Achim Blumensath blumensath@mathematik.tu-darmstadt.de is ...
Achim Blumensath blumensath@mathematik.tu-darmstadt.de is ...
Achim Blumensath blumensath@mathematik.tu-darmstadt.de is ...
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
<strong>Achim</strong> <strong>Blumensath</strong><br />
<strong>blumensath@mathematik</strong>.<strong>tu</strong>-<strong>darmstadt</strong>.<strong>de</strong><br />
�<strong>is</strong> documentwas las<strong>tu</strong>pdated2012-08-08.<br />
�e latestversion can be found at<br />
www.mathematik.<strong>tu</strong>-<strong>darmstadt</strong>.<strong>de</strong>/~blumensath<br />
Copyright2012 <strong>Achim</strong> <strong>Blumensath</strong><br />
Allrights arereserved.Perm<strong>is</strong>sion <strong>is</strong> grantedto d<strong>is</strong>tribute and copyth<strong>is</strong><br />
documen<strong>tu</strong>n<strong>de</strong>rthe followingterms:<br />
◆ �euse <strong>is</strong>private and non-commercial. Inparticular,youare not<br />
allowed tosellprinted copies ofth<strong>is</strong> document.<br />
◆ All changes,additions,and om<strong>is</strong>sionstothe document are clearly<br />
marked assuch.<br />
◆ All excerpts, quotes, and translations contain an acknowledgement<br />
ofthe original.
Contents<br />
A. Set�eory 1<br />
a1 Basic set theory 3<br />
1 Sets and classes. . . . . . . . . . . . . . . . . . . . . . . . . . . . 3<br />
2 Stages and h<strong>is</strong>tories . . . . . . . . . . . . . . . . . . . . . . . . . 11<br />
3 �e cumulative hierarchy . . . . . . . . . . . . . . . . . . . . . . 18<br />
a2 Relations 27<br />
1 Relations and functions . . . . . . . . . . . . . . . . . . . . . . . 27<br />
2 Products andunions . . . . . . . . . . . . . . . . . . . . . . . . 36<br />
3 Graphs andpartial or<strong>de</strong>rs . . . . . . . . . . . . . . . . . . . . . 39<br />
4 Fixedpoints and closure operators . . . . . . . . . . . . . . . . 47<br />
a3 Ordinals 57<br />
1 Well-or<strong>de</strong>rs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57<br />
2 Ordinals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64<br />
3 Induction and fixed points . . . . . . . . . . . . . . . . . . . . . 74<br />
4 Ordinal arithmetic . . . . . . . . . . . . . . . . . . . . . . . . . . 85<br />
a4 Zermelo-Fraenkel set theory 105<br />
1 �e Axiom of Choice . . . . . . . . . . . . . . . . . . . . . . . . 105<br />
2 Cardinals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113<br />
3 Cardinal arithmetic . . . . . . . . . . . . . . . . . . . . . . . . . 116<br />
4 Cofinality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122<br />
logic, algebra & geometry2012-08-08 — ©achim blumensath v
Contents<br />
5 �e Axiom ofReplacement. . . . . . . . . . . . . . . . . . . . . 131<br />
6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134<br />
B. Universal Algebra 137<br />
b1 Struc<strong>tu</strong>res and homomorph<strong>is</strong>ms 139<br />
1 Struc<strong>tu</strong>res . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139<br />
2 Homomorph<strong>is</strong>ms . . . . . . . . . . . . . . . . . . . . . . . . . . 146<br />
3 Categories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152<br />
4 Congruences andquotients . . . . . . . . . . . . . . . . . . . . 160<br />
b2 Trees and lattices 173<br />
1 Trees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173<br />
2 Lattices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181<br />
3 I<strong>de</strong>als and filters . . . . . . . . . . . . . . . . . . . . . . . . . . . 189<br />
4 Prime i<strong>de</strong>als andultrafilters . . . . . . . . . . . . . . . . . . . . 194<br />
5 Atomic lattices andpartition rank . . . . . . . . . . . . . . . . . 200<br />
b3 Algebraicconstructions 213<br />
1 Terms andterm algebras . . . . . . . . . . . . . . . . . . . . . . 213<br />
2 Direct andreducedproducts . . . . . . . . . . . . . . . . . . . . 224<br />
3 Direct and inverse limits . . . . . . . . . . . . . . . . . . . . . . 232<br />
b4 Topology 243<br />
1 Open and closedsets . . . . . . . . . . . . . . . . . . . . . . . . 243<br />
2 Continuous functions . . . . . . . . . . . . . . . . . . . . . . . . 248<br />
3 Hausdorffspaces and compactness . . . . . . . . . . . . . . . . 252<br />
4 �eProducttopology . . . . . . . . . . . . . . . . . . . . . . . . 259<br />
5 Densesets and <strong>is</strong>olatedpoints . . . . . . . . . . . . . . . . . . . 263<br />
6 Spectra andStone duality . . . . . . . . . . . . . . . . . . . . . . 272<br />
7 Stonespaces and Cantor-Bendixsonrank . . . . . . . . . . . . 279<br />
vi
Contents<br />
b5 ClassicalAlgebra 287<br />
1 Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287<br />
2 Group actions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291<br />
3 Rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 300<br />
4 Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 306<br />
5 Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314<br />
6 Or<strong>de</strong>red fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329<br />
C. First-Or<strong>de</strong>r Logic and its Extensions 343<br />
c1 First-or<strong>de</strong>r logic 345<br />
1 Infinitary first-or<strong>de</strong>r logic . . . . . . . . . . . . . . . . . . . . . 345<br />
2 Axiomat<strong>is</strong>ations . . . . . . . . . . . . . . . . . . . . . . . . . . . 356<br />
3 �eories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362<br />
4 Normal forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367<br />
5 Translations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374<br />
6 Extensions of first-or<strong>de</strong>r logic . . . . . . . . . . . . . . . . . . . 383<br />
c2 Elementarysubstruc<strong>tu</strong>res an<strong>de</strong>mbeddings 395<br />
1 Homomorph<strong>is</strong>ms and embeddings . . . . . . . . . . . . . . . . 395<br />
2 Elementary embeddings . . . . . . . . . . . . . . . . . . . . . . 400<br />
3 �e�eorem of Löwenheim andSkolem . . . . . . . . . . . . . 405<br />
4 �e Compactness �eorem . . . . . . . . . . . . . . . . . . . . 412<br />
5 Amalgamation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 422<br />
c3 Typesandtypespaces 427<br />
1 Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427<br />
2 Typespaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 433<br />
3 Retracts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 446<br />
4 Localtypespaces . . . . . . . . . . . . . . . . . . . . . . . . . . 456<br />
5 Stabletheories . . . . . . . . . . . . . . . . . . . . . . . . . . . . 461<br />
vii
Contents<br />
c4 Back-and-forthequivalence 473<br />
1 Partial <strong>is</strong>omorph<strong>is</strong>ms . . . . . . . . . . . . . . . . . . . . . . . . 473<br />
2 Hintikka formulae . . . . . . . . . . . . . . . . . . . . . . . . . . 482<br />
3 Ehrenfeucht-Fraïssé games . . . . . . . . . . . . . . . . . . . . . 485<br />
4 κ-complete back-and-forthsystems . . . . . . . . . . . . . . . . 494<br />
5 �etheorems of Hanf and Gaifman . . . . . . . . . . . . . . . . 501<br />
c5 General mo<strong>de</strong>ltheory 509<br />
1 Classifying logical systems . . . . . . . . . . . . . . . . . . . . . 509<br />
2 Hanf and Löwenheim numbers . . . . . . . . . . . . . . . . . . 513<br />
3 �e�eorem of Lindström . . . . . . . . . . . . . . . . . . . . . 520<br />
4 Projective classes. . . . . . . . . . . . . . . . . . . . . . . . . . . 532<br />
5 Interpolation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 542<br />
6 Fixed-point logics . . . . . . . . . . . . . . . . . . . . . . . . . . 553<br />
D. Axiomat<strong>is</strong>ation and Definability 581<br />
d1 Quantifierelimination 583<br />
1 Preservationtheorems . . . . . . . . . . . . . . . . . . . . . . . 583<br />
2 Quantifier elimination . . . . . . . . . . . . . . . . . . . . . . . 587<br />
3 Ex<strong>is</strong>tentially closedstruc<strong>tu</strong>res . . . . . . . . . . . . . . . . . . . 597<br />
4 Abelian groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . 603<br />
5 Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 608<br />
d2 Productsandvarieties 615<br />
1 Ultraproducts . . . . . . . . . . . . . . . . . . . . . . . . . . . . 615<br />
2 �etheorem of Ke<strong>is</strong>ler andShelah . . . . . . . . . . . . . . . . 620<br />
3 Reducedproducts and Horn formulae . . . . . . . . . . . . . . 633<br />
4 Quasivarieties . . . . . . . . . . . . . . . . . . . . . . . . . . . . 638<br />
5 �e�eorem of Feferman and Vaught . . . . . . . . . . . . . . 650<br />
viii
Contents<br />
d3 O-minimalstruc<strong>tu</strong>res 655<br />
1 Or<strong>de</strong>redtopologicalstruc<strong>tu</strong>res . . . . . . . . . . . . . . . . . . 655<br />
2 O-minimal groups andrings . . . . . . . . . . . . . . . . . . . . 661<br />
3 Cell <strong>de</strong>compositions . . . . . . . . . . . . . . . . . . . . . . . . . 663<br />
E. Classical Mo<strong>de</strong>l �eory 683<br />
e1 Sa<strong>tu</strong>ration 685<br />
1 Homogeneousstruc<strong>tu</strong>res . . . . . . . . . . . . . . . . . . . . . . 685<br />
2 Sa<strong>tu</strong>ratedstruc<strong>tu</strong>res . . . . . . . . . . . . . . . . . . . . . . . . . 690<br />
3 Projectively sa<strong>tu</strong>ratedstruc<strong>tu</strong>res . . . . . . . . . . . . . . . . . . 701<br />
4 Pseudo-sa<strong>tu</strong>ratedstruc<strong>tu</strong>res . . . . . . . . . . . . . . . . . . . . 705<br />
5 Definability inprojectively sa<strong>tu</strong>rated mo<strong>de</strong>ls . . . . . . . . . . 712<br />
e2 Prime mo<strong>de</strong>ls 721<br />
1 Isolatedtypes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 721<br />
2 �e Omitting Types�eorem . . . . . . . . . . . . . . . . . . . 723<br />
3 Prime and atomic mo<strong>de</strong>ls . . . . . . . . . . . . . . . . . . . . . 731<br />
4 Constructed mo<strong>de</strong>ls . . . . . . . . . . . . . . . . . . . . . . . . . 735<br />
e3ℵ0-categoricaltheories 743<br />
1 ℵ0-categorical theories and automorph<strong>is</strong>ms . . . . . . . . . . . 743<br />
2 Fraïssé limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 759<br />
3 Zero-one laws . . . . . . . . . . . . . . . . . . . . . . . . . . . . 765<br />
e4 Ind<strong>is</strong>cerniblesequences 773<br />
1 Ramsey�eory . . . . . . . . . . . . . . . . . . . . . . . . . . . 773<br />
2 Ramsey�eory fortrees . . . . . . . . . . . . . . . . . . . . . . 777<br />
3 Ind<strong>is</strong>cerniblesequences . . . . . . . . . . . . . . . . . . . . . . . 789<br />
4 �e in<strong>de</strong>pen<strong>de</strong>nce andstrict or<strong>de</strong>rproperties . . . . . . . . . . 799<br />
ix
Contents<br />
e5 Functorsan<strong>de</strong>mbeddings 809<br />
1 Local functors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 809<br />
2 Word constructions . . . . . . . . . . . . . . . . . . . . . . . . . 816<br />
3 Ehrenfeucht-Mostowski mo<strong>de</strong>ls . . . . . . . . . . . . . . . . . . 825<br />
e6 Abstract elementaryclasses 839<br />
1 Abstract elementary classes . . . . . . . . . . . . . . . . . . . . 839<br />
2 Amalgamation andsa<strong>tu</strong>ration . . . . . . . . . . . . . . . . . . . 848<br />
3 Limits of chains . . . . . . . . . . . . . . . . . . . . . . . . . . . 861<br />
4 Categoricity andstability . . . . . . . . . . . . . . . . . . . . . . 865<br />
F. Geometric Mo<strong>de</strong>l �eory 873<br />
f1 Geometries 875<br />
1 Depen<strong>de</strong>ncerelations . . . . . . . . . . . . . . . . . . . . . . . . 875<br />
2 Matroids and geometries . . . . . . . . . . . . . . . . . . . . . . 880<br />
3 Modular geometries . . . . . . . . . . . . . . . . . . . . . . . . . 887<br />
4 Strongly minimalsets . . . . . . . . . . . . . . . . . . . . . . . . 893<br />
5 Vaughtian pairs andthe�eorem of Morley . . . . . . . . . . . 901<br />
f2 Ranksand forking 913<br />
1 Morleyrank and ∆-rank . . . . . . . . . . . . . . . . . . . . . . 913<br />
2 In<strong>de</strong>pen<strong>de</strong>ncerelations . . . . . . . . . . . . . . . . . . . . . . . 927<br />
3 Preforkingrelations . . . . . . . . . . . . . . . . . . . . . . . . . 940<br />
4 Forkingrelations . . . . . . . . . . . . . . . . . . . . . . . . . . . 953<br />
f3 Simpletheories 965<br />
1 Dividing and forking . . . . . . . . . . . . . . . . . . . . . . . . 965<br />
2 Simpletheories andthetreeproperty . . . . . . . . . . . . . . . 975<br />
3 �e In<strong>de</strong>pen<strong>de</strong>nce�eorem . . . . . . . . . . . . . . . . . . . . 992<br />
x
Contents<br />
Litera<strong>tu</strong>re 1007<br />
Symbol In<strong>de</strong>x 1009<br />
In<strong>de</strong>x 1019<br />
xi
PartA.<br />
Set�eory
a1. Basicsettheory<br />
1. Setsandclasses<br />
In mathematicsthere are basically two waysto <strong>de</strong>finethe objectsun<strong>de</strong>r<br />
consi<strong>de</strong>ration. Onthe one hand, one can explicitly constructthem from<br />
already known objects. For instance, therational numbers andthereal<br />
numbers areusually introduced inth<strong>is</strong> way. Onthe other hand, one can<br />
takethe axiomatic approach,that <strong>is</strong>, one compiles a l<strong>is</strong>t of <strong>de</strong>siredproperties<br />
and one investigates any object meetingtheserequirements.Some<br />
well known examples are groups, fields, vector spaces, and topological<br />
spaces.<br />
Since set theory <strong>is</strong> meant as foundation of mathematics there are<br />
no more basic objects available in terms of which we could <strong>de</strong>finesets.<br />
�erefore, we will followthe axiomatic approach. We will present a l<strong>is</strong>t<br />
ofsix axioms and any objectsat<strong>is</strong>fying all ofthem will be called a mo<strong>de</strong>l<br />
of set theory. Such a mo<strong>de</strong>l cons<strong>is</strong>ts of two parts: (1) a collection S of<br />
objectsthat we will callsets, and(2)some method which, giventwosets<br />
a and b,tellsus whether a <strong>is</strong>anelement of b.<br />
We will not care what exactlythe objects inSare or howth<strong>is</strong> method<br />
looks like. For example, one could imagine a mo<strong>de</strong>l of set theory cons<strong>is</strong>ting<br />
of na<strong>tu</strong>ral numbers. If we <strong>de</strong>fine that a na<strong>tu</strong>ral number a <strong>is</strong> an<br />
element of the na<strong>tu</strong>ral number b if and only if the a-th bit in the binary<br />
encoding ofb<strong>is</strong> 1,then all but one of our axioms will besat<strong>is</strong>fied. It<br />
<strong>is</strong> conceivable that asimilar but more involved <strong>de</strong>finition might yield a<br />
mo<strong>de</strong>lthatsat<strong>is</strong>fies all ofthem.<br />
We will introduce our axioms in a stepw<strong>is</strong>e fashion during the following<br />
sections. To help rea<strong>de</strong>rs trying to look up a certain axiom we<br />
logic, algebra & geometry2012-08-08 — ©achim blumensath 3
a1. Basicsettheory<br />
inclu<strong>de</strong> a complete l<strong>is</strong>t below even if most ofthe nee<strong>de</strong>d <strong>de</strong>finitions are<br />
still m<strong>is</strong>sing.<br />
Axiom ofExtensionality. Two sets a and b are equal if, and only if, we<br />
have x∈ a⇔x∈ b, for allsets x.<br />
Axiom ofSeparation. If a <strong>is</strong> aset and φ apropertythen{x∈a∣φ} <strong>is</strong><br />
aset.<br />
Axiom ofCreation. For everyset athere <strong>is</strong> asetS suchthat S <strong>is</strong> astage<br />
and a∈S.<br />
Axiom of Infinity. �ere ex<strong>is</strong>ts asetthat <strong>is</strong> a limitstage.<br />
Axiom ofChoice. For everyset Athere ex<strong>is</strong>ts a well-or<strong>de</strong>r R over A.<br />
Axiom ofReplacement. If F <strong>is</strong> a function and dom F <strong>is</strong> a set then so <strong>is</strong><br />
rng F.<br />
Asking whether these axioms are true does make as much sense as<br />
the question of whether the field axioms are true, or those of a vector<br />
space. Instead, what we are concerned with <strong>is</strong>theircons<strong>is</strong>tency andcompleteness.<br />
�at <strong>is</strong>, there should ex<strong>is</strong>t at least one object sat<strong>is</strong>fying these<br />
axioms and all such objects should look alike. Unfor<strong>tu</strong>nately, one can<br />
provethatthere <strong>is</strong> no complete axiomsystem forsettheory. Hence, we<br />
will haveto <strong>de</strong>al withthe factthatthere are many different mo<strong>de</strong>ls ofset<br />
theory andthere <strong>is</strong> no wayto choose one ofthem asthe‘canonical one’.<br />
Inparticular, there <strong>is</strong> nosuchthing as‘thereal mo<strong>de</strong>l ofsettheory’.<br />
Moreseriously, it <strong>is</strong> even impossibletoprovethat our axiom system<br />
<strong>is</strong> cons<strong>is</strong>tent. �at <strong>is</strong>, it might be the case that there <strong>is</strong> no mo<strong>de</strong>l of set<br />
theory and we have wasted ourtimes<strong>tu</strong>dying a nonsensical theory.<br />
�e firstproblem <strong>is</strong> <strong>de</strong>alt withrather easily. It does not matter which<br />
ofthese mo<strong>de</strong>ls we are givensince anytheoremthat we can <strong>de</strong>rive from<br />
the axioms holds in every mo<strong>de</strong>l. Butthesecondproblem <strong>is</strong>serious. All<br />
we can do <strong>is</strong> to restrict ourselves to as few axioms as possible and to<br />
hopethat no one will ever be ableto <strong>de</strong>rive a contradiction. Of course,<br />
the weaker the axioms the more different mo<strong>de</strong>ls we might get andthe<br />
fewertheorems we will be abletoprove.<br />
4
1. Setsandclasses<br />
Inthe following we will assumethatS<strong>is</strong> an arbitrary but fixed mo<strong>de</strong>l<br />
of set theory. �at <strong>is</strong>, S <strong>is</strong> a collection of objects that sat<strong>is</strong>fies all the<br />
axioms we will introduce below. S will be called the universe and its<br />
elements are calledsets. NotethatSitself <strong>is</strong> not asetsince we willprove<br />
belowthat noset <strong>is</strong> an element of itself. By convention, if below wesay<br />
thatsomesetex<strong>is</strong>tsthen we meanthat it <strong>is</strong> contained inS.Similarly, we<br />
saythatallsets havesomeproperty if all elements ofSdoso.<br />
In<strong>tu</strong>itively, aset <strong>is</strong> a collection of objects called itselements. Ifa andb<br />
aresets, i.e., elements ofS, we write a∈bif a <strong>is</strong>anelement of b and we<br />
<strong>de</strong>fine<br />
a⊆b : iff every element x∈a <strong>is</strong> also an element x∈ b.<br />
If a⊆b, we call a a subset of b, and we say that a <strong>is</strong> inclu<strong>de</strong>d in b, or<br />
that b <strong>is</strong> a superset of a. We use the usual abbreviations such as a⊂b<br />
for a⊆band a≠b; a∋bforb∈a; and a∉bif a∈bdoes not hold.<br />
Since aset <strong>is</strong> a collection of objects it <strong>is</strong> na<strong>tu</strong>raltorequirethat aset <strong>is</strong><br />
uniquely <strong>de</strong>termined by its elements. Our first axiom can therefore be<br />
regar<strong>de</strong>d asthe <strong>de</strong>finition of aset.<br />
Axiom of Extensionality. Twosets aandbareequal if,and only if,<br />
x∈a iff x∈ b , forallsets x.<br />
Lemma1.1. Twosets aandbareequal ifand only if a⊆bandb⊆a.<br />
In or<strong>de</strong>rto <strong>de</strong>fine aset we havetosay what its elements are. Iftheset<br />
<strong>is</strong> finite we can just enumerate them. Otherw<strong>is</strong>e, we have to find some<br />
propertyφsuchthat an object x <strong>is</strong> an element of a if, and only if, it has<br />
theproperty φ.<br />
Definition1.2. (a) Let φ be aproperty.{x∣φ} <strong>de</strong>notestheset a such<br />
that, for allsets x, we have<br />
x∈a iff x haspropertyφ.<br />
5
a1. Basicsettheory<br />
If S does not contain such an object then the expression{x ∣ φ} <strong>is</strong><br />
un<strong>de</strong>fined.<br />
(b) Letb0,... ,bn−1 besets. We <strong>de</strong>fine<br />
{b0,... ,bn−1}∶={x∣x=bi forsome i
already been <strong>de</strong>fined,the nextstage<br />
HFn+1∶={x∣ x⊆ HFn}<br />
1. Setsandclasses<br />
cons<strong>is</strong>ts of allsetsthat we can construct from elements of HFn.<br />
Aset <strong>is</strong> called hereditary finite if it <strong>is</strong> an element ofsome HFn.�eset<br />
of all hereditary finitesets <strong>is</strong><br />
HF∶={x∣x∈ HFn forsome n}.<br />
Notethat we cannotprove atthe momentthatHFreally <strong>is</strong> aset.Since<br />
the emptyuniverseS=∅trivially sat<strong>is</strong>fies the Axiom of Extensionality,<br />
we even cannot show that the empty set ex<strong>is</strong>ts without additional axioms.<br />
Le<strong>tu</strong>s assume forthe moment that HF does ex<strong>is</strong>ts. Its firststages<br />
are<br />
HF0=∅<br />
HF1={∅}<br />
HF2={∅,{∅}}<br />
HF3={∅,{∅},{{∅}},{∅,{∅}}}<br />
...<br />
By induction on n, one can prove that HFn⊆ HFn+1 and each set a∈<br />
HFn+1 <strong>is</strong> of the form a = {b0,...,bk−1}, for finitely many elements<br />
b0,...,bk−1∈ HFn. Note that each stage HFn <strong>is</strong> hereditary finite since<br />
HFn∈ HFn+1⊆ HF, buttheirunion HF <strong>is</strong> not because HF∉HF.<br />
Exerc<strong>is</strong>e 1.1. Prove the following statements by induction on n. (Although<br />
we have not <strong>de</strong>fined the na<strong>tu</strong>ral numbers yet, you may assume<br />
for th<strong>is</strong> exerc<strong>is</strong>e that they are available and that their usual properties<br />
hold.)<br />
(a) HFn⊆ HFn+1.<br />
(b) HFn has finitely many elements.<br />
7
a1. Basicsettheory<br />
(c) Every set a∈HFn+1 <strong>is</strong> ofthe form a={b0,... ,bk−1}, for finitely<br />
many elements b0,... ,bk−1∈ HFn.<br />
HF can beregar<strong>de</strong>d as an approximationtothe class of allsets. In fact,<br />
all but one of the usual axioms of set theory hold for HF. �e only exception<br />
<strong>is</strong>the Axiom of Infinity whichstatesthatthere ex<strong>is</strong>ts an infinite<br />
set.<br />
We can enco<strong>de</strong> na<strong>tu</strong>ral numbers byspecial hereditary finitesets.<br />
Definition1.4. To each na<strong>tu</strong>ral number n we associatetheset<br />
[n]∶={[0],...,[n− 1]}.<br />
�eset of all na<strong>tu</strong>ral numbers <strong>is</strong><br />
N∶={[n]∣n a na<strong>tu</strong>ral number}.<br />
Notethat[n]∈HFn+1 but[n]∉HFn, and N∉HF. �<strong>is</strong> construction<br />
can beusedto <strong>de</strong>finethe na<strong>tu</strong>ral numbers inpurelysettheoretic terms.<br />
In the following by a na<strong>tu</strong>ral number we will always mean a set of the<br />
form[n].<br />
It would be nice if there were a universe S that contains all sets of<br />
the form{x∣φ}. Unfor<strong>tu</strong>nately, such a universe does not ex<strong>is</strong>ts, that<br />
<strong>is</strong>, if we add the axiom that claims that{x ∣ φ} <strong>is</strong> <strong>de</strong>fined for all φ,<br />
we obtain a theory that <strong>is</strong> incons<strong>is</strong>tent, i.e., it contradicts itself. In fact,<br />
we can even showthatthere areproperties φ suchthat no mo<strong>de</strong>l ofset<br />
theory contains a set of the form{x∣φ}. And we can do so without<br />
using asingle axiom ofsettheory.<br />
�eorem1.5(Zermelo-RussellParadox). {x∣x∉ x} <strong>is</strong> notaset.<br />
Proof. Supposethattheseta∶={x∣x∉x}ex<strong>is</strong>ts. Letx be an arbitrary<br />
set. By <strong>de</strong>finition, we have x∈a if and only if x∉x. In particular, for<br />
x=a,we obtain a∈a iff a∉a. A contradiction. ◻<br />
To betterun<strong>de</strong>rstand what <strong>is</strong> going on, le<strong>tu</strong>sseewhat happens ifwe<br />
restrict ourselves to hereditary finite sets. �e set{x∈ HF∣x∉x}<br />
8
1. Setsandclasses<br />
equals HF since no hereditary finite set contains itself. But HF∉HF <strong>is</strong><br />
not hereditary finite.�esame happens inrealsettheory.�e condition<br />
x∉ x <strong>is</strong>sat<strong>is</strong>fied by all sets andwe have{x∣x∉ x}=S,which <strong>is</strong> not a<br />
set.<br />
In general, an expression of the form{x∣φ} <strong>de</strong>notes a collection<br />
X⊆ S that may or may not be a set, i.e., an element X∈S. We will<br />
call objects ofthe form{x∣φ}classes. Classes that are notsetswill be<br />
called proper classes. If X={x∣φ} and Y={x∣ ψ} are classes and<br />
a <strong>is</strong> aset,wewrite<br />
a∈ X : iff a haspropertyφ ,<br />
X⊆ Y : iff every setwithproperty φ also haspropertyψ ,<br />
and X= Y : iff X⊆Y andY⊆ X.<br />
If X <strong>is</strong> a proper class then we <strong>de</strong>fine X∉ Y, for every Y. Note that, if<br />
X and Y are sets then these <strong>de</strong>finitions coinci<strong>de</strong> with the ones above.<br />
Finally, we remark that every set a <strong>is</strong> a class since we can write a as<br />
{x∣ x∈a}.<br />
When <strong>de</strong>fining classeswe haveto be a bit careful aboutwhatwe call<br />
a property. Let us <strong>de</strong>fine a property to be a statement that <strong>is</strong> build up<br />
from basicpropositions ofthe form x∈y and x=y by<br />
◆ logical conjunctions like‘and’,‘or’,‘not’,‘if-then’;<br />
◆ constructs of the form ‘there ex<strong>is</strong>ts a set x suchthat ...’ and ‘for<br />
allsets x it holdsthat...’.<br />
(Such statements will be <strong>de</strong>fined in a more formal way in Chapter c1<br />
wherewewill callthem‘first-or<strong>de</strong>r formulae’.)�ingswe are not allowed<br />
to say inclu<strong>de</strong> statements of the form ‘�ere ex<strong>is</strong>ts a property φ such<br />
that...’ or‘For all classes X it holdsthat...’.<br />
We have <strong>de</strong>fined a class to be an object of the form{x∣φ} where<br />
φ <strong>is</strong> astatement aboutsets.What happens ifwe allowstatements about<br />
arbitrary classes? Note that, if φ <strong>is</strong> a property referring to a class X=<br />
{x ∣ ψ} then we can transform φ into an equivalent statement only<br />
talking aboutsets byreplacing allpropositions y∈ X, X∈y, X=y, etc.<br />
bytheirrespective <strong>de</strong>finitions.<br />
9
a1. Basicsettheory<br />
Example. Let X={x∣∅∉ x}.We canwritethe class<br />
inthe form<br />
{y∣ y≠∅and y⊆ X}<br />
{y∣ y≠∅and∅∉ x for all x∈y}.<br />
�esi<strong>tu</strong>ation <strong>is</strong> analogoustothe case ofthe complex numberswhich<br />
are obtained fromthereal numbers by adding imaginary elements.We<br />
can translate any statement about complex numbers x+iy into one<br />
about pairs⟨x,y⟩ of real numbers. Consequently, it does not matter<br />
whetherwe allow classes inthe <strong>de</strong>finition of other classes.<br />
In<strong>tu</strong>itively, the reason for a proper class such as S not being a set <strong>is</strong><br />
that it <strong>is</strong> too ‘large’. For instance, when consi<strong>de</strong>ring HF we see that a<br />
set a⊆HF <strong>is</strong> hereditary finite if, and only if, it has only finitely many<br />
elements. Hence, if we can show that a class X={x∣φ} <strong>is</strong> ‘small’, it<br />
should form aset.What dowe mean by‘small’? Clearly, wewould like<br />
every set to be small. Furthermore, it <strong>is</strong> na<strong>tu</strong>ral to require that, if Y <strong>is</strong><br />
small and X⊆ Y then X <strong>is</strong> alsosmall. �erefore,we <strong>de</strong>fine a class X to<br />
besmall if it <strong>is</strong> asubclass X⊆aofsomeset a.<br />
Definition1.6. For a class A and aproperty φwe <strong>de</strong>fine<br />
{x∈ A∣ φ}∶={x∣x∈A and x haspropertyφ}.<br />
�<strong>is</strong> <strong>de</strong>finition ensures that every class of the form X={x∈a∣φ}<br />
wherea<strong>is</strong> aset <strong>is</strong>small. Conversely, if X={x∣ φ} <strong>is</strong>smallthen X⊆ a,<br />
forsomeset a, andwe have X={x∈a∣φ}. Oursecond axiomstates<br />
that everysmall class <strong>is</strong> aset.<br />
Axiom of Separation. If a <strong>is</strong>asetandφaproperty thentheclass<br />
<strong>is</strong>aset.<br />
10<br />
{x∈ a∣φ}
2. Stagesand h<strong>is</strong>tories<br />
With th<strong>is</strong> axiom westill cannot provethat there <strong>is</strong> any set. But if we<br />
have at least oneset a,we can <strong>de</strong>duce, for instance, that alsothe empty<br />
set∅={x∈ a∣x≠x} ex<strong>is</strong>ts.<br />
Definition1.7. LetA and B be classes.<br />
(a) �e intersection ofA <strong>is</strong>the class<br />
⋂A∶={x∣ x∈y for all y∈A}.<br />
(b) �e intersection ofA and B <strong>is</strong><br />
A∩B∶={x∣x∈A and x∈B}.<br />
(c) �edifference between A and B <strong>is</strong><br />
A∖B∶={x∈ A∣ x∉ B}.<br />
Lemma 1.8. Let a be aset and B aclass.�en a∩Band a∖B are sets.<br />
If Bcontainsat least oneelementthen⋂B <strong>is</strong>aset.<br />
Proof. �e factthat a∩B={x∈a∣x∈B} and a∖B aresets follows<br />
immediately from the Axiom of Separation. If B contains at least one<br />
element c∈Bthenwe canwrite<br />
⋂B={x∈ c∣x∈y for all y∈B}. ◻<br />
Notethat⋂∅=S<strong>is</strong> not aset.<br />
2. Stagesand h<strong>is</strong>tories<br />
�e construction ofHF above can be exten<strong>de</strong>dto one ofthe classSof all<br />
sets.We <strong>de</strong>fineSastheunion of an increasingsequence ofsetsSα , called<br />
thestages ofS. Again,westartwiththe emptysetS0∶=∅. IfSα <strong>is</strong> <strong>de</strong>fined<br />
thenthe nextstage Sα+1 contains all subsets of Sα. Butth<strong>is</strong>time,we do<br />
not stop when we have <strong>de</strong>fined Sα for all na<strong>tu</strong>ral numbers α. Instead,<br />
11
a1. Basicsettheory<br />
every time we have <strong>de</strong>fined an infinite sequence of stages we continue<br />
bytakingtheirunionto formthe nextstage.So oursequencestartswith<br />
S0= HF0 , S1= HF1 , S2= HF2 , ...<br />
�e nextstage a�er allthe finite ones <strong>is</strong>Sω∶= HF andwe continue with<br />
Sω+1={x∣ x⊆ HF} , Sω+2={x∣x⊆Sω+1} , ...<br />
A�erwe have <strong>de</strong>fined Sω+n for all na<strong>tu</strong>ral numbers n we againtakethe<br />
union<br />
Sω+ω={x∣ x∈ Sω+n forsome n},<br />
andso on.<br />
Unfor<strong>tu</strong>nately, makingth<strong>is</strong> constructionprec<strong>is</strong>e<strong>tu</strong>rns outto bequite<br />
technical since we cannot <strong>de</strong>finethe numbers α yet thatwe need to in<strong>de</strong>xthesequenceSα.�<strong>is</strong><br />
hastowai<strong>tu</strong>ntilSection a3.2. Instead,we start<br />
by giving a condition for some set S to be a stage, i.e., one of the Sα. If<br />
we or<strong>de</strong>r allsuchsets by inclusionthenwe obtainthe <strong>de</strong>siredsequence<br />
S0⊆ S1⊆⋅⋅⋅⊆Sω⊆ Sω+1⊆⋯ ,<br />
withoutthe needtoreferto its indices.<br />
First,we <strong>is</strong>olatesome character<strong>is</strong>tic properties ofthesets HFn which<br />
we would like that our stages Sα share. Note that, at the moment, we<br />
cannotprovethat any ofthesets mentioned below ac<strong>tu</strong>ally ex<strong>is</strong>ts.<br />
Definition 2.1. LetA be a class.<br />
12<br />
(a) We call Atransitive if x∈y∈A impliesx∈A.<br />
(b) We call A hereditary if x⊆y∈A impliesx∈ A.<br />
(c) �eaccumulation of A <strong>is</strong>the class<br />
acc(A)∶={x∣there <strong>is</strong>some y∈Asuchthat x∈y or x⊆y}.<br />
Notethat eachstage HFn of HF <strong>is</strong> hereditary andtransitive.
