The Bounded Proper Forcing Axiom Revised version, January 1994 5 Martin GOLDSTERN1 9 9 Free University of Berlin 1 n a Saharon SHELAH1,2 J 5 Hebrew University of Jerusalem 1 ] O L ABSTRACT. The bounded proper forcing axiom BPFA is the . h t statement that for any family of ℵ many maximal antichains a 1 m of a proper forcing notion, each of size ℵ , there is a directed 1 [ set meeting all these antichains. 1 v A regular cardinal κ is called Σ -reflecting, if for any regular 2 1 2 2 cardinal χ, for all formulas ϕ, “H(χ) |= ‘ϕ’” implies “∃δ<κ, 1 0 H(δ) |= ‘ϕ’” 5 9 We show that BPFA is equivalent to the statement that two / h nonisomorphic models of size ℵ cannot be made isomorphic t 1 a m by a proper forcing notion, and we show that the consistency : v strength of the bounded proper forcing axiom is exactly the i X existence of a Σ -reflecting cardinal (which is less than the ex- r 1 a istence of a Mahlo cardinal). Wealso show that thequestionofthe existenceofisomorphisms between two structures can be reduced to the question of rigid- ity of a structure. 1 The authors thank the DFG (grant Ko 490/7-1) and the Edmund Landau Center for research in Mathematical Analysis, supported by the Minerva Foundation (Germany) 2 Publication 507 Goldstern–Shelah: Forcing Axioms with Small Antichains Introduction The proper forcing axiom has been successfully employed to decide many questions in set- theoretic topology and infinite combinatorics. See [Ba1] for some applications, and [Sh b] and [Sh f] for variants. In the recent paper [Fu], Fuchino investigated the following two consequences of the proper forcing axiom: (a) If a structure A of size ℵ cannot be embedded into a structure B, then such 1 an embedding cannot be produced by a proper forcing notion. (b) If two structures A and B are not isomorphic, then they cannot be made isomorphic by a proper forcing notion. He showed that (a) is in fact equivalent to the proper forcing axiom, and asked if the same is true for (b). In this paper we find a natural weakening of the proper forcing axiom, the “bounded” proper forcing axiom and show that it is equivalent to property (b) above. We then investigate the consistency strength of this new axiom. While the exact consis- tency strengthoftheproperforcing axiomisstillunknown(but large, see[To]),itturnsout that the bounded proper forcing axiom is equiconsistent to a rather small large cardinal. For notational simplicity we will, for the moment, only consider forcing notions which are complete Boolean algebras. See 0.4 and 4.6. We begin by recalling the forcing axiom in its usual form: For a forcing notion P, FA(P,κ) is the following statement: 1 Goldstern–Shelah: Forcing Axioms with Small Antichains Whenever hA : i < κi is a family of maximal antichains of P, then there is a i filter G∗ ⊆ P meeting all A . i If f is a P-name for a function from κ to the ordinals, we will say that G∗ ⊆ P decides f if foreeach i < κ there is a condition p ∈ G∗ and an ordinal α such that p k− f(i) = α .e(If i i G∗ is directed, then this ordinal must be unique, and we will write f[G∗] forethe function i 7→ α .) Now it is easy to see that the FA(P,κ) is equivalent to theefollowing statement: i Whenever f is a P-name for a function from κ to the ordinals, then there is a filter G∗ ⊆eP which decides f. e This characterization suggests the following weakening of the forcing axiom: 0.1 Definition: Let P be a forcing notion, and let κ and λ be infinite cardinals. BFA(P,κ,λ)isthefollowing statement: Whenever f isa P-nameforafunction from κ to λ then there is a filter G∗ ⊆ P which deceides f, or equivalently: Whenever hA : i < κi is a family of maximal antichains eof P, each of size ≤ λ, i then there is a filter G ⊆ P which meets all A . i 0.2 Notation: (1) BFA(P,λ) is BFA(P,λ,λ), and BFA(P) is BFA(P,ω ). 