It has been proven by Blass [1973] that any two P-points which have a P-point as a common upper bound in the Rudin-Keisler (RK) ordering necessarily have a common lower bound (necessarily a P-point). Hence two nonisomorphic Ramsey ultrafilters have neither a common lower nor a common upper bound which is a P-point. So in a model of CH for example (or MA, P(c),…), the RK ordering restricted to P-points is neither upward nor downward directed, since it is well known that nonisomorphic Ramsey ultrafilters exist in such models. On the other hand, we will see that in the model for “near coherence of filters” (NCF) produced by Blass and Shelah [1985], the RK ordering of P-points is upward, hence downward directed. This shows that the question of directedness of the RK ordering of P-points, upward or downward, cannot be decided in ZFC.

There is a related question, asked by Blass in [1973], whether two P-points which have a common lower bound necessarily have a common upper bound which is a P-point. Our main result establishes the independence of this statement relative to ZFC. Its consistency will follow as soon as we show that the RK ordering of P-points is upward directed in the NCF model mentioned above, which we do in §2. But its independence will require a new construction, and will be given in §3.

Let T1 be the complete first-order theory of the additive group of the integers with 1 as distinguished element (in symbols, T1 = Th(Z, +, 1)). In this paper we prove that all models of T1 are ℵ0-homogeneous (§2), classify them (and lists of elements in them) up to isomorphism or L∞κ-equivalence (§§3 and 4) and show that they may be as complex as arbitrary sets of real numbers from the point of view of admissible set theory (§5). The results of §§2 and 5 together show that while the Scott heights of all models of T1 are ≤ ω (by ℵ0-homogeneity) their HYP-heights form an unbounded subset of the cardinal .

In addition to providing this unusual example of the relation between Scott heights and HYP-heights, the theory T1 has served (using the homogeneity results of §2) as an example for certain combinations of properties that people had looked for in stability theory (see end of §4). In §6 it is shown that not all models of T = Th(Z, +) are ℵ0-homogeneous, so that the availability of the constant for 1 is essential for the result of §2.

The two main results of this paper (2.2 and essentially Theorem 5.3) were obtained in the summer of 1979. Later we learnt from Victor Harnik and Julia Knight that T1 is of some interest for stability theory, and were encouraged to write up our proofs.

During 1982/3 we improved the proofs and added some results.

Let and be two countable relational models of the same first order language. If the models are nonisomorphic, there is a unique countable ordinal α with the property that

i.e. and are L∞ω-equivalent up to quantifier-rank α but not up to α + 1. In this paper we consider models and of cardinality ω1 and construct trees which have a similar relation to and as a above. For this purpose we introduce a new ordering T ≪ T′ of trees, which may have some independent interest of its own. It turns out that the above ordinal α has two qualities which coincide in countable models but will differ in uncountable models. Respectively, two kinds of trees emerge from α. We call them Scott trees and Karp trees, respectively. The definition and existence of these trees is based on an examination of the Ehrenfeucht game of length ω1 between and . We construct two models of power ω1 with mutually noncomparable Scott trees.

For e.g. λ = μ+, μ regular, λ larger than the continuum, we prove a strong nonpartition result (stronger than λ → [λ; λ]2). As a consequence, the product of two topological spaces of cellularity <λ may have cellularity λ, or, in equivalent formulation, the product of two λ-c.c. Boolean algebras may lack the λ-c.c. Also λ-S-spaces and λ-L-spaces exist. In fact we deal not with successors of regular λ but with regular λ above the continuum which has a nonreflecting stationary subset of ordinals with uncountable cofinalities; sometimes we require λ to be not strong limit.

The paper is self-contained. On the nonpartition results see the closely related papers of Todorčević [T1], Shelah [Sh276] and [Sh261], and Shelah and Steprans [ShSt1].

On the cellularity of products see Todorčević [T2] and [T3], where such results were obtained for (e.g.) cf and ; the class of cardinals he gets is quite disjoint from ours. In [Sh282] such results were obtained for more successors of singulars (mainly λ+, λ > 2cf λ). Also, concerning S and L spaces, Todorčević gets existence.

