An Invitation to Model Theory

Model theory begins with an audacious idea: to consider statements about

mathematical structures as mathematical objects of study in their own right. While
inherently important as a branch of mathematical logic, it also enjoys connections to
and applications in diverse branches of mathematics, including algebra, number theory
and analysis. Despite this, traditional introductions to model theory assume a
graduate-level background of the reader.
In this innovative textbook, Jonathan Kirby brings model theory to an undergraduate
audience. The highlights of basic model theory are illustrated through examples from
specific structures familiar from undergraduate mathematics, paying particular
attention to definable sets throughout. With numerous exercises of varying difficulty,
this is an accessible introduction to model theory and its place in mathematics.

j o n at h a n k i r b y is a Senior Lecturer in Mathematics at the University of East

Anglia. His main research is in model theory and its interactions with algebra, number
theory, and analysis, with particular interest in exponential functions. He has taught
model theory at the University of Oxford, the University of Illinois at Chicago, and the
University of East Anglia.
An Invitation to Model Theory

J O NAT H A N K I R B Y
University of East Anglia
 University Printing House, Cambridge CB2 8BS, United Kingdom
One Liberty Plaza, 20th Floor, New York, NY 10006, USA
477 Williamstown Road, Port Melbourne, VIC 3207, Australia
314–321, 3rd Floor, Plot 3, Splendor Forum, Jasola District Centre,
New Delhi – 110025, India
79 Anson Road, #06–04/06, Singapore 079906

Cambridge University Press is part of the University of Cambridge.

It furthers the University’s mission by disseminating knowledge in the pursuit of
education, learning, and research at the highest international levels of excellence.

www.cambridge.org
Information on this title: www.cambridge.org/9781107163881
DOI: 10.1017/9781316683002
© Jonathan Kirby 2019
This publication is in copyright. Subject to statutory exception
and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.
First published 2019
Printed and bound in Great Britain by Clays Ltd, Elcograf S.p.A.
A catalogue record for this publication is available from the British Library.
Library of Congress Cataloging-in-Publication Data
Names: Kirby, Jonathan, 1979– author.
Title: An invitation to model theory / Jonathan Kirby, University of East Anglia.
Description: Cambridge, United Kingdom ; New York, NY : Cambridge University
Press, 2019. | Includes bibliographical references and index.
Identifiers: LCCN 2018052996 | ISBN 9781107163881 (hardback ; alk. paper) |
ISBN 1107163889 (hardback ; alk. paper) | ISBN 9781316615553 (pbk. ; alk.
paper) | ISBN 1316615553 (pbk. ; alk. paper)
Subjects: LCSH: Model theory.
Classification: LCC QA9.7 .K57 2019 | DDC 511.3/4–dc23
LC record available at https://lccn.loc.gov/2018052996
ISBN 978-1-107-16388-1 Hardback
ISBN 978-1-316-61555-3 Paperback
Cambridge University Press has no responsibility for the persistence or accuracy
of URLs for external or third-party internet websites referred to in this publication
and does not guarantee that any content on such websites is, or will remain,
accurate or appropriate.
To Pirita, Lumia, Tapio, and Sakari
 Contents

Preface page ix
PART I LANGUAGES AND STRUCTURES 1
1 Structures 3
2 Terms 9
3 Formulas 13
4 Definable Sets 19
5 Substructures and Quantifiers 24
PART II THEORIES AND COMPACTNESS 29
6 Theories and Axioms 31
7 The Complex and Real Fields 37
8 Compactness and New Constants 42
9 Axiomatisable Classes 47
10 Cardinality Considerations 53
11 Constructing Models from Syntax 57
PART III CHANGING MODELS 63
12 Elementary Substructures 65
13 Elementary Extensions 70

vii
viii Contents

14 Vector Spaces and Categoricity 77

15 Linear Orders 83
16 The Successor Structure 88
PART IV CHARACTERISING DEFINABLE SETS 93
17 Quantifier Elimination for DLO 95
18 Substructure Completeness 99
19 Power Sets and Boolean Algebras 104
20 The Algebras of Definable Sets 109
21 Real Vector Spaces and Parameters 115
22 Semi-algebraic Sets 119
PART V TYPES 127
23 Realising Types 129
24 Omitting Types 133
25 Countable Categoricity 138
26 Large and Small Countable Models 142
27 Saturated Models 147
PART VI ALGEBRAICALLY CLOSED FIELDS 153
28 Fields and Their Extensions 155
29 Algebraic Closures of Fields 159
30 Categoricity and Completeness 163
31 Definable Sets and Varieties 167
32 Hilbert’s Nullstellensatz 173

Bibliography 177
Index 179
 Preface

This book is designed as an undergraduate or master’s-level course in model

theory. It has grown out of courses taught for many years at the University
of Oxford and courses taught by me at UEA. The choice of material and
presentation are based on pedagogical considerations, and I have tried to resist
the temptation to be encyclopedic.
In this book, the main programme of model theory is to take a familiar
mathematical structure and get an understanding of it in the following way.
First, find an axiomatisation of its complete theory. Second, if possible, classify
all the other models of the theory. Third, describe all the definable sets. As a
result, model theory is presented as a set of tools for understanding structures,
and the way the tools are applied to specific structures is as important as
the tools themselves. This gives motivation to the subject and connects it to
familiar material. Some readers may be more interested in the theory than in
the applications, but my view is that even those who eventually wish to work
in abstract model theory will get a better understanding of the basics by seeing
them applied to examples.
Historically, model theory grew as a branch of mathematical logic, and the
focus was mostly on logical issues, such as decidability. As model theory has
found more applications and connections to other branches of mathematics,
the study of definable sets has become more central.
I have tried to keep the book as self-contained as possible. Model theory
requires a level of mathematical sophistication in terms of abstract, rigorous
thinking, proofs, and algebraic thinking which students will normally have
developed through previous courses in algebra, logic, or geometry, but there
are few specific prerequisites. No topology is used, and almost no set theory is
used. A brief chapter explains the basic cardinal arithmetic methods needed for
some of the proofs, but with one or two exceptions, everything can be assumed

ix
x Preface

to be countable. A familiarity with the use of basic algebraic ideas, such as

bijections and homomorphisms or embeddings of groups, or rings, or vector
spaces, is essential, but when algebraic examples such as rings and vector
spaces are introduced, all the necessary definitions and facts are explained.
For the most part I have not given historical references for the material.
There are good historical remarks at the end of each chapter in the books of
Hodges [Hod93, Hod97] and Marker [Mar02], and I refer the reader to those.

Overview
The book is organised into six parts. Part I covers structures, languages, and
automorphisms, introduces definable sets, and proves the essential preservation
theorems.
Part II introduces theories and the programme of finding axiomatisations for
structures. Context is given with axiomatisations of the complex and real fields
and the natural numbers with addition and multiplication (Peano arithmetic).
The compactness theorem is then introduced, and examples of its use are given.
Part II concludes with the Henkin proof of the compactness theorem.
In Part III we show that even a complete theory can have many different
models with the Löwenheim–Skolem theorems. The notion of categoricity
is introduced via the example of vector spaces and is used to prove the
completeness of an axiomatisation. Further applications are given to dense
linear orders, where the back-and-forth method is introduced, and to the natural
numbers with the successor function.
The idea of quantifier elimination is introduced in Part IV, and used
to characterise the definable sets in dense linear orders and vector spaces.
Boolean algebras are introduced via an investigation into the theory of power
sets, partially ordered by the subset relation, and the definable subsets of a
structure in any number n of variables is shown to be a Boolean algebra, called
the Lindenbaum algebra of the theory. These Lindenbaum algebras are then
worked out in the examples of vector spaces and for the real ordered field.
Parts V and VI go in different directions and do not depend on each other.
Part V develops the notion of types, which is at the heart of modern model
theory. The first goal is the Ryll–Nardzewski theorem, which characterises
countably categorical theories as those for which there are only finitely many
definable sets in any given number of variables. There is then a brief discussion
of saturated models, leading to areas for further reading.
Part VI takes the model-theoretic techniques developed in the first four parts
and applies it to the theory of algebraically closed fields. Two chapters explain
 Preface xi

the necessary algebraic background, and a third proves categoricity, complete-

ness of the theory, and quantifier elimination. The next chapter explains how
the definable sets correspond to algebraic varieties and constructible sets, and
a final chapter gives a model theoretic proof of Hilbert’s Nullstellensatz, which
gives more information about the definable sets.

Suggestions for Using This Book as a Textbook

This book can be used for a course in many different ways. Figure 0.1,
illustrating the dependencies between chapters, can be used as a guide to
planning a course according to the lecturer’s preferences.
Each chapter is intended to be of such a length that it can be taught in one
hour (perhaps covering only the essentials) or in two hours (sometimes using
material from the exercises). Everything depends on the first three chapters,
so they should be covered first, unless the students have the prerequisite
knowledge from predicate logic. The emphasis is on semantic ideas, including
automorphisms, which is somewhat different from the usual emphasis in a first
logic course, and I have successfully taught this material to a class of students
of whom some had seen logic before and some not. For those who have not
seen this material before, Chapter 3 in particular goes rather quickly. Chapters
4 and 5 carry on the study of formulas to give students more opportunity to
consolidate their understanding.
Each chapter has several exercises at the end, which range from very easy
consolidations of definitions to more substantial projects. Some exercises need
more background than the other material in the chapter, and some exercises
develop extension material.
A short model theory section of a mathematical logic course could have
the compactness theorem and some applications as its goal and consist of
Chapters 1, 2, 3, and 6, Section 8.1, and Chapter 9. This could be filled out
with any of Chapters 4, 5, and 7, the rest of Chapter 8 or Chapter 11, or with
applications from Chapters 14, 15, 16, or 19. The back-and-forth method is
another highlight that could be reached quickly, by covering only the essential
sections of Chapters 1, 3, 6, and 17.
The heart of this book is the study of definable sets. A course centred on
those would consist of Parts I to IV, possibly omitting Chapters 5, 9, 11, and
16, and possibly adding Part VI.
A course aiming to cover the basics of types and the Ryll–Nardzewski
theorem, without the emphasis on definable sets, could consist of Chapters
1, 2, 3, 6, 8, 10–15, and 23–25.
xii Preface

31 32

30 5

29 27

7 13 28 26

25

24 20

22 23

21 7

20

18 19

17

16 15

14

13

12

10

7 19 9 8 11

6 10

4 5

Figure 0.1 Chapter dependencies: In order to make the diagram planar, Chapters
5, 7, 10, 13, 19, and 20 have been drawn in two (or even three) places. As the
diagram shows, Parts IV, V, and VI are largely independent of each other, with
much of Part V also independent of Part III. While Chapter 25 uses the concept of
Lindenbaum algebras from Chapter 20, that material does not rely on the rest of
Parts III or IV.
 Preface xiii

For research students, Part VI, on algebraically closed fields, gives a pattern
for carrying out the programme of understanding definable sets in a theory
which is similar to the pattern which works in many other theories of fields,
such as real-closed fields, differentially closed fields, separably closed fields,
algebraically closed valued fields, exponentially closed fields, algebraically
closed fields with an automorphism, and so on.

Further Reading
There are several more advanced textbooks in model theory, including those
of Marker [Mar02], Hodges [Hod93, Hod97], Poizat [Poi00], Tent and Ziegler
[TZ12], Sacks [Sac10], and Chang and Keisler [CK90]. Rothmaler [Rot00] is
more at the level of this book, but there is more emphasis on algebra, especially
groups. Väänänen [Vää11] introduces model theory via back-and-forth games,
which complements the approach in this book. Bridge [Bri77] was based on
the incarnation of the Oxford model theory course from the early 1970s, and
the emphasis is much more on logic. Other suggestions for further reading are
given through the book, particularly at the end of Parts IV, V, and VI.

Acknowledgements
Thanks to Boris Zilber for allowing me to take his lecture notes as a starting
point both for my course and for this book. I learned model theory initially
from the books of Wilfrid Hodges and David Marker, and from Boris. My
writing style was much influenced by the late Harold Simmons.
I would like to thank Cambridge University Press for their help in producing
the book and, in particular, Silvia Barbina for initially suggesting that I write it.
Many people have made helpful comments on drafts or have otherwise
helped me to write the book. I would like to thank Lou van den Dries,
David Evans, Åsa Hirvonen, Wilfrid Hodges, Ehud Hrushovski, Tapani Hyt-
tinen, Gareth Jones, Asaf Karagila, David Marker, Alice Medvedev, Charles
Steinhorn, Alex Wilkie, and Boris Zilber for sharing their expertise. Thanks
to Francesco Parente for reading two drafts and providing many considered
comments.
Thanks also to the students who attended my courses or read drafts of the
book and provided essential feedback. They include Abeer Albalahi, Michael
Arnold, Emma Barnes, Matt Gladders, Grant Martin, Oliver Matheau-Raven,
Audie Warren, and Tim Zander.
 Part I

Languages and Structures

In this first part of the book, we introduce the basic methods from predicate
logic which underpin the rest of the book. Mathematical objects are formalised
as L-structures, and statements about these structures are formalised in the
notions of L-terms and L-formulas. We introduce the central notion of a
definable set, a subset of an L-structure which is defined by an L-formula.
A key notion in mathematical logic is that of a recursive definition, and
both the notions of L-terms and L-formulas give examples of that. These
recursive definitions give rise to methods of proof by induction. Throughout
this part, these proofs by induction are used prove preservation theorems, most
importantly the preservation of definable sets by automorphisms of a structure.
From this result we get a method to show that certain subsets of a structure are
not definable, an essential tool which we will later exploit in the programme
of characterising the definable sets of a structure.

1
 1
Structures

In different mathematical contexts, familiar mathematical objects may actually

have different meanings. For example a reference to the integers Z in group
theory is likely to mean the infinite cyclic group, whereas in number theory it
is likely to mean Z as a ring. In model theory, when we specify a structure,
we have to be precise about such things. The integers as an additive group will
be written as Zadgp = Z; +, −, 0, and Z as a ring will be written as Zring =
Z; +, ·, −, 0, 1. The integers are also used to index the years in a calendar,
and an appropriate structure for that purpose is Z< = Z; <, because it is not
very meaningful to add or multiply calendar years, but the order is important.
In general, we capture these ideas with the notions of a language, L, and an
L-structure.

1.1 L-structures
Definition 1.1 A language L is specified by the following (sometimes called
the vocabulary or signature of L):

(i) a set of relation symbols,

(ii) a set of function symbols,
(iii) a set of constant symbols, and
(iv) for each relation and function symbol, a positive natural number called
its arity.

Definition 1.2 An L-structure A consists of a set A called the domain of the

structure, together with interpretations of the symbols from L:

(i) for each relation symbol R of L, of arity n, a subset RA of An ,

(ii) for each function symbol f of L, of arity n, a function f A : An → A, and
(iii) for each constant symbol c of L an element cA ∈ A.

3
4 1 Structures

Example 1.3 The language of rings Lring = +, ·, −, 0, 1, where + and · are
function symbols of arity 2 (we say they are binary function symbols), − is a
function symbol of arity 1 (a unary function symbol for negation rather than
the binary function of subtraction), and 0 and 1 are constant symbols. Zring is
the Lring -structure with domain the set Z of integers. We interpret the symbols
+, ·, and − as the usual functions of addition, multiplication, and negation of
integers and the constant symbols 0 and 1 are interpreted as the usual zero and
one. We can write this as Zring = Z; +Z , ·Z , −Z , 0Z , 1Z  if for example we want
to distinguish the symbol + from its interpretation +Z as a function from Z2
to Z. This distinction can be important in mathematical logic. For example we
also have the Lring -structure Rring with domain the set of real numbers R, and
the function +R : R2 → R is not the same function as +Z : Z2 → Z. However,
usually no ambiguity arises, and we just write + for the symbol and for its
interpretations in different structures.

Example 1.4 We write L< = <, where < is a binary relation symbol. Then
Z< is the L< -structure with domain
 the set Z of integers,
 and the symbol < is
interpreted as the set of pairs (a, b) ∈ Z2 | a < b . As above, we could write
this set as <Z , but usually we will not.

Examples 1.5 Many other languages can be built as variations on these two
languages:

(i) Lgp = ·, (−)−1 , 1 is the language of groups. Again, · is a binary function
symbol, (−)−1 is a unary function symbol representing the multiplicative
inverse function x → x−1 , and 1 is a constant symbol.
(ii) Ladgp = +, −, 0 is the language of groups written additively. It is a
sub-language of Lring , because every symbol in Ladgp is also in Lring (with
the same arity). We also say that Lring is an expansion of the language
Ladgp .
(iii) Lo-ring = +, ·, −, 0, 1, < is the language of ordered rings, consisting of
Lring ∪ L< .
(iv) A common language in which to consider the natural numbers is the
language of semirings, Ls-ring = +, ·, 0, 1. We will always use the
convention that 0 is a natural number.
(v) The language of monoids is Lmon = ·, 1 and the language of additive
monoids is Ladmon = +, 0.

Most of the structures we will consider as examples have a domain with

a commonly used notation such as Z, Q, or R, and the functions, relations,
and constants we consider will be those in the above languages. In this case
 1.2 Expansions and Reducts 5

we name the structure by putting the name of the language as a subscript, for
example Zring , Ro-ring , Q< , Ns-ring .
We can also specify structures directly, which is useful for creating simpler
examples. In this case we will often use a caligraphic letter for the name of the
structure and the corresponding Roman letter for the name of its domain, so
A would be the name of a structure with domain A. This convention of using
different notation for the name of a structure and its domain is useful when the
same set is the domain of different structures. However, where no confusion is
likely to arise, we will sometimes follow the common mathematical practice
of using the same notation for both.
Example 1.6 Take the language L = R, f, c, where R is a ternary
relation symbol, f is a unary function symbol, and c is a constant
symbol. We can define an L-structure A with domain A = {0, 1, 2, 3, 4},
by specifying the  interpretation of the symbols as
f A (x) = x+1 mod 5, RA = (x, y, z) ∈ A3  exactly two of x, y and z are equal
and cA = 3.

Remark 1.7 The key idea from mathematical logic here is that the symbols
of the vocabulary of a language L are separate from their interpretations in a
structure. It follows that the symbols do not have to be interpreted by their
usual meanings. For example we can make the set N into an Lgp -structure
by interpreting the symbol · as addition and (−)−1 as the identity function, so
x−1 = x for all x ∈ N, and interpreting the constant symbol 1 as the number 2.
However, this is perverse, and if we want an unusual interpretation, we will
choose to use a different symbol.
Given a mathematical problem you are trying to solve, or a statement you
want to understand, it is often a good exercise to work out what structure the
statement might be about. For example, the fundamental theorem of arithmetic
states that every positive integer can be written as a product of primes in a
unique way (up to reordering). An appropriate structure would have domain
the set N+ of positive integers and needs to have the multiplication function;
1 is a special case as the empty product, which is relevant to single out, so
we can regard the fundamental theorem of arithmetic as a statement about the
structure N+mon = N+ ; ·, 1. To actually prove the theorem, we might need more
than that, for example the order to do induction on and also addition.

1.2 Expansions and Reducts

When considering different structures with the same domain, there are two
useful pieces of terminology.
6 1 Structures

Definition 1.8 Let L be a language and L+ another language such that L ⊆ L+ ,

that is, every symbol of L is also a symbol of L+ . Let A+ be an L+ -structure
with domain A. Then the reduct of A+ to L is the L-structure A with domain
A, and every symbol of L interpreted in A exactly as in A+ . We also say that
A+ is an expansion of A to the language L+ .

For example Radgp = R; +, −, 0 is a reduct of Rring = R; +, −, ·, 0, 1,

which in turn is a reduct of Ro-ring = R; +, −, ·, 0, 1, <.

1.3 Embeddings and Automorphisms

Definition 1.9 Let A and B be L-structures. An embedding of L-structures
π
from A to B is an injective function A −→ B such that:

(i) for all relation symbols R of L, and all a1 , . . . , an ∈ A,

(a1 , . . . , an ) ∈ RA iff (π(a1 ), . . . , π(an )) ∈ RB ,

(ii) for all function symbols f of L, and all a1 , . . . , an ∈ A,

π( f A (a1 , . . . , an )) = f B (π(a1 ), . . . , π(an )), and

(iii) for all constant symbols c of L, π(cA ) = cB .

For any L-structure A, the identity function 1A on A is a embedding of A into

π
itself. An embedding A −→ B is an isomorphism iff there is an embedding
σ
B −→ A such that the composite π ◦ σ is the identity on B and the composite
σ ◦ π is the identity on A. We write π−1 for σ, as usual, and call it the inverse
of π. An isomorphism from A to itself is called an automorphism of A.

Examples 1.10 We have the obvious inclusion functions Z → Q → R → C,

which take a number to itself. These inclusion functions give us embeddings
of Lring -structures
Zring → Qring → Rring → Cring .

Definition 1.11 If we have an embedding A −→ B where the function is an

inclusion of the domain of A as a subset of the domain of B then we say that
A is a substructure of B, and that B is an extension of A.

Example 1.12 The only automorphisms of Zadgp are the identity and the map
x → −x. To see this, note that if π : Z → Z is an Ladgp -embedding, then
π(0) = 0, and if π(1) = n, then for any m ∈ N+ , we have
 Exercises 7

π(m) = π(1 + · · · + 1) = π(1) + · · · + π(1) = mn.

 
m m

Then also π(−n) = −π(n) = −mn. To have an inverse, π must be surjective,

which implies n = ±1.

Exercises
1.1 Let L =  f, c be a language with one unary function symbol and one
constant symbol. Describe all the possible L-structures on the domain
{1, 2}. How many of them are there up to isomorphism (which means
counting isomorphic structures as the same)?
1.2 For each of the following statements, give a structure which they say
something about.
(a) There is no largest integer.
(b) Every integer has an additive inverse.
(c) Square integers are always non-negative.
(d) Every complex number is the sum of a real number and i times a
real number.
(e) The exponential function is strictly increasing on the real numbers.
(f) A real quadratic equation ax2 + bx + c = 0 has real roots if and only
if b2 − 4ac  0.
(g) Euler’s identity eiπ + 1 = 0.
1.3 For each language L in Examples 1.5, say which of the sets from N, Z,
and R is the domain of an L-structure with the symbols interpreted with
their usual meanings. For example, there is no structure Ngp because
N is not closed under multiplicative inverses. Which of your structures
are expansions or reducts of each other? Which are extensions or
substructures?
1.4 What are all the automorphisms of Z< ?
1.5 Show that embeddings of N< into R< correspond to strictly increasing
sequences of real numbers.
1.6 Show that an embedding of L-structures is an isomorphism if and only if
it is surjective.
1.7 Explain what all the embeddings of Zadgp into Radgp are.
1.8 Find an automorphism π of the structure R; < such that π(0) = 1 and
π(1) = 5. Is your π also an automorphism of the structure R; +?
8 1 Structures

1.9 Find all the automorphisms of Qadgp .

1.10 Suppose that A+ is an expansion of A and π is an automorphism of A+ .
Show that π is also an automorphism of A. Give an example to show that
an automorphism of A may not be an automorphism of A+ .
1.11 Let A be an L-structure. Show that the automorphisms of A form a
group with the group operation being composition. We write this group
as Aut(A).
π
1.12 In model theory, a homomorphism of L-structures A −→ B is a function
π
A −→ B such that
(a) for all relation symbols R of L, and all a1 , . . . , an ∈ A,
if (a1 , . . . , an ) ∈ RA then (π(a1 ), . . . , π(an )) ∈ RB ,
(b) for all function symbols f of L, and all a1 , . . . , an ∈ A,
π( f A (a1 , . . . , an )) = f B (π(a1 ), . . . , π(an )), and
(c) for all constant symbols c of L, π(cA ) = cB .
Check that if L = Lgp or L = Lring , and A, B are groups or rings, then
L-homomorphisms are the same as group or ring homomorphisms. How
do L-homomorphisms compare with, and how do they differ from, the
notion of embedding of L-structures?
 2
Terms

A formal language is based on a collection of symbols, which consists of

variables such as x, y, or z, logical symbols such as =, ∧ , ¬, and ∃, brackets
(, ), [, and ], and then the vocabulary which is specific to the language:
the relation, function, and constant symbols. Apart from the collection of
symbols, a language also consists of various strings (lists) of the symbols,
which are called terms and formulas. These symbols, terms, and formulas
are then interpreted in structures. In the previous chapter we interpreted the
symbols from the vocabulary, and in the next chapter we will introduce the
formulas. Here we introduce and interpret the terms.

2.1 The Recursive Construction of Terms

Terms are certain strings of symbols that refer to elements of the structure or
to functions on the structure. For example, in the language Lring , terms include
0, (1 + 1), −((1 + 1) + 1), (x · (1 + 1)), (x + y), and ((x · x) + 1).
They can be interpreted in Zring as the numbers 0, 2, and −3 and as the functions
x → 2x, (x, y) → x + y, and x → x2 + 1, respectively.
Definition 2.1 (Terms) The set of terms of the language L is defined
recursively as follows.
(i) Every variable is a term.
(ii) Every constant symbol of L is a term.
(iii) If f is a function symbol of L of arity n, and t1 , . . . , tn are terms of L,
then f (t1 , . . . , tn ) is a term.
(iv) Only something built from the above three clauses in finitely many steps
is a term.

9
10 2 Terms

A closed term is a term which does not contain any variables.

In the examples we followed the usual practice of writing (1 + 1) rather

than +(1, 1) and so on. This is called infix notation and works only for binary
functions. When the function is written before its arguments, it is called prefix
notation.
As in the examples above, terms of L can be interpreted as elements of an
L-structure or functions on it.

Definition 2.2 (Interpretation of terms) Let A be an L-structure, t a term of

L, and x̄ = (x1 , . . . , xr ) a list of variables including all those which appear in
t. The recursive definition of terms is used to give a recursive definition of a
function tA : Ar → A as follows.
(i) If t = c, a constant symbol, then tA is the element cA of A, or as a
function it is the constant function which always takes the value cA .
(ii) If t = xi , the ith variable in the list x̄, then tA is the ith coordinate function
given by tA (x1 , . . . , xr ) = xi .
(iii) If t = f (t1 , . . . , tn ), then tA is given by composition

tA ( x̄) = f A (t1A ( x̄), . . . , tnA ( x̄)).

Remarks 2.3 (i) If t is a closed term, then in Definition 2.2 we can take
r = 0. In this case it is best to think of the interpretation tA as an element
of A. We can alternatively think of it as a function A0 → A. Since A0 is a
one-point set, this amounts to the same thing as an element of A.
(ii) Note that terms depend only on the language, but their interpretations
depend on the structure. For example, the term ((x + 1) + x) is interpreted
in the structure Rring as the function

R −→ R
x −→ 2x + 1,

but in Zring the same term is interpreted as a function Z → Z.

(iii) We can see that several terms can be interpreted as the same function.
For example, the terms ((−x + (x + x)) + (1 + 1)) and ((1 + x) + 1) are
both interpreted in Zring as the function

Z −→ Z
x −→ x + 2.

(iv) Where no ambiguity arises, we often do not write all the brackets which
should be there according to the recursive definition.
 Exercises 11

(v) In model theory we almost always deal with terms with a given list of
variables. So we will write that t( x̄) is a term, when we really mean that t
is a term and x̄ is a list of variables which includes all the variables
which appear in t.

2.2 Embeddings and Terms

By definition, an embedding of L-structures preserves the interpretations of the
constant and function symbols from L. We can use the recursive definition of
terms to give an inductive proof that embeddings preserve the interpretation
of all terms. When ā = (a1 , . . . , ar ) is a tuple of elements from A, we write
π(ā) as an abbreviation for the tuple (π(a1 ), . . . , π(ar )). Otherwise the notation
becomes too unwieldy.
π
Proposition 2.4 Suppose A −→ B is an embedding of L-structures, t is a term
of L, x̄ = (x1 , . . . , xr ) is a list of variables including all those which appear in
t, and ā ∈ Ar . Then
π(tA (ā)) = tB (π(ā)).
Proof We proceed by induction on the construction of terms.
If t is a constant symbol c, then
π(tA (ā)) = π(cA ) by Definition 2.2,
= cB by the definition of an embedding,
= tB (π(ā)) by Definition 2.2 again.
If t = xi , then π(tA (ā)) = π(ai ) = tB (π(ā)), using Definition 2.2 twice.
If t = f (t1 , . . . , t s ), then
π(tA (ā)) = π( f A (t1A (ā), . . . , tA
s (ā))),
= f B (π(t1A (ā)), . . . , π(tA s (ā))) by definition of an embedding,
= f B (t1B (π(ā)), . . . , tBs (π(ā))) by the inductive hypothesis,
= tB (π(ā)),
as required. 

Exercises
2.1 Write out a recursive definition of Lmon -terms.
2.2 Explain how the Lring -terms (−(x + 1) · y) and (x + x) + (y · −z) are built
up from the recursive definition of L-terms.
12 2 Terms

2.3 In the structure Rring , which elements of R are named by closed Lring -
terms?
2.4 Give three different Lring -terms which are interpreted as the same
polynomial function.
2.5 Define a polynomial function on R to be a function Rn → R of the form

m
d d
p(x1 , . . . , xn ) = ri x1i,1 · · · xni,n
i=1

for some ri ∈ R and some di, j ∈ N for i = 0, . . . , m and j = 1, . . . , n.

Show by induction on the construction of terms that every Lring -term
is interpreted in Rring as a polynomial function. Which polynomial
functions Rn → R are the interpretation of some Lring -term?
2.6 Give a complete proof (not using Proposition 2.4) that if A is an Lmon -
structure, π ∈ Aut(A), t is an Lmon -term, and ā ∈ Ar then π(tA (ā)) =
tA (π(ā)).
π
2.7 Show that if A −→ B is an L-homomorphism, then it preserves the
interpretations of L-terms.

2.8 An R-linear function is a function of the form f (x1 , . . . , xn ) = ni=1 ri xi ,

for some ri ∈ R. Design a language LR-VS in which the terms can
naturally be interpreted as R-linear functions.
2.9 Show how an R-vector space could be interpreted as an LR-VS -structure,
where LR-VS is the language of the previous question, and show that an
LR-VS -homomorphism of vector spaces is the same thing as a linear map.
 3
Formulas

Terms are certain strings of symbols from the language. As we have seen,
they are interpreted as elements of a structure or as functions on the structure.
Other strings of symbols of the language say something about structures or
about elements of the structure. For example, ((1 + 0) + 1) = 0 is a statement
about numbers (which is false in Zadgp but true in the cyclic group Z/2Z),
∃x[((x · x) + −(1 + 1)) = 0] asserts that there is a square root of 2 (which is
true in Rring but false in Zring ), and (x · x) < 1 is a statement about the number
x (which in Ro-ring will be true for some values of x and false for other values).

3.1 The Recursive Definition of Formulas

Definition 3.1 As for terms, the set of formulas of L is defined recursively.

(i) If t1 and t2 are terms, then (t1 = t2 ) is a formula.

(ii) If t1 , . . . , tn are terms and R is a relation symbol of L of arity n, then
R(t1 , . . . , tn ) is a formula.
(iii) If ϕ and ψ are formulas, then ¬ϕ and (ϕ ∧ ψ) are formulas.
(iv) If ϕ is a formula and x is a variable, then ∃x[ϕ] is a formula.
(v) Only something built from the above four clauses in finitely many steps
is a formula.

The formulas in clauses (i) and (ii) are called atomic formulas, because they
do not contain smaller formulas.

3.2 Free Variables and Scope

Variables play two different roles in formulas. For example, the formula

x<1+1

13
14 3 Formulas

says something about the variable x, namely that it is less than 2, whereas the
formula
∃x[0 < x]
says that there is some element which is greater than 0 and the variable x is just
a dummy. The formula has the same meaning as ∃y[0 < y], where x is replaced
by y. The difference is that in the second example, the variable x is quantified
over by the ∃ quantifier.
Definition 3.2 An instance of a variable x in a formula is a quantifier
instance if it occurs immediately after the ∃ symbol. Immediately following
the quantifier ∃x is the scope of the quantifier, which is enclosed in square
brackets [ ]. Any instance of x which occurs within the scope of the quantifier
∃x is called a bound instance of x, and the quantifier ∃x is said to bind it. Any
other instance of a variable is said to be a free instance. The variables which
have free instances in a formula are called its free variables.
For simplicity, we will usually assume that no variable occurs both free and
bound in the same formula. For example, the formula
(x  y ∧ ∃x[0  x])
should be rewritten as (x  y ∧ ∃z[0  z]).
Formulas with no free variables are particularly important, essentially
because they say something about a structure as a whole rather than about
some elements of it, so they have a special name.
Definition 3.3 A formula with no free variables is called a sentence.

3.3 Interpretation of Formulas

The purpose of formulas of L is to say something about L-structures. A
sentence should be either true or false in an L-structure. A formula with free
variables may be true for some interpretations of the variables and false for
others. We next explain how formulas are interpreted.
Definition 3.4 (Interpretation of formulas) Let ϕ be a formula of L and x̄ =
(x1 , . . . , xn ) a list of variables containing every free variable of ϕ. We also write
ϕ( x̄) for the formula with the list of variables. Let A be an L-structure and
ā = (a1 , . . . , an ) a list of elements of (the domain of) A. We define the notion
A |= ϕ(ā), read “A models ϕ(ā)” or “ϕ(ā) is true in A” or “A satisfies ϕ(ā)”
by recursion on formulas.
 3.4 Embeddings and Formulas 15

(i) A |= t1 (ā) = t2 (ā) iff t1A (ā) = t2A (ā), that is, iff the functions t1A and t2A
take the same value at ā.
(ii) A |= R(t1 (ā), . . . , tr (ā)) iff (t1A (ā), . . . , trA (ā)) ∈ RA .
(iii) A |= ¬ϕ(ā) iff A |= ϕ(ā), that is, A does not model ϕ(ā).
(iv) A |= (ϕ1 ∧ ϕ2 )(ā) iff A |= ϕ1 (ā) and A |= ϕ2 (ā).
(v) A |= ∃x[ϕ(x, ā)] iff there is some b ∈ A such that A |= ϕ(b, ā).

Definition 3.5 In the special case where ϕ is a sentence, if A |= ϕ, we say

that A is a model of ϕ. More generally, if Σ is a set of L-sentences, we say A
is a model of Σ and write A |= Σ if A is a model of every sentence in Σ.

This notion is where the name model theory comes from.

The definition above formally gives the formulas, which are strings of
symbols, the meaning that we already informally expect them to have. The
purpose of the formal definition is that we can use it to prove things that we
could not prove just from our informal understanding. Since formulas and
terms are defined recursively, the proofs go by induction on the recursive
definitions. The following results give good examples of this style of proof.

3.4 Embeddings and Formulas

π
Lemma 3.6 An embedding A −→ B of L-structures preserves atomic
formulas. That is, whenever ϕ( x̄) is an atomic L-formula and ā ∈ An ,

A |= ϕ(ā) if and only if B |= ϕ(π(ā)).

Proof Let ϕ be an atomic formula of the form R(t1 ( x̄), . . . , tr ( x̄)), and ā ∈
An . Let αi = tiA (ā), let βi = tiB (π(ā)), and write ᾱ = (α1 , . . . , αr ) and β̄ =
(β1 , . . . , βr ). Then Proposition 2.4 shows that π(ᾱ) = β̄. Then, by definition
of an embedding, ᾱ ∈ RA iff β̄ ∈ RB . Thus A |= R(t1 (ā), . . . , tr (ā)) iff B |=
R(t1 (π(ā)), . . . , tr (π(ā))), that is, A |= ϕ(ā) iff B |= ϕ(π(ā)) as required.
The proof is similar in the case when ϕ is an atomic formula of the form
t1 ( x̄) = t2 ( x̄). 
π
Proposition 3.7 An isomorphism A −→ B of L-structures preserves every
L-formula. That is, whenever ϕ( x̄) is an L-formula and ā ∈ An ,

A |= ϕ(ā) if and only if B |= ϕ(π(ā)).

Proof We proceed by induction on the construction of formulas. If ϕ is an
atomic formula, then the result is given by the preceding lemma.
16 3 Formulas

Suppose the result is true for ϕ and ψ.

¬ case: A |= ¬ϕ(ā) iff A |= ϕ(ā),

iff B | = ϕ(π(ā)) by induction hypothesis,
iff B |= ¬ϕ(π(ā)).
∧ case: A |= (ϕ ∧ ψ)(ā) iff A |= ϕ(ā) and A |= ψ(ā),
iff B |= ϕ(π(ā)) and
B |= ψ(π(ā)) by induction hypothesis,
iff B |= (ϕ ∧ ψ)(π(ā)).

∃x case: Suppose A |= ∃xϕ(x, ā). Then there is c ∈ A such that A |= ϕ(c, ā).
By induction, B |= ϕ(π(c), π(ā)). So B |= ∃xϕ(x, π(ā)). Conversely, suppose
B |= ∃xϕ(x, π(ā)). Then there is d ∈ B such that B |= ϕ(d, π(ā)). Since π is
an isomorphism, it has an inverse, π−1 . Let c = π−1 (d). Then π(c) = d, so, by
induction, A |= ϕ(c, ā). So A |= ∃xϕ(x, ā).
That completes all the necessary induction steps, one for each part of the
recursive definition of a formula. 

3.5 Abbreviations
Our formal language is very restricted. In practice, we adopt many
abbreviations which make it easier to use. We write

not equal (t1  t2 ) for ¬(t1 = t2 ),

or (ϕ ∨ ψ) for ¬(¬ϕ ∧ ¬ψ),
implies (ϕ → ψ) for (¬ϕ ∨ ψ),
iff (ϕ ↔ ψ) for ((ϕ → ψ) ∧ (ψ → ϕ)),
for all ∀x[ϕ] for ¬∃x[¬ϕ].
We could have taken all these symbols as part of our formal language, but
if we did that, then our proofs about the formal language would be longer, so
instead we introduce them as abbreviations.
We will omit brackets or sometimes add extra brackets where this improves
clarity and where no ambiguity can arise. We will also write ∃x1 · · · xn [ϕ] for

∃x1 [∃x2 [· · · ∃xn [ϕ] · · · ]] and ni=1 ϕi for ϕ1 ∧ ϕ2 ∧ · · · ∧ ϕn . We adopt the usual
convention of writing (x + y) instead of +(x, y), and similarly other standard
conventions from algebra. Our general philosophy is to use the usual notation
from mathematical practice where possible, but be careful to see that it can, in
principle, be translated into our formal language.
 Exercises 17

Exercises
3.1 Write down an Lring -formula in which the variables x, y, and z appear
free and the variables u and v are bound.
3.2 In the language L with a binary relation symbol < and constant symbols
for 0, 1, 2, 3, 4, 5, 6, write L-sentences expressing the following about the
L-structure on the integers, Z.

(a) There is no greatest integer.

(b) There is no integer between 0 and 1.
(c) For any two distinct integers, one is less than the other.
(d) There are exactly 2 integers between 3 and 6.

3.3 Let L= be the language with no relation, function, or constant symbols.

For n ∈ N+ , show there are L= -sentences ϕn and ϕ=n such that for any
structure A we have A |= ϕn iff A has at least n elements and A |= ϕ=n
iff A has exactly n elements.
3.4 Consider the structure N = N; +, ·. Write down

(a) a formula ϕ(x) in the language +, · with one free variable such that
for any n ∈ N, N |= ϕ(n) iff n = 2;
(b) a formula ξ(x, y) with two free variables such that for any
n1 , n2 ∈ N, N |= ξ(n1 , n2 ) iff n1 < n2 ;
(c) a formula δ(x, y) with two free variables such that for any
n1 , n2 ∈ N, N |= δ(n1 , n2 ) iff n1 |n2 ;
(d) a formula ψ(x) with one free variable such that for any n ∈ N,
N |= ψ(n) iff n is a prime number;
(e) a sentence stating that there are infinitely many pairs of prime
numbers which differ by 2.

You can use abbreviations, provided you explain what they are.
3.5 Consider the Lring -formula ϕ(x) given by ∃y[(y · y) + 1 = x]. For what
values of x do we have Rring |= ϕ(x)? For what x do we have Zring |= ϕ(x)?
3.6 Express each of the statements in Exercise 1.2 as L-sentences for an
appropriate choice of language L.
3.7 A formula is said to be quantifier-free if it does not use the symbol ∃ (or
the abbreviation ∀).

(a) Write down a recursive definition of quantifier-free formulas in the

style of Definition 3.1.
18 3 Formulas

(b) Prove by induction on your recursive definition that every

π
embedding A −→ B of L-structures preserves quantifier-free
formulas. That is, whenever ϕ( x̄) is a quantifier-free L-formula and
ā ∈ An ,
A |= ϕ(ā) if and only if B |= ϕ(π(ā)).
3.8 A positive quantifier-free formula is one defined using ∧ , ∨ , and atomic
formulas, but not ¬, → , or quantifiers. Write down a recursive definition
of positive quantifier-free formulas, and prove that they are preserved
under L-homomorphisms.
 4
Definable Sets

In this chapter we introduce the main idea of the book, that of a definable set.

4.1 Examples of Definable Sets

Definition 4.1 In an L-structure A, a subset S of An is said to be a definable
set if there is an L-formula ϕ(x1 , . . . , xn ) such that S = {ā ∈ An | A |= ϕ(ā) }.
The formula ϕ( x̄) defines the set S , and we sometimes write S as ϕ(A).

Example 4.2 In Ro-ring , the interval (0, 2) is defined by the formula

(0 < x) ∧ (x < (1 + 1)),

the pair {−1, 3} is defined by the formula

(x + 1 = 0) ∨ (x = (1 + 1) + 1),

and the formula

(0 = x ∨ 0 < x) ∧ (x < 1 + 1) ∧ (y < 0) ∧ (−(1 + 1) < y)

defines the rectangle [0, 2) × (−2, 0).

If t(x1 , . . . , xn ) is an L-term, then the formula y = t(x1 , . . . , xn ) defines the

graph of the function tA in the structure A. However, other functions may also
be definable.

