Module - Real Analysis

Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 233

Surigao State College of Technology

Narciso St., Surigao City

Real Analysis

Submitted by:
BS MATH 3A

Submitted to:
Dr. Louida P. Patac

0
Table of Contents
Sets and Functions ………………………. 2
Special Types of Function ………………………. 18
Mathematical Induction ………………………. 33
Finite and Infinite Sets ………………………. 48
Countable and Uncountable Sets ………………………. 61
The Algebraic and Order Property of Real Number ………………. 72
Absolute Value and Real Line ………………………. 85
Completeness Property of Real Number ………………………. 96
Applications of Supremum Property ………………………. 108
Intervals ………………………. 118
Sequences and their Limits ………………………. 129
Limit Theorem ………………………. 143
Monotone Sequences ………………………. 153
Subsequences ………………………. 164
The Cauchy Criterion ………………………. 177
Properly Divergent Sequence ………………………. 190
Introduction to Infinite Series ………………………. 201
Limits of Functions ………………………. 213

1
SETS and FUNCTIONS

JOHN KENNARD T. SEVILLE

2
SET

Definition:
A set is a well-defined collection of distinct objects. The objects that
make up a set (also known as the set's elements or members) can be
anything: numbers, people, letters of the alphabet, other sets, and so
on. A set can have any non-negative quantity of elements, ranging from
none (the empty set or null set) to infinitely many. The number of
elements in a set is called the cardinality, and can range from zero to
denumerably infinite (for the sets of natural numbers, integers, or q
rational numbers) to non-denumerably infinite for the sets of irrational
numbers, real numbers, imaginary numbers, or complex numbers).
Set Operations:
The set { x | some property of x } is the set of all objects x that satisfy the given
property. Formally, we have that w ∈ { x | some property of x } if and only if the
specified property holds for w.
The set A ∪ B is the set { x | x ∈ A or x ∈ B }. Equivalently, x ∈ A ∪ B if and only
if x ∈ A or x ∈ B.
The set A ∩ B is the set { x | x ∈ A and x ∈ B }. Equivalently, x ∈ A ∩ B precisely
if x ∈ A and x ∈ B.
The set A – B is the set { x | x ∈ A and x ∉ B }. This set is also sometimes
denoted A \ B.
The set A Δ B is the set { x | exactly one of x ∈ A and x ∈ B is true }.

Examples:
N : the set of natural numbers
W : the set of whole numbers
𝑍 + or I : the set of integers Z + :
𝑍 − : the set of negative integers
Q : the set of rational numbers
R : the set of real numbers
C : the set of complex number

REPRESENTATION OF A SET

There are two methods to represent a set.


(i) Roster method (Tabular form)
In this method a set is represented by listing all its elements, separating
these by commas and enclosing these in curly bracket. If V be the set of
vowels of English alphabet, it can be written in Roster form as : V = { a,
e, i, o, u}.
If A be the set of natural numbers less than 7. then A={1, 2, 3, 4, 5, 6}, is in the
Roster form.

3
Note:
To write a set in Roster form elements are not to be repeated i.e. all
elements are taken as distinct.
For example if A be the set of letters used in the word mathematics, then
A = {m, a, t, h, e, i, c, s}.

Examples:
(a) C = { x : x ∈ 𝑁 𝑎𝑛𝑑 50 ≤ 𝑥 ≤ 60}
(b) D = {x: x ∈ ℝ 𝑎𝑛𝑑 𝑥 2 − 5𝑥 + 6 = 0}

(ii) Set-builder form


In this form elements of the set are not listed but these are represented by
some common property.

Let V be the set of vowels of English alphabet then V can be written in the
set builder form as: V = {x : x is a vowel of English alphabet}
Let A be the set of natural numbers less than 7. then A = {x : x ∈ 𝑁 and 1
≤x <7}.

CLASSIFICATION OF SETS

Finite and infinite sets


Definition:
Let A and B be two sets where A = {x : x is a natural number} B = {x : x
is a student of your school} As it is clear that the number of elements in
set A is not finite (infinite) while number of elements in set B is finite. A
is said to be an infinite set and B is said to be is finite set. A set is said
to be finite if its elements can be counted and it is said to be infinite if it
is not possible to count upto its last element.

Empty (Null) Set


Definition:
Consider the following sets. A = {x: x ∈ ℝ 𝑎𝑛𝑑 𝑥 2 + 1 = 0 } B = {x : x is
number which is greater than 7 and less than 5}.Set A consists of real
numbers but there is no real number whose square is -1. Therefore, this
set consists of no element. Similarly, there is no such number which is
less than 5 and greater than 7. Such a set is said to be a null (empty)
set. It is denoted by the symbol void, 𝜙 or {}. A set which has no element
is said to be a null/empty/void set and is denoted by.

Singleton Set
Definition:
Consider the following set: A = {x : x is an even prime number} As there
is only one even prime number namely 2, so set A will have only one

4
element. Such a set is said to be singleton. Here A = {2}. A set which
has only one element is known as singleton.

Equal and equivalent sets


Definition:
Consider the following examples.
(i) A = {1,2,3}, B = {2,1,3}, (ii) D = {1,2,3} , E = {a,b,c} .
(i) Sets A and B have the same elements. Such sets are said to be equal sets
and it is written as A = B. In example (ii) set D and E have the same number
of elements but elements are different. Such sets are said to be equivalent
sets and are written as A » B. Two sets A and B are said to be equivalent
sets if they have same number of elements, but they are said to be equal if
they have not only the same number of elements, but elements are also the
same.

Definition:
The cardinality of a set is its size. For a finite set, the cardinality of a set is
the number of members it contains. In symbolic notation the size of a set
S is written |S|. We will deal with the idea of the cardinality of an infinite set
later.
Example:
Set cardinality for the set S = {1, 2, 3} we show cardinality by writing |S| = 3.
We now move on to a number of operations on sets. You are already familiar
with several operations on numbers such as addition, multiplication, and
negation.

Definition:
The intersection of two sets S and T is the collection of all objects that are in
both sets. It is written S ∩ T. Using curly brace notation S ∩ T = {x : (x ∈ S) and
(x ∈ T )} The symbol and in the above definition is an example of a Boolean or
logical operation. It is only true when both the propositions it joins are also true.
It has a symbolic equivalent ∧. This lets us write the formal definition of
intersection more compactly: S ∩ T = {x : (x ∈ S) ∧ (x ∈ T )
Example:
Intersections of sets Suppose S = {1, 2, 3, 5}, T = {1, 3, 4, 5}, and U = {2, 3, 4,
5}. Then:
S ∩ T = {1, 3, 5}, S ∩ U = {2, 3, 5}, and T ∩ U = {3, 4, 5}

Disjoint Sets
Definition:
Two sets are said to be disjoint if they do not have any common element. For
example, sets A= { 1,3,5} and B = { 2,4,6 } are disjoint sets.

Sub- Set:
Definition:

5
Let set A be a set containing all students of your school and B be a set
containing all students of class XII of the school. In this example each element
of set B is also an element of set A. Such a set B is said to be subset of the
set A. It is written as B ⊆ A.
Consider:
D ={1, 2, 3, 4,........}
E = {.....-3-2,-1, 0, 1, 2, 3, .......}
Clearly each element of set D is an element of set E also ∴ 𝐷 ⊆ E If A and B
are any two sets such that each element of the set A is an element of the set B
also, then A is said to be a subset of B.
Remarks
(i) Each set is a subset of itself i.e. A ⊆ A .
(ii) Null set has no element so the condition of becoming a subset is
automatically satisfied. Therefore, null set is a subset of every set.
(iii) If A ⊆ B and B⊆ A then A = B.
(iv) If A ⊆ B and A≠ B, then A is said to be a proper subset of B and B is said to
be a super set of A. i.e. A ⊂ B or B ⊃ A.
Examples:
1. If A = {x : x is a prime number less than 5} and B = {y : y is an even prime
number} then is B a proper subset of A ?
Solution : It is given that A = {2, 3 }, B = {2}. Clearly B ⊆ A and B A ¹ We
write B ≠ A and say that B is a proper subset of A.
2. If A = {1, 2, 3, 4}, B = {2, 3, 4, 5}. is A⊆ B or B ⊆ A ?
Solution : Here 1 ∉ A but 1 ∉ B⟹ A⊈ B. Also 5∈ B but 5 ∉ A ⟹ B ⊈ A
. Hence neither A is a subset of B nor B is a subset of A.
3. If A = {a, e, i, o, u} B = {e, i, o, u, a } Is A ⊆ B orB ⊆ A or both ?
Solution : Here in the given sets each element of set A is an element of set
B also ∴ A ⊆ B (i) and each element of set B is an element of set A also. ∴
B⊆ A. (ii) From (i) and (ii) A = B

Power Set
Definition:
Let A = {a, b} Subset of A are f , {a}, {b} and {a, b}. If we consider these subsets
as elements of a new set B (say) then B = {𝜙,{a},{b},{a,b}} B is said to be the
power set of A.
Notation:
Power set of a set A is denoted by P(A). Power set of a set A is the set of all
subsets of the given set.
Examples:
1. A={𝑥: 𝑥 ∈ ℝ 𝑎𝑛𝑑 𝑥 2 + 7 = 0}
2. B={𝑦: 𝑦 ℕ 𝑎𝑛𝑑 1 ≤ 𝑦 ≤ 3}

Solution:
(i) Clearly A = 𝜙 (Null set)
∴ 𝜙 is the only subset of given set

6
∴ P (A)={ 𝜙 }
(ii) The set B can be written as {1, 2, 3} Subsets of B are 𝜙 , {1}, {2}, {3},
{1, 2}, {1, 3}, {2, 3}, {1, 2, 3}. \ P (B) = { 𝜙 , {1}, {2}, {3}, {1, 2}, {1, 3}, {2, 3},
{1, 2, 3} } .

Universal Set
Consider the following sets.
A = {x : x is a student of your school}
B = {y : y is a male student of your school}
C = {z : z is a female student of your school}
D = {a : a is a student of class XII in your school}
Clearly the set B, C, D are all subsets of A.
A can be considered as the universal set for this particular example. Universal
set is generally denoted by U. In a particular problem a set U is said to be a
universal set if all the sets in that problem are subsets of U.
Remarks:
(i) Universal set does not mean a set containing all objects of the universe. (ii)
A set which is a universal set for one problem may not be a universal set for
another problem.
Example:
Which of the following set can be considered as a universal set?
X = {x : x is a real number}
Y = {y : y is a negative integer}
Z = {z : z is a natural number}
Solution:
As it is clear that both sets Y and Z are subset of X. ∴ X is the universal set for
this problem.

VENN DIAGRAM\
Definition:
British mathematician John Venn (1834-1883 AD) introduced the concept of
diagrams to represent sets. According to him universal set is represented by
the interior of a rectangle and other sets are represented by interior of circles.
For example, Let X = {1, 2, 3} and Let Y = {3, 4, 5}. Draw and label a Venn
diagram to show the intersection of sets X and Y.

The Venn Diagram in example 4 makes it easy to see that the number 3 is
common to both sets. So, the intersection of X and Y is 3. This is
what X and Y have in common. The intersection of X and Y is written as X ∩

7
𝑌 and is read as "X intersect Y". So, Intersection means "X and Y". In example
5 below, we will find the union of two sets. The union of two sets is the set
obtained by combining the elements of each.
Definition:
The compliment of a set S is the collection of objects in the universal set that
are not in S. The compliment is written 𝑆 𝑐 . In curly brace notation 𝑆 𝑐 = {x : (x ∈
U) ∧ (x /∈ S)} or more compactly as 𝑆 𝑐 = {x : x /∈ S} however it should be
apparent that the compliment of a set always depends on which universal set
is chosen. There is also a Boolean symbol associated with the
complementation operation: the not operation. The notation for not is ¬. There
is not much savings in space as the definition of compliment becomes 𝑆 𝑐 = {x :
¬(x ∈ S)}.
Example: Set Compliments
(i) Let the universal set be the integers. Then the compliment of the even
integers is the odd integers.
(ii) Let the universal set be {1, 2, 3, 4, 5}, then the compliment of S = {1, 2, 3}
is S c = {4, 5} while the compliment of T = {1, 3, 5} is T c = {2, 4}.
Definition:
The difference of two sets S and T is the collection of objects in S that are not
in T. The difference is written S − T . In curly brace notation S − T = {x : x ∈ (S
∩ (T c ))}, or alternately S − T = {x : (x ∈ S) ∧ (x /∈ T )}
Notice how intersection and complementation can be used together to create
the difference operation and that the definition can be rephrased to use Boolean
operations. There is a set of rules that reduces the number of parenthesis
required.
These are called operator precedence rules.
(i) Other things being equal, operations are performed left-to-right.
(ii) Operations between parenthesis are done first, starting with the innermost
of nested parenthesis.
(iii) All complementation’s are computed next.
(iv) All intersections are done next.
(v) All unions are performed next.
(vi) Tests of set membership and computations, equality or inequality are
performed last. Special operations like the set difference or the symmetric
difference, defined below, are not included in the precedence rules and thus
always use parenthesis.
Example: Operator precedence
Since complementation is done before intersection the symbolic definition of
the difference of sets can be rewritten: S − T = {x : x ∈ S ∩ T c }
If we were to take the set operations A ∪ B ∩ C c and put in the parenthesis we
would get
(A ∪ (B ∩ (C c )))

Definition:

8
The symmetric difference of two sets S and T is the set of objects that are in
one and only one of the sets. The symmetric difference is written S∆T. In curly
brace notation: S∆T = {(S − T ) ∪ (T − S)}
Example:
Symmetric differences Let S be the set of non-negative multiples of two that are
no more than twenty four.
Let T be the nonnegative multiples of three that are no more than twenty four.
Then S∆T = {2, 3, 4, 8, 9, 10, 14, 15, 16, 20, 21, 22}.
Another way to think about this is that we need numbers that are positive
multiples of 2 or 3 (but not both) that are no more than 24.
Another important tool for working with sets is the ability to compare them. We
have already defined what it means for two sets to be equal, and so by
implication what it means for them to be unequal. We now define another
comparator for sets.
Definition: For two sets S and T we say that S is a subset of T if each element
of S is also an element of T . In formal notation S ⊆ T if for all x ∈ S we have x
∈ T . If S ⊆ T then we also say T contains S which can be written T ⊇ S. If S ⊆
T and S 6= T then we write S ⊂ T and we say S is a proper subset of T .
Example:
If A = {a, b, c} then A has eight different subsets: ∅ {a} {b} {c} {a, b} {a, c} {b, c}
{a, b, c}
Notice that A ⊆ A and in fact each set is a subset of itself. The empty set ∅ is a
subset of every set. We are now ready to prove our first proposition. Some new
notation is required, and we must introduce an important piece of mathematical
culture. If we say “A if and only if B” then we mean that either A and B are both
true or they are both false in any given circumstance.
For example: “an integer x is even if and only if it is a multiple of 2”. The phrase
“if and only if” is used to establish logical equivalence. Mathematically, “A if and
only if B” is a way of stating that A and B are simply different ways of saying the
same thing. The phrase “if and only if” is abbreviated iff and is represented
symbolically as the double arrow ⇔. Proving an iff statement is done by
independently demonstrating that each may be deduced from the other.

Proposition 1:
Two sets are equal if and only if each is a subset of the other. In symbolic
notation:
(A = B) ⇔ (A ⊆ B) ∧ (B ⊆ A)
Proof:
Let the two sets in question be A and B. Begin by assuming that A = B. We
know that every set is a subset of itself so A ⊆ A. Since A = B we may substitute
into this expression on the left and obtain B ⊆ A. Similarly, we may substitute
on the right and obtain A ⊆ B. We have thus demonstrated that if A = B then A
and B are both subsets of each other, giving us the first half of the iff. Assume
now that A ⊆ B and B ⊆ A. Then the definition of subset tells us that any element
of A is an element of B. Similarly, any element of B is an element of A. This

9
means that A and B have the same elements which satisfies the definition of
set equality. We deduce A = B and we have the second half of the iff. ∎

Proposition 2:
De Morgan’s Laws Suppose that S and T are sets. DeMorgan’s Laws state that
(i) (S ∪ T) c = S c ∩ T c, and (ii) (S ∩ T) c = S c ∪ T c.
Proof:
Let x ∈ (S ∪ T) c; then x is not a member of S or T. Since x is not a member of
S we see that x ∈ S c. Similarly, x ∈ T c. Since x is a member of both these
sets, we see that x ∈ S c ∩ T c and we see that (S ∪ T) c ⊆ S c ∩ T c. Let y ∈
S c ∩ T c. Then the definition of intersection tells us that y ∈ S c and y ∈ T c.
This in turn lets us deduce that y is not a member of S ∪ T, since it is not in
either set, and so we see that y ∈ (S ∪ T) c. This demonstrates that S c ∩ T c
⊆ (S ∪ T) c. Applying Proposition 2.1 we get that (S ∪ T) c = S c ∩ T c and we
have proven part (i). The proof of part (ii) is left as an exercise. ✷ In order to
prove a mathematical statement you must prove it is always true. In order to
disprove a mathematical statement, you need only find a single instance where
it is false. It is thus possible for a false mathematical statement to be “true most
of the time”. In the next chapter we will develop the theory of prime numbers.
For now, we will assume the reader has a modest familiarity with the primes.
The statement “Prime numbers are odd” is false once, because 2 is a prime
number. All the other prime numbers are odd. The statement is a false one.
This very strict definition of what makes a statement true is a convention in
mathematics. We call 2 a counter example. It is thus necessary to find only one
counterexample to demonstrate a statement is (mathematically) false.
Disproof by counter example:
Prove that the statement A ∪ B = A ∩ B is false.
Let A = {1, 2} and B = {3, 4}. Then A ∩ B = ∅ while A ∪ B = {1, 2, 3, 4}. The sets
A and B form a counterexample to the statement.

Functions
Definition:
Let X and Y be sets. A function f from X to Y is a rule that assigns every element
x of X to a unique y in Y . We write f : X → Y and f(x) = y. Formally, using
predicate logic:
(∀x ∈ X, ∃y ∈ Y, y = f(x)) ∧ (∀x1, x2 ∈ X, f(x1) 6= f(x2) → x1 6= x2).
Then X is called the domain of f, and Y is called the codomain of f. The element
y is the image of x under f, while x is the preimage of y under f. Finally, we call
range the subset of Y with preimages.
Example:
Consider the assignment rule f : X = {a, b, c} → Y = {1, 2, 3, 4} which is defined
by: f = {(a, 2),(b, 4),(c, 2)}. We first check that this is a function. For every
element in X, we do have an assignment: f(a) = 2, f(b) = 4, f(c) = 2. Then the
condition that whenever f(x1) 6= f(x2) it must be that x1 6= x2 is also satisfied.
The the domain of f is X, the codomain of f is Y . The preimage of 2 is {a, c}

10
because f(a) = f(c) = 2. For the range, we look at Y , and among 1, 2, 3, 4, only
2 and 4 have a preimage, therefore the range is {2, 4}.
Example:
The rule f that assigns the square of an integer to this integer is a function.
Indeed, every integer has an image: its square. Also whenever two squares
are different, it must be that their square roots were different. We write
f : ℤ → ℤ, f(x) = 𝑥 2 .
Its domain is Z, its codomain is Z as well, but its range is {0, 1, 4, 9, 16, . . .},
that is the set of squares in Z.

Definition:
Let f be a function from X to Y , X, Y two sets, and consider the subset S ⊂ X.
The image of the subset S is the subset of Y that consists of the images of the
elements of S: f(S) = {f(s), s ∈ S}

Definition:
A function f is one-to-one or injective if and only if f(x) = f(y) implies x = y for
all x, y in the domain X of f. Formally: ∀x, y ∈ X(f(x) = f(y) → x = y). In words,
this says that all elements in the domain of f have different images.
Example.
Consider the function f : R → R, f(x) = 4x − 1. We want to know whether each
element of R has a different image. Yes, this is the case, why? well, visually,
this function is a line, so one may “see” that two distinct elements have distinct
images but let us try a proof of this. We have to show that f(x) = f(y) implies x
= y. Ok, let us take f(x) = f(y), that is two images that are the same. Then f(x) =
4x − 1, f(y) = 4y − 1, and thus we must have 4x − 1 = 4y − 1. But then 4x = 4y
and it must be that x = y, as we wanted. Therefore, f is injective.
Example.
Consider the function g: R → R, g(x) = x 2. Do we also have that two distinct
reals have distinct images? Well no... because 1 and −1 are both sent to 1...so
this function is not injective! If g(x) = g(y) = 1, we cannot conclude that x = y, in
fact this is wrong, it could be that x = −y.

Definition:
A function f is onto or surjective if and only if for every element y ∈ Y, there
is an element x ∈ X with f(x) = y: ∀y ∈ Y, ∃x ∈ X, f(x) = y. In words, each element
in the co-domain of f has a pre-image.
Example.
Consider again the function f: R → R, f(x) = 4x − 1. We want to know whether
each element of R has a preimage. Yes, it has, let us see why: we want to
show that there exists x such that f(x) = 4x − 1 = y. Given y, we have the
relation x = (y + 1)/4 thus this x is indeed sent to y by f.
Example.
Consider again the function g : R → R, g(x) = x 2 . Does each element in R
have a preimage? well, again no... Because R contains all the negative real

11
numbers, and it is not possible to square a real number and get something
negative... Formally, if y = −1, there is no x ∈ R such that g(x) = x 2 = −1.

Definition:
A function f is a one-to-one correpondence or bijection if and only if it is
both one-to-one and onto (or both injective and surjective). An important
example of bijection is the identity function.

Definition:
The identity function iA on the set A is defined by: iA : A → A, iA(x) = x.
Example 102. Consider the function f : R → R, f(x) = 4x − 1, which we have
just studied in two examples. We know it is both injective and surjective ,
therefore it is a bijection. Bijections have a special feature: they are invertible,
formally:

Definition:
Let f : A → B be a bijection. Then the inverse function of f, f −1 : B → B is
defined elementwise by: f −1 (b) is the unique element a ∈ A such that f(a) =
b. We say that f is invertible. Note the importance of the hypothesis: f must be
a bijection, otherwise the inverse function is not well defined. For example, if f
is not one-to-one, then f −1 (b) will have more than one value, and thus is not
properly defined. Note that given a bijection f : A → B and its inverse f −1 : B
→ A, we can write formally the above definition as: ∀b ∈ B, ∀a ∈ A(f −1 (b) =
a ⇐⇒ b = f(a)).
Example.
Let us look again at our two previous examples, namely, f(x) = 4x − 1 and g(x)
= x 2 . Then g(x), for g : R → R is not a bijection, so it cannot have an inverse.
Now f(x) is a bijection, so we can compute its inverse. Suppose that y = f(x),
𝑦+1 𝑦+1
then y = 4x − 1 ⇐⇒ y + 1 = 4x ⇐⇒ x = 4 , and f −1 (y) = 4 . We saw that
for the notion of inverse 𝑓 −1 to be defined, we need f to be a bijection. The
next result shows that 𝑓 −1 is a bijection as well.

Proposition 1.
If f : X → Y is a one-to-one correspondence, then f −1 : Y → X is a one-to-one
correspondence. Proof. To prove this, we just apply the definition of bijection,
namely, we need to show that f −1 is an injection, and a surjection. Let us start
with injection.
•𝑓 −1 is an injection: we have to prove that if f −1 (𝑦1 ) = f −1 (𝑦2 ), then y1 = 𝑦2 .
All right, then f −1 (𝑦1 ) = f −1 (𝑦2 ) = x for some x in X. But f −1 (𝑦1 ) = x means
that 𝑦1 = f(x), and 𝑓 −1 (𝑦2 ) = x means that 𝑦2 = f(x), by definition of the inverse
of function. But this shows that 𝑦1 = 𝑦2 , as needed.
• 𝑓 −1 is an surjection: by definition, we need to prove that any x ∈ X has a
preimage, that is, there exists y such that f −1 (y) = x. Because f is a bijection,
there is some y such that y = f(x), therefore x = f −1 (y) as needed. Suppose
that you have two functions f and g. It may be possible to compose them to
obtain a third function, here is how:

12
Definition:
Let f : A → B be a function, and g : B → C be a function. Then the composition
of f and g is a new function denoted by g ◦ f, and defined by: g ◦ f : A → C, (g
◦ f)(a) = g(f(a)). Note that the codomain of f is B, which is the domain of g.
Under this condition, the composition g ◦ f consists of applying first f, and then
apply g on the result. Therefore, g ◦ f 6= f ◦ g in general!
Example. Consider f : Z → Z and g : Z → Z defined by f(n) = 2n+3, g(n) = 3n
+ 2. We have (f ◦ g)(n) = f(g(n)) = f(3n + 2) = 2(3n + 2) + 3 = 6n + 7, while (g ◦
f)(n) = g(f(n)) = g(2n + 3) = 3(2n + 3) + 2 = 6n + 11. Suppose now that you
compose two functions f, g, and both of them turn out to be injective. The next
result tells us that the combination will be as well.

Proposition 2.
Let f : X → Y and g : Y → Z be two injective functions. Then g ◦ f is also
injective. Proof. What we need to do is check the injectivity of a function, so
we do this as usual: we check that g ◦ f(𝑥1 ) = g ◦ f(𝑥2 ) implies 𝑥1 = 𝑥2 . Typically,
to be able to prove this, you will have to keep in mind assumptions, namely
that both f and g are injective. So let us start. We have g ◦f(𝑥1 ) = g ◦f(𝑥2 ) or
equivalently g(f(𝑥1 )) = g(f(𝑥2 )). But we know that g is injective, so this implies
f(𝑥1 ) = f(𝑥2 ). Next we use that f is injective, thus 𝑥1 = 𝑥2 , as needed! Let us
ask the same question with surjectivity, namely whether the composition of
two surjective functions gives a function which is surjective too.

Proposition 3.
Let f : X → Y and g : Y → Z be two surjective functions. Then g ◦ f is also
surjective. Proof. The codomain of g ◦ f is Z, therefore we need to show that
every z ∈ Z has a preimage x, namely that there always exists an x such that
g ◦ f(x) = z. Again, we keep in mind that f and g are both surjective. Since g is
surjective, we know there exists y ∈ Y such that g(y) = z. Now again, since f
is surjective, we know there exists x ∈ X such that f(x) = y. Therefore there
exist x, y such that z = g(y) = g(f(x)) as needed.

Domain and Range of a Function

This section defines and gives examples of domains and ranges of functions.
These are important properties of a function and we will meet them in
subsequent sections. We write a function using the notation f(x). The notation
means that given a number x then the function gives another unique number
f(x). If we write y = f(x) then we say that x is the independent variable and y the
dependent variable. Note that a function can be written as f(x) = x 3 or as h(t)
= t 3 or as c(y) = y 3. These are all the same function - they all do the same
thing, cube a number.
Domain
Definition:

13
The domain of a function is the complete set of possible values of the
independent variable. The domain is the set of all possible x-values which will
make the function "work", and will output real y-values. When finding the
domain, remember:
The denominator (bottom) of a fraction cannot be zero,
The number under a square root sign must be positive in this section
Examples:
1. f(x) = x 2 . Defined for all values of x i.e. the domain is R.
2. f(t) = 2t . Defined for all values of t i.e. the domain is R.
3. f(z) = √𝑧. Defined for z ≥ 0 i.e. the domain is [0,∞) or R+.
4. f(x) =√4 + 3x − x 2 . We need 4 + 3x − x 2 ≥ 0 and solving this inequality
we see that this is only true for −1 ≤ x ≤ 4. So this function is only defined in this
range of values for x i.e. the domain is [−1, 4].
5. f(x) = sin(x). Defined for all values so the domain ℝ is .
6. f(u) = 1/u. The domain is defined for all u ≠= 0 and is denoted by ℝ − {0}.

Range
Definition:
The range of a function is the complete set of all possible resulting values of
the dependent variable (y, usually), after we have substituted the domain.
For example, consider the function defined by the rule that we take an input
and raise it to the third power. This can be

represented in equation form as y = x^3, and when this function is given input
values of {-2, -1, 0, 1, 2}, we can find the corresponding outputs by plugging
those inputs in for 'x' in the equation. For instance, if we input -2, we have y =
(-2)^3 = -8, so when the input is -2, the output is -8, and -8 is in our range. When
we find each of the corresponding outputs to our inputs, we have our range. In
our example, the range is {-8, -1, 0, 1, 8}, because these are the outputs
corresponding to -2, -1, 0, 1, and 2 respectively.
Graph of Domain and Range example:

14
15
Exercises on Sets and Functions

Exercise 1. Is it true that 0 ∈ {{0}}?


Exercise 2. If S = {{a}, {b}}, and T = {{a}, b}, is it true that S = T?
Exercise 3. If S = {A, B}, what is S ∪ ∅?
Exercise 3. If S = {A, B}, what is S ∪ ∅?
Exercise 4. For S as in the previous exercise, what is S ∪ S?
Exercise 5. If S = {A, B}, what is S ∩ ∅?
Exercise 6. If S = {A, B}, what is S ∩ S?
Exercise 7. Is it true that if S ⊂ T and T ⊂ S, then S = T?
Exercise 8. Find all the subsets of {{A, B}, {}}.
Exercise 9. For S = {1, 2}, what is S × S?
Exercise 10. What is the range of the identity function Id : S → S?
Exercise 11. For S = {A, B}, T = {A, B, C}, how many functions f : S → T are there with
the range of f equal to the range of the inclusion j : S → T?
Exercise 12. The example above is a bit special because the right hand side of the
equation is 0. Generalizing the above, explain how the solution to the equation f(x) =
g(x) can be interpreted as a subset of Z for any functions f : Z → Z, g : Z → Z.
Exercise 13. Let S = {1, 2} and T = {A, B, C}.
(i) How many functions are there from S to T?
(ii) How many 1-1 functions are there from S to T?
(iii) How many onto functions are there from S to T?
(iv) How many bijections are there from S to T?
Exercise 14. For which integers a does the function f : Z → Z given by f(x) = x + a
have an inverse? For each such a, what is the inverse?
Exercise 15. For which integers a does the function f : Z → Z given by f(x) = ax have
an inverse? For each such a, what is the inverse?
Exercise 16. Is {1, 2, 3} a partition of {1, 2, 3}?
Exercise 17. Find all partitions of {1, 2, 3}.

16
Answer Keys: Sets and Functions

1. Answer: No, because {{0}} is the set whose only element is {0}. And 0 6= {0}.
2. Answer: No, because the sets have different elements.
3. Answer: S ∪ ∅ = S. Many people wrote S = {A, B, {}}, but this is wrong; {} is not
an element of either S or {}.
4. Answer: S.
5. Answer: ∅.
6. Answer: S.
7. Answer: Yes
8. Answer: {{A, B}, {}}, {{A, B}}, {{}}, {}.
9. Answer: {(1, 1),(1, 2),(2, 1),(2, 2)}
10. Answer: S. (Some students assumed that S = {A, B}. Because of where the
question was placed, this was reasonable, though not what I intended, so I also
accepted the answer {A, B}.)
11. Answer: Most students got this wrong, at least in part. There are two functions.
One is given by f = j. The other is given by f(A) = B, f(B) = A.
12. Answer: the solutions are the set {z ∈ Z | f(z) = g(z)}.
13. (i) Answer: 9, because there are 3 options for where to map 1 to, and 3options
for where to map 2 to. 3 × 3 = 9.
(ii) Answer: 6. Of the 9 functions from (i), 3 are not 1-1.
(iii) Answer: 0
(iv) Answer: 0
14. Answer: All integers a. The inverse g : Z → Z is given by g(z) = z − a.
15. Answer: a ∈ {1, −1}. The inverse g : Z → Z is given by g(z) = z/a.
16. Answer: No. It is not even a set of subsets of {1, 2, 3}.
17. Answer: {{1, 2, 3}}, {{1}, {2}, {3}}, {{1, 2}, {3}}, {{1}, {2, 3}}, {{2}, {1, 3}}.

17
SPECIAL TYPES OF FUNCTIONS

JULIUS MARK M. PERTACORTA

18
Inverse and Composite Functions

Introduction to Inverse Functions


To find the inverse function, switch the x and y values, and then solve for y.

Learning Objectives
Calculate the formula of a function’s inverse by switching x and y and then
solving for y.

Key Points
An inverse function reverses the inputs and outputs.
To find the inverse formula of a function, write it in the form of y and x, switch
y and x, and then solve for y. Some functions have no inverse function, as a
function cannot have multiple outputs.

Key Terms
Inverse function: A function that does exactly the opposite of another.

Definition of Inverse Function


An inverse function, which is notated f−1(x), is defined as the inverse function
of f(x) if it consistently reverses the f(x) process. That is, if f(x) turns a into b,
then f−1(x) must turn b into a. More concisely and formally, f−1(x) is the
inverse function of f(x)

Below is a mapping of function f(x) and its inverse function, f−1(x).

Notice that the ordered pairs are reversed from the original function to its
inverse. Because f(x) maps a to 3, the inverse f−1(x) maps 3 back to a.

Thus, the graph of f−1(x) can be obtained from the graph of f(x) by switching
the positions of the x and y-axes. This is equivalent to reflecting the graph
across the line y=x, an increasing diagonal line through the origin.

Inverse functions:
mapping
representation: An
inverse function
reverses the inputs and
outputs.

Write the Inverse Function

In general, given a function, how do you find its inverse function? Remember
that an inverse function reverses the inputs and outputs. So, to find the
inverse function, switch the x and y values of a given function, and then solve
for y.

19
Inverse functions:

Graphic representation: The function graph (red) and its inverse function
graph (blue) are reflections of each other about the line y=x (dotted black
line). Notice that any ordered pair on the red curve has its reversed ordered
pair on the blue line. For example, (0,1) on the red (function) curve is reflected
over the line y=x and becomes (1,0) on the blue (inverse function) curve.
Where one curve is on the line y=x, the curves intersect, as a reflection over
the line leaves the point unchanged.

Example 1

Find the inverse of: f(x) = x2

a.: Write the function as: y = x2


b.: Switch the x and y variables: x = y2
c.: Solve for y, x = y2 ± √x = y

Since the function f(x) = x2 has multiple outputs, its inverse is not a function.
Notice the graphs in the picture below. Even though the blue curve is a
function (passes the vertical line test), its inverse would not be. The red curve
for the function f(x) = √x is not the full inverse of the function f(x) = x2

The inverse is not a function:

A function’s inverse may not always be a function. The function (blue)


f(x)=x2, includes the points (−1,1) and (1,1). Therefore, the inverse would
include the points: (1,−1) and (1,1) which the input value repeats, and
therefore is not a function. For f(x)=√x to be a function, it must be defined as
positive.

20
Example 2

Find the inverse function of: f(x) = 2x . As soon as the problem includes an
exponential function, we know that the logarithm reverses
exponentiation. The complex logarithm is the inverse function of the
exponential function applied to complex numbers.

Exponential and logarithm functions:

The graphs of y=2x (blue) and x=2y (red) are inverses of one another. The
black line represents the line of reflection, in which is y=x.Test to make sure
this solution fills the definition of an inverse function.

Composition of Functions and Decomposing a Function


Functional composition allows for the application of one function to another;
this step can be undone by using functional decomposition.

Learning Objectives
Practice functional composition by applying the rules of one function to the
results of another function.

21
Key Takeaways

Key Points:
Functional composition applies one function to the results of another.
Functional decomposition resolves a functional relationship into its constituent
parts so that the original function can be reconstructed from those parts by
functional composition.
Decomposition of a function into non-interacting components generally
permits more economical representations of the function.
The process of combining functions so that the output of one function
becomes the input of another is known as a composition of functions. The
resulting function is known as a composite function. We represent this
combination by the following notation: (f∘ g (x)=f (g(x) )
The domain of the composite function (f∘ g) is all x such that x is in the domain
of g and g(x) is in the domain of f.

Key Terms:
Codomain: The target space into which a function maps element of its
domain. It always contains the range of the function but can be
larger than the range if the function is not subjective.
Domain: The set of all points over which a function is defined.

Function Composition
The process of combining functions so that the output of one function
becomes the input of another is known as a composition of functions. The
resulting function is known as a composite function. We represent this
combination by this notation:

( f ∘ g ) (x) = f ( g (x) )

We read the left-hand side as “f “composed with g at x, and the right-hand
side as “f of g of x.” The two sides of the equation have the same
mathematical meaning and are equal. The open circle symbol, is called
the composition operator. Composition is a binary operation that takes two
functions and forms a new function, much as addition or multiplication takes
two numbers and gives a new number.

Function Composition and Evaluation


It is important to understand the order of operations in evaluating a composite
function. We follow the usual convention with parentheses by starting with the
innermost parentheses first, and then working to the outside.
In general, (f∘g) and (g∘f) are different functions. In other words, in many
cases f(g(x))≠g(f(x)) for all x.
Note that the range of the inside function (the first function to be evaluated)
needs to be within the domain of the outside function. Less formally, the
composition has to make sense in terms of inputs and outputs.

Evaluating Composite Functions Using Input Values

22
When evaluating a composite function where we have either created or been
given formulas, the rule of working from the inside out remains the same. The
input value to the outer function will be the output of the inner function, which
may be a numerical value, a variable name, or a more complicated
expression.

Example

If f(x)=−2x and g(x)=x2−1, evaluate f (g(3) ) and g (f(3) )

To evaluate f (g(3) ), first substitute, or input the value of 3 into g(x) and find
the output. Then substitute that value into the f(x) function, and simplify:

g(3) = (3) 2−1 = 9−1 = 8

f(8) = −2 (8 )= −16 . Therefore, f (g (3)) = −16

To evaluate g (f (3)), find f (3) and then use that output value as the input
value into the g(x) function:

f (3) =−2(3) = −6
g(−6) = (−6)2 −1 = 36 −1 = 35 . Therefore, g(f(3))=35

Evaluating Composite Functions Using a Formula


While we can compose the functions for each individual input value, it is
sometimes helpful to find a single formula that will calculate the result of a
composition f(g(x)) or g(f(x)). To do this, we will extend our idea of function
evaluation.
In the next example we are given a formula for two composite functions and
asked to evaluate the function. Evaluate the inside function using the input
value or variable provided. Use the resulting output as the input to the
outside function.

Example
If f(x)=−2x and g(x)=x2−1, evaluate f(g(x)) and g(f(x)).
First substitute, or input the function g(x), x2−1 into the f(x) function, and then
simplify:
f(g(x))=−2(x2−1)
f(g(x))=−2x2+2
For g(f(x)), input the f(x) function, −2x into the g(x) function, and then simplify:
g(f(x))=(−2x)2−1
g(f(x))=4x2−1

Functional Decomposition
Functional decomposition broadly refers to the process of resolving a
functional relationship into its constituent parts in such a way that the original
function can be reconstructed (i.e., recomposed) from those parts by function

23
composition. In general, this process of decomposition is undertaken either for
the purpose of gaining insight into the identity of the constituent components
(which may reflect individual physical processes of interest), or for the
purpose of obtaining a compressed representation of the global function; a
task which is feasible only when the constituent processes possess a certain
level of modularity (i.e., independence or non-interaction).

In general, functional decompositions are worthwhile when there is a certain


“sparseness” in the dependency structure; i.e. when constituent functions are
found to depend on approximately disjointed sets of variables. Also,
decomposition of a function into non-interacting components generally permits
more economical representations of the function.

Restricting Domains to Find Inverses


Domain restriction is important for inverse functions of exponents and
logarithms because sometimes we need to find an unique inverse.

Learning Objectives
Determine inverses of functions by restricting their domains

Key Takeaways

Key Points:
f−1(x) is defined as the inverse function of f(x) if it consistently reverses the
f(x) process.
f (x)=x2, without any domain restriction, does not have an inverse function, as
it fails the horizontal line test.

Key Terms:
domain: The set of points over which a function is defined.

Inverse Functions

f−1(x) is defined as the inverse function of f(x) if it consistently reverses the


f(x) process. That is, if f(x) turns a into b, then f−1x must turn b into a. More
concisely and formally, f−1x is the inverse function of f(x) if f(f−1(x))=x.

Inverse functions’ domain and range:


If f maps X to Y, then f−1 maps Y back to X.

Domain Restrictions: Parabola

Informally, a restriction of a function is the result of trimming its domain.


Remember that:

If f maps X to Y, then f−1 maps Y back to X. This is not true of the function
f(x)=x2.Without any domain restriction, f(x)=x2 does not have an inverse

24
function as it fails the horizontal line test. But if we restrict the domain to be
x>0

then we find that it passes the horizontal line test and therefore has an inverse
function. Below is the graph of the parabola and its “inverse.” Notice that the
parabola does not have a “true” inverse because the original function fails the
horizontal line test and must have a restricted
domain to have an inverse.

Failure of horizontal line test:


Graph of a parabola with the equation y=x2, the
U-Shaped curve opening up. This function fails
the horizontal line test, and therefore does not
have an inverse. The inverse equation,
y=√x (other graph) only includes the positive
input values of the parabola’s domain.
However, if we restrict the domain to be x>0,
then we find that it passes the horizontal line
test and will match the inverse function.

Domain Restriction: Exponential and Logarithmic


Functions

Domain restriction is important for inverse functions of exponents and


logarithms because sometimes we need to find a unique inverse. The inverse
of an exponential function is a logarithmic function, and the inverse of a
logarithmic function is an exponential function.

Example 1

Is x = 0 in the domain of the function f(x)=log(x)? If so, what is the value of
the function when x=0? Verify the result.

No, the function has no defined value for x = 0. To verify, suppose x = 0 is in
the domain of the function f(x) = log (x). Then there is some number n such
that n = log (0). Rewriting as an exponential equation gives: 10n = 0, which is
impossible since no such real number n exists. Therefore x = 0 is not in the
domain of the function f(x) = log(x).

Inverses of Composite Functions


A composite function represents, in one function, the results of an entire chain
of dependent functions.

Learning Objectives
Solve for the inverse of a composite function

Key Takeaways

25
Key Points:
The composition of functions is always associative. That is, if f, g, and h are
three functions with suitably chosen domains and co-domains, then f∘ (g∘ h) =
(f∘ g)∘ h.
Functions can be inverted and then composed, giving the notation of: ( f′∘ g′)
(x).
Functions can be composed and then inverted, yielding the following notation:
(f ∘ g)′ (x)

Key Terms
Composite function: A function of one or more independent variables,
at least one of which is itself a function of one or
more other independent variables; a function of a
function.

Composition and Composite Functions


In mathematics, function composition is the application of one function to the
results of another.
The functions g and f are said to commute with each other if g ∘ f = f ∘ g. In
general, the composition of functions will not be commutative.

A composite function represents in one function the results of an entire chain


of dependent functions. For example, if a school becomes larger, the supply
of food in the cafeteria must become larger. This would entail ordering more
sandwiches, which means ordering more ingredients, drinks, plates, etc. The
entire chain of dependent functions are the ingredients, drinks, plates, etc.,
and the one composite function would be putting the entire chain together in
order to calculate a larger population at the school.

Inverses and Composition

Let’s go through the relationship between inverses and composition in this


example. let’s take two functions, compose and invert them.

Find: (f∘g)′(x)(f∘g)′(x) ,f(x)=x2, g(x)=x+1

First, compose them by applying the function f to the result of applying


function g:

(f ∘ g) (x) = f (x+1) = (x+1)2y = (x+1)2√y = x+1√y−1 = x

Then, invert it by switching the x and y variables:


√x−1 = y√x −1 = (f ∘ g)′ (x).

ANOTHER SPECIAL FUNCTIONS


You can think of a function as a way of matching the members of a set "A" to
a set "B":

26
 Injective means that every member of "A" has its own unique matching
member in "B". You won't get two "A"s pointing to one "B", but you could have
a "B" without a matching "A"

 Surjective means that every "B" has at least one matching "A" (maybe more
than one).
 Bijective means Injective and Surjective together. So there is a perfect "one-
to-one correspondence" between the members of the sets.

Formal Definitions

Injective

 A function f is injective if and only if whenever f(x) = f(y), x = y.

Theorem:
If f: A→B and g:B→C are injective functions, then g∘f:A→C is injective
also.

Proof:
Suppose g (f (a)) =g (f (a′))
Since g is injective, f (a) = f(a′). Since f is injective, a=a′. Thus, (g ∘ f)
(a) = (g ∘ f) (a′) implies a=a′, so (g ∘ f) is injective.

Theorem:
Suppose f1, f2:A→B, g:B→C, h1,h2:C→D are functions.
a) If g is injective and g∘f1=g∘f2 then f1=f2
. b) If g is surjective and h1∘g=h2∘g then h1=h2.

Proof.
We prove part (b), leaving part (a) as an exercise. Suppose c ∈C. We
wish to show h1(c) =h2(c). By hypothesis g is surjective, so there is a b
∈B such that g (b)=c. So
h1(c)=h1(g(b))=(h1∘g)(b)=(h2∘g)(b)=h2(g(b))=h2(c),
as desired.

Example 1:
f(x) = x+5 from the set of real numbers  to  is an injective function.

27
This function can be easily reversed. for example:
f(3) = 8
Given 8 we can go back to 3
Example 2:
f(x) = x2 from the set of real numbers  to  is not an injective function
because of this kind of thing:
f(2) = 4 and
f(-2) = 4
This is against the definition f(x) = f(y), x = y, because f(2) = f(-2) but 2
≠ -2
In other words there are two values of "A" that point to one "B", and
this function could not be reversed (given the value "4" ... what
produced it?)
BUT if we made it from the set of natural numbers then it is injective,
because:
f(2) = 4 there is no f(-2), because -2 is not a natural number.

Note: Injective is also called "One-to-One", but this can be confusing


because it makes it sound like it is actually bijective which has "one-to-one
correspondence".

Surjective (Also Called "Onto")

A function f (from set A to B) is surjective if and only for every y in B, there is at


least one x in A such that f(x) = y, in other words f is surjective if and only if f(A)
= B.
So, every element of the range corresponds to at least one member of the domain.

Theorem:
Suppose f:A→B and g:B→C are functions.

a) If g∘f is injective then f is injective.


b) If g∘f is surjective then g is surjective.

Proof.
We prove part (a), leaving part (b) as an exercise. Suppose a,a ′∈ A
and f(a) = f(a′). We wish to prove a = a′. We have (g ∘ f) (a) = g (f (a) ) = g (f (a′) ) = (g
∘ f) (a′) and since g ∘ f is injective, we conclude that,
A = a′, as desired.
Bijective

A function f (from set A to B) is bijective if, for every y in B, there is exactly


one x in A such that f(x) = y

28
Alternatively, f is bijective if it is a one-to-one correspondence between
those sets, in other words both injective and surjective.

Theorem:
A function f:A→B has an inverse if and only if it is bijective.