Todorčević's work on cardinals like relies on [Sh68] (see more in [ShA2, Chapter XIII]) (the scales appearing in the proof of ). The problem was stressed in a preliminary version of the surveys of Juhasz and Monk. We give a detailed proof for one strong nonpartition theorem (1.1) and then give various strengthenings. We then use 1.10 to get the consequences (in 1.11 and 1.12).

We prove the existence of p-ideals that are nonmeagre subsets of (ω) under various set-theoretic assumptions.

In this work we give a complete answer as to the possible implications between some natural properties of Lebesgue measure and the Baire property. For this we prove general preservation theorems for forcing notions. Thus we answer a decade-old problem of J. Baumgartner and answer the last three open questions of the Kunen-Miller chart about measure and category. Explicitly, in §1: (i) We prove that if we add a Laver real, then the old reals have outer measure one. (ii) We prove a preservation theorem for countable-support forcing notions, and using this theorem we prove (iii) If we add ω2 Laver reals, then the old reals have outer measure one. From this we obtain (iv) Cons(ZF) ⇒ Cons(ZFC + ¬ B(m) + ¬ U(m) + U(c)). In §2: (i) We prove a preservation theorem, for the finite support forcing notion, of the property “F ⊆ ωω is an unbounded family.” (ii) We introduce a new forcing notion making the old reals a meager set but the old members of ωω remain an unbounded family. Using this we prove (iii) Cons(ZF) ⇒ Cons(ZFC + U(m) + ¬ B(c) + ¬ U(c) + C(c)). In §3: (i) We prove a preservation theorem, for the finite support forcing notion, of a property which implies “the union of the old measure zero sets is not a measure zero set,” and using this theorem we prove (ii) Cons(ZF) ⇒ Cons(ZFC + ¬U(m) + C(m) + ¬ C(c)).

We study increasing F-sequences, where F is a dilator: an increasing F-sequence is a sequence (indexed by ordinal numbers) of ordinal numbers, starting with 0 and terminating at the first step x where F(x) is reached (at every step x + 1 we use the same process as in decreasing F-sequences, cf. [2], but with “+ 1” instead of “−1”). By induction on dilators, we shall prove that every increasing F-sequence terminates and moreover we can determine for every dilator F the point where the increasing F-sequence terminates.

We apply these results to inverse Goodstein sequences, i.e. increasing (1 + Id)(ω)-sequences. We show that the theorem

(a combinatorial theorem about ordinal numbers) is not provable in ID1.

For a general presentation of the results stated in this paper, see [1].

We use notions and results concerning the category ON (ordinal numbers), dilators and bilators, summarized in [2, pp. 25–31].

It is consistent with ¬ CH that every universal theory of relational structures with the joint embedding property and amalgamation for -(3)-diagrams has a universal model of cardinality ℵ1. For classes with amalgamation for -(4)-diagrams it is consistent that and there is a universal model of cardinality ℵ2.

Weakly minimal sets were first invented by Shelah [S] to solve Łoś' conjecture for uncountable theories. The first major result about them, which is that any nonalgebraic strong type either is of Morley rank 1 or has locally modular geometry, was proved by Buechler [B1]. Recently, Hrushovski [Hr] showed that a locally modular weakly minimal set interprets an abelian group, a result which revolutionized the study of these sets. Also Hrushovski showed that the geometry on the connected component of a weakly minimal locally modular group is that of a vector space over a division ring. This is also implicit in the result of Pillay and Hrushovski (Theorem 1.3 from [HP]) quoted below. In a sense, this completes their study, as any vector space over any division ring is in fact strongly minimal if there is no other structure. But the question remains as to what other structure such a group could have. This is the issue we address here.

Some work has been done on this question; in fact, Pillay ([Pi], generalized in [HP]) has demonstrated, in the more general context of a weakly normal group A, that any definable subset of An is in fact a Boolean combination of almost 0-definable cosets of almost 0-definable subgroups. In the weakly minimal case this can be sharpened a fair bit. The case of a weakly minimal group of bounded exponent is dealt with in [HL]. Here we consider the case where the group has unbounded exponent.