Definition 4.3 Let A be an L-structure, n, m ∈ N+ , D ⊆ An , C ⊆ Am , and

f : D → C a function. Then f is a definable function (with respect to the
L-structure A) if D, C and the graph of f , that is {(a, f (a)) | a ∈ D }, are
definable sets.

19
20 4 Definable Sets

Note that if the graph of f is defined by the formula ϕ(x, y), then the domain
of f is defined by ∃y[ϕ(x, y)] and the image is defined by ∃x[ϕ(x, y)].
Example 4.4 In Ro-ring , the absolute value function R → R; x → |x| is defined
by the formula
((x < 0) ∧ (y = −x)) ∨ (¬(x < 0) ∧ (y = x)).
Lemma 4.5 If S 1 , S 2 ⊆ An are definable, then S 1 ∩ S 2 , S 1 ∪ S 2 , and An  S 1
are definable. Furthermore, if S ⊆ An+m is definable, then the projection
  
proj(S ) = ā ∈ An  There is some b̄ ∈ Am such that (ā, b̄) ∈ S
is also definable.
Proof Suppose S 1 is defined by ϕ( x̄) and S 2 is defined by ψ( x̄). Then S 1 ∩ S 2
is defined by (ϕ( x̄) ∧ ψ( x̄)) and so on. 
Figure 4.1 illustrates why existential quantification corresponds to projec-
tion. In the structure Ro-ring , let ϕ(x, y) be the formula
(x − 4)2 + (y − 3)2 < 4
which defines a disc of radius 2, centred at (4, 3) in the xy-plane. We have
used subtraction and squaring in the formula, although they are not symbols
in the language Lo-ring . Since they are definable functions, we can use them
as abbreviations. The formula ∃y[ϕ(x, y)] defines the projection to the x-axis,
which is the interval (2, 6).

0 1 2 3 4 5 6 7

Figure 4.1 Existential quantification as projection.

 4.2 Preservation of Definable Sets under Automorphisms 21

The subsets of Rn which are defined by Lo-ring -formulas without quantifiers,

that is, by equations and inequations of polynomials, are called semi-algebraic
sets. In fact, for the structure Ro-ring , using quantifiers does not give any more
definable sets. So, for example, one can prove that Z is not a definable subset
of R in Ro-ring . See Proposition 22.3 for more detail, although a complete proof
is beyond the scope of this book.

4.2 Preservation of Definable Sets under Automorphisms

A very important part of model theory is the study of the definable sets of a
structure A. To show a set is definable, we can just come up with a formula, but
it is more difficult to show that no formula defines a set. So it is important to
have methods to show that certain subsets are not definable. We now develop
one such method.
Proposition 4.6 If S ⊆ An is a definable set and π is an automorphism of A,
then π(S ) = S . We say that S is preserved under all automorphisms.
Proof By π(S ), we mean {π(ā) | ā ∈ S }, the image of S under the function π.
Suppose S is defined by the formula ϕ( x̄), and let ā ∈ S . Then A |= ϕ(ā),
so applying Proposition 3.7, we see that A |= ϕ(π(ā)) because π is an
automorphism, and hence π(ā) ∈ S . So π(S ) ⊆ S . Now π−1 is also an
automorphism of A, so π−1 (ā) ∈ S . But ā = π(π−1 (ā)), so ā ∈ π(S ). Thus
S ⊆ π(S ), and hence π(S ) = S , as required. 
Example 4.7 In the structure Radgp , the order < is not definable. That is,
the set of pairs (x, y) ∈ R2 | x < y is not definable. To see this, consider the
function π : R → R given by π(x) = −x. It is an embedding of Radgp into itself
because π(0) = 0 and

π(x + y) = −(x + y) = −x + (−y) = π(x) + π(y).

Furthermore, π ◦ π is the identity map on R, so π is its own inverse. Hence

π is an automorphism of Radgp . However, we have for example 0 < 1, but
π(0) ≮ π(1), and so < is not preserved under π, and hence by Proposition 4.6 it
is not definable in Radgp .

It is natural to ask if the converse to Proposition 4.6 is true. If S ⊆ An is a

subset which is not definable, must there be an automorphism π of A such that
π(S )  S ? The answer is no.
22 4 Definable Sets

Lemma 4.8 There are no non-trivial automophisms of Ro-ring .

Proof Suppose π is an automorphism of Ro-ring . Then π(0) = 0. Furthermore,
for any n ∈ N+ , we have π(1 + · · · + 1) = π(1) + · · · + π(1) = n, and then
 
n n
π(−n) = −π(n) = −n, so π fixes every integer. If q = n/m ∈ Q, we have
π(q) · π(m) = π(n), so π(q) · m = n, but q is the only solution to the equation
x · m = n, which is an Lo-ring -formula, so we must have π(q) = q. So π fixes
every rational number. The rationals are dense in the reals and π fixes the order
<, so it must fix every real number. Hence π is the identity map on R. 
There are uncountably many subsets of Rn , and only countably many
formulas, so only countably many definable subsets, so most of the subsets of
Rn are not definable. However, all subsets of Rn are preserved by the identity
automorphism and hence by all automorphisms of Ro-ring .

Exercises
4.1 Complete the proof of Lemma 4.5.
4.2 Show that addition and multiplication are not definable in R< .
4.3 Show that the order < is not definable in Zadgp but is definable in Nadmon .
4.4 Show that the order < is definable in Rring . Deduce that every subset of
Rn which is definable in Ro-ring is also definable in Rring .
4.5 Show that the set N and the order < are definable in Zring . [Hint: Use
Lagrange’s 4-square theorem.]
4.6 Show that negation is a definable function in R; +, 0 and in Z; +, 0
but that multiplication is not definable in either structure.
√
4.7 Show that the square root function R+ → R+ ; x → x is definable in
Ro-ring .
4.8 Give an example of a definable function in Zring which is not defined by
the interpretation of any term.
4.9 Show that the only automorphism of the structure Ns-ring is the identity
map.
4.10 Consider the expansion of the real ordered field Ro-ring by a unary
function f . Write down a formula ϕ(x) such that whatever the function
f is, ϕ(x) defines the set of real numbers at which f is continuous.
 Exercises 23

4.11 Find all the subsets of Q which are definable in Qadgp (by a formula with
one free variable) and prove that they are the only ones.
4.12 Sketch a proof of Proposition 4.6 starting from the relevant definitions
and including the statements and sketch proofs of any necessary lemmas.
You should explain what methods of proof are used, what all the key
ideas are, and how they fit together, but you do not need to give all the
details of the inductive proofs.
 5
Substructures and Quantifiers

In the previous chapter we showed that every definable set is preserved under
automorphisms. Here we give more subtle preservation theorems which show
that formulas constructed with only certain quantifiers are preserved under
taking substructures or embeddings.

5.1 Substructures
The field of rational numbers, Qring , is a subfield of the field of real numbers,
Rring . Similarly, Zring is a subring of Qring . More generally, recall from
Chapter 1 the notion of a substructure of an L-structure.

Definition 5.1 Suppose that A and B are L-structures. Then A is a

substructure of B and B is an extension of A if the domain of A is a subset of
the domain of B and all the symbols of L are interpreted in A as the restrictions
of their interpretations in B. More precisely:
A B
• for each constant symbol c, c = c ;
A B
• for each function symbol f and each ā in A, f (ā) = f (ā); and
A B
• for each relation symbol R and each ā in A, ā ∈ R if and only if ā ∈ R .
Equivalently, A is a subset of B, and the inclusion map of A into B is an
L-embedding. We write A ⊆ B to mean that A is an L-substructure of B.

5.2 Existential and Universal Formulas

For A ⊆ B, we now consider how the interpretations of formulas are related in
the two structures. Consider the following sentences, at first written in English
sentences, not as L-sentences, for a first-order language L.

24
 5.2 Existential and Universal Formulas 25

(i) ‘All numbers are greater than or equal to 0.’

(ii) ‘There is a number which when added to itself makes 1.’
(iii) ‘All numbers have square roots.’
The first sentence is true when we interpret “number” as an element of the
structure N; <, 0, and it is also true in the substructure E; <, 0 consisting of
the even natural numbers, but it is false in the extension Z; <, 0. Intuitively,
it makes sense, because if all natural numbers have a property, then certainly
all the even natural numbers have that property, but perhaps some numbers in
a larger set (the integers) might not have that property. For L = 0, <, we can
write the sentence as the L-sentence

∀x[0 < x ∨ x = 0]

with a universal quantifier. The second sentence is true in Qring and also in the
extension Rring but false in the substructure Zring . It makes sense, because once
the number (1/2) with this property is there in Q, it is still in any extension of
Q, but it might not be there in a substructure, in this case, Z. We can write it as

∃x[x + x = 1]

using an existential quantifier.

The third sentence is true in the complex field Cring . It appears to be
another universal statement since it is about ‘all numbers’, but it is false in
the substructure Rring . It is also false in the ring C[X] of polynomials over C,
which is an extension ring of C. In fact, when we write it as the Lring -sentence
∀x∃y[y · y = x], we see that it uses both universal and existential quantifiers.
The first two sentences are examples of a general phenomenon, that
universal statements true in a structure are also true in any substructure, and
that existential statements true in a structure are also true in any extension. The
pattern also holds for formulas with free variables. In our convention, universal
quantifiers are just abbreviations and can be replaced by existential quantifiers,
but only by also using negation symbols. So we need to be very careful about
exactly what we mean by universal and existential statements.

Definition 5.2 (Quantifier-free formula) A quantifier-free formula is a for-

mula which does not involve any quantifiers. In order to prove things about
them, it is useful also to give a recursive definition.
(i) An atomic formula is a quantifier-free formula.
(ii) If ϕ is a quantifier-free formula, then ¬ϕ is a quantifier-free formula.
(iii) If ϕ and ψ are quantifier-free formulas, then (ϕ ∧ ψ) is a quantifier-free
formula.
26 5 Substructures and Quantifiers

(iv) Only a formula constructed by the above rules in finitely many steps is a
quantifier-free formula.
Similarly, we can give recursive definitions of existential formulas and
universal formulas.
Definition 5.3 (Existential formula)
(i) A quantifier-free formula is an existential formula.
(ii) If ϕ is an existential formula and x is a variable, then ∃x[ϕ] is an
existential formula.
(iii) Only a formula constructed by the above rules in finitely many steps is
an existential formula.
Definition 5.4 (Universal formula)
(i) A quantifier-free formula is a universal formula.
(ii) If ϕ is a universal formula and x is a variable, then ∀x[ϕ] is a universal
formula.
(iii) Only a formula constructed by the above rules in finitely many steps is a
universal formula.
So any existential formula has the form ∃y1 · · · yn [ψ] where ψ is a quantifier-
free formula, and any universal formula has the form ∀y1 · · · yn [ψ]. From our
definition of ∀y as an abbreviation for ¬∃y¬, we can write ∀y1 · · · yn [ψ] as
¬∃y1 ¬¬∃y2 ¬ · · · ¬∃yn [¬ψ] and cancelling the double negations it becomes
¬∃y1 · · · yn [¬ψ], which is the negation of an existential formula.

5.3 Preservation Laws

Now we can prove that existential statements are indeed ‘preserved in exten-
sions’ and universal statements are ‘preserved in substructures’.
Lemma 5.5 If A is an L-substructure of B, ϕ( x̄) is a quantifier-free L-formula,
and ā ∈ A, then
A |= ϕ(ā) if and only if B |= ϕ(ā).
Proof If ϕ( x̄) is an atomic L-formula, the result is immediate from Lemma
3.6. More generally, it is Exercise 3.6. 
Proposition 5.6 Let B be an L-structure and A ⊆ B an L-substructure. Let
ϕ( x̄) be an existential L-formula, and suppose ā ∈ An is such that we have
A |= ϕ(ā). Then B |= ϕ(ā).
 Exercises 27

Proof We proceed by induction on the construction of existential formulas.

If ϕ( x̄) is a quantifier-free formula, then the result is Lemma 5.5. Suppose ϕ( x̄)
is ∃y[ψ( x̄, y)]. By assumption, A |= ∃y[ψ(ā, y)], so there is b ∈ A such that
A |= ψ(ā, b). Then, by induction hypothesis, B |= ψ(ā, b). So B |= ∃y[ψ(ā, y)],
as required. 
Proposition 5.7 Let B be an L-structure and A ⊆ B an L-substructure. Let
ϕ( x̄) be a universal L-formula, and suppose ā ∈ A is such that B |= ϕ(ā). Then
A |= ϕ(ā).
Proof We could do a simple inductive proof as for the existential case, but
instead we will deduce it from that case. Let θ be an existential formula such
that ϕ is ¬θ. Suppose that B |= ϕ(ā). Then B |= θ(ā), so, by Proposition 5.6,
A |= θ(ā). So A |= ϕ(ā), as required. 

Exercises
5.1 If L is a language with only relation symbols and A is an L-structure,
show that every subset of A is naturally an L-substructure of A.

5.2 If L has function and constant symbols, which subsets of the domain of
an L-structure correspond to L-substructures?

5.3 Show that an Lring -substructure of a field is a subring. Design a language

Lfield for fields such that Lfield -substructures of fields are subfields.

5.4 Write the axiom for a group which states that every element has an
inverse both as an Lgp -sentence and as an Lmon -sentence. Which of the
axioms is a universal sentence? Show that if G is a group considered
as an Lgp -structure, then its Lgp -substructures are its subgroups and its
Lmon -substructures are submonoids.

5.5 Show that every subgroup of an abelian group is also abelian.

5.6 Show that there is no universal Lring -sentence which asserts that every
non-zero element of a field has a multiplicative inverse.

5.7 Give a direct proof of Proposition 5.7, not using Proposition 5.6.

5.8 Suppose that A and B are L-structures and A ⊆ B. Let ϕ(x1 , . . . , xn ) be

an L-formula. Show that:
28 5 Substructures and Quantifiers

(a) if ϕ is quantifier-free, then ϕ(A) = ϕ(B) ∩ An ;

(b) if ϕ is existential, then ϕ(A) ⊆ ϕ(B) ∩ An ; and
(c) if ϕ is universal, then ϕ(A) ⊇ ϕ(B) ∩ An .
5.9 Suppose we have a chain of L-structures
A1 ⊆ A2 ⊆ A3 ⊆ · · · ⊆ An ⊆ · · ·
indexed by n ∈ N+ , and A is the union of the chain. Explain how A can
be made into an L-structure which is an extension of each An . Let ϕ( x̄)
be a ∀∃-formula, that is, an L-formula of the form ∀ȳ∃z̄ψ( x̄, ȳ, z̄), with ψ
quantifier-free. Show that if ā ∈ A1 such that for every n ∈ N+ we have
An |= ϕ(ā), then A |= ϕ(ā).
5.10 A positive L-formula is defined using ∧ , ∨ , ∃, and ∀, but without
negations (except implicitly in the abbreviations ∨ and ∀).
(a) Give a recursive definition of positive L-formulas.
(b) Show that positive L-formulas are preserved under surjective
homomorphisms of L-structures.
(c) What formulas are preserved under arbitrary (not necessarily
surjective) homomorphisms?
 Part II

Theories and Compactness

In this part of the book we introduce the first part of our programme for
understanding a mathematical structure: finding axioms for the theory of the
structure. The programme is illustrated with the Peano axioms for Ns-ring and
by the axioms for algebraically closed fields and for real-closed fields for Cring
and Ro-ring , respectively. At this stage, we lack the tools to give proofs for
these examples, or indeed for any other examples. We then introduce the most
powerful of the model-theoretic tools, the compactness theorem, together with
the method of changing the language by adding new constant symbols. These
methods are used to construct new models of the theory of an infinite structure.
Further applications of the compactness theorem are given in the context of
axiomatisable classes. A brief discussion of cardinal arithmetic and Henkin’s
proof of the compactness theorem complete Part II.

29
 6
Theories and Axioms

In this chapter we move the focus from individual formulas and sentences
to the set of al those sentences which are true in a structure, its theory. To
describe a theory, we need to have a subset of it we can write down, that is, an
axiomatisation.

6.1 Theories
Recall from Definition 3.5 that if A is an L-structure and Σ is a set of
L-sentences, we write A |= Σ, read ‘A models Σ’, to mean that every sentence
σ in Σ is true in A.

Definition 6.1 Let C be a class of L-structures. The theory of C, Th(C), is the

set of all L-sentences which are true in every A ∈ C. When C = {A}, a single
structure, we write Th(A) for Th({A}).

For any class of structures C, the theory Th(C) has the property of being
deductively closed, which we now explain.

Definition 6.2 (Entailment) Let Σ be a set of L-sentences and ϕ be an

L-sentence. We say Σ entails ϕ or ϕ is a logical consequence of Σ and write
Σ  ϕ if every model of Σ is also a model of ϕ. If Φ is also a set of sentences,
we write Σ  Φ to mean that for all ϕ ∈ Φ, Σ  ϕ.

Often, the symbol |= is used instead of  in this context, as well as when a

structure is on the left-hand side. Then the symbol  is used instead only for a
syntactic concept of provability within a given deduction system. Since we do
not use the concept of provability, we will follow Wilfrid Hodges [Hod93] and
use the symbols this way.

31
32 6 Theories and Axioms

Definition 6.3 A set Σ of L-sentences is deductively closed if, for any

L-sentence σ, if Σ  σ then σ ∈ Σ. The deductive closure of Σ is {σ | Σ  σ }.
A set Σ is satisfiable if there is an L-structure A such that A |= Σ, that is, if
Σ has a model, and is unsatisfiable otherwise.
Lemma 6.4 For any class C of structures, Th(C) is deductively closed.
Proof Suppose σ is an L-sentence and Th(C)  σ. If A ∈ C, then A |= Th(C),
so by the definition of , we have A |= σ. Hence σ ∈ Th(C). So Th(C) is
deductively closed. 
Definition 6.5 (Theories) A theory (more precisely, a first-order L-theory) is
a satisfiable and deductively closed set of L-sentences.

Some authors use the word theory to mean any set of sentences, and others
require satisfiability but not deductive closedness, or vice versa. In practice it
rarely matters. A minor quirk of our convention is that if C is the empty set of
L-structures, then Th(C) is the set of all L-sentences. So in this case (and only
in this case), Th(C) is not satisfiable and so is not a theory.

6.2 Examples of Axioms

In practice (and often even in principle) we cannot write down or even describe
all the sentences in Th(C). Instead we want to have a set of axioms for C, which
means a set of sentences Σ ⊆ Th(C) such that an L-structure A is in C iff
A |= Σ. To be useful, we should be able to write these axioms down, or at least
describe them. Often a class of structures is defined by a list of axioms, and for
illustration, we give some familiar algebraic examples.

Example 6.6 (Axioms for the theory of groups) We can write the following
axioms for groups in the language Lgp = ·, (−)−1 , 1:

G1. ∀xyz[(x · y) · z = x · (y · z)],

G2. ∀x[x · 1 = x ∧ 1 · x = x], and
G3. ∀x[x · x−1 = 1 ∧ x−1 · x = 1].

The theory of groups is the deductive closure of these three axioms. In other
words, it is the set of all Lgp -sentences which are true of all groups. So, for
example, the sentence

∀xy[x−1 · (x · y) = y]
 6.2 Examples of Axioms 33

follows easily from the axioms so is in the theory of groups, but the sentence

∃x[x  1 ∧ x · x = 1]

is true in some groups (for example C2 , the cyclic group of order 2), but not
in other groups (for example C3 , the cyclic group of order 3), so is not in the
theory of groups.

Example 6.7 (Abelian groups) A group is abelian if it satisfies also

AG. ∀xy[x · y = y · x],

so the theory of abelian groups is the deductive closure of {G1, G2, G3, AG}.

We have written four axioms for abelian groups, but we could write a single
axiom with the same meaning, the conjunction of those four axioms:

G1 ∧ G2 ∧ G3 ∧ AG.

The natural language in which to axiomatise rings or fields is Lring =

+, ·, −, 0, 1, the language of rings.

Example 6.8 (Rings) We have the axioms for abelian groups, but now in the
additive language Ladgp = +, −, 0:

R1. ∀xyz[(x + y) + z = x + (y + z)],

R2. ∀x[x + 0 = x ∧ 0 + x = x],
R3. ∀x[x + (−x) = 0 ∧ (−x) + x = 0], and
R4. ∀xy[x + y = y + x].

Then there are axioms for commutative monoids (the abelian group axioms
except for the inverse) in the language Lmon = ·, 1:

R5. ∀xyz[(x · y) · z = x · (y · z)],

R6. ∀x[x · 1 = x ∧ 1 · x = x],
R7. ∀xy[(x · y) = (y · x)],

and a distributivity axiom;

R8. ∀xyz[x · (y + z) = (x · y) + (x · z)].

Together these eight sentences axiomatise the theory of (commutative) rings.

Example 6.9 (Fields) A field is a ring satisfying two additional axioms. First,
an axiom to rule out the zero ring,

F1. 0  1,
34 6 Theories and Axioms

and second an axiom which says that every non-zero element of the field has a
multiplicative inverse:

F2. ∀x[x = 0 ∨ ∃y[x · y = 1]].

So the axioms R1–R8, F1, and F2 together axiomatise the theory of fields.

Remark 6.10 Notice that we put −, the additive inverse, as a function symbol
in the language. However, we cannot easily do the same for the multiplicative
inverse, because it is not defined at 0. We could put a unary function symbol for
it in the language, but we would have to define 0−1 somehow arbitrarily, and
then make sure that our axioms describing how (−)−1 works in fields took this
case into account. We follow the usual practice of not introducing a symbol
for it.

6.3 Peano Arithmetic and Complete Theories

One of the most important structures in mathematics is Ns-ring = N; +, ·, 0, 1,
the natural numbers with both addition and multiplication. In mathematical
logic, the study of this structure and its theory (and related theories) is often
called arithmetic, and Th(Ns-ring ) is sometimes called true arithmetic. We can
try to write down axioms for Th(Ns-ring ). The most common and useful axioms
are those written down by the Italian mathematician Peano. In modern form,
they are as follows:

P1. ∀x[x + 1  0],

P2. ∀x[x  0 → ∃y[x = y + 1]],
P3. ∀xy[x + 1 = y + 1 → x = y],
P4. ∀x[x + 0 = x],
P5. ∀xy[x + (y + 1) = (x + y) + 1],
P6. ∀x[x · 0 = 0],
P7. ∀xy[x · (y + 1) = (x · y) + x]

and for each L-formula ϕ(x), with one free variable x, the axiom

Pϕ. (ϕ(0) ∧ ∀x[ϕ(x) → ϕ(x + 1)]) → ∀xϕ(x).

The last axiom scheme tries to capture the idea of induction. Induction says
that if 0 has a certain property, and whenever n has that property, then so does
n + 1, then every natural number has that property. The axiom scheme says that
for each property which can be expressed as a first-order formula.
 6.4 Elementary Equivalence 35

This set of axioms, or its deductive closure, is called the theory of Peano
arithmetic, or PA. It is easy to check each axiom individually to see that
Ns-ring |= PA, and hence that PA ⊆ Th(Ns-ring ). Some of the axioms look
a little strange, for example (P5) is a very limited form of the associativity
axiom, and we could replace it with the full associativity axiom. In fact, the full
associativity of addition can be proved from all the axioms using the induction
scheme. The reason for using the limited form is that it is, at first sight, weaker,
and Peano was trying to write down the weakest set of axioms from which
everything else could be deduced. Can everything else about arithmetic be
deduced from the axioms of PA?

Definition 6.11 A set Σ of L-sentences is said to be complete if, for any

L-sentence σ, either Σ  σ or Σ  ¬σ.
Lemma 6.12 Let σ be any Ls-ring -sentence. Then either σ ∈ Th(Ns-ring ) or
¬σ ∈ Th(Ns-ring ). In particular, Th(Ns-ring ) is complete.
Proof If Ns-ring |= σ, then σ ∈ Th(Ns-ring ). Otherwise, Ns-ring |= σ, so
Ns-ring |= ¬σ, so ¬σ ∈ Th(Ns-ring ). Now Th(Ns-ring ) is complete, because if
σ ∈ Th(Ns-ring ), then certainly Th(Ns-ring )  σ. 
More generally, Th(A) is complete for any L-structure A. So the question
is whether PA is complete. One of the great theorems of twentieth-century
mathematics is Gödel’s first incompleteness theorem, which states that PA is
not complete. In fact, Gödel showed that whatever set of axioms about N we
write down explicitly, it will not axiomatise all of true arithmetic.

6.4 Elementary Equivalence

Definition 6.13 Two L-structures A and B are elementarily equivalent,
written A ≡ B, iff Th(A) = Th(B), that is, for every L-sentence σ, A |= σ iff
B |= σ.

So elementarily equivalent L-structures cannot be distinguished by formal

sentences in the language L.
Lemma 6.14 If A and B are isomorphic L-structures, then they are elemen-
tarily equivalent.
π
Proof Let A −→ B be an isomorphism, and let σ be an L-sentence. Since σ
has no free variables, we can apply Proposition 3.7 to σ with the empty list of
36 6 Theories and Axioms

variables to deduce that A |= σ if and only if B |= σ. This applies for every

L-sentence, so A ≡ B. 
For finite structures (structures whose domain is a finite set), isomorphism
and elementary equivalence are actually the same. (See the exercises.) How-
ever, we will see later that if A is an infinite structure, then there are infinitely
many other structures which are elementarily equivalent to A (and thus to each
other) but which are pairwise non-isomorphic.

Exercises
6.1 Given a set Σ of L-sentences, let Mod(Σ) be the class of L-structures A
such that A |= Σ. Show that Th(Mod(Σ)) is the deductive closure of Σ.
6.2 Show that a theory T is complete if and only if, whenever A and B are
models of T , then A ≡ B.
6.3 For each of the five structures Ns-ring , Zs-ring , Qs-ring , Rs-ring , Cs-ring , write
down a sentence which is true in that structure but not in any of the other
four structures.
6.4 Show that the theory of fields is not complete.
6.5 Show that there must be a model of Peano arithmetic which is not
isomorphic to the standard model, Ns-ring .
6.6 Show that in any model A of PA, the function +A is associative.
6.7 Let L = R, f, c, where R is a binary relation symbol, f is a binary
function symbol, and c is a constant symbol. Let A be an L-structure
with three elements. Show that there is a single L-sentence σA such that
for any L-structure B, if B |= σA , then B  A. How would you change
your proof to work for an arbitrary finite structure in an arbitrary finite
language?
6.8  -formula such that for every n ∈ N, the
Suppose that ϕ(x, y) is an Ls-ring
subset ϕ(Ns-ring , n) = x ∈ N  Ns-ring |= ϕ(x, n) of N is non-empty. Let
M |= Th(Ns-ring ), and let a ∈ M. Show that ϕ(M, a) has a least element.
6.9 Can you find two structures which are elementarily equivalent but not
isomorphic? [This is quite difficult without some theory. Later we will
see many ways to answer the question.]
 7
The Complex and Real Fields

One of the main problems addressed in this book is to start with a structure A
and to write down a complete axiomatisation of its theory. The general strategy
is simple. To find a candidate set of sentences Σ which might axiomatise
Th(A), start by writing down some sentences capturing properties of A. If
you can find another model B |= Σ and a sentence σ such that A |= σ and
B |= σ, then add σ to Σ to get a better candidate. When you can no longer do
this, you have a plausible candidate for a complete axiomatisation.
Actually proving that a given axiomatisation is complete is, in general,
a difficult problem. Gödel’s First Incompleteness Theorem shows that it is
impossible to give an explicit complete axiomatisation for Th(Ns-ring ). Using
this result, the same can also be shown for Th(Zring ) and Th(Qring ).
In this chapter we will describe complete axiomatisations for the complex
and real fields. In Part III of the book we will develop one method of proving
the completeness of an axiomatisation, the Łos–Vaught test. We will apply this
method to prove that our axiomatisation for the complex field is complete in
Chapter 30. The proof for the real field is similar but more involved and can
be found in the books of Marker [Mar02, Section 3.3] and of Poizat [Poi00,
Section 6.6].
The real field is a good illustration of the limitations of first-order logic.
The obvious axiomatisation is in second-order logic, where quantification over
subsets is allowed. We have to find a different axiomatisation to stay within
first-order logic.

7.1 The Complex Field

To axiomatise Th(Cring ), we start with the axioms of fields (see Example 6.9).
The field C has characteristic 0, which means that it satisfies the axioms

37
38 7 The Complex and Real Fields

+ 1 + ··· + 1  0
1
n

for each n ∈ N+ . Since there are fields with non-zero characteristic, we need
these axioms.
In Lring we can express statements about polynomials. The fundamental
theorem of algebra states that every non-constant polynomial with coefficients
from C has a root in C. This fact can be written as a list of sentences in Lring ,
taking the sentence σn given by

∀y0 . . . yn−1 ∃x[xn + yn−1 · xn−1 + · · · + y1 · x + y0 = 0]

for each n ∈ N+ , where we use the usual abbreviation of xn for multiplying n

copies of x together, and omit unnecessary brackets.
Theorem 7.1 The axioms of fields of characteristic 0 together with the axioms
σn above axiomatise the complete theory of the complex field Cring .
The proof will be given in Chapter 30. These axioms have become so useful
that they have a name.

Definition 7.2 A field satisfying the sentences σn for all n ∈ N+ is said to be

algebraically closed.

The complete theory of Cring is known as ACF0 , the theory of alge-

braically closed fields of characteristic 0. Since the theory is complete, any
two algebraically closed fields of characteristic 0 satisfy exactly the same
Lring -sentences.

7.2 Complete Ordered Fields

We now turn to the real field.

Definition 7.3 (Ordered field) The language of ordered rings is

Lo-ring = Lring ∪ {<}. In this language, the axioms of ordered fields are the
axioms of fields together with axioms stating that < is a linear order:
O1. (Transitivity) ∀xyz[(x < y ∧ y < z) → x < z];
O2. (Linearity) ∀xy[x < y ∨ x = y ∨ y < x];
O3. (Irreflexivity) ∀x[¬x < x];
and axioms relating the field structure with the order:
O4. (Additivity) ∀xyz[x < y → x + z < y + z]; and
O5. (Multiplicativity) ∀xyz[(x < y ∧ 0 < z) → x · z < y · z].
 7.3 First-Order Axioms for the Real Field 39

An ordered field F is necessarily of characteristic 0, since from the axioms

we can deduce
0 < 1 < 1 + 1 < 1 + 1 + 1 < ··· ,
so we have N sitting inside F as a subset in the obvious way.
The real field satisfies a further property: the completeness of its ordering.
Definition 7.4 An ordered set F is complete if every non-empty subset S ⊆ F
which has an upper bound has a least upper bound.
This is a different usage of the word complete from completeness of a
theory. A simple consequence of the completeness property for an ordered
field is that it satisfies the property of Archimedes.
Proposition 7.5 If F is a complete ordered field, then it is Archimedean, that
is, the subset N of F is not bounded above.
Proof Suppose for a contradiction that N is bounded above. Then, by the
completeness property, there is a least upper bound b for N. Now b − 1 < b
so b − 1 is not an upper bound for N; hence there is a natural number n ∈ N
with b − 1 < n. But then n + 1 ∈ N and b < n + 1, contradicting b being an
upper bound for N. 
A very important theorem for the foundations of analysis is that there is
only one complete ordered field.
Fact 7.6 Up to isomorphism, there is exactly one complete ordered field.
The unique (up to isomorphism) model of the axioms of complete ordered
fields is the field of real numbers. There are three common ways to build
models for these axioms, that is, to construct the field of real numbers. The
most familiar way is to construct R as the field of decimal numbers. It can also
be constructed via Dedekind cuts or by Cauchy sequences. These methods
give three different models of the axioms, but by the above fact, they are
all isomorphic, so it does not matter which of the three (or any other model
constructed by a different method) is used.

7.3 First-Order Axioms for the Real Field

The axioms for ordered fields are first-order sentences of the language Lo-ring .
We can write the completeness axiom in logical symbols. The formula
∀x[x ∈ S → x  y]
40 7 The Complex and Real Fields

states that y is an upper bound for S . Abbreviating it by UB(S , y), the formula
UB(S , z) ∧ ∀y[UB(S , y) → z  y]
states that z is the least upper bound for S . Abbreviating that by LUB(S , z), the
sentence

(∀S ⊆ F) ((S  ∅) ∧ ∃y[UB(S , y)]) → ∃z[LUB(S , z)]
expresses the completeness axiom. However, this is not a formula in the
language Lo-ring as we have defined it because we have also used the symbols ∈,
⊆, and ∅, but most importantly, we have a variable S which does not range over
the elements of F but over its subsets, and then we have quantified over that.
This is not allowed in first-order logic, but it is allowed in so-called second-
order logic. While we can make sense of what this axiom means, it is actually
not expressible by any set of first-order sentences. (See Exercise 9.5.) It follows
that there are models of the complete first-order theory of the real ordered field,
Th(Ro-ring ), which are necessarily ordered fields but which are not complete as
ordered sets.
However, it turns out that we can axiomatise the first-order theory
Th(Ro-ring ) in a way closer to the axiomatisation of algebraically closed fields.
Definition 7.7 The axioms for real-closed fields consist of the axioms of
ordered fields together with
RCF1. ∀x[0  x → ∃y[y2 = x]],
RCF2. ∀y0 . . . yn−1 ∃x[xn + yn−1 · xn−1 + · · · + y1 · x + y0 = 0]
for each odd n ∈ N.
In other words, a real-closed field is an ordered field in which every positive
element has a square root and every odd-degree polynomial has a root.
Fact 7.8 The axioms of real-closed fields axiomatise the complete theory RCF
of the real ordered field Ro-ring .
Using the basic ideas of real analysis, it is easy to verify that Ro-ring is a real-
closed field. However, the proof that the theory RCF is complete is beyond the
scope of this book.
Remark 7.9 Often the theory RCF is defined as an Lring -theory rather than
an Lo-ring -theory. The order is definable from the field structure as x < y ↔
∃z[z  0 ∧ x + z2 = y], so it would be possible to translate all the axioms we
have given involving < into Lring -sentences. However, other axiomatisations
are also possible. See [Mar02, Section 3.3] and [Poi00, Section 6.6] for more
details.
 Exercises 41

Exercises
7.1 Give an example to show that the axioms of ordered fields do not
axiomatise a complete theory.
7.2 Verify that Ro-ring does satisfy the of axioms for real-closed fields, using
standard theorems of real analysis.
7.3 Prove that in any ordered field F we have 0 < 1 and 1 < 1 + 1 and, more
generally, that N; +, 0, 1, < embeds in F; +, 0, 1, <.
7.4 Let Σ be the set of axioms of ordered fields. Show that Σ  ∀x[x · x  0].
7.5 Give a linear order on Cring satisfying axioms O1–O4. Show that there is
no linear order on Cring satisfying axioms O1–O5.
7.6 Suppose that S ⊆ R is non-empty and bounded above, and furthermore
that it is defined by an Lo-ring -formula ϕ(x). Now suppose that R is
another real-closed field, with domain R. Show that the subset ϕ(R) of R
defined by ϕ(x) has a least upper bound in R.
7.7 Is the usual definition of topological spaces given by first-order axioms
or second-order axioms? What about metric spaces?
7.8 [This exercise requires some knowledge of field theory.] A complex
number is algebraic if it is a zero of a non-trivial polynomial with integer
coefficients. Show that the set of algebraic numbers forms a subfield
of Cring which is algebraically closed. Show also that the set of real
algebraic numbers forms a subfield of R which is real-closed.
7.9 Look up the notion of Puiseux series, and find proofs that the field of
Puiseux series over C forms an algebraically closed field and that the
field of Puiseux series over R forms a real-closed field.
 8
Compactness and New Constants

If we start with a non-empty class of structures C and form a theory T =

Th(C), we know that T is satisfiable (that is, it has a model) because we have
some models to start with. However, we can write down a set of sentences Σ
without any particular model in mind and want to know that it is satisfiable.
It may be very difficult to describe a model of Σ explicitly. However, the next
theorem, which is the most important theorem in model theory, gives us at least
a practical way to show that models do exist and that when an infinite model
exists, then there are a lot of other models as well.
In the next chapter we will see that the same theorem helps us to understand
the class of models of a theory, and it will be used to show that some classes of
structures are not classes of models of any first-order theory. It has many other
uses as well.

8.1 The Compactness Theorem

Theorem 8.1 (The Compactness Theorem) Let Σ be a set of L-sentences. If
every finite subset of Σ has a model, then Σ has a model.
If every finite subset of Σ has a model, then we say that Σ is finitely
satisfiable. So the compactness theorem says that finite satisfiability is the same
as satisfiability.
There are at least three different approaches to the proof of the compactness
theorem. In Chapter 11 we will see a way of explicitly constructing a model
from the set of sentences Σ, which is due to Henkin. There is also a more
algebraic proof, which builds a model of Σ from given models of its finite
subsets using the technique of ultraproducts. See Section 4.1 of [CK90] or
Chapter 4 of [Rot00]. These methods of proof are reasonably self-contained

42
 8.2 The Use of New Constant Symbols 43

but not straightforward. By contrast, the original proof uses the idea of a formal
deduction and so is not self-contained. For those who have seen the notion of
a formal deduction, we give here a short proof of the compactness theorem.
Proof Given a set of L-sentences Σ and another L-sentence ϕ, we can write
Σ d ϕ to mean that there is a formal deduction of ϕ from Σ. A formal deduction
is a finite list of L-formulas, so it only uses finitely many of the sentences
from Σ. So if Σ d ϕ, then there is a finite subset Σ0 of Σ such that Σ0 d ϕ.
The Completeness Theorem of first-order logic states that Σ d ϕ if and only if
Σ  ϕ.
We write ⊥ for some contradictory sentence, for example ∃x[x  x], which
is not true in any L-structure at all. Now Σ  ⊥ means that every model of Σ is
a model of ⊥, and since there are no models of ⊥, it means there are no models
of Σ, that is, Σ is not satisfiable.
Now suppose that Σ is not satisfiable. Then Σ  ⊥, so Σ d ⊥, hence there is
a finite subset Σ0 of Σ such that Σ0 d ⊥. Then Σ0  ⊥, so Σ0 is not satisfiable.
Therefore, whenever every finite subset of Σ is satisfiable, Σ itself is satisfiable,
as required. 

8.2 The Use of New Constant Symbols

We will use the compactness theorem to show that if A is any infinite
L-structure, then there is another L-structure B which is elementarily equiv-
alent to A, that is, it satisfies exactly the same L-sentences, but which is not
isomorphic to A. The other tool we use is to change the language by adding
new constant symbols.
Before proving the general theorem, we do a special case with a simpler
proof which contains the main ideas.
Proposition 8.2 There is a model of true arithmetic, Th(Ns-ring ), which is not
isomorphic to the standard model Ns-ring .
Proof Let L+ = Ls-ring ∪ {c}, where c is a new constant symbol. Let Σ be the
set of L+ -sentences

Th(Ns-ring ) ∪ {c  0, c  1, c  1 + 1, c  1 + 1 + 1, . . .},

which consists of the complete theory of Ns-ring in the language Ls-ring , together
with sentences using the new constant symbol c which say that c is not equal
to any natural number. Now let Σ0 be a finite subset of Σ. Then there is
some number m ∈ N such that the sentence c  
1 + 1 + · · · + 1 does not
m
44 8 Compactness and New Constants

occur in Σ0 . We make a model of Σ0 by expanding Ns-ring by interpreting

the new constant symbol c as the number m. So Σ0 is satisfiable. Hence, by
+
compactness, there is a model M+ = M; +, ·, 0, 1, cM  of Σ. We consider the
reduct M = M; +, ·, 0, 1, which is an Ls-ring -structure. Then M |= Th(Ns-ring ),
which is a complete Ls-ring -theory, and so Th(M) = Th(Ns-ring ).
+
Suppose π : Ns-ring → M were an isomorphism. Let a = cM . Then there is
n ∈ N such that π(n) = a. There is a term t which is either 0, 1, or of the form
1 + · · · + 1, such that n = tNs-ring . Then a = π(tNs-ring ) = tM . But a is not equal to
any such tM because M+ |= Σ, so π is not surjective. Hence π does not exist,
and so M is not isomorphic to Ns-ring . 
We can make a similar argument with any infinite structure A in place of
Ns-ring , but we may need to use more than one new constant symbol.
Proposition 8.3 Let A be any infinite L-structure. Then there is another
L-structure B which is elementarily equivalent to A but not isomorphic to it.
Proof Any isomorphism is a bijection, so isomorphic models have the same
cardinality (size). So we will ensure that B has larger cardinality than A and
so cannot be isomorphic to it. To do this, let I be a set of cardinality larger
than |A|. Let LI be the language obtained from L by adding distinct constant
symbols ci for each element
 of I. 
Let Σ = Th(A) ∪ ci  c j | i, j ∈ I, i  j , a set of LI -sentences. We claim
that Σ is finitely satisfiable. So let Σ0 be a finite subset of Σ. Then only finitely
many of the ci , say, ci1 , . . . , cin , are used in Σ0 . Now choose a1 , . . . , an from A,
all distinct. This is possible because A is an infinite structure. Expand A to an
LI -structure A+ by interpreting ci j as a j for j = 1, . . . , n, and every other ci as
a1 . Then
 
A+ |= Th(A) ∪ ci j  cik | 1  j < k  n ,

so A+ |= Σ0 . Thus Σ is finitely satisfiable, and hence, by compactness, it has a

model, say, B+ , which is an LI -structure. Let B be the domain of B+ , and for
+
each i ∈ I, let ai = cB +
i . Since B |= Σ, the elements ai of B are all distinct, and
so |B|  |I| > |A|.
Let B be the reduct of B+ to L. Then B |= Th(A), and so B is elementarily
equivalent to A. The domain of B is B, the same as that of B+ . So B cannot be
isomorphic to A because their cardinalities are different. 
The compactness theorem opens the door to study infinite structures by
looking at different models of the same theory. We will explore this in Part
III of the book. Now we give an indication of what the compactness theorem
says about definable sets.
 Exercises 45

8.3 Applications to Definable Sets

Consider the closed intervals [0, 1/n] on the real line, for n ∈ N+ . The
intersection of all these intervals is {0}, so in particular, the intersection is
non-empty. If instead we consider the open intervals (0, 1/n) for n ∈ N+ ,
then the intersection is empty. However, if we take any finite sub-collection,
say, {(0, 1/n1 ), . . . , (0, 1/nr )}, then the intersection is the non-empty interval
(0, 1/N), where N = max{n1 , . . . , nr }.
Proposition 8.4 There is a model R of Th(Ro-ring ) in which the intersection of
the intervals {(0, 1/n) | n ∈ N+ }, as interpreted in R, is nonempty.
Proof The interval (0, 1/n) is defined by the Lo-ring -formula ϕn (x) given by
0 < x ∧ n · x < 1, where we use n as an abbreviation for 
1 + · · · + 1. Let c
n
be a new constant symbol, and let Σ = Th(Ro-ring ) ∪ {ϕn (c) | n ∈ N }, a set of
sentences in the language Lo-ring expanded by c. Let Σ0 be a finite subset of
Σ, and let N ∈ N+ be larger than any n such that ϕn (c) ∈ Σ0 . Expand Ro-ring
to R+ by interpreting c as 1/N. Then R+ |= Σ0 . So Σ is finitely satisfiable,
+
and hence, by compactness, it has a model, say, R+ . Let a = cR . The reduct
+
R of R to Lo-ring is a model of Th(Ro-ring ), and the element a ∈ R satisfies
R |= 0 < a ∧ n · a < 1 for every n ∈ N. Hence it lies in all the intervals (0, 1/n),
as interpreted in R. 