Proof:
Suppose g is an inverse for f (we are proving the implication ⇒). Since g∘f=iA
is injective, so is f (by 4.4.1(a)). Since f∘g=iB is surjective, so is f (by 4.4.1(b)).
Therefore, f is injective and surjective, that is, bijective.

Conversely, suppose f is bijective. Let g: B→A be a pseudo-inverse to f. From


the proof of theorem, we know that since f is surjective, f ∘ g = iB, and since f
is injective, g ∘ f = iA.
We have talked about "an'' inverse of f, but really there is only one.

Example:
The function f(x) = x2 from the set of positive real numbers to positive real
numbers is injective and surjective. Thus it is also bijective.

But not from the set of real numbers  because you could have, for example,
both
f(2)=4 and f(-2)=4.

29
EXERCISES

1.) Let 𝐴 and 𝐵 be two nonempty sets where 𝐴 = {1 ,2,3,4} and 𝐵 = {𝑎,𝑏,𝑐}. Consider
each of the following relations: 𝑇 = {(1,𝑎),(2,𝑏),(2,𝑐),(3,𝑐),(4,𝑏)} 𝑈 =
{(1,𝑎),(2,𝑏),(4,𝑏)} 𝑉 = {(1,𝑎),(2,𝑏),(3,𝑐),(4,𝑏)} Which of these relations (𝑇,𝑈 and
𝑉) qualify as functions?

2.) (i) Although the relation 𝑉 in Q1 is a function, it is not a one-to-one (or


injective) function. Why?
(ii) 𝑉 is an onto (surjective) function. Why?
(iii) Does the function 𝑓, defined by the relation 𝑉, have an inverse?

3.) For each of the relations {𝑄,,𝑆,𝑇,𝑈,𝑉} below, determine whether the relation is a
function. If the relation is a function, determine whether the function is injective
and/or surjective.
(i) 𝐴 = {1,2,3}, 𝐵 = {𝑎,𝑏,𝑐,𝑑} 𝑄 = {(1,𝑎),(2,𝑑),(3,𝑏)}
(ii) 𝐴 = {1,2,3}, 𝐵 = {𝑎,𝑏,𝑐} 𝑅 = {(1,𝑎),(2,𝑏),(3,𝑐)}
(iii) 𝐴 = {1,2,3}, 𝐵 = {𝑎,𝑏,𝑐} 𝑆 = {(1,𝑎),(2,𝑏),(3,𝑏)} (iv) 𝐴 = {1,2,3}, 𝐵=
{𝑎,𝑏,𝑐,𝑑}
𝑇 = {(1,𝑎),(2,𝑏),(2,𝑐),(3,𝑑)} (v) 𝐴 = {1,2,3}, 𝐵 = {𝑎,𝑏} 𝑈 =
{(1,𝑎),(2,𝑏),(3,𝑏)}
(vi) 𝐴 = {1,2,3}, 𝐵 = {𝑎,𝑏} 𝑉 = {(1,𝑎),(2,𝑏)}

4.) (i) Which of the relations in Q3 is a bijection?


(ii) For the relation that is a bijection, write down the elements of the inverse
function.

5.) The function 𝒇 is defined by: 𝒇:ℝ → ℝ:𝒙 ↦ 𝒙𝟐 +𝟐.


(i) Give an example to show that 𝑓 is not injective.
(ii) Give an example to show that 𝑓 is not surjective.

6.) The function 𝒇 is defined by: 𝒇:ℝ → ℝ:𝒙 ↦ 𝒙𝟐 −𝟔𝒙.


(i) Give an example to show that 𝑓 is not injective.
(ii) Give an example to show that 𝑓 is not surjective

7.) For each of the functions below determine which of the properties hold, injective,
surjective, bijective. Briefly explain your reasoning.
(i) The function 𝑓:ℝ → ℝ defined by (𝑥) = 𝑒𝑥. (ii) The function 𝑓:ℝ → ℝ+
defined by (𝑥) = 𝑒𝑥. (iii) The function 𝑓:ℝ → ℝ defined by (𝑥) = (𝑥 + 1)𝑥(𝑥 − 1).
(iv) The function 𝑓:ℝ → ℝ defined by (𝑥) = (𝑥2 −9)(𝑥2 − 4)

8.) Which of the following cubic functions have an inverse? [Hint: Finding the
derivative of the function may help!] (i) (𝑥) = 𝑥3 −6𝑥2 +3𝑥 +7, (ii) 𝑓(𝑥) = −𝑥3 −6𝑥2
−13𝑥 +4, (iii) 𝑓(𝑥) = 𝑥3 +3𝑥2 +4𝑥 +3, (iv) 𝑓(𝑥) = −𝑥3 +3𝑥2 −𝑥 −1

30
ANSWERS KEY

1.) Relation 𝑇, maps 2 (∈ 𝐴) to both 𝑏 and 𝑐 (∈ 𝐵). This violates condition 2 of the
definition.
Relation 𝑇, is not a function.
Relation 𝑈 is not defined for all elements of 𝐴.This violates condition 1 of the
definition. Relation 𝑈 is not a function.
Relation satisfies both conditions of the definition of a function.
Relation 𝑉 is a function.
If we call the function 𝑓 we have (1) = 𝑎,𝑓(2) = 𝑏,𝑓(3) = 𝑐 and 𝑓(4) = 𝑏.

2.) (i) (2) = 𝑓(4) = 𝑏, but 2 ≠ 4.


(ii) The range of 𝑓 is equal to the set 𝐵 (the codomain)
(iii) No, a function must be both injective and surjective to have an inverse.

3.) (i) The relation is a function. The function is injective. The function is not
surjective since 𝑐 is not an element of the range.
(ii) The relation is a function.The function is both injective and surjective.
(iii) The relation is a function.The function is not injective since (2) = 𝑓(3) but 2
≠ 3.The function is not surjective since 𝑐 is not an element of the range.
(iv) The relation is a not a function since the relation is not uniquely defined for
2.
(v) The relation is a function. The function is not injective since (2) = 𝑓(3) but 2
≠ 3. The function is surjective.
(vi) The relation is a not a function since the relation is not defined for 2.

4.) (i) Part (ii) 𝑓 = {(1,),(2,𝑏),(3,𝑐)}


(ii) 𝑓−1 = {(𝑎,1),(𝑏,2),(𝑐,3)}

5.) (i) 𝑓(−1) = 𝑓(1) = 3 but −1 ≠ 1, therefore the function is not injective.
(ii) There is no real number, 𝑥 such that (𝑥) = 1 therefore the function is not
surjective. Or the range of the function is 𝑦 ≥ 2. The range of the
function is not ℝ (the codomain) therefore the function is not surjective.

6.) (i) 𝑓(6) = 𝑓(0) = 0 but 6 ≠ 0, therefore the function is not injective.
(ii) (𝑥) = (𝑥 −3)2 −9 [by completing the square] There is no real number, 𝑥
such that 𝑓(𝑥) = −10 the function is not surjective. Or the range of the function
is 𝑦 ≥ 2. The range of the function is not ℝ (the codomain) therefore the function
is not surjective.

7.) (i) This function is injective, since 𝑒𝑥 takes on each nonnegative real value for
exactly one 𝑥. However, the function is not surjective, because 𝑒𝑥
never takes on negative values. Therefore, the function is not
bijective either.
(ii) The function 𝑒𝑥 takes on every nonnegative value for exactly one 𝑥, so it is
injective, surjective, and bijective.
(iii) This function is surjective, since it is continuous, it tends to +∞ for large
positive 𝑥, and tends to −∞ for large negative 𝑥. The function takes on
each real value for at least one 𝑥. However, this function is not

31
injective, since it takes on the value 0 at 𝑥 = −1, 𝑥 = 0 and 𝑥= 1.
Therefore, the function is not bijective either.
(iv) This function is not surjective, it tends to +∞ for large positive 𝑥, and also
tends to +∞ for large negative 𝑥. Also this function is not injective,
since it takes on the value 0 at 𝑥 = 3,
𝑥 = −3, 𝑥 = 4 and 𝑥 = −4. Therefore, the function is not bijective either.

8.) All cubic functions are surjective by their nature. So we check injective by seeing
if the function is always increasing or decreasing
(i) 𝑓′(𝑥) = 3(𝑥 −2)2 −9 this function is not always increasing or decreasing,
hence not injective
(ii) 𝑓′(𝑥) = −[3(𝑥 +2)2 +1] this function is always decreasing
(iii) 𝑓′(𝑥) = 3(𝑥 +2)2 +1 this function is always increasing
(iv) 𝑓′(𝑥) = −3(𝑥 −1)2 +2 this function is always increasing

32
MATHEMATICAL INDUCTION

SHAIRA S. BUCIO

33
Mathematical Induction

Definition:
This is a method of "pulling oneself up by one's bootstraps" and is
regarded with suspicion by non-mathematicians.
Example:
1) Suppose we want to sum an Arithmetic Progression: 1 + 2 +
1
3 +. . . + 𝑛 = 2 𝑛(𝑛 + 1).
Proof:
We define a sequence of "propositions" 𝑃(1), 𝑃(2), ... where 𝑃(𝑛) is "1 +
1
2 + 3 +. . . + 𝑛 = 2 𝑛(𝑛 + 1)"
First, we'll prove 𝑃(1); this is called "anchoring the induction".
Then we will prove that if 𝑷(𝒌) is true for some value of k, then so is
𝑃(𝑘 + 1) ; this is called "the inductive step".
If 𝑃(1) is OK, then we can use this to deduce that 𝑃(2) is true and then
use this to show that 𝑃(3) is true and so on.
So, if 𝑛 is the first value for which the result is false, then 𝑃(𝑛 – 1) is true
and we would get a contradiction.
So, let's look hard at the above example.
1
𝑃(1) is certainly OK: 1 = × 1 × 2.
2
Now suppose that 𝑃(𝑘) is true for some value of k.
Then try and prove 𝑃(𝑘 + 1):
1
Now 1 + 2 + 3 + . . . + 𝑘 + (𝑘 + 1) = 𝑘( 𝑘 + 1) + (𝑘 + 1) (using
2
𝑃(𝑘), which we are allowed to assume).
1 1
But this simplifies to (𝑘 + 1)( 2 𝑘 + 1) = (𝑘 + 1)(𝑘 + 2) and this is
2
exactly what 𝑃(𝑘 + 1) says.
Hence the result is true for all values.∎

2) Summing a Geometric Progression


Let r be a fixed real number.
1−𝑟 𝑛+1
Then 1 + 𝑟 + 𝑟 2 + 𝑟 3 + . . . + 𝑟 𝑛 = . This is 𝑃(𝑛).
1−𝑟
Proof:
Clearly 𝑃(0) is true. (Note that we can anchor the induction where we
like.
So, we suppose that 𝑃(𝑘) is true and we'll try and prove 𝑃(𝑘 + 1).
So, look at 1 + 𝑟 + 𝑟 2 + 𝑟 3 + . . . + 𝑟 𝑘 + 𝑟 𝑘+1 .
1−𝑟 𝑘+1
By 𝑃(𝑘) the term in brackets is and so we can simplify this to
1−𝑟
1−𝑟 𝑘+1 1−𝑟 𝑘+1 −𝑟 𝑘+1 1−𝑟 𝑘+2
1−𝑟
+ 𝑟 𝑘+1 = 1−𝑟
= 1−𝑟
which is what 𝑃(𝑘 + 1) predicts.∎

34
Theorem:
Mathematical induction ⇒ the well-ordering property.

Proof:

We prove this by contradiction.


Suppose that the well-ordering property were false.
Let 𝑆 be a counterexample: a nonempty set of nonnegative integers that
contains no smallest element.
Let 𝑃(𝑛) be the statement “𝑖 ∉ 𝑆 for all 𝑖 ≤ 𝑛.”
We will show that 𝑃(𝑛) is true for all n (which will contradict the assertion
that 𝑆 is nonempty).
Now 𝑃(0) must be true, because if 0 ∈ 𝑆 then clearly S would have a
smallest element, namely 0.
Suppose now that 𝑃(𝑛) is true, so that 𝑖 ∉ 𝑆 for all 𝑖 = 0, 1, 2, . . . , 𝑛.
We must show that 𝑃(𝑛 + 1) is true, which amounts to showing that
𝑛 + 1 ∉ 𝑆. If 𝑛 + 1 ∈ 𝑆, then 𝑛 + 1 would be the smallest element of
S, and this would contradict our assumption.
Therefore 𝑛 + 1 ∉ 𝑆.
Thus, we have shown by the principle of mathematical induction that
𝑃(𝑛) is true for all 𝑛, which means that there can be no elements of S.
This contradicts our assumption that 𝑆 = ∅ , and our proof by
contradiction is complete.∎

Theorem:
The well-ordering property ⇒ strong induction.

Proof:

To show that strong induction is valid;


Let us suppose that we have a proposition ∀𝑛, 𝑃(𝑛) that has been
proved using it.
We must show that in fact ∀𝑛, 𝑃(𝑛) is true (to say that a principle of proof
is valid means that it proves only true propositions).
Let 𝑆 be the set of counterexamples, 𝑖. 𝑒, 𝑆 = {𝑛 | ¬𝑃(𝑛)}.
We want to show that 𝑆 = ∅ .
We argue by contradiction.
Assume that 𝑆 ≠ ∅.
Then by the well-ordering property, 𝑆 has a smallest element.
Since part of the method of strong induction is to show that 𝑃(1) is true,
this smallest counterexample must be greater than 1.
Let us call it 𝑘 + 1.

35
Since 𝑘 + 1 is the smallest element of 𝑆, it must be the case that
𝑃(1) ∧ 𝑃(2) ∧ · · · ∧ 𝑃(𝑘) is true.
But the rest of the proof using strong induction involved showing
that 𝑃(1) ∧ 𝑃(2) ∧ · · · ∧ 𝑃(𝑘) implied 𝑃(𝑘 + 1) ; therefore since the
hypothesis is true, the conclusion must be true as well, i.e., 𝑃(𝑘 + 1) is
true.
This contradicts our assumption that 𝑘 + 1 ∈ 𝑆.
Therefore, we conclude that 𝑆 = ∅, so ∀𝑛, 𝑃(𝑛) is true. ■

Theorem:
Strong induction ⇔ mathematical induction

Proof:
If one has shown that 𝑃(𝑘) → 𝑃(𝑘 + 1), then it automatically follows
that [𝑃(1) ∧ · · · ∧ 𝑃(𝑘)] → 𝑃(𝑘 + 1).
In other words, ordinary induction ⇒ strong induction.
Conversely, suppose that 𝑃(𝑛) is a statement that one can prove using
strong induction.
Let 𝑄(𝑛) be 𝑃(1) ∧ · · · ∧ 𝑃(𝑛).
Clearly ∀𝑛, 𝑃(𝑛) is logically equivalent to ∀𝑛, 𝑄(𝑛).
We show how ∀𝑛, 𝑄(𝑛) can be proved using ordinary induction.
First, 𝑄(1) is true because 𝑄(1) = 𝑃(1) and 𝑃(1) is true by the basis
step for the proof of ∀𝑛, 𝑃(𝑛) by strong induction.
Now suppose that 𝑄(𝑘) is true, i.e., 𝑃(1) ∧ · · · ∧ 𝑃(𝑘) is true.
By the proof of ∀𝑛, 𝑃(𝑛) by strong induction it follows that 𝑃(𝑘 + 1) is
true. But 𝑄(𝑘) ∧ 𝑃(𝑘 + 1) is just 𝑄(𝑘 + 1).
Thus, we have proved ∀𝑛, 𝑄(𝑛) by ordinary induction.∎

Theorems above show that the well-ordering property, the principle of


mathematical induction, and strong induction are all equivalent.

Well Ordering Property

Definition:
The well-ordering principle is a concept which is equivalent to
mathematical induction. In your textbook, there is a proof for how the
well-ordering principle implies the validity of mathematical induction.
However, because of the very way in which we constructed the set of
natural numbers and its arithmetic, we deduced, in class, the validity of
mathematical induction directly from the axioms of set theory. In this
note, we show how mathematical induction, in turn, implies the well-
ordering principle.

36
Example:
1) The set ℕ is well-ordered.
The following sets are well-ordered:

(1) ℕ ∪ {0}
(2) ℕ ∪ {−1, 0}
(3) ℕ ∪ {−3, −2, −1}
(4) {𝑛 ∈ ℕ ∶ 𝑛 > 5}

The following sets are NOT well-ordered:

(1) ℝ (the open interval (0, 2) is a non-empty subset of ℝ but it has no


smallest element)
(2) ℤ (the set of negative integers is a non-empty subset of Z but with
no smallest element)
(3) the interval [0, 1] (because (0, 1) is a non-empty subset of [0, 1]
without smallest element)

Theorem:
Every non-empty subset of the natural numbers has a least element.
Proof:
Let A be a non-empty subset of ℕ.
We wish to show that A has a least element, that is, that there is an
element 𝑎 ∈ 𝐴 such that 𝑎 ≤ 𝑛 for all 𝑛 ∈ 𝐴. We will do this by strong
induction on the following predicate:
𝑃(𝑛): “If 𝑛 ∈ 𝐴, then 𝐴 has a least element.”
Basic Step: 𝑃(0) is clearly true, since 0 ≤ 𝑛 for all 𝑛 ∈ 𝑁.
Strong Inductive Step: We want to show that [𝑃(0) ∧ 𝑃(1) ∧···∧ 𝑃(𝑛)] →
𝑃(𝑛 + 1).
To this end, suppose that 𝑃(0), 𝑃(1),···, 𝑃(𝑛) are all true and that 𝑛 +
1 ∈ 𝐴.
We consider two cases.
CASE 1: ¬∃𝑚(𝑚 ∈ 𝐴 ∧ 𝑚 < 𝑛 + 1).
In this case, n + 1 is the least element of A.
CASE 2: ∃𝑚(𝑚 ∈ 𝐴 ∧ 𝑚 < 𝑛 + 1).
In this case, since P(m) is true, A has a least element.
Either way, we conclude that 𝑃(𝑛 + 1) is true.
So, by strong mathematical induction, we obtain that 𝑃(𝑛) is true for all
𝑛 ∈ 𝑁.
Since A is not empty, we can pick an 𝑛 ∈ 𝐴.
Moreover, since 𝑃(𝑛) is true, this implies that A has a least element.∎

37
Definition:
The validity of both the principle of mathematical induction and strong
induction follows from a fundamental axiom of the set of integers, the
well-ordering property. The well-ordering property states that every
nonempty set of nonnegative integers has a least element. We will show
how we can directly use the well-ordering property in proofs.

The Well-Ordering Property is an axiom about ℤ + that we assume to be


true. Following four axioms about Z +:
1. The number 1 is a positive integer.
2. If n ∈ Z +, then n + 1, the successor of n, is also a positive integer.
3. Every positive integer other than 1 is the successor of a positive
integer.
4. (The Well-Ordering Property) Every nonempty subset of the set of
positive integers has a least element.

Theorem:
(The Division Algorithm) Let 𝑎 be an integer and 𝑑 a positive integer.
Then there are unique integers 𝑞 and 𝑟, with 0 ≤ 𝑟 < 𝑑, such that 𝑎 =
𝑑𝑞 + 𝑟.

Proof:
Let 𝑆 be the set of nonnegative integers of the form 𝑎 − 𝑑𝑞, where 𝑞 is
an integer.
This set is nonempty because −𝑑𝑞 can be made as large as desired
(taking 𝑞 to be a negative integer with large absolute value).
By the well-ordering property, S has a least element 𝑟 = 𝑎 − 𝑑𝑞0 . The
integer r is nonnegative. It is also the case that 𝑟 < 𝑑.
If it were not, then there would be a smaller nonnegative element in 𝑆,
namely, 𝑎 − 𝑑(𝑞0 + 1). To see this, suppose that 𝑟 ≥ 𝑑. Because 𝑎 =
𝑑𝑞0 + 𝑟, it follows that 𝑎 − 𝑑(𝑞0 + 1) = (𝑎 − 𝑑𝑞0 ) − 𝑑 = 𝑟 − 𝑑 ≥ 0.
Consequently, there are integers q and r with 0 ≤ 𝑟 < 𝑑.
The proof that 𝑞 and 𝑟 are unique:
Suppose that we have two such pairs, say (𝑞, 𝑟) and (𝑞′, 𝑟′ ), so that 𝑎 =
𝑑𝑞 + 𝑟 = 𝑑𝑞′ + 𝑟′ , with 0 ≤ 𝑟, 𝑟′ < 𝑑. We will show that the pairs are
really the same, that is, 𝑞 = 𝑞′ and 𝑟 = 𝑟′ .
𝑑𝑞 + 𝑟 = 𝑑𝑞′ + 𝑟′ ⇒ 𝑑(𝑞 − 𝑞′ ) = 𝑟′ − 𝑟.
Therefore 𝑑 | (𝑟′ − 𝑟).
(0 ≤ 𝑟 < 𝑑) ∧ (0 ≤ 𝑟′ < 𝑑) ⇒ |𝑟′ − 𝑟| < 𝑑.
The only multiple of d in that range is 0, so we are forced to
conclude that 𝑟 ′ = 𝑟.
To show that 𝑞 = 𝑞′ :
𝑎 − 𝑟 𝑎 − 𝑟′
𝑞 = = = 𝑞′ . ∎
𝑑 𝑑

38
Principle of Mathematical Induction

Definition:
A mathematical technique which is used to prove a statement, a formula
or a theorem is true for every natural number.

The technique involves two steps to prove a statement, as stated below-


Step 1(Base step) − It proves that a statement is true for the initial value.
Step 2(Inductive step) − It proves that if the statement is true for the
nth iteration (or number n), then it is also true for (𝑛 + 1)𝑡ℎ iteration (or
number n+1).
Example:
1. 3𝑛 − 1 is a multiple of 2 for n = 1, 2, ...
Solution:
Step 1 − For 𝑛 = 1, 31 − 1 = 3 − 1 = 2 which is a multiple of 2
Step 2 − Let us assume3𝑛 − 1 is true for 𝑛 = 𝑘 , Hence, 3𝑘 − 1 is true
(It is an assumption)
We have to prove that 3𝑘+1 − 1 is also a multiple of 2

3𝑘+1 − 1 = 3 × 3𝑘 − 1 = (2 × 3𝑘 ) + (3𝑘 − 1)
The first part (2 × 3𝑘 ) is certain to be a multiple of @ and the second
part (3𝑘 − 1) is also true as our previous assumption.
Hence, 3𝑘+1 − 1 is a multiple of 2.
So, it is proved that 3𝑛 − 1 is a multiple of 2.∎

2. 1 + 3 + 5 + ⋯ + (2𝑛 − 1) = 𝑛2 for 𝑛 = 1,2, . . ..

Solution:
Step 1 − For 𝑛 = 1,1 = 12 , Hence, step 1 is satisfied.
Step 2 − Let us assume the statement is true for 𝑛 = 𝑘.
Hence, 1 + 3 + 5 + ⋯ + (2(𝑘 + 1) − 1) = (𝑘 + 1)2 also holds
1 + 3 + 5 + ⋯ + (2(𝑘 + 1) − 1 )
= 1 + 3 + 5 + ⋯ + (2𝑘 + 2 − 1)
= 1 + 3 + 5 + ⋯ + (2𝑘 + 1)
= 1 + 3 + 5 + ⋯ + (2𝑘 − 1) + (2𝑘 + 1)
= 𝑘 2 + (2𝑘 + 1)
= (𝑘 + 1)2
So, 1 + 3 + 5 + ⋯ + (2(𝑘 + 1) − 1) = (𝑘 + 1)2 hold which
satisfies the step 2.
Hence, 1 + 3 + 5+. . . +(2𝑛 − 1) = 𝑛2 is proved.∎

Definition:

39
Let 𝑃(𝑛) be a given statement involving the natural number n such that
(𝑖) The statement is true for 𝑛 = 1, i.e., 𝑃(1) is true (or true for any fixed
natural number) and
(𝑖𝑖) If the statement is true for 𝑛 = 𝑘 (where k is a particular but
arbitrary natural number), then the statement is also true for 𝑛 = 𝑘 + 1,
i.e., truth of 𝑃(𝑘) implies the truth of 𝑃(𝑘 + 1). Then 𝑃(𝑛) is true for all-
natural number n.

Weak and Strong Induction

Definition:
The difference between weak induction and strong induction only
appears in induction hypothesis. In weak induction, we only assume that
statement holds at k-th step, while in strong induction, we assume that
the statement holds at all the steps from the base case to k-th step.

Example:
(Weak Induction)
Prove the following statement is true for all integers 𝑛. The statement
𝑃(𝑛) can be expressed as below:

𝑛
𝑛(𝑛 + 1)
∑𝑖 =
2
𝑖=1
1) Base Case: Prove that the statement holds when n=1

1
1(1 + 1)
∑𝑖 = =1
2
𝑖=1
2) Induction Hypothesis: Assume that the statement holds when 𝑛 = 𝑘

𝑘
𝑘(𝑘 + 1)
∑𝑖 =
2
𝑖=1
3) Inductive Step: Prove that the statement holds when 𝑛 = 𝑘 + 1 using
the assumption above.
We already assumed the following fact:

𝑘
𝑘(𝑘 + 1)
∑𝑖 =
2
𝑖=1
Now we would like to consider the case of finding:

40
𝑘

∑𝑖
𝑖=1
We can interpret the formula above in this way:

𝑘+1 𝑘

∑ 𝑖 = ∑ 𝑖 + (𝑘 + 1)
𝑖=1 𝑖=1
From the induction hypothesis above, the right hand side of the
equation above can be written as:

𝑘+1 𝑘
𝑘(𝑘 + 1)
∑ 𝑖 = ∑ 𝑖 + (𝑘 + 1) = + (𝑘 + 1)
2
𝑖=1 𝑖=1
We can simplify this re-formatted right-hand side:

𝑘+1
𝑘(𝑘 + 1) (𝑘 + 1)(𝑘 + 2)
∑𝑖 = + (𝑘 + 1) =
2 2
𝑖=1
Now, we can conclude that when 𝑖 = 𝑘 + 1, the proposition we are
trying to prove holds, because under this case, 𝑛 = 𝑘 + 1 and
𝑛 + 1 = 𝑘 + 2.
Therefore, the proposition holds for all integers n.

(Strong Induction)

Prove by induction that every integer greater than or equal to 2 can be


factored into primes. The statement P(n) is that an integer n greater
than or equal to 2 can be factored into primes.

1) Base Case: Prove that the statement holds when 𝑛 = 2 We are


proving 𝑃(2). 2 itself is a prime number, so the prime factorization of
2 is 2. Trivially, the statement 𝑃(2) holds.
2) Induction Hypothesis: Assume that for all integers less than or equal
to k, the statement holds.
Note: In the previous example, the assumption was only about the
case when n = k.
3) Inductive Step: Consider the number 𝑘 + 1.
Case 1: 𝑘 + 1 is a prime number.
When 𝑘 + 1 is a prime number, the number is a prime factorization
of itself.
Therefore, the statement 𝑃(𝑘 + 1) holds.

41
Case 2: 𝑘 + 1 is not a prime number.
We know that 𝑘 + 1 is a composite, so 𝑘 + 1 = 𝑝 × 𝑞(𝑝, 𝑞 ∈ 𝑍 + ).
Intuitively, we can conclude that p and q are less than or equal to
𝑘 + 1.
From the induction hypothesis stated above, for all integers less than
or equal to k, the statement holds, which means both p and q can be
expressed as prime factorizations.
In this sense, because 𝑘 + 1 is a product of p and q, by multiplying
the prime factorizations of p and q, we can get the prime factorization
for 𝑘 + 1 as well.
Therefore, the statement that every integer greater than or equal to
2 can be factored into primes holds for all such integers.
The common mistake in this question was to prove the Case 2 in the
inductive step without using induction hypothesis by dividing the
cases further into even number and odd number, etc.
It works but does not fit into the notion of inductive proof that we
wanted you to learn.
For inductive step in inductive proof, you must reason your argument
based on induction hypothesis you yourself state.
Definition:
The strong induction principle says that you can prove a statement of
the form:
P(n) for each positive integer n.
As follows:
Base case: 𝑃(1) is true.

Strong Inductive Step:

Suppose K is a positive integer such that 𝑃(1), 𝑃(2), . . . , 𝑃(𝑘) are all true.

Prove that 𝑃(𝑘 + 1) is true. So, the key step is to show:


𝑃(1), 𝑃(2), . . . , 𝑎𝑛𝑑 𝑃(𝑘) =⇒ 𝑃 (𝑘 + 1).

So, to speak, the statement is true if you can prove that:


(1) The first domino has fallen.
(2) If k is such that the first k dominos have fallen, then the (𝑘 + 1)𝑡ℎ
domino has fallen. We now give an example of a proof related to prime
numbers using strong induction.

Example:
Every integer greater than 1 can be written as the product of prime
numbers.

Think of it this way:

42
Let 𝑃(𝑛) be the statement that 𝑛 can be written as the product of prime
numbers.
Then the proposition says: 𝑃(𝑛) is true for each integer greater or equal
to 2.

Proof (by strong induction).


(1) Base case: 2 is a prime, so it is the product of a single prime.
(2) Strong inductive step: Suppose for some 𝑘 ≥ 2 that each integer n
with 2 ≤ n ≤ k may be written as a product of primes.
We need to prove that 𝑘 + 1 is a product of primes.
Case (a): Suppose 𝑘 + 1 is a prime. Then we are done.
Case (b): Suppose 𝑘 + 1 is a not prime. Then by the fact stated above,
there exist integers a and b with 2 ≤ 𝑎, 𝑏 ≤ 𝑘 such that
𝑘 + 1 = 𝑎 · 𝑏.
By the strong inductive hypothesis, since 2 ≤ 𝑎, 𝑏 ≤ 𝑘, both a and b
are the product of primes.
Thus 𝑘 + 1 = 𝑎 · 𝑏 is the product of primes.
(3) We are done by strong induction.

43
Exercises on Mathematical Induction

𝑛(𝑛−1)(𝑛+1)
1. 𝑃(𝑛): ∑𝑛−1
𝑡=1 𝑡(𝑡 + 1) = , for all natural numbers 𝑛 ≥ 2.
3

1
2. Prove: 12 + 22 + ⋯ + 𝑛2 = 6 𝑛(𝑛 + 1)(2𝑛 + 1)

3. Prove by mathematical induction: The 𝑛𝑡ℎ odd number is 2𝑛 − 1.

4. Prove by mathematical induction: The sum of the first 𝑛 odd numbers is 𝑛2 .

5. Prove by mathematical induction: For all 𝑛 ≥ 1, (1 + 𝑥)𝑛 ≥ 1 + 𝑛𝑥, where 𝑥


is any real number greater than −1.
Note this means that (1 + 𝑥) > 0.

44
Answers Key: Mathematical Induction

1. Let the given statement 𝑃(𝑛), be given as

𝑛(𝑛−1)(𝑛+1)
𝑃(𝑛): ∑𝑛−1
𝑡=1 𝑡(𝑡 + 1) = , for all natural numbers 𝑛 ≥ 2.
3
We observe that

2−1 1
1∙2∙3
𝑃(2): ∑ 𝑡(𝑡 + 1) = ∑ 𝑡(𝑡 + 1) = 1 ∙ 2 =
3
𝑡=1 𝑡=1
2 ∙ (2 − 1) ∙ (2 + 1)
=
3
Thus, 𝑃(𝑛) is true for 𝑛 = 2.
Assume that 𝑃(𝑛) is true for 𝑛 = 𝑘 ∈ ℕ.

𝑘(𝑘−1)(𝑘+1)
i.e., 𝑃(𝑘): ∑𝑘−1
𝑡=1 𝑡(𝑡 + 1) = 3
To prove that 𝑃(𝑘 + 1) is true, we have

(𝑘+1−1) 𝑘

∑ 𝑡(𝑡 + 1) = ∑ 𝑡(𝑡 + 1)
𝑡=1 𝑡=1

𝑘−1
𝑘(𝑘 − 1)(𝑘 + 1)
∑ 𝑡(𝑡 + 1) + 𝑘(𝑘 + 1) = + 𝑘(𝑘 + 1)
3
𝑡=1

𝑘−1+3 𝑘(𝑘 + 1)(𝑘 + 2)


𝑘(𝑘 + 1) [ ]=[ ]
3 3

(𝑘 + 1)][(𝑘 + 1) + 1]
3
Thus, 𝑃(𝑘 + 1) is true, whenever 𝑃(𝑘) is true.
Hence, by the Principle of Mathematical Induction, 𝑃(𝑛) is true for all natural
numbers 𝑛 ≥ 2.■

1
2. Prove: 12 + 22 + ⋯ + 𝑛2 = 6 𝑛(𝑛 + 1)(2𝑛 + 1)
Proof:
𝑃(1) is true, so we assume that 𝑃(𝑘) holds and work on 𝑃(𝑘 + 1).
1
12 + 22 + ⋯ + 𝑘 2 + (𝑘 + 1)2 = 6 𝑘(𝑘 + 1)(2𝑘 + 1) + (𝑘 + 1)2 using P(k).
1
(𝑘 + 1)(𝑘 + 2) + (2𝑘 + 3) which is what 𝑃(𝑘 + 1) says it should be.
6

45
3. Prove by mathematical induction: The 𝑛𝑡ℎ odd number is 2𝑛 − 1.
Initial step: The first odd number is 1 which is 2(1) − 1 so the assertion is true
for 𝑛 = 1.
Inductive step:
Stage 1: The inductive hypothesis asserts that the 𝑛𝑡ℎ odd number is 2𝑛 − 1.
Stage 2: We want to prove that the (𝑛 + 1)𝑡ℎ odd number is 2( 𝑛 + 1) − 1, ie
2𝑛 + 1.
Stage 3: How can we get to Stage 2 from Stage 1?
By the inductive hypothesis, the 𝑛𝑡ℎ odd number is 2𝑛 − 1. \
The next odd number is 2 more than this, i.e. 2𝑛 − 1 + 2 = 2𝑛 + 1 as required.
Thus, the result is true for all 𝑛 ≥ 1.■
4. The sum of the first 𝑛 odd numbers is 𝑛2 .
Initial step: For 𝑛 = 1, the sum is 1 which is12 .
Inductive step:
Stage 1: The inductive hypothesis asserts that
1 + 3 + 5 +··· + (2𝑛 − 1) = 𝑛2 .
(Note, we are using the result of question 3 here.)
Stage 2: We want to prove that the sum of the first (𝑛 + 1) odd numbers is
(𝑛 + 1)2 .
Stage 3: Getting to stage 2 from stage 1.
The sum of the first (𝑛 + 1) odd numbers is equal to the sum of the first 𝑛 odd
numbers plus the next odd number, namely, 1 + 3 + 5 +··· + (2𝑛 − 1) +
(2𝑛 + 1), as each odd number is two more than the previous one.
So,
1 + 3 + 5 +··· + (2𝑛 − 1) + (2𝑛 + 1) = 𝑛2 + (2𝑛 + 1) the inductive
2
hypothesis = 𝑛 + (2𝑛 + 1)
= (𝑛 + 1)2 as required.
The result is therefore true for all n ≥ 1. ■
5. For all 𝑛 ≥ 1, (1 + 𝑥)𝑛 ≥ 1 + 𝑛𝑥, where 𝑥 is any real number greater than
−1.
Note this means that (1 + 𝑥) > 0.
Initial step: For 𝑛 = 1, (1 + 𝑥)1 = 1 + 𝑥 𝑎𝑛𝑑 1 + 𝑛𝑥 = 1 + 𝑥, so (1 + 𝑥)𝑏 =
1 + 𝑛𝑥 ≥ 1 + 𝑛𝑥 is true for 𝑛 = 1.
Inductive step: We assume that the proposition is true for 𝑛 = 𝑘. That is,
(1 + 𝑥) 𝑘 ≥ 1 + 𝑘𝑥, when 𝑥 ≥ −1.
We want to prove that(1 + 𝑥)𝑘+1 ≥ 1 + (𝑘 + 1)𝑥. (1)
Let us look at the inequality and try to relate it to the inductive hypothesis.
Now the left hand side of (1) is
(1 + 𝑥)𝑘+1 = (1 + 𝑥)(1 + 𝑥)𝑘
≥ (1 + 𝑥)(1 + 𝑘𝑥) using the inductive hypothesis and (1 + 𝑥) > 0
= 1 + 𝑘𝑥 + 𝑥 + 𝑘𝑥 2
≥ 1 + 𝑘𝑥 + 𝑥 since 𝑘𝑥 2 ≥ 0

46
= 1 + ( 𝑘 + 1)𝑥.
Hence (1 + 𝑥) 𝑘+1 ≥ 1 + (𝑘 + 1)𝑥,as required.
So, the result is true for all 𝑛 ≥ 1.■

47
FINITE and INFINITE SETS

FAVIOLA S. EDRADAN

48
Finite and Infinite Sets

Definition:
A set A is a finite set provided that A = ∅ or there exists a natural number
k such that A ≈ ℕk. A set is an infinite set provided that it is not a finite set.
If A ≈ ℕk, we say that the set A has cardinality k (or cardinal number k),
and we write card (A) = k.
In addition, we say that the empty set has cardinality 0 (or cardinal
number 0), and we write card (∅) = 0.
Notice that by this definition, the empty set is a finite set. In addition, for
each k ∊ ℕ, the identity function on ℕk is a bijection and hence, by definition,
the set ℕk is a finite set with cardinality k.

Theorem 1.1:
Any set equivalent to a finite nonempty set A is a finite set and has the
same cardinality as A.
Proof.
Suppose that A is a finite nonempty set, B is a set, and A≈B. Since A is a
finite set, there exists a k ∊ ℕ such that A ≈ ℕk. We also have assumed that
A≈B and so we can conclude that B≈A. Since A ≈ ℕk, we can conclude
that B ≈ ℕk. Thus, B is finite and has the same cardinality as A. ∎
The following two lemmas will be used to prove the theorem that states that
every subset of a finite set is finite.

Theorem 1.2:
If A is a finite set and x∈A, then A ⋃ {x} is a finite set and card A (A ⋃ {x})
= card (A)+1.
Proof.
Let A be a finite set and assume card (A) = k, where k = 0 or k ∊ ℕ. Assume
x ∊ A.
If A = ∅, then card(A) = 0 and A ⋃ {x} = {x}, which is equivalent to ℕ 1.
Thus, A ⋃ {x} is finite with cardinality 1, which equals card (A) + 1.
If A ≠ ∅ then A ≈ ℕk, for some k ∊ ℕ. This means that card(A) = k, and
there exists a bijection f: A ⟶ ℕk. We will now use this bijection to define
a function g: A ⋃ {x} ⟶ ℕk+1 and then prove that the function g is a
bijection. We define g: A ⋃ {x} ⟶ ℕk+1 as follows: For each t ∊ A ⋃ {x},

𝑓(𝑡) 𝑖𝑓 𝑡 ∊ 𝐴
g(t)= {
𝑘 + 1 𝑖𝑓 𝑡 = 𝑥.

To prove that g is an injection, we let x1, x2 ∊ A ⋃ {x} and assume x1 ≠ x2.

 If x1, x2 ∊ A, then since f is a bijection, f(x1 ) ≠ f( x2 ) and this implies


that g(x1) ≠ g( x2 ).

49
 If x1= x, then since x2 ≠ x1, we conclude that x2 ≠ x and hence x2 ∊
A. So g(x1) = 𝑘 + 1, and since f( x2 )= ℕk and g( x2 )= f( x2 ), we can
conclude that g(x1) ≠ g( x2 ).

This proves that the function g is an injection. ∎

Example 2:
For each natural number m, if A ⊆ ℕm, then A is a finite set and
card (A)≤ m.

Proof.

We will use a proof using induction on m. For each m ∊ ℕ, let


P(m) be, “If A ⊆ ℕm, then A is finite and card (A) ≤ m.”

We first prove that P(1) is true. If A ⊆ ℕ1, then A = ∅, or A= {1}, both of


which are finite and have cardinality less than or equal to the
cardinality of ℕ1. This proves that P(1) is true.

For the inductive step, let k ∊ ℕ and assume that P(k) is true. That is,
assume that if B ⊆ ℕk, then B is a finite set and card (B) ≤k . We need
to prove that P(k+1) is true.

So, assume that A is a subset of ℕk+1. Then A ⎯ { k+1} is a subset of ℕk.


Since P(k) is true, A⎯ {k+1} is a finite set and

card (A ⎯ {k+1}) ≤k.

There are two cases to consider: Either k+1 ∊ A or k+1 ∉ A.

If k+1 ∉ A, then A= (A ⎯ {k+1}). Hence, A is finite and

card (A) ≤ k ≤ k + 1.

If k+1 ∊ A, then A = A ⎯ { k+1} ⋃ { k+1}. Hence, by Theorem 2, A


is a finite set and
card (A)= card (A ⎯ {k+1}) + 1
Since card (A ⎯ {k+1}) ≤ k, we can conclude that card (A) ≤ k + 1.
This means that we have proved the inductive step. Hence, by
mathematical induction, for each m ∊ ℕ, if A ⊆ ℕm, then A is finite
and card (A) ≤ m. ∎

Theorem 1.3:
If S is a finite set and A is a subset of S, then A is a finite set and card(A)
≤ card (S).

50
Proof.
Let S be a finite set and assume that A is a subset of S. If A = ∅,
then A is a finite set and card (A) ≤ card (S). So we assume that
A ≠ ∅.
Since S is finite, there exists a bijection f: S ⟶ ℕk for some k ∊ ℕ. In this
case, card (S)= k. We need to show that A is equivalent to a finite set.
To do this, we define g: A ⟶ f(A) by
g (x) = f (x) for each x ∊ A.
Since f is an injection, we conclude that g is an injection. Now let y ∊ f
(A). Then there exists an a ∊ A such that f(a)= y. But by the definition of
g, this means that g(a)= y, and hence g is a surjection. This proves that
g is a bijection.
Hence, we have proved that A ≈ f(A). Since f(A) is a subset of ℕk, we
use Theorem 3 to conclude that f(A) is finite and card (f(A)) ≤ k. In
addition, by Theorem 1.1, A is a finite set and card (A) =card( f(A) ).This
proves that A is a finite set and card (A) ≤ card (S).∎
Example 2:
A finite set is not equivalent to any of its proper subsets.
Proof.
Let B be a finite set and assume that A is a proper subset of B.
Since A is a proper subset of B, there exists an element x in B− A. This
means that A is a subset of B− {x}. Hence, by Theorem 1.4, card (A) ≤
card( B− {x})
Also, by Corollary 1.41, (If A is a finite set and x ∊ A, then A− {x} is a
finite set and card (A− {x})= card (A)−1.),
card (B− {x})= card (B)−1.
Hence, we may conclude that card (A) ≤ card (B)−1 and that
card (A) < card (B).
Theorem 1.1 implies that B≉A. This proves that a finite set is not
equivalent to any of its proper subsets. ∎

Definition 2:
One of the property of finite sets that we will consider is often called the
Pigeonhole Principle. The “pigeonhole” version of this property says, “If
m pigeons go into r pigeonholes and m > r, then at least one pigeonhole
has more than one pigeon.”
In this situation, we can think of the set of pigeons as being equivalent to a
set P with cardinality m and the set of pigeonholes as being equivalent to a
set H with cardinality r. We can then define a function f : P⟶ H that maps
each pigeon to its pigeonhole. The Pigeonhole Principle states that this
function is not an injection. (It is not one-to-one since there are at least two
pigeons “mapped” to the same pigeonhole.)

Theorem 2.1:

51
(The Pigeonhole Principle).
Let A and B be finite sets. If card (A) > card (B), then any function f : A⟶
B is not an injection.
Proof.
Let A and B be finite sets. We will prove the contrapositive of the theorem,
which is, if there exists a function f : A⟶ B that is an injection, then
card (A) ≤ card (B).
So assume that f : A⟶ B is an injection. As in Theorem 1.4, we define a
function g: A⟶ f(A)
g(x) = f(x) for each x ∊ A.
As we saw in Theorem 1.4, the function g is a bijection. But then
A≈ f(A) and f(A) ⊆ B. Hence,
card (A) = card ( f(A) ) and card ( f(A) ) ≤ card (B).
Hence, card (A) ≤ card (B), and this proves the contrapositive.
Hence, if card (A) > card (B), then any function f : A⟶ B is not an
injection. ∎

Definition 3:
(a) The empty set ∅ is said to have 0 elements.
(b) If n ∈ ℕ, a set S is said to have n elements if there exists a bijection
from the set ℕn: = {1,2,...,n} onto S.
(c) A set S is sad to be finite if it is either empty or it has n elements for
some n ∈ ℕ.
(d) A set S is said to be infinite it is not finite.
Since the inverse of a bijection is a bijection, it is easy to see that a set
S has n elements if and only if there is a bijection from S onto the set
{1,2,…,n}. Also, since the composition of two bijections is a bijection, we
see that a set S1 has n elements if and only if there is a bijection from S1
onto another set S2 that has n elements. Further, a set T1 is finite if and
only if there is a bijection from T1 onto another set T2 that is finite.
It is now necessary to establish some basic properties of finite sets to be
sure that the definitions do not lead to conclusions that conflict with our
experience of counting. From the definitions, it is not entirely clear that a
finite set might not have n elements for more than one value of n. Also
it is conceivable possible that the set ℕ: {1,2,3, …} might be a finite set
according to this definition.

Theorem 3.1:
(a) If A is a set with m elements and B is a set with n elements and if A
⋂ B= ∅, then A ⋃ B has m + n elements.
Proof.
(a) Let f be a bijection of ℕm onto A, and let g be a bijection of ℕn onto
B. We define h on ℕm+n by h(i)= f (i) for i=1,…, m and h(i)= g(i−m)
for i= m+1, … , m+n. ∎
Example 3:

52
Suppose that S and T are sets and that T ⊆ S.
(a) If S is a finite set, then T is a finite set.
Proof.
(a) If T = 0, we already know that T is a finite set. Thus, we may
suppose that T≠ ∅. The proof is by induction on the number of
elements in S.
If S has 1 element, then the only nonempty subset T of S must coincide
with S, so T is a finite set.
Suppose that every nonempty subset of a set with k elements is finite.
Now let S be a set having k + 1 elements (so there exists a bijection f of
ℕk+1 onto S), and let T ⊆ S. If f (k + 1) ∉ T, we can consider T to be a
subset of S1 = S\ {f (k + 1)}, which has k elements. Hence, by the
induction hypothesis, T is a finite set.
On the other hand, if f (k + 1) ∈ T, then T1 = T\ {f (k + 1)} is a subset of
S1. Since S1 has k elements, the induction hypothesis implies that T1 is
a finite set. But this implies that T = T1 U {f (k + 1)} is also a finite set. ∎

Definition 4:
For any positive integer n, let Jn be the set whose elements are
the integers 1, 2, ..., n; let J be the set consisting of all positive integers.
For any set A, we say:

1. A is finite if A ∼ Jn for some n (the empty set is also considered to


be finite).
2. A is infinite if A is not finite.
3. A is countable if A ∼ J.
4. A is uncountable if A is neither finite nor countable.
5. A is at most countable if A is finite or countable.