2. Stagesand h<strong>is</strong>tories<br />
Exerc<strong>is</strong>e 2.1. By induction on n,showthattheset[n] <strong>is</strong>transitive. Give<br />
an example of a number n suchthat[n] <strong>is</strong> not hereditary.<br />
�e next lemmas follow immediately fromthe <strong>de</strong>finitions.<br />
Lemma 2.2. Let Abeaclass,andb,c sets. �e followingstatements are<br />
equivalent:<br />
(a) c∈b∈A implies c∈A,that <strong>is</strong>,A <strong>is</strong>transitive.<br />
(b) b∈A impliesb⊆A.<br />
(c) b∈A impliesb∩A= b.<br />
Lemma 2.3. Let AandBbeclasses.<br />
(a) A⊆ acc(A)<br />
(b) IfB<strong>is</strong> hereditaryandtransitiveand ifA⊆ B,then acc(A)⊆ B.<br />
(c) A <strong>is</strong> hereditaryandtransitive if,and only if, acc(A)= A.<br />
Lemma 2.4. IfAandBaretransitiveclassesthenso <strong>is</strong> A∩B.<br />
Exerc<strong>is</strong>e 2.2. Prove Lemmas2.2, 2.3, and 2.4.<br />
Definition 2.5. LetA be a class.<br />
(a) A minimalelement of A <strong>is</strong> an element b∈Asuchthat b∩A=∅,<br />
that <strong>is</strong>,there <strong>is</strong> no element c∈Awith c∈b.<br />
(b) Aset a <strong>is</strong> foun<strong>de</strong>d if everysetb∋a has a minimal element.<br />
(c) �e foun<strong>de</strong>dpart of A <strong>is</strong>theset<br />
fnd(A)∶={x∈ A∣ x <strong>is</strong> foun<strong>de</strong>d}.<br />
Example. �e emptyset∅andtheset{∅} are foun<strong>de</strong>d.Toseethat{∅}<br />
<strong>is</strong> foun<strong>de</strong>d, consi<strong>de</strong>r asetb∋{∅}. If{∅} <strong>is</strong> not a minimal element ofb,<br />
then b∩{∅}≠∅. Hence,∅∈ b <strong>is</strong> a minimal element ofb.<br />
Exerc<strong>is</strong>e 2.3. Provethat every hereditary finiteset <strong>is</strong> foun<strong>de</strong>d.<br />
13
a1. Basicsettheory<br />
Wewill introduce an axiom belowwhich impliesthat every class has a<br />
minimal element. Hence, everyset <strong>is</strong> foun<strong>de</strong>d andwe have fnd(A)= A,<br />
for all classesA. Althoughthe notions of a foun<strong>de</strong>dset andthe foun<strong>de</strong>d<br />
part of asetwill<strong>tu</strong>rn outto betrivial,westill needthemto <strong>de</strong>finestages<br />
andto formulatethe axiom.<br />
Lemma 2.6. IfB<strong>is</strong>ahereditaryclassand a∈Bthen fnd(a)∈fnd(B).<br />
Proof. For a contradiction suppose that fnd(a)∉fnd(B). Since B <strong>is</strong><br />
hereditary and fnd(a)⊆ a∈B, we have fnd(a)∈ B. Consequently,<br />
fnd(a)∉ fnd(B) impliesthatthere <strong>is</strong>somesetx∋ fnd(a)without minimal<br />
element. Inparticular, fnd(a) <strong>is</strong> not a minimal element of x,that<br />
<strong>is</strong>,there ex<strong>is</strong>tssomeset y∈x∩ fnd(a). But y∈ fnd(a) impliesthat y <strong>is</strong><br />
foun<strong>de</strong>d.�erefore, from y∈xit followsthatx has a minimal element.<br />
A contradiction. ◻<br />
Inthe language ofSection a3.1 the next theorem states that the membership<br />
relation∈<strong>is</strong>well-foun<strong>de</strong>d on every class oftransitive, hereditary<br />
sets.<br />
�eorem 2.7. Let Abeanonemptyclass. Ifeveryelementx∈ A <strong>is</strong> hereditaryandtransitive,then<br />
A hasaminimalelement.<br />
Proof. Choose an arbitrary element c∈A andset<br />
b∶={ fnd(x)∣ x∈ c∩A}.<br />
If b=∅ then c∩A=∅ and c <strong>is</strong> a minimal element of A. �erefore,<br />
we may assume that b≠∅. Since c∈A <strong>is</strong> hereditary, it follows from<br />
Lemma2.6thatb⊆fnd(c). Fixsomex∈ b⊆fnd(c).�enx <strong>is</strong> foun<strong>de</strong>d<br />
andx∈ b impliesthatbhas a minimal element y. By <strong>de</strong>finition ofb,we<br />
have y=fnd(z), forsomez∈c∩A.<br />
We claimthatz<strong>is</strong> a minimal element of A.Suppose otherw<strong>is</strong>e.�en<br />
there ex<strong>is</strong>tssome element u∈z∩A.Since c <strong>is</strong>transitive we have u∈c.<br />
Hence, u∈c∩Aimplies fnd(u)∈b. On the other hand, since z∈A<br />
<strong>is</strong> hereditary it follows from Lemma 2.6 that fnd(u)∈fnd(z). Hence,<br />
14
2. Stagesand h<strong>is</strong>tories<br />
fnd(u)∈fnd(z)∩b≠∅ and y= fnd(z) <strong>is</strong> not a minimal element ofb.<br />
A contradiction. ◻<br />
We would like to <strong>de</strong>fine that a set S <strong>is</strong> a stage if it <strong>is</strong> hereditary and<br />
transitive. Unfor<strong>tu</strong>nately, th<strong>is</strong> <strong>de</strong>finition <strong>is</strong> too weak to show that the<br />
stages can be arranged in an increasingsequenceS0⊆ S1⊆⋯⊆Sα⊆⋯.<br />
�erefore, our <strong>de</strong>finitionwill beslightly more involved.To eachstageSα<br />
wewill associate its h<strong>is</strong>tory<br />
H(Sα)={Sβ∣ β
a1. Basicsettheory<br />
Proof. (a) a⊆a∈Himplies a∈ acc(H)= S.<br />
(b) By <strong>de</strong>finition of a h<strong>is</strong>tory, we have a = acc(H∩a). Hence, if<br />
we can show that H∩a<strong>is</strong> a h<strong>is</strong>tory then its stage <strong>is</strong> a. Clearly, every<br />
element of H∩a⊆H<strong>is</strong> hereditary andtransitive. Letb∈H∩a.�en<br />
b⊆acc(H∩a)= a. It followsthat H∩b=(H∩a)∩b. Furthermore,<br />
since H <strong>is</strong> a h<strong>is</strong>torywe have<br />
b= acc(H∩b)=acc((H∩a)∩b) ,<br />
whichshowsthat H∩a <strong>is</strong> a h<strong>is</strong>tory.<br />
(c) Letb∈S.�e class<br />
a∶={s∈H∣b∈sorb⊆s}<br />
<strong>is</strong> nonempty becauseb∈S= acc(H). By�eorem2.7, it has a minimal<br />
elements∈a.<br />
If b∈s=acc(H∩s), there <strong>is</strong> some set z∈H∩s suchthat b∈z or<br />
b⊆z. It followsthat z∈a. But z∈s∩a impliesthats <strong>is</strong> not a minimal<br />
element of a. Contradiction.<br />
�erefore, b∉s which implies, by <strong>de</strong>finition of a, that b⊆s. For<br />
transitivity, notethatx∈ b implies<br />
x∈ b⊆s= acc(H∩s)⊆acc(H)= S.<br />
For hereditarity, let x ⊆ b. �en x ⊆ b⊆s∈H, which implies x ∈<br />
acc(H)=S.<br />
(d) By (c) we know that x⊆s∈S implies x∈S. For the other direction,<br />
suppose that x∈S= acc(H). �ere <strong>is</strong> some set s∈H such that<br />
x∈ s or x⊆ s. By (a), (b), and (c) it followsthat s∈S, s <strong>is</strong> a stage, and<br />
s <strong>is</strong> hereditary and transitive. By transitivity, if x∈ s then x⊆s. Consequently,<br />
we havex⊆ s∈Sin both cases andthe claim follows.<br />
(e) By (d),we have S= acc(H(S)). It remains to show that H(S) <strong>is</strong><br />
a h<strong>is</strong>tory. By (c), every element s∈H(S) <strong>is</strong> hereditary and transitive.<br />
Furthermore,since S <strong>is</strong>transitive we haves⊆Sand it followsthat<br />
H(S)∩s={x∈ s∣x<strong>is</strong> astage}.<br />
Sinces<strong>is</strong> astagewe know by(d)thats= acc(H(S)∩s). ◻<br />
16
2. Stagesand h<strong>is</strong>tories<br />
Notethat, by(a) and(b) above,we have H⊆ H(S), for all h<strong>is</strong>tories H<br />
of S. In fact, H(S) <strong>is</strong> the only h<strong>is</strong>tory of S but we need some further<br />
results beforewe canproveth<strong>is</strong>.<br />
Exerc<strong>is</strong>e 2.4. Prove, by induction on n, that{HF0,... ,HFn−1} <strong>is</strong> a h<strong>is</strong>torywithstage<br />
HFn.<br />
Exerc<strong>is</strong>e 2.5. Construct a hereditarytransitivity setathat <strong>is</strong> not astage.<br />
Hint. It <strong>is</strong>sufficientto consi<strong>de</strong>r sets HFn⊂ a⊂ HFn+1, for asmall n.<br />
A�erwe haveseen howto <strong>de</strong>finestageswe nowprovethatthey form<br />
astrictly increasingsequence S0⊆ S1⊆....Togetherwith�eorem2.7<br />
it followsthatthe class of all stages <strong>is</strong>well-or<strong>de</strong>red bythe membership<br />
relation∈(seeSection a3.1).<br />
�eorem 2.10. IfS andT arestagesthataresetsthenwe have<br />
S∈T or S=T or T∈ S .<br />
Proof. Supposethatthere arestages S and T suchthat<br />
(∗) S∉T, S≠T , and T∉ S.<br />
Define<br />
A∶={s∣s<strong>is</strong> astage andthere <strong>is</strong>somestage t suchthat<br />
s and t sat<strong>is</strong>fy(∗)}.<br />
By�eorem2.7,the class A has a minimal element S0. Define<br />
B∶={t∣t<strong>is</strong> astagesuchthat S0 and t sat<strong>is</strong>fy(∗)}.<br />
Againthere <strong>is</strong> a minimal element T0∈ B.<br />
Ifwe canshowthat H(S0)=H(T0), it followsthat<br />
S0= acc(H(S0))=acc(H(T0))= T0<br />
17
a1. Basicsettheory<br />
in contradiction to our choice of S0 and T0.<br />
Lets∈S0 be astage.�ens≠T0sinceT0∉ S0. Furthermore,we have<br />
T0∉ s since, otherw<strong>is</strong>e, transitivity of S0 would imply that T0∈ S0. By<br />
minimality of S0 it follows that s and T0 do not sat<strong>is</strong>fy(∗). �erefore,<br />
we haves∈T0.<br />
We haveshownthat H(S0)⊆H(T0). Asymmetric argumentshows<br />
that H(T0)⊆H(S0). Hence,we have H(S0)=H(T0) as <strong>de</strong>sired. ◻<br />
Lemma 2.11. Let S and T bestagesthataresets.<br />
(a) S∉ S<br />
(b) S⊆ T ifand only ifS∈ T orS=T.<br />
(c) S⊆ T or T⊆ S.<br />
(d) S⊂ T if,and only if, S∈T.<br />
Proof. (a)Suppose otherw<strong>is</strong>e. LetXbethe class of allstagesssuchthat<br />
s∈s. By �eorem 2.7, X has a minimal element s, that <strong>is</strong>, an element<br />
suchthats∩ X=∅. Buts∈s∩X. Contradiction.<br />
(b) If S=Tthen S⊆T, and if S∈Tthen S⊆T, bytransitivity of T.<br />
Conversely, if neither S=Tnor S∈ T then �eorem 2.10 impliesthat<br />
T∈ S. IfS⊆T then T∈ S⊆Twould contradict (a).<br />
(c) If S⊈Tthen (b) impliesthat S∉Tand S≠T. By �eorem 2.10,<br />
it followsthatT∈ S which, again by(b), impliesT⊆ S.<br />
(d) We have S⊂Tiff S⊆Tand S≠T. By (a) and (b), the latter <strong>is</strong><br />
equivalent toS∈ T. ◻<br />
3. �ecumulative hierarchy<br />
Intheprevioussectionwe haveseenthatwe can arrange allstages in an<br />
increasing sequence<br />
S0⊂ S1⊂⋅⋅⋅⊂Sα⊂⋯ ,<br />
whichwewill call thecumulative hierarchy. If S∈Tarestages then we<br />
willsaythatS <strong>is</strong>earlierthan T, orthat T <strong>is</strong> laterthan S.<br />
18
3. �ecumulative hierarchy<br />
From the axioms we have available we cannot prove that there ac<strong>tu</strong>ally<br />
are any stages. We introduce a new axiom which ensures that<br />
enoughstages are available.<br />
Axiom of Creation. Foreveryset athere <strong>is</strong>aset S∋awhich <strong>is</strong>astage.<br />
Inparticular, th<strong>is</strong> axiom impliesthat<br />
◆ for every stage S that <strong>is</strong> aset, there ex<strong>is</strong>ts a later stage T∋ S that<br />
<strong>is</strong> also aset.<br />
◆ theuniverseS<strong>is</strong>theunion of allstages.<br />
Of course, evenwithth<strong>is</strong> new axiomwe mightstill haveS=∅. But if at<br />
least oneset ex<strong>is</strong>ts,we can nowprovethat HF⊆S. Inparticular, S=HF<br />
sat<strong>is</strong>fies all axiomswe have introduced so far.<br />
Exerc<strong>is</strong>e 3.1. ProvethatS<strong>is</strong> astagewith h<strong>is</strong>tory<br />
H(S)={S∣S<strong>is</strong> astage}.<br />
Definition 3.1. (a) Wesay that a stage T <strong>is</strong> the successor of the stage S<br />
if S∈Tand there ex<strong>is</strong>ts nostage T ′ suchthat S∈T ′ ∈ T. A nonempty<br />
stage <strong>is</strong> a limit if it <strong>is</strong> notthesuccessor ofsome otherstage.<br />
(b) Let A be a class. We <strong>de</strong>note by S(A) the earliest stage such that<br />
A⊆ S(A).<br />
NotethatS(A) <strong>is</strong>well-<strong>de</strong>fined by�eorem2.7.We haveS(s)= s, for<br />
every stage s, in particular, S(∅)=∅. �e stages S and HF are limits<br />
and HFn+1 <strong>is</strong>thesuccessor ofthestage HFn.<br />
Lemma 3.2. a∈bimplies S(a)∈ S(b).<br />
Proof. Since a∈b⊆S(b)=acc(H(S(b))) it followsthatthere <strong>is</strong>some<br />
stage s∈S(b) such that a∈s or a⊆s. In particular, S(a) <strong>is</strong> not later<br />
than s which implies that S(a)⊆s∈S(b). As S(b) <strong>is</strong> hereditary we<br />
therefore have S(a)∈ S(b). ◻<br />
Lemma 3.3. S <strong>is</strong>the onlystagethat <strong>is</strong>aproperclass.<br />
19
a1. Basicsettheory<br />
Proof. Let S be a stage. If S≠ S, there <strong>is</strong> some set a∈S∖S. Hence,<br />
S(a)∉ S which impliesthat<br />
T∉ H(S) , for allstages T⊇ S(a).<br />
By Lemma2.9(e) and�eorem2.10,we have<br />
H(S)⊆{T∣ T <strong>is</strong> astagewith T∈ S(a)}= H(S(a)).<br />
Inparticular, H(S) <strong>is</strong> aset,which impliesthatso <strong>is</strong>S= acc(H(S)). ◻<br />
Lemma 3.4. Let Abeaclass.�e followingstatements areequivalent:<br />
(1) A <strong>is</strong>aproperclass.<br />
(2) S(A) <strong>is</strong>aproperclass.<br />
(3) S(A)=S.<br />
Proof. (3)⇒(1) By the Axiom of Creation, ifA <strong>is</strong> a set then so <strong>is</strong>S(A).<br />
(1)⇒(2) IfS(A) <strong>is</strong> a set then A⊆ S(A) implies that<br />
A={x∈ S(A)∣ x∈ A}<br />
<strong>is</strong> also a set.<br />
(2)⇒(3) follows by Lemma 3.3. ◻<br />
With the Axiom of Creationwe are finally able to prove most ‘obvious’<br />
properties ofsetssuchthat noset <strong>is</strong> an element of itself orthattheunion<br />
ofsets <strong>is</strong> aset.<br />
Lemma 3.5. If a <strong>is</strong>asetthen a∉a.<br />
Proof. Supposethatthere ex<strong>is</strong>tssomesetsuchthat a∈a.�en a∈a⊆<br />
S(a) and, by Lemma 2.9(d), there <strong>is</strong> some stage s∈S(a) with a⊆s.<br />
�<strong>is</strong> contradicts the minimality of S(a). ◻<br />
�eorem 3.6. Every nonemptyclassA hasaminimalelement.<br />
20
3. �ecumulative hierarchy<br />
Proof. By �eorem 2.7, we can choose some element b∈A such that<br />
S(b) <strong>is</strong> minimal. We claim that b <strong>is</strong> a minimal element of A. Suppose<br />
otherw<strong>is</strong>e.�enthere ex<strong>is</strong>tssome elementx∈ A∩b.Sincex∈ b⊆S(b),<br />
Lemma2.9(d) impliesthatthere <strong>is</strong>somestages∈S(b)suchthat x⊆ s.<br />
Hence, x <strong>is</strong> an element of A with S(x)∈ S(b) in contradiction to the<br />
choice ofb. ◻<br />
Wewillsee inSection a3.1 that �eorem 3.6 implies that there are no<br />
infinite <strong>de</strong>scending sequences a0∋a1∋... ofsets. (Ifsuch asequence<br />
ex<strong>is</strong>tsthentheset{a0,a1,...} has no minimal element.)<br />
Example. By induction on n, ittrivially followsthat, if a0∋⋯∋ ak−1 <strong>is</strong><br />
asequence of setsstarting with a0∈ HFn,then k
a1. Basicsettheory<br />
Proof. Let S0 and S1 be stages such that a∈S0 and b∈S1. We know<br />
thatS0⊆S1 orS1⊆S0. By choosing eitherS0 orS1we can find astageS<br />
suchthat S0⊆Sand S1⊆ S. Bytransitivity of S it followsthat<br />
⋃a={x∈ S∣x∈ b forsomeb∈a},<br />
a∪b={x∈ S∣x∈ a or x∈ b},<br />
{a}={x∈ S∣x=a},<br />
and ℘(a)={b∈S∣ b⊆a}. ◻<br />
Corollary 3.9. If a0,... ,an−1 aresetsthenso <strong>is</strong><br />
{a0,... ,an−1}={a0}∪⋅⋅⋅∪{an−1}.<br />
Inparticular,every finiteclass <strong>is</strong>aset.<br />
�e next <strong>de</strong>finitionprovi<strong>de</strong>s ausefultoolwhichsometimes allowsus<br />
to replace a proper class A by a set a. Instead of taking every element<br />
x∈Awe only consi<strong>de</strong>r thosesuchthat S(x) <strong>is</strong> minimal.<br />
Definition 3.10. �ecut of a classA <strong>is</strong>theset<br />
cutA∶={x∈ A∣ S(x)⊆ S(y) for all y∈A}.<br />
Exerc<strong>is</strong>e 3.2. What are cutS and cut{x∣a∈x}?<br />
Lemma 3.11. Everyclass ofthe form cutA <strong>is</strong>aset.<br />
Proof. IfA=∅then cutA=∅. Otherw<strong>is</strong>e, choose an arbitraryseta∈ A.<br />
�en cutA⊆ S(a)which impliesthat cutA<strong>is</strong> aset. ◻<br />
�e following lemmas clarify the struc<strong>tu</strong>re of the cumulative hierarchy.<br />
Lemma 3.12. �esuccessor ofastage S <strong>is</strong>℘(S).<br />
22
3. �ecumulative hierarchy<br />
Proof. By �eorem 2.7, there ex<strong>is</strong>ts a minimal stage T with S∈T. We<br />
havetoprovethat T=℘(S). a⊆S∈Timplies a∈T since T <strong>is</strong> hereditary.<br />
Hence,℘(S)⊆ T.<br />
Conversely, if s∈T<strong>is</strong> a stage then S∉s because T <strong>is</strong> the successor<br />
of S. By�eorem2.10, it followsthats∈Sors=S.�<strong>is</strong> impliess⊆S.<br />
We have shown that s ∈ T iff s ⊆ S, for all stages s. It follows by<br />
Lemma2.9(d)that<br />
T={x∣x⊆ s forsomestages∈T}<br />
={x∣x⊆ s forsomestages⊆S}={x∣ x⊆S}=℘(S). ◻<br />
Lemma 3.13. Let S be a nonempty stage. �e following statements are<br />
equivalent:<br />
(1) S <strong>is</strong>alimitstage.<br />
(2) S=⋃ H(S).<br />
(3) Foreveryset a∈S,thereex<strong>is</strong>ts somestages∈Swith a∈s.<br />
(4) If a∈Sthen℘(a)∈ S.<br />
(5) If a∈Sthen{a}∈ S.<br />
(6) If a⊆Sthen cuta∈S.<br />
Proof. (2)⇒(1) Suppose that S <strong>is</strong> the successor of a stage T. �en we<br />
have<br />
H(S)={T}∪ H(T).<br />
Sinces⊆T, for all s∈H(T), it followsthat<br />
⋃ H(S)= T≠ S.<br />
(1)⇒(2)Supposethat S <strong>is</strong> a limitstage. By Lemma2.9(d),we have<br />
S=⋃{℘(s)∣ s∈H(S)}<br />
=⋃{t∣t<strong>is</strong>thesuccessor ofsomestage s∈H(S)}<br />
=⋃{t∣t∈H(S)}<br />
=⋃ H(S).<br />
23
a1. Basicsettheory<br />
(1)⇒(3) Suppose that S <strong>is</strong> a limit and let a∈S. By Lemma 2.9(d),<br />
there <strong>is</strong>somestage s∈Swith a⊆s. Hence, a∈℘(s).Since T∶=℘(s) <strong>is</strong><br />
thesuccessor ofswe haveT∈ S.<br />
(3)⇒(4) For each a∈S, there <strong>is</strong> some stage s∈Swith a∈s. Since<br />
s <strong>is</strong> transitive it follows that x⊆a implies x∈ s. Hence,℘(a)⊆s. By<br />
transitivity of S,we obtain℘(a)∈ S.<br />
(4)⇒(5) Ifa∈S then{a}⊆℘(a)∈ S. SinceS<strong>is</strong> hereditary, it follows<br />
that{a}∈ S.<br />
(5)⇒(1) IfS<strong>is</strong> no limit, there <strong>is</strong> some stageT∈ SsuchthatS=℘(T).<br />
By assumption,{T}∈S=℘(T). Hence,{T}⊆ T which implies that<br />
T∈ T. A contradiction.<br />
(3)⇒(6) Letb∶= cuta. Ifa=∅thenb=∅andwe are done. Ifthere<br />
<strong>is</strong> some element x∈a then, by assumption, we can find a stage s∈S<br />
with x∈ s. By <strong>de</strong>finition, b⊆s, and it followsthatb∈S.<br />
(6)⇒(5) Let a∈S and set b∶={x∈S∣a⊆x}. Clearly, b⊆S. By<br />
assumption, we therefore have c∶= cutb∈S. Hence,{a}⊆ c implies<br />
{a}∈ S. ◻<br />
So far,westill might haveS=∅ orS=HF.To exclu<strong>de</strong>these caseswe<br />
introduce a new axiomwhichstatesthat HF∈S.<br />
Axiom of Infinity. �ereex<strong>is</strong>tsasetthat <strong>is</strong>alimitstage.<br />
We callthetheory cons<strong>is</strong>ting ofthe four axioms<br />
◆ Axiom of Extensionality<br />
◆ Axiom ofSeparation<br />
◆ Axiom of Creation<br />
◆ Axiom of Infinity<br />
basic set theory. Every mo<strong>de</strong>l of th<strong>is</strong> theory cons<strong>is</strong>t of a hierarchy of<br />
stages<br />
S0⊂ S1⊂⋯⊂Sω⊂ Sω+1⊂ ...<br />
whereSn= HFn, for finite n.�e differences between twosuch mo<strong>de</strong>ls<br />
can be classified accordingtotwo axes:the length ofthe hierarchy and<br />
thesize of eachstage.<br />
24
3. �ecumulative hierarchy<br />
LetSandS ′ betwo mo<strong>de</strong>lswithstages(Sα)α
a1. Basicsettheory<br />
26
a2. Relations<br />
1. Relationsand functions<br />
With basicsettheory availablewe can <strong>de</strong>fine most ofthe conceptsused<br />
in mathematics. �esimplest one <strong>is</strong>the notion of an or<strong>de</strong>red pair. �e<br />
character<strong>is</strong>tic property ofsuchpairs <strong>is</strong>that⟨a,b⟩=⟨c,d⟩ implies a=c<br />
and b=d.<br />
Definition1.1. (a) Let a andbbesets.�e or<strong>de</strong>red pair⟨a,b⟩ <strong>is</strong>theset<br />
⟨a,b⟩∶={{a},{a,b}}.<br />
(b) Let A and B be classes. �e cartesian product of A and B <strong>is</strong> the<br />
class<br />
A×B∶={c∣c=⟨a,b⟩ forsome a∈A and b∈B}.<br />
Le<strong>tu</strong>sshowthat or<strong>de</strong>redpairs havethe <strong>de</strong>siredproperty.<br />
Lemma1.2. If{a,b}={a,c}thenb=c.<br />
Proof. We have b∈{a,b}={a,c}. Hence,b=a or b=c. Inthe latter<br />
case we are done. Otherw<strong>is</strong>e, we have c∈{a,c}={a,b}={b}which<br />
impliesthat c=b. ◻<br />
Lemma1.3. If⟨a,b⟩=⟨c,d⟩then a=candb=d.<br />
Proof. Supposethat⟨a,b⟩=⟨c,d⟩.<br />
{a}∈{{a},{a,b}}={{c},{c,d}}<br />
logic, algebra & geometry2012-08-08 — ©achim blumensath 27
a2. Relations<br />
implies{a}={c} or{a}={c,d}. Inthe latter case,we have a=c=d.<br />
In both cases, wetherefore have{a}={c}. Bythepreceding lemma, it<br />
follows that{a,b}={c,d} and, applying the lemma again, we obtain<br />
b=d. ◻<br />
Remark. �e above <strong>de</strong>finition of an or<strong>de</strong>redpair⟨a,b⟩ does onlywork<br />
for sets. Nevertheless we will use also pairs⟨A,B⟩ where A or B are<br />
proper classes.�ere areseveral waysto makesuch an expressionwell<strong>de</strong>fined.<br />
Asimple one <strong>is</strong>to <strong>de</strong>fine<br />
⟨A,B⟩∶=({[0]}×A)∪({[1]}×B) (= A⊍B)<br />
whenever at least one ofA andB<strong>is</strong> aproper class.(�e operation⊍will<br />
be <strong>de</strong>fined more generally in the next section.) It <strong>is</strong> easy to check that<br />
withth<strong>is</strong> <strong>de</strong>finitiontheterm⟨A,B⟩ hastheproperties of an or<strong>de</strong>redpair,<br />
that <strong>is</strong>, A⊍B=C⊍DimpliesA= C and B=D.<br />
Definition1.4. (a) Forsets a0,... ,an we <strong>de</strong>fine inductively<br />
⟨⟩∶=∅ , ⟨a0⟩∶= a0 ,<br />
and ⟨a0,... ,an⟩∶=⟨⟨a0,... ,an−1⟩,an⟩.<br />
We call⟨a0,... ,an−1⟩ a<strong>tu</strong>ple of length n.⟨⟩ <strong>is</strong>theempty<strong>tu</strong>ple.<br />
(b) For a class A,we <strong>de</strong>fine its n-thpower by<br />
A 0 ∶={⟨⟩} , A 1 ∶= A , and A n+1 ∶= A n ×A , for n> 1.<br />
Definition1.5. Arelation, or apredicate, ofarity n <strong>is</strong> asubclassR⊆S n .<br />
IfR⊆A n , forsome class A,wesaythatR<strong>is</strong> over A.<br />
Notethat∅and{⟨⟩} arethe onlyrelations of arity0. In logicthey are<br />
usually interpreted as false andtrue. Arelation of arity 1 over A <strong>is</strong> just a<br />
subclassR⊆A.<br />
Definition1.6. LetRbe a binary relation.�edomain ofR<strong>is</strong>the class<br />
28<br />
domR∶={a∣⟨a,b⟩∈ R forsomeb},
and itsrange <strong>is</strong><br />
rngR∶={b∣⟨a,b⟩∈ R forsome a}.<br />
�e field of R <strong>is</strong> domR∪rngR.<br />
1. Relationsand functions<br />
Inparticular, domR andrngR arethesmallest classessuchthat<br />
R⊆domR×rngR.<br />
Definition 1.7. (a) A binary relation R <strong>is</strong> called functional if, for every<br />
a∈ domR,there ex<strong>is</strong>ts exactly onesetbsuchthat⟨a,b⟩∈ R.We <strong>de</strong>note<br />
th<strong>is</strong>unique elementbby R(a). Hence,we canwrite R as<br />
R={⟨a,R(a)⟩∣ a∈ domR}.<br />
A functional relation R⊆A×B<strong>is</strong> also called apartial function from A<br />
toB.<br />
(b) A function from A to B <strong>is</strong> a functional relation f ⊆ A× B with<br />
dom f= A and rng f⊆ B. Functions are also called maps or mappings.<br />
Wewrite f∶ A→ Bto <strong>de</strong>notethe factthat f <strong>is</strong> a function fromAtoB.<br />
A function ofarity n <strong>is</strong> a function ofthe form<br />
f∶ A0×⋯×An−1→ B.<br />
We will write x↦y to <strong>de</strong>note the function f such that f(x)= y.<br />
(Usually, y will be an expression <strong>de</strong>pending onx.)<br />
(c) For asetaand a classB,we <strong>de</strong>note byB a the class of all functions<br />
f∶ a→B.<br />
Remark. A 0-ary function f ∶ A 0 → B <strong>is</strong> uniquely <strong>de</strong>termined by the<br />
value f(⟨⟩).Wewill callsuch functionsconstants and i<strong>de</strong>ntifythemwith<br />
their only value.<br />
Sometimes wewrite binary relations and functions in infix notation,<br />
that <strong>is</strong>, for arelationR∈A×A,wewritea R b instead of⟨a,b⟩∈ R and,<br />
for f∶ A×A→ A,wewrite a f b instead of f(a,b).<br />
29
a2. Relations<br />
Definition 1.8. (a) For every class A, we <strong>de</strong>fine the i<strong>de</strong>ntity function<br />
idA∶A→ A by idA(a)∶= a.<br />
(b) If R⊆A× B and S⊆B×C are relations, we can <strong>de</strong>fine their<br />
composition S○R∶A×C by<br />
S○R∶={⟨a,c⟩∣there <strong>is</strong>someb∈Bsuchthat<br />
⟨a,b⟩∈ R and⟨b,c⟩∈ S}.<br />
(Note the reversal of the or<strong>de</strong>ring.) In particular, if f ∶ A→ B and g∶<br />
B→C are functionsthen<br />
(g○ f)(x)∶= g( f(x)).<br />
(c)�e inverse of arelation R⊆A×B <strong>is</strong>therelation<br />
R −1 ∶={⟨b,a⟩∣⟨a,b⟩∈ R}.<br />
Inparticular, a function g∶B→A <strong>is</strong>the inverse ofthe function f∶ A→<br />
B if<br />
g( f(a))= a and f(g(b))=b , for all a∈A and b∈B ,<br />
that <strong>is</strong>, if g○ f= idA and f○ g= idB.<br />
Forb∈B,wewillwrite<br />
R −1 (b)∶={a∣⟨a,b⟩∈ R}.<br />
Notethat, ifR −1 <strong>is</strong> a function,we have already <strong>de</strong>fined<br />
R −1 (b)∶= a fortheunique asuchthat⟨a,b⟩∈ R.<br />
Itshould always be clear fromthe contextwhich ofthesetwo <strong>de</strong>finitions<br />
we have in mindwhenwewriteR −1 (b).<br />
(d)�erestriction of arelation R⊆A×Bto a classC <strong>is</strong>therelation<br />
30<br />
R∣C∶= R∩(C×C).