Remarks 8.5 (i) The element a ∈ R in the above proof is called an

infinitesimal, and 1/a is an infinite element of R, that is, it is larger than
any natural number. So the model R of Th(Ro-ring ) is a non-Archimedean
ordered field.
(ii) The word compactness comes from topology, where it is a very useful
property of a topological space. The above proposition hints that
definable sets behave like the closed sets of a compact topological space,
at least if we are allowed to change models. In the exercises, this is
explored further, and we will revisit this idea when we consider types in
Part V. However, we will generally not use the language of topology.

Exercises
8.1 Is there a model M of Th(Ns-ring ) with an element m ∈ M satisfying
0 < m < 1?
8.2 Sketch a proof of Proposition 8.3. You should list all the key ideas and
explain how they fit together, without giving all the details.
46 8 Compactness and New Constants

8.3 Where does the proof of Proposition 8.3 go wrong if A is a finite

L-structure?
8.4 Show that there is a model M of Th(N< ) with an infinite descending
chain a1 > a2 > a3 > · · · in M.
8.5 Show that there is a model M of Th(Z< ) such that Q< embeds in M.
8.6 A linearly ordered set A = A; < is well-ordered if every non-empty
subset of A has a least element. Show that if A is any infinite linearly
ordered set, then there is B ≡ A such that B is not well ordered. Show
also that there is no well-ordered set A such that A ≡ Z< .
8.7 Suppose that A is an infinite L-structure and ϕ( x̄) is a formula such that
ϕ(A) is an infinite set. Show that there is a B elementarily equivalent to
A such that ϕ(B) is uncountable.
8.8 Suppose A is any L-structure and {S i | i ∈ I } is any collection of
definable subsets of A such that for every finite subset I0 ⊆ I, the

intersection i∈I0 S i is non-empty. Show there is B ≡ A such that,

interpreted in B, the intersection i∈I S i is non-empty.
8.9 Suppose Σ is a set of L-sentences and {ϕi (x) | i ∈ I } is a set of L-formulas
such that for every model A |= Σ and every element a ∈ A there is some
 A |= ϕi (a). Show that there is a finite subset I0 ⊆ I such
i ∈ I such that
that Σ  ∀x i∈I0 ϕi (x) .
8.10 The four-colour theorem states that for any map of finitely many
countries drawn on the plane, with each country connected, it is possible
to colour each country one of four colours such that no two adjacent
countries have the same colour. Assuming this theorem, prove that the
same holds for a map with infinitely many countries.
 9
Axiomatisable Classes

The axiomatic method has been important in mathematics since Euclid.

Euclid used axioms to capture fundamental properties of logic, geometry,
or arithmetic that were supposed to be self-evident and from which other
statements could be proved. In this book, we usually consider a structure such
as Ns-ring , Cring , or Ro-ring , which we take as given (perhaps by some second-
order description), and we look for axioms sufficient to capture its first-order
theory. These axioms may not be self-evident; indeed, it may be a difficult task
to find them and prove that they do hold.
In this chapter, however, axioms are used in a third way: to define a class
of structures. The compactness theorem will be used to give general properties
of classes which can be axiomatised by first-order sentences and to show that
some classes are not axiomatisable or are axiomatisable but only by an infinite
list of sentences.
We will use the statement of the compactness theorem from Chapter 8, but
otherwise, this chapter does not depend on that one.
Definition 9.1 If Σ is a set of L-sentences, we write Mod(Σ) for the class
of all L-structures which are models of Σ. The class Mod(Σ) is said to be
axiomatised by Σ and is an axiomatisable class. It is finitely axiomatisable
if it is axiomatised by a finite set of sentences.
For example, the class of all groups is axiomatised by the axioms G1, G2,
and G3 given in Chapter 6. To show a class of structures is axiomatisable, it
suffices to write down appropriate axioms.

9.1 Finite and Infinite Models

The class of all groups of size 3 is axiomatisable, since it is the class of all
models of the group axioms together with the sentence

47
48 9 Axiomatisable Classes

∃x1 x2 x3 [x1  x2 ∧ x1  x3 ∧ x2  x3 ∧ ∀y[y = x1 ∨ y = x2 ∨ y = x3 ]].

By a similar argument, the class of groups of any finite size is axiomatisable.

The class of all infinite groups is axiomatisable,
 as can be
 seen by considering
the sentences κn given by ∃x1 , . . . , xn 1i< jn xi  x j for n ∈ N+ . It is the
class Mod({G1, G2, G3} ∪ {κn | n ∈ N+ }). However, the class of finite groups is
not axiomatisable.
Proposition 9.2 Let C be an axiomatisable class with arbitrarily large finite
models. That is, for every n ∈ N, there is a model An ∈ C whose domain is a
finite set of size at least n. Then C contains an infinite model.
Proof Let Σ be a set of sentences such that C = Mod(Σ). Let Σ = Σ ∪
{κn | n ∈ N+ }, where the sentences κn are as given above, and let Σ0 be a finite
subset of Σ . Let N ∈ N be larger than any n such that κn ∈ Σ0 , which is
possible because Σ0 is finite. Then AN |= Σ0 . So Σ is finitely satisfiable, hence
by compactness it has a model, say, B. Then B is an infinite structure, and
B |= Σ, so B ∈ C. 
Corollary 9.3 The class of finite groups is not axiomatisable.
Proof There are arbitrarily large finite groups, for example the cyclic groups
Cn of order n for each n ∈ N+ . So any axiomatisable class containing all finite
groups also contains some infinite groups. 

9.2 Torsion in Abelian Groups

Consider an abelian group G written additively in the language Ladgp =
+, −, 0. For g ∈ G and n ∈ N+ we write ng as an abbreviation for g + · · · + g,

n
as usual.

Definition 9.4 We say that g is an n-torsion element of G if we have ng = 0.

It is a torsion element if it is an n-torsion element for some n.
A group is torsion-free if its only torsion element is 0.
An abelian group is a torsion group if all of its elements are torsion
elements.

For each n ∈ N+ the sentence ϕn given by ∀x[nx = 0 → x = 0] states

that there are no n-torsion elements, except for 0. The set of axioms for abelian
groups together with {ϕn | n ∈ N+ } axiomatises the class of torsion-free abelian
groups.
 9.3 Finite Axiomatisability 49

Now we consider torsion groups. Every finite abelian group is a torsion

group, but there are also infinite torsion groups, such as the direct sum of
infinitely many finite groups.
Proposition 9.5 The class of all torsion abelian groups is not axiomatisable.
Proof Suppose it were axiomatised by Σ. Expand the language Ladgp to L
by adding one new constant symbol c. Let ψn be the L -sentence nc  0, and
let Σ = Σ ∪ {ψn | n ∈ N+ }. Let Σ0 be a finite subset of Σ , and let N ∈ N+ be
larger than any n such that ψn ∈ Σ0 . Let A be the L -structure consisting of the
cyclic group C N with cA being a cyclic generator. Then A |= Σ, because C N
is a torsion abelian group, and also A |= ψn for each n < N, because cA has
order N. Hence A |= Σ0 , so Σ0 is satisfiable. By compactness, Σ is satisfiable.
Let G be a model of Σ . Then G is a torsion abelian group, since G |= Σ, but G
has an element cG which is not an n-torsion element for any n, because G |= ψn .
This contradiction shows that Σ could not exist, hence the class of all torsion
abelian groups is not axiomatisable. 

9.3 Finite Axiomatisability

Recall that an axiomatisable class is finitely axiomatisable iff there is a finite set
of sentences Σ which axiomatises the class. For example, the class of groups is
finitely axiomatisable. However our choices of axioms for algebraically closed
fields and for Peano arithmetic were not finite but involved infinite axiom
schemes, as did our axiomatisation of torsion-free abelian groups. It is often
useful to know whether a finite axiomatisation exists, or at least if it does not
exist, to prove it. We start with an easy lemma.
Lemma 9.6 If C is finitely axiomatisable, then it is axiomatised by a single
sentence.
Proof If C is axiomatised by {σ1 , . . . , σn }, then it is also axiomatised by
n
i=1 σi . 
Now if for example the class of torsion-free abelian groups were finitely
axiomatisable, how could we find a finite list of axioms? In fact we do not
have to look through all sentences.
Proposition 9.7 If C = Mod(Σ) and C is finitely axiomatisable, then it is
axiomatised by a finite subset of Σ.
Proof Suppose not. Then using Lemma 9.6, we have a set Σ of sentences
and a single sentence ϕ such that C = Mod(Σ) and C = Mod(ϕ), but C is
50 9 Axiomatisable Classes

not axiomatised by any finite subset of Σ. Let Σ0 be a finite subset of Σ. Then

Mod(Σ0 ) ⊇ Mod(Σ), but by assumption, Mod(Σ0 )  Mod(Σ), so there is
A |= Σ0 such that A |= Σ. Thus A  C, and so A |= ¬ϕ. Thus Σ0 ∪ {¬ϕ} is
satisfiable. This holds for any finite subset Σ0 of Σ, so, by compactness, there
is a model B of Σ ∪ {¬ϕ}. But then B |= Σ so B ∈ C, and B |= ¬ϕ, so B  C,
a contradiction. 
Corollary 9.8 The class of all torsion-free abelian groups is not finitely
axiomatisable.
Proof Recall that the class of torsion-free abelian groups is axiomatised by
{G1, G2, G3, AG}∪{ϕn | n ∈ N+ }. Suppose Σ0 is a finite subset of these axioms,
and let N ∈ N be larger than any n such that ϕn ∈ Σ0 . Let p be a prime such
that p  N. Then the cyclic group C p of order p is an abelian group and has
no n-torsion for n < p, so C p |= Σ0 . However, C p has p-torsion, so Σ0 does not
axiomatise the class of torsion-free abelian groups. By Proposition 9.7, this
class is not finitely axiomatisable. 
Our final application of the compactness theorem in this chapter is a result
about axiomatising the complement of an axiomatisable class.
Proposition 9.9 Let C be an axiomatisable class of L-structures, and let D =
{A an L-structure | A  C }. Then D is axiomatisable if and only if C and D
are finitely axiomatisable.
Proof First suppose that C is axiomatised by Σ and D is axiomatised by Φ.
Suppose for a contradiction that C is not finitely axiomatisable. Let Σ0 be a
finite subset of Σ, and let Φ0 be a finite subset of Φ. Then Mod(Σ0 )  Mod(Σ),
so there is A |= Σ0 such that A  C. So A ∈ D, so A |= Σ0 ∪ Φ. In particular,
Σ0 ∪ Φ0 is satisfiable. Then, by compactness, Σ ∪ Φ is satisfiable, so there is a
model M |= Σ ∪ Φ, so M ∈ C and M ∈ D, a contradiction. So C is finitely
axiomatisable.
Now suppose C is finitely axiomatisable. Then by Lemma 9.6 it is axioma-
tised by a single sentence, say, ϕ. Then D is axiomatised by ¬ϕ. 

Exercises
9.1 Let C be a non-empty class of L-structures. Show that C ⊆ Mod(Th(C))
and that C is axiomatisable iff equality holds.
9.2 Let Σ be a satisfiable set of L-sentences. Show that Σ ⊆ Th(Mod(Σ)) and
that Σ is a theory iff equality holds.
 Exercises 51

9.3 Show that the class of groups of size less than 100 is axiomatisable
and that the class of infinite groups is axiomatisable but not finitely
axiomatisable.
9.4 Show using the compactness theorem and the method of new constants
from the previous chapter that if C is an axiomatisable class with either
arbitrarily large finite models or an infinite model, then it has infinite
models of arbitrarily large cardinality.
9.5 Prove that the class of complete ordered fields is not axiomatisable.
[Hint: use Remark 8.5 (i).]
9.6 Is the class of cyclic groups axiomatisable? Is it finitely axiomatisable?
9.7 An abelian group G is said to be n-divisible for some n ∈ N+ if it satisfies
∀x∃y[ny = x], so each element can be divided by n (not necessarily
uniquely). G is divisible if it is n-divisible for all n ∈ N+ . Show that
the class of divisible abelian groups is axiomatisable but not finitely
axiomatisable.
9.8 Show that the group G = C  {0}; ·, 1 is divisible but not torsion-free.
9.9 Let DTFAG be the theory of divisible, torsion-free abelian groups. Show
that if V is a Q-vector space considered as an Ladgp -structure, then V |=
DTFAG.
9.10 The characteristic of a field F is the least number n ∈ N+ such that
1 + · · · + 1 = 0, if such an n exists, and 0 otherwise. The characteristic

n
of a field is always 0 or a prime number. Show that the class of fields of a
fixed prime characteristic p is finitely axiomatisable and that the class of
fields of characteristic 0 is axiomatisable but not finitely axiomatisable.
9.11 Show that the class of fields of positive characteristic is not
axiomatisable.
9.12 Suppose that σ is an Lring -sentence which is true in every field of
characteristic 0. Show that there is N ∈ N such that if F is a field of
characteristic p with p > N, then F |= σ.
9.13 A graph G consists of a set of vertices, and between each pair of distinct
vertices, there may or may not be an edge. We can consider a graph
as a structure for the language with one binary relation E. The vertices
of the graph are the elements of the structure, and vertices v, w have
an edge between them if E(v, w) holds. A graph is then a model of the
sentence
∀xy[¬E(x, x) ∧ (E(x, y) → E(y, x))].
52 9 Axiomatisable Classes

We say that a graph is connected if for every pair of vertices v, w,

there is a path from v to w, which means there is a sequence of vertices
v = v1 , v2 , . . . , vn = w with an edge between vi and vi+1 for each i =
1, . . . , n − 1.
Show that the class of connected graphs is not axiomatisable.
 10
Cardinality Considerations

The model theory we cover will use very few set-theoretic tools, essentially
only the basics of cardinal arithmetic. We give an intuitive explanation of what
little we need, omitting the proofs. In particular, we avoid the use of ordinals.
The underlying set theory we use is the standard one in mathematics, called
Zermelo–Fraenkel with Choice (ZFC). Details of the definitions and proofs
can be found in the first few pages of Kunen [Kun11] and Jech [Jec03]. For a
more gentle approach, Halmos [Hal74] is also suitable.

10.1 Cardinality
Definition 10.1 Two sets have the same cardinality (size) if and only if there
is a bijection between them. Given a set X, we write |X| for the cardinality of
X. A cardinal is a possible value of |X|.
We define |X|  |Y| to mean there is an injective function from X to Y.

We will make use of four basic facts about injective functions, as follows:
Facts 10.2 Let X and Y be sets.

(i) If there are injective functions X → Y and Y → X, then there is a

bijection between X and Y. (This is called the Schröder–Bernstein
theorem.)
(ii) Given two sets X and Y, there is either an injective function X → Y or
an injective function Y → X.
(iii) Suppose X  ∅. Then there is an injective function X → Y if and only if
there is a surjective function Y  X.

53
54 10 Cardinality Considerations

(iv) For any set X, there is a set Y (e.g., the power set PX) such that there is
an injective function X → Y but no injective function Y → X. (This is
Cantor’s theorem.)
From facts (i) and (ii), we see that  is a linear order on cardinals. Fact (iii)
is often useful as an alternative way to show that |X|  |Y|. Fact (iv) says there
is no greatest cardinal.
A set is finite precisely when its cardinality is some natural number n ∈
N; otherwise, it is infinite. There is a smallest infinite cardinal, which is |N|,
written ℵ0 . (ℵ, pronounced aleph, is the first letter of the Hebrew alphabet.) A
set is countable precisely when it is either finite or of cardinality ℵ0 ; otherwise,
it is uncountable. It is countably infinite if it is countable and infinite.
From Fact (iv), we see that the power set of the natural numbers, PN, is
uncountable. It is actually in bijection with the set R of real numbers, so R is
also uncountable.

10.2 Basic Cardinal Arithmetic

We can define addition and multiplication of cardinals. We define |X| · |Y| to
be |X × Y|, where X × Y is the usual cartesian product of the sets given by
{(x, y) | x ∈ X, y ∈ Y }. We define |X| + |Y| to be |X × {0} ∪ Y × {1}|. Doing basic
arithmetic with infinite cardinals is much easier than it is to do with natural
numbers. Using Facts 10.2, one can prove the following results about cardinal
arithmetic.
Fact 10.3 If both X and Y are finite, with |X| = n and |Y| = m, then |X| + |Y| =
n + m and |X| · |Y| = nm, the usual sum and product of natural numbers. If at
least one of X and Y is infinite and both are nonzero, then

|X| + |Y| = |X| · |Y| = max(|X|, |Y|).

Fact 10.4 Suppose that κ is an infinite cardinal and, for each n ∈ N, Xn is a

set such that |Xn |  κ. Let X = n∈N Xn . Then |X|  κ.
We give an application of this fact.
Proposition 10.5 If X is a nonempty set and S is the set of finite strings of

elements of X, that is, S = n∈N X n , then |S | = max(|X|, ℵ0 ).
Proof Take x ∈ X. Define f : N → S by taking f (n) to be the string of
length n with x in every place. Then f is injective and shows that |S |  ℵ0 .
 10.4 Further Set Theory 55

The inclusion map X → S which takes an element of X to itself as a string of

length 1 shows that |S |  |X|. So |S |  max(|X|, ℵ0 ).
For any set X, |X 0 | just consists of the unique string of length 0, so |X 0 | = 1.
If X is finite, then |X n | is finite for all n ∈ N; otherwise, by Fact 10.3 and
induction on n, we see that |X n | = |X| for all n ∈ N. So, by Fact 10.4, |S | 
max(|X|, ℵ0 ). Since cardinals are linearly ordered, we have |S | = max(|X|, ℵ0 ),
as required. 
That completes our brief survey of cardinal arithmetic.

10.3 The Cardinality of a Language

We now consider an application to our formal languages. We define the
cardinality |L| of a language L to be the cardinality of the set Form(L) of
L-formulas, where we only allow countably many variables, say, only xn for
n ∈ N. Write Symb(L) for the set of relation, function, and constant symbols
of L, that is, the vocabulary of L.
Proposition 10.6 |L| is equal to the maximum of ℵ0 and | Symb(L)|.
Proof The set Form(L) contains all the formulas (xn = xn ) for n ∈ N. This
gives an injective function from N into Form(L), so | Form(L)|  ℵ0 .
Considering the atomic formulas c = x0 , f (x1 , . . . , xn ) = x0 , and
R(x1 , . . . , xn ), we can see that for every symbol in Symb(L), there is a formula
containing it and no other symbol from Symb(L). So there is an injective
function from Symb(L) to Form(L), and hence | Form(L)|  | Symb(L)|. Thus
| Form(L)|  max(| Symb(L)|, ℵ0 ).
Now let X = Symb(L) ∪ {=, ∧ , ¬, ∃, (, ), [, ]} ∪ {xn | n ∈ N }. Then

|X| = | Symb(L)| + 8 + ℵ0 = max(| Symb(L)|, ℵ0 ).

Now Form(L) is a subset of the set of all finite strings of elements of X, and
hence | Form(L)|  max(| Symb(L)|, ℵ0 ) by Proposition 10.5. So | Form(L)| =
max(| Symb(L)|, ℵ0 ), as required. 

10.4 Further Set Theory

Many parts of model theory make use of ordinals and transfinite induction. A
proper account of the facts of cardinal arithmetic we have stated also requires
them. In this book we avoid ordinals and skirt around transfinite induction in
56 10 Cardinality Considerations

the one or two places it is needed, except in Part VI. Readers of that part of the
book should have no difficulty in learning how to do transfinite induction from
another source.
Some methods in model theory require an understanding of cardinal expo-
nentiation and issues relating to the continuum hypothesis and inaccessible
cardinals. These are briefly touched on in Chapter 27.

Exercises
10.1 What is the cardinality of the set of L-terms, given the vocabulary
Symb(L) of L?
10.2 Assume that for all infinite sets Y, we have |Y × Y| = |Y|. Then, using
Facts 10.2, deduce Facts 10.3 and 10.4.
10.3 Look up the definition of cardinal exponentiation and show that for any
set X, |PX| = 2|X| .
10.4 Prove Fact 10.2(i), the Schröder–Bernstein theorem.
10.5 Suppose that f : PX → X is an injective function. Let A =
{ f (S ) | S ⊆ X and f (S )  S }, and let a = f (A). By considering whether
a ∈ A, prove Fact 10.2(iv).
10.6 Show that Fact 10.2(iii) follows from the Axiom of Choice.
10.7 Using some form of the Axiom of Choice, for example Zorn’s lemma,
prove Fact 10.2 (ii).
 11
Constructing Models from Syntax

The goal of this chapter is to give a proof of the compactness theorem which
does not go via the notion of formal deductions.
In mathematical practice, there is a distinction between the language we
use to describe mathematical objects and the objects themselves. Mathematical
logic makes this distinction formal. Syntax is the word describing the formal
language we use, as distinct from the meaning, or semantics, which arises when
we interpret the syntax in mathematical structures. In logic it is important to be
able to distinguish between the two, so for example we can distinguish between
an element of a structure and a constant symbol which names the element.
However, mathematical logic also considers syntax in such a formal way
that we can reason mathematically about it and indeed build mathematical
structures out of it. This is precisely what we do in this chapter, where we
follow Henkin’s method to prove that a set Σ of sentences is satisfiable by
building a model of Σ using the sentences themselves. This proof is extended
in Chapter 24 to prove the omitting types theorem.

11.1 Lindenbaum’s Lemma

Recall the following definitions.

Definition 11.1 A set Σ of L-sentences is

finitely satisfiable if every finite subset of Σ has a model,

deductively closed if, for every L-sentence ϕ, if Σ  ϕ, then ϕ ∈ Σ, and
complete if, for every L-sentence ϕ, either Σ  ϕ or Σ  ¬ϕ.

57
58 11 Constructing Models from Syntax

Note in particular that if Σ is complete and deductively closed, then for each
L-sentence ϕ, either ϕ ∈ Σ or ¬ϕ ∈ Σ, and if Σ is also finitely satisfiable, then
it cannot contain both ϕ and ¬ϕ.
Lemma 11.2 (Lindenbaum’s lemma) If Σ is a finitely satisfiable set of
L-sentences, then there is a finitely satisfiable, deductively closed and complete
set Σ of L-sentences such that Σ ⊆ Σ .
The basic idea of the proof is to keep adding more sentences to Σ
maintaining finite satisfiability until, for every sentence ϕ, either ϕ or ¬ϕ is
included. We leave the details as an exercise.

11.2 Henkin Theories

Recall that an L-term is said to be closed if it does not contain any variables.
So if t is a closed L-term and A is an L-structure, then tA is an element of A.
Also recall that a theory is a satisfiable and deductively closed set of sentences.

Definition 11.3 A set Σ of L-sentences is called a Henkin theory if it is finitely

satisfiable, deductively closed, and complete and it has the witness property,
that is, for any L-sentence of the form ∃x[ϕ(x)] in Σ, there is a closed L-term t
such that the sentence ϕ(t) is in Σ.
Proposition 11.4 If Σ is a finitely satisfiable set of L-sentences, there is a
language L∗ which consists of L together with new constant symbols and a set
Σ∗ of L∗ -sentences such that Σ∗ is a Henkin theory.
In order to build our Henkin theory Σ∗ , we will expand the language to
introduce new constant symbols, which will be the closed terms needed for
the witness property. We will also use Lindenbaum’s lemma to ensure we get
a deductively closed and complete set of sentences. We have to iterate the
process to get all the properties we want, so in fact we will define a tower of
languages
L = L0 ⊆ L1 ⊆ L2 ⊆ · · · ⊆ Ln ⊆ · · ·

and a tower of sets of sentences

Σ ⊆ Σ0 ⊆ Σ1 ⊆ Σ2 ⊆ · · · ⊆ Σn ⊆ · · · ,

with each Σn being a finitely satisfiable, deductively closed and complete set
 
of Ln -sentences. Then L∗ = n∈N Ln and Σ∗ = n∈N Σn will have the desired
properties. With the explanation out of the way, we start the proof.
 11.3 Canonical Models 59

Proof Let L0 = L, and using Lindenbaum’s lemma, we can take Σ0 to

be a finitely satisfiable, deductively closed and complete set of L0 -sentences
extending Σ.
Assume inductively that we have Ln and Σn , which is a finitely satisfiable,
deductively closed and complete set of Ln -sentences. Let Ln+1 be Ln together
with a new constant symbol cϕ for each Ln -sentence of the form ∃x[ϕ(x)] which
is in Σn . Let Σn be Σn together with all sentences of the form ∃x[ϕ(x)] → ϕ(cϕ )
for the new constant symbols.
We claim that Σn is finitely satisfiable. To prove this, let S  be a finite subset
of Σn , and let S = S  ∩ Σn . By inductive hypothesis, Σn is finitely satisfiable, so
S has a model A which is an Ln -structure. Expand it to an Ln+1 structure A by
interpreting the constant symbol cϕ as an element a of A such that A |= ϕ(a) if
one exists, and as any element of A otherwise. Then A |= S  by construction,
so Σn is finitely satisfiable.
Applying Lindenbaum’s lemma again, we take Σn+1 to be some finitely sat-
isfiable, deductively closed and complete set of Ln+1 -sentences extending Σn .
 
Now take L∗ = n∈N Ln and Σ∗ = n∈N Σn . Clearly Σ∗ is a set of L∗ -
sentences. If ϕ is an L∗ -sentence, then for some n ∈ N it is an Ln -sentence,
so either ϕ or ¬ϕ is in Σn , because Σn is deductively closed and complete, and
hence either ϕ or ¬ϕ is in Σ∗ , so Σ∗ is deductively closed and complete. If
∃x[ϕ(x)] ∈ Σ∗ , then it is in some Σn , and then ∃x[ϕ(x)] → ϕ(cϕ ) ∈ Σn+1 ,
and then, since Σn+1 is deductively closed, ϕ(cϕ ) ∈ Σn+1 . So ϕ(cϕ ) ∈ Σ∗ , and
thus Σ∗ has the witness property. Finally, if S is a finite subset of Σ∗ , then it is
contained in some Σn , and Σn is finitely satisfiable, so S has a model, and thus
Σ∗ is finitely satisfiable. So Σ∗ is a Henkin theory, as required. 

11.3 Canonical Models

Definition 11.5 An L-structure A is said to be canonical if, for every element
a ∈ A, there is a closed L-term t such that tA = a. A canonical model of a
theory T is a model of T which is a canonical structure.

The structures Zring and Nsucc are canonical structures, but Ro-ring is not.
If A is an L-structure, then we can name every element of A with a new
constant symbol to get a canonical model in the expanded language LA . In
general, a theory will not have a canonical model unless it has enough closed
terms, which may involve adding new constant symbols. This is precisely what
Henkin theories are for.
60 11 Constructing Models from Syntax

Proposition 11.6 Every Henkin theory has a canonical model.

Proof Let T be a Henkin theory in the language L, and let Λ be the set of
closed L-terms. Define a relation ∼ on Λ by t1 ∼ t2 if the sentence t1 = t2 is
in T . Then ∼ is an equivalence relation (see Exercise 11.3). Let A be the set of
equivalence classes. It will be the domain of our model. We will write t˜ for the
equivalence class of t.
Now we have to make A into an L-structure A by giving interpretations
of the constant symbols, function symbols, and relation symbols of L. If c is a
constant symbol, we define cA to be the equivalence class of c. If f is a function
symbol of arity n and t1 , . . . , tn are closed terms of L, we define f A (t˜1 , . . . , t˜n )
to be the equivalence class of the closed term f (t1 , . . . , tn ). If s1 , . . . , sn are
closed terms with si ∼ ti for each i, since si = ti is in T and T is deductively
closed and complete, we have f (s1 , . . . , sn ) = f (t1 , . . . , tn ) ∈ T . So f A is well
defined. Finally, if R is a relation symbol of L of arity n and t1 , . . . , tn are closed
terms, we define (t˜1 , . . . , t˜n ) ∈ RA if and only if R(t1 , . . . , tn ) is in T . A similar
argument shows that RA is well defined.
By construction, A is a canonical L-structure. It remains to show that
A |= T . We prove by induction on the construction of L-formulas ϕ( x̄) that
for any (t˜1 , . . . , t˜n ) in An ,

A |= ϕ(t˜1 , . . . , t˜n ) if and only if ϕ(t1 , . . . , tn ) ∈ T.

For atomic formulas it holds by definition. The ∧ and ¬ inductive steps are
straightforward (see Exercise 11.4).
Suppose that ϕ( x̄) has the form ∃y[θ(y, x̄)]. Then

A |= ϕ(t˜1 , . . . , t˜n ) iff there is s ∈ Λ such that A |= θ( s̃, t˜1 , . . . , t˜n )

iff there is s ∈ Λ such that θ(s, t1 , . . . , tn ) ∈ T

by the inductive hypothesis.

Now θ(s, t1 , . . . , tn )  ∃y[θ(y, t1 , . . . , tn )], so if there is some s ∈ Λ
such that θ(s, t1 , . . . , tn ) ∈ T , since T is deductively closed, we have
∃y[θ(y, t1 , . . . , tn )] ∈ T . Conversely, if ∃y[θ(y, t1 , . . . , tn )] ∈ T , then, since T
is a Henkin theory, there is a closed term s such that θ(s, t1 , . . . , tn ) ∈ T . That
completes the inductive step.
So A |= T as required. 
Putting everything together, we get a proof of the compactness theorem.
Proof of Theorem 8.1 Let Σ be a finitely satisfiable set of L-sentences. By
Proposition 11.4, there is a Henkin theory Σ∗ extending Σ in a language L∗
 Exercises 61

expanding L. By Proposition 11.6, Σ∗ has a canonical model, say, A∗ . Then

the reduct A of A∗ to L is a model of Σ. 

11.4 The Strong Compactness Theorem

By considering the cardinalities of the languages involved, we will show that
Henkin’s method proves a stronger version of the compactness theorem.
Theorem 11.7 (Strong compactness theorem) If Σ is a finitely satisfiable set
of L-sentences, then there is a model A of Σ of cardinality at most |L|.
Proof We will show that in the proof of Proposition 11.4, the language L∗ has
the same cardinality as L. Then, in the canonical model A given by Proposition
11.6, every element of A is named by an L∗ -term, so |A|  |L∗ | = |L|.
For each n ∈ N, Σn is a set of |Ln |-sentences, so |Σn |  |Ln |. We have
| Symb(Ln+1 )|  | Symb(Ln )| + |Σn |, so by Proposition 10.6, |Ln+1 |  |Ln | + |Σn | =

|Ln |. So, by induction on n, |Ln | = |L0 | = |L| for all n ∈ N. Since L∗ = n∈N Ln ,
by Fact 10.4, |L∗ | = |L|, as required. 

Exercises
11.1 Let L be a countable language, and enumerate the set of L-sentences as
(ϕn )n∈N . Let Σ be a finitely satisfiable set of L-sentences, and show that
at least one of Σ ∪ {ϕ0 } or Σ ∪ {¬ϕ0 } is finitely satisfiable. Then build on
this idea to prove Lindenbaum’s lemma in the case that L is countable.
11.2 [For those who know some set theory] Use transfinite induction or
Zorn’s lemma to prove Lindenbaum’s lemma for a language of arbitrary
cardinality.
11.3 Show that the relation ∼ on the set of closed L-terms of a Henkin theory
T given by “t1 ∼ t2 if the sentence t1 = t2 is in T ” is an equivalence
relation.
11.4 Complete the proof of Proposition 11.6 by showing that RA is well
defined and that the ∧ and ¬ steps of the induction on formulas hold.
11.5 Summarise the proof of the compactness theorem; that is, list all of the
key ideas and explain how they fit together, without giving all the details.
 Part III

Changing Models

In this third part of the book, we extend the idea that a theory has different
models with the Löwenheim–Skolem theorems. The Downward Löwenheim–
Skolem theorem is proved using induction on formulas and the new idea of
Skolem functions. The Upward Löwenheim–Skolem theorem is proved by
extending the method of new constants to the method of diagrams.
The simplest possible description for the class of models of a theory is that
the models are determined by a single cardinal invariant, which is the case
for vector spaces which are determined by their dimension. In large enough
models, this dimension is equal to the cardinality of the model. A theory like
this with only one model of a certain cardinality is called categorical. The Łos–
Vaught test gives the completeness of an axiomatisation for such categorical
theories. So for several examples we are able to complete the first two parts of
our programme: to give an axiomatisation for the theory of a structure and to
classify the other models of the theory. The important back-and-forth method
is introduced to achieve the first aim for dense linear orders, although it turns
out to be impossible to classify all the models in that case. In the last chapter of
this part, the whole programme is given as an extended exercise for the natural
numbers with the successor function. This includes some investigation of the
definable sets, setting the scene for Part IV.

63
 12
Elementary Substructures

We now consider the problem of finding new models of the theory of a given
structure. In this chapter we start with a structure B and find substructures A of
B which are elementarily equivalent to it, and in fact cannot be distinguished
from B even by formulas applied to elements of A. The new model A
is built using Skolem functions, and the method of proof is induction on
the construction of formulas. The Tarski–Vaught test helps to organise the
induction proof efficiently.

12.1 Elementary Substructures

We first recall the notion of a substructure from Chapter 5.

Definition 12.1 Let A and B be L-structures, and suppose the domain of A is

a subset of the domain of B. We write A ⊆ B and say that A is a substructure
of B, and B is an extension of A iff the inclusion of A into B is an embedding
of L-structures.

For example, if B is a group considered as an Lgp -structure, then the

substructures of B are exactly the subgroups of B.
Recall that if A is a substructure of B, ϕ( x̄) is an atomic (or even quantifier-
free) formula, and ā ∈ A, then A |= ϕ(ā) if and only if B |= ϕ(ā). However, a
sentence with quantifiers may have different truth values in A and B.

Definition 12.2 We say that A is an elementary substructure of B and B is

an elementary extension of A, and write A  B, if the domain of A is a subset
of the domain of B and, for each formula ϕ( x̄) and each ā ∈ An ,

A |= ϕ(ā) if and only if B |= ϕ(ā).

65
66 12 Elementary Substructures

π
We say that an embedding of L-structures A −→ B is an elementary
embedding iff π(A)  B.

When A  B, then A and B are very similar; indeed, they are indistin-
guishable from the point of view of the truth of L-formulas applied to elements
of A. The word elementary is used because the structures look the same in
terms of their elements, that is, in terms of first-order logic.

Example 12.3 Consider Zadgp and the substructure E = E; +, −, 0 con-

sisting of the even numbers. Then E  Zadgp , because if ϕ(x) is the formula
∃y[y + y = x], then Zadgp |= ϕ(2) but E |= ϕ(2). However, Zadgp  E, because
the map π : Z → E given by π(n) = 2n is an isomorphism.

Separating out formulas into sentences and those with free variables, we
have two easy consequences of being an elementary substructure.
Lemma 12.4 Suppose A  B. Then A ≡ B, and for any L-formula
ϕ(x1 , . . . , xn ), we have ϕ(A) = ϕ(B) ∩ An .
Proof Both conclusions are immediate from the definitions. 

12.2 The Tarski–Vaught Test

We now turn to finding elementary substructures A of an L-structure B. The
key idea is that if B says that there is an element with some property, then there
must be such an element already in A. The surprising thing is that this is all
that is needed.
Lemma 12.5 (Tarski–Vaught test) Suppose A ⊆ B is an L-substructure and
that for every L-formula ϕ( x̄, y) and every ā ∈ A and b ∈ B such that B |=
ϕ(ā, b), there is d ∈ A such that B |= ϕ(ā, d). Then A  B.
Proof Suppose A ⊆ B, satisfying the condition in the statement of the
lemma. We prove by induction on formulas ϕ( x̄) that for each ā in A,

A |= ϕ(ā) iff B |= ϕ(ā).

For ϕ an atomic formula, it is true by Lemma 3.6, because the inclusion map
A → B is an embedding.
There are inductive steps for ∧ , ¬, and ∃. The first two are easy and left as
an exercise. So suppose ϕ( x̄) is ∃yψ( x̄, y) and ā ∈ A. If A |= ∃yψ(ā, y), then
there is d ∈ A such that A |= ψ(ā, d), and so by induction, B |= ψ(ā, d), hence
B |= ∃yψ(ā, y). Conversely, if B |= ∃yψ(ā, y), then there is b ∈ B such that
 12.3 The Downward Löwenheim–Skolem Theorem 67

B |= ψ(ā, b). Then, using the condition in the statement of the lemma, there is
d ∈ A such that B |= ψ(ā, d). Then, by induction again, A |= ψ(ā, d), hence
A |= ∃yψ(ā, y). That completes the ∃ inductive step and the whole proof. 

12.3 The Downward Löwenheim–Skolem Theorem

Now we show that elementary substructures exist.
Theorem 12.6 (The Downward Löwenheim–Skolem theorem) Let B be a
L-structure and S a subset of B. Then there is an elementary substructure
A  B such that S ⊆ A and |A|  max(|S |, |L|).
Proof If B is the empty L-structure, then we can take A = B. Otherwise, for
each formula ϕ( x̄, y), choose a function gϕ : Bn → B such that
⎧
⎪
⎪
⎨some d ∈ B such that B |= ϕ(b̄, d), if such a d exists,
gϕ (b̄) = ⎪
⎪
⎩any d ∈ B otherwise.

Now let S 0 = S , and for r ∈ N, define

  
S r+1 = S r ∪ gϕ (b̄)  ϕ( x̄, y) is an L-formula and b̄ ∈ S r ,

and let A = r∈N S r .
Note that if ϕ( x̄, y) is f ( x̄) = y for a function symbol f from L, then gϕ is
just the function f B . Also, if ϕ(x, y) is c = y for a constant symbol c, then gϕ
is the constant function with value cB . So A is closed under the functions and
constants in the language and hence is the domain of a substructure A of B.
Next we show that A  B. Suppose ϕ( x̄, y) is an L-formula, ā ∈ A, and
b ∈ B such that B |= ϕ(ā, b). Then also B |= ϕ(ā, gϕ (ā)), and gϕ (ā) ∈ A. So, by
the Tarski-Vaught test, A  B.
Finally, we must consider the size of A. Using Proposition 10.5, we see that
|S r+1 |  max(|S r |, |L|), and so by induction,
 |S r |  max(|S |, |L|) for each r ∈ N.
 
Then, using Fact 10.4, |A| =  r∈N S r   max(|S |, |L|), as required. 
Example 12.7 Let V |= ZFC, the axioms of set theory. The language is just
{∈}, so there is a countable elementary substructure V0  V. In particular, V0 |=
ZFC. Now, in ZFC, we can prove that there is an uncountable set. Thus there
is an apparent paradox (called the Skolem paradox) of a countable structure V0
containing an uncountable set. It is a good exercise for those who know some
set theory to see why this is not paradoxical at all.
68 12 Elementary Substructures

Example 12.8 There is a countable elementary substructure R of the real

ordered field Ro-ring . Since R is uncountable, R is not isomorphic to Ro-ring . In
particular, it cannot be a complete ordered field, so it must have subsets S ⊆ R
which are non-empty and bounded above but which do not have a least upper
bound in R. However, since S ⊆ R ⊆ R, these sets S do have a least upper
bound in R.

12.4 Skolem Functions

The functions gϕ which we defined in the proof of the Downward Löwenheim–
Skolem theorem are called Skolem functions. For the proof, it does not matter
how we choose the values of these functions. We may have to invoke some set
theory in the form of the axiom of choice, but the important thing is just that
they exist. In some cases, there is no way to choose these Skolem functions
to be definable. There are also structures which do have definable Skolem
functions for every formula.

Example 12.9 Consider the complex field Cring and the formula ϕ(x, y) given
by x = y · y. Then a Skolem function gϕ : C → C is a square-root function, so
for all x ∈ C we have gϕ (x)2 = x. In particular, we must have gϕ (−1) picking
out one of ±i, say, +i. But complex conjugation z → z is an automorphism of
Cring , and if gϕ were definable, it would be preserved under all automorphisms,
so we would have both gϕ (−1) = +i and gϕ (−1) = +i, that is, gϕ (−1) = −i, a
contradiction. So gϕ is not definable.

Example 12.10 Ns-ring has definable Skolem functions for every formula.
To prove it, we use the fact that every non-empty subset of N has a smallest
element. So for a formula ϕ( x̄, y) we can define a Skolem function by the
formula μϕ ( x̄, z) given by

(ϕ( x̄, z) ∧ ∀y[ϕ( x̄, y) → z  y]) ∨ (¬∃y[ϕ( x̄, y)] ∧ z = 0),

where we have  as an abbreviation, which is acceptable because  is definable

in Ns-ring .