Note that by convention, countable implies infinite (so, strictly speaking, we


do not say ‘countably infinite’, although you will hear this phrase from time
to time).

For two finite sets, we have A ∼ B if and only if A and B ’have the same
number of elements’. But for infinite sets this notion becomes vague, while
the idea of 1-1 correspondence (under which, given a mapping from A to
B, the image in B of x1 ∈ A is distinct from the image in B of x2 ∈ A whenever
x1 is distinct from x2) retains its clarity.

Example 4:
Let A be the set of all integers. Then A is countable. For consider the
following arrangement of the sets A and J:

A: 0, 1, −1, 2, −2, 3, −3, ...

53
J: 1, 2, 3, 4, 5, 6, 7, ...

We can, in this example, even give an explicit formula for a function f from
J to A which sets up a 1-1 correspondence:

𝑛
(𝑛 𝑒𝑣𝑒𝑛)
2
f(n) = { 𝑛−1
− (𝑛 𝑜𝑑𝑑)
2

Example 5:
Let R = {whole numbers between 5 and 45}
Then, R is a finite set and n(R) = 38.

Theorem 4.1:
Every infinite subset of a countable set A is countable.
Proof.
Suppose E ⊂ A, and E is infinite. Arrange the elements x of A in a
sequence {xn} of distinct elements. Construct a sequence {nk} as follows:
Let n1 be the smallest postive integer such that 𝑥𝑛 𝑘 ∈ E. Having chosen
n1, ..., nk−1 (k = 2, 3, 4, ...), let nk be the smallest integer greater than nk−1
such that xnk ∈ E.
Then, letting f(k) = 𝑥𝑛𝑘 (k = 1, 2, 3, ...), we obtain a 1-1 correspondence
between E and J. ∎
One interpretation of the theorem is that countability represents the
’smallest’ kind of infinity, in that no uncountable set can be a subset of a
countable set.

Theorem 4.2:
Let {En} be a sequence of countable sets, and put

S = ⋃∞
𝑛=1 𝐸𝑛

Then S is countable.
Proof.
Let every set En be arranged in a sequence {𝑥𝑛 𝑘 }, k = 1, 2, 3, ... and
consider the infinite array

x11 x12 x13 ...


x21 x22 x23 ...
x31 x32 x33 ...
... ... ... ...

54
in which the elements of En form the nth row. The array contains all
elements of S. We can arrange these elements in a sequence as follows:

x11; x21, x12 ; x31, x22, x13; x41, x32, x23, x14, ...

If any of the sets En have elements in common, these will appear more
than once in the above sequence. Hence there is a subset T of the set
of all positive integers such that S ∼ T, which shows that S is at most
countable. Since E1 ⊂ S, and E1 is infinite, S is also infinite, and thus
countable. ∎

Example 6:
Let S = {x : x ∈ ℤ and x2 – 81 = 0}.
Then, S = {-9, 9} is a finite set and n(S) = 2.
Definition 5:
A set S is said to be infinite if it is not finite.

Theorem 5.1:
Let A and B be sets. If A is infinite and A ≈ B, then B is infinite.

Proof.
We use a proof by contradiction and assume that A is an infinite set, A
≈ B, and B is not infinite. That is, B is a finite set. Since A ≈ B and B is
finite, Theorem 1.1 implies that A is a finite set. This is a contradiction to
the assumption that A is infinite. We have therefore proved that if A is
infinite and A ≈ B, then B is infinite. ∎

Example 7:
Set of all points in a line segment is an infinite set.

Example 8:
Set of all positive integers which is multiple of 3 is an infinite set.

Example 9:
𝕎 = {0, 1, 2, 3, ……..} i.e. set of all whole numbers is an infinite set.

Example 10:
ℕ = {1, 2, 3, ……….} i.e. set of all natural numbers is an infinite set.

55
EXERCISES

1. The set ℤ of integers is countably infinite, and so card (ℤ) = ℵ0.

2. Let A be a set and b ∉ A. Then A is infinite if and only if A bijects A ⋃ {b}.

3. The set of positive rational numbers is countably infinite.

4. If A is a countably infinite set, then A ⋃ {x} is a countably infinite set.

5. State, whether the following set is infinite or finite:

(i) Set of integers


(ii){Multiples of 5}
(iii) {Fractions between 1 and 2}
(iv) {Numbers of people in India}
(v) Set of trees in the world
(vi) Set of prime numbers
(vii) Set of leaves on a tree
(viii) Set of children in all the schools of Delhi}
(ix) Set of seven natural numbers
(x) { ……, -4, -, 2, 0, 2, 4, 6, 8}
(xi) {-12, -9, -6, -3, 0, 3, 6, …… }
(xii) {Number of points in a line segment 4 cm long}.

56
ANSWER KEY

1. The set ℤ of integers is countably infinite, and so card (ℤ) = ℵ0.

Proof. To prove that ℕ ≈ ℤ, we will use the following function f : ℕ ⟶ ℤ,


where
𝑛
if 𝑛 is even
2
f(n) = {1−𝑛
if 𝑛 is odd.
2

If n is an even natural number, then f(n) > 0, and if n is an odd natural


number, then f(n) ≤ 0. So it seems reasonable to use cases to prove
that f is a surjection and that f is an injection.

To prove that f is a surjection, we let y ∈ ℤ.

 If y > 0, then 2y ∈ ℕ, and


2𝑦
f(2y)= 2 = y.
 If y ≤ 0, then -2y ≥ 0 and 1−2y is an odd natural number. Hence,
1−(1−2𝑦) 2𝑦
f (1−2y) = = 2 = y.
2
These two cases prove that if y ∈ ℤ, then there exists an n ∈ ℕ such that
f( n)=y. Hence, f is a surjection.

To prove that f is an injection, we let m, n ∈ ℕ and assume that f(m)= f(n).


First note that if one of m and n is odd and the other is even, then one of f(m)
and f(n) is positive and the other is less than or equal to 0. So if f(m)= f(n), then
both m and n must be even or both m and n must be odd.

 If both m and n are even, then


𝑚 𝑛
f(m)= f(n) implies that =
2 2

and hence, m=n.

 If both m and n are odd, then


1−𝑚 1−𝑛
f(m)= f(n) implies that = .
2 2

From this, we conclude that 1− m = 1− n and hence that m=n. This


proves that f(m)= f(n), then m=n and hence that f is an injection.

Since f is both a surjection and an injection, we see that f is a bijection and,


therefore, ℕ ≈ ℤ. Hence, ℤ is countably infinite and card (ℤ)= ℵ0. ∎

57
2. Let A be a set and b ∉ A. Then A is infinite if and only if A bijects A ⋃ {b}.

Proof. Since A is not the same size as A ⋃ {b} when A is finite, we only have
to show that A ⋃ {b} is the same size as A when A is infinite.

That is, we have to find a bijection between A ⋃ {b} and A when A is


infinite.Here’s how: since A is infinite, it certainly has at least one
element; call it a0. But since A is infinite, it has at least two elements, and
one of them must not equal to a0; call this new element a1. But since A
is infinite, it has at least three elements, one of which must not equal
both a0 and a1; call this new element a2. Continuing in this way, we
conclude that there is an infinite sequence a0, a1, a2 , …, an , … of different
elements of A. Now it’s easy to define a bijection e : A ⋃ {b} → A:

e(b)= a0,
e(an )= an+1 for some n ∈ ℕ,
e(a)= a for a ∈ A− {b, a0, a1, …}. ∎

3. The set of positive rational numbers is countably infinite.

Proof. We can write all the positive rational numbers in a two-dimensional array
as shown in Figure 9.2.

The top row in Figure 9.2 represents the numerator of the rational
number, and the left column represents the denominator. We follow the
arrows in Figure 9.2 to define f : ℕ → ℚ+ . The idea is to start in the upper
left corner of the table and move to successive diagonals as follows:

 We start with all fractions in which the sum of the numerator and
1 1
denominator is 2 (𝑜𝑛𝑙𝑦 1). So f(1)= 1.
 We next use those fractions in which the sum of the numerator and
2 1
denominator is 3. So f(2)= 1 and f(3)= 2.

58
 We next use those fractions in which the sum of the numerator and
1 3 2 2 1
denominator is 4. So f(4)= 3, f(5)= 1. We skipped 2 since 2 = 1. In this
way, we will ensure that the function f is a one-to-one function.

We now continue with successive diagonals omitting fractions that are


not in lowest terms. This process guarantees that the function f will be
an injection and a surjection. Therefore, ℕ ≈ ℚ+ and card (ℚ+) = ℵ0. ∎

4. If A is a countably infinite set, then A ⋃ {x} is a countably infinite set.

Proof. Let A be a countably infinite set. Then there exists a bijection f :ℕ → 𝐴.


Since x is either in A or not in A, we can consider two cases.

If x ∈ A, then A ⋃ {x} = A and A ⋃ {x} is countably infinite.

If x ∉ A, define g: ℕ → A ⋃ {x} by

𝑥 𝑖𝑓 𝑛 = 1
g(n)= {
𝑓(𝑛 − 1) 𝑖𝑓 𝑛 > 1.

Since g is a bijection, we have proved that A ⋃ {x} ≈ ℕ and hence,


A ⋃ {x} is countably infinite. ∎

5. State, whether the following set is infinite or finite:

(i) Set of integers – Infinite

(ii){Multiples of 5} - Infinite

(iii) {Fractions between 1 and 2} - Infinite

(iv) {Numbers of people in India} - Finite

(v) Set of trees in the world- Infinite

(vi) Set of prime numbers- Infinite

(vii) Set of leaves on a tree- Finite

(viii) Set of children in all the schools of Delhi- Finite

(ix) Set of seven natural numbers - Infinite

(x) { ……, -4, -, 2, 0, 2, 4, 6, 8} - Infinite

59
(xi) {-12, -9, -6, -3, 0, 3, 6, …… }- Infinite
(xii) {Number of points in a line segment 4 cm long}. - Infinite

60
COUNTABLE and UNCOUNTABLE SETS

MIKAELLA G. PEDARSE

61
Countable and Uncountable Sets

We now introduce an important type of infinite set.

Definition:
(a) A set S is said to be denumerable (or countably infinite) if there
exists a bijection of N onto S.
(b) A set S is said to be countable if it is either finite or denumerable.
(c) A set S is said to be uncountable if it is not countable.

Countable Sets

Definition:
Countable set is a set with the same cardinality (number of elements)
as some subset of the set of natural number. A countable set is either a
finite set or a countably infinite set.
A map f between sets S1 and S2 is called a bijection if f is one-to-one
and onto. In other words
 If f(a) = f(b) then a = b. This holds for all a, b ∈ S1.
 For each b ∈ S2, there is some a in S1 such that f(a) = b.
We write S1 ∼ S2 if there is a bijection f : S1 → S2. We say that S1 and
S2 are equivalent or have the same cardinality if S1 ∼ S2. This notion of
equivalence has several basic properties:

1. S ∼ S for any set S. The identity map serves as a bijection from S to


itself.
2. If S1 ∼ S2 then S2 ∼ S1. If f : S1 → S2 is a bijection then the inverse.
map f−1 is a bijection from S2 to S1.
3. If S1 ∼ S2 and S2 ∼ S3 then S1 ∼ S3. This boils down to the fact that
the composition of two bijections is also a bijection.

These three properties make ∼ into an equivalence relation.

Definition:
Let N = {1, 2, 3...} denote the natural numbers. A set S is called count-

able is S ∼ T for some T ⊂ N.

Theorem:

62
If S is both countable and infinite, then there is a bijection between S
and N itself.

Proof.
For any s ∈ S, we let f(s) denote the value of k such that s is the kth
smallest element of S. This map is well defined for any s, because
there are only finitely many natural numbers between 1 and s. It is
impossible for two different elements of S to both be the kth smallest
element of S.
Hence f is one-to-one. Also, since S is infinite, f is onto. ∎

Theorem:
Every infinite subset of a countable set A is countable.

Proof.
Suppose E ⊂ A, and E is infinite. Arrange the elements x of A in a
sequence {xn} of distinct elements. Construct a sequence {nk} as
follows:
Let n1 be the smallest positive integer such that xnk ∈ E. Having chosen
n1, ..., nk−1(k = 2, 3, 4, ...), let nk be the smallest integer greater than
nk−1 such that xnk ∈ E.
Then, letting f (k) = xnk (k = 1, 2, 3, ...), we obtain a 1-1 correspondence
between E and J.
One interpretation of the theorem is that countability represents the
‘smallest’ kind of infinity, in that no uncountable set can be a subset of
a countable set.

Lemma:
If S is countable and S′ ⊂ S, then S is countable.

Proof.
Since S is countable, there is a bijection f: S → N. But then f(S′) = N′ is
a subset of N, and f is a bijection between S′ and N′. ∎

Theorem:
Let {En} be a sequence of countable sets, and put

S= ∪ En
n=1
Then S is countable.

Proof.
Let every set En be arranged in a sequence {xnk}, k = 1, 2, 3, ... and
consider the infinite array

63
x11 x12 x13 ...
x21 x22 x23 ...
x31 x32 x33 ...
... ... ... ...
in which the elements of En form the nth row. The array contains all
elements of S. We can arrange these elements in a sequence as follows:

x11; x21, x12; x31, x22, x13; x41, x32, x23, x14, ...

If any of the sets En have elements in common, these will appear more than
once in the above sequence. Hence there is a subset T of the set of all positive
integers such that S ∼ T , which shows that S is at most countable. Since E1 ⊂
S, and E1 is infinite, S is also infinite, and thus countable. ∎

Theorem:
Let A be a countable set, and let Bn be the set of all n-tuples (a1, ..., an)
where ak ∈ A (k = 1, 2, ..., n) and the elements a1, ..., an need not be
distinct. Then Bn is countable.
Proof.
That B1 is countable is evident, since B1 = A. Suppose Bn−1 is countable
(n = 2, 3, 4, ...). The elements of Bn are of the form
(b, a) (b ∈ Bn−1, a ∈ A)
For every fixed b, the set of pairs (b, a) is equivalent to A, and thus
countable. Thus, Bn is the union of a countable set of countable sets;
thus, Bn is countable, and the proof follows by induction on n.

Theorem:
Suppose that S and T are sets and that T ⊆ S.
 If S is a countable set, then T is a countable set.\

Theorem:
The following statements are equivalent:
a. S is a countable set.

b. There exists a surjection of N onto S.


c. There exists an injection of S into N.
Proof.
(a) ⟹ (b)
If S is finite, there exists a bijection h of some set Nn onto S and we
define H on N by
ℎ(𝑘) 𝑓𝑜𝑟 𝑘 = 1,· · · , 𝑛,
H(k) = {
ℎ(𝑛) 𝑓𝑜𝑟 𝑘 > 𝑛.
Then H is a surjection of N onto S.

64
If S is denumerable, there exists a bijection H of N onto S, which is also
a surjection of N onto S.
(b) ⟹ (c)
If H is a surjection of N onto S, we define H1: S → N by letting H1 (s)
be the least element in the set H-1(s) := {n ∈ N: H(n) = s}.
To see that H1 is an injection of S into N, note that if s, t ∈S and nst = H1
(s) = H1(t), then s = H(nst) = t.
(c) ⟹ (a)
If H1 is an injection of S into N, then it is a bijection of S onto H1 (S) ~
N.
By Theorem (a), H1 (S) is countable, whence the set S is countable.

Example 1:
Let A be the set of all integers. Then A is countable. For consider the
following arrangement of the sets A and J:
A : 0, 1, −1, 2, −2, 3, −3, ...
J : 1, 2, 3, 4, 5, 6, 7, ...
We can, in this example, even give an explicit formula for a function f
from J to A which sets up a 1-1 correspondence:

𝑛
f (n) = { 2 (𝑛 𝑖𝑠 𝑒𝑣𝑒𝑛)
𝑛−1
− (𝑛 𝑜𝑑𝑑)
2

Example 2:
The set of all rational numbers.
We apply the last theorem with n = 2, noting that every rational number
can be written as b/a, where b and a are integers. Since the set of pairs
(b, a) is countable, the set of quotients b/a, and thus the set of rational
numbers, is countable.

Example 3:
The Algebraic Numbers
A real number x is called algebraic if x is the root of a polynomial
equation c0 + c1x + ... + cnxn where all the c’s are integers.
For instance, √2 is an algebraic integer because it is a root of the
equation x2−2 = 0.
To show that the set of algebraic numbers is countable, let Lk denote
the set of algebraic numbers that satisfy polynomials of the form
c0+c1x+...+cnxn where n < k and max (|cj|) < k.
Note that there are at most kk polynomials of this form, and each one
has at most k roots.
Hence Lk is a finite set having at most kk+1 elements.
As above, we make our list L1, L2, L3 of all algebraic numbers and
weed out repeaters.

65
Example 4:
The set E := {2n : n ∈ N } of even natural numbers is denumerable,
since the mapping f : N → E defined by f (n) := 2n for n ∈ N, is a
bijection of N onto E. Similarly, the set O := {2n - 1 : n ∈ N} of odd
natural numbers is denumerable.

Example 5:
Countable Unions of Countable Sets

Lemma:
Suppose that S1, S2, ... ⊂ T are disjoint countable sets. Then S = ∪i Si
is a countable set.

Proof.
There are bijections fi : Si → N for each i. Let Lk denote the set of
elements s ∈ S such that s lies in some Si for i < k, and fi(s) < k.
Note that Lk is a finite set. It has at most k2 members.
The list L1, L2, L3... contains every element of S. Weeding out
repeaters, as above, we see that we have listed all the elements of S.
Hence S is countable. ∎
Indeed, if A = {a1, a2, a3, . . .} and B = {b1, b2, b3, . . .}, we can
enumerate the elements of A U B as
a1, b1, a2, b2, a3, b3, . . .

Uncountable Sets

Definition:
Uncountable Set (or uncountably infinite set) is an infinite set that
contains too many elements to be countable. The uncountability of a
set is closely related to its cardinal number: a set is uncountable if its
cardinal number is larger than that of the set of all natural numbers.
A set is called uncountable if it is not countable.

Lemma:
If S′ ⊂ S and S′ is uncountable, then so is S.

Proof.
This is an immediate consequence of the previous lemma. If S is
countable, then so is S′. But S′ is uncountable. So, S is uncountable as
well.

Cantor's Theorem:

66
If A is any set, then there is no surjection of A onto the set P (A) of all
subsets of A.

Proof.
Suppose that ᵩ : A → P(A) is a surjection. Since ᵩ (a) is a subset of A,
either a belongs to ᵩ (a) or it does not belong to this set. We let
D := {a ∈ A : a ∉ ᵩ (a)}.
Since D is a subset of A, if ᵩ is a surjection, then D = ᵩ (a0) for some
a0 ∈ A.
We must have either a0 ∈ D or a0 ∉ D. If a0 ∈ D, then since D = ᵩ (a0),
we must have a0 ∈ ᵩ (a0), contrary to the definition of D. Similarly, if a0
∉ D, then a0 ∉ ᵩ (a0) so that a0 ∈ D, which is also a contradiction.
Therefore, ᵩ cannot be a surjection.

Cantor's Theorem implies that there is an unending progression of


larger and larger sets. In particular, it implies that the collection P(N) of
all subsets of the natural numbers N is uncountable.

Theorem: Suppose that S and T are sets and that T ⊆ S.


 If T is an uncountable set, then S is an uncountable set.

Example 1:
The Set of Binary Sequences
Let S denote the set of infinite binary sequences. Here is Cantor’s
famous proof that S is an uncountable set. Suppose that f : S → N is a
bijection. We form a new binary sequence A by declaring that the nth
digit of A is the opposite of the nth digit of f−1(n). The idea here is that
f−1(n) is some binary sequence and we can look at its nth digit and
reverse it.
Supposedly, there is some N such that f(A) = N. But then the
Nth digit of A = f−1(N) is the opposite of the Nth digit of A, and this is a
contradiction.

Example 2:
The Real Numbers
Let R denote the reals. Let R′ denote the set of real numbers, between
0 and 1, having decimal expansions that only involve 3s and 7s. (This
set R′ is an example of what is called a Cantor set .) There is a bijection

67
between R′ and the set S of infinite binary sequences. For instance, the
sequence 0101001... is mapped to .3737337.... Hence R′ is uncounta-
ble.

Example 3:
The Transcendental Numbers
A real number x is called transcendental if x is not an algebraic number.
Let A denote the set of algebraic numbers and let T denote the set of
transcendental numbers. Note that R = A ∪ T and A is countable. If T
were countable then R would be the union of two countable sets. Since
R is uncountable, R is not the union of two countable sets. Hence T is
uncountable.
The upshot of this argument is that there are many more
transcendental numbers than algebraic numbers.

Example 4:
The Penrose Tiles
To each Penrose tiling P we can associate a tail end 𝜏 (P). Recall that

There is an infinite sequence P = P0, P1, P2, ... where Pn is the parent
of Pn−1. In other words, Pn is obtained from Pn−1 by the grouping
process discussed in class.
We say that the nth digit of 𝜏 (P) is a 0 if x is contained in a kite of Pn
and a 1 if x is contained in a dart of Pn. We might need to move x
slightly to avoid choosing a point that lies right on a crack. If we replace
x by x′, then only the initial part of the sequence changes. So, 𝜏(P) is
well defined.
By using the subdivision operation, we can produce a Penrose tiling P
that has any 𝜏 (P) we like. Hence, there are uncountably many different
Penrose tilings.

68
Exercises:
Part A
Prove or disprove that the following sets are countable:
1. A = {(m; n) ∈ N × N : m ≤ n}
2. A = Q100
3. The set of irrational numbers.

Part B
Prove that the following sets are countable or uncountable:
1. The set of all functions f : N → {0, 1}.
2. The set of all functions f : {0, 1} →N.
3. The set of all infinite sequences of integers.

69
ANSWER KEY

Part A
1. A = {(m; n) ∈ N × N : m ≤ n}

Solution:
A ⊆ N × N. Therefore |A| ≤ |N|. But since A is also infinite
(indeed, n is unbounded in the set), then |A| = N. Therefore A is
countable.

2. A = Q100
Solution:
Q100 = Q × Q × . . . × Q is a cross product of countable sets.
Since the finite cross product of countable sets is countable.
The set is countable.
3. The set of irrational numbers.
Solution:
Let A be the set of irrationals. Since every real number is
either rational or irrational, then A ∪ Q = R. Suppose for the
sake of contradiction that A is countable. Then since Q is
countable, A ∪ Q is also countable. Therefore R is countable.
This contradicts the fact that the set of real numbers is
uncountable. Therefore A is uncountable.
Part B
1. The set of all functions f : N → {0, 1}.
Solution:
UNCOUNTABLE.
Proof. Mapping a function f : N → {0, 1} to the
sequence (a1, a2, ) defined by ai = f(i) for each i yields a
bijection between functions of the given type and infinite binary

sequences. Since the latter set is uncountable, so is the given


set of all functions f : N → {0, 1}.
2. The set of all functions f : {0, 1} →N.
Solution:
COUNTABLE.
Proof. The functions f : {0, 1} →N are in one-to-
one correspondence with N × N (map f to the tuple (a1, a2)
with a1 = f(1), a2 = f(2)). Since the latter set is countable, as a
Cartesian product of countable sets, the given set is countable
as well.

70
3. The set of all infinite sequences of integers.
Solution:
UNCOUNTABLE.
Proof: The set of all infinite integer sequences is a superset of
the set of all infinite binary (i.e., 01) sequences. But an infinite
binary sequence can be interpreted as the binary expansion
of a real number in the interval [0, 1), and since the set of
such real numbers is uncountable, so is the set of all infinite
binary sequences.

71
THE ALGEBRAIC & ORDER PROPERTY OF
REAL NUMBER

MYCHA REYES

72
THE ALGEBRAIC AND ORDER PROPERTIES OF ℝ
Definition:
Algebraic Properties of ℝ on the set ℝ of a real number there are two
binary operations, denoted by + and called addition and multiplication,
respectively.
These operations satisfy the following properties
Field Axioms of ℝ: The real numbers are a field (as are the rational numbers ℚ and
the complex numbers ℂ). That is, there are binary operations +
and defined on ℝ ∋

(A1) a + b = b + a for all a, b in ℝ ( commutative property of addition);

(A2) (a + b) + c = + (b + c) for all a, b, c in ℝ (associative property of


addition);

(A3) there exists an element 0 in ℝ such that 0 + a = a and a 0 = a for


all a
in ℝ (existence of a zero element);

(A4) for each a in ℝ there exists an element –a in ℝ such that a + (−a)


=
0 and (−a) + a = 0 (existence of negative elements);

(M1) a ⋅ b = b ⋅ a for all a, b, c in ℝ (commutative property of


multiplication);

(M2) (a ⋅ b) ⋅ c = ⋅ a (b ⋅ c) for all a, b, c in ℝ (associative property of


multiplication);

(M3) there exists an element 1 in ℝ distinct from 0 such that 1⋅ a = a


and

73
a ⋅ 1 = a for all a in ℝ (existence of a unit element);

(M4) for each a ≠ 0 in ℝ there exists an element 1⁄𝑎 in ℝ such that a ⋅


(11⁄𝑎 ) = a and (1⁄𝑎 ) ⋅ a = 1 (existence of a reciprocals);

(D) a ⋅ (b +c) = (a ⋅ b) + (a ⋅ c) and = (b +c) ⋅ a = (b ⋅ a) + (c ⋅ a) for all


a,
b, c in ℝ (distributive property of multiplication over addition).

Theorem:
(a) If z and a are elements in ℝ with z + a = a, then z = 0.
(b) If u and b ≠ 0 are elements in ℝ with u ⋅ b =b, then u = 1.
(c) If a ∈ ℝ, then a ⋅ 0 = 0.
Proof.
(a) Using (A3), (A4), (A2), the hypothesis z + a = a, and (A4), we get
z = z + 0 = z + (a +(−a)) = (z +a) + (−a) = a +(−a) = 0.
(b) Using (M3), (M4), (M2), the assumed equality u ⋅ b = b, and (M4) again,
we get u = u ⋅ 1 = u ⋅ (b ⋅ (1⁄𝑏 )) = (u⋅ b) ⋅ (1⁄𝑏 ) = b ⋅ (1⁄𝑏 ) = 1.
(c) We have why?
a + a ⋅ 0 = a ⋅ 1 ⋅ (1 + 0) = a ⋅ 1 = a.

Therefore, we conclude from (a) and a ⋅ 0 = 0.

We next establish two important of multiplication: the uniqueness of


reciprocals and the fact that the product of two numbers in zero only when one of the
factors is zero.

Theorem: (a) If a ≠ 0 and b in ℝ are such that a ⋅ b = 1, then b = 1⁄𝑎 .


(b) If a ⋅ b = 0, then either a = 0 or b = 0.

Proof.

74
(a) Using (M3), (M4), (M2), the hypothesis a ⋅ b = 1, and (M3), we have
b = 1⋅ b = ((1⁄𝑎 )⋅ a) ⋅ b = (1⁄𝑎 )⋅ (a ⋅ b) = (1⁄𝑎 ) ⋅ 1 = 1⁄𝑎 .
(b) It suffices to assume a ≠ 0 and prove that b = 0. (Why?) We multiply
a ⋅ b by 1⁄𝑎 a and apply (M2), (M3) and (M3) to get
(1⁄𝑎 a) ⋅ (a ⋅ b) = ⋅ b = 1⋅b = b.
Since a ⋅ b = 0, by 2.1.2 (c) this also equals
(11⁄𝑎 ) ⋅ (a ⋅ b) = (1⁄𝑎 ) ⋅ 0 = 0.

Thus we have b = 0.

Rational and Irrational Numbers: We regard the set ℕ of natural numbers as a


subset ℝ, by identifying the natural numbers n 𝜖 ℕ with the n- fold sum of the
unit element 1 𝜖 ℝ. Similarly, we identify 0 𝜖 ℤ with the zero element of 0 𝜖 ℝ and we
identify the n- fold sum of −1 with the integer −𝑛.Thus we consider ℕ and ℤ to be
subset of ℝ.

Theorem: There does not exist of a rational number r such that 𝑟 2 = 2

Proof.
Suppose, on the contrary, that p and q are integers such that (𝑝/ 𝑞)2 = 2.
We may assume that p and q are positive and no common integers factors
other than 1. (Why?) Since 𝑝2 = 2𝑞 2 we see that 𝑝2 is even. This implies
that p is also even (because if p = 2n−1 is odd, then its square 𝑝2 = 2 (2𝑛2
−2𝑛 + 1) −1is also odd). Therefore, since p and q do not have 2 as a
common factor, then q must be n odd natural numbers.

Order Properties of ℝ: ℝ is an ordered field, i.e., the following properties are


satisfied.

(1) (Trichotomy) For a, b ∈ ℝ, exactly one of the following is true:


a < b, a = b, or a > b

(2) (Transitive) For a, b, c ∈ ℝ, if a < b and b <c, then a < c.

75
(3) For a, b, c ∈ ℝ, if a < b, then a + c < b + c.

(4) For a, b, c ∈ ℝ, if a < b and c > 0, ac < bc.

Theorem: If a, b ∈ ℝ, then a < b, ⟺ − a > − b.

Proof. a < b ⟺
a + [(−𝑎) + (−𝑏)] < b + [(−𝑎) + (−𝑏)] ⟺
[𝑎 + (−𝑎) + (−𝑏)] < b + [(−𝑏) + (−𝑎)] ⟺
0 + (−𝑏) < [𝑏 + (−𝑎)] ⟺
−b < 0 + (− a) ⟺ − b < − a > −b

Theorem: If a, b, c ∈ ℝ, then a < b and c < 0 ⟹ ac > bc

Proof.
c>0 ⇒ – c > 0.
𝑇𝐻4

Then a (− c) < b (− c) ⟹ −ac < − bc 0 ⇒ ac > bc


𝑇𝐻4

Theorem: If a ∈ ℝ, ∋ 0 ≤ a < 𝜖 ∀𝜖 > 0, then a = 0.

Proof.
Suppose a > 0.

1 1
Since 0 < 2 < 1, 0 < 2a < a.
1
Let 𝜖0 = 2a.
Then 0 < 𝜖0 < a, contradicting our hypothesis. Thus a = 0.

Theorem: By way of contradiction, suppose b < a.

[Need to find an ∈ that gives a contradiction.]

1
Let 𝜖0 = 2(𝑎 − b). then

1 1 1 1 1
𝑎 − 𝜖0 = a− 2(𝑎 − b) = 𝑎 + 2 𝑏 > 2 𝑏 + 2 𝑏 = b,
2

Contradicting our hypothesis. Thus a ≤ b.

76
Examples (1)
Determine the set A of real numbers x such that 2x + 3 ≤ 6.

We note that we have

2
× ∈ 𝐴 ⟺ 2x + 3 ≤ 6 ⟺ 2x ≤ 3 ⟺ x ≤ 3.
2
Therefore A = × ∈ ℝ :{ × ≤ 3}.

(b) Determine that set B: = { × ∈ ℝ : × 𝟐 + × > 2}.

We write the inequality so Theorem can be applied. Note that

× ∈ 𝐴 ⟺ × 𝟐 + × − 2 > 0 ⟺ (× − 1) (× + 2) > 0.

Therefore, we either have (i) × − 1 > 0 and × + 2 > 0, or we have


(ii)
× − 1 > 0 and × + 2 > 0. In case (i) we must have both × >
1 𝑎𝑛𝑑 × > − 2, which is satisfied if and only The symbol ⟺ should be read “If
and only If ”
if × > 1. In case (ii) we must have to × > − 2, which is satisfied if

and only if × > − 2.

(c) Determine the set

2x + 1
C: = {×∈ ℝ: < 1}.
x+2

We note that

2x + 1 x−1
×∈ 𝐶 ⟺ – 1 < 0 ⟺ x + 2 < 0.
x+2

Therefore we have either (i) × − 1 > 0 and × + 2 > 0, or (ii) × − 1 >


0 and × + 2 > 0. (Why?) In case (i) we must have both × > 1 𝑎𝑛𝑑 × > −

77
2, which is satisfied if and only if – 2 < x < 1. In case (ii), we must
have both x > 1 and × < − 2, which is never satisfied. We conclude that C =
{x ∈ ℝ: − 2 < x < 1}.

Examples (2)
(a) Let a ≥ 0 and b ≥ 0 Then

a < b ⟺ 𝑎2 < 𝑏 2 ⟺ √𝑎 < √𝑏


We consider the case where a > 0 and b > 0, leaving the case a = 0 to
the reader. It follows from (i) that a + b > 0. Since 𝑏 2 – 𝑎2 = (b − a) (b + a), it
follows from

(b) that b – a > 0 implies that 𝑏 2 − 𝑎2 > 0. Also, it follows that b – a > 0
implies that 𝑏 2 − 𝑎2 > 0. reader. It follows from (i) that a + b > 0. Since
𝑏 2 – 𝑎2 = (b − a) (b + a), it follows from (c) that b – a > 0 implies that 𝑏 2 − 𝑎2
> 0. Also, it follows that b – a > 0 implies that 𝑏 2 − 𝑎2 > 0.

If a > 0 and b > 0, then √𝑎 > 0 and √𝑏 > 0. Since a = (√𝑎 )2 and b =
(√𝑏 )2, We also leave it to the reader to show that if a ≥ 0 and b ≥ 0, then

a ≤ b ⟺ 𝑎2 ≤ 𝑏 2 ⟺ √𝑎 < √𝑏

1
(c) If a and b are positive real numbers, then their arithmetic mean is 2
(a + b) and their geometric mean is √𝑎𝑏. The Arithmetic-Geometric Mean
Inequality for a, b is
1
√𝑎𝑏 ≤ 2
(a + b)

with equality occurring if and only if a = b.

To prove this, note that if a > 0, b > 0, and a ≠ b, then √𝑎 > 0, √𝑏 > 0
and √𝑎 ≠ √𝑏. (Why?) Therefore it follows from (a) that (√𝑎 − √𝑏)2 >. Expanding
this square, we obtain

78
a−2√𝑎𝑏 + b > 0,

whence it follows that

1
√𝑎𝑏 < 2 (a + b).

Examples (3).
𝑎 −𝑎 𝑎
Prove that − ( ) = = if b ≠ 0.
𝑏 𝑏 −𝑏

𝑎
Proof. Since b ≠ 0 we have 𝑏 ab−1 . So

𝑎
− (𝑏) = −(𝑎𝑏−1 )

= − (𝑏 −1 a)
= ((−𝑏 −1 )𝑎)
= 𝑎(−𝑏 −1 )
𝑎
= −𝑏

Similarly,
−𝑎
−(𝑎𝑏−1 ) = (−𝑎) 𝑏 −1 = . ∎
𝑏

Theorem: Bernoulli’s Inequality. If x > −1, then

(1 + x)𝑛 ≥ 1 + n× for all n ∈ ℕ

Proof.
Uses Mathematical Induction. The case n = 1 yields equality, so that
assertion is valid in this case. Next, we assume the validity of the
inequality

79
for k ∈ ℕ and will deduce it if k + 1. Indeed, the assumptions that
(1 + 𝑥) 𝑘 ≥ 1 + k× and that 1 + × > 0 imply (why?) that.

(1 + 𝑥) 𝑘+1 = (1 + 𝑥) 𝑘 ⋅ ( 1 + ×)
≥ (1+ k×) ⋅ ( 1 + ×) = (k+1) × + k𝑥 2
≥1+(k+1)×.

all n ∈ ℕ Thus, inequality (4) holds for for n = 𝑘 + 1. Therefore, (4) holds .

Example (4).
Given x > −1, then

(a) (1+ x) 𝑟 ≤ 1 + 𝑟𝑥 for 0 < r < 1


(b) (1 + x) 𝑟 ≥ 1 + 𝑟𝑥 for r < 0 or r< 1

Firstly we give the proof that r is a rational number first.

Proof.
Use A.M. ≥ G.M.
𝑝
Since r ∈ Q, r=𝑞

(a) Let 0 < r <, ∴ p < q, q−p > 0. Also 1+ x > 0,

𝑞 𝑝(1 + x)+1 . (𝑝 −𝑞)


(1+ x) 𝑟 = (1+ x) 𝑝/𝑞 = √(1 + x) 𝑝 1 𝑝−𝑞 ≤ 𝑞

𝑝𝑥 + 𝑞 𝑝
= = 1 + 𝑞 𝑥 = 1 + 𝑟𝑥
𝑞

(b) Let r > 1,

(i) If 1 + 𝑟𝑥 ≤ 0, then (1+ x) 𝑟 > 0. ≥ 1 + rx.


(ii) If 1 + 𝑟𝑥 > 0, rx > −1.

80
1
Since r > 1, we have 0 < 𝑟 < 1. By (a) we get
1
(1+ x) 1/𝑟 ≤ 1 + 𝑟𝑥 = +𝑥
𝑟

∴ (1+ x) 𝑟 ≥ 1 + rx.
Let r > 0, then −𝑟 > 0.
Choose a natural number n sufficiently large such that 0 < − 𝑟/𝑛 < 1 and 1> rx/n >
−1.
1
By (a,) 0 < (1+ x) −𝑟/𝑛 ≤ 1 + 𝑥 > 0.
𝑟
𝑟 𝑟 𝑟 𝑟 𝑟
Since 1≥ 1 − ( 𝑥) 2 = (1 𝑥)(1 + 𝑥) ⇒ (1 𝑥) −1 ≥ 1+ 𝑥
𝑛 𝑛 𝑛 𝑛 𝑛

Hence by (11),

𝑟 𝑟
(1+ x) 𝑟 ≥ = (1 𝑥) 𝑛
≥ 1 + 𝑛 ( 𝑛 𝑥) = 1 + 𝑟𝑥.
𝑛

Again, equality holds if and only if x = 0.

Example (5).
Prove that f(x) = (1+ x) 𝑛 on [0,x]

(1 + x) 𝑛 −1 𝑛−1
= n (1+ c)
𝑥
𝑛−1
Because n. (1+ c) > n hence

(1 + x) 𝑛 −1 𝑛−1 𝑛−1 𝑛
= n (1+ c) > n (1+ c) > n ⇒ (1+ x) −
𝑥
𝑛
>n⋅ 𝑥 ⇒ (1+ x) − 1 + 𝑛𝑥
Which complete the proof.
The case for −1 ≤ 𝑥 < 0 is similar. The case for x = 0 is simple.

Example (6).
Prove that 3𝑛 ≥ 2𝑛2 + for n = 1, 2… using binomial
𝑛
Theorem, applied (1+ x) with x =2.

81
𝑛
(1+ 2) =
𝑛 𝑛 𝑛 𝑛
( ) + ( )2 + ( ) 22 + ⋯ ( ) ≥ 1 + 2𝑛! + 2𝑛
0 0 0 𝑛
𝑛 4𝑛!
( ) 22 = (−2)!2! = 2𝑛 (n+1).
2
Then you have, since all terms are positive,

𝑛
(1+ 2) = 1 + 2n (n−1) + … = 1
+ 2𝑛2 + ⋯ ≥ 1+ 2𝑛2

EXERCISES

1
Exercise 1. Find all × ∈ ℝ ⋺ 𝑥 < 𝑥 2 .

Exercise 2. Prove that. If a, b ∈ ℝ, then a < 𝑏 ⟺ −a > −𝑏.

Exercise 3. Let a, b ∈ ℝ, and suppose a ≤ 𝑏 + 𝜖 (or a − 𝜖 ≤ 𝑏 ∀𝜖 > 0. Then a ≤ 𝑏.

Exercise 4: Let a, b ∈ ℝ. Then a + × = b has the unique solution x = (−𝑎) + 𝑏.

Exercise 5: Prove that. If a ∈ ℝ is such that 0 ≤ a < 𝜀 for every 𝜀 > 0, then a = 0.

82
Answer Key:
Solution 1.
1
Proof. < 𝑥2 ⟺
𝑥
1
𝑥2 − 𝑥 > 0 ⟺
1 1
{𝑥 > 0 and 𝑥 3 − 1 > 0} 𝒐𝒓 {𝑥 < 0 and 𝑥 3 − 1 < 0} ⟺

{𝑥 > 0 and 𝑥 3 > 1} or {𝑥 < 0 and 𝑥 3 < 1} ⟺


{𝑥 > 0 and 𝑥 > 1} or {𝑥 < 0 and 𝑥 < 1} ⟺
x > 1 or x < 0.
Solution 2.
Proof 2.
a<b⟺
a + [(−𝑎) + (−𝑏)] < b + [(−𝑎) + (−𝑏)] ⟺
[𝑎 + (−𝑎)] + (−𝑏)< b + [(−𝑏) + (−𝑎)] ⟺
0 +(−𝑏) < [𝑏 + (−𝑏)] + (−𝑎) ⟺
−𝑏 < 0 + (−𝑎) ⟺ −𝑏 < −𝑎 ⟺ −𝑎 > −𝑏
Solution 3.
Proof 3. By way of contradiction, suppose b > a.
[ Need to find 𝜖 that give a contradiction.]
1
Let 𝜖0 = 2 ( 𝑎 − 𝑏). Then
1 1 1 1 1
𝑎 − 𝜖0 = − 2 ( 𝑎 − 𝑏) = 𝑎 + 2𝑏 > 𝑏 + 𝑏 = 𝑏.
2 2 2

Contradicting our hypothesis.


Solution 4.
Proof 4.
̿̿̿̿ [ a + (−𝑎) ] + b 𝐴4
a + [ (−𝑎) + 𝑏 ] 𝐴2 ̿̿̿̿ 0 + b 𝐴3
̿̿̿̿ 𝑏,

so (−𝑎) + b is a solution.
For uniqueness, suppose y is any solution of the equation, i,e., a + y = b. Then
y
̿̿̿̿
𝐴3 0 + 𝑦
̿̿̿̿
𝐴4 [ (−𝑎) + 𝑎 ] + y

83
̿̿̿̿
𝐴2 (−𝑎) + (a + y)
= (−𝑎) + b
Solution 5.
1
Proof 5. Suppose to the contrary that a > 0. Then if we take 𝜀0 : 2 𝑎, we
have
0 <𝜖0 < 𝑎.
Therefore, it is false that a < 𝜀 𝑓𝑜𝑟 𝑒𝑣𝑒𝑟𝑦 𝜀 > 0 and we conclude that a = 0.

84
ABSOLUTE VALUE and REAL LINE

ANIE PADILLA

85
Absolute Value of Real Line
Definition
The absolute value or magnitude of real number a is denoted by /a/ and
defined by
𝑎 𝑖𝑓 𝑎 ≥ 0
/a/ = { }
−𝑎 𝑖𝑓 𝑎 ≥ 0
Example 1:
5/ = 5 /0/ = 0
Since 5 > 0 since 0 ≥ 0
sign if the number is negative and to leave the number unchanged if it is
nonnegative
Example 2:
Solve / x-3 / = 4
Solution.
Depending on whether x-3 is positive or negative, the equation / x-3 / =
4 can be written as
x – 3 = 4 or x – 3 = -4
solving these two equations given x = 7 and x = -1.
Example 3:
Solve / 3x – 2 / = /5x + 4/
Solution. Because two numbers with the same absolute value are either
equal or differ in sign, the given equation will be satisfied if either
3x – 2 = 5x + 4 or 3x – 2 = -(5x + 4)
Solving the first equation yields x = -3 and solving the second yields x = -
¼; thus, the given equation has the solutions x = -3 and -1/4.
Definition
The distance is the same as the absolute value of their difference.

Relationship Between Square Roots and Absolute Value


Recall from algebra that a number is called a square root of a if its square
is a.
Recall also that every positive real number has two square roots, one
positive and one negative; the positive square root is denoted by √𝑎 and

86
the negative square root by - √𝑎. For example, the positive square root of
9 is √9 = and the negative square root of 9 is -√9 = - 3.

It is common error to replace √𝑎2 by a. Although this is correct when as a


is nonnegative, it is false for negative a. For example, if a = -4. Then

√𝑎2 = √(−4)2 = √16 = 4 ≠ a


A result that is correct for all a is given in the following theorem.
Theorem.
For any real number a,

√𝑎 = /a/
Proof.
Since 𝑎2 = (+𝑎)2 = (−𝑎)2, the numbers +a and -a are square roots of
𝑎2 .
If a ≥ 0, then +a is the nonnegative square root of 𝑎2 , and if
a < 0, 𝑡ℎ𝑒𝑛 − 𝑎 is the nonnegative square root of 𝑎2 .

Since √𝑎2 denotes the nonnegative square root of 𝑎2 , it follows that

√𝑎2 = +a if a ≥ 0

√𝑎2 = -a if a ≥ 0

That is, √𝑎2 = /a/. ∎

Properties of Absolute Value


Theorem.
If a and b are real numbers, then
(a) /-a/ = /a/
A number and its negative have the same absolute value.

(b) /ab/ = /a/ /b/


The absolute value of a product is the product of the values.

(c) /a/b/ = /a/∖/b/


The absolute value of the ratio is the ratio is the ratio of the absolute
values.

We will prove parts (a) and (b) only.


Proof
(a). From Theorem E.2,

87
/-a/ = √(−𝑎)2 = √𝑎2 = /a/
Proof
(b). From Theorem E.2 and a basic property of square roots,

/ab/ = √(𝑎)2 =√𝑎2 𝑏 2 = √𝑎2 √𝑏 2 = /a//b/. ∎


The result in part (b) of Theorem E.3 can be extended to three or more
factors.
More precisely, for any n real,𝑎1 , 𝑎2 , … , 𝑎𝑛, it follows that
/𝑎1 𝑎2 … 𝑎𝑛 / = /𝑎1 //𝑎2 / …/𝑎𝑛 / (1)
In the special case where 𝑎1 , 𝑎2 … , 𝑎𝑛 have the same value, a, it follows
from that
/𝑎𝑛 / = /𝑎/ (2)

Geometry Interpretation of Absolute Value


The motion of absolute value arises naturally in distances problems. For
example, suppose that A and B are points on a coordinate line that have
coordinates a and b, respectively. Depending on the relative positions of
the points, the distance d between them will be b – a or a-b (Figure E.1).
In either case the distance can be written as d = /b -a/, so we have the
following result.

A B
a b
b-a
(a)
B A
b a
a-b
(b)
Theorem (Distance Formula).
If A and B are points on a coordinate line with coordinate.
A and b, respectively, then the distance d between A and B is d = /b – a
/.
This theorem provides useful geometric interpretation of some common
mathematical expression.