1 (2) IfE isaclassorpropertyofforcingnotions,wewriteBFA(E)for∀P∈E BFA(P), etc. (3) BPFA = the bounded proper forcing axiom = BFA(proper). 0.3 Remark: For the class of ccc forcing notions we get nothing new: BFA(ccc,λ) is equivalent to Martin’s axiom MA(λ), i.e., FA(ccc,λ). ·–i· 0.3 0.4 Remark: If the forcing notion P is not a complete Boolean algebra but an arbitrary 2 Goldstern–Shelah: Forcing Axioms with Small Antichains poset, then it is possible that P does not have any small antichains, so it could satisfy the second version of BFA(P) vacuously. The problem with the first definition, when applied to an arbitrary poset, is that a filter on ro(P) which interprets the P-name (=ro(P)-name) f does not necessarily generate a filter on P. So for the moment our official definition of eBFA(P) for arbitrary posets P will be BFA(P) : ⇐⇒ BFA(ro(P)) In 4.4 and 4.5 we will find a equivalent (and more natural?) definition BFA′(P) which does not explicitly refer to ro(P). Contents of the paper: in section 1 we show that the “bounded forcing axiom” for any forcingnotionP isequivalenttoFuchino’s“potentialisomorphism”axiomforP. Insection 2 we define the concept of a Σ -reflecting cardinal, and we show that from a model with 1 such a cardinal we can produce a model for the bounded proper forcing axiom. In section 3 we describe a (known) forcing notion which we will use in section 4, where we complement our consistency result by showing that a Σ -reflecting cardinal is necessary: If BPFA holds, 1 then ℵ must be Σ -reflecting in L. 2 1 Notation: We use ⌣·i· to denote the end of a proof, and we write ·–i· when we leave a proof to the reader. We will use gothic letters A, B, M, ... for structures (=models of a first order language), and the corresponding latin letters A, B, M, ... for the underlying universes. Thus, a model A will have the universe A, and if A′ ⊆ A then we let A′ be the submodel (possibly with partial functions) with universe A′, etc. 3 Goldstern–Shelah: Forcing Axioms with Small Antichains 1. Fuchino’s problem and other applications Let E be a class of forcing notions. 1.1 Definition: Let A and B be two structures for the same first order language, and let E be a class (or property) of forcing notions. We say that A and B are E-potentially isomorphic (A ≃ B) iff there is a forcing P ∈ E such that k− “A ≃ B.” A ≃ B means E P P A ≃ B. {P} 1.2 Definition: We say that a structure A is nonrigid, if it admits a nontrivial automor- phism. We say that A is E-potentially nonrigid, if there is a forcing notion P ∈ E, k− “A P is nonrigid”. 1.3 Definition: (1) PI(E,λ) is the statement: Any two E-potentially isomorphic structures of size λ are isomorphic. (2) PA(E,λ) is the statement: Any E-potentially nonrigid structure of size λ is nonrigid. PI(E,λ) was defined by Fuchino [Fu]. It is clear that FA(E,λ) =⇒ BFA(E,λ) =⇒ PI(E,λ)&PA(E,λ) for all E, and Fuchino asked if PI(E,λ) implies FA(E,λ), in particular for the cases E=ccc, E=proper and E=stationary-preserving. We will show in this section that the the three statments BFA, PA and PI are in fact equivalent. Hence in particular PI(ccc,λ) is equivalent to MA(λ). In the next sections we will show that for E=proper, the first implication cannot be re- 4 Goldstern–Shelah: Forcing Axioms with Small Antichains versed, by computing the exact consistency strength of BPFA and comparing it to the known lower bounds for the consistency strength of PFA. 1.4 