Exercises
12.1 Prove Lemma 12.4.
12.2 Suppose A  B and ϕ(x) is an L-formula. Show that if ϕ(B) is finite,
then ϕ(A) = ϕ(B), and if ϕ(B) is infinite, then ϕ(A) is also infinite.
 Exercises 69

12.3 Why does Skolem’s paradox seem paradoxical, and why is it not in fact
contradictory?
12.4 In the real ordered field Ro-ring , write down Skolem functions for the
formulas x1 < y ∧ y < x2 , x1 < y, and y < x2 .
12.5 Suppose S ⊆ R is a finite subset which is definable in Ro-ring . Show that
S has a definable Skolem function. Then show that for every s ∈ S , the
singleton {s} is definable. [Hint: use induction on |S |.]
12.6 Suppose that B is a structure with Skolem functions given by function
symbols in the language. Show that any substructure of B is an elemen-
tary substructure.
12.7 Suppose that A ⊆ B ⊆ D are L-structures such that A  D and B  D.
Prove that A  B.
12.8 Sketch a proof of the Downward Löwenheim–Skolem theorem. You
should give all the key ideas and explain how they fit together, but you
do not need to give all the details.
 13
Elementary Extensions

In this chapter we consider the opposite problem to the previous chapter: given
a structure A, how can we find elementary extensions of it? In Chapter 8 we
used the compactness theorem in conjunction with new constant symbols to
find new models of Th(A). Here we combine those ideas with the method of
diagrams to find new models which are elementary extensions of A.

13.1 The Method of Diagrams

Recall that if A and B are L-structures with domains A and B, respectively,
π
then an elementary embedding A −→ B is an embedding such that for each
L-formula ϕ( x̄) and each ā ∈ An ,

A |= ϕ(ā) if and only if B |= ϕ(π(ā)),

and that A is an elementary substructure of B and B is an elementary extension

of A if A ⊆ B and the inclusion map is an elementary embedding.
The method of diagrams turns these formulas applied to tuples from A into
sentences by adding new constant symbols to the language. The terminology
diagram is unfortunate, because these diagrams have nothing to do with
pictures.
Given an L-structure A with domain A, we create a new language LA by
adding a new constant symbol ca for every element a ∈ A. These new constant
symbols must be distinct and different from any symbol in L. We then expand
A to an LA -structure A+ by interpreting the symbol ca as the element a.

Definition 13.1 The complete diagram of A is CDiag(A) = ThLA (A+ ), the

set of all LA -sentences which are true in A+ .

70
 13.2 The Upward Löwenheim–Skolem Theorem 71

Proposition 13.2 Let B be an L-structure. Then there is an elementary

π
embedding A −→ B if and only if B can be expanded to a model of CDiag(A).
π
Proof Suppose there is an elementary embedding A −→ B. Expand B to B+
+
by defining cB a = π(a). Any σ ∈ CDiag(A) is of the form ϕ(ca1 , . . . , can ) for
some a1 , . . . , an ∈ A such that A |= ϕ(a1 , . . . , an ). Since π is an elementary
embedding, B |= ϕ(π(a1 ), . . . , π(an )), and so B+ |= σ. Thus B+ |= CDiag(A).
Conversely, if B can be expanded to a model B+ of CDiag(A), define
π +
A −→ B by π(a) = cB a . A similar argument shows that π is an elementary
embedding. 
Remark 13.3 Given a structure A, the difference between an elementary
π
extension B of A and an elementary embedding A −→ B is not usually very
π
important. Given A −→ B with B = dom(B), let D be a set disjoint from A
such that there is a bijection f : B  π(A) → D. Let B = A ∪ D and define a
bijection g : B → B by

⎧
⎪
⎪ −1
⎨π (b) if b ∈ π(A),
g(b) = ⎪
⎪
⎩ f (b) otherwise.

We can make B into an L-structure B in exactly one way such that g is an

isomorphism. Then A  B .

The complete diagram has a useful variant called the diagram.

Definition 13.4 The diagram Diag(A) of an L-structure A is the set of

all atomic LA -sentences and negations of atomic LA -sentences which are true
in A+ .
Lemma 13.5 There is an embedding of A into an L-structure B if and only if
B can be expanded to a model of Diag(A).
The proof is left as an exercise.

13.2 The Upward Löwenheim–Skolem Theorem

Now we apply the compactness theorem with two lots of new constants: those
for the method of diagrams and those needed to ensure a model has large
cardinality.
72 13 Elementary Extensions

Theorem 13.6 (The Upward Löwenheim–Skolem theorem) For any infinite

L-structure A and any cardinal κ  max(|L|, |A|), there is an L-structure B of
cardinality equal to κ such that A  B.
Proof Let I be a set of cardinality κ. Let LA,I be the language obtained from
L by adding distinct constant symbols ca for each element of A, and ci for each
element of I.  
Let Σ = CDiag(A) ∪ ci  c j | i, j ∈ I, i  j , a set of LA,I -sentences. We
claim that Σ is finitely satisfiable. So let Σ0 be a finite subset of Σ. Then only
finitely many of the ci , say, ci1 , . . . , cin , are used in Σ0 . Now choose a1 , . . . , an
from A, all distinct. Expand A to an LA,I -structure A by interpreting ca as
a for each a ∈ A, ci j as a j for j = 1, . . . , n, and every other ci as a1 . Then
A |= Σ0 . Thus Σ is finitely satisfiable, and hence, by compactness, it has a
model, say, D , with domain D. Let D be the reduct of D to L. Since D |= Σ,

the elements cDi of D are all distinct, and so |D|  κ.
Since D |= CDiag(A), by Proposition 13.2, there is an elementary
π
embedding A −→ D. Using Remark 13.3, we can assume A  D.
We are nearly done, but the elementary extension D we have found has
cardinality at least κ, and we want an elementary extension of cardinality
exactly κ. Let S be a subset of D containing A and of cardinality κ. Then,
by Theorem 12.6, there is B  D of cardinality κ containing S and hence also
containing A.
We must show A  B. Let ϕ( x̄) be a formula and ā ∈ A. Then

A |= ϕ(ā) iff D |= ϕ(ā) iff B |= ϕ(ā)

because A  D and B  D. So A  B. 

13.3 Non-standard Natural Numbers

A proper elementary extension M of Ns-ring is called a non-standard model of
Th(Ns-ring ). An element of M  N is called a non-standard natural number,
while in this context, elements of N are called standard natural numbers. We
give a few examples of properties these non-standard natural numbers can
have.
Lemma 13.7 Suppose Ns-ring  M. Then any element a ∈ M  N is greater
than every standard natural number.
Proof Suppose not, say, a  n. We use n as an abbreviation for the term
(1 + · · · + 1), and we use  as an abbreviation for the formula defining it.
 13.3 Non-standard Natural Numbers 73

  
Now N |= ∀x x  n → ni=0 x = i , so since M ≡ N, we have M |=
   
∀x x  n → ni=0 x = i , so M |= ni=0 a = i, and then a is a standard natural
number, a contradiction. 
Proposition 13.8 Let M be any proper elementary extension of Ns-ring . Then
there is a non-standard prime number in M.
Proof Note that the formula ψ(y) given by

∀x[∃z[x · z = y] → (x = 1 ∨ x = y)] ∧ y  1

defines the set of prime numbers in Ns-ring . We have

N |= ∀x∃y[ψ(y) ∧ x < y]

and so also M |= ∀x∃y[ψ(y) ∧ x < y]. Thus, given a ∈ M  N, there is p ∈ M

prime such that a < p. Then p is greater than every standard natural number,
so it is non-standard. 
Proposition 13.9 There is an elementary extension M of Ns-ring with a number
β ∈ M with infinitely many distinct prime factors.
Proof We give a compactness argument. Let Lc = Ls-ring ∪ {c}, where c is a
new constant symbol. Let Σ = Th(Ns-ring )∪{∃x[n · x = c] | n ∈ N }, which states
that c is divisible by every (standard) natural number. Let Σ0 be a finite subset
of Σ, and let N be the largest natural number n such that ∃x[n· x = c] appears in
Σ0 . We make a model A of Σ0 by taking Ns-ring and interpreting c as N!. Then
A |= Σ0 , so Σ0 is satisfiable. By compactness, there is a model M+ of Σ. Let
M be the reduct to the language +, ·, 0, 1. By Proposition 13.2, Ns-ring  M,
+
and the element β = cM of M is divisible by every standard natural number,
in particular every standard prime number. 
We finish our brief discussion of non-standard natural numbers with an
observation about the twin prime conjecture, a famous problem in number
theory.
Conjecture 13.10 (Twin prime conjecture) There are infinitely many primes
P in N such that P + 2 is also prime. (Such a pair (P, P + 2) is called a pair of
twin primes.)
Proposition 13.11 The twin prime conjecture is true if and only if there is some
non-standard model of arithmetic M and at least one pair of non-standard twin
primes in M.
74 13 Elementary Extensions

Sketch proof Let ϕ(x) be the formula ψ(x) ∧ ψ(x + 2), where ψ defines the
set of primes as above. If the twin prime conjecture is true, then the subset of
N defined by ϕ(x) is infinite, and a compactness argument shows there is some
M with a non-standard realisation of ϕ(x). If the twin prime conjecture is false,
then we can list all the realisations of ϕ(x) as n1 = 3, n2 = 5, n3 = 11, . . . , nr ,
for some finite r, and we have

⎡ ⎤
⎢⎢⎢ 
r ⎥⎥
Ns-ring |= ∀x ⎢⎢⎣ϕ(x) → x = ni ⎥⎥⎥⎦ .
i=1

r
But then, whenever M |= Th(Ns-ring ), we have M |= ∀x[ϕ(x) → i=1 x = ni ],
and hence there are no non-standard twin primes. 
At the time of writing (2018), the twin prime conjecture is a long-standing
open problem, and we might hope that this reformulation of it would help to
solve it. Unfortunately, it does not, because there is no easy way to construct
explicit non-standard models of Th(Ns-ring ) and study them directly. However,
non-standard models are very useful for other theories where they can be
constructed.

13.4 Non-standard Real Numbers

Now we turn to elementary extensions of the real field Ro-ring .
Proposition 13.12 There is a non-Archimedean elementary extension of
Ro-ring .
Sketch proof Take a new constant symbol c and consider the set of sentences
Σ = CDiag(Ro-ring ) ∪ {n < c | n ∈ N }. A similar compactness argument to those
we have considered already shows that Σ is satisfiable, so let R+ be a model,
and let R be the reduct to the language Lo-ring . Then, as in the proof of the
+
Upward Löwenheim-Skolem theorem, R  R, and cR witnesses that R is non-
Archimedean. 
In particular, R is not a complete ordered field, yet every true first-order
statement about R is also true in R, including many theorems of analysis like
the intermediate value theorem, but restricted to definable functions.
In fact, one can prove that every proper elementary extension of R is non-
Archimedean. See Exercise 13.7(a).
 Exercises 75

Definition 13.13 If R is a non-Archimedean ordered field, we say an

element a of R is infinite if |a| is greater than every standard natural number,
finite if it is not infinite, and infinitesimal if |a|  0 but, for every n ∈ N, we
have |a| < 1n .

Warning: the notions of finite and infinite non-standard real numbers are
different from the notions of finite and infinite as applied to sets.

Observe that a is infinite if and only if 1a is infinitesimal. Let R be a proper

elementary extension of R, and let r ∈ R. There is a cloud of numbers in R
which are infinitesimally close to r, namely, all the numbers ρ ∈ R such that
|ρ − r| is infinitesimal or 0. We can use these non-standard real numbers to
do non-standard analysis, for example to define the derivative of a function
in the intuitive way it is sometimes done in a calculus course rather than the
usual way of doing it in analysis. Exercise 13.7(c) illustrates this method for
polynomial functions.

Exercises
13.1 Prove Lemma 13.5.
13.2 Show that there is a model A of Th(Z< ) such that Q< embeds in A. Can
the embedding be elementary?
13.3 Let T be a complete theory in a language L of cardinality λ, and suppose
that T has an infinite model. Show that T has models of all cardinalities
greater than or equal to λ and no finite models.
13.4 Complete the proof of Proposition 13.11 by writing down the compact-
ness argument.
13.5 A Mersenne prime is a prime number of the form 2n − 1. Prove that there
are infinitely many Mersenne primes if and only if there is a non-standard
Mersenne prime. [You may assume that f (n) = 2n is a definable function
in Ns-ring . For a more difficult exercise, prove that it is definable.]
13.6 Show that there is a countable model R of Th(Ro-ring ) which is non-
Archimedean. Show that neither of R nor Ro-ring can be embedded in
the other.
13.7 Let R be any proper elementary extension of the ordered field Ro-ring of
real numbers.

(a) Prove that there is an infinitesimal in R.

76 13 Elementary Extensions

(b) Let I be the set consisting of zero and all infinitesimals in R. Prove
that I is closed under addition and multiplication and that if α ∈ I
and r ∈ R, then α · r ∈ I.
(c) Let p(X) ∈ R[X] be a polynomial with standard real coefficients.
Show there is another polynomial q(X) ∈ R[X] such that whenever
x is a standard real and α is infinitesimal, there is β ∈ I such that
p(x + α) − p(x)
= q(x) + β.
α
(d) Explain briefly how we can use the above result to define the
derivative of a polynomial function. How does this method differ
from the usual approach in analysis?
13.8 Sketch a proof of the Upward Löwenheim–Skolem theorem. You should
give all the key ideas and explain how they fit together, but you do not
need to give all the details.
13.9 An elementary chain of L-structures is a chain
A 1  A2  · · ·  A n  · · · .

Let A = n∈N+ An as in Exercise 5.9.
(a) Show that for each n ∈ N we have An  A.
(b) Now suppose there is B such that each An  B. Prove that A  B.
13.10 This exercise uses the method of diagrams and compactness to prove
that an axiomatisable class C is axiomatised by universal sentences if
and only if it is closed under taking substructures. Suppose that C is
axiomatisable and that whenever B ∈ C and A is a substructure of B,
A ∈ C. Let Σ = Th(C), and let Σ∀ = {σ ∈ Σ | σ is a universal sentence }.
Let A |= Σ∀ , and let ϕ1 (c̄a ), . . . , ϕr (c̄a ) ∈ Diag(A).
 
(a) Show that Σ  ∀ x̄ ¬ ri=1 ϕi ( x̄) .
(b) Using this idea, show that Σ ∪ Diag(A) is finitely satisfiable.
(c) Deduce that C = Mod(Σ∀ ).
(d) For the converse direction, prove that if C is axiomatised by a set of
universal sentences, then C is closed under taking substructures.
 14
Vector Spaces and Categoricity

From the Downward and Upward Löwenheim–Skolem theorems, we can see

that if a first-order theory T has an infinite model, then it has at least one model
of each infinite cardinality at least as large as |L|. We now consider theories
where there is only one model in some infinite cardinality. Such theories are
very interesting, not just because the theory determines the models as much
as is possible, but also because, as we will see later, this property has strong
consequences for the definable sets. Our prototypical example is that of vector
spaces.

14.1 Vector Spaces as Structures

Let K be a field. The language LK-VS is Ladgp = +, −, 0 together with one
unary function symbol λ· for each λ ∈ K. (Note that we are using the two
characters λ and · as a single symbol of our formal language.) We interpret λ·
as scalar multiplication by λ.

Definition 14.1 The theory of K-vector spaces, T K-VS , is the LK-VS -theory
given by the axioms for abelian groups, written additively in the language
Ladgp , together with the following axioms which describe how scalar multi-
plication works:

VS1. ∀x[0 · x = 0 ∧ 1 · x = x];

VS2λ . ∀xy[λ · (x + y) = (λ · x) + (λ · y)];
VS3λ,λ . ∀x[(λ · x) + (λ · x) = μ · x], where μ = λ + λ in K;
VS4λ,λ . ∀x[λ · (λ · x) = ν · x], where ν = λλ in K.

A K-vector space is a model of T K-VS .

77
78 14 Vector Spaces and Categoricity

Note that axioms 2, 3, and 4 are not single LK-VS -sentences but families of
sentences. Axiom 2 is really a family of axioms indexed by elements of K,
and axioms 3 and 4 are families of axioms indexed by pairs of elements of K.
These are called axiom schemes. We cannot write these as single axioms in
LK-VS because that would involve a universal quantifier which quantified over
elements of K. However, that cannot exist in our language, which only allows
us to quantify over elements of our structure, which is the vector space, not the
field. (It is also possible to consider a language where some of the elements of
a model are elements of the field and some are elements of the vector space,
but that turns out to be like studying the field rather than studying the vector
space.)
For any field K, there is a 0-dimensional K-vector space with only one
element, 0. If K is a finite field, there are other finite K-vector spaces. Since we
are mostly interested in infinite structures, it is useful to consider the theory of
infinite K-vector spaces as well.

Definition 14.2 The theory of infinite K-vector spaces, T K∞-VS , is the theory
axiomatised by the axioms for K-vector spaces together with sentences saying
that there are at least n distinct elements for each n ∈ N+ .

14.2 Some Linear Algebra of Vector Spaces

Fix a field K. Concepts such as linear independence and bases are well
known for finite-dimensional vector spaces but less well known for infinite-
dimensional vector spaces, so we will review them.

Definition 14.3 Let V be a K-vector space. (We will use the same letter V for
the vector space and its domain.)

• A subset S of V is linearly independent iff whenever s1 , . . . , sn ∈ S are

distinct and λ1 , . . . , λn ∈ K are such that ni=1 λi si = 0, then each λi = 0.

• A subset S of V is a spanning set for V iff every element of V can be written
as a (finite) linear combination of elements of S . That is, for every v ∈ V,

there are n ∈ N, s1 , . . . , sn ∈ S and λ1 , . . . , λn ∈ K such that v = ni=1 λi si .

• A basis of V is a linearly independent spanning set of V.
Note that there is no notion of an infinite linear combination here. That
would require a notion of convergence of infinite series, which needs some
topological or metric information we do not have in our algebraic setting.
There are three important facts which we will use.
 14.2 Some Linear Algebra of Vector Spaces 79

Facts 14.4 (i) Let B be a basis for V, and v ∈ V. Then there are a unique
n ∈ N, unique distinct b1 , . . . , bn ∈ B and unique λ1 , . . . , λn ∈ K  {0}

such that v = ni=1 λi bi . That is, every element in V can be written

uniquely as a linear combination of elements of B.
(ii) Every linearly independent set can be extended to a basis. In particular,
every vector space has a basis.
(iii) Any two bases of V have the same cardinality.
Using (ii) and (iii), we can make the following definition.

Definition 14.5 The dimension dim V of a vector space V is the cardinality

of a basis.

We next show that for any field K, there are K-vector spaces of every
possible cardinal dimension.

Example 14.6 Let X be any set, and let K ⊕X be the set of functions f : X → K
of finite support, that is, such that there are only finitely many x ∈ X with
f (x)  0. We make K ⊕X into a K-vector space by pointwise operations:

( f + g)(x) := f (x) + g(x) and (λ · f )(x) := λ f (x).

For each x ∈ X, let b x be the function X → K given by

⎧
⎪
⎪
⎨1 if y = x,
b x (y) = ⎪
⎪
⎩0 otherwise.

Then b x has finite support, so b x ∈ K ⊕X . We claim that B := {b x | x ∈ X } is

a basis for K ⊕X . First we show that B is linearly independent. Suppose f :=

n
λi b xi = 0, with the xi distinct. Then for each j = 1, . . . , n, we have f (x j ) =

i=1
i=1 λi b xi (x j ) = λ j . But f is the zero function, so each λ j = 0. So B is linearly
n

independent. Now let f ∈ K ⊕X , and suppose that {x1 , . . . , xn } is the support

of f , that is, the set of elements of X on which f takes a non-zero value. Let

λi = f (xi ). Then f = ni=1 λi b xi , so B is a spanning set.

There is an obvious bijection X → B given by x → b x , so |B| = |X|. Thus
dim K ⊕X = |X|.
Theorem 14.7 Suppose that V1 and V2 are K-vector spaces. Then V1  V2 if
and only if dim V1 = dim V2 .
Sketch Proof If π : V1 → V2 is an isomorphism and B is a basis for V1 ,
then it is an easy exercise to show that π(B) = {π(b) | b ∈ B } is a basis for V2 .
Then, since π is a bijection, it restricts to a bijection between B and π(B), so
80 14 Vector Spaces and Categoricity

dim V1 = dim V2 . Conversely, suppose that B1 is a basis for V1 and B2 is a basis

for V2 , and α : B1 → B2 is a bijection. Define π : V1 → V2 by
⎛ n ⎞
⎜⎜⎜ ⎟⎟ n
π ⎜⎝ λi bi ⎟⎟⎟⎠ =
⎜ λi α(bi ).
i=1 i=1

Then π is a well-defined function on all of V1 because B1 is a basis (using Fact

14.4(a)). It is injective because B2 is linearly independent, and it is surjective
because B2 is a spanning set for V2 . It is also a linear map, that is, it preserves
addition and scalar multiplication, which means it is an LK-VS -isomorphism.
So V1  V2 . 

14.3 The Cardinality of a Vector Space

Proposition 14.8 If V is an infinite K-vector space, then

|V| = max(|K|, dim V).

Proof Let B be a basis for V. The inclusion B → V is an injective map, so

|V|  |B| = dim V. Since V is infinite, it is not the zero-dimensional vector
space, so B  ∅, and we can choose b ∈ B. Then there is an injective function
K → V given by λ → λ · b, so |V|  |K|. So |V|  max(|K|, dim V).
Let X = K × B, and consider the set S of all finite strings of X. We can
define a function f : S → V by

f ((λ1 , b1 ), (λ2 , b2 ), . . . , (λn , bn )) = λ1 b1 + λ2 b2 + · · · + λn bn .

This function is surjective because B is a basis, so, using Fact 10.2(iii), we have
|V|  |S |. By Proposition 10.5, we have |S | = max(|X|, ℵ0 ), and we also have
|X| = |K| · |B|. At least one of K and B is infinite (since otherwise V would be
finite), and so

|S | = max(|K|, |B|, ℵ0 ) = max(|K|, |B|).

Thus |V| = max(|K|, |B|), as required. 

Corollary 14.9 Up to isomorphism, there is exactly one model of the theory
of K-vector spaces of each infinite cardinality strictly greater than |K|. 
 14.4 Categoricity 81

14.4 Categoricity
We have seen that the theory of K-vector spaces has exactly one model (up to
isomorphism) of each infinite cardinality κ such that κ > |K|. There is a name
for such theories.

Definition 14.10 A theory T is categorical in cardinality κ or κ-categorical

iff there is a model of T of cardinality κ and any two models of cardinality κ are
isomorphic. ℵ0 -categorical is also called countably categorical. A structure A
is said to be κ-categorical iff Th(A) is.

Recall that an L-theory T is complete iff for every L-sentence ϕ, either T  ϕ

or T  ¬ϕ. One useful consequence of categoricity is completeness of a theory.
Lemma 14.11 (Łos–Vaught test for completeness) If an L-theory T is
κ-categorical for some κ  |L| and T has no finite models, then T is complete.
Proof Let ϕ be an L-sentence, and let M be the model of T of cardinality κ.
Then either M |= ϕ or M |= ¬ϕ. Suppose that M |= ϕ. Since every model of
T of cardinality κ is isomorphic to M, using Proposition 3.7, we see that there
is no model of cardinality κ of T ∪ {¬ϕ}. But then, by the Löwenheim–Skolem
theorems, there is no infinite model of T ∪{¬ϕ}. We assume that T has no finite
models, and hence there is no model of T ∪ {¬ϕ}. But then every model of T is
a model of ϕ, that is, T  ϕ. Otherwise, we have M |= ¬ϕ, and then the same
argument shows that T  ¬ϕ. So T is complete. 
Corollary 14.12 For any field K, the theory T K∞-VS is complete. 
Corollary 14.13 If K is an infinite field, for example R, then all K-vector
spaces of non-zero dimension are elementarily equivalent.
Proof They are models of the same complete theory. 
It may seem very unexpected that R1 , R2 , and R3 are all elementarily
equivalent as vector spaces. We think of geometry in 1, 2, and 3 dimensions as
being very different. The conclusion to draw is that very little of that geometry
is captured by the notion of an R-vector space. Essentially it is because our
language does not allow us to quantify over the scalars, R.

Example 14.14 There is no LR-VS -formula ϕ(x, y) which expresses that two
vectors are linearly independent.
Proof If such a ϕ(x, y) did exist, then we would have R2 |= ∃xy[ϕ(x, y)], but
R1 |= ¬∃xy[ϕ(x, y)], which is impossible, since R1 ≡ R2 . 
82 14 Vector Spaces and Categoricity

Exercises
14.1 If K is a finite field and V is a K-vector space of dimension d, show
that |V| = |K|d .
14.2 If K is an infinite field, show that T K∞-VS is axiomatised by the axioms
for K-vector spaces together with the single axiom ∃x[x  0].
14.3 If V is a one-dimensional K-vector space, show that Aut(V) is isomor-
phic to the multiplicative group K × of non-zero elements of K.
14.4 If V is an n-dimensional K-vector space, show that Aut(V) is isomor-
phic to GLn (K), the group of invertible n × n matrices with entries in
K.
14.5 For any field K and any non-zero vector space V, show that there are
exactly four definable subsets of V.
14.6 Show that if a1 , . . . , an and b1 , . . . , bn are each linearly independent n-
tuples from V, then there is an automorphism π of V such that π(ai ) = bi
for each i.
14.7 Write F3 for the field with three elements, and let V be an infinite-
dimensional F3 -vector space. What are all the definable subsets of V
and of V 2 ?
14.8 In the language L= with no relation, function, or constant symbols, let
T be the empty theory, that is, (the deductive closure of) the empty set
of L= -sentences. What is a model of T ? Show that T is κ-categorical
for every cardinal κ.
14.9 Sketch a proof of Corollary 14.13, including the results used to prove
it. You should give all the key ideas and explain how they fit together,
but should not give all the details.
14.10 Suppose that A is a model of the theory DTFAG of divisible torsion-
free abelian groups from Exercise 9.9. Show that for each λ ∈ Q,
there is an Ladgp -formula which defines multiplication by λ on A in
such a way that A becomes a Q-vector space. Deduce that the theory
DTFAG ∪ {∃x[x  0]} is complete.
 15
Linear Orders

In this chapter we will look at some theories of linearly ordered sets, aiming
to find complete axiomatisations using the method of categoricity. For dense
linear orders such as R< and Q< we will achieve this using the important back-
and-forth method.

Definition 15.1 A linear order on a set is a binary relation < satisfying:

Transitivity ∀xyz[(x < y ∧ y < z) → x < z];

Irreflexivity ∀x[¬x < x]; and
Linearity ∀xy[x < y ∨ x = y ∨ y < x].

A set equipped with a linear order on it is a linearly ordered set.

We will use the abbreviations x  y, y > x, and y  x, as usual.

Examples 15.2 The ordered set of real numbers is written R< . Likewise, we
have linearly ordered sets N< , Z< , Q< , and I< , where I is the closed unit interval
I = [0, 1] ⊆ R.

Now we consider ways to distinguish these examples.

15.1 Endpoints
Both N< and I< have a least element, so they are models of the sentence

∃x∀y[x  y],

while R< , Q< , and Z< do not have a least element. Similarly, I< has a greatest
element, so is a model of
∃x∀y[y  x],

83
84 15 Linear Orders

while the other four examples are not. Least and greatest elements of a linearly
ordered set are called endpoints, so a linearly ordered set is without endpoints
if it satisfies the sentence

∀x∃yz[y < x ∧ x < z].

15.2 Discretely and Densely Ordered Sets

Another property of Z< and N< is that they are discretely ordered, which means
that each element (except the greatest element, if it exists) has a next element
greater than it, and also (except for the least element) a next element below it.
This is captured by the sentences

∀x[∃w[w < x] → ∃y[y < x ∧ ∀z[z < x → z  y]]]

and
∀x[∃w[w > x] → ∃y[y > x ∧ ∀z[z > x → z  y]]].

Note that any finite non-empty linearly ordered set is discrete and has
endpoints, and indeed it must be isomorphic to {1, 2, 3, . . . , n}; < for some
n ∈ N+ . On the other hand, R< , Q< , and I< are dense linear orders, captured by
the axiom
∀xy[x < y → ∃z[x < z ∧ z < y]].

Clearly a linearly ordered set cannot be both dense and discrete, although it is
possible to be neither.
We have found sentences to distinguish three of our five examples from
the others, and we will show that it is impossible to distinguish the last
two, R< and Q< , by a first-order sentence. Let DLO be the theory of dense
linear orders without endpoints, as given by the axioms stating transitivity,
irrelexivity, linearity, no endpoints, and density, together with a non-triviality
axiom ∃x[x = x] to rule out the empty linear order. We have Q< |= DLO and
R< |= DLO.

15.3 Completeness of DLO

Theorem 15.3 Any two countable models of DLO are isomorphic to each
other.
The method of proof we will use is called back and forth, and it has become
very important in model theory. An early stage in the construction is shown in
Figure 15.1.
 15.3 Completeness of DLO 85

Figure 15.1 An early stage of the back-and-forth construction.

Proof of Theorem 15.3 Suppose A, B |= DLO and both are countable. Since
DLO has no finite models, A and B are both countably infinite. So we can
list A as (an )n∈N and B as (bn )n∈N . We will construct an isomorphism piece
by piece. In fact, for each n ∈ N, we will inductively construct finite subsets
An ⊆ A and Bn ⊆ B and a bijection πn : An → Bn such that:
  
• for each a, a ∈ An , a < a iff πn (a) < πn (a );
• An ⊆ An+1 , Bn ⊆ Bn+1 , and πn ⊆ πn+1 ; and
• an ∈ A2n+1 and bn ∈ B2n+2 .
In particular, πn is an isomorphism between An ; < and Bn ; <.
We start with A0 = B0 = ∅, and π0 the empty function. Now suppose n is
even, say, n = 2m, and we have constructed An , Bn , and πn . If am ∈ An , then
set An+1 = An , Bn+1 = Bn , and πn+1 = πn . By induction, the above conditions
are satisfied. Otherwise, am  An , and we define An+1 = An ∪ {am }. We must
find a suitable element in B to be πn+1 (am ). If n = 0, we choose π1 (a0 ) = b0 .
Otherwise, there are three cases. Either (i) am is less then every element of An
or (ii) am is greater than every element of An or (iii) there are c, d ∈ An with
c < am and am < d, and no elements of An strictly between c and d. In each
case, because B |= DLO, there is b ∈ B such that b is (i) less than every element
of Bn or (ii) greater than every element of Bn or (iii) between πn (c) and πn (d).
To be precise, we choose the br for the smallest possible value of r ∈ N with
the desired property. Then we define Bn+1 = Bn ∪{br } and πn+1 = πn ∪{(am , br )}.
Now suppose that n is odd, say, n = 2m + 1. If bm ∈ Bn , then set An+1 = An ,
Bn+1 = Bn , and πn+1 = πn . Otherwise, let Bn+1 = Bn ∪ {bm }. Then we use the
same process as above to find ar ∈ A.

Now let π = n∈N πn . Our construction ensures that π is defined on all of A
and its image is all of B. It also preserves < (in particular, it is injective), and
so it is an isomorphism from A to B. 
86 15 Linear Orders

Corollary 15.4 The theory DLO is complete.

Proof It is easy to see that DLO has no finite models. DLO is ℵ0 -categorical,
so, by the Łos-Vaught test (Lemma 14.11), it is complete.

Since every countable model of DLO is isomorphic to Q< , it is natural to
consider uncountable models. For example, is every model of DLO of the same
cardinality as R isomorphic to R< ? Consider a linear order M which consists
of a copy of Q and a copy of R, with every element of Q less than every element
of R. Then |M| = |Q| + |R| = |R|.
It is easy to check that M |= DLO, but we can show that M  R. Indeed,
suppose π : M → R were an isomorphism, and let r = π(0Q ). Then π
maps the set a ∈ M  a < 0Q bijectively to the set {b ∈ R | b < r }, but the
first set is countable and the second set is uncountable, so there cannot be a
bijection between them. This contradiction shows that M  R. So DLO is
not categorical in |R|, and in fact DLO is not categorical in any uncountable
cardinal.

Exercises
15.1 Give an example of a linearly ordered set which is neither discrete nor
dense.
15.2 Suppose that A is a discrete linear order without endpoints. Show that
the successor function which takes an element to the one immediately
above it is definable.
15.3 Let DLOep be the theory of dense linear orders with least and greatest
endpoints together with the axiom ∃xy[x < y]. Show that DLOep is
ℵ0 -categorical.
15.4 Let a, b ∈ Q. Show that there is an automorphism of Q< taking a to b.
15.5 Let A be a countable linearly ordered set. Show that there is an
embedding of A into Q< . [Hint: use the idea in the back-and-forth
proof, but only going forth, not back.]
15.6 Show that every non-empty finite linearly ordered set is discrete and
has endpoints. [Hint: use induction on the cardinality of the set.]
15.7 Show that the theory of non-empty discrete linear orders without
endpoints is not ℵ0 -categorical.
 Exercises 87

15.8 If A and B are two linear orders, their lexicographic product A ×lex B
is the set A × B, linearly ordered by setting (a, b) < (c, d) if and only if
either a < c or (a = c and b < d). Show that any discrete linear order
without endpoints can be written as A ×lex Z< , for some linear order A.
15.9 Show that each singleton {n} is definable in N< . Using this fact and a
compactness argument, show that Th(N< ) is not ℵ0 -categorical.
15.10 Show that ThZ; <, 0 is not ℵ0 -categorical. Hence or otherwise show
that Th(Z< ) is also not ℵ0 -categorical.
 16
The Successor Structure

This chapter is a guided extended exercise devoted to studying the structure

Nsucc = N; s, 0, the natural numbers with a constant symbol naming 0, and
the successor function s, a unary function defined by s(n) = n + 1. We will
write down axioms for its theory, classify all the models of our axioms, and
use the Łos–Vaught test to prove that the list of axioms is complete. We will
then show how to use the other models of T S to give information about the
definable sets in Nsucc .

16.1 Axioms
We first observe some basic properties of the successor function:
S1. s is injective;
S2. 0 is not in the image of s; and
S3. every element of N other than 0 is in the image of s.

Exercise 16.1 Write down first-order sentences to capture these properties.

Slightly less obvious is that there are no ‘cycles’, captured by the following
axiom scheme:
S4n . ∀x[sn (x)  x] for n ∈ N+ ,
where s0 (x) = x and sn+1 (x) = s(sn (x)) for each n ∈ N.
Let us write T S for the set of axioms {S1, S2, S3} ∪ {S4n | n ∈ N+ }.

Exercise 16.2 Write down a similar list of axioms that is satisfied by Zsucc =
Z; s, 0.

88
 16.2 Models 89

Exercise 16.3 Adapt the lists of axioms to the structures N; s and Z; s
which do not have a constant symbol naming 0.

16.2 Models
By design, Nsucc is a model of T S . By the Upward Löwenheim–Skolem
theorem, there are models of T S of every infinite cardinality. Now we will
investigate what these must look like.
Exercise 16.4 Show that every model of T S has a substructure which is
isomorphic to Nsucc . We will identify this substructure with N for notational
convenience.
Now suppose that M |= T S and M  Nsucc . Then there is an element
a ∈ M  N.
Exercise 16.5 Show that a lies in a substructure of M; s isomorphic to Z; s.
(We have taken a reduct to forget the constant 0; otherwise, it would not be a
substructure.) So any model M |= T S consists of Nsucc and a set of copies of
Z; s. We can be more precise.
For any set I, define MI = MI ; s, 0 to be the structure with domain MI =
N ∪ (I × Z), the constant symbol 0 naming the zero from N, and with s defined
by s(n) = n + 1 for n ∈ N and s((i, n)) = (i, n + 1) for each (i, n) ∈ I × Z.
Exercise 16.6 Verify that for any set I, MI |= T S .

Exercise 16.7 Show that if M is any model of T S , then there is a set I such
that M is isomorphic to MI .

Exercise 16.8 Show that MI1  MI2 if and only if I1 and I2 have the same
cardinality.
So now we have a classification of all the models of T S , up to isomorphism.
Exercise 16.9 Show that the cardinality of MI is |I| + ℵ0 .

Exercise 16.10 How many models are there of each cardinality, up to isomor-
phism? In which cardinalities is T S categorical?

Exercise 16.11 Prove that T S is a complete theory.

90 16 The Successor Structure

Exercise 16.12 What changes if we take Zsucc , N; s, or Z, s instead of
Nsucc ?

16.3 Definable Sets

Now we want to understand some of the definable sets in the original structure
Nsucc . It is possible to use the method of quantifier elimination to get a complete
understanding of the definable sets. See Exercise 18.4. Here we take a more
direct approach.
First observe that each element n ∈ N is named by the closed term sn (0). So
the singletons {n} are all definable.

Exercise 16.13 Show that every finite subset of Nsucc is definable and that
every cofinite subset (that is, the complement of a finite set) is also definable.

The technique we know to show that a subset is not definable is to show

that it is not preserved under some automorphism. However, Nsucc itself has no
automorphisms (except the identity), so we cannot use that technique directly.
However, we can use it together with our understanding of the other models of
T S to deduce that certain subsets of N itself are not definable.

Exercise 16.14 Suppose I  ∅, and let a, b ∈ MI  N. Show there is an

automorphism π of MI such that π(a) = b.

Exercise 16.15 Deduce that if S is a definable subset of MI , then we must

have either MI  N ⊆ S or (MI  N) ∩ S = ∅.

Exercise 16.16 Let A be any infinite structure and ϕ(A) ⊆ A and θ(A) ⊆ A
be infinite definable sets. Use a compactness argument to show there is B,
elementarily equivalent to A, such that the subsets ϕ(B) and θ(B) of B are
uncountable.

Exercise 16.17 Now suppose that ϕ(x) is a formula such that both ϕ(Nsucc )
and ¬ϕ(Nsucc ) are infinite. By considering ϕ(MI ) and ¬ϕ(MI ) for an
uncountable set I, find a contradiction. Deduce that the subsets of N which
are definable in Nsucc are exactly the finite and cofinite subsets. (S is cofinite
if its complement N  S is finite.)
 16.3 Definable Sets 91

Similar arguments using automorphisms of the models MI can be used to

understand what the definable subsets of Nn look like for n > 1. They are not
just finite or cofinite, for example {(a, a + 3) | a ∈ N } ⊆ N2 is defined by the
formula ϕ(x1 , x2 ) given by x2 = s(s(s(x1 ))). We restrict ourselves to showing
that Nsucc is genuinely different from N< .
Exercise 16.18 Take I = {i, j}, and show there is an automorphism π of MI
such that π((i, 0)) = ( j, 0) and π(( j, 0)) = (i, 0).

Exercise 16.19 Show that no formula can define a linear order on MI , and
deduce that the order < on N is not a definable subset of N2 .
 Part IV

Characterising Definable Sets

In this part of the book we reach the main theme: the study of the definable
sets of a structure. We first introduce the important method of quantifier
elimination. For those structures where it holds, it means that to understand
the definable sets, it is enough to consider the sets defined by formulas without
quantifiers. The basic back-and-forth method suffices to prove quantifier elim-
ination for dense linear orders, and we introduce the method of substructure
completeness, which works for many other examples, including vector spaces
and (in Part VI) algebraically closed fields.
We then show how definable sets form Boolean algebras and relate the
atoms of these Boolean algebras to principal formulas. Finally, we work out
several examples, including the important case of the real field, where the
definable sets are also known as semi-algebraic sets.

93
 17
Quantifier Elimination for Dense Linear Orders

In Chapter 15 we found an axiomatisation for the complete theory of Q< by

using the back-and-forth method to prove countable categoricity of the theory
DLO. In this chapter we will see that a minor variation of the same back-and-
forth method allows us to understand what all the subsets of Qn which are
definable in Q< are.

17.1 Principal DLO Formulas

Our first task is to understand which subsets of Qn are definable with
quantifier-free formulas ϕ( x̄). Typically, in any structure with a reasonably
chosen language, these are relatively easy, and it is quantifiers which cause
the difficulties. Quantifier-free formulas are Boolean combinations of atomic
formulas, which for the language L< are of the form (xi < x j ) and (xi = x j ).
Examples for n = 5 include

x1 < x2 ∧ x2 < x3 ∧ x3 = x4 ∧ x4 < x5

and
x3 < x1 ∧ x2 = x4 ∧ (x3 < x5 ∨ x4 = x5 ).

The first of these completely specifies the order of the variables x1 , . . . , x5 ,

while the second only partially specifies the order. For example, it does not
specify whether x2 < x3 , x2 = x3 , or x2 > x3 . We capture the idea of completely
specifying the order of the variables in the following definition.

Definition 17.1 A principal DLO formula for the variables x1 , . . . , xn is a

formula ψ(x1 , . . . , xn ) of the form x1 = x1 if n = 1 or

95
96 17 Quantifier Elimination for DLO

$
n−1
% &
xσ(i) i xσ(i+1) ,
i=1

where n > 1, σ is a permutation of {1, . . . , n} and each i is either < or =.

Evidently these formulas are quantifier-free L< -formulas.

Lemma 17.2 For each n ∈ N+ , there are only finitely many principal DLO
formulas for the variables x1 , . . . , xn . Furthermore, every n-tuple from Q<
satisfies a principal DLO formula.
Proof The reader should attempt this as an exercise. 