Its le�restriction <strong>is</strong><br />
R↾C∶= R∩(C×B) .<br />
1. Relationsand functions<br />
(e) �e image of a class C un<strong>de</strong>r a binary relation R⊆A× B <strong>is</strong> the<br />
class<br />
R[C]∶=rng(R↾C).<br />
Remark. �esetA A togetherwiththe operation○ forms a monoid,that<br />
<strong>is</strong>,○<strong>is</strong>associative<br />
f○(g○h)=( f○ g)○ h , for all f , g, h∈A A ,<br />
andthere ex<strong>is</strong>ts a neutralelement<br />
idA○ f=f and f○ idA= f for all f∈ A A .<br />
Exerc<strong>is</strong>e1.1. Is ittruethatR −1 ○R=idA, for allrelations R⊆A×B?<br />
Exerc<strong>is</strong>e1.2. Provethat○<strong>is</strong> associative andthat idA <strong>is</strong> a neutral element.<br />
Definition1.9. Let f∶ A→ B be a function.<br />
(a) f <strong>is</strong> injective ifthere <strong>is</strong> nopair a,a ′ ∈ A of d<strong>is</strong>tinct elementssuch<br />
that f(a)= f(a ′ ).<br />
(b) f <strong>is</strong>surjective ifrng f= B.<br />
(c) f <strong>is</strong> called bijective if it <strong>is</strong> both injective andsurjective.<br />
Lemma1.10. Let f∶ A→ Bbeafunction.<br />
(a) �e followingstatements areequivalent:<br />
(1) f <strong>is</strong>bijective.<br />
(2) f −1 <strong>is</strong>afunction B→A.<br />
(3) �ereex<strong>is</strong>ts a function g∶B→Asuchthat g○ f= idA and<br />
f○ g= idB.<br />
31
a2. Relations<br />
(b) �e followingstatements areequivalent:<br />
(1) f <strong>is</strong> injective.<br />
(2) f○ g= f○ h implies g= h, forall functions g, h∶C→ A.<br />
(3) A=∅ or there ex<strong>is</strong>ts some function g∶B→A such that<br />
g○ f= idA.<br />
(4) f −1 [ f[X]]= X, forall X⊆A.<br />
(c) �e followingstatements areequivalent:<br />
(1) f <strong>is</strong>surjective.<br />
(2) g○ f= h○ f implies g= h, forall functions g, h∶B→C.<br />
(3) f[ f −1 [Y]]=Y, forallY⊆ B.<br />
(d) Ifthere ex<strong>is</strong>ts some function g∶B→Asuchthat f○ g= idB then<br />
f <strong>is</strong>surjective.<br />
Proof. (a) (1)⇒(2) Let b∈B. Since f <strong>is</strong> surjective there ex<strong>is</strong>ts some<br />
a∈Asuchthat f(a)=b. Ifa ′ ∈ A <strong>is</strong>some elementwith f(a ′ )=bthen<br />
the injectivity of f implies that a ′ = a. We have shown that, for every<br />
element b∈B, there <strong>is</strong> a unique a∈A such that f −1 (b)= a. Hence,<br />
f −1 <strong>is</strong> functional and dom f −1 = B.<br />
(2)⇒(3) f −1 ∶ B→A <strong>is</strong> a function and we have f −1 ○ f = idA and<br />
f○ f −1 = idB.<br />
(3)⇒(1) If f(a)= f(b), for a,b∈ A, then<br />
a= idA(a)=(g○ f)(a)=(g○ f)(b)=idA(b)=b .<br />
Consequently, f <strong>is</strong> injective. To show that it <strong>is</strong> also surjective let b∈B.<br />
Setting a∶= g(b)we have<br />
f(a)=( f○ g)(b)=idB(b)=b.<br />
Hence,b∈rng f.<br />
(b)(1)⇒(4) Let X⊆ A. For every a∈X,we have f(a)∈ f[X] and,<br />
therefore, a∈ f −1 [ f[X]]. Consequently, X⊆ f −1 [ f[X]]. Conversely,<br />
32
1. Relationsand functions<br />
supposethat a∈ f −1 [ f[X]] and set b∶= f(a).Since b∈ f[X] there <strong>is</strong><br />
some c∈Xwith f(c)=b. As f <strong>is</strong> injective th<strong>is</strong> impliesthat a=c∈X.<br />
(4)⇒(3) If A=∅ then there <strong>is</strong> nothing to do. Hence, assume that<br />
A≠∅.We <strong>de</strong>fine g as follows. For everyb∈rng f ,there <strong>is</strong>some element<br />
a∈Awith f(a)=b.Since f −1 (b)= f −1 [ f[{a}]]={a} it followsthat<br />
th<strong>is</strong> element a <strong>is</strong>unique. Hence, fixing a0∈Awe can <strong>de</strong>fine g by<br />
⎧⎪ a if f<br />
g(b)∶= ⎨<br />
⎪⎩<br />
−1 (b)={a} ,<br />
a0 ifb∉rng f .<br />
(3)⇒(2) IfA=∅, there are no functionsC→A andthe claim holds<br />
trivially. Hence, assume that A≠∅ and let k be a function such that<br />
k○ f= idA.�en f○ g= f○ h implies<br />
g= idA○g=k○ f○ g=k○ f○ h= idA○h=h.<br />
(2)⇒(1)Supposethat f <strong>is</strong> not injective. �enthere aretwo elements<br />
a,b∈A with a≠b such that f(a)= f(b). Let C∶=[1]={0} be a<br />
set with a single element and <strong>de</strong>fine g, h∶C→A by g(0)∶= a and<br />
h(0)∶= b.�en g≠h but f○ g= f○ h.<br />
(c) (1)⇒(2) Suppose that g≠h. �ere <strong>is</strong> some element b∈B with<br />
g(b)≠ h(b). Since f <strong>is</strong> surjective we can find an element a∈A with<br />
f(a)=b. Hence,(g○ f)(a)= g(b)≠ h(b)=(h○ f)(a).<br />
(2)⇒(1)Supposethat f <strong>is</strong> notsurjective.�enthere <strong>is</strong>some element<br />
b∈B∖rng f. LetC∶=[2]={0, 1} be asetwithtwo elements and <strong>de</strong>fine<br />
g, h∶B→Cby<br />
⎧⎪ 1 if x= b ,<br />
g(x)∶= ⎨<br />
⎪⎩<br />
0 otherw<strong>is</strong>e,<br />
and h(x)∶=0 , for all x∈ B.<br />
�enwe have g≠h but g○ f= h○ f.<br />
(3)⇒(1) f[ f −1 [B]]=B implies that rng f= B.<br />
(1)⇒(3) Let Y⊆ B. If b∈ f[ f −1 [Y]] then there <strong>is</strong> some a∈ f −1 [Y]<br />
with f(a)=b. Hence, a∈ f −1 [Y] implies that b= f(a)∈Y. Consequently,<br />
we have f[ f −1 [Y]]⊆Y.<br />
33
a2. Relations<br />
For the converse, let b∈Y. Since f <strong>is</strong> surjective there <strong>is</strong> some a∈<br />
A with f(a)=b. Hence, a∈ f −1 [Y] and it follows that b= f(a)∈<br />
f[ f −1 [Y]].<br />
(d) Let k be a functionsuchthat f○k= idB.�en g○ f= h○ f implies<br />
g=g○ idB=g○ f○ k=h○ f○ k=h○ idB= h.<br />
By(c), it followsthat f <strong>is</strong>surjective. ◻<br />
Remark. �e converse of(d) also holds butwe cannotprove itwithout<br />
the Axiom of Choice, which we will introduce in Section a4.1 below.<br />
Ac<strong>tu</strong>ally one can prove that the Axiom of Choice <strong>is</strong> equivalent to the<br />
claimthat, for everysurjective function f ,there ex<strong>is</strong>tssome function g<br />
with f○ g= id.<br />
Remark. �esubset of all bijective functions f∶ A→ A forms a group<br />
since, bythepreceding lemma, every element f has an inverse f −1 .<br />
Exerc<strong>is</strong>e1.3. Let f∶ A→ B and g∶B→Cbe functions. Prove that, if<br />
f and g are(a) injective, (b)surjective, or(c) bijective thenso <strong>is</strong> g○ f.<br />
We conclu<strong>de</strong>th<strong>is</strong> section with two important results about the ex<strong>is</strong>tence<br />
of functions. �e first one can be used to prove that there ex<strong>is</strong>ts<br />
a bijection between two given sets without constructing th<strong>is</strong> function<br />
explicitly.<br />
Lemma 1.11. Let A⊆ B⊆C be sets. If there ex<strong>is</strong>ts a bijective function<br />
f∶ C→A,there <strong>is</strong>alsoabijection g∶ C→B.<br />
Proof. Let<br />
Z∶=⋂{X⊆ C∣ C∖B⊆ X and f[X]⊆X}.<br />
�enC∖B⊆Zand f[Z]⊆ Z.We claimthat<br />
34<br />
⎧⎪ f(x) if x∈Z ,<br />
g(x)∶= ⎨<br />
⎪⎩<br />
x otherw<strong>is</strong>e,
C<br />
B<br />
A<br />
Z<br />
Y<br />
1. Relationsand functions<br />
f<br />
id<br />
Figure 1..�eproof of Lemma 1.11.<br />
Z∩B<br />
<strong>is</strong> the <strong>de</strong>sired bijection g∶ C→B.<br />
LetY∶= C∖Z be the complement ofZ. Since g[Y]⊆ Y and g[Z]⊆ Z<br />
it <strong>is</strong>sufficienttoshowthattherestrictions g↾Y∶Y→ Y and g↾Z∶Z→<br />
Z∩B are bijections. Clearly, g↾Y= idY <strong>is</strong> bijective and g↾Z= f↾Z <strong>is</strong><br />
injective. �erefore,we only needtoprovethat f[Z]= Z∩B.<br />
By <strong>de</strong>finition of Z,we have f[Z]⊆Z∩rng f⊆ Z∩B. For the other<br />
inclusion, supposethat there ex<strong>is</strong>ts some element a∈(Z∩B)∖ f[Z].<br />
Since a∈B theset X∶= Z∖{a}sat<strong>is</strong>fies C∖B⊆Xand f[X]⊆ X. By<br />
<strong>de</strong>finition of Z, it followsthat Z⊆X. Contradiction. ◻<br />
�eorem1.12(Bernstein). Ifthereare injective functions f∶ A→ Band<br />
g∶B→Athenthere ex<strong>is</strong>tsabijective function h∶A→ B.<br />
Proof. We have g[ f[A]]⊆g[B]⊆A. Since f and g are injective so<br />
<strong>is</strong> their composition g○ f. When regar<strong>de</strong>d as function g○ f ∶ A→<br />
g[ f[A]] it <strong>is</strong> alsosurjective. Hence, bythepreceding lemma,there ex<strong>is</strong>ts<br />
a bijective mapping h∶A→ g[B].Since k∶= g −1 ↾ g[B]∶ g[B]→ B <strong>is</strong><br />
bijective it followsthatso <strong>is</strong> k○h∶A→ B. ◻<br />
�e second result <strong>de</strong>als with functions between a set and its power<br />
set.<br />
�eorem1.13(Cantor). Foreveryset a,thereex<strong>is</strong>tsan injective function<br />
a→℘(a)but nosurjective one.<br />
Y<br />
35
a2. Relations<br />
Proof. �e function f∶ a→℘(a)with f(x)∶={x} <strong>is</strong> injective.<br />
For a contradiction, suppose that there <strong>is</strong> also a surjective function<br />
f∶ a→℘(a).We <strong>de</strong>finetheset<br />
z∶={x∈ a∣x∉ f(x)}⊆ a.<br />
Since f <strong>is</strong> surjective there <strong>is</strong> some element b∈a with f(b)=z. By<br />
<strong>de</strong>finition ofz,we have<br />
b∈ z iff b∉ f(b)=z.<br />
A contradiction. ◻<br />
Corollary1.14. Forallsets a,thereare no injective functions℘(a)→ a.<br />
Proof. Suppose that f ∶℘(a)→ a <strong>is</strong> injective. We <strong>de</strong>fine a function<br />
g∶a→℘(a) by<br />
⎧⎪ f<br />
g(x)∶= ⎨<br />
⎪⎩<br />
−1 (x) if x∈rng f ,<br />
∅ otherw<strong>is</strong>e.<br />
Note that g <strong>is</strong> well-<strong>de</strong>fined since f <strong>is</strong> injective. Furthermore, we have<br />
g○ f= id ℘(A). Hence, Lemma 1.10(d) impliesthat g <strong>is</strong>surjective. �<strong>is</strong><br />
contradicts the�eorem of Cantor. ◻<br />
2. Productsandunions<br />
So far, we have <strong>de</strong>fined cartesian products of finitely many sets and<br />
<strong>tu</strong>ples of finite length. In th<strong>is</strong> section we will show how to general<strong>is</strong>e<br />
these <strong>de</strong>finitions to infinitely many factors.<br />
Remark. (a) �ere <strong>is</strong> a canonical bijection π∶A [n] → A n between the<br />
setA [n] of all functions[n]→A andthe n-thpowerA n of A.πmaps a<br />
function f∶[n]→Atothe<strong>tu</strong>ple<br />
36<br />
π( f)∶=⟨ f(0),..., f(n− 1)⟩ ,
2. Productsandunions<br />
and its inverseπ −1 maps a <strong>tu</strong>ple⟨a0,... ,an−1⟩tothe function f∶[n]→<br />
Awith f(i)= ai.<br />
(b)�ere <strong>is</strong> also a canonical bijection π∶(A×B)×C→A×(B×C)<br />
<strong>de</strong>fined by<br />
π⟨⟨a,b⟩,c⟩∶=⟨a,⟨b,c⟩⟩.<br />
(c) Finally, le<strong>tu</strong>s <strong>de</strong>fine a canonical bijection π∶A B×C →(A C ) B that<br />
maps a function f∶ B×C→Atothe function g∶B→A C with<br />
g(b)∶= hb where hb(c)∶= f(b,c) , forb∈B, c∈C.<br />
Inthetheory ofprogramming languages th<strong>is</strong>transformation of a function<br />
B×C→A into a function B→A C <strong>is</strong> calledcurrying.<br />
Part (a) of the above remark gives a hint on how to general<strong>is</strong>e finite<br />
<strong>tu</strong>ples. A<strong>tu</strong>ple of length n correspondsto a map[n]→A.�erefore,we<br />
<strong>de</strong>fine an infinite<strong>tu</strong>ple as mapN→ A.<br />
Definition 2.1. (a) Let A be a class and I aset. A function f∶ I→A <strong>is</strong><br />
called asequence, or family, over I. If f(i)= ai then we alsowrite f in<br />
the form(ai)i∈I.<br />
(b) Let I be aset,(Ai)i∈I asequence ofsets, and A∶=⋃{Ai∣ i∈I}<br />
theirunion.�eproduct of(Ai)i∈I <strong>is</strong>the class<br />
∏Ai∶={ f∈ A<br />
i∈I<br />
I ∣ f(i)∈ Ai for all i}.<br />
(c) Let(Ai)i∈I be asequence of sets and k∈I.�e projection tothe<br />
k-th coordinate <strong>is</strong>the map<br />
pr k ∶∏ i∈I<br />
Ai→ Ak with pr k ( f)∶= f(k).<br />
Remark. (a) IfAi= A, for all i∈I,then∏i∈I Ai= A I .<br />
(b) As we have seen above there <strong>is</strong> a canonical bijection between<br />
A0×A1 and∏i∈[2]Ai. Inthe followingwewill not d<strong>is</strong>tingu<strong>is</strong>h between<br />
thesesets.<br />
37
a2. Relations<br />
Le<strong>tu</strong>s introducesome notation and conventionsregardingsequences.<br />
To indicate that a certain variable refers to a sequence we will write it<br />
with a bar ā. Ifthesequence <strong>is</strong> over I,the components of ā will always<br />
be(ai)i∈I. Sometimes we will not d<strong>is</strong>tingu<strong>is</strong>h between a sequence ā=<br />
(ai)i∈I and its range rngā={ai∣ i∈I}. Inparticular, wewrite ā∪ ¯ b<br />
instead ofrngā∪rng ¯ b and, ifwe do notwanttospecifythe in<strong>de</strong>xset I,<br />
wewill write ā⊆A instead of ā∈A I . Finally, for a function f∶ A→ B,<br />
wewrite f(ā)to <strong>de</strong>notethesequence( f(ai))i∈I.<br />
Lemma 2.2. Let A be a set and(Bi)i∈I a sequence of sets. For every sequence(<br />
fi)i∈I of functions fi ∶ A→ Bi there ex<strong>is</strong>ts a unique function<br />
g∶ A→∏i Bi suchthat<br />
pr i ○ g= fi , forall i∈I.<br />
Proof. �e function<br />
g(a)∶=( fi(a))i∈I<br />
has obviously the <strong>de</strong>siredproperties.We havetoshowthat it <strong>is</strong>unique.<br />
Let h∶A→∏i Bi be another such function. If g≠h, there <strong>is</strong> some<br />
element a∈A such that g(a)≠ h(a). Let(bi)i∈I∶= h(a). For every<br />
i∈I,we have<br />
bi=(pr i ○ h)(a)= fi(a).<br />
Hence g(a)=( fi(a))i=(bi)i= h(a). A contradiction. ◻<br />
Definition 2.3. �e d<strong>is</strong>joint union of a sequence(Ai)i∈I of sets <strong>is</strong> the<br />
class<br />
⊍Ai∶={⟨i,a⟩∣ i∈I, a∈Ai}.<br />
i∈I<br />
Similarly, if A and B are classesthenwe can <strong>de</strong>finetheir d<strong>is</strong>joint union<br />
as<br />
38<br />
A⊍B∶=({[0]}×A)∪({[1]}×B).
�e k-th insertion <strong>is</strong> the canonical map<br />
ink∶ Ak→⊍ Ai with ink(a)∶=⟨k,a⟩ .<br />
i∈I<br />
Remark. IfAi= A, for all i∈I, then⊍i∈I Ai= I×A.<br />
3. Graphsandpartial or<strong>de</strong>rs<br />
Lemma 2.4. Let B be a set and(Ai)i∈I a sequence of sets. For every sequence(<br />
fi)i∈I of functions fi ∶ Ai → B there ex<strong>is</strong>ts a unique function<br />
g∶⊍i Ai→ Bsuchthat<br />
g○ ini=fi , forall i∈I .<br />
Proof. �e function<br />
g⟨i,a⟩∶= fi(a)<br />
has obviously the <strong>de</strong>siredproperties.We havetoshowthat it <strong>is</strong>unique.<br />
Let h∶⊍i Ai→ B be anothersuch function. If g≠hthenthere <strong>is</strong>some<br />
element⟨k,a⟩∈⊍i Ai suchthat g⟨k,a⟩≠ h⟨k,a⟩.We have<br />
h⟨k,a⟩=(h○ ink)(a)= fk(a)= g⟨k,a⟩.<br />
A contradiction. ◻<br />
3. Graphsandpartial or<strong>de</strong>rs<br />
When consi<strong>de</strong>ringrelations it <strong>is</strong> frequently necessarytospecifythesets<br />
they are over.<br />
Definition 3.1. A graph <strong>is</strong> a pair⟨A,R⟩ where R⊆A× A <strong>is</strong> a binary<br />
relation onA.<br />
More generally one can consi<strong>de</strong>rsetstogetherwithseveralrelations and<br />
functions. �<strong>is</strong>will leadtothe notion of astruc<strong>tu</strong>re in Chapter b1.<br />
Definition 3.2. Let⟨A,R⟩ be a graph.<br />
39
a2. Relations<br />
(a) R <strong>is</strong>reflexive if⟨a,a⟩∈ R, for all a∈A.<br />
(b) R <strong>is</strong> irreflexive if⟨a,a⟩∉ R, for all a∈A.<br />
(c) R <strong>is</strong>symmetric ifwe have⟨a,b⟩∈ R if, and only if,⟨b,a⟩∈ R, for<br />
all a,b∈A.<br />
(d) R <strong>is</strong>ant<strong>is</strong>ymmetric if⟨a,b⟩∈ R and⟨b,a⟩∈ R implies a=b.<br />
(e) R <strong>is</strong>transitive if⟨a,b⟩∈ R and⟨b,c⟩∈ R implies⟨a,c⟩∈ R, for all<br />
a,b,c∈A.<br />
Note that, for the <strong>de</strong>finition of reflexivity, it <strong>is</strong> important to specify<br />
the setA. If⟨A,R⟩ <strong>is</strong> reflexive and A⊂ B then⟨B,R⟩ <strong>is</strong> not reflexive.<br />
Example. (a) �e relation A×A <strong>is</strong> reflexive, symmetric, and transitive.<br />
It <strong>is</strong> irreflexive if, and only if,A=∅, and it <strong>is</strong> ant<strong>is</strong>ymmetric if, and only<br />
if,A contains at most one element.<br />
(b) �e diagonal idA={⟨a,a⟩∣ a∈A} <strong>is</strong> reflexive, symmetric, ant<strong>is</strong>ymmetric,<br />
and transitive. It <strong>is</strong> irreflexive if, and only if,A=∅.<br />
(c) �e empty relation∅⊆A×A<strong>is</strong> irreflexive, symmetric, ant<strong>is</strong>ymmetric,<br />
and transitive. It <strong>is</strong> reflexive if, and only if, A=∅.<br />
Definition 3.3. (a) A (non-strict) partial or<strong>de</strong>r <strong>is</strong> a graph⟨A,≤⟩where<br />
≤ <strong>is</strong> reflexive, transitive, and ant<strong>is</strong>ymmetric.<br />
(b) A strict partial or<strong>de</strong>r <strong>is</strong> a graph⟨A,
3. Graphsandpartial or<strong>de</strong>rs<br />
Similarly, if
a2. Relations<br />
the greatest element of X by maxA X and the least element by minAX,<br />
provi<strong>de</strong>dthese elements ex<strong>is</strong>t.<br />
(d) Let X⊆A. Wesay that a <strong>is</strong> an upper bound of X if x≤a, for all<br />
x∈X. If a <strong>is</strong> an upper bound of X and a≤b, for every other upper<br />
boundbof X,then a <strong>is</strong>the leas<strong>tu</strong>pperbound, orsupremum, of X. Ifthe<br />
leas<strong>tu</strong>pper bound of X ex<strong>is</strong>ts,we <strong>de</strong>note it bysup A X.<br />
�e notion of a (greatest) lower bound <strong>is</strong> <strong>de</strong>fined analogously. �e<br />
greatest lower bound <strong>is</strong> also called the infimum of X. We <strong>de</strong>note it by<br />
infA X. Ifthe or<strong>de</strong>r A <strong>is</strong>un<strong>de</strong>rstoodwewill omitthesubscript A andwe<br />
justwritesupX and inf X.<br />
(e) A linearly or<strong>de</strong>redsubsetC⊆A <strong>is</strong> called achain.<br />
Example. (a) Let Q∶=⟨Q,≤⟩. �e set I∶={x∈ Q∣x< √ 2} <strong>is</strong> an<br />
initial segment of Q. Every rational number y> √ 2 <strong>is</strong> anupper bound<br />
of I but I has no leas<strong>tu</strong>pper bound.<br />
(b) Consi<strong>de</strong>r⟨N,∣⟩. Its least element <strong>is</strong>the number 1 and its greatest<br />
element <strong>is</strong> 0. �e least upper bound of two elements[k],[m]∈N <strong>is</strong><br />
their least common multiple lcm(k, m), andtheir greatest lower bound<br />
<strong>is</strong>their greatest common div<strong>is</strong>or gcd(k, m).�eset P⊆N of all prime<br />
numbers has the least upper bound 0 and the greatest lower bound 1.<br />
�eset{2 n ∣ n∈N} of allpowers oftwo forms a chain.<br />
Exerc<strong>is</strong>e 3.1. Consi<strong>de</strong>r⟨B,⊆⟩where<br />
B∶={X⊆N∣ X <strong>is</strong> finite orN∖X <strong>is</strong> finite}.<br />
(a) Construct aset X⊆Bthat has no minimal element.<br />
(b) Construct aset X⊆Bwith lower bounds butwithout infimum.<br />
Lemma 3.5. Let⟨A,≤⟩beapartial or<strong>de</strong>r. IfA <strong>is</strong>aset,the followingstatementsareequivalent:<br />
42<br />
(1) Everysubset X⊆A hasasupremum.<br />
(2) Everysubset X⊆A hasan infimum.