Theorem: For any forcing notion P and for any λ, the following are equivalent: PI(P,λ) PA(P,λ) BFA(P,λ) Proof of PI ⇒ PA: This follows from theorem 1.13 below. Here we will give a shorter proof under the additional assumption that we have not only PI(P) but also PI(P ) for all p p ∈ P, where P is the set of all elements of P which are stronger that p: p Let M be a potentially nonrigid structure. So there is a a P-name f such that e k− “f is a nontrivial automorphism of M” P e We can find a condition p ∈ P and two elements a 6= b of M such that p k− “f(a) = b” P e Since we can replace P by P , we may assume that p is the weakest condition of P. So we p have that (M,a) and (M,b) are potentially isomorphic. Any isomorphism from (M,a) to (M,b) is an automorphism of M mapping a to b, so we are done. ·i· ⌣ PI⇒PA We will now describe the framework of the proof of the second part of our theorem: PA ⇒ BFA.Westartwithaforcing notionP. Recallthat (forthemoment) allourforcing notions are a complete Boolean algebras. Fix a small family of small antichains. Our structure will consist of a disjoint union of the free groups generated by the antichains. On each of the free groupsthetranslationby anelement ofthecorresponding antichain willbe anontrivial 5 Goldstern–Shelah: Forcing Axioms with Small Antichains automorphism, and if all these translating elements are selected from the antichains by a directed set, then the union of these automorphisms will be an automorphism of the whole structure. We will also ensure that “essentially” these are the only automorphisms, so every automorphism will define a sufficiently generic set. 1.5 Definition: For any set X let F(X) be the free group on the generators X, and for w ∈ F(X) define supp(w) = {Y ⊆ X : w ∈ hYi}, i.e. supp(w) is the set of elements x T of X which occur (as x or as x−1) in the reduced representation of w. (If you prefer, you can change the proof below by using the free abelian group generated by X instead of the free group, or the free abelian group of order 2, ...). 1.6 Setup: Let P be a complete Boolean algebra, and let (A : i ∈ I) be a system of λ i many maximal antichains of size λ. We may assume that this is a directed system, i.e., for any i,j ∈ I there is a k ∈ I such that A refines both A and A . So if we write i < j for k i j “A refines A ”, (I,<) becomes a partially ordered upwards directed set. (We say that A j i refines B if each element of A is stronger than some unique element of B, or in the Boolean sense if there is a partition A = A of the set A satisfying ∀b ∈ B a = b.) S b Pa∈Ab b∈B Assuming PA(P,λ), we will find a filter(base) meeting all the sets A . i 1.7 Definition: (a) Let (F ,∗) be the free group generated by A , and let M be the disjoint union i i of the sets F . i (b) For i ∈ I, z ∈ F let i R = {(y,z∗y) : y ∈ F } i,z i j (c) If i < j, then there is a “projection” function h from A to A : For p ∈ A , i j i j 6 Goldstern–Shelah: Forcing Axioms with Small Antichains j h (p) is the unique element of A which is compatible with (and in fact weaker i i j than) p. h extends to a unique homomorphism from F to F , which we will i j i j also call h . i 1.8 Fact: (1) The functions hj commute, i.e., if i < j < k then hk = hj ◦hk. i i i j (2) If i < j, p ∈ Aj, then p is stronger than hji(p). ·–i· 1.8 Now let M = (M,(F ) ,(R ) ,(hj) ), where we treat all sets F , R , i i∈I i,z i∈I,z∈Fi i i∈I,j∈I,i<j i i,z j h as relations on M. i 1.9 Definition: Let G ⊆ P be a filter which meets all the sets A , say G∩A = {y (G)}. i i i Define f : M → M as follows: If x ∈ F , then f (x) = x∗y (G) (here ∗ = ∗ is the group G i G i i