It is the purpose of this paper to make explicit the connection between J.-Y. Girard's “linear logic” [4], and certain models for the logic of quantum mechanics, namely Mulvey's “quantales” [9]. This will be done not only in the case of commutative linear logic, but also in the case of a version of noncommutative linear logic suggested, but not fully formalized, by Girard in lectures given at McGill University in the fall of 1987 [5], and which for reasons which will become clear later we call “cyclic linear logic”.

For many of our results on quantales, we rely on the work of Niefield and Rosenthal [10].

The reader should note that by “the logic of quantum mechanics” we do not mean the lattice theoretic “quantum logics” of Birkhoff and von Neumann [1], but rather a logic involving an associative (in general noncommutative) operation “and then”. Logical validity is intended to embody empirical verification (whether a physical experiment, or running a program), and the validity of A & B (in Mulvey's notation) is to be regarded as “we have verified A, and then we have verified B”. (See M. D. Srinivas [11] for another exposition of this idea.)

This of course is precisely the view of the “multiplicative conjunction”, ⊗, in the phase semantics for Girard's linear logic [4], [5]. Indeed the quantale semantics for linear logic may be regarded as an element-free version of the phase semantics.

One of the simplest and yet most fruitful ideas in forcing was the notion of Karel Prikry in which he used a measure on a cardinal κ to change the cofinality of κ to ω without collapsing it. The idea has found connections to almost every branch of modern set theory, from large cardinals to small, from combinatorics to models, from Choice to Determinacy, and from consistency to inconsistency. The long list of generalizers and modifiers includes Apter, Gitik, Henle, Spector, Shelah, Mathias, Magidor, Radin, Blass and Kimchi.

This paper is about generalizing Prikry forcing and partition properties to “simple spaces”. The concept of a simple space is itself the generalization of those combinatorial objects upon which the notions of “measurable”, “compact”, “supercompact”, “huge”, etc. are based. Simple spaces were introduced in [ADHZ1] and [ADHZ2] together with a broader generalization, “filter spaces”. The definition provided here is a small simplification of earlier versions. The author is indebted to Mitchell Spector, whose careful reading turned up numerous errors, some subtle, some flagrant.

In this first section, we review simple spaces briefly, including a short introduction to the space Qκλ. In §2, we describe our generalizations of partition property and Prikry forcing, and discuss the relationship between them. In §3, we find a partition property for the huge space [λ]κ, but show that Prikry forcing here is impossible. We find partition properties for Qκλ and show that Prikry forcing can be done here.

Given a finite lexicon L of relational symbols and equality, one may view the collection of all L-structures on the set of natural numbers ω as a space in several different ways. We consider it as: (i) the space of outcomes of certain infinite two-person games; (ii) a compact metric space; and (iii) a probability measure space. For each of these viewpoints, we can give a notion of relative ubiquity, or largeness, for invariant sets of structures on ω. For example, in every sense of relative ubiquity considered here, the set of dense linear orderings on ω is ubiquitous in the set of linear orderings on ω.

Before the remarkable theorem of Martin and Steel [6] showing that the existence of a supercompact cardinal κ implies L[R] ⊨ ZF + AD + DC, and the later theorem of Woodin [9] showing that Con(ZFC + There exists an ω sequence of Woodin cardinals) ⇔ Con(ZF + AD + DC), much set-theoretic research was focused upon showing that the consistency of fragments of AD + DC followed from more “reasonable” hypotheses such as versions of supercompactness. A good example of this is provided by the results of [1], in which the following theorems are proven.

C. Karp has shown that if α is an ordinal with ωα = α and A is a linear ordering with a smallest element, then α and α ⊗ A are equivalent in L∞ω up to quantifer rank α. This result can be expressed in terms of Ehrenfeucht-Fraïssé games where player ∀ has to make additional moves by choosing elements of a descending sequence in α. Our aim in this paper is to prove a similar result for Ehrenfeucht-Fraïssé games of length ω1. One implication of such a result will be that a certain infinite quantifier language cannot say that a linear ordering has no descending ω1-sequences (when the alphabet contains only one binary relation symbol). Connected work is done by Hyttinen and Oikkonen in [H] and [O].