3. Graphsandpartial or<strong>de</strong>rs<br />
Proof. We only prove (1)⇒(2). �e other direction follows in exactly<br />
thesameway. Let X⊆A andset<br />
C∶={a∈A∣ a <strong>is</strong> a lower bound of X}.<br />
By assumption, c∶= supC ex<strong>is</strong>ts. We claim that inf X=c. Let b∈X.<br />
By <strong>de</strong>finition, we have a≤b, for all a∈C. Hence, b <strong>is</strong> anupper bound<br />
of C andwe haveb≥supC= c. Asb was arbitrary it followsthat c <strong>is</strong> a<br />
lower bound of X. If a <strong>is</strong> an arbitrary lower bound of X,we have a∈C,<br />
which implies that a≤c. Consequently, c <strong>is</strong> the greatest lower bound<br />
of X. ◻<br />
Definition 3.6. Apartial or<strong>de</strong>r⟨A,≤⟩ <strong>is</strong>complete if everysubset X⊆A<br />
has an infimum and asupremum.<br />
Remark. Every complete partial or<strong>de</strong>r has a least element�∶= sup∅<br />
and a greatest element⊺∶= inf∅.<br />
Example. (a) Let A be aset. �epartial or<strong>de</strong>r⟨℘(A),⊆⟩ <strong>is</strong> complete. If<br />
X⊆℘(A)then<br />
supX=⋃X∈℘(A) and inf X=⋂X∈℘(A).<br />
(b)�e or<strong>de</strong>r⟨R,≤⟩ <strong>is</strong> complete.⟨Q,≤⟩ <strong>is</strong> notsincetheset<br />
{x∈ Q∣x≤π}<br />
has no leas<strong>tu</strong>pper bound inQ.<br />
(c) �e or<strong>de</strong>r⟨N,≤⟩ <strong>is</strong> not complete since inf∅ and supN do not<br />
ex<strong>is</strong>t.<br />
(d) Let A=⟨A,≤⟩ be an arbitrary partial or<strong>de</strong>r. We can construct a<br />
completepartial or<strong>de</strong>r C=⟨C,⊆⟩ containing A as follows. LetC⊆℘(A)<br />
betheset of all initial segments of A or<strong>de</strong>red by inclusion. �e <strong>de</strong>sired<br />
embedding f∶ A→ C <strong>is</strong> given by f(a)∶=⇓Aa.<br />
43
a2. Relations<br />
Nextwe <strong>tu</strong>rntothes<strong>tu</strong>dy of functions betweenpartial or<strong>de</strong>rs. Inparticular,wewill<br />
consi<strong>de</strong>r functions f∶ A→ A mapping onepartial or<strong>de</strong>r<br />
into itself.Tosimplify notation, wewillwrite<br />
f∶ A→B ,<br />
for partial or<strong>de</strong>rs A=⟨A,≤A⟩ and B=⟨B,≤B⟩, to <strong>de</strong>note that f <strong>is</strong> a<br />
function f∶ A→ B.<br />
Definition 3.7. Let A=⟨A,≤A⟩ and B=⟨B,≤B⟩ bepartial or<strong>de</strong>rs.<br />
(a) A function f∶ A→ B <strong>is</strong> increasing if<br />
a≤A b implies f(a)≤B f(b) , for all a,b∈ A ,<br />
and f <strong>is</strong>strictly increasing if<br />
a
3. Graphsandpartial or<strong>de</strong>rs<br />
Lemma 3.8. Let⟨A,≤A⟩and⟨B,≤B⟩bepartial or<strong>de</strong>rsand h∶A→ Ban<br />
increasing function. Let C⊆Aand a∈A.<br />
(a) If a <strong>is</strong>anupperbound ofCthen h(a) <strong>is</strong>anupperbound of h[C].<br />
(b) If a <strong>is</strong>alowerbound ofC then h(a) <strong>is</strong>alowerbound of h[C].<br />
Lemma 3.9. Let⟨A,≤A⟩and⟨B,≤B⟩bepartial or<strong>de</strong>rsand h∶A→ Ban<br />
embedding. LetC⊆Aand a∈A.<br />
(a) h(a)=sup h[C] implies a=supC.<br />
(b) h(a)= inf h[C] implies a= inf C.<br />
Proof. (a)Since h <strong>is</strong> an embedding it followsthat h(c)≤B h(a) implies<br />
c≤A a, for c∈C. Hence, a <strong>is</strong> an upper bound of C. To show that it <strong>is</strong><br />
the least one, suppose that b <strong>is</strong> another upper bound of C. �en c≤A<br />
b, for c∈C, implies h(c)≤B h(b). Hence, h(b) <strong>is</strong> an upper bound<br />
of h[C].Since h(a) <strong>is</strong>the leastsuch bound it followsthat h(a)≤B h(b).<br />
Consequently, we have a≤A b, as <strong>de</strong>sired.<br />
(b) h <strong>is</strong> also an embedding of⟨A,≥A⟩ into⟨B,≥B⟩. Hence,(b) follows<br />
from(a) byreversing the or<strong>de</strong>rs. ◻<br />
Corollary 3.10. Let⟨F,⊆⟩beapartial or<strong>de</strong>rwith F⊆℘(A)andC⊆F.<br />
(a)⋃C∈FimpliessupC=⋃C.<br />
(b)⋂C∈Fimplies inf C=⋂C.<br />
Proof. We can apply Lemma 3.9 to the inclusion map F→℘(A). ◻<br />
Corollary 3.11. Let A=⟨A,≤⟩beapartial or<strong>de</strong>r. IfB⊆A <strong>is</strong>anonempty<br />
setsuchthat<br />
inf AX∈ B and sup A X∈ B , forevery nonempty X⊆B ,<br />
then B∶=⟨B,≤⟩ <strong>is</strong> a complete partial or<strong>de</strong>r where, for every nonempty<br />
subset X⊆B,we have<br />
inf BX= inf AX and sup B X=sup A X.<br />
45
a2. Relations<br />
Proof. If X⊆B <strong>is</strong> nonemptythen, applying Lemma 3.9 to the inclusion<br />
map B→A, it followsthat<br />
inf BX= infAX and sup B X=sup A X.<br />
Inparticular, inf BXandsup B X ex<strong>is</strong>t. Forthe emptyset, it followssimilarlythat<br />
inf B∅=sup B B=sup A B∈B ,<br />
and sup B ∅=inf BB=inf AB∈B.<br />
Consequently, B <strong>is</strong> complete. ◻<br />
We haveseen that although increasing functions preserve the or<strong>de</strong>ring<br />
of elementsthey do not necessarily preservesupremums and infimums.<br />
Le<strong>tu</strong>stake a look at functionsthat do.<br />
Definition 3.12. Let⟨A,≤A⟩ and⟨B,≤B⟩ be partial or<strong>de</strong>rs. A function<br />
f ∶ A→ B <strong>is</strong> continuous if, whenever a nonempty chain C⊆A has a<br />
supremumthen f[C] also has asupremum andwe have<br />
sup f[C]= f(supC).<br />
f <strong>is</strong> called strictlycontinuous if it <strong>is</strong> continuous and injective.<br />
Remark. Every (strictly) continuous function <strong>is</strong>(strictly) increasing.<br />
Exerc<strong>is</strong>e 3.4. Provethat continuous functions are increasing.<br />
Example. (a) Let⟨A,≤⟩ bethe linear or<strong>de</strong>rwhereA=N⊍Nand<br />
46<br />
⟨i,a⟩≤⟨k,b⟩ : iff i
4. Fixed pointsandclosure operators<br />
�e function f∶ A→ A∶⟨i,a⟩↦⟨i,a+1⟩ <strong>is</strong> not continuous. Consi<strong>de</strong>r<br />
the initial segment X∶={0}×N=↓⟨1,0⟩⊆ A.We have supX=⟨1,0⟩<br />
but<br />
sup f[X]=⟨1,0⟩
a2. Relations<br />
Figure2.. Fixedpoints of f(x)= 1 4x 3 − 3 4x 2 + 3 4x+3 4<br />
Definition 4.1. Let f ∶ A→A be a function. An element a∈A with<br />
f(a)= a <strong>is</strong> called a fixed point of f. �e class of all fixed points of f <strong>is</strong><br />
<strong>de</strong>noted by<br />
fix f∶={a∈A∣ f(a)= a}.<br />
We <strong>de</strong>notethe least and greatest fixedpoint of f , if it ex<strong>is</strong>ts, by<br />
lfp f∶= min fix f and gfp f∶= max fix f .<br />
Example. (a) Let⟨R,
4. Fixed pointsandclosure operators<br />
hasthe fixedpoints{0},{1},{0, 1}. It has no least fixedpoint.<br />
(d) Consi<strong>de</strong>r⟨F,⊆⟩where<br />
F∶={X⊆N∣ X orN∖X <strong>is</strong> finite}.<br />
�e function f∶ F→ F <strong>de</strong>fined by<br />
⎧⎪ X∪{1+maxX} if X <strong>is</strong> finite,<br />
f(X)∶= ⎨<br />
⎪⎩<br />
X otherw<strong>is</strong>e,<br />
has fixed points<br />
fix f={X⊆N∣N∖ X <strong>is</strong> finite} ,<br />
but no least one.<br />
Exerc<strong>is</strong>e 4.1. Let A=⟨℘(N),⊆⟩. Construct a function f∶ A→A that<br />
has a least fixedpoint but no greatest one.<br />
Not every function has fixed points. �e next theorem presents an<br />
important special casewherewe always have a least fixed point. InSection<br />
a3.3wewill collect furtherresults aboutthe ex<strong>is</strong>tence of fixedpoints<br />
and methodsto computethem.<br />
�eorem 4.2 (Knaster, Tarski). Let⟨A,≤⟩ be a complete partial or<strong>de</strong>r<br />
where A <strong>is</strong> a set. Every increasing function f ∶ A→ A has a least fixed<br />
pointandwe have<br />
lfp f= inf{a∈A∣ f(a)≤ a}.<br />
Proof. Set B∶={a∈A∣ f(a)≤ a} and b∶= inf B. For every a∈B,<br />
b≤a implies f(b)≤ f(a)≤ a,since f <strong>is</strong> increasing. Hence, f(b) <strong>is</strong> a<br />
lower bound ofBand it followsthat f(b)≤inf B=b.�<strong>is</strong> impliesthat<br />
f( f(b))≤ f(b) and, by <strong>de</strong>finition ofB, it followsthat f(b)∈ B. Hence,<br />
f(b)≥inf B=b. Consequently, we have f(b)=b andb<strong>is</strong> a fixedpoint<br />
of f.<br />
Let a be another fixed point of f. �en f(a)= a implies a∈B and<br />
we haveb= inf B≤a. Hence,b<strong>is</strong>the least fixedpoint of f. ◻<br />
49
a2. Relations<br />
�eorem4.3. Let⟨A,≤⟩beacompletepartial or<strong>de</strong>rwhereA <strong>is</strong>asetand<br />
let f∶ A→ Abe increasing.�eset F∶= fix f <strong>is</strong> nonemptyand F∶=⟨F,≤⟩<br />
formsacompletepartial or<strong>de</strong>rwhere, for X⊆ F,<br />
infF X=sup A {a∈A∣ a≤infAX and f(a)≥ a} ,<br />
sup F X= inf A{a∈A∣ a≥sup A X and f(a)≤ a}.<br />
Proof. We have already shown inthepreceding theorem that F≠∅. It<br />
remains toprove that F <strong>is</strong> complete. For X⊆ A, let U∶=⇑sup A X⊆A<br />
betheset of allupper bounds of X. If Z⊆U then<br />
sup A Z≥sup A X and inf AZ≥sup A X.<br />
It followsthatthepartial or<strong>de</strong>r⟨U,≤⟩ <strong>is</strong> complete. Furthermore, ifa∈ U<br />
andx∈Xthena≥ximplies f(a)≥ f(x). Hence, f↾U <strong>is</strong> an increasing<br />
functionU→U. By�eorem 4.2, it followsthat<br />
sup F X= lfp( f↾U)=infA{a∈U∣ f(a)≤ a} ,<br />
as <strong>de</strong>sired. �e claim for infF X follows by applying the equation for<br />
sup F X tothe dual or<strong>de</strong>r A op . ◻<br />
Example. Consi<strong>de</strong>r a closed interval[a,b]⊆Rofthereal line.<br />
(a)Sincethe or<strong>de</strong>r⟨[a,b],
4. Fixed pointsandclosure operators<br />
As a special case of �eorem 4.3we consi<strong>de</strong>r complete partial or<strong>de</strong>rs<br />
obtained via closure operators.<br />
Definition 4.4. LetA be a class.<br />
(a) Aclosure operator on A <strong>is</strong> a function c∶℘(A)→℘(A)suchthat,<br />
for all x,y∈℘(A),<br />
◆ x⊆c(x) ,<br />
◆ c(c(x))= c(x) , and<br />
◆ x⊆y impliesc(x)⊆ c(y).<br />
(b) Aset x⊆A <strong>is</strong>c-closed ifc(x)= x.<br />
(c) A closure operator c has finite character if, for all sets x⊆A, we<br />
have<br />
c(x)=⋃{c(x0)∣ x0⊆ x <strong>is</strong> finite}.<br />
If c has finite characterwe alsosaythat c <strong>is</strong>algebraic.<br />
(d) A closure operatorc <strong>is</strong>topological ifwe have<br />
◆ c(∅)=∅and<br />
◆ c(x∪ y)=c(x)∪c(y) , for all x,y∈℘(A).<br />
Remark. Letcbe a closure operator on A.<br />
(a) �e class of c-closedsets <strong>is</strong> fixc=rngc.<br />
(b) Ifthe class A <strong>is</strong> asetthen it <strong>is</strong> c-closed.<br />
Example. (a) Let V be a vector space. For X⊆V, let⟪X⟫ be the subspace<br />
ofV spanned by X.�e function X↦⟪X⟫ <strong>is</strong> a closure operator<br />
with finite character.<br />
(b) Let X be a topological space. For A⊆ X, let c(A) be the topological<br />
closure ofA in X.�enc<strong>is</strong> atopological closure operator.<br />
(c) LetA be aset and a∈A.�e functions c,d∶℘(A)→℘(A)with<br />
c(X)∶= X and d(X)∶= X∪{a}<br />
are closure operators on A.<br />
51
a2. Relations<br />
Exerc<strong>is</strong>e 4.2. Let A=⟨A,≤⟩ be apartial or<strong>de</strong>r. For X⊆A,we <strong>de</strong>fine<br />
c(X)∶={supC∣ C⊆X<strong>is</strong> a nonempty chainwithsupremum}.<br />
(a)Provethatthe function c <strong>is</strong> atopological closure operator on A.<br />
(b) Let B be asecond partial or<strong>de</strong>r and d the corresponding closure<br />
operator.Prove that a function f∶ A→B <strong>is</strong> continuous if, and only if,<br />
every d-closedset X∈ fixd has ac-closedpreimage f −1 [X]∈fixc.<br />
Exerc<strong>is</strong>e 4.3. Let⟨A,≤⟩ be apartial or<strong>de</strong>r. Forsets X⊆A,we <strong>de</strong>fine<br />
U(X)∶={a∈A∣ a <strong>is</strong> anupper bound of X} ,<br />
L(X)∶={a∈A∣ a <strong>is</strong> a lower bound of X}.<br />
Provethatthe function c∶X↦L(U(X)) <strong>is</strong> a closure operator on A.<br />
Lemma 4.5. Let c beaclosure operator on Aandx,y⊆Asets.<br />
(a) c(x)∪c(y)⊆ c(x∪ y).<br />
(b) c(x∪ y)= c(c(x)∪c(y)).<br />
Proof. (a) By monotonicity of c, we have c(x)⊆ c(x∪y) and c(y)⊆<br />
c(x∪ y).<br />
(b) It follows fromx∪y⊆c(x)∪c(y) and(a)that<br />
c(x∪ y)⊆ c(c(x)∪c(y))⊆ c(c(x∪ y))=c(x∪ y). ◻<br />
Lemma 4.6. Let c be a closure operator on A with finite character. For<br />
everychainC⊆ fixc,we have<br />
c(⋃C)=⋃C.<br />
Proof. By <strong>de</strong>finition, we have⋃C⊆c(⋃C). Forthe converse, let x0⊆<br />
⋃C be finite.SinceC <strong>is</strong> linearly or<strong>de</strong>red by⊆there ex<strong>is</strong>tssome element<br />
x∈Cwith x0⊆x. Hence,we have c(x0)⊆c(x)= x⊆⋃C. It follows<br />
that<br />
52<br />
c(⋃C)=⋃{c(x0)∣x0⊆⋃C finite}⊆⋃C. ◻
4. Fixed pointsandclosure operators<br />
If c <strong>is</strong> a closure operator, the setC ∶= fixc of c-closed sets has the<br />
followingproperties.<br />
Definition 4.7. AsetC⊆℘(A) <strong>is</strong> called asystem ofclosedsets ifwe have<br />
◆ A∈C and<br />
◆ ⋂Z∈C, for every Z⊆C.<br />
Apair⟨A,C⟩whereC⊆℘(A) <strong>is</strong> asystem of closedsets <strong>is</strong> called aclosure<br />
space.<br />
Lemma 4.8. (a) If c <strong>is</strong>aclosure operator on Athen fixc forms asystem<br />
ofclosedsets.<br />
(b) IfC⊆℘(A) <strong>is</strong>asystem ofclosedsetsthenthe mapping<br />
c∶X↦⋂{C∈C∣X⊆ C}<br />
<strong>de</strong>finesaclosure operator on Awith fixc=C.<br />
�e followingtheorem statesthatthe family of c-closedsets forms a<br />
completepartial or<strong>de</strong>r.We canuseth<strong>is</strong>resulttoprovethat a given partial<br />
or<strong>de</strong>r A <strong>is</strong> complete by <strong>de</strong>fining a closure operatorwhose closedsets<br />
are exactlythe elements of A. An example ofsuch aproof <strong>is</strong>provi<strong>de</strong>d in<br />
Corollary 4.17.<br />
�eorem 4.9. Let A be a set and c a closure operator on A. �e graph<br />
⟨F,⊆⟩with F∶= fixc forms acomplete partial or<strong>de</strong>rwith<br />
inf X=⋂X and supX=c(⋃X) , forall X⊆ F.<br />
Proof. Since closure operators are increasingwe can apply�eorem 4.3.<br />
By Lemma 4.8 (b), it follows that<br />
supX=⋂{Z⊆A∣ Z⊇⋃X and c(Z)⊆ Z}<br />
=⋂{Z⊆A∣ Z⊇⋃X and c(Z)= Z}<br />
= c(⋃X) ,<br />
53
a2. Relations<br />
and inf X=⋃{Z⊆ A∣ Z⊆⋂X and c(Z)⊇ Z}<br />
=⋃{Z⊆ A∣ Z⊆⋂X}<br />
=⋂X . ◻<br />
Corollary 4.10. Let c be a closure operator on Aand set F∶= fixc.�e<br />
operator c <strong>is</strong>continuous ifweconsi<strong>de</strong>r itasafunction<br />
c∶⟨℘(A),⊆⟩→⟨F,⊆⟩ .<br />
Proof. For a nonempty chain X⊆℘(A),we have<br />
c(supX)= c(⋃X)⊆ c(⋃c[X])=supc[X]<br />
⊆sup{c(supX)}= c(supX). ◻<br />
As an application of closure operators we consi<strong>de</strong>r equivalence relations.<br />
Definition 4.11. (a) A binaryrelation∼⊆A×A <strong>is</strong> anequivalencerelation<br />
on A if it <strong>is</strong>reflexive,symmetric, andtransitive.<br />
(b) Let∼⊆A× A be an equivalence relation. If A <strong>is</strong> a set, we <strong>de</strong>fine<br />
the∼-class of an element a∈A by<br />
[a]∼∶={b∈A∣ b∼a}.<br />
Forproper classes A,weset<br />
[a]∼∶= cut{b∈A∣ b∼a}.<br />
Notethat, <strong>de</strong>spitethe name, a∼-class <strong>is</strong> always aset.We <strong>de</strong>notethe class<br />
of all∼-classes by<br />
A/∼∶={[a]∼∣a∈A}.<br />
Example. (a)�e diagonal idA <strong>is</strong>thesmallest equivalencerelation onA.<br />
�e largest one <strong>is</strong>the fullrelation A×A.<br />
(b)�e <strong>is</strong>omorph<strong>is</strong>mrelation≅ <strong>is</strong> an equivalencerelation onthe class<br />
of allpartial or<strong>de</strong>rs.<br />
54
4. Fixed pointsandclosure operators<br />
Lemma4.12. Let∼beanequivalencerelation on Aand a,b∈ A.�en<br />
a∼b iff [a]∼=[b]∼ iff [a]∼∩[b]∼≠∅.<br />
Remark. LetA be aset. Apartition of A <strong>is</strong> asetP⊆℘(A) of nonempty<br />
subsets ofAsuchthatA=⋃P and p∩q=∅, for all p,q∈Pwith p≠q.<br />
If∼<strong>is</strong> an equivalence relation on Athen A/∼ forms apartition on A.<br />
Conversely, given apartition P of A,we can <strong>de</strong>fine an equivalence relation∼P<br />
onAwith A/∼P= P bysetting<br />
a∼P b : iff there <strong>is</strong>some p∈Pwith a,b∈ p.<br />
Definition 4.13. Let A be a set and R⊆A× A a binary relation on A.<br />
�etransitiveclosure of R <strong>is</strong>therelation<br />
TC(R)∶=⋂{S⊆ A×A∣ S⊇ R <strong>is</strong>transitive}.<br />
Since the family of transitive relations <strong>is</strong> closed un<strong>de</strong>r intersections<br />
we canuse Lemma 4.8(b)toprovethat TC <strong>is</strong> a closure operator.<br />
Lemma4.14. Let Abeaclass. TC <strong>is</strong>aclosure operator on A×A.<br />
Exerc<strong>is</strong>e 4.4. Prove Lemma 4.14.<br />
Lemma 4.15. If R⊆A×A<strong>is</strong>asymmetricrelationthenso <strong>is</strong> TC(R).<br />
Proof. LetS∶= TC(R)∩(TC(R)) −1 .SinceR<strong>is</strong>symmetricwe haveR⊆S.<br />
We claimthat S <strong>is</strong>transitive.<br />
Let⟨a,b⟩,⟨b,c⟩∈ S. �en⟨a,b⟩,⟨b,c⟩∈ TC(R) and⟨b,a⟩,⟨c,b⟩∈<br />
TC(R). �erefore, we have⟨a,c⟩∈TC(R) and⟨c,a⟩∈TC(R). �<strong>is</strong><br />
impliesthat⟨a,c⟩∈ S, as <strong>de</strong>sired.<br />
We have shown that S <strong>is</strong> a transitive relation containing R. By the<br />
<strong>de</strong>finition of TC it follows that TC(R)⊆ S=TC(R)∩TC(R) −1 . �<strong>is</strong><br />
impliesthat TC(R) −1 = TC(R). Hence, TC(R) <strong>is</strong>symmetric. ◻<br />
Lemma4.16. Let R⊆A×Abeabinaryrelation.<br />
(a) �esmallestreflexiverelationcontainingR<strong>is</strong> R∪idA.<br />
55
a2. Relations<br />
(b) �esmallestsymmetricrelationcontainingR<strong>is</strong> R∪R −1 .<br />
(c) �esmallesttransitiverelationcontaining R <strong>is</strong> TC(R).<br />
(d) �esmallestequivalencerelationcontainingR<strong>is</strong>TC(R∪R −1 ∪idA).<br />
Proof. (a) R∪idA <strong>is</strong> obviously reflexive and it contains R. Conversely,<br />
supposethat S⊇R<strong>is</strong>reflexive. �en idA⊆ S impliesthat R∪ idA⊆S.<br />
(b) <strong>is</strong>proved analogously.<br />
(c) Let S⊇R be transitive. �en the intersection in the <strong>de</strong>finition<br />
of TC contains S. Hence, TC(R)⊆ S. Furthermore, we have R⊆ TC(R)<br />
by <strong>de</strong>finition. Itremainstoprovethat TC(R) <strong>is</strong>transitive.<br />
Let⟨a,b⟩,⟨b,c⟩∈ TC(R). �en we have⟨a,b⟩,⟨b,c⟩∈ S, for every<br />
transitive relation S⊇R. Hence, we have⟨a,c⟩∈ S, for each such relation<br />
S.�<strong>is</strong> impliesthat⟨a,c⟩∈ TC(R).<br />
(d)SetE∶= TC(R∪R −1 ∪ idA). Clearly,we haveR⊆Eand, ifS⊇R <strong>is</strong><br />
an equivalencerelationthenE⊆ S. Hence, it <strong>is</strong>remainstoprovethatE <strong>is</strong><br />
an equivalencerelation. It <strong>is</strong>transitive by(c),symmetric by Lemma 4.15,<br />
and E <strong>is</strong>reflexivesince idA⊆ TC(R∪R −1 ∪ idA). ◻<br />
Corollary4.17. LetAbeasetand F⊆℘(A×A)theset ofallequivalence<br />
relations on A. �en⟨F,⊆⟩ forms a complete partial or<strong>de</strong>r. If X⊆F <strong>is</strong><br />
nonemptythenwe have<br />
inf X=⋂X and supX= TC(⋃X).<br />
Proof. By Lemma 4.16,we have F= fixcwherec <strong>is</strong>the closure operator<br />
with<br />
c(R)∶= TC(R∪R −1 ∪ idA).<br />
�e relation E∶=⋃X <strong>is</strong> reflexive and symmetric since X <strong>is</strong> nonempty.<br />
Hence, we have TC(E∪E −1 ∪ idA)=TC(E). Consequently, the claim<br />
follows from�eorem 4.9. ◻<br />
56
a3. Ordinals<br />
1. Well-or<strong>de</strong>rs<br />
When <strong>de</strong>finingstageswe frequentlyusedthe factthat any class ofstages<br />
has a minimal element. In th<strong>is</strong> section we s<strong>tu</strong>dy arbitrary or<strong>de</strong>rs with<br />
th<strong>is</strong>property.<br />
Definition1.1. Let⟨A,R⟩ be a graph.<br />
(a) An element a∈A <strong>is</strong>R-minimal if⟨b,a⟩∈ R impliesb=a.<br />
(b) ArelationR<strong>is</strong> le�-narrow ifR −1 (a) <strong>is</strong> aset, for everyseta∈rngR.<br />
(c) R <strong>is</strong>well-foun<strong>de</strong>d if every nonempty subset B⊆A contains an Rminimal<br />
element. A le�-narrow, well-foun<strong>de</strong>d linear or<strong>de</strong>r <strong>is</strong> called a<br />
well-or<strong>de</strong>r.<br />
Example. (a)⟨N,≤⟩ <strong>is</strong> awell-or<strong>de</strong>r.<br />
(b)⟨N,∣⟩ <strong>is</strong> awell-foun<strong>de</strong>dpartial or<strong>de</strong>r.<br />
(c)�e membershiprelation∈<strong>is</strong> awell-foun<strong>de</strong>dpartial or<strong>de</strong>r onS. It<br />
<strong>is</strong> awell-or<strong>de</strong>r onthe class of allstages.<br />
(d)⟨℘(N),⊆⟩ <strong>is</strong> notwell-foun<strong>de</strong>d.<br />
(e) Apartial or<strong>de</strong>r⟨A,≤⟩ <strong>is</strong> le�-narrow if, and only if,⇓a <strong>is</strong> aset, for<br />
all a∈A.<br />
Exerc<strong>is</strong>e1.1. Provethat⟨℘(N),⊆⟩ <strong>is</strong> notwell-foun<strong>de</strong>d.<br />
Lemma1.2. If⟨A,R⟩ <strong>is</strong>awell-foun<strong>de</strong>d graphandB⊆Athen⟨B,R∣B⟩ <strong>is</strong><br />
alsowell-foun<strong>de</strong>d.<br />
Proof. Every nonempty subset C⊆B <strong>is</strong> also a nonempty subset of A<br />
and has an R-minimal element. ◻<br />
logic, algebra & geometry2012-08-08 — ©achim blumensath 57
a3. Ordinals<br />
Lemma 1.3. If ⟨A,≤⟩ <strong>is</strong> a well-foun<strong>de</strong>d and le�-narrow partial or<strong>de</strong>r,<br />
there ex<strong>is</strong>ts no infinite sequence(an)n∈N∈A N such that an≠an+1 and<br />
an+1≤ an, forall n.<br />
Proof. If there ex<strong>is</strong>ts such an infinite sequence then the class rngā=<br />
{an∣ n∈N} <strong>is</strong> nonempty and has no≤-minimal element. Furthermore,<br />
rngā⊆⇓a0 <strong>is</strong> asetsincethe or<strong>de</strong>r <strong>is</strong> le�-narrow. ◻<br />
�ereasonwhywell-foun<strong>de</strong>drelations are of interest <strong>is</strong>thatthese are<br />
exactly thoserelations that admitproofs by induction. Asthe theorem<br />
belowshowswe canprovethat every element of awell-foun<strong>de</strong>dpartial<br />
or<strong>de</strong>r⟨A,≤⟩ sat<strong>is</strong>fies a given property φ by showing that, if every elementb
1. Well-or<strong>de</strong>rs<br />
Example. Consi<strong>de</strong>r the well-or<strong>de</strong>r⟨N,
a3. Ordinals<br />
Proof. Suppose that there ex<strong>is</strong>ts some a∈Awith a> f(a). Let a0 be<br />
the minimalsuch element. By minimality of a0 we have<br />
f(a0)≤ f( f(a0)).<br />
Onthe other hand,since f <strong>is</strong>strictly increasing we have<br />
f( f(a0))< f(a0).<br />
Contradiction. ◻<br />
Lemma 1.8. Let⟨A,≤⟩ be a well-or<strong>de</strong>r and I⊆A. �e following statementsareequivalent:<br />
(1) I <strong>is</strong>aproper initialsegment ofA.<br />
(2) I=↓Aa, forsome a∈A.<br />
(3) I <strong>is</strong>an initialsegment ofAand I <strong>is</strong> non-<strong>is</strong>omorphicto A.<br />
Proof. (1)⇒(2) If I <strong>is</strong> a proper subclass of A then A∖ I <strong>is</strong> nonempty<br />
and has a least element a. Consequently, we have I=↓a.<br />
(2)⇒(3) Let I=↓a. Supposethere ex<strong>is</strong>ts an <strong>is</strong>omorph<strong>is</strong>m f∶ A→ I.<br />
By Lemma 1.7, we have f(a)≥ a. Hence, f(a)∉ I=rng f. Contradiction.<br />
(3)⇒(1) <strong>is</strong> trivial. ◻<br />
Corollary1.9. < <strong>is</strong>astrictpartial or<strong>de</strong>r onWo.<br />
Proof. We can see immediately from the <strong>de</strong>finition that
1. Well-or<strong>de</strong>rs<br />
Proof. Let f , g∶A→ B be <strong>is</strong>omorph<strong>is</strong>ms.�enso <strong>is</strong> g○ f −1 ∶ B→B. In<br />
particular, g○ f −1 <strong>is</strong>strictly increasing. By Lemma 1.7,we obtain<br />
f(a)≤(g○ f −1 )( f(a))= g(a) , for all a∈A.<br />
Similarly, we <strong>de</strong>rive g(a)≤ f(a), for all a. It followsthat f= g. ◻<br />
Westill have to provethat
a3. Ordinals<br />
�erestriction of h1 to↓a0 <strong>is</strong> an <strong>is</strong>omorph<strong>is</strong>m<br />
h1↾↓a0∶↓a0→↓h1(a0).<br />
Composing itwith h −1<br />
0 yields an <strong>is</strong>omorph<strong>is</strong>m<br />
(h1↾↓a0)○ h −1<br />
0 ∶↓b0→↓h1(a0).<br />
Butth<strong>is</strong> contradicts h1(a0)
1. Well-or<strong>de</strong>rs<br />
Proof. (⇒) By <strong>de</strong>finition, every continuous function sat<strong>is</strong>fies (2). Furthermore,<br />
a + =sup{a,a + } impliesthat f(a + )=sup{f(a), f(a + )}.<br />
(⇐) Forthe other direction,supposethat f sat<strong>is</strong>fies(1) and(2). First,<br />
we show that f <strong>is</strong> increasing. Suppose otherw<strong>is</strong>e and let a∈A be the<br />
minimal element such that f(b)> f(a), for some b
a3. Ordinals<br />
Proof. Ifa<strong>is</strong> the greatest element ofA,we can setb∶= a. Otherw<strong>is</strong>e,we<br />
have f(a + )> f(a)≥ a, by Lemma 1.7. Hence, there are elementsx∈A<br />
with f(x)>a. Let c be the least such element. We have c>� since<br />
f(c)> a≥ f(�). If c were a limitthen, by choice of c,wewould have<br />
f(c)=sup{ f(x)∣ x< c}≤ a< f(c).<br />
A contradiction. Hence, c <strong>is</strong> a successor and there ex<strong>is</strong>ts some b∈A<br />
with c=b + . By choice of c, we have f(b)≤a. Furthermore, if x>b<br />
then x≥c, which implies that f(x)≥ f(c)> a. �erefore, b <strong>is</strong> the<br />
<strong>de</strong>sired element. ◻<br />
2. Ordinals<br />
We have seen that there ex<strong>is</strong>ts a well-or<strong>de</strong>r on Wo if one does not d<strong>is</strong>tingu<strong>is</strong>h<br />
between <strong>is</strong>omorphic or<strong>de</strong>rs.Wewould liketo <strong>de</strong>fine asubclass<br />
On⊆Wo of ordinals such that, for each well-or<strong>de</strong>r A, there ex<strong>is</strong>ts a<br />
unique element B∈Onthat <strong>is</strong> <strong>is</strong>omorphicto A.<br />
Wewillpresenttwo approachesto doso.�eusual one – duetovon<br />
Neumann – hasthe d<strong>is</strong>advantage that itrequiresthe Axiom ofReplacement.<br />