17.2 Automorphisms of Q<

The back-and-forth method was used to produce isomophisms between any
two countable dense linear orders (without endpoints). Now we use it to
produce automorphisms.
Proposition 17.3 Suppose that ψ( x̄) is a principal DLO formula and ā, b̄ ∈ Qn
are such that Q< |= ψ(ā) ∧ ψ(b̄). Then there is an automorphism π of Q< such
that π(ā) = b̄.
Proof We have ā = (a1 , . . . , an ) and b̄ = (b1 , . . . , bn ). Extend both of these
tuples to enumerations (am )m∈N+ and (bm )m∈N+ of Q. Let An = {a1 , . . . , an }, let
Bn = {b1 , . . . , bn }, and let πn : An → Bn be given by πn (ai ) = bi for i = 1, . . . , n.
Since ψ( x̄) is a principal DLO formula, we have for each 1  i, j  n that
ai < a j iff πn (ai ) < πn (a j ). Now for each m ∈ N such that m > n we construct
πm by the back-and-forth method, exactly as in the proof of Theorem 15.3.

Then setting π = mn πm , we see that π is an automorphism of Q< such that
π(ā) = b̄. 

17.3 Quantifier Elimination for Q<

Theorem 17.4 For any n ∈ N+ , any non-empty subset of Qn which is definable
in Q< is defined by a finite disjunction of principal DLO formulas.
Proof Let n ∈ N+ , and suppose ϕ(x1 , . . . , xn ) is an L< -formula such that
ϕ(Q< )  ∅. Let P be the set of all principal DLO formulas ψ such that

Q< |= ∃ x̄[ϕ( x̄) ∧ ψ( x̄)], and let Ψ( x̄) = ψ∈P ψ( x̄). By Lemma 17.2, the set
P is finite, so Ψ( x̄) is a first-order L< -formula.
 17.4 Quantifier Elimination for Theories 97

We will show that Ψ( x̄) defines the same subset of Qn as ϕ( x̄). Suppose
Q< |= ϕ(ā). Then ā satisfies some principal DLO formula ψ( x̄), and then ψ ∈ P
by definition of P. Since Q< |= ψ(ā), we have Q< |= Ψ(ā).
Conversely, suppose that Q< |= Ψ(ā). Then there is ψ ∈ P such that Q< |=
ψ(ā), and since ψ ∈ P, there is b̄ ∈ Qn such that Q< |= ϕ(b̄) ∧ ψ(b̄). We have
Q< |= ψ(ā) ∧ ψ(b̄), so by Proposition 17.3, there is an automorphism π of Q<
such that π(ā) = b̄. Since Q< |= ϕ(b̄), by Proposition 3.7, we have Q< |= ϕ(ā).
So Ψ( x̄) and ϕ( x̄) define the same subset of Qn , as required. 
Definition 17.5 A structure A is said to have quantifier elimination if and
only if, for every n ∈ N+ , every definable subset of An is defined by a quantifier-
free formula.
Since disjunctions of principal DLO formulas are quantifier-free formulas,
Theorem 17.4 implies that Q< has quantifier elimination.

17.4 Quantifier Elimination for Theories

Quantifier elimination also makes sense for theories, not just structures. This
is a more general notion because it makes sense even for incomplete theories.
Definition 17.6 An L-theory T has quantifier elimination, if for every n ∈ N
(including n = 0), for every L-formula ϕ(x1 , . . . , xn ), there is a quantifier-free
L-formula θ(x1 , . . . , xn ) such that T  ∀ x̄[ϕ( x̄) ↔ θ( x̄)].
For n = 0 in this definition, that is, for sentences, there is a minor technical
problem. Consider the sentence ∀x∃y[x < y]. In the theory DLO, this sentence
is true, so for DLO to have quantifier elimination we should find a quantifier-
free L< -sentence which is a consequence of the DLO-axioms. However, in
languages, such as L< , which have no constant symbols, by our original
definition of L-formulas, there are no quantifier-free sentences at all! To solve
this problem, we introduce two new atomic sentences,  and ⊥, which are
interpreted as true and false, respectively, in any structure. For a language with
a constant symbol c we could use the sentences c = c and c  c instead. With
this convention we have DLO  ∀x∃y[x < y] ↔ .
Lemma 17.7 Let A be an L-structure. Then Th(A) has quantifier elimination
if and only if A has quantifier elimination.
Proof Suppose Th(A) has quantifier elimination, n ∈ N+ , and S ⊆ An is
defined by a formula ϕ(x1 , . . . , xn ). Then there is a quantifier-free formula
θ(x1 , . . . , xn ) defining the same set S . So A has quantifier elimination.
98 17 Quantifier Elimination for DLO

Now suppose that A has quantifier elimination and ϕ(x1 , . . . , xn ) is an

L-formula. If n > 0, then ϕ(x1 , . . . , xn ) defines a subset of An , which
is also defined by some quantifier-free formula θ(x1 , . . . , xn ). So A |=
∀ x̄[ϕ( x̄) ↔ θ( x̄)], and hence Th(A)  ∀ x̄[ϕ( x̄) ↔ θ( x̄)]. If n = 0, then ϕ is
a sentence, so it is either true or false in A. So either Th(A)  (ϕ ↔ ) or
Th(A)  (ϕ ↔ ⊥). 
Putting together Theorem 17.4 and Lemma 17.7, we have proved the
following.
Theorem 17.8 The theory DLO has quantifier elimination. 

Exercises
17.1 List all of the principal DLO formulas for n = 3.
17.2 Prove Lemma 17.2.
17.3 Let T =∞ be the theory of infinite sets in the empty language L= . Prove that
T =∞ has quantifier elimination.
17.4 Let A be a finite set considered as an L= -structure. Prove that A has
quantifier elimination.
17.5 Show that the empty L= -theory does not have quantifier elimination.
Observe that, with the previous two exercises, this gives an example of
a theory without quantifier elimination, even though all of its models do
have quantifier elimination.
17.6 For the language consisting of two constant symbols, c and d, show
that the theory T ∞ stating that there are infinitely many elements has
quantifier elimination. By considering the sentence c = d, show that the
theory has exactly two completions.
17.7 Suppose that L is a language without constant symbols and that T is an
L-theory with quantifier elimination. Prove that T is a complete theory.
17.8 Show that the structure I< from Examples 15.2 does not have quantifier
elimination. However, show that it does have quantifier elimination if
we expand the language by adding constant symbols 0 and 1 for the least
and greatest endpoints.
 18
Substructure Completeness

In the previous chapter we showed that the theory DLO of dense linear orders
without endpoints has quantifier elimination. The method we used relies on
the fact that every finite tuple from a dense linear order satisifies a special
quantifier-free formula (a principal DLO formula) which we could show
determines all the definable sets the tuple lies in. As we shall see in Chapter 25,
this method only works because DLO is countably categorical. In this chapter
we will give another method for proving quantifier elimination, which also
works for theories which are not countably categorical.

18.1 Substructure Completeness and Quantifier Elimination

Consider the substructures of N< with domains {0, 1} and {1, 3}. These
substructures are both linear orders of length 2 and so are isomorphic to each
other as L< -structures. However, they are embedded in N< differently in a
way which is captured by formulas with quantifiers: for example we have
N< |= ∃x[x < 1] but N< |= ¬∃x[x < 0]. In fact we can see that the pairs
(0, 1) and (1, 3) satisfy all the same quantifier-free L< -formulas, but not all the
same L< -formulas, and this shows that N< does not have quantifier elimination.
More generally, suppose ā is a tuple from an L-structure M and A is the
substructure of M generated by ā, that is, A consists of the interpretations of
all the L-terms t( x̄) applied to ā. Then the quantifier-free information about ā in
M determines the isomorphism type of A. So if M has quantifier elimination,
we would expect this to be all the information there is about how A embeds as a
substructure of M. We now give a theorem which captures this intuition about
how quantifier elimination, which is about formulas, relates to substructures of
models.

99
100 18 Substructure Completeness

Recall from Chapter 13 that to form the diagram Diag(A) of an L-structure

A, we first expand L to LA and A to A+ by adding constant symbols for each
element of A. Then Diag(A) is the set of all atomic LA -sentences and negations
of atomic LA -sentences which are true in A+ . Models of Diag(A) correspond
to L-structure extensions of A.

Definition 18.1 An L-theory T is said to be substructure complete if,

whenever A is a substructure of a model of T , the theory T ∪ Diag(A) is a
complete LA -theory.
Proposition 18.2 Let T be an L-theory. Then the following are equivalent.

(i) T is substructure complete.

(ii) Whenever A is a finitely generated substructure of a model of T , the
theory T ∪ Diag(A) is a complete LA -theory.
(iii) T has quantifier elimination.
Proof (ii) is just a special case of (i), so (i) ⇒ (ii) is immediate.
For (ii) ⇒ (iii), let ϕ( x̄) be an L-formula, let c̄ be new constant symbols, and
let L = L ∪ {c̄}.
Let Σ be the set of all quantifier-free L -sentences σ such that
T  ϕ(c̄) → σ.
We claim that T ∪ Σ  ϕ(c̄). Suppose not, for a contradiction. Then there is
an L -structure M such that M |= T ∪ Σ ∪ ¬ϕ(c̄). Let A be the substructure of
M generated by c̄. Then every element of A is named by a closed term of L .
So adding new constant symbols for elements of A does nothing, and we can
regard LA as the same as L , and Diag(A) as a set of L -sentences.
Then M |= T ∪Diag(A)∪¬ϕ(c̄). By assumption, T ∪Diag(A) is a complete
L -theory, so is equal to Th(M). So T ∪ Diag(A)  ¬ϕ(c̄).
By compactness, there is a finite subset {ψ1 , . . . , ψr } of Diag(A) such that

T ∪ {ψ1 , . . . , ψr }  ¬ϕ(c̄). Let ψ = ri=1 ψi . Then T  ψ → ¬ϕ(c̄). So, taking
the contrapositive, T  ϕ(c̄) → ¬ψ. Since ¬ψ is a quantifier-free L -sentence,
¬ψ ∈ Σ.
Now we have M |= T ∪ Σ ∪ Diag(A), with T ∪ Diag(A)  ψ and Σ  ¬ψ,
so M |= ψ ∧ ¬ψ, a contradiction. Thus T ∪ Σ  ϕ(c̄) as claimed.
By compactness again, there is a finite subset Σ0 of Σ such that T ∪Σ0  ϕ(c̄).

Let θ(c̄) = Σ0 , with θ( x̄) an L-formula. Then θ(c̄) ∈ Σ so T  ϕ(c̄) ↔ θ(c̄).
Since the constant symbols c̄ are not in L and T is an L-theory, we deduce
T  ∀ x̄[ϕ( x̄) ↔ θ( x̄)]. So T has quantifier elimination, as required.

Now to prove (iii) ⇒ (i), assume that T has quantifier elimination. Let M be
a model of T and A a substructure of M. Write MA for the expansion of M by
 18.3 Quantifier Elimination and Completeness 101

naming each element a ∈ A by a constant symbol ca . Let σ be an LA -sentence.

Then σ is of the form ϕ(ca1 , . . . , can ) for some L-formula ϕ(x1 , . . . , xn ). Since
T has quantifier elimination, there is a quantifier-free L-formula ψ(x1 , . . . , xn )
such that

T  ∀x1 , . . . , xn [ϕ(x1 , . . . , xn ) ↔ ψ(x1 , . . . , xn )].

If MA |= σ, then M |= ϕ(ā), so M |= ψ(ā), and so Diag(A)  ψ(ca1 , . . . , can ).

Then T ∪ Diag(A)  σ.
Otherwise, MA |= σ, so MA |= ¬σ, and similarly, T ∪ Diag(A)  ¬σ. So
T ∪ Diag(A) is a complete LA -theory. Thus T is substructure complete. 

18.2 Application to Vector Spaces

∞
Theorem 18.3 Let K be any field. Then T K-VS has quantifier elimination.
Proof We will use Proposition 18.2. Suppose that V is an infinite K-vector
space. Let A be a finitely generated LK-VS -substructure of V, that is, a finite
dimensional subspace. Let κ be a cardinal larger than |K|. We will show that
∞
T K-VS ∪ Diag(A) is κ-categorical. Then, since it has no finite models, by the
Łos–Vaught test, it is complete.
So suppose W and W  are models of T K-VS ∞
∪ Diag(A) of cardinality κ. That
is, they are K-vector spaces which contain A as subspaces. Let {a1 , . . . , an } be
a basis for A, and extend it to bases B of W and B of W  . By Proposition 14.8,
dim W = dim W  = κ, so |B| = |B | = κ. Choose a bijection α : B → B such
that α(ai ) = ai for i = 1, . . . , n. Then α extends to an isomorphism of K-vector
spaces π : W → W  . Furthermore, π restricts to the identity map on A, so π is
∞
an isomorphism of LK-VS (A)-structures. Thus T K-VS ∪ Diag(A) is κ-categorical,
as required. 

18.3 Quantifier Elimination and Completeness

Substructure completeness does not necessarily imply completeness. From
Exercise 17.7, we know that if T is a theory with quantifier elimination in a
language with no constant symbols then T is complete. On the other hand,
Exercise 17.6 gives an example where this fails in a language with constant
symbols. The point is that a complete theory has to specify the truth values of
all the quantifier-free sentences. This amounts to stating what the substructure
102 18 Substructure Completeness

of a model generated by the interpretations of all the constant symbols is, up

to isomorphism.
Proposition 18.4 Suppose an L-theory T has quantifier elimination and there
is an L-structure A which embeds into every model of T . Then T is complete.
Proof Let M1 , M2 be two models of T , and let σ be an L-sentence such
that M1 |= σ. Since T has quantifier elimination, there is a quantifier-free
L-sentence θ such that T  (σ ↔ θ). So M1 |= θ. By Lemma 5.5, M1 |= θ
implies A |= θ, which implies M2 |= θ. So every model of T is a model of θ,
and hence T  θ, so T  σ. Thus T = Th(M1 ), so T is complete. 
Together, Propositions 18.2 and 17.7 relate completeness and quantifier
elimination, giving methods of proving either from the other. For vector spaces
we deduced quantifier elimination from completeness.
So far, the only method we have for proving completeness of a theory is
the Łos-Vaught test: a theory which is categorical in some infinite cardinality
and has no finite models is complete. However, for theories which are not
categorical in any infinite cardinality, another method is needed.
A more sophisticated version of the back-and-forth method allows one to
prove quantifier elimination by building isomorphisms only between parts of
models, not between whole models as is needed for categoricity. Proposition
18.4 can then be used to prove completeness of the theory. It is not essential
for the theory to have quantifier elimination in the original language. For
example, Exercise 17.8 shows that DLOep has quantifier elimination after
adding constant symbols naming the endpoints.
One important application of this extended back-and-forth technique is to
the real field.
Fact 18.5 The theory RCF of real closed fields is complete and has quantifier
elimination in the language Lo-ring .
The proof is a little beyond the scope of this book. One can also prove
that the theory of non-empty discrete linear orders without endpoints is
the complete theory of Z< using the back-and-forth method. See [Poi00,
Section 1.2] for details. We will give a different proof of that theorem in chapter
27.
This more elaborate back-and-forth method is described very elegantly
in the first six chapters of Poizat [Poi00] and in Chapter 3 of Hodges
[Hod93, Hod97] and is also covered in the books by Marker [Mar02] and
Tent and Ziegler [TZ12]. The application to RCF can also be found in
 Exercises 103

these books. The method can be framed as a two-player game, called an

Ehrenfeucht–Fräissé game. This approach is stressed in Väänänen [Vää11].

Exercises
18.1 Show that the structure Z< does not have quantifier elimination.
18.2 Let L be the language with a single unary relation symbol R. Let T ∞
be the L-theory which says there are infinitely many elements, and for
n ∈ N, let T n∞ be the L-theory which also says that there are exactly n
elements in the subset named by R. Show that T ∞ is not substructure
complete but that each T n∞ is.
18.3 Let K be a field. Show that the incomplete theory T K-VS does not have
quantifier elimination but every completion of it does.
18.4 Describe all the finitely generated substructures of models of the theory
T S = Th(Nsucc ) from Chapter 16. Prove that T S is substructure complete.
Deduce that Nsucc has quantifier elimination.
18.5 Let N = N; s be the reduct of Nsucc without the constant symbol for 0.
Show that N does not have quantifier elimination.
18.6 Give two isomorphic subrings of Rring which illustrate that the real field
does not have quantifier elimination in the language Lring .
18.7 Suppose that T is a complete theory with quantifier elimination, that A
and B are models of T , and that π : A → B is an embedding. Show that
π is an elementary embedding.
18.8 An L-theory T is said to be model complete iff, whenever M |= T , the
theory T ∪ Diag(M) is a complete L M -theory.
(a) Show that T is model complete iff whenever M1 , M2 |= T and
M1 ⊆ M2 , then M1  M2 .
(b) Suppose that T ‘eliminates quantifiers to the level of existential
formulas’, that is, for every L-formula ϕ( x̄), there is an existential
L-formula θ( x̄) such that T  ∀ x̄[ϕ( x̄) ↔ θ( x̄)]. Show that T is
model complete.
(c) Show that if T is model complete, then it eliminates quantifiers to
the level of existential formulas. [Hint: try to adapt the proof of
Proposition 18.2.]
 19
Power Sets and Boolean Algebras

The notion of a Boolean algebra originally came from logic as a way to

capture the logical operations AND, OR, and NOT in an algebraic way. It
can also be used to capture the set theoretic operations of intersection, union,
and complement. In the next chapter we will see how these two ideas come
together when looking at the collection of definable subsets of a structure. In
this chapter we introduce Boolean algebras in a slightly roundabout way, by
trying to find the complete theory of the power set of an infinite set, considered
as a partial order. It is an example of a class of structures which is not first-
order axiomatisable, but for which the search for axioms is very fruitful. Even
though this is not historically how these axioms were found, it provides another
illustration of the process of finding an axiomatisation for the theory of a
structure. In particular, it shows how it is often useful to change the language
(without changing which sets are definable).

19.1 Power Sets as Partially Ordered Sets

Given a set X, the power set PX of X is the set of all subsets of X. We will
investigate the theory of power sets PX considered as a structure PX; ⊆,
where ⊆ is the usual subset relation. So the elements of our structure are subsets
∞
of X, not elements of X. Let T pow = Th({PX; ⊆ | X is a set }), and let T pow =
Th({PX; ⊆ | X is an infinite set }). The usual axioms of partial orders hold for
all power sets:

Transitivity ∀xyz[(x ⊆ y ∧ y ⊆ z) → x ⊆ z];

Reflexivity ∀x[x ⊆ x]; and
Antisymmetry ∀xy[(x ⊆ y ∧ y ⊆ x) → x = y].

104
 19.2 Lattices 105

However, not every partial order arises from a power set, so we need to
consider more properties.

19.2 Lattices
There are natural operations of intersection and union which we can capture.
Given a, b ∈ PX, we usually define a ∩ b as {x ∈ X | x ∈ a ∧ x ∈ b }. However,
this is not a definition by a formula in our chosen language. Nonetheless, there
is a formula which captures the notion of the intersection. We can see that
c = a ∩ b if and only if PX |= ϕ(a, b, c), where ϕ(x1 , x2 , y) is the formula
y ⊆ x1 ∧ y ⊆ x2 ∧ ∀w[(w ⊆ x1 ∧ w ⊆ x2 ) → w ⊆ y],
which says that y is the greatest lower bound of x1 and x2 . So in a power set,
every pair of elements has a greatest lower bound. Similarly, it has a least upper
bound, the union of the two sets. A power set PX also has a least element, ∅,
and a greatest element, X. So we can add the following axioms to the power
set axioms:
GLB ∀x1 x2 ∃y[y ⊆ x1 ∧ y ⊆ x2 ∧ ∀w[(w ⊆ x1 ∧ w ⊆ x2 ) → w ⊆ y]];
LUB ∀x1 x2 ∃y[x1 ⊆ y ∧ x2 ⊆ y ∧ ∀w[(x1 ⊆ w ∧ x2 ⊆ w) → y ⊆ w]];
Least element ∃y∀w[y ⊆ w]; and
Greatest element ∃y∀w[w ⊆ y].
Any partially ordered set satisfying these axioms is called a lattice.
It is convenient to introduce new constant symbols 0 for the least element
and 1 for the greatest element. The formula ϕ(x1 , x2 , y) above defines the
function y = x1 ∩ x2 , so we can use the binary function symbol ∩ as an
abbreviation, without changing which sets are definable. Similarly, we can use
∪ as an abbreviation. Such an expansion of the language, naming a relation,
function, or constant which is already definable, is called an expansion by
definitions.
We could then omit ⊆ from the language, since it is definable from ∩ or
from ∪ (see Exercise 19.3).
We note that ∩ and ∪ are commutative and associative in any lattice, and in
a power set, they also satisfy the following distributive laws:
Distributivity 1 ∀xyz[x ∩ (y ∪ z) = (x ∩ y) ∪ (x ∩ z)]; and
Distributivity 2 ∀xyz[x ∪ (y ∩ z) = (x ∪ y) ∩ (x ∪ z)].
Not all lattices satisfy these laws, but those which do are called distributive
lattices.
106 19 Power Sets and Boolean Algebras

19.3 Boolean Algebras

Within PX, we can also take the complement ac = X  a of any set a. It can be
characterized uniquely by the property that (a ∪ ac = 1) ∧ (a ∩ ac = 0). Not
every distributive lattice has complements, so we add the following axiom to
our list:

Complementation ∀x∃y[x ∩ y = 0 ∧ x ∪ y = 1].

Again, it is convenient to use the function symbol c as an abbreviation.
Definition 19.1 A Boolean algebra is a distributive lattice with complements.
The notion of Boolean algebra is important, but by no means is every
Boolean algebra isomorphic to, or even elementarily equivalent to, a power
set algebra. For linear orders, there was a natural division into discrete and
dense linear orders (although some linear orders are neither). A similar division
occurs for Boolean algebras.

19.4 Atoms
A fundamental property of sets is that they are determined by their elements.
In set theory this is known as the axiom of extension. An element x ∈ X is not
available directly to us, but the singleton set {x} is an element of PX. Singleton
sets are non-empty, but have no non-empty subsets.
Definition 19.2 An element a of a Boolean algebra B is called an atom if
B |= At(a), where At(x) is the formula
x  0 ∧ ∀w[w  x → w = 0].
So we can include a version of the axiom of extension in our axioms. In
fact, given the other axioms of a Boolean algebra, it is equivalent to something
simpler.
Lemma 19.3 In any Boolean algebra the following two conditions are
equivalent:

Extension property ∀xyz[(At(z) → (z ⊆ x ↔ z ⊆ y)) → x = y]; and

Atomic ∀x[x  0 → ∃y[At(y) ∧ y ⊆ x]].
Proof Suppose B satisfies the atomic property, suppose that b, d ∈ B have
the same set of atoms below them, and let a be an atom of B. If a ⊆ b, then
 19.4 Atoms 107

a ⊆ b ∩ dc . But if a ⊆ b, then a ⊆ d, so a ⊆ dc , so a ⊆ b ∩ dc . So b ∩ dc
has no atoms beneath it. By the atomic property, b ∩ dc = 0. Similarly, we can
show that every atom is beneath b ∪ dc . So no atoms are beneath (b ∪ dc )c , so
(b∪dc )c = 0, so b∪dc = 1. So dc satisfies the properties of being a complement
of b, but complements are unique, so dc = bc , and taking complements of both
sides, it follows that d = b. So the extension property holds. The converse
direction is immediate. 
Definition 19.4 A Boolean algebra satisfying the equivalent conditions of
Lemma 19.3 is called an atomic Boolean algebra.
Fact 19.5 The axioms for atomic Boolean algebras axiomatise the theory
T pow , and adding axioms saying the structure is infinite gives a complete
∞
axiomatisation of T pow .
∞
The theory T pow is not categorical in any infinite cardinal, so we cannot use
the Łos–Vaught test to prove completeness. The back-and-forth method can be
used as mentioned in Section 18.3. See [Poi00, Section 6.3] for details.
∞
Not every model of T pow is isomorphic to a power set algebra, so the class
of power set algebras is not an axiomatisable class.

Example 19.6 Let B = {a ⊆ N | a is finite or N  a is finite }, and let B =

B, ⊆. Then B is closed in PN under the operations ∩, ∪, and c , so (using
Exercise 19.6, or just by a direct proof) it is a Boolean algebra. Also, if a ∈ B
with a  0, then a is a nonempty set, so has an element, say, s. Then {s} ⊆ a.
So B is atomic. However, it is countably infinite, and no power set can be
countably infinite.

Remarks 19.7 (i) We have considered power sets and Boolean algebras in
the language L⊆ = ⊆, and we introduced the constants 0 and 1 and the
function symbols ∩, ∪, and c as abbreviations. However, we could also
use the language LBool = ∩, ∪, c , 0, 1. In this case the axioms for
Boolean algebras will be different. Chapter 2 of Cori and Lascar [CL00]
gives a clear account of this and many other topics on Boolean algebras.
(ii) We have used the symbols ⊆, ∩, and ∪ for the partial order, greatest lower
bound, and least upper bound in a lattice or Boolean algebra, respectively,
even when it is not a power set algebra. It is more conventional to use ,
∧, and ∨, but I have avoided these symbols because the latter two conflict
with our symbols for AND and OR, which sometimes appear in the same
formulas.
108 19 Power Sets and Boolean Algebras

Exercises
19.1 Show that any linearly ordered set with endpoints is a distributive lattice,
but that if it has at least three elements, it is not a Boolean algebra.
19.2 Give examples of a partially ordered set which is not a lattice and a lattice
which is not distributive.
19.3 Show that in a lattice, the partial order ⊆ can be defined just using ∩ or
using ∪.
19.4 A Boolean algebra B is complete if every subset of B (not just the
finite subsets) has a greatest lower bound. Show that a complete atomic
Boolean algebra B is isomorphic to the power set algebra P(At(B)).
19.5 Suppose that B is a finite Boolean algebra. Show that B is complete and
atomic and that if n is the number of atoms of B, then |B| = 2n .
19.6 Show that in the language ∩, ∪, c , 0, 1, the theory of Boolean algebras
can be axiomatised by ∀-sentences. Use Proposition 5.7 to deduce that if
B is a subset of PX containing ∅ and X which is closed under ∩, ∪, and
c
, then B is a Boolean subalgebra of PX.
19.7 A Boolean ring is a ring R; +, ·, −, 0, 1 which satisfies ∀x[x · x = x]. In
a Boolean ring, define x ∩ y = x · y, define x ∪ y = x + y + x · y, and define
xc = 1 + x. Show that R; ∩, ∪, c , 0, 1 is a Boolean algebra. Conversely,
if B is any Boolean algebra, show how to define ·, +, and − to make B
into a Boolean ring.
19.8 An ideal of a Boolean algebra B is a subset I ⊆ B such that 0 ∈ I; if
y ∈ I and x ⊆ y, then x ∈ I; and if x ∈ I and y ∈ I, then x ∪ y ∈ I.
(a) Show that an ideal of a Boolean algebra B is the same thing as an
ideal of B considered as a Boolean ring.
(b) Show that the quotient B/I is also a Boolean ring.
19.9 A Boolean algebra is atomless if it satisfies ¬∃x At(x) and 0  1.
(a) Show that any atomless Boolean algebra is infinite.
(b) Let B be an infinite Boolean algebra, and let Fin(B) be the ideal
generated by At(B). Show that B/Fin(B) is an atomless Boolean
algebra.
(c) Use the back-and-forth method to show that the theory of atomless
Boolean algebras is countably categorical and has quantifier
elimination in the language ∩, ∪, c , 0, 1.
 20
The Algebras of Definable Sets

In this chapter we will look more systematically at the definable sets in a model
A of a theory T . The definable subsets of An form a Boolean algebra, and we
study this in the examples of dense linear orders and vector spaces. We show
that the atoms of the Boolean algebras are related to certain formulas, called
principal formulas.

20.1 Lindenbaum Algebras

Definition 20.1 Let A be an L-structure. As usual, A denotes the domain of
A. We write Def n (A) for the set of all definable subsets of An .
Recall that An itself and ∅ are definable and that Def n (A) is closed under
the Boolean operations ∩ , ∪, and complement. So using Exercise 19.6, it can
be considered as a Boolean algebra.
Now we start from the syntactic direction: formulas and theories. Given a
language L, we have written Form(L) for the set of all L-formulas. Now define
Formn (L) to be the set of all L-formulas with free variables from x1 , . . . , xn
only. Let T be an L-theory. We define a relation ∼T on Formn (L) by
ϕ( x̄) ∼T θ( x̄) if and only if T  ∀ x̄[ϕ( x̄) ↔ θ( x̄)].
It is easy to see that ∼T is an equivalence relation on Formn (L). We write
[ϕ( x̄)] for the equivalence class of ϕ( x̄), and we write Lindn (T ) for the set
of equivalence classes.
Lemma 20.2 The logical operations ∧ , ∨ , and ¬ on formulas induce the
structure of a Boolean algebra on Lindn (T ). That is, if we define [ϕ( x̄)] ∧ [θ( x̄)]
to be [(ϕ ∧ θ)( x̄)], and if we define ∨ and ¬ similarly, we get well-defined
operations on Lindn (T ) which make it into a Boolean algebra.

109
110 20 The Algebras of Definable Sets

We leave the proof as an exercise. The algebras Lindn (T ) are called the
Lindenbaum algebras of T .
Lemma 20.3 Let A be an L-structure and T = Th(A). There is an
isomorphism of Boolean algebras between Lindn (T ) and Def n (A) given by

[ϕ( x̄)] → ϕ(A).

Again, we leave the proof as an exercise. An important consequence of
this lemma is that if A and B are models of the same complete theory T , the
Boolean algebras Def n (A) and Def n (B) are isomorphic to each other, because
they are both isomorphic to Lindn (T ), so as abstract Boolean algebras, they do
not depend on the choice of model.

20.2 Example: Dense Linear Orders

We take T = DLO and consider the model Q< . By Theorem 17.8, DLO has
quantifier elimination, so we only need to consider the quantifier-free definable
sets. More specifically, Theorem 17.4 shows that every definable set is defined
by a finite disjunction of principal DLO formulas.
For example, for n = 1, the only principal DLO formula is x1 = x1 , so the
definable subsets of Q are Q itself, defined by x1 = x1 , and ∅, defined by the
empty disjunction, which is equivalent to the negation x1  x1 . For n = 2, the
principal DLO formulas are x1 = x2 , x2 = x1 , x1 < x2 , and x2 < x1 . The first
two define the same subset of Q2 , so there are three different sets defined by
principal DLO formulas. Then there are 23 = 8 different definable subsets of
Q2 corresponding to the different possible disjunctions of these sets. In general,
if the principal DLO formulas for a given n define m different subsets of Qn ,
then the Boolean algebra Def n (Q< ) has size 2m and is isomorphic to the power
set algebra on a finite set of size m.

20.3 Principal Formulas

There are analogues of these principal DLO formulas in other theories.

Definition 20.4 Let T be a complete L-theory. A principal formula (with

respect to T ) is a formula ψ( x̄) such that
(i) T  ∃ x̄[ψ( x̄)], (we say ψ( x̄) is satisfiable), and
(ii) for every L-formula ϕ( x̄) with the same list of variables, either
 20.4 Vector Spaces over Finite Fields 111

T  ∀ x̄[ψ( x̄) → ϕ( x̄)] or T  ∀ x̄[ψ( x̄) → ¬ϕ( x̄)].

So principal DLO formulas are principal formulas for the theory DLO,
and every principal formula for DLO is ∼DLO -equivalent to a principal DLO
formula.
For any L-structure A, a formula ϕ( x̄) ∈ Formn (L) is a principal formula for
Th(A) if and only if ϕ(A) is an atom of Def n (A). So to understand Def n (A), a
good start is to try to understand all the principal formulas. When there are only
finitely many, and every n-tuple from A satisfies one of them, this is enough to
determine all the definable sets. The following proposition captures the method
we used for DLO.
Proposition 20.5 Suppose ψ1 ( x̄), . . . , ψm ( x̄) are all principal formulas in n
variables for the complete theory T = Th(A), defining different subsets of An ,

and m i=1 ψi (A) = A . Then every definable set in Def n (A) is a union of some
n

of the sets ψi (A), and |Def n (A)| = 2m . 

If there are only finitely many definable subsets of An , then they must follow
this pattern.
Lemma 20.6 If Def n (A) is finite, then every ā ∈ An satisfies a principal
formula.
Proof Given ā ∈ An , if Def n (A) is finite, then there is a minimal definable
subset containing ā, and any formula defining it is a principal formula. 
The converse of this lemma is false. For example, in Def 1 (Ns-ring ), every
m ∈ N satisfies a principal formula x1 = 0 or x1 = 
1 + · · · + 1, but Def 1 (N) is
infinite. m

20.4 Vector Spaces over Finite Fields

Recall the theory T F∞3 -VS of infinite vector spaces over the field F3 with three
elements: 0, 1, and 2. Let V |= T F∞3 -VS . From Theorem 18.3 we know that T F∞3 -VS
has quantifier elimination. So every definable set is defined by a Boolean
combination of atomic formulas. The atomic formulas in variables x1 , . . . , xn

are all equivalent under the theory T F∞3 -VS to formulas of the form ni=1 λi xi = 0
for scalars λi ∈ F3 .
For n = 1, this gives us atomic formulas 0 = 0, x1 = 0, and 2x1 = 0. The
latter two both define {0}, which is a singleton, so x1 = 0 is a principal formula
(see Exercise 20.6). The atomic formula 0 = 0 defines all of V and is not a
112 20 The Algebras of Definable Sets

principal formula. However, the negation x1  0 is principal, which we can see

because no atomic formula splits up the set it defines into proper subsets. Thus
Def 1 (V) consists of exactly the four sets ∅, {0}, V  {0}, and V.
Def 2 (V) is a bit more interesting, because for each λ ∈ F3 , we have the
formula x2 = λ · x1 , giving a linear equation. Suppose λ ∈ F3 and we have
(a1 , a2 ) ∈ V 2 and (b1 , b2 ) ∈ V 2 such that
V |= a1  0 ∧ a2 = λ · a1 and also V |= b1  0 ∧ b2 = λ · b1 .
Then there is an automorphism π of V such that π(a1 , a2 ) = (b1 , b2 ). So the
subset of V 2 defined by the formula x1  0 ∧ x2 = λ · x1 has no proper
non-empty definable subsets, and hence the formula is principal. (Here we
have used automorphisms directly, as in the proof of the quantifier elimination
theorem. We could also have used the theorem.)
We have the following five formulas in two variables:
x1 = 0 ∧ x2 = 0, x1 = 0 ∧ x2  0,

x1  0 ∧ x2 = 0, x1  0 ∧ x2 = x1 , x1  0 ∧ x2 = 2 · x1 ,
which are all principal. If (a1 , a2 ) satisfies one of these formulas, then a1 and a2
are linearly dependent, so there is also the possibility that (a1 , a2 ) is a linearly
independent pair of vectors. We can also capture this situation with a single
formula,
$
λ1 · x1 + λ2 · x2  0,
(λ1 ,λ2 )∈F23 {(0,0)}

which is first order because it is a finite conjunction. Every linearly indepen-

dent pair of elements of V satisfies the same atomic formulas, hence the same
quantifier-free formulas, hence (by quantifier elimination) the same formulas.
So this is again a principal formula.
So every pair of elements from V satisfies exactly one of these six principal
formulas. Hence Def 2 (V) has exactly 26 = 64 definable sets.

Exercises
20.1 Prove Lemma 20.2.
20.2 Prove Lemma 20.3.
20.3 Suppose that T is an incomplete theory and A |= T . Show there is a
surjective homomorphism of Boolean algebras Lindn (T ) → Def n (A)
as in Lemma 20.3, but that it is not injective.
 Exercises 113

20.4 Give two examples of principal formulas and two examples of formulas
which are satisfiable but non-principal in the theory of Zring .
20.5 How many definable sets are there in Def n (Q< ) for n = 3 and n = 4?
20.6 Suppose that ϕ(x) is a formula such that T  ∃=1 x[ϕ(x)]. Show that
ϕ(x) is a principal formula for T .
20.7 Consider the expansion A of Q< by constant symbols naming 0 and 1.
How many definable sets are there in Def n (A) for n = 1, 2?
20.8 How many definable sets are there in Def n (I< ) for n = 1, 2, where I is
the closed unit interval [0, 1] in the reals?
20.9 Let V be an infinite-dimensional F3 -vector space. Show that any pair
(a1 , a2 ) ∈ V 2 satisfies one of the six principal formulas given before
Definition 20.4.
20.10 How many definable sets are there in Def n (V) for n = 1, 2, 3, where V
is an infinite F4 -vector space?
20.11 Let A = Z; <, 0. Show that for every n ∈ N+ , Def n (A) is an atomic
Boolean algebra which is not complete.
20.12 Since Lindenbaum algebras are Boolean algebras, they satisfy the
distributive laws and de Morgan laws, which allow us to swap the
order of conjunctions, disjunctions, and negations. This allows us to
give normal forms for formulas which can be useful in understanding
definable sets. Let T be any L-theory, for example the empty L-theory.
Let ψ, ϕi , and ϕi j be any L-formulas for i = 1, . . . , r and j = 1, . . . , si .

(a) Prove the following logical equivalences:

 
(1) ψ ∧ ri=1 ϕi ∼T ri=1 (ψ ∧ ϕi ).
r 
(2) ψ ∨ i=1 ϕi ∼T ri=1 (ψ ∨ ϕi ).
r  si   
(3) i=1 j=1 ϕi j ∼T sj11 =1 · · · sjrr =1 ri=1 ϕi ji .
r  si  s1  sr r
(4) i=1 j=1 ϕi j ∼T j1 =1 · · · jr =1 i=1 ϕi ji .
 
(5) ¬ ri=1 ϕi ∼T ri=1 ¬ϕi .
 
(6) ¬ ri=1 ϕi ∼T ri=1 ¬ϕi .
(b) Conjunctive normal form theorem: suppose that ϕ is a
quantifier-free formula. Prove by induction on the construction of ϕ
that there is an L-formula θ such that θ ∼T ϕ, and θ is of the form
r  si
i=1 j=1 αi j , where each αi j is either an atomic formula or the
negation of an atomic formula.
(c) Disjunctive normal form theorem: show that every quantifier-free
formula ϕ is also logically equivalent to a formula ψ of the form
114 20 The Algebras of Definable Sets

r  si
i=1 j=1 βi j , where each βi j is either an atomic formula or the
negation of an atomic formula.
(d) Show that if ϕ is a positive quantifier-free formula, that is, it is
constructed using ∧ and ∨ , but without implications or negations,
then its conjunctive and disjunctive normal forms also do not use
negation.
 21
Real Vector Spaces and Parameters

In the previous chapter we characterised the Lindenbaum algebras of definable

sets for DLO and for vector spaces over finite fields, where these algebras
are finite. In this chapter we consider vector spaces over the real field where
the algebras are not finite. The pattern is the same for any infinite field. We
also consider the sets which become definable when new constants, called
parameters, are added to the language.

21.1 The Real Line as a Vector Space

Consider the real line R as a one-dimensional R-vector space. We write
it as RR-VS . By Theorem 18.3, RR-VS has quantifier elimination. Atomic
formulas in the variables x1 , . . . , xn are equivalent to formulas of the form

n
i=1 λi xi = 0 for scalars λi ∈ R. Conjunctions of atomic formulas therefore
define vector subspaces of Rn , and indeed all subspaces of Rn are definable.
Thus Def n (RR-VS ) consists of all the Boolean combinations of subspaces of Rn .
The only subspace of R1 is {0}, so Def 1 (RR-VS ) consists of the four sets ∅,
{0}, R  {0}, and R.
The subspaces of R2 are {(0, 0)}, R2 itself, and the one-dimensional
subspaces given by x2 = λx1 for each λ ∈ R, and by x1 = 0.

Lemma 21.1 For each n ∈ N+ , the formula ni=1 xi = 0 defining the origin
0 is a principal formula. If L is any 1-dimensional subspace of Rn , then any
formula defining L  {0} is a principal formula.

Proof It is immediate that ni=1 xi = 0 is principal since it defines a singleton
set. Suppose that L is a one-dimensional subspace of Rn and a, b ∈ L{0}. Then
there is λ ∈ R{0} such that b = λa. Multiplication by λ is an automorphism of

115
116 21 Real Vector Spaces and Parameters

RR-VS , so a, b must satisfy the same formulas. So any formula defining L  {0}
is principal. 
Proposition 21.2 For all n ∈ N+ , Def n (RR-VS ) is an atomic Boolean algebra.
It is not complete if n  2.
Proof Let S ⊆ Rn be definable and non-empty. We must show that there is
an atom of Def n (RR-VS ) contained in S . Let a = (a1 , . . . , an ) ∈ S . If a = 0, then
{0} ⊆ S . Otherwise, there is i such that ai  0. For convenience, assume a1  0.
Then for each j = 2, . . . , n, there is λ j ∈ R such that a j = λ j a1 . Let ϕ( x̄) be the

formula x1  0 ∧ nj=2 x j = λ j x1 . Then ϕ( x̄) is a principal formula by Lemma
21.1, and the set it defines is contained in S . So in either case, S contains an
atom of the Boolean algebra Def n (RR-VS ), so Def n (RR-VS ) is atomic.
If n  2, then the number of atoms in Def n (RR-VS ) is |R|, but the total
number of formulas is also |R|, so |Def n (RR-VS )| = |R|. However, a complete
atomic Boolean algebra with |R| atoms has size |PR|, which is strictly larger
than |R|. 
So as in the case of vector spaces over finite fields, principal formulas are an
important starting point in determining the definable sets. However, not every
definable set is a finite disjunction of these sets; indeed, R2 is not. So while
principal formulas are still important, they do not give the whole picture.
Now suppose that V is another model of T R∞-VS , for example V = R3 . For
each n we have Def n (V)  Lindn (T R-VS )  Def n (RR-VS ), so the definable
subsets of V n are not Boolean combinations of all the vector subspaces of
V n = R3n but only Boolean combinations of certain subspaces. This is because
the variables xi range over the vectors in V and not over the real numbers. In
the proof of Proposition 21.2, to find the scalars λ, we made essential use of
the fact that we were working in the one-dimensional model RR-VS of T R-VS .
However, the conclusion of the proposition also holds for Def n (V). This is an
example where we get a benefit from working in a particular model of the
theory.

