introduction - assaf rinot

17
HIGHER SOUSLIN TREES AND THE GCH, REVISITED ASSAF RINOT Abstract. It is proved that for every uncountable cardinal , GCH + ( + ) entails the existence of a cf()-complete + -Souslin tree. In particular, if GCH holds and there are no β„΅2-Souslin trees, then β„΅2 is weakly compact in GΒ¨ odel’s constructible universe, improving Gregory’s 1976 lower bound. Furthermore, it follows that if GCH holds and there are no β„΅2 and β„΅3 Souslin trees, then the Axiom of Determinacy holds in (R). 1. Introduction 1.1. Background. A tree is a partially ordered set (,< ) with the property that for every ∈ , the downward cone ↓ := { ∈ | < } is well-ordered by < . The order type of ( ↓ ,< ) is denoted by ht(), and the β„Ž -level of the tree is the set := { ∈ | ht()= }. Let denote a regular uncountable cardinal. A -Aronszajn tree is a tree of size having no chains or levels of size .A -Souslin tree is a tree of size having no chains or antichains of size . As tree-levels are antichains, any -Souslin tree is a -Aronszajn tree. The above concepts stemmed from a 1920 question of Mikhail Souslin [Sou20], asking whether every ccc dense complete linear ordering with no endpoints is isomorphic to the real line. 1 Kurepa realized [Kur35] that a negative answer is equivalent to the existence of (what we nowadays call) an β„΅ 1 -Souslin tree. However, all of Kurepa’s attempts to construct such a tree were unsuccessful. At one point, Kurepa told Aronszajn about his goal, and in response, Aronszajn came up with a construction of a poor man’s version of a Souslin tree; Indeed, Aronszajn constructed (what we nowadays call) an β„΅ 1 -Aronszajn tree. It was only three decades after [Kur35], in [Ten68], [Jec67], [Jen68], and [ST71], when it was discovered that β€” unlike β„΅ 1 -Aronszajn trees β€” the existence of an β„΅ 1 -Souslin tree is independent of the usual axioms of set theory (ZFC). As these objects proven incredibly useful and important, a systematic study of their consistency and interrelation was carried out. Jensen proved [Jen72] that in GΒ¨ odel’s constructible universe, , for every regular uncountable cardinal , the following are equivalent: βˆ™ There exists a -Souslin tree; βˆ™ There exists a -Aronszajn tree; βˆ™ is not a weakly compact cardinal. In another work, Jensen proved (see the monograph [DJ74]) that the existence of an β„΅ 1 -Souslin tree is independent of ZFC+GCH. 2 As for Aronszajn trees, Specker proved [Spe49] that GCH entails the existence of a + -Aronszajn tree for every regular cardinal , and Mitchell and Silver Date : February 21, 2017. 2010 Mathematics Subject Classification. Primary 03E05; Secondary 03E35. Key words and phrases. Souslin tree, microscopic approach, weakly compact cardinal, square. 1 Here, ccc is a consequence of separability, asserting that every pairwise-disjoint family of open intervals is countable. 2 GCH is an abbreviation for the Generalized Continuum Hypothesis, asserting that 2 β„΅ = β„΅+1 for every ordinal . 1

Upload: others

Post on 13-Nov-2021

7 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Introduction - Assaf Rinot

HIGHER SOUSLIN TREES AND THE GCH, REVISITED

ASSAF RINOT

Abstract. It is proved that for every uncountable cardinal πœ†, GCH+οΏ½(πœ†+) entails the existenceof a cf(πœ†)-complete πœ†+-Souslin tree. In particular, if GCH holds and there are no β„΅2-Souslin trees,then β„΅2 is weakly compact in Godel’s constructible universe, improving Gregory’s 1976 lower bound.Furthermore, it follows that if GCH holds and there are no β„΅2 and β„΅3 Souslin trees, then the Axiomof Determinacy holds in 𝐿(R).

1. Introduction

1.1. Background. A tree is a partially ordered set (𝑇,<𝑇 ) with the property that for every π‘₯ ∈ 𝑇 ,the downward cone π‘₯↓ := {𝑦 ∈ 𝑇 | 𝑦 <𝑇 π‘₯} is well-ordered by <𝑇 . The order type of (π‘₯↓, <𝑇 ) is

denoted by ht(π‘₯), and the π›Όπ‘‘β„Ž-level of the tree is the set 𝑇𝛼 := {π‘₯ ∈ 𝑇 | ht(π‘₯) = 𝛼}.Let πœ… denote a regular uncountable cardinal. A πœ…-Aronszajn tree is a tree of size πœ… having no

chains or levels of size πœ…. A πœ…-Souslin tree is a tree of size πœ… having no chains or antichains of sizeπœ…. As tree-levels are antichains, any πœ…-Souslin tree is a πœ…-Aronszajn tree.

The above concepts stemmed from a 1920 question of Mikhail Souslin [Sou20], asking whetherevery ccc dense complete linear ordering with no endpoints is isomorphic to the real line.1 Kureparealized [Kur35] that a negative answer is equivalent to the existence of (what we nowadays call)an β„΅1-Souslin tree. However, all of Kurepa’s attempts to construct such a tree were unsuccessful.At one point, Kurepa told Aronszajn about his goal, and in response, Aronszajn came up with aconstruction of a poor man’s version of a Souslin tree; Indeed, Aronszajn constructed (what wenowadays call) an β„΅1-Aronszajn tree.

It was only three decades after [Kur35], in [Ten68], [Jec67], [Jen68], and [ST71], when it wasdiscovered that β€” unlike β„΅1-Aronszajn trees β€” the existence of an β„΅1-Souslin tree is independentof the usual axioms of set theory (ZFC).

As these objects proven incredibly useful and important, a systematic study of their consistencyand interrelation was carried out. Jensen proved [Jen72] that in Godel’s constructible universe, 𝐿,for every regular uncountable cardinal πœ…, the following are equivalent:

βˆ™ There exists a πœ…-Souslin tree;βˆ™ There exists a πœ…-Aronszajn tree;βˆ™ πœ… is not a weakly compact cardinal.

In another work, Jensen proved (see the monograph [DJ74]) that the existence of an β„΅1-Souslintree is independent of ZFC+GCH.2 As for Aronszajn trees, Specker proved [Spe49] that GCHentails the existence of a πœ†+-Aronszajn tree for every regular cardinal πœ†, and Mitchell and Silver

Date: February 21, 2017.2010 Mathematics Subject Classification. Primary 03E05; Secondary 03E35.Key words and phrases. Souslin tree, microscopic approach, weakly compact cardinal, square.

1Here, ccc is a consequence of separability, asserting that every pairwise-disjoint family of open intervals is countable.2GCH is an abbreviation for the Generalized Continuum Hypothesis, asserting that 2℡𝛼 = ℡𝛼+1 for every ordinal 𝛼.

1

Page 2: Introduction - Assaf Rinot

2 ASSAF RINOT

proved [Mit73] that the nonexistence of an β„΅2-Aronszajn tree is equiconsistent with the existenceof a weakly compact cardinal.

Altogether, the four theorems crystallized the following question:

Question (folklore, see [KM78]). Does GCH entail the existence of an β„΅2-Souslin tree?If not, is the consistency strength of a negative answer a weakly compact cardinal?

Supporting the thesis of the question, Laver and Shelah [LS81] managed to construct a modelof ZFC + CH in which there are no β„΅2-Souslin trees, indeed assuming the consistency of a weaklycompact cardinal. However, GCH fails in their model.

More work connecting higher trees and weakly compact cardinals was then carried out in [Tod81],[SS82a], [SS82b], [SS88], [Tod89], yet the best known result on the original question remained thefollowing:

Theorem (Gregory, [Gre76]). If GCH holds and there are no β„΅2-Souslin trees, then β„΅2 is a Mahlocardinal in L.

In this paper, Gregory’s 1976 lower bound is increased to the anticipated value:

Theorem A. If GCH holds and there are no β„΅2-Souslin trees, then β„΅2 is weakly compact in 𝐿.

In [Tod81], Todorcevic proved that after Levy-collapsing a weakly compact cardinal to β„΅2 overa model of GCH: GCH holds, and every β„΅2-Aronszajn tree contains an β„΅1-Aronszajn subtree. Astrengthening of Theorem A, then, provides the following optimal result:

Theorem A’. If GCH holds, and β„΅2 is not weakly compact in 𝐿, then there exists an β„΅2-Souslintree with no β„΅1-Aronszajn subtrees.

We remind the reader that a regular uncountable cardinal πœ… is said to be Mahlo if the set ofregular cardinals below πœ… is stationary in πœ…. A regular uncountable cardinal πœ… is said to be weaklycompact if it satisfies the generalized Ramsey partition relation: πœ… β†’ (πœ…)22. By a theorem of Hanf[Han64], every weakly compact cardinal must have stationarily many Mahlo cardinals below it.

1.2. Some details. Recall that a coherent 𝐢-sequence (over a regular uncountable cardinal πœ…) isa sequence βŸ¨πΆπ›Ό | 𝛼 < πœ…βŸ© such that:

(1) for all 𝛼 < πœ…, 𝐢𝛼 βŠ† 𝛼;(2) for all limit 𝛼 < πœ…, 𝐢𝛼 is a club in 𝛼;(3) for all 𝛼 < πœ…, if οΏ½οΏ½ ∈ acc(𝐢𝛼), then 𝐢�� = 𝐢𝛼 ∩ οΏ½οΏ½.3

The easiest way to obtain such a sequence is to fix some club 𝐷 in πœ…, and let 𝐢𝛼 := 𝐷 ∩ 𝛼 for all𝛼 ∈ acc(𝐷), and 𝐢𝛼 := 𝛼 βˆ– sup(𝐷 ∩ 𝛼) for all other 𝛼. Of more interest is the following concept:

Definition 1.1 (Jensen, [Jen72]). οΏ½πœ†(𝐸) asserts the existence of a coherent 𝐢-sequence, βŸ¨πΆπ›Ό | 𝛼 < πœ†+⟩,such that otp(𝐢𝛼) ≀ πœ† and acc(𝐢𝛼) ∩ 𝐸 = βˆ… for all 𝛼 < πœ†+.

Write οΏ½πœ† for οΏ½πœ†(βˆ…).

Jensen proved [Jen72] that if there exists a stationary subset 𝐸 βŠ† πœ†+ for which β™’(𝐸) + οΏ½πœ†(𝐸)holds,4 then there exists a πœ†+-Souslin tree, and Solovay noticed that the existence of such an 𝐸follows from β™’(πœ†+) +οΏ½πœ†. Gregory proved [Gre76] that GCH implies β™’(πœ†+) for every cardinal πœ† ofuncountable cofinality, and Shelah [She79],[She81] improved this to every uncountable cardinal πœ†.Altogether:

3Here, acc(𝐢) stands for the set of accumulation points of 𝐢, that is, acc(𝐢) := {𝛽 ∈ 𝐢 | sup(𝐢 ∩ 𝛽) = 𝛽 > 0}.4The definition of the Diamond principle may be found at the beginning of Section 3 below.

Page 3: Introduction - Assaf Rinot

HIGHER SOUSLIN TREES AND THE GCH, REVISITED 3

Fact 1.2 (1970’s). For every uncountable πœ†, GCH +οΏ½πœ† entails the existence of a πœ†+-Souslin tree.

By results of Jensen and Solovay, the failure of οΏ½β„΅1 is equiconsistent with the existence of aMahlo cardinal, hence, in view of the goal of deriving a weakly compact cardinal, one should lookat weaker hypothesis than οΏ½β„΅1 .

One consequence of οΏ½πœ† is that every stationary subset of πœ†+ contains a nonreflecting stationarysubset. Baumgartner proved [Bau76b] that after forcing to Levy-collapse a weakly compact cardinal

to β„΅2, every stationary subset of 𝐸℡2=β„΅1

reflects, and in [Gre76], Gregory indeed managed to reduce

the hypothesis οΏ½πœ† into that of the existence of nonreflecting stationary subset of πΈπœ†+

=πœ†, providedthat πœ† is regular.

However, ten years later, Harrington and Shelah [HS85] proved that the reflection of every

stationary subset of 𝐸℡2=β„΅1

is equiconsistent with the existence of a Mahlo cardinal, sending some

people back to the drawing table.

So what should we do?At the end of [KS93], Kojman and Shelah suggested to try proving that GCH entails the existence

of an β„΅2-Souslin tree, under the hypothesis that there exist two stationary subsets of 𝐸℡2=β„΅1

which do

not reflect simultaneously. The point here is that Magidor proved [Mag82] that the failure of suchsimultaneous reflection is equiconsistent with the existence of a weakly compact cardinal. However,as of now, this idea appears to be infertile.

Lastly, during our visit to the Erwin Schroedinger Institute in 2009, B. Konig suggested to usto try deriving an β„΅2-Souslin tree from the combination of GCH and Todorcevic’s principle οΏ½(β„΅2).Here, οΏ½(πœ…) asserts the existence of a coherent 𝐢-sequence, βŸ¨πΆπ›Ό | 𝛼 < πœ…βŸ©, of the weakest possiblenontrivial form. That is, for every club 𝐷 in πœ…, there exists some 𝛼 ∈ acc(𝐷) for which 𝐢𝛼 = π·βˆ©π›Ό.This time, the key point is that by [Tod87], if πœ… is a regular uncountable cardinal and οΏ½(πœ…) fails,then πœ… is a weakly compact cardinal in 𝐿.