Without it we cannot prove that, for every well-or<strong>de</strong>r α, there<br />
ex<strong>is</strong>ts an <strong>is</strong>omorphicvon Neumann ordinal.�erefore,wewill adopt a<br />
different approach.�erelation≅ forms a congruence(seeSection b1.4<br />
below) on the class of all well-or<strong>de</strong>rs. A first try might thus cons<strong>is</strong>t in<br />
representing awell-or<strong>de</strong>ring by its congruence class.Unfor<strong>tu</strong>nately,the<br />
class of all well-or<strong>de</strong>rs <strong>is</strong>omorphic to a given one <strong>is</strong> not a set. Hence,<br />
withth<strong>is</strong> <strong>de</strong>finition one could not formsets of ordinals. Instead of consi<strong>de</strong>ringall<br />
<strong>is</strong>omorphicwell-or<strong>de</strong>rswewilltherefore onlytakesome of<br />
them.<br />
Definition 2.1. �e or<strong>de</strong>rtype of awell-or<strong>de</strong>r A <strong>is</strong>theset<br />
ord(A)∶=[A]≅= cut{B∣B<strong>is</strong> awell-or<strong>de</strong>r <strong>is</strong>omorphicto A}.<br />
�e elements of On∶=rng(ord) are called ordinals.<br />
64
2. Ordinals<br />
Instead of asubclass On⊆Wothe above <strong>de</strong>finition results in a function<br />
ord∶Wo→On. Belowwewillseethatthere ex<strong>is</strong>ts a canonicalway<br />
to associate with every ordinal α∈ On a well-or<strong>de</strong>r f(α)∈Wo. Using<br />
th<strong>is</strong> injection f∶ On→Wowe can i<strong>de</strong>ntify the class Onwith its image<br />
f[On]⊆Wo.<br />
First, let us show that the mapping ord∶Wo→On has the <strong>de</strong>sired<br />
property of character<strong>is</strong>ing awell-or<strong>de</strong>rupto <strong>is</strong>omorph<strong>is</strong>m.<br />
Lemma 2.2. Let A and B be well-or<strong>de</strong>rs that are sets. �ere ex<strong>is</strong>ts an<br />
<strong>is</strong>omorph<strong>is</strong>m f∶ A→B if,and only if, ord(A)=ord(B).<br />
Proof. If f ∶ A → B <strong>is</strong> an <strong>is</strong>omorph<strong>is</strong>m then a well-or<strong>de</strong>r C <strong>is</strong> <strong>is</strong>omorphicto<br />
A if, and only if, it <strong>is</strong> <strong>is</strong>omorphicto B. �erefore ord(A)=<br />
ord(B). Conversely, suppose ord(A)=ord(B).Since A <strong>is</strong> awell-or<strong>de</strong>r<br />
<strong>is</strong>omorphic to A, we have ord(A)≠∅. Fix an arbitrary element C∈<br />
ord(A). By <strong>de</strong>finition, C <strong>is</strong> <strong>is</strong>omorphic to A and to B. Consequently,<br />
A and B are <strong>is</strong>omorphic. ◻<br />
Remark. Wewillprove in Lemma a4.5.3with the help of the Axiom of<br />
Replacement that any two well-or<strong>de</strong>red proper classes are <strong>is</strong>omorphic.<br />
In particular, it followsthat inthe above lemmawe can droptherequirement<br />
of A and B beingsets.<br />
Definition 2.3. Let On∶=⟨On,
a3. Ordinals<br />
�eorem 2.4. On <strong>is</strong>awell-or<strong>de</strong>r.<br />
�e notions of a successor ordinal and a limit ordinal are <strong>de</strong>fined in<br />
thesameway as for arbitrarywell-or<strong>de</strong>rs.Recallthatwe <strong>de</strong>notethesuccessor<br />
of α by α + . Furthermore,we <strong>de</strong>fine<br />
0∶= ord⟨∅,∅⟩ , 1∶=0 + , 2∶= 1 + ,...<br />
�e first limit ordinal <strong>is</strong> ω∶= ord⟨N,≤⟩.<br />
Lemma 2.5. Let α,β∈On. Ifα≤βthen S(α)⊆ S(β).<br />
Proof. If α=β, the claim <strong>is</strong> trivial. �erefore, we assume that α
Lemma 2.7. On <strong>is</strong> notaset.<br />
2. Ordinals<br />
Proof. Suppose that On <strong>is</strong> a set. Since On <strong>is</strong> well-or<strong>de</strong>red there ex<strong>is</strong>ts<br />
some ordinal α ∈ On with α = ord⟨On,≤⟩. We have just seen that<br />
ord⟨↓α,≤⟩=α. �erefore, there ex<strong>is</strong>ts an <strong>is</strong>omorph<strong>is</strong>m f ∶↓α→On.<br />
But↓α <strong>is</strong> a proper initial segment of On. �<strong>is</strong> contradicts Lemma 1.8.<br />
◻<br />
Lemma 2.8. A subclass X⊆ On <strong>is</strong> a set if, and only if, it has an upper<br />
bound.<br />
Proof. (⇐) If X⊆ On has anupper bound α then X⊆⇓α.Since⇓α <strong>is</strong><br />
asetthe claim follows.<br />
(⇒) Suppose that X <strong>is</strong> a set. Since On <strong>is</strong> a proper class there ex<strong>is</strong>ts<br />
some ordinal α∈ On∖S(X).We claim that α <strong>is</strong> anupper bound of X.<br />
Suppose there ex<strong>is</strong>ts some β∈X with β≰α. �en α
a3. Ordinals<br />
Proof. (⇒) If A≤B then, by <strong>de</strong>finition, there ex<strong>is</strong>ts an <strong>is</strong>omorph<strong>is</strong>m<br />
f ∶ A→ I between A and an initial segment I of B. In particular, f ∶<br />
A→ B <strong>is</strong> astrictly increasing function.<br />
(⇐) Suppose that f ∶ A→ B <strong>is</strong> a strictly increasing function and<br />
letC∶=rng f.SinceC⊆B <strong>is</strong>well-or<strong>de</strong>redthere ex<strong>is</strong>ts an <strong>is</strong>omorph<strong>is</strong>m<br />
g∶C→ I⊆ On betweenCand an initialsegment of On.Similarly,there<br />
<strong>is</strong>some <strong>is</strong>omorph<strong>is</strong>m h∶B→J⊆ On.We claimthat<br />
k∶= h −1 ○ g○ f∶ A→ B<br />
<strong>is</strong>the <strong>de</strong>sired <strong>is</strong>omorph<strong>is</strong>m betweenA and an initialsegment ofB.Since<br />
f , g, and h −1 are <strong>is</strong>omorph<strong>is</strong>msso <strong>is</strong> k.Whatremainsto beshown <strong>is</strong>that<br />
k <strong>is</strong> in factwell-<strong>de</strong>fined,that <strong>is</strong>, I=rng g⊆rng h= J.<br />
We claim that g(c)≤ h(c), for all c∈C. Since I and J are initial<br />
segmentsth<strong>is</strong> impliesthat I⊆J. For a contradiction, supposethatthere<br />
<strong>is</strong>some c∈Cwith g(c)> h(c) and let c bethe minimal such element.<br />
Notethat, since g and h are strictly increasing and rng g and rng h are<br />
initial segmentswe must have<br />
g(c)= min(I∖rng(g↾↓Cc))<br />
and h(c)=min(J∖rng(h↾↓Bc)).<br />
By choice ofc,we haverng(g↾↓Cc)⊆rng(h↾↓Bc). But, bythe above<br />
equations, th<strong>is</strong> impliesthat g(c)≤ h(c). A contradiction. ◻<br />
In or<strong>de</strong>r tousethetheory of ordinals forproofs about arbitrary sets<br />
oneusually needsto <strong>de</strong>fine awell-or<strong>de</strong>r on a givenset. In generalth<strong>is</strong> <strong>is</strong><br />
only possible if one assumes the Axiom of Choice. Until we introduce<br />
th<strong>is</strong> axiomthe followingtheoremwillserve as astopgap. Oncewe have<br />
<strong>de</strong>fined the cardinality of a set in Section a4.2 it will <strong>tu</strong>rn out that the<br />
ordinalthetheoremtalks about <strong>is</strong> α=∣A∣ + .<br />
�eorem 2.12(Hartogs). For every set Athere ex<strong>is</strong>ts an ordinal α such<br />
thatthereare no injective functions↓α→A.<br />
68
2. Ordinals<br />
Proof. For a contradiction, suppose that there ex<strong>is</strong>ts a set A such that,<br />
for every ordinalα,there <strong>is</strong> an injective function fα∶↓α→A. LetAα∶=<br />
rng fα⊆ A andset<br />
Rα∶={⟨a,b⟩∈ Aα×Aα∣ f −1<br />
α (a)≤ f−1 α (b)}.<br />
By construction, fα∶⟨↓α,≤⟩→⟨Aα,Rα⟩ <strong>is</strong> an <strong>is</strong>omorph<strong>is</strong>m. Hence, by<br />
the <strong>de</strong>finition of an ordinal,we have<br />
S(α)⊆ S(⟨Aα,Rα⟩).<br />
Since Rα⊆ A×A∈℘ 3 (A)⊆℘ 3 (S(A)) it followsthat<br />
⟨Aα,Rα⟩={{Aα},{Aα,Rα}}⊆℘ 4 (S(A)).<br />
We haveshownthat<br />
α⊆S(α)⊆ S(⟨Aα,Rα⟩)⊆℘ 4 (S(A)) , for all α∈ On.<br />
Consequently, On⊆℘ 5 (S(A)),which impliesthat On <strong>is</strong> aset.�<strong>is</strong> contradicts<br />
Lemma2.7. ◻<br />
VonNeumannordinals<br />
We conclu<strong>de</strong>th<strong>is</strong>sectionwith an alternative <strong>de</strong>finition of ordinals.�<strong>is</strong><br />
<strong>de</strong>finition <strong>is</strong>simpler and the resulting ordinals have many niceproperties<br />
such that α=↓α and supX=⋃X. �e only d<strong>is</strong>advantage <strong>is</strong> that<br />
one needs an additional axiom in or<strong>de</strong>r to prove that every well-or<strong>de</strong>r<br />
<strong>is</strong> <strong>is</strong>omorphictosome ordinal. In<strong>tu</strong>itively, we <strong>de</strong>fine avon Neumann ordinalto<br />
bethe set of all smaller ordinals, that <strong>is</strong>, α∶=↓α. Asusual, the<br />
ac<strong>tu</strong>al <strong>de</strong>finition <strong>is</strong> moretechnical andwe havetoverify a�erwardsthat<br />
it hasthe <strong>de</strong>sired effect.<br />
Definition 2.13. Aset α <strong>is</strong> avon Neumann ordinal if it <strong>is</strong>transitive and<br />
linearly or<strong>de</strong>red by the membership relation∈. We <strong>de</strong>note the class of<br />
all von Neumann ordinals by On0 andweset On0∶=⟨On0,∈⟩.<br />
69
a3. Ordinals<br />
Example. �e set[n]={[0],... ,[n−1]} <strong>is</strong> a von Neumann ordinal,<br />
for each n∈N.<br />
Lemma 2.14. Ifα∈ On0 and β∈αthen β∈On0.<br />
Proof. First, notethat β∈α implies β⊆α. Asα<strong>is</strong> linearly or<strong>de</strong>red by∈<br />
ittherefore followsthatso <strong>is</strong> β⊆α.<br />
It remains to prove that β <strong>is</strong> transitive. Suppose that η∈γ∈β. By<br />
transitivity of α,we have η, γ,β∈α.Since α <strong>is</strong> linearly or<strong>de</strong>red by∈we<br />
know thatthe relation∈,restricted to α, <strong>is</strong>transitive. Hence, η∈γand<br />
γ∈β impliesthat η∈β. ◻<br />
Remark. Notethat, for α∈ On0,we have<br />
↓α={β∈On0∣ β∈α}.<br />
Hence, α=↓α and our <strong>de</strong>finition of avon Neumann ordinal coinci<strong>de</strong>s<br />
withthe in<strong>tu</strong>itive one.<br />
Exerc<strong>is</strong>e 2.1. Suppose that α={β0,... ,βn−1} <strong>is</strong> a von Neumann ordinalwith<br />
n
Lemma 2.16. On0 <strong>is</strong> notaset.<br />
2. Ordinals<br />
Proof. On0 <strong>is</strong>transitive andwell-or<strong>de</strong>red by∈. If itwere aset, itwould<br />
be an element of itself. ◻<br />
On0 <strong>is</strong> linearly or<strong>de</strong>red by∈.�e followingsequence of lemmas contains<br />
several character<strong>is</strong>ations of th<strong>is</strong> or<strong>de</strong>ring. In particular, we show<br />
thatthe mapping<br />
ord∶⟨On0,∈⟩→⟨On,
a3. Ordinals<br />
Proof. (1)⇔(2)was already shown in Lemma 2.17.<br />
(1)⇒(3) a∈bimplies S(a)∈ S(b), for arbitrary sets a andb.<br />
(3)⇒(1) If α ∉ β then, by Lemma 2.15, we either have α = β or<br />
β∈α. Consequently, either S(α)= S(β) orS(β)∈ S(α). It followsthat<br />
S(α)∉ S(β).<br />
(2)⇒(4) If α⊆β, the i<strong>de</strong>ntity idα∶ α→α⊆β <strong>is</strong> an <strong>is</strong>omorph<strong>is</strong>m<br />
fromα to an initial segment of β. Hence,α
2. Ordinals<br />
�ereason whythere might be lessvon Neumann ordinals than elements<br />
of On <strong>is</strong> that each von Neumann ordinal <strong>is</strong> contained in a new<br />
stage.�at <strong>is</strong>,we have exactly onevon Neumann ordinal for everystage.<br />
Lemma 2.22. �e function f∶ On0→ H(S)<strong>de</strong>finedby f(α)∶= S(α) <strong>is</strong><br />
an <strong>is</strong>omorph<strong>is</strong>mbetween On0 andtheclass ofallstages.<br />
Proof. By Lemma2.19 it followsthat f <strong>is</strong> injective and increasing. Suppose<br />
that it <strong>is</strong> not surjective. Let S be the minimal stage such that S∉<br />
rng f , andset<br />
X∶={α∈ On0∣ S(α)∈ S}.<br />
Since X ⊆ S, X <strong>is</strong> a set and, hence, a proper initial segment of On0.<br />
�erefore,there <strong>is</strong>some α∈On0 suchthat X=↓α.Since S(β)∈ S, for<br />
all β∈α, it follows that S(α)⊆S. By choice of S, we have S(α)≠ S.<br />
Hence, S(α)∈ S,which impliesthat α∈X=↓α. Contradiction. ◻<br />
Definition 2.23. For α∈ On0,weset Sα∶= S(α).<br />
Remark. In On0 we have finally found the indices to enumerate the<br />
cumulative hierarchy<br />
S0⊂ S1⊂⋯⊂ Sα⊂ Sα+1⊂⋯<br />
�e class of allstages can bewritten inthe form<br />
H(S)={Sα∣ α∈On0} ,<br />
andwe haveS=⋃{Sα∣ α∈ On0}.<br />
Definition 2.24. �erankρ(a) of aseta<strong>is</strong>thevon Neumann ordinalα<br />
suchthat S(a)= Sα.<br />
Remark. (a) For α∈ On0,we have ρ(α)= α.<br />
(b) Notethat<br />
cutA={x∈ A∣ ρ(x)≤ ρ(y) for all y∈A}.<br />
Lemma 2.25. AclassX<strong>is</strong>aset if,and only if,{ρ(x)∣ x∈X}<strong>is</strong>boun<strong>de</strong>d.<br />
Exerc<strong>is</strong>e 2.3. Provethepreceding lemma.<br />
73
a3. Ordinals<br />
3. Inductionand fixedpoints<br />
�e importance of ordinals stems from the fact that they allow proofs<br />
and constructions by induction.�e nexttheorem follows immediately<br />
from�eorem 1.5.<br />
�eorem 3.1(Principle ofTransfinite Induction). Let I⊆ Onbe an initialsegment<br />
of On. If X⊆I<strong>is</strong>aclasssuchthat, forevery α∈I,<br />
then X=I.<br />
↓α⊆X implies α∈X<br />
Usually one applies th<strong>is</strong> theorem in the following way. If one wants<br />
to prove that all ordinals sat<strong>is</strong>fy a certain property φ, it <strong>is</strong> sufficient to<br />
provethat<br />
◆ 0sat<strong>is</strong>fies φ;<br />
◆ if α sat<strong>is</strong>fies φthenso doesα + ;<br />
◆ if δ <strong>is</strong> a limit ordinal and every α
3. Inductionand fixedpoints<br />
If β
a3. Ordinals<br />
which impliesthat<br />
F↾↓α= fβ↾↓α , for all β≥α.<br />
�erefore, it followsthat<br />
F(α)= fα +(α)= H( fα +↾↓α)=H(F↾↓α).<br />
In particular, if F <strong>is</strong> a set then F= fα, for some α. Hence, we have<br />
dom F= dom fα =↓α. Since α∉dom F it follows that fα + does not<br />
ex<strong>is</strong>ts. Hence, H( fα)=H(F) <strong>is</strong> un<strong>de</strong>fined and F∉ dom H. If F <strong>is</strong> a<br />
proper classthenwetrivially have F∉ dom H. ◻<br />
Remark. A�er we have introduced the Axiom of Replacement we can<br />
ac<strong>tu</strong>ally showthat, if H∶S
y<br />
α+0 ∶= α ,<br />
α+β + ∶=(α+β) + ,<br />
3. Inductionand fixedpoints<br />
α+δ ∶=sup{α+β∣β
a3. Ordinals<br />
Remark. (a) Notethat, if A <strong>is</strong> asetthen, bythePrinciple ofTransfinite<br />
Recursion, there ex<strong>is</strong>ts a unique function F∶ On→A sat<strong>is</strong>fying the<br />
above equations provi<strong>de</strong>d we can show that, for every limit δ, the supremum<br />
sup F[↓δ] ex<strong>is</strong>ts. If, furthermore, we can prove that F(β + )≥<br />
F(β), for all β,then it followsthat f <strong>is</strong> inductive.<br />
(b) Every fixed-point induction F <strong>is</strong> continuous, by Lemma 1.13.<br />
Example. (a) �e function f∶ On→On∶ α↦α + <strong>is</strong> inductive. Its fixedpoint<br />
induction over0<strong>is</strong>the i<strong>de</strong>ntity function F∶ On→On∶ α↦α.<br />
(b) Let f∶S→S bethe functionwith f(a)∶=℘(a).�e fixed-point<br />
induction of f over∅<strong>is</strong>the function F∶ On0→Swith<br />
F(α)∶= Sα.<br />
(c)�e graph of addition<br />
A∶={(x,y,z)∈ N 3 ∣ x+y=z}<br />
<strong>is</strong>the least fixed point ofthe function f∶℘(N 3 )→℘(N 3 )with<br />
f(R)∶={(x,0,x)∣ x∈N}<br />
∪{(x,y+ 1,z+ 1)∣(x,y,z)∈ R}.<br />
Its fixed-point induction over∅<strong>is</strong>the function<br />
⎧⎪{(x,y,z)∣<br />
x+y=z ,y
and, generally, we have<br />
F(n)= ⋃k
a3. Ordinals<br />
Proof. F(α) <strong>is</strong> a fixed point of f since f(F(α))= F(α + )= F(α). ◻<br />
�us,we canusethe fixedpoint induction F of f to compute a fixed<br />
pointprovi<strong>de</strong>d F converges.<br />
Lemma 3.8. Let F bethe fixed-point induction ofafunction f. If F(α)=<br />
F(α + )then F(α)= F(β), forall β≥α.<br />
Proof. Weprovethe claim by induction on β. If β=αthenthe claim <strong>is</strong><br />
trivial. Forthesuccessorstep,we have<br />
F(β + )= f(F(β))= f(F(α))= F(α + )= F(α).<br />
Finally, ifδ> α <strong>is</strong> a limit ordinal,then<br />
F(δ)=sup{ F(β)∣ β
3. Inductionand fixedpoints<br />
Definition 3.10. Let f ∶ A→ A be inductive and F ∶ On→A the<br />
corresponding fixed-point induction. �e minimal ordinal α suchthat<br />
F(α)=F(α + ) <strong>is</strong> called the closure ordinal of the induction and the<br />
element F(∞)∶= F(α) <strong>is</strong>the inductive fixedpoint of f over a.<br />
Remark. IfA <strong>is</strong> aset, every inductive function f∶ A→ A has an inductive<br />
fixedpoint.<br />
Example. Let⟨A,R⟩ be a graph.�ewell-foun<strong>de</strong>dpart of R <strong>is</strong>the maximalsubsetB⊆Asuchthat⟨B,R∣B⟩<br />
<strong>is</strong>well-foun<strong>de</strong>d and, for all⟨a,b⟩∈<br />
R with b∈B,we also have a∈B.We can compute B as inductive fixed<br />
point over∅ofthe function<br />
f(X)∶={x∈ A∣ R −1 (x)⊆ X∪{x}}.<br />
Ifwewantto applythe above machineryto compute fixedpoints,we<br />
need methods to show that a given function f <strong>is</strong> inductive. Basically,<br />
there are two conditions a function f has to sat<strong>is</strong>fy. �e sequence obtained<br />
by iterating f hasto be linearly or<strong>de</strong>red and itssupremum must<br />
ex<strong>is</strong>ts.<br />
Definition 3.11. Let A=⟨A,≤⟩ be apartial or<strong>de</strong>r.<br />
(a) A <strong>is</strong> inductively or<strong>de</strong>red if every chain C⊆A that <strong>is</strong> a set has a<br />
supremum.<br />
(b) A function f∶ A→ A <strong>is</strong> inflationary if f(a)≥ a, for all a∈A.<br />
Remark. (a) Every inductively or<strong>de</strong>red set has a least element�since<br />
theset∅<strong>is</strong> linearly or<strong>de</strong>red.<br />
(b) Every completepartial or<strong>de</strong>r <strong>is</strong> inductively or<strong>de</strong>red.<br />
(c)⟨On,≤⟩ <strong>is</strong> inductively or<strong>de</strong>red.<br />
(d) If⟨A,≤⟩ <strong>is</strong> a well-or<strong>de</strong>r then according to Lemma 1.7 all strictly<br />
continuous functions f∶ A→ A are inflationary.<br />
Example. (a)�epartial or<strong>de</strong>r⟨F,⊆⟩where<br />
F∶={X⊆N∣ X <strong>is</strong> finite}<br />
81
a3. Ordinals<br />
<strong>is</strong> not inductively or<strong>de</strong>redsincethe chain<br />
[0]⊂[1]⊂[2]⊂⋅⋅⋅⊂[n]⊂⋯<br />
has noupper bound.<br />
(b) LetV be avector space overthe field K andset<br />
I∶={B⊆V∣ B <strong>is</strong> linearly in<strong>de</strong>pen<strong>de</strong>nt}.<br />
We claimthat⟨I,⊆⟩ <strong>is</strong> inductively or<strong>de</strong>red.<br />
LetC⊆I be a chain.WeshowthatsupC=⋃C. By Corollary a2.3.10,<br />
it <strong>is</strong>sufficienttoprovethat⋃C∈I.<br />
Suppose otherw<strong>is</strong>e. �en⋃C <strong>is</strong> not linearly in<strong>de</strong>pen<strong>de</strong>nt and there<br />
are elements v0,... ,vn∈⋃C and λ0,... , λn∈ K suchthat λi≠ 0, for<br />
all i, and<br />
λ0v0+⋅⋅⋅+ λnvn=0.<br />
For eachvi, fixsome Bi∈ C withvi∈ Bi.SinceC<strong>is</strong> linearly or<strong>de</strong>redso<br />
<strong>is</strong>theset{B0,... ,Bn}.�<strong>is</strong>set <strong>is</strong> finite and,therefore, it has a maximal<br />
element Bk, that <strong>is</strong>, Bi ⊆ Bk, for all i. It follows that v0,... ,vn∈ Bk,<br />
which impliesthat Bk <strong>is</strong> not linearly in<strong>de</strong>pen<strong>de</strong>nt. Contradiction.<br />
Lemma 3.12. Let A=⟨A,≤⟩be inductively or<strong>de</strong>red.<br />
(a) If f∶ A→ A <strong>is</strong> inflationary, f <strong>is</strong> inductive overeveryelementa∈A.<br />
(b) If f∶ A→ A <strong>is</strong> increasing, f <strong>is</strong> inductive overeveryelement awith<br />
f(a)≥ a.<br />
(c) If f∶ A→ A <strong>is</strong>continuous, f <strong>is</strong> inductive overeveryelementawith<br />
f(a)≥ a. Furthermore, ifthe inductive fixed point of f over a ex<strong>is</strong>ts, its<br />
closure ordinal <strong>is</strong>at most ω.<br />
Proof. (a) Bytransfiniterecursion,we construct an increasing function<br />
F ∶ I → A sat<strong>is</strong>fying the equations in Definition 3.6. Let F(0)∶= a.<br />
For the induction step, suppose that F(α) <strong>is</strong> already <strong>de</strong>fined. We set<br />
F(α + ) ∶= f(F(α)). Since f <strong>is</strong> inflationary, it follows that F(α + ) =<br />
f(F(α))≥F(α). Finally, suppose that δ <strong>is</strong> a limit ordinal. If F↾↓δ<br />
82
3. Inductionand fixedpoints<br />
<strong>is</strong> aproper class, we are done. Otherw<strong>is</strong>e, F[↓δ] <strong>is</strong> asetwhich, furthermore,<br />
<strong>is</strong> linearly or<strong>de</strong>red because F↾↓δ <strong>is</strong> increasing. As⟨A,≤⟩ <strong>is</strong> inductively<br />
or<strong>de</strong>red it followsthat F[↓δ] has asupremum andwe canset<br />
F(δ)∶=sup F[↓δ].<br />
(b) Again we <strong>de</strong>fine an increasing function F∶ I→A by transfinite<br />
recursion. Let F(0)∶= a. For the induction step, suppose that F(α) <strong>is</strong><br />
already <strong>de</strong>fined.Weset F(α + )∶= f(F(α)).Toprovethat F(α + )≥ F(α)<br />
we consi<strong>de</strong>r three cases. For α= 0we have F(1)= f(a)≥ a=F(0). If<br />
α=β + <strong>is</strong> a successor, we know by induction hypothes<strong>is</strong> that F(β + )≥<br />
F(β).Since f <strong>is</strong> increasing it followsthat<br />
F(α + )= f(F(β + ))≥ f(F(β))= F(β + )= F(α).<br />
If α <strong>is</strong> a limitthen F(α)=sup F[↓α] and<br />
F(α + )= f(sup F[↓α])≥ f(F(β))= F(β + ) , for all β
a3. Ordinals<br />
(a) f <strong>is</strong> inductive overα.<br />
(b) If F <strong>is</strong> the fixed-point induction of f over α then F(∞) ex<strong>is</strong>ts if,<br />
and only if, the set{f n (α)∣ n
4. Ordinalarithmetic<br />
�esecondtheorem <strong>is</strong> aversion ofthe�eorem of Knaster andTarski<br />
whichshowsthat we can computethe least fixed point of a function f<br />
by a fixed-point induction.<br />
�eorem 3.15. Let⟨A,≤⟩bean inductively or<strong>de</strong>red graphwhereA <strong>is</strong>aset<br />
and let f∶ A→ Abe an increasing function. If the least fixed point of f<br />
ex<strong>is</strong>ts then itcoinci<strong>de</strong>swith its inductive fixedpoint over�.<br />
Proof. Let F∶ On→ A bethe fixed-point induction of f over�.Suppose<br />
thata∶= lfp f ex<strong>is</strong>ts.Weprove by induction onαthat F(α)≤ a.�en it<br />
followsthat F(∞)≤ a andthe minimality of a impliesthat F(∞)= a.<br />
Clearly, F(0)=�≤ a. Forthe inductive step,supposethat F(α)≤ a.<br />
Since f <strong>is</strong> increasing it followsthat<br />
F(α + )= f(F(α))≤ f(a)= a.<br />
Finally, if δ <strong>is</strong> a limit ordinal,the induction hypothes<strong>is</strong> impliesthat<br />
F(δ)=sup{ F(α)∣ α
a3. Ordinals<br />
+ = ⋅ =<br />
andthe or<strong>de</strong>r <strong>is</strong> <strong>de</strong>fined by<br />
Figure 1..Sum andproduct of linear or<strong>de</strong>rs<br />
⟨i,a⟩≤C⟨k,b⟩ : iff i=k=0and a≤A b<br />
or i=k= 1 and a≤B b<br />
or i=0and k= 1.<br />
(b)�eproduct A⋅B<strong>is</strong>the graph⟨C,≤C⟩whereC∶= A×B andthe<br />
or<strong>de</strong>r <strong>is</strong> <strong>de</strong>fined by<br />
⟨a,b⟩≤C⟨a ′ ,b ′ ⟩ : iff b
(c) K (M) ≅⟨[k m ],≤⟩.<br />
4. Ordinalarithmetic<br />
Addition of linear or<strong>de</strong>rs <strong>is</strong> associative andthe empty or<strong>de</strong>r <strong>is</strong> a neutral<br />
element. Belowwe will give an example showing that, in general, it<br />
<strong>is</strong> not commutative.<br />
Lemma4.2. If A, B,and Care linear or<strong>de</strong>rsthen<br />
(A+B)+C≅A+(B+C).<br />
Proof. Let A=⟨A,≤A⟩, B=⟨B,≤B⟩, and C=⟨C,≤C⟩. We can <strong>de</strong>fine a<br />
bijection f∶(A⊍B)⊍C→A⊍(B⊍C) by<br />
f⟨0,⟨0,a⟩⟩∶=⟨0,a⟩ for a∈A ,<br />
f⟨0,⟨1,b⟩⟩∶=⟨1,⟨0,b⟩⟩ forb∈B ,<br />
f⟨1,c⟩∶=⟨1,⟨1,c⟩⟩ for c∈C.<br />
Sinceth<strong>is</strong> bijectionpreservesthe or<strong>de</strong>ring it <strong>is</strong>the <strong>de</strong>sired <strong>is</strong>omorph<strong>is</strong>m.<br />
◻<br />
As we want to <strong>de</strong>fine arithmetic operations on ordinals we have to<br />
showthat, ifwe applythe above operationstowell-or<strong>de</strong>rs,we again obtain<br />
awell-or<strong>de</strong>r.<br />
Lemma 4.3. If A and B are well-or<strong>de</strong>rs then so are A+B, A⋅B, and<br />
A (B) .<br />
Proof. Suppose that A=⟨A,≤A⟩ and B=⟨B,≤B⟩. We will prove the<br />
claim only for C∶= A (B) .�e other operations are le� as an exerc<strong>is</strong>eto<br />
therea<strong>de</strong>r.<br />
Let C=⟨C,≤C⟩. �e relation
a3. Ordinals<br />
that, respectively, f(b0)≠ g(b0) and g(b1)≠ h(b1). By <strong>de</strong>finition, we<br />
have f(b0)
<strong>is</strong> a bijection sincewe have<br />
g(c)= g ′ (c) for all g, g ′ ∈ Z and every c≥b.<br />
4. Ordinalarithmetic<br />
Furthermore, ρ preserves the or<strong>de</strong>ring, that <strong>is</strong>, it <strong>is</strong> an <strong>is</strong>omorph<strong>is</strong>m. It<br />
followsthat ρ −1 (h) <strong>is</strong>the minimal element of Z and of X. ◻<br />
Exerc<strong>is</strong>e 4.2. Show that, if A and B are well-or<strong>de</strong>rs then so are A+B<br />
and A⋅B.<br />
It <strong>is</strong> easy to see that A ≅ A ′ and B ≅ B ′ implies that the sums,<br />
products, andpowers are also <strong>is</strong>omorphic.�erefore,we can <strong>de</strong>finethe<br />
corresponding operations on ordinals bytaking representatives.<br />
Definition 4.4. For α= ord(A) and β=ord(B)we <strong>de</strong>fine<br />
α+β∶= ord(A+B) ,<br />
α⋅β∶= ord(A⋅B) ,<br />
α (β) ∶= ord(A (B) ).<br />
Example. �e following equations can beproved easily bythe lemmas<br />
below.We encouragetherea<strong>de</strong>rto <strong>de</strong>rivethem directly fromthe <strong>de</strong>finitions.<br />