88
Expression: Geometry Interpretation on a Coordinate Line
/x – a/ The distance between x and a
/x + a/ The distance between x and -a (since/x + a/ = /x – (-a)/)
/x/ The distance between x and the origin (since /x/ = /x – 0/)

Inequalities with Absolute Value

The absolute number of a number a is written as |a||a|.

And represents the distance between a and 0 on a number line.

An absolute value equation is an equation that contains an absolute value


expression.

The equation |x|=a | x |=a has two solutions x = a and x = -a because both
numbers are at the distance a from 0.

To solve an absolute value equation as

|x+7|=14|x+7|=14

You begin by making it into two separate equations and then solving them
separately.

x+7=14x+7=14
x+7−7=14−7x+7−7=14−7
x=7x=7

or

x+7=−14x+7=−14
x+7−7=−14−7x+7−7=−14−7
x=−21x=−21

An absolute value equation has no solution if the absolute value expression


equals a negative number since absolute number can never be negative.

The inequality

|x|<2|x|<2

Represents the distance between x and 0 that is less than 2

89
Whereas the inequality

|x|>2|x|>2

Represents the distance between x and 0 that is greater than 2

You can write an absolute value inequality as a compound inequality.

\left | x \right |<2\: or −2<x<2−2<x<2

This holds true for all absolute value inequalities.

|ax + b|<c, where c>0|ax+b|<c, where c>0

=−c<ax + b<c=−c< ax + b <c


|ax +b|>c, where c>0|ax+b|>c, where c>0
=ax + b <−c or ax + b >c=ax + b <−c or ax + b>c

You can replace > above with ≥ and < with ≤.

When solving an absolute value inequality, it's necessary to first isolate the
absolute value expression on one side of inequality before solving the
inequality.

Example.

Solve the absolute value inequality

2|3x+9|<362|3x+9|<36
2|3x+9|2<3622|3x+9|2<362
|3x+9|<18|3x+9|<18
−18<3x+9<18−18<3x+9<18
−18−9<3x+9−9<18−9−18−9<3x+9−9<18−9
−27<3x<9−27<3x<9

90
−273<3x3<93−273<3x3<93
−9<x<3

An inequality From Calculus


One of the most important inequalities in calculus is
0 < │x - a│ and │x - a│< 𝛿 (3)
where 𝛿 (𝐺𝑟𝑒𝑒𝑘 dela) is a positive real number. This is equivalent to the two
inequalities
0 < │x-a│ and │x - a│<𝛿
The first of which is satisfied by all x except x = a, and the second of which
is satisfied by all x that are within 𝛿 units of a on a coordinate line.
Combining these two restrictions, we conclude that the solution set of (3)
consists of all x in the interval (a - 𝛿, a + 𝛿 ) except x = a (Figured E.5).
Stated another way, the solution set of (3) is
(a – 𝛿, a) ∪ (a, a + 𝛿) (4)

The Triangle Inequality


It is not generally true that │a + b│ +│a│+│b│
Example
If a = 1 and b = -1, then │a + b│ = 0, here │a│+│b│= 2. It is true, however,
that the absolute value of a sum is always less than or equal to the sum of
the absolute values. This is the content of the following useful theorem,
called triangle inequality.
E.5 Theorem (triangle Inequality)
If a and b are any real numbers, then

│a + b│≤ │𝑎│ + │𝑏│ (5)


Proof.
Observe first that a satisfies the inequality

-│a│≤ 𝑎 ≤ │𝑎│
because either = │a│or a = -│a│, depending on the sign of a. the
corresponding inequality for b is
-│b│≤ b ≤│b│
Adding the two inequalities we obtain
-(│a│+ │b│) ≤ a + b ≤ a (│a│+│b│) (6)

91
Let us now consider the cases a + b ≥ 0 and a + b < 0 separately. In the
case, a + b =│a + b, so the right in (6) yields the triangle inequality (5).
In the second case, a + b = -│a + b│., so the left-handed inequality in (6)
can be written as
- (│a│+ │b│) ≤ --│a│+ │b│
which yields the triangle inequality (5) on multiplying by -1. ∎

Real Number

Definition:
If a is a real number, then we define the absolute value of the
number a
denoted ∣a∣ or abs(a) as:
∣a∣=⎧⎩⎨a0−aifa>0,ifa=0,ifa<0.

Example:
Suppose we want to find the absolute value of 5. Well since 5>0, we note
that ∣5∣=5.
If we wanted to find the absolute value of −5 then since −5<0 we note
that ∣−5∣=−(−5) = 5.
We will now look at some important properties of the absolute values of
real numbers utilizing the The Order Properties of Real Numbers.

Theorem 1:
If a is a real number then ∣a∣=∣−a∣.

Proof:
We will split this proof up into three cases.
Case 1:
Suppose that a>0. Then −a<0.
Therefore, by the definition of the absolute value of a number, ∣a∣=a,
and ∣−a∣=−(−a) = a, and so ∣a∣=∣−a∣
Case 2:
Now suppose that a=0.
Therefore −a=0 and clearly ∣a∣=0 and ∣−a∣=0, and so ∣a∣=∣−a∣.
Case 3:
Lastly suppose that a<0.
Then −a>0.
We obtain that ∣a∣=−a and ∣−a∣=−a, and so ∣a∣=∣−a∣.
In all three cases we get that ∣a∣=∣a∣. ■

Theorem 2:

92
If a and b are real numbers then ∣ab∣=∣a∣∣b∣.

Proof:

We will split this proof up into three cases.

Case 1:

Suppose that a=0 or b=0 or both a, b=0. Then a⋅ b=0, and so ∣ab∣=0.
Similari ty ∣a∣∣b∣ is either 0⋅∣b∣ or ∣a∣⋅0 or 0⋅0, all of which equal 0,
so ∣ab∣=∣a∣∣b∣.

Case 2:

Suppose that a, >0. Then ab>0 and so ∣ab∣=ab and ∣a∣∣b∣=ab,


so ∣ab∣=∣a∣∣b∣.

Case 3:

Suppose that one of a>0 and b<0. Then ab<0. So ∣ab∣=−ab,


and ∣a∣∣b∣=a⋅−b=−ab, so ∣ab∣=∣a∣∣b∣.

Case 4:

Suppose that a, b<0. Then ab>0 and so ∣ab∣=ab and ∣a∣∣b∣=−a⋅−b= (−1)
(−1) ab=ab. So ∣ab∣=∣a∣∣b∣.

In all four cases we get that ∣ab∣=∣a∣∣b∣. ■

Theorem 3: If a is a real number then ∣a∣2=a2.

Proof:

We know that a2>0 and there by applying Theorem 2 we get


that a2=∣a2∣=∣a⋅ a∣=∣a∣∣a∣=∣a∣2. ■

Theorem 4: If c≥0 then ∣a∣≤c if and only if −c ≤ a ≤ c

Proof:

⇒ If ∣a∣≤c then we have that both a ≤ c and −a ≤ c or rather a≥−c which


is equivalent to saying that − c≤ a ≤ c.

⇐ Suppose that −c ≤ a ≤ c. Then a ≤ c and −c ≤ a ⇔ c ≥−a so


then ∣a∣≤c. ■

93
Exercise

1. │3x + 12│+ 7 =7

2. │3x – 7 │+ 7 = 9

3. │𝑥 2 + 1 │= 5

4. │𝑥 2 + 5x + 4│= 0

5. │x + 3│= 𝑥 2 – 4x – 3

94
Answers Key

1.
│3x + 12│+ 7 = 7
│3x + 12│= 0
3x + 12 = 0
3x = -12
x = -4

2.
│3x – 7 │+ 7 = 9
│3x – 7 │ = 2
3x -7 = 2 or 3x – 7 = -2
3x = 9 or 3x = 5
5
x=3 or x = 3

3.
│𝑥 2 + 1 │= 5
𝑥 2 + 1 = 5 or 𝑥 2 + 1 = −5

𝑥 2 = 4 or 𝑥 2 = -6

√𝑥 2 = √4 or √𝑥 2 = √−6
x = ± 2 or x = imaginary
4.
│𝑥 2 + 5x + 4│= 0

𝑥 2 + 5x + 4 = 0
(x + 1) (x + 4) = 0
x + 1 = 0 and x + 4 = 0
x = -1 and x = -4
5.
│x + 3│= 𝑥 2 – 4x – 3

x + 3 = 𝑥 2 – 4x – 3
𝑥 2 – 5x -6
(x – 6) (x + 1) = 0

95
x – 6 = 0 and x + 1 = 0
x = 6 and x = -1

COMPLETENESS PROPERTY OF REAL


NUMBER

ZOCHEL BASLOT

96
COMPLETENESS PROPERTY OF REAL NUMBERS

The completeness property is also known as the least upper bound property.

Upper and Lower Bounds


Definition: Let S be a nonempty subset of R.
 The set S is said to be bounded above if there exists a number u ∈ R
such that s < u for all s E S. Each such number u is called an upper
bound of S.
 The set S is said to be bounded below if there exists a number w ∈ R
such that s > w for all s E S. Each such number w is called an lower
bound of S.
 A set is said to be bounded if it is both bounded above and below. A
set is said to be unbounded if it is not bounded.

Definition: Let S be a subset of R.


 S is said to be bounded above if there exists a number M in R such
that X < M for every x in S. M is called an upper bound for S.
 S is said to be bounded below if there exists a number m in R such
that x > m for every x in S. m is called a lower bound for S.
 S is said to be bounded if it is bounded above and below.
 S is said to be unbounded if it lacks either an upper bound or a lower
bound.

Theorem: Let S be a non-empty subset of R. An upper bound Mo of S satisfies o =


sup S, if and only if for each y < Mo, there exists an x in S for which
y < x < Mo

Proof: We need to prove both directions.


1. Let us assume that Mo = sup S . We need to prove that for each y < Mo,
there exists an x in S for which y < x < Mo. We do a proof by
contradiction .Let y < Mo be given and assume that there is no element x
of S such that y < x. Then, for every x in S, x in S, x < y. Thus, y is an

97
upper bound of S which is smaller than Mo which contradicts the fact that
Mo is the supremum.
2. Let Mo by an upper bound of S with the property that for each y < Mo,
there exists an x in S for which y < x < Mo. We need to show that Mo =
sup S. Since Mo is already an upper bound, it is enough to show it is the
smallest. If Y were an upper bound strictly smaller than Mo, then by
assumption, there would exists an x in S for which y < x < Mo. But then Y
would not be an upper bound of S, which contradicts our assumption.
Thus, there cannot be an upper bound of S smaller than Mo. It follows
that Mo = sup S.
Theorem: Every nonempty subset of R that is bounded below has an infimum in R.
Proof.
We do a direct proof. We will prove the infimum exists by finding it. Let S
be a nonempty subset of R which is bounded below. Define L to be the set
of lower bounds of S. Since S is bounded below, L = ∅. Furthermore, L is
bounded above by elements of S. By the supremum property, L has a
supremum. Call it 𝛼 that is 𝛼 = sup L. We will show that 𝛼 = inf S. To
prove that 𝛼 = inf S, we first prove that 𝛼 is a lower bound of S. We then
prove that no lower bound greater than 𝛼 can exist, making 𝛼 the greatest
lower bound of S.

First, we prove that is a lower bound of S. For this, we need to show


that every element of S is larger than 𝛼. Let s ∈ S. Then s is an upper
bound of L. Since 𝛼 = sup L, that is 𝛼 is the least upper bound of L, it
follows that 𝛼 ≤ s. We have proven that if s is an arbitrary element of S,
then we had s ≥ 𝛼. It follows that 𝛼 is a lower bound of S.

Next, we show that 𝛼 is the greatest of the lower bounds of S. This is


straightforward. If 𝛾 is another lower bound of S, then 𝛾 is an element of
L and therefore 𝛾 ≤ 𝛼 since 𝛼 is the least upper bound of L hence an upper
bound of L. Therefore 𝛼 is the greatest lower bound (or the infimum) of S.

Remark: In the first part of the proof, where we proved that 𝛼 is a lower bound of S,
it would have been wrong to say 𝛼 is a lower bound because 𝛼 = sup L
and L is the set of lower bounds of L. It is wrong because the supremum or
the infimum of a set do not necessarily below to the set. Thus 𝛼 is not
necessarily a lower bound of S. It turns out that it is. But we know this after
the proof we gave.

Lemma: Suppose that S is a nonempty subset of R, 𝛼 is a lower bound of


S and 𝛽 is an upper bound of S. Then, we must have 𝛼 ≤ 𝛽.

Proof. Since S ≠ ∅, we can choose x ∈ S. Because 𝛼 is a lower bound of S,


we have 𝛼 ≤ x. Because 𝛽 is an upper bound of S, we have x ≤ 𝛽. Using
the transitivity property of ≤, we obtain 𝛼 ≤ 𝛽.

98
Proposition: Let S be a subset of R.

1. If S has a smallest element, then min S = inf S.


2. If S has a largest element, then max S = sup S.

Proof. We prove part 1 and leave part 2 as an exercise. Let m = min S. By


definition, m ≤ s for any s ∈ S. Thus m is also a lower bound of S. If 𝛾 is
another lower bound of S, then 𝛾 ≤ m since m ∈ S. Thus m is the greatest
lower bound of S or m = inf S.

Bounded Sets

Maximum and Minimum of a Set

Definition: Let S be a subset of R

1. An element x0 of S is said to be a maximum of S if x0 ≥ x for every


other x in S. In this case, we say that x0 is the largest element of S
and we write x0 = max S.

2. An element x1 of S is said to be a minimum of S if x1 ≤ x for every


other x in S. In this case, we say that x1 is the smallest element of S
and we write x1 = min S.

3. An extremum is either a maximum or a minimum.

Remark: An extremum for a set S is always an element of S.

Completeness of R

Completeness

R is an ordered Archimedean field so is Q. What makes R special is that it


is complete. To understand this notion, we first need a couple of
definitions:

Definition: Given an ordered set X and A ⊂ X, an element x ∈ X is called an upper


bound of A if x ≥ a, ∀a ∈ A.

A special kind of upper bound is

Definition: s ∈ X is called the least upper bound of A, denoted by l.u.b. A or sup A


if

(i) s is an upper bound of A.


(ii) if x ∈ X is an upper bound of A, then x ≥ s.

99
Lemma: The least upper bound of a set A, if it exists, is unique.

Proof. Let s1 and s2 be two least upper bounds of A. Now since s1 is an upper
bound of A by (i) of definition and s2 is a least upper bound, (ii) of
definition shows that s1 ≥ s2. Similarly since s2 is an upper bound
of A and s1 is a least upper bound, we have s2 ≥ s1. Thus s1 = s2.

Theorem: N is unbounded.

Proof. If N is bounded, then by the completeness axiom, b = l.u.b N exists.


Since b − 1 < b there is an integer n ∈ N so that n > b − 1 (otherwise
b-1 would be an upper bound which is impossible). But then n + 1 > b, a
contradiction.

Corollary: For any x ∈ R there is a positive integer n so that n > x.

Proof. If not, x would be an upper bound for N contradicting previous theorem.

Definition: A set A ⊆ R is bounded above if there exists ab ∈ R such that a ≤ b


for all a ∈ A.

The number b is called an upper bound for A.

Definition: If an upper bound b for E is a member of E, then b is called the


maximum (or largest element) of E, and we write

b = max E

Similarly, if a lower bound of E is an element of E, then it is called the


minimum (or least element) of E, denoted by min E.

A set may have upper or lower bounds, or it may have neither. Note that a
set may have many upper and lower bounds, but if it has a maximum or a
minimum, those values are unique. Thus, we speak of an upper bound and
the maximum.

Definition: Let S be an ordered set and let E be a nonempty subset of S. Suppose


there is 𝛼 ∈ S with the following properties:

i) 𝛼 is an upper bound of E
ii) Ɐ𝛼 ∈S, 𝛾 < 𝛼 ⇒ 𝛾 is not an upper bound of E.

Then 𝛼 is called the least upper bound of E or the supremum of E, and


we write 𝛼 = sup E. Similarly, if there exists 𝛽 ∈ S such that

i) 𝛽 is a lower bound of E
ii) 𝛾 ∈ S, 𝛾 > 𝛽 ⇒ 𝛾 is not a lower bound of E.

100
Then 𝛽 is called the greatest lower bound of E or the infimum of E, and
we write 𝛽 = inf E.

As you have probably noticed, the difference between the maximum


and the least upper bound is that the least upper bound is not
necessarily an element of set E. The same is true between the
minimum and the greatest lower bound.

The Completeness axiom

Now we are in condition to introduce the main distinctive characteristic


of R, known as the completeness axiom

Definition. An ordered set S is said to satisfy the least -upper-bound property


(or the completeness axiom) if for every nonempty subset E ⊆ S that is
bounded above has a least upper bound. That is sup E exists in S.

While the completeness axiom refers only to sets that are bounded above,
the corresponding property for sets bounded below follows readily. The
following theorem tells you that every ordered set with the least-upper-
bound property also has the greatest-lower-bound property.

Theorem: Suppose S is an ordered set with the least-upper-bound property. Let


B be a nonempty and bounded below subset of S. Let L be the set of all lower
bounds of B. Then ∃ ⍺ ∈ S : ⍺ = sup L = inf B.

Proof. Since B is bounded from below, L is nonempty. Moreover L = {y ∈


S : y ≤ x for all x ∈ B} ⊆ S implies L is bounded from above. Finally,
since S has the least upper bound property, then ∃ ⍺ ∈ S : ⍺ = sup L.
Now, if 𝛾 < ⍺ then 𝛾 is not an upper bound of L; hence 𝛾 ∉ B and it
follows (by contrapositive) that ⍺ ≤ x for all x ∈ B. Thus ⍺ ∈ L.
Moreover, if ⍺ < 𝛽 then 𝛽 ∉ L since ⍺ = sup L. We have shown that ⍺
∈ L but 𝛽 ∉ L if 𝛽 > ⍺ which means that ⍺ is a lower bound of B and 𝛽
is not if 𝛽 > ⍺. Hence ⍺ = inf B.∎

Properties of Real Numbers

Let a, b, and c represent real numbers.

Property

Closure Property of Addition

a + b is a real number.

Closure Property of Multiplication

101
ab is a real number.

Commutative Property of Addition

a+b=b+a

Commutative Property of Multiplication

ab = ba

Associative Property of Addition

(a + b) + c = a + (b + c)

Associative Property of Multiplication

(ab)c = a(bc)

Distributive Properties

a (b + c) = ab + ac
(a + b) c = ac + ab

Additive Identity Property

a+0=0+a=a
Multiplicative Identity Property

a.1 = 1.a = a

Additive Inverse Property

a + (-a) = 0

Multiplicative Inverse Property

a.1/a = 1, a ≠ 0.

Example 1: The smallest element (or minimum) of [0, 1] is 0: Its largest


element (or maximum) is 1. More generally, if a and b are two real
numbers such that a ≤ b then min [a, b] = a and max [a, b] = b.

Example 2: The minimum of [0, 1) is 0. It does not have a maximum. To be a


maximum, a number 𝛼 would have to be in [0, 1). Thus, we would have
𝛼+1 𝛼+1
𝛼 < 1. But then 𝛼 < 2 < 1, so 2 is also an element of [0, 1) which
Is larger than 𝛼. This contradicts the fact that 𝛼 = max ([0, 1)).

Theorem: Any nonempty set of real numbers which is bounded above has
a supremum.
102
Proof.
We need a good notation for a real number given by its decimal
representation. A real number has the form

a = a0.a1a2a3a4... where a0 is an integer and a1, a2, a3, ... 2 {0, 1, 2, ...9}

To eliminate ambiguity in defining real numbers by their decimal


representation, let us decide that if the sequence of decimals ends up with
nines: a = a0.a1a2...an9999... (where an < 9) then we choose this number’s
decimal representation as a = a0.a1a2...(an + 1)0000.... (For example,
instead of 0.4999999.. we write 0.5.)

Let S be a nonempty set of real numbers, bounded above. Let us construct


the least upper bound of S. Consider first all the approximations by
integers of the numbers a of S: if a = a0.a1a2... collect the a0’s. This is a
collection of integer numbers. It is bounded above (by assumption). Then
there is a largest one among them, call it B0.

Next collect only the numbers in S which begin with B0. (There are some!)
Call their collection S0. Any number in S \S0 (number of S not in S0) is
smaller than any number in S0. Look at the first decimal a1 of the numbers in
S0. Let B1 be the largest among them. Let S1 be the set of all numbers in S0
whose first decimal is B1. Note that the numbers in S1 begin with B0.B1
Also note that any number in S \ S1 is smaller than any number in S1. Next
look at the second decimal of the numbers in S1. Find the largest, B2 etc.
Repeating the procedure we construct a sequence of smaller and smaller
sets S0, S1, S2, ...Sn, ...

S ⊃ S0 ⊃S1 ⊃ S2 ⊃... ⊃ Sn ⊃ ...

Note that every set Sn contains al least one element (it is not empty). At
each
step n we have constructed the set Sn of numbers of S which start with
B0.B1B2...Bn; the rest of the decimals can be anything. Also all numbers
S \ Sn are smaller than all numbers of Sn. (The construction is by induction!)

We end up with the number B = B0.B1B2...BnBn+1.... We need to show that


B
is the least upper bound. To show it is an upper bound, let a ∈ S. If a0 < B0
then a < B. Otherwise a0 = B0 and we go on to compare the first decimals.
Either a1 < B1 therefore a < B or, otherwise, a1 = B1. Etc. So either a < B or
a = B. So B is an upper bound.

To show it is the least (upper bound), take any smaller number t < B. Then t
differs from B at some first decimal, say at the nth decimal:
t = B0.B1B2...Bn−1tntn+1... and tn < Bn. But then t is not in Sn and Sn

103
contains numbers bigger than t.

(−1)𝑛
Example 3: Consider the set A = f{ : n ∈ N}.
𝑛

a. Show that A is bounded from above. Find the supremum. Is this supre-
mum a maximum of A?

SOLUTION:

Clearly, 1/2 is an upper bound of A: Let M > 0 be an upper bound of


A: We will show that 1/2 ≤ M: Suppose the contrary. That is, suppose
that
(−1)𝑛
M < 1/2: Since M is an upper bound of A, we have ≤ M for all n ∈
𝑛
N: In particular, letting n = 2 we obtain 1/2 ≤ M < 1/2 which is
impossible. Thus, 1/2 ≤ M so that sup {A} = 1/2: Since the supremum
is an element of A we conclude that 1/2 is also the maximum of A:

b. Show that A is bounded from below. Find the infimum. Is this infimum
a minimum of A?

SOLUTION:

Clearly, -1 is a lower bound of A: Let m be a lower bound of A: We will


show that m ≤ -1: Suppose the contrary. That is, suppose that m > -1:
(−1)
Letting n = 1 we find that -1 = 1 ≥ m > -1; which is impossible.
Therefore, we must have m ≤ -1: This establishes that inf {A} = -1: Since
-1 is in A; it is the minimum of A.

104
EXERCISES

1. Consider the set A = { x ∈ ℝ : 1 < x < 2 }.

(a) Show that A is bounded from above. Find the supremum. Is this
supremum a maximum of A?
(b) Show that A is bounded from below. Find the infimum. Is this infimum a
minimum of A?

2. Consider the set A = { x > 0 : x2 > 4 } = { x > 0 : x > 2 }.

(a) What is a lower bound of A?


𝐿+2
(b) Let L be a lower bound of A such that L > 2. Let y = 2 Show that
2<y<L.
(c) Show that y ∈ A and L ≤ y: Show that this leads to a contradiction.
Hence, we must have L ≤ 2 which means that 2 is the infimum of A:

3. Suppose that α = sup A < ∞. Let 𝜖> 0 be given. Prove that there is an
x ∈ A such that 𝛼 − 𝜖 < x.

4. Suppose that 𝛽 = inf A < ∞: Let 𝜖 > 0 be given. Prove that there is an
x ∈ A such that 𝛽 +𝜖 > x.

5. Consider the numbers s1, s2… where s1 = √2 and sn+1 = √2 + 𝑠𝑛 for n ∈ N:


Show that each of these numbers is irrational.

105
ANSWER KEY

1. Consider the set A = { x ∈ ℝ : 1 < x < 2 }.

(a) Show that A is bounded from above. Find the supremum. Is this
supremum a maximum of A?
(b) Show that A is bounded from below. Find the infimum. Is this infimum a
minimum of A?

Solution:

(a) Clearly, 2 is an upper bound of A: Let M > 1 be an upper bound of


A: We will show that 2 ≤ M: Suppose the contrary. That is, suppose
that 1 < M < 2: Let r be a rational number such that M < r < 2: Then
r ∈ A and M < r which contradicts the fact that M is an upper bound
of A. Hence, we must have 2 ≤ M so that sup {A} = 2. Since the
supremum is not an element of A we conclude that 2 is not a
maximum of A.
(b) Clearly, 1 is a lower bound of A. Let m be a lower bound of A. We
will show that m ≤ 1. That is, suppose that 1 < m < 2.

Let r be rational number such that 1 < r < m. Then r ∈ A and r < m
which contradicts the fact that m is a lower bound of A. Thus, we
must have 1 ≤ m so that inf {A} = 1. Since 1 is not in A, it is not a
minimum of A. ∎

2. Consider the set A = { x > 0 : x2 > 4 } = { x > 0 : x > 2 }.

. (a) What is a lower bound of A?


𝐿+2
(b) Let L be a lower bound of A such that L > 2. Let y = 2 Show that
2<y<L.
(c) Show that y ∈ A and L ≤ y: Show that this leads to a contradiction.
Hence, we must have L ≤ 2 which means that 2 is the infimum of A:

Solution:

(a) Since 2 ≤ x for all x ∈ A, 2 is a lower bound of A.


𝐿+2
(b) Since L > 2 we have L + 2 > 4 and this implies y = 2 > 2. Also,
𝐿+2 𝐿+𝐿
y = 2 < 2 = L.
(c) Since y > 2 we have y2 > 4 so that y ∈ A: But L is a lower bound of
A so we must have L ≤ y: But this contradicts y < L from (b). It follows
that 2 is the greatest lower bound of A. ∎

3. Suppose that α = sup A < ∞. Let 𝜖 > 0 be given. Prove that there is an
x ∈ A such that 𝛼 − 𝜖 < x.

106
Solution:

Suppose the contrary. That is, 𝛼 − 𝜖 ≥ x for all x ∈ A. In this case,


𝛼 − 𝜖 is an upper bound of A. Thus, we must have 𝛼 ≤ 𝛼 − 𝜖 which
` is impossible. ∎

4. Suppose that 𝛽 = inf A < ∞: Let 𝜖 > 0 be given. Prove that there is an
x ∈ A such that 𝛽 +𝜖 > x.

Solution:

Suppose the contrary. That is, 𝛽 + 𝜖 ≤ 𝑥 for all x ∈ A: In this case,


𝛽 + 𝜖 is a lower bound of A: Thus, we must have 𝛽 + 𝜖 ≤ 𝛽 which is
Impossible.

5. Consider the numbers s1, s2… where s1 = √2 and sn+1 = √2 + 𝑠𝑛 for n ∈ N:


Show that each of these numbers is irrational.

Solution:

We proof this claim by induction on n: For n = 1 we have s 1 = √2


which is an irrational number. Suppose that sk is irrational for
k = 1, 2, … n: We want to show that sn+1 is irrational. Suppose the
contrary, then sn = s2n+1 − 2 is rational which contradicts the
assumption that sn is irrational. Hence, sn+1 must be irrational

107
APPLICATIONS OF SUPREMUM PROPERTY

JEASON GALINDEZ

108
The Supremum Property

Definition:
Every nonempty set of real numbers that is bounded above has a
supremum which is a real number.

Definition:
Every nonempty set of real numbers that is bounded below has an
infimum which is a real number.

Theorem:
The Supremum Property and the Completeness Axiom are equivalent.
This is an if and only if statement.

Proof of the Supremum Property I

Proof.
Assume the Completeness Axiom and show that sup X and inf X exist
and are a real numbers.
Let X ⊆ R be a nonempty set that is bounded above.
Let U be the set of all upper bounds for X.
Since X is bounded above, U ≠ ∅. If x X and u U, x ≤ u, since u is
an upper bound for X.
So, x ≤ u ∀x X; u U
By the Completeness Axiom,
Ǝα R such that x ≤ α ≤ u ∀x X, u U
α is an upper bound for X, and it is less than or equal than every other
upper bound for X, so it is the least upper bound for X, so sup X = α
R.
The case in which X is bounded below is similar. (show it) Thus, the
Supremum Property holds.

Proof of the Supremum Property II

Assume the Supremum Property and show that the completeness axiom
holds.

Suppose L, H ⊆ R, L ≠ Ø H, and,
l ≤ h for al l L, h H

Since L ≠ Ø and L is bounded above ( by any element of H), sup L


exist and it is real number, so let α = supL

By the definition of supremum, α is an upper bound for L, so

l ≤ α for all l L

Suppose h H. Then h is an upper bound for L, so by the definition


of supremum, α ≤ h.

Therefore, we have shown that

109
l ≤ α ≤ h for al l L, h H
so the Completeness Axiom holds.

The Archimedean Property

The Supremum Property:


Every nonempty set of real numbers that is bounded above has a
supremum, which is a real number.
Every nonempty set of real numbers that is bounded below has an
inÖmum, which is a real number.
The supremum property is useful to prove other properties of real
numbers.

Definition:
For all x,y R, y > 0, Ǝn N such that x < ny = (y + . . . + y) → n times.

Theorem:
The set of real numbers (an ordered field with the Least Upper Bound
property) has the Archimedean Property.
This is the proof I presented in class. It is one of the standard proofs.
The key is the following Lemma.
Lemma:
The set N of positive integers N = {0, 1, 2, . . .} is not bounded from above.
Proof Reasoning by contradiction, assume N is bounded from above.
Since N ⊂ R and R has the least upper bound property, then N has a
least upper bound α ∈ R.

Thus n ≤ α for all n ∈ N and is the smallest such real number.


Consequently α − 1 is not an upper bound for N (if it were, since α − 1 <
α, then α would not be the least upper bound).
Therefore, there is some integer k with α − 1 < k. But then α < k + 1.
This contradicts that α is an upper bound for N.

Proof:

Since x > 0, the statement that there is an integer n so that nx > y is


equivalent to finding an n with n > y/x for some n, But if there is no such
n then n < y/x for all integers n.

That is, y/x would be an upper bound for the integers. This contradicts
the Lemma.

Maximum and Minimum

Given a set X ⊆ R we say a is a maximum for X denoted max X if

A X and a c x for all x X

110
And we say b is a minimum for X if,

B X and b ≤ for all x X

Maximum and minimum do not always exist if the set is bounded, but the
sup and the inf do always exist if the set is bounded.
If the sup and inf are also elements if the set, then they coincide with the
max and the min.

Simple Result

Theorem:
Given a set X ⊆ R, if max X exist it is equal to sup X.

Proof:

Let a = max X. We need to show that a equals sup X.

TO do this we must show: that (1) a is an upper bound, and (2) for
every other upper bound a’. we have a’ ≥ a.

is easy: for all y < sup X, Ǝx X such that x > y


a is an upper bound on X since a ≥ x for all x X by definition of max X.

is also easy: Let a’ ≠ a be and upper bound ( a’ ≥ x for all x € X) and


suppose a’ < a.

Since a X ( because a = max X), this yields an immediate contradiction


(a’ cannot be an upper bound).

Density of the Rationals in ℝ

Density of Rational Number

Definition:
An essential property of the natural numbers is the following induction
principle, which expresses the idea that we can reach every natural
number by counting upwards from one.

Axiom 1. Suppose that A ⊂ N is a set of natural numbers such that: (a)


1 ∈ A; (b) n ∈ A implies (n + 1) ∈ A. Then A = N. Before stating the next
theorem, we give a formal definition of the maximal and minimal
elements of a set.

111
Suppose that A ⊂ R is a set of real numbers. A maximal element of A is
a number M = max A such that M ∈ A and M ≥ x for every x ∈ A. A
minimal element of A is a number m = min A such that m ∈ A and m ≤ x
for every x ∈ A.

It follows immediately from the definition that if A has a maximal or


minimal element, then sup A = max A or inf A = min A. However, sup A
or inf A may exist, even when max A or min A don’t exist, if they don’t
belong to A. The following result states a fundamental property of the
natural numbers called the well-ordering property.

Theorem:
Every nonempty subset of natural numbers contains a minimal
element.

Proof:

We will use induction to show that if a set S ⊂ N contains no minimal


element, then S is the empty set, which proves the result.

Let
A = {n ∈ N : 1, 2, . . . , n /∈ S} .

If 1 ∈/ A, then 1 ∈ S, so 1 a minimal element of S since 1 ≤


n for every n ∈ N, which contradicts our assumption on S.
Hence, 1 ∈ A.

Suppose that n ∈ A for some n ∈ N, meaning that 1, 2, . . . ,


n /∈ S. If n+ 1 ∈/ A, then n+ 1 ∈ S, so n+ 1 is a minimal
element of S. This contradiction shows that n + 1 ∈ A. It
follows by induction that A = N, which implies that
S = ∅.

One can also prove that the well-ordering property of N


implies the induction axiom, so the well-ordering of N is
equivalent to induction. The integers Z are not well-ordered
(for example, Z itself has no minimal element), but a similar
property holds for sets of integers that are bounded from
below.

Corollary:
If a nonempty set S ⊂ Z is bounded from below in R, then S has a
minimal element

Proof.

Since S is bounded from below in R, the Archimedean property of R


implies that S is bounded from below by some integer a ∈ Z. Then
B = {n − a + 1 : n ∈ S} ⊂ N,

112
so it implies that B has a minimal element b ∈ B. It follows that m
= b + a − 1 ∈ S is a minimal element of S.
We can now prove the following (intuitively obvious) lemma,
which we will use to prove the density of Q in R.

Lemma:
If x, y ∈ R satisfy y − x > 1, then there exists m ∈ Z such that x < m < y.

Proof:

We construct m as the minimal integer strictly greater than x. Let


S = {n ∈ Z : x < n} .
Then S ⊂ Z is nonempty (by the Archimedean property) and
bounded from below (by x), so it has a minimal element m ∈ S
with m − 1 ∈/ S. It follows that x < m and x ≥ m − 1, so x < m ≤ x
+ 1 < y, which proves the result.
Theorem:
If x, y ∈ R and x < y, then there exists r ∈ Q such that x < r < y

Proof:
Let = y−x > 0. By the Archimedean property, there exists n ∈ N such
that 0 < 1/n < , which implies that ny − nx > 1. By Lemma 5, there
exists m ∈ Z such that nx < m < ny, which proves the result with r = m/n.

Repeated application of this theorem shows that there are, in fact,


infinitely many rational numbers between any pair of distinct real
numbers. One can also use the result to prove that the irrational
numbers R \ Q are dense in R.

The Density of the Rational/Irrational Numbers

We will now look at a theorem regarding the density of rational


numbers in the real numbers, namely that between any two real
numbers there exists a rational numbers.

Theorem 1:

(The Density of the Rational Numbers):

Let x,y ∈ R be any two real numbers where x < y.


Then there exists a rational number r ∈ Q such that x < r < y.
Proof:
Suppose that x > 0. Since x < y we have that y > 0 and furthermore we
have that y – x > 0.
Now we know by the Archimedean properties that since y – x > 0, then
there exists a natural number n ∈ N such that 1n < y − x.
If we multiply this out we get that 1 < ny − nx or rather nx + 1 < ny. Now
we know that since n > 0 and since x > 0, and by the Archimedean
properties that since nx > 0 then there exists a natural number, call it A
∈ N such that A – 1 ≤ nx < A or equivalently A ≤ nx + 1 ≤ A + 1.

113
Therefore nx ≤ A ≤ nx + 1 ≤ ny and so nx < A < ny and so the rational
number r = An works for x < r < y.

Theorem 2:

(The Density of the Irrational Numbers):

Let x,y ∈ R be any two real numbers where x < y. Then there exists an
irrational number q ∈ R ∖Q such that x < q < y.

Proof:

Consider the real numbers x / √2 and y / √2. By the theorem above there
exists a rational number r such that:

(1)
X / √2 < r < y / √2

We can assume that r ≠ 0 for if r = 0 we can apply theorem 1


again to get an r ∗ ∈ Q such that:

(2)
X / √2 < r∗ < r < y / √2 ⇒ x / √2 <r∗ < y / √2
And we can choose the r∗ instead of r. So we assume that r≠0.
We multiply the inequality at (∗) by √2 to get x < r √2 < y, and let q
= r √2. Since r ≠ 0 and r ∈ supQ we have that q is an irrational
number.

114
Exercises:

1.) Consider the set A = { (−1) / n n : n ∈ N}.


(a) Show that A is bounded from above. Find the supremum. Is this supremum
a maximum of A?
(b) Show that A is bounded from below. Find the infimum. Is this infimum a
minimum of A?

2.) Consider the set A = {x ∈ R : 1 < x < 2}.


(a) Show that A is bounded from above. Find the supremum. Is this supremum
a maximum of A?
(b) Show that A is bounded from below. Find the infimum. Is this infimum a
minimum of A?

3.) Consider the set A = {x > 0 : x 2 > 4} = {x > 0 : x > 2}.


(a) What is a lower bound of A?
(b) Let L be a lower bound of A such that L > 2. Let y = L+2 2 . Show that 2 < y
< L.
(c) Show that y ∈ A and L ≤ y. Show that this leads to a contradiction. Hence,
we must have L ≤ 2 which means that 2 is the infimum of A.

4.) Show that for any real number x there is a positive integer n such that n > x.

5.) Let a and b be any two real numbers such that a < b.
(a) Let w be a fixed positive irrational number. Show that there is a rational
number r such that a < wr < b.
(b) Show that wr is irrational. Hence, between any two distinct real numbers
there is an irrational number.

6.) Suppose that α = sup A < ∞. Let > 0 be given. Prove that there is an x ∈ A such
that α − < x.

7.) Suppose that β = inf A < ∞. Let > 0 be given. Prove that there is an x ∈ A such
that β + > x

8.) For each of the following sets S find sup {S} and inf {S} if they exist. You do not
need to justify your answer.
(a) S = {x ∈ R : x 2 < 5}.
(b) S = {x ∈ R : x 2 > 7}.
(c) S = {−1 n : n ∈ N}.

9.) (a) Show that for any positive numbers a and b we have a+b / 2 ≥ √ab.
(b) Let ai > 0 for i = 1, 2, · · · , n. Suppose that n√ a1a2 · · · an = 1. Use (a) to
show that (1 + a1)(1 + a2)· · ·(1 + an) ≥ 2 n .

10.) Consider the numbers s1, s2, · · · where s1 = √ 2 and sn+1 = √ 2 + sn for n ∈ N.
Show that each of these numbers is irrational.

115
Answers

1.) (a) Clearly, 1 / 2 is an upper bound of A. Let M > 0 be an upper bound of A. We will
show that 1 / 2 ≤ M. Suppose the contrary.

That is, suppose that M < 1 2 . Since M is an upper bound of A, we have (−1)n
n ≤ M for all n ∈ N. In particular, letting n = 2 we obtain 1 / 2 ≤ M < 1 / 2 which is
imposssibe. Thus, 1 / 2 ≤ M so that sup{A} = 1 / 2 .

Since the supremum is an element of A we conclude that 1 2 is also the


maximum of A. (b) Clearly, −1 is a lower bound of A. Let m be a lower bound of A. We
will show that m ≤ −1. Suppose the contrary. That is, suppose that m > −1.

Letting n = 1 we find that −1 = (−1) / 1 ≥ m > −1, which is impossible. Therefore,


we must have m ≤ −1. This establishes that inf{A} = −1. Since −1 is in A, it is the
minimum of A

2.) (a) Clearly, 2 is an upper bound of A. Let M > 1 be an upper bound of A. We will
show that 2 ≤ M.

Suppose the contrary. That is, suppose that 1 < M < 2. Let r be a rational
number such that M < r < 2. Then r ∈ A and M < r which contradicts the fact that M is
an upper bound of A.

Hence, we must have 2 ≤ M so that sup{A} = 2. Since the supremum is not an


element of A we conclude that 2 is not a maximum of A.

(b) Clearly, 1 is a lower bound of A. Let m be a lower bound of A. We will show


that m ≤ 1.

Suppose the contrary. That is, suppose that 1 < m < 2.


Let r be a rational number such that 1 < r < m. Then r ∈ A and r < m which
contradicts the fact that m is a lower bound of A. Thus, we must have 1 ≤ m so
that inf{A} = 1. Since 1 is not in A, it is not a minimum of A

3.) (a) Since 2 ≤ x for all x ∈ A, 2 is a lower bound of A.


(b) Since L > 2 we have L + 2 > 4 and this implies y = L+2 / 2 > 2.
Also, y = L+2 / 2 < L+L / 2 = L.
(c) Since y > 2 we have y 2 > 4 so that y ∈ A. But L is a lower bound of A so we
must have L ≤ y. But this contradicts y < L from (b). It follows that 2 is the
greatest lower bound of A.

4.) Let a = 1 and b = x in the Archimedean property

5.) (a) Since a < b, we have a w < b w ., there is a rational number r such that a w < r
< b w or a < rw < b.
(e) If rw = s with s rational then w = s r which is a rational, a contradiction. Hence,
rw is irrational

116
6.) Suppose the contrary. That is, α − ≥ x for all x ∈ A. In this case, α − is an
upper bound of A. Thus, we must have α ≤ α − which is impossible.

7.) (a) Note that S = {x ∈ R : − √ 5 < x < √ 5}. So √ 5 is an upper bound of S. Let M be
an upper bound of S.
Suppose that M < √ 5. Let r be a rational number such that M < r < √ 5.
Then r ∈ S and M < r.
But this contradicts the fact that M is an upper bound of S. Thus, √ 5 ≤ M so
that sup{S} = √ 5.
Likewise one can show that inf{S} = − √ 5.

(b) sup{S} = ∞ and inf{S} = −∞.


(c) sup{S} = 0 and inf{S} = −1

8.) (a) We have (√ a − √ b) 2 ≥ 0 → a + b ≥ 2 √ ab → a + b 2 ≥ √ ab.

(b) For i = 1, · · · , n we have 1+ai / 2 ≥ √ ai . Multiplying these n inequalitites


we find
( 1 + a1 / 2) (1 + a2 / 2) · · · (1 + an / 2) ≥ √ a1a2 · · · an = 1
…Hence, (1 + a1)(1 + a2)· · ·(1 + an) ≥ 2 n

9.) We proof this claim by induction on n.


For n = 1 we have s1 = √ 2 which is an irrational number. Suppose that sk is
irrational for k = 1, 2, · · · , n. We want to show that sn + 1 is irrational.

Suppose the contrary, then sn = s 2 n+1 − 2 is rational which contradicts


the assumption that sn is irrational. Hence, sn + 1 must be irrational.

10.) We have x = p3 1 + √ 5 =⇒ x 3 = 1+√ 5 =⇒ (x 3−1)2 = 5 =⇒ x 6−2x 3−4 = 0

117
INTERVALS

SALDIE G. INCINADA

118
INTERVALS

Interval Notation
Interval notation uses parentheses and brackets to describe sets of real numbers
and their endpoints.

Learning Objectives
Use interval notation to represent sets of numbers

Key Takeaways
Key Points
A real interval is a set of real numbers with the property that any number that lies
between two numbers included in the set is also included in the set.
The interval of numbers between a and b, including a and b, is denoted [a,b]. The
two numbers a and b are called the endpoints of the interval.
To indicate that an endpoint of a set is not included in the set, the square bracket
enclosing the endpoint can be replaced with a parenthesis.
An open interval does not include its endpoints, and is enclosed in parentheses.
A closed interval includes its endpoints, and is enclosed in square brackets.
An interval is considered bounded if both endpoints are real numbers. An
interval is unbounded if both endpoints are not real numbers.
Replacing an endpoint with positive or negative infinity—e.g., (−∞,b ]
— indicates that a set is unbounded in one direction, or half-bounded.

Key Terms
Interval: A distance in space.
Bounded interval: A set for which both endpoints are real numbers.
Open interval: A set of real numbers that does not include its endpoints.
Endpoint: Either of the two points at the ends of a line segment.
Half-bounded interval: A set for which one endpoint is a real number and the
other is not.
Closed interval: A set of real numbers that includes both of its endpoints.
Unbounded interval: A set for which neither endpoint is a real number.

FORMAL DEFINITION
A “real interval” is a set of real numbers such that any number that lies
between two numbers in the set is also included in the set. For example, the
set of all numbers x satisfying 0≤x≤1 is an interval that contains 0 and 1, as
well as all the numbers between them. Other examples of intervals include the
set of all real numbers and the set of all negative real numbers.
The interval of numbers between a and b, including a and b, is often denoted
[a,b]. The two numbers are called the endpoints of the interval.

Open and Closed Intervals


An open interval does not include its endpoints and is indicated with parentheses.
For example, (0,1) describes an interval greater than 0 and less than 1.
A closed interval includes its endpoints and is denoted with square brackets
rather than parentheses.
For example, [0,1]

119
- describes an interval greater than or equal to 0 and less than or equal
to 1.
To indicate that only one endpoint of an interval is included in that set,
both symbols will be used. For example, the interval of numbers between 1
and 5, including 1 but excluding 5, is written as [1,5)
.
The image below illustrates open and closed intervals on a number line.

Intervals: Representations of open and closed intervals on the real number line.

Bounded and Unbounded Intervals


An interval is said to be bounded if both of its endpoints are real numbers.
Bounded intervals are also commonly known as finite intervals. Conversely, if
neither endpoint is a real number, the interval is said to be unbounded.
For example,
The interval (1,10) is considered bounded; the interval (−∞,+∞) is considered
unbounded.

The set of all real numbers is the only interval that is unbounded at both ends;
the empty set (the set containing no elements) is bounded.

An interval that has only one real-number endpoint is said to be half-bounded,


or more descriptively, left-bounded or right-bounded. For example, the interval
(1,+∞) is half-bounded; specifically, it is left-bounded.

The Characterization Theorem for Intervals


We have just looked at Open and Closed Intervals. The following theorem will
verify that if we have an interval, then this interval contains all points in
between its endpoints, for example, the interval (2,4) must contain all values
of x such that 2<x<4, e.g., 3∈(2,4).

Theorem 1 (Characterization of Intervals):


Let S⊆R that contains at least two points. Then if S has the property such that
if x<y and x,y]⊆S, then S is an interval.