And indeed, the main result of this paper is the following.

Theorem B. If πœ† an uncountable cardinal, and GCH +οΏ½(πœ†+) holds, then:

(1) There exists a πœ†+-Souslin tree which is cf(πœ†)-complete;5

(2) There exists a πœ†+-Souslin tree which is club-regressive.6

So, Theorems A,A’ follow from Theorem B. As explained in Section 4, it also follows that thenonexistence of higher Souslin trees at two successive cardinals has considerably stronger strength:

Theorem C. If GCH holds and there are no πœ…-Souslin trees for πœ… ∈ {β„΅2,β„΅3}, then the Axiom ofDeterminacy holds in 𝐿(R).

By Fact 1.2 and the main result of [Ste05], GCH and the nonexistence of an β„΅πœ”+1-Souslin treeentails that the Axiom of Determinacy holds in 𝐿(R). So the surprise here is the move down intothe realm of the ℡𝑛 for finite 𝑛 (and away from the successors of singular cardinals).

1.3. Organization of this paper. In Section 2, we introduce a new ideal, which we denote by𝐽 [πœ…], and study its extent. It is established that if GCH holds, then for every uncountable cardinalπœ†, 𝐽 [πœ†+] contains a stationary set, yet, it is consistent that 𝐽 [β„΅2] is the nonstationary ideal over β„΅2.

In Section 3, we deal with the Diamond principle and a weak consequence of it β€” club hitting.

5A tree (𝑇,<𝑇 ) is said to be πœƒ-complete if any <𝑇 -increasing sequence of elements from 𝑇 , and of length < πœƒ, has anupper bound in 𝑇 .6The definition of a club-regressive tree may be found in [BR15]. For our purpose, it suffices to mention that aclub-regressive πœ…-tree contains no 𝜈-Aronszajn subtrees nor 𝜈-Cantor subtrees for every regular cardinal 𝜈 < πœ….

Page 4: Introduction - Assaf Rinot

4 ASSAF RINOT

In Section 4, we recall the principle οΏ½βˆ’(𝑆) from [BR15], and prove that for every regular un-countable cardinal πœ… and every stationary 𝑆 ∈ 𝐽 [πœ…], οΏ½(πœ…) + β™’(πœ…) entails οΏ½βˆ’(𝑆). This suffices toobtain an asymptotic version of Theorem B(2) that requires only a local instance of the full GCH:

Theorem D. For every cardinal πœ† β‰₯ iπœ”, CHπœ† +οΏ½(πœ†+) entails the existence of a πœ†+-Souslin treewhich is club-regressive.

Then, we introduce the principle οΏ½β€²(𝑆), and prove that οΏ½βˆ’(πœ…) + β™’(πœ…) entails οΏ½β€²(𝑆) for everystationary subset 𝑆 βŠ† πœ…, from which Theorem B(1) follow. Finally, Theorems A,A’,C are derivedas corollaries.

1.4. Notations and conventions. Throughout, by a cardinal, we mean an infinite cardinal. WriteCHπœ† to assert that 2πœ† = πœ†+. Denote πΈπœ…

πœƒ := {𝛼 < πœ… | cf(𝛼) = πœƒ}, and define πΈπœ…=πœƒ, 𝐸

πœ…>πœƒ and πΈπœ…

β‰₯πœƒ in a

similar fashion. Write [𝑋]πœƒ for the collection of all subsets of 𝑋 of cardinality πœƒ. A dense subfamilyof [𝑋]πœƒ is a collection β„± βŠ† [𝑋]πœƒ such that for all π‘Œ ∈ [𝑋]πœƒ, there exists some 𝑍 ∈ β„± with 𝑍 βŠ† π‘Œ .The least size of such a dense subfamily is denoted by π’Ÿ(𝑋, πœƒ).

For sets of ordinals 𝐢,𝐷, write 𝐷 βŠ‘ 𝐢 iff there exists some ordinal 𝛽 such that 𝐷 = 𝐢 ∩ 𝛽, thatis, 𝐢 end-extends 𝐷. Denote nacc(𝐢) := 𝐢 βˆ– acc(𝐢).

2. The regressive functions ideal

Recall that a function 𝑓 : 𝛼→ 𝛼 is said to be regressive iff 𝑓(𝛽) < 𝛽 for all nonzero 𝛽 < 𝛼.

Definition 2.1. For a regular uncountable cardinal πœ…, define a collection 𝐽 [πœ…], as follows.A subset 𝑆 βŠ† πœ… is in 𝐽 [πœ…] iff there exists a club 𝐢 βŠ† πœ… and a sequence of functions βŸ¨π‘“π‘– : πœ…β†’ πœ… | 𝑖 < πœ…βŸ©

satisfying the following. For every 𝛼 ∈ π‘†βˆ©πΆ, every regressive function 𝑓 : 𝛼→ 𝛼, and every cofinalsubset 𝐡 βŠ† 𝛼, there exists some 𝑖 < 𝛼 such that sup{𝛽 ∈ 𝐡 | 𝑓𝑖(𝛽) = 𝑓(𝛽)} = 𝛼.

It is easy to see that 𝐽 [πœ…] is a πœ…-complete normal ideal. We now turn to study its extent.

2.1. Positive results.

Proposition 2.2. Suppose that πœƒ < cf(πœ†) ≀ πœ† are cardinals, and π’Ÿ(πœ†, πœƒ) = πœ†.

Then πΈπœ†+

πœƒ ∈ 𝐽 [πœ†+].

Proof. If πœƒ is singular, then πΈπœ†+

πœƒ = βˆ… ∈ 𝐽 [πœ†+], so suppose that πœƒ is regular. Fix a family {𝐴𝛼 |𝛼 < πœ†+} βŠ† [πœ†]cf(πœ†) such that |𝐴𝛼 ∩ 𝐴𝛽| < cf(πœ†) for all 𝛼 < 𝛽 < πœ†+.7 Let {𝑔𝑖 | 𝑖 < πœ†} be a dense

subfamily of [πœ† Γ— πœ†]πœƒ. For all 𝛽 < πœ†+, fix a bijection πœ“π›½ : πœ† ↔ max{πœ†, 𝛽}. Finally, for all 𝑖 < πœ†,define a function 𝑓𝑖 : πœ†+ β†’ πœ†+ as follows. For all 𝛽 < πœ†+, if there exists a unique (𝛾, 𝛿) ∈ 𝑔𝑖 suchthat 𝛾 ∈ 𝐴𝛽, then let 𝑓𝑖(𝛽) := πœ“π›½(𝛿). Otherwise, let 𝑓𝑖(𝛽) := 0.

We claim that the club 𝐢 := πœ†+ βˆ– πœ† and the sequence βŸ¨π‘“π‘– : πœ†+ β†’ πœ†+ | 𝑖 < πœ†βŸ© together witness

that πΈπœ†+

πœƒ ∈ 𝐽 [πœ†+].8 To see this, fix an arbitrary 𝛼 ∈ πΈπœ†+

πœƒ ∩ 𝐢 along with a regressive function𝑓 : 𝛼 β†’ 𝛼, and a cofinal subset 𝐡 of 𝛼. Let 𝐡′ be a cofinal subset of 𝐡 βˆ– {0} of order-type πœƒ.Since πœƒ < cf(πœ†), for all 𝛽 ∈ 𝐡′, we may pick 𝛾𝛽 ∈ 𝐴𝛽 βˆ–

⋃{π΄πœ‚ | πœ‚ ∈ 𝐡′, πœ‚ = 𝛽}. Also, for all

𝛽 ∈ 𝐡′, since 𝑓(𝛽) < 𝛽 βŠ† Im(πœ“π›½), we may let 𝛿𝛽 := πœ“βˆ’1𝛽 (𝑓(𝛽)). Put 𝑔 := {(𝛾𝛽, 𝛿𝛽) | 𝛽 ∈ 𝐡′}.

As 𝛽 ↦→ 𝛾𝛽 is one-to-one over 𝐡′, we get that 𝑔 ∈ [πœ† Γ— πœ†]πœƒ, and hence we may find some 𝑖 < πœ†such that 𝑔𝑖 βŠ† 𝑔. Let 𝐡′′ := {𝛽 ∈ 𝐡′ | (𝛾𝛽, 𝛿𝛽) ∈ 𝑔𝑖}. Since 𝐡′′ βŠ† 𝐡′ and |𝐡′′| = otp(𝐡′),we get that sup(𝐡′′) = sup(𝐡′) = 𝛼. Thus, it suffices to verify that 𝑓 οΏ½ 𝐡′′ = 𝑓𝑖 οΏ½ 𝐡′′. Let

7See, e.g., [Bau76a, Theorem 2.3].8This is not a typing error; we simply settle for a sequence of length πœ†.

Page 5: Introduction - Assaf Rinot

HIGHER SOUSLIN TREES AND THE GCH, REVISITED 5

𝛽 ∈ 𝐡′′ be arbitrary. Since (𝛾𝛽, 𝛿𝛽) is the unique pair in 𝑔𝑖 to satisfy 𝛾𝛽 ∈ 𝐴𝛽, we conclude that

𝑓𝑖(𝛽) = πœ“π›½(𝛿𝛽) = πœ“π›½(πœ“βˆ’1𝛽 (𝑓(𝛽))) = 𝑓(𝛽), as sought. οΏ½

In particular, GCH entails that πΈπœ†+

<πœ† ∈ 𝐽 [πœ†+] for every regular cardinal πœ†. To deal with the caseof πœ† singular, let us recall the definition of Shelah’s approachability ideal :

Definition 2.3 (Shelah, [She93]). For a regular uncountable cardinal πœ…, define a collection 𝐼[πœ…],as follows. A subset 𝑆 βŠ† πœ… is in 𝐼[πœ…] iff there exists a club 𝐢 βŠ† πœ… and a sequence βŸ¨π‘Žπ‘– | 𝑖 < πœ…βŸ© ofbounded subsets of πœ… satisfying the following. For every 𝛼 ∈ 𝑆 ∩ 𝐢, there exists a cofinal subset𝐡 βŠ† 𝛼 of order-type cf(𝛼) < 𝛼 such that {𝐡 ∩ 𝑖 | 𝑖 < 𝛼} βŠ† {π‘Žπ‘– | 𝑖 < 𝛼}.

Proposition 2.4. Suppose that cf(πœ†) < πœƒ < πœ† are cardinals, and π’Ÿ(πœ†, πœƒ) = πœ†.

For every 𝑆 βŠ† πΈπœ†+

πœƒ , if 𝑆 ∈ 𝐼[πœ†+], then 𝑆 ∈ 𝐽 [πœ†+].

Proof. By Shelah’s celebrated theorem from [She94a, Theorem 1.5], let us fix a scale β„Ž = βŸ¨β„Žπ›Ό | 𝛼 < πœ†+⟩for πœ†. This means, in particular, that for all 𝛼 < πœ†+, β„Žπ›Ό is a function from cf(πœ†) to πœ†. Fix a bijectionπœ‹ : cf(πœ†) Γ— πœ†β†” πœ†, and let 𝐴𝛼 := πœ‹[β„Žπ›Ό] for all 𝛼 < πœ†+.

Suppose that we are given 𝑆 βŠ† πΈπœ†+

πœƒ in 𝐼[πœ†+]. To avoid trivialities, assume that 𝑆 is stationary.

In particular, πœƒ is regular. By 𝑆 ∈ 𝐼[πœ†+] and [CFM04, Corollary 2.15],9 there exists a club 𝐢 βŠ† πœ†+

such that every 𝛼 ∈ 𝑆 ∩ 𝐢 is good for the scale β„Ž. By [Eis10, Theorem 3.50], this means that forevery 𝛼 ∈ 𝑆 ∩ 𝐢 and every cofinal 𝐡 βŠ† 𝛼, there exists some cofinal 𝐡′ βŠ† 𝐡 of order-type πœƒ and𝑗′ < cf(πœ†) such that β„Žπ›½(𝑗) < β„Žπ›Ύ(𝑗) for all 𝛽 < 𝛾 both from 𝐡′ and all 𝑗 ∈ (𝑗′, cf(πœ†)).

Claim 2.4.1. Suppose that 𝛼 ∈ 𝑆 ∩ 𝐢, and 𝐡 is a cofinal subset of 𝛼.Then there exists some cofinal 𝐡′ βŠ† 𝐡 of order-type πœƒ for which the following is nonempty:∏

π›½βˆˆπ΅β€²

𝐴𝛽 βˆ–β‹ƒ

{π΄πœ‚ | πœ‚ ∈ 𝐡′, πœ‚ = 𝛽}.

Proof. By 𝛼 ∈ 𝑆 ∩ 𝐢, 𝛼 is good, so let us pick some cofinal 𝐡′ βŠ† 𝐡 of order-type πœƒ and 𝑗 < cf(πœ†)such that β„Žπ›½(𝑗) < β„Žπ›Ύ(𝑗) for all 𝛽 < 𝛾 both from 𝐡′. Then βŸ¨πœ‹(𝑗, β„Žπ›½(𝑗)) | 𝛽 ∈ π΅β€²βŸ© is an element of∏

π›½βˆˆπ΅β€² 𝐴𝛽 βˆ–β‹ƒ{π΄πœ‚ | πœ‚ ∈ 𝐡′, πœ‚ = 𝛽}. οΏ½

Now, continue as in the proof of Proposition 2.2, but using the sequence βŸ¨π΄π›Ό | 𝛼 < πœ†+⟩ we justconstructed. οΏ½

Corollary 2.5. Suppose that πœƒ is a regular cardinal.For every cardinal πœ† β‰₯ π’Ÿ(πœƒ, πœƒ), the following are equivalent:

(1) π’Ÿ(πœ†, πœƒ) = πœ†;

(2) there exists some stationary subset 𝑆 βŠ† πΈπœ†+

πœƒ with 𝑆 ∈ 𝐽 [πœ†+].