1+1=2 (3+6)ω= 9ω=ω
a3. Ordinals<br />
Ordinaladdition<br />
�e properties of ordinal addition, multiplication, and exponentiation<br />
aresimilarto, but notquitethesame asthose for integers.�e following<br />
sequence of lemmassummar<strong>is</strong>esthem.Westartwith addition.<br />
Lemma 4.5. Let α,β, γ∈On. If β
4. Ordinalarithmetic<br />
For a contradiction suppose that γ
a3. Ordinals<br />
�e next lemmasummar<strong>is</strong>esthe laws of ordinal addition.<br />
Lemma 4.9. Let α,β, γ∈On.<br />
(a) α+(β+γ)=(α+β)+γ.<br />
(b) α+β=α+ γ implies β=γ.<br />
(c) α≤β implies α+ γ≤β+γ.<br />
(d) If X⊆ On <strong>is</strong> nonemptyandboun<strong>de</strong>dthen<br />
α+supX=sup{α+β∣β∈X}.<br />
(e) β≤αif,and only if, α=β+ γ, forsome γ∈ On.<br />
(f) β
4. Ordinalarithmetic<br />
(e) If β
a3. Ordinals<br />
Proof. Fix representatives α= ord(A) and β=ord(B).<br />
(a) follows immediately from the fact that A⋅⟨∅,∅⟩=⟨∅,∅⟩.<br />
(b) �e canonical bijection<br />
given by<br />
A×(B⊍[1])→(A×B)⊍A<br />
⟨a,⟨0,b⟩⟩↦⟨0,⟨a,b⟩⟩ ,<br />
⟨a,⟨1,0⟩⟩↦⟨1,a⟩ ,<br />
induces an <strong>is</strong>omorph<strong>is</strong>m<br />
A⋅(B+⟨[1],≤⟩)→A⋅B+A.<br />
(c) Let X∶={αβ∣β
4. Ordinalarithmetic<br />
We can also show that ordinals allow a limited form of div<strong>is</strong>ion.<br />
Lemma 4.13. For all ordinals α,β∈On with β≠0, there ex<strong>is</strong>t unique<br />
ordinals γand ρα=βγ+ρ,<br />
which impliesthat ρβδ+σ= α.<br />
A contradiction. It followsthat γ <strong>is</strong>unique. Hence,theuniqueness of ρ<br />
follows from Lemma 4.8. ◻<br />
Lemma4.14. α <strong>is</strong>alimit ordinal if,and only if, α=ωβ, forsome β>0.<br />
Proof. (⇒) By Lemma 4.13,we have α=ωβ+n for some β∈On and<br />
n
a3. Ordinals<br />
Lemma4.15. Let α,β, γ∈On.<br />
(a) α(βγ)=(αβ)γ.<br />
(b) α(β+ γ)= αβ+αγ.<br />
(c) If α≠0andαβ=αγ then β= γ.<br />
(d) α≤β implies αγ≤βγ.<br />
(e) If X⊆ On <strong>is</strong> nonemptyandboun<strong>de</strong>dthen<br />
α⋅supX=sup{αβ∣β∈ X}.<br />
Proof. (b)Weprovethe claim by induction on γ. For γ=0,we have<br />
α(β+0)= αβ=αβ+0=αβ+α0.<br />
Forthesuccessorstep,we have<br />
α(β+ γ + )= α(β+γ) +<br />
= α(β+γ)+α<br />
= αβ+αγ+α<br />
= αβ+αγ + .<br />
Finally, if γ <strong>is</strong> a limit ordinalthen<br />
α(β+ γ)= α⋅sup{β+ρ∣ρ
Ordinalexponentiation<br />
4. Ordinalarithmetic<br />
Finally, we consi<strong>de</strong>r ordinal exponentiation. Again, the basic steps are<br />
thesame as for addition and multiplication.<br />
Lemma4.16. Let α,β, γ∈On. Ifα> 1and β
a3. Ordinals<br />
For a contradiction suppose that γ
4. Ordinalarithmetic<br />
Proof. By Corollary 4.18 and Lemma 1.14,there ex<strong>is</strong>ts a greatest ordinal η<br />
suchthat β (η) ≤ α, and, by Lemma 4.13, there ex<strong>is</strong>t ordinals γ and ρ<<br />
β (η) suchthat β (η) γ+ρ=α. If γ= 0,wewould have ρ=α≥β (η) > ρ.<br />
A contradiction. And, if γ≥β,wewould have<br />
α 1thenwe have β
a3. Ordinals<br />
Cantor normal form<br />
We can applythe logarithm to <strong>de</strong>compose every ordinal in a canonical<br />
way.<br />
�eorem 4.21. For all ordinals α,β∈On with β>1, there are unique<br />
finitesequences(γi)iηn−1 , and 0ρ1>... of ordinals which <strong>is</strong> impossible. Consequently,<br />
there <strong>is</strong>some number n suchthat ρn=0 andwe have<br />
α=β (η0) γ0+⋯+β (ηn−1) γn−1. ◻<br />
Definition 4.22. Letαbe an ordinal.�eunique <strong>de</strong>composition<br />
α=ω (η0) γ0+⋯+ω (ηn) γn ,<br />
with η0>⋯>ηn and 0
Proof. Supposethat β=α+ γ, for γ>0.We have<br />
ω (α) + ω (β) = ω (α) + ω (α+γ)<br />
= ω (α) + ω (α) ω (γ)<br />
= ω (α) (1+ ω (γ) )<br />
= ω (α) ω (γ)<br />
4. Ordinalarithmetic<br />
= ω (α+γ) = ω (β) . ◻<br />
Corollary 4.24. Let α,β∈Onbe ordinalswithCantor normal form<br />
α=ω (η0) k0+⋯+ ω (ηm−1) km−1 ,<br />
β=ω (γ0) l0+⋯+ω (γn−1) ln−1.<br />
If i <strong>is</strong>the maximal in<strong>de</strong>xsuchthat ηi≥ γ0 thenwe have<br />
α+β=ω (η0) k0+⋯+ ω (ηi) ki+ ω (γ0) l0+⋯+ω (γn−1) ln−1.<br />
Lemma 4.25. An ordinal α> 0 <strong>is</strong> of the form α=ω (η) , for some η, if,<br />
and only if, β+ γ
a3. Ordinals<br />
If k> 1,weset β∶= ω (η) (k− 1)+ρ0. Inth<strong>is</strong> casewe cansetβ∶= ω (η) andwe<br />
have<br />
β+β=ω (η) + ω (η) > ω (η) +ρ=α.<br />
Again a contradiction. ◻<br />
�e nexttwo lemmasprovi<strong>de</strong>the laws of multiplication and exponentiation<br />
of ordinals in Cantor normal form.<br />
Lemma 4.26. If γ>0,0≤ ρ
Example. Bythe above lemmaswe have<br />
(ω (ω(5) +ω4+2) + ω (5) ) (ω (2) 2+ω+1)<br />
=(ω (ω(5) +ω4+2) + ω (5) ) (ω (2) 2) ⋅(ω (ω (5) +ω4+2) + ω (5) ) (ω) ⋅<br />
⋅(ω (ω(5) +ω4+2) + ω (5) )<br />
4. Ordinalarithmetic<br />
=(ω ((ω(5) +ω4+2)ω (2) ) ) (2) ⋅ ω ((ω (5) +ω4+2)ω) ⋅(ω (ω (5) +ω4+2) + ω (5) )<br />
=(ω (ω(7) ) ) (2) ⋅ ω (ω (6) ) ⋅(ω (ω (5) +ω4+2) + ω (5) )<br />
= ω (ω(7) 2) ⋅ ω (ω (6) ) ⋅(ω (ω (5) +ω4+2) + ω (5) )<br />
= ω (ω(7) 2+ω (6) ) ⋅(ω (ω (5) +ω4+2) + ω (5) )<br />
= ω (ω(7) 2+ω (6) ) ⋅ ω (ω (5) +ω4+2) + ω (ω (7) 2+ω (6) ) ⋅ ω (5)<br />
= ω (ω(7) 2+ω (6) +ω (5) +ω4+2) + ω (ω (7) 2+ω (6) +5) .<br />
Exerc<strong>is</strong>e 4.5. Computethe cantor normal form of<br />
(ω (ω(2) 7+ω3+4) 3+ ω (ω6+3) 4+ω (4) 3+1) (ω (2) 5+ω7+2)<br />
Remark. We will prove in Lemma a4.5.6 that we can find, for every β,<br />
arbitrarily large ordinals α0,α1,α2 suchthat<br />
α0=β+α0 , α1=βα1 , and α2=β (α2) .<br />
Inparticular,there are ordinals εsuchthat ε=ω (ε) . By εα we <strong>de</strong>notethe<br />
α-th ordinal such that β (εα) = εα, for all β
a3. Ordinals<br />
ordinals are<br />
0, 1, 2, 3, . . .<br />
. . . , ω, ω+1, ω+2, . . .<br />
. . . , ω2, ω2+1, ω2+ 2, . . .<br />
. . . , ω3, . . . , ω4, . . . , ω (2) , . . . , ω (3) , . . .<br />
. . . , ω (ω) , . . . , ω (ω(ω) ) , . . .<br />
. . . , ε0, . . . , ε (ε0)<br />
0 , . . . , ε1, . . . , ε2, . . . , εω, . . .<br />
. . . , ω1, . . . , ω2, . . . , ωω, . . .<br />
�e ordinals ωα will be <strong>de</strong>fined in Section a4.2.<br />
104
a4. Zermelo-Fraenkelsettheory<br />
1. �eAxiom ofChoice<br />
We have seen that induction <strong>is</strong> apowerfultechniquetoprovestatements<br />
and to construct objects. But in or<strong>de</strong>r touseth<strong>is</strong> toolwe have torelate<br />
the sets we are interested in to ordinals. In basic set theory th<strong>is</strong> <strong>is</strong> not<br />
alwayspossible.�erefore,wewill introduce a new axiomwhichstates<br />
that, for everysetA,there <strong>is</strong> awell-or<strong>de</strong>r overA. Before doingso, le<strong>tu</strong>s<br />
present several statements that are equivalent to th<strong>is</strong> axiom. We need<br />
two new notions.<br />
Definition1.1. Aset F⊆℘(A) has finite character if, for all sets x⊆ A,<br />
we have<br />
x∈F iff x0∈ F , for every finitesetx0⊆ x.<br />
Lemma1.2. Suppose that F⊆℘(A) has finitecharacter.<br />
(a) F <strong>is</strong>an initialsegment of℘(A).<br />
(b) If X⊆ F <strong>is</strong> nonemptythen⋂X∈F.<br />
(c) IfC⊆ F <strong>is</strong>achainand⋃C <strong>is</strong>asetthen⋃C∈ F.<br />
Proof. (a) follows immediately from the <strong>de</strong>finition and (b) <strong>is</strong> a consequence<br />
of (a). For (c), let C⊆Fbe a chain such that X∶=⋃C <strong>is</strong> a set.<br />
If X0⊆X <strong>is</strong> finite, there ex<strong>is</strong>ts some element Z∈ C with X0⊆ Z∈F.<br />
Hence,X0∈ F, for all finitesubsetsX0⊆ X.�<strong>is</strong> impliesthatX∈ F. ◻<br />
Lemma1.3. If F has finitecharacterthen⟨F,⊆⟩ <strong>is</strong> inductively or<strong>de</strong>red.<br />
logic, algebra & geometry2012-08-08 — ©achim blumensath 105
a4. Zermelo-Fraenkelsettheory<br />
Proof. LetC⊆ F be a linearly or<strong>de</strong>redsubset of F. By Corollary a2.3.10<br />
and Lemma 1.2 (c), it follows thatsupC=⋃C∈ F. ◻<br />
Example. LetV be avector space overthe field K.�eset<br />
F∶={B⊆V∣ B <strong>is</strong> linearly in<strong>de</strong>pen<strong>de</strong>nt}<br />
has finite character.<br />
�esecond notionwe need <strong>is</strong>that of a choice function. In<strong>tu</strong>itively, a<br />
choice function <strong>is</strong> a functionthat, given someset A,selects an element<br />
of A.<br />
Definition 1.4. A function f <strong>is</strong> a choice function if f(a)∈ a, for all<br />
a∈ dom f.<br />
Exerc<strong>is</strong>e1.1. LetI betheset of all open intervals(a,b) ofreal numbers<br />
a,b∈Rwith a
1. �eAxiom ofChoice<br />
Proof. Let f∶℘(A)∖{∅}→A be a choice function. We <strong>de</strong>fine a function<br />
g∶℘(A)→℘(A) by<br />
⎧⎪ A if X=A ,<br />
g(X)∶= ⎨<br />
⎪⎩<br />
X∪{ f(A∖X)} if X≠A.<br />
Since g(X)⊇ X th<strong>is</strong> function <strong>is</strong> inflationary. Furthermore, the partial<br />
or<strong>de</strong>r⟨℘(A),⊆⟩ <strong>is</strong> complete. By �eorem a3.3.14, g has an inductive<br />
fixed point. Since g(X)≠ X, for X≠ A, it follows that th<strong>is</strong> fixed point<br />
<strong>is</strong> A. Let G ∶ On→℘(A) be the fixed-point induction of g over∅<br />
and let α bethe closure ordinal. For every β
a4. Zermelo-Fraenkelsettheory<br />
Proof. (2)⇒(3) If∏i∈I Ai <strong>is</strong> a proper class, it <strong>is</strong> nonempty and we are<br />
done. Hence,we may assumethat it <strong>is</strong> aset.�enA∶=⋃{Ai∣ i∈I}<strong>is</strong><br />
also aset. By(2)there ex<strong>is</strong>ts a choice function f∶℘(A)∖{∅}→ A. Let<br />
g∶I→A bethe function <strong>de</strong>fined by g(i)∶= f(Ai). Since g(i)∈ Ai it<br />
followsthat g∈∏i∈I Ai≠∅.<br />
(3)⇒(4) <strong>is</strong> trivial.<br />
(4)⇒(2) Let I∶=℘(A)∖{∅} and setAX∶= X×{X}, forX∈ I. Since<br />
∏X∈I AX≠∅ there ex<strong>is</strong>ts some element f ∈∏X∈I AX. We can <strong>de</strong>fine<br />
the <strong>de</strong>sired choice function g∶℘(A)∖{∅}→ A by<br />
g(X)= a : iff f(X)=⟨a,X⟩.<br />
(2)⇒(1)wasproved in Lemma 1.6.<br />
(1)⇒(5) Suppose that⟨A,≤⟩ <strong>is</strong> inductively or<strong>de</strong>red, but A has no<br />
maximal element. For every a∈A,we can find someb∈Awith b>a.<br />
By assumption,there ex<strong>is</strong>ts awell-or<strong>de</strong>rRover A. Let f∶ A→ A bethe<br />
functionsuchthat f(a) <strong>is</strong>the R-minimal elementb∈Awith b>a. By<br />
<strong>de</strong>finition, we have f(a)> a, for all a∈A. Hence, f <strong>is</strong> inflationary and,<br />
by�eorem a3.3.14, f has a fixed point a. But f(a)= a contradicts the<br />
<strong>de</strong>finition of f.<br />
(5)⇒(6) Let F be aset of finite character andA∈ F. It <strong>is</strong>sufficientto<br />
provethatthesubset F0∶={X∈ F∣ A⊆ X}<strong>is</strong> inductively or<strong>de</strong>red by⊆.<br />
By Lemma 1.3, we know that⟨F,⊆⟩ <strong>is</strong> inductively or<strong>de</strong>red. Let C be a<br />
chain in F0.�enC⊆F0⊆ F and C <strong>is</strong> also a chain in F. Consequently,<br />
it has a least upper bound B∈F.Since A⊆ X, for all X∈C, it follows<br />
thatA⊆ B,that <strong>is</strong>,B∈ F0 andB <strong>is</strong> alsothe leas<strong>tu</strong>pper bound ofC in F0.<br />
(6)⇒(2) Let A be a set. By Lemma 1.5(a), the set C of choice functions<br />
f with dom f ⊆℘(A)∖{∅} has finite character and, therefore,<br />
there <strong>is</strong> a maximal element f∈ C. By Lemma 1.5(b), it followsthat f <strong>is</strong><br />
the <strong>de</strong>sired choice function.<br />
(1)⇒(7) Fixwell-or<strong>de</strong>rsRandSon,respectively, A andB. By Corollary<br />
a3.1.12, exactly one of the following conditions <strong>is</strong> sat<strong>is</strong>fied:<br />
108<br />
⟨A,R⟩⟨B,S⟩.
1. �eAxiom ofChoice<br />
In the first two cases there ex<strong>is</strong>ts an injection A→ B and in the second<br />
and third case there ex<strong>is</strong>ts an injection B→A in the other direction.<br />
(7)⇒(1) LetA be a set. By �eorem a3.2.12, there ex<strong>is</strong>ts an ordinalα<br />
such that there <strong>is</strong> no injective function↓α→A. Consequently, there<br />
ex<strong>is</strong>ts an injective function f∶ A→↓α.We <strong>de</strong>fine arelation R onA by<br />
R∶={⟨a,b⟩∣ f(a)< f(b)}.<br />
Since f <strong>is</strong> injective andrng f⊆↓α <strong>is</strong>well-or<strong>de</strong>red it followsthatR<strong>is</strong>the<br />
<strong>de</strong>siredwell-or<strong>de</strong>r onA.<br />
(2)⇒(8) Let h∶℘(A)∖{∅}→ A be a choice function.We can <strong>de</strong>fine<br />
g∶B→A by<br />
g(b)∶= h( f −1 (b)).<br />
(8)⇒(4) Let(Ai)i∈I be a family of d<strong>is</strong>joint nonemptysets.We <strong>de</strong>fine<br />
a function f∶⋃{Ai∣ i∈I}→ I by<br />
f(a)= i : iff a∈Ai .<br />
SincetheAi are d<strong>is</strong>joint and nonempty it followsthat f <strong>is</strong>well-<strong>de</strong>fined<br />
and surjective. Hence, there ex<strong>is</strong>ts a function g∶I→⋃{Ai∣ i∈I}<br />
such that f(g(i))= i, for all i∈I. By <strong>de</strong>finition of f , th<strong>is</strong> implies that<br />
g(i)∈ Ai. Hence, g∈∏i∈I Ai≠∅. ◻<br />
Axiom of Choice. Foreveryset Athereex<strong>is</strong>tsawell-or<strong>de</strong>rRoverA.<br />
Lemma 1.8. A le�-narrow partial or<strong>de</strong>r(A,≤) <strong>is</strong> well-foun<strong>de</strong>d if, and<br />
only if,there ex<strong>is</strong>ts no infinitestrictly<strong>de</strong>creasingsequence a0>a1>....<br />
Proof. One directionwas alreadyproved in Lemma a3.1.3. For the other<br />
one, fix a choice function f∶℘(A)∖∅→A. Supposethatthere ex<strong>is</strong>ts<br />
a nonempty set A0⊆ Awithout minimal element. We can <strong>de</strong>fine a <strong>de</strong>scending<br />
chain a0>a1>... by induction. Let a0∶= f(A0) and, for<br />
k>0,set<br />
ak∶= f({b∈A0∣ b
a4. Zermelo-Fraenkelsettheory<br />
Notethat ak <strong>is</strong>well-<strong>de</strong>fined since ak−1 <strong>is</strong> not a minimal element of A0.<br />
◻<br />
Exerc<strong>is</strong>e1.2. We call asetacountable ifthere ex<strong>is</strong>ts a bijection↓ω→a.<br />
Provethat a le�-narrowpartial or<strong>de</strong>r⟨A,≤⟩ <strong>is</strong>well-foun<strong>de</strong>d if, and only<br />
if, every countable nonemptysubset X⊆A has a minimal element.<br />
Exerc<strong>is</strong>e 1.3. Let⟨A,R⟩ be a well-foun<strong>de</strong>d partial or<strong>de</strong>r that <strong>is</strong> a set.<br />
Provethatthere ex<strong>is</strong>ts awell-or<strong>de</strong>r≤onAwith R⊆≤.<br />
�e following variant of the Axiom of Choice (statement (5) in the<br />
abovetheorem) <strong>is</strong> known as‘Zorn’s Lemma’.<br />
Lemma1.9(Kuratowski,Zorn). Every inductively or<strong>de</strong>redpartial or<strong>de</strong>r<br />
hasamaximalelement.<br />
Example. We haveseenthatthesystem of all linearly in<strong>de</strong>pen<strong>de</strong>ntsubsets<br />
of a vector space V <strong>is</strong> inductively or<strong>de</strong>red. It follows that every<br />
vector space contains a maximal linearly in<strong>de</strong>pen<strong>de</strong>nt subset, that <strong>is</strong>, a<br />
bas<strong>is</strong>.<br />
�<strong>is</strong> example can be general<strong>is</strong>ed to a certain kind of closure operators.<br />
Definition1.10. Letcbe a closure operator on A.<br />
(a) c hastheexchangeproperty if<br />
b∈c(X∪{a})∖c(X) implies a∈c(X∪{b}).<br />
(b) Aset I⊆A <strong>is</strong> c-in<strong>de</strong>pen<strong>de</strong>nt if<br />
a∉c(I∖{a}) , for all a∈I.<br />
We call D⊆Ac-<strong>de</strong>pen<strong>de</strong>nt if it <strong>is</strong> not c-in<strong>de</strong>pen<strong>de</strong>nt.<br />
(c) Let X⊆A. Aset I⊆X <strong>is</strong> a c-bas<strong>is</strong> of X if I <strong>is</strong> c-in<strong>de</strong>pen<strong>de</strong>nt and<br />
c(I)= c(X).<br />
110
1. �eAxiom ofChoice<br />
Lemma 1.11. Let c be a closure operator on A and let F⊆℘(A) be the<br />
class of all c-in<strong>de</strong>pen<strong>de</strong>nt sets. If c has finite character then F has finite<br />
character.<br />
Proof. Let I∈Fand I0⊆ I. For every a∈I0,we have<br />
a∉c(I∖{a})⊇ c(I0∖{a}).<br />
Hence, I0 <strong>is</strong>c-in<strong>de</strong>pen<strong>de</strong>nt. Conversely, supposethat I∉F.�enthere<br />
<strong>is</strong>some a∈Iwith<br />
a∈c(I∖{a}).<br />
Since c has finite characterwe can find a finitesubset I0⊆ I∖{a}with<br />
a∈c(I0).�us, I0∪{a} <strong>is</strong> a finitesubset of I that <strong>is</strong> notc-in<strong>de</strong>pen<strong>de</strong>nt.<br />
◻<br />
Before proving the converse let usshow with the help of the Axiom<br />
of Choicethatthere <strong>is</strong> always a c-bas<strong>is</strong>.Westartwith an alternative <strong>de</strong>scription<br />
ofthe exchangeproperty.<br />
Lemma1.12. Letcbeaclosure operator onAwiththeexchangeproperty.<br />
If D⊆A <strong>is</strong>aminimalc-<strong>de</strong>pen<strong>de</strong>ntsetthen<br />
a∈c(D∖{a}) , forall a∈D.<br />
Proof. Let a∈D.Since D <strong>is</strong>c-<strong>de</strong>pen<strong>de</strong>ntthere ex<strong>is</strong>tssome elementb∈<br />
D with b∈c(D∖{b}). Ifb=a thenwe are done. Hence,supposethat<br />
b≠aand let D0∶= D∖{a,b}. By minimality of D we haveb∉c(D0).<br />
Hence,b∈c(D0∪{a})∖c(D0) andthe exchangeproperty impliesthat<br />
a∈c(D0∪{b}). ◻<br />
Proposition1.13. Letcbeaclosure operator onAthat has finitecharacter<br />
andtheexchangeproperty.Everyset X⊆A hasa c-bas<strong>is</strong>.<br />
111
a4. Zermelo-Fraenkelsettheory<br />
Proof. �e family F of all c-in<strong>de</strong>pen<strong>de</strong>ntsubsets of X has finite character.<br />
By the Axiom of Choice, there ex<strong>is</strong>ts a maximal c-in<strong>de</strong>pen<strong>de</strong>ntset<br />
I⊆X.We claimthat c(I)= c(X),that <strong>is</strong>, I <strong>is</strong> a c-bas<strong>is</strong> of X.<br />
Clearly, c(I)⊆ c(X). If X⊆c(I), it followsthat<br />
c(X)⊆ c(c(I))= c(I)<br />
and we are done. Hence, it remains to consi<strong>de</strong>r the case that there <strong>is</strong><br />
some elementa∈X∖c(I).We <strong>de</strong>rive a contradictiontothe maximality<br />
of I byshowingthat I∪{a} <strong>is</strong>c-in<strong>de</strong>pen<strong>de</strong>nt.<br />
Suppose that I∪{a} <strong>is</strong> not c-in<strong>de</strong>pen<strong>de</strong>nt. Since F has finite character<br />
there ex<strong>is</strong>ts a finite c-<strong>de</strong>pen<strong>de</strong>nt subset D⊆I∪{a} with a∈D.<br />
Suppose that D <strong>is</strong> chosen minimal. By Lemma 1.12, it follows that a∈<br />
c(D∖{a})⊆ c(I). A contradiction. ◻<br />
Proposition 1.14. Let c be a closure operator on A with the exchange<br />
property and let F⊆℘(A) be the class of all c-in<strong>de</strong>pen<strong>de</strong>nt sets. �en<br />
c has finitecharacter if,and only if, F has finitecharacter.<br />
Proof. (⇒) has already beenproved in Lemma 1.11.<br />
(⇐) For a contradiction, supposethatthere <strong>is</strong> aset X⊆Asuchthat<br />
Z∶=⋃{c(X0)∣ X0⊆X<strong>is</strong> finite}<br />
<strong>is</strong> apropersubset of c(X). Fixsome element a∈c(X)∖Z. ByProposition<br />
1.13 there ex<strong>is</strong>ts a c-bas<strong>is</strong> I for X. It follows that a∈c(X)= c(I).<br />
Since F has finite character we can find a finite subset I0⊆Isuchthat<br />
I0∪{a} <strong>is</strong> c-<strong>de</strong>pen<strong>de</strong>nt. By Lemma 1.12, it follows that a∈c(I0)⊆Z.<br />
A contradiction. ◻<br />
A more extensive treatment of closure operators with the exchange<br />
propertywill be given inSection f1.1.<br />
112
2. Cardinals<br />
2. Cardinals<br />
�e notion ofthe cardinality of aset <strong>is</strong> avery na<strong>tu</strong>ral one. It <strong>is</strong> based on<br />
thesame i<strong>de</strong>awhich ledtothe <strong>de</strong>finition ofthe or<strong>de</strong>rtype of awell-or<strong>de</strong>r.<br />
But instead ofwell-or<strong>de</strong>rswe consi<strong>de</strong>r justsetswithout anyrelation. Although<br />
concep<strong>tu</strong>allysimplerthan ordinalswe introduce cardinalsquite<br />
late inthe <strong>de</strong>velopment of ourtheorysince most oftheirproperties cannot<br />
beprovedwithoutresortingto ordinals andthe Axiom of Choice.<br />
In<strong>tu</strong>itively,the cardinality of asetA measures itssize,that <strong>is</strong>,the number<br />
of its elements. So, how dowe count the elements of aset?We can<br />
saythat‘A hasα elements’ ifthere ex<strong>is</strong>ts an enumeration ofA of lengthα,<br />
that <strong>is</strong>, a bijection↓α→A. For infinitesets,such an enumeration <strong>is</strong> not<br />
unique.We can findseveralsequences↓α→Awith differentvalues ofα.<br />
To get awell-<strong>de</strong>fined numberwethereforepickthe least one.<br />
Definition 2.1. �ecardinality∣A∣ of a classA <strong>is</strong>the least ordinalαsuch<br />
thatthere ex<strong>is</strong>ts a bijection↓α→A. Ifthere ex<strong>is</strong>ts nosuch ordinalthen<br />
wewrite∣A∣∶=∞. Let Cn∶= rng∣⋅∣⊆On betherange ofth<strong>is</strong> mapping.<br />
(We do not consi<strong>de</strong>r∞to be an element of the range.) We set Cn∶=<br />
⟨Cn,≤⟩.�e elements of Cn are called cardinals.<br />
Remark. Clearly, if∣A∣,∣B∣
a4. Zermelo-Fraenkelsettheory<br />
(2) �ereex<strong>is</strong>tsan injective function A→ B.<br />
(3) �ereex<strong>is</strong>tsansurjective function B→A.<br />
Proof. Set κ∶=∣A∣ and λ∶=∣B∣ and let g∶↓κ→A and h∶↓λ→B be the<br />
corresponding bijections.<br />
(1)⇒(2) Since κ ≤ λ there ex<strong>is</strong>ts an <strong>is</strong>omorph<strong>is</strong>m f ∶ ↓κ → I<br />
between↓κ and an initial segment I⊆↓λ. In particular, f <strong>is</strong> injective.<br />
�e composition h○ f○ g −1 ∶ A→ B <strong>is</strong>the <strong>de</strong>sired injective function.<br />
(2)⇒(1) For a contradiction, suppose that there ex<strong>is</strong>ts an injective<br />
functionA→ B butwe have∣A∣>∣B∣. By(1)⇒(2),the latter impliesthat<br />
there <strong>is</strong> an injective function B→A. Hence, applying �eorem a2.1.12<br />
we find a bijection A→ B. It followsthat∣A∣=∣B∣. Contradiction.<br />
(2)⇒(3) Let f ∶ A→ B be injective. By Lemma a2.1.10 (b), there<br />
ex<strong>is</strong>ts a function g ∶ B → A such that g○ f = idA. Furthermore, it<br />
follows by Lemma a2.1.10(d)that g <strong>is</strong>surjective.<br />
(3)⇒(2) As above, given a surjective function f ∶ B→A we can<br />
apply Lemma a2.1.10(and the Axiom of Choice) to obtain an injective<br />
function g∶A→ Bwith f○ g= idB. ◻<br />
For every cardinal, there <strong>is</strong> a canonical setwithth<strong>is</strong> cardinality.<br />
Lemma 2.4. Forevery cardinal κ∈Cn,we have κ=∣↓κ∣. It followsthat<br />
Cn={α∈On∣∣↓α∣= α}.<br />
Exerc<strong>is</strong>e 2.1. Let α and β be ordinals suchthat∣α∣≤ β≤α.Show that<br />
∣α∣=∣β∣.<br />
Exerc<strong>is</strong>e 2.2. Prove that α∈ Cn, for every ordinal α≤ω. Hint. Show,<br />
by induction on α, that there <strong>is</strong> no surjective function↓α→↓β with<br />
α
2. Cardinals<br />
Proof. By �eorem a2.1.13, there ex<strong>is</strong>ts an injective functionA→℘(A)<br />
but no surjective one. By Lemma 2.3, it follows that∣A∣≤∣℘(A)∣ and<br />
∣℘(A)∣≰∣A∣. ◻<br />
Cn <strong>is</strong> a proper class since it <strong>is</strong> anunboun<strong>de</strong>dsubclass of On.<br />
Lemma 2.6. Cn <strong>is</strong>aproperclass.<br />
Proof. For a contradiction, suppose otherw<strong>is</strong>e. By Lemma a3.2.8, it follows<br />
that there <strong>is</strong> someα∈ Onsuchthat κ∣↓α∣,<br />
which impliesthat λ>α. A contradiction. ◻<br />
Lemma 2.7. On0≤ Cn≤On.<br />
Proof. Since Cn⊆On it followsthat Cn <strong>is</strong> awell-or<strong>de</strong>r.�erefore,there<br />
ex<strong>is</strong>ts an <strong>is</strong>omorph<strong>is</strong>m h∶Cn→I, forsome initial segment I⊆ On.<br />
By �eorem 2.5 we know that the function f ∶ On0 → Cn with<br />
f(α)∶=∣Sα∣ <strong>is</strong> strictly increasing. Consequently, we have On0 ≤ Cn,<br />
by Lemma a3.2.11. ◻<br />
Remark. With the Axiom of Replacement which we will introduce in<br />
Section 5 we can ac<strong>tu</strong>ally prove that⟨On0,∈⟩≅⟨On,
a4. Zermelo-Fraenkelsettheory<br />
Note that, by our <strong>de</strong>finition of a cardinal, we have ωα = ℵα and<br />
ℵ0= ω0= ω. Furthermore,ℵ + α=ℵα+1.Sincewe have <strong>de</strong>fined the operation<br />
κ + differently for cardinals and ordinalswewill useth<strong>is</strong> notation<br />
only for cardinals intheremain<strong>de</strong>r ofth<strong>is</strong> book. Ifwe consi<strong>de</strong>rthesuccessor<br />
of an ordinal α wewillwriteα+ 1.<br />
3. Cardinalarithmetic<br />
Similarly to ordinals we can <strong>de</strong>fine arithmetic operations on cardinals.<br />
Notethat, except for finite cardinals,these operations are different from<br />
the ordinal operations. �erefore,we have chosen differentsymbolsto<br />
<strong>de</strong>notethem.<br />
Definition 3.1. Let κ, λ∈Cn be cardinals. We <strong>de</strong>fine<br />
κ⊕λ∶=∣↓κ⊍↓λ∣ , κ⊗ λ∶=∣↓κ×↓λ∣ , κ λ ∶=∣↓κ ↓λ ∣.<br />
�e following lemmas follows immediately fromthe <strong>de</strong>finition if one<br />
recalls that, for κ∶=∣A∣ and λ∶=∣B∣, there ex<strong>is</strong>t bijections A→↓κ and<br />
B→↓λ.<br />
Lemma 3.2. Let AandBbesets.<br />
∣A⊍B∣=∣A∣⊕∣B∣ , ∣A×B∣=∣A∣⊗∣B∣ , ∣A B ∣=∣A∣ ∣B∣ .<br />
Corollary 3.3. Forallα,β∈On,we have<br />
∣↓(α+β)∣=∣↓α∣⊕∣↓β∣ and ∣↓(αβ)∣=∣↓α∣⊗∣↓β∣.<br />
�e corresponding equation for ordinal exponentiation will be <strong>de</strong>layed<br />
until Lemma 4.4.<br />
Exerc<strong>is</strong>e 3.1. Prove that, if A <strong>is</strong> a set then∣℘(A)∣=2 ∣A∣ . Hint.Take the<br />
obvious bijection℘(A)→2 A .<br />
116<br />
For finite cardinalsthese operations coinci<strong>de</strong>withtheusual ones.