Proof:

120
We will consider four cases to this theorem.

Case 1: Suppose that the set S is bounded. Since this set is nonempty and
bounded, by the completeness property, this set contains an supremum (and
similarly, an infimum).
Let a=infS and let b=supS. Therefore, for all s∈S we have that a≤s≤b and so
S⊆[a,b]. Now we want to show that (a,b)⊆S. Let z∈(a,b). Therefore a<z<b.
Therefore, z is not a lower bound of the set S, so there exists an x∈S such
that x<z. Similarly, z is not an upper bound of the set S so there exists a y∈S
such that z<y. Combining these two inequalities we get that x<z<y. Now since
x<y and [x,y]⊆S, we thus have that z∈[x,y]. Therefore (a,b)⊆S. So S is an
interval. Either S=[a,b], S=[a,b), S=(a,b] or S=(a,b) depending on whether the
end points a,b are contained in S or not.

Case 2: Suppose that S is bounded above only. Then let b=supS. Therefore,
∀s∈S we have that s≤b and so S⊆(−∞,b]. We now want to show that
(−∞,b)⊆S. Let z∈(−∞,b). Then z<b, so z is not an upper bound to this set. So
there exists a y∈S such that z<y.
Furthermore, since this set is not bounded below, then there exists an x∈S
such that x<z. Combining these inequalities we have that x<z<y, and so x<y
and [x,y]⊆S so z∈[x,y], and once again, since z is arbitrary this implies that
(−∞,b)⊆S. Therefore S=(−∞,b] or S=(−∞,b) depending on whether b∈S..

Case 3: Suppose that S is bounded below only. Then let a=infS. Therefore
∀s∈S we have that a≤s and so S⊆[a,∞). We now want to show that (a,∞)⊆S.
Let z∈(a,∞). Then a<z, so z is not a lower bound to this set. So there exists an
x∈S such that x<z.
Furthermore, since this set is not bounded above, then there exists a y∈S
such that z<y. Combining these inequalities we have that x<z<y, and so x<y
and [x,y]⊆S so z∈[x,y], and since z is arbitrary this implies that (a,∞)⊆S.
Therefore S=[a,∞) or S=(a,∞) depending on whether a∈S.

Case 4: Suppose that S is not bounded above and not bounded below. Then
clearly S⊆(−∞,∞). We now want to show that (−∞,∞)⊆S. Let z∈(−∞,∞). Then
since this set is unbounded there exists x,y∈S such that x<z<y. We have that
x<y and [x,y]⊂S, so then z∈[x,y]. Since z is arbitrary this implies that
(−∞,∞)⊆S. Therefore S=(−∞,∞)

The Nested Intervals Theorem


We have just looked at what exactly a Nested Interval is, and we are about to
look at a critically important theorem in Real Analysis. Before we look at the
Nested Intervals Theorem let's first look at the following important lemma that
will be used to prove the Nested Intervals Theorem.

Lemma 1:
Let a<b and let c<d and let I=[a,b] and J=[c,d]. Then I⊆J if and only if
c≤a<b≤d.

Now let's look at the Nested Intervals theorem.

121
Theorem 1:
If the interval In=[an,bn] for n∈N is a sequence of closed bounded nested
intervals then there exists a real number ξ=sup{an:n∈N} such that ξ∈⋂∞n=1In.

Proof of Theorem:

We note that by the definition of nested intervals that In⊆I1 for all n∈N so then
an≤b1.
Now consider the nonempty set A={an:n∈N} that is bounded above by b1.
Thus, this set has a supremum in the real numbers and denote it supA=ξ so
that an≤ξ for all n∈N.
We now want to show that ξ≤bn for all n∈N. We will do this by showing that
an≤bk for all n,k∈N.

First consider the case where n≤k. We thus have that In⊇Ik by the definition of
nested intervals and so by lemma 1 we get that an≤ak≤bk≤bn. Here we see
that an≤bk.

Now consider the case where n>k. We thus have that Ik⊇In by the definition
of nested intervals and so by lemma 1 once again we have that ak≤an≤bn≤bk.
Once again we have that an≤bk.

So then an≤bk for all n,k∈N, and so bk is an upper bound to the set A and so
supA=ξ≤bk for all k∈N. Furthermore we have that ak≤ξ for all k∈N, and so
ξ∈Ik for every k∈N and thus ξ∈⋂∞n=1In and so the set theoretic union is
nonempty. ■

Theorem 2:
If the interval In=[an,bn] for n∈N is a sequence of closed bounded nested intervals
then there exists a real number η=inf{bn:n∈N} such that η∈⋂∞n=1In.

Proof:
We note that by the definition of nested intervals that In⊆I1 for all n∈N so then
a1≤bn.
Now consider the nonempty set B={bn:n∈N} that is bounded below by a1.
Thus this set has an infimum in the real numbers and denote it infB=η so that
η≤bn for all n∈N.
We now want to show that an≤η for all n∈N. We will do this by showing that
ak≤bn for all n,k∈N.
First consider the case where n≤k. We thus have that In⊇Ik by the definition of
nested intervals and so by lemma 1 we get that an≤ak≤bk≤bn. Here we see
that ak≤bn.
Now consider the case where n>k. We thus have that Ik⊇In by the definition
of nested intervals and so by lemma 1 once again we have that ak≤an≤bn≤bk.
Once again we have that ak≤bn.

122
So then ak≤bn for all n,k∈N, and so ak is an upper bound to the set A and so
ak≤η=infB for all k∈N. Furthermore we have that η≤bk for all k∈N, and so η∈Ik
for every k∈N and thus η∈⋂∞n=1In. ■

Theorem 3:
Let A:={an:n∈N} and B:={bn:n∈N}. If supA=ξ and infB=η then if the
interval In=[an,bn] is a sequence of closed bounded nested intervals then
[ξ,η]=⋂∞n=1In.

Proof:

We first prove that [ξ,η]⊆⋂∞n=1In. Now let x∈[ξ,η]={x∈R:ξ≤x≤η. Therefore


ξ≤x≤η. But we know that an≤ξ for all n∈N and we know that η≤bn for all n∈N
and so an≤ξ≤x≤η≤bn for all n∈N. Therefore x∈[an,bn] for all n∈N or in other
words x∈⋂∞n=1In. Therefore [ξ,η]⊆⋂∞n=1In.

We will now prove that ⋂∞n=1In⊆[ξ,η]. Let x∈⋂∞n=1In. Then an≤x≤bn for all
n∈N. We also know that an≤ξ≤η≤bn for all n∈N. Suppose that x is such that
an≤x≤ξ. Since ξ is the supremum of the set A then there exists an element
ax∈{an:n∈N} such that x<ax and so x∉[ax,bx] and therefore x∉⋂∞n=1In, a
contradiction. Now suppose that x is such that η≤x≤bn for all n∈N. Since η is
the infimum of the set B then there exists an element bx∈{bn:n∈N} such that
bx<x and so x∉[ax,bx] and so x∉⋂∞n=1In,
Once again, a contradiction. We must therefore have that an≤ξ≤x≤η≤bn and
so x∈[ξ,η] and so ⋂∞n=1In⊆[ξ,η].
Since [ξ,η]⊆⋂∞n=1In and ⋂∞n=1In⊆[ξ,η] we have that [ξ,η]=⋂∞n=1In

Set-Builder Notation

In this text, we use interval notation. However, other resources that you are likely
to encounter use an alternate method for describing sets called set-builder
notation. We have used set notation to list the elements such as the integers

The braces group the elements of the set and the ellipsis marks indicate that the
integers continue forever. In this section, we wish to describe intervals of real
numbers—for example, the real numbers greater than or equal to 2.

Since the set is too large to list, set-builder notation allows us to describe it using
familiar mathematical notation. An example of set-builder notation follows:

123
Here x∈R describes the type of number, where the symbol (∈) is read “element of.”
This implies that the variable x represents a real number.
The vertical bar (|) is read “such that.”
Finally, the statement x≥2 - is the condition that describes the set using
mathematical notation.
At this point in our study of algebra, it is assumed that all variables represent real
numbers.
For this reason, you can omit the “∈R” and write {x|x≥2}, which is read “the set of all
real numbers x such that x is greater than or equal to 2.”

To describe compound inequalities such as x < 3 or x ≥ 6, write { x Ι x < 3 or x ≥


6 }, which is read “ the set of all real numbers x such that x is less than
or x is greater than or equal to 6”.

Write bounded intervals such as -1 ≤ x < 3 as { x – 1 ≤ x < 3 }, which is read as “ the


set of all real numbers x such that x is greater than or equal to -1 and less than 3”.

Key Takeaways
Inequalities usually have infinitely many solutions, so rather than presenting an
impossibly large list, we present such solutions sets either graphically on a number
line or textually using interval notation.
Inclusive inequalities with the “or equal to” component are indicated with a closed
dot on the number line and with a square bracket using interval notation.
Strict inequalities without the “or equal to” component are indicated with an open dot
on the number line and a parenthesis using interval notation.

124
Compound inequalities that make use of the logical “or” are solved by solutions of
either inequality. The solution set is the union of each individual solution set.
Compound inequalities that make use of the logical “and” require that all inequalities
are solved by a single solution. The solution set is the intersection of each individual
solution set.
Compound inequalities of the form n<A<m can be decomposed into two inequalities
using the logical “and.” However, it is just as valid to consider the argument A to be
bounded between the values n and m.

125
Topic Exercises

Part A: Simple Inequalities

Graph all solutions on a number line and provide the corresponding interval notation.

1. x≤10

2. x>0

3. x≤−3

4. −4<x

5. x<−12

6. x≥−134

Part B: Compound Inequalities

Graph all solutions on a number line and give the corresponding interval notation

7. −2<x<5

8. −5<x≤20

9. 10<x≤40

10. 0<x ≤50

11. −58<x<18

12. −1≤x<112

126
ANSWERS
1: (−∞,10]

2: (0,∞)

3: (−∞,−3]

4: (−4,∞)

5: (−∞,−12)

6: [−134,∞)

7: (−2,5)

8: (−5,20]

9: (10,40]

127
10: (0,50]

11: (−58,18)

128
SEQUENCES and Their LIMITS

YASDAN DIAZ

129
Sequences and Their Limits

Definition:
A sequence of real numbers (or a sequence in R) is a function from N
into R.
Notation:
(1) The values of X : N ! R are denoted as X(n) or xn, where X is the
sequence.
(2) (xn : n ∈ N) or simply (xn) may denote a sequence — this is not the
same as {xn : n ∈ N}.
(3) (x1, x2, . . . , xn, . . .).
Examples:
(1) (3n) = (3n : n ∈ N) = (3, 6, 9, . . . , 3n, . . .).
(2) (1) = (1 : n ∈ N) = (1, 1, 1, . . . , 1, . . .).
(3) ((−2)𝑛 ) =(−2)𝑛 : n ∈ N =( -2, 4, -8, . . . , (−2)𝑛 , . . . .)
1 1 1 1
(4) (2 + 2 (−1)𝑛 )= (2 + 2 (−1)𝑛 : n ∈ N) = (0, 1, 0, 1, . . . , 0, 1, . . .).
𝑛 1 1(−1)𝑛 𝑛 1 1(−1)𝑛 𝑛
(5)( ( 2)2+2 )= ( ( 2)2+2 : n ∈ N )= ( 1, 1, 1, 2, 1, 3, 1, 4, . . , 1, 2 ,
...).

Definition:
Sequences may also be defined inductively or recursively.
Example:
(1) 𝑥1 = 5, 𝑥2 +1 = 2𝑥𝑛 −3 (n≥ 1) gives (5, 7, 11, 19, 35, . . .).
(2) Fibonacci sequence: 𝑥1 = 𝑥2 = 1, 𝑥𝑛+1 = 𝑥𝑛−1 + 𝑥𝑛 (n ≥ 2) gives (1, 1,
2, 3, 5, 8, 13, . . .).

Definition:
A sequence of real numbers converges to a real number a if, for every
positive number ∈ , there exists an N ∈ N such that for all n ≥ N, |an -
a| < ∈ . We call such an a the limit of the sequence and write lim 𝑎𝑛
𝑛→∞
an = a.

Proposition 1:
The sequence 1 n converges to zero.
Proof.
1
Let ∈ > 0. We choose N ∈ N such that N > 1 ∈. Such a choice is
always possible by the Archimedean property. To verify that this choice
1
of N is appropriate, let n ∈ N satisfy n ≥ N. Then, n ≥ N implies n > ∈
1 1
which is equal to 𝑛 = | 1 n −0 | < ∈ , proving that converges to zero
𝑛
by the definition of convergence.

130
Proposition 2.
An example of a sequence that does not converge is the
following:(1, −1, 1, −1, ...) If a sequence does not converge, it is said to
diverge, which we will explain later in the paper, along with the
explanation of why the above sequence does not converge.

Proposition 3.
If 𝑥𝑛 ≤ 𝑦𝑛 ≤ 𝑧𝑛 for all n ∈ N and lim 𝑥𝑛 = lim 𝑧𝑛 =l, then lim 𝑦𝑛 =l too
𝑛→∞ 𝑛→∞ 𝑛→∞
Proof.
Let ∈ > 0. We want to show there exists an N such that for all n > N, | yn
− l |< ∈. We know that xn → l. Therefore, there exists an N1 such that
for all n > N1, | xn − l |< ∈. Also, we know that zn → l. Therefore, there
exists an N2 such that for all n > N2, | zn − l |< ∈. Let N = max(N1, N2)
and n > N. Then, n > N1 so | xn − l |< ∈. Also, n > N2 so | zn − l |< ∈. We
want to show that | yn − l |< ∈. This is equivalent to showing that both yn
− l < ∈ and l − yn < ∈. We know that yn ≤ zn, so yn − l ≤ zn − l < ∈. Also,
we know that yn ≥ xn, so l − yn ≤ l − xn < ∈.
Theorem:
(Algebraic Limit Theorem).
Let lim 𝑎𝑛 = a and lim 𝑏𝑛 = b. Then,
𝑛→∞ 𝑛→∞
(i) lim 𝑐𝑛 = ca for all c ∈ R
𝑛→∞
(ii) lim ( 𝑎𝑛 +𝑏𝑛 ) = a+b
𝑛→∞
(iii) lim ( 𝑎𝑛 𝑏𝑛 ) = ab
𝑛→∞
(iv) lim ( 𝑎𝑛 /𝑏𝑛 = a/b provided b ≠ = 0
𝑛→∞
Example:
If (xn) → 2, then ((2xn - 1)/3) → 1.
Proof.
First, we will start with the information given in the example:
xn → 2.
4 3
Next, we simply use the fact that (3 )(2 ) = 2.
4 3
xn → (3 )(2 )
4 3 2
Now, let an = xn, and let a = (3 )(2 ), and let c = (3 ). From the Algebraic
2 2 4 3
Limit Theorem, we know that 𝑐𝑎𝑛 → ca. Then, (3 )(xn) → (3) (3 )(2 ), which
is equal to the following:
2𝑥𝑛 4
( ) →(3 )
3
4 1
The next step follows from the fact that (3 ) = 1 + (3 ) .
2𝑥𝑛 1
( ) →1 + (3 ) .
3
2𝑥 1 −1 −1 −1
Let ( 3𝑛 ) = an, let (1 + 3 ) = a, let bn = ( 3 , 3 , ...), and let b = 3 . Then,
by the Algebraic Limit Theorem, we know that an + bn → a + b.

131
2𝑥𝑛 −1 1 −1
Therefore, we know that ( ) + → (1 + 3 ) + , which is equal to the
3 3 3
following:
2𝑥𝑛 1
( )- →1
3 3
2𝑥𝑛 1 2𝑥𝑛−1
This last step follows because ( )- =( )=
3 3 3
2𝑥𝑛 1
( )- → 1.
3 3
Therefore, using the Algebraic Limit Theorem, we have shown that if
2𝑥 1
(xn) → 2, then ( 3𝑛 ) - 3 → 1.
Example:
2 +1 2
The following sequence converges to the proposed limit lim (5𝑛+4)= 5
𝑛

Proof.
2 +1 1 2 +1
Let 5𝑛+4 be an, let 5 be bn and let 5𝑛+4be cn, and cn = an + bn. By
𝑛 𝑛 +4 𝑛
Theorem 2.3, we know that lim(cn) = lim (an+bn) = lim(an)+lim(bn). We
must therefore determine what lim(an) and lim(bn) are. First, we will
1
show that lim(5 +4 ) = 0. Let ∈ > 0. By the Archimedean principle, there
𝑛
exists an N ∈ N such that N > 1/∈. Then, for n > N, 1 5n+4 < 1 5N+4 <
1/N < ∈. Therefore, the limit of 1 5n+4 is zero. Then, because lim(cn) =
lim(an+bn), lim(cn) = lim(an + 0) = lim(an). We will therefore find the limit
2 +1 2 2 +1
of an in order to prove lim(5𝑛+4 ) = 5. We now want to show that lim(5𝑛+4
𝑛 𝑛
2
) = 5. Let ∈ > 0. By the Archimedean Principle, there exists an N such
that 1/∈ < N. Let n > N. We then want to show the following:
2 +1 2 2 +1 2 −8
|5𝑛+4 − 5| < ∈, then |5𝑛+4 − 5|=|5(5 |
𝑛 𝑛 𝑛 +4)

We have to check the following:


−8
<∈
5(5𝑛 +4)
8
<∈
5(5𝑛 +4)
−8
We know that the inequality −5(5 < ∈ is true for every value of n
𝑛 +4)
because n > N > 1/∈ and ∈. Therefore we only need to show that the
8
inequality 5(5 +4)< ∈
𝑛

is true. Using the fact that N > 1/∈, we can say the following:
8 8 8∈
< 5(5(1/∈+4)= 25+20∈
5(5𝑛 +4)
8∈ 8∈ 8
Then 25+20∈ < 25< ∈. Therefore 5(5 < ∈.
𝑛 +4)

Example:
Let xn ≥ 0. If (xn) → 0, then (√𝑥𝑛 ) → 0.
Proof.
First, we have to prove that lim( √𝑥𝑛 ) exists. We know that xn is
decreasing but is greater than or equal to 0 for all values of n. The square

132
root of a positive number is also positive. Therefore, √𝑥𝑛 ≥ 0. Also, note
that if 0< a < b, then 0< √𝑎 < √𝑏. So if xn is decreasing, then so is √𝑥𝑛 .
Therefore, lim(√𝑥𝑛 ) exists. Next, we must prove that ( √𝑥𝑛 ) → 0. Let
lim(xn) = lim√𝑥𝑛 √𝑥𝑛 )) = 0. By the Algebraic Limit Theorem, we know that
if lim(an) = a and lim(bn) = b then lim((an)(bn)) = ab. By this theorem,
lim((√𝑥𝑛 )(√𝑥𝑛 )) = lim(√𝑥𝑛 )lim√𝑥𝑛 ) = 0. Thus, (lim(√𝑥𝑛 ))2=0. This implies
that lim(√𝑥𝑛 )=0.

Definition:
A sequence that does not have a limit or in other words, does not
converge, is said to be divergent.
Example:
Recall proposition 2, which says that the following sequence does not
converge: (3.3) (1, −1, 1, −1...).
1 1 1 1 1 1
(1, , , , , , … )
2 3 4 5 6 7
This converges to zero, as we proved earlier in this paper. However,
these sequences do have something in common. They are both
bounded.

Definition:
A sequence (xn) is bounded if there exists a number M>0 such that | xn
| ≤ M for all n ∈ N. Geometrically, this means we can find an interval
[−M,M] that contains every term in the sequence (xn).
Example:
Given the sequence xn = (1, 2, 1, 2, 1, 2 ...), we can see that the interval
[1, 2] contains every term in xn. This sequence is therefore a bounded
sequence.
Example:
Given the sequence xn = (10, 100, 1000, 10000, ...), we can see that
there is no real number that serves as an upper bound because lim(xn)
is infinity. Therefore, there does not exist any interval that contains every
term in the sequence xn, and xn is not a bounded sequence.
Theorem:
Every convergent sequence is bounded.
Example:
𝑛+1
Theorem being illustrated: Let xn = , which is the following sequence:
𝑛
2 3 4 5
( . , , )
1 2 3 4
We know this converges to 1 and can verify this using the same logic
used in the proof under the definition of convergence showing that 1 n
converges to zero. Therefore, as n becomes very large, xn approaches
1, but is never equal to 1. By the above theorem, we know that this
sequence is bounded because it is convergent. We can see that xn is a

133
decreasing sequence, so the x1 is the largest value of the sequence and
is the “upper bound.” The limit of the sequence, 1, is the lower bound.
An interval that contains every term in the sequence xn is (1,2].

Continuity

Theorem:
If f: R → R is continuous, xn → x implies f(xn) → f(x)
Example:
Theorem being applied: Let f(x) = 3x. This function is continuous. Let
lim(xn) = 5. In order words, xn → 5. By the above theorem, this implies
that f(xn) → f(5). This is equal to 3xn →(3)(5) which is also equal to 3xn
→ 15. Therefore, we are able to see what the limit of f(xn) is using this
theorem.
Example:
Theorem failing when function is non-continuous: Let f(x) be 1 x , a non-
continuous function. We know this is non-continuous because there is
an asymptote at x=0. Let xn be 1 n. We know this converges to zero
based on a previous proof. Let’s see if the continuity theorem fails for a
noncontinuous function f. The theorem states that f(xn) converges to f(x)
if xn → x. We know that xn → 0, so if the theorem works, then f(xn) →
f(0). But f(0) = 1 0 which does not exist. Therefore, f(xn) cannot converge
to 1/0, and the theorem fails for this non-continuous function.

Subsequences and the Bolzano-Weierstrass Theorem

Definition:
Let (an) be a sequence of real numbers, and let n1 < n2 < n3 < n4... be
an increasing sequence of natural numbers. Then, the sequence (an1 ,
an2 , an3 , an4 ...) is called a subsequence of (an) and is denoted by
(𝑎𝑛𝑘 ), where k ∈ N indexes the subsequence.
Example:
1 1 1
Let xn = 1 n = (1, 2 , 3 , 4...). Below are two examples of valid
subsequences:
1 1 1 1
( , , , )
3 6 9 12
1 1 1
( , , …)
20 200 2000
Bolzano-Weierstrass Theorem:
Every bounded sequence contains a convergent subsequence.
Example:
Given a sequence xn = (1,2,3,4,1,2,3,4...), a convergent subsequence
can be found.

134
Proof.
We know that this sequence is bounded by the interval [1, 4]. By the
Bolzano-Weierstrass Theorem, we can say that there indeed exists a
convergent subsequence of xn. Just by looking at this sequence, we can
see four convergent subsequences: (1,1,1...), (2,2,2...), (3,3,3...), and
(4,4,4...). These subsequences converge to 1, 2, 3, and 4 respectively.
Example:
Given an unbounded sequence xn = (1,2,3,4,5...), a convergent
subsequence of xn does not exist Proof. A convergent subsequence
does not necessarily exist because this sequence does not satisfy the
Bolzano-Weierstrass Theorem. Recall that any subsequence of a
sequence is non-repeating and in the order of the original entries of xn.
Notice that xn is increasing for all values of n and is divergent,
considering the sequence continues until infinity. Therefore, for any
subsequence an, the values will be increasing toward infinity as well, and
the subsequence will also be divergent.
Theorem:
Subsequences of a convergent sequence converge to the same limit as
the original sequence.
Example:
Let us return to the example of a divergent sequence that was given
under the definition of divergence. Recall that this sequence, xn, was (1,
-1, 1, -1, 1, -1...). One subsequence of xn is (1, 1, 1, ...). This
subsequence converges to 1. Another subsequence of xn is (-1, -1, -1,
...). This subsequence converges to -1. Now, we will prove that xn is
divergent by contradiction. Assume xn is convergent. Then, by the above
theorem, all its subsequences converge to lim(xn), implying that all its
subsequences converge to the same value. The two subsequences of
xn stated above converge to different values. Therefore, this contradicts
our original hypothesis that xn is convergent. We are then able to
conclude that xn is divergent. Acknowledgements. I would like to thank
my two mentors, Sean Howe and Abby Ward for meeting with me each
week for the duration of my stay at the program. Thank you so much for
suggesting this topic, explaining everything to me that I was confused
about, editing and helping with my paper, and calming me down when I
got overwhelmed. I would also like to thank Peter May for organizing this
wonderful REU and for giving me the opportunity to participate.
Theorem:
Let X = (xn) be a sequence in R, and let x ∈ R. The following are
equivalent:
(a) X converges to x.
(b) ∀ ∈ > 0, ∃ K ∈ N ∋ ∀ n ≥ K. |𝑥𝑛 − 𝑥| < ∈.
(c) ∀ ∈ > 0, ∃ K ∈ N ∋ ∀ n ≥ K , x- ∈< 𝑥𝑛 < x+ ∈.
(d) ∀ ∈-nbhd. 𝑉𝜖 (x) of x, ∃ K ∈ N ∋ ∀ n ≥ K , 𝑥𝑛 ∈ 𝑉𝜖 (x) .

The uniqueness of limits


135
The K-ǫ principle
Definition:
Can a sequence have more than one limit? Common sense says no: if
there were two different limits L and L ′ , the an could not be arbitrarily
close to both, since L and L ′ themselves are at a fixed distance from
each other. This is the idea behind the proof of our first theorem about
limits. The theorem shows that if {an} is convergent, the notation lim an
makes sense; there’s no ambiguity about the value of the limit. The proof
is a good exercise in using the definition of limit in a theoretical argument.
Try proving it yourself first.
Theorem:
A Uniqueness theorem for limits.
A sequence an has at most one limit: an → L and an → L ′ ⇒ L = L ′ .
Proof.
By hypothesis, given 𝜖 > 0, an ≈ ǫ L for n ≫ 1, and an ≈ ǫ L ′ for n ≫ 1
Therefore, given ǫ > 0, we can choose some large number k such that L
≈ ǫ ak ≈ ǫ L ′ . By the transitive law of approximation (2.5 (8)), it follows
that (4) given ǫ > 0, L ≈ 2ǫ L ′ . To conclude that L = L ′ , we reason
indirectly (cf. Appendix A.2). Suppose L 6= L ′ ; choose ǫ = 1 2 |L − L ′ |.
We then have |L − L ′ | < 2ǫ, by (4); i.e., |L − L ′ | < |L − L ′ |, a contradiction.
Remarks:
The line (4) says that the two numbers L and L ′ are arbitrarily close. The
rest of the argument says that this is nonsense if L 6= L ′ , since they
cannot be closer than |L − L ′ |.
Theorem:
{an} increasing, L = lim an ⇒ an ≤ L for all n; {an} decreasing, L = lim an
⇒ an ≥ L for all n. Proof. Both cases are handled similarly; we do the
first. Reasoning indirectly, suppose there were a term aN of the
sequence such that aN > L. Choose ǫ = 1 2 (aN − L). Then since {an} is
increasing, an − L ≥ aN − L > ǫ, for all n ≥ N, contradicting the Definition
3.1 of L = lim an.

The K-ǫ principle

Definition:
In the proof of Theorem 3 , note the appearance of 2ǫ in line (4). It often
happens in analysis that arguments turn out to involve not just ǫ but a
constant multiple of it. This may occur for instance when the limit
involves a sum or several arithmetic processes. Here is a typical
example. Example 3.2 Let an = 1 n + sin n n + 1 . Show an → 0, from
the definition. Solution To show an is small in size, use the triangle
inequality: ¯ ¯ ¯ ¯ 1 n + sin n n + 1 ¯ ¯ ¯ ¯ ≤ ¯ ¯ ¯ ¯ 1 n ¯ ¯ ¯ ¯ + ¯ ¯ ¯
¯ sin n n + 1 ¯ ¯ ¯ ¯ . At this point, the natural thing to do is to make the
separate estimations ¯ ¯ ¯ ¯ 1 n ¯ ¯ ¯ ¯ < ǫ, for n > 1 ǫ ; ¯ ¯ ¯ ¯ sin n n
+ 1 ¯ ¯ ¯ ¯ < ǫ, for n > 1 ǫ − 1 ; so that, given ǫ > 0, ¯ ¯ ¯ ¯ 1 n + sin n n

136
+ 1 ¯ ¯ ¯ ¯ < 2ǫ , for n > 1 ǫ . This is close, but we were supposed to
show |an| < ǫ. Is 2ǫ just as good?
The usual way of handling this would be to start with the given ǫ, then
put ǫ ′ = ǫ/2, and give the same proof, but working always with ǫ ′ instead
of ǫ. At the end, the proof shows ¯ ¯ ¯ ¯ 1 n + sin n n + 1 ¯ ¯ ¯ ¯ < 2ǫ ′ ,
for n > 1 ǫ ′ ; and since 2ǫ ′ = ǫ, the limit definition is satisfied. Instead of
doing this, let’s once and for all agree that if you come out in the end with
2ǫ, or 22ǫ, that’s just as good as coming out with ǫ. If ǫ is an arbitrary
small number, so is 22ǫ. Therefore, if you can prove something is less
than 22ǫ, you have shown that it can be made as small as desired. We
formulate this as a general principle, the “K-ǫ principle”. This isn’t a
standard term in analysis, so don’t use it when you go to your next
mathematics congress, but it is useful to name an idea that will recur
often.

Principle 3.2:
The K-ǫ principle. Suppose that {an} is a given sequence, and you can
prove that (5) given any ǫ > 0, an ≈ Kǫ L for n ≫ 1 , where K > 0 is a
fixed constant, i.e., a number not depending on n or ǫ. Then limn→∞ an
= L . The K-ǫ principle is here formulated for sequences, but we will use
it for a variety of other limits as well. In all of these uses, the essential
point is that K must truly be a constant, and not depend on any of the
variables or parameters.

Infinite limits

Definition:
Even though ∞ is not a number, it is convenient to allow it as a sort of
“limit” in describing sequences which become and remain arbitrarily
large as n increases. The definition is like the one for the ordinary limit.
Definition:
We say the sequence {an} tends to infinity if (6) given any M ≥ 0, an > M
for n ≫ 1 . In symbols: limn→∞ {an} = ∞, or an → ∞ as n → ∞.As for
regular limits, to establish that lim{an} = ∞, what you have to do is give
an explicit value for the N concealed in “for n ≫ 1”, and prove that it does
the job, i.e., prove that an > M when n ≥ N. In general, this N will depend
on M: the bigger the M, the further out in the sequence you will have to
go for the inequality an > M to hold. As before, it is not you who chooses
the M; the limit demon does that, and you have to prove the inequality in
(6) for whatever positive M it gives you. Note also that even though we
are dealing with size, we do not need absolute values, since an > M
means the an are all positive for n ≫ 1. One should not think that infinite
limits are associated only with increasing sequences. Consider these
examples, neither of which is an increasing sequence.

An important limit

137
As a good opportunity to practice with inequalities and the limit definition,
we prove an important limit that will be used constantly later on.
Theorem:
The limit of a n.
Proof.
We consider the case a > 1 first. Since a > 1, we can write a = 1 + k, k >
0. a n = (1 + k) n Thus , which by the binomial theorem = 1 + nk + n(n −
1) 2! k 2 + n(n − 1)(n − 2) 3! k 3 + ... + k n . Since all the terms on the
right are positive, a n (5) > 1 + nk ; > M, for any given M > 0, if n > M/k,
say. This proves that lim a n = ∞ if a > 1, according to Definition 3.3. ¤
The second case a = 1 is obvious. For the third, in outline the proof is:
|a| < 1 ⇒ 1 |a| > 1 ⇒ ³ 1 |a| ´n → ∞ ⇒ a n → 0 . Here the middle implication
follows from the first case of the theorem. The last implication uses the
definition of limit; namely, by hypothesis, given ǫ > 0, ³ 1 |a| ´n > 1 ǫ for
n large; by the reciprocal law of inequalities (2.1) and the multiplication
law for | |, |a n | < ǫ for n large.
Why did we begin by writing a = 1 + k? Experimentally, you can see that
when a > 1, but very close to 1 (like a = 1.001), a increases very slowly
at first when raised to powers. This is the worst case, therefore, and it
suggests writing a in a form which shows how far it deviates from 1. The
case a ≤ −1 is not included in the theorem; here the a n alternate in sign
without getting smaller, and the sequence has no limit. A formal proof of
this directly from the definition of limit is awkward; instead we will prove
it at the end of Chapter 5, when we have more technique.
THEOREM:
Suppose that an ≤ bn ≤ cn for all n > N, for some N. If limn→∞ an =
limn→∞ cn = L, then limn→∞ bn = L. And a final useful fact:
THEOREM:
limn→∞ |an| = 0 if and only if limn→∞ an = 0.This says simply that the
size of an gets close to zero if and only if an gets close to zero.
EXAMPLE:
Determine whether n n + 1∞ n=0 converges or diverges. If it converges,
compute the limit. Since this makes sense for real numbers we consider
limx→∞ x x + 1 = limx→∞ 1 − 1 x + 1 = 1 − 0 = 1. Thus the sequence
converges to 1.
EXAMPLE:
Determine whether ln n n ∞ n=1 converges or diverges. If it converges,
compute the limit. We compute limx→∞ ln x x = limx→∞ 1/x 1 = 0, using
L’Hˆopital’s Rule. Thus the sequence converges to 0.
EXAMPLE:
Determine whether {(−1)n } ∞ n=0 converges or diverges. If it converges,
compute the limit. This does not make sense for all real exponents, but
the sequence is easy to understand: it is 1, −1, 1, −1, 1 . . . and clearly
diverges.
EXAMPLE:

138
Determine whether {(−1/2)n } ∞ n=0 converges or diverges. If it
converges, compute the limit. We consider the sequence {|(−1/2)n |}∞
n=0 = {(1/2)n } ∞ n=0. Then limx→∞ 1 2 x = limx→∞ 1 2 x = 0, so by
theorem 11.1.4 the sequence converges to 0.
Sometimes we will not be able to determine the limit of a sequence, but
we still would like to know whether it converges. In some cases we can
determine this even without being able to compute the limit. A sequence
is called increasing or sometimes strictly increasing if ai < ai+1 for all i.
It is called non-decreasing or sometimes (unfortunately) increasing if ai
≤ ai+1 for all i. Similarly a sequence is decreasing if ai > ai+1 for all i and
non-increasing if ai ≥ ai+1 for all i. If a sequence has any of these
properties it is called monotonic.
EXAMPLE:
The sequence 2 i − 1 2 i ∞ i=1 = 1 2 , 3 4 , 7 8 , 15 16 , . . . , is increasing,
and n + 1 n ∞ i=1 = 2 1 , 3 2 , 4 3 , 5 4 , . . . is decreasing.
A sequence is bounded above if there is some number N such that an ≤
N for every n, and bounded below if there is some number N such that
an ≥ N for every n. If a sequence is bounded above and bounded below
it is bounded. If a sequence {an} ∞ n=0 is increasing or non-decreasing
it is bounded below (by a0), and if it is decreasing or nonincreasing it is
bounded above (by a0). Finally, with all this new terminology we can
state an important theorem.
THEOREM:
If a sequence is bounded and monotonic then it converges. We will not
prove this; the proof appears in many calculus books. It is not hard to
believe: suppose that a sequence is increasing and bounded, so each
term is larger than the one before, yet never larger than some fixed value
N. The terms must then get closer and closer to some value between a0
and N. It need not be N, since N may be a “too-generous” upper bound;
the limit will be the smallest number that is above all of the terms ai .
EXAMPLE:
All of the terms (2i − 1)/2 i are less than 2, and the sequence is
increasing. As we have seen, the limit of the sequence is 1—1 is the
smallest number that is bigger than all the terms in the sequence.
Similarly, all of the terms (n+ 1)/n are bigger than 1/2, and the limit is 1—
1 is the largest number that is smaller than the terms of the sequence.
We don’t actually need to know that a sequence is monotonic to apply
this theorem— it is enough to know that the sequence is “eventually”
monotonic, that is, that at some point it becomes increasing or
decreasing. For example, the sequence 10, 9, 8, 15, 3, 21, 4, 3/4, 7/8,
15/16, 31/32, . . . is not increasing, because among the first few terms it
is not. But starting with the term 3/4 it is increasing, so the theorem tells
us that the sequence 3/4, 7/8, 15/16, 31/32, . . . converges. Since
convergence depends only on what happens as n gets large, adding a
few terms at the beginning can’t turn a convergent sequence into a
divergent one.
EXAMPLE:

139
Show that {n 1/n} converges. We first show that this sequence is
decreasing, that is, that n 1/n > (n+1)1/(n+1). Consider the real function
f(x) = x 1/x when x ≥ 1. We can compute the derivative, f ′ (x) = x 1/x(1−ln
x)/x2 , and note that when x ≥ 3 this is negative. Since the function has
negative slope, n 1/n > (n + 1)1/(n+1) when n ≥ 3. Since all terms of the
sequence are positive, the sequence is decreasing and bounded when
n ≥ 3, and so the sequence converges. (As it happens, we can compute
the limit in this case, but we know it converges even without knowing the
limit; see exercise 1.)
EXAMPLE:
Show that {n!/nn } converges.
Again we show that the sequence is decreasing, and since each term is
positive the sequence converges. We can’t take the derivative this time,
as x! doesn’t make sense for x real. But we note that if an+1/an < 1 then
an+1 < an, which is what we want to know. So we look at an+1/an: an+1
an = (n + 1)! (n + 1)n+1 n n n! = (n + 1)! n! n n (n + 1)n+1 = n + 1 n + 1
n n + 1n = n n + 1n < 1.

Calculus with Power Series

Now we know that some functions can be expressed as power series,


which look like infinite polynomials. Since calculus, that is, computation
of derivatives and antiderivatives, is easy for polynomials, the obvious
question is whether the same is true for infinite series. The answer is
yes:
THEOREM:
Suppose the power series f(x) = X∞ n=0 an(x − a) n has radius of
convergence R. Then f ′ (x) = X∞ n=0 nan(x − a) n−1 , Z f(x) dx = C +
X∞ n=0 an n + 1 (x − a) n+1 , and these two series have radius of
convergence R as well.
EXAMPLE:
Starting with the geometric series: 1 1 − x = X∞ n=0 x n Z 1 1 − x dx = −
ln |1 − x| = X∞ n=0 1 n + 1 x n+1 ln |1 − x| = X∞ n=0 − 1 n + 1 x n+1.
when |x| < 1. The series does not converge when x = 1 but does
converge when x = −1 or 1 − x = 2. The interval of convergence is [−1,
1), or 0 < 1 − x ≤ 2, so we can use the series to represent ln(x) when 0
< x ≤ 2.

140
Sequences and Limits Exercises

2𝑛+4
1.Prove that the sequence 𝑎𝑛 = has a limit of 2. Also, calculate the terms whose
𝑛
distance from 2 is less than 0.1.

4𝑛+1
2. Prove that the sequence 𝑎𝑛 = 𝑛 has a limit of 4 and calculate how many terms
of the succession are not within (4 − 0.001, 4 + 0.001).

𝑛3
3. Prove that the sequence 𝑎𝑛 = 𝑛2 +3 has a limit of 1 and calculate how many terms
of the succession are not within (1 − 0.001, 1 + 0.001).

4. Take the sequence given by the rule x0 = 1, xn = xn−1 + 1 n! for n = 1, 2, 3, . . . .

(−1)𝑛
5. an = (−1)n n + 1
𝑛+1

6. Determine whether or not the sequence, xn defined for each integer n, by the finite
series xn = arctan 1 + arctan 2 + arctan 3 + ... + arctan n, for n ≥ 1, converges as n →
∞?

141
Answer Keys:
2𝑛+4 1
1.| − 2| < 10
𝑛
2𝑛+4−2𝑛 1
.| | < 10
𝑛
4 1 4 1
|𝑛| < 10 ; < 10 ; n> 40.
𝑛
4𝑛+1 1
2. | − 4| < 1000
𝑛
4𝑛+1−4𝑛 1
| | < 1000
𝑛
1 1 1 1
|𝑛| < 1000 ; < 1000 ; n> 1000.
𝑛
𝑛3 1
3. |𝑛2 +3 − 1| < 1000

−3 1
|𝑛2 +3| < 1000

3 1
< 1000 ; 𝑛3 + 3> 3000 ; 𝑛2 > 2997 ; n> 54. The first 54 terms are out.
𝑛2 +3

5 8 65 163 1957
4. 1, 2, 2 , 3 , 24 , , …
60 720
5. lim sup 𝑎𝑛 =+∞ ; lim inf 𝑎𝑛 = −∞, lim 𝑎𝑛
𝑛→∞ 𝑛→∞ 𝑛→∞
6. No, the limit cannot exist since arctan n → π/2 as n → ∞ and so the Divergence
Test for infinite series applies and this result follows.

142
LIMIT THEOREM

FERLY CANONAYON

143
Limit Theorem
Definition:
A sequence (𝑥𝑛 ) is bounded if there is a number 𝑀 > 0 such that
|𝑥𝑛 | ≤ 𝑀 𝑓𝑜𝑟 𝑎𝑙𝑙 𝑛 ∈ 𝑁.
Theorem:
Every convergent sequence is bounded.
Proof :
Assume that ( 𝑥𝑛 ) is a convergent sequence: there is a unique
𝑙𝑖𝑚
number 𝑙 such that 𝑥 =1
𝑛→∞ 𝑛
Theorem:
If 𝑎 = 𝑙𝑖𝑚 𝑎𝑛 and 𝑏 = 𝑙𝑖𝑚 𝑏𝑛 , then
(i) 𝑙𝑖𝑚(𝑐𝑎𝑛 ) = 𝑐𝑎 for all 𝑐 ∈ 𝑅,
(ii) 𝑙𝑖𝑚(𝑎𝑛 + 𝑏𝑛 ) = 𝑎 + 𝑏,
(iii) 𝑙𝑖𝑚(𝑎𝑛 𝑏𝑛 ) = 𝑎𝑏,
(iv) 𝑙𝑖𝑚(𝑎𝑛 /𝑏𝑛 ) = 𝑎/𝑏 provided 𝑏 ≠ 0.
Proof:
a. We consider two cases.
1. c ≠ 0
We are trying to show that 𝑐𝑎𝑛 → 𝑐𝑎. Hence we are interested in the
inequality
|𝑐𝑎𝑛 − 𝑐𝑎| < 𝜖

⇔|𝑐(𝑎𝑛 − 𝑎)| < 𝜖


⇔ |𝑐||𝑎𝑛 − 𝑎| < 𝜖
𝜖
⇔ |𝑎𝑛 − 𝑎| <
|𝑐|
𝜖
Let 𝜖 > 0 be arbitrary. Since (𝑎𝑛 ) converges to a, given |𝑐| ≥ 0, there exists
𝜖
𝑁 ∈ ℕ such that if 𝑛 ≥ 𝑁, then|𝑎𝑛 − 𝑎| < |𝑐| . Then we compute
𝜖
|𝑎𝑛 − 𝑎| <
|𝑐|
⟹ |𝑐||𝑎𝑛 − 𝑎| < 𝜖

⟹ |𝑐(𝑎𝑛 − 𝑎)| < 𝜖


⟹ |𝑐𝑎𝑛 − 𝑐𝑎| < 𝜖

Therefore for any 𝜖 > 0, we have found 𝑁 ∈ ℕ such that if ≥ 𝑁, then .


|𝑐𝑎𝑛 − 𝑐𝑎| < 𝜖
Hence, 𝑐𝑎𝑛 converges to 𝑐𝑎.
b. 𝑐 = 0. If 𝑐 = 0 then the sequence 𝑐𝑎𝑛 is simply the sequence
(0,0,0…)which consists only of 0’s. by the theorem, this sequence
converges to 0 = 𝑐𝑎, as desired.

Theorem:
144
Suppose that a = lim 𝑎𝑛 and b = lim𝑏𝑛 .
(i) If 𝑎𝑛 ≥ 0 for all 𝑛 ∈ 𝑁, then 𝑎 ≥ 0.
(ii) If 𝑎𝑛 ≤ 𝑏𝑛 for all 𝑛 ∈ 𝑁, then 𝑎 ≤ 𝑏.
(iii) If there exists 𝑐 ∈ 𝑅 for which c ≤ 𝑏𝑛 for all 𝑛 ∈ 𝑁, then
𝑐 ≤ 𝑏. Similiary, if 𝑎𝑛 ≤ c for all 𝑛 ∈ 𝑁, then 𝑎 ≤ 𝑐.
Proof.
(i) This will be proved by contradiction: we suppose that 𝑎 <
0 and we will show that 𝑎𝑛 < 0 for some n, a contradiction.
For 𝜖= |a| there exists 𝑁 ∈ 𝑁 such that |𝑎𝑛 − 𝑎| < = |𝑎| when
𝑛 ≥ 𝑁. Taking n = N, we have |𝑎𝑁 − a| < |a|, or unwrapping the
absolute value in the middle,
−|𝑎| + 𝑎 < 𝑎𝑁 < |𝑎| + 𝑎 = 0.

(ii) implies that the sequence (𝑏𝑛 − 𝑎𝑛 ) converges to 𝑏 − 𝑎.


