∶ 2κ ω → αω in N; and we also have a surjective 1 ( ) function s ∶ 2κ → 2κ ω. 0 ( ) Recall that in Chapter 2, we then took a surjection s ∶ αω V → α+ V from ̃2 ( ) ( ) our ground model V , which gave a surjection s ∶ αω N → α+ N in N. Then 2 ( ) ( ) s ○ s ○ s ∶ 2κ → α+ was a surjective function in N; hence, θN κ ≥ α++. 2 1 0 ( ) In a more general setting, where we can not refer to a ground model V , we try to use the constructible universe L = LN instead. Under the assumption that 0♯ does not exist, it follows by Jensen’s Covering Theorem ([DJ75]) that L does not diﬀer drastically from N: In particular, L and N have the same successors of singular cardinals; so if cf N α = ω, then α+ L = α+ N . ( ) ( ) ( ) This gives the following lemma:

89 Lemma. Let N be an inner model of ZF + DC with N ⊧ “ 0♯ does not exist”, and α ∈ CardN with cfN α = ω. Then there exists a surjective function s ∶ αω N → ( ) 2 ( ) α+ N in N. ( ) Proof. Let α i < ω denote a strictly increasing sequence in N which is coﬁnal ( i S ) in α. First, we construct in N an injection ι ∶ 2α L ↪ αω N , ι = ι ○ ι ○ ι , as ( ) ( ) 2 1 0 follows:

• Let ι ∶ 2α L → 2αi L denote the injection that maps any g ∶ α → 2, 0 ( ) ∏i<ω( ) g ∈ L, to the sequence of its restrictions g ↾ α i < ω . ( i) S • For any i < ω, there is in L an injection γ ∶ 2αi L ↪ α+ L; so with DC ( ) ( i ) in N, we can choose a sequence of injective maps γ i < ω such that ( i S ) γ ∶ 2αi L ↪ α+ L for all i < ω. Then we deﬁne in N an injection ι ∶ i ( ) ( i ) 1 2αi L → α+ L by setting ι X i < ω ∶= γ X i < ω . ∏i<ω( ) ∏i<ω( i ) 1( i S ) ( i( i) S ) • Finally, since α+ L ≤ α+ N < α for all i < ω, it follows that there is in N ( i ) ( i ) an injective map ι ∶ α+ L ↪ αω N . 2 ∏i<ω( i ) ( ) Thus, ι ∶= ι ○ ι ○ ι ∶ 2α L ↪ αω N is an injection in N; which yields a surjection 2 1 0 ( ) ( ) s ∶ αω N → 2α L, or s ∶ αω N → α+ L. ( ) ( ) ( ) ( ) Since we have assumed that N ⊧ “0♯ does not exist” and cf N α = ω, it follows by ( ) Jensen’s Covering Lemma in N that α+ L = α+ N . ( ) ( ) This gives our surjecion s ∶ αω N → α+ N in N as desired. 2 ( ) ( )

Corollary. Let N be an inner model of ZF + DC with N ⊧ “ 0♯ does not exist”, and cardinals κ, α such that θN κ = α+. Then cfN α > ω. ( ) ( ) Proof. Let s ∶ 2κ → α denote a surjective function in N, and assume towards a contradiction that cf N α = ω. As mentioned before, we have surjections s ∶ 2κ → ( ) 0 2κ ω and s ∶ 2κ ω → αω. By the previous lemma, it follows that there is also a ( ) 1 ( ) ω + surjection s2 ∶ α → α in N. Setting s ∶= s2 ○ s1 ○ s0, we obtain in N a surjective function s ∶ 2κ → α+. Contradiction.

Thus, without large cardinal assumptions, it is not possible to achieve θN κ = α+ ( ) for cardinals κ, α with cf N α = ω. ( )

Finally, we remark that our theorem gives a result about possible behaviors of the θ-function on a set of uncountable cardinals. Unfortunately, a straightforward generalization of our forcing notion to ordinal-length sequences κ η ∈ Ord , ( η S ) α η ∈ Ord does not result in a ZF -model: ( η S ) Denote by P the class forcing which canonically generalizes our forcing notion P to sequences κ η ∈ Ord , α η ∈ Ord of ordinal length; denote by G a ( η S ) ( η S )