Proof. (1) =β‡’ (2): Suppose that πœ† is a cardinal, satisfying π’Ÿ(πœ†, πœƒ) = πœ†. In particular, πœƒ = cf(πœ†).

I If πœƒ < cf(πœ†), then πΈπœ†+

πœƒ ∈ 𝐽 [πœ†+] by Proposition 2.2.

I If πœƒ > cf(πœ†), then πœ† is singular, and by [She93, S1], there exists some stationary 𝑆 βŠ† πΈπœ†+

πœƒ suchthat 𝑆 ∈ 𝐼[πœ†+]. Now, appeal to Proposition 2.4.

(2) =β‡’ (1): Suppose that πœ† β‰₯ π’Ÿ(πœƒ, πœƒ) is some cardinal, and 𝐽 [πœ†+] contains a stationary subset

of πΈπœ†+

πœƒ . In particular, there exists a sequence of functions βŸ¨π‘“π‘– : πœ†+ β†’ πœ†+ | 𝑖 < πœ†+⟩ and an ordinal

𝛼 ∈ πΈπœ†+

πœƒ with 𝛼 > πœ† such that for every regressive function 𝑓 : 𝛼 β†’ 𝛼, and every cofinal subset𝐡 βŠ† 𝛼, there exists some 𝑖 < 𝛼 such that sup{𝛽 ∈ 𝐡 | 𝑓𝑖(𝛽) = 𝑓(𝛽)} = 𝛼.

9See also [FM97, Claim 4.4].

Page 6: Introduction - Assaf Rinot

6 ASSAF RINOT

Fix a cofinal subset 𝐡 βŠ† 𝛼 of order-type πœƒ, with min(𝐡) > πœ†. For all 𝑖 < 𝛼, let π‘Œπ‘– := 𝑓𝑖[𝐡] βˆͺ πœƒ.As |π‘Œπ‘–| = πœƒ and π’Ÿ(πœƒ, πœƒ) ≀ πœ†, pick ℱ𝑖 of size ≀ πœ† which is a dense subfamily of [π‘Œπ‘–]

πœƒ. Then,β„± := [πœ†]πœƒ ∩

⋃{ℱ𝑖 | 𝑖 < 𝛼} has size ≀ πœ†. To see that β„± is a dense subfamily of [πœ†]πœƒ, let π‘Œ be an

arbitrary element of [πœ†]πœƒ. Pick a regressive function 𝑓 : 𝛼 β†’ 𝛼 such that 𝑓 οΏ½ 𝐡 is a bijection from𝐡 to π‘Œ . Fix 𝑖 < 𝛼 for which 𝐡𝑖 := {𝛽 ∈ 𝐡 | 𝑓𝑖(𝛽) = 𝑓(𝛽)} is cofinal in 𝛼. As otp(𝐡) = cf(𝛼),we have |𝐡𝑖| = πœƒ, so that |π‘Œπ‘– ∩ π‘Œ | = πœƒ. By π‘Œπ‘– ∩ π‘Œ ∈ [π‘Œπ‘–]

πœƒ and the fact that ℱ𝑖 βŠ† β„± , we infer theexistence of 𝑍 ∈ β„± such that 𝑍 βŠ† π‘Œ . οΏ½

Definition 2.6 (Cummings-Foreman-Magidor, [CFM01]). ADSπœ† asserts the existence of a sequenceβŸ¨π΄π›Ό | 𝛼 < πœ†+⟩ such that:

(1) 𝐴𝛼 is a cofinal subset of πœ† of order-type cf(πœ†);(2) For all 𝛽 < πœ†+, there exists some 𝑔 : 𝛽 β†’ πœ† such that the sequence βŸ¨π΄π›Ό βˆ– 𝑔(𝛼) | 𝛼 < π›½βŸ©

consists of pairwise disjoint sets.

Proposition 2.7. Suppose that ADSπœ† holds for a given singular strong limit cardinal πœ†.

Then πΈπœ†+

=cf(πœ†) ∈ 𝐽 [πœ†+].

Proof sketch. Since 𝐽 [πœ†+] is πœ†+-complete, it suffices to prove that πΈπœ†+

πœƒ ∈ 𝐽 [πœ†+] for every regularcardinal πœƒ < πœ† with πœƒ = cf(πœ†). Fix such a cardinal πœƒ. Since πœ† is a strong limit and πœƒ = cf(πœ†), wehave π’Ÿ(πœ†, πœƒ) = πœ† (cf. the proof of Claim 4.5.1). Then, continue as in the proof of Proposition 2.2,but using the sequence βŸ¨π΄π›Ό | 𝛼 < πœ†+⟩ witnessing ADSπœ†, instead. οΏ½

It follows that the model of [GS08] admits a singular cardinal πœ† for which 𝐽 [πœ†+] βŠ† 𝐼[πœ†+].

2.2. Negative results.

Proposition 2.8. Suppose that πœ† is an uncountable cardinal.

For every 𝑆 ∈ 𝐽 [πœ†+], we have that 𝑆 ∩ πΈπœ†+

cf(πœ†) is nonstationary.

Proof. Suppose not. In particular, there exists a sequence of regressive functions βŸ¨π‘“π‘– : πœ†+ β†’ πœ†+ | 𝑖 < πœ†+⟩and an ordinal 𝛼 ∈ πΈπœ†+

cf(πœ†) with 𝛼 > πœ† such that for every regressive function 𝑓 : 𝛼 β†’ 𝛼, and every

cofinal subset 𝐡 βŠ† 𝛼, there exists some 𝑖 < 𝛼 such that sup{𝛽 ∈ 𝐡 | 𝑓𝑖(𝛽) = 𝑓(𝛽)} = 𝛼.For all 𝛽 < πœ†+, fix a bijection πœ“π›½ : πœ†β†” max{πœ†, 𝛽}. For all 𝑖 < 𝛼, define 𝑔𝑖 : 𝛼→ πœ† by stipulating:

𝑔𝑖(𝛽) := πœ“βˆ’1𝛽 (𝑓𝑖(𝛽)).

Let βŸ¨πœ†π‘— | 𝑗 < cf(πœ†)⟩ be a strictly increasing sequence of ordinals converging to πœ†. Let βŸ¨π›½π‘— | 𝑗 < cf(πœ†)⟩be a strictly increasing sequence of ordinals converging to 𝛼, with 𝛽0 > πœ†. Now, pick some function𝑔 : 𝛼→ πœ† that satisfies for all 𝑗 < cf(πœ†) :

𝑔(𝛽𝑗) := min(πœ† βˆ– {𝑔𝑖(𝛽𝑗) | 𝑖 ∈ πœ“π›Ό[πœ†π‘— ]}).

Define a regressive function 𝑓 : 𝛼→ 𝛼 by stipulating:

𝑓(𝛽) :=

{0, if 𝛽 < πœ†;

πœ“π›½(𝑔(𝛽)), otherwise.

Now, by the choice of βŸ¨π‘“π‘– | 𝑖 < πœ†+⟩, we may find some 𝑖 < 𝛼 such that

sup{𝛽𝑗 | 𝑗 < cf(πœ†), 𝑓𝑖(𝛽𝑗) = 𝑓(𝛽𝑗)} = 𝛼.

In particular, sup{𝑗 < cf(πœ†) | 𝑔𝑖(𝛽𝑗) = 𝑔(𝛽𝑗)} = cf(πœ†). By 𝑖 ∈ πœ“π›Ό[πœ†], let us fix a large enough𝑗* < cf(πœ†) such that 𝑖 ∈ πœ“π›Ό[πœ†π‘—* ]. By definition of 𝑔, then, 𝑔(𝛽𝑗) = 𝑔𝑖(𝛽𝑗) for all 𝑗 ∈ [𝑗*, cf(πœ†)). Thisis a contradiction. οΏ½

Page 7: Introduction - Assaf Rinot

HIGHER SOUSLIN TREES AND THE GCH, REVISITED 7

Corollary 2.9. If non(β„³) > β„΅1, then 𝐽 [β„΅2] is the nonstationary ideal over β„΅2.10

Proof. Towards a contradiction, suppose that 𝐽 [β„΅2] is not the nonstationary ideal over β„΅2. By

Proposition 2.8, there exists a sequence of functions βŸ¨π‘“π‘– : β„΅2 β†’ β„΅2 | 𝑖 < β„΅2⟩ and an ordinal 𝛼 ∈ 𝐸℡2β„΅0

with 𝛼 > πœ”1 such that for every regressive function 𝑓 : 𝛼 β†’ 𝛼, and every cofinal subset 𝐡 βŠ† 𝛼,there exists some 𝑖 < 𝛼 such that sup{𝛽 ∈ 𝐡 | 𝑓𝑖(𝛽) = 𝑓(𝛽)} = 𝛼.

Let βŸ¨π›½π‘› | 𝑛 < πœ”βŸ© be a strictly increasing sequence of ordinals converging to 𝛼, with 𝛽0 > πœ”. Forall 𝑖 < 𝛼, define a real π‘Ÿπ‘– : πœ” β†’ πœ” by stipulating:

π‘Ÿπ‘–(𝑛) :=

{𝑓𝑖(𝛽𝑛), if 𝑓𝑖(𝛽𝑛) < πœ”;

0, otherwise.

By non(β„³) > β„΅1 = |𝛼| and [Bar87, S2], we may pick a real π‘Ÿ : πœ” β†’ πœ” such that for all 𝑖 < 𝛼,π‘Ÿ(𝑛) = π‘Ÿπ‘–(𝑛) for all but finitely many 𝑛 < πœ”. Pick a regressive function 𝑓 : 𝛼 β†’ 𝛼 such that𝑓(𝛽𝑛) = π‘Ÿ(𝑛) for all 𝑛 < πœ”. Put 𝐡 := {𝛽𝑛 | 𝑛 < πœ”}. Then, there exists no 𝑖 < 𝛼 such thatsup{𝛽 ∈ 𝐡 | 𝑓𝑖(𝛽) = 𝑓(𝛽)} = 𝛼. οΏ½

Proposition 2.10. For every inaccessible cardinal πœ…, 𝐽 [πœ…] is the nonstationary ideal over πœ….

Proof. Suppose not. In particular, we may fix an inaccessible cardinal πœ…, a sequence of functionsβŸ¨π‘“π‘– : πœ…β†’ πœ… | 𝑖 < πœ…βŸ©, and an uncountable limit cardinal πœ‡ satisfying the following. For every re-gressive function 𝑓 : πœ‡ β†’ πœ‡, and every cofinal subset 𝐡 βŠ† πœ‡, there exists some 𝑖 < πœ‡ such thatsup{𝛽 ∈ 𝐡 | 𝑓𝑖(𝛽) = 𝑓(𝛽)} = πœ‡.

Let 𝐴 denote the set of cardinals below πœ‡. Pick a regressive function 𝑓 : πœ‡β†’ πœ‡ such that for allπœ† ∈ 𝐴:

𝑓(πœ†+) := min(πœ†+ βˆ– {𝑓𝑖(πœ†+) | 𝑖 ≀ πœ†}).

Put 𝐡 := {πœ†+ | πœ† ∈ 𝐴}. Then 𝐡 is a cofinal subset of πœ‡. For all 𝑖 < πœ‡, we have

sup{𝛽 ∈ 𝐡 | 𝑓𝑖(𝛽) = 𝑓(𝛽)} ≀ |𝑖|+ < πœ‡,

contradicting the choice of βŸ¨π‘“π‘– | 𝑖 < πœ…βŸ©. οΏ½

3. Diamond and club hitting

Definition 3.1 (Jensen, [Jen72]). For a stationary subset 𝑇 of a regular uncountable cardinal πœ…,β™’(𝑇 ) asserts the existence of a sequence βŸ¨π΄π›Ύ | 𝛾 < πœ…βŸ© such that for every 𝐴 βŠ† πœ…, the intersection𝐺(𝐴) ∩ 𝑇 is stationary, where 𝐺(𝐴) := {𝛾 < πœ… | 𝐴 ∩ 𝛾 = 𝐴𝛾}.

Proposition 3.2 (folklore). Suppose that 𝑇 is a stationary subset of a regular uncountable cardinalπœ…, and β™’(𝑇 ) holds. Then there exists a matrix

βŸ¨π΄π‘–

𝛾 | 𝑖 < πœ…, 𝛾 < πœ…βŸ©, such that for every sequence

οΏ½οΏ½ =βŸ¨π΄π‘– | 𝑖 < πœ…

⟩of cofinal subsets of πœ…, the intersection 𝐺(οΏ½οΏ½) ∩ 𝑇 is stationary, where

𝐺(οΏ½οΏ½) := {𝛾 < πœ… | βˆ€π‘– < 𝛾(sup(𝐴𝑖 ∩ 𝛾) = 𝛾 & 𝐴𝑖 ∩ 𝛾 = 𝐴𝑖𝛾)}.