Because (𝑏𝑛 − 𝑎𝑛 )≥ 0 for all 𝑛 ∈ 𝑁, we apply part (i) to conclude
that 𝑏 − 𝑎 ≥ 0.
(ii) Take 𝑎𝑛 = c (or 𝑏𝑛 = c) and apply part (ii).
Theorem:
Two Fundamental Limits
𝑙𝑖𝑚
i. 𝑐 = 𝑐, where c is constant
𝑥→𝑎
𝑙𝑖𝑚
ii. 𝑥=𝑎
𝑥→𝑎
Examples:
𝑙𝑖𝑚
a. 10 = 10
𝑥→2
𝑙𝑖𝑚
b. 𝝅=𝝅
𝑥→6
𝑙𝑖𝑚
c. 𝑥=2
𝑥→2
𝑙𝑖𝑚
d. 𝑥=0
𝑥→0
Theorem:
Limit of a Constant Multiple
If 𝒄 is a constant, then
𝑙𝑖𝑚 𝑙𝑖𝑚
𝑐 𝑓(𝑥) = 𝑐 𝑓(𝑥)
𝑥→𝑎 𝑥→𝑎
𝑙𝑖𝑚 𝑙𝑖𝑚
a. 5𝑥 = 5 𝑥 = 5 · 8 = 40
𝑥→8 𝑥→8
𝑙𝑖𝑚 3 3 𝑙𝑖𝑚 3
b. (− 2) 𝑥 = (− 2) 𝑥 = (− 2) (−2) = 3
𝑥 → −2 𝑥 → −2
Theorem:

145
Limits of a Sum, Product, and Quotient
𝑙𝑖𝑚 𝑙𝑖𝑚
Suppose 𝑎 is a real number and 𝑓(𝑥) and 𝑔(𝑥) exist. If
𝑥→𝑎 𝑥→𝑎
𝑙𝑖𝑚 𝑙𝑖𝑚
𝑓(𝑥) = 𝐿1 and 𝑔(𝑥) = 𝐿2 , then
𝑥→𝑎 𝑥→𝑎
𝑙𝑖𝑚 [𝑓(𝑥) 𝑙𝑖𝑚 𝑙𝑖𝑚
i. ± 𝑔(𝑥)] = 𝑓(𝑥) ± 𝑔(𝑥) = 𝐿1 ± 𝐿2
𝑥→𝑎 𝑥→𝑎 𝑥→𝑎
𝑙𝑖𝑚 [𝑓(𝑥)𝑔(𝑥)] 𝑙𝑖𝑚 𝑙𝑖𝑚
ii. =( 𝑓(𝑥)) ( 𝑔(𝑥)) = 𝐿1 𝐿2 , and
𝑥→𝑎 𝑥→𝑎 𝑥→𝑎
𝑙𝑖𝑚
𝑙𝑖𝑚 𝑓(𝑥) 𝑥→𝑎𝑓(𝑥) 𝐿1
iii. = 𝑙𝑖𝑚 = 𝐿 , 𝐿2 ≠ 0
𝑥 → 𝑎 𝑔(𝑥) 𝑥→𝑎 𝑔(𝑥) 2

Theorem above can also be stated as


i. The limit of a sum is the sum of the limits
ii. The limit of a product is the product of the limits, and
iii. The limit of a quotient is the quotient of the limits
Theorem:
Limit of a Power
𝑙𝑖𝑚
Let 𝑥→𝑎 𝑓(𝑥) = 𝐿 and 𝑛 be a positive integer. Then
𝑛
𝑙𝑖𝑚 𝑙𝑖𝑚
[𝑓(𝑥)]𝑛 = [ 𝑓(𝑥)] = 𝐿𝑛
𝑥→𝑎 𝑥→𝑎
For the special case for 𝑓(𝑥) = 𝑥, the result given in theorem 2.2.4
yields
𝑙𝑖𝑚 𝑛
𝑥 = 𝑎𝑛
𝑥→𝑎
Theorem:
The Limit That Does Not Exist
𝑙𝑖𝑚 𝑙𝑖𝑚
Let 𝑓(𝑥) = 𝐿1 ≠ 0 and 𝑔(𝑥) = 0. Then
𝑥→𝑎 𝑥→𝑎
𝑙𝑖𝑚 𝑓(𝑥)
𝑥 → 𝑎 𝑔(𝑥)
does not exist.
Proof:
We will give an indirect proof of this result by using Theorem 2.2.3.
𝑙𝑖𝑚 𝑙𝑖𝑚
Suppose that 𝑓(𝑥) = 𝐿1 and 𝑔(𝑥) = 0 and suppose
𝑥→𝑎 𝑥→𝑎
𝑙𝑖𝑚 𝑓(𝑥)
further that exists and equal 𝐿2 . Then
𝑥 → 𝑎 𝑔(𝑥)
𝑙𝑖𝑚 𝑙𝑖𝑚 𝑓(𝑥)
𝐿1 = 𝑓(𝑥) = (𝑔(𝑥) · ) , 𝑔(𝑥) ≠ 0
𝑥→𝑎 𝑥→𝑎 𝑔(𝑥)

146
𝑙𝑖𝑚 𝑙𝑖𝑚 𝑓(𝑥)
=( 𝑔(𝑥)) = 0 · 𝐿2 = 0
𝑥→𝑎 𝑥 → 𝑎 𝑔(𝑥)
By contradicting the assumption that 𝐿1 ≠ 0, we have proved the
theorem.
Theorem:
Limit of a Root
𝑙𝑖𝑚
Let 𝑓(𝑥) = 𝐿 and 𝑛 be a positive integer. Then
𝑥→𝑎
𝑙𝑖𝑚 𝑛 𝑛 𝑙𝑖𝑚 𝑛
√𝑓(𝑥) = √ 𝑓(𝑥) = √𝐿
𝑥→𝑎 𝑥→𝑎
provided that 𝐿 ≥ 0 when 𝑛 iseven.
Theorem:
Existence Implies Uniqueness
𝑙𝑖𝑚
If 𝑓(𝑥) exists, then it is unique.
𝑥→𝑎
Theorem:
Suppose that f and g are functions such that f(x) = g(x) for all x in
some open interval interval containing a except possibly for a, then

Theorem:
Suppose that f and g are functions such that the two limits

exist, suppose that k is a constant and suppose that n is a positive


integer. Then

1.

2.

3.

4.

5.

6.

147
7.

8.

9. provided when n is even.

If for all then .


10.

11. If f is a polynomial then .

12. If f is a rational function then, for all a in the domain of f,

Theorem C.

The limit

if and only if the right-hand limits and left-hand limits exist and are equal to
M:

Theorem D.
(Squeeze Theorem)
Suppose that f, g and h are three functions such that f(x) < g(x) < h(x)
for all x. If

then

Theorem E.
Suppose that f and g are two functions such that
and

then the limit

does not exist.

Examples

148
𝑙𝑖𝑚 (4𝑥 2 𝑙𝑖𝑚 𝑙𝑖𝑚 𝑙𝑖𝑚
1. − 2𝑥 + 1) = 4𝑥 2 − 2𝑥 + 1
𝑥→2 𝑥→2 𝑥→2 𝑥→2
𝑙𝑖𝑚 2 𝑙𝑖𝑚 𝑙𝑖𝑚
=4 𝑥 −2 𝑥+ 1
𝑥→2 𝑥→2 𝑥→2
2
= 4(2) − 2(2) + 1
=9
𝑙𝑖𝑚 2 𝑙𝑖𝑚 2 𝑙𝑖𝑚
𝑙𝑖𝑚 √𝑥 2 −1 √𝑥→−3(𝑥 −1) √𝑥→−3𝑥 −𝑥→−31
2. = 𝑙𝑖𝑚 = 𝑙𝑖𝑚
𝑥 → −3 2𝑥 𝑥→−3
2𝑥
𝑥→−3
2𝑥
(−32 )−1 √8 √2
= = −6 = −
2(−3) 3

149
Exercises
𝑙𝑖𝑚
1. Evaluate (10𝑥 + 7)
𝑥→5
2. Evaluate
𝑙𝑖𝑚 3
a. 𝑥
𝑥 → 10
𝑙𝑖𝑚 5
b.
𝑥 → 10 𝑥 2
3.Using Theorem 2.2.3. Evaluate
𝑙𝑖𝑚
a. (𝑥 2 − 5𝑥+6)
𝑥→3
4. Using Theorem 2.2.4. Evaluate
𝑙𝑖𝑚
a. (3𝑥 − 1)10
𝑥→1
𝑙𝑖𝑚 3𝑥−4
5. Evaluate 2
𝑥 → −1 8𝑥 =2𝑥−2
𝑙𝑖𝑚 𝑥−1
6. Evaluate
𝑥 → 1 𝑥 2 +𝑥−2

150
Answers Key
𝑙𝑖𝑚 (10𝑥 𝑙𝑖𝑚 𝑙𝑖𝑚
1. + 7) = 10𝑥 + 7
𝑥→5 𝑥→5 𝑥→5
𝑙𝑖𝑚 𝑙𝑖𝑚
= 10 𝑥+ 7
𝑥→5 𝑥→5
= 10 · 5 + 7
= 57
𝑙𝑖𝑚 3
2. a. 𝑥 = 103 = 1000
𝑥 → 10
𝑙𝑖𝑚
𝑙𝑖𝑚 5 5
𝑥→4
b. 2 = 𝑙𝑖𝑚
𝑥 → 4𝑥 𝑥→4
𝑥2
5
= 42
5
= 16

𝑙𝑖𝑚 (𝑥 2 𝑙𝑖𝑚 2 𝑙𝑖𝑚 𝑙𝑖𝑚


3. − 5𝑥 + 6) = 𝑥 − 5𝑥 + 6
𝑥→3 𝑥→3 𝑥→3 𝑥→3
= 32 − 5(3) + 6
= 9 − 15 + 6
=0
𝑙𝑖𝑚
4. (3𝑥 − 1)10
𝑥→1
First, we see from Theorem 2.2.3 that
𝑙𝑖𝑚 𝑙𝑖𝑚 𝑙𝑖𝑚
(3𝑥 − 1)10 = 3𝑥 − 1=2
𝑥→1 𝑥→1 𝑥→1

It then follows from Theorem 2.2.4 that


𝑙𝑖𝑚 𝑙𝑖𝑚
(3𝑥 − 1)10 = [ (3𝑥 − 1)10 ]
𝑥→1 𝑥→1
= 210

= 1024
3𝑥−4
5. Solution: f(x)= 8𝑥 2 =2𝑥−2 is a rational function and so if we
identify the polynomials
𝑝(𝑥) = 3𝑥 − 4 and (𝑥) = 8𝑥 2 = 2𝑥 − 2 , then
𝑙𝑖𝑚 𝑙𝑖𝑚
𝑝(𝑥) = 𝑝(−1) = −7 and 𝑞(𝑥) = 𝑞(−1) = 4
𝑥→1 𝑥→1
Since, 𝑞(−1) ≠ 0 it follows that
𝑙𝑖𝑚 3𝑥 − 4 𝑝(−1) −7 7
= = = −
𝑥 → −1 8𝑥 2 = 2𝑥 − 2 𝑞(−1) 4 4
𝑙𝑖𝑚 𝑥−1 𝑙𝑖𝑚 𝑥−1
6. =
𝑥 → 1 𝑥 2 +𝑥−2 𝑥 → 1 (𝑥−1)(𝑥+1)

151
𝑙𝑖𝑚 1
=
𝑥 → 1 𝑥+1
𝑙𝑖𝑚
1
𝑥→1
= 𝑙𝑖𝑚
(𝑥+2)
𝑥→1
1
=3

152
MONOTONE SEQUENCES

CHRISTY IGNALIG

153
Monotone Sequence
Definition:
We begin by a definition. We say that a real sequence (an) is
monotone increasing if n1 < n2 =) an1 < an2 monotone decreasing if
n1 < n2 =) an1 > an2 monotone non-decreasing if n1 < n2 =) an1 6 an2
monotone non-increasing if n1 < n2 =) an1 > an2
Example:
Let an = n. Then (an) is monotone increasing. So is an = (2n + 1)2.

Definition:

A sequence (an) is monotonic increasing if an+1≥ an for all n ∈ N.

Remarks:

The sequence is strictly monotonic increasing if we have > in the


definition.
Monotonic decreasing sequences are defined similarly.
Then the big result is

Theorem:

A bounded monotonic increasing sequence is convergent.

Proof:

We will prove that the sequence converges to its least upper bound
(whose existence is guaranteed by the Completeness axiom).
So let α be the least upper bound of the sequence. Given ε > 0, we'll
show that all the terms of the sequence (except the first few) are in the
interval (α - ε, α + ε). Now since α + ε is an upper bound of the
sequence, all the terms certainly satisfy an< α + ε. Also
since α - ε is not an upper bound of the sequence, we must
have aN> α - ε for some N. But then all the later terms will be > α - ε also
and so (for n > this N) we have our condition for convergence.

Remarks:

A similar result is true for a bounded monotonic decreasing sequence


(which converges to its greatest lower bound).

There are sequences which are convergent without being monotonic.


For example, the sequences (-1 , 1/2 , -1/3 , 1/4 , ... ) and
(1/2 , 1/22 , 1/3 , 1/32 , ... ) both converge to 0.

For sequences given by recurrence relations it is sometimes easy to see


what their limits are.

154
Examples:

1. Define a sequence by an+1= √(2an- 1) with a1= 2.

Then suppose that this has a limit α. Then for large n, we


have an= α (approx) and an+1= α (approx) and so we must have α =
√(2α -1) and hence α2 = 2α - 1 and we get α = 1. To show that it does
indeed have a limit, we'll prove that it is monotonic decreasing and
bounded below.

Since the terms of the sequence are positive, the sequence is clearly
bounded below by 0.

2. Define an+1= (an2 + 1)/2 with a1 = 2.

Claim:

This is monotonic increasing.


2 2
an+1 - an= (an - an-1 )/2 = (an + an-1)(an - an-1)/2 and so (since an > 0) the
result follows by induction.

What is the limit?


The limit α satisfies α = (α2+ 1)/2 and hence α = 1.
But since the sequence starts at 2 and increases, it cannot converge to
1 and hence it has no limit.

One may show that if one takes the starting value a1 = 1/2 then the
sequence is bounded above and does converge to 1.

Definition:

The sequence is said to be increasing if 𝑠𝑛 ≤ 𝑠𝑛 +1 for all n ≥1, i.e.,


𝑠1 ≤ 𝑠2 ≤ 𝑠3
The sequence is𝑠𝑛 said to be decreasing if 𝑠𝑛 ≥+1 𝑠𝑛 for all n ≥ 1,
i.e., 𝑠1 ≥ 𝑠2 ≥ 𝑠3
A sequence is said to be monotone if it is either increasing or
decreasing.

Example:

The sequence n2 : 1, 4, 9, 16, 25, 36, 49, ... is increasing. The sequence
1/2n : 1/2, 1/4, 1/8, 1/16, 1/32, ... is decreasing. The sequence
(−1)n_1/n : 1,−1/2, 1/3,−1/4, 1/5,−1/6, ... is not monotone.

Example:

155
Let r 6= 0. Consider the geometric sequence sn = r n: 1, r , r 2, r 3, r 4, ....

Boundedness

Definition:
The sequence 𝑠𝑛 is bounded above if there exists a number M such that
𝑠𝑛 ≤ M for all n. M is called an upper bound for 𝑠𝑛 .
The sequence 𝑠𝑛 is bounded below if there exists a number K such that K
≤ 𝑠𝑛 for all n. K is called a lower bound for𝑠𝑛 .
A sequence is bounded if it is both bounded above and bounded below.
Example:
1
𝑠𝑛 𝑠𝑛 =𝑛 is bounded, since it is both bounded above (by 1, for example) and
bounded below
𝑠𝑛 = −n is bounded above (by 0, for example) but not bounded below.
𝑠𝑛 = 2𝑛 is bounded below (by 0, for example) but not bounded above.
𝑠𝑛 = (−1𝑛−1 )n is neither bounded above nor bounded below.
Remarks:
If a sequence is bounded above, then there are infinitely many upper
bounds, not just one. Any number greater than some upper bound is itself
an upper bound. However, among all the upper bounds, there is at least
one, the least upper bound, also known as the supremum.
If a sequence is bounded below, then there are infinitely many lower
bounds, not just one. Any number less than some lower bound is itself a
lower bound. However, among all the lower bounds, there is a greatest
one, the greatest lower bound, also known as the infimum.

Definition:
A sequence 𝑎𝑛 𝑛 ≥ 1of real numbers is called an increasing sequence if 𝑎𝑛 ≤
𝑎𝑛+1 for all n ≥ 1, and (𝑎𝑛 )𝑛 ≥ 1 is called a decreasing sequence if 𝑎𝑛 ≥
𝑎𝑛+1 for all n ≥ 1. A sequence that is increasing or decreasing is said to be a
monotone sequence.
Example:
The sequence (1, 1, 2, 2, 3, 3, 4, 4, . . .) is increasing, but (−1, 1, −1, 1,
. . .) is not monotone.
Exercise:
a) Show that any bounded above increasing sequence is bounded.
b) Show that any bounded below decreasing sequence is bounded.
Exercise:
a) Prove that (𝑛2−𝑛 )𝑛≥2 is a decreasing sequence.
b) Let (𝑎𝑛 )𝑛≥1 be an increasing sequence of positive numbers and
define (𝜎𝑛 )𝑛≥1 . Prove that (𝜎𝑛 )𝑛≥1 is also an increasing sequence.

156
Theorem:
Every bounded monotone sequence converges.
Proof.
We will prove the theorem for increasing sequences. The case of
decreasing sequences is left to Exercise 4.3. So, let a sequence
(𝑎𝑛 )𝑛≥1 increase. By the assumption of the theorem, (𝑎𝑛 )𝑛≥1 is
bounded, that is, there exists C ∈ R such that ⃓𝑎𝑛 ⃓ ≤ 𝐶 for all 𝑛 ≥ 1.
This implies that the set 𝐴 ≔ 𝑎𝑛 : 𝑛 ≥ 1is also bounded. Thus, by
Theorem 2.2 (i) there exists sup 𝐴 = : 𝑠𝑢𝑝 𝑛 ≥ 1 an denoted by a. Let
us prove that 𝑎𝑛 → 𝑎, 𝑛 → ∞. We first note that 𝑎𝑛 ≤ a for all 𝑛 ≥ 1,
since the supremum of A is also its upper bound (see Definition 2.6).
Next, we take an arbitrary ε > 0 and use Theorem 2.1 (i). So, there
exists a number m such that〖 a〗_m > a − ε. By the monotonicity, a −
ε < 𝑎𝑚 ≤ 𝑎𝑛 for all 𝑛 ≥ 𝑚. Thus, setting 𝑁 ∶= 𝑚, one has
𝑎 − 𝜀 < 𝑎𝑛 ≤ 𝑎 for all 𝑛 ≥ 𝑁 which implies |𝑎𝑛 − 𝑎| < 𝜀.
Exercise:
Prove Theorem above for decreasing sequences.
Theorem above remains valid if one requires the monotonicity of
(𝑎𝑛 )𝑛 ≥ 1 starting from some number m, that is, the monotonicity of
(𝑎𝑛 )n≥m = (𝑎𝑚 , 𝑎𝑚+1 , . . .).
Example:
𝑙𝑖𝑚 10𝑛
Prove that n! = 0, where 𝑛! ∶= 1 · 2 · 3 · . . .· 𝑛. Solution.
𝑛 → ∞ 𝑛!𝑛+1
10 10𝑛
First we note that(𝑛+1)! < 𝑛! ⇔ 10𝑛 + 1 ⇔ 𝑛 > 9. Hence, the sequence
10𝑛
( 𝑛! )𝑛≥10 is decreasing. Moreover, it is bounded below by zero. Thus,
10𝑛
( 𝑛! )𝑛≥10 is bounded, by Exercise 4.1 b). Using Theorem 4.1, one gets
𝑙𝑖𝑚 10𝑛 10𝑛+1 10𝑛
that there exists a ∈ R such that = 𝑎. But we can = •
𝑛 → ∞ 𝑛! (𝑛+1)! 𝑛!
10
. So,
𝑛+1

10𝑛+1 10𝑛 10
𝑎 = lim = lim · lim = 𝑎 ·
𝑛→∞ (𝑛 + 1)! 𝑛→∞ 𝑛! 𝑛→∞ 𝑛 + 1

This implies 𝑎 = 0

Exercise:
Show that
𝑛!
a) lim 2 =0
𝑛→∞ 2𝑛
𝑛
b) lim =0
𝑛→∞ 2√𝑛

157
Exercise 4.5.

Find a limit of the sequence (√2, √2 + √2, √2 + √2 + √2, … )

Exercise 4.6.
1
Let 𝑎1 = 1 and 𝑎𝑛+1 = 3 (𝑎𝑛+1 ) for all n ≥ 1.

a) Find 𝑎2, 𝑎3 , 𝑎4,


b) Use induction to show that 𝑎𝑛 > 1 2 for all 𝑛 ≥ 1.
c) Show that (𝑎𝑛 )𝑛 is a decreasing sequence. d) Show that limn→∞ an
exists and find it.
Exercise 4.7.
Let c > 0, a1 > 0 and let an+1 = 1 2 an + c an for all n ≥ 1.
a) Show that an ≥ √ c for all n ≥ 2.
b) Show that (an)n≥2 is a decreasing sequence.
c) Show that limn→∞ an exists and find it.
Theorem 4.2.
(i) If (an)n≥1 is an unbounded increasing sequence, then limn→∞ an
= +∞.
(ii) If (an)n≥1 is an unbounded decreasing sequence, then limn→∞ an
= −∞.
Proof.
We will prove only Part (i) of the theorem. The proof of Part (ii) is
similar. If (an)n≥1 is an unbounded increasing sequence, then it must
be unbounded above, since it is bounded below by a1. Taking any C
and using the unboundedness of (an)n≥1, one can find a number m ∈
N such that am ≥ C. Next, by the monotonicity of (an)n≥1, the
inequality an ≥ am ≥ C trivially holds for all n ≥ N := m. This proves
limn→∞ an = +∞ (see Definition 3.4).
Corollary 4.1.
If (an)n≥1 is a monotone sequence, then the sequence either converges,
diverges to +∞, or diverges to −∞. Thus limn→∞ an is always meaningful
for monotone sequences. Proof. The proof immediately follows from
theorems 4.1 and 4.2.
Exercise 4.8.
Let A be a bounded nonempty subset of R such that sup A is not in A.
Prove that there is an increasing sequence (an)n≥1 of points from A such
that limn→∞ an = sup A.

158
4.2 The number e In this section, we will consider two sequences of
positive numbers an := 1 + 1 n n n≥1 and bn := 1 + 1 n n+1! n≥1 (1) and
study their properties.
Theorem 4.3.
The sequences defined in (1) satisfy the following properties: 1) an < bn
for all n ≥ 1; 2) the sequence (an)n≥1 increases; 3) the sequence (bn)n≥1
decreases.
Proof.
Since bn = an 1 + 1 n = an + an n > an for all n ≥ 1, Property 1) is
proved. To prove 2), we are going to use Bernoulli’s inequality (see
Theorem 2.6). So, one has an an−1 = n + 1 n n n − 1 n n−1 = n n − 1 1
− 1 n2 n > n n − 1 1 − n n2 = 1, for all n ≥ 2. Thus, an > an−1 for all n ≥
2. For the prove of 3) we use the same argument. We consider bn−1 bn
= n n − 1 n n n + 1n+1 = n − 1 n n 2 n2 − 1 n+1 = n − 1 n 1 + 1 n2 − 1
n+1 > n − 1 n 1 + n + 1 n2 − 1 = 1 for all n ≥ 2. Hence, bn−1 > bn for all n
≥2
Theorem 4.3
yields the following inequalities a1 < a2 < . . . < an < . . . < bn < . . . < b2 <
b1. (2) Consequently, the sequences (an)n≥1 and (bn)n≥1 are monotone
and bounded. By Theorem 4.1, they converge. We set e := limn→∞ an =
limn→∞ 1 + 1 n n = 2, 718281828459045... It is known that e is an
irrational number. The number e is one of the most important constants in
mathematics. Since bn = an 1 + 1 n for all n ≥ 1, one has bn → e, n → ∞.
We also note that 1 + 1 n n < e < 1 + 1 n n+1 , (3) by inequalities (2).
Definition 4.2.
The logarithm to base e is called the natural logarithm and is denoted by
ln := loge , that is, for each a > 0 ln a is a (unique!) real number such that
e ln a = a. The inequality (3) immediately implies 1 n + 1 < ln 1 + 1 n < 1
n for all n ≥ 1.
Exercise 4.9. Show that limn→∞ n ln 1 + 1 n = 1.
Exercise 4.10. Prove that for each x > 0 the sequence 1 + x n n n≥1 is
increasing and bounded.
Subsequences
Definition4.3.
1 Subsequences and Subsequential Limits Let (an)n≥1 be a sequence.
We consider any subsequence (nk)k≥1 of natural numbers such that 1 ≤
n1 < n2 < . . . < nk < nk+1 < . . .. We note that nk ≥ k and nk → +∞, k →
∞.
Example 4.3.
1) nk = k, k ≥ 1; then (nk)k≥1 = (1, 2, 3, . . . , k, . . .);
2) nk = 2k, k ≥ 1; then (nk)k≥1 = (2, 4, 6, . . . , 2k, . . .);

159
3) nk = k 2 , k ≥ 1; then (nk)k≥1 = (1, 2, 9, . . . , k2 , . . .);
4) nk = 2k , k ≥ 1; then (nk)k≥1 = (2, 4, 8, . . . , 2 k , . . .).
Definition 4.3.
A sequence (ank )k≥1 = (an1 , an2 , an3 , . . . , ank , . . .) is said to be a
subsequence of (an)n≥1. Thus, (ank )k≥1 is just a selection of some
(possibly all) of the an’s taken in order.
Remark 4.2.
The following properties follows from the definition of subsequence. 1. If
a sequence is bounded, then every its subsequence is bounded. 2. If a
sequence converges to a (that could be +∞ or −∞), then every its
subsequence also converges to a.
Exercise 4.11.
Prove that a monotone sequences which contains a bounded
subsequence is bounded.
Exercise 4.12.
Prove that a sequence (an)n≥1 converges iff (a2k)k≥1, (a2k−1)k≥1 and
(a3k)k≥1 converge.
Definition 4.4.
A subsequential limit of a sequence (an)n≥1 is any real number or the
symbol +∞ or −∞ that is the limit of some subsequence of (an)n≥1. Let A
denotes the set of all subsequential limit of (an)n≥1.
Example 4.4.
a) For the sequence (1, 2, 3, . . . , n, . . .) the set of all subsequential limit
A = {+∞}.
b) For the sequence (−1, 1, −1, . . . ,(−1)n , . . .) the set of all
subsequential limit A = {−1, 1}.
c) If an → a, then A = {a}, by Remark 4.2.
Exercise 4.13. Prove the following statements. a) −∞ ∈ A ⇔ (an)n≥1 is
unbounded below. b) +∞ ∈ A ⇔ (an)n≥1 is unbounded above.
Exercise 4.14. Find the set A of all subsequential limits of the following
sequences.
a) (sin 3πn)n≥1;
b) (sin απn)n≥1 for α ∈ Q;
c) (an)n≥1, where an = ( (−1) n+1 2 + n, if n is odd, (−1) n 2 + 1 n , if n is
even.
4.3.2 Existence of Monotone Subsequence
Theorem 4.4.

160
A number a ∈ R is a subsequential limit of a sequence (an)n≥1 iff ∀ε > 0
∀N ∈ N ∃n˜ ∈ N : ˜n ≥ N, |an˜ − a| < ε. (4)
Proof.
We first prove the necessity. Let a ∈ A. Then there exists a subsequence
(ank )k≥1 such that ank → a, k → ∞. We fix an arbitrary ε > 0 and N ∈ N.
By the definition of the limit, ∃K1 ∈ N ∀k ≥ K1 : |ank − a| < ε. Similarly,
∃K2 ∈ N ∀k ≥ K2 : nk ≥ N. Thus, taking ˜k := max{K1, K2}, ˜n := nk˜, one
has ˜n ≥ N and |an˜ − a| < ε. To prove the sufficiency, we are going to
construct a subsequence of (an)n≥1 converging to a. Let (4) holds. Then,
by (4), for ε = 1 and N = 1 there exists n1 ≥ 1 such that |an1 − a| < 1.
Similarly, for ε = 1 2 and N = n1 + 1 there exists n2 ≥ n1 + 1 such that
|an2 − a| < 1 2 and so on. Consequently, we obtain a subsequence (ank
)k≥1 satisfying |ank − a| < 1 k for all k ≥ 1. Using Theorem 3.7 and
Exercise 3.5 a), one can see that ank → a, k → ∞
Exercise 4.15.
Show that +∞ ∈ A (−∞ ∈ A) provided ∀C ∈ R ∀N ∈ N ∃n˜ ∈ N : ˜n ≥ N
and an˜ ≥ C(an˜ ≤ C).
Theorem 4.5.
Every sequence of real numbers contains a monotone subsequence.
Proof.
We consider the set M := {n ∈ N : ∀m > n am > an}. If M is infinite, then
M can be written as M = {n1, n2, . . . , nk, . . .}, where n1 < n2 < . . . <
nk < . . .. By the definition of M, we have an1 < an2 < . . . < ank < . . ..
So, the subsequence (ank )k≥1 increases. If M is finite, then let n1 be
the smallest natural number such that ∀m ≥ n1 : m 6∈ M. Since n1 6∈
M, one can find n2 > n1 such that an1 ≥ an2 . Similarly, since n2 6∈ M,
one can find n3 > n2 such that an2 ≥ an3 and so on. Thus, the
constructed subsequence (ank )k≥1 decreases.
Corollary 4.2
For every sequence the set of its sub sequential limits is not empty.
Proof
The corollary immediately follows from Theorem 4.5 and Corollary 4.1.
Theorem 4.6
(Bolzano-Weierstrass theorem). Every bounded sequence has a
convergent subsequence
Proof.
The theorem is a direct consequence of theorems 4.5 and 4.1.

161
Exercise

Determine if the sequence is increasing or decreasing by calculating 𝑎𝑛+1 -𝑎𝑛


1
1.{4𝑛}−∞
𝑛=1
2𝑛−3
2.{3𝑛−2 }+∞
𝑛=1
𝑎𝑛+1
Determine if the sequence is increasing or decreasing by calculating .
𝑎𝑛
1
3.{.4𝑛}+∞
𝑛=1
𝑒 𝑛 −𝑒 −𝑛
4. {𝑒 𝑛+𝑒 −𝑛}+∞
𝑛=1
Determine if the sequence is increasing or decreasing by calculating the derivative
𝑎′𝑛 .
1
5. { 𝑛}−∞
𝑛=1
4
𝐼𝑛(2𝑛) +∞
6{ }
𝐼𝑛(6) 𝑛=1
Use an appropriate test for monotonicity to determine if
the sequence increases, decreases, eventually increases, or eventually decreases.
3𝑛 +∞
7.{ }𝑛=1
2𝑛+1
1
8.{𝑛 − 𝑛}+∞ 𝑛=1
𝑛2 +∞
9.{ 𝑛! }𝑛=1
2𝑛=1
10{. (2𝑛)! }+∞
𝑛=1
𝑒 √𝑛
11{. 𝑛 }+∞
𝑛=1
12.{𝑒 𝑛 𝜋 −𝑛 }+∞
𝑛=1
𝑛! +∞
13 {𝑛!}𝑛=1
2)
3(𝑛
14.{1000𝑛}+∞𝑛=1
15.{𝑛3 𝑒 _𝑛 }+∞
𝑛=1

162
Answers key

1. The sequence is (strictly) decreasing.


2. The sequence is (strictly) increasing.
3. The sequence is (strictly) decreasing.
4. The sequence is (strictly) increasing.;
5. The sequence is (strictly) decreasing.
6. The sequence is (strictly) increasing.
7. The sequence is (strictly) increasing.
8. The sequence is (strictly) increasing.
9. The sequence is eventually (strictly) decreasing.
10. The sequence is (strictly) decreasing.;
11. The sequence is eventually (strictly) increasing.
12. The sequence is (strictly) decreasing.
13. The sequence is (strictly) decreasing.
14. The sequence is eventually (strictly) increasing.
15. The sequence is eventually (strictly) decreasing.

163
SUBSEQUENCES

JOAN DIGOLTO

164
Sub- Sequences
Definition:
Let (an) be a sequence of real numbers, and let n1 < n2 < n3 < · · · be
a strictly increasing sequence of natural numbers. Then the sequence
an1 , an2 , an3 , . . . , is called a subsequence of (an) and is denoted
by (anj ), where j ∈ N indexes the subsequence. Notice that the order
of the terms in a subsequence (anj ) is the same as in the original
sequence
Example:
If an = 1/n2 , then 1, 1 9 , 1 25 , 1 49 , . . .
1 1 1
(1, 9,,25,49,…)
And
1 1 1 1
(4,16,64 , 256,…)
Are subsequences of (𝑎𝑛 ).
For the first of these we have (𝑎𝑛 ) where
𝑛𝑗 =2𝑗 -1
And for the second of these we have
𝑛𝑗 =2𝑗 .
Both of these forms of 𝑛𝑗 give strictly increasing sequences of positive
integers.

Theorem:
All subsequences of a convergent sequence converge to the same limit
as the original sequence.
Proof:
Let (𝑎𝑛 ) be a convergence sequence with limit l.
Suppose (𝑎𝑛 ) is a subsequence of (𝑎𝑛 ).
For 𝜖 > 0 we must find J ∈ N such that |𝐴𝑛̇ 𝑗 − l| < 𝜖 for all j ≥ J.
Since 𝑎𝑛 → l as n → ∞, there is N ∈ N such that |𝑎𝑛 − l| < 𝜖 for all n ≥ N.
By the nature of 𝑛𝑗 , there is a J ∈ N such that 𝑛𝑗 ≥ N for all j ≥ J.
Then because |𝑎𝑛 − l| < 𝜖 for all n ≥ N, and because 𝑛𝑗 ≥ N for all j ≥ J,
we have that |𝑎𝑛𝑗 − l| < 𝜖 for all j ≥ J.
Example:
For 0 < b < 1 we have b > b2 > b3 > b4 > · · · > bn > · · · > 0.
Thus the sequence (b n ) is decreasing and bounded below, and so it
converges by the Monotone Convergence Theorem.
A reasonable guess for the limit is 0, but we can confirm that by the
Algebraic Limit Theorem and a
If l is the limit of (𝑏 𝑛 ), then l is the limit of the subsequence (𝑏 2𝑛 ).
But 𝑏 2𝑛 = 𝑏 𝑛 𝑏𝑛 and so we have l = 12 , and thus l = 0 (why not 1?).
Can you extend this to −1 < b < 0? It is true.

165
Divergence Criterion for Sequences

Definition:
Since all subsequences of a convergence sequence converge to the
same limit as the original, then we can detect a divergence sequence
if we can produce two subsequences that converge to different limits.
The sequence ( −1)𝑛 is not convergent because it has two
subsequences (−1)2𝑛 and ( −1)2𝑛+1 which converge to 1 and -1
respectively.
But as we have seen, a bounded sequence might have a convergent
subsequence, like (−1)𝑛 does.
It is an amazing result that every bounded sequence has a convergent
subsequence.
Example:
Show that the sequence ((−1)n) is divergent.
To show this sequence is divergent, consider the subsequence of even
terms which is (1,1,1,...) which converges to the real number 1.
Now consider the subsequence of odd terms which
is (−1,−1,−1,...) which converges to the real number −1.
Since these two subsequences converge to different limits, we
conclude that ((−1)n) is divergent.
Example:
Show that the sequence (an) defined by an={1/nnifnisevenifnisodd.
Let's first look at a few terms of this sequence. We have
that (an)=(1,12,3,14,5,16,...). We can see this sequence is not bounded
above and hence not bounded, which we will prove.
Suppose that there exists an M∈N such that ∣an∣<M for all n∈N. By the
Archimedean property since M∈R there exists a natural
number nM∈N such that M≤nM. We also note that M≤nM<nM+1.
So, either nM or nM+1 is a term in the sequence (an) which contradicts
the sequence (an) from being bounded.
Example:
Show that the sequence (n) is divergent.
Once again, this sequence is unbounded. Suppose that instead (n) is
bounded, that is ∣n∣=n<M for some M∈R. But this contradicts the
Archimedean property which says that for any M∈R there exists
an nM∈N such that M≤nM, and so in fact (n) is not bounded and by the
divergence criteria, is divergent as well.

The Bolzano – Weierstrass

Theorem:
Every bounded sequence contains a convergent subsequence.

166
Proof.
Let (𝑎𝑛 ) be a bounded sequence.
Then there is ℳ Є ℝ such that |𝑎𝑛 |≤
м for all n Є N.
We will construct a convergent subsequence of (an) through a
bisection technique. Bisect the closed interval [−M, M] into the closed
subintervals [−M, 0] and [0, M].
Notice the midpoint is included in both subintervals, but as we shall
see, this does not complicate things.
Since there are infinitely many an, one of the two subintervals must
contain infinitely many of them; label this closed interval I1 and choose
n1 so that 𝑎𝑛1 ∈ 𝐼1 .
Now bisect the closed interval 𝐼1 into two closed subintervals that
overlap at the midpoint.
Since there are infinitely many 𝑎𝑛 for n > 𝑛1 , one of the two closed
subintervals must contain infinitely many of them; label this closed
interval 𝐼2 and choose 𝑛2 > 𝑛2 so that 𝑎𝑛2 Є 𝐼2 .
Notice that 𝐼1 ⊇ 𝐼2 .
We can repeat this step countably many times to obtain a nested
sequence of closed intervals
𝐼2 ⊇ 𝐼2 ⊇ 𝐼3 ⊇ 𝐼4 ⊇…
and positive integers 𝑛1 < 𝑛2 < 𝑛3 < 𝑛4 < · · · such that anj ∈ Ij for all j ∈
N.
By the Nested Interval Property
Theorem:
There is at least one x ∈ R contained in every 𝐼𝐽 .
Now the suspicion is that this x is the limit of the subsequence( 𝑎𝑛𝑗 ).
Let 𝜖 > 0.
1 𝐽−1
By the bisection technique, the length of 𝐼𝐽 is M (2) which converges
to 0. Choose J so that j ≥ J implies that the length of 𝐼𝐽 is less than 𝜖.
Then as 𝑎𝑛𝑗 and x are both in the closed interval 𝐼𝐽 of length less than
𝜖, we have |𝑎𝑛 − 𝑥|< 𝜖.
For all j ≥ 𝐽
This holds for all j ≥ 𝐽 because of the nested property of 𝐼𝐽 and
because 𝑎𝑛𝑗 ∈ 𝐼𝐽 .
Thus, we have that (𝑎𝑛𝑗 ) converges to x.
Definition:
Let (𝑥𝑛 ) be a sequence of points in X. Suppose that (𝑛𝑘 ) is a strictly
increasing sequence of natural numbers. Then we say that the
sequence (𝑥𝑛𝑘 ) of points in X indexed by k, is a subsequence of (𝑥𝑛 ).
In the following examples, we take the sequence:

167
Example:
1 1 1 1 1 1
The sequence 2 , 4 , 6 , 8 , 10 , 12 , … is a subsequence of (𝑥𝑛 ). For we take
(𝑛𝑘 ) = 2𝑘 then the sequence is (𝑥𝑛𝑘 ) = 1⁄2 𝑘 for 𝑘 = 1,2,3, …
Example:
1 1 1 1 1 1 1
The sequence 3 , 5 , 7 , 9 , 11 , 13 , 15 , … is a subsequence of (𝑥𝑛 ). For if we
take 𝑛𝑘 = 2𝑘 + 1 then the sequence is (𝑥𝑛𝑘 ) = 1⁄2 𝑘 for 𝑘 ∈ ℕ
Example:
1 1 1 1 1 1 1
The sequence 4 , 3 , 2 , 1 , 5 , 6 , 7 , … is not ha subsequence of (𝑥𝑛 ). For we
take 𝑛𝑘 such that (𝑥𝑛𝑘 ) gives these terms then 𝑛1 = 4 and 𝑛2 = 3,
which is enough to contradict the fact that 𝑛𝑘 must be increasing for a
subsequence.
Example:
1 1 1 1 1 1
The sequence 1, 2 , 3 , 4 , 5 , 6 , 7 , … is not ha subsequence of (𝑥𝑛 ). For we
take 𝑛𝑘 such that 𝑥𝑛𝑘 gives these terms then 𝑛4 = 𝑛5 = 4, which is
contradicts the fact that 𝑛𝑘 must be strictly increasing for a
subsequence.
Example:
Let 𝑝𝑘 denote the kth prime number. Thus 𝑝1 = 2, 𝑝2 = 3, 𝑝3 = 5, 𝑝4 = 7
etc. This is a strictly increasing sequence of natural numbers, so the
1 1 1 1 1 1 1
sequence 𝑥𝑝𝑘 , 2 , 3 , 5 , 7 , 11 , 13 , 17 , … is a subsequence of (𝑥𝑛 ).
One important reason subsequence is useful is because very often
even when a sequence does not converge itself, it does have
subsequences which converge.
Example:
Let 𝑥𝑛 = (−1)𝑛 . Then the sequences 𝑥2𝑘 = 1 and 𝑥2𝑘+1 = −1 both
converge, although the sequence itself does not converge.
On the other hand, if a sequence does converge then all its
subsequences also converge, and to the same thing.
Proposition:
If the sequence (𝑥𝑛 ) converges to x then all its subsequences also
converges to x.
Remark:
Although it seems a pretty unsurprising result, this last proposition does
give us a useful criterion for establishing that a given sequence
does not converge. All one has to do is to identify two subsequences
which converge to different limits. Since we have far more results at
our disposal which do prove convergence than we have which disprove
it, this is oftentimes much easier.
Example:
1
The sequence 𝑥𝑛 = (−1)𝑛 (1 − 𝑛) diverges.

168
Proof.
1
The subsequence 𝑥2𝑘 = 1 − 2𝑘 converges to 1. The subsequence
1
𝑥2𝑘+1 = 2𝑘+1 − 1 converges to -1. Thus, by the proposition above
(𝑥𝑛 ) does not converge. The following theorem shows that, at least in
the real numbers, many sequences have convergent subsequences.
So, this behavior is very common indeed.
Theorem:
Bolzano- Weierstrass Theorem
Every bounded sequence in ℝ hs a convergent subsequence.
Remark:
This proof of the Bolzano-Weierstrass theorem uses ideas based on
Ramsey's Theorem from combinatorics. Indeed the key step of the
proof, showing that every sequence of real numbers has a monotonic
subsequence can be proved using Ramsey's Theorem.
We can use Ramsey's theorem to prove again that every sequence of
real numbers has a monotonic subsequence:

Proof:
Let a sequence (𝑥𝑛 ) of real numbers be given. We define a collection
of 2-element subsets of ℕ as follows:
Let {𝑚, 𝑛} be a 2-element subset of ℕ. Then 𝑚 ≠ 𝑛, so one of 𝑚, 𝑛 is
smaller than the other. Assume that 𝑚 < 𝑛. We include {𝑚, 𝑛} in S if
𝑥𝑚 < 𝑥𝑛 , otherwise we exclude {𝑚, 𝑛} from S and use Ramsey’s
theorem to obtain an infinite set L with one or other of the two
properties described.
If the first case holds, then whenever 𝑚 < 𝑛 and 𝑚, 𝑛 ∈ 𝐿 then {𝑚, 𝑛} is
in S, and so 𝑥𝑚 < 𝑥𝑛 . Thus, if we list the elements of L in increasing
order as 𝑙𝑘 then (𝑥𝑙𝑘 ) is an increasing subsequence.
If the second case holds, then whenever 𝑚 < 𝑛 and 𝑚, 𝑛 ∈ 𝐿, then
{𝑚. 𝑛} ∉ 𝑆 and so 𝑥𝑚 ≥ 𝑥𝑛 . Thus, in this case, if we list the elements of
L in increasing order as 𝑙𝑘 then (𝑥𝑙𝑘 ) is decreasing subsequence.
Following on the Bolzano-Weierstrass Theorem, we can use obtain a
useful property of closed and bounded subsets of the real numbers:
Definition:
We say that a sequence of real numbers {an} is a Cauchy sequence
provided that for every > 0, there is a natural number N so that when
n, m ≥ N, we have that |an − am| ≤ .
Example:
Let x be a real number and 𝑡𝑛 (x) be the nth truncation of its decimal
expansion as in Lectures 2 and 3. Then if n, m ≥ N, we have that | 𝑡𝑛
(x)−𝑡𝑚 (x)| ≤ 10−N , since they share at least the first N places of their
decimal expansion. Given any real number 𝜖 > 0, there is an N(𝜖) so
that 10−N() < .Thus, we have shown that the sequencese {𝑡𝑛 (x)} is
a Cauchy sequence.

169
Example 1 was central in our construction of the real numbers. We got
the least upper bound property by associating to each sequence as in
Example 1, the real number x which is its limit. The class of Cauchy
sequences should be viewed as minor generalization of Example 1 as
the proof of the following theorem will indicate.
Theorem:
Every Cauchy sequence of real numbers converges to a limit.
Proof:
Let {𝑎𝑛 } be a Cauchy sequence.
For any j, there is a natural number Nj so that whenever n, m ≥ 𝑁𝐽 , we
have that |𝑎𝑛 − 𝑎𝑚 | ≤ 2 −j . We now consider the sequence {𝑏𝑗 } given
by 𝑏𝑗 = 𝑎𝑁𝑗 − 2−𝑗 .
Notice that for every n larger than 𝑁𝑗 , we have that 𝑎𝑛 > 𝑏𝑗 . Thus, each
bj serves as a lower bound for elements of the Cauchy sequence {𝑎𝑛 }
occuring later than 𝑁 𝑗 .
Each element of the sequence {𝑏𝑗 } is bounded above by 𝑏1 + 1, for the
same reason. Thus, the sequence {𝑏𝑗 } has a least upper bound which
we denote by L. We will show that L is the limit of the sequence e {𝑎𝑛 }.
Suppose that n > 𝑁𝑗 .
Then |𝑎𝑛 − L| < 2 −j + |𝑎𝑛 − 𝑏𝑗 | = 2−j + 𝑎𝑛 − 𝑏𝑗 ≤ 3(2−𝑗 ).
For every > 0 there is j(𝜖) so that 21−𝑗 < and we simply take N(𝜖) to
𝑁𝑗 (𝜖) .
The idea of the proof of Theorem 1 is that we recover the limit of the
Cauchy sequence by taking a related least upper bound. So, we can
think of the process of finding the limit of the Cauchy sequence as
specifying the decimal expansion of the limit, one digit at a time, as this
how the least upper bound property worked.
The converse of Theorem 1 is also true.
Theorem:
Let {𝑎𝑛 } be a sequence of real numbers converging to a linit L. Then
the sequence {𝑎𝑛 } is a Cauchy sequence.
Proof:
Since {𝑎𝑛 } converges to L, for every 𝜖 > 0, there is an N > 0 so that
𝜖
when j > N, we have |𝑎𝑗 − L| ≤ 2.
𝜖 𝜖
(The reason we can get 2on the right hand side is that we put 2in the
role of 𝜖 in the definition of the limit.) Now if j and k are both more than
𝜖 𝜖
N, we have |𝑎𝑗 − L| ≤ 2 and |𝑎𝑗 − L| ≤ 2 . Combining these using the
triangle inequality, we get
|𝑎𝑗 − 𝑎𝑘 | ≤ 𝜖,
so that the sequence {𝑎𝑗 } is a Cauchy sequence as desired.
Combining Theorems above, we see that what we have learned is that
Cauchy sequences of real numbers and convergent sequences of real
numbers are the same thing. But the advantage of the Cauchy criterion

170
is that to check whether a sequence is Cauchy, we don’t need to know
the limit in advance.
Example:
Consider the series (that is, infinite sum)
S = ∑∞𝑛=1 1𝑛2 .
We may view this as the limit of the sequence of partial sums
𝑗
𝑎𝑗 = ∑𝑛=1 1𝑛2.
We can show that the limit converges using Theorem 1 by showing that
{𝑎𝑗 } is a Cauchy sequence. Observe that if j, k > N, we definitely have
𝑗
|𝑎𝑗 − 𝑎𝑘 | ≤ ∑𝑛=𝑁 1𝑛2.
It may be difficult to get an exact expression for the sum on the right,
but it is easy to get an upper bound.
∞ ∞ ∞
1 1 1
∑ 1𝑛2 ≤ ∑ = ∑ − .
𝑛(𝑛 − 1) 𝑛(𝑛 − 1) 𝑛
𝑛=𝑁 𝑛=𝑁 𝑛=𝑁
1
The reason we used the slightly wasteful inequality, replacing by
𝑛2
1
that now the sum on the right telescopes, and we know it is exactly
𝑛2−𝑛
1
equal to 𝑁−1 . To sum up, we have shown that when j, k > 𝑁, we have
1
|𝑎𝑗 − 𝑎𝑘 | ≤ 𝑁−1 .
Since we can make the right hand side arbitrarily small by taking N
sufficiently large, we see that {𝑎𝑗 } is a Cauchy sequence. This example
gives an indication of the power of the Cauchy criterion. You would not
have found it easier to prove that the limit exists if I had told you in
𝜏2
advance that the series converges to 6 .
Let {𝑎𝑛 } be a sequence of real numbers. Let {𝑛𝑘 } be a strictly increasing
sequence of natural numbers. We say that {𝑎𝑛𝑘 } is a subsequence of
{𝑎𝑛 } . We will now prove an important result which helps us discover
convergent sequences in the wild.
Theorem 3:
(Bolzano-Weierstrass)
Let {𝑎𝑛 be a bounded sequence of real numbers. (That is, suppose
there is a positive real number B, so that |𝑎𝑗 | ≤ B for all j.) Then {𝑎𝑛 }
has a convergent subsequence.
Proof of Bolzano-Weierstrass
All the terms of the sequence live in the interval
𝐼0 = [−B, B].
We can make the right-hand side arbitrarily small by making N
sufficiently large. Thus, we have shown that the subsequence is a
Cauchy sequence and hence convergent.
A question you might ask yourselves is: How is the proof of the
Bolzano Weierstrass theorem related to decimal expansions?