operation on F ). i 1.10 Fact: If G is a filter which meets all sets A , then f is an automorphism of M i G without fixpoints. Proof: It is clear that the sets F and the relations R are preserved. Note that for i i,z j j i < j we have h (y ) = y , since y and y are compatible. Since the functions h are i j i i j i j j j j j homomorphisms, we have h (f (x)) = h (x ∗ y ) = h (x)∗ y = f (h (x)), so also h is i G i j i i G i i preserved. ·i· ⌣ 1.10 So M is potentially nonrigid. So by PA(P,λ) we know that M is really nonrigid. Finally we will show how a nontrivial automorphism of M defines a filter G∗ meeting all the sets A . i So let F be an automorphism. Let 1 be the neutral element of F , and assume F(1 ) = w . i i i i SincethesetsF arepredicatesinourstructure, wemusthavew ∈ F . Usingthepredicates i i i 7 Goldstern–Shelah: Forcing Axioms with Small Antichains j j h we can show: If i < j, then h (w ) = w , and using the predicates R we can show i i j i i,z that for all z ∈ F we must have F(z) = z ∗w . i i Therefore, as F is not the identity, we can find i∗ ∈ I such that w 6= 1 . From now on i∗ i∗ we will work only with I∗ = {i ∈ I : i∗ ≤ i}. Since every antichain A is refined by some i antichain A with j ∈ I∗ it is enough to find a directed set which meets all antichains A j j for j ∈ I∗. Let u = supp(w ). So for all i ∈ I∗ the set u is finite and nonempty (since hi [u ] ⊇ u 6= i i i i∗ i i∗ ∅). 1.11 Fact: (1) If J ⊆ I∗ is a finite set, then there is a family {p : i ∈ J} such that i (a) for all i ∈ J: p ∈ u i i j (b) for all i,j ∈ J: If i < j, then h (p ) = p . i j i (2) There is a family {p : i ∈ I∗} such that i (a) for all i ∈ I∗: p ∈ u i i (b) for all i,j ∈ I∗: If i < j, then hj(p ) = p . i j i (3) If {p : i ∈ I∗} is as in (2), then this set generates a filter which will meet all i sets A . i Proof of (1): As I∗ is directed, we can find an upper bound j for J. Let p be an element j of w such that p := h (p ) ∈ w for all i ∈ J. j i i j i (2) follows from (1), by the compactness theorem of propositional calculus. (Recall that all sets u are finite.) i (3): We have to show that for any i ,i ∈ I∗ the conditions p and p are compatible, 1 2 i1 i2 8 Goldstern–Shelah: Forcing Axioms with Small Antichains i.e., have a common extension. Let j be an upper bound of i and i . Then p witnesses 1 2 j that p and p are compatible, as hj (p ) = p and hj (p ) = p . ·i· ·i· . i1 i2 i1 j i1 i2 j i2 ⌣ 1.11 ⌣ 1.4 For the theorem 1.13 below we need the following definitions. 1.12 Definition: A tree on a set X is a nonempty set T of finite sequences of elements of X which is closed under restrictions, i.e., if η : k → X is in T and i < k, then also η↾i ∈ T. The tree ordering ≤ is given be the subset (or extension) relation: η ≤ ν iff η ⊆ ν iff T ∃i : η = ν↾i. For η ∈ T let Suc (η) := {x ∈ X : η⌢x ∈ T}. T For A ⊆ T, η ∈ T we let rk(η,A) be the rank of η with respect to A, i.e., the rank of the (inverse) tree ordering on the set {ν : η ≤ ν ∈ T,∀ν′ : η ≤ ν′ < ν ⇒ ν′ ∈/ A} In other words, rk(η,A) = 0 iff η ∈ A, rk(η,A) = ∞ iff there is an infinite branch of T starting at η which avoids A, and rk(η,A) = sup{rk(ν,A)+1 : ν a direct successor of η} otherwise. 1.13 Theorem: For any two structures A and B there is a structure C = C(A,B) such that in any extension V′ ⊇ V of the universe, V′ |= “A ≃ B ↔ C is not rigid.” Proof: Wlog |A| ≤ |B|. Also wlog A and B are structures in a purely relational language L, and we may also assume that A∩B = ∅. We will say that a tree T on A∪B “codes A” iff (1) Suc (η) ∈ {A,B} for all η ∈ T. T (2) Letting TA := {η ∈ T : Suc (η) = A}, the ranks rk(η,TA) are < ∞ for all T 9