Proof. Let βŸ¨π΄π›Ύ | 𝛾 < πœ…βŸ© be a witness to β™’(𝑇 ). Fix a bijection πœ‹ : πœ… Γ— πœ… ↔ πœ…. For all 𝑖 < πœ… and

𝛾 < πœ…, let 𝐴𝑖𝛾 := {𝛿 < 𝛾 | πœ‹(𝑖, 𝛿) ∈ 𝐴𝛾}. To see that

βŸ¨π΄π‘–

𝛾 | 𝑖 < πœ…, 𝛾 < πœ…βŸ©

works, let οΏ½οΏ½ =βŸ¨π΄π‘– | 𝑖 < πœ…

⟩be an arbitrary sequence of cofinal subsets of πœ…. For all 𝑖 < πœ…, 𝐷𝑖 := {𝛾 < πœ… | sup(𝐴𝑖 ∩ 𝛾) = 𝛾} isa club, and hence 𝐷 := {𝛾 ∈ △𝑖<πœ…π·π‘– | πœ‹[𝛾 Γ— 𝛾] = 𝛾} is a club. Put 𝐴 := {πœ‹(𝑖, 𝛿) | 𝑖 < πœ…, 𝛿 ∈ 𝐴𝑖}.

Then 𝐺 := 𝐺(𝐴) ∩ 𝑇 ∩𝐷 is stationary, and 𝐺 βŠ† 𝐺(οΏ½οΏ½) ∩ 𝑇 , so we are done. οΏ½

Fact 3.3 (Shelah, [She10]). For every uncountable cardinal πœ†, CHπœ† entails β™’(πœ†+).

10Here, β€œnon(β„³) > β„΅1” stands for the assertion that every β„΅1-sized set of reals is meager. This assertion is consistentwith ZFC + Β¬CH, as witnessed by the model of [ST71].

Page 8: Introduction - Assaf Rinot

8 ASSAF RINOT

Definition 3.4. Suppose that 𝑆 is a stationary subset of a regular uncountable cardinal πœ….A sequence βŸ¨πΆπ›Ό | 𝛼 < πœ…βŸ© is said to hit clubs at 𝑆 iff for every club 𝐷 βŠ† πœ…, there exists some 𝛼 ∈ 𝑆

such that sup(nacc(𝐢𝛼) ∩𝐷) = 𝛼.

Proposition 3.5. Suppose that πœ… is a regular cardinal β‰₯ β„΅2.Then οΏ½(πœ…) holds iff for every stationary 𝑆 βŠ† πœ…, there exists a coherent 𝐢-sequence, βŸ¨πΆπ›Ό | 𝛼 < πœ…βŸ©,

that hits clubs at 𝑆.

Proof. (⇐=) Suppose that βŸ¨πΆπ›Ό | 𝛼 < πœ…βŸ© is a coherent 𝐢-sequence that hits clubs at 𝑆 := acc(πœ…).Let 𝐢 be an arbitrary club in πœ…, and we shall find some 𝛼 ∈ acc(𝐢) such that 𝐢𝛼 = 𝐢 ∩ 𝛼.

Consider the club 𝐷 := acc(𝐢). Pick 𝛼 ∈ 𝑆 such that sup(nacc(𝐢𝛼) ∩ 𝐷) = 𝛼. Then 𝛼 ∈acc(𝐷) βŠ† acc(𝐢) and nacc(𝐢𝛼) ∩ acc(𝐢) = βˆ…, so that 𝐢𝛼 = 𝐢 ∩ 𝛼.

(=β‡’) Suppose that οΏ½(πœ…) holds, as witnessed by βŸ¨πΆπ›Ό | 𝛼 < πœ…βŸ©. Let 𝑆 be an arbitrary stationarysubset of πœ…. The proof is a simple combination of ideas from [She94b, S2] and [Rin14a, S3].

For every club 𝐸 βŠ† πœ… and 𝛼 < πœ…, denote

𝐢𝛼[𝐸] := {sup(𝐸 ∩ πœ‚) | πœ‚ ∈ (𝐢𝛼 βˆ– (min(𝐸) + 1))}.Notice that if 𝛼 ∈ acc(𝐸), then 𝐢𝛼[𝐸] is a club in 𝛼, and acc(𝐢𝛼[𝐸]) = acc(𝐢𝛼) ∩ acc(𝐸).

Claim 3.5.1. There exists a club 𝐸 βŠ† πœ… satisfying the following.For every club 𝐷 βŠ† πœ…, there exists some 𝛿 ∈ 𝑆 such that sup(nacc(𝐢𝛿[𝐸]) ∩𝐷) = 𝛿.

Proof. Suppose not. Then, we can recursively construct a βŠ†-decreasing sequence of clubs in πœ…,βŸ¨π·π‘– | 𝑖 < πœ”1⟩, such that 𝐷0 = πœ…, 𝐷𝑖 =

⋂𝑗<𝑖𝐷𝑗 for all nonzero limit 𝑖 < πœ”1, and for all 𝑖 < πœ”1 and

𝛿 ∈ 𝑆,

sup(nacc(𝐢𝛿[𝐷𝑖]) βˆ©π·π‘–+1) < 𝛿.

As πœ… = cf(πœ…) β‰₯ β„΅2, 𝐷 :=β‹‚

𝑖<πœ”1𝐷𝑖 is a club. We claim that 𝐡 := {𝛿 ∈ 𝑆 | sup((𝐷∩ 𝛿) βˆ–πΆπ›Ώ) = 𝛿}

is nonempty.Suppose not. Then there exists some πœ€ < πœ… for which

𝐻 := {𝛿 ∈ 𝑆 | sup((𝐷 ∩ 𝛿) βˆ– 𝐢𝛿) = πœ€} ∩ acc(𝐷 βˆ– πœ€)is stationary in πœ…. Consequently, for every 𝛼 < 𝛿 both in 𝐻, we have 𝛼 ∈ acc(𝐢𝛿) and hence 𝐢𝛼 βŠ‘πΆπ›Ώ. So {𝐢𝛿 | 𝛿 ∈ 𝐻} is an βŠ‘-chain, converging to the club 𝐢 :=

⋃{𝐢𝛿 | 𝛿 ∈ 𝐻}. Let 𝛼 ∈ acc(𝐢) be

arbitrary. Then 𝛼 ∈ acc(𝐢𝛿) for some 𝛿 ∈ 𝐻, and then 𝐢 ∩ 𝛼 = (𝐢 ∩ 𝛿) ∩ 𝛼 = 𝐢𝛿 ∩ 𝛼 = 𝐢𝛼. Thatis, 𝐢 contradicts the fact that βŸ¨πΆπ›Ό | 𝛼 < πœ…βŸ© is a οΏ½(πœ…)-sequence.

Thus, we have established that 𝐡 is nonempty. Pick 𝛿 ∈ 𝐡. For all 𝑖 < πœ”1, by the choice of𝐷𝑖+1, we have that 𝛿𝑖 := sup(nacc(𝐢𝛿[𝐷𝑖]) βˆ©π·π‘–+1) is < 𝛿. We consider two cases:I If cf(𝛿) > β„΅0, then by 𝛿 ∈ 𝐡, let us pick 𝛽 ∈ (π·βˆ©π›Ώ)βˆ–πΆπ›Ώ above sup𝑛<πœ” 𝛿𝑛. Put 𝛾 := min(πΆπ›Ώβˆ–π›½).

Then 𝛿 > 𝛾 > 𝛽, and for all 𝑛 < πœ”, since 𝛽 ∈ 𝐷𝑛, we infer that sup(𝐷𝑛 ∩ 𝛾) β‰₯ 𝛽. So min(𝐢𝛿[𝐷𝑛] βˆ–π›½) = sup(𝐷𝑛 ∩ 𝛾) for all 𝑛 < πœ”. Since {𝐷𝑛 | 𝑛 < πœ”} is a decreasing chain, there exists some 𝑛 < πœ”such that sup(π·π‘›βˆ©π›Ύ) = sup(𝐷𝑛+1βˆ©π›Ύ). Fix such an 𝑛. Then min(𝐢𝛿[𝐷𝑛]βˆ–π›½) = min(𝐢𝛿[𝐷𝑛+1]βˆ–π›½),and in particular, 𝛽* := min(𝐢𝛿[𝐷𝑛] βˆ– 𝛽) is in 𝐷𝑛+1.

βˆ™ If 𝛽* ∈ nacc(𝐢𝛿[𝐷𝑛]), then we get a contradiction to the fact that 𝛽* β‰₯ 𝛽 > 𝛿𝑛.βˆ™ If 𝛽* ∈ acc(𝐢𝛿[𝐷𝑛]), then 𝛽* = 𝛽 and 𝛽* ∈ acc(𝐢𝛿), contradicting the fact that 𝛽 ∈ 𝐢𝛿.

I If cf(𝛿) = β„΅0, then we may pick an uncountable 𝐼 βŠ† πœ”1, for which supπ‘–βˆˆπΌ 𝛿𝑖 < 𝛿.11 By𝛿 ∈ 𝐡, pick 𝛽 ∈ (𝐷 ∩ 𝛿) βˆ– 𝐢𝛿 above supπ‘–βˆˆπΌ 𝛿𝑖, and put 𝛾 := min(𝐢𝛿 βˆ– 𝛽). Then 𝛿 > 𝛾 > 𝛽,and min(𝐢𝛿[𝐷𝑖] βˆ– 𝛽) = sup(𝐷𝑖 ∩ 𝛾) for all 𝑖 < πœ”1. Since {𝐷𝑖 | 𝑖 < πœ”1} is a decreasing chain

11See Fact 1.3 of [Rin10].

Page 9: Introduction - Assaf Rinot

HIGHER SOUSLIN TREES AND THE GCH, REVISITED 9

and 𝐼 is cofinal in πœ”1, we may fix an 𝑖 ∈ 𝐼 such that sup(𝐷𝑖 ∩ 𝛾) = sup(𝐷𝑖+1 ∩ 𝛾). Thenmin(𝐢𝛿[𝐷𝑖] βˆ– 𝛽) = min(𝐢𝛿[𝐷𝑖+1] βˆ– 𝛽), and in particular, 𝛽* := min(𝐢𝛿[𝐷𝑖] βˆ– 𝛽) is in 𝐷𝑖+1.

βˆ™ If 𝛽* ∈ nacc(𝐢𝛿[𝐷𝑖]), then we get a contradiction to the fact that 𝛽* β‰₯ 𝛽 > 𝛿𝑖.βˆ™ If 𝛽* ∈ acc(𝐢𝛿[𝐷𝑖]), then 𝛽* = 𝛽 and 𝛽* ∈ acc(𝐢𝛿), contradicting the fact that 𝛽 ∈ 𝐢𝛿.

οΏ½

Let 𝐸 be given by Claim 3.5.1. For all 𝛼 < πœ…, define

𝐷𝛼 :=

{𝐢𝛼[𝐸], if 𝛼 ∈ acc(𝐸);

𝛼 βˆ– sup(𝐸 ∩ 𝛼), if 𝛼 ∈ acc(𝐸).

Claim 3.5.2. βŸ¨π·π›Ό | 𝛼 < πœ…βŸ© is a coherent 𝐢-sequence.

Proof. Let 𝛼 < πœ… be an arbitrary limit ordinal. It is easy to see that 𝐷𝛼 is a club in 𝛼. Supposethat we are given οΏ½οΏ½ ∈ acc(𝐷𝛼). We shall prove that 𝐷�� = 𝐷𝛼 ∩ οΏ½οΏ½.

Put 𝜏 := sup(𝐸 ∩ 𝛼), and consider two cases:I If 𝜏 < 𝛼, then 𝐷𝛼 is the interval [𝜏, 𝛼), and (𝜏, 𝛼)∩𝐸 = βˆ…. So, οΏ½οΏ½ ∈ (𝜏, 𝛼) and sup(𝐸 ∩ οΏ½οΏ½) = 𝜏 .

Consequently, 𝐷𝛼 ∩ οΏ½οΏ½ = 𝐷��.I If 𝜏 = 𝛼, then 𝐷𝛼 is the club 𝐢𝛼[𝐸], and by οΏ½οΏ½ ∈ acc(𝐷𝛼), we have οΏ½οΏ½ ∈ acc(𝐢𝛼) ∩ acc(𝐸),

meaning that 𝐢�� = 𝐢𝛼 ∩ οΏ½οΏ½ and 𝐷�� = 𝐢��[𝐸]. Consequently, 𝐷𝛼 ∩ οΏ½οΏ½ = 𝐷��. οΏ½

Finally, let 𝐷 be an arbitrary club in πœ…. By the choice of 𝐸, let us pick 𝛼 ∈ 𝑆 such thatsup(nacc(𝐢𝛼[𝐸])∩𝐷) = 𝛼. In particular, 𝐷𝛼 = 𝐢𝛼[𝐸], and sup(nacc(𝐷𝛼)∩𝐷) = 𝛼, as sought. οΏ½

4. Main Results

An examination of all β™’-based constructions of πœ…-Souslin trees in the literature (e.g., [Jen72],[Gre76], [Dev83], [She84], [BDS86], [Dav90], [KS93] ,[Cum97], [Sch05], [KLY07], and [Rin11b])reveals that they all involve an ingredient of sealing antichains at some nonreflecting stationary setof levels of the πœ…-tree. As here, we are interested in deriving more than a Mahlo cardinal, we mustdevise a method of deriving a πœ…-Souslin tree from principles that are compatible with reflection. Ayear ago, in a joint work with Brodsky, we came up with the following candidate:

Definition 4.1 (Brodsky-Rinot, [BR15]). For a regular uncountable cardinal πœ…, and a stationarysubset 𝑆 βŠ† πœ…, οΏ½βˆ’(𝑆) asserts the existence of a coherent 𝐢-sequence, βŸ¨πΆπ›Ό | 𝛼 < πœ…βŸ©, such that forevery cofinal subset 𝐴 βŠ† πœ…, there exists some 𝛼 ∈ 𝑆 for which sup(nacc(𝐢𝛼) ∩𝐴) = 𝛼.