90 V -generic filter on P, and N ∶= V G . Then N ⊭ P ower Set: Assume towards ( ) a contradiction that Z ∶= Ń ℵ ∈ N. Then there would be an ordinal γ and a ( 1) symmetric name Z˙ ∈ HS ∩ Name P ↾ γ with Z = Z˙ G↾γ, where P ↾ γ denotes the ( ) initial part of P up to κγ . Now, by an isomorphism argument similar as in the Approximation Lemma 21, one can show that any set X ∈ Ń ℵ is contained ( 1) ηm in an intermediate generic extension V ∏ G with ηm < γ, im < αη for [ m<ω im ] m all m < ω. Consider X ∶= Gγ+1 ↾ ℵ for some i < α . Then X ∈ Ń ℵ ; hence, i 1 γ+1 ( 1) γ+1 ηm X = G ↾ ℵ1 ∈ V ∏ G for a sequence ηm, im m < ω , such that ηm < γ, i [ m<ω im ] (( ) S ) γ+1 ηm im < αη holds for all m < ω. But this is not possible, since G is V ∏ G - m i [ m<ω im ] generic on P γ+1.

Broadly speaking, the point is that a class-sized version of our forcing construction never stops adding new subsets of ℵ1 (or any other uncountable cardinal). Although we can try and keep control over the surjective size of Ń ℵ , it is not ( 1) possible to capture Ń ℵ in an appropriate set-sized intermediate generic exten- ( 1) sion; and it remains a future project to settle this problem and find a countably closed forcing notion that is also suitable for sequences κ η ∈ Ord , α η ∈ Ord ( η S ) ( η S ) of ordinal length.

References

[AK10] Arthur W. Apter and Peter Koepke. The consistency strength of choice- less failures of SCH. Journal of Symbolic Logic, 75(3):1066–1080, 2010.

[BDL07] Andreas Blass, Ioanna M. Dimitriou, and Benedikt L¨owe. Inaccessible cardinals without the axiom of choice. Fund. Math., 194(2):179–189, 2007.

[Ber42] Paul Bernays. A system of axiomatic set theory. III. Inﬁnity and enu- merability. Analysis. J. Symbolic Logic, 7:65–89, 1942.

[Can78] Georg Cantor. Ein Beitrag zur Mannigfaltigkeitslehre. J. F. Math., 84:242–258, 1878.

[Coh63] Paul Cohen. The independence of the continuum hypothesis. Proc. Nat. Acad. Sci. U.S.A., 50:1143–1148, 1963.

[Coh64] Paul J. Cohen. The independence of the continuum hypothesis. II. Proc. Nat. Acad. Sci. U.S.A., 51:105–110, 1964.

[Dim11] Ioanna Dimitriou. Symmetric Models, Singular Cardinal Patterns, and Indiscernibles. PhD thesis, Rheinische Friedrich-Wilhelms-Universit¨at Bonn, 2011.

91 [DJ75] Keith I. Devlin and R. B. Jensen. Marginalia to a theorem of Silver. pages 115–142. Lecture Notes in Math., Vol. 499, 1975.

[Eas70] William B. Easton. Powers of regular cardinals. Annals of Pure and Applied Logic, 1:139–178, 1970.

[FK16] Anne Fernengel and Peter Koepke. An Easton-like theorem for Zermelo- Fraenkel set theory without choice. arXiv:1607.00205, 2016.

[Fri00] Sy D. Friedman. Fine structure and class forcing, volume 3 of de Gruyter Series in Logic and its Applications. Walter de Gruyter & Co., Berlin, 2000.

[G¨o40] Kurt G¨odel. The Consistency of the Continuum Hypothesis. Annals of Mathematics Studies, no. 3. Princeton University Press, Princeton, N. J., 1940.

[GH75] Fred Galvin and Andr´as Hajnal. Inequalities for cardinal powers. Ann. of Math. (2), 101:491–498, 1975.

[Git80] M. Gitik. All uncountable cardinals can be singular. Israel Journal of Mathematics, 35(1-2):61 – 88, 1980.

[Git89] Moti Gitik. The negation of the singular cardinal hypothesis from o κ = ( ) κ++. Annals of Pure and Applied Logic, 43(3):209–234, 1989.

[Git91] Moti Gitik. The strength of the failure of the singular cardinal hypothesis. Annals of Pure and Applied Logic, 51(3):215–240, 1991.