Lemma 3.4. For m, n
a4. Zermelo-Fraenkelsettheory<br />
(e) Ifthere <strong>is</strong> an injective function f∶ B→C,we can <strong>de</strong>fine an injective<br />
function A⊍B→A⊍C by<br />
⟨0,a⟩↦⟨0,a⟩ and ⟨1,b⟩↦⟨1, f(b)⟩.<br />
(f) If κ≥ℵ0= ω then κ=ω+α, for some α∈On. We can <strong>de</strong>fine a<br />
bijection↓ω→↓(ω+1) by0↦ω and n↦n−1, for n>0.�<strong>is</strong> function<br />
can be exten<strong>de</strong>d to a bijection↓ω⊍↓α→↓ω⊍↓α⊍[1]. Conversely, if<br />
κκ. ◻<br />
Lemma 3.6. Let κ, λ, µ∈Cn.<br />
(a)(κ⊗λ)⊗ µ=κ⊗(λ⊗ µ)<br />
(b) κ⊗λ= λ⊗κ<br />
(c) κ⊗0=0, κ⊗ 1=κ, κ⊗2=κ⊕κ.<br />
(d) κ⊗(λ⊕ µ)=(κ⊗ λ)⊕(κ⊗ µ)<br />
(e) λ≤ µ implies κ⊗λ≤κ⊗ µ.<br />
Proof. (a)�ere <strong>is</strong> a canonical bijection(A×B)×C→A×(B×C)with<br />
⟨⟨a,b⟩,c⟩↦⟨a,⟨b,c⟩⟩.<br />
(b)�ere <strong>is</strong> a canonical bijection A×B→B×Awith⟨a,b⟩↦⟨b,a⟩.<br />
(c)A×∅=∅.�ere are canonical bijections<br />
A×{0}→A and A⊍A=[2]×A→ A×[2].<br />
(d)�ere ex<strong>is</strong>ts a bijection A×(B⊍C)→(A×B)⊍(A×C)with<br />
⟨a,⟨0,b⟩⟩↦⟨0,⟨a,b⟩⟩ and ⟨a,⟨1,c⟩⟩↦⟨1,⟨a,c⟩⟩.<br />
(e) Given an injective function f ∶ B→C we <strong>de</strong>fine an injective<br />
function A×B→A×C by⟨a,b⟩↦⟨a, f(b)⟩. ◻<br />
Lemma 3.7. Let κ, λ, µ, ν∈Cn.<br />
118<br />
(a)(κ λ ) µ = κ λ⊗µ<br />
(b)(κ⊗λ) µ = κ µ ⊗ λ µ
(c) κ λ⊕µ = κ λ ⊗ κ µ<br />
(d) κ 0 = 1, κ 1 = κ, κ 2 = κ⊗κ.<br />
(e) If κ≤λand µ≤νthen κ µ ≤ λ ν .<br />
(f) κ
a4. Zermelo-Fraenkelsettheory<br />
β1<br />
9 10 11<br />
4 5<br />
1<br />
0<br />
3<br />
2<br />
8<br />
7<br />
6<br />
β0<br />
15<br />
14<br />
13<br />
12<br />
Figure 1.. Or<strong>de</strong>ring on↓κ×↓κ<br />
Exerc<strong>is</strong>e 3.2. Provethatℵ0⊗ℵ0=ℵ0 byshowingthatthe function<br />
<strong>is</strong> bijective.<br />
↓ω×↓ω→↓ω∶⟨i, k⟩↦ 1 2(i+k)(i+k+ 1)+ k<br />
Westart by computing κ⊗ κ by induction on κ≥ℵ0.<br />
�eorem 3.8. If κ≥ℵ0 then κ⊗ κ=κ.<br />
Proof. We have κ = κ⊗ 1≤κ⊗κ. For the converse, we prove that<br />
κ⊗κ≤ κ by induction on κ.<br />
Notethat,since κ <strong>is</strong> a cardinal we have α
3. Cardinalarithmetic<br />
<strong>is</strong> an initial subset of K. If ω≤α
a4. Zermelo-Fraenkelsettheory<br />
4. Cofinality<br />
Frequently, we will construct objects as the union of an increasing sequenceA0⊆<br />
A1⊆ ... ofsets. Inth<strong>is</strong>sectionwewills<strong>tu</strong>dythe cardinality<br />
ofsuchunions.<br />
Definition 4.1. For asequence(κi)i
4. Cofinality<br />
Since∣(↓α) n ∣≤∣↓α∣⊕ℵ0, for n
a4. Zermelo-Fraenkelsettheory<br />
Definition 4.7. (a) Let⟨A,≤⟩ be a linear or<strong>de</strong>r. AsubsetX⊆A <strong>is</strong>cofinal<br />
inA if, for every a∈A,there <strong>is</strong>some element x∈Xwith a≤x.<br />
We call a function f∶ B→A cofinal ifrng f <strong>is</strong> cofinal inA.<br />
(b) �ecofinality cf α of an ordinal α <strong>is</strong> the minimal ordinal β such<br />
thatthere ex<strong>is</strong>ts a cofinal function f∶↓β→↓α.<br />
Exerc<strong>is</strong>e 4.1. Provethat every linear or<strong>de</strong>r⟨A,≤⟩ contains a cofinalsubset<br />
X⊆ Asuchthat⟨X,≤⟩ <strong>is</strong>well-or<strong>de</strong>red.<br />
Lemma 4.8. Let⟨A,≤⟩ be a linear or<strong>de</strong>r. If X <strong>is</strong> cofinal in A and Y <strong>is</strong><br />
cofinal in X then Y <strong>is</strong>cofinal in A.<br />
We can restate the <strong>de</strong>finition of the cofinality in a moreuseful form<br />
as follows.<br />
Lemma4.9. If(αi)i
Proof. �e function g∶↓β→↓α with<br />
g(γ)=max{ f(γ), sup{ g(η)+1∣η h(γ)}.<br />
�<strong>is</strong> function <strong>is</strong> cofinal since, given η
a4. Zermelo-Fraenkelsettheory<br />
Example. ω andℵ1 areregularwhileℵω <strong>is</strong>singular.<br />
�e next lemma indicatesthatthe notion of cofinality <strong>is</strong> mainly interesting<br />
for cardinals.<br />
Lemma4.15. Everyregular ordinal <strong>is</strong>acardinal.<br />
Proof. Let α ∈ On∖Cn be an ordinal that <strong>is</strong> not a cardinal and set<br />
κ∶=∣α∣
4. Cofinality<br />
Cardinal exponentiation <strong>is</strong> the least un<strong>de</strong>rstood operation of those<br />
introducedso far.�ere are many openquestionsthattheusual axioms<br />
of set theory are not strong enough to answer. For example, we do not<br />
knowwhatthevalue of2 ℵ0 <strong>is</strong>. Given an arbitrary mo<strong>de</strong>l ofsettheorywe<br />
can construct a new mo<strong>de</strong>lwhere2 ℵ0 =ℵ1, butwe can also find mo<strong>de</strong>ls<br />
where2 ℵ0 equalsℵ2 orℵ3.<br />
Intheremain<strong>de</strong>r of th<strong>is</strong>section wepresentsome elementary results<br />
thatcan beproved.�e notion of cofinality appears atseveralplaces in<br />
these proofs. First, let us compute the cardinality of all stages Sα, by a<br />
simple induction.<br />
Definition 4.19. We <strong>de</strong>finethe cardinalℶα(κ)(‘bethalpha’), forα∈On<br />
and κ∈ Cn,recursively by<br />
ℶ0(κ)∶= κ ,<br />
ℶα+1(κ)∶=2 ℶα(κ) ,<br />
and ℶδ(κ)∶=sup{ℶα(κ)∣ α κ.<br />
127
a4. Zermelo-Fraenkelsettheory<br />
Proof. Fix a cofinal function f∶↓λ→↓κ. By�eorem 4.6,we have<br />
κ λ =∣(↓κ) ↓λ ∣=∣∏↓κ∣>∣⊍ ↓ f(α)∣≥∣↓κ∣= κ.<br />
α
(c) If λ
a4. Zermelo-Fraenkelsettheory<br />
Hence, κ=ℶδ. Sinceℶδ= κ>ℵ0 we have δ> 0. To show that δ <strong>is</strong><br />
a limit suppose that δ=α+1. �enℶα < κ impliesℶδ = 2 ℶα < κ.<br />
Contradiction. ◻<br />
We conclu<strong>de</strong> th<strong>is</strong> section with some results about sets of sequences<br />
in<strong>de</strong>xed by ordinals. Aswewillsee inSection b2.1,such aset formsthe<br />
domain of atree. Recall that asequence in<strong>de</strong>xed by an ordinal α <strong>is</strong> just<br />
a function↓α→A.<br />
Definition 4.29. IfA <strong>is</strong> aset and α∈ On,we <strong>de</strong>fine<br />
A α ∶= A ↓α<br />
and A
5. �eAxiom ofReplacement<br />
Lemma4.33. If κ <strong>is</strong>an infiniteregularcardinalthen κ
a4. Zermelo-Fraenkelsettheory<br />
�eorem5.2. �e followingstatements areequivalent:<br />
(1) If F <strong>is</strong>afunctionandA⊆dom F <strong>is</strong>asetthen F[A] <strong>is</strong>alsoaset.<br />
(2) If F <strong>is</strong>afunctionand dom F <strong>is</strong>asetthenso <strong>is</strong> rng F.<br />
(3) A function F <strong>is</strong>aset if,and only if, dom F <strong>is</strong>aset.<br />
(4) �ere ex<strong>is</strong>ts no bijection F∶a→B between a set a and a proper<br />
classB.<br />
(5) AclassA <strong>is</strong>aset if,and only if,∣A∣
5. �eAxiom ofReplacement<br />
(4)⇒(2) Let F∶A→ B be a function where A= dom F <strong>is</strong> aset. Let<br />
B0∶= rng F. Since the function F∶a→B0 <strong>is</strong> surjective there ex<strong>is</strong>ts<br />
a function G∶B0→a such that F○ G= idB0. Let A0∶= rng G. �e<br />
restriction F∶ A0→ B0 <strong>is</strong> a bijection. Since A0⊆ A <strong>is</strong> a set so <strong>is</strong> B0=<br />
rng F. ◻<br />
Axiom of Replacement. If F <strong>is</strong>afunction and dom F <strong>is</strong>aset then so <strong>is</strong><br />
rng F.<br />
Le<strong>tu</strong>s finallyprovetheresultsweprom<strong>is</strong>ed intheprecedingsections.<br />
First,upto <strong>is</strong>omorph<strong>is</strong>m, On <strong>is</strong>the onlywell-or<strong>de</strong>rthat <strong>is</strong> aproper class.<br />
Lemma 5.3. Let A=⟨A,≤A⟩and B=⟨B,≤B⟩bewell-or<strong>de</strong>rs. IfAandB<br />
areproperclassesthen A≅B.<br />
Proof. Suppose that A≇B. By �eorem a3.1.11, there either ex<strong>is</strong>ts an<br />
<strong>is</strong>omorph<strong>is</strong>m f ∶ A→↓b, for some b∈B, or some <strong>is</strong>omorph<strong>is</strong>m g∶<br />
↓a→B, for somea∈A. By symmetry,we may assumew.l.o.g.the latter.<br />
↓a <strong>is</strong> asetsince≤A <strong>is</strong> le�-narrow. Hence, bythe Axiom ofReplacement,<br />
B= g[↓a] <strong>is</strong> also aset. Contradiction. ◻<br />
It follows that it does not matter which of the two <strong>de</strong>finitions of an<br />
ordinalwe adopt.<br />
Corollary 5.4. On0≅ Cn≅On.<br />
Finally, we state the general form of the Principle of Transfinite Recursion.<br />
�eorem 5.5 (Principle of Transfinite Recursion). If H∶A
a4. Zermelo-Fraenkelsettheory<br />
Lemma 5.6. Every strictly continuous function f∶ On→On has arbitrarily<br />
large fixed points.<br />
Proof. For every α∈Onwe have to find a fixed point γ≥α. If F <strong>is</strong> the<br />
fixed-point induction of f over α then F[↓ω] ex<strong>is</strong>ts. By Lemma a3.3.13<br />
it follows that γ∶= F(∞)= F(ω)≥ α <strong>is</strong> a fixed point of f . ◻<br />
Corollary 5.7. �erearearbitrarily largecardinals κ suchthat cf κ=ℵ0<br />
an<strong>de</strong>itherℵκ= κ orℶκ= κ.<br />
Proof. �e functions f∶ α↦ℵα and g∶α↦ℶα arestrictly continuous.<br />
Furthermore, they are <strong>de</strong>fined by transfinite recursion. �erefore, �eorem<br />
5.5 implies that their domain <strong>is</strong> all of On. By Lemma a3.3.13 and<br />
Lemma 5.6, it follows that f and g have arbitrarily large inductive fixed<br />
points κ, andthese fixedpoints are ofthe form<br />
κ=sup{ f n (α)∣n
6. Conclusion<br />
ex<strong>is</strong>tence of certain sets. Infinity and Replacement ensure that the cumulative<br />
hierarchy <strong>is</strong> long enough. �ere are as many stages as there<br />
are ordinals. �e Axioms of Separation and Choice on the other hand<br />
makethe hierarchywi<strong>de</strong> by ensuringthatthepower-set operationyields<br />
enoughsubsets. Inparticular, every <strong>de</strong>finablesubset ex<strong>is</strong>ts and on every<br />
setthere ex<strong>is</strong>ts awell-or<strong>de</strong>ring.<br />
Finally, let us note that the usual <strong>de</strong>finition of ZFC <strong>is</strong> based on a different<br />
axiomat<strong>is</strong>ation wherethe Axiom of Creation <strong>is</strong>replaced by four<br />
other axioms and the Axiom of Infinity <strong>is</strong> stated in a slightly different<br />
way. Nevertheless,we are justified in callingthe abovetheory ZFCsince<br />
thetwovariants are equivalent: every mo<strong>de</strong>lsat<strong>is</strong>fying one ofthe axiom<br />
systems alsosat<strong>is</strong>fiesthe other one, andviceversa.<br />
135
a4. Zermelo-Fraenkelsettheory<br />
136
Litera<strong>tu</strong>re<br />
Settheory<br />
M. D.Potter,Sets.An Introduction, OxfordUniversityPress 1990.<br />
A. Lévy,BasicSet�eory,Springer 1979, Dover2002.<br />
K. Kunen,Set�eory. An Introductionto In<strong>de</strong>pen<strong>de</strong>nceProofs, North-Holland<br />
1983.<br />
T. J. Jech,Set�eory, 3rd ed., Springer 2003.<br />
Algebra<br />
P. M. Cohn,UniversalAlgebra,2nd ed.,Springer 1981.<br />
P. M. Cohn,BasicAlgebra,Springer2003.<br />
S. Lang,Algebra, 3rd ed., Springer 2002.<br />
S. MacLane,Categories fortheWorking Mathematician, 2nd ed., Springer 1998.<br />
Topologyand latticetheory<br />
R. Engelking, GeneralTopology, 2nd ed., Hel<strong>de</strong>rmann 1989.<br />
C.-A. Faure, A. Frölicher, Mo<strong>de</strong>rnProjective Geometry, Kluwer2000.<br />
G. Gierz, K. H. Hofmann, K. Keimel, J. D. Lawson, M. M<strong>is</strong>love, and D.S.Scott,<br />
Continuous LatticesandDomains, CambridgeUniversityPress,2003.<br />
Mo<strong>de</strong>ltheory<br />
D. Marker, Mo<strong>de</strong>l�eory:An Introduction, Springer 2002.<br />
K. Tent and M.Ziegler,ACourse in Mo<strong>de</strong>l�eory, CambridgeUniversityPress<br />
2012.<br />
logic, algebra & geometry2012-08-08 — ©achim blumensath 1007
Litera<strong>tu</strong>re<br />
W. Hodges, Mo<strong>de</strong>l�eory, CambridgeUniversityPress 1993.<br />
B.Poizat,ACourse in Mo<strong>de</strong>l�eory,Springer2000.<br />
C. C. Chang and H. J. Ke<strong>is</strong>ler, Mo<strong>de</strong>l�eory, 3rd ed., North-Holland 1990.<br />
General mo<strong>de</strong>ltheory<br />
J. Barw<strong>is</strong>e and S. Feferman, eds., Mo<strong>de</strong>l-�eoretic Logics, Springer 1985.<br />
J. T. Baldwin,Categoricity, AMS 2010.<br />
R. Diaconescu, Insti<strong>tu</strong>tion-in<strong>de</strong>pen<strong>de</strong>nt Mo<strong>de</strong>l�eory, Birkhäuser2008.<br />
H.-D. Ebbinghaus and J. Flum, Finite Mo<strong>de</strong>l�eory,Springer 1995.<br />
J. Adámek and J.Rosický, LocallyPresentableandAccessibleCategories,<br />
CambridgeUniversityPress 1994.<br />
Stabilitytheory<br />
S. Buechler,EssentialStability�eory, Springer 1996.<br />
E. Casanovas,Simple�eoriesand Hyperimaginaries, CambridgeUniversity<br />
Press2011.<br />
A.Pillay, GeometricStability�eory, OxfordSciencePublications 1996.<br />
F. O.Wagner,Simple�eories, Kluwer Aca<strong>de</strong>micPubl<strong>is</strong>hers2000.<br />
S.Shelah,Classification�eory,2nd ed., North-Holland 1990.<br />
1008
SymbolIn<strong>de</strong>x<br />
Chaptera1<br />
S universe ofsets, 5<br />
a∈b membership, 5<br />
a⊆b subset, 5<br />
HF hereditary finitesets,7<br />
⋂A intersection, 11<br />
A∩B intersection, 11<br />
A∖B difference, 11<br />
acc(A) accumulation, 12<br />
fnd(A) foun<strong>de</strong>dpart, 13<br />
⋃A union,21<br />
A∪B union,21<br />
℘(A) powerset,21<br />
cutA cut of A,22<br />
Chaptera2<br />
⟨a0,... ,an−1⟩ <strong>tu</strong>ple,27<br />
A×B cartesianproduct,27<br />
dom f domain of f ,28<br />
rng f range of f ,29<br />
f(a) image of a un<strong>de</strong>r f ,29<br />
f∶ A→B function,29<br />
B A set of all functions<br />
f∶ A→ B,29<br />
idA i<strong>de</strong>ntity function, 30<br />
S○R composition of relations,<br />
30<br />
g○ f composition of functions,<br />
30<br />
R −1 inverse ofR, 30<br />
R −1 (a) inverse image, 30<br />
R∣C restriction, 30<br />
R↾C le� restriction, 31<br />
R[C] image ofC, 31<br />
(ai)i∈I sequence, 37<br />
∏i Ai<br />
pri product, 37<br />
projection, 37<br />
ā sequence, 38<br />
⊍i Ai d<strong>is</strong>join<strong>tu</strong>nion, 38<br />
A⊍B d<strong>is</strong>join<strong>tu</strong>nion, 38<br />
ini insertion map, 39<br />
⇓X initial segment, 41<br />
⇑X final segment, 41<br />
↓X initial segment, 41<br />
↑X final segment, 41<br />
[a,b] closed interval, 41<br />
(a,b) open interval, 41<br />
maxX greatest element, 42<br />
minX minimal element, 42<br />
supX supremum, 42<br />
inf X infimum, 42<br />
logic, algebra & geometry2012-08-08 — ©achim blumensath 1009
Symbol In<strong>de</strong>x<br />
A≅B <strong>is</strong>omorph<strong>is</strong>m, 44<br />
fix f fixedpoints, 48<br />
lfp f least fixedpoint, 48<br />
gfp f greatest fixedpoint, 48<br />
[a]∼ equivalence class, 54<br />
A/∼ set of∼-classes, 54<br />
TC(R) transitive closure, 55<br />
Chaptera3<br />
a + successor, 59<br />
ord(A) or<strong>de</strong>rtype,64<br />
On class of ordinals,64<br />
On0 von Neumann ordinals,69<br />
ρ(a) rank,73<br />
A
Emb(Σ) category of embeddings,<br />
153<br />
Set∗<br />
category of pointed sets,<br />
153<br />
Set 2 category of pairs, 153<br />
Cat ∗ dual category, 155<br />
F≅G na<strong>tu</strong>ral <strong>is</strong>omorph<strong>is</strong>m, 158<br />
Cong(A) set of congruencerelations,<br />
161<br />
Cong(A) congruence lattice, 161<br />
A/∼ quotient, 165<br />
Chapterb2<br />
∣x∣ length of asequence, 173<br />
x⋅y concatenation, 173<br />
⪯ prefix or<strong>de</strong>r, 173<br />
≤lex lexicographic or<strong>de</strong>r, 173<br />
∣v∣ level of avertex, 176<br />
frk(v) foundationrank, 178<br />
a⊓b infimum, 181<br />
a⊔b supremum, 181<br />
a ∗ complement, 184<br />
L op dual lattice, 190<br />
cl↓(X) i<strong>de</strong>al generated by X, 190<br />
cl↑(X) filter generated by X, 190<br />
B2<br />
two-element boolean<br />
algebra, 195<br />
ht(a) height of a,200<br />
rkP(a) partitionrank,204<br />
<strong>de</strong>g P (a) partition <strong>de</strong>gree,208<br />
Chapterb3<br />
Symbol In<strong>de</strong>x<br />
T[Σ,X] finiteΣ-terms,213<br />
tv subterm atv,214<br />
free(t) freevariables,217<br />
t A [β] value of t,217<br />
T[Σ,X] term algebra,218<br />
t[x/s] substi<strong>tu</strong>tion,220<br />
SigVar category ofsigna<strong>tu</strong>res and<br />
variables,221<br />
Sig category ofsigna<strong>tu</strong>res,222<br />
Var category ofvariables,222<br />
Term category ofterms,222<br />
A∣µ µ-reduct of A, 223<br />
Str[Σ] class ofΣ-struc<strong>tu</strong>res, 223<br />
Str[Σ,X] class of allΣ-struc<strong>tu</strong>res<br />
withvariable<br />
assignments, 223<br />
StrVar category of struc<strong>tu</strong>res and<br />
assignments, 223<br />
Str category of struc<strong>tu</strong>res, 223<br />
∏i A i direct product,225<br />
⟦φ⟧ set of indices,227<br />
ā∼u ¯ b filter equivalence,227<br />
u∣J restriction of uto J,228<br />
∏i A i /u reducedproduct,228<br />
A u ultrapower,229<br />
lim<br />
�→ D direct limit, 237<br />
⋃i A i union of a chain,241<br />
lim<br />
←� D inverse limit,241<br />
Chapterb4<br />
cl(A) closure of A,245<br />
int(A) interior of A,245<br />
1011
Symbol In<strong>de</strong>x<br />
∂A boundary of A,245<br />
rkCB(x/A) Cantor-Bendixsonrank,<br />
267<br />
spec(L) spectrum of L,272<br />
⟨x⟩ basic closedset,272<br />
clop(S) algebra of clopensubsets,<br />
276<br />
Chapterb5<br />
Aut M automorph<strong>is</strong>m group,288<br />
G/U set of cosets,288<br />
G/N factor group,290<br />
Sym Ω symmetric group,291<br />
ga action of g on a,292<br />
Gā orbit of ā,292<br />
G(X) pointw<strong>is</strong>estabil<strong>is</strong>er,293<br />
setw<strong>is</strong>e stabil<strong>is</strong>er, 293<br />
G{X}<br />
dclG(U) G-<strong>de</strong>finitional closure,297<br />
aclG(U) G-algebraic closure,297<br />
⟨ā↦ ¯ b⟩ basic openset ofthe group<br />
topology,298<br />
<strong>de</strong>gp <strong>de</strong>gree, 302<br />
Idl(R) lattice of i<strong>de</strong>als, 303<br />
R/a quotient of aring, 305<br />
Ker h kernel, 305<br />
spec(R) spectrum, 306<br />
⊕i Mi directsum, 309<br />
M (I) direct power, 309<br />
dim V dimension, 312<br />
FF(R) field of fractions, 314<br />
K(ā) subfield generated by ā, 317<br />
p[x] polynomial function, 319<br />
Aut(L/K)automorph<strong>is</strong>ms over K,<br />
326<br />
1012<br />
∣a∣ absolutevalue, 329<br />
Chapterc1<br />
ZL[K,X] Zar<strong>is</strong>ki logic, 345<br />
⊧ sat<strong>is</strong>faction relation, 346<br />
BL(B) boolean logic, 346<br />
FOκℵ0[Σ,X] infinitary first-or<strong>de</strong>r<br />
logic, 347<br />
¬φ negation, 347<br />
⋀Φ conjunction, 347<br />
⋁Φ d<strong>is</strong>junction, 347<br />
∃xφ ex<strong>is</strong>tential quantifier, 347<br />
∀xφ universalquantifier, 347<br />
FO[Σ,X] first-or<strong>de</strong>r logic, 347<br />
A⊧φ[β] sat<strong>is</strong>faction, 348<br />
true true, 349<br />
false false, 349<br />
φ∨ψ d<strong>is</strong>junction, 349<br />
φ∧ψ conjunction, 349<br />
φ→ψ implication, 349<br />
φ↔ψ equivalence, 349<br />
free(φ) freevariables, 352<br />
qr(φ) quantifierrank, 355<br />
ModL(Φ) class of mo<strong>de</strong>ls, 356<br />
Φ⊧φ entailment, 362<br />
≡ logical equivalence, 362<br />
Φ ⊧ closureun<strong>de</strong>r entailment,<br />
362<br />
�L(J) L-theory, 363<br />
≡L L-equivalence, 364<br />
dnf(φ) d<strong>is</strong>junctive normal form,<br />
369<br />
cnf(φ) conjunctive normal form,<br />
369
nnf(φ) negation normal form, 371<br />
Logi$ category of logics, 380<br />
∃ λ xφ cardinality quantifier, 383<br />
FOκℵ0(wo) FOwithwell-or<strong>de</strong>ring<br />
quantifier, 383<br />
W well-or<strong>de</strong>ring quantifier,<br />
383<br />
QK Lindström quantifier, 384<br />
SOκℵ0[Σ,Ξ] second-or<strong>de</strong>r logic, 385<br />
MSOκℵ0[Σ,Ξ] monadic<br />
second-or<strong>de</strong>r logic, 385<br />
PO category of partial or<strong>de</strong>rs,<br />
390<br />
Lb Lin<strong>de</strong>nbaum functor, 390<br />
¬φ negation, 391<br />
φ∨ψ d<strong>is</strong>junction, 391<br />
φ∧ψ conjunction, 392<br />
L∣Φ restriction toΦ, 393<br />
L/Φ local<strong>is</strong>ation toΦ, 393<br />
⊧Φ consequence moduloΦ,<br />
393<br />
≡Φ equivalence moduloΦ, 393<br />
Chapterc2<br />
EmbL(Σ) category of L-embeddings,<br />
395<br />
QFκℵ0 [Σ,X] quantifier-free<br />
formulae, 396<br />
∃∆ ex<strong>is</strong>tential closure of∆, 396<br />
∀∆ universal closure of ∆, 396<br />
∃κℵ0 ex<strong>is</strong>tential formulae, 396<br />
∀κℵ0 universal formulae, 396<br />
∃ + κℵ0 positive ex<strong>is</strong>tential<br />
formulae, 396<br />
Symbol In<strong>de</strong>x<br />
⪯∆ ∆-extension, 400<br />
⪯ elementary extension, 400<br />
Φ ⊧ ∆ ∆-consequences ofΦ, 422<br />
≤∆ preservation of<br />
∆-formulae, 422<br />
Chapterc3<br />
S(L) set oftypes, 427<br />
⟨Φ⟩ types containingΦ, 427<br />
tpL (ā/M)L-type of ā, 428<br />
S ¯s L(T) typespace for atheory, 428<br />
S ¯s L(U) typespace overU, 428<br />
S(L) typespace, 433<br />
f(p) conjugate of p, 443<br />
S∆(L) S(L∣∆)with topology<br />
induced from S(L), 456<br />
⟨Φ⟩∆ closedset in S∆(L), 456<br />
p∣∆ restrictionto∆, 459<br />
tp∆ (ā/U)∆-type of ā, 459<br />
Chapterc4<br />
≡α α-equivalence, 473<br />
≡∞ ∞-equivalence, 473<br />
pIso κ (A, B) partial <strong>is</strong>omorph<strong>is</strong>ms,<br />
474<br />
ā↦ ¯ b map ai↦ bi, 474<br />
∅ the empty function, 474<br />
Iα(A, B) back-and-forthsystem, 475<br />
I∞(A, B) limit ofthesystem, 477<br />
≅α α-<strong>is</strong>omorphic, 477<br />
≅∞ ∞-<strong>is</strong>omorphic, 477<br />
m=k n equalityupto k, 480<br />
1013
Symbol In<strong>de</strong>x<br />
φ α A,ā<br />
Hintikka formula, 482<br />
EFα(A,ā, B, ¯ b)<br />
Ehrenfeucht-Fraïssé<br />
game, 485<br />
EF κ ∞(A,ā, B, ¯ b)<br />
Ehrenfeucht-Fraïssé<br />
game, 485<br />
I κ FO(A, B)partial FO-maps ofsize κ,<br />
494<br />
⊑ κ <strong>is</strong>o ∞κ-simulation, 495<br />
≅ κ <strong>is</strong>o ∞κ-<strong>is</strong>omorphic, 495<br />
A⊑ κ 0 B I κ 0(A, B)∶A⊑ κ <strong>is</strong>o B, 495<br />
A≡ κ 0 B I κ 0(A, B)∶A≡ κ <strong>is</strong>o B, 495<br />
A⊑ κ FO B I κ FO(A, B)∶A⊑ κ <strong>is</strong>o B, 495<br />
A≡ κ FO B I κ FO(A, B)∶A≡ κ <strong>is</strong>o B, 495<br />
A⊑ κ ∞ B I κ ∞(A, B)∶ A⊑ κ <strong>is</strong>o B, 495<br />
A≡ κ ∞ B I κ ∞(A, B)∶ A≡ κ <strong>is</strong>o B, 495<br />
G(A) Gaifman graph, 501<br />
Chapterc5<br />
L≤ L ′ L ′ <strong>is</strong> as expressive as L, 509<br />
(a) algebraic, 510<br />
(b) boolean closed, 510<br />
(b+) positive boolean closed,<br />
510<br />
(c) compactness, 510<br />
(cc) countable compactness,<br />
510<br />
(fop) finite occurrenceproperty,<br />
510<br />
(kp) Karpproperty, 510<br />
(lsp) Löwenheim-Skolem<br />
property, 510<br />
(rel) closedun<strong>de</strong>r<br />
1014<br />
relativ<strong>is</strong>ations, 510<br />
(sub) closedun<strong>de</strong>rsubsti<strong>tu</strong>tions,<br />
510<br />
(<strong>tu</strong>p) Tarskiunionproperty, 510<br />
hnκ(L) Hanf number, 514<br />
lnκ(L) Löwenheim number, 514<br />
wnκ(L) well-or<strong>de</strong>ring number, 514<br />
occ(L) occurrence number, 514<br />
prΓ (K) Γ-projection, 532<br />
PCκ(L,Σ) projective L-classes, 533<br />
L0≤ κ pc L1 projective reduction, 533<br />
RPCκ(L,Σ) relativ<strong>is</strong>ed projective<br />
L-classes, 537<br />
L0≤ κ rpc L1 relativ<strong>is</strong>ed projective<br />
reduction, 537<br />
∆(L) interpolation closure, 545<br />
ifp f inductive fixedpoint, 554<br />
lim inf f leastpartial fixedpoint, 554<br />
limsup f greatestpartial fixedpoint,<br />
554<br />
fφ function <strong>de</strong>fined byφ, 561<br />
FOκℵ0(LFP) least fixed-point logic,<br />
561<br />
FOκℵ0(IFP) inflationary fixed-point<br />
logic, 562<br />
FOκℵ0(PFP) partial fixed-point<br />
logic, 562<br />
⊲φ stage compar<strong>is</strong>on, 572<br />
Chapterd1<br />
tor(G) torsionsubgroup,603<br />
a/n div<strong>is</strong>or, 604<br />
DAG theory of div<strong>is</strong>ible<br />
torsion-free abelian
groups,604<br />
ODAG theory of or<strong>de</strong>red div<strong>is</strong>ible<br />
abelian groups,604<br />
div(G) div<strong>is</strong>ible closure,605<br />
F field axioms,608<br />
ACF theory of algebraically<br />
closed fields,608<br />
RCF theory ofreal closed fields,<br />
609<br />
Chapterd2<br />
(
Symbol In<strong>de</strong>x<br />
Chapter e3<br />
Subκ(A) substruc<strong>tu</strong>res of A,759<br />
atp(ā) atomictype,765<br />
ηpq extension axiom,765<br />
T[K] extension axioms forK,<br />
766<br />
Tran[Σ] randomtheory,766<br />
κn(φ) number of mo<strong>de</strong>ls,768<br />
Pr n<br />
M[M⊧φ] <strong>de</strong>nsity of mo<strong>de</strong>ls,768<br />
Chapter e4<br />
increasing κ-<strong>tu</strong>ples,773<br />
κ→(µ) ν<br />
λ partition theorem, 773<br />
pf(η,ζ) prefix ofζ of length∣η∣, 778<br />
T∗(κ
(mon) Monotonicity, 928<br />
(nor) Normality, 928<br />
(lrf) Le�Reflexivity, 928<br />
(ltr) Le�Transitivity, 928<br />
(fin) Finite Character, 929<br />
(sym) Symmetry, 929<br />
(rbm) Right Base Monotonicity,<br />
929<br />
(srb) StrongRight Boun<strong>de</strong>dness,<br />
929<br />
cl √ closure operation<br />
associatedwith √ , 934<br />
(inv) Invariance, 941<br />
(sfin) Strong Finite Character,<br />
941<br />
(ext) Extension, 941<br />
(lbd) Le� Boun<strong>de</strong>dness, 941<br />
(rbd) Right Boun<strong>de</strong>dness, 941<br />
lb( √ ) le� boun<strong>de</strong>dness cardinal<br />
of √ , 941<br />
rb( √ ) right boun<strong>de</strong>dness<br />
cardinal of √ , 941<br />
A df√ U B <strong>de</strong>finable over, 943<br />
Symbol In<strong>de</strong>x<br />
A at√ U B <strong>is</strong>olated over, 943<br />
A s√<br />
U B non-splitting over, 943<br />
A u√<br />
U B finitely sat<strong>is</strong>fiable, 948<br />
Av(u/B) average type of u, 948<br />
√<br />
∗<br />
forking relation to √ , 953<br />
A i√<br />
U B globally invariant over, 956<br />
Chapter f3<br />
A d√<br />
U B non-dividing, 965<br />
A f√<br />
U B non-forking, 965<br />
(ind) In<strong>de</strong>pen<strong>de</strong>nce �eorem,<br />
992<br />
ā∼ ls U ¯ b ind<strong>is</strong>cernible sequence<br />
startingwith ā, ¯ b,... ,<br />
993<br />
ā≡ ls U ¯ b Lascar strong type<br />
equivalence, 993<br />
1017
Symbol In<strong>de</strong>x<br />
1018
In<strong>de</strong>x<br />
abelian group,287<br />
abstract elementary class, 839<br />
abstract in<strong>de</strong>pen<strong>de</strong>nce relation, 928<br />
accumulation, 12<br />
accumulationpoint,266<br />
action,292<br />
acyclic, 420<br />
addition of cardinals, 116<br />
addition of ordinals, 89<br />
adjoint functors,220<br />
affine geometry, 881<br />
aleph, 115<br />
algebraic, 139, 712<br />
Aut-algebraic,712<br />
G-algebraic,297<br />
algebraic class, 840<br />
algebraic closure,297,712<br />
algebraic closure operator, 51<br />
algebraic diagram, 401<br />
algebraic elements, 322<br />
algebraic field extensions, 322<br />
algebraic logic, 389<br />
algebraic prime mo<strong>de</strong>l, 592<br />
algebraically closed, 712<br />
algebraically closed field, 322, 608<br />
algebraically in<strong>de</strong>pen<strong>de</strong>nt, 321<br />
almost strongly minimal theory, 900<br />
amalgamation class, 849<br />
amalgamation property, 760, 848<br />
amalgamation square, 549<br />
Amalgamation�eorem, 422<br />
ant<strong>is</strong>ymmetric, 40<br />
arity,28,29, 139<br />
associative, 31<br />
asynchronousproduct,651<br />
atom, 347<br />
atom of a lattice, 200<br />
atomic, 200<br />
atomic diagram, 401<br />
atomic struc<strong>tu</strong>re,721<br />
atomictype,765<br />
atomless,200<br />
automorph<strong>is</strong>m, 146<br />
automorph<strong>is</strong>m group,288<br />
averagetype,791<br />
averagetype of an<br />
Ehrenfeucht-Mostowski<br />
functor, 830<br />
average type of an ind<strong>is</strong>cernible<br />
system, 797<br />
average type of anultrafilter, 948<br />
Axiom of Choice, 109, 360<br />
Axiom of Creation, 19, 360<br />
Axiom of Extensionality, 5, 360<br />
logic, algebra & geometry2012-08-08 — ©achim blumensath 1019
In<strong>de</strong>x<br />
Axiom of Infinity, 24, 360<br />
Axiom of Replacement, 133, 360<br />
Axiom of Separation, 10, 360<br />
axiom system, 356<br />
axiomat<strong>is</strong>able, 356<br />
axiomat<strong>is</strong>e, 356<br />
back-and-forth property, 474<br />
back-and-forth system, 474<br />
Baire, property of —, 265<br />
ball, 244<br />
base, closed —, 246<br />
base, open —, 246<br />
basic Horn formula,634<br />
bas<strong>is</strong>, 110, 878, 881<br />
beth, 127<br />
Beth property, 544, 717<br />
bi<strong>de</strong>finable, 751<br />
