Introduction to mathematical thinking

Alexandru Buium
Department of Mathematics and Statistics, University of New Mex-
ico, Albuquerque, NM 87131, USA
E-mail address: [email protected]
 This is a (drastically) simplified version of my book:
Mathematics: a minimal introduction, CRC press, 2013.
For a more complete treatment of the topics one may refer to the book. However
one should be aware of the fact that there are some key conceptual differences
between the present text and the book, especially when it comes to the material
that pertains to logic (e.g., to witnesses, quantifier axioms, and the structure of
theories).
 Contents

Part 1. Logic 5

Chapter 1. Languages 7

Chapter 2. Metalanguage 15

Chapter 3. Syntax 21

Chapter 4. Tautologies 25

Chapter 5. Proofs 31

Chapter 6. Theories 39

Chapter 7. ZFC 49

Part 2. Set theory 53

Chapter 8. Sets 55

Chapter 9. Maps 59

Chapter 10. Relations 63

Chapter 11. Operations 69

Part 3. The discrete 75

Chapter 12. Integers 77

Chapter 13. Induction 81

Chapter 14. Rationals 85

Chapter 15. Combinatorics 87

Chapter 16. Sequences 89

Part 4. The continuum 93

Chapter 17. Reals 95

Chapter 18. Topology 97

Chapter 19. Imaginaries 101

3
4 CONTENTS

Part 5. Algebra 103

Chapter 20. Arithmetic 105

Chapter 21. Groups 109
Chapter 22. Order 113

Chapter 23. Vectors 115

Chapter 24. Matrices 117
Chapter 25. Determinants 121
Chapter 26. Polynomials 125

Chapter 27. Congruences 129

Part 6. Geometry 133

Chapter 28. Lines 135

Chapter 29. Conics 139

Chapter 30. Cubics 141

Part 7. Analysis 145

Chapter 31. Limits 147

Chapter 32. Series 151

Chapter 33. Trigonometry 155
Chapter 34. Calculus 157
Part 1

Logic
 CHAPTER 1

Languages

Mathematics is a theory called set theory. Theories are (growing) sequences of

sentences. The formation of sentences and theories is governed by (general) logic
(not to be confused with mathematical logic which is part of mathematics). Logic
starts with the analysis/construction of language.
Example 1.1. (Logical analysis of language) We will introduce here two lan-
guages, English and Formal, and we will analyze their interconnections.
Let us start with a discussion of English. The English language is the collection
LEng of all English words (plus separators such as parentheses, commas, etc.). We
treat words as individual symbols (and ignore the fact that they are made out of
letters). Sometimes we admit as symbols certain groups of words. One can use
words to create strings of words such as
0)“for all not Socrates man if”
The above string is considered “syntactically incorrect.” The sentences in the
English language are the strings of symbols that are “syntactically correct” (in
a sense to be made precise later). Here are some examples of sentences in this
language:

1) “Socrates is a man”
2) “Caesar killed Brutus”
3) “The killer of Caesar is Brutus”
4) “Brutus killed Caesar and Socrates is a man”
5) “Brutus is not a man or Caesar is a killer”
6) “If Brutus killed Caesar then Brutus is a killer”
7) “Brutus did not kill Caesar”
8) “A man killed Caesar”
9) “If a man killed another man then the first man is a killer”
10) “A man is a killer if and only if that man killed another man”

In order to separate sentences from a surrounding text we put them between

quotation marks (and sometimes we write them in italics). So quotation marks do
not belong to the language but rather they lie outside the language; they belong
to metalanguage, as we shall explain. Checking syntax presupposes a partitioning
of LEng into various categories of words; no word should appear in principle in two
different categories, but this requirement is often violated in practice (which may
lead to different readings of the same text). Here are the categories:
• variables: “something,...”
• constants: “Socrates, Brutus, Caesar,...”
• functional symbols: “the killer of,...”
7
8 1. LANGUAGES

• predicates: “is a man, is a killer, killed,...”

• connectives: “and, or, not, if...then, if and only if”
• quantifiers: “for all, there exists”
• equality: “is, equals”
• separators: parentheses “(,)” and comma “,”
The above categories are referred to as logical categories. (They are quite
different from, although related to, the grammatical categories of nouns, verbs, etc.
In general objects are named by constants or variables. (So constants and variables
roughly correspond to proper nouns.) Constants are names for specific objects while
variables are names for non-specific (generic) objects. The article “the” generally
indicates a constant; the article “a” generally indicates that a quantifier is implicitly
assumed. Predicates say/affirm something about one or several objects; if they
say/affirm something about one, two, three objects, etc., they are unary, binary,
ternary, etc. (So roughly unary predicates correspond to intransitive verbs; binary
predicates correspond to transitive verbs.) “Killed” is a binary predicate; “is a
killer” is a unary predicate. Functional symbols have objects as arguments but do
not say/affirm anything about them; all they do is refer to (or name, or specify,
or point towards) something that could itself be an object. (Functional symbols
are sometimes referred to as functional predicates but we will not refer to them as
predicates here; this avoids confusion with predicates.) Again they can be unary,
binary, ternary, etc., depending on the number of arguments. “The father of” is
a unary functional symbol. “The son of ... and ...” is a binary functional symbol
(where the two arguments stand for the mather and the father and we assume for
simplicity that any two parents have a unique son.) Connectives connect/combine
sentences into longer sentences; they can be unary (if they are added to one sentence
changing it into another sentence, binary if they combine two sentences into one
longer sentence, ternary, etc.). Quantifiers specify quantity and are always followed
by variables. Separators separate various parts of the text from various other parts.
In order to analyze a sentence using the logical categories above one first looks
for the connectives and one splits the sentence into simpler sentences; alternatively
sentences may start with quantifiers followed by variables followed by simpler sen-
tences. In any case, once one identifies simpler sentences, one proceeds by iden-
tifying, in each of them, the constants, variables, and functional symbols applied
to them (these are the objects that one is talking about), and finally one identifies
the functional symbols (which say something about the objects). The above type
of analysis (called logical analysis) is quite different from the grammatical analysis
based on the grammatical categories of nouns, verbs, etc.
A concise way of understanding the logical analysis of English sentences as
above is to create another language LF or (let us call it Formal) consisting of the
following symbols:

• variables: “x, y, ...”

• constants: “S, B, C, ...”
• functional symbols: “↓, ...”
• predicates: “m, k, †”
• connectives: “∧, ∨, ¬, →, ↔”
• quantifiers: “∀, ∃”
• equality: “=”
 1. LANGUAGES 9

• separators: parentheses “(, )” and comma “,”

Furthermore let us introduce a rule (called translation) that attaches to each symbol
in Formal a symbol in English as follows:

“x, y” are translated as “something, a thing, an entity,...”

“S, B, C” are translated as “Socrates, Brutus, Caesar”
“↓, ...” are translated as “the killer of,...”
“m, k, †” are translated as “is a man, is a killer, killed,...”
“∧, ∨, ¬, →, ↔” are translated as “and, or, not, if...then, if and only if”
“∀, ∃” are translated as “for all, there exists”
“=” is translated as “is” or “is a equal to”

Then the English sentences 1-10 are translations of the following Formal sentences;
equivalently the following Formal sentences are translations (called formalizations)
of the corresponding English sentences:

1’) “m(S)”
2’) “C † B”
3’) “↓ (C) = B”
4’) “(B † C) ∧ m(S)”
5’) “(¬(m(B))) ∨ k(C)”
6’) “(B † C) → (k(B))”
7’) “¬(B † C)”
8’) “∃x(x † C)”
9’) “∀x((m(x) ∧ (∃y(m(y) ∧ ¬(x = y) ∧ (x † y))) → k(x))”
10’) “∀x(k(x) ↔ (m(x) ∧ (∃y(m(y) ∧ ¬(x = y) ∧ (x † y))))”

In the above formalizations one first replaces 8), 9), 10) by

8”) “There exists something such that that something is a man and that some-
thing killed Caesar”
9”) “For any x if x is man and there exists a y such that y is a man, y is
different from x and x killed y then x is a killer”
10”) “For any x one has that x is a killer if and only if x is a man and there
exists y such that y is a man, x is not y, and x killed y”
Note that the word “exists” which has the form of a predicate is considered
instead as part of a quantifier. Sentences like “philosophers exist” and “philosophers
are human” have a totally different logical structure. Indeed “philosophers exist”
should be read as “there exists something such that that something is a philosopher”
while “philosophers are human” should be read as “for all x if x is a philosopher
then s is a human.” The fact that “exist” should not be viewed as a predicate
was recognized already by Kant, in particular in his criticism of the “ontological
argument.” On the other hand the verb to be (as in “is, are,...”) can be: i) part
of a predicate (as in “is a man”), ii) part of equality (as in “is equal, is the same
as”), iii) part of a quantifier (as is “there is”, an equivalent translation of ∃).
All of our discussion of English and Formal above is itself expressed in yet
another language which needs to be distinguished from English itself and which we
shall call Metalanguage. We will discuss Metalanguage in detail in the next chapter
(where some languages will be declared object languages and others will be declared
10 1. LANGUAGES

metalanguages). The very latter sentence is written in Metalanguage; and indeed

the whole course is written in Metalanguage.
Remark 1.2. (Naming) It is useful to give names to sentences. For instance
if we want to give the name P to the English sentence “Socrates is a man” we can
write the following sentence in Metalanguage:
P equals “Socrates is a man.”
So neither P nor the word equals nor the quotation marks belong to English; and
“Socrates is a man” will be viewed in Metalanguage as one single symbol. One can
give various different names to the same sentence. In a similar way one can give
names to sentences in Formal by writing a sentence in Metalanguage:
Q equals “m(s).”
Alternatively one can write
Q = “m(S)”
where = is equality in metalanguage.
Remark 1.3. (Syntax/semantics/reference/inference/truth of languages)
Syntax deals with rules of formation of “correct” sentences. We will examine
these rules in detail in a subsequent chapter.
Semantics deals with meaning. For us here the meaning of a sentence will be
defined (in a rather minimalist way) as the collection of its available “translations”
(where the latter will be understood in a rather general sense). For a person who
does not understand English establishing meaning of sentences such as 1,...,5 above
requires relating these sentences to sentences in another language (e.g., translating
them into French, German, a “picture language,” sign language, Braille, etc., or
using a correspondence to “deeper” languages, as in the work of Chomsky); the
more translations available the more definite the meaning. On the other hand the
meaning of 1’,...,10’ is taken to be given by translating them into the sentences
1,...,10 (where one assumes one knows English).
Reference (or universe of discourse) is “what sentences are about.” Words in
English may refer to the physical or imaginary worlds (including symbols in lan-
guages which are also viewed as physical entities); e.g., the English word “Socrates”
refers to the physical “man Socrates”; the word “Hamlet” refers to something in
the imaginary world. Metalanguage, on the other hand, refers to other languages
such as English or Formal; so the universe of discourse of Metalanguage consists
of other languages; such a reference will be called linguistic reference. Reference
to things other than languages will be called non-linguistic reference. Sentences in
Formal can be attached a reference once they are translated into English, say; then
they have the same reference as their translations.
Inference is a process by which we accept declarative sentences that have a
meaning based on other declarative sentences that are already accepted; see the
comments below on declarative sentences. There is a whole array of processes that
may qualify as inference from belief to mechanical proof.
We could also ask if the sentences 1,...,10, 1’,...,10’ are “true” or “false.” We will
not define truth/falsehood for sentences in any of our languages. Indeed a theory
of truth would complicate our analysis beyond what we are ready to undertake; on
the other hand dropping the concepts of truth and falsehood will not affect, as we
shall see, our ability to develop mathematics.
 1. LANGUAGES 11

Remark 1.4. (Definitions) A language may be enlarged by definitions. More

precisely one can add new predicates or constants to a language by, at the same time,
recording certain sentences, called definitions. As an example for the introduction
of a new predicate in English we can add to English the predicate is an astrochicken
by recording the following sentence:
Definition 1.5. Something is an astrochicken if and only if it is a chicken and
also a space ship.
Here are alternative ways to give this definition:
Definition 1.6. An astrochicken is something which is a chicken and also a
space ship.
Definition 1.7. Something is called (or referred to as) astrochicken if it is a
chicken and also a space ship.
Similarly, if in Formal we have a binary predicate ∈ and two constants c and s
then one could introduce a new predicate  into Formal and record the definition:
Definition 1.8. ∀x((x) ↔ ((x ∈ c) ∧ (x ∈ s))).
The two definitions are related by translating ∈, c, s, and  as “belongs to,”
“chicken,” “space ships,” and “is an astrochicken,” respectively. The word as-
trochicken is taken from a lecture by Freeman Dyson.
In a similar way one can introduce new functional symbols or new constants.
In the above discussion we encountered 2 examples of languages that we de-
scribed in some detail (English and Formal) and one example of language (Meta-
language) that we kept somehow vague. Later we will introduce other languages
and make things more precise. We would like to “define” now languages in general;
we cannot do it in the sense of Remark 1.4 because definitions in that sense require
a language to begin with. All we can do is describe in English what the definition
of a language would look like. So the remark below is NOT a definition in the sense
of Remark 1.4.
Remark 1.9. (Description in English of the concept of language) A first order
language (or simply a language) is a collection L of symbols with the following
properties. The symbols in L are divided into 8 categories called logical categories.
They are: variables, constants, functional symbols, predicates, logical connectives,
quantifiers, equality, and separators. Some of these may be missing. Also we assume
that the list of variables and constants may grow indefinitely: we can always add
new variables and constants. Finally we assume that the only allowed separators
are parentheses (, ) and commas; we especially ban quotation marks “...” from
the separators allowed in a language (because we want to use them as parts of
constants in Metalanguage). Given a language L one can consider the collection
L∗ of all strings of symbols. There is an “obvious” way (which will be explained
later) to define a syntactically correct string in L∗ ; such a syntactically correct
string will be called a sentence. The collection of sentences in L∗ is denoted by
Ls . (We sometimes say “sentence in L” instead of sentence in L∗ .) As in the
examples above we can give names P, ... to the sentences in L; these names P, ... do
NOT belong to the original language. A translation of a language L into another
language L0 is a rule that attaches to any symbol in L a symbol in L0 ; we assume
constants are attached to constants, variables to variables, etc. Due to syntactical
12 1. LANGUAGES

correctness such a translation attaches to sentences P in L sentences P 0 in L0 . The

analysis of translations is part of semantics and will be discussed later in more
detail. Actually the above concept of translation should be called word for word
translation (or symbol for symbol) and it is too rigid to be useful. In most examples
translations should be allowed to transform sentences into sentences according to
rules that are more complicated than the symbol for symbol rule. Translations and
reference are required to satisfy the following condition. If L and L0 are equipped
with reference (which is always the case but sometimes ignored) and if P is a
sentence in L whose translation in L0 is P 0 then we impose the condition that P
and P 0 have the same reference.
Remark 1.10. (Correspondences between languages) Translations are an ex-
ample of correspondence between languages. Other examples of correspondences
between languages are linguistic reference (a text referring to another text) and
disquotation (dropping quotation marks).
Example 1.11. (Fixed number of constants) English and Formal are examples
of languages. Incidentally in these languages the list of constants ends (there is a
“fixed number” of constants). But it is important to not impose that restriction
for languages. If instead of English we consider a variant of English in which we
have words involving arbitrary many letters (e.g., words like “man,” “superman,”
“supersuperman,” etc.) then we have an example of a language with “any number
of constants.” There is an easy trick allowing one to reduce the case of an arbitrary
number of symbols to the case of a fixed number of symbols; one needs to slightly
alter the syntax by introducing one more logical category, an operator denoted,
say, by 0 ; then one can form constants c0 , c00 , c000 , ... starting from a constant c; one
can form variables x0 , x00 , x000 , ... from a variable x; and one can do the same with
functional symbols, predicates, etc.; we will not pursue this in what follows.
Example 1.12. (Alternative translations) We already gave an example of
translation of Formal into English; cf. Example 1.1. (Strictly speaking that ex-
ample of translation was not really a word for word translation.) The translation
given there for connectives, quantifiers, and equality is called the standard transla-
tion. But there are alternative translations as follows.
Alternative translations of → into English are “implies,” or “by...it follows
that,” or “since...we get,” etc.
An alternative translations of ↔ into English are “is equivalent to,” “if and
only if.”
Alternative translations of ∀ into English are “for any,” or “for every.”
Alternative translations of ∃ into English are “for some” or “there is an/a.”
English has many other connectives (such as “before,” “after,” “but,” “in spite
of the fact that,” etc.). Some of these we will ignore; others will be viewed as
interchangeable with others; e.g., “but” will be viewed as interchangeable with
“and” (although the resulting meaning is definitely altered). Also English has
other quantifiers (such as “many,” “most,” “quite a few,” “half,” “for at least
three,” etc.); we will ignore these other quantifiers.
Remark 1.13. (Texts) Let us consider the following types of objects:
1) symbols (e.g., x, y, B, C, †, k, ..., ∈, ↓, ..., ∧, ∨, ¬, →, ↔, ∀, ∃, =, (, ));
2) collections of symbols (e.g., the collection of symbols above, denoted by L);
20 ) strings of symbols (e.g., ∃x∀y(x ∈ y));
 1. LANGUAGES 13

3) collections of strings of symbols (e.g., L∗ , Ls encountered above or theories

T to be encountered later);
30 ) strings of strings of symbols (such as the proofs to be encountered later).
In the above, collections are unordered while strings are ordered. The above
types of objects (1, 2, 20 , 3, 30 ) will be referred to as texts. Texts should be thought of
as concrete (physical) objects, like symbols written on a piece of paper or papyrus,
words that can be uttered, images in a book or in our minds, etc. We assume we
know what we mean by saying that a symbol belongs to (or is in) a given collec-
tion/string of symbols; or that a string of symbols belongs to (or is in) a given
collection/string of strings of symbols. We will not need to iterate these concepts.
We will also assume we know what we mean by performing some simple opera-
tions on such objects like: concatenation of strings, deleting symbols from strings,
substituting symbols in strings with other symbols, “pairing” strings with other
strings, etc. These will be encountered and explained later. Texts will be crucial
in introducing our concepts of logic. Note that it might look like we are already
assuming some kind of logic when we are dealing with texts; so our introduction to
logic might seem circular. But actually the “logic” of texts that we are assuming
is much more elementary than the logic we want to introduce later; so what we are
doing is not circular.
For the next exercises one needs to enrich Formal by new symbols as needed.
The translations involved will not be word for word.
Exercise 1.14. Find formalizations of the following English sentences:
1) “I saw a man.”
2) “There is no hope for those who enter this realm.”
3) “There is nobody there.”
4) “There were exactly two people in that garden.”
5) “The cat is on the mat”.
6) “A cat is on the mat”.
7) “If two lines have two points in common then the lines coincide.
8) “For any line and any point not belonging to the line there is a unique line
passing through the point and not meeting the first line.
Exercise 1.15. Find formalizations of the following English sentences:
1) “The movement of celestial bodies is not produced by angels pushing the bodies
in the direction of the movement but by angels pushing the bodies in a direction
perpendicular to the movement.”
2) “I think therefore I am.”
3) “Since existence is a quality and since a perfect being would not be perfect if
it lacked one quality it follows that a perfect being must exist.”
4) “Since some things move and everything that moves is moved by a mover
and an infinite regress of movers is impossible it follows that there is an unmoved
mover.”
Hints: The word “but” should be thought of as “and”; “therefore” should
be thought of as “implies” and hence as “if...then”; “since...it follows” should be
thought of, again, as “implies.”
Remark 1.16. The sentence 1 above paraphrases a statement in one of Feyn-
man’s lectures on gravity. The sentence 2 is, of course Descartes’ “cogito ergo sum.”
The sentence 3 is a version of the “ontological argument” (considered by Anselm,
14 1. LANGUAGES

Descartes, Leibniz, Gödel; cf. Aquinas and Kant for criticism). See 6.12 for more
on this. The sentence 4 is a version of the “cosmological argument” (Aquinas).
Remark 1.17. (Declarative/imperative/interrogative sentences) All sentences
considered so far were declarative (they declare their content). Natural languages
have other types of sentences: imperative (giving a command like: “Lift this
weight!”) and interrogative (asking a question such as: “Is the electron in this
portion of space-time?”). In principle, from now on, we will only consider declara-
tive sentences in our languages. An exception to this is the language called Argot;
see below
Example 1.18. For a language L (such as Formal) we may introduce a new
language called Argotic L (or simply Argot), denoted sometimes by LArgot . Most
mathematics books, for instance, are written in such a language. The language
LArgot has as symbols all the symbols of English together with all the symbols of
a language L, to which one adds one more category of symbols,
• commands: “consider,” “assume,” “let...be,” “let us...,” etc.)
Examples of sentences in Argot are
1) “Since (s ∈ w) → (ρ(s)) it follows that ρ(t)”
2) “Let c be such that ρ(c).”
We will not insist on explaining the syntax of Argot which is rather different from
that of both English and Formal. Suffices to note that the symbols in Formal do
not appear between quotation marks inside sentences of Argot; loosely speaking
the sentences in Argot often appear as obtained from sentences in Metalanguage
via disquotation.
 CHAPTER 2

Metalanguage

In the previous chapter we briefly referred to linguistic reference as being a

correspondence between two languages in which the first language L b “talks about”
a second language L as a language (i.e., it “talks about” the syntax, semantics, etc.
of L). We also say that L b refers to (or has as universe of discourse) the language
L. Once we have fixed L and L b we shall call L the object language and L b the
metalanguage. (The term metalanguage was used by Tarski in his theory of truth;
but our metalanguage differs from his in certain respects, cf. Remark 2.4 below.
Also this kind of correspondence between L b and L is reminiscent of Russell’s theory
of types of which, however, we will say nothing here.)
Metalanguages and object languages are similar structures (they are both lan-
guages!) but we shall keep them separate and we shall hold them to different
standards, as we shall see below. Sentences in metalanguage are called metasen-
tences. If we treat English and Formal as object languages then all our discussion of
English and Formal was written in a metalanguage (which is called Metalanguage)
and hence consists of metasentences. Let’s have a closer look at this concept. First
some examples.
Example 2.1. Assume we have fixed an object language L such as English
or Formal (or several object languages L, L0 , ...). In what follows we introduce
a metalanguage L.b Here are some examples of metasentences in L. b First some
examples of metasentences of the type we already encountered (where the object
language L is either English or Formal):

1) x is a variable in the sentence “∀x(x ∈ a).”

2) P equals “Socrates is a man.”

Later we will encounter other examples of metasentences such as:

3) P (b) is obtained from P (x) by replacing x with b.

4) Under the translation of L into L0 the translation of P is P 0 .
5) By ... the sentence P ∨ ¬P is a tautology.
6) c is a constant
7) The string of sentences P, Q, R, ..., U is a proof of V .
8) V is a theorem.
9) A theorem is a sentence for which there is a proof.
10) If x is a variable then x is a term.

The metasentences 1, 3, 6 are explanations of syntax in L (see later); 2 is a

definition (referred to as a notation or naming); 4 is an explanation of semantics
(see later); 5 is part of a metaproof; and 7, 8 are claims about inference (see later). 9
15
16 2. METALANGUAGE

is a definition in metalanguage (a metadefinition). 10 is an axiom in metalanguage

(a metaaxiom).
Here are the symbols in L:
b
• variables: something, thing,...
• constants: “Socrates,” “Socrates is a man,” “∧,” “=,” L, L∗ , Ls , ..., P, Q, R, ...,
• functional symbols: the variables in, the translation of, the proof of, ∧, ∨, ¬,
→, ↔, ∃x, ∀x,...
• predicates: is translated as, occurs in, is obtained from...by replacing...with...,
is a tautology, is a proof,..., follows from, by ... it follows that..., by ... one gets
that,...
• connectives: and, or, not, if...then, if and only if, because,...
• quantifiers: for all, there exists,...
• equality: is, equals,...
• separators: parentheses, comma, period.
Remark 2.2. Note that names of sentences in the object language become
constants in metalanguage. The texts of the object language, appearing between
quotation marks, also become constants in metalanguage. The connectives of the
object language become functional symbols in metalanguage. The symbols “∧, ...”
used as constants, the symbols ∧, ... used as functional symbols, and the symbols
and,... viewed as connectives should be viewed as different symbols (normally one
should use different notation for them).
Remark 2.3. The above metalanguage can be viewed as a MetaEnglish because
it is based on English. One can construct a MetaFormal metalanguage by replacing
the English words with symbols including:
• connectives: & (for and), ⇒ (for if...then), ⇔ (for if and only if)
• equality: ≡ (for is, equals)
We will not proceed this way, i.e., we will always use MetaEnglish as our meta-
language.
Remark 2.4. What Tarski called metalanguage is close to what we call meta-
language but not quite the same. The difference is that Tarski allows metalan-
guage to contain the symbols of original object language written without quotation
marks. So for him (but not for us), if the language is Formal, then the following is
a metasentence:
“∀x∃ys(x, y)” if and only if ∀x∃ys(x, y)
Allowing the above to be a metasentence helped Tarski define truth in a language
(the Tarski T scheme); we will not do this here.
Remark 2.5. (Syntax/semantics/reference/inference/truth in metalanguage
versus object language)
The syntax of object languages will be regulated by metalanguage. On the
other hand metalanguage has a syntax of its own which we keep less precise than
that of languages so that we avoid the necessity of introducing a metametalan-
guage which regulates it; that would prompt introducing a metametametalanguage
that regulates the metametalanguage, etc. The hope is that metalanguage, kept
sufficiently loose from a syntactic viewpoint, can sufficiently well explain its own
syntax without leading to contradictions. The very text you are reading now is,
in effect, metalanguage explaining its own syntactic problems. The syntactically
 2. METALANGUAGE 17

correct texts in metalanguage are referred to as metasentences. Definitions in meta-

language are called metadefinitions. It is crucial to distinguish between words in
sentences and words in metasentences which are outside the quotation marks; even
if they look the same they should be regarded as different words.
In terms of semantics sentences in object languages are assumed to have a
meaning derived from (or rather identified with) translations into other languages
but we will generally ignore this meaning. On the other hand, metasentences have
a metameaning derived from their own translations into other languages; we shall
assume we understand their metameaning (as it is rather simpler than the meaning
of sentences) and we shall definitely not ignore it.
Metasentences have a reference: they always refer to sentences in the object
language. On the other hand we will ignore the reference of sentences in object
language.
Metasentences, as well as sentences in object language, are assumed to have no
truth value (it does not make sense to say they are true or false).
For instance the metasentences
a. The word “elephants” occurs in the sentence “elephants are blue.”
b. The word “crocodiles” occurs in the sentence “elephants are gray.”
can be translated into the “metalanguage of letter searches” (describing how to
search a word in a sentence, say). Both metasentences have a meaning. Intuitively
we are tempted to say that (a) is true and (b) is false. As already mentioned we do
not want to introduce the concepts of true and false in this course. Instead we in-
fer sentences in object language, respectively, metasentences; inference of sentences
in object languages will be called proof; inference of metasentences in metalan-
guage will be called metaproof. The rules regulating proofs and metaproofs will
be explained as we go. As a general rule metaproofs must be finitistic, by which
we mean that they are not allowed to involve quantifiers in the metalanguage; in
particular we agree that no metasentence involving quantifiers can be metaproved.
Metaproving will also be called checking or showing.
For example we agree that (a) above can be metaproved; also the negation of (b)
can be metaproved. Metaproof is usually based on a translation into a “computer
language”: for instance to metaprove (a) take the words of the sentence “elephants
are blue” one by one starting from the right (say) and ask if the word is “elephants”;
the third time you ask the answer is yes, which ends the metaproof of (a). A similar
discussion applies to some other types of metasentences; e.g., to the metasentences
1-7 in Example 2.1. The metaproof of 5 in Example 2.1 involves, for instance,
“showing tables” whose correctness can be checked by inspection by a machine.
(This will be explained later.) The situation with the metasentences 8 and 9 in
Example 2.1 is quite different: there is no “finitistic” method (program) that can
decide if there is a metaproof for 8; neither is there a “finitistic” method that can
find a metaproof for 8 in case there is one. But if one already has a proof of U in
8 then checking that the alleged proof is a proof can be done finitistically and this
provides a metaproof for 8. The metasentence 9 is a definition in metalanguage.
Most metaproofs consist in checking that a definition applies to a given metasen-
tence. The rules governing the latter would be spelled out in metametalanguage;
we will not do this here.
The fine tuning of object language and metalanguage that we explained above
is based on the following “balance” principle best expressed in a table as follows:
18 2. METALANGUAGE

object language metalanguage

syntactic structure strong weak
semantic structure weak strong
ability to infer strong weak
ability to refer weak strong
truth banned banned

Remark 2.6. Since P, Q are constants in metalanguage and ∧, ∨, ..., ∃x are

functional symbols in metalanguage one can form syntactically correct strings
P ∧ Q, ..., ∃xP
in metalanguage, etc. If
P equals “p...”
Q equals “q...”
where p, ..., q, ... are symbols in the object language then we introduce the rule
(“metaaxiom”)
P ∧ Q equals “(p...) ∧ (q...)”
The above should be viewed as one of the rules allowed in metaproofs. Similar
obvious rules can be given for ∨, ∃, etc. Note that the parentheses are many times
necessary; indeed without them the string P ∨ Q ∧ R would be ambiguous. We will
drop parentheses, however, each time there is no ambiguity. For instance we will
never write (P ∨ Q) ∧ R as P ∨ Q ∧ R. Note that according to these conventions
(P ∨ Q) ∨ R and P ∨ (Q ∨ R) are still considered distinct.
Remark 2.7. Assume we are given a metadefinition:
P equals “p....”
Then we say P is a name for the sentence “p....” We impose the following rule for
this type of metadefinition: if two sentences have the same name they are identical
(as strings of symbols in the object language; identity means exactly the same
symbols in the same order and it is a physical concept). Note on the other hand
that the same sentence in the object language can have different names.
In the same spirit if
P (x) equals “p...x...”
is a metadefinition in metalanguage then we will add to the object language a new
predicate (still denoted by P ) by adding, to the definitions of the object language,
the following definition:
∀x(P (x) ↔ (p...x...)).
So the symbol P appears once as a constant in metalanguage and as a predicate
in the object language. (We could have used two different letters instead of just P
but it is more suggestive to let P play two roles.) This creates what one can call a
correspondence between part of the metalanguage and part of the language. This
correspondence, which we refer to as linguistic reference, is not like a translation be-
tween languages because constants in metalanguage do not correspond to constants
in the object language but to sentences (or to new predicates) in the object lan-
guage. In some sense this linguistic reference is a “vertical” correspondence between
 2. METALANGUAGE 19

languages of “different scope” whereas translation is a “horizontal” correspondence

between languages of “equal scope.” The words “vertical” and “horizontal” should
be taken as metaphors rather than precise terms.
Remark 2.8. (Disquotation) There is a “vertical” correspondence (called dis-
quotation or deleting quotation marks) that attaches to certain metasentences in
metalanguage a sentence in the object language. Consider for instance the metasen-
tence in MetaEnglish
1) From “Socrates is a man” and “all men are mortal” it follows that “Socrates
is mortal.”
Its disquotation is the following sentence in (object) English:
2) “Since Socrates is a man and all men are mortal it follows that Socrates is
mortal.”
Note that 1 refers to some sentences in English whereas 2 refers to something
(somebody) called Socrates. So the references of 1 and 2 are different; and so are
their meaning (if we choose to care about the meaning of 2 which we usually don’t).
If P equals “Socrates is a man,” Q equals “all men are mortal,” and R equals
“Socrates is mortal” then 2 above is also viewed as the disquotation of:
1’) From P and Q it follows that R.
Disquotation is not always performable: if one tries to apply disquotation to
the metasentence
1) x is a free variable in “for all x, x is an elephant”
one obtains
2) “x is a free variable in for all x, x is an elephant”
which is not syntactically correct.
Disquotation is a concept involved in some of the classical theories of truth,
e.g., in Tarski’s example:
“Snow is white” if and only if snow is white.
Since we are not concerned with truth in this course we will not discuss this
connection further.
We will often apply disquotation without any warning if there is no danger of
confusion.
Exercise 2.9. Consider the following utterances and explain how they can be
viewed as metasentences; explain the syntactic structure and semantics of those
metasenteces.
1) To be or not to be, that is the question.
2) I do not feign hypotheses.
3) The sentence labelled 3 is false.
4) You say yes, I say no, you say stop, but I say go, go, go.
1 is, of course, from Shakespeare. 2 is an English translation of a sentence in
Latin from Newton. 3 is, of course, a form of the liar’s “paradox.” 4 is from the
Beatles.
20 2. METALANGUAGE

From now on we will make the following convention. In any discus-

sion about languages we will assume we have fixed an object language
and a metalanguage. The object language will simply be referred to as
the “language.” So the word “object” will systematically be dropped.
Remark 2.10. (Declarative/imperative/interrogative) All metasentences con-
sidered so far were declarative (they declare their content). There are other types
of metasentences: imperative (giving a command like: “Prove this theorem!,” “Re-
place x by b in P ,” “Search for x in P ,” etc.) and interrogative (asking a question
such as: “What is the value of the function f at 5?,” “Does x occur in P ?,” etc.).
The syntax of metasentences discussed above only applies to declarative metasen-
tences. We will only use imperative/interrogative metasentences in the exercises
or some metaproofs (the latter sharing a lot with computer programs); these other
types of metasentences require additional syntactic rules which are clear and we
will not make explicit here.
 CHAPTER 3

Syntax

We already superficially mentioned syntax. In this chapter we discuss, in some

detail, the syntax of languages such as English or Formal. The syntax of metalan-
guages will be not explicitly addressed but should be viewed as similar. The syntax
of Argotic languages will be quite different and will be briefly addressed in a sepa-
rate chapter. All the explanations below are, of course, written in metalanguage.
As we saw a language is a collection L of symbols. Given L we considered the
collection L∗ of all strings of symbols in L. In this chapter we explain the definition
of sentences (which will be certain special strings in L∗ ). Being a sentence will
involve, in particular, a certain concept of “syntactic correctness.” The kind of
syntactic correctness discussed below makes L a first order language. There are
other types of languages whose syntax is different (e.g., second order languages, in
which one is allowed to say, for instance, “for any predicate etc....”; or languages
whose syntax is based on grammatical categories rather than logical categories; or
computer languages, not discussed in this course at all). First order languages are
the most natural (and are entirely sufficient) for developing mathematics.
In what follows we let L be a collection of symbols consisting of variables x, y, ...,
constants, functional symbols, predicates, connectives ∧, ∨, ¬, →, ↔ (where ¬ is
unary and the rest are binary), quantifiers ∀, ∃, equality =, and, as separators,
parentheses (, ), and commas. (For simplicity we considered 5 “standard” con-
nectives, 2 “standard” quantifiers, and a “standard” symbol for equality; this is
because most examples will be like that. However any number of connectives and
quantifiers, and any symbols for them would do. In particular some of these cat-
egories of symbols may be missing.) According to our conventions recall that we
will also fix a metalanguage L b in which we can “talk about” L.

Example 3.1. If L has constants a, b, ..., a functional symbol f , and a binary

predicate  then here are examples of strings in L∗ :
1) )(x∀∃aby(→
2) f (f (a))
3) ∃y((ax) → (ay))
4) ∀x(∃y((ax) → (ay)))
In what follows we will define terms, formulas, and sentences; 1 above will be
neither a term, nor a formula, nor a sentence; 2 will be a term; 3 will be a formula
but not a sentence; 4 will be a sentence.
For the rule (metaaxiom) below we introduce a unary predicate “is a term”
into metalanguage.
Metaaxiom 3.2.
1) If x is a variable then x is a term. If c is a constant then c is a term.
21
22 3. SYNTAX

2) If t, s, ... are terms and f is a functional symbol then f (t, s, ...) is a term.
Remark 3.3. Functional symbols may be unary f (t), binary f (t, s), ternary
f (t, s, u), etc. When we write f (t, s) we simply mean a string of 5 symbols; there
is no “substitution” involved here. Substitution will play a role later, though; cf.
3.12.
Example 3.4. If a, b, ... are constants, x, y, ... are variables, f is a unary func-
tional symbol, and g is a a binary functional symbol, all of them in L, then
f (g(f (b), g(x, g(x, y))))
is a term. The latter should be simply viewed, again, as a string of symbols.
For the next metaaxiom we introduce the predicate is a formula into metalan-
guage.
Metaaxiom 3.5.
1) If t, s are terms then t = s is a formula.
2) If t, s, ... are terms and ρ is a predicate then ρ(t, s, ...) is a formula.
3) If Q, Q0 are formulae then Q ∧ Q0 , Q ∨ Q0 , ¬Q, Q → Q0 , Q ↔ Q0 , ∃xQ, ∀xQ
are formulae.
Formulae of the form 1) or 2) above are called atomic formulae.
Recall our convention that if we have a different number of symbols (written
differently) we make similar metadefinitions for them; in particular some of the
symbols may be missing altogether. For instance if equality is missing from L we
ignore 1.
We introduce a predicate “is in Lf ” into metalanguage and we introduce the
metaaxioms: P is a formula if and only if P is in Lf .
Remark 3.6. Predicates can be unary ρ(t), binary ρ(t, s), ternary ρ(t, s, u),
etc. Again, ρ(t, s) simply means a string of 5 symbols ρ, (, t, s, ) and nothing else.
Sometimes one uses another syntax for predicates: instead of ρ(t, s) one writes tρs
or ρts; instead of ρ(t, s, u) one may write ρtsu, etc. All of this is in the language L.
On the other hand if some variables x, y, ... appear in a formula P we sometimes
write in metalanguage P (x, y, ...) instead of P . In particular if x appears in P
(there may be other variables in P as well) we sometimes write P (x) instead of P .
Formulas of the form (i.e., which are equal to one of) ∀xP , ∀xP (x) are referred
to as universal formulas. Formulas of the form ∃xP , ∃xP (x) are referred to as
existential formulas. Formulas of the form P → Q are referred to as conditional
formulas. Formulas of the form P ↔ Q are referred to as biconditional formulas.
Example 3.7. Let p and q be a ternary and a binary predicate respectively.
The following is a formula (∀y(∃x q(x, y)))) → ¬ p(x, y, z).
Example 3.8. Assume L contains a constant c, a unary predicate ρ, and a
unary functional symbol f . Then the following is a formula:
(∀x(f (x) = c)) → (ρ(f (x)))
For what follows we need to add a predicate “x is free in”. If x is free in P we
also say x is a free variable in P .
Metaaxiom 3.9.
 3. SYNTAX 23

0) x is free in the term x. x is not free in the term c if c is a constant. If x is

free in the term t or in the term s, etc., and f is a functional symbol then x is free
in f (t, s, ...).
1) If x is free in the term t or in the term s then x is free in the formula t = s.
2) If x is free in the term t or in the term s, etc., and if ρ is a predicate then x
is free in the formula ρ(t, s, ...).
3) If x is free in the formula P or in the formula Q then x is free in P ∧ Q,
P ∨ Q, P → Q, P ↔ Q, ¬P .
4) x in not free in ∀xP .
5) x is not free in ∃xP .
Example 3.10.
1) x is not free in ∀y∃x(ρ(x, y)).
2) x is free in (∃x(β(x))) ∨ ρ(x, a). (x has a “free occurrence” in ρ(x, a); the
“occurrence” of x in ∃x(β(x)) is not “free.”)
3) x is free in ∀x((∃x(β(x))) ∨ ρ(x, a))
4) The free variables in (∀x∃y(α(x, y, z))) ∧ ∀u(β(u, y)) are z, y.
Metadefinition 3.11. A string in L∗ is called a sentence if it is a formula
(i.e., is in Lf ) and has no free variables.
Note that
1) If P is in Ls then P is in Lf ;
2) If P is in Lf then P is in L∗ ;
3) all terms are in L∗ ; no term is in Lf .
Metadefinition 3.12. If x is a free variable in a formula P one can replace
all its free occurrences with a term t to get a formula which can be denoted by P xt .
More generally if x, y, ... are variables and t, s, ... are terms, we may replace all free
ts...
occurrences of these variables by t, s, ... to get a formula P xy... . A more suggestive
(but less precise) notation is as follows. We write P (x) instead of P and then we
write P (t) instead of P xt . Similarly we write P (t, s, ...) instead of P xy...
ts...
We will
constantly use this P (t), P (t, s, ...), etc. notation from now on.
Similarly if u is a term containing x and t is another term then one may replace
all occurrences of x in u by t to get a term which we may denote by u xt ; if we write
u(x) instead of u then we can write u(t) instead of u xt . And similarly we may
ts
replace two variables x, y in a term u by two terms t, s to get a term u xy , etc. We
will not make use of this latter type of substitution in what follows.
Example 3.13. If P equals “x is a man” then x is a free variable in P . If a
equals “Socrates” then P (a) equals “Socrates is a man.”
Example 3.14. If P equals “x is a man and for all x, x is mortal” then x is a
free variable in P . If a equals “Socrates” then P (a) equals “Socrates is a man and
for all x, x is mortal.”
Exercise 3.15. Is x a free variable in the following formulas?
1) “(∀y∃x(x2 = y 3 )) ∧ (x is a man)”
2) “∀y(x2 = y 3 )”
Here the upper indexes 2 and 3 are unary functional symbols.
Exercise 3.16. Compute P (t) if:
24 3. SYNTAX

1) P (x) equals “∃y(y 2 = x)” and “t” equals “x4 .”

2) P (x) equals “∃y(y poisoned x)” and “t” equals “Plato’s teacher.”
The following metadefinition makes the concept of definition in L more precise:
Metadefinition 3.17. A definition in L is a sentence of one of the following
types:
1) “c = t” where c is a constant and t is a term without variables.
2) “∀x((x) ↔ E(x))” where  is a unary predicate and E is a formula with
one free variable. More generally for several variables, “∀x∀y((x, y) ↔ E(x, y))”
is a definition, etc.
3) “∀x∀y((y = f (x)) ↔ F (x, y))” where f is a unary functional predicate and
F is a formula with 2 free variables; more generally one allows several variables.
If any type of symbols is missing from the language we disallow, of course, the
corresponding definitions.
A language together with a collection of definitions is referred to as a language
with definitions.
A definition as in 1 should be viewed as either a definition of c (in which case
it is also called notation) or a definition of t. A definition as in 2 should be viewed
as a definition of . A definition as in 3 should be viewed as a definition of f . More
general types of definitions will be allowed later.
Remark 3.18. Given a language L with definitions and a term t without vari-
ables one can add to the language a new constant c and one can add to the def-
initions the definition c = t. We will say that c is (a new constant) defined by
c = t.
Similarly given a language L with definitions and a formula E(x) in L with one
free variable x one can add to the predicates of L a new unary predicate  and one
can add to the definitions the definition ∀x((x) ↔ E(x)). We will say that  is (a
new predicate) defined by ∀x((x) ↔ E(x)).
One may introduce new functional symbols f by adding a symbol f and defini-
tions of type 3 or of type 1 (such as c = f (a, b, ...), for various constants a, b, c, ...).
When adding definitions of new constants, functions, or predicates one should
ask, of course, that these definitions be introduced in a sequence and at each step
in the sequence the symbol that is being introduced has not appeared before (i.e.
it is indeed “new”); this guarantees the predicativity of the definitions (i.e., essen-
tially their non-circularity), at least from a syntactical viewpoint. This device does
not get rid of the semantic impredicativity (which was one of the major themes
in the controversies around the foundation of mathematics at the beginning of the
20th century; cf. Russell, Poincaré, and many others.) However, since we chose
to completely ignore the meaning (semantics) of object languages, semantic im-
predicativity will not be an issue for us. To be sure, later in set theory, semantic
impredicativity is everywhere implicit and might be viewed as implicitly threatening
the whole edifice of mathematics.
 CHAPTER 4

Tautologies

We start now the analysis of inference within a given language (which is also
referred to as deduction or proof). In order to introduce the general notion of proof
we need to first introduce tautologies; in their turn tautologies are introduced via
certain arrays of symbols in metalanguage called tables.
Metadefinition 4.1. Let T and F be two symbols in metalanguage. We also
allow separators in metalanguage that are frames of tables. Using the above plus
arbitrary constants P and Q in metalanguage we introduce the following strings of
symbols in metalanguage (which are actually arrays rather than strings but which
can obviously be rearranged in the form of strings). They are referred to as the
truth tables of the 5 standard connectives.

P Q P ∧Q P Q P ∨Q P Q P →Q P Q P ↔Q
T T T T T T T T T T T T
T F F T F T T F F T F F
F T F F T T F T T F T F
F F F F F F F F T F F T
P ¬P
T F
F T
Remark 4.2. If in the tables above P is the sentence “p....” and Q is the
sentence “q....” we allow ourselves, as usual, to identify the symbols P, Q, P ∧ Q,
etc. with the corresponding sentences “p...,” “q...,” “(p...)∧(q...),” etc. Also: the
letters T and F evoke “truth” and “falsehood”; but they should be viewed as devoid
of any meaning.
Fix in what follows a language L that has the 5 standard connectives ∧, ∨, ¬,
→, ↔ (but does not necessarily have quantifiers or equality).
Metadefinition 4.3. Let P, Q, ..., R be sentences in L. By a Boolean string
generated by P, Q, ..., R we mean a string of sentences
P
Q
...
R
U
...
...
25
26 4. TAUTOLOGIES

such that for any sentence V among U, ..., W we have that V is preceded by V 0 , V 00
(with V 0 , V 00 among P, ..., W ) and V equals one of the following:

V 0 ∧ V 00 , V 0 ∨ V 00 , ¬V 0 , V 0 → V 00 , V 0 ↔ V 00 .

Example 4.4. The following is a Boolean string generated by P, Q, R:

P
Q
R
¬R
Q ∨ ¬R
P → (Q ∨ ¬R)
P ∧R
(P ∧ R) ↔ (P → (Q ∨ ¬R))

Example 4.5. The following is a Boolean string generated by P → (Q ∨ ¬R)

and P ∧ R:

P → (Q ∨ ¬R)
P ∧R
(P ∧ R) ↔ (P → (Q ∨ ¬R))

Remark 4.6. The same sentence may appear as the last sentence in two dif-
ferent Boolean strings; cf. the last 2 examples.

Metadefinition 4.7. Assume we are given a Boolean string generated by

P, Q, ..., R. For simplicity assume it is generated by P, Q, R. (When more or less
than 3 generators the metadefinition is similar.) The truth table attached to this
Boolean string and to the fixed system of generators P, Q, R is the following string
of symbols (or rather plane configuration of symbols thought of as reduced to a
string of symbols):
P Q R U ... W
T T T ... ... ...
T T F ... ... ...
F T T ... ... ...
F T F ... ... ...
T F T ... ... ...
T F F ... ... ...
F F T ... ... ...
F F F ... ... ...
Note that the 3 columns of the generators consist of all 8 possible combinations of T
and F . The dotted columns correspond to the sentences other than the generators
and are computed by the following rule. Assume V is not one of the generators
P, Q, R and assume that all columns to the left of the column of V were computed;
also assume that V is obtained from V 0 and V 00 via some connective ∧, ∨, .... Then
the column of V is obtained from the columns of V 0 and V 00 using the tables of the
corresponding connective ∧, ∨, ..., respectively.
 4. TAUTOLOGIES 27

The above rule should be viewed as a syntactic rule for metalanguage; we

did not introduce the syntax of metalanguage systematically but the above is one
instance when we are quite precise about it.
Example 4.8. Consider the following Boolean string generated by P and Q:
P
Q
¬P
¬P ∧ Q
Its truth table is:
P Q ¬P ¬P ∧ Q
T T F F
T F F F
F T T T
F F T F
Note that the generators P and Q are morally considered “independent” (in the
sense that all 4 possible combinations of T and F are being considered for them);
this is in spite of the fact that actually P and Q may be equal, for instance, to
a = b and ¬(a = b), respectively.
Metadefinition 4.9. A sentence S is a tautology if one can find a Boolean
string generated by some sentences P, Q, ..., R such that
1) The last sentence in the string is S.
2) The truth table attached to the string and the generators P, Q, ..., R has
only T s in the S column.
Remark 4.10. We do not ask, in the metadefinition above, that for any Boolean
string ending in S and for any generators the last column of the truth table have only
T s; we only ask this to hold for one Boolean string and one system of generators.
In all the exercises and examples below, P, Q, ... are specific sentences.
Example 4.11. P ∨ ¬P is a tautology. To metaprove this consider the Boolean
string generated by P ,
P
¬P
P ∨ ¬P
Its truth table is (check!):
P ¬P P ∨ ¬P
T F T
F T T
This ends our metaproof of the metasentence saying that P ∨ ¬P is a tautology.
Remark that if we view the same Boolean string
P
¬P
P ∨ ¬P
as a Boolean string generated by P and ¬P the corresponding truth table is
28 4. TAUTOLOGIES

P ¬P P ∨ ¬P
T T T
T F T
F T T
F F F
and the last column in the latter table does not consist of T s only. This does not
change the fact that P ∨ ¬P is a tautology. Morally, in this latter computation we
had to treat P and ¬P as “independent”; this is not a mistake but rather a failed
attempt to metaprove that P ∨ ¬P is a tautology.
Example 4.12. (P ∧ (P → Q)) → Q is a tautology; it is called modus ponens.
To metaprove this consider the following Boolean string generated by P, Q, R:
P
Q
P →Q
P ∧ (P → Q)
(P ∧ (P → Q)) → Q
Its truth table is:
P Q P → Q P ∧ (P → Q) S
T T T T T
T F F F T
F T T F T
F F T F T
Exercise 4.13. Explain how the table above was computed.
Exercise 4.14. Give a metaproof of the fact that each of the sentences below
is a tautology:
1) (P → Q) ↔ (¬P ∨ Q).
2) (P ↔ Q) ↔ ((P → Q) ∧ (Q → P )).
Exercise 4.15. Give a metaproof of the fact that each of the sentences below
is a tautology:
1) (P ∧ Q) → P .
2) P → (P ∨ Q).
3) ((P ∧ Q) ∧ R) ↔ (P ∧ (Q ∧ R)).
4) (P ∧ Q) ↔ (Q ∧ P ).
5) (P ∧ (Q ∨ R)) ↔ ((P ∧ Q) ∨ (P ∧ R)).
6) (P ∨ (Q ∧ R)) ↔ ((P ∨ Q) ∧ (P ∨ R)).
Metadefinition 4.16.
1) Q → P is called the converse of P → Q.
2) ¬Q → ¬P is called the contrapositive of P → Q.
Exercise 4.17. Give a metaproof of the fact that each of the sentences below
is a tautology:
1) ((P ∨ Q) ∧ (¬P )) → Q (modus ponens, variant).
2) (P → Q) ↔ (¬Q → ¬P ) (contrapositive argument).
3) (¬(P ∧ Q)) ↔ (¬P ∨ ¬Q) (de Morgan law).
4) (¬(P ∨ Q)) ↔ (¬P ∧ ¬Q) (de Morgan law).
 4. TAUTOLOGIES 29

5) ((P → R) ∧ (Q → R)) → ((P ∨ Q) → R) (case by case argument).

6) (¬(P → Q)) ↔ (P ∧ ¬Q) (negation of an implication).
7) (¬(P ↔ Q)) ↔ ((P ∧ ¬Q) ∨ (Q ∧ ¬P )) (negation of an equivalence).
Remark 4.18. 2) in Exercise 4.17 says that the contrapositive of an implication
is equivalent to the original implication.
Exercise 4.19. (VERY IMPORTANT) Give an example showing that the
sentence
(P → Q) ↔ (Q → P )
is not a tautology in general. In other words the converse of an implication is not
equivalent to the original implication. Here is an example in English suggestive of
the above: “If c a human then c mortal” is not equivalent to “If c is mortal then c
is a human.”
Metadefinition 4.20. A sentence P is a contradiction if and only if ¬P is a
tautology.
 CHAPTER 5

Proofs

In this chapter we discuss inference (which is the process by which we construct

proofs). As usual this chapter is written in metalanguage. Recall our language
Argot LArgot which is obtained from the symbols of English LEng , plus the symbols
of a language L (such as Formal), plus command symbols (such as “let,” “consider,”
etc.)
First we need to clarify the syntax of Argot. Rather than developing a detailed
explanation as for languages such as Formal (see the chapter in Syntax) we just
proceed by example.
Metadefinition 5.1. A sentence in Argot is a string of symbols of one of the
following forms:
1) We want to prove U .
2) Assume P .
3) Since P and Q it follows that R.
4) Since s = t we get P (s) = P (t).
5) So R.
6) We know P .
7) There are 2 cases.
8) The first case is A.
9) The second case is B.
10) We seek a contradiction.
11) Let c be such that P (c).
12) Assume there exists c is such that P (c).
13) By P there exists c such that Q(c).
14) This proves P .
Here c is a constant in L; t, s are terms in L; P, Q, R, U, A, B etc. are sentences
in L; etc. Various variants of the above sentences will be also called sentences in
Argot (cf. the examples that follow).
Metadefinition 5.2. A weak theory T in a language L is a collection of
sentences in that language that are grouped into axioms, definitions, theorems, and
proofs.
A weak theory T 0 in a language L0 is said to be a development of a weak theory
T in the language L if L0 is obtained by adding symbols to L and T 0 is obtained
by adding to T :
• definitions
• axioms
• theorems
subject to certain rules to be discussed below.
31
32 5. PROOFS

A weak theory that only has axioms is a theory. A weak theory that is a
development of a theory is a theory.
Axioms are specific to the theory and are referred to as specific axioms; they
are given either as a finite list or by a rule to form them which may create an
indefinitely growing list. When a definition is added to T one also needs to add to
L the newly introduced corresponding symbol. If ∃xP (x) is a theorem or an axiom
one can add a new constant cP to L and one can add to the theory the Theorem
P (cP ); cP is called an existential witness for the sentence ∃xP (x). Apart from the
above procedure of addition of theorems a theorem can be added to T only if a
proof for it is also being supplied; a proof of a sentence U in L is a sequence of
sentences in Argot that is formed as in the examples below (or “combining” the
examples below). There is a precise set of rules which one can use to check if a
sequence of sentences is a proof of a given theorem; we will not go into that but
rather present examples of applications of the various rules.
Metadefinition 5.3. A theory is inconsistent if it has a development that
contains a sentence of the form A∧¬A. A theory is consistent if it is not inconsistent.
A theory is complete if for any sentence A in L there is a development of the theory
in which either A belongs to the development or ¬A belongs to the development.
A theory is incomplete if it is not complete.
Remark 5.4. Clearly no metaproof can be given by inspection of the metasen-
tence that a given theory is consistent. However, if one finds a sentence of the
form A ∧ ¬A then a metaproof was found that the theory is inconsistent. Finally
note that no metaproof can be given in general for the metasentence saying that a
theory is complete or incomplete.
The concepts of logic developed here can be imitated by concepts in set theory
(i.e., in mathematics) and then theorems about set theoretic completeness and
consistency can be proved in set theory (Gödel’s theorems for instance); however
these latter theorems are not metatheorems in logic (i.e., about sentences) but
rather theorems in set theory (i.e., about nothing).
In what follows we clarify what proofs are by examples.
Example 5.5. (Direct proof). Fix a theory T in L with specific axioms A, B, ...
and definitions D, E, .... Say we want to prove H → C. Then H is usually called
hypothesis and C is called conclusion. A proof of H → C can presented as follows.
Theorem 5.6. H → C.
Proof. Assume H. Since H and A it follows that Q. Since Q and B it follows
that R. Hence C. 
The above counts as a proof if
H ∧ A → Q, Q ∧ B → R, R→C
are either tautologies or axioms or theorems previously included in the theory T .

Example 5.7. (Proof by contradiction). Another method of proving sentences

such as H → C is by contradiction and may look as follows.
Theorem 5.8. H → C.
 5. PROOFS 33

Proof. Assume H is true, C is false, and seek a contradiction. Since ¬C and A

it follows that Q. Hence ¬H, a contradiction. This ends the proof. 
The above counts as a proof if
(¬C) ∧ A → Q, Q → ¬H
are either tautologies or axioms or theorems previously included in the theory T .

Next we explain the rules that govern the use of quantifiers in proofs. Here
they are:
1) If in a proof one writes “∀xP (x)” at some point then anywhere after that
one is allowed to write “P (c)” where c is any constant (that has or has not been
used before in the proof).
2) If in a proof one writes “∃xP (x)” at some point then anywhere after that
one is allowed to write “Let c be such that P (c)” (where c is a NEW constant i.e.,
a constant that has NOT been used before in the proof, that needs to be added to
the language). Such a constant is called an existential witness.
3) If one wants to prove a sentence of the form “∀xP (x)” then it is sufficient
to proceed as follows. One first writes “Let c be arbitrary” (where c is a NEW
constant that needs to be added to the language). Such a constant is called a
universal witness. Then one proceeds to proving P (c).
Example 5.9. Here is an example of proof by contradiction that involves quan-
tifiers.
Theorem 5.10. (∀x(¬P (x))) → (¬(∃P (x))).
Proof. Assume ∀x(¬P (x)), ¬¬(∃xP (x)), and seek a contradiction. Since
¬¬(∃xP (x)) it follows that ∃xP (x) (a tautology). Let c be such that P (c). Now
since ∀x(¬P (x)) we get in particular ¬P (c), a contradiction. This ends the proof.

The constant c introduced in the proof is a new constant and is an existential
witness.
A proof can start as a direct proof and involve later an embedded argument by
contradiction. Here is an example.
Theorem 5.11. (¬(∃xP (x))) → (∀x(¬P (x))).
Proof. Assume ¬(∃xP (x)). We want to show that ∀x(¬P (x)). Let c be arbi-
trary; we want to show that ¬P (c). Assume ¬¬P (c) and seek a contradiction. Since
¬¬P (c) it follows that P (c) (a tautology). So ∃xP (x); but we assumed ¬(∃xP (x)),
a contradiction. This ends the proof. 
The constant c in the above proof is a universal witness.
Example 5.12. In order to prove a theorem U of the form P ↔ Q one first
proves P → Q and then one proves Q → P .
Exercise 5.13. Prove the following:
1) (¬(∀xP (x))) ↔ (∃x(¬P (x)))
2) (¬(∀x∀y∃zP (x, y, z))) ↔ (∃x∃y∀z¬P (x, y, z))
Example 5.14. Direct proofs and proofs by contradiction can be given to
sentences which are not necessarily of the form H → C. Here is an example of a
proof by contradiction for:
34 5. PROOFS

Theorem 5.15. C.
Proof. Assume C is false, and seek a contradiction. Since ¬C and A it follows
that Q. Since A and Q we get R. On the there hand since B and ¬C we get ¬R,
a contradiction. This ends the proof. 
The above counts as a proof if
(¬C) ∧ A → Q, A ∧ Q → R, B ∧ ¬C → ¬R
are sentences previously included in the theory T .
Example 5.16. (Case by case proof) Say we want to prove a theorem of the
form:
Theorem 5.17. (H 0 ∨ H 00 ) → C.
Proof. There are two cases: Case 1 is H 0 ; Case 2 is H 00 . Assume first that H 0 .
Then by axiom A we get P . So C. Now assume that H 00 . Since B it follows that
Q. So, again, we get C. So in either case we get C. This ends the proof. 
The above counts as a proof if
A ∧ H 0 → P, P → C, H 00 ∧ B → Q, Q→C
are sentences previously included in the theory T .
The above “case by case” strategy applies more generally to theorems of the
form
Theorem 5.18. H → C
Proof. There are two cases:
Case 1: W holds.
Case 2: ¬W holds.
Assume first we are in Case 1. Then by axiom A we get P . So C.
Now assume we are in Case 2. Since B it follows that Q. So, again, we get C.
So in either case we get C. This ends the proof. 
The above counts as a proof if W is any sentence and
A ∧ W → P, P → C, (¬W ) ∧ B → Q, Q→C
are either tautologies or axioms or theorems previously included in the theory T .
Note that, in the latter proof, finding a sentence W and dividing the proof in
two cases according as W or ¬W holds is usually a creative act: one needs to guess
what W will work.
Here is an example that combines proof by contradiction with “case by case”
proof. Say we want to prove:
Theorem 5.19. H → C.
Proof. Assume H and ¬C and seek a contradiction. There are two cases:
Case 1. W holds.
Case 2. ¬W holds.
In case 1, by ... we get ... hence a contradiction.
In case 2, by ... we get ... hence a contradiction.
This ends the proof. 
Example 5.20. Sometimes a theorem U has the statement:
 5. PROOFS 35

Theorem 5.21. The following conditions are equivalent:

1) P;
2) Q;
3) R.
What is being meant is that U is
(P ↔ Q) ∧ (P ↔ R) ∧ (Q ↔ R)
One proceeds “in a circle” by proving first P → Q then Q → R then R → P .
Example 5.22. In order to prove a theorem of the form P ∧ Q one first proves
P and proves Q.
Example 5.23. In order to prove a theorem of the form P ∨Q one may proceed
by contradiction as follows. Assume ¬P and ¬Q and one seeks a contradiction.
Example 5.24. In order to prove a theorem of the form ∀xP (x) one may
proceed as follows. One writes “Let c be arbitrary” (where c is a constant that has
never been used before in the proof) and then one proves P (c).
Example 5.25. In order to prove a theorem of the form ∃xP (x) it is enough
to find a constant c (that may or may not have been used before in the proof)
such that P (c). This is called a proof by example and it applies only to existential
sentences ∃xP (x) (NOT to universal sentences like ∀xP (x)).

Metadefinition 5.26. Let S be a sentence. A negation of S is a sentence R

such that ¬S ↔ R is a theorem. We say that a negation R of S is in reduced form
if the connective → does not appear in R and the connective ¬ only appears in R
in front of atomic formulae contained in R.
Example 5.27. Let P, Q be atomic sentences. Let S be the sentence
P ∧Q
0 00
Let S , S be the sentences
S 0 = “¬(P ∧ Q)”
S 00 = “¬P ∨ ¬Q”
0
Then S is a negation of S but is it not in reduced form because ¬ appears in front
of a non-atomic formula. On the other hand S 00 is a negation of S in reduced form.
It is a negation because
¬(P ∧ Q) ↔ (¬P ∨ ¬Q)
is a tautology (hence a theorem); it is in reduced form because ¬ appears in it only
in front of atomic formulae and → does not appear in S 00 .
Example 5.28. Let P (x, y), Q(x, y) be atomic formulae. Let S be the sentence
S = “∀x∃y(P (x, y) → Q(x, y))”
0 00 000
Let S , S , S be the sentences
S0 = “¬(∀x∃y(P (x, y) → Q(x, y)))”

S 00 = “∃x(¬(∃y(P (x, y) → Q(x, y))))”

S 000 = “∃x∀y((P (x, y) ∧ ¬Q(x, y)))”

36 5. PROOFS

Then S 0 is a negation of S because ¬S ↔ S 0 is a theorem, but S 0 is not in reduced

form because ¬ appears in S 0 in front of a non-atomic formula. Also S 00 is a negation
of S because ¬S ↔ S 0 is a theorem, but, again, S 00 is not in reduced form because
¬ appears in S 0 in front of a non-atomic formula. Finally S 000 is a negation of S in
reduced form. It is a negation because
¬(A → B) ↔ (A ∧ ¬B)
is a tautology (hence a theorem) and because
¬(∀xA(x)) ↔ ∃x(¬A(x))
and
¬(∃xA(x)) ↔ ∀x(¬A(x))
000
are theorems. S is in reduced form because it does not contain → and ¬ only
appears in it in front of atomic formulae.
Exercise 5.29. Find negations in reduced form of the following sentences. In
this exercise P, Q are atomic formulae involving two variables.
1) ∀x∀y(P (x, y) ∧ Q(x, y)).
2) ∀x∀y(P (x, y) ∨ Q(x, y)).
3) ∃x∀y(P (x, y) → (∃zQ(z, y))).
4) (∃x∀yP (x, y)) → (∀y∃xQ(x, y)).
Exercise 5.30. Formalize the following sentences. Find the negations in re-
duced form of those formalizations. Translate into English those negations.
1) If Plato is a bird then Plato eats nuts.
2) John is not Plato and Paul is not Aristotle.
3) For any line and any point that does not belong to the line there exists no
line passing through the point and parallel to the first line.
4) For any line and any point that does not belong to the line there exists at
least two distinct lines passing through the point and parallel to the first line.
5) For any  there exists δ such that for any x if |x−a| < δ then |f (x)−f (a)| < .
(Here f and | | are unary functions and < is a binary predicate. The sentence above
plays a role in analysis.)

We end by discussing fallacies. A fallacy is a logical mistake. Here are some

typical fallacies:
Example 5.31. Confusing an implication with its converse. Say we want to
prove that H → C. A typical mistaken proof would be: Assume C; then by ...
we get that ... hence H. The error consists of having proved Q → P rather than
P → Q.
Example 5.32. Proving a universal sentence by example. Say we want to prove
∀xP (x). A typical mistaken proof would be: By ... there exists c such that ... hence
... hence P (c). The error consists in having proved ∃xP (x) rather than ∀xP (x).
Example 5.33. Defining a constant twice. Say we want to prove ¬(∃xP (x))
by contradiction. A mistaken proof would be: Assume there exists c such P (c).
Since we know that ∃xQ(x) let c be (or define c) such that Q(c). By P (c) and Q(c)
we get ... hence ..., a contradiction. The error consists in defining c twice in two
unrelated ways: first c plays the role of an existential witness for P ; then c plays
 5. PROOFS 37

the role of an existential witness for Q. But these existential witnesses are not the
same.
Exercise 5.34. Give examples of wrong proofs of each of the above types. If
you can’t solve this now, wait until we get to discuss the integers.
Remark 5.35. Later, when we discuss induction we will discuss another typical
fallacy; cf. Example 13.7.
 CHAPTER 6

Theories

We analyze in what follows a few toy examples of theories and proofs of the-
orems in these theories. Later we will present the main example of theory in this
course which is set theory (identified with mathematics itself).
Example 6.1. The first example is that of group theory. The language L of
the theory has a constant e, variables x, y, ..., and a binary functional symbol ?.
We introduce the following definition in L:
Definition 6.2. z is a neutral element if
∀x((z ? x = x) ∧ (x ? z = x)).
So we added “is a neutral element” as a new unary predicate.
We also introduce the following
Definition 6.3. y is an inverse of x if x ? y = y ? x = e.
So we added “is an inverse of” as a new binary predicate.
The axioms of the theory are:
Axiom 6.4. For all x, y, z we have x ? (y ? z) = (x ? y) ? z.
Axiom 6.5. e is a neutral element.
Axiom 6.6. For any x there is y such that y is an inverse of x.
We prove the following:
Theorem 6.7. For any z if z is a neutral element then z = e.
The sentence that needs to be proved is of the form: “If H then C.” Recall that
in general for such a sentence H is called hypothesis and C is called conclusion.
Here is a direct proof:
Proof. Let f be a neutral element. Since e is a neutral element it follows
that ∀x(e ? x = x). By the latter e ? f = f . Since f is a neutral element we get
∀x(x ? f = x). So e ? f = e. Hence we get e = e ? f . So e = f . 
Here is a proof by contradiction of the same theorem:

Proof. Assume e 6= f where f is a neutral element and seek a contradiction.

So either e 6= e ? f or e ? f 6= f . Since ∀x(e ? x = x) it follows that e ? f = f . Since
∀x(x ? f = x) we get e ? f = e. So we get e = f , a contradiction. 
Here is another
Theorem 6.8. For any x, y, z if y and z are inverses of x then y = z.
39
40 6. THEORIES

Proof Let a, b, c be such that b and c are inverses of a. We want to show that
b = c. By Axioms 6.4, 6.5 and the definition of inverse we have
b ? e = b ? (a ? c) = (b ? a) ? c = e ? c = c.

Group theory can be developed independently of what later we will call set
theory (which is identified with mathematics); a better approach however is to
replace the group theory above with a series of theorems in set theory.

The next example is related to the “Pascal wager.”

Example 6.9. The structure of Pascal’s wager argument is as follows. If God
exists and I believe it exists then I will be saved. If God exists and I do not believe
it exists then I will not be saved. If God does not exist but I believe it exists I
will not be saved. Finally if God does not exist and I do not believe it exists then
I will not be saved. Pascal’s conclusion is that if he believes that God exists then
there is a one chance in two that he be saved whereas if he does not believe that
God exists then there is a zero chance that he be saved. So he should believe that
God exists. The next example is a variation of Pascal’s wager showing that if one
requires “sincere” belief rather than just belief based on logic then Pascal will not
be saved. Indeed assume the specific axioms:
A1) If God exists and a person does not believe sincerely in its existence then
that person will not be saved.
A2) If God does not exist then nobody will be saved.
A3) If a person believes that God exists and his/her belief is motivated only
by Pascal’s wager then that person does not believe sincerely.
We want to prove the following
Theorem 6.10. If Pascal believes that God exists but his belief is motivated by
his own wager only then Pascal will not be saved.
All of the above is formulated in the English language L0 . We consider a simpler
language L and a translation of L into L0 .
The new language L contains among its constant p (for Pascal) and contains 4
unary predicates g, w, s, r whose translation in English is as follows:
g is translated as “is God”
w is translated as “believes motivated only by Pascal’s wager”
s is translated as “believes sincerely”
r is translated as “is saved”
The specific axioms are
A1) ∀y((∃xg(x)) ∧ (¬s(y)) → ¬r(y)).
A2) ∀y((¬(∃xg(x))) → (¬r(y))).
A3) ∀y(w(y) → (¬(s(y)))).
In this language Theorem 6.10 is the translation of the following:
Theorem 6.11. If w(p) then ¬r (p).
So to prove Theorem 6.10 in L0 it is enough to prove Theorem 6.11 in L. We
will do this by using a combination of direct proof and case by case proof.

Proof of Theorem 6.11. Assume w(p). There are two cases: the first case is
∃xg(x); the second case is ¬(∃xg(x)). Assume first that ∃xg(x). Since w(p), by
 6. THEORIES 41

axiom A3 it follows that ¬s(p). By axiom A1 (∃xg(x)) ∧ (¬s(p)) → ¬r(p)). Hence

¬r(p). Assume now ¬(∃xg(x)). By axiom A2 we then get again ¬r(p). So in either
case we get ¬r(p) which ends the proof.


The next example is the famous “ontological argument” for the existence of
God (cf. Anselm, Descartes, Leibnitz, Gödel). The version below is, in some sense,
a “baby version” of the argument; Gödel’s formalization (which he never published)
is considerably subtler. Cf. (Wang 1996).
Example 6.12. The structure of the classical ontological argument for the
existence of God is as follows. Let us assume that qualities (same as properties)
are either positive or negative (and none is both). Let us think of existence as
having 2 kinds: existence in mind (which shall be referred to as belonging to mind)
and existence in reality (which shall be referred to as belonging to reality). It is
not important that we do not know what mind and reality are; we just see them
as English words here. The 2 kinds are not necessarily related: belonging to mind
does not imply (and is not implied by) belonging to reality. (In particular we do
not view mind necessarily as part of reality which we should not: unicorns belong
to mind but not to reality.) The constants and variables refer to things (myself,
my cat, God,...) or qualities (red, omnipresent, deceiving, eternal, murderous,
mortal,...); we identify the latter with their extensions which are, again, things (the
Red, th Omnipresent, the Deceiving, the Eternal, the Murderous, the Mortal,...)
In particular we consider the following constants: reality, mind, God, the Positive
Qualities. We also consider the binary predicate belongs to. We say a thing has
a certain quality (e.g. my cat is eternal) if that thing belongs to the extension of
that quality (e.g. my cat belongs to the Eternal). Assume the following axioms:
A1) There exists a thing belonging to mind that has all the positive qualities
and no negative quality. (Call it God.)
A2) “Being real” is a positive quality.
A3) Two things belonging to mind that have exactly the same qualities are
identical.
(Axiom A3 is Leibniz’s famous principle of identity of indiscernibles. It implies
that God is unique.) Then one can prove the following:
Theorem 6.13. God belongs to reality.
In other words God is real.
The above sentences are written in the English language L0 . Let us formalize
the above in a language L and prove a formal version of Theorem 6.13 in L whose
translation is Theorem 6.13. Assume L contains among its constants the constants
r, m, p and a binary predicate E. We consider a translation of L into L0 such that
r is translated as “reality”;
m is translated as “mind”;
p is translated as “the positive qualities”
xEy is translated as “x belongs to y”.
The specific axioms are:
A1) ∃x((xEm) ∧ (∀z((zEp) ↔ (xEz)))).
A2) rEp.
A3) ∀x∀y(((xEm) ∧ (yEm)) → ((∀z(xEz ↔ yEz)) → (x = y))).
42 6. THEORIES

Note that later, in set theory, we will have a predicate ∈ which, like E, will be
translated as “belongs to” (as in an object belongs to the collection of objects that
have a certain quality); but the axioms are different.
By A1 we can make the following:
Definition 6.14. g = c(xEm)∧(∀z((zEp)↔(xEz))) .
So g is defined to be equal to a certain existential witness.
Also we can add to the theory the following
Theorem 6.15. (gEm) ∧ (∀z((zEp) ↔ (gEz))).
We will translate g as “God”. We have the following theorem expressing the
uniqueness of God:
Theorem 6.16. ∀x(((xEm) ∧ (∀z((zEp) ↔ (xEz))) → (x = g))).
The translation of the above in English is: “If something in my mind has all
the positive qualities and no negative quality then that thing is God.”
Proof. A trivial exercise using axiom A3 only. 
Exercise 6.17. Prove Theorem 6.16.
On the other hand, and more importantly, we have the following Theorem
whose translation in L0 is “God belongs to reality”:
Theorem 6.18. gEr.
Proof. By axiom A1 we have gEm and
∀z((zEp) ↔ (gEz)).
Hence we have, in particular,
(rEp) ↔ (gEr).
By axiom A2, rEp. Hence gEr. 
The argument above is, of course, correct. What is questionable is the choice
of the axioms and the reference of L. Also recall that, in our notes, the question
of truth was not addressed; so it does not make sense to ask whether the English
sentence “God has existence in reality” is true of false. For criticism of the relevance
of this argument (or similar ones) see, for instance, (Kant 1991) and (Wang 1996).
However, the mere fact that some of the most distinguished logicians of all times (in
particular Leibniz and Gödel) took this argument seriously shows that the argument
has merit and, in particular, cannot be dismissed on trivial grounds.
Example 6.19. The next example is again a toy example and comes from
physics. In order to present this example we do not need to introduce any physical
concepts. But it would help to keep in mind the two slit experiment in quantum
mechanics (for which we refer to Feynman’s Physics course, say). Now there are
two types of physical theories that can be referred to as phenomenological and
explanatory. They are intertwined but very different in nature. Phenomenological
theories are simply descriptions of phenomena/effects of (either actual or possible)
experiments; examples of such theories are those of Ptolemy, Copernicus, or that of
pre-quantum experimental physics of radiation. Explanatory theories are systems
postulating transcendent causes that act from behind phenomena; examples of such
 6. THEORIES 43

theories are those of Newton, Einstein, or quantum theory. The theory below is
a baby example of the phenomenological (pre-quantum) theory of radiation; our
discussion is therefore not a discussion of quantum mechanics but rather it suggests
the necessity of introducing quantum mechanics. The language L0 and definitions
are those of experimental/phenomenological (rather than theoretical/explanatory)
physics. We will not make them explicit. Later we will move to a simplified language
L and will not care about definitions.
Consider the following specific axioms (which are the translation in English of
the phenomenological predictions of classical particle mechanics and classical wave
mechanics, respectively):

A1) If radiation in the 2 slit experiment consists of a beam of particles then the
impact pattern on the photographic plate consists of a series of successive flashes
and the pattern has 2 local maxima.
A2) If radiation in the 2 slit experiment is a wave then the impact pattern on
the photographic plate is not a series of successive flashes and the pattern has more
than 2 local maxima.
We want to prove the following
Theorem 6.20. If in the 2 slit experiment the impact consists of a series of
successive flashes and the impact pattern has more than 2 local maxima then in this
experiment radiation is neither a beam of particles nor a wave.
The sentence reflects one of the elementary puzzles that quantum phenomena
exhibit: radiation is neither particles nor waves but something else! And that
something else requires a new theory which is quantum mechanics. (A common
fallacy would be to conclude that radiation is both particles and waves !!!) Rather
than analyzing the language L0 of physics in which our axioms and sentence are
stated (and the semantics that goes with it) let us introduce a simplified language
L as follows.
We consider the language L with constants a, b, ..., variables x, y, ..., and unary
predicates p, w, f , m. Then there is a translation of L into L0 such that:
p is translated as “is a beam of particles”
w is translated as “is a wave”
f is translated as “produces a series of successive flashes”
m is translated as “produces a pattern with 2 local maxima”
Then we consider the specific axioms
A1) ∀x(p(x) → (f (x) ∧ m(x))).
A2) ∀x(w(x) → (¬f (x)) ∧ ¬m(x))).
Here we tacitly assume that the number of maxima cannot be 1. Theorem 6.20
above is the translation of the following theorem in L:
Theorem 6.21. ∀x((f (x) ∧ (¬m(x))) → ((¬p(x)) ∧ (¬w(x)))).
So it is enough to prove Theorem 6.21. The proof below is, as we shall see, a
combination of proof by contradiction and case by case.

Proof. We proceed by contradiction. So assume there exists a such that

f (a) ∧ (¬m(a))
44 6. THEORIES

and
¬(¬p(a) ∧ (¬w(a)))
and seek a contradiction. Since ¬(¬p(a)∧(¬w(a))) we get p(a)∨w(a). There are two
cases. The first case is p(a); the second case is w(a). We will get a contradiction
in each of these cases separately. Assume first p(a). Then by axiom A1 we get
f (a) ∧ m(a), hence m(a). But we assumed f (a) ∧ (¬m(a)), hence ¬m(a), so we
get a contradiction. Assume now w(a). By axiom A2 we get (¬f (a)) ∧ (¬m(a))
hence ¬f (a). But we assumed f (a) ∧ (¬m(a)), hence f (a), so we get again a
contradiction. .
Exercise 6.22. Consider the specific axioms A1 and A2 above and also the
specific axioms:
A3) ∃x(f (x) ∧ (¬m(x))).
A4) ∀x(p(x) ∨ w(x)).
Metaprove that the theory with specific axioms A1, A2, A3, A4 is inconsistent.
A3 is translated as saying that in some experiments one sees a series of successive
flashes and, at the same time, one has more than 2 maxima. Axiom A4 is translated
as saying that any type of radiation is either particles or waves. The inconsistency
of the theory says that classical (particle and wave) mechanics is not consistent
with experiment. (So a new mechanics, quantum mechanics, is needed.) Note that
none of the above discussion has anything to do with any concrete proposal for a
quantum mechanical theory; all that the above suggests is the necessity of such a
theory.
Example 6.23. The next example is a logical puzzle from the Mahabharata.
King Yudhishthira loses his kingdom to Sakuni at a game of dice; then he stakes
himself and he loses himself; then he stakes his wife Draupadi and loses her too.
She objects by saying that her husband could not have staked her because he did
not own her anymore after losing himself. Here is a possible formalization of her
argument.
We use a language with constants i, d, ..., variables x, y, z, ..., the binary pred-
icate “owns,” quantifiers, and equality =. We define a predicate 6= by (x 6= y) ↔
(¬(x = y)). Consider the following specific axioms:
A1) For all x, y, z if x owns y and y owns z then x owns z.
A2) For all y there exists x such that x owns y.
A3) For all x, y, z if y owns x and z owns x then y = z.
We will prove the following
Theorem 6.24. If i does not own himself then i does not own d.
Proof. We proceed by contradiction. So we assume i does not own i and i owns
d and seek a contradiction. There are two cases: first case is d owns i; the second
case is d does not own i. We prove that in each case we get a contradiction. Assume
first that d owns i; since i owns d, by axiom A1, i owns i, a contradiction. Assume
now d does not own i. By axiom A2 we know that there exists j such that j owns
i. Since i does not own i it follows that j 6= i. Since j owns i and i owns d, by
axiom A1, j owns d. But i also owns d. By axiom A3, i = j, a contradiction. 
Example 6.25. This example illustrates the logical structure of the Newto-
nian theory of gravitation that unified Galileo’s phenomenological theory of falling
bodies (the physics on Earth) with Kepler’s phenomenological theory of planetary
 6. THEORIES 45

motion (the physics of “Heaven”); Newton’s theory counts as an explanatory the-

ory because its axioms go beyond the “facts” of experiment. The language L in
which we are going to work has variables x, y, ..., constants S, E, M (translated into
English as “Sun, Earth, Moon”), a constant R (translated as “the radius of the
Earth”), constants 1, π, r (where r is translated as a particular rock), predicates
p, c, n (translated into English as “is a planet, is a cannonball, is a number”), a
binary predicate ◦ (whose syntax is x ◦ y and whose translation in English is “x
revolves around the fixed body y”), a binary predicate f (where f (x, y) is trans-
lated as “x falls freely under the influence of y”), a binary functional symbol d
(“distance between the centers of”), a unary functional symbol a (“acceleration”),
a unary functional symbol T (where T (x, y) is translated as “period of revolution
of x around y”), binary functional symbols :, × (“division, multiplication”), and all
the standard connectives, quantifiers, and parentheses. Note that we have no pred-
icates for mass and force; this is remarkable because it shows that the Newtonian
revolution has a purely geometric content. Now we introduce a theory T in L via
its special axioms. The special axioms are as follows. First one asks that distances
are numbers:
∀x∀y(n(d(x, y)))
and the same for accelerations, and times of revolution. (Note that we view all
physical quantities as measured in centimeters and seconds.) For numbers we ask
that multiplication and division of numbers are numbers:
(n(x) ∧ n(y)) → (n(x : y) ∧ n(x × y))
and that the usual laws relating : and × hold. Here are two:
∀x(x : x = 1).
∀x∀y∀z∀u((x : y = z : u) ↔ (x × u = z × y)).
It is an easy exercise to write down all these laws. We sometimes write
x
, 1/x, xy, x2 , x3 , ...
y
in the usual sense. The above is a “baby mathematics” and this is all mathematics
we need. Next we introduce an axiom whose justification is in mathematics, indeed
in calculus; here we ignore the justification and just take this as an axiom. The
axiom gives a formula for the acceleration of a body revolving in a circle around a
fixed body. (See the exercise after this example.) Here is the axiom:
 2

A) ∀x∀y x ◦ y) → (a(x, y) = 4π 2
d(x,y)
T (x,y) .
To this one adds the following “obvious” axioms
O1) ∀x(c(x) → d(x, E) = R),
O2) M ◦ E,
R) c(r),
K1) ∀x(p(x) → (x ◦ S)),
saying that the distance between cannonballs and the center of the Earth is the
radius of the Earth; that the Moon revolves around the Earth; that the rock r is
a cannonball; and that all planets revolve around the Sun. (The latter is Kepler’s
first law in an approximate form; the full Kepler’s first law specifies the shape of
orbits as ellipses, etc.) Now we consider the following sentences (NOT AXIOMS!):
46 6. THEORIES

G) ∀x∀y((c(x)
 ∧ c(y)) → (a(x,
 3E) = a(y, 3E)), 
K3) ∀x∀y (p(x) ∧ p(y)) → Td 2(x,S) d (y,S)
(x,S) = T 2 (y,S) ,
  
a(x,z) a(y,z)
N) ∀x∀y∀z (f (x, z) ∧ f (y, z)) → 1/d2 (x,z) = 1/d2 (y,z) .

G represents Galileo’s great empirical discovery that all cannonballs (by which
we mean here terrestrial airborne objects with no self-propulsion) have the same
acceleration towards the Earth. K3 is Kepler’s third law which is his empirical
great discovery that the cubes of distances of planets to the Sun are in the same
proportion as the squares of their periods of revolution. Kepler’s second law about
equal areas being swept in equal times is somewhat hidden in axiom A above.
N is Newton’s law of gravitation saying that the accelerations of any two bodies
moving freely towards a fixed body are in the same proportion as the inverses of
the squares of the respective distances to the (center of the) fixed body. Newton’s
great invention is the creation of a binary predicate f (where f (x, y) is translated
into English as “x is in free fall with respect to y”) equipped with the following
axioms
F1) ∀x(c(x) → f (x, E))
F2) f (M, E)
F3) ∀x(p(x) → f (x, S))
expressing the idea that cannonballs and the Moon moving relative to the Earth
and planets moving relative to the Sun are instances of a more general predicate
expressing “free falling.” Finally let us consider the definition
g = a(r, E)
and the following sentence:
4π 2 d3 (M,E)
X) g = R2 T 2 (M,E) .

The main results are the following theorems in T :

Theorem 6.26. N → X.
Proof. See the exercise after this example.
Theorem 6.27. N → G.
Proof. See the exercise after this example.
Theorem 6.28. N → K3.
Proof. See the exercise after this example.
So if one accepts Newton’s N then Galileo’s G and Kepler’s K3 follow, that
is to say that N “unifies” terrestrial physics with the physics of Heaven. The
beautiful thing is, however, that N not only unifies known paradigms but “predicts”
new “facts,” e.g., X. Indeed one can verify X using experimental (astronomical
and terrestrial physics) data: if one enlarges our language to include numerals
and numerical computations and if one introduces axioms as below (justified by
measurements) then X becomes a theorem. Here are the additional axioms:
g = 981 (the number of centimeters per second squared representing g).
π = 314
100 (approximate value).
 6. THEORIES 47

R =number of centimeters representing the radius of the Earth (measured for

the first time by Eratosthenes using shadows at two points on Earth).
d(M, E) =number of centimeters representing the distance from Earth to Moon
(measured using parallaxes).
T (M, E) =number of seconds representing the time of revolution of the Moon
(the equivalent of 28 days).
The fact that X is verified with the above data is the miraculous computation done
by Newton that convinced him of the validity of his theory; see the exercise after
this example.
Remark 6.29. This part of Newton’s early work had a series of defects: it
was based on the circular (as opposed to elliptical) orbits, it assumed the center
of the Earth (rather than all the mass of the Earth) as responsible for the effect
on the cannonballs, it addressed only revolution around a fixed body (which is not
realistic in the case of the Moon, since, for instance, the Earth itself is moving), and
did not explain the difference between the d3 /T 2 of planets around the Sun and
the corresponding quantity for the Moon and cannonballs relative to the Earth.
Straightening these and many other problems is part of the reason why Newton
postponed publication of his early discoveries. The final theory of Newton involves
the introduction of absolute space and time, mass, and forces. The natural way to
develop it is within mathematics, as mathematical physics; this is essentially the
way Newton himself presented his theory in published form. However, the above
example suggests that the real breakthrough was not mathematical but at the level
of (pre-mathematical) logic.
Exercise 6.30.
1) Justify Axiom A above using calculus or even Euclidean geometry plus the
definition of acceleration in an introductory physics course.
2) Prove Theorems 6.26, 6.27, and 6.28.
3) Verify that with the numerical data for g, π, R, d(M, E), T (M, E) available
from astronomy (find the numbers in astronomy books) the sentence X becomes a
theorem. This is Newton’s fundamental computation.
 CHAPTER 7

ZFC

Mathematics is a particular theory Tset (called set theory) in a particular lan-

guage Lset (called the language of set theory) with specific axioms called the ZF C
axioms (the Zermelo-Fraenkel+Choice axioms). The specific axioms do not form
a finite list so they cannot be included in the theory at the beginning. They will
have to be added one by one on a need to use basis to various developments of the
theory. By abuse of terminology we will refer to the various developments of set
theory again as set theory; this should introduce no confusion. We will introduce
all of this presently.
The origins of set theory are in the work of Cantor. Set theory as presented
in what follows is not Cantor set theory but rather an axiomatic version of it due
essentially to Zermelo and Fraenkel. The difference between Cantor’s paradigm and
the Zermelo-Fraenkel paradigm lies, essentially, in the ontological status of sets. In
contrast to Cantor, for whom sets were real collections of “things,” any set in the
definitions below is simply a constant (hence a single symbol) in a language.
Metadefinition 7.1. The language Lset of set theory is the language with
variables x, y, z, ..., no constants, no functional symbol, a binary predicate ∈, con-
nectives ∨, ∧, ¬, →, ↔, quantifiers ∀, ∃, equality =, and separators (, ), ,. As usual
we are allowed to add, whenever it is convenient, new constants, usually denoted
by
a, b, c, ..., A, B, C, ..., A, B, C, ..., α, β, γ, ...,
and new predicates to Lset together with definitions for each of these new symbols.
The constants of Lset will be called sets.
In particular we introduce new predicates 6=, 6∈, ⊂, 6⊂ and the following defini-
tion for them:
Definition 7.2.
∀x∀y((x 6= y) ↔ (¬(x = y))).
∀x∀y((x 6∈ y) ↔ (¬(x ∈ y))).
∀x∀y((x ⊂ y) ↔ (∀z((z ∈ x) → (z ∈ y)))).
∀x∀y((x 6⊂ y) ↔ (¬(x ⊂ y))).
Definition 7.3. x is a subset of y if x ⊂ y.
Remark 7.4. We recall the fact that Lset being an object language it does not
make sense to say that a sentence in it (such as, for instance, a ∈ b) is true or false.
Remark 7.5. Later we will introduce the concept of “countable” set and we
will show that not all sets are countable. On the other hand in set theory there
are always only “finitely many” sets (in the sense that one is using finitely many
symbols) although their collection may be increased any time, if necessary. Let
49
50 7. ZFC

us say that such a collection of symbols is “metacountable.” This seems to be a

paradox which is referred to as the “Skolem paradox.” Of course this is not going
to be a paradox: “metacountable” and “countable” will be two different concepts.
The word “metacountable” belongs to the metalanguage and can be translated into
English in terms of arranging symbols on a piece of paper; whereas “b is countable”
is a definition in set theory. We define “b is countable ↔ C(b)” where C(x) is a
certain formula with free variable x in the language of set theory.
Remark 7.6. There is a standard translation of the language Lset of set theory
into the English language as follows:
a, b, ... are translated as “the set a,” “the set b,”...
∈ is translated as “belongs to the set” or as “is an element of the set”
= is translated as “equals”
⊂ is translated as “is a subset” or as “is contained in”
∀ is translated as “for all sets”
∃ is translated as “there exists a set”
while the connectives are translated in the standard way.
Remark 7.7. Once we have a translation of Lset into English we can speak of
Argot and translation of Lset into Argot; this simplifies comprehension of mathe-
matical texts considerably.
Remark 7.8. The standard translation of the language of set theory into Eng-
lish (in the remark above) is standard only by convention. A perfectly good different
translation is, for instance, the one in which
a, b, ... are translated as “crocodile a,” “crocodile b,”...
∈ is translated as “is dreamt by the crocodile”
= is translated as “has the same taste”
∀ is translated as “for all crocodiles”
∃ is translated as “there exists a crocodile”
One could read mathematical texts in this translation; admittedly the English
text that would result from this translation would be somewhat strange.
Remark 7.9. Note that mathematics uses other symbols as well such as
Z
X df
≤, ◦, +, ×, an , Z, Q, R, C, Fp , ≡, lim an , f (x)dx, , ...
dx
These symbols will originally be all sets (hence constants) and will be introduced
through appropriate definitions (like the earlier definition of an elephant); they will
all be defined through the predicate ∈. In particular in the language of sets, +
or × are NOT originally functional symbols; and ≤ is NOT originally a predicate.
However we will later tacitly enlarge the language of set theory by adding predicates
(usually still denoted by) + or ≤ via appropriate definitions.
We next introduce the (specific) axioms of set theory.
Axiom 7.10. (Singleton axiom)
∀x∃y((x ∈ y) ∧ (∀z((z ∈ y) → (z = x))).
The translation in Argot is that for any set x there is a set y whose only element
is x.
 7. ZFC 51

Axiom 7.11. (Unordered pair axiom)

∀x∀x0 ∃y((x ∈ y) ∧ (x0 ∈ y) ∧ ((z ∈ y) → ((z = x) ∨ (z = x0 )))).
In Argot the translation is that for any two sets x, x0 there is a set that only has
them as elements.
Next for any formula P (x) in the language of sets, having a free variable x, we
introduce the following:
Axiom 7.12. (Separation axiom)
∀y∃z∀x((x ∈ z) ↔ ((x ∈ y) ∧ (P (x))).
The translation in Argot is that for any set y there is a set z whose elements are
all the elements x of y such that P (x).
Axiom 7.13. (Extensionality axiom)
∀u∀v((u = v) ↔ ∀x((x ∈ u) ↔ (x ∈ v))).
The translation in Argot is that two sets u and v are equal if and only if they have
the same elements.
Axiom 7.14. (Union axiom)
∀w∃u∀x((x ∈ u) ↔ (∃t((t ∈ w) ∧ (x ∈ t)))).
The translation in Argot is that for any set w there exists a set u such that for
any x we have that x is an element of u if and only if x is an element of one of the
elements of w.
Axiom 7.15. (Empty set axiom)
∃x∀y(y 6∈ x).
The translation in Argot is that there exists a set that has no elements.
Axiom 7.16. (Power set axiom)
∀y∃z∀x((x ∈ z) ↔ (∀u((u ∈ x) → (u ∈ y)))).
The translation in Argot is that for any set y there is a set z such that a set x is
an element of z if and only if all elements of x are elements of y.
For simplicity the rest of the axioms will be formulated in Argot only.
Definition 7.17. Two sets are disjoint if they have no element in common.
The elements of a set are pairwise disjoint if any two elements are disjoint.
Axiom 7.18. (Axiom of choice) For any set w whose elements are pairwise
disjoint sets there is a set that has exactly one element in common with each of the
sets in w.
Axiom 7.19. (Axiom of infinity) There exists a set x such that x contains some
element u and such that for any y ∈ x there exists z ∈ x with the property that y is
the only element of z. Intuitively this axiom guarantees the existence of “infinite”
sets.
Axiom 7.20. (Axiom of foundation) For any set x there exists y ∈ x such that
x and y are disjoint.
52 7. ZFC

One finally adds a technical list of axioms (indexed by formulas P (x, y, z))
about the “images of maps with parameters z”:
Axiom 7.21. (Axiom of replacement) If for any z and any u we have that
P (x, y, z) “defines y as a function of x ∈ u” (i.e., for any x ∈ u there exists a unique
y such that P (x, y, z)) then for all z there is a set v which is the “image of this
map” (i.e., v consists of all y’s with the property that there is an x ∈ u such that
P (x, y, z)). Here x, z may be tuples of variables.
Exercise 7.22. Write the axioms of choice, infinity, foundation, and replace-
ment in the language of sets.
Metadefinition 7.23. All of the above axioms form the ZFC system of axioms
(Zermelo-Fraenkel+Choice). Set theory Tset is the theory in Lset with ZF C axioms.
Unless otherwise specified all theorems in the rest of the course are understood to
be theorems in Tset . By abuse of terminology we continue to denote by Tset any
development of Tset .
Remark 7.24. Note the important fact that the axioms did not involve con-
stants. In the next chapter we investigate the constants, i.e., the sets.

From this moment on all proofs in this course will be written in

Argot. Also, unless otherwise stated, all proofs required to be given in
the exercises must be written in Argot.
 Part 2

Set theory
 CHAPTER 8

Sets

We will start here our discussion of sets and prove our first theorems in set
theory. Recall that we introduced mathematics/set theory as being a specific theory
Tset in the language Lset , with axioms ZF C described in the last chapter.
Recall the following:
Metadefinition 8.1. A set is a constant in the language of set theory.
Sets will be denoted by a, b, ..., A, B, ..., A, B, ..., α, β, γ, ....
In what follows all definitions will be definitions in the language Lset of sets.
Sometimes definitions are given in Argotic Lset . Recall that some definitions intro-
ducing new constants are also simply referred to as notation.
We start by defining a new constant ∅ (called the empty set) as being equal to
the witness for the axiom ∃x∀y(y 6∈ x); in other words ∅ is defined by
Definition 8.2. ∅ = c∀y(y6∈x) .
Note that ∀y(y 6∈ ∅) is a theorem. In Argot we say that ∅ is the “unique” set
that contains no element.
Next if a is a set we introduce a new constant {a} defined to be the witness for
the sentence ∃yP where
P equals “(a ∈ y) ∧ (∀z((z ∈ y) → (z = a))).”
In other words {a} is defined by
Definition 8.3. {a} = cP = c(a∈y)∧(∀z((z∈y)→(z=a))) .
The sentence ∃yP is a theorem (use the singleton axiom) so the following is a
theorem:
(a ∈ {a}) ∧ (∀z((z ∈ {a}) → (z = a))).
We can say (and we will usually say, by abuse of terminology) that {a} is “the
unique” set containing a only among its elements; we will often use this kind of
abuse of terminology. In particular {{a}} denotes the set whose only element is
the set {a}, etc. Similarly, for any two sets a, b with a 6= b denote by {a, b} the set
that only has a and b as elements; the set {a, b} is a witness for a theorem that
follows from the unordered pair axiom. Also whenever we write {a, b} we implicitly
understand that a 6= b.
Next, for any set A and any formula P (x) in the language of sets, having one
free variable x we denote by A(P ) or {a ∈ A; P (a)} or {x ∈ A; P (x)} the set whose
elements are the elements a ∈ A such that P (a); so the set A(P ) equals by definition
the witness for the separation axiom that corresponds to A and P . More precisely
we have:
Definition 8.4. A(P ) = {x ∈ A; P (x)} = c∃z∀x((x∈z)↔(x∈A)∧P (x))) .
55
56 8. SETS

Lemma 8.5. If A = {a} and B = {b, c} then A 6= B.

Proof. We proceed by contradiction. So assume A = {a}, B = {b, c}, and A =
B and seek a contradiction. Indeed since a ∈ A and A = B, by the extensionality
axiom we get a ∈ B. Hence a = b or a = c. Assume a = b and seek a contradiction.
(In the same way we get a contradiction by assuming a = c.) Since a = b we get
B = {a, c}. Since c ∈ B and A = B, by the extensionality axiom we get c ∈ A. So
c = a. Since a = b we get b = c. But by our notation for sets (elements listed are
distinct) we have b 6= c, a contradiction. 
Exercise 8.6. Prove that:
1) If {a} = {b} then a = b.
2) {a, b} = {b, a}.
3) {a} = {x ∈ {a, b}; x 6= b}.
4) There is a set b whose only elements are {a} and {a, {a}}; so
b = {{a}, {a, {a}}}.
To make our definitions (notation) more reader friendly we will begin to express
them in metalanguage as in the following example.
Definition 8.7. For any two sets A and B we define the set A ∪ B (called the
union of A and B) as the set such that for all c, c ∈ A ∪ B if and only if c ∈ A or
c ∈ B; the set A ∪ B is a witness for the union axiom.
Exercise 8.8. Explain in detail the definition of A∪B using a witness notation
as in 8.2, 8.3, 8.4.
Definition 8.9. The difference between the set A and the set B is the set
A\B = {c ∈ A; c 6∈ B}.
Definition 8.10. The intersection of the sets A and B is the set
A ∩ B = {c ∈ A; c ∈ B}.
Exercise 8.11. Prove that
1) A(P ∧ Q) = A(P ) ∩ A(Q);
2) A(P ∨ Q) = A(P ) ∪ A(Q);
3) A(¬P ) = A\A(P ).
Exercise 8.12. Prove that if a, b, c are such that (a 6= b) ∧ (b 6= c) ∧ (a 6= c)
then there is a set denoted by {a, b, c} whose only elements are a, b, c; in other words
prove the following sentence:
∀x∀x0 ∀x00 ∃y((x ∈ y)∧(x0 ∈ y)∧(x00 ∈ y)∧(z ∈ y) → ((z = x)∨(z = x0 )∨(z = x00 )))
Hint: Use the singleton axiom, the unordered pair axiom, and the union axiom,
applied to the set {{a}, {b, c}}.
Definition 8.13. Similarly one defines sets {a, b, c, d}, etc. Whenever we write
{a, b, c} or {a, b, c, d}, etc., we imply that the elements in each set are pairwise
unequal (pairwise distinct). Also denote by
{a, b, c, ...}
any set d such that
(a ∈ d) ∧ (b ∈ d) ∧ (c ∈ d)
 8. SETS 57

So the dots indicate that there may be other elements in d other than a, b, c; also
note that when we write {a, b, c, ...} we implicitly imply that a, b, c are pairwise
distinct.
Exercise 8.14.
1) Prove that {∅} =
6 ∅.
2) Prove that {{∅}} =
6 {∅}.
Exercise 8.15. Prove that A = B if and only if A ⊂ B and B ⊂ A.
Exercise 8.16. Prove that:
1) {a, b, c} = {b, c, a}.
2) {a, b} =6 {a, b, c}. Hint: Use c 6= a, c 6= b.
3) {a, b, c} = {a, b, d} if and only if c = d.
Exercise 8.17. Let A = {a, b, c} and B = {c, d}. Prove that
1) A ∪ B = {a, b, c, d},
2) A ∩ B = {c}, A\B = {a, b}.
Exercise 8.18. Let A = {a, b, c, d, e, f }, B = {d, e, f, g, h}. Compute
1) A ∩ B,
2) A ∪ B,
3) A\B,
4) B\A,
5) (A\B) ∪ (B\A).
Exercise 8.19. Prove the following:
1) A ∩ B ⊂ A,
2) A ⊂ A ∪ B,
3) A ∩ (B ∪ C) = (A ∩ B) ∪ (A ∩ C),
4) A ∪ (B ∩ C) = (A ∪ B) ∩ (A ∪ C),
5) (A\B) ∩ (B\A) − ∅.
Definition 8.20. For any set A we define the set P(A) as the set whose ele-
ments are the subsets of A; we call P(A) the power set of A; P(A) is a witness for
(a theorem obtained from) the power set axiom.
Exercise 8.21. Explain in detail the definition of P(A) (using the witness
notation).
Example 8.22. If A = {a, b, c} then
P(A) = {∅, {a}, {b}, {c}, {a, b}, {a, c}, {b, c}, {a, b, c}}.
Exercise 8.23. Let A = {a, b, c, d}. Write down the set P(A).
Exercise 8.24. Let A = {a, b}. Write down the set P(P(A)).
Definition 8.25. (Ordered pairs) Let A and B be sets and let a ∈ A, b ∈ B.
If a 6= b the ordered pair (a, b) is the set {{a}, {a, b}}. We sometimes say “pair”
instead of “ordered pair.” If a = b the pair (a, b) = (a, a) is the set {{a}}. Note
that (a, b) ∈ P(P(A ∪ B)).
Definition 8.26. For any sets A and B we define the product of A and B as
the set
A × B = {c ∈ P(P(A ∪ B)); ∃x∃y((x ∈ A) ∧ (y ∈ B) ∧ (c = (a, b)))}.
This is the set whose elements are exactly the pairs (a, b) with a ∈ A and b ∈ B.
58 8. SETS

Proposition 8.27. (a, b) = (c, d) if and only if a = c and b = d.

Proof. We need to prove that
1) If a = c and b = d then (a, b) = (c, d) and
2) If (a, b) = (c, d) then a = c and b = d.
Now 1) is obvious. To prove 2) assume (a, b) = (c, d).
Assume first a 6= b and c 6= d. Then by the definition of pairs we know that
{{a}, {a, b}} = {{c}, {c, d}}.
Since {a} ∈ {{a}, {a, b}} it follows (by the extensionality axiom) that {a} ∈
{{c}, {c, d}}. Hence either {a} = {c} or {a} = {c, d}. But as seen before {a} 6=
{c, d}. So {a} = {c}. Since a ∈ {a} it follows that a ∈ {c} hence a = c. Similarly
since {a, b} ∈ {{a}, {a, b}} we get {a, b} ∈ {{c}, {c, d}}. So either {a, b} = {c} or
{a, b} = {c, d}. Again as seen before {a, b} =
6 {c} so {a, b} = {c, d}. So b ∈ {c, d}.
So b = c or b = d. Since a 6= b and a = c we get b 6= c. Hence b = d and we are
done in case a 6= b and c 6= d.
Assume next a = b and c = d. Then by the definition of pairs in this case we
have {{a}} = {{c}} and as before this implies {a} = {c} hence a = c so we are
done in this case as well.
Finally assume a = b and c 6= d. (The case a 6= b and c = d is treated similarly.)
By the definition of pairs we get
{{a}} = {{c}, {c, d}}.
We get {c, d} ∈ {{a}}. Hence {c, d} = {a} which is impossible, as seen before. This
ends the proof. 
Exercise 8.28. If A = {a, b, c} and B = {c, d} then
A × B = {(a, c), (a, d), (b, c), (b, d), (c, c), (c, d)}.
Hint: By the above Proposition the pairs are distinct.
Exercise 8.29. Prove that
1) (A ∩ B) × C = (A × C) ∩ (B × C),
2) (A ∪ B) × C = (A × C) ∪ (B × C).
Exercise 8.30. Prove that
¬(∃y∀z(z ∈ y))
In Argot this says that there does not exist a set T such that for any set A we have
A ∈ T . (Intuitively there is no set such that all sets belong to it.)
Hint: Assume there is such a T and derive a contradiction. To derive a contra-
diction consider the set
S = {x ∈ T ; x 6∈ x}.
There are two possibilities. First possibility is S 6∈ S. Since S ∈ T , we get that
S ∈ S, a contradiction. The second possibility is that S ∈ S. Since S ∈ T , we get
that S 6∈ S, which is again a contradiction.
Remark 8.31. Before the advent of ZF C Russell showed, using the set S
above, that Cantor’s set theory leads to a contradiction; this is referred to as the
“Russell paradox.” Within ZF C Russell’s paradox, in its original form, disappears.
Whether there are other forms of this paradox, or similar paradoxes, that survive
in ZF C it is not clear.
 CHAPTER 9

Maps

The concept of map (or function) has a long history. Originally functions were
understood to be given by more or less explicit “formulae” (polynomial, ratio-
nal, algebraic, and later by series). Controversies around what the “most general”
functions should be arose, for instance, in connection with solving partial differ-
ential equations (by means of trigonometric series); this is somewhat parallel to
the controversy around what the “most general” numbers should be that arose in
connection with solving algebraic equations (such as x2 = 2, x2 = −1, or higher
degree equations with no solutions expressed by radicals, etc.). The notion of “com-
pletely arbitrary” function gradually arose through the work of Dirichlet, Riemann,
Weierstrass, Cantor, etc. Here is the definition:
Definition 9.1. A map (or function) from a set A to a set B is a subset
F ⊂ A × B such that for any a ∈ A there is a unique b ∈ B with (a, b) ∈ F . If
(a, b) ∈ F we write F (a) = b or a 7→ b or a 7→ F (a). We also write F : A → B or
F
A → B.
Remark 9.2. The above defines a new (ternary) predicate µ equal to “...is a
map from ... to ....” Also we may introduce a new functional symbol Fb by
∀x∀y((µ(F, A, B) ∧ (x ∈ A) ∧ (y ∈ B)) → ((Fb(x) = y) ↔ ((x, y) ∈ F ))).
Here F, A, B can be constants or variables depending on the context. We will
usually drop the b (or think of the Argot translation as dropping the hat). Also
note that what we call a map F ⊂ A × B corresponds to what in elementary
mathematics is called the graph of a map.
Example 9.3. The set
(9.1) F = {(a, a), (b, c)} ⊂ {a, b} × {a, b, c}
is a map and F (a) = a, F (b) = c. On the other hand the subset
F = {(a, b), (a, c)} ⊂ {a, b} × {a, b, c}
is not a map.
Definition 9.4. For any A the identity map I : A → A is defined as I(a) = a,
i.e.,
I = IA = {(a, a); a ∈ A} ⊂ A × A.
Definition 9.5. A map F : A → B is injective (or an injection, or one-to-one)
if F (a) = F (c) implies a = c.
Definition 9.6. A map F : A → B is surjective (or a surjection, or onto) if
for any b ∈ B there exists an a ∈ A such that F (a) = b.
59
60 9. MAPS

Example 9.7. The map (9.1) is injective and not surjective.

Exercise 9.8. Give an example of a map which is surjective and not injective.
Exercise 9.9. Let A ⊂ B. Prove that there is an injective map i : A → B
such that i(a) = a for all a ∈ A. We call i the inclusion map; we sometimes say
A ⊂ B is the inclusion map.
Exercise 9.10. (Composition) Prove that if F : A → B and G : B → C are
two maps then there exists a unique map H : A → C such that H(a) = G(F (a))
for all a. We write H = G ◦ F and call the latter the composition of G with F .
Hint: We let (a, c) ∈ H if and only if there exists b ∈ B with (a, b) ∈ F , (b, c) ∈ G.
Definition 9.11. (Restriction) If F : A → B is a map and A0 ⊂ A then the
composition of F with the inclusion map A0 ⊂ A is called the restriction of F to
A0 and is denoted by F|A0 : A0 → B.
Definition 9.12. (Commutative diagram) By a commutative diagram of sets
F
A → B
U↓ ↓V
G
C → D
we mean a collection of sets and maps as above with the property that G◦U = V ◦F .
Exercise 9.13. Prove that if F ◦ G is surjective then F is surjective. Prove
that if F ◦ G is injective then G is injective.
Exercise 9.14. Prove that the composition of two injective maps is injective
and the composition of two surjective maps is surjective.
Definition 9.15. A map is bijective (or a bijection) if it is injective and sur-
jective.
Here is a fundamental theorem in set theory:
Theorem 9.16. (Bernstein’s Theorem) If A and B are sets and if there exist
injective maps F : A → B and G : B → A then there exists a bijective map
H : A → B.
The reader may attempt to prove this after he/she gets to the chapter on
induction.
Exercise 9.17. Prove that if F : A → B is bijective then there exists a unique
bijective map denoted by F −1 : B → A such that F ◦ F −1 = IB and F −1 ◦ F = IA .
F −1 is called the inverse of F .
Exercise 9.18. Let F : {a, b, c} → {c, d, e}, F (a) = d, F (b) = c, F (c) = e.
Prove that F has an inverse and compute F −1 .
Exercise 9.19. Prove that if A and B are sets then there exist maps F :
A × B → A and G : A × B → B such that F (a, b) = a and G(a, b) = b for all
(a, b) ∈ A × B. (These are called the first and the second projection.) Hint: For G
show that G = {((a, b), c); c = b} ⊂ (A × B) × B is a map.
Exercise 9.20. Prove that (A × B) × C → A × (B × C), ((a, b), c) 7→ (a, (b, c))
is a bijection.
 9. MAPS 61

Definition 9.21. Write A × B × C instead of (A × B) × C and write (a, b, c)

instead of ((a, b), c). We call (a, b, c) a triple. Write A2 = A×A and A3 = A×A×A.
More generally adopt this notation for arbitrary number of factors. Elements like
(a, b), (a, b, c), (a, b, c, d), etc. will be called tuples.
Theorem 9.22. If A is a set then there is no bijection between A and P(A)
Proof. Assume there exists a bijection F : A → P(A) and seek a contradiction.
Consider the set
B = {a ∈ A; a 6∈ F (a)} ∈ P(A).
Since F is subjective there exists b ∈ A such that B = F (b). There are two cases:
either b ∈ B or b 6∈ B. If b ∈ B then b ∈ F (b) so b 6∈ B, a contradiction. If b 6∈ B
then b 6∈ F (b) so b ∈ B, a contradiction, and we are done. 
Remark 9.23. Note the similarity between the above argument and the argu-
ment showing that there is no set having all sets as elements (the “Russell para-
dox”).
Definition 9.24. Let S be a set of sets and I a set. A family of sets in S
indexed by I is a map I → S, i 7→ Ai . We sometimes drop the reference to S. We
also write (Ai )i∈I to denote S this family. By the union axiom for any such family
there is a Sset (denoted by i∈I Ai , called their union) such that for all x we have
that x ∈ i∈I T Ai if and only if there exists i ∈ I such that x ∈ Ai . Also a set
(denoted T by i∈I Ai , called their intersection) exists such that for all x we have
that x ∈ i∈I Ai if and only S if for all i ∈ I we have x ∈ Ai . A family of elements
in (Ai )i∈I is a map I → i∈I Ai , i 7→ ai , such that for all i ∈ I we have aQi ∈ Ai .
Such a family of elements is denoted by (ai )i∈I . One defines the product i∈I Ai
as the set of all families of elements (ai )i∈I .
Q
Exercise 9.25. Check that for I = {i, j} the above definitions of ∪, ∩, yield
the usual definition of Ai ∩ Aj , Ai ∩ Aj , and Ai × Aj .
Definition 9.26. Let F : A → B be a map and X ⊂ A. Define the image of
X as the set
F (X) = {y ∈ B; ∃x ∈ X, y = F (x)} ⊂ B.
If Y ⊂ B define the inverse image (or preimage) of Y as the set
F −1 (Y ) = {x ∈ A; F (x) ∈ Y } ⊂ A.
For y ∈ B define
F −1 (y) = {x ∈ A; F (x) = y}.
(Note that F −1 (Y ), F −1 (y) are defined even if the inverse map F −1 does not exist,
i.e., even if F is not bijective.)
Exercise 9.27. Let F : {a, b, c, d, e, f, g} → {c, d, e, h}, F (a) = d, F (b) = c,
F (c) = e, F (d) = c, F (e) = d, F (f ) = c, F (g) = c. Let X = {a, b, c}, Y = {c, h}.
Compute F (X), F −1 (Y ), F −1 (c), F −1 (h).
Exercise 9.28. Prove that if F : A → B is a map and X ⊂ X 0 ⊂ A are subsets
then F (X) ⊂ F (X 0 ).
Exercise 9.29. Prove that if F : A → B is a map and (Xi )i∈I is a family of
subsets of A then
F (∪i∈I Xi ) = ∪i∈I F (Xi ),
62 9. MAPS

F (∩i∈I Xi ) ⊂ ∩i∈I F (Xi ).

If in addition F is injective show that
F (∩i∈I Xi ) = ∩i∈I F (Xi ).
Give an example showing that the latter may fail if F is not injective.
Exercise 9.30. Prove that if F : A → B is a map and Y ⊂ Y 0 ⊂ B are subsets
then F −1 (Y ) ⊂ F −1 (Y 0 ).
Exercise 9.31. Prove that if F : A → B is a map and (Yi )i∈I is a family of
subsets of B then
F −1 (∪i∈I Yi ) = ∪i∈I F −1 (Yi ),
F −1 (∩i∈I Yi ) = ∩i∈I F −1 (Yi ).
(So here one does not need injectivity like in the case of unions.)
Definition 9.32. If A and B are sets we denote by B A ⊂ P(A × B) the set of
all maps F : A → B; sometimes one writes M ap(A, B) = B A .
Exercise 9.33. Let 0, 1 be two elements. Prove that the map {0, 1}A → P(A)
sending F : A → {0, 1} into F −1 (1) ∈ P(A) is a bijection.
Exercise 9.34. Find a bijection between (C B )A and C A×B . Hint: Send F ∈
(C ) , F : A → C B , into the set (map)
B A

{((a, b), c) ∈ (A × B) × C; (b, c) ∈ F (a)}.

 CHAPTER 10

Relations

A basic notion in set theory is that of relation; we shall investigate in some

detail two special cases: order relations and equivalence relations.
Definition 10.1. If A is a set then a relation on A is a subset R ⊂ A × A. If
(a, b) ∈ R we write aRb.
Remark 10.2. Exactly as in Remark 9.2 the above defines a new (binary)
predicate “... is a relation on ...” and we may introduce a corresponding new
predicate (still denoted by R).
Definition 10.3. A relation R is called an order if (writing a ≤ b instead of
aRb) we have, for all a, b, c ∈ A, that
1) a ≤ a (reflexivity),
2) a ≤ b and b ≤ c imply a ≤ c (transitivity),
3) a ≤ b and b ≤ a imply a = b (antisymmetry).
Definition 10.4. One writes a < b if a ≤ b and a 6= b.
Definition 10.5. An order relation is called a total order if for any a, b ∈ A
either a ≤ b or b ≤ a. Alternatively we say A is totally ordered (by ≤).
Example 10.6. For instance if A = {a, b, c, d} then
R = {(a, a), (b, b), (c, c), (d, d), (a, b), (b, c), (a, c)}
is an order but not a total order.
Exercise 10.7. Let R0 ⊂ A × A be a relation and assume R0 is contained in
an order relation R1 ⊂ A × A. Let
\
R= R0
R0 ⊃R0

be the intersection of all order relations R0 containing R0 . Prove that R is an order

relation and it is the smallest order relation containing R0 in the sense that it is
contained in any order relation that contains R0 .
Exercise 10.8. Let A = {a, b, c, d, e} and R0 = {(a, b), (b, c), (c, d), (c, e)}.
Find an order relation containing R0 . Find the smallest order relation R containing
R0 . Show that R is not a total order.
Exercise 10.9. Let A be a set. For any subsets X ⊂ A and Y ⊂ A write
X ≤ Y if and only if X ⊂ Y . This defines a relation on the set P(A). Prove that
this is an order relation. Give an example showing that this is not in general a total
order.
63
64 10. RELATIONS

Definition 10.10. An ordered set is a pair (A, ≤) where A is a set and ≤ is

an order relation on A.
Definition 10.11. Let (A, ≤) and (A0 , ≤0 ) be ordered sets. A map F : A → A0
is called increasing if for any a, b ∈ A with a ≤ b we have F (a) ≤0 F (b).
Exercise 10.12. Prove that if (A, ≤), (A0 , ≤0 ), (A00 , ≤00 ) are ordered sets and
G : A → A0 , F : A0 → A00 are increasing then F ◦ G : A → A00 is increasing.
Definition 10.13. Let A be a set with an order ≤. We say α ∈ A is a minimal
element of A if for all a ∈ A such that a ≤ α we must have a = α.
Definition 10.14. Let A be a set with an order ≤. We say β ∈ A is a maximal
element of A if for all b ∈ A such that β ≤ b we must have β = b.
Definition 10.15. Let A be a set with an order ≤ and let B ⊂ A. We say
m ∈ B is a minimum element of B if for all b ∈ B we have m ≤ b. If a minimum
element exists it is unique (check!) and we denote it by min B. Note that if min B
exists then, by definition, min B belongs to B.
Definition 10.16. Let A be a set with an order ≤ and B ⊂ A. We say M ∈ B
is a maximum element of B if for all b ∈ B we have b ≤ M . If a maximum element
exists it is unique and we denote it by max B. Again, if max B exists then by
definition it belongs to B.
Definition 10.17. Let A be a set with an order ≤ and let B ⊂ A. An element
u ∈ A is called an upper bound for B if b ≤ u for all b ∈ B. We also say that B
is bounded from above by b. An element l ∈ A is called a lower bound for B if
l ≤ b for all b ∈ B; we also say B is bounded from below by l. If the set of upper
bounds of B has a minimum element we call it the supremum of B and we denote
it by sup B; if the set of lower bounds of B has a maximum element we call it the
infimum of B and we denote it by inf B. (Note that if one of sup B and inf B
exists that element is by definition in A but does not necessarily belong to B.) We
say B is bounded if it has both an upper bound and a lower bound.
Exercise 10.18. Consider the set A and the order ≤ defined by the relation
R in Exercise 10.8. Does A have a maximum element? Does A have a minimum
element? Are there maximal elements in A? Are there minimal elements in A?
List all these elements in case they exist. Let B = {b, c}. Is B bounded? List all
the upper bounds of B. List all the lower bounds of B. Does the supremum of B
exist? If yes does it belong to B? Does the infimum of B exist? Does it belong to
B?
Definition 10.19. A well ordered set is an ordered set (A, ≤) such that any
non-empty subset B ⊂ A has a minimum element.
Exercise 10.20. Prove that any well ordered set is totally ordered.
Remark 10.21. Later, when we will have introduced the ordered set of integers
and the ordered set of rational numbers we will see that the non-negative integers
are well ordered but the non-negative rationals are not well ordered.
The following theorems can be proved (but their proof is beyond the scope of
this course):
 10. RELATIONS 65

Theorem 10.22. (Zorn’s lemma) Assume (A, ≤) is an ordered set. Assume

that any non-empty totally ordered subset B ⊂ A has an upper bound in A. Then
A has a maximal element.
Theorem 10.23. (Well ordering principle) Let A be a set. Then there exists
an order relation ≤ on A such that (A, ≤) is well ordered.
Remark 10.24. It can be proved that if one removes from the axioms of set
theory the axiom of choice then the axiom of choice, Zorn’s lemma, and the well
ordering principle are all equivalent.
Exercise 10.25. Let (A, ≤) and (B, ≤) be totally ordered sets. Define a rela-
tion ≤ on A × B by
((a, b) ≤ (a0 , b0 )) ↔ ((a < a0 ) ∨ ((a = a0 ) ∧ (b ≤ b0 ))).
Prove that ≤ is an order on A × B (it is called the lexicographic order) and that
(A × B, ≤) is totally ordered. (Explain how this order is being used to order words
in a dictionary.)
Definition 10.26. A relation R is called an equivalence relation if (writing
a ∼ b instead of aRb) we have, for all a, b, c ∈ A, that
1) a ∼ a (reflexivity),
2) a ∼ b and b ∼ c imply a ∼ c (transitivity),
3) a ∼ b implies b ∼ a (symmetry);
we also say that ∼ is an equivalence relation.
Exercise 10.27. Let R0 ⊂ A × A be a relation and let
\
R= R0
R0 ⊃R0

be the intersection of all equivalence relations R0 containing R0 . Prove that R is

an equivalence relation and it is the smallest equivalence relation containing R0 in
the sense that it is contained in any other equivalence relation that contains R0 .
Definition 10.28. Given an equivalence relation ∼ as above for any a ∈ A we
may consider the set
a = {c ∈ A; c ∼ a}
b
called the equivalence class of a.
a, one writes a or [a].
Definition 10.29. Sometimes, instead of b

a = bb if and only if a ∼ b.
Exercise 10.30. Prove that b
Exercise 10.31. Prove that:
a ∩Sbb 6= ∅ then b
1) if b a = bb;
2) A = a∈A b a.
Definition 10.32. If A is a set a partition of A is a family (Ai )i∈I if subsets
Ai ⊂ A such that:
1) if i 6=Sj then Ai ∩ Aj = ∅
2) A = i∈I Ai .
66 10. RELATIONS

Exercise 10.33. Let A be a set and ∼ an equivalence relation on it. Prove

that:
1) There exists a subset B ⊂ A which contains exactly one element of each
equivalence class (such a set is called a system of representatives. Hint: Use the
axiom of choice).
2) The family (bb)b∈B is a partition of A.
Exercise 10.34. Let A be a set and (Ai )i∈I a partition of A. Define a relation
R on A as follows:
R = {(a, b) ∈ A × A; ∃i((i ∈ I) ∧ (a ∈ Ai ) ∧ (b ∈ Ai ))}.
Prove that R is an equivalence relation.
Exercise 10.35. Let A be a set. Prove that there is a bijection between the
set of equivalence relations on A and the set of partitions of A. Hint: Use the above
two exercises.
Definition 10.36. The set of equivalence classes
{α ∈ P(A); ∃a((a ∈ A) ∧ (α = b
a)}
is denoted by A/ ∼ and is called the quotient of A by the relation ∼.
Example 10.37. For instance if A = {a, b, c} and
R = {(a, a), (b, b), (c, c), (a, b), (b, a)}
a = bb = {a, b}, b
then R is an equivalence relation, b c = {c}, and A/ ∼= {{a, b}, {c}}.
Exercise 10.38. Let A = {a, b, c, d, e, f } and R0 = {(a, b), (b, c), (d, e)}. Find
the smallest equivalence relation R containing R0 . Call it ∼. Write down the
equivalence classes b
a, bb, b b eb, fb. Write down the set A/ ∼.
c, d,
Exercise 10.39. Let S be a set. For any sets X, Y ∈ S write X ∼ Y if and
only if there exists a bijection F : X → Y . This defines a relation on S. Prove that
this is an equivalence relation.
Exercise 10.40. Let S = {A, B, C, D}, A = {a, b}, B = {b, c}, C = {x, y},
D = ∅. Let ∼ be the equivalence relation on S defined in the previous exercise.
Write down the equivalence classes A,
b B,
b C,
b Db and write down the set S/ ∼.

Definition 10.41. An affine plane is a pair (A, L) where A is a set and L ⊂

P(A) is a set of subsets of A satisfying a series of properties (which we call, by abuse,
axioms) which we now explain. It is convenient to introduce some terminology as
follows. A is called the affine plane. The elements of A are called points. The
elements L of L are called lines; so each such L is a subset of A. We say a point
P lies on a line L if P ∈ L; we also say that L passes through P . We say that two
lines intersect if they have a point in common; we say that two lines are parallel if
they either coincide or they do not intersect. We say that 3 points are collinear if
they lie on the same line. Here are the axioms that we impose:
1) There exist 3 points which are not collinear and any line has at least 2 points.
2) Any 2 distinct points lie on exactly one line.
3) If L is a line and P is a point not lying on L there exists exactly one line
through P which is parallel to L.
 10. RELATIONS 67

Remark 10.42. Note that we have not defined 2 or 3 yet; this will be done
later when we introduce integers. The meaning of these axioms is, however, clearly
expressible in terms that were already defined. For instance axiom 2 says that for
any points P and Q with P 6= Q there exists a line through P and Q; we do not
need to define the symbol 2 to express this. The same holds for the use of the
symbol 3.
Exercise 10.43. Prove that any two distinct non-parallel lines intersect in
exactly one point.
Exercise 10.44. Let A = {a, b} × {a, b} and let L ⊂ P(A) consist of all subsets
of 2 elements; there are 6 of them. Prove that (A, L) is an affine plane. (Again
one can reformulate everything without reference to the symbols 2 or 6; one simply
uses 2 or 6 letters and writes that they are pairwise unequal.)
Exercise 10.45. Let A = {a, b, c} × {a, b, c} . Find all subsets L ⊂ P(A) such
that (A, L) is an affine plane. (This is tedious !)
Definition 10.46. A projective plane is a pair (A, L) where A is a set and
L ⊂ P(A) is a set of subsets of A satisfying a series of axioms which we now
explain. Again it is convenient to introduce some terminology as follows. A is
called the projective plane. The elements of A are called points, P . The elements
L of L are called lines; so each such L ⊂ A. We say a point P lies on a line L if
P ∈ L; we also say that L passes through P . We say that two lines intersect if they
have a point in common; we say that two lines are parallel if they either coincide
or they do not intersect. We say that 3 points are collinear if they lie on the same
line. Here are the axioms that we impose:
1) There exist 3 points which are not collinear and any line has at least 3 points.
2) Any 2 distinct points lie on exactly one line.
3) Any 2 distinct lines meet in exactly one point.
Example 10.47. One can attach to any affine plane (A, L) a projective plane
(A, L) as follows. We introduce the relation k on L by letting L k L0 if and only
if L and L0 are parallel. This is an equivalence relation (check!). Denote by L
b the
equivalence class of L. Then we consider the set of equivalence classes, L∞ = L/ k;
call this set the line at infinity. There exists a set A such that A = A ∪ L∞ and
A ∩ L∞ = ∅. Define a line in A to be either L∞ or set of the form L = L ∪ {L}. b
Finally define L to be the set of all lines in A.
Exercise 10.48. Explain why A exists. Check that (A, L) is a projective plane.
Exercise 10.49. Describe the projective plane attached to the affine plane in
Exercise 10.44; how many points does it have? How many lines?
 CHAPTER 11

Operations

The concept of operation on a set is an abstraction of “familiar” operations

such as addition and multiplication of numbers, composition of functions, etc. Sets
with operations on them will be referred to as algebraic structures. The study of
algebraic structures is referred to as (modern) algebra and took the shape known
today through work (in number theory and algebraic geometry) done by Kronecker,
Dedekind, Hilbert, Emmy Noether, etc. Here we introduce operations in general,
and some algebraic structures such as rings, fields, and Boolean algebras. We prefer
to postpone the introduction of other algebraic structures such as groups, vector
spaces, etc., until more theory is being developed.
Definition 11.1. A binary operation ? on a set A is a map ? : A × A → A,
(a, b) 7→ ?(a, b). We usually write a ? b instead of ?(a, b). For instance, we write
(a ? b) ? c instead of ?(?(a, b), c). Instead of ? we sometimes use notation like
+, ×, ◦, ....
Remark 11.2. Exactly as in Remark 9.2 the above defines a new (binary)
predicate “... is a binary operation on ...” and we may introduce a corresponding
new functional symbol (still denoted by ?).
Definition 11.3. A unary operation 0 on a set A is a map 0 : A → A, a 7→ 0 (a).
We usually write a0 or 0 a instead of 0 (a). Instead of 0 we sometimes use notation
like −, i, ....
Example 11.4. Let S = {0, 1} where 0, 1 are two sets. Then there are 3 inter-
esting binary operations on S denoted by ∧, ∨, + (and called supremum, infimum,
and addition) defined as follows:
0 ∧ 0 = 0, 0 ∧ 1 = 0, 1 ∧ 0 = 0, 1 ∧ 1 = 1;
0 ∨ 0 = 0, 0 ∨ 1 = 1, 1 ∨ 0 = 1, 1 ∨ 1 = 1;
0 + 0 = 0, 0 + 1 = 1, 1 + 0 = 1, 1 + 1 = 0.
The symbol ∧ is also denoted by × or ·; it is referred to as multiplication. The
symbol + is also denoted by ∆. Also there is a unary operation ¬ on S defined by
¬1 = 0, ¬1 = 0.
Note that if we denote 0 and 1 by F and T then the operations ∧, ∨, ¬ on {0, 1}
correspond exactly to the “logical operations” on F and T defined in the chapter
on tautologies. This is not a coincidence!
Exercise 11.5. Compute ((0 ∧ 1) ∨ 1) + (1 ∧ (0 ∨ (1 + 1))).
Definition 11.6. A Boolean algebra is a tuple
(A, ∨, ∧, ¬, 0, 1)
69
70 11. OPERATIONS

where ∧, ∨ are binary operations, ¬ is a unary operation, and 0, 1 ∈ A such that

for all a, b, c ∈ A the following “axioms” are satisfied:
1) a ∧ (b ∧ c) = (a ∧ b) ∧ c, a ∨ (b ∨ c) = (a ∨ b) ∨ c,
2) a ∧ b = b ∧ a, a ∨ b = b ∨ a,
3) a ∧ 1 = a, a ∨ 0 = a,
4) a ∧ (b ∨ c) = (a ∧ b) ∨ (a ∧ c), a ∨ (b ∧ c) = (a ∨ b) ∧ (a ∨ c)
5) a ∧ (¬a) = 0, a ∨ (¬a) = 1.
Definition 11.7. A commutative unital ring (or simply a ring) is a tuple
(R, +, ×, −, 0, 1)
(sometimes referred to simply as R) where R is a set, 0, 1 ∈ R, +, × are two binary
operations (write a × b = ab), and − is a unary operation on R such that for any
a, b, c ∈ R the following hold:
1) a + (b + c) = (a + b) + c, a + 0 = a, a + (−a) = 0, a + b = b + a;
2) a(bc) = a(bc), 1a = a, ab = ba,
3) a(b + c) = ab + ac.
The element 1 is referred to as the identity; 0 is referred to as the zero element.
Definition 11.8. We write a + b + c instead of (a + b) + c and abc for (ab)c.
We write a − b instead of a + (−b).
Definition 11.9. An element a of a ring R is invertible if there exists a0 ∈ R
such that aa0 = 1; this a0 is then easily proved to be unique. It is called the inverse
of a, and is denoted by a−1 . A ring R is called a field if 0 6= 1 and any non-zero
element is invertible.
Definition 11.10. A Boolean ring is a commutative unital ring such that 1 6= 0
and for all a ∈ A we have a2 = a.
Exercise 11.11. Prove that in a Boolean ring A we have a + a = 0 for all
a ∈ A.
Exercise 11.12. Prove that
1) ({0, 1}, ∨, ∧, ¬, 0, 1) is a Boolean algebra.
2) ({0, 1}, +, ×, I, 0, 1) is a Boolean ring and a field (I is the identity map).
Exercise 11.13. Prove that if a Boolean ring A is a field then A = {0, 1}.
Definition 11.14. Let A be a set and let S = P(A) be the power set of A.
Define the following operations on S:
X ∧Y = X ∩Y
X ∨Y = X ∪Y
X∆Y = (X ∪ Y )\(X ∩ Y )
¬X = CX = A\X.
Exercise 11.15. Prove that
1) (P(A), ∨, ∧, ¬, ∅, A) is a Boolean algebra;
2) (P(A), ∆, ∧, I, ∅, A) is a Boolean ring (I is the identity map).
Hint: For any a ∈ A one can define a map ψa : P(A) → {0, 1} by setting
ψa (X) = 1 if and only if a ∈ X. Note that
1) ψa (X ∧ Y ) = ψa (X) ∧ ψa (Y ),
2) ψa (X ∨ Y ) = ψa (X) ∨ ψa (Y ),
 11. OPERATIONS 71

3) ψa (X∆Y ) = ψa (X) + ψa (Y ),
4) ψa (¬X) = ¬ψa (X).
Next note that X = Y if and only if ψa (X) = ψa (Y ) for all a ∈ A. Use these
functions to reduce the present exercise to Exercise 11.12.
Definition 11.16. Given a subset X ⊂ A one can define the characteristic
function χX : A → {0, 1} by letting χX (a) = 1 if and only if a ∈ X; in other words
χX (a) = ψa (X).
Exercise 11.17. Prove that
1) χX∨Y (a) = χX (a) ∨ χY (a),
2) χX∧Y (a) = χX (a) ∧ χY (a),
3) χX∆Y (a) = χX (a) + χY (a),
4) χ¬X (a) = ¬χX (a).
Definition 11.18. An algebraic structure is a tuple (A, ?, •, ..., ¬, −, ..., 0, 1, ...)
where A is a set, ?, •, ... are binary operations, ¬, −, ... are unary operations, and
1, 0, ... are given elements of A. (Some of these may be missing; for instance we
could have only one binary operation, one given element, and no unary operations.)
Assume we are given two algebraic structures
(A, ?, •, ..., ¬, −, ..., 0, 1, ...) and (A0 , ?0 , •0 , ..., ¬0 , −0 , ..., 00 , 10 , ...)
(with the same number of corresponding operations). A map F : A → A0 is called
a homomorphism if for all a, b ∈ A we have:
1) F (a ? b) = F (a) ?0 F (b), F (a • b) = F (a) •0 F (b),...
2) F (¬a) = ¬0 F (a), F (−a) = −0 F (a),...
3) F (0) = 00 , F (1) = 10 ,....
Example 11.19. A map F : A → A0 between two commutative unital rings is
called a homomorphism (of commutative unital rings) if for all a, b ∈ A we have:
1) F (a + b) = F (a) + F (b) and F (ab) = F (a)F (b),
2) F (−a) = −F (a) (prove that this is automatic !),
3) F (0) = 0 (prove that this is automatic !) and F (1) = 1.
Exercise 11.20. Prove that if F : A → A0 is a homomorphism of algebraic
structures and F is bijective then its inverse F −1 : A0 → A is a homomorphism.
Such an F will be called an isomorphism.
Definition 11.21. A subset A ⊂ P(A) is called a Boolean algebra of sets if
the following hold:
1) ∅ ∈ A, A ∈ A;
2) If X, Y ∈ A then X ∩ Y ∈ A, X ∪ Y ∈ A, CX ∈ A.
(Hence (A, ∨, ∧, C, ∅, A) is a Boolean algebra.)
Exercise 11.22. Prove that if A is a Boolean algebra of sets then for any
X, Y ∈ A we have X∆Y ∈ A. Prove that (A, ∆, ∩, I, ∅, A) is a Boolean ring.
Definition 11.23. A subset B ⊂ P(A) is called a Boolean ring of sets if the
following properties hold:
1) ∅ ∈ B, A ∈ B;
2) If X, Y ∈ B then X ∩ Y ∈ B, X∆Y ∈ B.
(Hence (A, ∆, ∨, I, ∅, A) is a Boolean ring.)
72 11. OPERATIONS

Exercise 11.24. Prove that any Boolean ring of sets B is a Boolean algebra
of sets.
Definition 11.25. A commutative unital ordered ring (or simply an ordered
ring) is a tuple
(R, +, ×, −, 0, 1, ≤)
where
(R, +, ×, −, 0, 1)
is a ring, ≤ is a total order on R, and for all a, b, c ∈ R the following axioms are
satisfied
1) If a < b then a + c < b + c;
2) If a < b and c > 0 then ac < bc.
We say that a ∈ R is positive if a > 0; and that a is negative if a < 0. We say
a is non-negative if a ≥ 0.
Exercise 11.26. Prove that the ring ({0, 1}, +, ×, −, 0, 1) has no structure of
ordered ring i.e., there is no order ≤ on {0, 1} such that ({0, 1}, +, ×, −, 0, 1, ≤) is
an ordered ring.
Remark 11.27. We cannot give examples yet of ordered rings. Later we will
see that the rings of integers, rationals, and reals have natural structures of ordered
rings.
Definition 11.28. Let R be an ordered ring and let R+ = {a ∈ R; a ≥ 0}. A
finite measure space is a triple (A, A, µ) where A is a set, A ⊂ P(A) is a Boolean
algebra of sets, and µ : A → R+ is a map satisfying the property that for any
X, Y ∈ A with X ∩ Y = ∅ we have
µ(X ∪ Y ) = µ(X) + µ(Y ).
If in addition µ(A) = 1 we say (A, A, µ) is a finite probability measure. We say
that X, Y ∈ A are independent if µ(X ∩ Y ) = µ(X) · µ(Y ).
Exercise 11.29. Prove that in a finite measure space µ(∅) = 0 and for any
X, Y ∈ A we have
µ(X ∪ Y ) = µ(X) + µ(Y ) − µ(X ∩ Y ).
Exercise 11.30. Let (A, ∨, ∧, ¬, 0, 1) be a Boolean algebra. For any a, b ∈ A
set
a + b = (a ∨ b) ∧ (¬(a ∧ b)).
Prove that (A, +, ∧, I, 0, 1) is a Boolean ring (I the identity map).
Exercise 11.31. Let (A, +, ×, −, 0, 1) be a Boolean ring. For any a, b ∈ A let
a ∨ b = a + b − ab
a ∧ b = ab
¬a = 1 − a.
Prove that (A, ∨, ∧, ¬, 0, 1) is a Boolean algebra.
Exercise 11.32. Let X be a set and (R, +, ·, −, 0, 1) a commutative unital
ring. Let RX be the set of all functions X → R. For F, G ∈ RX we define
F + G, F · G, −F, 0, 1 ∈ RX by the formulae
(F + G)(x) = F (x) + G(x), (F · G)(x) = F (x) · G(x),
 11. OPERATIONS 73

(−F )(x) = −F (x), 0(x) = 0, 1(x) = x,

for all x ∈ X. The operations F + G and F · G are called pointwise addition and
multiplication of functions. Prove that
(RX , +, ·, −, 0, 1)
is a commutative unital ring.
 Part 3

The discrete
 CHAPTER 12

Integers

In this Chapter we introduce the ring Z of integers and we prove some easy
theorems about this concept.
Definition 12.1. A well ordered ring is an ordered ring (R, +, ×, 0, 1, ≤) with
1 6= 0 having the property that any non-empty subset of R which is bounded from
below has a minimum element.
Remark 12.2. If (R, +, ×, 0, 1, ≤) is a well ordered ring then (R, ≤) is not a
priori a well ordered set. But if R>0 = {a ∈ R; a > 0} then (R>0 , ≤) is a well
ordered set.
We have the following remarkable theorem in set theory Tset :
Theorem 12.3. There exists a well ordered ring.
Remark 12.4. The above theorem is formulated, as usual, in Argot; but it
should be understood as being a sentence Z in Lset of the form
∃r∃s∃p∃o∃u∃l(...)
where we take a variable r to stand for the ring, a variable s for the sum, p for
the product, o for 0, u for 1, l for ≤, and the dots stand for the corresponding
conditions in the definition of a well ordered ring, written in the language of sets.
The sentence Z is complicated so we preferred to give the theorem not as a sentence
in Lset but as a sentence in Argot. This kind of abuse is very common.
Remark 12.5. We are going to sketch the proof of Theorem 12.3 in an exercise
below. The proof is involved. A cheap way to avoid the proof of this theorem is as
follows: add this theorem to the ZFC axioms and let ZFC’ be the resulting enriched
0
system of axioms. Then replace Tset by the theory Tset with axioms ZFC’. This is
what all working mathematicians essentially do anyway.
Definition 12.6. We let Z, +, ×, 0, 1, ≤ be the witnesses for the sentence Z
above; we call Z the ring of integers. In particular the conditions in the definition
of rings (associativity, commutativity, etc.) and order (transitivity, etc.) become
theorems for Z. We also set N = {a ∈ Z; a > 0} and we call N the set of natural
numbers. Later we will prove the “essential uniqueness” of Z.
Remark 12.7. The only predicate in the language Lset of sets is ∈ and the con-
stants in this language are called sets. In particular when we consider the ordered
ring of integers (Z, +, ×, 0, 1, ≤) the symbols Z, +, ×, 0, 1, ≤, N are all constants
(they are sets). In particular +, × are not originally functional symbols and ≤ is
not originally a predicate. But, according to our conventions, we may introduce
functional symbols (still denoted by +, ×) and a predicate (still denoted by ≤) via
77
78 12. INTEGERS

appropriate definitions. (This is because “the set + is a binary operation on Z” is

a theorem, etc.)
Exercise 12.8. Prove that 1 ∈ N. Hint: Use (−1) × (−1) = 1.
Exercise 12.9. Prove that if a, b ∈ Z and ab = 0 then either a = 0 or b = 0.
Give an example of a ring where this is not true. (For the latter consider the
Boolean ring P(A).)
Exercise 12.10. Prove that if a ∈ Z then the set {x ∈ Z; a − 1 < x < a} is
empty. Hint: It is enough to show that S = {x ∈ Z; 0 < x < 1} is empty. Assume
S is non-empty and let m = min S. Then 0 < m2 < m, hence 0 < m2 < 1 and
m2 < m, a contradiction.
Exercise 12.11. Prove that if a ∈ N then a = 1 or a − 1 ∈ N. Conclude that
min N = 1. Hint: Use the previous exercise.
In what follows we sketch the main idea behind the proof of Theorem 12.3. We
begin with the following:
Definition 12.12. A Peano triple is a triple (N, 1, σ) where N is a set, 1 ∈ N ,
and σ : N → N is a map such that
1) σ is injective;
2) σ(N ) = N \{1};
3) for any subset S ⊂ N if 1 ∈ S and σ(S) ⊂ S then S = N .
The conditions 1, 2, 3 are called “Peano’s axioms.”
Remark 12.13. Given a well ordered ring one can easily prove there exists
a Peano triple; cf. Exercise 12.14 below. Conversely given a Peano triple one
can prove there exists a well ordered ring; this is tedious and will be addressed in
Exercise 12.15. The idea of proof of Theorem 12.3 is to prove in set theory (i.e.,
ZFC) that there exists a Peano triple; here the axiom of infinity is crucial. Then
the existence of a well ordered ring follows.
Exercise 12.14. Prove that if Z is a well ordered ring, N = {x ∈ Z; x > 0},
and σ : N → N is σ(x) = x + 1 then (N, 1, σ) is a Peano triple.
Exercise 12.15. This exercise gives some steps towards showing how to con-
struct a well ordered ring from a given Peano triple. Assume (N, 1, σ) is a Peano
triple. For y ∈ N let
Ay = {τ ∈ N N ; τ (1) = σ(y), ∀x(τ (σ(x)) = σ(τ (x)))}.
1) Prove that Ay has at most one element. Hint: If τ, η ∈ Ay and S = {x; τ (x) =
η(x)} then 1 ∈ S and σ(S) ⊂ S; so S = N .
2) Prove that for any y, Ay 6= ∅. Hint: If T = {y ∈ N ; Ay 6= ∅} then 1 ∈ T and
σ(T ) ⊂ T ; so T = N .
3) By 1 and 2 we may write Ay = {τy }. Then define + on N by x + y = τy (x).
4) Prove that x + y = y + x and (x + y) + z = x + (y + z) on N .
5) Prove that if x, y ∈ N , x 6= y, then there exists z ∈ N such that either
y = x + z or x = y + z.
6) Define N 0 = {−} × N , Z = N 0 ∪ {0} ∪ N where 0 and − are two sets.
Naturally extend + to Z.
7) Define × on N and then on Z in the same style as for +.
8) Define ≤ on N and prove (N, ≤) is well ordered. Extend this to Z.
9) Prove that (Z, +, ×, 0, 1, ≤) is a well ordered ring.
 12. INTEGERS 79

From now on we accept Theorem 12.3 (either as a theorem whose proof we

summarily sketched or as an additional axiom for set theory).
Definition 12.16. Define the natural numbers 2, 3, ..., 9 by
2 = 1+1
3 = 2+1
....
9 = 8 + 1.
Define 10 = 2 × 5. Define 102 = 10 × 10, etc. Define symbols like 423 as being
4 × 102 + 2 × 10 + 3, etc. This is called a decimal representation.
Exercise 12.17. Prove that 12 = 9 + 3. Hint: We have:
12 = 10 + 2
= 2×5+2
= (1 + 1) × 5 + 2
= 1×5+1×5+2=5+5+2
= 5+5+1+1=5+6+1=5+7=4+1+7
= 4 + 8 = 3 + 1 + 8 = 3 + 9 = 9 + 3.
Exercise 12.18. Prove that 18 + 17 = 35. Prove that 17 × 3 = 51.
Remark 12.19. In Kant’s analysis, statements like the ones in the previous
exercise were viewed as synthetic; in contemporary mathematics, hence in the ap-
proach we follow, all these statements are, on the contrary, analytic statements.
(The definition of analytic/synthetic is taken here in the sense of Leibniz and Kant.)
Exercise 12.20. Prove that 7 ≤ 20.
Definition 12.21. For any integers a, b ∈ Z the set {x ∈ Z; a ≤ x ≤ b}
will be denoted, for simplicity, by {a, ..., b}. This set is clearly empty if a > b. If
other numbers in addition to a, b are specified then the meaning of our notation
will be clear from the context; for instance {0, 1, ..., n} means {0, ..., n} whereas
{2, 4, 6, ..., 2n} will mean {2x; 1 ≤ x ≤ n}, etc. A similar convention applies if there
are no numbers after the dots.
Example 12.22. {−2, ..., 11} = {−2, −1, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11}.
Recall that a subset A ⊂ N is bounded (equivalently bounded from above) if
there exists b ∈ N such that a ≤ b for all a ∈ A; we say that A is bounded by b
from above.
Exercise 12.23. Prove that N is not bounded.
Exercise 12.24. Prove that any subset of Z bounded from above has a maxi-
mum. Hint: If A is bounded from above by b consider the set {b − x; x ∈ A}.
Definition 12.25. An integer a is even if there exists an integer b such that
a = 2b. An integer is odd if it is not even.
Exercise 12.26. Prove that if a is odd then a − 1 is even. Hint: Consider
the set {b ∈ N; 2b ≥ a}, and let c be the minimum element of S. Then show that
2(c − 1) < a. Finally show that this implies a = 2c − 1.
Exercise 12.27. Prove that if a and b are odd then ab is odd. Hint: Write
a = 2c + 1 and b = 2d + 1 (cf. the previous exercise) and compute (2c + 1)(2d + 1).
80 12. INTEGERS

Exercise 12.28. Consider the following sentence: There is no bijection be-

tween N and Z. Explain the mistake in the following wrong proof; this is an
instance of a fallacy discussed earlier.
“Proof.” Assume there is a bijection f : N → Z. Define f (x) = x. Then f is
not surjective so it is not a bijection.
Exercise 12.29. Prove that there is a bijection between N and Z.
 CHAPTER 13

Induction

Induction is the single most important method to prove elementary theorems

about the integers. (More subtle theorems, such as many of the theorems of “num-
ber theory,” require more sophisticated methods.) Let P (x) be a formula in the
language Lset of sets, with free variable x. For each such P (x) we have:
Proposition 13.1. (Induction Principle) Assume
1) P (1).
2) For all n ∈ N if n 6= 1 and P (n − 1) then P (n).
Then for all n ∈ N we have P (n).
The above is expressed, as usual, in Argot. The same expressed as a sentence
in Lset reads:
(P (1) ∧ (∀x((x ∈ N) ∧ (x 6= 0) ∧ P (x − 1)) → P (x))) → (∀x((x ∈ N) → P (x))).
We refer to the above as induction on n. For each explicit P (x) this is a
genuine theorem. Note that the above Proposition does not say “for all P something
happens”; that would not be a sentence in the language of sets.
Proof. Let S = {n ∈ N; ¬P (n)}. We want to show that S = ∅. Assume S = ∅
and seek a contradiction. Let m be the minimum of S. By 1) m 6= 1. By Exercise
12.11 m − 1 ∈ N. By minimality of m, we have P (m − 1). By 2) we get P (m), a
contradiction. 
Exercise 13.2. Define n2 = n × n and n3 = n2 × n for any integer n. Prove
that for any natural n there exists an integer m such that n3 − n = 3m. (Later
we will say that 3 divides n3 − n.) Hint: Proceed by induction on n as follows.
Let P (n) be the sentence: for all natural n there exists an integer m such that
n3 − n = 3m. P (1) is true because 13 − 1 = 3 × 0. Assume now that P (n − 1) is
true i.e., (n − 1)3 − (n − 1) = 3q for some integer q and let us check that P (n) is
true i.e., that n3 − n = 3m for some integer m. The equality (n − 1)3 − (n − 1) = 3q
reads n3 − 3n2 + 3n − 1 − n + 1 = 3q. Hence n3 − n = 3(n2 − n) and we are done.
Exercise 13.3. Define n5 = n3 × n2 . Prove that for any natural n there exists
an integer m such that n5 − n = 5m.
Proposition 13.4. If there exists a bijection {1, ..., n} → {1, ..., m} then n =
m.
Proof. We proceed by induction on n. Let P (n) be the statement of the
Proposition. Clearly P (1) is true; cf. the Exercise below. Assume now P (n − 1)
is true and let’s prove that P (n) is true. So consider a bijection F : {1, ..., n} →
{1, ..., m}; we want to prove that n = m. Let i = F (n) and define the map
G : {1, ..., n − 1} → {1, ..., m}\{i} by G(j) = F (j) for all 1 ≤ j ≤ n − 1. Then
81
82 13. INDUCTION

clearly G is a bijection. Now consider the map H : {1, ..., m}\{i} → {1, ..., m − 1}
defined by H(j) = j for 1 ≤ j ≤ i − 1 and H(j) = j − 1 for i + 1 ≤ j ≤ m. (The
definition is correct because for any j ∈ {1, ..., m}\{i} either j ≤ i − 1 or j ≥ i + 1;
cf. Exercise 12.10.) Clearly H is a bijection. We get a bijection
H ◦ G : {1, ..., n − 1} → {1, ..., m − 1}.
Since P (n − 1) is true we get n − 1 = m − 1. Hence n = m and we are done. 
Exercise 13.5. Check that P (1) is true in the above Proposition.
Remark 13.6. Note the general strategy of proofs by inductions. Say P (n)
is “about n objects.” There are two steps. The first step is the verification of
P (1) i.e., one verifies the statement “for one object.” For the second step (called
the induction step) one considers a situation with n objects; one “removes” from
that situation “one object” to get a “situation with n − 1 objects”; one uses the
“induction hypothesis” P (n − 1) to conclude the claim for the “situation with
n − 1 objects.” Then one tries to “go back” and prove that the claim is true for
the situation with n objects. So the second step is performed by “removing” one
object from an arbitrary situation with n objects and NOT by adding one object
to an arbitrary situation with n − 1 objects. Below is an example of a fallacious
reasoning by induction based on “adding” instead of “subtracting” an object.
Example 13.7. Here is a wrong argument for the induction step in the proof
of Proposition 13.4.
“Proof.” Let G : {1, ..., n − 1} → {1, ..., m − 1} be any bijection and let F :
{1, ..., n} → {1, ..., m} be defined by F (i) = G(i) for i ≤ n − 1 and F (n) = m.
Clearly F is a bijection. Now by the induction hypothesis n − 1 = m − 1. Hence
n = m. This ends the proof.
The mistake is that the above does not end the proof: the above argument
only covers bijections F : {1, ..., n} → {1, ..., m} constructed from bijections G :
{1, ..., n − 1} → {1, ..., m − 1} in the special way described above. In other words
an arbitrary bijection F : {1, ..., n} → {1, ..., m} does not always arise the way we
defined F in the above “proof.” In some sense the mistake we just pointed out is
that of defining the same constant twice (cf. Example 12.28): we were supposed to
define the symbol F as being an arbitrary bijection but then we redefined F in a
special way through an arbitrary G. The point is that if G is arbitrary and F is
defined as above in terms of G then F will not be arbitrary (because F will always
send n into m).
Definition 13.8. A set A is finite if there exists an integer n ≥ 0 and a
bijection F : {1, ..., n} → A. (Note that n is then unique by Proposition 13.4.)
We write |A| = n and we call this number the cardinality of A or the number of
elements of A. (Note that |∅| = 0.) If F (i) = ai we write A = {a1 , ..., an }. A set is
infinite if it is not finite.
Exercise 13.9. Prove that |{2, 4, −6, 9, −100}| = 5.
Exercise 13.10. For any finite sets A and B we have that A ∪ B is finite and
|A ∪ B| + |A ∩ B| = |A| + |B|.
 13. INDUCTION 83

Hint: Reduce to the case A ∩ B = ∅. Then if F : {1, ..., a} → A and G : {1, ..., b} →
B are bijections prove that H : {1, ..., a + b} → A ∪ B defined by H(i) = F (i) for
1 ≤ i ≤ a and H(i) = G(i − a) for a + 1 ≤ i ≤ a + b is a bijection.
Exercise 13.11. For any finite sets A and B we have that A × B is finite and
|A × B| = |A| × |B|.
Hint: Induction on |A|.
Exercise 13.12. Let F : {1, ..., n} → Z be an injective map and write F (i) =
ai . We refer to such a map as a (finite) family of numbers. Prove that there exists
a unique map G : {1, ..., n} → Z such that G(1) = a1 and G(k) = G(k − 1) + ak for
2 ≤ k ≤ n. Hint: Induction on n.
Definition
Pn 13.13. In the notation of the above Exercise define the (finite)
sum i=1 ai as the number G(n). We also write a1 + ... + an for this sum. If
a1 = ... = an = a the sum a1 + ... + an is written as a + ... + a (n times).
Exercise 13.14. Prove that for any a, b ∈ N we have
a × b = a + ... + a (b times) = b + ... + b (a times).
Qn
Exercise 13.15. Define in a similar way the (finite) product i=1 ai (which
is also denoted by a1 ...an = a1 × ... × an ). Prove the analogues of associativity and
distributivity for sums and products of families of numbers. Define ab for a, b ∈ N
and prove that ab+c = ab × ac and (ab )c = abc .
Exercise 13.16. Prove that if a is an integer and n is a natural number then
an − 1 = (a − 1)(an−1 + an−2 + ... + a + 1).
Hint: Induction on n.
Exercise 13.17. Prove that if a is an integer and n is an integer then
a2n+1 + 1 = (a + 1)(a2n − a2n−1 + a2n−2 − ... − a + 1).
Hint: Set a = −b.
Exercise 13.18. Prove that a subset A ⊂ N is bounded if and only if it is
finite. Hint: To prove that bounded sets are finite assume this is false and let b
be the minimum natural number with the property that there is a set A bounded
from above by b and infinite. If b 6∈ A then A is bounded from above by b − 1
(Exercise 12.10) and we are done. If b ∈ A then, by minimality of b, there is a
bijection A\{b} → {1, ..., m} and one constructs a bijection A → {1, ..., m + 1}
which is a contradiction. To prove that finite sets are bounded assume this is false
and let n be minimum natural number with the property that there is a finite subset
A ⊂ N of cardinality n which is not bounded. Let F : {1, ..., n} → A be a bijection,
ai = F (i). Then {a1 , ..., an−1 } is bounded from above by some b and conclude that
A is bounded from above by either b or an .
Exercise 13.19. Prove that any subset of a finite set is finite. Hint: Use the
previous exercise.
Definition 13.20. Let A be a set and n ∈ N. Define the set An to be the set
{1,...,n}
A of all maps {1, ..., n} → A. Call
[
A? = An
n=1
84 13. INDUCTION

the set of words with letters in A.

Definition 13.21. If f : {1, ..., n} → A and f (i) = ai we write f as a “tu-
ple” (a1 , ..., an ) and sometimes as a “word” a1 ...an ; in other words we add to the
definitions of set theory the following definitions
f = (a1 , ..., an ) = a1 ...an .
Exercise 13.22. Show that the maps An × Am → An+m ,
((a1 , ..., an ), (b1 , ..., bm )) 7→ (a1 , ..., an , b1 , ..., bm )
(called concatenations), are bijections. They induce a non-injective binary opera-
tion A? × A? → A? , (u, v) → uv. Prove that u(vw) = (uv)w.
 CHAPTER 14

Rationals

With the integers at our disposal one can use the axioms of set theory to
construct a whole array of familiar sets of numbers such as the rationals, the reals,
the imaginaries, etc. We start here with the rationals.
Definition 14.1. For any a, b ∈ Z with b 6= 0 define the fraction ab to be the
set of all pairs (c, d) with c, d ∈ Z, d 6= 0 such that ad = bc. Denote by Q the set of
all fractions. So
a
= {(c, d) ∈ Z × Z; d 6= 0, ad = bc},
b
a
Q = { ; a, b ∈ Z, b 6= 0}.
b
Example 14.2.
6
= {(6, 10), (−3, −5), (9, 15), ...} ∈ Q.
10
Exercise 14.3. Prove that ab = dc if and only if ad = bc. Hint: Assume ad = bc
and let us prove that ab = dc . We need to show that ab ⊂ dc and that dc ⊂ ab . Now if
(x, y) ∈ ab then xb = ay; hence xbd = ayd. Since ad = bc we get xbd = bcy. Hence
b(xd − cy) = 0. Since b 6= 0 we have xd − cy = 0 hence xd = cy hence (x, y) ∈ dc .
We proved that ab ⊂ dc . The other inclusion is proved similarly. So the equality
a c a c
b = d is proved. Conversely if one assumes b = d one needs to prove ad = bc; we
leave this to the reader.
Exercise 14.4. On the set A = Z × (Z\{0}) one can consider the relation:
(a, b) ∼ (c, d) if and only if ad = bc. Prove that ∼ is an equivalence relation. Then
observe that ab is the equivalence class
[
(a, b)
of (a, b). Also observe that Q = A/ ∼ is the quotient of A by the relation ∼.
a
Exercise 14.5. Prove that the map Z → Q, a 7→ 1 is injective.
Definition 14.6. By abuse we identify a ∈ Z with a1 ∈ Q and write a1 = a;
this identifies Z with a subset of Q. Such identifications are very common and will
be done later in similar contexts.
a c ad+bc a c ac
Definition 14.7. Define b + d = bd , b × d = bd .
a a0 c c0
Exercise 14.8. Show that the above definition is correct (i.e., if b = b0 , d = d0
0 0 0 0
then ad+bc
bd = a db0+b
d0
c
and similarly for the product).
Exercise 14.9. Prove that Q (with the operations + and × defined above and
with the elements 0, 1) is a field.
85
86 14. RATIONALS

Definition 14.10. For ab , dc with b, d > 0 write a

b ≤ c
d if ad − bc ≤ 0. Also
write ab ≤ dc if ab ≤ dc and ab 6= dc .
Exercise 14.11. Prove that Q equipped with ≤ is an ordered ring but it is
not a well ordered ring.
Exercise 14.12. Let A be a non-empty finite set and define µ : P(A) → Q by
|X|
µ(X) = .
|A|
1) (A, P(A), µ) is a finite probability measure space.
2) Prove that if X = Y 6= A then X and Y are not independent.
3) Prove that if X∩Y = ∅ and X 6= ∅, Y 6= ∅ then X and Y are not independent.
4) Prove that if A = B × C, X = B 0 × C, Y = B × C 0 , B 0 ⊂ B, C 0 ⊂ C, then
X and Y are independent.
Exercise 14.13. Prove by induction the following equalities:

n(n + 1)
1 + 2 + ... + n = .
2
n(n + 1)(2n + 1)
12 + 22 + ... + n2 = .
6
n2 (n + 1)2
13 + 23 + ... + n3 = .
4
 CHAPTER 15

Combinatorics

Combinatorics is about counting elements in (i.e., finding cardinalities of) finite

sets. The origins of combinatorics are in the work of Pascal, Jakob Bernoulli, and
Leibniz; these origins are intertwined with the origins of probability theory and the
early development of calculus.
Definition 15.1. For n ∈ N define the factorial of n (read n factorial) by
n! = 1 × 2 × ... × n ∈ N.
Also set 0! = 1.
Definition 15.2. For 0 ≤ k ≤ n in N define the binomial coefficient
 
n n!
= ∈ Q.
k k!(n − k)!
One also reads this “n choose k.”
Exercise 15.3. Prove that
   
n n
=
k n−k
and    
n n
= 1, = n.
0 1
Exercise 15.4. Prove that
     
n n n+1
+ = .
k k+1 k+1
Hint: Direct computation with the definition.
Exercise 15.5. Prove that
 
n
∈ Z.
k
Hint: Fix k and proceed by induction on n; use Exercise 15.4.
Exercise 15.6. For any a, b in any ring we have
n  
X n
(a + b)n = ak bn−k .
k
k=0

Here if c is in a ring R and m ∈ N then mc = c + ... + c (m times). Hint: Induction

on n and use Exercise 15.4.
87
88 15. COMBINATORICS

Exercise 15.7. (Subsets) Prove that if |A| = n then |P(A)| = 2n . (A set with
n elements has 2n subsets.) Hint: Induction on n; if A = {a1 , ..., an+1 } use
P(A) = {B ∈ P(A); an+1 ∈ B} ∪ {B ∈ P(A); an+1 6∈ B}.
Exercise 15.8. (Combinations) Let A be a set with |A| = n, let 0 ≤ k ≤ n,
and set
Comb(k, A) = {B ∈ P(A); |B| = k}.
Prove that  
n
|Comb(k, A)| = .
k
 
n
In other words a set of n elements has exactly subsets with k elements. A
k
subset of A having k elements is called a combination of k elements from the set A.
Hint: Fix k and proceed by induction on n. If A = {a1 , ..., an+1 } use Exercise
15.4 plus the fact that Comb(k, A) can be written as
{B ∈ P(A); |B| = k, an+1 ∈ B} ∪ {B ∈ P(A); |B| = k, an+1 6∈ B}.
Exercise 15.9. (Permutations) For a set A let Perm(A) ⊂ AA be the set of all
bijections F : A → A. A bijection F : A → A is also called a permutation. Prove
that if |A| = n then
|Perm(A)| = n!.
So the exercise says that a set of n elements has n! permutations. Hint: Let
|A| = |B| = n and let Bij(A, B) be the set of all bijections F : A → B; it is enough
to show that |Bij(A, B)| = n!. Proceed by induction on n; if A = {a1 , ..., an+1 },
B = {b1 , ..., bn+1 } then use the fact that
n+1
[
Bij(A, B) = {F ∈ Bij(A, B); F (a1 ) = bk }.
k=1

For d ∈ N and X a set let X d be the set of all maps {1, ..., d} → X. We identify
a map i 7→ ai with the tuple (a1 , ..., ad ).
Exercise 15.10. (Combinations with repetition) Let
Combrep(n, d) = {(x1 , ..., xd ) ∈ Zd ; xi ≥ 0, x1 + ... + xd = n}.
Prove that  
n+d−1
|Combrep(n, d)| = .
d−1
Hint: Let A = {1, ..., n + d − 1}. Prove that there is a bijection
Comb(d − 1, A) → Combrep(n, d).
The bijection is given by attaching to any subset
{i1 , ..., id−1 } ⊂ {1, ..., n + d − 1}
(where i1 < ... < id−1 ) the tuple (x1 , ..., xd−1 ) where
1) x1 = |{i ∈ Z; 1 ≤ i < i1 }|,
2) xk = |{i ∈ Z; ik < i < ik+1 }|, for 2 ≤ k ≤ d − 1, and
3) xd = |{i ∈ Z; id−1 < i ≤ n + d − 1}|.
 CHAPTER 16

Sequences

Definition 16.1. A sequence in a set A is a map F : N → A. If we write

F (n) = an we also say that a1 , a2 , ... is a sequence in A or that (an ) is a sequence
in A.
Theorem 16.2. (Recursion theorem) Let A be a set, a ∈ A an element, and
let F1 , F2 , ... be a sequence of maps A → A. Then there is a unique map G : N → A
such that G(1) = a and G(n + 1) = Fn (G(n)) for all n ∈ N.
Sketch of proof. First one proves by induction on n that for any n there exists
a unique map
Gn : {1, ..., n} → A
such that for any k < n
Gn (k + 1) = Fk (Gn (k)).
Next by uniqueness one gets that for any n and any k ≤ n we have
Gn (k) = Gn+1 (k).
Finally, seeing Gn as a subset of {1, ..., n} × A one defines
[
G := Gn ⊂ N × A
and one proves G is a map with the desired properties. 
Here are some applications of recursion.
Proposition 16.3. Let (A, ≤) be an ordered set that has no maximal element.
Then there is a sequence F : N → A such that for all n ∈ N we have F (n) <
F (n + 1).
Proof. Let B = {(a, b) ∈ A × A; a < b}. By hypothesis the first projection
F : B → A, (a, b) 7→ a is surjective. By the axiom of choice there exists G : A → B
such that F ◦ G = IA . Then G(a) > a for all a. By the recursion theorem there
exists F : N → A such that F (n + 1) = G(F (n)) for all n and we are done. 
Exercise 16.4. (Uniqueness of the ring of integers) Let Z and Z 0 be two
well ordered rings with identities 1 and 10 . Prove that there exists a unique ring
homomorphism F : Z → Z 0 ; prove that this F is bijective and increasing. Hint: Let
Z+ be the set of all elements in Z which are > 0 and similarly for Z 0 . By recursion
0
(which is, of course, valid in any well ordered ring) there is a unique F : Z+ → Z+
0 0
satisfying F (1) = 1 and F (n + 1) = F (n) + 1 . Define F on Z by F (−n) − F (n)
for −n ∈ Z+ .
Definition 16.5. A set A is countable if there exists a bijection F : N → A.
A set is at most countable if it is either finite or countable.
89
90 16. SEQUENCES

Example 16.6. The set of all squares S = {n2 ; n ∈ N} is countable; indeed

F : Z → S, F (n) = n2 is a bijection.
Exercise 16.7. Any infinite subset of a countable set is countable. Hint: It is
enough to show that any subset A ⊂ N is countable. Let F ⊂ N × N be the set
F = {(x, y) ∈ N × N; y = min(A ∩ {z ∈ N; z > x})
which is of course a map. By the recursion theorem there exists G : N → N such
that G(n + 1) = F (G(n)). One checks that G is injective and its image is A.
Exercise 16.8. Prove that N × N is countable. Hint: One can find injections
N × N → N ; e.g., (m, n) 7→ (m + n)2 + m.
Exercise 16.9. Prove that Q is countable.
Exercise 16.10. Prove that P(N) is not countable.
Hint. Indeed this is a consequence of the more general theorem we proved
that there is no bijection between a set A and its power set P(A). However it is
interesting to give a reformulation of the argument in this case (Cantor’s diagonal
argument). Assume P(N) is countable and seek a contradiction. Since P(N) is in
bijection with {0, 1}N we get that there is a bijection F : N → {0, 1}N . Denote F (n)
by Fn : N → {0, 1}. Construct a map G : N → {0, 1} by the formula
G(n) = ¬(Fn (n))
where ¬ : {0, 1} → {0, 1}, ¬0 = 1, ¬1 = 0. (The definition of G does not need
the recursion theorem; one can define G as a “graph” directly (check!).) Since F is
surjective there exists m such that G = Fm . In particular:
G(m) = Fm (m) = ¬G(m),
a contradiction.
Exercise 16.11. Prove that if (An ) is a sequence of sets such that each An is
at most countable then the union
[
An
n∈N
is at most countable. Deduce that the set of words A∗ with letters in a finite
non-empty set A is countable.
Exercise 16.12. Let us call F the set of all functions f : N → N ∪ {∞} where
∞ 6∈ N. If f (n) = ∞ we say that f with input n does not terminate (runs forever
without giving an output). If f (n) = m ∈ N we say that f with input n terminates
and gives output m.
Let (fn ) be a sequence of elements of F. Its oracle is defined as the function
g ∈ F sartisfying:
1) g(n) ∈ {1, 2} for all n; in particular g with any input n always terminates.
2) g(n) = 1 if fn with input n terminates (i.e. fn (n) ∈ N),
3) g(n) = 2 if fn with input n does not terminate (i.e. fn (n) = ∞).
Given (fn ) one can define a function f ∈ F (which we refer to as the Turing
function attached to (fn )) as follows. Let g be the oracle of (fn ). Then we let
f (n) = fn (n) + 1 if g(n) = 1 and we let f (n) = 1 if gn (n) = 2.
Prove that if (fn ) is a sequence of functions in F then the Turing function f
attached to (fn ) is not a member of (fn ).
 16. SEQUENCES 91

Hint: if fn with input n does not terminate then f cannot be equal to fn

because f terminates with input n; on the other hand if fn with input n terminates
then again f 6= fn because they have different outputs.
(This is Turing’s proof that there is no “program” that decides if a given “pro-
gram” with a given input terminates. Indeed Turing defined the notion of “pro-
gram” as being a special type of function in F. One proves there are countably
many programs. Let (fn ) be the sequence of all programs. One proves that if the
oracle g of this sequence is a program then the Turing function attached to (fn ) is
a program. One concludes that g cannot be a program.)
Remark 16.13. Consider the following sentence called the continuum hypoth-
esis:

For any set A if there exists an injection A → P(N) then either there exists an
injection A → N or there exists a bijection A → P(N).

One can ask if the above is a theorem. Answering this question (raised by
Cantor) leads to important investigations in set theory. The answer (given by two
theorems of Gödel and Cohen in the framework of mathematical logic rather than
logic) turned out to be rather surprising.
 Part 4

The continuum
 CHAPTER 17

Reals

Real numbers have been implicitly around throughout the history of mathe-
matics as an expression of the idea of continuity of magnitudes. What amounts to
an axiomatic introduction of the reals can be found in Euclid (and is attributed
to Eudoxus). The first construction of the reals from the “discrete” (i.e., from the
rationals) is due to Dedekind.
Definition 17.1. (Dedekind) A real number is a subset u ⊂ Q of the set Q of
rational numbers with the following properties:
1) u 6= ∅ and u 6= Q;
2) u has no minimum;
2) if x ∈ u, y ∈ Q, and x ≤ y then y ∈ u.
Denote by R the set of real numbers.
Example 17.2.
1) Any rational number x ∈ Q can be identified with the real number
ux = {y ∈ Q; x < y}.
It is clear that ux = ux0 for x, x0 ∈ Q implies x = x0 . We identify any rational
number x with ux . So we may view Q ⊂ R.
2) One defines, for instance, for any r ∈ Q with ≥ 0,
√
r = {x ∈ Q; x ≥ 0, x2 > r}.
Definition 17.3. A real number u ∈ R is called irrational if u 6∈ Q.
Definition 17.4. If u and v are real numbers we write u ≤ v if and only if
v ⊂ u. For u, v ≥ 0 define
u+v = {x + y; x ∈ u, y ∈ v}
u × v = uv = {xy; x ∈ u, y ∈ v}.
Note that this extends addition and multiplication on the non-negative ratio-
nals.
Exercise 17.5.
1) Prove that √ ≤ is a total order on R.
2) Prove that r is a real number for r ∈ Q, r ≥ 0.
3) Prove that u + v and u × v are real numbers.
4) Naturally extend the definition of addition + and multiplication × of real
numbers to the case when the numbers are not necessarily ≥ 0. Prove that
(R, +, ×, −, 0, 1) is a field. Naturally extend the order ≤ on Q to an order on
R and prove that R with ≤ is an ordered ring.
95
96 17. REALS

Exercise 17.6. Define the sum and the product of a family of real (or complex)
numbers indexed by a finite set. Hint: Use the already defined concept for integers
(and hence for the rationals).
√
Exercise 17.7. Prove that for r ∈ Q, r ≥ 0, we have ( r)2 = r. (Hint: The
harder part is to show that if z ∈ Q satisfies z > r then there exist x, y ∈ Q such
that x ≥ 0, y ≥ 0, z = xy, x2 > r, y 2 > r. It is enough to show that there exists a
rational number ρ with z > ρ2 > r because then we can write z = xy with x = ρ
and y = z/ρ > ρ. To show this it is enough to prove that for any n natural there
exist rational numbers tn and sn such that t2n < r < s2n and sn −tn ≤ (s1 −t1 )/2n−1 .
For n = 1 take t1 = 0 and s1 such that s21 > r. Assuming the above is true for
n one sets tn+1 = tn and sn+1 = (tn + sn )/2 if t2n ≤ r < ((tn + sn )/2)2 and
one sets tn+1 = (tn + sn )/2 and sn+1 = sn if ((tn + sn )/2)2 < r < s2n ; the case
r = ((tn + sn )/2)2 is left to the reader.
Exercise
√ 17.8. Prove√that for any√r ∈ R with r > 0 there exists a unique
number r ∈ R such that r > 0 and ( r)2 = r.
√ √
Exercise 17.9. Prove that 2 is irrational i.e., 2 6∈ Q. Hint: Assume there
exists a rational number x such that x2 = 2 and seek a contradiction. Let a ∈ N
2
be minimal with the property that x = ab for some b. Now ab2 = 2 hence 2b2 = a2 .
Hence a2 is even. Hence a is even (because if a were odd then a2 would be odd).
Hence a = 2c for some integer c. Hence 2b2 = (2c)2 = 4c2 . Hence b2 = 2c2 . Hence
2c
b2 is even. Hence b is even. Hence b = 2d for some integer d. Hence x = 2d = dc
and c < a. This contradicts the minimality of a which ends the proof.
Remark 17.10. The above proof is probably one of the “first” proofs by contra-
diction in the history of mathematics; this proof appears, for instance, in Aristotle,
and√it is believed to have been discovered by the Pythagoreans. The irrationality
of 2 was translated √ by the Greeks as evidence that arithmetic is insufficient to
control geometry ( 2 is the length of the diagonal of a square with side 1) and ar-
guably created the first crisis in the history of mathematics, leading to a separation
of algebra and geometry that lasted until Fermat and Descartes.
Exercise 17.11. Prove that the set
{r ∈ Q; r > 0, r2 < 2}
has no supremum in Q.
Remark 17.12. Later we will prove that R is not countable.
Definition 17.13. For any a ∈ R we let |a| be a or −a according as a ≥ 0 or
a ≤ 0, respectively.
Exercise 17.14. Prove the so-called triangle inequality:
|a + b| ≤ |a| + |b|
for all a, b ∈ R.
Definition 17.15. For a < b in R we define the open interval
(a, b) = {c ∈ R; a < c < b} ⊂ R.
(Not to be confused with the pair (a, b) ∈ R × R which is denoted by the same
symbol.)
 CHAPTER 18

Topology

Topology is about geometric properties that are invariant under continuous

transformations. An early topological result is the formula of Descartes-Euler re-
lating the number of vertices, edges, and faces of a convex polyhedron. (We will
not discuss this here as it is surprisingly difficult to present things rigorously.) Af-
ter Riemann’s work on surfaces defined by algebraic functions, topology became
a key feature in geometry and analysis and nowadays topological ideas are to be
found everywhere in mathematics, including number theory. Here we will restrict
ourselves to explaining the basic idea of continuity.
Definition 18.1. A topology on a set X is a subset T ⊂ P(X) of the power
set of X with the following properties:
1) ∅ ∈ T and X ∈ T;
2) If U, V ∈ T then U ∩ V ∈ T;
3)SIf (Ui )i∈I is a family of subsets Ui ⊂ X and if for all i ∈ I we have Ui ∈ T
then i∈I Ui ∈ T.
A subset U ⊂ X is called open if U ∈ T. A subset Z ⊂ X is called closed if
X\Z is open. Elements of X are called points of X.
Example 18.2. T = P(X) is a topology on X.
Example 18.3. T = {∅, X} ⊂ P(X) is a topology on X.
Example 18.4. A subset U ⊂ R is called open if for any x ∈ U there exists an
open interval containing x and contained in U , x ∈ (a, b) ⊂ U . Let T ⊂ P(R) be
the set of all open sets of R. Then T is a topology on R; we call this the Euclidean
topology.
Exercise 18.5. Prove that T in Example 18.4 is a topology.
Exercise 18.6. Prove that if U, V ⊂ X then
T = {∅, U, V, U ∪ V, U ∩ V, X}
is a topology. Find the closed sets of X.
Exercise 18.7. Prove that if (Tj )j∈J is a family of topologies Tj ⊂ P(X) on
X then j∈J Tj is a topology on X.
T

Definition 18.8. If T0 ⊂ P(X) is a subset of the power set then the intersection
\
T= T0
T 0 ⊃T0

of all topologies T containing T0 is called the topology generated by T0 .

0

Exercise 18.9. Let T0 = {U, V, W } ⊂ P(X). Find explicitly the topology

generated by T0 . Find all the closed sets in that topology.
97
98 18. TOPOLOGY

Definition 18.10. A topological space is a pair (X, T) consisting of a set X

and a topology T ⊂ P(X) on X. Sometimes one writes X instead of (X, T) if T is
understood from context.
Definition 18.11. Let X and X 0 be two topological spaces. A map F : X →
X is continuous if for all open V ⊂ X 0 the set F −1 (V ) ⊂ X is open.
0

Exercise 18.12. If T is a topology on X and T 0 is the topology on X 0 defined

by T 0 = {∅, Y } then any map F : X → X 0 is continuous.
Exercise 18.13. If T is the topology T = P(X) on X and T 0 is any topology
on X 0 then any map F : X → X 0 is continuous.
Exercise 18.14. Prove that if X, X 0 , X 00 are topological spaces and G : X →
X , F : X 0 → X 00 are continuous maps then the composition F ◦ G : X → X 00 is
0

continuous.
Exercise 18.15. Give an example of two topological spaces X, X 0 and of a
bijection F : X → X 0 such that F is continuous but F −1 is not continuous. (This
is to be contrasted with the situation of algebraic structures to be discussed later.
See Exercise 11.20.)
Motivated by the above phenomenon, one gives the following
Definition 18.16. A homeomorphism between two topological spaces is a
continuous bijection whose inverse is also continuous.
Definition 18.17. If X is a topological space and Y ⊂ X is a subset then the
set of all subsets of Y of the form U ∩ Y with U open in X form a topology on Y
called the induced topology.
Exercise 18.18. Prove that if X is a topological space and Y ⊂ X is open
then the induced topology on Y consists of all open sets of X that are contained in
Y.
Definition 18.19. Let X be a topological space and let A ⊂ X be a subset.
We say that A is connected if whenever U and V are open in X with U ∩ V ∩ A = ∅
and A ⊂ U ∪ V it follows that U ∩ A = ∅ or V ∩ A = ∅.
Exercise 18.20. Prove that if F : X → X 0 is continuous and A ⊂ X is
connected then F (A) ⊂ X 0 is connected.
Definition 18.21. Let X be a topological space and let A ⊂ X be a subset.
A point x ∈ X is called an accumulation point of A if for any open set U in X
containing x the set U \{x} contains a point of A.
Exercise 18.22. Let X be a topological space and let A ⊂ X be a subset.
Prove that A is closed if and only if A contains all its accumulation points.
Definition 18.23. Let X be a topological space and A ⊂ X. We say A is
compact if whenever [
A⊂ Ui
i∈I
with (Ui )i∈I a family of open sets in X indexed by some set I there exists a finite
subset J ⊂ I such that [
A⊂ Uj .
j∈J
 18. TOPOLOGY 99

We sometimes refer to (Ui )i∈I as an open cover of A and to (Uj )j∈J as a finite open
subcover. So A is compact if and only if any open cover of A has a finite open
subcover.
Exercise 18.24. Prove that if X is a topological space and X is a finite set
then it is compact.
Exercise 18.25. Prove that R is not compact in the Euclidean topology. Hint:
Consider the open cover [
R= (−n, n)
n∈N
and show it has no finite open subcover.
Exercise 18.26. Prove that no open interval (a, b) in R is compact (a < b).
Exercise 18.27. Prove that if F : X → X 0 is a continuous map of topological
spaces and A ⊂ X is compact then F (A) ⊂ X 0 is compact.
Definition 18.28. A topological space X is a Hausdorff space if for any two
points x, y ∈ X there exist open sets U ⊂ X and V ⊂ X such that x ∈ U , y ∈ V ,
and U ∩ V = ∅.
Exercise 18.29. Prove the R with the Euclidean topology is a Hausdorff space.
Exercise 18.30. Prove that if X is a Hausdorff space, A ⊂ X, and x ∈ X\A
then there exist open sets U ⊂ X and V ⊂ X such that x ∈ U , A ⊂ V , and
U ∩ V = ∅. In particular any compact subset of a Hausdorff space is closed.
Hint: For any a ∈ A let Ua ⊂ X and Va ⊂ X be open sets such that x ∈ Ua ,
a ∈ Va , Ua ∩ Va = ∅. Then (Va )a∈A is an open covering of A. Select (Vb )b∈B a finite
subcover of A where B ⊂ A is a finite set, B = {b1 , ..., bn }. Then let
U = Ub1 ∩ ... ∩ Ubn
V = Vb1 ∪ ... ∪ Vbn .
Definition 18.31. Let X, X 0 be topological spaces. Then the set X × X 0
may be equipped with the topology generated by the family of all sets of the form
U × U 0 where U and U 0 are open in X and X 0 respectively. This is called the
product topology on X × X 0 . Iterating this we get a product topology on a product
X1 × ... × Xn of n topological spaces.
Exercise 18.32. Prove that for any r ∈ R with r > 0, the set
D = {(x, y) ∈ R2 ; x2 + y 2 < r2 }
is open in the product topology of R2 .
Definition 18.33. A topological manifold is a topological space X such that for
any point x ∈ X there exists an open set U ⊂ X containing x and a homeomorphism
F : U → V where V ⊂ Rn is an open set in Rn for the Euclidean topology. (Here
U and V are viewed as topological spaces with the topologies induced from X and
Rn , respectively.)
Remark 18.34. If X is a set of topological manifolds then one can consider the
following relation ∼ on X: for X, X 0 ∈ X we let X ∼ X 0 if and only if there exists a
homeomorphism X → X 0 . Then ∼ is an equivalence relation on X and one of the
basic problems of topology is to “describe” the set X/ ∼ of equivalence classes in
various specific cases.
100 18. TOPOLOGY

More properties of the Euclidean topology of R will be examined in the chapter

on limits.
 CHAPTER 19

Imaginaries

Complex numbers (also called imaginary numbers) appeared in work of Car-

dano, Bombelli, d’Alembert, Gauss, and others, in relation to solving polynomial
equations. The modern definition below is due to Hamilton.
Definition 19.1. (Hamilton) A complex number is a pair (a, b) where a, b ∈ R.
We denote by C the set of complex numbers. Hence C = R × R. Define the sum
and the product of two complex numbers by
(a, b) + (c, d) = (a + c, b + d)
(a, b) × (c, d) = (ac − bd, ad + bc).
Remark 19.2. Identify any real number a ∈ R with the complex number
(a, 0) ∈ C; hence write a = (a, 0). In particular 0 = (0, 0) and 1 = (1, 0).
Exercise 19.3. Prove that C equipped with 0, 1 above and with the operations
+, × above is a field. Also note that the operations + and × on C restricted to R
are the “old” operations + and × on R.
Definition 19.4. We set i = (0, 1).
Remark 19.5. i2 = −1. Indeed
i2 = (0, 1) × (0, 1) = (0 × 0 − 1 × 1, 0 × 1 + 1 × 0) = (−1, 0) = −1.
Remark 19.6. For any complex number (a, b) = a + bi. Indeed
(a, b) = (a, 0) + (0, b) = (a, 0) + (b, 0)(0, 1) = a + bi.
Definition 19.7. For any complex number z = a + bi we define its absolute
value p
|z| = a2 + b2 .
Exercise 19.8. Prove the so-called triangle inequality:
|a + b| ≤ |a| + |b|
for all a, b ∈ C.
Definition 19.9. For any complex number z = a + bi we define its conjugate
z = a − bi.
(The upper bar is not to be confused with the notation used in the chapter on
residues.)
Exercise 19.10. Prove that for any z, w ∈ C we have
1) z + w = z + w;
2) z · w = z · w;
3) z −1 = z −1 for z 6= 0;
4) z · z = |z|2 .
101
102 19. IMAGINARIES

Definition 19.11. For any complex number z = a + bi ∈ C and any real

number r > 0 we define the open disk with center z and radius r,
D(z, r) = {w ∈ C; |w − z| < r} ⊂ C.
A subset U ⊂ C is called open if for any z ∈ U there exists an open disk centered
at z and contained in U . Let T ⊂ P(C) be the set of all open sets of C.
Exercise 19.12. Prove that T is a topology on C; we call this the Euclidean
topology.
Exercise 19.13. Prove that C cannot be given the structure of an ordered
ring.
 Part 5

Algebra
 CHAPTER 20

Arithmetic

Our main aim here is to introduce some of the basic “arithmetic” of Z. In its
turn arithmetic can be used to introduce the finite rings Z/mZ of residue classes
modulo m and, in particular, the finite fields Fp = Z/pZ, where p is a prime. The
arithmetic of Z to be discussed below already appears in Euclid. Congruences and
residue classes were introduced by Gauss.
Definition 20.1. For integers a and b we say a divides b if there exists an
integer n such that b = an. We write a|b. We also say a is a divisor of b. If a does
not divide b we write a 6 |b.
Example 20.2. 4|20; −4|20; 6 6 |20.
Exercise 20.3. Prove that
1) if a|b and b|c then a|c;
2) if a|b and a|c then a|b + c;
3) a|b defines an order relation on N but not on Z.
Theorem 20.4. (Euclid division) For any a ∈ Z and b ∈ N there exist unique
q, r ∈ Z such that a = bq + r and 0 ≤ r < b.
Proof. We prove the existence of q, r. The uniqueness is left to the reader. We
may assume a ∈ N. We proceed by contradiction. So assume there exists b and
a ∈ N such that for all q, r ∈ Z with 0 ≤ r < b we have a 6= qb + r. Fix such a
b. We may assume a is minimum with the above property. If a < b we can write
a = 0 × b + a, a contradiction. If a = b we can write a = 1 × a + 0, a contradiction.
If a > b set a0 = a − b. Since a0 < a, there exist q 0 , r ∈ Z such that 0 ≤ r < b and
a0 = q 0 b + r. But then a = qb + r, where q = q 0 + 1, a contradiction. 
Definition 20.5. For a ∈ Z denote hai the set {na; n ∈ Z} of integers divisible
by a. For a, b ∈ Z denote by ha, bi the set {ma + nb; m, n ∈ Z} of all numbers
expressible as a multiple of a plus a multiple of b.
Proposition 20.6. For any integers a, b there exists an integer c such that
ha, bi = hci.
Proof. If a = b = 0 we can take c = 0. Assume a, b are not both 0. Then the
set S = ha, bi ∩ N is non-empty. Let c be the minimum of S. Clearly hci ⊂ ha, bi.
Let us prove that ha, bi ⊂ hci. Let u = ma + nb and let us prove that u ∈ hci. By
Euclidean division u = cq + r with 0 ≤ r < c. We want to show r = 0. Assume
r 6= 0 and seek a contradiction. Write c = m0 a + n0 b. Then r ∈ N and also
r = u − cq = (ma + nb) − (m0 a + n0 b)q = (m − m0 q)a + (n − n0 q)b ∈ ha, bi.
Hence r ∈ S. But r < c. So c is not the minimum of S, a contradiction. 
105
106 20. ARITHMETIC

Proposition 20.7. If a and b are integers and have no common divisor > 1
then there exist integers m and n such that ma + nb = 1.
Proof. By the above Proposition ha, bi = hci for some c ≥ 1. In particular c|a
and c|b. The hypothesis implies c = 1 hence 1 ∈ ha, bi. 
One of the main definitions of number theory is
Definition 20.8. An integer p is prime if p > 1 and if its only positive divisors
are 1 and p.
Proposition 20.9. If p is a prime and a is an integer such that p 6 |a then
there exist integers m, n such that ma + np = 1.
Proof. a and p have no common divisor > 1 and we conclude by Proposition
20.7. 
Proposition 20.10. (Euclid Lemma) If p is a prime and p|ab for integers a
and b then either p|a or p|b.
Proof. Assume p|ab, p 6 |a, p 6 |b, and seek a contradiction. By Proposition 20.9
ma + np = 1 for some integers m, n and m0 b + n0 p = 1 for some integers m0 , n0 . We
get
1 = (ma + np)(m0 b + n0 p) = mm0 ab + p(nm0 + n0 m + nn0 ).
Since p|ab we get p|1, a contradiction. 
Theorem 20.11. (Fundamental Theorem of Arithmetic) Any integer n > 1
can be written uniquely as a product of primes, i.e., there exist primes p1 , p2 , ..., ps ,
where s ≥ 1, such that
n = p1 p2 ...ps .
Moreover any such representation is unique in the following sense: if
p1 p2 ...ps = q1 q2 ...qt
with pi and qj prime and p1 ≤ p2 ≤ ..., q1 ≤ q2 ≤ ... then s = t and p1 = q1 ,
p2 = q2 , ....
Proof. Uniqueness follows from Euclid’s Lemma 20.10. To prove the existence
part let S be the set of all integers > 1 which are not products of primes. We
want to show S = ∅. Assume the contrary and seek a contradiction. Let n be the
minimum of S. Then n is not prime. So n = ab with a, b > 1 integers. So a < n
and b < n. So a 6∈ S and b 6∈ S. So a and b are products of primes. So n is a
product of primes, a contradiction. 
Exercise 20.12. Prove the uniqueness part in the above theorem.
Definition 20.13. Fix an integer m 6= 0. Define a relation ≡m on Z by a ≡m b
if and only if m|a − b. Say a is congruent to b mod m (or modulo m). Instead of
a ≡m b one usually writes (following Gauss):
a ≡ b (mod m).
Example 20.14. 3 ≡ 17 (mod 7).
Exercise 20.15. Prove that ≡m is an equivalence relation. Prove that the
equivalence class a of a consists of all the numbers of the form mb + a where m ∈ Z.
Example 20.16. If m = 7 then 3 = 10 = {..., −4, 3, 10, 17, ...}.
 20. ARITHMETIC 107

Definition 20.17. For the equivalence relation ≡m on Z the set of equivalence

classes Z/ ≡m is denoted by Z/mZ. The elements of Z/mZ are called residue
classes mod m.
Exercise 20.18. Prove that
Z/mZ = {0, 1, ..., m − 1}.
So the residue classes mod m are: 0, 1, ..., m − 1. Hint: Use Euclid division.
Exercise 20.19. Prove that if a ≡ b (mod m) and c ≡ d (mod m) then
a + c ≡ b + d (mod m) and ac ≡ bd (mod m).
Definition 20.20. Define operations +, ×, − on Z/mZ by
a+b=a+b
a × b = ab
−a = −1.
Exercise 20.21. Check that the above definitions are correct, in other words
that if a = a0 and b = b0 then
a + b = a0 + b0
ab = a0 b0
−a = −a0 .
Furthermore check that (Z/mZ, +, ×, −, 0, 1) is a ring.
Definition 20.22. If p is a prime we write Fp in place of Z/pZ.
Exercise 20.23. Prove that Fp is a field. Hint: Use Proposition 20.9.
 CHAPTER 21

Groups

Our next chapters investigate a few topics in algebra. Recall that algebra is the
study of algebraic structures, i.e., sets with operations on them. We already intro-
duced, and constructed, some elementary examples of algebraic structures such as
rings and, in particular, fields. With rings/fields at our disposal one can study some
other fundamental algebraic objects such as groups, vector spaces, polynomials. In
what follows we briefly survey some of these. We begin with groups. In some sense
groups are more fundamental than rings and fields; but in order to be able to look
at more interesting examples we found it convenient to postpone the discussion of
groups until this point. Groups appeared in mathematics in the context of symme-
tries of roots of polynomial equations; cf. the work of Galois that involved finite
groups. Galois’ work inspired Lie who investigated differential equations in place of
polynomial equations; this led to (continuous) Lie groups, in particular groups of
matrices. Groups eventually penetrated most of mathematics and physics (Klein,
Poincaré, Einstein, Cartan, Weyl).
Definition 21.1. A group is a tuple (G, ?,0 , e) consisting of a set G, a binary
operation ? on G, a unary operation 0 on G (write 0 (x) = x0 ), and an element e ∈ G
(called the identity element) such that for any x, y, z ∈ G the following axioms are
satisfied:
1) x ? (y ? z) = (x ? y) ? z;
2) x ? e = e ? x = x;
3) x ? x0 = x0 ? x = e.
If in addition x ? y = y ? x for all x, y ∈ G we say G is commutative (or Abelian in
honor of Abel).
Remark 21.2. For any group G, any element g ∈ G, and any n ∈ Z one defines
g n ∈ G exactly as in Exercise 13.15.
Exercise 21.3. Check the above.
Definition 21.4. Sometimes one writes e = 1, x?y = xy, x0 = x−1 , x?...?x =
n
x (n ≥ 1 times). In the Abelian case one sometimes writes e = 0, x ? y = x + y,
x0 = −x, x ? ... ? x = nx (n ≥ 1 times). These notations depend on the context and
are justified by the following examples.
Example 21.5. If R is a ring then R is an Abelian group with e = 0, x ?
y = x + y, x0 = −x. Hence Z, Z/mZ, Fp , Zp , Q, R, C are groups “with respect to
addition.”
Example 21.6. If R is a field then R× = R\{0} is an Abelian group with
e = 1, x ? y = xy, x0 = x−1 . Hence Q× , R× , C× , F×
p are groups “with respect to
multiplication.”
109
110 21. GROUPS

Example 21.7. The set Perm(X) of bijections σ : X → X from a set X into

itself is a group with e = 1X (the identity map), σ ? τ = σ ◦ τ (composition),
σ −1 =inverse map. If X = {1, ..., n} then one writes Sn = Perm(X) and calls this
group the symmetric group. If σ(1) = i1 , ..., σ(n) = in one usually writes
 
1 2 ... n
σ= .
i1 12 ... in
Exercise 21.8. Compute
   
1 2 3 4 5 1 2 3 4 5
◦ .
3 2 5 4 3 5 4 2 1 3
Also compute
 −1
1 2 3 4 5
.
3 2 5 4 3
Example 21.9. A 2 × 2 matrix with coefficients in a field R is a map
A : {1, 2} × {1, 2} → R.
If the map is given by
A(1, 1) = a
A(1, 2) = b
A(2, 1) = c
A(2, 2) = d
we write A as  
a b
A= .
c d
Define the sum and the product of two matrices by
  0 0  
a + a0 b + b0
 
a b a b
+ = ,
c d c0 d0 c + c0 d + d0
  0 0  
aa0 + bc0 ab0 + bd0
 
a b a b
· = .
c d c0 d0 ca0 + dc0 cb0 + dd0
 
a b
Define the product of an element r ∈ R with a matrix A = by
c d
   
a b ra rb
r· = .
c d rc rd
 
a b
For a matrix A = define its determinant by
c d
det(A) = ad − bc.
Say that A is invertible if det(A) 6= 0 and setting δ = det(A) define the inverse of
A by  
−1 −1 d −b
A =δ .
−c a
Define the identity matrix by  
1 0
I=
0 1
 21. GROUPS 111

and the zero matrix by  

0 0
O= .
0 0
Let M2 (R) be the set of all matrices and GL2 (R) be the set of all invertible matrices.
Then the following are true:
1) M2 (R) is a group with respect to addition of matrices;
2) GL2 (R) is a group with respect to multiplication of matrices; it is called the
general linear group of 2 × 2 matrices;
3) (A + B)C = AC + BC and C(A + B) = CA + CB for any matrices A, B, C;
4) There exist matrices A, B such that AB 6= BA;
5) det(AB) = det(A) · det(B).
Exercise 21.10. Prove 1), 2), 3), 4), 5) above.
Example 21.11. Groups are examples of algebraic structures so there is a well-
defined notion of homomorphism of groups (or group homomorphism). According
to the general definition a group homomorphism is a map between the two groups
F : G → G0 such that for all a, b ∈ G:
1) F (a ? b) = F (a) ?0 F (b),
2) F (a−1 ) = F (a)−1 (this is automatic !),
3) F (e) = e0 (this is, again, automatic !).
Here ? and ?0 are the operations on G and G0 ; similarly e and e0 are the
corresponding identity elements.
Definition 21.12. A subset H of a group G is called a subgroup if
1) For all a, b ∈ H we have a ? b ∈ H.
2) For all a ∈ H we have a−1 ∈ H.
3) e ∈ H.
Exercise 21.13. Show that if H is a subgroup of G then H, with the natural
operation induced from G, is a group.
Exercise 21.14.
1) Z is a subgroup of Q.
2) Q is a subgroup of R.
3) R is a subgroup of C.
4) If R is a field then the set
 
a b
SL2 (R) = { ; a, b ∈ R, ad − bc = 1}
c d
is a subgroup of GL2 (R); it is called the special linear group.
5) If R is a field then the set
 
a b
SO2 (R) = { ; a, b ∈ R, a2 + b2 = 1}
−b a
is a subgroup of SL2 (R); it is called the special orthogonal group.
Definition 21.15. If F : G → G0 is a group homomorphism define the kernel
of F ,
Ker F = {a ∈ G; F (a) = e0 }
and the image of F :
Im F = {b ∈ G0 ; ∃a ∈ G, F (a) = b}.
112 21. GROUPS

Exercise 21.16. Prove that Ker F is a subgroup of G and Im F is a subgroup

of G0 .
 CHAPTER 22

Order

We continue our investigation of groups and introduce the concept of order of

elements in a group. (This has nothing to do with the word order used in the phrase
order relations.)
Definition 22.1. Let G be a group and g ∈ G; we denote by hgi the set of all
elements a ∈ G for which there exists n ∈ Z such that a = g n .
Exercise 22.2. Prove that hgi is a subgroup of G. We call hgi the subgroup
generated by g.
Definition 22.3. We say that a group G is cyclic if there exists g ∈ G such
that G = hgi; g is called a generator of G.
Example 22.4. Z is cyclic. 1 is a generator of Z; −1 is also a generator of Z.
Exercise 22.5. Prove that Q is not cyclic.
Definition 22.6. Let G be a group and g ∈ G. We say the order of g is infinite
if g n 6= e for all n ∈ N. We say the order of g is n ∈ N if:
1) g n = e;
2) g k 6= e for all k ∈ N with k < n.
We denote by o(g) the order of g.
Definition 22.7. The order of a finite group G is the cardinality |G| of G.
Exercise 22.8. The order o(g) of g equals the order |hgi| of hgi.
Exercise 22.9. g has order n ∈ N if and only if:
1’) g n = e;
2’) If g N = e for some N ∈ N then n|N .
Hint: If 1’) and 2’) above hold then clearly g has order n. Conversely if g has
order n then 1’) clearly holds. To check that 2’) holds use Euclidean division to
write N = nq + r with 0 ≤ r < n. Then g r = (g n )q g r = g N = e. By condition 2)
in the definition of order r = 0 hence n|N .
In what follows we say that two integers are coprime if they have no common
divisor > 1.
Proposition 22.10. Assume a, b are two elements in a group such that ab = ba
and assume o(a) and o(b) are coprime. Then
o(ab) = o(a)o(b).
Proof. Set k = o(a), l = o(b). Clearly, since ab = ba we have
(ab)kl ≡ (ak )l (bl )k = e.
113
114 22. ORDER

Now assume (ab)N = e. Raising to power l we get aN l bN l = e, hence aN l = e,

hence, by Exercise 22.9, k|N l. Since k and l are coprime k|N (by the Fundamental
Theorem of Arithmetic). In a similar way raising (ab)N = e to power k we get
aN k bN k = e, hence bN k = e, hence l|N k, hence l|N . Again, since k and l are
coprime, l|N and k|N imply kl|N and we are done. 
Exercise 22.11. Prove that if o(a) = kl then o(ak ) = l.
Theorem 22.12. (Lagrange) If H is a subgroup of a finite group G then the
order of H divides the order of G: if n = |H|, m = |G| then n|m. In particular if
a ∈ G then the order o(a) of a divides the order |G| of the group. So if n = |G|
then an = e.
Proof. For each g ∈ G we let gH be the set of all elements of G of the form gh
with h ∈ H. Let π : G → P(G) be the map π(g) = gH ∈ P(G). Let X = π(G) and
let σ : X → G be any map such that π ◦ σ is the identity of X. (The existence of σ
follows by induction.) We claim that the map
(22.1) X × H → G, (X, h) 7→ σ(X)h, X ∈ X, h ∈ H
is a bijection. Assuming the claim for a moment note that the claim implies
|X| × |H| = |G|,
from which the theorem follows. Let us check the claim. To prove that 22.1 is
surjective let g ∈ G. Let g 0 = σ(gH). Then g 0 H = π(g 0 ) = π(σ(gH)) = gH. So
there exists h ∈ H such that g 0 h = ge = g; hence g = σ(gH)h which ends the proof
of surjectivity. We leave the proof of injectivity to the reader. 
Exercise 22.13. Check the injectivity of 22.1.
Theorem 22.14. (Fermat’s Little Theorem) For any a ∈ Z and any prime p
we have
ap ≡ a (mod p).
Proof. If p|a this is clear. If p 6 |a let a be the image of a in F×p . By Lagrange’s
theorem applied to the group F× p we have ap−1
= 1. Hence a p−1
≡ 1 (mod p). So
ap ≡ a (mod p). 
 CHAPTER 23

Vectors

Vectors implicitly appeared in a number of contexts such as mechanics (Galileo,

Newton, etc.), hypercomplex numbers (Hamilton, Cayley, etc.), algebraic number
theory (Dirichlet, Kummer, Eisenstein, Kronecker, etc.), and analysis (Hilbert,
Banach, etc.). They are now a basic concept in linear algebra which is itself part
of abstract algebra.
Definition 23.1. Let R be a field. A vector space is an Abelian group
(V, +, −, 0) together with a map R × V → V , (a, v) 7→ av satisfying the follow-
ing conditions for all a, b ∈ R and all u, v ∈ V :
1) (a + b)v = av + bv;
2) a(u + v) = au + av;
3) a(bv) = (ab)v;
4) 1v = v.
The elements of V are called vectors.
Example 23.2. Rn is a vector space over R viewed with the operations
(a1 , ..., an ) + (b1 , ..., bn ) = (a1 + b1 , ..., an + bn ),
−(a1 , ..., an ) = (−a1 , ..., −an ),
c(a1 , ..., an ) = (ca1 , ..., can ).
Definition 23.3. The elements u1 , ..., un ∈ V are linearly independent if when-
ever a1 , ..., an ∈ R satisfies (a1 , ..., an ) 6= (0, ..., 0) it follows that a1 u1 +...+an un 6= 0.
Definition 23.4. The elements u1 , ..., un ∈ V generate V if for any u ∈ V
there exist a1 , ..., an ∈ R such that u = a1 u1 + ..., an un . (We also say that u is a
linear combination of u1 , ..., un .)
Definition 23.5. The elements u1 , ..., un ∈ V are a basis of V if they are
linearly independent and generate V .
Exercise 23.6.
1) Show that (−1, 1, 0) and (0, 1, −1) are linearly independent in R3 but they
do not generate R3 .
2) Show that (−1, 1, 0), (0, 1, −1), (1, 0, 1), (0, 2, −1) generate R3 but are not
linearly independent in R3 .
3) Show that (−1, 1, 0), (0, 1, −1), (1, 0, 1) is a basis in R3 .
Exercise 23.7. If V has a basis u1 , ..., un then the map Rn → V , (a1 , ..., an ) 7→
a1 u1 + ... + an un is bijective. Hint: Directly from definitions.
Exercise 23.8. If V is generated by u1 , ..., un then V has a basis consisting of at
most n elements. Hint: Considering a subset of {u1 , ..., un } minimal with the prop-
erty that it generates V we may assume that any subset obtained from {u1 , ..., un }
115
116 23. VECTORS

does not generate V . We claim that u1 , ..., un are linearly independent. Assume
not. Hence there exists (a1 , ..., an ) 6= (0, ..., 0) such that a1 u1 + ... + an un = 0.
We may assume a1 = 1. Then one checks that u2 , ..., un generate V , contradicting
minimality.
Exercise 23.9. Assume R = Fp and V has a basis with n elements. Then
|V | = pn .
Theorem 23.10. If V has a basis u1 , ..., un and a basis v1 , ..., vm then n = m.
Proof. We prove m ≤ n; similarly one has n ≤ m. Assume m > n and seek
a contradiction. Since u1 , ..., un generate V we may write v1 = a1 u1 + ... + an un
with not all a1 , ..., an zero. Renumbering u1 , ..., un we may assume a1 6= 0. Hence
v1 , u2 , ..., un generates V . Hence v2 = b1 v1 + b2 u2 + ... + bn un . But not all b2 , ..., bn
can be zero because v1 , v2 are linearly independent. So renumbering u2 , ..., un
we may assume b2 6= 0. So v1 , v2 , u3 , ..., un generates V . Continuing (one needs
induction) we get that v1 , v2 , ..., vn generates V . So vn+1 = d1 vn + ... + dn vn . But
this contradicts the fact that v1 , ..., vm are linearly independent. 
Exercise 23.11. Give a quick proof of the above theorem in case R = Fp .
Hint: We have pn = pm hence n = m.
Definition 23.12. We say V is finite dimensional (or that it has a finite basis)
if there exists a basis u1 , ..., un of V . Then we define the dimension of V to be n;
write dim V = n. (The definition is correct due to Theorem 23.10.)
Definition 23.13. If V and W are vector spaces a map F : V → W is called
linear if for all a ∈ K, u, v ∈ V we have:
1) F (au) = aF (u),
2) F (u + v) = F (u) + F (v).
Example 23.14. If a, b, c, d, e, f ∈ R then the map F : R3 → R2 given by
F (u, v, w) = (au + bv + cw, du + ev + f w)
is a linear map.
Exercise 23.15. Prove that if F : V → W is a linear map of vector spaces
then V 0 = F −1 (0) and V 00 = F (V ) are vector spaces (with respect to the obvious
operations). If in addition V and W are finite dimensional then V 0 and V 00 are
finite dimensional and
dim V = dim V 0 + dim V 00 .
Hint: Construct corresponding bases.
Exercise 23.16. Give an example of a vector space that is not finite dimen-
sional.
 CHAPTER 24

Matrices

Matrices appeared in the context of linear systems of equations and were studied
in the work of Leibniz, Cramer, Cayley, Eisenstein, Hamilton, Sylvester, Jordan,
etc. They were later rediscovered and applied in the context of Heisenberg’s matrix
mechanics. Nowadays they are a standard concept in linear algebra courses.
Definition 24.1. Let m, n ∈ N. An m × n matrix with coefficients in a field
R is a map
A : {1, ..., m} × {1, ..., n} → R.
If A(i, j) = aij for 1 ≤ i ≤ m, 1 ≤ j ≤ n then we write
 
a11 ... a1n
A = (aij ) =  ... ... ...  .
am1 ... amn
We denote by
Rm×n = Mm×n (R)
the set of all m × n matrices. We also write Mn (R) = Mn×n (R). Note that R1×n
identifies with Rn ; its elements are of the form
(a1 , ..., an )
and are called row matrices. Similarly the elements of Rm×1 are of the form
 
a1
 ... 
 
 ... 
am
and are called column matrices. If A = (aij ) ∈ Rm×n then
   
a11 a1n
 ... 
u1 =   , ..., un =  ... 
 
 ...   ... 
am1 amn
are called the columns of A and we also write
A = (u1 , ..., un ).
Similarly
(a11 , ..., a1n ), ..., (am1 , ..., amn )
are called the rows of A.
117
118 24. MATRICES

Definition 24.2. Let 0 ∈ Rm×n the matrix 0 = (zij ) with zij = 0 for all i, j;
0 is called the zero matrix. Let I ∈ Rn×n the matrix I = (δij ) where δii = 1 for
all i and δij = 0 for all i 6= j; I is called the identity matrix and δij is called the
Kronecker symbol.
Definition 24.3. If A = (aij ), B = (bij ) ∈ Rm×n we define the sum
C = A + B ∈ Rm×n
as
C = (cij ), cij = aij + bij .
If A = (ais ) ∈ Rm×k , B = (bsj ) ∈ Rk×n , we define the product
C = AB ∈ Rm×n
as
k
X
C = (cij ), cij = ais bsj .
s=1

Exercise 24.4. Prove that:

1) Rm×n is a group with respect to +.
2) A(BC) = (AB)C for all A ∈ Rm×k , B ∈ Rk×l , C ∈ Rl×n .
3) A(B + C) = AB + AC for all A ∈ Rm×k and B, C ∈ Rk×n .
4) (B + C)A = BA + CA for all B, C ∈ Rm×k and A ∈ Rk×n .
5) AI = IA for all A ∈ Rn×n .
Exercise 24.5. If A ∈ Rm×k , B ∈ Rk×n , and the columns of B are b1 , ..., bn ∈
Rk×1 then the columns of AB are Ab1 , ..., Abn ∈ Rm×1 (where Abi is the product
of the matrices A and bi ). In other words
B = (b1 , ..., bn ) ⇒ AB = (Ab1 , ..., Abn ).
Definition 24.6.
     
1 0 0
 0   1
 , ...,  0 
  
 ...  ,  ...
  
  ... 
0 0 1
is called the standard basis of Rm×1
Exercise 24.7. Prove that the above is indeed a basis of Rm×1 .
Here is the link between linear maps and matrices:
Definition 24.8. If F : V → W is a linear map of vector spaces and v1 , ..., vn
and w1 , ..., wm are bases of V and W , respectively, then for j = 1, ..., n one can
write uniquely
m
X
F (vj ) = aij wi .
i=1
m×n
The matrix A = (aij ) ∈ R is called the matrix of F with respect to the bases
v1 , ..., vn and w1 , ..., wm .
 24. MATRICES 119

Exercise 24.9. Consider a matrix A = (aij ) ∈ Rm×n and consider the map
F : Rn×1 → Rm×1 , F (u) = Au (product of matrices).
Then the matrix of F with respect to the canonical bases of Rn×1 and Rm×1 is A
itself.
Hint: Let e1 , ..., en be the standard basis of Rn×1 and let f 1 , ..., f m be the
standard basis of Rm×1 . Then
   
  1 a11
a11 ... a1n  0   a21 
F (e1 ) = Ae1 =  ... ... ...   1
 ...  =  ...  = a11 f + ... + am1 f .
   m

am1 ... amn

0 am1
A similar computation can be done for e2 , ..., en .
Exercise 24.10. Let F : Rn×1 → Rm×1 be a linear map and let A ∈ Rm×n
be the matrix of F with respect to the standard bases. Then for all u ∈ Rn×1 we
have F (u) = Au.
Exercise 24.11. Let G : Rn×1 → Rk×1 and let F : Rk×1 → Rm×1 be linear
maps. Let A be the matrix of F with respect to standard bases and let B be
the matrix of G with respect to the standard bases. Then the matrix of F ◦ G
with respect to the standard bases is AB (product of matrices). Hint: F (G(u)) =
A(Bu) = (AB)u.
Definition 24.12. If A = (aij ) ∈ Rm×n is a matrix one defines the transpose
of A as the matrix At = (a0ij ) ∈ Rn×m where a0ij = aji .
Example 24.13.  
 t a d
a b c
= b e .
d e f
c f
Exercise 24.14. Prove that:
1) (A + B)t = At + B t ;
2) (AB)t = B t At ;
3) I t = I.
 CHAPTER 25

Determinants

A fundamental concept in the theory of matrices is that of determinant of a

matrix. The main results are due to Cauchy, Kronecker, and Weierstrass. In spite
of the computational aspect of this concept the best way to approach it is via an
axiomatic method as follows.
Definition 25.1. Let V and W be vector spaces over a field R and let
f : V n = V × ... × V → W
be a map. We say f is multilinear if for any v1 , ..., vn ∈ V and any i ∈ {1, ..., n} we
have:
1) If vi = vi0 + vi00 then
f (v1 , ..., vn ) = f (v1 , ..., vi0 , ..., vn ) + f (v1 , ..., vi00 , ..., vn ).
2) If vi = cvi0 then
f (v1 , ..., vn ) = cf (v1 , ..., vi0 , ..., vn ).
Example 25.2. f : R3×1 × R3×1 → R defined by
   
a d
f  b  ,  e  = ad + 3bf − ce
c f
is multilinear.
Definition 25.3. A multilinear map f : V n = V × ... × V → W is called
alternating if whenever v1 , ..., vn ∈ V and there exist indices i 6= j such that vi = vj
we have f (v1 , ..., vn ) = 0.
Example 25.4. f in Example 25.2 is not alternating. On the other hand
g : R2×1 × R2×1 → R defined by
     
a b a b
f , = 2ad − 2bc = 2 det
c d c d
is alternating.
Lemma 25.5. If f : V n → W is multilinear alternating and v1 , ..., vn ∈ V then
for any indices i < j we have
f (v1 , ..., vi , ..., vj , ..., vn ) = −f (v1 , ..., vj , ..., vi , ..., vn ).
Here v1 , ..., vj , ..., vi , ..., vn is obtained from v1 , ..., vi , ..., vj , ..., vn by replacing
vi with vj and vj with vi while leaving all the other vs unchanged.
Proof. We have
121
122 25. DETERMINANTS

f (v1 , ..., vi + vj , ..., vi + vj , ..., vn ) = f (v1 , ..., vi , ..., vi , ..., vn )

+f (v1 , ..., vi , ..., vj , ..., vn )
+f (v1 , ..., vj , ..., vi , ..., vn )
+f (v1 , ..., vj , ..., vj , ..., vn ).
Hence
0 = f (v1 , ..., vi , ..., vj , ..., vn ) + f (v1 , ..., vj , ..., vi , ..., vn ).

Exercise 25.6. Let σ : {1, ..., n} → {1, ..., n} be a bijection. Then there exists
(σ) ∈ {−1, 1} with the following property. Let f : V n → W be any multilinear
alternating map and v1 , ..., vn ∈ V . Then
f (vσ(1) , ..., vσ(n) ) = (σ) · f (v1 , ..., vn ).
Hint: Induction on n. For the induction step distinguish two cases: σ(1) = 1 and
σ(1) 6= 1. In the first case one concludes directly by the induction hypothesis. The
second case can be reduced to the first case via Lemma 25.5.
We identify (Rn×1 )n with Rn×n by identifying a tuple of columns (b1 , ..., bn )
with the n × n matrix whose columns are b1 , ..., bn . We denote I = In the identity
n × n matrix.
Lemma 25.7. There exists a multilinear alternating map
f : Rn×n → R
such that f (I) = 1.
Proof. We proceed by induction on n. For n we take f (a) = a. Assume we
constructed a multilinear alternating map
fn−1 : R(n−1)×(n−1) → R
such that fn−1 (In−1 ) = 1. Let A = (aij ) be an n × n matrix and let Aij be the
(n − 1) × (n − 1) matrix obtained from A by deleting the i-th row and the j-th
column. Fix i and define
n
X
fn (A) = (−1)i+j aij fn−1 (Aij ).
j=1

One easily checks that fn is multilinear, alternating, and takes value 1 on the
identity matrix In . 
Exercise 25.8. Check the last sentence in the proof above.
Lemma 25.9. If f and g are multilinear alternating maps Rn×n → R and
f (I) 6= 0 then there exists c ∈ R such that g(A) = cf (A) for all A.
Proof. Let A = (aij ). Let e1 , ..., en be the standard basis of Rn×1 . Then
!
X X X X
i1 in
g(A) = g ai1 1 e , ..., ain n e = ... ai1 1 ...ain n g(ei1 , ..., ein ).
i1 in i1 in
 25. DETERMINANTS 123

The terms for which i1 , ..., in are not distinct are zero. The terms for which i1 , ..., in
are distinct are indexed by permutations σ. By Exercise 25.6 we get
!
X
g(A) = (σ)aσ(1)1 ...aσ(n)n g(I).
σ
A similar formula holds for f (A) and the Lemma follows. 
By Lemmas 25.7 and 25.9 we get:
Theorem 25.10. There exists a unique multilinear alternating map (called de-
terminant)
det : Rn×n → R
such that det(I) = 1.
Exercise 25.11. Using the notation in the proof of Lemma 25.7 prove that:
1) For all i we have
n
X
det(A) = (−1)i+j aij det(Aij ).
j=1

2) For all j we have

n
X
det(A) = (−1)i+j aij det(Aij ).
i=1
Hint: Use Lemma 25.9.
We also have:
Theorem 25.12. For any two matrices A, B ∈ Rn×n we have
det(AB) = det(A) det(B).
Proof. Consider the multilinear alternating map f : Rn×n → R defined by
f (u1 , ..., un ) = det(Au1 , ..., Aun )
for u1 , ..., un ∈ Rn×1 . By Lemma 25.9 there exists c ∈ R such that
f (u1 , ..., un ) = c · det(u1 , ..., un ).
Hence
det(Au1 , ..., Aun ) = c · det(u1 , ..., un ).
Setting ui = ei we get det(A) = c · det(I) = c. Setting ui = bi , the columns of B,
we get det(AB) = c · det(B) and the theorem is proved. 
Exercise 25.13. Prove that (στ ) = (σ)(τ ) for any permutations σ, τ ∈ Sn ;
in other words  : Sn → {1, −1} is a group homomorphism.
Exercise 25.14. Prove that if A ∈ Rn×n is a matrix such that det(A) 6= 0
then A is invertible i.e., there exists B ∈ Rn×n such that AB = BA = I.
Hint: Define B = (bij ) where
bij = (−1)i+j det(Aji )
(notation as in Lemma 25.7). Prove that AB = BA = I using Exercise 25.11.
Exercise 25.15. Prove that if A ∈ Rn×n is a matrix then det(A) = det(At ).
Hint. Use Lemma 25.9.
124 25. DETERMINANTS

Exercise 25.16. Let R be a field.

1) Prove that the set GLn (R) = {A ∈ Rn×n ; det(A) 6= 0} is a group with
respect to multiplication; GLn (R) is called the general linear group.
2) Prove that the set SLn (R) = {A ∈ Rn×n ; det(A) = 1} is a subgroup of
GLn (R); SLn (R) is called the special linear group.
3) Prove that the set SOn (R) = {A ∈ Rn×n ; det(A) = 1, AAt = I} is a
subgroup of SLn (R); SOn (R) is called the special orthogonal group.
Check that for n = 2 the above correspond to the previously defined groups
GL2 (R), SL2 (R), SO2 (R).
Exercise 25.17.
1) Prove that if a linear map F : V → W is bijective then its inverse F −1 :
W → V is also linear. Such a map will be called an isomorphism (of vector spaces).
2) Prove that the set GL(V ) of all isomorphisms V → V is a group under
composition.
3) Assume V has a basis v1 , ..., vn and consider the map GL(V ) → GLn (R),
F 7→ AF where AF is the matrix of F with respect to v1 , ..., vn . Prove that
GL(V ) → GLn (R) is an isomorphism of groups.
 CHAPTER 26

Polynomials

Determining the roots of polynomials was one of the most important motivat-
ing problems in the development of algebra, especially in the work of Cardano,
Lagrange, Gauss, Abel, and Galois. Here we introduce polynomials and discuss
some basic facts about their roots.
Definition 26.1. Let R be a ring. We define the ring of polynomials R[x]
in one variable with coefficients in R as follows. An element of R[x] is a map
f : N ∪ {0} → R, i 7→ ai with the property that there exists i0 ∈ N such that for
all i ≥ i0 we have ai = 0; we also write such a map as
f = (a0 , a1 , a2 , a3 , ...).
We define 0, 1 ∈ R[x] by
0 = (0, 0, 0, 0, ...),
1 = (1, 0, 0, 0, ...).
If f is as above and g = (b0 , b1 , b2 , b3 , ...) then addition and multiplication are
defined by
f +g = (a0 + b0 , a1 + b1 , a2 + b2 , a3 + b3 , ...),
fg = (a0 b0 , a0 b1 + a1 b0 , a0 b2 + a1 b1 + a2 b0 , a0 b3 + a1 b2 + a2 b1 + a3 b0 , ...).
We define the degree of f = (a0 , a1 , a2 , a3 , ...) as
deg(f ) = min{i; ai 6= 0}
if f 6= 0 and deg(0) = 0. We also define
x = (0, 1, 0, 0, ...)
and we write
a = (a, 0, 0, 0, ...)
for any a ∈ R.
Exercise 26.2.
1) Prove that R[x] with the operations above is a ring.
2) Prove that the map R → R[x], a 7→ a = (a, 0, 0, 0, ...) is a ring homomor-
phism.
3) Prove that x2 = (0, 0, 1, 0, ...), x3 = (0, 0, 0, 1, 0, ...), etc.
4) Prove that if f = (a0 , a1 , a2 , a3 , ...) then
f = an xn + ... + a1 x + a0
where n = deg(f ). (We also write f = f (x) but we DO NOT SEE f (x) as a
function; this is just a notation.)
125
126 26. POLYNOMIALS

Example 26.3. If R = Z then

(3x2 + 5x + 1)(8x3 + 7x2 − 2x − 1) =
= (3 × 8)x5 + (3 × 7 + 5 × 8)x4 + (3 × (−2) + 5 × 7 + 1 × 8)x3 + ....
Definition 26.4. For any b ∈ R we define an element f (b) ∈ R by
f (b) = an bn + ... + a1 b + a0 .
So for any polynomial f ∈ R[x] we can define a map (called the polynomial map
defined by the polynomial f ):
R → R, b 7→ f (b).
(The polynomial map defined by f should not be confused with the polynomial f
itself; they are two different entities.) An element b ∈ R is called a root of f (or a
zero of f ) if f (b) = 0. (Sometimes we say “a root in R” instead of “a root.”)
Example 26.5. If R = F2 and we consider the polynomial f (x) = x2 +x ∈ R[x]
then f = (0, 1, 1, 0, ...) 6= (0, 0, 0, ...) = 0 as an element of R[x]; but the polynomial
2 2
map R → R defined by f sends 1 7→ 1 + 1 = 0 and 0 7→ 0 + 0 = 0 so this map is
the constant map with value 0. This shows that different polynomials (in our case
f and 0) can define the same polynomial map.
Exercise 26.6. Let R = R. Show that
2
1) x
p√ + 1 has no root in R.
2) 3 + 1 is a root of x4 − 2x2 − 1 = 0.
Remark 26.7. One can ask if any root in C of a polynomial with coefficients
in Q can be expressed, using (possibly iterated) radicals of rational numbers. The
answer to this is negative as shown by Galois in the early 19th century.
Exercise 26.8. Let R = C. Show that
1) i is a root of x2 + 1 in C.
2) 1+i
√ is a root of x4 + 1 = 0 in C.
2

Remark 26.9. Leibniz mistakenly thought that the polynomial x4 + 1 should

have no root in C.
Exercise 26.10. Let R = F7 . Show that:
1) x2 + 1 has no root in R.
2) 2 is a root of x3 − x + 1 in R.
Exercise 26.11. Let R = F5 . Show that:
1) 2 is a root of x2 + 1 in R.
2) x5 − x + 1 has no root in F5 .
The study of roots of polynomial functions is one of the main concerns of
algebra. Here are two of the main basic theorems about roots.
Theorem 26.12. (Lagrange) If R is a field then any polynomial of degree d ≥ 1
has at most d roots in R.
Theorem 26.13. (Fundamental Theorem of Algebra, Gauss) If R = C is the
complex field then any polynomial of degree ≥ 1 has at least one root in C.
In what follows we prove Theorem 26.12. (Theorem 26.13 is beyond the scope
of this course.) We need a preparation.
 26. POLYNOMIALS 127

Definition 26.14. A polynomial f (x) = an xn + ... + a1 x + a0 of degree n is

monic if an = 1.
Proposition 26.15. (Long division) Let f (x), g(x) ∈ R[x] with g(x) monic of
degree ≥ 1. Then there exist unique q(x), r(x) ∈ R[x] such that
f (x) = g(x)q(x) + r(x)
and deg(r) < deg(g).
Proof. Fix g (of degree m) and let us prove by induction on n that the statement
above is true if deg(f ) ≤ n. The case deg(f ) = 0 is clear because we can then take
q(x) = 0 and r(x) = f (x). For the induction step we may take f of degree n and
let f (x) = an xn + ... + a0 , an 6= 0. We may assume n ≥ m. Then
deg(f − an xn−m g) ≤ n − 1
so by the induction hypothesis
f (x) − an xn−m g(x) = g(x)q(x) + r(x)
with deg(r) < m. So
f (x) = g(x)(an xn−m + q(x)) + r(x)
and we are done. 
Proof of Theorem 26.12. Assume there exists a polynomial f of degree d ≥ 1
that has d + 1 roots. Choose f such that d is minimal and seek a contradiction.
Let a1 , ..., ad+1 ∈ R be distinct roots of f . By Long Division we can write
f (x) = (x − ad+1 )g(x) + r(x)
with deg(r) < deg(x − ad+1 ) = 1. So deg(r) = 0 i.e., r(x) = c ∈ R. Since
f (ad+1 ) = 0 we get r(x) = 0 hence c = 0. Since 0 = f (ak ) = (ak −ad+1 )g(ak )+c for
k = 1, ..., d it follows that 0 = (ak − ad+1 )g(ak ). Since R is a field and ak − ad+1 6= 0
for k = 1, ..., d it follows that g(ak ) = 0 for k = 1, ..., d. But deg(g) = d − 1 which
contradicts the minimality of d. 
Definition 26.16. A number α ∈ C is called algebraic if there exists a poly-
nomial f (x) = an xn + ... + a1 x + a0 of degree n ≥ 1 with coefficients in Q such that
f (α) = 0. A number α ∈ C is called transcendental if it is not algebraic.
√
Example 26.17. 2 is algebraic because it is a root of x2 − 2.
p√ √
Exercise 26.18. Prove that 3 + 4 + 5 7 is algebraic.
Remark 26.19. It is not clear that transcendental numbers exist. We will
check that later.
Exercise 26.20. Prove that the set of algebraic numbers in C is countable.
Remark 26.21. The main problems about roots are:
1) Find the number of roots; in case R = Fp this leads to some of the most
subtle problems in number theory.
2) Understand when roots of polynomials with rational coefficients, say, can be
expressed by radicals; this leads to Galois theory.
128 26. POLYNOMIALS

Definition 26.22. Let R be a ring. We defined the ring of polynomials R[x]

in one variable x with coefficients in R. Now R[x] is again a ring so we can consider
the ring of polynomials R[x][y] in one variable y with coefficients in R[x] which we
simply denote by R[x, y] and refer to as the ring of polynomials in two variables
x, y with coefficients in R. Again R[x, y] is a ring so we can consider the ring of
polynomials R[x, y][z] in one variable z with coefficients in R[x, y] which we denote
by R[x, y, z] and which we refer to as the ring of polynomials in 3 variables x, y, z
with coefficients in R, etc.
Example 26.23.
3x y z − x8 + x4 y 9 z 2 + 5xyz 2 = ((x4 )y 9 + (5x)y)z 2 − ((3x7 )y 4 )z − (x8 ) ∈ Z[x, y, z].
7 4
 CHAPTER 27

Congruences

We discuss here polynomial congruences which lie at the heart of number theory.
The main results below are due to Fermat, Lagrange, Euler, and Gauss.
Definition 27.1. Let f (x) ∈ Z[x] be a polynomial and p a prime. An integer
c ∈ Z is called a root of f (x) mod p (or a solution to the congruence f (x) ≡
0 (mod p)) if f (c) ≡ 0 (mod p); in other words if p|f (c). Let f ∈ Fp [x] be the
polynomial obtained from f ∈ Z[x] by replacing the coefficients of f with their
images in Fp . Then c is a root of f mod p if and only if the image c of c in Fp is
a root of f . We denote by Np (f ) the number of roots of f (x) mod p contained in
{0, 1, ..., p − 1}; equivalently Np (f ) is the number of roots of f in Fp . If f, g are
polynomials in Z[x] we write Np (f = g) for Np (f − g). If Zp (f ) is the set of roots
of f in Fp then of course Np (f ) = |Zp (f )|.
Exercise 27.2.
1) 3 is a root of x3 + x − 13 mod 17.
2) Any integer a is a root of xp − x mod p; this is Fermat’s Little Theorem. In
particular Np (xp − x) = p, Np (xp−1 = 1) = p − 1.
3) Np (ax − b) = 1 if p 6 |a.
4) Np (x2 = 1) = 2 if p 6= 2.
Proposition 27.3. For any two polynomials f, g ∈ Z[x] we have
Np (f g) ≤ Np (f ) + Np (g).
Proof. Clearly Zp (f g) ⊂ Zp (f ) ∪ Zp (g). Hence
|Zp (f g)| = |Zp (f ) ∪ Zp (g)| ≤ |Zp (f )| + |Zp (g)|.

Exercise 27.4. Consider the polynomials
f (x) = xp−1 − 1 and g(x) = (x − 1)(x − 2)...(x − p + 1) ∈ Z[x].
Prove that all the coefficients of the polynomial f (x) − g(x) are divisible by p.
Conclude that p divides the sums
p−1
X
a = 1 + 2 + 3 + ... + (p − 1)
a=1

and
X
ab = 1×2+1×3×...1×(p−1)+2×3+...+2×(p−1)+...+(p−2)×(p−1).
1≤a<b≤p−1

129
130 27. CONGRUENCES

Exercise 27.5. Assume p ≥ 5 is a prime. Prove that the numerator of any

fraction that is equal to
1 1 1
1 + + + ... +
2 3 p−1
is divisible by p2 .
Remark 27.6. Fix a polynomial f (x) ∈ Z[x]. Some of the deepest problems
and theorems in number theory can be formulated as special cases of the following
two problems:

1) Understand the set of primes p such that the congruence f (x) ≡ 0 (mod p)
has a solution or, equivalently, such that p|f (c) for some c ∈ Z.

2) Understand the set of primes p such that p = f (c) for some c ∈ Z.

In regards to problem 1) one would like more generally to understand the

function whose value at a prime p is the number Np (f ). In particular one would like
to understand the set of all primes p such that Np (f ) = k for a given k (equivalently
such that the congruence f (x) ≡ 0 (mod p) has k solutions in {0, 1, ..., p − 1}). We
note that if deg(f ) = 1 the problem is trivial. For deg(f ) = 2 the problem is already
highly non-trivial although a complete answer was given by Gauss in his Quadratic
Reciprocity Law (to be proved later). For the quadratic polynomial f (x) = x2 + 1,
for instance, we will prove below (without using quadratic reciprocity) that p|f (c)
for some c if and only if p is of the form 4k+1. For deg(f ) arbitrary the problem (and
its generalizations for polynomials f (x, y, z, ...) of several variables) is essentially
open and part of an array of tantalizing conjectures (part of the Langlands program)
that link the function Np (f ) to Fourier analysis and the theory of complex analytic
functions. This is beyond the scope of our course.
In regards to problem 2), by a theorem of Dirichlet, for any linear polynomial
f (x) = ax + b for which a and b are coprime there exist infinitely many integers
k such that f (k) is prime. But it is not known, for instance, if there are infinitely
many integers k such that f (k) is prime when f (x) is a quadratic polynomial such
as f (x) = x2 + 1. Problem 2) has an obvious analogue for polynomials in several
variables.
The following is a direct consequence of Lagrange’s Theorem 26.12:
Corollary 27.7. Assume p ≡ 1 (mod d). Then Np (xd − 1) = d.
Proof. By Lagrange’s Theorem Np (xd − 1) ≤ d. Assume Np (xd − 1) < d and
seek a contradiction. If p − 1 = kd then xp−1 − 1 = (xd − 1)g(x) where
g(x) = xd(k−1) + xd(k−2) + ... + xd + 1.
Since by Lagrange’s Theorem Np (g) ≤ d(k − 1) we get
p−1 = Np (xp−1 −1) = Np ((xd −1)g) ≤ Np (xd −1)+Np (g) < d+d(k−1) = dk = p−1,
a contradiction. 
Corollary 27.8.
1) If p ≡ 1 (mod 4) then Np (x2 = −1) = 2. Equivalently any prime p of the
form 4k + 1 divides some number of the form c2 + 1 where c is an integer.
2) If p ≡ 3 (mod 4) then Np (x2 = −1) = 0. Equivalently no prime p of the
form 4k + 3 can divide a number of the form c2 + 1 where c is an integer.
 27. CONGRUENCES 131

Proof. 1) By Corollary 27.7 if p ≡ 1 (mod 4) then Np (x4 − 1) = 4. But

4 = Np (x4 − 1) ≤ Np ((x2 − 1)(x2 + 1)) ≤ Np (x2 − 1) + Np (x2 + 1) ≤ Np (x2 + 1) + 2
hence Np (x2 + 1) ≥ 2 and we are done.
2) Assume p ≡ 3 (mod 4) so p = 4k + 3 and assume Np (x2 = −1) > 0 so
there exists c ∈ Z such that c2 ≡ −1 (mod p); we want to derive a contradiction.
We have (by Fermat’s Little Theorem) that cp ≡ c (mod p). Since p 6 |c we get
cp−1 ≡ 1 (mod p). But
cp−1 ≡ c4k+2 ≡ (c2 )2k+1 ≡ (−1)2k+1 ≡ −1 (mod p),
a contradiction. 
Exercise 27.9. Prove that:
1) If p ≡ 1 (mod 3) then Np (x2 + x + 1) = 2. Equivalently any prime p of the
form 3k + 1 divides some number of the form c2 + c + 1.
2) If p ≡ 2 (mod 3) then Np (x2 + x + 1) = 0. Equivalently no prime p of the
form 3k + 2 can divide a number of the form c2 + c + 1.
Definition 27.10. Let a be an integer not divisible by a prime p. The order
of a mod p is the smallest positive integer k such that ak ≡ 1 (mod p). We write
k = op (a). Clearly op (a) equals the order o(a) of the image a of a in Fp .
Definition 27.11. An integer g is a primitive root mod p if it is not divisible
by p and op (g) = p − 1, equivalently, if the image g of g in F×
p is a generator of the
group F×
p.

Exercise 27.12. Prove that g is a primitive root mod p if and only if it is not
divisible by p and
g (p−1)/q 6≡ 1 (mod p)
for all primes q|p − 1.
Exercise 27.13. Prove that 3 is a primitive root mod 7 but 2 is not a primitive
root mod 7.
The following Theorem about the existence of primitive roots was proved by
Gauss:
Theorem 27.14. (Gauss) If p is a prime there exists a primitive root mod p.
Equivalently the group F×
p is cyclic.

Proof. By the Fundamental Theorem of Arithmetic, p − 1 = pe11 ...pess with

p1 , ..., ps distinct primes and e1 , ..., es ≥ 1. Let i ∈ {1, ..., s}. By Corollary 27.7
ei ei −1 ei
Np (xpi − 1) = pei i and Np (xpi − 1) = piei −1 . So xpi − 1 has a root ci mod p
ei −1
which is not a root mod p of xpi − 1. So
ei
p
ci i ≡ 1 (mod p),
ei −1
p
ci i 6≡ 1 (mod p).
It follows that the order of ci is a divisor of pei i but not a divisor of pei i −1 . Hence
op (ci ) = pei i .
By Proposition 22.10
op (c1 ...cs ) = pe11 ...pess = p − 1
so c1 ...cs is a primitive root mod p. 
 Part 6

Geometry
 CHAPTER 28

Lines

We start exploring topics in geometry. Geometry is the study of shapes such

as lines and planes, or, more generally, curves and surfaces, etc. There are two
paths towards this study: the synthetic one and the analytic (or algebraic) one.
Synthetic geometry is geometry without algebra. Analytic geometry is geometry
through algebra. Synthetic geometry originates with the Greek mathematics of
antiquity (e.g., the treatise of Euclid). Analytic geometry was invented by Fermat
and Descartes. We already encountered the synthetic approach in the discussion of
the affine plane and the projective plane which were purely combinatorial objects.
Here we introduce some of the most elementary structures of analytic geometry.
We start with lines. Later we will look at more complicated curves.
Definition 28.1. Let R be a field. The affine plane A2 = A2 (R) over R is the
set R2 = R × R. A point P = (x, y) in the plane is an element of R × R. A subset
L ⊂ R × R is called a line if there exist a, b, c ∈ R such that (a, b) 6= (0, 0) and
L = {(x, y) ∈ R2 ; ax + by + c = 0}.
We say a point P lies on the line L (or we say L passes through P ) if P ∈ L. Two
lines are said to be parallel if they either coincide or their intersection is empty (in
the last case we say they don’t meet). Three points are collinear if they lie on the
same line.
Definition 28.2. We sometimes write L = L(R) if we want to stress that
coordinates are in R.
Exercise 28.3. Prove that:
1) There exist 3 points which are not collinear.
2) For any two distinct points P1 and P2 there exists a unique line L (called
sometimes P1 P2 ) passing through P1 and P2 . Hint: If P1 = (x1 , y1 ), P2 = (x2 , y2 ),
and if
m = (y2 − y1 )(x2 − x1 )−1
then the unique line through P1 and P2 is:
L = {(x, y) ∈ R × R; y − y1 = m(x − x1 )}.
In particular any two non-parallel distinct lines meet in exactly one point.
3) Given a line L and a point P there exists exactly one line L0 passing through
P and parallel to L. (This is called Euclid’s fifth postulate but in our exposition
here this is not a postulate.)
Hence A2 = R2 = R × R together with the set L of all lines (in the sense above)
is an affine plane in the sense of Definition 10.41.
Remark 28.4. Not all affine planes in the sense of Definition 10.41 are affine
planes over a field in the sense above. Hilbert proved that an affine plane is the
135
136 28. LINES

affine plane over some field if and only if the theorems of Desargues (Parts I and
II) and Pappus (stated below) hold. See below for the “only if direction.”
Exercise 28.5. Prove that any line in Fp × Fp has exactly p points.
Exercise 28.6. How many lines are there in the plane Fp × Fp ?
Exercise 28.7. (Desargues’ Theorem, Part I) Let A1 , A2 , A3 , A01 , A02 , A03 be
distinct points in the plane. Also for all i 6= j assume Ai Aj and A0i A0j are not
parallel and let Pij be their intersection. Assume the 3 lines A1 A01 , A2 A02 , A3 A03
have a point in common. Then prove that the points L12 , L13 , L23 are collinear
(i.e., lie on some line). Hint: Consider the “space” R × R × R and define planes and
lines in this space. Prove that if two planes meet and don’t coincide then they meet
in a line. Then prove that through any two points in space there is a unique line
and through any 3 non-collinear points there is a unique plane. Now consider the
projection R × R × R → R × R, (x, y, z) 7→ (x, y) and show that lines project onto
lines. Next show that configuration of points Ai , A0i ∈ R × R can be realized as the
projection of a similar configuration of points Bi , Bi0 ∈ R × R × R not contained in a
plane. (Identifying R × R with the set of points in space with zero third coordinate
we take Bi = Ai , Bi0 = A0i for i = 1, 2, we let B3 have a nonzero third coordinate,
and then we choose B30 such that the lines B1 B10 , B2 B20 , B3 B30 have a point in
common.) Then prove “Desargues’ Theorem in Space” (by noting that if Qij is the
intersection of Bi Bj with Bi0 Bj0 then Qij is in the plane containing B1 , B2 , B3 and
also in the plane containing B10 , B20 , B30 ; hence Qij is in the intersection of these
planes which is a line). Finally deduce the original plane Desargues by projection.
Exercise 28.8. (Desargues’ Theorem, Part II) Let A1 , A2 , A3 , A01 , A02 , A03 be
distinct points in the plane. Assume the 3 lines A1 A01 , A2 A02 , A3 A03 have a point in
common or they are parallel. Assume A1 A2 is parallel to A01 A02 and A1 A3 is parallel
to A01 A03 . Prove that A2 A3 is parallel to A02 A03 . Hint: Compute coordinates. There
is an alternative proof that reduces Part II to Part I by using the “projective plane
over our field.”
Exercise 28.9. (Pappus’ Theorem) Let P1 , P2 , P3 be points on a line L and
let Q1 , Q2 , Q3 be points on a line M 6= L. Assume the lines P2 Q3 and P3 Q2 are
not parallel and let A1 be their intersection; define A2 , A3 similarly. Then prove
that A1 , A2 , A3 are collinear. Hint (for the case L and M meet): One can assume
L = {(x, 0); x ∈ R}, M = {(0, y); y ∈ R} (explain why). Let the points Pi = (xi , 0)
and Qi = (0, yi ) and compute the coordinates of Ai . Then check that the line
through A1 and A2 passes through A3 .
Remark 28.10. One can identify the projective plane (A2 , L) attached to the
affine plane (A2 , L) with the pair (P2 , P̌2 ) defined as follows. Let P2 = R3 / ∼ where
(x, y, z) ∼ (x0 , y 0 , z 0 ) if and only if there exists 0 6= λ ∈ R such that (x0 , y 0 , z 0 ) =
(λx, λy, λz). Denote the equivalence class of (x, y, z) by (x : y : z). Identify a point
(x, y) in the affine plane A2 = R2 = R × R with the point (x : y : 1) ∈ P2 . Identify
a point (x0 : y0 : 0) in the complement P2 \A2 with the class of lines in A2 parallel
to the line y0 x − x0 y = 0. This allows one to identify the complement P2 \A2 with
the line at infinity L∞ of A2 . Hence we get an identification of P2 with A2 . Finally
define a line in P2 as a set of the form
L = {(x : y : z); ax + by + cz = 0}.
 28. LINES 137

So under the above identifications,

L = {(x : y : 1); ax + by + c = 0} ∪ {(x : y : 0); ax + by = 0} = L ∪ {L}
b
where L is the line in R2 defined by ax + by + c = 0. Then define P̌2 to be the set
of all lines L in P2 . We get an identification of P̌2 with L.
Some familiar concepts such as area and distance can be defined in the above
context. Assume in what follows that R is a field such that 2 6= 0 and identify R2
with R2×1 .
Definition 28.11. Let P1 , P2 , P3 ∈ R2×1 be 3 points in the plane. Define
1
area(P1 , P2 , P3 ) = det(P2 − P1 , P3 − P1 ) ∈ R.
2
Exercise 28.12. Prove that
1) area(Pσ(1) Pσ(2) Pσ(3) ) = (σ) · area(P1 , P2 , P3 ) for all permutations σ ∈ S3 .
2) area(P1 , P2 , P3 ) = 0 if and only if P1 , P2 , P3 are collinear.
3) Let F : R2 → F 2 be an isomorphism of vector spaces and A its matrix with
respect to the canonical basis. Then A ∈ SL2 (R) if and only if “F preserves areas”
in the sense that for all P1 , P2 , P3 ∈ R2 we have
area(F (P1 ), F (P2 ), F (P3 )) = area(P1 , P2 , P3 ).
4) For any P0 ∈ R2×1 area is “invariant under translation by P0 ” in the sense
that
area(P1 + P0 , P2 + P0 , P3 + P0 ) = area(P1 , P2 , P3 ).
Definition 28.13. Let P1 , P2 ∈ R2×1 be 2 points in the plane. Define the
distance squared between these points as
dist2 (P1 , P2 ) = (P2 − P1 )t (P2 − P1 ) ∈ R.
Exercise 28.14. Prove that
1) dist2 (P1 , P2 ) = dist2 (P2 , P1 ).
2) If R = R then dist2 (P1 , P2 ) = 0 if and only if P1 = P2 . (Show that this may
fail for other fields.)
3) Let F : R2 → F 2 be an isomorphism of vector spaces and A its matrix with
respect to the canonical basis. Then A ∈ SO2 (R) if and only if “F preserves areas”
and also “preserves distances” in the sense that for all P1 , P2 ∈ R2×1 we have
dist2 (F (P1 ), F (P2 )) = dist2 (P1 , P2 ).
4) For any P0 ∈ R2×1 , dist2 is “invariant under translation by P0 ” in the sense
that
dist2 (P1 + P0 , P2 + P0 ) = dist2 (P1 , P2 ).
 CHAPTER 29

Conics

So far we were concerned with lines in the plane. Let us discuss now “higher
degree curves.” We start with conics. Assume R is a field with 2 = 1 + 1 6= 0;
equivalently R does not contain the field F2 .
Definition 29.1. The circle of center (a, b) ∈ R × R and radius r is the set
C(R) = {(x, y) ∈ R × R; (x − a)2 + (y − b)2 = r2 }.
Definition 29.2. A line is tangent to a circle if it meets it in exactly one point.
(We say that the line is tangent to the circle at that point.) Two circles are tangent
if they meet in exactly one point.
Exercise 29.3. Prove that for any circle and any point on it there is exactly
one line tangent to the circle at that point.
Exercise 29.4. Prove that:
1) A circle and a line meet in at most 2 points.
2) Two circles meet in at most 2 points.
Exercise 29.5. How many points does a circle of radius 1 have if R = F13 ?
Same problem for F11 .
Exercise 29.6. Prove that the circle C(R) with center (0, 0) and radius 1 is
an Abelian group with e = (1, 0), (x, y)0 = (x, −y), and group operation
(x1 , y1 ) ? (x2 , y2 ) = (x1 x2 − y1 y2 , x1 y2 + x2 y1 ).
Prove that the map
 
a b
C(R) → SO2 (R), (a, b) 7→
−b a
is a bijective group homomorphism. (Cf. Exercise 21.14 for SO2 (R).)
Exercise 29.7. Consider the circle C(F17 ). Show that (3̄, 3̄), (1̄, 0̄) ∈ C(F17 )
and compute (3̄, 3̄) ? (1̄, 0̄) and 2(1̄, 0̄) (where the latter is, of course, (1̄, 0̄) ? (1̄, 0̄)).
Circles are special cases of conics:
Definition 29.8. A conic is a subset Q ⊂ R × R of the form
Q = Q(R) = {(x, y) ∈ R × R; ax2 + bxy + cy 2 + dx + ey + f = 0}
for some (a, b, c, d, e, f ) ∈ R × ... × R, where (a, b, c) 6= (0, 0, 0).
We refer to (a, b, c, d, e) as the equation of the conic and if the corresponding
conic passes through a point we say that the equation of the conic passes through
the point. We sometimes say “conic” instead of “equation of the conic.”
139
140 29. CONICS

Exercise 29.9. Prove that if 5 points are given in the plane such that no 4
of them are collinear then there exists a unique conic passing through these given
5 points. Hint: Consider the vector space of all (equations of) conics that pass
through a given set S of points. Next note that if one adds a point to S the
dimension of this space of conics either stays the same or drops by one. Since the
space of all conics has dimension 6 it is enough to show that for r ≤ 5 the conics
passing through r points are fewer than those passing through r − 1 of the r points.
For r = 4, for instance, this is done by taking a conic that is a union of 2 lines.
 CHAPTER 30

Cubics

Definition 30.1. Let R be a field in which 2 = 1 + 1 6= 0, 3 = 1 + 1 + 1 6= 0.

Equivalently R does not contain F2 or F3 . A subset Z = Z(R) ⊂ R × R is called
an affine elliptic curve if there exist a, b ∈ R with 4a3 + 27b2 6= 0 such that
Z(R) = {(x, y) ∈ R × R; y 2 = x3 + ax + b}.
We call Z(R) the elliptic curve over R defined by the equation y 2 = x3 + ax + b.
Next we define the projective elliptic curve defined by the equation y 2 = x3 + ax + b
as the set
E(R) = Z(R) ∪ {∞}
where ∞ is an element not belonging to Z(R). (We usually drop the word “pro-
jective” and we call ∞ the point at infinity on E(R).) If (x, y) ∈ E(R) de-
fine (x, y)0 = (x, −y). Also define ∞0 = ∞. Next we define a binary opera-
tion ? on E(R) called the chord-tangent operation; we will see that E(R) be-
comes a group with respect to this operation. First define (x, y) ? (x, −y) = ∞,
∞?(x, y) = (x, y)?∞ = (x, y), and ∞?∞ = ∞. Also define (x, 0)?(x, 0) = ∞. Next
assume (x1 , y1 ), (x2 , y2 ) ∈ E(R) with (x2 , y2 ) 6= (x1 , −y1 ) . If (x1 , y1 ) 6= (x2 , y2 )
we let L12 be the unique line passing through (x1 , y1 ) and (x2 , y2 ). Recall that
explicitly
L12 = {(x, y) ∈ R × R; y − y1 = m(x − x1 )}
where
m = (y2 − y1 )(x2 − x1 )−1 .
If (x1 , y1 ) = (x2 , y2 ) we let L12 be the “line tangent to Z(R) at (x1 , y1 )” which is
by definition given by the same equation as before except now m is defined to be
m = (3x21 + a)(2y1 )−1 .
(This definition is inspired by the definition of slope in analytic geometry.) Finally
one defines
(x1 , y1 ) ? (x2 , y2 ) = (x3 , −y3 )
where (x3 , y3 ) is the “third point of intersection of E(R) with L12 ”; more precisely
(x3 , y3 ) is defined by solving the system consisting of the equations defining E(R)
and L12 as follows: replacing y in y 2 = x3 + ax + b by y1 + m(x − x1 ) we get a
cubic equation in x:
(y1 + m(x − x1 ))2 = x3 + ax + b
which can be rewritten as
x3 − m2 x2 + ... = 0.
x1 , x2 are known to be roots of this equation. We define x3 to be the third root
which is then
x3 = m2 − x1 − x2 ;
141
142 30. CUBICS

so we define
y3 = y1 + m(x3 − x1 ).
Summarizing, the definition of (x3 , y3 ) is
(x3 , y3 ) = (y2 − y1 )2 (x2 − x1 )−2 − x1 − x2 , y1 + (y2 − y1 )(x2 − x1 )−1 (x3 − x1 )

if (x1 , y1 ) 6= (x2 , y2 ), (x1 , y1 ) 6= (x2 , −y2 ) and

(x3 , y3 ) = (3x21 + a)2 (2y1 )−2 − x1 − x2 , y1 + (3x21 + a)(2y1 )−1 (x3 − x1 )

if (x1 , y1 ) = (x2 , y2 ), y1 6= 0.
Then E(R) with the above definitions is an Abelian group.
Exercise 30.2. Check the last statement. (N.B. Checking associativity is a
very laborious exercise.)
Exercise 30.3. Consider the group E(F13 ) defined by the equation y 2 = x3 + 8̄.
Show that (1̄, 3̄), (2̄, 4̄) ∈ E(F13 ) and compute (1̄, 3̄) ? (2̄, 4̄) and 2(2̄, 4̄) (where the
latter is, of course, (2̄, 4̄) ? (2̄, 4̄)).
Affine elliptic curves are special examples of cubics:
Definition 30.4. A cubic is a subset X = X(R) ⊂ R × R of the form
X(R) = {(x, y) ∈ R ×R; ax3 +bx2 y +cxy 2 +dy 3 +ex2 +f xy +gy 2 +hx+iy +j = 0}
where (a, b, c, ..., j) ∈ R × ... × R, (a, b, c, d) 6= (0, ..., 0).
As usual we refer to the tuple (a, b, c, ..., j) as the equation of a cubic (or, by
abuse, simply a cubic).
Exercise 30.5. (Three Cubics Theorem) Prove that if two cubics meet in
exactly 9 points and if a third cubic passes through 8 of the 9 points then the
third cubic must pass through the 9th point. Hint: First show that if r ≤ 8 and
r points are given then the set of cubics passing through them is strictly larger
than the set of cubics passing through r − 1 of the r points. (To show this show
first that no 4 of the 9 points are on a line. Then in order to find, for instance,
a cubic passing through P1 , ..., P7 but not through P8 one considers the cubics
Ci = Q1234i + Ljk , {i, j, k} = {5, 6, 7}, where Q1234i is the unique conic passing
through P1 , P2 , P3 , P4 , Pi and Ljk is the unique line through Pj and Pk . Assume
C5 , C6 , C7 all pass through P8 and derive a contradiction as follows. Note that P8
cannot lie on 2 of the 3 lines Ljk because this would force us to have 4 collinear
points. So we may assume P8 does not lie on either of the lines L57 , L67 . Hence P8
lies on both Q12345 and Q12346 . So these conics have 5 points in common. From
here one immediately gets a contradiction.) Once this is proved let P1 , ..., P9 be
the points of intersection of the cubics with equations F and G. We know that
the space of cubics passing through P1 , ..., P8 has dimension 2 and contains F and
G. So any cubic in this space is a linear combination of F and G, hence will pass
through P9 .
Exercise 30.6. (Pascal’s Theorem) Let P1 , P2 , P3 , Q1 , Q2 , Q3 be points on a
conic C. Let A1 be the intersection of P2 Q3 with P3 Q2 , and define A2 , A3 similarly.
(Assume the lines in question are not parallel.) Then prove that A1 , A2 , A3 are
collinear. Hint: The cubics
Q1 P2 ∪ Q2 P3 ∪ Q3 P1 and P1 Q2 ∪ P2 Q3 ∪ P3 Q1
 30. CUBICS 143

pass through all of the following 9 points:

P1 , P2 , P3 , Q1 , Q2 , Q3 , A1 , A2 , A3 .
On the other hand the cubic C ∪ A2 A3 passes through all these points except
possibly A1 . Then by the Three Cubics Theorem C ∪ A2 A3 passes through A1 .
Hence A2 A3 passes through A1 .
Exercise 30.7. Show how Pascal’s Theorem implies Pappus’ Theorem.
Remark 30.8. An extended version of the Three Cubics Theorem implies the
associativity of the chord-tangent operation on a cubic. (The extended version, to
be used below, follows from the usual version by passing to the “projective plane”;
we will not explain this proof here.) The rough idea is as follows. Let E be the
elliptic curve and Q, P, R points on it. Let
P Q ∪ E = {P, Q, U }
∞U ∪ E = {∞, U, V }
V R ∪ E = {V, R, W }
P R ∩ E = {P, R, X}
∞X ∩ E = {∞, X, Y }.
Here ∞A is the vertical passing through a point A. Note that
Q ? P = V, V ? R = W 0 , P ? R = Y.
We want to show that
(Q ? P ) ? R = Q ? (P ? R).
This is equivalent to
V ?R=Q?Y
i.e., that
W0 = Q ? Y
i.e., that Q, Y, W are collinear. Now the two cubics
E and P Q ∪ W R ∪ Y X
both pass through the 9 points
P, Q, R, U, V, W, X, Y, ∞.
On the other hand the cubic
Γ = U V ∪ P R ∪ QY
passes through all 9 points except W . By a generalization of the Three Cubics
Theorem (covering the case when one of the points is ∞) we get that Γ passes
through W hence QY passes through W . The above argument only applies to
chords and not to tangents. When dealing with tangents one needs to repeat the
argument and look at multiplicities. So making the argument rigorous becomes
technical.
 Part 7

Analysis
 CHAPTER 31

Limits

We discuss now some simple topics in analysis. Analysis is the study of “passing
to the limit.” The key words are sequences, convergence, limits, and later differential
and integral calculus. Here we will discuss limits. Analysis emerged through work
of Abel, Cauchy, Riemann, and Weierstrass, as a clarification of the early calculus
of Newton, Leibniz, Euler, and Lagrange.
Definition 31.1. A sequence in R is a map F : N → R; if F (n) = an we denote
the sequence by a1 , a2 , a3 , ... or by (an ). We let F (N) be denoted by {an ; n ≥ 1};
the latter is a subset of R.
Definition 31.2. A subsequence of a sequence F : N → R is a sequence of the
form F ◦ G where G : N → N is an increasing map. If a1 , a2 , a3 , ... is F then F ◦ G
is ak1 , ak2 , ak3 , ... (or (akn )) where G(n) = kn .
Definition 31.3. A sequence (an ) is convergent to a0 ∈ R if for any real
number  > 0 there exists an integer N such that for all n ≥ N we have |an −a0 | < .
We write an → a0 and we say a0 is the limit of (an ). A sequence is called convergent
if there exists a ∈ R such that the sequence converges to a. A sequence is called
divergent if it is not convergent.
Exercise 31.4. Prove that an = n1 converges to 0.
Hint: Let  > 0; we need to find N such that for all n ≥ N we have
1
| − 0| < ;
n
it is enough to take N to be any integer such that N > 1 .
Exercise 31.5. Prove that an = √1 converges to 0.
n

1
Exercise 31.6. Prove that an = n2 converges to 0.
Exercise 31.7. Prove that an = n is divergent.
Exercise 31.8. Prove that an = (−1)n is divergent.
Exercise 31.9. Prove that if an → a0 and bn → b0 then
1) an + bn → a0 + b0
2) an bn → a0 b0 .
If in addition b0 6= 0 then there exists N such that for all n ≥ N we have bn 6= 0;
moreover if bn 6= 0 for all n then
3) abnn → ab00 .
Hint for 1: Consider any  > 0. Since an → a0 there exists Na such that for
all n ≥ Na we have |an − a0 | < 2 . Since bn → b0 there exists Nb such that for all
147
148 31. LIMITS

n ≥ Nb we have |bn − b0 | < 2 . Let N = max{Na , Nb } be the maximum between

Na and Nb . Then for all n ≥ N we have
 
|(an + bn ) − (a0 + b0 )| ≤ |an − a0 | + |bn − b0 | < + = .
2 2
Exercise 31.10. Prove that if an → a, bn → b, and an ≤ bn for all n ≥ 1 then
a ≤ b.
Definition 31.11. A sequence F is bounded if the set F (N) ⊂ R is bounded.
Definition 31.12. A sequence F is increasing if F is increasing. A sequence
F is decreasing if −F is increasing.
Definition 31.13. A sequence (an ) is Cauchy if for any real  > 0 there exists
an integer N such that for all integers m, n ≥ N we have |an − am | < .
Exercise 31.14. Prove that any convergent sequence is Cauchy.
Exercise 31.15. Prove the following statements in the prescribed order:
1) Any Cauchy sequence is bounded.
2) Any bounded sequence contains a sequence which is either increasing or
decreasing.
3) Any bounded sequence which is either increasing or decreasing is convergent.
4) Any Cauchy sequence which contains a convergent subsequence is itself con-
vergent.
5) Any Cauchy sequence is convergent.
Hints: For 1 let  = 1, let N correspond to this , and get that |an − aN | < 1
for all n ≥ N ; conclude from here. For 2 consider the sets An = {am ; m ≥ n}. If at
least one of these sets has no maximal element we get an increasing subsequence by
Proposition 16.3. If each An has a maximal element bn then bn = akn for some kn
and the subsequence akn is decreasing. For 3 we view each an ∈ R as a Dedekind
cut i.e., as a subset an ⊂ Q; the limit will be either the union of the sets an or the
intersection. Statement 4 is easy. Statement 5 follows by combining the previous
statements.
Exercise 31.16. Prove that any subset in R which is bounded from below has
an infimum; and any subset in R which is bounded from above has a supremum.
Definition 31.17. A function F : R → R is continuous at a point a0 ∈ R if for
any sequence (an ) converging to a0 we have that the sequence (F (an )) converges
to F (a0 ).
Exercise 31.18. ( and δ criterion). Prove that a function F : R → R is
continuous at a0 if and only if for any real  > 0 there exists a real δ > 0 such that
for any a ∈ R with |a − a0 | < δ we have |F (a) − F (a0 )| < .
Exercise 31.19. Prove that a function F : R → R is continuous (for the
Euclidean topology on both the source and the target) if and only if it is continuous
at any point of R.
Exercise 31.20. Prove that any polynomial function f : R → R (i.e., any
function of the form a 7→ f (a) where f is a polynomial) is continuous.
Exercise 31.21. Prove that R with the Euclidean topology is connected.
Hint: Assume R = A ∪ B with A, B open, non-empty, and disjoint, and seek
a contradiction. Let a ∈ A and b ∈ B. Assume a ≤ b; the case b ≤ a is similar.
 31. LIMITS 149

Show that there exists sequences (an ) and (bn ), the first increasing, the second
decreasing, with an ≤ bn and bn − an → 0. (To check this use recursion to define
an+1 , bn+1 in terms of an , bn by the following rule: if cn = an +b
2
n
then set an+1 = cn
and bn+1 = bn in case cn ∈ A; and set an+1 = an and bn+1 = cn in case cn ∈ B.)
Note that an → a0 and bn → b0 and a0 = b0 . Since A, B are open and disjoint they
are closed. So a0 ∈ A and b0 ∈ B. But this contradicts the fact that A and B are
disjoint.
Definition 31.22. For a, b ∈ R the closed interval [a, b] ⊂ R is defined as
[a, b] = {x ∈ R; a ≤ x ≤ b}.
Exercise 31.23. Prove that [a, b] are closed in the Euclidean topology.
Exercise 31.24. Prove that the open intervals (a, b) and the closed intervals
[a, b] are connected in R.
Exercise 31.25. (Heine-Borel Theorem) Prove that any closed interval in R
is compact. Hint: Assume [a, b] is not compact and derive a contradiction as
follows. We know [a, b] has an open covering (Ui )i∈I that does not have a finite open
subcovering. Show that there exists sequences (an ) and (bn ), the first increasing,
the second decreasing, with an ≤ bn and bn − an → 0, such that [an , bn ] cannot
be covered by finitely many Ui s. (To check this use recursion to define an+1 , bn+1
in terms of an , bn by the following rule: let cn = an +b 2
n
; then at least one of the
two intervals [an , cn ] or [cn , bn ] cannot be covered by finitely many Ui s; if this is
the case with the first interval then set an+1 = an and bn+1 = cn ; in the other case
set an+1 = cn and bn+1 = bn .) Note that an → a0 and bn → b0 and a0 = b0 . But
a0 = b0 is in one of the Ui s; this Ui will completely contain one of the intervals
[an , bn ] which is a contradiction.
 CHAPTER 32

Series
Pn
Definition 32.1. Let (an ) be a sequence and sn = k=1 ak . The sequence
(s
Pn∞) is called the sequence of partial sums. If (sn ) is convergent to some s we say
k=1 an is a convergent series and that this series converges to s; we write
∞
X
an = s.
k=1
P∞
If the sequence (sn ) is divergent we say that k=1 an is a divergent series.
Exercise 32.2. Prove that
∞
X 1
= 1.
n=1
n(n + 1)
Hint: Start with the equality
1 1 1
= −
n(n + 1) n n+1
and compute
N
X 1 1
=1− .
n=1
n(n + 1) N
Exercise 32.3. Prove that the series
∞
X 1
n=1
n2
is convergent.
Hint: Prove the sequence of partial sums is bounded using the inequality
1 1
≤
n2 n(n + 1)
plus Exercise 32.2.
Exercise 32.4. Prove that the series
∞
X 1
n k
n=1
is convergent for k ≥ 3.
Exercise 32.5. Prove that the series
∞
X 1
n=1
n
is divergent. This series is called the harmonic series.
151
152 32. SERIES

Hint: Assume the series is convergent. Then the sequence of partial sums is
convergent hence Cauchy. Get a contradiction from the inequality:
1 1 1 1 1 1
+ + + ... + n > 2n × n = .
2n + 1 2n + 2 2n + 3 2 + 2n 2 + 2n 2
Exercise 32.6. Prove that an → 0 if |a| < 1.
Hint: We may assume 0 < a < 1. Note that (an ) is decreasing. Since it is
bounded it is convergent. Let α be its limit. Assume α 6= 0 and get a contradiction
by noting that
1 an α
= n+1 → = 1.
a a α
Exercise 32.7. Prove that
∞
X 1
an =
n=1
1−a
if |a| < 1.
Exercise 32.8. Prove that the series
∞
X an
n=0
n!
is convergent for all a ∈ R; its limit is denoted by ea = exp(a); e = e1 is called the
Euler number; the map
R → R, a 7→ exp(a)
is called the exponential map. Prove that
exp(a + b) = exp(a) exp(b).
Exercise 32.9. Prove that the function exp : R → R is continuous.
Hint: Let a ∈ R and  > 0. It is enough to show that there exists δ > 0 such
that if |b − a| < δ then | exp(b) − exp(a)| < . Show that there is a δ such that for
any b with |b − a| < δ there exists an n such that for all m ≥ n
m n
X ak X ak 
| − |<
k! k! 3
k=0 k=0
m n
X bk X bk 
| − |<
k! k! 3
k=0 k=0
n k n k
X b X a 
| − |< .
k! k! 3
k=0 k=0
From the first two inequalities we get
n
X ak 
| exp(a) − |≤
k! 3
k=0
n
X bk 
| exp(b) − |≤ .
k! 3
k=0
Then
  
| exp(b) − exp(a)| < + + = .
3 3 3
 32. SERIES 153

Exercise 32.10. Let S ⊂ {0, 1}N be the set of all sequences (an ) such that
there exist N with an = 1 for all n ≥ N . Prove that the map
∞
X an
{0, 1}N \S → R, (an ) 7→
n=1
2n
is (well defined and) injective. Conclude that R is not countable.
Exercise 32.11. Prove that there exist transcendental numbers in R. Hint:
R is uncountable whereas the set of algebraic numbers is countable; cf. Exercise
26.20. This is Cantor’s proof of existence of transcendental numbers.
Real analysis (analysis of sequences, continuity, and other concepts of calculus
like differentiation and integration of functions on R) can be extended to complex
analysis. Indeed we have:
Definition 32.12. A sequence (zn ) in C is convergent to z0 ∈ C if for any real
number  > 0 there exists an integer N such that for all n ≥ N we have |zn −z0 | < .
We write zn → z0 and we say z0 is the limit of (zn ). A sequence is called convergent
if there exists z ∈ C such that the sequence converges to z. A sequence is called
divergent if it is not convergent.
Exercise 32.13. Let (zn ) be a sequence in C and let
zn = an + bn i,
an , bn ∈ R. Let z0 = a0 + b0 i. Prove that zn → z0 if and only if an → a0 and
bn → b0 .
Definition 32.14. A sequence (zn ) in C is Cauchy if for any real  > 0 there
exists an integer N such that for all integers m, n ≥ N we have |zn − zm | < .
Exercise 32.15. Prove that a sequence in C is convergent if and only if it is
Cauchy.
Exercise 32.16. Prove that:
1) The series
∞
X zn
n=0
n!
is convergent for all z ∈ C; its limit is denoted by ez = exp(z).
2) exp(z + w) = exp(z) exp(w) for all z, w ∈ C.
3) exp(z) = exp(z) for all z ∈ C.
4) | exp(it)| = 1 for all t ∈ R.
5) The map
C → C, z 7→ exp(z),
is continuous. This map is called the (complex) exponential map.
There is a version of the above theory in p-adic analysis (which is crucial to
number theory). Recall the ring of p-adic numbers Zp whose elements are denoted
by [an ].
Definition 32.17. Say that pe divides α = [an ] if there exists β = [bn ] such
that [an ] = [pe bn ]; write pe |α. For any 0 6= α = [an ] ∈ Zp let v = v(α) be the
unique integer such that pn |an for n ≤ v and pv+1 6 |av+1 . Then define the norm of
α by the formula |α| = p−v(α) . We also set |0| = 0.
154 32. SERIES

Exercise 32.18. Prove that if α = [an ] and β = [bn ] then

|α + β| ≤ max{|α|, |β|}.
Definition 32.19. Consider a sequence [an1 ], [an2 ], [an3 ], ... of elements in Zp
which for simplicity we denote by α1 , α2 , α3 , ....
1) α1 , α2 , α3 , ... is called a Cauchy sequence if for any real (or, equivalently,
rational)  > 0 there exists an integer N such that for all m, m0 ≥ N we have
|αm − αm0 | ≤ .
2) We say that α1 , α2 , α3 , ... converges to some α0 ∈ Zp if for any real (or,
equivalently, rational)  > 0 there exists an integer N such that for all m ≥ N we
have |αm − α0 | ≤ . We say α0 is the limit of (αn ) and we write αn → α0 .
Exercise 32.20. Prove that a sequence in Zp is convergent if and only if it is
Cauchy.
The following is in deep contrast with the case of R:

Pn Prove that if (αn ) is a sequence in Zp with αn → 0 then the

Exercise 32.21.
sequence
P∞ sn = k=1 αk is convergent in Zp ; the limit of the latter is denoted by
n=1 αn .
Exercise 32.22.P∞
1) Prove that n=1 pn−1 is the inverse of 1 − p in Zp .
2) Prove that if α ∈ Zp has |α| = 1 then α is invertible in Zp . Hint: Use the
fact that if p does not divide an integer a ∈ Z then there exist integers m, n ∈ A
such that ma + np = 1; then use 1) above.
3) Prove that for all n ≥ 1 and all a ∈ Zp with |a| < 1 there exists an element
n n
of Zp denoted by an! such that (n!) · an! = an . Hint: Use 2) above.
P∞ an
4) Prove that n=1 n! is convergent in Zp for all a ∈ Zp with |a| < 1. One
denotes the limit by expp (a).
 CHAPTER 33

Trigonometry

Trigonometry arose long before calculus mainly motivated by geometry and

astronomy. A rigorous approach to trigonometry requires some elements of analysis
that we already covered and hence can be used in what follows. We will define the
functions sin and cos and also the number π.
Definition 33.1. For all t ∈ R define cos t, sin t ∈ R as being the unique real
numbers such that
exp(it) = cos t + i sin t.
(This is called Euler’s formula but here this is a definition and not a theorem.)
Exercise 33.2. Prove the following equalities:
1) cos(t1 + t2 ) = cos t1 cos t2 − sin t1 sin t2 ;
2) sin(t1 + t2 ) = sin t1 cos t2 + cos t1 sin t2 .
Exercise 33.3. Prove that the map f : R → SO2 (R) defined by
 
cot t sin t
f (t) =
− sin t cos t
is a group homomorphism.
Exercise 33.4. Prove that if H is a closed subgroup of R and H 6= R, H 6= {0}
then there exists a unique T ∈ R, T > 0, such that
H = {nT ; n ∈ Z}.
Hint: One first shows that if T is the infimum of the set
{a ∈ H; a > 0}
then T 6= 0. In order to check this assume T = 0 and seek a contradiction. Indeed
from T = 0 we get that there exists a sequence (an ) with an ∈ H, and an → 0.
Deduce from this and the fact that H is closed that G = R, a contradiction. Finally
one shows that H = {nT ; n ∈ Z} using an argument similar to the one used to prove
Proposition 20.6.
Exercise 33.5. Prove that the map
F : R → C× , F (t) = exp(it)
is non-constant and non-injective. Conclude that there exists a unique real number
π ∈ R, π > 0, such that
Ker F = {2nπ; n ∈ Z}.
(This is our definition of the number π. In particular, by this very definition one
gets exp(πi) + 1 = 0 which is a celebrated formula of Euler; for us this is a trivial
consequence of our definition of π.)

155
 CHAPTER 34

Calculus

Calculus was invented by Newton and Leibniz, motivated by problems in me-

chanics and analytic geometry. The main concept of calculus is that of derivative
of a function which we briefly review here.
Definition 34.1. Let F : R → R be a map and a0 ∈ R. We say F is differ-
entiable at a0 if there exists a real number (denoted by) F 0 (a0 ) ∈ R such that for
any sequence an → a0 with an 6= a0 we have
F (an ) − F (a0 )
→ F 0 (a0 ).
an − a0
Exercise 34.2. Prove that if F is differentiable at a0 ∈ R then it is continuous
at a0 .
Exercise 34.3. Prove that if F is a constant function (i.e., F (x) = F (y) for
all x, y ∈ R) then F is differentiable at any a and F 0 (a) = 0.
Definition 34.4. We say F : R → R is differentiable if F is differentiable at
any a ∈ R. If this is the case the map a 7→ F 0 (a) is called the derivative of F and is
denoted by F 0 = dF 0
dx : R → R. If F is differentiable we say F is twice differentiable
00
and F is called the second derivative of F . One similarly defines what it means
for F to be n times differentiable (n ∈ N). We say F is infinitely differentiable (or
smooth) if it is n times differentiable for any n ∈ N. One denotes by C ∞ (R) the
set of smooth functions. We denote by D : C ∞ (R) → C ∞ (R) the map D(F ) = F 0 .
Exercise 34.5. Prove that for any F, G ∈ C ∞ (R) we have F +G, F ·G ∈ C ∞ (R)
and
1) D(F + G) = D(F ) + D(G) (additivity);
2) D(F · G) = F · D(G) + G · D(F ) (Leibniz rule);
here F + G, F · G are the pointwise addition and multiplication of F and G. In
particular C ∞ (R) is a ring with respect to + and ·; 0 and 1 are the functions
0(x) = 0 and 1(x) = 1.
Exercise 34.6. Prove that any polynomial function F : R → R is smooth and
Xn n
X
F (x) = an xn ⇒ F 0 (x) = nan xn−1 .
k=0 k=0

Hint: It is enough to look at F (x) = xk . In this case

akn − ak0 k−1
= ak−1
n + ak−2
n a0 + ... + a0 → kak−1
0 .
an − a0
Exercise 34.7. Prove that F (x) = exp(x) is differentiable and F 0 (x) = exp(x).
Hence F is smooth.
157
158 34. CALCULUS

Exercise 34.8. Prove that F (x) = sin(x) is differentiable and F 0 (x) = cos(x).
Prove that G(x) = cos(x) is differentiable and G0 (x) = − sin(x). Hence F and G
are smooth.
Exercise 34.9. (Chain rule) Prove that if F, G ∈ C ∞ (R) then F ◦ G ∈ C ∞ (R)
and
D(F ◦ G) = (D(F ) ◦ G) · D(G).
(Here, as usual, ◦ denotes composition.)
More generally one can define derivatives of functions of several variables as
follows:
Definition 34.10. Let F : Rn → R be a function. Let a = (a1 , ..., an ) ∈ Rn
and define Fi : R → R by
Fi (x) = F (a1 , ..., ai−1 , x, ai+1 , ..., an )
(with the obvious adjustment if i = 1 or i = n). We say that F is differentiable
with respect to xi at a if Fi is differentiable at ai ; in this case we define
∂F
(a) = Fi0 (ai ).
∂xi
We say that F is differentiable with respect to xi if it is differentiable with respect to
∂F
xi at any a ∈ Rn . For such a function we have a well defined function ∂x i
: Rn → R
which is also denoted by Di F . We say that F is infinitely differentiable (or smooth)
if F is differentiable, each Di F is differentiable, each Di Dj F is differentiable, each
Di Dj Dk F is differentiable, etc. We denote by C ∞ (Rn ) the set of smooth functions;
it is a ring with respect to pointwise addition and multiplication.
Definition 34.11. Let P ∈ C ∞ (Rr+2 ). An equation of the form
d2 F dr F
 
dF
P x, F (x), (x), 2 (x), ..., r (x) = 0
dx dx dx
is called a differential equation. Here F ∈ C ∞ (R) is an unknown function and one
i
defines ddxFi = Di+1 (F ) = D(Di (F )).
The study of differential equations has numerous applications within mathe-
matics (e.g., geometry) as well as natural sciences (e.g., physics).
Example 34.12. Here is a random example of a differential equation:
 2 5 !   3 
d F 3 dF d F
exp −x − x5 F 6 = 0.
dx2 dx dx3

The additivity and the Leibniz rule have an algebraic flavor. This suggests the
following:
Definition 34.13. Let R be a commutative unital ring. A map D : R → R is
called a derivation if
1) D(a + b) = D(a) + D(b) (additivity);
2) D(a · b) = a · D(b) + b · D(a) (Leibniz rule).
Example 34.14. D : C ∞ (R) → C ∞ (R) is a derivation.
 34. CALCULUS 159

Exercise 34.15. Prove that any derivation D : Z → Z is identically 0 i.e.,

D(x) = 0 for all x ∈ Z.
Hint: By additivity D(0) = 0 and D(−n) = −D(n). So it is enough to show
D(n) = 0 for n ∈ N. Proceed by induction on n. For the case n = 1, by the Leibniz
rule,
D(1) = D(1 · 1) = 1 · D(1) + 1 · D(1) = 2 · D(1)
hence D(1) = 0. The induction step follows by additivity.
Remark 34.16. Exercise 34.15 shows that there is no naive analogue of cal-
culus in which rings of functions such as C ∞ (R) are replaced by rings of num-
bers such as Z. An analogue of calculus for Z is, however, considered desirable
for the purposes of number theory. Such a theory has been developed. (Cf. A.
Buium, Arithmetic Differential Equations, Mathematical Surveys and Monographs
118, American Mathematical Society, 2005.) In that theory the analogue of x is a
fixed prime p and the analogue of the derivation D = dxd
: C ∞ (R) → C ∞ (R) is the
operator
d dx x − xp
: Z → Z, =
dp dp p
which is well defined by Fermat’s Little Theorem. For example,
d4 4 − 45
= = −204.
d5 5