21.2 Parameters
We have seen that definable sets in RR-VS correspond to Boolean combinations
of homogeneous linear equations, that is, equations which state that a linear
combination of the variables is equal to 0. Often in linear algebra, one wishes
to consider systems of inhomogeneous equations, that is, equations of the form

n
i=1 λi xi = a, for some a  0. We can do this by adding a constant symbol to
 21.2 Parameters 117

the language for a. In this case, constants representing elements of the structure
are called parameters.

Definition 21.3 Let M be an L-structure and A a subset of the domain M

of M. A subset S ⊆ M n is said to be definable with parameters from A,
or A-definable, if there is an L-formula ϕ(x1 , . . . , xn , y1 , . . . , ym ) and elements
a1 , . . . , am ∈ A such that

S = {(x1 , . . . , xn ) ∈ M n | M |= ϕ(x1 , . . . , xn , a1 , . . . , am ) } .

If A = M, then S is said to be definable with parameters, or parametrically

definable. If A = ∅, then S is sometimes said to be ∅-definable, null-definable,
zero-definable, or definable without parameters.

In principle, there is nothing very new here. We can expand the language
L to LA and expand the structure M to MA by adding constants naming each
element of A. Then ThLA (MA ) is a complete LA -theory, and the subsets of
M n which are A-definable with respect to the language L and the same as
the subsets which are ∅-definable with respect to the language LA . However,
in practice, it is very useful to consider M and MA not as totally different
structures, because they are much more closely related than the structures
you can get by changing the language in other ways. For example, quantifier
elimination is preserved.
Lemma 21.4 Suppose M is an L-structure with quantifier elimination and
S ⊆ M n is a subset definable with parameters. Then S is defined using the
same parameters by a quantifier-free L-formula.
Proof Suppose S = { x̄ ∈ M n | M |= ϕ( x̄, ā) }. By quantifier elimination for
M, there is a quantifier-free L-formula θ( x̄, ȳ) such that M |= ∀ x̄ȳ[θ( x̄, ȳ) ↔
ϕ( x̄, ȳ)]. So θ( x̄, ā) is a quantifier-free formula defining S with the same
parameters. 
Consider now the parametrically definable subsets of R2 with respect to
the structure RR-VS . As well as the origin (0, 0), any point (a, b) in R2 is now
definable by x1 = a ∧ x2 = b. As well as the straight lines through the
origin, formulas of the form λ1 x2 + λ2 x2 = a with λ1 , λ2 , a ∈ R now give
all the straight lines in R2 . So the parametrically definable subsets of R2 are
the Boolean combinations of points and straight lines.
118 21 Real Vector Spaces and Parameters

Exercises
21.1 Let V = R2 , considered as an R-vector space. Then both Def 2 (V) and
Def 4 (RR-VS ) are Boolean algebras of subsets of R4 . Give an example of
a subset S ⊆ R4 which is in Def 4 (RR-VS ) but not in Def 2 (V).
21.2 Let S ∈ Def 2 (RR-VS ). Show that either S or R2  S is defined by a
finite disjunction of principal formulas. Is the same true for every S ∈
Def 3 (RR-VS )?
21.3 Let A be the structure RR-VS expanded by a single constant symbol
naming 1. Find the atoms in the Lindenbaum algebra Def 2 (A).
21.4 Let M be an L-structure, suppose A ⊆ M is a set of parameters,
'
and let n ∈ N. Show that Def n (MA ) = Def n (MA0 ) | A0 is a finite
subset of A}.
21.5 Let M be an infinite set considered as an L= =structure, and let A ⊆ M.
How large is Def n (MA ) for n = 1, 2, when |A| = r, finite, and when
|A| = ℵ0 ?
21.6 Let M be an L-structure and A ⊆ M a subset. Recall that the
substructure of M generated by A is the smallest L-substructure of M
containing A. Write it as A. Show that Def n (MA ) = Def n (MA ).
21.7 The definable closure of a subset A of an L-structure M, written dcl(A),
is the set of all b ∈ M such that the singleton set {b} is definable in M,
with parameters from A. Show that A ⊆ dcl(A) and that Def n (MA ) =
Def n (Mdcl(A) ).
21.8 Suppose that S ⊆ R is parametrically definable with respect to the
structure RR-VS . Show that either S or R  S is a finite set. In the latter
case, we say that S is a cofinite subset. Conversely, show that every
finite subset and every cofinite subset of R is parametrically definable
with respect to RR-VS .
21.9 The Lindenbaum algebras of structures are not always atomic. Here
is one example. Let L be the language with unary relation symbols
Dn for n ∈ N+ , and consider the structure R with domain R and with
Dn interpreted as the set of real numbers whose nth decimal digit after
the decimal point is even. Show that R has quantifier elimination by
showing that if r, s ∈ R satisfy the same quantifier-free formulas, then
there is an automorphism of R sending r to s. Then show that Def 1 (R)
is atomless.
21.10 Recall T S = Th(Nsucc ) from Chapter 16. Using the methods of that
chapter or using Exercise 18.4, show that if M is any model of T S , then
every subset of M definable with parameters is finite or cofinite. Show
also that no linear order is definable on M, even using parameters.
 22
Semi-algebraic Sets

We consider now the subsets of Rn which are definable with parameters, with
respect to the structure Ro-ring . These subsets are also known as semi-algebraic
sets. As with many structures, the study of the definable sets combines model-
theoretic ideas with ideas from the branch of mathematics whence the structure
naturally comes, which in this case is elementary real analysis. For simplicity,
we will concentrate on the definable subsets of R and R2 , although most of the
ideas needed in higher dimensions are already present here.

22.1 O-Minimality
We will use Fact 18.5, which states that Ro-ring has quantifier elimination in the
ordered ring language. So, as in the previous chapter, we start by considering
the sets defined by atomic formulas, and then by quantifier-free formulas.
Lemma 22.1 Every atomic Lo-ring -formula, with parameters from R, defines
the same subset of Rn as a formula of the form f ( x̄) = 0 or f ( x̄) > 0, where
f ( x̄) is a polynomial with real coefficients.
The proof is left as an exercise.
Lemma 22.2 Every subset of Rn defined by a quantifier-free Lo-ring -formula,
with parameters from R, is a finite union of sets defined by formulas of the form

h( x̄) = 0 ∧ sj=1 g j ( x̄) > 0.
Proof Suppose S is defined by a quantifier-free formula ϕ( x̄). Then ϕ( x̄) is
a Boolean combination of atomic formulas. By the disjunctive normal form
theorem, Exercise 20.12, we can write ϕ( x̄) such that negation is applied only
to atomic formulas. But for any polynomial f ( x̄),

Ro-ring |= ∀ x̄[¬( f ( x̄) = 0) ↔ ( f ( x̄) > 0 ∨ − f ( x̄) > 0)]

119
120 22 Semi-algebraic Sets

and

Ro-ring |= ∀ x̄[¬( f ( x̄) > 0) ↔ ( f ( x̄) = 0 ∨ − f ( x̄) > 0)].

So S is defined by a positive Boolean combination of atomic formulas. Using

the disjunctive normal form theorem again, S is defined by a formula of the
form
⎛ t ⎞
 r ⎜$
⎜⎜⎜ i $
si ⎟⎟⎟
⎜⎜⎝ hi j ( x̄) = 0 ∧ gi j ( x̄) > 0⎟⎟⎟⎠ .
i=1 j=1 j=1

ti
Let hi ( x̄) = j=1 hi j ( x̄)2 . Then S is defined by
⎛ ⎞

r ⎜
⎜⎜⎜ $
si ⎟⎟⎟
⎜⎜⎝hi ( x̄) = 0 ∧ gi j ( x̄) > 0⎟⎟⎟⎠ ,
i=1 j=1

so it is a finite union of sets of the desired form. 

Now we restrict to the case n = 1. A formula of the form f (x) = 0 defines a
finite set unless f is a constant polynomial
√ when
√ it defines either all of R or ∅.
For example, x3 − 2x = 0 defines {− 2, 0, 2}.
We can work out what set is defined by a formula of the form f (x) > 0
by considering how the sign of √f (x) changes
√ between its zeros. For example,
x3 − 2x > 0 defines the set (− 2, 0) ∪ ( 2, ∞). In general, any finite union
of open intervals can be obtained this way. By combining these intervals with
points, we can get closed intervals and half-open intervals. Using quantifier
elimination and Lemma 22.2, we get the following.
Proposition 22.3 Every subset of R which is parametrically definable with
respect to the structure Ro-ring is a finite union of points and intervals.
Note that these subsets of R are also parametrically definable in the reduct
R< , although not necessarily with the same parameters. However, there are
definable sets in two or more free variables, such as that defined by y = x2 ,
which are not definable in R< .
The conclusion of Proposition 22.3 is a very useful condition that has been
shown to hold in many other structures and has been given a name.

Definition 22.4 An infinite structure M in a language containing < is said to

be o-minimal if M is densely linearly ordered by < and every parametrically
definable subset of M is a finite union of points and intervals.
 22.2 Continuity of Definable Functions 121

22.2 Continuity of Definable Functions

In o-minimal structures, the study of definable sets goes hand in hand with
the study of the definable functions. The parametrically definable functions
in Ro-ring are called semi-algebraic functions. Polynomial functions are semi-
algebraic, but there are also other semi-algebraic functions.

Example 22.5 The function f : R → R in Figure 22.1 is defined by the

formula

(x < −2 ∧ (y + 1)3 = x + 3) ∨ (y  0 ∧ (x + 1)2 + y2 = 1) ∨ (0 < x ∧ y2 = x).

This function is not a polynomial function, but it is at least continuous. Not

every definable function is continuous; for example the function
⎧
⎪
⎪
⎨0 if x < 0
g(x) = ⎪
⎪
⎩1 if x  0

is not continuous at 0, but it is continuous everywhere else in R. O-minimality

at once implies that the function
⎧
⎪
⎪
⎨0 if x ∈ Q
h(x) = ⎪⎪
⎩1 if x  Q

is not definable in Ro-ring . In fact, definable functions are piecewise continuous.

–4 –2 0 2

–2

Figure 22.1 A function R → R which is definable in Ro-ring but which is not a

polynomial function.
122 22 Semi-algebraic Sets

Proposition 22.6 Let (a, b) ⊆ R be an open interval, possibly with a = −∞

and / or b = +∞, and let f : (a, b) → R be a semi-algebraic function. Then
f is piecewise continuous, that is, there is m ∈ N and there are ak ∈ R with
a = a0 < a1 < · · · < am < am+1 = b such that for each k = 0, . . . , m, the
restriction f(ak ,ak+1 ) is continuous.
There is more than one proof of this proposition. We give a proof which
works for any o-minimal structure on the real line.
Lemma 22.7 Let f : (a, b) → R be a semi-algebraic function. Then there is
x ∈ (a, b) such that f is continuous at x.
Proof First suppose that some point c ∈ R has infinite preimage under f .
This preimage is defined by f (x) = c so, by o-minimality, it contains an open
interval U. Then fU is constant, so continuous.
Otherwise, we will construct a decreasing chain of intervals Un . Set U0 =
(a, b). Now assume n ∈ N+ , and we have constructed Un−1 . Since no point in
R has infinite preimage, the image of fUn−1 is infinite. This image is definable
and so, by o-minimality, contains an open interval, In . We can choose this In
to have length at most 1/n. Then the set Jn defined by x ∈ Un−1 ∧ f (x) ∈ In
is infinite and so contains an open interval. We choose Un to be some open
interval in Jn such that the closure U n is contained in Un−1 .
 
Then the intersection n∈N Un = n∈N U n , an intersection of nested closed

intervals, which is non-empty. Let c ∈ n∈N Un , and let  > 0. Take N ∈ N
such that N > 1/, and choose δ > 0 such that (c − δ, c + δ) ⊆ U N . Then, if
|x − c| < δ, we have x ∈ U N , so f (x), f (c) ∈ IN , so | f (x) − f (c)|  1/N < . So
f is continuous at c. 
Proof of Proposition 22.6 Let D ⊆ (a, b) be the set of points at which f is
discontinuous. Then D is defined by
(∃)[ > 0 ∧ (∀δ)[δ > 0 → (∃y)[|x − y| < δ ∧ | f (x) − f (y)| > ]]],
so by o-minimality, it is a finite union of points and open intervals. But by
Lemma 22.7, a semi-algebraic function cannot be discontinuous on an open
interval. So D is a finite set of points. 

22.3 Nash Functions

We now characterise the semi-algebraic functions more precisely.
Definition 22.8 A Nash function is an analytic function f : U → R where
U ⊆ Rn is a connected open semi-algebraic subset such that there is a non-zero
polynomial h( x̄, y) ∈ R[ x̄, y] such that for all x̄ ∈ U, we have h( x̄, f ( x̄)) = 0.
 22.3 Nash Functions 123

To say that f is analytic means that it is infinitely differentiable and,

furthermore, that for every point a ∈ U, the Taylor series of f at a converges
to f in a neighbourhood of a. A polynomial function f is analytic and is a
Nash function, taking h( x̄, y) to be f ( x̄) − y. The following consequence of the
implicit function theorem for analytic functions gives an easy way to check
analyticity.
Fact 22.9 Suppose that U ⊆ Rn is a connected open semi-algebraic subset,
that f : U → R is continuous, and that there is a polynomial h( x̄, y) ∈ R[ x̄, y]
such that for all x̄ ∈ U, we have h( x̄, f ( x̄)) = 0 and ∂h
∂y ( x̄, f ( x̄))  0. Then f is
a Nash function.
Using this fact, the function f from Example 22.5 can be split into the three
points (−3, −1), (−2, 0), and (0, 0) and four Nash functions:
⎧ √3
⎪
⎪
⎪ x+3−1 for x ∈ (−∞, −3) and for x ∈ (−3, −2),
⎪
⎨ (
⎪
y=⎪ ⎪ − 1 − (x + 1) 2 for x ∈ (−2, 0),
⎪
⎪
⎩ √x
⎪
for x ∈ (0, ∞).

The functions h(x, y) are then those appearing in the formula in Example
22.5. Note that f is not differentiable at x = −3, so is not analytic, and this
corresponds to the function h(x, y) = (y + 1)3 − x − 3 having ∂h
∂y (−3, −1) = 0.

Proposition 22.10 Let (a, b) ⊆ R be an open interval, possibly with a = −∞

and / or b = +∞, and let f : (a, b) → R be a semi-algebraic function. Then f
is a piecewise Nash function. That is, there is m ∈ N and a = a0 < a1 < · · · <
am < am+1 = b such that, for each k = 0, . . . , m, f(ak ,ak+1 ) is a Nash function.
Proof By quantifier
 elimination for Ro-ring (Fact 18.5) and Lemma 22.2, the
graph Γ f = (x, y) ∈ R2 | y = f (x) of f is defined by a formula of the form
⎛ ⎞
r ⎜
⎜⎜⎜ $
si ⎟⎟⎟
⎜⎝⎜hi (x, y) = 0 ∧ gi j (x, y) > 0⎟⎟⎠⎟ .
i=1 j=1

The functions hi cannot be the zero polynomial, because otherwise Γ f would

contain an open subset of R2 , which is impossible for the graph of a function.
Let H be the set consisting of the polynomials hi (x, y) and their iterated
y-derivatives ∂∂yhti (x, y) for all t ∈ N. Since the hi are polynomials, H is a
t

)finite set. For each h ∈* H, let Zh = {x ∈ R | h(x, f (x)) = 0 } and let Dh =

x ∈ Zh  ∂h (x, f (x))  0 . These sets are definable, so by o-minimality, they
∂y
are finite unions of points and open intervals. List all of the isolated points and
the endpoints of the intervals for all of the Zh for h ∈ H as a = a0 < a1 < · · · <
124 22 Semi-algebraic Sets

am < am+1 = b. Now we claim that each interval Ik := (ak , ak+1 ) is contained in
one of the Dh . By definition, if c, c ∈ Ik and h ∈ H, then either both or neither
of c and c lie in Zh . From the formula defining f , for some i = 1, . . . , r, we
have Ik ⊆ Zhi . If Ik ⊆ Dhi , then we replace hi by ∂h
∂y and repeat if necessary. If
i

∂d hi
d is the degree of hi (x, y) as a polynomial in y, then ∂yd
is constant in y and
∂d hi
non-zero, so ∂yd
(x, f (x)) is a non-zero polynomial in x which cannot vanish on
the interval Ik . So at some stage we stop and find an h ∈ H such that Ik ⊆ Dh .
Thus we have shown that f is a piecewise Nash function, as required. 
It is also true, and fairly easy to prove, that every piecewise Nash function
f : (a, b) → R is a semi-algebraic function, provided we use the definition
given above that piecewise means split into only finitely many pieces.

22.4 Cells
The definable subsets of R2 can be built from Nash functions as follows.

Definition 22.11 A 0-cell in R2 is a single point.

A (Nash) 1-cell is a subset of R2 defined by a formula of the form

x=c ∧ a<y ∧ y<b

or
a < x ∧ x < b ∧ y = f (x),

where c ∈ R, a < b with a ∈ {−∞} ∪ R, and b ∈ R ∪ {+∞}, and f is a Nash

function on the interval (a, b). Conditions of the form −∞ < x or y < +∞ are
automatically true and are not technically part of the Lring -formula defining the
cell. They are included above to streamline the definition of cells.
A (Nash) 2-cell is a subset of R2 defined by a formula of the form

a < x ∧ x < b ∧ f (x) < y ∧ y < g(x),

where a < b are as above and f and g are Nash functions on (a, b) such that, for
all x ∈ (a, b), we have f (x) < g(x), except that f may be the constant function
with value −∞ and g may be the constant function with value +∞.

Examples of cells are given in Figure 22.2.

It is clear that Nash cells, and finite unions of them, are semi-algebraic.
Using Proposition 22.10, one can prove the following.
Proposition 22.12 Every semi-algebraic subset of R2 is a finite union of Nash
cells.
 22.4 Cells 125

A 2-cell A 2-cell which is unbounded

above to the right

0-cells

1-cells 0 5

A 1-cell unbounded to the right

–5

Figure 22.2 Some examples of cells in R2 .

There are notions of Nash m-cells in Rn for all m and n ∈ N+ , and the
results analogous to Propositions 22.10 and 22.12 can be proved together by
an induction on n.
If we replace Nash functions by arbitrary continuous definable functions
in Definition 22.11, we get a more general notion of cells (continuous cells).
Then the graph of the function f in Example 22.5 is a single 1-cell, whereas it
is the union of seven Nash cells.
Remarkably, a generalisation of Proposition 22.12 to Mn for arbitrary n ∈
+
N , called the cell decomposition theorem, is true in any o-minimal structure
M when we consider continuous cells.
The study of o-minimal structures is an important branch of model theory.
The book [vdD98] by van den Dries is an excellent introduction to the
subject, and Chapter 2 treats the case of semi-algebraic sets in detail for
the continuous cells. Semi-algebraic sets are studied in detail from a more
geometric viewpoint in [BCR98]. More information about the model theory of
Ro-ring can also be found in Section 3.3 of Marker’s book [Mar02] and in the
first chapter of [MMP06].
126 22 Semi-algebraic Sets

Exercises
22.1 Prove Lemma 22.1.
22.2 Show that the singleton set {a} ⊆ R is definable in Ro-ring without
parameters if and only if a is a real algebraic number, that is, it is the
zero of a non-constant polynomial with integer coefficients.
22.3 Sketch a proof of Proposition 22.3, including any results which are
needed for the proof.
22.4 Let S ⊆ R2 be defined by y + xy = y2 + x ∧ 2y  1 + x. Show that S is
the graph of a function f : R → R, and give a decomposition of f as a
piecewise Nash function.
22.5 Show that R2 can be partitioned into 21 Nash cells such that the graph
of the function f from example 22.5 is the union of 7 of those cells.
  
22.6 Show how the closed unit circle (x, y) ∈ R2  x2 + y2  1 can be
written as a union of cells. How many cells are needed?
22.7 The definition of a cell in R2 depends on the order of the variables x
and y. Give examples of cells C1 , C2 ⊆ R2 such that if we exchange the
roles of x and y, then C1 is still a cell, but C2 is not.
22.8 Show that the proof of Proposition 22.3 generalises to prove that any
real-closed field is o-minimal.
22.9 Use the completeness of RCF to show that the conclusions of Lemma
22.7 and Proposition 22.6 hold for any real-closed field.
22.10 Find a proof of Proposition 22.12.
22.11 [Uniform finiteness for semi-algebraic sets] Let ϕ( x̄, y) be an Lo-ring -
formula. For each ā ∈ Rn , write Dā for the subset of R defined by
ϕ(ā, y). Show that Dā is infinite if and only if it contains an open
interval. Deduce that the set of ā such that Dā is infinite is definable.
Further deduce that there is N ∈ N such that for any ā ∈ Rn , either Dā
is infinite or |Dā |  N.
22.12 Consider Ro-R-VS , the real line as an ordered R-vector space, in
the language LR-VS ∪ {<}. Characterise the subsets of R2 which are
parametrically definable by quantifier-free formulas in this structure.
[This structure actually has quantifier elimination, and the definable
sets are known as semi-linear sets. See [vdD98, pp25–28] for more
details.]
  
22.13 Prove that (x, y) ∈ R2  y = x2 is not definable in R , nor even in
<
Ro-R-VS .
 Part V

Types

Types are a central concept in model theory. In this part of the book we start
by explaining how types generalise elements or tuples from a structure: they
are like potential elements which exist in elementary extensions. We then
explain how types also generalise definable sets. A key tool is the omitting
types theorem, which is proved by extending the method of Henkin that
we previously used to prove the compactness theorem. The first application
of types is the Ryll–Nardzewski theorem, which characterises the countably
categorical structures as those structures A such that An has only finitely
many definable subsets for each n ∈ N. The ideas from the proof of that
theorem naturally lead to an analysis of the countable models of other theories,
including the notions of prime, atomic, universal, and ℵ0 -saturated models.
Finally, we extend the idea of saturation to uncountable models and give a
method for proving the completeness of a theory which, unlike the Łos-Vaught
test, does not require the theory to be categorical in any cardinal. Here we need
more set theory than elsewhere in the book, and some ideas are only sketched.

127
 23
Realising Types

In Chapter 20 we saw that if (a1 , a2 ) is a pair of elements from a Q-vector space

that are linearly dependent, then they satisfy a principal formula, but if they are
linearly independent, then they do not. To analyse these tuples which do not
satisfy a principal formula, we need to consider infinitely many formulas at
once, which brings us to the notion of types.

23.1 Types
Recall that Formn (L) is the set of all L-formulas with free variables from
x1 , . . . , xn only.

Definition 23.1 Let T be a complete L-theory. An n-type of T is a subset

p( x̄) of Formn (L) such that p( x̄) is finitely satisfiable, that is, for any formulas

ϕ1 ( x̄), . . . , ϕr ( x̄) from p( x̄), T  ∃ x̄ ri=1 ϕi ( x̄) . The type p( x̄) is a complete
type if, for all ϕ( x̄) ∈ Formn (L), either ϕ( x̄) ∈ p( x̄) or ¬ϕ( x̄) ∈ p( x̄).

For example, we can consider the 2-type in the theory of infinite Q-vector
spaces consisting of all the formulas λ1 · v1 + λ2 · v2  0 such that (λ1 , λ2 ) ∈
Q2  {(0, 0)}. This is the type of a linearly independent pair of vectors.

Definition 23.2 Let T be an L-theory, let M be a model of T with domain

M, and let ā = (a1 , . . . , an ) ∈ M n . The type of ā in M is

tpM (ā) = {ϕ( x̄) ∈ Formn (L) | M |= ϕ(ā) } .

We should check that the type of a tuple really is a type, according to our
definition.
Lemma 23.3 Let M |= T and ā ∈ M n . Then tpM (ā) is a complete n-type of T .

129
130 23 Realising Types

Proof The set of formulas tpM (ā) is finitely satisfiable, because it is satisfied
by ā. For any ϕ( x̄) ∈ Formn (L), either M |= ϕ(ā) or M |= ¬ϕ(ā), so tpM (ā) is
complete. 
The type of a tuple is preserved under elementary extensions and auto-
morphisms.
Proposition 23.4 (i) Let π : M → N be an elementary embedding, and let
ā ∈ M n . Then tpM (ā) = tpN (π(ā)).
(ii) If π ∈ Aut(M) and ā ∈ M n , then tpM (ā) = tpM (π(ā)).
Proof For (i), we have ϕ( x̄) ∈ tpM (ā) iff M |= ϕ(ā) iff N |= ϕ(π(ā)) by
definition of an elementary embedding, and this is iff ϕ( x̄) ∈ tpN (π(ā)). Now
note that (ii) is a special case of (i). 
Usually we are only considering one model, or elementary extensions or
elementary submodels of a given model, so then, by the proposition, the type
of tuple does not depend on the particular model M. So usually we write just
tp(ā) without the subscript M.

23.2 Realising Types

Definition 23.5 Let T be a complete theory, M a model of T , and p( x̄) an
n-type of T . We say that the type p is realised in M iff there is a tuple ā ∈ M n
such that, for all ϕ( x̄) ∈ p( x̄), M |= ϕ(ā). In this case we write M |= p(ā). If p
is not realised in M, we say that it is omitted from M.

For example, the type of a Q-linearly independent pair of vectors is realised

in a Q-vector space of dimension at least 2 but omitted in a model of dimension
1. Using the compactness theorem and the method of new constants, we can
show that every type of T can be realised in some model of T .
Lemma 23.6 Suppose that p is an n-type of T , a complete L-theory. Then there
is a model M of T of cardinality at most |L| and ā ∈ M n such that M |= p(ā).
Proof Expand L to L by adding new constant symbols c1 , . . . , cn . Consider
the set of sentences Σ = T ∪ {ϕ(c̄) | ϕ( x̄) ∈ p( x̄) }. Then Σ is finitely satisfiable
since p( x̄) is a type, so by the strong version of the compactness theorem,
Theorem 11.7, it has a model M of cardinality at most |L |, which is equal to

|L|. Take ā = c̄M , and let M be the reduct to L. Then M |= p(ā). 
In fact, we can realise as many types as we like in the same model.
 23.3 Some Types in Th(Ns-ring ) 131

Proposition 23.7 Let P be any set of types of a complete theory T , and let
M |= T . Then there is an elementary extension of M which realises all the
types in P.
We leave the proof as an exercise for the reader.

23.3 Some Types in Th(Ns-ring )

We will investigate some 1-types in the structure Ns-ring . Firstly, for each n ∈ N,
we can consider tp(n). If n  m, then tp(n)  tp(m), because the formula ψn (x)
given by x = 0 
+ 1 + · · · + 1 is in tp(n) but not in tp(m). These are all the types
n
which are realised in Ns-ring itself, but there are many more types which are
not. Let pns (x) = {¬ψn (x) | n ∈ N }. Then pns is finitely satisfiable and so is
a type, so it is realised in some model of Th(Ns-ring ) other than the standard
model Ns-ring . Any such model N is called a non-standard model of arithmetic,
and an element of N which realises pns is called a non-standard natural
number.
Let θ(x) be the formula ∃y[y + y = x] which states that x is even. Then
pns (x) ∪ {θ(x)} and pns (x) ∪ {¬θ(x)} are each finitely satisfiable, so the type of a
non-standard natural number splits into the types of even and odd non-standard
numbers. In particular, pns is not a complete type. (The more meaningful
observation here is that the deductive closure of pns is not a complete type.
However, it is useful to identify types with their deductive closures, just as we
often identify a list of axioms with the theory it axiomatises.)
In fact, if S ⊆ N is any infinite definable subset, say, defined by the formula
ϕ(x), then p(x) ∪ {ϕ(x)} is finitely satisfiable and hence a type. Thus we can
talk about ‘non-standard members of S ’.
The formula ρns (x) given by ∀yz[x = y · z → (x = y ∨ x = z)] ∧ x 
0 ∧ x  1 defines the set of all prime numbers in N. Since there are infinitely
many primes in N, there are also non-standard primes. By definition, a non-
standard prime is not divisible by any standard natural number. The opposite
behaviour is also possible.
Lemma 23.8 There is a 1-type of Th(Ns-ring ) of a non-standard natural number
which is divisible by every standard prime number.
Proof The formula δ(x, y) given by ∃z[y · z = x] states that y divides x. 
Take p(x) to be the set of formulas δ(x, P)  P is a standard prime number .
If δ(x, P1 ), . . . , δ(x, Pr ) are finitely many formulas from p(x), then the number
+
N = ri=1 Pi satisfies them all. So p(x) is finitely satisfiable, hence it is a
type. 
132 23 Realising Types

Exercises
23.1 Suppose A is a substructure of B. Show that A  B if and only if, for
every finite tuple ā from A, tpA (ā) = tpB (ā).
23.2 Prove Proposition 23.7.
23.3 Show that there is a 1-type of Th(Ns-ring ) of a number whose only prime
factor is 5 but which is divisible by 5n for all n ∈ N. Show there is another
1-type of Th(Ns-ring ) of a non-standard number which is divisible by 3 but
not by 9 or by any other standard natural number greater than 3.
23.4 Suppose ā, b̄ ∈ Qn with a1 < a2 < · · · < an and b1 < b2 < · · · < bn . By
considering automorphisms of Q< , show that tpQ< (ā) = tpQ< (b̄). Deduce
that there are only finitely many complete n-types in DLO, for each n ∈
N+ .
∞
23.5 Show there are exactly two complete 1-types in T K-VS for any field K.
23.6 Let p( x̄) be an n-type of a complete L-theory T . Let L = L ∪ {c1 , . . . , cn },
with the ci being new constant symbols, and let T  be the deductive
closure of {ϕ(c̄) | ϕ( x̄) ∈ p( x̄) }. Show that T  is an L -theory and that p is
a complete type if and only if T  is a complete L -theory.
23.7 Recall that RCF = Th(Ro-ring ), and let p+∞ (x) = {x > n | n ∈ N }. Show
that p+∞ is a 1-type of RCF. For every polynomial f (x) ∈ Z[x], show
that exactly one of the formulas f (x) = 0, f (x) < 0, or f (x) > 0 is in (the
deductive closure of) p+∞ . Using quantifier elimination for RCF, deduce
that p+∞ is a complete type of RCF.
23.8 Recall from Exercise 18.4 that T S = Th(Nsucc ) has quantifier elimination.
Use this fact to determine all of the complete n-types in T S for n ∈ N+ .
 24
Omitting Types

In the previous chapter we saw that any type of a theory T can be realised in
some model of T . In this chapter we consider which types are realised in all
models of T and which types can be omitted. First we explain how types can
be considered as a generalisation of definable sets.

24.1 Types as Intersections of Definable Sets

Given a complete theory T and a model M |= T , a formula ϕ( x̄) defines the
subset ϕ(M) = {ā ∈ M n | M |= ϕ(ā) }. Likewise, given an n-type p( x̄) of T , we
can write p(M) = {ā ∈ M n | for all ϕ( x̄) ∈ p( x̄), M |= ϕ(ā) }. So p(M) is the

intersection {ϕ(M) | ϕ( x̄) ∈ p( x̄) }.
For example, if V is a Q-vector space, we can consider the definable sets
 
S λ1 ,λ2 = (v1 , v2 ) ∈ V 2 | λ1 · v1 + λ2 · v2  0

for each (λ1 , λ2 ) ∈ Q2  {(0, 0)}. Then the intersection of the sets S λ1 ,λ2 is the
set of linearly independent pairs of vectors in the model V.
More generally, we can consider the intersection of any set {ϕi ( x̄) | i ∈ I }
of definable subsets of M n . The condition that the ϕi ( x̄) are finitely satisfiable,
and so are actually a type, corresponds to the condition that any intersection of
finitely many of the sets ϕi (M) is non-empty.
The intersection p(M) may be empty, and this corresponds to the type p
being omitted by the model M. For example, in the case  of vector spaces

above, the intersection S λ1 ,λ2  (λ1 , λ2 ) ∈ Q2  {(0, 0)} is empty if V is a
one-dimensional Q-vector space but non-empty if dim V  2.
So we cannot necessary refer to the set of realisations in a particular model
as the type. However, by Proposition 23.7, there is a model M of T which

133
134 24 Omitting Types

realises all the n-types of T , for all n ∈ N+ . In this case, if p( x̄) and q( x̄) are
different n-types, then p(M) and q(M) are different subsets of Mn . If M is
such a model, then we refer to sets of the form p(M) as type-definable subsets
of M n .

Remarks 24.1 (i) Type-definable sets are a generalisation of definable

sets, and they show how we can think of types as a generalisation of
formulas. On the other hand, complete types are like elements (or tuples
of elements) from a model, or potential elements, and so types are
somehow a common abstraction of elements and of definable sets. This
idea is important in the branch of model theory known as stability theory.
(ii) Sometimes what we have called types are known as partial types, and the
word type is used to mean complete type.

24.2 Omitting Types

We now consider which types can be omitted. Roughly speaking, if a type
can be given by a single formula, then it must be realised in any model of T .
Otherwise, it can be omitted in some models. We can make this more precise.

Definition 24.2 An n-type p( x̄) of a complete theory T is a principal type

if there is a single formula ψ( x̄) such that T  ∃ x̄ψ( x̄) and, for every formula
ϕ( x̄) ∈ p( x̄), we have T  ∀ x̄[ψ( x̄) → ϕ( x̄)]. Any type which is not principal is
called a non-principal type.

It is easy to see that any principal type of T is realised in any model of T .

What is surprising is that the converse is also true (at least when the language
is countable).
Theorem 24.3 (Omitting types theorem) Let T be a complete theory in a
countable language L, and let p be a non-principal n-type of T . Then there
is a countable model of T which omits p.
We will adapt the proof of Proposition 11.4 which shows that any finitely
satisfiable set of L-sentences can be extended to a Henkin theory T ∗ in a
language with extra constant symbols. By Proposition 11.6, every Henkin
theory has a canonical model, that is, a model in which every element is named
by a closed term. We will construct our Henkin theory T ∗ in such a way that
every element of the canonical model M∗ is named by a new constant symbol,
and for every n-tuple c̄ of these constant symbols, there is a formula ϕ( x̄) ∈ p( x̄)
 24.2 Omitting Types 135

such that T ∗  ¬ϕ(c̄). It follows that no n-tuple from M∗ realises p, so the

reduct M of M∗ to L is a model of T which omits p.
Proof of Theorem 24.3 Let L∗ be L expanded by a countably infinite set C
of new constant symbols. Enumerate all the n-tuples of these new constant
symbols as (c̄r )r∈N , and enumerate all the L∗ -sentences as (σr )r∈N . We will
construct L∗ -sentences θm for m ∈ N such that for each m ∈ N we have θm+1 
θm , T ∪ {θm } is satisfiable and:

(i) if m = 3k + 1, either θm  σk or θm  ¬σk ;

(ii) if m = 3k + 2, θm−1  σk and σk is of the form ∃x[ϕ(x)], then θm  ϕ(c)
for some c ∈ C; and
(iii) if m = 3k + 3, there is some formula ϕ( x̄) ∈ p( x̄) such that θm  ¬ϕ(c̄k ).

To do this, we proceed by induction on m. Take θ0 to be ∀x[x = x]. Then

certainly T ∪ {θ0 } is satisfiable. Now suppose m ∈ N+ and we have constructed
θ0 , . . . , θm−1 .
If m is of the form 3k + 1 for some k ∈ N, take θm = θm−1 ∧ σk , if T ∪
{θm−1 ∧ σk } is satisfiable, and take θm = θm−1 ∧ ¬σk otherwise. Then T ∪ {θm }
is satisfiable.
If m is of the form 3k +2, θm−1  σk and σk is of the form ∃x[ϕ(x)], choose a
constant symbol c ∈ C which does not appear in θm−1 and set θm = θm−1 ∧ ϕ(c).
Then T ∪ {θm } is satisfiable because c is a new constant. Otherwise, we just let
θm = θm−1 .
If m is of the form 3k + 3, we claim there is a formula ϕ( x̄) ∈ p( x̄) such that
T ∪{θm−1 , ¬ϕ(c̄k )} is satisfiable. Then we can take θm = θm−1 ∧ ¬ϕ(c̄k ). To prove
the claim, we write θm−1 in the form ρ(c̄k , d̄), where ρ( x̄, ȳ) is an L-formula and
d̄ is a tuple of constants from C, all distinct from the elements of c̄k . Then,
if there is no ϕ( x̄) ∈ p( x̄) such that T ∪ {θm−1 , ¬ϕ(c̄k )} is satisfiable, for each
ϕ( x̄) ∈ p( x̄) we have T  ρ(c̄k , d̄) → ϕ(c̄k ), but none of the constants in c̄k or
in d̄ occur in T , so we must have T  ∀ x̄∀ȳ[ρ( x̄, ȳ) → ϕ( x̄)], or equivalently,
T  ∀ x̄[∃ȳ[ρ( x̄, ȳ)] → ϕ( x̄)]. But then ∃ȳ[ρ( x̄, ȳ)] is a principal formula for
p( x̄), a contradiction.
Thus we can produce our list of L∗ -sentences θm for m ∈ N satisfying all the
conditions.
Now take T ∗ to be the deductive closure of T ∪ {θm | m ∈ N }. Then T ∗
is finitely satisfiable, it is complete by condition (i), and it has the witness
property by condition (ii), and so it is a Henkin theory. Using Proposition 11.6,
let M∗ be a canonical model of T ∗ and M its reduct to L. By condition (ii),
every element of M is named by a constant symbol from C, and so M is
136 24 Omitting Types

countable. By condition (iii), no n-tuple of elements of M realises p. So M

omits p, as required. 
The theorem can be improved. More or less the same proof, with a little
more organisation, gives the following.
Theorem 24.4 (Vaught’s omitting types theorem) Let T be a complete theory
in a countable language L, and let P be a countable set of non-principal types
of T . Then there is a countable model of T which omits all the types in P.

Exercises
24.1 Let p( x̄) be an n-type of Th(M). Show that p is principal if and only if
there is a non-empty definable set S ⊆ M n such that S ⊆ p(M).
24.2 In Th(Ns-ring ), give an example of a complete principal 1-type and an
incomplete non-principal 1-type.
24.3 Show that if p( x̄) is a complete principal type, then there is a principal
formula ψ( x̄) in p( x̄). Find an example of an incomplete principal type
which does not contain a principal formula.
24.4 In Ro-ring , show that the family of intervals {(0, 1 + 1/n) | n ∈ N+ } is a
principal 1-type and the family of intervals {(1, 1 + 1/n) | n ∈ N+ } is a
non-principal 1-type.
24.5 Let r ∈ R be a real algebraic number, that is, a root of a non-zero
polynomial f (x) ∈ Z[x]. Show that tpRo-ring (r) is a principal type. Using
quantifier elimination for Ro-ring , show that if r ∈ R is a transcendental
number, then tpRo-ring (r) is non-principal.
∞
24.6 Let K be a field and let V |= T K-VS . Suppose (a1 , . . . , an ) ∈ V n is linearly
independent and (b1 , . . . , bn ) ∈ V is also linearly independent.
n

(a) Show that tp(ā) = tp(b̄).

(b) Suppose K is a finite field. Show tp(ā) is a principal type.
(c) Suppose K is an infinite field and n  2. Show tp(ā) is not principal.

24.7 Here is an example of a theory with no principal complete types. Let B

be the set of all binary sequences b = (bn )n∈N+ , with each bn ∈ {0, 1}. We
make B into an L-structure B by taking L to be the language with unary
relation symbols Zd for each d ∈ N+ , where ZdB = {b ∈ B | bd = 0 }. Show
that Th(B) has no principal complete 1-types.
 Exercises 137

24.8 Sketch a proof of the omitting types theorem, including any lemmas
which are needed for its proof. Give all the key ideas and explain how
they fit together, without giving all the details.
24.9 Prove Theorem 24.4.
 25
Countable Categoricity

In Chapter 15 we saw that the theory DLO of dense linear orders without
endpoints is countably categorical, that is, up to isomorphism, it has exactly
one countably infinite model, which is Q< . Then, in Chapter 17, we saw that
for each n ∈ N+ , only finitely many subsets of Qn are definable in Q< . In
the terminology of Chapter 20, the Lindenbaum algebras Lindn (DLO) are all
finite.
In this chapter we show that this finiteness property characterises countably
categorical theories. The proof uses types, and there are two more equivalent
properties involving types.

25.1 Stone Spaces

First we note that we do not lose much by considering only complete types.
Lemma 25.1 Let T be a complete theory. Then every n-type of T is contained
in a complete n-type of T .
Proof Let p( x̄) be an n-type of T . By Lemma 23.6, there is a model M of
T and ā ∈ M n such that M |= p(ā). Then tpM (ā) ⊇ p( x̄), and tpM (ā) is a
complete n-type of T . 
Definition 25.2 (Stone space) For a complete theory T , let S n (T ) be the set
of all complete n-types of T . It is called the nth Stone space of T .
Proposition 25.3 Let T be a complete theory and n ∈ N+ . The following are
equivalent:

(i) Every type in S n (T ) is principal.

(ii) The Stone space S n (T ) is finite.
(iii) The Lindenbaum algebra Lindn (T ) is finite.

138
 25.2 The Ryll–Nardzewski Theorem 139

Proof (i) ⇒ (ii): Suppose every type in S n (T ) is principal, but suppose for
a contradiction that S n (T ) is infinite, say S n (T ) = {pi | i ∈ I }. Let ψi be a
principal formula for pi . Let q = {¬ψi | i ∈ I }. We claim that q is finitely
satisfiable. So let ¬ψi1 , . . . , ¬ψir be a finite subset of q. Since I is infinite, there
is i ∈ I  {i1 , . . . , ir }. Let ā be an n-tuple in some model M such that tp(ā) = pi .