Fact 4.2 (Brodsky-Rinot, [BR15]). Suppose πœ… is a regular uncountable cardinal, and β™’(πœ…) holds.

(a) If οΏ½βˆ’(πœ…) holds, then there exists a club-regressive πœ…-Souslin tree;(b) If οΏ½βˆ’(πΈπœ…

β‰₯πœƒ) holds, and πœ†<πœƒ < πœ… for all πœ† < πœ…, then there exists a πœƒ-complete πœ…-Souslin tree.

By [Rin14b, Theorem A], if πœ† is an uncountable cardinal and CHπœ† holds, then οΏ½πœ† entails οΏ½βˆ’(𝑆)

for every stationary 𝑆 βŠ† πΈπœ†+

=cf(πœ†). One goal of the current section is to relax οΏ½πœ† to οΏ½(πœ†+).12

Theorem 4.3. Suppose that πœ… is a regular uncountable cardinal, and οΏ½(πœ…) + β™’(πœ…) holds.Then οΏ½βˆ’(𝑆) holds for every stationary 𝑆 ∈ 𝐽 [πœ…].

Proof. Suppose that 𝑆 ∈ 𝐽 [πœ…] is a given stationary set. Fix a club 𝐢 βŠ† πœ… and a sequence of regressivefunctions βŸ¨π‘“π‘– : πœ…β†’ πœ… | 𝑖 < πœ…βŸ© as in Definition 2.1. By β™’(πœ…), let us fix a matrix

βŸ¨π΄π‘–

𝛾 | 𝑖 < πœ…, 𝛾 < πœ…βŸ©

as in Proposition 3.2. For all 𝑖, 𝛽 < πœ…, put

𝑋𝑖𝛽 := {𝛽} βˆͺ𝐴𝑖

𝑓𝑖(𝛽).

12In light of Proposition 2.7, we mention that by [She82, p. 440],οΏ½πœ† entails ADSπœ†.

Page 10: Introduction - Assaf Rinot

10 ASSAF RINOT

Since 𝑆 is stationary, we infer from Proposition 2.8 that πœ… β‰₯ β„΅2. So, by οΏ½(πœ…), appeal toProposition 3.5 to obtain a coherent 𝐢-sequence, βŸ¨πΆπ›Ό | 𝛼 < πœ…βŸ©, that hits clubs at 𝑆 ∩ 𝐢. Finally,for all 𝑖, 𝛼 < πœ…, let

𝐢𝑖𝛼 := 𝐢𝛼 βˆͺ {min(𝑋𝑖

𝛽 βˆ– (sup(𝐢𝛼 ∩ 𝛽) + 1)) | 𝛽 ∈ nacc(𝐢𝛼) & 𝛽 > 0}.

Claim 4.3.1. Let 𝑖 < πœ… be arbitrary. ThenβŸ¨πΆπ‘–π›Ό | 𝛼 < πœ…

⟩is a coherent 𝐢-sequence.

Proof. Let 𝛼 < πœ… be an arbitrary limit ordinal. By 𝐢𝛼 βŠ† 𝐢𝑖𝛼, we have sup(𝐢𝑖

𝛼) = 𝛼. For allsuccessive points π›½βˆ’ < 𝛽 from 𝐢𝛼, the relative interval (π›½βˆ’, 𝛽)βˆ©πΆπ‘–

𝛼 contains at most one element,thus, as 𝐢𝛼 is closed below 𝛼, so does 𝐢𝑖

𝛼. Now, suppose that οΏ½οΏ½ ∈ acc(𝐢𝑖𝛼). Then οΏ½οΏ½ ∈ acc(𝐢𝛼), and

hence 𝐢�� = πΆπ›Όβˆ©οΏ½οΏ½. Then the local nature of the definition of 𝐢𝑖𝛼 makes it clear that 𝐢𝑖

π›Όβˆ©οΏ½οΏ½ = 𝐢𝑖��. οΏ½

Claim 4.3.2. There exists an 𝑖 < πœ… for whichβŸ¨πΆπ‘–π›Ό | 𝛼 < πœ…

⟩is a οΏ½βˆ’(𝑆)-sequence.

Proof. Suppose not. It then follows from Claim 4.3.1 that there exists a sequence of cofinal subsets

of πœ…, οΏ½οΏ½ =βŸ¨π΄π‘– | 𝑖 < πœ…

⟩, such that for all 𝑖 < πœ… and 𝛼 ∈ 𝑆, we have sup(nacc(𝐢𝑖

𝛼) ∩ 𝐴𝑖) < 𝛼.

Let 𝐺 be 𝐺(οΏ½οΏ½) as in the statement of Proposition 3.2. Then 𝐺 is a stationary subset of πœ…, and𝐷 := {𝛽 < πœ… | sup(𝐺 ∩ 𝛽) = 𝛽 > 0} is a club in πœ…. As βŸ¨πΆπ›Ό | 𝛼 < πœ…βŸ© hits clubs at 𝑆 ∩ 𝐢, let uspick 𝛼 ∈ 𝑆 ∩ 𝐢 such that sup(nacc(𝐢𝛼) ∩ 𝐷) = 𝛼. Put 𝐡 := nacc(𝐢𝛼) ∩ 𝐷. For all 𝛽 ∈ 𝐡, by𝛽 ∈ 𝐷, we know that the relative interval 𝐺∩ (sup(𝐢𝛼 βˆ©π›½), 𝛽) is nonempty. Consequently, we mayfind some regressive function 𝑓 : 𝛼 β†’ 𝛼 such that 𝑓(𝛽) ∈ 𝐺 ∩ (sup(𝐢𝛼 ∩ 𝛽), 𝛽) for all 𝛽 ∈ 𝐡. Pick𝑖 < 𝛼 and a cofinal subset 𝐡′ βŠ† 𝐡 such that 𝑓𝑖 οΏ½ 𝐡′ = 𝑓 οΏ½ 𝐡′. Fix a large enough πœ‚ ∈ 𝐢𝛼 such thatsup(𝐢𝛼 ∩ πœ‚) β‰₯ 𝑖. By omitting an initial segment, we may assume that 𝐡′ ∩ πœ‚ = βˆ….

Let 𝛽 ∈ 𝐡′ be arbitrary. Write 𝛾 := 𝑓𝑖(𝛽). Then 𝛾 ∈ 𝐺∩ (sup(𝐢𝛼 βˆ©π›½), 𝛽) βŠ† (𝑖, 𝛽). In particular,sup(π΄π‘–βˆ©π›Ύ) = 𝛾 and π΄π‘–βˆ©π›Ύ = 𝐴𝑖

𝛾 , so that 𝑋𝑖𝛽 = {𝛽}βˆͺ(π΄π‘–βˆ©π›Ύ), and min(𝑋𝑖

π›½βˆ–(sup(πΆπ›Όβˆ©π›½)+1)) ∈ π΄π‘–βˆ©π›Ύ.

Thus we have shown that for all 𝛽 ∈ 𝐡′, 𝐢𝑖𝛼 ∩ 𝐴𝑖 ∩ (sup(𝐢𝛼 ∩ 𝛽), 𝛽) is a singleton, contradicting

the fact that sup(nacc(𝐢𝑖𝛼) βˆ©π΄π‘–) < 𝛼 = sup(𝐡′). οΏ½

This completes the proof. οΏ½

Remark. Notice that since 𝐽 [πœ…] is an ideal, the conclusion of the preceding theorem may be strength-ened to: if 𝑆 ∈ 𝐽 [πœ…], then οΏ½βˆ’(𝑇 ) holds for every 𝑇 ∈ 𝒫(πœ…) such that 𝑆 ∩ 𝑇 is stationary. Comparethis with Kunen’s theorem [Kun80] that for 𝑆 βŠ† πœ…, β™’*(𝑆) entails that β™’(𝑇 ) holds for every 𝑇 ∈ 𝒫(πœ…)such that 𝑆 ∩ 𝑇 is stationary.

Corollary 4.4. Suppose that πœƒ < cf(πœ†) ≀ πœ† are cardinals, π’Ÿ(πœ†, πœƒ) = πœ† and 2πœ† = πœ†+.

For every stationary 𝑆 βŠ† πΈπœ†+

πœƒ , οΏ½(πœ†+) entails οΏ½βˆ’(𝑆).

Proof. Suppose that 𝑆 βŠ† πΈπœ†+

πœƒ is a given stationary set. By Proposition 2.2, πΈπœ†+

πœƒ ∈ 𝐽 [πœ†+]. Sincethe latter is an ideal, we get that 𝑆 ∈ 𝐽 [πœ†+]. By CHπœ†, πœ† β‰₯ β„΅1 and Fact 3.3, β™’(πœ†+) holds. So, byTheorem 4.3, οΏ½(πœ†+) entails οΏ½βˆ’(𝑆). οΏ½

Corollary 4.5. Suppose that GCH holds, and πœ† is an uncountable cardinal.

Then οΏ½(πœ†+) entails that οΏ½βˆ’(πΈπœ†+

πœƒ ) holds for every regular cardinal πœƒ < πœ†.

Proof. By Corollary 4.4, οΏ½(πœ†+)+GCH entails οΏ½βˆ’(𝑆) for every stationary 𝑆 βŠ† πΈπœ†+

<cf(πœ†). Thus, from

now on, suppose that πœ† is a singular cardinal, and πœƒ is a regular cardinal in [cf(πœ†), πœ†). If πœƒ = cf(πœ†),then let πœƒβ€² := πœƒ+; otherwise, let πœƒβ€² := πœƒ.

Claim 4.5.1. There exists a stationary 𝑆 βŠ† πΈπœ†+

πœƒβ€² such that 𝑆 ∈ 𝐽 [πœ†+].

Page 11: Introduction - Assaf Rinot

HIGHER SOUSLIN TREES AND THE GCH, REVISITED 11

Proof. By Corollary 2.5, it suffices to prove that π’Ÿ(πœ†, πœƒβ€²) = πœ†. Let Ξ£ : cf(πœ†) β†’ πœ† be an increasing

function whose image is cofinal in πœ†. Put β„± :=⋃{[𝛼]πœƒ

β€² | πœƒβ€² ≀ 𝛼 < πœ†}. By GCH, πœ† is a strong

limit, and hence |β„±| = πœ†. Let 𝐡 ∈ [πœ†]πœƒβ€²

be arbitrary. If there exists an ordinal 𝛼 < πœ† such that|𝐡 ∩ 𝛼| = πœƒβ€², then 𝐡 ∩ 𝛼 is a subset of 𝐡 that belongs to β„± , as sought.

Towards a contradiction, suppose that this is not case, and define β„Ž : cf(πœ†) β†’ πœƒβ€² by stipulatingβ„Ž(𝑗) := |𝐡 ∩Σ(𝑗)|. Since cf(πœ†) < cf(πœƒβ€²), there exists some large enough πœ€ < πœƒβ€² such that Im(β„Ž) βŠ† πœ€.But then |𝐡| ≀ max{cf(πœ†), |πœ€|} < πœƒβ€². This is a contradiction. οΏ½

Let 𝑆 be given by the preceding claim. Suppose that οΏ½(πœ†+) holds. By CHπœ†, πœ† β‰₯ β„΅1 and Fact 3.3,

β™’(πœ†+) holds. So, by Theorem 4.3, οΏ½βˆ’(𝑆) holds, let alone οΏ½βˆ’(πΈπœ†+

πœƒβ€² ). In particular, if πœƒβ€² = πœƒ, thenwe are done.

Suppose now that πœƒβ€² > πœƒ. Let βŸ¨πΆπ›Ό | 𝛼 < πœ†+⟩ be a witness to οΏ½βˆ’(πΈπœ†+

πœƒβ€² ). We claim that the same

sequence witnesses οΏ½βˆ’(πΈπœ†+

πœƒ ). To see this, let 𝐴 βŠ† πœ†+ be an arbitrary cofinal subset of πœ†+. Fix𝛼 ∈ 𝑆 such that sup(nacc(𝐢𝛼)∩𝐴) = 𝛼. As cf(𝛼) > πœƒ, we may pick some οΏ½οΏ½ ∈ acc(𝐢𝛼) of cofinalityπœƒ such sup(nacc(𝐢𝛼) ∩𝐴 ∩ οΏ½οΏ½) = οΏ½οΏ½. Then 𝐢�� = 𝐢𝛼 ∩ οΏ½οΏ½ and sup(nacc(𝐢��) ∩𝐴) = οΏ½οΏ½, as sought. οΏ½

Corollary 4.6. Suppose that πœ† is a successor of a regular cardinal πœƒ.