171
Our final topic for today’s lecture is the Squeeze theorem. It is a result
that allows us to show that limits converge by comparing them to limits
that we already know converge.
Theorem 4:
(Squeeze theorem)
Given three sequences of real numbers {𝑎𝑛 }, {𝑏𝑛 }, and {𝑐𝑛 }. If we know
{𝑎𝑛 } and {𝑏𝑛 }, that both converge to the same limit L and we know that
for each n we have
𝑎𝑛 ≤ 𝑐𝑛 ≤ 𝑏𝑛 ,
then the sequence {𝑐𝑛 } also converges to the limit L.

172
Exercises:

1. Given the sequence An = {1, 3, 5, 7, 9},


(a) what is the value of a3?
(b) Find the value of ∑5𝑛=1 𝑎𝑛

2. Expand the following series and find the sum: ∑4𝑛=0 2𝑛

3. List the first four terms of the sequence {an} = {n2}, starting with n = 1.

4. List the first four terms of the following sequence, beginning with n = 0.

5. Find the sum of the first six terms of An, where an = 2an–1 + an–2, a1 = 1, and a2 =1.

173
Answer Key: Subsequences
1. Given the sequence An = {1, 3, 5, 7, 9},
(a) what is the value of a3?
(b) Find the value of ∑5𝑛=1 𝑎𝑛

Solution:
(a) The index of a3 is n = 3 so they're asking me for the third term, which is "5".
(b) The funky symbol is the Greek capital letter "sigma", indicating a series. That
means that they're asking me here to do the addition of the terms of the sequence.
The "value" they're asking me to find is the total, the sum, of all the
terms an from a1 to a5; in other words:
a1 + a2 + a3 + a4 + a5
= 1 + 3 + 5 + 7 + 9 = 25
Then my answers are:
value of a3: 5
value of sum: 25

2. Expand the following series and find the sum: ∑4𝑛=0 2𝑛

They've given me a rule for each term of this series; the rule is to multiply the index by
two. So, to find each term, I'll plug the value of n into the formula; namely, I'll take the
index and multiply by two. I'll be starting with n = 0 and ending with n = 4. To find the
series sum, I'll be adding all the terms, like this:
2(0) + 2(1) + 2(2) + 2(3) + 2(4)
= 0 + 2 + 4 + 6 + 8 = 20

3. List the first four terms of the sequence {an} = {n2},


I'll just plug n into the formula, and simplify:
{a1, a2, a3, a4}
= {12, 22, 32, 42}
= {1, 4, 9, 16}
My answer is the simplified form of the sequence:
{1, 4, 9, 16}

4. List the first four terms of the following sequence, beginning with n = 0

An=(n+1)!(−1)n
Sequences and series are often the first place students encounter this exclamation-
mark notation. The notation doesn't indicate that the series is "emphatic" in some
manner; instead, this is technical mathematical notation. It indicates that the terms of
this summation involve factorials. (If you're not familiar with factorials, brush up now.)
A factorial symbol, k!, indicates that I need to find the product of all the whole numbers
from 1 through k. The first few factorial values are: 1! = 1

174
2! = 1 × 2 = 2
3! = 1 × 2 × 3 = 6
4! = 1 × 2 × 3 × 4 = 24
(Your graphing calculator can probably find factorials for you. Look for an appropriate
command, probably somewhere in a "Prob" or "Probability" submenu.)
I'll use these factorial values in my computations:
I'll use these factorial values in my computations:
1 1
𝑎0= (−1)0 =1!=1=1
(0+1)!
(−1)1 −1 1
𝑎1 =(1+1)!= 2! =− 2
(−1)2 1 1
𝑎2 = = =
(2 + 1)! 3! 6
(−1)3 −1 1
𝑎3 =((3+1)= 4! =− 24
So the first four terms are:
1 1 1
1, 2, 6,- 24

5. Find the sum of the first six terms of An, where an = 2an–1 + an–2, a1 = 1, and a2 =1.

This formula looks much worse than it really is; I just have to give myself some time,
and dissect the formula carefully.
They gave me the values of the first two terms, and then they gave me a formula that
says that each term (after the first two terms) is a sum formed from the previous two
terms. At each stage, I'll be taking the previous term and multiplying it by two; to this,
I'll be adding the term before that one. For instance, the third term will be twice the
second term, plus the first term.
Plugging into this formula, I get:
a3 = 2a3–1 + a3–2
= 2a2 + a1
= 2(1) + (1)
=2+1=3
a4 = 2a4–1 + a4–2
= 2a3 + a2
= 2(3) + (1)
=6+1=7
a5 = 2a5–1 + a5–2
= 2a4 + a3
= 2(7) + (3)
= 14 + 3 = 17
a6 = 2a6–1 + a6–2
= 2a5 + a4
= 2(17) + (7)

175
= 34 + 7 = 41
Now that I've found the values of the third through the sixth terms, I can find the value
of the series; the sum is:
1 + 1 + 3 + 7 + 17 + 41 = 70

176
THE CAUCHY CRITERION

KARLENE MAE GUINSISANA

177
THE CAUCHY CRITERION

Definition:
A sequence {𝑥𝑛 } of real numbers is said to be Cauchy sequence if for
every ε > 0 there exists N ∈N such that if n, m > N ⇒|𝑥𝑛 −𝑥𝑚 |< ε.
A sequence is Cauchy if the terms eventually get arbitrarily close to each
other.
Example:
1
The sequence {𝑛} is Cauchy. To see this let ε > 0 be given. Choose N ∈ N
1 𝜀 1 1 1 1 𝜀 𝜀
such that 𝑁 <2. Now, if n, m > N ⇒|𝑛 − 𝑚| ≤ 𝑛 + 𝑚 < 2 + 2 = ε.

Example:
𝑛
The sequence { } is Cauchy. To see this let ε > 0 be given. Choose N ∈N
𝑛+1
1 𝜀 𝑛 𝑚
such that 𝑁 <2. Now, if n, m > N ⇒| 𝑛+1 − 𝑚+1|=
𝑚+1−𝑛−1 𝑚−𝑛 1 1 𝜀 𝜀
| (𝑛+1)(𝑚+1)|≤| 𝑛𝑚 |< 𝑛 + 𝑚 < 2 + 2= ε.

Lemma:
Let sequence {𝑥𝑛 } be a Cauchy sequence of real numbers. Then {𝑥𝑛 } is
bounded.
Proof:
Since {𝑥𝑛 } is a Cauchy sequence, then there exists N ∈N such that if n, m > N
⇒|𝑥𝑛 −𝑥𝑚 |< 3.
if n, m > N ⇒|𝑥𝑛 −𝑥𝑚 |< 3
Let m = N +1, if n > N ⇒|𝑥𝑛 −𝑥𝑁 +1|< 3 Note: |𝑥𝑛 |−|𝑥𝑁 +1|≤|𝑥𝑛 −𝑥𝑁 +1|
⇒|𝑥𝑛 |−|𝑥𝑁 +1|≤|𝑥𝑛 −𝑥𝑁 +1|< 3
if n > N ⇒|𝑥𝑛 |< 3+|𝑥𝑁 +1|.
Let M = max {|𝑥1 |,|𝑥2 |,···|𝑥𝑁 |,|𝑥𝑁 +1|+3}
Now, if n > N ⇒|𝑥𝑛 |< 3+|𝑥𝑁 +1|≤M
Now, if n ≤ N ⇒|𝑥𝑛 |< max {|𝑥1 |,|𝑥2 |,···|𝑥𝑁 |}≤M
Thus ∀ n ∈ N, |𝑥𝑛 |≤ M. ∎

178
Theorem [Cauchy Convergence Criterion]

A sequence of real numbers is convergent if and only if it is a


Cauchy sequence.
Proof:
Let {𝑥𝑛 } be a sequence of real numbers.
(⇒) Suppose that limn→∞𝑥𝑛 = x ∈R. We want to show that {𝑥𝑛 } is
Cauchy sequence.
Give let ε > 0. Since 𝑙𝑖𝑚𝑛→∞ 𝑥𝑛 = x ∴ ∃ N ∈N 3
𝜀
if n > N ⇒|𝑥𝑛 −x|<2
𝜀
Also, if m > N ⇒|𝑥𝑚 −x|< 2
𝜀 𝜀
Now, if n, m > N ⇒|𝑥𝑛 −𝑥𝑚 |=|𝑥𝑛 –x + x−𝑥𝑚 |≤|𝑥𝑛 −x|+|𝑥𝑚 −x|<2 + 2= ε.

Thus {𝑥𝑛 } is a Cauchy sequence.

(⇐) Suppose that {𝑥𝑛 } is a Cauchy sequence. We want to show that {𝑥𝑛 } is
convergent.
Let ε > 0 be given. Since {𝑥𝑛 } is a Cauchy sequence, then by Lemma 1
it is bounded.
Hence {𝑥𝑛 } has a converge subsequence {𝑥𝑛𝑘 }.
Suppose lim 𝑥𝑛𝑘 = x∈R.
𝑘→∞
𝜀
There exist 𝑁1 ,𝑁2 ∈N 3 if n, m > 𝑁1 ⇒|𝑥𝑛 − 𝑥𝑚 |< 2
𝜀
And, if k > 𝑁2 ⇒|𝑥𝑛𝑘 −x|<2.
𝜀
Now, fix k > 𝑁2 such that 𝑛𝑘 > 𝑁1 and, if n > 𝑁1 ⇒|𝑥𝑛 −𝑥𝑛𝑘 |< 2 and |𝑥𝑛𝑘
𝜀
−x|<2.
𝜀 𝜀
Now, if n > 𝑁1 ⇒|𝑥𝑛 −x|=|𝑥𝑛 −𝑥𝑛𝑘 +𝑥𝑛𝑘 −x|≤|𝑥𝑛 −𝑥𝑛𝑘 |+|𝑥𝑛𝑘 −x|< 2 + 2= ε.

Thus {𝑥𝑛 } converges. ∃. ∎

179
Cauchy’s insight
Our difficulty in proving “𝑎𝑛 → l” is this: What is `? Cauchy saw that it was
enough to show that if the terms of the sequence got sufficiently close to
each other then completeness will guarantee convergence.
Remark. In fact, Cauchy’s insight would let us construct R out of Q if we
had time.

Definition
Let (𝑎𝑛 ) be a sequence [R or C]. We say that (an) is a Cauchy sequence if,
for all ε > 0 there exists N ∈N such that
m, n > N =⇒|𝑎𝑚 −𝑎𝑛 | < ε.
[Is that all? Yes, it is!

Cauchy ⇒ Bounded
Theorem:
Every Cauchy sequence is bounded [R or C].
Proof:
1 > 0 so there exists N such that m, n > N =⇒ |𝑎𝑚 − 𝑎𝑛 | < 1. So form > N,
|𝑎𝑚 |6 1+|𝑎𝑁 | by the ∆ law. So for all m
|𝑎𝑚 |6 1+|𝑎1 |+|𝑎2 |+···+|𝑎𝑁 |. ∎

Convergent ⇒ Cauchy [R or C]
Theorem:
Every convergent sequence is Cauchy.
Proof:
Let 𝑎𝑛 → l and let ε > 0. Then there exists N such that
k > N =⇒|𝑎𝑘 −l| < ε/2
For m, n > N we have
|𝑎𝑚 −l| < ε/2
|𝑎𝑛 −l| < ε/2

180
So
|𝑎𝑚 −𝑎𝑛 | ≤ |𝑎𝑚 −l| + |𝑎𝑛 −l| by the ∆ law
< ε/2 + ε/2 = ε ∎

Cauchy ⇒ Convergent [R]


Theorem:
Every real Cauchy sequence is convergent.
Proof:
Let the sequence be (𝑎𝑛 ). By the above, (𝑎𝑛 ) is bounded. By Bolzano-
Weierstrass (𝑎𝑛 ) has a convergent subsequence (𝑎𝑛𝑘 ) → l, say. So let ε > 0.
Then
∃N1 such that r ⩾ N1 =⇒ |𝑎𝑛𝑟 −l| < ε/2
∃N2 such that m, n ⩾ N2 =⇒ |𝑎𝑚 −𝑎𝑛 | < ε/2
Put s: = min {r|𝑛𝑟 ⩾ N2} and put N =𝑛𝑠 . Then
M, n⩾ N ⇒ |𝑎𝑚 −𝑎𝑛 |
⩽ |𝑎𝑚 −𝑎𝑛𝑠 | + |𝑎𝑛𝑠 −l|
< ε/2 + ε/2 = ε ∎

Cauchy =⇒ Convergent [C]


Theorem:
Every complex Cauchy sequence is convergent.
Proof:
Put 𝑧𝑛 = x + iy. Then 𝑥𝑛 is Cauchy: |𝑥𝑥 −𝑥𝑚 | 6 |𝑧𝑛 −𝑧𝑚 | (as |Rw| ⩽ |w|).
So 𝑥𝑛 → x, 𝑦𝑛 → y and so 𝑧𝑛 → x+iy. ∎

Cauchy Sequence in R
Definition:
A sequence 𝑥𝑛 ∈R is said to converge to a limit x if

181
• ∀² > 0, ∃N s. t. n > N ⇒ |𝑥𝑛 −x|< ².
A sequence 𝑥𝑛 ∈R is called Cauchy sequence if
• ∀², ∃N s. t. n > N & m > N ⇒ |𝑥𝑛 −𝑥𝑚 |< ².
Proposition:
Every convergent sequence is a Cauchy sequence.
Proof:
Assume 𝑥𝑘 →x. Let ² > 0 be given.
• ∃N s .t. n > N ⇒ |𝑥𝑛 −x|< ² 2.
• n, m ≥N ⇒ |𝑥𝑛 −𝑥𝑚 |≤ |x −𝑥𝑛 |+|x −𝑥𝑚 |< ² 2 + ² 2 = ².
Theorem:
Every bounded sequence in R has a subsequence that converges to some
point in R.
Proof:
Suppose 𝑥𝑛 is a bounded sequence in R. ∃M such that
−M ≤𝑥𝑛 ≤M, n = 1, 2,···. Select 𝑥𝑛0 =𝑥1 .
• Bisect 𝑙0 := [−M, M] into [−M, 0] and [0,M].
• At least one of these (either [−M, 0] or [0, M]) must contain 𝑥𝑛 for
infinitely many indices n.
• Call it I1 and select n1 > n0 with xn1 ∈I0.
• Continue in this way to get a subsequence 𝑥𝑛𝑘 such that
• 𝑙0 ⊃𝑙1 ⊃𝑙2 ⊃𝑙3 ···
• 𝑙𝑘 = [𝑎𝑘 ,𝑏𝑘 ] with |𝑙𝑘 |= 2−𝑘 M.
• Choose 𝑛0 < 𝑛1 < 𝑛2 <··· with 𝑥𝑛𝑘 ∈ 𝑙𝑘.
• Since 𝑎𝑘 ≤𝑎𝑘 +1 ≤M (monotone and bounded), 𝑎𝑘 →∃ x.
• Since 𝑥𝑛𝑘 ∈𝑙𝑘 and |𝑙𝑘 |= 2−𝑘 M, we have
|𝑥𝑛𝑘 −x|<|𝑥𝑛𝑘 −𝑎𝑘 |+|𝑎𝑘 −x|≤2−𝑘−1 M+|𝑎𝑘 −x|→0 as k →∞. 2

182
Corollary: (Compactness)
Every sequence in the closed interval [a ,b] has a subsequence in R that
converges to some point in R.
Proof:
Assume a≤𝑥𝑛 ≤b for n = 1, 2,···. By Theorem 1.4.3, ∃ a subsequence
𝑥𝑛𝑘 and a≤ ∃𝑥 ≤b such that 𝑥𝑛𝑘 →x.
Lemma. (Boundedness of Cauchy sequence)
If 𝑥𝑛 is a Cauchy sequence, 𝑥𝑛 is bounded.
Proof:
∃N s. t. n ≥N ⇒ |𝑥𝑛 −x|< 1. Then 𝑠𝑢𝑝𝑛 |𝑥𝑛 | ≤1 + max {|x1|, ···,|𝑥𝑁 |} (Why?)
Theorem. (Completeness)
Every Cauchy sequence in R converges to an element in [a, b].
Proof:
Cauchy seq. ⇒ bounded seq. ⇒ convergent sub seq.

Uniform Convergence
Definition:
A sequence of functions {𝑓𝑛 } is said to converge uniformly on an Interval [a,b]
to a function f if for any ε > 0 and for all x ∈ [a, b] there exists an integer N
(independent of x but dependent on ε) such that for all x∈[a, b]
|𝑓𝑛 (x) − f(x)| < ε, ∀n≥N (1)
Remark. Every uniformly convergent sequence is point wise convergent, and
the uniform limit function is same as the pointwise limit function. However, the
converse is not true. However non-pointwise convergence implies non-
uniform convergence.
Definition:
A series of functions ∑ 𝑓𝑛 is said to converge uniformly on [a, b] if the
sequence {𝑆𝑛 } of its partial sums, defined by
S n(x) =∑𝑛𝑖−! 𝑓𝑖 (x)

183
Converges uniformly on [a, b].
Definition:
A series of functions ∑ 𝑓𝑛 converges uniformly to f on [a, b] if for ∈ > 0 and all
x ∈ [a, b] there exists an integer N (independent of x and dependent on ε)
such that for all x in [a, b]
|𝑓1 (x) + 𝑓2 (x) + … + 𝑓𝑛 (x) − f(x)| < ε, for n ≥ N

Cauchy’s Criterion for Uniform Convergence


Theorem 1:
The sequence of functions {𝑓𝑛 } defined on [a, b] converges uniformly on [a, b]
if and only if for every ε > 0 and for all x ∈ [a, b], there exists an integer N such
that
|𝑓𝑛+𝑝 (x) − 𝑓𝑛 (x) | < ε, ∀ n ≥ N, p ≥ 1 … (1)

Proof:
Let the sequence {𝑓𝑛 } uniformly converge on [a, b] to the limit function f, so
that for a given ε > 0, and for all x ∈ [a, b], there exist integers n1, n2 such that
| 𝑓𝑛 (x) − f(x)| < ε/2, ∀ n ≥ n1
and
|𝑓𝑛+𝑝 (x) − f(x)| < ε/2, ∀ n ≥𝑛2 , p ≥ 1

Let N = max (𝑛1 , 𝑛2 ).


∴ |𝑓𝑛+𝑝 (x) − 𝑓𝑛 (x)| ≤ |𝑓𝑛+𝑝 (x) − f(x)| + |𝑓𝑛 (x) − f(x)|

< ε/2 + ε/2 = ε, ∀ n ≥ N, p ≥ 1


Conversely. Let the given condition hold so by Cauchy’s general principle of
convergence, {𝑓𝑛 } converges for each x ∈ [a, b] to a limit, say f and so the
sequence converges pointwise to f.
For a given ε > 0, let us choose an integer N such that (1) holds. Fix n, and let
p→∞ in (1). Since 𝑓𝑛+𝑝 →f as p → ∞, we get

|f(x) − 𝑓𝑛 (x)| < ε ∀ n ≥ N, all x ∈ [a, b]

184
Which proves that 𝑓𝑛 (x) → f(x) uniformly on [a, b].
Remark: Other form of this theorem is
The sequence of functions {𝑓𝑛 } defined on [a, b] converges uniformly on [a,
b]if and only if for every ε > 0 and for all x ∈ [a, b], there exists an integer N
such that
|𝑓𝑛 (x) − 𝑓𝑚 (x)| < ε, ∀ n, m ≥ N

Theorem 2:
A series of functions ∑ 𝑓𝑛 defined on [a, b] converges uniformly on [a, b] if
and only if for every ε > 0 and for all x∈ [a, b], there exists an integer N such
that
|𝑓𝑛+1 (x) + 𝑓𝑛+2 (x) +…+ (x)| < ε, ∀ n ≥ N, p ≥ 1 … (2)
Proof:
Taking the sequence {𝑠𝑛 } of partial sums of functions ∑ 𝑓𝑛, defined by
S n(x) = ∑𝑛𝑖−1 𝑓𝑖(𝑥)
In addition, applying above theorem, we get the result.

Example:.
Show that the sequence {𝑓𝑛 }, where
𝑛𝑥
𝑓𝑛 (x) =1+𝑛2𝑥2, for x ∈ [a, b].

is not uniformly convergent on any interval [a, b] containing 0.


Solution:
The sequence converges pointwise to f, where f(x) = 0, ∀ real x. Let {𝑓𝑛 }
converge uniformly in any interval [a, b], so that the pointwise limit is also the
uniform limit. Therefore for given ε>0, there exists an integer N such that for
all x∈ [a, b], we have
𝑛𝑥
| 1+𝑛2𝑥2-0 | < ε, ∀n≥N

185
1
If we take ε =3, and t an integer greater than N such that 1/t ∈ [a, b],

we find on
Taking n = t and x = 1/t, that
𝑛𝑥 1 1
= 2 ∢ 3 = 𝜖.
1+𝑛2𝑥2

Which is a contradiction and so the sequence is not uniformly convergent in


the interval [a, b], having the point 1/t. However, since 1/t→0, the interval [a,
b] contains 0. Hence the sequence is not uniformly convergent on any interval
[a, b] containing 0.

Definition:
We say that a sequence of real numbers{an}is a Cauchy sequence provided
that for every > 0, there is a natural number N so that when n, m ≥ N, we
have that |an −am|≤ .
Example 1 Let x be a real number and 𝑡𝑛 (x) be the nth truncation of its
decimal expansion as in Lectures 2 and 3. Then if n, m ≥ N, we have that
|𝑡𝑛 (x) −𝑡𝑚 (x) |≤ 10−N, since they share at least the first N places of their
decimal expansion. Given any real number > 0, there is an N (𝜖) so that
10−𝑁(𝜖) <𝜖. Thus, we have shown that the sequence {𝑡𝑛 (x)} is a Cauchy
sequence.
Example 1 was central in our construction of the real numbers. We got the
least upper bound property by associating to each sequence as in Example 1,
the real number x which is its limit. The class of Cauchy sequences should be
viewed as minor generalization of Example 1 as the proof of the following
theorem will indicate.
Theorem 1:
Every Cauchy sequence of real numbers converges to a limit.
Proof:
Let {an} be a Cauchy sequence. For any j, there is a natural number 𝑁𝑗 so that
whenever n, m ≥ 𝑁𝑗 , we have that |𝑎𝑛 −𝑎𝑚 |≤ 2−𝑗 j. We now consider the
sequence {𝑏𝑗 } given by

𝑏𝑗 = 𝑎𝑁𝑗 −2−𝑗.

186
Notice that for every n larger than 𝑁𝑗 , we have that an >𝑏𝑗 . Thus, each
𝑏𝑗 serves as a lower bound for elements of the Cauchy sequence {𝑎𝑛 }
occurring later than𝑁𝑗 . Each element of the sequence {bj} is bounded above
by b1 + 1, for the same reason. Thus, the sequence {𝑏𝑗 } has a least upper
bound which we denote by L. We will show that L is the limit of the sequence
{an}. Suppose that n >𝑁𝑗 . Then

|𝑎𝑛 −L| < 2−𝑗 +|𝑎𝑛 −𝑏𝑗 | = 2−𝑗 + 𝑎𝑛 −𝑏𝑗 ≤ 3(2−𝑗 ).

For every > 0 there is j (𝜖) so that 21−j < and we simply take N (𝜖) to 𝑁𝑗 (𝜖).

The idea of the proof of Theorem 1 is that we recover the limit of the Cauchy
sequence by taking a related least upper bound. Therefore, we can think of
the process of finding the limit of the Cauchy sequence as specifying the
decimal expansion of the limit, one digit at a time, as this how the least upper
bound property worked.
The converse of Theorem 1 is also true.
Theorem 2:
Let {𝑎𝑛 } be a sequence of real numbers converging to a limit L. Then the
sequence {𝑎𝑛 } is a Cauchy sequence.
Proof:
Since {𝑎𝑛 } converges to L, for every > 0, there is an N > 0 so that when j > N,
we have
𝜖
|𝑎𝑗 −L|≤ 2
𝜖 𝜖
(The reason we can get 2 on the right hand side is that we put in the role of
2
in the definition of the limit.) Now if j and k are both more than N, we have |𝑎𝑗
𝜖
−L|≤ 2 and
𝜖
|𝑎𝑘 −L|≤2. Combining these using the triangle inequality, we get

|𝑎𝑗 −𝑎𝑘 |≤ 𝜖.

So that the sequence {𝑎𝑗 } is a Cauchy sequence as desired.

Combining Theorems 1 and 2, we see that what we have learned is that


Cauchy sequences of real numbers and convergent sequences of real
numbers are the same thing. However, the advantage of the Cauchy criterion
is that to check whether a sequence is Cauchy, we do not need to know the
limit in advance.

187
EXERCISES


1. (A) Show that if {𝑎𝑛 } 𝑛=1
is Cauchy then {a𝑛2 }𝑛=1

is also Cauchy.

(B) Give an example of a Cauchy sequence {a𝑛2}𝑛=1


∞ ∞
such that {𝑎𝑛 }𝑛=1

Is not Cauchy.


2. Let {𝑎𝑛 }𝑛=1 be a Cauchy sequence such that 𝑎𝑛 is an integer for all n ∈ N. Show that
there is a positive integer N such that 𝑎𝑛 = C for all n ≥ N, where C is a constant.


3. What does it mean for a sequence {𝑎𝑛 }𝑛=1 to not be Cauchy?

4. Prove that if a subsequence of a Cauchy sequence converges to L, then the full


sequence also converges to L.

5. Consider a sequence defined recursively by 𝑎1 = 1 and 𝑎𝑛+1 = 𝑎𝑛 + (−1)𝑛 𝑛3 for all


n ∈ N. Show that such a sequence is not a Cauchy sequence. Does this sequence
converge?

188
ANSWER KEY


1. (A) since {𝑎𝑛 }𝑛=1 is Cauchy, it is convergent. Since the product of two convergent
sequences is convergent the sequence {a𝑛2}𝑛=1

is convergent and therefore is Cauchy.

(b) Let 𝑎𝑛 = (−1)𝑛 for all n ∈N. The sequence { 𝑎𝑛 } 𝑛=1 is not Cauchy since it is
divergent. However, the sequence {a𝑛2}𝑛=1

= {1,1,···} converges to 1so it is Cauchy∎

1 ∞
2. Let ∈ =2. Since {𝑎𝑛 }𝑛=1 is Cauchy, there is a positive integer N such that if m, n ≥ N
1
we have |𝑎𝑚 −𝑎𝑛 | <2. But am−an is an integer so we must have 𝑎𝑛 = 𝑎𝑁 for all n ≥ N∎


3. A sequence {𝑎𝑛 }𝑛=1 is not a Cauchy sequence if there is a real number > 0 such
that for all positive integers N there exist n,m ∈N such that n, m ≥ N and |𝑎𝑛 −𝑎𝑚 |≥ ∈

∞ ∞ ∞
4. Let {𝑎𝑛 } 𝑛=1 be a Cauchy sequence. Let {𝑎𝑛𝑘 } 𝑘=1 be a subsequence of {𝑎𝑛 } 𝑛=1

converging to L. By Exercise? The sequence {𝑎𝑛 }𝑛=1 is convergent say to a limit L0.
By Exercise? We must have L = L’ ∎

5. We will show that there is an ∈ > 0 such that for all N ∈N there exist m and n such
that m, n ≥ N but |𝑎𝑚 −𝑎𝑛 |≥ . Note that |𝑎𝑛+1 −𝑎𝑛 | = 𝑛3 ≥ 1. Let = 1. Let N ∈N. Choose
m = N + 1 and n = N. In this case, |𝑎𝑚 −𝑎𝑛 | = 𝑁 3 ≥ 1 =. Hence, the given sequence is
not a Cauchy sequence. Since every convergent sequence must be Cauchy, the given
sequence is divergent.∎

189
PROPERLY DIVERGENT SEQUENCE

MARIA MENNIE COMIDA

190
PROPERLY DIVERGENT SEQUENCES
Definition:
Let {𝑥𝑛 } be sequence of real numbers.
i. We say {xn } diverges to ∞ and we write lim𝑛→∞ 𝑥𝑛 = ∞, if for every
𝝀 > 0 there exists 𝑁 ∈ ℕ such that if 𝑛 > 𝑁 ⟹ 𝑥𝑛 > 𝝀.
ii. We say {xn } diverges to −∞ and we write lim𝑛→∞ 𝑥𝑛 = −∞, if for
every 𝝀 > 0 there exists 𝑁 ∈ ℕ such that if 𝑛 > 𝑁 ⟹ 𝑥𝑛 > −𝝀.

We say that {xn } is properly divergent if lim𝑛→∞ 𝑥𝑛 = ∞ or


lim𝑛→∞ 𝑥𝑛 = −∞.

Example 1:
The sequence {𝑛} is Properly divergent and lim𝑛→∞ 𝑥𝑛 = ∞. To see
this let 𝝀 > 0 be given. Choose 𝑁 ∈ ℕ such that 𝜆 < 𝑁. Now, if
𝑛 > 𝑁 ⟹ 𝑛 > 𝑁 ⟹ 𝝀. Hence, 𝑙𝑖𝑚𝑛→∞ 𝑛 = ∞.
Example 2:
𝑛2 𝑛2
The sequence {𝑛+1} is Properly divergent and lim𝑛→∞ 𝑛+1 = ∞. To see
this let 𝝀 > 0 be given. Choose 𝑁 ∈ ℕ such that 𝑁 > 2𝝀. Now, if
𝑛2 𝑛2 𝑛 𝑁 2𝜆
𝑛 > 𝑁 ⟹ 𝑛+1 > 2𝑛 = > > = 𝝀.
2 2 2

Lemma 1:
Let sequence {𝑥𝑛 } and {𝑦𝑛 } be two sequences of real numbers such
that 𝑥𝑛 ≤ 𝑦𝑛 ∀ 𝑛 ∈ ℕ.
i. If lim𝑛→∞ 𝑥𝑛 = ∞, then lim𝑛→∞ 𝑦𝑛 = ∞
ii. If lim𝑛→∞ 𝑦𝑛 = ∞, then lim𝑛→∞ 𝑥𝑛 = −∞

Proof:
i. Let 𝝀 > 0 be given. Since lim𝑛→∞ 𝑥𝑛 = ∞ ∴ ∃ 𝑁 ∈ ℕ such that if 𝑛 >
𝑁 ⟹ 𝑥𝑛 > 𝝀. Now, 𝑥𝑛 ≤ 𝑦𝑛 , then, if 𝑛 > 𝑁 ⟹ 𝑦𝑛 ≥ 𝑥𝑛 > 𝝀 .
Thus, lim𝑛→∞ 𝑦𝑛 = ∞.
ii. Let 𝝀 > 0 be given. Since lim𝑛→∞ 𝑦𝑛 = −∞ ∴ ∃ 𝑁 ∈ ℕ such that if 𝑛 >
𝑁 ⟹ 𝑦𝑛 < −𝝀. Now, 𝑥𝑛 ≤ 𝑦𝑛 , then, if 𝑛 > 𝑁 ⟹ 𝑥𝑛 ≤ 𝑦𝑛 < −𝝀 .
Thus, lim𝑛→∞ 𝑥𝑛 = −∞.
Notations:
Let {𝑥𝑛 } be sequence of real numbers.
Let 𝐴1 = {𝑥1 , 𝑥2 , … , 𝑥𝑛 , … } = {𝑥𝑘 ⃒𝑘 ≥ 1}

𝐴2 = {𝑥2 , 𝑥3 , … , 𝑥𝑛 , … } = {𝑥𝑘 ⃒𝑘 ≥ 2}

𝐴𝑛 = {𝑥𝑛 , 𝑥𝑛+1 , … , … } = {𝑥𝑘 ⃒𝑘 ≥ 𝑛}
Note 𝐴1 ⊇ 𝐴2 ⊇ ⋯ ⊇ 𝐴𝑛 ⊇ 𝐴𝑛+1 ⊇ ⋯.

191
Now, 𝑠𝑢𝑝𝐴1 ≥ 𝑠𝑢𝑝𝐴2 ≥ ⋯ ≥ 𝑠𝑢𝑝𝐴𝑛 ≥ 𝑠𝑢𝑝𝐴𝑛+1 ≥ ⋯,
and 𝑖𝑛𝑓𝐴1 ≤ 𝑖𝑛𝑓𝐴2 ≤ ⋯ ≤ 𝑖𝑛𝑓𝐴𝑛 ≤ 𝑖𝑛𝑓𝐴𝑛+1 ≤ ⋯.

Now, we have two sequences {𝑠𝑢𝑝𝐴𝑛 } which is decreasing, and


{𝑖𝑛𝑓𝐴_𝑛} which is increasing.
We define
𝑖𝑛𝑓 𝑖𝑛𝑓 sup 𝑥𝑘 𝑙𝑖𝑚 sup 𝑥𝑘
lim sup 𝑥𝑛 = sup 𝐴𝑛 = =
𝑛≥1 𝑛 ≥ 1𝑘 ≥ 𝑛 𝑛 → ∞𝑘 ≥ 𝑛
and

𝑠𝑢𝑝 𝑠𝑢𝑝 inf 𝑥𝑘 𝑙𝑖𝑚 inf 𝑥𝑘


lim 𝑖𝑛𝑓 𝑥𝑛 = inf 𝐴𝑛 = =
𝑛≥1 𝑛 ≥ 1𝑘 ≥ 𝑛 𝑛 → ∞𝑘 ≥ 𝑛

Example 3:
Let 𝑥𝑛 = 3 − (−1)𝑛 ∀ 𝑛 ∈ ℕ. Find lim 𝑠𝑢𝑝𝑥𝑛 and lim 𝑖𝑛𝑓𝑥𝑛 .
𝐴1 = {4,2,4,2, … }, sup 𝐴1 = 4 and inf 𝐴1 = 2.
𝐴2 = {2,4,2,4, … }, sup 𝐴2 = 4 and inf 𝐴2 = 2.

It is clear that sup 𝐴𝑛 = 4 and inf 𝐴𝑛 = 2 ∀ 𝑛 ∈ ℕ.
Hence, lim sup 𝑥𝑛 = lim sup 𝐴𝑛 = lim 4 = 4 and
𝑛→∞ 𝑛→∞
lim inf 𝑥𝑛 = lim inf 𝐴𝑛 = lim 2 = 2.
𝑛→∞ 𝑛→∞

Example 4:
1
Let 𝑥𝑛 = (−1)𝑛 + 𝑛 ∀ 𝑛 ∈ ℕ. Find lim sup 𝑥𝑛 and lim inf 𝑥𝑛 .
1 1 1 1
𝐴1 = {0,1 + 2 , −1 + 3 , 1 + 4 , … } , sup 𝐴1 = 1 + 2 , 𝑎𝑛𝑑 inf 𝐴1 = −1
1 1 1 1
𝐴1 = {1 + 2 , −1 + 3 , 1 + 4 , … } , sup 𝐴2 = 1 + 2 , 𝑎𝑛𝑑 inf 𝐴2 = −1

1 1 1 1
𝐴2𝑛 = {1 + 2𝑛 , −1 + 2𝑛+1 , 1 + 2𝑛+2 , … } , sup 𝐴2𝑛 = 1 + 2𝑛 , 𝑎𝑛𝑑 inf 𝐴2𝑛 =
−1
1 1 1
𝐴2𝑛+1 = {−1 + 2𝑛+1 , 1 + 2𝑛+2 , −1 + 2𝑛+3 , … } , sup 𝐴2𝑛+1 = 1 +
1
, 𝑎𝑛𝑑 inf 𝐴2𝑛+1 = −1
2𝑛+1
1
Now, sup 𝐴𝑛 = 1 + 𝑛 and inf 𝐴𝑛 = −1 ∀ 𝑛 ∈ ℕ
1
Hence, lim sup 𝑥𝑛 = 𝑙𝑖𝑚𝑛→∞ sup 𝐴𝑛 = 𝑙𝑖𝑚𝑛→∞ (1 + 𝑛) = 1 and
lim inf 𝑥𝑛 = 𝑙𝑖𝑚𝑛→∞ inf 𝐴𝑛 = 𝑙𝑖𝑚𝑛→∞ (−1) = −1

192
Definition:
Let 𝑥𝑛 be a sequence of real numbers.
1. We say that (𝑥𝑛 ) tends to −∞, and write 𝑙𝑖𝑚 𝑥𝑛 = +∞, if for
every 𝛼 ∈ ℝ there exists a natural number 𝐾(𝛼)such that if 𝑛 ≥
𝐾(𝛼), then 𝑥𝑛 > 𝛼.
2. We say that (𝑥𝑛 ) tends to +∞, and write 𝑙𝑖𝑚 𝑥𝑛 = −∞, if for
every 𝛽 ∈ ℝ there exists a natural number 𝐾(𝛽)such that if 𝑛 ≥
𝐾(𝛽), then 𝑥𝑛 < 𝛽.
We say that (𝑥𝑛 ) is properly divergent in case we have either
𝑙𝑖𝑚 𝑥𝑛 = +∞ or 𝑙𝑖𝑚 𝑥𝑛 = −∞.
Examples:
1. lim(𝑛) = +∞
2. lim(𝑛2 ) = +∞
3. If 𝑐 > 1 then lim(𝑐 𝑛 ) = +∞
Theorem:
A monotone sequence of real numbers is properly divergent if and only
if it is unbounded.
1. If (𝑥𝑛 ) is unbounded increasing sequence, then 𝑙𝑖𝑚 (𝑥𝑛 ) = +∞
2. If (𝑥𝑛 ) is unbounded decreasing sequence, then 𝑙𝑖𝑚 (𝑥𝑛 ) = −∞

Theorem:
Let (𝑥𝑛 ) and (𝑦𝑛 ) be two sequences of real numbers and suppose that
𝑥𝑛 ≤ 𝑦𝑛 for all 𝑛 ∈ ℕ.
a. If 𝑙𝑖𝑚(𝑥𝑛 ) = +1, then 𝑙𝑖𝑚(𝑦𝑛 ) = +1.
b. 2 If 𝑙𝑖𝑚(𝑦𝑛 ) = −1, then 𝑙𝑖𝑚(𝑥𝑛 ) = −1
Theorem:
Let (𝑥𝑛 ) and (𝑦𝑛 ) and be two sequences of positive real numbers and
𝑙𝑖𝑚 (𝑥𝑛)
suppose that for some 𝐿 ∈ ℝ, 𝐿 > 0, we have =𝐿
𝑛 → ∞ (𝑦𝑛)
Then 𝑙𝑖𝑚(𝑥𝑛 ) = +1 if and only if 𝑙𝑖𝑚(𝑦𝑛 ).