biinterpretable, 757<br />
bijective, 31<br />
boolean algebra, 184, 357, 392<br />
boolean closed, 392<br />
boolean lattice, 184<br />
boolean logic, 346, 364<br />
boundvariable, 352<br />
boundary,245,656<br />
κ-boun<strong>de</strong>d, 494<br />
boun<strong>de</strong>d in<strong>de</strong>pen<strong>de</strong>ncerelation, 941<br />
boun<strong>de</strong>d lattice, 181<br />
boun<strong>de</strong>d linear or<strong>de</strong>r, 479<br />
boun<strong>de</strong>d logic, 514<br />
boun<strong>de</strong>dness, 941<br />
box,656<br />
branch, 175<br />
branching <strong>de</strong>gree, 177<br />
Cantor d<strong>is</strong>continuum,253, 434<br />
1020<br />
Cantor normal form, 100<br />
Cantor-Bendixson rank, 267, 279<br />
cardinal, 113<br />
cardinal addition, 116<br />
cardinal exponentiation, 116, 127<br />
cardinal multiplication, 116<br />
cardinality, 113<br />
cardinality quantifier, 384<br />
cartesian product,27<br />
categorical,743<br />
category, 152<br />
¯δ-cell, 671<br />
cell <strong>de</strong>composition, 673<br />
Cell Decomposition �eorem, 674<br />
chain, 42<br />
L-chain, 403<br />
chain of struc<strong>tu</strong>res,241<br />
chaintopology,252<br />
chain-boun<strong>de</strong>d formula, 993<br />
Chang’s reduction, 432<br />
character, 105<br />
character<strong>is</strong>tic, 609<br />
character<strong>is</strong>tic of a field, 316<br />
choice function, 106<br />
Choice, Axiom of —, 109, 360<br />
class, 9, 54<br />
clopen set, 243<br />
=-closed, 413<br />
closed base, 246<br />
closed function,248<br />
closed interval,655<br />
closedset, 51, 53, 243<br />
closedsubbase,246<br />
closedunboun<strong>de</strong>d, 558<br />
closedun<strong>de</strong>rrelativ<strong>is</strong>ations, 510<br />
closedun<strong>de</strong>rsubsti<strong>tu</strong>tions, 510<br />
closure operator, 51, 110
closure ordinal, 81<br />
closurespace, 53<br />
closureun<strong>de</strong>rreverseultrapowers,<br />
633<br />
closure,topological —,245<br />
cocone,242<br />
codirected, 232<br />
coefficient, 302<br />
cofinal, 124<br />
cofinality, 124<br />
Coinci<strong>de</strong>nce Lemma, 217<br />
commutative,287<br />
commutativering, 300<br />
commuting diagram, 154<br />
comorph<strong>is</strong>m of logics, 380<br />
compact, 254, 509<br />
compact, countably —, 509<br />
Compactness�eorem, 416, 431<br />
compactness theorem, 618<br />
compatible, 375<br />
complement, 184<br />
complete, 364<br />
κ-complete, 494<br />
complete partial or<strong>de</strong>r, 43, 50, 53<br />
complete type, 427<br />
composition, 30<br />
concatenation, 173<br />
condition of filters, 621<br />
cone, 238<br />
congruencerelation, 161<br />
conjugacy class,293<br />
conjugation,293<br />
conjunction, 347, 391<br />
conjunctive normal form, 369<br />
connected, <strong>de</strong>finably —, 659<br />
consequence, 362, 390, 422<br />
In<strong>de</strong>x<br />
cons<strong>is</strong>tence of filterswith conditions,<br />
621<br />
cons<strong>is</strong>tent, 356<br />
constant, 29, 139<br />
constructedset,735<br />
construction,735<br />
continuous, 46, 134, 248<br />
contradictory formulae, 524<br />
contravariant, 157<br />
coset,288<br />
countable, 110, 115<br />
countably compact, 509<br />
covariant, 157<br />
cover,254<br />
Creation, Axiom of —, 19, 360<br />
cumulative hierarchy, 18<br />
cut,22<br />
<strong>de</strong>ciding a condition,621<br />
<strong>de</strong>finable,712<br />
Aut-<strong>de</strong>finable,712<br />
G-<strong>de</strong>finable,297<br />
<strong>de</strong>finable expansion, 375<br />
<strong>de</strong>finable struc<strong>tu</strong>re,751<br />
<strong>de</strong>finabletype, 466, 943<br />
<strong>de</strong>finablewith parameters, 657<br />
<strong>de</strong>finably connected, 659<br />
<strong>de</strong>fining a set, 349<br />
<strong>de</strong>finition of a type, 465<br />
<strong>de</strong>finitional closed, 712<br />
<strong>de</strong>finitional closure,297,712<br />
<strong>de</strong>gree of apolynomial, 302<br />
<strong>de</strong>nse class, 1001<br />
<strong>de</strong>nse linear or<strong>de</strong>r, 496<br />
κ-<strong>de</strong>nse linear or<strong>de</strong>r, 496<br />
<strong>de</strong>nse or<strong>de</strong>r, 357<br />
<strong>de</strong>nse set, 263<br />
1021
In<strong>de</strong>x<br />
<strong>de</strong>nse sets in directed or<strong>de</strong>rs, 233<br />
<strong>de</strong>pen<strong>de</strong>nce relation, 875<br />
<strong>de</strong>pen<strong>de</strong>nt, 875<br />
<strong>de</strong>pen<strong>de</strong>nt set, 110<br />
<strong>de</strong>rivation, 301<br />
diagram, 236, 241<br />
L-diagram, 401<br />
Diagram Lemma, 401, 530<br />
difference, 11<br />
dimension, 881<br />
dimension function, 882<br />
dimension of a cell,671<br />
dimension of avectorspace, 312<br />
direct limit, 237<br />
direct power, 309<br />
direct product,225<br />
directsum of modules, 309<br />
directed, 232<br />
directed diagram, 236<br />
d<strong>is</strong>continuum,253<br />
d<strong>is</strong>crete linear or<strong>de</strong>r, 479<br />
d<strong>is</strong>crete topology, 244<br />
d<strong>is</strong>integrated matroid, 888<br />
d<strong>is</strong>join<strong>tu</strong>nion, 38<br />
d<strong>is</strong>junction, 347, 391<br />
d<strong>is</strong>junctive normal form, 369<br />
d<strong>is</strong>tributive, 184<br />
dividing, 965<br />
dividing chain, 977<br />
dividing κ-tree, 985<br />
div<strong>is</strong>ible closure,605<br />
div<strong>is</strong>ible group,604<br />
domain,28, 141<br />
dual category, 155<br />
dual lattice, 190<br />
Ehrenfeucht-Fraïssé game, 485, 488<br />
1022<br />
Ehrenfeucht-Mostowski functor, 830,<br />
846<br />
Ehrenfeucht-Mostowski mo<strong>de</strong>l, 830<br />
element of a set, 5<br />
elementary diagram, 401<br />
elementary embedding, 395, 400<br />
elementary extension, 400<br />
elementary map, 395<br />
elementarysubstruc<strong>tu</strong>re, 400<br />
eliminationset, 588<br />
embedding, 44, 146, 237, 396<br />
∆-embedding, 395<br />
K-embedding, 839<br />
elementary —, 395<br />
embedding of a tree into a lattice, 206<br />
embedding of logics, 380<br />
embedding of permutation groups,<br />
752<br />
embedding, elementary —, 400<br />
endomorph<strong>is</strong>mring, 307<br />
entailment, 362, 390<br />
epimorph<strong>is</strong>m, 155<br />
equivalence class, 54<br />
equivalence of categories, 158<br />
equivalencerelation, 54, 357<br />
L-equivalent, 364<br />
α-equivalent, 473, 488<br />
equivalent formulae, 362<br />
Erdős-Rado theorem, 776<br />
Eukli<strong>de</strong>an norm,243<br />
even, 770<br />
exchange property, 110<br />
ex<strong>is</strong>tential, 396<br />
ex<strong>is</strong>tential closure, 597<br />
ex<strong>is</strong>tentialquantifier, 347<br />
ex<strong>is</strong>tentially closed, 597<br />
expansion, 145, 842
expansion, <strong>de</strong>finable —, 375<br />
explicit <strong>de</strong>finition, 544<br />
exponentiation of cardinals, 116, 127<br />
exponentiation of ordinals, 89<br />
extension, 142, 941<br />
∆-extension, 400<br />
extension axiom, 765<br />
extension of fields, 317<br />
extension, elementary —, 400<br />
Extensionality, Axiom of —, 5, 360<br />
factor<strong>is</strong>ation, 165<br />
Factor<strong>is</strong>ation Lemma, 148<br />
family, 37<br />
field, 300, 359, 400, 608<br />
field extension, 317<br />
field of a relation, 29<br />
field of fractions, 314<br />
field, real —, 329<br />
field, real closed —, 332<br />
filter, 189, 193, 430<br />
final segment, 41<br />
κ-finitary set of partial <strong>is</strong>omorph<strong>is</strong>ms,<br />
494<br />
finite, 115<br />
finite character, 51, 105, 929<br />
finite character, strong —, 941<br />
finite intersection property, 197<br />
finite occurrenceproperty, 509<br />
finite, being — over aset,674<br />
finitely axiomat<strong>is</strong>able, 356<br />
finitely branching, 177<br />
finitely generated, 144<br />
finitely sat<strong>is</strong>fiable type, 948<br />
first-or<strong>de</strong>r interpretation, 348, 377<br />
first-or<strong>de</strong>r logic, 347<br />
fixed point, 48, 81, 134, 553<br />
fixed-point induction,77<br />
fixed-pointrank, 572<br />
follow, 362<br />
forcing, 621<br />
forgetful functor, 157,220<br />
forking chain, 977<br />
forkingrelation, 941<br />
formalpowerseries, 301<br />
formula, 346<br />
foundationrank, 178<br />
foun<strong>de</strong>d, 13<br />
Fraïssé limit, 764<br />
free algebra, 218<br />
free extension of types, 942<br />
√<br />
-free extension of types, 942<br />
free mo<strong>de</strong>l, 639<br />
free struc<strong>tu</strong>res,648<br />
freevariables,217, 352<br />
function,29<br />
functional,29, 139<br />
functor, 157<br />
In<strong>de</strong>x<br />
Gaifman graph, 501<br />
Gaifman,�eorem of —, 507<br />
Galo<strong>is</strong>sa<strong>tu</strong>ratedstruc<strong>tu</strong>re, 855<br />
Galo<strong>is</strong>stable, 855<br />
Galo<strong>is</strong>type, 841<br />
game,79<br />
general<strong>is</strong>edproduct,650<br />
κ-generated, 239, 759, 809<br />
generatedsubstruc<strong>tu</strong>re, 144<br />
generated, finitely —, 144<br />
generating, 41<br />
generating an i<strong>de</strong>al, 304<br />
generator, 144, 639<br />
geometric dimension function, 882<br />
geometric in<strong>de</strong>pen<strong>de</strong>ncerelation, 929<br />
1023
In<strong>de</strong>x<br />
geometry, 881<br />
graduatedtheory, 596,681<br />
graph, 39<br />
greatest element, 41<br />
greatest fixed point, 553<br />
greatest lower bound, 42<br />
greatestpartial fixedpoint, 554<br />
group, 34, 287, 358<br />
group action,292<br />
group, or<strong>de</strong>red —,604<br />
guard, 349<br />
Hanf number, 514, 534, 847<br />
Hanf’s �eorem, 502<br />
Hausdorffspace,253<br />
height, 176<br />
height in a lattice, 200<br />
Henkin property, 724<br />
Henkin set, 724<br />
Herbrand mo<strong>de</strong>l, 412, 724<br />
hereditary, 12<br />
κ-hereditary, 760, 809<br />
hereditary finite, 7<br />
Hintikka formula, 482, 483<br />
Hintikka set, 414, 724, 725<br />
h<strong>is</strong>tory, 15<br />
homeomorph<strong>is</strong>m, 248<br />
homogeneous,685,773<br />
≈-homogeneous,779<br />
κ-homogeneous, 500,685<br />
homogeneous matroid, 888<br />
homomorphic image, 147,643<br />
homomorph<strong>is</strong>m, 146, 396<br />
Homomorph<strong>is</strong>m �eorem, 168<br />
homotopic interpretations, 756<br />
Horn formula,634<br />
1024<br />
i<strong>de</strong>al, 189, 193, 303<br />
i<strong>de</strong>ntity, 153<br />
image, 31<br />
implication, 349<br />
implicit <strong>de</strong>finition, 544<br />
inclusion morph<strong>is</strong>m, 393<br />
incons<strong>is</strong>tent, 356<br />
k-incons<strong>is</strong>tent, 965<br />
increasing, 44<br />
in<strong>de</strong>pen<strong>de</strong>nce property, 799<br />
in<strong>de</strong>pen<strong>de</strong>nce relation, 928<br />
in<strong>de</strong>pen<strong>de</strong>nce relation of a matroid,<br />
927<br />
In<strong>de</strong>pen<strong>de</strong>nce �eorem, 992<br />
in<strong>de</strong>pen<strong>de</strong>nt, 875<br />
in<strong>de</strong>pen<strong>de</strong>nt set, 110, 881<br />
in<strong>de</strong>x of asubgroup,288<br />
ind<strong>is</strong>cerniblesequence,789<br />
ind<strong>is</strong>cerniblesystem,796<br />
inducedsubstruc<strong>tu</strong>re, 142<br />
inductive,77<br />
inductive fixedpoint, 81, 553, 554<br />
inductively or<strong>de</strong>red, 81, 105<br />
infimum, 42, 181<br />
infinitary first-or<strong>de</strong>r logic, 347<br />
infinitary second-or<strong>de</strong>r logic, 385<br />
infinite, 115<br />
Infinity, Axiom of —, 24, 360<br />
inflationary, 81<br />
inflationary fixed-point logic, 562<br />
initial object, 156<br />
initial segment, 41<br />
injective, 31<br />
κ-injective struc<strong>tu</strong>re, 852<br />
innervertex, 175<br />
insertion, 39<br />
inspired by, 797
integral domain, 314, 611<br />
interior, 245, 656<br />
interpolant, 549<br />
interpolation closure, 545<br />
interpolationproperty, 543<br />
∆-interpolation property, 543<br />
interpretation, 346, 348, 377<br />
intersection, 11<br />
interval, 655<br />
invariance, 941<br />
invariant class, 1001<br />
invariant type, 956<br />
inverse, 30, 155<br />
inverse diagram, 241<br />
inverse limit, 241<br />
inverse reduct, 819<br />
irreduciblepolynomial, 320<br />
irreflexive, 40<br />
<strong>is</strong>olated point, 266<br />
<strong>is</strong>olated type, 721, 943<br />
<strong>is</strong>omorphic, 44<br />
α-<strong>is</strong>omorphic, 477, 488<br />
<strong>is</strong>omorphic copy, 643<br />
<strong>is</strong>omorph<strong>is</strong>m, 44, 146, 155, 158, 396<br />
<strong>is</strong>omorph<strong>is</strong>m, partial —, 473<br />
joint embedding property, 760, 849<br />
Jónsson class, 849<br />
Karp property, 509<br />
kernel, 148<br />
kernel of a ring homomorph<strong>is</strong>m, 305<br />
label, 213<br />
largesubsets,720<br />
Lascarstrongtype, 993<br />
lattice, 181, 357, 392<br />
leaf, 175<br />
least element, 41<br />
least fixed point, 553<br />
least fixed-point logic, 561<br />
least partial fixed point, 554<br />
leas<strong>tu</strong>pper bound, 42<br />
le� boun<strong>de</strong>d, 941<br />
le� i<strong>de</strong>al, 303<br />
le� reflexivity, 928<br />
le� restriction, 31<br />
le� transitivity, 928<br />
le�-narrow, 57<br />
length, 173<br />
level, 176<br />
level embedding function,779<br />
levels of a<strong>tu</strong>ple,779<br />
lexicographic or<strong>de</strong>r, 173, 869<br />
li�ing functions, 551<br />
limit, 59<br />
limitstage, 19<br />
Lin<strong>de</strong>nbaum algebra, 390<br />
Lin<strong>de</strong>nbaum functor, 390<br />
Lindström quantifier, 384<br />
linear in<strong>de</strong>pen<strong>de</strong>nce, 309<br />
linear matroid, 881<br />
linear or<strong>de</strong>r, 40<br />
linear representation, 585<br />
literal, 347<br />
local, 504<br />
local enumeration,670<br />
κ-local functor, 809<br />
local<strong>is</strong>ation morph<strong>is</strong>m, 393<br />
local<strong>is</strong>ation of a logic, 393<br />
locally compact, 254<br />
locally finite matroid, 888<br />
locally modular matroid, 888<br />
logic, 346<br />
In<strong>de</strong>x<br />
1025
In<strong>de</strong>x<br />
logical system, 387<br />
Łoś’ theorem, 615<br />
Łoś-Tarski �eorem, 584<br />
Löwenheim number, 514, 534, 537, 839<br />
Löwenheim-Skolem property, 509<br />
Löwenheim-Skolem-Tarski �eorem,<br />
421<br />
lower bound, 42<br />
lower fixed-point induction, 554<br />
map,29<br />
∆-map, 395<br />
map, elementary —, 395<br />
mapping, 29<br />
matroid, 880<br />
maximal element, 41<br />
maximal i<strong>de</strong>al, 314<br />
maximal i<strong>de</strong>al/filter, 189<br />
meagre, 264<br />
membership relation, 5<br />
minimal, 13, 57<br />
minimal element, 41<br />
minimal polynomial, 323<br />
minimal rank and <strong>de</strong>gree, 208<br />
minimal set, 893<br />
mo<strong>de</strong>l, 346<br />
mo<strong>de</strong>l companion, 597<br />
mo<strong>de</strong>l of a presentation, 639<br />
mo<strong>de</strong>l-complete, 597<br />
κ-mo<strong>de</strong>l-homogeneousstruc<strong>tu</strong>re, 852<br />
modular, 184<br />
modular lattice,200<br />
modular law,202,203<br />
modular matroid, 888<br />
modularity, 938<br />
module, 306<br />
monadic second-or<strong>de</strong>r logic, 385<br />
1026<br />
monoid, 31, 175, 287<br />
monomorph<strong>is</strong>m, 155<br />
monotone, 656<br />
monotonicity, 928<br />
monster mo<strong>de</strong>l, 720<br />
Morley <strong>de</strong>gree, 920<br />
Morley rank, 917<br />
Morley sequence, 958<br />
Morley-free extension of atype, 920<br />
morph<strong>is</strong>m, 152<br />
morph<strong>is</strong>m of logics, 380<br />
morph<strong>is</strong>m of matroids, 888<br />
morph<strong>is</strong>m of permutation groups,751<br />
multiplication of cardinals, 116<br />
multiplication of ordinals, 89<br />
na<strong>tu</strong>ral <strong>is</strong>omorph<strong>is</strong>m, 158<br />
na<strong>tu</strong>raltransformation, 158<br />
negation, 347, 391<br />
negation normal form, 371<br />
negative occurrence, 561<br />
neighbourhood,243<br />
neutral element, 31<br />
no<strong>de</strong>, 175<br />
normalsubgroup,289<br />
normality, 928<br />
nowhere <strong>de</strong>nse,264<br />
o-minimal,658, 802<br />
object, 152<br />
occurrence number, 514<br />
oligomorphic,292,743<br />
omitting a type, 428<br />
omitting types, 432<br />
open base, 246<br />
open cover, 254<br />
open <strong>de</strong>nse or<strong>de</strong>r, 357
open interval, 655<br />
open set, 243<br />
opensubbase,247<br />
orbit,292<br />
or<strong>de</strong>r, 356<br />
or<strong>de</strong>r property, 463<br />
or<strong>de</strong>r topology, 251, 656<br />
or<strong>de</strong>r type, 64, 789<br />
or<strong>de</strong>rable ring, 329<br />
or<strong>de</strong>red group,604<br />
or<strong>de</strong>redpair,27<br />
or<strong>de</strong>redring, 329<br />
ordinal, 64<br />
ordinal addition, 89<br />
ordinal exponentiation, 89<br />
ordinal multiplication, 89<br />
ordinal,von Neumann —,69<br />
package,737<br />
pair, 27<br />
parameter-<strong>de</strong>finable, 657<br />
partial fixed point, 554<br />
partial fixed-point logic, 562<br />
partial function,29<br />
partial <strong>is</strong>omorph<strong>is</strong>m, 473<br />
partial <strong>is</strong>omorph<strong>is</strong>m modulo a filter,<br />
626<br />
partial or<strong>de</strong>r, 40, 356<br />
partial or<strong>de</strong>r, strict —, 40<br />
partition, 55, 204<br />
partition <strong>de</strong>gree, 208<br />
partition rank, 204<br />
partitioning a relation, 674<br />
path, 175<br />
Peano Axioms, 386<br />
pinning down, 514<br />
point, 243<br />
In<strong>de</strong>x<br />
polynomial, 302<br />
polynomial function, 319<br />
polynomial ring, 302<br />
positive ex<strong>is</strong>tential, 396<br />
positive occurrence, 561<br />
positiveprimitive,634<br />
power set, 21<br />
predicate, 28<br />
predicate logic, 346<br />
prefix, 173<br />
prefix or<strong>de</strong>r, 173<br />
preforking relation, 941<br />
prelattice, 193<br />
prenex normal form, 371<br />
preor<strong>de</strong>r, 192, 390<br />
presentation, 639<br />
preservation by a function, 395<br />
preservation in products,634<br />
preservation insubstruc<strong>tu</strong>res, 398<br />
preservation inunions of chains, 399<br />
preserving fixed points, 551<br />
prime field, 316<br />
prime i<strong>de</strong>al, 194, 306<br />
prime mo<strong>de</strong>l, 734<br />
prime mo<strong>de</strong>l, algebraic, 592<br />
primitive formula, 597<br />
principal i<strong>de</strong>al/filter, 189<br />
Principle ofTransfiniteRecursion,75,<br />
133<br />
product,27, 37, 643<br />
product of categories, 156<br />
product of linear or<strong>de</strong>rs, 86<br />
producttopology,259<br />
product, direct —,225<br />
product, general<strong>is</strong>ed —,650<br />
product,reduced —,228<br />
product,subdirect —,226<br />
1027
In<strong>de</strong>x<br />
projection, 37, 532<br />
projective class, 532<br />
projective geometry, 887<br />
projectively reducible, 533<br />
projectively κ-sa<strong>tu</strong>rated,702<br />
proper, 189<br />
property of Baire,265<br />
pseudo-elementary, 533<br />
pseudo-sa<strong>tu</strong>rated,705<br />
quantifier elimination, 588,610<br />
quantifierrank, 355<br />
quantifier-free, 355<br />
quantifier-free formula, 396<br />
quasivariety,643<br />
quotient, 165<br />
Rado graph,766<br />
Ramsey’stheorem,774<br />
random graph,766<br />
randomtheory,766<br />
range,29<br />
rank,73, 178<br />
∆-rank, 917<br />
rank, foundation –, 178<br />
real closed field, 332, 609<br />
real closure of a field, 332<br />
real field, 329<br />
real<strong>is</strong>ing a type, 428<br />
reducedproduct,228,643<br />
reduct, 145<br />
µ-reduct, 223<br />
reflexive, 40<br />
regular, 125<br />
regular filter,617<br />
regular logic, 510<br />
relation,28<br />
1028<br />
relational, 139<br />
relationalvariant of a struc<strong>tu</strong>re, 821<br />
relativ<strong>is</strong>ation, 376, 510<br />
relativ<strong>is</strong>ed projective class, 537<br />
relativ<strong>is</strong>ed projectively reducible, 537<br />
relativ<strong>is</strong>ed quantifiers, 349<br />
relativ<strong>is</strong>ed reduct, 537<br />
Replacement, Axiom of —, 133, 360<br />
replica functor, 823<br />
restriction, 30<br />
restriction of a filter, 228<br />
restriction of a Galo<strong>is</strong> type, 859<br />
restriction of a logic, 393<br />
restriction of a type, 459<br />
retract of a logic, 446<br />
retraction of logics, 446<br />
reverseultrapower,633<br />
right base monotonicity, 929<br />
right boun<strong>de</strong>d, 941<br />
ring, 300, 359<br />
ring, or<strong>de</strong>rable —, 329<br />
ring, or<strong>de</strong>red —, 329<br />
root, 175<br />
root of a polynomial, 319<br />
Ryll-Nardzewski �eorem, 743<br />
sat<strong>is</strong>faction, 346<br />
sat<strong>is</strong>faction relation, 346, 348<br />
sat<strong>is</strong>fiable, 356<br />
sa<strong>tu</strong>rated,691<br />
κ-sa<strong>tu</strong>rated, 565,691<br />
κ-sa<strong>tu</strong>rated,projectively —,702<br />
Scott height, 483<br />
Scott sentence, 483<br />
second-or<strong>de</strong>r logic, 385<br />
segment, 41<br />
semantics functor, 387
semantics of first-or<strong>de</strong>r logic, 348<br />
semi-strict homomorph<strong>is</strong>m, 146<br />
semilattice, 181<br />
sentence, 352<br />
separated formulae, 524<br />
Separation, Axiom of —, 10, 360<br />
sequence, 37<br />
signa<strong>tu</strong>re, 139, 141, 221, 222<br />
simple struc<strong>tu</strong>re, 315<br />
simple theory, 975<br />
simply closed, 592<br />
singular, 125<br />
skew embedding, 786<br />
skew field, 300<br />
Skolem axiom, 406<br />
Skolem expansion, 843<br />
Skolem function, 406<br />
Skolemtheory, 406<br />
Skolem<strong>is</strong>ation, 406<br />
smallsubsets,720<br />
sort, 141<br />
spanning, 878<br />
special mo<strong>de</strong>l,705<br />
specification of a dividing chain, 977<br />
specification of a dividing κ-tree, 985<br />
specification of a forking chain, 977<br />
spectrum,272, 431, 434<br />
spectrum of aring, 306<br />
spine, 825<br />
stabil<strong>is</strong>er, 293<br />
κ-stable, 463<br />
stage, 15, 77<br />
stage compar<strong>is</strong>on relation, 572<br />
Stone space, 276, 431, 434<br />
strict homomorph<strong>is</strong>m, 146<br />
strict Horn formula,634<br />
strict∆-map, 395<br />
strict or<strong>de</strong>r property, 804<br />
strict partial or<strong>de</strong>r, 40<br />
strictly increasing, 44<br />
strictly monotone, 656<br />
strong γ-chain, 861<br />
strong γ-limit, 861<br />
strong finite character, 941<br />
strong limit cardinal, 705<br />
strong right boun<strong>de</strong>dness, 929<br />
strongly homogeneous,685<br />
strongly κ-homogeneous,685<br />
strongly local functor, 825<br />
strongly minimalset, 893<br />
strongly minimal theory, 900, 990<br />
struc<strong>tu</strong>re, 139, 141, 223<br />
subbase, closed —,246<br />
subbase, open —,247<br />
subcategory, 156<br />
subcover,254<br />
subdirectproduct,226<br />
subdirectly irreducible,226<br />
subfield, 316<br />
subformula, 352<br />
subset, 5<br />
subspacetopology,248<br />
subspace, closure —,248<br />
substi<strong>tu</strong>tion,220, 367, 510<br />
substruc<strong>tu</strong>re, 142,643, 759, 809<br />
∆-substruc<strong>tu</strong>re, 400<br />
K-substruc<strong>tu</strong>re, 840<br />
substruc<strong>tu</strong>re, elementary —, 400<br />
substruc<strong>tu</strong>re, generated —, 144<br />
substruc<strong>tu</strong>re, induced —, 142<br />
subterm,214<br />
subtree, 176<br />
successor, 59, 175<br />
successorstage, 19<br />
In<strong>de</strong>x<br />
1029
In<strong>de</strong>x<br />
sum of linear or<strong>de</strong>rs, 85<br />
superset, 5<br />
supremum, 42, 181<br />
surjective, 31<br />
symbol, 139<br />
symmetric, 40<br />
symmetric group,291<br />
symmetric in<strong>de</strong>pen<strong>de</strong>ncerelation,<br />
929<br />
syntax functor, 387<br />
Tarskiunionproperty, 510<br />
tautology, 356<br />
term, 213<br />
term algebra, 218<br />
term domain, 213<br />
term,value of a —,217<br />
term-reduced, 368<br />
terminal object, 156<br />
L-theory, 363<br />
theory of a functor, 815<br />
topological closure,245,656<br />
topological closure operator, 51,245<br />
topological group,298<br />
topologicalspace,243<br />
topology, 243<br />
topology of the type space, 433<br />
torsion element, 603<br />
torsion-free, 603<br />
total or<strong>de</strong>r, 40<br />
totally d<strong>is</strong>connected, 253<br />
totally ind<strong>is</strong>cernible sequence,790<br />
transcen<strong>de</strong>nce bas<strong>is</strong>, 322<br />
transcen<strong>de</strong>nce <strong>de</strong>gree, 322<br />
transcen<strong>de</strong>ntal elements, 322<br />
transcen<strong>de</strong>ntal field extensions, 322<br />
transfinite recursion,75, 133<br />
1030<br />
transitive, 12, 40<br />
transitive action, 292<br />
transitive closure, 55<br />
transitive <strong>de</strong>pen<strong>de</strong>ncerelation, 875<br />
transitivity, le� —, 928<br />
tree, 175<br />
treeproperty, 984<br />
tree-ind<strong>is</strong>cernible,797<br />
trivial filter, 189<br />
trivial i<strong>de</strong>al, 189<br />
trivialtopology,244<br />
<strong>tu</strong>ple,28<br />
Tychonoff,�eorem of —,261<br />
type, 459<br />
L-type, 427<br />
Ξ-type,702<br />
α-type, 428<br />
¯s-type, 428<br />
type of a function, 141<br />
type of arelation, 141<br />
typespace, 433<br />
type topology, 433<br />
type, average —, 791<br />
type, average — of an ind<strong>is</strong>cernible<br />
system, 797<br />
type, complete —, 427<br />
type, Lascar strong —, 993<br />
types of <strong>de</strong>nse linear or<strong>de</strong>rs, 429<br />
ultrafilter, 194, 430<br />
ultrahomogeneous,761<br />
ultrapower,229<br />
ultraproduct,229,695<br />
unboun<strong>de</strong>d class, 847<br />
uncountable, 115<br />
uniform dividing chain, 978<br />
uniform dividing κ-tree, 985
uniform forking chain, 978<br />
uniformly finite, being — over aset,<br />
674<br />
union,21<br />
union of a chain,241, 403, 586<br />
unit of aring, 314<br />
universal, 396<br />
κ-universal, 691,760<br />
universalquantifier, 347<br />
universalstruc<strong>tu</strong>re, 852<br />
universe, 139, 141<br />
unsat<strong>is</strong>fiable, 356<br />
unstable, 463<br />
upper bound, 42<br />
upper fixed-point induction, 554<br />
valid, 356<br />
value of aterm,217<br />
variable,222<br />
variablesymbols, 347<br />
variables, free —, 217, 352<br />
variety, 643<br />
Vaughtianpair, 901<br />
vectorspace, 306<br />
vertex, 175<br />
von Neumann ordinal,69<br />
In<strong>de</strong>x<br />
weak γ-chain, 861<br />
weak γ-limit, 861<br />
weak homomorphic image, 147,643<br />
weakly regular logic, 510<br />
well-foun<strong>de</strong>d, 13, 57, 81, 109<br />
well-or<strong>de</strong>r, 57, 109, 133, 494<br />
well-or<strong>de</strong>ring number, 514, 534<br />
well-or<strong>de</strong>ring quantifier, 383, 384<br />
winning strategy, 486<br />
word construction, 816, 821<br />
Zar<strong>is</strong>ki logic, 345<br />
Zar<strong>is</strong>ki topology, 244<br />
zero-dimensional, 253<br />
zero-div<strong>is</strong>or, 314<br />
Zero-One Law, 770<br />
ZFC, 359<br />
Zorn’s Lemma, 110<br />
1031
1032<br />
�e Roman and Frak<strong>tu</strong>r alphabets<br />
A a A a N n N n<br />
B b B b O o O o<br />
C c C $ P p P p<br />
D d D d Q q Q q<br />
E e E e R r R r<br />
F f F f S s S s +<br />
G g G g T t T t<br />
H h H h U u U u<br />
I i J i V v V v<br />
J j J j W w W w<br />
K k K k X x X x<br />
L l L l Y y Y y<br />
M m M m Z z Z z<br />
�e Greek alphabet<br />
A α alpha N ν nu<br />
B β beta Ξ ξ xi<br />
Γ γ gamma O o omicron<br />
∆ δ <strong>de</strong>lta Π π pi<br />
E ε epsilon P ρ rho<br />
Z ζ zeta Σ σ sigma<br />
H η eta T τ tau<br />
Θ ϑ theta Υ υ upsilon<br />
I ι iota Φ ϕ phi<br />
K κ kappa X χ chi<br />
Λ λ lambda Ψ ψ psi<br />
M µ mu Ω ω omega