If every stationary subset of πΈπœ†+

πœƒ reflects, then οΏ½βˆ’(πΈπœ†+

πœƒ ) entails οΏ½βˆ’(πΈπœ†+

πœ† ).

Proof. Suppose that every stationary subset of πΈπœ†+

πœƒ reflects, and that οΏ½οΏ½ = βŸ¨πΆπ›Ό | 𝛼 < πœ†+⟩ is a

witness to οΏ½βˆ’(πΈπœ†+

πœƒ ). We claim that οΏ½οΏ½ is also a witness to οΏ½βˆ’(πΈπœ†+

πœ† ).Let 𝐴 be an arbitrary cofinal subset of πœ†+. First, let us point out that

𝑆 := {οΏ½οΏ½ ∈ πΈπœ†+

πœƒ | sup(nacc(𝐢��) ∩𝐴) = οΏ½οΏ½}is stationary in πœ†+. To see this, notice that given any club 𝐷 in πœ†+, we may find some sparseenough cofinal subset 𝐴′ βŠ† 𝐴 such that for all 𝛾 < 𝛿 both from 𝐴′, the relative interval (𝛾, 𝛿) ∩𝐷is nonempty; then, by the choice of οΏ½οΏ½, we pick οΏ½οΏ½ ∈ πΈπœ†+

πœƒ such that sup(nacc(𝐢��) ∩ 𝐴′) = οΏ½οΏ½, andhence οΏ½οΏ½ ∈ acc(𝐷), so that οΏ½οΏ½ ∈ 𝑆 ∩𝐷.

As every stationary subset of πΈπœ†+

πœƒ reflects, let us pick 𝛼 ∈ πΈπœ†+

πœ† such that 𝑆 ∩ 𝛼 is stationary.Put 𝐡 := 𝑆 ∩ acc(𝐢𝛼). Then 𝐡 is stationary in 𝛼, and for all οΏ½οΏ½ ∈ 𝐡, we have sup(nacc(𝐢𝛼)∩𝐴) β‰₯sup(nacc(𝐢��) ∩𝐴) = οΏ½οΏ½. Consequently, sup(nacc(𝐢𝛼) ∩𝐴) = sup(𝐡) = 𝛼, as sought. οΏ½

Corollary 4.7. For every cardinal πœ† β‰₯ iπœ”, CHπœ† +οΏ½(πœ†+) entails οΏ½βˆ’(πœ†+), and hence the existenceof a club-regressive πœ†+-Souslin tree.

Proof. Suppose that we are given a cardinal πœ† β‰₯ iπœ”. By the main result of [She00] (see also [Koj15,S2]), there exists some regular cardinal πœƒ < iπœ” such that π’Ÿ(πœ†, πœƒ) = πœ†. So, by Corollary 2.5, let us

fix a stationary 𝑆 βŠ† πΈπœ†+

πœƒ which is in 𝐽 [πœ†+]. Assume that CHπœ† + οΏ½(πœ†+) holds. By CHπœ†, πœ† β‰₯ β„΅1

and Fact 3.3, β™’(πœ†+) holds. Then, by Theorem 4.3, οΏ½βˆ’(𝑆) holds, let alone οΏ½βˆ’(πœ†+).By Fact 4.2(a), then, there exists a club-regressive πœ†+-Souslin tree. οΏ½

Corollary 4.8. Suppose that GCH holds, and πœ† is an uncountable cardinal.

(1) If οΏ½(πœ†+) holds, then there exists a club-regressive πœ†+-Souslin tree;(2) If there are no club-regressive πœ†+-Souslin trees, then πœ†+ is weakly compact in 𝐿;(3) If there are no β„΅2-Souslin trees and no β„΅3-Souslin trees, then the Axiom of Determinacy

holds in 𝐿(R).

Proof. (1) By Corollary 4.5, GCH+οΏ½(πœ†+) entails οΏ½βˆ’(πΈπœ†+

πœ” ). In particular, οΏ½βˆ’(πœ†+) holds. By CHπœ†,πœ† β‰₯ β„΅1 and Fact 3.3, β™’(πœ†+) holds. Finally, by Fact 4.2(a), we infer the existence of a club-regressiveπœ†+-Souslin tree.

Page 12: Introduction - Assaf Rinot

12 ASSAF RINOT

(2) By [Tod87], if πœ… is a regular uncountable cardinal and οΏ½(πœ…) fails, then πœ… is weakly compactin 𝐿. Now, appeal to Clause (1).

(3) By [SS14], Β¬οΏ½(πœ”2)+Β¬οΏ½πœ”2+2β„΅1 = β„΅2 implies that 𝐿(R) |= AD. Now, appeal to Clause (1). οΏ½

The last goal of this section is derive Souslin trees which are complete to the maximal possibleextent. For this, we shall be considering the following finer concept:

Definition 4.9. A coherent* 𝐢-sequence (over a regular uncountable cardinal πœ…) is a sequenceβŸ¨πΆπ›Ό | 𝛼 < πœ…βŸ© such that:

(1) for all 𝛼 < πœ…, 𝐢𝛼 βŠ† 𝛼;(2) for all limit 𝛼 < πœ…, 𝐢𝛼 is a club in 𝛼;(3) for all 𝛼 < πœ…, if οΏ½οΏ½ ∈ acc(𝐢𝛼), then sup((πΆοΏ½οΏ½βˆ†πΆπ›Ό) ∩ οΏ½οΏ½) < οΏ½οΏ½.

Definition 4.10. For a regular uncountable cardinal πœ…, and a stationary subset 𝑆 βŠ† πœ…, οΏ½β€²(𝑆)asserts the existence of a coherent* 𝐢-sequence, βŸ¨πΆπ›Ό | 𝛼 < πœ…βŸ©, such that for every cofinal subset𝐴 βŠ† πœ…, there exists some 𝛼 ∈ 𝑆 for which sup(nacc(𝐢𝛼) ∩𝐴) = 𝛼.

The very same construction from the proof of Fact 4.2(b) demonstrates (cf. [BR16, S2]):

Proposition 4.11. Suppose that πœ… is a regular uncountable cardinal, and β™’(πœ…) +οΏ½β€²(πΈπœ…β‰₯πœƒ) holds.

If πœ†<πœƒ < πœ… for all πœ† < πœ…, then there exists a πœƒ-complete πœ…-Souslin tree.

Lemma 4.12. Suppose that πœ… is a regular uncountable cardinal, and β™’(πœ…) +οΏ½β€²(πœ…) holds.Then there exists a partition πœ… = 𝑇0 ⊎ 𝑇1 such that β™’(𝑇0) +οΏ½β€²(𝑇1) holds.

Proof. By β™’(πœ…) and Devlin’s lemma [Dev78], let us fix a partition πœ… = 𝑇0 ⊎ 𝑇1 such that β™’(𝑇0) +β™’(𝑇1) holds. Of course, it suffices to show that there exists some 𝑖 < 2 such that οΏ½β€²(𝑇𝑖) holds.

Towards a contradiction, suppose that this is not the case. Let βŸ¨πΆπ›Ό | 𝛼 < πœ…βŸ© be a οΏ½β€²(πœ…)-sequence.By β™’(πœ…), let us fix a matrix A =

βŸ¨π΄π‘–

𝛾 | 𝑖 < πœ…, 𝛾 < πœ…βŸ©

as in Proposition 3.2.Let 𝑖 < 2 be arbitrary. Put

βˆ™ 𝑋𝑖𝛾 := {𝛾} βˆͺ𝐴𝑖

𝛾 for all 𝛾 < πœ…, and then

βˆ™ 𝐢𝑖𝛼 := 𝐢𝛼 βˆͺ {min(𝑋𝑖

𝛾 βˆ– (sup(𝐢𝛼 ∩ 𝛾) + 1)) | 𝛾 ∈ nacc(𝐢𝛼) & 𝛾 > 0} for all 𝛼 < πœ….

The proof of Claim 4.3.1 makes it clear thatβŸ¨πΆπ‘–π›Ό | 𝛼 < πœ…

⟩is a coherent* 𝐢-sequence. So, since

οΏ½β€²(𝑇𝑖) fails, we may fix a cofinal subset 𝐴𝑖 βŠ† πœ…, such that sup(nacc(𝐢𝑖𝛼) ∩ 𝐴𝑖) < 𝛼 for all 𝛼 ∈ 𝑇𝑖.

Put 𝐺𝑖 := {𝛾 < πœ… | sup(𝐴𝑖 ∩ 𝛾) = 𝛾 & 𝐴𝑖 ∩ 𝛾 = 𝐴𝑖𝛾 & 𝛾 > 0}.

Now, by the choice of A, 𝐺 := 𝐺0 ∩𝐺1 is stationary. As βŸ¨πΆπ›Ό | 𝛼 < πœ…βŸ© is a οΏ½β€²(πœ…)-sequence, let usfix 𝛼 < πœ… such that sup(nacc(𝐢𝛼) ∩𝐺) = 𝛼. Let 𝑖 < 2 be the unique integer such that 𝛼 ∈ 𝑇𝑖. Let𝛾 be an arbitrary element of Ξ“ := nacc(𝐢𝛼)∩𝐺. Then 𝐴𝑖

𝛾 = 𝐴𝑖 ∩ 𝛾 is a cofinal subset of 𝛾, so that

min(π‘‹π‘–π›Ύβˆ–(sup(πΆπ›Όβˆ©π›Ύ)+1)) ∈ π΄π‘–βˆ©π›Ύ. Thus we have shown that for all 𝛾 ∈ Ξ“, 𝐢𝑖

π›Όβˆ©π΄π‘–βˆ©(sup(πΆπ›Όβˆ©π›Ύ), 𝛾)

is a singleton, contradicting the fact that sup(nacc(𝐢𝑖𝛼) βˆ©π΄π‘–) < sup(Ξ“) = 𝛼 ∈ 𝑇𝑖. οΏ½

Theorem 4.13. Suppose that πœ… is a regular cardinal β‰₯ β„΅2, and β™’(πœ…) +οΏ½β€²(πœ…) holds.Then οΏ½β€²(𝑆) holds for every stationary subset 𝑆 βŠ† πœ….

Proof. By Lemma 4.12, let us fix a partition πœ… = 𝑇0βŠŽπ‘‡1 such that β™’(𝑇0)+οΏ½β€²(𝑇1) holds. By β™’(𝑇0),let us fix a matrix M =

βŸ¨π΄π‘–

𝛾 | 𝑖 < πœ…, 𝛾 < πœ…βŸ©

as in Proposition 3.2. Let 𝑆 be an arbitrary stationarysubset of πœ…. Towards showing that οΏ½β€²(𝑆) holds, we prove the following.

Claim 4.13.1. There exist some 𝑖 < πœ”1 and a coherent* 𝐢-sequence, βŸ¨πΆπ›Ό | 𝛼 < πœ…βŸ©, such that forevery subset 𝐴 βŠ† πœ… and every club 𝐷 βŠ† πœ…, there exists some 𝛿 ∈ 𝑆 such that

sup{𝛾 ∈ nacc(𝐢𝛿) ∩𝐷 ∩ 𝑇0 | 𝐴𝑖𝛾 = 𝐴 ∩ 𝛾} = 𝛿.

Page 13: Introduction - Assaf Rinot

HIGHER SOUSLIN TREES AND THE GCH, REVISITED 13

Proof. Suppose not. Building on ideas from [KS93], we shall recursively construct a sequence of

triples ⟨(𝐴𝑖, 𝐷𝑖,βˆ’β†’πΆπ‘–) | 𝑖 < πœ”1⟩, such that for all 𝑖 < πœ”1:

(1) 𝐴𝑖 is a cofinal subset of πœ…;(2) 𝐷𝑖 is a club in πœ…;

(3)βˆ’β†’πΆπ‘– =

βŸ¨πΆπ‘–π›Ώ | 𝛿 < πœ…

⟩is a coherent* 𝐢-sequence. In addition, for all 𝛿 < πœ…:

(a) 𝐢𝑖𝛿+1 = {𝛿} and 𝐢𝑖

𝛿 βŠ† 𝐢𝑖+1𝛿 ;

(b) if 𝛼 ∈ acc(𝐢𝑖𝛿) and πœ‚ ∈ 𝐢𝑖

π›Όβˆ©πΆπ‘–π›Ώ satisfies sup((𝐢𝑖

π›Όβˆ†πΆπ‘–π›Ώ)βˆ©π›Ό) ≀ πœ‚, then sup((𝐢𝑖+1

𝛼 βˆ†πΆπ‘–+1𝛿 )∩

𝛼) ≀ πœ‚;

(c) if 𝑖 is a limit nonzero ordinal, then 𝐢𝑖𝛼 =

⋃𝑗<𝑖𝐢

𝑗𝛼 and acc(𝐢𝑖

𝛿) =⋃

𝑗<𝑖 acc(𝐢𝑗𝛿 ).

Whenever 𝐴𝑖 and 𝐷𝑖 will be defined, we shall also derive the set 𝐺𝑖, by letting:

𝐺𝑖 := {𝛾 ∈ 𝐷𝑖 ∩ 𝑇0 | 𝐴𝑖𝛾 = 𝐴𝑖 ∩ 𝛾}.

Here comes the recursion:

I Put 𝐴0 := πœ… and 𝐷0 := πœ…. Letβˆ’β†’πΆ0 =

⟨𝐢0𝛿 | 𝛿 < πœ…

⟩be some witness to οΏ½β€²(𝑇1). Of course, we

may also assume that 𝐢0𝛿+1 = {𝛿} for all 𝛿 < πœ….

I Suppose that 𝑖 < πœ”1, and ⟨(𝐴𝑗 , 𝐷𝑗 ,βˆ’β†’πΆπ‘—) | 𝑗 ≀ π‘–βŸ© has already been defined. By the indirect

hypothesis, let us fix some cofinal subset 𝐴𝑖+1 βŠ† πœ… and some club 𝐷𝑖+1 βŠ† πœ… such that for all 𝛿 ∈ 𝑆,we have:

sup(nacc(𝐢𝑖𝛿) βˆ©πΊπ‘–+1) < 𝛿.

Next, defineβˆ’βˆ’β†’πΆπ‘–+1 =

βŸ¨πΆπ‘–+1𝛿 | 𝛿 < πœ…

⟩by recursion on 𝛿 < πœ…, as follows. Let 𝐢𝑖+1

0 := βˆ…, and

𝐢𝑖+1𝛿+1 := {𝛿} for all 𝛿 < πœ…. Now, if 𝛿 < πœ… is a nonzero limit ordinal and

βŸ¨πΆπ‘–+1𝛾 | 𝛾 < 𝛿

⟩has already

been defined, let:

(⋆)𝑖 𝐢𝑖+1𝛿 := 𝐢𝑖

𝛿 βˆͺ⋃

{𝐢𝑖+1𝛾 βˆ– sup(𝐢𝑖

𝛿 ∩ 𝛾) | 𝛾 ∈ nacc(𝐢𝑖𝛿) βˆ–πΊπ‘–+1}.

It is easy to see that Clauses (3)(a) and (3)(b) hold for 𝑖. So, asβˆ’β†’πΆπ‘– is a coherent* 𝐢-sequence,

we get thatβˆ’βˆ’β†’πΆπ‘–+1 is a coherent* 𝐢-sequence.

I Suppose that 𝑖 < πœ”1 is a nonzero limit ordinal, and ⟨(𝐴𝑗 , 𝐷𝑗 ,βˆ’β†’πΆπ‘—) | 𝑗 < π‘–βŸ© has already been

defined. Put 𝐴𝑖 := πœ… and 𝐷𝑖 := πœ…. For all 𝛿 < πœ…, let 𝐢𝑖𝛿 :=

⋃𝑗<𝑖𝐢

𝑗𝛿 . As Clause (3) holds for

all 𝑗 < 𝑖, to see thatβˆ’β†’πΆπ‘– =

βŸ¨πΆπ‘–π›Ώ | 𝛿 < πœ…

⟩is a coherent* 𝐢-sequence, it suffice to show that for all

𝛼 < 𝛿 < πœ… such that sup(𝐢𝑖𝛿 ∩ 𝛼) = 𝛼 > 0, there exists some 𝑗 < 𝑖 such that 𝛼 ∈ acc(𝐢𝑗

𝛿 ).

Suppose that 𝛼 < 𝛿 are as above. By Clause (3)(a), ⟨min(𝐢𝑗𝛿 βˆ– 𝛼) | 𝑗 < π‘–βŸ© is a weakly decreasing

sequence of ordinals, and hence must stabilize at some 𝑗* < 𝑖. Denote 𝛾 := min(𝐢𝑗*

𝛿 βˆ– 𝛼) and

π›Ύβˆ’ := sup(𝐢𝑗*

𝛿 ∩ 𝛾). If 𝛾 βˆˆβ‹‚{nacc(𝐢𝑗

𝛿 ) | 𝑗* ≀ 𝑗 < 𝑖}, then we get by induction on 𝑗 ∈ [𝑗*, 𝑖) β€”

using (⋆)𝑗 at successor stages and Clause (3)(c) at limit stages β€” that 𝐢𝑗𝛿 ∩ (π›Ύβˆ’, 𝛾] = {𝛾} for all

𝑗 ∈ [𝑗*, 𝑖), contradicting the fact that π›Ύβˆ’ < sup(𝐢𝑖𝛿 βˆ©π›Ό) = 𝛼 ≀ 𝛾.13 Thus, pick 𝑗 ∈ [𝑗*, 𝑖) such that

𝛾 ∈ acc(𝐢𝑗𝛿 ). As 𝛾 = min(𝐢𝑗

𝛿 βˆ– 𝛼), this means that 𝛼 = 𝛾, and hence 𝛼 ∈ acc(𝐢𝑗𝛿 ), as sought.

At the end of the above process, we have obtained ⟨(𝐴𝑖, 𝐷𝑖,βˆ’β†’πΆπ‘–) | 𝑖 < πœ”1⟩. Put 𝐺 :=

⋂𝑖<πœ”1

𝐺𝑖. By

the choice of the matrix M, we know that 𝐺 is a stationary subset of 𝑇0. Asβˆ’β†’πΆ0 is a οΏ½β€²(𝑇1)-sequence,

13Note that 𝛾 ∈ nacc(𝐢𝑗*

𝛿 ) =β‡’ π›Ύβˆ’ < 𝛼.

Page 14: Introduction - Assaf Rinot

14 ASSAF RINOT

for every 𝛽 < πœ…, there exists some 𝛼 ∈ 𝑇1 such that sup(nacc(𝐢0𝛼) ∩ (𝐺 βˆ– 𝛽)) = 𝛼. Consequently,

the following set is cofinal in πœ…:

𝐴 := {𝛼 ∈ 𝑇1 | sup(nacc(𝐢0𝛼) ∩𝐺) = 𝛼}.

Let 𝛼 ∈ 𝐴 be arbitrary. Trivially, the set Γ𝛼 := nacc(𝐢0𝛼) ∩ 𝐺 is cofinal in 𝛼. Suppose that

𝑖 < πœ”1, and Γ𝛼 βŠ† nacc(𝐢𝑖𝛼). Then, by Γ𝛼 βŠ† 𝐺 βŠ† 𝐺𝑖+1 and (⋆)𝑖, we also have Γ𝛼 βŠ† nacc(𝐢𝑖+1

𝛼 ).Recalling Clauses (3)(a) and (3)(c), we altogether get that for all 𝑖 < πœ”:

(⋆⋆)𝑖 Γ𝛼 βŠ† nacc(𝐢𝑖𝛼) βˆ©πΊπ‘–+1.

As 𝑆 is stationary in πœ… = sup(𝐴), we now pick 𝛿 ∈ 𝑆 such that sup(𝐴 ∩ 𝛿) = 𝛿. For all 𝑖 < πœ”1,by the choice of the pair (𝐴𝑖+1, 𝐷𝑖+1), the following ordinal is strictly smaller than 𝛿:

𝛿𝑖 := sup(nacc(𝐢𝑖𝛿) ∩ 𝐺𝑖+1).

At this stage, the analysis splits into two main cases and a few subcases:

Case 1. cf(𝛿) > β„΅0: In this case, 𝛿* := sup𝑖<πœ” 𝛿𝑖 is below 𝛿, so let us fix some 𝛼 ∈ 𝐴∩ (𝛿*, 𝛿).

Denote 𝛾𝑖 := min(𝐢𝑖𝛿 βˆ– 𝛼) for all 𝑖 < πœ”. By Clause (3)(a), βŸ¨π›Ύπ‘– | 𝑖 < πœ”βŸ© is weakly decreasing,

so let us fix some 𝑖 < πœ” such that 𝛾𝑖 = 𝛾𝑖+1. Now, there are three cases to consider, each ofwhich yielding a contradiction:Case 1.1. 𝛾𝑖 > 𝛼: By 𝛾𝑖 = min(𝐢𝑖

𝛿 βˆ–π›Ό) and 𝛾𝑖 > 𝛼, we have 𝛾𝑖 ∈ nacc(𝐢𝑖𝛿). By 𝛾𝑖 > 𝛼 >

𝛿* β‰₯ 𝛿𝑖, we moreover have 𝛾𝑖 ∈ nacc(𝐢𝑖𝛿) βˆ–πΊπ‘–+1. Denote π›Ύβˆ’π‘– := sup(𝐢𝑖

𝛿 ∩ 𝛾𝑖). By (⋆)𝑖,we have

𝐢𝑖+1𝛿 ∩ [π›Ύβˆ’π‘– , 𝛾𝑖) = 𝐢𝑖+1

𝛾𝑖 ∩ [π›Ύβˆ’π‘– , 𝛾𝑖).

By 𝛾𝑖 = min(𝐢𝑖𝛿 βˆ– 𝛼) and 𝛾𝑖 > 𝛼, we have π›Ύβˆ’π‘– < 𝛼, and hence

𝐢𝑖+1𝛿 ∩ [𝛼, 𝛾𝑖) = 𝐢𝑖+1

𝛾𝑖 ∩ [𝛼, 𝛾𝑖).

So 𝛾𝑖+1 = min(𝐢𝑖+1𝛿 βˆ– 𝛼) = min(𝐢𝑖+1

𝛾𝑖 βˆ– 𝛼) < 𝛾𝑖, contradicting the choice of 𝑖.

Case 1.2. 𝛾𝑖 = 𝛼 ∈ acc(𝐢𝑖𝛿): In this case, sup((𝐢𝑖

π›Ώβˆ†πΆπ‘–π›Ό)βˆ©π›Ό) < 𝛼 and hence by (⋆⋆)𝑖, we

get that sup(nacc(𝐢𝑖𝛿) βˆ©πΊπ‘–+1) β‰₯ 𝛼, contradicting the fact that 𝛼 > 𝛿* β‰₯ 𝛿𝑖.

Case 1.3. 𝛾𝑖 = 𝛼 ∈ nacc(𝐢𝑖𝛿): Write π›Όβˆ’ := sup(𝐢𝑖

𝛿 βˆ©π›Ό). As 𝛼 ∈ 𝐴 βŠ† 𝑇1 and 𝐺𝑖+1 βŠ† 𝑇0,

we get from (⋆)𝑖 that 𝐢𝑖+1𝛿 ∩[π›Όβˆ’, 𝛼) = 𝐢𝑖+1

𝛼 ∩[π›Όβˆ’, 𝛼). In particular, by (⋆⋆)𝑖+1, we have

Ξ“π›Όβˆ© [π›Όβˆ’, 𝛼) βŠ† nacc(𝐢𝑖+1𝛿 )βˆ©πΊπ‘–+2. But then sup(nacc(𝐢𝑖+1

𝛿 )βˆ©πΊπ‘–+2) β‰₯ 𝛼, contradictingthe fact that 𝛼 > 𝛿* β‰₯ 𝛿𝑖+1.

Case 2. cf(𝛿) = β„΅0: Pick an uncountable 𝐼 βŠ† πœ”1 and some 𝛿* < 𝛿 for which supπ‘–βˆˆπΌ max{𝛿𝑖, 𝛿𝑖+1} ≀𝛿*. Pick 𝛼 ∈ 𝐴 ∩ (𝛿*, 𝛿), and denote 𝛾𝑖 := min(𝐢𝑖

𝛿 βˆ– 𝛼) for all 𝑖 < πœ”1. By Clause (3)(a) andsince 𝐼 is cofinal in πœ”1, let us fix some 𝑖 ∈ 𝐼 such that 𝛾𝑖 = 𝛾𝑖+1. As before, there are threecases to consider:Case 2.1. 𝛾𝑖 > 𝛼: By 𝛾𝑖 > 𝛼 > 𝛿* β‰₯ 𝛿𝑖, we have 𝛾𝑖 ∈ nacc(𝐢𝑖

𝛿) βˆ– 𝐺𝑖+1. Then, by (⋆)𝑖,

we have 𝐢𝑖+1𝛿 ∩ [𝛼, 𝛾𝑖) = 𝐢𝑖+1

𝛾𝑖 ∩ [𝛼, 𝛾𝑖). So 𝛾𝑖+1 = min(𝐢𝑖+1𝛿 βˆ–π›Ό) = min(𝐢𝑖+1

𝛾𝑖 βˆ–π›Ό) < 𝛾𝑖,contradicting the choice of 𝑖.

Case 2.2. 𝛾𝑖 = 𝛼 ∈ acc(𝐢𝑖𝛿): In this case, sup((𝐢𝑖

π›Ώβˆ†πΆπ‘–π›Ό)βˆ©π›Ό) < 𝛼 and hence by (⋆⋆)𝑖, we

get that sup(nacc(𝐢𝑖𝛿) βˆ©πΊπ‘–+1) β‰₯ 𝛼, contradicting the fact that 𝛼 > 𝛿* β‰₯ 𝛿𝑖.

