ANNALS OF PURE AND APPLIED LOGIC Annals of Pure and Applied
The number
Logic 9X (1999)
69-86
of types in simple theories’ Enrique Casanovas *
Received
27 April 1998; wxived
in revised form
Communicated
I September
1998
by A. Wilkic
Abstract We continue work of Shelah on the cardinality of families of pairwise incompatible types in simple theories obtaining characterizations of simple and supersimple theories. WC develop a local analysis of the number of types in simple theories and we find a new example of a simple unstable theory. @ 1999 Elsevier Science B.V. All rights reserved. ,-lAfS
cbss~jicntion:
03C45
Ke~~~o~ds: Model theory; Simple theories;
Low theories
1. Introduction In [9] Shelah introduced simple theories as a generalization of stable theories where forking has still some nice properties. In Kim’s thesis [2] this class of theories was studied and it was shown that in simple theories independence is symmetric and transitive and forking = dividing. As a consequence a substantial part of the forking machinery developed by Shelah for stable theories is valid in the wider context of simple theories. More surprisingly, Kim and Pillay (see [2, 31) were able to show that simple theories are in fact the class of theories where a notion of independence satisfying some basic propcrtics can bc defined. Morcovcr there arc several cxamplcs of simple unstable theories that are of interest in their own right, mainly the random graph, pseudofinite fields, smoothly approximable structures and algebraically closed fields with a generic automorphism. So it seems now that stable theories are going to be replaced by simple theories in many foundational investigations in pure model theory. The interested reader can find a good overview of the situation in [6]. The present paper is devoted to extending some early work of Shelah about the number of types in simple theories. ’ Partially supported by grant PB94-0854 * E-mail: [email protected].
of DGICYT
016%0072/99/$-see front matter @ 1999 Elsevier PII: SOl68-0072(98)00050-5
and grant HA1996-0131
Science B.V. All rights reserved
E. Casanovusl
70
obtaining
Annals of Pure and Applied Logic 98 (1999)
this way several new characterizations
well as a new example The number
of a simple unstable
of simple and supersimple
q= cp(x,y)
by the condition
theories as
theory.
of types played a major role in Shelah and subsequent
theory, both from a local or a global point of view. Locally of a formula
6946
work in stability
one can define stability
that for any infinite
set A there are at
most (A] many complete q-types over A, i.e, IS,(A)I d IA]. Globally, one can define A-stability of a theory T by the condition that for any set A of cardinality 62 there are at most II many complete n-types over A, i.e, I&(.4)1 63,. Moreover, the classes of theories in the stability hierarchy are usually defined in terms of the number of types. A small number of types corresponds to definability of types, to the nonexistence of certain combinatorial structures like orderings or trees and to the nonexistence of long forking chains. Simple unstable theories have the independence property and therefore for every cardinal /1 there exists a set A with IAl = ,? and IS( = 2’. But there are nonsimple theories with the independence property, so we cannot expect to find a dividing line between simple and nonsimple theories by looking at the cardinality of S(A). However there is a different way of counting types, taking in consideration both the cardinality 2 of the set of parameters and the cardinal@ K of the (partial) types. Shelah discovered in [9] how to use this notion to find a different behavior between simple and nonsimple theories. In [9], Theorem 0.2, we can read: Theorem 1.1. The following conditions on T are equivalent: (1) T is nonsimple;
(2) for every 2, K such that 35” = 2, there are /1”pairwise contradictory m-types of power ICover a set A of cardinality 3,; (3) there is a set A, and a set S of m-types over A ofpower x such that ISI > IAll’l+z+ 21T1+%and no C E 6 realizes > x types from S. A proof can be found in [8], Theorem III 7.7, or, with some simplifications, in [2], Proposition 2.20. The main ingredient in the proof is an application of the Erdiis-Rado theorem. proof.
Many results
in the present
paper are obtained
by developing
ideas of this
In this paper we study the function NT(rc, 2) which assigns to every pair K, J. the supremum of the cardinalities IP] of families P which consists of pairwise incompatible partial types of size
E. CasanovasI Annals of Pure und Applied Logic, 98 ( 1999) 6946 Definition
2.5).
This allows
us to find in Theorem
71
2.7 a local characterization
of
simplicity. The well-known plies NT,(x,
examples
of simple unstable
theories
A) < i. + 2” for every 40. The questions
NT: (K, A)
are supersimple
and this im-
arise of what is the meaning i) > ,! + 2” is possible
is given in Proposition
of in a
3.4. In Section 5
of a simple theory with NT,(K, i) > i + 2” for some cp.ti. A. This
example has a property which seems to be stronger:
the formula
cp divides c( times for
every c(
of a certain kind of trees and to the non existence
of
2. Counting types In the following T is a complete theory with infinite models and L is the language of T. By consistent we always mean consistent with T. We work inside a very big, very saturated model of T, the monster model of T. All the parameters needed belong to the monster model and all the models considered are elementary submodels of the monster model. We use x, y,z, . . , for finite tuples of variables and a, b, c.. . . for finite tuples of parameters. If a is an ordinal number and (b,: i
of the same length,