[GK12] Moti Gitik and Peter Koepke. Violating the singular cardinals hypothesis without large cardinals. Israel J. Math., 191(2):901–922, 2012.

[Hau07] Felix Hausdorff. Untersuchungen über Ordnungstypen. Berichte über die Verhandlungen der Königlich Sächsischen Gesellschaft der Wissenschaften zu Leipzig, mathematisch-physische Klasse, 59:84–159, 1907.

[Hau08] Felix Hausdorﬀ. Grundz¨uge einer Theorie der geordneten Mengen. Math. Ann., 65(4):435–505, 1908.

[Hau14] Felix Hausdorﬀ. Grundz¨uge der Mengenlehre. Veit and company, Leipzig, 1914.

[Hau02] Felix Hausdorff. Felix Hausdorff – gesammelte Werke. Band II. Springer-Verlag, Berlin, 2002. “Grundzüge der Mengenlehre”. [“Foun- dations of set theory”], Edited and with commentary by E. Brieskorn, S. D. Chatterji, M. Epple, U. Felgner, H. Herrlich, M. Huˇsek, V. Kanovei, P. Koepke, G. Preuß, W. Purkert and E. Scholz.

92 [HKL+16] Peter Holy, Regula Krapf, Philipp L¨ucke, Ana Njegomir, and Philipp Schlicht. Class forcing, the forcing theorem and Boolean completions. J. Symb. Log., 81(4):1500–1530, 2016.

[Jec73] Thomas J. Jech. The Axiom of Choice. North-Holland Publishing Co., Amsterdam, 1973. Studies in Logic and the Foundations of Mathemat- ics, Vol. 75.

[Jec06] Thomas Jech. Set Theory. Springer Monographs in Mathematics. Springer-Verlag, Berlin, 2006. The Third Millenium Edition, revised and expanded.

[Kar14] Asaf Karagila. Embedding orders into the cardinals with DCκ. Fund. Math., 226(2):143–156, 2014.

[Kun06] Kenneth Kunen. Set theory. An introduction to independence proofs, volume 102 of Studies in Logic and the Foundations of Mathematics. Elsevier, Amsterdam, 2006.

[Mag77a] Menachem Magidor. On the singular cardinals problem. I. Israel J. Math., 28(1-2):1–31, 1977.

[Mag77b] Menachem Magidor. On the singular cardinals problem. II. Ann. of Math. (2), 106(3):517–547, 1977.

[Mer07] Carmi Merimovich. A power function with a ﬁxed ﬁnite gap everywhere. J. Symbolic Logic, 72(2):361–417, 2007.

[Mit01] William J. Mitchell. The covering lemma. Handbook of Set Theory, 2001.

[Mit10] William J. Mitchell. The covering lemma. In Handbook of set theory. Vols. 1, 2, 3, pages 1497–1594. Springer, Dordrecht, 2010.

[Moo11] Gregory H. Moore. Early history of the generalized continuum hypothesis: 1878–1938. Bull. Symbolic Logic, 17(4):489–532, 2011.

[She94] Saharon Shelah. Cardinal arithmetic, volume 29 of Oxford Logic Guides. The Clarendon Press, Oxford University Press, New York, 1994. Oxford Science Publications.

[She97] Saharon Shelah. Set theory without choice: not everything on coﬁnality is possible. Arch. Math. Logic, 36(2):81–125, 1997.

[She10] Saharon Shelah. Pcf without choice. arxiv:math/0510229, 2010.

[She16] Saharon Shelah. ZF + DC + AX4. Arch. Math. Logic, 55(1-2):239–294, 2016.

93 [Sil75] Jack Silver. On the singular cardinals problem. In Proceedings of the International Congress of Mathematicians (Vancouver, B. C., 1974), Vol. 1, pages 265–268. Canad. Math. Congress, Montreal, Que., 1975.

[Zar73] Andrzej Zarach. Forcing with proper classes. Fund. Math., 81(1):1– 27, 1973. Collection of articles dedicated to Andrzej Mostowski on the occasion of his sixtieth birthday, I.

Anne Fernengel, Mathematisches Institut, Rheinische Friedrich-Wilhelms-Universitat,¨ Bonn, Germany E-Mail address: [email protected]

Peter Koepke, Mathematisches Institut, Rheinische Friedrich-Wilhelms-Universitat,¨ Bonn, Germany E-Mail address: [email protected]