Then M |= rj=1 ¬ψi j (ā). So q is finitely satisfiable and hence is a type. By
Lemma 25.1, there is a complete type q containing q. But q cannot be any of
the pi for i ∈ I, which is a contradiction. So S n (T ) is finite.
(ii) ⇒ (i): Suppose S n (T ) is finite, say, S n (T ) = {p1 , . . . , pr }. For each i, j ∈
{1, . . . , r} with i  j, there is a formula ϕi j ( x̄) such that ϕi j ( x̄) ∈ pi and ¬ϕi j ( x̄) ∈

p j . Let ψi = rj=1, ji ϕi j . Then T  ∃ x̄ψi ( x̄) because pi is a type and so finitely
satisfiable. Now suppose M |= T and ā ∈ M n with M |= ψi (ā). Then, for each
j  i, M |= ϕi j (ā), so tpM (ā)  p j . Thus tpM (ā) = pi . In particular, ψi is a
principal formula for pi . So every p ∈ S n (T ) is principal.
(iii) ⇒ (ii): Each type in S n (T ) corresponds to a subset of Lindn (T ), and
different types correspond to different subsets. There are 2|Lindn (T )| subsets of
Lindn (T ), so |S n (T )|  2|Lindn (T )| . Thus, if Lindn (T ) is finite, so is S n (T ).
(ii) ⇒ (iii): Suppose S n (T ) is finite. Let r = |S n (T )|. Then, by (ii)
⇒ (i), every type in S n (T ) is principal. Say the principal formulas are
ψ ( x̄), . . . , ψr ( x̄). Then, for each subset J of {1, . . . , r}, the formula ϕ J ( x̄) =
1
j∈J ψ j ( x̄) defines a different subset of M . But if S is a definable subset of
n
n
M , and p is a type, then either every realisation of p is in S or no realisation
of p is in S , so S must be defined by one of the formulas ϕ J ( x̄). Hence
|Lindn (T )| = 2|S n (T )| , so |Lindn (T )| is finite. 

25.2 The Ryll–Nardzewski Theorem

Now we can state and prove the main theorem characterising countably
categorical theories. We use the results on realising and omitting types and
also the back-and-forth method.
Theorem 25.4 (Ryll–Nardzewski, Svenonius, Engeler) Let T be a complete
theory with infinite models in a countable language L. The following are
equivalent.

(i) T is countably categorical.

(ii) For all n ∈ N+ , every type in S n (T ) is principal.
(iii) For all n ∈ N+ , the Stone space S n (T ) is finite.
(iv) For all n ∈ N+ , the Lindenbaum algebra Lindn (T ) is finite.
140 25 Countable Categoricity

Proof Conditions (ii), (iii), and (iv) are all equivalent by Proposition 25.3.
(i) ⇒ (ii): Suppose that T is countably categorical but, for a contradiction,
assume there is a non-principal type p ∈ S n (T ) for some n ∈ N+ . By the
omitting types theorem, Theorem 24.3, there is a countable model A of T
which omits p. By Lemma 23.6, there is a countable model B of T and ā ∈ Bn
such that tp(ā) = p. Since T is countably categorical, there is an isomorphism
π : B → A. Then, by Proposition 23.4, we have tpA (π(ā)) = tpB (ā) = p.
But A omits p, so we have a contradiction. So every complete type of T is
principal.
(ii) ⇒ (i): This part of the proof uses the back-and-forth method and is very
similar to the proof that DLO is ℵ0 -categorical. Firstly, since the language is
countable and T has infinite models, by the strong compactness theorem, there
is at least one model of cardinality ℵ0 .
Suppose A and B are both models of T of cardinality ℵ0 . Enumerate A as
(an )n∈N+ and B as (bn )n∈N+ . We will construct new enumerations (αn )n∈N+ of A
and (βn )n∈N+ of B such that for every n ∈ N and every L-formula ϕ(x1 , . . . , xn ),
we have

A |= ϕ(α1 , . . . , αn ) if and only if B |= ϕ(β1 , . . . , βn ). (∗)

We proceed by induction on n. For n = 0, condition (∗) says that every L-

sentence is true in A iff it is true in B. This is correct, because A and B are
models of the same complete L-theory.
Now suppose we have αi and βi for all i < n. If n = 2m − 1, odd, then let
αn = am . Let ψn (x1 , . . . , xn ) be a principal formula for (α1 , . . . , αn ). Then we
have A |= ψn (α1 , . . . , αn ), and so we get A |= ∃xn [ψn (α1 , . . . , αn−1 , xn )]. Then
by induction using (∗), it follows that B |= ∃xn [ψn (β1 , . . . , βn−1 , xn )]. Choose
βn ∈ B such that B |= ψn (β1 , . . . , βn ). Then, since ψn is a principal formula,
condition (∗) holds for n.
If n = 2m, even, then let βn = bm . Repeat the above argument, swap-
ping the roles of A and B, to choose αn such that condition (∗) holds
for n.
Note that every am appears as some αn . It may appear more than once, but
if αn = αl , then this is witnessed by a formula, so also βn = βl by (∗). Thus we
may define a function π : A → B by π(αn ) = βn . This function is surjective,
since every bm appears as some βn . It is injective, since if βn = βl , then also
αn = αl by (∗). Now (∗) also tells us that π preserves all formulas, in particular
it preserves all atomic formulas and their negations, so it is an embedding. The
same argument shows that π−1 is an embedding, so π is an isomorphism. So
A  B, as required. Thus T is countably categorical. 
 Exercises 141

Exercises
25.1 Let T be any complete theory. Show that a formula ψ( x̄) is a principal
formula for T if and only if [ψ( x̄)] is an atom of the Boolean algebra
Lindn (T ).
25.2 Use the idea from the proof of Lemma 20.6 to show that if Lindn (T ) is
finite, then every type in S n (T ) is principal.
25.3 Suppose that every complete n-type of a complete theory T is principal.
Show that every incomplete n-type of T is also principal.
25.4 Show that the theories ACF0 = Th(Cring ) and RCF = Th(Ro-ring ) are not
ℵ0 -categorical.
25.5 List all of the complete 2-types in the theory T F∞3 -VS . How many complete
3-types are there?
25.6 Sketch a proof of the Ryll–Nardzewski theorem. Explain all the key ideas
in the proof and how they fit together, without writing out all the details.
25.7 Suppose that T = Th(A), that |A| = ℵ0 , and that every finite tuple ā from
A realises a principal type. By adapting the last part of the proof of the
Ryll-Nardzewski theorem, show that for any model B |= T , there is an
elementary embedding π : A → B.
25.8 Countably categorical theories can also be characterised in terms of
permutation groups. This exercise and the next one briefly explain the
idea. Suppose A is a countable L-structure whose theory is countably
categorical. Let ā, b̄ ∈ An have the same type. Use the back-and-forth
method to construct π ∈ Aut(A) such that π(ā) = b̄.
25.9 Using the Ryll–Nardzewski theorem, we can deduce that the permutation
action of Aut(A) on An has only finitely many orbits, for each n ∈ N+ . A
group G acting on an infinite set A such that the induced action on An has
only finitely many orbits is called an oligomorphic permutation group.
Let G be an oligomorphic permutation group on a countably infinite
set A. Name each G-orbit O of An by an n-ary relation symbol RO , to
make a structure A in a language L. Show that every quantifier-free
definable subset of An is a finite union of orbits and that the projection
of such a set to An−1 is still quantifier-free definable, so A has quantifier
elimination. Show that A is countably categorical.
 26
Large and Small Countable Models

In the previous chapter we considered theories with exactly one countable

model. There are other theories, such as T Q∞-VS , where we can make sense of
the idea that some countable models are smaller than others. In cardinality
they are the same size, but the vector space Q1 of dimension 1 is in a sense
smaller than the the space Q3 of dimension 3, which in turn is smaller than
the vector space Q⊕ℵ0 of dimension ℵ0 . One way to see this is that there are
elementary embeddings Q1  Q3  Q⊕ℵ0 , but not the other way around. In
this chapter we consider notions of the smallest and largest countable models
and give a sufficient condition on a theory T for them to exist. Most of the
proofs involve constructing an elementary embedding inductively, one element
at a time, sometimes using the back-and-forth method. These proofs are very
similar to each other and to proofs we have done earlier, so they are left as
exercises.

26.1 Atomic and Prime Models

Let T be a complete theory with infinite models.

Definition 26.1 A model A |= T is prime if, for every model M |= T , there

is an elementary embedding A  M.

Definition 26.2 A structure A is atomic if, for every finite tuple ā from A,
the type tpA (ā) is a principal type.

Warning The word atomic is related to the principal formulas which are
atoms in the Boolean algebras Lindn (T ); it has nothing to do with atomic
formulas.

142
 26.2 Saturated and Universal Models 143

Both these definitions capture some aspect of a model being small. By

the omitting types theorem, atomic models realise as few different types as
possible. This gives a way to relate the two notions.
Theorem 26.3 Let T be a complete theory in a countable language, and
suppose A |= T . Then A is a prime model of T if and only if A is countable
and atomic.
Proof By the strong version of the compactness theorem, Theorem 11.7,
there is a countable model B of T . So for A to embed in B, it must also be
countable. By the omitting types theorem, given any non-principal type p, we
can choose B to omit p. If A is a prime model, then by Proposition 23.4, A
must also omit p. So A can only realise principal types and so is atomic. So
prime models of T are countable and atomic.
For the converse direction, see Exercise 25.7. 
Theorem 26.4 Suppose that A and B are both countable atomic models of T .
Then A  B.
Proof idea The proof uses the back-and-forth method and is very similar to
the proof of Theorem 15.3 and (ii) ⇒ (i) from Theorem 25.4. 
Examples 26.5

(i) Every model of DLO is atomic. More generally, by the Ryll–Nardzewski

theorem, every model of a countably categorical theory is atomic.
(ii) If V |= T Q∞-VS , then V is atomic if and only if dim V = 1.
(iii) Zring is a prime model of its theory.

26.2 Saturated and Universal Models

A prime model of T embeds elementarily in every model of T . By analogy,
one might define a universal model of T to be one in which every model
of T embeds elementarily. By cardinality considerations and the Upward
Löwenheim–Skolem theorem, such a model cannot exist. However, if we are
more modest and merely ask for models of sufficiently small cardinality to
embed, then we can hope to find such models.

Definition 26.6 A model M |= T is universal if for every model A |= T such

that |A|  |M|, there is an elementary embedding A  M.
144 26 Large and Small Countable Models

Analogously to the situation for small models, we also consider the property
of a model realising as many types as possible. However, there is a subtlety
here in that we need to consider types over parameters, that is, types in
expansions of T by constant symbols.

Definition 26.7 A structure M is ℵ0 -saturated if, for every expansion M of

M by naming finitely many elements of M by constant symbols, every n-type
of Th(M ) is realised in M . These new constants are called parameters, and
types of Th(M ) are called types over parameters of Th(M).
Theorem 26.8 Suppose that M is countable and ℵ0 -saturated. Then M is a
universal model of its theory.
Proof idea The proof is very similar to Exercise 25.7. 
Theorem 26.9 Suppose that M and N are both countable and ℵ0 -saturated
models of the same complete theory. Then M  N.
Proof idea As for the proof of Theorem 26.4, the back-and-forth method is
used. 

26.3 0-Stable Theories

Not every theory has prime models or countable ℵ0 -saturated models. We can
give a sufficient condition for both in terms of counting types.

Definition 26.10 A complete theory T is 0-stable (zero-stable) if, for every

n ∈ N+ , the Stone space S n (T ) is countable. Many authors say T is small rather
than 0-stable.

By the Ryll–Nardzewski theorem, every countably categorical theory is

0-stable, since finite sets are countable.
Lemma 26.11 Suppose T is a 0-stable L-theory and T  is a complete theory
extending T in a language L = L ∪ {c1 , . . . , cr } with finitely many new constant
symbols. Then T  is also 0-stable.
Proof For n ∈ N+ , let f be the map f : S n (T  ) → S n+r (T ) given by
f (p) = {ϕ(x1 , . . . , xn+r ) | ϕ(x1 , . . . , xn , c1 , . . . , cr ) ∈ p }. It is straightforward to
check that f (p) really is a complete (n + r)-type of T and that the map f is
injective. So |S n (T  )|  |S n+r (T )|  ℵ0 . 
 26.3 0-Stable Theories 145

Theorem 26.12 Let T be a complete theory in a countable language, L. Then

T has a countable ℵ0 -saturated model if and only if T is 0-stable.
Proof First suppose that T has a countable ℵ0 -saturated model M. Then M
realises every n-type in S n (T ) for every n ∈ N+ . Since M is countable, it has
only countably many n-tuples. So each S n (T ) must be countable.
Now suppose that T is 0-stable. Let A1 be any countable model of T , and
expand it to A+1 by naming each element a ∈ A by a new constant symbol ca .
For any finite subset S ⊆ A1 , let PS1 be the set of all n-types (for all n ∈ N+ )
of the expansion of A1 just by the constant symbols ca for a ∈ S . By Lemma
26.11, PS1 is countable. There are only countably many finite subsets of A1 , so
 S
P1 = P1 | S a finite subset of A1 is also countable. By Proposition 23.7,
there is a countable elementary extension A+1  A+2 which realises all the
types in P1 . Then the reduct A2 of A+2 to L is an elementary extension of A1 .
We iterate this process to get a chain of elementary extensions

A 1  A2  A3  · · ·  A r  · · ·

indexed by natural numbers. Let M be the union of the chain. Then, by

Exercise 13.9, M is an elementary extension of each An . As a countable
union of countable sets, it is countable. If S is any finite subset of M, there is
r ∈ N+ such that S ⊆ Ar . Then, by construction, A+r+1 realises every type of T
expanded by constant symbols for the elements of S . Since M is an elementary
extension of Ar+1 , M+ also realises these types. Hence M is an ℵ0 -saturated
model of T . 
There are theories with prime models which are not 0-stable. For example,
Th(Ns-ring ) has uncountably many 1-types, but Ns-ring is a prime model. So
0-stability is not a necessary condition for the existence of prime models.
However, it is sufficient.
Theorem 26.13 Suppose T is 0-stable. Then T has a countable atomic (hence
prime) model.
Proof Let P be the set of all non-principal complete n-types of T , for all
n ∈ N+ . Since T is 0-stable, P is countable. Then, by Theorem 24.4, there is a
countable model A |= T which omits all the types in P and hence is an atomic
model. 
Remark 26.14 It is possible to give a necessary and sufficient condition for
the existence of a prime model of T using the topology of the type spaces.
146 26 Large and Small Countable Models

Exercises
26.1 For which K and d is a K-vector space of dimension d an atomic LK-VS -
structure? For which K and d is it prime, universal, or ℵ0 -saturated?
26.2 Suppose that T is a complete theory in a countable language and T has
an uncountable atomic model A. Show that T has a prime model.
26.3 Show that a complete theory T is countably categorical if and only if
every model of T is atomic.
26.4 Suppose that for every expansion A of A by finitely many constant
symbols, every 1-type of Th(A ) is realised in A . Show that A is ℵ0 -
saturated.
26.5 Complete the proofs of all the theorems in this chapter.
26.6 A structure A is strongly ℵ0 -homogeneous if, whenever ā, b̄ are n-
tuples of A and tpA (ā) = tpA (b̄), then there is an automorphism π ∈
Aut(A) such that π(ā) = b̄. Using the back-and-forth method, show
that countable atomic models and countable ℵ0 -saturated models are
strongly ℵ0 -homogeneous.
26.7 Suppose that A and B are countable, strongly ℵ0 -homogeneous models
of a complete theory T and they realise the same sets of n-types for all
n ∈ N+ . Show that A  B.
26.8 Suppose that A is countable, universal, and strongly ℵ0 -homogeneous.
Show that A is ℵ0 -saturated. Give an example of a countable universal
structure which is not ℵ0 -saturated.
26.9 Let P be the set of all prime numbers in N. Given any subset Q of P,
show there is a 1-type of Ns-ring of an element which is divisible by all
the primes in Q but no other primes in P. Deduce that S 1 (Ns-ring ) is
uncountable.
26.10 Suppose that M is countable and ℵ0 -saturated and that S ⊆ M n . Show
that S is preserved under all automorphisms of M if and only if S is a
union of type-definable sets.
 27
Saturated Models

The only method we have used in this book to prove that a given axiomatisation
of a theory is complete is the Łos–Vaught test: a theory with no finite models
which is categorical in some infinite cardinality is complete. In Chapter 18
we briefly discussed how the back-and-forth method can be adapted to give a
second method. In this chapter we give a third method using saturated models
and apply it to the theory of discrete linear orders without endpoints. This
works in the same spirit as the Łos–Vaught test, but we need to use more set
theory than before: at least transfinite induction, and for some theories, more.

27.1 Saturation
Theorem 26.9 is suggestive. It states that countable ℵ0 -saturated models of a
complete theory are unique up to isomorphism. Suppose a theory T has no
finite models and we can prove that any two countable ℵ0 -saturated models of
T are isomorphic. Then we want to conclude that T is complete. The problem
is that T might not have any countable ℵ0 -saturated model or that it might have
a completion with no countable ℵ0 -saturated model. So we would still need to
prove that every completion of T is 0-stable to deduce that T is complete. We
can address this problem by generalising the notion of ℵ0 -saturation.

Definition 27.1 Let κ be an infinite cardinal. A structure M is κ-saturated if,

for every subset A ⊆ M such that |A| < κ, writing MA for the expansion of M
by constant symbols for the elements of A, every n-type of Th(MA ) is realised
in MA .
M is saturated if it is |M|-saturated.

As in Exercise 26.4, it is sufficient to consider 1-types rather than n-types.

147
148 27 Saturated Models

If we use the same back-and-forth proof as for Theorem 26.9, but with
transfinite induction in place of induction on natural numbers, we get the
following.
Theorem 27.2 Suppose A, B are saturated models of the same complete
theory T , of the same cardinality. Then A  B. 

27.2 Stability
We still have the problem of the existence of these saturated models. As in
the countable case, the only obstacle is that there may be too many types to
fit into a model, without making the model too large. There is an analogue of
0-stability.

Definition 27.3 Let κ be an infinite cardinal. A complete L-theory T is κ-

stable if, over any set of at most κ parameters in any model of T , there are at
most κ complete types.
T is stable if it is κ-stable for some infinite κ. Otherwise, it is unstable.

If T is κ-stable, then a transfinite version of the construction for the proof

of Theorem 26.12 shows that there is a saturated model of T of cardinality κ.
Surprisingly, stability is a much more robust notion than 0-stability, and it
has several other equivalent definitions which, on the face of it, are unrelated
to counting types. For example:
Fact 27.4 Suppose M is a structure in which some infinite linear order is
definable. Then Th(M) is unstable.
A related but more complicated condition called the Order Property is
in fact equivalent to instability. Such conditions sometimes make it possible
to prove that the theory of a structure is stable without first proving the
completeness of any axiomatisation.
Stability theory and its generalisations are a major area of research in
model theory. The beginnings of the subject are covered in Marker [Mar02],
Tent and Ziegler [TZ12], and Baldwin [Bal88]. Poizat [Poi00] and Pillay
[Pil83] offer a different approach. More advanced books taking the subject
in different directions are Pillay’s [Pil96], which highlights the connections
between stability theory and geometric ideas, and Shelah’s great work [She90],
which is still the best guide to the subject for those with the perseverance to
work through it.
 27.4 Discrete Linear Orders 149

27.3 Strongly Inaccessible Cardinals

If a theory is unstable, which by Fact 27.4 the theory of discrete linear orders
is, we can find a saturated model by working harder with the set theory.
Given a subset A of a model of an L-theory T with |A| = κ, we have
| Form(LA )| = κ + ℵ0 . A type over A is a subset of Form(LA ) so there are at
most |P Form(LA )| types.

Definition 27.5 A cardinal κ is strongly inaccessible if, for a set X of size κ,

whenever Y ⊆ X with |Y| < κ, then also |PY| < κ, and furthermore, if (Yi )i∈I
is a family of subsets of X such that |I| < κ and |Yi | < κ for each i ∈ I, then

i∈I Yi  X.

Again, a transfinite version of the proof of Theorem 26.12 gives us saturated

models.
Theorem 27.6 Let T be an L-theory with infinite models, and let κ be a
strongly inaccessible cardinal such that κ > |L|. Then T has a saturated model
of cardinality κ. 
So we have the following analogue of the Łos–Vaught test.
Theorem 27.7 Suppose T is an L-theory with no finite models and κ is a
strongly inaccessible cardinal such that κ > |L|. Then T is complete if and only
if any two saturated models of T of cardinality κ are isomorphic.
Proof Combine Theorems 27.2 and 27.6. 

27.4 Discrete Linear Orders

We give an application of Theorem 27.7.
Theorem 27.8 The theory DiscLO of nonempty discrete linear orders without
endpoints is complete.
Proof From Exercise 15.8, any discrete linear order without endpoints is
isomorphic to a lexicographic product A × Z< , where A is a non-empty linear
order. Conversely, for any non-empty linear order A it is easy to see that
A × Z< |= DiscLO.
Suppose that M = A × Z< is saturated (or even just ℵ0 -saturated).
By Exercise 15.2, the function s taking an element of M to its immediate
successor is definable, and likewise the predecessor function s−1 is definable.
Given α ∈ M, consider the set of formulas with parameter α given by
150 27 Saturated Models

{sn (x) < α | n ∈ N+ }. This is finitely satisfiable, so a 1-type, so by saturation,

it is realised in M. Thus A has no least endpoint. Similarly, A has no greatest
endpoint. Given a < b in A, we have α = (a, 0) and β = (b, 0) in M. Then
{α < s−n (x) ∧ sn (x) < β | n ∈ N+ } is a 1-type over the parameter set {α, β} and
so is realised in M. Thus A is a dense linear order. Using quantifier elimination
for DLO, we can understand that 1-types of DLO over any set of parameters
just correspond to cuts in the order. It follows that since M is saturated, so is A.
Now suppose that M = A × Z< and N = B × Z< are two saturated models
of DiscLO of the same strongly inaccessible cardinality κ. Then A and B are
saturated models of DLO of cardinality κ. Since DLO is a complete theory,
A  B. Thus M  N. So by Theorem 27.7, DiscLO is complete. 

27.5 Discussion
Note that the proof above of Theorem 27.8 is almost entirely about under-
standing the models of the theory, in fact, countable parts of models, and
relating them to something we already understand (countable parts of models
of DLO). The set theory behind the saturated models is not visible. Any
proof that DiscLO is complete needs to use at least some comparable level of
understanding of the models. In practice, this is exactly how saturated models
are used to simplify proofs, which could in principle be done by working
entirely inside small, often countable, models.
We have focused on proving completeness of theories here, but another
advantage of saturated models is that they have a lot of automorphisms, so
the methods we have used for quantifier elimination and for understanding
definable sets or types more generally can be extended easily.
One minor disadvantage to the method of saturated models, at least for
unstable theories, is the requirement to work with strongly inaccessible
cardinals. Under the usual set theoretic axioms (ZFC), you cannot prove that
strongly inaccessible cardinals do or do not exist. There are a number of
different ways to get around this. You can just work in this slightly stronger
set theory with strongly inaccessible cardinals or assume the generalised
continuum hypothesis, which gives another reason for saturated models to
exist. Then you can use set-theoretic methods to show that the theorems you
prove with these extra assumptions could also be proven in ZFC. Chang and
Keisler [CK90] use special models, which are similar to saturated models but
do always exist, and (for countable recursive theories) they also use recursively
 Exercises 151

saturated models. Tent and Ziegler [TZ12] and Hodges [Hod93] take different
approaches again.

Exercises
27.1 Show that R< is not a saturated model of DLO.
27.2 Show that DiscLO is 0-stable. What is its countable saturated model?
27.3 Let M |= DLO and let A be a substructure with domain A. Using
quantifier elimination for DLO, show that 1-types of M over A
correspond to cuts in the order of A, that is, subsets C ⊆ A such that if
c ∈ C and a ∈ A with a < c, then a ∈ C.
27.4 Sketch a proof of Theorem 27.8, including any theorems and lemmas
which are used.
27.5 Show that, for each n ∈ Z, the function fn given by x → x + n is
definable in Z< , and hence in any model of DiscLO. Then show that
Z; <, ( fn )n∈Z  has quantifier elimination.
27.6 Characterise the models of the theory of infinite discrete linear orders
with endpoints, and show the theory is complete.
27.7 Suppose we alter Definition 27.3 to say that T is κ-stable if over any
set of at most κ parameters there are at most κ + ℵ0 types, and we
remove the condition that κ be infinite. Show that the new definition
is equivalent to the old one when κ is infinite and that for any n ∈ N
the newly defined notion of n-stable is equivalent to Definition 26.10
of 0-stable.
27.8 This exercise, analogous to Exercise 26.6 in the countable case, indi-
cates how saturated structures have a lot of automorphisms. A structure
A is strongly κ-homogeneous if, whenever ā and b̄ are potentially
infinite tuples of length strictly less than κ such that tp(ā) = tp(b̄),
there is an automorphism π of A such that π(ā) = b̄. Make sense of
this definition, and use transfinite induction to prove that a saturated
structure of cardinalty κ is strongly κ-homogeneous.
27.9 Referring to another source, such as [Poi00], [Mar02], or [TZ12],
use the back-and-forth method to give another proof that DiscLO is
complete.
27.10 Assuming the generalised continuum hypothesis, show that every
theory has saturated models.
 Part VI

Algebraically Closed Fields

One of the largest areas of applications of model theory is to the model

theory of fields. These may be fields considered just in the ring language
or fields with operators such as derivations, automorphisms, or valuations;
fields equipped with certain analytic functions; or fields equipped with relation
symbols naming certain subgroups or subfields. In many cases the fields under
consideration are algebraically closed, but even when they are not, the theory
of algebraically closed fields usually plays a role. In this final part of the book
we will study the theory of algebraically closed fields. Two chapters give the
basic algebra of fields and of algebraic closures, then we prove categoricity,
completeness, and quantifier elimination by the methods developed in Parts III
and IV, in particular the Łos–Vaught test and substructure completeness. We
show how definable sets are algebraic varieties and give a model-theoretic
proof of Hilbert’s Nullstellensatz. The connection between the Stone spaces
of types and the Zariski spectra of prime ideals is given as an exercise.
In this part of the book, every structure will be a ring considered as an
Lring -structure, so rather than writing a ring as Rring or R, we will use the same
symbol R for both a ring and its underlying set.

153
 28
Fields and Their Extensions

In this chapter we review the basic concepts of field extensions we need,

up to the Hilbert basis theorem. Details and proofs can be found in basic
textbooks on rings and fields, and Lang [Lan02] is a standard reference. To
understand fields with extra structure, one usually has to develop or understand
the analogous concepts.

28.1 The Basic Algebra of Rings

Recall the language of rings is Lring = +, ·, −, 0, 1. The axioms for rings and
for fields were given in Chapter 6.

Definition 28.1 A domain (sometimes called an integral domain) is a ring

with no zero divisors, that is, it is a model of the sentence

∀xy[x · y = 0 → (x = 0 ∨ y = 0)].

Every field is a domain, and since the ring axioms and the property of being
a domain are given by universal sentences, every Lring -substructure of a field is
a domain. The converse is also true: every domain is a substructure of a field,
in particular its field of fractions. The following lemma captures the property
we need.
Lemma 28.2 If R is a domain, then it has a field of fractions, which is a field
FR ⊇ R such that any embedding π : R → F of R into a field F extends
uniquely to an embedding FR → F.
Examples 28.3 We write R[x1 , . . . , xn ] for the ring of polynomials in the
variables x1 , . . . , xn , with coefficients from a ring R. If R is a domain, so is
R[x1 , . . . , xn ]. If R is a field F, the field of fractions of F[x1 , . . . , xn ] is written as

155
156 28 Fields and Their Extensions

F(x1 , . . . , xn ). It is called the field of rational functions over F in the variables

x1 , . . . , xn .

Definition 28.4 A homomorphism of rings π : R → S is a function that

preserves +, ·, −, 0, and 1. The kernel of π is ker(π) = {a ∈ R | π(a) = 0 }.
An ideal of a ring R is an additive subgroup I ⊆ R such that, for all r ∈ R
and all a ∈ I, we have r · a ∈ I.
The ideal generated by a subset A ⊆ R is the intersection of the ideals of R
which contain A.
Lemma 28.5 Suppose I ⊆ R is the ideal generated by f1 , . . . , fr , and f ∈ R.

Then f ∈ I if and only if there are g1 , . . . , gr ∈ R such that f = ri=1 gi fi .

Lemma 28.6 The kernel of a ring homomorphism π : R → S is an ideal of R.
If I is an ideal of R, there is a quotient ring R/I whose elements are the
cosets a + I = {a + b | b ∈ I } for a ∈ R. The map π : R → R/I defined by
π(a) = a + I is a surjective ring homomorphism with kernel I.
Definition 28.7 An ideal I of R is proper if I  R. It is prime if it is proper
and for all a, b ∈ R, if ab ∈ I, then a ∈ I or b ∈ I. An ideal I is maximal if it is
proper and there is no proper ideal J of R with I  J.
Lemma 28.8 Let R be a ring and I an ideal of R. Then the quotient R/I is a
domain if and only if I is prime, and it is a field if and only if I is maximal.
Example 28.9 The ring Z/nZ is a domain if and only if n is a prime p, and
then it is a field written F p .

Definition 28.10 The characteristic of a field F is the smallest p ∈ N+ such

1 + · · · + 1 = 0 if such a p exists, and 0 otherwise.
that 
p
The prime subfield of a field F is the intersection of all the subfields of F.

If F has characteristic p > 0, then the prime subfield of F is F p and if F has

characteristic 0, then its prime subfield is the field of rational numbers Q.

28.2 Simple Field Extensions

Algebraically closed fields arise as extensions of any field. To understand this,
we need to understand the finitely generated field extensions. We start with the
extensions generated by just one element.
 28.3 Finitely Generated Field Extensions 157

Definition 28.11 A simple extension of a field F is a field extension K

generated by a single element, say, b. Then we write K = F(b). We write F[b]
for the subring of K generated by F and b. So F(b) is the field of fractions of
F[b].
Lemma 28.12 Let F ⊆ F(b) be a simple field extension. There is a unique
homomorphism evb : F[x] → F(b) such that evb (a) = a for all a ∈ F and
evb (x) = b.
The kernel of evb is Ib := { f (x) ∈ F[x] | f (b) = 0 }, and it determines the
field extension F(b) of F and the choice of generator up to isomorphism. So to
classify the simple extensions of F, it is enough to classify the prime ideals of
F[x].
Proposition 28.13 Every ideal of F[x] is generated by a single polynomial,
which is unique up to multiplication by a non-zero element of F. (That is, F[x]
is a principal ideal domain.) The ideal is prime if and only if the generating
polynomial is irreducible.
A generating polynomial g(x) for the ideal Ib is called a minimal polynomial
for b over F. We write the ideal Ib as g. So simple extensions of a field F are
determined up to isomorphism fixing a choice of generator, by the minimal
polynomial of the generator, which is an irreducible polynomial in F[x].

√ √
Examples 28.14 Suppose √ F 2= Q and b = 2 ∈ Q(√ 2) ⊆ C. Then the
minimal√polynomial of 2 is x − 2. The same √ field Q( 2) is also generated
by c = 2 + 1. The minimal polynomial of 2 + 1 is x2 − 2x − 1.
The number π is transcendental, that is, it does not satisfy any polynomial
equations with coefficients from Q. So its minimal polynomial over Q is the
zero polynomial, and the ideal Iπ is the zero ideal {0}. Then the quotient map
evπ : Q[x] → Q(π) is injective, and the field Q(π) is isomorphic to the field
Q(x) of rational functions in one variable over Q.

28.3 Finitely Generated Field Extensions

If F ⊆ K is a field extension generated by finitely many elements, say, b̄ =
(b1 , . . . , bn ), then we write K = F(b̄) as in the simple
 case. Again,
 K is the field
of fractions of the quotient of F[ x̄] by the ideal f ( x̄) ∈ F[ x̄]  f (b̄) = 0 , which
is a prime ideal. So again, to classify the finitely generated field extensions of
a field F, it is enough to classify the prime ideals of F[x1 , . . . , xn ], for each
158 28 Fields and Their Extensions

n ∈ N. For n > 1, these are not always generated by a single polynomial, but
we have the next best thing.
Theorem 28.15 (Hilbert’s basis theorem) For any ideal I of F[x1 , . . . , xn ],
there is a finite set of polynomials which generates I.
Definition 28.16 If F ⊆ K is a field extension and b ∈ K, then b is said to be
algebraic over F if b satisfies some non-zero polynomial over F. Otherwise, it
is transcendental over F. If every element of K is algebraic over F, then K is
said to be an algebraic extension of F.
Finitely generated algebraic extensions have a special form.
Proposition 28.17 If K = F(b̄) is a finitely generated algebraic extension of
F, then there is a single element a which generates K as an extension of F.
Furthermore, the ring F[a] is actually all of K.
Finally, we give two useful lemmas.
Lemma 28.18 If F ⊆ K ⊆ L is a tower of field extensions, L is algebraic over
K, and K is algebraic over F, then L is algebraic over F.
A zero of a polynomial f (x) is a solution to the equation f (x) = 0.
Lemma 28.19 Let F be a field and f (x) ∈ F[x] a polynomial in one variable,
of degree d. Then there are at most d zeros of f in F.
This chapter is intended as a reminder or reference, but it covers the material
rather quickly to learn it from. Thus there are no exercises, other than to read
up on any parts of the chapter you are unfamiliar with from a suitable reference
and to do the exercises from there.
 29
Algebraic Closures of Fields

In this chapter we study algebraically closed fields from an algebraic point of

view. In particular, we prove that every field has an algebraic closure, which is
unique up to isomorphism.
There are actually two different notions called algebraic closure relating to
fields. They are closely related, and often the difference is not made explicit,
but they play different roles in the model-theoretic approach. Both roles are
important.

29.1 Relative Algebraic Closure

Definition 29.1 Suppose that F ⊆ K is an extension of fields. The relative
algebraic closure of F in K, written aclK (F), consists of all the elements of K
which are algebraic over F.

For example, aclR (Q) is the field of real algebraic numbers. It is not an
algebraically closed field because it is a subfield of R and so, for example, the
polynomial x2 + 1 has no zero in aclR (Q).
Lemma 29.2 For any extension F ⊆ K of fields, aclK (F) is a subfield of K
of cardinality |F|  | aclK (F)|  max{|F|, ℵ0 }. In particular, if F is an infinite
field, then | aclK (F)| = |F|.
Proof We leave the proof that aclK (F) is a subfield as an exercise. For the
cardinality result, note that each element of aclK (F) is a zero of a polynomial
f (x) ∈ F[x], and f has at most deg( f ) zeros, a finite number. So | aclK (F)| 
|F[x]| · ℵ0 . But |F[x]| = max{|F|, ℵ0 }, so | aclK (F)|  max{|F|, ℵ0 }. Also F ⊆
aclK (F), so |F|  | aclK (F)|. 

159
160 29 Algebraic Closures of Fields

29.2 (Absolute) Algebraic Closure

Recall that a field F is algebraically closed if every non-constant polynomial
f (x) ∈ F[x] has a zero in F. A first-order axiom scheme capturing this property
was given in Chapter 7. The fundamental theorem of algebra states that the
complex field Cring is algebraically closed. Now we show that there are many
other algebraically closed fields.
Proposition 29.3 For any field F, there is an extension F ⊆ K such that K is
an algebraically closed field.
We give a proof using the compactness theorem.
Proof Expand Lring to LF by adding constant symbols naming each element
of F. Recall from Chapter 13 that we have Diag(F), the set of all atomic LF -
sentences and negations of atomic LF -sentences which are true in F. Let Σ =
Diag(F)∪{∃x[ f (x) = 0] | f (x) ∈ F[x] and f (x) is non-constant }. Then a model
of Σ is an extension field of F in which every non-constant polynomial over F
has a zero. Let Σ0 be a finite subset of Σ, and let { f1 (x), . . . , fr (x)} be the finite
set of polynomials f (x) such that the axiom ∃x[ f (x) = 0] appears in Σ0 . Let
g1 (x) be an irreducible component of f1 (x), and let F1 = F[x]/g1 . Now let
g2 (x) be an irreducible component of f2 (x) considered as a polynomial over F1 ,
and let F2 = F1 [x]/g2 . Iterating, we get a field extension Fr of F in which all
the polynomials f1 (x), . . . , fr (x) have zeros. Thus Σ is finitely satisfiable. By
compactness, Σ has a model, say, K1 .
Now iterate the process to get a chain of field extensions

F = K0 ⊆ K1 ⊆ K2 ⊆ · · · ⊆ Kn ⊆ · · ·

for n ∈ N such that every non-constant f ∈ Kn [x] has a zero in Kn+1 . Let

K = n∈N Kn . Then K is an algebraically closed field extension of F. 
Now we come to the second notion of algebraic closure of F. Unlike the
relative algebraic closure, it is defined without reference to a previously given
extension field K.

Definition 29.4 An algebraic closure of a field F is an algebraic field

extension of F which is algebraically closed. Very rarely, this notion is called
the absolute algebraic closure of F to distinguish it from the relative algebraic
closure. Usually we let the context make the distinction.

Putting together the previous proposition with the relative algebraic closure,
we can show the existence of an algebraic closure of a field F. With some more
work, we can also show uniqueness of the algebraic closure.
 29.2 (Absolute) Algebraic Closure 161

Theorem 29.5 Every field F has an algebraic closure. If K and L are both
algebraic closures of F, then there is an isomorphism K  L which restricts to
the identity on F.
Proof Let F be any field. By Proposition 29.3, there is an algebraically closed
field extension K of F. Take K0 = aclK (F). Suppose that f (x) is a non-constant
polynomial in K0 [x]. Then since K is algebraically closed, there is b ∈ K
such that f (b) = 0. Then b is algebraic over K0 , which is algebraic over F,
so, by Lemma 28.18, b is algebraic over F, and hence b ∈ K0 . So K0 is an
algebraically closed field which is an algebraic extension of F. So we have
shown that F has at least one algebraic closure.
Now suppose that K and L are both algebraic closures of F. We must find an
isomorphism π : K → L fixing F pointwise. We give a proof using transfinite
induction.
List K as (aα )α<λ for some ordinal λ. (If K is countable, then just take the
α to be natural numbers.) We define subfields Kα of K and embeddings πα :
Kα → L for α  λ as follows:

• K0 = F and π0 : F → L is just theinclusion of F as a subfield of L.

• If α is a limit ordinal, take Kα = K
β<α β and πα = β<α πβ .
• If α = β + 1 is a successor ordinal, let Kα = Kβ (aβ ). Let f (x) be the minimal
polynomial of aβ over Kβ . Then since aβ is algebraic over F (and hence
over Kα ), we have Kα  Kβ [x]/  f . Let Lβ = πβ (Kβ ), a subfield of L, and
let g(x) ∈ Lβ [x] be the polynomial obtained from f (x) by applying πβ to all
the coefficients. Since L is algebraically closed, g(x) has a zero in L. Let b
be such a zero. Since f (x) is irreducible in Kβ [x], g(x) is irreducible in
Lβ [x], so it is the minimal polynomial of b over Lβ . So we have

Kβ (aβ )  Kβ [x]/  f   Lβ [x]/ g  Lβ (b) ⊆ L

via a map taking aβ to b, which extends πβ . Let πα be this embedding of Kα

into L.

Let π = πλ . Then π is an embedding of K into L, which extends π0 , the identity

map on F. It remains to show that π is surjective.
Let b ∈ L and let f (x) be its minimal polynomial over F. By Lemma
28.19, f (x) has only finitely many zeros in K, say, a1 , . . . , an . Since π is an
Lring embedding fixing F, we have f (π(ai )) = 0 for each i = 1, . . . , n, and the
f (ai ) are all distinct because f is injective. So L has at least n zeros of f . But
we can swap the roles of K and L, and the above argument shows there is also
an embedding of L into K. So L has at most n zeros of f . Hence L has exactly
162 29 Algebraic Closures of Fields

n zeros of f , so for some i, f (ai ) = b. So π is surjective, and hence it is an

isomorphism. 
We usually write F alg for the the algebraic closure of F.

Exercises
29.1 Show that if F is an algebraically closed field and f ∈ F[x] is a non-
zero irreducible polynomial, then f has the form f (x) = x − a for some
a ∈ F. Deduce that an algebraically closed field has no proper algebraic
extensions.
29.2 Show that for any field F, there is no algebraically closed proper subfield
of F alg which contains F.
29.3 Let F ⊆ K be an extension of fields. Explain how K can be regarded as
an F-vector space. Let α ∈ K. Show that α ∈ aclK (F) if and only if the
field F(α) is finite-dimensional as an F-vector space.
alg
29.4 For p prime, show that F p is the union of its finite subfields.
29.5 Let F ⊆ K be an extension of fields. Prove that aclK (F) is a subfield of
K.
29.6 The use of the compactness theorem in the proof of Proposition 29.3 is
actually hiding a use of the axiom of choice or transfinite induction. Give
a proof not using compactness when the field F is countable. Where is
the axiom of choice used in the proof of the compactness theorem?
29.7 Give another proof of the uniqueness part of Theorem 29.5 using the
back-and-forth method, avoiding a separate proof that the function π is
surjective.
 30
Categoricity and Completeness

In this chapter we prove that the theory of algebraically closed fields of

a given characteristic is categorical in uncountable cardinalities, even after
adding parameters for a subfield. We deduce that the theory is complete and
has quantifier elimination. The new algebraic concept we use is that of a
transcendence base, which is an analogue of a basis for a vector space.

30.1 Transcendence Bases

The notions of a basis of a vector space and its dimension have analogues for
fields, which are called transcendence base and transcendence degree.