Definition:
A sequence of real numbers (𝑎𝑛 ) is said to be properly divergent to ∞ if
𝑙𝑖𝑚
𝑎 = ∞ that is ∀𝑀 ∈ ℝ there exists an 𝑁 ∈ ℕ such that if 𝑛 ≥ ℕ
𝑛→∞ 𝑛
then 𝑎𝑛 > 𝑀. Similarly, (𝑎𝑛 ) is said to be properly divergent to −∞ if
𝑙𝑖𝑚
𝑎 = −∞ that is ∀𝑀 ∈ ℝ there exists an 𝑁 ∈ ℕ such that if 𝑛 ≥ ℕ
𝑛→∞ 𝑛
then 𝑎𝑛 < 𝑀.
Theorem 1:
An increasing sequence of real numbers (𝑎𝑛 ) is properly divergent to ∞
if it is unbounded. A decreasing sequence of real numbers (𝑎𝑛 ) is
properly divergent to −∞ if it is unbounded.
Proof:

193
Suppose that (𝑎𝑛 ) is a sequence of real number that is increasing.
Since, (𝑎𝑛 ) is unbounded, then for any 𝑀 ∈ ℝ there exists a term 𝑎𝑀
(dependent on 𝑀) such that 𝑀 < 𝑎𝑀 . Since (𝑎𝑛 ) is an increasing
sequence, then for 𝑛 ≥ 𝑀 we have that 𝑀 < 𝑎𝑛 and since 𝑀 is arbitrary
we have that lim 𝑎𝑛 = ∞.
𝑛→∞
Similarly suppose that (𝑎𝑛 ) is a sequence of real number that is
decreasing. Since, (𝑎𝑛 ) is unbounded, then for any 𝑀 ∈ ℝ there exists
a term 𝑎𝑀 (dependent on 𝑀) such that 𝑀 > 𝑎𝑀 . Since (𝑎𝑛 ) is an
decreasing sequence, then for 𝑛 ≥ 𝑀 we have that 𝑀 < 𝑎𝑛 and since
𝑀 is arbitrary we have that lim 𝑎𝑛 = −∞.
𝑛→∞
Theorem 2:
Let (𝑎𝑛 ) and (𝑏𝑛 ) be sequences of real numbers such that 𝑎𝑛 ≤ 𝑏𝑛 for
all 𝑛 ∈ ℕ. Then if lim 𝑎𝑛 = ∞ then lim 𝑏𝑛 = ∞.
𝑛→∞ 𝑛→∞
Proof :
Let (𝑎𝑛 ) and (𝑏𝑛 )be sequences of real numbers such that 𝑎𝑛 ≤ 𝑏𝑛 for
all 𝑛 ∈ ℕ, and let lim 𝑎𝑛 = ∞. Then it follows that for all 𝑀 ∈ ℝ that
𝑛→∞
there exists an 𝑁(dependent on 𝑀 such that if 𝑛 ≥ ℕ then 𝑎𝑛 ≥ 𝑀). But
we have that 𝑏𝑛 ≥ 𝑎𝑛 for all 𝑛 ∈ ℕ and so for 𝑛 ≥ ℕ we have that
𝑏𝑛 ≥ 𝑀. Since 𝑀 is arbitrary it follows that lim 𝑏𝑛 = ∞.
𝑛→∞
Theorem 3:
Let (𝑎𝑛 ) and (𝑏𝑛 ) be sequences of real numbers such that 𝑎𝑛 ≤ 𝑏𝑛 for
all 𝑛 ∈ ℕ. Then if lim 𝑏𝑛 = −∞ then lim 𝑎𝑛 = ∞.
𝑛→∞ 𝑛→∞
Proof:
Let (𝑎𝑛 ) and (𝑏𝑛 )be sequences of real numbers such that 𝑎𝑛 ≤ 𝑏𝑛 for
all 𝑛 ∈ ℕ, and let lim 𝑏𝑛 = −∞. Then it follows that for all 𝑀 ∈ ℝ tht
𝑛→∞
there exists an 𝑁(dependent on 𝑀 such that if 𝑛 ≥ ℕ then 𝑏𝑛 ≤ 𝑀). But
we have that 𝑎𝑛 ≤ 𝑏𝑛 for all 𝑛 ∈ ℕ and so for 𝑛 ≥ ℕ we have that
𝑎𝑛 ≤ 𝑀. Since 𝑀 is arbitrary it follows that lim 𝑎𝑛 = ∞.
𝑛→∞
Theorem 4:
Let (𝑎𝑛 ) and (𝑏𝑛 ) be sequences of positive real numbers that for some
𝑎
real numbers 𝐿 > 0 that lim 𝑏𝑛 = 𝐿. Then lim 𝑎𝑛 = ∞ if and only if
𝑛→∞ 𝑛 𝑛→∞
lim 𝑏𝑛 = ∞.
𝑛→∞
Proof:
𝑎𝑛
Suppose (𝑎𝑛 ) and (𝑏𝑛 )are convergent sequences and that lim =𝐿
𝑛→∞ 𝑏𝑛
𝐿
for 𝐿 ∈ ℝ and 𝐿 > 0. Then for 𝜖 = 2 > 0 we have that for some 𝑛 ∈ ℕ if
𝑎 𝐿
𝑛 ≥ ℕ then |𝑏𝑛 − 𝐿| < 2 or equivalently
𝑛
𝐿 𝑎𝑛 3𝐿
< <
2 𝑏𝑛 2
𝐿 3𝐿
𝑏 < 𝑎𝑛 < 𝑏𝑛 .
2 𝑛 2

194
3𝐿
If lim 𝑎𝑛 = ∞ then since 𝑎𝑛 < 𝑏𝑛 it follows that lim 𝑏𝑛 = ∞. Similarly
𝑛→∞ 2 𝑛→∞
𝐿
if lim 𝑏𝑛 = ∞ then since 2 𝑏𝑛 < 𝑎𝑛 it follows that lim 𝑎𝑛 = ∞.
𝑛→∞ 𝑛→∞
Theorem 5:
If (𝑎𝑛 ) is a properly divergent subsequence then there exists no
convergent subsequences 𝑎𝑛𝑘 of 𝑎𝑛 .
Proof:
We will first deal with the case where (𝑎𝑛 )is properly divergent to ∞.
Suppose instead that there exists a subsequences 𝑎𝑛𝑘 that converges
to 𝐿. Then ∀𝜖 > 0 ∃𝐾 ∈ ℕ such that if 𝑘 ≥ 𝐾 then 𝑎𝑛𝑘 − 𝐿 < 𝜖, and so
for 𝑘 ≥ 𝐾 then 𝐿 − 𝜀 < 𝑎𝑛𝑘 < 𝐿 + 𝜖 and so 𝑎𝑛𝑘 < 𝐿.
Now if (𝑎𝑛 ) diverges to ∞ then for 𝐿 + 𝜖 ∈ ℝ ∃𝑁 ∈ ℕ such that if 𝑛 ≥ 𝑁
then 𝑎𝑛 > 𝐿 + 𝜖. So for 𝑛𝑘 ≥ 𝑚𝑎𝑥{𝐾, 𝑁},we have that
𝐿 + 𝜀 < 𝑎𝑛𝑘 < 𝐿 + 𝜖 which is a contradiction. So our assumption that
𝑎𝑛𝑘 converges was false, and so there exists no convergent
subsequences 𝑎𝑛𝑘 .
Definition 2:
Let (𝑠𝑛 ) be a sequence of real numbers. We say that (𝑠𝑛 ) diverges to
∞ and write lim 𝑠𝑛 = ∞ or 𝑠𝑛 → ∞ as 𝑛 → ∞ provided that for every
𝑛→∞
number 𝑀 there is an integer 𝑁 so that 𝑠𝑛 ≥ 𝑀 whenever 𝑛 ≥ 𝑁.
Example:
𝑛2 +1
Let us prove that → ∞ using the definition.
𝑛+1
If 𝑀 is any positive number we need to find some point in the sequence
after which all terms exceed 𝑀. Thus, we need to consider the
inequality
𝑛2 + 1
≥𝑀
𝑛+1
After some arithmetic we see that this is equivalent to
1 𝑛
𝑛+ − ≥𝑀
𝑛+1 𝑛+1
𝑛
Since 𝑛+1 < 1 we see that, as long as 𝑛 ≥ 𝑀 + 1 this will be true.
Thus, take any integer 𝑁 ≥ 𝑀 + 1 and it will be true that
𝑛2 + 1
≥𝑀
𝑛+1
for all 𝑛 ≥ 𝑁. (Any larger value of 𝑁 would work too.)
Definition:
Let (𝑥𝑛 ) be a sequence.

a. We say that (𝑥𝑛 ) tends to +∞ and write lim 𝑥𝑛 = +∞ if


∀𝛼 ∈ ℝ ∃𝐾(𝛼) ∈ ℕ ∋ ∀𝑛 ≥ 𝐾(𝛼), 𝑥𝑛 > 𝛼
b. We say that (𝑥𝑛 ) tends to −∞ and write lim 𝑥𝑛 = −∞ if
∀𝛽 ∈ ℝ ∃𝐾(𝛽) ∈ ℕ ∋ ∀𝑛 ≥ 𝐾(𝛽), 𝑥𝑛 < 𝛽

195
We say (𝑥𝑛 ) is properly divergent in either case.
Example:
For 𝐶 > 1, 𝑙𝑖𝑚𝑐 𝑛 = +∞
Proof:
Let 𝛼 ∈ ℝ be given. [How to express 𝐶 > 1.]
𝑐 = 1 + 𝑏 where 𝑏 > 0. By the Archimedian property,
𝛼
∃𝐾(𝛼) ∈ ℕ ∋ 𝐾(𝛼) > 𝑏 . Then ∀𝑛 ≥ 𝐾(𝛼),
𝑐 𝑛 = 1 + 𝑏 𝑛 ≥ 1 + 𝑛𝑏 > 1 + 𝛼 > 𝛼
Thus, 𝑙𝑖𝑚𝑐 𝑛 = +∞
Definition:
Let (𝑥𝑛 ) be a sequence in real numbers
i. We say that 𝑥𝑛 tends to +∞ and 𝑙𝑖𝑚𝑥𝑛 = +∞ if ∀𝛼 ∈ ℝ, ∃𝐾(𝛼) ∈
ℕ such that ∀𝑛 ≥ 𝐾, we have that 𝑥𝑛 > 𝛼.
ii. We say that 𝑥𝑛 tends to −∞ and 𝑙𝑖𝑚𝑥𝑛 = −∞ if ∀𝛽 ∈ ℝ, ∃𝐾(𝛽) ∈
ℕ such that ∀𝑛 ≥ 𝐾, we have that 𝑥𝑛 < 𝛽.
iii. We say that 𝑥𝑛 is properly divergent if either 𝑙𝑖𝑚𝑥𝑛 = +∞ or
𝑙𝑖𝑚𝑥𝑛 = −∞
Theorem:
A monotone sequence in ℝ is properly divergent if and only if it is
unbounded, and
a. If (𝑥𝑛 ) is unbounded and increasing sequence then 𝑙𝑖𝑚𝑥𝑛 = +∞
b. If (𝑥𝑛 ) is unbounded and decreasing sequence then 𝑙𝑖𝑚𝑥𝑛 = −∞
Proof
Suppose (𝑥𝑛 ) is monotone increasing sequence.
⟹ Suppose ( 𝑥𝑛 ) is bounded, then it must be convergent by MCT,
which is a contradiction.
Theorem:
Let 𝑥𝑛 ≤ 𝑦𝑛 ∀𝑛 ∈ ℕ.
a.If 𝑙𝑖𝑚 𝑥𝑛 = +∞ then 𝑙𝑖𝑚 𝑦𝑛 = +∞
b. If 𝑙𝑖𝑚 𝑥𝑛 = −∞ then 𝑙𝑖𝑚 𝑦𝑛 = −∞
Proof
(b) let 𝛽 ∈ ℝ, ∃𝐾(𝛽) ∈ ℕ such that ∀𝑛 ≥ 𝐾, then 𝑦𝑛 < 𝛽 ⟹ 𝑥𝑛 ≤ 𝑦𝑛 <
𝛽.
⟹ 𝑙𝑖𝑚 𝑥𝑛 = −∞
Theorem:
Let (𝑥𝑛 ) , (𝑦𝑛 ) be two sequences of positive real numbers and suppose
that
𝑥𝑛
𝑙𝑖𝑚 ( ) = 𝐿 > 0
𝑦𝑛
Then 𝑙𝑖𝑚𝑥𝑛 = +∞ if and only if 𝑙𝑖𝑚𝑦𝑛 = +∞
Proof:

196
𝑥
Since 𝑙𝑖𝑚 (𝑦𝑛 ) = 𝐿
𝑛
𝐿 𝐿
Let 𝜖 = 2 > 0, then ∃𝐾 , (2) ∈ ℕ such that ∀𝑁 ≥ 𝐾,
𝑥 𝐿
|𝑦𝑛 − 𝐿| < 2⟹
𝑛
𝐿 𝑥 𝐿
− 2 < 𝑦𝑛 − 𝐿 < 2⟹
𝑛
𝐿 𝑥𝑛 3𝐿
( )< <( )
2 𝑦𝑛 2
𝐿 3𝐿
⟹(2) 𝑦𝑛 < 𝑥𝑛 < ( 2 ) 𝑦𝑛
3𝐿
Now, if 𝑙𝑖𝑚𝑥𝑛 = +∞ ⟹ 𝑙𝑖𝑚 ( 2 ) 𝑦𝑛 = +∞
⟹𝑙𝑖𝑚𝑦𝑛 = +∞
𝐿
and if 𝑙𝑖𝑚𝑦𝑛 = +∞ ⟹ 𝑙𝑖𝑚 (2) 𝑦𝑛 = +∞
⟹ 𝑙𝑖𝑚𝑥𝑛 = +∞

197
Exercises
1. Prove that
a. lim 𝑛 = ∞
b. lim(−𝑛2 ) = −∞
c. If 𝑐 > 1, then 𝑙𝑖𝑚𝑐 𝑛 = +∞
2. Determine if the following series is convergent or divergent. If it
converges determine its value.
d. ∑∞𝒏=𝟏 𝒏
3. Determine if the following series converges or diverges. If it
converges determine its sum.
e. ∑∞𝒏=𝟎(−𝟏)
𝒏

4. Let 𝑏𝑟 = √𝑟 + 1 − √𝑟

198
Answers Key
1.
a. lim 𝑛 = ∞
Let 𝛼 ∈ ℝ be given. Then , be given. Then , ∃𝐾(𝛼) ∈ ℕ such
that 𝐾 > 𝛼. So ∀𝑛 ≥ 𝐾, we have 𝑥𝑛 = 𝑛 ≥ 𝐾 > 𝛼
b. lim(−𝑛2 ) = −∞
Let 𝛽 ∈ ℝ be given. Then , ∃𝐾(𝛽) > −𝛽 and so ∀𝑛 ≥ 𝐾, we
have
𝑛2 ≥ 𝑛 ≥ 𝐾 > −𝛽 ⟹ 𝑥𝑛 = −𝑛2 < 𝛽
c. If 𝑐 > 1, then 𝑙𝑖𝑚𝑐 𝑛 = +∞
Let 𝑐 = 1 + 𝑏, 𝑏 > 0 ⟹ 𝑐 𝑛 = (1 + 𝑏)𝑛 ≥ 1 + 𝑛𝑏 > 𝑛𝑏
(Bernoulli’s Inequaliity)
𝛼 𝛼
So, if 𝛼 ∈ ℝ is given . then ∃𝐾 (𝑏 ) such that 𝐾 > 𝑏 . So if 𝑛 ≥
𝐾, then we have
𝛼
𝑥𝑛 = 𝑐 𝑛 ≥ 1 + 𝑛𝑏 > 𝑛𝑏 > 𝐾𝑏 > 𝑏 · 𝑏 = 𝛼.
2. To determine if the series is convergent we first need to get our hands
on a formula for the general term in the sequence of partial sums
𝑛

𝑠𝑛 = ∑ 𝑖
𝑖=1
This is a known series and its value can be shown to be
𝑛
𝑛(𝑛 + 1)
𝑠𝑛 = ∑ 𝑖 =
2
𝑖=1
So, to determine if the series is convergent we will first need to see if
the sequence of partial sums,
𝑛(𝑛 + 1) ∞
{ }
2 𝑛=1
is convergent or divergent. That’s not terribly difficult in this case. The
limit of the sequence terms is,
𝑙𝑖𝑚 𝑛(𝑛 + 1)
=∞
𝑛→∞ 2
Therefore, the sequence of partial sums diverges to ∞ and so the
series also diverges.
3. In this case we really don’t need a general formula for the partial sums
to determine the convergence of this series. Let’s just write down the first
few partial sums.
𝑠0 =1
𝑠1 =1−1=0
𝑠2 =1−1+1=1
𝑠3 =1−1+1−1=0

So, it looks like the sequence of partial sums is,

199
{𝑠𝑛 }∞
𝑛=0 = {1,0,1,0, . . . }
𝑙𝑖𝑚
and this sequence diverges since 𝑠 doesn’t exist. Therefore,
𝑛→∞ 𝑛
the series also diverges.
4. 𝑏𝑟 = √𝑟 + 1 − √𝑟
√𝑟 + 1 + √𝑟
𝑏𝑟 = (√𝑟 + 1 − √𝑟) ( )
√𝑟 + 1 + √𝑟
𝑟+1−𝑟
𝑏𝑟 =
√𝑟 + 1 + √𝑟
1 1
𝑏𝑟 = ≥
√𝑟 + 1 + √𝑟 2√𝑟 + 1
∞ ∞
1 1
∑ 𝑏𝑟 ≥ ∑
2 √𝑟 + 1
𝑛=2 𝑟=1
∞ ∞
1 1
∑ 𝑏𝑟 ≥ ∑ , {𝑛 = 𝑟 + 1}
2 𝑛
𝑛=2 𝑟=1
1
𝑤ℎ𝑒𝑟𝑒 ∑∞
𝑟=2 is divergent
√𝑛

∑ 𝑏𝑟 is divergent
𝑛=2

∑(−1)𝑟−1 (√𝑟 + 1 − √𝑟) is divergent.


𝑛=1

200
INTRODUCTION TO INFINITE SERIES

JOVIE AGUE

201
Introduction to Infinite Series

Definition:
A series or infinite series is the sum of all the terms in a sequence.

Example:
(Examples of infinite series).
1. ∑∞
𝑛=1 𝑛 = 1 + 2 + 3 + ⋯

1 1 1
2. A geometric series ∑∞
𝑛=1 2𝑛=1+2 + 2 + ...

Every term in this series is obtained from the previous one


1
by multiplying by the common ratio . This is what geometric means.
2

1 1 1
3. The harmonic series ∑∞
𝑛=1 𝑛=1+ 2 + 3 + ...

4. An alternating series ∑∞
𝑛=0(−1)𝑛=1+(−1)+1+(−1)+ ...

Notes

1. For now these infinite sums are just formal expressions or arrangements
of symbols. Whether it is meaningful to think of them as numbers or not is
something that can be investigated.

2. A series is not the same thing as a sequence and it is important not to


confuse these terms. A sequence is just a list of numbers. A series is an
infinite sum.
3. The “sigma” notation for sums : sigma (lower case σ, upper case Σ) is a
letter from the Greek alphabet, the upper case Σ is used to denote sums.
The notation ∑𝑖𝑛=𝑖 𝑎𝑛 means: i and j are integers and i ≤ j. For each n from
i to j the number an is defined; the expression above means the sum of the
numbers an where n runs through all the values from i to j, i.e.

∑ 𝑎𝑛 = 𝑎𝑖 + 𝑎𝑖 + 1 + 𝑎𝑖2 + ⋯ + 𝑎𝑗 − 1 + 𝑎𝑗.
𝑛=1

For example

∑5𝑛=2 𝑛2 = 22 + 32 + 42 + 52 = 54.
For infinite sums it is possible to have −∞ and/or ∞ (instead of fixed integers i
and j) as subscripts and superscripts for the summation.

What does it mean to talk about the sum of infinitely many numbers? We
cannot add infinitely many numbers together in practice, although we can (in
principle) at least, add up any finite collection of numbers. In the examples
above we can start from the beginning, adding terms at the start of the series.
Adding term by term we get the following lists.

202
1. ∑∞
𝑛=1 𝑛 = 1 + 2 + 3 + ...

1, 1 + 2, 1 + 2 + 3, 1 + 2 + 3 + 4, 1 + 2 + 3 + 4 + 5, · · · : 1, 3, 6, 10,
15, . . .

Since the terms being added on at each stage are getting bigger, the
numbers in the list above will keep growing (faster and faster as n
increases) - we can’t associate a numberical value with this infinite
sum.

2. A geometric series
1 1 1
∑∞
𝑛=1 = 1 + 2 + 22 + ⋯
2𝑛

1 1 1 1 1 1
1,1+2 , 1 + 2 + 22 , 1 + 2 + 22 + 23 …

These numbers are


3 7 15 31 63
1,1+2 , 4 , , 16 , 32 …
8

1
In this example the terms that are being added on at each step ( 2𝑛 )
are getting smaller and smaller as n increases, and the numbers in the
list appear to be converging to 2.

3. The harmonic series


1 1 1
∑∞
𝑛=1 𝑛 = 1 + 2 + 3 + ⋯

1 1 1 1 1 1 3 11 25 137
1, 1 + 2 , 1 +2 + 3 , 1 + 2 +3 + 4 . . . : 1, 2, , 12, ,...
6 60

It is harder to see what is going on here.

4. An alternating series

∑∞
𝑛=0(−1) n = 1 + (−1) + 1 + (−1) + . . .

1, 1 − 1, 1 − 1 + 1, 1 − 1 + 1 − 1, 1 − 1 + 1 − 1 + 1 . . . : 1, 0, 1, 0, 1, .
.
The terms being “added on” at each step are alternating between 1 and
−1, and as we proceed with the summation the “running total”
alternates between 0 and 1. So there is no numerical value that we can
associate with the infinite sum ∑∞
𝑛=0(−1)n.

Note

203
he series in 2. above converges to 2, the series in 1. and 4. are both
divergent and it is not obvious yet but the series in 3. is divergent as well. Our
next task is to give precise meanings to these terms for series. In order to do
this we need some terminology. Bear in mind that we know what it means for
a sequence to converge, but we don’t yet have a definition of convergence for
series.

Definition:
For a series ∑∞
𝑛=1 an, and for k ≥ 1, let

sk =∑𝑘𝑛=1 an = a1 + a2 + a3 + · · · + ak.

Thus s1 = a1, s2 = a1 + a2, s3 = a1 + a2 + a3 etc.


Then sk is called the kth partial sum of the series, and the sequence
(sk)∞ k=1 is called the sequence of partial sums of the series.

If the sequence of partial sums converges to a limit s, the series is said


to converge and s is called its sum. In this situation we can write

∑∞
𝑛=1 𝑎𝑛 = 𝑠.

If the sequence of partial sums diverges, the series is said to diverge.

Example:
(Convergence of a geometric series). Recall the second example
above :

1 1 1
∑∞
𝑛=0 = 1 + 4 + 22 + ⋯
2𝑛

In this example, for k ≥ 0,

1 1 1 1
Sk=∑𝑘𝑛=0 2 = 1 + 2 + 4 + ⋯ 2𝑘

1 1 1 1 1 1
𝑠𝑘 = ∑𝑘𝑛=1 2𝑛+1 = + 4 + ⋯ 2𝑘 + 2𝑘+1
2 2
Then
1 1 1 1
Sk-2 𝑠𝑘 = 2 𝑠𝑘 = 1 − 2𝑘+1 ⇒ 𝑠𝑘 = 2 − 2𝑘.

1
So the sequence of partial sums has kth term 2 −2. This sequence
converges to 2 so the series converges to 2; we can write
1
∑𝑛 = 0∞ = 2.
2𝑘

204
General geometric series :
Consider the sequence of partial sums for the geometric
series

∑∞
𝑛=0 𝑎𝑟𝑛 = 𝑎 + 𝑎𝑟 + 𝑎𝑟2 + ⋯

(This is a geometric series with initial term a and common


ratio r.) The kth partial sum sk is given
By

Sk=∑𝑘𝑛=0 𝑎𝑟𝑛 = 𝑎 + 𝑎𝑟 + ⋯ 𝑎𝑟𝑘

rsk=∑𝑘𝑛=0 𝑎𝑟𝑛 + 1 = ar+ar2+ …ark+1

𝑎(1−𝑟𝑘+1)
Then (1 − r)sk = a − ark+1 ⇒ sk = 1−𝑟 . If |r| < 1, then
rk+1 → 0 as k → ∞, and the sequence of partial sums
𝑎
(hence the series) converges to 1−𝑟. If |r| ≥ 1 the series is
divergent.
Next we show that the harmonic series is divergent.

Theorem:
1
The harmonic series ∑∞
𝑛=1 𝑛 is divergent. We give two proofs.

Proof 1:
1 1
Think of 𝑛 as the area of a rectangle of height 𝑛 and width 1,
sitting on the interval [n, n + 1] on the x-axis. So the first
1
term 1 corresponds to
a square of area 1 sitting on the interval[1, 2], the term 12
1
corresponds to a rectangle of area 2 sitting on the interval [2, 3] and so
on, as in the following picture.The total area accounted for by these
triangles is the sum of the harmonic series, and this exceeds the area
accounted for by the improper integral
∞1
∫1 𝑋
dx.

From Section 1.5 we know that this area is infinite, hence the series is
divergent.

Proof 2:

205
We show that the sequence of partial sums of the harmonic
series is not bounded above.

• The first term is 1.


1
• The second term is 2
1
• The sum of the 3rd and 4th terms exceeds 2.

1 1 1 1 1 1
+ 4 > 4+4 + 4 = 2.
3

1
• The sum of the 5th, 6th, 7th and 8th terms exceeds2:

1 1 1 1 1 1 1 1 1 1
+ 6 + 7 + 8 > 8 + 8 + 8 + 8 + 8 = 2.Type equation here.
5

the same reason, the sum of the next 8 terms (terms 9 through 16) also
1
exceeds 2

1 1 1
• In general the sum of the 2𝑛−1 terms 𝑛−1 +1 𝑡ℎ𝑟𝑜𝑢𝑔ℎ exceeds2.
2 2𝑛
12n exceeds 12
.
So, as we list terms in the sequence of partial sums of the harmonic
series, we keep accumulating non-overlapping stretches of terms that
1
add up to more than 2. Thus the entire series has infinitely many non-
1
overlapping stretches all individually summing to more than 2. Then the
sum of this series is not finite and the series diverges.
Note:
A necessary condition for the series
∑∞𝑛−1 𝑎𝑛 an to converge is that the sequence {𝑎𝑛 }∞𝑛=1 converges to 0;
i.e. that 𝑎𝑛 → 0 as n → ∞. If this does not happen, then the sequence
of partial sums
has no possibility of converging.
The example of the harmonic series shows that the condition 𝑎𝑛 → 0 as
n → ∞ is not sufficient to guarantee that the series ∑∞
𝑛=1 will
converge.

LEARNING OUTCOMES FOR SECTION

After studying this section you should be able to

• explain what an infinite series is and what it means for an infinite


series to converge;
• Give examples of convergent and divergent series;
• show that the harmonic series is divergent;
• Use the “sigma” notation for sums.

206
Many of our in_nite sequences, for the remainder of the course, will be
de_ned by sums.
For example, the sequence
Sm :=
Xm
n=112n : (1) is defined by a sum. Its terms (partial sums) are 1 2; 1 2
+14=34;12+14+18=78;12+14+18+116=1516;:::
These in_nite sequences de_ned by sums are called in_nite series.
Review of sigma notation
The Greek letter _ used in this notation indicates that we are adding
(\summing") elements of a
certain pattern. (We used this notation back in Calculus 1, when we
_rst looked at integrals.) Here our
sums may be \in_nite"; when this occurs, we are really looking at a
limit.
Resources
An introduction to sequences a standard part of single variable
calculus. It is covered in every calculus
textbook. For example, one might look at
* section 11.3 (Integral test), 11.4, (Comparison tests) , 11.5 (Ratio &
Root tests), 11.6 (Alternating,
abs. conv & cond. conv) in Calculus, Early Transcendentals (11th ed.,
2006) by Thomas, Weir,
Hass, Giordano (Pearson)
* section 11.3 (Integral test), 11.4, (comparison tests), 11.5 (alternating
series), 11.6, (Absolute conv,
ratio and root), 11.7 (summary) in Calculus, Early Transcendentals (6th
ed., 2008) by Stewart
(Cengage)
* sections 8.3 (Integral), 8.4 (Comparison), 8.5 (alternating), 8.6,
Absolute conv, ratio and root, in
Calculus, Early Transcendentals (1st ed., 2011) by Tan (Cengage)
Integral tests, comparison tests,
ratio & root tests.
* section 9.4 (Convergence Tests), 9.5 (Comparison, ratio, root tests),
9.6 (Alternating, abs. conv &
cond. conv) Calculus, Early Transcendentals (11th ed., 2009) by
Anton, Bivens, Davis (John Wiley
& Sons) p. 645 of Anton has a nice list.
* section 10.3 (integral test), 10.4 (alt series), 10.5 (comparison), 10.6
(absolute convergence), 10.7 (ratio and root) in the Whitman College
online textbook: http://www.whitman.edu/mathematics/multivariable/*
Whitman's online textbook:
http://www.whitman.edu/mathematics/multivariable/calculus_
10_Sequences_and_Series.pdf
1 3.2.1 What is a series?
Given an infinite series 1X n=1 an we de_ne the partial sum Sm :=Xm
n=1 an: Thus, in the series 1X n=1 12n we have the partial sums
S1 =1 2
S2 =3 4

207
S3 =7 8
S4 =15 16
We mean, by the expression, 1X n=1, the limit, as n ! 1, of the partial
sums Sn:
In this case, the partial sums appear to have the pattern Sn = 1 􀀀12n :
So 1X n=1 1 2n really means limm!1 Xmn=112n = limm!11 􀀀12m =
1:Since, in this case, the limit is 1, we say that1Xn=112n = 1:3.2.2 An
easy divergence test
Intuitively, if a series is to converge to a _nite limit L we would expect
that eventually the terms we are
adding up are contributing very little to the series { that at some point
the sum is close to L and each
new term is not changing that. This argument can be made precise (but
we won't do that here.)
This gives us a theorem, the \n-th term divergence test":
Theorem.
(n-th term test) If a series X
an converges then the limit, as n goes to in_nity, of the terms an, must
be zero.
Although stated in terms of convergence, the theorem is really a
statement about divergence, for it is equivalent to the statement:
The n-th term divergence test: If in a series X the terms an do not go to
zero then the series does not converge! (This statement is the
contrapositive of the statement in the theorem.)

3.2.3 Geometric series


The complexity of our investigation into series (initiated long ago by the
Bernoulli brothers and Euler) can be displayed by examining the family
of geometric series { central to our understanding of series {and two
interesting \sporadic" series, the harmonic series and the alternating
harmonic series.
2
In an earlier section, we examined the series 1Xn=11 2n = 2+14+18+1
16+ : : : 12n + : : : :
We concluded, just by observation, that the partial sums had the form
1􀀀 1
2n and thus the series converged to 1. Note that this series has this
property: each term added on is exactly 12 of the previous term.A
series which has a \ratio" r such that each new term is exactly r times
the previous term, is said
to be geometric. (See http://en.wikipedia.org/wiki/Geometric_series for
a general discussion of these series, including modern applications.)
The main idea.
There is a nice way to work out a formula for the partial sum of a
geometric series. In general, a geometric series has form a + ar + ar2 +
ar3 + : : : + arn􀀀1 + : : : = 1X n=0 arnwhere a is the _rst term and r is
the common ratio between terms.
Let us write
Sm := a + ar + ar2 + ar3 + : : : + arm􀀀1 = mX􀀀1 n=0 arn:

208
Notice that rSm := ar + ar2 + ar3 + : : : + arm􀀀1 + arm = Xm n=1 arn
and so rSm 􀀀 Sm := arm 􀀀 a: (Notice how most terms cancel!)
Therefore Sm = a(rm 􀀀 1) r 􀀀 1 = a(1 􀀀 rm) 1 􀀀 r :
If jrj > 1 then the expression a(1 􀀀 rm) 1 􀀀 r does not converge and so
the geometric series does not converge.
If r = 1 then the expression a(1􀀀rm) 1􀀀r is unde_ned but it is easy to
check that the partial sums are
Sm = ma and so the series diverges to in_nity.
If r = 􀀀1 then the a(1􀀀rm)
1􀀀r does not converge and so the geometric series does not converge.
The partial sums alternate between a and 0:
But if jrj < 1 then the a(1􀀀rm) 1􀀀r converges to a 1􀀀r : Therefore the
geometric series converges to a 1􀀀r :
This is important enough to emphasize as a theorem.
Theorem.
(Geometric series)

If jrj < 1 the a + ar + ar2 + ar3 + : : : + arn􀀀1 + : : : = 1X n=0 arn = a 1 r:


But if jrj _ 1 then 1X n=0 arn diverges.3
The geometric series are central to the study of in_nite series. We will
see later in this course that if a
series is not geometric, we will attempt (in a certain way) to \pretend" it
is geometric anyway! Sometimes
this \pretense" gives us very useful information. (This will motivate the
\ratio" and \root" tests.)
3.2.4 The Harmonic Series
The n-th term divergence test says that the terms of a series must go
to zero if there is any hope of the
series converging.
Warning! It is tempting to believe in the converse statement. Is it true
that if the terms go to zero
then the series converges? That would be nice { but it is not true in
general.
Here is an example { a classical one { where the terms go to zero but
the series diverge. The series is called the \harmonic series". The
harmonic series 1Xn1n= 1 +12+13+14+ : : : +1n+ : : : :
(See http://en.wikipedia.org/wiki/Harmonic_series_(mathematics) for
Wikipedia's discussion of
this series, including an explanation for its name.)
The terms 1n go to zero as n goes to in_nity. Yet this series diverges!
We give one argument that this series diverges. This argument is a
\comparison test". Another
argument will be given later.
Compare the series 1Xn1n= 1 +12+13+14+15+16+17+18+19+110+
::n+ : : : and the series 1 + 1 2 + (14+14) + (18+
18+18+18 + (116+116+ : : :116) + : : :In the second series, any
expression of the form 1n , 2s < n _ 2s+1, we replace n by 2s+1: Since
12s+1_1n , we have a series which is \smaller" than the harmonic
series. Each partial sum of the second
series no bigger than the partial sum of the harmonic series.

209
But notice that, since in the second series there are 2s terms of the
form 12s+1 then we can collect them together to form the term12:So
the second series becomes1+12+(14+14)+(18+18+18+18)+(116+116+
: : :116)| {z }8 terms+: : : = 1+12+(12)+(12)+(12)+: : :+(12)+: : :
We continue, forever, to add 12 in the sum. So this second series
diverges (by the \n-th term test”, if you will.) But since the harmonic
series is forever larger than this diverging series, then the harmonic
series diverges! The harmonic series forms a nice counterpoint to the
n-th term test for divergence. If the n-th term of a series goes to zero,
the series still might not converge. The harmonic series is a nice
example of thisphenomenon.43.2.5 The Alternating Harmonic Series
One more interesting series: Consider the alternating harmonic series
1Xn(􀀀1)n+1n= 1 􀀀12+􀀀14+15+ : : : +(􀀀1)n+1n+ : : : :
Note the elect of the expression (􀀀1)n+1; it forces the signs to
alternate, so that we are adding positive, then negative terms.
It turns out that an alternating series, where the n-th term does to zero,
does converge!
The alternating harmonic series converges { indeed, we will see later
that its sum is ln 2 _ 0:693147180559945.
3.2.6 Telescoping series
There is another type of geometric series we run into from time to time,
where a little trick will give not
only convergence, but the exact value of the limit of the series.
Consider, for example, the series1X11n (n + 1)=12+16+112+ : : : ::::
We can use the algebraic concept of partial fractions (remember that?!)
to rewrite 1n(n + 1)=1n􀀀1n + 1: So1X11n(n + 1)=1X 1 1n􀀀1n + 1=
(11􀀀12) + (12􀀀13) + (13􀀀14) + : : : ::::
Notice how the terms begin to cancel! We can rewrite the sum
as=1+(􀀀12+12) + (􀀀13+13) + (􀀀14+14) + : : : ::::In this case, the series
collapses (\telescopes") and only the _rst term survives. So1X11n(n +
1)= 1:5

210
EXERCISES
𝒆𝒏
1. 𝒂𝒏 = 𝟑𝒏
2. 𝒂𝒏 = (−𝟏)𝒏 √𝒏
𝟏
3. 𝒂𝒏 = (−𝟏)𝒏
√𝒏
4. 𝒂𝒏 = (−𝟏)𝟐𝒏+𝟏
(−𝟏)𝒏 +𝒏
5. 𝒂𝒏 = (−𝟏)𝒏 −𝒏
𝒏
𝒍𝒏((𝒆𝟒 ) )
6. 𝒂𝒏 = 𝟑𝒏
𝝅
7. 𝒂𝒏 = (−𝟏)𝒏 𝒄𝒐𝒔( 𝟐 (𝒏 + 𝟏))
𝝅
8. (−𝟏)𝒏 𝒄𝒐𝒔( 𝟐 (𝟐𝒏 − 𝟏))

211
ANSWERS KEY

1. Converges to 0
2. Diverges
3. Converges to 0
4. Converges to −1
5. Converges to −1
4
6. Converges to3
7. Diverges
8. Converges to 0

212
LIMITS OF A FUNCTION

BRIX LAPIAD

213
Limits found Graphically

Definition: A limit is the idea of looking at what happens to a function as you


approach particular values of x. Left-hand and right-hand limits are
the idea of looking at what happens to a function as you approach a
particular value of x, from a particular direction.
The limit of f(x) as x approaches the value of a from the left is written

lim f ( x)
x a

and the limit of f(x) as x approaches the value of a from the right is
written

lim f ( x)
x a

Let’s explore these ideas with the graph of f(x) in Figure 3.1 below.

Figure 3.1

Looking at f(x) when x = -2, you notice there is a “break” in the function.
However, if you approach x = -2 “from the left” (Figure 3.2a) you can see
that the function values are getting closer and closer to 1. On the other
hand, if we approach x = -2 “from the right” (Figure 3.2b) you can see
that the function values are getting closer and closer to 3.

214
Figure 3.2a Figure 3.2b

Looking at f(x) when x = 1, you notice there is a hole in the function. If


we approach f(x) from the left or from the right (Figure 3.3), you can see
that the function values are getting closer and closer to 2.

Figure 3.3

Therefore, the following statements are true.

lim f ( x)  1 lim f ( x)  3 lim f ( x)  2


x 2 x 2 x 1

lim f ( x)  2
x 1

215
Example 1
Using the given graph of g(x), find the following left- and right-hand
limits.

a. lim g ( x)
x  0
b. lim g ( x)
x  0
c. lim g ( x)
x 1
d. lim g ( x)
x 1

Solution

a. This asks us to look at the graph of g(x) as x approaches 0 from the left. You
can see that the function values are getting closer and closer to -1. So,

lim g ( x)  1
x  0

b. This asks us to look at the graph of g(x) as x approaches 0 from the right.
You can see that the function values are getting closer and closer to -1. So,

lim g ( x)  1
x0

216
c. This asks us to look at the graph of g(x) as x approaches 1 from the left. You
can see that the function values are getting closer and closer to -2. So,

lim g ( x)  2
x 1

d. This asks us to look at the graph of g(x) as x approaches 1 from the right.
You can see that the function values are getting closer and closer to -2. So,

lim g ( x)  2
x 1

Notice that in the solutions to parts (c) and (d) above, the function value
g(1)=1 does not play a role in determining the values of the limits. A limit
is strictly the behavior of a function “near” a point.

217
Example 2
Using the graph of h(x) below, find the following left- and right-hand
limits.

a. lim h( x)
x  4

b. lim h( x)
x  4

Solution
a. Looking at the graph of h(x), as x approaches 4 from the left, you can see that
the function values keep getting more and more negative, without end. Thus,
we say that the function values approach negative infinity, written

lim h( x)  
x  4

b. Looking at the graph of h(x), as x approaches 4 from the right, you can see
that the function values keep getting more and more positive without end.
Thus, we say that the function values approach positive infinity, written

lim h( x)  
x  4

By considering both the left- and right-hand limits of a function as you


approach a particular value of x, you can determine whether or not the
limit of the function at that point exists.

Definition of a Limit at a Point:

If lim f ( x)  L and lim f ( x)  L, then lim f ( x)  L.


x a  x a  x a

Therefore, if the left-hand limit does not equal the right-hand limit as x
approaches a, then the limit as x approaches a does not exist.

218
Example 3
Using the graph of f(x) below, find the following limits.

a. lim f ( x)
x 1

b. lim f ( x)
x 2

Solution

a. In previous investigations of this function, we found that

lim f ( x)  2 and lim f ( x)  2 . Therefore, by definition, since


x 1 x 1

lim f ( x)  lim f ( x)  2 , then lim f ( x)  2 .


x 1 x 1 x 1

It is important to notice that this limit exists even though f(1) does not
exist.

b. In previous investigations of this function, we found that

lim f ( x)  1 and lim f ( x)  3 . Therefore, by definition, since


x 2 x 2

lim f ( x)  lim f ( x) , then lim f ( x) does not exist (DNE).


x 2 x 2 x 2

Example 4

219
Using the given graph of g(x), find lim g ( x) .
x 1

Solution:
In previous investigations of this function, we found that

lim g ( x)  2 and lim g ( x)  2 . Therefore, by definition, since


x1 x1

lim g ( x)  lim g ( x)  2 , then lim g ( x)  2 .


x 1 x 1 x 1

Example 5
Using the given graph of h(x), find lim h( x) .
x4

220
Solution
In previous investigations of this function, we found that

lim h( x)   and lim h( x)   . Therefore, by definition, since


x 4 x 4

lim h( x)  lim h( x) , then lim h( x ) DNE.


x 4 x 4 x4

So far we have been focusing on what is happening with functions at


particular values of x by looking at what is happening to the function
values corresponding to values very near to the x value. Let’s now
explore what happens to the function values when we allow x to
approach positive and negative infinity.

Figure 3.4
Figure 3.4a Figure 3.4b Figure 3.4c

In Figure 3.4a, you can see that if you move to the right on the graph
and allow x to continually become larger (approach infinity), the function
values also become larger and larger. If you move to the left on the
graph and allow x to become more and more negative (approach
negative infinity), you can see that the function values are again
becoming larger and larger. Thus, we have

lim k ( x)   and lim k ( x)  


x  x 

In Figure 3.4b, you can see that if you allow x to approach infinity, the
function values go towards negative infinity. If you allow x to approach
negative infinity, you can see that the function values go towards positive
infinity. Thus, we have

221
lim m( x)   and lim m( x)  
x  x 

In Figure 3.4c, you can see that if you allow x to approach either positive
or negative infinity, the function values approach zero. Thus, we have

lim p ( x)  0 and lim p( x)  0


x  x 

When a function approaches a numerical value, say L, as x  or as


x  , we say that the function has a horizontal asymptote at y = L.
Thus, we have just found that p(x) has a horizontal asymptote at y=0,
while k(x) and m(x) have no horizontal asymptotes.
(Note: You can only approach positive infinity from the left and you can
only approach negative infinity from the right so there is no discussion
of left- and right-hand limits at infinity.)
Example 6
Using the graph of f(x) below, find the following limits.
a. lim f ( x)
x 5
b. lim f ( x )
x 
c. lim f ( x)
x 0
d. lim f ( x)
x 

Solution
a. We need to find and compare the left- and right-hand limits of f(x) at x= -5. As
x approaches -5 from the left, f(x) approaches 1 and as x approaches -5 from
the right, f(x) also approaches 1. Therefore,

lim f ( x)  1
x 5

b. As x  , the function values get more and more positive without end.
Therefore,
lim f ( x)  
x 

c. As x approaches zero from the left, f(x) approaches 5. Therefore,

222
lim f ( x)  5
x 0

d. As x  , the function values get closer and closer to zero. Therefore


lim f ( x)  0
x 

Limits found Numerically and Algebraically

While almost all limits can be found graphically, as we have been


discussing, it is not always practical or necessary if the function is
defined algebraically.

x2  9
For instance, say we are given that f ( x)  . If we are looking for
x 3
lim f ( x) , instead of having to graphically search for the answer, we can
x 3

find both the left- and right-hand limits by using tables. By choosing x
values that get closer and closer to x = 3 from both sides, we can analyze
the behavior of f(x).

Table 3.1

Limit from the left Limit from the right


   

x 2.99 2.999 2.9999 3 3.0001 3.001 3.01

f(x) 5.99 5.999 5.9999 ? 6.0001 6.001 6.01

Notice that when we chose values on either side of x = 3, they were


values that were very close to x = 3. It seems that as x approaches 3
from either side, the function values are approaching 6. Therefore, it
seems reasonable to conclude that

x2  9
lim 6
x 3 x  3

We can check this conclusion by looking at the graph of f(x) near x = 3,


as shown below.

223
Example 7

 4, x  1
Using tables, find the following limits given that f ( x)   2
x , x  1

a. lim f ( x)
x2

b. lim f ( x)
x 1

Solution

a. We need to construct a table with x-values approaching 2 from both sides.


Since all of these x-values are in the domain of x >1, we will use the part of
the function defined by x 2 to determine the function values in our table.

Limit from the left Limit from the right


   

x 1.99 1.999 1.9999 2 2.0001 2.001 2.01

f(x) 3.96 3.996 3.9996 ? 4.0004 4.004 4.04

Approaching x=2 from both the left and the right sides shows that the function
values are approaching 4. Thus, lim f ( x)  4 .
x2

b. We need to construct a table with x-values approaching 1 from both sides. All
x-values approaching x = 1 from the left are in the domain x < 1, so we will be
using the part of the function defined by 4 when finding these function values.
All x-values approaching x = 1 from the right are in the domain x > 1, so we
will use x 2 to find these function values in our table.

Limit from the left Limit from the right


   

x 0.99 0.999 0.9999 1 1.0001 1.001 1.01

f(x) 4 4 4 ? 1.0002 1.002 1.0201

224
As x 1 from the left, f(x) seems to be approaching 4, while as
x 1 from the right, f(x) seems to be approaching 1. Since these are
not equal, by definition, lim f ( x) does not exist.
x 1

Making tables can still be as time-consuming as graphing, so we will use


the following rules to algebraically evaluate limits more efficiently. Most
of these rules can intuitively be verified from looking at the previously
worked examples.

Table 3.2 – Limit Rules

If a, c, and n, are real numbers, then

1) lim c  c (The limit of a constant real number is that number.)


xa

2) lim p( x)  p(a) where p(x) is any polynomial (The limit value of a


x a

polynomial is the function value at that point.)


3) lim c  f ( x)  c  lim f ( x)
x a x a

(The limit of the product of a constant and a function equals the


constant times the limit of the function.)

4) lim  f ( x)  g ( x)  lim f ( x)  lim g ( x)


x a x a x a

(The limit of the sum or difference of two functions equals the sum
or difference of the limits of the functions.)

5) lim  f ( x)  g ( x)  lim f ( x)   lim g ( x) 


xa  xa   xa 
(The limit of the product of two functions is the product
of the limits of the functions.)

f ( x) lim f ( x)
6) lim  x a , if lim g ( x)  0
x a g ( x) lim g ( x) x a
x a

(The limit of a quotient is the quotient of the limits of the numerator and
denominator if the limit of the denominator is not zero.)

225
n
7) lim  f ( x)  lim f ( x)  (provided this is defined)
n

x a  xa 
(The limit of a function raised to a power equals the limit of the function
raised to the power provided the math makes sense.)

Example 8

Evaluate

a. lim x2  2 x  4
x 1

x2  9
b. lim
x 3 x  3

1
c. lim
x 4 x4
Solution

a. lim x2  2 x  4  (1)2  2(1)  4  5 (Rule 2)


x 1

x 2  9 lim x2  9
b. lim  x 3
(Rule 6)
x 3 x  3 lim x  3
x 3

0
 (Rule 2)
0

When you get 0/0 you have what is called an indeterminant form and you
must try other techniques to determine the limit. In this case, factor both the
numerator and denominator and cancel common terms to remove the zero in
the denominator. Then, apply the limit rules to the simplified expression.

x2  9 ( x  3)( x  3)
lim  lim (Factor)
x 3 x  3 x 3 ( x  3)

 lim x  3 (Cancel common terms)


x 3

=3+3=6 (Rule 2)

1 lim1
c. lim  x 4 (Rule 6)
x 4 x  4 lim x  4
x 4

1
 (Rules 1 and 2)
0

226
This is not defined and whenever you get a result of a non-zero number over
zero, there are no common factors in the numerator and denominator which
can be cancelled. Therefore, there is no way to rid the denominator of its zero
term, meaning that the limit does not exist. Looking at the graph of the
function near

x = 4 , we can see what is happening.

GRAPH with MAPLE

Notice this is the same function we analyzed when finding limits graphically.
There, we also found that the limit does not exist.

While these rules also apply when looking for limits at infinity (or negative
infinity), it is almost always necessary to algebraically manipulate the
expression of the function to determine the limit.

Example 9

2x2  7
Evaluate lim
x  6  3 x 2

Solution

2 x 2  7 lim 2 x2  7 
lim  x 

x  6  3 x 2 lim 6  3x 2 
x 

  
or or are all also known as indeterminant forms. When this form
  
occurs when finding limits at infinity (or negative infinity) with rational functions,
divide every term in the numerator and denominator by the highest power of x
in the denominator to determine the limit.

Since x 2 is the highest power of x in the denominator of our function, we have

227
2 x 2  7 lim  2 x2  7   x2
lim  x 
x  6  3 x 2 lim  6  3x 2   x 2
x 

 7
lim  2  2 

x 
 x 
 6 
lim  2  3 

x  x

2  0 2
 
03 3

Continuity

Definition:

A function f(x) is continuous at x = a, if all of the following are true:

1. f(a) is defined (A function value exists at x=a.)


2. lim f ( x) exists (A limit value exists as you approach x=a.)
xa

3. lim f ( x)  f (a) (The function value equals the limit value at


x a

x=a.)
Graphically, this means that a function is continuous wherever the graph
of the function has no holes, gaps, or jumps. A function is said to be
discontinuous at x = a, if a hole, gap, or break occurs in the graph at x =
a, meaning the function violates one of the three items above.

Notice that the third item in the definition of continuity and Rule 2 of the
Limit Rules show that all polynomial functions are continuous for all real
values of x.

Example 10

Using the graph of f(x) below, find all values of x where f(x) is
discontinuous and state why f(x) is discontinuous at these points,
according to the definition of continuity.

228
Solution

 x = -3, f(-3) is undefined


 x = -2, f(-2) is undefined
 x = 1, while the function is defined by f(1)=5 and lim f ( x)  1 , these are not
x 1

equal and thus the third item of the definition is violated


 x = 4, lim f ( x) does not exist
x4

Example 11

x4
Is f ( x)  continuous at x =1? At x =3?
x 3

Solution

For both values of x, we must check each of the three items in the
definition of continuity. If one of the items fails, f(x) not continuous at
that particular x-value.

x = 1:

1 4 5
First, check to see if a function value exists: f (1)  
1  3 2

229
x  4 lim x4 5
Next, check to see if a limit value exists: lim  x1 
x 1 x  3 lim x  3 2
x 1

Last, check that the function value equals the limit value, which it does
in this case. Therefore, f(x) is continuous at x =1.

x = 3:

3 4 7
First, check to see if a function value exists: f (3)   is
33 0
undefined

Therefore, since the first item of the definition is violated, f(x) is


discontinuous at x = 3.

As previously stated, polynomial functions are continuous for all real


values of x. This is also true for exponential functions. Moreover, it is
true that rational functions are continuous for all real values of x that do
not make the denominator zero and logarithmic functions are
continuous for all x-values in their domains.

230
Exercises

Use the given graph of f(x) to answer Exercises 3.1-3.4.

Exercise 3.1 Evaluate lim f ( x)


x 1

Exercise 3.2 Evaluate lim f ( x)


x 

Exercise 3.3 Evaluate lim f ( x)


x 2

Exercise 3.4 For what values of x is f(x) discontinuous?

Exercise 3.5 Fill in the given table and use it to find lim f ( x) .
x2

x 1.99 1.999 1.9999 2 2.0001 2.001 2.01

x2
f ( x)  ?
x2

Exercise 3.6 Fill in the given table and use it to find lim g ( x) .
x 0

x -0.01 -0.001 -0.0001 0 0.0001 0.001 0.01

231
 x  1, x  0

g ( x)  3, x0 ?
 x 2  1, x  0

x 2  4 x  21
Exercise 3.7 Evaluate lim 2
x 7 x  5 x  14

x 1
Exercise 3.8 Evaluate lim
x 1 x  2x 1
2

x 2  2 x  x3
Exercise 3.9 Evaluate lim
x  3x 4  7

 x  2, x  1
Exercise 3.10 Is f ( x)   continuous at x = -1? At x=1?
 x , x  1

232

You might also like