Case 2.3. 𝛾𝑖 = 𝛼 ∈ nacc(𝐢𝑖𝛿): Write π›Όβˆ’ := sup(𝐢𝑖

𝛿 ∩ 𝛼). As 𝛼 ∈ 𝐺𝑖+1, we get from (⋆)𝑖

that 𝐢𝑖+1𝛿 ∩[π›Όβˆ’, 𝛼) = 𝐢𝑖+1

𝛼 ∩[π›Όβˆ’, 𝛼). In particular, by (⋆⋆)𝑖+1, we have sup(nacc(𝐢𝑖+1𝛿 )∩

𝐺𝑖+2) β‰₯ 𝛼, contradicting the fact that 𝛼 > 𝛿* β‰₯ 𝛿𝑖+1. οΏ½

Fix a coherent* 𝐢-sequence βŸ¨πΆπ›Ό | 𝛼 < πœ…βŸ© and some 𝑖 < πœ”1 as in Claim 4.13.1. Put:

Page 15: Introduction - Assaf Rinot

HIGHER SOUSLIN TREES AND THE GCH, REVISITED 15

βˆ™ π‘Œπ›Ύ := {𝛾} βˆͺ𝐴𝑖𝛾 for all 𝛾 < πœ…, and then

βˆ™ 𝐷𝛼 := 𝐢𝛼 βˆͺ {min(π‘Œπ›Ύ βˆ– (sup(𝐢𝛼 ∩ 𝛾) + 1)) | 𝛾 ∈ nacc(𝐢𝛼) & 𝛾 > 0} for all 𝛼 < πœ….

The proof of Claim 4.3.1 makes it clear that βŸ¨π·π›Ό | 𝛼 < πœ…βŸ© is a coherent* 𝐢-sequence. Finally, let 𝐴be an arbitrary cofinal subset of πœ…. Put 𝐷 := {𝛾 < πœ… | sup(𝐴 ∩ 𝛾) = 𝛾 > 0}. Pick 𝛿 ∈ 𝑆 such that

Ξ“ := {𝛾 ∈ nacc(𝐢𝛿) ∩𝐷 ∩ 𝑇0 | 𝐴𝑖𝛾 = 𝐴 ∩ 𝛾}

is cofinal in 𝛿.Let 𝛾 be an arbitrary element of Ξ“. Then 𝐴𝑖

𝛾 = 𝐴 ∩ 𝛾 is a cofinal subset of 𝛾, so that min(π‘Œπ›Ύ βˆ–(sup(𝐢𝛿 ∩ 𝛾) + 1)) ∈ 𝐴 ∩ 𝛾. Thus we have shown that for all 𝛾 ∈ Ξ“, 𝐷𝛿 ∩ 𝐴 ∩ (sup(𝐢𝛿 ∩ 𝛾), 𝛾) is asingleton, and hence sup(nacc(𝐷𝛿) ∩𝐴) = 𝛿, as sought. οΏ½

Corollary 4.14. For every uncountable cardinal πœ†, GCH +οΏ½(πœ†+) entails the existence of a cf(πœ†)-complete πœ†+-Souslin tree.

Proof. By Corollary 4.5, GCH + οΏ½(πœ†+) entails οΏ½βˆ’(πΈπœ†+

πœ” ). In particular, οΏ½β€²(πœ†+) holds, and then,

by Theorem 4.13, οΏ½β€²(πΈπœ†+

cf(πœ†)) holds. By CHπœ†, πœ† β‰₯ β„΅1 and Fact 3.3, β™’(πœ†+) holds. Finally, by GCH

and Proposition 4.11, we infer the existence of a cf(πœ†)-complete πœ†+-Souslin tree. οΏ½

Remark. The preceding provides an affirmative answer to Question 9 from the survey paper[Rin11a].

Acknowledgements

I thank A. M. Brodsky, M. Kojman, M. Gitik, C. Lambie-Hanson and D. Raghavan for theirinput that improved the exposition of this paper. Special thanks go to Y. Hayut for proofreading apreliminary version of this paper, and the anonymous referee for a thorough reading of this paper.

The results of this paper were presented at the Mini-Symposia of the 7π‘‘β„Ž European Congressof Mathematics, Berlin, July 2016, and the Set Theory and its Applications in Topology meeting,Oaxaca, September 2016. I thank the organizers for the invitations, and the participants for theirfeedback.

This work was partially supported by the Israel Science Foundation (grant #1630/14).

References

[Bar87] Tomek Bartoszynski. Combinatorial aspects of measure and category. Fund. Math., 127(3):225–239, 1987.[Bau76a] James E. Baumgartner. Almost-disjoint sets, the dense set problem and the partition calculus. Ann. Math.

Logic, 9(4):401–439, 1976.[Bau76b] James E. Baumgartner. A new class of order types. Ann. Math. Logic, 9(3):187–222, 1976.[BDS86] Shai Ben-David and Saharon Shelah. Souslin trees and successors of singular cardinals. Ann. Pure Appl.

Logic, 30(3):207–217, 1986.[BR15] Ari Meir Brodsky and Assaf Rinot. A microscopic approach to Souslin-tree constructions. Part I. arXiv

preprint arXiv:1601.01821, 2015. submitted December 2015.[BR16] Ari Meir Brodsky and Assaf Rinot. More notions of forcing add a Souslin tree. arXiv preprint

arXiv:1607.07033, 2016. submitted July 2016.[CFM01] James Cummings, Matthew Foreman, and Menachem Magidor. Squares, scales and stationary reflection.

J. Math. Log., 1(1):35–98, 2001.[CFM04] James Cummings, Matthew Foreman, and Menachem Magidor. Canonical structure in the universe of set

theory. I. Ann. Pure Appl. Logic, 129(1-3):211–243, 2004.[Cum97] James Cummings. Souslin trees which are hard to specialise. Proc. Amer. Math. Soc., 125(8):2435–2441,

1997.[Dav90] R. David. Some results on higher Suslin trees. J. Symbolic Logic, 55(2):526–536, 1990.[Dev78] Keith J. Devlin. A note on the combinatorial principles β™’(𝐸). Proc. Amer. Math. Soc., 72(1):163–165,

1978.

Page 16: Introduction - Assaf Rinot

16 ASSAF RINOT

[Dev83] Keith J. Devlin. Reduced powers of β„΅2-trees. Fund. Math., 118(2):129–134, 1983.[DJ74] Keith J. Devlin and Havard Johnsbraten. The Souslin problem. Lecture Notes in Mathematics, Vol. 405.

Springer-Verlag, Berlin, 1974.[Eis10] Todd Eisworth. Successors of singular cardinals. In Handbook of set theory. Vols. 1, 2, 3, pages 1229–1350.

Springer, Dordrecht, 2010.[FM97] Matthew Foreman and Menachem Magidor. A very weak square principle. J. Symbolic Logic, 62(1):175–196,

1997.[Gre76] John Gregory. Higher Souslin trees and the generalized continuum hypothesis. J. Symbolic Logic, 41(3):663–

671, 1976.[GS08] Moti Gitik and Assaf Sharon. On SCH and the approachability property. Proc. Amer. Math. Soc.,

136(1):311–320 (electronic), 2008.[Han64] W. Hanf. Incompactness in languages with infinitely long expressions. Fund. Math., 53:309–324, 1963/1964.[HS85] Leo Harrington and Saharon Shelah. Some exact equiconsistency results in set theory. Notre Dame J.

Formal Logic, 26(2):178–188, 1985.[Jec67] Tomas Jech. Non-provability of Souslin’s hypothesis. Comment. Math. Univ. Carolinae, 8:291–305, 1967.[Jen68] Ronald B Jensen. Souslin’s hypothesis is incompatible with V=L. Notices Amer. Math. Soc, 15(6), 1968.[Jen72] R. Bjorn Jensen. The fine structure of the constructible hierarchy. Ann. Math. Logic, 4:229–308; erratum,

ibid. 4 (1972), 443, 1972. With a section by Jack Silver.[KLY07] Bernhard Konig, Paul Larson, and Yasuo Yoshinobu. Guessing clubs in the generalized club filter. Fund.

Math., 195(2):177–191, 2007.[KM78] A. Kanamori and M. Magidor. The evolution of large cardinal axioms in set theory. In Higher set theory

(Proc. Conf., Math. Forschungsinst., Oberwolfach, 1977), volume 669 of Lecture Notes in Math., pages99–275. Springer, Berlin, 1978.

[Koj15] Menachem Kojman. Splitting families of sets in ZFC. Adv. Math., 269:707–725, 2015.[KS93] Menachem Kojman and Saharon Shelah. πœ‡-complete Souslin trees on πœ‡+. Arch. Math. Logic, 32(3):195–201,

1993.[Kun80] Kenneth Kunen. Set theory, volume 102 of Studies in Logic and the Foundations of Mathematics. North-

Holland Publishing Co., Amsterdam-New York, 1980. An introduction to independence proofs.[Kur35] Duro Kurepa. Ensembles ordonnes et ramifies. Publications de l’Institut Mathematique Beograd, 4:1–138,

1935.[LS81] Richard Laver and Saharon Shelah. The β„΅2-Souslin hypothesis. Trans. Amer. Math. Soc., 264(2):411–417,

1981.[Mag82] Menachem Magidor. Reflecting stationary sets. J. Symbolic Logic, 47(4):755–771 (1983), 1982.[Mit73] William Mitchell. Aronszajn trees and the independence of the transfer property. Ann. Math. Logic, 5:21–46,

1972/73.[Rin10] Assaf Rinot. A relative of the approachability ideal, diamond and non-saturation. J. Symbolic Logic,

75(3):1035–1065, 2010.[Rin11a] Assaf Rinot. Jensen’s diamond principle and its relative. In Set theory and its applications, volume 533 of

Contemp. Math., pages 125–156. Amer. Math. Soc., Providence, RI, 2011.[Rin11b] Assaf Rinot. On guessing generalized clubs at the successors of regulars. Ann. Pure Appl. Logic, 162(7):566–

577, 2011.[Rin14a] Assaf Rinot. Chain conditions of products, and weakly compact cardinals. Bull. Symb. Log., 20(3):293–314,

2014.[Rin14b] Assaf Rinot. The Ostaszewski square and homogeneous Souslin trees. Israel J. Math., 199(2):975–1012,

2014.[Sch05] Ernest Schimmerling. A question about Suslin trees and the weak square hierarchy. Notre Dame J. Formal

Logic, 46(3):373–374 (electronic), 2005.[She79] Saharon Shelah. On successors of singular cardinals. In Logic Colloquium ’78 (Mons, 1978), volume 97 of

Stud. Logic Foundations Math., pages 357–380. North-Holland, Amsterdam-New York, 1979.[She81] Saharon Shelah. Models with second order properties. III. Omitting types for 𝐿(𝑄). Arch. Math. Logik

Grundlag., 21(1-2):1–11, 1981.[She82] Saharon Shelah. Proper forcing, volume 940 of Lecture Notes in Mathematics. Springer-Verlag, Berlin-New

York, 1982.[She84] Saharon Shelah. An β„΅2 Souslin tree from a strange hypothesis. Abs. Amer. Math. Soc., 160:198, 1984.

Page 17: Introduction - Assaf Rinot

HIGHER SOUSLIN TREES AND THE GCH, REVISITED 17

[She93] Saharon Shelah. Advances in cardinal arithmetic. In Finite and infinite combinatorics in sets and logic(Banff, AB, 1991), volume 411 of NATO Adv. Sci. Inst. Ser. C Math. Phys. Sci., pages 355–383. KluwerAcad. Publ., Dordrecht, 1993.

[She94a] Saharon Shelah. β„΅πœ”+1 has a Jonsson Algebra. In Cardinal Arithmetic, volume 29 of Oxford Logic Guides.Oxford University Press, 1994.

[She94b] Saharon Shelah. There are Jonsson algebras in many inaccessible cardinals. In Cardinal Arithmetic, vol-ume 29 of Oxford Logic Guides. Oxford University Press, 1994.

[She00] Saharon Shelah. The generalized continuum hypothesis revisited. Israel J. Math., 116:285–321, 2000.[She10] Saharon Shelah. Diamonds. Proc. Amer. Math. Soc., 138(6):2151–2161, 2010.[Sou20] Mikhail Yakovlevich Souslin. Probleme 3. Fundamenta Mathematicae, 1(1):223, 1920.[Spe49] E. Specker. Sur un probleme de Sikorski. Colloquium Math., 2:9–12, 1949.[SS82a] S. Shelah and L. Stanley. Generalized Martin’s axiom and Souslin’s hypothesis for higher cardinals. Israel

J. Math., 43(3):225–236, 1982.[SS82b] S. Shelah and L. Stanley. 𝑠-forcing. i. a β€œblack-box” theorem for morasses, with applications to super-souslin

trees. Israel J. Math., 43(3):185–224, 1982.[SS88] Saharon Shelah and Lee Stanley. Weakly compact cardinals and nonspecial Aronszajn trees. Proc. Amer.

Math. Soc., 104(3):887–897, 1988.[SS14] Ralf Schindler and John Steel. The core model induction. 2014. unpublished.[ST71] R. M. Solovay and S. Tennenbaum. Iterated Cohen extensions and Souslin’s problem. Ann. of Math. (2),

94:201–245, 1971.[Ste05] John R. Steel. PFA implies AD𝐿(R). J. Symbolic Logic, 70(4):1255–1296, 2005.[Ten68] S. Tennenbaum. Souslin’s problem. Proc. Nat. Acad. Sci. U.S.A., 59:60–63, 1968.[Tod81] Stevo B. Todorcevic. Trees, subtrees and order types. Ann. Math. Logic, 20(3):233–268, 1981.[Tod87] Stevo Todorcevic. Partitioning pairs of countable ordinals. Acta Math., 159(3-4):261–294, 1987.[Tod89] Stevo Todorcevic. Special square sequences. Proc. Amer. Math. Soc., 105(1):199–205, 1989.

Department of Mathematics, Bar-Ilan University, Ramat-Gan 52900, Israel.URL: http://www.assafrinot.com