An Introduction to Extenders and Core Models for Extender Sequences
Logic Colloquium '87 H a . Ebbinghaus et el. (Editors) 0 Elsevier ScienLe PublishersB.V.(North-Holland), 1989
I37
AN INTRODUCTION TO EXTENDERS AND C...
Logic Colloquium '87 H a . Ebbinghaus et el. (Editors) 0 Elsevier ScienLe PublishersB.V.(North-Holland), 1989
I37
AN INTRODUCTION TO EXTENDERS AND CORE MODELS FOR
EXTENDER SEQUENCES Peter Koepke Universitit Freiburg, West Germany Abstract. This article surveys the theory of core models for non-overlapping coherent sequences of extenders. These models of set theory arise canonically when one wants to construct inner models for strong cardinals. Strong cardinals are defined in terms of elementary embeddings of V. and we define extenders a s a way of coding elementary embeddings. The natural inner model for a strong cardinal is of the form L[E] , where, in L[E], E is a coherent sequence of extenders. We obtain such sequences E = recursively if each EKV is a suitable extender on the core model K[Ep]. We give a definition of the core model K[F] from iterable premice, which are small "L[ E] -like" structures. Without proofs, we cite the fundamental properties of this family of core models. If a model L[E] for a strong cardinal does not exist (''7L[E]'') we can define a canonical core model K[ Fcan]. K[Fcan] is the largest core model. Assuming ?L[ El, # K[Fcan] satisfies analogues of the properties of L when 0 does not exist: K[Fcan] is rigid, i.e.. there is no non-trivial elementary embedding II :K[ Fcan] -t K[ Fcan] , and "[ 'can1 satisfies a weak covering theorem. The coarse ( = non-finestructural) characterisation of core models together with some fundamental facts on core models suffice for certain applications. We show: If the existence of a successor cardinal which is Jonsson is consistent. then the existence of a strong cardinal is consistent (relative to ZFC).
138
P. Koepke
Introduction. The study of consistency strengths in axiomatic set theory involves the construction of certain models of set theory out of given models. The method of forcing allows to extend ground models by generic sets. Smaller models are obtained by forming inner models of set theory within given universes. The well-established program of reducing consistency strengths to large cardinal axioms (see [6]) typically proceeds as follows: To show that a principle A is equiconsistent with a large cardinal axiom B (relative to ZFC) one (a) gives a forcing construction of a model of ZFC+A from a ground model of ZFC+B, and (b) shows that within any model of ZFC+A we can define an inner model of ZFC+B. Part (b) corresponds to the intuition that by suitably restricting the universe large cardinal properties should grow stronger. a d e l ' s model L of constructible sets is the paradigm of an inner model. It is distinguished as being the smallest inner model of ZF. L is definable as the union of a hierarchy whose structure is exceedingly uniform. L is a model a of the generalized continuum hypothesis (GCH). Jensen's finestructure analysis of the L=-hierarchy produced many important combinatorial principles in L. Unfortunately, L ' s role as an inner model for large cardinals is restricted since there are no measurable cardinals in L [lo]. The decisive step towards larger inner models with a constructible structure was taken by Dodd and Jensen in defining the core model K [4]. Large cardinal properties below measurability are compatible with "V = K" . The definition of K readily generalizes to include measurable cardinals. and also measurable cardinals of higher order (see [91). In this paper we consider the notion of a strong cardinal which was introduced by R.Jensen. Like measurables. these cardinals are defined using elementary embeddings of the universe. We define an inner model K[Fcan] which captures the
Extenders atid Core Models for Extender Sequences
139
large cardinal strength of the universe below a strong cardinal. K[ Fcan] relates to a strong cardinal much like K relates to a measurable cardinal. K[Fcan] is the canonical member of a whole family of inner models. the family of c ~ r e models for non-overlapping extender sequences. Extenders are a generalization of the concept of a normal measure on a measurable cardinal. They were invented by A.Dodd, R.Jensen, and W.Mitchel1. This family of core models satisfies natural embedding properties, and a weak covering theorem holds for K[Fcan]. We apply these in establishing a lower bound for a consistency strength: Theorem. Assume that the existence of a successor cardinal which is Jonsson is consistent with ZFC. Then the existence of a strong cardinal is consistent with ZFC. Although the core models for extender sequences are defined here without using any finestructure. the deeper results depend heavily on finestructural techniques. So far, these techniques are of much greater complexity than the finestructure for L. However, ongoing research indicates that by convenient re-structuring of K[Fcan] as the union of some hierarchy one can get along with an "L-like" finestructure theory. This paper is structured as follows: First we introduce strong cardinals and show, that they are of maximal order of measurability. Extenders are introduced in 5 2 . In 53 we consider extendability. i.e. whether an extender possesses an extension map with a transitive target model. Next we show how extenders can be used to approximate arbitrary elementary embeddings. This is used to re-formulate the notion of strongness within ZFC. $ 5 introduces the model L[E] for a strong cardinal. To test the "L-likeness" of L[E] we begin to prove the continuum hypothesis (CH) in L[E] and are led to the consideration of iterable premice. S6 studies iterable structures, and 57 develops (part of) the theory of iterable premice. With this we are able to conclude the proof of CH in
P. Koepke
140
L[E]. $ 8 contains an informal argument why one is lead to consider core models if one wants to recursively build a model L[E] for a strong cardinal. The next paragraph quotes the fundamental theorems on these core models, and finally in 5 1 0 we prove the result on the consistency strength of successor Jonsson cardinals, strengthening a result of [7]. The theory of extenders presented here is basically that of R.Jensen in [5]. Throughout this paper we use standard set theoretical notation. We work in ZFC, i.e., Zermelo-Fraenkel set theory with the axiom of choice. By ZFC- we denote the system ZFC but without the powerset axiom. Relative constructibility is done with the J-hierarchy. If d is a finite sequence of predicates let rudx(X) denote the closure of X under the rudimentary functions and the functions x xnAi (see [ 21 ) . The J[d]-hierarchy is defined as: