The Theory of Single-Variable Calculus

Math 421, Section 001, Spring 2021

May 5, 2021

Abstract
These are collaboratively TeXed notes for our section of Math 421 at UW-Madison
this semester. Each student will volunteer to be the official notetaker for one week.
When it’s your turn, please feel free to consult with each-other and with Alex if you
need help getting started with LaTeX.

Contents
1 Introduction to sets, logic, and proofs 2
1.1 Introduction to set theory (Jan 26) . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Logical statements, implications, and quantifiers (Jan 28) . . . . . . . . . . . . 5
1.3 Proofs (Feb 2) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

2 Set operations and functions 16

2.1 Intersection and union (Feb 4) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2 Collections of sets (Feb 4) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3 More on set operations (Feb 9) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.4 Functions (Feb 11) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.5 Cartesian product and graphs of functions (Feb 16) . . . . . . . . . . . . . . . . 22

3 Infinity 25
3.1 Finite and infinite sets (Feb 16) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.2 Countable sets (Feb 18) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.3 Uncountable sets (Feb 23) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.4 Another look at Cantor’s proof (Feb 25) . . . . . . . . . . . . . . . . . . . . . . . 33

4 Number systems 35
4.1 Rules of Arithmetic (March 9) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.2 Equivalence relations (Mar 11) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.3 The rational numbers (Mar 11) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
 4.4 Properties of Q (Mar 16) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.5 The real numbers (Mar 18) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.6 Properties of R (Mar 23) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.7 Supremum and infimum (Mar 23) . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

5 Sequences and limits 57

5.1 The definition of a limit (Mar 25) . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.2 Limit rules (Mar 30) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.3 Monotone sequences (Mar 30) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
5.4 lim sup and lim inf (Apr 1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.5 Subsequences and the Bolzano-Weierstrass Theorem (Apr 8) . . . . . . . . . . . 67
5.6 Accumulation points (Apr 13) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

6 Functions and limits 73

6.1 Limits of functions (Apr 13) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
6.2 Continuity of functions using sequences (Apr 15) . . . . . . . . . . . . . . . . . . 76
6.3 Extreme values (Apr 20) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
6.4 A word or two about calculus (Apr 22) . . . . . . . . . . . . . . . . . . . . . . . . 81

1 Introduction to sets, logic, and proofs

1.1 Introduction to set theory (Jan 26)
A set is simply a collection of elements.
Example 1.1.1. Consider the set

S = {0, 1, π, tomato}.

We write
1∈S
to say that the element “1” belongs to S. On the other hand, we write

carrot ∈/ S

to say that the element “carrot” does not belong to S.

We say that two sets are equal if they contain the same elements. The order in which
elements are listed, and any repetitions, do not matter. For instance, the set

S = {1, 1, 0, π, tomato, π}

is equal to the set of Example 1.1.1 above.

2
Example 1.1.2. Consider the two sets

A = {Utah, Arizona, Colorado, New Mexico}

and
B = {states in the U.S.A. that share a corner with three other states}.
By consulting a map, one can observe that every element of the set A shares a corner with
three other states (indeed, with the other three elements of A). Hence, every element of A
belongs to B. By studying the rest of the map exhaustively, one checks that none of the
remaining 46 states has a 4-way corner. Therefore A = B.
But what is a state? And, more importantly, what is a tomato? (Vegetable? Fruit?
Berry?) These questions are outside the purview of this class.
In mathematics, we generally restrict ourselves to a very limited “universe” of sets. In
fact, we will start with only the following two sets: the natural numbers

N = {1, 2, 3, 4, 5, . . .}

and the integers

Z = {. . . , −3, −2, −1, 0, 1, 2, 3, . . .}.
Everything else we need (rational numbers, real numbers, limits, derivatives), we will con-
struct from these raw ingredients.1
Let’s continue discussing basic set terminology.
Definition 1.1.3. Let S and T be sets. We say that S is a subset of T if every element of
S also belongs to T. In this case, we write

S ⊂ T.

For example, we have

{0, 2, 4} ⊂ {0, 1, 2, 3, 4, 5}.
Also, notice that the natural numbers form a subset of the integers:

N ⊂ Z.

In fact, the former is precisely the set of “positive integers.” We can thus specify the subset
of natural numbers using the following notation:

N = {n ∈ Z ∣ n > 0}.
1
In fact, I encourage you not even to take the familiar properties of the natural numbers (a + b = b +
a, a(b + c) = ab + ac, etc) for granted. These are consequences of Peano’s axioms — see Wikipedia.

3
Example 1.1.4. Consider the subset of Z defined by

E = {n ∈ Z ∣ n = 2m for some m ∈ Z}.

In English, this reads:

“E is the set of all integers, n, such that n = 2m for some natural number, m.”

In fact, this is just the set of even integers:

E = {. . . , −8, −6, −4, −2, 0, 2, 4, 6, 8, . . .}.

For instance, why is 0 ∈ E? Because there does exist an integer (namely, m = 0) such that
2 ⋅ m = 0.
Example 1.1.5. The set of odd integers is given by

E ′ = {n ∈ Z ∣ n = 2m + 1 for some m ∈ Z}.

Definition 1.1.6. We say that an integer p divides an integer n if and only if there exists
an integer m such that
n = p ⋅ m.
For instance, given an integer n ∈ Z, we have n ∈ E if and only if 2 divides n. (In this case,
m = n/2 is an integer).
Example 1.1.7. Consider the subset of N defined by

Q = {n ∈ N ∣ m divides n ⇒ m = 1 or m = n}.

The symbol ⇒ means “implies;” in other words, the set Q is defined to be “the set of all
positive integers such that if m divides n, then m = 1 or m = n.” (See Section 1.2.2 below for
more on implications.)
What precisely is the set Q? Let’s translate the definition into more familiar terms. An
integer that divides n is also called a “factor.” So the statement is that “if” m is a factor of
n, then m must equal either 1 or n itself. In other words:

“the only factors of n are 1 and n itself.”

Hence
Q = {1, 2, 3, 5, 7, 11, 13, 17, 19, . . .}
is just the set of all prime numbers (including 1).
Next week, we will continue discussing “set operations,” i.e., things one can do with sets
(or collections of sets) to obtain new sets. But first, we need to discuss logical statements
and implications, and go over what it means to prove a statement mathematically.

4
1.2 Logical statements, implications, and quantifiers (Jan 28)
References for this section and the next are Hutchings’s notes and Chapter 3 of “Transition
to higher math,” both available on Canvas.

1.2.1 Statements

A statement is a sentence that is either true or false, but not both.

Examples 1.2.1.

1. P = “6 is an even integer”. True.

2. Q = “3 is an even integer”. False.

3. “The time has come.” (The time for what, exactly?) Not a statement, at least as far
as math is concerned.

Given one or more statements, there are several standard logical “operations” that we
can perform on them. These are:
Negation: ¬P = “6 is not an even integer” (False.)
The negation of a statement is true if and only if its negation is false. A double negation
gives us back the original statement:

¬(¬P ) = “6 is not not an even integer ” = P (true).

Conjunction: P ∧ Q = “6 is an even integer and 3 is an even integer” (false).

Disjunction: P ∨ Q = “6 is an even integer or 3 is an even integer” (true).
Here is a “truth table” summarizing the above:

P Q ¬P P ∨Q P ∧Q
T T F T T
F T T T F
T F F T F
F F T F F

It’s also important to know how to negate statements involving and/or.

Example 1.2.2. Consider the two statements:

P = “Liz is taking math”

Q = “Liz is taking organic chemistry.”

5
The conjunction of these would be:

P ∧ Q = “Liz is taking math and o. chem”

whereas the disjunction would be:

P ∨ Q = “Liz is taking either math or o. chem.”

Then the statement

¬(P ∧ Q) = “Liz is not taking math and o. chem”

is logically equivalent to:

“Liz is not taking math or not taking o. chem” = (¬P ) ∨ (¬Q).

On the other hand, the statement

¬(P ∨ Q) = “Liz is not taking either math or o. chem”

would be phrased in better English as:

“Liz is taking neither math nor o. chem”

which is just the same as saying:

“Liz is not taking math and not taking o. chem” = (¬P ) ∧ (¬Q).

We have just demonstrated what are know as “DeMorgan’s Laws:”

¬(P ∧ Q) = (¬P ) ∨ (¬Q)

¬(P ∨ Q) = (¬P ) ∧ (¬Q).

Note: Although it’s good to write them down once for clarity, it’s not at all necessary to
memorize these kinds of formal logical rules; you will become very familiar with them just
by doing your homework over the next few weeks.

1.2.2 Implications

Suppose that P and Q are two separate statements (as above). We say P implies Q, and
write
P ⇒Q
to mean
“if P, then Q”
or, in other words,
“P is true only if Q is true.”

6
Example 1.2.3. Let
P = “Harry is a Packers fan”
Q = “Harry is a human being.”
Does P imply Q? Certainly....if Harry is a Packers fan, then he must be a human being. If
Harry is a household cat, he might potentially be a mascot, but calling him a “fan” would
be difficult to justify. In other words, Harry can be a Packers fan only if he is also a human
being.
On the other hand, does Q imply P ? In other words, if Harry is a human being, does it
follow that he is a Packers fan? Blasphemous as it may sound, there do exist human beings
that do not actively root for the Packers. So for all we know, Harry may think and breathe,
but not be a Packers fan. Therefore Q does not imply P, in which case we write

Q⇒
/ P.

So here, P ⇒ Q is an example of a true implication whose converse implication, Q ⇒ P, is

false.
Lastly, let’s suppose that Harry is not a Packers fan, and is also not a human being
(for instance, Harry is a cat). Does the implication P ⇒ Q hold true in this case? Well, it
certainly isn’t false: since Harry is not a Packers fan, the fact that he is a cat is completely
irrelevant. Therefore P ⇒ Q is vacuously true in this case. Here’s another example:
Example 1.2.4. “If you’re the king of France, then I’ll be a monkey’s uncle.”
Notice that both claims—that you’re the king of France, and that I could ever have a
niece or nephew not belonging to homo sapiens—are patently false. (The last king of France
met his end in 1793, so we needn’t check the identity of the “you” in the statement.) But
is the whole sentence, i.e. the “if, then” implication, a true statement? Yes....in fact, it is
vacuously true, since the “P ” statement can never be true.
Now, back to Harry of Example 1.2.3 for a minute. Let’s consider the implication

¬Q ⇒ ¬P = “If Harry is not a human being, then Harry is not a Packers fan.”

This is known as the contrapositive of the implication P ⇒ Q.

Is the statement true? Quite certainly...in fact, we already used it above to justify
P ⇒ Q. A funny (and convenient) fact about implications is that an implication and
its contrapositive are always logically equivalent.2 In other words, the statement
“if Harry is a Packers fan, then Harry is a human being” is completely equivalent to the
statement “if Harry is not a human being, then Harry can hardly be a Packers fan.” (Both
implications happen to be true, in this case.)
2
i.e., they are either both true or both false.

7
Example 1.2.5. Let S and T be sets. The statement:

“If x ∈ S, then x ∈ T ”

is logically equivalent to the statement:

“If x ∈/ T, then x ∈/ S”.

Why? Because if the first one is true, then the second one is definitely true. On the other
hand, if the first one is false (i.e. there does exist an x ∈ S such that x ∈/ T ), then the second
one is also obviously false. Indeed, both implications are equivalent to the statement that S
is a subset of T (i.e. S ⊂ T )—see Definition 1.1.3 above.
It takes some getting-used-to; but once we’ve been doing proofs for a little while, the
freedom to change between an implication (P ⇒ Q) and its contrapositive (¬Q ⇒ ¬P ) will
become second nature.
This discussion can be summarized in the following truth table:

P Q P ⇒Q Q⇒P (¬Q) ⇒ (¬P )

T T T T T
F T T F T
T F F T F
F F T T T

1.2.3 Quantifiers

Things get more interesting when we combine logic with set theory (as we have already
begun to do in Example 1.2.5 above).
Let S be a set (for instance, S = Z is the set of integers), and suppose that P (x) is a
sentence about some element x ∈ S. For example, we might take the sentence

R(x) = “x > 0 ⇒ x2 + 1 > 0.”

This isn’t quite a statement (yet), because we don’t know which x we’re talking about. But
there are two common ways to make it into a statement:

∀x ∈ S, R(x) = “For all x ∈ S, if x > 0 then x2 + 1 > 0”

and
∃x ∈ S, R(x) = “There exists an x ∈ S, (such that) if x > 0 then x2 + 1 > 0.”
The two symbols ∀ (“for all”) and ∃ (“for some,” or, “there exists”) are known as quanti-
fiers. They determine the range of x’s that a statement needs to be true for: either for all
of them, or just for one.

8
 The two statements above are both true for S = Z (in fact, the first one implies the second
one). But here is another example:
Example 1.2.6. Consider the statement

P = “∀x ∈ Z, x2 + 2x + 1 > 0.”

Is this statement true or false? It’s a bit more difficult to tell...because of the “∀” symbol,
we are somehow asked to check all possible integers x. But in fact the statement is false,
because for x = −1, we have
(−1)2 + 2(−1) + 1 = 0 >/ 0.
Since P is false, we have established that the negation ¬P is true.
Notice that to establish ¬P, for a statement involving “∀,” we had to show that there
“∃” an x for which it fails. (Such an x is sometimes called a counterexample.)
Example 1.2.7. Consider the statement

∃x ∈ Z, x2 + 2x + 1 > 0.

This statement is clearly true: for instance, we can just take x = 1, and note that 12 + 2 + 1 =
4 > 0.
Example 1.2.8. Consider the statement:

∃x ∈ Z, x2 = 2.

This statement asserts that there exists an integer whose square is 2; but of course, this is
√
false, since 2 is not an integer (indeed, not even a rational number....to be proved later).
To show this, we can check that the set

{y ∈ Z ∣ y = x2 for some x ∈ Z} = {0, 1, 4, 9, 16, 25, . . .}

does not contain 2, or in other words, that

∀x ∈ Z, x2 ≠ 2.

In other words, the negation of a “∃” statement is a “∀” statement.

The moral of the story is that negation changes “∀” to “∃,” and vice-versa (much as in
DeMorgan’s laws, where negation changes ∧ to ∨). In short:

¬ (∀x, P (x)) = ∃x, ¬P (x)

¬ (∃x, P (x)) = ∀x, ¬P (x).

Again, it’s not wise to try and memorize these rules; we’re planning to “learn by doing.”
Which raises the question: what do we plan on doing for the rest of the class?

9
1.3 Proofs (Feb 2)
A proof is a translation of statements that we know (or assume) to be true into a new
true statement, using the rules of logic discussed above. As our first example, we read a
dramatic dialogue from Hutchings’s notes, in which the following proposition3 and its proof
were “discovered.”
Proposition 1.3.1. If x ∈ Z is an even number, then x2 is also even.
Proof. If x is even, then we know by definition that

x = 2y

for some integer y. Squaring both sides, we obtain

x2 = (2y)2 = 4y 2 .

Now, let z = 2y 2 , which is again an integer. By the last equation, we have

x2 = 4y 2 = 2(2y 2 ) = 2z.

Therefore x2 = 2z, so x2 is even.

In this section, we’ll discuss several different proof “techniques.”

1.3.1 Direct proof

A direct proof is usually carried out by starting from a true statement, P, and establishing
a chain of implications
P ⇒ P1 ⇒ P2 ⇒ P2 ⇒ ⋯ ⇒ Q,
from which we conclude that Q is also true.
The above proposition was our first example of a direct proof. Here is another one:
Proposition 1.3.2. For every x ∈ Z, if x is odd then x + 1 is even.
Proof. Let x ∈ Z.4 If x is odd, then by definition,

x = 2y + 1

for some integer y ∈ Z. Adding 1 to both sides, we obtain:

x + 1 = (2y + 1) + 1
= 2y + 2 (1.1)
= 2(y + 1).
Therefore x + 1 is also even.
3
A proposition is a true statement that, while not completely obvious, is not quite as deep or universal
as a theorem.
4
This is how to begin the proof of a “for all” or “for every” statement.

10
 Here is another example, this time involving set theory instead of arithmetic.

Proposition 1.3.3. Let A, B, and C be sets. If A ⊂ B and B ⊂ C, then A ⊂ C.

Proof. We need to show that if x ∈ A, then x ∈ C; i.e., for all x ∈ A, x ∈ C.

Let x ∈ A. Since A ⊂ B, x ∈ A ⇒ x ∈ B. So x ∈ B.
Since B ⊂ C, x ∈ B ⇒ x ∈ C. So x ∈ C.
Since x ∈ A was arbitrary, we’re done.

1.3.2 Proof by cases

Proof by “cases” is a technique that one does very often in combination with the other proof
techniques.

Proposition 1.3.4. For any integer x, x(x + 1) is even.

Proof. Let x be an integer. Then x is either even or odd.

Case 1. If x is even, then x = 2y for some y ∈ Z. So

x(x + 1) = 2y(2y + 1)
(1.2)
= 2 (y(2y + 1)) .

Therefore x(x + 1) is even.

Case 2. If x is odd, then by Proposition 1.3.4, x + 1 is even. So

x + 1 = 2z

for some z ∈ Z. Multiplying both sides by x, we obtain

x(x + 1) = x(2z) = 2(xz).

So x(x + 1) is even.
Since we’ve covered both cases, we’re done.5

1.3.3 Proof by contrapositive

This is where instead of proving an implication P ⇒ Q directly, we instead prove its con-
trapositive ¬Q ⇒ ¬P. As discussed above, we are free to do this, since the contrapositive of
any statement is always logically equivalent to it (i.e. if one is true then the other is true,
and vice-versa).

Proposition 1.3.5. For all integers a, b ∈ Z, if ab ≠ 0, then both a ≠ 0 and b ≠ 0.

5
Technically, we did not show that every integer is either even or odd. This can be proved by induction—
see Example 2.3.2 below.

11
Proof. We prove the contrapositive of the statement, namely:
“If a = 0 or b = 0, then ab = 0”.
This also breaks up into cases:
Case 1. If a = 0, then ab = 0 ⋅ b = 0 and we’re done.
Case 2. If b = 0, then ab = a ⋅ 0 = 0 and we’re also done.

Here is a more interesting example.

Proposition 1.3.6. If the average of four distinct integers is 10, then at least one of them
is greater than 11.
Proof. We show the contrapositive. Note that the negation of the statement “x > 11” is
“x ≤ 11.” So the contrapositive, as a whole, is:
For distinct integers a, b, c, d, if a, b, c, d ≤ 11 then Avg(a, b, c, d) ≠ 10.
Now, since a, b, c, and d are distinct, we may assume without loss of generality6 that
a < b < c < d.
Otherwise, we could relabel the four integers so that they are ordered in this way, without
changing their average.
Then, since d ≤ 11 and c < d, we have
c ≤ d − 1 ≤ 10
so c ≤ 10. But then
b≤c−1≤9
and
a ≤ b − 1 ≤ 8.
So we have shown that in fact
a ≤ 8, b ≤ 9, c ≤ 10, and d ≤ 11.
Adding together these four inequalities, we obtain
a + b + c + d ≤ 8 + 9 + 10 + 11 = 38.
Dividing by 4, we get
a + b + c + d 38 40
≤ < = 10.
4 4 4
In particular, we have
Avg(a, b, c, d) < 10.
But this is precisely the desired statement.
6
This phrase means that we will prove a special case that actually covers the general case of the statement,
with a little more thought.

12
1.3.4 Proof by induction

Suppose that P (n) is a statement that involves a natural number n ∈ N. The Principle of
Mathematical Induction (PMI) states that:
if

1. P (1) is true (“base case”), and

2. For all n ∈ N, the implication P (n) ⇒ P (n + 1) is true (“inductive step”)

then

P (n) is true for all n ∈ N.

The logic behind the principle is as follows. Suppose that the above two assumptions are
true; we claim that P (n) must be true for all n ∈ N.
Let n ∈ N. First, we know that P (1) is true, by (1). Then, by (2), we know that if P (1)
is true then P (2) is also true. But if P (2) is true, then by (2), P (3) is also true. Continuing
in the same fashion, we obtain a chain of implications:

P (1) ⇒ P (2) ⇒ ⋯ ⇒ P (n − 1) ⇒ P (n)

that eventually gets us all the way up to P (n); so P (n) is true.

But since n ∈ N was arbitrary, we get to conclude that P (n) is actually true for all n ∈ N.
The wonderful thing about induction is that you avoid having to check the statement
P (n) separately for n = 1, 2, 3, 4, . . .; instead, you just have to establish (1) and (2) above,
and you’re done.

Proposition 1.3.7. If a1 , . . . , an ∈ Z are such that the product a1 ⋯an = 0, then ai = 0 for
some i with 1 ≤ i ≤ n.

Proof. We use induction on n.

For the base case (n = 1), we have only one number, so the assumption is that a1 = 0.
Then the conclusion is a tautology.
For the inductive step, assume that the claim is true for n (this is known as the inductive
hypothesis). To prove it for n + 1, let a1 , . . . , an+1 be integers such that

a1 ⋯an+1 = 0.

We may rewrite this as

(a1 ⋯an ) ⋅ an+1 = 0.
By Proposition 1.3.5, we know that either a1 ⋯an = 0 or an+1 = 0.

13
 Case 1. If a1 ⋯an = 0, then by the inductive hypothesis, we know that ai = 0 for some
1 ≤ i ≤ n, so we’re done.
Case 2. If an+1 = 0, then we’re also done.
In either case, we have established the claim for n + 1, completing the inductive step. By
the PMI, it follows that the proposition is true for all n ∈ N, as desired.

Here’s another example:

Proposition 1.3.8. The formula

n(n + 1)
1+2+⋯+n=
2
holds for all n ∈ N.

Proof. For the base case, we have

1(1 + 1) 2
1= = =1
2 2
so the formula is true.
For the inductive step, suppose that the formula is true for n. We need to show that the
left-hand side (LHS) of the formula with n + 1 equals the right-hand side (RHS). The LHS
is:
1 + 2 + ⋯ + n + (n + 1).
By the induction hypothesis, we have:

n(n + 1)
1 + 2 + ⋯ + n + (n + 1) = +n+1
2
n
= (n + 1) ( + 1)
2
(1.3)
n+2
= (n + 1) ( )
2
(n + 1)(n + 2)
= .
2
But this is precisely the RHS of the formula for n + 1, completing the inductive step. By the
PMI, the formula holds for all n ∈ N, as desired.

1.3.5 Proof by contradiction

Proof by contradiction is similar to proof by contrapositive, but the hypothesis “P ” can be

any true statement. What you do is assume ¬Q, i.e., that your desired conclusion fails, and
use this to show ¬P. But if P is a priori a true statement, then ¬P is a “contradiction,” and
you get to conclude that ¬Q must be false, i.e., Q is true.

14
 Contradiction arguments should only be used as a “method of last resort,” when a direct
proof or proof by contrapositive is infeasible. When you do get to do one, though, it can be
quite fun.
Below, we will use a contradiction argument to prove a very famous theorem. First we
need the following lemma.7

Lemma 1.3.9. Let p ∈ N be a natural number, and n ∈ Z be an integer. If p > 1, then p does
not divide both n and n + 1.

Proof. We will prove the contrapositive statement: supposing that p ∈ N divides both n and
n + 1, then p = 1.
If p divides both n and n + 1, then there exist integers m, m′ ∈ Z such that

n=p⋅m and n + 1 = p ⋅ m′ .

Subtracting these two equations, we obtain

1 = (n + 1) − n = pm′ − pm
(1.4)
= p(m′ − m).

But then p divides 1, which implies that p = 1.

Theorem 1.3.10 (Euclid, ca. 300BC). There exist infinitely many prime numbers.

Proof. Suppose the contrary. Then there exist only finitely many prime numbers, which we
label as
p1 , p2 , . . . , pm
for some m ∈ N. Here, we assume that pi ≥ 2 by definition.
Consider the number
n = p1 p2 ⋯pm .
By construction, n is divisible by every prime number pi . By Lemma 1.3.9, it follows that
n + 1 is not divisible by any prime numbers. But since n + 1 > 1, it must have at least one
prime factor. This is a contradiction.
It follows that our assumption must have been false, i.e., infinitely many prime numbers
exist.

Note: Later on, we will discuss what it means for a set to be “finite” or “infinite” in much
greater detail.

7
A lemma is a “simple” proposition that will be used afterwards to prove a theorem.

15
2 Set operations and functions
2.1 Intersection and union (Feb 4)
Definition 2.1.1. Let A and B be sets.

• The union A ∪ B is the set of all elements belonging either to A or to B (or both).

• The intersection A ∩ B is the set of all elements belonging to both A and B.

Example 2.1.2.

• {1, 2} ∪ {2, 3, 4} = {1, 2, 3, 4}

• {1, 2} ∩ {2, 3, 4} = {2}

• {0, 2, 4}∩{1, 3} = {} = ∅. This set, which has no elements at all, is known as the empty
set.

• {n ∈ Z ∣ n is divisible by 2}∩{n ∈ Z ∣ n is divisible by 5} = {n ∈ Z ∣ n is divisible by 10}.

This follows by considering the last digit in the decimal expansion of n.

Proposition 2.1.3. Here are several basic properties of intersection and union.

1. A ∪ ∅ = A

2. A ∩ ∅ = ∅

3. A ∩ B ⊂ A ⊂ A ∪ B

4. A ∪ B = B ∪ A, A ∩ B = B ∩ A (commutativity)

5. A ∪ (B ∪ C) = (A ∪ B) ∪ C, A ∩ (B ∩ C) = (A ∩ B) ∩ C (associativity)

6. A ∩ (B ∪ C) = (A ∩ B) ∪ (A ∩ C) (distributivity)

7. A ∪ (B ∩ C) = (A ∪ B) ∩ (A ∪ C).

Proof. We will prove #6.

Recall that two sets are equal if every element of one is also an element of the other, and
vice-versa. Equivalently, we will show that

A ∩ (B ∪ C) ⊂ (A ∩ B) ∪ (A ∩ C)

and
A ∩ (B ∪ C) ⊃ (A ∩ B) ∪ (A ∩ C).

16
 (⊂) Let x ∈ A ∩ (B ∪ C). Then x is in A, and x is either in B or in C. If x ∈ B, then
x ∈ A ∩ B, so x ∈ (A ∩ B) ∪ (A ∩ C). If x ∈ C, then x ∈ C ∩ A = A ∩ C, so x ∈ (A ∩ B) ∪ (A ∩ C).
This shows that x ∈ (A ∩ B) ∪ (A ∩ C), as desired.
(⊃) Let x ∈ (A ∩ B) ∪ (A ∩ C). Then either x ∈ A ∩ B or x ∈ A ∩ C. In the first case, we
have x ∈ A and x ∈ B, so x ∈ B ∪ C. So x ∈ A ∩ (B ∪ C). In the second case, we have x ∈ A
and x ∈ C, so x ∈ C ∪ B = B ∪ C. This shows that x ∈ A ∩ (B ∪ C), as desired.
Since each set is a subset of the other, they are equal.

It is actually easy to picture the above result using Venn Diagrams; still, at the end of
the day, one has to write out the proof formally.

2.2 Collections of sets (Feb 4)

This seems like a good time to discuss the following point about sets and set theory. Some-
times, the elements of a set are themselves sets, as in the following example:
Example 2.2.1. Let S = {{1, 2}, {3, 4}}. This set has two elements, {1, 2} and {3, 4}.
A set whose elements are also sets is sometimes called a collection of sets, although
not always.
Question. Is 1 an element of S?

Answer. No. S has only the two elements listed above.

Question. Is {1, 2} a subset of S?

Answer. No; it is an element.

Question. Is {{1, 2}} a subset of S?

Answer. Yes. This is a set with the single element {1, 2}, which is also an element of S.
Therefore, every element of {{1, 2}} is also an element of S.

Question. Is ∅ a subset of S?

Answer. Yes; in fact the empty set is a subset of any set. (Why?)

Example 2.2.2. Let S be as above and consider T = {{1, 2}, 3}. We have:

• S ∪ T = {{1, 2}, {3, 4}, 3}.

• S ∩ T = {{1, 2}}.

• S ∪ {S} = {{1, 2}, {3, 4}, {{1, 2, }, {3, 4}}}.

Question. Is 2 an element of S ∪ T ?

17
Answer. No; but 3 is.

Question. Is {1, 2} a subset of S ∩ T ?

Answer. No; but it is an element of it.

Question. Is S a subset of S ∪ {S}?

Answer. Yes; by construction, S ∪ {S} contains every element of S.

2.2.1 A word of caution

Now that we have started making collections of sets, it is tempting to become a bit too
grandiose, as follows.
Example 2.2.3 (Russell’s Paradox). Let U be the set of all sets. Define the subset

R = {S ∈ U ∣ S ∈/ S}.

Question. Is R an element of R?

Answer? Suppose that R ∈ R. Then by the definition of R, R ∈/ R, which is a contradiction.

But if R ∈/ R, then R is not an element of itself. By the definition of R, we have R ∈ R,
which is again a contradiction.
Therefore, both the statement R ∈ R and its negation R ∈/ R are false, i.e., R ∈ R is both
true and false.

It is possible to avoid these kinds of problems by being more strict about what kinds of
sets one allows in one’s theory: this is called axiomatic set theory (as opposed to naive set
theory, which we have been practicing). But for everyday use, naive set theory is almost
always good enough.

2.3 More on set operations (Feb 9)

Definition 2.3.1. Suppose A is a subset of a larger “universal” set U . Then

A∁ = {x ∈ U ∣ x ∈/ A}

is called the complement of A.

Example 2.3.2. Let A = {even integers}, U = Z. Then A∁ = {odd integers}.

18
Proof. We need to show that each x ∈ Z is either even or odd. We assume without loss of
generality that x ≥ 0 and use induction.
Base Case: x = 0 is even since 0 = 2 ⋅ 0.
Inductive Step: assume that all integers up to n = k are either even or odd. We want to show
that n = k + 1 is either even or odd.
Case 1: k = 2m, m ∈ Z is even. Then k + 1 = 2m + 1 is odd.
Case 2: k = 2m + 1, m ∈ Z is odd. Then k + 1 = 2m + 1 + 1 = 2m + 2 = 2(m + 1) is even.
This completes the inductive step, and the proof.
Theorem 2.3.3. Suppose A and B are subsets of U . Then (A ∪ B)∁ = A∁ ∩ B ∁
Proof. (⊂) Let x ∈ (A ∪ B)∁ . Then x ∈/ A and x ∈/ B (otherwise x ∈ A or x ∈ B, hence
x ∈ A ∪ B). But this just means x ∈ A∁ ∩ B ∁ .
(⊃) Let x ∈ A∁ ∩B ∁ . Then x ∈/ A and x ∈/ B Ô⇒ x ∈/ A∪B (otherwise, x ∈ A or x ∈ B).
Note: This is “the same” as De Morgan’s Law!
Now, what about A ∩ B ∩ C? We have:
A ∩ B ∩ C = {x ∣ x ∈ A and B and C}.
Similarly,
A ∪ B ∪ C = {x ∣ x ∈ A or B or C}.
Suppose we have a collection of sets {A1 , A2 , A3 , ⋯}. We normally don’t use {} and simply
write
A 1 , A2 , A3 , ⋯
We say that this collection is “indexed” by the natural numbers N. Their intersection is
∞
⋂ An = A1 ∩ A2 ∩ A3 ∩ ⋯ = {x ∣ x ∈ An ∀n}.
n=1
Their union is
∞
⋃ An = A1 ∪ A2 ∪ A3 ∪ ⋯ = {x ∣ x ∈ An for some n}.
n=1
Example 2.3.4. For each n ∈ N, let
An = {1, 2, 3, . . . , n} = {x ∈ N ∣ x ≤ n}
Question. What is ⋃∞
n=1 An ?

Answer. ⋃∞ n=1 An = N. Why? Because given m ∈ N, m ∈ An for any n ≥ m. So m ∈ ⋃n=1 An .

∞

Since m ∈ N was arbitrary,

∞ ∞
⋃ An ⊂ N, ⋃ An ⊃ N.
n=1 n=1
Therefore, n=1 An
⋃∞ = N.
Question. What is ⋂∞
n=1 An ?

Answer. ⋂∞
n=1 An = {1}.

19
2.4 Functions (Feb 11)
Consider two sets A (the domain) and B (the codomain). A function (written f ∶ A → B) is
an assignment of a value f (x) ∈ B for each element x ∈ A. A “function” can also be called a
“map”. The set of all outputs

f (A) = {y ∈ B ∣ y = f (x) for some x ∈ A}

is called the image of f (Note: the image of f is a subset of B, i.e. f (A) ⊂ B).
Given a subset Y ⊂ B, the pre-image (or inverse image) is the set of all points that map
to Y . I.e.
f −1 (Y ) = {x ∈ A ∣ f (x) ∈ Y }.
Note that the image of a subset of A is a subset of B, and the pre-image of a subset of B is
a subset of A.
Example 2.4.1. Let A = {1, 2, 3}, B = {0, 2, 4, 6, 8}. Let

f ∶A→B
x z→ 2x.

Question. What is f (A)?

Answer: f (A) = f ({1, 2, 3}) = {2, 4, 6}

Question. What is f ({1, 3})?

Answer: {2, 6}

Question. What is f −1 ({4, 6})?

Answer: {2, 3}

Question. What is f −1 ({8})?

Answer: ∅, because 4 ∈/ A.

Example 2.4.2. Let

f ∶Z→Z
x z→ x2

Using the same sets A and B as above,

f ({1, 2, 3}) = {f (1), f (2), f (3)} = {1, 4, 9}

f ({−2, −1, 0, 1, 2}) = {0, 1, 4}

20
Question. What is f −1 ({1})? What about f −1 ({−1})? f −1 ({2, 3, 5})?

Answer: f −1 ({1}) = {x ∈ Z ∣ x2 = 1} = {−1, 1}

f −1 ({−1}) = ∅,
f −1 ({2, 3, 5}) = ∅

Question. What is f −1 ({N})?

Answer: The pre-image of N is the set of all numbers x ∈ Z such that x2 ∈ N. This is every
integer except 0. So f −1 ({N}) = Z ∖ {0}.

Note: We will sometimes use the set difference between two sets:

A ∖ B = {x ∈ A ∣ x ∈/ B} = A ∩ B ∁ .

Definition 2.4.3. A function f is injective (one to one) if f maps distinct elements of the
domain to distinct elements of the codomain. I.e. for each y ∈ B, there is at most one x ∈ A
such that f (x) = y.
If a function f is injective, we can say that if x1 , x2 ∈ A and that x1 ≠ x2 , then f (x1 ) ≠
f (x2 ). Example 2.4.1 is injective but Example 2.4.2 is NOT injective (because f (n) = f (−n)
for all n ≠ 0).
Definition 2.4.4. A function is surjective (onto) if f (A) = B (the image of the domain
equals the codomain). I.e. for every y ∈ B, there exists at least one x ∈ A such that f (x) = y.
f (A) = B ⇐⇒ f −1 (y) ≠ ∅
for all y ∈ B. Neither Example 2.4.1 nor Example 2.4.2 is surjective.
Example 2.4.5. f ∶ Z Ð→ N ∪ {0}, x ↦ ∣x + 2∣. Is this surjective? Yes!

Proof. Given y ∈ N ∪ {0}, we let x = y − 2. Clearly x ∈ Z.

Furthermore,
f (x) = f (y − 2) = ∣y − 2 + 2∣ = ∣y∣ = y.
Since y ∈ N ∪ {0} was arbitrary, f is surjective.

While f is surjective, it is not injective. Why?

Proof. Consider y = 1. Clearly, y ∈ N ∪ {0}. f (−3) = 1, and f (−1) = 1. But −3 ≠ −1. So f is

not injective.

Definition 2.4.6. A function f is bijective if it is both injective and surjective.

Example 2.4.7. Construct a bijection f ∶ Z Ð→ N. Is this possible?

21
Yes!
⎧
⎪
⎪
⎪
⎪ 0 x=0
⎪
⎪
f (x) = ⎨2x + 1 x > 0
⎪
⎪
⎪
⎪
⎪
⎪ 2∣x∣ x<0
⎩
Although Z seems to have more elements than N, we can map every element in Z to every
element of N. Interesting...

If a function f is bijective, there exists a unique inverse map f −1 such that

f −1 (y) = x ⇐⇒ f (x) = y.

Since f is surjective, for any y, there exists an x such that f (x) = y. Since f is injective,
that x is unique. Consequently, (f ○ f −1 )(x) = x , and (f −1 ○ f )(x) = x.
Note: The notation “f −1 ” has two different meanings! It is both used to denote the
“inverse image” of a set, and to denote the “inverse function” of a bijective function. It
should always be clear from the context which one we mean.

2.5 Cartesian product and graphs of functions (Feb 16)

Definition 2.5.1. Given two sets, S and T , the Cartesian Product is

S × T = {(x, y) ∣ x ∈ S and y ∈ T }

Example 2.5.2.

S = {0, 1, 2}
T = {3, 4}

S × T = {(0, 3), (0, 4), (1, 3), (1, 4), (2, 3), (2, 4)}
S ×T ≠T ×S
Example 2.5.3. N × N =
{(x, y) ∣ x ∈ N, y ∈ N}

⎧
⎪ (1, 1) (1, 2) (1, 3) (1, 4) ...⎫
⎪
⎪
⎪
⎪ ⎪
⎪
⎪
⎪
⎪(2, 1) (2, 2) (2, 3) (2, 4) ...⎪
⎪ ⎪
⎪
⎨ ⎬
⎪
⎪
⎪ ⋮ ⎪
⎪
⎪
⎪
⎪
⎪ ⎪
⎪
⎪
⎪
⎩ ⋮ ⎪
⎭

22
N×N=

Definition 2.5.4. Given a function f ∶ S ↦ T , the graph of f is the subset of S × T given

by {(x, y) ∈ S × T ∣ y = f (x) for some x}.
Example 2.5.5. Graph of f ∶ N Ð→ N
x z→ x2

{(1, 2), (2, 4), (3, 9), (4, 16), ⋯}

Question. Let S = {1, 2, 3} and T = {4, 5}. Which of the following subsets of S × T are equal
to the graph of some function f ∶ S Ð→ T ?
Example 2.5.6. {(1, 4), (2, 4), (3, 5)}

5
4

1 2 3

Answer: Yes... b/c every x ∈ S is mapped to exactly one value y ∈ T .

f (1) = 4
f (2) = 4
f (3) = 5
Example 2.5.7. {(1, 4), (1, 5), (2, 4), (3, 5)}
Answer:
5
4
1 2 3

No! f (1) = 4, f (1) = 5???

⇒ f is not a function!
FAILS THE VERTICAL LINE TEST!
Example 2.5.8. {(1, 4), (2, 5)}

1 2 3

23
f (3) = ???
Answer: Not a function, since f (3) = ”undefined.”

24
3 Infinity
3.1 Finite and infinite sets (Feb 16)
Let
Pn = {1, 2, 3, . . . , n} ⊂ N
be the set of all integers from 1 to n.
Definition 3.1.1. We say that a set is finite if there exits a bijection from Pn to S, for
some n ∈ N. In this case, we write
∣S∣ = n,
and say “the order of S is n.”
Otherwise, we say that S is infinite.
Example 3.1.2. Consider the set

S = {∗, †, @, x, Ivanka Trump, potato}

Is S finite? Let’s define a map as follows:

1 z→ ∗
2 z→ @
3 z→ †
(3.1)
4 z→ Ivanka
5 z→ x
6 z→ potato.

This gives us a bijection

∼
→ S.
P6 Ð
We conclude that S is a finite set of order 6 (i.e. ∣S∣ = 6).
Note: We will sometimes use the following notation for maps:
↪ = injection
↠ = surjection
∼
→ = bijection
Ð
Example 3.1.3. The sets

N, N × N, N × Z, Z × N, and Z × Z

are all infinite.

25
Example 3.1.4. If A and B are finite, and
∣A∣ = n = ∣B∣,
then there exists bijections
∼
→A
Pn Ð
∼
→ B.
Pn Ð
∼
But any bijection is invertible. So we also have A Ð
→ Pn , and the composition
∼ ∼
AÐ →B
→ Pn Ð
is a bijection.
Informal definition. Two sets S and T have “the same size” if there exists a bijection
(i.e. a 1-1 correspondence) between them.
Example 3.1.5. Z N
Do these have the same size?
{..., −4, −3, −2, −1, 0, 1, 2, 3, 4, ...}
{1, 2, 3, 4, ...}
We saw last time that there does exist a bijection:
∼
ZÐ
→N
∼
and Z ←
ÐN
So they have the same size!
Question. Do all the sets of Example 3.1.3 have the ”same size?” I.e., is it possible to
construct bijections between them?
More generally, are all infinite sets the “same size?”

3.2 Countable sets (Feb 18)

Definition 3.2.1. Suppose that S is an infinite set. We say that S is countably infinite
∼ ∼
if there exists a bijection N Ð
→ S (or S Ð
→ N).
The reason for this definition is as follows. If we are given a bijection f ∶ N → S, this
gives us the ability to label each element of S as
sn = f (n).
Then we can list all the elements of S as:
S = {s1 , s2 , s3 , s4 , ...}.
So, a countably infinite set is precisely one whose elements can be “counted” (i.e., indexed)
using the natural numbers.

26
Definition 3.2.2. We say that a set S is countable if it is either finite or countably infinite.
∼
Example 3.2.3. We saw last time that there is a bijection g ∶ N → Z (specifically, the inverse
∼
of the bijection Z → N of Example 2.4.7). Therefore Z is countably infinite.
Example 3.2.4. Let

E = {x ∈ Z ∣ x = 2n for some n ∈ Z} = {. . . , −4, −2, 0, 2, 4, . . .}

be the set of even integers. To show that E is countable, we just need to construct a bijection
∼
N → E.

To do this, we first write down a bijection:

f ∶Z→E
(3.2)
n ↦ 2n.

To obtain a bijection N → E, we just take the composition with the map g of the previous
example:
g f
N Ð→ Z Ð→ E.
Since the composition of two bijections is a bijection, f ○ g ∶ N → E is the required bijection.
We will now prove a few lemmas that will save us from having to construct bijections
explicitly every time we want to show that a set is countable.

Lemma 3.2.5. Any subset A ⊂ N is countable.

Proof. Let’s order the elements of A ∶

a1 < a2 < a3 < ...

Either A contains a largest element, or not.

Case 1. A contains a largest element. Then the list above

a1 < a2 < a3 < ... < ak

ends at the k’th element, ak . The map

Pk → A

n z→ an
is a bijection, so A is finite (⇒ countable).
Case 2. A does not contain a largest element (i.e ∀a ∈ A, ∃ a′ ∈ A s.t a′ > a). Then the
above list
a1 < a2 < a3 < ...

27
goes on forever. So the map
N Ð→ A
n z→ an
is a bijection. ⇒ A is countably infinite (⇒ countable).

Example 3.2.6. Let

Q = {m ∈ N ∣ m ∈ N divides n ⇒ m = 1 or m = n}
= {1, 2, 3, 5, 7, 11, 13, 17, 19, 23, ...}.

This is the set of all prime numbers (including 1). We proved in Theorem 1.3.10 above that
this set is infinite! By the previous Lemma, Q is in fact countably infinite.

Lemma 3.2.7. Suppose there exists an injective map f ∶ S ↪ N. Then S is countable.

Proof. Define the subset A = f (S) ⊂ N. By the previous Lemma, A is countable. But we
can define a map
g ∶ S Ð→ A
x z→ f (x).
Since f is injective, g is also injective. Since A = f (S), g is also surjective. Therefore g is a
bijection. Since A is countable, S is also countable.

Lemma 3.2.8. Let A ⊂ S be a subset of a countable set S. Then A is countable.

Proof. Since S is countable, there exists an injection f ∶ S ↪ N. (If S is finite, we have

∼ ∼
S → Pn ⊂ N, which is an injection; if S is countably infinite, we have a bijection S → N.) But
then the “restriction”

f ∣A ∶ A → N
x ↦ f (x)

is also an injection. So, by the previous lemma, A is countable.

Lemma 3.2.9. Suppose that there exists a surjective map N ↠ A. Then A is countable.

Proof. Given f ∶ N ↠ S, we can choose a subset M ⊂ N such that the restriction

f ∣M ∶ M → A
(3.3)
x ↦ f (x)

is bijective. (For each y ∈ S, choose exactly one x ∈ f −1 ({y}), and let x ∈ M.)
∼ ∼
Case 1. M is finite. Pn Ð→M Ð → S is bijective ⇒ S is finite.
∼ ∼
Case 2. M is infinite. N Ð →M Ð → S is a bijection. ⇒ S is countably infinite.

28
Theorem 3.2.10. Let A and B be countable sets. Then A ∪ B is countable.

Proof. We may assume without loss of generality that A∩B = ∅. (If not, replace B by B ∖A,
which is countable by Lemma 3.2.8.)
Since A and B are countable, there exist injective maps

f ∶A↪N

g ∶ B ↪ N.
Consider the map
h ∶ A ∪ B Ð→ Z
⎧
⎪f (x) x ∈ A
⎪
h(x) = ⎨
⎪
⎩−g(x) x ∈ B.
⎪
∼
Then h is injective. So the composition A ∪ B ↪ Z Ð→ N is a composition of injective maps,
so is injective.
Therefore, by the lemma above, A ∪ B is countable.

Question. Is N × N countable?
Answer: Yes. We showed two ways to “count” the points...either by using “boxes,” or by
using “diagonals.”
We also showed how to count Z × Z. In the picture below, we label the 1st, 2nd, 3rd, ...,
37th dot, etc.

...
37 31
...
17 13
4
5 3

... ... ... 1 2 ... ...

10

21 26
...

More generally, we have the following result.

29
Theorem 3.2.11. Let A and B be countable sets. Then the Cartesian product A × B is
countable.

Proof of Theorem 3.2.11. Since A and B are countable, there exist surjective maps

f ∶N↠A

g ∶ N ↠ B.
(Why?) Consider the “product map”

f × g ∶ N × N Ð→ A × B

(x, y) ↦ (f (x), g(y)).

We claim that this is surjective. Given (z, w) ∈ A × B, there exists x ∈ A such that f (x) = z
and y ∈ B such that f (y) = w. Then, by definition, we have

(f × g)(x, y) = (z, w).

Since (z, w) ∈ A × B was arbitrary, this shows that f × g is surjective.

But then the composition
∼ f ×g
NÐ→ N × N ÐÐ→ A × B
is surjective. Therefore, by the lemma, A × B is countable.

Corollary 3.2.12. Let A1 , . . . , An be a finite collection of countable sets. Then the cartesian
product
A1 × A2 × ⋯ × An
is countable.

Proof. This follows by induction from the previous theorem (see homework).

Example 3.2.13. We now know that the sets N, Z, Z × N, Z × Z, Z × Z × Z × N, etc., are all
countable.
Question: Do there exist uncountable sets??

3.3 Uncountable sets (Feb 23)

Above we discussed countable sets. An infinite set S is said to be uncountable if it is not
countable, i.e., there does not exist any bijection from N to S. The goal of this section is to
show that indeed, uncountable sets exist. This fact was discovered by Georg Cantor in 1878.

30
 Given a set S and a positive integer n ∈ N, we shall denote the n-fold cartesian product
by
n times
³¹¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹·¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ µ
S n =S × S × S × ⋯ × S
= {(a1 , a2 , . . . , an ) ∣ ai ∈ S for i = 1, . . . , n}.

The elements of this set are sometimes called “n-tuples.” (For n = 2, we say “pairs,” and for
n = 3 we say “triples,” etc.)
Example 3.3.1. Take S = {0, 1}. Then ∣S∣ = 2. It was shown on homework that ∣S × S∣ = 4,
by constructing an explicit bijection with P4 . More generally, we have:

Proposition 3.3.2. For S = {0, 1}, ∣S n ∣ = 2n .

Proof. The base case is clear. For induction, assume the result for n. We have

S n+1 = S n × S = {0, 1}n × {0, 1}

= {0, 1}n × {0} ∪ {0, 1}n × {1}
= {(a1 , a2 , . . . , an , 0} ∪ {b1 , b2 , . . . , bn , 1}.

Since the union is “disjoint,” i.e. the sets do not overlap, we have:

∣S n+1 ∣ = ∣{0, 1}n × {0}∣ + ∣{0, 1}n × {1}∣

= 2n + 2n
= 2n+1 .

This completes the induction.

Notice that the order of this set grows “exponentially” with n; although of course if S is
finite, then S n is still finite for each n ∈ N. What we need is a much bigger set.
Definition 3.3.3. Given any set S, we denote the “infinite” cartesian product by

SN = S × S × S × ⋯
= {(a1 , a2 , a3 , ⋯) ∣ ai ∈ S for i ∈ N}.

An element of S N is an “infinite string” or “sequence” of elements of S.

Example 3.3.4. Let S = {0, 1}. Then the set S N is the set of all infinite binary sequences.
Here are a few examples of elements of this set.

• (0, 0, 0, 0, 0, . . .) = sequence of all 0’s

• (1, 1, 1, 1, 1, . . .) = sequence of all 1’s

31
 • (1, 0, 0, 1, 0, 0, 1, 0, 0, . . .) sequence of 1, 0, 0 repeating ad infinitum

• (1, 1, 1, 0, 1, 0, 1, 0, 0, 0, 1, 0, 1, 0, 0, 0, 1, 0, 1, 0, 0, 0, 1, . . .) = the sequence defined by

⎧
⎪1 n is prime
⎪
an = ⎨
⎪
⎩0 n is composite.
⎪

• (1, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, . . .) = digits in the binary expansion of π (essentially a ran-

dom sequence of 0’s and 1’s).

• (1, 0, 0, 1, 0, 1, 0, 0, 0, 1, 1, 0, . . .) = transmission data of Thursday’s lecture video over the

internet...

The point here is that the set {0, 1}N contains an absolutely vast array of possible se-
quences....in fact, it turns out to be even bigger than the set of natural numbers N.

Theorem 3.3.5 (G. Cantor, 1878). The set {0, 1}N is uncountable.

Proof. Suppose, on the contrary, that {0, 1}N is countable. Then it is possible to list (i.e.
index) all of its elements by the natural numbers, as follows:

{0, 1}N = {s1 , s2 , s3 , . . .}.

For example, our list might consist of the following sequences:

s1 = 0, 1, 0, 1, 0, ⋯
s2 = 1, 0, 1, 0, 0, ⋯
s3 = 1, 1, 1, 0, 1, ⋯
s4 = 0, 1, 1, 0, 0, ⋯
s5 = 1, 0, 0, 1, 1, ⋯
⋮ ⋮ ⋱

We now construct a new sequence, s, as follows. If the first digit of s1 is 0, let the first
digit of s be 1, and vice-versa. If the second digit of s2 is 1, let the second digit of s be 0, and
vice-versa. Continuing, if the n’th digit of sn is 0, let the n’th digit of s be 1, and vice-versa.
So, in the above example, we would change each digit on the diagonal,

s = 1, 1, 0, 1, 0, ⋯
1, 1, 1, 0, 0, ⋯
1, 1, 0, 0, 1, ⋯
0, 1, 1, 1, 0, ⋯
1, 0, 0, 1, 0, ⋯,

32
to obtain the sequence s = (1, 1, 0, 1, 0, . . .).
By construction, the n’th element of s is different from the n’th element of sn , for each
n ∈ N. Therefore, as elements of {0, 1}N , we have

s ≠ sn for all n ∈ N.

But this means that the sequence s appears nowhere on the list s1 , s2 , s3 , . . . , which contra-
dicts our assumption.
We conclude that no such list can include all the elements of {0, 1}N , i.e., this set is
uncountable.

3.4 Another look at Cantor’s proof (Feb 25)

Above, we viewed the set
{0, 1}N = {(a1 , a2 , a3 , . . .)}
as the set of all sequences of 0’s and 1’s. Equivently, we may think of any such sequence as
a function from N to {0, 1} ∶ given such a sequence, we can define a function by

f (n) = an ∈ {0, 1}

and vice-versa. So it is completely equivalent to define:

{0, 1}N = {f ∶ N → {0, 1}}

as the set of all possible functions from N to {0, 1}. More generally, given two sets S and T,
we can define
T S = {set of all functions from S to T }.
We should note that two functions f and g are equal if and only if

f (x) = g(x) ∀ x ∈ S.

So, to show that f ≠ g, you just have to make sure that there exists at least one x ∈ S such
that
f (x) ≠ g(x).
(Above, we treated sequences in the same way.)
It’s worth reviewing the proof of Cantor’s theorem from this perspective.

Proof of Theorem 3.3.5, redux. Assume that the set of all functions from N → {0, 1} can
be indexed by the natural numbers:

{0, 1}N = {f1 , f2 , f3 , . . .}.

33
Define a new function by

f ∶ N → {0, 1}
⎧
⎪0 if f (n) = 1
⎪
n↦⎨
⎪
⎩1 if f (n) = 0.
⎪

Since each n ∈ N is mapped to exactly one value, this is a function. And, by construction,
we have
f (n) ≠ fn (n) ∀ n ∈ N,
and so
f ≠ fn
as functions, for all n ∈ N. But this contradicts our assumption.

This proof can easily be adapted to show that for any set S with ∣S∣ ≥ 2, the set S N is
uncountable. For instance, with S = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9}, we could just define
⎧
⎪
⎪
⎪ 1 if fn (n) = 0
⎪
⎪
⎪
⎪
⎪
⎪ 2 if fn (n) = 1
⎪
⎪
⎪
f (n) = ⎨ ⋮
⎪
⎪
⎪
⎪
⎪
⎪ 9 if fn (n) = 8
⎪
⎪
⎪
⎪
⎪
⎩0
⎪ if fn (n) = 9.

In fact, any choice so that f (n) ≠ fn (n) works.

The next question one might ask is if there exist even bigger sets than {0, 1}N . In fact,
we have the following:

Theorem 3.4.1. For any set S, there does not exist a bijection between S and {0, 1}S .

Proof. Adapt the proof of Cantor’s theorem (exercise!).

So, given any set S, one can always make a strictly larger set just by taking the set of
all functions from S to {0, 1}. This shows that “infinity” actually comes in infinitely many
different sizes.

34
4 Number systems
4.1 Rules of Arithmetic (March 9)
Below is the set of all integers
Z = {. . . − 3, −2, −1, 0, 1, 2, 3, . . .}
This is more than just a set, it comes equipped with arithmetical operations.

4.1.1 Addition

What is “addition”? It is a map:

f+ ∶ Z × Z Ð→ Z
(a, b) z→ f+ (a, b) = “a + b”(Just a notation)
This map satisfies the following axioms, for any a, b, c ∈ Z.

A1: a+b=b+a i.e. f+ (a, b) = f+ (b, a)

A2: (a + b) + c = a + (b + c) i.e. f+ (f+ (a, b), c) = f+ (a, f+ (b, c))
A3: a+0=a
A4: a + (−a) = 0
Note: −(−a) = a by definition
. Note: You can prove these axioms by induction from even simpler axioms (Peano’s axioms).
It is good to see how to prove the familiar laws of arithmetic from the above axioms.
Proposition 4.1.1. if a + x = a, then x = 0.

Proof. if x + a = a, then,
(−a) + (x + a) = (−a) + a (Because f+ is a function). But
(−a) + a = a + (−a) (A1)
=0 (A4)
Therefore
(−a) + (x + a) = 0
= (−a) + (a + x) (A1)
= ((−a) + a) + x (A3)
=0+x (A4)
=x+0 (A1)
=x (A3)
∴ x = 0.

35
Definition 4.1.2. “a − b” = a + (−b). In the last proof, we just ”subtracted a from both
sides”.

4.1.2 Multiplication

We can also “multiply” two integers together: It is a map:

f⋅ ∶ Z × Z Ð→ Z

(a, b) z→ f⋅ (a, b) = “a ⋅ b”
This map satisfies the following axioms, for any a, b, c ∈ Z.

M1: a⋅b=b⋅a i.e. f⋅ (a, b) = f⋅ (b, a)

M2: (a ⋅ b) ⋅ c = a ⋅ (b ⋅ c)
M3: 1 ⋅ a = a(= a ⋅ 1)
M4: a ⋅ (b + c) = a ⋅ b + a ⋅ c i.e. f⋅ (a, f+ (b, c)) = f+ (f⋅ (a, b), f⋅ (a, c))
M5: If x ⋅ y = x ⋅ z and x ≠ 0, then y = z

Note: You can also prove these by induction using addition.

1⋅a=a
2⋅a=a+a
3 ⋅ a = 2a + a
...
(n + 1) ⋅ a = n ⋅ a + a

Proposition 4.1.3. a ⋅ 0 = 0.

Proof.

a ⋅ 0 + a ⋅ 0 = a ⋅ (0 + 0) (M 4)
=a⋅0 (A3)

Subtracting a ⋅ 0 from both sides, we obtain

a⋅0=0 (M 4)

36
Proposition 4.1.4. (−a) ⋅ (−b) = a ⋅ b.

Proof. We first show that

(−a) ⋅ b = −(a ⋅ b)
(−a) ⋅ b + a ⋅ b = ((−a) + a) ⋅ b (M 4)
=0⋅b (A3)
=0 (last proposition)
Subtracting −(a ⋅ b) from both sides, we now calculate

(−a) ⋅ (−b) − (a ⋅ b) = (−a) ⋅ (−b) + [−(a ⋅ b)]

= (−a) ⋅ (−b) + (−a) ⋅ b (just showed)
= (−a) ⋅ ((−b) + b)
= (−a) ⋅ 0
=0

∴ (−a) ⋅ (−b) = a ⋅ b

Note: we just showed that the product of two negative numbers is positive!

4.1.3 Ordering

In addition to “+” and “⋅”, Z comes with an “ordering”, < or >.

O1: ∀a, b ∈ Z, exactly one of the following is true:

a < b, or
a > b, or
a=b
O2: if a < b and b < c, then a < c
O3: if b < c, then a + b < a + c
O4: if a, b > 0, then a ⋅ b > 0

We write a ≤ b if a < b or a = b, etc. Note: Rather than a function on Z × Z, “<” is

represented as a subset
S< ⊂ Z × Z
where
(a, b) ∈ S< iff “a < b”.

37
Rewriting O2 and O3 from before:
O2: if (a, b) ∈ S< and (b, c) ∈ S< , then (a, c) ∈ S<
O3: if (b, c) ∈ S< , then (f+ (a, b), f+ (a, c)) ∈ S<

Proposition 4.1.5. The following properties are true of “<”.

i) if x > 0 and y < z, then x ⋅ y < x ⋅ z
ii) if x < 0 and y < z, then x ⋅ y > x ⋅ z
iii) if x ≠ 0, then x2 > 0

Proof. i) if z > y then

z + (−y) > y + (−y) = 0 (O3)
⇒z−y >0
Since x > 0,
x ⋅ (z − y) > 0 (O4)
⇒x⋅z−x⋅y >0 (M4)
⇒x⋅z−x⋅y+x⋅y >x⋅y (A2, A3, A4)
⇒ x ⋅ z > x ⋅ y.

ii) if x < 0 and z > y, then

−[x ⋅ (z − y)]
= (−x) ⋅ (z − y) (by proof of proposition above)
Since x > 0, −x < 0
So by (i), we have:
(−x) ⋅ (z − y) > 0
⇒ −[x ⋅ (z − y)] > 0
⇒ 0 > x(z − y)
⇒ 0 > xz − xy
⇒ xy > xz.

iii) By 01, either x > 0 or x < 0 (because x ≠ 0)

if x > 0, then x2 = x ⋅ x > 0 by O4
if x < 0, then
x2 = (−x) ⋅ (−x) > 0 (by proposition above)
since −x > 0.

38
4.1.4 Division
p
We now want to “extend” Z to contain fractions , where p, q ∈ Z.
q
p
We want to build a set that contains these elements. Notice that is a pair of integers,
q
where q ≠ 0.

First Attempt: take the Cartesian product:

Z × (Z ∖ {0}).
p
We think of a pair (p, q) as a fraction . How do we add, subtract, multiply these pairs?
q
p p′ pq ′ + qp′
Definition 4.1.6. + ′ =
q q qq ′

(p, q) + (p′ , q ′ ) = (pq ′ + qp′ , qq ′ ) ∈ Z × (Z ∖ {0})

Does this “+” equation also satisfy A1 - A4 above? Yes.

A1: (p, q) + (p′ , q ′ ) = (pq ′ + qp′ , qq ′ )

= (p′ q + q ′ p, q ′ q)
= (p′ , q ′ ) + (p, q) (by the axioms for Z)

A2: ((p, q) + (p′ , q ′ )) + (p′′ , q ′′ ) = (pq ′ + qp′ , qq ′ ) + (p′′ , q ′′ )

= (p′ q + q ′ p, qq ′ ) + (p′′ , q ′′ )
= ((pq ′ , qp′ )q ′′ + qq ′ p′′ , qq ′ q ′′ )
= (pq ′ q ′′ + p′ q ′′ q + q ′ p′′ q, qq ′ q ′′ )

In fact, all axioms work!

p p′ pp′′
Definition 4.1.7. ⋅ ′ = ′′′ .
q q qq

(p, q) ⋅ (p′ , q ′ ) = (pp′ , qq ′ )

1
But we want: p ⋅ =1
p
1
whereas, in Z × (Z ∖ {0}), we have p ⋅ = (p, 1) ⋅ (1, p) = (p, p). But in the set Z × (Z ∖ {0}),
p
(p, p) ≠ (1, 1).

So what we need to do is define a new set, called Q, the rational numbers, by “declaring”
that (p, p) = (1, 1). And more generally, that (p, q) = (p′ , q ′ ) if and only if ∃r ∈ Z s.t.

39
 p ⋅ r = p′ and q ⋅ r = q ′
or
p = r ⋅ p′ and q = r ⋅ q ′ .

4.2 Equivalence relations (Mar 11)

Goal: construct the set of rational numbers, Q. This should be composed of pairs of integers
p
(p, q), with q > 0, representing a fraction .
q

We need to declare that

(2, 3), (4, 6), (6, 9), etc.
2 4 6
all represent the same number. (Because = = = . . . )
3 6 9

We’re going to write this as:

(2, 3) ∼ (4, 6)
but:
(5, 8) ≁ (16, 30).
Why? Because
(30 ⋅ 5 = 150) ≠ (16 ⋅ 8 = 128).
We’re now going to describe a new set operation that allows us to “declare different elements
to be the same.”
Definition 4.2.1. We say that a relation ∼ is “an equivalence relation” if
1) ∀x ∈ S, x ∼ x(reflexive)
2) ∀x, y ∈ S, if x ∼ y then y ∼ x(symmetric)
3) ∀x, y, z ∈ S, if x ∼ y and y ∼ z, then x ∼ z(transitive)
Example 4.2.2. Consider the relation on Z given by

x ∼ y ⇐⇒ x2 = y 2 .

Is this an equivalence relation? Need to check 3 properties:

1) x ∼ x ? x2 = x2 (reflexive)
2) x ∼ y ⇒ x2 = y 2
⇒ y 2 = x2
⇒ y ∼ x (symmetric)
3) x ∼ y, y ∼ z
⇒ x2 = y 2 and y 2 = z 2

40
 ⇒ x2 = z 2
⇒ x ∼ z (transitive)

This is an equivalence relation.

1 ∼ −1
2 ∼ −2
...
n ∼ −n
0∼0
Example 4.2.3. Consider the relation on Z defined by:

x ∼ y ⇐⇒ x ≤ y

Reflexive: x ∼ x?
x ≤ x.

Transitive: if x ≤ y and y ≤ z, then x ≤ z.

by O3, this is true.

Symmetric: if x ≤ y, is y ≤ x?
2 ≤ 3 but 3 ≰ 2.

So this is not an equivalence relation!

Example 4.2.4. Let S be a course with 9 students, sitting in a lecture hall. The relation
“a is sitting next to b”.
Symmetric? Yes.
Reflexive? Yes.
Transitive? No.
This is not an equivalence relation.
However, let’s try: “a is sitting in the same row as b”.

a f i

e b d
g h c
Symmetric? Yes.
Reflexive? Yes.
Transitive? Yes.
So this is an equivalence relation!

41
Definition 4.2.5. Given an equivalence relation on a set, and an element a ∈ S, we write

[a] = {b ∈ S ∣ a ∼ b}

to denote the equivalence class of a.

[a] ⊂ S.
Sometimes, a is called a “representative” of [a].
In example 1, x ∼ y ⇐⇒ x2 = y 2 .
We have:
[1] = {1, −1}
[2] = {2, −2}
...
[n] = {n, −n}
[0] = {0}
In example 4.2.4 (lecture hall),

[a] = {a, f, i}
[f ] = {f, a, i} = [a]
[g] = {g, h, c}

Proposition 4.2.6. Suppose that ∼ is an equivalence relation on a set S.

i) if x ∼ y, then [x] = [y] as a subset of S.

ii) if x ≁ y, then [x] ∩ [y] = ∅.

Proof. i) Assume x ∼ y. Let us show that [x] ⊂ [y].

Given z ∈ [x], we have z ∼ x.
By transitivity, this is the same as x ∼ y,
z ∼ x ∼ y ⇒ z ∼ y ⇒ z ∈ [y].
Since z ∈ [x] was arbitrary, we’re done.
ii) We show the contrapositive:
Assume [x] ∩ [y] ≠ ∅.
Then let z ∈ [x] ∩ [y]. Then z ∼ x, and z ∼ y.
But then by transitivity, x ∼ z ∼ y ⇒ x ∼ y.

Definition 4.2.7. We write S/ ∼ to denote “S modulo ∼”, the set of all equivalence classes
of ∼ on S. This is a collection of disjoint subsets of S.

42
 if A ∈ X/ ∼ and B ∈ X/ ∼ ,
then if A ≠ B, then
A∩B =∅
Also,
⋃ A=S
A∈S/∼

(because every element x ∈ S is contained in its own equivalence class, A = [x]).

In example 1, we have
Z/ ∼= {{0} , {1, −1} , {2−, −2} , . . . }
We could write
{[0], [1], [2], [3], . . . }
There is an obvious bijection
N ∪ {0} z→ Z/ ∼
n z→ [n]
Z/ ∼ is countably infinite. In example 4.2.4 (lecture hall), we had

S/ ∼= {{a, i, f } , {g, h, c} , {e, b, d}}

= {[a], [b], [c]}
In S/ ∼, we have:
[a] = [f ], and
[g] = [c], etc.

4.3 The rational numbers (Mar 11)

Now, consider the following relation on Z × N

(p, q) ∼ (p′ , q ′ ) ⇐⇒ pq ′ = p′ q.

Note in particular that if p′ = q ⋅ r, for r ∈ N, then

p ⋅ q ′ = p ⋅ (q ⋅ r) = (p ⋅ r) ⋅ q = p′ ⋅ q.

So [(p, q)] = [(p′ , q ′ )]. (This corresponds to “cross-multiplication!”)

p p′
= ⇐⇒ pq ′ = p′ q
q q′

43
Proposition 4.3.1. ∼ is an equivalence relation.

Proof.
1) (p, q) ∼ (p, q)? Yes.
2) if (p, q) ∼ (p′ , q ′ ), does (p′ , q ′ ) ∼ (p, q)?
pq ′ = p′ q
⇒ p′ q = pq ′ (M1)
⇒ (p′ , q ′ ) ∼ (p, q)
3) if (p, q) ∼ (p′ , q ′ ) and (p′ , q ′ ) ∼ (p′′ , q ′′ ), does (p, q) (p′′ , q ′′ )?
Have: pq ′ = p′ q and p′ q ′′ = p′′ q ′ .
Need: pq ′′ = p′′ q.
pq ′ q ′′ = p′ qq ′′
⇒ pq ′′ q ′ = q(p′ q ′′ )
⇒ pq ′′ = qp′′

Since q ′ ≠ 0, by (M5), we have pq ′′ = qp′′

⇒ (p, q) ∼ (p′′ , q ′′ )

Definition 4.3.2. The rational numbers, Q, are defined as:

Q = (Z × N)/ ∼

where ∼ is the equivalence relation:

(p, q) ∼ (p′ , q ′ ) ⇐⇒ pq ′ = p′ q

So a natural number is an equivalence class of pairs

[(p, q)] = {(p, q), (2p, 2q), (3p, 3q), . . . }

p
These all represent the “fraction” .
q

We can define the equations “+” and “⋅”, and the order “<”, as follows:
pq ′ + p′ q
[(p, q)] + [(p′ , q ′ )] = [ ]
qq ′
[(p, q)] ⋅ [(p′ , q ′ )] = [(pp′ , qq ′ )]
[(p, q)] < [(p′ , q ′ )] if pq ′ < qp′ .
Moreover, if p ≠ 0, we can also define the inverse:
⎧
⎪(q, p)
⎪ p>0
[(p, q)] −1
=⎨
⎪
⎩(−q, −p) p < 0.
⎪

44
This satisfies:
⎧
⎪[(p, q) ⋅ (q, p)] = [(pq, pq)] = [(1, 1)]
⎪ p>0
[(p, q)] ⋅ [(p, q)]−1 = ⎨
⎪
⎩[(p, q) ⋅ (−q, −p)] = [(−pq, −pq)] = [(1, 1)] p < 0
⎪

Note: Strictly speaking, we need to check that these operations are “well-defined”, i.e. if

[(p, q)] = [(p′ , q ′ )],

then
[(p, q)] + [(p′′ , q ′′ )] = [(p′ , q ′ )] + [(p′′ , q ′′ )].
i.e, that “f+ ” is a function on Q. I encourage you to check this as an exercise.

Theorem 4.3.3. The rational numbers, Q, also satisfy the axioms A1-A4, M1-M5, O1-O4.

Proof. We actually checked A1 and A2 already last class, and you have M4 and O3 on your
homework. The rest can be checked in a similar way.

4.4 Properties of Q (Mar 16)

From last lecture, we defined the set of rational numbers, Q. We know that this number
system satisfies the same axioms as Z.

Proposition 4.4.1. The map

Z Ð→ Q
n z→ [(n, 1)]
is injective.

Proof. We have to check that if m ≠ n, then [(n, 1)] ≠ [(m, 1)], i.e. (m, 1) ∼ (n, 1). But if
m ≠ n, then m ⋅ 1 ≠ n ⋅ 1, so by definition of ∼,

(n, 1) ≁ (m, 1).

Therefore the above map is injective.

We will now switch to the following “standard” notation for rational numbers:

p
= [(p, q)]
q
p
= [(2p, 2q)]
q
p
= [(100p, 100q)]
q

45
 p p q
= [( , )] if p and q are both even.
q 2 2
n
n = = [(n, 1)]
1
0
0 = = [(0, 1)]
1
1
1 = = [(1, 1)].
1
Theorem 4.4.2. Let x, y ∈ Q with x < y. There exist infinitely many rational numbers, z ∈ Q,
with
x < z < y.

Proof. Take the average of the x and y values:

x+y
z= ∈Q
2
Then, since x < y, we can write:

x+x<x+y <y+y
2x < x + y < 2y
x+y
x< <y
2
x < z < y.
If we continue this method for more z-values, we can choose:

x+z
z′ =
2
Thus, we know that
x < z ′ < z < y.
Continuing, we write
x + z′
z ′′ =
2
⇒ x < z < z′ < z < y
′′

⇒ x < ... < z ′′ < z ′ < z < ... < y.

Question: How big is the set Q? Is it countably infinite or uncountably infinite?

Proposition 4.4.3. Q is countably infinite.

46
Proof. Q is certainly infinite, since there is an injective map from Z to Q (by the last
proposition).
Note that Z × N is countable, and there exists a surjective map:

Z×N↠Q

(p, q) z→ [(p, q)]

Therefore, Q is countable.

Problem: Q is full of holes!

Theorem 4.4.4. There does not exist x ∈ Q such that x2 = 2.

Proof. Suppose that x = pq satisfies the equation x2 = 2. We may assume, without loss of
generality, that p and q are not both even. (This can be achieved this by cancelling their
common factors of 2). We can write:

x2 = 2 ⇐⇒ p2 = 2q 2
Ô⇒ p is even
Ô⇒ p = 2r for some r ∈ Z.

But then

p2 = (2r)2
= 4r2 , and so
4r2 = 2q 2 .

Cancelling a factor of 2, we obtain:

2r2 = q 2
Ô⇒ q 2 is also even
Ô⇒ q is even.

But, this contradicts our assumption that p and q are not both even.

r r rr r r rrr r r
0 1 2 3
The blue set of numbers satisfy r2 < 2. The green set of numbers satisfy r2 > 2. Between
√
these two sets, there should exist a number called “ 2.” But we do not yet know that such
a number exists!
√
We need to define a number system in which 2 exists. This number system will be called
the real numbers, R. More generally, we want R to have the following special property.

47
Definition 4.4.5. (i) Given a set E, we say that M is an upper bound for E if for all
x ∈ E, x ≤ M.
(ii) We say that E is bounded above if an upper bound exists.
(iii) We say that M is the least upper bound for E if M is itself an upper bound for
E and for any given upper bound, M , we have M ≤ M.
(iv) We say that a number system (for instance, R) has the least upper bound prop-
erty if, for any subset E that is bounded above, there exists a least upper bound M for E.
In this case,
M = sup E
is called the supremum of E.
Example 4.4.6. Let E = {x ∈ Q ∣ x2 < 2}. Is this set bounded above? Yes. Let M = 2.

If x2 < 2, then x < 2

Ô⇒ M = 2 is an upper bound for E.

BUT there does not exist a least upper bound for E within the natural numbers, Q. This
is because, given any x ∈ Q with x2 < 2, there exists y ∈ Q with y > x and x2 < y 2 < 2...see
homework.
Ô⇒ Q does not have the least upper bound property!
Idea: Instead of just rational numbers, we will consider entire subsets of Q and make a
number system out of those!

4.5 The real numbers (Mar 18)

Goal: Construct a number system, R, that has the least upper bound property.
Recall that given any subset E ⊂ R, if E is bounded above, then there exists a least
upper bound (supremum) for E, in R.
Then for any upper bound M , M ≤ M .
Note that this is not true for the set Q. We can consider the set E = {r ∈ Q ∣ r2 < 2} .
Definition 4.5.1. A subset α ⊂ Q is said to be a Dedekind Cut if it satisfies the following
three properties:
1. α is a proper subset of Q (i.e. α ≠ 0, Q)
2. If r ∈ α and s < r, where s ∈ Q, then s ∈ α
3. If s ∈ α , ∃r ∈ α with s < r.
Example 4.5.2. Let 0∗ = {r ∈ Q ∣ r < 0} Is this a Dedekind cut?
1. −1 ∈ 0∗ , 1 ≠ 0∗ , therefore 0∗ ≠ 0, 0∗ = Q✓
2. Let r ∈ 0∗ . Then r < 0 . Given s < r , s < 0

Ô⇒ s ∈ 0∗ ✓

48
3. Let s ∈ 0∗ . Then s < 0. Let r = 2s . Then,
s
s< <0
2
Ô⇒ s < r and r ∈ 0∗ .✓
Therefore, 0∗ is a Dedekind cut.
Example 4.5.3. Given any rational number t ∈ Q, we define

t∗ = {r ∈ Q ∣ r < t}.

This is called a ”rational” Dedekind cut. Then we must check if this set follows the param-
eters of a Dedekind cut:
1. t − 1 ∈ t∗ and t + 1 ≠ t∗
Ô⇒ t ≠ 0, Q
2. Given r ∈ t∗ , we have r < t. So if s < r, then s < r < t

Ô⇒ s ∈ t∗ ✓

3. Given s ∈ t∗ , we take
s+t
r= .
2
Since we have s < t, we write
s<r<t
Ô⇒ r ∈ t∗ and r > s.✓
So this is in fact a Dedekind cut!
Example 4.5.4. Let α = {r ∈ Q ∣ r2 < 2} ∪ 0∗ This is a Dedekind cut (this will be proved
√
on HW). But, it does not equal t∗ for any rational number t! This is true since 2 is not
rational. This would be called an irrational Dedekind cut.
Example 4.5.5. The set {r ∈ Q ∣ r ≤ 1} is not a Dedekind cut. Why is this the case?
Property (III) fails for s = 1.
Definition 4.5.6. Given two cuts α and β, we define the ”sum”

α + β = {r + s ∣ r ∈ α, s ∈ β}

Is this a Dedekind cut?

1. If α ≠ 0 and β ≠ 0, then α + β ≠ 0. Given r ∉ α and s ∉ β, we must have

r+s≠α+β

49
If r ≠ α, then ∀ r′ ∈ α, we have r′ < r. This is true since if r ≤ r′ , then r ∈ α by property (2.).
Also, if s ∉ β, then ∀ s′ ∈ β, we have s′ < s. So, ∀ r′ ∈ α and s′ ∈ β, we have

r′ + s” < r + s

Ô⇒ r + s ∉ α + β
Ô⇒ α + β ≠ Q✓
2. Given r ∈ α + β, we have r = t + t′ , for t ∈ α, t′ ∈ β. Given that s < r, let t′′ = t′ + s − r < t′

Ô⇒ t′′ ∈ β

But then, we see

t + t′′ = t + t′ + s − r
=r+s−r
=s
Ô⇒ s ∈ α + β✓
3. Given that s ∈ α + β, we have s = t + t′ and t ∈ α, t′ ∈ β. By property (2.), for β, ∃t′′ ∈ β
such that t′′ > t′ . Then,
t + t′′ ∈ α + β
and
t + t′′ > t + t′ > s
Therefore
r = t + t′′ > s
and
r ∈ α + β✓
Therefore, we conclude that α + β is a Dedekind cut!
Next, we define the order “<” on two Dedekind cuts by:

α < β ⇐⇒ α ⊊ β.

This means that α is a proper subset of β. We can also write that

α > β ⇐⇒ α ⊋ β
α = β ⇐⇒ α and β are equal and subsets of Q.

50
Definition 4.5.7. The real number system, R, is defined to be the set of all Dedekind cuts,
with the operation “+” and the order “<” defined above.
We have the following inclusions:

Z⊂Q⊂R

n ∈ Z ↦ [(n, 1)] ∈ Q
r ∈ Q ↦ r∗ ∈ R

Theorem 4.5.8. R satisfies all of the axioms A1-A4, M1-M5, O1-O4.

Proof. (A1)
α + β = {r + s ∣ r ∈ α, s ∈ β}
= {s + r ∣ s ∈ β, r ∈ α}
= β + α✓
(A2) Same
(A3) We need to show:
α + 0∗ = α
(⊂) Given that r ∈ α + 0∗ , we have that r = t + t′ for t ∈ α, t′ ∈ 0∗

Ô⇒ t′ < 0

Ô⇒ r < t
Ô⇒ r ∈ α (by (2.))✓
(⊃) Given that r ∈ α, by (3.), ∃E ∈ α with r < t. Then, r − t < 0

Ô⇒ r − t ∈ 0∗ .

But then, we have,

r = t + r − t ∈ α + 0∗ .
Ô⇒ r ∈ α + 0∗ .✓
(A4) Define the following:
−α = {s ∈ Q ∣ ∃t > 0, −s − t ∉ α}
Claim: α + (−α) = 0∗
(⊂) Given that r ∈ α and s ∈ −α, ∃t > 0 such that −s − t ∉ α

−s − t > r

Ô⇒ 0 > r + s + t

51
 Ô⇒ 0 > r + s
Ô⇒ r + s ∈ 0∗ ✓
(⊃) Given q > 0, we must show that

−q ∈ α + (−α)

Let t = 2q . We choose n ∈ N such that

n ⋅ t ∈ α but,

(n + 1) ⋅ t ∉ α
Next, let r = −(n + 2)t
Ô⇒ −r − t = (n + 2)t − t
= (n + 1)t ∉ α
Ô⇒ r ∈ −α.
But then, we see that
r + (n ⋅ t) = −(n + 2)t + nt
= −2t = −q
Ô⇒ Since r ∈ −α and nt ∈ α, we conclude that − q ∈ α + −(α)✓
The rest of the axioms can be proved similarly.

Note that
Q⊂R
r z→ r∗
(r + s)∗ = r∗ + s∗
r < s Ô⇒ r∗ < s∗ .
The real numbers fill the “holes” in the number system Q! In fact, we have:

Theorem 4.5.9. R satisfies the least upper bound property.

Proof. Let E ⊂ R be a nonempty subset that is bounded above (i.e. ∃β ∈ R, ∀γ ∈ E, γ < β).
We need to construct the least upper bound of E. We write

α = ⋃ γ ⊂ Q.
γ∈E

52
 We claim that α is a Dedekind cut, and is the least upper bound of E.
(1.) α ≠ 0 since γ ∈ E is a Dedekind cut, and then it is nonempty as a subset of Q. Since
γ ⊂ β, ∀γ ∈ E, we see that α ⊂ β ⊊ Q

Ô⇒ α ⊊ Q✓

(2.) If r ∈ α, then r ∈ γ for some γ ∈ E. Given that s < r, s ∈ r

Ô⇒ s ∈ α✓

(3.) Same as (2.) ✓ Therefore α is the least upper bound!

- α is an upper bound for E, because given any γ ∈ E, we have γ ⊂ ⋃ γ = α

Ô⇒ γ ⊂ α

Ô⇒ γ < α✓

-Let α′ be any other upper bound for E. We need to show that α ≤ α′ , meaning that α ⊂ α′
Let r ∈ α. Then, r ∈ γ for some γ ∈ E. Since α′ is an upper bound for E, γ ⊂ α!

Ô⇒ r ∈ α′

Since r ∈ α was arbitrary, we know that α ⊂ α′ .

4.6 Properties of R (Mar 23)

We proved above that R satisfies the least-upper-bound property. This will be crucial to
developing limits, calculus, etc. There are two more properties of R that are easy to prove
from the above construction.

Theorem 4.6.1. (i) Given any α > 0∗ in R, and β ∈ R, there exists n ∈ N such that

n⋅α>β

This is called the Archimedean property of R.

(ii) Given any α, β ∈ R such that α < β, there exists some r ∈ Q such that

α < r∗ < β

This is sometimes referred to as the density of Q in R.

Proof. (i) Pick a rational number s ∈ α with s > 0. Then

p
s= for p, q ∈ N.
q

53
 ′
Pick r = pq′ ∉ β, and let n = q ⋅ p′ ∈ N.
Claim: n ⋅ s ∉ β
For,
p p′
n ⋅ s = (q ⋅ p′ ) ⋅ = pp′ > ′ = r.
q q
Since r ∉ β and n ⋅ s > r, n ⋅ s ∉ β (by contrapositive of (II)) but n ⋅ s ∈ n ⋅ α.

Ô⇒ n ⋅ α ⊃ β

Ô⇒ n ⋅ α > β
as real numbers.
(ii) If α < β, then α ⊊ β by definition. So ∃r ∈ β, r ∉ α, so α ⊂ r∗ (by contrapositive of
(II)). Also r∗ ⊊ β (because r ∈ β, but r ∉ r∗ )).
So α ⊊ r∗ ⊊ β i.e. α < r∗ < β as real numbers.

√
Also, for any α > 0, we can define n α = {t ∈ Q ∣ t > 0, tn ∈ α} ∪ 0∗ .
Question: How big is R? Is it countably infinite or uncountably infinite?

Recall that {0, 1, 2, ..., 9}N is the set of all sequences with digits 0 through 9. Given
a = (a1 , a2 , a3 , ...), we want to define

“.a1 a2 a3 a4 ...,′′

i.e., the real number with decimal expansion a.

Given n ∈ N, we first define the rational number
a1 a2 a3 an
sn = + 2 + 3 + ... + n
10 10 10 10
“ = .a1 a2 ...a′′n .
To get the whole string as a real number, define:

f (a) = {t ∈ Q ∣ t < sn for somen ∈ N}.

It’s not hard to see that this is a Dedekind cut!

Moreover, the only way for two sequences a and a′ to have f (a) = f (a′ ) is if

a = .a1 a2 ...an 1000...

and
a′ = .a1 a2 ...an 0999,
or vice versa.

54
Theorem 4.6.2. The set R is uncountable.

Proof. Assume, for contradiction, that R is countable. Above, we defined a map

f ∶ {0, 1, ..., 9}N → R

a ↦ .a1 a2 a3 ...
Given y ∈ R, there exist at most two sequences, a and a′ , with f (a) = f (a′ ). Therefore

∣ f −1 ({y}) ∣≤ 2 ∀y ∈ R.

In particular, f −1 ({y}) is countable. But, by definition, we have {0, ..., 9}N = ⋃ f −1 ({y}).
y∈R
So if R were countable, this would be a countable union of countable sets Ô⇒ countable.
But this contradicts Cantor’s theorem that {0, 1, ..., 9}N is uncountable.

4.7 Supremum and infimum (Mar 23)

We will now stop using Dedekind cut notation. We’ll write 0 instead of 0∗ , α < β instead of
α ⊊ β, etc.
Definition 4.7.1. [0, 1] = {x ∈ R ∣ 0 ≤ x ≤ 1} and (0, 1) = {x ∈ R ∣ 0 < x < 1}.
Recall that sup E = least upper bound of a set E that is bounded above.
If E is not bounded above, we write sup E = ∞. This is just a notation to say that E
is not bounded above.
Definition 4.7.2. Given a subset E ⊂ R that is bounded below, we write
inf E = ”infimum of E” = greatest lower bound of E.
(In other words, this is a number N ∈ R such that N ≤ x ∀x ∈ E and, given N ∈ R with
N ≤ x ∀x ∈ E, N ≤ N = inf E.)
If E is not bounded below, then we write inf E = −∞.
Example 4.7.3.
sup(0, 1) = 1 = sup[0, 1]
inf(0, 1) = 0 = inf [0, 1].
Example 4.7.4. sup N = ∞ (because N is not bounded above).
Example 4.7.5. Fix x > 0. Let E = {n ⋅ x ∣ n ∈ N}. Then sup E = ∞.

Proof. Let m ∈ R. By the Archimedean property, ∃n ∈ N such that n ⋅ x > M. Therefore, M

is not an upper bound for E. Since M ∈ R was arbitrary, E is not bounded above.

Example 4.7.6. sup Z = ∞, inf Z = −∞ (because Z is bounded neither above nor below.)

55
Theorem 4.7.7. For any set E ⊂ R such that E is bounded below, the infimum inf E ∈ R
exists (∞ ∉ R).

Proof. Let −E = {x ∈ R ∣ −x ∈ E} then −E is bounded above. Let N = − sup(−E). One can

check that this is the infimum of E!

56
5 Sequences and limits
5.1 The definition of a limit (Mar 25)
Definition 5.1.1. A sequence is an infinite string of real numbers a = (a1 , a2 , a3 , ...). In
our earlier notation, this is just an element of RN = {functions from N to R}, where we write
f (n) = an .
Example 5.1.2. 1, 21 , 13 , 41 , 51 , ...
an = n1
Example 5.1.3. 1, 2, 3, 4, 5, 6, ...
an = n ∀n ∈ N
Example 5.1.4. 1, 0, 1, 0, 1, 0, 1, 0, ...
⎧
⎪1 n odd
⎪
an = ⎨
⎪0 n even
⎪
⎩
Example 5.1.5. 1, 1.4, 1.41, 1.414, 1.4142, 1.41425, ...
√
This is the decimal expansion of 2.
Example 5.1.6. 1, 0, 1, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, ...
Example 5.1.7. 3, 1, 4, 1, 5, 9, 2, 6, ...
This is the sequence of digits in the decimal expansion of π. (Essentially just a random
sequence of integers in {0, ..., 9}.)
Let L be a real number. Informally, a sequence a converges to L if an ”becomes arbitrarily
close to L as n becomes arbitrarily large.” In the above examples:
Example 5.1.2: 1, 21 , 13 , 41 , ... This converges to L = 0.
Example 5.1.3: 1, 2, 3, 4, 5, ... This diverges to infinity.
Example 5.1.4: 1, 0, 1, 0, 1, 0, ... This diverges (all of the points have to get close to the
same L).
√
Example 5.1.5: 1, 1.4, 1.41, 1.414, ... This converges to 2.
Example 5.1.6: 1, 0, 1, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, ... Diverges.
We now make the formal definition of a limit, as follows.
Definition 5.1.8. We write an → L, or

lim an = L
n→∞

if the following statement is true: for each ε > 0, ∃N ∈ N such that

∣an − L∣ < 

for all n > N.

If no such L ∈ R exists, we say that a diverges.

57
 Note: Order of quantifiers is essential.
1
Proposition 5.1.9. The sequence an = n converges to L = 0, i.e.

1
lim = 0.
n→∞ n

Proof. Let ε > 0. We need to find N such that ∀n ≥ N , ∣an − 0∣ < ε.

We have:
1
∣an − 0∣ = ∣an ∣ =
n
1
and < ε ⇔ 1 < n ⋅ ε.
n
By the Archimedean property of R (Theorem 4.6.1i), ∃N ∈ N such that 1 < N ⋅ . Then for
any n > N , we have
1
1 < N ⋅ ε < n ⋅ ε Ô⇒ 1 < n ⋅ ε Ô⇒ < ε,
n
which is the desired inequality. Therefore, N has the required property.
Since ε > 0 was arbitrary, we’re done.

Note: By a similar proof, one can show

1
lim = 0 ∀p > 0.
n→∞ np

n
Proposition 5.1.10. The sequence an = 1 + (−1)
n converges to L = 1.
3 2 5 4 7 6
0, 2 , 3 , 4 , 5 , 6 , 7 , . . .

Proof. Let ε > 0. We have

(−1)n 1
∣an − 1∣ = ∣ ∣= .
n n
1
By the proof of the previous proposition, we know ∃N > 0 such that n < ε ∀n > N

1
Ô⇒ ∣an − 1∣ = < ε,
n
which is the desired statement. Since ε > 0 was arbitrary, we’re done.

Proposition 5.1.11. The sequence

⎧
⎪0 n even
⎪
an = ⎨
⎪
⎩1 n odd
⎪

is divergent.

58
Proof. To prove that a diverges, we need to show that no possible N ∈ R can satisfy the
definition, let L ∈ R. We must show that an ↛ L, i.e. in ∃ε > 0 such that ∀N ∈ R, ∃n > N
with ∣ an − L ∣≥ ε. In this case, we let ε = 12 .
Case 1: L ≥ 12
For n even, we have an = 0 Ô⇒ ∣ an − L ∣ = ∣ 0 − L ∣ = ∣ L ∣≥ 12 = ε. So ∣ an − L ∣≮ ε
∀n even. Since, for any N, ∃n > N that are even, this shows that an ↛ L
Case 2: L < 12
For n odd, we have an = 1 Ô⇒ ∣ an − L ∣ = ∣ 1 − L ∣> 21 .
Since L < 12 Ô⇒ −L > 12 Ô⇒ 1 − L > 12 Ô⇒ ∣ an − L ∣≮ 12 = ε. Since ∀N, ∃n > N with n
odd, this shows that an ↛ L
This covers all possible L, so we’re done.

The following simple lemma is very useful for doing limits.

Lemma 5.1.12 (Triangle inequality). For x, y, and z ∈ R we have
∣x − z∣ ≤ ∣x − y∣ + ∣y − z∣.
Proof. Let a = x − y, b = y − z
a + b = x − y + y − z = x − z so the above inequality is equivalent to:
∣a + b∣ ≤ ∣a∣ + ∣b∣
.
Case 1: a, b ≥ 0. Then
∣a + b∣ = a + b = ∣a∣ + ∣b∣
so we’re done.
Case 2: a, b ≤ 0. Then
∣a + b∣ = −(a + b) = −a − b = ∣a∣ + ∣b∣
so we’re done.
Case 3: a < 0, b > 0 and 0 < −a ≤ b. This implies that a + b > 0. So we have:
∣a + b∣ = a + b = −∣a∣ + ∣b∣ < ∣a∣ + ∣b∣
and we’re done.
There are a few more similar cases, which we omit.
Theorem 5.1.13. The limit of any convergent sequence is unique.
Proof. Suppose that an → L, but also an → L′ .
We need to show that L = L′ .
Let ε > 0. We know that there exists N such that if n > N then ∣ an − L ∣< 2ε , and ∃N ′
such that ∀n > N ′ , ∣ an − L′ ∣< 2ε . But then by the triangle inequality,
∣L − L′ ∣ ≤ ∣L − an ∣ + ∣an − L′ ∣ < 2ε + 2ε = ε, for n > max (N, N’) Ô⇒ ∣L − L′ ∣ < ε
Since ε > 0 was arbitrary, this implies that L = L′

59
Definition 5.1.14. We say that a sequence is bounded above (and respectively below) if
the set {an ∣ n ∈ N} is bounded above.
Note: A sequence can be bounded above (and below) and still diverge.

Proposition 5.1.15. Any convergent sequence is bounded above (and below).

Note: In other words, boundedness is a necessary, but not sufficient condition for con-
vergence.

Proof. Let L ∈ R such that an → L. Let  = 1. Since an → L, there exists N such that ∀n > N ,
∣an − L∣ < 1. Since an − L + L = an , ∣an ∣ ≤ ∣an − L∣ + ∣L∣ by the triangle inequality for
n > N Ô⇒ ∣an ∣ < 1 + ∣L∣ Ô⇒ {∣an ∣ ∣ n > N } is bounded above.
But the set {∣an ∣ ∣ n ≤ N } is also bounded above, since this is a finite set.
So {∣an ∣ ∣ n ∈ N} = {∣an ∣ ∣ n ≤ N } ∪ {∣an ∣ ∣ n > N } is also bounded above.
If {∣an ∣ ∣ n ∈ N} is bounded above, then {an ∣ n ∈ N} is bounded.

5.2 Limit rules (Mar 30)

Theorem 5.2.1. Suppose that a and b are convergent sequences with an → a and bn → b.
Then
1.an + bn → a + b
2.can → ca
3.an ∗ bn → a ∗ b
4.If bn ≠ 0∀n, and b ≠ 0, then an /bn → a/b.

Proof of 1. Given ε > 0, ∃N1 , N2 such that

n ≥ N1 ⇒ ∣an − a∣ < ε/2

and
n ≥ N2 ⇒ ∣bn − b∣ < ε/2.
Let N = max(N1 , N2 ).
Then for n ≥ N , we have

∣an + bn − (a + b)∣ = ∣(an − a) + (bn − b)∣.

By the Triangle Theorem,

≤ ∣an − a∣ + ∣bn + b∣ < ε/2 + ε/2 = ε

by construction. Since ε > 0 was arbitrary, we are done.

60
Proof of 2. Given ε > 0, choose N such that, ∣an − a∣ < ε/c, for n > N ,

∣c ∗ an − c ∗ a∣
= ∣c ∗ (an − a)∣
= ∣c∣∣an − a∣ < ∣c∣ ∗ ε/∣c∣
= ε.

Since ε > 0 and was arbitrary, we’re done.

Proof of 3.
an ∗ b n → a ∗ b
We first do the special case that a = 0. We need to show that

an ∗ b n → 0 ∗ b = 0

⇒
an ∗ bn → 0.
Let ε > 0. By last class, since bn is convergent, it is bounded by some number B > 0.

∣bn ∣ < B, ∀n ∈ N
Since an → 0 by assumption, Choose N such that

∣an − 0∣ = ∣an ∣ < ε/B

∀n > N . But then,

∣an bn − 0∣ = ∣an bn ∣ = ∣an ∣∣bn ∣
ε
< B = ε.
B
Since ε > 0 was arbitrary, we are done with this case.
So, if an → 0 and bn → b, then an bn → 0.
For the general case, consider the sequence

(an − a)bn .

Note that an − a → a − a = 0 by the first property. By the previous case, we have

(an − a) ∗ b → 0.

By the second property, we also have

abn → ab

61
(because a is just a constant). Therefore, the sequence

an bn = an bn − abn + abn = (an − a)bn + abn .

By the first property,

an bn → 0 + ab = ab
as desired.
The proof for number 4 is skipped.

Example 5.2.2. an = 1 + ((−1)n )/n

limn→∞ 1 = 1 and
limn→∞ (−1n )/n = 0
So, limn→∞ an = 1 + 0 = 1.
√ √
Example 5.2.3. limn→∞ (100/n2 ) ∗ (5 + 10/ n + 8/ 3 n)
The limits all exist, so the limit of it is

0 ∗ ∣5 + 0 + 0∣ = 0.

Example 5.2.4.

1 + (−1)n 1
an = = (1 + (−1)n ) .
n n
We can show directly that limn→∞ an = 0 or use the Squeeze Theorem:
if 0 < an < n2 as 0 → 0
as n2 → 0
an → 0.

5.3 Monotone sequences (Mar 30)

Recall that being bounded is a necessary but not sufficient condition for a general sequence
to converge. However, there IS an important exception.
Definition 5.3.1. A sequence a is monotonically increasing if an ≤ an + 1∀n.
A sequence a is monotonically decreasing if an ≥ an + 1∀n.
A sequence a is monotone if it is either monotonically increasing or decreasing.

Theorem 5.3.2. Suppose that a sequence a is monotone and bounded, then a is convergent.

Proof. Suppose a is monotonically increasing and bounded above. Let E = an ∣n ∈ N . Since

an is bounded, E is bounded. We may therefore let a = sup E.
We now claim that
an → a.

62
By definition, we have
an ≤ a∀n
since a is an upper bound for E. Let ε > 0. Since a is the least upper bound, there must exist
N such that
an > a − ε.
But since a is monotone, we have
aN ≤ an
∀n ≥ N.
a − ε ≤ aN ≤ an ≤ a
⇒ a − ε < an ≤ a
⇒ a − an < ε,
an − a ≤ 0 < ε
∣a − an ∣ < ε
Since ε > 0 was arbitrary, we are done.

Example 5.3.3. Consider the sequence

1
an = (1 + )n .
n
We have

1
an = (1 + )n
n
n+1 n
=( )
n
n+2 n
an + 1 = ( ) +1
n+1
So,

an + 1 (n + 2)n + 1 (n + 1)n
= ∗
an (n + 1)n + 1 nn
(n + 2)n + 1
=
(n + 1) ∗ nn
n+1 n n+2
=( ) ×
n n+1
n+2 n+2 n
>1⇒( ) >1
n n

63
 n+2
>1
n+1
an+1
⇒ >1
an
Therefore, the sequence an is monotonically increasing. We will see later that an is bounded.
Therefore, the previous Theorem implies that

lim an
n→∞

exists!
This limit is just the number “e” (by definition).

5.4 lim sup and lim inf (Apr 1)

In the last class we proved the following theorem.
Theorem 5.4.1. A bounded, monotone sequence is convergent.
Proof. Let L = sup{an ∣ n ∈ N}.
Then for each ε > 0, ∃aN such that

L − ε < aN ≤ L.
(If not, L − ε is an upper bound; but then L is not the least upper bound!)
Since an is increasing, if n > N, then

L − ε < aN ≤ an ≤ L

⇒ ∣an − L∣ < ε.

What about a sequence that is bounded, but not necessarily monotone? We want to
compare such a sequence with a monotone sequence (in fact, two of them.)
Let
En = {an ∣ m ≥ n}.
Define
An = sup En = sup am .
m≥n

Lemma 5.4.2. An is a decreasing sequence.

Proof. Since En+1 ⊂ En , the number An = sup En is an upper bound for En+1 . Since An+1 is
the least upper bound of En+1 (by definition), we have

An ≥ An+1

as desired.

64
 Similarly, one has,
Bn ≤ Bn + 1
⇒ Bn is an increasing sequence.
By the previous theorem, both sequences An , Bn are convergent! We can therefore make
the following definition:
Definition 5.4.3.
lim sup an = lim An
n→∞ n→∞

lim inf an = lim Bn .

n→∞ n→∞
Example 5.4.4. a = (1, 0, 1, 0, 1, 0, ...)

lim sup an = 1
n→∞

lim inf an = 0.
n→∞
Example 5.4.5. an = (−1)n ∗ (1 + 1/n)

lim sup an = 1
n→∞

lim inf an = −1.

n→∞
Example 5.4.6. a = (1, 0, 2, 0, 3, 0, 4...)

An = sup am = ∞∀n,
m≥n

An = (∞, ∞, ...)
lim An = ∞
n→∞

lim sup an = ∞.
n→∞

On the other hand, we have,

lim inf an = 0.
n→∞

Example 5.4.7. (0, -1, 0, -2, 0, -3...)

lim sup an = 0
n→∞

lim inf an = −∞.

n→∞

(−1)n
Example 5.4.8. Let an = n . Then

lim sup an = 0 = lim inf an .

n→∞ n→∞

In fact, this is just a convergent sequence, with limn→∞ an = 0.

65
 More generally, we have the following:

Theorem 5.4.9. Let a be a bounded sequence. Then a is convergent if and only if

lim sup an = lim inf an .

n→∞ n→∞

In this case, we have

lim an = lim sup an = lim inf an .
n→∞ n→∞ n→∞

Proof. We shall write

L̄ = lim sup an
n→∞

L = lim inf an .
n→∞

(⇒) Suppose that a is convergent, with limn→∞ an = L. Then, given ε > 0, ∃ N such that

∣an − L∣ < ε
(∀n > N )
⇒ an − L < ε
⇒ an < L + ε
(∀n > N )
⇒ sup an ≤ L + ε
n≥N +1

⇒ lim sup an ≤ L + ε
N →∞ n≥N +1

L̄ ≤ L + ε.

But also,

L − an < ε
⇒ L − ε < an
(∀n > N )
⇒ L − ε < inf an
n≥N +1

⇒ L − ε ≤ lim inf an
N →∞ n≥N +1

⇒ L − ε ≤ L.

Therefore,
L̄ ≤ L + ε = (L − ε) + 2ε
L̄ ≤ L + 2ε.

66
By homework, we always have
L ≤ L̄
⇒ L ≤ L̄ ≤ L + 2ε
⇒ ∣L̄ − L∣ ≤ 2ε.
Since ε > 0 is arbitrary, we conclude that L = L̄
(⇐) Suppose that
L = L̄.
Let L = L = L̄. We will show that
lim an = L.
n→∞
Let ε > 0. By definition of the limsup, ∃N1 such that
sup am < L̄ + ε.
m≥N1

By definition of the liminf, there exists N2 such that

L − ε < inf am .
m≥N2

For n > N = max(N1 , N2 ), we have

inf am ≤ an ≤ sup am .
m≥N m≥N

Combining with the above, we have

L − ε < inf ≤ an ≤ sup an .
m≥n m≥n

But
L = L̄ = L
L − ε < an < L + ε ∀n > N
⇒ ∣an − L∣ < ε.
Since ε > 0 was arbitrary, we are done.

5.5 Subsequences and the Bolzano-Weierstrass Theorem (Apr 8)

The following notion can be very helpful for dealing with sequences.
Definition 5.5.1. Given a sequence an and an increasing sequence of positive integers
n1 < n2 < n3 < ⋯,
the sequence ani , i.e.
an1 , an2 , an3 , ⋯
is called a subsequence of an .

67
 ⎧
⎪0 n even
⎪
Example 5.5.2. an = ⎨ . This has two obvious subsequences:
⎪
⎪ 1 n odd
⎩
(a1 , a3 , a5 , . . .) = (1, 1, 1, . . .)

and
(a2 , a4 , a6 , . . .) = (0, 0, 0, . . .).
Note that while the original sequence is divergent, each of these subsequences is conver-
gent. In this way, passing to a subsequence can help you “fix” the divergence of a sequence.
In fact, for bounded sequences, this is always possible.

Theorem 5.5.3 (Bolzano-Weierstrass). Any bounded sequence has a convergent subsequence.

This is a very well-known theorem, and there are many different ways to prove it. Fortu-
nately, with the technology we have already developed, we can give a very fast proof. In fact,
we shall prove the following more precise theorem, which clearly implies Bolzano-Weierstrass
as a corollary.

Theorem 5.5.4. Given any bounded sequence an , there exists a subsequence ani such that

lim ani = lim sup an .

i→∞ n→∞

There also exists a subsequence amj such that

lim anj = lim inf an .

j→∞ n→∞

Proof. Let L = lim supn→∞ an . We know that the sequence

An = sup an
m≥n

converges to L.
Let ε = 2i1 . Since An → L, there exists N such that

1
∣AN − L∣ < .
2i
Since AN = supm≥N am , there must exist m ≥ N such that
1
∣am − AN ∣ < .
2i
Let ni = m. By the triangle inequality, we have
1 1
∣ani − L∣ < ∣ani − AN ∣ + ∣AN − L∣ < +
2i 2i
68
and
1
∣ani − L∣ < .
i
1
But i → 0 as i → ∞. Therefore (by HW), the sequence ani satisfies

lim ani → L.
i→∞

So, ani is the desired subsequence.

The subsequence amj converging to lim inf n→∞ an is obtained in a similar way.

Note that the Bolzano-Weierstrass Theorem follows immediately from this theorem (since
the lim sup and lim inf of any bounded sequence exists as a real number). We can also make
the following corollary:

Corollary 5.5.5. A bounded sequence an is divergent if and only if there exist two (or more)
convergent subsequences, ani and amj , such that

lim ani ≠ lim amj .

i→∞ j→∞

Proof. This follows from Theorem 5.4.9 and the previous theorem.

Note: This corollary applies only to bounded sequences. If a sequence is unbounded, it may
not have any convergent subsequences (see HW).
Example 5.5.6. Consider the sequence of integers a = (3, 1, 4, 1, 5, 9, 2, 6, 2, 7, . . .). This is
the sequence of digits of π, so an lies in the finite set [0, 9] ∩ Z for each n. In particular, the
sequence is bounded, so we know from the Bolzano-Weierstrass Theorem that it must have
a convergent subsequence. In this case, however, this is obvious. Why? Because there are
only 10 digits, at least one of them (say d) must occur infinitely many times. Then

d, d, d, d, d, . . .

is a subsequence of an , which (clearly) converges to d.

Example 5.5.7. Consider the following sequence of real numbers:

a1 = π = 3.141592 . . .

a2 = 1.415926 . . .
a3 = 4.1519262 . . .
a4 = 1.5192627 . . .
⋮.

69
This is a sequence of real number between 0 and 10. By the Bolzano-Weierstrass Theorem,
it must have a convergent subsequence. We discussed a somewhat more explicit way to
construct such a subsequence, by “freezing” one digit at a time:

d1 . ∗ ∗ ∗ ∗∗

d1 .d2 ∗ ∗ ∗ ∗∗
d1 .d2 d3 ∗ ∗ ∗ ∗
d1 .d2 d3 d4 ∗ ∗ ∗ .
This is Frank Morgan’s (very informal) proof of Bolzano-Weierstrass, on p. 38 of the text.

5.6 Accumulation points (Apr 13)

Let E ⊂ R be a subset and p ∈ R.
Definition 5.6.1. We say p is an accumulation point of E, if there exists a sequence an ∈ E
with an → p, an ≠ p, ∀n. We denote by E ′ the set of all accumulation points of E.
Example 5.6.2. E = Q, E ′ = R!

Formal proof. Let x ∈ R, we need to construct a sequence of rationals: an → x (with an ≠ x).

By density of Q in R, for any ε > 0, ∃ r ∈ Q, such that

x − ε < r < x.

Choose ε = n1 . There exists r = an , such that

1
x− < an < x.
n
Do this for each n ∈ N, we obtain a sequence an with
1
∣an − x∣ < ,
n
⇒ an → x,
(also, an < x ⇒ an ≠ x, ∀n)

Informal proof. Take the decimal expansion of x:

x = d1 d2 .d3 d4 d5 d6 . . .

Let

a1 = d1 0.0000 . . .

70
 a2 = d1 d2 .0 . . .
a3 = d1 d2 .d3 0 . . .
a4 = d1 d2 .d3 d4 00 . . .

d3 d4 dn
⇒ an = 10.d1 + d2 + + + ⋯ + n−1
10 100 10
10n d1 + 10n−1 d2 + ⋯ + 10dn−1 + dn
=
10n−1
⇒ an ∈ Q for each n, ∣an − x∣ = 00.00000dn+1 dn+2 ⋯ < .000001
⇒ an → x!

Example 5.6.3 (From Homework). E = {r2 ∣ r ∈ Q}, E ′ = [0, ∞). Let x ≥ 0. How to
construct a sequence an ∈ Q s.t. a2n → x?

Proof. By the previous example, we know that there exists a sequence an ∈ Q, s.t. an →
√
x ∈ R!
√ √
By limit rules, a2n = an ⋅ an → x ⋅ x = x! ⇒ a2n is a sequence of element of E that tends to
x.
On the other hand, if x < 0, I claim there is no sequence an ∈ Q with a2n → x, a2n ≥ 0 = bn , ∀n.
By problem 3 of Homework 8,

lim a2 ≥ 0 > x ⇒ lim a2n ≠ x, x ∉ E ′ , ∀x < 0.

n→∞ n n→∞

E ′ = [0, ∞).

Theorem 5.6.4. Let E ⊂ R and p ∈ R, then p is an accumulation point of E if and only if

for every ε > 0, the set E ∩ ((p − ε, p + ε) ∖ p) is nonempty.

Proof. (⇒) If P ∈ E ′ , then there exists a sequence an ∈ E, such that an → p. Given ε > 0,
∣an − p∣ < ε, for n sufficiently large, i.e. ∃ N , such that true for n ≥ N , particularly, (p−ε, p+ε)
contains an . Since an ≠ p,

((p − ε, p + ε) ∖ p) ∋ an , but an ∈ E

⇒ E ∩ ((p − ε, p + ε) ∖ p) ≠ ∅.
(⇐) Let ε = n1 . By assumption, ∃ x ∈ E, such that

p − ε < x < p + ε, x ≠ p.

71
Let an = x. Do this for each n ∈ N,
1
⇒ 0 < ∣an − x∣ <
n
⇒ an → x.

Example 5.6.5. E = Z (set of integers), E ′ = ∅.

Proof. Case 1: p ∈ Z. Let ε = 12 , (p − 21 , p + 12 ) ∩ Z = p. ⇒ (p − 21 , p + 21 ) ∖ p = ∅. By previous

theorem, p ∉ Z.
Case 2: p ∉ Z. Then p ∈ (n, n + 1), for some integer n, ε1 = p − n > 0 and ε2 = p − (n + 1) > 0.
Let ε = min (ε1 , ε2 ) ⇒ (p − ε, p + ε) ∩ Z = ∅
⇒ p is not an accumulation point of Z!
Overall, we get Z′ = ∅.

Example 5.6.6. E = { n1 ∣ n ∈ N} , E ′ = {0}, clearly, 0 ∈ E ′ .

Theorem 5.6.7. Let an be a convergent sequence with an → L, an ≠ L for all n. Let E =

{an ∣ n ∈ N} ⇒ E ′ = {L}.

Proof. Proof is on Homework.

Compare with ”Extra Credit” from this week’s homework, how to construct a sequence
with subsequences tending to any x ∈ R ?
Idea: Make a sequence out of Q !

{an ∣ n ∈ N} = Q.
Q′ = R.

Accumulation points of this set are just limits of subsequence.

⇒ A divergent sequence can have all manners of subsequence limits! (i.e. accumulation
points of {an ∣ n ∈ N}). Whereas, convergent sequences have only one accumulation point.

72
6 Functions and limits
6.1 Limits of functions (Apr 13)
Let E ⊂ R be a subset and p ∈ E ′ (an accumulation point of E). Consider a function
f ∶ E → R with domain E.
Definition 6.1.1. We say lim f (x) = L, if for any ε > 0, there exists δ > 0, such that
x→p
∣f (x) − L∣ < ε, for all x ∈ E with 0 < ∣x − p∣ < δ.
Note: p may not be a point of E. This is why we put
0 < ∣x − p∣
in the definition. You only have to consider x ≠ p.
Example 6.1.2. f (x) = 2x. Claim: for all p ∈ R, lim f (x) = 2p = L.
x→p

Proof. Given ε > 0, let δ = 2ε , if ∣x − p∣ < δ = 2ε , then ∣f (x) − L∣ = ∣2x − 2p∣ = 2 ∣x − p∣ < 2δ = ε,
definition is satisfied.

Example 6.1.3. f (x) = x2 . Claim: lim f (x) = 1.

x→1

Proof. Side Calculation:

f (x) − 1 = x2 − 1 = (x − 1)(x + 1).
If ∣x − 1∣ < δ, then
∣x2 − 1∣ = ∣x + 1∣ ∣x − 1∣
< δ ∣x + 1∣ .
If δ < 1, then
∣x − 1∣ < 1 ⇒ 0 ⩽ x ⩽ 2
⇒ ∣x + 1∣ ⩽ 3
⇒ ∣x2 − 1∣ < 3δ.
Let ε > 0, if ε ≥ 3, choose δ = 21 ; if ε < 3, choose δ = 3ε . By the previous side calculation, since
δ < 1, we have
ε
∣f (x) − 1∣ = ∣x2 − 1∣ < 3δ = 3 ⋅ = ε, if ε < 3.
3
1 3
If ε ≥ 3, then δ = 2 ⇒ 3δ = 2 < 3 ≤ ε.
In either case, we have ∣x2 − 1∣ < ε.

Note: We could make an equivalent definition of Limits, where we only consider 0 < ε < 1.
1
”∀0 < ε < 1, ∃ δ...”, ∣f (x) − L∣ <
< 1 < ε, if ε > 1.
2
Or, we could have, at the beginning of the proof, said: ”Without loss of generality, we assume
that ε < 1.”

73
 2
Example 6.1.4. f (x) = xx−2−4
, E = R ∖ {2}, p = 2 is accumulation point of R ∖ {2}(i.e. E),
1
an = 2 + n ⇒ It makes sense to write lim f (x) =?
x→2

Proof. Side Calculation:

x2 − 4 (x − 2)(x + 2)
=
x−2 x−2
= x + 2, for x ≠ 2
But since ∣x − 2∣ > 0 in the definition, we get

lim f (x) = lim(x + 2) = 4.

x→2 x→2

Definition 6.1.5. Suppose that p ∈ E, we say that f (x) is continuous at p if either lim f (x) =
x→p
f (p); or p is not an accumulation point of E. E = domain of f ⇒ f is continuous on E.
Example 6.1.6. f (x) = xn is continuous everywhere, for all p ∈ R.
Example 6.1.7. sin(x), cos(x), ax (a > 0) are also continuous everywhere.
Example 6.1.8.
1 x∈Q
f (x) = {
0 x∉Q
E = R. But f (x) is discontinuous everywhere. (∀p ∈ R, f (x) is discontinuous at P ).
Recall from last time: Let E ⊂ R, and p ∈ E ′ (set of all accumulation points of E ). Given
a function f ∶ E → R, we say f (x) → L as x → p, or lim f (x) = L, if: ∀ ε > 0, ∃ δ > 0 such
x→p
that 0 < ∣x − p∣ < δ ⇒ ∣f (x) − L∣ < ε, we say f (x) is continuous at x = p ∈ E if either

lim f (x) = f (p), or

x→p
p ∉ E ′.

Example 6.1.9. x, x2 , ⋯ are all continuous. sin(x), cos(x), ax (a > 0) are continuous ev-
erywhere (i.e. on E = R).
Example 6.1.10.
x x < 0.
f (x) = {
x + 1 x ≥ 0.
Claim 1: f (x) is continuous at any p ≠ 0.

Proof. Let p ≠ 0 and ε > 0, we may assume without loss of generality that ε ≤ ∣p∣. (If ε ≥ ∣p∣,
use the δ that works for ε = ∣p∣.)
Case 1: p > 0, we choose δ = ε. if ∣x − p∣ < δ = ε ≤ ∣p∣, i.e. ∣x − p∣ ≤ ∣p∣.

⇒x≥0

74
 ⇒ f (x) = x + 1
⇒ ∣f (x) − f (p)∣ = ∣x + 1 − (p + 1)∣ = ∣x − p∣ < δ = ε
⇒ ∣f (x) − f (p)∣ < ε.
Case 2: p < 0, choose δ = ε, if ∣x − p∣ < δ ≤ ∣p∣, then x < 0.

⇒ f (x) = x

⇒ ∣f (x) − f (p)∣ = ∣x − p∣ < δ = ε

⇒ ∣f (x) − f (p)∣ < ε.

Claim 2: f (x) is discontinuous at x = 0.

Proof. We choose ε = 1, we must show ∄ δ > 0, s.t. ⋯, i.e. ∀δ > 0, 0 < ∣x − 0∣ < δ ⇏
∣f (x) − f (0)∣ < ε. i.e. ∀ δ > 0, ∃ x with 0 < ∣x∣ < δ such that ∣f (x) − 1∣ ≥ ε. Given δ > 0, let
x = − 2δ ,
δ δ δ
⇒ 0 < ∣x∣ < δ and f (x) = f (− ) = − because − < 0
2 2 2
δ δ δ
⇒ ∣f (x) − 1∣ = ∣− − 1∣ = ∣1 + ∣ = 1 + ≥ ε.
2 2 2

Question: How discontinuous can a function be?

1 x∈Q
Proposition 6.1.11. The function f (x) = { is discontinuous everywhere. (Do-
0 x∉Q
main of f ∶ E = R!)

Proof. Let p ∈ R, we’ll show that for ε = 1, ∄ δ as required by definition.

Case 1: p ∈ Q, then f (p) = 1. Given δ > 0. denote x ∉ Q, such that ∣x − p∣ < δ,

⇒ ∣f (x) − f (p)∣ = ∣0 − 1∣ = 1 ≥ ε.

But R ∖ Q is dense in R, so such an x must exist.

Case 2: p ∉ Q, then f (p) = 0. Given δ > 0, ∃ x ∈ Q, such that p − δ < x < p by density of Q.
But then
f (x) = 1 ⇒ ∣f (x) − f (p)∣ = ∣1 − 0∣ = 1 ≥ ε.

75
6.2 Continuity of functions using sequences (Apr 15)
You can test for whether a function is continuous using sequences, in the following way:

Theorem 6.2.1. A function f: E → R is continuous at a point if and only if: for any
sequence an ∈ E with an → p as n → ∞, we have f (an ) → f (p) as n → ∞.

Proof. First note that if p ∉ E ′ , then there are no sequences an → p with an ∈ E. So the
Theorem is vacuously true. So we can assume without loss of generality (i.e. ”Wlog”) that
p ∈ E ′.
(⇒) Assume that f (x) is continuous at x = p. Given an → p as n → ∞, we need to show
that f (an ) → f (p). Let ε > 0, since lim f (x) = f (p), there exists δ > 0. such that
x→p

∣x − p∣ < δ ⇒ ∣f (x) − f (p)∣ < ε.

Since an → p as n → ∞, ∃ N such that ∣an − p∣ < δ for all n ≥ N. ⇒ ∣f (an ) − f (p)∣ < ε for all
n ≥ N . Since ε > 0 was arbitrary, we conclude that f (an ) → f (p) as n → ∞.
(⇐) We’ll show that if f (x) is discontinuous at x = p, then there exists a sequence an → p
with f (an ) ↛ f (p). Since f (x) is discontinuous at x = p, ∃ ε > 0, such that ∀ δ > 0.∃ x ∈ E
with 0 < ∣x − p∣ < δ but ∣f (x) − f (p)∣ ≥ ε. Let δ = n1 , choose an = x for this δ, so: 0 < ∣an − p∣ < n1 ,
but ∣f (an ) − f (p)∣ ≥ ε > 0. Then clearly an → p, but f (an ) ↛ f (p).

Corollary 6.2.2. Let f (x) and g(x) be continuous function at x = p. Then:

(i) f (x) + g(x) is continuous at x = p.
(ii) f (x) ⋅ g(x) is continuous at x = p.
(iii) f (x)/g(x) is continuous at x = p, if g(p) ≠ 0.

Proof. Just need to check an arbitrary sequence an → p. Since f and g are continuous,
f (an ) → f (p), g(an ) → g(p). By the limit rules for sequences, we have

f (an ) + g(an ) → f (p) + g(p)

f (an ) ⋅ g(an ) → f (p) ⋅ g(p)

f (an ) f (p)
→ , if g(p) ≠ 0.
g(an ) g(p)
Since an → p was arbitrary, by the previous Theorem, we are done.

Corollary 6.2.3. Let f, g, h be 3 functions with f (x) ≤ g(x) ≤ h(x), ∀x ∈ E, if lim f (x) =
x→p
L = lim h(x), then lim g(x) = L.
x→p x→p

Proof. Let an → p with an ∈ E, ⇒ f (an ) → L and h (an ) → L. By ”squeeze Theorem” for

sequences, g (an ) → L.

76
Corollary 6.2.4. Let g ∶ E → R, f ∶ H → R with g(E) ⊂ H. (f ○ g ∶ E → R is a well-defined
function). Suppose that g(x) is continuous at x = p and f (u) is continuous at u = g(p).
Then f ○ g is continuous at x = p.

Proof. Let an → p with an ∈ E, since g is continuous at x = p ⇒ g (an ) = g(p); since f is

continuous at u = g(p) ⇒ f [g(an )] → f [g(p)].
√
Example 6.2.5. Show that sin x is continuous on E = [0, π].

Proof.
√
Let g(x) = sin x f (u) = u
√
(f ○ g)(x) = sin x
sin(x) is continuous on [0, π]
sin([0, π]) = [0, 1]
Domain of f (u) = [0, ∞) and f (u) is continuous on [0, ∞). Since [0, 1] ⊂ [0, ∞), the previous
theorem applies.
√
⇒ sin x is continuous on E = [0, π).

Note: Yon can also use the above theorem to prove discontinuity! You just need to find
an → p, such that
f (an ) ↛ f (p),
prove that f (x) is discontinuous.

Proof. Case 1: p ∈ Q. Can I find an → p, such that f (an ) ↛ p ?

Let an = p + πn . Since π ∉ Q, πn ∉ Q
π
⇒p+ ∉ Q (since p ∈ Q)
n
⇒ an ∈ R ∖ Q
⇒ f (an ) = 0.
But an = p + πn → p because πn → 0 as n → ∞. So f (an ) = 0 ↛ f (p) = 1, ⇒ f is discontinuous
at p.
Case 2: p ∉ Q. One can make a similar argument.

77
6.3 Extreme values (Apr 20)
Next topic: max’s and min’s of function.
Definition 6.3.1. Let f ∶ E → R.
(a) f (x) is bounded above if f (x) < M for some M ∈ R, and for all x ∈ E. (Bounded below
is similar).
If f (x) is bounded above, we write

sup f (x) = sup f (E),

x∈E

where f (E) = {f (x) ∣ x ∈ E}. This is a real number! (Similarly, if f is bounded below,
inf x∈E f (x) = inf f (E).)
(b) Suppose that f (x) is bounded above. We say that f (x) attains its supremum if
∃ p ∈ E, s.t.f (p) = supx∈E f (x).
(Similarly, ”attains its infimum”.)
Example 6.3.2.
f (x) = 1 − x2 , E = [−1, 1]
Then supx∈E f (x) = 1. (Why? f (x) = 1 − x2 ≤ 1, ∀ x ∈ E and f (0) = 1)
inf x∈E f (x) = 0. (Why? If − 1 ≤ x ≤ 1, then

1 − x2 = (1 − x)(1 + x) ≥ 0

⇒ inf f (x) ≥ 0
x∈E

But f (−1) = f (1) = 0 ⇒ inf x∈E f (x) = 0 and is attained at x = −1 (and x = +1).)
Note: If E = R, then f (x) would not be bounded below!

⇒ inf f (x) = −∞
x∈E
⇒ inf can never be attained!
Example 6.3.3.
1
f (x) =
x2
E = (−∞, 0) ∪ (0, +∞) = R ∖ {0}.
supx∈E f (x) = +∞, f (x) is not bounded above! inf x∈E f (x) = 0.
(Why?
1
> 0, ∀x ∈ E
x2
⇒ inf f (x) ≥ 0
x∈E

78
In fact, inf x∈E f (x) = 0, because lim = 0!)
x→+/−∞
Given any N > 0, ∃x s.t. x12 < N
(E.g. x = √2N ⇒ x12 = N4 < N ).
Is the infimum attained on E?
1
No! inf f (x) = 0, x ∈ E, but f (x) = x2 > 0, ∀x ∈ E

⇒/∃ p ∈ E s.t. f (p) = inf f (x).

x∈E
Example 6.3.4.
f (x) = arctan(x), E = R.
Domain of arctan(x) = R! But values of arctan(x) are always between − π2 and π2 .
π
sup arctan(x) =
x∈R 2
π
inf arctan(x) = −
x∈R 2
⇒ arctan(x) is bounded above and below !
But neither the sup nor the inf are attained!
Example 6.3.5.

f (x) = arctan(x), E = [−1, 1] (different domain from previous example!)

π
⇒ sup f (x) = f (1) =
x∈E 4
π
⇒ inf f (x) = f (−1) = −
x∈E 4
Moral: Attainment of sup and inf depend both on continuity of f (x) and on the domain E.

Theorem 6.3.6. (Extreme Value Theorem). Let f (x) be a continuous function on a closed
interval E = [a, b]. Then f (x) is bounded both above and below, and attains both its sup and
its inf on E.
Note: In common language, supx∈[a,b] f (x) = ”absolute max” and inf x∈[a,b] f (x) = ”absolute min”.

Proof. Let
L = sup f (x) = sup{f (x) ∣ a ≤ x ≤ b}.
x∈[a,b]

We choose a sequence an ∈ [a, b] s.t. f (an ) → L, as follows:

If L < ∞, then for each n, let an be s.t.
1
L− < f (an ) ≤ L.
n

79
If L = ∞, then for each n, choose an s.t.

f (an ) > n.

In either case, we have:

a ≤ an ≤ b, and
f (an ) → L
(We don’t know if an is convergent or not!)
By the Bolzano–Weierstrass Theorem, since an is a bounded sequence of real numbers, it
has a convergent sub-sequence, ani s.t.

ani → p as i → ∞, for some a ≤ p ≤ b.

This still satisfies

f (ani ) → L as i → ∞!
But since p ∈ [a, b] and f (x) is continuous at p,
By Definition 2 of continuity, we have

lim f (ani ) = f (p).

i→∞

Since the limit of any sequence is unique, and

lim f (an ) = L = lim (ani )

n→∞ i→∞

⇒ L = f (δ).
Therefore, f (x) is bounded above (L < ∞), and attains its supremum at x = p.
The argument for inf is similar.

More Examples:
Example 6.3.7.
√
f (x) = sin4 (x) + cos18 (x) + 16x4 on E = [−1000, 1000].

(a) Is f (x) bounded on E?

(b) If so, does f (x) attain its sup/inf?
Answer: Let’s do (a) and (b) together.
By the Theorem, we just have to check:

• Is E a closed interval? (Yes! Since E = [−1000, 1000])

• Is f (x) continuous on E?

80
 √
g(x) is on [0, ∞)
g(x) = sin4 (x) + cos18 (x) + 16x4 ≥ 0 (since sin4 (x) ≥ 0, cos18 (x) ≥ 0, and 16x4 ≥ 0).
⇒ g(x) is on R, and g([−1000, 1000]) ⊂ [0, ∞)
By Corollary 6.2.4, we conclude that
√
f (x) = g(x) is continuous on E!

By the Extreme Value Theorem, it must be bounded and attain its sup / inf somewhere on
E!
Example 6.3.8.
x x∈Q
f (x) = {
0 x∉Q
Claim 1: p = 0, use squeeze theorem!

−∣x∣ ≤ f ∣x∣ ≤ ∣x∣

Claim 2: p ≠ 0, looks like

1 on Q
f (x) = {
0 on Qc

x x∈R
Example 6.3.9. Consider f (x) = { , and E = [−π, π].
0 x∉R
(a) Is f (x) bounded?
(b) Does it attain inf / sup?
Can we apply the T heorem? ⇒ No! f (x) needs to be for all p ∈ [−π, π](E)! It is only at
p = 0!
Answer:
(a) −π ≤ f (x) ≤ π is bounded!
(b)
sup f (x) = π
x∈[−π,π]

inf f (x) = −π
x∈[−π,π]

But there are not attained!

Because π is irrational, so f (π) = 0, f (−π) = 0.

6.4 A word or two about calculus (Apr 22)

Recall the main theorem of last class:

81
Theorem 6.4.1 (Extreme Value Theorem). Let f (x) be a continuous function on a closed
interval E = [a, b] (i.e. “∀p ∈ [a, b], f (x) is continuous at x = p”). Then f (x) is bounded
and attains its supremum and infimum.
In other words,
∃p ∈ [a, b] s.t. f (p) = sup f (x),
x∈[a,b]

and ∃q ∈ [a, b] s.t. f (q) = inf f (x).

x∈[a,b]

How to find these points p and q explicitly?

Definition 6.4.2. Let f ∶ E → R and p ∈ E ′ . We say that f (x) differentiable of x = p if the
limit
f (x) − f (p)
lim
x→p x−p
exists. In this case, the limit is denoted by f ′ (p), and is called the derivative of f (x) at
x = p.
Example 6.4.3.
f (x) = x2
Answer:
x2 − p 2
f ′ (p) = lim
x→p x − p

(x − p)(x + p)
= lim
x→p (x − p)
= lim(x + p) = p + p = 2p.
x→p

Thus, f ′ (x) = 2x.

Theorem 6.4.4. If f (x) is differentiable at p, then f (x) is continuous at p.
Proof. Note that f (x) is continuous at p iff

lim[f (x) − f (p)] = 0.

x→p

But we know that

f (x) − f (p)
lim
x→p x−p
exists.
Applying limit rule, we get
x−p
lim (f (x) − f (p))
x→p x−p
f (x) − f (p)
= lim ⋅ lim x − p
x→p x−p x→p

= 0 ⋅ f ′ (p) = 0. (Since lim x − p = p − p = 0)

x→p

82
 Note: Continuity is a necessary, but not sufficient condition for differentiability.

differentiability ⇒ continuity,

but
differentiability ⇍ continuity.
In other words, there exist continuous functions that are not differentiable.
Example 6.4.5.
f (x) = ∣x∣
Answer: This is continuous at x = 0.

x x≥0
f (x) = { .
−x x < 0

Then,
∣x∣ − 0
lim
x→0 x
∣x∣
= lim
x→0 x

For x > 0, let an = n1 , then

lim an = 1.
n→∞

For x < 0, let bn = − n1 , then

lim bn = −1.
n→∞

Therefore, the limit in the definition of f ′ (0) does not exist! So the f (x) is not differentiable
at x = 0.
Question: How non-differentiable can a continuous function be?
More specifically, can f ′ (x) fail to exist for infinitely many points x ∈ [a, b]?
Example 6.4.6.
1
x ⋅ sin = f (x)
x
We have:
1 x 1
f ′ (x) = sin − 2 ⋅ cos
x x x
1
1 cos x
= sin −
x x
By imitating this example, it is possible to construct a function that fails to be differentiable
on a whole sequence an → 0!
f ′ (an ) = dne.

83
⇒ Can get a countably infinite set of x where

f ′ (x) = dne.

In fact, one can construct a function f (x) that is continuous everehere, but differentiable
nowhere. (Preview of Math 521.)
Let’s say that f (x) is differentiable at all points on E. How can we use this fact in
finding min’s and max’s?
Definition 6.4.7. We say that f (x) has a local maximum at p if ∃δ > 0 s.t. ∀x ∈ (p − δ, p +
δ), f (x) ≤ f (p). Local minimum: f (x) ≥ f (p).
local max ≠ absolute max

Theorem 6.4.8 (Fermat’s Theorem). Let f ∶ (a, b) → R. If f (x) has a local maximum/minimum
at p ∈ (a, b), and f (x) is differentiable at x = p, then f ′ (p) = 0.
Note: If the derivative at x = p does not exist, then the Theorem does not apply.

Corollary 6.4.9 (“Closed interval method”). If f (x) is a continuous function on [a, b],
then the absolute maximum and minimum can be found among the set of points a ≤ x ≤ b
where:

• x = a or x = b

• f (x) = 0

• f ′ (x) = DN E.

Point: we reduce from checking an infinity of points x ∈ [a, b] to (usually) a finite list!

Proof of Fermat’s Theorem. Since f ′ (p) = lim f (x)−f

x−p
(p)
exists, we can use any sequence an →
x→p
p to compute it.
Recall that a local maximum satisfies f (x) ≤ f (p) ∀p − δ ≤ x ≤ p + δ (for some δ).
Let an = p − n1
⇒ an → p, p − δ < an < p
⇒ f (an ) ≤ f (x), f (an ) − f (x) ≤ 0.
But an − x < 0, so
1
<0
an − x
f (an ) − f (p)
⇒ ≥0
an − p
f (an ) − f (p)
⇒ lim ≥ 0 (we assume that this exists!)
n→∞ an − p

84
 ⇒ f ′ (p) ≥ 0.
Next, let bn = p + nδ .
δ
bn − p = > 0, f (bn ) − f (p) ≤ 0
n
f (bn ) − f (p)
⇒ ≤0
bn − p
f (bn ) − f (p)
⇒ lim ≤0
n→∞ bn − p
⇒ f ′ (p) ≤ 0
Since,
0 ≤ f ′ (p) ≤ 0,
Thus,
f ′ (p) = 0

Corollary 6.4.10 (Rolle’s Theorem). If f (x) is on [a, b] and differentiable on (a, b), and
f (a) = f (b), then ∃x ∈ (a, b) s.t. f ′ (x) = 0.

Proof. Since f (x) is on [a, b], it has a maximum and minimum.

Case 1: The absolute maximum is attained at x ∈ (a, b). By F ermat, f ′ (x) = 0.
Case 2: The absolute minimum is attained at x ∈ (a, b).
Case 3: Both the maximum and minimum occur at the end points.
Then f (a) and f (b) are the maximum/minimum, on vice-versa.
But f (a) = f (b)!
So f (a) ≤ f (x) ≤ f (b) = f (a) ∀x ∈ [a, b].

⇒ f (x) = constant

⇒ f ′ (x) = 0, ∀x ∈ [a, b].

Corollary 6.4.11 (Mean Value Theorem). Let f (x) be on [a, b] and differentiable on (a, b).
∃x ∈ (a, b) s.t.
f (b) − f (a)
f ′ (x) = .
b−a
Proof. Let
f (b) − f (a)
g(x) = f (x) − (x − a) ⋅ .
b−a
Then g(x) is differentiable on (a, b). And,
f (b) − f (a)
g(a) = f (a) − (a − a) = f (a) − 0 = f (a)
b−a
85
 f (b) − f (a)
g(b) = f (b) − (b − a) ⋅ = f (b) − (f (b) − f (a)) = f (b) − f (b) + f (a) = f (a) = g(a)
b−a
⇒ g(a) = g(b)!
By Rolle′ s, ∃x ∈ (a, b) s.t. g ′ (x) = 0.
But
f (b) − f (a)
0 = g ′ (x) = f ′ (x) − .
b−a
Thus,
f (b) − f (a)
f ′ (x) = .
b−a

Corollary 6.4.12. Let f (x) be continuous on [a, b] and differentiable on (a, b). If f ′ (x) = 0
for all x ∈ [a, b], then f (x) is constant on [a, b].
f (x)−f (a)
Proof. If f (x) ≠ f (a) for some x ∈ [a, b], then ∃y ∈ (a, x) s.t. f ′ (y) = x−a ≠ 0.

⇒ f ′ (y) ≠ 0.

But this is the contrapositive.

Note: This last result is used in an essential way during the proof of the Fundamental
Theorem of Calculus.

86