Definition 30.1 Let F ⊆ K be a field extension. A subset B of K is said

to be algebraically independent over F if, for every non-zero polynomial
f (x1 , . . . , xn ) ∈ F[x1 , . . . , xn ], for every n ∈ N, if b1 , . . . , bn ∈ B are distinct,
then f (b1 , . . . , bn )  0.
A transcendence base for the extension F ⊆ K is a subset B of K which is
algebraically independent over F and such that K is algebraic over the subfield
generated by F ∪ B.
Lemma 30.2 Let F ⊆ K be a field extension, and let B ⊆ K be any subset.
Let F[(xb )b∈B ] be the ring of polynomials over F with one variable xb for each
element b ∈ B. Consider the ring homomorphism evB : F[(xb )b∈B ] → K,
which is the identify on F and satisfies evB (xb ) = b. Then B is algebraically
independent over F if and only if the map evB is injective.
Proof By definition, B is algebraically independent over F if and only if the
kernel of evB is the zero ideal if and only if evB is injective. 

163
164 30 Categoricity and Completeness

Proposition 30.3 Let K be an infinite field, let F ⊆ K be a subfield, and let B

be a transcendence base for K over F. Then |K| = max{|F|, |B|, ℵ0 }.
Proof We write F(B) for the subfield of K generated by F ∪ B.
alg
Then we have F(B) ⊆ K ⊆ K alg , and aclK (F(B)) = K alg . So by Lemma
29.2, we have |F(B)|  |K alg | = |K|, with equality if F(B) is infinite. If F is a
finite field and B = ∅, then F(B) = F is finite, so |K alg | = ℵ0 , and we are done.
Otherwise, F(B) is infinite and |F(B)| = max{|F|, |B|, ℵ0 }, so we are done. 
Proposition 30.4 Any two transcendence bases of a field extension F ⊆ K
have the same cardinality.
Proof We leave the proof as an exercise, except in the case when the
cardinality of some transcendence base B is greater than |F| + ℵ0 . In
that case, for any transcendence base B , we have |B | = |K| = |B| by
Proposition 30.3. 
Definition 30.5 We define the transcendence degree of a field extension to
be the cardinality of any transcendence base. The transcendence degree of a
field F is the transcendence degree of F considered as an extension of its prime
subfield.

30.2 Categoricity
Recall from Chapter 7 that we write ACF p for the theory of algebraically
closed fields of characteristic p, axiomatised by the axioms for ACF and either
1 + · · · + 1 = 0 or, if p = 0, the set of sentences
the sentence χ p given by 
  p
¬χq | q ∈ N+ .
Theorem 30.6

(i) For each p, prime or 0, the theory ACF p is categorical in all uncountable
cardinals.
(ii) For any field F, the theory Diag(F) ∪ ACF is categorical in all
uncountable cardinals λ such that λ > |F|.
Proof First we deduce (i) from (ii). For p prime, K |= Diag(F p ) ∪ ACF if
and only if K is an algebraically closed field of characteristic p, with constant
symbols naming 0, 1, 2, . . . , p − 1, if and only if K (without the extra constant
symbols) is a model of ACF p . For p = 0, the same applies with Q in place of
F p . So (i) follows from (ii).
 Exercises 165

Now we prove (ii). Suppose that K and L are both models of Diag(F) ∪
ACF of the same uncountable cardinality λ, with λ > |F|. Then K and L are
both algebraically closed field extensions of F, and by Proposition 30.3, the
transcendence degrees of K and L as extensions of F are both λ. So let B
be a transcendence base for F ⊆ K, and let B be a transcendence base for
F ⊆ L. Then there is an isomorphism π between the subfields F(B) of K and
F(B ) of L, restricting to the identity on F, since by Lemma 30.2, they are both
isomorphic to the field of rational functions in λ variables. Then, by Theorem
29.5, this isomorphism π extends to an isomorphism π : K → L. So the theory
Diag(F) ∪ ACF is categorical in λ. 
Corollary 30.7 For each p, prime or 0, the theory ACF p is complete and has
quantifier elimination.
Proof By the Łos–Vaught test, Lemma 14.11, it follows from part (i) of
Theorem 30.6 that the theories ACF p are complete.
Applying the Łos–Vaught test to part (ii) of the same theorem, we deduce
that ACF is substructure complete. By Proposition 18.2, it has quantifier
elimination, and thus so does each completion ACF p . 
Putting together results from this chapter, we have the following classifica-
tion of algebraically closed fields.
Theorem 30.8 For each cardinal λ and each p, either a prime number or 0, up
to isomorphism, there is exactly one algebraically closed field of characteristic
p and transcendence degree λ. 

Exercises
30.1 Prove that the characteristic of any field is either 0 or a prime number.
Deduce
  that ACF0 is axiomatised by the axioms for ACF together with
¬χ p  p is prime .
30.2 Use the compactness theorem to deduce that if σ is any Lring -sentence,
then ACF0  σ if and only if there is N ∈ N such that for all primes
p > N, ACF p  σ.
30.3 Given a set X, a function cl : PX → PX is a closure operator if, for all
A, B ⊆ X and all a, b ∈ X: if A ⊆ B, then cl(A) ⊆ cl(B), A ⊆ cl(A), and
cl(cl(A)) = cl(A).
A closure operator cl has finite character if, whenever a ∈ cl(A), then
there is a finite subset A0 ⊆ A such that a ∈ cl(A0 ).
166 30 Categoricity and Completeness

A closure operator cl has the exchange property if, whenever b ∈

cl(A ∪ {a})  cl(A), then a ∈ cl(A ∪ {b}).
A pregeometry is a closure operator with finite character which
satisfies the exchange property.
Show that aclK is a pregeometry on any field K and that that linear
span is a pregeometry on any vector space.
30.4 Using the exchange property for aclK , prove Proposition 30.4.
30.5 Give definitions of independent set, spanning set, and basis for a
pregeometry, by analogy to those for vector spaces. Prove that every
pregeometry has a basis and that any two bases have the same cardinality.
30.6 Give an outline of the proof that the theory ACF0 is complete and has
quantifier elimination from first principles, that is, including all the main
results in this book which are used in the proof.
30.7 We outline a proof due to Ax [Ax69] of a theorem of Bailynicki-Birula
and Rosenlicht [BBR62]. In Exercise 31.10, we will give Ax’s more
general theorem.
If f1 , . . . , fn ∈ K[x1 , . . . , xn ], then f = ( f1 , . . . , fn ) defines a poly-
nomial map K n → K n . Let Ψ(K, n, d) denote the statement ‘for all such
f , if each fi has degree at most d and f is injective then f is surjective.’
(a) Show that Ψ(K, n, d) holds when K is a finite field.
alg
(b) Using Exercise 29.4, show that Ψ(F p , n, d) is true for p prime.
(c) By quantifying over the coefficients of the polynomials, for each
fixed n and d, show that there is an Lring -sentence ψn,d such that for
any field K, K |= ψn,d if and only if Ψ(K, n, d) is true.
(d) Using the completeness of ACF p , deduce that Ψ(K, n, d) is true for
all algebraically closed fields of positive characteristic.
(e) Deduce that Ψ(K, n, d) is true for all algebraically closed fields.
(f) Give an example of a field K and a map f showing that Ψ(K, 1, 3)
fails.
(g) Give an example of a surjective polynomial map f : C → C which
is not injective.
 31
Definable Sets and Varieties

In this chapter we fix an algebraically closed field K. The quantifier elimination

theorem proved in Chapter 30 shows that every definable subset of K n
is quantifier-free definable. In this chapter we explore what the quantifier-
free definable sets are, relating them to affine algebraic varieties and, more
generally, to constructible sets.

31.1 Varieties
Quantifier-free formulas are, by definition, Boolean combinations of atomic
formulas. The language Lring has no relation symbols, so the only atomic
formulas are of the form t1 ( x̄) = t2 ( x̄), where t1 and t2 are terms. Lring -terms
are interpreted in K as polynomials f ( x̄) ∈ Z[ x̄]. If we allow parameters
from a subfield F of K, that is, constant symbols naming the elements of
F, then Lring (F)-terms are interpreted as polynomials in F[ x̄]. So atomic
formulas are of the form f1 ( x̄) = f2 ( x̄), for polynomials f1 and f2 . Setting
f ( x̄) = f1 ( x̄) − f2 ( x̄), the atomic formulas are all equivalent to formulas of the
form f ( x̄) = 0.
Before moving to arbitrary Boolean combinations of these equations,
we consider the positive Boolean combinations, that is, those built using
conjunctions and disjunctions but not negations.
By the conjunctive normal form theorem (see Exercise 20.12), a positive
Boolean combination of equations is equivalent to a formula of the form

$
r 
si
fi j ( x̄) = 0.
i=1 j=1

167
168 31 Definable Sets and Varieties

i
Since K is a field, it has no zero divisors, so sj=1 f ( x̄) = 0 if and only if
+ si + si ij
f
j=1 i j ( x̄) = 0. So writing f i ( x̄) for the product j=1 fi j ( x̄), every positive
quantifier-free formula is equivalent to a finite conjunction of polynomial

equations ri=1 fi ( x̄) = 0.
The subsets of K n defined by these systems of polynomial equations are
known as varieties.
Definition 31.1 Let P be a set of polynomials from K[x1 , . . . , xn ]. The zero-
set of P is V(P) = {ā ∈ K n | for all f ( x̄) ∈ P, f (ā) = 0 }. The subset V(P) ⊆ K n
is called an affine algebraic variety, which we will abbreviate to variety. It is
also called a Zariski-closed subset of K n .
Remark 31.2 In algebraic geometry, the word variety is sometimes reserved
for a Zariski-closed subset which is irreducible. See Exercise 31.3. Apart from
affine varieties, there are also projective varieties, quasi-projective varieties,
and abstract varieties. They can also be treated as definable sets, but we will
keep to affine varieties.
In the definition of V(P), the set P of polynomials does not need to be
finite. So it appears that varieties are more general than positive quantifier-free
definable sets. We will show that in fact they are the same thing.
Varieties are closely related to ideals of the polynomial ring. Given a subset
P ⊆ K[ x̄], we have defined the variety V(P) ⊆ K n . We can also define an
operation in the opposite direction.
Definition 31.3 Given any subset S ⊆ K n , let I(S ) be the set of polynomials
in K[x1 , . . . , xn ] which vanish at all points in S . That is,
I(S ) = { f ( x̄) ∈ K[ x̄] | for all ā ∈ S , f (ā) = 0 } .
Lemma 31.4 For all S , S 1 , S 2 ⊆ K n and all P, P1 , P2 ⊆ K[x1 , . . . , xn ], the
following hold:
(i) I(S ) is an ideal in K[ x̄].
(ii) If S 1 ⊆ S 2 ⊆ K n , then I(S 1 ) ⊇ I(S 2 ).
(iii) If P1 ⊆ P2 ⊆ K[x1 , . . . , xn ], then V(P1 ) ⊇ V(P2 ).
(iv) V(I(S )) ⊇ S .
(v) I(V(P)) ⊇ P.
(vi) I(V(I(S ))) = I(S ).
(vii) V(I(V(P))) = V(P).
(viii) If I is the ideal generated by P, then V(I) = V(P).
The proof is left as an exercise. The characterisation of exactly which ideals
are I(S ) for some S is given by Hilbert’s Nullstellensatz, which is the subject
 31.2 Constructible Sets 169

of Chapter 32. With this relation between varieties and ideals, we can apply
Hilbert’s basis theorem.
Proposition 31.5 Every variety V ⊆ K n is the zero set of a finite set of
polynomials.
Proof Suppose V = V(P). Then, by Lemma 31.4, V = V(I(V(P))). By
Theorem 28.15, there is a finite set { f1 , . . . , fr } of polynomials which generates
I(V(P)). Using the lemma again, V = V( f1 , . . . , fr ). 
The following corollary summarises what we have proved.
Corollary 31.6 Let K be an algebraically closed field and n ∈ N+ . The
varieties V ⊆ K n are exactly the subsets of K n which are defined by positive
quantifier-free formulas (using parameters from K). 

31.2 Constructible Sets

Constructible sets are what geometers call subsets of K n which are finite
Boolean combinations of varieties. So they are exactly the subsets of K n
which are quantifier-free definable, with parameters from K. By quantifier
elimination for algebraically closed fields, they are also exactly the definable
sets. We will give a normal form for them and then explain how they can be
considered as varieties.
By the disjunctive normal form theorem (see Exercise 20.12), every
quantifier-free formula is equivalent to a formula of the form
⎛ s ⎞
r ⎜$
⎜⎜⎜ i $
ti ⎟⎟⎟
⎜⎜⎝ fi j ( x̄) = 0 ∧ gi j ( x̄)  0⎟⎟⎟⎠
i=1 j=1 j=1

for some natural numbers r, si , ti and some polynomials fi j ( x̄) and gi j ( x̄). For
+i
each i, let gi ( x̄) be the product tj=1 gi j ( x̄), let Vi be the variety defined by
 si i
f i j ( x̄) = 0, and let Wi be the variety defined by sj=1 fi j ( x̄) = 0 ∧ gi ( x̄) = 0.
j=1 r
Then the set defined by the formula above is i=1 (Vi  Wi ). Thus we have
proved the following.
Proposition 31.7 Let K be an algebraically closed field and n ∈ N+ . Every
subset of K n which is Lring -definable with parameters from K is a finite union
of sets of the form V  W, where V and W are varieties in K n . 
It is generally much easier to deal with systems of equations than to
deal with negations of equations as well. The Rabinowitsch trick, introduced
170 31 Definable Sets and Varieties

in a one-page paper in 1930 [Rab30], turns the negated equations into

equations, at the cost of introducing new variables. The observation is simply
that the negated equation g(x1 , . . . , xn )  0 is equivalent to the equation
xn+1 g(x1 , . . . , xn ) = 1, because in a solution a1 , . . . , an+1 of this equation,
g(a1 , . . . , an ) = 1/an+1 , which cannot be 0.
Theorem 31.8 Let S ⊆ K n be a constructible set. Then there is r ∈ N and
a variety Y ⊆ K n+r such that the projection p : K n+r → K n onto the first n
coordinates restricts to a bijection from Y to S .
Proof As described above, we may assume that S is defined by a formula of
the form
⎛ s ⎞
r ⎜$
⎜⎜⎜ i ⎟⎟⎟
⎜⎜⎝ fi j (x1 , . . . , xn ) = 0 ∧ gi (x1 , . . . , xn )  0⎟⎟⎟⎠ .
i=1 j=1

Take Y to be defined by
⎛ s ⎞
r ⎜$
⎜⎜⎜ i ⎟⎟⎟
⎜⎝⎜ fi j (x1 , . . . , xn ) = 0 ∧ xn+i gi (x1 , . . . , xn ) − 1 = 0⎟⎟⎟⎠ .
i=1 j=1

This is a positive quantifier-free formula, so by Corollary 31.6, Y is a variety.

It is immediate that p restricts to a bijection Y → S . 
If we do not want the extra free variables, we can note that the projection is
the same thing as existential quantification and get the following corollary.
n
Corollary 31.9 Every definable subset of K  is defined by a formula of the
form ∃xn+1 . . . , xn+s i=1 fi (x1 , . . . , xn+s ) = 0 , for some s ∈ N and some
s

polynomials fi . 

31.3 Chevalley’s Theorem

As an application of quantifier elimination for algebraically closed fields, we
prove a theorem of Chevalley.
Theorem 31.10 (Chevalley’s theorem) Let K be an algebraically closed field,
let S be a constructible subset of K n , and let f : K n → K m be a polynomial
map. Then the image f (S ) ⊆ K m is a constructible set.
Proof The polynomial map f has components f1 , . . . , fm , with each fi ∈
K[x1 , . . . , xn ]. The constructible set S is defined by a quantifier-free Lring (K)
 Exercises 171

formula ϕ(x1 , . . . , xn ). The image f (S ) is defined in the free variables y1 , . . . , ym

by the formula
⎡ ⎤
⎢⎢⎢ $
m ⎥⎥
⎢
∃x1 , . . . , xn ⎢⎣ϕ(x1 , . . . , xn ) ∧ yi = fi (x1 , . . . , xn )⎥⎥⎥⎦ .
i=1

By quantifier elimination for algebraically closed fields, Corollary 30.7, there

is a quantifier-free Lring (K)-formula θ(y1 , . . . , ym ) (with the same parameters)
which also defines f (S ). So f (S ) is constructible. 

Exercises
31.1 Suppose that I1 ⊆ I2 ⊆ I3 ⊆ · · · ⊆ Ir ⊆ · · · is an ascending chain of
ideals of F[ x̄]. Using the Hilbert basis theorem, show there is s ∈ N+
such that for all t > s, It = I s . [We say that F[ x̄] has the ascending
chain condition for ideals.]
31.2 Show that the collection of varieties in K n satisfies the descending
chain condition, that is, if

V1 ⊇ V2 ⊇ V3 ⊇ · · · ⊇ Vr ⊇ · · ·

is a descending chain of varieties then there is s ∈ N+ such that for all

t > s, Vt = V s .
31.3 A variety V ⊆ K n is said to be irreducible if it cannot be written as
V = V1 ∪ V2 for varieties V1 and V2 , both proper subsets of V. Show
that any variety is a finite union of irreducible varieties.
31.4 Show that a variety V ⊆ K n is irreducible if and only if I(V) is a prime
ideal of K[ x̄].
31.5 Let K |= ACF p with p > 0. Show that the map K → K given by
√
x → p x is a definable (single-valued) function. Deduce that if F ⊆ K
is a subfield, then a variety V is definable with parameters from F if
and only if it is defined by polynomials
 with coefficients
 in the perfect
p−∞ 1/pr
closure of F, that is, in F = a | a ∈ F, r ∈ N .
31.6 Let K be an algebraically closed field and F ⊆ K be a subfield. Show
that b ∈ aclK (F) if and only if b lies in a finite subset of K which is
definable with parameters from F.
31.7 The property of the relative algebraic closure from Exercise 31.6 makes
sense in any structure and is called the (model-theoretic) algebraic
closure. Let M be any L-structure, and let A ⊆ M be a subset and
172 31 Definable Sets and Varieties

b ∈ M. We define b ∈ acl(A) iff b is contained in a finite subset of M

which is definable with parameters from A. Show that acl is a closure
operator on M with finite character (see Exercise 30.3).
31.8 Let K be an algebraically closed field, and suppose that S ⊆ K is
definable (possibly with parameters). Show that S is either finite or
cofinite, that is, K  S is finite.
31.9 A complete theory T such that the property of the previous exercise
holds for every model of T is called strongly minimal. Show that if T
is strongly minimal and M |= T , then the closure operator acl on M is
a pregeometry (see Exercise 30.3).
31.10 We outline Ax’s generalisation of the result from Exercise 30.7. This is
also known as the Ax–Grothendieck theorem.
(a) Write down an Lring -sentence ϕn,t,d which states that ‘for any variety
V = V(g1 , . . . , gt ) ⊆ K n and any polynomial map f : K n → K n ,
where the fi and g j have degree at most d, if f restricts to a map
V → V and this map is injective, then it is surjective.’
(b) Use the method of Exercise 30.7 to prove that if K is any
algebraically closed field and n, t, d ∈ N, then K |= ϕn,t,d .
(c) Formulate and prove a similar statement for a constructible set S
and a constructible map f : S → S .
31.11 In this exercise we relate the Stone space of types over parameters to
the Zariski spectrum of a polynomial ring.
Let K be an algebraically closed field, F a subfield, and ā ∈ K n .
Define IF (ā) = { f ( x̄) ∈ F[ x̄] | f (ā) = 0 }.
(a) Show that IF (ā) is a prime ideal and that tp(ā/F) = tp(b̄/F) if and
only if IF (ā) = IF (b̄).
(b) Use this idea to construct a bijection from the Stone space
S n (ACF, F) of complete n-types with parameters from F to the
Zariski spectrum Spec(F[x1 , . . . , xn ]), the set of all prime ideals of
F[x1 , . . . , xn ].
(c) Show that V(IF (ā)) is the smallest variety containing ā which is
defined by polynomials with coefficients from F.
(d) If you know about the topologies on the Stone space and the Zariski
spectrum, prove that this map is continuous but that its inverse is
not.
 32
Hilbert’s Nullstellensatz

In this final chapter we use quantifier elimination to prove Hilbert’s Nullstel-

lensatz (theorem on zero sets), which shows that the ideals corresponding to
algebraic varieties are exactly the radical ideals. We prove the theorem in three
stages to highlight the different techniques used in the proof. First, quantifier
elimination is used to prove a theorem about systems of equations. Then we
show that any ideal can be extended to a maximal ideal and thereby use the
first theorem to prove the weak nullstellensatz. The strong Nullstellensatz can
be proved in a similar way, but because the Rabinowitsch trick is so useful, we
illustrate its use by deducing the strong Nullstellensatz from the weak version.
This was actually Rabinowitsch’s original application of the trick.

32.1 Systems of Polynomial Equations

The definition of algebraically closed fields only requires that each individual
non-constant polynomial in one variable have a zero in the field. The power
of the quantifier elimination theorem is that we can deduce a consequence
for systems of polynomial equations (and negations of equations) in several
variables.
Theorem 32.1 Let K be an algebraically closed field. Then any finite system
of polynomial equations and negations of equations with coefficients from K,
in any number of variables, which has a solution in some field extending K
already has a solution in K.
Proof Let c̄ be a tuple from K consisting of all the coefficients of all the
polynomials in the finite system of equations and inequations. Then we can

173
174 32 Hilbert’s Nullstellensatz

capture the system with an Lring -formula ϕ( x̄, c̄) where ϕ( x̄, ȳ) has the form

$
r $
s
fi ( x̄, ȳ) = 0 ∧ gi ( x̄, ȳ)  0
i=1 i=1

and the fi and gi are polynomials in Z[ x̄, ȳ].

Suppose that B is a field extension of K such that B |= ∃ x̄ϕ( x̄, c̄). Then
Balg |= ∃ x̄ϕ( x̄, c̄) by Proposition 5.6.
Since ACF has quantifier elimination, there is a quantifier-free formula θ(ȳ)
such that ACF  ∀ȳ[∃ x̄ϕ( x̄, ȳ) ↔ θ(ȳ)].
So Balg |= θ(c̄) and, since θ is a quantifier-free formula and K is a
substructure of Balg , by Lemma 5.5, K |= θ(c̄). Then, since K |= ACF, we
have K |= ∃ x̄ϕ( x̄, c̄). 
This theorem says that to consider systems of polynomial equations, we
do not have to look beyond an algebraically closed K to see if solutions are
possible. However, it does not say explicitly which systems of equations do
have solutions. For that we need Hilbert’s Nullstellensatz.

32.2 The Weak Nullstellensatz

Theorem 32.2 (The weak Nullstellensatz) Let K be an algebraically closed
field, and let I ⊆ K[ x̄] be an ideal. Then I is a proper ideal if and only if
V(I)  ∅.
Proof First note that I is a proper ideal if and only if it does not contain
the constant polynomial 1. Since 1 does not vanish anywhere, if 1 ∈ I, then
V(I) = ∅.
Now suppose that I is a proper ideal. First we extend I to a maximal proper
ideal J in K[ x̄]. To do this, index all the polynomials in K[ x̄] as ( fα )α<λ for
some ordinal λ. If K is countable, we can just index by natural numbers. Then
set I0 = I, and for each α in turn, set Iα+1 to be the ideal generated by Iα and
fα if this ideal is proper, and set Iα+1 = Iα otherwise. For a limit ordinal β, take

Iβ = α<β Iα . Then we can take J to be Iλ .
Let B = K[ x̄]/J, which is a field extension of K because J is a maximal
ideal. Let b̄ be the image in B of x̄ under the quotient map. Then, for each
f ( x̄) ∈ J, so in particular for each f ( x̄) ∈ I, we have f (b̄) = 0.
Using the Hilbert basis theorem, Theorem 28.15, there is a finite set

f1 , . . . , fr of polynomials which generates I. So B |= ∃ x̄ ri=1 fi ( x̄) = 0.
 32.3 The Strong Nullstellensatz 175


By Theorem 32.1, K |= ∃ x̄ ri=1 fi ( x̄) = 0. Let ā ∈ K n be a witness, so
ā ∈ V(I). So V(I)  ∅, as required. 

32.3 The Strong Nullstellensatz

Example 32.3 Let f (x1 , x2 ) = (x1 − 6) and let g(x1 , x2 ) = f (x1 , x2 )2 . Then
V( f ) = V(g) = {(6, x2 ) | x2 ∈ K }, but the ideals generated by f and g are
different.
Definition 32.4 An ideal I of a ring R is called a radical ideal if, for all
m ∈ N+ , for all√ f ∈ R, if f m ∈ I, then f ∈ I. The radical of an ideal I of a ring
R is the ideal I = { f ∈ R | for some m ∈ N+ , f m ∈ I }.
,√ -
Lemma 32.5 For any ideal I ⊆ K[ x̄] we have V I = V(I).
Proof Suppose ā ∈ K n and f m (ā) = 0, that is, f (ā)m = 0. Then, since K is a
domain, f (ā) = 0. 
So to understand varieties, we can restrict our attention to radical ideals.
The next theorem, Hilbert’s Nullstellensatz, shows that different radical ideals
do correspond to different varieties. For this it is essential that the field K be
algebraically closed. For example, in the real field R, let I1 = R[x] as an ideal
of itself, and let I2 ⊆ R[x] be the ideal generated by x2 + 1. Then I1 and I2 are
radical, but V(I1 ) = V(I2 ) = ∅. Taking the complex field C in place of R, we
have V(I1 ) = ∅ and V(I2 ) = {±i}.
Theorem 32.6 (Hilbert’s Nullstellensatz) Suppose that K is an algebraically
closed field and I ⊆ K[ x̄] is a radical ideal. Then I(V(I)) = I.
Proof We will show that if I is any ideal of K[x1 , . . . , xn ] and √ f ∈ I(V(I)),
then there is m ∈ N+ such that f m ∈ I. So, in particular, if I = I, then f ∈ I.
Using Hilbert’s basis theorem, Theorem 28.15, choose a generating set
f1 , . . . , fr for I. Then f ∈ I(V(I)) means that for all ā ∈ K n , if fi (ā) = 0
for each i = 1, . . . , r, then f (ā) = 0. Using the Rabinowitsch trick, we consider
f and the fi as polynomials in x1 , . . . , xn+1 . Then f1 , . . . , fr and (1− xn+1 f ) have
no common zeros in K n+1 . So, by the weak Nullstellensatz, Theorem 32.2, the
ideal of K[x1 , . . . , xn+1 ] generated by f1 , . . . , fr and (1 − xn+1 f ) is not proper.
So by Lemma 28.5, there are g0 , g1 , . . . , gr ∈ K[x1 , . . . , xn+1 ] such that the
equation

r
1 = g0 (1 − xn+1 f (x1 , . . . , xn )) + gi (x1 , . . . , xn+1 ) fi (x1 , . . . , xn )
i=1
176 32 Hilbert’s Nullstellensatz

holds in the polynomial ring K[x1 , . . . , xn+1 ].

We cancel the first term by substituting 1/ f (x1 , . . . , xn ) for the new variable
xn+1 to get an equation

r
1= gi (x1 , . . . , xn , 1/ f (x1 , . . . , xn )) fi (x1 , . . . , xn ) (32.1)
i=1

in the field K(x1 , . . . , xn ) of rational functions. Let m be the maximum

degree in the variable xn+1 of the polyomials g1 , . . . , gr . Then we can
,...,xn )
write each gi (x1 , . . . , xn , 1/ f (x1 , . . . , xn )) as fh(xi (x11,...,x n)
m for some polynomials

hi (x1 , . . . , xn ) ∈ K[x1 , . . . , xn ]. Clearing denominators, Equation (32.1)

becomes

r
f (x1 , . . . , xn )m = hi (x1 , . . . , xn ) fi (x1 , . . . , xn ))
i=1

in K[x1 , . . . , xn ], which by Lemma 28.5 says that f m ∈ I, as required. 

The model theory of fields is now a large subject, and we have just scratched
the surface. Introductions to various aspects of it can be found in [MMP06],
[HV02], [Bou98], and [HHM08], as well as in the books mentioned earlier.

Exercises
32.1 Let K be an algebraically closed field, and let I ⊆ K[ x̄] be an ideal.
Show that I = I(V(I)) if and only if I is a radical ideal. Give an example
to show that this fails if K is not algebraically closed.
32.2 Show that an ideal I ⊆ K alg [ x̄] is maximal if and only if it is generated
by polynomials of the form x1 − a1 , . . . , xn − an . Give an example to show
that this fails if K is not algebraically closed.
32.3 Give a complete proof of Hilbert’s Nullstellensatz along the lines of the
proof of Theorem 32.2, not using the Rabinowitsch trick.
32.4 Give a sketch proof of Hilbert’s Nullstellensatz. Explain what all the key
ideas in the proof are, how they fit together, and how they depend on
ideas from earlier in the book.
 Bibliography

[Ax69] James Ax. Injective endomorphisms of varieties and schemes. Pacific J.

Math., 31:1–7, 1969.
[Bal88] John T. Baldwin. Fundamentals of Stability Theory. Perspectives in Mathe-
matical Logic. Springer, Berlin, 1988.
[BBR62] Andrzej Białynicki-Birula and Maxwell Rosenlicht. Injective morphisms of
real algebraic varieties. Proc. Amer. Math. Soc., 13:200–203, 1962.
[BCR98] Jacek Bochnak, Michel Coste, and Marie-Françoise Roy. Real Algebraic
Geometry, volume 36 of Ergebnisse der Mathematik und ihrer Grenzgebiete
(3) [Results in Mathematics and Related Areas (3)]. Springer, Berlin, 1998.
Translated from the 1987 French original, revised by the authors.
[Bou98] E. Bouscaren, editor. Model Theory and Algebraic Geometry, volume 1696
of Lecture Notes in Mathematics. Springer, Berlin, 1998.
[Bri77] Jane Bridge. Beginning Model Theory. Oxford Logic Guides. Clarendon
Press, Oxford, 1977.
[CK90] C. C. Chang and H. J. Keisler. Model Theory, volume 73 of Studies in
Logic and the Foundations of Mathematics. North-Holland, Amsterdam,
third edition, 1990.
[CL00] René Cori and Daniel Lascar. Mathematical Logic. Oxford University Press,
Oxford, 2000. A course with exercises. Part I, Propositional calculus,
Boolean algebras, predicate calculus, translated from the 1993 French
original by Donald H. Pelletier, with a foreword to the original French
edition by Jean-Louis Krivine and a foreword to the English edition by
Wilfrid Hodges.
[Hal74] Paul R. Halmos. Naive Set Theory. Undergraduate Texts in Mathematics.
Springer, New York, 1974. Reprint of the 1960 edition.
[HHM08] Deirdre Haskell, Ehud Hrushovski, and Dugald Macpherson. Stable Domi-
nation and Independence in Algebraically Closed Valued Fields, volume 30
of Lecture Notes in Logic. Association for Symbolic Logic, Chicago, IL,
2008.
[Hod93] Wilfrid Hodges. Model Theory, volume 42 of Encyclopedia of Mathematics
and Its Applications. Cambridge University Press, Cambridge, 1993.

177
178 Bibliography

[Hod97] Wilfrid Hodges. A Shorter Model Theory. Cambridge University Press,

Cambridge, 1997.
[HV02] Bradd Hart and Matthew Valeriote, editors. Lectures on Algebraic Model
Theory, volume 15 of Fields Institute Monographs. American Mathematical
Society, Providence, RI, 2002.
[Jec03] Thomas Jech. Set Theory. Springer Monographs in Mathematics. Springer,
Berlin, Third millennium edition, revised and expanded. 2003.
[Kun11] Kenneth Kunen. Set Theory, volume 34 of Studies in Logic (London).
College Publications, London, 2011.
[Lan02] Serge Lang. Algebra, volume 211 of Graduate Texts in Mathematics.
Springer, New York, third edition, 2002.
[Mar02] David Marker. Model Theory: An Introduction, volume 217 of Graduate
Texts in Mathematics. Springer, New York, 2002.
[MMP06] David Marker, Margit Messmer, and Anand Pillay. Model Theory of Fields,
volume 5 of Lecture Notes in Logic. Association for Symbolic Logic, La
Jolla, CA, second edition, 2006.
[Pil83] Anand Pillay. An Introduction to Stability Theory, volume 8 of Oxford Logic
Guides. Clarendon Press, Oxford, 1983.
[Pil96] Anand Pillay. Geometric Stability Theory, volume 32 of Oxford Logic
Guides. Clarendon Press, Oxford, 1996.
[Poi00] Bruno Poizat. A Course in Model Theory. Universitext. Springer, New York,
2000. An introduction to contemporary mathematical logic, translated from
the French by Moses Klein and revised by the author.
[Rab30] J. L. Rabinowitsch. Zum Hilbertschen Nullstellensatz. Math. Ann.,
102(1):520, 1930.
[Rot00] Philipp Rothmaler. Introduction to Model Theory, volume 15 of Algebra,
Logic and Applications. Gordon and Breach, Amsterdam, 2000. Prepared
by Frank Reitmaier, translated and revised from the 1995 German original
by the author.
[Sac10] Gerald E. Sacks. Saturated Model Theory. World Scientific, Hackensack,
NJ, second edition, 2010.
[She90] Saharon Shelah. Classification Theory and the Number of Nonisomorphic
Models, volume 92 of Studies in Logic and the Foundations of Mathematics.
North-Holland, Amsterdam, second edition, 1990.
[TZ12] Katrin Tent and Martin Ziegler. A Course in Model Theory, volume 40 of
Lecture Notes in Logic. Association for Symbolic Logic, La Jolla, CA, 2012.
[Vää11] Jouko Väänänen. Models and Games, volume 132 of Cambridge Studies in
Advanced Mathematics. Cambridge University Press, Cambridge, 2011.
[vdD98] Lou van den Dries. Tame Topology and o-Minimal Structures, volume 248
of London Mathematical Society Lecture Note Series. Cambridge University
Press, Cambridge, 1998.
 Index

algebraic (element of a field), 158 Boolean algebra, 106, 109

algebraic closure Boolean ring, 108
model-theoretic, 172
canonical model, 59, 134
of a field, 160
Cantor’s theorem, 54
relative, 159
cardinal, 53
algebraically closed field, 38, 160
cardinal arithmetic, 54
algebraically independent, 163
strongly inaccessible, 149
Archimedean ordered field, 39, 74, 75
cardinality, 53
arithmetic, 34
of a language, 55
arity, 3
of a vector space, 80
ascending chain condition, 172
categorical, 81
atom (of a Boolean algebra), 106, 116
atomic Boolean algebra, 107 κ-categorical, 81
atomic formula, 13 categoricity
atomic structure, 142 countable categoricity, 81, 138, 143
atomless Boolean algebra, 108 of ACF p , 164
automorphism, 6, 96, 146, 151 of atomless Boolean algebras, 108
axiom of extension, 106 of DLO, 84
axiomatisable class, 47, 76 of DTFAG, 82
axioms, 32 of the successor function, 89
for abelian groups, 33 of vector spaces, 81
for algebraically closed fields, 38 cell decomposition theorem, 125
for divisible torsion-free abelian groups, 51 cells, 124
for fields, 33 characteristic of a field, 38, 51, 156
for groups, 32 Chevalley’s theorem, 170
for linear orders, 83 closed term, 10, 58
for ordered fields, 38 cofinite subset, 90
for real-closed fields, 40 compactness theorem, 42, 71, 76, 90, 130,
for rings, 33 160, 162
for the successor function, 88 proof of, 43, 57–61
for vector spaces, 77 strong version, 61
complete Boolean algebra, 108
back-and-forth method, 85, 96, 108, 139, complete diagram, 70
142–144, 146, 147, 162 complete ordered field, 39
basis of a vector space, 78 complete theory, 35, 38, 40, 57, 81, 101,
binary, 4 147–150

179
180 Index

complete type, see type field of fractions, 155

completeness finite
of ACF p , 165 non-standard real number, 75
of DLO, 85 set, 54
of the theory of discrete linear orders, 150 finitely axiomatisable, 47, 49
of the theory of real-closed fields, 102 finitely satisfiable, 42, 57, 133
of the theory of the successor function, 89 set of formulas, 129
complex field, 37, 68, 160, 176 first-order logic (comparison with
conjunctive normal form, 113, 167 second-order), 40
connected graph, 52 formula, 13
constant symbols, 3 atomic, 13
constructible set, 169, 170 existential, 26, 170
countable, 54 positive, 28
positive quantifier-free, 167, 169,
deductively closed, 32, 57
170
definable function, 19
quantifier-free, 17, 25
definable sets, 19, 45, 90, 109–126, 133, 150
universal, 26
∅-definable, 117
free variables, 14
definable with parameters, 117
function symbols, 3
in Nsucc , 90
fundamental theorem of algebra, 160
in Ro-ring , 119
in algebraically closed fields, 167 Gödel’s first incompleteness theorem, 35
in dense linear orders, 110 graph, 51
in real vector spaces, 115
Henkin theory, 58, 134
in vector spaces over finite fields, 111
Hilbert’s basis theorem, 158, 169, 172, 175,
dense linear order, 84, 95–98, 110, 150
176
descending chain condition, 172
Hilbert’s Nullstellensatz, 174
diagram, 71, 100, 160, 164
homogeneous, 146, 151
complete diagram, 70
homomorphism, 8
dimension of a vector space, 79
of rings, 156
discrete linear order, 84, 149
disjunctive normal form, 113, 169 ideal of a Boolean algebra, 108
distributive lattices, 105 ideal of a ring, 156, 168
distributive laws, 105 maximal, 156
divisible (group), 51 prime, 156
DLO, see dense linear order proper, 156, 175
domain (integral domain), 155 radical, 176
domain (of a structure), 3 infinite
Downward Löwenheim–Skolem theorem, 67 non-standard real number, 75
DTFAG, 51, 82 set, 54
infinitesimal, 45, 75
elementary embedding, 66, 70
infix notation, 10
elementary equivalence, 35
integral domain, 155
elementary extension, 65, 70, 130
interpretation
elementary substructure, 65, 70
of formulas in a structure, 14
embedding, 6, 15, 65
of symbols in a structure, 3
endpoints, 84
of terms in a structure, 10
entails, 31
irreducible, 157
expansion
irreducible variety, 172
of a language, 4
isomorphism, 6
of a structure, 6
extension, 6, 24, 65, 100 kernel, 156
 Index 181

L-structure, 3 of terms, 11
language, 3 of types, 130
lattice, 105 of universal formulas, 27
least upper bound, 105 prime model, 142
lexicographic product, 87, 149 prime subfield, 156
Lindenbaum algebra, 110, 116, 139 principal DLO formula, 95, 110
Lindenbaum’s lemma, 58 principal formula, 110, 115
linear order, 83, 148 principal ideal domain, 157
linearly independent, 78, 133, 136 principal type, 142
logical consequence, 31
quantifier elimination, 97, 100, 117, 150
Łos–Vaught test, 81, 85, 101, 147, 149, 165
for Nsucc , 103
Mersenne prime, 75 for Q< , 97
method of diagrams, 70, 72, 76, 160 for T =∞ , 98
method of new constants, 43–45, 71, 130, 134 for ACF, 165, 169, 172, 174
method of preservation under automorphisms, for atomless Boolean algebras, 108
21, 90, 97, 130, 146, 150 for DLO, 98, 150
minimal polynomial, 157, 161 for real-closed fields, 102, 132, 136
model, 15 for structures, 97
model completeness, 103 for theories, 97
for vector spaces, 101
n-type, 129 quotient ring, 156
Nash function, 122
piecewise Nash function, 123 Rabinowitsch trick, 169, 176
non-standard rational functions, 156
analysis, 75 real algebraic number, 136
model, 72 real field, 38, 45, 74, 176
natural number, 72, 131 real-closed field, 40, 102, 132
real number, 75 realise a type, 130
realising types, 139
o-minimal, 120 recursive definition
oligomorphic permutation group, 141 of existential formulas, 26
omitting types, 130, 139 of formulas, 13
omitting types theorem, 134 of quantifier-free formulas, 25
of Vaught, 136 of terms, 9
Order Property, 148 of universal formulas, 26
parameters, 117, 144 reduct, 6
parametrically definable, see definable with relation symbols, 3
parameters relative algebraic closure, 159
partial orders, 104 ring, 155
Ryll–Nardzewski theorem, 139, 143, 144
partial type, 134
Peano arithmetic, 35 satisfiable, 32
polynomials, 155, 167 formula with respect to a theory, 110
power set, 104 saturated, 147
prefix notation, 10 ℵ0 -saturated, 144
pregeometry, 166 κ-saturated, 147
preservation Schröder–Bernstein theorem, 53
of atomic formulas, 15 second-order logic, 37
of definable sets, 21 second-order logic (comparison with
of existential formulas, 26 first-order), 40
of formulas, 15 semantics, 57
of quantifier-free formulas, 18, 26 semi-algebraic sets, 21, 119
182 Index

sentence, 14 transfinite induction, 161, 162

signature, 3 twin prime conjecture, 73
simple extension of a field, 157 type, 129, 133, 150
Skolem functions, 68 complete type, 129, 134, 148
Skolem paradox, 67 in Ns-ring , 131
small theory, 144 in Ro-ring , 136
spanning set, 78 in DLO, 132
stability theory, 134, 148 in vector spaces, 136, 141
stable, 148 non-principal type, 134
κ-stable, 148 of a tuple in a model, 129
0-stable, 144 partial type, 134
Stone space, 138 principal type, 134, 138
strongly ℵ0 -homogeneous, 146 type over parameters, 144
strongly κ-homogeneous, 151 type-definable sets, 134
strongly minimal, 173
unary, 4
structure, 3
uncountable, 54
sub-language, 4
universal axiomatisation, 76
substructure, 6, 24, 65, 99
universal model, 143
substructure completeness, 100, 165
universal sentences, 155
syntax, 57
unstable, 148
Tarski–Vaught test, 66, 67 Upward Löwenheim–Skolem theorem, 72, 89,
term, 9 143
closed, 10
variety, 168, 169
theory, 32
irreducible, 172
first-order, 32
vector space, 77, 101, 111, 133
of a class of structures, 31
vocabulary, 3
torsion (in groups), 48
transcendence base, 163 witness property, 58
transcendence degree, 164
Zariski-closed subset, 168
transcendental, 136, 157, 158